You are on page 1of 16

The Electric Arc


Synopsis.-DEFINITION OF ARC. An arc is a discharge of

electricity, between electrodes in a gas or vapor, which has a voltage
drop at the cathode of the order of the minimum ionizing or
minimum exciting potential of the gas or vapor.
ARC CHARACTERISTICS. The relation of arcs to glow discharges and coronas is illustrated by discussion of "generalized"
curve of the gas discharge characteristic. Empirical equations for
arc characteristics are interpreted, and a dependence on the boiling
temperature of the anode is shown. Seeliger's experiments on
the transition from glow to arc, accompanied by the development
of a cathode spot, show that the mechanism of the current at the
cathode is fundamentally. different in the two types of discharge.
CATHODE SPOT. An analysis based on heat conduction in
the cathode shows that the cathode spot has no sharp thermal definition, but does have a sharp boundary if defined by visual brightness
or by thermionic emission. The phenomenon of moving cathode
spots presents the problem of accounting for the observed

THEORIES OF CATHIODE FALL. Compton's theory is

based on space charge considerations and the assumption that
the thickness of the fall space is equal to the electronic mean free
path. Langmuir's theory differs from. Compton's in assuming
this thickness to be considerably less than a free path. Considerations of energy balance at the cathode definitely support Langmuir's
rather than Compton's theory.
ENERGY BALANCE AT CATHODE. Calorimetric measurements permit an estimate of the fraction of the current at the
cathode which is carried by electrons. Though uncertain, the
data are accurate enough definitely to support Langmuir's theory
and to indicate that, in many cases, thermionic emission of
electrons from the cathode is supplemented by a "pulling out"
of electrons by the electricfield vhich is concentrated at the cathode
Factors which determine the anode drop and the potential fall
and ionization in the negative glow and the positive column are
briefly discussed.

IN the brief space at the writer's disposal, it is impossible to treat the subject in a comprehensive
manner and there will therefore be discussed principally certain recent developments which have added
much to our understanding of the processes involved
in arc discharges.
One isstruck,inreadingtheliteratureofthissubject,
at finding no precise definition of an arce. This is due
to the fact that, although we readily distinguish common forms of arcs from sparks, glow discharges, and
coronas, yet there are gradations from one form to
another so that the distinction is sometimes diffieult to
make. Child' describes an arc as 'a continuous current
of several amperes or more, passing through a gas and
having a cathode drop which is comparatively small."
Hagenbach' says, "In order to be able to define the arc,
the cathode fall must be taken to be characteristic. As
compared with the glow discharge, it is small
The arc is characterized by a larger current and a
lower voltage than any other type of gas discharge.
It is generally obtained in gases or vapors whose
density, at the cathode, corresponds to a pressure of the
order of a millimeter of mercury, or more. Every arc
has a region of luminous gas near the cathode.
Whether or not there is another region of luminosity,
the positive column, depends on the gas pressure, the
distancefromthcathode to the anode, thcurrent, and
the shape of the containi-ng vessel if the arc is enclosed,
High pressure, large distance, and constricted container
favor the appearance of the positive column in an arc.
*Professor of Physics, Princeton University, Princetonl, N. J.
1. For all numbered references, see Bibliography,

The total voltage across the arc is the sum of (1) the
cathode drop, which has a value characteristic of the
gas; (2) the anode drop, which depends on the size and
shape of the anode as well as the nature of the gas and
its degree of ionization, and which may be positive or
negative in sign; (3) the drop along the positive column,
which is generally proportional to the length of the
positive column and depends on the current and the
nature and density of the gas; (4) a voltage drop,
generally negative but usually small, between the region
of the cathode fall and the beginning of the positive
column. Of these parts it is only the cathode drop
which appears to have a definite characteristic value;
the other three may be altered by altering the current,
the pressure, or the geometry of the arc path. Hence
the total are voltage is not particularly significant,
although it may be considered as a characteristic
parameter if the arc conditions are specified as; for
example, an arc between plane parallel electrodes of
ll arge extent, with more than a minimal separation, and
placed in a gas at a given pressure. The volt-ampere
characteristics of an arc is generally negative, i. e., the
voltage across the are falls as the current is increased.
It may, however, be zero (voltage drop independent of
current) as would be the case if the cathode drop constituted the entire voltage drop in the arc. Probably a
slight positive characteristic could be obtained in an arc
whose anode is of very small dimensions and is located
within the region of negative glow just beyond the
boundary of the cathode fall space.
In view of these considerations, which will be amplifled later, the following definition of an arc is proposed:

Detroit, Mich., June 20-24, 1927.

potential of the gas or vapor.

An arc is a discharge of electricity, between electrodes in a

gas or vapor, which has a voltage drop at the cathode of

Presented at the Sumlmer Convention of the A. I. E. E., the order of the minimum ionizing or minimum exciiing


June 1927

It must be remembered, however, that there are

transition stages between typical arcs and typical glow
discharges which cannot be defined as either. These
transition stages are generally extremely unstable, so
that they are seldom encountered in practise. Unless
an extremely high series resistance and a correspondingly high source of e. m. f. are used, the transition from
arc to glow or from arc to nothing is abrupt and these
transition stages are not obtained.3


The relation of the arc to other types of gas discharge
is well illustrated by the generalized discharge characteristic4. The simplest gas discharge circuit consists of a
source of e. m. f., E, a discharge between electrodes, D,
and a series resistance, R, together with an ammeter
and a voltmeter to measure the current i and the



\ 9 ~~G:

FIG. 1

voltage drop V across the discharge. The external

circuit characteristic is obviously
V + R t.
The nature of the discharge apparatus itself is given by
an internal characteristic
V f (i) *
In Fig. 1 the external characteristic is represented by a
straight line of slope - 1/R and voltage intercept E.
The internal characteristic is represented by the curve
O' G G' A A'. The possible values of current and
voltage are given by the intersection points of these
curves (1) and (2).
If, keeping B constant, E is gradually increased from
an initial value 0, the current is first small and the voltampere characteristic is positive. This is the region of
corona, or Townsend currents 0 0 '. Beyond 0 ' the
current rapidly increases, the volt-ampere characterE =


istic is negative and we have the region of the glow

discharge G G'. The current jumps discontinuously
from point B to point C on the curve. Obviously, the
entire glow discharge may be skipped over if the
resistance R is small, i. e., the slope of the line E B C
large. In this case the discharge passes abruptly from
the corona to the arc type. On the other hand, if the
resistance R is very large and the line E B C almost
horizontal, the entire change from corona through
glow to arc may be passed through continuously.
At G' A there is a transition from glow to arc.
Sometimes this transition is gradual and sometimes
abrupt, in which cases the curve is rounded or sharp
at the transition region G' A. If the transition is
abrupt, there is evidence that the glow and are characteristics intersect and may be prolonged as G' G" and
A' A, and it is then possible to have either an arc or a
glow discharge at the same voltage, or at the same
current, and we have, within a small range, the anomalous situation of a glow discharge carrying larger currents than the arc at the same voltage.
In general, it is possible to determine the entire discharge characteristic of any given type of discharge
apparatus by using sufficiently large ballast resistance,
R, and correspondingly large e. m. f., E. Once this
characteristic is known, the various changes in the
discharge, which will be found when any variations of
E or R are made in the circuit, may be predicted.
In<<this connection, mention only may be made of a
very complete discussion of the question of stability of
gas discharges given by Dallenbach5 and summarized
by Bar'. The fundamental condition for stability7 is
that- di < R, i. e., that the slope of the internal

characteristic curve (2) be greater than that of the

external characteristic curve (1). In addition to this,

inductance and capacity and inertia of ions must be

taken into consideration.


Several empirical equations have been proposed to
describe the current-voltage characteristics of arcs.
The best known of these are
V = a + bI
c + dl
V = a +b1 +


V = a+

c (l + d)


These may be interpreted as follows. The constant

term a is the sum of the cathode and anode drops.
The term b 1 is the voltage drop in the positive column,
whose length is taken to be equal to the total arc length
1. The terms involving current i in the denominators



Transactions A. I. E. E.

ditions in which the arc is produced. Thus it is possible

to produce arcs in which the anode drop in potential is
practically eliminated, in which the potential gradient
in the gas is nearly zero or is reversed, in which there is
no chemical action or consumption of the electrodes,
or in which the gas or anode temperature is low. The
cathode drop and its emission of electrons are indispensable, however. Theoretically any mechanism or
process for supplying electrons from a cathode in sufficient numbers to produce, at low voltages of the order
of the minimum critical potentials of the gas, enough
ionization to give a positive space charge should suffice
to maintain an arc. Actually, however, only two
emission processes seem capable of supplying electron
emission in sufficient amount: thermionic emission and
iX _AR?cA B n
the pulling of electrons from the cathode by the large
X 5 4.03Z.0 4309 065
field in the cathode fall space, or a combination of these
RN e <\U
_7 3 7.5 S2.9 6775
two. J. J. Thomson12 and Stark13 first suggested the
former theory and Langmuir4 the latter one. The
present evidence, some of which we shall now review,
to the truth of each in particular cases, and generpints
~~~~~20~ ~ ~<tt
-__ - nl.< to a combination of both. We shall proceed, therefore, to an examination of the conditions at the cathode.

of (4) and (5) take care of the negative characteristic

feature of the are. In (4) the term cli involves the
negative characteristic of the negative glow and perhaps of the anode drop, while the term d 1/i gives the
negative characteristic of the positive column. These
same features are described somewhat differently in (5).
In any case, these equations are known to be only
empirical approximations.
Recently Nottingham11 has shown that a new equation
V = A + B/li
is very accurate for all the large number of arcs tried,
_____ =__ _____ _ COP- EUCTRODE3


In all arcs, except those in which the cathode has
1 1.5 2 I A3MPERES 5 6 7 8 3'small area and cannot lose heat rapidly by metallic
FIG. 2
conduction (as in arcs with hot filament cathodes as
used in Tungar rectifiers) the current at the cathode is
provided one of the electrodes, usually the anode, concentrated in a small area which is generally called
reaches a definite temperature which is usually fixed the ";cathode spot."^ To study the physical condition
by its boiling point. In this case, n is found to depend
on the boiling temperature T of the metal through the
n = 2.6 2 (1)iT
and A and B are constants for a given metal and arc
length. The experimental determination of n is
illustrated by the case of copper, Fig. 2. Here V and i
t I
are plotted logarithmically, so that n is given by the
slope, which is seen to be the same for all are lengths. ,
The accuracy of relation (7) over the entire range of arc
temperatures is shown by Fig. 3. Recent unpublished
M0 2000 AT 3000TO K. 4000
work has extended this curve to cadmium at T = 1051
deg. K. The significance of this dependence of n on the
maximum electrode temperature is not yet understood, of the cathode, we must therefore examine this cathode
but the fact cannot be doubted.
spot. This is extremely difficult, however, owing to
its small size, its frequent rapid motion, and the diffiFUNDAMENTAL IMPORTANCE OF PHENOMENA
culty in defining it. Until very recently there were no
All lines of evidence indicate that the essential feature measurements of the area of the cathode spot except in
of an arc is the emission of electrons from the cathode the case of carbon arcs, but recently measurements also
which produces sufficient ionization of the surrounding have been made on several metallic ares. These
gas to give a positive space charge just outside the results are shown in Table I.
cathode, thus facilitating ionization and permitting a
In the case of carbon, the spot is stationary and the
large, generally saturation, electron emission at rela- chief source of errcor is pro}bably in the measurement of
tively low vroltage. All other characteristics of arcs the photographic plate, owing to photographic broadenappear to be either consequences of this emission or ing with overexposure and to failure to use reliable
prerequisites to it under the particular physical con- methods in correlating distribution of photographic



June 1927

dehsity with intensity of light. It should be remarked,

moreover, that different grades of carbon give different
results, presumably due to the effect of alkaline impurities on the amount of thermionic emission. In the
metal arcs, the spot usually wanders rapidly, so that
Gtintherschulze photographed it as a band after reflection from a revolving mirror. Seeliger" was unable to
repeat Guntherschulze's work. Nottingham has used

per cm.2


Carbon in air




Mercury vacuum. 4000
Iron in air. ....7200
Tungsten in air... 3200


Cadmium in air..., 5000





Phys. Zeits. 7, 73, 1906.

Phys. Zeits. 7, 79, 1906.

Guntherschulze Zeits. f. Phys. ll, 71,

Zeits.f. Phys. 11, 74,
Zeits. f. Phys. li, 74,
Ann. d. Phys. 60, 95,
Nottingham To be published.


To be published.


where heat is liberated at a rate Q per cm.' at a fixed

circular area A on the plane surface of a metal block of
indefinite extent. This corresponds roughly to the
heated region of a cathode surface. Because of heat
conduction in the metal, a temperature gradient is set
up radially outward on the surface of the block as well
as into its depth. The final steady surface temperature distribution is calculated by the known theory of
heat conduction and is found to be of the form shown
by the curve T, Fig. 4A. It is quite obvious that there
is nothing in the nature of a sharply defined "hot" spot.
The spot is observed, however, by means of the light
radiation from it, and it is well known that visual
brightess L increases as a high power of the temperature T, being given approximately by the relation


5.367 - log L
in the temperature range involved here. From this,

an accurate photometric method of measuring his,

photographic plates, but did not use a revolving mirror;
the internal evidence in his work, however, justifies
considerable confidence in its correctness.
Similarly, the temperature of the cathode spot is not
very accurately known. The best determination for
carbon arcs is probably that of Reich'6, who gives 3413


deg. K, although other observers give values from

2903 deg. K. to 3593 deg. K"7. Hagenbach and Langbein18 give for iron 2430 deg. K., nickel 2365 deg. K.,
tungsten 3000 deg. K., silver and copper below 1800
deg. K. Nottingham, however, has found fusion of
tungsten in the cathode spot, which would prove its
temperature to be at least 3643 deg. K. It is quite
possible that the small size of cathode spots has led to an
underestimate of their maximum temperature.
The cathode spot of a mercury arc has been estimated

between 2000 and 3000 deg. K. on account of a continuous spectrum emitted from it and ascribed by
Stark'3 to local high temperature, in spite of the much
lower boi]ing temperature of mercury, thus supporting
his theory of the thermionic origin of the electron emission from the cathode. This spectrum, however, is not
characteristic of so high a temperature and may be
otherwise accounted for, and there is no certain evidence that the temperature is so high. Seeliger15
applied Knudsen's equation to rate of evaporation as a
function of temperature, using Gtintherschulze's measurements of rate of evaporation18, and calculated a lower
limit of 673 deg. K. We shall present evidence below,
however, indicating that the mercury loss measured by
Gtuntherschulze was partly in the form of a spray
rather than true evaporation, so that Seeliger's lower
limit should be considerably less than 673 deg. K.
Thus we really know very little regarding the temperature of the cathode spot in mercury arcs.
A very illuminating study of the theory of the cathode
spot has been made by Seeliger'5. Consider first a case






FIG. 4

and assuming that the maximum temperature in curve

T, Fig. 4A, is 3300 deg. K., the visual brightness curve
L is found to be as shown. This does limit quite a
sharply defined region, which does not difer much in
area from A. Thus, as seen by the eye, the cathode
spot is sharply defined.

it thermionic
the emission
em perature
which iS
i the

I - A T"2 e-b/r
Taking b = 6 (10)4 for carbon, this leads to curve I for
the current density of thermionic emission from various
regions of the spot. Here again the spot is quite sharply
defined and has approximately the dimensions of A,
although it is somewhat smaller than the "visual" spot.




It is certain that considerations such as these are

involved in stationary cathode spots, especially if the
electron emission is primarily of thermionic origin.
The problem is further complicated, however, in cases
where evaporation or sublimation tends to cool the
cathode, and of course electron emission is itself a cooling process. Thus these cooling agencies, acting in
addition to conduction through the body of the cathode,
must tend to limit the temperature in the hottest
regions of the spot, and thus to alter the distributions
of Fig. 4A to something like the form of Fig. 4B, which
helps to explain the fact that the area of the spot is so
nearly directly proportional to the current.
In the case of metallic arcs whose spot wanders
rapidly over the cathode surface, there is a real difficulty
in explaining the high temperature of the spot, since the
. a . .
a mercury
time available for heating is very short. In~~~~~~~~~~~~b
arc, for instance, the spot frequently wanders at a rate
of at least 300 cm. per sec.9"0 and may move 30 times
this fast. Guntherschulze2l, Stolt2e, and Seeliger" have
attempted to calculate the maximum possible rise in
temperature if all the energy i V, liberated at the
cathode goes into the metal and is carried away by heat
conduction into the body of the cathode. Both calculations are rough approximations and they lead to
opposing conclusions; i. e., Gtmntherschulze concludes
that the cathode spot even in mercury arcs rises to
temperatures above 2000 deg. cent., while Seeliger and
Stolt conclude that the temperature rises to only a few
hundred degrees in mercury and copper arcs. The
evidence is that Guntherschulze's conclusion is right,
for at least in copper the metal is found to be fused
where the hot spot passes, although both Seeliger and
Stolt criticize his computations. It is difficult to
estimate the value of these computations, not only
because of uncertainty regarding the data and the
constants (such as heat conductivity at elevated temperatures) but also because the spot may not wander
continuously, but jump from point to point, remaining
at each point long enough to heat it.
It is evident from this brief survey that, in spite of
the attention which has been focussed on the cathode
spot since its crucial importance in the theory of the arc
has been realized, there is as yet no agreement as to
whether the cathode spot always reaches such temperatures as to warrant a purely thermionic explanation of
the electron emission from it.

Transactions A. I. E. E.

very considerably. An upper limit for are carbons

would be the values given for lime-impregnated carbon23.
The actual thermionic emission from an arc carbon
must lie between the values 26.7 and 4400 amperes per
cm.2 Further than this we cannot say at present.
But this makes it evident that much, and possibly
practically all, of the arc current, see Table I, is simply
thermionic emission of electrons from the cathode.
Similarly, in the case of the tungsten arc in air, the
thermionic emission at the temperature of the cathode
spot is adequate to account for the arc current, if
Nottingham's values are correct.

A = 1.49 (10)25

= 48,700







A = 3.3 (10)26
b = 42,000

T deg. K. per cm.'

Impregnated carbon


T deg K.

per cm2



A = 1.55 (10)26
b =


Td K





The most accurate identification of arc current with

thermionic emission from the cathode is obtained in arcs
from a small non-vaporizing cathode, such as in Tungar
rectifiers or Pointilite lamps in which there is no "hot"
spot, but the entire cathode is at practically unifo-:m
temperature. In these cases, the temperature may be
measured with an optical pyrometer and the thermionic
emission current rather accurately estimated. In such
cases, the are current is generally found to be accounted
for by thermionic emission24, although in very intense
arcs in gas at high pressure the arc current is somewhat
larger than the calculated thermionic emission.
In the case of arcs from more easily volatilized cathodes the data, as we saw above, are too uncertain to
support any very positive statement regarding the
adequacy or inadequacy of thermionic emission in
accounting for the arc currents. On the whole, the
writer is inclined to the opinion that in these cases, as
well as in the intense high pressure arcs above, the
ordinary thermionic emission is increased by an effect
of the intense electric field at the cathode in actually
pullinq electrons away from the cathode which would not
otherwise be emitted. This theory, due principally to
Langmuir, is discussed later. It is significant that some
agency in addition to thermionic emission appears to be
Table II gives thermionic emission values calculated needed to account for arc currents in just those cases in
from Richardson's equation (9). In comparing these which conditions for such a "pulling out" effect would
values with current densities at the cathode spot in be most anticipated.
arcs, certain facts should be kept in mind. The emission vallues for carbon were given by Langmuir"2 for
carbon as pure as could be obtained and with great care
Seeliger" has recently made an instructive exto avoid contamination. Such purity is utterly im- periment on the development of an arc from a glow
possible in arc carbons, and the impurities which are discharge, and the relation of this to the formation of
known to be present are such as to increase the emission the cathode spot. He used a very high resistance



Jun e 1927

to stabilize the discharge and very pure electrodes,

and followed through the variations in current density
as the total discharge current was increased, beginning
with a normal glow discharge covering only part of the
cathode. The results are shown diagrammatically in
Fig. 5.
In interval, a b, the current density was constant, the
cathode fall of potential was constant at about 300 volts,
and the glow did not completely cover the cathode.
At b, the glow completely covered the cathode. Further increase in current was accompanied by increase
in current density and large increase in the cathode
drop, which may rise to several thousand volts. At c
there was first observed a tendency for the cathode glow
to concentrate into a hot spot, which tendency increased
with further increase in current. Simultaneously the
current density increased at an accelerated rate, while
the cathode drop began again to diminish. At the
point d, the cathode drop had fallen to a value less than
the original normal cathode drop, and it was falling so
fast and the current density was rising so fast that the
series resistance was insufficient to stabilize the discharge and it passed abruptly to condition e, from which

to be new in the theory

recently been applied with
normal glow discharge26.
should be attempted.

of gas discharges and has

success to the theory of the
Further applications of it


The existence of the cathode fall of potential is proof

that the space charge near the cathode is positive, i. e.,
that the concentration of positive ions exceeds that of
electrons. Let i be the electron current density and
J + j be the positive ion current density. j is that part
of the positive ion current which just neutralizes the
space charge of the electrons, and J is the excess, which
accounts for the positive space charge, whose density
we shall call p. By Poisson's equation,
d2 V
d 2 =-4 p = -

where v is the average velocity of advance of the positive ions in the field - d VId x. If this field were uniform, and if the ion made numerous collisions with
atoms, and if these collisions were either "head on" or
inelastic, v would be given in terms of ionic charge e,
mass M and mean free path L by 98
f1 7r |T,e d V
M ~dx
If we consider the impacts to be elastic, not all "head
on," but rather as if made by a sphere moving under
constant force among similar spheres distributed at
changes to very nearly
random, the factor \/
.96 -\226, which is a little larger. If, however, the field
is not constant, but is increasing toward the cathode,
)wrnw2 b+<
as is almost certainly the case, this factor is less, but
cannot be less than .96 V/ 2/X 2 and is certainly much
nearer the higher value if the positive ions make at
least two or three collisions while traversing the cathode
FIG. 5
fall space. With these uncertainties in mind, we cannot
point on the discharge was a true arc, the glow was be far wrong if we take equation (11), as was done in the
replaced entirely by the hot spot, and the cathode drop original publication of the theory.
Substituting for v in equation (10) from equation
was in the neighborhood of 10 volts.
This illustrates the fact that the mechanism of current (11) and integrating, we obtain
transfer in the glow and arc discharges is quite different.
7r e
In the glow discharge the current at the cathode is ( d V)
x + Cl, where B = 4 a/
carried principally by positive ions, and the electron
2 M
emission from the cathode is "secondary" emission due
to positive ion bombardment and photoelectric action.
In the arc discharge the current at the cathode is carried The integration constant C, is determined by the conprincipally by electrons, which are probably liberated dition that d Vld x = 0 at the outer boundary of the
thermionically, assisted by the "pulling out" action of cathode fall space. Taking x = 0 at the cathode
the field. The thickness of the cathode fall space in the surface and x = c at the boundary of the fall space, we
arc is certainly thousands of times smaller than that in have
the normal glow discharge.
Qualitatively, the progressive stages in the development of an arc may be "explained" by the principle that
the potential distribution in a gas discharge adjusts itself Integrating again, and putting V = 0 when x = 0,
so as to give maximum current, subject to the limitations V = Vc when x = c, and solving for the cathode fall of
imposed by Poisson's equation. This principle appears potential, we have




V. =

Since j = n+ e v+ and i = n- e v-, and since n+ e = n- e

for exact compensation of space charge, we have

3 13 \2/3
i( 2 BJ)J c513.

Resubstituting the value of B from equation (12) gives


3 (6

7r))213 J2/3 C513 13

2 M

ing We
of the cathode
fall space in arcs, except
for the knowledge that it is extremely small. It seems
certain that it does not exceed the electronic mean free
path 1, since the electrons have their best chance to
ionize at their first impact owing to the fact that electric
intensity diminishes with distance from the cathode.
In the present theory it is assumed that c = 1, though
it may be that this is an upper limit. Now the ionic
free path L is V 2 times the molecular free path X,
since the ions have a higher order of speed than do the
molecules. Also, the electron free path 1 is 4 V/ 2
times X, and hence 4 times L, owing to the negligibly
small dimensions of an electron. Thus, writing c = 1
and L = 1/4, and solving equation (13) for J, we find
e 1/2 /5 \3/2 1
J =T12 4 2 M i t 3 V J) 12 in c. g. s. units.
= 0.76 (10)-'

Ml12 12

in amperes per cm.2, with

M in ordinary molecular
weight units.
If c in equation (13) should have been taken to be less
than 1, then J would be larger than calculated by
equation (14).
Consider now that part of the positive ion current
density j required to neutralize the electron space
charge. An exact calculation of the relative times
required for an electron and a positive ion to pass
through the fall space appears impossible, but the following approximation is probably accurate enough for
the present purpose. Assume first that the field in the
fall space is uniform and hence equal to V,/l. The time
t required for an electron, starting from rest at the
cathode, to traverse the fall space iS given by
e V,
= at2, where a = m = m
t-= 1 +

ev .

Transactions A. I. E. E.


The positive ions, on the other hand, move forward

with average velocity given approximately by equation
(11). Putting d V/d x = V. 1, and L = 1/4, and taking the time t+ as distance 1 divided by mean velocity
u, we have
t+- \'

= V



In the actual case, however, the field is not constant

but varies from a maximum value at x = 0 to zero at
x = 1. Thus the electrons always move faster and the
positive ions slower than we have assumed, and the
i.] is larger than the value given by equation (17).
A graphical integration, using the actual field distributtion aas given by equation (12) to find t- and t+, led to a
value of i/ not much different from


= 4 / 2 -m
which was the relation taken in the original statement of
the theory27, but derived there in a manner quite inconsistent with the actual physical conditions in the fall
space. We shall use equation (18), therefore, in the
belief that it is at least a fair approximation to the
requirements of the theory.
Expressing currents in amperes, potential drop in
volts and ionic mass M in ordinary atomic units, we
have the results of this theory expressed by the


Total current density

+j + J
Total currentdest
I = 2+ +
Neutralizing current density i = 242 v/ M j
Space charge current density J = 0.76 (1)-7 -lt


Applications: Carbon Arc. At atmospheric pressure
and 3300 deg. K., which is close to the cathode temperature 1 = 0.66 (10)-3 Cm., Vc is given as about 8.6
volts29 although no determinationby areliable method
has ever been made, and the true value is probably
several volts higher. Substitution in equation (19)
gives J = 1.6 amperes per cm.2 Since the total current density I is of the order of 320 amperes per cm.2,30
J/I = 0.005. Similarlyj/I = 0.001. Thus altogether
about 0.006 of the total current is carried by positive
Mercury Arc. The vapor density at the cathode is of
the order of an atmosphere3" and its temperature is at
least 400 deg. K., and may reach 2000 deg. K., although
reasons are given later which weigh against this high
value. We shall not be far wrong as to order of magnitude if we take 600 deg. K., which gives 1 = 0.000040
cm. Vc lies between 5.5 and 10.3, and is probably
about 8.6. This leads to J = 162 amperes per cm.2
Guintherschulze finds the current density I at the
cathode to be 4000 amperes per cm.2, whence J 'I =
0.040. Similarly
= 0.0003. Thus about 0.04 of
the total current isj/I
carried by positive ions.
~~~~Other cases agree in indicating that only a small

June 1927



fraction of the total current,at the cathode is carried

by positive ions.
A test of this theory is afforded by comparing these
calculated values for the fraction of current carried by
positive ions with the values calculated from considerations of thermal equilibrium at the cathode. Before
doing this, however, we shall consider an alternative
theory of the cathode fall space which has been proposed by Langmuir.

B. Cooling of Cathode. (1) By escape of electrons,

f s _, where sp - is the electron "work function," or heat
of evaporation; (2) by conduction through the body of
the cathode, C; (3) by gaseous conduction and convection, C'; (4) by radiation, R; (5) by evaporation of
cathode material, E.
Grouping all these items, we find the equilibrium
condition to be given by


On this theory, the cathode fall space is simply the
positive ion sheath produced around the cathode by the
incoming positive ions. If it is assumed that the positive ions traverse this fall space without colliding with
gas molecules, i. e., if d < 1, the space charge equation
of Child32 and Langmuir33 may be applied in the form,
VX/2 4 e V!3/2
d2 in c.9w\g. s.M - d2
9 1r@ M
= 0.543 (1
in ordinary electrical and



V, + ss + s -- F (VI + Vi)
An experimental determination of the factors in this
equation should therefore permit a calculation of the
fraction f of the current at the cathode which is carried
by electrons. Let us consider the various terms in
this equation:
The cathode fall of potential V, in various arcs has
been measured with the following typical results:



Carbon (impregnated) in air ........

Magnetite in air ..............
Copper in air at reduced pressure .............

molecular units. (20)

Thisdoesnotappeartodiffermuchfrom equation (19) of
Compton's theory,
thzeory, but in realtyit may be quite different
since it does not assume d to be equal to the electron
free l pathT 1, but1 leaves dz undetermined.
use this equation, information from some other source
must be obtained regarding either J or d.
Two courses are open for finding independently the
positive ion current density, J + j, in order to test
Compton's theory or to complete the information
necessary for Langmuir's theory. J + j may perhaps
be measured directly by Langmuir's exploring electrode
method34, although this has never been done near the
cathode and presents experimental difficulties, or
J + jlI may be calculated from considerations of
thermal equilibrium at the cathode, as follows:

but in reality it maybe quite

Vc (volts)

Carbon in air (current I) .................... 7.6 +

Mercury in vacuum..........................

*Argon gas and heated non-vaporizing cathode.

*Helium gas and heated non-vaporizing cathode

*Mercury vapor and heated non-vaporizing


~ tde...............
~ h



13 .7




5. 5



As all values except those marked * were obtained by

the old probe method which is known to give incorrect
results34, they are only approximate and are probably
several volts too low. More accurate values are
greatly needed.
The heat of neutralization, or condensation, of positive ions o + was formerly calculated from a theoretical
relation so + = Vi L- - derived by Schottky and
von IssendorffS4 and by Compton3". Recent experimental measurements42 have shown that the true value
is much less than this, and nearly zero. The discrepENERGY BALANCE AT CATHODE35
ancy must be due to the fact that some of the energy
Let f be the fraction of the current at the cathode liberated at the electrode surface during neutralization
which is carried by electrons and 1 -e that carried by of an ion is radiated away and hence does not contribute
positive ions. Then, per ampere of current, we have to the heating of the electrode. Compton and Van
the following rates of heat development, in watts:
Voorhis43 give reasons for modifying the above equaA. Heating of Cathode. (1) By incoming positive tiontotheform
ions, which fall through the cathode drop V,, (I - f)
s+ = r VT (L)-so (22)
(V, + +)4 where p + is the heat of neutralization of where r is a "radiation factor" a little less than 0.5,
positive ions at the surface of the cathode; (2) by out- and L is the latent heat of condensation of the neugoing electrons, some of whose energy may be returned tralized ion on the electrode, in case the ion remains
to the cathode, [f(V - (1 -f) Vi] F; here f V. is the there deposited. If the material of the ion does not
energy gained by the electrons in the cathode fall space, remain on the electrode after neutralization, L is to be
(1 -f) Vz is the energy of these electrons which is used omitted from equation (22).
in ionizing the gas whose ionizing potential is Vt, and F
In this connection, the writer would suggest that the
is the fraction of the remaining energy which returns luminosity of the cathode in mercury arcs, which has
to the cathode in the form of radiation, etc.; (3) by heat been taken to indicate high local temperatures exceedsupplied-by an external heating source, if there be ing 2000 deg. K., may be simply this radiation acone, H.
companying ion neutralization at the cathode surface,



and showing as a continuous spectrum because -of the

intense field at the surface. Evidence that electron
emission in this case is not of ordinary thermionic
origin will be presented below.
The ionizing potential Vi is accurately known for
most gases and vapors. In the case of arcs in air between vaporizing electrodes, there is uncertainty regarding the type of gas which is being ionized.
The fraction F of the excess energy of the electrons
which is returned to the cathode is unknown. It
cannot exceed 0.5. It is probably much nearer 0.0,
expecially in the case of a rapidly vaporizing cathode,
where a blast of atoms would tend to carry away any
high speed atoms which might have been indirectly
accelerated by the electrons. Radiation of energy
back to the electrode from the gas would be a small
positive factor.
H, C, C', R, and E may all be measured or computed.
In considering the cooling E by evaporation of cathode
material one must be cautious, however, since there
is evidence in cases like the mercury arc that not all
material lost is by true evaporation, but part of it is by
mechanical loss as a "spray" which does not contribute
to the cooling. Also, the radiation loss R may be
partially compensated by radiation gain from the
anode, which must be taken into account.
Evidently our present knowledge and our experimental technique are too limited to permit us to use
equation (21) for accurate results. It may be used,
however, to show orders of magnitude and to set certain upper and lower limits which permit us to draw
some important conclusions.
Applications: Carbon Arc. Take, for a 10-ampere
are, V, = 9.0 volts, o - = 3.9 volts, sp + = 0, Vi = 16
volts. A rough estimate of conductivity loss based on
a hot spot area of 0.04 cm.2, temperature gradient 2500
deg. per cm. and conductivity 0.01, gave C = 0.04 volt,
though the data are uncertain and the writer is inclined
to believe that the result is too low. Net loss by
radiation, calculated as if cathode and anode hot spots
were black body radiators at 3140 deg. K., and 3700
deg. K., respectively, gave R = 0.75 volt. E is relatively small, and so is C', provided the are is not cooled
by an air blast. With these values we find
f = 0.64, assuming F = 0;
f = 0.63, assuming F = 0.25.
f could be raised as high as 0.70 by neglecting all heat
losses, C + C' + R + E, which is clearly an upper
limit. No reasonable value of so+ differing from 0
would produce much change in f. The assumed value
of V. is probably several volts too small, but no reasonable increase would increase f greatly. s - could only
be given a smaller value if the electrons were pulled out
of the cathode by the field, rather than spontaneously
emitted thermionically, and we have previously seen
that the evidence proves certainly that no large effect of
this kind can be important in the carbon arc. We thus
seem forced from energy considerations to conclude that

Transactions A. I. E. E.

the fraction of current carried by electrons at the

cathode is of the order of 60 to 70 per cent, rather than
99.4 per cent as predicted by Compton's theory. The
fact that an earlier calculation35 appeared to support
Compton's theory was due, first, to the use of a value of
<+ now known to be inadmissible42 and second, to the
use of an impossibly high value for F.
Mercury Arc. In this case, recent experiments by
Giintherschulze44 give apparently accurate data for
most of the quantities involved, except for minor corrections pointed out by Seeliger"5 and included here.
The data are, in watts (volts) per ampere of are current,
C = 2.68; E = 2.8 to 3.9, depending on the assumed
temperature of the cathode spot; R = 0.04. Taking
S- = 3.9, Vi = 10.4, + = 0, Vc = 8.6 volts (as a
reasonable value owing to the fact that it must lie
between lower and upper limits of 5.3 and 10.3 and
probably nearer the upper value 4) knowing C' to be
negligible and H zero, and assuming F = 0, we find
f = 0.25 to 0.16. If F is taken to be greater than zero,
f becomes still smaller.
Even if cooling by radiation R and evaporation E is
entirely neglected, which could only be justified if all
mercury were lost from the cathode mechanically rather
than by evaporation, and even if the cooling o_ by
electron emission were neglected, which would be
justified if the emission were due entirely to the "pulling out" effect of the field, still equation (21) gives only
f = 0.70. In any case, therefore, the fraction of current carried by electrons must be less than 70 per cent,'
whereas Compton's theory predicted 96 per cent.
From this consideration of energy balance at the
cathode, therefore, it would appear that Compton's
assumption that the thickness of the cathode fall space
is equal to the electron mean free path is incorrect, and
that this thickness is much smaller. If it is much smaller, the positive ions must move through it generally
without colliding, and we have exactly the space charge
condition leading to equation (20) of Langmuir's theory.
We must therefore consider the evidence as strongly
supporting Langmuir's theory.
Further than this, these energy considerations lead
us to some conclusions regarding the mechanism of
electron emission from the cathode of a mercury arc.
Since almost certainly the cathode drop does not exceed
the ionizing potential Vi = 10.4 volts, it is obvious that
no electron can ionize more than once near the cathode.
The fraction f cannot, therefore, be less than 0.5 and
could only be that small in case the probability of
ionization were unity, which cannot be so. From this
consideration, f must exceed 0.5. An examination of
equation (21) in connection with Guintherschulze's
data shows that a value off> 0.5 can only be obtained
if * < 3.9 and E K 2.8 by large margins. In other
words, the field at the cathode surface acts to pull
out electrons which would not otherwise be liberated,
and some of the mercury is lost from the cathode

June 1927



mec_hanically, rather than by evaporation. The former ments. A. W. Hu1149 reports that calorimetric work
of these possibilities was suggested by Langmuir, whose on high-power mercury arcs at the General Electric
measurements of positive ion current densities led him laboratory is also in quantitative agreement with these
to estimate the field at the cathode of a mercury arc ideas.
From the preceding discussion it will be seen that
to be of the order of 106 volts per cm. Such fields are
known to pull electrons from metal surfaces in the much progress in the understanding of are phenomena
presence of gases or vapors, and would probably be has been made during the past few years, and that
especially effective if the metal surface is heated, as there are at present numerous possibilities for
in an are, so that many electrons need only the addi- further experimental research, guided by theoretical
tional assistance of the field to permit their escape45.
(1.) Complete discussions of earlier work on arcs,
This region, generally called the negative glow, is a
region in which the concentration of ions is maximum.
be found in
may Mrs.
nd iss often
trenth and
inium strength
of minimum
field isisof
The The
Ayrton (The Electrician
Electric Arc,"
reversed in direction, the current being by diffusion of 1902)
"The Electric Arc," Child (Van Nostrand 1913)
electrons in the direction of decreasing concentration39 .
Lichtbogen, "Handbuch der Radiologie," Vol. IV,
Probably much of the radiation from this part of the arc
is the result of recombination of ions and electrons47. pp. 211-444, Hagenbach (Akademische VerlagsgesellCONDITIONS IN THE POSITIVE COLUMN
schaft, Leipzig, 1917)
L'Are Electrique, Leblanc Fils (Journal de Physique
Here ionization occurs to just a sufficient extent to
balance the loss of ions by recombination or diffusion 1922)
Lichtbogen, "Handbuch der Physik," Vol. XIV,
to the walls, if the arc be enclosed. This ionization may
be produced thermally, by electron impact, photo- Hagenbach (Springer 1927)
(2) In terms of mechanism, the arc may be defined
electrically, or by a combination of these. There are
reasons for ascribing much of it to high temperature in as a gas discharge in which the ionization near the
the carbon are35, while this certainly plays no role in the cathode is produced by electrons which have fallen
mercury arc, where the ionization is due to electron through the cathode fall of potential and thereby gained
impacts, probably of a cumulative nature. The light the energy necessary for ionization, whereas in the glow
from the positive column is almost certainly due to discharge the ionization is produced while the electrons
are falling through the cathode fall space. In the glow
excitation rather than to recombination47.
discharge the ionization increases exponentially with
The anode drop in potential may be positive or distance from the cathode: in the arc there is no exnegative according to conditions first explained by ponential building up of ionization. This definition is
Langmuir and Mott-Smith34 as follows: Surrounding equivalentrtothe onesalready given.
(3) Different arc types are sometimes found under
the anode is an atmosphere of ions and electrons moving
with more or less random motion. If, in this random conditions in which transitionsfrom one form to another
motion, the excess of electrons over positive ions striking may occur. It is suggested that the primarily "therthe anode would be greater than the total current in the mionic" arc and the primarily "pulling out of electrons"
circuit, then a negative, or reverse, anode drop is set up arc may be two such types. In Table I those arcs
so as to hold back enough electrons to keep the current whose current densities are thousands of amperes are
to the value demanded by the constants of the circuit. probably of the latter type and those with smaller
On the other hand, if the number naturally striking the current densities of the former type. Both types are
anode is insufficient to carry the current, then a positive shown for tungsten in Table I. There is some evidence
anode drop is set up so as to draw in more electrons. of still another arc mechanism (Dr. Slepian,
From these considerations, it is evident that anode drop unpublished)
decreases with increasing anode area and with increased
Are," p. 2 et seq.
ion concentration,
der Physik," XIV, p. 324,
anode by promoting
The heating of the anode depends on three factors: Springer.
3 Seeliger, "Handbuch der Radiologie," IV, pp. 257-262.
(1) the heat of condensation of electrons s _; (2) the
average energy V _ of the electrons in their initial ran5. Dtillenbach, Phys. Zeits, 27, 101, 448, 1926.
6. Bar,"Handbuch derbPhysikd"XiV,pp.l75-182,2Springer.
dom motion; (3) the anode drop Va, if this be4positive.
7. Kaufmann, Ann. d. Phys., 2, 158, 1900.
Although this subject has been studied calorimetrically45
S. Artonic h Elertchn,l Zei ts., 4,190, 1583.
and the order of magnitude of these predictions always
verified, thus far only Van Voorhis42 has measured
101oSenet,TA-.A.I .E,196 .82
11. Nottingham, JoITR. A. I. E. E., 42, 12, 1923; Phys. Rev.,
the quantities necessary to make an accurate quantitative test, which has exactly verified the above state- 28, 764, 1926.



Transactions A. I. E. E.


12. J. J. Thomson, "Conduction of Electricity through

Gases," 2 ed., p. 604.

13. Stark, Phys. Zeits., 5, 51, 750, 1904.
14. Langmuir, G. E. Rev., 26, 735, 1923; Science, 58, 290, 1923.
15. Seeliger, Phys. Zeits., 27, 22, 1926.
73, 1906.**
16. Reich, Phys. Zeits., 7, 73,

Joseph Slepian: There is hardly

a scientist on whose work

I lean more than Professor Compton's.
1 remember as one of the best of Professor Compton's papers
the one on the theory of the are which he gave in the Physical
Review in 1923, in which he so ably defended the thermionic
de Radl,"
theory which he has discussed tonight. Today, however, Pro17.
192fessor Compton apparently feels that perhaps this theory may
19. Dufoulr, Jour. de Phys., 1, 109,
la hdafeal
20. Stolt, "Inaugural-Dissert.," Upsala, 1925; Zeits. f. Phys., not always hold after all.
Remarkable also have been his papers on abnormally low26, 95, 1924; ibid. 31, 240, 1925.
voltage arcs. The first deals with the theoretical difficulties,
21. Ganthersehulze, Zeits. f. Phys., 28, 325, 1924.
22. Langmuir, Trans. Am. Electrochem., Soc. 29, 125, 1916. almost proving that an are with a voltage less than the ionizing
23. Richardson, "Emission of Electricity from Hot Bodies," voltage of the gas is impossible. A later paper, however, demonstrates the existence of arcs with voltage less than the ionizing
p. 69.
24. Compton, Phys. Rev., 21, 270-273, 1923; W de Groot, potential but greater than the resonance potential., Then
Physica, 5, 121, 234, 1925; also unpublished work at the Researoh followed a paper showing that where arcs with voltages lower
than the resonance potential had apparently been obtained,
Laboratory of General Electric Co., Schenectady, N. Y.
oscillations had been present which momentarily would raise
25. Seeliger, Phys. Zeits, 27, 730, 1926.
the voltage above the resonance potential. Then came still
and Morse (to a
n PAys. Rev.)
27. This discussion is based on a paper by Compton, Phys. another paper showing that arcs with total voltage less than the
Rev., 21, 266, 1923, but includes some corrections and extensions. resonance potential were possible under suitable conditions
Another paper which I found exceedingly valuable, stimulating,
28. Compton, Phys. Rev., 7, 489, 1916.
and instructive, is that one on the Mobility of Ions in Discharges
29. Ayrton, "The Electric Arc," p. 222.
30. Hagenbach, "Handbuch der Physik," XIV, p. 349, in which he very boldly sets out to calculate the ways in which
ions will move in gases of considerable density under the action
31. Gutntherschulze, Zeits. f. Phys., 11, 74, 1922; Langmuir, of the electrical fields, taking account of various kinds of collisions which electrons can have with molecules. This was an
Science, 58, 290, 1923.
exceedingly difficult problem, and I marveled that anyone had
32. Child, Phys. Rev., 32, 492, 1911.
the temerity to tackle it. Yet with a few skillful manipulations
33. Langmuir, Phys. Rev., 2, 457, 1913.
34. Langmuir and Mott-Smith, G. E. Rev., 27, 449, 538, 616, and ideas, Professor Compton derived equations which are
quite easy to understand and exceedingly valuable.
762, 810, 1924.
There are some experiences which I have had in connection with
35. The use of this method was first pointed out by J. J.
Thomson, "Cond. Elec. through Gases," 2 ed., p. 614. The my engineering work which I think will be interesting in relation
details were first worked out and applied by Compton, Phys. Rev., to the theory of the are. The various theories of the cathode of
21, 281, 1923. Gfintherschulze, Zeits. f. Phys., 11, 74, 1922; 31, the are mentioned by Professor Compton require that a con509, 1925, used first an incomplete analysis of the problem and siderable portion of the current be carried by electrons leaving
later partially corrected it. W. de Groot, Physica, 5, 121, 234, the cathode. The question then arises as to how these electrons
1925, has made a partial application to arcs of the Pointilite type get out of the cathode, as ordinarily electrons will not pass from
a metal into an adjoining gas. One agency which will assist
in neon, argon, mercury.
electrons in escaping from a metal is heat. When its temperature
36. Child, "The Electric Are," p. 61.
37. Hagenbach, "Handbuch der Physik," XIV. p. 335. is sufficiently high electrons can pass freely out of a cathode.
This is essentially the thermionic theory of the cathode of the
38. Stark, Retschinsky, and Schaposchnikoff, Ann. d. Phys., arc, which Professor Compton advocated a number of years ago.
Another possibility which Professor Compton has mentioned
18, 243, 1905.
is that a very high electric gradient may develop at the cathode
39. Compton and Eckart, Phys. Rev., 25, 139, 1925.
surface in the are a gradient so high that the electrons are pulled
40. Compton, (unpublished), by the method of (39).
out of the cathode even though it is not hot enough for thermionic
41. Schottky and von Issendoriff, Zeits. f. Phys., 26, 85, 1924. emission.
At the time of the experiments which I am going to describe,
42. Van Voorhis (to be published in Phys. Rev.)
43. Compton and Van Voorhis, Proc. Nat. Acad. Sc., 13, I, along with almost everybody else, believed in the thermionic
theory of the cathode; that is, that in an are it was necessary to
XXX, 1927.
44. Gfintherschulze, Zeits. f. Phys., 11, 74, 1922; 31,509, 1925. have a cathode hot enough for thermionic emission. If the
45. Schottky, Phys. Zeits., 15, 827, 1914; 20, 220, 1919; cathode was not hot enough for this, an are discharge would be
impossible, and if anv discharge was obtained it would have to be
Zeits. f. Phys.,14, 63, 1923.
or other high-voltage form. I tried to apply these ideas
Emeleus'Proc Camb.Phil. Soc.23,531,1927;McCurdv atoglow
46. Emeleus,
the developm ent of the are which follows the breakdown of a
Phil. Mag., 48, 898, 1924; MeCurdy and Dalton, Phys. Rev., 27, spark-gaP by application of high voltage.
163, 1926. This work dealt with glow discharges, but there is Spr-a by aliction of hi vlage.
even more reason for expecting the same phenomenon in arcs.
Sicah lcrdso h pakgpaeiiilycl,i
seemed necessary that the discharge should start as a glow and
47. Compton, Turner, and McCurdy, Phys. Rev., 24, 614,.only after
some point of the cathode reached a sufficiently high
1924. The discussion given here of the relation of glow to arc emeauesolthdicrgcanenoanr.Itid
discarg isnotineneal orrct.
he uthrs t tat imehad to calculate the time for the heating up of the cathode spot, and
in mnd oly te
"lw votage arc wit hotcathdes ith therefore the time for the flow to change into an
aUrc, using data
which theyhad
been working.
the watts input at the cathode of a glow on copper obtained
48. Schottky and von Issendorif, Zeits. f. Phys., 26, 85, 1924; from other experiments. I found it would take seconds before
Penning, Physica, 5, 217, 1925.
the copper would get to the melting point, let alone a temperature
49. IIull, in discussion of paper by Van Voorhis a.nd Compton, sufficient for thermionic emission. But the experiment showed
at Washington meeting of the Am. Phys. Soc. in April, 1927.
that the arc struck almost at once. Immediately after the gap

119, 74,1

26. Cmiptrn




June 1927

broke down, the voltage dropped to 20 volts, which is too low

for a glow.
I had been of the opinion that the cathode had to be hot in
order to maintain an are; yet here, where the electrodes did not
have time to get hot, I was getting a discharge with only 20 volts.
More recently, this experiment has been repeated, using the
DuFour Oscillograph, and it has been found that the time for the
discharge to change from glow to are is of the order of a microsecond.
Since the cathode couldn't have become hot in so short a time,
this experiment made me feel that the thermionic theory of the
cathode couldn't be correct; at least not all of the time.
Another experience in connection with my engineering work
which made me believe that probably the cathode didn't have to
be hot was in studying the operation of switches. I have seen
switches in which the arc was blown rapidly along the arcing
horns operate, and examined the horns afterward, finding
stretches on the arcing horn absolutely free of burning. There
might be some oxidation but no evidence of a very high
This seemed strange and would be hard to explain if the cathode
had to be hot enough for thermionic emission. I looked into this
a little more closely, and considered the hypothesis that perhaps
the arc hopped from point to point, without passing over the
intervening stretch, so that this stretch might not appear burned
because the arc had not actually played on it.

20,000 amperes; that is, I have moved a 20,000-ampere arc, over

a cathode surface so rapidly that there was no melting of the
copper but merely a trace of oxidation. Incidentally, the current
density in these experiments was of the order of 30,000 amperes
per Cm.2
The application of the method of energy balance at the
cathode which Professor Compton has used in his paper for estimating the fraction of the current carried by electrons is certainly very interesting, and the values f = 0.25 to f = 0.16
obtained for the mercury arc seem significant. If the ionization
of the gas next to the cathode is primarily due to collisions from
electrons coming from the cathode, f could not be less than 0.50.
Some time ago, I suggested in the Physical Review that perhaps no part of the current at the cathode was carried by electrons, but that all of the current was carried by positive ions
coming from the highly ionized gas next to the cathode. The
cause of the high state of ionization in the gas was to be sought
in the very intense energy concentration there. The values of f
which Compton finds indicate that this suggestion may be near
the truth. Indeed, if a necessary correction to the energy balance
equation is applied, the value of f comes even closer to zero in
accordance with my suggestion. The correction is as follows.
As item (1) under A, "Heating of the Cathode," Professor
Compton has "By incoming positive ions, which fall through the
cathode drop B0, (1 - f) (B, + 4+)." But all the positive
ions do not fall through the cathode drop unimpeded. Some

Cooli~~~~~~~~~~~~~~~~~~oln Watate

/1^\\ in6in


\1~~jl// /Distance



FIG. 1

I eliminated this possibility by bringing the electrodes very

close together, and also took photographs with a high-speed
camera. I found that even at those spots where the >arc had
played, as indicated by the photographs, there was no burning,

will collide with molecules and lose energy to the gas. Also,
many positive ions will be reflected from the cathode thus increasing the chances of collision with molecules. Let a be the
fraction of the energy acquired by falling through the cathode
A German, H. Stolt, has also c>arried out similar experiments, drop, which a positive ion, on the average gives up to the cathode,
published in the Annalen der Phytsik. Stolt caused an arc to Then item (1), under A, becomes (1 -f) (at B0 + k).
Equation (21) then becomes
move over a cathode so rapidly that apparently there was no
as BC + - F Bt + H - C-C' -R-E
heating of the cathode. The claims of Stolt were criticized by
a B-+~
B ~
f =
Gttntherschulze, who is mentioned frequently in this paper of
+_-F( B
Professor Compton's; but Stolt replied quite well to the criticism
of Guntherschulze, and I believe that Stolt's conclusions are If we substitute the nlumerica.l values used by Compton we get
8.6 a - 6.6
8.6 a - 5.5
fairly well established. Stolt did get a low--voltage discharge
f = 8 t+39or f =
from copper and other metals, which moved so rapidly over the
copper surface that no spot of the copper surfa,ce became hot8.
enough to melt, let alone ha'xe thermionic emission. This If we take at I wre get, of course, the vahles of Compton,
seemed to me to disprove definitely the theory of the necessity for f = 0.25 and f = 0.16. If, however, a is as low as 0.61 b}y the
thermionic emission. I have carried experiments similar to first formnula, orO0.77 by the second formula, we get f-=0.
Now what is a reasonable estimate of the value of at? We
Stolt's somewhat further, and have used currents as high as



may obtain some idea by determining the number of positive ions

which cross. the cathode space without colliding with a molecule.
The electronic mean free path as given by Compton in this paper
for the mercury-are cathode is 4 X 10 -5 cm. and therefore the
ionic mean free path is 1.0 X 10-5 cm. On the other hand from
equation (20) taking - J = 4000 as given by G-intherschuilze,
we find, for d, the cathode-fall space d = 4.95 X 10-6. The
fraction of the positive ions which will have free paths greater
than the cathode-fall space will therefore be
4.95 X10-6

e 1.0X10-5


495 =


That is only 61 per cent of the positive ions crossing the cathodefall space fail to collide with a gas molecule. Hence a value of
a equal to 0.77 does not look altogether unreasonable.
J. C. Lincoln: At the plant of The Lincoln Electric Company, we have run across a new phenomenon which has to do
with the nature of the arc and which has changed my notion of
what happens in the are. This phenomenon occurs in a device
which we call an "electric torch."
The illustration herewith, shows the arrangement of the parts
making up the torch. The copper electrode holderfor the negative
terminal. is water-cooled as well as the copper positive terminal
where the path of the cooling water is indicated. The copper
anode has a tapered hole of the dimensions indicated on the


in it




n cth in

dimensions are those of a section through the center of the

opening in the anode.
After the parts are set up an are is started between the earbon
cathode and copper anode by short circuiting them with a
carbon pencil. While the are is maintained, a flame projects
from the anode as shown in the illust,ration.
Observation of the are shows that the size of the flame projecting from the anode is roughly proportional to the amperes
across the are. as might be expected.
Furthermore, with a given current, the flame is larger when the
distance D is smaller, the size of the flame decreasing as the
distance D is increased.
A change of current causes a much greater difference in the
size of the flame than is caused by a proportionate change in
the distance D.
The flame is apparently due to very hot carbon particles.
If the flame is cooled, carbon is deposited on the cooling surface just as it would be if the flame from a wick vere cooled.
A series of tests was mhade varying the current and the spacing
D. The current was furnished by one of the company's 200ampere, d-c. welders and was adjusted to 50, 75, 100, and 150
amperes across the arc. The distance D was adjusted to 3/16,
3/8, and 9/16 in. For the results shown in the accompanying
table, the flame was above the torch, as shown in the illustration.
To determine the amount of heat in the flame, the amount of
cooling water passing through the torch and its rise in temperature were measured. In the tests the initial temperature of the
cooling water was 19.1 deg. cent. and 5.375 lb. per minute was
used. From these measurements, the heat absorbed by the
water was calculated. The rate at which this heat was absorbed
was then expressed in watts. It was assumed that the watts
input, minus the watts carried off by the water, equals the heat
energy in the flame, and the table shows the percentages of heat
in the water and the flame respectively.
The results indicate in general that the greater the current,
the greater is the energy in the flame; also, that the smaller D is,
the greater is the energy in the flame.
The table shows that from 1/2 to 2/3 of the heat appears in the
flame, and I believe that if none of the heat developed in the flame
was radiated and absorbed by the copper and waater, an even
larger proportion of the total heat in the arc would appear in
the flame.
The direction of the flame can be affected by a magnet. By
presenting the south-seeking pole of a bar magnet to the arc, the

D in
of an
inch Amperes

Transactions A. I. E. E.
rise of




coolinig Watts

dater, absorbed




Percee tage
of heat
water flame

42.5 57.5






5 9






















46. 1


*5.375 lb. of water passed per




333.6 66. 4

55.5 44.5




flame is pushed to one side. bo far as I could judge the flame

itself is not affected by the magnet. The direction of the flame
is a function of the current from the carbon cathode to the anode.
iTo put it another way, the magnet had no effect on the direction
of the flame except wzhen close to the a,re between the carbon and
ofte amee
copper anode.
The current between the carbon and copper anode is effected
in just the way one would expect from the laws governing
electromagnetic action.
When there is no external magnetic field at the are, the current
flows radially between the carbon and the copper anode. When
the arc is subjected to an external magnetic field, the current is
forced to only a part of the radial path between carbon and copper
and at the same time the flame is deflected so that it is more at
right angles to the current.
What bearing do these results have on our conception of
what takes place in the arc?
The present view is that the voltage across the are is made up
of three portions: (1) the drop at the negative terminal, which
must be great enough to heat the terminal to the point where it
will throw off ions readily, (2) the I- R drop due to the resistance
of the gas stream between anode and cathode, and (3) the drop
at the positive terminal which is fixed by the nature of the
material and in the carbon are is much greater than the drop at
the negative terminal. The results of the measurements would
indicate that in the carbon arc there is a drop at the positive
terminal that may be fixed by the nature of the material, but
that this drop is not nearly so great as has been supposed. The
heat at the positive terminal in the ordinary carbon are is the
sum of the heat due to the inherent drop and the heat of the flame
or blast from the negative terminal. The heat due to the flame
or blast has been separated largely from the inherent anode drop
in the electric torch and measured. The measurements indicate
that the heat in the flame or blast is greater on the average than
the sum of the anode drop, the cathode drop, and the I- R drop
due to the resistance of the gaseous part of the are.
I do not think it is far from the truth to say that two-thirds
of the energy in the carbon arc appear as heat in the flame or
blast from the cathode. The question naturally occurs-What
is the nature of the flame? Two things can be said of it. First,
pa.rticles of very hot carbon are shot off the end of the cathode
and these draw the air with them so that the flame from a match
is sucked downward through the opening in the anode when the
a,ppa,ra,tus is set up so that the fla,me is below the torch. The
current of air was doubtless much stronger when the apparatus
was turned over, for it was not possible to get the 9/16-in. reading with 50 amperes, for the arc would not persist long enough
to permit of measurement of the heat.

June 1927



I assumed that this was due to the stronger current of air

I should like to ask Professor Compton if experiments have
through the opening in the anode when the apparatus was set been made to determine the various ionizing potentials at high
up to take the measurements contained in the table.
temperatures, because knowledge of these potentials is quite
The second thing is that the actinic value of the flame near the essential to the correct interpretation of arc phenomena.
opening in the anode is very much greater than the value in most
V. Karapetoff: Dr. Compton's paper is mainly concerned
of the flame. When the flame was focused on the ground glass with simple, steady arcs, and it is only right that an involved
the image of the flame covered nearly the whole plate, but a phenomenon should first be studied in its simplest form. In
short-time photograph showed a very small figure on the plate. practical applications, we have mostly variable arcs, and our
The pictures were taken in 1/500 to 1/1000 sec. and at this problem is two-fold: (1) To make an arc as steady as possible;
speed, not more than 10 per cent of the flame that showed on the for example, in are furnaces, in electric welding, in arc lamps,
ground glass plate appeared on the photograph.
rectifiers, etc.; or else, (2) to make an are as unstable as possible
A picture of the flame was taken with the camera behind a so as to extinguish it quickly; for example, in switches, sparkpiece of i8-in. thick pasteboard to see if the active part of the gaps, relay contacts, flashovers, etc.
flame contained X-rays. The results were negative.
In either group of problems, it is of importance to know the
When the current was reversed in direction, the apparatus factors which contribute both to the stability and instability of
refused to work as a torch and the are apparently tried to run up an are, so as to intensify the desirable factors at will. This
the carbon when it was made the positive terminal of the are. means that engineers will have to pay more and more attention
This is a most noteworthy fact, for it depends on something be- to the physical nature of the arc, and Dr. Compton's paper, with
side the electromagnetic forces. In any piece of apparatus with its references to literature, should prove a valuable introduction
which I am acquainted, the direction of motion is independent of to the subject as well as a guide to future investigators.
the direction of current, for the reversal of current reverses the
Dr. Compton quotes several empirical equations for the obflux and with both flux and current reversed, the direction of served relationship between the voltage and the current in a
motion is unchanged.
steady are. In a transient arc, or spark-over, both the current
What is this blast or flame from the cathode? Apparently and the voltage are functions of time, and the apparent total
it is not a stream of electrons, for if it were, it would be affected resistance of the arc is variable. Dr. Max Toepler' has proposed
by a magnetic field. At the same time it must be remembered the following funetion for this resistance:
that approximately two-thirds of the total energy in the arc
Rt = k F/At
appears in this flame. It is my opinion that the blast from the
Here k is an empirical constant, F, the length of the arc, and
cathode in the carbon arc is due to vaporized carbon from the
Ai the total quantity of electricity which has passed through the
carbon pencil.
We do not know much about the latent heat of carbon, but it arc from the instant t = 0, when it was struck, to the instant t
is possible and even probable that it is very high. If I am correct under consideration.
For a transient are, there is some reason for Toepler's formula,
in the opinin that most of the energy of the carbon arc is
expanded in vaporizing carbon from the carbon cathode and in that the ionized state of the gas is established only gradually,
most of ,theth het tand may be considered a function of the quantity of electricity
that most of the heat that appears at the positive
the solidification of the vaporized carbon at the anode, this which has passed through the are, the conductance increasing
would be evidence of a large amount of energy required as latent with this quantity.
On the other hand, Toepler's formula has some serious defects;
heat to vaporizecar'bon.
It is my belief that to get a more accurate conception of what
.yT r
occurs in the arc, we shall have to substitute the idea of the blast at the instant of striking, no finite voltage should be able to start
from the negative terminal as being the central and important an arc;
thing which occurs in the are, for the idea that there are inherent
2. Should the arc continue over an indefinite period of
anode and cathode drops.
its resistance, according to Toepler, shouldis drop to zero;
The tests described in this paper show that the flame is a
The ratio of the voltage to the current assumed to be
phenomenon associated with the negative terminal,
to the length of the arc; in reality there is a conproportional
The old way of looking at it would be to say that the inherent siderable and concentrated
fall of potential at the cathode, and
drop at the positive terminal was great enough to produce the some drop at the anode.
heat that actually appears there. This old conception has,
It is proposed, therefore, to generalize Toepler's formula as
I think, been shown to be wrong by these tests.
The old idea was that the current passed across the are in a follows
solid stream and that a cross-section of the current in the arc
RS = (k F + kc)/(Aj + q) + r
would be a circle.
In this expression, kl, q, and r are additional constants, introThe experiment with the torch, as well as some others not duced for the purpose of correcting the above-mentioned defects
described, indicate that the core of the arc is the blast from the of the original formula. When Q t = 0, i. e., at the beginning
negative terminal and that the current flows outside of the of the discharge, Rt is no more infinitely large, but has a high
blast and that the section of the current across the arc would be finite value, Ro = (k F + kl) /q + r. With a steady arc, when
an annulus and not a circle. In such a cross-section, the inner Qt = co, the resistance is no 'more equal to zero, but has the
circle would be the cross-section of the blast from the negative limiting low value of R c = r. Furthermore, the resistance is
terminal and the annulus outside of this inner circle would be the assumed to increase more slowly than the length F of the are,
cross section of the current. There is no doubt that this is there being a correction term kl.
the true picture of the cross-section of the current in the case of
Dr. Otto Mayr has given a general theory of condenser disthe torch, and I believe it is the true picture in any carbon arc.
charge through a resistance and a sphere-gap, using Toepler's
P. P. Alexander: I should like to ask Professor Compton formula for the resistance of a traJnsient arc2. He has also
to say a few words about the ionizing potentials of different gases. determined some va.lues of k from the available experimental
These aUre well known at ordina,ry temperatures, but at the da.ta. The next step should be to extend his theory on the basis
temperature of the arc core, apparently, they are entirely of generalized formula (b), and to determine the numerical values
different. For instance, the ionizing potential of nitrogen at of the constants which it contains.
Ordinary temperature is something like 11 volts; a.t the tempera1. Arch7ivfutr Elek., 1925, Vol.-14, p. 306.
ture of the arc core, it appears to be several hundred times less.
2. Archie fur Elekc., 1926, Vol. 17, p. 53.



E. C. Starr: I should like to ask Dr. Compton and the

gentlemen who have discussed his paper if they have any data on
the order of magnitude of the transient resistance of an are.
I have reference to the type of arc that is initiated by a potential
of ,everal thousand volts between electrodes in air at normal
pressure and temperature.
The size and shape of the electrodes, as well as the spacing,
no doubt affect the resistance considerably. For example, the
ionized path between the points of a needle-gap is not uniform
in intensity of ionaization and the effective cross-sectional area
of the path is relatively small compared to the area of the path
between large spheres or parallel disks. Hence it is to be expected that the resistance of an arc between the latter type of
electrodes should have a lower value throughout the entire period
of the transient than in the case of a needle-gap of the same
Dr. Slepian spoke of measuring the voltage transient of an arc.
Perhaps he also recorded the current transient and could therefore readily determine the resistance characteristic.
The transient resistance equation suggested by Prof. Karapetoff
should be of considerable value in the calculation of transients in
circuits containing spark-gaps if the values of the constants can
be determined.

len en:

mnatedas ter

Transactions A. I. E. E.

now know that electrons can be shot across vacuous spaces.

When an arc is struck by the two separating electrodes, there
must be present a host of ions or ionized particles, as well as free
electrons which serve as the carrier for the are. Is it not possible
to picture a condition under which these carriers may be provided
entirely from the electrodes, and not by a surrounding gas?
K. T. Compton: The work to which Dr. Slepian calls
attention is, I think, some of the most interesting in connection
with the theory of the arc. I made a reference to it in the paper,
but I should like to call attention to just one thing for fear of
being misunderstood.
There were two theories of Langmuir that have been discussed.
One theory has no reference to the origin of the electrons. It
apply independently
. whether the electrons have thermionie or any other origin. It is merely a space-charge theory.
As to the other theory, i. e., that electrons may be pulled out
of the eathode by high electric fields, I think that we have there a
possibility of two types of electric arcs. Of course there are often
two or even more types of arcs. The character may change
from one to the other, but both are recognized as arcs. It seems
to me we have brought out in this discussion opportunity at least
for two of these.
Dr. Slepian mentioned the case of a copper are in which the
was not melted, and obviously didn't get to the melting
nteohrh,d eoranyd aecpe
On the other hand, we certainly do have copper
in Table Ihe notie to tu
In Table I we notice two tungsten arcs, one with a current
amperes per sq. cm., and the other
with 700. hIn

should like Sor

to supplement what has been said by telling some of
the interesting things relating to arcs that Dr. Millikan and I
have found, as we have endeavored to produce a non-arcing
swt, fo u se on elcti circis At Caiori Intiut of
uptn. A high-voltage,
Technology ue
we have
electric circuits by means of switches enclosed in a vacuum
chamber. To date, we have been very successful in our at- temperature that the are reached. The tungsten melted, and at
the melting point of tungsten. In that case (see
tempts to do this, largely because the arc at the opening of the least reached
emission, calculated from the constants
switch very small, and apparently removes a negligible amouAt
of material from the switch terminal when the arc is struck. of tungsten, comes as near 700 amperes per sq. cm. as the purity
We have some switches showing practically no burning or pitting of the tungsten would justify. On the other hand, 3200 is
of contacts after 4000
Also, by means of relatively clearly too high to be accounted for by thermionic emission. In
small contacts, currents of several thousand amperes at approxi- tungsten we have these two different types, one evidently of
mately 50,000 volts, have been successfully interrupted. In thermionic origin and the other of different origin,-perhaps with
out (Langmuir) or perhaps arising from intense
not electrons pulled
performing these interruptions, the switch terminals.have
in front of the cathode (Slepian).
been unduly pitted and since there is no pitting of the metal,
Nottingham used a special type of are, especially designed to
it is rather difficult to account for the energy dissipation at the
reduce heat conduction so that cathode temperature could rise.
switch during the time of opening.
Dr. Compton has defined an arc "as a discharge of electricity He got a large cathode spot. In the case of carbon also we have
between electrodes in a gas or vapor, which has a negative or pretty good evidence of the large proportion of the emission
practically zero volt-ampere characteristic and a voltage drop thermionically, and in the case of Pointilite lamps and Tungar
at the cathodes of the order of the minimum ionizing or minimum rectifiers where we can use a pyrometer to determine the temperaexciting potential of the gas or vapor," all of which may be true ture of the various parts of the cathodes one gets pretty good
but we have found from our many experiences that if gas or vapor agreement with the theory of thermionic emission
In Table I, in the cases where we have current densities running
is required to maintain an are, the amount required is indeed
very small. Perhaps if we could hypothecate a liquid or gas, into thousands of amperes per sq. cm. I think, with Dr. Slepian,
which will not vaporize at arc temperature, it would still be possi- that thermionic emission is not adequate to account for currents
ble, though it may appear improbable, to start an are in such a of that order. Another agency must be operative in such cases.
There are two ways in which the temperature may affect the
liquid. I should like to ask Dr. Compton how it would affect
his definition to leave out the words "gas or vapor" and have his ionizing potential. In the first place, the gas to be ionized may
definition read "an arc is a discharge of electricity between be dissociated from its molecular state into an atomic state as a
electrodes, which has a negative or practically zero volt-ampere result of high temperature. That such action is possible has
characteristic and a voltage drop at the cathode of the minimum been shown in experimental eases where it is possible to produce
ionizing or minimum exciting potential of the material stripped this dissociation under conditions that can be controlled, namely,
from an electrode at the temperature of the arc." In other in hydrogen, iodine, etc. In those cases the effect is always to
words, is it essential that there be a surrounding medium of gas reduce the ionizing potential. This action of temperature is
or vapor in order that an arc may be struck by the electrodes,
If for the moment we assume a medium of gas or vapor not
As regards any direct effect of temperature on the ionizing
essential to the establishment of an arc, we must, of course, look potential of the gas, this effect will probably be rather small
for some other means of expla,ining the process by which an arc because translating degrees centigrade into volts, about 8000 deg.
between electrodes is sustained even for aU very short period of cent. correspond to only one volt. There are no laboratory
time. This presents a difficulty, which, however, may not be experiments that reach a temperature as high as 8000 deg. So
insurmountable. Contrary to public opinion, the best known the average energy imparted to electrons as the result of high
vacuum is not a perfect insulator, in the sense that no electric temperature or to the molecules by high temperature would in
current can be made to pass across such a vacuum, because we genera,l be only a fra.ction of a volt.


June 1927


I don't know whether or not there are other ways in which the
ionizing potential of the gas might be affected than these. The
only ones I know that have been directly investigated have been
dissociation of molecular gases into their constituents, and the
direct thermal ionization of alkaline vapors in electric furnaces,
as done at the Mt. Wilson Observatory.
With regard to the question by Mr. Starr, I am sorry that I
cannot give the desired information because I have made no
study of transient arc phenomena.
Professor Sorensen's suggestion that the ionization of materials
stripped from the electrodes at the temperature of the are be
substituted for that of a surrounding gas or vapor appears to me
to be quite permissible as including the interesting discharges
which he described as true arcs. In fact such material is ineluded in the term "vapor" in the sense that I have used. The
important thing, as I see it, is the presence of some ionizable
material in the space between the electrodes.
The great success of this current interrupter seems to be due to
the fact that, at such low gas or vapor pressures, the mobility of


the ions is so great that they effectively disappear from the are
space during the time of low voltage between voltage reversals.
In high-pressure arcs as in oil-immersed circuit breakers, on the
other hand, the ion mobility is so small that ions remain in
in sufficient concentration to re-strike the arc after the voltage
In answer to Prof. Karapetoff I wish to say that I never discuss the question of an electric are with anyone who has had any
real practical experience with an electric arc without feeling how
limited is the experience which we have in the laboratory. As I
said, we physicists work with arcs on a small scale, and the attention of physicists has been devoted to ares under the simplest
conditions in order to find out something about the things going
on in the arc. Unfortunately those aren't the arcs met with in
engineering practise, where simplicity and even understanding
of the phenomena are not the prime considerations. It may be,
I am afraid, another generation of physicists which will be able
to answer some of the questions which are uppermost in the
minds of engineers.