(1967, R. Holm) Electric Contacts

RagnarRolm
Electric Contacts
Theory and Application
Fourth Edition.
Springer-Verlag Berlin Heidelberg GmbH 1981

ISBN 978-3-642-05708-3 ISBN 978-3-662-06688-1 (eBook)
DOI 10.1007/978-3-662-06688-1
Library of Congress Cataloging in Publication Data (Revised):
Holm, Ragnar, 1879-
Eiectric contacts.
The first English language ed. (Stockholm.1946) was largely
a translation ofthe author's Die technische Physik der elektrischen Kontakte.
Bibliography: p.
1. Electric contacts. I. Holm, Else,joint author.
TK2821.H613 1967 621.31'6 66-29437
This work is subject to copyright All rights are reserved, whether the whole or part
ofthe material is concemed, specifically those oftranslation, reprinting, re-use of
illustrations, broadcasting, reproduction by photocopying machine or similar means,
and storage in data banks.
Under §54 of the German Copyright Law where copies are made for other than
private use, a fee is payable to the publisher, the amount ofthe feetobe determined
by agreement with the publisher.
© by Springer-Verlag Berlin Beideiberg 1958 and 1967.
Originally published by Springer-Verlag Berlin Beideiberg New York in 1967
Softcover reprint of the hardcover 4th edition 1967
The use of registered names, trademarks, etc. in this publication does not imply,
even in the absence of a specific statement that such names are exempt from the
relevant protective laws and regulations and therefore free for general use.
2061/3014-5432
Electric Contacts
Theory and Application
By
Ragnar Holm
Ph. D., hon. Dr. Techn.
in collaboration with Else Holm Ph. D.

both of St. Marys (Pa.)
Fourth completely rewritten edition
With 218 Figures
Springer-Verlag Berlin Heidelberg GmbH 1967

Preface
This book is a completely revised and rewritten edition of "Electric
Contacts Handbook" published in 1958. A large number of new in-
vestigations are considered, and many of the basic theories are revised
in detail and even in general.
The body of information had to be limited as it was not advisable
to increase the volume of the book. In particular, no attempt was made
to cover all of the practical applications. They appear as examples
following concentrated explanations of basic phenomena.
As in several branches of technology, the solutions of problems ari-
sing in the field of electric contacts involve insight into various disci-
plines of physics. It is feit that reviews of some of those topics, especi-
ally adapted to electric contact phenomena, are welcome to many
readers. For example, chapters have been devoted to the structure of
carbon, the band theory of electric conduction in solids, certain pro-
blems in statistics, and the theory of the electric arc. As regards arc
problems, new ideas have been introduced. In order to make the main
text less cumbersome, such reviews are presented as appendices.
Throughout this edition, the mksa-unit system is used in accord with
the latest recommendation for standardization of units in scientific and
technical writings. The chapter "History of Early Investigations on
Contacts" forming Part IV in the preceding edition of 1958 has not
been repeated in this book.
The bibliography is restricted and so fashioned that the literature
index together with citations in specifically relevant papers essentially
covers the entire modern literature on electric contacts. We also point
to citations in the earlier editions and to the "Bibliography on Elcctric
Contacts" published by the American Society for Testing Materials.
For valuable information, the author is endebted to Professor
S. MRozowsKI, University of Buffalo; Dr. A. BAUER, Privatdozent at
the Technische Hochschule in Karlsruhe, Germany; Dr. T. E. BROWNE,
Westinghouse Electric Corporation; and to bis colleagues of the Stack-
pole Carbon Company, particularly Dr. E. I. SHOBERT II, G. A. MEYER
Sr., and W. KRELLNER.
The research in the field of electric contacts by both author and
collaborator was essentially carried out during their tenure with the
IV Preface
Siemens-Werke in Berlin, Germany, and with the Stackpole Carbon

Company in St. Marys, Pennsylvania. The support of the manage-
ment.s of these corporations is gratefully acknowledged.
Finally, it is a pleasure to express particular thanks to Springer-
Verlag in Berlin, Germany, for t.he kind invitation to write this book,
and for the excellent printer's performance.
St. Marys (Pa.), January 1967 R.Holm

Contents
Page
List of frequently used symbols and abbreviations XIII
Part I
Stationary Uontacts
§ 1. Introduction, A simplified summary of the theory of stationary electric
contacts..................... ................................. .. 1
§ 2. The contact surface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
§ 3. The contact resistance. General theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
§ 4. Calculation of constriction resistances with constant resistivity !.! in alt
isotropic material . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
Problem A. 13- Problem B. 14- Problem C. Constriction resistance of an
elliptic spot. 14 - Problem D. 15 - ProblemE. 17 - Problem F. The in-
fluence of the elliptic shape of the contact area on the constriction resist-
ance expressed by a shape factor. 18
§ 5. Constriction resistances when conditions deviate from those in § 4, but
with e still a constant . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
Problem A. 20 - Problem B. The constriction resistance of one member
when the a-spot is covered with a film. 20 - Problem C. Multispot metallic
contact referred to a semi-infinite member. 21 -Problem D. 23- Problem
E. 24- Problem F. 24- Problem G. Distorted constrictions. 25- Problem
H. 26 - Problem I. 26
§ 6. Introduction to thin films on contacts. Conta<:t cleaning . . . . . . . . . . . . . . . 27
§ 7. The Ioad boaring contact area as a function of load and elastic and plastic
properlies of the members . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
.-\. Introduction. 29- B. Flat Contact. 30- C. Experimentaldetermination
of A 6 • 32 - D. Persistence of asperities in indentations. 37- E. Infl.uencc
of temperature and contact duration on the contact area. 37- F. Special
application of Eq. (I,23) concerning creep in contacts. 39
§ 8. The relation between contact load and resistance, particularly at moderate
and high load . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
A. Introduction with description of Fig. (8.01). 40- B. Crossed rod con-
tacts. 42- C. Explanation of the dashed lines in Fig. (8.01). 44- D. Di-
versified resistance measurements. 46- E. Use of Diagram (8.01) in prac-
tice. 47- F. Practically clean contacts, preloaded with a high P. 47
§ 9. Contact resistance on freshly cleaned rods in air at very small contact loads 48
A. Observations on gold and silver. 48 - B. Observations on carbon con-
tacts. 50- C. Method of wire probes for exploring contact surfaces. 52
VI Contents
§ 10. The inductance of a current constriction. Skin effect 52

A. Inductance. 52 - B. The skin effect. 54
§ 11. Electrodynamic repulsion in a symmetric contact of non-magnetic material 55
§ 12. The capacitance of a contact. Electrostatic attraction in a contact . . . . . 57
Example A. Crossed rod metallic contact. 58 - Example B. 59 - Ex-
ample C. JOHNSEN-RAHBEK effect. 59
§ 13. The relationship between electric potential and temperature in a current
constriction which is symmetric with respect to the contact surface; that
is, the tpO-relation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
A. Introduction. 60- B. Proof of the tpO-relation in a symmetric con-
striction. 61. C. Justification of assumptions. 63 - D. The tpO-relation
in metallic constrictions. 63
§ 14. The tpO-relation in cases of dissymmetry . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
Case A. 64- Case B. Contact between electrodes differering in e and J!. 65 -
Case C. 65- Case D. Heat fl.ows across contact surface. 67- Case E. 67-
Case F. Thomson effect. 67 - Case G. 68
§ 15. Infl.uence of a thin film in the contact on the tpß-relation. KüHLER effect 69
§ 16. The infl.uence of the JoULE heat on constriction resistences insymmetric
contacts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
Example A with W-F law. 74- Example B. Heat fl.ows across contact
smface. 75- Example C. Integration of Eq. (16.09) without using WIEDE-
MANN-FRANz law. 77
§ 17. Distribution of the temperaturein a symmetric constriction with circular
contact surface at given current . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
Example A. 78 - Example B. 79 - Example C. 80
§ 18. The (equilibrium) temperature distribution in the constriction of a con-
tact between two metals with different conductivities, both obeying
WIEDEMANN-FRANZ law. Thermoelectric effects . . . . . . . . . . . . . . . . . . . . . 80
A. Features of the thermoelectric effects. SO - B. The problern of tem-
perature distribution in the case of PELTIER heat in the contact. 82
§ 19. Temperature distribution in the constriction of a contact between mem-
bers of very different conductivities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
§ 20. Resistance-voltage characteristics of clean symmetric contacts. Softening
and melting voltages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
A. Introduction. 87- B. Discussion of Diagram (20.01). 87- C. Softening.
88- D. Melting. 89- E. Measurements of RU-characteristics. 91
§ 21. Development of the temperature in a metallic current constriction . . . . 92
A. Introduction. 92 - B. Survey of the problems. 93 - C. Diseuseion of the
diagrams for the different cases. 96 - Case A. Symmetrie contact heated
by the current throughout the constriction. 96 - Case B. Contact heated
by the current. The members have very different conductivities. 96 -
Case S. A semi-infinite body at rest, with a circular heat source. 97 -
Case (A and S) for very small z-values. 98- Case M. Heat source moving
on tbe surface of a semi-infinite body. 99 - D. Cooling of a previously
heated contact constriction. 100 - E. Examples. 100
Contents VII
§ 22. Growth of films on metals used for contacts, near room temperature. . . . 102
A. Introduction. 102 - B. Rate of adsorption. 103 - C. Development
of tarnish from a chemisorbed monolayer. 104 - D. Protective films.
107 - E. Tarnishing of some relatively poor contact metals. 108- F. Tar-
uishing of noble metals. 109 - G. Contact breathing. 111
§ 23. Growth of visible oxide films on metals at moderate to high temperatw·e.

Decomposition at still higher temperature. . . . . . . . . . . . . . . . . . . . . . . . . . . 111
§ 24. Water film, local cells and ru~ting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1H

A. Thickness of water films. 114- B. Rusting by means of electrochemical
attack. 115
§ 25. Alien ·solidified films on eontaets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
§ 26. Tunnel effect .................................................... 118
A. Introduction. 118- B. General results in the case of room temperature
and less. 122 - C. Details of observations on the tunnel effect at room and
lower temperature. 123 - D. Evaluation of film thickness, based on
measured a, for the caRe of smalls. 125 - E. Observations on tunneling
at room temperature through relatively thick films, 25 A and more. 126 -
F. Tunneling between dissimilar metal members. 128- G. The supercon-
ductivity of contacts, particularly contacts with a film which is penetrable
via tunnel effect. 128 - H. Type III, field emission. Infiuence of relatively
high temperatures on the tunnel effect. - 130- I. Tunnel effect when both
electrodes are of the same semiconducting material. 131 - J. Tunnel cur-
rent across a gap that surrounds a metallic contact consisting of a circular
spot with the radins a. 132- K. Tunneling within a semiconductor. 134
§ 27. Fritting of tarnish films . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
A. Introduction. 135 - B. Theory of fritting. 135 - C. Widening of an a-
spot by electrical forces. 137- D. Cessation of anormal A-fritting. 139-
E. A-and B-fritting; defiuition. 139- F. ObservationsonA-fritting. 140-
G. Observations on B-frittings. 146- H. Applied Fritting. 147- I. Metallic
or quasi-metallic contact through an opening in a relatively thin contact
film. 148- J. Polarity dependence of frittings of tarnish films on silver and
copper. 150
§ 28. Adherence in dry contaets which are not heated to any infiuential extent
by the current. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
A. Introduction and theory. 153 - B. Experimental verification. 155 -
C. Location of the weid break. 158- D. The practical significance of the ad-
herence in non-heated contacts. 159
§ 29. Adherence in contacts that are heated by the current passing through
them. Resistance welding ........................................ 160
§ 30. About stationary contacts in practice. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164

A. Introduction. 164- B. Type I, permanent contacts. 165- C. Separable
contacts of high mechanicalload. 166- D. Resistance measurements of
clamped contacts with force produced by a screw. 167.- E. Medium and
low load contacts in air. Printed circuit boards. 169- F. Medium and low
load contacts in sealed chambers. 171- G. Microcontacts. 172- H. Stored
lubricated contacts. 174- I. Contacts in measuring apparatus. 175-
J. Contact assemhly. 17;)
VIII Contents
§ 31. Dimensioning a contact with respect to its heating 176
§ 32. Contact effects in carbon granular microphones . . . . . . . . . . . . . . . . . . . . . . 180

A. Introduction. 180- B. Configuration of the transmitter. 180- C. Cal-
cula.tions of the a.c. power of a microphone. 181 - D. Deterioration by
"burning" of microphone granu1es during the life of the transmitter. 184
§ 33. Contact with semiconductors. Rectification. Static electrification . . . . . . . 185

A. Introduction.185- B. Contact between a meta! and a semiconductor,
and its rectification property. 185- C. The p-n-junction. 187- D. The
silicon controlled rectifier. 188- E. Remark concerning the contact be-
tween silicon carbide crystals. 189 - F. Static electrification. 189
§ 34. Carbon-pile rheostats. Electric resistance of pressed meta! powders ...... 190
Part II
Thermal Contacts
§ 35. Thermal metallic contacts ......................................... 193
A. Introduction. 193- B. Bilateral heat current. 193 - C. Thermal resis-
tance of nominally flat bolted contacts. Examples. 194
Part III
Sliding Contacts
§ 36. Survey of fundamentals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199

A. Introduction. 199 - B. The geometry of a nominally flat sliding con-
tact. 199 - C. Experiments. Type II. 202 - D. Experiments when Type I
wou1d be approached. 204
§ 37. Sliding contacts in air • . .. .. .. . . . . . .. . . .. . . . . . . . . .. . . .. . . . . . . . .. . 205
§ 38. Boundary lubrication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209
A. Features of boundary lubrication. 209- B. Discussion of two competing
theories of boundary lubrication. 213. - C. Lubricating practice. 218 -
D. The BEILBY Iayer. 218 - E. Boundary-lubricated separable contacts
including some sliding metal-metal contacts. 219- F. Properlies required
of bearing materials. 220- G. Ball bearings. 221
§ 39. Theory of friction and wear of carbon contacts. Lubrication by means of
solid lubricants as graphite and molybdenum disulfide . . . . . . . . . . . . . . . . 221
A. Introduction. 221 - B. Bond strength in a sliding contact between
graphite members. 223 - C. The high altitude effect. 225 - D. Adjuvants.
227
§ 40. Stick-slip motion. The temperature in currentless sliding contacts . . . . . . 228
A. Stick-slip or jerky motion. 228- B. The temperaturein currentless slid-
ing contacts. 230
§ 41. Frictional wear in metallic .contacts without current A. Types of wear.
232 - B. Details of the formation of wear detritus. 234 - C. The influence
of adhesive wear on friction in clean metallic contacts. 235 - D. Why
Contents IX
liquids, even the deposit from air humidity, are able to strongly infiuenee
wear, without appreeiably affecting the frietion eoeffieient. 235 - E. Frie-
tional wear in eurrentless sliding eontaets; represented by Z whieh is eal-
eulated aeeording to Eq. (41.03) as a funetion of the hardness, H, of the
softer member. 236- F. Size and frequenee of wear fragments appearing
during periods of adhesive wear. 240- G. Behavior of sliding eontaets
of measuring apparatus. 242
§ 42. Electrical performanee of earbon brushes on rings and commutators when
arcing is excluded. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 242
A. Introduetion. 242 - B. Early investigations about the conduction
meehanism in the brush-ring (eopper) eontact. 243- C. Polarity effect in
the eontaet between a graphite brush and a copper ring. 245
§ 43. The temperature in the sliding eontaet between a carbon brush and a
copper ring or eommutator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 249
§ 44. Fl;'ietion and wear with a carbon-brush eolleetor contaet .............. 251
A. Meehanieal wear with sliprings, no current. 252 - B. Infiuenee of the
eurrent on the wear (absenee of ares). 252- C. Brush wear during eommu-
tation with ares. 253 - D. Abrasion of the slip ring. 257 - E. Friction be-
tween an eleetrographite brush and a ring. 257- F. Brush lifting by hydro-
dynamie air pressure. 259
§ 45. Theory of eommutation with special regard to voltage fiashes and arcs . . 260
A. Fundamental equations. 260- B. A first integral. 262- C. Arcing during
unbalaneed eommutation. 263 - D. Energy dissipated in ares and voltage
fiashes. 264- E. Each brush makes contact with more than two segments.
266- F. The importanee of the elastieity of the brush for the eommuta-
tion. 266 - G. About so-ealled short-eircuit current in the brush. 267
§ 46. Current eolleetors for trolley cars . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 268
§47. Eleetric noise in eontacts ......................................... 269
A. The thermal noise. 269 - B. Shot noise. 270 - C. The speeifie eontaet
noise. 270 - D. Contaet trembling or agitation. 272 - E. Noise generated
by sudden ehanges in, or interruptions of, the contact. 272 ·
Part IV
Electric Phenomena in S"itching Contacts
§ 48. Introduction ..............................•..................... 274

§ 49. Ignition of ares in switehes ........................................ 275
A. Introduction. 275 - B. Electrical breakdown in the gap between
metallic electrodes at atmospheric pressure. 275 - C. Dtawing of ares.
278. - D. Are ignition followed by fioating. 278
§50. VI-eharacteristies of the stationary arc in air; their use for ealculation of
the duration of drawn short ares . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 279
A. Introduetion. 279 - B. Observations on breaking eontaets in an ohrnie
eircuit. 281 - C. Simplified VI-eharaeteristics for different metals with
known I,. and for small eurrents. 283 - D. Use of the resistanee line
together with arc characteristics for the problern of how a eonstant eur-
rent is shared between an arc and an ohrnie resistanee both in parallel. 285
X Contents
§51. Vacuum arc; particularly its extinction ............................. 286

A. Introduct.ion. 286- B. The duration of the vacuum arc. 286. - C. Cur-
rent ehopping by vacuum arcs. 289
§52. Interrupting an a.c. current ........................................ 290
A. Introduction. 290- B. Vacuum switch. 291 - C. High power current
breakers. 291 - D. Medium and low duty breakers. 295 - E. Contact recti-
fier and synchronaus switch. 296
§ 53. Breaking direet eurrent. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 297
§ 54. Elcctric oscillations generated by d.c. arcs . . . . . . . . . . . . . . . . . . . . . . . . . . 298
§ 5i). Bouncing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 300
§ 56. Material tmnsfer in switching contacts. A survey . . . . . . . . . . . . . . . . . . . . 304
A. Definition of the major types of material transfer. 304- B. Concerning
t.he amount of arc transfer, particularly from the cathode. 306- C. Are
quenching. 311 - D. Mechanical wear produced by the hammering effect
of switching contaet.s. 312 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 313
§ 57. Discharge transiPnts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 313
§ :38. Are duration during contact closure with voltages below 200 to 300 V ... 316
A. Introduction. 316- B. Calculation with respect to the wiring diagram
(58.02) with initially charged capacity. 316 - C. Inductance l = 0, (thus
ß = 0) in the circuit (58.02). 319 - D. A bat.tery instead of C, equivalent
to C = oo ; finite J. :l20
§59. Floating ....................................................... 320
§ 60. Are duration on breaking contact without quenching. Circumstances
common to Communication Engineering . . . . . . . . . . . . . . . . . . . . . . . . . . . . 323
_\, Ohrnie circuit according to Fig. (60.01). Are extinction in air during a
steady increase of the gap. 323 - B. Inductive circuit in air according to
Fig. (60.04). Capacity of the Ieads is neglected. 323 - C. The quantity of
electricity that flows through a drawn arc. 325
§ 61. Queuehing a drawn arc by means of a capacitor, parallel to the contact .• 325
A. General theory. 325 - B. A short description of the process. 327 - C.
~ote conccrning the position of r. 328
§ 62. Capacitive queuehing when an arc with a very small duration or no are is
drawn .......................................................... 329
A. Equations. 329- B. Condition for no breakdown of the gap between the
separating electrodes. 331 - C. Example. 331
§ 63. Queuehing of ares by a resistanee parallel to the operating contact or to
the inductive eoil . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 332
A. Queuehing with ,. in position a. 332- B. Are queuehing with r in posi-
tion b. 334
§ 64. Details about the types of arc in relay contaets and the material transfer
produced by them ....... , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 334
A. Survey. 334 - B. Result of measurements. 335 - C. The eriticallength,
s.,. of the änode dominated arc. 336
Contents XI
§ 65. Bridge transfer and short are tram;fer at contact separation . . . . . . . . . . . . . 338
A. Introduction. 338- B. A classical example of a long liquid bridge. 339-
C. Methods of determination of the rate of transfer. 341- D. Results of
measurements and their interpretation. 342
§ 66. Theory of thc bridge transfer . . . . . .. . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . 347

A. Introduction. 347. - B. Bridge transfer because of the THoMSON effect.
Theory. 348 - C. THoMSoN effect. Comparison with measurcments. 351 -
D. Konr.ER cffect. 352 - E. Comparison with measurements. 353 - F. For
completeness wc calculate the material transfer as it could be produced by
the PELTIER effect. 3i;:J - G. Remark. 31)4 - H. Rcmark. 354
§ 67. Bridge material tra.n><fer in thc shape of pips and spires ................ 354
§ 68. Mereury switches . .• . . . . . . . . . .• . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 356
A. Introduction. 356 - B. Medium duty switches with a pool of mercury
forming one electrode. 357- C. Mercury wetted contaets. 358- D. Mercury
between the members of a sliding contact. 358
§ 69. Application of statistics to contact operations • . . . . . . . . . . . . . . . . . . . . . . 359
A. Reliability of eontacts. 359- B. Some numerical results of tests. 361 -
C. Effeet of twin contacts. 361
§ 70. The choice of contact material; contact shape for practical applications 362
A. Permanent contacts. 363- B. Microcontacts. 363-C. Relay contacts with
high repetitive operation. 364- D. Light duty relays for medium frequen-
cies. 364- E. Medium duty circuit breakers and contactors. 365- F. Heavy
duty circuit breakcrs with up to thousands of amperes and volts. 365 -
G. Slirling contacts for resistors and apparatus. 366
Appendh1es
§I. Elasticity, plasticity and hardness ............................... 367
A. Introduetion. 367-B. HERTZ' formulas forideallyelastie indentations.
367 - C. Plastic deformations. Dislocations. 369 - D. Mathematics of
plastic yielding. 370- E. Indentation in an isotropic semi-infinite body,
produced by a spherical indenter (ball). 371 - F. The ball and pyramid
indentation tests. Hardness. 372- G. lnfl.uence of friction on the inden-
tation. 377 - H. Diffusion effeets. 377 - I. The work consnmed by a plas-
tic deformation. Examples. 378
§ U. Electronic conduction in solids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 380
A. Introduction. 380 - B. General concepts concerning conduction.
380 - C. Quantization of the electrons in a crystal and the constitution
of a band. 381 - D. Infl.uence of the temperature on the distribution (or
partition) of electrons on cells in metals. 382- E. Current carriers in se-
miconductors. 384 - F. Mobility of current carriers. Resistance. 386 -
G. Surfaee potential barrier of a metal. Thermionic emission of electrons
387- H. Equilibrium in a contact between metalsA and Bwithdifferent
work functions iP, iP..- > iP8 389 -I. Metal-semieonduetor contact. 390
§ III. Heat conduction. WIEDEMANN-FRANZ law (WF-law) .....••.....•.. 393
A. Heat capacity. 393- B. Thermal conduction. 395-C. WIEDEMA.NN-
FRA.NZ law (WF-law). 396
XII Contents
§IV. Probability. Noise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 397

A. Normal distribution. 397- B. Probabilitypaper. 399- C. The PoiS-
soN distribution. 400 - D. Electrical noise. 402
§ V. Structure, electrical and thermal conductivity of carbons. . . . . . . . . . . . 404
A. Introduction. Graphite lattice. 404 - B. Carbon grades. 406 -
C. Graphitization. 407- D. Electrical conductivity of carbons. 408- E.
Heat conductivity of carbons. 410
§VI. Hydrodynamic or thick illm lubrication ............................ 412
§ VII. Metal whiskers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 417
A. Whiskers formed from the solid metal. 417 - B. Whiskers rapidly
formed from metal vapor. 417 - C. Mechanical and magnetic properties
of whiskers. 418- D. Metallic dendrites in semiconductors. 418
§VIII. Some fundamental formulas concerning electric discharges . . . . . . . . . . 419
A. Introduction. Kinetic fundamentals. 419 - B. Drift velocity. 420 -
C. Thermal ionization. SAHA's equation. 420- D. Plasma. 421 - E. Cur-
rent in vacuum restricted by the space charge of the current carriers 422
§ IX. Theory of the electric arc . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 423
A. Introduction. 423- B. Cathode layer. 425- C. Emission of primary
electrons from the cathode. 426 - D. The ion current in the cathode
layer. 429- E. The rate of evaporation, w, m3 fcoul. 429- F. Summary
of the results presented in Table (IX,5). 430- G. Power balance at the
cathode. 430- H. Powerbalance at the anode. 433 -I. Voltage-current
characteristics of arcs. 433- J. Pinch effect. 435
§ X. Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 436
§ XI. Diagram XI in two copies, one detachable . . . . . . . . . . . . . . . . . . . . . . . 441
Author and literature index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 443
Subject index ...................................................... 477
Errata .... ···· ...................................................... 483
List of frequently used symbols and
abbreviations
a and c are indices, referring to anode and cathode, § 64, IX.
a-spot discrete, continuous and conducting contact area. A contact can have many
a-spots, § 2 to 5, 7, 30.
a radius of an a-spot.
b radius of a perfectly conducting sphere replacing the contact surface in model
(1.02), § 1, 10, 11, 12, 21.
c thermal capacity, J m-3 deg- 1 , § 21, TII.
e charge of the electron, 1.60 · 10-19 coul.
eV electron volt.
emf electromotoric force.
f friction coefficient § 36.
h relative humidity: h = 0.4 means 40%, § 31, X.
h pitch of a screw, § 30.
h PLANCK constant, 6.63. 10-27 ergsec = 4.14. 10-lii eV sec.
k BOLZTMANN constant, 1.38 -10- 23 J deg- 1 = 8.6.5 · 10-5 eV deg- 1
l small inductance, §58, 60, 61.
m mass of the electron, 0.91 . 10-ao kg.
p pressure.
q quantity of electricity passing through a single discharge, § i58, 60.
r radius.
r resistance in quench-circuit, §58, 61.
r latent heat of vaporization, J m-3 , §IX, X.
8 distance of travel of sliding contact member, § 41, 45.
8 gap, also arc length, §50, 52, 54, 61, 62.
8., arc length where anode dominated arc changes into cathode dominated, § 64.
time.
ta. duration of arc, §50, 51, 52, 58, 60, 61.
v velocity.
z=
ca
4t dimensionlass substitute for time, § 21.
A. apparent contact area.
A6 Ioad bearing area.
A. conducting contact area.
A. end-surface, § 3 etc.
A0 contact surface in several deductions.
A Angström unit = 10-10 m.
B radius of end-surface in model (1.02).
C electric capacity, § 4.
D volume of meta! deposited from vapor in an arc, § 64.
E electromotive force.
E Young's modulus, § 7, 8, I.
XIV List of frequently used symbols and abbreviations
F.L. research laboratory of the Siemens-Cancern, division Holm.

G material transfer caused by an arc. § 64.
H magnetic field intensity.
H ha.rdness, § I, X.
I electric current intensity.
I. a.rc current.
I, minimum current necessary for floating, § 59.
I". minimum currentnecessa.ry for an arc on a fl.at cathode, §54. lX. X.
I6 final current through a contact bridge, § 65.
J Current density.
L inducta.nce.
L ooefficient for the law of \VIEDEMAN:s--FRANZ-LoREXZ, § 13, IIT.
N newton = 0.102 kg force.
P mechanical contact load.
Q qua.ntity of electricity.
R constriction resistance in one oontact member.
R total constriction resistance.
R.6 and R~6 measured quantities for det~rmining R, § 3 etc.
R 1 film resistence, § 3, 8.
R"' resistance between the A 0 surface and another equipoteutial surface
Aw
RU-cha.racteristic giving the contact resistance plotted against the voltage, § 20.
T absolute temperature, °K.
U oonstriction voltage due to one contact membe1·.
U total contact voltage, § 1, 3, 4 etc.
U 61 boiling voltage.
U 1 sometimes U1 liquefying or melting voltage. § 13, X.
U 8 Fritting voltage, § 27.
U, sometimes U, softening voltage, § 20. X.
V voltage in general.
V,. arc voltage.
V". minimum arc voltage, § 58, 63. IX.
VI- cha.rakteristic of the arc, §50.
W thermal resistance, § 35.
W voltage induced by interpoles, § 45.
W frictional wear, § 41, 44.
W. and W. volume of metal tha.t, during the Jife time of an arc, evaparateR from
the cathode or anode respectively. § 65.
X electric field strength, § IX.
X 8 fritting field strength, § 27.
Z number used for the classification of frictional wear, § 44.
oc
y2 = 7f ratio between the axes, ar: a.nd p, of an elliptic a-spot, § 4.
y transfer of material, volume per coulomb, caused by the arc. § 58.
6 material deposit, volume per coulomb, during a.rcing, § 66.
w material evaporated, volume per coulomb, during arcing. § 66.
Q =ohm.
e0 permittivity (or dielectric constant) of vacuum = 8.8!). 10-12 farad/m in the
mkssystem.
e, relative permittivity.
17 ooefficient of internal friction, § VI.
{} temperature, particularly supertemperature, centidegrees.
List of frequeutly used symbols and abbreviations XV
A supertemperature of the warmest isothermal surface in a constriction (the

contact surface in a symmetrical contact), centidegrees. § 13 ek.
% electric conductivity, § IIL
i. thermal conductivity, wm-1 deg- 1 , § 13, 14, 3ö.
A mean free path, § IX.
flo magnetic permeability of vacuum = 1.257 . 10-6
H
m.
p, relative magnetic permeability, § 10.
11 parameter of equipotential surfaces of a symmetrical contact constriction
§ 4 etc.
1• frequency.
~ ratio pfU, § il. I.
II Peltier coefficient, § 18, 66.
(! = e0 (1 + a{}) electric resistivity, where q0 iH associated with the temperature
of the end -surfaces.
J 1 THOMSON coofficient, § 18. 66.
a tunnel resitivity, § 26.
<p eleetric potential.
tP work funetion, § 26, 3:3.
1p friction force per unit sliding area, § 3ü.
this sign is used as a symbol for chapter and appenclix; thus § 15 mean..,
chapter 15. and ~ III means appendix IIJ.
N.n.
Propositions, equat.ions, and figures are numbered consecuti,·ely
within each chapter in the following way: in Fig. (18.04), 18 refers to
§ 18 and 04 is the number in that chapter.
Numbers in square braekets refer to the lü;t. of Iiterature a t the end
of t.he book.
Part I
Stationary Contacts
§ I. lntroduction. A sim.plifled summary of the tbeory of

stationary electric contacts
An attempt to present the concepts of contact theory in such a
sequence that none of them is used before being thoroughly defined
would be unwieldy. For example, the fundamental concept of contact
surface can not be treated in detail before the constriction resistance
has been defined, since determinations ofthe conducting areas arc based
on the measurement of constriction resistances. Again, the concept of
constriction resistance can not be thoroughly treated without reference
to the contact areas. Therefore, it seems preferable to introduce, in a
provisory manner, some important concepts in an opening paragraph,
allowing a more elastic and agreeable exposition in those which follow.
The introduction is given a quantitative character by means of some
calculations, based upon an artificial model of the current ßow lines
which highly simplifies the treatment. In the later paragraphs the same
model is used for the calculation of particular problems. Some of the
concepts are de.fined solely in this introduction.
The term electric contact means a releasable junction between two
conductors which is apt to carry electric current. These conductors may
be called contact members, or simply contacts, when no misinterpretation
is likely. The member from which the positive current enters the contact,
is called anode; the other member is the cathode. When the contact
members are separated by an insulating layer, it is conventional to
speak of an open contact.
The force that presses the contact members together is the mecha.
nicalload or simply the load, P. If the contact members were infinitely
hard, the load could not bring them to touch each other in more than
three points. But since actual materials are deformable, the points be.
come enlarged to small areas and simultaneously new contact points
may set in. The sum of all these areas or spots ist the load bearing area,
Ab, upon which the pressure, p, is finite. Ab can be generated merely
1 Electric Contacts, 4th Ed.
2 Stationary Contacts
by elastic deformation. But, because of the unevenness, the contact
members, even though they may be nominaJly flat, actually touch each
other in areas that are more or less plastically generated. They then
satisfy Eq. (1.01)
(1.01)
whereusually0.2 <; < 1 andH is "the contact hardness" 1 . Itisdifficult
to polish contact members so as to attain an average pressure assmall
as 0,05H.
The area Ab usually is much smaller than was supposed until about
40 years ago. In the case of nominally flat bodies placed on top of each
other (as a brush on a ring), the whole covered area was often called
the contact surface. It is more correct to call it the apparent contact
area, as we shall do in the following. Ab may be of a much smaller
order of magnitude than the apparent contact area.
The expression contact resistance will often be used. This term was
coined at a time when it was believed that the metallic contact surface
itself accounted for the observed resistance. Actually, what is measured
as contact resistance R, always implies or is a constriction resistance Re,
which is the consequence of the current flow being constricted through
small conducting spots. Of course, the constriction extends into both
contact members. An alien film in the contact may give rise to an
additional resistance R1.
One may theoretically expect that the discontinuity of the crystal
lattice order in the contact will refl.ect electrons and produce a tran-
sition resistance. However, such resistances are extremely small and
similar to the resistances of grain boundaries in a polycrystalline ma-
terial. Measurements are unavailable. According to a rough computa-
tion these transition resistances per m 2 might be of the order of a
= 10-s (] Q m 2 where (] Q m is the resistivity ofthe metal; cf. Eq. (3.04).
The surface Ab usually ist partly covered by insulating tarnish films
and then only a fraction of Ab has metallic or quasimetallic contact.
A conducting contact area is called quasimetallic when it is covered with
a thin film (20 A or less) that is penetrable for electrons by means of
the tunnel effect 2 , and consequently produces a relatively small film
resistance.
Summing up, we see that not only is the load bearing contact area
very small, but also that only a fraction of it may be electrically con-
ducting. In any case, the current lines of flow are bent together through
narrow areas, causing an increase of resistance beyond the case of a
fully conducting, apparent contact surface. This increase of resistance
is the constriction resistance.
1 Cf. § 1F. 2 See § 26.

§ 1. Introduction. Theory of stationary electric contacts 3
The main part of the constriction resistance is localized in the
immediate neighborhood of the small conducting spots. The probes
measuring the contact voltage will always be at macroscopic distanees
from the contact surface. In most cases this distance covers the total
range in which the lines of flow are deviated from a straight path
by the constriction. Thus, the resistance measurement does not tell
directly whether the measured resistance ist essentially located within
t.he contact surface or, more or less, has the character of a constriction
resistance. Weshallshow that the constriction resistance can be calcu-
lated as a function of the conducting contact area, and when the mea-
sured resistance is greater than calculated for a known area, we conclude
that the area is covered by a filmthat produces an additional resistancc.
Conventionally the expression contact resistance is preserved irrespective
of whether there is merely a pure constriction resistance, or whether a film
also contributes by a rAsistance at the conducting area.
We illustrate the constriction
resistance by a calculation1 under
simplified conditions: assume a. cir-
cular contact surfac<>, samematerial
in both contact members and per-
fect symmetry, i. e., no disturbance
by any thermoelectric effects. The
contact members are considered as
semi-infinite. Even with these as-
sumptions, the exact calculation is
circumstantial (cf. § 4). However,
by using the imaginary model illus-
trated in Fig. (1.02), we approach
the real shape of the lines of flow
with a much simpler calculation.
Fig. (1.02). Model to illustrate a currcnt cun-
In the model the contact surface striction with spherical symmetry. 'fhe resi·
has been replaced by a sphere, K , stance betwecn two consecutive equipotential
surfaces is 1/6 of the constriction resistancc
of infinite conductivity, whose rain one rnember
dius, b, is chosen slightly smaller
than the radius, a, of the circular area; cf. Eq. (1.06). The lines of
current flow start radially and are symmetrically distributed around
K so as to arrange the equipotential surfaces as hemispheres concen-
tric to K.
Consider the constriction resistance in one contact member. The
resistance d R between the hemispheres with the radii r and r + dr is
1 R. HoLM (1].
1*
if e is the resistivity of the contact material. We integrate through a

hemisphere which passes through the point where the potential is tapped
for measuring the contact voltage. The radius, B, of this hemisphere
willbelarge compared with b. More distaut parts add very little to the
integral. We may therefore choose infinity as the upper Iimit and thim
obtain the constriction resistance in one contact member
(1.03)
The total constriction resistance R is twice as great, thus1
(1.04)
Eq. (1.04) ist based on the artificialmodel (1.02) and thereforeisapprox-

imate only. The correct value of the resistance belanging to a flat
circular contact area with the radius a is
(1.05)
as will be proven in § 4. Eqs. (1.04) and (1.05) define the same R if
2a= nb (1.06)
a relationship that plays a part in making deductions in later chapters.

The contact is heated by the current. If both contact members are
of the same material, the highest temperature is localized in the contact
surface and distaut points are at correspondingly lower temperatures.
Thus, if the supertemperature in the contact surface above the bulk of
the contact members is e, and the temperature coefficient of resistivity
is a., the constriction resistance will be increased by a factor somewhat
less than (1 + cx@). This would be the factor in the case of a uniform
temperature distribution. The factor for the resistance of the constric-
e
tion, in which is the maximum temperature, is approximately
(1.07)
cf. § 16.
1 It may be asked why the formula (1.05) contains the factor 1/a instead of
1/a2 • The following simple consideration gives the answer. An electric resistance
varies in proportion to a length and in inverse proportion to a cross-section. The
dominating part of the constriction resistance is found in the neighborhood of the
contact surface. The order of magnitude is for its length a and for the cross-section
a 2 ; thus the factor in question is afa2 = 1/a.
§ 1. Introduction. Theory of stationary electric contacts 5
If we want to check this formula, the problern would be how to
measure the temperature e. Perhaps the first idea would be to try
thermoelements fitted near the contact. This does not work because
the elements never get close enough to the contact surface. The error
would amount to the order of the measured value.
A simple, indirect method, that enables the determination of the
supertemperature from the contact voltage exists because of the inter-
esting relationship between the supertemperature ein the contact sur-
face of a monometallic contact and the conta.ct voltage, U = RI. This
relationship is a consequence of the heat flowing along the same paths
as the electric current, irrespective of the fact that the surroundings of
the contact which insulate electrically do not insulate thermally. Be-
cause of the symmetry, there is no reason for heat transfer from one
member to the other.
As is proved in § 13, saidrelationshipfor a monometallic contact with
electric and thermal conductivities 1/e and ). respectively is
(1.08)
in the equilibrium state. Table (1.09) is calculated for copper. How-

ever, the table has a general validity for metals used in contacts, for
according to the WIEDEMANN-FRANZ law, e). is nearly the same for
different metals 1 .
Table (1.09). Copper
softenlng melting
lJ 0.03 0.12 0.3 0.41 V

e 16 190 700 1063 centideg.
2
1 +-ae 1.04 1.5 2.8 3.8
3
The table gives supertemperatures, 0, related to contact voltages,

U, with the bulk at about 20 oc; and, below, the factor [1 + 2/3 (1X0)].
It gives the answer to the question raised concerning the method of
determining the supertemperature g in a contact. We simply calcu-
late it from the contact voltage using Eq (1.08).
It is particularly interesting that the voltage teils us whether the
m.elting point is reached. The melting temperature is, of course, the
highest temperature possible in a solid contact, and every attempt to
1 Cf. § IIIC.
increase the voltage beyond the melting voltage would result in the
contact sinking together and forming a greater contact area that
carries the current without further melting. Ifthis happens, the voltage
usually drops to a value somewhat below the melting voltage. Melting
voltages of various materials are given in Table (XI.2).
The temperature can surpass the melting point only if the contact
members are mechanically kept from approaching each other, for ins-
tance, in an opening contact. Then the boiling point of the metal may
be reached. For copper, the boiling point corresponds to 0,8 V and for
tungsten, 2.1 V. Boiling in the last moment of opening may imme-
diately precede the ignition of an arc with an arc voltage of 10 to 15 V;
since the arc demands a certain minimum gap, we conclude that boiling
here appears like an explosion which at once produces a gap and fills
it with overheated, ionized vapor.
In order to check the factor [1 + 2/3 (o:0)] one must, of course, in-
vestigate a contact with a constant contact area. Usually, the contact
area enlarges with increasing current because of softening of the metal
by the heat. Therefore, in order to secure a constant area, one should
begin the test with a high current and proceed to lower values. Or,
better still, vary the current up and down and record reversible varia-
tions (see § 20). If such a reversible series of resistance records shows
that the resistance increases with rising current (rising R U-character-
istic, § 20), this would prove that the contact is metallic. Conversely,
a falling R U-characteristic would indicate that the resistance belongs
to either a semiconducting film, say a tarnish film, or to a thin, tunnel-
conducting film. It happens that visibly tarnished contacts have a re-
versible R U-characteristic that rises with increasing current. Such a
behavi_or indicates that the current :flows mainly through (invisible)
metallic contact spots which have been produced either by rupture of
the film at contact make or by fritting.
A-fritting is an electric breakdown that occurs when the electric
field reaches the order of 108 Vfm and may result in a metal bridge
through the film 1 or even in a small conducting contact spot. There is a
commonplace example of fritting. If you investigate an ordinary plug
and socket contact by using a small emf, such as 1 V, you may fre-
quently find it insulating. But, this feature is not noticeable in service
since the ordinary line voltage is able to frit the contact.
W e also distinguish B-fritting which Ieads to enlargement of the
conducting areas at relatively low voltages across the conducting spot
that is limited by a surrounding film. For details see § 27.
When the contact film is thin enough to be permeable to electrons
1 Cf. § 27.
§ 2. The contact surface 7
by means of the tunnel effectl, usually no fritting occurs. A contact
spot with such a film has been called quasi-metallic. The tunnel resis-
tance may be negligible at high Ioads; but if the load is of the order
of 0.01 N and the contaot area correspondingly small, this resistance
may surpass the constriction resistance and render measurements in-
consistent. Many investigators have observed that it is necessary to
submit such contacts to small vibrations before measuring in order to
secure reproducible values. HoLM called this action aging the contaot 2 •
The explanation of the aging seems tobe as follows. When a clean
metallic surface is exposed to air, it soon becomes covered by two lay-
ers of oxygen atoms. The outer layer is bound much weaker than the
inner one but contributes to the tunnel resistance by a far greater
amount. Later, these layers may develop into an oxide tarnish. It
seems that aging results in the mechanical breakdown of the outer
oxygen layer.
Wehave sketched problems of stationary contacts which constitute
the very fundamentals of the theory of electric contacts. An extended
treatmentwill be presented in Part I, while Parts (III and IV) will be
devoted to problems of moving contacts. They concern: Part III slid-
ing contaots, and Part IV make and break contacts, including the
theory of the arc, the chief enemy of the switches. Although the elec-
tric conduction through sliding contacts ist physically the same as
through stationary contacts, a significant difference may result from
the fact that single contacts serve too short a time for reaching temper-
ature equilibrium. This means that in sliding contacts Eq. (1.08) is
no Ionger valid. Part II is devoted to thermal resistances.
§ 2. The contact surface
Concepts such as contact members, mechanical contact Ioad P, Ioad

bearing area Ab, apparent contact area Aa differentiated from the true
contact area A., have been defined in the Introduction. The relation
between the load-bearing contact area, the contact Ioad, and the average
pressure, p, is
(2.01)
The local pressure, p, may vary from point to point with elastic
deformation in some spots and plastic deformation in others. In many
1 Cf.§ 26.
2 In Germanit was called normieren; see R. HOLM [29] p. 69; cf. § 20B.
cases however, the entire area Ab will yield plastically; for example, in
contacts between two cylinders placed crosswise (cross rod contact)
at an ample Ioad. Then, the average pressure
will be close to the hardness lP. Even in mac-
roscopically flat contacts, p usually ist not
lower than about 0.2 of the hardness because a
considerable fraction of the Ioad-hearing eleva-
tions become plastically deformed; cf. § 7. It is
to be noted that the hardness ist not a strictly
unambiguous concept, and the reader is referred
to Appendix I and § 7 for detailed information
on hardness and Ioad-hearing area.
Fig. (2.02). Apparent con· From the aspect of current conduction, the
tact surface .4 load-bea·
4 ;
Ioad-hearing area may consist of three differ-
ring contact area .db, con·
taining insulating spots rent parts; cf. Fig. (2.02):
(shaded) and conducting 1. Portions with metallic contact. The cur-
spots, i. e., a-spots (dotted)
rent passes through them without perceptible
transition resistance in the interface, just as it does between different
crystallites in a compact meta!; cf. Introduction.
2. Quasi-metallic spots. Theseare film-covered areas with films suffi-
ciently thin to be easily permeable by the electron current by means
of the tunnel effect, irrespective of the resistivity of the film material.
Typical films of this kind are chemisorbed layers of oxygen atoms which,
in air, are formed on any meta! surface. This will be discussed in §§ 3,
6, 20, and 26.
3. Areas covered by relatively thick films; particularly, visible tar-
nish films (oxides, sulphides, etc.). As a rule, such areas are pratically
insulating.
The short name a-spot for the conducting contact areas, referring
to the radius a of a circular contact area, is a widely accepted term.
Fig. (2.02) illustrates different kinds of contact surfaces. The ratios
between these surfaces are diversified, and it may even happen that
Aa = Ab = A 0 • For example, if two clean meta! cylinders, whose dia-
meter is a few mm, are placed crosswise in contact with an ample Ioad,
50 N, plastic deformation Ieads to a Ioad-hearing area which is simul-
taneously the conducting area. This area is circular within the Iimits
of the irregularities of the surfaces. It is a difficult problern to deter-
mine Ab and Ac in cases where these surfaces are small compared to
Aa, as in the contact of a carbon brush on a collector ring. The present
state of our knowledge concerning Ab and Ac is summarized in § 7.
The contact between mercury and a solid metal has particular fea-
tures, since the deformation of the solid member is perfectly negligible
1 As for hardness, see § IF.
§ 3. The contact resistance. General theory 9
and tarnish or chemisorbed films on it remain undamaged. Here the
Ioad-hearing area and the apparent contact area are equal, i. e., Aa = Ab.
The contact between the film covered meta! and mercury provides
means for investigating the conduction of the films because we may
regard Ac = Aa.
§ 3. The contact resistance. General theory

An introductory description of the basic ideas of constriction and
contact resistance has been presented in § 1. We shall now enrich and
widfln the knowledge by forming the definitions on a broader basis, and
it. is feit that. a typical example may illustrate their content without
limiting the range of vision. Since thermoelectric
forces are irrelevant for the concepts, we neglect
them in the first instance and take them into con-
sideration later. See § 18.
In Fig (3.01), the cylinders 0 1 and 0 2 are the
contact members. They are placed on each other,
base Aa 1 against base Aa 2 , thus Aa constituting the
apparent cont.act surface.To begin with, we as-
sume that the faces Aa 1 and Aa 2 are clean metallic
but, due to their uneveness, have contact only in
Fig. (3.01). Theappar·
a small portion of Aa namely in Ac. Because of ent contact surface A4
the current flow being constricted through A c, a is the common basesec-

tion of the cyllnders
"constriction resistance" appears; cf. § 1. It will be which have real con-
shown that this resistance not only depends on the tact only in a portion
of A.
size of the area Ac but also on its shape. Ac often
consists of several spots, a-spots; cf. § 2. The voltage Uab may be mea-
sured between the points a and b when a current I passes through
the contact. Consequently the resistance between the equipotential
surfaces containing a und b respectively is
R ab = [Uab
W e now imagine one single solid cylinder similar to the model

except that the entire area Aa is perfectly conducting so that the lines
of current flow go straight through it. In this case Iet the resistance
between the equipotential surfaces containing the points a and b, the
same as in the model, be ~b· Then by definition, the constriction res1~s
tance and constriction voltage are
and
R = Rab- Rgb} (3.02)
U=Rl
In the case of a clean meta! contact, i. e., with no film in the contact,
the contact resistance R, is simply a constriction resistance. See § 1. If
a film is present and both sides have different metals, the contact re-
sistance R is the sum of the constriction resistances R 1 and R 2 in the
two members respectively and the film resistance R,, thus
(3.03)
According to this definition, the contact resistance is not a transition

resistance, as was believed earlier, but a surplus resistance in the body
of the contact members, caused by the narrowness of the current paths
through the a-spots, to which the resistance of a film on the a-spots may
add. It is shown in § 5 B that R 1 and R 2 arenot quite independent of R,.
As may be understood from the introduction, the order of magni-
tude of R 1 and R2 is __!?.!__
na
and -~
na
, where a is an averagelinear dimen-
sion of the a-spots, n their number, and g1 and g2 the resistivities
of the members C1 and C2 • A more accurate calculation of these resis-
tances is given in § 4. The corresponding expression for Rr is
(]
Rt=-
A, (3.04)
in the case of a conducting film uniformly distributed over the conduct-

ing area Ac, where a is the resistance across one cm 2 ofthat film. With a
varying along the film, one applies
~~= Jd:_c (3.05)
With r!t being the resistivity of the film material and s being its thick-
ness we have
(3.06)
Through very thin films, the tunnel effect furnishes a current inde-
pendent of e1 even if e1 is "infinite". For this case, a is defined and
calculated in § 26. In later chapters the picture of the resistances R 1 ,
R 2 , Rr will be completed by the study of details under different condi-
tions.
Those regions within the contact members C1 and C2 (Fig. 3.07)
where, owing to the smallness of the contact spots, the lines of current
flow noticeably deviate from the straight course, are called constriction
regions or simply constrictions. Within the constriction region, the po-
tential gradient is relatively great but relatively small outside the con-
striction; in other words, R: b in Eq. (3.02) is relatively small, usually
even negligible as compared to Rab· As a consequence, an exact de-
finition of the positions of the probes a and b ist not necessary. This
§ 4. Calculation of constriction resistances with constant resistivity e 11
fact obviously contributed to the impre:"<sion that the resistance between

equipotential surfaces containing probes as a and b is locaterl within
the contact surface and to the designat.ion of
contact resistance for something that usually
is a constriction resistance, localized in a very
small but finite volume within the contact
c.
members.
Fig. (3.07) shows the system of equipoten-
tial surfaces and current flow line:"< when both
members are of the same metal, and A,. repre-
sents a single circular a-spot (c means conrluc-
ting) in the middle of Aa. In practice, the con-
striction may be regarded as limited in thc
bulk of the members by certain surfaces, as for
example A. in the figure. Such surfaces we call
end-surfaces. According to the figure. t.he dis-
tance of A. from the a-spot. is of the same
order of magnitude as the radius r of the cy-
linder; and since R:b is of the order g BfnB2
= efnB, it is negligible compared with Rab if
B is very much greater than a. If the constric- ~
tion volume compared with the volume a 3 is so
great that R:b may be neglected, we speak of
a long constriction; cf. Eqs. (4.09, (4.15) and Fig. (3.07). Lines of current.
(4.21). flow and equipotential sur-
faces of a current constriction
As already described, the smallness of R~ b
follows from the voltage gradient being small at distances from thC'
a-spot which are much greater than the radins a of the spot. When,
lwwever, a constriction is limited to a distance comparable with the
radius a, R: b is no Ionger negligible and the constriction is called short.
Concerning the problem of rleterming the quantities A c, a, anrl the
ratio aJB which are implied in the formlllas of t.llis chaptf'r, we refer
to §§ 7 and 26.
§ 4. Calculation of constriction resistances with <'Onstant

resistivity ~ in an isotropic material
We shall consider long constrictions; i. e., the conducting area Ar is

small compared with the dimensions ofthe "semi-infinite" contact mem-
bers, and R:b in Eq. (3.02) is negligible. The constriction resistance R
is a function of the resistivity e of the mat erial and the dimensions and
shape of A c. The general theory of the calcnlations of R will be eluci-
dated, but the detailed evaluation of the forumlas will only be given for
three relatively simple problems: 1) Ac is a (hemisphere or) circle;
2) Ac is an ellipse and 3) Ac is composed of several circular or elliptic
a-spots. Allproblemsare treated under the conditions that the material
is the same in both members and that (!, thus also the temperatur, is
the same in all points of the constriction. The assumption of a spatially
homogeneous structure concerning (! at constant temperatures is no
considerable Iimitation in most applications because structural Varia-
tions usually are insignificant in so small a region as a constriction.
vVe shall generally assume that the conducting area Ac is equipoten-
tial, as is true, due to symmetry, if both membcrs consist of the same
material and thus ha ve equal constrictions. Exceptions which are treated
in § 5 show that deviations from the equipotentiality of A" have only
a moderate inßuence on the constriction resistance.
We seek the resistance between Ac and another equipotential sur-
face A1 belanging to the same contact member, and for this purpose
require a mathematical expression for the current ßow in the space
between A" and A1 . It is immaterial whether or not this expression
describes the state outside this region correctly. A method often applied
is the following: Ac is considered as a current source and A1 as a sink,
the region between them being free from sources and sinks. Then, the
current vector can be expressedas the gradient of a potential function,
q;, which in the respective region satisfies the Laplacian equation 1
and the mathematical problem is to find solutions of this equation which

fit certain boundary conditions.
The Laplacian equation appears in the electrostatic problern of the
capacitance between electrorles Ac and A1 in vacuum with boundary
conditions which are mathematically similar to the conditions of the
current problem. In handbooks on electricity, the electrostatic problern
usually is treated with mathematical details, whereas the description
of the current problern is shortened by reference to the electrostatic
one 2 • Since the picture of the electrostatic potential being caused by
charges is relatively easy to visualize, we shall deduce important re-
sistance formulas by reference to known expressions for capacitances.
Fig. (4.01) illustrates, in a general manner, the part of the contact
member which is bounded by the surfaces Ac, A1 and the mantle sur-
1 With e varying, the electrostatic picture would contain space charges and the
Laplacian equation no Ionger be valid.

2 Cf. forinstance RIEMANN-WEBER [IJ chapteron Stationäre Ströme, or SMYTHE
[1] § 6.14.
§ 4. Calculation of constriction rPsist.ances with constant resistivity e 13
face M of the body. In the capacitance problem, we consider Ac and

A 1 as charged conducting sheets with the potentials q;c and q;1 , situated
in an infinite space of perfect vacuum so that.
the potential q; is calculat.ed without any re-
gard to the boundary M. If the current prob-
lern shall correspond to the electrostatic one,
it is necessary that the condition (dq;fdn)u = 0
is automatically satisfied at M, where n is the
normal to M. This may seem to be some-
Fig. (4.01). Symbols
what excessive, but one can easily verify that
the condition is secured by the following examples. Using cgs unitsl
the >mperficial charge on Ac is
Q = 4n
- 1/lßq;i
on - ~ dA '. (4.02)
Ae
and on A 1 sits an equal charge, only with opposite polarity.

The capacity is
(4.03)
In the resistance problern (now with n being th!:' normal to A,.) the
current density is ~ j ~: \ , and the total current. is
I = ~! I ~: IdA , (4.04)
Ae
From Eqs. (4.02) to (4.04) the respecti\·e resist.ance is obtained as

R _ I ({Je - f{!t I _ _e_ (4.05)
"1 - I -- 4n0
The application of boundary conditionl" will be elucidated by examples.

Eq. (4.05) has been deduced in cgs units but remains valid if we mea-
sure R in Q, q; in V, I in A, ein Q cm, and C in electrostatic cm. In the
following we shall apply Eq. (4.05) to distinct problems.
Problem A. In Fig. (4.06) K is a metallic sphere with a diamter D
cm and resistivity e Q cm. The hemispherical surfaces Ac and A 1 , both
equipotential surfaces with the radius b cm, const.itute the inlet and
outlet of the current. We assume that bis much smaller than D. We
seek the resistance between Ac and A1 . In order to simplify the calcu-
lations we first imagine Ac and A 1 to be replaced, in the electrostatic
case, by whole spheres Sc and Sl' The capacity between them is known
1 In this deduction, cgs units are used for the following reasons: 1. because
with them the expression for 0 is particularly simple, and 2. because they are nsed
in the references.
14 Stationa.ry Contacts
to be C
= 2( ! _~) cgs umts
1 .
For t.he resistance problern we first imagine perfectly conducting spheres

S~and S~ in an infinite conductor with constant resistivity f2· Accord-
ing to Eq. (4.05) the resistance between these
spheres is
R' = 2en ( ~ - ~)
Because b is small compared wirh D, the lines of
flow start radially in spherical symmetry from s;
and S~. The sphere K intersects S~ and S~ along
l<'ig. (-!.Oß). Metalsphere meridians. It follows that exactly half the number
K with hemispherical of lines of ßow begin inside K and the other half
currcnt inlet and outlet h f ß
surfaces at boths end of outside K. Now if t e lines o ow starting ex-
a diameter actly along the intersection between K and S~
continued their course along the surface K, exactly half the current
flux and consequently half the conductance, 1fR, would be in K.
Although the comse is not quite as simple as this, the existing devi-
ations do not perceptibly influence 1 the resistance of the sphere K.
Thus with good approximation
R = 2 R' = _g__ (..!._ - _:!__) (4.07)
n b D.
It is evident that Eq. (4.07) and the following formulas for resistances
remain valid if (] is expressed in Q m and the linear dimensions, here b
and D, in m.
Problem B. Recalling the calculation in connection with Fig. (1.02),
we now seek the resistance between the concentric hemispheres with
t.he radii b and r. We obtain
(4.08)
Hence, for a long constriction, i.e., with r » b
(4.09)
In the problems A 'Rnd B, no area deserves the designation of"appa-

rent contact area". Const.riction exists because b is small compared
with r and D.
Problem C. Constriction resistance of an elliptic spot on the flat sur-
face of a semi-infinite body is of basic importance in the contact theory.
1 Cf. KoTTLER [1], Abschn. 60.

§ 4. Calculation of constriction resistances with constant resistivity e 15
The mathematical problern is treated in many textbooks. A nry

clear general presentation of the electrostatic problern is given by
KoTTLER [1]. A profound and elegant, though more difficult exposi-
tion is given by SMYTHE1 . Considering the availability of excellent treat-
ments and the rather protracted details of the calculations, we confine
ourselves to citing the resultant equations.
The basic assumptions are the following. The conducting surface Ac
isaplane ellipse with the semi-axes o: anrl ß. Another equipotential sur-
face A1 is characterized by a certain parameter t-t· In the capacity prob-
lern, one begins by calculating the capacity between an ellipsoid, of
which the ellipse (o:ß) is the major section, and some distant equipoten-
tial surface. Then one allows the axis normal to Ac to approach zero, so
that the ellipsoid changes into a flat disk in the Ac surface.
It is found that the equipotential surfaces in the contact members
are semi-ellipsoids with the equation
(4.10)
where f-l is a parameter and the axes of the ellipse coincide with the
directions of x and y. It is obvious from Eq. (4.10) that the height of a
- ---
semi-ellipsoid (normal to the xy-plane) is ltt and that vo: 2 + f-l is its
axis in the x direction, cf. Fig. (4.16). The capacity between the flat
elliptic base surface A 0 and the semi-ellipsoid (4.10) is, in cgs electro-
static units
[/vl•' + -~~ß' + •>,]_,

Thus with Eq. (4.05) the sought resistance is
JJ
R,.t = R
· P
= _g_f y(oc• + p.)dp.(ß" + p.) p.
4n
(4.11)
0
The current density in Ac, J = ~ ~: , is combined with the density of

the charge in the capacity problernandin the point (x, y) is found tobe
(4.12)
Evalutions of Eq. (4.11) are discussed in problems D and E.

Problem D with verification of the constriction resistance. Assump-
tions as in C, but with o: = ß = a, i.e., a circular conducting surface A 0 •
1 Sl\IYTHE [1] Chapter V, pp. 109-112.

Eq. (4.11) is reduced to

y#
R Lj
#
dp = _g_~~ _f!_arctg V/.i (4.13)

,..
=
4n
0
(a2 + p)tp, 2n
0
a2 + z2
=
2na a
For a long constriction, i .e., for ft = oo, one obtains for one contact
member the exact expression1
R= _g_ (4.14)
4a
Hence for the total constriction including both members
~I (4.15)
R. HoLM and STÖRMER [6] have experimentally verified this for-

mula within error Iimits of observation of ± 1.5% .
Note: Weshall find in § 5 and problern F that the constriction re-
sistance changes only little when Ac ceases to be equipotential, and
rather slowly when Y. und ß deviate from a, conserving their product
aß = a 2 • Because of this, and also due to their simplicity, Eqs. (4.14)
and (4.15) are frequently employed in computations.
The equipotential surfaces are rotational semi-ellipsoids. Fig. (4.16)
shows sections of such surfaces chosen so that the resistance between
Fig. (4.16). Lines of current tlow and equipotential surfaces in the vicinity of a circular a-spot
in a symmetric ccntact. The resistance between two consecutive equipotential surfaces of the
ftgure is 1/6 of the total constriction resistance in one contact member
two consecutive surfaces is Rf6. Their foci, c, lie on the circumference

of Ac.
1 Cf. Problem A of § 5. A direct derivation of Eq. (4.14) using the HANKEL trans-
form is given in LLEWELLYN JONES [4], p. 13-15.
§ 4. Calculation of constriction resistances with constant resistivity (! 17
The ellipsoid halving the constriction resistance has the height a,
that is, it corresponds to f1 = a 2 , as is easily verified by inserting this
f1 into (4.13) giving for one member
(4.17)
The semi-axes of this ellipsoid are aV2, aV2, and a.

On Fig. (4.16) the lines of current flow are marked by numbers
indicating the portion of the total current that flows within an axial
distance smaller than that of the marked solid line of flow. The calcu-
lation of the current densities in the contact surface are based on Eq.
(4.12) which, for a circular conducting area, assumes the form
I 1
J(r) =- (4.18)
2:n:a Va 2 - r2
giving the current density in Ac at a distance r from the center of A,.
Integration shows that half the current flows through a circle with the
radius 0.866a.
ProblemE. Returning to the general problem with IX =1= ß we have
to perform the integration in (4.11). The integral is elliptic and is suit-
ably treated as follows. Putting
f1 =
1- t
1Xß 1 + t
l
and l (4.19)
(X - ß r
rx+ß=k J
!
changes Eq. (4.11) into
R _ e
u - 2:n:(rx + ß)
1
t/(1-
dt
t2) [1- (: ~ ~rt2]
j
«P _,. (4.20)
~ ~:
.. + ß;{;(.) 'F IF (•· '"''" 1= ~ :m I} J
with the same sign before the last term as with P.- rxßß. The symbols
P.+rx
k, F(k, q;), K(k) are those used by JAHNKE and EMDE [1] 1, k signi-
fying a modulus, F(k,q;) the elliptic integral of the first kind, i.e.,
-j }'1-k'sin
tp
F(k ) dq;
,t:p- q; 2
0
1 JAHNKE and EMDE use k = siniX in the tables. This IX shall not be confused
with our IX for one of the axes of the elliptic A •.
and
the complete integral.

In the case of a long con8triction, i. e. {t = oo, one obtains for one
member, with Eq. (4.20)
(4.21)
which is twice the value that is obtained with {t = cxß. Thus the con-
striction resistance is halved by the equipotential surface that has
,u = rxß (4.22)
The current density in Ac is given by Eq. (4.12).
Problem F. The infiuence of the elliptic shape of the contact area on
the constriction resistance expressed by a shape factor. Putting rx = y a
and ß = afy in ProblemE , one obtains the area of the elliptic contact
surface :n; cxß = :n;a2 , equal to a circular area with radins a. Here y is
a measure of the ellipticity. W e now express the constriction resistance
R (cx,ß) for an elliptic surface by means of the resistailce R (a, a)
= ef4a for a circular surface with the same area as the elliptic one by
writing for one member
R(rx,ß) = R(a,a)f(y) = 4eaf(y) (4.23)
The shape factor f, calculated in accordance with Eqs. (4.14) and

(4.21) is shown in Fig. (4.24).
In the case of a great y (i.e., cx » ß and k"-* 1) available tables for
1.0
........
..........
r-..
r-....
s
.........
f(J
?--
["'"--..
/ftJ
--
Fig. (4.24). The functlon f(y), a form factor
F(k,cp) are not suitable due to Iack of small increments. But, the fol-
lowing approximation is advantageously applicable:
1 16
K(k) ~ 2 In 1 _ k 2 (4.25)
§ 5. Constriction resistances when conditions deviate from those in § 4 19
or, since 1 - k 2 "'=' 4fy2 (as is readily checked),
K(k) :~dn(2y) (4-.26)
giving
4
f(y) = -ln(2y)
:ny
(4.27)
While the direct use of Eq. (4.20) is protracted, the calculation can
be much simplified by aid of Figs. (4.24) and (4.28), both based on this
equation. In the latter figure, :(~) is plotted against y with ttfa.ß as a
parameter. Since R(co) = R(aa) f(y) we have
R(p,)
R(tt) = R(a,a) R(co)f(y) (4.29)
where the first factor is obtained from Eq. (4.14), the second from
Fig. (4.28) and the third from Fig. (4.24).
_ An important feature of Fig. (4.28) is that it illustrates how ]arge
VP •the length ofthe constriction, must be as compared with the aver-
age radius, a = ~~aß, of
-
--
the contact spot in orrler
lO
~-tt«fJ - <»
r--:-_wo I
Long co~·lriclinn I
to make the constric- ~
tion long. Even with y - --r---..; :+- -- r-:- --
I~ ~G- -
= 1, i. e., a circular spot,
lflfa has to be larger
than 10 in order to ha ve
R (fl) a pproach R (co) to
---
,- 1-t-
I
!
~
p
I
:----r-- --:.. -
~-1
--
within less than 5 % of ~ - -4-: ::--
- I
I 1 i
R(co); with y = 20 and ~OS
1pfa =
- I
-
10, R ({l) devi-
ates
20% .
from R(co) by
-+ ~
I - -~-
Some disagreements
betwcen observation
' I I I
'
I
I
and calculation in the

I
l Ii
Iiterature result from
observations on short 0
1 z J 5710
I ZOJO SQ
constrictions beingcom- r-
pared to calculations for Fig. (4.28). R(!J)/R( oo) plotted as a function of the a-spot 's
long constrictions. cllipticity expressed by y with IJ/tz {! as parameter
§ 5. Constriction resistances when conditions deviate from those

in § 4, but with Q still a constant
The following problems (A and B) refer to cases when the contact
surface is circular but not strictly equipotential.
2*
Problem A. Cf. Fig. (5.01). The often used designation of spreading

resistance initially concerned the connection between a cylindric mem-
ber, 0 1 , with a circular cross section, and a semi-infinite member, 0 2 , of
the same material as 0 1 . When the ohm was determined as the resist-
ance of a mercury cylinder with certain dimensions, the investigators
employed a capillary filled with mercury, which constitutes 0 1 • The
capillary ended in large electrodes of mercury, 0 2 , The resistance was
measured between these electrodes. It comprises the resistance in 0 1
and two spreading resistances in the
0 2 electrodes. Let 0 1 have t he len gth
l and the cross section na2 • Then the
resistance of the cylinder 0 1 would be
elfna2 , if the current flow were uni-
form all along l. At a cursory glance
of Fig. (5.01) one could expect each
of the spreading resistances t o b e e f4a
a ccording to Eq. (4.14). At the ends
Fig. (5.01). Current tlow near the boun- of 01 the current becomes denser to-
dary between a cylindric conductor c,
and a semi-inßnite body c,. Above the wards the walls. However, the distri-
section a-a, the current density is not
uniform in c, becau•e the lines begin to bution of the current density required
deviate from their straight Course by Eq. (4.18) is only approached . This
means an increase o f the resistance within 0 2 compared with the con-
striction resistance a ccording to Eq. (4.14). The resistance of 0 1 is in-
creased as compared to the case of a straight flow. For the calculation
of the total resistance one may represent the resistance in 0 1 by
elfna2 but then1 , one has to use the formula.
(! (! 10 (5.02)
3.82a = 4a · 5
as representing a spreading resistance; i. e., 5 % greater than the con-
striction resistance according to Eq. (4.14). Certainly, Eq. (4.14) is
correct for a symmetric contact. Those authors who have believed that
the MAXWELL-RAYLEIGH correction should be applied also for the
symmetric contact problern have obviously not considered that Eq.
(5.02) deviates from (4.14) because Eq. (4.18) ist not satisfied in case of
a spreading resistance.
Problem B. The constriction resistance of one member when the a-
spot is covered with a film. We confine the discussion to the c ase of a,
circular contact area A of radius a, uniformly covered with a homo-
geneous film. One might be tempted to believe that the constriction
resistances R1 and R2 illustrated in § 3, remain unchanged whether A
is met allic or covered with a film. But in reality the problem i s more
1 Fir8t calculated by MAXWELL and with improved approximation by LoRD
R-\YLEIGH ; see MAxwELL [1] p. 434.
complicated because the ffim alters the distribution of the current

density from that of Eq. (4.18).
Webegin by assuming that the film resistance is very large com-
pared with the constriction resistance in the metal. Then, the interface
A between metal and film is no Ionger an equipotential surface on the
side of the metal. But, because of the high conductivity of the metal,
all potential differences along this interface are negligible compared
with the gradient within the film perpendicular to the interface. There-
fore, the current density in the film is approximately distributed as if
the boundaries between film and meta] were equipotentials. This means
a uniform distribution of the current in the boundary and thus, in the
electrostatic problem, a uniform distribution of the electric charge in
the interface A.
In order to find the capacitance, C, to be used in Eq. (4.05), we
imagine a non-conducting disc with area A and radius a, tobe uniform-
Iy charged with an electric density, a. The potential differencel between
its center and infinity then is 2naa electrostatic units, and since the
total charge ist na 2 a, the capacitance, C, is af2 electrostatic units.
Returning to the problern of the current, defining the constriction
voltage as the potential difference between the center of A and infinity,
and considering that the current fiow is limited to half the space (semi-
infinite body), we obtain from Eq. (4.05) the constriction resistance
(in one member) with subindex f referring to the presence of a ffim
R c r = _g_
na
(5.03)
which is somewhat Iarger than the value of yf4a according to Eq. (4.14).
If the film resistance is of the same order as the constriction resist-
ance, R1 willlie somewhere between e/na and ef4a, and the film resist-
ance R1 can no Iongerbe calculated for a constant current density with-
in the film. Thus the circumstances have become rather complicated,
but as we shall not require great accuracy when dealing with film-
eovared contacts, only this rough discussion ofthe subject will be made.
Problem C. Multispot metallic contact referred to a semi-infinite
member. To begin with, we simplify the problern by assuming all of
the a-spots to be circular (radii ~. ~. a 3 , etc.) and to lie at distances
from each other which are large compared to the radii, thus permit-
t.ing the assumption of a long constriction for every 8ingle a-spot. Their
1 1
conductances add. Thus R = ~ e/ 4 a. , and
(5.04)
. 1 . The potential at a point on the symmetry axis, perpendicular to the disc, is
eaßily calr.ulated by using concentric ring-shaped elements of the disc and then
integrating.
is the total constriction resistance in one member. In Fig. (5.05) a num-

ber of a-spots is shown nearly equally distributed over a circular appar-
ent contact surface, A,.
Next we suppose that the a-spots lie close to each other in such a
fashion that the constricted lines of fl.ow from different a-spots deflect
each other. Then Eq. (5.04) is no Ionger
valid. The following is a useful approxi-
mation if the distribution of the a-spots
can be regarded as essentially uniform.
The shape of the contact member usually
has a minor infl.uence. Imagine n circular
a-spots, with equal radii a, uniformly
distributed over the apparent contact sur-
face A,. at an average distance of 2l be-
tween neighbouring centers. Then n is ap-
Fig. (5.05). n a-spots dlstrlbuted proximately the number of squares with
over A, supposedly clrcular wlth sides 2l contained in Ar. Thus
radlus r
n = A,f4l2 (5.06)
Now imagine all a-spots tobe covered concentrically by hemi-ellip-
soids given by Eq. (4.10) with cx = ß = a, whose circular bases touch
each other; cf. Fig. (5.07) where the
plane E is a tangent to all ellipsoids.
The resistance between a single a-spot
and its respective ellipsoid is given by
Fig. (5.07). Problem C. Model of a R(a,l) = -2

e- a r c tvz2-
g--
a2
multispot contact conatrlctlon in :na a
a semlln1lnite member
in accordance to Eq. (4.13) with a 2 +
p, = Z2 • Thus the resistance between the surface E and all a-spots in
parallel is approximately
Rc(n,a,l)= -2 Vl21i2
e arctg---
:nna a
Referring to Eq. (3.02) one has to subtract a certain R!b in order to
obtain the constriction resistance RcE (n, a, l) in the volume between A,.
and the plane E. We therefore write
R E(n a l) =
e a r c tvz-
-- g -- -
a 2 VF -- -a
-- 0 6 n
2
(5.08)
c ' ' 2nna a ' o: A,
where the coefficient 0.6 is chosen so as to make RcF: (n, a, l) approaching 0
for z...... a without assuming any negative values.
Eq. (5.08) is approximate, particularly because the current is not
uniformly distributed among the a-spots but the spots are all the more
favored the closer they lie to the rim of A,; cf. Eq. (4.18). In Eq. (5.08),
§5. Constriction resistances when conditions deviate from those in § 4 23
this effect could cause an error of up to 20 % when the spots are very
close to each other; but the error decreases with increasing lfa. The un-
certainty is taken care of by the shaded area in Figs. (5.12) and (5.13).
Problem H provides the opportunity to compare Eq. (5.08) for n = 1
with an exact calculation.
However, since the height l/12 - a 2 is a microscopic length, the vol-
tage between A, and E can not be measured directly. Assuming the vol-
tage probes to be at a distance from the contact surface that is great
compared with the diameter of A,, the measurement involves a long
constriction on the basis E ""' A,. Therefore, with
A,. = nr2 = 4nl2 (5.09)
according to (5.06), a resistance ef4r has tobe added 1
to Rcr;(n, a,l).
The total constriction resistance in one member resulting from the
group of a-spots on the Lase of a semi-infinite member then is
(! Vl2-a2 eVP-a2
R(n,a,l) = --arctg
2 nna ---0,6
a
A
r
+ 4-(!r (5.10)
Actually the a-spots are seldom circular; but their shape can be
taken into account by calculating aR if they were ellipses, characterized
by a certain y in ac-
-t-
lO
cordance with Problem I I I I
F in § 4. Then the first
term on the right side
of Eq. (5.10) becomes
Oß\ I
I I
I
I
I
I
I
I
fl& ~
I i :
gj_('Y) arctg vz2 - a2 ~~
2nna a
(5.11) 09 ~ IQ:~ I I
~~ n-10
Problem D. We con-
sider an apparant area f1Z
~~
A,. on which equal circu- --= """" I 700
lar a-spots are uniform- 0 I

ly distributed, and com- 2 J 5710 20JO so 70 700
l/a,-
pare the constriction re-
Fig. (5.12). Problem D. The constrlctlon resistance R (n, a, l)
sistance in this case with of an assembly of n circular a-spots (average radins = a, dis-
the constriction resistance between nelghbouring centers = 2l) is compared with
the constriction resistance 1!14 r 0 of a circular area = the sum
tance when the spots are R(n, a , l)
of the a-spots. ~ referred to one ·member, is plotted
unitedintoatotallycon-
against l!a with n and a remalning constant, and wlth l varying
ducting area A 0 = nrij.
The distance between the centers of two neighbouring a-spots is termed
2l. The comparison is facilitated by Fig. (5.12) where R(~~a,l) is
(! ro
plottedas a function of lfa with n as a parameter.
1 The corresponding R~b according to Eq. (3.02) is zero for a long cdnstriction.
Fig. (5.12) demonstrates the influence on the constriction resistance

of disuniting the contact area A 0 into n equal a-spots and increasingly
spreading them.
Problem E. Consider a given area nr2 , first totally conducting with
the constriction resistance e/4r in one member. Then assume that uni-
formly distributed surface asperities Iimit the conduction to n a-spots.
The a-spots are regarded as circles
(radius a) with an average distance 2l
between the centers of neighbouring
a-spots. Thereby n and l sa.tisfy Eq.
(5.09) . R(n, a, l) is then the long con-
striction resistance of the assembly of
~
~J ~~~~~~~~--~~ the a-spots in accordance with (5.10).
<:3" Fig. (5.13) gives the ratio R (ni a, l)
~
'>::: z~--I-,.Qg4J!---+~~::::.._~-I e 4r
plotted against lfa with n as a para-
meter. Evidently l remains constant
but a decreases. The ratio shows how
the constriction resistance connected
z with the given area nr 2 is increased
0 b 8
Zfa____,.- by the asperity.
BARKAN and TuoHY [I] also treat
Fig. (5.13). ProblemE. Constriction re-
sistance R (n, a, l) of an assembly of n this problern but with a rougher ap-
circular a-spots (average radius = a, dis-
tance between neighbouring centers proximation. In addition, they point
= 21) uniformly distributed over a cir- to the fact that the area enveloping the
cular area with radius r. R (n, a, l) is
compared with the constriction resis- a-spots in the case of rough surfaces
tance e/4 r . R (n, a, I) referred to one extends slightly beyond the HERTZ
e/4r
member, is plotted against l/a area that is calculated for ideally
smooth surfaces, elastically deformed.
Problem F. In this case, the constriction resistance R (n, - a, l) of
a contact area is interrupted by n equal and insulating circular spots .
This is the conversion of ProblemE. In Fig. (5.05) now the small
circles represent insulating spots and their surroundings are conduc-
ting. In order to find the solution of the problern we use the mathe-
V
matical trick of expressing the conductance A,fg l 2 - a 2 of the cylinder
with the height }l 2 - a 2 on the base A, (see Fig. (5.07]) as the sum of
2nna [
the conductance -e- arctg --a-
VP- a2]-1 through the spots and the con-
ductance through the intermediate area. Calling the latter conductance
1/Ri (n,- a, l) we have with a somewhat rougher approximation than in
Eq. (5.08),
1 = A, - - [ arc t g Vz2
-2nna _ a2]-1 (5.14)
R!{n,- a,l) eVl 2 - a2 e --a-
and for a long constriction (cf. Eq. [5.10])

R ( n, - a, l) = R;( n, - a, l) -
eVlz- az + -e (5.15)
:n:r2 4r
The ratio ~ a, l) tells of how the constriction resistance is in-
R(n,
e/ r
creased when the conductance looses the help of the spots considered.
Fig. (5.16) illustrates the results for
lfa > 1.5. The curves are not drawn
3.0
I
below lfa = 1.5 because there the ac- I
curacy of these simplified calcula-

I I
tions becomes unsatisfactory. Any
I
I I
curve is obtained by varying a only;
cf. Probiem E.
i l
Fig. (5.16) shows that insulating
I I
\ '' I
spots small enough to make lfa > 4 I
to 7, have practically negligible in-

fluence on the constriction resist-
\.\1
['.... ;.~
zo I '2SO
ance. I
Problem G. Distorted constric· 0 G 8 70

l/a -
tions. The pro blem concerns a thin
disc-Iike contact member (illustrated Fig. (5.16). Problem F. The constriction
resistance of an area (radius r) interrupted
in Fig. [5.17]), instead of an extend- by n insulating spots (average radins ~ a,
ed one sketched in Fig. (4.16). The distance between neighbouring ecnters
R(n, -a,l)
~ 2l) is called R (n, - a, l). 1114r
disc has a circular contact area of
referred to one member, is plotted against
radius a, similar to that in Fig. l/a
(4.16), is a~3 thick and 6a broad.
The contact area is located in the center of the lower disc surface. The
current ftow lines bend to the right and finally continue with uniform
density. Now imagine Fig. (4.16) to be placed on Fig. (5.17) in such
a way that the contact areas
.!:!
coincide. The upper surface of -.;..-------...---.--.---,,-.---
the disc would then touch the ~ ~
ellipsoid of Fig. (4.16) which ~
is characterized by ft = 3a2 • ;J.._____ ~~~:-'-'""---'--'--'---'--
The equipotential surfaces
in Fig. (5.17) are no Ionger Fig. (5.17). Problem G. Equipotential surfaces and
ellipsoids but are deformed,
lines of rucrent flow in a distorted constriction
and the formulas for Iong constrictions are, of course, no Ionger valid.
The calculation of a deformed constriction resistance may be ob-
tained by an approximate, step by step, construction of equiresistant
space elements 1 , bounded by current tubes and equipotential surfaces.
1 The practical details of such constrictions have been described, for instance,
by KÜPFMÜLLER [1] § 19.
With l the average length and A the average cross section of such an
element, it is required that all elements satisfy lfA = C where C is a
chosen constant. If n such elementary tubes intersect an equipotential
surface (the same tubes intersecting all equipotential surfaces) and if
there are rn elements along each line of flow within the constriction ,
then the resistance of the distorted constriction is
(5.18)
where e is the resistivity of the metal.

Problem H. A special distorted constriction has been treated with
exactness by RoEss [1]. It concerns the contact between two cylinders
in a central a-spot as illustrated in Fig. (3.07). The cylinder walls
cause the system of flow lines to approach the type of parallel lines
through a totally conducting Aa corresponding to a vanishi.ng constriction
resistance. Let R(a, B) be the constriction resistance, for one member,
whichforafB->Oisef4a and fora = Bis zero. For intermediate values
of afB =
0.1 0.3 0.5 0.6 1
according to RoEss:
e2 1
1.31 3.81 6.85
nB R(a, B) = 0 ·296
00
and with R (a, B) = RcE according to Eq. (5.08) with n = 1, l = B

and Ar= nB 2 :
0.296 1.29 3.79 6.82 00
Problem I. Another kind of distortion appears when the current

enters an infinite slab (thickness h, resistivity e) through a circular
contact area A 1 (radius a), with the bottom surface A 2 of the slab in
perfect contact with an ideal conduc-
tor; cf. Fig. (5.19). Let R(h, a) be the re-
Za-
sistance in a slab characterized by h and
Ideal
condvclar 1 a. Approximate values R(h, a)fR(oo, a)
can then be obtained from Fig. (4.28).
Fig. (5.19). Problem I. Current ftow The following table compares such calcu-
through a slab (thickness h) entering
through the citcular area A1 (radius a), lations with measurements by FoxHALL
leaving through the infinite ba.se A 2 and LEWIS [J] :
h/rl 10 5 2 1 0.5
R (h, a)f R (oo, a) accord. to Fig. (4.28) 0.94 0.88 0.7 0.5 0.3
accord. to FoxHALL et al. 0.96 0.90 0.80 0.64 0.43
§ 6. Introduction to thin films on contacts. Contact cleaning 27
§ 6. Introduction to thin fl.lms on contact!'l. Contact cleaning

Several following chapters are essentially devoted to perfectly clean
contacts, but some short passages will refer to the beha vior of thin,
contact films. The main treatment of tarnish and other eontact films
appears in § 23-26.
The thin films now to be considered, particularly oxygen deposits,
are of two kinds: physisorbed and chemisorbed.
Physisorbed particles (atoms or molecules) are bound to the metal
by means of v AN DER W AALS forces with fee ble bonds of the order of
0.05 eV. They are easily rubbed away at contact make, are to some
degree volatile, and therefore can be pumped away at room temper-
ature.
Ohemisorbed atoms are bound to Yalences of the metal surfacc atoms
by covalent bonds. In addition, the atoms may carry charges and be
ionically bound. The bond strength is 1 to 8 eV. Two groups of chemi-
sorbed atoms can appear on a surface, one stronger bound than the
other. The difference is obviously caused by different sites on the
pattern of the surface lattice. Chemisorbed films endure some friction
and elevated temperatures without being removed from the meta! sur-
face. However, they are not thicker than about 10 A and therefore
easily penetrated by means of the tunnel effect (see § 26). As seen by the
slopes of the associated solid and dashed curves in Fig. (8.01), the
degree of influence of the tunnel effect on the contact resistance becomes
smaller with increasing load P, aud is for crossed rod contacts practi-
cally negligible at P > 10N. But, the influence of these thin films on
cold welding and friction can be considerable.
It evidently is importaut to define the concept of a clean metallic
contact and to have reliable methods to accomplish t.he cleaning.
Clean metallic surface; definition

With respect to contacts, a metal surface is "clean" when it does
not contain any contaminant that perceivably influences contact tests;
viz. measurement of electric contact resistance, cold weid or specific
friction force. This requirement is harder to accomplish, the smaller the
Ioad. In any event, "clean" means that t.he surface shall not be covered
with more than a small fraction of a monofilm and, of course, carry no
alien insulating particles, as dust.
M ethods to test the cleanness

A. When the Ioad bearing area Abis known and is a circle (example:
cross rod contact under certain conditions at not too small P) essential
cleanness is stated if the measured contact resistance equals the con-

striction resistance associated with Ab according to Eq. (4.15).
B. A friction coefficient of about f = 1 in air is an often used cri-
terion of a fairly acceptable cleanness, cf. § 37.
C. At a cleanness similar to B, water wets the metal surface, viz.
water drops spread on it.
Notice that oxygen from the air interferes so rapidly with clean
surfaces that cleavage of graphite is much easier in air than in vacuum1 ;
cf. § 39A.
A high degree of cleanness can be achieved in vacuurn alone.
Test methods are described in RoBERTS [1]. One sensitive pheno-
menon, for example, is the emission of electrons indicating whether
the work function corresponds to the clean metal, and also secondary
electron yield.
Cleaning metho ds
Grease and lubricants are removed with the aid of acetone, carbon-
tetra-chloride and trichlor-ethylene (not as toxic as CC14 ). So much
liquid must be used that the fatty solution becomes very diluted. Final
eleaning is accomplished by washing in ethyl alcohol, and rinsing in
tap water, or, better, in boiling distilled water.
Among the several recommended procedures for combined mechan-
ical and chemical cleaning we cite 2 : Polish with 1/4 mi9ron diamond
powder or alumina; extract with benzene for four hours in a Soxhlet
Extractor; soak for 15 minutes in hot chromic acid; rinse in destilled
water; heat to 600 to 635 oc in helium atmosphere; use immediately
to prevent contamination by the ambient atmosphere.
The cleanness of the surfaces and the density of the adsorbed films
e
were determined by measuring the contact angle of methylene iodide
and water. For clean palladium0max = 0 to 3°. For organic monolayers,
Bmax was in the range of 37 to 93° depending upon the materials and
eoherence of the coverage.
Another fairly intricated method for cleaning is recommended by
BLAKE [1].
For noble metals strong etching processes may be used. CHAIKIN [ 1]
recommends for palladium, treating in cold aqua regia for 15 seconds,
rinsing in 1 : 1 hydrochloric acid, and then rinsing in distilled water;
this followed by a 15 minute treatment in sulfuric-chromic acid clean-
ing solution at 80 °C. Finally, after a thorough rinsing in double-
distilled water, the contacts are dried in a desiccator containing magne-
1 BRYANT et al. [1].

2 Recommended by J. R. ANDERSON, Stanford Research Institute.
§ 7. The load bearing contact area as a function of load 29
sium perchlorate. The gold cleaning procedure may be similar except

that warm (50 °C) aqua regia is used, the hydrochloric acid step may
be eliminated, and the contact oven-dried at 110 °C.
In many cases cutting the surface with a clean tool has been used.
However, CHAIKIN has shown that an iron (or copper) tool can leave
invisible traces of its metal on the treated surface. These may oxidize
and cause trouble in micro contacts.
H eating1 in vacuum. Atomically clean surfamis of high melting point
metals can be generated by heating to temperatures, at which surfacc
contaminants vaporize. It has been shown 2 that heating wolfram to
2200 °K for several seconds produces a clean surface, provided no
contaminant is soluble in wolfram.
Ultrasonic cleaning 3 • Ultrasonic action provides efficient means to
bring soluble and weakly adhering eontaminants into solvent fluids.
The action penetrates into crevices and pores. But, as a certain redepo-
sit occurs, it is necessary to repeat the cleaning process in renewed clean
solvents.
Cleaning before assembly. MAHLER [1] describes bis great experience
on making sealed-in contacts. He recommends: 1. to free every part
from contaminants before its assembly; 2. to heat-stabilize organic
components (if they are unavoidable) before they are used in the
assembly; 3. to remove all impurities before the switch itself is placed
in the bousing; 4. to conduct all steps under aseptic conditions.
§ 7. The Ioad bearing contact area as a function of Ioad and

elastic and plastic properties of the members
A. Introduction. The theory of indentation in § I treats the forma-
tion of calculable load bearing areas, Ab, for members with smooth
surfaces and cases when Ab is a circle or an ellipse, particularly the
cases when the members are spheres, including a sphere and a plane,
or crossed cylinders.
Imagine the indentation to be produced purely elastically at suffi-
ciently smallload. Fora circular Ab, the radius is given by HERTz's
formula (I, 1); for ellipses, formulas are given in RoARK [1]. These for-
mulas are deduced for ideally smooth surfaces. Real surfaces have
microscopic elevations and depressions, with tbe deep depressions re-
maining as voids in the contact. Those usually much spread voids
1 Cf. RoBERTS [1].
2 Cf. HAGSTRUM et al. [2].
3 Expert rules for the operation are given in ScHROEDER [1] and L. K. JONES
[1]. A description of the method and action is given in the article "Ultrasonics" in
Encyclopedia of Electronics, Reinhold, New York 1962; cf. McCornick [1].
have a negligible influence on t.he contact resistance 1 • This is one reason

why the formulas for smooth surfaees can widely be applied in this book.
When HERT'z formulas are applied to single asperity elevations,
the concern ean only be a rough computation since the curvature can
not be well known. It is described in § I how plastic deformation begins
and develops when the Ioad on a contact with simple geometry increases
and finally produces a "specific depth" of 0.03 or more. Then, the
average pressure p attains H, the hardness. Interesting pictures of
contact patterns with voids have been published by HöFT [1] and
GREENWOOD et al. [8].
B. Contact between nominally flat contact members, for example a
carbon brush on a collector.
If the members were perfectly flat, there would be no constriction
resistance. But actual surfaces have certain roughness (humps and
grooves) and certain waviness (with macroscopic radii of curvature).
At contact make, protuberances of any one member indent in the other
member, and so the formation of contact areas resembles the formation
of indentations described in § I.
At small average pressure, p, the indentations may be formed
elastically. With increasing p, more and more indentations become
plastically produced. Finally, nearly all indentations could have spe-
cific depths, D, of 0.03 or more. Then the Ioad bearing area Ab would
satisfy
(7.01)
with H chosen so as to count for probable strain hardening.
However, this is an extreme and improbable situation. Actually,
when some indentations deepen, other protuberances obtain the oppor-
tunity to make contact. These initially generate shallow, elastic inden-
t.ations. The average pressure will be smaller than H, say
(7.02)
with ~ < 1. Hence
(7 .03)
cf. Eq. (1,17).
Theoratically, any value of ~ between 0 and 1 is possible; but accord-
ing to measurements, values between 0.1 and 0.3 aremostfrequent for
apparent contact pressures which arenot too small, say for
p = loadP > 1QSN/m2
apparent area A.
R. HoLM 2 has reported ~-values as small as 0.02, obtained in a
carbon-carbon contact after extended polishing ofthe members against
I Cf. Fig. (5.16).
2 R. HoLM [37] p. 35; cf. R. HOLM [1] p. 323.
§ 7. The Ioad bearing contact area as a function of Ioad 31
each other, steadily back and forthin the samestraight path. Recently,
CLAUSING and CHAO [1] attained g < 0.02 between extremely smooth
contact members of stainless steel which had a curvature radius of
between 50 and 100m. The Ioad was about 1000 N.
In § 36B the explanatioil of CouLOMB's law of friction is discussed.
Thereby the question arises whether g can be independent of P . .As to
this particular problern we point to the fact that ~ is proportional to
the average pressure p (see 1,17 and the remark connected with this
equation). Therefore, g is a function of the specific depth D of the in-
dentations constituting the Ioad bearing area. Thus, a constant g means
a constant average specific depth D. Since nature evidently presents
cases of D = constant, it is important to investigate relevant condi-
t.ions.
According to Eq. (I, 17) a constant g is equivalent to Ab oc: P.
GREENWOOD and WILLIAMSON1 have investigated a mathematical
model of a very general type that satisfies Ab oc P. They characterize
the unevenness of a surface by asperities each of them rcpresented by
its height, z, and the radius, r, of curvature of its top. They show by
profilometric measurements that usually a Gaussian distribution of z
is essentially realized. The distribution of r is skew.
The authors assume a contact between nominally flat members
{without macroscopic waviness) with the asperities so far apart that
they deform independently of each other. When the higher asperities
deform, lower ones touch, and a variety of contact areas with different
sizes is produced. It is shown that probably both the nurober of spots
and the total contact area are fairly proportional to the Ioad P, whereas
the density of the spots is proportional to the apparent pressure PJA.,
where Aa is the covered (apparent) area.
The authors introduce the concept "plast.icity index", lJ', that we
slightly modify to
with a = standard deviation of z, and

2 1 1
and H = H1 + H2
where E = YoUNG's modulus of elasticity and H = hardness. Indices
refer to the member 1 and 2. When lJ' < 1 and the apparent pressure
< 103 Nfm 2, all deformations are practically elastic.
1 J . .A. GREENWOOD et al. [4], [5], [7], [8], where earlier contributions, parti-
cularly that of ÄRCHARD [2] are discussed.
Remark referring to § 350 and Fig. (35.10). There, it is madeevident

that, in clamped contacts with normal asperity, the a-spots are clustered
in a preferred region with the effect that the constriction .resistance is
defined by P largely independent of Aa and of the usual degree of
polish.
Estimates of the amplitudes of the surface asperities after different
surface treatments (mostly steel) are given in the table below, based on
data by GREENWOOD and WILLIAMSON [4], and D'YACHENKO et al. [1].
The standard deviation for amplitudes on any sample was generally
about 25%.
Average amplitudes, z, of asperities in w-& m when sample is

polished and dressed clean drawn ground turned punched
with cloth wheel
0.1 1 6 10 100
The slope of any asperity is probably not greater than 5 degrees,

which means that the average radius of curvature of an asperity is a
few to 100 times its height z.
C. Experimental determination of Ab. 1. Measurement of Ab when
it is a circle or an ellipse. Satisfactory determination of these Ioad
bearing areas is obtained microscopically when the Ioad P is high
enough to produce remaining indentation; cf. § I.
At small P, generating a merely elastic indentation, Ab may be
calculated from observed electric resistances, provided that Abis per-
fectly clean. Measurements of clean contact.s in vacuum are represented
in Fig. (8.01). They confirm the formulas (1,1) and (1,16) with an accu-
racy of a few per cent. Unfortunately, many other measurements that
have been since published have not been made with requisite consider-
ation of contact disturbances by alien films.
2. Measurement of Ab in nominally flat contacts is difficult, and al-
ways involves considerable uncertainties. This means that even the
better data on the factor ~ ofEq. (7.03) are afflicted with uncertainties
ofup to 30%.
Weshall now describe two methods ofmeasuring Ab on a nominally
flat contact. They are typical and the most advanced that have been
published to date. The difficulties to be overcome will be evident.
R. HoLM1 tried to base the determination of Ab on measurements
of contact resistances. Primary difficulties arise from the necessity to
determine three independent variables: number, n, size and shape of
the contact spots. In addition, the influence of contact disturbances
have to be considered.
1 R. HoLM [22]. particularly [23] p. 44.

§ 7. The Ioad bearing contact area as a function of Ioad 33
In order to secure a constriction resistance much larger than any

likely remaining film resistance, HoLMchose a contact with at least one
carbon member. His method of determining n required a sliding con-
tact, and so he was led to investigate an electrographite brush against
a copper, silver, or electrographite ring. The thoroughly cleaned and
ground copper surface was 8lightly amalgamated in order to be least
disturbed by oxide films.
The number n was determined by a statistical method of counting
the a-spots with probes fitted into the brush and insulated from it. The
probes were cylinders of the same material as the brush; they were
cemented into holes in the brush and ended in the sliding surface. Thus,
it was important to have the insulating cement so rigid that all motion
relative to the brush was avoided. The currents through the probes as
weil as the current through the brush itself were recorded. A probe
carried current as long as it had contact in at least one a-spot. There
were intermittent currentless intervals. From the probability of such
Fig. (7.04). Record of the currents in three insulated probes in the brush and of the current in the
matrix brush
intervals, the average number of a-spots in momentary contact with

the probe can be calculated if the a-spots are uniformly distributed
over the sliding surface of the ring. Therefore, only registrations which
indicated a fairly uniform distribution of the a-spots, by giving the
same mean result for the three probes in the same pf'obe..-- - - - - ,
brush, were employed for calculations. 1 O
Fig. (7.04) shows a part of an accepted regis-
tration, containing the currents in the probes 1, 2 8 0
and 3. On top of these records is the current in the J O
rest of the brush. One sees that the probes had
currentless intervals and, in transient moments, Fig. (7.05). Position of the
carried the entire current since the brush current probes in the brush
had then decreased to zero. The average probability of no current in
a probe was 0.69. At the top of Fig. (7.04) is a 50 cycle timing wave,
showing that the ring rotated very slowly. The position of the probes
in the brush is illustrated in Fig. (7.05).
As for the shape of the spots, HoLM estimated them to be on an
3U Electrlc Contacts, 4th Ed.
average three times Ionger than wide. Thus, y = V3

giving /(y) = 0.9
according to Fig. (4.24).
The contribution of the copper member to the resistance can be
neglected. For the constriction resistance in the brush, one can use
Eq. (5.04) since the a-spots were sufficiently distant from each other
to keep the mutual disturbance of the constrictions negligible.
Hence, since n had been determined, the formula for the measured
contact resistance to be used was
R = 0.9e (7.06)
4na
with (! = resistivity ofthe brush material, and f(y) = 0.9.

Now, with R measured and n determined, Eq. (7.06) gives the
average radius a of the a-spots. One could be tempted to calculate the
load bearing area as Ab = Ac = nna2 ; but HoLM showed (see the orig-
inal paper) that, with respect to probable differences in the size of
the a-spots, a more likely value is
Ab= Ac= 1.1nna2 (7 .07)
Hence, the average presstue p in the contact was
_ 0.9P
P = nna 2 (7.08)
HoLM's final results are summarized in Table (7.09). The data vary
considerably. This may be a consequence of the complicated nature
of contact formation and, on the other band, may lie in the meas-
uring method. The contact resistance between brush and ring
Table (7.09). Number of a-spots on A. and their mean diameter in a contact o.f an
electrographite brush on a copper or graphitering (No. 5).
Brush material is characterized by e = 4.31. w-• Q m and H 1 = 2. 108 N/m 2
Worn-in
Dia- Average Number
area Contact Contact Diameter
meter number of of Contact
of the Ioad resistance of a-spots
No. of the brush p a-spots in a-spots pressure
probe R 2a in.A 6
Aa contact
with probe
cm cm 2 N n 10- 3 cm n 10• N/m'
1 0.38 2.1 11.0 0.012 to 0.015 1 9 to 7 18 0.9 to 1.3

2 0.38 1.8 11.0 (0.018) 0.7 10 11 1.2
3 0.38 1.7 5.0 0.03 0.37 12 5.5 0.8
4 0.27 1.5 4.35 0.025 0.43 7 11 0.93
5 0.5 2 6.0 0.022 0.44 12.6 7 0.63
1 The hardness noted in the original paper was measured with too shallow an
indentation. Gonverted to the right specific depth, the hardness was about 2 • 108
Nfm2.
§ 7. The load bearing contact area as a function of load 35
varies largely, occasionally even in the ratio 3 to 1, depending on the

position of the brush on the ring. Table (7 .09) contains mean values of
the lowest resistances with the greatest deviations being omitted for
they were presumably caused by contamination of the surfaces. The
number of a-spots under the brush also varies from one position of the
brush on the ring to another.
According to the table, the av~rage pressure p lies between 30 and
60% of the hardness , which means ~ = pjH = 0.3 to 0.6 in Eq. (7.02).
The contact resistance R was measured with a ring at rest and for
the determination of n , the speed was about three revolutions per
minnte. As for the method of preparing the ring1 we refer to the orig-
inal paper.
lnvestigations by DYSON and HIRST [ 1] provide a very conspicuous
picture of the Ioad bearing area Ab on ground metal and support the
results by HoLM that the a-spots are numerous and small in size; cf.
Figs. (7.10) to (7.13). But, for a determination ofthe size of Ab the pres-
ent accuracy is unsatisfactory, as will be discussed below.
DYSON and HIRST employed a disc of glass with two polished faces,
one side coated with an opaque metal film. The metallized side is placed
in contact with the specimen being studied and is examined through
the glass by a microscope with phase contrast illumination. At the spots
of contact the glass is deformed and with positive phase contrast these
spots are revealed as bright areas. This is shown in Figs. (7.10) to (7.13),
all with a magnification of 69 times. Figs. (7.10) and (7.11) show long
Fig. (7.10), refers to groundsilversteel Fig. (7.11). Same as (7.10) but with 30 K
( = 1 % carbon steel). P = 20 N
narrow Ioad bearing spots on a specimen of ground "silver steel" (high

carbon steel). A number of contact spots (the figures show only about
one tenth of the total spot-carrying area) appear distributed along
1 The observations with a graphitering are communicated in R. HoLM [30] § 17.
ridges of grinding marks. Fig. (7.12) reveals a HERTZ area 1 ofa polished
contact surface whose size (according to the authors) reasonably agrees
with what is calculated with elasticity theory.
Fig. (7.12). Polished contact surface of Fig. (7.13). Steel polished on corrosil finishing
silver steel. P = 20 N paper. P = 30 N
Fig. (7.13) of a specially prepared steel specimen is comparable to

Fig. (7.11) but exhibits a much greater nurober of very small spots,
dustered together within a kind of HERTZ area.
The accuracy of the determination of Ab may be estimated as fol-
lows. It is known that Ab is proportional to F« with ~ ~ 1 . But the
total area of the photographed spots seems to be proportional to pt. 6 ,
thus with IX> 1.There may be two reasons for the discrepancy. As the
authors remark, the immediately surrounding region ofstronglystressed
Fig. (7.14). Due to A. J. W. MooRE [1]. Taper section of a deep indentation in originally annealed
grooved copper, showing the persis tence of the grooves in the indcnt.ation
1 As to HERTZ area, cf. Section D and Eq. (1,1).

§ 7. The Ioad hl'aring contact area as a function of Ioad 37
spots will also be deformed so that the true size of these spot~ may
be smaller than the revealed areas. On the other band, it seeroH that
slightly stressed spots do not show up at all.
D. Persistence of asperities in indentations. Fig. (7.14) shows a rc-
markable persistence of asperities during plastic bulk deformation of
originally annealed copper. That the asperities arenot totally flattened
is understandable if one considers that they yielded plastically with
about the same percentage of their height as the underlying material
does with respect to its much greater thickness. In addition, the asperi-
ties strain-harden during the deformation, thus increase their persist-
ence. A similar picture was obtained with copper originally strain-
hardened to ultimate hardness. If the thickness of the sample is com-
parable with the amplitudes of the asperities, the entire sample is
flattened by a high Ioad; see GREENWOOD et al. [5].
Fig. (7.15a) illustrates a HERTZ area with a-spots on asperities.
The shaded circle represents elastic deformation in a cross-rod contaet .
Because of the asperities, real contact exists only at
their tops, sketched as black spots. The tops are re-
garded as plastically deformed and hardened. They
actually carry the Ioad P and are carried them-
selves by elastic counter forces that require a de-
•
formed area of size according to HERTZ equation i.L
(1,1). We therefore call this pot"ential are<~ a HERTZ
area. It is obvious that here the HERTZ area (shaded)
represents the apparent. contact area whereas t-h<'
h
Ioad bearing area Ab is the sum of the black arcus.
The HERTZareais not strictly given by HERTZ equa- .Fig.(7.t5).Twot;ve-'
of contact. with the
tions because asperities can touch the other mem- same size of the Ioad
her outside the ideal H ER1'Z area where the ga p is very bearing area (black)
bnt different clastic-
smalJ: see GREENWOOD Pt al. r7]; HARKAN et al [1]. ally strained areas
Fig. (7.15b) shows a continuous contact area for- (shaderl)
med by total plastic deformation with a size equal to thE' sum of the
black areas in Fig. (7.15a). It may be regarded as produced between
rods thinner than those for thP upper figure.
E. Influence of t.emperature and contact duration on the contact. area.
As is described in Appendix I, the measured hardness, H, depends 011
the temperature, and on the time of the stress due to continued ther-
mal diffusion of atoms under the influence of the stress. Table (7.16)
shows how the constriction resistance, R, slowly decreases with time.
The essential cause for this decrease will be that the strain in the struc-
ture, produced around the contact by the stress at contact make, more
or Iess vanishes, enlarging the contact surface. Table (7.16) shows that
3E
silver contacts attain a rather permanent resistance after some days.

Various other metals (not zinc; cf. §22,E and §I,H) behave similarly.
Table (7.16). Change of oontact reBislance R 1 with timet
Contact Observation-
Contact Ioad P temperature (J t = age of the contact in hours, and resistance B in
N •c w-• !.l
Ag-Ag 350 18 f= 0 0.55 144 1100
R= 17 16.2
15.84 15.2 -
Ag-Ag 100 20 R= 34.7 33 31.5 29.5 28.3
Ag-Ag 350 165 t= 0 4 9 72 210
R= 22 14 12.5 9.5 8
Ag-Ag 100 18 t= 0 1 24 360 1200
R= 34.7 32.6 30.7 28.9 28.3
Cu-Cu 350 150 t= 288 960 2640 - -
R= 37 34.3 32.6 - -
In Table (7.17) the following notations have been used. The hard-
ness H({}a, {}, t), measured as PfAb is expressedas a function ofthe va-
riables {}a, {}, t. Here {}a means the temperature at which the specimens,
two cylindrical rods, were annealed before being brought into contact.
They were kept at the temperatures {} a until hardness tests indicated a
permanent state. Then
Table (7.17). Change of hardness of eilver with time
after having attained the
(hours)and temperature
observation temperature {},
{}. I Observation-~
temperature (J H((J
a. •
1160)
(J the
H((J 0 , (},
specimens
co) were
oc •c 10• N/m• 10• N/m2 brought into contact and
left in contact for the time
18 18 8.2 6.8
130 18 7.8 -
t. Finally the contact was
130 130 - 1.45 opened and the indenta-
165 18 6.3 - tions measured microsco-
165 165 4.5 0.72 pically, from which the
:WO 18 4.6 - noted hardness H({}a, {},
200 200 - 0.35
t) was calculated. The
symbol t = oo means that members have been in contact at the
temperature {} for such a time t that a practically permanent contact
resistance was observed with a small current. The contact duration
before the initial reading was one minnte.
Ta ble (7.17) shows that the initial hardness (after the preceding heat-
ing) diminishes from 8.2 to 4.6 108 Nfm2, if {}a is increased from 18 to
1 Agitation of the contact was essentially eliminated. Without precautionagainst
agitation, the contact resistance varies more rapidly particularly at a small Ioad.
The variation of the resistance of tungsten contacts with time observed by FuxuBOI
and MuTO (1] was probably accelerated by small vibrations.
§ 8. The relation between contact Ioad and resistance 39
:wo oc. A considerably greater reduction has been noted for t = oo,
namely, from 6.8 to 0,35 108 Nfm 2 •
F. Special application of Eq. (1,23) concerning creep in contacts.
Three crossed silver rod contacts were investigated each under a con-
tact Ioad of 350 N at the temperatures T = 473, 438 and 403 °K. One
observed a steady yielding of the contacts. This is explanable by
atomic diffusion.
Weshall compare "states" which may be defined by the size of the
contact area which is the mouth of an indentation. It is just as weH
accomplished by the contact resistance, mathematically reduced to
a chosen environment temperature {) by the formula 1 : rxiJ· During
Observation at the temperature Tl °K, Iet R,.(Tl) be the initialresistance
and Rb(T1 ) the resistance after t1 hr both reduced to the chosen temper-
ature. We represent the change between the states by the number x of
atoms displaced among x 0 atoms belonging to the initial indentation. In
another experiment, at T 2 °K, the respective change of the reduced re-
sistances is from Ra ( T 2 ) to Rb ( T 2 ) in t 2 hr. The experiments are made
in such times that
Then we may as ume that x0 fx is the samein both cases. Hence apply-
ing Eq. (1,24) with v = constant, we obtain
11 600 11 600
---<p ---rp
t1 e T, = t2 e T2
or
(i.18)
Table (7.19) contains values of observed t and T and calculated

values of q;; for example q;23 is obtained by inserting t2 , T2 and t3 , T 3 in
Eq. (7.18). Note that the rate of the process is strongly sensitive to the
temperature. The variation of the q;-values is probably essentially due to
observation errors and less to the approximations involved in Eq. (7.18).
Table (7.19). Same amount of creep meaBured by the reduced ratioR1/Rz, taking place
at different temperatur68 T°K requiring tim68 t hr; rp in eV iB calculated UBing Eq. ( 7.18}
Observation I T, e, '1'12 T, e. '1'23 I Pa ta
1 473 0.25 1.38 438 4.0 1.42 403 99.0

2 473 0.47 1.46 438 8.5 1.46 403 230.0
3 473 6.0 1.57 438 138.0
The q;-values in Table (7.19) Iie between those given for silver by
SEEGER [ 1]p. 4 72: 2.0 e V for self-diffusion and 1,2 eV for diffusion of
vacancies.
Stationary Contacts
§ 8. The relation between contact Ioad and resistance,

particularly at moderate and high Ioad
A. Introduction with description of Fig. (8.01). In a clean contact
between snfficiently smooth members in the shape of two balls, a ball
and a plate, or two eqnal cylinders that are crossed, the load bearing area
Ah = Aa conforms with HERTZ eqnation (1,1) ifthe indentation is purely
elastic. When the deformation becomes plastic, Ab will correspond to
Eq. (7.03), and to Eq. (7 .01) when theindentationattains aspecific depth
above 0.03 (totally plastically deformed). In all these events, and be-
tween isotropic materials, the eontact snrface is circnlar (elliptic be-
tween crossed rods ofdifferentdiameters); andifitslineardimensionsare
less than 1/20 the dimensions ofthe contact members, the constriction is
long and its resistance can be calculated according to Eq. (4.14).
Imagine curved members with perfectly smooth surfaces, say, two
eqnal cylinders with radins r pressed cross wise into contact by a Ioad
P prodncing purely elastic deformation. The contact area is a circle,
Ab. With a greater r, a greater Abis formed by the same P. Theoreti-
cally, with the cnrvature radins infinitely increasing, Ab would also
infinitely grow. Bnt, a practicallimit is defined by the always existing
waviness of the surfaces. This implies, that with nominally flat mem-
bers, ~ of Eq. (7.03) decreases with decreasing waviness ofthe snrfaces,
and, if the waviness ist not known, ~ is nncertain between wide Iimits,
say between 0.02 and 1; cf. § 7 B.
Nevertheless, bolted jnnctions, for instance between bus bars and
many other contacts between apparently flat members, show resistances
that scatter astonishingly little at given Ioad P when the surfaces are
clean. The faces of the members may have been fairly flat before being
bolted, bnt the Ioad usually deforms them. This has the effect of con-
centrating the contact spots to a rather small area. This means an
approach to rod, contacts of a relatively large rod diameter, implying a
fairly nneqnivocal relation between P and R. That is, a representing
cnrve can be drawn for any metal in Fig. (8.01), around which observed
points gather with scarcely any deviations towards higher R bnt with
considerable deviations below the cnrve in cases of very thick members
(resistant against deformation) with extremely small waviness. On
Fig. (8.01) the curves marked Cu-plates and Ni-plates illnstrate this
relation between R and P. A more thorough treatment of nominally
flat contacts is given in § 35 on thermal contacts.
Fig. (8.01) has been designed1 to meet the practical demand of an
easily read graphical illustration of the relation between P and R in
1 Fig. (8.01) essentially constitutes an extract from several similar diagrams
in R. HoLM [4].
§ 8. The relation between contact load and resistance 41
metallic contacts under conditions given in the introduction of thi;;
chapter. The figure refers to: A) contacts between crossed cylindrical
.rods (labeled rods); B) junctions between nominally flat members, fOl'
10 ' 1 t:----+-~·_,-+""-<,---l
lw~~--~~--r---~~·~dp
Flg. (8.01). Vlrgin contact resistances R plotted against the contact Ioad P. Solid llnes are for
clean contacts. In alr the metals soon become covered by a monolayer of oxygen. The increased
resistances are represented by the dashed lines, associated wlth solid lines by shading. Practical
contacts have thicker alien films. Their curves have a steeper slope and often lie higher above
the dashed lines as the latter above the solid lines. The measurements for the carbon haw hren
carried out in air, but also hold for vacuum, except for thcir dashed part.
example contacts between bolted bars with reetangular cross section

(labeled plates).
For any material, four curves are drawn, namely: 1. solid for clean
crossed rods; 2. dashed and connected with its associated curve No. 1 by
shading for the same rods when covered by a chemisorbed oxygen film:
3. solid for clean nominally flat members; 4. dashed forthe same nominally
fl.at contacts when covered with a chemisorbed oxygen layer, the associa·
tion with the corresponding solid curve again being indicated by shading.
Every observationwas made with a new contact. The dashed curves .
are entirely based on measurements on contacts which were closed after
being exposed to air for some minutes (up to one hour) after a thorough
cleaning. Under these circumstances, we can expect a chemisorbed
oxygen layer to cover the faces. This seems to be con:firmed because the
deviation, R1, from the associated solid curves can be correctly calcu-
lated from Eq. (8.02) assuming reasonable values for the tunnel re-
sistivity a of the film
(8.02)
The solid curves for nominally flat contacts (plates) as weil as all
curves for carbon are also based on measurements. However, the
solid curves for metallic rods have been calculated and checked in
vacuum by measurements in only a few points.
The basic data for the calculations are summarized in Table (8.09)
which is followed by an example. :Notice that Fig. (8.01) refers to cir-
cumstances that are idealized in certain respects. Particularly, the solid
curves concern perfectly clean and very smooth surfaces. In practice,
deviations are caused by surface contamination and roughness. General
rules for these changes can not be given, but it is important to know the
optimum that can be attained. This is presented in Fig. (8.01).
B. Crossed rod contacts. Preceding the discussion of the results, we
first indicate by means of Figs. (8.03) to (8.06) a suitable method for
measuring contact resistances which is the same method that was used,
in 1928, by R. HOLM [4]. Fig. (8.03) illustrates the wiring diagram,
while Figs. (8.04) and (8.05) show an appropriate construction of the
holders for the contacts. A circular contact area Ac = Ab is obtained
under conditions described in the introduction of this chapter. Fig.
(8.06) pictures the equipotential surfaces in one of the cylinders. In the
vicinity of the contact area, represented by a point in the figure, the
equipotential surfaces are nearly concentric ellipsoids of the type
shown in Fig. (4.16).
The surface A. is considered as end surface of the constriction. Since
it intersects the end of the cylinder, the voltage measured according to
diagram (8.03) will be the voltage between the end surfaces, and conse-
quently be what is called the contact voltage, U. With I being the
current, the ratio Ufl is the contact resistance. R, tobe measured with
sufficiently low U to avoid heating of the constriction. If the radins of
the cylinders is sufficiently large, say more than 20 times !arger than
that of the contact surface, (cf. Fig. [4.28]), the constriction is long
and its resistance can be calculated according to Eq. (4.15).
Results of the resistance measurement8 with crossed rod contacts. To
begin with, we consider a cylinder material of a moderate hardness, for
example copper. The diameter may be 2r = 5 mm a.nd we suppose the
conta.ct load to be a.bove 100 N. Then the pressure p will atta.in the plas-
Fig. (8.04). Balance used for resistance measuremcnts

Fig. (8.03). Wiring diagramm for on contacts between crossed rods. Each rod is kept in
measurement of contact resistance position by a holder H 1 or H 2 in such a manner that
between crossed rods bending is avoided. H 2 is tlxed, H 1 is connected t.o
the balance Iever
Fig. (8.05). Form of the holder H, and H, , used Fig. (8.06). Equipotential surfaces in the en-
for the device shown in Fig. (8.04) vironment of a contact between crossed rods
ticity Iimit in nearly the whole contact surface, making the impression
circula.r. The constriction will be fairly long a.nd Eq. (4.15) applicable
with good approximation.
We then consider a smaller Ioad, for example P = 0.1 N, making
it possible for merely elastic deformation to produce the contact sur-
face; note that this is true only for sufficiently smooth rod surfaces.
The contact area will again be a circle, and its radius a is given by
Eq. (1,1). Combining (1,1) with (4.15) and (7.01) with (4.15) we easily
find in the case of a small P and purely elastic deformation
R o:: p-'1• (8.07)
and in the case of a. la.rge P and purely plastic deformation

Roc ~ 1 /t (8.08)
Both equations are true, of course, only for clean metal surfaces and,
beyond this, in the case of (8.07) for sufficiently smooth surfaces.
Although surfaces of real bodies are never perfectly smooth but
affected with a certain asperity1 , the contact surface at }arge P will be
1 Amplitudes of the asperities0.1 to 0.01 mm, cf. ScHMALTZ [1], and p. 32
essentially coherent because the protuberances of one face are pressed

into the counter face 1 . The contact resistances, under these circum-
stances, are weil represented by Eqs. (4.15) or (4.21). As for the in-
fluence of persisting grooves within the contact surface on the re-
:;:istance, cf. Fig. (5.13).
\Ve now suppose the contact loadtobe sufficiently small to produce
with perfectly smooth members a contact area A by merely elastic
yielding, where A is the so-called HERTZ area corresponding to Eq.
(1,1) with r being the macroscopic curvature. However, with asperity
being superposed on A, it may happen that only discrete a-spots are
generated, mainly plastically as discussed in § 7 with Fig. (7.15a).
Then, the contact resistance is given by Fig. (5.13).
It is obvious that this case may be realized with microphone carbon
contacts at P < 0,2 N. Here the contact resistance is largely independ-
ent of the curvature of the contact members, i.e., independent of the
size of the HERTZ area, so that even the curve for nominally flat mem-
bers coincides with the curve for rods in Fig. (8.01). The explanation 2
for this fact is as follows. The contacts in question have a HERTZ area
with discrete contact spots. For a very small load, P, there may be
only three spots. With rising P the size of these spots increases only
slightly. At the same time, new small a-spots are added with the re-
sult that the average radius of an a-spot, thuR the resistance per spot,
remains essentially constant. Such a resistance phenomenon is fairly
independent of the curvature of the members as long as the a-spots
remain as discrete spots on the HERTZ area in the fashion ofFig. (7.15a).
In other words, the resistance is independent of the HERTZ area, and
as a consequence, independent of the curvature of the members.
C. Explanation of the dashed lines in Fig. (8.01). As already men-
tioncd in this paragraph, the deviation of the dashed lines from the
Table (8.09). Data for diagram (8.01)
r = rarlius of rod, E = YouNG's modulus, H = hardness measured with the specific
depth D ""' 0.04, e = resistivity, and a = tunnel resistivity
Material 2r E H (} a
cm 1010 Ntm• 10• X/rn' w-s n m 10-12 n m•
Al 6 2.4 2.8
Cu 0.5 11 5 1.76 1.5
Pt 0.58 13 6.5 13
Ni 0.3 20 14 9 2 to 8
Graphite 0.5 0.8 1.4 4300
baked carbon 0.3 to 0.6 0.5 3.5 7000 7
-~-·---
See, for instance, R. HoLM [28] Fig. 2, and I. MING FENG [1] particularly
1
Figs. (5) and (8).
2 R. HOLM [1] § 7 and R. HOLM [4] p. 243.
course of the associated solid lines is explained by the existence of a
film penetrable for tunneling electrons. Table (8.09) contains basic data
for Fig. (8.01).
We calculate a point on the copperrod curve obtained with very
smooth and clean surfaces. At P = 0.1 N we are in the region
where the yielding is elastic. Using the data of Table (8.09) and Eqs.
(1,2) as weil as (4.15), i.e., long constriction, we obtain
a = 1.11 V 0.1-2.5-10-3
11-toto = 1.46 ·10-sm
and
o-4 0
R = _!L = 1. 76 . to-s = 6 . 1
2 a 2.92. 10-s
R can be read on the solid line. The indentation is elastic since the
average pressure p = Pfn a 2 = 1.5 ·108.Njm2 < Hj3 is too small to pro-
duce a plastic indentation, cf. §I. The error resulting from assuming a
long constriction is < 1 % in the actual case of : = 1. 4~·~:0 3 = 170,
i.e-, Vß= 170 a. The additive resistance, L1 R (leading to the da!!hed
line), which we assume to be the resistance of a thin uniform film,
obtained from Eq. (8.02), is
L1R = ...!!._
:n:a2
= 2.24 · to-s 0
Thus, the total resistance corresponding toP= 0.1 N is found tobe
R = 0.6. 10-3 + 2.24 . to-s = 2.84 . 1Q-S n
differing very little from 2,9 · 10-3 as given by the dashed line.
The slope of the solid line in the region P = 0.1 N is - 1/3, accord-
ing to Eq. (8.07). Evidently, R is dependent on r in the case of elastic
deformation between relatively smooth surfaces according to Eq. (1,2),
but independent of r when the deformation is plastic.
Applying P = 50 N would bring about plastic deformation and R
has to be calculated according to Eq. (7.01). The slope then is -1/2,
according to Eq. (8.08). The portions of the graph with the slopes
-1/3 and -1/2 are connected by a slightly curved line.
The influence ofthe waviness has not been taken into account when
dra wing the solid lines, otherwise they would ha ve turned upwards a little
at their left end. But, the measured dashed lines show this tendency.
Notice that the dashed curves have a greater slope than the solid
ones. In the event of films thicker than those for Fig. (8.01), the slope
of the resistance lines approaches proportionality to P- 1 . The reason
is that the dominating film resistance varies as A~1 according to Eq.
(8.02), and Ac is nearly proportional to P.
D. Diversified resistance measurements. Diagrams with R plotted

against P (R-P-characteristics), measured under different conditions,
have been given by many investigators. We cite R. HoLM [4] § 12,
CoNTIUS [1], MÜLLER-RILLEBRAND [1], KAPPLER et al. [1], BowDEN
and TABOR [2], SnoBERT [1], CocKs (1], FuKUROI and MuTo (1], and
FAIRWEATHER (1]. FAIRWEATHER carefully avoided artificial aging.
Nevertheless, his results agree very weil with Fig. (8.01) which means
that aging has little inHuence on clean contacts ü Pis not too small, say
> 1N. According to KAPPLER et al., silver and gold, freshly scraped,
did not show any alien film resistance, and the results on crossed-rod
resistances coincide very closely with the solid copper rod curve in
Fig. (8.01); but after the members were exposed to air for some weeks,
the contacts became covered by a film with a = 5. 10-1s Q m 2. Also,
on platinum such a film appears minutes after scraping, even though it
is somewhat non-coherent. On nicke!, a film with a = 2 . 10-12 Q m2
was observed.
CocKs determined average values of a for various metals in air
after practical cleaning. He found, for example, with both members of
nicke! or wolfram, a of the order of 10-1° Q m2. With wolfram against
gold, a was of the order of 10-9 Q m 2. The gold very likely did less
darnage to the film.
MILLIAN and RIEDER [1], COMPTON and BAKER [1], cf. ANGUS [1]
published measurements on resistances in cross-rod contacts which had
been cleaned in different ways and were observed in air. The significant
conclusions can be so expressed: On freshly cleaned contacts a is of the
order of 10-12 Q m 2. Even after a month, and with P > 10N, one finds
a ""' 2 . 10-12 on silver and electropolished wolfram. But on copper, a
grows to more than 10-n after a month and 10- 9 after 3 months, all
n
with the unit m 2.
An interesting feature of the rod- and Hat contact curves for
carbon, as has already been described, is that they coincide for P <
0.02 N. This is explained above as a result of the asperity of the faces,
which shapes the contact area as a group of discrete a-spots, fairly alike,
whether the memhers are rod-shaped or Hat. Alien films exist, very
likely chemisorbed oxygen, but they interfere imperceptihly at P >
2 ·. 10-4 N, since their resistance is small compared with the consider-
able constriction resistance in carbon. Their role at smaller Ioads will
be discussed in § 9.
Short, distorted constrictions. Until now, we have calculated with
long constrictions and assumed the thickness of the rods or bars to be
great compared with the diameter of the individual contact spots, and
the spots to lie sufficiently apart to prevent distortion of the con-
strictions. An interesting exception is noted by the dotted line in the
lower right of Fig. (8.01). The measurements in question were marle
with crossed bars pictured in Fig. (8.10). The bars had been greased,
brushed by means of a steel
wire brush, and wiped clean
without completely removing
the last layer of grease, and
they were then immediately
clamped together. This proce-
dure produced fairly clean
contacts.
The bars were 15 mm broad
and 3 mm thick. It is obvious Fig. (8.10). Cross bar contact with negative R!b
that the constriction was not long and the tangential resistance in the
bars became comparablewith the constriction resistance. This can cause
the quantity ~b, defined in § 3, to assume negative values, andin fact
negative quantities for Rab were measured with P = 104 N. To under-
stand this, we consider the extreme case in which Pis !arge enough to
bring the whole covered surface into actual contact. Then the equipo-
tential surfaces will intersect the contact as indicated in Fig. (8.10).
We see that the potential tapped at point 2 will be that of an earlier
equipotential surface than is tapped at point 1. Thus, the voltmeter
measures a voltage with reversed polarity.
E. Use of Diagram (8.01) in practice. The diagram is very helpful
for determining: 1. the contact resistance corresponding to a certain
Ioad, or 2. the Ioad needed if the resistance is prescribed. The causes
for deviation from the curves in Fig. (8.01) are often found to be in a
faulty wiring or in films thicker than those met in the dashed curves.
It is of no great disadvantage that only some few metals are repre-
sented, since the curves for allmetals have a similar inclination. There-
fore, in most cases it is sufficient to calculate one point in the elastic and
one point in the plastic region, and then draw the characteristic p_a rallel
to the plotted ones. The formulas for the calculation are given in §§ 4,
5, 7, 8, and I.
For a rough computation of the resistance in a contact with 0.1
< P < 100 N, practically clean members with resistivities [?! and g2 and
the contact hardness,H, ofthe softer member, one may use the follow-
ing rule of thumb:
R e1 + e2 vH
=
2 p
(8.11)
F. Practically clean contacts, preloaded with a high P 1• Comparison

of the resistance R at the light Ioads P in Fig. (8.12) with that of the
1 R. HOLM et al. [8] p. 61. Fig. 14a.

nickel curves in Fig. (8.01) at the same Ioads shows that the films on
the practically clean nickel rods were relatively thick, but still tunnel
conducting. Submitting these contacts to great Ioads reduces R quicker
than according to (8.01). Va-
rying P between high and low
1
(without vibrations) results in
a constantl R for a pressure
\;
,.__ interval between 1.9and0.25N.
0.0.
This means that metallic spots 2
have been produced (most pro-
bably on the elevations) in
which the members were weld-
ed with a strength that could
~~
(tJ)
-
(10)
withstand the elastic counter-
1
r,~,
force during unloading. The
"'
""
__"., average pressure attained a-
{10/ bout half the hardness of the
(1{){) l10.
ZO}
(150)-'0.
nicke!, which makes it very
0.00~
likely that complete plastic de-
az as 1 N 2 formation was produced in the
Conlocf Ioad contact at the top of elevations,
Fig.(8.12).Cycling of the Ioad P (50 to !50 times) on disrupting the continuity of
a crossed nicke! rod contact; freshly cleaned rods of
0.3 cm diameter; exposed to air two days the film in some spots.
§ 9. Contaet resistanee on freshly eleaned rods in air at very

small contaet Ioads
A. Observations on gold and silver. Instructive measurements were

made in F. L. 3 with the aid of a modified moving-coil instrument. The
pointer, provided with a very good bearing, carried one contact mem-
ber, the other being stationary, see Fig. (9.01). The contact Ioad, or
sometimes tension was generated by means of the current through the
coil. Fig. (9.02) shows observations on crossed gold and silver rods. The
conductance 1/R is plotted against the Ioad P. The freshly cleaned rods
were washed with soap and water and finally with alcohol. Sometimes
1 R. HoLM [8] p. 61. Aluminum contacts behave differently due to the rapid
oxidation of clean surfaces. A metallic spot which was welded oxidizes very quickly
when the contact is broken and the spot exposed to air. When the contact is remade
this spot no Ionger exhibits adherence.
2 HoLM [26] pp. 334 and 335.
3 R. HoLM [4] § 13 and R. HoLM [30] Fig. (16.02) for which the measurements
were also made in F. L.

§ 9. Contact resistance on freshly cleaned rods in air 49
they were also scraped with a clean tool, but there always was suf-
ficient time to allowadeposit ofthe normal (chemisorbed) oxygen layers
on silver. Without the film (perfectly clean) contacts would yield a
conductance at P = 10-4 N several times }arger than was observed.
An intere&ting feature of the ailver curves in
Fig. (9.02) are the sudden changes of the conduc-
tance to higher values. We contribute these jumps
to the intricate nature of the formation of a metallic
spot as follows. We assume one rod to have contact
with a single protuberance. First, the contact is
formed merely elastically and the conductance
through the film is due to the tunnel effect. With
Fig. (9.01). Device for
increasing Ioad the Ioad carrying hump breaks measuring contact re-
down, disrupting the film and producing a clean sistances at very small
Ioads P
a-spot. This is evidenced because after such a yield
the contact shows a cold weld during unloading: cf. Fig. (9.02). Gold
contacts sbow less jumps; but whenever a jump occurs the contact ex-
hibits a cold weld. The metallic spots producedare, ofcourse, verysmall.
N-
Conf<rcf Ioad--
Fig. (9.02). Conductance 1/R of gold and silver contacts plotted against the Ioad fing = 0.01 N.
Hardness H lsilver) - 7 . to• NJm•. The reversible branches which have been run several times,
are marked wlth r
At further increasing Ioad, there may be new rupturing of the film

on the same hump due to new plastic yielding which would result in
enlargement ofthe already existing metallic spot. Or, other humps may
collapse too, and new spots may be generated. Very likely both phe-
4 Electric Contscts, 4th Ed.
nomena \\'ill happen during one jump. It may even be that the effect of
sudden enlargements of a-spots is involved to a greater extent than
the generation of new spots with the consequence that the number of
jumps does not equal the number of a-spots. The suddenness of the
jumps indicates that the yielding protuberances were overloaded, i.e.,
loaded with a higher pressure than the hardness H. F AIRWEATHER [ 1]
has observed similar jumps.
The reversible branches (symbol r) evidently are due to elastic varia-
tion ofthat portion of the contact area which is quasimetallic, whereas
the produced metallic spots may remain constant due to adherence.
The adherence or cold welding of contacts is often troublesome in
service. In Fig. (9.02), a minor adherence is indicated as an example
in the curve Au2 • Mter having attained the Ioad 2.2 · 10-4 N, P was
decreased, but the resistance did not increase. The contact area kept
adhering without altering its size until a tension of 8 · 10-6 N sepa-
rated the contact members. The platinum contacts did not adhere to
any ]arge extent and did not offer as good and uniform a conduction
at small Ioads as gold contacts. Nickel contacts at small Ioads could
compete with platinum contacts were it not for their magnetic proper-
ties which produce a particular kind of adherence; see p. 59.
The measured adherence indicates that clean meta] surfaces stick
to each other as if welded together, cf. § 28. lf the contact area remains
unchanged on removal of the Ioad, one can expect the adherence force
per m 2 to be equal to the tensile strength Z ""' 0,3 H of the metal in
question. Then at contact break the tension would reach 0.3 of the
initiaJ pressure. But even with perfectly clean surfaces, one measures
less because elastic forces help to Iift the contact; cf. § 28.
The kind of adherence described above must not be confused with
another kind, very common in practice, produced by liquid contami-
nations on the contact surface. This kind is prevented by cleaning.
B. Observations on carbon contacts. The curves for carbon contacts,
in Fig. (9.03), show the influence of alien filrns because they do not begin
to ri~::~e at the zero Ioad point. The jumps are strongly marked. We ex-
plain them as caused by a sudden generation of clean a-spots by dis-
rupting of the surface films when the underlying carbon yields plastic-
ally on suitably shaped Ioad bearing humps, or the enlargement of exist-
ing a-spots or both. After any sudden change, a gradually rising branch
appears. It is re-versible (symbol r) and consequently belongingto merely
elastic deformations with negligible adherence. This is the reason why
the reversibles have a smaller slope than the curves, on an average.
Fig. (9.04) reproduces the reversibles with other co-ordinates and
compare!l them with curves from Fig. 21 in GoucHER [1]. These. were
also obtained with iterated Ioad variations. .
§ 9. Contaot resistanoe on freshly oleaned rods in air 51
Attempts by GoucHER1 to explain the effect in microphone contacts
by assuming hemispherical humps on the grain surfaces do not lead
to increased understanding. The shape of the humps is obviously other
than hemispherical.
As mentioned above, IV-
:f)~ 1 1 t()-+
we expect the resistance
H'" -=-
mho
of the chemisorbed film
on carbon to be mea- ,.,/
V
V f.-.1
aurable at P < 2 -10- 4 N.
In fact, observations of ,Y ~ ';;-"
WRIGHT and MARSHALL
/ j/ Mk~tl orbi;n
[ 1] show this effect. They

investigated the contact f·f()-J /r
between two carbon fila- ... l-> ~ ~~
ments from incandescant
/
---~~
~.A---
lamps, both after degas-
sing at 1700 ac in vacuum _[_ ;:. _......
&
J
r-
,.... _t:......
..."
and after exposing them r0;r f
to air. The measurements

give, atP = 10- 4 N:ef2a tmg
67' r f·ffl
- ;:'
g z f(J-z
= 80Q and afna2 = (125 Conluct louo'-
- 80)Q = 45Q. Substitut- Flg. (9.03). Conductance 1/R of contacts of baked car·
ing the resistivity (! = 3.5 bon, 2 plotted against the ·Ioad P. Hardness H = 3. 7 · 1o•
N /m • The reversible br nches which have been run through
. 10-6 Qm of the carbon several times, are marked with r. They have a smaller slope
than the other solid lines, which are virgin curves for ln-
used we calculate a = 6.7 creasing P. Irreversible curves for decreasing P are dashed
·10-I2i~m 2 . Withincreas-
ing P, the influence ofthe film resistance dimishes. One might expect this
resistanceto be proportional to 1/ P; buta greaterrate was observed. The
reason for this discrepancy might have been that the films were dam-
aged at the greater loads, making clean carbon contact spots possible.
The film on the carbon in-
f~()r---~r------.-----,-----,
creases the slope ofthe r-branches SJ
in Fig. (9.04) to some degree; this
might ha ve a slight influence on
the sensitivity of microphones,
but by no means is decisive for
the microphone effect, as was once
believed.
At elevated temperatures (500
to 600 °0) an oxide, say a tarnish, 100 11,------zl:-----!:-
f ---!-:10,....".0y_n_e_3-JO.
can be generated on the carbon
Flg. (9.04). Reversible branches (the dashed
1 GoUCHER [ 1), of. CmuSTENSEN and ones belong to GouCHER) of R P·charl\cterlstics
PEARSON [1]. of microphone carbon contacts
4•
according to investigations by MACKLEN (1], [2], [3]. His results will

be discussed in § 32.
C. Method of wire probes for exploring contact surfaces. Fig. (9.05)
shows the circuit with the exploring loop; cf.
1~ ;2?~cp
Fig. (8.03). A 5 fl wire of platinum or gold
forms the loop. The Ioad is indicated by the
compression of the loop and can be calibrated
Sumpfe
in mg.
Fig. (9.05). Wlre loop The method was used, in 1954, by SAVAGE
probe method for testlng and FLOM [3] and SHOBERT [2] . It was im-
contact points at P of
the order of 1o-• N proved by CHAIKIN et al. [1]; cf. § 30 G.
§ 10. The inductance of a current constriction. Skin effect
A. Inductance. The constriction of the current through the contact

not only generates an increase of the resistance but also produces an
additional inductance. In order to compute its order of magnitude, we
investigate the simplified current fl.ow as illustrated in Fig. (10.02) and
follow the deduction given by BusoH1 . We compare Fig. (10.01), show-
ing lines of current fl.ow in a cylindrical conductor, with Fig. (10.02)
Fig. (10.01). Parallel current llow Fig. (10.02). Model of F lg. (1.02) lllustratlng
the current flow lnto a slngle central a-spot
between two cyllnders
where two cylinders are held against each other as in Fig. (3.07), The
base, Aa, is the apparent contact surface with a relatively small con-
ducting area, A. , in its center. As in Fig. (1.02), the conducting con-
tact surface is replaced by an artifact consisting of a perfectly conduct-
ing sphere with radius b. The larger sphere in the figure has the same
radius, B, as the cylinders. We imagine a radial current to fl.ow out
through this sphere from b to B. In the surrounding body the lines of
fl.ow are assumed to be parallel to the axis of the cylinders as in Fig.
(10.01).
Let L1 be the inductance in the large sphere B according to the cur-
rent fl.ow of Fig. (10.01}, and let L 2 be the inductance of the correspond-
1 According to a communication by HANS BusCH to the author, not otherwise
published than in the author's books about contacts.

§ 10. The inductance of a current constriction. Skin effect 53
ing sphere Bin the case of Fig. (10.02); then the increment of induct-
ance is L= L 2 - L1 . We may confine our calculations to the magnetic
fields generated by the flowlines within the spheres B, since the other
fields are identical in both cases. If W be the magnetic field energy
and I the current, we have with easily understandable indices1•
(10.03)
Using mks units we proceed as follows :

Oalculation of 2 W2 /I 2 • The current density through the equipoten-
tial sphere with r = 1 is J = I j2n. The current within the cone with
the vertexangle {} (between the axis ofthe cylinders and the generatrix
of the cone) is
lt
1 (v) = f2nJsinvdv = I(1- cos{})
0
Thus the magnetic field intensity, H, in the mantle surface ofthat cone
at the distance r from the center is given by
2nrsin{}H({},r) =I({})
Hence
I {}
H =-tan-
2nr 2
(10.04)
Integrating through the hemispheres of both electrodes gives the

magnetic field energy
Wq• = {l~JH
2
2 dv
where tto =- 1.257 . w-s Hyjm anrl tt is the relative permeability. Hence,
after inserting H from (10.04)
We put {}j2 = cp and find
-2
B
w2 =4~~ojd r J n/4
1- COS2 <p d( cosm )
l
J2
--
n b 0 COB <p T J (10.05)
= 1.545tt10-7 (B- b) Hy
with Bandbin meters.
Oalculation of 2 W1 JI 2• Here it is convenient to imagine the B sphere
dissected into cylindrical sheets with radius x, thickness d x, and length
1 Cf. the weil known equation W = 1/2 (Ll2),

2 ~,B2- x 2 • The volume ofsuchanelementisdv = 2:n:xdx2. B 2 - x 2•

The magnetic field intensity in it is
I X
H= --
2n B 2
Thus we obtain
B
2 Wl
f"
= _! l'-f-'ojH2dv =
f2 2
1-'n 1-'of(!!-.)a
B
d 1/1 - (!!-.)2
VX
B
0
Putting z = 1 - x 2 f B 2 we easily evaluate the integral and find with
length in m
1
2 ~1 = 5.33 ·10-8 fl B Hy (10.06)
and according to (10.03), (10.05) and (10.06)
L = 21~2 - 21~1 = [1.545,u(B- b)- 0.533,u B] 10-7 Hy (10.07)

Since b is small compared with B, it may be neglected. The induct-
ance, L, is similar to the known expression for that part, L 3 , of the
inductance of a straight wire, length l, which is caused by the magnetic
field within the wire. L 3 being equal to 0.5 ,ul · 10-7 Hy, we may inter-
pret Eq. (10.07) in the following way.
The increment of inductance resulting from the constriction of the
Iines of flow through a circular contact area approximately equals the
inner inductance of a wire of arbitrary thickness, with length equal to
the mean diameter of the apparent surface. Since this length will be a
very small part of the length of the whole circuit, the inductance of the
current constriction may generally be neglected.
B. The skin effect arises from an induction by its own magnetic field
on an alternating current. The lines of current flow are pressed towards
the outside of the conductor, thus dimishing the effective conducting
cross section andincreasing the resistance. In a cylindrical non-magnetic
wire (radius B m), the resistance increases by a factor of about
fB =IX+ ! where IX= 1~3 V; (10.08)
Here v is the frequency of the A-C and (! Q m the resistivity of the wire.
W e are concerned with the skin effect in a current constriction. For
computing its order of magnitude one may proceed as follows. Refer-
ring to Fig. (10.02) we consider two non-magnetic cylinders (radius
B m) connected by the infinitely conducting sphere with the radius
b m according to (1.02).
We imagine the volume between the spheres band B, of one contact
member, as being subdivided into concentric shells, each ofwhich (with
§ 11. Electrodynamic repulsion in a symmetric contact 55
the radius r and the thickness d r) has the d - c resistance:: ~. In every

shell, the skin effect moves the current outward from the axis towards
the apparent contact surface, increasing the resistance by a factor simi-
lar to fB where B must be replaced by n/2 (r). Thus the a- c resistance
of one shell is about
::;2 [; 1~3.~~ + ! ]
Integrating through all shells we obtain an a- c resistance of
2~[2.~oa ~In:+!(!- i)]

where 1/B may be neglected in comparison with 1fb.
Dividing by the d- c resistance R = ef2:n; b we find that the a- c
resistance of the constriction is larger than the d - c resistance by the
factor
VeIn b + 4
:nb lfv B 1
2 ·101 (10.09)
Formulas (10.08) and (10.09) are valid1 only for
!
1 3 Vf> 1 and
:os
In the case of 2 ~ ~ ~ 1 the skin effect in the constriction is
negligible.
The skin effect is counteracted by the repulsion which is treated in
§ 11, and is expressed by Eq. (11.04). This force is independent of the
frequency of the current, and even though it can compete with the skin
effect at small frequencies, it will be negligible at high frequencies.
§ 11. Electrodynamic repulsion in a symmetric contact

of non-magnetic material
We simplify the problern by again using the artifact of Fig. (10.02).
Electrodynamic forces are active where the lines of current fl.ow have
components parallel to the surface Aa. These mechanical forces are
generated by the interaction of the electric current with i~s own mag-
netic field. Wehave already studied this field in §10. The magnetic lines
offorce are concentric circles around the cylinder axis, and for the magne-
tic field intensity on a circle r, [cf. Fig. (11.02)], we have with mks units
H = _!__ (1- cos~) (11.01)
2:nr sin~
1 Cf. deduction of the skin effect, per example JAHNKE-EMDE [1] p. 142-146.
W e imagine a conical ring volume with the cross section dimensions

rdß and dr along the circle (r, ß). The current flow through the ring
may be subdivided into a series of current elements dl dr, with the
intensity dl and the length dr . A mechanical
force I, perpendicular to r and H, acts on each
such element, and its component in the direc-
tion of the cylinder axis is 11 = floH di dr sinß
newtons. Thus the whole ring volume, carry-
ing a differential current I sint? dt?, is subjected
to a force p0 H I sin 2 t? dr dt? in the same direc-
tion. Integrating through the whole constric-
tion of one contact member from b to B, we
find after eliminating H by Eq. (11.01)
J:r J
contoct plane B n/2
Fig. (11.02). The shaded
element has co-ordinates r,
Ir= P~~2 (1- cos#)sin#d#
e and dlmensions dr, r dll b 0
in the plane of the paper. where Ir is the total repulsive force between
A small extension perpen-
dicular to it may be lmag- the electrodes. The integration can easily be
ined; f = mechanical
force generated by the in-
carried out, and we find with Ir expressed in
teractlon of the magnetic newtons
fteld and the current ele-
B
ment dl dr; / 1 = compo-
nent of that force parallel
fr = 10-7 I2ln -b N (11.03)
to the cylinder axis
Eq. (11.03) is approximate. SNOWDON [2]
shows that within the limits of the approximation , one may put for
b, the radius a of the flat contact area; hence
fr = 10-7 12 In B
a
N (11.04)
and that Irisindependent of the face curvature of the members. Be-

cause of the repulsion , a is decreased thus influencing I,. This must be
calculated by successive approximation. Heating and softening in the
contact also has to be considered; cf. § 30. SNOWDON presents nomo-
grams, based on experience, for calculation of the Ioad P that is re-
quired to keep contacts from blowing up. This is valuable, as the
determination of b for Eq. (11.03) or a for (11.04) is hardly feasible in
practice.
In order to elucidate the importance of the force I, for practical
contact measurements, we first compute it for a contact between crossed
metallic rods, simplifying by the application of the artifact according
to Fig. (1.02). The whole constriction resistance is efnb according to
Eq.(4.09) ·and consequently I= U nbfe. We want b tobe expressed in
terms of the contact Ioad P and the contact hardness 1 H. From Eq.
1 See§I.
§ 12. The capacitance of a contact. Electrostatic attraction in a contact 57
(7.01) we have H na 2 = p. The relation between a and b is obtained

from Eq. (1.06) and we find
Substitution in Eq. (11.03) gives

4 _7 U 2 P B
fr ~ -10
:n;
H
'(!
2 In -b N :11.05)
We represent the metal by (! = 2. 10-8 Q m and H = 7. 108 Nfm 2• The

quantity In Bfb has the order of magnitude of 5 to 10. Substituting
these expressions in Eq. (11.05) we find with U in volts and P in newtons
(11.06)
showing that for a single contact spot, 'Iris comparable to P if 3.5 U 2

is of the order of 1. However, for a contact with several a-spots, this
equation is no Ionger true. The force Ir then diminishes, since the con-
tribution of each a-spot decreases at the same rate as the square of
the current through it, according to Eq. (11.04)
§ 12. The capacitance of a contact. Electrostatic attraction

in a contact
The capacitance, 0, between two essentially parallel equipotential

surfaces both with the area A m 2 and distant from each other by s m,
which may vary from point to point, is, with the relative dielectric
permittivity = 1,
(12.01)
where
e0 = 8.85 · 10-12 F/m
The electrostatic attraction in such a capacitor being charged to
V volts is
Vfd0
I a = 22 8 = e20 v2 8 2 N ftlA
(12.02)
Due to the curvature of the equipotential surfaces of a constriction,

the average distance between the charges below and above a metallic
a-spot is much greater than the distance between opposite points of the
members in an influential surrounding of the spot. Therefore, ,in the
integrals of Eqs. (12.01) and (12.02), the contact area may be neglected
for a clean metallic contact. It is otherwise if the contact carries a film
with a high resistance, in which case the cotact constitutes a capacitor

with the film as dielectric. Weshall now apply Eqs. (12.01) and (12.02)
to some typical examples.
Example A. Crossed rod metallic contact.
Radius of rods = r m, radius of contact sur-
face = a m; contact voltage = UV. Fig. (12.03)
a
gives a qualitative picturt~ of a section of the
contact. The distance between opposite points
of the members is 8, of which the voltage dif-
Fig. (12.03). Vertlcal sec- ference V (x) varies with x, the distance from
tion through a crossed rod the axis of the contact. On the basis of Eq.
contact (radlus a) and the
surroundlng interspace (4.08)· we find
with the variable thick-
x-a
ness 8 V(x)=U-
x
(12.04)
where U = total contact voltage. Hence, the capaeity between the con-
tact members is
r
O= -e0 f2:n:xdx Ux- a
U 8 X
a
x2 _ a,2
or, after inserting 8 = --
r
0 = 2ne0 rjr.l:!!_ = 2ne0 rln 2r

x +a a
l (12.05)
=5.5·10- rln 2ra
a 11
F J
I
The electrostatic attraction in the same contact is
r
f = eo U2r2j 2:n:xdx (x- a)2
1I
a 2 \X~ - a2)2 x2
a
=neo U2r 2 1 ::)2x ~neo

(x U2 (: nln 2 _ 0.5
(12:06)
= 5.35 . 10-12 U2 (: r N
Applications of Eqs. (12.05) and (12.06) to the example after Table

(8.09) with (rja) = 170 and say U = 0.1 V gives
0 = 0.61 · 10-12 F
Ia = 1.54·10-9 N
showing that this attraction was quite negligible in the experiments
with crossed rods, and that the capacity is extremely small.
§ 120 The capacitance of a contacto Electrostatic attraction in a contact 59
Example B. In nominally flat contacts, larger, non-conducting areas
lie around the a-spots with such small distances between the members
(some few atomic diameters) that a noticeable attraction will occuro
In the contact of a graphite brush on a copper ring, for instance, we
must consider almost the whole ground area, say 10- 4 m2, as being
markedly attracted, while the conducting contact area is much smallero
We assume, around any a-spot, an ~rea of a size equal to the "Conduct-
ing spot, where the average thickness of the interspace between the
members ist 5o 10-9 mo The rest of the interspace might have the
average thickness 8 >5o 10-7 mo At a load of P N the total area hav-
ing the small distance 8, will be of the order of 2 P 10-8 m2o The area
with the larger distance will be about 10- 4 m 2 0 Inserting in Eqo (12o02)
gives
fa = 1.8 10- 3 U 2 (2 P - 1)""' 306 10-3 U 2 P
o o (12007)
Usually U is of the order of 1 V and P of the order of 5 No Thus,
according to Eqo (12o07), the electrostatic attraction in the brush con-
tact contributes to the Ioad with less than 1 % 0 The corresponding
capacity is Oo02 fl.Fo At about 105 cycles per second, its impedance
approaches the value of the contact resistanceo WILMAN [1] investi-
gated the contact resistance, R, between flat nicke} members, initially
magnetized to produce an attraction force of P and another time de-
magnetized and the same force applied mechanicallyo He found that in
the :first event, R was always smaller by up to 50% than in the second
evento
Examp1eC.John8en-Rahbek effecto The so-called JoHNSEN-RAHBECK
effect is the electrostatic attraction between a semiconductor, for exam-
ple agate, and an adjacent metal electrode, when a current passes through
the contacto The original experiments were made with a metal foilo
The contact resistances are great, and, due to the great value of
(?A of the semiconductor, the contact endures voltages of the order
of 100 V without reaching very high temperatureso The average
distance between the positive charge on one side of a contact spot and
the negative charge opposite to it in the other member is known tobe
equal to the thickness, 8, ofthe barrier Iayer, see § Ilo In the surround-
ing of the spots, the average distance also includes the interspace be-
tween the memberso Application of Eqo (12002) with A = 10- 4 m 2 ,
V= 100 V and 8 =10- 6 m yields Ia = 4.4N = 0,45 kg per cm 2 apparent
contact surface, which is a practicalmaximumattractiono STUCKES1 was
able to attain 10 Nfcm 2 with 100 V between a magnesium titanate disc
and a metal piece by polishing the disc to a flatness within 0,25 o to-s mo
1 AuDREY STUCKES [l]o She cites important earlier literatureo

Recalling that metals do not endure contact voltages above 0,5 to

1 V (see § 20) one may wonder why semiconductors stand 100 V. The
reason isthat gÄ is solarge that Eq. (15.16) defines relatively low tem-
peratures for a given contact voltage. For instance, el for black silicon
carbidel is about 0,4 V2 rK
and for magnesium titanate 2 greater than
104 units, both measurements at room temperature (decreasing with
rising temperature). It is instructive to compare these values with
eÄ "" 10-5 units for metals.
Because the voltage is squared in Eq. (12.02), workers have ex-
pected an adherence proportional to the square of the applied voltage.
This would be true only if the geometry of the contact and also the
distance 8 were independent of V. This is not the case. STUCKES [1]
found that the force-voltage relationship is a steadily varying function,
the force ultimately becoming constant at high V.
With 100 V at 8 = 10-6 m the average field strength is 108 Vfm
and in many points it will reach the order of 109 Vfm. Under such
circumstances electrons tunnel through the boundary hill, where the
semiconductor has contact with the metal (see § 26, L) and a decrease
of the contact resistance results. STUCKES has proved the existence of
this effect. She also made tests concerning applicability of the JoHN-
SEN-RAHBECK effect in a clutch. She found that the adherence force
decreases with the nurober of operations, evidently because wear dam-
ages the polish of the contact surfaces. Even a change in surface flatness
that seems to be small has a great influence on the adherence. The
applicability of the effect is therefore limited. It is historically inter-
esting that EDISON 3 , in 1877, and FESSENDEN, in 1909, used similar
effects for inventions.
§ 13. The relationship between electric potential and

temperaturein a current eonstriction which is symm.~tric with
respect to the contact surface; that is, the p.a.-relation
A. lntroduetion. The thermal and electric currents o bey similar
laws. Heat energy flows by virtue of temperature differences that cor-
respond to voltages ( = potential differences) in the electric case.
With symmetric contacts the generated heat flows in the same path
as the electric current. Therefore, a simple relation exists between volt-
ages and temperatur differences. This relation is often called qJ{}-
relation and is the subject of this Section.
1 ELSE HoLM [6].

2 AUDREY STUCKES (1].
3 English patent 2909 of July 30, 1877 with addition of 1878.
§ 13. The relationship between electric potential and temperature. .. 61
In non-symmetric contacts, usually only small deviatiom; appear

from the common path of the electric and thermal currents. For the
general proof of the qJt?-relation we will assume perfect common paths
to simplify the proof.
The qJt?-relation was discovered in 1900 by F. KOHLRAUSCH (1]
and served as the base for his measurements on heat conductivity.
DIESSELHORST (1], also GREENWOO,D and WILLIAMSON [2] pi.Iblished
strict proofs under defined conditions. They assume a conductor with
current inlet and outlet areas which areisothermal and equipotential,
the rest of the boundary being insulated electrically and thermally. A
common path of the currents appears as a consequence. DAVIDSON [1]
made a proof fairly similar to the following but with the common path
presupposed. A particular proofthat the electric and thermal currents
use the same paths in contact constrictions is presented below, so as
R . HoLM did earlier1 • DIESSELHORST [ 1] presented a special proof for
the single-validity ofthe solution as it is made for the LAPLACE equation
in the theory of potentials.
In the application to the theory of contacts, the contact area is
treated as the inlet. The outlet is assumed to be sufficiently distant so
that the thermal and electric flow in the vicinity of the contact are
negligibly influenced by features of the flow at the outlet. The condi-
tions of contacts motivate the special deduction presented below.
B. Proof of the cp&-relation in a symmetric constriction. Our assump-
tion of a symmetric contact im plies: 1. Both members are of the same
material which is isotropic in thesensethat the electrical resistivity, (!,
and the thermal conductivity, A., depend on the temperature (T °K)
alone; 2. Both members have equal
/.,..,---- ........,
systems of niveau-surfaces; the
contact surface, A 0 , is equipoten- I
, '\
tial and isothermal with the high- a !JI a \
I I
est temperature, T 8 , ofthesystem; \ I
' ___
\ I
3. In the intermediate space no ................
..,..",/
heat flows from one member to the
other because opposite points of the Fig. (13.01). Symmetry around contact
spots
members have the same tempera-
ture. See Fig. (13.01) where the thick lines, a, represent contact areas
and the dashed curve an isotherm.
The deduction is independent of the shape of A 0 which may be
coherent or consist of several contact spots.
Fig. (13.02) gives a qualitative picture referring to one contact
member, which for simplicity is considered as having a single contact
1 See R. HoLM [29], [30], and [37].

62 Sta.tionary Conta.cts
spot, A 0 • It is assumed that at a great distance from A 0 the tempera-

ture is T 0 (the bulk temperature) in both members. Differences T- T0
~" {} are called supertemperatures (above T0 ). The maximum supertem-
perature, namely that of A 0 is called
8. Thus, A 0 is a section acrosf' which
no heat ßows. The deduction gains
simplicity by stipulating rp = 0 in this
K section.
Consider an electric niveau-surface
A and a shell between A and a con-
secutive niveau-surface. Let potential
A,. and temperature of A be rp and T,
.......-.;;r--.ct-- - A, drjJ 8dl7>
L-~___J~- A"oe r; with rp + dqJ and T + d T belonging
Fig. (13.02). Symbols
to the consecutive niveau-surface.
Furthermore, Iet d R and d W be the
electrical and thermal resistances of the shell against the flow. The
heat I rp is developed between A 0 and A. Since the electrical and
thermal currents have the same paths, this energy is moved through
the shell by the temperature difference d T.
Hence, considering that according to Eq. (35.01) dW = dRfe).,
- dT = lqJdW = lqJdRfe}. (13.03)
Since
dR = drp/1 (13.04)
- f!Ad T = f!Jdf!J (13.05)
and integrated
To 6
Je).dT= Je).d{}= ~<t (13.06)

T fJ
Notice: A condition for Eqs. (13.03) and (13.06) isthat no heat is accu-
mulating anywhere in the constriction, and the process has reached the
equilibrium state.
Eq. (13.06) is the general expression for the qJO-relation. Applied
to a clean contact with the total voltage U between the end surfaces
of a long constriction, thus the end surfaces at the potential ± U/2,
we have according to Eq. (13.06)
(13.07)
Often a suitable mean value (!A of (!A can be computed and then the
simple form lJ2
eJ8=- s (13.08)
can be used.
§ 13. The relationship between electric potential and temperature. . . 63
C. Justification of the assumption that the electrical and thermal
currents fl.ow in the same patlws, implying that any equipotential sur-
face also is isothermal: Consider the shell on the equipotential surface
A and try the assumption that it is isothermaL Let flow tubes be chosen
so as to cut out, from the shell, elements with the same electrical re-
sistance, dR, thus carrying the same current. These elements then,
also have equal thermal resistances, dRfgA.. Consequently the same
heat is produced in each. Begin with a shell between A 0 and a conse-
cutive surface A 1 . It is evident that A 1 is isothermaL Now applying
the deduction to the shell between A 1 and a consecutive A 2 one con-
cludes that A 2 is isothermaL Repeating the procedure one finds that
all equipotentials are isothermal, which means that the assumption is
reasonable. There is no reason to expect another solution.
Remark. Consider the condition when the heat flow has wider crose
sections than the electric current. When heat flows over the mantle
surface, for example, then, only a portion of the heat q;I reaches A;
l
consequently Eqs. (13.03) and (13.07) change to
- gJ..dT < q;dq;

and
To+e (13.09)
jeJ..a T <
To
~2 f
expressing that Eq. (13.07) gives the highest supertemperature pos-
sible (namely fJ) that can be connected with the voltage U. This is
true not only for contacts but for any form of a conductor that is
heated by the current produced by the voltage U between two arbitrary
isotherms with the temperature T 0 •
The q;O-relation when heat enters the constriction through A 0 is
treated in § 14,D.
D.The p,'}-relation in metallic constrictions. Here the law of WIEDE-
M.ANN-FRANZ, treated in § III, C can be employed as expressed in Eqs.
(13.10) and (13.11):
(!Ä = LT (13.10)
where L is regarded as a constant, independent of both the metal and
the temperature. Actually, L is not quite the same for allmetals used
in contacts; cf. Table (X,2). But, a fairly good approximation is ob-
tained with
L = 2.4 ·10-s (V;oK) 2 (13.11)
Substituting (13.10) in (13.06) we obtain
J
e T8
2jeJ..a1J = 2 LTd T = L(T~ _TZ)= q;2 (13.12)

{J T
For a long constriction with the total voltage U
L(T~- T~) = ~2 (13.13)

hence with T~ > T~
T9 """ 3200 U (13.14)
Applications. Let T 0 be room temperature (295 °K), Te= soften-
ing, melting or boiling temperature of a metal, and call u., U1 and Ubz
the corresponding voltages defined by Eq. (13.07). We name these
voltages softening, melting and boiling voltages respectively. They can
easily be observed by critical changes of the contact resistance. This
was done for most of the values in Table (X,2). Some of the voltages
in this table have been calculated under the assumption ofthe validity
of the WIEDEMANN-FRANZ law; namely by inserting Te (taken from
tables) and T 0 """ 300 °K in Eq. (13.13). The situation is less clear
concerning ub! because the validity of the law is dubious beyond the
melting point.
Several workers have reported deviations from Eq. (13.13).
F. LLEWELLYN-JONES and R. H. JoNES [3] recorded on an oscillograph
a limiting voltage less than the calculated Ubz in Pt, Pd, Cu and'''Sn
contacts and slightly higher in Fe, W and Mo contacts. GREENWOOD
and WILLIAMSON [2] observed that a gold contact collapsed at the
expected 0.43 V, but soon after the voltage dropped to 0.38 V. The
authors tried to explain this step as a consequence ofEq. (16.31) which
defines a stable maximum below the melting point under certain con-
ditions; cf.§ 16. BowDEN and WILLIAMSON [11] observed that gold
contacts (annealed and then subjected to a number of contact cycles)
adapted themselves to about 0.13 V under current pulses (1.2 ms) up
to 160 A maximum current. It is remarkablc that this voltage is higher
than the softening voltage U8 """ 0.08 V usually measured under statio-
nary conditions.
§ 14. The 91~-relation .in cases of dissymmetry

Six cases of dissymmetry, A, B, C, D, E and F shall be treated.
Case A. Deviations from symmetry are limited to regions where
the distance, r, from the contact is great compared with the dimension
a of the contact surface Ac.
We know from the theory of constrictions that the strength of the
voltage- and temperature-fields is concentrated to the neighborhood
of the contact area Ac. We expect that there the g>B--relation essentially
governs the situation, and the problern is to compute to what degree
this is done in the surrounding of Ac.
§ 14. The tp{}- relation in cases of dissymmetry 65
Since any electric or thermal fiux tube begins in a charge and ends
in a charge that certainly is not larger (including an electric charge of
the opposite sign), the total charge (or quantity of heat) in great dis-
tances from A 0 is not greater than the total charge in, and near, A •.
In the neighborhood of A 0 (a distance = the order of a) any charge
located there infiuences the potential about rfa times stronger than does
a similar charge at the distance r. We conclude that deviation from
Eq. (13.06) by virtue of distant dissymmetries will be characterized by
a factor (1 ± ac) where a is of the order of ajr.
Case B. Contact between electrodes diHering in Q and ;., The es-
sential deviation implies that the locus for maximum temperature is
no Ionger the contact surface but is placed within the member with the
greater fl· Assurne members with e1 and g2 , g2 > ev and Iet h be the
distance between the top of the warmest isothermal surface and the
center of the contact. The constriction resistances on both sides of the
warmest surface are essentially equal, i.e.,
~- ..RLarctg!!:... = ..RLarctg!!:... + _h
4a 2na a 2na a 4a
or
e2 ( 1 - ! arc tg : ) = (11 (14.01)
giving hfa. For example with members of platinum and copper one
finds hfa = 0.593.
A consequence of this dissymmetry is that opposite points of the
members at the interspace between them no Ionger have equal tempe-
ratures. This disturbance is smaller between metals than it is in a
carbon-metal contact. This is treated in Oase C and it is found that the
dissymmetry in the carbon-metal contact causes an error in e of less
than 10%. In a metallic contact the error is much smaller, i.e., practi-
cally negligible.
The calculation of voltage and temperature differences in a bi-metal
contact is discussed in § 20.
Case C. Conta.ct between electrodes whose conductivities dilfer largely;
for instance, copper and graphite. The poor conductor is primarily re-
sponsible for the contact resistance, while the other, in this case copper,
plays the role of a perfectly good conductor. Therefore, with respect to
the graphite, the contact surface is approximately equipotential and
isothermaL It behaves as Ao in Fig. (13.02). However, the condition
that opposite points in the intermediate space have equal temperatures
is no Ionger satisfied.
Example. Consider the contact in one of the a-spots (radius a) be-
tween a carbon brush and a stationary copper ring. The thermal con-
ductivity in the brush may be Äb = 25 Wfm 0 K. We assume that the
interspace width is ;;:;; 10-7 m in the region considered. Now, 10-7 m

is about the mean free path, A, of the air molecules under normal cir-
cumstances. According to a law for heat conduction in gases the con-
ductance across a distance smaller than A does not depend on how
this distance decreases below A. Thus the thermal conductance per m 2
across the interspace is oftheorderof(thermalconduc tivityofair)/A =
0.025/10-7 = 2.5 · 1()6 Wfm 2 °K; cf. footnote to Example I of § 35.
The heat current is devided between a path along the constriction,
path X, and a path across the interspace, path Y, having heat con-
ductances Kx and Ky respectively. Not all the heat, (IU/2) that is
generated between the contact surface A 0
and the hottest section (located within
the carbon) reaches A 0 , and the supertem-
perature in the hottest section will be about
(1 - Ky/K:xJ@ instead of @ according to
(13.06).
We find 1 Kx > 10aA.b. For the calculation
of Ky we refer to Fig. (14.02) which shows
the ring-shaped area elements between the
contact boundary E 1 and the intersection
Flg. (14.02). Elements of the E 2 with the hottest equipotential surface.
interspace around a clrcular con·
tact spot of radlos a The temperature difference between oppo-
site points of both members at the element
with the radius r may be {}(r). For our rough estimation it suffices
to put
{}(r) = g r - a
a(lf2- 1)
namely so that t'Jo(a) = 0 and V
t'Jo(a 2) = @. The heat current through
the element 2nr dr is
2nrdr · 2.5 ·105 #(r)
Integrating through all elcments we obtain a heat current of
ay2
2.5 ·10 2n e f(r- a)rdr = 4.16 ·105 @a2
5
a}'2 -1
IJ
This means that the temperature difference, @, causes a heat cur-

rent against the thermal resistance of
Ky@ with Ky = 4.16·105a 2
Carrying through the calculation with a = 5 . 10-5 m in accordance
with Table (7.09) one finds
Ky/Kx< 0.08 (14.03)
1 See (35.07) that gives 10 < K; < 16
a,.~
§ 14. The fi"''-relation in (1ases of dissymmetry 67
i.e., the highest supertemperature, e, in the hot.test section ofthe poor
conductor is less than 8% smaller than the contact temperature in
a symmetric graphite contact with the same a and I.
Case D considers a heat eurrent, q, :ßowing aeross the eontaet surface
Ao into the member regarded. Therefore, more heat :ßows through
the area A [see Fig. (13.02)], than assumed for Eq. (13.03), with the
consequence that Eqs. (13.03) and (13.05) are no Ionger valid.
A relation similar to (13.05) is 6btained by introduoing the expres-
sion
q= YI (14.04)
where Y is a voltage. We now assume the density of the heat

current, q, through elements of A0 to be fairly proportional to the den-
sity, J, of the electric current through the respeotive elements.
(usually no considerable constraint on· the problem). Heat and electri-
city are considered to :ßow in the same paths. But, the heat :ßowing
a,cr.oss the surface A in Fig. (13.02) is no Ionger rpl, as in § 13, but is
(rp + Y)I. Therefore Eq. (13.03) changes into ·
dR
- dT = (rp + Y)I --· (14.05)
eA
and Eq. (13.05) into
- eld T = (rp + Y)dq; (14.06)
lntegrating from A0 to A one obtains

e
feldD
(J
= ! rp + Yrp
2 (14.07)
Case E1 is confronted in a contact between different metals when a

Peltier heat (11 I) is produced in the junction. If this heat :ßows to one
member only, the heat current is
/I I (14.08)
Then evidently 11 plays the same part as Y in (14.06). Therefore, Eq.

(14.07) is applicable to the member into which the PELTIEB heat :ßows.
In §§ 18 and 66,F such a situation is treated.
Case F1 considers a Thomson eftect by means of which heat is given
o:lf by current carriers coming from warmer to cooler regions of the
current path in a conductor. Tobegin with a conductor is considered
irrespective of whether it belongs to a contact or not.
1 Cf. the discussion on thermo-electric eft"ects in § 18.

The following calculations are applied to a positive THOMSON effect,

i.e., behaving as if the carriers were positive. Otherwise a negative
coefficient is to be used.
An electric current I ßowing from a section with the temperature
T to an adjacent section with the temperature T - d T liberates
THOMSO:N heat at a rate of
I a 1d T watt (14.09)
between these sections; a1 denotes theTHOMSO:N coefficient. It is obvious
that this heat adds to the RI 2 heat, changing Eq. (13.03) into
T
- dT = (cpl + IfatdT)dW
Te
or, with d W = ::/ and integrated,
J J J
a ~ e
eÄdif = ~ cp2 - dcp atd# (14.10)
6 0 6
Case G. The 9>..'}-relation when the current passe<J in a slanting path

through a boundary Ac between different material M1 and M 2 • Fig.
(14.11) illustrates two equipotential surfaces A 1 and A2 in M 1 and
equally marked continuations in M 2 where the dis-
tance between A1 and A2 is now smaller than in M 1
if e2 > e1 • For simplicity we assume that A1 is iso-
thermal in both M 1 and M 2 • However, the tempe-
rature difference d{} between A 1 and A 2 will, in gen-
eral, not be the same in M 1 and · M 2 • This means
that the isothermal continuation of A 2 in M 2 does
not coincide with the equipotential continuation of
A2 in M 2 • We conclude that the equipotential and
isothermal surfaces obtain a relative inclination
when passing the contact surface between M 1 and
M 2 • The condition for the validity of Eq. (13.06) is
Fig. (14.11). Bimetal only satisfied if WIEDEMANN-FRANZ law is valid
contact. The contact
surface Ae lntersects which means
the equlpotentlal
surfaces .41 and .4 2 ~ !?2 (14.12)
~=~
Then the distance between two subsequent equipotential surfaces is
altered in the same proportion as the distance between the isotherms
when the current traverses the contact from M 1 to M 2 , and Eq.
(13.06) remains valid across the contact. Evidently Eq. (13.06) endures
anisotropy of e and Ä as long as eÄ remains constant.
§ 15. Influence on the tp&-relation. KoHLER effect 69
§ 15. Influence of a thin film in the contact on the p..9--relatiou.

li:ohler e:ffect
We consider a contact with a thin film that is penetrated by the
conduction electrons by means of the tunnel effectl. The electrons
do not alter their energy Ievel whe:r'J. tunneling; but, since they land in
an anode with a lower negative potential than the cathode, they have
a surplus kinetic energy there. The increment of kinetic energy is given
off as heat, making the anode warmer than the cathode. A portion of
the liberated heat flows back to the cathode across the film whereas
t.he remainder flows into the constriction of the anode.
M. KüHLER calculated these heat portions, and thus made it
püssible to ca.lculate the temperature difference, {}1 , between anode
and cathode 2 at the boundaries between the film and the electro-
des. This effect influences the material transfer in opening contact.s,
cf. § 66.
The calculations in the present chapter are based on KüHLERS theo-
retical discovery that an analogue to Eq. (13.10) holds for the electric
and thermal currents through the film. Let 1fx1 = 111 be the electric
tunnel resistivity per unit area of the film against the voltage, U1 , that
is applied to the film, and let 1/Al be the corresponding thermal resisti-
vity against the temperature difference, &1 . Then, as has been proved
by KüHLER IJ r
where
~ = f1t Ä.t = L T 1 (15.01)
L""' 2.4 · 10-s (VtK) 2 J
In spite of the similarity of the formulas (13.10) and (15.01) the

appearance ofT has essentially different grounds in either of them. In
the first formula it results f:rom the temperature dependence of p, and
in (15.01) it results from the temperature dependence of ;..1.
Weshall now calculate {)1 which is the difference between the super-
temperature, e+' of the anode and, e-' of the cathode, and we do it
under simplified conditions, by treating the coefficients ;..1 and a1 of the
film as weil as the coefficients Ä. and (} of the contact member as con-
stants; the value of ;,.1, a1, Ä and e shall correspond to approximately
the median temperature in the constriction. With this assumption one
omits the complication of somewhat differing constriction voltages in
1 For details about the tunnel effect, see § 26.
2 KüHLER has not yet published bis calculations. Only a provisory formula has
been cited by JusTI and SCHULTZ [J] p. 96.
70 Stationa.ry Contacts
anode and cathode, and the condition for continuity of the electron
current through the ·contact is automatically satisfied by the pictures
of the equipotential surfaces in the electrodes. The voltage in the end
surface of either member may be called ± U.
Another simplification will be made with respect to the film. The
existence of a film causes an approach to Problem B in § 5 and it makes
the inhomogeneity ofthe current density within the contact area, A 0 ,
smaller than in a clean contact. The second simplification consists in
assuming U1, {}1 and the electric current density, J, tobe constant all
over the film. Then, the heat which fiows into the anode, is
I U1 - ).11J1A,. (15.02)
where the first term is the heat carried by the tunneling electrons and
the second term is the portion that fiows back to the cathode through
the' film. According to the assumption above, the heat fiow that is ex-
pressed by (15.02) has a constant density, i. e., it is proportional to J,
which is the condition for Eq. (14.04).
Evidently U1 - A,~A. plays the same part as Y in § 14. Thus,
applying Eq. (14.07) to the conditions in the anode we have
! U + (Ur- A,~A·) U
e+
fe).d{} = e).8+ = 2 (15.03)
0
The corresponding equation for the cathode is
e).8_ = ! U2 + Ät~A. U (15.04)
Subtraction of (15.04) from (15.03) yields
eUJ1 = ( U1 - 2 Ät~A·) U (15.05)
In Eqs. (15.03) to (15.05) we transform Ä,~A. as follows. First we

write it a1).1()1 1;,, where a1 is the tunnel resistivity per unit area. Then
we recall that according to Eq. (15.01) a1).1 = LT and that for metals,
according to WIEDEM4NN-FRANZ-LORENZ law, L T is equal to eil, thus
atÄt = (!Ä, and that ~~ = U1• Finally one obtains
rur-=e
). .Q
I -11,U1
Li, ).
(15.06)
Inserting in (15.05) gives
U1 U
f}f = (15.07)
1 n
II" 2U + u,
§ 16. The infiuence of the Joule heat on constriction resistances 71
or assuming U1 < 2 U = U, the total contact voltage, roughly
#r "., Uf_ ( 1 __
2le
!!!)
v (15.08)
Inserting (15.08) in (15.03) and (15.04) finally gives
eÄ e+ ! U + ; u, u + ! uj 1
"., 2
(15.09)
eÄ8_ "., ~ U + ; U U- ! Uj J
2
1
Eq. (15.07) corresponds to the first ofKoHLER's equations, which were

published in JusTI and SCHULTZ [1] p. 96.
An experimental confirmation of the KüHLER effect is presented in
§ 66D and E.
Example corresponding to the problern in § 66,E. Assumptions:
Pt contact carrying a film with tunnel resistivity a = 5 · 10-9 Q cm 2 ;
1=10A; the average (JÄ>:::4·10- 5 V2tK; melting temperatureinthe
anode, thus 8+ = 1750°. Eqs. (15.08) and(15.09)aresatisfied by U ""'0.3V,
U1 "., 0.14 V, A. "., 3.6 ·10- 7 cm 2 , and
(15.10)
§ 16. The inßuence of the Joule heat on constriction

resistances1 in symmetric contacts
If a contact constriction is heated by the current flowing through it,

its resistance will differ from that ofthe unheated constriction. One may
expect that the contact resistanceR (8)where 8 is the supertemperature
n the contact surface, can be expressedas
R(8) "., R(O) (1 + 1X1 8)

Here R (0) means the resistance of the unheated constriction. In fact,
such a linear relation is true with surprisingly high accuracy. Due to
the very nature of the constriction, the coefficient a1 in the relation will
differ from the normal temperature coefficient, a, that appears in the
1 The validity of OBlll's law is presupposed. It was confirmed, in 1930, by R.HoLM

and STÖRMER [12] for platinum up to 4. 1010 Afm2 and by BARLOW [1] for plati-
num and gold up to 1.3. 1010 and 2. 1010 A/m 2 respectively.
binominal (1 + af) of a conductor that is uniformly heated, and it is

obvious that a1 will be smaller than a because the average temperature
in the constriction is smaller than 8; as a matter of fact, IX1 turns out
tobe
(16.01)
that is
(16.02)
For the following deductions, including proof of (16.02), we compare

an actual constriction K (e Ä) (where e and). are temperature dependent)
with a geometrically equal constriction K (f!oAo) the material of which
is imagined to have constant e = eo and constant). = Ao· Both con-
strictions are traversed by equal currents and have equal eo and Äo
belonging to their end-surfaces A •. Thus, R (eÄ) is the actual resistance
against I, and R(e0 A.o) is the constriction resistance that one would
measure with a very low current.
We consider1 a series of niveau-surfaces A 0 , A1 , A 2 ... in K(eA.)
and the homologous surfaces in K (e0 A.o). They are equipotential and
isothermal in both bodies; cf. § 13 and Fig. (13.02). The potentials
fJJ (eÄ) and fJJ (e0 A.o) are measured from the surface A 0 where the poten-
tial is zero. The supertemperatur, f, in K (f!Ä) has its maximum 8 in
A 0 • .Actually also K (e0 Ao) has its maximum temperature there although
a lower one. If the potential differences between consecutive niveau-
surfaces in the two hoclies are dfJJ ((!Ä) and dfJJ (e0 A.o) we have
d!Ji(lloÄo)
d!Ji(eA) = e
llo
(16.03)
Approximate deduction. The following deduotion, although being

approximate and valid only for af-«: 1, is given beoause of its simpli~
oity and conspicuity. We infer from (16.03)
or approximately (1 - a(}) dfJJ ((!Ä) = dfJJ (l!oAo)· Here we substitute for

(} an approximate expression obtained from (13.06) with an average
value of e and ).
fJ = e - _!_ 'P'(e A> with

2 'i1
1 The deduction is independent of the shape of the contact surface, as in § 13;'
§ 16. The inß.uence of the Joule heat on constriction resistances 73
Integration from A 0 to the end-surface A. in one member gives
U(IIA)
/ [ 1- r:t.8 + cc~z~.t)] dq;(eA) = U(eA)- r:t.8 U(eJ.) + 6 ~ U3 (el)

0
= U(el)[1-rx.8+; e] = U(eJ.)[1-: rx.eJ

J
U(lloAo)
= dq;(eoAo) = U (eoAo)
0
With rx.8 sufficiently small to allow [1 - 2/3 (r:t.8)]-1 to be re-

placed by 1 + 2/3 (o:8), one obtains
U(e.t) = R(e.t) ~ 1 + .!rx.e
U (!!0 Äo) R (e 0 Ä0) 3
or, since R (el)f R (e0 ~) equals the ratio between the resistance R(8)
of the actual body at the respective current and its resistance R (0)
at small current, we have for a long bi-constriction
(16.04)
which conforms to (16.02).

Aceurate calculation. In (16.03) if we substitute dq; (e ).) from (13.05):
eoM,fJ
dq;(eoAo) = - IJI(eÄ) (16.05)
To make (16.05) integrable, we express q; (e,A) as a function of the

temperature according to (13.06)
(16.06)
and obtain1
(16.07)
We define R,.. as the resistance between the equipotential surfaces

A 0 and A,.., where the index f1. is merely a denotion. As said above,
the shape of A 0 is indifferent; A0 may even consist of several a-spots
of different shape. The only requirement isthat Ao and A,. are niveau-
1 Cf. DIESBELHORST (J] p. 317.

surfaces. In the special case of .A0 being an ellipse lt may be the para-
meter of Eq. (4.10). It is obvious that in general
R,.{eoÄu) IP{l'oÄu) (16.08)
R,.{el) = IP{el)
and by substituting in (16.07) we finally find
(16.09)
Note(16.10). Because Ä. enters into both the numerator and denomi-

natorofEq. (16.09), its effect is partly cancelled and the ratio R;;r~;)>
is only slightly dependent on Ä. at low temperatures.
Weillustrate Eq. (16.09) by typical example~:~, A, B, C;
Example .A. Assumptions: The WIEDEMANN-FRANZ-LoRENZ law is
valid, i.e., ()Ä. = LP according to Eq. (13.10), and Ä is constant, Ä. = .a.,.
Wehave
(16.10)
and
eoÄo=LPo
Substitution in Eq. (16.09) yields
Rp{l'oÄu)
-"-""'-"--.::.:. =
R,.(eA) VTlt-
Po
Tl
I VT&-
Te
d.T
PI
=
VT~- PI
T0 . T (16.11)
arccos-
Te
T
which also can be written

R,.(eoÄu) = Po arctg VTL- pz (16.12)
Rp((.IÄ) VTL- pz T
In the case of a long bi-constriction
u = 2}/L vP~ - P~
Inserting this in (16.12) we obtain
R(~?Ä) = [2ToVLarctg~]·-1 (16.13)

R (l'0 Äu) U 2 T0 VL
. . R( ~
In F1g. (16.14), where P 0 = 293 °K, R( l' ~) according toEq. (16.13)
. 1 l'o"'
1s p otted against U as solid curve. The dashed curve shows e accord-
§ 16. The influence of the Joule heat on constriction resistances 75
ing to Eq. (13.13) plotted against U. If R (eA.)/R (eoAo) is plotted against

e with both taken for the same u (from Fig. 16.14), it turns out that,
1&00 8
!WO /I /
~ 1tQO I V
~
~1000
i V
/
Flg. (16.U) presnpposes a mono-
metallic contact and valldity of ~800 1- - o.l 1Ar11'-J/Rft~o'-oJ _1-
WIBDIIHANN·FRANZ law. The &olld lcoo / / I
cnrve shows R (11 .1.)/ R (Qo Äo), the
dashed one the supertemperatnre e
~
'fOO
V I
in the contact surface; both plotteä
agalnst the total voltage U zoo / -1 I 1
/ I I 0
00
.0.1 az Q.J 0., OS a& 0.7 Oß Q9 tOV l1
u-
with great accuracy
R(eA.) = R(e0 A.0 )(1 + 0.00227 @) (16.15)
where the coefficient before e agrees with
(ofEq. 16.02)
W. MEISSNER [1] deduced the samerelationship in the form
R(eA.) ~ R (e0 A0 ) [ 1 + ; 8 (~ :~)" ~ 0 ] (16.16)
Observing tha.t
9'((.JÄ) = VL(T~- T 2 ) =I
Rp (!? Ä) R~' (e J.)
one obtains Eq. (16.17) from (16.11)
R~' (eo A.~) =

VL
T T 0 arc cos Po
T (16.17)
This is often significant in relating the temperature t.o the location of

the isotherm considered; see Example B.
Example B. Heat, q watts, enters the constriction .across the surface
A0 • We solve the problern under simplified conditions: 1. A. = Ao =
constant and q is proportionaltoT with WIEDEMANN-FRANZ law being
valid. 2. The heat ftows in a manner to allow calcula.tion with Eq.
(14.05) where we have put.
q= rI
A particularly tangible solution is not obtained by directly using
Eq. (14.05) but modifying the problern as follows. As for symbols see
Fig. (13.02). We imagine the contact member tobe extended[seeF ig.

16.18] to a cross sect.ion A~ with the following property: Between A~
K and A0 the JouLE heat I Y is generated,
and no heat is flowing across A~ which
is both equipotential and isothermal
(which could be achieved if the mating
;4)4.1' contact member were extended in the
\ I~ f
\
- ' - -'---L----'-11
I , same way) and has the fictive tempera-
Fig. (16.18). Abody K, which carries ture T x. Then, A~ plays the same part
current, lmaglned to be elongated for the following
to A~ so that .d~ obtalns the charac·
deductions as A 0 in § 13
ter of .II 0 in Fig. (13.02) and we can apply (13.12) obtaining
(16.19)
and
(16.20)
giving T 9 if rp , T and Y are known. Let R~ be the resistance between
A~ and A 0 , and R' that between A~ and the isothermal surface in
question with the temperature T. Then, a.ccording to Eq. (16.17)
1 1fL Te
R0 ((>0 A.0 ) = T T 0 arccos T,
and
R' (Q0 A0 ) =V{' T arccos ~.
0
The difference R' (e0 .f.o) - R~ (e0 .f.o) is the resistance R,_. (e0 Ä.o) in the actual
constriction when unheated. Hence, we finally have with Rl'(e0 Ä.o) and
Rl' (eA.) referring t.o t.he actual, not extended member
(16.21)
or
Rp (t?o Äo) = 1fL T o (arc cos .!.__ ·.- arc cos _Te] (16.22)
Rp(t?l) 'P(t?l) T, T.
where according to (16.19) and (16.20)
rp(eA.> = VP + L(T~ - T Y 2)- (16.23)

and consequently
u = VP + L(T~- T~)- y (16.24)
Assuming T 9 , T0 , T and Y tobe given, Rl'(eoÄ.o), Rl'(eA.) and U
can be calculated.
Thus far we have imagined the heat, I Y, tobe positive, represent-
ing an inflow of energy. But, if the thermo-electric PELTIER effect
brings forth I Y, it may be negative as weil, dependent on the direc-
tion of the current. In the case of a n egative I Y, the section A~, where
§ 16. The influence of the Joule heat on constriction resistances 77
the temperature has its maximum, lies within the contact member con-
sidered, and R~((>0 Äo) and R' (eoÄo) add. Eqs. (16.19), (16.21) and (16.23)
change into
(16.25)
valid for cp > Y,
(16.26)
(16.27)
with corresponding changes of (16.22) and (16.24).
Example C. Integration ol Eq. (16.09) without using Wiedemann-
Franz law. This is valuable because some metals, nicke! for instance,
deviate considerably from this law. For non-metals the law is not valid
at all. The Integration is easily obtained1 with the assumption that
both members are of the same material with
el = eolo(1 + e1J)
and (16.28)
where the coefficients e and ß are regarded as constants and where the
index 0 refers to the temperature of the end-surface A •.
Note that {} is measured from the temperature of the end-surfaces,
thus e and ß may differ from the values given in standard tables.
According to Eq. (13.06)
e
cp~(1J)=2eolos/(! +D)dfJ=eoloe((! +8r -(! +DrJ (16.29)
(I
Substituting this in Eq. (16.07) gives

9
cp eo o - v-e V(-+9)-
( l ) - 1/eo'-of
1
(1+{J{})iJ{}
(I
2 (-+{}
1 )2
e e
--- !. + {}
=8 {J 1/eoÄoarccos-
8 - - + _f!__P(el)
V
-
e e 1 a 8
-+""
e
= 8 - fJ 1/lio'-oarctg
8 V e
VH- + ey- (-} + {}Y + i!__cp(el>
1 {} e
-+
e
=e-{J veoÄoarctg(.11 e p({}) )+_!!__cp(el)

8 e VeoÄo 1 + e{} e
---
1 Deduction first given in R. HoLM and STöRMER [12].
Hence,
IP(eoAo) = R,.(eoAo) = e- ß veo~arctg

q.>(eA) R,.(eA) eq.>(eA) e
[liVeoÄu 1IP(eA)
s
+
J + }_
e{) e
(16.30)
and by integrating to the end-surfaces, where {} = 0 and rp (eA.) = UJ2,

we have in the case of a symmetric contact, considering that the ratio
R (e0 J.o)JR (!?Ä) equals the ratio R (e0 J.o)/R (eÄ) for a long bi-constric-
tion as defined for (16.04)
R(eoÄu)=e-ß 2 l;eo;~arct [lCs

R((IÄ) e u Ve g VeoÄu
u] +}_=B(eoÄu>
2 e B(eÄ)
(16.31)
§ 17. Distribution of the temperature in a symmetrie

constrietion with circular contact surface at given current
The aim of the present problem is to determine the position of any
isotherm characterized by its temperature, T, or its supertemperature,
{}. The position can be defined geometrically by the parameter, fl• in
Eq. (4.10), or, in cases when the simplified calculation according to
Fig. (1.02) is employed, by the distance r from the center of the contact
to the respective isotherm. It may be more convenient and just as
satisfactory to define the position by R,.(e0 Äo)/R (e0 Äo), the ratio be-
tween the resistance in the unheated partial constriction between the
contact surface and the equipotential characterized by p, and R (eoÄ.o)
the total constriction resistance in the unheated member considered1 .
However, R1, (!?oAo) and R (!?oAo) will have to be calculated, since it is
likely that only the potentials cp ((!A) and U ((!A)/2 pertaining to the
. 1 U(el)
current, I, are known, 1. e., that R,. (eA.) = cp (eA.)jl and R (eA.) = 2 - 1 -
are given. The conversion from R,. (eA.) and R ((?Ä) to R,. (!?oÄo) and
R (!?oAo) is deduced in § 16.
Examples A, B and C will show the modes of calculation, which are
based on the formulas of § 16 using the same symbols.
Example A. Here !? and A. are considered as constants, thus
R,. ((?0 A6 ) = R,. (e Ä), both = R for fl = oo
Wehave
(~r
and
= 2eoA.oe
1 As for the symbols R11 (eoÄo), R11 (el) eto., see § 16.
§ 17. Distribution of the temperature in a. symmetric constriction. . . 79
and by division
(17.01)
Limiting the subject to the simplification which is represented by

the model in Fig. (1.02), position co-ordinates are introduced on the
basis of Eq. (4.08) by putting
giving
(17.02)
where b is the radius of the contact sphere representing the contact

surface, and r is the distance from the contact center to the respec-
tive isotherm.
Example B, corresponding to Example A in § 16. We assume the
validity of the WIEDEMANN-FRANZ law and Ä. = A.o to be constant.
FormingR"(e0 Ä.o) andR,.,(eoÄ.o) = R(e0 Ac,) according toEq . (16.17) and
dividing, one obtains
Rp(l?0 Au) _ arc cosT/Ta
(17.03)
Roo(e 0 Ä.0) - arccos T 0 /T a
whereR" (eoÄ.o)/ Roo (l!oÄ.o)is used to define the location in the constriction.
Disc·ussion of Fig. (17 .04). Eq.
lO
(17.01) is illustrated by curve 1 in ~
the figure . Its course is the same ~
for any e. Eq. (17.03) is illustra- ~
ted by curve 2, calculated for ~'<!
e = 1000° and To = 293 °K. ~as 2~
Although a considera bly !arge ~ ~
value of e has been chosen for '\
the calculation of curve 2, there is i'\
only a small deviation between 1\
curve 2 and curve 1. This indicates \
that in many cases fairly good com- 0 as 1.0
HpJ9o '-oJ/R<»(9o 'J-.oJ-
putations can be based on curve 1.
In case of a circular, flat con- Flg.(17.04). Curvel illustrates Eq. (1 7.0l)
and curve 2 Eq . (17.03)
tact surface with radius a, the
introduction of position co-ordinates is generally made by means of
Rp <eo Au) = _!_ arc tg 1fP (17.05)

Roo (eo Äo) n a
which is a consequence of Eq. (4.13).

Example C concerns the temperature within a film in a symmetric

contact with metallic members. The film is assumed to be thick enough
that the tunnel current through it is small. Let the voltage across
the film be 2 Y and across the mono-constriction, U volt; cf. Eqs.(16.19
and 16.20). With QA belanging to the metal, the supertemperat.ure {} 111
in the contact surface betweeri metal and film satisfies
IJm
2j eÄd# = (U + Y) 2 - P (17.06)
0
or
IJm
f eÄd# = u Y
0
(17.07)
if U2 can be neglected against 2 U Y. Eq. (17.07) gives {}m·

The maximum supertemperature 0 - {}m in the film above {}m of
the metal is according to (13.06)
(17.08)
where (eÄ)t that belongs to the film, is of the order of 10"' with o: > 1.
Even with a considerable Y, of for example 5 V, 0 - {}m is small. In
other words, the film is practically isothermal up to high Y-values.
This surprising conclusion results from the fact that the high value of
(eÄ) 1 is based on a high e that keeps the current and U very small; Ä is
of the order of 1 watt per m °K.
§ 18. The (equilibrium) temperature distribution in the

constriction of a contact between two metals with different
conductivities, both obeying Wiedemann-Franz law.
Thermoelectric effects
A. Features of the thermoeleetrie effeets. Consider a current carrier
that is moved from position (1) to another one (2) whereby the carrier
in (1) was on an energy level higher than it meets in (2). To restore
equilibrium it has to give off the surplus energy at (2). Thermoelectric
effects are such energy transports.
It is relati vely simple to explain both PELTIER and SEEBECK elfects
between semiconductors. Fig. (18.01) demonstrates features of a con-
tact between semiconductors B 1 and B 2 , where B 1 has a higher density
of free electrons than B 2 • In order to restore equilibrium (i. e., equal
density on equal energy Ievels) the conducting band of B2 is lifted so
that B 2 assumes a higher negative potential than B 1 . Such voltage
§ 18. The (equilibrium) temperature distribution in the constriction... 81
differences constitute the SEEBECK effect; see below. Now let a current
move electrons from B2 to B1 • These electrons evidently belong to
higher average niveaus than the average niveau in B 1 • In order to be
adapted to the average Ievel in B1 , they give off energy in the form of
heat. In other words, the electric po-
Negolife
tential energy is converted into heat at potenliol
the junction. An opposite current pro-
duces a corresponding cooling. This is
the PELTIER etfect.
In the case of metals, the distribution
of electrons on energy Ievels is of the
FERMI type, which complicates the picture
8oHon:
although the effects still resemble the
description above. However, in metals, Boliom ----"' - - -- '--L
both PELTIER and SEEBECK effects are
relatively weak because, owing to the
degenerated partitionl, the carriers Fig.(18.01). PELTIER effect. Equilib·
change their average energy very little rium between two semiconductors
B 1 and B2 where B, has a higher
with the temperature. electron density than B,. The den-
of electrons on any energy Ievel
It is evident from the foregoing that sltyls plotted as abscissa to the le!t !or
the PELTIER and SEEBECK effects are B 1 and to the right for B 2
related to each other. The relationship
shows up in the formulas (18.02)and (18.03) which refer to Fig. (18.04).
The conductors a and b of different materials constitute a circuit that
contains a voltmeter v8.
If the junctions are held at different temper-
atures T and T 1 °K with T > T 1 , the voltmeter indicates a SEEBECK
voltage that may be written
T Eba(T)- T 1Eba(T1 ) (18.02)
Eba is a slowly varying function of T and is called the differential
SEEBECK coefficient. It is termed positive when Eba is directed to move
a positive current from the lower temperature T 1 to the higher tem-
'perature T in the conductor a.
The PELTIER heat at a junction having a temperature T has the
amount
II T Eba ( T) I watt (18.03)
This junction is cooled when the SEEBECK voltage T Eba (T) tries to
move its own current in the direction of the circuit current2 I. With
respect to the PEL'I'IER effect, T Eba(T) is labeled Ilba(T) the PELTIER
coetficient.
The THOMSON etfect appears when the carriers are moved along a
temperature gradient within a conductor, if the distribution of carriers
1 See §II. 2[ cancome essentiallyfroma battery, notsketched inFig. (18.04).
6 Electric Contacts, 4th Ed·
on energy levels varies with the temperature, i. e., the carriers have
heat capacity. In the shell between a section of the conductor at the
Fig. (18.04). SEIIBECK effect; a and bare two

different conductors in a circuit that con·
tains a Voltmeter. If the junctions have
different temperatures, a thermo·electric
voltage appears
temperature T and an adjacent one at the temperature (T - d T),

carriers, running in the direction of decreasing temperature, give off
the heat at the rate
(18.05)
within the respective shell; see Eq. (14.08). The THOMSON coefficient
fit is positive when the carriers run in the direction of the current I,
viz., are holes (as in zink).
The THOMSON coefficient fit is zero when either the distribution of
carriers is independent of the temperature or when positive and nega-
tive carriers cancel their THOMSON effects. If this happens in both con-
ductors of Fig. (18.04) the differential SEEBECK coefficient Eba is in-
dependent ofT, cf. § 66B.
B. The problern of temperature distribution in the case of Peltier
heat in the contact. We will consider t.he condition that Eba =~ constant
which means (as said above) absence ofthe THOMSON effect. Regarding
the rough approximation
Conr/ucfor made below, we state that
it makes little difference
I
'I whether or not both PELTIER
and THOMSON effect are in-
lAe volved, which affects only the
I average distance the thermo-
I
I electric heat flows.
{/ Fig. (18.06) illustrates po-
Q sitions of niveau-surfaces
Fig. (18.06). Symbols for niveau surfaces around the contact between
the metallic members M and
M1 ; the latter being the poorer conductor. Symbols are marked on
the figure. The maximum temperature, 8, is located within the poor
conductor. On this niveau-surface the potential is put = 0.
§ 18. The (equilibrium) temperature distribution in the constriction... 83
Fig. (18.07) shows the temperature distribution in the contact
illustrated in (18.06) . Because both members obey WrEDEMANN-FRANZ-
law, the distribution is essentially symmetric around the hottest niveau-
surface. We assume that a PELTIER effect develops in the contact sur-
face at the rate of I T.E watt where I is the electric current and Tc the
actual temperature in the contact surface. E is the differential SEE-
BECK coefficient that, for sim-
plicity is regarded tobe independ- T
["K]
ent of the temperature. In the
Te
event of PELTIER heat consump-
tion, E is negative.
The thermal gradient is di-
rected from the contact surface
into the good conductor. The
total PELTIER heat therefore
0
ßows into M, constituting a I
thermal current that causes a 0 wll iJ
temperature increment Z at the Fig. (18.07). Temperature distribution in the
contact surface. This adds to the constriction of a contact between different
temperature produced by JoULE metals both obeylng WIEDEMANN-FRANZ law.
The symmetric parabolatype of the distribution
heat, whereby ls correct if no thermoelectric effect interferes.
The existence of a positive PELTIER effect lilts
Z =I (T.- T 0 ) E W (18.08) the contact and maximum temperatures to
higher niveaus, causiug a small dissymetry that
with W = the thermal resistance is neglected in the ftgure.
in the constriction M. Thus,
l
only T. - Z is produced by JOULE heat. The following equations1
result from the cr*-relation according to Eq. (13.12).
U2 = L (~ - T~)
V 2 = L(T~- T~) (18.09)
(V+ tp) 2 - V2 ,." L[(T.- Z) 2 - T~j
with the total constriction voltage
U = U + V + 'P + T.
( - T 0) E (18.10)
where (T 0 - T 0 )E is the total SEEBECK voltage.

We first disregard the PELTIER effect putting Z = 0. Then Eqs. (18.09)
and (18.10) give
U0 = V0 + VJo and U 0 = 2 U0 (18.11)
where the subindex 0 refers to Z = 0.
1 The last of Eqs. (18.09) is an approximation becaul!e a small deviation from
the parabola is neglected.

The maximum temperature Te is obtained from Eq. (13.12) (with

U = cp), and an approximate temperature distribution is represented
by a parabola in conformity with curve 1 in Fig.(17.04). Tc, the tem-
perature in the contact surface, is read from Fig. (18.07) as the ordi-
nate at the point B that makes ABfBO = ehh·
With the PELTIER effect represented by Z, we obtain the following
from Eqs. (18.09) and (18.10) by neglecting the term with Z 2
Noticingthatgenerally W = RfeÄ.and that the average f!Ä. = L T.; To

we obtain according to (13.10)
lJ'
W ""I L(T. + T 0 )/2
Hence
(18.13)
and
or
(18.15)
An 'additional equation is obtained considering that T 0 is the or-

dinate in Fig. (18.07) where ABfBO equals the ratio between the
electrical resistivities of the metals M and M 1 at room temperature.
The factor (Tc- T 0 )/U is of the order of (Te- T 0 )jU = Ve12e Ä..
where@ is the maximum supertemperature inM1 , and e~ is valid for a
ve
supertemperature of about 8/2. Thus, the relative influence of the E-
term on U and R approaches zero as does. Since Eis of the order
of to-s VtK, we see that a measurement of the resistance of a bimetal-
lic contact is generally only slightly influenced by the PELTIER effect..
However, the additional temperature Z can reach the order of 10% of
the contact temperature Tc. Therefore, the PELTIER heat plays a part.
for resistance welding.
The contribution of the PELTIER effect in resistance welding was
first described by S. SCHOLZ [1]. He observed with gold-palladium
contacts a change of UfU0 of the same order as calculated in the above
§ 19. Temperature distribution in the constriction of a contact ... 85
example. He deduced a relation between the PELTIER and JouLE heat

for the case of resistance welding. With two different methods of meas-
uring, he found that the PELTIER energy amounted to about 10% of
the total heat produced.
§ 19. Temperature distribution in the constriction of a contact

between members of very different conductivities
For simplicity, we will calculate as if the constriction belonged to
a single a-spot whose dimensions, however, do not appear in the for-
mulas. The formulas are generallyvalid and independent of the shape
of the contact surface by virtue of the character of the gJ-0.-relation.
Cf.§ 13.
Fig. (19.01) with less symmetry than Fig. (18.07), illustrates the
temperature distribution around the contact between members M and
MI> the latter being the poor
conductor not obeying WIEDE-
MANN-FRANZ' law. The abscissas
have the same meaning as in
Fig. (18.06) . The maximum tem-
perature, e, is located within
the poor conductor. At the cor- 0
responding niveau surface, the -(" tV} -V 0 voll U
potential is chosen 9? = 0. The --+----- ~ (Ro?.
J /R,., (qo'-oJ
temperature in the contact sur- Fig. (19.01). Temperature distribution around
face ist?. In M1 , the distribution the contact between graphite and nicke!
curve is parabolic. Exact cal-
culations for M 1 Iead to transeendental equations which have to be
solved by successive approximation. However, if M 1 is a conductor
muck poorer than M a simplification can be applied that Ieads to an
elementary treatment. This can be used as a first step to, sometimes
even resulting in, a final solution, as in the example below.
The simplified equations are obtained by assuming : 1. the heat to
ßow from- M 1 to M through only the electrically conducting areas ;
2. that Ä1 and Ä are constants; 3. that (h is a constant and (! has the
temperature coefficient cc Using Eq. (13.06) and Fig. (19.01) we find:
V2 = 2(!1Ä1 (8- 0.) (19.02)
U2 = 2e1 ;.1 e (19.03)
Considering the heat development in M as negligible so that the

entire heat current in M originates in the carbon, the average tempe-
rature in M becomes ()j2 and we have1

V'(el) = (1 + a.{}j2)V'(e0 Ä0 )
Because here V plays the same part as Y in Eq. (14.07), we use this
equation and obtain
tf(el) + 2V'fJ(eÄ) = 2g0 Ä.(1 + a.{}f2){}
and neglecting V'z
(19.04)
It is obvious that
V' (1/o.Äo) = l?o (19.05)
v+u e1
Substituting from (19.02) to (19.04) in (19.05) Ieads to
l
2~ye 1 ~ AB
(9- {}) + 2~ + A v8 - () = 2~ + .a (19:06)
and regarding this as 8 quadratic equation of ve -{}

1~=-A_1fe
V""-v 2~+AV""
or (19.0'1)
-o = ~A~<A + ~> e
(2~ + .A)2
If either 8 or () is known, Eq. (19.06) provides the solution. If a.lone,
the total voltage U (gÄ) = U (gÄ) + V (eÄ) + 'IJ.' (gÄ) (with V' usually
being negligible) is given, one obtains 8 1 according to (19.03) as a first
appro:ximation to 8 from
(~r = 2g1Ä.l 81
Further successively improved appro:ximations to V, (U- V),{} and
8 must be calculated until Eq. (19.05) is satisfied.
Example that has a bearing on Table. (27 .11 ). Assumptions: M con-
sists of nicke! with e = eo(1 + 0.004{}) where l.>o = 8 .1Q-SQ m, and
Ä = 80 m~K ; M1 consists of graphite with et = 9 · 10-6 n m and
Ä.1 = 40 m~K ; the contact temperature is near melting, i. e., {} = 1400°.
With the method described above one finds directly 8 = 1870°;
{} = 1400°; u = 1.16 V; V= 0.582 V; V' (e0 Äo) = 0.0155; V' (e.l.)
= 0.059. Fig. (19.01) shows the temperatures plotted against volt-
ages. The total voltage is U + V + V' = 1.8 V.
1 For the meaning of V'((), Ä) and '1'(1/oÄo) see § 16.
§ 20. Resistance-voltage characteristics of clean symmetric contacts 87
§ 20. Resistance-voltage characteristics of clean symmetric

contacts. Softening and melting voltages1
A. Introduction. The resistance-voltage characteristic of a contact,
called R U-characteristic, isadiagram where the resistance R is plotted
against the voltage U. In Fig. (16.14) the ratio R (e/..)fR (e0 /..o) (accord-
ing to Eqs. [13.13] and [16.13] for a long constriction) is plotted against
e
both U and the supertemperature in the contact surface. The solid
curve represents a RU-characteristic that is typical for clean metallic
contacts under the condition that the metal satisfies the WIEDE-
MANN-FRANZ law and that the contact area remains constant. The
calculation is carried through for a bulk temperature of 293 °K. Since
the resistivity increases with rising temperature, the ratio R (e).)/
R (e0 /..o) increases with rising U. We term such a characteristic ideally
metallic.
B. Discussion of Diagram (20.01). In Fig. (20.01) curve ABO was
calculated according to Eq. (16.13). Curve A BDEF refers to a cop-
per crossrod contact with
P = 10 N. It shows the soft- 10•10
ening drop, BD, and the Q •
5 /
melting drop, EF. FG and y /
--
D H arereversible branches 2 •
In order to attain a t~ /
certain perspicuity, we often Q:: A ~ V 1

--
1
plot the R U-characteristics /
in logarithmic co-ordinates. 7T /
Q5
This has been done, for G V
OJ
example, in Fig. (20.01)
where the upper curve, when 001 ODJ ODS at t12 OJ o.5 OJV
referred to the scale on the u-
right side, is obtained with so 100 zoo taJ 7t1J 10fK) ZIJ(J(J"
the same data as used for Superkmperofurt fJ
Fig. (16.14). In Fig. (20.01) Fig. (20.01). Typical R U·characteristics of a symmet-
ric clean metallic contact. Curve .A B 0 is calculated
the temperature scale which according to Eq. (16.13). Curve .A BDEF refers to a
is unequivocally associated copper crossrod contact with P = 10 ·N. It shows the
softening drop, BD, and the melting drop, EF. FG
with the voltage scale by and D H are reversible branches. U in the tlgure means u
virtue of the cp-t?-relation,
is for convenience plotted below the voltage scale. If we assume that
the material is copper, for which the WIEDEMANN-FRANZ law is valid,
1This chapter is mainly based on R. HoLM [4], and R. HoLM and E. HOLM [5].
2 ÜHOUDHURI et al. [1] give good information applicable to Fig. (20.01) and its
discussion.
the upper curve of this figure will correspond to measurements on

crossed rod contacts with P = 10N (when referred to the scale on the
left side). A greater load lowers the curve with respect to the left scale,
but it is easily seen that the scale on the right side moves down the
same step. The consequence is that the whole curve preserves its form.
This is because the formulas for the ratio R ((! A)/ R (l?o ).0 ) do not
contain the dimensions of the contact area, although R ((!, ).) and
R (e0 Ao) themselves do. It is this constancy of form that constitutes
the essential advantage of the logarithmic representation of a RU-
characteristic.
Now let us discuss how the R U-characteristics are obtained by
measurements. It is advisable to use crossed rod contacts and the de-
vices described in § 8 and Figs. (8.03) to (8.05). We must be aware that
the contact area and consequently R ((!0 ).0 ) remain constant through-
out for the ideal course of the curve A-B-0 in Fig. (20.01). This is
not automatically realized. It has to be achieved by keeping the con-
tact closed for some time or otherwise aging it.
There will always be typical deviations from the ideal characteristic
A-B-0. This curve may be realized by measurement up to the point
B. At higher voltage, the observed points willlie along the line B-D-E
and finally the measurements will not surpass a limiting contact volt-
age which is bound tobe on the verticalline E-F. This branch is com-
posed of voltages, rising and immediately dropping when the contact
area inreases by melt.ing. Two such steps are indicated.
On the other hand, after having reached the point F, the current,
thus also the voltage, is decreased, and the branch F-G may be meas-
ured. This branch has the form of the ideal curve A-B-0 and will be
discussed below.
The sharp deviations from the ideal course A-B-0 are caused by
softening of the meta] within the contact constriction at certain con-
tact temperatures (9 (at D), and finally by melting. The latter process
determines a Iimit for the voltage at E-F. Both effects shall now be
discussed in detail.
C. Softening. Suppose, for simplicity, that the measurement is made
with annealed members under a load that forces the contact to yield
plastically. At contact closure the deformation around the contact
surface will produce a certain degree of strain hardening. With in-
creasing current passing through the contact, the supertemperature,
e, in the contact surface, A0 , will rise by a corresponding amount.
When the softening temperature is reached, strain hardening disappears
in and around A 0 (cf. § 7 and §I); the contact area increases, and the
contact resistance, R, decreases, as indicated by the drop commencing
at the point B. This decrease of R is the sojtening drop. There is no
§ 20. Resistance.voltage characteristics of clean symmetric contacts 89
sharp voltage limit, but rather a narrow range of the voltage that
corresponds to the temperature range ofrecrystallization. Wespeak of
this as the softening voltage . For copper, the softening voltage is about
0.12 V, with 8 = 180° as is indicated in Fig. (20.01).
After the softening drop, the characteristic rises again between D
and E. However, the slope is very seldom ideal, since some further
softening seems to be effective. In order to obtain the right resistance
it is also necessary to use fairly short current transients since creep
may cause resistance drop. Only with transients of about 2 seconds
was it possible to obtain the steep branch of curve II in Fig. (20.02)
that corresponds to D-Ein Fig. (20.01). For the soft gold in the same
figure , the transients obviously have not been short enough.
Notice that the curves of Fig. (20.02) and also those of Fig. (20.03)
were obtained with small loads that would produce only elastic de.
formation in a contact
afJS
with smooth surfaces. The Ohm
fact that nevertheless the
P- __... V
A""'o.
softening drop which is
always caused by plastic
*'9 I
~/
yielding, appeared , can be tfectn?{ytkl1~ nfcki:jJhteri Ni·~ /
explained by the manner
of preparing the rods. To P-tg I -o.
guarentee clean contacts,
the members were scraped
pvt't"j'g oltl
~ ~
with a clean hard tool and ~-

v
therefore bad small pro. aara?f aot tzf as
Contuct wltflg'e V-
tuberances. These yielded
Fig. (20.02). R U·characteristics of nicke! and gold con·
plastically 1 at the soften- tacts measured with current transients of about 2 sec.
ing point. E .HOLM showed The branches of the curves marked by an arrow pointing
to the left, are so·called reversible characteristics. Their
that with highly polished slope corresponds weil to calculated curves in Fig. (20.01)
nickel rods (lapped with
finest diamond paste) and loads of the order of 0.01 N the softening
dropwas small but not avoided 2 •
D. Melting. The melting temperature forms a doonite upper limit
for the temperature in solid contacts. We call the corresponding volt-
age the melting voltage. Every attempt to increase the contact voltage
beyond this limit results in the contact sinking together. This forms
a greater contact area, capable of carrying the current in the resulting
solidified state. The greater area is associated with a smaller contact
resistance and the effect appears as a melting drop in the R U-charac-
1 Cf. § 7, the discussion of Fig. (7.15a).

2 Cf. Fig. 5a in R. HOLM and E. HoLM [5], and Fig. (27.14).
teristic. In Fig. (20.01) the melting drop E-F appears at U = 0.43 V,

the melting voltage of copper.
After having reached F , we decrease the current and obtain the
branch F--G, parallel to C-B-A. This branch can be followed in both
directions and is thus a reversible characteristic having the ideal form .
The condition for reversibility is a constant contact area. But, if the
contact area conforms with the hardness or the elasticity of the metal,
it will not remain constant during the variation of the temperature or
~a~~-+---+--~--r---+--+--~~~--1
~
·~a~~~~~---h~---+--,_~~~r-_,
'
Fig. (20.03). R U-characteristics of a nicke! contact measured in vacuum, 10-• Torr. The upper
setwas measured with the temperature 220 •c in the bulk of the contact members; consequently
the slope of the reversible curve is comparatively small
the current since the hardness changes with the temperature. Actually,
the contact has been welded during stage F, and for this reason the
area remains cons.tant during the current variation. It is even possible
to obtain fairly good reversible characteristics when starting from a
contact temperature far below the melting point, for instance the dashed
curve D-H of Fig. (20.01). This is explained by the cold weld which
appears in every clean metallic contact; see § 29. Thus the contact of
the dashed line may also be regarded as welded, although it has not
been subjected to a temperature higher than the softening tempera-
ture. Such contacts are very sensitive to agitation because of the small-
ness of their load carrying areas. Extremely delicate experimenting is
necessary torecordreversible ideal characteristics.
§ 20. Resistance-voltage characteristics of clean symmetric contacts 91
During these tests the Ioad shall be constant. This requires that at
least one contact member is movable, so that it can approach the other
member during softening or melting in the contact. The current must
also be maintained1 for at least 1 ms in order to produce equilibrium.
E. Measurements of RU-characteristics. Figs. (20.02) and (20.03)
show measured R U-characteristics with reversible curves indicated
either by arrows to the left or in both directions. Every nieasurement
was made with a current transient of about 2 seconds. In that short
time creep in the nicke! was ineffective. Consequently, even a new
contact has a fairly ideal characteristic. Gold creeps more rapidly and
therefore shows initial and reversible characteristics with very different
slopes.
The lowcr curves of :fig. (20.03) illustrate reversible characteristics
commencing at different voltages. 'l'his does not appear to greatly in-
fluence their slope.
In Fig. (20.03), characteristics calculated according to (16.31) have
also been 'plotted and it is obvious that they are not quite parallel to
the measured curves. Very likely the reason for this is that (!, Ä., a. and
pwere not determined for the nicke! used. However, the variation ofa
with {} was taken into account.
In order to show an optimum agreement between measurement and
calculation, when (!, Ä., a. and p are exactly determined for the metal
used, we cite Table (20.04) taken from R. HoLM and STÖRMER [6]. The
calculations are made with Eq. (16.31) which enables the individual
variations of (! and Ä. to be considered.
Table (20.04). Calculated and obaerved resislancea cm platinum contacta2

Contact valtage U V 0.122 0.200 02745 0.348 0.42 Remarks
Supertemperature (/J 200° 400° 600° 800° 1000°
R(ei.)
calculated 1.415 1.84 2.233 2,6 2.945
R(eo~)
R(eA> observed on a
}·~
1.42 1.83 2.22 2.62 2.96
R(eo~) welded contact
observedonanot per- reversible
R(eA)
ceivably adhering 1.42 1.83 2.24
R(eo~) contact
Softening and melting voltages for commercial metals are given

in Table (X,2).
1 Cf. § 21, Eti· (21.02) and Fig. (21.04).

2 HoLM, R. and STÖRMER [6], Table (3b), also STÖRMER [1].
The R U -characteristic of a symmetric carbon contact. R. HoLM and

STÖRMER [12] used a carbon grade with properlies as described by the
manufacturer as follows:
(! = 9 -to-s (1 - 0.0003 #) n m and Ä. = 2.0 (1 + 0.0005 #) W/m °K
Allowance for the individual variations of (! and Ä. was made possible
by using Eq. (16.31) for the calculation of R ((!Ä.)/R (e0 ~).
Table (20.05). Symmetrie carbon eontaets
According to tl>= 200° 700° 1000°
Observation U= 0.62 1.12 1.37

Eq. (16.29) U= 0.54 1.04 1.26
Observation R((IÄ) 0.97 0.89 0.87

R((loÄo) ±0.01 ±0.01 ±0.01
Eq. (16.31) R((IÄ)
0.96 0.89 0.85
R(eoÄul
§ 21. Development of the temperature in a metallic1

current constriction
A. Introduction. We shall consider contacts that are heated either
by the current which they transport or by heat developed in the con-
tact surface by friction for instance. In the case of heat generated by
friction, we anticipate certain problems that, really, belong to Part III.
This has the purpose to achieve a certain unity in the exposition.
Equilibrium states have been treated in the foregoing chapters, but
in many problems, for instance those concerning moving contacts,
one deals with .contacts of very short transient duration. Under this
condition, the temperature does not reach equilibrium in the constric-
tion. In order to obtain a clear conception oftbis behavior, it is impor-
tant to know how the temperature is distributed in the constriction
of the a-spots at various stages of heating.
We shall pursue a formulation of the solutions in a simple graph-
ical representation, even at some expense of accuracy. Fortunately,
no great accuracy is required since the observation of contact area, time
of heating etc. are affl.icted with considerable uncertainty due to the
difficulty to perform the measurements.
1 In a semiconducting member, the space charge of the barrier readjusts itseH
by means of diffusion. This may require more time than heating the constriction
in the member.
§ 21. Development of the temperature in a metallic current constriction 93
B. Survey of the problems. We assume that the heating begins at

the time t = 0 and continues either at constant rate, or by virtue of
constant applied voltage. One problern belonging to this Section has
been treated exactly; namely, the temperature development in a cir-
cular area when a constant and uniformly distributed thermal current
enters this area.
It is now appropriate to survey the different problems we will meet
in this Section and also the publications upon which the treatment
is based.
Olass I - deductions dealing with the assumption of a flat contact
surface in close agreement with actual cases.
Class II- deductions dealing with the simplifying b-model of Fig.
(1.02), in which the actual curved lines of flow are replaced by straight
lines of flow emanating from a hemispherical surface, K.
From another aspect we distinguish the following cases:
Oontact constriction heated by the current passing through it:
Oase A, symmetric contact.
Oase A,, symmetric contact, when temperatures are calculated
solely for the contact surface.
Oase B, contact between a member of moderate conductivity and a
member whose conductivity is considered as infinite.
The heat source lies in the contact surface and the process within one
member is discussed:
Oase 8, contact surface regarded as stationary during the time of
heating.
Oase M, treatment with particular consideration of the tempera-
ture distribution in and around a moving contact.
A survey of the publications which have contributed to solving
the present problems is given in Table (21.01). They all are based on
treatments of heat conducting problems in such standard books as
RIEMANN-WEBER [1] and 0ARSLAW-JAEGER [1].
The important problern of case A, involves a particular difficulty.
Because the current density is relatively small in the center of the
contact area and increases closer to the rim, cf. Eq. (4.18), the initial
heating is concentrated in the vicinity of the rim. Therefore, the iso-
therms during the initial heating do not coincide with the equipoten~
tial surfaces.
This has been demonstrated with a !arge contact surface by apply-
ing a current strong enough to cause rapid welding at the rim. The cur-
rent is then quickly interrupted. The welded annulus may be observed
after breaking the weld. Cf. HJ:LGARDT [1] and [2], and GREENWOOD
and WILLIAMSON [2], [3]. Oontinued heating with the same current Ieads
to melting over the entire contact area. This area will increase until
Table (21.01). Survey ofthe relevant literature, hereby briefly characteri?-ing the methods
and aims of the papers. The following Diagrams (21.04) to (21.06) account for the
information from all the papers
Position of the Shape of Class and case Author and year of

heat source contact surface publication
Joule heat b-model1 II; A, R. HoLM [2] 1925

throughout the b-model II; A, R. HoLM [17] 1935
constriction b-model II; A, AVRAMESCU (1] 1937 2
b-model and
circular II;A,B R. HOLM [29] 1941
circular I; A, KüHLER and ZIELASEK
[2] 1952
b-model and
circular IIA,B,S,M R. HOLM [36] 1952
Heat generated in circular I;S,M BLOK (1) 1937

contact surface reetangular I;M JAEGER (1] 19423
circular I;M JAEGER (2] 1952
circular I;S JAEGER [3] 1953
b··model II; S R. HoLM [32] 1948
circular I; S ÜOSTERKAMP (1) 1948
various shapes II; A, B.S,M R. HoLM [36] 19524
it can carry the Ioad and the current without further melting. The
heating reaches an equilibrium state characterized by a contact super-
temperature e
in accordance with Eq. (13.06), and a temperature
distribution on equipotential and isothermal surfaces in accordance
with Fig. (17.04).
During the initial heating, isotherms within the constriction(par-
ticularly those near the contact surface) deviate from equipoten-
tial surfaces and steadily change their form while in distances r, great
compared with the radins a of the contact area, any isotherm coincides
fairly well with an equipotential surface.
The initial calculations by R. HoLM for the case A, can not indicate
the particularly rapid increase of the temperature near the rim of the
contact area because they are based on the b-model where the current
density at the b-sphere is uniform and the equipotential surfaces remain
isotherms also during the initial heating period.
R. HoLM, nevertheless, used the b-model throughout the initial
calculations for all cases A, B, M and S. This means a great simpli-
1 b-model refers to the artifact in Fig. (1.02) with b = radius of the infinitely
great conducting sphere.
2 The solution is given in the form of a series without application to numerical
calculations.
3 Important for the treatment of temperature distribution around a moving
heat source (friction).
4 Gives HOLM's solutions in a handy form.
fication of the mathematics making it feasible to calculate not only

for the contact surface but also for points within the bulk of the con-
strictions. With the exception of case M, other calculations are not
available. The curves in the Diagrams (21.04), (21.05) and (21.06)
even when based on HoLM's calculations with the b-model, all are
referred to a flat contact surface. This means a conversion_ from the
sphere of the b-model to a flat contact surface and application of cor-
rections. These manipulations imply: first, a transformation of quan-
tities expressed in b, the radius of the sphere in the b-model, to funtions
of a, the radius of the flat contact area where b = 2 afn according to
Eq. (1.06); second, corrections ofthe ordinates so as to agree with values
fort= oo (i. e., equilibrium) corresponding to Eqs. (17.05) and (17.03);
third, a correction to bring the results in agreement with certain direct
calculations concerning the flat contact surface. The latter correction
requires changes of less than 5% of Off) where {} is the temperature in
the reference point.
In addition, we must consider the inaccuracy which has to be ex-
pected froni the fact that all calculations are made with constant (!
and A.. This correction is also not considerable. We know from Fig.
(17.04) that the temperature distribution at equilibrium, expressed
by fractions {}jf) of the maximum temperature e, is largely independ-
ent of the variation of resistivity (! with the temperature (similarity of
curves 1 and 2). A similar independence may be expected during the
heating period. This means that the influence of the temperature on
(! is essentially taken care of, when the finale is correctly determined.
This addition could not be considered for the diagrams. The other three
corrections have been applied.
C. Discussion of the diagrams for the different cases. The Diagrams
(21.04) to (21.06) contain the ratio {} (p,, t)jfJ plotted against time t
expressed by a dimensionless variable
Ä
z = -ea2 t (21.02)
Here {} (p,, t) is the (median)supertemperature in the equipotential

surface, characterized by p, in Eq. (4.13), and e
is the equilibrium
supertemperature (for t = oo) in the contact surface. e
has to be cal-
culated according to Eq. (13.08), i. e.,
(21.03)
where e A. isamedium value during heating up to 0 (p,, t). Here, is a e

mathematical quantity, and this temperature may never be reached.
Therefore, Eq. (13.10) can not be used, since it would imply e.A.-values at
temperatures thathave no inßuence on the transient states in question.

In the Case A below, the median value (! A. has been taken 1.2 times
greater than e A. for the bulk temperature (""' 20 °C).
The primary deductions are published in the references given in
Table (21.01) and to repeat them in extenso would exceed the scope
of this book. Instead, we have confined the exposition to a description
of important features of the methods used.
Case A. Symmetrie contact beated by the current througbout the
v-;
constriction. In Diagram (21.04) the curves for the contact surface
and its neighborhood (say for = 0, 0.5a, a, 2a) give only median
temperatures at different z, as long as z is relatively small. The iso-
v-;
therms in this region deviate considerably from the equipotential sur-
faces but very little 1 when z grows beyond 100. Curves for greater
than 2a may represent isotherms in cases when z > 10 . The case A 8
is included in Diagram (21.04).
1.0
- -
I--
v__....
A\,\1>~\\o;l).
/_,
V ?
lll'-" - r-
/ /
I zo. v
/ 3(1. I--
-
~.L
~ ""/
v_ f- --
/ / nsa. .....-
0
I/" :..-/
07 10
rr 100
z-
Flg. (21.04). Case A: Symmetrie contact heated by the current through the constrictlon
Experimental tests by FINK and KöRNER [2] confirm the rate of

heating in accordance with Diagram (21.04). They observed, by means
of oscillograms, the decrease of the resistance of a contact between
graphite rods during the heating period. Considering the variation as
caused by the increasing temperature in the constriction, it could be
shown that the average constriction temperature varied with time as
was to be expected from Diagram (21.04).
Case B. Contact heated by the current. The members have very
different conductivities. Diagram (21.05) represents a case 2 where the
contact surface constantly has {} = 0 and the final hottest section has
y-; = a. The material between this section and the contact surface
has a small volume. Thus, it has a small heat capacity and changes its
1 This statement is supported by measurements by HrLGARDT [1] .
2 One member is infinitely conducting.
temperature rapidly in the beginning. That is why, atz < 2, the curve
for v;= 0.5a runs higher than the curve for = a which is the iso- -y;
therm that finally represents the hottest section .
1.0
I t stct;on - r-
--,__
I \\0~~
..~ 0.
~~~·
-~-
...-:17 vV
-
// ~0. /
_/ / ./
ov at
_"/
/ ~,.; V
.....
V
V
~
f-
10~ -
1 10 100 1000
z-
Fig. (21.05). Case B: Asymmetrie contact heated by the current through the constriction. One
member. has moderate conductivity and the other an inftnitely great conductivity
Case S. A semi-infinite body at rest, with a circular heat source.

Diagram (21.06) concerns a semi-infinite body that receives a steady
ß.ow of heat across a circular heat source, S, begimiing at t = 0. This
may be a contact area. None or a n egligible e lectric current through
the body is assumed. All calculations have been made with the
assumption that the heat inß.ow [q wattjm 2] is constant and un-
iformly distributed over S. As is known from the electric current
1.0
I /
I tiSllrfo&e
~r;.lf>C ::::;- -
4"~
/ f-f-
/ J!..
V V
~f-
-~--"
_,_.
0
1 10 1()0 1000
z-
Fig. (21.06). Case S: Semi-inftnite body at rest wlth a clrcular heat source
problem, the area S could be a niveau-surface only if the current

distribution were of a certain non-uniform type. Thus, in the pre-
sent problem, neither S is an isotherm nor any of those ellipsoids
which are characterized by the parameter f-t in § 4. Nevertheless,
Diagram (21.06) refers to such ellipsoids defined by their height Vß.
7 Electric Con tacts. 4th Ed.
Then the temperature at a time t in the surface with the height

means only the maximum temperature at t in that ellipsoid (at its top).
v;
For this reason, one must use @ for the equilibrium maximum tempe-
rature1 in the body in accordance with Eq. (21.07)
8 = al (21.07)
If one confines the calculations to the maximum temperature in

S alone, Diagram (21.06) can be applied to an elliptical heat source.
Let S have the semi-axes o: = ay and ß = afy. Then the thermal re-
sistance of the constriction belonging to S is according to § 35 and Eq.
(4.23)
1
W = 4al f(y)
with f(y) according to Fig. (4.24); and Eq. (21.07) changes to
8 = a).q t (y) (21.08)
Diagram (21.06) starts at z ""' 0.04 where the value {) (0, t)/8 is
already considerably great. For smaller z, the following approximation
is recommended
D~t) = 1.6 (z- 2.24z (21.09)
Case (A and S) for very small z-values. Diagram (21.10) shows
{) (p,, t)/8 as a function of z for very small z-values. The lower curve
v;
is a continuation of the curve for
= 0 in Fig. (21.04). The two
upper curves are such continu-
t a~ ations concerning Fig. (21.06).
General remark. The Diagrams
Q:, aJ r--- - - t--7''--
~ (21.04), (21.05}, (21.06) and (21.10)
~ provide a handy method to solve
~~~-+--t~--r--1--~ many typical problems on the devel-
opment of the temperature in con-
tacts without the need for calcula-
tions other than these of a, 8 and
ato ms 010
z. As said above, the curves for the
z- contact surface are calculated di-
Flg. (21.1 0). Concerns cases (A and S)
rectlyfor a ßat circular area whereas
for very small z-values the curve for interior niveau-sur-
faces have been initially calculated
with the artifact of Fig. (1.02) and then geometrically corrected. They
have not been published before with their present accuracy.
1 See JAEGER [1] Eq. (31).
Case M. Beat source moving on the surface of a semi-infinite body.

This problern has been thoroughly treated by JAEGER [1] and [2),
particularly for a reetangular heat source = contact surface. Figs.
(21.11) and (21.12), based on his formulas, give the temperature distri-
bution within the contact surface along its axis in the direction of
motion. Notice the co-ordinates and the fact that the maximum tem-
08
tz11/.1!·1 i
/ '\ ".......,
\
"
coo/Z?. ·/031G "z. v
1/ __\
1/ /' \ J.1C
'\
/ I \ 02 1\ ~
\ \ 31.C
01 /
"', '\.i
1--·-..l
01
&or fronf
I ~ 0
-z -1 0 1 z
-1 0
r/tL -
1 r/« -r/ra -
Fig. (21.11 ). Temperature distributlon in the Fig. (21.12). As Flg. (21.11) but referring to a
symmetry section wlthln and around a square reetangular heat sourcc moving in the direction
heat souree moving with the veloelty 11 on of its major extension. The diagram remains
a semi-inftnlte body. The dlagramm remalns unehanged when referred to an elllptic source
unchanged when referred to a clreular souree, with y1 = 10 except that the factor 1.1 mu~t
except that the faetor 1.1 must be omltted be omltted in the ordinate which then reads
11(1") II (r, t)
in the ordlnate whieh then reads a q/ i /(y). aq/i.. The reetangle is 10 times Ionger
than wide
perature is shifted toward t.he rear end of the contact. The calculat.ion
of such figures must be made by aid of graphical int.egration and is very
time consuming. Fairly good computations of the maximum tempera-
ture in the moving heat source can be made by application of Diagram
(21.06). The principle is to compare the heating in the source with the
transient heating of a stationary contact. Due to the motion the heat
source covers successively new points of its trail, and any point is con-
tacted only a limited time, t (v), where v refers to the sliding speed v.
The average contact time with a circular source is
- :n; a
t(v) =-- (21.13)
2 "
It is plausible that t (v) plays a similar part for the maximum tem-
perature in the moving contact as does the heating timet for a station-
ary contact. Diagram (21.06) should then give {}v (0, t) with the abscissa
z(v)
Ä.-
= ca 2 t(v) = 2
1t
cav
Ä.
(21.14)
7*
The confirmation is given by Table (21.15) where Oov·values, deter-

mined in the manner just described, are compared with results obtained
from Figs. (21.11) and (21.12) that are based on JAEGER's formulas.
The function of the heat source presents a particular problem. For
instance, if all heat is produced by friction, the question arises how
much of it enters either member. The distribution of the heat between
the members is discussed in JAEGER [1]; cf. Example B below.
Diagram (21.06) can be successfully used even in the case of an
elliptical shape of the moving heat source. One would expect that a
modified z (v) had to be employed; but, as Table (21.15) again con-
firms, Diagram (21.06) together with formulas (21.08) and (21.14) yield
good solutions. Details ofthe calculation will be elucidated in SectionE.
Table (21.15). Moving hea~ Source
o.<o. t.>
cav
e
z(v) according to
2Ä
Diagram (21.06) together
with Eq. (21.14) Figs. (21.11) and (21.12)
Circular heat source

2 0.39 0.44 0.44
1 1.94 0.66 0.68
0.2 9.7 0.835 I 0.85
Elliptic heat source with y 2 = 10
31.6
3.16
0.0613
0.613
0.22
0.50
I 0.25
0.52
0.316 6.13 0.79 0.75
D. Cooling of a previously heated contact constriction. The problern

of cooling a previously heated contact is of the same importance as the
heating process. R. HoLM has proved1 that, if cooling starts from a
steady state, both the previous process ofheating and the cooling follow
the same pattern. The cooling process can be expressed by
(21.16)
where {) refers to the formulas for heating, and {)c and tc to the cooling.
Consequently, the foregoing rules can also be applied to the problern
of the cooling starting from an equilibrium state.
E. Examples: Example A. A symmetric long con!'triction with a
circular contact surface of radius a = 5 . 10- 5 m is heated by the cur-
rent that is passed through the constriction by the voltage U = 0.34 V.
Wauted is the average supertemperature, {} (p, t), in the surface char-
1 R. HoLM [30] Appendix IV and [32] p. 363, concerning the problems II,A
and II,S respectively.
acterized by VP
= 3a at the time t = 10-4 sec. The contact material
is copper with A. = 380 Wjm°K, c = heat capacity = 3.4. 10s Jjm3°K,
and e = 1.75 · 10-s Q m at room temperature. We anticipate that, be-
cause of the elevated temperature, the average resistivity during the
process is about 1.2 times higher; namely, (} = 2.1 . to-s m. n
First we use Eq. (21.02) to obtain
380 · .10-4
z = 3.4 .1Q6 -25 . 1Q-10 = 4.47
ff(p., t> .
Th e ordi nat e e- m
n·1agram (21.04) corresponding to z = 4.47 on
the curve v;
= 3a reads about 0.07. Forthis problern the number e is
e
obtained with Eq. (21.03): = U2 /8eA. = 1810°. Hence
#(p, , t)"" 0 07 ·1810 = 127°
e
Notice that = 1810° lies beyond the melting point of copper. This
is no discrepancy since the high temperature does not appear during
the time considered; it is only a mathematical quantity.
Example B. We want to find the steady, maximum supertempera-
ture within a moving heat source of uniformly distributed heat gener-
ation under the following circumstances. 1. The heat is produced by
friction within a circular spot (radius 10- 4 m) that is fixed to a senii-
infinite slider and glides with the velocity v = 2 mfsec on a stationary
ßat semi-infinite member. 2. The friction coefficient is 0.9 and the Ioad
3 N. With subindex 1 referring to the stationary member and 2 to the
slider, we assume A.1 = 60and~ = 170Wfm °K,c1 = c2 = 3.9 ·106 Jfm 3°K.
The heat is produced at the rate of 2 · 3 · 0.9 = 5.4 watt. The next
problern is to determine the fraction of the heat that enters the sta-
tionary member. This problern has been solved by JAEGER1 . The sought
fraction I is
I= (1 + ~: cf (21.17)
where C is the ordinate at to ~ I

n Al / !
z (v) = -2 -
c1 av
on Fig. (21.18); / I
cf. Eq. (21.14). One finds I
z(v) = 0.12, thus C= 0.31; the ~ l

/
fraction f = 0.53 of the heat _,.,_
V _l
I
enters the stationary member 0 I
and q = 5.4 · 0.53/n. 10-8 = 9.1 . om 01 1 10 flO
e
107 Wfm 2 ; and = 152° accord-
Zu -
Fig. (21.18). Cas a function of Z(t>) where v
is the velocity of the moving member
1 JAEGER [J) p. 216ff. z. means z( v)
ing to (21.07). Entering Fig. (21.06) with z(v) = 0.12 one reads
fJ(p, t)f@ = 0.4. Thus finally
{) = 152 . 0.4 = 61°

is the supertemperature asked for.
§ 22. Growth of films on metals used for contacts,

near room temperature
A. lntroduction. Films on contacts create an electric resistance that
can cause failure in contact applications. On the other hand, contact
films diminish both cold welding and friction and therefore are des-
irable in many cases providing they do not harm the desired electric
performance·.
One property by which to distinguish these films is composition: A,
Pre-tarnish films which on base metals can develop to tarnish, for
example, a chemisorbed monolayer of oxygen. B, Tarnish films (Ger-
man: Anlaufschicht) in which one constituent is the atom of the matrix
metal, for example, an oxide. C, Alien films, for example, a lubricant
or water film (treated in § 24).
With respect to thickness we distinguish :
D, Microfilm,one or a few atoms thick.A specialtype is the monolayer.
E, Protective or passivating films which stop growing at small thickness;
for example, the film that makes stainless steel resistant against chemi-
cal attack. F, Tarnish films which arevisible and continue to grow.
Among microfilms we distinguish:
1. Physisorbed films, fixed to the metal by VAN DER WAALs forces with
bonds of the order of 0.05 eV. They do not involve metallic valences.
2. Ohemisorbed films, fixed to the metal by aid of its free valences with
probably more or less covalent bonds ofthe order of 1 to 8 eV. A special
kind is ad-ion films, the atoms of which are ionized and thus add an
electrostatic component to the bonding.
In regard to the free valence, any clean metallic surface constitutes
a defect of the lattice due to free, unoccupied valences. It is in a
"state of high energy". From cleavage work it is estimated that peeling
consumes about 30% ofthe heat required to sublimate one atom layer.
Physisorption of alien atoms is brought about by VAN DER WAALS
forces, explained as follows. The electron cloud belonging to any atom
(or molecule) constitutes together with the nucleus a weak, rapidly
ßuctuating dipole. Any instantaneous dipole induces correlated dipoles
in surrounding a.toms and therefore attracts them. The VAN DER WAALS
§ 22. Growth of films on metals near room temperature 103
attraction between a pair of atoms has the form: constantfr7 • Here r
is the separation1 · of the two atoms. These forces correspond to
bonds of the order of 0.05 eV. The bonds depend on the site on the
lattice arrangement, but depend very little on whether or not the
lattice is metallic. This is understandable since the free valences of the
surface are not involved; neither are the free electrons of the metal
because they can not follow the rapid fl.uctuations of the atomic cloud.
EHRLIOH [ 1] and [2] describes recent informative experiments with
physisorbed noble gas atoms on refractory metals. However, concern-
ing electric contacts it is sufficient to know that physisorbed mono-
layers produce tunnel resistivities of the small order of a = 10-13 Qm2,
which is practically negligible.
Chernisorption is a more complicated phenomenon. Weshall discuss
the chemisorption of oxygen only. The process is similar with halogenes
and also with nitrogen in part. Initially, molecules may be phJbisorbed,
but in fractions of a second they are dissociated into atoms which are
chemisorbed2 • How far their binding is covalent (occupying free valen-
ces of the metal) or ionic has been investigated only slightly and shall
not be discusseda. The bond is ofthe order of 1 to 8 eV; viz., as strong
as normal covalent bonds. Atoms bonded with 4 eV can be pumped
off at 1500 °K.
Usually more than 10% of the gas molecules hitting free sites will
adhere4 at once. The rate of their impact on a surface is so great that
even at an oxygen pressure of 10-6 Torr (in not too small a vessel) a
monolayer is produced in seconds. The binding energy of the adhering
molecules seems sufficient to dissociate them since such a layer appears
chemisorbed in fractions of a second even at room temperature.
Quasi-metallie contact areas are those covered by films which add
tunnel resistances smaller than the constriction resistances.
B. Rate of adsorption dnfdt = number of molecules adsorbed per
second and cm 2 from a gas with the molecular weight M having the
pressure p Torr at T °K. Let cp eV be the activation energy (cp = 0 for
physisorption) and /(f9) = a function of the fraction e
of the areas
already covered. One obtains6
dn = 3.52. 1022 P f(f9) exp (- 11600cp/T)

iH VMT
1 See for instance SLATER [2] p. 251.
2 Chemisorbed atoms prefer special crystallographic sites on the metallic sur-
face; see GERMER and McRAE [14].
a If the film consists of negative ions the work function of the metal is increased;
cf. J. ßECKER [1).
' See for instance J. BECK.ER [1].
5 See for instance DE BoER [1] p. 7 and p. 45.
Recently the migration of chemisorbed atoms over different metallic
crystal"faces could be studied by means of :field ion microscopy; see
EHRLICH [1] and [2]. The barrier agairrst migration is of the order of
1/5 of the binding energy. Migration enables atoms to meet each other
and to recombine so thatthe particles evaporating at increased temper-
atures are molecules.
C. Development of tarnish from a ehemisorbed monolayer. The
mdde will grow when the bonds between metal and oxygen in the oxide
represent a greater energy than the preceding bonds between the oxy-
gen molecules and in the metal. The oxide develops into a regular lat-
tice except at the metal-oxide interface where the atoms and their
bondswill have a special distribution. Littleis known about the lattice
of the :first oxide layer except that it is chemically bound to the metal.
The growth usually proceeds by metal ions migrating through the oxide
and meeting chemisorbed oxygen ions at the outer oxide surface, where
the growth then proceeds. In some oxides, oxygen ions migrate in the
opposite direction. The consequence of the direction of migration will
not be discussed1 •
The bond between oxygen and gold ist too weak to produce oxide
formation, and the film does not grow beyond a chemisorbed mono-
layer. According to KlsLIUK [4] and [5], nitrogen on wolfram behaves
in the same way although the nitrogen atoms are bound to the metal
surface with nearly 4 eV (half of the strength of the wolfram-oxygen
bond); seealso EHRLICH [2] p. 350.
With respect to the growth beyond the chemisorbed monolayer
we distinguish two cases: Oase I, as long as the oxide layer is only a
few atoms thick and the temperature is low; its further growth rate dep-
ends on many circumstances, and measurements of the thickness, a,
are difficult. Oase II 2 , at high temperatures (several 100 °0) and fairly
thick films (s > 100 A), the rate of growth is essentially determined
by the diffusion of the ions, viz., it is proportional to the ion density
gradient (itself inversely proportional to s) and to the diffusion coeffi-
cient, D (itself proportional to exp [- 11600 cpf T] where cp e V is the
activation energy of diffusion). It therefore obeys the famous para-
bolic rate law of oxidation at least approximately:
d 8 = constant exp ( _ 11600 '~~) (22.01)
dt 8 T
and integrated
s2 - s~ = constant · t exp ( - -11600tp)

-T- (22.02)
1 See for instance the information in EVANS [2].
2 As for details see § 23.
where s0 is the thickness at t :::::: 0. This law has been confirmed by
many exact measurements. For information about its validity and ex-
ceptions see HAUFFE [1], also GuLBRANSEN [2] and RöNNQUIS'l' [1}
and [2].
lt is unfortunate that in the study of contacts, the interest neces-
sarily concentrates on Oase I where measurements of the film thick-
ness are particularly difficult and infl.uential variables are- numerous.
Let us disCllSS measurements on contacts pre·viously cleaned in
vacuum so that neither the electric resistance nor a diminished ability
to cold welding at low load (:::::: 10-4 N) indicates any films; these
contacts were studied during exposure to air. The first measure-
ments of this kind (R. HoLM et al. [8] p. 53 to 57) give data on devel-
oping resistances (evidently tunnel resistances) and on change in weld-
ing, from which the existence and ayerage thickness of deposited films
can be assessed. Later, a different film growth on different crystal faces
was reported, for example in GERMER et al. [11] and [12], and RHODIN
[1].
Table (22.03) summarizes typical determinations of film thickness
after exposure of the clean metal to air. The table refers to different
methods of investigation labeled a, b, c, d and e. For instance, (a2 )
refers to method a and to the second author cited for this method.
(a): Measurements of film resistance considered due to tunneling,
and computation of film thickness (3 to 10 A); R. HoLM et al. [8];
KISLIUK [2], [4] and [5]; ANGUS [1]; KLAUDY [2].
(b): Electrolytic reduction with film as cathode in a potassium chlo-
ride solution in absence of air; EvANS and MILEY [1]; CAMPBELJ, and
THOMAS [1]; TonT et al. [1].
(c): Electron diffraction and influence of film on polarization of
light; WHITE and GERMER [1], where observations ofkind (b) are criti-
cised; HASS [1].
(d): Measurements of cold welding and friction assessing the ex-
istence ofa film; R. HOLMet al. [8]; MAOHLIN and YANKEE [1]. Darn-
age of the film at contact movement is probable. Therefore, compu-
tation of the film thickness is unreliable.
(e): Microbalance technique; RuKA [l]; RHODIN1 [1].
The scattering of the data in the Cu-column 2 indicates existance of
1 RHODIN investigated the oxide growth on different crystal faces and found
a rate about 3 times larger on 100-faces than on 111-faces. Table (22.03) contains
averages from all faces.
2 Concerning copper, VERNON [1] reports that copper which was exposed to
clean air at room temperature acquired an almost transparent tarnish that was
considerably protective. However, if unfilmed copper was exposed to air contain-
ing as little as one part in 35. 1()8 of H2S the tarnish was not protective.
.....
0
~
Table (22.03). Number of chemisorbed atomic layers of oxygen1 on initially clean melallic polycrystalline surfaces after exposure to air
the time t. Within brackest are references to the Iist above. Room temperature. All data are averages
time t I Au j Ag I Cu I Pt I Ni I W j . _ T_i _l Al j_ Mo . I Steel
20sec 2.5 (e2 )

7.0 (c2 )
1.0 (b.)
1-2min 0.3 (b 3 ) 0.5 (a1 )

1.0 (b.)
4.0 (b 3 ) 3.0 (b 3) 0.5 (a 1) 1. 7 (a2 ) 2.5 (a 4 ) 2.5 (a4 )
2.0 (ct)
3.5 {e2 ) j
10min 0 (a1 )? 10 (b 3)
1 (a3 )
1-2 h 6.0 (c1 ) 2.0 {e1) 2.0 (a2 ) 2.0 (d.) 2.5 (a4 ) 3.0 (a 4 )
f
2 days 1 (a 1 ) 2.0 (a 1 )
10.0 (c1 ) 2.0 (c1 ) :to (a4 ) 3.5 (a4 )
3.5 (e2 )
weeks 10.0 (c1 )

I
1 A "fractionallayer" does not cover the total conducting area.

§ 22. Growth of ffims on metals nea.r room tempera.ture 107
uncontrolled in:O.uences, which would appear in all columns if the ob-
servations were as numerous.
The growth ofthe oxide beyond the thickness noted in Table (22.03)
is very slow. Notice that the table shows averages. There seem tobe
considerable differences on different crystal faces at the beginning of
growth at room temperature. This is similar to the results by YouNG,
CATHCART and GWATHMAY (1]. They state 1. that at 130 °0 a thick-
ness of 320 A develops on 100-faces in 200 minutes, but only 90 A on
311-faces in the same time. During continued growth, the difference
levels off. 2. The film on a single crystal face is not uniform, apparently
due to imperfections in the face.
ANGUS [1] has measured contact resistances between a hard drawn
gold wire and different noble metals at a Ioad of 0.05 g. The samples
were initially ground, scraped or fractured. During exposure to air for
periods up to two months the average resistance of platinum and iri-
dium against the probe did not change.
D. Protective films. Certain tarnish films stop growing at thicknesses
ofthe order of 10 to 100 A, protecting the metal against. further attack
by oxygen and usually against various other chemicals. Those films are
called "protective". The very thin ones are called "passivating" films.
Stainless steel and aluminum have very effective protective films. The
oxides on nicke! and copper do not really stop growing but grow slowly
and therefore serve to some extent as passivating.
The existence of a maximum thickness, 8max, of the protecting films
may be explainedl as follows. Their structure has very few imperfec-
tions, and therefore diffusion of ions through them is obstructed and
requires the help of an electrical field. A field is generated between the
negative oxygen ions on the external surfaces of the film and a corre-
sponding positive charge at the interface film-metal. The total voltage of
this field has a given value of 1 to 2 volts because the FERMI Ievel of the
ions is lifted into coincidence with the FERMilevel ofthe metal. The cor-
respondingfieldstrengthisabout (1 to2)/s Vfm. Thus the field decreases
when 8 increases and will eventually be unable to maintain diffusion
and growth when 8 reaches a certain 8max· This is about 60 A for Al 20 3 .
Applying 2 an additive field of 7 . 108 Vfm, or rising the temperature
to about 500 °0 restores the growth of the Al 20 3 film 3 • Unfortunately,
it has not been discussed, how the potential "weil" around an ion must
look4 in order to react as assumed for a field of 7 . 108 V fm, or a tem-
perature > 400 °0.
1 See CABRERA and MoTT [1]. Their ideas arenot consistent with those of Mader[1].
2According to experience with electrolytic oxidation.
a Cf.
HAUFFE [1] p. 20.
'Other objections are discussed by HAUFFE [1] p. 97ft'.
108 Sta.tiona.ry Contacts
The high activation energy for diffusion in protective films (perhaps

2.5 eV in Al20 3 , see HELDT and HAASE [1]), seems tobe a result ofthe
metal ions all ha ving the same valence higher than 1. This is not so in
copper or iron-oxides where metal ions of different valencies can appear.
The alumina film is practically insulating and aluminum electric con-
tacts can be used only when the film is ruptured. The mechanical
strength of AI:.Pa films is so considerable that they may be used as
specimen holders in the electron microscope; see HAss and KEHLER [2].
The film on stainless steel and various other kinds of steel is rela-
tively thin, 10 to 20 A, and thus penetrable for tunneling electrons;
cf. Fig. (26.08).
Valuable information about the tunnel resistivity and indirectly
about the thickness of protective films has been obtained by measure-
ments of the contact resistance between pure mercury and a solid
film-covered metal; see R. HoLM [30], KoBEL [1], and particularly
KLAUDY [1] and [2]. In such contacts, the deformation of the solid
member is perfectly negligible and the film on the meta! remains un-
damaged. It is important that the mercury is very pure because in air
a layer of contamination forms on the mercury. This has the ability
to adhere to the solid metal and when the solid member is dipped into
the mercury this layer follows it without rupturing.
By varying the mercury-immersed contact area it was found that
the protective film on the metal was essentially coherent and uniform.
Among other results of this kind of observation we mention: On meta!
surfaces of steel and molybdenum cut under mercury, no transition
resistance was measured. After exposing them to air (normal humidity)
during one minute KLAUDY observed a tunnel resistivity a :::::: 2 . 10-10
Q m 2 corresponding to a thicknessl of about 7 A. After 15 to 30 hs,
(] had increased to 10-8 n m 2 on steel and to 10-9 n m 2 on molybdenum,
both films corresponding to a thickness of about 10 A. In humid air
the further growth of the film on molybdenum is much slower than on
steel. R.HoLM tested freshly etched rods of copper, iron, stainless steel
and freshly cleaned piano chords. He found a = to-s to 10-s Q m2
corresponding to a probable film thickness of 10 to 30 A.
E. Tarnishing of some relatively poor contact metals. NiO formed
on nickel. The initialfilm of a few atom layers is considerably protec-
tive in a dry atmosphere; cf. Table (22.03). In a humid atmosphere,
oxidation is relatively rapid. Experiments 2 show ·that when nicke!
'with a thin oxide tarnish is stored in rooms, the tarnish does not grow
perceivably in wintertime at a low humidity (< 25% ). It grows notice-
ably, however, during spring and summer. Hygroscopic dust on nickel
1 Cf. Fig. (26.08). 2 CAMPBELL [5].
in humid air may promote electrolytic attack by means of local cells;
cf. § 23. The NiO-film is as strong as the metal. It scarcely ruptures at
contact make and therefore is disturbant in contacts in spite of its
slow growth. Nickel with its protective oxide films is very little attacked
by H2S.
Zinc is also covered by a protective film which usually insulates
in contacts. Zinc is as soft as gold. It has efficient slipplanes in the 001-
direction of its hexagonallattice so that metallic spots may be produced
at contact make by rupturing the film, thus producing fissures with the
possibility for a-spots to develop. Zinc has no ultimate rigidity but
steadily creeps under the force, and the load bearing spots become
larger with the consequence that the pressure diminishes on the initial
metallic spots. In addition, breathing during vibrations may readily
oxidize these spots and increase the resistance. Then the temperature
rises accelerating tarnishing. Some alloys of zinc behave better.
Bronzeisapassable contact material at low temperature. On brass
chemisorbed layers begin to develop about as on copper according to
Table (22:03); but the rate of growth continues unchanged and does
not slow down as on copper; seeVERNON [1] Fig, 12 p.141.
Wolfram is covered by an oxide film that usually is not considered
to be protective in a broad sense. To judge from observations by GuL-
BRANSEN et al. [1] the oxide seems to remain at a thickness of about
50 A at room temperature. Probably the upper layers are damaged to
a large extent when making contact with a wolfram counter contact
because new wolfram contacts can be reasonably conducting, appar-
ently by means of electrons tunneling through the few remaining müde
layers. When contact is made with a soft gold up to 1 N Ioad, rever-
sible poor conduction is measured at low currents, before the thicker
film around the tunnel conducting area on a hump is fritted. The same
experience is made with molybdenum and freshly cleaned steel contacts. 1
Rusting, the fatal enemy of iron, has features other than the pro-
cesses described above; cf. § 24.
F. Tarnishing of noble metals. Gold does not oxidize in air and it
is inefficient in catalytically promoting the formation of polymerized
hydrocarbon deposits; cf. § 25. It is attacked by chlorine at 180 oc
forming water-soluble AuCI 3 • A chemisorbed monolayer of oxygen is
deposited on gold in air as on every metal but does not grow beyond
this layer. Such a layer is well penetrable for tunneling electrons and
thus is harmless in contacts.
Silver oxidizes to Ag 20 at room temperature only in presence of
ozone. Ag 20 is soft, easily removed mechanically and decomposes at
t An interesting comparison between Wand Rh contacts is given in CooPER
et al. [1].
200 °0. It seldom disturbs the performance of contacts. In a normal

atmosphere silver obtains a dark, resistive film of Ag 2S. Since silver
is an ideal contact material except for its tendency to sulfidize, the
silver sulfide has been the subject of much attention. Ag 2S is not pro-
tective. It grows steadily but slowly because the concentration of the
attacking gas usually is small. Thereby sulfidation is controlled by
the scarce rate of arrival of active sulfur atoms to the surface. A concen-
tration of H2S in the rate of 1/109 in the air is considered1 tobe harmful
for contacts. Cantamination of the atmosphere by industrial activity
considerably enhances tarnishing of silver but, air :filtered through
activated carbon powder does not attack it.
The following common experiences have been noted. Dry H2S
does not attack silver. But H2S or S02 and air does attack silver. Since
the tarnish rate at room temperature was observed tobe independent 2
of the seasons, it was concluded that even a low humidity suffices for
tarnishing. It is assumed that H2S forms an acid with H20 or S02, which
attacks the silver3 • The fact that sulfur gas sublimating from a solid
piece of sulfur at room temperature attacks silver more rapidly than
H2S in considerable concentration in the atmosphere has been known
a long time.
The resistivity of Ag 2S at room temperature lies between 105 and
1010 Q m, depending on the state of its lattice structure. This is the
resistivity range of copper oxide. Barely visible film.s with such resis-
tivities are practically insulating where they cover contact spots.
Unlike almost all known tarnish products, Ag 2S is softer than the
matrix metal. Its hardness is about halfthat of pure silver (about
1.5 · 108 Nfm 2). It may weil be that the relative softness is related to
the weak binding' between silver and sulfur, and the high mobility ofthe
positive ions, which may be responsible for two singular phenomena ..
U. B. THOMAS [2] tested the contact between a silver member with
a Ag 2S tarnish and a metallic member without tarnish. He found a pro-
nounced unilaterality. With the tarnished silver member as anode the
contact was kept conducting with voltages of the order of 0.1 V,
whereas the current in the opposite direction required up to 1 V.
E.HoLM's analysis and explanation of t.he phenomenon is described in
§ 27.
The second phenomenon was observed by EGAN a~d MENDIZZA [1]
gn a silver plate half of which was gold-plated. The specimen was held in
1 CAMPBELL [5]. 2 See CAMPBELL [5].
3 Ln.IENFELD et al. [1] assume that H2S oxidizes to H10 and S, a.nd that the la.tter
a.tta.cks silver.
' The Ag-S bonds, 0.33 eV. a.re several times (a.bout 3) wea.ker tha.n other metal-
S bonds; see liAUFFE [1] p. 311.
§ 22. Growth of visible oxide films on metals a t moderate to high temperature 111
humid air saturated with sulfur vapor. Mter an overnight exposure at

60 °Ü, Ag 2S had crept 1 mm over the gold surface. Ag 2S also moves
through pores in the gold-plating, leaving cravities in the silverl. No
such creepage appears on rhodium and palladium. lLSCHNER-GENSCH
and WAGNER [1] observed a similar effect with Agl; they assume that
in the latter case a local cell acts with silver as anode and the other
metal as cathode to which positive silver ions migrate in the adhering
water layer.
Gold-plated silver is used for micro contacts; see § 30 G. It is evi-
dent that the plating must be poreless. Otherwise, conduction disturb-
ing Ag 2S can form when silver creeps through the pores.
Palladium is not attacked by sulfur at room temperature. Alloyed
with silver in the ratio 50/50, it protects against sulfidation2 •
Thoroughly cleaned and annealed wires and pieces of platinum
give clean or quasi-metallic contacts even after being kept in air for
months. But on commercial wires HOLM 3 , and FLoM and SAVAGE [1]
found more or less insulating spots. Evidently the wires were contam-
inated by the tools used for their manufacture; cf. ÜHAIKIN [1].
G. Contact breathing. At contact make, clean metallic a-spots may
be generated where protuberances are plastically deformed and surface
films are ruptured; see § 8 B. Such spots are cold welded and therefore
have a tendency to resist opening forces; see Fig. (9.02). Diffusion of
oxygen into closed spots is as slow as oxygen diffusion into a solid metal.
However, if the contact is opened and closed again, the air has an
opportunity to deposit a physisorbed layer of oxygen on these spots.
At least a part may chemisorb and remain there, preventing a new cold
weid from forming. We say that the contact has breathed.
During short breathing of gold or silver contacts, only a physisorbed
:film can be generated. This is usually wiped away at contact make. On
aluminum, however, a strong Al 20 3 -film of about 20 A will be generated
during a few seconds breathing. Thus, aluminum contacts do not with-
stand breathing.
§ 23. Growth of visible oxide fllms on metals at moderate to

high temperature. Decomposition at still higher temperature
Visible oxide films on metals insulate contacts. They have to be
ruptured, better yet avoided by use of noble metals, or by operating
contacts of base metals in vacuum or in a non-reactive atmosphere.
1 See KEIL [9] who cites other pertinent experinlents and literature.
2 This ratio corresponds to TAMMANN's "Resistenz-Grenze"; cf. MASING (1]
p. 555.
a HoLMet al. [8) p. 55.
Details of the growth of oxide films beyond 100 A at elevated tempe-

rature is of minor interest in contact work. A short summary will
suffice.
As mentioned in § 22, the growth of the oxide requires that ions
diffuse through the already existing film. Elevated temperature ·in-
fiuences the diffusion: 1. By providing activation energy for the step-
wise progression of diffusion. 2. By generating lattice defects (as
vacancies) into which the diffusion steps progress. Domination of the
first effect leads to the parabolic law as expressed in Eq. (22.02). Oxi-
dation on copper and nickel has been extensively studied.
RöNNQUIST et al. [1] and [2] summarize measurements by many
observers on the thickness of Cu 20 and CuO films on copper as a func-
tion of the temperature (100 to 900 °C) and time. RöNNQUIST hirnself
investigated the role of oxide nuclei and other reasons for the growth.
His results show more complicated relations than are expressed in Eq.
(22.02). However, we use Eq. (23.01) as a rule of thumb. It is of the
same type as (22.02). Errors are smaller than by a factor of 2,
s2 = 202 + t. 10(14.9- 5100/T) (23.01)
with s in A, t in hours and T in °K. The exponent corresponds to an

activation energy of q; = 1 eV. The preponderant oxide developed in
films below 400 oc is Cu 20; at higher temperatures CuO is formed.
The growth of NiO on nickel at 200 to 600 °C is expressed with Eq.
(23.02). It is adapted to reasonably conform with data presented in
Table (23.08) of R. HoLM [37],
s2 = 52 + t ·10(13.3- 6000/T) (23.02)
Here the activation energy is 1.2 eV (literature gives values between

0.88 and 1.5 eV.)
The so-called noble metals, except gold, lose their resistance against
oxidation at high temperature. Oxidefilms will develop. However, these
Table (23.03)
Metal temperature interval in

which oxidt>a are stable
Pd 400 to 870 °C
Rh 600 to 1200 oc
Pt 900 to 1200 oc
oxides are stable only within relatively narrow temperature limits; see
Table (23.03) which is based on KEIL [8]. As to palladium see also
ÜHASTON [1].
Certain oxides are volatile at the higher limit.
§ 23. Growth of visible oxide films on metals at moderate to high temperature 113
At the high temperature developed by the electric arc in switches,

oxidation has to compete with the decomposition which prevails above
certain temperature limits; for example as given above for platinum
10.
olm
702
~
~
~ 70°
~
~10
!:>.
.g 7tr'
7frG
0 soo
Fig. (23.04). Oxygen equilibrium pressure plotted against temperature for the following processes
2 Ag 2 0 +=:± 4Ag + Oz 2 Cu,O +=:± 4 Cu +02
4 CuO +=:± 2 Cu,O +01 2 NiO +=± 2 Ni +0 2
metals in normal atmosphere. Taking into account both the temper-
ature and the oxygen pressure, Fig. (23 .04) gives useful information.
The calculations were made using Eq. (23 .05)
F = - 4.58 T log p (23.05)
where F = free energy in calfmole, p = oxygen pressure in atmospheres

and T = temperaturein °K. On the curves plotted,. any point (0 0, p)
defines circumstances at which the oxidation and dissociation of the
oxide balance each other and the thickness of the film remains con-
stant. The curves have been calculated using free energy data given
in CouGHLIN [1] except for the lower part of the curve for CuO which
is due to HoNJO [1]. In the region to the left of each curve, oxidation
is promoted; to the right the existing oxide is decomposed. For example
in normal atmosphere with 0.2 atm of 0 2 , Ag 20 begins to dissociate at
180 °0. At higher temperature it rapidly gives off the oxygen and no
Ionger disturbs the contact. Ag 2S begins to decompose at about 360 °0.
Copper oxides exhibit a more complicated behavior. In normal
atmosphere CuO is transformed to Cu 2 0 above 1100 °0; above 2000 °0
the oxides dissociate altogether. This explains the surprisingly clean
appearance of the arc track on copper. On contacts of switches with
magnetic blowout, the central highly heated portion of the track will
often be relatively free from oxide, whereas on both sides thick oxide
layers of variou8 colors are generated.
Wolfram behaves in a special way. It is oxidized to W03 in the arc.
This substance, a pale yellow-green powder, is known to sublimate1 at
1 GMELINS Handbuch der anorganischen Chemie 1935, article on wolfram,
p. 122.
about 1700 °K. It deposits around the arc spot, whereas the arc spot.
itself is covered only by a thin porous relatively easily fritted dark
oxide. The dark oxide results also from oxidation of the bot bridges
formed between the members in separating contacts. Wolframcarbide
(WC) behaves similarly, perhaps more favorably as regards conduction.
It is of importance in contacts whether or not the generated oxide
is a powder and consequently can be easily removed; e. g., the oxide
of very pure iron appears in contacts as a powder under circumstances
where a solid oxide would develop on commercially pure iron.
The degree to which the oxidation of metals is disturbant for contact
applications does not depend solely on the proneness of the meta! to
oxidize, but also on how easily the oxide is mechanically damaged in
the contact.
§ 24. Water film, local cells and rusting
A. Thickness of water films. The fact that water is adsorbed on

solid surfaces that are exposed to humid air has been known more than
100 years. The effect of the adsorbed water on friction and wear is
significant1 , as will be discussed in §§ 41 and 44. lt is, however, not
yet clari:fied whether water adheres preferably at the outside of che-
misorbed oxygen or directly on clean metallic surfaces2 •
The adsorption of water on clean metallic surfaces has been investi-
gated by BownEN and THROSSELL [5] employing two methods. They
measured, on a microbalance, the increase in weight of a metal foil in
a vessel when water vapor was admitted to the vessel. Prior, the foil
bad been heated by electron bombardment and then allowed to cool
in vaccum. Gonfirmatory experiments were carried out using the ellip-
tical polarization of light for measuring the thickness of the film ad-
sorbed. Making allowance for the difference between true (rough) and
apparent surface they found no evidence of a film formation of more
than two monolayers of water on preheated platinum and gold foils,
even at 90% saturation. A platinum foil, cleaned but not preheated,
became covered with a water film of roughly (Dh) molecular layers
where {} is ambient temperature in °0 and h is the relative humidity
(h ~ 1 means saturation), in good agreement with ?ata given by
STRÖMBERG [1]. AB a matter of fact, all earlier investigations seem
to have been made on somewhat contaminated surfaces. This corre-
1 Ct. the interes•ing paper by LANDSBERG [1], whose ideas were far in advance
of his time.
2 About the extension of the attraction between surfaces, see A. BAIT.EY [2].
§ 24. Water films, local cells and rusting 115
sponds to the circumstances in practical devices; we cite observations

by RoHMANN [1] and [2] which may be regarded as typicall.
RoHMANN used an electrical method, employing a device that per-
mitted him to measure as small as 10 A a change in the distance be-
tween the two electrodes. The electrodes were fixed on two prongs of
a solid piece of steel. One of the prongs could be slightly bent by an
applied force. On moving the electrodes towards each other, lie noticed
the first contact between the water films by means of an indicating
current. According to his observations and some reproductions of such
experiments by R. HoLMet al. 2 , the thickness ofthe water film depends
on the metaland the humidity of the air. On platinum and silver, the
film often appears to be less than 50 A thick. On base metals films of
100 A and more in thickness have been measured.
On separating the contact the water film does not follow the moving
electrode by forming a bridge. It ruptures at about the same distance
where the first conductance was observed on approaching the elec-
trodes. Thus, these thin water films have a certain rigidity contrary to
thick water films. Of course, they are not strong enough to stand the
contact pressure. The upper layers certainly are pressed away at con-
tact make. One would expect that a monolayer remains where the
metal is only elastically deformed, but experimental data are missing.
It is inherent to RoHMANN's method that he measures the film thickness
at points where it has attained a maximum. He does not determine
an average thickness as BowDEN does.
B. Rusting by means of electrochemical attack 3• Local cells. Weshall
find (§ 31) that a passivating film tagether with a thin outer layer of
grease is capable of keeping, for a long time, the surface of iron and
steel in such a state as is required for quasimetallic contacts. This is
not true if iron carries a water film that is not too thin.
Imagine water in contact with a metal. The water molecules are
strong permanent dipoles 4 (this is the reason for the high dielectric
constant of the water) which disturb the binding field of molecules
tonehing them. This is why molecules are dissociated in water (ARRHE-
NIUS). A meta! crystal is a great molecule and water on its surface pro-
duces dissociation, i. e., looses metal ions which form a charged layer
in the water at the metal surface. The ion layer assumes such a den-
sity that the valtage between it and the meta! reaches a characteristic
value. When two metals with unequal such voltages dip into the
water, a current is generated. Its strength depends on resistances and
1 Other investigations of an early date are cited in BowDEN and THROSSEL [5]
2 Cf. R. HoLM [J) §§ 15 and 16; [4] § 13; [7].
3 Cf. EvANS, Corrosion Handbook p. 5, also Evans [2].
4 See for instance v. HIPPEL [J] p. 152.
8*
on the delivery of agents for the chemical process involved; for example
upon the delivery of oxygen as a cathodic stimulator.
It is not necessary that these electrodes be of different metals. Pure
metallic spots on an iron surface can serve as anodes with other spots
covered with oxide or containing exceptionally large amounts of carbon
serving as cathodes. A water film must be present for carrying the cur-
rent between the different spots. Such a current-producing combination
is called a local cell. The current in the water is carried by positive iron
ions starting from the anode and by negative OH ions running in the
opposite direction. If many of them were to meet at the anode, thereby
forming a protecting film, all spots would soon become cathodic and
the current would cease. However, it happens that the iron and OH
ions meet between the electrodes and combine to form a hydroxide
with a spongy structure which is deposited as rust. The anodic spots
remain clean and continue to loose ions i. e., to corrode, and the hygro-
scopic rust aids the maintenance of the water film, the electrolyte.
This process works much more rapidly than the growth ofthe oxide
film described in § 22. That is why rusting is so detrimental.
The action of local cells is dependent on the presence of a water film
that is sufficiently thick and continuous to conduct electricity. It has
been found that rusting does not proceed at relative humidities1 below
70 to 80% because the waterfilm then is either too thin or not conti-
nuous. It is known that hygroscopic dust particles promote the for-
mation of local cells.
Sulphur dioxide in the air is a stimulator of rusting activity that
can increase its rate by powers often; seeVERNON [3], Fig. 6.
§ 25. Alien solidifi.ed films on contacts

Films of water and liquid lubricants influence contact resistance
only slightly because their upper layers are squeezed away at contact
make and the remaining monofilm is penetrable for tunneling electrons.
The remaining film may obtain certain qualities of a solid material if
it is intimately connected to the basic metal. In this paragraph we are
concerned with solid deposits that are thicker than 10 A. They increase
the resistance and even lead to contact failures. We are particularly
interested in deposits which originate from organic vapors that are
transformed into amorphous solids of high molecular weight; they accu-
mulate to become insulating particles around and on the contact area
specially in case of slide. Our knowledge on this field is essentially due
1 See EvANs [3].
§ 25. Alien solidified films on contacts 117
to a decade of thorough sturlies in Bell Telephone Laboratories, parti-

cularly by GERMER et al. [1], [2], [3] and by HERM.ANCE and EGAN [2].
In order to exclude contaminations from the surroundings and in-
hibit sulfur attack, relays with Pd-contacts were used and enclosed
in sealed glass chambers. Nevertheless, a dark-brown organic deposit
formed around the switching surface between palladium contacts caus-
ing contact failure. HERM.ANCE an~ EGAN [1] discovered its origin and
revealed its character with a replica technique consisting of pressing the
warmed contact into a thin piece of a clear thermoplastic material such
as Vinylite. When cooled, the resulting imprint accurately reproduces
surface details. Foreign deposits adhere to the Vinylite and are lifted
off in their original configuration, enabling microchemical study. Exa-
mination of the imprint film in the electron microscope1 permits easy
distinction between organic and inorganic matter. Palladium particles
are opaque to electrons and organic matter is translucent.
The dark-brown and insulating powder formed on palladium con-
tacts (with and without current) when ruhhing accompanied contact
closure (0.17 mm stroke at P = 0.3 N). Quantities sufficient for study
were obtained after 105 operations. The substance was found to be
organic and to originate from vapors of organic insulators used in the
relay structure. It is essentiaHy soluble in strong organic bases as pyri-
dine and combustible at read heat. Since it is solid, it must be poly-
merous. The authors proposed to term it "frictional polymer" since it
was not observed at contact make without rubbing.
Not only Pd but also Pt, Ru, Mo, Ta, and Cr serve as catalysts for
the brown deposit. Gold is verylittle effective and Ag not at all. Nickel,
Cu, Fe, W behave similar to Ag.
The property of polymerization as described above is quite general
among hydrocarbons. Only the lower members of the methane group
do not polymerize. Acetylene and the aromatic group, including the
much tested benzene, are most efficient. A variety of structural insu-
lators were found to develop polymerizing vapor, see HERMANCE et al.
[l]p.32.
The polymer influences the mechanical wear during rubbing, which
in one case was 30 times smaller with the polymer than without it (va-
pors avoided). The yield of deposit was proportional to f"P, i. e., pro-
portional to the width of the track where P = mechanicalload on the
contact. In the case of benzene, the yield corresponded to about one
monolayer on the Pd-track per wipe.
If the contact is arcing, the deposit may be partly burnt away and
partly transformed by carbonization. The carbonized material supports
the ignition of arcs at closure, see § 49 B.
1 A modified method is dl'l!cribed by PF:rsTEREB et al. [1].
DIETRICH et al. [6], [7], using C14 as a tracer of the amount of depo-
sit, confirm essentially the results of HERMANCE and EGAN, although
their contacts supposedly did not rub. They found that alloying Pd
with 55% Ag diminshes the deposit by 90% .
Further investigations on " frictional polymer" formation have been
carried out particularly by ÜHAIKIN [1], [2], [3]. He states that the
formation also proceeds without friction, but very slowly. He found
Tetraethyl-lead (TEL) mixed into the benzene to be a promising in-
hibitor of the process. Cf. § 32. An informative survey of relay contam-
inants is given in LISH [1].
Recently, CHAIKIN claimed1 the use of vapors from compounds of
iodine and particularly of lower hydrocarbyl iodides as inhibitors of fric-
tional polymerization.
§ 26. Tunnel effect

A. Introduction. Because oftheir quality of being DE ßROGLIE waves,
conduction electrons can penetrate thin contact films which would be
insulating accot:ding to classical physics. This process of penetrating
potential barriers is called the " tunnel effect".
In §§ 6, 8, 16 and 22, it has been stated that physisorbed and pas-
sivating films on metals are sufficiently thin to be penetrable by the
tunnel effect. This assertion is based on calculations and observations
that will be described in the present chapter.
Both the tests and calculations are afflicted with considerable un-
certainties. The tunnel effect is extremely sensitive to the width 8
( = thickness of the film) and height of the potential barrier to be pene-
trated. These are quantities that never are exactly known , and the
calculations are made with simplifying
E.l ---0\--~
assumptions. This situation leaves its
marking on the following treatment.
__I Fermi I~Yel We are concerned particularly with
Plofet -s-=- flolet the current density of the tunneling elec-
Flg. (26.01). Potentlai hlll in the trons. To begin wit h, we elucidat e funda-
narrow gap (wldth B) between t wo mental aspects of the theory with refer-
plates of the same meta!
ence to Figs. (26.01) and (26.04). Imagine
two pieces of the same metal with their perfectly plane surfaces a
.small distance s apart. The interspace can be vacuum or a film.
Fig. (26.01) illustrates potentials within and between the metal
pieces. The FERMI Ievels coincide since no voltage is applied. The bound-
ary barriers at the surface of each metal extend into the interspace
1 United Sta.tes Patent 3 238 352 of March 1, 1966.
§ 26. Tunnel effect 119
and superpose. They form a potential hill whose height would be equal
to the work function (/> for a great width s, but is lower than (/> when s
is sosmallthat the barriers interfere to a finite amount with each other.
The height of the hill in any distance x from the left body is cp. We im-
agine an electron (mass m) moving to the right in metal1 and approach-
ing the barrier with the velocity component vx. We attempt to visual-
ize what will happen to it when it reaches the boundary. We denote
the component energy ~ mv! in mks units by eEx, where e coulomb
is the charge of an electron and Ex is measured in volts,
(26.02)
If Ex is larger than the height of the potential hill the electron will
jump over the hill and enter metal 2: thermionic emission. If Ex is
lower than the crest of the potential hill, classical physics precludes
finding the· electron in the interspace and predicts reflection with
certainty. The quantum theory informs us otherwise. The SCHRÖDINGER
function, 'lfJ, has finite values in the interspace even though the hill is
higher than Ex. This tells that there is a certain probability for the
electron to penetrate into the hill. Within the hill "P decreases with
approach to metal2 and finally, one gets the probability D (Ex) for find-
ing the electron in metal 2. D(Ea) is called the transmission coetficient,
and a well known approximate expression1 for it is
J
8
D(Ex) = exp [- \n lf2m[cp(x)- Ex]dx] (26.03)

0
Eq. (26.03) is based on the approximate BRILLOUIN-WENTZEL-

KRAMER solution of the ScHRÖDINGER equation and is further simpli-
fied by replacing a factor by the number 1 which it approaches for
Ex~ ip where ip is the average height of the hill.
When the electron from metal1, with Ex lower than the hill, appears
in metal2, one says that it has tunneled through the potential hill. This
expresses that the trespass was not the result of a jump over the hill
and was not accompanied by any change of the total energy of the
electron. The conservation of the energy is expressed in the solution
ofthe ScHRÖDINGER equation by the fact that its DE BROGLIE frequency
remains constant during the tunneling.
In the case of Fig. (26.01) electrons tunnel just as often to the
right as they do to the left and no observable tunnel current results.
1 See for instance SoMMERFELD and BETHE (1] §§ 19 t{) 21, or RoJANSKI [1]
§§ 34 to 38.
But now imagine metal 2 being V volt below1 metal 1 as is indicated

in Fig. (26.04). An observable tunnel current of electrons then flows
from metal 1 to metal 2. The figure shows the FERMI Ievels (1!'-L) of
the metals (now differing by V volt) and the potential hill in the
interspace. To the left and right are indicated the distribution of the
electrons on energy Ievels E. The energy scales are directed upward
and the y-scale gives the density of the electrons; to the left for meta I
1 and to the right for metal 2.
The shaded area marks where occupied Ievels in metal 1 are oppo-
site to empty Ievels in metal 2. In other words, a tunnel current from
this area is not compensated by tunneling clectrons from metal 2.
It constitutes the observable current. Clearly, the tunnel current is
proportional to the product of the available nurober of electrons and
the frequency with which any of them approaches the boundary. For
the calculation of the frequency, we refer to SoMMERFELD and BETHE
[1], Ziffer 21.
It is to be noticed that, for the calculation, the energy Ievels of
Fig. (26.04) are used in two senses. In the distribution curve, any Ievel
[ E refers to the total energy.
But we shall also refer the
figure to Ex levels2, and an
electron can belong to a very
low Ex Ievel though its total
energy E may be great be-
P/ofe 1 cause of v11 and vz being large.
y 11 The reference to Ex Ievels is
made in order to show the
Fig. (26.04). Potential hill in the gap (width s) bet- depth of Ex under the height
ween two plates of the same metal, the right plate
being on a V volt lower Ievel than the left one. The q; of the potential hill at a
distribution of the electrons on energy Ievels is lndi- pertinent position x; and we
cated by the nurober 11 on the respective Ievel E
have seen that this depth
(q; - Ex) appears in the transmission coefficient D(Ex) of Eq. (26.03).
Fig. (26.05) is drawn for the case that V is large and the tunneiing
from metal 1 proceeds above the vaccum-leveP of meta! 2, moving
the electrons into the space. This tunneling is called field emission of
electrons.
Weshall distinguish three types oftunneling and two classes of calcu-
lation :
Type I is characterized by small V-values, say V~ 0. Here the
tunnel effect is ohmic; i. e., the current density, J, is proportional to V.
1 Notice that we countnegative voltages upwards.
2 Cf. for inst. R. HoLM [35].
3 The vacuum Ievel is higher than the FERMI Ievel by f/i, the work function.
Type III is "field emission" with J strongly dependent on the field;
i. e., on V, namely according to Eq. (26.12).
Type II belongs to intermediate V-values with J rising more rapidly
than V and at great V approaching the field emission.
The convenient concept of tunnel resistivityl, a = VJJ in Q m2, is
used throughout this book. The quantity a is a function of the gap
width 8, the work functions
<P1 and <1>2 for electron emis-
sion from metals 1 and 2 into -~
the gap material, the applied
voltage V, and the relative per-
mittivity c, of the material in
the gap. When </J1 = <P2 = ifJ, we
write a = a (8, <P, c,, V) and for
the field emission (ca.Iculated f-L
forc, = 1) a = a1(1J,X) , wbere
Fig. (26.05). The potential hill when the difference
X is the field strength. The between the FERMI Ievels, V, is !arger than 'I + <1>.
theory shows thata is strongly As t.o 'I see Eqs. (11,5) and (11,6) and Fig. (11,7)
dependent on s, <!> and the
shape of the hill.
The clas8e8 are characteriz- E
ed by a different degree of
a pproximation. In clas8 A,
the potential hill is consid-
ered trapezoidal, as indicated
on Fig. (26.06) by the dashed
line, (in case of great 8
t ou ching the r eal hill con- y
tours) and the distribution Fig. (26.06). The potential hill apqroximated by a tra-
of electrons in metals on pezoid. The e lectron distribution corresponds toT = 0
energy Ievels is assumed to
correspond to T = 0; viz, it finishes exactly at the FERMI Ievels. This
simplification of the distribution representation is fairly good for
room temperature as only a small percentage of the tunneling elec-
trons are lifted above the FERMI Ievel. The simplified shape of the
hill provides easy integration, but introduces considerable errors of J
as a function of 8 . Nevertheless, calculations of class A are valuable
because they give a picture of how J depends on V.
Ola88 B comprises all attempts to account for the image force, that
rounds off the corners of the potential hill, and comprises also attempts
to account for the in:fiuence of the temperature, which, when high, Iifts
a considerable number of electrons to Ievels above the FERMI Iimit. This
1 About the heat corresponding to u J 2 see KOHLER effect, § 15.
has the effect that these electrons have a relatively easy tunneling
through a low hill, namely the hill above their own Ievel.
B. General results in the case of room temperature and less.
For type I, SoMMERFELD and BETHE [1] Eq. (21.18), and also
HoLM and KIRSCHSTEIN [19] and [20] approached the rounded off hill
by means of a parabola. It is evident from Fig. (26.01) that this can
be a good approximation in case of a smalls. HoLM and KIRSCHSTEIN
used a differently defined parabola that better approximated the po-
tential hill. This resulted iri a correction of J by about a power of 10.
In case of a great 8 the hill assumes a skew shape as in Fig. (26.04).
Evidently this cannot be approximated by a parabola. Calculations
which have been made with parabolas make a too high when 8 is great.
Calculations of type II were first carried out by R. HoLM [35] and
later repeated and extended by SIMMONS [1] and [2] with deletion of
a certain term that HoLM had introduced in order to achieve consistency
with the formula for V ---+ 0 by SoMMERFELD and BETHE. This means
~onsistency with the deduction of Eq. (26.07) which we regard as more
reliable than any other formula presented for small 8. The respective
term becomes negligible for V> 0.5 V.
HoLM's essential calculations of type II belong to clas8 A with
er= 1, and adaptation to the imageforcewas achieved by a graphical
approximation. SIMMONS tried an algebraic method for a (8, (/J, er, V)
even for the case witb (/J1 =F (/)2 , but his formulas have deficiencies. They
do not approach a1(ifJ, X) correctly when e, = 1, andin case of small 8
they even make J negative.
Other calculations of a (8, (/J, e,., V) with comparison to experiments
have been published by FISHER and GIAEVER [1], STRATTON [1],
MEYERHOFER and OcHs [1], HARTMAN and ÜHIVIAN [3], and HART-
MAN [2]. A special category concerning the field effect will be treated
at the end of this chapter.
The published experiment8 on the tunnel effect through contact
films belong to two di8tinct group8: 1. concerning films with a thick-
ness of the order of 5 to 10 A; 2. concerning films 20 to 50 A thick.
The initial observations by HoLM and MEISSNER1 belong to group 1.
So do several other studies of films that are highly penetrable for the
tunnel effect and therefore are of practical interest. Of course, it is
desirable to test the tunnel effect on films with a uniform thickness
that can be measured. It is not possible to accurately d~ this with the
very thin layers. Aluminum oxide (AlPs) films on Al have been re-
garded as fairly uniform, and the thickness of AlPs in an AI-AI 20s-AI
sandwich could be measured by determination of the capacity between
the Al-electrodes. Therefore several experiments were carried out with
1 HoLM and MEISSNER [10] and [13]; R. HoLM [9].
such sandwiches. The film thickness of group 2 is a consequence of the
fact that Al20 3 films stop to grow after reaching a thickness of about
50 A, as we learned in § 23.
Eq. (26.07) of ('lass B for a at V ---+ 0 was given by R. HoLM and
KIRSCHSTEIN1 with 8 in A, <P in e V and s,. = relative permittivity
. V .
hmy = a(8, <P, s"O) = - 2-. tABeABQcm2
V=O
t0-22
+
A2 l
with (26.07)
A = 7.32 ·10 5 ( 8 - $)
72
and B = 1.265 ·10- 6 V
<P- :~I
To cite other formulas by different investigators would exceed the
scope of this book; they are all more complicated than (26.07). It may
suffice to recommend Eq. (26.07)
for small 8, and Eqs. (48) to (50) of ~~----.---,----r--~8
QmZ
SIMMONS fl] for 8 > 15A.
On Fig. (26.08) the four upmost
curves are calculated using SIMMON's
formulas, the curves for 8 = 5 A and
the lower ones for 8 = 10A are ac-
cording to Eq. (26.07) and R. HoLM
[37] Fig. (III, 21). Intermediate
curves represent averages of differ-
ent calculations. The curves on Fig. -2
(26.08) are solid up to voltages at
which the film certainly would break -f
down. The dashed continuations of
the solid lines merely represent cal- -C
culations.
C. Details of Observations on the -8
tunnel effect at room and lower
temperature. -10
In 1928, GAMOW, GuRNEY and
CoNDON simultaneously explained -72
the a-radio activity as tunneling of
a-particles through the barrier of -100 -1'1
3 ~ V s
nuclei. The same year FoWLER anfi
Fig. (26.08). Tunnel resistivity a 11 plotted
N ORDHEIM 2 calculated the field against the voltage V across the film in a
emission as tunneling of electrons. metallic contact. Values of ~ and s are in-
dicated as parameters. Calculatlons of
In 1931, FRENKEL [1] propounded class II
1 R. HoLM and KIRSCHSTEIN [19], R. HoLlll [37] Eq. (III,19).

2 See Goon and MüLLER [1] p. 178.
the idea that the tunnel effect would appear across minute con-
tact gaps. However, a quantitative observation of this effect was first
made, in 1931 and 1932, by HoLM and MEISSNER1 and interpreted by
R. HoLM [9]. Because of the fundamental importance of these mea-
surements for the contact theory, they will be discussed to some
extent 2 • ·
Observations of the tunnel effect was not included in the initial

intentions of HoLM and MEISSNER. The problern they were pursuing
was whether or not clean metallic contacts exhibited any transition
resistance caused by the mere existence of a contact between two
members. They were also inquiring the problern of superconducting
contacts. Both problems required the employment of very low tem-
peratures.
The following idea was basic for the method of determining the
transition resistance in metallic or quasi-metallic contacts. Since the
resistance tobe measured was expected to be about the order of in-
accuracy of the measurement of a constriction resistance at room
temperature, the idea was to make the constriction resistance essen-
tially vanish by using a very low temperature. This idea proved to be
fruitfuL
All contacts were mounted in glass containers, heated and evacuated
before being sealed off, and usually filled with pure helium for better
heat exchange. The aim was to keep deposits from the airsmall or non-
existent. Evidently, at least a monolayer of oxygen always had the
opportunity to remain on the contacts. Platinum contact members could
be heated several hours at 1100 °0 until all deposit had disappeared.
Their contact showed that within small errors of observation, any
transition electric resistance could be made to disappear.
Crossed rod contacts between cylinders of 2 mm diameter were used.
The contact area, na2 , had to be calculated using either the elasticity
formula (1,2) or the plasticity formula (1,16) with coefficients valid for
the temperaturein question. These coefficients were determined for the
very materials employed at the respective temperature.
With sufficient accuracy the · contact resistance R is expressed as
(26.09)
where e is the resistivity of the metal, ef2a the constriction resistance,

and a the resistance per m 2 of the film, here called tunnel resistivity.
1 R. HoLM and MEissNER [10], [11], [13].

2 The report on these measurements in R. HoLM [37] § 26 contains additions
to the initial publications.
It will be evident from Table (26.10) that it did not matter for the con-
clusions on the tunnel effect whether Eq. (I,2) or Eq. (I.16) is used for
the calculation of a. When in the table two data are given for the same
quantity, the value on the right side refers to Eq. (I,2), the other to
Eq. (I,16) which is used for all other events.
In Table (26.10) subindices r and l refer to room and low temper-
ature. The tunnel resistivity a varies strikingly little on Au contacts.
Table (26.10). Film resistance an m2 at room temperature (subindex r) and at low
±
temperature (subindex l) 4.2 or 20.4 °K indicated by *
I Contact Contact
resistance
Constriction
resistance
Film
resistance
Contact
Ioad
p
R,
I
R, er
2ar
I ·PI
2a1
a,
na:
I nafa,
g 10-• n 10-13 n m2
Au-Au 1.15 74 88 22;20 0.32;0.17 52; 54 88 I 4.6;5.71 3.7; 8
Au-Au 11 14.2 14.4 7.0 0.08 7.2 14.3 6.1 5.8
Au-Au 35 5.8 1.8 3.94 0.045 1.86 1.75 5.0 2.3
Au-Au 400 1.44 0.73* 1.16 0.014 0.28 0.72 8.6 10.6
Cu-Cu 35 94 610 7 0.45 87 610 4.6 23
Ni-Ni 35 199 305 45;36 20 154; 163 285 54; !)2 53
It was also stated that a for gold was independent of the contact
voltage U within the limit.s 10-5 < U < 0.018 V anrl for nicke]- and
copper contacts within the limits 10-3 < U < 0.035 V.
D. Evalution of film thickness, based on measured a, for the case
of small s. For his initial calculations HoLM assumed that the thin film
merely kept the contact members apart, electrically acting as a vacuum
gap. This certainly is not far from being so concerning the work func-
tion <P. A chemisorbed film does not diminsih </J as a solid semiconduc-
tor does when contacting the meta!. It rather increases <P or does not
change it. However, it seems incorrect to put the relative permittivity
er = 1 as it is for vacuum. The thin film certainly is able to make e,.
greater than 1.
Applying Eq. (26.07) with s = 5A = the thickness of a double
monofilm, </J = 5e V, e, = 3, we find a = 4 . 10- 12 Q m 2 in good agree-
ment with the data for copper and nicke! in Table (26.10). The smaller
a-value for gold is expected because on gold the film in the contact
may not be quite uniform but covering only some portions.
Because true surfaces are uneven one could expect a strongly vary-
ing thickness of the film. Actually the surfaces in contact will have
been weil adapted to each other by virtue of the repeated number of
contact closures in nearly the same spot (crossed rod contacts); and the
film, which here is too thin and also too strongly bound to the metal for
flowing away at contact make, will define an interspace of modera-
tely uniform thickness.
It was important that the investigations by HOLM and MEISSNER
were completed by showing that the active film could be alternatively
evaporated at elevated temperaturein vacuum and redeposited. This
was done by WENT [ 1], [2] and particularly by KISLIUK [2]. They work-
ed with the hard metals molybdenum and wolfram which provide very
small contact areas and therefore dominating tunnel resistances at a
temperature as high as 78 °K. KISLIUK observed a a of a few 10-14 n m 2
after exposing a clean metal to one torr-minute oxygen at room tem-
perature; with increased exposure time a slowly increased. To remove
such a film within one minute required 1700 ° K. It is not probable
that a continuous film can generate assmall a aaswas measured. Cf. § 6.
E. Observations on tunneling at room temperature through rela-
tively thick films, 25 A and more. In 1952 I. DIETRICH(4) studied the
10
Q
I
V" ßi -
10
I
tot tJ1 I'
tJ ~'0:3
I I
1(JS
V' I
I
1()3
; I/
v;
1()2
tr' 11'
JI/ / ~ ~
11
lT /: y V !
~
I/
l•
lU
1fT'
I
'
II
0 20 ltO GO 80 1J() 1ZIJ 1'10 A 1GO
8-
Fig. (26.11). Dependence of the contact resistance n and the tnnnel restßtivity an m• on the ftlm
thickness s. Curve I ( x )with average values; curve II ( 0 ) with maximum values; curve 111 cal·
culated with resistlvity proper, r!f, of the film material; r!t "." 10' Cl m
transition from tunneling through very thin films of Ti02 (8 = 5A)

to tunneling through relatively thick films (s = 20 to 80 A). Fig. (26.H)
summarizes her results. At 8 ""' 5 A the film acts as a vacuum with a
of the order of 10-12 .Q m 2 • With increasing 8, the film eventually
achieves the ability to define a conducting band with a gap L1 e V from
the valence band (cf. § II) and also to define the work function <P for
emission from the metal into the film. Hereby <P decreases to a value of
<Pc""' L1f2. It is reasonable to assume that the films were not quite
uniform but had effective thickness of about half of those noted on
Fig. (26.11). With 8 = 20A, <Pc = 0.7 eV, er= 4, m*fm = 1, one cal-
culates1 when using SIMMON's [2] formulas the same a as given in Fig.
(26.11) for 8 = 40A. Using m*fm = 1/2 the same a is obtained with
8 = 26A.
The decrease of <Pc when 8 increases as indicated by DIETRICH's
data appears also in FISHER and GI.A.EVER [1] Fig. 7 but is not pur-
sued there 2 •
Measurements by DIETRICH, H. MEISSNER and GI.A.EVER on tunnel-
ing through thick films between superconducting members will be dis-
cussed in Section G. Certain results by GI.A.EVER stimulated the inter-
est for the tunnel effect particularly for tunneling through thick films,
the thickness ofwhich could be computed from capacity measurements.
GIAEVER [1], and FISHER and GI.A.EVER [1] investigated tunneling
through Al 20 3 sandwiched between thin Al-sheets, one of which had
been oxidized. Al 20 3 was chosen because it is known to grow to a fairly
uniform film of 30 to 100 A. Supplementary measurements were pub-
lished by MEYERHOFER and OcHs [1]3, HARTMAN and CHIVIAN [3] ;
and further theoretical work aiming at the interpretation of these
Observations by STRATTON [1] and HARTMAN [2].
Capacity measurements made to determine the film thickness gave
about 48 A, which is regarded as sufficient to give the film the character
of a semiconductor. The "forbidden" gap LI. of Al 20 3 is known to be
high, and a barrier height of at least <Pc = 2 eV is expected.
It turned out that any theory predicted high er a for 8 = 48 A and
<Pc = 2 than was measured. FISHER and GIAEVER hoped to overcome the
difficulty by assuming a small m*fm, but MEYERHOFER and OcHs
remarked that the large gap L1 indicates a small interaction between the
bands leading to an effective mass near m. However, agreement with
the theory can be restored by assuming an effective thickness of 25 A,
considering that the film certainly is not uniform, and that the tunnel
effect is much more sensitive to the thickness than the capacity is. It
seems reasonable to choose a middle course. The curve for e, = 3,
<P = 2 eV, 8 = 25 A (and m*fm = 1) on Fig. (26.08) represents calcu-
1 m* is the effective electron mass, cf. § Il,D.
2 Schmidlin [1] investigates tunneling enhanced by ion defects in the film.
He cites recent literature.
3 They also studied sandwiches with BeO as the insulator film.
lations by MEYERHOFER and ÜCHS. The same curve is obtained by

changing m*fm to 1/2 and s to 35 A, whereby the infl.uence of m* is
computed from the fact that it adds a factor Vm* fm toB in Eqs. (26.07).
That the conduction occurs by means of tunneling is proven by
the fact that a dependence on the temperature (below room tempera-
ture) is hardly observed and that the resistance is much smaller than
according to the resistivity ofthe film material; cf. Fig. (26.11)
F. Tunneling between dissimilar metal members. Fig. (26.08) shows
a V-characteristics for symmetric contacts. If the metals have differ-
ing barrier heigbts f/>1 and f/>2 (say f/>2 > f/>1 ) a rectification appears
according to calculations by SrMMONS [2]; in other words the tunnel
current is dependent on the direction of the current. In the range of
f/>1 < Ve < f/>2 a greater current fl.ows when the electrode with the
lower barrier is positively biased. For Ve ~ f/>1 the tunnel resistivity
turns out in agreement with a curve for (/> = 1/2 (1>1 + f/> 2 ) fitting
into Fig. (26.08), essentially without rectification. But at further in-
creasing V the characteristic parts into two, for a 2 and a 1 • The difference
a 2 - a 1 reaches a maximum of about a power of 10 Q m 2 at Ve = 1/2
(f/>1 + f/>2 ). It decreases with further increasing V and finally changes
its sign. Experimental confirmation exists for IJ = 25 to 30 A, see
T. E. HARTMAN [1] and SrMMONs1 .
G. The superconductivity of contacts, particulary contacts with a
film which is penetrable via tunl!el effect. Tbe investigation of the super-
conductivity of contacts was another problern HoLM and MEISSNER
were pursuing. In 1914, KAMERLINGH-ÜNNES [1] observed that a Iead-
lead contact was capable of constituting a superconducting switr:h. He
did not investigate whether or not the contact resistance implied all
the properties of superconductivity, neither did he consider the pos-
sible interference by a contact film. It is dear that a condition for
observable superconductivity in the contact is that the members are
superconducting. HOLM and MEISSNER proved that not only tin-tin and
Iead-lead but also tin-Iead contacts can possess all the properties of
superconductivity, especially the variation of the critical temperature
with the current and its magnetic field.
In addition to this it was found that a film that has a small normal
tunnel resistivity (order of magnitude 10-12 Q m 2) does not suppress
the supercond11Cting ability of the contact, but only lowers the critical
temperature 2 • Of course, it had to be proved that the supercU'l'rent
indeed traversed filmsandwas not at all or not wholly carried through
metallic spots where the initial film had been eliminated. The proof was
based on the fact that the superconductivity ceases in a magnetic field
1 SIMMONS (2] footnote 12, p. 2585.
2 Clearly, this does not mean that the film material was superconducting.
that surmounts a certain strength. This is valid also for the field that
t.he current itself generates, which is stronger the narrower the current
path.
Let us consider an example: A lead-tin contact 1 that became super-
conducting at 1.2 °K, had at 4 °K a resistance that would require a
contact radins of 6 · 10-7 m tobe explained as a constriction resistance.
According to SILSBEE's formula, this is 1/4 of that necessary for a
superconduction that endures its own magnetic field with the current
used (I = 0.59 A). Thus, the current of the superconducting Iead-tin
contact flowed at least partly across the contact film which at 4° was
responsible for most of the resistance.
Twenty years passed without anything added to the introductory
observations by HoLM and MEISSNER. Then, in 1952, DIETRICH [4]
reported superconduction in sandwiches [Ta-Ti02-Ta] and [Ta-Ce02-
Ta] with a film thickness of 15 to 40 A; and, in 1958 and 1960, HANS
MEISSNER [l] and [2] published the discovery that sandwiches [Sn-
M-Sn] (where M stands foranot superconducting metallic film of the
metals Cu, Ag, Au, Pt plated on one superconducting member) could
be superconducting with a thickness of the plating up to 3500 A for
Cu, Ag, Au, and 750 A for Pt. Even ferromagnetic metal films, if
thinner than 200 A seemed to be penetrable for the tunneling super-
conducting electrons. DIETRICH and MEISSNER report on critical cur-
rents and temperatures also.
GIAEVER [1], also together with J. C. FISHER [1], made another dis-
covery on contacts of great value for the theory of superconductivity.
The samples consisted of a thin insulator layer sandwiched between
metal sheets which were kept at a temperature where the metal was
in the superconducting state. An applied voltage could produce a
tunneling current only when it surmounted a certain voltage of the
order of m V. The explanation for this behavior is based on the fact
(previously known) that an energy gap exists around the FERMI-level
of any superconductor. The superconducting electrons can tunnel only
if they are lifted on one side so that they can flow above the gap. The
width of the energy gap and the distribution of electrons near the
FERMI-level could be determined. Devices which use the tunneling be-
tween dissimilar mcmben; and display a negative dynamic resistance
(as EsAKI's tunnel diode does) have been suggested.
These discoveries about the penetration of superconducting electrons
into different materials and over the energy gap at the FERMI-level
are of fundamental interest for the presently very flourishing theory
of superconduction but lie beyond the scope of this book 2 •
1 Cf. R. HOLM [37] p. 129.
2 Readable surveys: E. HARTMANN [2], DIETRICH [9].
9 Electric Oontacts. 4th Ed.
H. Type III, field emission. Infiuence of relatively high tempera-

tures on the tunnel eft'ect. In 1928, FowLER and NORDHEIM deduced
the fundamental equation for the field emission on the basis of SoM-
MERFELD's electron theory of metals1 • They calculated with a distri-
bution of electrons on energy levels as belonging to T = 0 i. e. neg-
lecting the small influence of the temperature up to room temperature.
Their formula, completed by considering the image force with cor-
rection given by BuRGESS, KROEMER and HousTON [1], is for the field
emission current density J 1
J 1 = 1.54 ·1014 ! exp [- 0.685 (/) xf)] A jm

2 1 ·5 2 (26.12)
with ~ = 3. 79 V: . Here X is expressed in V fA and <P, the work

function of the metal for emission into vacuum, in eV. The function
X(~) is given in Table (26.13).
lnfluence of the temperature on the tunnel etfect. At about room tem-
perature or less few electrons tunnel above the FERMI level of the
Table (26.13)
E x (E)
~
0 1.0000 10 11
z
0.05 0.9948 ~ Alm'
" \.""-
0.1 0.9817 "'-.. 10 70
0.15 0.9622 ~ r---_ ~~?(
0.2 0.9370 ....;,;..
'
0.25 0.9068 I
0.3 0.8718 ~
0.35
0.4
0.8323
0.7888 z ~ ~
0.45 0.7413 ~
0.5 0.6900
10 ~
0.55 0.6351
0.6 0.5768 &
0.65 0.5152 z ~
0.7
0.75
0.4504
0.3825 -~
0.8 0.3117 1 z J t A.V 5
0.85 0.2379 1/X-
0.9 0.1613 Fig. (26.14). Computed current denslty J veraus recipro·
0.95 0.0820 cal fteld 1/X at various temperature for tl> = 4.5 eV
1 0
emitting electrode, viz., the electrons tunnel through similar hills as

at T = 0 °K. Therefore, as mentioned above, the tunnel current varies
only slightly in this range. But when T approaches the order of 1000 oK
1 For complete history with references see GooD and MÜLLER (1].
a considera ble percentage of the electrons tunnel on high Ievels through
relatively low potential hills with increased transmission coefficient.
How the field emission increases at high temperatures according to
calculations by DoLAN and DYRE [1], is shown in Fig. (26.14) 1 •
In the range of field and temperatures represented in Fig. (26.14)
the tunnel (field) emission is greater than the thermionic emission.
Below about X= 0.1 Vf A and at T > 2000 oK the thermionic emission
is the greater one.
MuRPHY and Goon [1] published calculations ofthe entire emission
phenomenon at strong fields including field emission with its dependence
on temperature, and thermionic emission with ScHOTTRY-term. Dia-
grams expressing results of these formulas, obtained by means of an
IBM digital computer, were published by T. H. LEE [2].
Although the variation of the tunnel current across films between
metal electrodes at temperatures up to room temperature is insigni-
:ficant, it could be measured. HARTMAN and CmVIAN [3] report a
variation of up to a few per cent in agreement with calculations by
STRATTON. Another illustration of this effect is given by NARAI anrl
MIYAZARI [1], Fig. 4, p. 679.
I. Tunnel efl'ect when both electrodes are of the same semiconduct-
ing material. When the semiconductor has less than 3 · 1025 electrons
per m 3 at room temperature, the distribution is Maxwellian; cf. § IID.
Formula (26.15) has been derived 2 for semiconductors with Maxwellian
distribution. The transmission coefficient D(Ex) has been calculated
as if all electrons did belong to the same Ex Ievel; namely, the bottom
of the conducting band. Since the temperature influences the frequency
with which any electron approaches the boundary it causes the factor
[1 - exp ( -11600 VfT)]- 1 to appear in Eq. (26.15) where V = voltagc
applied across the contacta, n = nurober of electrons per m 3 and s = gap
width in A. The tunnel resistivity is
ri = ww V exp[0.683~{cpt.s- (C/>- V) t.s}] !2m2 (26.15)

2.5n 1fT [1 - exp (- 11600 V/T)] V
where zero is to be substituted for (C/>- V) 1 ' 5 when V ~ (/>. With

V< 0.5</J the last exponential factor may be simplified to
exp [ 1.025 s V(/> - ! V] (26.16)
The term exp ( -11600 VJT) is usually negligible compared with unity.
1 Fig. (26.14) is Fig. 3, p. 191 in Goon and MüLLER [1]. About the significance
of Fig. (26.14) for the theory of the arc, see §VIII and IX.
2 See BRAUN and BuscH [1], Drr..woRTH [1] and R. HoLM [35].
a Since Vacts on electrons it is numerically equal to the energy in eV that Iifts

an electron V volt.
9*
In the interspace between semiconducting members no image force

is active; but the width of the boundary regions would require a cor-
rection that has been neglected in (26.15).
Fig. (26.17) has been obtained according to Eq. (26.15) for T
= 300 °K and n = 10~ electrons per m 3 • The curves show a maximum
,r-:::
at V= 0.4 to 0.5 V, a trait that
i"-..J may be caused by the approxima-
Q~
1(/ fl)
. tions involved. Therefore, the ini-
Q· cm~
~ tialhorizontal portions of the curves
.. ""'-... r"\.
"'
~'0' have been indicated by the solid
Iines, justified since an ohrnie region
G "<. ~
11 of a according to these lines has been
'\ stated experimentally by A.BRAUN
and BuscH [1].
10
, '
i\ S•2flA
1 J. Tunnel current across a gap
.. ·· -...._J
..........
\ that surrounds a metallic contact
10 2
b-_~ 10 -t consisting of a circular spot with
......... s-1(}1. the radins a. The gapwill be narrow
enough in some environment
.JV around the contact spot as to enable
a tunnel current to flow across it.
Fig. (26.17). Due to R. HOLM [35]. Tunnel
resistivity, a, plotted agalnst the voltage This current adds to the current
across the film of an insulating material in through the conducting contact
a contact between semiconductlng members
spot. In order to calculate the addi-
tional current we must estimate the thickness of the interspace. The
estimation will always be crude, and therefore rough computations will
do .
As an example, imagine a sphere with radius r 0 contacting a plane
in a spot of radius a. Within the contact plane in the distance r from
thc center of the spot the gap is
8 ( r) = r_2_-_a_2 = (,__r_+_a=--)_,_(r_-_a_:_)
2r0 2 r0
or, since it turns out that r values to be considered differ from a only
slightly
a
s(r)""-(r-a) (26.18)
ro
According to Eq. (27.02) the voltage U(r) at r is
u (r) = u 12
:re
vr- a
a
where U is the total constriction voltage.

The tunneling current density J (r) at r can be computed from Eq.
(23) in R. HoLM [35]. With all distances expressed in A, we obtain J(r)
Vif) -
in A/A 2 not greater than 3 .w-s 8 U exp (-sVa>). The tunnel current
surrounding the spot is readily found to be
J JV:o
00 00
lt.::: 2:rr:rJ(r)dr = 1.8 .fo-• ~ exp (- s~)dr

a o
or using Eq. (26.18)
11 .::: 1.8 ·10-• - rl.s

0
a2
U J 00
ds
exp (- 2s)-
(a
(26.19)
0
Introducing z 2s the integral is transformed to [constant · erf z]

2 =
with erf z .::: 1 ; finally

rl.5
11 .::: 4.5·10-•~ U A (26.20)
a
with r0 and a in A. The constriction current is, with fl in 0 m and a in A

2a
lc = 10101! U (26.21)
Hence
I ru
.....!.. :5: 2.25 • 105 .JL n (26.22)
I. - 8 a <::"
Applying this equation to a crossrod contact with

r0 = 5 · 107 A; a = 105 A; and e = 10-7 Om
we obtain
;;~8·10• 6
showing that here the tunnel current through a gap around the contact
spot is completely negligible.
The circumstances may be morefavorable for tunneling in case of
a nominally :fl.at contact. With the reasonable assumption of humps
with the radius 1 cm = 108 A and a = 104 A, we obtain
4-:- = 0.0225
which is still negligible.

K. Tunneling within a semiconductor. With a strong field applied

in a semiconductor, its band picture becomes as shown in Fig. (26.23) •
An electron in a. at the top of the :filled band Br will have a possibility
to tunnel along 8 into the conduchng band
l'o/1 Be . The tunneling proceeds through the
potential hill ab c where the gap width LI
plays the part of a work function. By means
of such tunneling , Be would obtain intrinsic
electrons , and the top of B 1the correspond-
ing hol es; and intrinsic conduction would
' - - - - - -- - -'!:::x be generated. This so-called Zener effect is
Fig. (26.28). Tunneling within a
similar to the field emission into a v acuum,
semiconductor.The fllled and con- and the current density J is given by a for-
ducting bands (B 1an dBe respec-
tively) are inclined by a strong mula somewhat similar to Eq. (26.12) with
field so that tunnellng is possible LI substituting for if>, X represented by'Lthe
between points a and c (belonging
to the same energy Ievel) 11 their slope of the bands. As to X (~) see below.
distance issmaU enough Now Iet the line ab in Fig. (26.23) de-
fleclron note the boundary between a meta! and the
energy semiconductor, and LI be replaced by if>e.
Then, the figure illustrates tunneling from
the meta} into the Be-band of the semi-
conductor.
n A tunneling between B 1 and Be bands as
illustrated in Fig. (26.23) is a major proceas
in EsAKI's tunnel diode1 • A field of the order
of 108 voltfm appears between strongly doped
(for example by 5 . 1026 m- 3 dopinp atoms)
p- and n-regions. Fig. (26.24) shows the
band picture. The tunneling depends on
whether electrons find sites in the opposite
band into which to tunnel. The opportunity
for tunneling varies with applied voltage in
such a manner that the device has a
Fig. (26.24). Inclined bands in a negative-resistance region in the forward
(p-n)-junction, particularly re- characteristic, to which it owes its useful-
ferring to an EBAKI diode
ness in the applications.
However, the measurement of J was never madeundersuch pure
circumstances as pictured in Figs. (26.23) and (26.24). The material
was doped (heavily in the tunnel diode) and the gap was deformed so
that tunneling was easy, as if a relatively small X (~) had existed, in
some cases as low as 0.2.
1 Instructive descriptions: CoNWELL [1] and BRODY (1].

§ 27. Fritting of tarnish films 135
§ 27. Fritting of tarnish films

A. Introduction. Wiring connections are often made without pre-
viously Iiberating the contacting wires from their tarnish films, al-
though these may be visible as a color that is slightly different from the
color of a clean metal. According to the conventional experience it is
taken for granted that these films, which actually are insulators when
coherent, do not essentially disturb the conduction through the con-
tact. However, the experienced conduction is not due to a proper qua-
lity of the films but to the fact that they either become mechanically
ruptured in some spots at contact make or are electrically broken down
\vhen enough voltage is applied. The breakdown considered is called
fritting.
Fritting is a kind of electrical breakdown different from the break-
down that operates with electron avalanches. When pure fritting
appears in contact films, these are too thin for any form of avalanche.
In an avalanche an electron must move through several volts (in alkali
halides up to 10 volts) between two opportunities to ionize. Fritting
of previously insulating films has been observed at a total contact
voltage of one volt.
The interest for the physics of electric breakdown in thick samples,
where avalanches are essential, got an impetus by the investigations
ofv. HIPPEL1 Ollalkali halides, beginning in 1931, shortly after R. HOLM
had begun his study of fritting 2 • Since then the avalanche dependent
breakdown has been the subject of numerous investigations, however,
without any reference to fritting, which attracted the interest of only
a few workers on electric contacts. Still diverse theories on the ava-
lanche dependent breakdown are discussed, and recently DwYER [1]
S"!Jms up: "not only might the breakdown mechanism be complex but
it might be different in experiments which ;~.re, as far as possible, iden-
tical". The theory of fritting is due to R. HoLM and E. HoLM. In its
present formulation the theory differs from earlier conceptions, parti-
cularly by introducing a new explanation for the widening of the cur-
rent path with increasing current intensity, see below.
B. Theory of fritting. Consider a symmetric contact containing
a high-resistivity film, a few 100 A thick say 500 A. A low beginning
voltage is applied and gradually increased. Initially, because of the
high resistivity of the film, a very feeble current flows. But, when a
potential gradient of the order of 108 V Jm, i. e. a fritting valtage UF of
1 See FRANZ [1] p. 158 etc., also v. HIPPEL [1].

2 R. HoLM [4] p. 258, [7] p. 7, and R. HoLMet al. [8].
about 5 V is reached, the usually suddeu increase of the current and

decrease of the contact voltage indicate that fritting has generated
an a-spot. This spot is able to carry the current at a contact voltage
below the melting voltage but above the softening voltage of the
metal. We call this voltage the cessation valtage of the fritting and
Iabel it U h, where h refers to halt.
It is generally assumed that the first stage of any high resistivity
film breakdown is the injection of electrons into the film by a kind of
field emission or ZENER effectl. The strong field makes the boundary
barrier steeper and consequently thinner so that electrons can tunnel
through the hill. The injection will surely happcn at a preferred point
of the film, for such a film never is quite uniform. The injected elec.
trons produce a strongly enhanced current flow within a narrow
path with diverse consequences. The material within the path is
strongly heated by the current and thereby the cohesion of the film
material is diminished with the result that a channel through the
film is produced. R. HoLM assumed that molten metal was attracted
into the channel. This may sometimes happen but observations by
E. HoLM [15], ULSH [1] and ÜHOUDHURI et al. [1] show that in the
event of copper and silver members, usually only the softening tem-
perature is approached.
If the softening of the metal has an influence on the generation of
the initial a-spot, one may infer that plastic yielding of the metal helps
to widen the channel. Hereby film material is taken along with the
yielding metal which initiates at least a quasi-metallic a-spot. Such
a conception requires that the channel is formed on surface humps 2 of
the one member and that the surface finish has an influence on the
fritting. However, this influence has not yet been clearly stated. Statis-
tical investigations as illustrated by Fig. (27.07) would be neccessary
for a proof.
Yielding of humps is certainly not always responsible for a second
kind offritting, namely, widening of initially fritted a-spots. Take, for
instance, a graphite brush-collector contact where probably no plastic
yielding is involved. The sustained adaptation of the a-spots to rising
and decreasing current indicates a reversible competition between oxi-
dation and a widening agent. This continues after the contacts are run
in and the humps certainly are flattened. Adding to the mechanical
widening, an electric effect must exist. Section C is devoted to the des-
cription of what we assume this effect to be. Such an effect certainly
is involved in the beginning of the breakdown also.
The breakdown described above will be called a regular A -fritting;
1 See§ 26K.
2 Cf. CHOUDHURI et al. ( J).
see also Sections D and E. It will be shown in Section J that relati-
vely thin films of copper oxides and particularly of silver sulfides be-
have differently when the matrix meta! is the anode. Evidently, posi-
tive metal ions from the matrix meta! enter the films, there forming
conducting paths that prevent the normal A-fritting.
C. Widening of an a-spot by electrical forces. The confinement of
the current to a channel means that current flow lines are constricted
just as they areinan a-spot according to Fig. (4.16). The cross section
of the channel plays the part of the bottom of a current constriction
or a-spot. It is a common place that the current follows the potential
gradient. Its bent course is determined by charges maintained on the
surface of the film.
At the commencement of fritting, the voltage becomes smaller than
U F and is too small for the current flow to be continued unchanged
through the surrounding film. Impinging electrons are stopped there
leaving a negative charge of negative ions on the one side of the film.
Similarly, positive ions are placed on the other side of the film. These
charges force the current lines to bend into the injection path, forming
a constriction; cf. Fig. (27 .01 ). The gradient X, from the instance of
its existence, excerts a push outwards on the ions in the surface, mov-
ing them. During the continued current flow, loss and substitution of
charges balance each other.
We shall calculate this push 1 that is directed parallel to the film
surface, radially away from the channel. The calculation can bc re-
ferred to the bottom of any constric-
tion, irrespective of whether this is
an a-spot or the mouth of the ~
channel now beingconsidered. We will ~
imagine a channel with a mouth of
radius a, which also can be an a-
spot; cf. Fig. (27.01).
We calculate the average volt-
age gradient X radially along the Fig. (27.01). Gradient lines of the field
(strength X) bend through an a·spot
film at the rim of the spot. Consider surrounded by a film
a point on the film at the distance
a + /j from the center of the spot with /j « a. According to Fig. (4.16)
and Eq. (4.13) the ratio between the voltage U,. in this point (the volt-
age in the spot being zero) and the total constriction voltage U is
Uu R,. 1 lfp
u =n=-;rarctg 11
1 R. HOL.'II presented a provisory calculation in the Proc. Internat. Symposium
on Electric Contact,ij ~ov. 1961, Univ. of Maine; seealso E. HoLM [15] and E . HOLM
[14].
and
A simplified expression for fl is obtained as follows; a + Ö, the semi-

axis of the equipotential surface characterized by p, can be expressed
as Va 2 + fl according to Eq. (4.10). From (a + ())2 = a2 -+- !I we obtain
neglecting the small quantity ö2
p=2ao
Inserting this ,u in the equation for U"jU we find
C! I' = nu arct g ~a- 1121/'a6 u

l/2ao ""'n (27.02)
Hence the average voltage gradient
X= U" = 112 _!!_ V/m (27.03)

o n lfao
We may regard X as the average gradient that acts on the ions,
subjecting any one of them to a force F = X e. The calculations aim
merely at the order of magnitude. A reasonable value of ö would be
the thickness of a protecting tarnish, say = 25 A. Even = 100 A o o
would introduce no greater change than by a factor of 2. With ö = 25 A
= 2.5 . 10-s m we obtain
- 1/2
F - -- u 1.6 .1()-19 -- 1·45 ' 1o--15 -u N"/IOn (27.04)
n Va v2.5 . 1Q-9 Va
Examples: 1. With a channel width of 2a = 1000 A, and U = 0.8V
(as it well can be an instance during the fritting action) we find
F1 ::::: 5.2 · 10-12 Nfion (27.05)
2. At the rim of a metallic or quasimetallic a-spot with a = 10- 7 m

and U = 0.15 V
F2 = 7 ·10-13 Njion (27.06)
whereby the supertemperature in the spot is about 300°; cf. sec-

tion G.
To get an idea of the ability of these forces to move the ions, we
compare them with the average force on a surface atom in the event of
a specific friction force 1 of 1p = 5 . 107 N jm 2 • This is a ble to produce a
1 See Table (37.01).
§ 27. Fritting Qf tarnish films 139
small wear at room temperature. Since the nurober of atoms per m 2 is
about 1019 , the force per atom is 5 . 10-12 Nfm 2 • This is about equal to
Fi but more than F 2 • However, when F 2 is active, a high temperature
facilitates the movement of the ions.
We get the following picture of the wideni~g process. The gradient
moves ions away from the spot, mostly supported by a high temper-
ature. If one ion is removed, a neighbor atom is at once changed by
the current into an ion; soon this is pushed away and so on until suf-
ficient film is displaced.
Notice. There are powers of 10 moreneutral atoms than ions in the
respective film surface, and the force F' would not be able to move
the whole assembly of atoms and ions simultaneously.
D. Cessation of a normal A-fritting. The a-spot that results when
an A-fritting has ceased, must be able to carry the current without
being heated to melting. Consequently, the melting voltage U 1 is an
upper Iimit for the cessation voltage U11 • We do not expect this limit
to be quite reached and we never measure it in the equilibrium state.
In the presence of an easily removed film, the widening of the channel
may continue to a voltage near softening. We shall find that nickel
contacts can give cessation voltages which approach the melting volt-
age. Copper contacts have cessation voltages as low as 0.2 V.
A process as decribed above requires a finite time; cf. below. How-
ever, an increased U (overvoltage) increases the speed of widening in
proportion to U2 since both the field strength and the number of the
ions grow.
E. A· and B-fritting; definition. It is customary to speak of B-frit-
ting when the fritting starts from an existing a-spot, and constitutes
the adaptation of this spot to increasing currents through widening
of the conducting area. The widening can involve plastic yielding; but
it surely also involves the effect characterized by Eq. (27.04).
We speak of normal A-fritting when the process 'begins with injec-
tion of electrons into a path in the undamaged film. However, the
electron current alters the constitution of the film in the path pro-
ducing a "channel". In the continuation, the conduction through the
channel is adapted to the current by a widening of the channel; that
is by a B-process.
Earlier, it was supposed that an A-fritting always led to t.he melting
temperature in the final a-spot, i. e., ulo = ul at least for a moment,
and this assumption was included in the definition. As motivated above
we abandon this assumption.
When a high temperature is reached between metal members during
an A-fritting, the final conducting spot is almost metallically clean.
This can be proven by taking a reversible R U-characteristic of this
spot; i. e., decreasing and again inereasing the current below the value
it bad at the end of the A-fritting. The characteristic shows a positive,
metallic temperature coefficient cx. After B-fritting, starting from exist-
ing a-spots, R U-characteristics with a positive temperature coefficient
cx are also measured. However cx may be relatively small indicating
some contamination on the spot; see for example Fig. (27 .13). E. HoLM
had indications of films remaining in the spot. after A- and B-frittings;
cf. Section J.
F. Observations on A-fritting. A thorough study of the initial pro-
cess of fritting has to be made with a large series resistance R,. Typical
measurements 1 carried out on copper oxide films with R. :::::: 108 n are
shown in Fig. (27.07). The crossed
rod contact (each rod 3 mm dia-
meter) was formed between copper
-- 0 carrying the tarnish film (thickness
I
""""'-
8 \
~
several100 A) and soft gold. With
the copperrod as anode, Ris essen-
s \ tially constant at small voltages.
t
)\ [\ With increasing current, R may de-
crease by apower of ten until frit-
v· ) ting sets in. This is indicated by a
s C· / sudden drop of film resistance and
.f"O voltage, as is marked by the branch
C. The branch begins at a point
(1.8 V; 5 . 105 Q) when the voltage
s ! is still increasing. Sometimes one
1 succeeds in following the characte-
1 ristic t.hrough a maximum (at
~a()f o.tJt a<!l' at tu as f t s V f() a bout 3 V in the figure) and fritt-
l<'ig, (27 .07). R U -characterlstlcs of a crossrod ing appears first when the voltage
meta! contact (copper agalnst gold) witb a
film, llnaUy subjected to A·frltting. Hard· has begun to decrease with further
ness of Cu-rod was 5 · 10• NJm' and tbat increase of current; see brauch D.
of Au 1.6 · 10• Nirn'. P = 3.2 · w-a N
Up to ariy point before fritting
sets in, the characteristic is rever-
sible within 10 to 20% of the resist.ance. Certainly, the voltage drop
along the reversible A is caused by tunneling between the metal and
the conductive band in the oxide; cf. Section B, and Fig. (26.23) where
the forbidden gap width L1 has to be replaced by the work function (/Je
of Fig. (33.01). The drop can not be caused by a temperature effect,
since no high temperature appears in the undamaged film. This is dis-
cussed in§ 17, Example C.
1 R. HoLM [29] Fig. (23.01).
With the parent copper as cathode, the dashed Iine B was measured.
This is near the Iine A for which copper was anode. Thus, rectification
is insignificant in such a relatively thin film as investigated here. The
branch C in Fig. (27.07) demonstrates t.hat A-fritting starts at Ur
= 1.8 V and R = 5 · 105 with a series resist.ance R, about 108 Q. When
the process has ceased, R has dropperl t.o 5700 Q and U to about 0.3 V
~~ Uh. It. is essentially a constriction voltage. R would have been di-
minished still more in t.he ennt. of a smaller R,.
Fig. (27.08) summarizes in a statistical fashion observations of the
fritting valtage U F at many different point.s of a nickeloxide film I about
1250 A t.hick. The nicke} rod
carrying the oxidewas crossed I •
with a soft gold rod, both of fi!SOA
the same dimensions. The

observat.ions were made on I I
different occasions, when the I I
1 I
film had the age of 3 hours, I
3 days and finally 4 weeks.
I
Each curve in Fig. (27.08) )I I
represents a certain age of II I
the film and about 80 single I ~
measurements. Two discrimi- I

I
native maxima stand out, one
at 13.5 and the other at 18
to 19 V. In the young film
the maximum at 13.5 V pre-
vails. In the older film nearly
all points cluster around
18.5 V.
The influence of the film
thickness on UF was deter-
mined by the method of
measuring such accumulation
Flg. (27.08). Statistics of frit voltages, u, (Uf on
values of UFfor various thick- figure), between crossed Ni-rods. Frequency of fritt-
nesses of the NiO film. In ings ls plotted against UF. Two maxima are reglste-
red, namely, at 13.5 and 18.5 V. The former predomi·
Fig. (27.09), values of the nate• at young films, the latter at aged films
1 The thickness of this and other oxidefilmswas determined by aid of NEWTON
colors; see R. HoLMet al. [8]. This method has not always been employed with
accurate estimation of the reflecting power of nicke!, an error that was avoided in
this paper. In order to increase the accuracy the calibration was controlled with a
NiO film on glass (obtained by completely oxidizing a nicke! deposit) once ~ith
the NiO film uncovered and another time with the same NiO film covered by a
nicke! film sputtered on it. Exactly the same thickness was found without and with
the sputtered film, of course with different colors.
higher and the lower accumulation regions are plotted against film
thickness. The points arrange themselves along straight lines through
the origin, indicating that the field strength is decisive for the frit-
ting. With the upper line in Fig. (27.09) is associated a fritting field
of XF = 1.42. 108 Vfm and with the lower line a field of 1.1 . 108 V jm.
~-------.-------.------~-----,-------,
50(J f/)')7 150fl

Thi'ckness or film -
l<ig . (27.09). Values of the higher and lower accumulatlon regions of .A-fritting voltages u1 found
by a statistlcal method according to Flg. (27 .08). Ur ( U1 on ftgure) is plotted against
the film thickness
It should be noticed that the described change in the statistic from

the lower to the higher maximum of fritting voltage has nothing to do
with the growth of the film thickness during the time of storing the
samples. According to Eq. (23.02) not more than 3 molecular layers
could have been added to the existing oxide during the time in question.
Their influence on the fritting voltage ist negligible.
In Table (27.10) tests 8 to 13 refer to the statistical determination
of the higher accumulation values of Up. All the other observations
listed have not been treated statistically, and are simply averages. The
observations on Cup roughly confirm the result that the fritting re-
quires a certain field strengtb , X F , independent of the film thickness.
The strength of the fritting field that one calculates from the averages
for contacts Cu-Cup againstAu ist XF = 0.6. 108 Vfm. Wehave an
indication that this nurober is fairly low, and that damaged spots of
the film have influenced the average as was shown with soft electro-
graphite instead of gold as counter electrode. A stronger field, XF
= 108 Vfm , was then measured. We suppose that the soft graphite
'vas less able than the barder gold to darnage the film.
It is also possible that among the single measurements on NiO some
belonged to damaged spots; the darnage certainly does not affect the
accumulation voltages. It increases only the spread at lower voltages
and therefore the data ofthe tests 15 to 20 in Table (27.10) are secured.
As mentioned above, observations in several contact positions on
t.he tarnished nickel showed resistances and fritting voltages of so
unusually high values that they were omitted for the statistics as being
abnormal. Similar abnormal spots have been reported by FLOM and
SAVAGE [1], and CHAIKIN [2]. It must be considered that these spots
might have carried alien particles that separated the contact members
by about 10- 4 cm. Particles invisible to the observers would be capable
of doing this.
Table (27.10) 1 Normal film resistances R 1 (ohmic region) in the RU-characteristic,

cf Fig. (27.07), and fritting voltages U F (direct current) across !arnish films on Cu,
Ag, Ni rods generally in contact with a soft gold rod; contact load P = 0.036 F;
Au: H = 1.4·108 N/m 2• Atmospheric air pressure.
~"
"00
Load Film Fritting Frit·
No.
Tar·
nish g:;.:
·~" bearing
area
resistance voltage
Film
new Polar·
ting
field Remarks
ftlm or old ity
8 Ab Rf UF XF
A 10-,.m• 10• n V 108 V/m
1 cu.o• 680 - - 4.05 - - 0.6
2 940 2.5 - 5.45 - - 0.58
3 " 240 2.5 4 2.5 new Cu+ 1.0 graphiteascount<r
" electrode
4 240 2.5 4 4.1 new Cu- 1.6 do.
5 " 940 2.5 7 5.5 new Cu- 0.58 do.
6 " 1480 2.5 (12) 4.2 new Cu+ - do.
"
7
"
1110 2.5 0.5 ± 0.3 3 ± 2 new Ag- - Cu 2 0 deposited on
Ag
8 NiO 125 3.3 1.6 1.5 old Ni± 1.2
9
"
275 3.3 - 4 old - 1.45
10 300 3.3 (5) 3 to 6 old Ni± 1.5
11 " 1250 3.3 10 to 50 17.5 old Ni± 1.4
"
12 1250 3.3 - 13 new Ni± 1.05
13 " 2400 3.3 50 to 100 33 old Ni± 1.38
"
14 cu.o• 450 1.5 5.5 - new Cu± in air, 20 °C
15
"
450 1.5 6 - new Cu± after 1 day in
vacunm at 20 oc
16
"
450 1.5 250 - new Cu± after heating
50 hrs. at 150 oc
in vacuum
17 450 1.5 6.2 - new Cu± again in air, 20 'C
18 NiO 1250 3.3 6 17 old - in dry air, 20 oc
19
"
1250 3.3 20003 34 old - 2.7 in vacuum at
18 °C to 100 oc
after heating
at 300 oc
20
"
1250 3.3 0,073 1.1 old - in vacuum at300'C
1 The greater part of the data of this table are taken from figures and tables in
R. HoLMet al. [8]. The rest are based on other measurements in Siemens' research
laboratory.
2 Cf. critics on p. 142.
3 Cf. HOLM and collab. [8] Table (5j).
Table (27.10) summarizes not only fritting voltages but also obser-
vations on film resistances R1 before fritting showing their dependence
on the film thickness. The NiO films were generated by heating the
metal in air at high temperatures; the Cu 20 films were prepared1 elec-
trolytically. The thickness of the NiO filmswas determined by means
of interference colors 2 and that of the Cu 20 films by means of a titration
method after dissolving the film in a suitable liquid. The Ioad bearing
areas of the contact are greater than according to Eq. (7.01) because
they are partly produced by elastic deformation.
The R,-values recorded in Table (27.10) are denoted as normal since
they belong to a clear group of most frequently appearing readings.
Markedly small breakdown voltages are noted and explained in
Section J.
We revert the discussion to some phenomena that are exemplified
by Table (27.10). Heating the film in vacuum deprives it of excessive
oxygen that is, of acceptors3 (see Appendix Il). This manifests itself in
an increased resistivity, cf. nos. 16 and 19. Heating also seems to in-
crease Xp, cf. no. 19. But, according to tests in the Stackpole Labo-
ratory, simply varying the temperature of NiO in dry air between
- 150 and about 100 oc affects XF very little. If there is an increase at
the low temperature, compared with room temperature, it certainly
is below 20%. In vacuum, beyond 100 to 200 °C, Up and thus Xp be-
gins to drop rapidly with rising temperature, see no. 20.
The humidity ofthe air seems to have no infl.uence on XF. The con-
tact duration affects X F, though not to that extent as is indicated in
Fig. (6r) of HoLM at el. [8]. It may suffice to say that one finds Xp
lower the Ionger the duration of the contact. Possibly, this is an effect
of fatigue by vibrations, since Xp is sensitive to mechanical stress. If
both contact members possess tarnish films the fritting Yoltage, UF,
usually corresponds to the entire film thickness.
Film resistance. Calculating 12 = R1Abfs and XF= Upfs with data
from Table (27.10) one finds that the evalues have a much !arger spread
than the X 11 values. This possibly means that fields from space charges
interfered with the measurements of R1, which were carried out at
small voltages; and that the same fields could not noticeably have
affected the much stronger fritting fields.
Fritting with alternating current. When applying alternating current
to thin films, it was found4 that at frequencies of 500 up to 106 Hz the
rms fritting voltages agree with the corresponding values for direct
cnrrent. At low frequencies of, for instance, 50 Hz the crest voltages
1 Prepared by R. STÖRMEB. 2 Cf. footnote p. 143.
3 Cu 20 and NiO are p-conductors.
~ HoLM, R. et al. [8] § 6.
of the a. c. agree with the d. c. fritting voltages. Evidently the duration
near the maximum suffices for the fritting process. Notice that the de-
scribed fritting with alternating current is without overvoltages. How-
ever, with more rapidly increasing tension overvoltages may be em-
ployed and then fritt.ing may proceed in microseconds or less.
Oessation voltages at A-fritting. Table (27.11) is the basis for the-
statement (Section B) that with NiO-films the cessation voltage can
approach U1 • Correspondingly the cessation temperature in the con-
tact surface of an Ni-NiO-C contact approaches the melting point of
Ni. This is indicated by the contact voltage approaching the value of
1.8 as is calculated in the example of § 19.
Table (27.11)1. fJesaation voltagea UA. Direct currenf witk matrix nickel anodic
.R, = aeriea reaiatance
Contact I Ni against
NO on Ni
I Graphite against
NiO on Ni
Au agalnst NiO on Ni
Pg = 3.6 3.6 3.6 3.6 3.6 3.6 3.6 3.6 1.6 80

.R,n= 380 90 380 90 15 810 90 15 90 90
Film thick-
ness 8 A
100 -· - - - - - - - 0.52 -
170 - - - - - - 0.38 ·- 0.32 -
220 - - - - -- 0.36 - - - -
360 - - - - - 0.25 0.36 0.48 0.61 -
600 - 0.51 1.67 1.57 - - - 0.42 - 0.33:
1100 to 1270 0.35 0.39 1.6 1.7 1.5 0.40 0.51 - - -
2180 - - - - - 0.50 - - - 0.47.
With the reversed current direction, 10 to 20% higher Uh-values were

observed. The case when one electrode is of carbon is treated also in
Section J and § 42.
With Ag 2S- and Cu 20-films, Uh lies considerably below U1 accord-
ing to E. HoLM [15], also (Section J) and ULSH [1].
In Table (27.12) values of field strength XF leading to A-frittings.
are compared with those observed at other kinds of breakdown.
The X-values in Table (27.12) are strikingly confined to a narrow
range, at least if exception is made for varnish. This suggests the idea.
that the initial process always is the same. It is assumed that the·
process is tunneling from the metal through the barrier into the-
conducting band of the film, i.e., a kind of ZENER effect. One encoun-
ters the objection that the pure ZENER effect provides too small cur-
1 Table (27.11) is due to R. HoLMet al. [8] p. 46.

10 U Electric Contacts. 4th Ed.
rent densities for a breakdown. But the effect will be supported by

irregularities in the lattice. Then the X-values of Table (27.12) are
reasonable.
Table (27.12}. A1rittings compared with other kinds of breakdown
X 11 = fritting field in 1()8 V/m
Sampie
Insulator x, thickness Kind of Author
material ws V/m A breakdown
NiO 1.1 to 1.4 125-2400 A-fritting Table (27.10)

Cu20 1 to 1.6 240 A-fritting Table (27.10)
Ta20 11 2 JusT ct al. [1]
Na Cl 1.5 macroscopic leading to FRANZ [1] p. 250
avalanches
KCI 1 macroscopic leading to· FRANZ [1] p. 250
avalanches
Ge 0.6 macroscopic leading to FRANz [1] p. 252
avalanches
Si 1.4 macroscopic leading to FRANz [1] p. 252
avalanches
Cd-arachinate 4-5 100-1000 A-fritting V. GILLHAUSSEN
Ba-stearate [1]
Varnish 7-10 macroscopic leading to S.WHITEHEAD [1]
and avalanches p.72
Lacquer 7-10 450-600 A-fritting R.HoLM [8]
Nb-oxide 2 200 A-fritting1 ÜHOPRA [1]
G. Observations on B·lrittings. Fig. (27.13) ist informative for B-

frittings, i. e., widening of existing a-spots. One series of B-frittings
commences at 0.02 V with 20 0 and continues until 0.3 0 is reached.
A second series begins at 0.15 V with 0.140. In both series the essen-
tially metallic character of the a-spot after any B-fritting is evidenced
by reversibles. Assuming one single a-spot in either case, the radii of
the sp()ts as computed from Eq. (4.15) are 1.5 · 10-9 and 6.3 . 10-8 m
respectively. Inserting these volt- and radii values in Eq. (27.04), one
obtains the force F that produces the B-frittings;
at 0.02 V, F = 7.5 ·10-13 Nfion at about room temperature
and
at 0.15 V, F = 7.0 ·10-13 Nfion at about 300 oc
thus values of the same order as in (27.06). It is reasonable to assume
that fritting appears only in one spot; but it should be considered that
in the event of several a-spots, stronger forces would result for the
widening·effect of these spots.
1 ÜHOPRA believes that he observes avalanche eifects, but a thickness of 200 A
and a breakdown voltage of 4 V are too little for avalanches to develop.

Fig. (27.14) shows R U-curves obtained with highly polished and
thoroughly cleaned nickel rods in air measured at the light Ioad of
0.12 N. Curve I shows a quasi-metallic contact with resistance drop
between 0.09 to 0.17 V due to the widening effect of B-fritting, with
reversible branches of essentially metallic character. The minimum
film recession force, i. e.,
for one single a-spot is 100
!2 P· "l 7g -
~
-
estimated at F """ 8 .10-1 3
so
""
Nfion at a contact tem-
perature of 130 to 300 °C. zo I'>
Surpassing 0.17 V with a c['"_
more metallic a-spot on 70
~
the same rod and under s
the same conditions as for
curve I exhibits an addi- z I
tional resistance drop, in 1
the voltage range of 0.19 P- ZG.7g ~
to 0.27 V at about 500 oc
when the material within
L~ ""-- ~
the constriction softens. 7 )..
This is marked U8 on the 1 -
figure. Softening will surely h
enlarge the contact area;
probably only the smaller o.oz
Scroped u
Jl,
areas on elevations1 (5 to P-1.1g _.o,
OJJ1
10 of them) can obtain an
average pressure equal to 0.00'S - -- - '
the respective hardness. ~/
O.fXJ.'Z
Thus even lapping with -:.", t1" 10G3'J
- _,_ -ö - "-l>t.--<1'-"" ~ ~ml

the finest diamond paste 0.001
seems to have left humps
I-
t<-
on the HERTZ area. These Colculaled ~ Zt\'1

so·c w
yield plastically at the re- 01J01 0/XJZ OiJOS 0.01 aoz aos 0.1 02 V OS
spective temperature and Conlucl WJ!Iu;e U
partly rupture the film. Fig. (27 .13). B-frittings of copper oxide on copper rods.
The rods were polished, cleaned and protected against
H. Applied Fritting. further OXidation by a Jubricant that was tirst washed
Fritting has been used for olf before the start of the measurements. A~ about
0.02 V and 0.15 V marked B-frittings occur. The B-
a long time to keep inter- frittings at 0.18 to 0.2 V occur within the range of sof-
communication switch- tening of the constriction at a contact temperature of
about 400 °C. Here B-frittings certainly are supported
and selector contacts in by plastic yielding
1 The curvature radius r of the elevations can roughly be computed on the basis
of Eqs. (4.15) and (1,2). With hardness = 19 . 108 N/m 2 and modulus of elasticity
= 20. 101 0 Nfm 2 one finds r tobe about 10-3m.
conducting condition, or to "wett" them as the term has been in the

U. S. A. The films concerned remain so thin that their A-fritting volt-
age seldom exceeds a few volts. Consequently such contacts are easy
to frit.
1
2
I
'i..Ue
J.
I US N.I
.
L.
05 crossed
~
~~
Ot
......,-·"" r:;--~ ..u.u.
-· /~
---- / ::?
][
r - - -·
01
tooo·c
I
700
I
~ 300
I
JOO"C
00!
001 0.0! 005 01 0.2 05 V 1
u-
Fig. (27.14). R U-curves, due to E. HOLM, private communication, with nicket rods (0.32 cm dla·
.m eter) lapped with flnest diamond paste. P ""'0.12 N. Curve l i is transferred from a U I-oscillo·
gram taken with quickly varied current I.
The fritted state is recognizable by the contact voltage dropping to

about 0.3 V immediately after A-fritting has ceased, independent of
the series resistance R.; cf. Table (27.11). A tension of 0.3 V is the
normal cessation voltage U" after A-fritting between metallic contact
members. Since different parts of an ohrnie circuit assume voltages
proportional to their resistance, the contact res:stance at Uh with a gi-
ven emf is proportional to R •. It therefore can easily be calculated.
To keep a selector contact sufficiently conducting for telephone
connections, it is inserted in a high ohrnie circuit (order 105 Q) contain-
ing an emf of 48 to 60 V, which maintains a normal current of the
order of mA through the contact. In the event the contact resistance
rises too high, the contact voltage rises to a value that can frit the high
resistivity film. It is not necessary to use D. C.; A. C. and pulses have
been employed. However, after a thorough investigation, BoRCHERT
and RAu [2] recommend D. C.
I. Metallic or quasi-metallic contact through an opening in a rela-
tively thin contact film. Imagine an opening formed by A- and B-frit-
ting through the film as demonstrated in Sections B and C. There, it
was assumed that the two members are pressed into contact within
the hole. This is self-evident if the pressure between the members ex-
ceeds the hardness of at least one member. But it will be shown that
by virtue of the thinness of the film, elastic deformation is able to pw-
duce a good contact with much less pressure. It is sufficient to treat
a numerical example, although considerably idealized, by referring to
Fig. (27 .15). This computation is rough and gives merely t.he order of
magnitude.
We assume a lower infinitely hard member, li, carrying an equally
hard film of thickness y with the circular opening A B of diameter 2 a 1 •
The Bat semi-infinite upper mem-
ber, I,is pressed against the film ,
bulging through the opening and
hitting member li at the circle
CE with the force P 2 • Let PI
be that part of the pressure on
the upper member that is not Fig. (27.15). Forming of an a-spot after A-frit-
ting has ceased. Notice: film thickness y ls wry
carried by the film . Then the much exaggerated
actual bulge is produced by
PI - P 2 • Such a force would produce the bulge AC FE B if the hole
were deep enough. Actually the force P 2 flattens the bulge CF E into
ODE.
Both AC FE B and CF E can be roughly computed by meam: of
an inversion of HERTZ' formulasi given in Appendix I. Denoting
AB = 2 a1 and DE = a2
with r = the radius ofthe circle AF Bto be determined and E = Y o uKG

modulus of elasticity of the material of member I , we have
(27.16)
and , as r eadily read from two similar triangles

a_1 -_ a2 r
"",_
(27.17)
y al
For the numerical example we choose PI = 0.1 N; y = 3- to-s m;

a 1 = 10 5 m corresponding to H = 3. 108 Nfm 2 ; E = 1011 Nfm 2 • Anti-
cipating that a 1 and a 2 differ only slightly, we infer from Eqs. (27.16)
that P 1 ""' 2 P 2 • Then each of the equations can be written
to-s = 1.11 Vo.o5. r;1on

giving
r = 1.46 -10-s m
1 Whether the deformation is an indentation or a bulge does not change the
mathematics except for the interpretation of the direction of the deformation.

IOE
Substituting the values for y, a 1 and r in (27.17) yields

a1 - a2 = 4.4 · 10 7 m
which is 4.4% of a 1 . This shows that even elastic yielding provides con-
tact in nearly the entire opening.
J. Polarity dependence of frittings of tarnish fllms on silver and
copper1• Tobegin with, the phenomena are described with silver sulfide
films (Ag 2S) because in this case they are particularly pronounced.
Consider the crossed rod contact between two polished silver rods, the
one clean, the other covered with an Ag 2S tarnish with a thickness of
a few hundred A (matrix rod). The diameter of the rods was 0.32 cm
and the Ioad about 0.1 N. Symbols as u+and u- for contact voltages
refer to the matrixrod as being anodic or cathodic, respectively.
With a cathodic matrix rod the film breaks down by means of a
normal A-fritting at u; ~ 5 V. But, with an anodic matrix rod, the de-
crease of the film resistance is produced by another physical process.
TuBANDT [1] discovered that positive ions are moved quickly by
an electric field through halogens as Agl, AgCI and sulfides as Ag 2 S
and CuS, and that the ions
have the ability to conglom-
10
,,MJinxAg~l I ~ ~~
~ Sulfitfer/ Ag IJS cleon Ag
erate, forming electronically
conducting bridging filaments
from electrode to electrode.
,
70 w This is evidently what hap-
I / pens when the matrix rod is
3 / anodic. The matrix rod is
able to deliver ions into the
t I'
/I
/ film that has grown on it,
but the clean silver rod, when
1
j {jht
~
I ......_
used as anode, is unable to
give off ions to the film on
~
;~ p the other rod, at least not in
'?> a sufficient number for the
10 0
.~ r'
effect described.
7 ;-:-r IF
u
~
1tT " I Fig. (27.18) demonstrates

001 0.02 aos ot oz as 2 V s the process. With anodic ma-
u- trix rod, Ag+, the intially high
Jfig. (27.18). Due to E. HOLM [15]. Typical RU-
curves between crossed rods of polihed silver, one resistance of between 105 and
clean, the other one sulftded. The fritting valtage 106 n, breaks down at about
Ui is about 100 times higher than UJ;
0.05 V after a short delay
because bridging filaments are generated by positive ions . During ~
following increase of the current, B-frittings (perhaps supported by
1 This Section is based on the paper E. HoLM [15].
ion-conglomerations) adapt the diameter of the conducting filaments

to the current, somehow forming normal a-spots that endure up to
0.06 V. The reversibles indicate that the a-spots are fairly metallic, at
least quasi-metallic.
Going back to the curve for matrixrod cathodic, Ag- , we notice
that the A-fritting leadstoanormal cessation voltage of about 0.12 V.
However, the reversible rises when the current decreases, indicating
that the conducting spot (or spots) is not clean but carries a thin film
that must have remained in the A-fritted spot and that seems to con-
10 G
Ag ~ w
~ ........ Sulfided s11rer usgrophile
I
t
~g l!l """-<
IJU/
I
J
~~~
I .
1/
, ~u;;
~~
~
10 ' ~ "::;
~
01 1 V 10
ll-
Fig. (27.19). Due to E. HOLM [15]. Typical R U-curves between crossed rods, one sulftded silver,
the other clean graphite. The dashed curve is calculated (Eqs. [19.02] to [19.06]) in accordance with
the temperature coefftclent of the resistivity of the graphite'.
duct by a tunnel effect. We know from Fig. (26.08) that the tunnel
resistance, a, increases with decreasing voltage. Later increase of the
current beyond the cessation value leads to B-frittings apparently
without affecting the respective thin film in the contact.
Fig. (27.19) refers to the same sulfided silver rod as was used for
Fig. (27.18). Butthis time a counter electrode of graphite is used1 • By
this change, neither the fritting voltages u;. and Uj;. nor the cessation
1 The graphite used is characterized by eo = 5.2 . 10-5 0 m at room temper-
ature and about 2.6. 10-5 n m at the elevated temperatures; heat conduction
~ "" 27 wattfm °K which probably decreases to 16 wattfm °K at high tempera-
tures; hardness H = 1.8. 108 Nfm 2 and modulus of elasticity E = 52. 108 Nfm 2•
voltage 1 U~ are affected. The only difference is that the cessation volt-
age Uh, after A-fritting is high, namely about 2 V. This is simply the
consequence of the graphite member taking over the greater part of
the contact voltage; cf. the example in § 19. Again, the reversible in-
dicates a thin film remaining in the-a-spot produced by A-fritting.
Contrary to this, with Ag+, a fairly ciean final contact seems to be pro-
duced. The straight line indicates that the contact is not heated.
Evidently the matrix rod alone is responsible for the difference
between u-p and U"F which are independent of the material in the
counter electrode. This was stated with different materials in the latter
electrode.
The nominally flat contact between a graphite brush and a silver
ring behaves quite in accordance to Fig. (27.19). This has a bearing
on the theory of the brush-ring contact, treated in § 42.
The contact between polished copper rods, one clean and the other
one carrying an oxide tarnish shows qualitatively similar effects as
described above, only this time, the ration U"FfU"F is about 3 instead
of about 100.
Remarkl. The portionoftheR U-curves in Figs. (27.18) and (27.19)
until the breakdown occurs is reversible. The conduction seems to be
carried for both polarities by electrons tunneling at low temperatures
from the metal into the conducting band ofthe semiconductor whereby
the resistance, R, decreases with increasing voltage. This is similar to
the reversible portion (A) in Fig. (27.07). Cf. the remark made there
about a low temperature within the undamaged film.
Remark I I. It is significant for the correct judgement of the phe-
nomena described in this section, that films of 50 to 400 A, or even
several 1000 A thick do not have sufficient semiconducting bulk to
deliver the carriers for a rectification effect; cf. text following Eqs.
(33.02) and (33.03) about their validity for rectification. Furthermore,
U. B. THOMAS[2] and E. HOLM[15] observed that a current reversal does
not at once show the effect characteristic for the reversed polarity while
a rectification process is instantaneous. Fig. 5 in E. HoLM [15] shows
that the respective effect on current reversal needs about 1/3 sec to
be produced. This is a clear indication that here semiconducting recti-
fication is not essentially involved.
1 The fact that U h with Ag+ is the same whether the counterrod is of silver or
graphite makes it apparent that the metal diffuses into tlte graphite making the
constriction resistance small.
§ 28. Adherence in dry contacts which arenot heated... 153
§ 28. Adherence in dry contacts1 which are not heated to any

inßuential extent by the current
A. lntroduction and theory. The fact that hot metal pieces are
capable of welding when forced into contact is an ancient experience
of the blacksmiths, but any attempt to analyze the role the high tem-
perature plays in the phenomenon is of a recent date. Anticipating the
results of both the present and the next chapter we state that the high
temperature softens the metal forming !arge, metallic contact areas
by plastic deformation. Furthermore, a high temperature readily de-
livers activation quanta for formation of covalent bonds strengthening
the weid. The present Section does not consider the infl.uence of the
temperature on adherence but is devoted to "cold welding" . This weid
is produced at room temperature (and lower) immediately on contact
make.
Cohesion in a metal is only partly based on directed, saturated
covalent bonds, more in hard metals as wolfram, less in copper2. In
copper the bonds are prevailingly "unsaturated covalent'' , here called
"metallic", produced by socalled free, conducting electrons.
Fig. (28.01) illustrates the action of a covalent bond. Orbitals of
the pair of ions involved provide a great probability for the electrons
to appear between the ions, forming an "electron
cloud". This cloud attracts both ions thus binding
them together. The bond has its direction strictly
oriented in the lattice and therefore a distinct alin-
8·1~ (~
E!eclron cloud
ing of surrounding ions is required for full develop-
ment3. Fig. (28.01). The ion
attracting "'electron
Also, the "free" conducting electrons form clouds cloud" in the orbital,
which produces a
between the ions but less dense and less effective. covalent bond bc-
The total binding of any atom or ion to the lattice in a tween two ions
solid metal crystal is ofthe order of 3to 8 e V (depend-
ent on the metal) when based on covalent bonds ; only 0.5 to 2 e V how-
ever when based on metallic bonds. The small value of 0.5 e V belongs
to alkalimetals which have about a 50% }arger spacing than copper, or
iron, and correspondingly weaker bonds.
Consider contact make between perfectly clean metallic surfaces. We
assume both members tobe of the same metal, although the case of a
1 A detailed discussion on adherence in contacts by means of liquid alien con-
tact films is given in BownEN and TABOR [4] p . 299ff. The adherence between
grains of fine powder and the supporting body which influence granulation processes,
is essentially due to adhering water films ofthe kind described in§ 24; see BATEL [1].
2 See for instance SLATER [2].
3 We do not discuss ionic bonds since they do not appear in metals.
bimetallic contact is only quantitatively different. In the instance of

contact make, conducting electrons move across the contact surface
about as frequently as across the boundary between crystallites in the
bulk of the metal. On their way , they produce metallic bonds which are
certainly weaker than they would be after some rearrangement of the
ions by means of diffusion; still however of the same order of magni-
tude as normal metallic bonds.
Metallic bonding in a contact is responsible for the immediate
"cold weiding". In contacts at rest, the welds will grow stronger by
means of rearrangement of atoms and development of covalent bonds.
In sliding contacts there certainly is no time for covalent bonds to
develop. We conclude that hard1 and soft metals will behave essentially
different with respect to cold weid and friction. Soft metals, as copper
and silver, which owe their coherence largely to metallic bonds, will
form cold welds which are able to tear out debris during sliding, mak-
ing the surfaces rough and scuffing. Hard metals as wolfram , owe their
hardness to covalent bonds and their cold welds during sliding are
much weaker than the normal bulk coherence. Therefore, there is little
y
or no reason for the shear-
01. ing and sliding to proceed
Vmale
in a surface different from
the initial contact surface.
V I~
V
The adhesion wear is small,
but wear by virtue of inter-
Au /"( locking may occur.
~oSn Cu
W e shall pursue these
so ideas into numerical details .
In Fig. (28.02), the vapori-
~~ zation heat, r, in calories
0 s 10 TS ZIJ ·tJBNfm• t.f
per mole (i. e., the total
Ball intlenlulion hortlness ofonniJQ/td put: melul bond strength) is plotted
Fig. (28.02). Relation between vaporization heat and against ball indentation
hardness of annealed metals
hardness of annealed met-
als. Notice that the ball
indentation hardness actually measures shearing strength, although
it is directly observed as the average pressure in the indentation; cf. §I.
One reads from Fig. (28.02) that the hardness of the annealed meta!
is roughly proportional to that part of the vaporization heat which
adds to the vaporization heat of the softest metals. As remarked above,
1 Plasticity and hardness of metals are known as effects of dislocations; see § I.
However, the strength of the anchoring of dislocations is so dependent on the

strength of the bonds in annealed metals, that the reference directly made to the
bonds is justified.
§ 28. Adherence in dry contacts which are not heated. . . 155
sodium and potassium are exceptions. We are induced to make the
following conclusions; cf R, HoLM [42].
Pure metallic bonds produce small shearing strengthl, and most what
i8 mea8ured a8 strength and hardne8s in hard metals, derives from covalent
bonds.
Surely, strain hardened or alloyed metals with a particular anchor-
ing of dislocations do not fit into Diagram (28.02).
For a rough quantitative computation we may represent the small
hardness of the annealed soft metals Sn, Au, Cu by H 1 where
(28.03)
Then the hardness H 2 of the barder metals is according to Fig. (28.02)
H2 :::::: 3 + 0.184(r- 70) (28.04)
where the la8t term refers to covalent bonds.

Now apply Eqs. (28.03) and (28.04) to the problern of cold weld.
Since the initially active bonds are of the same kind as those for H 1
we expect any cold weld to begin with a tensile strength 2 of the order
of H 1 /3 Nfm 2 • In soft metal contacts, the strength grows only slightly
higher with time. In hard metal contacts, a steady rearrangement of
atoms will gradually introduce covalent bonds3 and develop a higher
weld strength.
This has a bearing on the theory of sliding contacts. We expect
that only metallic bonds have time to develop in sliding contacts. Thus,
the shearing strength (equal to the tensile strength according to § IE)
will generally be tp = H 1 /3. But, the contact area is proportional to
H;I; anticipating Eq. (36.08) with ~ = constant we expect the friction
coefficient f to be proportional to
H1 3
(28.05)
H2 = 3 + 0.184 (r - 70)
Cf. § 70 C and D.
B. Experimental verifi.cation. Relevant experiments on friction and
wear will be treated in § 36. A thorough investigation of how a cold
weid in a soft metal contact depends on the temperature seems to be
missing; but an informative demonstration was made long ago by
1 It seems that the force with which the contacts are pressed together plays a
minor part for the strength. Cf. the discussion about experiments by HOLM and
KIRSCHSTEIN in § 37.
2 See the end of § I E.
a If free valences are not present, the development of the covalent bonds seems
to depend on activation energies of the order of 1 eV. See, for example, ERDMANN·
JESNrrZER [J].
R. HoLMl. Heuseda degassed vacuum switch fitted to a stethoscope.

On switching off (without current) at room temperature a cracking noise
was heard - a result of ~he sudden break of the adhering contact. Then
the end of the stethoscope, tagether with the switch containing tube,
was immersed into liquid air and kept there long enough (say an hour)
to obtain the low temperature. At repeated switching off, the cracking
noise was heard again with the same intensity as before, even in cases
when the contact had been closed only for a second. Warming the con-
tact up and cooling it again was repeated without any change of the
noise. On admittance of air into the tube that contained the switch,
the noise disappeared after about one minute, evidently because an
adhesion film formed in the switching surface that essentially prevented
the adherence between the contact members. Since the amplitude of
the vibration of the switch will be proportional to the adherence or
elastic deformation at rupture, the noise at rupture is a fairly sensitive
indicator of the degree of adherence.
Another confirmation that elevated temperature is not essential
for a cold weld is given by HoFMANN and RuGE [1]. They produced
at -150 oc (at high pressure) a cold weld, between clean aluminium
contacts having the same strength as obtained at room temperature.
It is obvious that the welding discussed with reference to Fig. (28.02)
refers to perfectly clean, metallic contact surfaces. The slightest con-
tamination would change the density ofthe electron clouds, diminishing
or annihilating the bonds between the clean surfaces. BRY.ANT et al. [1]
report a very instructive observation on the influence of the oxygtln
deposit on adherence in a solid contact. MoS2 and graphite were cleaved
in airandin ultrahigh vacuum of 10-13 Torr. From the cleavage work,
the bondage could be estimated to 0.2 to 0.3 eV for graphite in vacuum
and about 6 times less for MoS2 • This is for graphite about an order of
magnitude greater than previously reported values obtained in air 2 •
No similarly thorough investigations about the influence of oxygen
deposit on the tensile strength in contacts between metallic mernbers
have been published. Adhesion measurements made in air are too much
influenced by the presence ofremaining films (cf. § 29) to permit calcu-
lations about adhesion per unit area.
Cold welds play a significant part in the theory of friction and wear
(see §36). For practical purposes, it is more directly important to know
about its appearance in contact spots which have been rubbed clean
in gaseous environment than to know about welds in a perfect vacuum.
This may be the reason why pertinent investigations have been made
1 R. HOLM [30] p. 148.

2 BRYANT et al. (Jl p. 124.
§ 28. Adherence in dry contacts which are not heated... 157
in air. For examples of cold welds in contacts just scraped and observed
in air, see Fig. (9.02).
Elastic peeling of cold welds. It is an old experience that the ad-
herence against lifting a contact apart is smaller than expected from
comparison with the shear strength of the contact, i. e., with friction.
Therefore, many earlier investigators did not even believe in the ex-
istence of cold welds. The inconsistency is due to the interference of
elastic forces. An instructive example of how these forces act is ob-
tained from the ball indentation test.
Compare Fig. (28.06) with Fig. (I,10). A hard ball has been pressed
by the load P against a flat member into the position ABC. After re-
moval of the ball the indentation would
recover to ADC if adhesion were absent:
ABC DA represents the elastic deforma-
tion that canied the load. If the surface
had been clean, a cold weld would have
been produ0ed in the entire indentation Fig. (28 .06). A ball having been in·
dented to ABC by the Ioad P, is
ABC. Butwhen theload is removed,elastic released from P so as to be lifted.
forces interfere with the adherence in the The bottom of the indentation re-
covers to A D C whereby the ball
weld. Fig. (28.06) demonstrates a moment adheres only to EDF
during the lifting of the ball. In the region
near the rim of the indentation, elastic tension has prevented material
of the flat member to follow the ball through the gap designed. In the
middle region, EDF, adhering material has been lifted to a small
deviation from A D C. Then the corresponding small elastic counter-
forces have not been able to peel the weld. The observed weld strength
corresponds to the final small area E DF. In hard metals the elastic
forces usually dominate over those in the cold weld and the adherence
is too small tobe observed. The very soft indium represents the other
extreme. It may adhere to the indenter with a force that is stronger
than its own elastic reaction, and the weld strength corresponds to the
total indentation; cf. BowDEN and TABOR [4] p. 306ff. In addition:
In the case of a soft flat member, the radius of ADC differs relatively
little from the radins of the indenter. This means that the peeling force
is weak. This has been calculated by McFARLANE and TABOR [1]
p. 231ff, K. L. JoHNSON [1] , and particularly BowDEN and TABOR [12]
p. 93ff.
There has been speculation on whether a weld between different metals
depertds on their mutual solubility. MACHLIN and YANKEE [1] report
that iron and silver do not solidphase weld. However, in their experi-
ments, the friction coefficient f was below 1 in all cases. This indicates
the presence of physi- or chemisorbed films. A more extensive explora-
tion of the weldability of various metals with iron has been reported
in summary by RoACH and GooDZEIT [2]. They claim that for a metal
t.o form a good bearing against iron, i. e., to seize only a little, it is
required that their atomic diameters differ greatly. The idea is that
relatively few atoms in the contact will have mutualpositionsfav orable
for adhesion. However, one has to wait for a more complete report on
the experiments before taking a stand with respect to this theory.
C. Location of the weid break. So far, the discussion has been con-
ducted as if the weid would break in the initial contact surface. This is
Fig. (28.07). Due to E. RABINOWICZ [2). Autoradiographs ol copper surlace alter indentation by
radioactive copper hemisphere under a Ioad of 100 A. (a) Clean. Mctal transfer ~ 6 · l o-a N.
(b) Burface covered with cetane. Metal transfer = w-u N (exposures have been adjusted to give
equal intensities of balckening)
no rule. In portions of the welded area, the cohesion can be stronger

than within one of the members. An impressive demonstration that
welded spots can be stronger than the base meta! below them has been
given by RABINOWICZ and TABOR [1]. They obtained autoradiographs
of copper transferred from a radioactive contact member to a non-
radioactive one. This shows that the adherence in parts of the welded
contact surface was stronger than the cohesion in some interior sur-
face . Fig. (28.07) is an example that is particularly interesting. It could
be inferred from exposure times and blackening intensities that with
boundary lubrication, 6 times less material was transferred from one
electrode to the other, than in the event of a dry contact. This probably
means that the ratio between the total areas of metallic spots was
about 6; cf. § 38. Any considerable accuracy in the determination of
the conducting areas, as was intended, is not achieved. Also, other
investigators, for exa.mple HoFMANN and RuDE [1], report material
transportat the opening of cold welds. Pictures in BoWDEN and TABOR
§ 28. Adherence in dry contacts which arenot heated. .. 159
[ 12] p. 110 show that several 10-6 g are transferred from a hammer
during a single blow on a flat copper surface.
D. The practical significance of the adherence in non-heated con-
tacts. The easy peeling of welded contacts, a nuisance for the in-
vestigator of adherence, surely is favorable totheusage of contacts in
relays. Actually, adherence causes trouble only in very sensitive relays;
cf. § 30 G on microcontact.s. It is otherwise if the contact assumes a
high temperature.
The generation of clean metallic surfaces in the contact is the basic
process of the cold (press) welding that has been practiced since 1947.
The surfaces to be welded are cleaned, for instance, by brushing with
a rotating wire brush. It is more important that considerable plastic
ueformation takes place at contact make, causing rupture of disturb-
ing films and the joining together of freshly liberated metallic spots.
To that purpose, the surfaces to be welded are furrowed and put to-
gether with crosswise riffles which strongly deform when pressed against
each other. For details about cold welding by pressing sheets or pieces
of metal together (deformed to about half their thickness), see, for
instance, BARNES [1] and an article 1 by the "Coldweld Corp". A thor-
ough review of solid phase welding is MILNER and RowE [1].
An ingenious application of cold welding is realized by the Bell Tel.
Co in the solderless connection between a terminal with reetangular
cross section and a wire wrapped arounditwithhightension 2 • The great
plastic deformations appear at the edges of the terminal, where welded
spots surely are responsible for the good and Iasting connection. How-
ever, they are too small to appreciably withstand unwinding of the
wire. Therefore, it is important that the tension in the wrapped wire
does not totally relax. Actually, according to MASON [1], there are
indications that half of the initial tension in the wrapped copper
wires may remain in a copper wire after 40 years. Meanwhile, creep
causes an increase of the welded areas as was stated.
It is surprising that cold welding has been specially successful with
aluminum although in air it has a thick protecting Al 20 3 film. The
probable reason is that this film breaks so as to leave broad fissures.
Other metals with thinner films, as copper and iron, are less prone to
pressure welding. Evidently, their filmsrupturein less suitable fissures.
It has been expected that a great ratio of (film hardness)f(metal hard-
ness), as in the event of Al, is favorable for pressure welding. Investi-
gations by TYLECOTE et al. [1] do not support this idea.
1 Iron Age, october 1950 p. 58, where other Iiterature is cited.
2 See MALLINA et al. [1]; ELLIOTT [1]; and MILLS (1].
§ 29. Adherence in contacts that are heated by the current

passing through them. Re~istance welding
The current itself does not affect the adherence in contacts. The in-
crease of adhesion that is observed after passage of a high current is
the direct result of the elevated temperature which softens the metal
around the contact. The temperature may even produce melting, and
in addition, activate covalent bonds.
The behavior of contact members that carrytarnish films is of parti-
cular practical interest. If at contact make the metal merely yields
elastically, different points in the contact surface increase their dis-
tance extremely little and the film may follow the deformation with-
out rupturing. However, as we have seen in § 28, plastic deformation
may be produced by the pressure where the members contact each
other. Groups of atoms thereby move from their positions, rupturing
the continuity of the initial metal surface. An increased surface is
generated. The film cannot follow this movement and extension, and
will break open, providing the opportunity for the formation of metallic
spots. This process is enhanced in its consequences in the event the
a-spot constriction becomes softened by the current heat.
In order to produce metallic spots, fissures must evidently occur in
the tarnish films on both members. If at the temperature attained, one
member is sufficiently hard for yielding merely elastically, its tarnish
film will prevent noticeable metallic contact and adherence.
Some instructive measurements1 concerning welding, when the
softening contact voltage is reached, are summarized in Table (29.01).
The device that was used for the observations in air is pictured in Fig.
(9.01). The contact, closed at a Ioad P, was subjected to a current 11
for a certain time (a few seconds) producing a contact voltage U1 • Then
the current through the contact was diminished to a very small value
that merely was kept as an indication that the contact was conducting.
Mechanical tensionwas then applied by means of inverse coil current.
If 11 and U1 were small enough, rupture occurred without any notice-
able tension. The experimentwas repeated in steps with I> 11 giving
U > U1 . Ultimately, a Iimit U.a was reached when adherence became
noticeable; when this Iimit was surpassed,adherence rose quickly. We
call this Iimit adherence voltage U ad.
Table (29.01) shows that clean gold contacts begin to adhere in the
very beginning of softening. The results with bimetal contacts. confirm
what has been said above. A condition for adherence isthat the member
with the higher U, must also be softened so as to offer fissures where
1 R. HOLM [29] p. 133.
§ 29. Adherence in contacts that are heated by the current passing. .. 161
Table (29.01). Voltage u.,
where adherence ris€8 quickly compared with sojtening
voltage U, jor mono- and bimetal contacts
Material
I Fe Cu IBrass I Ag
Iw Pt
I Au wc Ag onPt Pt on WC
Adherence
begins at 0.10 0.35 0.60 0.25
uoil 0.35 0.10 0.35 0.24 0.60 0.40 0.060 1.0 0.30 0.70to0.90
the softer member can touch the harder one; that is, the contact volt-
age must attain the value U8 of the harder member1 • Silver on plati-
num adheres when the softening temperature of platinum, 530 °C, or
[l, = 0.25 V is reached. The contact of platinum on wolfram carbide
adheres at the softening temperature of tungsten carbide, 1000 to
1-500 °C.
Table III in BowDEN and TABOR [12] p. 100 shows that the onset
of strong adhesion between thoroughly clean metallic members in air
occurs between 0.4 T 1 and 0.5 T 1 (°K) where T 1 is the melting point
of the metal. This is the temperature of strong softening.
Gontact adherence after the melting point has been reached or nearly
reached in the contact surface has been inv€stigated by E. HoLM and
R. HoLM [7], and MAu [1]. Before discussing MAu's results we present
a theory for the event that 1) the melting temperature has been reached
in the contact; 2) the contact has had the possibility to adapt its area
to the current assuming a temperature only slightly less than the melt-
ing point. This surely is an idealized picture since in reality the contact
adapts its area to a temperature that often lies considerably below the
melting point.
In order to produce the melting temperature in the contact, the
current must exceed a certain limit I(P) that is determined by the
quality of the metal and the initial contact area which in its turn
depends on the load P. Thus, we assume that when the melting point
is reached, the material yields and the contact members are free to follow
the yield until the contact area A becomes sufficiently large to conduct
the current in the solidified state, that is, without being heated beyond
a limit slightly below the melting point. We denote the ultimate area
by A (I) since, evidently, it is a function of I alone. The role of the load
Pis largely to move the movable contact member in the moment when
melting occurs. To meet this purpose, it is required that P be larger
than the electromagnetic repulsion according to Eq. (11.04). If P cor-
responds to this requirement, its exact value is of little importance,
since the molten metal does not resist any pressure.
I The softening temperature appears in the hottest section of this member,
which is just as e:lfective for the yield as softening in the contact surface; cf. § 18.
It is readily shown that the ultimate contact surface A (I) is pro-

portional to the square of the current I. To prove it, let R(eA.) be the
constriction resistance when I is flowing, i. e., when the contact has
the high temperature. Since R (eA.) is inversely proportional1 to (I) VA
one finds the ultimate contact voltage U (which is equal to the melting
voltage U 1):
U[ = ri R(eA.)j 2 = const A~~)

or
A (I) = consti 2 (29.02)
If A (I) is known we have at room temperature
Z = SA(I) (29.03)
where Z is the adherence force and S is the tensile strength of the ma-
terial. Since 2 S = H/3,
Z = ! HA(I) (29.04)
Actually, the knowledge of A (I) is no prerequisite for the calcu-

lation of Z. The following procedure is easier. Let R (eA.) be the resist-
ance that is actually measured with the current I immediately after
the welding. The temperature is about the melting temperaturein the
contact (see above). Let R (e0 A.0 ) be the resistance that the same
welded contact would exert against a small current. Then the ratio
:(~~) can be read on Fig. (16.14) for 8""" the melting point. Thus
R (e0 Ao) can be calculated. The load P (I) corresponding to R (e0 Ao)
which would be able to generate the ultimate contact area A (I) at room
temperature can be obtained 3 from Fig. (8.01).
Assuming that P(I) is large enough to produce a specific depth
D > 0.03, the hardness is defined as in § I and thus
P(I) = HA(I)
We obtain by inserting this expression in (29.04)
(29.05)
The measurements differ principally from each other. E. HoLM and

R. HoLM heated the contact by several cycles of an a. c. current
1R(QA) is equal to (}/2a times a temperature factor given in Fig. (16.14).
2See Eq. (J,18) and end of § IE.
3 As described in §SE, Fig. (8.01) is easily completed for any metal and size
of the members not contained in this figure.
§ 29. Adherence in contacts that are heated by the current passing. . . 163
while MAU used the pulse of one single half cycle. However, in case
the repulsion force was weil below the Ioad and the current high
enough, a reasonable confirmation of Eq. (29.05) was obtained by the
HoLMS (using silver contacts) and by MAu (using copper contacts).
Taking into account the shortness of time for the high voltage, one
calculates that only about 90% of the melting temperature was reached.
Table (29.06). Typical re&Ults on Ag [E. HoLM and

R. HoLM [7])
p I Weidstrengtbin N
N A measured I accord. to (29.05)
4.5 2000 4 5.1
11.2 3000 8 11.3
18.0 3360 0 14.0
In the third test at P = 18 N, no weid appeared because the great

P produced a contact area so great that 3360 A could not heat it close
to melting.
In the experiments by the HoLMS with base metals, the heating
time expressed by z of Eq. (21.02) was so short that only a tempera-
ture half way between the softening and melting point was reached .
It is obvious that disturbing films remained in the contact. Although
~' ig. (29.07). Due to MAU [J]. l!'ig. (29.08). Two plates, b, b,
Weid strength as a function of tbe mecbanicalload P to be welded are pressed be·
(in N); amplitude Im= 7270 A tween the electrodes E, E
the contact area became large enough to endure thc current, as seen
from the resistance, the adhesion was small; namely, if expressed as
percentage of the value given by Eq. (29.05)
for Ni steel Mo w
about 35 10 4 3%
Fig. (29.07) is due to MAu. It demonstrates that the weld becomes
weak when either the repulsion approaches the Ioad P or, at relatively
n•
high load, the current is unable to heat the total area to melting during
the pulse. The observation is typical for variation of P at a constant
current.
Resistance welding1 is the only deliberate welding method with high
temperature that is related to the contact theory as it is treated in this
book. For details ofthe methods, we refer to Welding Handbook 2 chap-
ter 30. We add interesting conclusions from application of the <pD-rela-
tion and ofFig. (21.04). A typical arrangement is shown in Fig. (29.08)
where b, b are the plates to be welded together and E, E the water-
cooled electrodes which produce pressure and conduct the current.
They are softer than the plates in order to provide good contact.
In the beginning there is metallic contact only in very small spots,
giving a contact resistance favorable for rapid luiating. Mter a milli-
second, the resistance in the weld-surface will have reached its mini-
mum. Further heating is required in order to obtain the greatest ten-
sile strength of the weld.
During the whole welding process the original contact spot remains
the warmest section, not only because it offers the greatest resistance
in the circuit, but still more because this section is most distant from
the cooled electrodes.
The high temperature entails a high resistivity that is favorable for
concentration of the heating to the weld.
Table (29.09). Spot-welding of rnetal plates
Amplitude of contact voltage U V
I
Contact Amplitude
Welding Ioad of current
contact between electrode
N and plate between tile plates A
Fe plates
Al plates
700
700
0.7
0.3
0.3
0.1
I 5000
13000
A thorough calculation of the temperature distribution is obstructed
because no temperature equilibrium is reached during welding. There
is also insuf:ficient space for long constrictions to develop. Therefore,
the temperature cannot be computed from the voltage using the <p{}-
relation. Typical a. c. amplitudes of voltages during welding are given
in 'l'able (29.09).
§ 30. About stationary contacts in practice3

A. Introduetion. While the preceding chapters deal with the phy-
sics of stationary contacts seen from a theoretical aspect, this chapter
1 Survey: DIXON and TAYLOR [1].
2 Published in 1963 by American Welding Soc., New York.
3 A completion of this Chapter as to moving contacts is given in § 70.
§ 30. About stationary contacts in practice 165
gives examples of how to use the knowledge for understanding statio-
nary contacts as they appear in practice. Here the word stationary
expresses that switching during applied voltage appears either seldom
or not. at all and that the wear, which occurs at switching current by
the contact, need not be considered. Typical examples of a stationary
contact in the sense of this chapter are bolted bus bars, the terminal
0ontacts of an ammeter, even the plug-socket contact, and also many
relay contacts.
The word practice in the caption of this chapter expresses that tlw
main concern is how to avoid obstructions of the contact operation.
for instance by alien particles, films, etc. We shall even base the de-
finition of contact types on the kinds of obstruction they are especiall,\r
subjected to. An appropriate subdivision must apply not only to static
contacts but also to contacts which switch during current ßow, especially
since the difference between "static" and "switching" contacts is som<'-
t.imes vague.
W e rlistinguish
I. Permanent contacts as clamped bus bars and the solrlerless wrapperl
contacts of Bell Telephone mentioned on p. 159.
li. High load (P = several10 N) interruptible contacts as terminal
contacts of ammeters, plug and socket, knife switches.
III. Medium load contacts particularly in air, P = 0.3 to 1 X, as
in intercommunication switches.
IV. Low load contacts with P around 0,3 N.
V. Microcontact8 with P < 0.2 N and open contact volt.age < 0.05 V.
The wear and other phenomena connected with the moving of con-
tact.s is treated in Parts III and IV of this book.
B. Type I, permament contacts, cf. Section D. The reliability of
these contacts is always due to cold welded spots produced at contact
·make and then protected by the Ioad against (transient) openings,
which otherwise could lead to breathing and loss of the weid, see § 22 G.
Into a base metal contact, that has breathed, oxygen diffuses and
oxide continues to develop. The resulting rise of the contact resistance
is often evident from increasing contact temperature. Loss of load in
clamped contacts because of creep may lead to elastic peeling of welds,
cf. Fig. (28.06). Therefore, it is often advisable to rlamp with spring
wa.shers.
A means to protect aluminum bus bars with their relatively great
contact spots which are sensitive to even very short breathing is to
silver-plate the overlapping portions of the bars. CoNNOR and WILSON
(1] describe experiences with silver-coated joints of aluminum con-
ductors. They and also BURLEY [1] and P. QuiNN [1] discuss means
for the plating. Good platings showed a greater tensik ;-;tl"ength than
the basicmetals Aland Ag.
When "permanent contacts" are constructed so that they do not
breathe, their life is practically infinite. RICHTER ancl ScHADE [2] ob-
served that clamped Al-contacts which were good the first year re-
r
mained fit for use 20 years more, at least. MASON 1] computes that
Bell Telephone's wrapped contacts will well have a life of 40 years. His
conclusion is supported by MILLS [1].
A remark about plastic flow in a nominally flat contact. If the con-
tact members approach perfect smoothness, plastic flow would require
a pressure about equal to the yield point in the total area. The much
smaller apparent pressure applied in practice would not be able to rup-
ture the tarnish film. That normal a-spots with metallic portions
develop is due to the actual waviness and asperity of the surfaces.
Fig. 5 in WILLIAMSON [2] illustrates this fact by showing that low con-
tact resistances (little varying with the roughness) appear at rough-
ness amplitudes above 0.5 [LID; but with smoother surfaces, tarnish films
remain because the resistance rapidly increases with decreasing rough-
ness amplitude. One sees that very smooth contacts can be unwelcome.
Metallic contact members can be adapted to each other by repeated
make and break of the contact, without current, and in the same posi-
tion. This results in flattening the surfaces. At further contact make,
the members meet in points with a relatively small curvature, and the
contact spots may be produced by merely elastic yielding. In this state,
their films endure the deformation without rupturing; they add their
resistance to the constriction resistance and are able to prevent any weid.
A note on the elfect ofthe thickness of alien films in the formation of
fissures. Fissures probably appear as easily in thick alien films as in thin
ones provided the basic metal yield plastically. Does this imply that the
thickness of the film has no influence on the final contact resistance? The
most elementary experience contradicts this supposition. It is self-evi-
dent that a good contact is not seenred by fissures alone. but that metal
from the members must be squeezed into the fissures. This surely is more
difficult, the thicker the film. In very thick films, the fissures are merely
channels into which only air can enter, promoting further oxidation.
C. Separable contacts of high mechanicalload. Plug-socket contacts
often have visible oxide films which would insulate if undamaged at
make. The actual conduction is produced either by mechanical rupture
of the film at closure or by fritting. The available voltage of 100 V or
more guarentees frittings. The final adaptation ofthe a-spots may leave
the contact voltage at the order of 0.1 V. The contact is correspond-
ingly heated rendering the socket warm to the touch. This, however,
can be tolerated.
Terminal contacts of ammeters, etc. must have sma1J resistances ,
i. e., their contacts must have metallic or quasi-metallic a-spots in
both members. The clamps are made of brass and are usually slightly
greased in order to prevent oxidation . A Ioad of several hundred
Newtons is reached; cf. Table (30.03). This Ioad and a twist secure a
sufficiently small resistance between clamp and copper wire.
Plug and socket connections in 'Telephone Selectors with a load of
.1 to 3 N and voltages of the order of 5 V provide insufficient means to
break tarnish films. Palladium plated, they function satisfactory; cf.
Section E.
D. Resistance measurements of clamped contacts 1 with force produced
by a scrcw. The contact between a conducting wire and a tPrminal screw
is suitably i.nvestigated in a cricuit sketch-
ed in Fig. (30.01) in which the resistance
of the wireisnot contained in the measured I
resistance. Asufficiently exact calculation -----l
of the screw pressure can be made accord- '--~-'
ing to Eq. (30.02). M in Nm denotes the
torque, h m the pitch and 2 r m the diameter Fig.(30.01). Wiring diagram for mea-
suring the cont.a ct resistance be-
ofthe thread 2 and / 1 and / 2 the coefficients tween a wire and a terminal screw
of friction in the thread and screw head
respectively. In one revolution, the torque does the work 2nM which
suffices for: 1. The displacement h m of the head (equal to the com-
pression of the wire) corresponding to the work hP, where Pis the
contact Ioad; 2. the work of friction in the thread of about 2nr1 P f1 ;
and 3. the friction work between the head and the compressed wire
2nr2 Pf2 , where r1 and r 2 are average values of the axial distance of
the contact spots at the thread and head respectively. The contact load
in both the thread and the wire is assumed to be constant during
one twist. If during this time P varies, one may consider a sufficiently
small part of one turn whereby the factor introduced finally cancels.
Since r1 is somewhat smaller, and r 2 somewhat greater than the radius
of the screw, say r 1 = rf1.1 and r 2 = 1.1 r, we finally obtain
2nM = Ph + 2nrP/~ + 2nrP·1.1f2 Nm

O l'
P= 2nM N (30.02)
2 nr u.~ + 1.1/ + h
2)
See R. HoLM [29] p. 145, cf. R. HoLM et al. [8] § 12.

1
American screws are commonly characterized by two numbers. For instance

2
in 8 by 32 the number 8 is a conventional reference to a major diameter of 0.1640"

and the number 32 means 32 threads per inch. In Germany a screw with a major
diameter = 4 mm is called a M4 screw.
This equation is suitable for computations. Substitution of conveu-

tional values shows that friction makes the clamping force about ten
times smaller than it would be without friction. Exact precalculation.<;
areimpossible since / 1 and / 2 vary considerably from case to case. Fric-
tion coefficients assmallas 0.15 to 0.25 may occur if some oil adheres
to the screw. The coefficients increase the more the oil film is squeezed
out of the contact, for instance, by screwing back and forth; e. g .. on
brass and steel to f = 0.4 and on aluminum and zinc to 0.7.
Table 30.03 gives typical examples of the behavior of contacts
clamped with screws in normal atmosphere1 and kept clamped during
the measured times.
The contacts were kept at a temperature of 100 °0 from the middle
of the first month to the end of the second (with the exception of some
days when measurements were made). At the end of the first month
they were all drawn tighter at room temperature by the same torque
M as was used initially. This was possible because, by creep of the
metal of the wire, the contact Ioad had given way to a certain degree2.
It is strange that as a rule, the resistance did not change on tightening
the contacts. However, during the noted time, a tenth of the contacts
increased their resistances considerably while 1% decreased them mar-
kedly. It. is evident that the contact still adhered in original metallic
points, and that new metallic spots were not usually generated on
tightening the contact.
Table (30.03). Extractfrom mea.mrements an screw contacts carried out in F.L.
Wlre Screw Resistance at room tem-

and lts 1\Iethod of dia- Torque p perature to-• n; t in
Test Clamp cleaning• .lC about months after clamping
diameter ledg~ meter
the wire
mm mm Nm N t = oj o.5 j 2
I 1:!
1 Cu 0.5 Ni oil 2 0.13 280 0.08 0.05 0.05 0.04

2 Ag 0.5 Ni no cleaning 2 0.057 120 0.18 0.20 0.25 0.43
3 Al 0.28 Ni cotton 2 0.057 90 1.3 1.9 2.2 2.2.
4 Al 0.28 Ni oil 2 0.057 110 0.27 0.30 0.28 0.13
5 Al 0.28 Zn oil 2 0.057 90 0.8 18 ' 400 1600
6 Al 0.28 Ag no cleaning 2 0.207 350 0.2 0.2 0.4 1.1
7 Al 0.28 Zn no cleaning 2 0.207 280 1.0 4.5 18 23
8 Al 0.28 Zn oil 2 0.207 320 1.7 3.2 0.84 0.83
9 Al 0.6 Ni no cleaning 4 0.207 130 0.43 0.48 0.50 0.58
10 Al 0.6 Ni no cleaning 4 0.407 200 0.14 0.15 0.15 0.15
11 Al 0.6 Ni oil 4 0.407 250 0.09 0.09 0.08 0.08
R. HoLM and collab. [8] § 12.
1
Cf.§I.
2
3 Oil signifies that the contact members, after abrasion with emery in oil, \lo-er~
clamped together in the oily sta.te; and cotton means that the members were only
rubbed with cotton soaked in alcohol.
Discussion of Tablc (30.03). The importance ofthe pressure from the
screw is remarkable. Really poor contacts appear only at P"" 100 N,
while P "" 200 ~ and more produces Iasting conductance. Only zinc
contacts form an exception 1 • Consequently it is necessary to choose as
great a clamping force as the conducting wire can endure without
losing the necessary tensile and bending strength outside the terminal 2 •
The thinnest aluminum wire in the table, for instance, did not endure
more than P "" 300 N.
Hardly any influence of the increase of temperature is noticeable.
This experience is based on experim.ents with contacts at high temper-
atures carried out during a year. On copper contacts (cleaned with
P.Otton) at 150 °Ü and P = 350 N, only a decrease of the resistance
followed; on aluminum contacts (oil) at 150 °0 and P = 50 N an in-
crease of the resistance of only about 10% was observed. A varying
temperature impairs a contact more than a constant temperature,
equal to the crest of the temperature variations.
Among observations on aluminum in Germany we mention those
by RICHTER, together with ScHADE3 , and those of GEBAUER [1] and
VIEHMANN [1] and [2] should be noted. The first mentioned investi-
gators found that screw contacts which were good during the first year
remained fit for use for 20 years more. Rivet and twist contacts on the
other hand were unreliable. The crimping method which VIEHMANN
recommended for stranded aluminum wire is widely used in pressure-
type crimps and also for terminating either solid or stranded copper
wire. For instance, the strands are enclosed in a metal sleeve, preferably
of copper, on which a Iasting pressure is exerted. Since the wires can
not escape, strong plastic deformationsandcold welding appear in the
mutual contacts. Of course, the sleeve material must also yield in the
contact spots in order to provide a good connection. Wires of alloyed
aluminum that creeps less than the pure material are suitable. Not-
withstanding the diminished creep and the cold weid, the use of spring
washers or similar elastic joints procuring a Iasting pressure is recom-
mended, since a disturbing gradual yielding will always exist.
E. Medium and low Ioad contacts in air. Printed circuit boards. In
early intercommunication switches, base metal contacts of beryllium-
copper, phosphor-bronze, nickel-silver, have been used (many still in
use). In Germany, much bronze slider on brass lamellas were used. All
were disposed to tarnishing. However, the relatively high load and
1 Cadmium behaves Iike zinc.
z Experience has shown that a three to ten times greater contact Ioad is needed
to give fairly clean, fiat contacts the same conducting surfaces as are obtained with
wires.
3 RICHTER [1]; RICHTERand ScHADE [2].
high voltage, about 1 N and about 50 V, sufficed to produce a-spots

either mechanically or by frittings.
For better performance, one resorts to noble metals. Silver the
cheapest of them, has the disadvantage of tarnishing in a sulphur con-
tainingatmosphere. Palladium, the next in price, is resistant to sul-
phur but it catalyzes the formation of polymers when sliding in atmos-
pheres containing organic vapors; see § 25. As a compromise, alloys
of silver with 30 to 50% palladium are expedient 1 •
In modern telephone techniques the requirements have become
morestringent as the trend goes toward the use of miniaturized multiple
contact connectors and printed circuit boards. Here load and voltage
are too low to secure a-spot with base metals. With respect to costs,
one is restricted to noble-plated material.
The Bell Telephone Laboratories 2 have conducted experiments with
various platings exposed (unmated and sheltered against dust) 1/2 year
to industrial and maritime atmospheric evironment. With P = 0.25 N
and contact voltage below 0.01 V, which is too low to rupture the films
mechanically or by. frittings, only the following coatings were found to
be satisfactorily protective :
A 2.5 fJ.m Au plating and the duplex coating of 0.5 !J.m Rh over
0.75 fJ.m Ag. Au over Ni, Rh over Ni and Au over Ag (sulfiding environ-
ment) are unsatisfactory because Ni and Ag migrate through pores in
a thin Au coating.
For higher medium contact forces, the following coatings were pro-
tective: 0.5 fJ.m Rh over 7.5 fJ.ID Ni, 0.5 fJ.ffi Rh over 5.0 fJ.ffi Ag, also
2.5 flm hard gold. The composite coating over Cu requires 7.5 fJ.m Ni
because a deforming Ni Ieads to cracks in Rh. Au-coatings lead read-
ily to galling, and lubrication at least for the wearing-in process is re-
commended; cf. p. 175.
In recent years, new successful plating baths have been developed;
see for instance VINES [1]. Pd-platings are widely accepted when poly-
mers can be prevented. Massive Pd is very soft compared with Pd-plat~
ings (300 D.P.N.); 5 flm thick layers of Pd can be made sufficiently
porefree and crack-resistant. They give a satisfactory resistance against
wear3 •
The structure of the deposits is difficult to control. ANGUS [3] pres-
ents the following picture. The growth seems to start from a series of
nuclei on the surface of the substrate. The unplated spots give rise to
a network of fine submicrosc0pic pores. Their existence is deduced
1 See BoRCHERT [1].
• See U. B. THOMAS [1], [3], and R. G. BAKER [1]. Cf. the valuable review
by KEEFER and GuMLEY [1].
3 Cf. ANGUS [3], [4].
from chemical evidence and from the fact that as the thickness in-
creases beyond 1 [LID, the coa.tings become progressively more protec-
tive. However, when they grow be,vond :~ to 5 [JID, microscopic cracks
very often develop.
F. Medium and low Ioad contacts in sealed chambers. During World
War II, an attempt was made by the Bell Telephone Laboratories to
"achieve the ultimate in relay contact reliability by operating relays in
sealed containers". It was hoped that sealed containers would protect
against undesirable atmospheric conditions and dust. However, the
result was a failure. Corrosion increased and insulating deposits ap-
peared on the contacts, now known to be polymer deposits mentioned
in § 25 and § 49B. The deposits originated essentially from vapors of
insulating mat.erials; also solder flux, which is difficult to remove, caused
trouble.
These disturbances are avoided in the modern construet.ion of reed
relays1 (in German: Schutzrohr-Relais). The reeds are fused into glass
containers where a good vacuum is maintained during the manufacture,
and necessary welds are performed by means of the electron beam
method without solder. The basic idea of a reed relay is illustrated
in Fig. (30.04). The magnetizable reeds 2 have their outer ends fused
into the glass tube and overlap one
or two mm at the inner ends, sepa-
rated with a gap of about 0.15 rum.
They are magnetized by the oper-
ating coil (with 50 to 100 ampere
turns) to a flux density of the order
Flg. (30.04). Reed relay
of0.5 Wbfm 2 • The ends then attract
each other and make contact in a few ms. Wben the coil current i::;
interrupted, the elasticity of the reeds snaps open the contact.
It has been necessary, at least in present manufacturing, to cover
the contacting areas with a metal of no or little tendency to tarnish such
as gold or wolfram. In order to secure a reasonable heat dissipation, the
container is filled with an indifferent gas (N 2 plus a few % of H2 at a
pressure of about 1 atm). Nevertheless, the maximum current is only
1 A. The contact producing load is 0.05 to 0.2 N. The life of reed r elays
with arc queuehing can reach 105 to 109 operations 3 . It is usttally limited
by material transfer with interlocking; sec § 65.
About constructions with magnetically biased relays, assernblies
ofrelays etc. see HowGAARD et al. [1] and DuMBAULT [1] p. 80.
1 SeeHOWGAARD etal. [l],DUMBAULD[l], WAGAR[l] and WoLAK[l], REUCH [1 ].
2 An iron-nickel alloy that can be fused to glass.
3 ÜETTL [1] gives an interesting review of the number of Operations reached
with different contact types.
When considering microcontacts, notice that the range of P places

the reed contacts on and below t.he border between low Ioad and
microcontacts.
G. Microcontacts, particularly with very low voltage, less than
0.05 V and P smaller than 0.2 N. In the presence of disturbing tarnish,
no metallic portion can be generated in the contacts, neither mechanic-
ally nor by frittings. The contact surfaces must be initially clean, which
means noble meta! contacts if a neutral atmosphere is not provided.
Frictional polymer deposit must surely be avoided, just as was re-
quired for low Ioad contacts : cf. the discussion on polymers in § 25.
Here we concentrate our interest on another disturbance, namely on
the influence of small particles which may be invisible for the naked
Pye but which frequently are present, for instance, air-borne dust parti-
cles.
A microcontact. with P below 0.2 N can be kept open by such a
:o:ingle particle; cf. WILLIAMSON et al. (1).
Fig. (30.05) represents the analysis given by WrLLIAMSON et al. A
dust particle is pictured between the contact members I and II. In
t.he upper figure, the member I (solid line) ,
considered as having a hard spherical sur-
face, is carried by the area A on the top of
particle C, and no contact exists between
I and IJ. The dashed line pictures the
member I in such a position that it is still
partly carried by the particle but has
access to a feeble contact at B . AB is the
radius of the "region of infl.uence" of the
particle. Within this region no contact
Fig. (30.05). Effect of dust ean appear. But with member I moved
particle in a contact. Vertical
scale is exaggerated slightly to the right a full contact would
develop at B .
So far, it was assumed that the Ioad P has deformed member I
(and the particle) only slightly at the contact spot. The area A of the
spot is in accordance with
P;:::;A.H (30.06)
where 11 is the hardness of the :softer member. Eq. (30.06) holds true
provided that the deformation is "deep plastic"; cf. § I. The maximum
A will essentially equal the cross section AP of the particle.
The lower figure pictures the event that the Ioad is able to deform
the member I as to produce contact areas A on the particle and A 1
with member II. Hence
P""' (A ·· A 1 )1l (30.07)
where now H may refer to member I. Tlw particle can not prevent thc
contact at A1 when the load Pis great enough to satisfy (30.07).
We conclude from this discussion that a partide with the cross
seetion AP is likely to cause open contact if
p < APH l
but will not be able to do it if ~ (30.0S)
P>APII J
Fig. 7 in WILLIAMSON et al. [1] shows that Eq. (30.07) is essentially
true. In practice ll is of the order of 109 N jm2 and dust particlesl have
AP between the limits 2 · 10- 12 and 2 . 10-lo m 2, corresponding to about
2 · 10-s < P < 0.2 N l

or in other units (:~0.09)
0.2 < P< 20g j

Meaning: with P > 20 g, a contact is not likely to be kept open by a
single dust particle, but with P < 20 g this can happen. That is the
reason why we let P = 20 g differentiate the contact classes for
micro- and low-medium load. The significance of P = 20 to 30 g
for the surety of contact make has been known for a long time; cf. for
C'xample R. HoLM [37] p. 392.
It might be questioned whether a special name should be given
to contacts with extremely low load; for instance when P ~~ a fraction
of a gram. This seems nottobe advisable because all microcontacts re-
quire noble metals or vacuum or an indifferent atmosphere independ-
ent of P in the range below 0.2 N. Only the maximum current must be
adapted to the load so that the contact voltage remains below the soft-
ening limit. Otherwise considerable deformation could be producPd,
jeopardizing the adjustment of the contact.
· We now pass to particles that do not come from the air. ÜHAIRIN
[ 1], [ 3] has evidenced that during machining and incorporation of the
parts in the relay, base material particles can be transferred from the
tools to the contact members (cf. Fig. (28.06]). He uses sensitized test
paper pressed against the surface concerned to show the presence of
the transferred particles; see for example Fig. 1 in ÜHAIKIN [1].
The insulating power of such deposite should not be tested by slid-
ing the wire probe2 along the surface because it can easily gather several
disturbing particles and experience a higher insulating power than
could derive from one particle.
1R. J. HAMILTON [1], examples also in PFU:l'l'ERER et al.[l].
"The probe is described in § 90.
174- Stationary Contacts
When small particles are very numerous, the discussion referring

t.o Fig. (30.05) may fail to describe their influence on contacts because
many particles may cover the Ioad bearing area and tagether carry
the Ioad. The tumble-polish that is connected with electroplating seems
likely to produce this kind of contamination; see, for instance, LISH [ 1].
H. Stored lubricated contacts. If open contacts of base metals shall
be stored for a long time, and must be available for use at any moment,
it is customary to protect them with a lubricant. Of course, the lubri-
cant must not contain, or in the course of time, form ingredients which
corrode the metal. Thick films of the lubricant protect somewhat better
than thin ones but have the tendency to gather dust and form a dis-
turbant mash. This disadvantage is avoided when a monolayer of the
lubricant is used. Still the protection is considerable as shown in Table
(30.10) based on measurements by R. HoLMet al. see R. HoLM [37],
p. 163 and 164.
The samples, initially practically clean, were tested against a gold
mating member with P == 1 N without impact or sliding, in order not
to darnage the tarnish. The monolayer (sometimes called epilamen) was
placed on the sample in the following way. One drop of a very dilute
::~olution of a Iong-ehain polarized compound in trichlorethylene, carbon-
tetrachloride, or a similar solvent was put on the specimen and then
wiped off. Mter drying, a monolayer remains and sticks firmly to the
sample. The tested epilamen1 was a compound of a Iong-ehain paraf-
finic hydroxan acid with straight-chain paraffinic hydrocarbon, giving
the epilamen a particularly dense structure. It is seen that the final
contact resistances are much greater without than with protection,
which means that the tarnish grew relatively slowly under the epilamen.
However, testing the samplesagainst members ofthe same metal, showed
practically no change during the time of the experiment. This means
that all tarnish concerned was very vulnerable. The table contains data
averages. The spread of the data was great, with a standard deviation
of about 50%.
Table (30.10)
Resistance in 10-• Q
Material without protection with epilamen

against. Au
initial
I alter
'2 weeks I
after
1/2 year I
after
1/2 year
Ag I 1 3 20 3
Cu 10 I 100 >1000 10
Ni 10 100 200 15
w I 3000 I 5000 6000
1 The epilamen was manufactured in the organic laboratory of Siemens & Halske
A. G. Berlin, Germany.
With an epilamen formed of stearic acid, a resistance increase
similar to that noted for epilamen-covered contacts after 1/2 year in
Table (30.10) appeared within 2 months.Thus different lubricants have
different protecting power. This has been stated also by ÜHIARENZELLI
et al. [l] operating with thin lubricating films. Whether they were
monomolecular or not was not discussed. These authors found a mix-
ture of paraffin and syntetic oil to be a promising protector. They made
the tests with sliding contacts.
I. Contacts in measuring apparatus 1 • In this context, it will be
appropriate to mention lever resistance contacts (Kur belkontakte) and
wire rheostats, although they have certain qualities which would motiv-
ate their treatment in Part III of this book. For obvious reasons the
resistance material can not be of noble metals. The good conduction
in the contact has to be secured by the slider scratching the tarnish
film, if the device is used in air without a protecting lubricant. If a very
low contact resistance is required, greasing is recommended. Lubri-
cation should be renewed once a month because the grea;se deteriorates
in the air; perhaps partly because of catalytic action of the metal.
The wear of such contacts is of interest. Typical data are given in
Table (30.11).
Table (30.11). Typical numbers for sliding contacts in measuring apparatus. Both
wear, expressed by 2 Z, and hardness Hin 10 8 Nfm 2 refer to the softer member
Rotary switch I p R
z
slider 1 stationary contact 1 N to-a a
Bronze or silver Brass or silver
H=8 H= 10 10 0.3 0.8 greased
11 = 8 H= 10 10 0.3 16 dry
Wire rheostat
slider 1 wire
-Cu, 2.5% Be Constantan

H = 25 to 33 I H= 14 30 6
slightl.v
greased
Terminal contacts for instruments need ruhhing closure, which

means a certain wear. W ear shall he of the small, not of the micro-type,
nearly of the order noted in Table (30.11); cf. BAYER et al. [1].
J. Contact assembly. The method of contact assembly is an im-
portant factor in the final performance of electric contacts. It would
require a large chapter to give a survey of the many methods of rivet-
ing, brazing, welding, etc. However, these are manfacturing methods
and lie beyond the scope of this book that is devoted to explain contact
phenomena. We emphasize the importance of the contact assembly,
1 Cf. R. HOLM [34]. 2 As to Z see Eq. (41.03).
but confine the information to recommendation of relevant Iiterature:

KEIL [4) p. 280, KEHOE [1], MüüRADIAN (1], and GRABBE [1] for a
special new method to interconnect a network of electrical contacts.
GRABBE who constructs printed circuit boards, found the following
method: The holes and the pins to be inserted are overplated with
copper and a thin layer of indium. During heating beyond the melting
point ofindium (156 °C} but below a temperature that would darnage
the fiberglass board, an alloy between copper and indium forms that
serves as a solder 1 . The holes are wetted with a suitable flux.
Also ultrasonic welding seems to work without heating the parts too
much; see publications by Sonobond Corp.
§ 31. Dimensioning a contact with respect to its heating

The underwriters' rules for dimensioning electric line contacts are
usually founded on the principle that the RI 2 heat must not raise the
temperature of the contact beyond a certain limit. Fixing this limit is a
problern in itself. Stationary contacts can endure 100 to 150 °Ü for a
long time if they are good from the beginning. In practice the tempe-
rature limit is usually lowered to gain greater security in the event the
contact connections are poorly made as well as for other reasons. We
shall discuss the suitable temperature limit at the end of this chapter
but now only remark that it would be possible to increase the tempe-
rature limits which are prescribed if said connections were made with
greater care.
For the moment, we shall assume that the temperature limits used
in practice are given, and dimension a bus-bar contact according to
the requirements. We shall illustrate the principle of such a calculation,
assuming that the RI2 heat produces exactly the permitted increase
of temperature.
Presuming the current tobe 1000 A, we will design dimensions for
the conducting bars and their contact under the following assumptions.
\ In stationary air the free parts of
f'------b-!:1_ _.1 1.. ____7 ---"-~ the bars shall not ha ve more than
0 d:c x-
Fig. (31.01). Bars butt-jointed at x = 0
fA = 35° above the ambient tempe-
rature. In the apparent contact area,
Aa, {}2 = 35° may be added. The temperature of the end surfaces of
the contact constriction is regarded as equalling the temperature of
Aa, and the temperature in the conducting surface, Ac, is go higher.
Even though the temperature 6 is actually not prescribed, we shall
make certain calculations with the sum {}1 + {}2 + e.
1 Cf. e. S. patents 3,175,181 and 3,233,034.
Fig. (31.01) illustrates an assumption simplifying the calculation.

According to this figure the bars do not overlap but are butt-jointed at
x = 0. Let (! be the electric resistivity, A. the heat conductivity, b the
breadth and y the thickness of the bar; h is the coefficient of the heat
transfer to the surrounding air which is presumed tobe stationary. To
obtain the differential equation we consider an element dx of the bar
having the supertemperature, {}, with respect to the surroundings. The
heat, {}h 2b dx, flows out sideways over the surface 2b dx (the small
edge surfaces are neglected). In the stationary state this heat must be
d2(9
balanced by the excess A. b y d x2 d x of the inflow along the bar over the
corresponding outflow, together with the JouLE heat ebl2 dx that is

y
generated in the element b y d x. Consequently we obtain the differen-
tial equation
d2{} (!
A. b y - - 2 h b {} + -- 12 = 0 (31.02)
dx 2 by
A boundary condition stipulates that, at the joint, half the heat

developed in the contact resistance R flows into each bar. This is mathe-
matically expressed by
[dfJJ
-Aby- 1
=-Rl 2 (31.03)
dX X= 0 2
Eqs. (31.02) and (31.03) define the distribution of the temperature
along the bar. Owing to the linearity, Superposition is valid, and we seek
the solution in two steps. First we disregard the heat generation in the
2
joint, and have : : = : ; = 0. Eq. (31.02) then is simplified to
f}l = 2h~2 y ]2 (31.04)

As we intend to calculate numerically it is necessary to know the
value of h. For the cooling of a metal surface in air with the (tangential)
.
velocity v mfsec and wlt h h m . 10-4 watt h
m2 oK we ave
1
h = (7 + 400v) (31.05)
Since one centimeter is about the conventional thickness for bus-bars
we assume this dimension for y. Inserting into Eq. (31.04), J = 1000 A,
(! = 1.8 . 10- 8 Q m, valid for copper, y = 10- 2 m, {}1 = 35°, and with
v = 0, one obtains
b = 0.061 m = 6.1 cm
This is the order of magnitude adopted in practice.
1 Cf. e. g. Handbuch der Experimentalphysik, Bd. IX, Teil I, p. 311, Leipzig
1929.
1:.! Elertric Contacts, 4th Ed.
Now we turn to the distribution of the temperature around the

joint. The influence of the JOULE heat, ~ ~2 d x,
generated in the bar it-
self has already been taken into consideration. Thus we omit the term
bey J2 and solve the remainder of Eq. (31.02) together with (31.03) for
the additional temperature increase {} 2 due to the heat flow from the
contact, and we obtain
{} (x)--
1 RJ2
e
V
-X
'2h
:;;:-
Y
2 - 2 V2hAb2y
Consequently for x = 0, the contact surface,
{} (0) = 1 RJ2
2 2 V2kAb 2 y
The quantity {}2 (0) was assumed to be 35°, and inserting the other
. 1 va1ues an d 11.,
numeriCa =
380
m watt
oK , t h e va 1ue c~or copper, we ob ta~n
.
R = 3.1 · 10- 5 Q
R defines the quality required of the contact; i. e., the cleanness

of the contact and the contact load must be chosen such that R =
3.1·10- 5 Q. Weshallreturn to this condition later.
First we deal with the problern of 6 which is the supertemperature
over distaut portions of the members in a metal contact with the re-
sistance R, when the contact carries a current I, i. e., has the contact
voltage Rl. For the calculation of 6 we use Eq. (13.08), simplifying
by regarding (! and Aas constants, viz.
U2 = (Rl) 2 = 8gll0 (31.06)
Substituting the data of the example we find
e= 17.6 centidegrees
For the modification ofthe l]l{}-relation in the event of quasi-metallic
a-spots see § 15. Since the true contact area is of the order of 10-6 m 2 ,
or more, the error in e does not exceed 1% if films with a tunnel resis-
tivity of less than 10-12 Q m 2 are present. Thus the supertemperature
of the conducting contact surface in relation to the surrounding air is
The required contact resistance, R = 3.1 · 10-5 Q is obtained by

choosing a suitable contact load P. The relation between P and R is
evident from Fig. (8.01), though not quite for the kind of contact and
low degree of cleanness of the surfaces which we now assume. In Fig.

(8.01) a curve is drawn for nominally flat clean copper contacts and
a similar curve for clean copper rod contacts. A clamped bar joint is
something between the extremities of flat and rod contacts. Its load
is concentrated more than in the case of nominally plane contacts but
less than in crossed rods. In the case of bars with metallic clean surfaces,
the curve at P = 10000 N would lie at about 3 · to-s Q. However, con-
sideration must be given to the disturbing films on the surfaces, which
carry the greatest part of the contact load; i. e., the care with which
the bars have been cleaned before joining must be taken into account.
If we raise the curve at P = 10000 N to 10-6 Q, it will correspond to
easily mounted contacts. Assuming the slope of the R U-characteristics
tobe 45°, we read P = 3000 N at R = 3.1 . 10-6 Q. For practical pur-
poses it is conventional to raise P another power of ten. If this were
done solely with regard to the contact resistance, it would mean that
the mounting standard of the bar contact would be very low, since
P = 3000 N would be sufficient for tolerably cleaned bans. But in prac-
tice, other circumstances play a part, which we have not taken into
consideration. The contact must have a certain mechanical strength
to prevent loosening by various strains causing brt'athing. Breathing 1
leads to oxidation, thus increases the resistance to an indeterminant
degree and causes unduly high temperatures. This shows that the
practical rule for the temperature fh + {} 2 is based on very intricate
circumstances. Of course, clearer information than furnished by the
temperature measurement would be obtained by the determination of
the contact resistance if this can be arranged.
We now return to the calculation of e, which played a secondary
role in our example. lf, instead of being thin bars, the members con-
sisted of heavy and well-cooled masses suchthat the heat produced in
the contact does not change the mean temperature in the end-surfaces
of the current constrictions, e
will be decisive for the temperature in
the contact. It is instructive to calculate the above example with
I = 1000 A and with a R P-characteristic (see Fig. [8.01]) with a slope
of 45°. Beginning at the point P = 10000 N and R = 10-6 Q, suppos-
ing that we have only to take Eq. (31.06) into consideration. The super-
temperature of the contact in relation to the surroundings may still be
allowed to be 87.6 centidegrees, i. e., e
= 87.6°. Substitution in Eq.
(31.06) gives
(1000 R) 2 = 8 · 3.8 · 1.8 ·10-6 • 87.6 Q 2
thus
R = 6.9 · 10-5 Q
1 Breathing is described in § 22 G.
12*
We readfrom the assumed RP-characteristic a corresponding con-

tact Ioad P of 1600 N; but in the case of a clean contact it would lie
lower by about two powers of ten .
§ 32. Contact effects in carbon granular ~icrophones
A. lntroduction. The carbon granular microphone, or transmitter,

is universally used in telephone communication 1 . Other microphone
types used for radio, loudspeakers etc. provide much higher fidelity
to the audio waves, but they require costly additional amplifiers. As we
shall see, the carbon transmitter itself amplifies the sound energy by
a factor of several hundred. It has the great merit of low cost.
Fig. (32.01) shows the basic principle of the transmitter. Sound
waves entering through openings in a protecting cover strike a mem-
brane that conveys a varying pressure to the
Membrone assembly of carbon granules. These carry the
r current I from the membrane to the base car-

fhlleclion bon electrode. An average hydrostatic presstue
between the granules is determined by the gra-
Eleclrode
vity of the grain assembly in a way similar to
that in a liquid . The sound causes superimposed
pressure variations which vary the resistance
Fig. (32.01). Basic prin- R ofthe microphone. This produces an alternat-
ciple of the transmltter. ing modulation (amplitude L1 J) of the current
Not copied after any par·
tict!lar model which is driven by a voltage V applied at the
microphone. The transmission power is V · L1 I.
Particularly because the factor V can be considerable, this power is
much high er than the modulating so und power; see Section C.
B. Confl.guration of the transmitter. ]'ig. (32.01) indicates a some-
what complicated configuration of the electrodes. Several configura-
tions are in use; all aimed at making the function of the microphone
independent of its space orientation. The infiuence of resonance in
the membraneandin the cavities is also diminished. The modern micro-
phone has a nearly constant semdtivity over a broad band offrequencies.
An instructive, historical review of how the sensitivity spectrum of the
microphone has been stepwise improved is given in DEHN and LANGs-
DORFF [J).
Modern microphones have a fairly uniform spectrum between 250
and 4500Hz. This range is sufficient for the formants of the vowels 2 •
1 For the history of telephony and microphones see GoucHER [1] , GREEN [1].
and DEHN and LANGSDORFF [1].
2 The formants of the sound i (ee) reach up to about 3700Hz.
§ 320 Cont.act effects in carbon granular microphones 181
The fact that it cuts off consonant formants between 4500 and 10000
Hz does not render the speach unintelligible
C. Calculations of the a. c. power of a microphonc. Thorough calcu-
lations referring to published extended invest.igations of German micro-
phones (as used in the 1930's) are given by R. HoLM [37]0 The literaturf'
about modern microphones is relatively meager in this respecto We
t.herefore confine the treatment to rough approximationso For example,
we use the average hydrostatic pressure P and its variation LI P in
that part of the assembly of grains where the main resistance of the
microphone is located, and do not apply formulas expressing that
grains in a distance from the membrane experience smaller impulses
than those adjacent to the membraneo We are forced to form conjectures
in two points, where clear measurements are missing: 1. the fraction
ofthe sound pressure force on the membrane that is transferred to the
grains; 20 the influence of resonanceo However, since the assumptiont:
Iead to the same ao Co power of the microphone as measured, we may
eonsider the condusions reached by conjecture as essentially confirmed_
The calculations refer most closely to a Siemens microphone bul
apply in principle to allmicrophones now in common usageo
The grains (average diameter about 0017 mrn) are obtained by
earbonizing crushed and sieved anthraciteo There are some 6 graim; in
series in the active region that has a cross section of about0o6 cm 2 ver-
tical to the currento The hydrostatic pressure of this grain assembly is
circa P = 0.4 gfcm2 =- 40 Nfm 2 giving any contact between two graim: a
constriction resistance of circa 4 ° 104 Q and for the ent.ire microphone
t.he resistance of about 125 Qo
f' 'J.
conilensel'
fööoooo· choke
=t
.Fig. (32o02)o Wiring diagram for a micropbonc Figo (32o03)o Wiring diagram for a mlcrophoue
circult with local battery circuit with central station battery
The resistance of 4 104 Q per constriction corresponds to a curve

°
lying about a power of ten above that for baked carbon in Figo (8001)0
In order to assess the energy that the rnicrophone delivers to the
eornmunication circuit (electrie Ioad) we discuss Figso (32002) and
(32003)0
Microj>hone circuits belong mainly to two types illustrated in .Figso
(32002) and (32o03) where R is the microphone resistance and R 1 the
Ioad resistanceo The terminal voltage across the microphone may be U,
and LII, LI R etc. denote variations caused by the sound waves. We

assume the variations signified by LI in terms of percentage small
enough tobe treated as differentials.
In the local battery circuit shown in Fig. (32.02) the basic current
I=~E~ (32.04)
Rz+ R
is superposed by the audio current with the amplit.ude
LJRE
LI I= (Rz+ Rf (32.05)
The Ioad resistance R 1 consumes the power
1 R ( A I) 2 _ 1 J2 Rz ( R) 2
(R 1 + R)2
A
2 1 LJ - 2 LI
which, with R = R 1 , attains its maximum value

J2(LJR)2
(32.06)
--sR
for given I, Rand LIR.
In the central station battery case shown in Fig. (32.03) where the
impedance of the choke coil is dominating, an essentially constant cur-
rent :flows through this coil. Let the terminal voltage be U + LI U. Then
I=U =U+LJU_LLJu_ (32.07)
R R + LJ R ' R,
giving
LJR
LI U =I 1 + R/Rz (32.08)
and the current in the Ioad
The Ioad resistance consumes the a. c. power
which with R = R1 for a given I, R and LI R, attains a maximum that

is expressed by the same equation (32.06) that was derived for the
local battery circuit.
Now a major task is to find LI P and LI Ras functions of the pressure
variation Llp of the sound wa ves that hit the membrane. W e assume LI p =
10 fLbar1 = 1 N /m2 which corresponds to the normalsoundspoken into the
1 One 1-1-bar is one dyne per cm 2 = 0.1 N/m 2•
§ 32. Contact effects in carbon granular microphones 183
micropbone. Tbe pressure tbat is transferred from tbe membrane to

tbe carbon grain assembly is greater tban Lip because tbe sound force
on tbe total membrane is conveyed to grains by its small central part
tbat alone bas contact witb tbe grains. Tbe value of tbe ratio in mod-
ern micropbones bas not been determined, at least not publisbed. Re-
ferring to R. HoLM [37] p. 178 and applying a certain correction we
assume tbat tbe sound would produce
N
L1P=4L1p=4-
m2 (32.09)
However, tbis L1 Pis calculated witbout considering any resonance in

tbe membrane. Tbe resonance is small in modern micropbones, as is
evident from tbe broad band of tbeir frequencies. Yet resonance exists,
otberwise tbe band would not be limited. It seems reasonable to assume
a resonance magnification of 1.5, cf. R. HOLM [37] p. 179. Tben Eq.
(32.09) is altered to
(32.10)
The next task is to find L1 R tbat corresponds to L1 P given in Rq.

(32.10). Fig. (9.04) would directly provide tbe answer if tbe material
of tbe micropbone grains were tbe same as tested for Fig. (9.04). Yet
it is barder and bas 10 times tbe constriction resistance for a given
load. Fig. (9.04) indicates tbat tbe curves become steeper witb barder
carbon. It gives for tbe lower curves
R = const p-0 ·45 or LI R = - 0.45 LI P

R p
and for tbe upmost curve
LIR=_ 06 LIP
R . p
Since tbe granules are very bard it is reasonable to assume

LIR LIP
7r= -o.7-y (32.11)
or witb R = 125 Q, P = 40 Nfm2 and L1 P = 6 Njm 2

L1 R = 13 n (32.12)
Inserting in Eq. (32.06) witb I = 0.035 amp, tbe normal current,
\ve obtain an effective power of
12 ~RR) 2 = 2.07. 10- 4 watt (32.13)
Measurements give on an average 2.17 · 10- 4 watt confirming tbe calcu-

Jations.
The compliance in the contact between two grains is surprisingly

small, approximately 0.3 A under the circumstances here considered,
according to calculation in R. HoLM [37] p. 182 and 183 on the basis
of measurements by GoucHER [ 1], HAYNES [1] and SALINGER [1].
D. Deterioration by "burning" of micropbone grannies during tbe
life of tbe transmitter. The granules are of selected anthracite, crushed,
and then "roasted" at about 1100 oc in hydrogen. The surface .layer is
barder and less porous than the bulk, although perhaps not typically
pyrolytic, as was assumed earlier.
GRISDALE [1], [2] investigated specially deposited pyrolytic layers
(see R. HOLM [37] p. 176) and the normal "burning" of microphone
granules. He found that, if the single contact voltage e.xceeds a critical
value, the contact resistance rises and the contact becomes noisy. The
eritical voltage lies between 2 and 3 V for a single heating; repeated
heating at lower voltages seems to produce the same effect.
In older transmitters the hydrostatic pressure in the grain assembly
was much dependent on the orientation of the transmitter; and it
easily happened that contacts with too small a Ioad became heated to
"burning". Modern transmitters are so constructed that variation of
their space orientation causes only small variations of the hydrostatic
pressure. However, in order to gain desired magnifications, also they op-
erate so elosetothe critical voltagethat the grains deteriorate with time.
GRISDALE [2] associated the burning with oxidation. MACKLEN [1],
[2],[ 3],[ 4] has thoroughly investigated tbis effect.The oxidation manifests
itself in weight increase and in increased contact resistance, which at
530 oc rises to a sharp maximum. On granu]es roasted at 900 oc the
resistance reached powers of ten above the normal resistance. At still
bigher temperatures, particularly above 1100 °0, de-oxidation causes
the resistance to decrease again. (It is conceivable that the resistance
increases because the oxide provides traps for tbe conducting electrons.)
The modulation efficiency ofthe granules deteriorates. HIRABAYASHI
et al. [1], and MACKLEN [3] discovered and studied a purely mechanical
deterioration of the modulation efficiency. This effect can be produced
by tumbling the granules.
The oxidation effects become relatively small if the carbonization
temperature is 1100 oc or higher. With respect to the adaptation of
the transmitter resistance tothat of the line (see Eq. [32.08]) MACKLEN
proposes a high baking temperature, perhaps 1350 °0, and an oxidizing
treatment prior to the filling of the microphone.
The chemical character of the oxidation and its infiuence on the
electricaHy conducting:n: electrons have been studied recently by several
workers. However, their studies lie beyond the scope of this book1 •
1 Relevant references are cited in UBBELOHDE and LEWIS [1].
§ 33. Contact with semiconductors. Rectification. Static electrification 185
§ 33. Contact with semiconductors. Rectification.

Static electrification
A. Introduction. Rectification effects are the most important topic
of this chapter. The contact between a meta} and a semiconductor is
responsible for the rectification in Cu 20- and Se-rectifiers. Although
these devices are not very old, they have given way to other types.
Only the Se-rectifiers preserve a limited a pplication field of their own 1.
Their semiconducting layers are polycrystalline and do not provide
current density and back-voltage ratings (Sperrspannung) as high as
do the monocrystalline silicon rectifiers in whose interior, p-n junctions
are the rectifying agent.s. The latter effects do not belong to t.he theory
of contacts and t.herefore are not treated here.
We confine this chapter essentially to describing the principal fea-
tures of a rectifying metal-semiconductor contact. Although the p- n
junction does not involve contacts, it is made available for some re-
ferences by means of a short description in Section C. A description
of the transistor treated in an earlier edition, is now omitted.
In all events, we give precedence to n-conductors 2 and treat n- a1Hl
p-conductors as if they were doped solely with donors or solely with
acceptors, thus being pure n- or p-conductors. In practice, an n-con-
ductor can be generated by diffus-
ing donors into a conductor that
has acceptors. ,I\
B. Contact between a metal and ~\
, \
a semiconductor, and its rectification \ ', b V
property. In Fig. (33.01) the solid \ '-+--- -0.1
curve represents the potential distri-
hllt.ion around the boundary when no \
net current passes it (equilibrium). \ 0
1 J 5'
This figure is essentially a copy of · , . fo - 'cm
'""'+·-
the upper portion of Fig. (11,14),
MefrJ! I J'cmi-COfldi/CfOI'
• +0.3
the only difference being a certain
spacial extension (exaggerated) of Flg. (38.01). Potential dlstribution (negative
potential upwards) around the boundary
the potential hill on the side of the between a meta! and a semlconductor. Solid:
metal to remind the reader that the equillbrium.Dashed: bias in torward direction.
Dash-pointed :bias in backward direction
slope in question is finite.
The FERMI Ievels in the metaland semiconductor coincide at equi-
librium . As is explained in § II, the slope of the potential hill is rela-
tively gentle within the semiconductor and therefore the breadth of
1 See H. MILLER [1].
2 Similar features are met with n- and p-oonductors; cf. § II.
the hill is considerable on that side, 3.15 · 10- 5 cm according to the

figure 1 .
We now assume that a negative voltage, V = -0.3 V, is applied,
lifting the FERMI level of the semiconductor in the negative direction
(- 0.3 V) above the FERMIlevel of the metal2 • The bottom of the con-
ducting band is lifted by the same amount, as is indicated by the dashed
curve in Fig. (33.01). This roeans that the shape of the slope is changed
in such a roanner that the field strengthin the barrier region of the semi-
conductor has becoroe weaker; or in other words, a sroaller nurober
of ionized donors accoroplishes its generation. Since the density of the
donor ions reroains approxiroately constant, the diroished nurober roust
be corobined with a sroaller thickness of the barrier layer. Applying
Eq. (II,23) we find that its width has decreased from 3.15. 10-5 to
2.57 . 10-5 cro. On the other hand, the application of V = + 0.3 V
lowers the conducting band and broadens the hill to 3.64 . 10- 5 cro, as is
indicated by the dash-dot curve.
In equilibriuro, there is a sroall electron current density J. olimhing
over the boundary hill <Pc in the roetal and a siroilar opposite current
density in the seroiconductor diffusing uphill because of the teropera-
ture. The latter current consists essentially of extrinsic electrons froro
the conducting band. W e neglect the sroall nurober of intrinsic carriers
which are generated in the seroiconductor within the barrier.
We obtain a suitable expression for J. by considering the electron
flow froro the roetal. Let J 0 be the current density of all elec.trons in
the bulk of the roetal, which have a velocity coroponent versus the
boundary. Aroong those only the fraction
J. = J 0 exp (- 11600<Pc/T)
is able to climb over the hill, and in equilibriuro an equal current
J. leaves the semiconductor. With a forward bias, V, the hill height
in the semiconductor is <P0 - V and now the current density entering
the metal is
J 0 exp[ -11600((1),.- V)JTJ (33.02)
whereas the current froro the roetal reroains = J •. Thus, the aroount
of the net electron current density is
J(l')=J0 exp[- 11; 00 ((1)"-V)]-Jc l

~J,exp(- u;"'<P,) [exp(u;"' v) -!Jl
(33.03)
1 The figure refers to the same numerical assumptions as have been used for
Fig. (11,22).
2 The lifted FERllfi Ievel is not marked on Fig. (33.01)
A back voltage V changes the factor in the last bracket to

11600 )
1- exp ( - T -V
Example. With T = 300 oK and V = 0.3 volt as in Fig. (33.01) t,he
ratio between forward and back eurrent is a strong n'ctifieation:
11600 )
exp ( ~0.3 -1 .
J(-0.3)
J(0.3) 11600 ""' iO·'
1 - exp ( - ~ 0.3)
Films of semiconducting materiab a few 100 A or less in thickness1

have no bulk or too little bulk for the application of Eqs. (33.02) and
(33.03). The high resistivityfilm on commutators and sliprings of copper
are of this kind and normally are not more than 100 A thick, usually
much less under the anodic brush (individual track). For these films
Eqs. (33.02) and (33.03) cannot be used as KuczoGI [1] does it.
Within the region of the diffusion current the density of electrons
issmall compared with the density in the bulk of the semiconductor.
Therefore ScHOTTKY called this region an exhaustion region, a word
weshall employ.
We have now described in principle the (diflusion) rectification
effect of one contact. In service the semconductor will have a second
metallic terminal that also exerts some rectification The rectification
of the entire unit is then brought about if the effects at one terminal
are greater than at the other. The reason for this difference may be:
1. One terminal electrode has a greater work function than the other;
2. The density of the donors (or acceptors) is different on both sides;
3. The contact areas between semiconductor and terminals are of diffe-
rent sizes.
It is easily seen that the blocking or· reverse direction of a rectifying
contact always coincides with the direction for moving the majority
current carriers from the metal to the semiconductor. The opposite
direction is called forward direction or direction for easy flow. Thus, at
a p-conducting semiconductor, the blocking appears with the positive
current entering into the semiconductor.
C. The p-n-junction. Imaginc a p-conductor in contact with an n-
conductor. Their barriers may fit together in a way similar to that
shown in Fig. (33.04). The combined voltage barrier pushes holes from
the n-into the p-conductor, and electrons in the reverse direction. Thus,
the majority carriers, holes from the left and electrons from the right,
can move only by diffusion by means of their temperature energy. An
applied voltage that lowers thE' barriPr increases the diffusion current
1 Cf. HALPERN fl].
and rectificat.ion is produced just as discussed above with reference to

Fig. (33.01 ).
In practice p-n- junctions are not made as contacts but constitute
the boundary between a p- and a n -d oped region, within one single
crystal. For example: a plate of silicon may be provided with an alu-
minum electrode on one side and an antimony sheet on the other side.
Then, the whole assembly is heated
I
tf1egolive
polef11i'al
so as to canse atoms of the electrodes
t.o diffuse into the silicon lattice,
I wherey the aluminum atoms become
acceptors and the antimony atoms be-
come donors.
D. The silicon controlled rectifler
SCR (in German: Silizium-Strom-
tor) has 4 differently doped layers
p-n-p-n. The first three constitute a
t.ransistor. The last p-layer is con-
Ä nected to the metallic gate Iead . The
'
p- comluclol' n.-comluclol' voltage applied to it has a triggering
effect, so that the device can serve as
Fig. (33.04). Potential distrlbut.ion
around a p-n-junctlon an on and off switch beside being a
rectifer. The trigger current of 50 mA
applied for only a few microseconds may initiate conduction for any
current from a few mA to hundreds of amperes . The SCR is a kind
of solid state equivalent of thyratrons.
Table (33.05). Properties of important rectifiers1
ReetHier type Se·contact Si-junction
Maximum current density 0.2 A/cm 2 200 A/cm 2

Voltage rating (effective) 30V 400V
Tamperature range -55 to + 150°C -55 to + 175°C
Efficiency 90% 98 to 99%
Relative volume for
same wattage 15 1
The rectifiers work by means of extrinsic current, and do not endure

temperatures that generate intrinsic conduction. This determines the
upper limit of their temperature range.
Data characterizing a few technicaUy important semiconductors
are given in Table (II,24).
1 For details see manuals or handbooks about silicon rectifiers of the manufa.c-
turers: General Electric, Westinghouse, Internat. Rectif. Corp., etc., also Siemens
ZS, vol. 36 (April1962) 211 ff.
E. Remark concerning the contact between silicon carbide crystals.

Investigations have revealed two singularities that we shall discuss.
First, the contact voltage attains a maximum of 50 to 100 V simulta-
neously with a heat glow in the contact constriction. E.HoLM[6] proved
that this is in accordance with the qn?-rt>lation1 ; it is a consequence of
the high values of elt that great contact voltages ofthe order mentioned
are connected with the highest temperature e
(glowing temperature)
that this contact endures. So long as the temperature is of this magni-
tude the crystal material is intrinsically !'onducting and does not ex-
hibit any exhaustion layer.
Second, at low contact voltages a boundary layer has been estab-
lished2 that has no perceivable conductivity ofits own beside its tunnel
conductivity. E. HoLM [6] showed that the observahle conductivity
is a function ofthe applied voltage in agreement with Fig. (26.17), which
proves it to belong essentially to the t.unnel effect; cf. HEYWANG [1].
F. Static electrification3• Though THALES, 2500 years ago, knew how
to electrify amber by rubbing, the explanation of the effect is of a very
recent date. It is simple enough. When two bodies with different work
functions, C/>1 and C/>2 , are brought into contact, the equilibrium state
requires that their FERMI Ievels coincide with the consequence that
their surfaces then attain the potential difference of C/>1 - C/>2 , cf. § IIH.
Insulators will become charged solely in the contact surface, and within
a narrow vicinity of it. Of course, the charged region and the total
charge will increase when the bodies are rubbed against each other,
making contact in ever new areas. This alone, and no new electric effect,
seems to be the reason why the ruhhing intensifies the electrification.
The contact voltage will seldom attain the exact value of C/>1 - f/J2
since many factors are apt to preYent it. It may be that the time of
contact is too small for a maximum effect, or surface currents may
spread out the charges. Still more influential are electric discharges
through the air. These appear between the separating charged members
and Iimit the voltage that otherwise would increase with t.he distance
between the charged surfaces.
A simple computation shows that such discharges are probable, and
their appearance has been proved in many instances. Particularly, the
maximum charge obtainable depends on the ambient gas pressure as
the theory predicts. For said computation we notice t.hat the charged
layer in a semiconductor will have a thickness of the order of 10-5 cm,
cf. Fig. (II,24). If the voltage between the surfaces of the ruhhing
1 See§ 13.
"BRAUN and Busen [1]; JoNES, ScoTT and SILLARS (1]; E. HoL.l\1 (6].
3 See LoEB [3].
bodies is of the order of 1 V, the field strength at the surface of the

non-conductor will be of the order of 107 VJm during and immediately
after opening the contact. Such a field is amply sufficient for producing
an electric discharge through the air in the moment of contact sepa-
ration.
Electrostatic electrification can have an unsettling effect on manu-
facturing processes and may even be hazardous in industrial establish-
ments. Much attention has been given to it in England, and, in 1953,
it. was the subject of a symposium in London. The report1 on this meet.-
ing is recommended for much detailed information on the phenome-
non, both on its use in electrostat.ic machines and on its detrimental
effects.
§ 34. Carbon-pile rheostats.

Eleetric resistance of pressed metal powders
This section deals with resistances between carbon or metal grains
or plates in series, particularly with their dependence on the Ioad that
produces the pressure between the individual conductors. The de-
8Cribed devices play the part of examples.
Variable carbon pile rheostats are generally composed of discs
stacked in series in a refractory tube. The pile may contain 30 to 80
di8cs. The change in resistance is effected by applying a different Ioad
to the column of carbon discs. The normal regulating load range is
bet.ween 1 and 40 N, varying the total resistance R in the ranges from
1.3 > R > 0.05 Q up to 13 > R> 0.5 Q, depending on the carbon
material. Common practice is to allow up to 0.5 V per junction. The
reproducibility of the resistance as a function of the Ioad can be ± 7%.
A stiicter reproducibility, of ± 3% or better, can be achieved by
replacing the discs with windings of a helical spring made of a carbon
strip. By allowing the thickness of this strip to decrease from winding
to winding, more and more windings are short circuited by their mutual
contacts at increasing Ioad. In addition, the contact resistances de-
crease, thus enabling a great range of resistances, for instance, two
powers of ten to be attained.
The carbon contacts do not weid, and after repeated closures the
contact spots may be formed merely elastically, cf. the r-branches in
Fig. (9.03). Non-welding and elastic deformation are the conditions for
a good reversibility of the resistances; in other words, for a fairly single-
valued relation between resistance and Ioad.
1 Static elect.rifieation. Brit. J. Appl. Phys. Supplement No. 2, 1953; seealso
WALKUP [1].
§ 34. Carbon-pile rheostats. Electric resistance of pressed meta! powders 191
The discs appear flat; but, surely, the waviness of the surfaces is a
condition for the variation of the contact resistance in accordance ·with
the Ioad. Dust within the pile must be avoided because it would change
the resistance.
Metal powder at high pressure1 • Imagine the powder pressed by
means of a punch in a cylindrical hole. With rising hydrostatic pressure,
p, the contact areas become !arger, disturbing films are damaged and
the conductivity of the powder increases. This effect usually is partly
compensated for by the breaking of the grains into pieces, thus in-
creasing the number of contact spots in series. At high pressure, the
contact spots will have dimensions of the same order as the cross-
section of the grains. This consequently does not provide enough
volume for long constrictions 2 • ThusEqs. (4.15), (8.07) and (8.08) are no
Ionger valid.
In view ofthe complications arising from the varying size and shape
of the grains, the following computations will be sufficient. As the
simplest case we imagine clean spherical grains (average diameter D)
and apply Problem A of § 4, designating b the radins ofinlet and outlet
contact. In accordance with Eq. (!,17), we suppose b ~ l/pfH. Eq. (4.07)
.
g1ves th e res1s
. t ance m . R = e (b-
. one gram, 1
n
D1 ) wh ereg1s
· th . .
. eresistiV·
ity of the grain; we find the resistance per cm 3 of the powder to be
o
-P
= const [AVIi- _.!._]
p D
= const VH [1- -
p
1
AD
Ii-~]
V H
where A is a constant and the index p refers to powder. The last term
of the expression within brackets is very small. Thus
:p ""const V~ l1 + A~ ~] = const Vi, + /1 (p)
where / 1 (p) is a relatively small quantity.

In reality the grains do not have clean contact surfaces, and with
r
rising p the conducting areas increase at a greater rate than the Ioad-
hearing areas, say proportional to (~ where 1/2 < 8 < 1' instead of
proportional to VJ,. . Consequently we expect
(!~ = const(J,.)"+ / 2 (p) (34.01)
1 The numerical data of this section are given by SKAUPY and KANTORO·
WlCZ [1], HoLM et al. [8] and [29], KANTOROWICZ [1], BRUNNER and HAMMER-
SCHMID [1]. Earlier investigations are cited in these papers.
2 .As for Iong and short constrictions see end of § 4.
where / 2 (p) is a small quantity of the same order as fdP) . In fact Eq.
(34.01) expresses the results of KANTOROWICZ [1], and of BRUNNERand
HAMMERSCHMID [1].
Mter having varied the contact load repeatedly between a high
and a low Iimit, KANTOROWICZ found an increase of resistance of the
powder, which was prob-
Table (34.02). ably caused by breathingl
Ratio of resiativitiea of powder and Bolid material8
of the contacts, i. e.,
in relation to tke kardnea8 Hof tke Bolid material.
D designates average diameter of the grains oxidation of contact areas
to which the air had free
B D
Material 10• Ntmt access during the period of
low pressure and which
Fe 185 14 25 remained covered by oxide
Ni 18 to 150 16 30
layers at the renewed large
Cu 90 6 20
Ag 7 2.5 10 pressure. At decreasing
Sn 1.1 0.3 10 pressure a hysteresis of
Sb 10 to 25 2.5 100 the resistance was obser-
w 114 to 420 20 5 ved. It is explained by
Pt 100 7.5 2
Au 7 to 12 2.5 8 adherence in metal con-
Pb 1.7 to 3 0.3 10 tact points, which was
Bi 8 0.5 50 discussed in connection
Graphite 3.2 1.5 8 with Fig. (8.12). Without
citing further examples, it
seems likely that other observed details can be explained by the
contact theory expounded in this book.
Table (34.02) compares the resistivities (!p of powders with the re-
sistivities e. of the solid metals.
There is an evident parallelism between the colums (!p/(! 8 and H.
Certain deviations are easily explained. The e11 /e, values for Fe, Cu, Sb,
and W have been somewhat increased by oxide films. The platinum
powder had very small grains and therefore relatively large resistivity.
The reason for the relatively small (!p/(!, values for silver and gold is,
of course, the absence of oxide films.
1 See §22G.
Part li
Thermal Contacts
§ 36. Thermal metallic contacts

A. Introduction. As pointed out in § III, the mathematics for heat
flow is the same as for the electric current. The electric concepts of volt
and ampere correspond to temperature and energy flow per sec (watt).
Assuming geometrically equal electric and thermal flow lines with the
electric resistivity (! and the thermal conductivity A. both being constant
within the regarded volume, the relation (35.01) holds bctween the
electric resistance R Q and the thermal resistance W
R °K
W= --
e}. watt (35.01)
and with the WrEDEMANN-FRANZ law being valid
R
W=-
LT
(35.02)
More specifically, when the WrEDEMANN-FRANZ law (see § IIIC)

holds with A. = A.0 , a constant, the thermal resistance is always the
same as at room temperature T0 • Thus, the resistance in one member isl
W' = R (eo Äo) (35.03)
eoAo
The thermal constriction resistance against the bilateral heat cur-
rent resuJting from the heat produced by the electric current through
the constriction. The problern is solved for the case W-F law being
valid with A. = A.o, a constant, as follows.
We apply Eq. (13.13) and
T8 - T 0 = W" · (heatcurrent) = W" R%2A) (35.04)

and find
W" = R(eA) (35.05)
4L(Te + T0 )
1 The prime ' or " refer to different types of heat fl.ow in a constriction, as de-
scribed in the text.
194 Thermal Contacts
and introducing the factor f = R(eÄ)jR((! 0 Ä0 ) in accordance with Fig.

(16.14)
W" = I · R Ü?,, Ao) (35.06)
4L(Te+ T 0 )
In the particular case of a single circular heat-conducting a-spot,
W" = _1_ I· To (35.07)

SaÄTe+To
C. Thermal resistance of nominally ftat bolted contacts. The simi-

larity between the electric and thermal problern is valid for a clean
metallic contact in vacuum, but there is a fundamental difference.
Films and other contaminations insulate more ör less in the electrical
contact, but thermally they produce considerable short circuits in the
surrounding of a-spots, sometimes even dominating. The order of
magnitude of such short circuits is shown in Section D. Approximate
computations will suffice since in practical cases surface asperities, air
gap and other geometrical features never are exactly determinable.
We begin by considering a bolted, quasi-flat clean metallic con-
tact in vacuum. The load bearing area, which here is the conducting
area also, is an assembly of a-spots which are produced with different
pressures p and thus have different specific depths 1 D. In flat contacts,
the average pressure, p, is smaller than the hardness H, namely p
= ~H where 1; < 1 according to Eq. (I,17). Theoretically no lower
limit for ~ exists, and thus the problern of predicting the thermal (or
electric) resistance between nominally flat members seems to be in-
determinate.
Fortunately, the practical problern is a better situation. It has been
found, empirically, that certain resistances are much more probable
than others. In particular, the spread around them tends to smaller re-
sistances, but scarcely higher. In other words, a function exists of the
metal, M, and of the load, P, which defines an upper probable resistance
W(M,P) for any M and P. In addition, the function has the merit of
representing the most probable resistance, as stated within a great
range of experiments. Designers have to calculate with such Iimits.
We write
W(M,P)=f!!(M)·F(P) oKJwatt (35.08)
where (/! is a dimensionless function of the metal M; and F, with the

dimension of a thermal resistance, is a function of the total mechanical
load P independent of M. In vacuum, the apparent contact area A and
1 About "specific depth" see Eq. (I, 12).

§ 35. Thermal metallic contacts 195
the normal roughness of the surfaces have a fairly small infiuence, in
striking contrast to their effect in fluid environment. Only contacts which
have been prepared with exceptional care (not possible in practice) ex-
hibit a thermal resistance that is considerably smaller than according
to Eq. (35.08).
Fig. (8.01) indicates that some metals, within the range of the high
Ioads here concerned, have an electrical resistance for fiat contacts
nearly a power of ten below the resistance of crossed-rod contacts. In
crossed-rod contacts the resistance is proportional to VJi,
where His
the indentation hardness of the metal. We make the reasonable as-
sumption that this dependence on H is generally valid for fiat contacts
also. Consequently, tbe funl>tion q; shall have the factor VHfH 0 where
H belongs to the metal considered and H 0 to a reference metal. Silver
was chosen as refP-rence metal with H 0 = 4.8 · 108 Njm 9 and the heat,
conductivity Äo = 418 wattjm 0 K. The choice is motivated by the fact
that good electric measurcments on silver exist1 • These are readily
converted to the thermal analysis by means of Eq. (35.01).
That q; also contains the factor Äo/Ä is self-evident. Wfl see that it
is sensible to put
Ä
q;(M) = -~ 1-
Ä
·tiH
H 0
= a pure number; or, with the values of Äo and H 0 of silver inserted.
rp(M) = 0.019 v:'

We turn to the function F(P) which we present as an empirically
determined curve in Fig. (35.10). The left part of this figure, for P
up to 100 N, is based on the electrical measurements on silver. The
rest of the curve is based on thermal measurements reported in the
following references:
BoESCHOTEN et al. [1], Fig. 1, aluminum contacts.
CLAUSING et al. [1], Fig. 31, series 2 S and 3 S, stainless steel; Fig. 29,
series 2Mand3M, magnesium; Fig. 25, series 2B and 3B, brass.
FENECH et al. [1], Fig. 15, stainless steel.
FRIED et al. [1], Fig. 9, aluminium.
1 It is remarkable that there is complete agreement, at equal lol\ds and normal
asperity, between the measurements of electric resistances of quasiflat Ag-contacts

by R. HoLM [4] Fig. 12b, BuRKHARD[l] Fig. 5 and HARADA a. MANo [1] Fig.12
although the apparent oontact areas certainly were considerably different.
13°
FRIED [2], Fig. 6.1 lower curve, stainless steel; Fig. 6.3 average, alu-
minum.
RüGERS [1], Fig. 3 upper curve, aluminum.
SHLYKOV et al. [1], Fig. 4 curve 2, stainless steel; Fig. 5, duraluminum.
STURSTAD [1], Fig. 27, stainless steel; copper; aluminum.
The observed W-values are converted to the silver standard by

means of division by the pertinent q;(M) and then plotted in Fig.
(35.10) where the plots are marked by small circles. Generally, they
arrange themselves nicely along a curve that we regard as giving F(P)
°Kfwatt.
Some of the points, marked by crosscs, constitute exceptions. They
require a special discussion, particularly those lying far below the curve
F (P). The crosses represent the theoretically interesting paper of CLAU-
SIN a and CHAO [1]. The points far below the F (P) curve belong to tests
made with extra-ordinary smooth contacting surfaces which were
spherically curved with a large radius of curvature, of the order of 50
to 100m. Evidently the contact area was elastically formed. It is
safe to say that under practical circumstances deviations of this order
are not reached in vacuum.
The crosses marked by interrogation signs represent measurements
with brass members. The reason why they lie far above the F (P) curve
is not understood presently. In any case, they seem to be rare excep-
tions.
The fact that the function F (P) is unequivocal and independent
of the apparent contact area, Aa, creates a theoretical problem. The
contact model given by GREENWOOD and WILLIAMSON (cited in§ 7B)
provides a kind of explanation for the independence of Aa and Ieads to
an approximation of F(P) for a given shape of the asperities (defined
by their amplitudes and the curvature of their tops). But, that the
points from different observers on different materials preferrably
cluster around the same F (P) curve, requires a modification of the
model. We shall not try to construct the complete model presently,
but confine ourselves to emphasize that it seems, that in practice the
bolting Ioad P usually concentrates its effects on a portion of Aa and
there produces dominant plastical deformation of the asperities, which
is largely independent of the curvature of their tops.
We add fi ve instructive exarnples:
Exarnple I. Between nominally flat members of a metal, with A
o~ 300 Wfm °K and ll = 109 Nfm 2 , bolting is produced in air by P
= 100 N. The apparent contact area is Aa = 10- 4 m 2 • In the air filled
gap the averageinverse "'idth 1/s may be 1/10- 6 m- 1 , a common order
of magnitude. Air, in normal atmosphere, has A.::::. 0.025 Wfm °K, a
§ 35. Thermal metallic contacts 197
value that is independent of the air pressure as long as the mean free
path is smaller than the gap width 8. This A is still valid in the gap
assumed; cf. Exam ple II. Thus,
the conductance, Gair• through P-
~- 10 0 10 ' k 1(JJ
tbe gap is Gair = 0.025 · 10- /10-
4 6 10
~ 2.5 WtK. According to Eq. i</W !'-.
(35.08) and Fig. (:35.10), the con- !'... -
ductance of the true contact is
1.8 WtK. A liquid, for example
an oil deposit in the gap, would
111
I I"". ""'
~ X
"
)....
I
1- f--1-
1- 1-
fr 1
produce a conductance of about ~ N. ·-
"
~
lO · Gair = 25 W t K. Thus, a ...::: ~
......
fluid in the gap will dominate
the heat conductance. However,
the effect of the air can be made
1Q-l
I ~
' 1-
negligible by diminishing the 3

.air pressure as will be demon-
1(f100
N 70'
st.rated in Example II.
Example 11, concerning the l<'ig. (35.10). Plots of the function F(P) °K/watt,
or, otherwise expressed, thermal resistances of
same contact as described in nominally flat contacts in vacuum reduced to
Example I, but this time tested the conductivity of silver and plotted again,t.
the contact Ioad P.
at a low pressure p at which the
free mean path in the air, A ,
satisfies A ~ 8. The kinetic theory provides a simple approximate for-
mula for the thermal conductance GA through the gap. One substi-
tutes A for s and obtains
0 1 = 0.025 -10- 4 /A Wt K
independent. of s for 8 < A. Let us require that GA shall be only 1% of
the conductance through the metallic junction, i.e., G;1 = 1.8. 10- 2 • We
find A = 1.4 . 10- 4 m which corresponds to an air pressure of 0.35 Torr
·at room temperature. The condition 8 < A is satisfied at this pressure.
Example 111. Clean looking meta} surfaces in air are covered by
some deposit from the air or, perhaps a thin tarnish. The thickness will
be as much as about 50 A = 5. 10- 9 m. Such films are disturbing in
electric contacts but play only a small part in thermal contacts. The
thermal resistivity of the film material can be as high as 500 times
that for copper, producing a film resistance corresponding to 5 · 10- 9
= 2.5 · 10- 6 m copper only. Even this highest probable resistanco
may be regarded as negligible.
Example IV, showing how to use the information given above for
a practical construction. A clamped junction between stainless steel
members shall be made by means of a screw so as to secureathermal
contact conductance of at least 0.5 WtK. What torque M Nm shall

be applied on the screw ?. The respective steel has the hardness H
= 18. 10s N(m2 (corresponding to VICKERs' number 180) and the ther-
mal conductivity .II. = 17 W (m 0 K.
We have q;(M) = 0.019 V18 · 10S(17 = 47.4; thus 47.4 · F(P)
= 1(0.5 = 2 °K(W or F(P) = 0.042 °K(W. The corresponding mecha-
nical Ioad as read on Fig. (35.10) is 1250 N. The relation between M
and the pressing force P N excerted by a screw on which the torque,
M Xm is applied, is given by Eq. (30.02). In the first approximation the
piteh, h, can be neglected and Eq. (30.02) simplifies to
P= JJt
2r f
whcn the screw has the diameter 2.2 r m and I is the friction coefficient
between screw and nut. A normal value for I on a screw that still has
a thin layer of grease left on its threads is I = 0.3. Hence in our example
1250 = M(0.6r or
M = 750r Nm
Notice that it may be necessary to apply the pressure across a
washer in order to prevent decrease of the pressing force and the con-
ductance when the material creeps.
Example V, showing that the short-circuiting radiation in a metallic
contact is negligible.
Consider a poorly conducting metal as stainless steel. Let P be
p c= JOO N and the temperature Oll both sides of the gap be T and
T -- -&. With the emissivity being s and the gap area 10-4 m 2 , the
radiation across the gap amounts to
according to STEPHAN-BOLTZMANN's law. The resistance ofthe contact

is (according to [35.08], [35.09] and [35.10]) 0.25. 47.4 = 12 °K(W which
by virtue of the temperature difference # conducts #(12 W. The
emissivity s will probably be 0.1, but assume s = 0.5 as maximum value
possible, and take T = 500 °K. Then the radiation is 0.00145# W or
only 1.7% of the conducted heat current.
Part III
Sliding Contacts
§ 36. Survey of fundamentals

A. Introduction. The performance of sliding contacts, for example
the contact between a brush and commutator, is intimately connected
with friction and wear. Therefore Part III must not only deal with the
electrical proporties of sliding contacts but also with the phenomena
of friction and wear. The text and also the Iiterature index can be
shortened by reference to the informative book1 BoWDEN and TABOR
[12]. But, we are inclined to look at some fundamentals in a special
manneras shall be shown in the following.
The friction force F can have four components: F. the force that
produces shearing in the contact surface; Fa the force that produces
shearing off interlocking asperities adjacent to the contact surface; F P
the force that produces plastic deformation; F. the force for those elastic
deformations whose energy is not restored to the slider. In order to
elucidate what is meant with restoration of energy, we recall that the
elastic deformation energy of the rubber tires of an automobile is retur-
ned to the tires to such an extent that the rolling friction becomes
very small.
B. The geometry of a nominally ßat sliding contact: For example a
brush-collector contact is characterized by the load bearing contact
area, Ab, being constituted by several spots which all together cover
only a fraction of the apparent contact area Aa, cf. § 7. The waviness
a:nd asperity of the surfaces of the contact members is responsible for
the generation of those spots. One often finds this fact illustrated as in
Fig. (36.01), where the spots aa are formed on meeting protuberances
ofboth members. Thefigure suggests that the protuberances make deep
indentations, say with specific depths 2 D > 0.03, so that the average
pressure p in the load bearing area Ab can be considered as equal to H,
1 Other recommendable reviews are RABINOWICZ [7]; HIRST [4]; R. HOL:\1 [37]
Part. II.
2 For D see § IF.
200 Sliding Contacts
the indentation hardness. This means an agreeable simplification for

many calculations, but the probability that asperity tops meet each
other is very small. Fig. (36.02) is a more realistic picture. At A, a
blunt protuberance of the upper member touches a shallow valley on
the lower member. At B, another type of contact is produced. The cur-
vatures drawn with different signs in A and about zero in B , defint>
Fig. (36.01). Contact spots formed Fig. (36.02). Contact spots between a hnmp
where tops of asperities meet on one member and a valley on the other
greater contact areas for the same Ioad than according to Fig . (36.01) .
The specific depth may be much smaller thau 0.03 and therefore Jl
much smaller than H , sayl
where
p = UI l (36.0:J)
~< 1 I
In very smooth contacts values of g down to 0.02 have been found .
cf. § 7. One reason for g < 1 will be expressed in Eq. (36.15). We
also refer to contact models by ARCHARD [2], and GREENWOOD and
WILLIAMSON [4] , [8] , constructed so that they provide a small and
constant g, particularly a contact without any plastic deformation ;
cf. p. 31.
Another very useful concept is the specilic lriction lorce, 1p , in
points of the true sliding Ioad bearing area Ab . When the specific fric-
tion force is Iabelied f{! , its average character is emphasized. We have
(36.04)
Ab appears in a similar manner as in Eq. (I,17) which we now repeat
as (36.05)
p = gnA 1, (36.05)
By definition, the lriction coellicient I is

I=_!__ = F, + F. + Fp + F, (36.06)
p p
In many cases we shall find

Fa + FP + F.~F. (36-07)
1 Cf. Fig. (1,13) for D's smaller than 0.03, Eqs. (1,17) and (7.02) . R. HoLM [25]
p. 77 introduced this factor in 1939, initially denoting it by y.
§ 36. Survey of fundamentals 201
particularly for contacts which have run repeatedly in the same track
under normal atmospheric conditions. Thus,
(36.08)
A remark about the explanation of the AMONTON-ÜOULOMB law.

In the literature, one finds the following explanation. :First, assume
that the loa.d bearing area is stressed to the hardness Jimit with P
e
= HAb. This means = 1. Second, that not only H but also 1f' is a
material constant, or better, a contact constant. Since the friction
forre ist F = '!'Ab, one has
f = PF = 'P
H = constant
i. e., independent ofthe apparent contact areaAa. ÜOULOMB's law states

f tobe constant.. However, e = 1 is not a necessary condition for Cou-
LüMB's law. Evidently, it suffices that e = constant; making f = tp(eH
= constant.
The models mentioned above have the merit of showing theoretical
e
possibilities for a constant that is smaller than 1; as it is usually founrl
tobe.
Unfortunately measurements ofAband ~ are scarce. They are diffi-
cult to obtain and drastic changes in the nature of the sliding surface,
for instance when mild wear changes into severe wear (the change one
wants to discover), can be ascertained by corresponding variations of
the friction coefficient. Theseare the consequences of such great changes
in tp, that simultaneaus variations of ~ can be neglected in a first ap-
proximation.
As examples of astonishingJy strict validity of CouLOMB's law,
we cite J. R. WmTEHEAD [1]. He found with a silver slider on a silver
flat surface I = 0.95 ± 0.15, and with aluminum on aluminum
I = 1.3 ± 0.15: when observed in air at 2 . 10-4 < P < 20 N.
We distinguish between two idealized concepts of gliding without
liquid lubricant. I. Gliding in the initial interface between two bodies
that is realized when the adherence across the interface is smaller than
the coherence within the bodies. Therefore, the interface is predeter-
mined as the gliding surface. In this ideal case, no wear would be pro-
duced. II. Gliding in the case of cold welded contacts where the material
behaves as one solid with plastic properties represented by Eq. (1,8).
G liding will proceed where the total amount of shear stress has its
maximum, not necessarily in the initial contact surface. Wear will be
produced.
The process is particularly complicated in Oase II by the action
of dislocations, which is responsible for the shear strength being powers
of ten smaller than in an ideal lattice. This means that tp is not uni-
quely connected to properties referring to the ideallattice, as are the
melting point and the surface energy.
Inhomogencity of the materials, hardening etc. pre,·ent the types
of gliding defined above from being purely realized.
C. Experiments where type I I gliding is expected to be a ppcoached :
Contacts between clean outgassed metal::; are cold welded and sliding
is expected to approach type II. The static tp should approximate the
normal shear strength of the material and the kinetic tp should be
smaller. This is particularly true with hard metals because the kinetic
tp is generated by metallic bonds, whereas the covalt>nt bonds have
t.ime to contribute before static friction is observed: cf. § 28.
Sliding experiments with clean , outgassed metal~ were performed
in nlCcuum by HoLMet al. and by BowDEN et aP . They showed high
friction because the friction force caused
material flow and new contact spots
accord.ing to Fig. (36.09). This preceded
an increase ofthe contact area. the reason
Fig. (36.09). A is an Initial contact
area. The applied friction force F for which these workersdid notdetermine.
bends thc upper member wlth th~
effect that an additional contact spot However, CocKs' [1]. [.2], [3], [4] recent
at R can appear experiments with nominally flat mem-
bers provid.e 2 the possibility to determine
the real sliding area and its tp (as CocKs
hirnself indicates in CocKs [2] p. 814).
Weshall use hisexperimental results for
interest.ing calculat.ions. CocKs did not
F'ig. (36.10). Schematic picture of a employ vacuum, but he cleaned the
wedge in a sllding contact as ob·
served by COCKS members to wettability and used a Ioad
sufficiently high (order of 10 X) to scrat.ch
away films that could interfere. Disct·ete lumps or wedges of a rela-
tively simple shape were built up between t-he members. This was
transferred material which had. hardened by d.eformation. It could be
photographed. In Fig. (36.10) a wedge, W , is schematically pictured.
The members are slightly separated by the wedges and glide over them.
Consider that a wedge as in (36.10) approximates the shape of a
short cylinder, and Iet its cross section be A. It is loaded by a vertical
force P and acted on by the tangential force F. The total contact Ioad
1 See R. HoLM [37] § 38 where references are given. The contact in all cases
was a hollow cylinder hanging on a straight wire and being able to glide along it.
2 ANTLER [2] has also observed the building-up of wedges and prows on sliding
contacts, but under circumstances not equally suitable for the present theoretical
treatment - a.s are the Observations by CocKs. Some of his results are discussed in
the addendmu to this section.
§ 36. Survey of fundamentals 203
'f.. P and the total friction force 'f..F are carried by such lumps. They all
will be rleformed and hardened to about the same degreP: thus
EF F
{co=-o.o-- (:36.11)
' EP P
We can base a simple computation of lf' on Eq. (36.11) discussing one

individual wedge. Evidently the regarded lump. as weil as any other,
is strained in a manner similar to a cylindrieal samplt' during a com-
pression test. Notice, that the comprPs~<iou is Hot produced hy P alone.
but by the resultant of P and F.
This force presses on a surface approximately equal to the cross sPe-
tion, A, of the lump. Thus with Y . thc tensile or C'Ompression strengt h
of the lump material we have
CocKs measured the friction coefficient f P/P"" 0.9. Henct'

P = A Y/1.34 (:~6.1:{)
lfurthermore, since F = A Y we have

AVJ
FfP = 0.9-=>< 1.34yy
or
lf' ""o.67 r
Assuming that the friction is mainly of the shearing type we conclude

that Eq. (36.14) conforms weil with the fact that the ratio between shear
strength and tensile strength of metals is about 2/3, as we read 011
pages 2-62 to 2-6S of Am. Instit. of Physics Handbook, McGraw Hili
1963. We regard this as a proofthat our computations are sound.
The value of ~ in Oock's experiments: Inserting P from Eq. (36.13)
into Eq. (36.0.5) and observing that H = 3 Y since both Y and the hard-
ness Tl refer to the hardened 1 lump material, we have
p = A y I1.34 - ' ~HA ~ $3 y A

or
(3ti.15)
Eq. (36.14) deviates slightly from the theory of v. MISES. His equn.-
tion (1,8), gives tp = Y instead of 0.67 Y: see end of § ID.
Addendum, concerning special observationt' by ANTLF.R [ 4] on tlw
generation of wedges.
1 Hardened during the formation of the lump.

ANTLER operated a spherical rider on a flat member, both of noble

metals. A prow usually formed at the leading edge of the rider taking
its new material from the flat. A soft rider may initially smear material
on the flat, then form a prow with this material. However, this prow
becomes about twice as hard as its matrix member and therefore may
be a ble to wear the flat. When the flat consisted of very hard rhodium,
a gold prow would be too soft to attack it. The regularfront prow then
did not develop.
D. Experiments when Type I would be approached: Imagine two
hard metallic members gliding over each other on perfectly flat contact
areas. According to the theory propounded in § 28, we expect the
adhesions across the contact interfacetobe produced by metallic bonds,
that is, by weak bonds as compared to the essentially covalent cohesion
within the members. We would therefore expect a specific friction
force about equal to the shear strength of silver, and no wear.
However, such an event has never been rei"Jized in an experiment
ht>cause
1. perfectly flat contact areas cannot he provided and interlocking
will produce wear;
2. lattice defects can produce regions, within the member:s, which
a re bonded with a weaker adhesion than tha t in the initial contact arel'l.
However, with a contact between monolayers of calcium stearate
on mica, BAILEY and CouRTNEY-PRATT [1] seem to have realized, or
at least approached 1 , Type I. The monolayer was (leposited by the
LANGMUIR-BLODGETT method on ";heets of mica cleaved to molecular
smooth surfaees (irregularities below 2 A, which is far less than the
length of a stearate moleeule. 25 A). Although the stearatc is liquid
in thick layers, the molecules of the monolayer arc fixed t.o their posi-
tions on the metal surface, and t.hus are a solid body.
Selected sheets of mica were bent like a bmv and mounted as crossed
cylinders. The area between them was observed by means of optical
interference techniques. The friction force at different Ioads was deter-
mined; on the one hand, in the dry contact, and on the other hand, when
the cylinders were covered by the monolayers. The specific friction force
between the monolayers was found tobe 1p = 0.025 . 108 Njm 2 • This is a
very low value. Why is it so low? The answer is given by the generally
accepted picture of the monolayer. Its molecules have a polar group
that binds them to the meta!, whereas the outer ends are ehemically
indifferent.
The outer ends of the lubricant molecules on one member meet the
outer ends of the lubricant on the other member. The attraction be-
1 Their experiments are described in BowDEN and TABOR [12] chapt. XI (XIX)
and XX, where also the character of lubricant monolayers is analyzed.
§ 37. Slidings contact in air 205
tween them is due to VAN DER WAALS forces. Actually measured value
1p = 0.025. 108 Njm 2 is of the right order of magnitude for VAN DER
WAALS bonds.
ZISMAN [1] and SHAFRIN [1] have investigated gliding contacts of
metaland glass with lubricating monolayers. However, here the sur-
faces can not have been so perfectly smooth as on mica, and therefore
the character of the monolayer nöt so secured as in the experiments
cited above. Particularly interesting are observations and conclusions
concerning the ability of the film layers to endure several traverses
without damage. It is important that there is a good intermolecular
cohesion within the monolayer. Paraffinic structures with one or more
covalfmtly bonded halogen atoms show a high intermolecular cohesion.
In one experiment, during 15 successive unidirectional traverses of a
steel ball on lubricated glass the frirtion coefficient remained between
0.05 and 0.06.
It is an interesting problern whether, during sliding, the lubricating
molecules stand vertical or are bent more or less parallel to the metal
surface, gliding so to say on each others back. Expe!'iments by R. W.
WILSON [1] indicate that with platinum members lubricated by mono-
layers of stearic acid, bending is realized when the pressure exceeds a
certain value. In his experiments, Pmin = circa 2 · 106 Njm 2 • The fric-
tion coefficient was low, I = 0.2, and the electrical conduction was high.
With pressures smaller than cited above, the friction was relatively
high, I = 0.6, and the lubricant imulated the contact. The high friction
at small pressures is attributed to interlocking of lubricant molecule~
standing essentially upright on the members. At higher pressures, all
lubricant molecules in the track seem to be bent during sliding and
interlocking does not appear. Due to bending, the film has become
thin, and tunneling can succeed.
WILSON found 1p similar to the value cited above for the experiments
with lubricated mica. It thus seems that he also had very smooth con-
tact surfaces and that Type I ol sliding was approached.
§ 37. Sliding contacts in air

The title of this chapter could ha ve been: sliding contacts inßuenced
by deposits from the air, oxides, etc. in such a manner that no portion
or only small fractions of the load bearing area have metallic contact.
In order to distinguish this chapter from § 36 we add that in § 37 we
do not consider surfaces so extremely :smooth as were realized by
BAILEYand CoURTNEY-PRATT. Interlocking cumbined with some wear
and spots with metallic contact will always be present. Therefore the
average specific friction force tp has values between tp of the order of a

few 10s Nfm 2 valid for clean metals and ofthe order of0.05 · 108 N/m 2
valid in cases of gliding when the adherence is due to VAN DER W AALS
forces alone.
Jt, is interesting that a contact with "orange-peel" type of surfaces
can move essentially without interloclring and with a small tp at small
load; but interlocking, with tp a power of ten high er, a ppears in this con-
tact when the load is great enough to press asperities of one member
into rlepressions on the other one1 . However, also in the event of inter-
locking the producti011 of wear debris is still a. rare event compared
with the number of encounters between asperities.
It seems that, at least in air environment, a kind of ramps are
generated in front of the asperities, which enables encountering humps
to glide up on each other without damage. Certainly this is possible
only when the deformations during the encounter are elastic. In case
of plastic deformation the lattice lamina, the glide of which constitutes
the deformation, protrude step-like from the surfaces and make cor-
responding steep impressions in the counter member, as FENG [1] and
[2] has shown. This means an interlocking that necessarily leads to
wear. Unfortunately, nobody has succeeded in making pictures of the
ramps.
Typical examples of contacts belonging to this chapter are summa-
rized in Table (37.01). Only such examples have been chosen that allow
a reasonable determination of the average tp, whereby the average is
meant to be extended over areas without wear as well as areas where
interlocking pushes wear grains forward.
The type of gliding is characterized by tp, and it is interesting to
find two examples (14 and 15 of Table [37.01]) demonstrating the
failure of the friction coefficient I to define the type. The reason for
this failure is, of course, that I is a function of tp and the contact area,
that can vary much without a change of "P·
Several authors, for example J. R.WmTEHEAD [ 1] and R. W. WILSON
[1], using a pin slider on a plate observed P, I and trackwidth b, and
found Pjb 2 ~ H. However, I· Pfb 2 , which should be equal to tp if the
sliding area were equal to b2, turns outmuch greaterthan H. Evidently,
a prow formerl before the slider, making the sliding area greater than
b2 ; cf. § 36C. Therefore, their data could not be used for Table (37.01).
Explanatory comments on Table (37.01): E signifies a very thin
probably not continuous film remaining on an initially lubricated con-
tact member after cleaning it with cotton soaked in alcohol. Symbol
* signifies crossed-rod experiments. The letter st before a number
refers to static friction.
1 Cf. BowDEN and TABOR [12] p. 40.
§ 37. Sliding contacts in air 207
Table (37.01). Friction measurements
"' I I
p
I p H 0.11 H
Contact
I
f
Test
I N 10• NJm•
1 Cu with E =ll= 31.5 0.13 0.58 4.5 4.5 0.5

2 Cu with E =ll= 61. 0.16 0.57 3.5 4.5 0.5
3 Cu with E =ll= 61. st 0.18 - - - -
4 Al with E 12. 0.12 0.38 3.2 3.0 o.:t3

5 Steel with E =lf 12. 0.12 -· - - -
6 Steel with E =ll= 61. 0.12 3.4 28. 43.5 4.8
7 Steel with E =ll= 61. st 0.135 - - - -
to
0.15
8
9
Ni with E *
Ni with E =ll=
61.
61.
0.16
st 0.19
1.7
-
10.
-
14.
-
Ui
-
10 Pt with E * 12. 0.28

to
-- - 7.5 -
0.33 I
11 Cu plates partly with E 1. 0.12 0.48 4.0 7.5 O.il

to to
(0.2) (0.8)
12 Cu plates partly with E 20. 0.12 0.62 5.2 7.5 0.8
to to
(0.2) (1.0)
13 Steel on copper with Sn
film 4. 0.15 1.4 9.4 9.4 1.1
14 Graphite brush on gra-

phite ring, run-in
2 days 5. 0.13 0.21 1.6 2.6 0.283
phite ring, run-in 1 / 2
year 5. 0.61 0.21 0.35 2.6

phite ring at 40 oc 2.4 0.24 0.21 0.6 2. 0.:!

phite ring at 95 °0 2.4 0.15 0.13 I 0.6 2.
Examples 1 to 10, crossed-rod experiments by R. HoLM [28] with

a velocity of about 1 cmfsec. Apparatus: On a base plate two
parallel wires were stretched, forming a pair of rails on which the
slider moved. The slider was also provided with two such wires fixed
underneath it and at right angles to the rails ofthe lower member. The
coefficient of friction was determined by the method of the inclined
plane. The wires obtained Iasting deformations, enabling the con-
tact area to be measured microscopically. This area is the sum of the

fourdeformed areas on the slider's rods or the four momentary corre-
sponding impressions on the rails. The totalload (on the 4 contacts) is
noted. It was high enough to produce plastic deformation during the
first glide. After running in, no further change of the scar appeared, and
I bad become smaller than initially. The ratio (F. + Fp)/(F8 +Fa)
could be calculated1 • On Cu-wires it was 0.1 to 0.2; on Ni-wires it was
negligible.
Chemical cleaning led to I ""' 0.5 for all of the various kinds of wires.
With a lubricant, base metals gave I""' 0.12. With especially thick
greasing, f could be diminished by 10% at the most, thus for instance,
from 0.12 to 0.11, probably because recovery of the darnage done on
the lubricating film during sliding was facilitated. On lubricated plati-
num I""' 0.3 was measured. In the original paper, the contact surface
was calculated as if being elliptical, but certainly it was nearly quadran-
gular. Therefore, a correction factor of 4/:n: has been introduced, as was
done in R. HoLM [30].
Examples 11 and 12 are due to HoLM [28]. The number of a-spots
was not directly determined but set up in accordance with Table (7.09).
Then, the bearing area was calculated by the aid of Eqs. (7 .06) and
(7 .08).
Example 13 is due to HEATON et al. [1]. It refers to the contact be-
tween a steel hemisphere and a copper plate lubricated with tin (prob-
ably with a thickness of about to-s m).
Examples 14 and 15 are due to HoLM [28]. The contact area was
computed from the contact resistance during slidipg. Values of I, ijJ
and the average pressure, p, at the contact are given in Table (37.01)
for two occasions. First, after running the device two days; then, after
running it 1/2 year. Meanwhile, the contact area increased by the
factor of 4.6. This is expressed in the table by the corresponding de-
crease of p since the load P was constant. Because I, which is propor-
tional to the area, also increased by the factor 4.6, it is concluded
that ijJ remained constant, namely ijJ = 0.21 . tos Nfm 2 •
Examples 16 and 17 are due to E. HoLM [12], [13]. The 1-values
are averages after sliding in forseveral days. The ring and brush mate-
rial bad a resistivity of 4.5 . 10-5 Q m.
Unfortunately, no example of a clean metallic contact in vacuum
could be given in Table (37.01) because no such measurement with a
reliable determination of the sliding area has been published.
Discussion ol Table (37.01). Employing Eq. (36.05) to Nos. 1 to 13,
one states that ~ = pjH varies between 1 a.nrl 0.5 whereas the well run-
1 For definitions of F, etc. see § 36 A.

§ 38. Boundary lubrication 209
in nominally flat contacts of examples 14 to 17 give; = 0.13; 0.11 and
0.065 respectively.
Finally notice the striking fact that 1p:::::; 0.11 H throughout, evi-
dently not perceivably dependent on the lubricant. This fact strongly
supports Hypothesis (38.11).
§ 38. Boundary lubrication

A. Features of boundary lubrication. In §§ 36 and 37 it is concluded
that the specific friction force between monolayers of lubricants is of
the small order of 1p:::::; 0.03 · 108 Nfm 2 • This has a bParing on the theory
of boundary lubrication to be discussed in the present Section.
When journal bearings are lubricated with a liquid and run at full
speed, metallic contact does not appear and there is no mechanical
wear. The lubrication functions hydrodynamically. These conditions
do not prevail during starting and stopping of machinery. On some
elements, for instance, pistons, hydrodynamic conditions will even be
absent during fractions of every operating cycle. The variable wbich
is critical to the maintenance of the hydrodynamicallubrication, with
a given viscosity, is the speed or the relative velocity w between the
sliding members.
When we speak about hydrodynamic lubrication conditions, we have
in view conditions which are realized between nominally flat members.
The hydrodynamic wedging effect (see §VI) that produces the Ioad-
hearing force is generated at the front of small elevations which always
exist. In Appendix VI simplification of the calculations has been gained
by imagining the elevations as semicylinders with their axes perpen-
dicular to the movement. Forthis case the load carrying ability is pro-
portional to wfh1 where ~ is the smallest gap between the sliding con-
tact surfaces across the interspace that is filled with a lubricant. With
another model, the load carrying ability will depend in a more compli-
cated manner on w and h1 , but the difference is of no importance for
the following discussion1 •
A decrease of w is automatically compensated for by a decrease of
h1 , as long as such a decrease is possible. But finally, at a certain low
value of w, h1 will attain the order of magnitude of the amplitude of the
asperities of the sliding surfaces. Then the pressure at the tops of the
elevations becomes too great for maintenance of hydrodynamic lubri-
cation. As a consequence, wear appears, evidently because metallic
contacts are established.
1 Notice that our model surely approaches practical conditions much closer
than the inclined plane slider model does.

The state when the lubricant no Ionger performs hydrodynamically,

but nevertheless helps in carrying the load and keeping the wear smalL
is called boundary lubrication or mixed lubrication 1 • We shall char-
acterize the boundary lubrication by the friction coefficient I which
is an easily measured variable. Its order of magnitude I = 0.03 to 0.2
is typical and whenever I appears of this order one may conclude that
with great probability boundary lubrication is present. The same range
of 1 is characteristic for lubrication with graphite and molybdenum
sulfide. The lubricants considered in this section (also the soft meta!>:'
as tin and indium) flow under the high contact pressure while graphite
and molybdenum sulfide behave in a different manner, and therefore
shall be treated in a separate chapter, namely in § 39.
The fact that boundary lubrication is accompanied by wear does
not mean that metallic contact appears on all asperity tops where the
hydrodynamic film collapses. It is imaginable that a monolayer on each
member remains in the contact at the top of many asperities. The meas-
ured friction force will be composed mainly of the friction between
such monolayers and friction in metallic contact spots. This is a tri-
viality. But in the Iiterature there is a controversy with respect to the
question which of the components is preponderant. One group of wor-
kers propound the theory that the greater portion of the friction force
is located in the lubricant, i. e., on those contact spots which are covered
by perbaps two monolayers of lubricant, one adhering to eacb of the
contact 'members. The present writer has emphasized evidence tbat, in
cases of a friction coefficient I ""' 0.1, the friction in the lubricant is
negligible in the first approximation.
We are now going to display typical characteristics of boundary
lubrication and then discuss the efficacy of the two theories proposed
with respect to the facts. For easy reference, tbe following features of
boundary lubrication will be presented in numbered propositions.
Proposition (38.01). The boundary friction coefficient is about
I = 0.03 to 0.2, say of the order of I = 0.1 independent of the vis-
cosity 'YJ of the lubricant and of the hardness of the contact metal.
The distinction from dry friction (f""' 1 between clean metals,
I""' 0.4 between oxides) and from hydrodynamic lubrication (f""' 0.001)
is unmistakable.
Basis lor Proposition (38.01). It has been known for a long time
1 Strictly speaking, "boundary lubrication" should be referred to the case that
in portions of the contact area the members are covered with monomolecular liquid
films and that no portions are completely uneavered by the liquid. On the one hand,
the hydrodynamic wedging can not work with as little as two adhering monolayers
of liquid and, on the other hand, it is im probable that a pure boundary lubrication
without uncm·ered Bf')ts ever appears. Instead, the lubrication will be mixed.
past that?} can be varied through a range of more than one order of
magnitude without causing any considerable influence on the friction
coefficient I in boundary lubrication. BuRWELL and STRANG [2] found
I independent of both the Ioad (after a small correction) and 'Y), with
17 being varied from 0.08 to 3.9 poise. 0. H. ÜLARK et al. [1] report a
similar constancy of I with 17 varying between 0.04 and 0.5 poise, and
the relative velocity varying between 1 and 16 mfsec. BoYD and
RoBERTSON [1] made interesting experiments with nominally llat con-
tact members at high Ioad. They measured I ""' 0.05 to 0.1 with various
kinds oflubricants as castor oil (YJ = 10 poise), machine oil (?} ~ 1 poise).
oleic acid and Iead oxide; the apparent pressure varying between
1 and 25. 108 Nfm 2 had little influence on the friction coefficient. ThP
true prcssure will be discussed later.
Even with a soft metalas lubricant, I appears in the order of 0.1,
if the metal film is sufficiently thin. The film must be so thin that the
Ioad bearing area is defined by the elasticity and the hardness of the
bulk metal alone. Such a film is unable to exert any measurable groov-
ing resistance against sliding. Examples: Films of cadmium or of tin on
copper surface, as reported by HEATON et al. [1]. Possibly Fig. 44 in
BoWDEN and TABOR [4] referring to a film of indium on steel furnishes
another example of this kind, although it is not evident that the film
of indium was sufficiently thin.
Proposition(38.02). The wPar during boundary lubrication is often
about 1/10 of the wear that appears on dry smooth contacts sliding in
humid (relative humidity ;;:;:;70%) air and having 0.7 <I< 1. This is
concluded from direct measurements ofthe volume that is worn away;
represented by Nos.12 and 13 in Table (41.09). The wear noted corre-
sponds to 1 to 3 ·to-s gfcm in Fig. 1 of RABINowrcz [4], and consti-
tutes the normal amount of wear under the circumstances considered.
With very good boundary lubricants, it can happen that the wear is
10 times smaller.
l-MING FENG [3] points to an interesting phenomenon that has
been studied only a little and shall not be discussed here. He states
that a lubricant performs better ifit contains dissolved gas (as oxygen)
which maintains a surface film to some degree.
Proposition ( 38.03). In case of a good boundary lubricant providing
a smooth motion, I usually rises slightly when the relative velocity of
the members increases at starting within a range of small velocity as
is shown by the lower curve of Fig. (38.05), which is due to KLUGE [1],
or in Fig. (38.06) due to RABINOWICZ1 .
Proposition (38.04). The electric conduction of a lubricated contact
under boundary condition is considerable.
1 RABINOWICZ [3] where earlier Iiterature is cited.
14*
With boundary lubricated cross-rod contacts R. HoLM 1 proved

that the lubricant added no more to the constriction resistance than
physisorbed films contribute. This causes the difference between the
solid and the dashed curves in Fig. (8.01). The good conduction at
boundary lubrications has also been reported by R. W. WILSON [1].
Importent conculsions can be based directly on Propositions (38.01)
through (38.04). First, the increase of f with increasing velocity at
commencing movement, which is ex-
aJ
I
steel againsI tempet>ofut>c ZO't: pressed in (38.03), shows that hydro-
~~ 1cost Iran 1 dynamic lubrication does not de-
i millet>a/ Oll velop in more than negligible por-
fulfy uil tions of the contact within the
IF
range of velocity concerned since
0
I
i
1 2 J
I
a strong effect would reduce j. The
Hmfsec.f exclusion of the hydrodynamic
Fig. (38.05), Due to KLUGE [1). The coeffl. effect as being the Ioad-hearing
cient of frictlon I between contact members
at boundary lubrication, at low speed3 agent allows the conclusion to be
made that the Ioad-bearing area is
determined by the hardness of the contact bulk meta I in conformit.y to
Eq. (1,17), since a liquid lubricant would yield to any static Ioad by
flowing away.
Second, Prop. (38.04) indicates that, in the boundary state,
metallic or quasimetallic 2 spots exist. Thus, if one makes the hypothesis
that a film of lubricant covers an essential portion of the a-spot.s (mak-
z o milli .rleel on milli sfeel
aro
o tool.rleel
tJ. filonium RC !JQA
II
I 6
V
v lilonium RC' 1308 o oo o
~ lungslt:n coPIJide
V
I,P~c~~ _.o'! 0--q~
ß'
~
Q(l8
0::
t
~ a:;l>
0
~
0
au6' n
Q05
lJ.<
/ 0
..-- ~
0
0 ~
~
11.0'1
I
10 V 10 -·
10 -
velocity-
10 cm.fsec
Fig. (38.06). Due to RABINOWICZ [3]. Frictlon-velocity plot for varlous riders (all as hard as mild
steel or harder) slldlng on mild steel. Load 10 N. Lubrlcant copper palmihte•
1 R. HoLlll [28] Section VII. 2 The term quasimetallic is defined in § 2.
3 The friction coefficient f is labeled lt-
ing them quasimetallic) it has to be theoretically proved that it is

plausible that the tunnel current density through them can be capable
of reaching that which has been observed. Data for such a proof are
not available. However, it is reasonable to assume that the work func-
tion for electron emission from the metal into the lubricant is as Jo,,·
as 0.5 V making the required tunneling probable. Cf. § 26E.
ß. Discussion of two competing theorics of boundary lubrication.
Investigators have expressed the opinion that, with good lubricants,
the resistance to sliding is due mainly to the shearing of the lubricant
itselfl. We may add asselfevident that the authors refer to areas where
the lubricant is in the boundary state, consisting of about one mono-
layer on each contact member. This hypothesis of a considerable shear
strength of the lubricant, which we shall call
Hypothesis (38 07 ),Ieads to a theoretical difficulty that shall nOI\'
be debated.
Let us compare two sliding contacts M and N in the boundary
lubricated state, both between equal metal members with equal shape
and Ioad P, both with 2 f = 0.1, but with different har·dness 3 , Hm and
H,., of the bulk meta! where
11,,, = qlln (38.08)
H follows from the assumptions that the friction force is equal in

both cases, namely, F = f P = 0.1 P. In each case, tP,e load bearing
area, A 111 or A" is determined by Eq. (1,17), where we may assume the
same ~ for both contacts. Hence
qA,11 =A 11
Since Ioad and friction force respectively arc the same in both cases,
the average pressures satisfy
p,/1 = qp"
and the specific friction forces
q'tp"
'ip", =
Hence, generally when varying the metal

tp is supposed to be proportionaltop (38.09)
where ffJ would be the average shear strength of the lubricant film,
provided that this film essentially covers the area Ab.
A general validity of Prop. (38.01) and Hyp. (38.07) together with
Eq. (38.09) requires the extremely unlikely assumption which we call
1 BoWDEN and T.ABOR [4] Chap. X, and addendum p. 350.
2 According to (38.01) we assume f = 0.1.
a This case is represented by Nos. 4 to 6 of Table (37.01).
Assumption (38.10) containing two statements: 1. tlmt alllubricants

exhibit about the same ip on a given metal; 2. that on two metals (1) and
(2) characterized by the hardness H 1 and H 2 , ip has values fjj 1 and fjj,_
satisfying
Such a twofold accomodation of ip for any lubricant on any of thl'

metals is extremely improbable. Nevertheless, some workers seem to
believe in it, particularly because they feel that experimental proofs
for (38.09) existand are given in observations by EoYD and RoRERTSO!i! 1 •
Let us scrutinize the conclusions drawn from the ER experiments.
They have been considered as being of a particular type because very
great contact Ioads and nominally :flat contact members were used.
providing relatively great load bearing areas. It seems to have been
believed that the Ioad bearing area of the lubricated ER-contact com-
prised the entire face of the anvils that served as contact members and
that the average Ioad bearing pressure ji on the true contact would vary
proportionally to the Ioad P. Eut this is at variance with facts.
On the basis ofreasonable assumptions aboutthefinishofthe contact
surfaces, it is shown in R.Hor M [ 37] §IX that the true Ioad bearing area,
Ab, was much smaller than the apparent contact area Aa. The conse-
quence is that the performed variations of the Ioad have essentially
led to a variation of Ab. The average pressure ji on Ab changed less tkan
the third root of the total Ioad, contrary to the postulation of the
authors mentioned. We conclude that the ER-experiments do not
prove more about the in:fluence of the pressure on the shear strength
than the other experiments which show that f""' 0.1 is typical for
boundary lubrication independent of the metal. The ER-experiments
do not constitute any exception.
It is remarkable that workers have been astonished to find average
pressures of the order of the hardness of metals under boundary con-
ditions. EOYD and ROBERTSON [1] as weil as 0. H. CLARK et al. (1)
emphasize in the title of their papers that the pressure was high. As
a matter of fact, the pressure in the Ioad bearing area was ofthat order
of magnitude in all experiments for Prop. (38.01).
It is now appropriate to attempt another theory 2 • To that purpose
Hypotkesi.<J (38.11) shall be described with reference to Fig. (38.12)
that may be regarded as illustrating a cross-rod, lubricated contact
between cylindrlcal members3 • It is further assumed that the contact
1 BOYDand ROBERTSON [1], abbreviated BR in the following.
2Cf. R. HoLM [37] § 39B.
3 In this case the Ioad bearing area is circular and essentially coherent. In a
nominally flat contact it may be divided into non-coherent portions.

can be considered as fresh, which means that the shape of the average
elevations ofthe members has not changed considerably by wear. These
assumptions involve simplifications for the discussion, but do not
severely endanger the conclusions. Because of the present provisory
state of the theory, the simplifications may be allowed, and modi-
fications of them are not discussed.
First we imagine the contact as being dry . Then the load bearing
area is defined by P = ~ Aill, i. e., Eq.(36.05). We shall prove t hat a
lubricant does not alter Ab . The
factor ~ will be essentially inde-
pendent of the metal. On Fig.
(38.12) the diameter of Abis (ah).
For simplicity the figure has Fig. (38.12). Diagrammatic scct.ion through con·
tact area a t boundary lubrication
been designed for an ideally hard
and smooth member C, and the actual unevenness is represented by the
waves on member B. The tops of these waves have been fiattened by
the pressure from member C. The fiattening action is limited in a
certain manner by the counter-pressure of the lubricant which we
imagine to be locked up in the valleys around and between the eleva-
tions . It is of fundamental importance for the discussion that., at equi-
librium, the hydrosta.tic pressure must be the same in the lubricant
and on the asperity tops, as is evident from the following.
The pressure on the tops is ~H, which is a function of the mechan-
ical strength of the metal. However, it is evident tha.t the tops alone
are not able to carry the total load, that in the dry contact would
generate a load-bearing area with the diameter (ah) in Fig. (38.12): That
the lubricant assists them is easily seen if we imagine what happens a
short time before the equilibrium is established. There will be a moment
when the pressure on the lubricant is !arger than ~H, and this of course en-
ables the lubricant to fiow out from the valley over some elevations, a
process that continues until the pressure in the lubricant attains the value
of the pressure on the tops. Then , at established equilibrium, the overall
pressure is the same as in the dry contact, and we conclude that the
load bearing area preserves its diameter (ah) also in the lubricated state.
It is to be noted that this argument and conclusion is indcpendent
of the hardness of the contact members. The tops have been fiattened
so far that the depressions containing lubricant are closed ; i. e., none
of the lubricant fiows out of them. This means that, on an a verage, the
tops ha ve been lowered by a certain percentage of their initial height;
this percentage probably is independent of amplitude and wavelength
of the elevations. The deformation is defined purely geometrically and
therefore is independent of the hardness of the metal. Hyp. (38.11) can
now be expressed as follows .
Hypothesis (38.11) assumes: First, that the ratio between the sum
of the top areas and the totalload bearing area (ah) in Fig. (38.12) is
Iargely independent of amplitude and wavelength of the surface une-
venness. Second, that this ratio is of the order 1/10 to 1/20. Third, that
the friction on areas which are covered by the lubricant is negligible
to the first approximation. Fourth, that the friction on the tops is of
the kind that in dry contacts1 .makes f"" 1. All of these assumptions
are plausible, and if Hyp. (38.11) be factual, it would explain Props.
(38.01) to (38.04) as shall be shown below.
It may be appropriate to begin with citing observations by RABI-
xowwz and TABOR2 which support the second statement ofHyp. (38.11).
These authors made one contact member radioactive and observed by
means of autoradiographs that metal was transferred to the other mem-
ber; on one band, when the contact is only closed and opened ; on the
other band, when sliding occurs. The authors compare the transfer in
dry and boundary lubricated contacts and find pick ups which are
6 to 50 times smaller in lubricated than in dry contacts. This indicates
that the lubricant reduces the metallic areas by a factor of the ordet'
that is assumed in Hyp. (38.11).
Furthermore, we mention that a determination of the average spe-
cific shear force fji between two monolayers of a lubricant (calcium
stearate) has been made by BAILEY and CoURTNEY-PRATT3 • They find
fji = 0.025. 108 Nfm2 which lies about two orders of magnitude below
the shear strength of those metals that are used for the contacts dis-
cussed above. Where the lubricant has a greater thickness or perhaps
acts hydrodynamically, its contribution to the friction is still smaller,
and we conclude that the total friction force on the lubricant is about
10% (or less) ofthe friction force on the metallic tops. We thus justify
the statement that the friction on the lubricant is negligible compared
with that on the metallic tops in Fig. (38.12), the case considered in
Hyp. (38.11).
Fig. (38.12) has been described particularly for the case of contact
closing, but it also applies reasonably weil to the formation of any
momentary contact during sliding without galling. We are now pre-
pared to show how the Props. (38.01) to (38.04) are explained on the
basis of Hyp. (38.11).
Concerning Prop. (38.01). Since the friction force is mainly referred
to the top areas it must be reduced by the same factor of 1/10 to 1/20
by which the top areas are smaller than the Ioad bearing area. This ex-
This is a typical friction coefficient for non-deformed metallic contacts.
1
RABINOWlCZ and TABOR [1] and RABINOWlCZ (2]; cf. Fig. (28.06); cf. also
2
pictures by MooRE [1].

3 See§ 36D.
plains Prop. (38.01). Prop. (38.02) also is a direct consequence of the

rubbing area being 10 to 20 times smaller than in the dry contact. In
order to explain Prop. (38.03) we have to add the plausible hypothesis
that, to a small degree, the sliding cleans, i. e., increases the top areas
which are responsible for the friction. Finally, the fact that the con-
ductivity of the contact is reduced only little, even though the total
area of the a-spots is much smaller than the Ioad bearing area, is ex-
actly what is to be expected on the basis of Fig. (5.13). Even with the
a-spots (20 to 50 of them) covering only 1/4 of the Ioad bearing area
Ab, they conduct together only about 20% less than a fully conduct-
ing Ab according to Fig. (5.13). Thus, also Prop. (38.04) is explained.
Not included in the propositions is an experience to which atten-
tion should be paid. First, it is a commonplace that a lubricant in order
to be good at boundary conditions must have fairly long moleculeR
(containing, say, 10 to 12 carbons in the chain) which shall adhere
strongly to the contact members 1 . Unfortunately, a high ability to
adhere is a result of chemical bonds 2 and is combined with the tend-
ency of the lubricant molecules to detach meta! atoms from the mem-
bers, i. e., to produce chemical corrosion. This is why the engineers
have to compromise by using mixture lubricants; a chief ingredient
being one that lubricates weil hydrodynamically, and a sparse ingre-
dient that is st.rongly adhering; but, because of its sparsity producing
lit,tle corrosion.
After this remark, we go back to discussing the relation between
Hyp. (38.11) and the fact that long and strongly adhering molecules
are important for a good lubrication. Hyp. (38.11) refers the Ioad
carrying ability of the lubricant to its being captured in the valleys
bet,ween the superficial elevations. Obviously, the Ionger the lubricant
molecules, the more perfectly they are prevented from escaping through
fine gaps and the more they adhere to the members.
To summarize: We have discussed two hypotheses - Hyp. (38.11)
and Hyp. (38.07) together with Assumpt. (38.10). They have been pro-
pounded to explain the behavior of boundary lubricated contacts, and
we have emphasized that no acceptable proof of either one exists. How-
ever, Hyp. (38.11) seems to be moreplausible and to have greater effi-
1 For instructive pictures of the arrangement of the long lubricant molecules
on a metal surface, see GoDFREY et al. [1].

2 The ability of lubricants to adhere to metallic contact members has been
studied by BowDEN and collaborators. They have emphasized that the adherence
is relatively good on base metals, particularly in presence of oxygen and humidity,
but is relatively poor on noble metals. See BoWDEN and TABOR [4], Chapter X and
[12] Chapter XVIII, and later publications as MENTER and TABOR [1], BoWDEN
and MoORE [6] and DANIEL [1].
cacy in explaining the subject than Hyp. (38.07). Therefore, we repeat

that Hyp. (38.11) is recommended for serious consideration.
C. Lubrieating pratiee. It remains to Iook at our picture of bound-
ary lubrication in connection with aspects of lubricating practice. A short
discussion that is confined to a typical example suffices, because we
would be led to repeat conclusions as are given in available treatments
of lubrication.
Typical boundary lubrication occurs in journal bearings at the
start of motion. It is important at this stage that galling in metallic
spots does not pull out grains, thereby introducing asperities. These
would dispose interlocking and continued wcar. Fortunately, the bound-
ary wear in contacts with such a finish, ao;; is common in new bearings,
does not increase the roughness of the surfaces. On the contrary, it
smoothes out their asperities, if suitable materials are used. One aspect
ofthe process seems tobe a plastic deformation that continously moves
material from elevations into depressions; it is possible that a granular
wear also plays a part. Very small particles seem to have the ability
to adhere strongly where they are deposited on the matrix material,
probably because of their relatively large surfaces. The effect of smooth-
ing out the surfaces is taken into account with the choice of materials
for bearings. However, before discussing the materials a note about the
transformed surface is appropriate. It is a typical BEILBY layer.
D. The Beilby layer. The top layer on a mechanically polished metal
has a structure different from the underlying material. Because BEILBY
[1] was the first to pointout the existence of such a layer, it is usually
called the Beilby layer. BEILBY hirnself made only optical observations.
For thorough information, investigations by means of electron diffrac-
tion were necessary. Such have been carried through particularly by
RAETHER1 and collaborators and by FrNCH 2 and collaborators.
The main results may be summarized as follows. Mechanical polish-
ing or any dry abrasion breaks up crystallites to a certain depth below
the surface depending on the severity of the rubbing. Particularly in
the topmost layers, the debris becomes extremely small. On a brittle
body (for example rocksalt or spar), the diffraction pattern may in-
dicate crystallite dimensions of the order of 100 A. On the surface of
metals, the elements become so small- of the order of 10 A- that the
layer ma.y be called amorphous. Probably the polishing process has
moved these amorphaus elements into previous depressions. It is parti-
cularly this amorphous top layer, caused by a smoothing-out action,
that is called Beilby layer. It is plausible that a temperature rise by
friction heat can support the formation of the BEILBY layer, but
1 RAETHER [2] and [3] where earlier literature is cited; seealso RAETHER [1].
2 See the survey paper FINCH [2].
§ 38. Boundary lubrication 2Hl
RAETHER has observed its generation under cireumstanees that ex-

clude any considerable inerease of the temperature above room tem-
perature. The amorphous, partly oxidized BEILBY layer exhibits a great
electric resistance and seems tobe harder than the underlying material.
BowDEN and TABOR [4] p. 58ff. present a discussion on the BEILBY
layer in which they particularly treat cases where melting seems to
have been produced. SAlVIUELS [1] arrives at the conclusion that the
fine polish, including a possible BEILBY layer, results from plastic flow
for which a softening temperature may be favorable.
Recently, H. RICHTER [1] showed that the BEILBY layer on tin and
lead, that had been polished against glass, consisted of SnO + Sn and
PbO + Pb with an essentially amorphaus strurture exhibiting a re-
latiunship to the oxide lattiee.
E. Boundary-lubricated separablt> contacts inclutling somt> sliding
metal-metal contacts. Noble metalf' are requircd for maximum reliabi-
lity at contact make in air. However, the reliability may be gained at
the expense of resistance against adhesion and wear. The wear must
be considered if P > 0.1 N. With the idea that boundary lubrication
could prevent the wear without disrupting electric continuity,investi-
gators are looking for suitable lubricants which can serve during a
great number of operations without impeding the electric conduction
or producing disturbing noise.
Several investigators, connected with Stanford Research Institute
or with IBM, have worked on this problem: J. R. ANDERS ON and SAUN-
DERS [1]; ANTLER [1], [2]: SPERGEL, GoODWIN and STEINBERG [1]; ÜHIA-
RENZELLI and HENRY [1]; STEINBERG [1]. The standard test condi-
tions were: Ioad P = 1 to 3 N; speed""' 1 cm(sec; voltage at open contact
::::: 0.01 V. Details of the results fit in with Section B above. The lubri-
cantl (physically adsorbed) fills depressions in the contact surfaces
(depths of the order of a few 1000 A) .At the speed tested, smallmetallic
or quasimetallic, and thus eonducting a-spots are left on elevations,
even with the lubricant. A certain asperity of the contacts is required
for this. Very smooth contacts are prone to insulate. The depressions
serve as good lubricant reservoirs for back and forth sliding, but are
soon emptied when sliding continues in Oll(' direction.
A high viscosity of at least 17 = 3 poise is required to keep the frie-
tion eoefficient f = 0.1 to 0.3, as desired. With still greater 11 the contact
is transiently lifted by hydrodynamic lubrication, thus very noisy.
When the electric conduction was acceptable it was close to that
of a clean contact. We know from Fig. (5.13) that this does not contra-
dict the idea of separated a-spots. The good electric contact is considered
( STEINBERG) as a consequence of the low yield point of the adhering
1 For methods of application of the lubricant, 8ee especially STEINBERG [1].
lubricant tbat sball not be cbemisorbed. Tbere was evidence tbat

(ODA-HCI) corresponded totbis condition (STEINBERG [1]).
According to tbeir present experience, tbe autbors recommend for
lubrication, particularly on gold: octadecylamine-bydrochloride (ODA-
HCI); see SPERGEL et al. [1]; on silver: a mixture of paraffin and a
synthetic oil; see CmARENZELLI and HENRY [1].
F. Properlies required of bearing materials are primarily the follow-
ing: A. Anti-seizure; B. Ability to become smoothed during bound-
ary conditions; C. Load carrying capacity and fatiguc resistivity. We
shall now consider bow tbese requirements are met.
Concerning A, one member sball be barder than the otber, initially
a little tarnished as a prote9tion against seizure, and the softer member
shall not wet the barder one. Many ideas are being considered con-
cerning bow to obtain little or no wetting. According to one recent idea,
seizure is particularly small between metals with very different atomic
spacings of the lattice, and the materials of a bearing should consist
of such metals1 • LUNN [1] aud [2] concludes from bis experiences that
a seizure-resistant, very plastic film is produced on good bearing ma-
terials by a cbemical reaction between the bearing metal, the lubri-
cant and probably oxygen. No such specific reaction nor a general rule
for a suitable combination of lubricant and metal can be named. The
importance of the oxygen is demonstrated by the observations by
ROACH et al. [2] that seizure appeared when a good lubricant oil was
deprived of its content of oxygen. Granted, differences in the weld-
bility between pairs of metals exist, but it seems true that all of them
weid to some degree; cf. MACHLIN and YANKEE [1].
Concerning B, the surfaces of both contact members should become
smootbed by tbe initial wear so as to form BEILBY layers. It is believed
that tbe softer member sball be fairly plastic and that its tendency to
strain barden sball be as small as possible. Tbe corresponding require-
ments on bard members are more difficult to meet since tbe cboice
of metal is limited to materials with a high mecbanical strength, for
instance steel. However, it would be satisfactory if a thin surface layer
were produced on tbe hard metal whicb was prone to become smootbed
by forming a BEILBY layer. It seems that the so-called high pressure
additives to lubricants, used for hypoid gears, contain radioals (chlorine,
phospborous, etc.) that react witb the steel, producing an alloy film
with the quality required2.
Concerning C, tbere exists a lower admissible Iimit for the hardness
of the softer bearing member since it must be able to carry the Ioad.
Because of tbe B requirement, Iead or tin or recently indium, have
1 ROACH [1].
2 BoWDEN andTABOR [4] Chapt.. XI; FINCH [2].
§ 39. Theory of friction and wear of carbon C'OiltllC'ts. Lubrication... 221
becn the major constituents of bearing materials. In order to meet re-

quirement C, the strength of the material is increased by alloying, pre-
ferably with copper and antimony.
A duplex structure1 consisting of hard particles embedded within a
softer matrix has been considered as favorable for satisfying require-
ments B and C. However, this theory seems now to be opposed as it
cannot explain that some homogeneous materials are superior to the
duplex structure alloys.
It is obvious that this field is very much in flux and it is out of the
question here to give an extended account of the theories of bearing
materials. We refer to specific
literature as for instance Bow- Table (38.13). Electric reBiBtance acro88 a ball
DEN and TABOR [4] and [12], bearing
(mfd by MRC, outside diameter 19 mm)
TOWLE [1].
G. Ball bearings Whether rpm Plb V resist. n
the lubrication in a ball bear-
0 0 0.03 2
ing works hydrodynamically or
in the boundary state can be 170 0 0.1 600
170 0 0.5 170
answered by observing the
170 50 0.1 6
electric resistance across the 170 50 0.5 2
bearing. When run without
1800 0 0.1 00
mechanical load, the bearing 1800 50 0.1 600
usually insulates, indicating hy- 1800 50 0.5 200
drodynamic lubrication. But, 1800 85 0.1 30
when a load is applied, there is
a certain amount of conduction which is an indication of boundary
lubrication. Table (38.13) gives examples of observations. Electric
current through the bearing produces corrosion. MECKE [1] observed
a considerable corrosion when the current reached the order of 10 A.
In trolley cars care is always taken to prevent currents from pass-
ing through the ball bearings.
§ 39. Theory of friction and wear of carbon contacts.

Lubrication by means of solid lubricants as graphite and
molybdenum disulfide
A. Introduction. The properties of carbon ma.terials as sliding con-
tacts are outstanding in their application as electrical brushes. Because
of the importance of this application, the brush contact will be the
primary subject of this section. Although during the 80 years that
carbon brushes have been used and immense experience on their per-
1 BowDEN and TABOR [4] where bearing alloys ofthistype are discussed.
formance has been gathered, the theory of the brush contact, never-
theless, is still ineomplete. Equally incomplete is the theory of lubri-
cation by means of powders, e. g. graphite. Various materials qualify
for easy sliding in a manner similar to graphite, for instance, MoS2 ,
WS 2 , Cdl 2 , Pbl 2 , CdCI 2 , Hgl 2 • Allare characterized by a laminar st.ruc-
t.ure with relatively weak bonds between the basal planes. This ob-
viously constitutes a primary but not a sufficient condition for the
lubrieat.ion ability.-It also plays apart that some planes in a crystal
glide eat:~ier than others; see below. The idea of laminar lubrication re-
quires parallellamina to glide over each other.
We envisage the circumstances in the contact of a carbon brush
sliding on a collector ring. On the surface of both the brush and ring,
carbon films are deposited1 • They have been investigated by several
workers by means of electron diffraction and microscopy. We parti-
cularly mention JENKINS [1] whose results have been eonfirmed by
MIDGLEY and TEAR [1], and T.F.J.QUINN [1]. The wear fragments of
carbon brushes seem to be disintegrated into platelets of single cry-
stals. The dimen~üon of their basal planes can be as small as 25 A.
Their thickness certainly is smaller than the basal dimension. The
platelets are mainly orientated with their cleavage plane fairly parallel
to the sliding interfact> with an average inclination of up to 5° against
the sliding direction. It seems very likely that they are bent during
8liding so that actually cleavage crystal planes glide over each other.
Surely, the coherence is not the same between allhexagonal (basal)
planes within a crystal. That is why wear particles never eonsist of
singlebasal planes but ofplatelets containing many ofthem. We expect
that in those partial regions of the crystal where the bonding is rela-
tively weak, the distance is increased by interstitial alien atoms or by
dislocation 2 • Based on theoretical calculations by BRENNAN [l] and
observations on cleavage forces in ultrahigh vaeuum by BRYANT et al.
[ 1] we conclude that the normal bond between basal planes in a gra-
phite crystal is about 0.3 eV per earbon atom. In air, oxygen atoms
seem to enterat least some interspaces and diminish the average bond
to about 0.06 eV. Cf. Section B.
Rem.ark. Run-in copper collectors are covered by a film of orien-
tated gra.phite platelets over a thin layer of copper oxide. The brush
glides mainly on the graphitefilm at least during smooth sliding. Since
the brush carries a similar film, sliding is essentially between basal
planes of graphite. The orientated graphite film on both brush and
1 The carbon film deposits on a layer of copper oxide; see v AN BRu:sT and
SAVAGE [1].
2 Cf. TsuzuKu [1], PATELand BAHL [1] where earlier Iiterature is cited, and
HENNIG [1).
§ 39. Theory of friction and wear of carbon contacts. Lubrication... 22:3
copper collector can reach a thickness 1 of the order of 1000 A. The
insulating oxide film does usually not exceed 2 100 A. When a pre-
viously cleaned collector ring begins to rotate, the very first readings
show the low friction of f::::: 0.1, indicating that the deposition of an
efficient film proceeds in seconds. This means that a film much thinner
than mentioned above is sufficient for easy gliding.
B. Bond strength in a sliding contact between graphite members.
E. HoLM computes 3 the bond strength in a (nominally flat) sliding
contact between graphite members from the dependence of the fric-
tion coefficient on the contact temperature under circumstances4 that
guarantee a constant contact sarface, A 0 • The experimental results
support the assumption that the friction is essentially due to adhesion
forces and that the temperature diminishes the average bond strength.
The average bond strength perpair of atoms across a contact sur-
face, A 0 , is assumed tobe !peV at temperatures low enough to havc a
negligible influence on the bond. At influential temperatures, the
average thermal energy of those two degrees of freedom (kinetic and
potential) that are vertical to A 0 , subtract from (/J· Thus, at the tcm-
perature T, the average bond energy at A 0 is
qJ-kT (39.01)
where k = 1/11600 e V per °K. If the friction force, F, is entirely used
to shear the bonds in A 0 , the friction coefficient, f, is strictly propor-
tional to the specific friction force 1p = FjA 0 = ffi, where p = PjA 0 is
the average pressure with P =' mechanicalload. Eq. (39.01) then leads
to
/(T) = 1p(T) = c(rp- T/11600) (39.02)
where c is a constant.
The validity of Eq. (39.02) requires that sliding occurs in the coH-
tact surface A 0 • This clearly is the case after the mem bers ha ve been
run in a relatively long time because the wear then is negligible; namely
about 10- 12 cm 3 per revolution on a sliding track 0.4 cm wide aml
220 cm long, independent of the current.
The necessary requirements were realized when the interdependent
and simultaneously measured variables j, T and U remained constant;
1 VANBRUNT and SAVAGE[1]; G.L.CLARK etal. [1], they also measured a
bardness of 0.6. 108 N /m 2 of the film deposited on a brush manufactured on a car-

bon black basis.
2 See E. HoLM [14].
3 Cf. E. HoLM [13] where a different computation gives somewhat higher values
of the bond strength.

4 As for friction device, accuracy of measurements, environment, and calcu-
lation of contact temperature see E. HoLM [13].

namely I= 10 , T = T 0 and U = U0 , where U = contact voltage. This

stage is markedi0 (T0 ) on the curves in the figures below. It is the stage
when ambient heating of the members starts.
The dashed parts of the curves in Fig. (39.03) demonstrate the
sliding-in process. In all tests, I starts at 0.1 and U at about 0.8 V.
A
8
I I
fofloJ flrophd~ -grophde
I
10.15
I
'O.W r-~~------r-----~~~~~~
ODS~~~~----~W~
0----~~~0---.~~~
T-
Fig. (39.03). Due to E. HoLM [JJ). Friction coefficient, f, and contact voltage, U, vs contact tem·
perature, between an electrographite brush (0.3 cm• sliding face) slidlng at 10 A and a ring (7 cm
diameter) of the same material. Formaterial constants see p. 151, footnote I. Curve A: run-in time
"" I 0 hours; curve B: run-in time "" 60 days. P = 2.4 N, speed about 2 rn/sec. The end of sliding
in and start of ambient heating is marked by / 0 (T0 )
Then I increases and finally Ievels off to a maximum lo (T0 ) while U

decreases to a certain minimum, U0 , at given current. Curves A and
B refer to very different sliding-in times.
At the sliding stage lo (T0 ) a surface film has developed on both
membertracksand thendeformationhas become merely elastic. The film
consists of graphite platelets with their basal planes almost horizontally
orientated along the sliding track. It is reasonable to conclude that this
fo(TqJ
~ r-N Grophile-grop/Jile
()
foflo)
·;.
'S
~~
0 ~ =--- _""_
r=-
0.0
so 100 150 "C zoo
r-
Fig. (39.04). Due to E. HOLM [12) and [13). Simllar to Flg. (39.03) but wlth extended measure-
ments. Curws A and Bare reproduced from Fig. (39.03). Curve C: I (current) = 5 A; curve D:
I = zero
§ 39. Theory of friction and wear of carbon contacts. Lubrication... 225
film with a sporarlic gas coverage from a clean atmosphere at constant

humidity remained constant during the respective changes of the tem-
perature because U remained constant.
Calculation. Fig. (39.04) was used for the calculation. The different
curves in this figure refer to different currents. With graphite members
t.he conducting area equals the load bearing area as is evident from the
contact resistance being independent. of the current. The current ex-
erts no other inftuence on A 0 than to increase the temperature. The
high er the current, the more lo (T 0 ) is shifted to high er T 0 and thus low-
ered. The curves show that during rising temperature, I steadily de-
creases and finally Ievels off.
Using Eq. (39.02) for two different points on the curyes in Fig.
(39.04) with / 0 ( T 0 ) > I (T), one obtains
(/!-T 0 /11600
q;- T/11600
or
Tf 0 (T 0 ) - T 0 f(T) 1
rp = fo(T 0 ) - f(T) 11600
(39.05)
I. For graphite members in air according to Fig. (:39.04), for

example, with the points lo (T 0 ) = 0.25 at T 0 = 322 °K, and I (T) = 0.08
at T = 498 °K, one obtains
rp = 0.052 eV (39.06)
somewhat smaller than the value obtained in E. HoLM [13].
II. For outgassed graphite members heated to about 2000 oc and
then rapidly cooled to room temperature1 • The friction is then measured
in vacuum between room temperature and 2000 °C. It is assumed that
the contact surface remained fairly constant during sliding only once
at the various test temperatures. Eq. (39.05) yields for two different
points on the respective friction-temperature curve, namely lo = 0.43
at T 0 ~ 300 °K and I = 0,19 at T = 2273 °K,
rp = 0.33 eV (39.07)
A bond strength of about 0.3 e V has been measured during cleaving
a graphite crystal in ultrahigh vacuum 2 at room temperature. Clea-
vage of graphite crystals could explain the extremely fine carbon dust
obtained under high altitude conditions.
C. The high altitude efl'ect on brush wear and its significance for
the theory of graphite friction. The high altitude effect on brush wear,
1 See D. M. KENYON, PH. D. Thesis, University of Cambridge, England (1956).
His friction-temperature curve is reproduced in BowDEN and TABOR [12] Fig. on

p.189.
2 See BRY ANT et al. [1].
15 Electric Contacts. 4th Ed.

HAE, is a name that aircraft equipment engineers have given to the

se.vere increase of the wear at high altitude that caused much trouble
beginning in World War II. Carbon brushes that are able to serve for
a long time on a copper collector at sea Ievel conditions, may wear out
in minutes in the dry air at altitudes above 7000 m. The same effect
is observed in a dry vacuum. Admission of water vapor eliminates the
high wear provided a certain critical humidity is attained. The critical
amount of water vapor1 seems tobe 1 to 3 gfm 3 .
We recall from Section A that the bonds between normal basal
planes of a crystal are much stronger than those bonds where cleavage
occurs. Assuming that the bonds between platelets gliding over each
other in vacuum are similar to the bonds within a crystal, we have
an explanation for the increased friction and wear in vacuum; see
(39.07). It is also plausible that the high friction destroys the orien-
tation of the platelets.
However, other effects may be involved. Earlier theories of HAE
converged on malring free valences at the edges of graphite platelets
responsible for HAE. Any graphite crystal offers free valences at the
edges unless they are occupied by criss-cross bonds to neighbor crystals
or otherwise neutralized 2 • With unoccupied edge valences (a- electrons)
one expected the platelets in the surface of the collector film to be
prone tostick to similar platelets in the brush surface and cause HAE.
During friction,platelets are steadily broken and newedges are generated.
This explanation presupposes instantaneous development of coval-
ent bonding by a-electrons 3 • According to § 28A it seems doubtful
that this process has time to develop in a sliding contact. Therefore,
it is to be considered whether n-electrons moving along basal planes
could be numerous enough to produce considerable adherence by
"metallic" bonds 4 at platelet edges where the exchange of n-electrons
can be abundant 5 •
If HAE is to be avoided, the surrounding gas must supply radicals
quick enough to neutralize freed edges and block emission of n-electrons
there. SAVAGE [2] and [4] found that the least humidity required to
prevent HAE with carbon brushes on a copper collector, namely 0.1
Torr, is of the order necessary to maintain a monola~ er of chemically
bound water on the free surfaces of platelets and edges. Actually, not
water itself but its dissociation products6 are chemisorbed. SAVAGE et
al. showed that various vapors as CH 30H, C5 H11 , C5 H12 , CC1 4, Br and
others prevented HAE.
If the theory outlined above is correct, these vapors, including
1 VAN BRUNT and SAVAGE [I]; Par dee [1].
z by radioals or by the effect that produces access holes; see p. 410.
3 See§ V. 4 See§ 28, p. 153. 5 See SHOBERT [8]. 8 See STUDEBAKER [1].
§ 39. Theory of friction and wear of carbon contacts. Lubrication. . . 227
water, should have the ability both to neutralize freed valences anrl
also lubricate the basal planes.
DEACON and GoODMAN [l] emphasize (with respect to lubrication)
the superior role of edge valencies getting occupied. BrssoN et al. [1], [2]
observed a good lubrication with non-outgassed graphite at 500 °C,
although the orientation of the platelets was lost. At 300 °0 the Iubri-
cation was poor inspite of a good platelet orientation. They be1ieve
that oxide developing on the copper enhances lubrication by facili-
tating the adherence of lubricating graphite platelets.
It is evident that the theory of HAE has not reached its final stage.
Since MoS2 has a laminar lattice structure, it should lubricate simi-
lar to graphite. But, it has been shown that MoS2 lubricates up to a
merlium vacuum and V. R. JoHNSON [1] and [2] found that humidity
even impairs lubrication. HALTNER [l] could not confirm the bene-
ficial role assumed to be played by adsorbed sulphur vapor in the
lubrication mechanism, since sulphur evaporizes at temperatures when
MoS2 still lubricates. They think that surface contaminat.ions are re-
sponsible for the neutralization of edge valence. MoS2 lubricates up to
500 °0 in vacuum but in the presence of oxygen it oxidizes to MoO~.
This is a poor lubricant and welds form.
An important property of any solid lubricant is its adherence to
its meta} member in order to remain in the contact during sliding. MoS2
is superior to graphitein this respect; but talc, although slippery, fails
as a lubricant because it is too loosely bound to the members1 •
D. Adjuvants. Adjuvants are mixed into the brush material in order
to diminish the severe wear at high altitude conditions. Lubricating
organic impregnants that were initially tested, evaporated or in any
case did not remain in the contact. Then ELSEY [1] introduced impreg-
nants of metallic halides with promising effects. He was guided by the
idea that the adjuvants become decomposed in the sliding contact
.and help to build up a convenient collector film. However, only un-
changed halides have ever been detected in the collector films. Actually
the chemical reactivity of the adjuvants seems to have no influence on
their lubricating and wearing quality. For instance, Cdl 2 although
having a lower heat of formation than BaF2 , is nevertheless less effi-
cient as an adjuvant than BaF2 • It seems that the action of the ad-
juvants is mechanical.
MoS 2 is the best known high altitude brush adjuvant. MoS 2 powder
is mixed with a binder, rolled and then carbonizcd. This results in a
relatively hard mass where MoS2 is distributed as fine streaks within a
1 Reviews ofliterature on MoS 2 are given in BowDEN and TABOR [12] p.196ff.,
STOCK [1], McCABE [1], and Molysulfide Newsletter published by Climax Co.,
New York.
15*
22S Sliding Contacts
more massive hard skeleton. Fine part.icles or larger chunks of this

material are then mixed into a graphitematerial where they at'e firm!~·
held in place after baking. Brushes with MoS2 adjuvant immediately
film the collector ring and therefore can be used in high alt.itude without
a run in at sea Ievel. Due to their excellent filming property the wear on
copper and steel rings is only somewhat high er than with plain brushes
at sea Ievel.
Their action seems tobe as follows. At sea Ievel where brushes wit.h
MoS 2 are used in fractional horse power motors, the chunks in the
hrush wear about as much as the rest of the brush. At high altitude,
the severe wear of the graphite can not develop because the relatively
hard chunks, which do not much change in wear and take over enough
load, prevent the graphite face from wearing quicker than the chunks.
§ 40. Stick-slip motion.

The temperature in currentless sliding contacts
A. Stick-slip or jerky motion. Schoolboys can manage to make chalk
squeak when writing on the blackboard. The squeak is created by a
stick-slip motion. Chattering of motor brushes is another example of
the phenomenon. Actually, sliding has the tendency to take place with
discontinuities and if the system has a suitable natural frequency this
squeak tone may be generated by stick-slip vibrations. BowDEN,
LEBEN and TABOR in particular investigated stick-slip on metallic
contacts. Results of their and other workers investigations are re-
ported in BowDEN and TABOR [4] p. 105 and [12] p. 78.
SHOBERT [4] studied the particular kind of stick-slip that appears
as brush whistling and chatter on electric machines. We are concerned
with this brush behavior. Decicive features of the
a process are illustrated in Fig. (40.01) where several
simplifications are applied. The shaded area is a
rart of the ring that moves to the right. The line ab
represents the brush 1 . It is inclined by an angle e
0. from the normal of the ring surface because such an
inclination was shown to be necessary to create
Fig.( 40.0 I). Simplifted whistling. The Ioad, with the direction ab, is ap-
sketch to illustrate plied at a. During whistling, the lower end of the
whistling of brushes
brush vibrates along bc. To begin with, we assume
contact taking place between brush and ring only in b. In the moment
of contact this end of the brush gets an impulse to the right from
1 a being the center of the Ioad and b the center of the friction forces.
§ 40. Stick-slip motion. The temperature in eurrentless sliding contacts 229
the friction force. This compensates for the damping which occurs
during each cycle of the vibration and thus keeps the vibration going.
Let us now eliminate some of these simplifications. The real brush
face can not be represented by the single point b. Because of its ex-
tension, it may have some contact during the entire oscillation, but
with diminished pressure during the movement between b and c. Tbe
amplitude of the vibration of the brush endissmall (order of magnitudE'
10-s m and less). On slip rings in practice the maximum vibration
velocity of the brush end is small compared witb tbe velocity of thE'
ring, up to 50 or 100 times smaller in SHOBERT's experiments. Th11s
tbere is never any static friction.
The explanation of the whistling given above, does not include or
require that the friction coefficient decreases witb rising relative veloc-
ity between brush and ring, but it requires f) > a. certain minimum.
The investigations have stated the existence of a minimum for @. A
maximum, 6"" also exists, beyond which the impulse in b would Iw
directed essentially along ba and contribute too Iittle to tbe mainte-
nance of tbe vibration.
Testing brushes with the face narrowed to simulate the line ab.
SHOBERT could determine the angle fJ fairly weil. He found, with
different brusb grades, P ""' 1 N and v = 0.5 to 24 mfs, t.bat whistling
appeared in a fJ-range of little over 0° < fJ < 10 to 15°.
The frequency, about 3100Hz, corresponded to transverse vibra.
tions of a bar clamped at one end, baving the mechanical properties of
the brush, and being clamped where t.hc brush is fixed at tbe holder
(at its upper end).
This is true for ""whistling" contacts. As running time increases,
whistling degenerates to "chatter" with several frequencies super-
posed, for instance, a ·'wabbling" frequency defined by tbe loading
spring and the moment of inert.ia of the brush; see SHOBERT [7] p. 84
and VOLKMANN [1].
BoWDEN and TABOR [12] p. 79, referring to RABINOWICZ [5], diseuss
stick-slip under circumstances corresponding to fJ = 0. That is, under
the assumption of different static and kinetic friction coefficients. The
tests were made with a speed so low (order of 0.1 cmjs) that "stick"
evidently meant no relative velocity between the slider and base mem-
ber.
Investigators have observed the electric conduction through the
contact during the stick-slip. Just as one would expect. t.he conduction
is highest at moments of strongest friction. If during the slip a real
separation occurred in the contact, this would be apparent from the
contact voltage reaching values as high as that in arcs. Brush chattcr
is usually observed wit.h voltage peaks of the order of 2 t.o 3 volts, but
actual separations have also been recorded 1 • Also the contact temper-
ature has been observed with contact members that constitute a ther-
mocouple (see below). The temperature increases during the high speerl
slip 2 •
A condition for regular periodicity of the stick-slip is that the pro-
cess is about the samein every period. Plastic flow can obliterate this
condition, and make the phenomenon very irregular. Another change
occurs when a certain crit.ical speed is attained. Then, the sliding pro-
ceeds relatively smoothly.
The Contraposition of a heavy basal member and a light slider is not
a principal condition for stick-slip. This combination was chosen only
to simplify the description of the phenomenon. Stick-slip may also
occur between massive members. Then the elasticity of their bulk
materials may replace the elastic bending of the slider.
Contacts with a strongly adsorbed lubricant film slide with a smooth
motion. But if the temperature of the lubricant exceeds a certain limit3
its surface melting point - the motion proceeds in jerks.
B. The Temperature in currentless sliding contacts. Stimulating in-
vestigations are due to BowDEN and collaborators. A survey of their
results is given in BOWDEN and TABOR [4] chapter II and BowDEN [7]
and [9] with reference to the literature. Quite high temperatures ofthe
order of 100 to 1000 °0 appear surprisingly often in sliding currentless
metallic contacts in spite of the good heat conductivity of the metals.
The reason is that the heat generation is concentrated in small contact
spots. Because of their size, theyexhibit considerable thermal resistance.
BowDEN and RIDLER [1] used bimetal contacts as thermocouples for
indicating the rise of temperature generated by the sliding work in
contact surfaces. With rising speed the temperature increased, andin
some cases an upper Iimit was found. A further increase of speed gave
no change in the thermo-current. It was suggested that the melting
point of one member in the contact bad been reached. Fig. (40.02)
shows two of the curves that were recorded. The temperatures plotted
are averages. In a later work cathode-ray oscillograms4 showed that
the friction may be able to cause temperature flashes of high values.
Several attempts have been made to compare the measurements of
BowDEN and RIDLER with calculations of the kind described in § 21.
However, the available experimental data ar not sufficiently complete
for such calculations. Particularly, one does not know which fraction
1 In the Stackpole Laboratory.
2 MoooAN, MusKAT and REED [1] and [2].
3 It is remarkable that this limit seems to depend on the device used, see
T. P. HUGHES and WHITTINGHAM [1) p. 23.
4 BoWDEN and TABOR [ 4] Plate III p. 33.
§ 40. Stick-slip motion. The temperature in currentless sliding contacts 231
of the contact area is metallic and which is covered by oxide. Consider
for example the Observations that are represented by Fig. (40 002). With
f = 1, ~ = 1 in Eqo (I,17) for the determination of the circular contact
area and employing Eqso (21.09) and (21.14) under the assumption that
70 to 80% of the heat dissipates into the iron, one calculates tempera-
tures about half as high as were measuredo This was in spite of the
assumptions being chosen as favorably as possible to give high temper-
a ture valueso The main reason for the
discrepancy probably is that the friction
work in reality was not uniformly distri-
buted over t he bearing contact area as is
assumed for the formulaso Better agree-
mt-nt between observations and calcula-
tion has been reported by Eo HoLM and 't;
R 0 HoLM 1 working with contacts of nickel ~rooH-7--T----+---~
against brass and nicke} against tino Under (:3
the circumstances, realized in their experi-
ments and also in the investigations by 0 1100 800 !6C 1200
Vr:locify
MoRGAN et al. [1] the supertemperature
Figo ( 40o02)o Temperature iu a slid-
in the contacts never surpassed 50° 0 It ing contact between Iead and mild
should be noticed that the high temper- steel at different velocities, accord·
ing to BOWDEN and RIDLER [1 ]
ature flashes reported by BowDEN and
collaborators have been obtained with high sliding speed (several
meters per second) and are not general characteristics of sliding
metallic contactso
ARCHARD [7] made measurements with a steel pin gliding on a
steel ring (hardness 25. 108 Njm 2).A marked decrease in the wear rate
was attributed to the formation of martensite. This requires a quench
following a temperature rise of at least 700° and was observed with
P = 10 N and v = 2 mfs. Assurne steel to have a heat conductivity of
47 Wjm °K, a friction coefficient of f = 0.7 and all heat entering the
ring . Using Eq. (I,17) (with ~ = 1/2}, Fig. (21.04) and Fig. (21.10}, we
calculate a temperature rise of 1300° in the hottest point of the contact
area. This, according to Fig. (21.11}, corresponds to an average tem-
perature of about 1000° in the same areao ARcHARD finds 900° with
rougher approximationso
BowDEN and HuGHES 2 proved that the sliding of skis on snow is
due to a lubricating film of water that is melted by friction heato
Recently BowDEN and FREITAG [8] demonstrated an analogous pheno-
menon between steel and copper contact memberso In the latter case
1 Eo and R. HoLM [4] and R. HOLM [32].

~ See BoWDEN and TABOR [4] p. 65, and BowDEN [7] and [9]0
232 8liding Contacts
t.he rate of heat production must be very high, viz., the sliding veiocity
must be extremely high, namely of t.he order of 1000 mjsec. Table-
(40.03) summarizes some typical result.s, particularly that at sufficiently
high veloeity the frict.ion coefficient is very small in both kinds of con-
tact..
Table (40.03). Approximate f-values
Contact
material Air temp. - 130 - ;) 20 20 cc
Velüf·ity ~low 5 150 •'iOO m fseC'
~ki-snow l ....· 0.4 0.02
Cu-steel i ~ 1.5 0.2
§ 41. Fridional wear in metallic contacts without current

A. 'fypes of wear. The current itself does not intrinsically influence
t.he wear in sliding electrical contacts. Its influence arises from R I"
heat. that causes softening of the metaland supports chemical corrosion
m·, from the particularly strong heating of arcing followed by evapora-
tion. The high temperature also accelerates chemical corrosion. We
distinguish the following types of wear: First, with respect to the pro-
eess; mechanieal wear (abrasive and adhesive) and chemical we~n
(corrosion). Second, with respect to the amount. ; micro, mild, medium,
severe wear.
Abrasive wear is produced by protrusions and by particles attaehed
to one contact member cutting into the other member, forming grooves.
The material from the grooves may gather to ]arger wear fragment ...
that finally drop out.
We speak of adhesil'e wear when t.he removed material was seiz,,d
by means of cold welding without interlocking. The pl'ocess is illustrat.ed
in Fig. (41.01). Adhesion or cold welding
in a momentary contact spot .40 may
happen to be stronger t.han cohesion in a
surfaee A 1 wit.hin the member C. Then·-
fore , the grain betwecn A 0 and A1 adheres
to thc mem her B and follows it with t.he
consequence that sliding occurs in tlw
l<'ig. (4I .P:). Sliding contact with th~
membcr B moving to the right. Cold surface A1 • The grain may finally disen-
weid in the original contact area .11 0 ; gage from B aml become a free metallic
sliding occurs in the surface A 1
within member C wear fragmeut . We call this a prefrag-
ment because the observed wear particle;;
are composed of many prefragments that have been kneaded together.
This was shown by HIRST and LANCASTER [3], CocKs [1] , [2] and
§ 41. Fril'tional wear in metallic contacts without currPnt 233
STEIJN [1]. Those final fragments which constitute the major portion
of the wear become greater than the a-spots where the initial weid
occurred. During kneading, heavy deformation makes base metallic
fragments pronc:- to oxidize. This explains why analy?.Nl rlehris largely
consists of oxide according to KERRIDGE [1].
Adhesive wear Ieads t.o roughness of the contact. surfaces, providing
the opportunity for interlocking and further wear of the abrasive type.
There is an important exception from this development. Boundary
lubricatcd pistons and cylinder walls are smoothed as if loosened par-
ticles, protected against oxidation by the lubricant., have thP oppot·-
tnnity to fill depressions and adherc there.
Wear types d·istinguished with respect to the depth of the wear 8ra1'.
lt has heen shown experimentally by R. HoLM (see Tahle [41.071},
HmsT and LA:~WASTER [3], ARCHARD and HmsT [3] aml others tlmt
for a given type of contact the amount of wear is fairly weil propor-
tional to the Ioad. P, to the distance of sliding, s, and essentially itt-
Jependent of the apparent area of contact, A". The iatter fcaturP is
:mrprising. One would expect that the nurober of spots incrPascH atHl
t he average specific depth D decreases with increasing A", then•hy
reducing the wear. However, we recall a counterpart to this indepeud-
ence of A," from the behavior of the resistances of nominally tlat eou-
tacts. They also appear to be fairly unequivocal functions of P, inde-
pendent of A", as is illustrated by the curvcs for nomina.lly ftat piates
in Fig. (8.01) and by Fig. (:~5.10). Considering that under simiiar con-
ditions, hard materials wear less than soft oneR, we expeet 1/11 to lw a
eharacteristic factor in any wear formula.
Summarizing: In any type of wear we expect. the amount of \war.
W, to bP exprPssiblP approximat.ely as
where
W oZ~ l (41.02)
z = WH
Pa
f
Z is a dimensionless number. R. HoLM1 has shown that 2. tooz }1as
an illustrative significanee. Imagine the total volume W of the frag-
ments that. are worn off on passing the slider once, as being uniformiy
spread on the true sliding track as a coherent film. We express the
thickness of the imagined film as a number, Z, of atomic layers. That
is, we express Z in the unit of one ntomic spacing. Then Z equals
approximately 2 · 106 Z. We use Z to signify the wear types with re-
spect to t.he character of t.he wear scar and call it a dimensionleAs rept·P-
1 R. HoL~r [.31'] p. 24 7.
sentor of the wear, defined by Eq. (41.03)

Z=2·10 5 WH (41.03)
Ps
Eq. (41.03) i:.; approximately valid for a bimetallic contact if His tlw
hardness of the softer member determining the width of the scar.
ARCHARD [1] presented a model of the wear process leading to an
interesting interpretation of Z, although with a wry restricted appli-
cation. He assumes that every time two asperities (considered as hav-
ing circular cross sections) meet, there is a probability p that an ad-
hesive fragnwnt (considered as a hemisphere) is formed. He finds
p = Zj3 = Z/6 · 10 5 (41.04)
Examples to Table (41.05) are given in Table::-; (4t.09) and (41.11).
Kotice: If Z is the same on metals with different H. viz. different width
of the tracks under otherwise equal
Table (41.05). Types of wear signified circumstances, the total wear vol-
by a dimensionless wear rate repre· ume W is greater on the softer
sentor Z metal; namely in proportion to 1JH.
z wear type B. Details of the formation of
wear detritus. KERRIDGE [1] made
<2 micro
2 to 25 small experiments with a radioactive,
25 to 100 medium annealed steel pin rubbing against a
>100 severe hardened steel ring, and found that
the softer material was transferred
to the ring, Hidently by welding, and that it formed a surface layer
which attained an equilibrium state, i. e., a certain average thickness.
This did not mean that the wear process ceased. If now the radioactive
pin was replaced by a non-radioactive one, without disturbing the layer
at the exchange, the radioactivity of the layer soon disappeared but its
visible appearance was not altered. This sho\vs that the wear from the
layer still was balanced by transfer from the pin, this time the non-
radioactive one.
Evidently, the metal film experiences a steady wear and renovation.
In addition, it is also oxidized. The oxide breaks and the wear debris
consists of fragments from the oxidized film. Thus, it is not directly
detached from the body of the members. KERRIDGE, in his first tests,
used an annealed steel pin against a tool steel ring; the wear was small,
particularly in vacuum where the film could not oxidize. KERRIDGE
and LANGASTER made similar experiments with brass pins and meas-
ured about 20 times greater wear as analyzed by the function Z. The
wear was predominantly metallic 1 , STEIJN [1] ancl RowE [7] worked
1 Cf. Table (41.07).
§ 41. Frictional wear in metallic cont.acts without current 23i'l
with brass against steel, confirming the earlier result.s with the added
feature that a small back t.ransfer complicates the proct>ss. In argon,
strong brass lumps are built up on the steel. In air the brass layer is
weak and interlaced with slightly oxidized surfaces.
It is remarkable that, in argon, thP transfPr is great, but the n<>t
wear relatively small; see STEin< [1].
C. The inßuence of adhesive wear on frietion in elean metallic con-
tacts. Naively thinking, one would expect wcar to always increase
friction. However, this is not generally true. As illustrat.ed in Fig.
(41.01) and described in Section A, wear grains can be produced by
cold welds in contact spots when the weid need a grt'ater force for
shearing than any other surface. Then, sliding will occur in t.he surface
of easiest shearing, for example, in A1 in the figure. Later, the particle
between A0 and A1 (double shaded) may hreak off from Band form a
wear grain. If the exchange of thc sliding surfaees, A0 and A1 , does
not Iead to an increase of the shear force beyond that in the initial
spot, the friction coefficient f will not be affect.ed. Besides t.he work
for shearing, the friction work will comprise the deformation energy that
is required to break off grains. Howeyer, the latter may be relatively
small. Where an increase of f, combined with wear is obserYed, it will
often be a secondary phenomenon resulting from a change of the cha-
racter of the contact surfaces by the wear. For inst.ance, this happens
when the contact surfaces become roughened by wear which Ieads to
interlocking.
D. Why liquids, even the deposit from air humidity, are able to
strongly inßuence wear, without appr~ciably aß'ecting the friction coeffi-
cient. The transfer of a fragment from one contact member to thc
other, which is the introduction to adhesive wear, is illustrated in Fig.
(41.01). It is supposed that the shear force is smaller in the section
A1 than in the initial contact spot A0 • Let us now imagine the picture
changed so that A0 no Ionger represents an undisturbed weid. A number
of alien molecules or groups of such molecules, for inst.ance water mo-
lecules, may in some way be inserted in the interface A 0 • Even if they
cover only a small fraction of A0 , t.hey will diminish the ewerage shear
strengthin A0 • Very likely a moderate decrease of t.his strengt.h will
suffice to make A0 the sliding surface inst.ead of A 1 • The loosening of
the wear grain between A0 and A1 thereby is prevented. It is plausible
that a sufficient number of water molecules is provided from the nor-
mal humidity of the air; cf. No. 3 in Table (41.09).
It may be noticed that RÄDEKER [J] and [2] observed a minimum
of metallic wear at temperatures which are high enough to aceeierate
oxidation but sufficiently low to prevent softening of the metal (for
iron between 200 and 350 °C). At higher temperatures, when the metal
236 ~liding Contacts
becomes soft., wear increases. Probably, the cause for the mmmmm
wear is an oxide film so thin that it contaminates the contact surfaces
just enough to diminish their tendency to weid; thicker oxide films
t.hat are able to produce abrasive wear debris seem to increase wear.
E. Frictional wear in currentless sliding contacts; represented by z
which is calculated according to Eq. ( 41.03) as a function of the hardness,
11, of the softer member. Tables (41.07), (41.08) and (41.09) summarize
observations of wear expressed by Z. The values of P, Hand Z are
t.abulated, and thus the rate of wear, Wfs, that has been observed, can
readily be calculated from the tabulated data. All observations refer
t.o the run-in state, meaning that Wfs was considered as remaining essen-
t.ially constant during the test period 1 .
Thc major portion of the data are due to R. HoLl\1 2 • He used a
device that is sketched in Fig. (41.06). R is a rotating ring against
which the slider B glides. The
holder of B is carried by the
Iever H that is controlled by the
springs Y, Y. The friction force
can be determined from the de-
viation of H along the scale S, or
from a compensating regulation
of thc springs so as to keep H
horizontal. In order to facilitate
Fig. (41.06). Devicc for measurement of wear and
thc determination of the wear
friction coefftcient in the contact between a brush somet.imes such means was used
Band a ring R. Tbc edges of the brush are faceted
in ordcr to facilitak thc det.ermination of the wear
as : 1. faceting the slider in its
edges, thus providing a small
area subjected t.o wear for the geometrical determination of the wear,
2. making the slider glide along a wire stretched round the wheel R.
The wire could easily be removed from the wheel for the determination
of its wear by weighing. A correction was introduced in most cases for
the initial wear, which usually was greater than the wear in the run-in
stat.e. Data of inYestigators other than HoLM are signified by foot-
notes .
Disc-u8sion of Tables (41.0i) and (41.08). The following major in-
format.ion is to be noticed.
1. Under otherwise equal conditions, Z is seen to be independent of
P. In other words, the wear rat.e is proport.ional toP, cf. ARCHARD and
HIRST [3].
1 ARCHARD [5] used crossed·cylinder sliding contacts moved so as to steadily

expose virgin contact spots on both members. By this method he always measured
the relatively high "initial" wear.
2 R. HoLM [30] § 44.
§ 41. Frictional wear in metallic contacts without current 237
Table (41.07). Mechanic(tl wear in cu1·rentleaa alid·ing contacta rep1·eaented by Z for the
softer contact member; wire or ring. Run-in state. Speed v, of t.he order of a few m/s
In air at room temperature

Slider material meta! of wire or ring, and hardness, H, in 10• N/m•
Contact Ioad P
and hardness
Hin 10• N/m• Fe Cu
I
Ag
IAg ~oft I Al Zn
N 15 to 24
I 6 5 2.7
I 2.4
steel 150. 5.2; 32 2 I

62 45. 41 ; 28" 7.2 3 (7)3
11. 13.6 Ul
1. 8 6.8 1.8 0.4 128
.15 8 6.0 2.4 132 368
- - - --- --- - - -
glass 5.3 4 1.2
25 1. 3 13.6
.15 4.8 13.2 1.6 196 400
- - - - - - --- ---
Ag 1. 36 40 24 12; 36
9 .15 64 68 68 6.8 320
in hydrogen
steel, 62 1.2 16 12 52 330

Ag, 9 1.2
I 4 136 I I 112
I 240
in H 20 or alcohol
steel, 62 1.23 6 30
Ag,9 1.23
I I I I I 10
steel, 22, in hexadecane
steel 270. 0,054 Pa= 3.5 • 108 N/m 2

55 860. 1,14
I Pa = 11 • 108 N/m 2
steel, 90, in air5
brass, 14 40. 130.

I
steel, 90, in 10-s Torr5
brass, 14 40. I 1.1 I

1 BONDI [1].
2 MAILÄNDERand DIEs [1]. The !arge values of Z, viz. 28 or 32 are exceptions,
the character of which has not yet been explained.
3 SIEBEL and KOBITZSCH [1].
4 ßURWELL and STRANG (3] Fig.11 and 15.
5 ÜORNELIUS and ROBERTS (J).
Table (41.08). Z for the karder contact member, the slider
Cf. Table (41.07)
In alr at room temperature
Slider material meta! of wlre or ring, and hardness, H, 108 N/m2
Contact Ioad P
and hardness
Hin 10• N)m• Fe Cu Ag IAg 3soft I Al Zn
:N 15 to 24
I 6
I 6 2.7
I 2.4
steel 11. 0.12 1.2

62 1. 10 0.12 1.2 0.01 +
.15 6 0.04 0.02 0.6 +
- -- -- -- -- -
glass 1. 0.4 16
25 .15 4 6 0 0.6
- -- -- -- -
Ag 1. 30 10 7 4
9 .15 60 10 + 1.0 0
in hydrogen
Rteel 1.2 10 0 0 +
--
Ag 1.2
I1 I8 I I1 I 1
in water or alcohol
~teel 1 •) 2 0.6
I I
.~
Ag 1.2 I I I 0.4
2. The barder contact member wears much less than the softer one 1 •
If both members have about the same H they have fairly equal Z-
value~>, forexample Ag (H = 9 ·108 Nfm 2 ) against Cu (H = 6. 108 Nfm2).
The more pronounced, but still small difference between Ag (H
= 9. 108 N fm 2 ) and Ag (H = 5. 108 N fm 2 ) suffices to diminish the wear
of the barder member by a factor of about 4.
3. The data by BuRWELL and STRANG [3] on steel in hexadecane are
represented by only two points on their curve. The purpose of citing
these observations is to show the order of magnitude of the observed
wear. It lies in the range ofmicrowear a.ccording to Table (41.05), prov-
ing that the liquid has acted as decribed in Section D, above2. Partic-
ularly noticeable is the following result. As long as the average pres-
sure Pa, referred to the apparent surface (i. e., load P divided by the
worn-in surface) in the contact, was below H/3""' 7-108 Nfm 2, Z was
independent of P and Z ""' 0.05. But raising Pa (by increase of P) be-
1 BoWDEN and HUGHES, sec BowDEN and TABOR [4] p. 62, emphasize that the
hardness data must refer to the temperature of the contact. It happens that a
material, that is harder than another one at room temperature, is the softer one
at a high temperature.
2 The coefficient f probably was of the order 0.5 to 1 or otherwise the authors
would have remarked on it.

Table (41.09). Jlechanieal wear represented by Z. Run-in state.
I I I
~0-1
Type
Contact Hardness Speed Contact
loadP z Remark Observer of
material
108 Ntm• rn/sec N
I I wear
iron 15 1 1 1800 in dry air HOLM
1 severe
on iron (15) et al. [24]
- ---
2
iron 15 1 1 5400 in dry air HOLll severe
on iron (15) at 200 oc et al. [24)
-- 15 1 1 in damp air
---
3 iron 12 HOLM small
on iron (15) et al. [U)
--- ---
brass 15 220 110 KERRIDGE
4 on 0.045 and LAN· severe
steel 80 CASTER [2)
---
brass 10 1.8 10 40 ARCHAl\D
5 Oll and mellintn
steel 80 HIRST [3)
brass 10 1.8 3 7 ARCHARD
6 on
steel I 80
and
Hrnsr [31
small
gold RABINO•
7 Oll 0.1? 0.5 180 WICZ [7) severe
gold Fig, (6.27)
-- ---
gold RABIXO·
8 Oll 0.1? <0.04 ""'0 WICZ [7) micru
gold Fig. (6.2i)
---
brass 14 air CORXELIUS
9 Oll 3 40 160 760 Torr and severe
steel 90 ROBERTS [1)
Fig. (2)
---
brass 14 CORNELIUS
10 Oll 3 40 1.1 10-• Torr and
ROBERTS [1) tnicro
steel 90
Fig.(2)
-- ---
iron wire 10 2 in dry air R.HOLll[28) micro
11 Oll 0.03 1.5 and F. L.
white meta! 3 24
Wl\1 80 small
---
iron wire 10 0.6
12 Oll 0.03 1.5 in damp air HOLJ( [28) micro
white meta! 3 20
WM80 small
---
steel on 20 12.3 1.2 boundary KLUGE [2) micro
13 lubr.
cast iron
---
iron 15 boundary
14 Oll 0.2 1 0.4 lubr. HOLM [28) micro
iron (15)
---
electrographite 2.6 1. w-a HOLM
15 brush on 10 5 et al. [24) micro
copper ringt 8 0.1. 10-• Table (44.03)
metalgraphite 1" 0.05 HOLM
16 brush on
copper ringt 8
110 5
0.05
et al. [24)
micro
1 In the case of hydrodynamic brush lifting, the air pressure force has to be
subtracted from the applied Ioad P in order to obtain the accurate load for Z.
Cf.§ 44G.
yond H/3 caused a rapid increase of Z. The point of Z = 1.1 lies on

the rising branch. It is not surprising that the critical Pa equals the
yield point of the metal, which is ""'H/3, cf. §I.
4. Not only hexadecane, but also alcohol and water are capable of
diminishing the wear. A hydrogen atmosphere as weil as a moderate
vacuum diminishes wear, evidently because oxidation of wear debris
is prevented.
5. The soft metals Al and Zn exhibit great adhesion wear even
against the hard steel. The reason is that the soft metals are smeared
out on the harder member ( + in the table indicates that the member
in question has gained weight) and that the sliding proceeded on rela-
tively great load bearing areas.
The contacts between steel or glass and Fe or Cu had I ""' 0.5; but
with Ag (H = 9.108 Nfm 2) against Cu, or with Al or Zn against any
harder metal I""' 1 was found, indicating that welding appeared.
Discussion ol Table ( 41.09). It is commonplace that the dry friction
wear is particularly great if both members are of the same metal. The
explanation is as follows. The specific friction force '1fl, the average pres-
sure, and consequently the contact area, are prescribed by the strength
of the softer member. But, the harder member prescribes the structure
of the surface. If its own surface is smooth, the contact surface remains
smooth without interlocking. It is otherwise when both members are
of the same metal. Then, p and '1{J can produce plastic deformation in
both members; the contact surface is in a labile state and will attain
a wavy structure, leading to interlocking and high wear. Table (41.07)
does not contain any contact with the same metal in both members,
but in Table (41.09) Nos. 1 and 2 are typical examples. They are parti-
cularly interesting because they refer to cases where both members are
hard. Their great hardness is not able to prevent the interlocking and
high wear when the members can deform plastically. No. 3 deq1onstrates
the statement of § 41 D. In No. 12 the humidity influences only the
iron.
The influence of P on Z is demonstrated by Nos. 4, 5, 6, and still
more impressive by Nos. 7 and 8.
CoRNELIUS and RoBERTS [ 1] state a surprising decrease of wear
in vacuum, evidently because of diminished oxidation: cf. the obser-
vation in H 2 mentioned in Table (41.07).
F. Size and frequence ot wear fragments appearing during periods
of adhesive wear. The size of wear fragments has been computed from
autoradiographs and more directly observed by means of electron micro-
graphs. Table (41.10) summarizes typicallinear dimensions belonging
to the group of relatively great fragments which constitute the major
part of any arlhesive wear. Around these values, one finds a statistical
spread with a standard deviation of the order of about 30%. A great

number of very small fragmentsalso appear; they, however, contribute
little to the total wear volume.
The fragments from adhesive wear are not arranged in rows along
the sliding tracks, as they would be when ploughed out. Along any
track line they are so sparsely distributed that the production of any
wear fragment per encounter between elevations ofthe contact members
evidently is a very rare event. ARCHARD [8] computes that during
severe wear, about 1% ofthe encounters contribute to the wear, where-
as insmall wear, less than 0.01% may be active. Another computation
in R. HoLM [ 37] p. 244 concerning medium wear yielded 0.5%.
Table (41.10) shows the average linear dimension of dominating
wear fragments under different sliding circumstances.
Wear particles are subjected to cold welding joining a contact mem-
ber. RABINOWICZ [7] assumes that the main reason why such once
Table (41.10)
Linear dimen-
Bardness sion of typical
No. Contact Speed Load z
II wear fragments Author
material
10' N/m' rn/sec N w-· m
brass 15 .0045 220 110 9 KERRIDGE
1 on and
steel 80 LANCASTER [2]
brass 15 .0045 2.5 140 1.6 KERRIDGE

2 on and
steel 80 LANCASTER [2]
brass 15 4 · 10-• 220 1350 3 BIRST

3 on and
steel 80 LANGASTER [J]
4
brass
on
15 0.65 220 160 1 I and
BIRST
steel 80 LANCASTER [J]
brass 15 1.8 10 40 1 ARCHARD

5 on resist. and
steel 80 metallic BIRST [3]
brass 15 1.8 3 7 0.1 ARCHARD

6 on resist. and BIRST
steel 80 ""10 [J] alld [4]
gold - - 0.5 180 (1) RABINO-

7 Oll WICZ [7]
gold Fig. (6.27)
gold - - 0.1 40 (0.5) RABINO-

8 on WICZ [7]
gold Fig. (6.27)
9
gold
Oll
- - <0.04 9 - RABINO-
WICZ [7]
gold I Fig. (6.27)
16a Electric Colltacts, 4th Ed.
welded fragments become loosened is peeling off by virtue of stored

elastic deformation energy. Wear particles can be so small that the
energy stored in their volume is insufficient to peel off the welded area.
Then the wear is very small. No. 9 in Table (41.10) exemplifies this
effect at P below a critical value of about 0.04 N on sliding noble
metals: No overalllass of weight occurs and no wear fragments form;
allegedly because the wear particles are too small to be peeled off. Base
metals "cheat" by forming small oxide particles that do not weid. One
would recommed wire brushes of noble metal with a Ioad per wire of
less than 0.01 N. Confirmation fo this idea from the practice is still
missing 1 •
G. Behavior of sliding contacts of measuring apparatus. Data for
switch rheostats and resistance bridges are given in R. HoLM [ 34]. Typ-
ical cases are represented in Table (30.11). In Table (30.11) contacts
between base metals are represented. For many modern instruments
contacts between noble metals are chosen. At a smallload of the order
of P = 1 g = 0.01 N, Au-Au contacts function satisfactorily in air. But
at Ioads of P > 0.1 N, cold welding effects a noisy contact and harmful
wear. Concerning lubrication of Au-Au contacts see p. 219.
BAYER et al. [1] report good results with a Rh slider on Au in air at
P = 0.01 to 1 N and unlubricated. The wear of the gold seemed tobe
of the type small to medium as termed in Table (41.05). The wear of
the rhodiumsliderwas too small as tobe measured. During sliding the
coutact resistance decreased to about 0.3 to 0.1 of its initial value and
then became equal tothat of a Au-Au contact in air.
§ 42. Electrical performance of carbon brushes on rings and

commutators when arcing is excluded
A. Introduction. Electrical performance of sliding contacts in generat,
Although electric conduction through a given metallic contact area is
independent of whether the contact is stationary or moving, there are
nevertheless many reasons why metallic contacts actually behave differ-
ently under stationary or sliding conditions. With clean metallic sur-
faces thf'1 contact area usually increases at the commencement of sliding.
Among other phenomena causing a difference is the differing tempera-
ture rise because of the Rl 2 heat that affects the resistivity. The in-
fl.uence of the RI2 heat will be relatively low in a sliding contact be-
cause any a-spot is merely transiently active and the temperature will
1 Cf. § IJ and a remark about experiments by SMEKAL[l] in R. HoLM [37] § IC.

§ 42. Electrical performance of carbon brushes on rings and commutators 243
not have time to reach equal values as in the stationary case with the
same current. However, the sliding contact is heated not only by the
current but also by friction.
The particular contact between a carbon brush and a copper collector1 in
air. The greater part of this cha pter is devoted to the discussion of this
important contact. Thereby, the problern of the familiar darkfilm on
the copper collector comes to the fore. Although the film is regarded
as essential for the performance of the brush, no general accepted name
has been given to it. W e shall call it collector film. It will be proved
that the theory 2 which R. HoLM proposed in the early thirties is sub-
stantially correct. He said that the film is practically insulating where
it is undamaged and that the conductivity through the brush contact
is essentially a result of fritting the collector film. The fritting Ieads to
metallic spots the size of which is dependent on the current in such a
manner that the contact voltage assumes values of 0.5 to 1 V.
B. Early investigations about the conduction mechanism in the
brush-ring (copper) contact. A question that was the subject of
h 9
Fig. (42.01). Voltage records of two electrographite brushes (anodic and in parallel) running in
the same track on a copper ring. Under one brush a portion of the collector film on the copper
ring bad been strongly fritted. The other brush did not respon<l to the fritted regions
argument in the thirties was whether the current transiently changes

the resistance in contact spots or produces Iasting changes. 1. BAKER
and HEWITT [5] applied A. C. on a brush against a film covered copper
ring and rotated it synchronously with the current. Thus, certain spots
under the brush experienced only low currents and certain other spots
the maximum amplitude of the current. The authors then analyzed
the track with a weak D. C., registering the contact resistance. They
found that the resistance was high where the forming current had been
low and relatively low where the forming current had been high. Now-
1 Weshall use the term collector as denoting both slip-ring and commutator.
2 R. HOLM [29] p. 195.
adays we would say that the a-spots had adapted themselves to the
respective current by B-frittings.
A more complete demonstration1, due to R. HoLM, is given in
Fig. (42.01). It records contact voltages of two electrographite brushes
A and B (both anodic) placed diametrically opposite each other on
the film covered copper ring, but running in the same track. On a short
portion of this track brush A had been loaded with a strong current,
producing a low resistance. Every time brush A, now carrying a much
smaller current, passed this position (labelled fritted in the record) its
contact voltage feil to nearly zero. One expected the brush B to behave
in the same manner as it passed this discrete region; but it did not.
This shows that every brush essentially2 met its individual group
b 272
Fig. (42.02). Voltage record of a sliding contact between a graphite brush and a copper ring
covered by a highly resistive film. P = 4 N , I = 0.8 A, 810 rpm. The records 2, 5, 270 and 272
(figures refer to the number of revolutions) are placed one under the other for comparison. The
initial voltages before fritting are recorded in the part called start. Repeated imptovement of the
conductlon by fritting can be followed at a, a, a, a. At (b, b. b) a returning peak indicates a portion
on the ring that obviously did not frit
1 R. HoLM [30] Fig. (41.04).
2 Only the voltage drops labeled g and h appear in both records.
of current adapted a-spots on the elevations of the ring. It was con-

cluded that the individual waviness of the brushes was responsible for
this effect.
It was likewise surprising to find that the picture of the topography
of the ring is weil repeated after hundreds of revolutions. This is shown
in Fig. (42.02) 1 by 5 sectors of a voltage record of a graphite brush on a
film covered copper ring. The thick line labeled A in the second revo-
lution, corresponds to the equally labeled ring portion ofthe first revo-
lution. Each revolution reaches as far as the thin base line. At the start,
the voltage is about 1 V, and at the repeated fritting of spots in the ring
region labeled a, it has dropped to about 0.5 V. The record at a, a, a
shows that conduction increases in steps obviously because B-frittings
(i. e. adaptation of a-spots to the current) are incomplete during one
revolution because of the short activity of an a-spot during any
pass. At b, b, b a returning voltage peak indicates an asperity on the
ringthat had too small a top area for adaptation to the current.
Summing up: The experiments cited above prove that the current
produces Iasting conductance through the highly resistive collector
film and that the conductance improves as the current increases, viz.,
that the a-spots become larger so as to keep the contact voltage at
about 0.5 to 1 V. The generation of a-spots and their widening as the
current increases are typical fritting phenomena and HoLM's hypo-
thesis, indicated in the Introduction, means that an equilibrium 2 exists
between steadily competing oxidation and B-fritting.
HUNTER-BROWN [1] in 1919 and ScHRÖTER [1] in 1927 reported
that the film in the respective contact had to be broken down to pro-
vide practically needed conduction.
C. Polarity e:ft'ect in the contact between a graphite brush and a
copper ring. This Section is particularly due to investigations by
E. HoLM[10], [11], [12], [14], [15]. It essentially concerns the sliding of a
fine grain electrogra phite grade 3, either against rings of the samematerial
or rings of gold and copper.As to the cleaning ofthe members, and the
device for simultaneous measurements of the friction coefficient, f, and
the contact voltage, U, see E. HoLM [ 13].
Testing a cleaned copper ring it was found that it carried oxide
films, about 5 A thick. This was estimated from resistances being grea-
ter than those with a gold ring under the same conditions. Electrons
tunnel through such a film. Indexes + and- (for example U+ and U-)
always refer to the ring being anodic or cathodic respectively.
1 R. HoLM [29] Fig. (38.03).
2 R. HOLM [29] p. 195 and R. HOLM [30] p. 225ff.
3 This grade is manufactured at Stackpole Carbon Company; its material con-
stants are given on p. 151, footnote 1.

Sliding occurred pratically without vibrations. After a short time

of sliding on a copper ring, a difference between u~ and u- appeared.
This effect, is shown in Fig. (42.03): u~ shows fluctuations between a
minimum of 0.4 V and a maximum of about 1.0 V. On the other hand,
10
( ~M aA !AJI\\
as - --- -· ·~· ...... .
:::, 0
-05
-1.0
Fig. (42.03). Due to E. HOLl! [15] Fig. 1. Simultaneous L"·l·rec:orclR with a strong polarity effect
having developed after 1 hour of sliding on oxidized copperring (film about 150 ..\ thick). Voltage
tluctuations ( U+) appear under the brush on anodic ring between 0.4 and 1 V indicating strong
oxidation anrl thus fritting. Under the brush on cathodic ring oxidation can not develop and c-
is almoet enmtant = 0.4 V
1
Q
as l
I
I
i
I
I
! I'
2 I
i'
! ! ·JA
t:
,
0
1 '-., SA
<>:: !
I
I
I
..........
00. ....
0
I I-10A
l o ..............~
z1A 0 ....
0
00'2
O.D1 I
I i
I
01 01 o.s 2 s
P-
Fig. (42.04). R P-curves between nominally flat members; - - - b etween graphite membe" at
rest; o o obetween graphite brush and copper ring freshly cleaned, at rest; x x ~ sliding I km
between graphite brush and copper ring previously cleaned; the solid horizontal lines - refer to
t
the brush on cahodic ring, sliding at different currents (separate tests) until a polarity difference
in U bad developed as r ecorded in F ig. (42.03). See Fig. (42.05)
u- is practically without fluctuations; it is almost constant = 0.4 V

(the same as the minimum voltage with the other polarity). It did not
matter whether the oxide film on the copper was thin1 or thick; the
only difference with a thick film was that the respective etfect needed
a Ionger distance of sliding to develop.
1 Fig. 5 in E. HOLM [H] is akin to Fig. (42.03) but refers to a ring freshly
cleaned (5 A thick film).

How polarity difference develop8 during sliding is demonstrated by

Fig. (42.04). The dashed curve, measured between nominally flat mem-
bers of graphite , serves merely as a eomparative curve. During sliding
the first km on a previously cleaned copper ring , conducting spots are
available on the ring for both polarities. They produce almost ohrnie
resistance, R, equivalent to R with the members at rest; cf. the crosses
with the circles. However, the voltage, U, thus also R soon rise for
both current directions. This indicates that the tunnel-eonducting oxide
films on the spots have beeome thicker, and perhaps also the a-spots
smaller than the load bearing area.
Finally, tunnel conduction essentially ceases under the brush on
the anodic ring , and voltage peaks often up to 1.5 V prevail in air.
:F' ig. (42.05). Due to E.HOLM f 14], Fig. S. Simulatneous U·l·records with graph ite brushes (deftned
on p.151, footnote 1) after slid ing about 10 km on copper ring, previou s l~· cleaned for each measure-
ment. The pictures are thc middle parts of completc records aki n to th" rccords in Fig. (42.03)
Under the brush on the cathodic ring , voltage peaks also appear, but
they are lower and soon disappear; u- decreases to an almost constant
value of 0.4 V. The polarity effect as shown in Fig. (42.03) ha8 devel-
oped. This effect appears at various Ioad P , various brush grades 1 and
currents I. Fig. (42.05) shows the effect with different I at constant P.
The solid horizontallines in Fig. (42.04) show the long known fact that
the contact resistance becomes smaller with increasing current and
independent of P.
The explanation of the polarity effect is given by E. HOLM as fol-
lows: Frictional oxidation obviously is considerable under the brush
on the anodic ring indicated by the voltage fluctuations above 0.4 V.
It seems that many contact spots are insulating at encounter with the
brush and have to be fritted for adequate conduction. It has been des-
cribed in § 27 J that positive ions from the matrix metal 8,re moved
through the resistive film forming metallic filaments between the mem-
bers. This time dependent process requires obervoltages during sliding.
1 See Fig. 9 in E. HoLM [14]. It shows the polarity effect for brush grades on
a lamp black base as weil as on a petroleuro coke base, also for a pure natural gra-
phite grade.
During sliding, the metallic filaments break and and their clean
ends being rubbed for a short time, rapidly oxidize. This can explain
the occurance of transient high voltages whenever fritting is needed.
Mter current reversal, the appearance of the polarity effect in ac-
cordance with Fig. (42.03) takes an appreciable time dependent on the
current (both sliding and stationary). This indicates that rectification
due to semiconducting films is not involved.
In vacuum, the voltage fiuctuations disappear, andin addition, the
voltage ofboth brushes decrease below 0.4 V .Fig. (42.06) shows that U
of both brushes has decreased
7.0 to about 0.25 V~ The de-
crease below 0.4 V indicates
that a protecting film has been
worn off the a-spots resulting
in an enlargement of the con-
ducting area.
-ID
The existence of a thin tun-
nel-conducting film an the a-
Time - spots during sliding (when the
Fig. ( 42.06). Simultaneous U ·l·records with the temperature in the contact is
brushes in vacuum on oxidized copper ring (film
about 150 A thick). Both U+ and u- decrease to about 50 °0) always is indi-
about 0.25 V ( < 0.4V in air) similar to Fig. 10
in E.HOLM (1 5].
cated by the shape of the re-
versibles; see E. HoLM [14]
Figs. (5) to (7). This film is not worn off during smooth sliding 1 in air.
B-frittings leading to adaptation of the a-spots to the C'urrent. When
sliding has developed to the stage pictured in (42.03), a sufficient num-
ber of a-spots are produced on the ring and adapted to the current. The
lowest adaptation voltage in air is about 0.4 V corresponding to a
temperature in the contact surface far below the softening temperature
of copper (I< 20 A per 0.3 cm 2 brush face). Thus adaptation occurs
by virtue of the force F according to Eq. (27.04). This force seems to
be just able to permanently maintain conducting spots (under the
brush on the cathodic ring) by recessing overlying disturbing films
except a thin oxide film adjacent to the copper.
Under the brush on the anodic ring adequate a-spots characterized
by a low valtage are present only occasionally. This is seen in Fig.
(42.03) when u+ drops to about 0.4 V.
1 Only for relatively short periods (occurring at random)

this protecting film
seems to be more seriously damaged; see E. HoLM [12] Fig. (7), and M!DGLEY
and TEER [1] who observe roughening effects during these periods.
§ 43. The temperature in the sliding contact between a carbon brush... 249
§ 43. The temperature in the sliding contact between a carbon

brush and a copper ring or commutator
The supertemperature in the brush-collector contact is harmless

when it corresponds to the average contact voltage. But, there are
moments when the brush makes inadequate contact. particularly just
before leaving a segment. In such moments, transient high voltages
and temperatures appear. These events cause wear. The complicated
features of the problem, with computations of temperatures and their
effects, will be indicated by two numerical examples. Several simpli-
fications are applied.
The brush has simultaneously contact with many spots on the
collector. Any electrically conducting spot (a-spot) constitutes only
a portion of its Ioad bearing area, A. The a-spots under the brush are
regarded as far enough apart from each other that their constrictions
do not infiuence each others fiow lines.
Weshall consider a single a-spot and its Ioad bearing surroundings.
The warmest isotherm is located within the brush. Therefore, the heat
fiow across the contact is directed into the copper, and the total fric-
tion heat fiows into the collector. The question of which portion of
the RI2 heat that is developerl in the carbon will fiow each way, requires
a special consideration.
Examples: Material constants for copper: Am= 380 Wfm °K; c",
= 3.4. 106 Jjm 3 ; for brush: e = 4.5 · 10- 5 Q m; A = 30 Wfm cK; c =
1.5 ·106 Jjm3; H = 2. 108 Nfm 2 ; ~ = 0.3 for Eq. (1.17).
Example 1: Normal average electrical conditions with a contact
voltage U = 0.8 V; a load bearing spot, A, (on the copper, with the
average radius a1 = 7 . 10- 5 m); on A an elliptical a-spot with the
semi-axes r:t. = 1.2 . 10-4 , ß = 1.3 · 10- 5 m, thus the average radius
ac = V~ß = 4 ·10- 5 m and y = 3; see § 4F. The a-spot carries 4 A.
The peripheral speed of the collector is 30 mfsec and the a-spot on the
copper has contact with the brush during t = 10- 4 sec 1 • The friction
coefficient is 0.2. The Ioad P on A is ~ H times the area A, i. e.,
0.3 · 2 · 108 n (7 . 10- 5 )2 ~ 0.9 N.
In the first instance, we disregard the friction heat. Assuming brush
and collector to have the same bulk temperature, a directive for the
ratio between the maximum supertemperature e
in the brush and
the supertemperaturein the contact is given by Eq. (19.07). Although
this equation is valid for a thermal equilibrium (t = oo ), it is reason-
able to assume that the ratio {}j@ is about the same for a finite t. In-
1 Cf. STEBBENS [1].

16 b Electric Contacts, 4th Ed.
serting the data given above into (19.07}, we obtain
!!_- 4 ·30 (30 + 380) = 0 253
e- (2 .3o + 380) 2 ·
All heat generated between the isotherms, e

and {}, fl.ows into the
copper. With {} ~ 0.250, this means almost 1/2 U I. But, anticipating
that the friction heat increases the temperature in the contact, we as-
sume that about 1/4 ofthe R/2 heat, developed "\\-ithin the brush, flows
into the collector.
We now meet the following complication. The electrically generated
heat does not fl.ow through the electric a-spot alone but also through
the surrounding of this spot and thus experiences a smaller resistance
than from the a-spot alone. The difference is mitigated by the fact that
the heat current density issmall in the surrounding.
The heat generated by friction fl.ows fairly uniformly from the total
load bearing area A. Considering the electrically and frictionally gene-
rated heat together, we calculate with a common thermal constriction
area (assumed to be a circle) of radius a2 = 5 . w-s m that is !arger
than a0 but smaller than a1 •
The further calculation runs as follows. The R 12 heat from the car-
bon, which fl.ows per second into the copper, is
1 1
4 U I = 4 0.8 · 4 = 0.8 watt
and the friction heat is
f Pv = 0.2 · 0.9 · 30 = 5.4 watt
The friction heat dominates. Applying § 21, case S, it is found that
6.1 5 5 108 watt
q = n (6 · 10-5) 2 = · · m2
and according to Eq. (21.08)
e= 5 .10-~~~.5 ·1os o.75 = 550
Furthermore according to (21.02)
380 10-4 45
z = 3.4. 106 (5. 10-5) 2 • = .
This is a large z. From Diagram (21.06) one reads that 86% of the equi-
e
Iibrium is reached. Hence
{} = e. o.86 = 47°
This is a harmless supertemperature.
§ 44. Friction and wear with a carbon-brush collector contact 251
The supertemperature of the warmest isotherm in the brush can
be calculated by means of Eq. (13.06) and a figure similar to (19.01).
It is found tobe about 70° which is morethan {}. This result confirms the
statement that the thermal gradient at the contact is directed into the
collector.
Exarnple 11: Contact during a valtage flash. A flash is alwayscaused
by a conducting area which is decreased below adaptation size because
of reduction of the load. Assumptions: Flash time of 2 . 10- 5 sec. The
conducting area decreases by loss of contact spots so that one remain-
ing a-spot has to carry the whole current (8A). In addition the area
of this one spot decreases, say to a clean circular area of radius 1/3 ac
= 1.33 . 10- 5 m. As is readily checked, this means a constriction voltage
of U = 6.8 V. This voltage is called a flash: see § 45.
The electrically produced power is now 58 watt (instead of 0.8 watt
in Example I), of which at least 14 watt flow to the collector. The fric-
tion heat is diminished to approximately 2 watt. Therefore assuming
16 1 27 10to watt
q= (2. 10-5)2 = . . m2
we have e = 2 . 10-5 . 1.27 . iolo = 67o

380
z = 3.4 . 10~~~. 10-5)2 2 . 10-5 = 5.6
{} = e .o.ss = 590°
{} is the maximum supertemperature near the center of the spot.
The average supertemperature during the flashtime is smaller, namdy
about 500° as judged from Fig. (21.11). With a bulk temperature of
70 °C, the contact temperature can reach 843 °K. This suffices to harm-
fully soften the copper if the duration of the flash exceeds a certain
minimum. This minimum is 7.5 . 10- 5 sec according to Eq. (17) of
R. HOLM [38]. This is more than the assumed flashtime of 2. 10- 5 sec.
However, flashes may occur several times in the same spot and the
softening effects partly add. Five consecutive flashes of the assumed
strength in the same spot can render the respective spot soft and prone
to increased wearl.
§ 44. Friction and wear with a carbon-brusb collector contact

Any wear rate treated in this chapter is of the microtype; cf. clas-
sification Table (41.05). In spite of the small rate, the wear can be in-
tolerable due to great distances of travel. Fairly subtle discriminations
1 See R. HoLli [38].
are important in practice. The present theory can not help much in this
situation. Therefore, the exposition will be mainly confined to summa-
rizing the empirical results obtained under fairly well defined condi-
tions. Threading (grooving) and copper picking of brushes as well as
blistering of the collector film will not be treated. Concerning these
phenomenas we refer to R. HoLM [37] p. 271, SHOBERT [7], and J. E.
THOMPSON [2] where further references are noted.
A. Mechanical wear with sliprings, no current. We suppose smooth
sliding, i. e., a mechanical system that does not cause such disturbing
factors as for instance vibrations. They could increase the wear tenfold.
We also suppose a graphite material without special grinding compo-
nents added. Then, friction and wear with a graphite brush and a cop-
per ring covered with its normal collector film are much the same as
with the same graphite brush and a ring of the same material. It does
not matter that a copper oxide film is adjacent to the copper. The
upper layers of the collector film always consist of graphite platelets
with their basal planes, placed essential horizontally along the sliding
track of the ring, either graphite or copper. A similar film also prevails
on the brush face.
When the brushes on a copper slipring have run in, the sliding sur-
faces are very smooth1 . This occurs when the three interdependent vari-
ables, friction coefficient, contact temperature and contact voltage,
have become almost constant (cf. Fig. [39.03] and text) and when the
platelet film with its characteristic coverage has been formed on both
members. The wear may then be represented by Z"" 10-3 • In prac-
tice, on commutators, Z "" 10- 2 is common.
B. Inßuence of the current on the wear (absence of arcs). The cur-
rent does not directly inßuence the brush wear except for an occasional
small electrolysis; but it may be the cause of a high temperature that
produces oxydation and eventually softening of the ring material. This
leads to increased wear. If brushes slide smoothly on a ring of graphite
or a non oxidizing metal (as gold) the wear remains largely independent
of the current and dependent on the load alone. However, when the
ring consists of oxidizing material such as copper, oxidation plays a
great role. It is well known that the cathodic brush weBrs about 2 to
4 times more than the anodic brush (individual tracks); cf. E. HoLM [15]
Table 1. The wear of the anodic brush shows little dependence on the
current. lt can even happen that the wear is less with current than
without 2 •
The difference in wear can be explained, according to § 42 C, by
1 See in E. HoLM [12] the replica pictures (10) and (12).
2 See R. HoLMet al. [24] chapter 5, also R. HoLM [30] § 42, and HESSLER [1],
[2], [3].
the different conduction mechanism under brushes with different

polarity, with reference to Fig. (42.03). It has been shown that under
the brush on the anodic ring, the process of fritting, breaking of metallic
filaments and oxidation go steadily on. Though the quantity of copper
that is transported is quite small, it seems to be the source of the for-
mation of copper oxide grains and of roughening of the respective slid-
ing track. Copper grains, with carbon usually attached to them, are
found in the wear debris and on the cathodic brush face. They wear the
cathodic brush by grinding.
Under the anodic brush, metallic filaments can not form because
the field works against diffusion of positive ions; see Fig. 6 in E. HOLM
[1-5]. The a-spots covered and protected by the thin oxidefilm nearest
to the copper, experience negligible further oxidation; see § 42 C. That
is why the wear of this brush is very small.
It is obvious from the description above that wear will depend on
the atmosphere. This was found early by BAKERand HEWITT [4]. They
showed that electrographite brushes (slightly abrasive and running in
the same track) had higher wear in oxygen than in air but much less
in hydrogen (not dried and with about 1% oxygen).
Metal graphite 1 brushes on a copperring behave similar in principle
to graphite brushes; the cathodic brush wears more than the anodic
brush. But the wear of both brushes is increased. A copper graphite
brush has pronounced wear when the brush is cathodic (about 10 times
more) while silver graphite brushes show a small difference with the
polarity and have a relatively small wear on a steel ring.
C. Brush wear during commutation with arcs. The wear plays a
major role in the problems of commutation. It is a common beliefthat
the detrimental wear results primarily from arcing. Butthis view has
not been based on clear cut measurements. It is weil known that in the
brush-collector contact transient voltages appear which are much
higher than the average contact voltage, but lower than the arc voltage.
We shall call these transients "voltage flashes".
R. HoLM et al. [43] show that the flashes strongly contribute to
the wear, even more than the arcs under conditions as are often met
in practice. Investigations under artificial conditions 2 do not reveal this
fact. The relative influence of arcs and flashes had to be studied statis-
tically, and could be performed with reasonable expenditure of time
only when automatic counting was available.
HoLM and collab. used a Hewlett-Packard Company electric coun-
ter. A small auxiliary brush was so attached to the brush holder as to
1 The grades discussed here are manufactured by Stackpole Carbon Company
in St. Marys, Pennsylvania.
2 See for example BINDER [1] and LANCASTER [2].
make contact with the commutator at a spot 0.5 mm behind the trail-
ing edge of the brush under test. The voltage between the normal and
the auxiliary brush were scanned by the counter. Fig. (44.01) illustrates
the voltage variation during contact with one segment. It is essentially
equal to the brush-bar voltage. Each testwas characterized hy records
of the number of flashes and arcs, and their duration. At the end of each
test, the wear of the brushes was
1Z
V measured. Several tests were per-
10 formed under normal commutation
circumstances; but it was necessary
to run other tests with a poor com-
mutation (altered strength of the
J. interpoles) in order to obtain a dis-
z rV\ I tinct influence of the flashes.
1/ \. _/ The sign of the current i (in
0 z 10sec J Fig. (45.01]) atthe final point of con-
Time tact between the brush and a com-
Fig. (44.01). Variation of the brush-segment
voltage during 3 · 10-• sec. A transient llash
mutator segment is dependent upon
appears between 1 · 10-• and 1.6. 10-• sec. the machine compensation. The
Another llash beginning at 2.6 · 10-• sec Ieads
to an arc with the duration of 4 · 10-• sec. cathode ofthe arc and the flash con-
At arc ignition, the brush has lost contact nected with the arc will be different
with the segment. As to t 1 see Fig. (45.01)
for over-commutation and under-
commutation, as is evident from Fig. (45.01). Because of the unsym-
metrical conditions in machines, some of the armature coils may be
und er- and others over-compensated; cf. SHOBERT et al. [3]. Thus,
evidently the arc and the last flash on a segment are not bound to the
particular main polarity of the brush, in other words, they do not follow
the general signofthe brush under consideration. Ares may occur under
the main brush with either their cathode on t he brush or with their
cathode on the ring.
The investigation by R. HoLMet al. [43] is regarded as introductory.
In tabulating arcs and flashes, the sign of the current was disregarded
Averages of the observations for all brushes independent of their polari-
ty, were used for the presentation in Table (44.03). The following vari-
ables are used :
P = total mechanical Ioad on a brush; I = current per brush ;
Q = charge in coulomb transported by arcs during 1 km of sliding;
Qw = volume evaporated from the brush under the influence of
the arcs. From earlier measurements, w was found to be about
2.5 ·10-6 cm 3 fcoul. This value ist used in the calculations. r (2) and r (5)
are fractions of the test time during which the voltage was over 2 and
over 5 V respectively.
Under normal commutating conditions, it is not feasible to vary
any of the variables P, I, r (2), r (5) and Q without influencing the
others. In other words, a direct separation of the variables was im-
possible and it, therefore, was necessary to express their mutual relation
by a formula. The chosen empirically based formula for the wear per
km tra vel distance is
W = P[W0 + 0 1 ('r(2) + 2r(5))I + gVQ] + wQ cm3 /km (44.02)
The term P [ ... ] represents adhesional and interlocking wear,

essentially due to the roughening of the surfaces by flashes and arcs.
The product wQ represents the brush volume evaporated under
t.he influence of the arcs. The factor 2 before r(5) is motivated by the
relatively high efficiency of high voltage flashes. The terms within the
bracket represent the following components of brush wear:
1. P W0 = wear without current. 2. P 0 1 [ T (2) + 2 r (5)] I = wear
due to surface roughening by the flashes. 3. Pg V"Q
= wear because of
surface roughening by arcing. The square root is motivated by earlier
measurements 1 ; g is a coefficient that is determined so as to provide
optimum agreement between formula (44.02) and measurements.
The current, j, at which an arc ignites, is related to the observed
arc time, ta, according to 2
ta = 0.03 Lj
where L is the inductance of the tested armature coil in the position of

commutation. Q is equal to the average arc current of j/2 times the
total arc time per km of sliding.
Typical results are shown in Table (44.03). They are calculated with
g = 3 10- 7 , 0 1 = 11 . 10- 7 for nos. 5, 6 and 7, and with the same g
and 0 1 = 8 · 10-7 for nos. 1, 2, 3 and 4.
The coefficient W0 for wear without current was obtained by oper-
ating the machines without electricalload. Considering the great spread
of the measured W0 , averages were used, namely, for a particular
Westinghouse 3 H.P. generator: W0 = 2.3. 10-7 cm3 /N km; for other
machlnes: W0 = 6 · 10-7 cm 3 /N km.
Eq. (44.02) facilitates the calculation of the average volume of wear.
More specific problems appear in practice, however; as for example the
following. In the moments of starting of locomotive motors, very high
current intensities pass certain segments of the commutator, and it
happens that they are heated to such a high temperature that their
strain hardness is lost. As a consequence, those segments will wear more
1 R. HoLM [37] Eq. (47.05) and LANCASTER [2].

2 See Eq. (45.11).
~
Cl
~
Table (44.03). W ear due to flashes and arcs during commutation with Stackpole brushes 1
Current Distance Number Are Are of igni- Wear

No. per of p T (2) T(5) of duration Q tion with j calcul. measur. R
brush travel arcs I
A km N perbar w-• sec coul A in w-• cm3 /kg
General Electric aircraft generator; L = 5. 10- 6 H; Stackpole brush 124; brushface 1.1 X 2.54 cm2; speed 15 mfsec
1 20 10350 9 0.03 0 0 0 1.0 0
2 50 1 10400 9 0.13 0.01 0.9 5.9 5.9 0.05
[;!2
Westinghouse 25 kW Exciter; L = 9. 10- 6 H; brush 124; brush face 1.9 X 4.55 cm 2; speed 35 mfsec s:
~·
3 50 I 26500_116 I o.o31 I o.oo12 I o I o I o I o 1 2.8 1 2.4 0
Westinghouse 25 kW Exciter; L = 9. 10- 6 H; brush 189 resin treated; brush face 1.9 X 4.55 cm 2
~~
4 l_:oo . l_18ooo 116 1 o.oo861 __ <).000351 o.o16 1 o.8 1 o.ooo231 3.0 1 3.4 4.0 0.003
Westinghouse 3 HP generator; L = 10- 4 H; brush 124; brush face 0.95 X 1.9 cm 2; speed 8 mfsec
5 10 9500 4.5 0.065 0.004 0.025 1 0.001 0.3 0.47 0.55 0.15
6 10 7600 4.5 0.081 0.0025 0.03 2 0.005 0.3 0.55 0.53 0.15
7 10 4330 4.5 0.3 0.055 0:7 6.5 1.03 2.2 2.51 1.885 0.18
1 Both brushes, 124 and 189, have lampblack base and are molded; 124 has density 1.51 gfcm 3 , 11 = 1.9. 108 Njm 2 and E
(YouNo's modulus) = fi3. tos N/m 2; 189 (resin treated) has density 1.6 gjcm3, H ~ 5. tos Njm2 and E ~ 80. tos Njm2.
t.han other segments and the commutator becomes unsymmetric, caus-
ing vibrations of the brushes and other detrimental effects1 •
R, listed in the last column of Table (44.03), is the ratio between
the wear due to arcs and the wear due to fiashes. It is computed
according to
PgQ + wQ (44.04)
R = P01 L•(2l + 2.(5)]1
Applying Eq. (44.04) to the observations in Table (44.03), one .finds
that always R < 1, i. e., the wear due to fiashes was always greater than
the wear due to arcs. Both effects are called sparking. Strong arcing is
always accompanied by strong fiashing resulting in high wear. Thus
visible sparking indicates high wear or, as one says, poor commu-
tation.
Table (44.05) shows properties of brush grades resulting in fairly
good performance with respect to f, U, and W0 when used on the kind
of machines noted.
D. Abrasion of the slip ring. The total volume worn from the ring
by electrographite brushes is smaller than that of the brushes sliding
on it (about 10 times smaller). Since the wear is distributed over the
whole ring track, it is very small referred to the unit area. In the
case of metal-graphite brushes, the total volume worn from the ring 2 ,
is about equal to that worn from the brushes. A consequence of these
facts is that the ring surface changes very slowly. Therefore, even a
brush that is sliding a long time (days) touches the ring only insmall
bearing areas on the humps and the indentation deformation of these
is merely elastic.
E. Friction between an electrographite brush and a ring. The dis-
cussion is con.fined to normal conditions with the friction coef.ficient, f,
within the range of 0.05 < I< 0.25. Engineers have ample expe-
riences concerning I, but the results give incomplete understanding of
the physical process.
The friction coef.ficient is proportional to the product of the load
bearing area, Ab, and the average speci.fic friction force, ip; cf. § 37.
A great 1 does not tell whether this value is caused by a change in Ab
or in ip; i. e., in the nature of the sliding surface. In (37.01) nos. 14
and 15 show variation of Ab with time at constant 'lfJ, while nos. 16
and 17 show variation3 of 'lfJ with temperature at constant Ab. No
further determinations of 'lfJ in brush contacts are available. As for
1See for instance TöFFLINGER [2].
2Cf. Table (41.09).
a The result was obtained with a contact between an electrographite brush of
H = 2. 108 Nfm 2 (face about 0.32 cm2 ) and a ring ofthe samematerial at P ~ 2.4N.
1 7 Electric Contacts. 4th Ed.
'l'able (44.05) Average values 1 referring to several similar brush grades for the application noted ~
H = indentation hardness (average of values for different directions}, E = modulus of elasticity (a great difference refers to different di- 00
rections}, S = breaking strength. The friction coefficient, f, refers to a contact temperature of 50-70 °C. W0 is the wear without current.
If the averagetotal voltage from brush to brush is 0.5 to 1.5, or 1.5 to 2, or 2 to 3, it is called low, medium or high respectively
Brush type and applicatlon Density II I E I s Resistivity f Wo u

to•~g/m" 10• N/m2 w-• m to-"m"/km V
Resin bonded artificial graphite, some abrasive;
on small commutating machines (fractional
horsepower) 1.65 4 70 0.36 70-120 0.2 6 high
Baked carbon; on distributors, old fiush mica
motors; cranes 1.6 2.5 103-140 0.28 4.5 0.2 low Ul
Natural graphite with carbonized binder; on
turbomachines 1.3 1 35-125 0.14 2.5 0.15 4 medium
s·s:
OQ
Lampblack base electrographite impregnated; 1.6 1.6 46 0.18 4.5 0.15 1 medium
on many d. c. machines 4.5 82 0.4 4.5 0.18 1.5 medium i
Lampblack base electrographite; large power ~
machines at high speed 1.4 1.5 43 0.14 6.5 0.2 0.5 high
Lampblack base electrographite, usually with
impregnation; on large industrial motors and
generators (steel mills), also on main generators
on Rail Roads 1.6 4 88 0.42 . 6.0 0.23 high
Graphite with adjuvants, carbonized hinder;
high altitude (about 27 km) 1.65 3.5 0.27 2.0 0.25 1 medium
Metal graphite (copper or bronze) ; low voltage
a. c.and d. c.machines, commutators and sliprings 3.3-5.0 3.5 100-600 0.6-1 0.1-0.5 0.23 0.7-1.0 low
1 The values werc obtained by mcasurement~ in the Labaratory of the Stackpole Carbon Co. Cf. SnoBERT [7].
I, a few average values taken from service, are represented in Table

(44.05).
The following data about brush friction are cited from R. HoLM
et al. [24]. On a contact between a non-impregnated electrographit.e
brush and a copper ring or commutator, I was measured with P
=6N, 5<I<30A, 5<v<30mfsec: 1=0.17±0.03 under the
anodic brush, I= 0.21 ± 0.04 under the cathodic brush, and I= 0.18
to 0.30 without current. A blast of tobacco smoke against the contact
could produce a transient increase to I = 0.35.
Measurements on a contact between a metal graphite brush and a
steel ring gave:
I= 0.20 with I= 0; I= 0.15 with I= 7 A under anodic brush and
f= 0.22 under cathodic brush.
With impregnated brushes smaller f-values (f = 0.15 to 0.1) are
obtained, provided the brush temperature is high enough for the im-
pregnation to act as lubricant.
F. Brush Iifting by hydrodynamic air pressure. The gap surrounding
the a-spots is filled with air, but not steadily the same. Air enters
the gap at the brush front and escapes at the rear end. However,
wear debris can produce a difference between front and rear as
debris deposites preferably at the rear end, more or less blocking the
air from escaping. In this way a considerable air pressure can be
generated between a brush and a ring. On a commutator, the effect is
small or not existent due to the slots acting as air receivers. The pres-
sure can be measured by means of a manometer connected to a tube
through the brush, ending in the brush face. In brush-ring contacts,
STINE [J] observed up to 35 Torrand HoLMet al. [24] up to 100 Torr
at a speed between 10 and 20 mfsec. The corresponding pressure force
subtracts from the load.
If the direction of rotation is reversed, "sucking" is observed. After
a run of some days, wear fragments again obstruct passage of air at
the rear end of the brush.
A certain tangential length of the brush face is necessary for tlw
effect. Lifting becomes insignificant when the brush is shorter than
4 mm or if the ring is slotted with slots spaced by 4 mm.
The lifting effect is of the same kindas REYNOLDS' hydrodynamic
wedge effect; see § VI. However, the effect is not the result of an in-
clination ofthe brush, as some authors have believed. Both the friction
and the hydrodynamic air pressure tend to incline the brush in the
direction for production of a "sucking".
17•
§ 45. Theory of commutation with speeial regard to voltage

flashes and arcs
A. Fundamental equations. In § 44, data of wear caused by voltage
flashes and arcs are given. The present problern is to investigate how
these phenomena depend on the qualities (inductance, resistance, speed)
ofthe machine. Westart from well known differential equations. They
are of a type that is familiar to engineers, and no e:l!.:tended description
will be necessary. The figures imply considerable simplifications. For
instance, it is assumed that the brush contacts only two segments at
a time. A tjpical armature coil (RL) in Fig. (45.02) is regarded tobe
in its situation of current commutation.
The simplification is justified because the aim is merely to point
to phenomena which emerge from the modern theory of contads.
They were neglected or little under-
- -1 stood in earlier literature, even in
Jfindings such a modern textbook as BöDEFELD
Riser
and SEQUENZ [1] or the interesting
C.Segmenk paper of DREYFUS [1]. The case when
rh->"777777:..-rh ' - - - - - - ' the brush contacts more than two
segments will be outlined in Section E.
In the configuration ofFig. (45.01) ,
-zl segment (1) at the time t = 0 carries
l
the total current, 2 I , to the brush.
Mter r seconds, segment (2) is in this
·S very position, and the current i has
changed from - I to + I, viz., the
current i has been commutated. The
0 lower diagram in the figure illustra-
tes how the currents i in the windings
Fig. (45.01). Schematic, showlng symbols and i 1 in the riser (1) vary with time;
and variatlon of current in a coll durlng
commutatlon the time scale conforms with the
corresponding positions of the front
of the brush in the top figure. If the variation of i and i 1 respectively
follows the solid curve that ends exactly at i = +I (or i 1 = 0) at the
time r, the commutation is called ideal1 • In reality, the brush will
leave its last contact with segment (1) at a time t1 before r when i 1 is
not zero. The current Variation may be similar either to the dashed
curve in the event of undercommutation or to the dashed-dotted curve
at overcommutation.
1 The realization of ideal commutation succeeds only under special fll.vorable
condition; see for example BINDER [2] p. 47.

§ 45. Theory of conunutation with special regard to voltage fiashes and arcs 261
At t1 when the contact opens between the brush and segment {1),
the current, i 1 (t1 ) has a finite value j. This value is positive at over-
commutation and negative at undercommutation. The Separation of
the contact memhers usually does not mean a sudden interruption of
i 1 . An arc may ignite through which current continues to flow a short
time. The current i 1 (t) for t > t 1 decreases rapidly accompanied by a
corresponding swift change of i .
Three circuits are considered. They are sketched in Fig. (45.02).
First is the commutating coil (RL), R being its resistance , L its self-
inductance, and i the current.
Second, the remainder of the
armature.Togetherwith the Ioad,
it is represented as (R1 L 1 ) with
the resistance R 1 , the self-induc-
tance L 1 , the mutual inductance
. .lf1 ver sus (RL), and a current
y1 ± I whose values will be dis-
cussed below.
Third is a circuit (R2L2 ) re-
presenting the pole coils. They
have the mutual inductance M 2
versus (RL) and the current
Y + y2 ; Y is the average steady \.
current driven by the voltage E
of the machine ; y2 is added be-
cause of inductance from (RL). sidered Fig. (45.02). Diagram indicates the clrcuits con-
in Eqs. (45.05). 1t shows symbols and di·
The mutual inductance between rection of positive currents and relative motion
(R1L 1 ) and (R2L 2 ) is neglected.
It is assumed that. the coil (RL) experiences overcommut.ation
(undercommutation leads to mathematics similar to that for over-
commutation). W is the voltage induced into (RL) when it mows
across the interpole field. Meanwhile, the voltage E of the machine
does not produce any current through (RL). The variation of i causes
induction into (R1L 1 ) increasing the current beyond +I on the left
side (from the Ioad) and decreasing the current on the right side. During
the interval between t = 0 a nd t = ~ , the process is fairly symmetric
so that one may s et the currents in (R1L 1 ), on the left and right sides
respectively
and
+ I+ ydt)) (45.03)
- I + YI (t)
and use y1 (0) = 0 as initial condition.
Quantities R, L, R1 , L1 , M 1 , R 2 , L 2 , and M2 are considered as

constant during the short commutation time. The contact resistances
between brush and segments (1) and (2) are labeled r1 and r2 respectively.
They are the sum ofthe constriction and tunnel resistances combined
with the a-spots. Thus
(45.04)
Applying KIROHROFFS rules to the circuits on Fig. (45.02) yields
the following equations valid for the interval between t = 0 and t = T,
when there is no arcing,
(45.05)
J
with the initial conditions i (0) = -1, y1 (0) = 0, y2 (0) = 0.
The emf E (of the machine) does not drive a current through the
coil just commutating as it is in a so-called neutral position. In the
circuit R 1 L 1 , equal but ·opposite currents are driven by E through the
halves of this circuit cancelling each other with respect to the fiow
direction here considered and marked in Fig. (45.02). For these reasons
E does not appear in the first two equations of (45.05).
The integration of(45.05)is carried out in R. HoLM [41]. There it is
shown that the third of the equations can reasonably be simplified to
L2 dyz = - M2 di
dl dt
B. A ßrst integral, slightly simplified, is as shown in R. HOLM (41),
(45.06)
where L, L 1 , M 1 , and M 2 are average values for the different coils.

W (voltage induced into [RL]) and U are averages, also with respect
to time. Introducing v = peripheral velocity of the commutator and
s = distance traveled during tx. Eq. (45.06) is transformed into
-W -. U-(·1 + --2
M)
L1
""' (1 - -M21
LL1
-
M2)
-2
LL2
v
2 I L ---
s
(45.07)
If Eq. (45.07) shall remain satisfied when v varies, W must be

essentially proportional to v. This actually is the case in good machines.
The minus sign before the terms with U in Eqs. ·(45.06) and (45.07)
gives the impression that because of U an increased W is needed, i. e.,
§ 45. Theory of commutation with special regard to voltage flashes and arcs 263
that U constitutes a difficulty for commutation. However, the nega-

tive sign before U is merely a consequence of the choice of positive
directions (in the figures) for the calculation. It is readily proved that
U actually assumes such a direction as to approach the commutation
towards the ideal event. It is evident that U supports W on the whole.
lt produces the so-called resistance commutation.
It is advantageous that the· high resistivity of the brush material
and the (tunnel) resistances of the collector films produce relatively
high contact resistances r1 and r 2 rendering fJ a substantial factor im-
proving commutation. Adaptation of the a-spots to the current by
means of B-frittings causes U to remain about 0.4 V = constant
(except in flashes), largely independent of the brush material, the
current and the mechanicalload; cf. § 420.
C. Arcing during unbalanced commutation. Current and lifetime of
tke arc. During the essential portion of the lifetime, ta, of the arc1 , the
arc voltage V a is practically constant, only slightly higher than the
cathode drop. When the contact separates, the current continues to
flow across the gap, and the arc voltage Va appears in Eqs- (45.05)
instead of the contact voltage U = i 1 r 1 • In spite of all simplifications
used, the deduction of the solution is still too complex to be cited here.
lt is carried out in R. HoLM [41]using the LAPLACE transformation. It
is shown there that the time constant of the arc, Ta, is always large
enough to make
for t < ta (arc life). This means i. a. that the arc current decreases
Jinearly until the minimum arc current Im is reached. At Im the arc
extinguishes and I suddenly drops to zero.
However,in the event of considerable arcing, Im is so much smaller
than j that one may calculate with an arc current decreasing linearly
with time from j to zero. In this event, the arc life is expressed by the
equations
where
(45.08)
1 The arc life is of the order of 10-s sec; see Table (44.04). Special arcs burning
about to-a sec as studied by SWINNERTON et al. [1] grow in length increasing
their voltage.
The energy dissipated in the arc is Ea = l j Va ta. With ta from

the first equation of (45.08)
(45.09)
This shows that Ea is essentially independent of Va.

To numerically evaluate Eqs. (45.08) and (45.09) it is necessary
to know Land j. Other quantities as R, R1 , L 1 , M 1 etc. appear in such
a fashion that only certain ratios between them are needed for the
calculations. The ratios vary only slightly from machine to machine,
and are relatively easy to compute. Computations carried out in
R. HoLM: [41] lead to
~= 0.5to0.6 (45.10)
2
Va is about 12 volt under the positive and 24 volt under the negative
brush. Hence the rule of thumb
0 ' 55 L.
ta"" 18 1 = 0 • 03L.1 (45.11.)
D. Energy dissipated in arcs and voltage flashes. According to Eqs.

(45.09) and (45.10) the energy that the arc consumes is
Ea"" 0.3Lj2 J (45.12)
The magnetic energy simultaneously liberated in (RL) is
Em= !Lr(J~j) 2 -i(ta) 2 J J

which is found to be "" 1.5 LI j.
Hence
(45.13)
showing thatE", is many times greater thanEa. The differenceEm- Ea

is consumed in (R1 L 1 ) and (R2 L 2 ).
The essential constancy of the arc voltage facilitates calculations
concerning the arc. The mathematics of the voltage flaskes is more in-
tricate. However, for rough computations one may treat the average
flash voltage, V1, as a constant and then apply formulas deduced for
arcs.
Now consider the following problem. Can the last flash [between
2.7and3 ·10-4 sec onFig. (44.01)] consume all available energy and thus
prevent arc ignition 1 Let t1 be the time from the beginning of the flash
at the current i 1 = it to the interuption of the contact. If during t1 the
fiash has consumed only a fraction of the available inductive energy, the
§ 45. Theory of commutation with special regard to voltage fiashes and arcs 265
current is finite ( = j) when the contact opens; accordingly, an arc ignites

with the current j. But if the ßash has consumed the whole availahle
energy, j is zero, and no arc appears. The ahility of the ßash to con-
sume the inductive energy depends on L, j 1 , t1 and the average ßash
voltage V1. We shall make a computation with reasop.ahle values. To
avoid the arc is fortunate hecause with ßash and arc energy ahout equal
the arc produces the much greater wear. Let us consider the condition
that the energy is consumed in the ßash.
A reasonahle length of the last a-spot is ahout 10-4m, as assumed
for Example 1 of § 43. With the peripheral speed of the collector heing
v mfs, t1 is 10-4 /v sec and the energy that the ßash consumes is
E1 = ! i v t = 5 ·10- v, ~
1 11 5 (45.14)
This we compare with the energy EL that during t1 is made availahle

to the ßash hy L. It can he computed with equations deduced for the
arc, hy using V1 instead of Va. Eq. (45.12) is evi:m independent of the
voltage and is direcf.ly applicahle. Hence, the rule of thumh
EL = 0.3LjJ (45.15)
Evidently t.he condition for no arc is
EL:;;:;;; E 1 or Lj1 v:;;:;;; 1.67 ·10-4 • V1
and for V1 = 5 volt

(45.16)
Measurements hy BINDER [1] permit a check of Eq. (45.16). BIN-
DER used a specially constructed collector which produced interruption
of a determined current between hrush and segment. In Tahle (45.17)
ohservations are compared with calculations using Eq. (45.16). Con-
sidering that (in this equation) V1 means a kind of average ßash voltage
and also a maximum voltage that appears without consecutive arc
ignition in BmDER's experiments, the agreement hetween observation
and Eq. (45.16) is satisfactory.
Table (45.17). OompariiJO'TI, between observation (BINDER) and Eq. (45.16)

L tl ;, v, L;v in 10-«J
10- 8 H rn/sec A V observed 1 nsing Eq. (45.16)
I
3.8 12.9 19 5 9.3 8.3
6.25 12.9 12 5 9.6 8.3
3.8 12.9 23 10 11.5 16.6
E. Each brush makes contact with more than two segments. When
a brush contacts more than one segment, more than one coil is short
circuited. Two situations appear: 1. when a coil (RL) is engaged in
commutation, a second coil (R,L.) lying in the same slot is short cir-
cuited by the brush; and 2. no coil in the same slot is short circuited
but a coil (R,L8 ) in an adjacent slot is. The discussion is confined to
one coil (R,L,). Extension of the problem to several coils of this kind
is easily conceived.
In case 1, the magnetic connection between the coils (RL) and
(R 8 L 8 ) is strong. Practically the total energy, not commutated in
(R L), is tranferred by induction to (R8 L 8 ), thus lost for the arc. This
energy is soon consumed in the resistance of (R8 L.}. DREYFUS [1] intro-
duced the designation of dependent (in German: unselbständig) commu-
tation for this process.
In case 2, the magnetic connection between these coils is weaker
than in case 1. As shown in R. HoLM [41], this has the effect that the
inductanceof(RL)appears tobe reduced to about 0.75LandEq. (45.11)
to
ta = 0.75·0.03Lj = 0.0225Lj (45.18)
Clearly, these effects require that the brush not only covers the seg-
ments concerned but also that the brush has true contact spots on each
of them simultaneously; cf. Section F.
F. The importance of the elasticity of the brush for the commutation.
Since the same interpole field shall produce the commutation on every
segment, it is necessary for good com-
mutation that the brush makes contact
for an equal time on each segment. This
means that the a-spots must be fairly
uniformly distributed over the entire
face of the segment, when brush and
~
segment are positioned as illustrated
I
in Fig. (45.01). In other words, the brush
I
I
I
I
must fit very weil to the curvature of
I
\
I
I
the commutator1 • The requirement of a
good adaption is particularly apparent in
Fig. (45.19). In the point Ba brush the case when it is required that the brush
with thc radius of curvature r,
tonches a ring with the radius '• contacts several segments simultaneously.
Recalling that a commutator never is
perfectly round, and that the wear can adapt the brush only to the
least curved portions of the ring, one understands that the unstrained
brush face always has a radins r 1 that is greater than the averageradins
1 Cf. TöFFLil'WER [2].
§ 45. Theory of com.mutation with special regard to voltage flashes and arcs 26i
r2 of the commutator. The difference will be increased by unavoidable
wabbling ofthe brush. From these considerations it is evident that the
reasonably good fitting of the brush which actually occurs must be
produced by elastic deformation. A formula for the elastic deformation
of the brush shall now be deduced.
Fig. (45.19) shows the unstrained brush just tonehing the ring in
a line at B. It is supposed that .the load P N (per meter of the brush
width) deforms it so as to produce contact spots from A1 to A 2 , i. e.,
along the distance 2 b. This corresponds to a lowering of the brush face
at A1 and A2 by <52 - <51 • Closing the gap requires1
(45.20)
where Eis YouNG's modulus of elasticity of the brush material. We

neglect the deformation of the ring, i. e., assume its modulus to be
infinitely great. Actually, it is about 25 times greater than E of the
brush.
A simple calculation yields
bl
151- -
- 2rl
and
hence
(45.21)
Combining (45.20) and (45.21) we obtain
(45.22)
Splitting a brush axially might have an infl.uence (beside the elas-

ticityeffect)on the adaptation ofthe brush to the curvature ofthe ring
by virtue of the vertical mobility of its plates 2 •
Example. Supposing P = 180 Nfm and E = 5 · 109 Nfm 2 (which are
values frequently met) we seek r1 - r2 that makes 2b = 6 -10- 3 m when
r 1 ",., r 2 = 5-10-2 m. Insertion in Eq. (45.20) gives r 1 - r2 = 2.3 · 10-s m.
The requirement of 2b = 1.2. 10- 2 m makes r1 - r 2 = 5.8 ·10-6 m.
These examples demonstrate that the requirement on the round-
ness of the commutator and on the deformability of the brush are
considera ble.
G. About so-called short-circuit current in the brush. The contact
resistances r1 and r2 in Fig. (45.02) are located within the respective
smal1 constriction volumes with dimensions of the order of 3 · 10- 4 cm
.1Cf. RoARK [J] p. 288, No. 6.

' Cf. TöFFLINGER [J].
(a few a-spot radii). In the rest of the brush the current density and
voltage gradient are small. This means that the essential voltage be-
tween the spots a,_ and l':! is located within the two constrictions, and
that the current fiowing in the intermediate regions consumes a negli-
gible energy compared with that of the contact resistances. Cf.
R. HoLM [37] p. 285.
That it may be suitable to provide the brush with a relatively high
resistance in the rear portion is another question. This means increas-
ing the constriction resistance r1 in relation to r2 • The voltage r1 i 1 is
then able to dominate the quantity U in Eq. (45.06). That this is
favorable is evident from the discussion above on r1 i 1 achieving
the right sign for resistance commutat.ion. Cf. R. HoLM [ 41].
§ 46. Current collectors for trolley cars

Current collection always operates with sliding contacts. Even a
trolley wheel slides considerably on the wire. The reason for intro-
ducing the wheel was the hope of Iess wear; but this was not gained.
Actually, the wire usually wears more when the counter contact is a
metallic wheel than when it is a carbon shoe.
Arcing increases the trolley wear in the same manner as it increases
the wear in the carbon brush-collector contact. The arc causes evap-
oration which roughens the surface of the contact members. I seems
that arcs are more frequent the thicker the wires are, obviously be-
cause the thicker wire is less flexible. Ice on the wires causes an un-
steady motion and increases arcing. Rain also increases the wear, but
Table (46.01). Wear of current coilectors per contact

Distance of travel, 1000 km
Test Contact Load Hardness Speed Current Wear
locallty material
N to• Nim• km/h A cma/10" km
Shoe
Germany Carbon 75. 8 30 150 2
Sweden Al+ 10% Cu 40. 8 80 0 27
Sweden Al+ 10% Cu 40. 8 80 70 58
Baltimore Carbon 160. 18 32 120 2.5
Wheel
Baltimore Bronze 160. 8 32 120 1
Wire
Baltimore against shoe 160. 8 32 120 1.5
Baltimore I against wheel 160. 8 32 I 120 8
§ 47. Electric noise in contacts 269
details of its action are not known. Table (46.01) shows results of
wear measurements under fine weather conditions.
Measurements of the kind presented in Table (46.01) have not been
made with sufficient variation of the variables. Therefore, the effects
of arcs and voltage fiashes can not be separated, as was achieved for
the brush-collector contact by means of Eq. (44.02).
A rough estimation of the wear of the carbon collector Ieads to the
following values for Eq. (44.02)
W0 = 25 ·10- 7 cm3 /Nkm: 0 1 = 8 ·10- 7 ; -r(2) + -r(5) = 0.2
§ 47. Electric noise in contacts

Electric noise is current variations which interfere with desired os-
cillations; for example, with the soundresponsein a microphone or with
radio waves in a radio receiver.
W e shall discuss five types of noise : A. Thermal noise; B. Shot
noise; C. Specific contact noise; D. Contact trembling; E. Noise
generated by sudden changes or interruptions of a contact, for in-
stance, during defective commutation and in switches.
A. The thermal noise, also called JoHNsoN1 noise, is a continuous
hum that can never be avoided since it is inherent in the discreteness
of the electronic charges and the random motion of the electrons (their
Brownian motion). It appears in both contacts and resistors, independ-
ent of their material, and it has nothing to do with the current.
The simplest derivation of a formula for the thermal noise is that
given by NYQUIST [1]. Even though the deduction is neither general
nor strict, more thorough calculations have stated the validity of his
formula: A noise source having the resistance R, develops within the
frequency range v2 - v1 a thermal noise power of
(47.01)
where k is BoLTZMANN's constant; hence at room temperature about
Halfofthis power can be scanned in an analyzer if the analyzer has the

same resistance R. According to Eq. (47.01) the thermal noise is
"white", i. e., independent of where the interval v2 - v1 lies on the
frequency scale. Actually, beyond about 1012 Hz quantum effects cut
1 Named after the first observer J. B. JOHNSON [1].
down the intensities of still high er frequencies 1 • At audible frequencies

the noise of types II and III are much stronger.
The thermal noise limits the ultimate threshold sensitivity of
amplifiers, since signals are indiscriminate if they are not stronger
than the noise. In contact applications treated in this book, the ther-
mal noise is insigificant compared to other types of noise.
B. Shot noise arises because any emission of electrons is a statistical
process; for example, the emission from a hot cathode and the diffusion
current through a pn-junction. Let, on an average, y electrons be
emitted per time interval LI t. According to Eq. (IV,12) CF = V"Y is the
standard deviation of the emission during LI t. Starting from this idea,
one finds that an average noise current i" is produced per frequency
interval LI v, where
i = 5.65. 1o-lo ·w LI V A (47.02)
if the emitted current is I. Eq. (47.02) was first deduced by ScHOTTKY,
in 1918.
C. The specifi.c contact noise, also called modulation noise, is signi-
ficant in contacts where at least one member is semiconducting. In the
audible range, the contact noise may be up to 105 times stronger than
the JoHNSON noise. It is considered tobe consubstantial with the cur-
rent noise in resistors. Its intensity in the frequency interval LI v is
generally expressed by
- J2
V 2 = A -Liv V2 (47.03)
V"
where A is a constant characterizing the sample, and a "'"" 1.

The theory is far from being complete, but it is generally believed 2
that the agent is the variation ofthe space charges in boundary regions
(exhaustion layers). This causes large changes, on a percentage basis,
of the work function for the minority carriers (holes in n-conductors),
and thus indirectly produces fl.uctuation of minority carrier current.
That it is the current that varies is seen from the proportionality be-
V
tween V 2 and I. The appearance of v inthe denominator ofEq.(47.03)
indicates the spectrum as corresponding to the FoURIER expansion of
relatively sudden variations, cf. § IVD.
A similar noise appears in stationary, not acoustically energized,
microphones when they are electrically highly loaded. The heated con-
ta:cts between grains produce varying expansions and resistance
changes. It is a problern how the partly paralleland partly series con-
tacts co-operate. RETTICH [1] has treated this problem.
1 See for instance v. D. ZIEL [1], and BITTEL [1].
2 See VAN VLIET et al. [1] and GRüSCHWITZ [1] who cites earlier literature.
§ 47. Electric noise in contacts 2i1
We shall deduce one of bis formulas on a very elementary basis,
namely with the following simplifying assumptions.
The microphone, see Fig. (47.04), has (m - 1) grain layers in series
in the direction of the current and every layer contains n grains. Thus,
with the contacts against the electrodes included, m-confocls-
there are m contact planes in series and in any
layer n contacts in parallel. The resistance in any
single contact is r(1 ± e), where r is constant and
eis the fractional variation of the resistance, which t
causes the noise, r and (! having the same amounts
in all contacts, but (! having random phases. Since
g « 1 we may assume the same current j through /.
CUf'f'ent-
any contact, letting the single contact voltage
Fig. (47.04). Simplificd
vary with increments jre. Thus the noise volt- picture of the arrangc-
age of a horizontal row of contacts is ment of the grains in a
microphone
(47.05)
As for the deduction of how parallel rows co-operate, we assume a
set of n parallel equal resistances q. First we imagine that only one of
these resistances varies, then we consider how random variations among
all of them co-operate. If the first considered resistance changes from
q to q (1 + a) the conductance of the total set becomes
q (1 ~ a) + ~ + ··· ,." ~ (1 - :)
In other words, when one of the resistances varies by the small frac-
tion a, the total resistance varies by the smaller fraction afn. With
random phases of a the n parallel resistances co-operate so as to make
the total resistance vary on an a verage by the fraction
~rn=~
n y-n
Substituting jerV;, for a we find that the total noise voltage of the
microphone is
. lfm
Jer V-n (47.06)
But, since the current in the single row is j = I fn, we finally obtain the
result that the total noise voltage i&
(47.07)
which corresponds to Eq. (33) of HETTICH [1].
Actually, r will not be the same in all single contacts, and should
be considered as an average value.
The fractional Variation e has appeared in the order of a few 10-4.
D. Contact trembling or agitation produces an electric noise in
current carrying contacts. This kind of noise plays a great part in
practice. The elastic counterforce against the load in the contact, to-
gether with movable masses, define mechanical eigen-frequencies which
cause variations of the contact resistance and, indirectly, variations
of the currentl.
Let L1 P be the amplitude of mechanically generated load variations.
Then, in a contact, L1 PfP (and according to Fig. [8.01]) also L1 RfR)
usually is greater, the smaller the load P. On the other band, at great
P, friction force and wear are great. To find an optimum for L1 Rf R
means to compromise.
E. Noise generated by sudden changes in, or interruptions of, the
contact. The concept of the contact area as composed of discrete spots
suggests the idea that the fl.uctuations of the conduction and of the
friction of sliding contacts is referred to the random appearance of the
spots. Since every conducting spot performs a solely transient ser-
vice, (which is true for the a-spots ofboth members) the beginnings and
endings of the activity of the discrete a-spots appear as more or less
sudden changes in registrations of the voltage of a sliding contact.
Calculations concerning features of noise of this type are cited in
R. HoLM [37] p. 240-242. The basic frequency, Iahelied 1/T in § IV,
should equal the number of a-spots passed per second, and the noise
spectrum should shift to higher frequencies with increasing speed.
EPPREOHT [1], investigating carbon-brush-copper-ring contacts, states
that the shift does not happen. Evidently, the noise observed did not
correspond to the exchange of a-spots, but was mainly caused by
effects independent of 1fT, as contact trembling and fritting, which
may have overshadowed the a-spot effect.
Radio noise is of partiewar interest. It is electrical noise that inter-
feras (usually by means of induction) with radio reception in the fre-
quency range around 106 to 107 Hz. Its generation by sparking commu-
tators and switches has recently been the subject of several investi-
gations2. Fig. (47.08) shows a typical noise spectrum of a switch malring
and breaking 1 to 4 A at 30 V. The bandwidth of the analyzer was
1000 cycles. The similarity with Fig. (IV,20) is evident. The spectrum
for 0 < nfT;;;;; 1 in Fig. (IV,20) 3 is seen between 0 and 2. 107 Hz in
1 BRANDMÜLLER and HEUMANN [1], and VoLKMANN [1].
• See Final Radio Interf. lEE ReportNo.4012; BENEDICK andMERCHANT [1];
MoTTER [1); FOBSTEB [1].
3 For Symbols t: and T see Eq. (IV,19).
§ 47. Electric noise in contacts 273
Fig. (47.08), rendering Tf-r = 2 · 107 • The maximum around v-rfT = 1.4
corresponds to that around 2.8 · 108 Hz in (47.08) and belongs to the
radio range proper. Its strength indicates sudden current changes as
they appear when arcs chop the current.
When an arc "chops", its current drops suddenly from a finite
value to zero 1 . The preceding variation has a relatively insignificant
noise effect. Therefore, the specific
radio noise practically disappears if ttr 7
the switch operates on a current V
below the minimum arc current. to·t
Vacuum arcs chop at relatively great
I and therefore produce a very ro- Jl \
disturbing radio noise.
Commutators produce noise spec- 1Q· '~
tra similar tothat ofFig.(47.08). The
exchange of the a-spots is of minor t/J!i ~ V\Z.
influence, but chopping by the com-
mutation arcs is a major distur- 7/JG
~ ,/ -
..
bance.
Possible means to reduce the 7
l
noise are: 1. To use several brushes 10 -7fl 1 70
·1 1
70 10 10 MHz 10'
in parallel; 2. To avoid arcing; 3. To Fig. (47.08). Typical noise spectrum of a
contact switching 1 to 4 A at 30 Y
nullify the causes of brush bounce.
This is more difficult, the higher the peripheral speed of the armature.
Using n equal contacts in parallel is expected to lower the noise
Ievel by the factor 1/V;;; cf. Eq. (47.06). GLOSSBRENNER et al. [1] found
a somewhat stronger effect on miniature sliding contacts.
Nowadays a speed of the order of 150 mfsec appears on slip rings
(not on commutators) in turbo-prop engines aimed at supersonic
velocities. As FoRSTER [1] and [2] reports, a new brush material has
been developed for these machines. The brushes are made of a com-
pliant material - molybdenum wool - that consumes impact energy
and therefore absorbs mechanical shock 2 • These brushes have a good
conductivity. Running on pure silver rings they show a negligible tend-
ency to weid with the ring. About the use of mercury contacts on high
speed machines see § 68.
1 See §51.
2 Thin wires werc felted together and compressed to form circular pads.
18 U Electric Contacts. 4th Ed.

Part IV
Electric Phenomena in Switching Contacts
§ 48. Introduction
We term a contact a switch when we wish to emphasize that its
beha vior on breaking or malring is of primary interest. The general
problern in Part IV is confined to cases where the members move at
right angles to the contact surface; very little attention is paid to
tangential components of the movement. The treatment is also limited
as regards the energy of the circuits. The ma.jor interest is concentrated
on relay contacts, where the switching energy is small enough to be
taken up by a condenser of manageable size.
In most switches arcing is significant. This motivates the occupation
with the theory of the arc in this book. Reference to any Handbook
would not do because important parts of the theory are too little agreed
upon to be adequately represented in handbooks concerning arcs in
switches. As in earlier editions of this book, we begin with a short
survey ofthe appearance and treatment of arcs in a few typical medium
and high-duty switches. No complete discussion about types 1 of these
switches and contactors is pursued.
The arc produces vaporization ofthe electrodes. Apart ofthe vapor
again deposits on the electrodes, but at least one member loses material.
We call this effect material transfer by arcing. It is treated in § 64.
Preventing arcing does not mean absence of any wear. The bridge trans-
fer of material caused by the current remains. It appears mostly as
cavitation in the anode and a corresponding elevation on the cathode,
and is especially inconvenient when the elevation interlocks in the cra-
ter. §§ 65, 66, 67 are devoted to these phenomena.
1 As to types see Standard Handbook for electrical engineers, McGraw-Hill
Book Co., 1957. A valuable, easily read theoretical survey is given in KESSEL·
RING (J].
§ 49. lgnition of arcs in switches 275
§ 49. lgnition of arcs in switches

A. Introduction. An arc can be directly drawn at its typical voltage
in an operring contact as weil as it can result from the transition of a
high-voltage glowlike breakdown in a gap to the low-voltage arc type.
These facts are commonplace experiences. Going into detail and nume-
rical information, we shall keep strictly to phenomena in switches. This
means that ignition of long sparks and the pertinent theory of streamers
is not treated here.
B. Electrical breakdown in the gap between metallic electrodes at
atmospheric pressure. The electrical breakdown of the gas in the gap
between electrodes generally requires electron avalanches. On an aver-
age, any avalanche produces means for at least one new avalanche
to run, until a self-sustaining discharge is established. This kind of
breakdown involves the buildup of a suitable space charge that deter-
mines the cathode fall. The typical process is as follows. At the in-
stance t = 0 a voltage V is applied across the gap (length 8). In this
gap, or at the cathode surface, electrons are occasionally liberated by
cosmic radiation or light. One or some such electrons will be present
at t = 0, or appear soon after1 and will initiate avalanches. Out of
their path the highly mobile electrons dissipate into the anode leaving
behind slow positive ions. The space charge of those ions enhances the
field strength on their cathodic side. Meanwhile new electrons have
been liberated at the cathode by means of photons2 that were generated
in the first avalanche. They produce secondary electron avalanches
which now run in a stronger field than the first ones and thereforc are
more efficient. The formation of a glow discharge is rapidly completed.
This glow transforms to an arc if the current supply suffices.
The following conditions evidently are required for the process:
1. The breakdown voltage V must be greater than the cathode fall of
the glow in the gas that fills the gap, viz. for air Vmin "" 300 V, for the
discharge comprises the cathode fall. 2. The gap, 8, must contain enough
gas to feed the avalanches.
A 8imilitude law exists for the electric breakdown of a gas-filled gap
with a uniform electric field. In a modern form, it reads: The sparking
or breakdown potential, Vs, is a function of the gas and the product
~8, where bis the density of the gas and 8 the gap width. For the case
1 Interesting observations on the lag between impulses initiated by single field-
extracted electrons are described in LLEWELLYN-JONES [2].

2 Later, when the self-substaining discharge is established, ionswill be more
efficient at the cathode for the Iiberation of electrons than the photons. The ions
work less by impact than by the so-called Auger-effect, see LOEB [2] Chapter IX,
p. 775.
276 Electric Phenomena in Switching Contacts
of room temperature, the product <5s is often replaced by p8 with 8 in

cm and p in Torr or atm.
The similitude law was found empirically by F. PASCHEN in 1889.
The theory was given by J . S. ToWNSEND about 12 years later. The
relationship between V, and ps in air is shown in Fig. (49.01) in the
form as it results from many investigations 1 . This relationship usually
is called PASCHEN's law.
Fig. (49.01) shows a minimum at V, :::::: 330 V and ps :::::: 7 . 10- 4
atm cm. With p8 decreasing below this value, V, increases again.
However, the increase was not
10000 stated at atmospheric pressure
V
sooo because of experimental difficul-
zooo I ties. The investigations were

made at reduced pressure with
t
1000 ll 1/ 8 > 10- 2 cm. Only R. HoLM et
al. [27] Fig. 2a continued to a
500
_'\ / gap down to 10- 3 cm, but still
\__ / above the gapthatcorresponded
to the minimum2.
zoo
Recently, GERMER 3 et al. [1],
100tJ5
10"' 'f(rJ 'Kr" otman v -7 [2], [3], [4], [5] succeeded in
ps- making observations with gaps
Fig.(49.01). Breakdown voltagesfor air fllled gaps down to 1000A. They discovered
between plane and parallel electrodes at 20 °0 that at atmospheric pressure the
rising branch at the left side of
the minimum was often missing. Are ignition occurred with V, smaller
than 330 V, even down to 50 V. This is inconsistent with PASCHEN's
law because the field strength in the narrow gap evidently produces
e:ffects that were not considered in the initial theory of PASCHEN's law.
Which e:ffects these are is not yet quite clear.
Earlier, R. HoLM assumed that the arc at closure is initiated when
contact is made with an asperity peak. The peak metal immediately
evaporizes, whereby the arc ignites in the heated and ionized vapor.
Fig. (58.01) illustrates such an ignition. A transient voltage drop to
zero indicates the contact make with the peak.
However, GERMER and collaborators observed ignition without any
transient zero voltage. They concluded that the ignition was performed
1 A thorough review of these investigations is given in ScHUMANN [ 1]; see also
COBINE [1] p. 163.

2 Notice the general observation that consistent results are obtained only after
some arcing.
3 See the survey article GERMER [12]. Cf. HAWORTH [1], [2], ATTALLA [1].
KISLIUK (1).
§ 49. Ignition of arcs in switches 277
without a preceding contact. Further significant Observations stat~d
that arc ignition during closure appears much more readily and at low
voltages (300 V down to 50 V), if the contacts were "activated". Acti-
vation is generally caused by carbon from organic vapors or deposits.
Their carbonization is produced by previous discharges. Polymer de-
posits of the kind described in § 25 Iead to activation.
For the explanation of arc ignition at low valtage without a preced-
ing contact, several ideas were propounded. For example, it was as-
sumed that the strong field in the gap moves carbon particles, gathering
them to conglomerates with high peaks, at which the field is enhanced
by a very high factor. lt was furthermore supposed that the field
emission at the peak produces a current that is able to heat and evapa-
rate a portion of the conglomerate and that an arc ignites in the hot
and ionized vapor.
KISI.IUK [3] considered a phenomenon called 1-effect in §IX as
essentially aiding the ignition. Fig. (49.02) elucidates the 1-effect. The
boundary barrier of the metal, and the wall of the potential well of
an adjacent ion (on its way to the cathode) define a narrow potential
hill through which electrons from the metal can tunnel. On the energy
level that is indicated by a dashed line, the electron is able to escape
over the right side hill of the ion well, which is bent down by the f>trong
field . This electron is thus emitted by aid
of the ion. An electron tunneling on a Ek~~{
level lower than the right side hill, is
captured and neutralizes the ion.
The present writer has objections
agairrst these theories as explaining the arc Melu/
ignition. It is not probable that a field
emission which certainly cools the peak,
0
would nevertheless be able to produce such
3: current in the solid asperity as to heat
· · · F
it to evaporation and 10n1zatwn. or means of a positive ion adiacent by
Fig, ( 49.02). Electron emission
to
the 1-effect, it will be shown in § IX the metat
that this effect aids the electron emission
in a regular arc, where enough ions are available for the effect. But it
is not probable that the 1-effect can be decisive during the beginning
establishment of an arc.
However, since the time the cited papers were published, we have be-
come familiar with whiskers; see §VII. It is reasonable to assume that
in GERMER's experiments the following happened. Activation powder
and metal vapor produced at a contact operring or perhaps at a make
elongate an asperity peak into a hair-like whisker directed by the field.
When hit at the next consecutive closure, it is immediately heated to
explosion with ignition of an arc, just as the asperity peak functions in

HoLM's experiment. However, there is an important difference. By virtue
of itb minute volume, the whisker has a high resistivity and can not
transiently lower the contact voltage to zero, and the oscillogram is
bound to assume the type observed by GERMER.
Crucial experimental verification of the ideas of the piling up of
carbon grains and of the piles growing to whiskers are still missing.
Among many accompanying observations by GERMER and collabo-
rators we emphasize the fact that the cathode spot moves, preferably
along scratches, although not gradually but in jumps so that spot mark-
ings partially cover each other like roofing-tiles; see particularly Fig.4
in GERMER [12]. The current density in the arcs at closure as calculated
by GERMER, is much higher than we assume for regular arcs.
The arcs at closure produce material transfer from the anode if they
are very short, and transfer from the cathode when they are long
enough, in agreement with the discussion of Fig. (64.01).
LLEWELLYN-JONES [1] confirms the observations by GERMER,
HAWORTH and others, adding that a thin tarnish increases electron
emission by powers of ten. Such a film also shortens the time lag of
arc ignition.
We close this section, emphasizing the fact that arcs which ignite
across apparent gaps of 10- 5 to 10- 4 cm have a bearing on the per-
formance of relays. With a closure speed of 1 to JO cmjsec this means
arcing during about 10- 5 sec. The material transfer caused by such
arcs is sufficient to be disturbing; cf. §§ 64 and 65.
C. Drawing of arcs. When a contact with sufficient voltage (V> V,.)
and current (I> Im) supply is opened, an arc ignites in the gap. Evi-
dently, the process is as follows. With decreasing Ioad, the contact
area diminishes and the contact resistance increases. Finally, the RI2
heat is concentrated on a very small volume of metal that is heated to
a high temperature and evaporates explosively. A kind of highly con-
ducting plasma is formed in which a short arc develops, probably in
10-s sec. Certainly, such an arc requires a finite gap length s0 , probably
produced immediately at the explosive rupture of the contact bridge.
A voltage oscillogram of a drawn arc is pictured in Fig. (49.03). It
refers to an opening silver contact. It is a characteristic feature of the
drawn arc that the ignition voltage V m is not higher than the voltage
of the burning arc immediately following ignition. The peak at ignition
seen on the figure is due to the inertia of the mirror of the observing
oscillograph. The voltage rise after ignition is by virtue of the lengthen-
ing of the plasma.
D. Are ignition followed by ßoating. It happens that the production
of vapor in the arc is so abundant that the electrodes are kept apart
§50. FJ-characteristics of the stationary arc in air 279
against the Ioad P. This phenomenon, called floating, was first observed
at contact closure when the initial contact constriction was heated to
such a violent boiling and vapor developing as to keep one member
floating on the vapor cloud. However, floating can also be produced in
an initially closed contact, namely by a sufficiently strong current
pulse on a material (as carbon) that is not able to increase the contact
area by melting during the pulse (no welding). There is no principal
difference between these two modes of arc ignition. They both have
sufficient individual character tobe treated as a special kind of ignition.
These phenomena are described in more detail in §59.
time-
Fig. (49.03). Voltagc oscillogram of a drawn arc between silver electrodcs
Also in the arcs at closure, investigated by GERMER et al., the vapor

pressure counteracts the load, delaying the closure. However, this
action is negligible in GERMER's experiments because the momentum
of the moving contact is at least 104 times greater than the counter
momentum generated by the pressure of the vapor developed during
the short life of the arc.
§ 50. VI-characteristics of the stationary arc in air; their use

for calculation of the duration of drawn short arcs
A. Introduction. The voltage-current charateristics are termed VJ-
characteristics. They are curves in which the voltage, V, of the arc is
plotted against the current, I, with the arc length as a parameter. They
constitute an endorsed means of illustrating properties of an arc1 . In
order to avoid complications by insignificant circumstances, we shall
describe the characteristics under slightly simplified conditions. We
suppose that the arc at every instant is in a stationary state : Thus,
1Characteristics of drawn short arcs have been published by IvEs [J], and
GAULRAPP [J]where earlier works are cited and criticized; by FINK [1], R. HoLM [27]
and R. HoLM [37].
its dependence on the rate of change of current and the opening speed
may be neglected1 • The infl.uence of the shape of the electrodes is also
neglected2 • These simplifications are permissible owing to the rough
approximations. The approximations are in turn justifiable due to the
rather great spread of observed arc durations and with respect to the
uncertainty of the circumstances met with in the practical work on
switches for which we wish to predict the average behavior of arcs in
switches. The observations refer to arcs shorter than 2 mm. Longer
arcs seldom appear in communication reiays delt with here.
80r----------,---------- -.-----------.----r----- ----- .
cunYel'flence(/f
Vol.t __._
flOVol.t Au.
Fig. (50.01). V I-characteristics of arcs drawn between gold electrodes
The theory of short arcs given in § IX explains why an arc of a

given length requires a greater voltage the smaller the current. In the
same chapter the concepts of shortest arc are introduced along with
. .
1 FINK[1] finds that the infiuence of the contact opening speed, v, is negligible
for v < 20 ern/sec.
2 It was shown in F. L. that arcs of the length s < 2 mm behave in the same
manner between fiat electrodes and normal electrodes (according to definition §56)
and even between somewhat pointed electrodes,
§ 50. VI -characteristics of the stationary arc in air 281
minimum current, I",, and minimum voltage, V 111 , both functions of the
cathode material, I 111 also of the contact shape; see § IX,I.
It is shown that the lines I= Im and V= Vm are asymptotes to
VI-characteristics and that the point (Im, Vm) does not belong to any
characteristic. The asymptote I= Im is common for the family of
VI-characteristics1 belanging to the metal with this particular-1"., the
different curves for one metal being' discriminated by a different arc
length, 8. The line V= V". is an asymptote merely concerning the
shortest arc. For all arcs with finite length 8 the characteristics run at
voltages higher than V".. Fig. (50.01) is presented as a typical example
for a family of VI-characteristics. To begin with, we describe how the
plotted points were determined by measurements.
B. Observations on breaking contacts in an ohrnie circuit. Current, I,
voltage, V, and length, 8, are simultaneously measured. The rnethod is
to draw the arc slowly enough to simulate stationary conditions at
each state passed. An opening speed v < 20 cmfsec was found 2 to be
sufficiently slow. The observations can be made with oscillograph
records similar to the voltage record in Fig. (49.03). Because the circuit
is ohmic, the current oscillogram is the reversed image of the voltage
record. Therefore, it suffices to trace either one of these records. Pairs
of values ( V,I) for times t1 , t2 etc. are read from the records. The corre-
sponding arc lengths are approximately
8 = vt (50.02)
provided the opening speed v may be considered as constant. The
determination of v and 8 will be somewhat inaccurate, particularly at
incipient opening.
Fig. (50.01) contains straight lines beginning on the voltage axis
and ending on the ampere axis; for example the one beginning in the
point (60 V, 0 A) and ending at (0 V, 2 A). Such so-called re8i8tance
line8 play a considerable part in the following. Let us analyze what
they mean. .
Assurne an ohrnie circuit with the emf E = 60 V and the short-
circuit current I 0 = 2 A. In this circuit an arc is drawn. The observed
{V,I)-points referring to the arc arrange thernselves on the resistance
line. This line is called so because it is characterized by the ohrnie
resistance of the circuit; namely R = E fI 0 = 60/2 . The theoretical
evidence of this fact is proved by the statement that the equation
E = RI + V (valid for the circuit) is satisfied by (E- V)fi = Efi0
= R, as the result of the geometrical position of the resistance line.
1 The famous equation by Mrs. AYRTON is formed considering I,. as being zero,
The equation is not applicable for such short arcs as treated here,
2 See footnote 1 on p. 280,
Two points on the line are particularly significant. The open circle
at 8 ~0 and I~ 1.45 A marks the arc ignition. The filled circle at
8 ~ 0.37 and I~ 0.65 indicates the situation when the arc is ruptured.
In this point the resistance line touches a characteristic corresponding
to 8 ~ 0.37 mm in the example. This 8 pertains to the greatest arc length
attainable by the line. Thus, it is the greatest arc length that can appear.
The idea of the resistance line was coined by W. KAUFMANN, in 1900.
The lowest arc voltages, on any resistance line, pertain to ignition
voltages ofthe arc at a gap whose width is too small to be determined.
The curve that connects such points is regarded as the characteristic
of the 8horte8t arc.
As remarked above, the voltage peak with which record (49.03)
begins, is not real. It is caused by an oscillation of the moving system
of the DunDEL oscillograph. A better value of the ignition voltage is
obtained from the intersection of the nearly verticalline of the incipient
jump and the smoothed curve averaging the consecutive arc voltages.
Fig. (50.01) is due to R. HoLM et al. [27]. The circuit was ohmic 1
with an emf < 250 V. The electrodes were not particularly cleaned
but can be regarded as representing service conditions (metallic clean
contacts were employed by FINK2 ). V", is read from the horizontalline
and is fairly well defined, but the position of the asymptote I = Im is
indistinctive. It is sensitive to many circumstances, some of which are
uncontrollable.
The minimum current Im is found to be dependent on the relative
humidity, h, of the air and properties of the electrodes. Usually, it
seems to have a minimum at h ~ 0.4 and assumes higher values (often
up to 50% higher) at very low or at great humidity3 ; cf. Table (X,3).
Evidently, it must have been the cleanness of the contacts that was
responsible for Im-values in FINK's measurements having been higher
than those judged as normal in.Table (50.03). A further influence on Im
has been observed by BuRSTYN [1_] who succeeded in increasing the
minimum current from 0.4 to 0.95 A on silver contacts by switching at
slowly increased currents. Determinations of I". and Vm made by various
observers with practically clean electrodes are summarized in Table
(50.03); cf. § IX,I.
The I 111 -values are to be regarded as averages during the noted
circumstances. Actually, variations in the cathodic arc basis steadily
change the conditions for Im resulting in a random character of the
duration of the arc. This will be discussed in §51.
1A resistance mat, SaHNIEWIND asbestos grid, was used.
2FINK [1] drew each arc from freshly clean spots of the electrodes and thereby
secured a small spread in the records under equal circumstances.
3 ,Accordin$ to observations br R. HoLM and collab., h = 0,4 means 40%,
§50. VI-characteristics of the stationary arc in air 283
Table (50.03). Determinations of Im and V min normal atmosphere, by various observer8;
electrode diameter ::P diameter of cathode spot; cf. Table (X,3)
Im Vm
A y
)Jaterial
I lYES I FIXK
I HOL)! lVES I GAULRAPP I FIXK
I HOL)!
c 0.02 I 0.01 i5.5 20

Al I 18.3 14
Fe 0.73 0.35 to 0.55 8.0 13 to 15
Ni 0.2 0.5 8.0 14
Cu 1.15 0.43 12.5 8.5 13
Zn 0.36 (0.1) 10.9 9.0 10.5
Ag 0.9 0.4 12.3 8 12
Cd (0.1) 9.8 11
Sb 9.9 10.5
Ta 0.59 8
w 1.75 1.27 1.0 to 1.1 15.2 10 15
Pt 0.67 1.0 0.7 to 1.1 15 15.3 13.5 17.5
Au 0.38 0.42 0.38 11.5 12.6 9.5 15
Pb 0.52 9.1 7.5 I
REMARK concerning arcs between electrodes of different metal. V". is
essentially determined by the cathode metal: I", approaches to some
degree the I",- value for the anode metal.
The line I= I", plays the part of an asymptote to the V I-charac-
teristics only as long as the emf lies below 200 to 250 V. With higher
emf, one observes transient arcs below I",. In a circuit with an induct-
ance L, the inductive voltage, Ldijdt, may play the part of an in-
crement ofthe emf E. Working with E = 200 V and L = 0.87 H, HoLM
and GüLDENPFENNIG observed acring on breaking Mo- and Pt-con-
tacts as follows :
I Electrodes I Mo Pt
Arcing at a short circuit current 0.3 0.5A

whereas according to § X the normal Im
is approximately 0.75 0.9A
C. Simplified VI-characteristics fordifferent metals with knownlm and

for small currents. Any diagram of the kind represented in Fig. (50.01)
is valid solely for a certain metal and a certain Im. If we wish to use
the method of the resistance line for different metals and different I",,
a large album of diagrams would be necessary. In order to avoid this
burden, we present a simple method that enables valuable compu-
tations of the arc duration in normal atmosphere to be made by aid
of one single diagram, namely the one presented as Diagram XI in
§XI, referring to Fig. (50.04).
The method is based on a similarity between the V I-characteristics

for different met.als. The asymptotes I = Im and V= V"' are charac-
teristic only for the particular metal, but the co-ordinate differences
(I - I ml and (V - V",) for any point (I, V) are fairly independent of
70
- 50Vo1.t.+~
I
Im.+2Am:p -- ~,tOVo1.t.+~
--JOVo1t+~
20Vo1.t.+~
20
0 J Amp
Fig . (50.04) . Illustration of the m e of Diagram XI
the contact metal. They are defined by the plasma that, at small cur-
rents, burns in an atmosphere which is essentially characterized by air1 .
Diagram XI contains curves with such co-ordinat.es (I - I",) and
(V- V",). It has to be placed on a co-ordinate diagram so that its ordi-
nate axis coincides with the asymptote 1 =1111 and its abscissa axis with
V = Vm for the respective metal. Then the family ofits curves represents
approximately a diagram of the kind in Fig. (50.01), but this time for
the metal chosen.
1 The plasma is not. strictly independent of the meta!. In R. HOLM [37] p. 313
rules are given for certain relevant corrections. Because of the poor accuracy of the
method these rules are not repeated here.
§ 50. VI ·characteristics of the stationary arc in air 285
On Fig. (50.04), a basic system of co-ordinates is seen. We begin by
marking its scales on a mm-squared paper where an ordinate of25 mm
correspond to 10 V, and an abscissa of 50 mm to 1 A. Then we apply
the cut out Diagram XI for representing the characteristics of silver
at normal atmosphere (relative humidity "" 0.5). The asymptotes for
this case are Vm = 12 V and Im= 0.4 A'. Diagram XI is placed upon the
mm-squared paper so that its asymptotes coincide with V = 12 V and
I= 0.4 A, as it is seenon Fig. (50.04). It is now ready for application
of the resistance line. Fig. (50.04) illustrates the employment of the
resistance line, represented by a ruler, that is shown in a position
corresponding to E = 60 V and I 0 = 2 A, the current with the contact
closed. We read an ignition current of 1.6 A from the point where the
ruler intersects the lowest characteristic. We also read an arc length
s1 = 0.44 mm at rupture for the very characteristic that the ruler would
touch (not drawn). Supposing a constant v = 10 cmfsec, the duration
of the arc would be t = 0.0044 sec.
Table (53.08) in R. HoLM [37] shows that measurements of arc
durations at different currents and opening speed can be weil consistent
ifthey are madeundersuch circumstances that Imremains constant. It
is quite otherwise if Im can vary. In the text following the table, a
drastic example is given. It demonstrates how a possible variation of
Im can influence the arc duration. Another example is obtained from
the curve for 0 2 of Fig. 5 in FARRAL and CoBINE [2]. The change in
the slope of the curve could be referred to a change in Im from about
0.4A to about 0.7 Afor the short circuit current I 0 > 1 A. An Im= 0.7
lies between the values reported in Table (50.03).
D. Use of the resistance line together with arc characteristics for the
problern of how a constant current is shared between an arc and an
ohrnie resistance both in parallel. In the foregoing we
have assumed the absence of any inductance in the
contact circuit. Now we treat the other extreme
case where a !arge inductance, L, is inserted (see
Fig. [50.05]) in the circuit, so that the total current,
I through L, is practically constant throughout the
life time of the arc to be considered. As indicated
by the figure the arc is shunted by the ohrnie resist-
ance R. We shall find in § 63 that such cases are Fig. (50:05). Are
of practical interest. We ask in which portions I is ghunted by the re-
sistance R. Constant
shared by the ohrnie resistance and the arc, and current
when the arc ruptures.
Suppose for example that the short circuit current is I 0 = 2 A,
R = 30 Q, and that the arc has been drawn between silver electrodes
at t = 0. Fig. (50.04) can be applied. The ruler isplaced so as to inter-
sect the abscissa axis at I 0 = 2, having a slope corresponding to R

= 30 Q, i. e., cutting the ordinate axis at E = 60 V. An intersection
point between the resistance line (the ruler) and a certain characteristic
(say that for 8 = 0.3 mm) now has the following meaning: The point
pertains to I = 1.06 A and V = 28.2 V giving current and voltage of
the arc when it has the lenght of 0.3 mm. The current in R is I 0 - I
= 2 - 1.06 = 0.94 A. This is in accordance with the slope of the
resistance line since 28.2;0.94 = 30. Again, the arc extinguishes at a
length 8 1 providing the characteristic for 8 1 touches the resistance line.
HAMILTON and SILLARS [1] in their Fig. 3 have employed another
method to find the current at which an arc extinguishes after burning
under conditions illustrated by Fig. (50.05). The method described
above is simpler.
§ 51. Vacuum arc; particularly its extinction

A. Introduction. We speak of a vacuum arc when vacuum existed
around the electrodes at arc ignition and no gas was admitted during
the lifetime of the arc. Evidently, the arc burns in the metal vapor
issued from the electrodes. The rate of evaporation is essentially the
same as with arcs burning in air. In Fig. (56.09) the material trans-
ferred per CouLOMB is plotted against current intensity for both arcs
in air and arc!J in vacuum. The plots arrange themselves along the same
curve.
The absence of a gas which is different from the metal vapor, has
the following consequences. When the arc current is interrupted the
metal vapor rapidly condenses on the electrodes and the gap quickly
assumes a high dielectric strength. Another feature in the vacuum arc
is a low pressure in the plasma region, at least in case of a moderate
current, and the voltage gradient in the plasma is so small that all
V I -characteristics coincide with the single straight line V = V m.
Any resistance line then indicates only the point where the arc ignites
and nothing about the interruption. The arc duration is deter-
mined by a phenomenon that usually is called the instability of the
vacuum arc.
B. The duration of the vacuum arc has been investigated in many
publications by members of the scientific staff of General Electric
Company. We refer particularly to CoBINE and FARRALL [2], [3],
FARRALL et al. [1], [2], H. C. M:rLLER and FARRALL (1), and KINGDON
[ 1] where other pertinent literature is cited.
The experiments show that the duration of the vacuum arc has
a random character with a very great spread. This points to an inter-
§51. Vacuum arc; particularly its extinction 287
vention of individual random happenings which KINGDON called
bursts hecause of the short duration of the single event.
The process of the hursts is not yet clarified. The AuGER effect as-
sumed hy KINGDON is too weak (see §IX, A) while the 1-effect at points
of enhanced field strength (illustrated hy Fig. [49.02] and discussed in
§IX, C) can he much more efficient. It therefore seems reasonahle to
assume the 1-effect as heing responsihle ~or the hursts.
We supposethat particularly the current in the vacuum arc is depend-
ent on hursts, as suggested hy KINGDON. Thus, we expect that the arc
extinguishes in a moment when there are too few hursts, or none at
all. A random appearance of hursts would explain the random charac-
ter of arc extinction. A complete theory is missing.
We confine the discussion to pointing out some interesting relations
which, in the form given, may he included in a future theory. Therehy,
the hehavior of vacuum arcs is compared with arcs in air. We start
from the idea that there are two different reasons for arc extinction.
First, during increasing gap the applied valtage may hecome insufficient
for the arc characteristic. Then the arc extinguishes at once. This is
the phenomenon which we predict hy means of Diagram XI for arcs jn
air in accordance with KAUFMANN's stahility rule. Second, there is the
dependence on the hursts. A situation with insufficient hursts can happen
hefore an arc in air would die hecause of lacking voltage. However, it
seems that such happenings have not much interferred in the experi-
ments leading to Fig. (50.01). But, when the gap stops to increase he-
fore its criticallength is reached, the arc would continue to hurn if it
were not for a momentary hurst deficiency. The time of such a happen-
ing is unpredictahle.
The vacuum arc which, as we have seen, is not subjected to KAUF-
MANN's rnle, often extinguishes owing to hurst deficiency. This arc
is moresensitive to hurstsl than an arc in a gas and, therefore, has a
relatively short average life.
Table (51.01). Ou-electrodes
Are drawn at an opening speed of 5.6 cmfsec, either with steadily increasing gap or
to a maximum gap of 0.56 cm. Ohroie circuit with emf = 125 V and at the short
circuit current / 0 A; t. = arc duration in sec
Vacuum arc• durlng 'steadlly Are in alrl wlth constant Are in alr durlng steadUy
increasing gap gap (0.56 cm) lncreaslng gap
2 to 3.5 1.2 1.2

0.12 to 0.20 0.023 0.0124
1 COBINE and FABALL [3] Fig.12.
2 COBINE and FARRAL [3] Fig. 4 and 12.
3 F ARRALL and CoBINE [2] Fig. 5.
4 R. HoLMetal [27] Bild 3b, emf. 60 V.
Numerical data for the comparison are obtained from three refer-
ences and noted in Table (51.01).
We shall perform some instructive calculations concerning the
following concept of bursts. An ion in a suitable position at the cathode
surface provides passes for electrons from the metal. It may happen
that it accommodates several electrons, but it may also be that no one
uses the opportunity to escape. On an average the number of electrons
that passes through an ion will be smaller than the ratio (electron
current)/(ion current). It can not be much less, since (as shown in §IX)
the ions passing through dominate the emission of primary electrons1 •
Oalculalion. The following assumptions are made for both, vacuum
arcs, and arc$ in air with constant gap:
L All bursts have the same duration, A sec, equal to the duration
of a suitable ion position.
2. The observation time is divided into intervals A; we calculate
as if each hurst starts at the beginning of an interval.
3. The average number of bursts per time interval A is y = fj I for
the vacuum arc and y = I · fJ I for an arc in air; I is a numerical factor,
and I is the current.
4. Tlie probability p(x) of x bursts on one interval A is given by
PorssoN's formula
11'
p(x) = zlexp(y)
Hence
p(O) = exp(- y)
5. The arc extinguishes when x = 0 and consequently has the aver"
age duration
ta = Aexp(y) (51.02)
The condition that fJ shall satisfy the observations for the vacuum
arc noted in Table (51.01), yields
fJ = 2.27 and L1 = 5 · 10-s sec (51.03)
Note that the rather rough calculation is based on measurements
that are afflicted by a great spread. KrNGDON. [1] p. 1356 uses a more
sophisticated calculation and computes the hurst duration on indium
to be 5 · 10- 7 sec. The difference between the two different compu-
tations lies within the Iimits of uncertainty.
Using the values in Eqs. (51.02) and (51.03) with the data of
Reference 2 in Table (51.01), one finds that I varies from 2 to 4. This.
1 Kingdon refers the bursts to the AuGER e:ffect. This e:ffect, however, as we
know from the theory ofthe glow (cf. § IXA) would pass less than 0.1 electron per
ion; i. e., much too little for the hurst e:ffect.
§51. Vacuum arc; particularly its extinction 289
is always great enough to indicate that the air ions produce at least
as many bursts as the metal ions, in spite of the air pressure being many
times (about 15 times) smaller than the pressure of the metal vapor.
This makes it comprehensible that gas dissolved in the metal has
a great infl.uence on the stability of the arc. The contact metal of a
good vacuum switch must be degassed.
C. Current chopping by vacuum arcs. Consider an opening vacuum
switch in which an arc ignites. The current decreases during the lifc
of the arc and induces voltages in circuits that are magnetically con-
nected with the circuit of the arc. Much higher voltages are induced
when the arc finally extinguishes and the current suddenly drops to
zero. In this event the arc is said to chop the current. The high voltage
can be harmful in many respects, i. a., as noise impulse or in damaging
insulation. The chopping induces less distarbing energy, the smaller
the chopping current. It is of great practical interest to find means to
make this current small. We gain clearification of the description by
employing concepts and formulas of Section B.
Consider the case that the switch arc is ignited at t = 0 with thc
current I = I 0 and the arc current decreases as I (t) = 10 exp (- tf-r)
where -r is the time constant of the circuit.
At the average duration time, t", of tbe arc, the current is
I (t.) = I 0 exp (- ta/T) (51.04)
with [according to (51.02)]
ta =LI exp(fJ I) (51.05)
Evidently I(tu), the chopped current, is smaller thegreater ta: i. e.,

the greater LI and fJ· The quantities ta and fJ depend on the metal.
KESAEV [1] has given evidence to the idea that a positive ion which
has .created a hurst soon diffu8es into the metaland so looses its emis-
sive power. In our terminology: the quicker the ions diffuse the shorter
is the duration LI of the bursts. KESAEV and KINGDON discuss qual-
ities of metals which would prolong LI. For instance, the diffusion is
about 6 times slower in Pb and Zn than in Cu and Al; thus LI is greater
and I (t J as defined by (51.04) is smaller with Pb and Zn electrodes
than with those of Cu orAl. However, Pb and Zn are unsuitable for
contacts in other respects. According to FARBALL et al. [1] Bi-Cu
seems to be a material with a fairly low chopping current; this material
is also wellsuited for vacuum contacts.
The favorable quality of Bi-Cu was verified with A. 0. and D. 0. The
problern to what extent LI or fJ is responsible for the usefulness of Bi-Cu
is not solved.
§ 52. Interrupting an altemating current

A. Introduetion. When an a. c. arc is maintained through a long
series of cycles, this means re-ignition in reverse direction after every
passage of the current through zero. Switching out does not mean
killing the arc that carries a normal current, but preventing re-ignition
after a normal current zero. It would be incorrect to call the arc anni-
hilated in the moment of zero current since its plasma may still be
hot and highly ionized, i. e., prone to carry current and quickly to
recover to a true arc as soon as sufficient voltage reappears. Certainly,
the thin regions of cathode and anode fall are de-ionized instantaneously.
The positive ions which create the cathode fall diffuse into the cathode
in about 10- 8 sec, and electron emission becomes negligible when the
cathode temperature sinks below about 2500 °C. But, the plasma cools
and de-ionizes slower. If an electron density of more than 1015 to 1016
electrons per m 3 is still in the plasma1 when a suitable high voltage
reappears, the plasmawill carry a current that may heat the plasma so
efficiently as to deliver ions and electrons for again building up the cath-
ode fall and for carrying current to the electrodes. Re-ignition that starts
from remaining ionization in the plasma is called thermal re-ignition2 •
Another mode of re-ignition is the spark breakdown of the gap by
means of electron avalanches. lt is of the type described by PASCHEN's
similitude law in §49 and Fig. (49.01). For
this breakdown a special remaining ioni-
zation is not required .
Time It is evident that in order to avoid
thermal re-ignition when switching off an
a. c. current, the arc plasma must be cooled
and considerably de-ionized during the
short period of nearly zero current, be-
fore the electrodes again assume a high
Fig. (62.01). Illustration of the race
between lncreaslng dielectrlc strength voltage. Fortunately, the reappearance
of the gap and returnlng voltage of high voltage takes time because the
d urlng the resplte Interval
electrodes and their Ieads possess capaci-
tance and inductance. The respite time is usually of the order of
10- 4 sec; but in some cases it may be only 10-5 sec or even less.
The race (Wettlauf) between the increasing dielectric strength of
the gap and the returning voltage is illustrated in Fig. (52.01). The
1 The recent publication LEE et al. [4] followed by comments contains experi-
ments and discussions concerning this condition.

1 The initial theory was given by SLEPIAN [3] in 1928, completed in SLEPIAN [5].
§ 52. Interrupting an a. c. current 291
figure 1 pictures the happening in a short interval around the time t0

for the normal passing of the current through zero. The switch has
opened during the preceding half-cycle (or a few half-cycles earlier)
whereby an arc was drawn. This arc burns at the voltage V ... It ex-
tinguishes at ft, shortly before t0 • The switch capacitance takes over
the current while the gap current becomes zero. The voltage across
the gap (solid line) rises to vl, then pendulates through a maximum
V2 (with sign opposite to V1) oscillating around the returning voltage V,
of the system. Meanwhile the gap current remained practically nil
(thick solid line). The increasing dielectric strength of the gap is re-
presented by the dashed curves (1) and (2). If (1) is valid the inter-
section with the V-curve means re-ignition at ~. But if the dielectric
strength recovers as rapidly as shown by curve (2), the arc remains dead
and the switch has succeeded in breaking the a. c. current.
B. Vacuum switcb. We first discuss the process in the vacuum switch.
The process ofrecovery ofthe dielectric strengthin the gap is extremely
rapid and efficierit and not accompanied by fire hazard. Even the
erosion of the contacts is relatively small because it is proportional to
the duration of the arc 2 which is one half-cycle at the most.
The swift recovery of the insulation is the consequence of the rapid
evacuation of the gap by means of condensation of the metal vapor on
the electrodes and enclosing shields when the current has become zero.
Condensation is performed in microseconds, particularly swiftifthe gap
diameter, D, is great and its width, s, is small. RICH and FARRALL L2]
give the following data for Ag electrodes and 250 amp A. C.
D = 1.27 cm D = 5cm
cm recovery time in microseconds
0.076 7 1
0.23 11 4
0.46 25 12
When current chopping is no problern and currents of only up to

1000 A are to be interrupted, pure W and Mo are good contact rnate-
rials. According to Ross [1], they even act as a getter, self-correcting
.the vacuum. However, current and voltage limits were moderat~ until
the staff of the General Electric Cornpany recently succeeded in raising
thern (in tests) to 23000 A, 34500 V and recovering rates up to
6000 Vfflsec using Bi-Cu contacts; cf. LEE et al. [3].
C. High power current breakers. First, we consider gas-blast breaktrs.
As motivated in § 48, the problern of the present section lies beyond
the proper scope of this book. We therefore confine the presentation
1 Cf. TER HoRSTet al. [1]. s Cf. Fig. (52.01).
111'"
to short remarks and refer the reader to pertinent literature. Concern-

ing constructions and 'interrupting ratings, see articles called "switch-
ing of extra-high-voltage cirucits" by several authorsl, also recent
articles about switchgear 2 • For survey of the theory see articles 3 by
T. E. BROWNE jr., H. EDELS, D. TH. J. TER HoRsT; further: BROWNE
[1], RIEDER et al. [2] and CASSIE [1]. In these papers the pioneer
investigations by KESSELRING (and collaborators), A. M. CASSIE and
0. MAYR are cited. See also KESSELRING [1].
The most powerful breakers work with air- and SF6-blast delivered
from a compressor with 10 to 20 atm (Preßgasschalter). The blast
velocity approaches sound velocity. The plasma is long and its voltage
is great compared with the cathode and anode falls. As long as the arc
burns, its plasma. responds to the strong cooling action of the blast, di-
minishing its surface; viz., plasma becoming narrower. The temperature
rises, but increases the conductivity only little. Instead, the potential
gradient increases.
This has a bearing on the remanent plasma that exists during the
respite time. Because its initial volume is small, it is rapidly cooled. Its
decay time can be brought down to the order of 10- 5 sec. This is nec-
essary for avoiding re-ignition. Without efficient cooling by means of
the blast, the decay time would be up to 0.1 sec. In some switches, it
may happen that the currentless plasma is cut into parts by the blast
(KESSELRING [1) p. 123).
A quantitative theory of the cooling is important for switchgear
designers. The processes to be considered involve on the one hand heat
conduction, radiation and convection; on the other hand, ionization
and heating by·the ourrent. Any of these phenomena is very compli-
cated. l!'or example, the heat oonduotion is not simply of the common
kinetio type. It involves the effect that partielas consume energy by
moleoular dissooiation, ionization and exoitation in central parts of
the plasma, and after diffusing to peripheral regions liberate energy
by recombination and exoitation decay. Details of the oonveotion pro-
cess are inoompletely known. It is expected that modern digital com-
puters will provide means to carry out numeriaal results4 of reasonable
assumptions; but until now, engineers have feit the need for a highly
simplified mathematioal model for estimation of switch performance.
Basic models, in part theoretical, in part empirical, have been con-
trived by CASSIE and MAYR. They have been completed by several
1 Published in IEEE Trans. pwr apparatus and syst. 83 (1964) p. 1187ff.

2 IEEE Trans. pwr apparatus and syst. 84 (1965) p. 851ff.
a Published in Trans. Internat. Symp. on electr. contact phenomena, Univ.
Maine, Nov. 1961.
4 PHILLIPS [1] presents examples of such results.
other workers, part.icularly those mentioned above. We cite typical
assumptions.
1. The plasma is regarded as a cylinder with radial heat conduction.
CASSIE has taken pictures of the plasma in a nozzle at the moment of
zero current and noted its threadlike strncture. This is surprising. But
jn a current carrying plasma, such a formation is understandable be-
cause this plasma is highly viscous. In its tight arrangement of positive
ions and electrons the ions are bound together in a manner similar to
the metallic bonding1 • At 10000 °K and 1 atm, the viscosity of the air
is about 0.003 poise. At the high pressure in the nozzle, with high ion
density, the viscosity may rise beyond 0.01 poise which is the visco-
sity of water at room temperature. The REYNOLDS' number then
guarantees a lamjnar movement. But, in the cooling currentless plasma
the viscosity swiftly decreases. However, there is no time for displace-
ments to reach easily observable amounts· during the short respite
interval. The strongly luminous central part of the plasma may appear
undamaged, wherea,s peripheral parts are resolved by eddies; see below.
2. It is commonly assumed that the plasma temperature and gra-
dient X remain practically constant as long as the arc carries current.
Consequently, the cross section of the plasma is considered as varying
proportional to the current, I, as is indicated by photographs of arcs by
KIRSCHSTEIN et al. [1] Tafel 3.
3. The power loss, p, by convection per unit length of a plasma
with the axial gradient Xis just balanced by the power production X I
as long as the arc burns. Viz.,
p=XI
Let p 1 = X I 1 be its value at t1 when X starts its rise before the arc
looses its current. The speed of the continued cooling has been computed
as follows. It is assumed that the cooling follows the formula
(52.02)
where T 1 is the temperature at beginning respite time.

Let Q be the heat content per unit length of the arc, and Q prop. to
T. Then Eq. (52.02) is equivalent to
rH
dQ = - Qe = - pdt
Hence, the time constant 8 is
(52.03)
t Cf.§ 28A.
Both Q and p decrease with time and it is reasonable to assume for a

e
first (rough) approximation that is a constant. Consequently
(52.04)
So much about the assumptions 1, 2, 3.

Note that we are interested in the cooling down to about 3000°K
where the electron density, 1016/m 3 , becomes too small to support ther-
mal re-ignition.
It is evident that Eqs. (52.03) and (52.04) are roughly approximate.
e
BROWNE and collab. 1 have tried to measure under weil defined cir-
cumstances, unfortunately in free atmosphere and not the circumstances
existing in the nozzle of the switch. LINGAL, STROM and BROWNE [1]
bad earlier discovered that SF6 -gas 2 had good "arc extinguishing"
ability, and now it was found that it makes every small. The ex-
planation for this property has been given by FRIND [1] . FRmn cal-
culates p as the rate of heat conducted radially from the interior of the
plasma to its cool mantle. The greater part is carried by particles disso-
ciated in warmer regions and reoombined in cooler regions. The thermal
conduotivity varies with the distanoe, x, from the axis of the plasma.
The caloulation requires an integration for whioh FRIND uses a method
given by MAEOKER.
Fig. (52.05) illustrates the temperature distribution. There is a
plateau with a very small temperature gradient. Here the the mal con-
duotivity is enhanced where the dissociation-
T Core recombination prooess is most effeotive. The
central core is represented by a steep tem-
perature hill at the top of the plateau. This
hill has a steep gradient because recombi-
nation is prevented by the high temperature,
in spite of a high degree of ionization.
Fig. (62.05). Tamperature dlstrl- This picture refers to the burning static
butlou in a plaama wlth a pla·
teau where the thermal conduc- plasma. During the respite interval the cir-
tlvity ls high and a steep core cumstances are otherwise. Peripheral parts
where the thermal conductivlty
' islow of the plasma with relatively low temperature
and ion density, and therefore low visco-
sity, are rapidly cooled by conveotion eddies. The problern is how
the hotter parts are cooled. This depends on the hot volume. In SF6 -gas
the dissoeia.tion-recombination temperature is below 3000 °K. Thus
the vulnerable, practically non-conductive plateau is low, and the
narrow highly oonductive core is readily cooled owing to its small
1 References mentioned in FRIND [1] .
2 SF4 isastabile gas as little reactive as N2 •
volume. In air the plateau temperature is high and therefore the con-
ducting volume to be cooled in the respite interval is large. In addition,
it is relatively slowly cooled because of the small temperature gradient.
To the merits of SF6 may be added that it attaches electrons and thus
prevents them from serving thermal re-ignition. However, the electrons
are captured only below 1200 °K according to RIOKAM et al. (1].
The use of SF6-circuit breakers is rapidly increasing. For achieved
ratings see LEEDS and FRIEDRICH [1]; and BATTAGLIA [1].
Post arc currents. The previous discussion has been held · as if the
current remained completely zero during the respite interval, until the
arc re-ignites. Evidently this is a simplification. As long as the rest-
plasma contains electrons and ions, any voltage across the gap pro-
duces a current. This current is termed post-arc current, particularly
when it is strong enough to increase the luminosity ofthe plasma. These
currents noticeably infl.uence the damping of oscillations in the system
which are illustrated in Fig. (52.01). Among several investigators of
post-arc currents, . we mention BROWNE [2], NösKE [1], EDELS [1],
KOPPLIN and SOHMIDT (1] and TESZNER et al. in CIGRE 1954.
In oil brakers cooling of the plasma is effected by the surrounding
liquid. This is heated by the arc and partly evaporates forming high
speed vapor jets which impinge upon the plasma. The power limit is
below that of the blast breakers. The oil has not only the mission of
developing vapor but it also improves insulation in the stationary
state, enabling smaller and more manageable breakers to be employed.
Concerning "Expansion switches" see KESSELRING [1], EINSELE [1],
and remarks by KoPPLIN and SoHMIDT [1].
D. Medium and low duty breakers. Rela- orc electrode
tively high volta.ge a.rcs are treated with
magnetic blowoutl. Hereby the plasma is
extended and blown against and between
insulotion
cooling barriers. The famous Deion breaker2
belongs to this type. Although it js pre-
sently no Ionger manufactured3 in U. S.
it will be discussed in some detail be-
cause thereby the opportunity is given to
emphasize features that are important for Fig. (52.06). Principle of de-ion
the problern of arc extinction. Its principle structure in an Deion breaker
is illustrated in Fig. (52.06). The bar-
riers are metal plates insulated from each other. The plasma
meander has been cut into a series of short subarcs, each having
1 Cf. COBINE [1] Chapter X.
2 lnvented and described by SLEPIAN [3], [4] and [5].
a due to economic reasons.
the adjacent plates as electrodes. Each subarc has its cathode and
anode fall. Together they greatly increase the voltage that the sub-
divided arc requires. During zero current interval the short subarcs
are swiftly de-ionized so that each gap requires about 200 V for re-igni-
tion by means of electron avalanches; cf. § 52A. In order to have a
safety factor, enough barriers are usually used so that the total voltage
tobe mastered per gap corresponds to about 100 peak volt. As regards
the heavy de-ion breakers (>15000 V) see ÜOBINE [1] p. 394 and
DICKINSON [1].
The rate of cooling within the gaps (width = s) is proportional to
8- 2 (FRIND [2]). The de-ionization proceeds essentially by ambipolar
diffusion of carriers to the walls, when the temperature is < 3000 °K
and s < 4 mm. A higher temperature is able to produce ionization that
compensates for the losses (RIEDER [1)).
In case of short arcs as those between the barriers of a Deion, it is
necessary that the arc spots cool quicker than the plasma. The cooling
of W and C is too slow, particularly due to their relatively great arc
spots and thus small z in (21.02). This has been shown experimentally
by Tonn and BROWNE [1], MACKEOWN et al. [2] and others1 • See also
R. HOLM [ 37] § 51 E.
Recently, the cooling influence of the electrodes was investigated
for gaps ofthe order ofmm by WHITTAKER and EnELS [1], EnELS and
ETTINGER [1] and NösKE [1]. WmTTAKER and EDELS show that a gap
(between carbon electrodes) with the width of 2 mm and the dia-
meter of 4 mm can cool a currentless plasma from 5500° to 3000 °K
(the temperature of the electrodes) in about 10- 4 sec. NösKE finds gaps
of 1 mm width particularly efficient. The reliability of the switching
process was not investigated.
Electrodes of silver containing cadmium oxide are known to ha ve
a kind of arc extinguishing ability; see KEIL [4] p. 152 and 184.
SNOWDON [1] and [3] showed that contactors with electrodes of sin-
tered Ag + 10% CdO had an interrupting capacity of 200 V at 300 A
up to 350 V at 60 A, while electrodes of silver or coin silver without
CdO, managed only half these powers. Internally oxidized cadmium
material was about 10 '7;, less effective than the sintered metal. No con-
vincing theory of this effect with CdO has been published.
E. Contact rectifter2 and synchronous switch. Considering the extent
of the difficulties in preventing re-ignition, one may wonder whether
or not it would be possible to arrange a "respite interval" around the
moment of zero current, and let the contact open just in this interval.
1 See CoBINE [1] p. 352 etc. where various references are made.
2 References: KESSELRING [3], KoPPELMANN [1], [2], [3]; seealso R. HoLM [37]
§57 D and RoLF [1].
§ 53. Breaking direct current 297
Actually, this is done in the "contact rec#fier" that mechanically
blocks every second half-cycle of an a. c. current. By ingenious inventions
the current is quenched to almost zero during a "respite interval" at
which time the contacts move together bouncelessly in less than
6. 10- 4 sec. Several such rectifiers are in use, but it seems that semi-
conducting rectifiers wHI provide simpler means.
§ 53. Breaking direct current1

In the case of D. C., the arc has no natural weak phase as the d. c.
arc when its current passes zero; therefore the switch has to extinguish
an unweakened arc at full current. The active phenomena are cooling
and lengthening the arc until it requires a greater voltage than is
available in the circuit. Are lengthening is mostly produced by mag-
netic blowing. A magnetic field perpendicular to the arc (very often
generated by the current which also fiows through the arc) deviat.es
the arc plasma forcing it between cooling walls or into the gap be-
tween diverging horns. In the de-ion breaker 2 , the arc is out by the
walls and thus replaced by a number of relatively short arcs in series,
each dependent on its own cathode drop. These drops constitute an
essential adflition to the tension of the plasma.
The function of the switch arc can be described mathematically as
the generation of an emf Va (the arc voltage) and a current component
I 11 both of opposite direction to the emf E and the current I of the
system. If ta be the life time of the arc, the available energy, W, of the
system shall be consumed by the arc during this time with the conse-
quence that, at t = ta, the current through the switch is zero. This
requirement is expressed by Eq. (53.01).
f f
t. t.
W= Vaiadt= ~ LI2 + (EI- RI2 )dt (53.01)
0 0
t.
where _!_ LI2 is the inductive energy of the system, JEI dt is the
2 ~ 0
energy that the system produces during i., and J RP.dt is the energy
0
that is consumed in the resistance, R, of the system. Evidently, during
a sufficient interval, V. must be greater than E. The greater Va, the
smaller is ta. This means the advantage of a relatively small heat devel-
oped by the arc. However, the dielectric strength of the insulations
defines an upper Iimit of va.
1 Informative literature: BoEHNE and JA.NG [1], KESSELRING [J].
2 Cf. Fig. (52.07).
Under favorable circumstances the integral on the right side of

Eq. (53.01) can be held at the order of LI2J4. BoEHNE and JANG [1]
calculate switching time (order of 0.01 to 0.02 sec) and total heat devel-
oped by the arc under several typical circumstances.
§ 54. Electric oscillations generated by d. c. arcs

Fig. (54.01) shows a circuit containing a capacitor 0, a resistance R,
and an inductance L in series with an arc. The source of the d. c. cur-
rent through the arc is not designated. It could be the main circuit of
Fig. (61.01) while the quench circuit would corre-
spond to the circuit of Fig. (54.01). As to this
circuit, it merely carries a current I that is an
a. c. or a transient damped current; but, because
of 0, the D. 0. that maintains the arc does not
Flg. (64.01). Circuit in enter this circuit. We consider a moment when
whlch osclllatlons can be the current in the circuit of Fig. (54.01) is I, the
malntalned by the arc.
The d. c. circult contaln- capacity has the voltage Vc and the arc has the
ing the source of the voltage Va. The electric state of the circuit is
current through the arc
is not designated expressed by Eq. {54.02)
(54.02)
A short time dt later
holds. Subtraction of the first equation from the Jatter one gives
d2 I
dVc = L a,,2 dt + Rdi + dV.
We introduce Q the charge of 0
- dQ =I
dt
Hence
{54.03)
which has the form of the familiar equation of oscillations in a damped

a
electric circuit. Vafß I is written as a partial differential quotient since
Va will depend also on the length s of the arc. The second term in
§ 54. Electric oscillations generated by d. c. arcs 299
(54.03) represents the damping. Writingl
,t_ . a v.
. _ R .-BI (54.04)
the condition for aperiodic variation becomes
2L)2
1 ( r L 1
LO< or ---- < -· (54.05)
Or~ 4
and the condition for oscillations

1
Lo> 2L )2
( r or (54.06)
In §50, we have secn that stationary VI-characteristics of the arc

have negative oVafo I. The influence of this fact on oscillations in a
circuit as demonstrated in Fig. (54.01) may become clear by numerical
examples. Tobegin with we refer to static characteristics and to Diagram
XI, and afterwards discuss actual deviations from the conditions as-
sumed.
Assumptions: R = 9 !l; L = 10- 6 H, 0 = 2.5. 10-8 F, constant
length, s, of the arc and the following three cases represented by points
on arc characteristics
(1) I - I".= 1A, 8 = 0.03om
(2) I - I".= 0.64A, 8 = 0.04om
(3) I - I".= 0.4A, 8 = 0.05cm
Diagram XI gives
for oase (1) (2) (3)
av. -3 -9 -230
aT
henoe rl 6 0 -140
and acoording to aperiodio undamped swinging up to }arger
Eqs. (54.05) and (54.06) osoillation osoillation and larger amplitude
In case (3) the amplitude of the oscillation would increase by the

factor of 100 after one period - if the arc obeyed the static VI-cha-
racteristic. But it does not. In order to adapt itself to a change of volt-
age and current, the arc must change the heat content of its gas and
vapor, adjust the cathode and anode spots, and so on; and the adjust-
ment consumes time. Therefore, against variations of a very high fre-
quency the arc tends to keep its resistance constant. If the arc actually
succeeded in doing so the slope of its dynamic characteristic would be-
come positive, oVfoi = Vfi. Against a frequency of 104Hz, the slope
a Vjo I mayremain negative hutwill besmallerthan thatin Diagram XI.
The phenomenon of the oscillations swinging up to !arger and !arger

amplitude is significant with respect to arc quenching, see § 60. How-
ever, calculations on the basis of dynamic characteristics are too dif-
ficult for practical purposes; and we simply conclude from the above
discussiou that oscillations with increasing amplitude will come about
by aid of the arc, only when its dynamic characteristic has a consider-
able negative slope, which is realized only for small values of I - I". .
§ 55. Bouncing
Contact make is usually not finished at the first touch, but as a con-
sequence of bouncing the members make and break their contact several
times before they reach a permanent state of contact. The bouncing of
relay contacts brings many disturbances with it. The exactness of con-
t act make is lost, and the material transferred by arcs and bridges is
increased since each bounce means a n ew switching operation. A clos-
ing contact is particularly exposed to darnage by rebounce when the
current begins with a high inrush as is the case of a circuit containing
wolfram filament lamps. The cold filament has a much smaller resistance
than the glowing one, and the initial current may be 8 times the normal
lamp current.
However, the contact break at a bounce may not entail the same
wear a& anormal switching operation. \Ve imagine, for instance, bounc-
ing on switching a circuit accord-
Impad forct- ing to Fig. (61.01). Let the first
tlllSiic contact make be at t = 0. At t1
t:DIInkrforce
the first rebounce commences.
With t1 < L j R , the current I 1 at
Time the moment t1 has attained only
a fraction of the short-circuit
current I 0 • We assume the time
Fig. (55.01). Variation of the impact force during constant r 0 to be much smaller
the time of an impact
than ti . Consequently at the
moment t1 the condenser is completely discharged and therefore has
its total ability to quench the arc. Thus only the relatively feeble cur-
rent I 1 which is interrupted at t1 , could possibly cause arcing which,
however; would be quenched. The ratio I 1 ji 0 depends on the circuit.
The impact has two phases, penetration and restitution. Fig. (55.01)
illustrates how the elastic counterforce varies. It rises to a maximum
at maximum indentation and reaches again zero when the contact
opens at the rebounce. The momentum of an impinging body with
mass m and velocity v varies from v m to zero during penetration and
§ 55. Bouncing 301
from zero to v.m during restitution where v. is the velocity of begin-
ning rebounce. The quantities vm and v.m are given by the respective
areas between the curve of Fig. (55.01) and the abscissa axis. In case
of perfect elastic deformation these areas would be equal, thus v. = v.
Actually, the curve is unsymmetric as indicated in the figure; v.< v.
This means that some impact energy is cousumed by plastic defor-
mation or lost because restitution of some portions of the elastic de-
formation is delayed.
The total impact time, t;, usually is of the order of 10- 5 to 10- 4 sec.
Eq. (1,6) shows how to calculate t; for a ball hitting a heavy anvil, both
members perfectly elastic. With members
shaped otherwise, the coefficient in (1,6)
changes but little, whereas the factors within
the brackets remain essentially unchanged.
Note that the velocity appears as v116 and
thus usually has a; small influence.
The elastic counterforce that is able to
return the impinging body during the short
time t;, is quite great. often 3 magnitudes Ami/
greater than the weight of the body.
As far as the above discussion goes, v. is Fig. (55.02). Movable holder H
with ball m. When m hits the
finite and thus a bounce will always appear. anvil and looses contact with
In order to oppress the bounce, a special the latches of the holder, the
spring Sp excerts a Counter-
pressure must act during the restitution. Fig. force against bounce
(55.02) demonstrates the principle by means
of an example. The figure shows details of a device that was con-
structed by N. C. SH.AW [1] for the study of bouncing. The ball m
is held in a holder H between the spring Sp and latches ofthe holder H.
In a test the holder, fixed to a turnable Iever, is released and falls until
the ball hits the anvil indenting it. The holder is then stopped by a
Jatch 1 and the ball soon tries to bounce. But its movement is braked
by the "stored" force in the spring that starts to act just during the
impact time. Means are provided to record the movement. lf the stored
energy is great enough, the ball is not able to open its contact with
the anvil. This means that the contact remains closed. It becomes
stationary after some damped undulations of the contact pressure.
In SH.Aw's device for analyzing bouncing, the spring Sp can be
adjusted to no bounce or to Iet the ball bounce with more or less energy.
However, practice desires simpler means against bouncing even though
they would be less effective.
Friction is often applied to dissipate impact energy of spring contact.
The principle is shown for the butting type in Fig. (55.03). The movable
1 Not seen on the figure.
arm A carries two plate springs / 1 and / 2 • The moving contact member
k 1 is attached to / 1 . On its impact with ~ the springs, / 1 and / 2 , become
bent and slide with friction against each other, thus
consuming a portion of the impact energy1 •
At closure of a knife type contact the friction
effect is considerable, particularly if the blades be
preloaded by elastic pressure 2 • But, of course, a
good symmetry must be provided in order to make
the closure instantaneous.
Another method consists in increasing the Ioad
during rebound. It can be materialized by using a
hollow movable member containing a heavy powder
in the cavity. At contact impact, the powder conti-
nues to move a short distance and then exerts an
Fig. (55.03). Spring
impact pressure at a moment when the contact
contact closing wlth without the powder would bounce. Cf. the effect of
sliding between the
springs / 1 and / 2
Spin SHAW's device, Fig. (55.02).
KEsSELRING avoids the darnage of bounce by
using several contacts in parallel. The contacts are operated so that
when the first one rebounces the second just penetrates and so on. He
also used the method 3 , illustrated in Fig. (55.04). The upper mass M is
exposed to the downward directed force F (for instance, a magnetic
attraction). After moving the distance h under the inßuence of F, M
hits another mass m that is carried by the ßat spring f. The strength
of f is defined so that a deviation x of m from the position designed
corresponds to the application of a force - 0 x on m. During impact,
M and m move in contact with each other.
The contact willremain closed (pro ba bly loos-
ened) during the rebound when m is able to
follow M up again . This is achieved when the
relationship
mF(
O<kM i + 2M
m) (55.05)
+ :X is satisfied4 , as verfied by experiment.

Fig. (55.04). Contact wlth one Fig. (55.06) illustrates a means to avoid
member attached to a spring of
prescrlbed strength high frequency bouncing5 • ABis a ßat spring
carrying a weight B. When contact is made
a.t 0 the weight has a velocity to the right, and it continues to move in
this direction for some time, bending the spring and producing an i.Dcreas-
Cf. RECHE [1] Fig. (1).
1 2 RUSSELL and KEILIEN [1]. 3 KESSELRING (3].
& This relationship has been calcula.ted by KEssELRING in co-work with
M. BELAMIN, seeKESSELRING [2]. 5 BREHM [1]
§ 55. Bouncing 303
ing pressure in the contact which may prevent high frequency bouncing.
Mter half its own period, B commences to move back and the contact
opens. The device is applicable when the frequency of B equals the
desired frequency of closing and opening the contact.
The described methods, except that referring
to Fig. (55.02), are not infallible means. A great
deal of regulating skill is required to produce a
switch that will operate without bounce.
Sometimes a kind of bouncing is recorded in
breaking contacts, viz. a transient, renewed con-
A
tact shortly after the first break. It is caused by
vibrations in the contact holder which have been Fig. (55.06). Device
excited by the mechanism that made it move. In for avoding bounce
case of frequent switching it happens ·that vibrations in with case of swJt.chlng
a constant
initiated at a closure continue and cause bounce at frequency
the consecutive opening.
In practice, the system that defines the bounce is complex. The
moving cotitact member will be solidly fixed to a Iever. The inertia of
contact member, Iever and springs together carry the bouncing energy
instead of the simple mass of the baU in SHAW's device in Fig. (55.02).
The counter member also constitutes a complex system that can os-
cillate. The pattern of the bouncing of such complex systems has been
calculated by ERK et al. (1] and compared with observations. Fig.
(55.07) shows a typical oscillo-
gram of their observations.
The upper curve registers the mm
02
I
movement of the system, the ~01 7 r\
~ I V\
lower curve the voltage across
the gap. There are pictured
0
•
IJ VV
'"" w -
....,..".
3 impacts with durations of

J
0
about 5 · 10-4 sec and 3 boun- ] J1
ces with arcs each burniug V
about 2 · 10- 3 sec. The system 8 1Z msec 1G

finally wobbles without Joosing t-
contact. Fig. (5().07). Due to ERK and F_INKE [I]. Simul·
taneous records of movement of a complex sys-
SHAW, and ERK et al. ob- tem and contact voltage ·across (Ag, CdO) con-
served welding after arcing at tacts durlng closure
contact make. The strength of
the weid is very erratic; it certainly depends on the particular .dis-
tribution of the molten metal at that time as weil as on the kind
of metal. ERK found welds usually strong enough to withstand a
Iifting force of 50 N in certain silver contacts when the current
was 200 A. (Ag, CdO)-contacts· opened until above 600 A. SHAW
observed welding after an arc at the first impact. The weid prevented
bouncing. He has indications that the closing contact was hitting some
protruding asperity which melted (occasionally owing to an arc) and
provided the material for the weid.
MERL [6] using SHAW's device (cf. Fig. [55.02]) investigated welds
after discharge of a capacitor (C = 1.7 · 10- 3 and 10- 2 F, each charged
to 100 V) at closing Ag contact (velocity = 12 cmfsec). The strength
of the welds was very erratic. Averages were of the order of 0.2 and
1 N at initial current of 25 and 95 A respectively, little dependent on C.
§ 56. Material transfer in switching contacts. A survey

A. Definition of the major types of material transfer. We generally
define material transfer as any removal of material from a contact
member, irrespective of whether a portion of the removed material has
been redeposited either on its matrix member or on the other one, or
whether it is lost to the surroundings. We distinguish four types of
material transfer in switching contacts.
1. M echanical disintigration and displacement of material without
the aid of the current. This kind of wear is usually negligible in relays,
but it has to be prevented in the contact rectifier in which the closure
must occur at precisely defined moments. This type of wear and means
to avoid it are described in Section D below.
2. Bridge material transfer, which is caused by the formation of
bridges of metal that is melted by the current in the last phase of con-
tact opening: The reason for this transfer is that the hottest section of
the bridge, where it ruptures, is displaced from the center. The volume
transferred per operation is very small, but after numerous operations
it may produce disturbing deformations of the electrodes as are
shown in Fig. (56.01). A more harmless deformation is pictured in
Fig. (56.02). Because of the smallness of the volume displaced per oper-
ation, this type of transfer was originally called fine transfer. Now,
it is usually called bridge transfer.
3. and 4. Material transfer caused by arcs, often simply called arc
transfer: The main agent is vaporization of metal by the arc. At high
currents (thousands' of amperes), the violently produced vapor also
blows out droplets of liquid metal from the contact1 .
With respect to the direction of the transfer, we distinguish two
types of arcl:
1 The ear!ier used terms "plasmaless arc" and "plasma arc" are abandoned
as misleading since type I mostly has a plasma although a short one.

§ 56. Material transfer in switching contacts. A survey 305
I. The anode dominated arc with dominant evaporization from the

anode.
II. The cathode dominated arc with dominant evaporization from
the cathode.
The difference between I and II results from the current fiow (and
plasma) diverging from the cathode toward the anode. The anode spot
,oIotinum- iritlium
onotle colllode
platinum
0,5mm
onode cotl!ode
Fig. (56 .01). The effect of bridge transfer enhanced by the action of anod~ dominated arcs, after
millions of operations. A pip on the cathode, a crater on the anode
always receives more heat than the cathode (cf. Eqs. (IX,8] and [IX,11])
and consequently evaparates more, provided this spot is not consider-
ably larger and thus much more cooled than the cathode spot. Since
the anode spot increases with the gap, the transition from type I to
type II appears when the gap length surpasses a critical value.
306 Electric Phenomena in Switching Coiltacts
Remark 1: The concepts of anode and cathode dominated arcs do

not coincide with GERMER's "anode and cathode arc", cf. § 64E.
Remark 2. Instead of the usual cathode loss in a long arc a domi-
nating anode loss appears when a sufficient anode fall is developed;
for instance, owing to a thinly shaped anode, thus a small heat con-
duction. Accordingly, the anode spot is poorley cooled, and a strong
Flg. (56.02). Bridge and short arc material transfer in the shape of a core or dome
on the cathode, magnifted 8 fold.
evaporation is produced in it, obstructing the electron current and

therefore causing a high anode fall. Typical phenomena of this kind
are realized in the "flaming arcs" that have been studied by FINKELN-
BURG et al.l: The anode is a relatively thin carbon rod and the current
is 100 A or more. The vapor emerges at a velocity of the order of
100 mjsec, forming a highly ionized and luminating jet or flame.
The following designations will be used. W means a volume of
material evaporated from a contact member. G means a transferred
volume of material, no matter whether it derives from evaporation or
from other processes (for instance, bridge transfer). In the case of arcing,
G means a difference: G = W - D, where D is a redeposit. Where it is
needed, reference will be made to the anode or cathode respectively
by subindices a or c.
With respect to the different types of arc, we speak of anode domi-
nated transfer and cathode dominated transfer.
B. Concerning the amount of arc transfer, particularly from the
cathode. The heat produced in the cathode by the arc essentially derives
from the cathode fall region. Therefore, the power causing evaporation
is ßVJ, where V< is the cathode fall, I the current through the arc,
and ß (< 1) a factor (dependent on I). Let r .b e the sum of the latent
heats of melting and evaporation of 1 ms volume of the solid metal.
Then, an arc evaporates metal at the rate of .
1
-,:-ß V,I m 8 jsec
1 F'mxELNBURG and MuCKER (1) p. 293 and 432.
Putting 1/r ßVc = w (as in earlier editions of this book) we have wl

= volume 1 evaporated per sec from the cathode. When the arc Jives
ta sec, with the current varying, the total evaporated volume is
J
ta
W= wldt=wq (56.03)
0
where q coul is the quantity of electricity that has passed through the·
arc, and iö is a certain average value of w, which is a function of the
current (as will be seen).
The only directly observed quantity is the difference, G, between
the total evaporated volume and the redeposited volume; hence
G= W -·D (56.04}
It is an agreeable experimental result for the calculation that G
is fairly proportional to w so that we may write
G= yq (56.05)
where y depends on I, especially for silver in accordance with Fig. (56.09).

Observations have often been madeundersuch circumstances that
the current through the arc varied only a little during most of its
life. It is then permissible to register y as a function of the average
arc current.
In order to calculate W or w from Eq. (56.04), one has to know the
redeposit D. Only the paper by E. HoLM; and R. HoLM [3] deals with
this problem. Ares were drawn be-
tween electrodes of different types
as shown in Fig. (56.06). It was
found necessary to consider partly
a neutral deposit and partly a I II 111 lY
deposit of positive ions always Fig. (56.06). Types of electrodes. II is the-
migrating to the cathode. The normal type
following assumptions were applied.
Fora given q, w is independent of the electrode type. In type I , the
evaporated neutral meta! atoms mainly migrate to the Sl,Jrroundings. In
type II electrodes, the neutral deposit is equally shared by both elec-
trodes. Electrodes ofthistype are called normal electrodes. In types III
and IV the greater portion of the evaporated meta! condenses on the
cup; and it is supposed that the ratio of deposits on the anode and
cathode is the same as the ratio between the electrode surfaces that
are exposed to the arc.
1 Notice that w is a function of I; cf. Fig. (56.09).
20*
In this manner, a suf:ficient number of equations could be set up

for the determination of the variables. From the results it can be con-
cluded that (under the circumstances concerned with I ~ 2.5 A) w is
of the order of 2.5y at normal electrodes. In the absence of more pre-
dse results, we shall generally assume
Q) = 2.5y (56.07)
An upper limit for w can be computed theoretically. R. HoLM [33]
and W. R. WILSON [2] have shown that at very high currents, nearly
.all the energy entering the cathode is used for evaporation of cathode
material. The theoretical proof is based on the following facts: 1. With
given current density at the cathode, the current is proportional to
the cathode spot (radius a), thus proportional to a 2 ; 2. The heat con-
ducted from the cathode spot is proportional to its radius a (accord-
ing. to § 35). The energy entering the cathode is proportional to VJ.
Thus, when I and a are very great, the entering energy is proportional
to a2 and is much greater than the heat conducted from the spot. Thus
nearly the entire amount is used for evaporation.
V. is of the order of 8 V. Heating up and evaporating 1 ma of silver
-consumes about 2.6 · 1010 Joule. The volume wi ~ 2.5 yi evaporates
per sec. Hence with ß ~ 1
ßV. I = ß8 I = 2.5 y I. 2.6 . 1010
Dr
y = ß·123 ·10-1 2 m 3 /coul (56.08)
Since the upper Iimit of ß is 1, y = 123 · 10- 12 m 3Jcoul
is the upper
limit for y with respect to a silver cathode.
At very high currents, liquid metal is sputtered from the electrodes
into the surroundings. This process adds to the evaporation and
makes y appear increased; for silver even beyond the Iimit of 123 · 10- 12.
W. R. WILSON [2] has reported on sputternig.
Fig. (56.09) summarizes measurements by several workers. The
references with pertaining markings on the figure are: 1. Measure-
ments in the research Lab. Siemens (Forschungslab.), F. L.; 2. E. HoLM
[9], E. H.; 3. KoLLER [1], Ko; 4. W. R. WILSON [2], Wi; 5. SWINE-
HART [2] and [3] Fig. 4 and private communication, Sw; 6. CoBINE
et al. [4], Oob; 7. HIRSCHBRUNNER [1], Hr; 8. FREUDIGERet al. [l],Fr.
First note the observations with copper electrodes in air. They
arrange themselves along a straight line (dashed) with the slope 1,
if!.ciuding observations in vacuum. A poi~t concerning silver in nitro-
genalso lies on this line.
Then look at the observations with silver electrodes in air. The
points gather around a curve (solid) that runs about an order of ten
below the straight line for copper. This shows it is the oxygen that
considerably affects the silver arc because, in nitrogen and vacuum,
y is as high as with copper in air and vacuum. An additional (maxi-
SwAI
Jfi
I ~\Nz
EH I / V
Yoculm {><~ A~
1/
r;;:,< 11Ag.15%CdO
. 1 Olr 0
1 I" Cob-_ lsw ,- f--
I/' Au NA?%~ I lsinler
Kox~~ V{. I _I 1
~~~~ / H't Ag, 15'1. CdO,pofenled

moleriol -
/\,Ii
Df---- -Fl. X /
sw _V
f-- e- fL.
0 f- Silr Olf 0 A~, M1%0z
1
tOl 703 A w>
I-
Fig. (56.09). Rate of evaporation in m3 /coul of the cathode as a function of the arc current; 0 refers
to silver, --o to (Ag, CdO), x to copper, 0 to gold. Air means normal atmosphere, 100 % N 2 means
pure N2 • Drawn arcs.
100
'KJ"G 1 To2WJO
g/lilul
I
JJ_
lJt)
\
1\
t)f-----
"§
t
!\>
· 11otts
·~· \
I
I \ I
11
\ ~ V
- ~~er1~--- I ~ -----
[.2::.pneS!..,-
10
0
N
10 20
i""
1 Air
30 WJ 50
_:;...-
f1}
--;c\\0350~
II
70 80 90% 10.'0
Oxygen (remoinder f111rogen)
Fig. (66.10). Due to SWINBHART [3). Effect of oxygen on arc erosion,
gas pressure = 1 atm. Fine silver
mum) lowering of y by about half an order of magnitude appears in an

atmosphere with 40% 0 2 . The compound (Ag, 15% CdO) brings ero-
sion of silver to the same rate.
It is significant that also gold electrodes are sensitive to oxygen,
see the verticalline in the center of the figure.
The point marked Wi indicates the amount of cathode loss after

subtraction of a sputtering effect assumed to correspond to 60% of y.
The papers SwiNEHART [2], [3] contain many observations that
-certainly will be important for an explanation of the behavior of silver
in arcs. Fig. (56.10) shows that the effect of oxygen on the erosion
Flg. (56.11). Due to SWINEHART [3). Fine silver in 100% N•• Total arcing Operations 80; aver-
age arc amperes = 160; each. arc duration = 3.5 millisecond ; arc voltage = 28 at arc Iength
= 0. t 27 cm. The surfaces are very rough. AfinP black powder completely coats the cylindrlcal part
()f the anode (not visible on the ftgure) while the corresponding part of the cathode (right slde)
remalns in its original condition
Flg. (66.12). Due to SWINEHART [3). Fine silver in 100 % oxygen. Total arcing operations 16;
average arc amperes = 160; each arc duration = 3.5 millisecond; arc voltage = 19 at 0.127 cm
length. Very smooth arced surfaces. The areas adjacent to the arced surfaces have a black deposit
of silver in arcs varies strongly with the percentage of oxygen in the

ambient atmosphere. There is a marked minimum of erosion at 30 to
40% oxygen when the arc current is of the order of 200 to 300 A.
(Ag, CdO) produces nearly the same effect as an atmosphere with
40% oxygen. Fig. (56.10) shows the total erosion rate of both the
anode and the cathode. For the distribution on the electrodes see
other figures in SWINEHART ( 3).
Figs. (56.11) and (56.12) demonstrate that the erosion process is

qualitatively different in 100% N2 and 100% 0 2 • In 100% N2 the elec-
trodes lose metal at a tremendous rate, some of it leaving as partielEis
or droplets. At 100% 0 2 the erosion is considerable but leaves the
surfaces fairly smooth.
SWINEHART observes that the trace of the arc on the cathode is
wider in 0 2 than in N 2 • He states that the current density, J, is about
100 times less in 0 2 than in N2 • Since a small J means a relativ'ely weak
field strength at the cathode1 , it seems that the oxygen diminishes the
work function of the metal so that enough thermionic emission is pro-
vided at a relatively low field strength. The arc spots are relatively
great and thus weil cooled. This may explain the decrease of the
evaporation. But how the oxygen manages to diminish the work func-
tion is not explained.
Wolfram and the compound (W, Ag) wears about 5 to 10 times less
than silver alone. This shows that the framework of wolfram not only
strengthens the material but also obstructs evaporation and particul-
arly sputtering of silver. The reason for this is twofold. 1. The W-
skeleton obstructs sputtering mechanically. 2. The primary electrons are
emitted from the refractory metal by a combined thermionic and field
effect. The necessary field is generated in the silver vapor with a pressure
smaller than needed in a normal silver arc. The wolfram does not melt
at the respective temperature for thermionic emission. Thereby, the
current density in the cathode spot becomes relatively low, and the
silver is dispensed from providing the higher pressure that it needs
for a normal silver arc.
MERL et al. [5] investigated material transfer produced by damped
oscillating arc current from a capacitor. The initial amplitude reached
1.2 · 105 A at about 8000 V. AllSfirst (great) half-cycles carried about
21 coul, which could be referred to an effective current of 5 . 10' A.
The result can be expressed as:
w ::::::: 900 ·10-12 m 3 fcvul for silver
w::::::: 110 ·10-12 m 3 fcoul for-(W, 20 Ag)
w ""' 70 ·10-12 m 3 fcoul for (W, 20 Cu)
The greater part of the silver-contact loss was because·of sputtering.
C. Are quenching. Arcing can be avoided by operatiilg with a cur-
rent below Im, the minimum current required by the lirc 2• However,
even if this is done, bridge transfer can cause trouble; as is discussed
in § 67. Relays usually need a current higher than Im and harmful arcing
has tobe prevented by quenching the arc. Weshall find that induct-
1 Cf.§ IX.
s Cf. Table (50.03).
ances always exist which oppose a sudden change of the current, and
therefore arcing begins with the current I 0 that was flowing in the
moment of opening. The purpose of the quenching is to quickly di-
minish the current during the life of the arc. Quenching methods are
described in §§ 61 and 62.
D. Mechanical wear produced by the hammering e:lfect of switching
contacts. Besides the electrically produced wear (discussed above), a
purely mechanical wear appears in switching contacts. It is a minor
effect and usually negligible, but, in the contact rectifier it is of
great significance. KESSELRING and coworkers1 , when engaged in the
development of a contact rectifier~, used fine silver because a small
contact resistance was required. In order to keep bridge tranRfer3 as
small as possible, the switching was made in moments when the cur-
rent, I, was very small; namely I~ 0.3 A <Im. Usually material was
transferred to the cathode indicating that bridge transfer nevertheless
was involved. But, it also occurred that material was transferred to the
anode, which showed the presence of a new phenomenon. When the
contacts operated without current at all, elevations appeared at random
on the one or the other member. The amount oftransfer was about the
same as with the small current.
The very light movable member had merely to close and open a
narrow gap between two terminals. Nevertheless, by virtue of the high
velocity of about 0.5 mjsec at contact closure, the members hit each
other with a kinetic energy of about 5 · 10- 3 Nm. A material transfer
of up to 10-10 m 3 was produced in 108 operations.
The workers suggested the following picture ofthe process. Fatigue,
because of repeated impact and of some sliding, leads to loosening of
silver scales at the contact surface. Such particles assemble, probably
by means of a suction at contact opening and form a compaction on one
member under the influence of hammering. The other member (oppo-
site to the elevation) suffers an increased stress, and therefore the trans-
fer continues in the same direction, enlarging the elevation. The small
bridge transfer (at small currents) merely defines a preference for the
direction of the start of mechanical transfer. The compaction could
easily be removed as a whole from the u.nderlying silver. No sign of
weid could be. detected. However, a strong adherence, at least in very
small spots, must have taken place because a lubricant, even the thin
water deposit from a very humid atmosphere diminished the material
transfer 4• The relative loosely attached particles can bc removed by
i See HÄMMERLI [1].
2 Cf.§ 52E.
3 See§ 65.
'Cf. the inHuence of very thin liquid films on frictional wear; Table (41.09).
§ 57. Discharge transients 313
air blasts at contact opening. No compaction is then formed, and any
visible transfer by hammering is avoided. In vacuum where suction is
excluded, the surfaces become roughened, but no concentrated build
up appears. The authors ' succeeded in making rectifier contacts ser-
viceable for 108 operations by means of lubrication and blasting.
§ 57. Discharge transients

In circuits operating with small steady currents that themselves are
unable to ignite or sustain any arc, transient arcing may nevertheless
appear during switching operations, and may produce harmful wear
when frequently repeated. These discharges have a very short duration
because they are maintained by small capacities and inductances
whose energy is quickly consumed. An' extended investigation of tran-
sients in telephone switching circuits was published by CuRTIS [ 1] in
1940. For details, the reader is referred to his clear and concise article.
Figs. (57.01)" to (57.04) are due to CuRTIS. They represent typical
discharge transients.
The circuit for Fig. (57.01) contained a battery (50 V), an induct-
ance (relay) and a two wire line (100m) to the contact. The line con-
-600
-0(){)
+200
0.1 0.2 0.3 0.~ as ().ß 0.7

milliseconds-
Fig. (57.01). Aseries of ·discharge transients during relay opening, each starting with a glow at
about 300 V, and immediately developing to an arc. After the second, fourth and fifth arc re-
ignition occurs during the reversed phase of the electric oscillations in the Ieads
stitutes a capacitance, an inductance and a resistance. At t = 0, the

contact starts to open and is completely separated after 0.08 milli-
seconds. The steady current of 0.1 A is too small to ignite an arc at
incipient separation. Therefore, the current from the inductance is
utilized to charge the capacity of the Ieads, causing the voltage of the
separating contact to increase at a rate of about 1.5 · 107 V jsec. From
t = 0.08 to 0.1 millisecond the inductance charged the Ieads to about
300 V. This sufficed to ignite a glow in the gap. This glow immediately
developed to an arc which was maintained by the charge of the Ieads

for about 5 · 10- 5 sec. After this time, the arc had discharged only the
nearest part of the line and then extinguished due to a current defi-
ciency.
..
~ 2
6
~
~0
ti 2
~
6
8
0 at 02 03 a~
milliseconds
Fig. (57.02) records the current of a series_of discharge transients similar to those of Fig. (57.01)
~
....;:: -3001
~ 0
+200
0 zo 30
microseconds
Fig. (57.03). Voltage record of a single discharge transient with a resolution high~r than the re-
cords in Flg. (57.01). The ftgure shows the formation of a. sudden bridge over the gap (proba.bly
a Whisker)
0 10 20 30 so
micl'oseconds
Fig. (57.04) shows detalls of such transients as in Fig. (57.01) but with a higher resolution
Then, the inductance charged the Ieads anew until a breakdown

at about 300 V again appeared in the gap. Also, this glow immediately
developed to an arc which extinguished at zero current during an
oscillation. However, this time the plasma of the arc remained con-
ducting long enough to enable re-ignition of a glow in the reversed
direction during the next halfcycle of oscillation in the line. The pro-
cess was repeated several ti~es until the gap became too large for a
§ 57. Discharge transients 315
breakdown and the remaining capacitive and inductive energy dissi-
pated in damped low frequency oscillations.
Fig. (57.02) depicts currents of a series oftransients similar to those
in Fig. (57.01). The traces of the arc currents appear as spots.
Fig. (57.03) depicts a typical transient ofFig. (51.01) with a higher
resolution. Mter the contact is separated, the voltage across the gap
rises at a constant rate, being closed for a moment by a bridge.
Fig. (57.04) shows the voltage of some transients .as are depicted
in Fig. (57.01) but with a higher resolution.
It is evident that the presence of inductance and capacitance in
the contact circuit has Iead to arcing that otherwise would not have
appeared. The inductance provided voltages higher than the battery
could do. The capacitance could deliver current pulses beyond Im, the
minimum current of the arc, although the steady current was below
Im. The circuit used in recording Figs. (57..01) to (57.04) is principally
similar to that in Fig. (61.01). The line with its capacitance and re-
sistance corresponds to the quench circuit, and at the same time is
part of the main circuit containing the battery and the relay.
MARTIN and STAuss (1], using a wiring similar to Fig. (61.01),
produced the same types of transients as described by CuRTIS. They-
also found that capacities<10- 9 F ignited a single glow at 300 V which
did not develop into an arc but burned until final interruption. Laterl,
STAUSS observed transients in air at low pressureandin helium. The
single glow was then the normal appearance.
The transition from a glow to an arc has been described as happen-
ing as soon as the current is strong enough for the maintenance of an
arc. This implies two conditions; 1. that there exists enough insta-
bility to frequently produce a concentration of the current at the glow
cathode; 2. that such a strong concentration is accompanied by a de-
crease of the discharge voltage with the consequence that the total
current fiows into the path of minimum voltage. An arc is thereby for-
med, provided the total current is strong en«rugh. It seems that condi-
tions 1 and 2 were always realized in CuRTIS' experiments. An instabi-
lity of the described type has to be produced by an effect that increases
the efficiency of electron emission per ion at the cathode; cf. § 49.
Recently, BoYLE and HAWORTH [1] discussed cases where the in-
stability seemed to derive from the average field strengthat the glow
cathode, when it is reasonably close to the field strength at an arc
cathode. Cf. § IXA.
An instructive discussion of the infiuence of the Iead capacity on
the transient arcs is given in GERMER [13] p. 244.
1 -According to a lecture given at a meeting of ASTM in 1952.

§ 58. Are duration during contact closure with voltages below

200 to 300 V
A. Introduction. Arcing at contact make takes place because of
several reasons. 1. Ares may be drawn at rebounds. This event shall
not be discussed here. 2. The contact members may first meet solid sur-
face asperities. These are heated to vaporization by the current, thus
interrupting the contact. In the superheated, highly ionized vapor, an
arc ignites. Such a happening is apparent from Fig. (58.01). 3. Germer
and collaborators assume (see § 49 B,) that an arc can ignite during clo-
sure in the narrow gap at points of enhanced field emission. In this
book another idea is represented, namely that a whisker forms and, when
5000
c.p.s.
time
Fig. (58.01). Due to R. HoLM. Voltage record of an arc during contact closure: using rough silver
electrodes at P = 1 N and v ""0.1 rn/sec. The voltage drops from 110 V to zero at the first touch
with a protuberance. The tJIOtuberance evaporates producing a vapor-tllled gap in which an arc
ignites at 22 V. This arc extinguishes after w-• sec at 11 V
hit, shortcircuits the gap, thereby exploding and igniting an arc1 . This
means an ignition similar to that of caEe 2. It is a special question if the
arc finally produces floating as decribed in §59.
The final closure after arc ignition according to 1, 2 or 3 without
floating often takes place after the movable member has travelled about
10- 5 cm to solid contact. In relays the movement is usually not per-
ceptibly braked by the vapor, and the closure may appear 10- 6 to
10- 5 sec after arc ignition. This is then the duration of the arc.
Section B contains formulas that are useful to have at band for
trea.tment of several pertinent contact problems (not only those treated
in this book). Thereby an important task is to calculate the quantity
of ele?ti-icity, q, transferred by the arc because q determines the amount
of material transfer according to Eq. (56.05).
B. Calculation with respect to the wiring diagram (68.02) with
initially charged capacity. This case is encountered in the problern of
1 See p. 277.
§ 58. Are duration during contact closure with voltages below 200 to 300 V 317
the quench circuit of § 61. The arcs at contact closure are short and have
a voltage that differs only slightly from the minimum arc voltage V m·
Therefore, calculations can be made with the assumption of a con-
stant arc voltage = V".. The wiring diagram in Fig. (58.02) to which
the calculations are referred, may have no other inductance than the
small inductance l resulting from inevitable loops in the wiring. How-
ever, the function of this circuit when it is used
r l
as a quenching circuit, is surprisingly dependent
on this small inductance.
It is assumed that prior to closure the ca- ~
pacity has the voltage E. Due to l, the current Fig.(58.02). Wiringdiagram
is una.hle to begin with a finite value, but rapidly
grows from zero tu a maximum I max. We are particularly interested in
the ratio of this maximum to the current I 0 that would initially appear
in the arc if l = 0. This ratio I maxiI 0 is a function of lj0r2 which we
labe! ß. The .calculations as weil as Fig. (58.09) are formed so as to
emphasize this relationship.
We investigate the variation of the voltage, V, across the capa-
citor, 0, and of the current, I, in the circuit during the time the arc
is burning in the closing switch. Although the mathematical problern is
weil known, it is appropriate to arrange the formulas so as to fit the
present problems. The initial conditions are: in the moment the arc
ignites in the closing contact, i. e., at t = 0, V = Vm and I :::::: Im (mini-
mum arc current). Simplifying, we put Im = 0. Let the momentary
charge on the capacitor be Q.
The capacitor voltage obeys the equation
Q dl
(58.03)
V= 0 =ldt+ri+Vm
Since
dV
l = ·- 0 - (58.04)
dt
Eq. (58.03) can be transformed to
~ r d 1
dt2 (V- Vm) + T dt(V- V'")+ Cl(V- Vm) = 0 (58.05)
the well-known equation for damped oscillations. With the initial con-
ditions mentioned, and using the following symbols
1 r
e=--
2 l
(58.06)
m- Ve - Ci -- Tr 1'2 .,
- lr::;:-1 I _ E-Vm.
o- r '
tm =time when I has its maximum, Imax' and I 0 = initial current that
would ßow through the arc if l = 0, we obtain the f.ollowing solutions:
In the case of oscillations which require ß> !

V- V m = E ~~m e-<~cos (wt- tan-1 .!.._)
wvCl ())
E-V . I ___!___..!_. t
I= ---"'e-etsinwt = _!!e 2 fi Or sm 1'1 -
wl y1 ß Gr
Imax = I ~exp ( - tan-
1 2y) 1 for~
t ß
= -tan- 1 2y 1
Vß, 21'1 Gr y1
In the case of aperiodicity with ß = !
Imax= 2 Io at ~ = ..!_ (58.07)

e Gr 2
In the case of ß~ 1 we o btain
Crt = ßln ß;
1 .
lmax = I 0 ß/J
At t = 0 we have generally
di = E- V,. thus _!-_[__ = I 0 (58.08)
at z a(d.) ß
The solutions are illustrated in Fig. (58.09). However, in the figure

the curves hav.a bcen extended to cover greater values of tjO r than are
§58. Are duration during contact closure with voltages below 200 to 300 V 319
possible under the condition of the arc burning. This has been done in
order to make the figure available for other contact problems leading
to the same differential equation. The straight dashed lines in Fig.
(58.09) show slopes at zero current according to Eq. (58.08).
In arc quenching circuits l will be smaller than 10-5 H and ß
smaller than 0.1. Thus one would not expect oscillations with I pas-
tOrr--~--~---r---r---.--~----r---,
-azo~--~1--~~--~3--~~~--~5--~~~--~7~~
t/Cr-
Fig. (58.09). Variation of the current in the circuit of Fig. (58.02) after closing the contact at
t = 0. In case of {J = 0 the current would begin with the value J 0 •
ing its zero value during the life of the arc. However, we shall find that
the arc introduces a negative resistance (cf. §54) that subtracts from r
and renders oscillations possible, see § 61.
The solutions (58.07) and Fig. (58.09) are applicable to the case of
an instant contact make at t = 0 without previous arcing, by putting
V m = 0. Here the curves are valid throughout their whole extent.
C. Inductance l = 0 (thus fJ = 0) in the circuit (58.02). This means
infinitely rapid arc ignition, i. e., the initial condition is I = (E - V m)fr
= 10 • The sequent current decreases super-aperiodically according to
Eq. (58.10) and to the curve that begins at the ordinate 1 on Fig. (58.09)
t
_E--V,. -er
I - ---e (58.10)
r
Consequently an arc with the duration ta will be passed by the quan-

tity q of electricity, where
E V ta - -
t ( --Ia)
q =~Je Cr dt = C(E- Vm) 1- e Cr (58.11)
0
If t.fOr 4:. 1, Eq. (58.10) may be simplified to

E- v... t
q<=:::--r- a (58.12)
D. A battery instead of C, equivalent to C = oo; finite l. Then Eq.

(58.03) becomes simplified to
dl
E = l dt + r I + V",
with the solution
I= E: v".(1- e--Tt) = I0 (1- e- *dr) (58.13)
Hence
(58.14)
Particular cases :
If l = 0 we ha ve
(58.15)
r
If T t. 4:. 1 we have
(58.16)
§ 59. Floating
With respect to arcs, we speak of "fioating" when evaporation pro-
duced by the arc is so abundant that, for a while, a certain gap can be
maintained in spite of the mechanical contact Ioad. In this gap the
arc burns. This phenomenon has been observed during contact closure
by R. HoLM et al. [16]. NEUKIRCHEN [lJ observed a similar pheno-
menon, but with initially closed contacts. In the first event, the voltage
was not exactly measured but merely stated tobe ofthe order of 10 V.
As soon as the current decreased below a certain Iimit (that depended
on the load), the contact closed. The loadwas 0.2 to 1 N. HoLM used
the term floating 1 to express that the electrodes are held separated by
the vapor produced by an arc.
1 R. HoLlii [30] p. 280. The electrodynamic repulsion according to Eq. (11.03)
was negligible, namely less than 0.01 N.
§ 59. Floating 321
In NEUKIRCHEN's experiment, carbon members made contact in
the center of their slightly curved faces. The contact was subjected
to transient pulses of 6.4 V and about 90 A with the duration of a few
seconds. When the mechanical load was less than 2.7 N, the contact
was pushed open by the vapor developed in the overheated contact
Anode Cafhode
Flg. (59.01 ). Due to NEUKIRCHEN [1]. Appearance of the carbo .. electrodes

after arcing during floating; magniflcation 3.5
and by the arc that burned at 5.6 V. One may ask why the contact
area did not yield and adapt itselfto carry the current without boiling.
It is evident that the totalload bearing area conducted. Since carbon
does not melt before sublimating, no increase of the co,n tact ·area hap-
pened. As indicated by Fig. (59.01), the materialtransferwas directed
from the anode. A Ioad of 5.1 N was needed to prevent any evaporation
at 90 A.
The vapor pressure during floating is tremendous. In Fig. (59.01a)
the two pits on the anode (both certainly did not simultaneously serve
as anodes) show a deposit beyond their area of 0.04 cm diameter.
Since the material is carbon, the deposit must derive from vapor. Thus,
21a Electrlc Contacts, 4th Ed.
the high pressure is located only within the pit area. This pressure is
L15/n (2 . 10- 4 ) 2 ""' 107 Nfm2 or about 100 atm.
In the gap between the metal electrodes in HOLM's experiment,
the vapor pressure probably was even higher. It is evident that this
arc has other properties than the normal arcs described in § IX. The
small arc voltage of 5.6 V in the carbon arc during floating indicates
that a normal cathode fall did not exist. It is very reasonable to assume
that the whole arc consisted of a plasma-like, highly heated and dense
vapor. The current transit through this vapor is, nevertheless, here
said to be carried by an arc because a voltage of 5 to 10 V is required.
In HoLM's experiments the source was a capacitor. In order to cal-
culate the charge q that passes through the arc, such oscillograms as
300 illustrated by the Dia-
Amp gram (59.02)were usedl. A
Z50 current of 300 A passes
zoo
'( through the initial un-
adapted contact spot. A
\ very short time later a
150 1\.
lf '
'\ kind of arc burns requiring
100
' ~Is
" ........_
a certain voltage andlower-
ing the current to 240 A .
Mter 3.2 · 10- 5 sec, at
50
--.... r-- r- 1 8 = 85 A, this arc is no
0 1 z J 'f 5 8 7 8 •10. 5sectQ Ionger able to keep the
Flg. (59.02). Cu.rrent varlatlon in ca se of tloatlng movable contact member
floating, and the contact
closes as is perceptible by the sudden rise of the current.
Using the oscillogram (59.02), q is calculated as follows. We apply
the symbols of Diagram (58.02) and neglect l since it has here practi-
cally no influence. At the instant of I = I8 , the capacitor voltage is
V= I,r + Va where Va is the voltage of the arc, about 6 V; the
charge is then
q=C(E-I,r-V.) (59.03)
HoLM observed with normal electrodes2 (Cu, Ag, Pt) that I , = 60

to 80 A under a Ioad of P = 1 to a few newton. With pointed electrodes,
18 was about 140 A. With wolfram electrodes, I. was 140 A at normal
electrodes.
1 "Fig. (46.10) in HoLM [29] showssuch an oscillogram.

2 For normal electrodes see Fig:'(56.06).
§ 60. Are duration on breaking contact without quenching 323
§ 60. Are duration on breaking eontaet without quenehing.

Cireumstanees eommon to Communieation Engineering
For the case that no suitable oscillograms are available for in-
formation on the arc duration t4 and the variation of both the voltage
and current of the arc, the following method of calculating may be
applied. When not otherwise stated, it is assumed that voltage and
current satisfy the conditions for drawing an arc, and that the static
V I-characteristics of Diagram XI are applicable; cf. §50.
A. Ohrnie eireuit aeeording to Fig. (60.01). Are extinetion in air
during a steady inerease of the gap. This problern may be solved by
application of Diagram XI in the manner de-
H
scribed in § 50 a$ the method with the resist-
~ ance line. The intersections of this line with a
Tf sw/fc/1~ series of characteristics gives (V, I)-points char-
acterizing the state of the arc as a function of
Fig. (60.01). Ohmic circuit
with arcing switch its length 8. If V is the Velocity of Separation,
8 can be converted to time, t, with 8 = J vdt;
or, in the case of constant v, by 8 = vt. Thus, V and I may be ex-
pressed as functions of time.
In the following, we are particularly interested in the quantity q
of electricity that passes an arc. An approximate calculation of q can
be made in the following way. With the aid of the resistance line and
Diagram XI, we determine the lifetime, ta of the arc, the current I ; at
ignition, and I, at rupture (touching point between resistance line and
a characteristic). Then, we put q equal to the product of ta and the
average current through the arc, which is approximately (I; + /,)/2 .
Thus,
(60.02)
When suitable oscillograms exist

Ia
q= j I dt (60.03)
0
can be obtained by graphical integration.

Both formulas (60.02) and (60.03) are also valid for vacuum arcs,
but, as shown in §51, the use of Diagram XI is then not applicable.
B. Induetive circuit in air aeeording to Fig. (60.04). Capacity of the
Ieads is neglected. With the capacity being equal to zero, the same
current passes the whole circuit. On breaking the circuit, the induct-
ance prevents any discontinuity of the current. Thus an arc ignites1 at

10 = (E - U)f R where U is the last voltage across the metallic con-
R l tact. Mter the arc has been ignited, the current
~ I varies according to
TE swifcl! ? E = RI + L :: + V(J,s) (60.05)
where V (l,s), the arc voltage, is a function of

Flg.(60.04).1nductlve circuit
with arclng swltch I and the arc length s.
If Iris the current at arc rupture, the energy dissipated in the arc
during its life time ta is given by the following first integral to Eq.
(60.05).
fV to t.
f Ir l
0
Idt = j(E- Rl)ldt-
0 Io
Lid!
j (60.06)
[(E-
t.
= RI)ldt +! L(I~- P,)
t.
The integral J (E- Rl) I dt gives the energy supplied to the arc by
0
the battery during ta. Eq. (60.06) shows that the arc also dissipates the
inductive energy 1f2L(Po- .Pr)
which will be nearly the whole
energy of the inductance if
I~~ I~; cf. §53.
The general integration of
Eq. (60.05) is difficult. A sim-
plified stepwise, algebraic me-
thod combined with graphic
integration is proposed in
R. HoLM [37] § 59C. For the
examples, given there, the time
intervals chosen are too great
to give exact results. Amended
calculations Iead to the curves
\ in Fig. (60.07). By virtue of L,
\
the arc ignites with the current
0 as 10 1.S z.o E f R that flows at the begin-
Fig. (60.07). Are characteristicsln case of inductive ning of opening (we neglect U,
circuits; silver electrodes, FJ = 54 V, R = 54 n, se- the contact voltage). But, if the
paratingspeed = 5cmtsec;I.. = 0.4A, Vm = 12V,
Curve A : L = 0.1 H; curve B: L = 5 · 10-a H time constant is very small, sa y
1 Here the ignition current is Iabelied 10 instead of 11 in order to obtain agree-
ment with Fig. (58.09).
§ 61. Quenching a drawn arc by means of a capacitor... 325
Lf R < 3 · 10- 4 sec, the influence of L has a short duration and the nor-
mal resistance line describes the continued process as is indicated by the
curve B of Fig. (60.07). Curve Ais obtained with LfR = 1/540. Both
curves refer to an opening speed of s = 5 cmfsec, Im= 0.4 A and V m
= 12 V, the normal values for silver cathodes.
C. The quantity of electricity that ßows through a drawn arc with
the life time ta. Calculation of the material transfer caused by arcs
requires a knowledge of q. For rough approximations Eq. (60.02) can
be used. If ta is unknown, the following method is often convenient.
During ta, the current source and the inductance deliver the energies
qE+ ~ L(Ji -.Pr}
On the other hand, the arc and the resistance R dissipate the energies
t.•
qV +f I 2 Rdf
0
where V is an average arc voltage. Hence
qE+ 21 Lm-Pr)=qV+
- f t.
I 2 Rdt (60.08)
0
The last term often is small enough to be treated as a correction for
which an approximation suffices. Putting I= I;(1 - tfta) in the re-
spective term, one obtains
t.
ji Rdt=!Rfita
2
Introducing here the approximation ~ l;ta "." q
f
t.
renders I 2 Rdt = ~ R/l = : qE
0
and after substistution in (60.08)
r'- E ;3
L J?- J2
q "." -2- (60.09)
§ 61. Quenching a drawn arc1 by means of a capacitor, parallel

to the contact
A. General theory. Fig. (61.01) shows a main circuit connected with
a queuehing circuit. The main circuit contains the battery with the
1 About "quenching" in case of a current below Im, see § 62.
emf E, tne inductance L, the resistance R, and the contact. It may

represent a relay. The quenching circuit parallel with the contact,
includes the capacitor 0 and the resistance r. Although one will strive
to make this circuit inductionless, a small inductance, l, will always
result from its geometrical dimensions. We shall find that this inevit-
able inductance in the quench circuit is an important variable of the
problem.
Positions a or b of r will be discussed below.. At the time this
quenching circuit was deviced about 100 years ago1 , the presence of
the small l was not considered,
and the Ieading idea was as follows.
Since the inductance L prevents
any sudden change of the current,
moin the current in L will be I 0 = E j R
T
clrcult
immediately after contact opening,
and V.fr A would fiow into the
capacitor, where v.is the (varying)
Fig. (61.01). Inductive main circuit, and 1 b h 1 d
quench circuit with capacltor c, reslstance r vo tage etween t e e ectro es.
and the arclng contact. Two possible positions Thus, the current a vaila ble for
of r in the circuit are 1ndicated as a and b.
· v. = arcvoltage any spark would be ! - :· = I8•
The spark can develop to an arc only if V.> Vmand I.> Im (when
Vm = minimum voltage and Im = minimum current of the arc). lf
! ,. .- ~"' < Im, no arc can develop.
As mentioned above, l never is nil and therefore the current avail-
.able for the arc immediately after contact opening is I 0 = EfR. An arc
then ignites if EjR >Im, as it is supposed tobe in this chapter. Thus,
quenching by a capacitor does not prevent the ignition of the arc. Its
purpose is to shorten the duration of the ignited arc. This fact is often
neglected in the literature. It is evident that the queuehing is more
ef:ficient as r becomes smaller.
Earlier, one used a constant r and had to consider that, at contact
closure, a small r permits 0 to deliver a strong discharge current which
may cause a detrimental material transfer. and welding. One had to
compromise. The optimum is achieved when the material transfer is
the same at opening and closure. The problern to calculate the condi-
tions for this optimum is treated in R. HoLM [37] § 67. We do not
repeat .it here because nowadays one usually does not apply a con-
stant r. Instead, one uses a semiconducting device that exerts a small
resistance, r, against the current at opening and a high resistance
against the current at clo"'ure 2 •
1 See R. HoLM [.17] p. 327. 2 Cf. PARRISH [1] and GonsAY [1].
§ 61. Quenching a drawn arc by means of a capacitor... 327
How much the arc life is shortened by the quench is treated. in

R. HoLM [37] § 60F. A mixed theoretical-empirical method is used.
It is not repeated here, partly because the theoretical solution is now
in little demand, and partly becausc the method used is limited to
the narrow region of variables that was tested. Test results are given
in Table (61.02).
Table (61.02). Duraticm t. of arcs quenched by a capacitor circuit according to

Fig.(61.01), during contact break. The arcs have been observed by taking oscillograms.
10 = short circuit current. G. = total cathode loss per Operation and y = cathode
loss per coulomb transported by the arc
E Io L 0 r V t. lO•Ge y
obs.
10-6 cm3
V A H ~>F n -cm
sec
ms 10-6 cm3 - -
coul
Silver contacts
114 1.14 0.1 10 33 6 8 110 0.22
110 2 0.1 150 15 6.5 1.7
61 1.5 6.2. 10- 5 2 15 6.5 1.5
61 3 0.0012 300 22.5 6.7 3.5 180 0.17
109 2 0 140 19.5 6.7 2.6 110 0.38
8.5 4.05 0.0012 10 3.5 8 Oto0.4 22
Platinum contacts
110 2 0.1 2 23 6 7.5 1000 1.4
113 2.02 1.0 77 25 7 2.5 300 1.6
110 2 0.3 150 40 6.1 5.0 440 1.0
110 2 0.1 150 65 4.8 2.7 445 1.0
110 2.3 0 2 20 6 10
8.6 4.05 0.0012 10 5.21 8 Oto0.2 38
110 3 0 70 16 7 Oto3
110 9.3 0 1 0.13 7 (2.2) 370 1.5
B. A short deseription ot the proeess in the circuits (C, R + r, L)

and (C, r, l) during the life ofthe arc will be appropriate. At the moment
of opening, when the arc .ignites, changes begiii in both circuits. There
usually are oscillations with frequencies VLJl times higher in (0, r, l)
than in (0, R + r, L). Charging of 0, in the circuit (0, R + r, L), leads
to a decrease of the arc current whereby the situation in (0, r, l) be-
comes favorable for oscillations with increasing amplitude1 • Mter a
number, n, of cycles, the arc current reaches zero and the arc extin-
guishes. The process is shown in Fig. (61.03). However, these oscillo-
grams were obtained with enhanced l and show a greater n tha:n usu-
ally appears under service conditions. The fact that the arc (in the
manner described) shortens its own life, has an important bearing on
1 Cf.§ 54.
practice. It makes it possible to construct quench circuits with suffi-

cient efficiency using relatively small and cheap capacitors. In the
examples noted in Table (61.02), the quench has diminished the
material transfer Ge by a factor of the order of 0.1.
500 cps
.
0
.
0,07 6!C
.
0 O,Oisee
()' - - -- - -
Fig. (61.03). Due to R. HOLM. Oscillations excited by an arc in the quench circuit, increase their
amplitude until they are able to cancel the arc current
The number of oscillations in the circuit (C,r, l) until the arc ex-
tinguishes is of the order of 10, usually during about 10-4 sec. The
capacity C should be great enough to be only partly charged by the
current E ! R during this time, say to the voltage of E j2. Hence
!o".,10-4 ! say 0 > 5 · 10-4/R

This agrees with tests by GonsAY [1].
C. Note concerning the position of r. In wiring diagram (61.01) two
positions of r arP. noted, namely a and b. Wehave treated position a.
It is the mostfavorable position as regards the queuehing effect; but
position b is only slightly inferior. BuRSTYN1 regarded this independ-
ence as a criterion for the existence of oscillations. And actually the
development of oscillations leading to the extinction of the arc must
be independent ofwhere r is located in the oscillating circuit (O,l,r,arc).
1 ßURSTYN (2] p. 26,
§ 62. Capacitive quenching when no arc is drawn 329
§ 62. Capacitive queuehing when an arc with a very small

duration or no arc is drawn
A. Equations for the voltage v. in the gap between the electrodes.
With a short circuit current that is smaller than the minimum, cur
rent, Im, of the arc, no arc is drawn. Nevertheless, a quench can be
valuable because it can prevent a glow discharge from igniting and
developing to transients of the kind described in § 57.
The following calculations refer to the event when no arc is drawn
They arealso approximately valid when an arc is drawn ifits life-time
is very much shorter than the timetm defined below. For simplicity, we
calculate with a constant L. In reality, L varies considerably with the
armature gap and a suitable average value of L has tobe employedl in
the calculation.
Before contact opening, the current in L is EfR, with no current
:O.owing in the quench circuit and the voltage of the capacitor is zero.
At t = 0, immediately after contact opening without an arc, the total
circuit (E, R, L, r, C) experiences the current 11 = E / R. Since l is very
small compared with L, it can be neglected. Using LAPLACE transforms,
one obtains:
V• (t) = E -L
'
"w11 (~-
Gr
R)e-IXtsin(wt- m)
L r
where
R+r
Qt=--u· (62.01)
and
L- -~lR2 + r2)
.
wLO(R- r)
smrp= L-ORr ' cos rp = L - 0 Rr
l
The maximum of V. is
v.' ,. - E_ ...L
VG Rr e-«t,. -_ E [ 1 + L-V0 Rr e-"'tm]
, 11 L-
LG R LO
with (62.02)
1 [:n; - -;ac + VL/0
00 tan- -;- + rp ""' 00 2 1
t111 =
1 [ w ] 1 R- "]
At the time of contact break, t = 0,
V.(O) = 11 r and (d....V,]e t= o=

11
fJ
(1 - 0L"
2
] (62.03)
The complete solution is illustrated in Figs. (62.04) to (62.07). The first

two diagrams correspond to the most common cases with r < R.
1 See for instance BORCHERT [3).
21b Electric Contacts, 4th_ Ed_
330 Electric·Phenomena in Switching Contacts
8
0 ß J 'I s
t .-
C(H+rl
Flg. (62.04.) Fig. (62.05)
><E '
r-... I
S.'J
• I 1\
I I
'1. f -fJrC'
'f:O
I; r-.. \ 1- i- JH -7'
J.S Jl \ i\
\,~-1
1\ IV-:~ \
'\_p-J \! '
1.!i
E
~},,,1\\ -- -;v~-7 (
~'/J-f .\
I
). ~8 t/trV
91- 10 n/ 1/J
O.!i jl-o
1 i ß .J
i
'I • 1Z I'·
r
0
.I f .J_Lf~7Ai
I t:/H+r - L
" ' 3
____l_-
f:(H+r)
'I !i G
(Flg. 62.06) Fig. (62.07)
Figs. (62.04) to (62.07). Variation of the voltage between the electrodes after contact break in
the clrcult shown in Fig. (61.01), if no arc lgnites.
§ 62. Ca.pacitive queuehing when no arc is dra.wn 331
.lfig. (62.07) shows a particular case characterized by r = Rand by the
aperiodic curve being a straight line V= E, obtained if
L
R2 = 75 (62.08)
Eq. (62.08) is a well-known condition for the parallel impedances,

R + jwL and R + . 1 0 , constituting a purely ohmic impedance.
JW
Remark concerning Eqs. (62.02) and Fig. (62.05). The maximum
voltage that appears between the electrodes at opening a switch with-
out arc is given by Eqs. (62.02) if
C is regarded as the capacity be- Volt
tween the electrodes together
with adjacent Ieads, and r as the
resistance(increased by skin effect)
of the same assembly.
B. Condition for no breakdown
V
of the gap between the separating
eleetrodes. A breakdown occurs if
v. reaches or surpasses the di- time -
electric breakdown voltage of the (/ ~ 8 16 IG tll 6'1 _.ft._
or eleclrotle rlislonce - ,.;7:
gap. Fig. (62.09) illustrates the
Fig. (62.09). The curve V gives the voltage be·
boundary case in which no break- tween the electrodes as a function o! the time
according to Fig. (62.04), or as a function of in-
down occurs. The curve {11 r, V max, creasing distance s, if s = fit. No breakdown
V) taken from Fig. (62.04) gives occurs if the course of this eurve, as on the
the development of the voltage flgure, lies below the straight llne V • that re-
presents the breakdown voltage as a function
between the separating electrodes. of Bor t
The straight line represents the
breakdown voltage, Va, as a function of the increasing gap. Using
s = vt and measuring Vd in volt, v in cmfsec, and t in second, the
equation of this line is
vd = 300 + 6.9 ·104 vt (62.10)
l
Denoting the voltage between the electrodes by V., one obtains as a
condition for no breakdown
V•. < 300 + 6.9 · 104 vt

at the point where (62.11)
dd~' = 6.9 · 104 V
C. Example. Consider a reed relay with a current of the arc lower

than Im, first with quench then without.
Assumptions: The main circuit has E = 45 V, R = 150 Q, L
= 0.1 H. The quench circuit has r = 50 Q, l is negligibly small, C
= 2 · 10- 7 F. These conditions correspond to the curve for ß = 200

of Fig. (61.05), from which is read that the maximum gap valtage is
45 · 4.65 = 210 V. This is lower than the smallest glow ignition volt-
age, 220 V, of the gas1 in the container. Thus, no glow ignites at con-
tact opening.
Without quench a glow ignites. Supposing that this happens at
250 V in the gap, the fundamental equation is L dijdt + R i + 250 = E
with the solution
.
l =
250-E
R r1 - exp (- t Rfl)J + ER exp (- t R/L)
The glow burns t sec, until i """ 0. Inserting assumed values we find
t = 1.32 · 10-4 sec
In this time, 0; 3 1.32 · 10- 4 = 2.10- 5 coul pass through the glow. In the
glow, the volume of cathode material evaporated per coulomb depends
on several <;:ircumstances 2 • A reasonable value 3 is 10- 6 cm3fcoul. Applying
this we find that the wear per operation would be 2 · 10- 11 cm3fcoul.
This is about 1000 times more than the bridge transfer could produce
(independent ofwhether the quench is used or not). The wear would be
detrimental after less than 106 operations. An optimum would require
an arrangement that causes the glow to burn away bridge deposit.
§ 63. Quencbing of arcs by a resistance parallel to the

Operating contact or to the inductive coil
The wiring diagram is given in Fig. (63.01). The small unavoidable
inductance that is associated with r does not interfere in any perceivable
degree and will be neglected.
· A. Quenching with r in position a. This kind of arc quenching is
often used in Tnm.L valtage regulators. Here E is not a battery as in
Fig. (63.01) but a generator, and (R, L) is its field coil. At the set valt-
age of the generator the field current flows through the closed contact.
But if the generator increases its voltage above the set Iimit (say, be-
cause of increased speed) a relay opens the contact and thus r is in-
serted in series with the field coil. This minimizes its current and as a
consequence diminishes the generator voltage. When the valtage has
then decreased below its set value, the relay closes the contact. The
~e gas is not air.
2 See R. HoLM [87] § 641 a.nd FRANCIS [1] p. 158.
3 Cf. HAWORTH [1].
§ 63. Quenching of arcs by a resistance parallel to the operating contact 333
average field current and consequently the average voltage of the

generator, at a given speed, depends on the ratio between the inter-
vals of open and closed contact. Contact opening is accompanied by
the ignition of an arc. In order to minimize its corroding effect on the
contact members, the problern
r in posifion b
arises to choose the circuits
rr vr-h
with a given contact material r
I
----c===r-----1 I
so as to keep the duration of
the arc as short as possible.
This problern will now be dis-
cnssed.
I
EL
~pos:ton
rin
In the TIRILL regulator,

t.he movement of the operat- Fig. (63.01). Wiring diagram illustrating arc quench-
lng by means of a shunt
ing contact members is slow
and the gap remains small. Then the arc current rises only little above
I",. Diagram XI shows that the average arc voltage is ""' 2 V", under
;such circumEtances. Considering this, and labeling I = current in L,
i = current in . r, ia and 2 V", current and voltage respectively in the
arc , we obtain the following equations for the considered cases with r
in position a. The equations are valid during the life of the arc. As in
l
§ 62 we calculate with L = constant, chosen as a suitable average.
E = R: + L ddl'. + ~ V", (63.02)

Vm=r~: l=~ + ta
with the condition that, for t = 0,

E
J(O) = R and i(O) = 0 (63.03)
The solution is
. R )
i = E _ V,._2V,.(t-e-Lt (63.04)
a R r R
The arc life, t., will be the time until i. becomes too small for the arc.
This time usually is so small that ~ ta ~ 1 and (63.04) can be written
i = E _2V,._2V,.t (63.05)
a R r L
giving for the duration of the arc
t
a
= .!!._
V,. R
(E _ i _ 2rV"')
r
(63.06)
where ir is the current at which the arc ruptures.

It is important to note in these equations, that L is the inductance

opposing superimposed undulations of the current in the field coil. This
inductance will be several times smaller than the inductance measured
with A -0 in the demagnetized coil.
As for i" it is always greater than Im. This is seen from Diagram
(60.07) according to which
(63.07)
B. Are quenehing with r in position b. The same apparatus as in
Section Ais considered. The difference being that r is in position b. The
arc voltage is assumed to be about 2 V m. Instead of the second line in
(63.02), we now have
E - 2 V m= ri and ia = I + i
with the solution
ia = ~- 2 V..r- E- 2 ~"'[1- exp(- tRJL)] (63.08)
where the second term is smaller than the corresponding term in (63.04),
making ia greater. Hence, we conclude that as far as the amount of
i 4 is concerned, position a is morefavorable than position b. But, this
may not be true in other respects. For instance, if it is required that
no current shall flow in the main circuit in times of open circuit, r must
be placed in position b.
§ 64. Details about the types ol arc in relay contacts

and the material transler produced by them
A. Snrvey. Chapter 56A defines the concepts of anode and cathode
dominated arcs. It also introduces symbols that we here repeat with
some additions: w.
and wa
= volume evaporated per operation from
the cathode and anode respectively; ru = Wfq where q is the charge
(coulomb) that has passed through the arc; D = redeposit from the
metal vapor; and G= net volume loss from an electrode per Operation.
In the case of arcing, one usually measures the difference between
evaporated and redeposited volumes. For example
G, = W. - D. or, per couJ: y ~ ru - t1
Imagine the drawing of an arc. It begins with decreasing Ioad, P,
causing a decrease of the contact area. The constriction resistance
steadily increases as well as the temperature produced by the current.
Finally, a liquid metal bridge is generated. The bridge suddenly rup-
§ 64. Details about the types of arc in relay contacts . . . 335
tures whereby the rupturing section may be heated to boiling. Boiling

and rupture have often the character of an explosion that may splash
metal droplets into the surrounding of the bridge and push apart the
liquid halves of the bridge.
This process has been studied with different metals and atmos-
pheres by the LLEWELLYN-JONES' group. See PRICE [1] who ~sed· a
prism camera capable of obtaining up to 16000 pictures per sec. His
film indicates turbulance within some bridges. Its role for. the pheno-
mena here described is not clear at present.
The initial vapor column is strongly overheated, ionized and suffi-
ciently conducting to carry an incipient discharge that immediately
uevelops to a regular arc. This is evident from its typicalvoltage of the
order of 10 V appearing without any noticeable delay; and a current
density1 of a few 1010 Afm2.
~\11 the arcs produced by inductance energy and discussed below
in § 64 as weH as in § 65, behave in this manner as far as is known.
We distinguish the following regularly occurring stages or periods
during a steadily increasing separation of the electrodes:
Period I. Development of the liquid metallic bridge up to its rup-
ture, accompanied by "bridge transfer".
Period II, belonging to an arc with a length shorter than a critical
length, s•r• (defined below) and characterized by a material transfer
from the anode: the so-called anode dominated arc.
· Period III, during which material transfer from the cathode con-
sumes the previous transfer of the opposite direction.
Period IV, with an arc that produces net transfer from the cathode.
Periods III and IV are referred to the cathode dominated arc.
Chapter 65 will be devoted to the process during periods I and II.
The current lines diverge from the cathode spot, A., towards the anode
thus rendering the anode spot, Aa, !arger than A,. This means that
heat is better conducted from Aa into the anode than from A. into the
cathode. However, as long as the length of the gap is smaller than the
critical value, 8 00 the ratio AafA. is not great enough to cause Aa to
be cooler than A.. For this reason, period II extends until the gap
length scr is reached.
When s.r is passed, A. becomes the considerably hotter electrode
and loses more material than Aa. At the end of period III, the material
transfer from A. has compensated the previous transfer from Aa. From
then on, the cathode loses material.
B. Result of measurements. In Fig. (64.01), the ordinates show
the net transfer to the cathode until the time given by the respective
1 Calculated on the basis of the bridge dimensions.

abscissa. The figure corresponds roughly to measurements by RIDDLE-

STONE [1] on palladium contacts with I = 5 A and a separation speed
of up to 2.5 cmfsec. The solid curve is designed for the assumed case
of a constant current. Actually, the current decreases with time causing
the slope to decrease, as indicated by the dashed curve. In addition,
the slope dGcfdt = rJ (where the Subindex c of r refers to cathode
dominated arc) decreases not only because the factor I diminishes, but
also by virtue of 'Yc being a function of I becoming smaller with de-
creasing I, as is shown by Fig. (56.09).
The branch representing period II in Fig. (64.01) starts at a point
P near the ordinate axis, which measures the preceding bridge tran~;fer1 .
When the cathode gain has reached the rnaximum of 50 . 10- 12 cm3 ,
period III begins. On the figure, this period ends at about 24 11-sec
after the rupture of the con-
so tact. During period II the
r;:,fl"-. average slope dGufdt is very
I ~~~ steep. With a symbol used
JO ~ ~.,.~ ·
!() ~A in R. HOLM [37] this slope
~ ~~ is expressed as dGufdt = r al
where the subindex a of r
10
0 "-'" ~
~......_ ....... refers to anode dominated
-10
0...___. s 10 15 ;;o zs 30 J.!oSeCM arc 2 • According to measure-
][ 1 JTperiods ments at currents of the
order of 20 A, r./rc is of the
Fig. (64.01). Material transfer as a functlon of the
Iife time of the arc. A principle presentatlon order of 10, as is also indi-
cated in Fig. (64.01).
Only some few measurements of 'Yc and 'Ya exist for most metals.
They scatter widely, and the values cited in Table (X,3) are intended
for rough computations. The 'Yc·values of a few special metals which
are recorded in Fig. (56.09) are of a higher quality.
C. The eritieal length, scr, of the anode dominated are. No re-
gular arc can burn with a length that is smaller than the thickness of
the cathode layer. Considering Table (IX,5) we assume the shortest
possible arc length to be about
A = 5·10-8 m (64.02)
A gap of at least this order will always be generated at the explosion
of the contact bridge. Hence, at the time tm of the maximum cathode
gain 3 the arc length is, provided v is constant,
(64.03)
1 The point P usually belongs to cathode gain but to cathode loss in platinum.
2 In R. HoLM [37] r. and y" are termed y" and y, reapectively.
3 At about t = 4 (J.Bec on Fig. (64.01).
§ 64. Details about the types of arc in relay contacts . . . 337
Measurements of v, during the life time of short arcs, have been

published only by RmnLESTONE [1], namely v:::. 2 cmfsec for her
observations with palladium electrodes and v :::. 1 cmfsec for silver
electrodes (in both cases crossed rods). For the range 5 <I< 10 A,
her measurements between palladium electrodes give
s.,:::. (5 + 2 • 4) to-s.= 13 ·to-s m (63.04}

On the basis of Fig. (8) in GERMER and BOYLE [6], one computes
for palladium electrodes
80 r = 34 • 10-s m (64.05)
It seems likely that the difference between (64.04) and (64.05) large-
ly derives from inaccuracy in the v-values.
Note that s.,, the criticallength for the anode dominated arc, marks
the limit between periods II and III and not the limit between periods
III and IV which is the limit between "anodic" and "cathodic" arcs
according to GERMER's terminology. The limit between periods III
and IV has been investigated by HAASE et al. [1], and (at I= 20,~o
50 A) was found to be 4 · 10- 6 m at Ag and Cu, and 2 • 10-6 m at Pt
and W, both values with a great spread. It is surprising that the greater
value belongs to low melting metals which have particularly small
cathode spots.
In §IX the thickness of the cathode layer (at moderate currents)
is computed to be about 5 · 10- 8 m which is the mean free path of the
atoms in this layer. The primary electrons have Ionger mean free paths,
about 5.7 times longer. Comparing these distances with those of Eq.
(64.04) we conclude that, during the life of the anode dominated arc,
a fraction of the primary electrons will be able to reach the anode
with velocities they have gained in the cathode fall. These electrons
contribute to the heating of the anode spot, but their contributi0 n does
not account for the entire materialloss by means of evaporation during
the short arc; cf. Eq. (IX,U).
We are now prepared to give an answer to a question concerning
the amount of the ratio s.,fa. where a. is the radins of the cathode
spot. Assuming I = 5 to 10 A and a current density of J = 5 · 1-010 A/~
one obtains a0 ~ 7 · 10- 6 m. For scr between palladium electrodes, we
use the average between the values in (64.04)and(64.05),i.e., 24 ·10- 8 m.
With this value one obtains
(64.06)
Note that (64.06) is valid only for currents less than 10 A.

22U Electric Contacts, (th Ed.
§ 65. Bridge transfer and short arc transfer

atcontactseparation1
A.. Introduction. Bridges in opening metal contacts have been
a
observed long time ago. SUNDORPH [J], in 1903, and particularly
ANGELICA SzEKELY [1], in 1924, studied such bridges. The latter wor-
ker stated that the material of the bridge was derived from the anode,
and that the bridge always attained such dimensions that its voltage
finally remained constant, of the order of 0.5 to 1 V. Visible bridges
were easily obtained between iron electrodes. Her publication was little
observed. In 1938, the bridges were rediscovered and studie~ by
BETTERIDGE and LAIRD [1]. They succeeded in photographing stabil-
ized molten bridges between platinum contacts. They showed that
the final bridge voltage, before rupture, was of the order of 0. 7 to 2 V,
depending on the contact metal. They also observed that the hottest
section is not in the midst of the bridge. They assumed the THOMSON
effect tobe responsible for·the displacement. However, R. HoLM [29]
remarked that the THOMSON effect, at that time measured on solid
metals only, ·had the wrong direction for an explanation of the dis-
placement in platinum. In the following, the THOMSON effect can be
accepted as the cause ofthe displacement in platinum because LLEWEL-
LYN JONES has shown that with platinum contacts the effect changes
its sign when the metal melts; sec Fig. (66.02).
The study of bridges in air between oxidizing electrodes (as nicke!
and iron), initiated by SzEKELY [1], was continued by PAETOW [1],
LLEWELLYN JONES et al. [4], [5]. SCHRAGet al. [J] were able to keep
the bridges in air under constant current up to one hour, and observed
that their volume steadily increased. However, it should be noted that
this particular growth was due to the formation of oxide, a secondary
phase of the process. The dimensions of those bridges have little in
common with the volume of the short living metallic bridges of noble
metals. Only bridges of noble metals are treated in the following
sections.
In the case that the experiment involves an arc (periods II, III, IV),
its time can be computed as follows: The regular short arc requires a
voltage of the order of 10 V. Observers arrange their tests so that the
emf of the circuit is too small (emf between 2 and 6 V is recommended)
to maintain an arc. The arc then depends on voltage and energy deliv-
ered by the inductance, L, of the circuit. The arc duration, ta, satisfies
10 1; ta ~ ~L.ij or ta ~ 1~ Lh (65.01)
1 Cf.§ 56.
§ 65. Bridge transfer and short arc traDl!fer at contact separation 339
where Ibis the current when the arc ignites; thus lbf2 is a good approxi-
mation ofthe average arc current (the subindex b means bridge). One
sees that ta is proportional to L for given lb. Therefore, ta can be re-
presented by L which is measured easier than ta; observers usually
plot the transfer as a function of L.
The time of the development of the bridge can not be calculated
in the samesimple manner because thf) bridge voltage varies until the
rupture. Nevertheless, in Figs. (65.03} and (65.04} we Iet L represent
the time to a certain extent, also during period I, being ·weil aware
that this is a very rough representation.
We introduce the concepts of slwrt and long bridges. A bridge is
called long when it attains a length similar to its diameter; it is short
when rupture occurs at a length small compared with the diameter. We
shall see that the length has a significant infl.uence on the bridge trans-
fer, even on its direction.
The long bridge can attain a kind of equilibrium whereby its shape
is dominated by·the surface tension and similar to the liquid equilibrium
figures calculated by GAuss and studied by PLA'I'EAU more than 100
years ago. So shaped bridges have been photographed by DAVIDSON [2].
Their rupture can be caused by boiling in the hottest section, but prob-
ably more often mechanically. Also then, boiling and ignition of an
arc in the hot vapor may occur, but this boiling is an attendant pheno-
menon.
The skort bridge does not attain an equilibrium shape. Its rupture is
not always accompanied by boiling or by an arc, according to obser-
vations by the LLEWELLYN JONES' group and by HENTSCH [2]. He
presents statistics showing a considerable spread of the contact volt-
age at rupture with one maximum at the melting point and another
maximum at the boiling point.
In the following we frequently refer to Fig. (65.04} which will be
di!Jcussed with reference to average, somewhatsimplifiedcircumstances.
B • .A classical example of a long liquid bridge between platinum
members is pictured in Fig. (65.02} due to DIETRICH and RüCHARDT
[1]. The hottest section is displaced toward the cathode. Reversing
the current, reverses the displacement. Accordingly, the transfer
was found to be directed from the cathode; it was observed that
rupture of the bridge resulted in a cathode loss immediately after
switching off the current. It was also shown that with either external
heating of the anode, or with a certain non-symmetry in the apparatus
this electrode was much less cooled than the other. Then the transfer
in the contact was diminished or even reversed as was confirmed by
JUSTI and SCHULTZ [1].
The shape of a molten metal hridge is seldom visible to that extent
t -
distance
d!slonce
OllOde cgi!Joo'e Ollode
Fig. (65.02). Dne to DIETRICH

a.nd RttCBARDT [I). Distribution
of the tcmpcrature in a liquid
platinum bridge
•I
--
d!slollce
(Jil(lt/e cgl/;ode -
tilslonce
OllOde coi!Joo'e
§ 65. Bridgetransfer and short arc transfer at contact separation 341
as on Fig. (65.02). Often only its central portion can be seen and the
luminous bases are hidden behind the material of the members.
C. Methods of determination of the rate of transfer. Two distinct
methods are used:
I. Determination of the volume of pips and mounds formed by the
transfer, either microscopically or by weighing. R. HOLM [15] described
the formation of pips and spires (cf. § 67) by bridge transfer and intro-
duced the term "Feinwanderung" for this phenomenon. The present
chapter summons only the most thorough relevant measurements made
after World War II: GERMER et al., particularly LANDERand GERMER
[2]; E. and R. HOLM [8]; RIDDLESTONE [1]; lTTNER [1]; MERL [1], [2],
[ 3]; see also KEIL and MERL [3].
II. Determination of the material transfer by means of radioactive
tracers. One electrode is previously bombarded by neutrons so as to
produce radioactive atoms in it. Using this electrode as anode (or ca-
thode respective!y) allows the amount of metal migration to the other
electrode tobe measured by means of a GEIGER-MÜLLER counter. The
net transfer is the difference between the migrations in both directions.
This method was used by LLEWELLYN JoNES et al., see especially
LLEWELLYN JONES [4] p. 142ff., (5]; HOPKINS [3], [4]; DIETRICH et
al. [7), [8]; HENTSCH [1], [2].
The amount of transfer during periods I and II is very small with
current intensities as they appear in relays. In order to obtain reliable
averages, it is necessary to observe the effect of a great number of
operations. The number usually was of the order of 106 with method I,
and 50 to 1000 with method IL
Method II reveals peculiarities of the migratory process that are
not accessable to method I, particularly the fact that the movement of
the molten material is not confined to the net gain of one electrode.
About one order of magnitude more material is usually exchanged by
tränsfer in both directions1 during any one operation. This means that
the visible pip may contain material from both electrodes and the
amount transferred to the pip from one electrode alone cannot be re-
liably measured by the difference between anode and cathode losses.
Only in the case of a sharp transfer peak that grows at its top into a
crater of the counter electrode does the latter electrode alone deliver
all material migrating to the pip.
For the function of relays, the height of a peak is a serious reason
for trouble, whereas the amount transferred in the form of a fiat mound
is relatively harmless. Therefore, a determination of the transferred
volume alone is not sufficient to characterize the phenomenon. Its
1 See particularly HOPKINS [4].

22E
geometry should also be studied in its dependence on material and

circumsta.nces ; see § 67.
tr'J
~ Wper (JJtrolion l I .1
Oielricli el. ol [ 7]
v/ I I I
/ /. .1/ien/i [1~
frt1
/ .
l!J·t.GA
_,.... 3.S3
uv Iv·/
1.$.
",
~ '/
~ Niddlesfone [1}
I
tr1Z
.....
~ :.-:.c
13.2"' ~
1--' -- ~- ~lrins et ol [3}
I
//
_",lo" ··s Iot t(nd f(q'm(al: fHne([1]
1 [ondN.Ho!m[6};londeronrllltrmer{t j
" tr' tJD 1Q1
lnduclonce offlle circuif
Fig. (66.08) Materlai transfer per operatlon in openl1111 allver contact& expressed 88 cathode galn in
cm•, dlvtded by Il (lt gtven by Eq. [66.06]) and plotted agalnst the lnductance L of the circult
D. Results ol measurements and their interpretation. Features of

the transfer during periods I and II are illustra.ted in Figs. (65.03)
to (65.05). In these figures
the ordinates show the vol-
I
cml/A zptr operolion lr5.GA ume transferred per oper-
./ ation, devided by Ig. This
.d Ib is the current in the mo-
11
""""'
r. ~ 9.1 \ / V / ment when boiling begins in
I J\ ~ ~ s.c( / the bridge. It is given by Eq.
(65.06) where I is the current
13 s\ \ I
through the closed (solid) con-
2 \'-, tact, E the emf in the circuit
,_\ Hir/r/l;lone [ 1j -
and ub the voltage at contact
\ \
rupture. We have generally
I s put ub = ubl• the boiling
to·
" \
...f:!:lones ef o/.{S]
r--+ i
Fig. (66.04). Material transfer per oper·
atlon in opening platlnum contact& ex-
1!1
10
.I [8 ,_
· - ~ Eond1.R. Holm
pressed as cathode gain or cathode loss
respectlvely incm•devided by IlUtgiven
by Eq. [65.06]) and plotted against in-
U7 Nz ductance L of tbe circuit. The horizontal
--~-- Une at 10-u devides the diagram into
an upper half for cathode gain and a
20A \ lower part for cathode lose. Solid lines
10 -l~ 11)•1 100 1-LH 10 1
refer to RIDDLBS!rONB, dashed lines to
HOPKINB and pointed llnes to E. and
R. HOLlll
§ 65. Bridge transfer and short arc transfer at contact separation 343
voltage, for the possible error is insignificant for the conclusions.
11 = 1 E- Ub ".,I E- Ubl (65.06)

I E E
The reduction by the factor l'b 2 , used in the figures, is appropriate

in order to render transfer observations reasonably comparable when
made at different currents,
l
especially when arcing occurs.
Figs. (65.03) and (65.04) show X 0 n
+V
a clear consistency between + 0 g
0 0
measurements under different 0

0
circumstances and with diffe- 0

rent methods. This conformity -0.1
would not have been apparent f
without the reduction. It is, •
however, true that considerable t I
differences still exist. It would

be reasonable to apply also in 0
the horizontal direction a kind · I
of reduction, namely witb re- •

-o.s
spect to the influence of Ib,
and not only of L, on the time
scale; cf. Eq. (65.01). This re-
duction would have been some-
what complicated and has not p/11/invm •
been applied, evidently with
little disadvantage for the use
ofthe respective figures. 0 5 10 15 Amp 20
It may be noted that besi- 4-
des the inductance and the Fig. (65.05). Due to E. and R. HOLK [8). Ma.terial
(per operatlon) devided by Il (lt glven by
current, the capacity in the transfer
Eq. [65.06)). In a.ddltlon: mea.surements by LAIWER
circuit (for instance the capa- and GBRißR [2], m~ked +. For symbols circles and
squares, see text p. 344.Piatlnum electrodes
city between the electrodes)
can have an influence on the duration of the current during opening.
The investigators used the following emf in V in their circuits:
DIETRICH, HENTSCH, HOLM, ITTNER , LL. JOilES et al., LANDER et nl., RIDDLBBTONE with
2 6 6.5 3 2 3 6 Ag
6.5 4 6 Pt
LAND ER and· GERMER ·obtained no detectable transfer in a Pt con-

tact when the emf was 1.4 V or less.
The curves of Fig. (65.03) concerning silver are relatively simple.
The material transfer always is in the direction to the cathode and
decreases to the smallest L tested. However, the explanation is diffi-

cult; cf. § 66.
Fig. (65.04) concerning platinum contacts Iooks more complex, but
is more appropriate to the theme because the THOMSON effect is known
(approximately) for molten platinum but not for other metals in the
molten state. The curves of this figure display a surprising variation up
and down, indicating that different effects are involved in the transfer.
We point to a well in the curve of Fig. (65.04) with its bottom at
about 1 fJ.H. Evidently, the slope ofthe well at the right side (represent-
ing the transfer process to the cathode) is due to the anode dominated
arc appearing after rupture. But the slope downward at the left side of
the well (representing the transfer process from the cathode) can not
be caused by a discharge. The discharges appearing under the cir-
cumstances considered ha ve been analyzed and are arcs with the
typical arc voltage of about 10 V (see papers mentioned at the top of
p. 346). Since they are very short arcs, they are anode dominated and
can not explain the transfer from the cathode which is shown by the
slope at the left side of the well.
We accept the idea that the material transfer indicated by the left
side of the well is caused by the THOMSON effect in long liquid bridges.
In § 66, it will be shown that this effect is able to produce transfer of
the right order of magnitude.
Questions now arise as to why there is any transfer at such a small
value of L (L < 0.1) that only a very short liquid layer could be gener-
ated for an extremely short rupture time, and why this transfer is
directed to the cathode. The responsible effect is not localized within
the metal since it appears when the surrounding atmosphere contains
oxygen but does not appear in pure nitrogen, as shown by E. and
R. HoLM [8]. In § 66, it will be evidenced that the agent is probably
the KüHLER effect, as propounded by J USTI and ScHULTZ (1]. The
observations by E. and R. HoLM on platinum both in air and nitrogen
are plotted in Fig. (65.05). Data obtained in air with L = 0.2 and
0.8 flH are marked by open circles and squares respectively. Crosses
refer to data by LAND ER and GERMER [2] in air at L ""' 0.3 flH. Solid
squares represent observations by E. and R.HOLM with ambient nitrogen.
It remains to explain how the KüHLER effect acting in the solid
contact can be able to influence the temperature distribution in the
short liquid bridge. Evidently, the process is as follows: When the liquid
bridge. begins to form, it takes over the temperature distribution from
the solid contact with the hottest section toward the anode, causing
cathode gain. Certainly, if the bridge lives Ionger and grows, a reversal
of the initial displacement of the hottest section develops by virtue of
the THOMSON effect.
§ 65. Bridge transfer and short arc transfer at contact separation 345
We also read from Fig. (65.05) the following: For the case of pla-
tinum in air and L = 0.2 fLH, all plott.ed points indicate cathode gain.
We believe that the respective bridges ruptured when they were so
I
tpcm/sec
Fig. (66.07). Due to HENTSCH [1]. Material transfer per operation as a fnnction of the inductance
L in the circuit in openlng sßver contacts, at different speed. I~ = 3.5 A, emf = 6 V.
short that they still possessed the temperature distribution due to the
KoHLER effect generated by an oxygen film in the -previous solid state
ofthe contact (as pointed out above). At L = 0.8 f.I.H the bridge became
long with dominating migration
from the cathode when Ib sur- JO
I..-
10- llr:rJ
passed 7 A. 25 /
o%
In pure nitrogen at 0.14 fLH )
only transfer from the cathode

was observed evidently because no
I
J
chemisorbed film appeared. The
v_ "?:!%
-
respective data are also plotted
"'
~ ..........
iQ Fig. (65.04). F::::::: -...;::
I"
Finally we emphasize that
every extremum of the transfer ~ ?.*'% 1\
curve (positive at about 0.1 fJ.H,
\ \
negative at 1 fLH, positive again 17.3'1\ \
beyond 10 f.I.H) marks a state
where a different process begins to
\ \
cut down the volume transferred I
\ \
\
by the previous process.

Several Ph. D. theses at the
-J"aos o.t o.2 o.s 1 1 5 1Qp.H20
Intlucfonce
University of Wales discuss the Flg. (65.08). Due to MERL [1]. Material trans-
respective role of the molten fer as a functlon of the inductance of the
circuit.. Au-Ni alloys at 4.3 A and 6 V. Per·
metal bridge and the anode domi- centage nicke! is indlcated
nated arc in producing transfer at contact rupture of the following

metals: Pt, Ir, Pd, W, Ni, Ptlr. We refer to the publications men-
tioned in HoPKINS [4]. HENTSCH [2] also publishes oscillograms of
anode dominated arcs.
Material transfer with alloys. Fig. (65.08) is unlike Fig. (65.04) in
two respects. There is no weil (cf. Fig. [65.09]); perhaps the THOMSON
coefficient in the alloys is zero. And stranger still, the curves for 7.4
and 10.3% Ni do not indicate any anode dominated arc. The reason is
01 S11sec
'10
1fT12
lo/lice .1 i
cm 3
J'0
I ,
Jrregulor k
I q _/
~ I
\
\ I
'0
f
~
Hegu/or 1_.,.... I
~
1 _L .I./
0
'0
I I \~
_,_ AuCo 97/J 9oo"G \
- AuCo 97/J soo-c
emfi GV ,lrJA I
'0
'0
01 0.2 0.5 1 l
l-
Fig. (65.09). Duc to DtlRRW.iCHTER and MERL [I). Materialtransferper operatlon at contact open
ing as a function of the inductance L of the circuit. The uppermost scale refers to the duratlon t.
of the arc
not known. MERL has found another alloy with an extremely flat
characteristic, namely (Au, 3 Co). According to Fig. (65.09), this alloy
produces a transfer not higher than about 2. t0-13 cm 3JA2 per
operation with 4.3 A. This is true in the entire range of L between 0.1
and 10 fl.H when the alloy has been heat treated at 500 °0, resulting
in a regular lattice of Au and Co.
If the Co atoms are in random positions in the Au lattice, the trans-
fer is up to 10 times higher. Simultaneously, thermal and electric con-
ductivities are about 4 times lower than inthe case of a regular lattice.
KEIL, DüRRWÄCH1'ER and MERL1 find an explanation for the low trans-
fer in the high thermal conductivity 2 •
The flat characteristic of (Au, 3 Co) was observed in air. PuLLEN [1]
also reports fairly flat characteristics for Re and W in air although with
1 See KEIL [4] p. 243, and DÜRRWÄCHTER and MERL [1].
2 Cf.§ 66.
§ 66. Theory of the bridge transfer 347
a transfer about 10 times higher than with (Au, 3 Co). The flatness
continues into the region of the cathode dominated arc. But, with a
nitrogen atmosphere, PULLEN finds that the cathode dominated arc
produces the normal transfer, i. e., from the cathode.
Investigations by E. and R. HOLM with alloyed contact materials
with pronounced flat transfer characteristics led to a patent! concern-
ing voltage regulators and the like. One ofits claims aimed at a balance
between anode and cathode Iosses in an arc. It is proposed to use a
noble metal as anode and a cathode formed from about 2% Pb, a few
percent of material selected from the group Cd, Bi, Sb, Mn, Cr and the
balancenoble metal. This material is able to produce a high anode fall.
Transfer of material from the cathode may dominate during an initial
series of switching operations. This results in a deposit on the anode,
which promotes a high anode fall and thus a migration of the deposit
back to the cathode establishing a balance on an average.
§ 66. Theory of the bridge transfer

A. Introduction. The main task of this chapter is to prove that
features of the observations on platinum contacts, as they are shown
in Fig. (65.04), can be essentially explained by the THOMSON and KüH-
LER effects. As stated in § 65, the experimental proof of the theory is
I -10
~~0 ~--~~------~
Fig. (66.01). Determination of the THOl!lSON elfect by LANDBR [1], solid curves, and EOKBRS et al
[1], dashed curve. The THOHSON coefftcient a 1 !s plotted aga!nst KBLVIN temperature
confined to platinum because the THOMSON effect has not been investi-
gated for other metals in the molten state. The PELTIER effect could
also be involved and will therefore be treated for completeness. Owing
1 U. S. A. patent 2,770,700 ofNov.13, 1956.
to the considera ble differences between observations of the cited author::;,

it will be sufficient to calculate with rough approximations.
B. Bridge transfer because of the Thomson effect. Theory. We
recalP that a positive THOMSON effect works as if positive current car-
riers coming from warmer sections of a conductor give off heat to cooler
sections. In other words: The carriers act as having heat capacity. The
given-off heat adds to the R [2
heat of the current, thus increas-
ing the thermal current and the
temperature gradient, where ther-
mal and electric gradient direc-
tions coincide.
The effect can be demon-
strated on a wire that is heated by
the current through it, while the
ends of the rod are kept on the
same low temperature. Without
07 OB V 0.9 this effect, the hottest section
Applied YO!Ioge
would be in the middle of the rod.
Fig. (66.02). Due to LLEWELLYN JONES (4)
Fig. 8. The temperature at two sections of a But with the positive effect, the
short platinum rod, equidistant from the centre hottest section is displaced in the
for different values of voltage across the wires.
The temperature asymmetry reverse~ at the direction of the current fiow, i. e.,
melting point. (Note: The temperature along towards the cathode since the
the ordinates is not the temperature at the
hottest point) cathode side has the greater ther-
mal gradient.
Fig. (65.02) corresponds to the effect of a positive THOMSON coef-
ficient. However, in the solid state, platinum has a negative THOMSON
coefficient, a1 , according to Fig. (66.01). If the actual displacement
of the hottest section is caused by the THOMSON effect, a1 must change
its sign from negative to positive when the meta! is molten. LLEWELLYN
JoNES [4] showed that this change really takes place; see Fig. (66.02).
The plots in the figure are compatible with the assumption that a1 is
generally proportional to T except for the small interval where a
change occurs from negative to positive. According to Eq. (13.10) the
product eA. is also proportional to T. Therefore, we may put
I
for solid platinum an = - r 1 e A.
(66.03)
for molten platinum a 12 = r2e Ä
where -r1 and -r2 are constants.

For the following calculations we imagine the bridge in a separat-
ing contact to be a cylinder with radins a. Let the contral cross section
1 Cf.§ 14E.
be A~ and, at a distance x from it, the hottest section be A0 ; see Fig.
(66.04). In this figure, A0 is displaced by virtue of the THOMSON effect.
We denote by 1 and 2 the portions of the
I \
constriction on both sides of A0 • Subindexes I \
I \
1 and 2 refer to portions 1 and 2. Since A0 I \
I I
is displaced towards the cathode according I I
to Fig. (65.02) the electric current flows I
I I
I
'
from 1 to 2 in Fig. (66.04). We seek the vol- I I
ume 1ta2 x that represents the transfer per I
I
''
operation. I
\
I
I
I
'
As in § 16, we introduce a model with con- \
4e \ ~e
stant eo and A.0 , and refer to it by symbols
l
R (e0 , A.o), U (l?o, A.ol etc. The real bod y . has Fig. (66.04). Sketch of a
molten meta! bridge in an
R(e,A.), U (e,A.) etc. In the model the resistance opening contact. For sym·
r (eo, j\0 ) of the sought volume n a2 x is bols see the text
r {eo, Ao) =: ~ =~1 [R1 (eo• Ä.o) - R2 (l?o• Ao)]

(66.05)
= 21 b [Ul {eo, Ao) - U2 (eo, Au)]
It will be seen that [U1 ((!0 ,A.o)- U2 ((10 ,A.o)] can be approximately
expressed by the actual voltages U1 (e,;!) and U2 (e,;!). To accomplish
this we use Eq. (66.06) which is another expression for Eq. (14.10),
and Iabel T 0 = ambient temperature, T 1 = the melting point, U1 =
melting voltage and Te = the highest temperature in the bridge. In-
serting in (14.09) gives
~ ~~~
! Ui(e,.?.)-
T
{e.t.dT = { dfP { atdT (66.06)

~ ö ~
and a similar equation for portion (2). Using (66.03) Ieads to
T T
Jatd T = J're.?.d T
To To
Since the term with • is small (see Eq. [66.09]), the last integral may
be replaced by expressions of the type ! '1:(/) 2 according to Eq. (13.06).
Eq. (66.06) then takes the form
~ ~ ~~~ l
fe.t.dT = ; Ui(Q,A) + ~1 ~fP2dqJ- ;a f f!J2df!J~
'l'o 0 Uz (66.07)
= ~ Ui(e, .?.) - ~1 u~ (e, Ä.) + Tl ! m T2

Based on Fig. (66.01) with eA. according to Table (X,2), one calculates
for platinum, T1 = 0.36. It is consistent with Fig. (66.02) to put T2
= T1 • If boiling occurs in the hottest section,
U1 ( A.)- Uu = 1.5 and Uz~ u, = 0.7

e. - 2 2 2 2 (66.08)
These values are good enough to be inserted in the correction terms

and Eq. (66.07) transforms to
Te
j eA.dT =! u~(e,A.)- T~ o.8 UHe.A.)
To
and to a similar equation for portion (2)

T9
je;. d T = ! UHe. A.) + T~ o.8 UHe. A.) (66.09)

To
Hence
~(e,Ä)- u~(e.A.) = o.5a-r2 uHe.A.) = o.53T2 (~'r (66.10)
However,
U~((!,Ä)- UHe,A.) = (U1- U2HU1 + U2) = [Ul(e,Ä)- Uz(e,A.)] Ubz
and since in the case ofboiling the voltage~; in the real platinum body are
higher than in the model by a factor {J = 10
{J [U1 (eo. Äo) - u2 (eo• Äo)] = 0.066. Ta llfz (66.11)
and
(66.12)
Considering that
(66.13)
the radius, a, of the bridge can be eliminated from (66.05), and one
finally obtains
= 0.325 ß T, ;o
s
325 Tz eo
8J3 SJB
:7t a2 X b = b (66.14)
ubl Uf,
Notice that ~q. (66.14) refers to an equilibrium state of a long liquid
':Jridge. The time to attain this state is not represented in (66.14) and
has no special influence on a 2 x.
It will be understood from the deduction that Eq. (66.14) is valid
for different metals except for a minor change of the factor 325. In the
e:
case of silver, 1: is smail and is extremely small. This means that with
silver the inßuence ofthe THOMSON effect may be expected tobe negli-
gible, as Fig. (65.03) actually indicates.
Remark concerning the values of a and x. From electrode marks
left after single operations with Pd-electrodes, RIDDLESTONE computes
a::.:: 10- 6 Ib. Assuming the same formula for Pt, we calculate for / 6
= 10A:
a = 10-s ID and X = 3 · 10-7 ID
C. Thomson eftect. Comparison with measurements. In Figs. (65.03)

to (65.05) the ordinates are transferred material devided by I: or,
using the symbols of this chapter, represent values of
(66.15)
where C = 325 for platinum and slightly different for other metals.
We expect that within a range of smail inductances L, the
THOMSON effect is responsable for the drop from higher ordinates into
the ordinate weil on Fig. (65.04). In Fig. (66.16) such values of drops
taken from RIDDLESTONE's curves in
NT"
Fig. (65.04) are marked by point.s • · cm 3/AJ
The circles o represent observations by
E. and R. HoLM on Pt-electrodes in
N 2 ; see Fig, (65.05). There is a clear
tendency of the transfer, reduced by
the factor I!/ , to vary proportional to
.
1U'
Ib as required by (66.15).For/b= 10A , 0
one reads on Fig. (66.16) an average 0
ordinate value of 10- 12 cm3jamp 2 =

10-16 ma;amp2. Formula (66.15) gives
7 . i0-1 3 cm 3Jamp 2 . This is as good an
agreement as can be expected consider- tJ1l
2 S 10 20ASO
ing the spread of the measurements Io -
a.nd the approximations made during Fig. (66.16). Depth of the weil in cur-
ves as drawn in Fig. (65.04) : e e refer
t.he deduction of (66.15). to RIDDLESTONE, measured in air;
We conclude that the THOMSON effect 0 0 r efer to Fig. (65.05) measured in
nitrogen
causesthewellofthecurvesinFig.(65.04).
What is the reason for the transfer to the cathode at L smaller than
L values leading to a weil ? It is reasonable to assume that here no boil-
ing appears, and that the rupture of the bridge is a mechanical process
at the hottest section of the bridge. The liquid bridge here is now very
short and comprises a relatively smail temperature interval. One may
be tempted to assume that the THOMSON effect in the solid metal pre-
dominates and causes the observed cathode gain. But the measure-
ments plotted in Fig. (65.05) show that the cathode gain observed with
ambient air, does not appear with ambient nitrogen. This fact compells
us to seek the reason for the cathode gain in the KüHLER effect which
has the right direction and would not appear in nitrogen. The reason for
the difference between air and nitrogen is that the oxygen of the air
chemisorbes1 whereas nitrogen does not chemisorb on platinum2 • That
the KüHLER effect has the right magnitude for the explanation, is shown
in Section D.
D. Kohler e:lleet. If the KüHLER effect 1 causes material transfer in
repeated opening of contacts, we expect the process to be as follows.
During open contact an oxygen (mono-) film deposits and chemisorbs
on the members. It remains in the closed contact. When during a sub-
sequent opening the current density becomes high in
the last contact area, the KüHLER effect develops
heat in the anode. It is to be considered firstly, that
subsequent closures on exactly the same spot are very
improbable, secondly that a chemisorbed monofilm
of oxygen develops rapidly, cf. Table (22.03).
Calculating the amount of the bridge transfer that
the Kahler effect might produce. With reference to
Fig. (66.17). Illu- Fig. (66.17) we describe the temperature distribution
stration of symbols
close to the film (situated in the contact A 0 ) as follows.
On the cathodic side of the film the temperature may be T. Then the
temperature on the anodic side 3 is T + fJ1. In A 1 that has the distance
x from A 0 , the temperature may again be T. Thus the volume xna2
of the anode is hotter than any portion of the cathode and is likely to
be twice the transferred volume.
We calculate xna2 by stating that the temperature difference {}1
{}, 1
moves the heat current (I U f - eA. u, I)~ 2 I U 1 through the thermal
resistance x j1ea2 A.. Hence
or
We insert
o 2a
u,=-1
2 na
and -1=
e ub,
1 See§ 15.
2 EHRLICH[1].
For the definition of {}1 see § 15. The subindex f refers to the KoBLER e:lfect a.e
3
being caused by a film.
in which reasonable values are
e = 5e 0 = 5.5 ·10- 7 Om; Ub = 0.6 for platinum,
and a is the tunnel resistivity of the film. We obtain
~66.18)
Eo Comparison with measurementso According to Fig. (65.04) the

cathode gain is about 10- 12 J~cm3 per operation. A reasonable value of
n
a is 5 ·10- 13 m 2 • Substitution of the numerical data in (66.18) with I
= 10 A gives
1
2 xna2 = 3 ·10-17 m 8 = 3 ·10-11 cm8 , per operation
whereas according to Fig. (65.03) the cathode gain at 10 A is between
5. 10-11 and 2 o 10- 10 cm 3 , per operation. The agreement between
theory and observation is regarded as satisfactory.
Fo For eompleteness we ealeulate the material transfer as it eould be
produeed by the Peltier e:O'eeto In the boundary between a liquid and
a solid mass, both of the same metal, a PELTIER effect is activei, just
as between two different solid metals. The liquid bridge in an opening
contact has one boundary (i) against solid metal where the current
enters the bridge, and another such boundary (0) where the current
leaves the bridge. If the PELTIER effect I1 = T oE 81 (with E 81 being
the differential SEEBECK effect at the boundary between solid and
liquid metal) is positive, then (0) is heated and (i) is cooled by the
effect. This means that the cathodic side becomes hotter than the an-
odic side. We notice that the effect can develop only when a liquid
bridge e:xists. Using Fig. (66.04) with the current from (1) to (2), x
denotes the displacement of the hottest section.
It is expected that the volume xna2 is transferred and we are going
to -calculate this volume. Considering the average resistivities to be
ßeo in the constriction and 1.5ßeo in the liquid bridge, the electric
resistances can be written
fJ f!o + 1.5 fJ f!o ". and fJ f!o 1.5 fJ f!o (66.19}
4a :na2 "" 4a-~x
respectively. The thermal resistances against Rl2 heat flow are
1 x and 1 x (66.20}
8aÄ + 2na2 Ä 8aÄ- 2na2 Ä
respectively. The thermal resistance against the PELTIER heat flow is,
with sufficient accuracy 4 l. !
t See BARDEEN et al. [1].
23a Electric Contacts, 4th Ed.
The heat current in (1) and (2) is 1:

times the respective electric
resistance plus the heat produced per second by the PELTIER effect.
The supertemperature, e, in the hottest section A 0 is the product of
the heat current and the thermal resistance. Hence
.ij (!·: + ~:ßa~o x) (s!A + 2n: ,t)-

0
2 Els (@ + 300) 4~,t = 0 (66.21)
for (1) and a similar equation for (2) with negative x and E. The sum
(8 + 300) is the absolute temperature in A 0 • Subtracting the equation
of type (66.21), valid for (2), from the corresponding equation, valid
for (1), neglecting x2 and considering that
8 = llßeo
32a2 A
we obtain
I
4.45@x = (8 + 300) 2 ~ Els
.and :finally
(66.22)
G. Remark concerning the infiuence of the resistivity of alloys

on the bridge transfer, which has been reported by KEIL, DüRRWÄCHTER
and MERL, mentioned in § 65.
All three effects (THOMSON, KoHLERand PELTIER), treated in§§ 65
and 66, produce a larger bridge transfer the greater the resistivity f!o·
The infiuence of eo does not refer solely to the quality of the hottest
section of the bridge because the calculations with the qn?-relation in-
volve integration along the bridge.
H. Remark. The calculations concerning THOMSON, PELTIER and
KoHLER effects are performed by means of formulas that presuppose
ihermal equilibrium. Certainly, complete equilibrium does not exist
during contact opening. But, it can be checked on the basis of § 21
that the temperatures in the bridge have reached more than 90% of
equilibrium values in all cases represented by :figures of § 65 at opening
velocities of a few cmfsec, corresponding to the momentary geometry
and heating.
§ 67. Bridge material transfer in the shape of pips and spires

'Ehe formation of pip and crater by bridge transfer is illustrated in
Fig. (56.01) where it is also pointed out that this kind of transfer can
lead to interlocking in the contact. This :figure shows a fairly course
pip. Finer pips are seen in Fig. (67.01). Still thiriner pips, called spires,
§ 67. Bridge material transfer in the shape of pips and spires 3J55
can be detrimentalin modern servo mechanism, worlring with milliampere

as was shown by DALLAS and STUELPNAGEL [1]. They write : "The first
dilemma confronting the designer of sensitive contact systems occurs
when he discovers that the elimination of arcs can result in a critical
bridging problem. The second dilemma occurs when he starts to reduce
bridge transfer by reducing current". Reducing the current diminishes
the volumes of the spires but not their height; even the very thin
spires are able to make sensitive relays unreliable.
Fig. (67.01). Due to F UJIMOTO [1]. Material transfer in form of

pips or spires on the cathode (osmium-rhodium c.o ntact.s)
The process of pip and spire formation is evidently as follows. At the

first contact opening the bridge transfer produces a small protuberance,
usually on the cathode, and a corresponding crater in the counter elec-
trode. At the consecutive closure this protuberance is likely to make
contact, within a small region, at the bottom of the crater, provided
that no arc has burnt it away. A new opening extends the protuberance,
and the process is repeated with the result that the pip grows Ionger and
longer. The authors mentioned found that the current intensity essen-
tially influences the diameter of the pip (making it fairly proportional
to the current) but affects the final height of the pip only slightly.
Evidently the condition for the pip-shape is that, on making contact,
the top of the pip always touches the bottom of the crater and that the
top is the last point released on breaking the contact. Consequently,
the formation of the pip can be avoided by arranging so that t he elec-
trodes steadily (or even occasionally) make contact in new spots.
DALLAS et al. [1] have avoided spires by using a rotary cathode.
In case of currents > ! 111 (thus not in servo mechanism relays) a means1
to avoid the pips is to Iet an arc burn away the protuberance left by
1 A pip formed during period I is not always cutdown during period II. ln-
8tead ot this, it happens that during period II a moat is digged around the pip.
the bridge transfer. Another method is to employ metals or alloys

which oxidize in the hot stumps of the ruptured bridge. This makes
the stumps incapable of carrying current at the next touch, thus
forcing the current to prefer other spots for the fl.ow. When the pip
formation is inhibited by such means the material transfer results in
broad domes, which may be harmless for the service.
RAUB [1], [2] has observed that base metal electrodes plated with a
noble metal have a certain ability to limit the height of pips in such a
manner that they penetrate the noble metal but do not continue into
the base metal.
Earlier when one was looking formeans to avoid pips in telephone
communication relays working with currents of the order of 1 amp,
the method using oxidizing metals was frequently employed.
Another idea is to employ bimetal contacts. One would expect an
advantage (in a certain range of current and inductance) by using a
metal as anode, which in a symmetric contact transfers from the cath-
ode, and vice versa concerning the cathode metal. However, it happens
that during incipient operations one member deposits a layer of its
metal on the other member. The contact then becomes symmetric and
pip transfer develops. Means to avoid the troublesome initiallayer are
not yet thoroughly analyzed. However, PULLEN [1] states that Pt+ vs
Ag- behaves well in a 12.5 A, 6 V, 2.3 fl.H circuit. R. HoLM1 reported
that the tendency to formation of pips in air is small or nil with C vs
C, WC vs WC and Ag- vs Ni+ as contacts.
§ 68. Mercury switches
A. Introduction. FARADAY recognized the reliability of the contact

between amalgamated copper wires and mercury. The so-called RUHM-
KORFF's induction coil has often been operated by an interrupter work-
ing with a pin dipping in mercury. The mercury is kept in a cup, and its
surface is protected against the infl.uence of the air by a layer of dilute
alcohol or creosot. However, the protection is incomplete, and a dis-
turbant emulsion of mercury oxide is formed in the liquid layer during
operating. The switching power is small.
The turbine interrupter works better. Here a mercury jet is ejected
from one electrode and directed upon the periphery of a revolving
toothed wheel. The circuit is closed as the jet strikes a tooth and bro-
ken as it passes through a slot between the teeth. In some constructions,
1 R. HOLM [29] p. 292 and [37] p. 376.

§ 68. Mercury switches 357
the jet rotates and the wheel is stationary. The jet moves through an
insulating liquid.
B. Medium duty switches with a pool of mercury forming one elflc-
trode. The mercury is enclosed in an hermetically sealed chamber.
Metal electrodes are sealed in, and the mercury can form a bridge be-
tween them. Rupturing the liquid connection (by inclining the vessel)
breaks the circuit. Because of the vapor pressure of the mercury, such
interrupters cannot work as vacuum switches, and arcs are easily
drawn out. In order to give the arc a high plasma gradient, causing it
to be extinguished at a smalllength, the vessel is filled with hydrogen
or other neutral gas to a pressure of one half up to some few atmospheres.
The arc-extinguishing power of hydrogen is especially appreciated for
d. c. switches. SuiT~ [l] found it possible to interrupt 100 A d. c. by
means of a switch filled with hydrogen, whei'eas with nitrogen or helium
only up to 1.4 A could be interrupted. Though such figures depend very
much on the construction of the switch (for instance, on the quantity
of mercury moved) they illustrate the high interrupting ability of
hydrogen, caused by its large plasma gradient in consequence of the
great heat conductivity. Arcing occurs in mercury switches also on
making contact. The mercury vapor generated at the first touch keeps
the liquid electrodes separated for some little time and causes a kind
of ßoating; cf. § 59. The mercury switch is operated by inclining the
vessel (tilt angles of a few down to one degree are practical); or by a
plunger which displaces the mercury and causes it to flow over to the
other electrode ; or by heat-
ing the filling gas and thus
altering its pressure on the
mercury, etc.
If the mercury is not
quite pure, or the walls of
the · vessel not quite clean,
products are generated
under the influence of the
arc, which form conduct-
ing, disturbing films on the
walls. Therefore, it is impor-
tant to purify the mercury
from alien metals, oxygen, Fig. (68.01). A type of mercury switch
sulphur, carbon, etc.; and
the vessel must initially be carefully degassed. In order to protect
the glass walls against the arc, protecting porcelain trays are some-
times used as shown in Fig. (68.01).
Small currents can be interrupted by means of plunging eJectrodes
of platinum, iron or wolfram, but for currents of some amperes up to

100 A it is better to use mercury for both electrodes.
The mercury switch offers the following advantages:
1. The electrodes experience no wear since the mercury evaporated
by the arc redeposits and returns to the liquid. 2. The conductivity of
the contact does not decrease. 3. The arc smoothes the abruptness of
the current break. 4. The switch can be operated by small forces.
5. Retarded switching is easily obtained.
A drawback is the sensitiviness to agitation; another is the slow
operation.
C. Mercury wetted contacts. Relays for up to 250 VA using
mercury wetted contacts are manufactured by C. R. Clare and Com-
pany, Chicago. The mercury is fed to the platinum contact along
the capillary paths between parallel platinum wires. The contact is
moved magnetically by a permalloy armature. The bridging time is of
the order of milliseconds. Advantages: no wear, no bounce, lifetime of
more than 109 operations. Some other applications of mercury wetted
contacts are described in SPEAR et al. [1].
D. Mercury between the members of a sliding contact. The content
ofthis section is entirely due to KLAUDY [1] to [5]. Hisspecial problern
is to construct sliding contacts for unipolar generators with great
current and high peripheral speeds (50 to 250 mfsec). At such speeds,
carbon brushes have a tendency to chatter; they also exhibit too high
contact resistances, partly because the time of passing any contact
spot is so short that fritting of the alien films forming in air, is more
difficult than under a brush at normal speed, and partly owing to
hydrodynamic air pressure under a brush1 . A smooth movement and
easy fritting is expected from mercury between the sliding members.
Since base metals (Ni and Mo) in air are used, a surface film is una-
voidable. But due to the extended coverage by the mercury, the pro-
bability to touch easily fritted and even tunnel conducting spots is
much greater for mercury than for a carbon brush on discrete eleva-
tions. The probability of good conduction is increased by the fact
that any conducting spot between one electrode and the mercury can
always carry current since this can also occur when the opposite spot
on the other electrode is insulating, because the current has the oppor-
tunity to flow in the mercury to other suitable outlets 2 •
The initial arrangement is illustrated in Fig. (68.02). The mercury
is held. by centrifugal forces within a groove (with the bottom on the
outer side) belonging to the rotor. The friction was found to be pro-
1 The hydrodynamic effect can be diminished by using many relatively small
brushes. Experiences by R. A. MARsHALL [1] seems to verify this conclusion.
2 See the indication of the current path on Fig. (68.03).
§ 69. Application of statistics to contact operations 359
portional to v3 A where A = contact surface and v = peripheral speed.
At the desired high v, the friction became too great for practice. In
order to compensate for the effect of the high v3 , KLAUDY made A small
by applying separated "jet contacts" as illustrated in Fig. (68.03). The
jets are forced through channels within the stator into the contact
gap. With this arrangement, A is diminished but still much greater
than the entire true contact area under a carbon brush.
Fig. (68.02). Mercury contact of the unipolar

generator. The mercury is held in a groove Fig.(68.03). Mercury-jet contact; a and b are
by the centrifugal force current constrictions of a-spots
It is suitable to let necessary frittings develop at low v before the

final run at high speed. The fritted areas on nicke! and molybdenum
then remain conducting when v is increased, although they are tran-
siently exposed to air between encounters with jets. Mter initial frit-
tings, the contact voltage could be kept smaller than 0.1 V. The gap
width was 0.4 mm ; currents up to 20000 A were tested. The papers
cited discuss the efficiency of the arrangements and the pressure needed
to eject the jets.
§ 69. Application of statistics to contact operations

A. ReHability of eontaets. Fallures of contacts are of different kind.
1. At closure the contact resistance surmounts the preset maximum
value; 2. Opening is prevented by interlocking or welding; 3. closure
or opening do not occur at the right phase of the operation, usually
owing to deformation of the members.
As to the time of appearance we distinguish two typical causes:
I. a cause of failure that does not exist at the beginning of usefullife
of the contact but develops with time, becoming crucial after a relia-
bility time t.,.. Examples of this kind are deformation due to material
transfer and formation of polymers or other high resistivity films.
rr. a cause of failure that always is immanent with a certain probability
of appearance; for instance, dust particles.
A histogram of kind I is similar to Fig. (IV,1) where l would re-

present the averagetime tr without failure, whereas individual t,.-values
differ from tr in accordance with a standard deviation a; a can be cal-
culated by means of Eq. (IV,6) on the basis of a series of tests. It can
also be determined from a plot of test results on probability paper
according to Fig. (IV,10).
Failures of type II usually obey the rules of the PorssoN distri-
bution; see § IVC. One chooses the intervallength LI (say a month)
and determines y for this LI. Instead of extending the test over a great
number of Ll's, one tests simultaneously many equal devices during
relatively few intervals LI. The concept of "reliability" is mathematic-
ally expressed by (IV,17) but a more tangible concept is MTBF1 in
accordance with Eq. (IV,16).
Microcontacts for certain servo-mechanisms must perform entirely
without failure for a certain predetermined time t,. A single failure
could entail disastrous consequences. This means that t, is chosen
much smaller than tr, and that failures of type II must not appear at
all. The contact must operate in a dustfree enclosure. If the ambient
atmosphere contains oxygen, only gold or some noble alloys with gold
meet the requirements.
The Iiterature contains many sophisticated rules for calculation
with probabilities; for instance, the determination of a skewness of the
distribution resulting from
'15
personal errors of the observer.
!t . Often a simple record of read-
IV,
!~ .
ings immediately reveals an
"'
m~~iln '{
I " '·
""00'
~
V ,.. '
'""' I>""
external inß.uence, as illustra-
ted by Fig. (69.01). At orders
lO
of readings greater than 11 the
variation around the median
!5° 5 '0 %,.o'eru~reol~ JO JS 110 are reasonably random. This
Fig.(69.01). Due to CAMPBELL [2]. Contact resistance can be easily checked by
measurements at random points on a clean silver plotting the readings with
surface
higher numbers than 11 on
a probability paper; see § IV,B. For orders of reading 1 to 11 on
Fig. (69.01), resistances were observed which clearly are out of sta-
tistical control. They belong to a long run above the median. The reason
for this particular run was that the movement of the movable contact
member was impeded by friction. Frequently, runs of high resistance
as in Fig. (69.01) are caused by a dust particle that enters the contact
and eventually is removed by the mechanical action of the contact.
1 Mean time between fa.ilures.
§ 69. Applica.tion of statistics to contact operations 361
B. Some numerical results of tests. Under the aegis of ASTMt, fai-
lures of relays have been investigated statistically. Fine silver and
coin silver contacts, when tested in air immediately after cleaning in
sequences of 1000 operations, have behaved as well as gold contacts,
with less than 0.1% failure at P = 8 g and emf = 0.07 V. Probably
the failures were caused by dust particles. Oxidation of the coin silver
one hour at 300 oc produced insulating films which required fritting
voltages between 0.5 and 2 V.
Wolfram contacts are poor when operating against a soft member,
for instance, gold. Such contacts required fritting voltages of 2 V and
then exhibited less than 1% failure. For these tests the load played a
secondary role because the conduction was produced by fritting. Oxi-
dizing wolfram one hour at 300 oc had a small influence on the fritting
voltage. Wolfram operating against itself or a hard platinum alloy
behaves better because films are then mechanically damaged.
When the data of the ASTM tests were investigated with respect
to the distribution of the deviations it was found that they were not
controlled by anormal distribution. Evidently, external agents as dust
particles or vibrations of the testing device exerted an influence which
appeared to be different in various laboratories. The previous ex-
perience was confirmed that the disturbance by dust particles is
largely eliminated when the contact load exceedes 20 g. At loads
higher than 20 g, the contact is apparently able to produce conduction
in a spot adjacent to the dust particle; see § 30 G.
C. E:lfect of twin contacts. Imagine a relay contact consisting of a
stationary member and a movable member carried on the end of a
fl.at spring. The spring defines the load P that may be small enough
to make the contact sensitive to single dust particles. Thus, a dust
particle can cause a failure of contact make. A failure is called an open
when it appears electrically as if the contact were not closed. In the
following we shall discuss only dust particles as the cause for opens.
The consideration of other agents would not essentially change the
theorypresented.
Imagine a second relay, similar to the first one, but with the mov-
able member replaced by two contact pieces (twins), each on its own
spring producing the load P /2, the total load for the twins together
being P. Such a pair of contacts is called a twin contact. A dust particle
in the surface of one of the twins may cause an open, but not an open
of the other twin. Evidently the probability of a total failure is smaller
than with the single contact; but the gain is diminished by the fact
that, because of the relatively small load of P/2, each twin is more
1 American Society for Testing Materials.

23 b Electrlo Contaots, 4th Ed.
362 Electrio Phenomena in Switohing Contaote
sensitive to dust than the initial single contact. We shall now give
mathematical expressions to these facts.
Assurne that the dust particle somehow leaves the contact after the
time -r:(P) (on an average) where -r: is written as a function of the Ioad
P. An attempt to determine the function -r:(P) has been made by
ATTALA and Cox (2). The observations showed a skew distribution of
the duration, strongly preferring long durations.
A quantitative treatment of the twin problern is indicated below.
The probality p(P) of an open at contact closure is considered to
be p = 1 (i. e. 100% failure) at P""' 0, and p = 0 for a great P, and
to vary between these values for 0 < P < great. The empirical formula
(69.02) satisfies these conditions
p(P) = exp [- (;J] (69.02)
where P 0 is a constant with dimension offorce. This function resonably

conforms to the following observations: 1. by GRANDSTAF.F [1], re-
presenting fairly normal conditions in telephone stations with P 0 = 0.45 g
and e = 0.6; 2. among ASTM tests, with P 0 = 1.8 g and e = 0.44.
A measure of the gain by using twin contacts is the ratio
p(P) = exp [.! (!._)•- (!.-)"] = p(Pt<21-•-o (69.03)

'P2 ( : ) 2' Po P0
E xample. Assuming P = 20 g, we find under the conditions of

GRANDSTAFFS experiments the ratio 22.8. It seems that the inventors
of the twin contacts had expected the twin contacts to reduce the pro-
bability of opens from p(P) to p 2 (P), making the above ratio to
1/p(P), which is a much greater number than observed.
KEEFER et al. [1] find that twin contacts diminish failures by a
factor of 25 to 80. AwEIDA [1] describes a knife type twin contact. The
knife is common for the contacts with the two outer members that
press with equalload. However, the probability of good contact is less
than according to p 2 ( P) .
§ 70. The choice of contact material; contact shape

for practical applications
This chapter presents a survey1 of many informations given in pre-
vious chapters, thereby adding special items of information.
1 Reoommendable surveys of oontact materials: KEIL [4]; MooRADIAN [1];
SoHREINER [1]; VINEs [1] with many references.

.A. Permanent contacts, a.s plug a.nd socket, cla.mped or bolted con-
tacts. As discussed in § 30 the requirement for a.lasting good conduction
through a. conta.ct is tha.t metallic spots are genera.ted by plastic de-
forma.tion a.t the initial contact make and that subsequently, the con-
ta.cting area.s a.re not exposed to air, for instance, by breathing which
implies a. tra.nsient partial opening of the contact; see § 22 G..Silver is
practically insensitive to short brea.thing, and silver plating of contact
members is an effective mea.ns to provide Iasting contacts1 . Silver
when alloyed with less than 9% copper is so fa.vora.bly hard as to di-
m.inish strong welds, a.nd loses only slightly of its high conductivity.
Alloyed with more than 30% palladium, silver is fa.irly resistant
a.ga.inst a.tta.ck of sulfur. This alloy has a. considerable ha.rdness; see
Table (X,1). About Pd-plating see BEATTIE [1].
Plating with relatively hard silver alloys is to be preferred to plat-
ing with pure silver. Silver deposits, produced electrolytically or by
other means, a.re used on contacts in the radio industry. The thickness
ofthe deposit is less than 10- 3 cm. Unfortunately, ba.se metals are able
to diffuse through these thin films of silver (or gold) and produce a
disturbant tarnish on the outer surface of the noble meta.l. Certainly,
the silver itself can tarnish when exposed to an atmosphere containg
sulfur. The use of members of different metals for permanent contacts
is to be avoided when electrolytic corrosion is to be expected.
B. Microcontacts. The concept of microconta.cts ha.s been defined
in§ 30A and discussed in § 30G, where it also is emphasized that only
gold a.nd noble a.lloys on a. gold basis give satisfa.ctory surety of contact
ma.ke in dust-free air 2 • Dust on the conta.ct must be prevented. Con-
tact opening can be impeded by a.dhesion (cf. §§ 9 and 28) and by inter-
locking between a. spire and a cratP-r, see § 67. There is mentioned, as
a. remedy a.ga.inst these disturbances, the proposa.l by DALLAS a.nd
STUELPNAGEL [1] to use a. rotating conta.ct member, which tearH apart
the a.dhering contact without the a.id of the small Iifting force, and in
addition prevents anyformation of spires. STUELPNAGEL [1] has shown
that contact oscilla.tions (called contact frying) can be produced by
electrosta.tic a.ttraction in sensitive relays working with 100 V. As a
remedy against frying, contact members so shaped that the apparent
contact a.rea is very small, are recommended; then the electrostatic
attraction is correspondingly small.
As rega.rds sensitive relays, it is a difficult problern to prevent
harmful contamination by vapors (e.g., backstreaming from vacuum
pumps) by lubricants and other deposits, left by tools used durng manu-
1 Cf. W. R. WILBON [I] and P. QUINN [I]. About both tin plating and silver
plating on aluminum, see p. 165 and R. HoLM [37] p. 160.
• AA for platinum, see also end of § 7 A.
facturing. Interesting experiences concerning this problern are published

by Ll:SH [1], and NUNN [1], [2]. Cf.§§ 7 and 22, F.
C. Belay contacts with high repetitive operation. In these rela.ys
several operations per second may be performed with currents below
5 A and voltages between 5 and 100 V. Examples: automotive regu-
lators used for generator voltage control, distributors, current regulat-
ing thermostats. Disturbances arise f:rom high resistive contact films,
adherence and material transfer. In the case that arcing can be essen-
tially avoided, for instance, by means of quenching, a wolfram cathode
against platinum iridium, platinum ruthenium or platinum palladium
is often recommended. An advantage gained with these hard materiale
is that the minimum arc current Im is high (Im~ 1 A) and that there-
fore the arc can be avoided at currents higher than with soft metals
where Im is about 0.5 A. For the purpose of fritting the emf should be
>6V.
The combination of wolfram (or molybdenum) against hard nobel
metals is chosen in order to prevent adherence owing partly to the
hardness (see § 28,A), partly to the slight oxidation of the wolfram.
Initial metallic contact, evidently helonging to very small a-spots, is
observed.
Ares produce welding only when they appear immediately hefore
closure so that liquid metal still exists at closure, as often is the case
during bouncing. Generally, whether arcs interfere or not, the
following means against welds are advisable. 1. The use of very hard
contact materials, as for instance platinum-iridium and wolfram,
which form relatively small and brittle welds that may be negligible;
cf. § 28. 2. The use of oxidizahle metals whose welds are brittle be-
cause of the intermixed oxides. Here, a compromise between the
requirement for good conduction and that for non-stielring must be
made. The welds of sintered compounds of silver (or gold) with gra-
phite or CdO are easily broken, because the added constituent pre-
vents the formation of large a-spots. In case of arcing, CdO is de-
composed or sublimated at about 1000 °C without severely impairing
the conduction through the contact. Compounds between silver and
2 to 5% graphite prevent noticeable weldingl.
D. Light duty relays for medium frequencies, as in telephone sta-
tions. As to the reed-relays, now much used in telephone stations, and
their contacts, see § 30F.
Neither weid nor hridge huildup appears between WC electrodes;
WC carries a protective film. The comhined film of both members is
readily fritted by the line voltages (120 to 160 V) used; and if the final
1 See KEILet al. [3) and KEIL [4] p.1541!Jld pp. 234-239.
resistance of 3 Q does not matter, as in a Western Union relay work-

ing with currents of 35, up to 250 mA, the WC contact renders good
service.
WC is representative of materials that have been called hardmetals
because they are very hard and have conductivities in the metallic
range (see Table Ill,9). ANTLER et al. [.5] have investigatedl several
hardmetals. They are chemically stable and do not weid because a
protective film develops in air and stops growing at a thickness cha-
racterized by a fritting voltage of a few volt. If the surfaces are rough,
contact resistances below 1 Q could be obtained at P = 5 to 10 N,
even against gold. Similar properties were observed with tin-nickel
platings obtained by electrodeposition on printed circuit boards.
E. Medium duty circuit breakers and contactors work with currents
of a few 100 A, and under such circumstances that, at break, arcs
appear so strong that they have tobe blown into an arc cutter or cooler
by means of a magnetic blowout. Because of the motion of the arc spot,
the material transfer is not concentrated and becomes of secondary
importance as compared to oxidation and deformation resulting from
melting and welding. The contact members usually are small and
difficulties arise from overload and accidental short-circuit currents.
To meet these difficulties the contact material must have a high conduc-
tivity and endure strong arcs without essentially being deformed. Silver
is too soft and silver alloys will have, for many applications, too poor a
conductivity. Sintered compounds of silver and wolfram or molyb-
denum combine a good conductivity with resistance against deforma-
tion and welding. The conductivity corresponds essentially to the
volume content of silver whose lattice remains normal. It is otherwise
in alloys where atoms intermixed in the parent lattice shorten 2 the
mean free path of the electrons (due to additional disturbances in the
lattice) and increase the resistivity. The wolfram or molybdenum
framework is responsible for the hardness of the compound; it also
previmts the silver from fiowing out and thus prevents extended welds8 •
The wear of the silver containing compounds, under the influence
of arcs, is several times smaller than the wear of silver; see § 56 B.
F. Heavy duty circuit breakers with up to thousands ol amperes and
volts. Here again sintered compounds of a good conductor (Ag or Cu)
in a. framework of a refractory material (W, Moor even WC) are used4 •
Adding to other erosion phenomena, liquid drops of the low-melting
1 In ANTLER et al. [5] earlier Iiterature is cited.

2 See§ li.
8 Experiences on the in1luence of the a.tmosphere on (sintered) contacts a.re
reported by PALME [1].
'More than 60% re:fractory material renders ma.chining difficult.
366 Electric Phenomena. in Switching Contacts
metal may be blown out-sputtering- during tbe intense boiling in

tbe arc spots1• Coberence oftbe refractory diminisbes tbe sputtering. It is
common to use silver compounds in air blast breakers and sintered
copper compounds in oil breakers. Tbe idea is tbat tbe silver will resist
oxidation in tbe air2, whicb tbe copper does not. Unfortunately, less
is gained tban expected since silver togetber witb wolfram or molyb-
denum oxidizes to silver-wolframate or -molybdenate3 , glassy slugs
whicb are able to produce insulation at contact make.
G. Sliding eontaets for resistors and apparatus'. Different hardness
in rider and stationary member is used in order to minimize wear. In
many cases it is particularly important to prevent wear of tbe statio-
nary member, for instance, on film potentiometers wbere a relatively
soft gold wire may slide on tbe barder metal deposited on glass. For
some servo-mecbanisms relatively soft brusbes of metal-graphite type
may be used on silver or silver-alloy rings.
Wire-wound potentiometers usually bave wires of oxidizing mate-
rials, as manganin, constantan or chromium-nickel-alloys, on whicb a
somewbat barder rider slides; consisting of bronze, nicke! or bard plati-
num alloy. The rider must be able to abrade thin oxide films. Oxidation
is kept low by lubrication, while for purposes ofhigb precision the wbole
potentiometer may be kept submerged in oil.
1 W. R. Wn.soN [2].
11Oxidation of silver occurs above 550 °C.
a KEIL [ 4] p. 223.
4 Cf.§ 301.
Appendices
Appendix I. Elasticity, Plasticity and Hardness
A.. Introduction. The concepts of elasticity, plasticity and hardness

are discussed with respect to their application in the theory of contacts,
viz. for the determ.ination of actual contact areas. Simplification is
gained by reference to more complete but easily read treatments by
TABOB. [1], KUBLMANN-WILSDOB.F [1], and BOWDEN and TABOB. [12],
Chapter XVI.
TWo ideally hard bodies cannot touch in more than three points. In
these cont act pom . t s th e pressure = mechanical Ioad would be m
. finite.
area
But in reality, the material yields and thereby defines contact
areas. Additional contact spots may then be generated. The yielding is
elastic if the initial shape of the bodies is restored when the stress is
removed, and plastic if a deformation remains. It is essential to note
that a plastic deformation is always surrounded by an elastic deforma-
tion which delivers the Ioad carrying counter-force.
The mathematics for this phenomenon can be worked out only
approximately. The most elaborate calculations concem ideally elastic
or plastic contact members with simple shapes, particularly in the
cases where one or both members are spherical or both are cylinders
ma~g a cross-rod contact. Also, perfect isotropy is assumed, with
HooxE's law valid for the elasticity, and no strain hardening. We can
then extend the discussion to nominally Hat members whose actual
sUrface roughnass can be represented by spherically curved protu-
berances. The time dependent phenomena will be treated in Section H.
Certainly, no crystal has spherical symmetry. However, contact
materials usually are polycrystalline and behave essentially isotropic-
ally in as much as the dimensions of the crystallites are much smaller
than any curvature radius concemed.
B. Hertz' lormulas lor ideally elastic indentations. Hertz' formulas
(HERTZ [1]] with many additions are cited in Ro.AB.K [1] p. 287 :lf. They
are repeated here to the extent they are used in the present book. Be-
tween two spheres the contact surface is practically Hat and circular
368 Appendices
with the radius a, given by Eq. (1,1). Eq. (1,1) gives the general for-
mula for different members (1) and (2}, with r = radius of curvature
(positive for a convex and negative for a concave surface), p = PoiSSON
ratio between lateral and longitudinal strain under the condition of
longitudinalstress, E = YouNG's modulus of elasticity and P = Ioad.
Fora~ r1 and r 2 (cf. below the reference to STOREY}
a= V 3
-P
4
1 - -"'~+1-
(-
E 1
"'~) ( -
- - 1 ,' -
E2
1
r1 r2
)-1 (1,1)
gtvmg Eq. (!,2} with r1 = r and r2 = oo, E 1 = E 2 = E and p 1 = p 1

= 0.3, i. e., for a ball agairrst a semi-infinite plane body of the same
V-
material with PoiSSON ratio1 of 0.3.
a = 1.1 ~ r (!,2)
Here, in a point of contact at a distance x from the center of the load
1.5Pp
bearing area A 11 , the pressure is
P =na
--
3 a-x2 (I,3)
In the center, the pressure has its maximum

1.5 p
Pmax = na2 (I,4)
With an elastic indentation producing A 11 = na2 , the centers of

the spheres approach each other by
(1,5)
where yj2 is the sum of the indentations and the other half of y is due
to elastic deformation in the bulk of the spheres. When not complete
spheres but only spherical humps produce the indentations, the elastic
deformation in the bulk is smaller than yj2. The distance y is often
labeled "compliance". Eq. (1,5} is also valid for negative r.
For a contact between a sphere and a plate, the plate is required
to be at least 6y thick; otherwise a deviates considerably from (1,1).
The formulas (1,2 to 1,5} are also valid for the contact between two
crossed cylindrical rods with the same r and E, and p = 0.3 for both
members.
The more complicated formulas for the elliptical contact area of
crossed rods of different radii and materials are given in RoARK [1]
p. 289. He also treats the contact between parallel cylindrical rods, a
case which is difficult to realize in practice.
1 Actual values of J.': 0.28 for Fe; 0.36 for Cu and Ag; 0.39 for Pt.
§ I. Elasticity, Plasticity and Hardness 369
According to SHTAERMAN (see C. STOREY [1]), Eq. (1,2) is theoretic-
ally valid up to considerable elastic deformations, even beyond the
actual Iimits for elastic deformation. For instance, when {1,2) gives
a = 0.4r, the complete formula gives a = 0.396 r.
Eq. (1,6) gives the time, ti, of a perfectly elastic impact1 of a sphe-
rical indenter (radius r and mass m) hitting a heavy flat anvil with the
velocity v mfsec:
ti =
m2)0.2 (-1 +-1 )0.4 sec
(vr
2.74 - {1,6)
E1 E 2
where E 1 and E 2 are the YouNo's moduli of elasticity.

C. Plastic delormations. Dislocations. Plastic deformation of crys-
talline solids proceeds by slips (in German Gleitung), and the under-
standing of slips is based on the theory of dislocations (in German Ver-
setzung) which will now be sketched. At least in rough outlines, these
facts are rather commonplace. This includes the fact that any mechan-
ical strength (i. e., the Iimit where the deformation passes from elastic
to plastic) such as tensile strength and hardness is ultimately shearing.
It is equally weil known that slips follow along crystallographic planes
and are essentially directed with a 45° incli-
nation against the direction ofprincipal stress. ~OOOOOQOOO
force o oooöobooo
Reasonable calculations indicate that the X !J
stiffness of an ideal lattice against the start- Fig. (I, 7). Edge dislocation
ing of slip should be hundreds of times
higher than the observed shear strength. Thus, the actual behavior
of slips proves that lattices are not ideal. The essential irregulari-
ties in question are dislocations. The simplest type of a dislocation (an
edge dislocation) is illustrated in Fig. (1,7) which pictures a section
through the boundary between two atomic layers. Within the region
xy, the lower layer has one row of atoms more than the upper one. The
lateral extension of the irregularity is usually so small compared with
the surrounding regular lattice that it is reasonable to speak of a dis-
location line being perpendicular to the plane of the figure.
Evidently, bonds between upper and lower neighboring atoms are
directed to the left on one side and to the right on the other side of the
dislocation center as indicated by arrows. A shear force pressing the
upper layer to the right is aided by the right side bonds. This results
in a smaller stress being required to move such a dislocation than to
shear an ideal crystal. This is true even though the dislocation is some-
what anchored at its ends. The edge dislocation moves perpendicular
to its axis. When it fi.nally reaches the crystal boundary, the effect is
a slide of a cross section of the crystal by the length of one spacing
1 Eq. (1,6) has been deduced by TABOR [1] p. 131 using fornmlas by HERTZ.
For equal spheres the coefficient is "" 3.
370 Appendices
(or one BuRGERS vector). Certainly, a macroscopic slip requires the

increments of many dislocations.
The easy move of dislocations is impeded by their anchoring in
lattice irregularities such as vacancies, interstitial atoms or rather
clusters of them, crystal boundaries, and intersections with crossing dislo-
cation lines. Strain-(work-) hardening results when abnormally numerous
anchoring obstacles are produced, for instance, by subdivision of crys-
tallites.
It is understandable that a material can appear with strengths as
different as strain hardened and annealed copper. It is rather sur-
prising that a common quality can be characterized within small stand-
ard deviations to make reasonable the presentation of normal hardness
values (in Table X,1). A plot of such numbers vs the strength of co-
hesion bonds, as judged from evaporation energies, suggests that the
dislocation-anchoring forces are roughly proportional to the strength
of those bonds.
One of the reasons for the existence of a normality is that disloca-
tions are impeded from gathering in very tight groups by the action
of parallel and equally directed dislocations repelling each other. The
repulsion is connected with the fact that adding two dislocations would
double the BuRGERS vector and increa&e the total dislocation energy.
This is proportional to the square of the BuRGERS vector. Usually the
distortion energy in a dislocation is about 2 eV per length of 1 A. The
normal distance between dislocations is some 1000 A. However, dist-
ances down to 100 A have been recorded and the most regular Ge
crystal produced had a dislocation spacing of about 1 mm.
Dislocation lines are made visible in the microscope by the so-called
decoration method, obtained by alloying the material with a substance
that gathers in dislocations and colors them. In an etched surface,
mouths of dislocation lines appear as small pits; for illustrations see
KuHLMANN [1] Figs. (4) and (14).
Evidently, dislocations will move along (001), (111) etc. crystallo-
graphic surfaces in single and polycrystalline bodies. This means that
the plasticity can be approximately "ideal" only if there are slip sur-
faces in all directions. For example, in face-centered cubic crystals there
are altogether 12 slip systems. In contrast to this, only the (001) plane
type is a slip surface in graphite. For generation of dislocations see
KUHLMANN [1).
D. Mathematics of plastic yielding. It has been emphasized that
plastic yielding is tbe result of shearing. A hydrostatic pressure does
not produce any plastic deformation. For our purposes, these facts are
expressed in v. MisEs' Eq. (I,8) with s~fficient exactness:
(0"1 - 0"2) 2 + (u2 - O"s) 2 + (O"s - 0"1) 2 = 2 Y2 (1,8)
It gives the criterion for plastic deformation by expressing tha t plastic
deformation will occur in an isotropic material when the principal
stresses a 1 , a2 , a3 reach values which satisfy (1,8).
From the principal shear stresses
1 1 1
812 = 2(0'1- 0'2); s2S = 2(0'2- O's); SSl = 2 (O's- 0'1)
it follows that Eq. {1,8) requires a certain amount of total shear stress.
Eq. (1,8) can not be satisfied by a hydrostatic pressure; i. e., by
a 1 = a2 = a 3 which would make the left side equal to zero.
The meaning of Y manifests itself when oonsidering the tensile or
compression test of a cylinder. At the start of yielding, the axial stress
is a 1 whereas a2 = a 3 = 0. From Eq. {1,8) it then follows that a 1 = Y;
consequently, Y is the yield stress, yield point. Y would be a constant
in absence of strain hardening. However, if
the material hardens during the deforma-
tion, Y increases correspondingly. It can
also happen that the material looses its
symmetry. Both events would effect an
equation more complicated than {1,8).
According to Eq. (1,8) also the shear F!g. (1,9). Principal stresse•
dur!ng pure shear!ng
strength would be Y.
To prove this, consider pure shearing, for simplicity without a
hydrostatic pressure1 , as illustrated in Fig. {1,9). Evidently a 1 = - a2
and a3 = 0. lnserting in Eq. (1,8) we find a 1 = I a 2 1 = Y. The shear
force in the diagonal area of the elementary cube of the figure is
V~ (0'1 + Ia 2 1} = ß Y giving the shear strength = a1 = Y.
This does not quite agree with measurements, according to which
the shear strength is 20 to 30% smaller than the tensile strength 2 •
E. Indentation in an isotropic semi-inftnite body, produced by a
spherical indenter (ball). Fig. (1,10) illustrates the indentation after
Fig. (1,10). Ball Indentation with a = rad!us of the mouth area and h = depth
plastic yielding. The heavy line represents the indentation before re-
moval of the ball. The finally remaining deformation of the surface is
1 According to Eq. (I,S) a hydrostatic pressure has no influence on the shear
force. It adds the same quantity to any stress a1 , a 2 and a3 which cancels out in
Eq. (I,S).
2 See Am. Inst. of Physics Handbook (1963) p. 2-62 to 2-68. Cf. § 36C.
24*
372 Appendices
indicated by the dashed linet. The space between these two lines cor-
responds to the elastically deformed portion of the indentation. The
stress is most inhomogeneous at the rim, and plastic deformation in the
contact surface begins close to the rim2 • That is why the mouth ofthe
indentation remains visible and can be measured even though the
Ioad was insufficient to produce plastic deformation overall in the
load bearing area Ab.
Fig. (!,11) sketches the slip lines under the ball3 for the case when
the plastic deformation is either still in progress or just at the end,
::.... ::.... :.... finally reaching the broken line at
..,
Fressure '7 'i".... a d epth of about 1f2a. The pressure
at the ball surface is not uniform,
but distributed according to the
values marked; cf. TABOR [1]
Fig. (I, II ). Slip llneslnan indentatlon, Fig. 27.
and pressures vertical to the wall Imagine that we follow the slip -
of the Indentation
line xy in Fig. (!,11) and note the
changes in the vertical principal pressure p along it. At x this pres-
sure is zero. The horizontal pressures at x , which are responsible for
the slip, are obtained from Eq. (!,8) to be Pt = Y . The horizontal pres-
sures vary slowly along xy and still are of the order of Y at point y.
There the plastic deformation requires p > Pt , namely at least sat-
isfying 2 (p - Y) 2 = 2 Y2 or p = 2 Y.
This roughly simplified analysis suggests why (in the ball inden-
tation test) the average pressure against the indenter turns out to be
larger than the yield point Y. Actually, in the event of a specific depth4
of about 0.05, pisnot only 2 Y but about 3 Y; cf. HILL [1] chapter IX,
also TABOR [1] pp. 37, 55, 73 and 104.
F. The ball and pyramid indentation tests. Hardness. The ball inden-
tation test; see Fig. (!,10). A ball with the radius r is pressed against a
flat sample producing an indentation. To begin with, regard the ball
as infinitely hard. Then r is the radius of curvature of the indentation.
Call a the radins of the circle defining the mouth of the indentation, A
the area of the mouth, and S the area of the curved surface of the inden-
tation; dA and dS are elements of A and S. The normal to dS makes
an angle oe with the direction of the load P. If the pressure force per-
1 About the effect of strain hardening on the indentation, see BoWDEN and
TABOR.(l2) p. 336.
2 The pla.stic deforma.tion ca.n not begin exactly at the rim when still p = 0
according to Eq. (1,3), cf. Fig. (1,11).
3 under the condition tha.t friction between the ball and the base member does
not produce considera.ble horizontal forces.

4 See (1,12).
§I. Elasticity, Plasticity and Hardness 373
pendicular to dS is pdS, its component in the direction of Pis
pdS cosrx = pdA
Integration gives
P = fpdA=pA
where p is a kind of average of the pr.essure against the wall of the
indentation. Hence, since A = na2
- p
P""' n;a2
Let the depth ofthe indentation be h; see Fig. (1,10). We introduce

the specific depth hfr as a dimensionless characteristic of the geometry
of the indentation and Iabel it D. It is readily seen that
D = !!:__ ""' !:.._ (!!.-)2 (1, 12)

r 2 r
showing that D can be calculated without measuring h.
It is apparent that indentations with the same D are geometrically
similar and have the same degree of deformation (as defined by [1,8])
in homologous points 1 . Notice J()
that Eq. (1,8) .does not con- 1I

-- ----- -- ----$;,;,;;----
tain dimensions.
/
I
Actual balls are not per-
fectly hard and therefore r
in Eq. (1,12) should be cor-
rected as shown in E. HoLM
et al. (5] p. 231.
Typical measurements
j~ I/ t
1/
with the average pressure p
plotted against the specific I I I
depth D are shown in Figs. f J s f() - z
(1,13) and (1,14) 3• The defor-
o--
Fig. (1,13). Due to E. HOLK, R. HOLK and SHOBBRT
mation is purely elastic up [5]. Average pressure fJ plotted agaln speclllc depth;
to the yield point p = Y, fJ along the thlck llnes is called coutact hardness. Steel,
straln hardened: I, allghtly annealed; II; not annealed
see Fig. (1,13). The corre-
sponding small variation of D is not marked on this figure. All
points refer to a visible indentation mouth; thus to plastic yielding
at least at the rim of the indentation. The pressure, p, increases stead-
ily with D and does not have any strict maximum; but after D has
reached a value of about 0.03, the further increase in p is small.
1 This is an expression of MEYER's [l] similarity law.
2 Cf. TABOR [l] Fig. 34.
374 Appendices
With respect to the moderate accuracy attainable in calculations
about contacts, p in the range of D > 0.03 may be regarded as de-
fining the hardness of the material. The definition must be completed
by the stipulation that the indentation pressure shalllast about 20 sec
f.-- ....-
JO
--
'10
-
2.--
--
Jtl
!- ~
- 30'~'fi.
...
t
!i1!
·~
I~
- fO
aos O.f ai! o.s f.IJ u S.O·V -z
o--
Fig. (1,14). Same measurements as for Flg. (1 ,13) but plotted in logarithmic co·ordinates
because this time is needed for completion of the (initially relatively

rapid) creep. Other definitions of hardness attain a seemingly unequi-
vocal character by referring to the curved indentation S instead of the
mouth :n:a2 ; see below. In order to distinguish the definition used in
this book, we speak of contact hardness or MEYER' s hardness defined by
Eq. (1,15).
H - p
=P= :n;a2 (1,15)
This requires that contact is made without impact, impression time is

20 sec, and D > 0.03. .
The mouth area :n;a2 is practically equivalent to the Ioad bearing
area Ab. Thus, for any shape of Ab, it is reasonable to put
(1,16)
provided it can be assumed that plastic deformation is reached over-
all in Ab.
The Ioad bearing subareas of a nominally flat contact are inden-
tations of different shapes. Types are illustrated in Figs. (36.01) and
(36.02). Some of these subareas may be rather deep indentations with
D > 0.03; others will be shallow with smaller D's (even with no re-
maining indentation). Consequently, p will be smaller than H. This
situation is expressed by
(1,17)
According to the second expression of (1,17), ~ as a function of D is

represented by a curve similar to those of Fig. (1,13) if the ordinate
scale marks ~ = 1 where p ~ H .
As remarked at the end of Section E, the indentation hardness is
greater than the yield point Y, normally H = 3 Y if no strain hardness
is produced during making the indentation; otherwise H is up to 4 Y
or even higher. Recalling that Y = tensile strength S, we have
H = 3 Y to 4 Y with Y =S (1,18)
At the beginning of plastic deformation, (in the ball indentation test):
1 1
p= Y or ~ =-to-
3 4
(1,19)
Remark concerning the duraiion of the load P on an indenter: Imme-

diately after load application, the area of indentation begins to increase
by creep. The time until a practically final area is reached depends on
the metal. The time is short with steel an~ relatively long with soft
contact metals. At room temperature, p may decrease 6 to 14% in
20 to 60 sec corresponding to a decrease of the constriction resistance
of 3 to 7%. At the temperature of liquid air, no creep is observed1 •
During sliding, the formation time of any momentary indentation
is very short and creep can not develop. Therefore p is larger than for
stationary contacts.
In the literature, BRINELL's, VICKERS' or KNOOP's hardness num-
bers are usually given instead of MEYER's. Fortunately, they are suf-
ficiently near (about 7% below) the MEYER's or contact hardness values
tobe used directly in Eq. (1,15) when they are expressed in N/m 2 •
BRINELL [1] and [2] used ball indenters (from ball bearings) and
defined hardness as
(1,20)
where S is the curved surface area of the indentation (not the mouth).
The reference toS is an artifice without a physical meaning. However,
it p:J;ovides a maximum of p, as seen on Fig. (1,13), and it is this maxi-
mum which BRINELL used for his hardness definition. The correspond-
ing D lies between 0.05 and 0.07.
VICKERS' diamond pyramid indenter has the merit of always
producing the same specific depth (D:::: 0.06) practically independent
of the load P. VICKERS calculates with Eq. (1,20) and obtains hardness
numbers nearly equal to HB as a consequence of the choice of the
angle of 136° between opposite planes of the pyramid. KNOOP uses a
pyramid which produces an in dentation that is 7 times Ionger than
wide 2 •
1 See MULHEARN and TABOR [1].
2 For details concerning VICKERS and KNOOP hardness see TABOR [1] p. 164 and
p. 100 respectively.
376 Appendices
One may ask why the ball indentation hardness and not the pyramid
type is chosen as "contact hardness". The reason is that the load bear-
ing areas are generated by surface elevations acting similar to indent-
ing balls but not to pyramids.
Z.O r--------.-------.~------~-----.
108
N/ml
+
1.5
II:>.,
s KJ
o-
Fig. (1,21). Average pressure fJ on artlftcial graphite plotted against speciftc depth D; contact
hardness""' 0.85 ·10• Ntm• at D o:. 0.10. Ball radlus r : o o ,0.079 ; x x , 0.158; + +. 0.316 and
o o, 0.632 cm
The BRINELL and VICKERS hardness numbers are expressed in

kgfmm 2 • The conversion factor reducing these numbers into Nfm 2 • is
107. For example, a meta! with a BRINELL hardness of 80 kgfmm 2 has
a contact hardness of about 8. 108 Nfm 2 •
Any yielding proceeds at a finite speed. Therefore, BRINELL pre-
scribed orie minute for the indenting t esttobe finished. Usually, 20 sec
is employed.
A contact hardness can also be assigned to polycrystalline graphite
in spite of the slip being restricted to the 001-planes. However, slip can
only proceed as long as voids exist into which slipping platelets of gra-
phite can move. When the depth of the indentation reaches about
0.18 mm does p rise rapidly with further increasing P because then
the pertinent por~s are essentially closed and the slip almost stops. This
effect resembles a hardening effect. It has been measured1 on various
electrographite and impregnated carbon grades. An example with arti-
ficial graphite is shown in Fig. (1,21). One has to use small balls in order
to reach D = 0.07 to 0.10 at a reasonable P before the hardening effect
sets in.
1 Measured by E. HoLM.
G. lnßuence of friction on the indentation. The friction against the
indenter obstructs the increase of the indentation area from na2 to S
by acting with a component in the direction of the axis of the inden-
tation. Therefore, the counter pressure is increased from p, that would
appear without friction, to about
(I,22)
with friction. Formula (I,22) is a readily obtained approximation of

formulas deduced by FAIRWEATHER et al. [3]. The e:ffect is small, parti-
cularly with a diamond pyramid because ofits small friction coefficient.
With f < 0.3 and D = 0.06, the measured hardness numbers are 5.2%
higher than they would be without friction.
H. Diffusion etlects. Any dislocation has excess energy and there-
fore does not belong to a thermodynamic equilibrium. It is metastable,
and many of the lattice irregularities that constitute the strain har-
dened state (including anchored dislocations) can be obliterated by
annealing1 •
Any annealing process is atomic di:ffusion. It is usually very slow
at room temperature, but becomes increasingly accelerated after pass-
ing a certain temperature threshold. Every atomic di:ffusion, whether
recovery or recrystallisation, ultimately Ieads to atom positions more
probable than the preceding ones, i. e., to more homogeneaus lattices
with fewer hardening peculiarities. Briefly, atomic di:ffusion Ieads to
softening of the metal.
One can follow the increase of atomic di:ffusion with rising temper-
ature by observing the decrease ofthat portion of the hardness which
is due to strain hardening. The strain hardening may essentially dis-
appear in a few minutes if the temperature surpasses the threshold de-
monstrated for various metallic contacts by the resistance drops in
Figs. (20.01) to (20.03), also (27.14). The temperature range of the
threshold is narrow. That is why one can speak of a certain softening
temperature. Numerical values of softening temperatures for various
metals are given in Table (X,1). Expressed in °K, they are roughly one
third of the melting temperatures.
The relatively slow atomic di:ffusion called creep, is cauliled by steady
yielding to a constant stress. In copper and silver, for example, the
creep eventually ends. In aluminum, and particularly in pure zinc the
yielding to the stress never stops.
Recovery, recrystallization and creep, being results of atomic dif-
fusion, are dependent on both the temperature and the time in a
fashion which is expressed approximately in Eq (I,23). The diffusion
1 Precipitation hardening can not be obliterated simply by heating.
378 Appendices
of atoms proceeds in jumps and any such jump requires: 1. that the
thermal motion of the atom has a suitable phase which will reappear
with a certain frequency v, so that, during the observationtime oft,
this phase appears v t times; 2. that in order to jump, the atom must
be in a state with sufficient energy to overcome a certain activation
energycp, for which the probability1 is exp (- cpfk T), where T = KELVIN
temperature and k = BOLTZMANN constant. Combining both factors, we
obtain for the probability, B, that a certain atom jumps from one posi-
tion to a more probable one in the interval dt:
B = vexp (- cp/k T)dt (I,23)
We now imagine the following simplified picture: 1. The surplus
hardness, which results from strain hardening of a body, is at the time
t = 0 characterized by x 0 atoms being in strained situations. 2. If each
of these atoms makes a single suitable jump, the surplus hardness dis-
appears. 3. In an interval dt, at the timet, with x atomsstill in strained
positions, every atom has the probability B of jumping. This means
that xv exp (- cpfk T) dt jumps will occur during dt. This is the decrease
d x during dt or
- d x = x v exp (- cp / k T) d t
Integration yields
ln Xo = vtexp (- cpfk T)
X
or with cp measured in eV
In Xo = vtexp (- 11600cp/T) {I,24)

X
Actual phenomena will depend on more than one activation energy.

However, in spite of all simplifications including the confinement to
only one cp, Eq. {I,24)·is reasonably satisfied by observations on creep
in contacts, as shown in Table (7.19).
I. The work consumed by a plastic deformation. We consider the
work done by an indenter (ball, pyramid etc:) when producing a deep
indentation, e. g. with D = 0.05. The momentary, curved contact sur-
face is S with elements dS. The pressure on dS is (during the plastic
deformation) of the order of H and is here considered as a constant.
Let any dS move the distance dn perpendicular to dS. Then the total
work W in producing the final indentation volume V= dS dn is J
W "..jHdSdn = HV (I,25)
1 Seeper example ScHAEFER [1] chapter on statistical mechanics; or SLATER [1]
chapter IV.
Tangential forces, i. e., friction forces along the surface of the inden-
tator are not taken into account since they contribute very little to
the deformation considered, cf. section G.
RABINOWICZ 1 has pointed out that a term must be added to W of
Eq. (!,25). This term represents the increase of surface energy when
wear grains are produced. Because it represents a surface effect,- the
part it plays compared to W (a volume effect), becomes greater the
smaller are the wear grains. At very smallloads ( P ~ 10- 5 N) the grains
would be very small, and the additive term can become so dominant
that any formation of grains is prevented, as stated by experiments.
Example 1. To calculate the work done by a perfectly hard ball
indenter, radius r, when producing an indentation with the final mouth
radius a and depth hin a ßat body, with finalload being P, see Fig.
(!,10). During indenting, the load increases from zero to P; the average
value being Pj2. Thus the total work W is PJ2 h or, since P = na 2H and
a 2 "., 2rh
W = nrh2 11 (I,26)
Considering that the volume of the indentation is V= n rh2 we state
that Eq. (I,26) conforms to (!,25).
Example Il. Grooving friction coetficient fur.
Let the indenter of example I be in the position when immerged
to the depth h. Imagine this indenter to be moved tangentially along
the ßat body the distance s producing a groove with the depth h and
the width 2a. The average pressure against the front of the indenter
is aboutH.Consideringthatthe cross section ofthe groove is ~ :3 , we
find the work done according to Eq. (I,25)
u·-
rr -
2 as
3 -.,-s H
This can be expressedas W = furPs with P = na2H giving
Igr = 0.1~
r VH
lfP (I,27)
/gr is the grooving component of the friction, which also contains a

gliding component /gz making the total friction coefficient
(!,28)
Usual1y grooving does not appear in the total load bearing area and
/gl > /gr•
Cf. R. HOLM [22].
1 RABINOWICZ [6], and [7] p. 153 to 163.
380 Appendices
Appendix II. Electronic conduction in solids

A. Introduetion. The description of electronic conduction in solids
bas received intriguingly simple mnemonics in the energy band scheme.
It is the aim of the present chapter to make this scheme easily avail-
able for its application in this book, thereby only a qualitative use of
the basic theory 1 will be pursued. We shall not discuss complications
which arise from the fact that around any atom of a crystal the po-
tential distribution has no spherical symmetry. We also shall not
dwell much on the concept of the so-called effective mass of the con-
ducting electrons. We will simply distinguish between electrons and
holes without deducing the properties of the holes 2 •
B. General concepts concerning eonduetion. Let us review some of
the general concepts of electric conduction. The current carriers can
be electrons or ions. In gas discharges, both kinds are important. Pre-
sently, we are interested in electronic conduction only. The carriers are
so-called free electrons2 • We obtain vivid information about their num-
ber and other characteristics from the band theory. Of course the cur-
rent carriers move kinetically in all directions and the effect of the
applied field strength, X Vjm, is to superpose a driftvelocity,vd, which
is proportional to X,
(II,i)
where {t is the mobility of the carriers, that is proportional to their
mean free path A. In its turn A depends on the regularity of the lattice
and on the temperature. But we better leave these problems until some
properties of semiconductors have been described.
Generally the current density J is
(II,2)
where e is the elementary charge, 1.6 · 10- 19 coul, n is the number of
carriers per unit volume and "is the conductivity. In this book, the
resistivity
1
is usually used.
-
o=-
"
(II,3)
Corresponding units are

for X va e
V m
unit
m 8
Om
1 Recommended survey: KITTEL [1].
2 About the concept of "holes" see Section E.
§ li. Electronic conduction in solids 381
C. Quantization of tbe electrons in a crystal and tbe constitution of

a band. In a crystal the atoms are so close together that the orbits of
the outer electrons (valence electrons) overlap. These electrons become
capable of moving from atom to atom or better, between the atoms.
Therefore they are called free. The remainder of the atoms are posi-
tive ions. Speaking of orbits we refer to BoHR's theory. Their counter-
part in the modern quantum theory are orbitals, defining regions in
the field of the atom where electrons appear with considerable prob-
ability. In a crystal the overlapping orbitals do not belong to single
atoms but more or less to the whole assembly of atoms of the crystal.
For simplicity we consider a body with only one kind of atoms. The
overlapping orbitals are altogether governed by one ScHRÖDINGER
equation with the consequence that any single BoHR orbit energy of
a free atom is split into a "band" of energy Ievels around the BoHR
Ievel. The bands derived from two consecutive BOHR Ievels can be
wide enough to overlap, but in the first instance let us imagine that
there is left a part of the "forbidden" BoHR gap remaining between
them.
Certain properties of the bands are better referred to phase co-ordi-
nates than directly to energies. The "phase space" has 6 co-ordinates:
x, y, z and P.r,Py,P: defining site and momentum components respect-
ively of the particle considered. Actually, it is in these Co-ordinates
that the quantization is carried out. Weshall call an elementary volume
dx, dy, dz, dpx, dpy, dp, a "cell" if its "volume" is h 3 where h is
PLANCK's constant. Concerning the electrons, the quantization together
with PAuLr's principle means that any cell can accomodate not more
than two electrons. These electrons must have opposite spins. If there is
no (or a sufficiently weak) electric field in the crystal considered, we
may choose dx dy dz = the volume unit, 1m3 , and dpx = dp!l = dp:
= hfmeter. Evidently the cell then has the right phase volume. Note:
with m being the mass Of a particle, its kinetic energyis 1/2 m (p! + p; +
+ p~) and many cells with differently directed although equal p-vec-
tors represent the same energy. These cells may form a spherical mono-
layer in the p-space. Call such a layer a "shell" with the radins p. Its
thickness is dp = one hfmeter. Its phase volume is then 4 np 2 hjmeter
and it contains
4np2hTr_!_npzV ll (Il,4)
ha - hz ce s
where V ms is the volume of the body. The ScHRÖDINGER equation
defines as many cells for a band (deriving from one BoHR orbit) as
there are atoms in the body considered, namely about 5. 1028 per m 3 •
Evidently these cells fill a greater number of shells, the smaller p is.
In other words, the bandiswider the greater the radins of its BoHRorbit.
382 Appendices
The valence electrons are the outmost ones in a non-excited atom.

If they belong to n BoHR orbits, and the body has N atoms, the rele-
vant valence band has n N cells. For instance, a silver atom has one
valence electron and thus a piece with N atoms has a valence band
with N cells which can accomodate twice as many electrons. A ger-
manium atom has four valence electrons, all occupied in covalent bonds
in the interior of a crystal. The relevant band has 2 N cells which are
completely filled with electrons.
We return to the problern of electric conductivity. A field generates
a current by accelerating electrons along its direction. The acceleration
means Iifting the electrons to slightly higher energy Ievels. In the com-
pletely :filled Ge-band, all niveaus are occupied; Iifting and accelerating
are not possible. The material will insulate provided the temperature
does not change the situation. This is otherwise in silver, where the
partially :filled band provides abundant opportunities for acceleration.
The valence band of some metals would be filled, but because it over-
laps with the next band, sufficient Ievels exist into which electrons
may be lifted. These examples show the difference between insulator
and metal. At T = 0 °K, (cf. below) the valence band is completely :filled
in an insulator and is surrounded by forbidden gaps. In metals, con-
duction exists even at zero temperature.
D. Inßuence of the temperature on the distribution (or partition) of
electrons on cells in metals. The following discussion aims at the in-
fluence of the temperature on several properties of the current carriers.
A particular result is that the number, n, of carriers per m 3 available
for the current is in semi-conductors strongly dependent on the tem-
perature but constant in metals. At T = 0 °K, the electrons gather on
the lowest energy Ievels (or cells) available.
In the case of metals, this means that the cell shells (cf. [II,4]) are
filled with electrons up to an energy Ievel called the Fermi-level which is
h
'7 =Sm
2 (3n'
n) 2/3
~ 3.65 .1Q-t9 n2/s eV (II,5)
If the electrons have an effective mass m* that is smaller than the nor-
mal mass m, m* must be substituted form in Eq. (II,5). The effective
mass m* is a concept introduced to account for the influence of certain
circumstances that affect the mobility of electrons. These electrons act
as if they had a mass m* smaller than the real mass m, cf. Section E.
Examples of n and '7 values of metals
for Na Cu Ag and Au Al
n 2.5 8.4 5.9 ·1ozs m-a
3.12 7.04 5.51 11.6 eV
§ II. Electronic conduction in solids 383
At finite temperatures, several electrons are lifted to Ievels above fJ,
and the electron distribution on Ievels E e V is changed from 2 per cell to
2
per cell (IL6)
1 + exp 11600(E-7])
T
This expresses the "FERMI DmAc" distribution.

Formula (II,6) is illustrated in Fig. (II,7) by solid curves for T
= 300 and T = 1500 °K. Low energy Ievels (E < f}) have nearly 2 elec-
f(E)
Elccfrvnsper ce/1
ZOt----""""'"--.,....;.---~
lO
s TJ-C eV 7
ferrni /er8/
~'ig. (II,7) is designed for n = 6.5 · 10•• electrons per m• giving the FERm Ievel = 6 eV. From
this Ievel on the work function is counted
trons per cell; but around the FERMI-level the nurober of electrons per
cell decreases with increasing energy. This is shown by the S-shaped
branches of the curves. At T = 0, the distribution is represented by
the solid reetangle with the height 2 ending at E = fJ. This corresponds
to the distribution in a metal at zero temperature.
ÜQly the tails of the curves, far to the right side, play a part for
many calculations1 • There, the exponential function is much greater
than 1 and the nurober 1 can be neglected. Formula (II,6) then assumes
the simple form
2exp[- 11600(E- f})/ TJ (IL8)
For 1500 °K, formula (II,8) is represented by the dashed curve.

This curve begins with the ordinate 2 at E = 'YJ and nearly coincides
(far to the right side) with the tail of the solid curve for the same tem-
perature. It follows that the relatively simple application of the tail
of the dashed curve can be substituted for the more difficult use of
the solid curve referring to the same temperature.
1 For example, the calculation of thermionic emission.

384 Appendices
Let (E - '7) in (II,8) mean not an energy difference, but directly an

energy = Y counted as zero at the FERMI Ievel;
2exp (- 11600 YJT) (II,9)
then represents the "Maxwellian distribution" on cells with v elect.rons
per m 3 where
V=
2 (2nm* k T)l.5
hS = 4.85 ·1021 (m*
m T )1.5 (II,10)
Remark: Assume a semiconductor with the electron density n per m 3

in its conducting band (bottom at Y = 0) where n < v. Then, these
electrons have a "Maxwellian distribution" according to (II,11). At
the upper Iimit, n = v, Eq. (II,11) assumes the form of (IL9).
2!:
V
exp (- 11600 YfT) (11,11)
It is customary to call the distribution according to (II,6) "degene-

rate" as compared with the earlier known type of distribution in (11,11).
If v from Eq. (11,10) with m* fm = 1 is inserted in Eq. (11,5), we obtain
'7 = 3/2 k T which is the average
Be energy per free electron at T °K.
! ;~:·~,.,w~ Conrfuclion bonrf All electrons are then on the zero
1 ..",
1~ ------------ Donorfereis energy Ievel at T = 0, just as
1<:) MAxWELL assumed the atoms to
' . } - - - - - - - ----- Accepfor/mls be. This is characteristic for the
t""l
non-degenerated distribution, and
is true for n ~ v. With n a bove the
Fig. (II,12). Energy bands B 1 (ftlled) and Be Iimit v, thedistribution is degener-
(conducting). Donor and acceptor Ievels in ated. In metals, the distribution
the energy gap between the bands
is always strongly degenerated.
E. Current carriers in semiconductors. The electric current in semi-
conductors is essentially dependent on lattice defects. The band theory
visualizes this situation as it is shown in Fig. (II,12). B1 is the valence
band that is filled with electrons at T = 0; Be, the "conducting" band,
is located L1 eV higher than B1, where L1 is the "gap width"1 • The Be
band is called conducting because electrons which are in this band can
be accelerated and serve as current carriers. There can be conduction
also in B1, but the carriers then are electrons and "holes"; see below.
Within the gap, "donor" and "acceptor" Ievels are marked. They re-
sult from lattice defects: they are near the bottom of Be or the top of
B 1 respectively. Electrons can easily be )ifted (thermally for instance)
1 On p. 381 the gap width LJ eV was considered as the remainder of a forbidden
BoHRga.p. We indica.te another definition referring to an a.tom in a. Ge crysta.l,
which is bonded to 4 neighboring a.toms ~y 0.75 eV totally. The gap then is LJ
"., 0.75 eV.
from donors into Be, or from B 1 into acceptors. In both events, this
means a generation of current carriers by energies smaller than those
needed to lift electrons from B 1 over the large gap j into Be.
Electrons lifted directly from B 1 to Be produce intrinsic conduc-
tion. If the carriers are liberated by means of donors or acceptors, the
conduction is extrinsic. Examples of donors and acceptors will be given
after having introduced the concept of holes. We designate by jd eV
the gap between a donor and Be, and by L1a the gap between Br and
an acceptor, as marked on Fig. (II,12).
The energy Ievel of any band border defines a de BROGLIE wave
length A. of electrons such that the crystal spacing equals nA. where n
is an integer. Therefore, a kind of reflection takes place that prevents
electrons in the band from leaving this band across the border. Also
close to the border within the band strange effects appear. Firstly.
electrons may be more easily accelerated than according to their nor-
mal mass m; namely, as if they had an "effective mass" m* that is
smaller than their normal mass. Another effect results in the generation
of current carriers called holes. These behave like positive electrons.
They appear near the top of B 1 as electrons are lifted from it into accep-
tors or directly into Be. A lifted electron leaves behind an uncompen-
sated vacancy in the electron assembly of B 1 . Evidently, the vacancy
plays the part of a positive elementary charge. The proofthat it be-
haves as a current carrier with the same mass as an electron, must be
left out here.
Conduction by means of electrons (negative) is called n-conduction,
that by means of holes (positive) is called p-conduction.
Example of donors produced by alien atoms is given by As-atoms
in the Ge-lattice. An As atom is capable of replacing a Ge atom in the
Ge-lattice producing 4 covalent bonds. However, since As belongs to
the fifth group of the periodic system of atoms, it possesses one excess
valence electron. This electron remains unoccupied in the covalent
bonding. We may imagine it as rotating in a BoHRorbit around the
un-neutralized positive charge e of the As atom. This orbit is located
within the Ge-lattice. There the relative dielectric constant of s, = 16
diminishes the potential in the orbit by the factor e~ 2 or from 13.6 eV
of the H atom to 13.6/162 = 0.053 eV. Ionization of the donor there-
fore means Iifting the rotating electron from its orbit by 0.053 eV.
This calculation indicates the reason for the gap jd being so small.
Example of acceptors produced by alien atoms. Anatom belonging
to the third group of the periodic table (for instance Al) can replace a
Ge atom in the Ge lattice and serve as an acceptor. In order to satisfy
the covalent bonds, it borrows one electron from the Ge lattice and
becomes a negative ion. It attracts a positive hole which will rotate
386 Appendices
around it in a weakened energy orbit. A small energy quantum of the
order of 0.1 eV is able to move the hole to the top of B 1 . For the hole,
this means being lifted to a higher positive energy Ievel. In other words,
Iifting a hole is equivalent to Iifting an electron from the top of B 1 to
the acceptor Ievel.
Example of acceptors produced by lattice vacancies. In the lattice
of Cu 20 (used in rectifiers), many Cu+ ions are missing, i. e., their place
is vacant. These vacancies play the part of negative centers around
which a hole will be rotating (as around the Al ion). An energy of about
0.13 eV is required to Iift this hole to the top of B 1.
At any temperature, the number n of electrons per m 3 simultane-
ously located in B c is determined by the balance between the frequency
of Iifting events and probability of falling into ionized donors, or into
holes in B 1 . One finds for intrinsic electrons
5800 ) m-s
n =vexp ( -~LI (11,13)
with v according to Eq. (11,10), and for extrinsic electrons
11600 )
-n2- = vexp ( ---Lid (11,14)
N-n T
where N is the number of donors per m 3 , and Lide V is the gap between
them and Be.
Whether the conduction is of n- or p-type is shown by the sign of
the HALL effect. The amount of the effect indicates the number of
carriers. In the event of pure intrinsic conduction, the number of
electrons and holes would be equal. No HALL effect would appear in
the event of equal mobilities. But, the electrons usually have the higher
mobility and therefore dominate and display n-conduction. In some
carbons with intrinsic conduction it seems that many electrons become
trapped and lost for conduction; cf. §V.
F. Mobility of current carriers. Resistance. Theory shows that the
"free electrons" would move without resistance in a perfectly regular
crystal. But no such crystal exists. There are always lattice defects as va-
cancies, alien atoms, dislocations, crystallite boundaries and in particu-
lar, disturbances of the ideallattice by virtue ofthermal movement of
the atoms. All these circumstances produce scattering, combined with
energy loss for the electrons, thus a resistance to the current.
At room temperature and higher, the scattering by phonons, re-
presenting the thermal movement, is dominating by far in most pure
metals. The scattering is proportional to the cross section of the acting
subject. Thus, the scattering ability of atoms oscillating around regular
sites is proportional to the square of their amplitude, i. e., proportional
to T. Consequently, the resistivity (! of these metals is fairly propor-
tional to T since the average velocity of the electrons (proportional to
v;J, see Fig. [II, 7]) is constant and does not interfere. In semiconductors
the average thermal velocity varies with '1' and the mobility therefore
varies more "V~-ith T than in metals.
To understand this behavior of the electrons in metals, we have to
keep in mind that the field accelerates the whole assembly of conduct-
ing electrons and thereby changes the average energy (which is about
0.6 71) very little.
The resistivity of a meta} can be written (! = (!1 T + (!2 where (!1 T
is due to the phonons and p2 to constant lattice defects as alien atoms,
dislocations etc. Notice that the
product of the temperature coeffi. 0~mcm fQ.1(i
1 ae
. d
Cient an (!, name1y e d T • (! = f!t
A
is independent of (!2 • This

"MATTIDESSEN rule" was disco-
vered experimentally.
Clearly, in semiconductors the
,
temperature not only infl.uences t
7r\
the mobility but also the number 1\
of carriers, and the resistivity be- ~
comes a complicated function of
i.'
1\
the temperature. Fig. (Il,15) illus-
trates some typical features of the
~V
f/}17
\
Y:18
resistivity (! of semiconductors.
The figure is reproduced from
E. HoLM [6] and refers to n-con- llo
~
ffiQO K &()()(,
0
ducting grades of silicon carbide. 81l/Jr-'i!tltl

The numbers beside the curve are Fig. (11,15). Due to E. HOLM (6). Resistivity,
numbers n of electrons per cm3 • 11, of a green-yellow SiC crystal plotted against
temperatnre T
The . drop of (! from A to B is
caused by the increasing number of extrinsic electrons that have been
lifted into the conducting band from donors. At B the donors are ex-
hausted. Since they cannot deliver a greater number of electrons, (!
of the electrons now present in the conduction band increases with
rising T along BC due to decreasing mobility. Finally, intrinsic con-
ductivity sets in and determines the branch 0 D.
About the WIEDEMANN-FRANZ law for the conduction in metals, see
§ III that treats thermal conduction.
G. Sudace potential barrier of a metal. Thermionic emission of
electrons. It is evident that a potential barrier exists at the surface of
any conductor. This barrier prevents electrons from escaping the con-
25*
388 Appendices
ductor. The atoms in the surface define the lower part of the barrier in
the same fashion as free atoms define forbidden zones keeping their
electrons bound to them. The higher part of the potential barrier of a
metal is due to the so-called image force. Imagine an electron ernerging
from a plane metallic surface into vacuum and being x meter from the
surface. The electron induces a positive charge in the surface. It is
attracted by this charge in the same manner as the positive image of
E!t:ctron volts (ne_;alive)
the electron in the metal would
attract it. The distance to the
image is 2 x and the attrac-
tion, as is readily calculated,
corresponds to a potential of
3.6jx volt, with x in A units.
This potential means that
3.6 · ejx joule is required
to move an electron from x
to infinity.
Fig. (11,16). Potential barrler at the surface of a In Fig. (11,16), the curve
meta! and its deformatlon by an exterlor fteld.
<1> = work function. The FERMI Ievel 'I is the point
termed "normal barrier"
zero on the energy scale (eV). From there on the illustrates the potential hill as
work functlon counts
it appears at the surface of
the metal above the ~ERMI level. In the figure, it has a total height,
of 4.5 eV which is a typical value for metals.
The height of the barrier above the FERMI Ievel (</>in Fig. [11,7]
and [11,16]) is termed the work function, f/J. Thus f/J = 4.5 eV is a fairly
normal value. Weshall describe the role which the work function plays
for thermionic emission.
Let the border of an emitting body be a plane yz. Those electrons
that hit the plane with the partial energy mv7ef2 > fJ + ifJ are then able
to escape over the barrier, thereby constituting thermionic emission.
The calculation is based on Eq. (11,9). This simplification is feasible,
since the respective electrons belong to the far end of the distribution
tail (see solid curve in Fig. [11,7]). One finds the current density J of
the emission to be (with <P in e V)
J = A p2 e- 11600<1>/T ~ (II,17)
m2
where A would be the same for all metals were it not for certain compli-
cations as reßection ofthe electrons at the boundary surface. Actually,
A lies .between 6.106 and 106 2 Ad for various metals. Eq. (11,17) is
m eg 2
called the RICHARDSON-DUSHMAN equation1 •
1 See for instance S11UTZ [1] p. 165.
On Fig. (ll,16) the dash-dot curve illustrates how the barrier is
bent down by an applied strong field. The number of electrons able to
climb over the lowered barrier is increased by a factor deduced by
SCHOTTKY [1], [3], and is given by Eq. (II,18) with the field strength
being X V f A ;e, is the relative permittivity.
11600 ( if>- V14.3t.J

J=AT2exp [ -----;_r- l A
m2
r;;-:;Y\)] (II,18)
where A is the same constant as in Eq. (II,17) .

Remark. Fig. (26.01) illustrates the potential hill in the narrow gap
(width s) between two plates of the same meta!. The height of the hill
is lower than the work function of the meta! by virtue of the image
effect. Therefore, thermionic emission across the gap is increased. Let
T 1 and T 2 be the temperatures of the plates, if> = the work function
and V = the applied voltage. Then, the thermionic current density is
approximate]yl
J = 11600
A T 2 exp [ - - - ( if> - -V - -10)]
I T1 2 se.
- if> - V
- A T 22 exp [ - -11600( - - 10)]
- A/m 2 (II,19)
T2 2 SE,
H. Equilibrium in a contaet between metals A and B with different

work funetions tP; tP A > tP B· Electrons generally move unrestricted across
Ne!lfllire
Mergy
~..:>...::-- Bol/r;mDc
.__...._,~-+-T--------=:.""---ferrni
2 2 lerel
Eleclrons ~r ce/1
Semiconducfor
!
Bose A !.....--;;---::--'
BoseB 0
Fig. (11,20), Illustration of equili· Fig. (11,21). Electron distribution in the bulk around
brium between two metals with a metal·semlconductor contact. ""' counted to the
different work functions '1>, conta.ct· left and v, counted to the rlght slde are numbers of
ing each other electrons per cell in the phase space on any Ievel E
eV in the meta! and the semlconductor respectively
t Cf. Eq. (26.07).
390 Appendices
a metallic contact. At equilibrium, the FERMI Ievels coincide so that
equal numbers of electrons move in both directions. The difference be-
tween (/) A and (/JB does not inßuence the actual exchange. But, the free
surface of B assumes a positive potential V= (f/JB- f/JA)fe, the so-
called Volta potential above the surface potential of A. Without the
VOLTA potential the equilibrium would be destroyed by the thermionic
emission from B being greater than that from A due to a smaller (/).
The VoLTA potentialisnot directly observable. For its determina-
tion, the metals are arranged as plates of a capacitor with a thin di-
electric between them. Equilibrium is produced by tonehing both
plates simultaneously with a wire. Mter removing the wire, the plates
are separated and their charges generate an observable potential differ-
ence; cf. static electrification in§ 33F.
I. Metal-semiconductor contact. At first, we discuss the equilibrium
conditions for the bulk of the two members; after that, the features of
the boundary barriers.
Equilibrium between the bulk of the members requires that the
distribution per cell is the same in both members on energy Ievels
where electron exchange can proceed. This particularly concerns Ievels
that reach above the barrier. They are represented by the distribution
far to the right side in Fig. (II,7). The metal there has (according to
Eq. [11,9]) the same distribution as a semiconductor with the Be-bot-
tom on the FERMIlevel of the metal, providing the Be-band contains v
electrons as given by Eq. (II,10) 1 . If, however, the Be-band contains
less electrons, equilibrium requires that Be be "lifted" by a negative
potential, V1• This is above the FERMI Ievel of the metal with which
the FERMI level of the semiconductor shall coincide according to defi-
nition. Fig. (11,21) illustrates this situation: the FERMI Ievel of the
semiconductor lies eV1 eV below the bottom of Be, somewhere between
this bottom and the donor Ievels. (/)" is the height of the boundary
barrier in the contact. It is smaller than the work function (/) m of the
metal against vacuum. The condition for equilibrium is that the distri-
butions (shaded) above (/Je are equal.
The region of the boundary barrier. The barrier of the metal is not
annihilated as when · a metal to meta I co·ntact is made, but its
height is decreased to the order of (/Je~ LJj2, where LJ eV is the gap
between B 1 and Be of the semiconductor. (/Je is nearly equal to f/J 111 - C
with (/) m = work function of the metal for emission into vacuum. The
quanti~y Cis the electron affinity ofthe semiconductor. It is the energy
required to remove an electron near the surface of the body, from the
bottom of Be to infinity, viz, Cis about the width ofthe band Be. The
1 In this case, the FERMI-level of the semiconductor is just the bottom of B•.
reduced work function, if>c, is typical for this contact. It is independent
of the potential distribution in the barrier of the semiconductor. The
image force contributes only little to if>c because its value (in vacuum)
is here reduced by the factor 1/s;.
Discussion of the potential distribution in the boundary region of a
semiconductor. We consider the boundary hill in an n-type semiconduc-
tor. The hill is formed between negative .charges in the metaland posi-
tive charges in the hill region of the semiconductor. The latter charges
originate from ionized donors. For simplicity following ScHOTTKY [2], we
assume that every donor in the barrier region is monovalent and
ionized (i. e., has given off one electron). Note that the donors are
immobile and thus are no current carriers.
The simplification is based on the fact that the donors have a great
probability to remain ionized. The barrier field moves the electrons
(appearing in this region) downhill into the bulk of the semiconductor.
Since the delivery of electrons from
the metal and over the potential
hill is small, there exists little op-
portunity for donor ions to become
recombined with electrons.
Fig. (II,22) is a picture of the po-
tential hill between a metal and a
semiconductor at equilibrium, i. e., Fermi Be boHom
without applied voltage. On the t.T'er~el;:--'--!o,----f--,.:----,J!:---·,10::-_,..-m-----:
:r:
side of the metal, the hill is very Xo
steep because sufficient lattice ions Melo/ Semiconrluclor
participate in its generation. This Fig. (11,22). Shape of the boundary hill at
side of the hill is represented by a the contact between a meta! and a semlcon-
ductor; with numerlcal values as glven in
verticalline. In the semiconductor the text
the slope is gentle.
Fig. (II,22) has been designed for the case of pure n-conduction
with the following assumptions: The height of the hill is assumed to be
if>,, = 1 eV; the FERMI levellies 0.1 eV below the bottom of the con-
ducting band; the density ofthe monovalent donors is n = 1022 per ma;
all donors are ionized in the barrier region; the relative dielectric con-
stant is er = 10.
The calculation is based upon the PoiSSON differential relation be-
tween the divergence of the potential gradient and the density (! of
the space charge. In mks units:
d,2 rp (! 1.6 . iQ-19 n
d :~;2 = - 6,6 0 = - 8.85 ·io-12 Er
with s0 = 8.85 · 10- 12 (permittivity of the vacuum). Integration from
392 Appendices
a point (x, q;) to x = x0 with q; = 0.1 and dq;fdx = 0, yields
q; - 0.1 = 9.05 · 10- 9 .!!:. ( x - x 0 ) 2 e V

e,
where 0.1 eV is considered tobe about the difference between FERMI

Ievel and bottom of B •.
Hence for the total barrier q; = ifJ c,
(/Je- 0.1 = 0.9 = 9.05 ·10-9 ~ ~ = 9.05 ·10-l2 X~ (II,23)

e,
or the width x 0 of the barrier (for the figure)
x0 = 3.2 · 10-7 m
Note that q; eV in a point with the distance x from the bound-

ary is added to any otherwise defined energy Ievel in the point
x. This means Iifting the band
Ievels. Both, the bottom of
Be and the top of B 1 are bent
upwards in the boundary region
as shown in Fig. (II,24) which is
designed forthe same conditions
as are assumed for Fig.(II,22). An
Fermi 8 additional assumption is made
/erel here, namely that the gap L1 is
c '-~---
'Oonor
1.2 eV. This means that electrons
Im/
from B 1 must surpass 0.2 eV in
order to enter the metal. They
are close to the FERMI Ievel
where they find free Ievels.
D <P. = 1 eV plays the part of
Melal a work function for electrons
SemiCIJnducfor
going from the metal into E. of
the semiconductor, and f/Jh = 0.2
Fig. (11,24). Potential barrier between a meta! eV plays the same part for elec-
and a semlconductor with numerical values
given in the text trons moving from the filled
band into the metal. It is note-
worthy that the sum (/Je + f/Jh equals the difference between the
top of· the filled band and the bottom of the conducting band, thus
(/Je + <P,. = LJ. The values 1 and 0.2 have no general significance.
They arerather a consequence of the figure being drawn under spe-
cified conditions with the FERMI Ievel fairly close to the bottom
of B e. One frequently meets cases with a FERMI Ievel lower than
§ III. Heat conduction. Wiedemann-Franz law (WF-law) 393
that in Fig. (II,24), and with another ratio between the work func-
tions. How the picture changes when an applied voltage causes a
current to pass the contact is discussed in § 33B.
Table (11,25). Data characterizing Bome BemiconductorB, valid at room temperature
LI-symbols as in Fig. (11,12). A is the mean free path of the electrons
Si Ge lnSb cu.o
LI eV 1.1 0.7 0.27 1.5

Ll,ev 0.05 0.012
Ll.ev 0.07 0.012 0.3
AA 200. 750. 1500. 50.
"! watt/m °K 150. 70. 17.
m*/m 0.25 0.3 0.03
Appendix 111. Heat eonduetion. Wiedemann-Franz law

(WF-law)
A.. Heat eapaeity. Heat is random kinetic energy of elementary par-
ticles: atoms, molecules, electrons, ''holes". According to classical phys-
ics, any degree of freedom has on an average the kinetic energy 1/2 k T
Joule, where k is BoLTZMANN's constant and T the temperaturein °K.
Thus, a particle that can move along 3 co-ordinates, but has no definable
rotational momentum (for example atoms and electrons) has on
an average the kinetic energy 3f2kT. If the particle is bound to an
equilibrium position by a potential, it has also potential energy, namely
1f2 k T per co-ordinate. Thus, in the classic theory the N atoms of 1 m 3
of a piece of metal have the heat content
(III,1)
The·kinetic energy Q, of the free electrons adds to Q4 and then the total
heat content is
(III,2)
The quantum theory makes the functions Q4 and Q, more compli-

cated than according to (III,2). This will be discussed later.
The heat capacity per m 3 is
dQ
C=a,p (III,3)
According to (III,1) and (III,2), it is
C = Ca + C, = 3 N k + :~ (III,4)
394 Appendices
In metals N has a value between 3 · 1028 and 8.5 · 1028 jma. We

shall find that at room temperature dQ.fd T is small and that the heat
capacity per m 3 of a metal is between
1.2 · 106 and 3.5 · 106 Joulejm 3 °K (111,5)
cf. Table (X,2). Note that c (per unit volume) is only little different
for different metals. However, the concept of capacity expressed per
mass unit conceals this similarity. Therefore, this latter unit is in several
respects inconvenient for our calculations.
Free (conducting) electrons contribute to the heat content with
3j2 k T per electron if they obey a Maxwellian cnergy distribution, (cf.
§ II) as they usually do in semiconductors. But, in the "degenerate"
assembly ofthe free electrons in metals (cf. § liD) only a small percent-
age of them, namely those with energies close to the FERMI Ievel, are
affected by the temperature, i. e., consume heat when T increases. This
is the reason why Q., the contribution of the assembly of conducting
electrons to Q, is small at room and medium temperatures. At very
low temperatures, a quantum effect diminishes the contribution of the
atoms so that (compared to it) the influence of the electrons means
more. At very high temperatures, the nurober of electrons above the
FERMI Ievel increases and therefore the electrons assume a higher in-
fluence on c; see for example the data for wolfram in Table (X,2).
The quantum effect on the atoms is as follows. The probability of
a particle energy, q, is proportional to exp (- qj k T). This is small
for low T if q is not also very small. Consequently only the smallest
quanta among those required for
v- -ru- ---
c
calfm oledeg the classic average of 1j2 k T really
s are incited at low T. Therefore, the
/W average particle energy and also ca
3 I decrease with Tat low temperatures.

Ca is a function of Tj8 where e, the
I~ "Debye temperature", is a material
2
I
1.0 constant; for example1 ' e is for
copper 343°; for Wolfram 400°, for
J O.C 08 lO 12
0 graphite about 1500 °K.
Fig. (111,6) 2 shows the heat capa-
T/8-
city, ca, of the atoms plotted against
Fig. (111,6). Thermal capaclty of sollds at the ratio Tj8. The heat capacities
dl1ferent temperature°K plottedagalnst the
ratlo T/9 where 9 is the "DEBYB tempera- for C, W and Cu are marked on the
ture" that is characterlstlc for the meta!
curve. Because of the high 8 the
heat capacity of graphite is low at room temperature and does not
approach the final value until the temperature is of the order of
1 Cf. KITTEL [1] p. 129ff. 2 Scale Jjm3 deg for rough computations.
§ ill. Heat conduction. Wiedemann-Franz law (WF-law) 395
1500 °K; cf. Table (X,2). At room temperature Cu and Ag approach
their final heat capacity fairly close.
Concerning the contribution of c. to c, it is generally small at room
temperature. In semiconductors it is small because of the small number
of free electrons. In metals, only few electrons belong to levels close
to the FERMI Ievel where they can be affected by the temperature :·for
example at room temperature in Cu and Ag, c,jc ~ 0.008, in Fe and
Pt, c.fc ~ 0.05 andin Ni, c.fc ~ 0.08.
B. Thermal conduction. The mathematics for thermal and electric-
al conduction are the same. To the electric concepts of voltage (volt)
and current (ampere) correspond temperature and energy moved per
second (say watt) in the thermal problems. The unit forthermal con-
duction is watttK, for thermal resistance °Kjwatt. Thermal conducti-
vity, ;., is measured in wattjm°K and thermal resistivity in m°Kjwatt.
In order to apply contact formulas to thermal problems, one simply
replaces the electrical resistivity f2 by the thermal resistivity 1 j;..
Thereby Eq. (III,7) becomes evident when R is the electrical and W
the thermal resistance of a given contact geometry with constant f2
and ;. within the body considered,
R
W = eJ. (III,7)
Passing to a property of the thermal conductivity, interesting con-
sequences are met. Regarding phonons and electrons as the carrier of
the thermal flow, some fundamental features of the process can be
approximately expressed by Eq. (III,8) 1 • In this equation, c is the heat
capacity per m 3 of the carriers (phonons or electrons), v is their root-
mean-square velocity and A their mean free path,
Ä = 1/3 (Ce v.A. + Cph VphAph) = Äe + Ä~h (III,8)
Subindices ph and e refer to phonons and electrons respectively.
Table (III,9) gives typical values. The table compares observed .il.-
values with rough calculations. The data for W, WC and TiC are
added in order to indicate how a hard lattice can compete with elec-
tronic conduction with respect tothermal conduction; cf. § 70D.
Table (III,9)
Vph I Ve Cph I Ce Aph I Ae Äph I Äe I Acalc I
Äobser;ved hardness H
rn/sec J/m 8 °K A watt/m °K 108 N/m2
Cu 3200 10• 3.4. 10• .008Cph 60 420 21 380 401 390 5

Si 4500 10• 2 · 10• 10-7 cph 350 200 105 10-3 105 110 450
w 40 140 180 30
wc 24 16 40 \00
TiC 25 11 35 110
1 See for instance, KITTEL [1] p. 138ff.

396 Appendices
How Ae is limited by the scattering of the electrons is discussed in

§ 1I. The mean free path,Aph• ofthe phonons is greatly Iimited by the
mutual impacts between phonons. These impacts are effective only if
the forces between adjacent atoms are inharmonious, so that the pho-
non amplitudes do not simply superimpose. The disharmony decreases
with the amplitude of the
1(JS
deviations from equilibrium;
W/m "K and consequently it decreases
•
10
/
V
" "'~"
;--.,~u
when
creases.
the temperature de-
Therefore Aph and
"'
/
'\. ~ the thermal conductivity, Ä,

increase when T goes to
f".._ I""-. -
.'\. ~ A.[-==
........ ~w -'--- lower values (below room
temperature), until finally
silve~f-
~ the decrease of c causes Ä to
~ I' sink again at the very lowest
70 1 ~Q(\
~tt ? '-"S'ttel temperatures. This behavior
/ l-§of~ltSS
700
/ of Ä is illustrated by some of
/
the curves of Fig. (II1,10) .
'KJ1 10~ 1{ 1)3 Notice that for medium
l!'ig (111•10). Heat conductvity of metals as a func- and high temperatures, A. of
tion of temperature, in oK pure metals remains essen-
tially independent ofT.
C. Wiedemann-Franz' Iaw (WF-Iaw). In its modern formulation
the WF-law states: The ratio Äe/u between the contribution Äe of the
free electrons to the thermal conductivity and the electric conduc-
tivity, u, satisfies Eq. (1Il,11)
Ä..=LTV 2 (1Il,11)
X °K
where the so-called LORENZ function L is
L = 2.45 -1o·s (o~f (111,12)
Because Äe is the dominant part of the thermal conductivity of

metals, it follows that
,1,
-=Äe:::::LT (111,13)
X
with L somewhat smaller than according to (1Il,12). One usually uses

the value
L = 2.4. 10-s (o~r (ill,14)

§ IY. Probability. Noise 397
Eq. (III,13)is the original empirically based expression for the WF-
law, and is the formula that is used in this book for several simplified
calculations. Eq. (III,11) is a consequence of the fact that the mobility
of the same electrons is responsible for Ae and u, as shall now be dis-
cussed in some detail.
In the case of Maxwellian energy distribution among the conduct-
ing electrons, as in a semiconductor, the electric conductivity is given
by Eq. (III,15},
1 Ne 2 A,
U=-=-- (III.15)
• e 2mv
where N = number of electrons per m 3 , m = electron mass, Ae mean
free path and v, as above, the velocity of the electrons. Considering
thatl ~ mv 2k T and Ce = : Nk, we obtain for the ratio be-
= :
tween the thermal and electric conductivities caused by the electrons
--;- k )2 T = 2.23 · 1o-s T

;.. = 3 ( e (III,16)
Degeneration of the electrons by the FERMI distribution changes the

ratio, but not drastically, although now the capacity loses its in-
dependence of T. Thls is because the factorthat increases v, compen-
sates the factor that diminishes Ce. Instead of (III,16) one obtaines
Eq. (III,11).
Because Ae drops out of (III,16) and (III,11}, lattice irregularities
have no inßuence on the WF-law. This is somewhat surprising since
these irregularities diminish the temperature coefficient of (!· However,
this effect on the WF-law is compensated by the appearance of a tem-
perature coefficient of A; cf. the curves for stainless steel and German
silver in Fig. (III,10).
It has been shown that the ratio m* fm between effective and normal
electron mass does not inßuence the FW-law.
According to data gathered by PowELL [1] it even seems that the
WF-law is valid for molten metals, however with
(III,17)
Appendix IV. Probability. Nois~
A. Normal d.istribution. Consider, as an example, the determination

of a length l. Let the single measurements be l + x1 , l + x2 ••• , where
tis the average length measured, close to the reallength l, and Xp x2 ...
398 .Appendices
are deviations (errors). Fig. (IV,1) shows small equal steps around the
endpoint of i. Each step is :figured as a reetangle with the base LI, whose
height measures the nurober r (x) of deviations that end on the step
containing x. Such a :figure is called a histogram.
1----- - - - l
-:c • I .
+.:Z:
Fig. (IV,l). Hlstogram of measurements of a length
Imagine N measurements, and the height of any reetangle to be

represented by r(x)JN. Evidently
+oo r (:t)
~ y=1 (IV,2)
:t--oo
which for N --> oo corresponds to
lim r(x) = p(x)LI (x) (IV,3)

N->ao N
where Ll(x) is the base of the ractangle with the height r(x). If LI (x)
is chosen greater r(x) increases in the same proportion. Thus p(x) is
independent of the length of LI (x). Moreover, since
00
j p(x)dx = 1 (IV,4)
-00
by virtue of (IV,2), p (x) dx may be regarded as the probability of

deviations between x and x + dx. The function p (x) follows the
histogram as shown by the continuous curve on Fig. (IV,1).
It is a well known fact that measurements and many events approach
the normal distribution given by Eq. (IV,5)
'JI2i exp [- _!2 (~)z]

p(x)dx = - 1
a
dx
a
(IV,5)
where a is the standard deviation given by Eq. (IV,6)

1
a2=N-1~x2 (IV,6)
We remind of features of the deduction of Eq. (IV,5). One assumes

that the deviations are caused by a number of agents independent of
each other and occurring at random. To illustrate their actions, imagine
small pins to be thrown on a table and counted as positive when in-
§IV. Probability. Noise 399
clined to the right and the others as negative. The sum of the pin
lengths of one throw, considering the sign, represent one x. Another
throw gives a second x, and so on. With each throw many pins will
cancel in the sum and a very preponderant nurober being positive (or
negative) will be relatively rare. This means that a quantity I x I has
a probability p(x) that issmalleras x becomes greater. The histogram
of the experiment will be similar to Fig. (IV,i), and the continuous
curve, calculated for a great nurober of throws, takes the form of p (x}
in (Eq. IV,5).
The randomness of the deviations is the basis for p (x) of the nor-
mal distribution. This basis is so generally present in events that a
clear skewness of a histogram is regarded as indicating the presence of
a systematic error.
Notice that Cf is the sole parameter in Eq. (IV,5) and that no in-
fl.uence of l on Cf is expressed in this equation. The integral of Eq. (IV,5)
(IV,7)
is called the normal distribution function. The normal densitiy function

from Eq. IV,5) is
-3
(Fig. IV,8). The normal density function
Both functions 9' (:) and <P (:) are illustrated in Figs.(IV ,8) and (IV ,9)
respectively.
B. Probability paper. Checking whether or not a histogram indicates
a normal distribution is relatively easy with the aid of probability
400 Appendices
paper. This paper has a special abscissa scale that we explain referring
to Fig. (IV,10). On this figure the curve of Fig. (IV,9) is reproduced
(dashed) with changed directions of the axes. For the solid line the
abscissa scale has been ex-
panded at its ends so that
the dashed curve becomes
transformedintoa straight
line. The probability paper
has tbe a bscissa scale of the
solid line.
Directions for tke use of
the probahility paper to test
whetber or not a set of
measurements has a nor-
-3 0 z 3 :c.'d mal distribution.
Consider a series of n
Fig. (IV,9). The normal distributlon funetlon
measurements with obser-
vation points l + X;. Every
measured point is plotted
on tbe probability paper
l•2d J, X
with its l + X; as ordinate,
lJ I X
l+ d
4ft 80
for slfT!i§hl /int
S'J~
whereas the abscissa, tP;,
is the fraction (percentage)
of the nurober of observa-
BQ 100% lt
Ooshetf CUI'Yt tions that are smaller than
-
l +X; .
The plotted points cor-
05 11J <1 respond to anormal distri-
bution if, on the proba-
Flg. (IV,10). Demonstration of the co-ordinates
of probabll1ty paper bility paper, they lie on
the straight line that con-
nects the points (ordinate = [, t/J = 0.5) and (ordinate = l + a,
t/J = 0.84) or spread around this line with ordinate deviations smaller
than ± afVn. The spread afVn
represents the uncertainty of when a
n = total nurober of points. See, for instance, LEE [ 1].
The probability paper has no point for t/J = 100%. The last point
to plot has an inherent uncertainty of (1/n)% and may reasonably be
plotted at t/J = 100 - (1/n)%.
C. The Poisson distribution. In the applications here considered, the
problern concerns the frequency of random events (for example, ope-
rational failures among a nurober ofrelays) along a time scale. In order
§IV. Probability. Noise 401
to define the frequency, we imagine the time scale tobe cut into equal
intervals L1, suitably chosen. The events are assumed to appear in-
dependently of each other, but in such a manner that it is reasonable to
count with an average, y, of events per .1. The problern is to predict
how often x events will appear during .1, i. e., to determine the prob-
ability p (x) of x for this L1.
Notice, there is nothing that here corresponds to l of Fig. (IV,1)
although the deduction is similar to the deduction of the normal distri-
bution. The deduction Ieads to
p(x) = yz )
x!e'
hence (IV,11)
p(O) = e-Y
The standard deviation with y as the parameter is
(IV,12)
Notice that y depends on the choice of L1 and is proportional to L1.
N otice also that
a 1
-y=yy (IV,13)
or, in words: the percent amount of the deviations decreases as 1/VY

with increasing y.
x-
Fig. (IV,14). Function p(x) = xf:v for 11 = 3, 6, 15
Fig. (IV,14) illustrates the function p (x) of Eq. (IV,11) for y ,= 3,

6, 15 and also for y = 0.1. In the case of y being a fraction, a Gaussian
11-function must be substituted for the factorial x! Notice that the p (x)
2fl Electric Contacts, 4th Ed.
402 Appendices
curves of Fig. (IV,14) approach the normal density function of Fig.

(IV,S) for the standard deviation, 0' = VY when y ~ 15. A probability
paper can then be used to check the absence ofunilateral inßuences.
It is evident that the PorssoN distribution formula is applicable
only when a risk of failure is steadily present for a long time. Thus, it
is not applicable to the behavior of a graphite brush on a ring because
this contact does not fail until the brush is wom out.
A disturbance that can lead to y "" constant, for a time, is the
presence of dust particles falling on the contact area between closures
and causing transient insu-
lation. Fig. (IV,15) refers to
this kind of interference on
a set of contacts. The verti"
cal lines measure the num-
ber of failures per interval
,1. During the "useful life"
Fig. (IV,16). Vertical llnes glve nurober of contact
faUures per Interval 4. Durlng the useful lüe, faU· with tolerable failurefrequen-
ures happen according to the PorssON dlstrlbution cy, y is essentially constant.
The behavior of a set of con-
tacts during a useful life is often characterized by the mean-time-
between-failures, MTBF. From the definitions of y and Ll, it follows
that
Ll
MTBF=-
y
(IV,16)
This concept has· the merit of being independent of the choice of Ll,
since y is proportional to ,1.
Another often used concept is the reliability R (Ll) for a chosen
interval Ll. It is the fraction of the nurober of intervals without fail-
ure, hence
R(Ll) = p(O) = e-u (IV,17)
D. Electrical noise. The term electrical noise is applied to electrical
variations which produce acoustic or radio disturbance, particularly
when these variations have random character. The connection with
probability is evident. We shall not discuss how the electric energy
is converted to sound, but only the noise as electrically measured. The
variations may be registered as a voltage oscillogram; but more often
the noise is invcstigated by qetectors acting by means of resonance,
and apt to be adjusted to limited "bands of frequencies". They can
be used to scan the spectrum of the noise. Such a. detector contains. a
wave filterthat passes a certain frequency band, and finally an energy
indicator (for example a bolometer) with an amplifier between them.
§IV. Probability. Noiae 403
The filter acts by means ofresonance. Clearly, resonance to a frequency

requires a sequence of waves with the same frequency. Such sequences
might not be directly evident on an oscilloscope, but the FOURRIER
analysis shows that they
exist. The ear and also detec-
tors are analyzersofFoURRIER
frequencies. a
Fig. (IV,18) illustrates a
FoURRIER analysis of a se-
quence of reetangular pulses,
all with the amplitude a (volts)
and the duration T sec. The
(IV,IS). FOURRIBR approximation (9 terms) of re·
interval between the centers Fig.gular polses (amplltude a), pulse time{!, T = time
of two consecutive pulses is between pulse starts
T sec. The time t is counted
from the center of a chosen pulse. The contour f (t) of the pulse-figure
is represented by the FoURRIER expansion1
l
I(
f(t)= ~ a0 +a1 coswt+a2 cos2wt ... a.cos11wt ...
where
2n 2T IV 1
w = -T' ao =Ta 9)
and
sin(vnr/T}
a" = ao r/T
Fig. (IV,18) is drawn for T = 4 't'. The wavy curve approaching the
reetangular contour, is obtained as the sum of the FoURRIER terms up
to v = 9. The approach to
the reetangular pulses im- a7 fa 0
proves (of course) with in- 1.0
creasing v.
The spectrum of the
amplitudes a. is shown in 0.5
Fig. (IV,20). Consider
the spectrum range of
0 ~ nfT ~ 1. There, the 0
highest frequency 111 v gen- Fig. (IV,20). Envelop(spectrom) of the amputodes a. in
erally is 11111 W = 2 nf•, inde- the FOURRIBR series (IV,19)
pendent ofT. Buttbenum-

ber of terms, V"'' increases as T. Simultaneously ,the amplitude aodecreases
as 1fT. Consequently, the energy in this range remains constant if the
1 About FoURRIER expansion see for example PIPES [1] chapter m, section 4;
or V. MANGOLDT-KOPP (1] p. 522.

26°
404 Appendices
product aT: is constant. The spectrum has its first a. = 0 at the fre-
quency
vw = 2nf•
independent of T, even at T ~ oo, i. e., with a single pulse
If the pulses are not reetangular but rounded, a change appears:
the amplitudes become smaller at high frequencies, nfT > 1. This
means that in practice the radio-disturbances also become small.
Surely, the volt variations of noise (in practice) never are of such
a regular type as the sequence of pulses in Fig. (IV,18), and the
.FouRRIER expansion appears otherwise than according to Eqs. (IV,19),
but act11ally only slightly different. In the noise from commutators
the amplitude drops from an a0 at audible frequencies (1000 to 10000
Hertz) to about zero at 105 to 108 Hertz, corresponding to nfT = 1
on Fig. (IV,20), and then rises to finite amounts at higher ranges of
nJT; cf. § 47.
Appendix V.
Structure, electrical and thermal conductivity of carbons1
A. Introdnction. Graphite lattice. Carbon is an important contact
material. It, therefore, is discussed in various parts of this book. For
different purposes very different grades of carbon are used, and the
manufacturers have known for a long time how to prepare various
useful grades. But, the physical explanation of the differences is of
recent date.
Carbon isabasicmaterial in organic chemistry, that owes its intri-
cacy largely to the manifold bonding-possibilities of the carbon atom
which thereby uses its four electrons of the L-shell. In the free carbon
atom two of these electrons belong to 2s-orbitals, the other two are in
p-orbitals. At chemical bonding, the involved orbitals are altered. The
electrons can be "hybridized". For instance, in CH 4 and in diamond
all four L-electrons are hybridized in such a manner that they produce
four equal bonds in a tetragonal arrangement. In the aromatic carbon
rings and in the graphite lattice only three electrons per atom are
hybridized into similar orbitals, namely the so-called u-electrons. These
orbitals belong to three bonds in a plane in directions 120° apart. The
fourth, the so-called n-electron retains a certain liberty. In an aromatic
molecul~ it runs around along lthe ring increasing the negative charges
1Monographies: UBBELOHDE et al. [1], the survey article WALKER [1] and
SHOBEBT [6]. For the mechanical properlies of carbons, the reader is referred to
these books and to HoWE [1]. Those properlies are not treated in the present book,
except in the remark on the hardness of graphitein §I, see Fig. (1,21).
§V. Structure, electrical and thermal conductivity of carbons 405
through which the ions are bound. This increase means strengthening
the bonds. That the n-electrons actually move around the ring is evi-
dent from the increased magnetic moment of such a ring. In the graphite
lattice, the 1t-electrons behave otherwise inasmuch as they are active
also perpendicular to the rings.
The regular lattice of a p erfect graphite crystal is illustrated in
Fig. (V,1). The 001 and 002 planes (basal planes) with their hexagonal
structure are typical for
this lattice. Any 001 plane ~ ~ -- --....;? -
pattern can be comprehend-
OOT-plot7e
•-' I
•
I
ed as the projection of the I
I
pattern of the other 001 I
planes but with respect to -- ....... ~

~ ....... 7
them the 002 planes are dis- ~
~
placed as indicated in the

figure. There are 1.13 . 1029 ....... .......
...2 -- ~ -:....,..-
C-atoms per m in an ideal
3 ;2 """
./'
I
~ ...?'
. I\
graphite crystal and 3.8 -101 9
atoms per m 2 within a basal

Fig. (V, I). Atomic lattice of graphite
plane.
The ideal kind of layer structure does not appear exactly for all
pairs of planes. Many of them will be restrained from ideal posit ions
with respect to each other by lattice defect s and then may be able to
glide easily over each other. A graphite crystal may be regarded as a
pile of thin crystallic platelets1 (thickness 30 to 300 A). The specific
friction force between such platelets can be as low as 0.006 · 108 N jm 2 ,
to judge from BRYANT' s [1] measurements. of the shear strength of
graphite crystals. Wear grains in air seem to consist of such platelets.
Of course, the ease of slipping is basic for the writing ability of the
"lead" pencil and, at least partly, for the lubricating property of the
graphite; cf. § 44.
In the regular hexagonal planes every carbon atom is bound to
three neighbors by covalent bonds of 3.5 eV per atom pair. Thus, any
atom is completely fixed by 3 · 3.5 = 10.5 eV. It is significant that this
is nearly the same energy that binds a diamond atom to its four neigh-
bors (4 · 2.6 = 10.4 eV). In the diamond the single bond is feebler than
in the graphite (2.6 instead of 3.5 eV) since the atom distance is greater
(1.54 instead of 1.42 A) but its tridimensional isotropic structure causes
the high density and unique hardness. The regular bonds from one gra-
phiteplane to the next one have relatively weak forces. However, they
are stronger than VAN DER W AALS forces because n-electrons partici-
1 See TsuzuKo [1], AKAMATO et al. [1], BaLLMANNet al. [1].

406 Appendices
pate in the bonding in a particular manner. These forces 1 correspond

to about 0.25 eV per atom in the perfect lattice and less where lattice
defects cause a greater separation of the planes2 •
B. Carbon grades. In order to understand the differences between
carbon grades, one has to consider the history of their formation 8 : In
the very beginning of their history we meet aromatic hydrocarbons.
Fig. (V,2) indicates how two pairs of hexagonal rings (trunks of naph-
thalene molecules) become bound together. It is assumed that the
b
Flg. (V,2). Two palrs of hexagonal rings become bound together
temperature is high enough to split off two pairs of hydrogen atoms

(Fig. V,2a) which combine forming two H2 thus providing the oppor-
tunity for the freed carbon valencies to directly bind the hexagons to-
gether (Fig. V, 2b). By means of repeated reactions of this kind at
temperatures between 450° and 800 °C !arger planes - they may be
called mono-planes'..,. with diameters of 20 A or more may be composed.
Fig. (V ,2) which is simplified, does not indicate that hydrogen or groups
of hydrocarbons may be linked to the periphery of the planes, an ar-
rangement that frequently occurs.
Electron-diffraction patterns evince the singular fact that the mono-
planes are able to pile in stacks. The reasons for this behavior are not
yet understood; it seems that residuals of hydrocarbon chains provide
criss-cross bonds between peripheral atoms of different planes in the
same stack thus keeping the stack together. ·Such stacks play the part
of crystallites in low carbons, even though there is no crystalline
order between the mono-planes other than their parallelism, the a-
and b-axes of a·parallellayer' group being randomly oriented around
1 BBYA.NT [1].
2 See§ 39B.
3 Cf. MBozowSXI [1], WALKEB [ 1].
'Analogons to mono-layer..
§ V. Structure, electrical and thermal conductivity of carbons 407
the c-axis. Due to this disorder, the stacks are called turbostratic cristals.
In them the average distance1 between adjacent mono-planes is slightly
larger than the spacing of 3.348 A in a perfect crystal.
At about 800 oc a large part of the initial hydrogen has been de-
tached and the substance treated has been transformed into a so-called
calcined or baked carbon, mainly consisting of turbostratic crystallites
with remnants of hydrocarbon chains bonded to the periphery. Evo-
lution of hydrogen or hydrocarbon gas still continues at higher tem-
peratures and apparently only at about 1300 oc is the whole periphery
of the momoplanes bare of hydrocarbon groups.
C. Graphitization 2• At temperatures of the order of 2000 to 3000 oc
the atom diffusion is lively. At least in so-called graphitizing carbons,
the mono-planes attain the ability to grow at the expense of the bond-
ing chains and finally even at the expense of smaller mono-planes in
a neighboring stack. However, according to FRANKLIN [1] and to the
experience of manufacturers, there exist also carbons that are resistant
to graphitization even at temperatures between 2500 and 3000 °C. The
carbon black and lamp black which are condensed from gaseous hydro-
carbons or from evaporized oil respectively remain turbostratic up to
3000 °C. TARPINIAN et al. [1] estimate the activation energy for graphi-
tization to 2 to 2.3 eV.
A physical picture of the constitutional difference between graphi-
tizing and non-graphitizing carbons has not been generally agreed up-
on. We, therefore, mention only two ideas that have been propounded.
RILEY [1] and [2] has propounded the idea that lamp black contains
many tridimensional cross-linked hexagons which mutually occupy
each others peripheral bonds, and therefore are unable to be built into
monoplanes. This idea is not generally agreed upon. FRANKLIN 3 finds
that also non-graphitizing carbons consist mainly of stacks of mono-
planes which she assumes to be randomly orientated. Due to the small
size ofthe particles and their dis-orientation, the crystallites are arrested
in their development even at high heat treatment. Graphitizing carbons
have their stacks initially orientated in such a way that adjacent stacks
have fairly parallel mono-planes and have relatively weak criss-cross
bonds. The non-graphitizing, so-called luJrd carbons have a glassy
structure and do not write on paper.
In order to graphitize carbons, it has been found necessary to start
1 See HousKA and WARREN [1], also FRANKLIN [1]. According to FRANKLIN,
neglect of certain corrections is responsible for higher values of the c-spacing in

turbostratic crystals, which are found in the literature.
2 About raw materials and manufacture of carbons and graphites see WALKER[1]
and SHOBERT [6].

a FRANKLIN [2]; see particularly p. 212.
408 Appendices
with substances which remain fluid to above 400 °0, then solidify by
formation of three dimensional polymer bonding in a narrow temper-
ature range. Petroleum coke, after calcination, shows a flow pattern
apparently caused by the movement of gases liberated during heating 1 •
Platelike aromatic molecules are aligned turbostratically with their
c-axes vertical to these flow Iines. When upon heating above calcination
temperatures the mono-planes reach diameters of 100 A or more 2 , the
total directing force between them, which is proportional to the area,
is capable of aligning many pairs of mono-planes with respect to their
a- and b-axes inspite of obstructing bonds at the periphery. The planes
then attain the mutual position that is characteristic of a perfect graph-
ite lattice with the c-spacing between them diminished to 3.348 A.
With time and with increasing temperature more and moreplanes obey
this right-dress. The growth of crystals and their alignment in all direc-
tions is graphitization 3 • However, alignment does not proceed simul-
taneously for all planes within a stack and therefore the average c-
spacing as measured with X-rays, undergoes a continuous change 4 from
3.44 down to 3.348A.At the sametime the a-spacing increases from less
than 1.40 to about 1.42 A. It is conceivable that this is the result of
the action of the peripheral bonds being able to pull the mono-planes
closer to each other; the smaller the mono-planes, the greater the re-
action. Graphitization at 2500 to 3000 °0 may Iead to crystallite dia-
meters of the order of 1000 A. The material is then called artificial
graphite or electrographite since it usually is heated in electric furnaces.
D. Electrical conductivity5 of carbons. The interest of this book
with respect to carbon is particularly focused on the electrical conduc-
tion, the discussion of which shall be led with reference to the work of
S. MROZOWSKI and his collaborators6 .
The conduction is always performed by 1t-electrons lifted out of, or
holes left behind in the "filled" band of the 7t-electrons. The 7t-elec-
trons have this band on a fairly high Ievel with a small distance to the
conducting band; and the conduction of carbon is exclusively due to
7t-electron current in the conducting band and hole current in the
"filled" 7t-band.
1 WALKER [1].
The average crystai dimension in the c-axis is· about half of this; see G. L.
2
CLARK et al. [1].

3 The definition of graphitization is due to WARREN; see HousKA and WAR-
REN [1]. The firstX-ray observations on the development from amorphaus to gra-
phitic carbon is due to E. HoLM [1].
4 FRANKLlN [1].
For concepts of the conduction process see § II.
5
6 See particularly MRozowsKI [1] and KMETKO [1] who refers to earlier liter-
ature.
Measurements1 on single graphite crystals have shown that at room

temperature the electric conduction is about 250 times higher along the
basal planes (about 2.6 · 106 [!1 m]- 1 ) than perpendicular to them
[about 2 · 104 (Q m)- 1 ]. This difference means that the conducting
electrons meet more and higher barriers when moving in the c-direc-
tion than when moving in a direction perpendicular to it. Along the
basal planes the crystals are all but "ideal" and therefore provide high
electron mobility; cf. § IIF. That is why the conductivity along the
planes is of the same order as in many metal alloys at room temper-
ature, although the number offree n-electrons is relatively small. How-
ever, this number is not reduced to zero at T = 0 °K because the filled
and conducting bands of a near perfect crystal overlap slightly 2 •
In practice, one always deals with polycrystalline carbons. Several
phenomena (for instance, transfer of electrons from the filled to the
conducting band, appearance of holes, scattering of electrons and
hol es) complicate the process of conduction in the carbons; it is to be
noted that the influence of any of them depends on the structure that
the carbon has received by heat treatment, the temperature of which we
denote by Ht.
The number of electrons in the conducting band, thus also the gap
between the filled and conducting band3 , varies with the degree of re-
gularity and size of crystals. MROZOWSKI 4 describes the situation as it
appears at the present time as follows: There is a gap of some tenths
of an eV after low heat treatments but the gap almost closes at the
time hydrogen is driven out. This happens around 1300 °C. From there
on, the bands probably touch each other. In graphitizable materials,
around 2100 °C, they start overlapping at corners of the BRILLOUIN
zones, the overlap increasing as the graphitization proceeds.
The electronsenteringthe conductingband are intrinsic, and to each
of them corresponds a hole; cf. § II. Both electrons and holes contri-
bute to the conduction. However, since the holes have a relatively
small mobility, each of them contributes less to the current than an
electron. Nevertheless, hole conduction dominates in some carbons
because electrons are trapped in a manner described below.
Fig. (V,3) shows features of the electric resistivity, (!, of carbons
distinctive to the structure they have obtained at the highest temper-
ature, Ht, at which they have been treated. The dash-pointed curve
Ht shows the irreversible variation of (! during a gradual increase of
Ht.
1 See PRDIAK and FucHs [1] where earlier Iiterature is cited.
2In the BRILLOUIN zone corners.
s Cf. Eqs. (11,12) and (11,14).
4 Private communication.
410 Appendices
The branches starting at various points of Ht and extending to
the left, are reversibles. These branches reveal strarige features of the
resistivity e at a constant structure. After a highHt (> 2200 oc in the
range CD) the conduction is clearly electronic. But, the HALL effect
has revealed that, in the range BC, p-conduction exists; in the range
AB, n-conduction; and in the range of the lowest Ht, p-conduction
again. Other peculiarities are apparent in the dependence of the revers-
ibles on the temperature.
MRozowsKI ·explains the alternation between p- and n-conduction
by introducing the concept of "excess holes". When during Ht a cova-
lently bound hydrogen atom (or a hydrocarbon group) is Iiberated, it
takes with itone electron from the binding electron pair. The other
electron that belongs to the carbon, becomes "odd" and its orbital is
able to accommodate a second electron. Soon ah available electron
will be captured. At low Ht, it usually is an electron from the filled
n-band as the conducting band is practically empty. The captured
electron leaves a hole in the respective n-band. These holes (excess
holes) can produce a p-conduction that exceeds the in:ßuence of the
relatively few remaining free electrons. Evidently, the free valences
having been left after the rupture of covalimt bonds act as traps for
electrons. After capturing an electron, such a trap is saturated and has
lost its property of a free, reactive valence.
We do not enter into details of further alterations between p- and
n-conduction in the respective regions mentioned above.
Now Iook again at the reversible branches of Fig. (V,3). CUrves e
and d are of the semiconducting type. When the temperature increases
above 0 oc the number of free electrons increases, lowering e. Saturation
appears at about 500 °C. Further rising temperatures produce scatter-
ing by means of phonons1 and e increases as in metals. The curve c
requires another theory. This reversible branch belongs to a p-conduc-
tion and its current carriers are excess holes. They do not alter their
number as long as the temperature varies below the respective Ht. To
explain the drop of e in this region, one could imagine a kind of therm-
ionic emission within the carbon which becomes more efficient the
higher the temperature.
E. Beat conductivity ot carbons. Since knowledge of the heat con-
ductivity is required for application of the rp{} relation, a short report
on this subject is appropriate.
Becaqse of their extreme Jtrength the basal planes have a great
ability. to transport elastic waves along themselves producing a so-
direoted heat oonduotivity that exoeeds that of copper at room tem-
perature. In a polycrystalline graphite the conductivity is much smaller,
particularly due to the scattering of the heat waves (phonons), which
takes place at crystal boundaries. In addition, the waves reßect each

other. While the scattering at the boundaries is independent of the
temperature, the mutual scattering of the waves is proportional to the
KELVIN temperature, T.
The temperature has a further effect on the thermal conduction.
Evidently, heat conduction requires that the conductor be able to
carry a certain heat per m 3 and therefore the heat conductivity con-
tains a factorthat is proportional to the heat capacity. Now according
to Fig. (III,6) the heat capacity of a material is small at low tem-
s_
-1rr1 .Q·cm iA
i
\
, ~ \
J ~~ !'\.
\Ht
\
' .,
I B'-
"""" -------
' r'--\
\ C
------ -------
----- .P. ·':.:- .~
- -----
--
f- ~-
~
-
'-......~
f ------
------- I
Q - ~ --- ,
Q f OW
I
ZWfl <;J(}(}(J
r-
Flg. (V,S). Due to MRozowsKI [1]. Electric reslstivlty, 11. as function of temperature for a carbon
rod treated to temperatures ·: a, 1000; b, 1100; c, 2200; d, 2600; e, 3000 °C. Fullllnes mark the
reglon in whlch the actual measurements were performed (reversible); dashed llnes represent the
expected trend outside of thls region. Curve Ht glves the varlatlon of the resistivity during the
heat treatment of the carbon (irreversible)
perature, and approaches saturation at about the DEBYE t emperature.

That is why the thermal conductivity of carbons is low at very low T.
At such low temperatures only the scattering at the crystal boundaries
prevails. Since the temperature does not affect the scattering at the
boundaries, it exerts its inßuence on the thermal conductivity merely
via the heat capacity (at low T). It is otherwise at high temperatures.
Here the capacity varies little and the mutual scattering between the
412 Appendices
phonons prevails, with the result that the conductivity is proportional

to 1fT.
Fig. (V,4) illustrates the heat conductivity, .?., in carbon grades1
obtained with different Ht, namely
Curve G H B
Ht at 2700 2000 1100 °C
zoo
Wa J
m.iiiög ('.
\
\
-,r-,
II
/
/
, ... --
~~
---
/
.c
I
0
I
/
....
0
-----
.fOO
8
1000 1500 3000 8500
TemperQ/111'(!
Fig. (V,4). Duc to 1\IROZOWSKI [2). Thermal conductivity for various heat-treated carbon. Treat-
ment temperature : for G 2700, for H 2000, for B 1100 •c
Appendix VI. Hydrodynamic or thick film lubrication

The notations hydrodynamic lubrication or thick film lubrication are
used to signify a mode of lubrication, at which the lubricant, where it
is thinnest, still has a thickness of several molecules, thus preventing
any direct contact between the solid contact members. In a stationary
contact a fluid lubricant is not able to preserve this thickness of the
film; the film is soon squeezed out leaving the solid members in a more
or less complete solid contact. A relative movement between the mem-
bers is necessary to sustain a thick lubricating ·film.
The thick film lubrication is characterized by a very small friction
and by absence of wear since all sliding friction is localized in the in-
terior of the liquid lubricant, whereas the monolayer proximate to the
solid a~eres to the solid sliders. In any mode of hydrodynamic lubri-
cation the Ioad bearing ability of the contact is produced by a wedging
efjeot. Imagine, for instance, a stationary contact member and a slider,
1 MROZOWSKI [2).
§ VI. Hydrodynamic or thick film lubrication 413
both flat, and biased so that there is a wedge-like interspace between
them. A lubricant fills this interspace as the slider moves in the direction
in which the interspace becomes wider, and thus has a tendency to
glide up on the lubricant, i. e., to be lifted by it. It is known that the
total upward force in this plane slider bearing is proportionaP to 1/hi,
where h1 is the minimum distance between the contact members.
When, in this book, we meet the problern of hydrodynamic lubri-
cation, it is the question of nominally flat, parallel contact members
with wavy surfaces whereby the amplitude of the waves is micros-
copical. The wedge-like interspaces where the lubricant exerts the
Iifting appear at the front side of the elevations. These elevations will
generally be curved,in contrast to the shape discussed above, and there-
fore we shall not refer to the plane-slider bearing but to curved surfaces
according to Fig. (VI,5). For the case considered we shall find that
the total upward force is proportional to 1/h~, see Eq. (VI,19). It is
important to emphasize this because workers have been led to false
conclusions by trying to explain hydrodynamic Iifting of brushes by
the plane-slider bearing theory. The respective theory can be presented
with reference to Figs. (VI,1) and (VI,5) as follows.
The movement in the liquid is imagined to be laminar in the x-
direction without any z-component. Since mass forces may be neglected
the pressure p depends on x but neither on y nor z. The fact that the
mass forces are negligible makes the old designation of hydrodynamic
ZUbrication a misnomer, and it is often replaced by thick film lubrication.
Xevertheless, we shall keep to the traditional term.
Westart from an equation which we obtain from considering a small
right-angled element of the lubricant with the sides dx, dy, Llz = 1,
in Fig. (Vl,1). The arrows indicate the direc- _k
tion of the forces, which are, with p being the
hydrostaticpressure,pdyand - (P +:~dx)dy ~
on the f?ides 1 and 2 respectively. The shearing
forces at the sides 3 and 4 are - '7: vy d x and Fig. (VI,t),-;;::ions of the
'7 (!; + ::~ d y) dx

forces on a element of the
where v is the velocity of Iubricant
the liquid in the x-direction and '1 its coefficient of internal friction,
also called dynamic viscosity.
The condition for equilibrium between these forces , i. e., under the
assumption of negligible mass forces, is
fPv dp
(VI,2)
'1ay 2 = dx
1 See for instance SHAw-MAcKS [1] Fig. 5-15.
414 Appendices
As in Fig. (VI,5) we consider an upper stationary and a lower moving

member, the velocity of which is w. Because of the adherence between
the liquid and the members, we have
v = 0 at
and (VL3)
V =W at
Integration of Eq. (VI,2) under the boundary conditions (VI,3)
yields
(VI,4)
The further treatment will be referred to (Fig. VI,5) which, as said

above, depicts a plane lower member in motion and an upper curved
member at rest. For simplicity a cylindrical curvature is considered.
The symbol xis a position co-ordinate
increasing to the left, which is also the
direction of w, whereas
I y I
w •~ U=X 1 -x (Vl,6)
lfisf. X ::C1 X 0
lfisf. u. o ~ ;{: increases to the right; a; is the position
~':f:ssuM ~ 4/u o of the smallest gap, ~; k is the gap at x .
ttpfti;c 0 The wedging effect is a result of the
interspace between the surfaces of the
Fig. (VI,6). A Iubricant ftlls the gap be-
tween a flat contact member moving to
members becoming narrower as x varies
theleft andacurvedstatlonarymember
from 0 to x1 , in the direction of the
motion. The cylinder may represent an arbitrary hump on a nominally
ßat contact member. We assume that the following is true with suffi-
cient approximation
at X=0 is p=O and k= h0 )
dp =0 (VI,7)
at X= x1 is
dx
and h= hl
J
The curvature is such that
(VI,8)
i. e., 2y represents the curvature at small u.
W e refer to a length in the z-direction of 1length unit of the cylin-
drical member. With its curvatur~ in accordance with (VI,8), the picture
represents in a fairly general manner thick-film lubricated interspaces
ofsuch a.form that it can be predicted where the section with dpfdx= 0
willlie. Applying the expression for constant ßow
h 11,
J vdy j vdy
0
=
0
§ VI. Hydrodynamic or thick film lubrication 415
where the integral on the left side is to be evaluated at the position x
and that on the right side at the position x1 , we transformEq. (VI,4) to
(VI,9)
Starting from this equation we calculate the pressure p as a func-

tion of x, and later will obtain the Ioad P per cylinder length z = -1.
First we find
(VI,10)
l ' I
\
I
I
1l = q vh~ (VI,11) t l(J
I
I
I I
giving \
h (1 .:_yq2 )hl
II
= ~
Doing so we obtain 1\X I
3 f)V t(\
"f'''"-
p= 4 k-{•yO·•X(q) (VI,12)
:/ :-- -
X (q) is plotted as the full drawn curve on 0
Fig. (VI,13). Fig. (VI,IS). The functiotul X(q)
A similar calculation Ieads to an expres- and Y(q)
sion for the specific friction force 'lfJ. By differ-
entiatingEq. (VI,4)oneobtains dvfdy and an expression can be written
for 'lfJ,
dv
f]- = 'ljJ
dy
One finds different values of the specific friction force in. the vici-
nities of both members, namely,
4h- 3hl f
'lfJ= -f]W h2 or y =O
and
2h- 3h1
'lfJ = fJ w h2 for y = h
The difference is a consequence of the assumption (VI, 7) that arbitra-

rily associates p = 0 with x = 0 not considering the exact influence of w.
416 Appendices
l
We employ as a kind of average
1jJ =
k- k 1 = -.",-;
4fJWk2 4'7W Y(q)
where
Y(q) = (1 +q2q2)2 I (VI,14)
The function Y (q) is plotted as a dashed curve in Fig. (VI,13)

We want the Ioad per unit length z = 1
J pdu
Zt
P= (VI.15)
U=O
and the friction force per z = 1
J tpdu
Zt
F = (VI,16)
U=O
Graphical integration by means of Fig. (VI,13) yields
JXdu~ J Vf
Zt 00
Xdu= 2 (VI,17)
U=O U=O
and
J Ydu~ J Ydu=~
Zt 00
(VI,18)
U=O U=O
Finally employing Eqs. (VI,12), (VI,15), (VI,17) we obtain
p = 1.5'1W (VI,19)
kly
and employing Eqs. (Vl,16), (VI,14), (VI,18)
F=3.2 1 ~ (VI,20)
ykly
Thus the friction coefficient is
(VI,21)
Example. Sliding contact with nominally flat members. As repre-

senting the unevenness of the members we consider a cylindrical hump
on one of them with its axis perpendicular to the relative motion. Its
c1lrvature is supposed tobe in accordance with Eq. {VI,S) with a coeffi-
cient y that will be calculated from the assumptions that x1 = half the
=
diameter of the hump, Zt 10- 5 m, and the height h0 = 26 h1 , h0 being
§ VII. Metal whiskers 417
compared with the smallest gap which is h1 = 2 · 10- 8 m. The relative
velocity of the members is w = 0.1 mjsec. The lubricant is water with
t'J = 10- 3 N sec m-2 = 10- 2 poise1 .
Eq. (Vl,8) applied to h = h0 and u = x 1 gives y = 5000. According
to Eqs. (VI,19) and (VI,21) we find
P = 1.5N per z = 1m
and
t= 0.021
The example shows that in case of a very small h1 , a noticeable Ioad
can be carried even by such a poor lubricant as water, and even at a
very small relative velocity w if the member opposite to the hump is
sufficiently smooth to correspond reasonably to Fig. (VI,5) 2 •
This example illustrates a statement by VoGELPOHL3 [1]. He em-
phasizes that the wedging effect will be particularly strong on the top
of humps on nominally plane and parallel contact members which,
with a lubricant in the interspace, slide on each other. Consequently,
the probability that such humps become bare oflubricant is small. But,
of course, if the humps really come into metallic contact they adhere
and are likely tobe worn away. In the view of VoGELPOHL, the tran-
sition from a thick film lubrication over boundary conditions to dry
friction is continuous.
Appendix VII. Metal whiskers

Concerning earlier observations by TUBANDT, PEARSON, SzEKELY,
DöPFERT, P AETOW and ScHAFFS see R. HoLM [37] Appendix VI.
BRENNER [1] traces observations on whiskersback to the year 1580.
A. Whiskers formed from the solid metal4 • Designers of micro-elec-
tronic equipment have been confronted with the annoyance of unex-
pected short-circuiting caused by threadlike formations that connected
the e~ectrodes over an otherwise insulating gap. Theseformations were
called whiskers. They grow slowly on Sn, Zn and Cd-coatings. It can
take months for a whisker to proceed to the length of 0.03 cm. It is
stated that the whiskers preferrably issue from screw dislocations.
However, it seems that whiskers also form by agglomeration of
powder grains; see experiences with activated contacts p. 277. Cf.
PEARSON [1].
B. Whiskers rapidly formed from metal vapor. LAwsoN [1] working
with a.c. switches at 15A and 120 V, observed an occasional accelerated
1 Poise is the cgs-unit.
2 Cf. BURWELLet al. [1].
3 VoGELPOHL's discussion is cited in TINGLE [1].
4 See particularly .ARNOLD [1], BRENNER [1], KEIL and MEYER [7].
418 Appendices
contact erosion. He could trace back this happening to the formation

of fast-growing whiskers during arcing. They only form in a quiet
atmosphere; conveotion prevents their formation. LAWSON describes
the process as follows: An initially drawn arc produces metal vapor
in which whiskers form in fractions of a second, seemingly arranged
along field lines. They short-circuit the gap causing explosive break-
down and ignition of a new arc during a following cycle. A succession
of arcs could occur. These arcs were the cause of the increased corro-
sion. This kind of whisker formation can be avoided by blowing
away the metal vapor.
· LAWSON found that the alloy (Ag, 26% Cd) is particularly apt to
produce whiskers, while fine silver and the alloy (Ag, 30% Pd) sparcely
produce whiskers. Whiskers were not obtained with either copper or
brass as arcing contacts; wolfram produced whiskers only with d.c.
current.
C. Mechanical and magnetic properties of whiskers 1• By virtue of
the small diameter of the whiskers and the fashion by which they grow,
they contain very few or no dislocations. That is why their strength
can approach that of ideal crystals, which is 100 to 1000 times that of
common structural materials. The elastic elongation lhnit is correspond-
ingly increased.
A ferromagnetic whisker can represent a single row of domains and
therefore be magnetically extremely "hard".
D. Metallic dendritlls in semieonductors, generated by silver migratio-n
at high humidity, have been observed in American telephone stations.
The process has been investigated by KoRMAN, HERMANCE and DoWNES
[1]. They found that silver ions, dissolved in the water film, migrate
und er the infl.uence of the electric field into the insulator, provided they
find through-paths. They are discharged at the cathode and there build
dendrites. A typical example is provided by a step-by-step-dial banlc
where the silver contacts are separated by phenolic fiber bars. When
the bank is not in action but still under a tension of 48 V, silver
migrates through this bank, building dendrites on the cathodic side.
The dendrites eventually reach the anode and short-circuit the bank.
It was found that a high humidity is necessary for the e:ffect. In-
sulators that are neither porous nor hygroscopic resist the migration.
Other metals than Rilver also e~it ions at high humidity. They, how-
ever, are soon stopped, for instance, by forming insoluble oxides or
hydroxides.
The e:ffect described by the authors above is an electrolytic e:ffect.
It is different from that described in § 27 J. There the ions are
delivered without the help of ·humidity, for the e:ffect has also been
1 See BRENNER [1].
§VIII. Some fundamental formulaS concerning electric discharges 419
observed in vacuum. Only the matrix metal, on which the tarnish has
grown, delivers ions noticeably, and not the clean metal when used as
anode against a tarnished cathode.
Appendix VIII.
Some fundamental formuias conceming electric discharges
A. lntroduetion. Kinetie fundamentals. An assembly of independ-
ently1 moving particles, for instance, a gas composed of molecules, ions
and electrons in thermal equilibrium is characterized by a M axwellian
energy distribution among the particles. According to this law, the
probability thst a particle in a given moment has at least the energy
rp eV is
exp (- rpfk T) = exp (- 11600rp/T) (VIII,1)
where T °K is the temperature of the assembly.
The average energy 2 per degree of freedom of the particles is (with
k = BoLTZMANN constant)
~kT (VIII,2)
2
giving for a particle with 3 degrees of freedom
ip=:kT
Hence with ip in eV
_ 3 T
rp = 2 11600
or T = 7730ip (VIII,3)
The free paths ofthe particles also obey a distribution law, but with
respect to the application in this book it suffices to recall that the mean
free p_ath A is inversely proportional to the density of the gas, and to
notice the data of the Table (VIII,4).
Table (VIII,4). Mean free path A of a gas moleeule at 0 °0
He H, air Hg
at 1 mm Hg, A =
at 760mm Hg, A = I 170013.7 I 11208.5 I 4.6
610
I 1.7
220
10-sm
A
The mean free path A. of a fast electron (with a velocity about
sufficient to ionize gas molecules) is 5.65 times greater than A of the
gas mo]ecules; at lower electron speed A. approaches 4 A.
1 Thus, not a "degenerate" assembly; cf. § liD.
2 Cf. §illA.
27•
420 Appendices
B. Drift velocity. If the assembly is subjected to an electric field with

the strength X V fm, electrons and ions are accelerated and will have
another velocity at the end of a free path than they had at its com-
mencement. It may happen that then an electron receives more energy
from the field than it loses at impacts, and on one or several free paths
gathers energy enough to ionize gas particles. Strange as it may seem,
this does not necessarily mean that the energy distribution among the
electrons ceases to be essentially random. It usually remains random
and can thus be referred to a certain temperature, which, however,
may be higher than the temperature of the gas, see Section D about
plasma. But the field superimposes on the random movement of the
electrons a drift velocity in the direction of -X. Ions, even in a fairly
strong field, usually do not attain ionization energy and remain at the
temperature of the gas, the fi.eld giving the positive ions a drift velocity
in the direction of the fi.eld.
Any drift velocity can be written
v=KX (VIII,5)
where K, the mobility, is a coeffi.cient referring to the current carriers

considered and to the gas in which they move. At the gas density of
y times the atmospheric density (which is 2.7 · 1020 moleculesfm 3 ) K
for single-charged atom ions is of the order of 5/y -10- 4 mjsec per Vfm.
For electrons with the same temperature asthat of the ions, the mobility
is about 200 times greater than for the ions. But if the electrons have
their own temperature T. which is higher than T of the gas, their
V
mobility is lowered by the factor TJT•.
In normal atmosphere and weak fields, the ions are able to attach
molecules forming clusters which move several times slower than free
ions.
In a gas composed of N _-electrons per m 3 with the mobility K_ and
N +positive ions of a valence n with the mobility K+, subjected to thE>
fi.eld strength X V fm, the current density J is
(VIII,6)
When N + ~ N _, which then is apparent from the space charge being

small or nil, the term N _K_ dominates, i. e., the electrons are the
essential carriers of the current.
C. Thermal ionization. Saha's equation. It is a consequence of the
Maxwellian distribution of the energy that is expressed in Eq. (VTII,1)
that, in the assembly, energy quanta appear which are sufficient for
ionization. The resulting degree of thermtil ionization can be computed,
§ VIII. Some fundamental formulas concerning electric discharges 421
at least for an atomic gas, by means of Saha' s equation1 • The restriction

to atomic gas is insignificant since we shall employ the equation solely
for monoatomic metal vapor or for temperatures at which any gas is
largely dissociated into atoms. Let p be the gas pressure in atmospheres,
(/Ji the ionization potential of the atoms in volt, and
number of electrons
x = number of neutral atoms
then SAHA's equation is

x2
-1 - 2 p = 3.1·10- 7 T 2 · 5 exp (- 11600cp;/T) (VIII,7)
-X
Table (VIII,8). Example.s of X·values

Copper vapor with
air with 'l'l = 14 eV
'I'«= 7.8eV
Tin °K 5000 10000 15000 10000
p in atmospheres
X=
1
2 • 10- 6
1
0.016
I 10
0.005
1
0.38
1
0.51
20
0.13
D. Plasma. A gas volume that is highly ionized with N _ ~ N + is

called a (quasineutral) plasma. Examples are the positive columns in
glow discharges and electric arcs 2 •
In a neutral gas the Maxwellian energy distribution is produced
by means of energy exchange at impacts between the molecules. Elec-
trons and ions are able to exchange energy without coming in touch
with each other, namely by means of their far-reaching electric fields,
either repelling or attracting. The electrons, because of the smallness
of their mass, give off energy in very small quanta at a time to heavier
particles and therefore experience a relatively slow energy exchange
with atoms and even with ions. But by means of electrostatic repulsions
among themselves they rapidly develop a random energy distribution
which ·may belong to another temperature than that of the gas. For
instance, in the positive column of a glow discharge where the electrons
are accelerated by the field to a considerable velocity they nevertheless
develop a random energy distribution among themselves. Thereby their
temperature may be 70000 °K whereas the gas and the ions have only
slightly more than room temperature.
It is otherwise in an arc plasma, because here the density of charged
particles very much enhances the energy exchange. In such a plasma
atoms, ions and electrons will have about the same teinperature, which
is the condition for SAHA's equation tobe applicable.
1 See e. g. UNSÖLD [1] chapter IV, or LOEB [1] p. 618.
2 Cf. FINKELNBURG and MAECKER [1].
422 Appendices
E. Current in vacuum restricted by the space charge of the current

carriers. Fig. (VIII,9) illustrates the basic assumptious for the relations.
The lower figure depicts two parallel" electrodes in vacuum the dis-
tance 8 apart from each other. The anode emits positive ions whose
0
initial velocity1 we neglect. They are attracted
i
by the charge of the electrons (small circles)
localized in the cathode face and thus move
::",.
towards the cathode. Lines of force are design-
ed as connecting one positive ion and one elec-
tron. The ions are accelerated in the described
z-
J field and thus constitute a moving space charge
and a current density J +· They are supposed
to be so numerous as to just occupy alllines
-~ of force issuing from the cathode, thus making

the field strength vanish at the surface of the
anode. Calculations proceeding from the .fol-
- -- s - lowing conditions:
Fig. (VIII,9). Illustration of the
potential fteld between parallel
electrodesforthecaseof a Satu-
ration current of positive ions and a = J+ (VIII,10)
tl
flowing from anode to cathode
where x = distance from the anode, a = density of space charge,

e0 = 8.85 · 10- 12 Ffm, v = velocity of the ions and n = valence, lead 2
to CHILD's equation
Vl··
J. = 0.545 · 10- 7 - = A/m 2 (VIII,11)
s2VM!n
where M = atomie weight of the ions, 8 is measured in m, and V volt
is the potential difference between the electrodes.
It is to be noticed that Eq. (VIII,11) is independent of the rate. of
supply of ions from the anode beyond J + as given in (VIII,11). No
greater current density than J + can be transferred by V across 8 pro-
vided that, as is assumed above, the ions do not leave the anode with
a velocity to be conside.red. But, of course, whe.n the supply is smaller
than the current density J + given by (VIII,11), then the current is
no Ionger given by (VIII,11) and just equals the supply.
Similar calculations as above ~J.re valid for the emission of electrons
frori:t a cathode. The solution is again (VIII,11) with M = 1/1840, hence
J = 2 3 ·10-s VI·::. Afm2 (VIII,12)

- ' 82
1 As for the influence of the initial velocity see, for instance, NoTTINGHAM [2]
p. 36ff.
2 See, for instance, LOEB [1] p. 317ft".
§ IX. Theory of the electric arc 423
In both cases, for ions and electrons as weH, represented by Eqs.
(VIII,11) and (VIII,12) the field strength X in immediate vicinity of
the electrodes to which the carriers move is given by
4 V
X=--
3 8
(VIII,13)
We refer to the lower picture of Fig. (VIII,9) also for expressing the
following expedient formula. Consider the field strength in the imme-
diate vicinity of the cathode. It is represented by parallel force lines
which end in equal numbers of electrons and ions. Let this number be
n per m 2 of the cathode surface. Then, with X in Vfm, a simple calcu-
lation shows that
108
n=-X
1.8
(VIII,14)
It is immaterial whether the electrons are situated in the cathode sur-

face or spread as the ions are.
Appendix IX. Theory of the electric arc

A. Introduction. In this appendix, details of the theory of the arc
are treated with the purpose of unburdening the description of switch
phenomena. The presentation is roughly outlined but suffices for our
purpose. It also fits to the fact that the theory still is in a fluid state,
and many experimental data, for instance concerning the current den-
sity at the cathode, have only a preliminary character. The following
more complete surveys of the modern theory of the arc are recom-
mended: BAUER (2], [3], and MAECKER (2].
However, our deductions and results deviate from those author's
representation in several respects.
One distinguishes two main types1 of self-sustaining discharges: the
glow and the arc. Their most conspicuous characteristics are the fol-
lowing. The glow has a cathode fall of a few 100 V or more, a moderate
cathode temperature, and a relatively small current density. The arc
has a cathode fall of the order of 10 V, a very high cathode tempera-
ture, and a very high current density at the cathode.
It is the type of emission of primary electrons from the cathode that
determines the di:fference. In the glow, the cathode temperature is too
low to influence electron emission. Emission is largely due to the AuGER
e:ffect at the normal cathode fall of a few 100 V and also to ion impact
at a higher cathode fall.
1 The discharge during fl.oating (see §59) is a special kind of arc that has no
proper name and is not treated in Handbooks.

424 Appendices
The AuGER effect uses the energy that is freed when an ion recom-
bines with an electron from the cathode surface1 • With this energy, a
second electron is liberated from the cathode. However, any primary
electron is bound to produce many ions (10 or more) in order to secure
one single secondary electron liberated from the cathode by the AuGER
effect. ·This explains the need forahigh cathode fall in the glow. The
AUGER effect is negligible for the arc.
The arc makes use of its high cathode temperature. In the case of
refractory materials such as wolfram and carbon, enough primary elec-
trons are thermionically emitted from cathodes which remain solid.
But, low melting cathodes require a complex process. The temperature
must rise beyond the normal boiling point, and energy must be pro-
vided to maintain the evaporation during boiling. This is possible
only when the cathode spot is small and thus the current density high;
namely several powers of ten higher than on refractory materials.
The high density of primary electron emission, in its turn, cannot
be produced by the temperature alone. A strong field at the cathode
surface is required. The average field, although strong itself, does not
suffice for the effect. But the field is enhanced at asperity peaks on the
cathode, and its inßuence on the emission is increased by the I-effect;
see below. In addition, the work function (/J of the boiling metal in the
cathode spot certainlyis smaller than that of a solid metal versus vacuum.
A very high efficiency is required from the aiding effects as they
have to explain about three orders of magnitude of the current
density. The theoretical situation is unsatisfactory because any basic
experimental data still Iack the necessary accuracy, and only rough
computations can be made. Field enhancing is an evident phenomenon.
But an equally direct knowledge about the mentioned diminution of
(/J has not been considered so far, and authors have preferred to cal-
culate with an unchanged (/J (namely (/J = 4.5 eV, the value for elec-
tron emission into vacuum) and accordingly have tried to make the
other effects responsible for the entire three orders of magnitude. For
this purpose, factors greater than probable had tobe assumed.
It will be shown that there are good reasons to assume an inßuential
decrease of (/J with the consequence that in doing so, a far reaching con-
sistency with experimental data is gained.
One distinguishes 3 sections o:f the arc, which will be treated sepa-
rately:
1. The- catkode layer identical with the region of the cathode fall;
2. The arwde layer; 3. The plasma, also called positive column, com-
prising all of the arc between the terminallayers.
1 For complete description see, for instance,. LOEB [1] p. 771ff., or LITTLE [ll
p. 651.
§IX. Theory of the electric arc 425
B. Cathode layer. The features of tbe terminal layers are largely
independent of the existence of a plasma between them. This is parti-
cularly true for the cathode layer. This layer receives heat conducted or
radiated from a thin, bot layer bordering the plasma where the primary
electrons give offtheir energy. But, the more distaut parts of the plasma
do not much affect the balance at the cathode. The border region bas
a negligible voltage. It will therefore suffice to treat the cathode layer
as belonging to a short arc with the total voltage V(::::::: Vm= minimum
arc voltage) equal to the sum ofthe cathode fall , V0 , and an anode part,
va' while the voltage of the plasma can be neglected.
Fig. (IX , I). Illustration of the positive space charge and the fteld in the cathode fall region
Fig. (IX,1) depicts a short arc between circular cathode and anode
spots with radii a,
and aa respectively1 • The cathode fall acts across
the cathode layer of thickness, 8 . The difference, V- V 0 , between the
total arc voltage and the cathode fall may be an anode fall, V4 ; but
it is of little significance whether or not this voltage deserves the
name of anode fall2. The only requirement isthat the arc is flat , i. e.,
much shorter than the diameter of the spots ; thus, with negligible la-
teral heat losses. This has the consequence that the arc voltage is of
the order of 10 V.
Fig. (IX,1) is schematic, and 8 is designed too !arge compared with
a0 • Actually the ratio a0 j8 may be several hundred. The plus signs in-
dicate· positive ions on their way to the cathode through the cathode
layer. It is weil known that the cathode fall is generated between those
ions and the negative charge of the cathode. This is indicated by force
lines having their bases in the ions. All force lines end at the cathode.
representing a strong field there. No Iines are reaching the anodicside
of the cathode layer, pointing to the fact that there the field strength
is negligible compared with the field within the cathode layer. The simi-
larity between this picture and Fig. (VIII,9) is evident.
1 The theory, here presented, applies to arcs with one single cathodic and one
single anodic spot, as is realized at drawn arcs.

2 The anode fall is no indispensable part of the arc. It is a consequence of Iack
of positive ions to neutralize the space charge of the electron stream proximate to
the anode; See FINKELNBURG and MAECKER (J].
426 Appendices
W e carry out calculations on the basis of the following assumptions:

1. The ions enter the cathode layer with negligible velocity1 . 2. The ions
are accelerated through this layer by the cathode fall, essentially un-
disturbed by impacts. Then, Eqs. (VIII,11) 2 and (VIII,13) can be
applied. Eq. (VIII,11) can be used for calculation of the thickness s of
the cathode layer, if the ion current density, J +• in the layer is known.
We put J +""' Jj200 within the plasma, where J is the total current
density, and reason as follows. The plasma contains the same number
of electrons and ions (quasi-neutrality). Since the electrons have a
mobility about 200 times higher than that of the ions, J + has the value
mentioned above. About the same J + (possibly slightly higher) is valid
for the cathode layer. One could expect that the ionization by impact
of the primary electrons might produce an influential number of ions
in a plasma region bordering the cathode layer. But then a positive
charge would appear in this region causing a strong field gradient and
an increase of the cathode fall. However, essential effects of this kind
are not indicated by experimental data. Therefore, we calculate with
J +""' Jj200 in the cathode layer.
C. Emission of primary electrons from the cathode. The problern is
relatively simple when the cathode consists of refractory material as
wolfram and carbon. Here the normal cathode spot reaches tempera-
tures of about 3500 °K without causing melting on the cathode. This
temperature suffices for thermionic emission of the observed amount,
namely J _ = 5 · 106 to 10 · 106 Afm 2 • For more information see
FINKELNBURG and MAECKER [1].
Great theoretical difficulties arise when the behavior of low melting
cathodes (copper, silver, etc.) shall be explained. With them, the cur-
rent density J _ is several powers of ten greater than at refractory
cathodes, indicating that pure thermionic emission plays a minor part.
We begin by discussing how the strong field, X, at the boiling front
of the cathode enlarges the emission of primary electrons. The emission
is influenced in three respects : 1. The strong field provides the ScHOTTKY
correction that increases thermionic emission in accordance with Eq.
(II,18). 2. It causes field emission in accordance with Eq. (26.12). This
emission is enhanced by the high temperature as described with re-
ference to Fig. (26.14). 3. A stropg field is a condition for electron
emission with the aid of individual ions. This effect is described with
reference to Fig. (49.02).
It is common to use the symbol, T-F, for the combined processes
1 and 2, and the symbol, 1-effect, for the process 3. The modern for-
1 The relatively few ions generated by the high temperature within the ca-
thode layer are neglected.
2 With a correction treated in Section D.
mulas for T-F were derived by DoLAN and DYKE [1] with additions
by MURPHY and Goon [1]. Numerical values based on their theories
have been calculated by LEE [2] with the aid of a digital computer
and presented in diagrams1 •
In the first instance, we consider an arc without 1-effect at a smooth
plane silver cathode with the current I = 20 A, the total current den-
sity J = 1010 Ajm 2 , the cathode fall= 8 V, and the atoniic weight
= 108. In this case, J +""' 0.5 · 108 Ajm 2 • Inserting pertinent values in
Eqs. (VIII,11) and (VIII,13) we obtain X 0 = 2.3 · 108 Vfm and 8
= 4.8 · 10- 8 m for the thickness of the cathode layer where X 0 refers to
a smooth cathode spot.
The value 8 = 4.8 . 10-8 m is of special interest. Whe shall find good
reasons for the assumption that, in an arc of" 20 A, the cathode spot
has a temperature of the ordec of 4000 °K, more precisely 3500° corre-
sponding to a vapor saturation pressure 2 of 20 atm. Under these circum-
stances the mean free path of the ions isA = 4.8 · 10-8 m, i.e., equal to 8.
This seems to be a very reasonable coindicence3 • Decisive facta behind
the assumptions are discussed below with reference to Table (IX,5) and
Eq. (IX,8). We add the following. The emitted vapor exerts a repulsive
force on the cathode. Measurements of such forces are summarized in
BAUER [1] p. 571. Using the repulse measurements on copper arcs at
20 A in vacuum4 , one finds that the rate of evaporation, w = 5 . 10-1 2 m 3
per coul as given in Table (IX,5), corresponds to the vapor leaving the
cathode at 4200 °K. This is the expected order of magnitude. The
deviation from 3500° may weil be explained by the w-value being set
somewhat too high, or to some other inaccuracy.
For the arc at 20 A, weshall keep to the values X 0 = 2.3 · 108 Vjm
and T = 3500 °K. Tobegin with we apply the work function f/J = 4.5 e V
as is customary. Entering Lee's diagram with these values, one obtains
the emitted current density
J_ = 3 ·107 A/m 2
which is about 3 magnitudes less than observed.

However, the cathode surface is not smooth, certainly not during
boiling. The field is strengthened at the asperity peaks so that the
emission proceeds as if the cathode were smooth and the field strength
increased by a "field enhancing factor". T. J. LEWIS [1] showed that
1 See§ 26H.
2 Much higher pressure appears during floating; see § 59.
3 It lies beyond the scope of the present survey to discuss how a moderate
deviation of s from A would influence Table (IX,5) and Eq. (IX,S).

' Measurements in ambient air of 1 atm are marred by an effect that is dis-
cussed in BAUNR [1] p. 572.
428 Appendices
this factor could not be greater than 2. Application of this factor,
i. e., using X= 4.6 · 108 Vfm does not lead to a satisfying value of J _,
We now introduce the 1-effect. According to BAUER [4], this effect
would add to J _ a factor of 10, if J + were as large as 0.2 J and <P
= 4.5 eV. But, in the case of a silver electrode, J + is only Jj200.
However, the diminution of the work function (see below) supports
the 1-effect, and we shall compute as if both the field enhancing effect
and the 1-effect produced what a field enhancing factor of 3 would
accomplish. Hence, we apply an effective field of
Xeu = 6.9 ·10 8 V/m
Entering LEE's diagrams wit.h X elf= 6.9 . 108 Vjm' T = 3500 °K,
and <P = 4.5 eV, we obtain
still 100 times too small.

We now consider the probable diminution of the work function. It
can be computed if we compare the contact between the cathode and
the ionized vapor in the arc with the contact between a metaland a
semiconductor. According to ScHOTTKY's theory, the work function
(/) for electron emission from the metal into the semiconductor is the
difference (/) m - C where (/) m = work function for emission from the
metal into vacuum and C= the electron affinity1 of the semiconductor.
In our case, C corresponds to the average attraction between an elec-
tron and a positive ion in the plasma (including the cathode layer)
bordering the cathode. Iri this region the temperature is very high,
partly due to the acceleration and impacts of the carriers. The ioni-
zation will correspond to a temperature of the order of 10000 °K and
will be 10 to 50%. Then the average distance between ions is of the
order of 50 A. The most probable distance between electrons and ions
is smaller than 25 A; namely of the order of 10 A. Accordingly, the
attraction between electrons and positive ions corresponds to an electron
affinity ofabout 1.5eV. This means that the work function for emission
of electrons into the arc is no Ionger 4.5 eV but <P = 4.5- 1.5 = 3 eV.
W e shall use this value·.
It will be seen that by using <P = 3 eV, quite reasonable values of
J _ are calculated. Beyond this, one obtains a good agreement with
the equation for the power bala:rice at the cathode. However, this does
not mean that the problems are definitely solved, in part due to the
great spread of the observed values of the current density, from which
the calculations start, and partly due to the uncertainty of the assump-
tions.
1 Cf. p. 390.
For simplicity we confine the treatment to two examples. Assump-
tions for the examples: the metal of the electrodes is silver with the
density e = 10.5 · 103 kgfm 3 and the cathode fall= 8 V. In Example I,
the current is I = 20 A, and the temperature in the boiling cathode
spot is 3500 °K. The corresponding values for Example lJ are I= 10 4 A
and 2500 °K. The temperatures are set so as to lead to the observed
average emission current densities 1 • They are boiling temperatures. The
pertinent vapor pressures, obtained by extrapolation of a curve for
silver 2 , are noted in Table (IX,5).
D. The ion current in the cathode layer, and w, the rate of evaporation.
Following MACKEOWN [1], we distinguish two different parts of the
ion current density: One part 1:,
the charge of which is compensated
by the electrons that consitute J _; the other part, J~, is given by
CHILD's formula, the same as Eq. (VIII,11).
J'' = 0.545 • 10-7 V~· 5 Ajm2 (IX,2)

+ A2VM!n
where M is the atomic weight of the metal, with the valence n = 1

for silver.J:
is given by
J' = J_ (IX,3)
+ V1850M
We find for Example I
J: + J~ = 2.25 ·10 7 + 5.25 • 107 = 7.5 ·10 7 Ajm 2
and for Example II
J: + J~ = 1.34 · 106 + 4.6 · 10 6 = 5.9 ·106 Ajm 2
Only the not compensated part, J~, is responsible for the cathode
fall and the field in front of the cathode 3 • The total J: + J~ is involved
in the I-effect. Eqs. (IX,2) and (IX,3) are used for calculations con-
cerning Table (IX,5). J: + J~ turnsout somewhat greater than J _f200,
but all values lie witbin the uncertainty margin.
E. The rate of evaporation, ro, m 3fcoul. The metal vapor issued from
the cathod spot is heated by the current to a temperature high above
the temperature of the spot. This obstructs the evaporation to some
degree. Exact formulas for evaporation under such circumstances
are not available. Only a rough approximation can be obtained as
follows. Let p atm be the saturation pressure of the metal vapor at
1 Observations particularly by FRoOME [J], GERMER et al. [6] and SWINEHART,
private communication 1963.
2 American Institute of Physics Handbook (1963) p. 4--286.
3 Cf. Eqs. (VIII,11) and (VIII,13).
430 Appendices
the temperature of the cathode surface, and p 1 atm the actual vapor
pressure in front of the cathode. Then the rate of evaporation would
be1
ro = J440~
(!,
T (p- p 1 ) m 3 /coul (XI,4)
where e. = density of the metal in kgjm 3 , J = current density in Ajm2

and M = atomic weight of the metal, here silver. Weshall apply Eq.
(IX,4) with p = saturation pressure at T and p 1 = 0 to obtain the
order of magnitude.
Fig; (56.09) presents measurements of the coefficient y defining
the cathode loss, which satisfies 2.5y = ro, in accordance with Eq.
(56.07). For comparison with ro calculated for I= 20 A, we should
consider the value of y for silver observed in N 2 • This value is y ...,
2 · 10-12 mBfcoul, thus ro = 2.5y = 5. 10-12 m 3fcoul. The low posi-
tions of the plotted y-values of silver in.air (noted in Fig. [56.09]) have
not been explained. In the case of very high currents as I= 104 A in
Example II, the maximum rate 2 is ro ..., 125 · 10-12 m 3fcoul. The cal-
culated values for ro are 20 · 10- 12 and 58 · 10- 12 m 3fcoul respectively.
The agreement between calculated and observed ro-values is poor (cf.
Table IX,5) as is to be expected if one considers the limited validity
of Eq. (IX,4).
F. Summary of the results presented in Table (IX,o).
Table (IX,5)
Example
I T
Cathode
p A J+ I J_ x. w
A •K atm 10-8 m 107 Atm• t()l• .tOOV/m 10- 11 m•tcoul

(given) (assumed) calc calc calc• calc• calc 1observ
I
II I 20
10'
I 2500
3500 I 20
4
I 164.8 I 0.59
7.5
1100.06 I 2.3
0.43
I 20
58
I 1235
G. Power balanee at the eathode. We consider a silver cathode spot
with the radins a•. The total arc voltage is V= 12 V with the cathode
fall V.=8V.
Energy gain. Any positive ion that enters the cathode gives off both
its kinetic and its excitation energy. Thus, a current /+ of single-
charged Ag+ ions gives off V)+ watt kinetic energy and VJ+ ioni-
1 Cf. 'nE BoER [1] p. 17 and 18.
2 Cf. p. 308.
3 The J _-values noted in the Table, are often measured. However, values up
to about 10lll A/m 2 have been verified with short arcs by GERMER et al. [6], see
their Fig. (3b); cf. "floating" in§ 59.
' Acoording to (VIII,13).
zation energy, where V; is the ionization voltage. Since V;~ Vc, we
note the total power as 2 VJ+· This term is of minor importance in
Eq. (IX,8).
We shall find that a considerable fraction of the total arc power,
V I, remains unbalanced if we do not assume that it is transferred by
conduction or radiation to the electrodes. Transfer to the surroundings
certainly is negligible since the space of 'the short arc has a diameter
of about 2ac which may be 100 times larger than the width ofthe gap.
The total power transferred to the electrodes in this way is equal to the
total arc power V I minus the energies given off by positive ions to the
cathode and by electrons to the anode; i. e.,
(IX,6)
where Va represents the average energy of an electron impinging per
sec on the anode. It is reasonable to assume that the anode normally
receives a smaller portion of this power than the cathode by virtue of
its greater distance from the particularly heated section at the border
of the cathode layer. We calculate with the fraction LI to the cathode
and with (1- LI) to the anode.
Power losses. On calculating the electron emission from the cathode,
we apply formulas containing the work function (/) as a variable. This
does not meaq that each escaping electron can climb over the top of
a potential hill of (/) eV. The greater part of the electrons tunnel
through the hill. Therefore, the average evaporation energy of the elec-
trons is smaller than (/), on an average ß(/); and the total power the
cathode loses, owing to the electron emission, is ß(/) I. We do not try
to calculate ß. However, Table (IX,10) shows that the assumption of
ß = 1/3, which is reasonable, leads to a very good confirmation of the
equation for the energy balance at the cathode.
Weshall now express the heat conducted into the bulk of the cath-
ode. First, consider the heat fiow that starts in the_ cathode spot. This
spot is usually so small compared with the cross section of the cathode
that the thermal constriction resistance, W, is well given by the formula
W = 1J4a0 Ä. for a long constriction in one fiat member. Thus, the power
conducted from the cathode is about 4acÄ. ( T - T1), where T 0 _K is
the actual temperaturein the spot, and T1 (about 300°) is the ambient
temperature.
It is evident that RP heat is also developed within the constriction.
It contributes to T. However, we find that this contribution can be
neglected. The thermal resistance against the fiow of this heat is W/2.
The electric constriction resistance is fef4a 0 , where fe is the average
resistivity at the high temperature in the cathode spot with (} = 1.65
10- 8 Q m and the factor f ~ 6 in accordance with Fig. (16.14).
432 Appendices
Hence, the power conducted by the current I within the constriction

is feJ2f4ac. This power(= heat per sec) flows through the thermal re-
sistance W/2 and generates the supertemperaturein the spot
(IX,7)
This gives for I= 20 A, J = 1010 Afm2 and). = 300 wattjm °K
and for I = 104 and J = 6 · 108
Evidently, the temperature increment developed by the current

within the cathode is fairly negligible1 in both examples. We neglect
it in the following.
Finally there is the power spent on evaporation. With r being the
heat per ms for both melting and evaporating, the respective power
is wr I. In case of silverl, r = 2.6 · 1010 joulefm3 •
According to the preceding discussion, the balance equation for the
cathode is
As pointed out above, Eq. (IX,S) is valid not only for short arcs but
also for arcs with a long plasma, if V means the sum of the cathode
and anodefalls alone, i. e., V= Vc + Va.
Transformation of Eq. (IX,S). From I= na~ Jwe obtain ac = V :niJ .
Furthermore, I+= 1~.Putting V= 12 V, means calculating with

Va = 4 V. The average heat conductivity in the hot, partly liquid
cathode is).""' 300 wattjm 0 K. Let us try LI= 0.7.
Inserting in Eq._(IX,S) the values LI= 0.7, V= 12, Vc = 8,
Ä = 300, r = 2.6. 1010, we obtain
4.8I:; + 5.6I =I+ (:)80 Vi(T- T 1 ) +I w2.6 ·1010 (IX,9)
With the numerical values of I, J, J+, T, T1 (cf. Table [IX,5]), and

w = 5 · 10- 12 and 123 · 10-12 in m 3fcoul for Examples I and II respec-
Note that this result corrects diverging enunciations in the literature, as by

1
RICH [1]. Furthermore note that the calcUlation above is not valid in the case
of a narrow cathode, see Section (IX,I).
tively, we calculatetheleft side and right side members ofEq. (IX,9),
as noted in Table (IX,10).
Table (IX,lO)
Left Right side

Example side member member
I
II
112.7 watt
5.6. 10'watt
I 4.21.
119.6 watt
10' watt
The agreement is satisfactory. This means that both Eq. (IX,8) and the
assumptions are reasonable, with all numbers nearly correct for a cer-
tain realizable arc. But not every arc has L1 = 0.7 and ß = 1/3. A very
short arc may have L1 = 0.5.
H. Power balanee at the anode. An equation for the power balance
at the anode is similar to Eq. (IX,8) and is readily composed. It
reads with Ta = anode temperature
V.I +(I- L1)(V I- 2 Vcl+- V.I) +([).I= 4a.J.(T.- T 1 ) +I war
. (IX,11)
The term l/Jal measures the condensation heat of the electrons. (/Ja
is not necessarily equal to l/J. The plasma can have somewhat differing
properties at the anode and the cathode. But (/Ja=(/)= 3 eV is cer-
tainly a good approximation. We notice that Eq. (IX,11) has two
important terms on the "gain" side: Valand l/Jal, whereas the only
corresponding term in (IX,8) is the negligible 2 V. I+· This means that
the anode receives more heat than the cathode. Whether this renders
the anode the warmer electrode depends on the amount of the term
4a.J. (Ta- T1 ). In case of a very short arc, aa is not far from being
equal to a0 • The factor 4aaÄ is then not so much greater than
the corresponding term in Eq. (IX,8) that it could compensate for the
stroriger heating of the anode. Consequently, Tabecomes greater than
the temperature of the cathode spot. But, as decribed in § 64C, when
the arc length exceeds a certain critical value, aa becomes much greater
than a•. The anode is correspondingly cooled and the cathode is the
hotter electrode.
I. Voltage-eurrent eharacteristics of ares. Are life. We begin with
discussing the minimum arc current. In Eq. (IX,9) we neglect the first
term, devide the rest by I and obtain
680
4.6- w 2.6 ·1010 = , , - (T- T1) (IX,12)
viJ
Let I decrease. The bala.nce expres.'3ed by Eq. (IX,12) a.nd (IX,8)
is automa.tica.lly maintained by cha.nges of the other variables, pa.rtic-
434 Appendices
ularly by inorease of J. But it is evident that the respause is restricted

and tha.t therefore a lower limit, Im, of I exists. In other words, the
maintena.nce of the arc is more difficult the closer I approa.ches I", .
There is an easily observed reaction of the arc against this difficulty,
which is not explicit in the simplified Eq. (IX, 12); namely the increase
of the ca.thode fall, Vc, when I approaches I ,n· This has a bearing on
the VI-cha.racteristics, as will be discussed immediately.
But first we point to the fact that the validity of Eqs. (IX,8) and
(IX,12) is limited. The term
refers to a cathode that has a much larger diameter than the cathode
spot. However, when the cathode cross section is of the order of the spot,
the heat conduction into the electrode is small with the consequence
thatthe aro can be maintained by a current smaller than Im. Cf. §49B.
In problems concerning relays, the current I belongs to the range
where the cathode fall is sensitive to I. Calculations are difficult. Valu-
able and subtile information is obtained from empirical voltage-current
characteristics of the arcs. How they are recorded and employed for
the determination of the duration of arcs, is described in § 50. Indi-
cations of the explanation of features of the VI-characteristics will
now be given.
The minimum current, Im, of the arc is never exactly attained but
plays the part of an asymptote; see below.
The valtage also has an asymptotic lower limit, V"., termed the
minimum voltage of the arc. The arc valtage can approach V m when the
arc uses only negligible valtage for maintaining a plasma. We there-
\
\
'
Flg. (IX,13). IllW!tration of the V 1-characteristic of the shortest arc
fore expect that V", is best approached when the arc is plasmaless. The
dependence of the arc on the cathode material is indicated by V 111 being
a function of the cathode material.
Imagine the following experiment. An arc has been drawn by open-
ing a. current-carrying contact and is kept burning at I >Im in a gap
sufficiently short to make V~ V m. Mterwards, I is diminished by means
of a. series resistance. The discussion about the existence of Im implies
that the arc experiences increasing difficulties as I approaches Im. We
expect that these difficulties require an increase of the arc voltage, V.
Ac~ually, this is what happens. The characteristic of the shortest arc
does not consist simply of the straight lines V= V m and I= Im in
Fig. (IX,13), but, of a curve (dashed) that looks like a hyperbola. As
IvEs [1] has shownl, the characteristic ofthe shortest arc is approached
by a hyperbola
(V - V",) (I - 1111 ) = C (IX,14)
where C is a material constant, referring essentia.lly to the cathode

material a.nd the atmosphere. If the ga.p is extended and a pla.sma
develops, its voltage VP adds to Vm. Then, a.pproximately
[V- Vm- V]J(s,l)J(I~- I,..)= O(s) (IX,15)
where Vp(s,I) and O(s) are functions ofthe arc length s, the current I
and also depend on the a.tmosphere.
We make no use of algebraic expression for VP and 0; instead, em-
pirical diagrams are discussed in§ 50 where it is also shown that a simpli-
fied, diagrammatic representation can be based on the fact that V P
is a. function of the atmosphere but widely independent of the material
of the electrodes. On this empirical representa.tion is ba.sed a. simplified
methode to determine the life of the arc, also described in §50.
J. Pinch e:llect. We have seen that the arc needs a. strong current
concentration a.t the cathode spot. The pinch effect is one of the pro-
moters of the concentra.tion. In a. cylindrical current fl.ow with the
radius ac (= the spot radius) the pinch effect generates a squeezing
effect with a. pressure of
(IX,16)
where /Lo = 4n · 10-7 Hfm. In the event of Example I this means 0.2 atm.
This effect produces a. gas flow from the cathode2 •
1 The hyperbolic shape of the characteristic had also been observed early by
other workers, for instance NoTTINGH.UI [11.

2 See MAECKER (1].
28*
'fable (X,1) t
~
(! ;. E H
Material Density Electrical
"'
temp. coeff. of IJ thermal conduc- YOUNG'S contact Softening Bolling
resistivity between 0 and tivity at 18 oc modulus hardness
I Melting I
at 18 oc 100 °C temperature
103 kgtm• 10-• !l m to-• watt/m °K 10'" N/m' 108 N/m' oc
------ ---·-- ------ ---- ----- -
Metals
Ag 10.5 1.65 4 418 7 3 to 7 180 960 2193 Ag
Al 2.7 2.9 4 210 7 1.8 to 4 150 660 2447 Al
Au 19.3 2.3 4 310 8 2 to 7 100 1063 2817 Au
Cr 7 20 2.5 67 7 to 13 1903 2642 Cr
Cu 8.9 1.8 4 380 11 4 to 7 190 1083 2582 Cu
Fe 7.8 10 6.5 60 20 6 500 1539 2887 Fe
Hg 13.6 96 1 11 -39 357 Hg ~
Ir 22.4 5.5 4.5 6 27 2443 4127 Ir
Mo 10 5.8 4.5 140 30 18 900 2610 3700* Mo
1
Ni 8.8 8 7 70 20 7 to 20 520 1453 2837 Ni ~
~
Pb 11.3 21 4 35 1.5 0.5 200 327 1750 Pb
Pd 12 10.8 3.3 70 12 4 to 10 1552 3560* Pd
Pt 21.4 11 3.8 70 13 4 to 8 540 1772 4010* Pt
Sn 7.3 12 4.5 64 5 0.45 to .6 100 232 2507 Sn
Ta 16.6 15 3 54 19 10 to 30 800 2996 5427 Ta
Ti 4.5 42 3 17 11 1668 >3000* Ti
w 19.3 5.5 5 180 30 12 to 40 1000 3380 5527 w
Zn 7.1 6.1 3.7 111 11 3 to 6 170 420 908 Zn
Alloys
Au, 24 Ag, 7 Pd 16 15 10 780
Ag, 10 Cu 10.3 3 300 6 I I 900
Ag, 3 Pd 10.5 3 300 4
Ag, 40 Pd 11 21. 30 9.5 1200
Ag, 60 Pd 11.3 42 10 I I 1300
Cu, 40 Zn 8.4 8.5 100 !) 10 950
Cu, 10 Sn 8.7 18 50 10 7 I I 1000
Constanta.n 8.8 50 23 17 8 to 20
German Silver 8.8 33 30 8.2 15
Cu, 60 Pd 10.5 35 12 I I 1200
Cu, 2.4 Be 8.2 10 110 13 35
Pt, 10 Ir 21.6 25 40 15 6 to 20 1780
Pt, 25 Ir 21.7 35 30 12.5 to 24 1830
Pt, 10 Rh 20 19 10 9 1830
Stainless Steel 7.9 55 to 70 17 15 to 30 16 1400
Steel 7.8 14 50 20 10 to 30 1430
Sintered Oompounda
W, 15 Ag 17 4
W, 25 Ag 15.8 3.5 230 20
W, 50 Ag 13.3 2.6 10
Mo, 25 Ag 10.3 3.2 18 '""'
Mo, 40Ag 10.3 2.9 13 I I I I ~
W, 35 Cu, 0.5 Ni 13.6 5.3 150 15
W, 70 Cu, 0.5 Ni 3.3
Carbides and Oarbona1
WC, 13Co 14 20 70 70 20 >2800
>1000 I
wc 15 43 40 55 20 >1000 >2800
TiC, Co 11 43 40 55 20 I >1000 >2800
Baked Carbon 1.6 20000 iRoom temp. 2 0.7 5
Electrogr. brush 1.7 1000to6000 Roomtemp. 30 0.3 1 to 2
Graphite 1.8 700 Room temp. 160 0.3
Graphite 550 500 °0
Graphite 700 1000 °0 55
Graphite 800 2000 oc 40
Graphite 3000 °0 20
Graphite 3500 oc 10 I I I I I I
~
1 Cf. Fig. (V,3). Note: Cu&RIE, HAliiiSTER and MoPHERSON [Z] call attention to the fact that for graphites and electrographiteR ooCI
(treated at > 2800 °0) e~ = 1 to 1.2. 10-a V2 /deg at room temperatnre.

438 Appendices
Table (X,2)
r
Bulk at room temperature heat of evapo- LORENZ con-
Materia.l c
heat capaclty
u,
softenlng
I u,
meltlng
I Uu
boüing1
rizatlon at stants when
normal boil- much deviatlng
ing point from 2.4 · 10-•
10' JJm3 •K voltage 1010 JJm3
Ag 2.5 0.09 0.37 0.75 2.5

Al 2.4 0.1 0.3 2.8
Au 2.5 0.08 0.43 0.9 3.1
Cu 3.4 0.12 0.43 0.8 4.3
Fe 3.6 0.21 0.6 4.9 2.1
Ir 2.9 8.8
{2.6auo•c
Mo
3.4 at 1 500 •c 0.25 0.76 1.1 5.6
Ni 3.9 0.22 0.65 5.7 2
Pb 0.12 0.19 t.O
Pd 2.7 0.67 1.3 4.7
r-6ato•c;•
Pt 2.8 0.26 0.71 1.3 i\.3 3.0 at 1 ooo•c
3.1 at 15oo•c
Sn 1.65 O.o7 0.13
r.7at20°C
w 3.5 at 1400 •c 0.6· 1.1 2.1 8.1 3.6
3.9 at 2100 •c
Zn 0.1 0.17 1.3
Cu, 40Zn 3.2 0.2
Cu, 40Sn 3 0.15
Stainless
steel 3.9 0.6
WC,Co 3 0.6
r.3at 20°C
Electro- 2.5 at soo•c (11)
graphlte3 3.5 at t ooo •c
s.s at 2 ooo •c
1 Cf. end of § 13.
2 According to HoPKINS [1] and [2].
8 Values for the density 1.8 · 103 kg/m3.
§X. 439
Appendix X
This appendix contains such physical constants characterizing con-
tact materials that are often used for computations on electrical con-
tacts. Considering that the experimental data which serve as basis for
such calculations· usually have rather wide uncertainty Iimits, the
tables contain averages of values which are met in practice. No difference
is marked between metals of high and low purity.
The data of the tables are expressed in mksa units and fit directly
into the equations of this book, without conversion factors.
In accordance with the choice of the units the contact hardness H
is expressed in 1()8 Nfm2 • BRINELL and VICKERS hardness numbers,
when given in kgfmm2 are to be multiplied by 107 in order to express
Hin Nfm2 •
Fig. (111,10) illustrates the thermal conductivity of some metals as
a function of the temperature. Several metals have a slightly higher
heat conductivity (up to 10%) at 600 °C than at room temperature.
It will be noticed that heat capacity, c, and evaporation heat, r,
are referred to unit volumes1 in this book, and that this choice provides
simplicity to the respective formulas. Melting and boiling tempera-
tures are cited from American Institute of Physics Handbook 1963,
except values marked by *; these are cited from KoHLRAUSOH: Prak-
tische Physik, 2 (1956).
Table (X,3) presents material constants or coefficients, Im, V"''
y for arcs at opening and closing contacts respectively, which can be
used for rough computations on the material transfer in arcs with
normal electrodes; cf. §§ 56 and 64. The values of r for anode loss in
cloliling contacts are valid for I < 100 A, those of r for cathode loss in
opening contacts are valid for I < 15 A. As to r at higher currents see
Fig. {56.09).
The Ietter n denotes normal atmosphere at roöm temperature with
a humidity of k = 0.35 to 0.60 (i. e., 35 to 60%). A humidity different
from this is always noted, e. g., k = 0.8. N 2 denotes dry nitrogen with
an oxygen content of < 0.01%.
Due to the considerable spread of observations on V m, the values
given have uncertainty Iimits of ± 0.5 V.
At closing contact, y is independent of the atmosphere because
the short arc bums in metal vapor.
1 This involves a dependence on density deviations. The values for graphite
refer to the density of 1.8. tos kg/ms.

440 Append.ices
Table (X,3). Mean values of coeffieients characterizing the aro material transfer on
making or breaking contact during a lang series of operations
V., and I., in oase of eleotrode diameter > cathode spot diameter
y, anode loss at closing
Material Atmosphere Vm Im
"
cathode 1> 20A 1 < IOA
loss at during
opening, floating,
I< t5A'
volt amp 10-10 m•tcoul
c n 20 0.03 2 to 6 0.8
Al n 14 - 10 5
Ca No 10 - - 85
Cr n 16 - - 1.5
Fe, pure n 13 to 15 0.35 to 0.55 2 4
Ni N• 14 0.5 2 2.3
Ni, Carbonyl n 14 0.4 to 0.5 1.4 -
Cu n 13 0.43 0.8 6
Cu h = 0 and 0.8 13 0.6 - -
Cu No 14 0.6 0.08 -
Cu Ho 18 or 32 1.3 1 -
Cu oil ·15 0.4 LI -
Zn n 10.5 (0.1) 5.4 6
Mo No 17 0.75 0.3 1
Mo Ho - - L4 -
Rh n 13 (0.35) 0.5 -
Ag n 12 0.4 0.4 6 to 18 10
Ag h = o and 0.8 12.5 0.55 - -
Ag 120 •c 12 0.25 0.1 -
Ag Na 13 0.8 0.14 -
Ag H• 21 - 1 -
Ag oll 15 0.25 0.9 -
Cd No (H) (0.1) 12 29
Sn n 13.5 - - 31
Sb
W, pure
n 10.5 - (150) 150
4
n 15 1.0 to LI 0.04
W, pure No 16 0.9 0.04 -
Pr n 17.5 0.9 0.8 6 6
Au n and h = 0.9 15 0.38 LI 18
Au Ho 20 - 2 -
Stainless steel n 15 0.5 2 -
Stainless steel h = 0.8 15 0.2 - -
Bronze (Cu 8, Sn) n 13.5 0.31 (2) 25
Constantan n 14 (0.4) LI 11
Ag+ 40% Ni n 13 - 0.6 20
Ag+ 10% Pd n 11 0.3 0.15 15
Ag+ 10%Au n H 0.25 L1 15
Ag r (>50% Au) n 14 0.4 2 15
Widia(WC) n 14.5to16 0.65 to 1 2 2
Widia(WC) Nt 14.5 1.15 0.2 -
Pr-Ir n 201 0.74 1 9
Pr+ 8% Ni n 16 - 0.9 -
Au+ 5% Ni. n 15 0.38 2 -
Sintered W + Cu oll 18 - 0.5 -
w+ agalnst Ag- n 13 0.35 0.4 0.4
w- agalnst Ag+ n 17 0.2 -2 (3)
Ni+ agalnst Ag- n 13 - 0.3 -
Ni- agalnst Ag+ n 13.5 - o.s -
1 Cf. ITTNER et al. [2].
§X. 441
Appendix XI
This Appendix contains two copies of Diagram XI. One of them is
intended to be separated from the book and to be used according to the
+ I
I
I
\
Diagram XI
\
\
\
\ - SOVt-Vm..
\ \
'
\
\ \
\ \
\ \
\ \
\ \
\ \
\ \
\ \
\ \
:-.. \
\
\
\
''
' ' :-... -JPV+i/m
- aPV+Vm
tls
tlfl
()J
(}2 - tDVrVm
at
aos
+ -Vm
442 Appendicee
directions which are given in§§ 50C and 60.A. In order to facilitate the
separation the sheet is perforated.
Directions for t'IIR. aeparation of t'IIR. copy on M.avy pwper: Cut the dia-
gram along the two straight ünes :
1. Vertically through the centers of the crosses.
2. Horizontally through the center of the lower cross and the da.sh
beside vm.
The da.shed lines refer to arcs between carbon electrödes.
Diagram XI
+ I
\
\
\
\
\
\
\
\ \ -SOVt-1/m,
\ \
\
\
\
\
' \
\
\
\ \
\ \
\ \
\ \ - ' IOVt-Y""
\ \ Im~2 Amp
~ \
\
\
'' '
'' :-....
-JOV +It/ n
- aoV+I/m
((S
0.~
O.J
ti2 -tov~vm
at
aos
+ -Ym
Holm , Elect rlc Conta cts, 4th. Ed. Sprln aer·Ver!&~~ , Berlln /HeidelbcrgfNew
York
.Author and Iiterature index1
page
ÄDKINS, R. M. : Dry circuit contacts and the applica.ble theoretical phenomena.
Proc. Internat. Symposium on Electric Contact Phenomena. Univ. of Maine
(1961) 119
AIUMATU, H., H. KuRODA: On the substructure and the crystallite growth in
carbon. Proc. Fourth Carbon Conference, New York, Pergarnon Press 1960,
355 ..............................................•.............. . 405
ALBIN, A. L., E. BuscH: Miniaturized suppression components for relay inter-
ference reduction. Proceedings Relay Conference, Stillwater, Okla., April
1965, paper 21
ANDERSON, J. R., J. SAUNDERS: Stahle sliding connector contacts. Report,
Contract DA 36-039, sc-89150, Spt. 1963 ............................. 219
- mentioned . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
ANDERSON, 0. L.: see MALLINA
.ANDERSON, W. J.: see BISSON
.ANous, H. C.: [J] Surface illms on precious-metal contacts. Brit. J. appl. phys.
13 (1962) 58 .............................................. 46, 105, 107
- [2] Properties and behavior of precious-metal electrodeposits for electrical
contacts. Transact. Instit. Metal Finishing 39 (1962) 20 ................• 437
- [3] Platinum-metal electrodeposits as contact surfaces. Proc. Internat. Sym-
posium on Electr. Contact Phenomena. Techn. Hochschule Graz, Austria
(1964) 89 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
[4] Mechanical wear of lightly loaded precious metals. Instrument Practice
(March 1966) 241 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 437
ANTLER, M. [1] The lubrication of gold. Wear 6 (1963} 44 .................. 219
- [2] Metal transfer and the wedge forming mechanism. J. Appl. Phys. 34
(1963) 438 .............................................•.......... 202
-, L. V. AULETTA, J. CONLEY: [3] Automatie contact resistance probe. Rev.
Scientif. Instruments 34 (1963) 1317
- [4] Processes of metal transfer and wear. Wear 7 (1964) 181 .......... 202, 203
- S. J. KRUMBEIN: [5] Contact properties of conductive hard metals and of
tin nickel plate. Proc. Seminar on electr. contacts. Univ. of Maine (1965)
103 ............................................................. . 365
- [6] Wear and friction of the platinum metals. Platinum Metals Rev. 10
(1966) 2 .........•••...................•.........•........... 202, 219
ARCHARD, J. F.: [1] Contact and ruhhing of flat surfaces. J. Appl. Phys. 24
(1953) 981 ........................................................ 234
- [2] Elastic deformation and the laws of friction. Proc. Roy. Soc. London
.A 243 (1957) 190 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31, 200
1 A single bracket [ indicates that the numbers to the right of it refer to pages
in this book. "Mentioned" means mentioned in the text without referring to any
publication. Several informative papers are included without being cited in the
text.
444 Author and Iiterature index
ARCHARD, J. F., W. HIRST: [3] The wear of metals under unluhricated condi-
tions. Proc. Roy. Soc. A 236 (1956) 397 . . . . . . . . . . . . . . . . 233, 236, 239, 241
-, W. HmsT: [4] An examination of a mild wear process. Proc. Roy.
Soc. A 238 (1957) 515 ............................................. 241
- [5] A crossed-cylinders friction machine. Wear 2 (1958) 21 ............... 236
-, W. HmsT: [6] The wear of metals. Scientific luhrication (Nov. 1958) 3
- [7] The temperature of ruhhing surfaces. Wear 2(1959) 438 ............... 231
- [8] Single contacts and multiple encounters. J. Appl. Phys. 32 (1961) 1420 .. 241
ARNOLD, S. M.: Metal whiskers. Elec. Mfg. 04 (1954) 110 ................... 417
ATALLA, M. M.: [1] Arcing of electrical contacts in telephone switching circuits.
Bell Syst. techn. J., I: 32 (1953) 1231; II: 32 (1953) 1493; III: 33
(1954) 535; IV: 34 (1955) 203; V: 34 (1955) 1081 ....................... 276
-, Miss R. E. Cox: [2] Theory of open-contact performance of twin contacts.
Bell Syst. techn. J. 33 (1954) 1373 ....................................362
AULETTA,L. V.: see ANTLER
AvRAMESCU, A.: Beiträge zur Berechnung der Kurzschlußerwärmung. Disser-
tation. Dresden 1937 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
AwEIDA, J. I.: Analysis of contact design. Proc. Seminar on electric. contacts.
Univ. of Maine (1959) II ............................................ 362
AYRTON, H., mentioned· ............................................... 281
BAILEY, ANITA T., J. S. CouRTNEY-PRATT: [1] The area of real contact and
the shear strength of monomolecular layers of a boundary lubricant. Proc.
Roy. Soc., Lond. A 227 (1955) 500 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204,. 216
- [2] Oberflächenenergie und Adhäsion in lamellaren Feststoffen. Dechema
Monographie 51, 21 ................................................ 114
BAKER, R. G.: Studies of static low voltage contacts at the Bell Tel. Lahora-
tories. Proc. Internat. Symposium on Electric Contact Phenomena. Techn.
Univ. Graz, Austria 1964, 545 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
- see COMPTON
BAKER, R. M.: [1] The effect of mercury vapor on sliding contacts. Electr.
J. 29 (1932) 64
- [2] Electrical sliding contacts. Electr. J. 31 (1934) 359 und 448
- [3] Sliding contacts. EI. characteristics. Electr. Engng. 55 (1936) 94
- and G. W. HEWITT: [4] Brush wear in hydrogenandin air. Electr. J. 33
(1936) 287 0 0 0 0 0 0 0 0 0 0 0 0 253
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0. 0 0- 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0- 0 0 0 0
- G. W. HEWITT: [5] Contact drop and wear of sliding contacts. Electr.

Engng. 56 (1937) 123 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243
BARDEEN, J. M., B. S. CHANDRASEKHAR: Peltier heat at the interface between
a metaland its melt. J. Appl. Phys. 29 (1958) 1372 . . . . . . . . . . . . . . . . . . . . . 353
BARKAN, P., E. J. TuoHY:Acontact resistance theory forroughhemispherical
silver contacts in airandin vacuum. IEEE Trans. PAS-84 (1965) 1132 .. 24, 37
BARLOW, H. M.: Experiments on the apparent deviation from Ohm's law for
metals at high current densities. Phil. Mag. 9 (1930) 1041 ................. 71
BARNES, W. A.: Progress in "C~dwelding". Wire and wire products 30 (1955) 219 159
BATEL, W.·: Über die Haftfähigkeit trockener feinkörniger Stoffe. Chemie-Ing.-
TechH. 31 (1959) 343 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
BATTAGLIA, D.: L'hexafl.uorene de soufre, nouveau moyen d'extinction de l'arc
dans les disjoncteurs a haute tension. Bull. Soc. Franc. Electr. Ser. 83
(1962) 523 0 0 0 0 0 0 0 0 0 0 0 0 0 295 0 0 0 0 0 0 0 0 0 0 0 0 0-0 0 0 0 0 0 0 0 0 0 0 0 0 0--0 0 0 0 0 0 0 0 0 0 0 0 0
BAUER, A.: [1] Zur Feldbogentheorie bei ~lten verdampfenden Kathoden.

I and II. Z. Phys. 164 (1961) 563 and 165 (1961) 34 ..................... 427
Author and Iiterature index 445
BAUEB, A.: [2] The arc cathode. Discharge and plasma physics, Univ. New
England, Australia 1963, 319 ....................................... 423
- [3] Der Mechanismus vor Bogenkathoden. Neuere Ergebnisse. Phys. Ver-
hand!. DPG 4 (1964) 343 ........................................... 423
- [4] Der Einfluß des individuellen Ionenfeldes auf die Thermo-Feldemission.
Habilitationsschrüt, Karlsruhe 1964 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 428
BAUEB, R. M.: see R. HoLM
BAYER, R. G., D. D. ROSHON: Some design considerations for low voltage
contacts. Microelectronics and ReHability 4 (1965) 131 .............. 175, 242
BEATTIE, R. W.: Palladium plating on telephone plugs and sockets. Platinum
Metals Rev. 6 (1962) 52 ............................................ 363
BECKEB, J. A.: The life history of adsorbed atoms, and molecules. Ann. N.
York Acad. Sei. 68 (1954) 723 • . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
BEDWOBTH, R. E.: see PlLLING
BEILBY, Sir GEORGE: Aggregation and fiow of solids. London: Macmillan
1921 ....................•......................................... 218
BELAllliN, M.: mentioned . . • . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .• . . . . . . . 302
BENEDICH:, M. H., C. C. MEBCHA.NT: Interference studies, investiga~ions and
tests concerning electric arc-producing devices. Navy Contract NO bsr
72681, Nov. 30, 1960 ..•............................................ 272
BETHE, H.: see SoMMERFELD
BE'l"l'EBIDGE, W., J. A. LAIBD: The wear of electrical contact points. J. Instn.
electr. Engrs. 82 (1938) 625 ..........•.............................. 338
BnmEB, K.: [1] Grenzbedingungen der Lichtbogenbildung bei Kommutierung.
ETZ-A 81 (1960) 558 . . . . . . .• . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253, 265
- [2] Beitrag zur Dynamik des Kontakts Kohlebürste- Lamelle. ETZ-A 82
(1961) 46 ................................................•......... 260
- [3] Verschleiß von Kohlebürsten durch Kommutierungslichtbögen. ETZ-A
86(1964)248
BISSON, E. E., R. L. JoHNSON, M. A. Sww:KEBT, D. GoDll'BEY: [1] Friction,
wear and surface darnage of metals as a:lfected by solid surface films.
NACA, TN 344. May 1955 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227
-, R. L. JOHNSON, W. J. ANDEBSON: [2] Friction and lubrication with solid
lubricants at temperatures to 1000 °F with particular reference to graphite.
Inst. Lubricat. Engnrs. Conference on Lubrication, London 1957 ......... 227
-, see GoDFBEY
BiTT~. H.: Schwankungserscheinungen bei der Elektrizitätsleitung in Fest-
körpern. Ergebn. d. Exakt. Naturwiss. Sl (1959) 84 ...........•.......• 270
BLAKE, B. C.: Summary report of the ASTM Section G, contact field tests.
Proc. Internat. Symposium on Electric Contact Phenomena. Techn. Univ.
Graz. Austria (1964) 531 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
BLOK, H.: [1] Theoretical study of temperature rise at surfaces of actual con-
tact under oilinesa lubrication conditions. Gen. discuss. on lubrication and
lubricants. Proc. Instn. mech. Engrs., Lond. 2 (1937) 222 ......... ~...... 94
- [2J Fundamental mechanical aspects ofthin-film lubrication. Ann. N. Y.
Acad. Sei. OS (1951) 779
BwK, J.: see VAN VLIET
BöDEFELD, Th., H. SEQUENZ: Elektrische Maschinen, Wien: Springer 1962 ... 260
BoEHNE, E. W., M. J. JANG: Performance criteria of D. C. interruptors.
Trans. AIEE 66 (1947) 1172 ........... .' ..............•.............. 297
BoESCHOTEN, F., E. T. M. VAN DEB HELD: The thermal conductance of con-
tacts between Aland other metals. Physica XXIll (1957) 37 ............ 195
BoLLMANN, W., J. SPREADBO:ROUGR: Action of graphite as a lubricant. Nature

186 (1960) 29 . .• • . .• . • . .• . • . .• . . . . . .• • • . . . . . . . . . . . . . . . . . .• . .• • • . . . 405
BONDI, W.: Beiträge zum Abnutzungsproblem. Berlin: VDI-Verlag 1927 .... 237
Bo:ROREBT, L.: [1] Über die Auswahl von Kontaktwerkstoffen für Fernsprech-
verbindungen. Nachrichtentechn. ZS. 14 (1961) 175 •...............•... 170
-, K. L. RAu: [2] Verglt>ichende Untersuchungen verschiedener Frittarten.
Nachrichtentechn. ZS. 1o (1962) 119 . . . • . .• . • . • . . . . . . .• . . . . . . . . . . .• • . 148
- [3) tiber die rechnerische und experimentelle Ermittlung einer Funken-
löschung. NTZ (1963) 304 . . .• . . . . . . . .• • . . . . . .• . . . . . . . . . . . . . . . . . . . . . . 329
BOWDEN, F. P., K. E. w. RIDLER: [1] Phys. properties of surfaces. m. Tbe
surface temperature of sliding metals. The temperature of lubricated sur-
faces. Proc. Roy. Soc., Lond. A. 104 (1936) 640 . .• • . . . . . .• . . . . . . . . . .• . • 230
-, D. TABOR: [2] The area of contact between stationary and between moving
surfaces. Proc. Roy. Soc., Lond. A.169 (1939) 391 . . • • . . .• . . .• . .• . .• . .• • • 46
-, A. J. W. MooRE, D. TABOR: [3] The ploughing and adhesion of sliding
metals. J. Appl. Phys. 14 (1943) 80
-, D. TABOR: [4] The friction and lubrication of solids. Oxford: Clarendon
Press, 1950. Second edition brought up to date by accounts of recent work
in addenda, 1954 ...••.••.... 153,157,211,213,217, 219ff., 228,230,231, 238
-, W. R. TR:ROSSELL: [.5] Adsorption of water vapor on solid surfaces.
Proc. Roy. Soc., Lond. A 209 (1951) 297. Cf. Nature 167 (1951) 601 and
1037....... . . . . . . . . . • . .• . .• . . . .• . . . . . . . . . . .• . . . . . . . . . . . . . . . .• 114, 115
-, A. C. MoORE: [6] Physical and chemical adsorption of long chain com-
pounds on radioactive metals. Trans. Faraday Soc. 47 (1951) 900 .•.••.••. 217
- [7] Recent studies of metallic-friction. Engineer (Dec. 24, 1954) 886 and
(Dec. 31) 902 • . .• . .• . • • • . • . .• • . . .• • . .• . . . . . . . . • . . . . .• . • . . . . .• 230, 231
-, E. H. FluTAG: [8] Some recent experiments in friction. Nature, Lond.176
(1955) 944 • . . . . . . . . . . . . . . . . . . . . . . . . . . .• . . . . . . . . . .• • • . . . . . . . . . . . . .• 231
- [9] Adhesion and friction. Endeavour 16 (1957) 5 . . . . . . . . . . . . . . • . . . 230, 231
- [10] A review of the friction of solids. Wear 2 (1958) 333
-, J. B. P. WILLIAMSON: [11] Influence of the passage of current on the con-
tact between solids. Proc. Roy. Soc. A 246 (1958) 1 . . . .• • • • . . . . . . . . . . . . . 64
-, D. TABOR: [12] The friction and lubrication of solids, Part II. Oxford
University Press, London 1964 [157, 161,199,204,206,221,225, 227ff., 367, 372
BoWDEN,F. W: seeMACKEOWN
BoYD, J., B. P. RoBERTSON: The friction properties of various lubricants at
high pressures. Trans. Amer. Soc. mech. Engrs. 67(1945) 51 . . • • . • . • 211, 214
BoYLE, W. S., F. E. HAwoRTR: The glow to arc transition. Phys. Rev. 101
(1956) 935 ..................••............••.•.................... 315
- see GERMER
BRANDliiÜLLER, J., H. HEUMANN: Untersuchungen über das Kontaktrauschen
z.
und das Kontaktbeben. angew. Phys. 1 (1948) 139 and 1 (1949) 454 . • . . 272
BRAUN, A., G. Busen: "Ober den Mechanismus spannungsabhängiger Wider-
stände. Helv. phys. Acta 1ö (1942) 571 • • . .• • . . . .• . . . . . . .• . • 131, 132, 189
BRERlll, K: "Ober Prellschwingungen bei elektrischen Kontakten. AEG-Mitt.
(1951) 302 ••............................•..••••••..............•.. 302
BRENNAN, R. D.: The interlayerhinding in graphite. J. chem. Phys. 20 (1952) .
40 ................•••••••.....•..•................•....•.......•. 222
BRENNER, S. S.: Metal whiskers. Scientif. American 203 No. 1 (1960) 64 ..... 417
BRINELL, J. A.: [1] EinVerfahren zur Härtebestimmung nebst einigen An-
wendungen desselben. Baumaterialienkunde o (1900) 276, 294, 317, 364, 392,
412 .. •.• .......................................................... 376
Author a.nd Iiterature index 447
BBINELL, J. A., G. DILLNER: [2] Die Brinellsche Härteprobe und ihre prak-
tische Verwertung. Internat. Verbindung für Materialprüfung. Brüsseler
Kongreß 27 (1906) 1 ............................................... 376
BRISTOW, J. R.: see HEATON
BBODY, T. P.: Nature of the valley current in tunnel diodes. J. Appl. Phys.
33 (1962) 100 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
BROWNE, T. E. Jr.: [1] An approach to mathematical analysis of a. c. arc .
extinction in circuit breakers. Pwr. apparatus and syst. (Febr. 1959) 1508 .. 292
- [2] Extinction of long a. c. arcs. Proc. Internat. Symposium on electric
contact phenomena. Univ. of Maine (1961) 399 ........................ 295
-, see LINGAT
-, mentioned . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 292, 294
BBUNNER, J., H.IIAMMEBSCHMID: Über die elektrische Leitfähigkeit gepreßter
Graphitpulver. Z. Elektrochem. 40 (1934) 60 ..................... 191, 192
BRYANT, P. J., P. L. GUTSHALL, L. H. TAYLOR: A study of mechanisms of
graphite friction and wear. Wear 7 (1964) 118 ..... 28, 156, 222, 225, 405, 406
BUCKLEY, D. H., R. L. JoHNSON: Mechanism of lubrication for solid carbon
materials in vacuum to 10-9 mm of mercury. ASLE Transact. 7 (1964) 91
BURGESS, R. E., H. KBoEMER, J. M. HousTON: Corrected values of Fowler-
Nordheim field emission functions v(y) ands(y). Phys. Rev. 90 (1953) 515 .. 130
BUR.KIIARD, G., H. M. WEDELL, H. HILBIG [1]: Zur Kontaktmaterialwanderung
unter dem Einfluß des Lichtbogens und deren Messung mittels radioaktiver
Isotope. Wiss. ZS Elektrotechn. 2;2 (1963) 92 ........................• 195
-, E. WEBNER [2]: Über das Verhalten von Kontaktlenkverbindungen bei
großen Strömen. Kontakte in der Elektrotechnik. Berlin: Akad. Verlag
1965,252
BUBLEY, C. E.: Silver plated aluminium bus conductor. Pwr apparatus and
Syst. (Dec. 1958) 1024 . . . . . . . . . . . . . . . . . . . . . . . . . . . . .• . . . . . . . . . . . . . .• 165
BUBSTYN, W.: [1] Neue Beobachtungen an SilberkontaktenETZ 62 (1941) 149 282
- [2] Elektrische Kontakte und Schaltvorgänge. Berlin: Springer 1950,
enlarged 1956 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 328
BURTON, R. L.: see CLABK
BURWELL, J. T., J. KAYE, D. W. VAN NYMEGEN, D. A. MoRGAN: [1] E:ffects
of surface finish. J. appl. Mech. 8 (1941) A-49 .......................... 417
-, C. D. STRANG: [2] The increment friction ooefficient- a nonhydrodynamic
component of boundary lubrication. J. appl. Phys. 20 (1949) 79 . . . . . . . . . . 211
-, C. D.·STRANG: [3] On the empiricallaw of adhesive wear. J. Appl. Phys. 23
(1952) 18 .................................................... 237, 238
BusCH, G.: soo BRAuN, A.
BusCH, H.: mentioned . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
CABBERA, N., N. F. MoTT: Theory of the oxidation of metals. Rep. Phys. Soc.
(Progr. Phys.) 12 (1948-49) 163 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . • . . . 107
CAMPBELL, W. E., U. B. THoMAs: (1] The electrolytic reduction method for the
analysis of films on metal surfaces. Trans. electrochem. Soc. 76 (1939) 303;
likewise: Bell Teleph. Syst. Techn. Publ. B-1170 ....................... 105
- [2] Use of statistical control in corrosion and contact resistance studies.
Trans. electrochem. Soc. 8 (1943) 377; likewise: Bell Teleph. Syst. Tech.
Publ. B-1350 (1942) ................................................ 360
-,RosE KozAK: [3] Studies in boundary lubrication. The wear of carbon
brushes in dry atmospheres. Trans. Amer. Soc. mech. Engrs. 70 (1948) 491
- [4] Solid lubricants. Lubr. Engng. (Aug. 1953) 195
CAMPBELL, W. E.: [.5] Lectures on tarnishing, friction and wear in contacts.

Proc. Seminar on electric contacts, Penna. State University (June 1956)
108, 110
CANNON, P.: Mechanism of the vapor lubrication of graphite. J. Appl. Phys.
35 (1964) 2928 ............................... .................. Cf. 226
CARSLAW, H. S., J. C. JAEGER: Conduction of heat in solids. Oxford: Clarendon
Press 1948 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
ÜASSIE, A. M.: [1] Introduction to the theory of circuit interruption, in Circuit
Breaking, edited by H. Trencham, London, Butterworth, 1953, 46 ....... 292
- [2] Some theoretical aspects of arcs in nozzles under forced convection.
Ionization Phenomena in Gases, edited by H. Maecker. N-Holland Publ. 6,
Amsterdam (1962) 1837
- mentioned . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 292
C.ATHCART, J. V.: see YouNG
CHA.IKIN, S. W.: [1] Study of effects and control of surface contaminations on
electrical materials. Stanford Res. Inst. PU3145, June1961 [28, 52, 111,
118, 173
-, J. R. ANDERSON, G. J. SANTOS: [2] Improved probe apparatus for measur-
ing contact resistence. Rev. Scient. Instr. 32 (1961) 1294 . . . . . . . . . . . 118, 143
- [3] Organic contact deposits under rubbing and non rubbing conditions.
Proc. Internat. Symposium on Electric Contact Phenomena. Univ. ofMaine
1961 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118, 173
- [4] Electrical-contact failure caused by surface contamination. Electro-
Technology, Aug. 1961
- [.5] Inhibition of frictional polymer formation on rubbing contacts. Internat.
Conf. on Electromagn. Relays, Sendai, Japan 1963, paper A-17
CH.Ao, B. T.: see CuusiNG
ÜH.ASTON, J. C.: Oxidation of palladium. Platinuni Metals Rev. 9 (1965) 126 . . 112
CmARENZELLI, R. W., B. C. HENRY: Lubricating separable electric conta.cts:
tarnish prevention. ASLE Trans. March 1965 . . . . . . . . . . . . . . . . . 175, 219, 220
ÜHIVIAN, .J. S.: seeHARTMAN ·
ÜHOPRA, K. L.: Avalanche-induced negative resistance in thin oxide films. J.
Appl. Phys. 36 (1965) 184 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
CHOUDIIURI, S. P. R., K. P. PILL.AI: An analysis of.resistance-characteristics
of static electric contacts. IEEE Trans. IGA-1 (Jan., Febr. 1965) 9 .... 87, 136
CHRISTENSEN, C. J., G. L. PE.ABSON: Spontaneous resistance fiuctuations in
carbon Inicrophones and other granular resistances. Bell Syst. techn. J. 15
(1936) 197 . . . . . . . . . . . • . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
CL.ARE, C. R. : mentioned . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 358
ÜLARK, G. L., A. C. ECKERT, JR., R. L. BURTON: Commercial andexperimental
carbon blacks. Industr. Engng. Chem. 41 (1949) 201 ................ 223, 408
ÜLARK, 0. H., W. W. WooDs, J. R. WHITE: Lubrioation at extreme pressures
with mineral oil films. J. Appl. Phys. 22 (1951) 474 . . . . . . . . . . . . . . . . 211, 214
ÜLAUSING, A. M., B. T. CH.Ao: Thermal contact resistance in a vacuum environ-
ment. University of lliinois. ME Techn. Report 242-1, Aug. 1963 . . . . . 31, 196
ÜOBINE~ J. D.: [1] Gaseous conductors. New York: Mc Graw Hill1941 [276, 295,296
-, G. A. FA.RR.ALL: [2] Experimental study of arc stability. J. Appl. Phys. 31
(1960) 2296 ............................... ..............•.......... 286
-, G. A. FARR.ALL: [3] Are stability. Proc. Internat. Symposium on electric
contact phenomena. Univ. of Maine (1961) 263 ................... 286, 287
-, T. A. VANDERSLICE: [4] Electric erosion and gas evalution of vacuum arcs.
IEEE Trans. Communicat. and Electroriics 82 (1963) 240 ................ 308
ÜOBINE, J. D.: see M.ACKEOWN
Cocxs, M.: [1] The effect of compressing and shearing forces on the surface
films present in metallic contacts. Proc. phys. Soc., London B 67 (1954)
238 ...........................................•.......... 46, 202, 232
- [2] Role of displaced metal in the sliding of fl.at metal surfaces. J. Appl.
Phys. 80 (1964) 1807 . . . . . . . . . . .• . . . . . . . . . . . . . . . . . . . . . . . .• . . . . . 202, 232
- [3] The formation of wedges of displaced metal between sliding metal sur-
faces. Wear 8 (1965) 85 ..............................................·202
- [4] Shearing of junctions between metal surfa~es. Wear 9 (1966) 320 ...... 202
Co:Ml'TON, K. G., R. N. BAXER: The use of electroplated metals in static
contacts. Proc. Electric Contacts Seminar, State Univ. Pennsylvania
(1960) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
CONDON, E. U.: mentioned ............................................ 123
CoNLEY, J.: see ANTLER
CoNNoR, T. J., W. R. WILSON: Performance of electric joints utilizing new
silver coating on aluminium conductors. Power Apparatus and Systems
(August 1953) 702 •......................... :. . . . . . . . . . . . . . . . . . . . . . 165
CoNTIUs, E. : Der Einfluß der Größe des Druckes und der Fläche auf den
Kontaktwiderstand. Dissertation Dresden 1929 . . . . . . . . . . . . . . . . . . . . . . . . 46
CoNWELL, ESTHER: Basic theory of the tunnel diode. The Sylvania Technolo-
gist 12 (1959) 121 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
CooPER, R. I. B., J. PuLLEN: The erosion of tungsten and rhenium contacts
by arcing and by spark breakdown. E. R. A. Report U;T 144 (1961) ...... 109
CoRNELIUs, D. F., W. H. RoBERTS: Friction and wear of metals in gases up
to600°C.ASLEPaper60LG-9,1960 ....................... 237,239, 240
CosTELLO, F. A.: see FRrnn
ÜOUGBLIN, J. P.: Heats and free energies of formation of inorganic oxides.
Bull. 542 of Bureau of Mines. Washington: 1954 ....................... 113
CoURTNEY-!'RATT, J. S., E. EISNER: Contact of metallic bodies. Engineering
(Febr. 7. 1958)
- seeBAILEY
Cox, R. E.: see ATALLA
CRoucn, D. W.: see LEE
CURR!E, L. M., V. C. IIAMisTER, H. G. MAcPHERSON: The production and
properlies of graphite for reactors. Internat. conference on the peaceful
uses of atomic energy. 8/P/534, USA (June 1955) ...................... 437
CURTIS, A. M.: Contact phenomena in telephone switching circuits. Bell Syst.
techn. J. 19 (1940) 40; likewise: Electr. Engng. 59, Transact. (1940) 360 .. 313
DALLAs, J. P., T. R. STUELPNAGEL: Control of metal build-up in minimum

pressure sensitive contact systems. Applications and industry (Jan. 1954)
398 ......................................................... 355, 363
DANIEL, S. G.: The adsorption on metal surfaces of long chain polar com-
pounds from hydrocarbon solutions. Trans. Faraday Soc. 47 (1951) 1345 .. 217
DAVIDSON, P. M.: [1] The theory of the THOMSON effect in electrical contacts.
Proc. Inst. Electr. Engrs. 00, Part I (1949) 293 . . . . . . . . . . . . . . . . . . . . . . . . 61
- [2] The growth of the liquid bridge in an electrical contact. Brit. J. Appl.
Phys. 5 (1954) 189 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 339
DEAOON, R. F., J. F. GooDMAN: Lubrication by lamellar solids. Proc. Roy.
Soc. A 243 (1958) 464 .............................................. 227
DEBOER, J. H.: The dynamical character of adsorption. Oxford, Clarendon
Press 1953 • • . . . . . . . . . . . . . . . . . . . . . . . .• . . . . . . . . . . . . . . . . . . . . . . .• 103, 430
29 U Electric Contacta, 4th Ed.
450 Author and Iiterature i.Jidex
DEHN, G., W. LA.NGSDOBFF: [1] 100 Jahre Telephon. Frequenz-Sonderausgabe

16 (1961) 9 ............•.....•........•.......•................... 180
Th:CKINSON, R. C.: High-power "De-ion" air oircuit breaker for central-station
service. AIEE. Trans. 68 (1939) 421 .....•............................ 296
DnmL, J. E.: see SHOBERT
DIES, K.: see MA.n.XNDER
Th:ESSELHORST, H.: [1] Vber das Problem ei.Jies elektrisch erwärmten Leiters.
Ann. Phys. Lpz. 1 (1900) 312 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
Dm'IRICH, !SOLDE, E. RüCHARDT: [1] Fei.Jiwanderung an Abhebekontakten. Z.
angew. Phys. 1 (1948) 1 . . . . . . . . . . .• . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71, 339
- [2] Thermospannungen an Platinkontakten. Z. Naturforsch. 4a (1949) 482
- [3] Messung des Widerstandes dünner isolierender Schichten zwischen Gold-
kontakten im Bereich des Tunneleffektes. Z. Phys. 182 (1952) 231 ........ 126
- [4] Versuche zur Supraleitung an Kontakten. Z. Phys. 188 (1952) 499 .... 129
- [5] Untersuchungen zum kurzen Abreißbogen an schwach belasteten Kon-
takten. Ann. d. Physik 7. Folge 4 (1959) 167
-, M. HoNRATH-B.ARKHAUSEN: [6] Zur Bildung widerstandserhöhender Beläge
organischen Ursprungs auf elektrischen Kontakten, Z. angew. Physik 11
(1959) 399
- [7] Untersuchungen des Materialtransportes an elektrischen Abhebekontak-
ten mit Hilfe radioaktiver Isotope. Z. angew. Physik 12 (1960) 538 [118, 341ft.
- [8] Investigations with the radioactive tracer method i.Jicluding material
transfer measurements. Proc. Internat. Symposium on Electric Contact
Phenomena. Univ. of Mai.Jie (1961) 215 ............................. 341ft'.
- [9] Tunneleffekt an supraleitenden Kontakten. Proc. Internat. Symposium
on Electrio Contact Phenomena. Techn. Univ. Graz. Austria (1964) 159 ... 129
Dn.woRTH, Miss C. C.: The infl.uence of surface films on the electrioal behavior
of contacts. Proc. phys. Soc., Lond. 60 (1948) 315 ......••••••••........ 131
DIXON, H. E., H. G. TAYLOR: Resistance Welding. Sheet Metals lndustries 80,
No. 313 (May 1953) 385 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
DoLAN, W. W., W. P. DYKE: Tamperature and field emission of electrons from
metals. Phys. Rev. 96 (1954) 327 ........................... : . . . 131, 427
DoWNES, G. H.: see KoHMAN .
DREn us, L. A. : Die Stromwendung großer Gleiohstrommaschi.Jien. · Berlin:
Springer 1929, and with the same title in Acta Polytechnioa 146, also IVA
Handlingar 212 (1954) . . . . . . • . . . . . . . . . . . . . . .• • . . . . . . . . . . . . . . .• 260, 266
DUMBAULD, L. D.: Dry reed switches. Control Engi.Jieering (July 1963) 75 . . 171
DüRRWÄCHTER, E., W. MERL: Der Einfluß der Legierungsstruktur auf die
Stoffwanderung von Schwachstromkontakten. Proc. DAW-Tagung "Kon-
takte iJi der Elektrotechnik" Berli.Ji: Akademie-Verlag (1965) 53 ........• 346
DWYER, J. J. 0.: The theory of dielectric breakdown of solids. Oxford, Claren-
don Press (1964) ....................•...•••. .' ....•..•.......••.•..• 135
D'YACHENKO, P. E., N. N. TOLKA.CHEVA, G. A. ANDREEV, T. M. KABPovA: The
actual contact area between touching surfaces. Library of Congress Catalog
Card Number 64--13145 (1)}64) Russian paper published 1963 . . • . . • • • . • • • 32
DYsoN, J;, W. HmsT: The true contaot area between solids. Proc. Phys. Soo.,
Lond. B 47 (1954) 309 • . .• • • . . .• • . . .• . • . . . . . . . . . . . .• . • • . . . . . .• • . . . . . 35
EOXERT,A. C.: see CL.um.

EDELS, H., Y. ETTINGER: Are i.Jiterruption and thermal reignition. Inst. Electr.
Engin. Paper 3769 S (Febr. 1962) ...... , . . . . . . . . . . . . . . . . . . . . . .• • 295, 296
- see WHITTAKER
Author and literature index 451
EDEU!, H.: mentioned . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 292
EDISON, T. A.: mentioned . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
EGAN, T. F., A. MENDIZZA: Creeping silver sulfide. J. Electro-chem. Soc. 107
(1960) 353 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
- see HERMANCE
Emu.!CH, G.: [1] Molecular processes in adsorption on metals. Trans. 8th
National Vacuum Symposium. New York, Per~amon Press, 1962 [103, 104, 352
- [2] An atomic view of adsorption. Brit. J. Appl. Phys.19 (1964) 349 . . 103, -104
EINSELE, A.: Ein neuer Expansionsschalter. Siemens ZS 39 (1961) 747 .....• 295
ExxERB, G. J., A. FARNEB, R. KLÄUI: Der Thomsonkoeffizient von Metallen
bei hohen Temperaturen. Albiswerk-Berichte 1 (1949) 20 ............•... 347
ELuoTT, S. J.: Evaluation of solderlese wrapped Connections for central office
use. Bell System Techn. J. 38 (1959) 1033 ............................. 159
ELSEY, H. M.: [1] Treatment of high-altitude brushes by application of metallic
halides. Trans. AIEE 64 (1945) 576 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227
Ema:TT,D.F.: see R.HoLM
EPPR:ECHT, G. W.: Current fiuctuation phenomena Ül current-carrying sliding
contacts. J. Appl. Phys. 29 (1954) 1473 ............................•... 272
ERDMAN-JESNITZER, F.: Metallphysikalische Grundlagen der Preßverschwei-
ßung. Aluminium 33 (1957) 730 . . . . . . . . . . • . . . . . . . . . . . . . . . . . . . . . . . . . . 155
ERB:, A., H. FINKE: Über die mechanischen Vorgänge während des Prellans
einschaltender Kontaktstücke; Über das Verhalten unterschiedlicher Kon-
taktwerkstoffe beim Einschalten prellender Starkstrom-Schaltglieder.
E.T.Z.-A (1965) Heft 5, 129 and Heft 9, 297 respectively ................ 303
ESAKI: mentioned . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
ETTINGEB, Y.: see EDELS
EvANs, U. R., H. A. Ml:LEY: [1] Measurements of oxide films on copper and
iron. Nature 139 (1937) 283 ........................•................ 105
- [2] Anintroduction to metallic corrosion. London: Arnold and Co.1948 [104, -115
- [3] Metallic oxidation. Research 6 (1953) 130 ...................•.....• 116
- mentioned ........................................................ 115
F AIRWEATHER, A.: [1] The closure and partial separation of a metallic contact.
J. Inst. Electr. Engrs., Lond. 92 (1945) 301 . . . . . . . . . . . . . . . . . . . . . . . . 46, 50
- [2] The behavior of metallic contacts at low voltages in adverse environ-
ments. Proc. Inst. Electr. Engrs., Lond. 100 (1953) 174
-, D. G. M. SHIRLEY, R. E. FuDGE: [3] Contact closure. Internat. Symposium
on Electric Contact Phenomena. Graz, Techn. Hochschule, Austria (1964) 40 [377
FARRALL, G. A., J. M. LAF1!ERTY, J. D. COBINE: [1] Electrode materials and
their stabilit.y characteristics in the vacuum arc. IEEE Trans. Communi-
cation and Electronics 82 (1963) 253 ............................. 286, 289
-, J. D. CoBINE: [2]Stability of arcs in gases. J. Appl. Phys. 36 (1965) 53 [285, 286, 287
FARNEB, A.: see EKKERS
FENECH, H., W. M. RoHSENOW: Prediction ofThermal conductance of metaJ.Jlc
surfaces in contact. ASME paper No. 62-HT-32 Sept. 1962 . . . . . . . . . . . . . . 195
FENG, I. Mnm: [1] Metal transfer and wear. J. Appl. Phys. 23 (1952) 1011 [44, 206
-, B. G. RIGHTMIRE: [2] The mechanism of fretting. Amer. Soc. Lubr. Eng.
Meeting (April1953) . . .• . . . . . . . . . . . . . . . . . . . . . . . . .• . • . • . . . . .• • . . . . . . 206
-, [3] The infiuence of surface activity on friction and surface damage. Wear
4 (1961) 269 ........•..............................•............•• 211
FEssENDEN, R. A.: mentioned . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . • . . . . . . . 60
FINCH, G. I.: The sliding surface. Proc. phys. Soc. Lond. 63 B (1950) 473 [218, 220
29E
FINK, H. P.: [1] Untersuchung über die Entstehung .von Kontaktbögen. Wiss.
Veröff. Siemens-Werk 17 (1938) 45 .......................... 279, 280, 282
-, H. KöRNER: [2] Messungen des zeitlichen Verlaufs der Erwärmung in einem
Kontakt infolge eines plötzlich einsetzenden Stromes. Wiss. Veröff. Siemens-
Werk 19 (1940) 280 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
- seeR. HoLM
F!NKE, H.: see ERK
F!NKELNBURG, W., H. MAEcKER: [1] Elektrische Bögen und thermisches
Plasma. Handbuch der Physik. XXII. Springer 1956, 254ff. [306, 421, 425ff.
- [2] Structure of matter. Berlin: Springer 1964 and Acad. Press, New York
F:rscHMEISTER, H.: see RöNQUIST
FISHER, J. 0., I. GlA.EVER: Tunneling through thin insulating layers. J. Appl.
Phys. 32 (1961) 172 ....................................... 122, 127, 129
FLOM, D. G., R. H. SAVAGE: Detection ofthin insulating films on metals. Gen.
Elect. Report No. RL-1188 (Oct. 1954). Shorter in Arm. New York Acad.
Sei. o8 (1954) 946 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111, 143
FoRSTER, G. A.: [1] Investigation of noise generated by electrical slip rings.
W ADC techn. report 54---125 (Apr. 1954) . . . . . . . . . . . . . . . . . . . . . . . . . 272, 273
- [2] Better electrical brushes. The Frontier 19 (1956) No. 3, i 7 . . . . . . . . . . . . 273
FOWLER, R. H., L. NoRDHEIM: Electron emission in intense electric flelds.
Proc. Roy. Soc., Lond. A 119 (1928) 173 .............................. 130
Fox, R. E.: see HwKAM
FoXHALL, G. F., J. A. LEWIS: The resistance of an infinite slab with a disk
electrode. Bell Syst. Techn. J. 43 (1964) 1609 . . . . . . . . . . . . . . . . . . . . . . . . . . 26
FRANCIS, G.: The glow discharge at low pressure. Handbuch der Physik XXII.
Berlin: Springer 1956,53 ........................................... 332
FRANKLIN, RosALIND, E.: [1] The structure o' graphitic carbons. Acta Cryst. 4
(1951) 253 ................................................... 407, 408
- [2] Crystalline growth in graphitizing and non-graphitizing carbons. Proc.
Roy. Soc., Lond. A 209 (Oct. 23, 1951) 196 ............................ 407
FRANz, W. : Dielektrischer Durchschlag. Handbuch der Physik XVII. Berlin:
Springer 1956, 155 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135, 146
FREIER, R.: see TonT
FRENKEL, J.: On the electrical resistance of contacts between solid conductors.
Phys. Rev. 36 (1931) 1604 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
FREUDIGER, E., E. JosT: Kinetics and thermodynamics of the internal
oxidation of silver cadmium. Proc. Internat. Symposium on Electric Con-
tact Phenomena, Univ. of Maine (1961) 177 ........................... 308
FRIEn, E., F. A. ÜOSTELLO: [1] Interface thermal contact resistance problern
in space vehicles. ARS J. 32 (1962) 237 ............................... 195
- [2] Study of interface thermal contact conductance. Document No. 64
SD 652, General ElectriQ Spacecraft Department . , . . . . . . . . . . . . . . . .• . . . 196
FRIEDRICH, R. E.: see LEEDS
FRIND, G.: [J] Über das Abklingen von Lichtbögen. ZS. angew. Physik 12
(1960) 231 and 515 ................................•.............•.. 294
- [2] Time constants of fiat ar95 cooled by thermal conduction. IEEE Trans.
PAS-84 (1965) 112.5 .•.............................................. 296
FRITAG, E. H.: see BownEN
FRooME, K. D.: The behavior of the cathode spot on an undisturbed liquid
surface of low work function. Proc. Phys. Soc., Lond. B 63 (1950) 377 .... 429
FucHs, E.: see PFISTERER
F-uCHs, L. H.: see PRIMA l{
FuDGE, R. E.: see F .AIRWHEATHER
FuJIMOTO, M.: Microstructure of metal transfer in electric contacts (Japanese).
Trans. Inst. Electr. Engrs. Japan 4 (1943) 477 ......................... 355
FUKUROI, T., Y. l\IUTO: Electrical contact resistance between metallic surfaces
subjected to various treatments. Sei. Rep. Res. Inst. Töhoku Univ. A 3
(1951) 281 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38, 46
FURMIDGE, J. E.: see TYLECOTE
G.ulow, G.: mentioned ............................... ................ 123

GANIN, E. A.: see SHLYKOW
GAULB.APP, K.: Untersuchung der elektrischen Eigenschaften des Abreiß-
bogens. Ann_ Phys., Lpz. 25 (1936) 705 ............................... 279
GEBAUER, W.: Versuche zur Entwicklung einer neuen Aluminiumkabelab-
zweigklemme für Hausanschlüsse. Elektrizitätswirtsch. 35, (1936) 694 .... 169
GERMER, L. H., F. E. fuwoRTH: [1] A low voltage discharge between very
close electrodes. Phys. Rev. 73 (1948) 1121 ....................... 117, 276
- - [2] Erosion of electrical contacts on make. J. Appl. Phys. 20 (1949)
1085 ............................... ......................... 117, 276
- [3] Arcing at electrical contacts on closure. Part I. J. Appl. Phys. 22 (1951)
955 ............................... .......................... 117, 276
-, J. L. SMITH: [4] Arcing at electrical contacts on closure. Part III. Develop-
ment of the arc. J. Appl. Phys. 23 (1952) 553 .......................... 276
- [5] Activation of contacts by organic vapor. J. Appl. Phys. 25 (1954)
332
-, W. S. BoYLE: [6] Two distinct types of short arcs. J. Appl. Phys. 27 (1956)
32 ............................... ........................... 337, 429
-, J. L. SMITH: [7] Activation of electrical contacts by organic vapors. Bell
Syst. techn. J. 36 (1957) 769
-, J. L. SMITH: [8] Organic vapor and relay contacts. Bell Lab. Record 36
(1958) 122
- [9] Physical processes in contact erosion. J. Appl. Phys. 29 (1958) 1067
- [10] Electrical breakdown between close electrodes in air. J. Appl. Phys.
30 (1959) 46
- C. D. HARTMAN: [11] Oxygen on nicke!. J. Appl. Phys. 31 (1960) 2085 .... 105
- [12] The erosion of relay contacts. Wear 3 (1960) 188 ............... 105, 276
- [13] Erosion of separating electrical contacts. Proc. Internat. Symposium
on Electric Contact phenomena, Univ. of Maine (1961) 239 .............. 315
-. A. J]. MAc RAE: [14] Oxygen-nickel structures on the 110 face of clean
nicke!. J. Appl. Phys. 33 (1962) 2923 . .. .. . . . . .. .. . . . . .. . . . . . . . . . . . . .. 103
- see LANDER
- seeWHITE
GIAEVER, I.: Electron tunneling between two superconductors. Phys. Rev.,
Letter 5 (1960) 464 .............................. ............. 127, 129
GILLHAUSSEN, HASSO Frhr v.: Elektrische Durchschlagfestigkeit von Einfach-
und Vielfach-Schichten organischer Stoffe. Dissertation Berlin 1940 ...... 146
GLOSSBRENNER, E. W., J. K. SuN: Effects of parameters on noise in miniature
sliding contacts. Proc. Seminar on electr. contacts, Univ. of Maine 1963 .. 273
GoDFREY, D., E. E. BrssoN: NACA studies of mechanism of fretting and prin-
ciples of migration. Lubricating Engng. (Oct. 1952) 241 ................. 217
- see BrssoN
GoDSEY, W. J.: Are suppression for relay contacts in d-c service. IRE Trans.
on component parts (June 1957) 36 ............................. 326, 328
29°
Author and Iiterature index
Goon, R. H. JR., E. W. MÜLLER: Field emission. Encyclopedia of Physics.

Berlin: Springer 1956, XXI, 176 ............................ 123, 130, 131
- see MURPHY
GooDMAN, J. F.: see DEACON
GoonzEIT, C. L.: see RoAcH
GouCHER, F. S.: The carbon m.icrophone: An account of some researches bear-
ing on its action. J. Franklin Inst. 217 (1934) 407; likewise: Bell Syst.
Techn. J.13 (1934) 163 .•.............•.....••.•......... 50, 51, 180, 184
GRABBE, D.: Permanent contacts in electronic assemblies. Proc. Internat.
Symposium on Electric Contact Phenomena. Techn. Hochschule Graz,
Austria 1964,492 .........•..•..••.........•....................... 176
GRANDSTAFF, 0. D.: Double contactspringsforsignal circuits, Strowger, Techn.
J. 4 (1935) 1 ....•......•.•......•...•..•••.••.•.••..•............. 362
GREEN, E. 1.: Telephone. Bell Syst. Techn. J. 37 (1958) 289 ................ 180
GREENWOOD, A. N.: see LEE
GREENWOOD, J. A., D. TABOR: [1] Deformation properties of friction junctions.
Proc. Phys. Soc. Lond. B 68 (1955) 609
-, J. B. P. WILLIAMBON: [2] Theory of temperatura-dependent conductors.
Proc. Roy. Soc. A 246 (1958) 13 .....•................•....... 61, 64, 93
-, J. B. P. WILLIA.MSON: [3] Thermal transients in graphite-copper contacts.
Brit. J. Appl. Phys.ll (1960) 389 • . . . . . . . . . . . . . .• . . . . . . . . . . . . . . . . . . . . 93
-, J. B. P. WILLIAMSON: [4] The contact of nominally :flat surfaces. Proc.
Intern. Symposium on Electric Contact Phenomena. Techn. Hochschule
Graz, Austria 1964, 24 ................•...•.........••...... 31, 32, 200
-, J. H. TRIPP: [5] Static contact of rough spheres. Burndy Corp. Res.
(25. Febr. 1965) 1 . . . . . . . . . . . . . . . • • • • • . • . . . . . . . . . . . . . . . . . . . . . . . . 31, 37
-, G. W. RoWE: [6] Deformation of surface asperities during bulk plastic ßow.
J. Appl. Phys. 36 (1965) 667
- [7] Area of contact between a rough surface and a plane. ASME--ASLE
Lubric. Conf., San Francisco 1965 . . . . . . . . . . . . . . . . . . . • . . .• . . . . . . . 31, 37
-, J. B. P. WILLIAMSON: [8] Contact of nominally ßat surfaces. Will be pub-
lished in Proc. Roy. Soc•.......•............................ 30, 31, 200
GRIFFITH, R. L.: see HoPKINS
GRISD.A.LE, R. 0.: [1] The formation of black carbon. J. Appl. Phys. 24 (1953)
1082 ........•..................•....................•..••.. ; ..... 184
- [2] The properties of carbon contacts. J. Appl. Phys. 24 (1953) 1288 • . . . . . 184
GROSCHWITz, E. : Übet die physikalischen Ursachen des Rauschans in Halb-
leitern. Phys. Blätter 11 (1955) 121 . . . . . . . . • . • . . . . . . . . . . . . . . . . . . . . . . . . 270
GuLBRA.NSEN, EARL A., W. S. WYSONG: [1] Thin oxide films on tungsten. Am.
lnst. Mining. Techn. Publ. No. 2224 (1947) . . . . • . . . . . . . . . . . . . . . . . . 105, 109
- [2] Classical theory of diffusion and the oxidation of metals. Ann. New
York Acad. of Sei. OS (1954) 830 ·
·GuMLEY, R. H.: see KEEFER
GURNEY, R. W.: mentioned . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
GuTSHALL, P. H.: see BRYA.NT
GwATMEY, J. V.: see YoUNd
H.u.sE, G., A. KLEINLE: Zu den kurzen Lichtbögen in elektrischen Kon-

takten. ZS. angew. Physik 18 (1964) 116 . . . . . . • . . .• • . . . . . . . . . . . . . . . . . . 337
·- seeHELDT
:HAGSTRUM, H. D., C. D'AMico: Productio:r~; a.nd demonstration of atom.ically
clean metal surfaces. J. Appl. Phys. 31 (1960) 715 . . . . . . . . . . . . . . . . . • . • . . 29
HALPEBN, V.: The electron structure of thin films of non-metallic materials.
Confer. on non-metallic thin films, Chelsea (1965) ...................... 187
IIALTNEB, A. J.: An evaluation of the role of vapor lubrication mechanism in
MoS 2 • Wear 7 (1964) 102 ............................................ 227
lliM:n.ToN, A., R. W. SILLABS: Spark quenching at relay contacts interrupting
D. C. circuits. Proc. Instn. Elect. Engrs. 00 Part I (1949) 64 ............. 286
IIAIIILTON, R. J.: The relation between free falling speed and particle size of
airburne dusts. Brit. J. appl. phys., Suppl. 3 (i954) S 90 . . . . . . . . . . . . . . . . 173
H.unsTEB, V.: see CURBIE
HÄMMEBLI, S. : Kontaktprobleme an schnellen Schaltsystemen. Bulletin
Schweiz. El. techn. Vereins. (1956) No. 26 ............................. 312
HAMMEBSCBMID, H.: see BRUNNE:&
IIABADA, 8., K. MANo: The effects of surface roughness on contact resistance.
Proc. Internllt. Symposium on Electric Contact Phenomena, Univ. of
Maine (1966) 37 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
HAim!s, J.: see WILLIAMBON
J!ARTMAN, C. D.: see GERMER
HARTMAN, T. E.: [1] Tunneling through asymmetric barriers. J. Appl. Phys.
35 (1964) 3283 . . . . . . . . . .• . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
- [2] Electron tunneling. Internat. Science and Technology (June 1964)
74 ................•......................................... 122, 127
-, J. S. ÜHIVIAN: [3] Electron tunneling through thin aluminum oxide films.
Phys. Rev.134 (1964) p. 1094 .. . .. . .. . .. .. .. .. .. . .. . .. 122, 127, 128, 131
HAss, G.: [1] Struktur und Optik aufgedampfter Metallschichten. Ann. Phys.
Lpz. 31 (1938) 245 ................................................. 105
-, KEHLER: [2] Korrosion, Passivität und Metallschutz. Verh. dtsch. phys.
Ges. 3. Reihe 22 (1941) 1 ........................................... 108
HAUFFE, K.: Oxydation von Metallen und Metallegierungen. Berlin: Springer
1956 .................................................... 105, 107, 110
HAWOBTH, F. E.: [1] Electrode reactions in the glow discharge. J. Appl. Phys.
22 (1951) 606 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 276, 332
- [2], [3] Breakdown fields of activated electrical contacts. J. Appl. Phys. 28
(1957) 381 ........................................................ 276
- seeBOYLE
- see GERMER
HAYNEs, J. R.: Measuring displacement of microphone contacts. Bell Labor.
Rec. 13 (1935) 337 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184
HEATON, J. L., J. R. BBISTOW, G. WHITTINGHAM, T. P. HuGHES: Frictional
properties of bearing metals. Nature loO (1942) 520 ............... 208, 211
HELDT, K., G. HAAI!IE: Der elektr. Widerstand von reinem hoch vacuum-
gesintertem Aluminiumoxyd. Z. angew. Physik 6 (1954) 157 ............. 108
HENNIG, G. R.: Screw dislocations in graphite. Science 147 (1965) 733 ...... 222
HENRY, B. C.: see CHIARENZELLI
HENTSCH, A.: [1] Materialwanderung an Ag-Abhebe-Kontakten bei relativ
niedrigen Schaltkreisinduktivitäten. Wiss. ZS der Elektrotechnik 3 (1965)
129 ..•......................................................... 341ff.
- [2] Beiträge zum Materialwanderungsverhalten von Silberkontakten. Proc.
Internat. Symposium on Electric Contact Phenomena. Univ. of Maine
(1966) 345 ......... - ................................... 339, 341ff., 346
HERMANCE, H. W., T. F. EGAN: [1] The examination of electric contacts by the
plastic replica method. A.I.E.E. Trans. Communic. and Electronics 34
(1958) 756 ........................................................ 117
HERMANOE, H. W., T. F. EGAN: [2] Organic deposits on precious metal con-
tacts. Bell Syst. techn. J. 37 (1958) 739 .............................. 117
- seeKoRMAN
HERTZ, H.: Gesammelte Werke 1. Leipzig: Barth 1895 .................... 367
HESSLER, V. P.: [1] The effect of various operating conditions upon electrical
brush wear and contact drop. Jowa State College XXXIV (1935) No. 25,
cf. Table (42.04) in R. HoLM [30] •.................•........•.•.....• 252
- [2] Electrical brush wear. Electr. Engng. 04 (1935) 1050, cf. Table (42.04)
inR. HOLM [30] ... 0 •••••••••• 0 •••••••••••••••••• 0 0 252 •••• 0 •••••••••• 0
- [3] Abrasion, a factor in electrical brush wear. Electr. Engng. 56 (1937) 8,

cf. Table (42.04) in R. HoLM [30] .................................... 252
-, HETTIOH, A.: Geometrische Dimensionen und Widerstandsrauschen. Fre-
quenz 4 (1950) 14 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 270ft'.
HEUMANN, H.: see BRANDMÜLLER
HEWITT, G. W.: see BAKER
HEYWANG, W.: Zum Mechanismus des Spannungsabhängigen Kontaktwider-
standes von Siliziumkarbid. Z. angew. Physik 8 (1956) H. 8 . . . . . . . . . . . . . 189
HIOKAM, W. M., R. E. Fox: Negative ion formation using monoenergetic
electrons. 7th ann. Gaseous Electronics Conf. New York, Oct. 1954, paper
A-1 ............................................................. . 295
IIILBIG, H. : see BURKHA.RD
HILGARDT, G.: [1] Über die Grenzstromstärken ruhender Starkstromkontakte.
ETZ-A 78 (1957) 211 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93, 96
- [2] Zusammenhang zwischen Kontaktspannung und Kontaktstellentempe-
ratur bei kurzzeitig belasteten Starkstromkontakten. ETZ-A 79 (1958) 464 . 93
HILL, R.: The Mathematical Theory of Plasticity. Oxford, Clarendon Press
1950 and 1964 .................................................... 372
HIPPEL, A. v.: Dieleetrios and waves. New York: Wiley & Sons; London:
Chapman and Hall1954 . . . . . . . . . . . . . . . .. . .. . . . . . . . . .. . . . . . . . . . 115, 135
limA:BAYASm, H. H., ToYADA, H., SHIBATA: Electrical properties of amor-
phous carbon. Tanso 3 (1953) 81 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184
HmsBRUNNER, H. : Are erosion properlies of silver and silver cadmium oxide.
Proc. Seminar on Electric Contacts. Univ. of Maine, 1962 . . . . . . . . . . . . . . . 308
HIRST, W.: [1] Surface film formation and metallic wea.r. J. Appl. Phys. 27
(1956) 1057
- [2] The mechanical wear of metals. Brit. J. Appl. Phys. 9 (1958) 125
-, J. K. LANOASTER: [3] The infiuence of speed on metallic wear. Proc. Roy.
Soc. A 259 (1960) 228 ..................................... 232, 233, 241
- [4] Wear. Metallurgical Rev.lO (1965) 145 ............................ 199
-see DYSON
HLASNIK, 1., J. SomLDER: Einfluß der Anisotropie kohlenstoffhältiger Stoffe
auf die Größe des Stromenge- und Querwiderstandes von Bürsten elek-
trischer Maschinen. Proc. Internat. Symposium on electric contact pheno-
mena. Techn. Hochschule, Graz, Austria (1964) 452
HoDGSON, B.: see MAoKLEN
HOFMANN; W., J. RuGE: Versuche über Kalt-Preß-Schweißung von Metallen.
Z. Metallkunde 43 (1952) 133 .. .. .. .. .. . .. .. .. .. . .. .. .. . . .. .. .. . 156, 158
HöFT, H.: Die wahre Berührungsfläche punktförmiger Kontakte. Proc. Internat.
Symp. on Electric Contact Phenomena. Techn. Hochschule Graz, Austria
(1964) 150 ...... 0 0 0 0 • 0 • 0 ••••••• 0 • 0 ••••••••••• 0 • • 30
• • • • • • • • • • • • • • • • •
HoLM, ELSE: [1] Strukturuntersuchungen der kristallinan Kohlensto:ffreihe.

Wiss. Veröff. Siemens-Werk6/1 (1927) 188 ............................ 408
HoLM, EU!E: [2] Investigation interrupted through war circumstances about
1940
-, R. HoLM: [3] Über die vom Relais-Lichtbogen erzeugte Stoffwanderung.
Ark. mat., astr., fysik M A (1946) 1 .................................. 307
-, R. HoLM: [4] The temperature generated by friction in the contact surface
of a sliding contact. Festskrift tillägnad Arvid Hedvall, Göteborg'(1948)
271 ................................. ·.· .......................•. :. 231
-, R. HoLM, E. I. SHOBERT II: [.5] Theory of hardness and measurements
applicable to contact problems. J. Appl. Phys. 20 (1949) 319 ............. 373
- [6] Contribution to the theory of the silicon carbide contact. J. Appl. Phys.
23 (1952) 509 ............................................. 60, 189, 387
-, R. HoLM: [7] The fundamentals of the welding of electric contacts under
heavy current conditions. ASTM Bulletin No.188 (Febr.1853) 39 .•.••• 161ff.
-, R. HoLM: [8] Die Stoffwanderung in Abhebekontakten aus Silber und
Platin. Z. angew. Phys. 6 (1954) 352 ..........•..................... 341ff.
- [9] Lecture on material transfer, given at Electrical Contacts Seminar,
Pennsylv. State University, June 1954 .........•.•.•................. 308
- [10] Contribution to the theory of the contact between a carbon brush and
a copper collector ring. J. Appl. Phys. 28 (1957) 1171 .............. 208, 245
- [11] Contribution to the theory of the brush-collector contact. Trans.
A.I.E.E. Part III, 78 (1959) 431 .................................... 245
- [12] Tamperature effect on the specific friction force in contacts where at
least one member is graphite. Proc. Internat. Symposium on Electric
Contact Phenomena. Univ. of Maine (1961) 69 ....... 208, 224, 245, 247, 252
- [13] Specific friction force in a graphite brush contact as a function of
the temperature in the contact spots. J. Appl. Phys. 33 (1962) 156 ... [223ff.,
245, 246
- [14] Einfluß der Stromrichtung auf die Kontaktspannung im Schleifkontakt
Graphit-Kupfer. Proc. Internat. Symp. on electr. contact phenomena.
Techn. Hochschule Graz, Austria (1964) 375 . . . . . . . • . . . . . . . . 137, 223, 245ff.
- [1.5] Dependance of the conduction mechanism on polarity in stationary and
sliding contacts when high-resistivity film is present in the contact. Trans.
IEEE PAS-84 (1965) 404 ..................... 136, 137, 150ff., 245ff., 252ff.
.-.., R. HoLM: patent .................................................. 347
- seeR. HoLM
HoLM, R.: [1] V'I?er Kontaktwiderstände, besonders bei Kohlekontakten. Z.
techn. Phys. 3 (1922) 290, 320, 349 .••................•...... 3, 30, 44, 115
- [2] Beitrag zur Kenntnis der Kontaktwiderstände. Z. techn. Phys. 6 (1925)
166 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
- [3] Über Kontaktwiderstände. Z. techn. Phys. 9 (1928) 455
- [4) Über metallische Kontaktwiderstände. Wiss. Veröff. Siemens-Werk 7/2
(1929) 217 ........................... 40, 42, 44, 46, 48, 87, 115, 135, 195
-, E. HoiM: [.5] Charakteristiken von Kontaktwiderständen. Wiss. Veröff.
Siemens-Werk 7/2 (1929) 272 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
-, R. STÖR.'\I:ER: [6] Eine Kontrolle des metallischen Charakters von gereinigten
Platinkontakten. Wiss. Veröff. Siemens-Werk 9/2 (1930) 323 . . . . . . • . . . . . . 91
- [7] Zur Theorie der ruhenden, metallischen Kontakte mit und ohne Fremd-
schicht. Wiss. Veröff. Siemens-Werk 10/4 (1931) ..•............... 115, 135
-, F. GÜLDENPFE~'NIG, ELSE HoLM, R. STöRMER: [8] Untersuchungen über
· ruhende, gestört metallische Kontakte und über Kontakte mit Fremd-
schichten. Wiss. Veröff. Siemens-Werk 10/4 (1031) 20 [47, 48, 105, 111, 135, 141,
143ff., 167, 168, 192
458 Author and literature index
HoLM, R.: [9] Vorläufige Mitteilung über Metallkontakte mit sehr dünner
Fremdschicht. Z. techn. Phys.12 (1931) 663 ...................... 122, 124
-, W. MEissNER: [10] Kontaktwiderstand zwischen Supraleitern und Nicht-
supraleitern. Messungen mit Hilfe von flüssigem Helium. Z. Phys. 74 (1932)
715 • . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122, 124
- - [11] Einige Messungen über den Fließdruck von Metallen in tiefen
Temperaturen. Z. Phys. 74 (1932) 736 .............................••. 124
-,STRÖME:&: [12] Über Kontakte zweier verschiedener Leiter. WiBB. Verö:lf.
Siemens-Werk 12/1 (1933) 61 ..........................••..... 71, 77, 92
-, W. MEisSNER: [13] Einige Kontaktwiderstandsmessungen bei tiefen
Temperaturen. Z. Phys. 86 (1933) 787 ......................•.... 122, 124
-, A. LoTz [14] Messungen der Gesamtstrahlung der Säule eines Wechsel-
strombogens in Luft. Wiss. Verö:lf. Siemens-Werk 13/2 (1934) 87
- [15] Die Elektrodenzerstäubung in Abhebekontakten. Z. techn. Phys. 16
(1934) 483 ........................................................ 341
-, F. GÜLDENP11ENNIG, R. STRÖME:&: [16] Die Materialwanderung in elek-
trischen Abhebekontakten. WiBB. Verö:lf. Siemens-Werk 14/1 (1935) 30 ..•• 320
- [17] Der zeitliche Verlauf der Erwärmung eines metallischen Kontaktes.
Arch. Elektrochem. 29 (1935) 207 . . . . . . . . . . . . . . . . . . . . .• . . . . . . . .• . . . . . 94
-, F. GÜLDENPFENNIG: [18] Die Materialwanderung in elektrischen Aus-
schaltkontakten, besonders mit Löschkreis. WiBB. Verö:lf. Siemens-Werk
14/3 (1935) 53
-, B. KmscHSTEIN: [19] Über den Widerstand dünnster Fremdschichten in
Metallkontakten. Z. techn. Phys. 16 (1935) 488 • . . . . . . . . . . . . .• . . .• 122, 123
- - [20] Über den Widerstand dünnster Fremdschichten in Metallkontakten.
Phys. Z. 36 (1935) 882 •....................................... , ...•• 122
- - [21] Über das Haften zweier Metallflächen aneinander im Vakuum und
die Herabsetzung des Haftens durch gewisse Gase. Wiss. Verö:lf. Siemens-
Werk 19/1 (1936) 122
- [22] Über die auf die wirkliche Berührungsfläche bezogene Reibungskraft.
Wiss. Verö:lf. Siemens-Werk 17/4 (1938) 38 ........................ 32, 379
- [23] Eine Bestimmung der wirklichen Berührungsfläche eines Bürstenkon-
taktes. Wiss. Verö:lf. Siemens-Werk 17/4 (1938) 43 . . . . . . . . . . . . • . . . .• . . . 32
-, H. P. FINK, F. GüLDENPFENNIG, H. KöRNER: [24] Über Verschleiß und
Reibung in Schleifkontakten, besonders zwischen Kohlebürsten und Kup-
ferringen. Wiss. Verö:lf. Siemens-Werk18/1 (1939) 73 .............. 239, 259
-, B. KmscHSTEIN: [25] Die Reibung von Ni auf Ni im Vakuum. WiBB.
Verö:lf. Siemens-Werk 18/2 (1939) 73 ................................• 200
- [26] Grundsätzliches zum metallischen Kontakt. Z. techn. Physik 20
(1939) 332 .......... : .........•....•........ ; . . . . . . . . . . . . . . .• . . . . . 48
-, H. P. FINK, F. GüLDENPFENNIG: [27] Beiträge zur Lehre der Stoffwande-
rung in Abhebekontakten. WiBB. Verö:lf. Siemens, Werksto:lf-Sonderh. 1940.
103 ••.•.............•............................•...... 279, 282, 287
- [28] Bejtrag zur Kenntnis/der Reibung. Wiss. Verö:lf. Siemens-Werk 20/1
(194~) 68 ....•..............•..................... 44, 207, 208, 212, 239
- [29] Die technische Physik der elektrischen Kontakte. Berlin: Springer
1941 ••................... 7, 61, 94, 140, 160, 167, 191, 243, 245, 338, 356
- [30] "Electric Contacts". Stockholm: Hugo Geber 1946 [35, 48, 61, 100, 108,
156, 208, 236, 244, 245, 252, 320
- [81] Über die Stromdichte in den ersten Momenten eines gezogenen Bogent:t.
Ark. mat., ast., fysik, Stockholm, 34 B (1946) 1
HoLM, R.: [32] Calculation of the temperature development in a contact
heated in the contact surface, and application to the problern of the tem-
perature rise in a sliding contact. J. Appl. Phys. 19 (1948) 361 . • • 94, 100, 231
- [33] The vaporization of the cathode in the electric arc. J. Appl. Phys. 20
(1949) 715 • . . . . .• . . . . . . . . . . . . . . . . . . . . .• . . . . . . . . . . . . . . . . . . . . . . . .• • . 308
- [34] Elektrische Kontakte in Meßgeräten. ATM. J. 04-4 (Aug. 1950) .. 1'75, 242
- [35] The electric tunnel effect across thin insulator fil.ms in contacts. J. Appl.
Phys. 22 (1951) 569, with errata J. Appl. Phys. 22 (1951) 1217 [120, 122,
131, 133
- [36] Tamperature development in a heated contact, with application to
sliding contacts. J. Appl. Mech. 19 (1952) 369......................... 94
- [37] Electric Contacts Handbook, Berlin, Springer, 1958 [30, 61, 112, 123,
129, 174. 181ff., 199, 202, 214, 233, 241ff., 252, 255, 272, 276, 284ft'., 296,
332,336,356, 363
- [38] Contribution to the theory of commutation on d-c machines. AIEE
Power Apparatus and Systems (Dec. 1958) 1124 . . . . . . . . . . . . . . . . .• . . . . . 251
- [39] Electron cloud outside a metalsurface. J. Appl. Phys. 30 (1959) 792
- [40] Theorie der ungleichen Stromverteilung im Kontakt zwischen Kohle-
bürsten und Schleifringen aus Kupfer und Stahl in Luft. ETZ-A 81 (1960)
570
- [41] Theory of the sparking during commutation on dynamos. AIEE
Power Apparatus and Systems (Dec. 1962) 588 .•.••......••..... 262ff. 268
- [42] Relation between cold weid and hardness of metals. Proc. Internat.
Symposium on Electric Contact Phenomena Techn. Hochschule Graz,
Austria (1964) 19 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .• . . . . . . . . . . . . . .• . 155
- R. M. BAuEB D. F. EMMETT W. E. YoBE: [43] Brush wear during com-
mutation. IEEE Trans. PAS-85 (1966) 84 .•...................... 253, 254
- seeE.HOLM
HoNJO, G.: Electron diffraction studies on oxide films formed on metals and
alloys. I Oxidation of pure copper. J. Phys. Soc. Japan (1949) 330 . • . . . . . 113
HoPXINS, M. R.: [1] The thermal and electrical conductivities of metals at
high temperatures. Z. Phys.147 (1957) 148 •.••••....•.•••.....•....•.. 438
- R. L. GRD!'l!'ITH: [2] The determination of the Lorenz number at high
temperatures. Z. Phys. 100 (1958) 325 • . .• • . . . .• • . .• . • . . . . . . . . . . . . . . . . 438
- C. H. JoNEs: [3] The use of radioactive contact electrodes in the measure-
ment of metal migration in electric contacts. Brit. J. Appl. Phys. 14 (1963)
137 ••.....•......•.••....•.....•..••••...•••..•••.•.......•...•.. 341
- [ 4] The physical phenomena leading to metal transfer between the electrodes
of electrical contacts. Proc. Relay Conference, Oklahoma State Univ.
(1965) paper 28 ••........................................... 341ff., 346
- see LLEWELLYN-JONES
HoWE, J. E.: Graphite as a high temperature material. Trans. Metallurgy.
Soc. of AlME Feh. 1958, 7 .• • • • . • . • • • • • • . .• • . • • • . • • . • • • • . • . • . • • . . . . 404
HovGAABD, 0. M., G. E. PElmAULT: Development of reed switches and relays.
Bell Syst. Techn. J. 34 (1955) 309 • .. . .. . .. . . .. .. . .. .. .. .. . . .. . .. . .. . 171
HousKA, C. R., B. E. WARBEN: X-Ray study of the graphitization of carbon
black. J. Appl. Phys. 29 (1954) 1503 ............................ 407, 408
HOUSTON, J. M.: see ßURGESS
Hown, D.: see TYLECOTE
HuGHES, T. P., G. WHITTINGHAM: The infiuence of surface fil.ms on the dry
and lubricated sliding of metals. Trans. Faraday Soc. 38 (1942) 9 ........ 230
- seeHEATOX
HUNTER-BROWN, P.: Carbon brushes. J. Inst. electr. Engrs. 67 (1919) 193 •.. 245
HUTCHEON, I. C.: Contact resistance effects in mechanical choppers. J. Brit.
I.R.E.21(1961)153
ILscHNER-GENBCH, CHRISTA, C. WAGNER: Local cell action during the scaling

of metals. J. El. Chem. Soc. 106 (1958) 198 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
!TTNER, W. B.: [1] Bridge and short arc erosion of copper, silver, andpalladium
contacts on break. J. Appl. Phys. 27 (1956) 382 ...................... 341ft".
-, H. B. ULBH: [2] The erosion of electrical contacts by the normal arc. Proc.
Inst. El. Eng. 104 (1957) 63 .....................................•.•• 440
lvEs, H. E.: Minimallength arc characteristics. J. Franktin Inst. 198 (1924)
437 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .• • • 279, 435
JACOB, CH.: Über gleitende Reibung. Ann. Physik Lpz. 38 (1912) 126
JAEGER, J. C.: [1] Moving sources of heat and the temperature at sliding con-
tacts. J. Roy. Soc. NewS. Wales 1)6 (1942) 203 ..................•. 94, 98ft".
- [2] Approximation in transient surface heating Austral. J. Sei. Res. 5
(1952) 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . • 94, 99
- [3] Pulsed surface heating of a semi-infinite solid. Quart. Appl. Math. 11
(1953) 132 . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . • . . . 94
- see CARSLAW
J.AHNKE-EMDE: Funktionstafeln. Leipzig: Teubner 1938 ............... 17, 55
JENKINS, R. 0.: Electron diffraction experiments with graphite and carbon
surfaces. Phil. Mag.12 (1934) 457 ..................................•• 222
JoBLING, H. B.: A review of the manufacture and use of carbon brushes.
Trans. S. African Inst. Electr. Engin. 06, Sept. (1965) 211
JoHNSON, J. B.: Thermal agitation of electricity in conductors. Phys. Rev. 32
(1928) 97 ................................................•........ 269
JoHNSON, K. L.: A note on the adhesion of elastic solids. Brit. J. Appl. Phys.
9 (1958) 199 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .• . . . . . . . . 157
JoHNSON, R. L.: see BISBON
- see BuCKLEY
JOHNSON, V. R.: [1] Survey of the nature of the friction forces in molybdenum
disulphide lubrication. MRI Project No. 129-P-65 . . . . . . . . . . . . . . . . . • • • . . 227
-, G. W. VAUGHN: [2] Investigation of the mechanisni of MoS2 lubrication in
vacuum. J. Appl. Phys. 27 (1956) 1173 ............................... 227
JoNES, C. H.: see HoPXINs
JoNEs, C. R.: see LLEWELLYN-JONES
JoNEs, F. LL.: see LLEWELLYN-JONES
JONES, L. K.: Materials and process variables and .their effect on conta.ct
resistance. Proc. Seminar on Electric Contacts. Pennsylvania State Univ.
(1960) 52 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .• • • 29
JoNES, R. H.: see LLEWELLYN-JONES
JoNEs, T. K., R. A. ScoTT, lt. W. SILLARS: The structure and electrical
prope~ies of surfaces of semiconductors. Proc. Phys. Soc. Lond. A 62
(1949) 333 ..... '. . . . . . . . • . . . . . . . . . . .• • . . . . . . . .• • • • . . . .• . • . . . . . . . . . 189
JusT, G.: Die Durchschlagsfestigkeit äußerst dünner Ta20 1-Schichten in Ab-
hängigkeit von der Schichtdicke. Z. Phys. 82 (1933) 119 .••....••....••.• 146
JusTI, E., H. SCHULTZ: Neue Versuche zur))entung derFeinwanderung in
elektrischen Abhebekontakten. Abh. Braunschw. Wiss. Ges. t (1949) 89 [69,
71, 339, 344
KAMERLINGH-ÜNNES, H.: Further experiments With liquid helium. The per-
sistence of currents without electromotive force in supraconducting circuits.
Commun. Leiden 141 b (1914) 15 . . . . . . . . . . . . • . . . . . . . .• . . . . . . . . . . . . . . . 128
KANTOROWICZ, 0.: Zur Leitfähigkeit gepreßter Metallpulver. Ann. Phys., Lpz.
12 (1932) 1 0. 0 0. 0 0 0 0 0 0 0 19.!, 192
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
- see SKAUPY ,
KAPPLER, E., E. RecHARD'I, R. ScHLÄFER: Kontaktwiderstand in Abhängig-
keit von der Kontaktlast. Z. angew. Phys. 2 (1950) 313 . . . . . . . . . . . . . . . . . 46
KAUFMANN, W.: mentioned ...•..............•........................ 282
KEEFER, H. J., R. H. GUMLEY: Relay contact behavior under non-eroding
circuit conditions. Bell Syst. techn. J. 37 (1958) 777 . . . • . . . . . . . . . . . 170, 362
KEHOE, J. W.: Attaching contacts. Proc. Seminar Electric Contacts. Pennsylv.
State Univ. (June 1954) ........................................•... 176
KEIL, A., C. L. MEYER: [1] Die Feinwanderung an Kontakten aus Legierungen
mit Überstruktur. :Metallkunde 44 (1953) 22
-, W. MERL: [2] Über die Materialwanderung an elektrischen Unterbrechungs-
kontakten. Metallkunde 48 (1957) 16
- , - [3] Über das Schweiß- und Abbrandverhalten von Unterbrecherkontakten
aus Silber-Graphit. Metall12 (1958) 619 ......................... 341, 364
- [4] Werkstoffe für elektrische Kontakte. Berlin: Springer 1960 [176, 296,346,
362, 364, 366
-, C. L. MEYER: [5] Die mechanische Deformation von Kontaktstücken durch
den Schaltlichtbogen. ETZ, B 12 (1960) 309
-, C. L. MEYER: [6] Kristallwachstum bei Schwefeleinwirkung auf Silber und
beim Zerfall von Silbersulfid. Metallkunde ol (1960) 253
-, C. L. MEYER: [7] Über die Entstehung haarförmiger Kristalle auf me-
tallischen Oberflächen. ETZ B, 14 (1962) 697 .......................... 417
- [8] Über spec. ehern. Reaktionen an Edelmetall-Oberflächen. Metall 15
(1961) 655 0 0 0 0 0 •• 0. 0. 0 0 0 112 0. 0 0 0 •• 0 0 0 0. 0. 0 0 0 0 0 0 0 0 0 0 0 0 0. 0 0 0. 0 •• 0 0 0. 0. 0
- [9] Migration ot silver as a sulfide, deformation of contacts by arcs, con-

ductivity of condensation products. Materials, ASTM 3 (1963) 489 ....... 111
KEILmN, S.: see RussELL
KENYON, D. M.: mentioned ........................................... 225
KERRIDGE, M.: [1] Meta! transfer and the wear process. Proc. phys. Soc., Lond.
B68 (1955) 400 .............................................. 233, 234
-, J. K. LANCASTER: [2] The stages in a process of severe metallic wear. Proc.
Roy. Soc. A 236 (1956) 250 .................................... 239, 241
KESAEV, I. G.: Interna! instability of an arc with a mercury cathode. Sowj.
Phys. Doklady 3 (1958) 967, and Tech. Phys. 4 (1960) 1351 ............. 289
KESSELRING, F.: [1] 'l;heoretische Grundlagen zur Berechnung der Schaltgeräte
Samml. Göschen 711. Berlin 1950 ..................... 275, 292, 295, 297
- [2] Technische Kompositionslehre, Berlin: Springer 1954 . . . . . . . . . . . . . . . 302
- [3] Erfahrungen mit elektromagnetisch gesteuerten Großgleichrichtern.
Scientia Electrica 11 (1956) 140 ............... .'................ 296, 302
- [4] Der Einfluß der Feinwerktechnik auf die konstruktive Gestaltung von
Hochleistungsschaltern. VDI-Z 106 (1964) 1573
KINGDON, K. H.: The arc cathode spot and its relation to the diffusion of ions
within the cathode material. J. Appl. Phys. 36 (1965) 1351 .... 286, 288, 289
KmsCHSTEIN, B., F. KoPPELMANN: Photographische Aufnahmen elektri-
scher Lichtbögen großer Stromstärke. Wiss. Veröff. Siemens Werk 13/3
(1934) 52 . 0 • 0 0 0 0 0 • 0 •• 0 0 293
0 0 • 0 • 0 ••••• 0 • 0 0 0 0 0 ••• 0 • 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 • 0 0 0 0
- see R. HoLM
KisLrux., P.: (1] Arcing at electric contacts on closure. Part V. The cathode
mechanism of extremely short arcs. J. Appl. Phys. 25 (1954) 897 ....•.... 276
- [2] Using contact resistance to measure adsorption of gases on metals. Bell
Syst. techn. J. 37 (1958) 925 ...............................•... 105, 126
- [3] Electron emission at high fields due to positive ions. J. Appl. Phys. 30
(1959) 51 . ·......•............................................•.•.. 277
- [4] Chemisorption of nitrogen on tungsten. J. Chem. Phys. SO (1959) 174 [ 104, 105
- [5] Calorimetric heat absorption-nitrogen on tungsten. J. Chem. Phys. 31
(1959) 1605 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .• . . . . . . . .• . . . . . . . 104, 105
Kr.rTEL, CH.: Introduction to solid state physics. Second ed. New York: Wiley
& Sons, 1956 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . • . . 380, 395
KI.AunY, P.: [1] Eigenschaften und Anwendungsmöglichkeiten von Flüssig-
keitskontakten. ETZ 76 (1955) 525 ...................•.•...•.... 108, 358
- [2] Stromabnahme bei hohen Geschwindigkeiten. Maschinenbau und
Wärmewirtschaft 11 (1956) 315 ............................ 105, 108, 358
- [3] Fortschritte im Bau von Unipolarmaschinen durch Anwendung von
Flüssiglreitskontakten. Elektrotechn. und Maschinenbau 78 (1961.) 128 .... 358
- [4] Liquid sliding contacts. Proc. Internat. Symposium on Electric Con-
tact Phenomena. Univ. of Maine (1961) 49 ......................••..•. 358
- [5] Einigeneuere Untersuchungen an Flüssigkeitskontakten. Proc.Internat.
Symposium on Electric Contact Phenomena. Techn. Hochschule, Graz,
Austria (1964) 3. About the same subject: 'Ober Flüssigkeitskontakte, ATM
(1965) R97 ....................................................... 358
KLXm, R.: see ExxERS
KLE!NLE, A.: see HAASE
KLuGE, J.: [1] Schmierung als Grenzfl.ächenvorgang. Metallkunde 40 (1949)
386 ...•........................................................•. 211
- [2] Communication to the author . . . . .. .. .. . . . . .. . . .. . . . . . . . . .. . .. . .. 239
KMETXO, E. A.: Electronic properlies of carbons and of their interstitial com-
pounds. J. Chem. Phys. 21 (1953) 2152 ...••..............•........•.. 408
KOBEL, E.: Zusammenhang zwischen der Benetzung und dem elektrischen
Vbergangswiderstand zwischen Eisen und Queclrsilber. Schweizer Arch. für
angew. Wiss. Techn. (1948) 326 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
KoBITZSCH, R.: see SIEBEL
KoHLER, MAx.: [1] Thermischer Kontaktwiderstand von Metallen und der da-
durch verursachte Temperatursprung im Kontakt. Ann. Phys., Lpz. 38
(1940) 542 .....••. ·.........•.....•..........•.•..••••.....•...•••. 69
-, G. ZIELA.SEX: [2] Der zeitliche Temperaturverlauf in elektrischen Kontak-
ten. Abh. Braunschw. Wiss. Ges. 4 (1952) 117 . . . . . . . . . . . . . . . . . . . . • . . . . 94
- mentioned .. .. .. • . . ... .. . .. .. . . . . . .. .. . . . . . .. . . . . . . . . . . . . . . . . . . . . . 69
KoBLRAUSCH, F.: Vber den stationären Temperaturzustand eines elektrisch
geheizten Leiters. Ann: Phys., Lpz.l (1900) 132 :. . . . . • • . . . . . . . .• . .• • . . 6i
KoBliiA.N, G. T., H. W. HERMANOE, G. H. DOWNES: Silver migration in electri-
cal insulation. Bell Syst. techn. J. 34 (1955) 1115 •..•...•.•.•...•.•.... 418
KoLLER, R.: FundameTJ.tal properties of the vacuum switch. Trans. Amer. Inst.
electr•.Engrs. 65 (1946) 59'7 .......................................... 308
KOPPE{.MANN, F.: [1] Der Kontaktumformer. ETZ 62 (1941) 3 ...........• 296
- [2] Der Kontaktgleichrichter der AEG. ETZ 4B (1952) 224 •...........• 296
- [3] Kontakt-Stromrichter für 16000A, ETZ-B (1953) 395 .......•.•.•.• 296
KOPPLIN, H., E. 8omm>T: Beitrag zum Dynamischen Verhalten des Licht-
bogens in Ölarmen Hochspanimngs-Leistungsschaltern. ETZ-A 80 (1959)
805 ...... ;· ...........•.....•....•..•............................. 295
KÖRNER, H.: see FINK
- see R.HoLM
KOTTLER, F.: Elektrostatik der Leiter. Handbuch der Physik von GEIGER und
SCHEEL, Bd. 12. Berlin: Springer 1927 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14, 15
KoZAK, RosE: see CAMPBELL
KUHLMANN-WILSDORF, Dorus, H. G. F. WILSDORF: Dislocation movements in
metals. Science 144 (3 Apr. 1964) 17............................. 367, 370
KÜPFMÜLLER, K.: Einführung in die theoretische Elektrotechnik. Berlin:
Springer 1965 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
KuozoGI, E.: Die Halbleitertheorie der Schleifkontakte elektriscrer Maschi-
nen. Acta Techn. Hung. 49 (1964) 191 ................................ 187
LAFFERTY, J. M.: see FARRALL

LANoAS'l.ER, J. K.: [1] The influence of temperature on metallic wear. Proc.
Phys. Soc. London B, 70 (1957) 112
- [2] The influence of arcing on the Vlear of carbon brushes on copper. Wear,
6, Sept. 1963, 341 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253, 255
- [3] The relationship between the wear of carbon brush materials and their
elastic moduli. Brit. J. Appl. Phys. 14 (1963) 497
- see HmsT
LL'IDER, J. J. [1] Measurements of THOMSON coefficients for metals at high
temperatures and of PELTIER coefficients for solid-liquid interfaces of metals.
Phys. Rev. 74 (1948) 479 ...........................•............... 347
-, L. H. GERMER: [2] The bridge erosion of electrical contacts. Part. I. J. Appl.
Phys. 19 (1948) 910 .............................................. 341ff.
- [3] Chemisorption and ordered surface structures. Surface Science 1 (1964)
125
LANDSBERG, K. E.: tlber die physikalischen Vorgänge bei der gleitenden Rei-
bung fester Körper. Poggendorfs Ann. 1 (1864) 283 ..................... 114
J..A.NGSDORFF, W.: seeDEHN
LANYON, H. P. D.: see SPEA.R
LAWSON, G. R.: Generation of fast-growing whiskers in the neighborhood of
arcing metallic contacts. Proc. Internat. Symposium on Electric Contact
Phenomena. Univ. of Maine (1961) 285 ............................... 417
LEE, T. H.: [1] Testing randomness of experiments theory and application to
ASTM surety of make tests. Data Folder R 53 CO 605, Gen. Electr. Co .... 400
- [2] T-F theory of electron emission in high current arcs. J. Appl. Phys. 30
(1959) 166 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131, 427
-, A. N. GREENWOOD, D. W. CRouOH, C. H. TITUS: [3] Development of power
vacuum interruptors. AIEE Trans. 81 (1963) 629 ...................... 291
-, A. N. GREENWOOD, D. R. WHITE: [4] Electrical breakdown of high tem-
perature gases and its implications in post-arc phenomena in circuit breakers.
IEEE Trans. PAS-84 (1965) 1116 ............................... ; ... 290
LEEDS, W. M., R. E. FRIEDRIOH: Recent developments in the use of SF8 for
power circuit breakers. Proc. Americ. Power Conf. 24 (1962) 833 ......... 295
LEWis, F. A.: see UBBELOHDE
LEWIS, J. A.: seeFOXHALL
LEWIS, T. J.: High field electron emission from irregular cathode surfaces. J.
Appl. Phys. 26 (1955) 1405 ......................................... 427
LlLIENFELD, S., C. E. WHITE: A study of the reaction between hydrogen sul-
fide and silver. J. Am. Chem. Soc. 52 (1930) 885 ...................... 110
464 Author aud Iiterature index
Ll:NGAL, H. J., A. P. STROM, T. E. BROWNE, JR.: Are quenching in sulfur hexa-

fiuoride. Power Apparatus & Systems, No. o (April 1953) 242 . . . • • . . . . .• . 294
LrsH, E. F.: Relay contaminants and how to identify them. Conf. on Electro-
magn. Relays. Stillwater, Oklahoma, April1961. Paper 35 . . . . . • . . . 118, 364
LITTLE, P. F.: Second.ary effects. Handbuch der Physik, Bd. XXI, 574. Berlin:
Springer 1956 .............................•........•.............• 424
LLEWELLYN-JONES, F.: [1] Arcing phenomena at electrical contacts as used in
communication engineering. Proc. Inst. Electr. Engineers 00 (1949) 305. • • 278
- [2] Initiation of discharges at electrical contacts. Proc. Instn. Electr.
Engrs.100 (1953) 169 ...•...............•....................•..... 275
-, R. H. JONES: [3] The information and rupture of molten metal bridges in
electrical contacts. Z. Physik 147 (1957) 43 . . . • . . . . . . . . . . . . . . . . . . . . . . . 64
- [4] The Physics of Electrical Contacts. Oxford, ClarendonPress 1957 [16,338, 341
-, M. R. HoPKINS, C. R. J ONES: [5] Measurements of metal transfer on electri-
cal contacts bJ the radioactive isotop method. Brit. J. Appl. Phys.12 (1961)
485 ....•.......•....•..............•...•..............•...•• 338, 341
- [6] Matter transfer in contacts and the microscopic molten metal bridge.
Proc. Internat. Symposium on Electric Contact Phenomena. Techn. Hoch-
schule Graz, Austria (1964) 180
- mentioned . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 339ff.
LoEB, L. B. [1] Fundamental proeesses of electrical discharge in gases. New
York: Wiley & Sons 1939 ............•............................ 421ff.
- [2] Basic processes of gaseous electronics. Berkeley: University California
Press, 1955 ........................................................ 275
- [3] Statik electrifi.cation. Berlin Springer, 1958 .......................• 189
LoTz: see R. HoLM
LuNN, B.: [1] Die Beilby-Schicht. Metallkunde. 40 (1954) 92 .............. 220
- [2] Evaluating bearing materials under bound.ary lubrication. Lubr. Engng.
July-Aug. 1955 .•....•.........•.............•••..........•.......• 220
- [3] Friction and wear under boundary lubrication. Wear 1 (1957-58) 25
McCoRNIOK, W. G. A.: Cleaning by ultrasonic. Industrial Electronics 1 (August

1963) 551 . . • • . • . . . . . . . . . . .• . . . . . . . . . . . . . . . . . . . .• . . . . . . . . . . . . . . .• . . 29
McCABE, J. T.: Molybdenum disulfide- its role in lubrication. Abrief sum.mary
of the state of the art. Scientific Lubrication (1965) . . . . . . . • . • . . . . . . . . .• 227
McFARLANE, J. S., D. TABOR: Adhesion of solids and the effect of surface
films. Proc. Roy. Soe., Lond. A 202 (1950) 224 ...................•.... 157
MACHLIN, E. S., W. R. YANKEE: Friction of clean metals and oxides with spe-
cial reference to titanium. J. Appl. Phys. 29 (1954) 576 •...••• 105, 157, 220
MA.CKEOWN, S. S.: [1] The cathode drop inanelectric arc. Phys. Rev.M (1929)
611 •...•...........•.•........•............ ·.· ...............•...• 429
-, J. D. CoBINE, F. W. BoWDEN: [2] Reignition of an· arc at low pressures.
Electr. Engng. ii3 (1934) 1081 •.•.•••.••...••••••.................... 296
MAOKLEN, E. D.: [1] Investigation of the electric contact properlies of granular
carbon aggregates. Brit. J.~ppl. Phys. 12 (1961) 443 .•..•....•.•.•• 52, 184
-, B. HoDGSON: [2] Correlation between contact resistance and oxygen con-
tent. 'Brit. J. appl. phys. 18 (1962) 171 . . . . . . . . . . .• . . . . . . . . . . . . . . . 52, 184
- [3] Electric contact properlies of granular carbon aggregates. Part 3- ·
Effects of prolonged air oxidation. Brit. J. appl. phys. 14 (1962) 28 •. 52, 184
- [4] Electrical contact properlies of granular carbon aggregates. Part. 4: ln-
vestigation of mechanical ageing. Brit. J. appl. phys. 6 (1965) 69 . . .• • . . . 184
McRAE, A. U.: see GERMER
llicxs, E. F.: see SHAW
)!AcPHERSON, H. J.: see CURRIE
!\<!ADER, H. J.: Elektrische Stromleitung und Struktur von dünnen technisch
herstellbaren Aluminiumoxydschichten auf AluminiumträgermetalL Z. an-
gew. Phys. 14 (1962) 542 ............................................ 107
:i.\<IAECKER, H.: [1] Plasmaströmungen in Lichtbögen. Z. Phys. 141 (1955) 198 [435
- [2] Fortschritte in der Bogenphysik. Ionization phenomena in gases. Amster-
dam, N.-Holland Publ. Co. ( 1962) 1793 .............................. 423
- see FlNKELNBURG
- mentioned . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 294
:i\lAHLER, P.: On reliability and what must be done to attain it. Proc. Seminar
on electr. contacts. Univ. of Maine (1963) paper 7 . . . . . . . . . . . . . . . . . . . . . 29
)iAILANDER, R., K. DIEs: Beitrag zur Erforschung der Vorgänge beim Ver-
schleiß. Tech. Mitt. Krupp, 5. Jahrgang, 14, 209. Essen 1942 ............ 237
::IIALLINA, R. F., W. P. MAsoN, T. F. ÜSMER, R. H. VAN HoRN, 0. L. ANDER-
soN: Solderless wrapped connections. Bell Syst. techn. J. 32 (1953) 523-610
and 1093 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
v. MANGOLD-KNOPP: Einführung in die höhere Mathematik 3. Leipzig: Hirzel
1933 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 4030 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
lliNo, K.: see HARADA

MARsHALL, M. J.: see WRIGHT
3-!ARSHALL, R. A. : The design of brush gear for high current pulses and high
ruhhing velocities. IEEE Trans. paper 31TP66-11 (1966) ............... 358
::\IARTIN, F. E., H. E. STAUSS: Contact transients in simple electric circuits.
Trans. AIEE 70 (1951) 304 .......................................... 315
l\IASING, G.: Lehrbuch der allgemeinen Metallkunde. Berlin: Springer 1950 .. 111
:NIASON, W. P.: Adhesion between metals and its effect on fixed and sliding con-
tacts. ASLE Trans. 2 (1952) 39 ............................. 154, 159, 166
- see MALLINA
::\IAu, H. J.: Über die Schweißhaftkraft von ruhendenKontakten bei Belastung
durch einen Halbwellenstrom. Tagung über Kontakte in der Elektrotechnik.
Berlin: Akademie-Verlag 1965 ..•................................. 161ff.
MAXWELL, J. C.: A treatise on electricity and magnetism, I. Oxford: Clar-
endon Press 1904, also Dover Publications: 1954 . . . . . . . . . . . . . . . . . . . . . . 20
}!AYR, 0.: mentioned . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 292, 293
MECKE, H.: Stromdurchgang bei Rollenlagern. AEG-Mitt. (Feh. 1925) 74 ..• 221
::\-IEISSNER, H.: [1] Measurements on superconducting contacvs. Phys. Rev. 109
(1958) 686 0 0 0 0 0 0 0 0 0 0 0 0 0 0129 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
- [2] Superconductivity of contacts with interposed barriers. Phys. Rev. 117

(1960) 672 0 0. 0 0 0 0 0 0 0 0 0 0 0129 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
~lEISSNER, W.: [1] Thermische und elektrische Leitfähigkeit einiger Metalle

zwischen 20 und 373° abs. Ann. Phys., Lpz. 47 (1915) 1001 . . . . . . . . . . . . . 75
- see R. HoLM
~lENDIZZA, A.: see EGAN
MENTER, J. W., D. TABOR: Grientation of fatty acid and soap ffims on meta!
surfaces. Proc. Roy. Soc., Lond. A 204 (1951) 514 ...................... 217
MERcluNT, C. C.: see BENEDICK
MERL, W.: [1] Stoffwanderung an Gold und Goldnickel-Kontaktstücken. ETZ A
77 ( 1956) 201 0 0 0 0 0 0 0 0 0 0 341 ff.
0 0 0 • 0 0 0 0 0 0 0 0 0 0 0 • 0 0 0 0 0 0 0 0 0 0 0 0 0 0 •• 0 0 0 0 0 0 0 0 0
- [2] Der elektrische Kontakt. Pforzheim, E. Dürrwächter (1959) .......... 341

- [3] Infl.uence of alloy structure on the transfer of matter. Brit. J. Appl. Phys.
12 (1961) 447 0 0 0 0 0 0 0 0 0 0 0341 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
Mmu., W. : [4] Material transfer on make in a low voltage circuit. Proc.

Internat. Symposium on Electric Contact Phenomena. Univers. of Ma.ine
(1961) 201
-, E. VINARICKY: [5] Abbranduntersuchungen an Sinterwerkstoffen auf Wolf-
rambasis mit einer Kondensatorbatterie. Proc_ Internat. Symposium on
Electric Contact Phenomena. Techn. Universit. Graz, Austria (1964) 206 311
- [6) Weidtests on silver contacts with ASTM weid test fixture. Doduco In-
formation [1] 10 (1965) ............................................. 304
- see DfumWACHTER
MEYER EuGEN: Untersuchungen über Härteprüfung und Härte. Phys. Z. 9
(1908) 66 ......•....•.....................•...•••.•.••..........•. 373
MEYERHOFER, D., S. A. OCHS: Current ßow in very thin films of .AJs03 and BeO.
J. Appl. Phys. 34 (1963) 2535 ...............••..•....•...•....•...•• 122
MmGLEY, J. W., D. G. TEER: An investigation of the mechanism of the fric-
tion and wear of carbon. Trans. ASME (Dec. 1963) 488 . . . . . . . .• • • 222, 248
MlLLER, H. C., G. A. FARRALL: Polarity eft'ect in vacuum breakdown electrode
conditioning. J. Appl. Phys. 88 (1965) 1338 ....•.•.•..•••.•..••••..•• 286
MILLER, H.: Selenium device use •.• Electronic News: May 16 (1965) • . • . • . • • 185
Mn.u.A.N, K., W. RIEDER: Kontaktwiderstand und Kontaktoberßäche. Z. an-
gew. Phys. 8 (1956) 28 ...... ·.......•................•............... 46
Mn.Ls, G. W.: A comparison of permanent electrical connections. Bell Syst.
Techn. J. 43 (1964) 1067 . . . . • • . . . . . . .• . . . . . . . . . . . .• • . . . . . .• . • • 159, 166
MILNER, D. R., G. W. RowE: Fundamental of solid phase welding. Metallurg.
Rev. 7 (1962) 433 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . • . . . . . .• • . . . 159
M:rsEs, v. R.: mentioned ..................................•..•....•... 370
MIYAZAKI, T.: see NAKAI
MooRADIAN, V. G.: Selecting materials for electric contacts. Materials and
Methods 44 (1956) 121 ........................................ 176, 362
MooRE, A. J. W.: Deformation of metals in static and sliding contacts. Proc.
Roy. Soc., Lond. A 196 (1948) 231 . . . . . . . . . . . . . • . . . . . . . . . . . . . . . . . 36, 216
- see BOWDEN
MORGAN, F., M. MuSKAT, D. W. REED: [1] Friction phenomena in stickslip
process. J. Appl. Phys. 12 (1941) 743 . • • . . . . .• • . . . .• • • . . . .• • . . . . 230, 231
- - - [2] The friction of dry and lubricated surfaces. Lubrication Engng.
April 1949, 75 .... , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 230
MoRGAN, D. A.: see BURWELL ·
MoRT, J.: see SPEAR
MOTTER, D. P.: Commutation of D-C-machines and its eft'ects on radio in-
ßuence voltage generation. Trans. AIEE 8 (1949) 491. .......•...•••...• 272
MRozowsKI, S.: [1] Semiconductivity and diamagnetiBm of polycrystalline
graphite and condensed ring systems. Phys. Rev. 8ö (1952) 609, with correc-
tion. Phys. Rev. 88 (1952) 1056 ....•.......•.........•••....••. 406, 408
- [2) Diseuseion of Howe paper. J. Amer. ceram. Soc. 3ö (1952) 282. Fig.1 ... 412
- [3) St~dies of carbon powders under compression. Proc. 3rd Conference on
Car~on in Buft'alo. Pergamon Press (1959) 495 ·
- mentioned ... , . ... .. .. . . . . . .. .. . .. . . . . . .. .. .. .. . . .. .. . . . .. .. . . .. . • 409
MULHEARN, T. 0., D. TABOR: Creep and hardness of metals. J. Instit. Metals
(Sept. 1960) 7 . • • . . . . . . .• . • . . . . . . . . . . . . . . . . . . . . . . . . .• . • . • . • • . • • • . . . 375
MÜLLER-Hn.I.E'BRA1W, D.: Aluminium als Leiter in Schaltgeräten. ETZ 61
(1940) 1117" •..•..•.•............... :. . . . . . . . . . . . . . . . . . . . . . . . . . . . . • 46
MÜLLER, E., i!ee. GooD
MUBPHY, E. L., R. H. GooD, JR.: Thermionic emission, field emission and the
transition region. Phys. Rev. 102 (1956) 1464 .................... 131, 427
MuSKAT, H.: see MoRGAN
MUTO, Y.: see FUKUBOI
NAKAI, J., T. MIYAZAKI: On the tunneling current through thin aluminium

oxide films. Japan. J. appl. phys. 3 (1964) 677 ............•.........•.. 131
NEUKIBCHEN, J.: Feinwanderung an elektr. Kohle-Kohle Kontakten. Abhandl.
Braunschw. Wiss. Gesellsch. o (1953) 63 ............•......••..•••.... 320
Noble metals: Information in "Platinum Metals Revue", Johnson Matthey
Group, Malvem, Pa; and in "Precious Metal Electrodeposits", Internat.
Nickel Co., published 1966
NoRDHEIM, L.: see FoWLER
NösKE, H.: Untersuchungen an kurzen Wechselstrom-Lichtbögen in Luft.
Z. angew. Phys.10 (1958) 327, 382 .............................. 295. 296
NoTTINGHAM, W. B.: [1] Normalare chara.cteristic curves. Phys. Rev. 28 (1926)
764 ............................................................... 435
- [2) Thermionic emiBSion. Handbuch der Physik XXI, 1. Berlin: Springer
1956 .................•..•........................................ 422
NuNN, C. P.: [1] A study of discharge transients in rela.ys with grounded ca.ses.
Conf. on electrom.agn. rela.ys. Stillwater, Oklahoma., April1961. Paper 34 .. 364
- [2] Minimum current testing and studies. Conf. on electroma.gn. rela.ys. Still-
water, Okla.homa., April1964, Paper 20 ............................... 364
NYQVIST, H.: Thermal a.gita.tion of electric charge in conductors. Phys. Rev.
32 (1928) 110 ..•....•.......................................•..... 269
OCHS, S. A.: see MEYERHOFER
OETTL, K.: Die betrieblichen Ansprüche an den Schwa.chstromkontakt. Inter-
nat. Symposium on electric conta.ct phenomena.. Techn. Univ. Gra.z, Austria.
(1964) 214 . . . . . . . . . . . . . . . .• . • . . . . . . . . . . . . .• . .• . . . . . .• . . . . . .• . • . . . . 171
OosTERKAMl', W. J.: The hea.t diBSipation in the a.node of an x-ra.y tube.
Philips Res. Rep. 3 (1948) 49. Cf. J. appl. Phys.19 (1948) 1180.......... 94
OSMER, T. F.: see MALLINA
PAETOW, H.: Kontaktschmelzbrücken undFeinwanderung. ETZ 70 (1949) 227 [338
PALME, R.: Der Einfluß der oxidierende Wirkung atmosphärischer Luft auf
die Wirkungsweise elektrischer Kontakte. Schweizer Arcliiv 19 (1953) 177 [365
P ARDEE, R. P.: Moisture dependenoe of silver-graphite brushes in a.ir, nitrogen,
h.älium and ca.rbon dioxide. IEEE Transa.ct. paper 31 TP 66-148 (1966) ... 226
PARRISH, F. W.: Are suppreBSion with semiconducting devioes. Elecr. Mfg, 97
(June 1956) 127 ....................•......................••...... 326
PASCHEN, F.: mentioned .........•..........................••........ 276
PATBL, A. R., 0. P. BAHL: Evidenoe of screw disloca.tions in gra.phite. Brit. J.
appl. Phys. 16 (1965) 169 .•.•.••....•....•......•...............•..•. 222
PEARsoN, G. L.: The forma.tion of metallic bridges between separated oonta.cts.
Phys. Rev. 01 (1937) 1015 and 06 (1939) 471 .................•........ 417
Plm.RAULT, G. E.: see HovGAARD
PlnsTERER, H., E. FuCHS: Identifizierung von Kontaktverunreinigungen mit
elektronen-optischem Verfahren. Siemens Zsohr. 34 (1960) 484 ....•. 117, 173
Pn.LAI, K. P. P.: see CHOUDHUBI
Plm.I.n>s, R. L.: The behavior of dyna.mic electr. a.rcs. ARL 64-150 (1964) .. 292
PIPEs, L. A.: Applied ma.thema.tics for engineers and physicists, New York:
McGraw-Hill 1946 • . . . .• . . . . . . . . . . . . . . . . . . . . . .• . . . . . . .• . . . . . . . .• . . . 403
PoWELL, R. W.: Thermal conductivities of molten metals and alloys. J. Iron.

Steel Inst. 162 (1949) 315 ..........................................• 397
Plu.TL, J.: see RmnER
PmcE, M. J.: The role of the microscopic molten metal bridge in the electrical
erosion of contacts. Proc. Relay Conf. Oklahoma State Univ. (1965) Paper 26 [335
Plmux, W., L. H. FuCHS: Electrical conductivities of natural graphite
crystals. Phys. Rev. 90 (1954) 22 ....•........•.•...........•..•..... 409
PuLLEN, J.: [1] El.ectrical contact erosion. A theoretical study of pip and cra-
ter shape. E.R.A. report U/T 145 (1961) . . . . . . . . . . . . . . . . . . . . . . . . . 346, 356
- [2] Some methods of reducing erosion in light-duty electrical contacts. Proc.
Internat. Symposium on electric contact phenomena. Techn. Univ. Graz,
Austria (1964) 276, and similsr Proc. Univ. of Maine (1966) 305
Qunm, P.: Electroplated contacts. Engl. El.ectr. J.14 (1955) 48 ....... 165, 363
Qunm, T. F. J.: The preferred orientations withip. a contact film formed by·
the repeated sliding of an electrographite brush on a copper ring. Brit. J.
appl. Phys. 16 (1964) 513 ••.•••••...•...........•.....••..••••...... 222
RABINOWICZ, E., D. TABOR: [1] Metallictransfer between sliding metals: An

autoradiographic study. Proc. Roy. Soc., Lond. A 208 (1951) 455 . . . 158, 216
- [2] Metal transfer during static loading and impacting. Proc. Phys. Soc. B 66
(1952) 630 ....................................................... ·. 216
- [3] The boundary friction of very welllubricated surfaces. Lubr. Engng.
(July, Aug. 1954) 205 . . . . . . . . . . . . . . . . . . . . . . . . . . ... . . . . . . • • . . . . . 211, 212
- [4] The relation between friction and wear for boundary-lubncated surfaces;
Proc, Phys. Soc., Lond. B 68 (1955) 603 ............................. 211
- [6] A study of the stick-slip process. Friction and wear, 149. Proc. Sympos.
in Detroit 1957. Amaterdam, Elsevier, 1959 •.........•...•.••..•...... 229
- [6] Inftuence of surface energy on friction and wear phenomena. J. Appl.
Phys. 32 (1961) 1440 ....... : . .......................•.......•...... 379
- [7] Friction and wear of materiale. New York, J. Wiley & Sons, 1965 [199, 239,
- see KlNGSBURY 379
R.lDEKER, W.: [1] Der Verschleiß bei metallischer 'Gleitreibung, besonders·
seine Beeinfi.UBBung durch die Wärme. Arch. Eisenhüttenw.16 (1942) 453.. 235
- [2] Verschleißforll!en an Nichteisenmetallen bei trockener Gleitreibung.
Metallwirtsch. 23 (1944) 202 .................................... _. . . 235
RAETHER, H.: [1] Über die Struktur mechanisch bearbeiteter Oberflächen und
deren Eigenschaften. Z. Phys. 124 (1948) 286 .......................... 218
- [2] Elektroneninterferenzen und ihre Anwendung. Ergebn. exakt. Naturw.
24 (1951) 54 .............................•• ~ •.....•............••. 218
..... ·[3] Zur Struktur einer Metalloberftäche. Feinwerktechnik 67 (1953) H. 5 . . . 218
RAUB, E.: [1] Plattierte Edelmetallkontakte.Metallforsch. 1 (1946) 71 ....•• 356
- [.2] Die Kontaktwanderung bei Abhebekontakten aus Unedelmetalle ent-
haltenden Goldlegierungen. Metallforsch. 2 (1947) 281 ••.......•..•..•.• 356
RAYLEIGH, J. W.: mentioned................................. ... . . . . . ... 20
RECHE, K.: Fortschritte der Relaisentwicklung. ETZ 60 (1939) 753 • • • • • • . • . 302
REUSCH, H.: Magnetkreise/von hermetisch abgeschlOBBenen Kontakten in
Sohutzgasatmosphäre. NTZ (1959) 625 ....... : .. .................... 171
RHonlN, T. N.: Low temperature oxidation of copper. J. Am. Chem. Soc. 78
(1951) 3143 • : . ................•.............•••.•.... : . .......••.. '105
RICH, J. A.: [1] Resistance heating in the arc spot zone. J. Appl. Phys. 82
(1961) 1023 ..••••..••••.....•.•.....•.•.•...••....••...••...•••••..• 432
RICH, J. A., G. A. FARRALL: [2] Vacuumarcrecoveryphenomena. Proc. IEEE
52 (1964) 1293 ......•...................•.......................... 291
RICHTER, H.: Amorphe Strukturen bei Polier- und Einlaufvorgängen. Z. an-
gew. Physik8 (1956) 585 ........................................... 219
RICHTER, R.: [1] Untersuchungen über die Größe und Beständigkeit von Kon-
taktverbindungen unter besonderer Berücksichtigung des Aluminiums. ETZ
41 (1920) 345, 386, 409, 433, 448 ..................................... 169
-, G. SCHADE: [2] Untersuchungen über die Beständigkeit von Kontakt-
verbindungen unter besonderer Berücksichtigung von Aluminium. ETZ
59 (1938) 1321 ....................... '....................... 166, 169
- [3] Kontakte aus Kadmium-Silber und Silber-Kadmium-Oxyd. ETZ 11
(1959) 38
RIDDLESTONE, JANET: [1] Meta! transfer between palladium and silver con-
tacts at low inductances. Techn. report U/T 133. Brit. electrical research
Assoc. (1955) and Proc. Inst. electr. Engrs C 102 (1955) 29 ... 336,337, 341ff.
- [2) Short duration discharges between separating contacts in a 6 V circuit.
Brit. J. appl. Phys. 8 (1957) 105 .................................... 346
RIDDLER, K. E. W.: see BoWDEN
RIEDER, W., H. SCHNEIDER: [1] On the reignition of a-c arcs. Proc. Internat.
Conference on Ionization Phenomena in Gases. Uppsala 1959 . . . . . . . . . . . . 296
- [2] Switching arcs and related plasma physics. Proc. Internat. Sympos. on
Electric Contact Phenomena. Univ. of Maine (1961) 225 ................ 292
-, J. PRATL: [3] Das Löschen und Wiederzünden von Schaltlichtbögen.
Scientia Electrica 11 (1965) 33
- see MILLIAN
RIGHT.MIRE, B. G.: see FENG
RIEMANN-WEBER-FRANK-v. MlsES: Differentialgleichungen der Physik 2.
Braunschweig: Vieweg & Sohn 1927 .............................. 12, 93
RILEY, H. L.: [1] Chemical and crystallographic factors in carbon combustion.
J. de chim. phys. 47 (1950) 565 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 407
- [2] Discussion of Rosalie Flanklin's paper. J. de chim. phys. 47 (1950) 573 . 407
R1s, S.: see VAN VLIET
RoACH, A. E., C. L. GoonzEIT: Why bearings seize. Gen. Motors Engin. J.
(Sept.-Oct. 1955) 25 .......................................... 158, 220
RoARK, R. J.: Formulas for stress and strain. New York: McGraw-Hill1954 [29,
267, 367ff.
ROBERTS, R. W.: Generation of clean surfaces in high vacuum. Brit. J. Appl.
Phys. 14 (1963) 537 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
RopERTS, W. H.: see CoRNELIUS
ROBERTSON, B. P.: see BoYD
RoEss, C. L.: Theory of spreading conductance. Unpublished report, Beacon
Laborat. ofTexas Co.................................... ·. . . . . . . . . . . 26
RoGERS, G. F. C.: Heat transfer at the interface of dissimilon metals. Int. J.
Heat Mass Transfer. 2 (1961) 150 .................................... 196
RoHMANN, H.: [1] Elektrische Kontakte. Phys. Z. 21 (1920) 417, 699 ....... 115
- [2] Über den elektrischen Kontakt zwischen glühender und kalter Elektrode
im Vakuum. Phys. Z. 34 (1925) 94 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
ROHSENOW, W. M.: see FENECH
RoJANSKI, V.: Introductory quantum mechanics. New York: Prentice-Hall1949 [ 119
RoLF, E.: Der Kontaktumformer. Berlin: Springer 1957 .................. 296
RöNNQUIST, A., H. FiscHMEISTER: [1] The oxidation of copper, a rev. of data.
J. Inst. of Metals (Oct. 1960) 65 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105, 112
RöNNQUIST, A.: [2] The oxida.tion of copper. J. Instit. Metals 91 (1962) 89 [105, 112
RoSHON,D.D.: sooBAYER
Ross, H. C. : Switching in high vacuum environment. IEEE Trans. Professio-
nal techn. group on component parts. CP-10 (1963) ••.................. 291
RowE, G. W.: Friction and metal-transfer of heavily-deformed sliders. Wea.r
7 (1964} 204 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 234
- see G&EENWOOD
RücHARDT, E.: see DIETRICH
- see KAPPLEE
Ruu, R. J.: A study of nickel surfaces employed in the catalytic decomposi-
tion of formic acid vapor. Thesis Univ. of Michigan, Apr. 1954 ........... 105
RusSELL,M.N., S.KEILIEN: The vibration of electr. contacts. Trans. AIEE 83
( 1944) 153 . . . . .• . . . . . . . . . .• . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 302
RuTGERS, C. A. W.: see ter HoRST
SALINGER, H.: Beobachtungen am Kohlemikrophon. Elektr. Nachr.-Techn. 6

(1929) 395 . . . . . . . . . . . . . . . . . . . . . . . . . . .• . . . . . . . . . . . .• • . . . . . . . . .• • . . . 184
S.AMUELS, L. E.: [1] The nature of mecha.nica.lly polished meta.l surfaces. J. In-
stit. Metals (Oct. 1956) 51. . . . . . . . . . . . . . . . . . . . . . . . . • . . . . . . . . . . . . . . .• . 219
- [2] Damaged surface la.yers, meta.ls. The surfa.ce chemistry of metals
and semiconductors. Edited by H. C. Gatos. J. Wiley & Sons, New
York 1959
SANTOS, G. J.: see CHAIKIN
SAUNDERS, J.: see J. ANDERSON
SAVAGE, R. H.: [1] Carbon-brush conta.ct films. Gen. Electr. Rev. 48 (1945) 13
- [2] Graphite lubrica.tion. J. Appl. Phys.19 (1948) 1 .................... 226
-, D. G. FLoM: [3] Exploration of metal surfa.ces with fine wires. Ann. New
York Aca.d. Sei. 68 (1954) 946 •.•.....•........•............•....•..• 152
-, D. L. ScHAEFER: [4] Vapor lubrica.tion of sliding contacts. J. appl. Phys.
27 ( 1956) 136 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 226
- see FLoM
- soo VAN BBUNT
SCHADE, G.: 800 RICHTER
SOHAEFER, CL.: Einführung in die theoretische Physik. 3. edit., 2., Berlin: de
Gruyter 1929 . . . . • . . . . . . . . .• • . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .• 378
ScHAEFER, D. L.: see SAVAGE
ScHLll!ER, R. : 800 K.AFPLER
ScnnnLIN, F. W.: Enha.nced tunneling through dielectric films due to ionic
defects. J. Appl..Phys. 37 (1966) 2823 ................................ 127
ScHMIDT, E.: see Kol'l'LIN
ScHNEIDER, H.: see RIEDE&
ScHOLTZ, S.: Un.tersuchungen zum Einfluß des Peltier-Effektes auf Wider-
stands-Schweißungen. Z. angew. Phys. 12 (1960) 111 . . • • • . . . .• . • . . . .• . . 84
ScHOTiXY, W.: [1] Vber kalte und warme Elektronenentla.dungen. Z. Phys.
14 (1923) 63 ...................•..............•......•..........•. 389
- [2] Zur Halbleitertheorie der Sperrschicht- und Spitzengleichrichter. Z.
Phys.113 (1939) 367 .................•..••••...•...•.............•.• 391
- [3J Artikel "Elektronenbefreiung" in: Das freie Elektron in Physik und
Technik. Editor C. Ramsa.uer. Berlin: Springer 1940 ................... 389
- edited [4] Halbleiterprobleme. Bd. I (1954), Bd. II (1955), Braunschweig:
F. Vieweg & Sohn
- mentioned . . . • . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 270
ScBRAG, G., et al.: Beobachtungen an flüssigen Kontaktbrücken. Metallkunde
1,38 (1947) 25; 11,42 (1951) 24; m, 42 (1951) 243 ..................... 338
ScHREINER, H.: Pulvermetallurgie elektrischer Kontakte. Berlin: Springer 1964 [362
SCBROEDER, C. F. JR.: Getting the most from your ultrasonic cleaner. SC. TM
345---63 (11). Sandia Lab. (1963) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
SCHRÖTER, F.: [1] Zur Physik des Schleifkontaktes. Arch. Elektrotechn. 18
(1927) 111 ............................... ......................... 245
- [2] Der Einfluß der Temperatur auf einen Gleitkontakt. ETZ-A 85 (1964) 568
SCHULTZ, H.: see JusTI
SCHUMANN, W. 0.: Elektrische Durchbruch-Feldstärke von Gasen, Berlin:
Springer 1923 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 276
SCHWARZ, W.: see TonT
ScoTT, R. A.: see JoNES, T. K.
SEEGER, A.: Theorie der Gitterfehlstellen. Handbuch der Physik. VII/1 ( 1955)
383 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
SEITZ, F.: The modern theory of solids, New York and London: McGraw-Hill
Book Co. 1940 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 388
SHAPRIN, ELAINE G.: The lubrication properlies of monomolecular films ad-
sorbed on solid surfaces.ReportNRL. U.S.NavalRes. Lab., Wash. July 1958 205
SHAW, M. C., E: F. MACKS: Analysis and lubrication of bearings. New York,
McGraw-Hill Book Co. 1949 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 413
SHAw, N. C.: Investigations of contact bounce and contact welding. Internat.
Symposium on electr. contact phenomena. Graz, Austria (1964) 286 301
SHIBATA, H.: see HmABAYASm
SHIBLEY, D. G. M.: see FAmWHETHER
SHLYKOV, Yu. P., E. A. GANIN: Thermal resistance of metallic contact. Inter-
national J. Heat Mass transfer 7 (1964) 921 .......................... 196
SHO:BER'l, ERLE I.: [1] Calculation of electrical contacts under ideal conditions.
Amer. Soc. for Testing Materials, Proc. 46 (1946) 1126.................. 46
- [2] Electrical resistance of carbon brushes on copper rings. Pwr. Apparatus
and Syst. (Aug. 1954) 788. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
-, J. E. DrEHL: [3] A new method of investigating commutation as applied
to automotive generators. Pwr. Apparatus and Syst. (Febr. 1955) 1592 ... 254
- [4] Carbon brush friction and chatter. Pwr. Apparatus and Syst. (June
1957) 268 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 228
- [5] Commutation. Pwr. App. and Systems (Dec. 1962) 594
- [61 Carbon and Graphite. New York, Academic Press 1964 ......... 404, 407
- [7] Carbon brushes. New York, Chemical Publ. Co. Inc. 1965 ... 229, 252, 258
- [8] Recent advances in sliding contacts including space applications. IEEE
Trans. on P. M. and P. 1966 .......................•....... .......... 226
SIE:BEL, E., R. Ko:siTZSCH: Verschleißerscheinungen bei gleitender trockener
Reibung. Berlin: VDI-Verlag 1941 ............................... .... 237
SILLABS, R. W.: see T. K. JoNES
SnmoNs, J. G.: [1] Generalized formula for the electric tunnel effect between
similar electrodes separated by a thin insulating film. J.Appl. Phys.34 (1963)
1793 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122, 123
- [2] Electric tunnel effect between dissimilar electrodes separated by a thin
insulating film. J. Appl. Phys. 34 (1963) 2581 . . . . . . . . . . . . . . . . . . . . 122, 128
SKAUPY, F., 0. KANTOROWICZ: Die elektrische Leitfähigkeit pulverförmiger
}letalle unter Druck. Metallwirtsch. 10 (1931) 45 ...................... 191
SLATER, J. C.: [1] Introduction to chemical physics. New York and London:
:YicGraw-Hill Book Co. 1939 ............................... ......... 378
30*
472 Author and Iiterature fudex
SLATEB, J. C.: [2] Quantum theoryofmatter.NewYorkandLondon:McGraw-

Hill Co. 1951 . . . . . . . . . • . . . . . . . . . .• . . . . . . . . . . . . . . . . . .• . . .• • • . • 103, 153
SLEJ.>IAN, J.: [1] Tamperature of a contact and related current interruption
problems. J. Amer. electr. Engrs. 40 (1926) 930
_ [2] Theory of current transference at the cathode of an arc. Phys. Rev. 27
(1926) 407
- [3] Extinction of an A.C. arc. Trans. AIEE 47 (1928) 1398 ........ 290, 295
- [4] Theory of the Deion circuit breaker. J. AIEE 48 (1929) 93 ....•....• 295
- [5] Extinction of a long a.c. arc. AIEE Trans.119 (1930) 521 ...... 290, 295
_ [6] Displacement and diffusion in fl.uid-fl.ow arc extinction. Electr. Engng.
Tra.nsact. 60 (1941) 162
SMEKAL, A.: Notices about "Mikroplastizitä.t und Mikrofl.ießhärte" in Sitzungs..
berichte der Osterreichischen Akademie der Wissensch. June 11, 1953; Jan.
28, 1954; April 21, 1955 • . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 242
SMITH, G. P.: see GwATHMEY
SMITH, J. L.: see GERMER
SMYTHE, W. R.: Static and dynamic electricity. New York and l.ondon:
McGraw-Hill 1939 . . . . . . . . . . . . . . . . . . . . • . . . . . . . . . . . . . . . . . . . .• . . .• 12, 15
SNOWDON, A. C.: [1] Some factors infl.uencing a.lternating current interruption.
Proc. Seminar on electrical contacts, Penn. State Univ. June 1957 . . . • • . 296
- [2] Studies of electrodynamic forces occurring at electrical conta.cts. Appl.
and Industry (March 1961) 24 . . . . . . . . . . . . . . . . . . . .• . . . . . . . . . .• • • . . . . . 56
- [3] Low power a.-c interrupting characteristics of ca.dmiumoxide silver con-
tacts. Internat. Symposium on electric Contact Phenomena. Univ. of Ma.ine
(1961) 367 ............•..........•....................•.....••.... 296
SoMMERFELD, A., H. BETHE: Elektronentheorie der Meta.lle. Handbuch der
Physik von GEIGE& und SCHEEL, 2. Auf!.. 24/2. Berlin: Springer 1933 .••. 119ff.
SPEAR, W. E., H. P.D. LANYoN, J. MoRT: Someapplicationsofmercurywetted
conta.ct relays. J. Sei. Instrum. 89 (1962) 81. ••....•...•••..••. ·•.....•• 358
SPERGEL, J., E. GooDWIN, G. STEINBERG: A thin film lubricant for connector
contacts. Proc. Electronics components conf., IEEE, May 1965 ...... 219, 220
S~,J.A.: seeGREGORY
STAUBS, H. E.: see MARTIN
STEBBENS, A. E.: Wear and current tra.nsfer processes at the carbon-copper
interface. Internat. Symposium on Electric Conta.ct Phenomena.. Techn.
Hochschule Graz, Austria. (1964) 396 .••...•.•.........•.•......•....• 249
STEIJN, R. P.: Sliding wea.r and metal transfer under unlubricated conditions.
Tra.nsact. ASME. J. Basic Engineering (March 1959) 67 . . • . . . . 232, 234, 235
STEINBERG, G.: High reliability connective devices. Sta.nford Res. lnst. Semi-
a.nnua.l Report No. 11, Feb. 1965.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219, 220
- see SPERGEL
STILWELL, N. A., D. TABOR: Elastic revovery of. conica.l indentations. Proc.
Phys. Soc. 78 (1961) 169
STINE, W. E.: Brush friction greatly affected by conta.ct air pressure. Electr.
Wld. 88 (1926) 67 . . . • . . . . . ... • • . • • . • . . . . . . . . . . . .• • . .• • . • . .• • • • • • • • • 259
STOCK, A. J.: Graphite, IJWlybdenum disulfide and PTFE. Lubrica.tion En-
gineering 19 <1963) 333 ••..•..••...•••...... ~ . . . • . • • . .. . • . . . . . .• • • • . 221
:SToimY, C.: Investigation into one of the assumptions of Hertz theory of con-
ta.ct. Brit. J: Appl. Phys.ll (1960) 67 •.••••.•••••......•.•••...• 368~ 369
:STÖBliiER, R.: [1] Eine Methode zur Messung der Wärmeleitfähigkeit und des
Temperaturkoeffizienten der Wärmeleitfähigkeit elektrisch leitender Kör-
per. Wisft.. v",~öff, Siemens-Wc"k 18/1.(1934) 30 • . . . . . . . . . . . . . . .• . . . . . . 91
STäRllfEB, R.: see R. HoLM
- mentioned ....•................................................... 144
STRANG, C. D.: see BUBWELL
STBA'l"l'ON, R.: Volt-current characteristics for tunneling through insulating
:films. J. Phys. Chem. Solids 23 (1962) 1177 •.................... 122, 127
STROM, ,A.. P.: see L!NGAL
STBÖMBERG, R.: Adsorptionsmessungen mit einer verbesserten Mikrowaage.
Kgl. Svenska Vet. Akad. Handlingar Stockholm 6 Nr. 2 ( 1928) - also: Phys.
Ber. 11 (1930) 602 ....•..•...•...............•.........•............ 114
STUBSTAD, W.: Thermal contact resistance between thin plates in vacuum.
Thesis, 1964. State Univ. of Iowa . . . . . . . • . . . . . . . . . . . . . . . . . . . . . . . . . . . 196
STUCKES, AUDBEY D.: Some theoretical and practical considerations of the
Johnsen-Rahbekeffec'".Proc.Instn.Electr.Engrs.,Lond.103 (1956) 125 [59, 60
STUDEBAKEB, M. L.: Observations on the reaction between carbon black and
waterat lowtemperatures. Third Carbon Confer. Univ. of Buffalo (1957) 289 [226
STUELPNAGEL, T. R.: Electrostatic oscillation in sensitive contact mechanisms.
Application and Industry (Jan. 1953) 397 ............................. 363
- seeDALLAS
SUITs, C. G.: Arcing phenomena in mercury switches. Gen. Electr. Rev. 43
(1940) 120 .............•.......................................... 357
SUN, J. K.: see GLOSSBBENNEB
SuNDOBPH, TH.: Über die Bil.dung leitender Brücken an der Stelle, wo ein
Strom von geringer Spannung unterbrochen wird. Ann. Phys., Lpz. 10
(1903) 198 .......••..•......•...................................•. 338
SWIOKEBT, M. A.: see BISBON
SWINEHABT, M. R.: [1] Heavy current welding in electrical contacts. Proc.
Electrical Contact Seminar, Pennsylv. State Univ. (June 1956)
- [2] Ares and arc erosion in high current contacts. Proc. Electrical Contacts
Seminar. Pennsylv. State Univ. (1959) 55 ..••.•.•................ 308, 310
- [3) The effect of oxygen on the erosionproperlies of silver contacts. Proc.
Internat. Symposium on Electric Contact Phenomena. Univ. of Maine ( 1961)
301 • • . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 309, 310
- mentioned • .. .. .. .. .. .. .. .. .. .. . . . .. . . . . . .. . . . . .. . . . . .. . .. . . .. . .. . 429
SWINNEBTON, B. R. G., M. J. B. TuRNER, J. E. THoMPSON: Arcing at Copper-
graphite interfaces. Proc. Conf. on Commutation, Inst. Electr. Engineers,
London (1964) 68 .••.•••••••..•.......•..•••••.••...•......•••••••• 263
SZEKELY, ANGELICA: Über die Art ·des Elektrizitätsüberganges zwischen
Metallen, die sich lose berühren. Z. Phys. 22 (1924) 51 ••...••.......... 338
TABOB, D.: The hardness of metals. Oxford: Clarendon Press 1951, [367,369, 372,
- see BOWDEN 373, 376
- see GREENWOOD
- see McFABLANE
- see MENTER
- see MULHEARN
- see RABINOWICZ
- see 8Trr.WELL
TABPINIAN, A., C. TEDMON: An analysis of carbon crystallite growth kinetics
during graphitization. Techn. Raport Wal TR 851 Ö/1 (1962) Watertown
Arsenal. . . . . . . . .• . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 407
TAYLOB, L. H.: see BBYANT
TAYLOB, W. G. A.: see CoLLIER
TEDHOLM, C. E.: see RABINOWICZ

TEDMON,C.: see TARPINIAN
TE:& Ho:&sT, D. TH. J., G. A. W. RuTGE:&S [1]: Current-zero phenomena in a.c.
arcs in air-blast circuit breakers. CIGRE (1956) paper 122 ....•....•..... 291
- [2] Current-zero phenomena in a.c. air-blast circuit breakers. Proc. Internat.
Symposium on Electric Contact Phenomena. Univ. of Maine (1961) 385
- mentioned . ... .. . . .. .. .. .. • .. . . .. .. .. .. . .. . .. .. .. .. . .. .. . . . .. . .. . 292
TESZNE:&: mentioned . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 295
THoMAS, U. B.: [1] Contact resistance characteristics of various metals before
and after tarnishing for 12 weeks in Barclay central office. Bell. Tel. report.
Apr. 1937 ......................................................... 170
- [2] Bell Tel. report 1937. Personal communication ................ 110, 152
- [3] The use of electroplated metal in static contacts. Internat. Symposium
on Electric Contact Phenomena, Univ. Maine 1961, 109 ................ 170
- see CAMPBELL
THOMl'SON, J. E., M. J. B. TURNE:&: (1] The part played by a metal oxidein
determining the characteristics of a graphite-meta} interface. Wear 6 (1963)
30
- [2] A review of work at the Nelsonresearch laboratories on current trausfer
and commutation in electrical machines. Internat. Symposium on Electric
Contact Phenomena. Techn. Hochschule Graz, Austria (1964) ........... 252
see SWINNE:&TON
TH:&OSSEL, W. R.: see BownEN
TINGLE, E. D.: Fundamental work on friction, lubrication, and wear in Ger-
many during the war years. J. Inst. Petrol. (Oct. 1948) 743 ............. 417
TITus, C. H.: see LEE
Tonn, F. C., T. E. B:&OWNE, J:&.: Restriking of short A. C. arcs. Phys. Rev.
36 (1930) 732 ..................................................... 296
ToDT, F., R. FREIER, W. SCHWARZ: Die Messung der Oxydationsgtischwin-
digkeit und Oxydschichtdicke von Metalloberfiächen sowie der Lokalelement-
tätigkeit zwischen Metall und Metalloxyd. Z. EI. chem. und angew. phys.
Chemie 53 (1949) 132 .............................................. 105
TÖFFLINGE:&, K.: [1] Unterteilte Kohlenbürsten. ETZ 73 (1952) 573 • . . . . . . . 267
- [2] Der Bürstenkontakt beim Wechselstrombahnmotor. Elektr. Bahnen 24
(1953) Heft 2 ................................................. 257, 266
- [3] Neue Erkenntnisse bei Wechselstrombahnmotoren. Siemens Zschr. 27
(1953) 73
TOWNE, A.: Some problems in lubrication and the substances called additives.
Proc. Automob. division. Inst. mech. engrs. No. 2 (1954-55) 57 ............ 321
ToWNSEND, J. S.: mentioned . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 276
ToYADA, H. H.: see HI:&.ABAYASID
TruPP, J. H.: see G:&EENWOOD
TSUZUKO, T.: Dislocations in graphite crystals. Proc. 3rd Conference on Carbon,
Buffalo. Pergamon Press New York (1959) 433 .................... 222, 405
TUBANDT, C.: tlber einseitige Ionen- und gemischte Stromleitung in Kristal-
len. Z. Elektrochem. 26 (1920) 358 . ... . . . .. .. .. . .. .. .. .. . .. . . . . . .. .. . 150
TUBNER; M. J. B.: see SWINNERTON
TYLEOOTE, R. F., D. HoWD, J. E. E'u:&MmGE: The infiuence af surface films or:
the pressure welding of metals. Brit. weldg. J. (Jan. 1958) 21. ........... 159
UBBELOHDE, A. R., F. A. LEWIS: Graphite and its crystal compounds. Oxford,

ClarendonPress, 1960 ......................................... 184, 404
ULSH, H. B.: Electric characteristics of oxide films on copper base materials
before electrical breakdown and contact resistance after electrical break-
down. Proc. Seminar on electric. contacts. Pennsylvania State Univ. 1957. 136
UNSÖLD,A.: Physik.der Sternatmosphären. Berlin: Springer 1938 ......... 421
VAN BRUNT, C., R. H. SAVAGE: Carbon-brush contact films. Gen. Elektr. Rev.
47 (1944) 16 ............................................. 222, 223, 226
VAN DER HELD, E. F. M.: see BoESCHOTEN
VANDERSLICE, T. A.: see CoBINE
VAN DER ZIEHL, A.: Thermal noise at high frequencies. J. Appl. Phys. 21
(1950) 399 ........................................................ 270
VAN HoRN, R. H.: see MALL!NA
VAN LEEUWEN, C. J.: see VAN VLIET
VAN NYMEGEN, D. W.: see BuRWELL
VAN VLIET, K. M., C. J. VAN LEEUWEN, J. BLOK, C. R1s: Measurements on
current noise in carbon resistors andin thermistors. Physica 20 (1954) 481 . 270
VAUGHN, G. W.: see JoHNSON, V. R.
VERNON, W. H. J.: [1] Second experimental report on the atmospheric corro-
sion research committee. Trans. Faraday Soc. 23 (1927) 113 ........ 105, 109
- [2] A laboratory study of the atmospheric corrosion of metals. Trans.
Faraday Soc. 27 (1931) 255, 582; 31/2 (1935) 1668
- [3] Metallic corrosion. Research 5 (1952) 54 ........................... 116
VIEHMAN, H.: [1] Klemmverbindungen von elektrischen Leitungen aus Leicht-
metallitze. Jb. dtsch. Luftf.-Forschg. Teil III (1938) 24 . . . . . . . . . . . . . . . . . 169
- [2] Leitungen-, Klemmen- und Funkenstörung im Flugzeug. VDE-Fach.
Ber. 11 (1939) 70 .................................................. 169
VINARICKY, E.: Thermokraft und elektrischer 'Widerstand einiger Kontakt-
werkstoffe auf Edelmetallbasis. Metall 20 (1966) Heft 7
- see MERL
VINES, R. F.: Progress on noble meta! contact materials. Proc. Internat. Sym-
posium on Electric Contact Phenomena, Techn. Univ. Graz, Austria (1964)
475 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170, 362
VoGELPOHL, G.: Zur Klärung des Gleitreibungsvorganges. Öl u. Kohle 37
(1939) 720 ........................................................ 417
VoLKMANN, W.: Über Kohlebürsten-Schwingungen mit kleiner Amplitude.
ETZ A 86 (1965) 151 ......................................... 229, 272
VON MISES, R. : mentioned . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 370
WAGAR, H. N.: Some measurement techniques for characterizing sealed reed

contact performance. IEEE Transact. on component parts. (Dec. 1964) 8 . . 171
WAGNER, C.: see ILSCHNER
WALKER, P. L. JR.: Carbon- an old but new material. Amer. Scientist (June
1962) 259 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 404, 406ff.
WALKUP, L. E.: Renaissance in Electrostatics. Batteile techn. Rev. 5 (1956)
No. 2, 3 ........................................................... 190
WARBEN, B. E.: see HousxA
WEDELL, H. M.: see BURKHARD
WENT, J.J.: [1] Adsorptionserscheinungen beiMetallkontakten. Philips techn.
Rdsch. 5 (1940) 245 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
- [2] Adsorption phenomena on massive meta! surfaces measured by means
of eleetrieal contact resistances. Physica 8 (1941) 233 .................. 126
WERNER, E.: see BuRKHARD
WHITE, A. H., L. H. GERMER: The rate of oxidation of copper at room tem-

perature;Trans. electrochem. Soc. 81 (1942) 305 ....................... 105
WHITE, C. E.: see LILIENFELD
WHITE, D. R.: see LEE
WHITE, J. R.: see 0. H. CLARK
WHITEHEAD, J. R.: Surface deformation and friction of metals at light loads.
Proc. Roy. Soc., Lond. A 201 (1950) 109 ......................... 201, 206
WHITEHEAD, S.: Dieleetrio breakdown of solids. Oxford: Clarendon Press 1951 [146
WHITTAKER, D., H. EDELS: Electrode infl.uence on arcing and gap recovery.
Proc. Internat. Symposium on Electric Contact Phenomena. Techn. Univ.
Graz, Austria (1964) 239 ............................................ 296
WmTTINGHAM, G.: see HEATON, also HuGHES
WILLIAMSON, J. B. P., J. A. GREENWOOD, J. HARRIS: (1] Theinfl.uenceofdust
partieleB on the contact of solids. Proc. Roy. Soc. A 237 (1956) 560 .. 172, 173
- [2] Significance of surface topography in electric contacts. Internat. ·conf.
electromagn relays, Oct. 1963, Sendai, Japan, paper A. 7 . . . . . . . . . . . . . . . 166
- [3] Basic properties of electric contacts. Proc. Seminar on electr. contacts.
Univ. of Maine June 1965
- see GREENWOOD
WILMAN, C. W.: Electric contact resistance between magnetically attracted
bodies. Nature 172 (1953) 917 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
Wn.soN, R. W.: The contact resistance and mechanical properties of surface
films on metals. Proc. phys. Soc., Lond. B 68 (1955) 625 ........ 205ff., 212
Wn.soN, W. R.: [I) Life of silver-surfaced contacts on repetitive arcing duty.
Pwr Apparatus and Syst. (Deo. 1953) 1236 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 363
- [2] High current arc erosion of electric contact materials. Pwr Apparatus and
Syst. (August 1955) 657 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 308, 366
- see CoNNoR
WoLAK, K.: Schutzgaskontakte und Schutzgaskontakt-Relais. Siemens Zschr.
32 (1958) 845 ..........•..............•••........••............... 171
WoLLENEK, A.: Die Veränderung der Kontaktberührungsftäche bei hohen
Temperaturen. Zschr. angew. Phys. 12 (1960) 360 ..................... 436
WooDs, W. W.: see 0. H. CLARK
WRIGHT, R. H., M. J. MARBHALL: The effect of adsorbed gas on the contact
resistance of carbon. Trans. electrochem. Soc. 54 (1928) 149 . . . . . . . . . . . . . 51
WYSONG, W. S.: See GULBRANSEK
YANKEE, W. R.: see MA.cHLIK

YoHE, W. E.: see HoLM
YoUNG, F. W. JR., J. V. CATHCART, A. T. Gw.ATHMAY: The rates of oxidation
of several faces of a single crystal of copper as determined with elliptically
polarized light. Acta 'metallurgica 4 (1956) 145 . . . . . . . . . . . . . . . . . . . . . . . . ·107
ZIELABEK, G.: see KüHLER

ZISMAN, W. A.: Friction, durability and wettability properties of nionomole-
cular.films on solids. Friclion and wear, Amsterdam Elsevier 1959, 110 .... 205
Subject index
abrasive wear 232 are and ohrnie resistance in parallel 285
aeceptors 187, 384 -, oseillations 298
aetivated contaet 277 -, plasma 290ff.
aetivation energy 39, 104, 112, 153 - queuehing 311, 325
adaptation of eontaet surface 139, 244, - reignition 290ff.
248 - with semieondueting resistor in
additives to lubrieants 220 parallel 326
adherence 50, 153, 160ff. -, short 280
- foree 162 - shortest 280, 282
- in eurrent-heated eontaets 160 - types in relays 334
- in vaeuum 153ff. - vacuum 286ff.
- voltage 160ff. - VI-charaeteristies 279ff.
adhesive wear 232 asperity, amplitude 24, 32, 43
adjuvants 227 a-spot (a-Fläehe) 8
affinity 390, 428 -, determination 34
A-fritting 136ff. - generation by fritting 6, 135
aging (normieren) 7, 88 - generation by plastic deformation 6,
air blast breaker 291 48, 135
alien illms 102 -, number and shape 33ff.
alumina 107 -, widening by fritting 137-
aluminum 106ff., 164, 168, 188 ASTM361
anode 1, 424, 433 attraetion in contaet 57
- are 306, 337 Auger effect 275, 287ff., 423
- dominated arc 305ff., 337 avalanehe 135, 146, 275
- - transfer 306
antimony 188, 192 ball bearing 221
apparatus 175, 242 band theory 380ff.
apparent contaet area, barium fluoride 227
see contaet area barrier, see boundary barrier
are, anode fall 426 BEILBY layer 218
-, blow out 295 B-fritting 137ff., 147ff.
-, at bouneing 303 bimetal contacts 65ff., SOff.
-, at contact elosure 276 blocking direetion 187
-, eathode fall423ff. blow-out of arcs 295, 365
- eurrent below I,. 433 boiling voltage 64
-,drawn 278 bolted junction 40, 194
- duration 310, 323 - - thermal 194
- during com.mutation 263 bond 153 ff., 156, 223 ff.
- extinetion 297 bouncing 300
-, flaming 306 boundary barrier 388, 391
- ignition 275 - lubrieation 209ff.
- instability 286 brass109
-, minimum eurrent 281, 283 T, 433 breakdown of dielectries 135, 146
-, minimum voltage 281, 283 T, 434 - of gap 275ff., 331
478 Subject index
breakdown of films, see fritting oompliance 368

breathing of contacts 111, 179, 192 compound materials 311
bridge transfer 304, 338 constantan 242
- liquid 339 oonstriction (Stromenge) 9ff.
- molten 344, 349 -, distorted 20ff.
- short 339 - long- 11, 18
- long 339 - b-model3
BRINELL hardness nurober 375, 439 - resistence 2ff., 9ff.
brittle materials 218 - - calculation 3ff., 11ff.
bronze 109 - - and Joule heat 71ff.
brush types 256, 258 -, shape factor 18
Burgers vector 370 - -, thermal193ff.
hurst 287 ff. - -, at ultra low temperature 125
bus bar 165, 176 - -, verification 76
-, short- 11
Cadmium oxide 296, 309, 364 - voltage 9
calcium stearate 216 contact 1
capacitance in contact 57 contact area 2 ff., 7
carbon 404ff. - -, apparent 2
- baked 407 - -, conducting 2
- against copper 85 ff. - -, Ioad bearing 1ff.
- -nicke! 86, 145 -, change with time and temperature 39
- brush 33, 150, 221ff., 258 - hardness 2, 374
- brush chatter 228 - Ioad 1, 7
- contacts 41 - material362ff.
- grain microphone 51, 183 - member (Kontaktglied) 1
-, hard 407 - micro 360
-, thermal conductivity 410 - models 3, 11
- rheostat 190 - noise, see noise
castor oil 211 - open 1
cathode 1, 423ff. -, preloaded 4 7
- arc 306 - rectifier 185ff., 296
- dominated arc 305 - resistance 3, 9ff., 47
- fall, see arc - -, calculation 11ff.
- spot 305 - -, change with temperature 71ff.
- sputtering 310, 396 - -, change with time 37 ff.
cathodic layer 424 - - measurements 3, 41 ff.
cell 381 - -in slinding contact 242ff.
centra! battery 181 - - at very smallload 48ff.
cessation voltage of fritting 136 - - vs Ioad 40 ff.
chemisorbtion 27, 102 - - rule of thumb 47
chopping current 273, 289 - pressni-e 34, 207
circuit breakers 290, 365 - reliability 359
clean contacts, practically- 27 - shape 307, 362ff.
cleaning of contacts 28 - spot 1, 8
ooin silver contacts 361 - -, multispots 22ff.
cold weid 48, 153ff. - surface, see contact area
- - in sliding contacts 201 - temperature, see supertemperature·
oollector film 2.t3 ff. and also temperature
oommutation 253ff., 260 - type, bimetallic 68
- dependent 266 - -, bolted 194, 363
- ideal260 ...:_ -, bus-bar 165
Subject index 479
contact type, carbon brush against oop- electron emission, see field and ther,
per ring 242ff. mionic emission
- -, crossed rod 41ff., 368 end surface of a constriction 11
- -, :flat 30, 194 energy band scheme 380ff.
- -, graphite brush on graphitering 207 epilamen 174
- -, Imife switch 302 evaporation in an arc 309
- -, mercury and solid metal108 excess holes 410
- -, between metal and semioonductor ex4austion layer 187
185ff. expansion switch 295 [385ff.
- -, permanent 363 extrinsic conduction (Störstellenleitung)
- -, plug and socket 363
- -,repetitive 364 FERMI Dirac distribution 383
- -, twin- 361 - Ievel in metals 382
- valtage 3ff., 9 - - in semicondilctor 185, 390
- - and supertemperature, see tp{}- tp{}-relation 60ff.
relationship field emission 120, 130
cooling of oonstriction 100 films 27, 69, 102, 197, 222, 245ff.
copper106, 112,113 fine transfer 304
COULOMB law of friction 201 :flash, see voltage flash
creep (Kriechen) 39,. 89, 109, 377 floating 279, 320
critical arc length 336, 337 forward direction 187
crossed rod oontact 42 ff. free electrons 381
cupraus oxide 112, 393 friction 200ff., 377
current density 15, 71 -, adhesion, see adhesion friction
current collector 268 - coefficient 200
- oonstriction, see constriction - force (total) 199
frictional polymer 117
Debye temperature 394 fritting 135ff.
debris 240ff. - applied 147
deformation, elastic 30, 367 ff. -, cessation voltage 145
-, plastic 30, 367ff. -, field strength 142 ff.
de-ion breaker 295 frying 363
diagramm XI 284, 441 r 307, 309, 431, 440
dielectric strength recovery 290 ff. gas-blast brea.ker 291
diffnsion effects 377 Gaussian distribution ( = normal distri-
dimensioning a contact 176 bution) 31, 398
direct current, break'ng 297 germanium 370, 382, 384ff., 393
discb.arge transients 313 glow discharge, seealso discharge
dislocations 369 gold 106, 109
distribution 382, 397 ff. - plating 111
donors 187, 384ff. - against wolfram 361
drift velocity 380 graphite 404ff.
dryness and wear 225 - hardness 375ff.
dynamic characteristic of the arc 299 - lattice 405
- platelets 222, 252, 405
effective mass 385 - with chemisorbed film 46, 51
elastic deformation 30, 367 graphitization 407
electrical breakdown of gas 275ff. graphitizing carbons 407
electric contact 1 grooving friction 379
- oscillations 298ff.
electrographite 408 Hall effect 386
electron affinity 390,428 · hardening, see strain hardening
480 Subject index
hardmetal 365 mercury switch 356ff.

hardness 154, 372ff. - contact 8
- number 375, 435 - jet contact 359
heat capacity 393ff. - sliding contact 358ff.
- conduction 393ff. metal graphite brush 258, 259
- treatment of carbons 409 - powder 192
HEBTZ area 36 metallic conduction 382, 397
high altitude effect 225 - contact, see also a-spot 2, 9, 11ff.
histogram 398 MEYER hardness 374
hole conduction, see p-conduction mica 204
holes (Defektelektronen) 385 micro contacts 165, 172, 355, 360
humidity 114 -film 102
hydrodynamic lubrication 412 ...;.. relay 363
hydrodynamic air pressure under a microphone 180ff.
brush 259, 358 migration of silver in insulators 150
minimum current, see arc
1-effect 277, 287 - voltage, see arc
image force 121, 388 mixed lubrication, see boundary
indium 211 lubrication
- antimony 393 mobility 386
inductance of constriction 52 moisture, see humidity
instability of the arc 286 molybdenum 186
interuption of A.C. 290ff. - wool273
intrinsic conduction (Eigenleitung) - sulfide 156, 227
385ff. mono-plane 406
MTBF 360, 402
IoHNBON noise, see noise multispot contact 21 ff.
JOHNSEN-RAHBEK effect 59
n-conduction 385
Knoop hardness 375 nickel 106, 108ff., 113
KoHLER effect 69, 344 noble metals 363ff., 437
- - see also Author Index 467
lacquer 146 - -, tarnishing 112
Ioad 1, 7 noise 269ff., 402
- bearing area 7, 29 ff. - in brush-ring contact 229, 269
local battery 181 - spectrum 273, 402
- cell (Lokalelement) .114ff. , non-graphitizing carbons 407
long bridge 339 normal electrodes 307
- constriction 11 - distribution 398ff.
LORENZ function 396
lubrication 209ff. (I) 307
oil breaker 295

magnetic blow out 295, 365 oscillation, see arc
material transfer 304, 334ff. over-commutation 260ff.
- - by arcing 306
- - on making contact 440 palladium 112
- - on Opening contact 308ff., 440 parabolic law of film growth 104
MATHIESSEN rule 387 partition 382
MAxWELLIAN distribution 384, 419ff. Paschen's law 276
mean free path 197, 397, 426, 430 passivating films 102, 107
mechanical wear 312 p-conduction 385
melting drop 89 periods dlning contact opening 335
- voltage 6, 89, 438 p-n-junction (p-n-Vbergang) 188
Subject index 481
peeling effect 157 respite interval 290ff., 297
Pl'.:IlriER effect 67, 76, SOff., 353 RU-characteristics of stationary
permanent contacts 363 contacts S7
g;>11-relationship 5, 60ff. resonance in microphones 1S1, 1S3
physical constants of contact materials reversible RU-characteristics 6, 50ff.,
436 90ff.
physisorbed film 102 rhodium 112
:n:-electrons in carbon 226, 404ff. RJCHARDSON-DUSHMAN equation 3SS
pip formation 354ff. rusting 115
plasma 290ff.
plastic deformation 30, 369 SAHA equation 420
plated metals 170, 356 ScHOTTKY-term 3S9
platinum 106, 112 srew 167
- bridge 340 SEEBECK effect SOff.
plug-socket contact, see contact type selenium rectifier 1S8
Poisson distribution 400 semiconducting contacts 185ff.
- ratio 368 semiconductor (Halbleiter) S1, 131, 384
polarity effects 150ff., 245ff. servomechanism 355
polymer 117, 277 SF6 294
potential barrier 3S~ ff. shape factor, f (y) 18
- - around p-n-junction 1SS shear strength 154ff., 203, 405ff.
- -, semiconductor and metal 185 shell 381
post arc current 295 short constriction 11
potentiometer 366 short arc 280 ff.
power ba]ance in arc 431 ff. short circuit currents during commuta-
preloaded contact 4 7 tion 267
pretarnish film 102 a-electrons 404
pressure 34 silicon 188
primary e]ectrons 423 - carbide 387
probability 397 - rectifier 188
- of contact failure 359 silver 106, 109
- paper 399 - cadmium-oxide 309, 364
protecting films 20, 107ff. - graphite 364
- oxide 109
quantize 3S1 - plated contacts 165, 363
quasimetallic contact 2, Sff. - sulfide HOff., 150ff.
quenching circuits 325ff., 329ff., 332ff. - wolfram 311, 365
- wolframate_ 366
radio noise 272 similitude law 275ff.
radioactive tracers application 341 skin effect 54
recovery of dielectric strength 291 skis on snow 232
rectifier 1S5 ff. softening 377
reed relay 171 - drop 8S
reignition of arc 290ff. - temperature (Erweichungsintervall)
relay 361, 364 377,436
reliability 359, 402 - voltage 89, 136, 161, 438
replica 117 solderlass wrapped connection 159
repulsion in contact 55, 161 spark 290
resistance commutation 263 specific depth 30, 373
- line 2S1 - friction force 200, 207, 223
- Ioad relation 40ff. split brush 267
- welding 164 spot welding 164
482 Subject index
spreading resistance 20 tunnel effect between semiconduoting

sputtering 310, 396 members 131
stainless steel108 - resistivity, definition 121
static electrification 189 turbin interruptur 356
- friction 164 turbo prop engines 273
stationary contacts in practice 164 turbostratic structure of carbon 407
steel106 twin oontacts 361
sticking, see adherence
stick-slip 228ff. ultra low temperature measurements
strain hardening (Kaltverfestigung) 370 124
superconductivity of oontacts .128 ff. ultrasonic cleaning 29
supertemperature 5, 62ff. under-commutation 260
surety of contact make 359
vacuum arc 286ff.
- switch 291
talc 227
valence band = filled band 384
tarnish film 102, 111ff., 132ff.
van der Waals force 102
tarnishing 104ft'., 111ft'.
varnish film 146
TEL 118
Vxcu&'s hardness number 376, 439
temperature boiling C 436 viscosity 210ff.
- development in constriction 92ff. voltage flash 253ff., 264
- distribution 78ft'. volta potential 390
- in warmest isotherm of carbon on
- regulator 332
copper 65, 259
-, melting 436
water film 114
- in sliding contact 99, 223 ff., 249 wear 232ff.
tensile strength 203 -, brush-ring contact 251ff.
terminal contact 193ff. -, mechanical 312
thermal conductivity 193ft'., 395 wedge effect 259, 412
- resistance 193ff. welding in contacts, see also adherence
thermionic emission 119, 387ff. whiskers, meta! 277, 417 ff.
thermoelectric effects SOff. WICKER's hardness number 375
thick-film lubrication, see hydrodynamic widening of a-spots 137
lubrication WIEDEMANN, FRA.Nz, LoRENz law 63,
THOMSON effect 67, 338, 344, 348 396
Tirri1l regulator 332 wolfram 106, 109, 113, 155
titanium 106 - arc 424
Transients 313 - carbide 114, 365
transistor 185 work hardening, see strain hardening
transition resistance 2, 124 - function 388
transmission coefficient 1.19
trolley car contacts 268 yield point 371
tungsten, see wolfram Yowa's modulus 368
tunnel current 118
- - around a metallic contact 132 z 233ft".
- dependence on T 130 Zener effect 134, 145
- diode 129, 134 zinc 109, 377
Errata
p. 60: Line 3, read: Eq. (13.07).
T(J
p. 62: In Eq. (13.06), read: f

'T
p. 74: Instea.d. of Note (16.10), read: Note.
p. 75: At the end in Eq. (16.17) read: arc cos ::8
p. 84: Line 5 from bottom. Instead of tempera.ture Tc read: Supertemper·
ature Tc - T 0 •
p. 91: In Ta.ble (20.04), read: Supertempera.ture e.
p. 92: In Table (20.05) instead of q, read: e.
p.126: In Section E the reference to DmTRICH is [3] not [4].
p. 129: Line 17 from bottom, read: where one metal was.
Lines 12 and 11 from bottom, read: of any superconductor. The
electrons from the normally conducting meta.l ca.n tunnel into the
superconductor only ü they are lüted above its gap.
p. 150: Line 2, read: 4.4·10-7 m.
p. 177: Line 10 from bottom, read: k in ;a.!~
p. 197:Line 6 from bottom, read: 5 ·10-9 m film or.
p. 202:Footnote 2 refers to ANTLER [2], [4], [6].
p. 219:In about the middle read: ANTLER [1], [6] instead of ANTLER [1], [2].
p. 233:Line 21 read: nominally fla.t, bolted.
6.2
p.250: Correctto: q= :n;(5 ·10-6)Z =7.9·108 wa.tt/m2
5·10-6·7.9·108
e= 380
o.75 = 78°
{} = 8·0.86 = 67°
p. 258: In Table (44.05), in column for resistivity, read: 10-1 Om.
p. 333: In Eq. (63.02) read: 2 V m = ri; I""; i +
ia
. = R-
E q. (63.04:) rea d: $a E -2V
rm-- ---x-
2V m [
1 - exp ( - LR t )]
p. 336: Eq. (64.03) shall not conta.in A but the distance 8 between the bot-
toms of the craters appearing in the electrodes after the rupture of
the liquid bridge, accompa.nied by an arc. The high pressure of
metal vapor in the arc punchces the stumps of the liquid bridge apart
generating the cra.ters. (Pietures taken by T. A. Da.vies, J. Materials
Science 3 (1968) 314, show that "er is of the order of ac.) The simul-
ta.neous movement of the electrodes is usually negligible.
Footnote 2, read: Ya instea.d. of Yp•
p. 339: Lines 7, 8, 9, 10 change into: in the sa.me simple ma.nner. Here the
energy! LI2 is not essentia.lly ~ponsible for the current through the
liquid bridge, but it sta.bilizes the bridge by opposing every current
cha.nge.
Line 15, read: fer, even on its direction. Evidently, whether a bridge
ca.n grow "long" depends on the sta.bilizing L.
p; 346: In the third line from a.bove, RIDDLESTONE (2] is missing;
p. 349: Before Eq. (66.06) read: (14.10) gives.
p. 350: Eq. (66.13) read: Ubt ""' ßeo Ib

a
OT(}~ Ib . Tflo Ib
p. 351: Eq. (66.15) read: U2 mstea.d of 0 -w;-
bl bl
Line 14 from bottom, read: 10-1s m8 fa.mp2 instea.d of 10-18 m8fa.mp2
p. 352:
e 2a
Bottom line, read: -2 I = U bl instea.d o f - I = U b.
a (}
p. 353: Read: U bl instea.d of U b·
Line 10, read: Fig. (65.04) instea.d of Fig. (65.03)
p. 358: Line 2, read: 100 A it is better to move the mercury a.wa.y from both
electrodes.
p. 381: Disrega.rd the last two sentences.
p. 384: Line 4 after Eq. (li, 11) read: 1J ~ : kT instea.d of 1J = : kT.
p. 386: In Eq. (II,13) read: ""'instea.d of =
p. 389: In Eq. (II,19) lowerterm, read: + V/2. Betterrea.tmentinR. Holm,
J. Appl. Phys. 39 (1968) 3294.
p. 394: Line 6 from bottom after capa.cties, a.dd: at room tempera.ture.
p. 396: Line 3 from bottom, read: with L forthebest conductors somewha.t•••
p. 397: Line 3 a.fter Eq. (II,16) rea.d: the fa.ctor tha.t increa.ses c of the elec-
. trons compensa.tes the fa.ctor tha.t diminishes A.
p. 400: Line 16 from bottom, read: or equa.l to i + xi.
p. 424: Line 14 from bottom, read: rp """4.5 eV.
p. 430: In Table (IX, 5) under J - read: 1 instea.d of 10.
p. 433: in Eq. (IX, 11) read: (1 -LI) instea.d of (I- LI).
p. 434: Line 14 read: §50 B instea.d of § 49 B.

(1967, R. Holm) Electric Contacts

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

(1967, R. Holm) Electric Contacts

Uploaded by

Copyright:

Available Formats

RagnarRolm

Springer-Verlag Berlin Heidelberg GmbH 1981

in collaboration with Else Holm Ph. D.

Fourth completely rewritten edition

With 218 Figures

Springer-Verlag Berlin Heidelberg GmbH 1967

Siemens-Werke in Berlin, Germany, and with the Stackpole Carbon

St. Marys (Pa.), January 1967 R.Holm

§ 10. The inductance of a current constriction. Skin effect 52

§ 23. Growth of visible oxide films on metals at moderate to high temperatw·e.

§ 24. Water film, local cells and ru~ting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1H

§ 30. About stationary contacts in practice. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164

§ 32. Contact effects in carbon granular microphones . . . . . . . . . . . . . . . . . . . . . . 180

§ 33. Contact with semiconductors. Rectification. Static electrification . . . . . . . 185

§ 36. Survey of fundamentals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199

§ 48. Introduction ..............................•..................... 274

§51. Vacuum arc; particularly its extinction ............................. 286

§ 66. Theory of thc bridge transfer . . . . . .. . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . 347

§IV. Probability. Noise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 397

F.L. research laboratory of the Siemens-Cancern, division Holm.

A supertemperature of the warmest isothermal surface in a constriction (the

§ I. lntroduction. A sim.plifled summary of the tbeory of

1 Cf. § 1F. 2 See § 26.

if e is the resistivity of the contact material. We integrate through a

The total constriction resistance R is twice as great, thus1

Eq. (1.04) ist based on the artificialmodel (1.02) and thereforeisapprox-

as will be proven in § 4. Eqs. (1.04) and (1.05) define the same R if

a relationship that plays a part in making deductions in later chapters.

in the equilibrium state. Table (1.09) is calculated for copper. How-

lJ 0.03 0.12 0.3 0.41 V

The table gives supertemperatures, 0, related to contact voltages,

§ 2. The contact surface

Concepts such as contact members, mechanical contact Ioad P, Ioad

§ 3. The contact resistance. General theory

the current flow being constricted through A c, a is the common basesec-

W e now imagine one single solid cylinder similar to the model

According to this definition, the contact resistance is not a transition

in the case of a conducting film uniformly distributed over the conduct-

~~= Jd:_c (3.05)

fact obviously contributed to the impre:"<sion that the resistance between

§ 4. Calculation of constriction resistances with <'Onstant

We shall consider long constrictions; i. e., the conducting area Ar is

and the mathematical problem is to find solutions of this equation which

Laplacian equation no Ionger be valid.

face M of the body. In the capacitance problem, we consider Ac and

and on A 1 sits an equal charge, only with opposite polarity.

From Eqs. (4.02) to (4.04) the respecti\·e resist.ance is obtained as

The application of boundary conditionl" will be elucidated by examples.

For t.he resistance problern we first imagine perfectly conducting spheres

Hence, for a long constriction, i.e., with r » b

In the problems A 'Rnd B, no area deserves the designation of"appa-

1 Cf. KoTTLER [1], Abschn. 60.

The mathematical problern is treated in many textbooks. A nry

[/vl•' + -~~ß' + •>,]_,

The current density in Ac, J = ~ ~: , is combined with the density of

Evalutions of Eq. (4.11) are discussed in problems D and E.

1 Sl\IYTHE [1] Chapter V, pp. 109-112.

Eq. (4.11) is reduced to

dp = _g_~~ _f!_arctg V/.i (4.13)

R. HoLM and STÖRMER [6] have experimentally verified this for-

two consecutive surfaces is Rf6. Their foci, c, lie on the circumference

The semi-axes of this ellipsoid are aV2, aV2, and a.

the complete integral.

The shape factor f, calculated in accordance with Eqs. (4.14) and