Advances in Cryogenic Engineering

A Cryogenic Engineering Conference Publication
Advances in
Cryogenic Engineering
VOLUME 3
Proceedings of the 1957

Cryogenic Engineering Conference
National Bureau of Standards
Boulder, Colorado
August 19-21,1957
K. D. TIMMERHAUS, Editor
Chemical Engineering Department
University of Colorado
Boulder, Colorado
Distributed by
PLENUM PRESS, Inc., NEW YORK
1960
ISBN-13: 978-1-4684-3107-0 e-ISBN-13: 978-1-4684-3105-6
DOl: 10.1007/978-1-4684-3105-6
Softcover reprint of the hardcover I st edition 1960
FOREWORD
By popular request, the National Bureau of Standards

was again a host to a conference on cryogenic engineering on
August 19-21, 1957. Similar meetings were held here in 1954
and 1956. The acceptance of over forty papers for this
conference was certainly a sign of the increasing activity and
interest in this engineering field. There seems little doubt that
it will continue to grow, justifying the need for annual meetings.
To make the Proceedings more interesting an attempt

was made to include as much as possible of the general
discussion which followed each paper.
To obtain individual reprints of anyone particular

paper, please contact the authors directly.
1957 CRYOGENIC ENGINEERING CONFERENCE COMMITTEE
B. W. Birmingham National Bureau of Standards
S. C. Collins Massachusetts Institute of

Technology
E. F. Hammel Los Alamos Scientific
Laboratory
R. B. Scott National Bureau of Standards
K. D. Timmerhaus University of Colorado
W. T. Ziegler Georgia Institute of Technology
i
ACKNOWLEDGMENTS
The 1957 Cryogenic Engineering Conference Committee

gratefully acknowledges the continued support and interest of
the following organizations who have made the 1957 Cryogenic
Engineering Conference and the publication of this Proceedings
possible.
L' Air Liquide

Air Products, Inc.
Allison Division, General Motor s
American Messer Corporation
Aro Equipment Corporation
Beech Aircraft Corporation
Bell Aircraft
Boeing Airplane Company
Cambridge Corporation
Convair
Curtiss- Wright Corporation
Garrett Corporation
General Electric Company
Herrick L. Johnston, Inc.
Hofman Laboratories
Linde Company
A. D. Little, Inc.
The Martin Company
National Bureau of Standards
Northrop Aircraft
Pesco Products --Division of Borg Warner
Pioneer Centra1--Division of Bendix Aviation
Ronan and Kunz1, Inc.
Stearns-Roger Manufacturing Company
Superior Air Products, Inc.
Thompson Products
United Aircraft
Westinghouse Electric Corporation
ii
CONTENTS
PAGE
1957 CRYOGENIC ENGINEERING CONFERENCE

COMMITTEE .. i
ACKNOWLEDGMENT . . . .. . . . . . . . . . . . . . . . . ii
PAPERS PRESENTED
Cryogenic Proces se s
A-I Hydrogen Liquefaction Cycles . • . . . • . • 1

J. Macinko, D. B. Chelton, and J. Dean
CEL National Bureau of Standards
A-2 Removal of Nitrogen from Hydrogen

with Silica Gel at Low Temperatures. 11
V. J. Johnson
A-3 Preliminary Design Studies of Low

Temperature Refrigeration Plants 19
D. Aronson
Worthington Corporation
A-4 Power Consumption and Thermodynamic

Reversibility in Low Temperature
Refrigeration and Separation Processes. 32
J. B. Gardner and K. C. Smith
British Oxygen Research and Develop-
ment Ltd.
A-5 The Reversible Separation of Multi-

component Mixtures . • . • • • • . • • . 47
H. M. Scofield
Linde Company
iii
PAGE
A-6 Separation of Hydrogen Isotopes by

Multicomponent Distillation • . • . . . 58
T. M. Flynn, K. D. Timmerhaus,
D. H. Weitzel, and J. W. Draper
Cryogenic Processes (continued)
B-1 Efficient Utilization of Ortho-Para

Catalyst . . . . . . I ••••••••••• 64
C. R. Class, R. P. Spero, and
G. E. McIntosh
B-2 Design Data for Ortho-Parahydrogen

Converters • • . • . • 73
D. H. Weitzel, C. C. Van Valin,
and J. W. Draper
B-3 Vapor Phase Ortho-Para Conversion

in the Large CEL-NBS Hydrogen
Liquefier . . . . . . . . . . . . . . . . . . 85
W. A. Wilson and D. H. Weitzel
B-4 Design and Application of a Gas

Liquefier . . . . . . . . . . . . . . 92
T. W.· Schaffers
Philips Electronics, Inc.
B-5 The Thermodynamic Theory of a

Liquid Nitrogen Generator Using a
Norelco Refrigerator-Liquefier .• 99
F. G. Brickwedde
North American Philips Company
iv
PAGE
Cryogenic Applicactions
C-l Design and Performance Data on a

Laboratory Size Liquid Nitrogen
Plant . . . . . . . . . . . . . 106
G. Y. Robinson and J. T. Selldorf
Arthur D. Little, Inc.
C-2 Economic and Other Aspects of the

Distribution of Navy Helium in
Liquid Form •••••• 114
R. E. Brown
Bureau of Aeronautics
Department of the Navy
C-3 Technical Aspects of Large Scale

Liquid Helium Liquefaction and
Transportation ••••••••••••• 125
D. B. Mann, B. W. Birmingham,
and P. C. Vander Arend
C-4 Dilution of Cryogenic Liquid Rocket

Propellants During Pressurized
Transfer • • • • • • • • • • . • • •• 136
S. Greenfield
Rocketdyne
C-5 Liquid Oxygen for Missile Testing 149

W. J. Scharle
Air Products, Inc.
C-6 Unattended Oxygen Plants. .. .... 160

W. W. Aton
Linde Company
Cryogenic Equipment
D-l On the Design of a Cryogenic

Facility for Testing Aircraft
Components ••••••••••••••• 172
N. Meyers, H. Rind, and P. Gomez
Stratos
v
PAGE
D-2 Liquid Helium Temperatures in an
Atomic Reactor .173
J. T. Howe, R. R. Coltman, and

T. H. Blewitt
Oak Ridge National Laboratory
D-3 Integral Probe Capacitance Gaging of

Liquid Gas Container Contents 179
H. J. Smith
The Aro Equipment Corporation
D-4 A Unique Thermal Conductivity Gas

Analyzer ••••••.••..• 191
J. R. Purcell, J. W. Draper, and
D. H. Weitzel
D-S Piston Type High Pressure Aircraft

Cryogenic Pumps• • • • . • . • • • • • • • 196
R. H. Ball
Sundstrand - Denver
D-6 Operation of Bearings and Pumps at

Low Temperatures•••••••••••• 209
K. B. Martin, R. B. Jacobs, and
R. J. Hardy
D-7 1000 Gallons Per Minute Liquid Oxygen

Transfer Unit . . . . . . . . . . . . . . . . . . 217
D. W. Seavey
D-8 Contaminants in Liquid Oxygen as Related

to Safety in Liquid Oxygen Production
and Distribution Equipment •••••.•• 218
C. D. Arrick
Linde Company
vi
PAGE
Cryogenic Equipment (continued)
E-l A Cryostat for Liquid Nitrogen Cool-

ing of Rocket-borne Photomultiplier
Tubes .. t •••••• 226
J. E. Tracey
F. K. Dearborn
Air Fa rce Cambridge Research Center
E-2 Beech Aircraft 6000 Liter Liquid

Hydrogen Dewar • • • • • • • . . • • . 232
V. E. Isakson, C. D. Holben, and
C. V. Fogelberg
E-3 Vacuum Insulated Dewars for Liquid

Oxygen and Liquid Nitrogen •••.•• 238
H. A. Eichstaedt
Ronan and Kunzl, Inc.
E-4 A Precision Low Temperature X-Ray

Camera. . . . . . . . . . . . . . . . . . . . 248
E. G. Eeles
National Research Council of Canada
E-5 Design of Simple DC Resistance

Thermometer Bridges for Wide
Range Temperature Control ••• 254
R. D. Goodwin
E-6 Calibration of Thermocouples at

Low Temperatures •••• 269
M. D. Bunch and R. L. Powell
vii
PAGE
Cryogenic Propertie s
F-l Vapor-Liquid Equilibria in the System

Hydrogen-Nitrogen-Carbon Monoxide .. , 275
W. W. Akers and L. S. Eubanks

The Rice Institute
F-2 Vapor-Liquid Equilibrium Studies on

the System Argon-Oxygen •••••••• 294
D. I. J. Wang
Linde Company
F-3 Epoxy Resins as Cryogenic

Structural Adhesives 305
R. M. McClintock and M. J. Hiza
F-4 The Impact Testing of Various Alloys

at Low Temperatures ••• , ••••• 316
R. p, Mikesell and R. P. Reed
F-5 Performance Prediction of Low

Temperature Air-cooling Extended
Surface by Several Comparative Methods 325
C. F. Kayan and R. G. Gates
Columbia University
F -6 Temperatur e -Pre s sure - Time Relationships

in a Closed Cryogenic Container ••••• 342
S. C. Huntley
National Advisory Committee for
Aeronautics
viii
PAGE
Cryogenic Propertie s (continued)
G-l Calculation of Gaseous Heat Conduction

in Dewars . . . . . . . . . . . . . . . . . 353
R. J. Corruccini
G-2 Measurement of Thermal Stresses at

-320 o F by Strain Gages ••.•• 367
J. P. Hunsaker and A. Pastuhov
G-3 Heat Transfer to Boiling Liquid Nitrogen

and Hydrogen Flowing Axially Through
Narrow Annular Passages • • • • • • • • • •• 375
R. J. Richards, R. F. Robbins,
R. B. Jacobs, and D. C. Holten
G-4 Friction Studies of Various Materials

in Liquid Nitrogen ••.••••• 390
D. W. Wisander, W. F. Hady, and
R. L. Johnson
National Advisory Committee for
Aeronautics
G-5 Properties of 9 Per Cent Nickel Low

Temperature Steel ••••••••••••• 407
J. G. Althouse
Lukens Steel Company
G-6 Thermal Conductivities of Common

Commercial Aluminum Alloys 408
W. J. Hall, R. L. Powell, and
H. M. Roder
G-7 A Comparison of Insulating Materials. 416

J. E. Schrodt, M. M. Fulk,
W. A. McReynolds, and W. S. Flanery
ix
A-I
HYDROGEN LIQUEFACTION CYCLES
J. Macinko, D. B. Chelton, and J. Dean

Boulder, Colorado
Introduction
Because of increased activities in cryogenic engineering,

liquefaction cycle s, . in particular hydrogen cycles, are under-
going a more thorough thermodynamic analysis than had been
justified previously.
This is a brief summary of the results of the cycle

studies conducted at the Boulder Laboratories of the National
Bureau of Standards.
In the thermodynamic analysis of cycles, certain

assumptions are necessary in order to provide a consistent basis
for calculations.
The following assumptions were made in this particular

work.
1) The heat leak from the surroundings into the system

was neglected in all cases.
2) Expansion engines, where used, were assumed to

have an efficiency of 80 per cent.
3) A minimum heat exchanger warm end ~ T of 10 K

was set for comparison of cycles.
4) Again for the comparison only, ortho to para

conversion was restricted to the precoolant bath and
the liquid receiver.
5) The work of compression per unit volume of liquid

produced has been taken as the standard for
comparison.
-1-
Parahydrogen
Concerning parahydrogen, considerable variation in cycle

performance results from different catalyst locations. In
addition, several methods of correction for para concentration
have been used in the past when setting up heat balances in
parahydrogen liquefiers. Figure 1 illustrates the methods used
in our analysis and points out a significant factor in the use of
a catalyst.
Since the enthalpy of hydrogen is a function of pressure

P, temperature T, and para concentration C, a correction must
be made to the enthalpy of normal hydrogen for para concentra-
tions other than normal. Since the heat of conversion from ortho
to para is 168 cal/gm, and normal hydrogen is 25 per cent para,
the enthalpy, in cal/gm, of hydrogen with a para concentration of
C per cent is equal to the enthalpy of normal hydrogen minus
1. 68 (C-Z5).
The schematics represent two methods of using the

conversion catalyst. In system A the catalyst is enclosed and
only the liquid drawn off passes through the catalyst. In system
B, an open bed catalyst is used resulting in partial conversion
of the return stream. Comparing heat balances, the fraction
yield in A is 19 per cent, in B the yield is 16 per cent. The
difference occurs because of the increase in para concentration
of the return stream of system B.
Simple Linde Cycle
The simple Linde cycle shown in Figure 2 is one of the

most elementary methods of liquefying hydrogen and is the cycle
used in most present day hydrogen liquefiers. The compressed
high pressure gas passes through exchanger I, and is precooled
by liquid nitrogen boiling at atmospheric pressure in reservoir A.
Further cooling is obtained by passing through exchanger II, then
through reservoir B containing liquid nitrogen boiling under
reduced pressure. The high pressure gas then enters exchanger
III • expanding isenthalpically through the J -T valve into the
liquid receiver, C. The unliquefied portion returns through the
low pressure side of the exchangers, giving up refrigeration to
the high pressure stream.
For each cycle one parameter has been chosen to

demonstrate some of the performance characteristics of the
given cycle. For the simple Linde cycle the parameter is the
-2-
He (col / gm) • f (P, T, C) • Hn- 1.68 (C-25)
.,-" .,-"
PRECOOL.ANT PRECOOLANT
!50". PARA ~% PARA ~O.,. PARA

Hl- 18!5 Ho· 241 Hl -IB!5
A B
PRECOOLING TEMPERATURf; I_ se K
EXCEPT AS NOTED
SIMPLE l1NOE CYCLE
60 00
YIELD' 19~ YIELD· 16~ INLET PRESSURE - ATM
Figure 1. Figure 2.
, 15ATIoI H, ,1
"'IroAl.RA Hz
,
"'.9''Iro~VERS'i
0' I'-.. AT 20K
.. He PRESSURES
20~
"AT
10AT
.0
~
•
.0
•
. . .
HELIUM· HYDROGEN
CONDENSING CYCLE CLAUDE CYCLE
0 2
Ht PRESSURE· ATM
00
" 10101040110 00
INLET PRESSURE· ATM
Figure 3. Figure 4.
-3-
inlet pressure, plotted along the abscissa. The work of
liquefaction in kw. hr. / gal. of liquid is plotted as a function of
the inlet pressure for several conditions of para conversion
and heat exchanger .6. TIs.
All curves indicate a decreasing work figure for increas-

ing pressures.
The bottom curve is for normal hydrogel1:, 6S o K pre-

cooling, and a 10 AT at the warm end of the final heat
exchanger.
The next curve shows the effect of producing 95 per

cent parahydrogen when partial conversion is carried out in
the precoolant bath and final conversion in the liquid receiver.
If all of the conversion is done in the liquid receiver,

the work increases to the values indicated by the next curve.
Comparison of the top curve with the bottom curve

indicates the effect of increasing the inlet temperature. The
top curve is for BOoK while the bottom curve is for 6S o K.
Helium-Hydrogen Condensing Cycles
From this point on, the discussion will be limited to

that portion of each cycle below the precoolant bath.
Figure 3 shows the schematic of a cycle which we have

called the helium-hydrogen condensing cycle. In principle it is
a helium refrigerator used to condense hydrogen.
Helium, at a pressure of 10 to 20 atmospheres, is pre-

cooled by liquid nitrogen, undergoes further cooling in the heat
exchanger, then expands through an expansion engine to a
temperature low enough to condense the hydrogen in exchanger
III. The condensed hydrogen expands to atmospheric pressure
into the liquid receiver. The vapor formed in flashing returns
through the heat exchangers and provides additional refrigeration
to the incoming gas.
The performance curves have several significant charac-

teristics. The work of liquefaction for several helium pres sures
is plotted as a function of the hydrogen pressure. The three
lower curves for 10, 15, and 20 atmospheres respectively, are
-4-
almost superimposed on each other indicating the insensitivity
of the cycle to fluctuations in helium pres sure in this range.
Also of interest is the flatness of the curve between 3 and 10
atmospheres of hydrogen pressure indicating extremely stable
performance in this region.
Note the sizeable increase in the work required to pro-

duce parahydrogen a,s compared to the work for normal hydrogen
Most cycles have an increase of 20 to 25 per cent going from
normal to para but in this cycle the increase is about 50 per
cent. The reason or reasons for this large increase are not
yet fully understood.
Claude Cycle
Although the true Claude cycle employs no precoolant, if

we consider only the portion of this cycle (Figure 4) below the
precoolant bath it is identical to the Claude cycle as used in
liquid air plants. Due to this similarity we have chosen to apply
the Claude name to the cycle.
The high pressure inlet gas of the Claude cycle is pre-

cooled in the same manner as in the simple Linde cycle,
attaining a temperature essentially the same as the precoolant
bath.
At point 1 the £low is split and a fraction passes through

the expansion engine where it is cooled and exhausts into the low
temperature end of exchanger II.
After passing through exchanger III, the high pressure gas

expands through a J -T valve. The unliquefied portion returns
through the low pressure side of the exchangers, cooling the
incoming high pressure stream.
Plotting the work of liquefaction as a function of inlet

pressure, variation of the engine inlet temperature results in
the family of curves shown. The curves are for inlet temper-
atures from BOoK for the top curve down to 50 0 K for the
bottom curve in 50 increments.
Temperatures to 65 0 K are readily obtainable by pre-

coolant baths. However, in order to obtain inlet temperatures
-5-
below 65 0 K it is necessary to use a heat exchanger between the
precoolant bath and the engine inlet.
The region below the dotted line results in two phase

discharge from the engine. This presents serious design
problems but has been included to show the trend of the
curves.
Dual Pressure Cycle
What we have chosen to call the dual pressure process

shown in Figure 5 has a two stage throttling arrangement
(Valves I and II). The high pressure gas first undergoes
expansion to an intermediate pressure where the flow is split.
A small portion just slightly more than is actually liquefied
is then expanded to atmospheric pressure. The rest is returned
to the compressor at the intermediate pressure. The reduction in
the work of compression is greater than the reduction in the
fraction liquefied, resulting in a more favorable figure for the
overall work of liquefaction.
Linde used a dual pressure process on air liquefaction

but did not include a heat exchanger between the intermediate
pressure pot and the liquid receiver. When applied to hydrogen
liquefaction a gain in yield on the order of 20 per cent is
experienced by introducing heat exchanger II.
The performance curves give the work of liquefaction

for several inlet pressures as a function of the intermediate
pressure. There are several significant features to be noted.
1) The work of liquefaction is at a minimum when the

intermediate pressure is about 1/2 that of the inlet
pressure, and increases rapidly as the intermediate
pressure approaches the inlet pressures.
2) The higher inlet pressures, 80 to 100 atm, have a

wide range of intermediate pressures over which
the work of liquefaction remains fairly constant.
For example, between 15 and 40 atmospheres
intermediate pressure the variation in work is less
than 3 per cent.
-6-
3.0
95't,PARA HZ
58 -95'1.CXJNVERSION I50 ATM '60 ATM\
·t
40ATM ,
AT 20'K~ ..
) l ......" ..
7~
5
-
......[7'" 80 ATM
:;,:::;;
, 100ATM
0
HEAT EXCHANGER AT. I' K
PRECOOLING TEMPERATURE· 60' K
PRODUCING NORMAL H2
0
DUAL PRESSURE CYCLE
o 10 20 30 40 50 60
INTERMEDIATE PRESSURE -ATM
Figure 5,
2.0
90'l.PARA H2 ~:~.) ......
- ~ ~:O'K
ca-9S'l.CXJNVERSION
AT 20·K-~"'II'" ... ot."" .. ""1
",. 30-K
6
~ '7'
2
8
INLET PRESSURE' 100 ATM
PRECOOLING TEMPERATURE '65'K
PRODUCING NORMAL H2
.4
DUAL PRESSURE CYCLE

WITH EXPANDER
o 10 30 40 60 75 90
INTERMEDIATE PRESSURE· ATM
Figure 6,
7.5
I
~
...J
<3..... VSIMPLE LINDE
I
----P~ESSURE
a: 6.0
l:
~, ILHe-~ CON::ING
/OUAL
t5 4.0
~
~
~
o
3.0
/ r---. ................
[CLAUDE - 6S'K ENGI NE
\
DUAL PRESSURE
:J WITH EXPANDER
u.
o
:!It: I.S
I
a: PRECOOLING TEMPERATURE· 6S'K
~ HEAT EXCHANGER AT, I'K
WORK INCLUDES PRECOQLANT
o 20 40 60 80 100 120
INLET PRESSURE -ATM
Figure 7.
-7-
3) For the production of parahydrogen the work of
liquefaction increases about 25 per cent while
retaining the same general characteristics.
The dotted lines are for 80 atmospheres inlet, the top

one for parahydrogen and the bottom one for normal hydrogen.
Dual Pressure Cycle With Expander
Because of the additional cooling obtainable by isentropic

expansion, a dual pressure cycle using an expansion engine in
place of the first throttling valve (Figure 6) has a thermodynamic
advantage over the two valve system. The second valve is
retained to avoid operating an expansion engine in the two-
phase region. The flow pattern is the same as the dual preSlllre
cycle just described. The difference in performance is due to
the greater amount of refrigeration provided in the first expan-
sion. This increases the yield and results in a reduction in the
work of liquefaction.
The performance curves give the work of liquefaction

for several engine inlet temperatures (30, 40, and 4S o K) as a
function of the intermediate pressure. The inlet pressure is
100 atmospheres for all curves. The performance of the
cycle is more sensitive to changes in intermediate pressure
than the cycle without an engine as is shown by the slope of
the curves. The minimum work figures all occur in the region
of 30 atmospheres intermediate pressure.
Conclusions
In Figure 7, a comparison of the cycles is made on the

basis of the production of 95 per cent parahydrogen, including
the work required to produce the precoolant.
Comparing minima, the simple Linde is about 50 per

cent higher than the Claude, the helium-hydrogen 25 per cent
higher, the dual pressure cycle 20 per cent, and the dual
pressure with expander about 4 per cent higher.
Although a thermodynamic comparison is of prime

importance, other factors enter into the final selection of the
most practical system such as initial cost, reliability,
simplicity, and maintenance costs.
-8-
When these other factors are considered,the dual pressure
cycle becomes more interesting. It has the advantage of
simplicity combined with a moderate efficiency. There are no
moving parts at low temperatures, and the system is relatively
insensitive to minor fluctuations in pressure, temperature,
and flow.
Substitution of an expansion engine for the first

expansion valve reduces the work of liquefaction to a value
comparable to the Claude cycle. Although this requires moving
parts at low temperatures the problems are not as serious as
in the Claude cycle. The high inlet pressure and the small
pressure ratio of about 2.5: 1 for the dual pressure cycle lends
itself readily to a reciprocating expander, and engine efficiency
is not a critical factor in the performance of the cycle.
The design of a liquefier incorporating a parallel system

comprised of a valve and an expansion engine for the first
stage of expansion provides even greater flexibility. Since the
optimum operating pressures are similar for both systems, a
plant operating on an engine would not be rendered inoperative
due to engine hilure. The engine can be by-pas sed by the
expansion valve and operation continued at a rate of about 70
per cent of original capacity. When the practical aspects of a
dual pressure cycle are evaluated. the cycle warrants serious
consideration in the choice of a hydrogen liquefaction system.
e specially in a system of a size below which expansion engines
can be effectively utilized.
Discussion
Question by A. Latham. Jr •• Arthur D. Little. Inc.: Is the

power required for the production of liquid nitrogen used for
precooling included in the values shown for the production of
liquid hydrogen?
Answer by author: The graphs for the individual cycles were

drawn up to show the characteristics of the cycles and did not
include the power required to produce the precoolant. However.
the final graph giving a comparison of the optimum values of
the cycles includes the work of liquefaction for nitrogen. The
figure used was 1.91 kw. hr. / gal. for the liquid nitrogen at
65 0 K in the pumped bath.
-9-
Question by M. P. Sweeney: Did the Claude cycle shown on
the last graph expand into the two phase region?
Answer by author: The solid line of the Claude cycle

represents gaseous discharge from the expansion engine. The
dotted line represents the region of two phase discharge.
-10-
A-2
REMOV AL OF NITROGEN FROM HYDROGEN WITH

SILICA GEL AT LOW TEMPERATURES
V. J. Johnson
Boulder, Colorado
This paper gives the results of about twenty full scale

purification tests made on the silica gel purifiers installed in
the large CEL-NBS hydrogen liquefier for final purification of
electrolytic hydrogen. These results are compared with
predicted capacities made by W. T. Ziegler of the Georgia
Institute of Technology. The increasing interest in purification
of hydrogen for many uses, including liquefaction, prompted
this evaluation program.
Most of the impurities found in hydrogen can be

removed quite successfully by condensing and freezing out on
heat exchanger surfaces using conventional methods of refrig-
eration. Carbon dioxide, if present, is usually removed,
however, in caustic scrubbers. Oxygen can be removed by
combining with hydrogen using a palladium catalyst at room
temperature or by some other means a.t elevated temperatures
forming water. Nitrogen and carbon monoxide are not so
easily removed since their condensing temperatures are below
the level of conventional sources of refrigeration and also
because suitable heat exchanger schemes have not been
developed. Adsorption on charcoal or silica gel at liquid
air or liquid nitrogen temperatures is the method commonly
used.
In the CEL-NBS hydrogen liquefier system (Figure 1)

which uses electrolytic hydrogen for supply gas, a catalyst
consisting of palladium deposited upon alumina is used to
combine the oxygen with hydrogen to form water. After
compression, an oil and water separator is used to remove
condensate. The remaining oil and water vapors are con-
densed and frozen out in a refrigeration drier which is a
counterflow heat exchanger enclosed in a high pressure shell
with one end at about -190 0 F and the other end at about room
temperature. All the impurities except nitrogen are thus
removed. After the drier the gas enters another heat
exchanger and is cooled to near liquid nitrogen temperature.
It then enters the silica gel adsorber which removes the
-11-
LIQUID N.... LIQUID H.
I LIQUID
I HZ
~t:~I-~I------j l
I
l
VACUUM
REFRIG.
DRIER
ADSORPTION
PDR u:
LIQUEFIER I
~S
H I
r ----J
8'
OIL a
WATER I
SEPARATOR
I~:
-~I~II
__ ~_ ...JI
I-COMPRESSOR NZ VACUUM PUMP
I
• r....,
A~4~fr 4IE~~i,
FRO TO
COM. LIQUEFI ER
REFRIGERATION ADSORPTION
GAS SUPER- HYDROGEN SUPPLY DRIER PURIFIER
HOLDER CHARGER
Figure 1. GEL-NBS hydrogen Figure 2. GEL-NBS refrigeration

liquefier system and adsorption purifier
1000
"ii
0'
0
(.)
......-....
I/)
I-- ~
100
Go
.... ....
I/)
....
......-
<II
Z
10
EXPERIMENTAL DATA
--EXTRAPOLATION BASED ON DATA
> OF DREVING ET AL.
I 11111111 I I 1111111 I I II f
1.0
RELATIVE PRESSURE. P/po
Figure 3. Adsorption isotherm for nitrogen on silica gel
-12-
nitrogen impurity. The pure gas then returns through the
heat exchanger and goes to the liquefier. It will be noted
in Figure 2 that the total flow in this system thus passes
through the purification train. In some systems the purifica-
tion is done on just the supply gas. The recompression and
removal of oil from the recirculating gas is handled
separately.
The liquefier was by-passed during the adsorption

measurements. The system was filled with pure gas and
measured quantities of nitrogen gas were injected into the
circulating stream just ahead of the compressors. Saturation
of the silica gel coil was then monitored by sampling the gas
at the inlet. outlet, and at three intermediate taps which were
located 1/4, 1/2, and 3/4 the distance along the coil. The
samples were continuously analyzed on a sensitive thermal
conductivity instrument developed and built by E. R. Weaver
at the National Bureau of Standards in Washington. The
3-3/8 inch 1. D. adsorption coil is 51 feet long. It holds
approximately 3. 1 cubic feet of silica gel. The experimental
program was arranged to determine the effects of three
variables:
1) Impurity level,
2) Temperature of silica gel bed, and
3) Pressure.
Figure 3 shows the amount of pure nitrogen adsorbed

at equilibrium by a unit volume of silica gel. The solid line
represents data of Davison Chemical Company. The dashed
part is an extrapolation based on the findings of Dreving et al. 1
The pressure of the nitrogen is denoted by p and the pressure
of saturated nitrogen vapor at the chosen temperature by Po'
Thus this graph represents an infinite number of adsorption
isotherms. one for each value of Po' Accordingly it may be
used to determine the amount of nitrogen that will be adsorbed
within its range of applicability. However, it is not valid
when a second gas is present. The presence of high-pressure
hydrogen has the effect of reducing the amount of nitrogen that
the silica gel will adsorb.
1Dreving, Kiselev and Likhacheva. Zhur. Fiz. Khim, 25,

710 (1951); Chern. Abst. 47. 4164 (1953) •
• 13-
Figure 4. Enhancement factors for nitrogen vapor in
hydrogen as computed by W. T. Ziegler.
Figure 5. Computed adsorption capacities of silica gel for

nitrogen in hydrogen at a concentration of 1000 ppm.
The dashed curves are computed simply from the
adsorption isotherm (Figure 3) on the as sumption that
the capacity is unaffected by the presence of high
pressure hydrogen. The solid curves utilize the
enhancement factor, reducing the corre sponding value s
of pIp by the appropriate factor from Figure 4.
o
Figure 6. Concentration of nitrogen in hydrogen for equal

amounts of nitrogen adsorbed by silica gel. A,
computed without taking account of the enhancement;
B, observed in the CEL-NBS purifier for 109 ft3 of
nitrogen per ft3 of silica gel; C, computed using
~nhancement factors for 119 ft3 of nitrogen per ft 3
of silica gel.
Figure 7. The adsorptive capacity of the silica gel used in

the CEL-NBS purifier compared with the capacity
computed by two different methods. A, 900 psi, 70 o K;
B, 900 psi, 77. 4 o K; C, 1800 psi, 70 o K; D, 1800 psi,
77.4 o K. Observed data are shown as circles.
-14-
.
/
-'~r---r---.----r---.---.----r---.---_r---r---'
t,
1
~-----j'-'--'---'-
•
II I---I--l---+--+---t---I-----l-".
'0
" .1.0.-::'
.
1:1
..
~ 1~1----+_--_+--~~.---,-~.-~~~~--~----+---~.-_-.--.-4---1
I
,0
..... - .....
'/ "~ 125r_--+-~·~"t----r~_,~-+.-~·~--'r_--+_--_+----I_--+_--_J
.0
/ V ~
"C--..-,. .-+-_.Jr-- r--....!2:.!.
".'>., ,/
" - I11
100
/ ,/ S ~ r___
i,/;""- 't!:!! I'---
~,:::::--
~k~ kr-::::::
. --
:
10
~
1 l."."V
V ~ 15111f'~,'L-+----+--~1---+---~--~----+---~---4----4
0 ~O~--~---L--~~--~.--~,OL---~IZ----1---~..L---l---~.O
• 10 12 II zo
PRESSURE. psi )( 162 " PRESSURE. pil X 1d'1
Figure 4. Figure 5.
__r_--_.
, .........
ZOOr---r----.----.----~--_,---_r--
............. ,.
". f-----I------Ih-·.:"c.'--·+----h ",.",--'I------l------t------l
"
II
1
"" ">/
~
I'-...
~ ............
- --
\0--.. V .-/'
• ../' V
"'-- < 11."---
o
o •
PRESSURE • psi X ICS1
10 12 ." I. I. zo " S
CONCENTRATION OF NITROGEN. ppm X 10. 1
• 7
Figure 6. Figure 7.
-15-
It has been found that a condensed phase (liquid or
solid) exhibits a higher vapor pressure when it is exposed to
a high pressure gas. This results in a higher concentration
of the vapor in the surrounding high pressure gas. Dokoupil 2
has made theoretical studies and measurements of this effect
for nitrogen and carbon monoxIDe in hydrogen and has
expressed his results in terms of lIenhancement factor,lI the
ratio of the observed concentration of the nitrogen in the vapor
phase to that computed from the ordinary vafor pressure-
temperature relations for nitrogen. Ziegler has extended
Dokoupil ' s theory to consider the enhancement of the equilibrium
pressure of adsorbed nitrogen in the presence of high pressur e
hydrogen.
Figure 4 shows the enhancement factors for nitrogen

in hydrogen gas at 77. 4 0 K and 70 0 K and at various pressures.
Figure 5 shows the expected capacity of a silica gel adsorber
for a nitrogen impurity of 1000 ppm of hydrogen for various
pressures at 77. 4 0 K and 70 o K. It will be noted that the
capacity is much less than that computed from Figure 3 under
the assumption that the high pressure hydrogen has no effect
on the equilibrium pressure of the adsorbed nitrogen. Figure 6
shows a plotting of the first real evidence in our tests of
confirmation of the enhancement theory. At the end of a 6000
ppm saturation test at 77. 4 0 K the coil was isolated, being
careful to maintain the 1800 psi pressure and 77 0 K temperature.
Since an excess of nitrogen impurity accumulated just as the
coil reached full saturation, the total nitrogen impurity in the
free hydrogen gas rose to about 10,000 ppm. While slowly
venting the coil (about 1 hour from 1800 psi to 1000 psi and
2 hours from 1000 psi to 200 psi) it was observed that the
silica gel not only adsorbed the excess nitrogen in the
surrounding gas down to 6000 ppm as might be expected but
continued to purify it down to a level of about 4000 ppm between
400 and 800 psi. From this the conclusion was drawn that
had the test been run at 400 to 800 psi instead of 1800 psi the
coil would have adsorbed the same amount of nitrogen for an
impurity level of 4000 ppm as it did for 6000 ppm.
2 Dokoupil, van Soest and Swenker, App!. Sci. Res. A5,

182 (1955); Commun., K. Onnes Lab., Univ. Leiden, No. 297.
3Ziegler, W. T., Unpublished data, CEL-NBS, Boulder

(1957) .
-16-
Figure 7 plots the adsorptive capacity of silica gel for
nitrogen in hydrogen in concentrations up to 8000 ppm for four
specific conditions of temperature and pressure. The solid
lines are the capacities computed using the enhancement
theory. Note the order of capacities, i. e., A> B> C>D. The
dashed line s are the capacities computed from the adsorption
isotherm, Figure 3, for a value of p equal to the vapor
o
pressure of nitrogen at 77.4 o K. Note here D>B, i.e.,
greater capacity for greater total pressure. The circled points
are measured capacities of the CEL-NBS purifier resulting
from some twenty tests at the conditions indicated by the
identifying letter. The dotted lines show the trend of the tests.
In all cases the relative positions of the points agree with the
enhancement theory and not with the partial pres sure considera-
tion. The fact that the total amount adsorbed did not come up
to the predicted amount for high capacity silica gel could
possibly be from aging of the gel (used since 1952), not as
much silica gel in the coil as calculated, or possibly full
adsorption equilibrium was not reached.
The principal conclusion from the tests to date is that

the most efficient pressure range for maximum nitrogen
adsorption on silica gel is from about 300 to 900 psi and that
a reasonable increase in amount of nitrogen adsorbed results
as the temperature of the adsorbent is lowered. However, at
1800 psi the increase due to lowering the temperature is not
as significant as at the lower pressures because of the
rapidly increasing enhancement factor at the higher pres sures
and lower temperatures.
The author wishes to thank W. T. Ziegler for his

valuable advice and suggestions in evaluating the tests. In
addition acknowledgement is given to W. A. Wilson and others
on the liquefaction staff for running the tests and to J. R.
Purcell for his work in rebuilding and operating the Weaver
thermal conductivity analyzer.
Discussion
Question by W. H. Williamson, General Electric Company:

Is there a relation between the adsorption capacity and the
freshnes s of the silica gel? Doe s regenerated silica gel have
equal adsorptivity?
Answer by author: No conclusive evidence has been obtained

conce rning the deterioration of silica gel with time.
-17-
Question by L. Garwin, Kerr-McGee Oil Ind.: Is the
enhancement factor calculable from the effect of increased
pressure per se on the fugacity of the nitrogen impurity,
since the nitrogen is regarded as adsorbed as a condensed
phase?
Answer by author: Dokupil's equation takes into consideration

both the enhanced fugacity of the condensed phase and the non-
ideal behavior of the vapor phase.
Question by D. White, Cooper Bessemer Corporation: Did

you take any data showing the variation of the nitrogen
composition at the outlet of the purifier with the amount of
gas passed through the purifier?
Answer by author: Yes. Both inlet and outlet as well as

other sample taps were analyzed with respect to the same
reference gas.
-18-
A-3
PRELIMINARY DESIGN STUDIES OF LOW TEMPERATURE

REFRIGERATION PLANTS
D. Aronson
Worthington Corporation
Harrison, New Jersey
Introduction
The design of a plant for the liquefaction of gases at

extremely low temperatures can be investigated with many
variations in cycle arrangements, choice of working fluid,
operating pressures and component characteristics. One has
the problem of judging the reasonableness of the choices made.
In a recent paper Gifford l discusses in general terms the
requirements for the optimum plant for the liquefaction of
parahydrogen, pointing out that:
1) Vapor cycle refrigeration should be used for

cooling to approximately l40 0 K.
2) Expansion device refrigeration should be used

for cooling from approximately 140 0 K to lower
temperatures.
3) Care should be exercised to avoid use of

refrigerants generated at low temperature for
cooling at a much higher temperature.
4) As many steps of refrigeration levels should

be used as is feasible.
5) Great care should be taken to optimize the system

for the cooling below aOoK since most of the
compressive work will be required there.
lW. E. Gifford, Proceedings of the 1956 Cryogenic

Engineering Conference, p. 1, Boulder, Colorado, September
1956.
-19-
Cycle Analysis
The present paper suggests a method of applying

numerical values to an evaluation of a given plant performance
to that of an ideal refrigeration plant. The latter is necessar-
ily one which has all processes carried out reversibly with
zero losses. A cycle for such a plant might well be the
Brayton cycle. also known as the "dense air." or Claude cycle.
It is shown on a temperature-entropy representation in Figure
I and in terms of physical components in Figure Z. The
refrigeration load is shown as the cooling down of a process
gas of constant heat capacity. Cpt from temperature Th to
temperature T g'
For the purpose of simplifying the analysis. yet

holding to a broad approach. the refrigerant and process
gases are shown as separate streams. The procedures
suggested here can readily be modified to apply where the
refrigerant and process fluids are one and the same.
In order to simplify the analysis still further. the

refrigerant gas will be treated as an ideal gas--a reasonable
approximation for the case of helium at moderate pressure and
down to a temperature of ISoK. For such an ideal gas. the
Joule- Thomson coefficient is zero. ( aT / ap)H • O. and C 6 T
• 6H. P
The author has chosen to use the logarithm of the

absolute temperature as the Y ordinate. since it offers some
advantages in the presentation. For an ideal gas (or a real
gas of constant C ) constant pressure lines are then straight
lines with slope e~ual to the reciprocal of heat capacity. In
(TZ/T 1) • 6 S / C p ' In addition, the representation of absolute
temperature on a logarithmic scale rather than an arithmetic
scale gives a good physical picture of the magnitude of the
cooling problem at different temperature levels. since the
physical dimensions of the Y ordinate are proportional to the
power required for refrigeration over a particular temperature
range.
Using the Brayton cycle one cannot give consideration

to the first principle advanced by Gifford. The following
analysis is most suited to cycle studies at low temperatures.
regardless of whether the limit be 1400 K or some other higher or
-20-
1000
900
v
800
q,¥
700 I MULTI-STAGE
COMPRESSORS
II 1t
600 -
0
1/ ,:;
Jf
500
TEMPER ~TURE
il,nKin =oF.4160
tl
II
tIl
~l
300
fI
~ A~ 1/
V rfj' !I
tJ W
Tc
V
;r--r-- EXPANDER
!\ / J
200
Te
/
V ) ~--- 7 8
Figure 2. Low temperature
100 35
/ 4.0 4.5
/~
5.0 5.5 6.0 6.5 7.0
gas cooling flow diagram
ENTROPV-S-Bt.u/(lbOR)
Figure 1. Low temperature

gas cooling system
600 60
Torr ~ (:7 Tertii ~. ..~-
/ r_Th 55 0
500 IJ.V Jil-r..
500
rJ.y /1
,
/ ~/-
~ 45 0 c'
I 40
7 ~7
400
T E~PER! TURE / / o rt~~PE ,,,=OF+
35 o nKi
ATUR
~ 7 \ /
anKine =OF... 4 0
/ V / 300
/ \V/
0
I
/ / J
If
--;f&
/ ii/I /
250
/ V/
-
200
-/ i
/ L
20 0 If
I(
T· I TI. J I"-1f
II
150
II 8l-S""
~'
V
~ li'l.uid
\/
liquid 8 a S.
100 / 100
/ V
90 I Ie 90
80
II II 80 f
it
b
703.5 4.0 4.5 5.0 5.5 6.0 6.5
703.5 4.0 4,5 5.0 5.5 6,0 6.5 7.0 7.0
E NT ROPY-S- Btu/(ibOR) ENTROPY-S-Btu/(ib OR)
Figure 3. Low temperature Figure 4. Low temperature

gas liquefaction system gas liquefaction system
-21-
lower temperature limit. However, the extra power associated
with the use of the Brayton cycle for the higher temperature
part of the cycle is small compared to the overall plant power
requirements. In such a case the designer may choose to
omit the vapor cycle refrigeration for the sake of equipment
simplicity.
The cycle to be considered (see Figures land 2) is one

having a compressed gas expand adiabatically through some form of
turbo or reciprocating engine, path ceo The discharge of the
expander cools down the process stream in a process heat
exchanger. The working fluid then enters the refrigerant heat
exchanger where it cools down the high pressure feed going to
the expander, being itself heated to T. From the heat exchan~
the low pressure refrigerant stream g'tes to a compressor which
is generally multi-stage with intercooling and after cooling. One
or more coolers are necessary since they constitute the heat sink
for the cycle. From the aftercooler the high pressure refrigerant
at temperature Tb enters the refrigerant heat exchanger where it
is cooled down to the expander inlet temperature, T. The process
stream is cooled from temperature Th to T g' Liqu~faction of
the gas will be considered as a separate phase of the analysis.
The efficiency of this cooling down of the process gas without
change of phase will now be considered as compared with an ideal
refrigeration cycle.
Ideal Cycle Performance
Based on pure thermodynamic considerations the

coefficient of performance of a refrigeration system is given by
Heat Abstracted from Process

(Ideal) Coefficient of Performance =- - -____________
Net Power to Process
( 1)
C. O. P. • (la)
where T is the mean temperature of heat rejection from the

system a~d Tp is the mean temperature of heat absorption from
the process.
If one neglects the mechanical work available from the

expander, the relationship between the work supplied to the
-22-
compressors and the heat abstracted from the process,
assuming an ideal refrigeration cycle having no irreversibilities,
equation (1) reduces to
(Ideal) Coefficient of Performance = Power Supplied to Compressors
( 1b)
C.O.P.= Tp (Ic)
To
where T P is the mean temperature of heat removal from the

process.
Assuming an ideal gas as process fluid, the mean is

given by the log mean temperature,
(2)
In the general case,
T
P
= (3 )
and To is the temperature at which heat is rejected from

compressor coolers operating isothermally.
Actual Cycle Performance
The performance of an actual cycle can be expressed

by multiplying the ideal cycle performance by a number of
efficiency factors to represent the departure of the actual
cycle conditions from ideal conditions. The final expres sion
for coefficient of performance can then be written as
(4)
-23-
Efficiency Terms
The several efficiency terms of equation (4) are

defined in the following manner.
Eco = adiabatic efficiency of compressor stage s

between intercoolers, provided the value of
To is properly specified. Where several
intercoolers are used, as indicated on (5)
Figure 2, the value of To will be very close
to the log mean of the temperature of the gas
leaving the intercooler and the discharge
temperature from the compressor. The first
stage of compression starts at a slightly lower
temperature, but this can be neglected.
E ex = adiabatic efficiency of expander (or stage of

expandeI), (6)
Eex = (A Hadiabatid / (AHisentropic) (6a)
and for the conditions of Figure I
(6b)
= ratio of mean temperature associated with

isentropic expansion to mean temperature (7)
associated with actual adiabatic expansion.
Ead = [(T c - T d ) / In (Tc/Td)] / [(T c - Te)/ln(Tc/Te)]

(7a)
= efficiency of heat exchanger as related to

the net refrigeration available from the (8)
expander. Following the notation on the
process flow path shown on Figures I and 2.
-24-
Ehe = (8a)
E • T / T is the ratio of the mean temperature at which

pr r p
refrigeration is supplied to the mean temperature required
for cooling down the process fluid. For the case of an (9)
ideal gas as refrigerant.
(9a)
T p is given by equations (2) or (3). Where

refrigeration is carried out in stages. T p is calculated for
the appropriate stage.
efficiency of the cycle as affected by the pressure

losses in the piping. heat exchangers. inter- (10)
coolers and aftercoolers. The pressure drops are
too small to be indicated meaningfully on the T-S
diagrams.
(lOa)
where ~ (~p/p) represents the summation of all pressure

drop ratios in the cycle.
Heat Infiltration + Heat Abstracted from Process

(11)
-25-
The significance of some of these efficiency terms is
almost self-evident. Compressor efficiency is usually dictated
by choice of type of compressor, the volume of gas flow,
specific speed, and similar considerations which permit only
limited choice. Obviously, the highest efficiency consist ent
with capital costs is sought. By limiting the tempe rature
rise per stage, one can hold Toto the minimum practical
temperature level and so keep down compressor power, as
indicated in equation (1).
Likewise, the highest expander efficiency is desired.

Typical values of Eex have been given by Jekat 2 . In addition
to its own value in setting cycle efficiency, E affects the
ex
value of Ed' The value of Ead in turn depends on another
parameter ~esides the expander efficiency, namely the
temperature (or pressure) ratio across the expander. Expanders
having a low temperature ratio will, for a given expander
efficiency, have a higher value of Ead than a high temperature
ratio expander. However, other efficiency terms, such as
Ehe and E dp are adversely affected by a low temperature (or
pressure) ratio.
The pressure drop term Ed includes the effect of

losses due to resistance in ducts ang valves and due to friction
in heat exchangers. The optimum resistance allowed for in
ducts and valves depends largely on a balance between increased
capital investment for large size equipment and reduced operati~
cost for compression work. The pressure los s in heat
exchangers, however, is related to heat transfer performance.
An increase in pressure loss might actually reduce the plant
power requirement by a reduction in refrigeration losses asso-
ciated with end temperature differences in the heat exchangers.
Guides to designing heat exchangers with regards to heat
transfer effectiveness and optimum pressure losses are given
by Aronson 3 and Kays and London 4 •
2W. K. Jekat, Proceedings of the 1956 Cryogenic

1956.
3D • Aronson, Heat Transfer and Fluid Mechanics
Institute, Stanford University Press, Palo Alto, California (1954).
4W. M. Kays and A. L. London, The National Press,

Palo Alto, California (1955).
-26-
Process Efficiency
The process efficiency term. E • is one where

considerable variation is to be expected. pr For supplying
refrigeration over a wide range. a high value of Epr calls
for either a high pressure ratio with one expander. or lesser
pressure ratios. but several expanders. each operating over
a part of the total temperature range. as indicated in Figures
4. 5 and 6.
Before considering the multi-expander cycle. we shall

calculate the proces s efficiency of a single' expander cycle
handling a refrigeration load comprising cooling down a gas
and then liquefying it. as indicated in Figure 3. The values
of the process efficiencies for the conditions given are.
154-92 1 120.2
(liquefaction)E pr = (In (154/92») ( 150.0) = 150.0 = 0.802
(12)
This compares with the proces s efficiency for the cooling

down process as indicated in Figure 1 of.
"
( coo I lng d own)E =(222 - 135) (In(500/150) ) = 174.6 =0.600
pr In (222/135) 500 - 150 291
( 13)
The cooling down process has a much lower process efficiency

than the liquefaction process because the refrigeration is
supplied at a much lower temperature than required. This
adverse condition is often alleviated by cooling down the gas
from Th to some intermediate temperature by means of a
supplementary refrigeration cycle. generally of the vapor
compression and condensation (Rankine) cycle as indicated by
Gifford's first principle. However. if we are willing to accept
the use of two expanders. the overall performance can be
improved by operating both expanders at higher temperature
levels. This is possible if we allow the expander supplying
refrigeration for liquefaction to supply also refrigeration for
part of the cooling down process. The actual heat loads per
pound of fluid can be calculated from the plot on Figure 3.
-27-
ENTROPY,S ,Btu 1(lbOR) for PROCESS GAS
1000 -100 102.030 405060 7.08090100 ILO 12013014015016017.0180 190
800
600 ...'0 -~.
'<11
IiV"'
J'
17
I~
A
400 c;. ..J /

ION E",,\ ~~ /
200
'1
/
J V"8
Up
b L L
I
I-t-
~
J
B
TEM ~ERAT ~I :E, -<- iTw
°Ran '\ine" ~ 6 p.~ V;
100
V~ ,KV- ~. I
I
I ~(
~RO :ES S c: AS
80 IT ID
I PJ l'l ....
1\ V
60
IV
"
'Ii ...t I I /
.,~ VJ ",-0
40
F06'R~ ~~ I I IF E
~.
I I
20
2.0
II r.
3.0
I I
4.0 5.0
I6.0I 7.0 8.0
ENTROPY, S,Btu/(ibOR) foy REFRIGERANT GAS
Figure 5, Multiple expander cycle

Performance of low temperature
gas liquefaction plant
b
BATTERY OF COMPRESSORS, Psu.
INTERCOOLERS.
AND AFTERCOOLERS
~---~~~~~~~~~---~ Pdi
To
u. h c a
TWO ONE
J d b
(f)
0::: W
""00:
z::» THREE
<0
Q.,,-
)(
e
"" w v
F COOLER
A
Figure 6, Multiple expander cycle

Performance of low temperature
gas liquefaction plant
-28-
As shown here, the gas has a constant specific heat over the
range of temperature ~rom 500 0 to 150 0 R, and liquefaction
occur s at constant temperature of 1500R.
Heat to cool down = S TdS = Tp.6.S = (291) (6.75 - 5.75)
= 291 BTU/lb ( 14)
Heat to liquefy =Tp.6.S = (150)(5.75 - 3.75) = 300 BTU/lb

( 15)
It should be noted that the value of T p for the cooling

down process is the log mean temperature, as given by
equation (2).
From these heat values one can calculate the average

value of the process efficiency, Epr as follows,
65hg(ideal) + .6. Sgj(ideal) 1.00+2.00

Average E
pr = ~hg(ideal) .6.Sgj (ideal)
= 1. 00 2.00
+ ---t---
Epr Epr 0.6000.802
= O. 722 \16)
Equation (16) indicates a method of averaging efficiency terms

on the basis of the ideal entropy change associated with each
phase of the proces s.
A 100 per cent process efficiency for liquefaction

requires an isothermal expander. This can be approximated by
a multi-stage expander with reheat between stages. In practical
installations, liquefaction is effected by the expansion at
constant enthalpy of a non-ideal gas or by comparable expan-
sion of saturated liquids at one or more pressure levels. By
condensing the high pressure liquid at different pressures (hence
over a range of temperatures) the overall process efficiency
can be made fairly high.
-29-
Multiple Expanders
Comparable to the approach to 100 per cent process

efficiency for liquefaction by multi-stage expansion with reheat,
is the use of a series of expanders for handling the cooling
down part of the refrigeration load. The chief limit on number
of expanders is the increase in capital investment and the
increased maintenance. Of course, one should never use
more expanders than are required to cover the temperature
range, otherwise efficiency will actually be lowered. While
maintenance may be increased by use of multiple expanders,
plant reliability may be enhanced because part load operation
is usually possible with a reduced number of expanders. For
both normal and emergency operation, an approach to equal
process efficiency for each expander stage should be sought.
Figure 4 illustrates how the use of a logarithmic temperature
plot facilitates laying out a cycle for equal proces s efficiency
in each expansion stage. A similar geometrical pattern of
temperature lines gives the desired result.
Heat Exchanger Efficiency
The heat exchanger efficiency, E he , must be carefully

distinguished from the conventional usage of the term which
relates the actual heat transferred to that possible with an
exchanger having infinite heat transfer surface. The efficiency
term proposed in this paper related the refrigeration made
available to the process to that which would be available with
an ideal heat exchanger. The refrigeration comes from the
enthalpy drop in the expander. Hence at low absolute tempera-
tures, where the enthalpy drop is comparatively small, the end
temperature difference in the heat exchanger must be kept
small (for a given heat exchanger efficiency). Gifford l recom-
mends that the end temperature difference be proportional to
the square root of the absolute temperature. A method of
achieving this variation in end temperature is given in Figures
5 and 6. A small bypass stream is cooled down by the
refrigeration process in a manner similar to the process stream.
This unbalances the mass flow in the refrigerant heat exchangers
and permits a small temperature difference at the cold end and
a larger difference at the warm end. The heat transfer size of
the refrigerant gas heat exchangers is usually many times that
of the bypass or process gas exchangers. This heat transfer
size is given by N tu= (UA/WC p )' For the refrigerant gas
-30-
heat exchanger s with almost equal mas s flow of the two
streams,
UA (Temperature Change of One Fluid) (17)

WC p = (Overall Mean Temperature Difference)
The last three references 4,5,6 offer simplified methods of

calculating the design of these gas to gas heat exchangers.
The original version of this paper (the foregoing being a
condensed version) contains a more detailed study of the
conditions existing in the heat exchangers, and considers
an approach to optimum design.
The ideas presented here are just a few of the many

possible approaches to the analysis of the performance of
components of very low temperature refrigeration systems.
For cycles not based primarily on Cle use of expanders,
other methods of analysis should prove more suitable.
The author is grateful to his associates at Worthington

Corporation and others outside the company who have helped
in various ways in the preparation and presentation of this
paper.
4W. M. Kays and A. L. London, The National Press,

Palo Alto, California (1955).
5 C • Simpelaar and D. Aronson, Trans. A.S.M.E.,

Vol. 72, 955 (1950).
6
D. Aronson, Trans. A.S.M.E., Vol. 72, 675 (1952).
-31-
A-4
POWER CONSUMPTION AND THERMODYNAMIC REVERSIBILITY

IN LOW TEMPERATURE REFRIGERATION AND
SEPARATION PROCESSES
J. B. Gardner and K. C. Smith

British Oxygen Research and
Development Limited
London, England
Introduction
The design of low temperature refrigeration and

distillation plants, in particular those associated with the air
separation industry, is relatively well developed and improve-
ments in processes or techniques are not now so obvious.
A most useful approach, from the point of view of

establishing the scope and incentive for further improvements,
is by a thermodynamic analysis in which the irreversible
entropy increments occurring in each section or component of
a particular process or cycle are determined. The product
of these increments and the ambient temperature gives a
quantitative measure of the penalty in energy which is being
paid for thermodynamic inefficiency.
A similar analysis for an ideal cycle, essentially

the same as the actual cycle but having theoretically perfect
components, is made. This provides a basis for comparison
and indicates what further improvements in a particular cycle
may theoretically be possible, what the incentive is for such
improvements, and how far improvements are limited by the
type of cycle used. The thermodynamic inefficiency, expressed
as an equivalent power consumption, in each plant component
is simultaneously also obtained. This can serve as a con-
venient basis for determining the most economic size.
The use of this method is not necessarily always the

most convenient for determining the optimum size of compo-
nents, but in certain cases it can save a considerable amount
of trial and error calculation. Moreover, the thermodynamic
analysis does enable a true appreciation of the relative
importance of the various inefficiencies to be made.
-32-
Method of Analysis
The basic thermodynamic data for air used is that

recently derived by Din. 1 For oxygen and nitrogen the
pressure-enthalpy charts (with isentropes indicated) prepared
at Leiden and issued in 1940 have been used.
In analysing a particular cycle, the following steps

are taken:
1) Calculation of the reversible thermodynamic work

of separation and/ or liquefaction.
2) Calculation of the entropy increments occurring at

each step for the actual cycle under consideration.
3) Calculation of the operating pressure, heat and

mass balance for an ideal cycle, which is the
same in character as the actual cycle but in which
theoretically perfect equipment is as sumed, i. e. ,
isothermal air compression, zero temperature
difference at one point in every heat exchanger,
infinite number of trays in the distillation column,
etc. Then for this ideal cycle. determination of
the entropy increments still occurring at each step
in the process.
The basic equations used are well known,
WR = AH - ToASR (1)
W = W R + :I;ToASIRR (2)
where
reversible work of separation
and/ or liquefaction
W = actual work
IF. Din, Thermodynamic Functions of Gases. Butter-

worth Scientific Publications, London (1956).
-33-
60H = difference in enthalpy between
final and initial states
60S = difference in entropy between

R final and initial states for
reversible process
= irreversible entropy increment
in any particular step
o
T , K - heat sink into which energy is
o
pumped, usually the atmosphere
From such an analysis, three types of efficiency can 2

be distinguished, as has been previously pointed out by Dodge.
Thermodynamic efficiency ( TJT ) =--------

reversible work
actual work
reversible work
Process efficiency (TJp) =
work for "ideal"
cycle
work for "ideal"
cycle
Practical efficiency (TJA) =--
TJT
= actual work
TJP
TJT represents the fundamental efficiency without any

regard to the process or cycle, TJp is an indication of the
best possible performance for a given cycle, and TJA is a
measure of how near the actual process approaches this best
performance.
These efficiencies are important in that if, for exam-

ple, TJP is low, it may well be better to attempt to devise a
more efficient cycle than to strive for the ultimate in
component efficiency.
2 B • F. Dodge and H. Bliss, Chem. Eng. Prog. 45,

51, (1949).
-34-
A number of air separation cycles have been analysed
in the above manner and it is proposed in the next section, as
an example, to consider the Heylandt cycle for liquid oxygen
production.
Thermodynamic Analysis of Heylandt Cycle
The basic cycle is sh~wn in Figure 1 and is well known.

It need not be described in detail here.
Typical performance data, and the calculated perform-

ance of an ideal cycle producing the same products, are
summarized in Table 1.
TABLE I
Ideal Actual
Cycle Cycle
Operating pressure (atm) 51 150

Liquid oxygen yield (gros s plant) 20.2 19.04
mols per 100 mols air (at 1 atm (at 1. 4 atm
saturated) subcooled)
Liquid oxygen purity in per 100 99.6

cent
Oxygen content of waste 1 2.5

nitrogen in per cent
The ideal cycle assumes air compressors of 100 per

cent isothermal efficiency, 100 per cent isentropic expansion
in the engine, infinite heat exchangers, an infinite number of
trays in the rectification columns, and zero pressure drop
through heat exchangers, etc. The figures used for the yield
from the ideal column may be somewhat arbitrary and depend
on the vapor-liquid equilibria data used in the column calcula-
tions, but small changes in the yield will not significantly
affect the subsequent comparisons.
-35-
1------
AIR
~lillWLH~+L. o.
Figure 1. Basic flow diagram of Heylandt cycle
---- ......
o·s 1·0 1·5 2·0 25 3·0 3·5 4.0 4·5 5-0

TEMP NSER-EVAPORATOR ~
Figure 2. Variation with size of annual capital

plus power charges for main condenser-
evaporator of a large liquid oxygen plant
-36-
In the following table, the results of thermodynamic
analyses of the ideal and actual cycles are tabulated.
TABLE II
Breakdown of calculated energy

requirements expressed as kw.hr./
1000 cu. ft. of liquid oxygen*
Ideal Actual
Cycle Cycle (A - I )
(I ) (A)
Reversible work of
separation and lique-
faction 8.00 8.09 + o. 09
Irreversibilities due to:
Heat exchange A T' s 0.82 0.98 +0.16

Main expansion valve 0.09 0.95 +0.86
Expansion engine nil 1. 96 +1. 96
Rectification column
system (including
condenser evaporator
and liquid nitrogen sub-
cooler) 2.65 4.21 +1. 56
Compressor and motor nil 17.23 +17.23
Heat leak nil 0.74 +0.74
Other s (mixing of fluid

streams, air losses,
etc. ) 0.02 1. 22 +1.20
3.58 27.29 +23.71
Total 11.58 35.38 +23.80
~cIn this paper liquid oxygen is expres sed as gas measured

at 15.6°C and 1 atmosphere absolute.
-37-
From a study of Table II, it is apparent that,
1) 1') T is approximately 23 per cent
1')p is approximately 69 per cent
1') A is approximately 33 per cent
2) For an ideal cycle, 74 per cent of the

irreversibility is in the rectification
column system.
3) For the actual cycle, 15.4 per cent of the

total irreversibility is in the rectification
column system.
4) Considering the third column, (A-I), which

shows the differences between the correspond-
ing inefficiencies suffered in the actual and
ideal cycles, by far the greatest, as might
be expected, is in the compression system.
5) There would on first inspection appear to be

relatively little incentive for attempting to
improve the efficiency of the plant items
other than the air compression system,
since, for example, even if a rectification
column were used having an infinite number
of trays, zero pressure drop and zero
temperature difference across the condenser-
evaporator, the savings would, on the basis
used, appear to be only of the order of 1.6
kw. hr. / 1000 cu. ft. of liquid oxygen product.
One can, however, consider the figures in a different

way by treating the cycle proper as a separate system, distinct
from the air compression side, so that any inefficiencies occur-
ring in the cycle are accentuated by the compression inefficiency
In this case, if it is assumed that compressor efficiencies are
not likely to markedly increase, the incentive for improvements
in the efficiencies of the cycle components is greater. This is
illustrated in Table III where the items in the cycle proper only
are considered and the theoretical power equivalent of the
inefficiencies due to thermodynamic irreversibility are divided
by the effiCiency of the air compression step.
-38-
TABLE III
Inefficiencies due to irreversibilities

in cycle proper, assuming compressor
efficiency of about 72 per cent
adiabatic and motor efficiency of 95
per cent approximately, kw. hr /1000
cu. ft. liquid oxygen
Ideal Actual A - I
Cyc1e Cycle
(I) (A)
Reversible work of
separation and lique-
faction 15.60 15.77 +0.17
Irreversibilities . due to:
Heat exchange AT' s 1. 60 1. 91 +0.31

Main expansion valve 0.18 1. 85 +1. 67
Expansion engine nil 3.82 +3.82
Rectification column
system 5.16 8.21 +3.05
Heat leak nil 1. 44 +1. 44
Other s (mixing of
fluid streams, etc.) 0.04 2.38 +2.34
6.98 19.61 12.63
Total 22.58 35.38 12.80
For example, the total theoretical possible saving

by improvements to the rectification system is then approxi-
mately 8.2 kw. hr. /1000 cu. ft. of liquid oxygen product,
though replacement with an ideal version of the present type
of column would still lead to a power reduction of only
3.1 kw. hr. /1000 cu. ft. liquid oxygen produced.
Note that the largest difference in irreversibility

between any individual component in the actual and ideal
cycles is in the expansion engine, although the efficiency of
-39-
the practical engine (calculated on the basis of actual
enthalpy drop divided by the enthalpy drop for isentropic
expansion) is approximately 80 per cent.
Irreversibility in Low Temperature Distillation Columns
It is of interest to consider briefly the fundamental

factors affecting the efficiency of the conventional double colwnn,
as frequently used in air distillation, in rather more detail.
Although, in effect, heat enters at the base of each column and
is removed at the top, the section in between containing the
contact trays is essentially adiabatic, except for heat influx
from the atmosphere.
As an example, itemized data are presented in Table
IV for the estimated irreversibilities incurred in the double
column system of a Heylandt plant producing liquid oxygen.
TABLE IV
Estimated irrever- Work equivalent in kw. Work equivalent,

sibilityarising hr./lOOO cu. ft. liquid taking compres-
from various sources oxygen, assuming ideal sion inefficiency
compres sion factor into
account, kw.
hr./lOOO cu.ft.
liquid oxygen
Ideal Actual Ideal Actual
Case Case B-A Case Case BI -AI
A B AI BI
Adiabatic rectification 2.22 2.22 4.33 4.33

Reflux above minimum 0.55 0.55 1. 07 1. 07
Expansion of reflux
streams from lower to
upper column 0.21 0.36 0.15 0.41 0.71 0.30
Liquid nitrogen reflux

sub.cooler 0.02 0.09 0.07 0.04 0.180.14
Condenser -reboiler
6T's 0.46 0.46 0.89 0.89
6 P across column
trays 0.53 0.53 1. 03 1. 03
Heat leak 0.54 0.54 1. 05 1. 05
Total 3.00 4.75 1.75 5.85 9.26 3.41
-40-
It will be observed that the figure for total irreversi-
bility suffered in the ideal case differs slightly from that
given in Tables II and III. This is because in Table IV
separation into the actual products has been as sumed for the
ideal unit, in order to facilitate direct comparison with the
actual rectification system, whereas previously, in comparing
the complete cycle, an ideal separation was assumed, giving
rise to a higher liquid oxygen yield.
Commenting first on the ideal case, since this

reveals the nature and extent of the intrinsic irreversible
factors present, it will be noted that much of the largest
penalty is attributable to the adiabatic nature of the rectifica-
tion process, even under minimum reflux conditions. This is
true because reversibility implies phase equilibrium at all
levels in the column and this in turn require s a varying liquid
and vapor flow in the column instead of the approximately
constant molal flow which is characteristic of the adiabatic
column. At minimum reflux, generally only at a single point
in the column, usually the feed point, where the operating
and equilibrium lines coincide, are reversible conditions
established. The continuously variable reflux ratio required
for fully reversible rectification can theoretically be achieved
by adding appropriate quantities of heat at all levels below the
feed entrance to the column and removing heat at all levels
above it, instead of merely at each end of the column as in
normal adiabatic systems. It would still be necessary to
perform all the requisite heat pumping operations in an ideal
manner in order to obtain 100 per cent rever sible rectification.
Referring again to the ideal column, there is an

additional irreversibility contribution due to reflux in exces s
of the minimum. The characteristics of the double column
being what they are, the total reflux provided is essentially
fi xed, although there is a certain limited flexibility in the
distribution of this reflux between the lower and upper
columns. The fact that the excess of reflux and, Itherefore,
the associated irreversibility is not very large, is to some
extent fortuitous. Other irreversibilities in the ideal case
are small. Of the total irreversibilities, equivalent to
3.00 kw. hr./lOOO cu. ft. liquid oxygen product, some 74 per
cent are due to the adiabatic nature of the column.
Turning now to the actual double column. all

penalties in addition to those incurred in the ideal column are
-41-
attributable to design factors, e.g., heat leak, pressure
drop, temperature difference across the condenser-evaporator.
The largest single penalty is, however, again due to the
adiabatic nature of the rectification process, the magnitude
of which should be virtually identical with that suffered in
the ideal case, as is the penalty due to greater than
minimum reflux.
There is an increase in irreversibility equivalent to

O. 15 kw. hr. /1000 cu. ft. liquid oxygen in the expansion of
fluid streams from the lower to the upper column due to the
higher expansion ratio. This is largely a secondary effect
arising from the effect of the condenser-reboiler temperature
duference, but pressure drops across the columns are also
partly responsible.
Therefore the total irreversible losses in the actual

column are estimated as equivalent to 4.75 kw. hr. /1000 cu. ft.
liquid oxygen. Intrinsic factors having been shown in the
ideal column to be responsible for 3.00 kw. hr. /1000 cu. ft.
liquid oxygen. Design factors are thus responsible for a total
of 1.75 kw. hr./1000 cu. ft. liquid oxygen. This assumes that
compression of the air is performed reversibly. If the
compression inefficiency factor is taken into account, the
total penalty due to irreversibility is increased to 9.26 kw. hr./
1000 cu. ft. liquid oxygen and the contribution from design
factors to 3.41 kw. hr. /1000 cu. ft. liquid oxygen.
As far as power consumption is concerned, the scope

and incentive for improvement in the rectification step of a
liquid oxygen plant having a total power consumption of about
36 kw. hr. / 1000 cu. ft. of liquid oxygen is therefore not
very great.
In a large plant producing gaseous products only,

however, the power consumption may be only about one third
of that of a liquid oxygen plant, but it can be shown that the
work penalty due to the adiabatic nature of the rectification
process is of the same order. The incentive for increased
reversibility in the rectification process is therefore
conSiderably greater in this case if it can be achieved without
a proportionate increase in capital charges. It may well be
that the next major improvement in the large scale separation
of air into gaseous products by a low temperature process will
result from a thermodynamically more reversible approach to
the distillation step.
-42-
Use of Thermodynamic Analysis for Optimising Plant Items
From a consideration of the irreversible entropy

increments occurring, a useful picture can be obtained of the
power penaltie s paid for thermodynamic inefficiencie s in a
particular cycle, and hence of the maximum limit to the
savings theoretically achievable by attention to design factors.
In order, however, to arrive at optimum designs, from an
economic standpoint, for individual plant items, power
charges must be balanced against capital charges.
For certain items, e. g., the compressor and motor

system, the power and cost data required to carry out
optimisation can be obtained readily without recourse to
thermodynamic evaluation. For most items within the cycle
proper, however, to obtain a relationship between component
performance and overall power consumption by standard
techniques often involves a considerable number of tedious
trial and error calculations, requiring heat and mass balances
over the complete cycle. In such cases a method based upon
the changes in the irreversible entropy increments which
accompany an alteration to a component is often useful in
enabling a rapid and sufficiently accurate appreciation to be
made of the effect of such an alteration on overall power
requirements.
Any alteration in the performance of a single com-

ponent will in many instances result in changes in the entropy
increments incurred both in the component itself and also in a
number of other plant items, and to obtain the effect on power
consumption exactly, all of these increments would need to be
considered. However, most of the secondary entropy increment
changes are usually negligible or mutually compensating. For
example, for small alterations in the rectification circuit, the
overall performance of the refrigeration section is generally
virtually constant.
It has been found that a sufficiently accurate result

can usually be otained by considering only the change in
irreversibility in the compotent itself which results from an
alteration to its design, and also in certain instances the
accompanying changes in one or two closely associated items.
The selection of the secondary entropy increments which must
be considered is to some extent a matter of experience.
Let us consider, as an example, the condense r-

evaporator in a Heylandt double column.
-43-
The effect of increasing or decreasing the surface
area available, other things being equal, will be to decrease
or increase the pressure in the lower column. This will
result in a lower or higher exhaust pres sure from the
expansion engine and also after the main expansion valve,
and assuming that the expansion engine efficiency remains
constant, the pressure of the air delivered by the compressor
which is necessary to provide the required amount of
refrigeration will vary.
The operating pressure required, and hence the

effect on power consumption, for any particular condenser-
evaporator surface area can be derived by the method of
trial and error calculations based on heat balances. In such
calculations allowance would be made for the fact that the
ratio of the amount of air passing through the engine to that
passing through the final heat exchanger will vary slightly
with operating pressure for optimum refrigeration performance.
Accurate calculations tend, therefore, to be rather tedious.
The specific effect of changes in the surface area

can also, however, be derived reasonably accurately from
an entropy increment analysis. From a consideration of
Table IV, altering the size of the condenser-evaporator and
hence the lower column pressure will, in the column system,
alter the irreversibility due to temperature difference across
the heat transfer surface, and also that due to expansion of
the reflux streams from lower to upper column pressure.
Other irreversibilities in the column system are practically
unaffected. Therefore by simply calculating To 6.SIRR for any
particular condenser area, where
To' OK = ambient temperature
6.SIRR = sum of the irreversible entropy

increments due to 6.T across the
condenser evaporator, and the
expansion of the reflux liquids
from the lower to the upper
column pressure, which are
easily derived.
The variation of the power consumption with size,

due to irreversibilities in the above items, can then be
compared with the variation in capital charges and the most
economic size estimated.
-44-
Results from an application of this technique are
shown graphically in Figure 2, where, for illustrative pur-
poses:
1) The temperature difference across the condenser-

evaporator is assumed proportional to surface
area. (This is not strictly correct since the
overall heat transfer coefficient is a function of
the temperature difference. A correction for
this will, however, be necessary whatever method
is employed and for the purpose of illustration
this correction has been ignored.)
2) As a basis for capital costs, that for a condenser-

evaporator for a 2 0 C temperature difference is
taken as $16,000, which is representative of a
unit from a fairly large plant.
3) Depreciation, maintenance and charges for

return on capital are assumed for the purpose of
illustration to total 30 per cent of the capital cost.
4) The expansion engine efficiency is assumed to be

constant over the operating pressure range
considered.
5) Power has been taken at 1.2 cents/kw. hr.

(representing United Kingdom conditions).
For the particular example and conditions considered,

the most economic size would appear to be that corresponding
to a temperature difference of approximately 1.4o C. It is of
interest to note that, for the same capital costs and charges
but assuming a power cost of O. S cents/kw. hr., which is
more typical of conditions in the United States, the most
economic temperature difference in the example cop.sidered is
shifted to 1. SoC., although over the range of 1.4-2. zoe the
variation in total charges is small.
-45-
Discussion
Question by D. Aronson, Worthington Corporation: Does

ex pander provide refrigeration over range of temperature
associated with both sensible cooling and liquefaction of
oxygen for liquid plants?
Answer by author: In the Heylandt cycle the exhaust temper-

ature of the air leaving the engine may be about _160 o c,
which is somewhat above the saturation temperature for air
at the engine exhaust pressure.
Although the combined air streams from the expander and

the second heat exchanger pass through the separation
columns, the essential effect is that the engine refrigeration
is used for sensible cooling of the incoming high pres sure
air.
-46-
A-5
THE REVERSIBLE SEPARATION OF

MULTICOMPONENT MIXTURES
H. M. Scofield
Linde Company
Division of Union Carbide Corporation
Tonawanda, New York
Introduction
The concept of reversible rectification of binary gas

mixtures is well known and has been applied often to define
the thermodynamic limits of binary separations. An extension
of this concept to multicomponent mixtures is quite possible,
and a preliminary paper on the reversible separation of
multicomponent mixtures was given at this conference last
year. 1 It is the purpose of the present paper to clarify and
extend some of the ideas presented at that time.
A reversible rectifying column is defined as one in

which liquid-vapor equilibrium exists for all components at
every level in the column. The Carnot work required to
maintain the ideal liquid flows in such a column is equivalent
to the work given by the semi-permeable membrane concept.
Equations are given which show that in any tray section when
the throughput of anyone component of a multicomponent
mixture is assigned, then the requirement of liquid-vapor
equilibrium for all components can be met by a single and
uniquely defined heat flow at every column level. This is
entirely analogous to the binary case, and no additional
principles are required. A temperature-liquid flow plot is
given for a particular reversible separation of an ethylene-
ethane -propylene mixture.
The question of finite approach to reversible separa-

tion is considered. Calculations are given which indicate that
the separation in a column with a finite number of trays, and
with heat adjustment made in a stepwise fashion on each tray,
can be made to approach as closely as desired to a reversible
separation.
1 J. F. Grunberg, Proceedings of the 1956 Cryogenic

1956.
-47-
The effect of preliminary cocurrent condensation of
the feed before introduction into the reversible column has also
been studied. Calculations are presented which show that a
feed which has been 60 per cent condensed before separation
results in larger column liquid flows and more overall heat
transfer requirements than a saturated vapor feed. This
particular phase of reversible separation requires more study.
Defining Equations
Figur e 1 shows a schema tic view of the top of a

r ectifica tion column handling q components. The three
defining equations of reversible separation are also shown.
Equation (1) is the material balance equation for component n;
equation (2) is the equilibrium equation for the nth component;
and equation (3) expresses the unit-sum nature of the vapor
mol fractions. Equations (1) and (2) must, of course, be
written for each of the q components of the mixture, and the
three equations together impose (2 q + 1) independent conditions
on the relevant variables. The variables in general are q
values of y, q values of x, the temperature implied in the
K values, and the reflux ratio L/V, giving a total of (2 q + 2)
variables. The selection of anyone of these variables is
sufficient to define the remaining variables by algebraic
solution of the (2 q + 1) independent equations. In computing a
particular reversible separation, the equations are most
conveniently applied in two steps. By applying the equations
first at the feed point where the q values of the feed composi-
tions (yF) are known, we find that selection of the throughput
of a single component (y W) serves to define the throughput of
all other components. Wffh throughputs of each component
known, the equations can then be applied at selected tempera-
ture levels between the feed and the top to define the entire
course of the reversible rectification. A more detailed
analysis of the application of the defining equations is given
in the latter part of this paper.
An Example of Reversible Rectification
The equations shown on Figure 1 have been applied to

the reversible separation of a ternary mixture of ethylene-
ethane-propylene at 1 atmosphere total pressure. Figure 2
shows the feed composition used, the equilibrium relationships
assumed, and a sketch of the three columns necessary for a
-48-
(V -L)
Material Balance:
T
Yn VYn = LX n + (V -L) Y~
(n =1,2,3 ... , q)
Equilibrium Condition:
Yn =K
xn n
L,xn V'Yn (n • I, 2, 3 ••• , q)
Mol Fraction Condition:

q
1;
n=l n
Y =1
Figure 1. Defining equation for reversible
multi component separation.
TABLE I
Vapor Feed Separation
Liquid Liquid
Condensed Evaporated
Ternary Column 0.5684 0.0121

Top Binary Column 0.6579 0.6226
Bottom Binary Column 0.0515 0.6431
Totals 1.2778 1.2778
Mixed Feed Separation
Cocurrent Condenser 0.6000

Ternary Column 0.1188 0.0507
Top Binary Column 0.5483 0.5192
Bottom Binary Column 0.0551 0.7523
Totals 1.3222 1. 3222
-49-
/'~<!O COH.-oSI710N
SHTU",,,ZtSO l/""PM
CTHYLtSlVc eo;/. (COHPDIV.!Al7.3)

£rl1l"1A1E
r'IfOPYt.UJl ;,c".
.;10 ';I.. (COHPO.A.tEAlT 2)
(COMPON£JJT I)
MD4 .1Ib
£rHY,IfAlt! rltl'fc.
cqUIt.JSI(luM ErN/9AJ'! )(e .SO
/{ct."ZI0IV5HIP5 ?t'fDPyt.EAlc.
.~
1(3 :: 1/ 1$ "'V 0.6814
..90
I'I~: 5,., XI O·7~81 A£Vt!"'fSI8L£ CASE
.ID /TNO SeTs R;J
/03)C.· ~.S694-~
r ./~
00 .10 .20 .30 .1'4 .R) .M> .70

1'1tN. .r-M&. XI
Figure 2. Rever sible separa- Figure 4. Liquid composition

tion flow diagram. paths in ternary column
-110,---------------------,
-100
",,,,-
.'I -..,~--------=-=-~--=.-:::...-~-
...
:It I
K""" \
------------- -, ~ \
\
/
I
~ -70 \
\
/ ~ ~-------------------
,,
~--------------~'
---.:1lfru-"lfr4D v",,",,'" h4D -.sa
----MIXED ,F4EO
~--------------------~
-~~~~--~--~--~--~-~.S~--.~.~--.~7--~.8 -~0'L--~./L--~.e~--.~~~--.4~--.S~--.~6~--J.7
LlqUII:J ~QW t.,QUID how
(a..) (b)
Figure 3. Temperature-liquid flow diagrams for reversible

separation of an ethylene-ethane-propylene mixture
-50-
complete reversible separation. The ternary column
separates the feed into a binary mixture of ethylene-ethane
at the top and a binary mixture of ethane-propylene at the
bottom. The top and bottom binary columns then complete
the reversible separation of the respective binaries.
A plot of temperature versus liquid flow for the

ternary column is shown as the solid line on Figure 3a. The
corresponding plots for the two binary columns are shown as
the solid lines on Figure 3b. The liquid composition changes
in the ternary column are shown as the solid line on Figure
4 where they are plotted as xl versus x2'
Finite Approach to Rever sibility
The liquid flows and composition changes shown as

solid lines on Figures 3 and 4 are those that must result in
a column with infinite plates and with continuous heat flow
adjustment in order to bring about a reversible separation.
The question that immediately arises is whether it is possible
to approach reversible separation in a column with a finite
number of trays where the heat flow is adjusted in a stepwise
fashion on each plate. In order to answer this question,
three sets of standard plate-to-plate calculations were made
for the ternary column. In each set, the top and bottom
purities and the throughputs were taken the same as for the
rever sible case. The liquid flows, however, were adjusted
in the following three manners:
1) The liquid flow was held constant throughout the

column at a value 10 per cent greater than the
liquid flow at the feed point of the reversible
column. This is the case that most closely
approximates normal rectification practice where
liquid is boiled at the bottom of the column only
and vapor is condensed at the top.
2) The liquid flow was adjusted on each plate to be

10 per cent greater than the liquid flow that
would exist in the reversible column at the same
purity level.
3) The calculations of step 2 were repeated for 5

per cent excess liquid flow on each plate.
-51-
The liquid composition changes resulting from these
sets of calculations are shown as dashed lines on Figure 4
where the individual lines are labeled with the set number.
With constant liquid flow throughout the column, the course
of the composition changes lies well above the reversible
case at every column level. When the liquid is adjusted to
10 per cent greater than the reversible value (Step 2), the
composition changes in the top section are already equal to
those for the reversible case within the accuracy of this
plot. The composition changes in the bottom section for the
three sets show a steady approach to the reversible values.
It may be concluded that it is indeed possible to approach as
closely as desired to reversible conditions by increasing the
number of plates and the number of stepwise adjustments of
heat flow.
Effect of Feed Condition
In all of the preceding calculations, the feed to the

reversible column was considered to be a saturated vapor.
It has been stated in the literature that for this particular
ternary mixture there exists an optimum condition for which
the flow of heat to be transferred to the fluids being separated
is a minimum when the feed is initially cocurrently and
reversibly condensed to 60 per cent liquid before entering the
reversible separation column. In order to study the effect,
therefore, of preliminary co current condensation of the feed, a
complete reversible separation analysis was made for the case
of 60 per cent condensation. The condition of the feed afte r
condensation is
Per Cent Per Cent

Vapor Liquid
Ethylene 37.41 8.39

Ethane 44.07 20.62
Propylene 18.52 70.99
Flow 0.400 0.600
A plot of temperature versus liquid flow for the various

columns handling a mixed feed are shown as dashed lines on
Figures 3a and 3b. A comparison of this case with that for a
saturated vapor feed shows that the liquid flows required in
the ternary column and in the bottom binary column to
reversibly separate the mixed feed are about 20 per cent
greater than those required in the original case. The flow in
-52-
the top binary column is about 16 per cent less. This in
itself is possibly not significant, although it seems to indicate
that the mixed feed is harder to separate than the saturated
vapor feed. A check on total heat transferred to the fluids
being separated irrespective of temperature level is shown in
Table 1. (The quantities given in Table I are change in
liquid flow through the various sections as these are
essentially proportional to total heat transferred. )
Table I shows that on the basis of total heat trans-

ferred, the separation which includes preliminary cocurrent
condensation of the feed requires about 3.5 per cent more
heat transfer than the separation of a saturated vapor feed.
The existence of a minimum at any value of partial lique-
faction is doubtful, and it is the author's feeling that the
separation of a saturated vapor feed represents the most
direct separation with the minimum of heat transfer. The
effect of feed condition on the reversible separation is an
aspect of this subject which requires more study.
Conclusions
To conclude, three points have been considered:
I) The equations given in this paper define the

reversible separation path of a multicomponent
mixture when the throughput of a single component
in each tray section is assigned. This is entirely
analogous to the binary case.
2) An actual rectification with a finite number of trays

and a finite number of stepwise heat flow adjust-
ments can be made to approach arbitrarily close to
a reversible separation.
3) The reversible separation proceeds in the most

direct manner when a saturated vapor is fed to the
column. It is doubtful that any minimum in heat
flow requirements exists for a partially liquefied
feed. More study along this line is indicated
however.
-53-
Application of the Defining Equations of
Reversible Multicomponent Separation
The defining equations of reversible multi component

separation are shown on Figure 1 in the body of the paper.
In order to demonstrate the application of these equations,
consider the single column below having a saturated feed F
of q components and from which we are withdrawing a
product P, and a waste W.
F
n = I, 2, 3, ..... , q
x
F
n -t--,-
L,xn V'Y n
I P
For reversible rectification, where equilibrium condi-

tions must exist at each column level, the composition balance
above the feed is written as
W
= LXn + WYn (l)
where Yn is in equilibrium with x n ' i. e. ,
= Kn x n (2)
-54-
Realizing that W = V -L and defining L/V =R equations
(1) may be solved for R giving
W
Yn - Yn
R =
W
n = 1, 2, 3, ... q (1 a)
xn - Yn
For purposes of calculation it is more convenient to

write the system of equations (la) as
W W
Yn - Yn Yq - Yq
= n = 1, 2, 3, • . . (q-l) (lb)
W
Xq - Yq
Substituting equations (2) in (lb) and adding the mol

fraction condition we get a set of working equations,
_ yW
= n = 1, 2, 3, . . . (q-l)
n
(3)
q
where L: x = 1
n=l n
The system of equations (3) represents q conditions on

the (q + 1) variables (q values of x and the temperature implied
in the K values) when the values of yW are known. The
n
throughputs YWare determined by applying the system of
n
equations (3) to the feed point where K! and x F are known.
n
KF F _ yW KF x F _ yW
x
n n n q q q
=
_ yW
xF
n n
xF
q - yW
q
n = 1, 2, 3, • • . (q-l) (3a)
Equations (3a) impose (q -1) conditions on the q values

of throughput, so that the specification of the throughput of a
single component, say y W is sufficient to define the throughputs
q
-55-
of all other components. In general we are interested in
the complete reversible separation of a given feed so that
above the feed we would specify the elimination of the
heavy component, i. e. ,
yW
H
= 0
That no other component (say p) can be eliminated completely

may be seen from equations (3) since that would require the
condition,
at every column level. This means that the components are

identical from a rectification standpoint.
Having determined the values of yW from (3a),

. n
equatlons (3) may be solved at selected temperature levels
between the feed and the top of the column to completely
define the composition course of the reversible rectification.
The required liquid flow at each temperature level is
then obtained from anyone of equations (]a) plus the condition
W = V-L.
Applying a similar analysis below the feed point, the

following set of working equations is obtained,
= n = 1,2,3, ••• (q-l) (4)
q
where 1:
n= 1
Xn =1
Again the application of equations (4) at the feed point

serves to define the throughput of (q-l) components when the
throughput of a single component is specified (usually
xE = 0).
The values of Wand P are obtained by overall mass

balance on any given component, say n p, then =
-56-
p w
(mp - xp ) (Yp - mp)
W = F and P = F (5)
(yW _ x P ) w P
p p (y P - x p )
where m is the average composition of component p in the

p
feed. (m = yF for a saturated vapor feed.)
p p
Nomenclature
F = feed flow rate

W = waste flow rate
P = product flow rate
L = liquid flow rate at any column level
V = vapor flow rate at any column level
R = L/V = reflux ratio
y = mol fraction In vapor
x = mol fraction in liquid
K = equlibrium constant (y /x)
Subscripts and Superscripts
n = general component of a q -component mixture

q.p = some particular component
H = heaviest component
L = highest component
W = waste level of column
F = feed level of column
P = product level of column
-57-
A-6
SEPARATION OF HYDROGEN ISOTOPES BY

MULTICOMPONENT DISTILLATION
T. M. Flynn, K. D. Timmerhaus, D. H. Weitzel,

and J. W. Draper
Boulder, Colorado
Separation of liquefied gas mixtures by fractional

distillation may be accomplished in either a continuous or
batch operation. Continuous distillation is preferred for large
feed mixtures of nearly constant compositions. On the other
hand, the batch method is more suitable for separating
relatively small feed mixtures of widely varying compositions.
Steady state conditions with regard to temperature, pressure,
and composition characterize the continuous system while in
a batch system all of these parameters change continuously
with time. As an extension of earlier work l on the separation
of hydrogen-deuterium mixtures, this paper deals with the
batch separation of a mixture containing all six possible
isotopic modifications of hydrogen.
Having once chosen the type of distillation, the means

of achieving the vapor-liquid contact necessary to effect the
separation must be selected. In general, either plates or
packing may be used. If plates are used, whether they are
bubble-cap, screen, perforated, or grid plates, the vapor
flowing up the column bubbles through a pool of liquid flowing
across the plate. In a packed column liquid reflux flows
down over the surface of the packing material while the vapor
flows up through the voids.
Since the design parameters for screen and perforated

plates were presented at the last conference and the perform-
ance of bubble-cap plates are reported elsewhere 2 , this study
is principally concerned with parameters for a packed column
in order to compare it to a plate column designed on the basis
1
T. M. Flynn, D. H. Weitzel, K. D. Timmerhaus,
P.C. Vander Arend, and J. W. Draper, Proceedings of the
1956 Cryogenic Engineering Conference, p. 39, Boulder,
Colorado, September, 1956.
2E. S. Sellers and D. R. Augood, Trans. A. I. Ch. E.34,
(London), No.1 (1956).
-58-
of our earlier work. Two parameters especially important
in the design of packed towers are Height Equivalent to a
Theoretical Plate, or HETP, and holdup.
The HETP may be loosely defined as the height of

packing that will cause the same separation as a theoretical
plate. Thus the total height of packing required is the
product of the HETP and the number of theoretical plates
required for the given separation.
Holdup, caused by the wetting of the packing and the

column walls, is defined as the volume of liquid held up per
volume of packing. The effect of holdup in batch distillation
is not completely clear because a satisfactory method of
analyzing the operatiorl has not been developed. If the charge
to a batch still is small, however, the holdup may represent
a substantial fraction of the original charge which is not
available for distillation, since the operation ceases when the
reboiler runs dry.
The experimental apparatus used to investigate these

two parameters, HETP and holdup, was built around a 1-1/2
inch O. D. stainless steel column, shown in Figure 1, packed
with 15 inches of stainles s steel helices approximately 3/16
inch long. This packing material, known as Podbielniak
Heli-Pak 3013, was held in place by 40 mesh screens at both
ends of the packed zone. Boil-up in the column was provided
by an electric heater fastened to the bottom of the re boiler.
Ascending vapors were condensed to provide total reflux for
all runs by means of a finned condenser cooled by liquid
hydrogen. Insulation was achieved by the usual method of an
isothermal radiation shield at 20 0 K and high vacuum.
Charges to the column were synthesized mixtures of

deuterium in hydrogen, containing 10 to 20 mol per cent
deuterium. Samples were taken at five points along the packed
section and analyzed by means of a thermal conductivity cell
which in turn activated a continuously recording potentiometer.
These concentration values were treated in the usual

manner by the McCabe -Thiele 3 method of analysis to determine
the number of theoretical plates represented by a given length
of packing. With this information the HETP was calculated to
be approximately 5/8 inch. This value is comparable to that
of higher boiling systems.
3 W . L. McCabe and E. W. Thiele, Ind. Eng. Chern. 17,

605 (1925). -59-
To determine holdup, the charge was allowed to
escape slowly until no more liquid was present in the reboiler.
This was indicated by a thermocouple attached to the reboiler
which remained at a constant temperature as long as liquid
was present, but warmed rapidly once the reboiler became
dry. The liquid remaining in the column at this time was
assumed to correspond to the holdup adhering to the packing
material. Holdup calculated in this manner was 0.18 cc of
holdup per cc of packing with an average deviation of 3 per
cent. This value is considerably larger than that associated
with higher boiling systems.
A packed column may now be designed using these

parameters and compared under similar operating conditions
with a plate column of equal diameter. This comparison is
shown in Figure 2. Both columns are designed to handle an
average equilibrated batch feed of 100 cc of liquid containing
all six isotopic modifications of hydrogen in the proportions
shown by this figure. Since both columns contain the same
number of theoretical plates, the pos sible separation in each
case wi1l be identical.
This comparison shows, however, that the plate column

with a I inch plate spacing is nearly twice as ta1l as the
packed column and for a 1/8 inch weir has a holdup of 38
per cent of the charge as compared to 24 per cent for the
packed column. The advantages from the standpoint of both
column length and holdup seem under these conditions to be
in favor of a packed column.
Two charts shown in Figure 3 are useful in describing

the product. Both sets of curves are calculated assuming
ideality. One shows the instantaneous composition of the
distillate as a function of the fraction of the charge that has
been distilled. An integration of these curves yields the
average disti1late composition over any given time interval.
For example, the first 10 per cent of the charge yields a
distillate whose composition is essentia1ly 0.1 atom per cent T.
The last 24 per cent of the charge, which is holdup and there-
fore not available for distillation, has an average composition
of 50 atom per cent T. As should be expected, the components
are removed as product in order of decreasing volatility and
dominate the composition in proportion to their concentration in
the feed.
The second curve is the temperature profile of the

distillate, also as a function of the fraction of the charge
removed. The plateaus exhibited are typical of batch
-60-
HEIGHT HOLDUP
CHARGE PLATE PLATE
FINNED
CONOENSER
1- l"
PACKED PACKED
43" 38ee
A.B.C.D.E.
SAMPLE TAPS
30 THEORETICAL PLATES
3/4' DIAMETER
100ee LIQUID FEED
THERMO' ~
COUPLE~ -- --REBOILER
~:"'::::=::i.::::t
HEATER
1.2% HZ
PACKED COLUMN 2.7 "I. HT
AVERAGE FEED MIXTURE

Figure 1.
Figure 2.
-
25
1 1 _I T I.
J
--
AVG. COMP. LESS THAN 0.1% atomic T
24 I - - -
V
2
'~
HOLDUP
AVG. COMP.
50% atomic T
22
2 II---
2 0'---
'"
!;( 1.0 H
oJ _
HD ..,..............11....... ....... DT"/~
I " Tz!.!
-J
oJ , Z
!;i .........
"" •••••
............ Dz --+--
is
...o
\
0.8Il---+~\.--Ih/...=-+--+--t--i---t...:.,.........
\ 1/
ZZ
00
Et
0.6~-I--\lri·\--I--+--+--I---+-\T\I--t--H
~: )
~ ; O.4Il----l--...-j+----t--+---If---+--t--j+t.\--+--
(112 i / \.02. •
~ ~~~~-+--+--+--+-_+r_~~-'
~; 0.2Ji~~z~~··~~····~·E"-..~tHT~~~~~:j;:D:!.J-'lv~~~\~;.
~" .~ !tiD \.~~~j
..... !.g.••••
~
(II
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Z
FRACTION OF CHARGE DISTILLED
Figure 3.
-61-
distillation, and correspond to intervals of slowly changing
distillate composition. These plateaus may be considered
as individual products. Thus a temperature sensing element
may be used to indicate when a new product is coming off.
More work on this ba!ch distillation problem is anti-

cipated in the future. Our current work is directed toward
the construction and operation of a continuous distillation
column of pilot plant size. This column will increase the
concentration of HD in normal hydrogen from its natural
abundance of I part in 3500 to 3 per cent. It is 6 inches
in diameter and contains 30 plates made of 40 mesh screen.
Plate spacing is 4 inches. The column will handle 18.5
SCFM of natural hydrogen which is equivalent to a heavy
water production of about 60 pounds per year. The various
components of the pilot plant are currently being assembled,
and it is anticipated that experiments will begin early next
year.
Discussion
Question by J. F. Grunberg, L' Air Liquide: Our company is

engaged in the separation of hydrogen isotopes by cryogenic
distillation on a fairly large scale. Am I to understand that
for the separation of the isotopes of hydrogen, small packed
columns are used to advantage by NBS? Would this still be
true for large separation units?
Answer by author: No generalization can be made accurately

concerning which is better, plate columns or packed columns.
Each case must be evaluated on its own merits. The example
presented in this paper does show that for very small
throughputs packed columns may be preferred. It is doubtful
that this would be true for large throughputs since HETP is
known to increase with diameter and because channeling is
likely to occur.
Comment by R. Katzen, Air Products, Inc.: HETP for

packed columns is affected by the size of the column, as well
as by the size and type of packing. It generally increases
with increased size of the column. In an example of rectifi-
cation, a linch HETP in a 1 inch diameter column became a
2 to 3 foot HETP in a 6 foot diameter column.
-62-
Comment by author: This effect, increase in HETP with
increase in diameter, is fairly well known. HETP is, after
all, an empirical value and should not be expected to hold
for systems radically different from those for which it was
measured.
Question by J. T. Selldorf, A. D. Little, Inc.: Would you

care to estimate a plate efficiency for the separation described
in the paper?
Answer by author: The design parameters for plate columns

were presented by us at the last conference and have been
published in both the Proceedings of the 1956 Cryogenic
Engineering Conference and in NBS Report 5074. However,
to answer your question, plate efficiency is a function of vapor
velocity in the column and consequently no general answer
can be made except that 65-70 per cent is a safe design
figure.
Question by D. Wang, Linde Company: Is there any reason

for believing effectiveness of mass exchange of multicomponent
mixtures to be different from that of binary mixtures?
Answer by author: With regards to HETP, there is some

reason to believe that the value may be different for multi-
component mixtures from that of binary mixtures. This
follows since HETP is purely an empirical value with no
theoretical basis and should be measured for each system.
However, it is quite reasonable to expect the HETP to vary
little between similar systems. Thus a HETP measured for
a H2 - D2 system should be reasonably valid for a
H2 - HD - D2 system.
-63-
B-1
EFFICIENT UTILIZATION OF ORTHO-PARA CATALYST
C. R. Class, R. P. Spero, and G. E. McIntosh

Boulder, Colorado
Introduction
Several methods are available to increase the efficiency

and lower the cost of hydrogen liquefaction. Recent investi-
gations aimed at more efficient hydrogen liquefaction have
concentrated on the application of expansion engines, particu-
larly turbo expanders, more efficient purification, and
development of and effective use of ortho-para catalysts to
improve liquefaction cycle efficiency. The purpose of this
paper is to show the relative effects of these liquefaction
parameters, with special emphasis on the effect of ortho-para
converter applications. This will be done by discussing the
results of theoretical calculations for two simple cycles which
indicate the effects of cycle modifications, expansion engine
efficiency, and ortho-para catalyst location. For this study,
purification was not treated as a cycle parameter.
Ortho-Para Catalyst Considerations
Emphasizing the function of ortho-para catalyst in

hydrogen liquefaction cycles invites a technical catalysis
discus sion which is beyond the scope of this paper due to the
complexity of the process. For this, the interested reader
is referred to the literature and to other papers found in
these Proceedings. However, it is appropriate to make two
statements regarding the application of ortho-para catalyst
to hydrogen liquefaction cycle s. Fir st, for thermodynamic
efficiency, a source of refrigeration is required at the point
of catalysis. To avoid pumping heat from unnecessarily low
temperatures the source of refrigeration must originate at
temperatures only slightly below the catalyzing temperature.
Second, with respect to economy and design, gas phase
catalysis requires a greater amount of catalyst and more
elaborate heat exchanger design, even though it offers
theoretical performance advantage s.
-64-
Cycle Calculation Criteria
The conditions governing the liquefaction cycle calcula-

tions are listed in Table 1. Examination of these conditions
indicates that the results obtained would apply to an actual
liquefier designed and built with current techniques and
equipment.
TABLE I
Minimum heat exchanger temperature

approach
Precooler bath temperature approach
Temperature of gas leaving lower
pre cooler
Gas phase ortho-para approach 95 per cent of equi-

librium at catalysis
temperature
Para content of product liquid Equilibrium (99.8 per
cent para)
Expansion Engine Efficiency Per cent of isentropic
Hydrogen compressor efficiency 75 per cent of isen-
tropic, defined as
W = Wisentropic
.75
Work of nitrogen liquefaction 2.2 kw hr I gal of
saturated liquid at
one atmosphere
Hydrogen Liquefaction Cycles with Three Expanders
Four variations of a three expander hydrogen lique-

faction cycle are illustrated in Figures 1 and 2. Figure 1
indicates the basic and modified forms of the cycle with
ortho-para conversion in the liquid hydrogen receiver.
Figure 2 shows the schematic for the basic and modified
cycles with both gas and liquid phase ortho-para conversion.
The significant features of each version of the three expander
cycle are as follows:
-65-
HYDROGEN LlOUEFACTION CYCLES WITH THREE EXPANDERS
PART A
EXPANDER
I
«
EXPANDER
BASIC
Figure 1.
HYDROGEN LIQUEFACTION CYCLES WITH THREE

EXPANDERS - PART B
EXPANDER
EXPANDER
EXPANDER
EXPANDER
EXPANOER
EXPANDER
BASIC WITH
CONTI~ERSION
Figure 2.
THEORETICAL PERFORMANCE OF CYCLES

WITH THREE EXPANDERS
"f--+--+--+--+---1
(XPA-OER EFFICIENCY EXPAHD£R EFFICIENCY
Figure 3.
-66-
1) Basic Cycle: The basic three expander cycle is a
moderately efficient cycle which offers the advantage
of low gross work at low pressure (calculated
results shown in Figure 3 are for 30 atmospheres
working pressure) and no external refrigerant
requirement. Possible disadvantages of this cycle
are the complexity involved in the use of three
expansion engines, and its relatively low yield
which results in large flows and equipment. The
functioning of the basic cycle is conventional, with
each expander supplying refrigeration required to
balance the heat exchanger which it brackets.
Final cooling is effected in a Joule-Thomson heat
exchanger, and liquefaction is by Joule-Thomson
expansion. A dual Joule-Thomson valve arrange-
ment is employed to prevent recirculation of
partially converted gas.
2) Modified Three Expander Cycle: The modified

three expander cycle differs from the basic cycle by
the addition of a heat exchanger between the
compressor and the first expander heat exchanger.
This addition gives flexibility in applying refrigera-
tion from the expanders and results in better
performance. Functioning of the modified cycle is
the same as that of the basic cycle in all re spects.
3) Basic Cycle with Gas Phase Conversion: As shown

in the left half of Figure 2, the basic cycle can
incorporate continuous gas phase conversion which
is accomplished in the two lower expander heat
exchanger s only, since conver sion in the upper
heat exchanger would be of little value due to its
temperature level. It will be noted that the high
pressure stream is split at the beginning of gas
phase conversion. The purpose of this split is to
conserve refrigeration by converting only the gas
which is to be liquefied. Liquid phase conversion
is retained to bring the para content of the delivered
liquid to approximately the equilibrium value at
20.4 o K.
4) Modified Cycle with Gas Phase Conversion: The

right half of Figure 2 illustrates the use of gas
phase conversion in the two lower expander heat
exchanger s of the modified cycle. Comment applying
to the basic cycle with gas phase conversion applies
equally to the modified cycle with continuous conversion.
-67 -
Performance of the four cycle variations is illus-
trated in Figure 3. The left half of Figure 3 relates the
yield of the cycle as a function of expander efficiency and
the right half shows the calculated gros s work of the cycle,
also as a function of expander efficiency. Analysis of
these calculated curves reveals two significant features. First,
on both plots, the curves for the basic cycle with gas phase
conversion cross the curves for the modified cycle without
gas phase conversion. This indicates that for the same
expander efficiency the effect of gas phase conversion is of
the same magnitude as modifying the cycle. Second, from
both plots it is evident that gas phase conversion materially
improves the performance of both the basic and modified
cycles. For the two cycle variations, continuous conversion
is equivalent to an increase in expander efficiency of 6 to
10 per cent.
Hydrogen Liquefaction Cycles with One Expander
Figures 4 and 5 illustrate four variations of a cycle

which uses nitrogen precooling and one expansion engine. The
basic and modified versions of this cycle with liquid phase and
liquid nitrogen temperature gas phase ortho-para conversion
are shown in Figure 4. Figure 5 indicates the cycle arrange-
ments when continuous gas phase conversion is added to the
expander heat exchanger. The features of each cycle
variation are described in the following paragraphs.
1) Basic Single Expansion Engine Cycle: This is an

efficient cycle which offers high yield with low
gross work requirements and low pressure opera-
tion. (The operating pressure assumed for the
calculated results shown in Figure 6 is 25
atmospheres.) The principal disadvantage of the
cycle is that it requires liquid nitrogen. This
disadvantage is somewhat balanced by the fact that
only one expander is required; thus retaining much
of the simplicity of the classical Linde cycle. The
relationship to modern versions of the Linde cycle
is quite close, the only difference being the
addition of the expander and its heat exchanger.
Split precoolers to reduce nitrogen vacuum pump
work, gas phase conversion at liquid nitrogen
temperature, and dual Joule-Thomson expansion
valves have all been incorporated into present day
Linde hydrogen liquefiers.
-68 -
HYDROGEN LlOUEFACTION CYCLES WITH ONE EXPANDER
EXPANDER
Figure 4.
EXPANDER
BASIC WITH
CONT~ERSION
~
CONTINUOUS CONVERSION
Figure 5.
~ OF LlaUID NITROGEN
I
DER CYCLES
..
YIELD VS !u WORK VS EXPANOER
'4- I ~~
/IilOOlFIEO-
CONTINUOUS
" / II
.~
~ 21
/
--
/MODIFIED
S ~slc
.
.0
c~~
-- ~ 4.5
/ /e&~~~~U s ~
"
'0
--
/ ~
" / / /
• 4~
--:
'0
".
//USIC
~
...
~ CONTINUOUS
MOOIFI
f
/'
"
" . .. '.0 I
EXPANOE ..
70
" 00
EffiCIENCY
os 00 O•
tKPANDER
10
"
EFFICIENCY
.0 O.
Figure 6.
-69-
2) Modified Single Expansion Engine Cycle: This
cycle 1S obtained by the addition of a heat exchanger
between the low temperature nitrogen precooler and
the expander heat exchanger. The modification
allows the expander heat exchanger to "float",
permitting application of auxiliary refrigeration from
the expander at the most efficient point. Functioning
of the modified cycle is the same as that of the basic
cycle.
3) Basic Single Expander Cycle with Gas Phase

Conversion: The basic cycle with gas phase conver-
sion in the expander heat exchanger is illustrated in
the left half of Figure 5. As discussed previously,
gas phlse conversion is applied only to the heat
exchanger bracketed by the expander which provides
refrigeration, and the high pressure stream is split
to prevent- recirculation of partially converted gas
and the resultant possibility of losses by back-
conversion at higher temperatures. Liquid phase
conversion is required to bring the para fraction to
the equilibrium value at 20.40 K.
4) Modified Single Expander Cycle with Gas Phase

Conversion: This cycle with gas phase conversion
in the expander heat exchanger is shown in the right
half of Figure 5. Except for the addition of the
modifying heat exchanger, this cycle is identical to
the basic cycle with gas phase conversion.
Calculated performance curves for the four variations

of the precooled single expander cycle are shown in Figure 6.
The left half of Figure 6 relates cycle yield to expander
efficiency, and the right half is the plot of gros s liquefaction
work versus expander efficiency. Comparative analyses
indicate two important points. First, it is seen that modifica-
tion has a greater effect on this cycle than does the addition
of gas phase convers ion. This is principally because gas
phase conversion in the basic cycle results in an excessive
expansion engine flow. Second, the modified cycle is
extremely susceptible to continuous conversion. Both yield
and work figures indicate that for the modified cycle
continuous conversion is worth as much as 10 per cent
improvement in expander efficiency.
-70-
Conclusions
Two significant conclusions are drawn from these

theoretical calculations. First, proper use of ortho-para
catalyst in hydrogen liquefaction cycles can bring about
major improvements in the cycle performance without major
modifications of the process equipment. Second, effective
use of catalyst in hydrogen liquefaction cycles in some cases
is more effective in improving performance than either cycle
modification or major improvement in expansion engine
efficienc y.
In summary, for hydrogen liquefaction cycles employing

expanders, it is believed that gas phase ortho-para conversion
offers more cycle improvement per dollar invested than any
other area of process equipment development.
Discussion
Question by D. B. Crawford. Air Products. Inc.:

What effect would result from using more than one expander
below the liquid nitrogen refrigeration level?
Answer by author: With the precooled expansion engine cycle

as shown. use of more than one expansion engine would result
in expanding into the two-phase region. A lower pressure
cycle with more than one expander should result in a slight
improvement in cycle efficiency because a portion of the heat
would be removed at temperatures higher than the present
design exhaust temperature. However, it is doubtful that this
increase in performance would compensate for the added cost
and complexity of the cycle.
Question by J. B. Gardner, British Oxygen Research and

Development, Ltd.: With regard to ortho-para conversion in
the liquid phase at liquid hydrogen temperature, why does
a split expansion give a higher yield than a single expansion
with ortho-para conversion of all the liquid produced?
For a fixed ortho -para conversion. say to 95 per cent. one

has a closed hydrogen system into which normal hydrogen is
fed and the same amount of 95 per cent para liquid withdrawn.
The amount of ortho-para conversion is therefore the same
by either system and the heat liberated is at liquid hydrogen
temperature in both cases. Why should not both systems
have the same liquefaction coefficient?
-71-
Answer by author: If the only source of catalytic conver-
sion in the cycle were in the liquid receiver, then both the
split expansion and single expansion arrangements would
result in the same yield. However, there are possible
extraneous sources of ortho-para catalysis in process piping,
compressors, heat exchangers, and, particularly, in
adso rption purifiers. With the single expansion, tre result
is that converted gas is recirculated through the cycle wrere
part of it back converts at some point and raises the overall
heat load. The advantage of the double expansion arrange-
ment is that only the product liquid is catalytically converted,
thus eliminating any chance of back conversion in the cycle.
-72-
B-2
DESIGN DATA FOR ORTHO-PARAHYDROGEN CONVERTERS
D. H. Weitzel, C. C. Van Valin, and J. W. Draper

Boulder, Colorado
Of the many parameters which must be considered in

the design of an efficient ortho-parahydrogen converter. the
most obvious is choice of an effective catalyst. After this
choice there remain as variables the conditions of tempera-
ture. pressure, feed composition, and space velocity (or
contact time) under which the conversion proceeds. The
effect of these variables constitutes the kinetics of the
conversion. Studies of conversion catalyzed by unsupported
hydrous ferric oxide granules have resulted in sufficient data
for reasonably good converter design and have also made
pos sible a first approach to the kinetics problem.
Figure 1 shows liquid phase conversion of normal and

partially converted hydrogen. when catalyzed by hydrous ferric
oxide granules in 30-50 mesh size. These curves were
obtained by flowing hydrogen through a small capsule of
catalyst immersed in a liquid hydrogen bath. They indicate,
for example, that one can convert liquid hydrogen from 25
to 95 per cent para at a space velocity of 2000 STP volumes
per minute per unit volume of catalyst or 51 to 95 per cent
para at a space velocity of about 3200 STP volumes per
minute per unit volume of catalyst.
The liquid phase converter in the CEL-NBS 240 liter

per hour hydrogen liquefier contains 1. 5 liters of 30-100 mesh
hydrous ferric oxide catalyst and converts 240 liquid liters
per hour to about 94 per cent parahydrogen. This perform-
ance is represented by the large cross on the figure, and falls
very close to the curve obtained with 25 per cent para feed.
Actually, the feed to the converter in our liquefier is around
30-35 per cent para because of high para gas returning from
the storage dewar and transfer siphon.
The fact that catalyzed ortho to para conversion is

rapid and highly exothermic presents the problem of maintain-
ing a constant temperature (or at least a known temperature)
-73-
100
"-.
~
- ..........
.......
~I :
" '"
90
%PARA
80
25'1'. PARA FEED/
~ )-
5L5'Y,PARA FEED/
70
60
o 1000 2000 3000 4000 5000
SPACE VELOCITIES, MIN-I
(STP GAS VOLUMES PER MIN PER UNIT VOL OF CATALYST)
Figure 1. Liquid phase conversion at 19.7 o K, 15 psig.

(30-50 mesh hydrous ferric oxide catalyst)
70
r-- -- .....
-
60
j"-Oo..
% PARA ~<"1r
......
-.
-
50
~/(
r-- r- r- r--.
""- 8S~6"1r
-....
- !--
300
400 800 1200 1600 2000 2400
(STP GAS VOLUMES PER MIN PER UNIT VOL OF CATALYST)
Figure 2. Vapor phase conversion of 25 per cent para-

hydrogen at 20 psig. (30-50 mesh hydrous ferric
oxide catalyst)
TO ANALYSIS
CONTROL VALVES DEWAR

FOR COOLING
He
CATALYST
THERMO-
COUPLE LIQUID
N2
Figure 3. Test chamber for isothermal conver sion at

various temperatures.
-74-
throughout the catalyst bed. This problem is not so difficult
in the case of liquid phase conversion because heat of
conversion merely vaporizes some of the liquid without any
change in temperature. The equilibrium para concentration,
and therefore the driving force for conversion, remain
essentially constant. Heat of conversion can then be
relTIoved downstream of the converter to recondense the vapor
fraction.
In the case of vapor phase conversion, however,

removal of heat is entirely by thermal conduction, and
temperature gradients in the bed are unavoidable. The
closest practical approach to isothermal conversion is a
design which minimizes the path from the center of the
catalyst bed to a constant temperature source of refrigeration.
This condition is approximated by slender copper tubes of
catalyst surrounded by a boiling liquid bath, i. e., a design
similar to a tube and shell heat exchanger unit. Other
possibilities such as copper fins or fingers extending into the
catalyst bed also exist.
At best, however, the isothermal condition can only be

approached, and for a large installation it may be desirable to
place the catalyst in fairly thick beds, so that the conversion
which takes place near the center of a large bed of catalyst
may be almost an adiabatic process. In this case the space
velocity data obtained from an isothermal experiment will not
be very useful unless heat transfer through the catalyst is
analyzed.
We have tried to simulate both of these extremes, i. e.,

isothermal and adiabatic, in laboratory experiments. For the
isothermal case we first used a .326 inch 1. D. copper tube
in liquid nitrogen baths boiling at three different pressures.
Figure 2 gives space velocity curves obtained in this way.
These curves show, for example, that under ideal conditions
at 76°K one can convert from 25 to 50 per cent para with a
space velocity of 500 STP volume s of hydrogen per minute
per unit volume of catalyst.
The number of low temperatures available with boiling

liquids is quite limited, however, so a test chamber, shown
in Figure 3, was designed to be cooled with controlled streams
of cold hydrogen vapor. One stream spiraled around the
chamber from bottom to top and the other spiraled around the
chamber from top to bottom. The 1/8 inch copper tubes
-75-
carrying these streams, as well as the tube carrying feed
hydrogen to the catalyst, were soft soldered to the converter
with an excess of solder, so that essentially the coils were
an integral part of a heavy copper and solder wall surrounding
the catalyst. A thermocouple imbedded in the solder at a
point opposite the center of the catalyst gave the temperature
of the converter wall.
Figure 4 shows a series of curves obtained with this

heavy walled chamber, and also shows again the three curves
obtained with liquid nitrogen baths at 63.2 0 ,75.8 0 , and
85.6°K. Note that the 75.8 0 K curve is now traced by both
white and black dots. The white dots were obtained with a
liquid nitrogen bath and the black dots by cooling the heavy
walled chamber to 75.8 0 K with cold hydrogen vapor. It
is apparent that the two methods of cooling are Equivalent to
each other.
The other curves of Figure 4 cover temperatures

ranging from 23.4 0 K to 59.4 0 K, and were all obtained by
cooling the heavy walled chamber with cold hydrogen vapor.
From these curves one can determine the quantity of catalyst
required for isothermal conversion of normal hydrogen vapor
at any temperature from 23 0 K to 85 0 K.
We now have in one of the large eEL-NBS liquefier s

a vapor phase converter which is described in detail in the
following paper. This converter operates at 70 0 K, 1800 psig,
and has a feed of about 47 per cent para. We made a
special run with the converter, however, using 25 per cent
para feed, and obtained an average of 46.5 per cent para at
a space velocity of 1890 per minute. This performance is
represented by a cross near the dotted line in Figure 5. The
dotted line representing conversion at 70 0 K was sketched in
between the experimental curves, which are the same as
those shown on previous figures.
Actually, the activity of the catalyst used in this

converter was not as high as that used in our laboratory
measurements. The cross falls above the line of our
laboratory measurements primarily because of the high
pressure in this part ci. the liquefier cycle.
Figure 6 is another family of curves, also obtained

with the heavy walled test chamber, but with 51.5 per cent
instead of 25 per cent parahydrogen feed gas. From these,
-76-
100
~ r-.., ........ I'--.
..... r--....
90
~~
.....
""""
r-. r---... .......
.......
~
t-....
...... ~
~
i'- '"
.0
;-. r-.... ........ .........

.......... ~ ...""" ~ i'-. -
,
BO 0 '<t
...... J:. t - -
'"'"""' ............ """ ~ ~ ....... ~
i'-... r....... ~O i"-
-..:
0
r- r.... r-... ............ ;:: ~ ......:, r....... ~<. i"- ...

,"PARA
70
....... ......... ..........

~ .......... .......... r--... -.....; ~ ~~O i"- .
I-- r-- ::'k ...::: . . . . t-- ~ r-....
.......... '11.0
r"- '-... ....... . . . . . r- ..:::: t:- "«'I R

60 "'1·8 '"--
r- F::
3/.~ L-
......::.. "'6a.:!,/( 3ot.8
50
-.: ....... .........., r-=: - ~.
-... r- .........
40
30
- r-
-~
r-
85.6'/(
?S.81oo.
.p 400 eoo 1200 1600 2000 2400

Figure 4. Vapor phase conversion of 25 per cent para-
hydrogen at 20 psig. (30-50 mesh hydrous ferric
oxide catalyst)
10
r---o
60
~ r---- .........
.....
........ ....... ......... i"o..
..........
r-
-
.....•. ....... ....... r-:.. 39.'~ ........
50 ...... 6a.:!,/(
....... .......
-.:
- --r-r-
"I. PARA ~ """-0 ~ ........ .±
r--- -... r- •••.
7~-:~ t--
-I(
40 ~
!-85,6'/(
30
o 400 eoo 1200 1600 2000 2400
SPACE VELDCITIES, MIN-I
Figure 5. Comparison of laboratory experiments with

performance of gas phase converter in 240 liter
per hour liquefier.
=-
100
--
~~ -'r-::: F-
I--- ~ ~ ~
I~ f-.., ~
""- ......... r-.. '/(
-
:!3.S~
90
..... r-; :::::
-
r-::- ~ <6.3
-
~
loA ~ ~ "-
r-. ........ r-.... r--- -3<.,/ ~ =-~
;!Q:!
---.....
--
BO ~
7.0
"-. ~ "---~
,"PARA " - ~ r---t.
r- -....:;
f-e... r- 7.-
a.,
I"- '--I-
-- ---
~
70
-;- r-.. r- Sto -I-
...... r--. .S
60
-.. 89.S
'I(
400 eoo 1200 1600 2000 2400

SPACE VELDCITIES, MIN-I
Figure 6. Vapor phase conversion of 51. 5 per cent para-

hydrogen at 20 psig. (30-50 mesh hydrous ferric oxide
catalyst) -77-
and the previously shown curves, it is possible to calculate
apparent contact times within the catalyst volume and arrive
at some approximation for the order of the reaction. Although
data at other feeds and at smaller fractions of complete
reaction should also be examined, the Curves which we have,
suggest that the reaction is approximately second order.
Figure 7 shows a chamber which was designed to

simulate conversion taking place deep within a large bed
of catalyst, i. e., an approximately adiabatic reaction. The
chamber contained 20 cc of catalyst and was insulated by
means of helium filled commercial powder. and further
isolated from room temperature by means of a liquid
nitrogen jacket outside the powder. Refrigeration for cooling
the incoming feed gas to any desired temperature was again
provided by cold hydrogen vapor. Thermocouples on the
central axis of the chamber gave the temperature of the gas
just before it entered the catalyst, the temperature at the
center of the catalyst bed, and the temperature of the gas
emerging from the chamber.
The space velocity curves obtained in this way were,

as one would expect, considerably different from those
obtained with the approximately isothermal chamber. The
upper curves of Figure 8 show space velocity versus para
concentration of exhaust gas for 25 per cent para feed at
four inlet temperatures, and compare the adiabatic and
isothermal curves for 75.8 0 K. The importance of efficient
cooling of a converter is evident from this comparison. To
obtain 45 per cent para. for example, takes more than seven
times as much catalyst in the adiabatic case as is required
if the gas is kept close to the incoming temperature of
75.8 0 K.
The lower curves of Figure 8 show how the tempera-

tures on the axis of the catalyst chamber varied with space
velocity through the catalyst for constant inlet temperatures
of 41.0 0 and 51.6°K. The middle and exhaust temperatures
were fairly close together. showing that, especially at low
flows, most of the conversion takes place in the first half
of the catalyst. The maximum heating effect came at a
space velocity of about 400 per minute. At this velocity. gas
entering the catalyst at 41 0 K emerged at about 64 0 K, and
gas entering at 51.6°K emerged at about 72 0 K.
-78-
FLOW METER
ANALYSIS IN
He
THERMO- CATALYST -SANTOCEL

COUPLES POWDER
LIQUID
N2
Figure 7. Test chamber for adiabatic conversion at various

inlet gas temperatures.
100
90
\
SO \
,\
~
-- 26.7°K INLET
\ ....... ,
70
% PARA
60 \. "-
.. -. • -
"~ jl.o"~ IN~ET i
.-. 51.6"K INLET

50
1"- '-
....... I'--
~'''- .... 1- 75.J"K (ISOT~ER~ALl
40 .... - -'-1'-'~'_'
T5,S"I K I~LET
I',
~XH~USTI -
60 r--
MIDDtE I
40
.I. J
IjLETj- r--
o 200 400 EiOO 800 1000 _I 1200 1400

SPACE VELOCITY, MIN
Figure 8. Adiabatic vapor phase conversion of 25 per cent

parahydrogen at 20 psig. (10-30 mesh hydrous ferric
oxide catalyst)
-79-
If the process is adiabatic, the temperature of the
exhaust gas can, of course, be calculated by equating the
heat of conversion (as determined from the space velocity
versus para curves) to the heat required to warm the gas.
This provides a check to determine how nearly adiabatic
the system actually is. If this check is made for the two
inlet and exhaust temperature curves shown in Figure 8, the
calculated and observed exhaust temperatures are within
two degrees of one another. We feel, therefore, that these
data do fairly well represent ortho to para conversion under
adiabatic conditions.
The performance of an actual converter will, of

course, be somewhere between the isothermal and the
adiabatic case. To predict temperatures within a catalyst bed,
it is important to have some knowledge of the thermal
conductivity of the hydrous ferric oxide granules in a
hydrogen gas atmosphere. M. M. Fulk of this laboratory
conducted some measurements of this property for us, using
the equipment and techniques developed by his group in their
studies of evacuated powder insulations. For 10-30 mesh
granules containing one atmosphere of hydrogen gas, the
thermal conductivity between boundary temperatures of 76 0 and
300 0 K was 1.7 milliwatts/ cm oK. Hydrogen gas conductivity
in this interval is about 1.3 milliwatts/ cm oK, so the
catalyst increased the conduction by about 30 per cent.
Between boundary temperatures of 76°K and 20 o K, the total
conductivity was 0.76 milliwatts / cm oK, which is about
double the conduction, at these temperatures, of hydrogen
gas alone. These values make it pos sible (if one is able to
set up the heat flow equations) to calculate the temperature,
and therefore the equilibrium ortho-para composition, at
any point in a catalyst bed. *
*For the gas phase converter in the liquefier there

was, during a special test with 25 per cent para feed,
sufficient heat of conversion to raise the temperature of the
gas stream in one unit about 4 0 K if no heat were conducted
through the catalyst to the bath. The distance from the
center of the catalyst to the bath was 1. 5 inches, and the
observed temperature rise of the gas stream was about
1. 5 0 K.
-80-
Discussion
Question by W. J. Scharle, Air Products, Inc.: With

respect to reactivation of the catalyst, have any comparisons
been made on the effectiveness of reactivation by just
heating or heating and also pulling a vacuum on the converter
bed?
Answer by author: Althollgh no controlled tests of this have

been made, we feel that it is necessary to remove from the
immediate vicinity of the catalyst all the impurities which
are driven from the surface by heating. If the impurities
are not taken away they will simply recondense on the
surface when the catalyst is cooled. It is possible that
purging with dry clean hydrogen or helium could be substi-
tuted for evacuation, but we have not studied this comparison.
Question by J. V. Fetterman, Air Products, Inc.: State

reason for initial activation of catalyst.
Answer by author: The surface of the catalyst must be

clean. Exposure to atmospheric gases "poisonsll the surface
by covering it with a layer of' adsorbed gas and water
molecules. These must be removed before the catalyst will
effectively contact and convert hydrogen molecules.
Question by J. V. Fetterman, Air Products, Inc.: Is it

necessary to reactivate the catalyst during normal operation?
Answer by author: So long as exposure to air, water vapor,

or other poison.s does not occur, the catalyst can apparently
be used indefinitely without reactivation. See page 18 of the
1956 Cryogenic Engineering Conference Proceedings and paper
B-3 of the 1957 Cryogenic Engineering Conference Proceedings.
-81-
Question by R. Katzen, Air Products, Inc.: Has any study
been made of the effect of the configuration of the catalyst
chamber on ortho-para conversion. Would a high L/ D
vessel show any difference from a low L/D vessel?
Answer by author: Although no detailed study of chamber

geometry has been made, we feel that this is important only
insofar as it influences the heat transfer from the catalyst.
Either a very high or a very low LID ratio would, for a
given amount of catalyst, minimize the path from the center
of the catalyst to the refrigerated wall, thus resulting in
more nearly isothermal operation. In the case of a high
LID, however, pressure drop would be relatively high,
and in the case of a low LID ratio, it would be necessary to
guard against channeling in order to utilize the entire
catalyst volume.
Question by E. R. Goodfelt, Olin Mathieson Chemical

Corporation: What is the effect of pressure on the
catalyst efficiency?
Answer by author: The attached curves (Figures A and B)

give results of experiments in which we varied the pressure
from 20 to 400 psig. These are the only laboratoryexperi-
ments which have been done. Although we have no reason to
doubt the validity of these experiments, the vapor phase
converter described in paper B-3 showed somewhat less than
the expected gain in effectiveness at high pressure. The
gain which the large converter showed in going from 20 to
1800 psig was about the same as our laboratory experiments
showed in going from 20 to 400 psig.
Question by E. R. Goodfelt, Olin Mathieson Chemical

Corporation: What is the time required for reactivation of
the catalyst?
Answer by author: This depends on the condition of the

catalyst. For initial activation of a fresh batch of catalyst
we recommend at least 48 hours. After this, the contamina-
tion introduced by a brief exposure to air can usually be
removed by 16 to 24 hours of activation.
-82-
49
48
~
z
w
u PARA ~ PRESSURE
a:
w
..
Hydrous Iron Oxide,
0..
7cm. 3 , 20-30 Mesh,
25"0 Pora Feed, 76°K,
a: Constont Flow 7.4 Liters/min (STP)
~
47
46
0 50 100 150 200 250 300 350 400
PRESSURE, PSIG
Figure A.
52
PARA '!! FLOW

50
Hydrous Iron Oxide,
7cm',20-30Mesh,
25% Pora Feed,
76°K.
48
46
~
z
w
u
a:
w 44
0..
..
.;
a:
0..
42
40
38
36
0 5 10 15 20 25
FLOW, LITERS IMIN. (STP)
Figure B.
-83-
Question by L. T. Hendrix, Richfield Oil Corporation: Have
uses for parahydrogen, other than the advantage in storage,
been found?
Answer by author: Parahydrogen liquid provides an

excellent, stable, reference temperature for calibration of
low temperature thermometry equipment. The boiling
temperature of normal liquid hydrogen will slowly change as
conversion proceeds. Both the mutual transformation of the
two hydrogen modifications and their different properties
have been applied to the investigation of various physical and
chemical problems. For example, the self-diffusion of
hydrogen has been investigated in this way. For other
examples see IIOrthohydrogen, Parahydrogen and Heavy
Hydrogen" by A. Farkas. Cambridge University Press. 1935.
We occasionally receive requests for small quantities of pure

parahydrogen to be used by other laboratories engaged in
basic research.
-84-
B-3
VAPOR PHASE ORTHO-PARA CONVERSION IN THE

LARGE CEL-NBS HYDROGEN LIQUEFIER
W. A. Wilson and D. H. Weitzel

Boulder, Colorado
The use of hydrous ferric oxide catalyst for

converting hydrogen from the ortho to the para state during
liquefaction was developed over a year ago here at the
National Bureau of Standards and was reported in papers A-3
and A-4 of the 1956 Cryogenic Engineering Conference
Proceedings. Since that time, this type of catalyst has been
used in the large CEL-NBS hydrogen liquefier. The original
1500 cc of catalyst is still in the liquefier and has performed
quite satisfactorily. It has, during the past year, converted
a little more than 170,000 liters of liquid hydrogen to around
94 per cent para at an average rate of about 240 liquid liters
per hour. During this time we have investigated the feasibil-
ity of intermittent operation without reactivating the catalyst
and have produced as much as 48,500 liters of liquid during
eight separate runs without reactivating the catalyst.
Studies of liquefaction cycles and catalyst perform-

ance, as reported in the two preceding papers, indicated that
an appreciable increase in the efficiency of the large hydrogen
liquefier could be realized by the installation of a vapor phase
converter. This modification is quite feasible in the CEL-NBS
liquefier if it is used in conjunction with the closed cycle,
two-expansion valve arrangement which was described by
V. J. Johnson in paper A-4 of the 1956 Cryogenic Engineering
Confe'rence Proceedings. With this arrangement only the
liquid which is transferred to the storage dewar is converted.
By supplying this liquid phase converter with hydrogen having
a higher para level, larger liquefaction rates can be
realized.
The only convenient place for a gas phase converter

in the CEL-NBS liquefier is immediately downstream of the
purifier. The total flow of gas at this point is 750 SCFM,
but only one-sixth of this is liquefied. The remaining five-
sixths is recycled to the compressors. This recycled gas,
-85-
together with new make-up gas, passes through the gas phase
converter again. The net result of this process is to raise
the para level of the entire system, including the feed gas to
the vapor phase converter. This effect complicated our
estimation of the quantity of catalyst required because we do
not as yet have space velocities for a wide variety of feed
concentrations. We wanted to avoid over design since this
was to be an experimental test as well as a practical
improvement in our liquefier capacity.
From the available data it appeared that three liters

was about the correct quantity of catalyst. It was decided
to use three catalyst chambers, each containing one liter,
with sections of cooling coil between the chambers. In
order to compa"l'e the pressure drops and activities of
different mesh sizes, the chambers were to be filled
respectively with 10-30, 30-50, and 50-70 mesh granules.
Space velocities of representative samples for each chamber
were measured in the laboratory at 76°K and one atmosphere
for conversion from 25 per cent to 48 per cent para. These
were as follows:
lO-30 mesh 580 STP volumes per minute per

unit volume of catalyst
30-50 mesh 625 STP volumes per minute per

50-70 mesh 640 STP volumes per minute per

To compute the pressure drop across the catalyst

chambers, the formula for obtaining pressure drop through
beds of granular solids, given in Perry's Chemical Engineers
Handbook,was used. This formula states that:
531-L LVoAf
AP =
D 2
P
where
AP = pressure drop, lb./sq. ft.
I-L = viscosity of fluid, lb./ft. sec.
-86-
L = depth of packing, ft.
v = velocity of fluid through empty tower,

o
ft./sec.
= wall effect factor
= nominal diameter of particle, ft.
Based upon this formula, the volume of the catalyst

required, and the physical limitations within the purifier
dewar, chambers fourteen inches in length were constructed
from 3-1/2 inch nominal, schedule 80 stainless steel tubing.
The resulting chambers were constructed as shown in Figure
1. The catalyst was contained within the six inch effective
length by means of 100 mesh screen. Each chamber contaired
three thermocouple wells evenly spaced along the length of
the catalyst bed. Pressure drop taps were placed at both
ends of each chamber and sample taps were located at the
discharge of each chamber. The chamber was all welded
construction and extreme care was taken to avoid overheating
the catalyst during the final welding operation, since it has
been determined that this catalyst will change its structure
and be completely ruined if heated much beyond 140 o C. The
theoretical pressure drop across the catalyst beds was
computed to be 0.25 psi for the 10-30 mesh, 0.85 psi for the
30-50 mesh, and 1.89 psi for the 50-70 mesh catalyst.
Figure 2 shows a schematic diagram of the CEL-NBS

liquefaction system and the location of the catalyst chambers
relative to the overall liquefaction process. The chambers
were installed in series at the discharge of the silica gel
coil in the hydrogen purifier and were submerged in liquid
nitrogen along with the silica gel coil. They were installed
with the coarser mesh catalyst coming first in the line of
gas flow, the intermediate mesh second, and the finest mesh
last, with a twelve foot length of 5/8 inch O. D. copper
tubing separating each catalyst chamber to serve as a heat
exchanger. These heat exchangers were installed to insure
that the temperature of the hydrogen gas entering each
catalyst chamber would be the same in order to compare the
conversion qualities of the three chambers. During normal
operation, a vacuum is maintained over the liquid nitrogen
bath resulting in a liquid nitrogen temperature of approxi-
mately 70 o K.
-87-
PRESSURE PRESSURE
SAMPLE
THERMOCOUPLE WELLS
Figure 1. Vapor phase catalyst chamber
LIQUEFIER PURIFIER REFRIG ERATION

DRIER
Figure 2. Schematic flow diagram liquefaction process
-88-
Since installing this vapor phase, ortho-para converter,
a number of liquefaction runs have been made with quite
satisfactory results. Analysis of the hydrogen gas at
various stages of the process revealed that the following
average para levels were obtained:
A nalysis Point Per Cent
Inlet to purifier 47.1

Discharge of silica gel coil 47.4
Discharge of first catalyst chamber 48.2
Discharge of second catalyst chamber 49.1
Discharge of third catalyst chamber 49.9
Inlet to liquid phase converter 49.9
Discharge of liquid phase converter 97.7
The equilibrium para concentration at the temperature of the

vapor phase converter is 56 per cent while at the tempera-
ture of the liquid phase converter it is 99 per cent. The
above results show that each chamber is doing about the
same amount of converting so there is no particular advan-
tage in using the finer mesh catalyst. In fact, from an
economical standpoint it would be desirable to use the
coarser mesh catalyst.
Prior to the installation of the vapor phase converter,

the para level of hydrogen at the inlet to the liquid phase
converter ranged from 31 to 43 per cent, depending upon the
amount of high para return gas obtained from the storage
dewar during the run. This level is now very nearly 50 per
cent, so it is evident that the para level throughout the
entire liquefier is raised because of the vapor phase
converter. The higher para at the inlet to the liquid phase
converter now raises the liquid product from a previous
average of 94 to almost 98 per cent para. We could
actually increase our production rate a little more if we
would remove part of the 1.5 liters of catalyst from the
liquid phase converter and return the para level of the liquid
product to 94 per cent.
During the liquefaction runs using the vapor phase

converter, the actual pressure drop measured across the
three catalyst chambers was 0.5 psi in the first chamber,
1.1 psi in the second chamber, and 1. 5 psi in the third
chamber. The calculated values based on standard methods
-89-
of analysis were, as mentioned before, 0.25 psi, 0.85 psi,
and 1.89 psi. Since the normal total pressure drop across
the purifier and drier is in the order of 150 psi, this small
additional drop acros s the catalyst chambers is negligible.
Calculations show that under normal operating

conditions the heat of conver sion in one chamber is sufficient
to raise the temperature of the gas roughly lOC if no heat
is transferred from the chambe r to the bath. The thermo-
couples gave temperature s of about 1/4o C above that of the
bath, showing that some heat was transferred through the
catalyst bed.
The liquefaction runs resulted in a production

increase of 7 per cent over that obtained when not using the
vapor phase converter. In addition, the para level of the
liquid hydrogen increased from approximately 94 to 98 per
cent. If the 94 per cent liquid para level were maintained,
one could expect a production increase of about 12-1/2 per
cent.
The liquid nitrogen consumption when using the

vapor phase converter increased by approximately 3 per cent
or 13 liters per hour. Here at the National Bureau of
Standards the cost of producing liquid nitrogen is 10 per cent
of the cost of liquid hydrogen. Taking into consideration the
increase in hydrogen gas liquefied and the additional consump-
tion of liquid nitrogen, the output of the plant from an econom-
ical standpoint is increased by 6-1/2 per cent.
Discussion
Question by M. A. Dubs, Linde Company: How rugged has

the catalyst been in actual service? Has it behaved well
mechanically?
Answer by author: The catalyst has proven to be quite

durable when in service. Care is taken in the reactivation
process to avoid overheating the catalyst. The reactivation
process used involves heating and evacuation. There has
been no occasion to remove the catalyst since its initial
installation, consequently, it has not been manhandled during
its use.
-90-
Apparently the catalyst has not broken down or dusted during
its use. This conclusion is made from the fact that the
degree of conversion and liquefying rate has not changed
during its use indicating that the quantity of catalyst is the
same as it was initially.
Que stion by D. B. Burkhardt, E. 1. du Pont de Nemour s

and Company, Inc.: Would the same end result be achieved
by combining the total quantity of catalyst used in the liquid
phase converter and the vapor phase converter into a single
liquid phase conve rter?
Answer by author: No. The efficiency of the liquefier

depends on the heat load within the liquefier which is a
function of the amount of ortho-para conversion.
Performing part of the conversion in the hydrogen purifier by
virtue of the vapor phase converter reduces the heat load on
the liquefier and thus increases its efficiency.
If all the catalyst were placed in the liquid phase converter,

the degree of conversion would be higher, but the efficiency
of the liquefier would be lower due to the additional heat
load imposed on the liquefier.
-91-
B-4
DESIGN AND APPLICATION OF A GAS LIQUEFIER
T. W. Schaffers
Philips Electronics, Inc.
Mount Vernon, New York
Introduction
It is the purpose of this paper to describe the non-

conventional design features of the Norelco gas liquefier
which is based upon the principle of the Stirling cycle and
the demonstrations of J. W. L. Kohler l at the last confer-
ence. The Stirling cycle consists of two isothermal steps
for compression and expansion and two isochors for cooling
and reheating. A schematic and diagrammatic representation
of the process in four phases is shown in Figure 1, illus-
trating the piston and displacer position at the beginning of
each phase.
The design is an extremely versatile, single cylinder,

water cooled apparatus for gas liquefication. It is unique in
that the gas to be liquefied does not pass through the working
parts. It merely condenses on the surface of the cylinder
head. This permits the production of an exceptionally pure
product, free from lubricating oil or similar contamination.
This new apparatus uses the cold gas principle of refrigeration
and efficiently reaches a range from -80 to _200 0 C in a
single stage.
The Norelco gas liquefier is capable of supplying

diverse liquefied gases. i. e., those which liquefy at
temperatures above -2000 C. The rate of production is
determined by the characteristics of the gas or gas mixture.
In the case of liquid air, production of 5 quarts per hour can
be obtained under the following conditions:
1 J. W. L. Kohler,Proceedings of the 1956 Cryogenic

Engineering Conference, p. 243, Boulder, Colorado,
September 1956.
-92-
1 PHASE 1. The medium is in the
compression chamber where it
is compressed by the piston.
PHASE 2. The displacer drives the

medium from the compression area to
2 the expansion chamber through the
cooler, where heat generated by the
compression is absorbed by circulating
water; passing next through the regen-
erator, where further cooling occurs,
to the head.
PHASE 3. Most of the medium is now in

the expansion chamber. The tempera-
3 ture of the medium is further reduced
by the expansion resulting from the
combined downward movement of both
the main piston and the displacer.
PHASE 4. The cold medium is driven

back to the compression chamber by
the displacer. During this phase, heat
is also removed from the "freezer"
head. In passing back to the compres-
sion chamber, the medium absorbs the
heat stored in the regenerator during
the second phase.
Figure 1. Operating cycle of the

Norelco gas liquefier
-93-
Running speed, 1725 rpm
Cooling water, 220 gal/hr
Cooling temperature, 50 0 F
Dew point, 50 0 F
Average helium pressure in working
area, 310 psi
Variance of these values will affect the output to

some extent. Using hydrogen instead of helium as the com-
pression gas will increase the output by approximately 10 per
cent.
Depending upon the amount of moisture in the air,
the machine can produce liquid air for periods up to 24 hour s
without interruption. After running for such periods, defrost-
ing is required. A special hot air blower is delivered with
the machine for this purpose.
The apparatus is designed primarily for liquid air

production. However, with modifications it can be used for
liquefying other gases as well as for low temperature cold
boxes.
Mechanical Design
The mechanical design is unusual and compact.

Major components are the crankcase and crankshaft, oil
pump and filter, connecting rods, piston, displacer, cylinder
housing, water cooler, regenerator, condenser and ice
separator (which removes moisture and carbon dioxide by
freezing) •
Special design features were necessary to prevent
lubricating oil from reaching the compression area and
eventually blocking the regenerator. While the displacer is
not lubricated in the liner, the piston is. Due to the careful
design in selected dimensioning and very high precision
tolerances and superfinishes, oil contamination in the
compression area has practically been eliminated.
The piston and displacer are moved by a single

crankshaft. The displacer connecting rod is located between
the two parallel connecting rods of the piston and passes
through the center of the piston. The angle of contact for
the displacer and piston connecting rods to the crankshaft is
selected in such a way that the movement of the displacer is
in exact phase with respect to the movement of the piston.
-94-
The crankshaft which passes through the crankcase
on one side is coupled to a lO horsepower motor with a
flywheel and clutch. A liquid seal on this side prevents
gas pressure losses in the crankcase.
A detailed description of the various components of

the gas liquefier is given opposite Figure 2.
Installation
The complete gas liquefier including the control

instrument panel is shown in Figure 3. Provisions have
been made for maintenance free operation and complete
safety controls. The unit will automatically shut off under
the following conditions:
1) No cooling water
2) Cooling water less than 2 gallons per minute
3) Cooling water temperature above 100 o F
4) Gas pressure below 150 psig or above 375 psig
5) Under pressure in the condenser below 15 mm

of mercury
6) No oil pressure lubrication
-95-
Figure 2. Discription and function of the various components
of the gas liquefier
Piston (1) rides in cylinder (2).

Displacer (3) causes a periodic flow of gas coolant back
and forth between spaces (4) and (5).
This gas goes through the annular exchangers surround-
ing the displacer. Parallel connecting rods (6) couple the
piston cranks (7), of the crankshaft (8) with piston (1).
Displacer rod (9) is linked to connecting rod (10) and
crank (11) of the crankshaft (8).
Port (12), cooler (13) and regenerator (14) connect the
compression area with the head or "freezer" (15). Displacer
piston (16) contains insulation (l 7) •
Condenser (18) is the surface upon which the gas

becomes liquefied. Annular channel (19), tapping pipe (20),
permit liquid collection and draw off.
Insulation (21) surrounds plates, past which incoming
gas flows, within the outer enclosure (22).
Gas inlet (23).
Plates of the ice separator (24) are joined by the
tubular structure (25) to the top of the head (15).
Gas-tight shaft seal prevents medium loss (26).
Cylinder supplying gas medium (27).
Medium supply pipe (28) with one-way valve (29).
Driving motor (30).
Figure 3. Completed Norelco gas liquefier, including

instrument panel.
-96-
Figure 2.
o ,.
Figure 3.
-97-
Applications
Many uses for this gas liquefier are visualized. Below

are just a few of the possible applications.
I} Small scale liquid air production.

Z} Liquefication of gases between _80 o C and _ZOOoC,
such as nitrogen, oxygen, methane, etc.
3} Removal of carbon dioxide from air and drying of
protective gases.
4} Recovery of escape gases from liquid gas containers,
such as liquid nitrogen, oxygen, and liquefied natural
gas.
5} Separation of gases such as nitrogen from hydrogen;
argon from neon; krypton, xenon from argon.
6} Cold box application, such as stabilization of metals,
shrink fit applications, conservation of serum in
animal husbandry.
7} Single unit pre-cooler in obtaining temperatures
between _ZOOoC and absolute zero.
8) For use as a refrigerator for a condenser on a
distillation column; such as separating hydrocarbon
gases in oil refineries and separating nitrogen or
oxygen in a pure state from atmospheric air.
Discussion
Question by R. Katzen, Air Products, Inc.: Is it necessary to

remove water and carbon dioxide from air to make unit oper-
able? Can you use external absorbers, scrubbers, or
reversing regenerators?
Answer by author: The water and carbon dioxide must be

removed from the air for satisfactory operation of the gas
liquefier. External absorbers can be used. Scrubbers and
reversing regenerators, however, cannot be used.
Question by H. W. Altman, Herrick L. Johnston, Inc.: Is

this liquefier suitable for liquefaction of oxygen gas 7
Answer by author: Yes, it is.
-98-
B-5
THE THERMODYNAMIC THEORY OF A LIQUID NITROGEN

GENERATOR USING A NORELCO
REFRIGERA.TOR-LIQUEFIER
F. G. Brickwedde
North American Philips Company
Mount Vernon, New York
A design of a liquid nitrogen generator using a

Norelco refrigerator is shown in Figure 1. The cycle was
suggested by S. C. Collins of M.1. T. and has been developed
by Arthur D. Little, Inc. It has three essential parts
consisting of a heat exchanger, a rectification column, and a
refrigerator. The refrigerator condenses the vapor flowing
from the top of the rectifier, and through the liquefied vapor
returned to the rectifier as liquid reflux furnishes the
refrigeration needed for the rectifier and heat exchanger.
The heat exchanger functions as a refrigeration-purifier
removing moisture and carbon dioxide from the feed air
before entering the rectifier.
An important difference between a conventional

liquid nitrogen generator and one using a Norelco refrigerator
is in the pressure of the air feed to the generator. In the
conventional generator high pressure air is needed because
the refrigeration for the operation of the generator is obtained
from the air feed. In a design like Figure 1 in which the
refrigeration is supplied externally by a refrigerator,
pressure of the air is needed only to overcome the resistance
to flow of air and waste gases through the refrigerator.
The operator of a generator like Figure 1 has only

two variables to control for steady state operation. These are
the pres sure and rate of supply of the air feed to the
generator. It was the purpose of this analysis to determine
the effect of these two operating variables on the liquid
nitrogen producing capacity of the generator, the composition
of the liquid product and waste gases, and the temperature of
the air entering the rectifier.
-99-
Air Feed
"-(20·C)
(rio)
Liq. N2
Product·
Rote, ~
Purity, ~
Liq. Woste
02+N2 tA
x.
Figure 1. Schematic diagram of liquid nitrogen generator

1.0 1.0
0.9 0,9
~
'0 0,8 Liquid Air_
0,8
liquid
~ 0,7 Concentration
Ranoe of
0,7
~ 0,6 liquid in Rectifier

Waste
0,6
i -----.
Equilibrium
0,5 with Air 0.5
0,4 293·
250·
200·
150·
100·
100 192 300 338 404

T·K
Ra'e of Air Feed to Generator (orom moles/hr.)
Figure 2. Rate of air feed to generator (no heat loss) as

a function of rectifier inlet temperature and molar
composition of liquid product and waste gas.
1,0 1.0
0,9 0,9
~ LIquid
Concentration 0.8
'0 08
Rano! of
~ 07 Rectifier 0,7
! 0.6 06
~
'" 05 0,5
04 293·
Proportional
250"
to Insulation
loss
200-
I!!IO·
100-
300 396 t
T·K
Role of Air Fud to Generator (Qrom moles/hr.)
Figure 3. Rate of air feed to generator (40 per cent heat

loss) as a function of rectifier temperature and molar
composition of liquid product and waste gas.
-100-
Figures 2 and 3 are diagrams of the composition
of liquid product and waste gases, and the temperature of
the air entering the rectifier. Figure 2 is for an ideal heat
exchanger, i.e., one without any losses, leaving the entire
refrigerating capacity of the refrigerator for liquefaction
of oxygen and nitrogen. Figure 3 is for a generator with a
heat exchanger purifier having a refrigeration los s equal to
40 per cent of the refrigerating capacity of the refrigerator.
A comparison of Figures 2 and 3 shows the effect of heat
exchanger-purifier loss on the operation of a generator.
The mathematical analysis of this paper is based on

the principles of mass and energy balance. Since the only
work done by the air in the generator is in pushing gas
directly ahead of itself, the energy balance is really an
enthalpy balance. Interesting features of this analysis are
discussed in the following paragraphs.
The generator is operable over a wide range of air

feed rates, extending on the low side from a rate just
sufficient to absorb all the refrigeration of the Norelco
liquefier to a maximum that the rectifier column and condenser
will accept before diverting some cold unliquefied air directly
from the air inlet of the rectifier to the waste drain through
which the exces s air leaves as gas mixed with waste liquid
draining from the rectifier. These limits mark the beginning
and end of the curves in Figures 2 and 3.
There is a critical rate of air feed (indicated by the

breaks in the composition curves of Figures 2 and 3) above
which the purity of the liquid nitrogen from the generator is
determined by the efficiency of the rectifier, and below which
the composition is determined by thermodynamic principles
and not by rectifier efficiency. These thermodynamic
principles are mass balance of oxygen and nitrogen and a
limitation on the maximum concentration of oxygen possible
in the waste liquid draining from the rectifier. This maxim un
concentration of oxygen is 51 mol per cent, the concentration
of oxygen in the liquid phase in equilibrium with air having
the composition of the atmosphere.
Over the wide range of air feed rates which the

rectifier and condenser can handle, the liquid nitrogen
capacity of the generator is only indirectly influenced by the
-101-
rate of air feed, i. e., through the heat exchanger-purifier
losses. For generators with ideal heat exchangers, and
generators with exchangers having constant refrigeration
losses (independent of rate of air feed) the liquid product
delivery rate of the generator is independent of the rate of
air feed. As the rate of air feed to the generator is varied,
it is the amount of waste gases that varies--not the liquid
product capacity of the generator which is determined by the
refrigerating capacity of the Norelco liquefier.
Over the whole range of air feed rates for which

steady state operation of the generator, with liquid reflux
flowing to the rectifier, is possible, the following operating
characteristics of the generator are all constant: the rate
of flow of vapor from the rectifier to the condenser, the
rate of condensation of vapor in the condenser, and the rate
of flow of liquid reflux from the condenser to the rectifier.
The temperature of the cold air flowing from the

heat exchanger to the rectifier falls as the rate of air feed
to the generator is increased (see Figures 2 and 3). The
maximum rate of air feed that can be accommodated in the
rectifier column is reached when the temperature of the air
flowing into the rectifier is the same as the temperature of
the waste liquid draining from the bottom of the rectifier.
Increasing the pressure of the air feed to the

generator increases the production of liquid nitrogen. Raising
the pressure from 1 atmosphere to 1-1/3 absolute atmos-
pheres increases the delivery rate 10 per cent, or to 1-2/3
atmospheres increases it 18 per cent, or to 2 atmospheres
25 per cent. This increase results from the increase in
refrigerating capacity of the Norleco liquefier with increase
of the condensing temperature which is equivalent to a 3.8
per cent increase in refrigerating capacity per l)C increase
in condensing temperature.
The heat exchanger-purifier losses which may total

40 per cent of the refrigerating capacity of the refrigerator-
liquefier, in extreme cases, are strongly dependent on the
rate of air feed. The loss arising from condensation of water
and carbon dioxide in the air feed is proportional to the rate
of air feed. The insulation los s is proportional to the
difference between the temperatures of the air feed stream at
the warm and cold ends of the exchanger (indicated in Figure 3).
-102-
The 6. T los s at the warm end of the exchanger (6. T ew) is
proportional to the rate of discharge of waste gases. The
maximum driving force for transfer of heat in the exchange
(6. T at the cold end) decreases as the rate of air feed is
increased. Table I shows this dependence of the losses on
rate of air feed, and also the relative importance of the
different contributing losses. The calculated data in Table I
are for an air feed at 20 0 C with a 100 per cent relative
humidity, for a 2 inch thickness of hairfelt insulation around
the heat exchanger which is built around the rectifier column,
and for a 6. T of 100C at the warm end of the heat exchanger.
The losses are compared for flow rates of 400 and 300 mols
of air per hour. The higher rate corresponds to the
maximum that the rectifier column can accommodate, using a
Norelco refrigerator that can liquify 5.7 liters of liquid air
per hour under favorable laboratory conditions. The lower
rate of air feed (300 mols/hr) is still in the range in which
nearly pure liquid nitrogen is obtainable from the generator.
Losses are expressed as fractions of the refrigerating capacity
capacity Q c of the Norelco liquefier used as the condenser.
TABLE I
Rate of Refrigeration Loss
Various Losses Air feed--400 Air feed- -300

mols/hr. mols /hr.
Condensation of 0.19 Q c 0.14 Q c

Moisture
Heat Leak through 0.11 Q c 0.08 Q c

Insulation
6.T Loss 0.04 Q c 0.02 Q c
Total Loss 0.34 Q c 0.24 Q c
A 15 per cent increase, (1-0.24)/(1-0.34), in the

production of liquid nitrogen from a generator, operating
under the assumed conditions, would result from reducing
the rate of air feed from 400 to 300 mols / hr.
-103-
Question by G. C. Szego, General Electric Company:
Would it not prove advantageous for a purer bottoms product
to insert the feed above the bottom and provide the necessary
boil up by an auxiliary heat exchange loop with the incoming
air feed?
Answer by author: The useful product of the generator

(liquid nitrogen) flows from the top of the rectifier as vapor.
The liquid draining from the bottom of the rectifier column
is discarded as waste gas.
This suggestion would be most advantageous in the range of

low rates of air feed (less than 338 mols/hr for Figure 2,
and 203 mols/hr for Figure 3) where the nitrogen purity of
the liquid product from the generator is limited by the
comparatively low concentration of oxygen (51 per cent) in the
waste liquid draining from the bottom of the rectifier.
Increasing the oxygen concentration in the waste liquid from
the rectifier would increase the nitrogen concentration in the
liquid product of the generator. Your suggestion would
accomplish this and would shift the critical rates of air feed
for the liquid product curves in Figures 2 and 3 to lower
values (approximately the minimum air feed rate for the
Figure divided by 0.78).
Question by D. B. Crawford, Air Products, Inc.: Could you

tell us approximately what fraction of the air feed comes out
as product and what fraction comes out as waste gas?
Answer by author: The answer to this question depends upon

the rate of air feed to the generator.
The rate of delivery of liquid product (liquid nitrogen) from

the generator is approximately constant, that is, it is
independent of rate of air feed so long as this rate is in
excess of a minimum rate. All the air feed that is not
delivered as liquid product is discharged as waste gas.
The rates of delivery of liquid product are approximately 192

mols/hr for Figure 2 and 115 mols/hr for Figure 3. The
fraction of the air discharged as liquid product, is therefore
-104-
approximately 192/(rate of air feed, mols/hr) for Figure 2
and l15/(rate of air feed, mols/hr) for Figure 3. Figures 2
and 3 may be consulted for the further evaluation of these
fractions.
The fraction of the air feed discharged as waste gas is one

minus the fraction discharged as liquid product.
Question by J. F. Grunberg, L'Air Liquide: Assuming no

external heat los ses, if the machine is able to produce a
given amount of liquid air without any separation, how much
liquid nitrogen production would be expected starting from
air?
Answer by author: Starting with dry and carbon dioxide free

gas at 20 0 e and a pre ssure of 1 atmosphere absolute, 344
joules of refrigeration are required to condense 1 ml of
liquid nitrogen, and 388 joules to condense 1 ml of liquid air.
Hence, for the assumed condition of equal amounts of
refrigeration for condensing nitrogen and air, the liquid
nitrogen capacity of the generator in liters/hr should be
1. 13 times the liquid air capacity. A refrigerator -liquefier
with a capacity of 5 liters of liquid air per hour would under
the assumed conditions have a capacity of 5-2/3 liters/hr
of liquid nitrogen.
In practice, the refrigeration consumed in condensing the

moisture and carbon dioxide in the air, and heat leaks
through the thermal insulation around the rectifier are
important considerations.
-105-
C-l
DESIGN AND PERFORMANCE DATA ON A LABORATORY

SIZE LIQUID NITROGEN PLANT
G. Y. Robinson and J. T. Selldorff

Cambridge, Massachusetts
S. C. Collins
Massachusetts Institute of Technology
Cambridge, Mas sachusetts
The need for a simple laboratory size liquid nitrogen

generator is increasing rapidly as more and more organiza-
tions enter the cryogenic field, and research efforts in low-
temperature applications increase. Arthur D. Little, Inc.
has developed for the North American Philips Company, Inc.
a compact liquid nitrogen plant employing as its source of
refrigeration the Norelco gas liquefier.
The cycle, whose essential feature is simplicity, was

conceived by S. C. Collins of M.1. T., and is shown on
Figure 1. Feed air near its dewpoint is introduced at the
bottom of a rectification column, and 99.5 per cent nitrogen
removed at the top. The thermal economy of the cycle is
kept high, to the point where 75 per cent or more of the
refrigeration supplied by the Norelco air liquefier is removed
as liquid nitrogen product, and production rates of 3-1/2 to 4
liters per hour of 99.5 per cent nitrogen are obtained.
The major components of the unit are a small

diaphragm compressor, a counter current heat exchanger, a
rectifying column and the Norelco gas liquefier used as a
condensing unit.
Because air is a mixture, reflux is required in the

distillation column to produce nitrogen purities of 99.5 per
cent or better.
The reflux ratio necessary is about 0.5 to 0.6, that

is, about one-half of the feed stream is withdrawn as product and
-106-
......
.-.-~~
Figur e 1.
. /"
~\-~/ / D~
/ \ /.-'----
~>--.
'- ..
11M ..
v..1II~"Id.~. ,... ~t_~
*"M'VN ,..,. ,.-J
r•. .....,. ..... I ·~ ...,.
""",...,..,CIAII,. ........,'/ .., .... ~T <-*,'... ~I 4""ltl-
1\"'-'" ...... M - ,.0.
.. .
Figure 2.
-107-
half is returned counter cur rent to the feed in the column,
and in the exchanger. Since the waste stream ente rs the
heat exchanger in the liquid state, the refrigeration available
from it will produce almost saturated feed air at the inlet
to the distillation column. Such a design insures adequate
purification of the inlet air during normal operation.
The pressures in the unit are on the order of 2 to

3 psig.
Heat Exchange
During normal operation, the moisture and carbon

dioxide in the air are frozen out on the walls of the inlet
passage of the exchanger. The cooling medium is the waste
stream. In the selection of this exchanger, a compromise
must be struck between heat transfer efficiency and adequate
volume so that ice and carbon dioxide snow may collect
without causing excessive pressure drop or entrainment
in the inlet stream.
Two types of heat exchangers were considered. One

was a double tube type depending on a thick solder bond
between the two tubes for heat transfer and the other was a
fin and tube type with the inlet air passing along the outside
of the fins and the return air through the central tube.
Heat transfer coefficients for the two types of

exchangers in actual service were determined experimentally.
See Figures 2 and 3. On the freeze out side, the coefficient
was found to be lower for the finned tubing than for the plain
tube. This condition is due largely to the lower superficial
velocity used. Nevertheless, the large area ratio of the
finned design, about 12 to 1, made this design more desirable
from the standpoint of length of run before defrosting, weight
of exchanger (and consequently cool down time), and
reduction of entrainment of ice and carbon dioxide snow.
Column
In order to conserve space and weight, we chose wire

mesh saddles for the packing in the column. This packing,
-108-
discussed by A. W. Fisher and R. J. Bowen, 1 was
developed by H. O. McMahon for air borne oxygen plant?
and other applications where little space, light weight and
high efficiency are considerations. HETP's obtainable with
McMahon packing are on the order of 2 to 3 inches, as shown
on Figure 4.
In our column, which was 1-1/2 inches in diameter

with 28 inches of packed height, no difficulty was experienced
in obtaining nitrogen purities of 99.5 per cent and higher.
When the packing is free of moisture and carbon dioxide, the
pressure drop across the entire column is on the order of 10
to 14 inches of water. Frost formation on the packing,
however, increases its resistance to flow. and seriously
interferes with the operation of the unit. Such increase in
resistance in the column packing causes flooding and loss of
capacity since the gas to be rectified is drawn through the
column by the pressure differential created by condensation of
nitrogen in the head.
Start-up
We have paid special attention to the design of the

unit for efficient start-up and thawout. Three schemes were
found which result in satisfactory operation.
The first method is to use pure dry nitrogen from a

cylinder to cool the unit and the heat exchanger before
admitting room air. The second scheme is to provide for
auxiliary surface area at the coldest spot in the system to
freeze out impurities in the relatively small amount of air
required for start-up. The auxiliary cooling surface area
is valved out of the system after normal operation has begun.
The third employs a silica gel or molecular sieve adsorbent
bed which removes the moisture from the air during start-up
and which is regenerated by the dry waste gases during
normal operation of the unit.
lA. W. Fisher and R. J. Bowen. Chern. Eng. Prog.

45, 9. 359 (1949).
-109-
-_ ..
.-.:I~.'"
,_,,...,oH.(",.,.
...-, 4"
, .....
.l-
- .................
(N , .
".y
_ t ... ~ . _ _ 1
~ . ~ , ---..-
.J~""""",,"...-Lo,t
'-1 '" - I I _1-
_ 0 ___ • _ __ ' _ . - . ---
l.'-_ _ _ _ _ __ _ __ _
• ~ • j • , I ~ • • ., ...
1,1,\\ . '~.'-
-\\\,~ ,\, ~\\, ~ .\,'
Figure 3.
1-
!
.,.
...
.. l
- f ..-
... -..!.~
-
__ ______
- -
o~---~---~- ~
o
MoP"l6ItTU ~
""~I"IA"" 9!' gt:!I~1:J - ..
... ""'"
~f!t,c"....,.
Figure 4 .
.... "'.....
ItwR\ ...... f.J, ......
~-~- ~ ~
Figure 5. Perspective sketch of liquid nitrogen unit.

(Access plate and powder insulation removed)
-110-
This last scheme has proven to be a satisfactory
compromise in operation and has been chosen for the
present design.
Room air, heated to about 250 0 F, is used to drive

out all of the moisture in the unit. This air is admitted at
several points to insure rapid and complete thawout. We
considered using electrical resistance wires to speed up the
thawout in critical locations. However. our experience has
indicated that the temperature control and the difficulty of
repairing or replacing burned out elements would make this
approach less desirable.
Control
The entire quantity of refrigeration produced by the

Norelco machine is removed as product with but three
exceptions. These are losses at the warm end of the heat
exchanger. heat leak, and heat content of the water and
carbon dioxide frozen out on the heat exchanger surfaces.
In order to maintain constant conditions in the unit,

the liquid nitrogen produced in the condenser must be split
between reflux and product. This split is reflected in the
temperature gradient in the heat exchanger. For this reason,
the temperature gradient is used for the control of the unit.
The product is withdrawn from the condenser through

a manually set needle valve in series with a solenoid valve.
The needle valve is set to withdraw the product at approxi-
mately the correct rate. The solenoid valve is actuated by
the temperature at the control point in the heat exchanger
through a thermocouple circuit and closes periodically to
maintain the refrigeration balance of the unit. Control of the
heat exchanger temperature continuously within ± 10 0 F has
been possible by this means.
We are investigating the use of a continuously

modulating type of control valve for product withdrawal. At
present such a valve would have to be specially built for this
application. Since the scheme with an inexpensive solenoid
valve appears to be quite satisfactory, the cost of a custom
built modulating valve may not be justifiable.
-111-
Further Development
We are currently studying several auxiliaries for

use with the basic liquid nitrogen unit. These would be used
for continuously dehumidifying the air before it is fed to the
heat exchanger. There are three advantages to such an
arrangement. First, an appreciable extension in operating
time is possible, particularly in warm humid areas, since
the limitation on operating time is largely the plugging of
the inlet heat exchanger with moisture and carbon dioxide.
Second, an improvement of the thermodynamic characteristics
of the heat exchanger is possible. The size and weight of
the exchanger might be decreased. This would result in a
shorter start-up and thaw-out time, as well as reduced
initial cost. Finally, the liquid nitrogen capacity of the unit
would be increased as much as 1/2 liter per hour in areas
where the humidity of entering air is high.
In addition, we are considering alternating adsorbent

beds in which the adsorbent is continuously being regenerated
by the dry heated waste gases. Such a system shows
considerable promise for the removal of water and some
carbon dioxide. Use of a small auxiliary Freon forecooling
unit, merely to freeze out the water, also is being studied.
It should be noted, however, that the extent of auxiliary
forecooling is limited by the minimum refrigeration available
from the column reflux. It has been determined that cooling
of the feed air below about OOF would not be economical.
Commercial Unit
A perspective sketch of the prototype commercial unit

in shown in Figure 5. The design has aimed at a compact
unit, easily acce ssible for inspection of all the major parts.
Withdrawal of liquid nitrogen into 25 liter metal dewars by
means of a swinging withdrawal pipe, is planned, at least
for the present.
A method for conveniently producing up to 4 liters per

hour of high purity liquid nitrogen has thus been developed.
-112-
Acknowledgement
The authors wish to express their thanks to the

North American Philips Company for their encouragement
and support of the development of this liquid nitrogen unit.
Discussion
Question by D. Aronson, Worthington Corporation: In

considering precooling of the feed stream for water removal,
would there be justification for using the cool air to remove
some of the heat to be rejected from the Norelco refriger-
ator?
Answer by author: Because of the extremely low tempera-
ture at which refrigeration is produced in this unit, about
7 HP of heat must be removed from the warm end of the
Norelco Refrigerator to produce about 1 HP of refrigeration
at liquid air conditions.
To effect a given increase in capacity, a considerably greater

amount of external refrigeration is required when cooling the
heat sink than when precooling the entering air. Therefore,
use of the cool air to remove the heat rejected from the
Norelco refrigerator would be less economical than intro-
ducing the cool air directly to the unit.
Use of external refrigeration to cool the water before it

enters the Norelco unit does not seem warranted except in
very special instances where the increase in capacity is
needed and cannot be obtained by less expensive means.
Question by R. Katzen, Air Products, Inc.: What were heat

transfer coefficients and operating aT's for double tube and
finned tube heat exchangers?
Answer by Author. The operating ~ T' s were on the order of

SooF for the log-mean aT. The approach at the warm end
was about SOF and at the cold end about lSoF for both
exchangers. The large value for the logarithmic mean
temperature difference resulted from the fact that part of the
heat recovered from the waste stream was latent heat and
was thus released at essentially liquid air temperature.
-113-
C-2
ECONOMIC AND OTHER ASPECTS OF THE DISTRIBUTION

OF NAVY HELIUM IN LIQUID FORM
R. E. Brown*
Bureau of Aeronautics
Department of the Navy
Washington, D. C.
Introduction
Helium is now shipped compressed in steel cylinders.

The ratio of helium weight to total loaded weight ranges from
around one per cent of the total in the case of tank cars to a
little over two per cent in the case of the best standard
cylinders. Therefore most of the transportation charges are
for transporting steel, and a small portion is for transporting
helium. This is economically a very unsatisfactory condition,
but until now there has been no remedy. The primary object
of the liquid helium program herein discussed is to increase
the above stated ratio so that a much greater portion of tbe
transportation dollar will be paid for helium transportation.
A secondary and up to now entirely incidental objective is
to put the Navy into a position to furnish, more economically,
helium for end use in liquid form.
Basic Physical Facts
Safety valve closure of liquid helium container vents

to prevent escape of vaporized liquid until a substantial
pressure has built up prevents loss of gas during long
shipping periods, without reliquefaction en route and usually
without replenishment of the shielding gas. Such closure
makes possible shipping liquid helium in containers of any
size, by rail, highway, or air, for distances such as those
within the continental United States, without evaporation loss
of helium during transit and for a reasonable time after
* The views, opinions, and recommendations presented

herein are those of the author and not necessarily those of
the Bureau of Aeronautics or the Department of the Navy.
-114-
arrival, without reliquefaction of evaporated liquid. Such
closure also makes possible such no-loss shipment possibly
by water, but certainly by air, from the United States to the
farthest overseas locations. Liquid heliwn so transported
can be kept indefinitely after arrival if reliquefaction of
evaporated helium is eventually applied when the pres sure in
the container due to evaporation closely approaches the
safety device relief pressure.
The reliquefaction mentioned above will not be

possible at all helium-using locations, and if liquid helium
supply to these is considered desirable, the helium will have
to be used before the relief pressure is reached or an
evaporation loss will be incurred thereafter. The evaporation
rate before the relief pressure is reached, the evaporation
loss rate thereafter, and the reliquefaction rate necessary to
prevent los s, is not expected to exceed one per cent of the
container capacity per day.
Helium liquefiers with a daily liquefaction capacity

as great as 300,000 cubic feet are now under consideration
and considered practicable.
Basic Transportation Cost Principles
Freight classifications for liquid helium for all forms

of transportation have not yet been obtained, but it is expected
that the clas sifications will follow a pr inciple such that the
freight charges per unit weight of package freight containers
of liquid helium will be about the same as. but not more than,
are now paid on compressed helium in the corresponding type
of package, and such that the price paid for hauling a tank car
or semi-trailer of liquid heliwn between two particular points
will be about the same as, but not more than, is now paid on
compressed helium in the corresponding type of vehicle. This
means that, if the ratio of helium weight to gross shipping
weight is x-times as great for liquid helium as for compressed
helium, the transportation cost for the same shipment will be
1 Ix as much in liquid form as in compres sed form. This
program is based on this simple fact, and the object of the
program is to obtain all possible benefit from this fact.
-llS-
General Economic Principles
In so far as savings in transportation costs are

concerned liquid helium transportation should be adopted
only if the total cost of such transportation and the cost of
liquefaction is less than the present transportation cost of
compressed helium. In making this comparison the trans-
portation cost should include, in each case, on a cost per
unit volume basis, maintenance, depreciation, interest on
the investment, and any other applicable overhead costs of
charges of the transportation equipment involved. In the case
of liquid helium, the same must be considered for the
liquefaction equipment and the shielding gas manufacturing
equipment used.
In terms of costs per 1, 000 cubic feet of helium, the

cost of transportation of compressed helium in cylinders from
its Texas source to Lakehurst, N. J., a principal use point,
is about ten times the liquefaction cost. The corresponding
cost of transportation of compressed helium in tank cars is
about twice the liquefaction cost. In either case about five
times as much helium could be carried with the same shipping
weight in liquid form as in compressed form.
The foregoing brings out some interesting facts. If

the present shipment is in cylinders for the considerable
distance mentioned, it is easy to obtain a very great saving
by liquefaction and it is not particularly necessary to strive
for a very low cost of liquefaction in order to obtain this
great saving. In the case assumed, compressed helium tank
cars have already saved four - fifths of the cost of compressed
cylinder helium transportation, and any further saving with
Liquid helium is very difficult. To obtain any further such
saving, great effort must be made to liquefy at the lowest
possible cost.
The above illustration applies to the long distance

transportation which is typical of most Navy helium. The
next illustration will assume the opposite extreme of very
short distance transportation. It is assumed that the cost
of compressed cylinder helium transportation per 1, 000 cubic
feet is five times the liquefaction cost. The corresponding
cost of compressed tank car helium transportation is equal
to the cost of liquefaction. The other assumptions of the
-116-
previous illustration apply. By the same processes, in the
case of compressed cylinder helium and its liquid counterpart,
with liquid helium the cost of transportation plus that of
liquefaction would be twice the liquefaction cost, in compari-
son with five times the liquefaction cost if shipped com-
pressed. Three-fifths of the former shipping cost is saved.
On the other hand, in the case of compressed tank car
helium and its liquid counterpart, with liquid helium the cost
of transportation plus that of liquefaction would be 1. 2 times
the cost of liquefaction, in comparison with equal to the cost
of liquefaction if shipped in compressed form. In this case
liquefaction might be worth while for cylinder helium, but
would not be as economical as compressed helium for tank
car helium.
The above brings out the fundamental principle that

the advantages of liquid helium increase and decrease with
distance transported, and that below a certain minimum
distance, greater for tank cars than for cylinders, compressed
helium shipment is more economical than liquid helium ship-
ment.
These illustrations are based on rail transportation

in the United States. The economics of air transportation have
not been thoroughly investigated. However, a few general
remarks can be safely made. In the case of tank car trans-
portation the cost of transportation is, at most, about equal
to the cost of liquefaction, and is usually less. In air
transportation as so far studied the conditions are very much
the opposite of this. For the long distance air transportation
in which the Navy is mainly interested, the cost of compressed
cylinder helium transportation is likely to be around 25 times
as great as the cost of liquefaction. By the processes used
above the cost of transportation of liquid helium plus the cost
of liquefaction would be around six times the cost of lique-
faction. Even with this great reduction this cost for liquid
helium is very much greater than the very low cost of water
transportation of compres sed helium in cylinders. The
conclusion is that air transportation of liquid helium is of no
interest for economizing in transportation costs. If, however,
air transportation is needed for strategic reasons, to get
helium to its destination quickly, the comparison is between
air transportation of compressed helium and air transportation
of liquid helium, and the latter is likely to be only about one-
fifth to one-quarter as costly as the former.
-117-
Strategic Advantages
Using the same means of transportation, liquid helium

will save about five-sixths of the present container material,
in so far as it is applied. Using semi-trailer s instead of
tank cars will still further increase material savings, due to
faster travel. Support of overseas operations by air trans-
portation of liquid helium instead of water transportation of
compres sed helium would save a phenomenal amount of
container weight.
Savings in weight would be encouraging to air

transportation and would often justify it where present ship-
ping weights do not. Quick supply by air would be very
advantageous in cases of surprise requirements at distant
locations.
Shipping one liquid helium tank car instead of five

present compressed helium cars would reduce the burden on
the railroads in that proportion. This would be of value in a
war emergency when railroads are loaded to capacity. Liquid
helium semi-trailers can relieve the railroads of the helium
tank car burden entirely at a transportation cost about the sar
as that now paid on compressed helium in tank cars.
Suggested Navy Liquid Helium Planning
The Navy uses helium for airship inflation at stations

along the coasts. Cylinder helium is used by Navy shore
stations all over the world and by most Naval vessels. Heliun
presently is transported in tank cars from the sources to the
stations along the coasts, which serve as bulk storage stationE
and which transfer from tank cars or storage to cylinders for
distribution through the Navy supply system to stations and
vessels throughout the world.
As a first application, liquid helium transportation

should be applied only where economy in transportation is
attained, leaving other advantages for later attainment. For
economy in tank car transportation, helium must be liquefied
in very large quantities, such as the entire production of a
helium plant.
-118-
During a war emergency coastal stations are called
upon for large helium shipments which might justify lique-
faction for economy in transportation if rapid shipment by
air is wanted. In times of peace, however, there is little
such bulk shipment and most shipments are in small lots
to many completely separate and uncoordinated end use
destinations. Liquefiers at these stations might be moderately
successful in attaining desired objectives during a war
emergency. Under normal conditions, though, it would be
virtually impossible to regiment enough of the many end use
activities to want liquid helium to obtain the large liquefaction
rate necessary to make liquefaction economically worth while.
For these reasons liquefaction at coastal Naval Air Stations
is not being currently recommended.
On the other hand, nearly all Navy helium is shipped

in tank car s to only a few stations, from which it is further
distributed as described. Naval Air Station, Lakehur st, N. J. ,
alone took nearly half of the production of the Exell, Texas,
helium plant prior to a recent expansion. Liquefaction at
this plant primarily for tank car shipment to Naval Air
Station, Lakehurst, and similar activities does not require
regimentation of a multitude of practically unrelated interests,
and it is considered that tank car liquid helium shipment
between the se points would yield an attractive economic gain.
It is accordingly recommended that alar ge liquefier, in
successively erected units, be erected at Exell for this
purpose. This plant will also be available to meet whatever
demand may develop for small package liquid helium ship-
ments.
By direct flight Exell is not significantly farther from

a typical overseas destination than is a coastal distribution
point. Liquid helium should not be shipped from Exell to
Lakehurst, vaporized and stored there, and then reliquefied
and shipped, as a liquid, over seas. Liquid helium should be
shipped in the final overseas containers from Exell to the
overseas destination, direct if pos sible, and with transhipment
at the coast or elsewhere if necessary, but without relique-
faction or transfer between containers enroute.
-119-
Tank Car Transportation
This is the most attractive application for the initial

venture in liquid helium transportation. A liquid helium tank
car will carry nearly six times as much helium on the same
total weight as will a present compressed helium car.
However, the corresponding helium content on which freight
is paid will, as in the present car, be less than the 40,000
pounds minimum weight on which f reig:u: is paid in both type s.
Liquid helium is very bulky in comparison with o,ther liquids.
Therefore the liquid car must be of considerably greater
dimensions (length) than the present car to utilize the same
total loaded weight and enable it to carry the above
mentioned relatively greater helium content. The present
compressed helium cars are about 40 feet long. As the first
venture a liquid helium tank car 60 feet long has been
designed, to carry 1,200,000 cubic feet, or five times the
capacity (240,000 cubic feet) of the present cars, without
utilizing the limiting weight available. An 85 foot (coupled
length) car could be used, and would have somewhat greater
capacity and give somewhac better economy.
The containers for the se cars will follow established

liquid helium practice. There will be three shells with a
liquid gas shield attached to the middle shell. The cars will
carry no refrigeration machinery. With the closed vents
already mentioned they will provide no-loss transportation to
the farthest extremes of the continuous railway system of
North Anlerica.
Helium Plant Liquefaction Facilities
Liquefaction cost decreases with increase of lique-

faction ratio. The unit liquefaction cost for several small
liquefiers operating in parallel to produce a certain total
production, though not as low as for a single liquefier of that
total capacity, is lower than for one of the small liquefiers
operating alone. This is due to the several small liquefiers
sharing supervision and other overhead items. For the first
venture the liquefaction cost plus liquid transportation cost
should not exceed the present compressed transportation cost.
This can be accomplished with a liquefier unit of capacity
sufficient to supply Naval Air Station, Lakehurst. Use of such
-120-
a unit, with an appropriate number of tank cars, is
recommended as the first venture. Additional units can be
added as found desirable.
Liquid helium storage will be necessary to permit

continuous plant operation despite irregular car arrivals.
Liquid nitrogen or liquid air will be used as the

shielding gas in shipping and for cooling containers and for
refrigeration in liquefaction. Facilities will be required for
producing this gas. Use of liquid air is currently planned.
Bulk Storage Tank Car Terminal Equipment
Liquid helium can be handled in several ways at

these stations, but it is considered best, initially at least, to
vaporize the liquid helium on arrival and store the gas in the
compressed helium storage tanks now available. An attractive
procedure is to draw the liquid helium, with a pump, from the
tank car, and pass it through a non-return check valve against
storage pressure, through a heat exchanger designed for
storage pressure, and thence as a gas into the storage, the
pressure of which would be the result of vaporization in the
heat exchanger, without use of a compressor. This would be
an excellent way to fill cylinders with Grade A helium, with-
out compressor contamination. No other equipment will be
required for these stations. initially at least.
Stations with relatively small storage capacity will

require additional capacity to accommodate the very large
capacity of a liquid helium tank car. Whether or not the
gains from liquid helium will more than offset the cost of
additional storage will have to be individually studied in each
case.
Semi-Trailer Transportation
Liquid helium semi-trailer design involves essentially

the same problems as does liquid helium tank car design. A
liquid helium semi-trailer will have about twice the capacity of a
compressed helium tank car. Since a semi-trailer can make two
trips while a tank car is making one, one liquid helium
semi-trailer will be equal to four compressed helium tank
-121-
cars, or eighty per cent of the planned liquid helium tank
car, in transportation capacity. It is believed that the
charges for hauling a semi-trailer will be about equal to
those for hauling a tank car. In that case the unit cost for
hauling liquid helium by semi-trailer would be one-half that
for compressed helium tank car transportation and two and a
half times that for liquid helium tank car transportation.
Adding the liquefaction cost in each case, including amortiza-
tion and maintenance charges of the liquefaction plant,
results, for liquid helium semi-trailers, in a total liquefaction
and transportation cost which seems to be somewhat higher
than that for liquid helium in tank cars and somewhat lower
than the transportation cost for compressed helium in tank
car s for the long trip, Exell to Lakehur st.
The cost of liquid helium semi-trailers per unit of

capaCity is probably somewhat greater than that of tank cars,
but it is believed that this is more than offset by the smaller
total semi-trailer capacity requirement resulting from the
faster turn-around of the semi-trailers. It is hardly believed
that the liquid helium semi-trailer is in a position to displace
the liquid helium tank car, on an economic basis, as the
mainstay of bulk helium transportation, but it is believed
that it comes so close to it that its superior strategic
advantages might make it worth while to handle a part of the
transportation with semi-trailers.
Containers for Overseas Bulk and Other Retail Distribution
Except for safety valve closure of vents these container.

will have no features not already well established. Shielding
gas appropriate to the transportation time will be carried.
For overseas service air transportation is expected. For this
the lightest practicable shipping weight will be emphasized.
This will probably favor large containers. For domestic
transportation of end use liquid helium smaller and relatively
heavier container s will be acceptable.
The problems of overseas destinations are fundamen-

tally the same as those of tank car terminals already
discussed. Vaporization and storage, compressed in steel
tanks, is probably the best disposition from the standpoint of
the safety of the supply, which is very important at a
distant overseas base.
-122-
Conclusion
A recommendation that all liquefaction, including

that for overseas shipment, be at a Bureau of Mines helium
plant has already been developed, and the equipment for
that plant already discussed will suffice for charging over-
seas and retail containers. As the matter is now viewed,
there is no prospect for liquefaction at coastal stations unless
the recommended liquefaction at a Bureau of Mines helium
plant should fail to materialize. In that case liquefaction at
these stations would be recommended for the immediate
purpose of supplying end use liquid helium within the United
States and on a cautious basis for economy in bulk transporta-
tion overseas. The factors governing the proper capacity of
such a plant are almost totally unknown, but it is thought
that for an initial venture a daily liquefaction capacity of
about 50, 000 cubic feet would be appropriate.
Discussion
Question by J. V. Fetterman, Air Products, Inc.: Have you

any idea of the cost of a liquid helium tank car? Have you a
design for the liquid helium tank car?
Answer by author: The first 1,200, 000 cu. ft. liquid helium
tank car will probably cost $180, 000. Later cars in lots of
three or more each will probably cost $165, 000 each. Presert
compressed helium cars, holding only one-fifth as much
(240, 000 cu. ft.) cost over $100, 000 each.
A general design for a 1,200, 000 cu. ft. liquid helium tank
car has been developed to the stage necessary for bidding
purposes and for calculating performance.
Question by O. Heustis, The Martin Company: At what

pressure is liquefied helium shipped in tank cars? What is
the cost of liquefying helium?
Answer by author: The gage pressure will range from as

near zero as possible at the start of the trip to whatever
pressure results from evaporation during the trip, within the
safety valve release pressure. The release pressure selected
-123-
is 10 atmospheres, and this is sufficient for no-loss operation
within the United States.
Liquefaction cost depends on the process used and

the various governing factors selected. As calculated in
this study the liquefaction cost per 1000 cu. ft. ranges from
$4.42 for lOa, 000 cu. ft. per day (the Naval Air Station,
Lakehurst, requirement) down to $2.89 for 300, 000 cu. ft.
per day including all plant operating and plant amortization
costs.
Question by D. R. Cheyney, Kaiser Aluminum: Since very

large tank cars are required to take advantage of liquefaction
of helium for shipment, has the use of aluminum alloys been
considered? Aluminum alloys would be suitable at these low
temperatures, and allow savings in cost and weight.
Answer by author: There is so much that is experimental

in this project that, to insure satisfactory results, each
material will be restricted to uses for which cryogenic
experience has firmly established it as the most satisfactory
material. On that basis aluminum alloy has been selected
for the nitrogen shield, carbon steel for the outer shell and
saddles, stainless steel for the inner shell and swash plates,
and copper for the cool down shield. Perhaps an aluminum
alloy could be argued for some of these other parts, but it
has not been considered best. Use of aluminum alloy in car
chassis parts has not been considered. That development is
left to projects involving less new development above the
chassis.
Question by M. P. Sweeney, Process Associates: How long

would your liquid helium tank car take before the safety valve
opened at the 10 atmosphere design point?
Answer by author: Such a car can be designed for almost any

period to safety release pressure, within reason. The longer
the period the higher the pressure, or the heavier the
insulation or other protection, the heavier the car, the less
the capacity, and the higher service for which designed.
Helium tank car transit time to Naval Air Station, Lakehurst,
New Jersey is about eleven to fourteen days, and this period
also covers travel to all four corners of the United States. The
cars are designed, with some margin, for no-loss operation
within this period and distance.
-124-
C-3
TECHNICAL ASPECTS OF LARGE SCALE LIQUID

HELIUM LIQUEFACTION AND TRANSPORTATION
D. B. Mann, B. W. Birmingham, and P. C. Vander Arend*

Boulder, Colorado
R. E. Brown has just presented some of his thoughts

on the economics of large scale liquid helium transportation
as applied to the shipment of Navy helium. CEL-NBS, as
part of a study sponsored by the Bureau of Aeronautics,
Department of the Navy, has investigated the technical
aspects of a liquefaction and transportation program of the
magnitude proposed by the previous paper.
The results of this investigation can be broken down

into four general areas of interest: liquefaction of the
helium, transfer of the liquid to transportable storage, the
transport of the liquid to the destination (i. e., Lakehurst
Naval Air Station), and finally the vaporization of the liquid
and the warming of the gas to ambient temperature at the
destination, where it is then introduced into the existing
compres sion and storage equipment.
We shall first discuss the liquefaction process.
In order to produce liquid helium cheaply, an

efficient liquefaction process should be chosen. Two basic
processes have been considered for a helium liquefaction plant
of the scale proposed. They are first, a cycle in which the
helium gas performs external work to provide the necessary
refrigeration for the liquefaction and second, a cycle in which
other liquefied gases, such as liquid nitrogen and hydrogen,
provide the necessary refrigeration for helium liquefaction.
It is possible, of course, to integrate these two basic cycles
and to use a cycle which employs both external work and
other liquefied gase s.
*Present address-Air Products, Inc., Allentown,

Pennsylvania.
-125-
Cycles with expansion engines in which the gas
performs work are in principle more efficient than cycles
in which the gas does not perform external work. A
cascade cycle with liquid nitrogen (or liquid air) and liquid
hydrogen also is somewhat less efficient, because of the
los s e s in handling of both liquid pr ecoolants.
Calculations confirmed the belief that a liquefaction

process with liquid hydrogen precooling of the helium gas
requires more electrical energy per liter of helium produced
than a process with expanders. In addition to a high energy
requirement, more equipment is necessary, and therefore
it is expected that more manpower would be required for
operation.
For economic reasons the liquefaction plant should

be located at or near a Bureau of Mines helium separation
plant such as those in West Texas, Kansas, and New Mexico.
A suitable liquefaction process for a Bureau of Mines

installation is shown in Figure 1. In this process helium gas
is compressed at room temperature to a pressure of approxi-
mately 15 atmospheres. The gas is then cooled in a series of
heat exchangers. A portion of the gas is diverted through an
o
expander which operates at approximately 20 K, is cooled by
performing external work, and provides the necessary
refrigeration for liquefaction. The balance of the high
pressure gas is further cooled and then expanded through a
Joule-Thomson expansion valve at which time a part of it is
liquefied. The non-liquefied portion joins with the gas from
the expander and returns through the heat exchangers to the
intake of the compressor.
It should be noted that another expander is used in

this liquefaction process. The use of this expander, which is
quite important to the proce s s since it provides a significant
amount of refrigeration, is made pos sible by the availability
of pure high pressure (170 atm) helium gas at the Bureau of
Mines installation. This is the feed gas to the liquefier and
in performing external work in this expander it cools to a
temperature of approximately 190 0 K. This feed stream then
enters the cycle at a pressure of 15 atmospheres.
-126-
COMPRESSOR SYSTEM
, . . - - - - VENT
EXPANDER NO. I
LIQUID AIR PRECOOLING SATH
Figure 1. Schematic flow diagram of

a helium liquef~ction cycle
'2 = 15 otm
I+d
O.03L--~'----7---6O------;:-----:';:IO'--""'1!;:2-
X
Figure 2. Helium cycle efficiency
DESIGN CRITERIA
I) THERMODYNAMIC EFFICIENCY
2) COMPONENT AVAILABILITY
3) COMPONENT RELIABILITY
4) SIMPLICITY
5) DOWN -TIME
6) DIRECT LABOR COSTS
7) INDIRECT LABOR OR OVERHEAD COSTS
B) CAPITAL INVESTMENT a MAINTENANCE
Figure 3. Design criteria
-127-
Liquid nitrogen or liquid air precooling of the high
pressure helium gas provides refrigeration at a temperature
of the liquid ' s boiling point at one atmosphere and makes it
possible to reduce the quantity of helium circulated per unit
volume of liquid produced. The use of this precoolant liquid
is somewhat in conflict with the previous statement that
elements of a cascade cycle reduce the efficiency of the
process. However. liquid nitrogen or liquid air can be
handled with relatively small losses and can also be pro-
duced fairly efficiently. Rather large quantities are required
to insulate the liquid helium during shipment in the transport
container; its use in the helium liquefier. therefore. does
not mean that the liquid precoolant is to be produced for the
helium liquefaction process only.
The temperature region below this precoolant bath is

of primary interest, as it influences the efficiency of the
cycle in respect to gas circulated per unit volume of liquid
produced. To examine this region, a heat balance is
determined and the results are as shown in Figure 2. This
graph shows the cycle efficiency (~). the quantity liquefied
a. divided Dy the quantity of gas X + I circulated by the
compressor (x + 1) versus the quantity of gas that flows
through the low temperature expander. The other variables
are the expander efficiency and the expander exhaust
temperature. We can see from this plot that for the optimum
conditions. a 60 per cent efficient expander should pass about
two-thirds of the total compressor flow. For any size cycle
we can then determine the compres sor size and the cycle
power requirement.
The power requirement, a reflection of the cycle

efficiency. is important, but is only one of several points
that must be considered when determining the overall cost of
operation.
Figure 3 shows the other criteria established for

design purposes. To procure a large scale plant it is
desirable to have all components available and reliable. For
the cycle just described. replies to exploratory inquiries
indicate that all components are available and in most cases
are shelf items. Expanders suitable for use at 20 0 K have
been developed and are available for liquefiers of sizes
-128-
contemplated in this study. The cycle is extremely simple
and should require little maintenance and downtime. It
should be pointed out here that the cycle can be made more
efficient by the addition of expanders but it was felt that
the added complexity is not warranted to achieve a decrease
in the power cost since this cost was only a small percent-
age of the total. Direct labor costs and overhead costs are a
large item in the overall cost picture and were estimated for
the area in which the plant was to be erected. Amortization
of the capital investment, which includes depreciation and
interest, was also charged to the process and included in the
cost of liquefaction. Initial maintenance costs were estimated
to be somewhat higher than in a well established production
process. The cycle described is one that, when evaluated
using these design criteria, gave a low production cost and
still required a relatively low initial capital investment. The
total estimated cost to condense helium showed that the
project is economically feasible for the application West
Texas to Lakehurst.
Efficient transfer of the liquid to transportable con-

tainers can be accomplished using liquid nitrogen or liquid
air cooled transfer lines.
A problem arises when cooling the container using

cooldown procedures for cryogenic tanks. Normal practice is
to purge a container by pouring the liquefied gas into it. The
liquefied gas evaporates immediately upon contact with the
warm walls of the container and' the cold gas leaves the
container through suitable vents. However, the low heat of
vaporization of helium makes this practice extremely wasteful.
Therefore it is planned to employ a technique well known to
low temperature research workers and utilize the cold gas that
has just been vaporized in the cooling of the tank. An
arrangement that will accomplish this is shown in Figure 4.
To conserve liquid helium the helium container is first cooled
with liquid air to liquid air temperature and then purged with
helium gas. Liquid helium flows through the fill line to the
bottom of the container where it immediately vaporizes. The
gas formed is channeled between the cooldown shield and the
wall of the container, thus cooling the container efficiently.
When cooled to liquid helium temperature the vent line used
during cooldown is then valved off and the normal vent line is
used for the remainder of the filling operation. This arrange-
ment when used on large containers has the effect of reducing
the liquid helium required for cooldown by a factor of ten.
-129-
LIQUID FILL
FILL VENT COOL DOWN

VENT
SHIELD
Figure 4. Cross section of helium container
10
'"'"
II:
'"
X .
...
:l;
o
'"
I-
'!;
'"
II:
HEAT INPUT:: 10 WATTS
~ 4
'"g:
5 ~ 15 W ~
TIME AFTER FILL IN CAYS
Figure 5. Transport time of isolated 16,320

liter liquid helium container
Figure 6. Liquid helium tank car
-130-
The actual large scale transport of liquid helium
will be accomplished in tank cars. The thermodynamic
properties of helium near the normal boiling point make it
possible to seal off the transport container during the entire
time of shipment. The pressure will rise in the container
because of the heat transfer from the atmosphere, but the
final pressure can be kept within reasonable limits using
existing insulation techniques. Figure 5 shows a plot of
pressure rise versus transport time for a 16,320 liter
container filled to various levels. It is estimated that a
one way trip from West Texas to Lakehurst, New Jersey,
will require a minimum of 12 days. At the end of this
period the pressure in a 90 per cent full conto.iner will be
about 6 atmospheres. If the container is designed for a
pressure of 10 atmospheres, there would be a factor of
safety of almost two to take into account delays en route.
The proposed tank car is shown in Figure 6. The

helium storage consists of three separate containers to
minimize losses should failure occur. The three tanks are
mounted on a 60 foot flat car. Longer car s can be used, but
they would possibly require special handling.
After delivery to the destination, the helium can be

transferred from the car using either of two general methods.
If the 10 atmosphere pressure available in the container is
used, the helium could be forced through a suitable heat
exchanger and warmed to ambient temperature and then piped
to existing compression equipment. Otherwise, a low
temperature liquid helium pump could force the helium through
a suitable heat exchanger directly to high pressure storage.
In any case, pure helium gas in quantities of 1,200,000 SCF
per tank car could be delivered to the destination.
These are only some of the highlights of the study of

large scale liquid helium production. Technically, the process
plant could be built at the present time using presently avail-
able equipment. Economically, the liquefaction and shipment
of helium as a liquid for end use as a gas will depend on the
quantity desired by the user and the distance from the helium
liquefaction plant.
-131-
Discussion
Question by L. Garwin, Kerr-McGee Oil Industries, Inc.:

Would you care to comment on the design features of the
second expander, the one which operates at about 20 0 K?
Answer by author: For detailed design features it is

suggested that the manufacturers of this equipment be
co ntacted.
Question by J. 1. Nestel, Consolidated Airborne Systems,

Inc.: What type of volumetric or mass weight measurement
system has been considered for the liquid helium vessel
being discussed? In the case of the liquid level capacitance
type, will it read volume or weight of liquid helium?
Answer by author:· Liquid level measurement based on

hydrostatic principles has been considered during the dewar
filling operation. If the dewar is promptly sealed off, the
liquid level measurement has no further significance since
the contents will soon be a single phase. Except near the
critical point, the capacitance measurement of the contents
is approximately proportional to the length of the liquid
column and the density of the liquid. In cases where the
density varies this must be taken into consideration in the
instrument calibration.
Question by J. V. Fetterman, Air Products, Inc.: Can you

comment on the cost and performance of liquid helium transfer
lines?
Answer by author: Transfer lines of the type contemplated

here should be quite satisfactoJ:'Y. Development lines of this
type have cost up to $100 per foot. However, it is hoped that
in their present state of development there would be a sub-
stantial cost reduction for these lines.
Question by W. Blackman, United Aircraft Corporation: What

material is used in the construction of the tank containing
helium?
-132-
Answer by author: Several materials are suitable.
Examples are stainless steel types 304L, 316, 321 and 347.
Aluminum or copper are also pos sibilities.
Question by G. H. Caine, Pesco Products, Division of Borg

Warner: For the three liquid helium containers is one liquid
nitrogen source used for cold gas shielding, or does each
container have its own liquid nitrogen shield?
Answer by author: Each liquid helium container has its own

liquid nitrogen shield.
Question by G. E. McIntosh, Beech Aircraft Corporation:

With regards to the second expansion engine, how long will
this engine run without stopping for repair or replacement"i
Answer by author: This time is not precisely known,

although engines of a design similar to that needed in this
application have reportedly operated in excess of 1000 hours.
Question by J. D. Sudbury, Continental Oil Company: What is

the insulation material used in tank car design?
Answer by author: Evacuated perlite is used.
Question by D. N. Lyon, University of California Radiation

Laboratory: Has the thermal oscillation phenomenon, so
commonly observed in laboratory liquid helium vessels which
are closed off, which increases the heat leak by enormous
amounts been considered in estimating the heat leak during
transport?
Answer by author: It has not been considered because we don~t

expect it to OCCur. This phenomenon has never been observed,
to our knowledge, in pipes of the size we would expect to use
on the large transport dewar s. If this phenomenon proves to
be a problem we feel that it can be solved by using various
methods. For instance, cold valves located next to the liquid
helium shell should eliminate the trouble.
-133-
Question by M. A. Dubs, Linde Company: What is the
relative volume of liquid nitrogen carried on the 12 day trip
and 90 per cent full container?
Answer by author: 1200 liters of liquid nitrogen are used

to shield 16,320 liters of liquid helium.
Question by M. A. Dubs, Linde Company: Does the feed gas

all enter through the first expander? What is the pressure
of the feed gas?
Answer by author: Yes, it does. The pressure of the feed

gas is 170 atmospheres.
Question by D. B. Crawford, Air Products: Can you explain

the change in rate of pressure rise with time,! For example
it rises slowly at first but then rises rapidly later on.
Answer by author: The process is one of heating the

contents of a constant volume container. The change of slope
occurs when the contents change from liquid and vapor phase
(a mixture) to a single phase.
Comment by R. B. Scott, CELNational Bureau of Standards:

It is my understanding that the break in the pressure time
curve occurs when the liquid in the vessel expands until it
completely fills the container.
Question by V. J. Johnson, CEL National Bureau of Stand-

ards: What other considerations were involved in choosing
three vessels on the tank car instead of two or one? Does
a single large tank have structural or expansion problems?
Answer by author: The choice of three vessels. mounted on

the flat car was an arbitrary one. Multi-vessels are used
in an effort to minimize loss in the event of an accident.
-134-
Question by L. Lazare, AInerican Cyanamid Company:
Liquid helium transported will probably arrive at its restina-
tion in a supercritical dense phase. Are pumps contem-
plated, able to handle this dense phase material as well as
liquid helium?
Answer by author: Yes, that is the hope. Other methods

of handling the helium at the destination have been considered
and are mentioned in the paper.
Question by R. Katzen, Air Products, Inc.: What is the

effect of impurities in the feed gas on the liquefaction
system?
Answer by author: Because the feed gas is delivered

directly from the helium production facility its purity can
be carefully controlled. The expected impurities based on
Bureau 'of Mines figures will be water vapor, 10- 3 to 10- 4
mol per cent and hydrogen, less than 10- 3 mol per cent.
Of course, these impurities could possibly be concentrated
over a period of time by freezing out in a low temperature
region of the liquefier and be rather troublesome in respect
to heat exchanger and expander performance. Present
thinking on this problem involves the placing of regenerator
traps in the cycle which will accumulate the frozen impurities.
The operation of these traps will be such as to allow warm up
in the case of the hydrogen or blowdown in the case of water
vapor impurities.
Question by M. P. Sweeney, Proces s Associates: What is

the cost of liquid helium?
Answer by author: For the size plants considered in our

studies, we feel that 1000 standard cubic feet of helium can
be liquefied for somewhat less than 5 dollars.
Question by L. B. Nesbitt, General Electric Company: What

is the plant size associated with the 5 dollars per 1000 cubic feet
of helium?
Answer by author: Liquid helium can be delivered to storage at
the rate of lOa, 000 standard cubic feet per day for somewhat less
than 5 dollars per 1000 cubic feet of helium.
-135-
C-4
DILUTION OF CRYOGENIC LIQUID ROCKET PROPELLANTS

DURING PRESSURIZED TRANSFER
S. Greenfield
Rocketdyne
A Division of North American Aviation
Canoga Park, California
Introduction
In the testing of rocket engines and engine compo-

nents, such as thrust chambers and gas generators, the
rocket engineer is regularly faced with the job of transferring
cryogenic liquid propellants, using pressure feed from a
storage tank, through a flow system and into the combustion
device to be tested. At Rocketdyne, liquid oxygen is quite
commonly used as a propellant. In the testing of engine
components inert pressurants are usually used, for example,
helium or nitrogen. However, because of the cost and
logistics, nitrogen is preferred. When gaseous nitrogen is
used as the pressurant, we have repeatedly run into testing
conditions which have resulted in condensation of nitrogen and
subsequent mixing in the tanked liquid oxygen. This dilution
of the liquid oxygen effects a loss of performahce and
imposes mechanical difficulty in maintaining a regulated
propellant tank pressure.
This paper will attempt to describe the conditions

affecting the dilution of liquid oxygen with nitrogen and to
present a program suggested at Rocketdyne to evaluate the
parameters, and if possible, eliminate the causes of this
dilution, without necessitating a substitution for nitrogen as
the pres surant.
Rocket Test Equipment
Before discussing the conditions under which dilution

of liquid oxygen occurs, I would like to acquaint those of you
who are unfamiliar with the test equipment which we use for
-136-
the evaluation of rocket engine components. In Figure I,
a photograph of a typical small thrust rocket research test
stand is shown during a firing of a thrust chamber. This
test stand uses pressurized feed systems shown schematically
in Figure 2. In contrast to this research test stand, which
is designed to accommodate thrusts of less than 10,000
pounds, Figure 3 shows a test firing on a typical large
rocket engine component test stand. The thrust chamber
shown here develops well over 75,000 pounds of thrust.
Again, this system for large engine component testing
ut ilizes pressurized feeding of both the oxidizer and the
fuel. To a lesser degree liquid nitrogen dilution can occur
during the tests of complete rocket engine systems. Here
the propellants are fed to the thrust chamber by the use of
a set of turbine driven pumps, one for each of the oxidizer
and fuel. A slight pressurization of the tanked propellants
in the missile airframe is required in order to provide
sufficient suction head to the pumps. But in this case, the
pressures are low enough so that a problem with nitrogen
dilution does not usually exist.
Effects of Dilution on Performance
Condensation of gaseous nitrogen over liquid oxygen

is usually evidenced during rocket component testing by a rise
in the temperature of the delivery liquid oxygen and a discrete
drop-off in liquid oxygen tank pressure. When a thrust
chamber is being tested, we usually experience a decrease
in thrust and chamber pressure. During tests of gas generator
combustors, which are used to supply the driving fluid for the
engine turbopumps, we find that the temperature and corre-
spondingly the available energy of the gases decreases.
In Figure 4 a curve is presented which shows the

anticipated loss in thrust chamber specific impulse as a result
of liquid nitrogen dilution of the liquid oxygen. It can be
seen that for the propellant combination, oxygen and JP-4, at a
mixture ratio of oxidizer to fuel of 2.35, that a dilution of the
oxygen to the extent of 10 per cent by weight, results in a
seven second decrease in performance for the propellant
combination. For oxygen and 92.5 per cent ethyl alcohol at a
mixture ratio of 1. 4, a similar decrease in performance is
calculated. This loss in specific impulse performance, on the
surface, does not seem to be significant. However, when
-137-
--~I-r(~---.L...---{><h-!>:r-- ..Iff
UOUID fut:L
OXYG(N
"'Hie
'AN.
flU. --!>c:)----i I---{><l-- mL
IL". --!>:r--I
Figure 1. Typical small thrust

rocket stand
COOLANT IN
COOI..oVIT OUT
Figure 2. Typical research

test stand
-r--------,---------.------~
Figure 3. Typical large thrust

chamber firing
~~~-~~~~~~~~--±~-~-,-~
f'£ItaHT NIl'RC\ICIPI tH OJ(IDlUA
Figure 4. Effect of nitrogen dilution

of oxygen on specific impulse
-138-
translated into the effect on payload and range of a missile,
we find that such small losses in specific impulse are most
important. As an example, it can be calculated that a one
per cent decrease in specific impulse of a typical ballistic
missile can result in a nine per cent decrease in payload,
or a six per cent decrease in range.
In addition to the problem of loss in engine perform-

ance due to liquid nitrogen dilution, we find that from the
test ~tandpoint, dilution will cause inaccurate metering of the
liquid oxygen and consequently inaccurate measurements of
specific impulse. Furthermore, the difficulty mentioned
previously in maintaining tank pres sure can result in the
negation of a very expensive rocket test. It can be seen
that the elimination or drastic reduction of liquid nitrogen
dilution could constitute a very urgent problem in the rocket
engine test field.
Solving the Dilution Problem
There are, of course, several alternates to the use

of gaseous nitrogen for pressurizing liquid oxygen. One
which is most prevalently used is substitution of the nitrogen
with helium ~s. For test stand use, when large volumes of
pressurizing gas are needed, helium has become a logistic
and cost problem.
During the early stages of planning for Rocketdyne's

test facilities, gaseous oxygen was considered for pressuriza-
tion of liquid oxygen. At that time the use of high pressure
(up to 3000 psi) gaseous oxygen was considered an unnecessary
hazard compared to the use of gaseous nitrogen. Furthermore,
it was not felt that the problem of liquid nitrogen condensation
in the liquid oxygen would be a significant problem. In fact,
the dilution problem did not appear until the testing of rocket
engine components had advanced to a point where higher
pressures and larger sized units were being used. Under
these conditions, condensation of nitrogen and subsequent
dilution became a noticeable and significant factor.
Mixtures of gaseous nitrogen and helium have been

found to reduce the condensation and dilution of the liquid
oxygen. However, the use of such gas mixtures represents
only a partial cure for the problem.
-139-
Another scheme, which has been suggested and used,
involves the use of helium for the initial pressurization of
the liquid oxygen to the desired tank pressure level and then
switching to gaseous nitrogen for pressure feeding during
the test run. This procedure reduces the extent of the
condensation and dilution. However, it involves a complica-
tion of the pressure feed system to a point where the gains
are not economical.
The most satisfactory solution to the problem of

dilution was felt to lie in a program which would clearly
define the significant parameters affecting dilution, and
with this knowledge to reduce the extent of dilution to an
acceptable minimum.
During the early months of 1957 the Rocketdyne

Research Section began an analytical and experimental
program to qualitatively, or semi-quantitatively, define the
significant parameters affecting the dilution of liquid oxygen.
To conduct this program, experience was required in the
techniques for accurately sampling and analyzing the tanked
liquid oxygen or liquid oxygen flowing in a feed system to a
combustion device. Through such a program we hoped to
establish a firm foundation from which we could build a more
comprehensive and quantitative solution to our problem. The
results which are presented represent an initial phase of the
overall program. Unfortunately, time and funds were not
available to pursue the program to its end.
Evaluation of Pressurant Diffusers
The physical nature of the condensation of pressurant

nitrogen over a volume of liquid oxygen at atmospheric
pressure is quite straight forward. As the nitrogen pressure
level within the tank is increased, the normal boiling point
of nitrogen similarly increases until it is equal to or higher
than the normal boiling point of the liquid oxygen and the
temperature of the metal surfaces of the tank. Under these
conditions the nitrogen gas can condense. If the system is
physically agitated, the condensed nitrogen will then mix with
the liquid oxygen resulting in a physical and chemical
dilution. As the tank is emptied. additional condensation can
occur at the walls of the tank. although this will be limited
-140-
by the heat capacity and conductivity of the metal walls.
On emptying the tank, such effects as sloshing or vortexing
of the fluids will naturally enhance the condensation and
dilution.
The next figure, Figure 5, shows a compilation of

the parameters which we have considered as affecting the
condensation of nitrogen when it is used as a pressurant for
transferring liquid oxygen. It was immediately apparent to
us that the majority of these parameters could not be
investigated with the time and money allotted. The feature
which we felt could most inexpensively be attacked was the
mechanical entry of the pressurizing gas into the liquid
oxygen tank. Experience has shown that the position and
nature of the inlet line for the pressurant gas was most
important. If the pressurizing gas is permitted to enter
the tank directly from the inlet line and impinges freely with
great force on the surface of the liquid oxygen, we find
that, generally, a condition of condensation and dilution
results. Mechanical diffusers were designed which consisted
of a length of tubing attached to the inlet line within the
liquid oxygen tank. This tube is pierced on its circumference
with several hole s whose area is at least equal to or greater
than the cross-section of the inlet line. This diffuser causes
the entry of the pressurizing gas to flow radially from the
inlet line over the top of the liquid oxygen tank and on to
the surface of the liquid oxygen. Figure 6 shows a photo-
graph of a typical drilled tube diffuser.
Tests to evaluate the efficiency of mechanical

diffusers for reducing diffusion with liquid nitrogen were con-
ducted in the equipment shown schematically in Figure 7. In
effect we have duplicated a liquid oxygen flow system typical
of those which would be found on a rocket test stand. The
liquid oxygen effluent line is branched into a sampling system.
The sampled liquid oxygen which is suspected of containing
a diluent of nitrogen is continuously bled through a heat
exchanger. Here the liquid is converted to gas, and, at the
appropriate time, one or more of the sampled bottles can
be opened remotely to receive a gas sample to a prescribed
pressure level within the sample flask.
After a test run is completed and the sample flasks

are filled, they can be detached and then taken to a chemistry
laboratory where a gas chromatographic analysis is made.
-141-
PARAMETERS AFFECTING NITROGEN DILUTION OF OXYGEN
OXYGEN TANI< PRESSURE
RATE OF PRESSURIZATION
HOLD TIME AT PRESSURE
OXYGEN TANI< TEMPERATURE
ULLAGE VOLUME
NITROGEN INLET CONDITIONS
VIBRATION
SLOSHING
VORTEXING
OXYGEN TANK GEOMETRY
Figure 6. Photo of a drilled

tube diffuser
Figure 5.
. .
I.ClIiI " .... --O'<r---...L..--l!~i>'<J-..,.--:....:;=----------rrn--
~ ~-""""'"
10
\ 1IIIOCllffl.e{.1
I
Figure 7. Test equipment for

nitrogen dilution of oxygen
0
10 to "- )0
TM: "'-S!UI'IIZAnc)H (S(.CI

40 .
Itt &AL 01l"I'IOI' ".,.,
1'4110 ,... taM . . -...v
Figure 8. Average per cent nitrogen

in oxygen vs. time of pres surization
-142-
The gas chromatography analysis involves passing the gas
through a column filled with an adsorbing material consisting
of small, closely packed solid particles. As the sample
passes through the solid packing, each component is adsorbed
to the packing preferentially. The column is then desorbed
of the sample, and, since the oxygen and nitrogen were
adsorbed at different rates, they will be desorbed in the
reverse order. As the gases leave the column, the thermal
conductivity is measured and recorded. In this manner, the
amount of each component present in the sample can be
determined to an accuracy and precision of ±. 1 per cent.
Using the equipment shown in Figure 7, a series of

tests was conducted using the drilled deflector tube described
earlier. The liquid oxygen tank was filled to the same level
for each test. This level was maintained below the end of
the deflector tube and amounted to an ullage space of about
10 per cent in the 25 gallon liquid oxygen tank. Gaseous
nitrogen was admitted to the tank through the regulator at a
rate which was linear with respect to time until a pressure of
300 psig was obtained. After waiting about two minutes,
the time usually consumed during a rocket test for making
final adjustments, the oxygen was removed from the tank at
about 10 pounds per second. In Figure 8 we have plotted the
maximum per cent of nitrogen found in the effluent liquid
oxygen from the lower half of the tanked liquid oxygen versus
the time of pressurization. Note that with the standard or
drilled tube diffuser the amount of nitrogen diluent in the
liquid oxygen increases very rapidly as the time of pressuriza-
tion to a 300 psi tank pressure is decreased below 20 seconds.
In the same figure a single point is shown which represents
the amount of dilution which occurs when the standard diffuser
is removed and the nitrogen gas enters the tank directly from
the inlet line. In this case the dilution at a 48 second
pressurization time corresponds to the dilution obtained using
the standard diffuser with a pressurization time of 16 seconds.
A design of a mechanical diffuser which would pro-

vide even greater mechanical dispersion of the entering
pressurant gas was sought. A search of commercial vendors
of metallic filter elements was made. A sintered metallic
mesh was found, under the trade name of Poroloy*, which
could be manufactured in the desired shape for a diffuser
and could be fabricated from stainless steel. In Figure 9, a
*Poroloy Equipment Company, Van Nuys, California.
-143-
Figure 9. Photo of Poroloy diffuser
element
OJ .. os
~{. .i / / 1/ /
.~~.;> I V V
I---- f- ,. "~4. V
I, /
V
7
/ I)
r-~< I..'~ v e' I 1/ v . J
V
..
1
~~ ~/;' ~"
t="~
' #.
.
I
~
- ~~ ~ I. V
/
..oj " V II
.!'s:~. I. / /
~~
V
/
/
7 l
I
,{!<' '/1 V V V1I •
~ // )
VI
.Ii i . ' .I.. /;1-

IOPOO
~
· ~. l lillL
L*rr-
/ 1 1
V L
/ V L L Ii
I, / J
V
I V /1
AIR FLOW (Gl- 'M,I .. Mot.
Figure 10. Relation between strength,

flow, relative density and
pressure for Poroloy with 35 0
crossover angle
-144-
photograph of a typical diffuser manufactured from the
Poroloy element is shown. Figure 10 gives the relation
between strength, flow, relative density, and pressure for
Poroloy with 35 0 crossover angle. This unit is cylindrical
in shape and is about eight inches long and one and one-
fourth inches in diameter. The mesh of the metallic
portion provides for 50 per cent open area with an interstitial
opening of 10- 4 to 10- 5 square inches. At about the time
that the Poroloy diffuser elements were obtained, the nitrogen
dilution program was interrupted by more pressing problems
and, unfortunately, the results of dilution using such diffusers
cannot be reported at this time. However, since the installa-
tion of these elements have been made in the Rocketdyne
Research Test Area, we have had no incidents of nitrogen
dilution of liquid oxygen under the conditions of test with
which we are now involved. Pressurization times much less
than those reported in Figure 8 for the standard diffuser can
be obtained with the Poroloy diffuser, with the end tank pres-
sure level reaching as high as 1500 psi without qualitative
evidence of nitrogen dilution.
Incidental to the investigation of the effect of

mechanical design of pressurant diffusers, a number of
interesting observations were made. If the amount of ullage
in the liquid oxygen tank was small enough so that the level
of the oxygen was above the end of the diffuser, it was found
impos sible to avoid liquid nitrogen dilution of the oxygen.
Furthermore, it was found that the initial rate of pressuriza-
tion was the critical time during the pressurizing operation.
If the initial charge of gaseous nitrogen was allowed to enter
at a high rate, the amount of dilution was found to be
increased. On the other hand, if the first charge of nitrogen
gas was admitted slowly, the remainder of the pressurizing
operation could be conducted at a much higher rate.
Apparently, the initial charge of nitrogen gas acts as a
cushion and absorbs most of the momentum of the remaining
gas needed to achieve the desired tank pressure level.
Concluding Remarks
The test program which has been described was

conducted on research test stands using liquid oxygen tankage
which varied between 20 and 200 gallons capacity. The
problem of liquid nitrogen dilution appears to be most
prevalent in the smaller capacity containers where the ratio of
-145-
surface area of the liquid oxygen is large compared to the
volume of the tank. On the large rocket component test
stands at Rocketdyne, where liquid oxygen capacities will
run to several thousand gallons, we find that the problem of
dilution with nitrogen is not quite as prevalent. Nevertheless,
incidents have been reported which bear all of the earmarks
of nitrogen dilution characterized by an unusual increase in
liquid oxygen temperature and a corresponding decrease in
the combustion device performance. Measurements have been
made of nitrogen dilution on these large test stands and
values of less than five per cent have been recorded. The
results of our research program on mechanical diffusers
have been applied to both the large test stands and our
smaller research test stands, and the incident of nitrogen
dilution has been measurably reduced.
In conclusion, we have demonstrated that the use of

properly designed pressurant gas inlet diffusers can markedly
reduce the amount of nitrogen dilution while pressurizing
liquid oxygen. The use of metallic mesh Poroloy elements
as described in this paper has semi-quantitatively, at least,
provided an interim solution to the problem.
We have observed that the initial tanked level of

liquid oxygen must be kept below the end of the inlet diffuser;
otherwise, dilution with nitrogen is unavoidable. The initial rate
of pressurization is the most critical portion of the pressurizi~
operation. If this rate is kept to a minimum, it is possible
thereafter to pressurize at a much higher rate and still
maintain a low and acceptable amount of nitrogen dilution.
A nwnber of parameters still remain to be investi-

gated. All are believed to affect the amount of nitrogen
dilution while pressurizing liquid oxygen. It is hoped that
additional testing can be undertaken at Rocketdyne in the near
future to fully define the significance of these parameters.
The use of a continuous oxygen-nitrogen analyzer is antici-
pated in our future work. In this manner an expanded program
which will provide more quantitative results can be achieved.
We at Rocketdyne are aware of a similar program

concerned with nitrogen dilution which has been conducted at
the Lewis Flight Propulsion Laboratory of the NACA at
Cleveland, Ohio. The results of this progratn have not been
formally published, although, through informal communications
with NACA personnel, we have learned that dilutions of the
-146-
oxygen of over 50 per cent by weight were obtained under
selected conditions. It would be interesting to hear from
representatives of NACA as to the results that they feel may
be reported at this time.
Discussion
Question by D. B. Burkhardt, E. 1. du Pont de Nemours

Company: Have insertable electrical heaters been used to
vaporize liquid oxygen to obtain pres surization?
Answer by author: Insertable electrical heater s have not

been used to obtain pressurization over liquid oxygen on our
research static test stands; however, the heat exchanger
principle wherein liquid oxygen is vaporized for pressurizing
low pressure missile tankage is quite commonly used. The
control and equipment requirements for electrical heating
pressurization, I believe, would be uneconomical for our
research test use.
Question by G. C. Szego, General Electric Company: Have

you considered the application of a diaphragm or other means,
such as a float, of limiting the effective interface area
between the liquid oxygen and the nitrogen?
Answer by author: The application of a diaphragm, or float,

or other means, for physically separating the liquid oxygen
from the pressurizing gas has been considered. The high
pressure liquid oxygen tanks now in use at Rocketdyne are of
an all-welded construction except for a comparatively small
diameter flange connection at the outlet of the tank. Modifi-
cation of these tanks to accept the float or diaphragm would be
extremely expensive and time consuming. Furthermore, we
have felt that the use of such diaphragms, or floats, would
introduce a complexity which would reduce the reliability of
the flow system. To be more explicit, we envision that
float mechanizations would tend to jam and diaphragm
materials would require checking to be certain of their
integrity; otherwise, the failure rate associated with the given
test would undoubtedly go up. Floats and diaphragms would
definitely be effective in reducing the dilution problem;
however, we have not made application of such devices because
of the reasons listed above.
-147-
Development, Ltd.: Did you study the effect of the tempera-
ture of the pressurizing nitrogen?
Answer by author: We did not study the effect of the

temperature of the pressurizing gas up to the time of the
report to this Conference. It certainly would be one of the
more influencing parameters and should be investigated
during future work.
Question by M. Koshar, The Martin Company: On a test

of a complete engine assembly, what portion of a full
duration run was unaffected by dilution?
Answer by author: In this paper I have not referred to

dilution which has occurred at Rocketdyne on complete
engine tests. My remarks referred particularly to pressure-
fed static engine component testing.
Question by M. Koshar, The Martin Company: Did the

sintered filter affect pressurization system accuracy?
Answer by author: The sintered filter element did not affect

pressurization accuracy in any way. There was sufficient
open area in the element to reduce the pressure drop to an
amount which could not be measured within the precision of
our instrumentation (plus or minus one per cent).
Question by C. A. Harrison, Aerojet-General Corporation:

How soon before fire switch do you pressurize?
Answer by author: Generally during our small thrust research

test firings, the liquid oxygen and fuel tanks are pressurized
from two to five minutes before the fire switch.

Development, Ltd.: Did the figures given for nitrogen dilu-
tion during the test runs refer only to the lower layers of
liquid oxygen in the test tank?
Answer by author: Yes, the figures given for nitrogen dilution

during our test runs refer only to the lower one-half of the
contents of the tank.
-148-
C-5
LIQUID OXYGEN FOR MISSILE TESTING
W. J. Scharle
Air Products, Inc.
Allentown, Pennsylvania
Modern defense measures have necessitated the

advancement of missile testing programs from the stage of
small experimental units to an all-out race to test and
perfect full scale prototype missiles. Fuel components must
now be available in large quantity and convenient form at all
missile testing and launching stations. This prototype missi~
testing program has established the first large scale direct
usage of liquid oxygen.
Liquid oxygen is considered the workhorse of fuel

oxidizers being used for missile propulsion. It has high
energy release with most fuels, and its non-toxic products
of combustion offer no hazard to operating personnel.
Liquid oxygen production is safe, and its handling and
storage follow well proven methods. Since the only raw
material required for liquid oxygen production is air,
there are no direct raw material limitations on size and
location of producing facilities.
Liquid oxygen requirements, dating back to early

experiments on rocket engines and missiles, were supplied
from commercial sources with delivery by railroad tank car
or highway vehicle. As the United States military missile
progTam gained momentum, it soon became apparent to the U. S.
Air Force that the quantity of liquid oxygen needed was so
large that commercial producers could not curtail the liquid
oxygen sU2ply to industry sufficiently to meet the large and
unpredictalie demand of the missile testing program. Also,
the liquid oxygen would be required in specific areas
generally remote from large liquid oxygen producing stations,
some ar eas inacces sible by railroad. This necessitated an
evaluation of alternative methods of liquid oxygen supply.
The logical answer, to insure an adequate, continued

supply, and also reduce the cost of liquid oxygen, was to use
-149-
on-site liquid oxygen generators. The U. S. Air Force,
therefore, decided to install its own liquid oxygen producing
facilities at points of contemplated large consumption.
The low temperature industry was confronted with the

problem of supplying large liquid oxygen producing facilities
having the following basic requirements:
1) High cycle efficiency, with consequent low

operating cost.
2) Minimum of utilities.
3) Reliability of equipment for continuous operation.
4) Transportability.
Large liquid oxygen producing facilities which meet

the above requirements have been developed for the U. S.
Air Force. Each facility is made up of individual liquid
oxygen generating units which are considered to be a maxi-
mum size compatible with railroad and highway transporta-
bility. Each liquid oxygen generating unit produces a rated
75 tons per day of liquid oxygen along with 7.5 tons per day
of liquid nitrogen, each at 99.5 per cent purity.
Multiple liquid oxygen generating unit facilities are

presently in operation and under construction at U. S. Air
Force testing stations located in California, Florida, and
Colorado.
High Cycle Efficiency
The 75 T /D liquid oxygen generator employs a high

pressure cycle, with certain process and equipment innova-
tions designed to increase the overall cycle efficiency.,
resulting in a minimum power input for the required produc-
tion. A condensed flow diagram for the cycle is shown in
Figure 1.
Filtered air is compressed in a multi-stage compressor

to 2000- 3000 psig. Between the second and third stages of
-150-
I
I
I
I
I
,I ,
~~
CA"ION DIOXIDE
SCRUIBER
fii DIITILLATlOli "TO"DCAItION

COLUM" ADSGItIUS
Figure 1. Simplified flow diagram

75 T /D liquid oxygen generator
IAI~
1\ 1\
3000
V
~ 1\ / ~ ~v 2000
\ 1\ 16 \ 1500
~ 1'- il 1\ V IPsE~ V
1000
900
~ X 800
t-
~~g: t-
TOO
II &00
.--<>---- ~o
'00
.,/ '-II II 1I\k' ~+'vI
CASE 1- NO REHEAT ,,,,v IX rY ~ ~...'"t'./ ~
400
300
..
~ ~~ ~~
RECIPROCATING EXPANSION
ENGINE 6H' 1480 ']1 200
iii
Q.
V\ II IYV '(/11 W
«
::>
..
150
ct?,,~ 'J /
CI)
CASE n - REHEAT ... CI)
'"«
RECIPROCATING EXPANSION
ENGINE .. AH' 1260
..II!
"
~
100
90
80
Q.
TO
CENTRIFUGAL ~ 10
--
I-
EXPANDER . . AH· 420 II '0
1680
..
~ r-t 40
/~
~ ~
~
b
~I
~
0
! ~ co ..
I t- I ~ r-" 30
,
INCREASED REFRIGERATION I
I I
USING REHEAT 11.1 20

Filure Z
~'I. 14 Effect of Reheat on
1480 Expander ReCrileration "
10
Z600 . 28 ·30 32 34 31 :sa 40 42 44 4& .8 50 5200
ENTHALPY (H) BTU I MOL
Figure 2. Effect of reheat on

expander refrigeration
-151-
compression, the air is passed through a caustic scrubbing
system to remove carbon dioxide from the air stream.
Air leaving the fifth stage of the compressor passes

through a desiccant type drier system to remove moisture
from the air. After the drier, the air stream is split. One
stream passes through the main heat exchangers, being
cooled by effluent gases from the distillation column, and is
expanded through a valve to the high pressure column
pressure. The other portion of the air stream is expanded
from compressor discharge pressure to an intermediate
pressure through a reciprocating expander, warmed through
coils in the main heat exchanger, and expanded to the high
pressure column pressure in a centrifugal expander.
A double rectification column is used. Crude liquid

oxygen from the high pressure column is pas sed through
hydrocarbon adsorbers before being expanded to the low
pressure column. Liquid nitrogen product is withdrawn from
the stream which also provides liquid nitrogen reflux to the
low pressure column. Liquid oxygen is withdrawn from the
bottom of the low pressure column. Effluent nitrogen rich
gas from the low pressure column is warmed in the main
heat exchangers counter currently with incoming air and
discharged to atmosphere. A portion of the warm, effluent
gas is used to load the centrifugal expander and is then used
for reactivating the air drier, hydrocarbon adsorbers, and
reciprocating expander oil filters, since dual assemblies of
each are provided to insure continuous operation of the liquid
oxygen generator.
An important cycle refrigeration efficiency increase,

with consequent power input saving, is obtained by using the
dual reciprocating and centrifugal expander setup with inter-
mediate reheat of the air stream. This scheme can be
effectively used where air flows are relatively large, since a
centrifugal expander is more practical at the lower expansion
pressures. The refrigeration increase is due to increased
enthalpy drop resulting from the higher inlet temperature to
the centrifugal expander, and higher reciprocating expander
efficiency realized when operating at a higher exhaust
pressure due to smaller los ses in the expansion engine
ports, valves, and cylinder. This is illustrated in Figure 2,
by comparison against the conventional method of a single
reciprocating engine expanding from compressor discharge
-152-
pressure to high pressure column pressure. As seen from
Figure 2, an increase in total expander refrigeration of about
14 per cent can be realized by this scheme. Also, very
significant, is the fact that the reciprocating expansion
engine is considerably smaller in physical size because of
the higher density of the exhaust air from the engine.
Actual 75 T/D liquid oxygen generator operation has

resulted in an overall cycle efficiency of approximately 780
kw hrs/ton of liquid oxygen plus liquid nitrogen. Actual
operating costs for large quantity production rates have
been reduced to approximately 0.7 cents per pound of
delivered liquid oxygen and liquid nitrogen.
Minimum of Utilities
The 75 T /D liquid oxygen generator is designed to

operate with electrical power as its only essential utility.
It is required that these liquid oxygen generators be

suitable for operation in remote areas, often where no
cooling water is available. To accomplish this, the air
compressor is equipped with large air to air inter- and
aftercoolers, which are some of the largest in operation for
this service. A photograph of the se large air to air coolers
is shown in Figure 3.
Reliability of Equipment for Continuous Operation
Becaus e of the extreme importance for the applica-

tion of the liquid oxygen, it is important for these liquid
oxygen generators to have a high reliability factor.
Several units have been in operation for over a year

with very high onstream factors. Actual production has been
about 10 per cent over the rated capacity.
Because the use of the liquid oxygen may be cyclic

with high peak requirements and also periods of full storage,
it is very important that the generators can be taken off-
stream and put back on-stream with a minimum delay in
production. Experience has shown that approximately five
hours are required for a warm liquid oxygen generator unit
to be placed into full production. Approximately one-half
hour is required to return to production if the generator unit is
shut down cold.
-153-
Occasionally, any high pressure cycle liquid oxygen
generating unit requires defrost of the main heat exchangers.
This unit is so designed that the heat exchangers can be
defrosted independent of the main cold equipment. The total
time required to defrost the main heat exchangers, and cool
down again to normal liquid oxygen production is four hour s.
Transportability
As mentioned before, the size of the unit liquid oxygen

generators developed for the U. S. Air Force missile program
was dictated by the transportability factor. Even though the
liquid oxygen generator is large, the components, including
"factory packaged" cold boxes, are all capable of transport
by rail or low-bed trailer, which permits a minimum of
erection time and reduces erection costs.
Equipment
A photograph of a typical multiple 75 T / D liquid oxygen

generator unit facility is shown in Figure 4. The compres-
sors are installed inside a building while all other operating
equipment is outside. A brief description of major equip-
ment for a 75 T /D liquid oxygen unit follows:
Air Compressor: The air compressor is a five-

stage, 6-cylinder, horizontally opposed piston-type
compres sor, driven by a 4000 Hp electric motor, and with
air to air cooling system. A photograph of a typical
compres sor, also showing the reciprocating expander
directly coupled to the motor shaft is shown in Figure 5.
The large fan cooled inter- and aftercoolers are

mounted outside the compressor building. A photograph of
the inter- and aftercoolers is shown in Figure 3. A special
dampening system is provided to keep pulsation to a
minimum.
Reciprocating Expander: The reciprocating expansion

engine is a single cylinder. horizontal unit directly coupled
to the air compressor motor. Views of this expansion
engine are shown in Figures 5 and 6. Inlet pressure is a
maximum of 3000 psig. The isentropic efficiency of this
expander is approximately 85 per cent. The reciprocating
-154-
Figure 3. Compressor alr to air Figure 4. 75 T /D Lox generator,
inter- and aftercooler multiple unit facility
Figure 5. Air compressor, motor Figure 6. 75 T /D Lox generator,

and reciprocating expander main operating area
Figure 7. Cold box and yard Figure 8. Liquid oxygen semi-

equipment area trailer
-155-
expansion engine is directly coupled to the compressor motor
so that power is returned directly to the system.
Air Separation Cold Box: That portion of the equip-

ment operating at sub-ambient temperatures, including
distillation column, main heat exchangers, subcoolers, and
hydrocarbon adsorbers, are contained within two "factory
packaged l1 cold boxes. The cold boxes are located adjacent
to the compressor building, with the panel containing the
main control instruments extending within the building.
All major controls for the liquid oxygen unit, including

air compressor, reciprocating expander, and the air separa-
tion cold box, are confined within a small area, as shown in
Figure 6, so that operating personnel is kept to a minimum.
Yard Equipment: All accessory yard equipment,

including the air drier with reactivation equipment, centrifugal
expander, etc., are located adjacent to the air separation
cold boxes. A photograph of this area is shown in Figure 7.
The centrifugal expander is loaded by means of a

direct coupled centrifugal blower which pressurizes effluent
gases for use in regeneration of the air drier, hydrocarbon
adsorbers, and reciprocating expander oil separator, thus
eliminating the necessity for an auxiliary motor driven blower.
It is significant that no auxiliary equipment, such as

an electrical generator, is required to return either the
reciprocating or centrifugal expander power to the system.
LO X Storage Facilities: Transfer and evaporation

losses are frequently cited as disadvantages inherent to the
production and use of liquid oxygen. With the use of on-site
generators and well designed storage and transfer facilities,
losses are reduced to comparatively low rates.
The liquid oxygen and liquid nitrogen product are

stored in a group of large, stainless steel, cylindrical tanks.
These tanks are mounted on perlite concrete saddles, and are
encased within a single perlite filled, gas tight steel building.
An outside view of this storage tank is shown in Figure 4. It
is estimated that for large liquid oxygen storage facilities,
this method of construction is approximately 40 to 50 per cent
cheaper than individual vacuum type storage containers.
-156-
It is estimated, for the subject facilities, that
storage tank evaporation losses are about 1 per cent per
day. Although it may be possible to reduce the storage
tank evaporation losses to about 1/2 per cent per day by the
use of vacuum type storage containers, the use of on-site
liquid oxygen generators and the low production cost of
liquid oxygen make the payout period for the increased
storage equipment cost too long. It is also of importance
that, with the use of on-site generators, the evaporation
losses normally associated with liquid oxygen semi- trailer
travel time are eliminated.
Flow of liquid oxygen into the storage tanks and from

the tanks to the liquid .oxygen semi-trailers is by gravity,
although a pump transfer system from the tanks to the semi-
trailers is employed where geographically flat sites are
encountered. A photograph of a typical 3600 gallon semi-
trailer being filled is shown in Figure 8. Approximately
twenty-five minutes are required to fill a semi-trailer
from the main storage tank system and about 30 minutes
are required to unload a trailer into a customer's tank,
including time to hook up and unhook the trailer flexible
transfer hose.
Conclusion
The importance of liquid oxygen in this country's

missile program is firmly established. It is now being
produced in large quantities at U. S. Air Force missile
testing stations by the most modern and efficient type of
equipment available and at a cost that assures its extended
use.
Discussion
Question by M. P. Sweeney, Process Associates: What is the

efficiency of your centrifugal expander?
Answer by author: About 75 per cent adiabatic efficiency.
Question by N. Meyers, Stratos-Division of Fairchild Engine

and Airplane Corporation: What type of dessicant was used in
the dryer?
-157-
Answer by author: Activated alumina was used.

What is the approximate cost of the 75 ton liquid oxygen
installation which has been illustrated?
Answer by author: Because of the type of contract entered

with the U. S. Air Force, it would be difficult to give any
cost figures at this time that could be compared with other
liquid oxygen plant costs.

Development, Ltd.: What fraction of the high pressure air
passes through the expansion engine system?
Answer by author: About 35 to 40 per cent of the total air

feed passes through the expansion engine system.

Development, Ltd.: Are any special precautions taken to
damp out oscillations in the expander exhaust before
entering the turbine?
Answer by author: Yes, a low temperature oil adsorber is

used after the reciprocating expander which serves as a
dampener. Also, a separate surge bottle is located
immediately downstream of the reciprocating expanders.

Development, Ltd.: What is the expander exhaust pressure,
i.e. at inlet to the turbine?
Answer by author: About 250 psig.

Development. Ltd.: Does the power figure of 780 kw hr
per ton of liquid oxygen and liquid nitrogen refer to the
overall figure plus all auxiliaries?
-158-
Answer by author: Yes, this figure includes all auxiliary
power requireInents.
Question by J. T. Selldorff, Arthur D. Little, Inc.: Why

use caustic scrubbers for carbon dioxide reInoval rather
than freeze-out purification?
Answer by author: The selection of caustic scrubbers or

carbon dioxide freeze-out systeIn is an econoInic probleIn
for the particular cycle selected. Air Products, Inc.
eInploys both Inethods. The generator discussed in the
subject paper, however, eInploys caustic scrubbers.
-159-
C-6
UNATTENDED OXYGEN PLANTS
W. W. Aton
Linde Company
Tonawanda, New York
Introduction
An unattended oxygen plant is defined as one which

operates automatically and entirely without operating labor.
The building is normally locked to exclude all personnel and
the plant is only checked weekly or daily depending on the
size. Current models produce gaseous oxygen and range in
capacity from less than 10 million cubic feet per month to
more than 100 million cubic feet per month. The plants have
a safety control system which maintains safe conditions and
advises a supervisor through a unique communication system
if trouble does occur. In addition, the plant is normally
backed up by a reserve liquid oxygen unit to as sure the
customer of a continued supply during an outage. The plants
are usually located on the customer's property.
General Plant Description
Referring to Figure I, we have the essential

components for the unattended plant. In the plant proper, we
have an air compressor system supplying oil free compressed
air to the air separation unit. Oxygen is separated from the
air which has been cooled, cleaned, and liquefied. Oxygen
is compressed to pipe line pressure by a base load oxygen
compres sor which is sized to handle the customer's average
demand, although the demand may vary considerably during
any day. During periods of reduced demand, the booster
compressor pumps oxygen into storage receivers. When the
customer's demand rises above plant capacity, the receivers
will first be placed in service through the pressure reducing
valve. If this will not meet the demand, then vaporized
liquid oxygen is supplied from the liquid oxygen unit. The
liquid unit also supplies oxygen during plant outage. When
-160-
To
D , llr"'- ;6U/";DW
OT"ut U..4 P""AJTS
1.1</,,1() OJtyq4JJ $_,,'1" 8y #oJJ-'<.

lI<UCI( o~ T"""", c:;..~ /D SW·"".HIA/T
,Pt.If/tJr ~DO(jC'/~DAI ~ND 41"1G*f- v~
~"'Alr Ou.lft' AM OurAq&
/'
/
,/ '"
/
/
'"
,PD~/T;¥t:
O'.:I~4"'~4M~T
8~t"AJt:; /1QU(,
Figure 1. Unattended oxygen plant
Figure 2. Typical unattended oxygen plant
-161-
absolute continuity of supply is not critical, no liquid oxygen
unit is provided. Oxygen is billed to the customer by a
positive displacement meter having ~ 2 per cent accuracy
over a 50 to I flow range.
The plant is equipped with a "fail-safe" monitoring

system which can detect defective conditions, shut down the
plant if necessary, and transmit a signal to a commercial
monitoring service. The commercial monitoring service
advises one of a number of supervisors on the call list that
a signal has been received. The supervisor then corrects
the difficulty and returns the plant to service.
The liquid oxygen unit is supplied by liquid from a

strategically located large liquid oxygen plant. These large
mass plants also supply oxygen to many use points and have
considerable reserve stock always on hand.
The photograph of an actual plant is shown on Figure

2. The air separation unit, compressor building, gas
receivers, and billing meter are clearly visible. The liquid
oxygen unit is to the rear of the compressor building.
Safety
Recognizing the hazards as sociated with gaseous oxygEn

plants of improper design, great emphasis was placed on
safety in the engineering of the unattended plant. This plant
must be able to cope safely with unusual conditions with
complete certainty. A greater margin of safety is achieved
than in the usual attended plant. The important features
contributing to safe operation are:
1) The design was based on two years of pilot plant

operation.
2) The process equipment was engineered to remove

dangerous contaminants, such as acetylene I from
the process streams and not just control to a
supposedly safe level. This assumed a typical
steel mill atmosphere.
3) Site surveys are made for each plant in an effort

to select the best site available. If unusual air
contamination is evident and cannot be eliminated,
additional contaminant handling measures are
engineered.
-162-
4) All process equipment and controls are designed
on a "fail-safe" basis.
5) Appropriate national and local codes govern

mechanical and electrical equipment design. For
example, the ASME Code for Unfired Pressure
Vessels governs pressure vessel design.
6) Cleanliness of equipment and selection of mater-

ials is based on fifty years of experience in the
oxygen industry.
Air Separation Process
Discussion of specific process equipment is beyond

the scope of this paper. Therefore, the low temperature
process will be described using as a basic reference a
simplified flow diagram for an 80 million cubic feet per
month plant. Cycles for plants of different sizes are
modified to provide the best economy and reliability.
Referring to Figure 3, air is compressed to about

75 psig in a centrifugal compressor, cooled in an aftercooler
and freed of liquid water. It then enters a regenerator where
it is cooled to the saturation temperature. The air is then
processed to remove small amounts of contaminating solids
as well as all the vapor phase carbon dioxide and some
h)d.rocarbons. The air then enters the lower column.
To maintain self cleaning conditions in the regenera-

tors and to provide a preheated stream to the turbine, some
air is withdrawn at the -100 0 to _120 0 C level. This air
is cleaned of carbon dioxide and its temperature adjusted to
a preset turbine inlet temperature to assure a turbine exit
temperature close to the liquefaction temperature. In detail
this is more complex, of course, than shown.
With one exception, the double column and condenser-

reboiler operate in the conventional manner with the rich
liquid air and the liquid nitrogen from the lower column being
subcooled prior to transfer to the upper column. The
important exception is the hydrocarbon removal trap which
receives a continuous recirculation of liquid from the
-163-
Figure 3. Simple process diagram, air
separation unit, 80 million cu. ft. /month 02
Figure 4. Air separation unit, 20

million cu. ft. / month 02
Figure 6. Ins trument panel
Figure 5. Air separation unit, 80

million cu. ft. / month 02
-164-
condenser- reboiler. Operating experience has shown that
this system, properly designed, provides 100 per cent safe
conditions.
The nitrogen product is warmed to within a few

degrees of the inlet air temperature in the regenerators.
The oxygen product, normally 99.5 per cent, is warmed in
coils passing through the regenerators. The entire process
is completely automated.
Air Separation Packages
Figure 4 shows an air separation unit rated at 20

million cubic feet per month of oxygen. This is a single
package unit which uses reversing heat exchangers. This
unit is attached to the compressor building through an
operating shed which houses the operating balconies, ventila-
tion equipment, and the instrument panel. Connections to the
instrument panel are manifolded in the lower left hand corner
(near the mechanic). The enclosures covered by wire mesh,
house the exchanger reversing valves.
Figure 5 shows an air separation unit rated at 80

million cubic feet per month of oxygen and made up of three
cold box assemblies. Two of the cold boxes are regenerator
assemblies with contaminant removal equipment located on
top and reversing valves located below. The third assembly
houses the column and other equipment. The operating shed,
which is attached to the compressor building, houses the
expansion turbine, instrument panel. controls. and operating
balconies.
Instrumentation
Instrumentation Requirements: Sufficient instrumenta-

tion is provided to operate and control the process, match
oxygen production with customer demand, monitor critical
process variables, and protect equipment in case of mal-
function. All of the process controls are basically pneumatic
while monitoring equipment is essentially electrical.
Requirements for an unattended plant are much more

stringent than for the automatic plant with an attended central
-165-
control room. Succes sful unattended operation requires
fir st rate instrumentation engineering and a generous develop-
ment program preferably on a pilot plant.
Considerations in Instrument Selection: The prime

consideration in instru~ent selection was reliability. Where
possible, instruments which had proven reliable in other field
applications were used. Extensive bench and pilot plant tests
were conducted on unproven components, particularly electri-
cal components such as pressure switches, time delay
relays, etc. The simplest instrument which would perform
the desired function was chosen. For example, mercury
manometer type differential pres sure controllers were used
instead of differential pressure transmitters and receiver-
controllers. A minimum of electronic instruments were
used. These were employed on some temperature and
analytical applications where an acceptable alternative was
not available.
Proces s Controls: The number and type of controls

depend, to some extent, on plant size. In the larger plants,
controls are provided for three liquid levels, three tempera-
tures, four flows and/or flow ratios, and six pressures.
Instrumentation is also provided to control the capacity of the
air and oxygen compres sors and the reversal of regenerators
and reversing heat exchangers. In some plants it is neces-
sary to control the switching and regeneration of purification
traps. An analyzer -controller is provided to control product
purity. Figure 6 shows a typical instrument panel for a
small plant. A purity controller, three liquid level
controllers, monitor or safety panel, and pressure gages
are visible. Equipment for controlling the rever sing heat
exchangers and air compressor and oxygen compressor
capacity as well as miscellaneous primary monitoring elements
are located elsewhere.
Process Monitors: Process monitors, the key to an

unattended plant, are provided for items which would normally
be observed by the operator. These include such items as
product purity, cooling water pressure, oil pressures on the
compressors, compressor suction, interstage, and discharge
temperatures and pressures, and certain critical temperatures
and pressures in the air separation package. On the larger
plants monitors are provided for as many as 90 items. In
most cases the monitors are temperature and pres sure
switches. High and/ or low level monitors are provided for
-166-
some liquid levels. In applying monitors to critical
variables which are controlled, it was assumed that a
controller (with alarm contacts) cannot monitor its own
control and a separate instrument for this purpose was
provided.
Referring to Figure 7, all process monitors are

connected into a safety or monitor panel to provide visual
indication of actuation. The annunciators are simple
indicating circuit breaker s. This panel contains 40 annun-
ciators and is used on plants of 7 and 20 million cubic feet
per month oxygen capacity. To explain the operation of the
panel, suppose the plant had been shut down due to high
discharge temperature on the oxygen compressor. Upon
arrival at the plant the supervisor would find No. 856
actuated and the base load compressor circuit pilot light
glowing bright. (To guard against burned-out bulbs, pilot
lights are dimmed when there is no signal.) The difficulty
is therefore clearly indicated. Upon restarting the plant,
it is necessary to by-pass certain of the monitors until the
plant is running normally. The by-pas s pilot light glows
brightly when a by-pass is actuated and if the supervisor does
not acknowledge an audible alarm every twenty minutes by
pressing the by-pass reset, the by-pass will be automatically
nullified and an alarm sent to the monitoring service if there is
an improper condition. This prevents the plant from being
improperly monitored if the supervisor should become
careless. Three circuits on the panel contain annunciators
which indicate a condition that shuts down a compressor. The
fourth circuit contains annunciators which indicate an
undesirable condition requiring attention but not serious enough
to shut the plant down.
A simple safety circuit without the complicated by-pass

system is shown in Figure 8. Each of the monitoring devices
provides a normally open and normally closed contact. The
normally open* contacts are wired in series to provide an
energized circuit of the IIfail-safel! design. All circuits are
designed to be safe in caSe of loss of power. In caSe of a
mal-function, the circuit is deenergized causing the safety
control relay to drop out. At the same time the normally
closed contact opens, causing the indicating breaker to actuate,
*Open when deenergized, closed when energized.
-167-
•
.
't.Ol L CiIol' _ _ _
• •
Figure 7. Safety panel
1 - - - - - - - - - 1 1 0 V . I I . C . - - - - - - - - -- -t
/lUXI(. 11'/1'. Y
COAJT"I(:T 0.<.1 7iHi!E COAJT"'CTo,i(
OeLl''''' )feL"'Y (/.5eo )'O~
WHIC# Ol"'e.<.l$ 3 ShuTT/Alt; Dow"';
S~CCNOS /9'TI!H CG/UIPM,E,..,,:
S~uToOWAl
SIt/AlAioion.)<i
MCAII TC/'t/A./t:;
54Avlce, eTC.
/NOICnT,Alq~~--~---~----------i
C/~C:LI/T
B,I(el9xtV?,
Figure 8. Typical safety control cir cuit
-168-
but power is disrupted after a short time delay to prevent
other breakers from actuating due to secondary factors.
When the safety control relay drops out it first opens the
compres sor motor circuit to effect the shutdown and cau&es
all electrical and mechanical controls to be deener gized and
subsequently adjusted automatically to the safe position. It
then opens the circuit to the comrnercial monitoring service
causing a signal to be transmitted, lights a pilot light, and
finally opens a contact in its own coil circuit to prevent
possible reenergization due to the monitoring device closing
again.
Instrument Air Supply: In all pneumatic control

installations the cleanliness and reliability of the air supply is
an important consideration. In these plants the air for
operation of the instruments and valves is withdrawn from
the process at a point where it is oil free and has a dew-
point of _150 0 F.
Field Operation
General: A total of seventeen plants are currently in

operation. Plants under 40 million cubic feet per month
capacity operate completely unattended with weekly service
checks which require one man for about six hours. Plants
above 40 million cubic feet per month capacity are attended
by one man eight hours per day with the other two shifts
being unattended. During 1957 and 1958 about twenty
additional plants will be placed in operation.
Operating Personnel: High caliber men with previous

oxygen plant experience were selected to operate or service
these plants. The men were given an intensive training
program by the Engineering Department to familiarize them
with the plant. About two-thirds of this course was devoted
to instrumentation. They received further training during
checkout and initial operation of the plant.
Operating Experience: Seventeen plants have been

placed in service and a total of about 230 months of operating
experience has been obtained. During the first month of
operation of a new plant, outage is slightly higher than normal
but rapidly decreases as adjustments are refined. However,
because the plants are guaranteed 100 per cent, the customers'
-169-
oxygen supply is always maintained. The plants were
designed with an expected outage of less than 4 per cent
of which one-half of this is scheduled. The average outage
time for all plants is currently below 4 per cent and it is
expected that with further operation the outage will be
further reduced. Complete safety has been demonstrated and
the plants have been accepted with great enthusiasm by the
operating personnel and the customer.
Discussion
Question by L. Lazare, American Cyanamid Company: Are

continuous monitoring devices used for hydrocarbon and
acetylene content in process streams?
Answer by author: No. The plant is equipped with sufficient

hydrocarbon and acetylene removal equipment to remove the
contaminants. A routine manual test is conducted weekly in
the normal case as a precautionary measure.
Question by J. F. Grunberg, LIAir Liquide: Would a plant

remotely operated in normal operation require the presence
of personnel at the plant site when starting the unit from
warm conditions?
Answer by author: Yes.
Question by M. P. Sweeney, Process Associates: How often

do you defrost the plant? Do you regenerate the adsorbers
manually or automatically? How often?
Answer by author: Defrost time depends on plant design but

one time per year is reasonable. Adsorbers are regenerated
both manually and automatically depending on the specific
design.
Question by R. Katzen. Air Products, Inc.: I do not under-

stand the logic of the signal alarm and monitor system in
an unattended plant. Don't you require standby personnel in
all areas? Isn l t there lost time in getting this standby
personnel to the plant site?
-170-
Answer by author: The monitoring service notifies one of
our employees listed on the call list. Time for a man to
travel to the plant is relatively short.
Que stion by J. J. Brehany, California Oxygen Company: What

is the advantage of controlling hydrocarbon concentration in
the liquid system by a continuous recirculating liquid bleed
from the condenser through a filter rather than filtering the
rich liquid transfer? The latter is simpler to do.
Answer by author: The condenser is the most likely spot

for hydrocarbon concentration and we have found it economi-
cally feasible and technically sound to effect final cleanup
at this point.
Question by E. Johnson, American Messer Corporation:

What is the minimum operating level? How fast can plant
capacily be increased or decreased?
Answer by author: We would normally consider half load as

the minimum operating level when considering all technical
and economical aspects. Plant capacity can be increased from
one-half to full load in less than 15 minutes. Decreasing
plant capacity is limited by the fact that oxygen purily will
decrease temporarily during the transient. Going from full
to one -half load in incremental steps would require about
15 to 30 minute s.
Question by V. J. Johnson, CEL National Bureau of

Standards: In the more than 200 months of plant operation
have there been any incidents of equipment damage that could
have been prevented with operators present?
Answer by author: I know of none.
-171-
D-I
ON THE DESIGN OF A CRYOGENIC FACILITY FOR

TESTING AIRCRAFT COMPONENTS*
N. Meyers, H. Rind, and P. Gomez

Stratos
Division Fairchild Engine and Airplane Corporation
Bayshore, L. I., New York
Testing of aircraft cryogenic components such as

valves, pumps, etc., requires the usual torture testing
plus the use of various cryogenic fluids. Consequently,
many of the probelms are ordinary but some are extremely
difficult to solve. The difficulties are greatest for small
companies with limited resources who must either use the
services of specialized testing agencies or do it themselves.
The use of outside test laboratories is economically justified
in many cases, although problems exist in scheduling,
setting-up, delays, etc. This paper discusses the do-it-
yourself approach.
*Abstract of paper which was not presented.
-172-
D-2
LIQUID HELIUM TEMPERATURES IN
AN ATOMIC REACTOR
J. T. Howe, R. R. C oltlTI an , and T. H. Blewitt

Oak Ridge National Laboratory~'
Oak Ridge, Tennessee
FundalTIental studies of the basic nature of radiation

effects have been conducted in the Oak Ridge National
Laboratory graphite reactor by the Low TelTIperature Group
of the Solid State Division. It was discovered that lTIany
radiation induced defects were not observed at rOOlTI telTIpera-
ture because they therlTIally anneal at lTIuch lower telTIpera-
tures. MeasurelTIents of radiation induced changes in
electrical resistivity, yield stress, and stored energy in
copper and alulTIinUlTI single crystals were lTIade at salTIple
telTIperatures as low as l5 0 K. Subsequent annealing studies
show that about 50 per cent of the change in electrical
resistivity is recovered between 28 and 50 o K. This appears
to be the result of radiation induced defects escaping frOlTI
low energy traps. 1,2 It is particularly advantageous, when
lTIeasuring electrical resistance of these lTIetals, to be able
to bOlTIbard thelTI below 20 0 K because the change of resistance
with telTIperature is slTIall for lTIany lTIetals in this telTIpera-
ture region. It is desirable, however, to extend these
studies to lower telTIperatures to perlTIit lTIore precise studies
of other lTIetals, such as gold, for which the telTIperature
dependence of resistivity is still appreciable at l5 0 K.
*Oak Ridge National Laboratory is operated by

Union Carbide Nuclear COlTIpany, a Division of Union Carbide
and Carbon Corporation, for the AtolTIic Energy COlTIlTIission
under Contract No. W-7405-eng-26.
1
T. H. Blewitt, R. R. ColtlTIan, C. E. Klabunde and
T. S. Noggle, Journal of Applied Physics June 1957.
2
T. H. Blewitt, R. R. ColtlTIan, D. K. HollTIes and
T. S. Noggle, Creep and Recovery, ASM (1957).
-173-
Cooling for these experiments was supplied by a
refrigerator using helium gas as the refrigerant and employ-
ing a regenerative heat exchanger and two expansion engines
to remove energy from the gas and convert it to work in a
hydraulic mechanism. The refrigerator was designed for
another purpose by Arthur D. Little, Inc. and was modified
for our use by the designers. A detailed description of the
operating rrinciples of the refrigerator was given by A.
Pastuhov. The refrigerator has operated over 8000 hours
with very little maintenance.
The in-pile cryostat was built at ORNL and is

described by R. R. Coltman and T. H. Blewitt. 4
Figure 1 shows what the refrigerator looks like. The

cold case, in which a regenerative exchanger and two expan-
sion engines are located, is the large tank. The smaller
tank is the expansion tank and other objects are related
vacuum equipment and warning and emergency controls for
shutting down the reactor in case of trouble that would
endanger reactor operation. Compressors are on another
skid which is located outside the room.
Figure 2 is a picture of the inside of the cold case

and shows the heat exchanger with the expansion engines in
the center.
Figure 3 is a schematic of the refrigerator. The

original circuit will be modified in order to utilize the
compressor and filter system and to bring the temperature of
the liquefier gas down below the Joule-Thompson inversion
temperature. It is estimated that less than two pounds of
helium per hour will be needed to produce approximately a
watt of cooling at about 4 0 K in the reactor. Since the main
refrigerator circulates about 120 pounds per hour. the
quantity needed for the liquefier circuit should have little
effect on the main system.
3
A. Pastuhov. Proceedings of the 1954 Cryogenic
Engineering Conference. p. 21. Boulder. Colorado. September
1954.
4 R • R. Coltman and T. H. Blewitt. Proceedings of
the 1956 Cryogenic Engineering Conference. p. 110. Boulder.
Colorado. September 1956.
-174-
Figure 1. Figure 2.
COI.~"" "'".... !.1IfNtI".

_ r,.oot.al(I:(.HA""'IIl*
fIIIIXIo'IU.'.'-
-·-·-or.c.o "'~oo(L
Figure 3. Figure 4. Helium liquefier

main heat exchanger
-175-
In the liquefier section the high pressure gas enters
a regenerative exchanger at room temperature where it
gives up its heat to the gas returning to the compressors;
this exchanger is a main heat exchanger used in the Collins
helium cryostat except that the standard tap-off points were
omitted. The gas then passes through a short length of
exchanger in which the temperature is reduced to between
11 and 15 0 K by gas from the low pressure side of the
main refrigerator.
From the liquefier cold case the gas pas ses through
another exchanger where it will be cooled further by the gas
returning from the expansion valve. This exchanger will be
made up of small, thin stainless steel tubes, one within the
other. Because of the small flow of gas it is necessary
to provide a low longitudinal conductivity path for the
exchanger to operate satisfactorily. This stainless steel
exchanger is located within the cryostat in the shield section
of the reactor. This shield is of concrete and is seven feet
thick. The exchanger assembly will consist of four parallel
exchangers about 5 feet long. The gas will then enter a
J -T expansion valve for further cooling after which it will be
routed to a small gas chamber surrounding the sample
chamber. The sample chamber will be about 1/2 inch in
diameter and the cooled portion will be about 6 inches long.
The gas chamber will be about 12 feet below the J -T valve
and will be in the center of the reactor lattice.
A high gamma flux exists in a nuclear reactor. The

absorption of this radiation by materials results in a substan-
tial source of heat and is an important factor in the design of
cryogenic devices to be installed in a reactor. In the ORNL
graphite reactor this heat source amounts to about 5 mw/g.
In reactors of higher flux, particularly in those moderated
by natural water, where a substantial portion of the neutrons
are converted to gamma rays by a capture process with the
hydrogen, the gamma ray heating can be extremely high;
for example, the Oak Ridge research reactor will have a
gamma heating of about 1 w/ g. Therefore in designs of
cryogenic equipment for a reactor, the mass of those
components located within the reactor lattice must be held to
a minimum. The combined mass of the sample chamber,
gas chamber, helium lines, and sample in this case will be
only 160 grams.
-176-
Since the gamma heat input to a radiation shield will
exceed the heat input from thermal radiation, no thermal
radiation shields are incorporated in the reactor portion of
the circuit.
The largest heat load in this liquefier system will

occur in the supply and return line s to the gas chamber,
since vaporization of the majority of the liquefied helium will
occur here. It is hoped that the helium within the gas
chamber can be maintained at 4 0 K or lower; it may not,
hov.ever, be in the liquid form because the pres sure of the
gas chamber will be maintained as flr below 15 psia as is
possible. To do this an oil sealed mechanical vacuum pump
is to be utilized. Mechanical vacuum pumps of this type are
designed to discharge at atmospheric pressure. Therefore, a
small auxilliary compressor will be used to maintain the
vacuum pump exhaust at or below atmospheric pressure and
to compress it to 30 psia for return to the suction side of
the refrigerator compressors.
The sample chamber will be filled with helium under

pressure so that, when the temperature in the gas chamber
becomes low enough, the sample chamber gas will liquefy
and the sample will be immersed in liquid helium.
Figure 4 shows a schematic diagram of the liquefier

cold case assembly. Several problems were encountered in
the construction of the cold case. The Collins main heat
exchanger has a thin stainless steel shell designed for less
than 5 psi differential pressure with the surrounding gas;
the problem of possible pressure buildup was solved by
permitting the return gas to pressurize the chamber surround-
ing the Collins exchanger.
It was desirable to locate the short exchanger which

supplied cooling from the main refrigerator in the space
within the Collins exchanger. Since this ~hamber should be
evacuated and the chamber in which the Collins exchanger is
located may at times be subjected to slight pressure, the wall
separating the two should be thick enough to withstand the
pressure difference; on the other hand, a thick wall would
conduct heat more rapidly than a thin one. This problem was
solved by machining a 6 inch O.D. x 1/4 inch wall stainless
steel tube to about .020 inch so as to leave' fins about every
-177-
2 -1/4 inches thus reducing the effective LID ratio and at
the same time keeping longitudinal conduction to a minimum.
We would like to thank C. E. Klabunde of ORNL for

his many valuable suggestions in the design and his aid in
building the apparatus.
-178-
D-3
INTEGRAL PROBE CAPACITANCE GAGING OF

LIQUEFIED GAS CONTAINER CONTENTS
H. J. Smith
The Aro Equipment Corporation
Bryan, Ohio
Introduction
The need for a simple, reliable method of determining

the volume of liquid in a liquefied gas container became
apparent early in the existence of a suitable container for
liquid oxygen used as a source of breathing oxygen aboard
high altitude aircraft.
Of the several methods of gaging common to this

application (differential pressure indicators, differential
pressure actuated transducers, etc.) the capacitance gaging
system was selected because it more nearly conformed to the
rigid requirements of accuracy, continuous indication and light
weight.
Capacitance gaging supplies a means for accurately

and continuously measuring the liquid volume in stainless
steel liquefied gas containers. The only connecting links
between the indicating device and the container are two low
power electrical leads.
The first gaging system utilized by The Aro Equipment

Corporation was designed by The Aero Division of Minneapolis-
Honeywell Regulator Company in conjunction with The Aro
Equipment Corporation. The liquefied gas gaging system is
based on the fuel gaging systems used satisfactorily for
many years.
This report describes a null-seeking, servo balanced

capacitance gaging system utilizing an integral probe as the
transmitter assembly. Included are: the general theory of
system operation, a description of the integral probe, its
function and a discussion of its practicality from the viewpoint
of the container manufacturer and the container user.
-179-
The complete gaging system consists of the indicating
device, intermediate device, an integral capacitance probe,
and interconnecting cables. The usual configuration has
the indicating and intermediate devices combined in one
package.
Basic Operation of Capacitance Gaging System
Capacitor gages operate on the basic principle of a

capacitance bridge circuit maintaining a continuous null
balance. Figure 1 illustrates a simplified block diagram of
the capacitance gaging system under discussion. On one side
of the bridge is the integral probe that, in effect, is a
variable capacitor in which the capacitance is determined by
change s in the level of the liquefied gas. This variable
capacitor will unbalance the bridge circuit and provide an
input signal to the amplifier which in turn, controls a motor
driven balance potentiometer in the indicator. The poten-
tiometer wiper reapportions the voltage drop across a
reference capacitor in the side of the bridge opposite the
tank unit to restore the bridge to balance. An indicator
pointer, rotating on a common shaft with the potentiometer
wiper provides a continuous indication of the liquid volume
remaining in the containe r •
Integral Probe
The integral probe provides the variable capacitance

value that causes the bridge to become unbalanced. The
purpose of the probe is to provide a capacitance value that
corresponds to predetermined values indicating liquid quantity.
The probe is physically a multi-element capacitor

whose elements are concentric tubes. Since the integral
probe does not require an entry port bridging the vacuum, it
is not restricted to small diameters. The overall diameter
is governed by the amount of capacitance desired and the
curvature of the container. The diameter of the integral
probe for a 5 liter spherical container is approximately 1
inch with approximately lIS inch spacing between its three
elements. The probe is constructed of the same type of
stainless steel as the container with high alumina ceramic
insulation between elements and between the sensing assembly
-IS0-
r------l r--- -------- - - - - - - - - - - - - - - ,
I CONTAINER, I I ..'~OICATOR. LlQUEFIEO GAS QUANTITY I
I LIOUEFIEO GAS I I I
I I I I
I I I I
I I I I
I I I
I
I : I
I I I
I _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ .JI
L
Figure 1. Block diagram

capacitance gaging system
LEAOY,IR[ CONNECTION
~OVAIILE
/
/
Figure 2. Integral probe, side Vlew
Figure 3. Integral probe, top view
-181-
and the mounting brackets. Figures 2 and 3 show side and
top views respectively of the integral probe.
An active capacitance of 12 mmf for each liter of

liquid to be sensed was set by military specifications and
a value of 3.5 mmf of dead capacitance due to spacers and
electrode connections was set by agreement of probe
designer s.
The measured capacitance of a I1dryl1 probe for a 5

liter container is then 63.5 mm£. This is measured with
air or gaseous oxygen as the dielectric. When the liquid
oxygen level rises in the probe, the increased dielectric
constant causes an increase in measured capacitance. The
value of the measured capacitance may be calculated by
the following equations,
(1)
or (2)
where C = measured capacitance in mmf.

m
Ce = empty or dry capacitance of the probe in mmf.
Ko = 1. 483 = dielectric constant of liquid oxygen

at -2970F and atmospheric pressure.
12 mmf = active capacitance per liter
Va = volume sensed in liters.
The first term on the right hand side of this equation

is a constant for the liquid capacity of the container used.
The term (Ko - 1) is also a constant having a value of 0.483
at -2970F and one atmosphere. Combining this constant with
the 12 mmf per liter we arrive at an added capacitance value
of 5.796 mmf per liter of liquid volume. Equation (2) may
then be written as follows:
-182-
C = C
e
+ 5.796 V0 ( 3)
m
By the use of equation (3) and Table I, which lists

the empty capacitance of the common container volumes,
the capacitance variation for any listed container may be
calculated.
TABLE I
Container Volume Empty Capacitance

Liters mmf
5 63.5
10 123.5
25 303.5
Table II shows the liquid height variations for a 5

liter container for each major volume division. The heights
listed do not include the displacement of the probe or
fittings. This illustrates that the capacitance variation is
linear while the liquid height is a non-linear variation.
TABLE II
Volume (Liters) Measured Capacity Liquid Height
mmf (inches)
0 63.5 0
1 69.30 2.35
2 75.09 3.55
3 80.89 4.65
4 86.68 5.75
5 92.48 7.30
-183-
The capacitance of the probe is dependent upon the
dielectric constant and height of the liquid present between
its electrodes. Therefore, to give a linear change of
capacitance with respect to liquid volume the capacitance
of the probe must be characterized to the configuration of
the container. Figure 4 shows the military requirements of
measured probe capacitance with respect to liquid volume.
Figure 5 shows the relationship of height and liquid volume
for a spherical 5 liter container.
As previously stated, the probe capacitance varies

only as the dielectric constant and height of the liquid
present in its elements. This then requires that the
corre sponding heights of liquid in the container and heights
of liquid in the sensing elements yield compatible volume
and capacitance relationships. The characterization of the
probe provides this necessary requirement. The probe is
caused to yield a capacitance signal that is linear with
respect to volume by varying the area of the middle electrode.
For a low (dv/dh) the middle electrode will have a small
area thus yielding a lesser (del dh). The largest common
area of the middle electrode will occur where (dv/dh) is
maximum •
Since the integral probe is a characterized element

and the liquid level must yield compatible capacitance and
volume values, the location of the probe in the container
must be closely controlled.
A variation of 0.03 of an inch in shell diameter would

cause a 0.05 liter change in volume. Therefore, to insure
probe accuracy, a means of mounting adjustment must be
provided before assembly of the container. To compensate
for container shell diameter variations, the probe is
accurately located with respect to lower end relationship
with the inner diameter of the shell. The accuracy of this
location is dependent upon the tolerances given for the
capacitance variation, but will be in the area of..±. .005 of
an inch for conventional tolerances. Reference to Figure 5
will show that (dv/dh) is greater at the volumetric center
of the container than at the extreme ends of the probes. This
is an additional reason for close location of the probe with
respect to end clearance of the inner diameter of the
container.
-184-
90.0
80.0
u
z
V
~ 70.0
LIQUIO -VOLUME-LITERS
Figure 4. VolUll1e and measured capacitance

relationship, 5 liter container
HEIGHT {INCHES)
Figure 5. VolUll1e and liquid height relation-

ship, 5 liter spherical container
.s
.. .8 1.0 1_2
BAFFLE DISTANCE - INCHes
I.' 1.5 1.8
Figure 6. Probe capacitance reduction curve,

5 liter spherical container
-185-
The need for close tolerance of the probe placement
is further accented when we consider the reduction of probe
capacitance due to the electric field interference that occurs
whenever a ground potential surface approaches the extreme
ends of the probe. The capacitance reduction for a probe
used in a spherical 5 liter container is shown as a function
of the proximity of the fill or vent fittings in Figure 6. The
same relationship is shown for a "flat" 5 liter container in
Figure 7. A comparison of the two reduction curves show
that the capacitance reduction is a function of fitting or shell
wall proximity and will vary with the container configuration.
This requires close coordination of the probe manufacturer
and the container designer with regards to the expected
end clearance of the probe. This then may be interpreted
in terms of capacitance reduction and proper compensation
made in the manufacture of the probe.
From this it can be seen that a compromise must be

made between probe capacitance reduction and sensed volume.
Typical lower end clearance of the probe is one-fourth of an
inch. This means that the last .01 of a liter is not sensed
or indicated by the gaging system for a 5 liter spherical
container. This is not an undesirable condition when the
safety of the aircraft and its occupants is considered.
Because the integral probe is subjected to temperature

extremes in processing and exposure to liquefied gas, the
stability of the electrical properties of the insulating mater-
ials is a critical requirement. The thermal expansion of
the insulating materials must also be comparable to the type
of stainless steel used in the probe electrodes and mounting
brackets.
The behavior of electrical properties of the insulating

materials at liquefied gas temperature requires rigid quality
control of materials used as well as the incorporation of
compensating factors for changes of characteristics from
room temperature to liquid temperature.
A common occurrence is a decrease of dry or empty

capacitance measured at liquid temperatures as compared to
room temperature. One of the more common causes of
this trouble is the insulating spacers between the probe
sensing elements. Impurities in the insulating material may
cause the insulating spacer to have a smaller than normal
-186-
$
~.5"
~
<,~
~A
~
~
~'
~
~Z
~
".~
~ J
.2 .if B 1.0
BAFFLE DlSTANC.E -INCHES
I.il I.' 1$ I.e
Figure 7. Probe capacitance reduction

curve, 5 liter "flat" container
EMPTY CAPACITANCE· 63.5 JJJJF {NOMINAL)
ADDEO CAPACITANCE. 28.9 J.llJF
FIVE LITER SENSOR BY M'H
~
"-
0 .•
~
~
! 0.'
u
z
~
Of
~
.
!
0.'
~
0.1 U
DISSIPATION FACTOR
Figure 8. Probe capacitance reduction

due to dissipation factor
INTEGRAL PROBE
TO CAPACITANC£ GAGING
I .. uuo ..., '
REMOVABLE PROBE
Figure 9. Comparison of heat leak

paths, integral probe versus
removable probe
-187-
resistance. The exposure of the spacer to liquid tempera-
ture causes the resistance of the spacer to approach infinity.
The smaller insulation resistance will cause the capacitor
to have a high dissipation factor and therefore the measured
capacitance will be higher than the true capacitance of the
probe. This means that the measured capacitance of the
probe may be satisfactory when measured at room tempera-
ture and prior to installation in the container. During
exposure to liquid temperature, at the time the pattern
conformance is checked, the probe will be below tolerance
when the container is empty and at liquid temperature.
Figure 8 shows the expected decrease of capacitance at
liquid temperature from room temperature as a function of
the dissipation factor. There are two ranges shown on this
curve. One range is shown as probable, the other as a safe
range. The data for this curve was taken from a representa-
tive group of 5 liter spherical liquid oxygen container s.
Therefore, the data expressed should not be used as an
accurate means of determining the capacitance decrease due
to the dis sipation factor of the capacitance element.
Practicality of the Integral Probe as a Gage Sensing Element
The prime consideration of container designers and

manufacturers is simplicity with sound engineering. The use
of the integral probe answers both aims well. The elimina-
tion of the large and direct heat leak path associated with a
removable probe makes the gaging system a secondary
design problem and not a major element. Figure 9 shows a
comparison of the heat leak path for an integral probe and a
removable probe. The probe leads may be routed through
the vacuum space with little concern for heat leak paths. The
elimination of the container seal necessary for removable
probes is a further step in simplification. The integral probe
mounting is a simple method of fastening to the container.
This is accomplished by weldments of the same type that are
used for the container construction. This, in effect,
combines the probe and the container into a single structure
with more desirable vibration characteristics.
The use of an integral probe simplifies the system for

the user. This is true since the container may be manu-
factured, tested and calibrated as a system. Once the calibra-
tion is accomplished by an accurate location of the probe, the
-188-
probe should have the same service life as the container.
A complete unit that does not require adjustment or

sensing element replacement periodically simplifies the
system use for military application. The life of the integral
probe should be the same as the container. This is stated
because the probe is made of the same material as the
container and subjected to the same processes. The fact
that a more sturdy mount is obtainable and that there are
no moving parts necessary adds further basis for this state-
ment.
Discussion
Question by O. Heustis, The Martin Company: How closely

can the capacitance probe determine the liquid level in a
liquid oxygen container?
Answer by author: In an insulated liquid oxygen container,

the level of the liquid may be determined within a 2 per
cent maximum. This error is in regard to volume and may
be transformed to a height variation by consideration of the
volume sensed and the container shape.
If you are specifically questioning the accuracy obtainable in
an uninsulated tank during rapid filling this would be some-
what higher than 2 per cent. I would not like to specify an
accuracy for this type of application due to the many
variables present. The amount of turbulence of the liquid
would have a decided effect as well as the change in
dielectric constant which may occur as the temperature
changes in the liquid.
Question by W. A. Wilson, CEL National Bureau of

Standards: What method is used to obtain a vacuum tight
seal around the electrical leads which extend from the
integral probe to the outside?
Answer by author: The method used to obtain a vacuum tight

seal around the electrical leads which extend from the integral
probe to the outside is by weldments. The probe leadwire has
a solid stainless steel sheath insulated from the inner
conductor which allows the 1eadwire sheath to be welded to
the walls of the inner and outer container to insure vacuum
tightness.
-189-
Question by J. P. Stechmeyer, Pesco Products-Division
of Borg Warner: Is a metal to glass seal used in taking
the leads through the container wall?
Answer by author: A metal to glass seal is not used in

taking the leads through the container wall. The probe
leadwire is a single length of shielded conductor from
the probe to the outside of the container.
-190-
D-4
A UNIQUE THERMAL CONDUCTIVITY GAS ANALYZER
J. R. Purcell, J. W. Draper, and D. H. Weitzel

Boulder, Colorado
For many years the thermal conductivity method of

gas analysis of binary mixtures has been very popular.
Recent improvements in the field of chromatography have led
to the use of thermal conductivity for analysis of multicom-
ponent mixtures. Using this system the mixture is passed
through an adsorption column. Or. passing a carrier gas
through the column, the components are recovered in
sequence, being determined in turn by means of a suitably
calibrated thermal conductivity cell.
Figure 1 shows a conventional thermal conductivity

analyzer. The measuring cell generally consists of a bras s
block with four chambers containing resistance thermometers.
The resistance thermometers are connected in a Wheatstone
bridge and supplied with sufficient current to heat them
appreciably above their surroundings. A reference gas is
passed by two of the thermometers in opposite arms of the
bridge and a sample gas is passed by the other pair. The
temperature, and hence the resistance of each thermometer,
depends on the thermal conductivity of the gas surrounding it.
Thus the degree of bridge unbalance will vary with the
composition of the sample.
In conventional systems 1 the thermal conductivity cell

is excited with direct current and the resulting d-c signal is
measured with a potentiometer. A d-c system becomes very
difficult to use with very small signals, in the order of less
than 0.5 mv full scale, due to zero drift caused by thermal
emf's. The system has many unavoidable bimetallic connec-
tions which make it extremely temperature sensitive. Another
disadvantage of a d-c system is the expense and size of the
measuring equipment involved.
lD. H. Weitzel and L. E. White, Rev. Sci. Instr.,

26, 290 (1955).
-191-
REC. +
r--- ---l
B+ I
I
CELL
I B-
PUSH TO l---
CAL. I
A B
Figure 2. Cell connections
TO CELL
TO VENT
PURGE
Figure 1. Conventional
thermal conductivity
gas analyzer
REFERENCE A
, '
, \ .. ~J)
REFERENCE - - - ~ , NORMALIZER
SOURCE l<~.!.:
NORMALIZER
Figure 4. Gas flow
50.
HEll POT
3.3k Ik
L..JW'-~MJ""""--+
TO RECORDER
L-_ _ _..;.IO;..;.O II.V.
Figure 3. Phase-sensitive
rectifier
-192-
If the cell is excited with alternating current
several advantages become apparent for very low level use.
Small a-c signals can be measured in the presence of la-rger
parasitic d-c potentials. The amplifier is very simple with
an a-c system since a chopper is not required and standard
components can be used as in any conventional, low level,
a-c amplifier. Peaking the amplifier at the bridge excitation
frequency improves the signal to noise ratio, improving the
low level characteristics even further. The output of the
amplifier passes through a phase-sensitive rectifier to
operate a d-c recorder or output meter. The large magni-
tude of the d-c output allows the use of an inexpensive
recorder. A wiring diagram showing cell connections is
shown in Figure 2 and the phase-sensitive rectifier is shown
in Figure 3.
Stray pickup presents a small problem but can be

minimized by using a special transformer for the input, and
proper placement of the low level leads. Some stray pickup
can be tolerated since the cell can be adjusted to cancel
small strays. The gain or amplification of the amplifier will
affect the readings, as this is not a self-nulling device. It
is therefore important that the amplifier gain be very stable.
A great deal of negative feedback must be used to obtain this
required stability.
The cell is excited with 60 cycle current from a

constant voltage source supplied by a Raytheon constant
voltage transformer and is monitored by a milliampere
meter which reads the cell current. The output of the cell is
fed into a Peerles s low level transformer, carefully shielded
against magnetic and static stray pickup. The transformer
provides impedance matching from the cell to the amplifier
and also allows both the input and the output of the cell to
be grounded. This eliminates troubles arising due to
capacitance to ground in the excitation sour ce. The gain of
the amplifier must be constant and reproducible. To achieve
this, a small signal is taken from the constant voltage source
and is injected into the amplifier by means of a push switch.
The gain of the amplifier is then adjusted to give a predeter-
mined reading of the output. Using this scheme the gain can
be checked at any time during operation.
-193-
The d-c output is obtained from the phase -sensitive
rectifier, which gives polarity reversal when the bridge
passes through balance. The resulting signal thus obtained
is the same as though the cell were excited with d-c.
Without this phase-sensitive rectifier it is impossible to tell
whether it is the sample or the reference which has the
higher thermal conductivity as all deflections will be in the
same direction. A conventional phase-sensitive rectifier
using triodes will show some drift. To minimize this a
rectifier circuit is used which requires only diodes for
operation. This type of phase-sensitive rectifier has very
small zero drift and is linear.
If an instrument of this type is to be used for

hydrogen impurity analysis, ortho-para normalizers must be
used in both sample and reference lines, to eliminate any
effect due to difference in ortho-para composition of the two
streams. A possible flow system is shown in Figure 4.
An analyzer constructed as described in the preceding

paragraphs, and using a standard Gow-Mac four filament
brass thermal conductivity cell, has the following character-
istics: Using a miniature 100 mv recorder to indicate the
output, and using mid scale as zero, the unit can have a
range of as little as 200 parts per million of nitrogen in
hydrogen for full scale deflection, or one per cent of para-
hydrogen in pure or tho -parahydrogen mixtures, and 700 parts
per million deuterium in hydrogen. Zero stability and nois e
level are equivalent to about 10 parts per million nitrogen in
hydrogen, or 0.05 per cent para differential, and 35 parts
per million deuterium in hydrogen. Time required to
complete a reading on a sample is about 30 seconds.
Discussion
Question by G. H. Caine, Pesco Products: What is the

source of stray a-c signal?
Answer by author: Most of the stray a-c pickup comes from

the constant voltage transformer. The constant voltage
transformer produces a very strong magnetic field and thus a
strong stray voltage. The stray pickup is all internal and
constant, however, and once it is cancelled out it remains stable.
-194-
Question by G. H. Caine, Pesco Products: Would there be
any advantages in using a frequency other than 60 cycles per
second, say 1,000 or 2,000 cycles per second?
Answer by author: Yes, there would be an advantage but it

is doubtful that the added stability would be worth the
increase in cost.
-195-
D-5
PISTON TYPE HIGH PRESSURE AIRCRAFT

CRYOGENIC PUMPS
R. H. Ball
Sundstrand-Denver
Denve r, Colorado
There are requirements in the aircraft industry for

high pressure, light weight, cryogenic pumps. Sundstrand-
Denver, a division of the Sundstrand Machine Tool Company,
is currently developing a piston type, high pressure aircraft
pump for cryogenic fluids. This type of pump has proven
itself in hydraulic systems as an efficient means of obtaining
moderate flow rates at high pressure. Development testing
to date has shown that similar performance is obtainable
pumping liquid nitrogen.
Early in the testing program it was decided to test

some existing pumps making only such modifications as
necessary. In this manner it was hoped to overcome some
of the more fundamental problems before testing a pump
designed specifically for liquid nitrogen. This paper deals
p:Limarily with tests of existing pumps and the modifications
required to obtain satisfactory performance with liquid
nitrogen.
The first unit selected for testing as a cryogenic

pump was a Sundstrand 3 cubic inch displacement industrial
hydraulic motor, since a hydraulic motor can be driven as a
pump without change. Figure 1 shows the as sembly drawing
of this unit. This is a 9 cylinder axial pump with a shaft
driven wobbler plate. A fixed clearance, oscillating disc
type valve is incorporated. The shaft is supported by roller
bearings. The hydraulic rating at 1800 rpm is 2000 psi, and
22 gallons per minute flow. The displacement is 3 cubic
inches per revolution. The pistons are .614 inches in
diameter. Figure 2 shows a photograph of the disassembled
pump. Note the valve plate, the wo bbler plate containing both
a ball and roller bearing, and the three anti-friction bearings
on the pump shaft. This unit is constructed almost entirely
of iron and steel so that no modification was required to allow
for differential thermal contraction.
-196-
A satisfactory test facility for the pumping program
presents many problems. Figure 3 shows the typical piping
arrangement used during all tests. The pump is driven by
a variable speed drive pad. Since most of the pumps tested
have free floating pistons. an inlet pressure of about 50 psi
is needed to force the pistons to follow the wobbler plate
during the inlet stroke. This is accomplished by pres surizing
the tank from a nitrogen gas cylinder. The liquid nitrogen
flows from the 15 gallon insulated tank to the pump inlet.
The pump discharge is adjusted and held at high pressure by
throttling with a needle valve. Flow is measured with a
Rotameter type flow meter before returning to the tank.
The system is operated until the pumping energy heats the
liquid nitrogen to the boiling point which can cause severe
pumping instability. The tank is then depressurized to cool
the liquid to -320 o F by boiling. After repressurization the
testing is resumed. Figure 4 shows a test in progress. The
tank is just above the view shown. Styro foam wrapped with
plastic tape is used to insulate the tank and piping.
The first attempt at pumping liquid nitrogen with

the 3 cubic inch pump soon ended in seizure of the valve
disc. The only damage was valve surface scoring which was
repaired by lapping. In the next test, seizure was eliminated
by running a bypass line with valve from the pump inlet to
the pump casing and filling the case with liquid before
starting to pump. In this manner thermal equilibrium was
obtained prior to start-up, resulting in less warping of the
close tolerance valving surfaces due to rapid temperature
changes when flow commences. The pump turned freely in
subsequent testing.
Figure 5 shows the test results finally achieved on

this first pump. The volumetric efficiency obtained was 65
per cent at 800 rpm and 290 psi. The maximum pressure
obtained was 290 psi, limited by the available driving torque
at the time. The large leakage was attributed to the fixed
clearances in the valve mechanism. Later experience
indicates that higher volumetric efficiency would have been
gained by running at higher input rpm since leakage is a
function of pressure rather than rpm. The final test run was
1/2 hour at 400 rpm, then 1/2 hour at 800 rpm. No
mechanical damage occurred during this test. The pistons.
all but one anti-friction bearing, the wobbler plate surface,
-197-
Figure 1. Assembly drawing of Sundstrand 3 cubic inch
pump
Figure 2. Sundstrand 3 cubic inch pump, disas sembled
Figure 3. Piping diagram, liquid nitrogen pump test
Figure 4. Liquid nitrogen pumping test on Sundstrand

3 cubic inch pump
Figure 5. Test curve on 3 cubic inch pump
Figure 6. Photograph of disassembled Sundstrand .71

cubic inch pump
Figure 7. Photograph of disassembled other vendors .506

cubic inch pump
Figure 8. Assembly drawing of Sundstrand .013 cubic inch

pump
Figure 9. Photograph of disassembled Sundstrand .013 cubic

inch pump
Figure 10. Liquid nitrogen pumping test on Sundstrand .013

cubic inch pump
Figure n. Early test curve on .013 cubic inch pump
Figure 12. Piston pick-up, .013 cubic inch pump
Figure 13. Laminated plastic valve surface, .013 cubic

inch pump
Figure 14. Burned valve face, .013 cubic inch pump
Figure 15. Scored piston heads, .013 cubic inch pump
Figure 16. Scored wobbler bearing plate, IJ 13 cubic inch pump
Figure 17. Failed ball bearing, .013 cubic inch pump
Figure 18. Recent test curve on .013 cubic inch pump
-198-
Figure 2.
Figure 1.
--.... ,....-----
Figure 4 .
Figure 3.
- ~ ...L.
· ~. ~ .
·
· ~ ..
:s ~. .
~
~
· ....
I
· ....... ,..,
· 1"-
· .
· ~ un
· .
Figure 6.
Figure 5.
-199-
I
,",
,,' I
I
Figure 7. Figure 8.
",
Figure 9. Figure 10.
-· -
,I··
)I....
.;: tt--..
· ~") ~ ...
~
.
~.
\
II ··
~.
.]>•."
.. \
· ......
.........."!;. ~
· .r - ....
· ......
· ,
Figure 11.
Figure 12.
-200-
Figure 14.
Figure 13.
JS .
-
· """ .. ~
· -ro- r-'
/ ~~
·
.
0
.-~
1--. -- :..- -..,
IP:~ e'_
°F_
-~ ;;= 1--
I-
0
·
·
-201-
and valve surfaces were all in good condition. The pump
shaft roller bearing nearest the mounting pad showed signs
of warpage and was lightly discolored. This was caused by
its location in a dead space with no circulation of cold gas
or liquid.
The next pump selected was a .71 cubic inch

displacement modified Sundstrand aircraft hydraulic motor.
Better volumetric efficiency was expected because this pump
utilized a spring and pressure loaded valve surface. The
pressure loaded, lapped valve face allows less valve leakage
than the fixed clearance type previously tested. Figure 6
shows a photograph of the disassembled parts. Note the
pressure loaded valve face, the wobbler plate, the shaft ball
and roller bearings and the free pistons.
The first two tests were unsuccessful due to sump

leakage caused by galling of the cylinder block bronze valve
surface against the hard steel valve surface. A linen base
laminated phenolic valve surface was cemented on, and
immediate improvement in pumping was achieved. The anti-
friction bearings, pistons, wobbler plate. and plastic valve
surface were undamaged in the testing that followed.
At this time a .506 cubic inch displacement pump

manufactured by another vendor was tested after slight
modification. Figure 7 shows a photograph of the disassembled
pump parts. This pump incorporated two ball bearings
supporting the input shaft. a pressure loaded valve. while the
7 pistons are connected with ball sockets. The cylinder bores
were slightly enlarged to prevent piston seizure due to
differential thermal contraction of its bronze cylinder block on
the steel pistons. Also. a plastic laminate valve surface was
cemented on. Succes sful pumping was achieved with 70 per
cent volumetric efficiency at 600 psi and 2000 rpm. In the
final test, the pump was inadvertently run at 2000 rpm for
several seconds with no liquid flow due to vapor lock condi-
tions. Piston seizure occurred and two piston rods were
pulled from their sockets.
In the testing to this point. no undue difficulties were

encountered with pumps designed for hydraulic fluids.
Clearances could be adjusted to handle the differential contrac-
tion encountered. Conventional materials in ball and roller
-202-
bearings, hard steel pistons, iron and aluminum castings,
and shafts appeared operable. The laminated plastic was
found to be a satisfactory rubbing surface material for the
valve faces, and leather shaft seals were satisfactory.
Some corrosion was encountered due to intermittent pump
exposure to air and moisture between test runs. It should
be stated that a series of occasional difficulties prevented
any of these pumps from pumping liquid nitrogen with
reliability.
From this point on, a smaller pump model was

tested in order to consume less liquid nitrogen, less
power, and operate at higher speeds with the available test
equipment. All the following testing reported has been
accomplished using modified Sundstrand .013 cubic inch
displacement aircraft hydraulic motors. This unit weighs
2.5 pounds. Its 9 pistons are .254 inches in diameter. The
pump hydraulic rating would be .56 gallons per minute per
1000 rpm at 3000 psi. Figure 8 is the assembly drawing of
this pump. Note the free pistons, the pressure loaded
cylinder block, the 18 degree angle wobbler plate, and the
ball and roller bearings which support the pump shaft.
Figure 9 is a photograph of the disassembled pump parts.
The greatest bulk of testing and development has

been done on this pump model. Figure 10 shows the test
set-up used for this pump. The pump is mounted on the
variable speed drive and is insulated with fiberglass wool.
The tank and piping are Styrofoam insulated. Recording
instrumentation is being used in addition to a visual gage
or meter for each parameter to observe the results during
the test.
There have been several basic problem areas. Most

of these have been resolved; however, they are worthy of
still further study and development. Testing and. instrumenta-
tion has involved many minor problems. At first there were
leaks at connections and valve pac kings , but this was readily
corrected. Commercial valves with Teflon packing are used.
Piping is generally 37 0 flared tube aircraft fittings and some
heavy duty industrial piping. Standard pressure gages are
used with long lines to warm the gas before it reaches the
gage. Flow is measured quite successfully by an all stainless
steel Rotameter. Recording instrumentation has been added to
record flow, pressures, rpm, and temperature on a
-203-
continuous chart. The recorder requires a digital to voltage
conve rter for rpm, strain gage type pressure pick-ups, an
electrical recording head on the flow meter, and copper-
constantan thermocouples. Two types of pump drives are
used, one a hydraulic motor regulated by a flow control
valve, and the other a variable speed belt drive unit.
Typical early test results on this pump are shown

in Figure 11 which plots both flow and volumetric efficiency
against pump discharge pressure. Note that zero flow can
occur at less than 3000 psi for this pump indicating the
existence of experimental problems. It should be mentioned
that the earlier pump tests were not run at this high a
pressure.
Since the pressure loaded valve has nearly zero

clearance, it was suspected that the largest leakage was
across the pistons. Therefore, considerable attention was
directed to decreasing piston clearance and in developing
piston rings. According to leakage theory, flow varies as the
cube of the clearance, so emphasis was placed on close
clearance. A diametrical clearance of .00025 inch for .250
inch diameter pistons is a typical value successfully operated.
To further combat leakage aggravated by the very low viscosity
of liquid nitrogen, several types of piston rings were tried
such as Teflon, laminated plastic, and steel. Good rings
are quite difficult to design and manufacture in such a small
size.
Piston pick-up with the original bronze cylinder

block was at first a problem. The side load on the pistons
causes the cylinder material to pick-up or rub-off onto the
piston surface. Figure 12 shows this condition. The dis-
coloration is material picked up from the cylinder block.
Tests with 52100 steel piston and various cylinder block
materials such as cast iron, bronze, ductile IIni-resist,"
hardened 440 stainless, and hard 52100 steel showed that
bronze gave bad pick-up while the hard steels appeared
to be good. The 52100 pistons and block are now used.
The use of laminated plastic against hard steel as

the valve materials has been satisfactory. Figure 13 shows
the plastic valve surface. Proper pres sure loading is
accomplished by careful selection of the port size and the land
-204-
width. Considerable trouble was encountered in getting a
good adhesive bond between the .032 inch plastic and the
metal cylinder block. Several thermo setting and epoxy
adhesives were tried until one was found that was adequate.
In conjunction with this problem, it has been necessary to
initiate a basic study of possible valve face materials and
bonding methods. A study contract covering this phase has
been assigned to the Denver Research Institute. This
program is still in progress and involves literature search,
investigation of materials and methods, and strength testing
of samples.
A serious problem in pumping has been suction

instability. Lowering the inlet pressure to the pump suffi-
ciently causes loud mechanical noise in the pump which is
partially attributed to cavitation conditions, although it is
possible that other effects could contribute. The same
condition can occur at the higher pumping pressures when a
positively overbalanced valve face is used. This is believed
to be caused by the greater friction heat generated at the
higher pressure. Overbalancing allows pump pressure to
force the valve surfaces together. When this unstable
operation is allowed to continue for more than a few seconds,
damage is likely to be incurred. Typical damage is piston
pick-up, a scored wobbler plate, badly scored piston heads,
and/or a burned valve face. Figure 14 shows a burned valve
face, Figure 15 shows scored piston heads, and Figure 16
a scored wobbler bearing plate. Careful pressure loading
design has practically eliminated pumping instability due to
valve friction.
Ball and roller bearings were used in all pumps

tested and surprisingly have been almost free of trouble. The
most critical bearing is the wobbler bearing which is under
variable loads at speeds up to 5000 rpm. In approximately
50 total hours of test runs, no failures or excessive wear
have occurred other than one unit in which the bronze
retainer rubbed on the race and produced bronze powder.
Under conditions of piston sticking, the wobbler face can be
badly scored as was shown in Figure 16. This is repaired
by lapping the face flat which restores the bearing to a
usable condition.
Bearing failure can easily happen,as shown in Figure 17.

This is a 440 stainless steel bearing. This pressed steel retainer
-205-
failed after only 4 hours of operation. Our overall exper-
ience using several types, sizes, and loading conditions on
anti-friction bearings would indicate this to be one of the
lesser probleIn areas except for corrosion resistance.
Two basic types of shaft seals were tested with

successful operation. The original cOInposition, spring
loaded, lip type shaft seal in the first pUInp soon leaked due
to cracking of the COInposition. These were replaced with
siInilar seals containing genuine leather sealing eleInents and
have operated satisfactorily to date. The latest pUInp
incorporates a spring and pressure loaded axial shaft seal
which utilizes two static O-ring type seals and a lapped face
sealing surface. The bronze on hard steel lapped surface
quickly galled and hiled. The saIne seal has been Inade
operable by using two leather O-rings instead of rubber and
by ceInenting a lapped laIninated plastic face on the sealing
surface.
The overall corrosion probleIn is one which will

involve a detailed developInent prograIn. In aircraft
cryogenic PUInpS, the unit is required to stand idle in a
corrosive atInosphere for extended periods. Another require-
Inent is to run without daInage pUInping only gas, at least for
periods of a few Ininutes. Although the austenitic stainless
steels are corrosion resistant, they are too soft for use as
ball bearings or pistons. The harder Inartensitic steels used
so Inuch in conventional pUInpS are brittle at liquid nitrogen
teInperatures. However, this effect has caused no failures
as yet.
Most of these probleIns have been solved and the

latest pUInping test results indicate good operation. Figure 18
shows recent pUInp perforInance. The voluInetric efficiency is
73 per cent at 3000 psi. This particular test was run with
no piston rings which indicates that the valving leakage has
been Inuch reduced and that piston leakage is less than 25 per
cent of total flow. The use of prope r pis ton rings is
expected to increase the voluInetric efficiency to a very
satisfactory level. The eliInination of suction instability and
piston pick-up has allowed test runs of Inore than one hour
without wear. At present, the Inanufacture of a pUInp
designed specifically for liquid nitrogen is proceeding.
Study of the data published by the National Bureau of

Standards, Ohio State University, the University of California
-206-
Cryogenic Laboratories and the NACA has been of great
value. In the design of mechanical devices involving stress,
friction and wear, heat transfer, corrosion, instrumentation,
etc., many unanswered questions have been raised. In
several cases finding the answer by test has been difficult.
We would like to suggest and encourage an even more
extensive pursuit of basic cryogenic engineering data and the
further dissemination of information such as this conference
provides.
Discussion
Question by B. S. Tucker, C. H. Wheeler Manufacturing

Company: Are standard bearings used in the pumps?
Answer by author: The ball and roller bearings used were

the same as originally used in the pumps. As stated, very
little difficulty was had with anti-friction bearings. I
believe that slight modifications, especially of the retainers,
could make better cryogenic bearings.
Question by R. L. Johnson, NACA: Were the leather lip seal

materials specially treated or handled to prevent moisture
absorption prior to use?
Answer by author: The leather lip seals obtained were

standard and come impregnated with some type of grease, I
believe. We washed the bearings as free of grease as
pOSSible with solvent. We did not use any special handling
except to avoid touching the leather with our fingers.
Question by B. S. Tucker, C. H. Wheeler Manufacturing

Company: Could you give an estimate of the bearing
loading?
Answer by author: Most of the pump bearings are rather

lightly loaded except for the wobbler bearings which can be
highly loaded at 3000 psi. On the last pump there are a
maximum of 5 active pistons of 1/4 inch diameter. The
force per piston is 150 pounds times 5 or 750 pounds divided
by the cosine of 18 degrees, the wobbler angle·. The pitch
dlameter of the balls is about 1-3/8 inches.
-207-
Question by J. E. Boretz, Stratos -Division Fairchild Engine
and Airplane Corporation: What was the suction specific
speed for this pump and what was the minimum net positive
suction head required?
Answer by author: I do not know the suction specific speed

but this parameter has almost no meaning for positive
displacement pumps. It can be figured from the flow of
2 gpm at 4000 rpm.
We have not measured the net positive suction head yet

because our millivolt potentiometer is only accurate to
about 3 0 F. We have just procured a microvolt potentiometer
for this purpose. However, we have run the pump many
times with the liquid nitrogen in the tank boiling at a level of
8 feet above the pump inlet. From this I would guess that
our NPSH can be lower than 8 feet of liquid.
Question by J. M. Novajosky, Allison Division General

Motors Corporation: What materials were found to be most
compatible with the 52100 pistons? What was the hardness
of these parts?
Answer ~y author: The best cylinder block material tried

was also 52100 steel. Both the pistons and block are 60 to
64 Rockwell C scale in hardness. This is the same material
and hardness used in ball bearing races.
Question by M. P. Sweeney, Process Associates: Would you

tell us more about the leather or composition seals? Who
supplies them?
Answer by author: I would like to point out that the leather

seals used were merely for the purpose of completing tests
on the existing pumps. The axial pressure balanced shaft
seal mentioned is probably a better type for aircraft appli-
cation.
The seals we used were the spring loaded lip type, genuine
leather sealing element type. They are identical to the more
commonly used composition sealing element units generally
used in automobile front wheel bearing seals. These are
available at hardware or automotive supply houses and are
standardized sizes. At least three companies manufacture these.
-208-
D-6
OPERA TION OF BEARINGS AND PUMPS AT

LOW TEMPERATURES
K. B. Martin, R. B. Jacobs, and R. J. Hardy

Boulder, Colorado
Two years ago the Cryogenic Engineering Laboratory

of the National Bureau of Standards started a program to
study pumping of liquid hydrogen at low heads. The primary
application was for a pumping unit to circulate 100 gpm of
liquid hydrogen with a pressure drop of 5 psig through the
cooling passages of an electromagnet. Preliminary work
involved the modification of water pumps for pumping
cryogenic fluids and these investigations have been described
earlier. 1
The program was broadened to study performance

and cavitation characteristics of centrifugal pumps in other
liquefied gases as well as liquid hydrogen and to evaluate
bearings and seals. Some of the results of these investiga-
tions were reported at the 1956 Cryogenic Engineering
Conference. 2
The purpose of the present paper is to discuss the

results of the study during the past year. The report
includes: bearing data obtained from both the bearing tester
and the pump test facility; the perfection of a reliable test
facility by the addition of accurate instrumentation to obtain
head, capacity, efficiency, Net Positive Suction Head, and
liquid level; and some data concerning the operation of the
test facility.
lR. B. Jacobs, K. B. Martin, G. J. Van Wylen, and

B. W. Birmingham, National Bureau of Standards Report No.
3569.
2 K . B. Martin, R. B. Jacobs, and R. J. Hardy,
Proceedings of the 1956 Cryogenic Engineering Conference,
p. 295, Boulder, Colorado, September 1956.
-209-
Bearing Data
Data have been obtained under actual bearing load

conditions on the ball bearings used in an aircraft fuel
booster pump that is being tested in liquid hydrogen and
liquid nitrogen. These data indicate that a ball bearing
with a non-metallic retainer has longer life than a similar
ball bearing with a metallic retainer. The same results have
been obtained from the special bearing tester under essentially
no load conditions, as shown in Figure 1. In the pumping
apparatus, a bearing was replaced if the bearing appeared
rough or if destructive failure of the bearing occurred. In
the bearing tester, when the two bearings being tested
transmitted a frictional torque of 1. 8 inch-pounds, the
bearings were said to have failed. However, destructive
failure has occurred before the bearing friction reached this
value. All the bearings tested so far have been submerged
in liquid.
In the pumping apparatus, Figure 2, a total of 12

pairs of bearings with metallic retainers have been used for
an average length of time of 8 hours and 50 minutes at speeds
ranging from 3200 rpm to 6100 rpm while 6 pair s of the same
type of bearings with non-metallic retainers have been used for
an average of 20 hours and 32 minutes at speeds ranging from
3600 rpm to 8000 rpm. All the bearings used, however,
operated under very severe conditions. Between some runs,
the cold apparatus was exposed to the atmosphere and
moisture collected on all of the cold surfaces including the
bearings. The moisture was evaporated by warm sunlight
and/or an electric heater as rapidly as possible, but some
rusting occurred. Some bearings were replaced even if they
only appeared rough; the others sustained destructive failure.
The only destructive failures so far have occurred

with bearings having metallic retainers. In the pumping
apparatus the only destructive failure occurred in a bearing
having clipped retainers, but destructive failures in the bear-
ing tester have occurred with bearings having both clipped and
riveted retainers. Bearings having clipped retainers have
failed more rapidly than bearings having riveted retainers.
The wear begins first between the ball and the retainer. The
destructive failure occurs from wear through the retainer
and/or vibration fatigue of the brittle rivets. No actual
failure has been recorded in bearings with a non-metallic
-210-
DRIVE
PULLEY
SHAFT -;---t-I-...-
PUMP
DISCHARGE
VALVE
ORIFICE
HEAT
EXCHANGER
BYPASS
VALVE
CATALYST
CHAMBER
TEST B~~RIIIIGS·1'-..
Figure 2. Pump test

facility schematic
Figure 1. Low tempera-

tur e bear ing te s te r
-211-
retainer. In fact, a set of these bearings is still being
used in the pumping apparatus. This set of bearings was
installed February 18, 1957, and the lower bearing was
changed on March 12, 1957. The upper bearing has operated
about 60 hours while the lower bearing has operated a little
over 52 hours. The bearings still operate very smoothly.
Two upper bearings used in the pump with non-

metallic retainers are also being tested in the special bearing
tester. To date these bearings have run for over 1500 hours
at 3250 rpm in liquid nitrogen with no appreciable increase in
torque above one inch-ounce. These bearings are, however,
running relatively unloaded.
The maximum life of a pair of similar bearings with

riveted metallic retainers was 651 hours; the retainer s of
these units failed completely. Our tentative conclusion is
that unloaded ball bearings of the R-12 type with metallic
retainers should not be expected to run unloaded at about
3000 rpm for more than 200 to 300 hours.
This preliminary testing indicates that the non-

metallic retainers are more suitable for operation in liquid
nitrogen and liquid hydrogen than similar bearings with
metallic retainer s.
Instrumentation of the Pump Facility
The pumping apparatus instrumentation has been

improved by the use of Foxboro recording instruments for
measuring head, flow through an orifice, NPSH, and
hydrostatic liquid level. Three instruments are utilized.
The first instrument is a type 9 single pen differential
pressure instrument with a range of 0 to 2.5 inches of water.
This instrument is used to measure liquid level and NPSH
below 35 inches of liquid hydrogen to an accuracy of ± 1/4
per cent of the full scale deflection. The other two instru-
ments are type 27 double pen instruments, recording one
gage pressure and one differential pressure each. These
instruments are used interchangeably to indicate and record
head, capacity, and liquid level with a spare or adjustable
sensing device to permit changing the differential pres sure
ranges. the ranges were selected so that pump characteristics
-212-
in liquid nitrogen, hydrogen, and oxygen could be investi-
gated. The type 27 instruments have been very satisfactory
and have the same full scale accuracy as the type 9 instru-
ment.
The efficiency measurement is obtained from the

boiloff of the heat exchanger through alar ge capacity gas
meter. This indicates the power going into pump work 2.S
shown in Figure 1.
Instrumentation for determining NPSH with cryogenic

fluids is perhaps the most difficult problem encountered. By
definition the NPSH of a pump is the equivalent total head at
the pump centerline minus the vapor pressure. The total
head at the pump centerline is reasonably easy to determine,
but an accurate vapor pressure measurement is more difficult.
Gold-cobalt thermocouples were used at first to measure the
temperature of the liquid being pumped; then the vapor
pressure was determined from vapor pres sure data. How-
ever, as these thermocouples are only accurate enough to
correct vapor pressure to 16 feet of liquid hydrogen, a more
accurate determination is required. A vapor pressure
thermometer or bulb has the necessary accuracy.
It is the vapor pressure of the liquid in the bulb that

is actually desired, so the vapor bulb should not be considered
a thermometer since a temperature reading is not obtained.
Two methods of determining the vapor pressure were investi-
gated. In the first method, two vapor bulbs could be Ufed with
a differential sensing device to indicate the magnitude of the
pressure differences. One bulb would of necessity have to be
located in a reference bath of liquid hydrogen boiling at
atmospheric pressure. The other bulb would be located near
the pump suction. In the second method, the bulb located
in the reference bath is replaced by an open tube which
measure s the total liquid head at the pump suction. Again, a
vapor bulb is located close to the pump suction to indicate
the vapor pressure of the surrounding liquid. The differential
reading here, however, is a direct measurement of NPSH.
Therefore, the second method was selected. The instrument
is described fully in an earlier publication. 3 The limiting
factor of this type of device is the range and accurracy of the
differential sensing device.
3R . J. Hardy, K. B. Martin, R. B. Jacobs, National

Bureau of Standards Report No. 5500.
-213-
In the preceding discussion it was assumed that the
liquid being pumped and the liquid in the vapor bulb have
the same composition. This is not necessarily true; diffi-
culty was experienced in this respect with both liquid
hydrogen and liquid nitrogen. With liquid hydrogen a number
of runs indicated a negative NPSH. Therefore, the vapor
pressure in the vapor bulb was greater than the pressure of
the liquid head above its pump suction. The liquid level
tube and the static NPSH tube were interchanged to check
proper functioning of the instrumentation lines; the same
negative NPSH reading was observed. This negative reading
was believed due to a difference in ortho-para composition
of the liquid hydrogen in the vapor bulb and pump dewar.
Catalyst chambers were therefore installed in the vapor bulb
and the pump dewar to assure that liquid of the same compo-
sition was in both places. A small negative NPSH value was
still observed indicating that complete conversion did not
occur.
An ortho-para analysis system was therefore needed

to measure small ortho-para change.s. The analyzer was
built using a d-c system described by Weitzel and White. 4
With this equipment, the difference in para concen-

tration between the vapor bulb and the circulated liquid was
measured during liquid hydrogen runs. As the liquid from the
pump discharge was being flashed to atmospheric pressure,
the NPSH should have been equal to the liquid level. After
the resulting difference, however, was corrected for ortho-
para conversion, the NPSH reading differed from the liquid
level reading by only I /10 of an inch of hydrogen. This
agreement was perhaps fortuitous, but indicated that our
deductions were correct. This analyzer was not, however,
basically accurate enough for our work. Para changes below
1 per cent could not be seen accurately with the d-c system,
and at least 1/10 per cent was required to correct to less
than 1/4 of an inch of liquid hydrogen.
A new analysis system was developed using a-c

across the bridge and this has been described fully in paper
D-4 of this conference.
4
D. H. Weitzel and L. E. White, Rev. Sci. Instr.
26, 290 (1955).
-214-
Operation of the Test Facility
The aircraft fuel booster pump being used in the

pumping apparatus has been operated for more than 147
hours in cryogenic fluids at speeds from 3200 rpm to 8000
rpm. Of this total, the pump operated over 22 hours in
liquid hydrogen. During this testing, the data obtained have
only served to indicate the tendencies of cryogenic fluids to
cavitate. The runs made, however, were instrumental in
perfecting the instrumentation accuracy es sential to precise
measurement of NPSH and composition in cavitation studies.
More complete liquid hydrogen data will become

available after installation of the new ortho-para analysis
equipment.
Discussion
Question by R. L. Johnson, NACA: Was there any evidence

or quantitative data of ball skidding or wear in no load
bearing runs? Under low load conditions friction of
retainers can exceed tractive force on rolling elements and
cause skidding that is apt to result in wear of rolling element.
Answer by author: The bearings were only inspected visually

with a magnifying glass, no quantitative data were obtained
from the bearings. It is reasonable to believe that skidding
of the balls did occur as some wear was evident between the
balls and races.
Question by R. L. Johnson, NACA: What type of non-

metallic retainer was used?
Answer by author: It is not known what kind of retainer

material was used, but this information can be obtained by
contacting the manufacturer.
Question by R. L. Johnson, NACA: What loads were imposed

on the bearings during the 22- l/2 hours at 3600 to 8000 rpm
described as very severe conditions?
-215-
Answer by author: The loading on these bearings is not
exactly known but about 88 per cent of the total running
time on the bearings with non-metallic retainer was done
in liquid nitrogen where the motor was pulling between 500
and 600 watts. The severe conditions described, however,
are those described in the conference paper where moisture
is condensing on the bearings causing some corrosion. In
addition, any particles of moisture which are not evaporated
freeze to the bearing surfaces and cause the bearing to be
very rough.
Question by S. Greenfield, Rocketdyne: Did I understand you

correctly that the gold-cobalt thermocouple permitted meas-
urements of temperature equivalent to 16 feet of liquid
hydrogen head?
Answer by author: Yes, from vapor pressure data for

hydrogen. If you can read accurately to the nearest 2 micro-
volts this is equal to approximately 16 feet of liquid hydrogen
head. (1 microvolt equals 8.5 feet of liquid hydrogen.)
-216.
D-7
1000 GALLONS PER MINUTE LIQUID

OXYGEN TRANSFER UNIT*
D. W. Seavey
Lowell, Massachusetts
The Cambridge Corporation has designed and manu-

factured a transportable, compact skid mounted liquid oxygen
transfer unit with a pumping rate of 1000 gpm at 100 psig
discharge pressure. As a secondary function, the unit supplies
liquid oxygen at variable flow rates from 10 to 250 gpm. The
main flow is provided by two 500 gpm single stage centrifugal
pumps operating in parallel. A smaller third pump provides
the variable flow.
Design criteria include remote electrical control of

pumps and valves, a requirement for flow in either direction
through the unit, and limitations on horsepower and space
envelope. The required testing program included operational
and life tests using liquid oxygen. Hence, a significant
proportion of the design effort went into testing equipment and
procedures.
The desire for continuous test operations at the 1000

gpm flow rate led to the use of a static head tank vented to
atmosphere. The tank is elevated to provide the necessary
NPSH for pump operation. A remotely actuated throttle valve
in the liquid oxygen return line controls unit operation and
permits the liquid to enter the top of the head tank at
atmospheric pressure.
The test loop is instrumented to measure liquid oxygen

vapor pressure at various points and to record pressures and
flow rates.
>'''Abstract of paper presented. For additional informa-

tion contact the author directly.
-217-
D-8
CONT AMINANTS IN LIQUID OXYGEN AS RELATED TO

SAFETY IN LIQUID OXYGEN PRODUCTION AND
DISTRIB UTION EQUIPMENT
C. D. Arrick
Linde Company
Tonawanda, New York
Introduction
If suitable precautions are taken in production and

distribution, there is very little hazard from the traces of
combustible contaminants present in liquid oxygen. However,
the se contaminants could contribute to serious fire s or
explosions if they are allowed to increase to dangerous levels.
Combustible materials may get into liquid oxygen from

the atmospheric air processed, from oil lubricated air
compressors, from improperly cleaned equipment, and from
equipment parts such as gaskets, packings, etc. This
discussion is limited to the contaminants introduced from
the air processed in a low temperature air separation plant.
The bulk of the combustible materials in the liquid oxygen
product usually comes from this source if reasonable atten-
tion has been paid to the other sources.
It is quite practical and desirable to keep the concen-

tration of combustibles in liquid oxygen far below dangerous
levels. It is difficult and unnecessary to eliminate them
completely.
Clas sification of Combustible Contaminants
The number of combustible contaminants, in the air

processed, is practically unlimited at industrial sites. A
large amount of information has been obtained on individual
contaminants and their potential hazard. Fortunately it has
not been neces sary to consider them individually in the
design and operation of equipment. A simple, practical
-218-
classification of them has been developed with an eye toward
the potential hazard from the various categories. They have
been divided into three classes called the "volatiles," the
"non-volatiles, \1 and the "acetylenic" hydrocarbons.
The lower limits of flammability of highly soluble and

relatively volatile combustibles (for example, methane,
ethane, and ethylene) in liquid oxygen are very roughly in the
same range as their limits in gaseous air on a mol per cent
basis. Their solubilities are so high that even at the most
polluted plant sites, their solubility limits or lower limits of
flammability are never approached. Our main concern is
with materials which are hazardous in the parts per million
range. The heavier, saturated, and relatively non-volatile
hydrocarbons and the unsaturated hydrocarbons have low
solubilities in liquid oxygen. When the solubility limit is
exceeded, even though the average concentration in a body of
liquid oxygen is considerably below the lower limit of
flammability, there are local concentrations which can burn
rapidly or explode. The highly unsaturated (acetylenic)
hydrocarbons are the least soluble and least stable and there-
fore, potentially the most dangerous.
TABLE I
Classification of Hydrocarbons
Potential Hazard
Class Materials Included in Typical
Liquid Oxygen
Volatiles Methane, ethane, Very little

ethylene
Non-vo la tile s Propane and heavier Some

saturates, propylene
and heavier unsaturates,
etc.
Acetylenic Acetylene, propyne, etc. Considerable
-219-
Table 1 shows the simple classification of the hydro-
carbons with respect to their hazard. These categories were
established by test. The analysis for volatiles consists of
the determination of all hydrocarbons having appreciab Ie vapor
pressures (above a few microns) at about _180 o C. The non-
volatiles are those hydrocarbons with low vapor pressures at
-180 o C. The acetylenic hydrocarbons are specifically deter-
mined separately.
Contaminants in Process Air
Table II shows the results of analytical surveys made

on the atmospheric air at plant sites. In essentially all
liquid oxygen plants and in distribution equipment the volatile
hydrocarbons from the process air present no problem. The
non-volatiles may require some attention in addition to their
normal reduction in pas sing through the low temperature air
separation cycle. A portion of these materials is prevented
from reaching the product by freeze-out in the low tempera-
ture heat exchangers and elimination by periodic heating and
purging. Additional reduction may be obtained by scrubbing
all the process air by the liquid fraction formed, and by
mechanical filtration of the liquid fraction. This is much
more effective than filtration of the liquid fraction without
first using it to scrub contaminants out of all the process air.
TABLE II
Hydrocarbons in Atmospheric Air at Plant Sites
Class Range Average

---
Volatiles 1-20 ppm 2 ppm
Non-volatiles 0.2-4 ppm 0.5 ppm
Acetylenic 0.001-0.100 ppm 0.01 ppm
Where this is not sufficient to keep concentrations of non-

volatiles well below the solubility limits in the liquid oxygen,
additional removal by adsorption on silica gel is employed.
The latter method is also used where the acetylenic hydro-
carbon concentration is high in the process air. An effort
-220-
is made to keep extraneous combustible contaminants out of
the process air by strategic location of the plant air intake
and by regulations preventing contamination of the process
air with combustibles such as paint or solvent fumes, etc.
Contaminants in Product Liquid Oxygen
The hydrocarbon content of a typical liquid oxygen

product and the solubilities of the classes of hydrocarbons
is shown in Table III. Comparing the typical hydrocarbon
contents to their solubility in liquid oxygen, it is obvious
that it would be necessary to concentrate this material by a
very large factor to produce a hazard. If the acetylenic
hydrocarbon content of 0.05 ppm was all acetylene (solubility
5. 7 ppm), the liquid would have to be concentrated by a
factor of 114 to produce a hazard. This is, however, a
distinct possibility in improperly designed or operated liquid
oxygen vaporizers.
TABLE III
Hydrocarbons in Liquid Oxygen
Typical Solubility
Class Concentration Limits
(At about _l80 o C
Volatiles 10 ppm > 4 per cent
Non-volatile s 2 ppm From traces

up to 4 per
cent
Acetylenic 0.05 ppm < 10 ppm
Production Equipment
The most critical component in the air separation plant

is the main condenser or condenser-reboiler. Its purpose is
to condense nitrogen vapor at the top of the high pressure
column, against boiling liquid oxygen, for use as reflux in the
-221-
high and low pressure columns. It operates with a small
temperature difference between the condensing nitrogen and
boiling oxygen. The liquid oxygen product, constituting
about 20 per cent of the process air, is withdrawn from the
body of boiling oxygen in the main condenser. This limits
the average concentration of hydrocarbons in this liquid to a
maximum of about 5 times that in the air processed,
assuming no reduction prior to the main condenser. This
is basically a safer system than in a gaseous oxygen plant
where the product is withdrawn as vapor and additional
hydrocarbon removal provisions, such as silica gel traps,
are required to limit the combustibles in the main condenser
liquid. Concentration by a factor approaching five normally
is quite safe and results in the typically clean liquid shown
in Table III. There is, however, the pos sibility of unlimited
local concentrations of combustibles in this heat exchanger.
The liquid must not be evaporated to dryness or unduly con-
centrated at any point since solubilities at that point will be
exceeded and there is the danger of local accumulations of
solid hydrocarbons. It should be stressed that even in a
liquid oxygen plant, where a substantial quantity of liquid is
continuously drained from the main condenser, considerable
experience is required to establish safe design and operating
procedures. At no point should a condition be created which
would permit the concentrating of combustibles to levels
appreciably above the average concentration in the body of
liquid oxygen.
Figure 1 illustrates two major types of main condensers

used in liquid oxygen plants. The earlier design had the
condensing nitrogen inside the condenser tubes. The tube
bundle was then immersed in the body of boiling oxygen
contained by the condenser shell. With unusual air contamina-
tion, operating with too low an oxygen liquid level (such that
boiling to drynes s at the top ends of the tubes was likely), or
with temporary high boiling rates, accumulation of solid
combustibles ill the relatively large body of liquid oxygen could
occur. A few explosions occurred in the early years which
were violent enough to rupture the vessel. More recent
designs have the boiling liquid oxygen inside the tubes. They
are operated so that the tube walls are always completely
wetted and liquid is discharged from the tubes and recircu-
lated with the liquid entering the reboiler to obtain good
re-mlxlng. If combustible solids are deposited, due to opera-
ting with too low a liquid level or the presence of other
-222-
02 Vflf"'O~
h~ ,.d~
R. E X,'9MP,"cS Or ,P~ope~,"y 7ft~ppe.:J ilQ"leJ CtJA./A./cCT/(i)Al5

B. eX/7Mf"',cS Or IMF'ROpe~ Llt((.//L) COAlNcC7/0,uS
Figure 1. Major types of main condensers

in liquid oxygen plants
-223-
foreign material such as solid carbon dioxide, water, etc.
(which can plug a condenser tube) the type of equipment
failure would be relatively mild and would rupture only a
few tubes, with no danger to personnel.
Attention must be paid to the proper trapping of

pIpmg on a vessel containing liquid oxygen. There must be
no trapped pockets where liquid may boil and the non-volatile
hydrocarbons concentrate. Figure 1 shows some examples
of proper and improper piping connections. In a proper
connection liquid can enter only a small section of the pipe
next to the vessel. If the liquid tends to proceed further
into the pipe, heat leak into the pipe will vaporize the liquid
and force all of it back into the vessel. In an improperly
designed connection, liquid will enter the pipe and not be
forced back by the vapor.
Periodic analysis of the hydrocarbon concentrations

in an oxygen plant must be exercised. If attention to the
various means of keeping their content low does not correct a
high hydrocarbon content situation, the main condenser is
drained rapidly or as the ultimate measure, the plant is shut
down. Periodic, complete warm -up and purging of the equip-
ment is practiced as an additional plant safety measure. The
final precaution on the liquid oxygen product consists of
passing it through mechanical filters before it enters storage
or distribution equipment.
Normal external sources of ignition, such as

smoking, open flames, sparks, and impacts are not allowed
near oxygen equipment. However, it has not been neces sary
to eliminate all ignition sources which have been suggested as
inherent in the boiling oxygen system, such as the catalytic
effect of traces of the oxides of nitrogen or ozone. The
philosophy is to maintain the combustible contaminants in the
equipment well below the concentrations which can react, and
to pay reasonable attention to the elimination of sources of
ignition.
Distribution Equipment
Similar precautions must be taken in distribution

equipment. Liquid oxygen in a tank is not permitted to boil
off to dryness. The bulk of the product withdrawal is made
in the liquid phase. In vaporizer s, the principle of flash
-224-
evaporation is used. Large temperature differences in
stearn, hot water, or atmospheric vaporizers flash the
liquid and the contaminants so that the latter are not concen-
trated and solid combustible s cannot be deposited on the heat
transfer surfaces.
With proper equipment, and if only a typically clean

liquid is handled, the distribution equipment is quite safe
from the standpoint of hazard from the trace s of contaminants
present in the liquid oxygen product.
Discussion
Question asked by R. Hampton, Lox Equipment Company:

What is the reason for having the oxygen boil inside the
tubes of the reboiler?
Answer by author: One reason is that this design is

basically more safe. If hydrocarbons do build up in the
reboiler and are ignited, usually only a few tubes would be
ruptured with no hazard to personnel. The small tubes can
contain an explosion better than a large diameter shell.
Also in this design the quantity of material (liquid oxygen
plus hydrocarbons) is relatively small per unit of reboiler
capacity and consequently there is less total material to
burn.
Question asked by W. A. Wilson, CEL National Bureau of

Standards: Is the typical liquid oxygen product you described
also the liquid oxygen used for breathing purpose s?
Answer by author: Yes, it is.
-225-
E-l
A CRYOSTAT FOR LIQUID NITROGEN COOLING OF

ROCKET-BORNE PHOTOMULTIPLIER TUBES
J. E. Tracey
F. K. Dearborn
Air Force Cambridge Research Center
Bedford, Massachusetts
The improved performance of certain electronic

devices due to operation at low temperatures has been known
for some time. One such device is the multiplier phototube.
This tube is approximately 1 inch in diameter and 3 inches
long- very similar in appearance to an ordinary radio tube.
Sensitivities in photomultiplier tubes can be greatly

increased by operation of the cathode at liquid nitrogen
temperatures-some 400 0 F colder than customary operating
conditions. Reduction to practice of this laboratory observed
phenomena has resulted in a unique, compact, liquid nitrogen
cooled photomultiplier cryostat.
The cathode of the photomultiplier tube (lP2l) is

coated with a light sensitive oxide which is excited by
incident light and emits electrons. These emissions are
amplified to produce a measurable current.
The extremely high amplification fac;:tor of the photo-

multiplier tube, principally responsible for its success,
permits the extension of measurement to lower light fluxes
than is pos sible with an ordinary photo tube, yet this same
factor acts as a deterrent to accuracy when measuring very
low light fluxes.
The Air Force Cambridge Research Center is involved

In investigations of the upper atmosphere. One method of
investigation has been the construction of a spectrum of that
portion of the atmosphere in which they are most interested.
This has been achieved by placing a multiplier photo tube in
-226-
the nose of an "Aerobee" rocket. A weak light signal is
passed through the atmosphere and received by the photo
tube while the rocket is in flight, and from the signal
emissions of the phototube a spectrum can be constructed.
Figure I shows portions of film from a rocket-

borne scanning spectragraph using multiplier photo tubes at
ambient temperatures. The fir st bank was obtained before
flight. The spectrum is clear and the synchronization line
(the dark line on the right) is steady. The second bank was
filmed during the start of the flight and the spectrum is still
readable. However, this is not the area of the atmosphere
of greatest interest to the researchers. The third area
apparently results from the phototube being adjusted auto-
matically to detect the lower light levels. The unintelligible
blurriness is caused by thermionic emission and effectively
covers up whatever might be present in the light spectrum
of the upper atmosphere. It is thi s most important area
that will be improved by using a refrigerated tube.
In any application of the multiplier photo tube to

very low light levels one of the most important considerations
is of three types of dark current. Each of these is respon-
sible for background noise which acts to conceal the signal
emissions of a low light flux. They are due to ohmic leakage,
regenerative ionization, and amplified thermionic emission.
The first two, ohmic leakage and regenerative ionization have
been virtually eliminated by improved production methods and
controlled voltage. Thermionic emission, amplified by the
same factor that makes the multiplier photo tube desirable,
remains as the only obstacle before operation at maximum
sensitivity can be obtained.
Thermionic emission at room temperature amounts

to approximately 5000 electrons / sec/ cm 2 • The reduction of
this emission to near zero is thus the main objective.
R. W. Engstrom 1 of R. C. A. has discussed experi-

ments to determine if thermionic emissions could be reduced
1
R. W. Engstrom, J. Opt. Soc. Am., 36, 6 June
1947.
-227-
0
VOlTSPCASTAG[
TU8E T'fPC IP21
~IOO r _l I .! Iotre
j
10
8" OW1 0l7 or r:'tAl.'fSEJtf!j= I, ~ Us.
IIlIb R['-C~[NCt Odb =1 IIIICR0\0I'0\.1 ACROSS I
~ La...
L.04t)
.....
....
"!t1GN.Al.. OUTPVT • I ,
or- .cATlooc ,LU-"..1,lA1J
IV :I
0.
LUM[N,!.". 10
C.OLOR TE t.'P[fU,TURC :.a81cf'1'C
CHopprO . 90CJ'A f ~ ,
.....
....
f
O-
----
I/~ ,
0 ~
V'
/
'0
Figure 1. -, 0. . ~o -100 .... 110 -120 -eo -.0 Q 40 ao 120

TueC U:: ~PIERA;TuRE - D[CREES C[ NTICAAO£
Figure 2. Signal and noise output

of photomultipliers versus temperature
, 3 · "1"
"IfC
",Noo.
tlASltc T
Figure 3.
' 4 0/1 IIUT
(T.O LlNESI
Figure 4. Photo-
multiplier cryostat
-228-
significantly by refrigeration of the tube. The curve of
Figure 2 is a result of that work. It shows that this form
of dark current may be virtually eliminated, and except
for random emis sions, there will be little or no nois e to
interfere with accurate measurement of very low light fluxes.
A direct comparison of low light signals amplified by

cooled and non-cooled photo tubes can be seen in Figure 3.
The signal did not pass through the same medium in each case
but the thin lined, distinct peaks and valleys of the cooled
photo tubes (on the right) are an obvious improvement over
the thick, indistinct curves of the warm tubes as shown on
the left.
The first approach to cooling the multiplier photo tube

was study of the feasibility of extracting heat through leads
extending from glass to metal seals at the base of the tube.
Immersion of the leads in liquid nitrogen proved that this
method was not a satisfactory solution. The glass to metal
sealing materials and the leads, because of their composition
and small cross sectional area were very poor conductors.
With the leads held at -320 o F, the equilibrium temperature of
the cathode was +32 o F, under conditions of ambient radiation.
Next the cathode was cooled directly with a good

thermal conductor. Use of a good heat conductor for direct
cooling of the cathode has the advantage of not requiring
additional glass walls as vapor barriers. The experimental
feasibility of cooling the cathode by this means was defini~ely
established. Although modifying the tube itself is not
desirable, the thermal conductor approach was appealing
enough to warrant consideration of manufacturing a special
tube of this type. A prominent manufacturer indicated that
a conductively cooled tube would be both expensive and long
in the making.
Abandonment of the thermal conductor approach left

only one path to follow-radiant cooling. With some selected
tubes from Air Force Cambridge Research Center, construc-
tion of a special dewar cell to contain the tube with a window
for light transmission was started. In the first test of this
dewar, thermal measurements indicated temperature gradients
throughout the tube. However reduction of thermal emis sions-
the real objective - was apparent, and in an operational test
the dewar proved successful.
-229-
In removing the refrigerated tube apparatus from the
laboratory to the nose of an llAerobee ll rocket, the construc-
tion requirements such as those for vibration, shock,
acceleration, weight, and particularly space were strict.
This entailed some changes in the laboratory model, the
r~sults of which are shown in Figure 4.
Because of space restrictions the unit was designed

primarily from the outside, working inward. High vacuum,
powder, and glass wool were considered as possible insula-
tions. High vacuum insulation in itself was discarded· in
favor of a powder or glass wool in conjunction with a vacuum
in the order of 10 to 100 microns. Thus if a leak occurred
in flight, the dewar would remain insulated and the tube cool.
Glass wool was selected because when properly

oriented (with the fiber s perpendicular to the heat path) it is
superior to a powder insulation. The thickness of the
insulation is 3/8 inch.
As shown in Figure 4, a double walled cylinder

surrounds the tube to be cooled. This vessel is filled through
one of the two tubes which serve both as supports and as the
vent and fill lines, respectively. These lines are identical
and either may be used as the vent or fill line. A third tube
pierces the vacuum jacket only. This is the vacuum pump-
down line.
The window for incident light to the photo tube is

made vacuum tight by compression of a neoprene gasket. The
light path is kept free of glas s wool by a thin walled micarta
tube. A spring riveted to a centering washer at the bottom
of the tube well, isolates the tube from shock and acceleration.
Electrical connections are made through an octal type

socket. A heat station to circumvent heat leakage through the
leads is located approximately I inch back from the face of
the socket. The relative position of the socket and heat
station is maintained by filling the interim with an epoxy resin.
This also serves as an electrical insulator. The access hole
for the wiring and placing of the phototube is made vacuum
tight by crushing a .040 diameter indium wire.
Since the multiplier photo tube is manufactured with

its cathode in line with the key on its base, the positive
alignment of. the cathode, with the window in the cryostat has
-230-
been assured by the addition of a locating pin to the heat
station and proper assembly procedure of the cryostat.
Tube replacements can be made quite easily by

unbolting the flange and removing a bat of glass wool, and
the retaining ring which holds the tube captive. The only
disadvantage in this scheme is that the indium wire vacuum
seal has to be replaced each time the flange is unbolted.
This will be eliminated in a later unit by the addition of an
o -ring and an increase in the diameter of the flange.
The method of increasing the sensitivity of a
multiplier photo tube through refrigeration has been well
known in the electronics industry for many years, but the
removal of this observed phenomena from the laboratory to
the field, has awaited the need of a specific application.
Discussion
Question by T. Stearns, Cambridge Corporation: Was a

replenishment of refrigerant necessary during flight?
Answer by author: No. The capacity of the unit is I liter

and it requires approximately two hours of cooling to reduce
the cathode of the photomultiplier tube to the desired
operating temperature.
A gum rubber hose supplies the refrigerant to the cryostat

before flight. When the rocket takes off a shear block breaks
the rubber hose and the liquid boils off into the atmosphere.
It takes approximately one-half hour for the tube to warm up
and since the flight only lasts three minutes this method of
operation is more than sufficient to obtain the desired
results.
Question by B. W. Birmingham, CEL National Bureau of

Standards: Can you comment further on the apparent
superiority of the oriented fiber glass insulation?
Answer by author: I do not have a definite figure for a

comparison of the performances of evacuated oriented glass
wool and evacuated powder, but it has been observed at
Arthur D. Little, Inc. that evacuated glass wool does
outperform the evacuated powders.
-231-
E-2
BEECH AIRCRAFT 6000 LITER LIQUID HYDROGEN DEWAR
V. E. Isakson, C. D. Holben, and C. V. Fogelberg

Boulder, Colorado
A relatively large dewar designed specifically for

liquid hydrogen has now been produced and tested by the
Boulder Division of Beech Aircraft Corporation. This dewar
has a total volume equivalent to 6000 liters with some 5400
liters pay load.
The dewar and trailer, shown in Figure 1, is over

24 feet long, 7.5 feet wide and a maximum of 12.5 feet high.
The dewar proper is 12 feet in length and 7.5 feet in diam-
eter.
Figure 2 is a cross sectional view of the unit. The

inner tank is made of 304 stainless steel. It is 10 feet long
with an I. D. of 65 inches. Two aluminum baffles reduce
splash, strengthen the shell and help reduce temperature
stratification. Flanged and dished A. S. M. E. heads form
the cylinder ends.
The outer vessel is made of carbon steel. It is

designed to withstand an external pressure of 15 psi in
addition to loads imposed by highway accelerations. Two
rolled steel I-beams strengthen the shell and absorb loads
transmitted from the inner tank through the support rods.
The inner tank is supported in all directions with high

strength low area rods. The rods have right and left hand
threads for making adjustment. Correct adjustment is made
at the factory and safety wired to prevent maladjustment while
in service.
Approximately 12 inches of evacuated powder insulates

the liquid from ambient heat. No radiation shield of any kind
was used. The opaque quality of the perlite, together with
the near spherical shape of the dewar gives satisfactory
-232-
POWDER PIPING ACCESS L UID LINE
"[SERVOtR PluG
Figure 1.
£'lJACUAT[D VACUUM COU.E:ClIOH

P[ARlITE INSUI.ATION PIPING
Figure 2.
....i
r HEAT lEA~ TRENC
6000 LITER DEWAR
"'lilt \" .... ,,,",,au," ftol,
tlllr.lL,&fIO.ll ;IIIr.T
OPERATING PEfitFORIt4-A CE
BEECH AIRCftAf" T CORPOIltA TlQN
6000 LITER DEWAR
BEECH AIRCRAFT CORPOUTION

AUIU.' I, ..,
O.'::o-----,~:-;.-- .O'--"':-;---'~O;--.~
[Ptt ER vESSEL r)IAMETE:R
•.--------;,.O
rUt ·}·--~"-T-'~-~~·'-.-~-'''~'--~·O'--~~'--~'O'--"~
Figure 3. Figure 4.
-233-
performance without the added complexity of a radiation
shield.
The shape of the dewar has a significant effect on

the operating performance. The effect is demonstrated by
the curves shown on Figure 3, This graph represents
theoretical heat leak as a function of inner vessel diameter
and with various insulation thicknesses. It is interesting to
note that a 5. 5 foot diameter tank with twelve inches of
insulation will give less boil-off than a 4 foot diameter tank
with 18 inches of insulation. Still larger inner tank diameter
would have improved the performance. A larger diameter
was not selected because of height limitations and higher
center of gravity.
The inner tank was designed for a maximum operating

pressure of 150 psi. The A. S. M. E. Code for Unfired
Pressure Vessels was followed for design. The tanks are
hydrostatically tested to 250 psi prior to assembly. They
are protected against over pressure by a spring loaded
safety valve tapped into the vent valve.
A second pressure test is conducted on the inner

tank and connecting piping after completion of the unit. This
test is performed at low temperature using liquid nitrogen for
a refrigerant and nitrogen gas for pressurization. Liquid
hydrogen was used to test the first unit to prove the design
but was not used on later units to reduce cost.
Strict quality control is exercised on all phases of

manufacture to reduce gas leak into the vacuum space to an
absolute minimum. Leak rates have been surprisingly low
for this size ves sel. The total leak into the insulation space
from internal as well as the external container has been as
low as 2 x 10- 9 cc per second with rates of 5 x 10- 8 cc
per second being common. Some units have been in service
for over six months without re -evacuation.
The vent line, liquid line, and all gage lines enter
the inner tank through a 10 inch 1. D. access manhole. All
lines are thus readily acces sible in case repair or alteration
is necessary. This feature is especially desirable for an
experimental vessel as changes are often required.
-234-
The large access manhole is a potential source of
considerable heat leak. The heat leak is minimized by
channeling boil-off gas over a series of baffle plates below
the manhole plug. Heat entering the vessel is absorbed by
the boil-off gas as it is discharged from the vessel. This
system has proved very effective in reducing loss through
the manhole.
Instrumentation includes a thermocouple type vacuum

gage, three pressure gages, and a liquid level gage.
The instruments and panel lights are powered with a

6 volt battery carried on the unit. A battery charger
operated from 120 volt a-c power is also included to
recharge the battery when the dewar is not connected to the
truck circuit.
Each dewar is furnished with a 15 CFM vacuum pump

to re-evacuate the insulation as required. The pump is
driven with a one horsepower 120 volt motor. Power for the
motor and battery charger is supplied through a service cord.
The dewar and all auxiliary components are all

mounted into one complete package. This package is mounted
to a separate semi-trailer framework for highway transporta-
•
tion but can be removed for skid mounting if desired. The
trailer is suitably constructed to allow unrestricted highway
travel.
Performance of this unit has been most encouraging.

Measured boil-off rates have been better than expected. The
increased performance can probably be attributed to the heat
absorbed by the boil-off gas. This system appears to be
more efficient than predicted.
Figure 4 is an average curve of boil-off rates on

several vessels. All the units tested were within 10 per cent
of this curve. It can be seen that boil-off rates of approxi-
mately 1. 7 per cent can be expected on a stabilized unit.
Discussion
Question by R. Katzen, Air Products, Inc.: At what vacuum

were the heat leaks measured?
Answer by author: Off scale with a thermocouple gage.

-235-
Question by J. J. Brehany, California Oxygen Company: Was
the inner vessel ASME code stamped and did you run impact
tests on the inner vessel shell at liquid hydrogen tempera-
tures as called for by the code?
Answer by author: The code was followed in design but the

vessel was not code stamped.
Comment by C. D. Holben, Beech Aircraft Corporation:

The ASME Code Committee has ruled that cooling of the
specimen for impact test to 10 0 below operating temperature
has been waived. (Refer to Case # 1239. )
Question by W. W. Aton, Linde Company: What are transit

los ses of 6000 liter liquid hydrogen dewar?
Answer by author: This has not been measured.
Question by J. J. Brehany, California Oxygen Company: Was

this a double or single axle trailer?
Answer by author: It is a single axle trailer.
Question by L. R. Stoecker, Beech Aircraft Corporation: How

much stratification is there in the liquid phase in the 6000
liter liquid hydrogen dewar?
Answer by author: There appears to be considerable

stratification when the vessel is still.
Question by V. J. Johnson, CEL National Bureau of Standards:

You mentioned in your talk that you took advantage of cool
boil-off gas in shielding the manhole. Could this account for
the larger than expected pressure rise experienced on the road
while sealed off?
Answer by author: Yes it could.
-236-
Question by V. J. Johnson, CEL National Bureau of
Standards: What was the maximum pressure build-up in the
dewar? Was the first unit tested with liquid hydrogen?
Answer by author: The dewars were tested with cold

hydrogen gas to 250 psi. The maximum possible pressure
build-up could not exceed 150 psi due to pressure relief
devices.
Question by R. S. Warner, Cambridge Corporation: What are

the weights of the units involved?
Answer by author: 15,000 lbs.
Que stion by T. Stearns, Cambridge Corporation: You

mentioned that the manhole heat leak was reduced by baffling
vent gas through it. What is the difference between heat
leak when venting and not venting?
Answer by author: We have no test data at this time from

which to make a comparison. It will be considerably highe r,
however.
Question by H. L. Johnston, Herrick L. Johnston, Inc.: Is

it possible from the design of your vessel to measure the
heat leak into the liquid by allowing vapor to boil off through
a meter without passing through the cold baffling? Such a
boil-off, if measured, would give the heat leak corresponding
to the static condition with vent valve closed.
Answer by author: No. The dewar has only one vent system,
making this procedure impossible.
Question by V. L. Biggs, Douglas Aircraft: Was an attempt

made to take advantage of the higher steel strength at the
lower temperatures in the ASME code design?
Answer by author: The code makes no allowance for

increased strength. We followed the code.
-237-
E-3
VACUUM INSULATED DE WARS FOR LIQUID OXYGEN

AND LIQUID NITROGEN
H. A. Eichstaedt
Ronan and Kunzl, Inc.
Marshall, Michigan
This paper will present some of the problems and

their solutions for production of an efficient vacuum and
powder insulated container for liquid oxygen and liquid
nitrogen. The manufacture of a vacuum insulated liquid
oxygen container on a production line basis and how it was
done is best handled by defining each major problem that had
to be met and how it was solved to insure continuous quality.
Military specifications and the quantities required to

meet the tactical requirements of the armed services call for
removing the fabrication of vacuum and powder insulated
liquefied gas containers from a job shop to a production line.
Military requirements for a liquid oxygen container

can be divided into five major parts. The first was to
achieve the lowest attainable evaporation rate. The require-
ment for this evaporation loss was dictated by the tactical
situations in which the Air Force would be operating. Many
areas are far removed from commercial sources of liquid
oxygen and in areas accessible only by air. Islands in the
Pacific Ocean fit this category.
This leads to the second requirement for a military

container which was a container light in weight and suited for
air transport. This required the dewar to be built light
enough to keep the tare weight to a minimum and strong
enough to withstand constant handling abuse.
The third requirement was a tough and durable piece

of equipment, light in weight, capable of holding a vacuum.
The fourth part of the military requirement was peak

service efficiency at all times. The container had to have a
long service life. It had to hold vacuum not only for months,
-238-
but years, without re-evacuation. Repairing had to be a
simple and easy operation. Trouble shooting had to be kept
to a minimum.
The last factor of paramount importance was safety.

This factor required the units to be as safe as possible since
personnel with no knowledge of liquefied oxygen and its
characteris tics would be required to handle the equipment at
all times and in all places.
This many-sided problem was first approached by

considering the fact that pieces of equipment which should
be adapted to a production line cannot be produced by
craftsmanship (meaning that if a piece does not fit, the
craftsman stops long enough to shape it to fit). Production
flow, at as sembly line speed, had to continue if the
requirements of more than 100 units per month would be met.
The fabrication of vacuum and powder insulated dewars

wherein each unit would be as good as the first, presented a
challenge which was successfully met.
The problem of a container with the lowest evapora-

tion rate possible to attain was overcome by solving the
vacuum problem. New devices were required for incorporation
in the vacuum system. New vacuum techniques were devel-
oped for pumping the vacuum in the powder insulation.
Santocel "A" is used as the powder in the containers.
Since no leak in the vacuum system could be

tolerated, parts were designed which could be produced in
quantity with tolerances allowable for production work, yet
with tolerance s close enough to guarantee holding a vacuum.
This meant no lost working time on the production line
because of improperly fitting pieces.
An interesting side light on vacuum holding is that

the Air Force Quality Control Inspector at the plant has sent
in a suggestion to the Air Force that the part of the specifi-
cation which allows a leak rate of not more than 1 micron per
hour be eliminated and that the specification be re-worded to
allow for no increases in pressure in the vacuum space during
the 48 hour vacuum and pre ssure test.
-239-
The use of aluminmn made possible a substantial
reduction in the weight of the container. The low tempera-
ture properties of aluminum and the low working pressure
requirement (50 psig) of the Air Force admirably suited
aluminum for this weight requirement.
The porosity of metal, interpreted here as out-

gassing of atmospheric gases on the surface of the metal,
appears to be only a negligible factor in holding a vacuum
on the size of dewars fabricated for the Air Force. Many
units which have been in the field and in daily use since
January, 1954, and have not been re-evacuated although there
has been a slight increase in the absolute pressure in the
annular space which could be attributed to an out-gassing of
the impurities in the Santocel powder type insulation used
as the radiation shield.
The requirement for an Air Force container capable

of standing the abuse encountered during transport and shift-
ing from place to place was considered by using stainless
steel suspension tubes to support the inner container within
the vacuum jacket. Stainless steel tubes maintain their
low heat leak characteristics and maintain their strength at
the low temperatures. A skid and hoisting frame built
around the 500 and 150 gallon sizes protects the vacuum
jacket from possible damage. The trailer on which the 50
gallon unit is mounted protrudes far enough on all sides to
protect the vacuum jacket.
The requirement for greatest possible efficiency while

in service was fulfilled by designing the valves and gages so
that they could be mounted into a control panel, or as we call
it, control hood, which would be detachable from the vacuum
jacket without loss of vacuum. Figure 1 is a view of the
valves and gages on a Ronan and Kunzl, Inc. container. Two
couplings, lower right, connect the controls to the container.
An end view showing support tubes and vent (left) and fill-
drain line (right) of a container is given in Figure 2.
The design of this control hood has allowed personnel

to change the controls from one unit to another with a minimurr
of time lost and has allowed them to remove the control hood
from the units for easy interchanging of valves or gages, whic
may have been leaking or damaged in service.
-240-
Figure 1. Valve and gage unit Figure 2. End view of
viewed from the container unfinished container
Figure 4. High vacuum tight

foil type rupture disc
Figure 3. Vacuum sealed container

and detachable control hood
-241-
Figure 3 shows how the valves, gages and piping
can be removed from the container without disturbing the
vacuum and guaranteeing interchangeability from vacuum
jacketed container to vacuum jacketed container.
Trouble shooting is kept to a minimum through this

method. Nearly all the repairs can be made on the line by
using the removable control hood.
The components in the control hood are divided into

two major assemblies: one, the vent valve assembly, and
the other the fill-drain and pressurizing valve assembly. The
vent assembly has the pop-type pressure relief valve which is
a manually operated pres sure relief valve incorporated into
one unit, which can be removed or repaired and pressure
tested prior to assembly of the unit. The same holds true
for the fill-drain and pressurizing valve. The pressurizing
coil is attached to these two asseD1blies through the control
hood and is readily removed from the container.
The reD10vable hood pictured in Figure 3 is suited

for production lines and has enabled the company to reach a
production level of 135 units per month without any difficulty.
Valve assemblies are prefabricated, cleaned and pressure
tested prior to as sembly to the unit for the functional tests.
The use of screwed fittings facilitates the fabrication of the
valve asseD1blies on an assembly line basis.
Incorporating D1axiD1uD1 safety into each unit required

that each pressure relief valve be backed up by a frangible
disc. The transfer ho se between the fill-drain valve and the
aircraft, or receiving container, also had to be equipped with
pressure relieving safety devices because in a time of
eD1ergency, the operator could have his attention distracted
momentarily and trap liquid in the line causing possible line
rupture.
Aluminum foil discs were developed to meet safety

requireD1ents. They are used both for vacuum and pressure
and are inexpensive and easily replaced. A foil type rupture
disc on the outer shell of the vacuum space is shown in Figure
4. The disc holds vacuum to zero microns indefinitely.
The pressure relief valves are rebuilt for manual

operation and the operator is requested, through the Operating
-242-
Manual, to pull the handle to make sure the pressure relief
valve is free prior to pressurizing the unit.
This sums up briefly the problems for mass

produced vacuum insulated liquid oxygen containers for
the armed services, and how they were surmounted by
the aluminum vacuum and Santocel insulated liquid
oxygen container currently manufactured.
Recently the question has arisen as to why we do not

use automatic welding in the fabrication of these containers,
Even though automatic welding has its place and could possibly
be adapted to our production facilities, at the time we were
getting into production we had many other problems which had
precedence and had to be solved. Thus the use of automatic
welding was dropped for the time being and all seams are
welded by hand where the skill of the welder is readily
evident.
Some facts on the performance of the units fabricated

on a production line basis may be of interest here. Fifty,
150, 500, and 1000 gallon containers have held vacuums as
low as 0 microns when measured using a Consolidated Vacuum
Corporation thermocouple gage, TG-025, Ten microns and
lower is the range of pressure in the annular space during
functional test, All units are tested. Since the accuracy of
the gage is plus or minus a few microns, we feel we are
achieving maximum efficiency and lowest absolute pressure
in the annular space.
Evaporation rates for the units at ten microns or less

pressure in the vacuum space, using Santocel II A" as a
radiation shield, have been as low as three and one-half per
cent of volume per day in the 50 gallon size; one and one-
quarter per cent of volume per day in the 150 gallon size;
seven-tenths of one per cent of volume in the 500 gallon size
and six-tenths of one per cent on the 1000 gallon size, These
percentages are based on test results using liquid oxygen at
standard temperature and pressure, Measurements were made
by weight, as well as by vented gas volume through a gas
meter,
Figure 5 shows a completed 50 gallon USAF type

liquid oxygen container mounted on a trailer,
-243-
Figure 5. Completed 50 gallon
USAF type liquid oxygen
container, trailer mounted
-244-
Discussion

Standards: Do you buy extended stem valves or manufacture
them:
Answer by author: When Ronan and Kunzl, Inc. first pro-

duced liquid oxygen containers, extended stem valves for use
with low temperature liquefied atmospheric gases were not
available on the market and the company proceeded to design
and manufacture their own valve s. This method has been
followed to this date and we now have valves in 3/4, 1, and
1-1/2 inch sizes available. A standard valve is purchased
and the extended stem is inserted. A Teflon seal is used as
packing.
Question by D. R. Cheyney, Kaiser Aluminum: What is the

weight of the 1000 gallon dewar?
Answer by author: The 1000 gallon container weighs 5.050

pounds when empty. The container and its vacuum jacket
are made from aluminum and the skid is steel.
Question by B. W. Birmingham. CEL National Bureau of

Standards: What insulating vacuum do you achieve warm 7
Answer by author: The insulating vacuum achieved when the

containers are warm ranges from 1000 to 750 microns. This
has proven to be a satisfactory level of vacuum during manu-
facture. This vacuum will go as low as 0 microns, using a
Consolidated Vacuum thermocouple type vacuum gage when the
containers are filled with liquid oxygen or liquid nitrogen.
Question by G. E. McIntosh, Beech Aircraft Corporation: Why

do you use Santoce1 instead of perlite?
Answer by author: Santoce1 is used in the fabrication of the

containers to conform with the requirement of the Military
Specification and to insure the minimum heat loss through
the insulation when the container no longer holds a vacuum.
-245-
Comment by H. L. Johnston, H. L. Johnston, Inc.: Our
observations at the Ohio State University some years ago
established the superiority of Santocel over other insulations
then tested. However, since that time we have found that
some perlites may be obtained that considerably exceed the
performance while other samples of perlite are inferior.
This is understandable since both perlite and Santocel are
chemically and physically almost alike. Both are exploded
silica gels. Santocel is produced by release of pressure from
silica gel forms with high boiling alcohols in an autoclave.
Perlite is formed in nature by volcanic action. As may be
expected the natural product may vary widely in its
properties.

Standards: Are any special cleaning methods used during
assembly of dewars?
Answer by author: The cleaning methods used d~ring the

manufacture of the liquid oxygen containers for the Air Force
consist of a thorough de-greasing of every piece during
fabrication and a purging with oxygen gases prior to the
testing with liquid oxygen.
Question by R. B. Scott, CEL National Bureau of Standards:

What fraction of the total heat leak is through the insulating
powder?
Answer by author: To answer this question, I will turn to

R. Rabbitt, Project Engineer for Ronan and Kunzl, Inc.
Answer by R. Rabbitt, Ronan and Kunzl, Inc.: We estimate

the heat leak through the insulation to be approximately one-
half of the total heat leak. The remaining heat leak would be
through the fill-drain and vent line and the suspension system.
Question by R. S. Hampton, Lox Equipment Company: Under

what conditions was the 48 hour vacuum and pressure test
made?
-246-
Answer by author: The 48 hour vacuum pres sure test is
conducted conforming to the requirements of Air Force
Specification MIL-T-4829, which states that each tank will
have a partial filling of liquid oxygen with an absolute pres-
sure in the inner tank of 65 psi, and the annular space shall
show a vacuum of 100 microns or less. The container under
pressure must then stand for 48 hours without showing an
increase of pressure in the vacuum space to exceed 1
micron an hour. As an aside here, may I state that our
Air Force Quality Control Inspector has suggested to the Air
Force that this be changed to a micron increase per hour
during the vacuum test, as he bases his observation on more
than 100 containe rs which showed no increase of pres sure in
the vacuum space while the containers were undergoing the
48 hour vacuum and pressure test.
Question by R. S. Hampton, Lox Equipment Company: What

density of perlite has lower thermal conductivity values than
Santocel?
Answer by author: In order to answer this question, I will

have to turn to M. M. Fulk from the dEL National Bureau of
Standards, who may be able to supply the answer.
Answer by M. M. Fulk, CEL National Bureau of Standards:

Any denSity of perlite • 06 gram per cc or greater will
have a lower heat transport than Santocel (both under
evacuated conditions).
-247-
E-4
A PRECISION LOW TEMPERATURE X-RAY CAMERA
E. G. Eeles*
National Re search Council of Canada
Ottawa, Canada
Introduction
In many fields of cryogenic work the design of the

equipment depends solely on the low temperature require-
ments and in many cases where actual physical measurements
are required, the method of measurement can readily be
adapted to the low temperature design. However, this is
not the case in measuring lattice parameters by the Debye-
Scherrer powder method as the X-ray optical requirements
of the instrument must remain, in essence, the same as
.
those of conventional room temperature cameras. In general
it is best to consider the problem as one of adapting a
conventional room temperature design to take in low tempera-
tures as well. The basic problem is therefore one of locating
and rotating a specimen, of accurately controlled and measured
temperature, in a collimated X-ray beam with provision for
recording the diffracted X-ray beams over nearly the whole
angular field of diffraction.
Because of these X-ray considerations some of the

cryogenic aspects of the design are different from those
normally encountered in low temperature physics. In the first
place it is not possible for opacity reasons, to surround the
specimen, itself less than 0.020 inch in diameter, by solid
metallic radiation shields, and secondly it is not possible to
attach the temperature measuring device directly to the area
of the specimen irradiated. It is therefore neces sary to
locate the specimen inside a zone of controlled temperature and
to prevent all external thermal radiation from reaching this
zone.
*Present address -Physical Metallurgy Division, Mines

Branch, Department of Mines and Technical Surveys, Ottawa,
Ontario, Canada.
-248-
The pattern of diffracted X-rays can be recorded
either by a counter spectrometer or by photographic film.
While for speed of operation the former is to be favored,
a film technique offers simplicity and at least an equal
accuracy. As an ability to detect weak diffraction line s,
a characteristic of the spectrometer, is not normally
required for parameter measurements, a film technique is
used for the instrument described in this article. It should be
noted, however, that no modifications to the design are
necessary to use a counter as the recording device.
X-Ray Camera
The camera lis built within the dimensions of a 19

cm diameter Unicam film cassette. For reasons of record-
ing speed this diameter cannot be any greater and, accordingly,
the use of film places some dimensional restrictions on the
design, particularly on the capacity of the primary refrigerant.
The camera (Figure 1) consists of three sections.

The base plate, which carries the film cassette, provides the
main window housing and X-ray collimator. The vacuum
pumping line for the camera is located in this section which is
mounted on three adjustable legs. The window is covered
with 0.003 inch thick Mylar sheet, sealed to the housing with
black wax. This window housing carries the center portion
of the camera, sealing being accomplished by an 0 -ring.
The concentric refrigerant containers and radiation shields
form the upper section and are all carried by the outermost
casing which is positioned by adjusting screws in the center
section. The center, prime refrigerant, container is held
only at its upper end. being located by three pins at its lower
end. The heat leak into the refrigerant thus becomes less
as the level drops and the overall consumption is less than for
a rigid construction. Any movement of the specimen which may
occur during cooling is corrected by the external adjusting
screws and this type of construction is therefore considered
superior to one in which the rigidity of the instrument is relied
upon to maintain a fixed specimen position.
lUnicam Instruments. Cambridge, England.
-249-
HEATER +THERMOCOUPLE WIRES
~~ AND GAS THERMOMETER CAPILLIARY
SEALED AT EXIT.
~ LIQUID N2 DEWAR
LIQUID H. DEWAR
NEEDLE VALVE ALLOWING

BY PASS OF ADJUSTABLE
THERMAL CONTACT
WORM AND WHEEL
FOR ROTATING DEWAR SPECIMEN CENTERING
AND SPECIMEN
O-RING SEAL IN WHICH DEWAR
AND SPECIMEN ROTATES
THERMOCOUPLE
GAS THERMOMETER
VOLUME
THRUST
BEARING
RADIATION
SHIELDS
HEATER
Figure 1. Precision low

temperature X-Ray camera
-250-
The specimen holder is connected to the base of the
refrigerant container through the thermally adjustable coupling
shown. This coupling is adjustable through a wide range of
conductivities by means of the brass and copper sleeves and
by the pressure of helium gas in the annular volume. By
control of the voltage applied to a heater which is wound on
the specimen holder any specimen temperature above that of
the primary refrigerant can be maintained with the minimum
consumption of the refrigerant.
If possible solid specimens are mounted in the holder

by means of indium or tin; if these metals are inapplicable,
then a thin film of silicone grease provides adequate thermal
contact. For measurements on materials which are liquid or
gaseous at room temperature a closed end thin wall silica
capillary is sealed into the specimen holder by cold setting
Araldyte 2 • For thermal conductivity reasons a copper wire,
indium soldered to the holder is contained in the capillary.
This capillary and the thin wall tube soldered into the back of
the holder then form a separately evacuable system and by this
means most liquid or gaseous materials can be frozen into a
suitable specimen. In this respect mention should be made
that the quality (that is the grain size and the degree of
preferred orientation) of such specimens is critically dependent
on the rate of solidification and here the heater and the
adjustable coupling provide adequate means of control.
The specimen holder carries its own radiation shield

to provide, as far as possible, the constant temperature
zone already mentioned. Of the other two radiation shields,
the outer one is connected to the secondary refrigerant while
the other is connected directly to the center container. The
specimen radiation shield is thus surrounded by a shield which
is at a temperature the same or les s than that of the speci-
men. If the thermal radiation through the windows in these
shields can be eliminated or lowered sufficiently, then the
external heat leak should not be sufficient to cause anyappre-
ciable error in the indicated temperature of the specimen.
The circumferential windows cut in the radiation shields

are covered with 0.001 inch thick cellophane thermally
anchored by grease to the copper shield. Absorption
2 Araldyte, Ciba Company Limited.
-251-
measurements show that, at room temperature, this thick-
ness of cellophane transmits about 10 per cent of the thermal
radiation falling on it. Experiment has shown that in service
the cellophane attains a temperature not more than 100C
above that of the shield. The use of cellophane as the window
material would appear to lower the heat leak to satisfactory
limits, the as sumption being that the absorption coefficient
of the cellophane has a value at lower temperatures similar
to that at room temperature. Experiment suggests that this
is so since removal of the cellophane from the outer windows
results in a measurable increase in the temperature of the
center radiation shield.
Three equiangular pillars in the windows are used to

support the base part of the .shield. The camera rotation is
cam controlled such that the incident X-ray beam is shut off
and rotation speeded whenever a pillar approaches the inter-
section of the beam.
The specimen temperature is measured as that of the

specimen holder by a copper-constantan thermocouple, using
a liquid helium cold junction calibration, and by a gas
thermometer the bulb of which is an annular volume in the
specimen holder. The various electrical leads together with
the gas thermometer capillary are brought out through the
central tube and the camera rotation is automatically reversed
on each revolution to prevent snarling of these leads.
Conclusions
By using a calibrated film cassette which is at room

temperature together with a camera geometry of proven
design there is no reason to believe that the accuracy of any
results obtained should be inferior to that of measurements at
room temperature. The error of measurement has in fact
been similar to that encountered at room temperature, dependent
to varying degree on the quality of the specimen and the
highest Bragg angle of diffraction obtained. With an accuracy
of ± 10C temperature control, which is easily obtainable and
more than sufficient for parameter measurement, the overall
reproducibility under adverse conditions can be as low as
1:2000 but with suitable high angle reflections relative errors
of 1 :50,000 are encountered.
-252-
The work described here was performed during the
tenure of a postdoctorate fellowship of the National Research
Council of Canada.
Discussion
Question by J. F. Eichelberger, Monsanto Chemical Company:

Can grain sizes be determined?
Answer by author: No absolute measurements of grain size

are possible due to the several thicknesses of cellophane
between the specimen and the film. However, reasonably
accurate relative measurements should be possible.
Question by L. E. Cox, Convair: Is the cassette in or out-

side and how do you allow for film shrinkage?
Answer by author: The cassette is completely separate from

the main body of the camera. Film shrinkage is corrected by
the calibrated knife edges in the cas sette as in normal room
temperature cameras.
-253-
E-5
DESIGN OF SIMPLE DC RESISTANCE THERMOMETER

BRIDGES FOR WIDE RANGE TEMPERATURE CONTROL
R. D. Goodwin
Boulder, Colorado
Introduction
This paper is concerned with the design of simple

resistance thermometer bridges for the control of temperature
over very wide ranges.
One objective is to obtain maximum practical sensitiv-

ity by selection of the best practical relationships between
thermometer resistance, ratio resistance and the input
resistance of a detector or amplifier.
Another objective is to determine the actual value of

the temperature sensitivity under any particular conditions.
Most of us recall that maximum sensitivity may be

obtained from a Wheatstone bridge under some conditions when
all four arms are equal. This simple rule ignores the
electric power developed in the thermometer. But low
temperature resistance thermometry usually is limited by this
power. In the following, therefore, the criterion for designing
a simple bridge will be taken as the ratio of overall sensitivity
to the required thermometer power.
Two Bridges and Two Thermometers
Figure 1 illustrates the simple resistance thermometer

bridges. Thermometer resistance is R, and detector resistance
is g. Note from this figure that two, different, simple bridges,
A and l3 exist. The two circuits labeled B are identical. The
different bridges may be interchanged either by suitable
selection of the adjustable arm, as in passing from A to the
lower, left figure B, or merely by interchanging battery and
detector connections, as in passing from A to the lower, right
-254-
figure. 1 The two bridges, of course, become identical when
all arms are equal. In the following discussions, the ratio
arms always will be taken of equal resistance. This condi-
tion simplifies the problem of compensation of resistances in
the connecting wires between bridge and thermometer.
Now there exist, also, two basically different types

of resistance thermometer: first, the pure metals, such as
copper2 and platinum; and second, the semiconductors or
thermistors, including the carbon composition type useful at
the lower temperatures. 3,4
A pertinent question is whether a particular type of

bridge is preferable for a particular type of thermometer and
if so, what are the be st relations among the several resis-
tance s.
These questions may be attacked by solving the six

simultaneous, linear equations for the brid~e, which are
provided by application of Kirchoff's laws.
Qualitative Properties
Let us first review the situation in a qualitative way.

Figure 2 includes a representation of the behavior of thermom-
eters versus temperature. The resistance R of a metal
increases, whereas that of a semiconductor decreases with
increase of temperature. The specific sensitivity, (d ln R/dT),
of each type of thermometer, however, decreases with
increasing temperature.
1 F. R. Kotter, R. P. 1884, NBS J.R. 40, 401 (1948).
2Dauphinee and Preston-Thomas, Rev. Sci. Instr. 25,

884 (1954).
3S. A. Friedberg, Temperature, Vol. 2, Reinhold
Publishing Corporation, New York, New York (1955).
4H. B. Sachse, Electronic Industries 16, (I), 55 (1957);
16, (9). 58 (1957).
5
L. Page and N. 1. Adams, Jr., Electricity, D.
Van Nostrand, New York, New York (1949).
-255-
f }® I·
A IJ
/ ~ ~
~
~ .e!
t ~®r.
~
~~.
~
~ !! m
G:f
tr
4
10
Figure 1. Two different bridges

10 100 1000
TEMPERATURE, -K
Figure 2. General characteristics
I02r----.-----,----~
I~'YII
lo'l-----t--- +d
~ A a
rEl
a
g
a> R
?
IWR(2. (I +.8. +~) 8E I3T
.!! " B- o a g 81n R / aT
a b
Figure 3. Preferred Figure 4. Bridge for

bridge connection metallic thermometer s
-256-
The line, W, repre sents the powe r dis s ipa ting
capacity of any thermometer, which results from the thermal
conduction behavior of insulating materials. For very wide
ranges of temperature it would be both convenient and prac-
tical to employ a bridge connection which would produce this
qualitative power behavior automatically with temperature.
The dashed line, S, represents the effective sensitiv-

ity contribution of the bridge circuit alone. This bridge
sensitivity depends upon the circuit employed, upon the
symmetry of the bridge and upon the relative resistances of
the battery and detector. The useful, overall sensitivity will
be a function of this bridge sensitivity, of the thermometer
sensitivity and of the detector or amplifier sensitivity.
The slope of S shown on this figure represents

merely our guess that S should be greatest when the thermom-
eter sensitivity, (d In R/dT), is least, that is at the highest
temperature to be controlled. It remains to be seen whether
this will be consistent, even qualitatively, with the more
important criterion of thermometer power.
Preferred Thermometer Bridge Circuit
Not shown on this figure is the method for selecting

one of the two possible bridges which will require the smaller
thermometer power for a given sensitivity at the detector
input. Selection is based upon results presented by Figure 3.
The ordinate is the ratio of thermometer power requirements
of the two bridge circuits, A and B, producing a given signal
sensitivity from the same thermometer. The abcissa is
thermometer resistance relative to ratio arm resistance.
Individual curves correspond to relative detector resistance, g.
The conclusion from this figure is summarized as follows:
Bridge circuit A (Figure 1) is preferable when thermometer
resistance is less than ratio arm resistance; circuit B is
preferable when thermometer resistance exceeds ratio arm
resistance.
Bridge for Pure Metal Thermometers
Let us now examine a particular bridge for metallic

thermometers, (Figure 4). The resistance of a pure metal may
become exceedingly small at low temperature. We therefore
-257-
select bridge circuit A, with ratio arm resistance greater
than the maximum thermometer resistance. The properties
of this circuit are as follows:
l) The thermometer power required for a given

signal strength always will be the smaller of the
two possible circuits.
2) As the temperature and thermometer resistance

decrease, the thermometer power will decrease.
3) The symmetry of the bridge decreases with

decreasing temperature, and therefore does vary
in a sense opposite to that of the specific
thermometer sensitivity.
4) The required thermometer power in this circuit is

decreased by an increase of the ratio arm
resistance. The equation shown on this figure is
the derived relationship between thermometer
power, W, thermometer characteristic, (d InR/dT).
desired signal strength. dE/dT. and the resistance
parameters. R. a. and g. To achieve a low
thermometer power. it is clear that the ratio arm
and detector resistances. a and g. should be
greater than maximum thermometer resistance. R.
The algebraic conclusion that ratio arm resistance

should be infinite arises merely because no restriction was
placed on ratio arm power. A rough practical value for ratio
arm resistance could be twice the maximum thermometer
resistance.
Application of this circuit with a nominal 100 ohm

platinum thermometer affords the power requirements given by
Figure 5. The abcissa is temperature. The ordinate is
power in milliwatts required in the thermometer to produce a
sensitivity at the detector input of one millivolt per degree K.
Note that the power requirement decreases rapidly

with temperature despite an accompanying decrease in
symmetry of the bridge and also the large effect of detector
input resistance upon the power requirement.
-258-
100,----.----,----,----~--~
10~---+----4----+----4----J
THERMOMETER ./
POWER, m w B b
rE1
9
1.0 t-++---?L-f-----,~~---t--~ b < R b RsEMI
R273 < loon PT.

a <200n
8E /8T<0.001 V'OK
(WRl 12
b
= 2(1+~)
2g
~
81n R 1ST
Figure 6. B~idge
50 100 200 300
for thermistors
TEMPERATURE. oK
Figure 5. Metallic
thermometer power
a:
~102~______' -______- r______- '
R=104n BRIDGE B
b=loon
THERMOMETER
POWER, m W
I L -__~~~~__~~~__~
BE/BT =0.001 V;oK d ~4
I DETECTOR I NPUT IMPEDANCE, n
4 10 20 40
TEMPERATURE, OK
Figure 8. Relative
Figure 7. Carbon thermistor power
thermometer power
-259-
Bridge for Thermistors
Let us now consider a thermistor bridge, Figure 6.

As the resistance of the thermistor may become exceedingly
large at low temperature, we select circuit B with ratio arm
resistance less than the minimum thermometer resistance.
This bridge then has the following properties:
1) The thermistor power required for a given

sensitivity always will be the smaller of the two
possible bridge circuits.
2) As the temperature decreases and thermistor

resistance increases, the thermistor power will
decrease.
3) The syrrmetry of the bridge decreases with

decreasing temperature and therefore does vary
in a sense opposite to that of the specific
thermometer sensitivity. The change of symmetry
would have no effect upon the sensitivity of this
bridge circuit if a detector of infinite resistance
were employed.
4) The required thermometer power in this circuit is

decreased by a decrease of the ratio arm
resistance. The equation for this circuit shows
that ratio arm resistance, b, profitably may be
taken less than minimum thermometer resistance.
Detector resistance would be taken greater than
maximum thermometer resistance.
The algebraic conclusion that ratio arm resistance

should become zero arises, again, because no power restriction
was imposed. A rough practical value for ratio arm resistance
could be one- half the minimum thermometer resistance.
Application of this circuit with a nominal 56 ohIn

carbon composition thermometer affords the approximate
power requirements given by Figure 7. Again the abcissa is
temperature and the ordinate is power in milliwatts required
to produce a sensitivity of one millivolt per degree K at the
detector input. The qualitative behavior of this circuit is
roughly similar to that of the platinum thermometer with the
other bridge connection.
-260-
Optimum Detector-Amplifier Input Impedance
Consider, finally, the problem of the detector or

amplifier input impedance. We have seen that the highest
possible input resistance would be best from the point of view
of loading the bridge circuit. The criterion we used, how-
ever, was the signal from the bridge, not the observable,
amplified signal. A limitation of some detectors and ampli-
fiers is that their gain approaches zero as their input
impedance is increased without limit. Under such conditions
the useful signal from the system would be nil.
This behavior describes the voltage sensitivity of the

d'Arsonval galvanometer and the emf gain of various electronic
amplifiers which employ an input transformer. The gain of
each of these instruments is inversely proportional to the
square root of the input resistance. The response of each is
proportional to the square root of the input power. But
maximum transfer of power between source and load occurs.
when the respective impedances are equal. For a resistance
thermometer bridge, therefore, the input impedance of these
power sensitive detectors should be matched to the output
impedance of the bridge. Unfortunately, the bridge output
impedance is not a constant. A qualitative rule, therefore,
would be to match impedances at the temperature of lowest
sensitivity per unit thermometer power. This usually will
occur at the highest temperature to be measured.
Figure 8 illustrates the effect of impedance matching

upon the thermometer power requirements of a thermistor
bridge, for a fixed overall sensitivity. The abcissa is the
detector input resistance. The ordinate is thermometer power
relative to the minimum power required when the impedances
are matched. Each curve is for a particular thermometer
resistance. You will see that in this case a detector impedance
mismatch by a factor of ten requires a thermometer power
increase roughly by a factor of three to preserve a given
overall sensitivity.
-261-
Summary of Re sults
When the design criterion for a simple resistance

thermometer bridge, for control over wide ranges at low
temperature, is an optimum practical ratio of sensitivity to
thermometer power, the following conditions are found:
I) For pure metal thermometers, use the connection

which places the thermometer and one of the fixed ratio arms
in series with the battery. Select the resistance of each of
the equal ratio arms to be greater than the greatest thermom-
eter resistance.
2) For thermistors, use the connection which places

the thermometer and the adjustable arm of the bridge in
series with the battery. Select the resistance of each of the
equal ratio arms to be less than the smallest thermometer
resistance.
3) In each case, match the input resistance of power

sensitive detectors to that bridge output impedance which
corresponds to the most difficult conditions of measurement or
control.
Mathematical Appendix
Circuit and Solutions. In Figure 9 let resistances in

ohms be represented by lower case letters, currents in
amperes by corresponding capital letters. Exceptions will be
battery potential, v, and equivalent source emf, e, in volts.
By Kirchoff l slaws
R - C - A =A - G - X = C +G - B = 0, (la, lb, lc)
rR + cC + bB = rR + aA + gG + bB =
rR + aA + xX = v, (ld, Ie, 1£)
which readily are combined to the following set of coefficients,

chosen to include the currents G and X,
-262-
Figure 9. General bridge circuit
G-
Figure 10. Equivalent source circuit
-263-
B + G + X = const.
r a (a+ x+ r) v (2a)
(b+ c+ r) -c r v (2b)
(b+ r) (a+ g) (a+ r) v. ( 2c)
The determinant, D, of equations (2) is useful in the following

forms,
D = M+ Ng =P + Qx ( 3)
where the coefficients are defined as
M = abc + (a+ c) (b+ x) r + (ab+ bc+ calx (3a)

N = (a+ x) (b+ c) + (a+ b+ c+x)r (3b)
p = abc + (a+c)br + [a(b+c) + (a+b+c)r]g (3c)
Q = (ab+ bc+ cal + (a+ c)r + (b+ ctr)g (3d)
The thermometer and detector currents, X and G, are given

by the solutions
XD/v = (a+c)b + (b+c)g ( 4)
GD/v = cx - ab ( 5)
Source Impedance. In Figure 10 the emf arising within

the bridge network is represented by e and the associated
impedance m. From this figure alone.
-ml g = I + d In gl d In G ( 6)
From equations (5) and (6) the bridge source impedance in

general is
m = MIN ( 7)
but at balance, when x = abl c, this becomes independent of

battery resistance
mo = x(l + c/a)/(1 + b/c) (7a)
-264-
Bridge Sensitivity. A specific bridge sensitivity may
be defined as
S = (l/v)dE/d In x
where E = gG ( 8)
From equation (5) in general

2
S = (gcx) (P + Qab/ c) /D ( 9)
or for a balanced bridge,
S = gcx/D (9a)
o
Collection of terms in equation (9a) shows that So depends

upon a symmetry factor, U, a battery resistance factor, 1/ Z,
and a detector loading factor, l/L,
So = U/ZL (10)
where U, Z, and L are defined mathematically by
U = (b/c)/(I+b/c)2 (lOa)
(lOb)
L =I + m o /g (IOc)
Particular Bridges at Balance. Bridge A is defined by

fixed arms a and c of Figure 9. With c = na, arm b is
adjusted to b = nx. Hence,
Ua = (x/a)/(l + x/a)2 (Ua)
Za = I + i(2r/a)/(1 + x/a) (lIb)
La =1 + j(2x/g)/(1 + x/a) ( lle)
where
i = (1 + n)/2n; and j =(1 + n)/2 (lId)
-265-
Bridge B is defined by fixed arms band c of Figure
9. With c= nb, arm a is adjusted to a nx. Hence, =
Ub = n/(l + n)2 = 1/4ij (1Za)
Zb =1 + (r /2 b)( 1 + b / x) /j ( 12b)
Lb =1 + (b/2g) (1 + x/b) /i (12c)
Thermometer Power. For balanced bridges,

thermometer power, W 0' from equation (4) is
(13)
becoming in particular,
(Wo)a = (v 2 /a)(x/a)/ (1 + x/a + 2ir/a)2 ( 13a)
(W o)b = (v 2 /b)(x/b)/ [2jx/b + (1 + x/b)r/b] 2 (13b)
Net Sensitivity and Power. From equations (8), (10)

and (13) for balanced bridges with equal ratio arms
W
o
= H(dE/dT)2/ (d In x/dT)2 (14)
where the resistance parameters respectively are
= (l/x) (1 + x/a + 2x/g)2 ( 14a)

2
= (4/x) (1 + b/2g + x/2g) ( 14b)
Preferred Bridge Connection. The ratio of power

requirements of the two circuits is Ha/Hb' Interchanging
connections cannot alter ratio arm resistance. Placing b a =
in this ratio produces
(15)
where o <k = (1 + g/a)/(l + 2g/a + x/a) < 1 (l5a)
-266-
It follows that
1 when (x / a) ~ 1. (ISb)
>
Optimum Detector Input Resistance. Let the emf

gain, J, of a detector-amplifier be inversely proportional to
the square root of the input impedance, g,
1/2
J = Jo/ g ( 16)
then the amplified output signal will be
dEI/dT = (J)dE/dT (17)
From equations (14), (16) and (17),
1/2 1/2 -1
dEI/dT = J o W 0 K[ g (1 + l/q)] d In x/dT (IS)
where q = g/F and for the respective circuits,
1/2
K
a
=x /(1 + x/a) Fa = 2x/(1 + x/a) (1Sa)
1/2
Kb =x /2 Fb = (b + x)/2 (lSb)
For each circuit, maximum amplified output as a function of g

occurs for a given thermometer power when
q = 1. (19)
Reference to equation (7A) shows that F is identical with the

bridge source impedance, m •
o
For estimating effects of impedance mismatch, a

relative amplified signal, s, is defined and given by
s=(dE'/dT)/(dEI/dT) = 2ql/2/(1+q) (20)

max
-267-
when the relative power requirement therefore becomes
(21 )
Discussion
Question by E. T. Benedikt, Douglas Aircraft: The expres sion

(WR)1/2 was used in this paper. Since W = E2/R where W is
the power, R the resistance, and E the potential difference or
voltage, then (WR) 1 /2 = E. Does this voltage represent the
amplitude of the noise in the signal?
Answer by author: The left sides of the equations given on

Figures 4 and 6 contain the thermometer power, Wand the
thermometer resistance, R. The square root of their product
therefore, is the potential across the thermometer.
Question by G. H. Caine, Pesco Products -Division of Borg

Warner: You mentioned an output of 1 mv per oK sensitivity.
Is this a reasonable figure? What is the maximum sensitivity
available practicably at 20 o K?
Answer by author: Figure 5 for a platinum thermometer

shows that an output of 1 mv/oK requires a thermometer
power ranging from roughly 0.1 milliwatt at 70 0 K to ten
milliwatts at room temperature. This power range is
practical for a resistance thermometer wire which is in good
thermal contact with the object to be measured, as achieved
by wrapping and varnishing the wire, as described by
Dauphinee and Preston-Thomas. In my opinion, but not my
experience, this power range is quite high for a glas s
encapsulated thermometer.
Question by J. 1. Nestel, Consolidated Airborne System,

Inc.: Did you find the thermistor stable in low temperature
measurements?
Answer by author: I have no experimental results on low

temperature thermistors. From published reviews, to which
I refer, one gathers that thermistors generally are stable
if maintained at the low temperature.
-268-
E-6
CALIBRATION OF THERMOCOUPLES
AT LOW TEMPERATURES
M. D. Bunch and R. L. Powell

Boulder, Colorado
Introduction
Scientists and engineers have been confronted with the

problem of a low cost, accurate thermometer for low tempera-
tures. The thermocouple is one of the most logical answers.
However, conventional thermocouples have inadequate sensi-
tivities at 20 0 K and below. From previous investigations by
the Kamerlingh Onnes Laboratories in Leiden, it has been
found that dilute alloys of transition metals in Group lB can
have high thermoelectric power near absolute zero and the
alloy of gold-2. 1 atomic per cent cobalt has been suggested as
a practical thermoelement. Consequently, the purpose of this
investigation has been to find the sensitivity at temperatures
from 4 0 to 300 0 K and to study the stability of the thermo-
element. An apparatus is required that is capable of maintain-
ing one junction of the thermocouple at a fixed reference
temperature (4 0 , 20 0 , and 76°K) , while allowing the tempera-
ture of the other junction to be varied.
Apparatus
The cryostat consists of three containers. These are

shown in Figure 1. The outer container is immersed in a
liquid nitrogen bath for thermal shielding and the two small
containers comprise the actual calibration apparatus. The
lower container holds the refrigerant liquid in which is
immersed the reference junctions of all thermocouples to be
measured, and a calibrated platinum resistance thermometer.
The copper lead wires for the thermocouples and resistance
thermometer are wrapped on the lower tempering block.
The upper container incorporates another tempering

block and the measuring block, the temperature of which is
-269-
20000
EXCHANGE
GAS,-+---f.
~rt-+-- UPPER
TEMPERING
10000 I
BLOCK 6000
4000 i7
UPPER VENT
TUBE
2000
i
:/'
, i
GAS
1000
iii
THERMO-
METER +--*-4--1+ 600
400
1-/~I,-H-'lI--- MEASURING
BLOCK l 200
iI
)
~ 100
"- 0
7
~
FILL UJ 40
TU BE -+-----jP"l ~-+-- VACUUM
I)
1
0
)
0
VACUUM THERMOELECTRIC
CONTAINER
6 I, FORCE OF
4 u-2.1 AT". Co VS Cu
2 V 4 6 10 20
T TT
40 60 100 200 300
~-+-+-- LOWER TEMPERATURE. oK
TEMPERING
BLOCK
Figure 2.
Figure 1. Thermometer
calibration apparatus
6= /
o
I 10-
ELECTRICAL l
RESISTIVITY
E
:.: 50 .r: - COPPER WIRE
o
o
.... /
,:
~_ 40 -
a:
UJ
-
-
i--
I--
I--
THERMOELECTRIC
POWER OF
Au - 2.1 AT%.
Co VS Cu
-- - t::
>
;::
'!? 10- 7
il
3: (J)
o0.. - - UJ
30 a:
u
I
a:
~ /
~
.J
UJ
o
20
/ 8 /
:;; 1/
~ 10
./
r
~
_I""
o 9
I 34 6810 203040 6080100 200 300
10 50 100 300
TEMPERATURE. °K TEMPERATURE. OK
Figure 3.
Figure 4.
-270-
adjusted by a heater and to which are attached the measuring
junctions of the thermocouples. The measuring block contains
a gas thermometer bulb and a re-entrant well for a capsule
type platinum resistance thermometer. In addition, certain
temperatures can be established by measuring the vapor
pressure of pure liquid that can be introduced into the upper
chamber. The wall temperature of the upper chamber is
controlled to about the same value as the measuring block
by a differential thermocouple and heaters. This container
is neces sarily filled with exchange gas from the refrigerant
liquid. With exception of the lower tempering block, which is
of brass, the other major components are of copper. However,
the tube connecting the upper and lower containers is of
stainless steel to minimize heat flow to the refrigerant. A
copper heat short is connected from the refrigerant container
to the upper vent tube in order to bypass to the refrigerant
bath any heat leak from the surroundings to the upper
container.
The helium gas thermometer is a part of the measur-

ing block with a small capillary tube leading to the manometer
system. The gas space itself is polished to reduce adsorption
on the walls. The quantity of gas contained in the various
IUlisance volumes is negligible for the temperature range 4 to
20 0 K. Manometric measurements were made with a cathetom-
eter to 0.1 mm.
Experiment and Re suIts
Measurements were made from 4 to 300 0 K. However,

two methods of measurement were neces sary. A calibrated
platinum resistance thermometer was used as a standard for
the range 20 to 300 0 K, and the helium gas thermometer for
the range 4 to 20 0 K. Temperature equilibrium is first
attained by manual control of the heaters and regulation of the
differential thermocouples. The resistances of the platinum
thermometers were measured with a Mueller bridge, and the
emfl s of the thermocouples were measured with a five dial
Wenner potentiometer.
The thermoelectric force of gold-2.l atomic per cent

cobalt versus copper shown in Figure 2 is about 10,000
microvolts at 300 0 K for reference temperature of OOK as
compared to about 6,000 microvolts for constantan versus
-271-
copper. It should be noted that no measurements were made
below 4 o K. Figure 3 showing the thermoelectric power versuS
temperature is more informative. This shows the sensitivity
of the couple at various temperatures. The sensitivity of the
gold-2.l atomic per cent cobalt versus copper couple is 15
microvolts/oK at 20 0 K as compared to 6 microvolts/oK for
constantan versus copper couple at the same temperature.
Table I gives the thermoelectric force and the thermoelectric
power for the gold-2.l atomic per cent cobalt versus copper
couple for the temperature range from 0 to 300 o K.
The apparatus was also used to measure the electrical

resistance of a pure copper wire in the regions indicated by
Figure 4. The sample wire was wrapped on the measuring
block.
Conclusion
The gold-2.l atomic per cent cobalt versus copper

couple is suitable for many engineering uses. It has a high
sensitivity and the reproducibility is satisfactory for most
engineering applications. The thermoelectric force variations
of different melts of the gold-2.l atomic per cent cobalt
element prepared by the same manufacturer have been as high
as seven per cent. It is expected that the product will be more
uniform with additional experience. For wires of the same
melt, there are variations of about one per cent. The wire
is cold drawn to the necessary diameter without any annealing.
Thus, for a wire of 0.005 inch diameter, the drawing reduces
the diameter to approximately 1/400 of the original diameter
(about 2 inches).
Our plans for the future are to continue more research

on the present series. One of the main purposes will be to
make periodic checks of the wire to find if there is a time
variation of the emf caused by room temperature annealing.
An analysis of the thermoelectric properties of the copper-
constantan couple will also be forthcoming.
-272-
TABLE I
Temp. emf dE/dT Temp. emf dE/dT
( OK) (mv) (mv/oK) ( -K) (mv) (mv/oK)
0 0 0 100 2553. 37.0

2 1.9 1.9 110 2927. 37.9
4 7.7 3.8 120 3310. 38.5
6 17. 1 5.6 130 3698. 39. 1
8 29.9 7.3 140 4090. 39.5
10 46.0 8.9 150 4487. 39.8
12 65.3 10.4 160 4886. 40.0
14 87.6 11. 9 170 5287. 40.2
16 113. 13.2 180 5690. 40.4
18 140. 14.5 190 6095. 40.5
20 171. 15.8 200 6501. 40.6
25 257. 18.7 210 6908. 40.7
30 358. 21. 3 220 7315. 40.7
35 470. 23.6 230 7722. 40.7
40 593. 25.6 240 8129. 40.7
45 726. 27.4 250 8536. 40.7
50 867. 28.9 260 8942. 40.6
60 1169. 31. 4 270 9348. 40.6
70 1493. 33.3 280 9753. 40.5
80 1833. 34.7 290 10158. 40.4
90 2187. 36.0 300 10 561. 40.3
Reference temperature OOK.
Discussion
Question by B. Welber, NACA: What is the name of the

manufacturer of the gold-2.l atomic per cent cobalt wire?
Answer by author: Sigmund Cohn Manufacturing Company,

Inc. Mount Vernon, New York.
Questicn by B. Welber, NACA: Is this the thermocouple with

the highest expected sensitivity at 20 o K?
Answer by author: Yes, the sensitivity is about 15 microvolts

per oK.
-273-
Question by E. T. Benedikt, Douglas Aircraft Company: How
long doe s the helium procedure of calibration take?
Answer by author: About one week.
Que stion by E. T. Benedikt, Douglas Aircraft Company: For

how long is the calibration valid?
Answer by author: For as long as the thermocouple is not

heated above ambient temperatures or for as long as the
thermocouple is used in the calibrated range.
Question by D. A. Burgeson, Pesco Products: Are any

unusual mechanical properties of the gold-2.l atomic per cent
cobalt wire experienced?
Answer by author: Yes, one must be careful of soldering as

the cobalt is in solution with the gold and could possibly come
out of the solution upon being heated excessively.
Question by D. A. Burgeson, Pesco Products: What pre-

cautions are required in handling?
Answer by author: One should be careful of cold working the

wire by knotting, bending, etc.
Question by D. A. Burgeson, Pesco Products: How is

junction joining completed?
Answer by author: We have had success with a soft solder

connection, making sure the iron is only hot ~nough to melt
the solder.
Question by B. Welber, NACA: Do you have any data on the

residual resistivity of the pure copper mentioned in the
abstract?
Answer by author: Our copper wire data begins at ISO or

19° K, so we have not done any work on the residual
resistivity of this sample.
-274-
F-1
V APOR-LIQUID EQUILIBRIA IN THE SYSTEM

HYDROGEN -NITROGEN -CARBON MONOXIDE
W. W. Akers and L. S. Eubanks

The Rice Institute
Houston, Texas
Introduction
The purpose of this investigation was to determine the

equilibrium compositions of mixtures of hydrogen, nitrogen
and carbon monoxide in a portion of the region where the
vapor and liquid coexist, particular emphasis to be placed on
the regions of low concentration of carbon monoxide in
hydrogen-nitrogen mixtures.
Equilibrium compositions were determined at three

temperatures, _310 0 , -280 0 and -240 o F, and at pressures
from 300 psig to pressures very near the critical. The upper
limit of the temperature range is very near the critical
temperature s of nitrogen and carbon monoxide which marks the
upper boundary of the existence of the liquid phase. Lower
temperature runs were not feasible because of the refrigeration
problems involved.
Earlier Studies of Hydrogen-Nitrogen-Carbon Monoxide Systems
The hydrogen-nitrogen-carbon monoxide system, or

portions thereof, have been investigated by several authors over
limited ranges of conditions. Yuskevich and Torocheshnikov 1
investigated the binary system, nitrogen and carbon monoxide.
The hydrogen-nitrogen system was investigated by Verschoyle 2 ,
Steckel and Zinn3 and Ruhemann and Zinn':l:. Ruhemann. and
IN. F. Yuskevich and N. S. Torocheshnikov, Zhurn.

Khim. Prom., l3, No. 21, 1273 (1936).
2 T. T. H. Verschoyle, Phil. Trans. Roy. Soc. A,
230,189,(1932).
3Steckel and N. Zinn, Zhurn. Khim. Prom. 16, No.8
(1939) .
4M. Ruhemann and N. Zinn, Phys. Z. d Sowjeturioni,
12, 389 (1937).
-275-
Zinn and Yerschoyle also investigated the hydrogen-carbon
monoxide system and obtained a limited amount of data on
the ternary system. More recently, Dokoupil, et al. 5
investigated the solid-vapor region for the binary systems,
hydrogen-nitrogen and hydrogen-carbon monoxide, with a few
experimental points on the ternary system.
Refinements in the experimental techniques of vapor-

liquid equilibria and improved methods of analyses make it
possible now to determine the equilibrium compositions more
accurately, and the desirability of more complete data is
obvious.
Hydrogen-light hydrocarbon systems have been exten-
sively investigated. The hydrogen-methane-ethane system was
investigated by Levitskaya 6 • Freeth and Yerschoyle 7 and
Fastowsky and Gonikberg 8 also investigated the hydrogen-
methane binary. Katz and Williams 9 investigated the binary
systems of hydrogen with ethane, ethylene, propane and
propylene and included some ternary data. All of these
investigations were at temperatures ranging from room
temperature to -300 0 F. Numerous investigations were
carried out with systems containing hydrogen at room tempera-
ture and above, but will not be mentioned here.
5 Z . Dokoupil, G. Van Soest, and M. D. P. Sevenker,

Appl. Sci. Res. A, 5, 182 (1955).
6 E. P. Levitskaya, J. Tech. Phys. (U.S.S.R.), 11,

197 (1941).
7F . A. Freeth and T. T. H. Yerschoyle, Proc. Roy.

Soc. London, l30A, 453 (1931).
8 y . G. Fastowsky and M. G. Gonikberg, J. Phys.

Chem. (U.S.S.R.) • .!i, 427 (1940).
9 D . L. Katz and R. B. Williams, Ind. Eng. Chem.,

46. 2512 (1954).
-276-
ExperiTI1ental Methods
There are several possible TI1ethods of obtaining

vapor -liquid equilibriuTI1 data. These TI1ay be divided into two
principal approaches, the" static" TI1ethod and the "flow"
TI1ethod.
One forTI1 of the "static" TI1ethod is the dew point-
bubble point TI1ethod. In this TI1ethod a TI1ixture of known
cOTI1position is placed in a cell at a controlled teTI1perature ani
the pressure is slowly raised until the first drop of liquid is
forTI1ed. The pressure at this point, the dew point, is
recorded and the pressure is raised further until the last bit
of vapor is condensed. This pressure is recorded as the
bubble point. This is repeated for several different TI1ixtures
and at several different teTI1peratures. EquilibriUTI1 vaporiza-
tion ratios can be obtained only for binary systeTI1s.
The second forTI1 of the "static" TI1ethod consists in

placing a two phase TI1ixture in a container at a controlled
teTI1perature and agitating the TI1ixture in SOTI1e way, usually
by ro c king the containe r. Aite race rtain pe rio d of tiTI1e the
TI1ixture should reach equilibriuTI1. The pressure is then read
and a portion of the vapor and liquid are withdrawn and
analyzed. Usually, however, in this method no provision is
TI1ade for keeping the pressure constant while the samples are
being withdrawn. A drop in pressure during sampling, of
course, disturbs the equilibrium.
There are also two principal type s in the "flow"

method approach to vapor-liquid equilibriUTI1. The first of
these is to take a homogeneous mixture and by reducing
pressure and/or temperature reach the two-phase region.
The two phases are then separated and samples of each are
analyzed. The intiTI1ate contact between the vapor and liquid
during condensation provides a rapid approach to equilibriUTI1.
The principal difficulties encountered in this method are
pressure and temperature control in the vessel where conden-
sation occurs. Also, in the near critical region, the
separation of the two phases becomes difficult. This method
has also been applied to solid-vapor equilibria by Dakoupil,
et al. although no attempt was TI1ade to withdraw or analyze the
solid phase s.
-277-
The other type of "flow" system is called the vapor-
recycle method. In this approach a two-phase mixture is
placed in a vessel at controlled temperature. Vapor is then
withdrawn from the cell and recirculated through the liquid.
This contact between vapor and liquid provides a fairly rapid
approach to equilibrium. After equilibrium is reached,
samples of the liquid and vapor are withdrawn and analyzed.
In this method also, some provision must be made to keep
the pres sure constant during sampling. The vapor -recycle
method was used in this investigation.
A schematic diagram of the equilibrium apparatus

appears in Figure 1.
Nomenclature for Figure 1
G , 0-2000 psi Heise gage; G 2 , 0-10,000 psi Heise gage;

1
G 3 differential gage; G 4 , 0-200 psi gage for sampling system;
W, windowed cell; P, magnetic pump; V, variable volume

cell; E, equilibrium cell; M, mercury pump; B l , B 2 , blocking
valves; 8 1 , liquid sample valve; 8 2 , vapor sample valve; D,
Drierite container; A, Ascarite container; C, cold trap for
additional drying.
Experimental Apparatus
Equilibrium Cell. The cell used in this investigation

was a piece of 2 inch O.D. type 303 stainless steel bar 10
inches long, drilled along its axis with a 1-1/8 inch drill. The
ends were threaded inside and plugs were made, also from 2
inch O.D. 303 stainless. The seal between the plug and the
cell was made with an annealed copper gasket. This provided
an excellent seal and no leaks ever developed here.
In the bottom plug three holes were drilled to provide

a liquid sample outlet, an inlet for the recycled vapor and
a tap for a thermowell. In the top only two holes were
drilled, one for the vapor outlet and one for a thermowell. A
Barton differential gage with a range of 0 to 20 inches of
water and rated at 5000 psi was used to indicate the level
of liquid in the cell. Taps for the differential gage were
provided by tees in the vapor inlet and outlet lines. The
differential gage was intended only au a qualitative indication of
the amount of liquid present.
-278-
caUl TlIAP
SCHEMATIC DIAGRAM OF EQUIPMENT TEMPERATURE CONT1IOL SYSTEM
Figure 1. Figure 2.
PRESSURJ
DIA~~
NITROG
.':'''! ~
SYSTE
2 I
/ ''\
f: ~ \
REGlLAT~
' ,
IDOD
V '\ "\
l~ )
vr 1--- LJ
SCHEMATIC DIAGRAM aF
EQUIPMENT
CHRCIIIOTOGAAfIHIC 0
MOL FRACTION HVDROGEN
--- 08 1.0
Figure 3. Figure 4.
-279-
It was necessary to dampen the movement of the
differential gage to prevent rapid movement of gases in and
out of the differential gage. This was done by inserting about
two feet of 17 gauge wire into the tubing leading to the differ-
ential gage. This made the gage completely insensitive to the
pressure surges caused by the recirculating pump but still
permitted it to indicate the liquid level.
Type 303 stainless steel tubing rated at 10,000 psi

was used throughout the system. Sample lines and pressure
lines were of liB inch O. D. tubing. In the charge system,
recirculation system and in the coil used to cool the recycled
vapor before reintroducing it to the equilibrium cell, 3/16
inch O. D. tubing was used in order to decrease clogging
possibilities from small amounts of unremoved contaminants,
such as water or carbon dioxide.
Insulation for the cooling system and equilibrium cell

was provided principally by a block of Styrofoam. A hole was
drilled in the Styrofoam and the cooling system and cell placed
inside. The high pressure tap for the differential gage was
led horizontally out of the insulation to avoid a vertical liquid
leg build-up. All other tubing and the thermocouple leads
were led out the top of the insulation. The void space in the
hole was filled with insulation and rodded as it was added for
even distribution. The top of the hole was closed with a two-
piece Styrofoam plug. Attempts to seal the remaining cracks
with putty to prevent water from entering the cold space failed.
The vapor was circulated by a magnetically operated

pump, rated at 7500 psi working pressure and designed to
deliver 250 cc per minute. The magnetically operated pump is
desirable because it eliminates the necessity of a rotating or
sliding seal against high pressure. Vapor recirculation was
controlled by a by-pass valve around the pump.
Theoretically, the pump piston required no lubrication;

however, experience proved that the pump would operate for
only a few hours unlubricated. Therefore, the piston and
cylinder wall were coated with a thin layer of Apiezon high
vacuum stopcock grease. This allowed smooth operation for a
long period. One further difficulty was experienced with the
pump. The check valves had to be reseated periodically in
order for the pump to meet design specifications.
-2BO-
FrolTI the pUlTIP the vapor continues to a stainless
steel cell, terlTIed herein the "variable volulTIe cell," designed
for 10,000 psi. The vapor passes into the cell through a
perforated tube and rises through a series of baffles to exit
at the top. These baffles provide turbulence necessary
for periodic flushing of the vapor in the cell. The volulTIe
of this cell is approxilTIately 350 cc and provisions were lTIade
for adding lTIercury to the cell through the bottolTI, providing
a lTIeans of pressure control.
FrolTI this cell the vapor passes back to the equilib-

riulTI cell through a coil of tubing which is wound about the
constant telTIperature bath. Good therlTIal contact with the
copper walls of the telTIperature bath was lTIade by lead-
soldering the tubing to the wall. The recycle enters the
bottolTI of the cell and then bubbles through the liquid. The
total volUlTIe of the systelTI external to the cell is a lTIaxilTIulTI
of 540 cc. With the pUlTIP supplying a flow rate of 250 cc per
lTIinute the entire alTIount of vapor external to the cell is
replaced approxilTIately every two lTIinutes.
TelTIperature Control. A bath surrounding the walls

of the equilibriulTI cell provided control of the telTIperature.
The bath liquid was boiling nitrogen or oxygen, depending
on the telTIperature desired. Liquid was supplied frolTI the
condenser placed above the cell. The nitrogen or oxygen for
the bath was transferred by air pressure frolTI a cylinder
through a Hoke two-stage regulator and was dried in a dry
ice-alcohol cold trap before adITIission to the condenser.
The telTIperature was controlled by controlling the
pressure on the boiling liquid. By inserting a surge drulTI
with a volulTIe of about one cubic foot in the line connecting
the regulator and the condenser, and by providing a slTIall but
continuous bleed of gas, the pressure could be controlled quite
accurately. It was found, however, that at the higher telTIper-
atures there was a pronounced tendency for the liquid to
subcool in the condenser. This fed an unsaturated liquid to
the bath and lTIade it ilTIpossible to control the telTIperature.
The difficulty was corrected by enclosing the cold side of the
condenser and lTIetering and controlling the rate of evaporation
of liquid nitrogen.
-281-
By controlling the pressure difference between the
storage flask and the surge drum in the evaporated nitrogen
line, the rate of refrigeration could be controlled quite
accurately.
The combination of the two controls gave quite good

control of the temperature for the runs at -280 o F. For the
runs at -3l0 o F the refrigeration control was not used, because
there was no problem of sub-cooling or conduction between
the condenser and the cell and the pressure control on the
cooling system was adequate.
Because of the comparatively small mass of the cell,

control of the temperature was limited to ± 0.1 of. Temper-
atures were measured with copper-constantan thermocouples
inserted in thermowells which extended into the vapor and
liquid space of the equilibrium cells. The thermocouples
were calibrated by use of the normal boiling point of nitrogen
and by comparison with the vapor pressure curve of nitrogen
at higher temperatures. The temperatures are believed to be
as accurate as they are precise. A diagram of the temper-
ature control system appears in Figure 2.
Pressure Control. The main items of the pres sure

control system were the stainless steel cell external to the
equilibrium cell, which was described earlier, and the mercury
pump. The pressure was adjusted by injecting mercury into
the cell, thus changing the overall volume of the system. After
the system was charged, final adjustment of the pressure was
made by adding or removing mercury. Adjustments were also
made in this manner for small changes in pressure due to
gases going into or out of solution while equilibrium was being
approached. This also provided adjustment for drops in
pressure due to the minor leaks in the system, which seemed
to be unavoidable due to the multiplicity of connections in the
system.
Pressures below 2000 psia were read on a 0 - 2000

psi Heise gage (2 psi subdivisions) which was calibrated at the
factory shortly before the equilibrium runs were begun. The
factory calibration indicated that there was no error as great
as 2 psi.
Pressures of 2000 psia and over were read on a

o- 10,000 psi Heise gage (10 psi subdivisions). This gage was
-282-
calibrated up to 2000 psia by comparison with the lower
range gage, and also showed no error as large as one scale
division.
Sampling System. The samples were collected in

evacuated metal container s at room temperature, thus both
the liquid and vapor samples were collected as a gas. A
pressure gage was placed in the sample system to indicate
the amount of sample taken and the rate of sampling.
The liquid sample was withdrawn directly from the

bottom of the equilibrium cell. The tubing from the equilibrium
cell to the sample valve had a .045 inch diameter wire inside
it to minimize the dead volume. The line from the sample
valve to the sample container was evacuated before sampling
and then purged with a small portion of liquid sample.
Mercury was added to the variable volume cell as the liquid
sample was taken in order to hold the pressure constant.
The vapor sample was withdrawn from the system

external to the equilibrium cell after the equilibrium cell was
closed off. Since the equilibrium cell was closed off, it was
unnecessary to keep the pressure constant during sampling of
the vapor phase. The first bit of vapor sample was also
rejected in order to purge the lines. An amount of sample
equivalent to approximately 5 liters at atmospheric conditions
was usually taken.
Materials. The hydrogen and nitrogen used were of

99.7 per cent minimum purity. The carbon monoxide was
approximately of 97 per cent purity, with hydrogen, nitrogen
and carbon dioxide being the principal contaminants. A wave
length scan of the carbon monoxide in the infra-red region
indicated that no hydrocarbons were present. For calibration
of the analytical devices, carbon monoxide was prepared in the
laboratory by the reaction of formic and sulfuric acid.
The hydrogen, nitrogen and carbon monoxide used in

the equilibrium runs were passed over Drierite and Ascarite
and further dried in a dry ice-alcohol cold trap before
charging to the system. This purification process was carried
out at pressures from 500 to 2000 psi, depending on the
pressure of the equilibrium run.
-283-
Operation of Equipment. While still at room temper-
ature, the system .;as evacuated for 1/2 to I hour and then
charged with either nitrogen or carbon monoxide to five or
six hundred pounds. The equilibrium cell was then cooled to
the temperature desired and more nitrogen or carbon monoxide
was added until the differential gage indicated that the equilib-
rium cell was 1/2 to 3/4 full. Hydrogen was then charged
until the desired pressure was approached. All the gases
were charged at a slow rate to insure as complete a removal
of contaminants as possible.
The magnetic pump was then started and circulation

was continued until rapid changes in pressure stopped. Usually
the pressure would drop appreciably as the hydrogen went into
solution. 1£ the pressure had dropped appreciably (30 psi or
more), more hydrogen was charged and recirculation was
restarted. This procedure was repeated until no appreciable
changes in pressure were noted. Final adjustment of pressure
was made by addition or withdrawal of mercury from the
variable volume cell and recirculation was continued.
The equilibrium run was then started with pressure,

temperature at the top and bottom of the equilibrium cell,
differential gage reading, pressure on the cooling system and
rate of evaporation of liquid nitrogen from the condenser being
read at five minute intervals or less, and being recorded at
ten minute intervals.
Circulation was continued for at least one hour under

conditions of constant temperature and pressure. Ten minutes
before the end of the run the sample containers were connected
and then evacuated during the final ten minutes of the run.
Before the magnetic pump was turned off, the valve in

the recycle return line was closed. This was done to prevent
the liquid in the cell from running back into the recycle cooling
coil. With the pump turned off, the refrigeration rate was cut
back to compensate for the heat load of the recycle. This heat
load amounted to approximately 25 per cent of the total heat
transfer to the equilibriu.rn cell.
The liquid sample line was then purged and the liquid
sample was collected. The liquid sampling time was usually
about five minutes. The valve in the recycle outlet line was
-284-
then closed and the sample system from the containers to
the sample valves was re - evacuated. The vapor sample
valve was then opened and the pressure on the external
portion of the system was allowed to drop about ten psi
before closing the valve again. The sample system and
vapor sample container were re-evacuated and then the
sample was collected.
Preparing for the next run could be done in several

ways, and all of these were used at one time or another.
Perhaps the simplest method was to simply add hydrogen
until the next highest pressure at which data were to be taken
was reached. Usually the addition of hydrogen decreased the
liquid level in the cell slightly and to use this method of
changing the conditions, it was only necessary to have s'uffi-
dent liquid in the cell initially. This method was used
almost exclusively in obtaining the binary data.
Another way to change the conditions was to add

either nitrogen or carbon monoxide to the system, thus
changing the liquid composition. The amount added was meas-
ured by the change in pressure while charging to the external
system only. The amount needed to give a desired composi-
tion change was determined by experience.
The third approach was to go to a lower pressure by

bleeding the system. Because of the extremely toxic nature
of carbon monoxide, and the explosion hazard of hydrogen,
arrangements were made to bleed the gases to the outside of
the building.
It should be noted that for runs above 2000 psia the

differential gage was disconnected from the cell in order to
reduce the total volume of the system. Since the maximum
cylinder pressure available was 2000 psi, a method for
compressing hydrogen above this pressure was needed. This
was provided by a stainless steel cylinder designed for a
5000 psi working pressure. This cylinder had an internal
volume of 560 cc and was so situated that mercury could be
added to it from the mercury pump. Hydrogen from the
cylinder was charged to this container at 2000 psi, then by
the addition of mercury to the container, a limited volume of
hydrogen could be delivered up to 5000 psi.
-285-
The equilibrium cell and cooling system described
above were used only to obtain the data at -310 0 and -280 0 F.
For the data at -240 0 F, the equilibrium cell and cooling
system described by Price lO were used.
Analytical Methods
Thermal Conductivity Analysis. There are several

differences in the physical properties of hydrogen compared
to nitrogen and carbon monoxide that could be utilized to
analyze a mixture for hydrogen content. Of these it was
decided that thermal conductivity offered the best combination
of simplicity of equipment, reliability and accuracy.
A Gow-Mac improved standard model thermal conduc-

tivity cell was used for this purpose. These cells measure
only relative and not absolute thermal conductivity, but that is
quite sufficient for this purpose. For this analysis, the
thermal conductivity of the sample gas was compared to the
thermal conductivity of hydrogen. The actual measurement
made was tre emf from the cell and this was compared to the
emf's from the cell wheI;L pure nitrogen and pure hydrogen
were the sample gases.
To calibrate the cell, several known mixtures of

hydrogen and nitrogen, and hydrogen and carbon monoxide were
prepared in a gas mixing apparatus. These known mixtures
were then run on the thermal conductivity cell and the emf from
the cell was measured. At the same time, the emf's from the
cell were measured with pure nitrogen and pure hydrogen as
the sample gases.
The data from the known mixtures were plotted as per

cent hydrogen versus per cent of full scale deflection. Full
scale deflection was defined as the emf from the cell with
nitrogen on the sample side and hydrogen on the reference side
minus the emf from the cell with hydrogen on both sides. This
full scale deflection was of the order of fifty millivolts and
would vary slightly from day to day. However, the same
lOA. R. Price, Ph.D. Thesis, The Rice Institute

(1957).
-286-
sample of gas would reproduce the "per cent of full scale
deflection" to within one part per thousand even though the
zero point of the cell and/or the full scale deflection varied
slightly. The full scale deflection and zero point would be
determined each day that samples were to be analyzed.
The hydrogen-nitrogen mixtures gave a curve that

was slightly different from the hydrogen-carbon monoxide
curve, with the greatest difference toward the low hydrogen
concentration end. For simplification, the data were fitted
to equations and the hydrogen concentration of a sample was
calculated from both equations. A linear interpolation was
used when the nitrogen-carbon monoxide ratio had been
determined. The maximum difference in hydrogen concentra-
tion as calculated from the two equations was about 1. 2 per
cent. Assuming no error in the linear interpolation, the
error in the hydrogen analysis is always less than 4
parts per thousand and this error would be reached only on
hydrogen concentrations as low as 3 per cent. For hydrogen
concentrations of the order of 90 per cent, the error is of
the order of 1 part per thousand.
Chromatographic Analysis. The presence of nitrogen

in a gas mixture that is to be analyzed usually presents a
special problem. It cannot be determined by ordinary means
and is usually, therefore, determined by difference. However,
if nitrogen in low concentration is to be determined with any
accuracy, analyses for the other components must be extremely
precise.
Recent developments in the techniques of gas chroma-

tography have made available a simple and direct method for
determining essentially all the components in a gas mixture.
There have been many recent discussions on the theory

and practice of gas analysis by chromatographic means.
Podbielniak l l provides a good summary of these discussions.
Since a good determination of hydrogen was available,

it was decided that using hydrogen as a carrier gas would
llW. J. Podbielniak and S. T. Preston, Petroleum

Refiner, 34, No. 11, 165 (1955).
-287-
simplify the separation of the components as well as the
determination of composition. The nature of the two compo-
nents indicated that adsorption chromatography offered the
best answer to the separation. Activated charcoal (60-80
mesh) was chosen as the separating medium after several
adsorbents were investigated. The gas chromatography
apparatus is diagrammed in Figure 3.
A flow rate of approximately 130 cc per minute of

hydrogen was used through the column, and 60 cc. per
minute was used through the reference side of the thermal
conductivity cell. The sample was introduced to the space
.
between the two-way glass stopcocks, marked liS" on the
diagram, by purging a generous portion of the sample gas
through the line. The stopcocks were then turned so as to
permit flow of the carrier gas stream through the sample
container, the by-pass stopcock was closed and the sample
was thus carried to the column.
Since it was difficult to reproduce with the preClslOn

desired such operating variables as flow rate, bath temper-
ature, etc., a reference area for nitrogen was determined
before each series of analyses. It should be pointed out
that because of the nature of the relationships of thermal
conductivity of mixtures of hydrogen and nitrogen, and hydro-
gen and carbon monoxide, in the region of high hydrogen
concentrations, the reference areas for nitrogen and carbon
monoxide were identical.
The method was checked by analyzing binary mixtures

of hydrogen in nitrogen and hydrogen in carbon monoxide that
had previously been analyzed by thermal conductivity measure-
ments. The determination of nitrogen and carbon monoxide by
this method are expected to have a relative accuracy of 1 to 2
per cent, limited primarily by the measurement of the areas
under the peaks. An excellent check on all the analyses is
provided by the fact that the determination ."')f each component
is independent, thus the analyses are assumed correct if the
three compositions total 100 per cent within the accuracy of
the determinations.
-288-
Experimental Results
The binary data are presented in Figures 4 and 5 as

pressure-composition diagrams with parameters of constant
temperature. They are also presented in Figures 6 and 7 as
log K versus log pressure plots.
The pressure-composition curves for the binary

systems below 315 psia were drawn by extrapolating the liquid
portion of the curve as a straight line. This straight line
must of necessity, intersect the composition axis at the vapor
pressure of the pure constituent, nitrogen or carbon monoxide,
as the case may be. The consistency of the vapor pres sure
of the pure component with the straight line extrapolation was
interpreted as a check on the method.
With the liquid composition line established, the vapor

composition could then be calculated using Henry1s law. It
was found that the system obeyed Henry1s law at _280 0 and
_310 0 F up to the pressure where dp/dy became infinite.
The ternary data are presented in Figures 8 through

10 essentially as log K versus mol fraction nitrogen in the
liquid, with the exception that at _280 0 F for nitrogen and
carbon monoxide the product Kp was plotted versus mol frac-
tion of nitrogen in the liquid. This was done in order to
separate the curves which crossed one another and led to
confusion when plotted as K versus m·ol fraction of nitrogen
in the liquid.
The points indicated are the experimental points and

the curves represent the smoothed data. The data were
smoothed by first plotting the hydrogen compositions, vapor and
liquid, versus composition of nitrogen in the liquid on expanded
scales and drawing smooth curves through the points. The
nitrogen and carbon monoxide composition in the liquid phase
was then calculated so that their ratio was the same as exper-
imental ratios and the three compositions totaled one.
The nitrogen and carbon monoxide compositions in the

vapor were smoothed by x-y plots consistent with the previously
determined hydrogen composition in the vapor. Consistent in
this case means that the three vapor compositions totaled one.
The binary data were smoothed by the p-x plots. The smoothed
binary and ternary data are presented in Table 1.
-289-
4000
'"
i'
PRESSUR - COM OSITION
DIAGRAM QROGEN-
CARBON M~~~IO~ SYSTEM
50
f - - - --
/ \ ~~t. ~M P:OE~S~~"GEN
NITRpGEN BINARY I
!
3000
1\ ~
ill 200
0.. f
,
iO
- \ I \~
, ...
0.
~
\ \
0 \
/~
\
I
:11'
",
t ..
I.
'<!
"~...
- OOF
/
1000
f '7
" '.to
~
",
" r---
V
-V
0.1 f--
V LJ
o0 0.2 0.' 0.6 0 1.0 o.0 20 50 100
MOL FRACTION HYDROGEN 1000 5000
PRESSURE - PSIA
Figure 5. Figure 6.
I--
HYDROGEN Ie - \ALUE VS MOL
FRACTION NITROGEN IN LIQUID AT -~
-
"'""-.).. ~
0
30
~~
...
0
~ --. r----c o I
Sf A
i
'"3
'\
\.
'\, '\,
\ "\ 10 '-l!. --. I
~
"
.0
0 --, --
0'--
-~ 800 SIA
0
'.
. "- /"
I,
"" I .
I I
--,-I--
--
~ ~
o 1100 SfA. I I
"-" / i
f -~
0
0.1
t- PSIA
2D
i
I"" '0' .20 0.40 0.00 0.80 1.00
PRESSURE PSI A MOL FRACTION NITROGEN IN UQUID
Figure 7. Figure 8.
-290-
(KP) FOR NITROGEN VS MOL FRACTION
NITROGEN IN LIQUID AT -280· F
1000
90 0 I
X-KP AT INFINITE DILUTION EQ(l5
I
-
800
1400 ~SIA
700
SO 0
j I
I I I
500 '"
I IIbo
I
z SIA
III 40 I
o i I
I
0
a: I
~300 I
~~ISIA i •
a: ! .,
...
o
I !
a:: 20O r - r--.:
!
~ r-- ' 5)0 P IA
"ID'
",.
C', 315 P51A

01
I I
100
90
0 I
8 0 020 0.40 O.SO 0.80 1.00
MOL FRACTION NITROGEN IN LIQUID
Figure 9.
(KP) FOR CARBON MONOXIDE VS MOL FRACTION

NITROGEN IN LIQUID AT -280· F
1000
900
800
700
~SOO
x l\~ -A-
~500
F
o ,....---:-
:::0 40
z ~
o
m \\~ ~
~300.....-
---
o
a: -'""
~ 800 PSIA
~
+
_20
-
Q.
'" D"
50~ 1----"7" 0
I 315 PSIA ~
0
0
100
90 X-KP AT INFINITE
80 DILUTION EQ. ( J
70
60
o 0.20 0.40 0.60 0.80 1.00
MOL FRACTION NITROGEN IN LIQUID
Figure 10.
-291-
TABLE I Smoothed Data
Code No. Temp. Press. Liquid Compos1t1on Vapor Composition K-Value
op psla. m.f.H2 m.f.N 2 m.f.CO m.f.~ m.f.N 2 m.f.CO H2 N2 CO
1 -310 315 0.0398 0 0.9602 0.9084 0 0.0916 22.8 .202* .0954

2 -310 315 0.0404 0.2186 0.7410 0.8865 0.0416 0.0719 22.0 .190 .0969
3 -310 315 0.0447 0.7483 0.2070 0.8671 0.0181 0.0242 19.4 .145 .120
4 -310 315 0.0460 0.8270 0.1270 0.8662 0.1l69 0.0169 18.8 .141 .133
5 -310 315 0.0469 0.8776 0.0805 0.8658 0.1234 0.0108 lS.§ .141 .134
6 -310 315 0.0487 0.9513 0 0.8655 0.1345 0 17. .141 .135*
-310 500 0.0609 0 0.0301 0.9320 0 0.0680 15.3 .134* .0725
~ -310 500 0.0658 0.ll65 0.8177 0.9250 0.0158 0.0592 14.1 .136 .0725
9 -310 500 0.0672 0.2170 0.7158 0.9198 0.0285 0.0517 13.7 .131 .0725
10 -310 500 0.0694 0.3120 0.6186 0.9050 0.0401 0.0449 13.2 .129 .0725
II -310 500 0.0706 0.4030 0.5~64 0.9020 0.0495 0.0385 12.9 .123 .0731
12 -310 500 0.0726 0.4784 0.4490 0.9073 0.0583 0.0344 12.5 .122 .0766
13 -310 500 0.0745 0.6195 0.3060 0.9018 0.0~1O 0.0272 12.1 .1l5 .0889
14 -310 500 0.0755 0.7185 0.2060 0.8987 0.0 19 0.0194 1l.9 .1l4 .0942
15 -310 500 0.0761 0.8176 0.1063 0.8964 0.0932 0.0104 1l.8 .1l4 .0978
16 -310 500 0.0763 0.9237 0 0.8948 0.1052 0 1l.7 .1l4 .103*
17 -310 1400 0.1745 0 0.8255 0.9143 0 0.0857 5.24 .176* .104
18 -310 1400 0.1815 0.0940 0.7245 0.9080 0.0168 0.0752 5.00 .179 .104
19 -310 1400 0.1828 0.1l22 0.7050 0.9067 0.0201 0.0732 4.96 .179 .104
20 -310 1400 0.1943 0.2265 0.5292 0.8956 0.0470 0.0574 4.61 .170 .108
21 -310 1400 0.1950 0.2860 0.5190 0.8950 0.0483 0.0567 4.59 .169 .109
22 -310 1400 0.2094 0.4796 0.3ll0 0.8831 0.0810 0.0359 4.22 .169 .1l5
23 -310 1400 0.2190 0.6212 0.1598 0.8765 0.1049 0.0186 4.00 .169 .1l6
24 -310 1400 0.2231 0.6864 0.0905 0.8731 o. :159 0.0110 3.91 .169 .122
25 -310 1400 0.2300 0.7700 0 0.8700 0.1300 0 3.78 .169 .136"
26 -310 2000 0.2490 0 0.7510 0.8790 0 0.1210 3.53 .161
27 -310 2000 0.2815 0.2270 0.4915 0.8508 0.0440 0.1052 3.02 .194 .214
28 -310 2000 0.3000 0.3500 0.3500 0.8356 0.0743 0.0901 2.79 .212 .257
29 -310 2000 0.3317 0.5644 0.1039 0.8090 0.1550 0.0360 2.44 .275 .346
30 -310 2000 0.3446 0.6554 0 0.7977 0.2023 0 2.31 .309
31 -280 315 0.0389 0 0.96ll 0.6803 0 0.3197 17.5 .485" .333

32 -280 315 0.08ll 0.8796 0.6682 0.0392 0.2925 17.0 .483 .333
33 -280 315 g:g~§§ 0.1970 0.7632 0.6504 0.0958 0.2538 16.3 .486 .333
34 -280 315 0.0400 0.2980 0.6620 0.6357 0.1390 0.2253 15.9 .467 .3 1'0
35 -280 315 0.0397 0.5493 0.4ll0 0.5919 0.2563 0.1518 14.9 .467 .369
36 -280 315 0.0392 0.7102 0.2506 0.5680 0.3315 0.1005 14.5 .467 .401
§~
39
-280
-280
-280
315
315
315
0.0385
0.0380
0.0377
0.8651
0.9413
0.9623
0.0964
0.0207
0
0.5563
0.5519
0.5509
0.4037
0.4393
0.4491
0.0400
0.0088
0
iU
14.6
.467
.467
.467
.415
.425
.433"
40 -280 500 0.0674 0 0.9326 0.7619 0 0.2381 11. 3 .448" .255
41 -280 500 0.0689 0.0766 0.8545 0.0335 0.2182 10.9 .437 .255
42 -280 500 0.0707 0.1908 0.7385 g:+m 0.0797 0.1885 10.4 .418 .255
43 -280 500 0.0720 0.2968 0.6312 0.7190 0.1l33 0.1677 9.99 .382 .266
44 -280 500 0.0739 0.5201 0.4050 0.6974 0.1886 0.1140 9.44 .363 .281
45 -280 500 0.0744 0.6576 0.2680 0.6868 0.2352 0.0780 9.23 .358 .291
46 -280 500 0.0745 0.6761 0.2494 0.6851 0.2422 0.0727 9.20 .358 .291
47 -280 500 0.0744 0.8342 0.0914 0.6740 0.2985 0.0275 9.06 .358 .301
48 -280 500 0.0744 0.8455 0.0801 0.67§2 0.3023 0.0245 9.05 .35d .306
49 -230 500 0.0741 0.9259 0 0.66 6 0.3314 0 9.02 .358 .310"
50 -280 800 0.1163 0 0.8837 0.7844 0 0.2156 6.74 .352" .244
51 -280 800 0.1281 0.1342 0.7377 0.7708 0.0492 0.1800 6.02 .367 .244
52 -280 800 0.1342 0.3074 0.5584 0.7539 0.1099 0.1362 5.62 .358 .244
53 -280 800 0.1366 0.4826 0.3808 0.n98 0.1627 0.0975 5.42 .337 .256
54
55
-280
-280
800
800
o .13~8
0.13 1
0.6556
0.7230
0.2066
0.1389
0.7290 0.2120
0.7256 0.2340
0.0590
0.0404
5.29
5.25
32
..32 a .286
.291
5u -280 Boo 0.1383 0.8313 0.0304 0.7213 0.2693 0.0094 5.22 .324 .309
57 -280 800 0.1384 0.8616 0 0.72795 0.2721 0 5.21 .324 .310*
58 -280 1100 0.1727 0 0.8273 0.7848 0 0.2152 4.54 .373" .260
59 -2Bo 1100 0.1860 0.0711 0.7~3 0.7760 0.0257 0.1983 4.17 .358 .267
GO -280 1100 0.2030 0.3452 0.4518 0.7448 0.1237 0.1~15 3.67 .358 .291
61 -280 1100 0.2074 0.4995 0.2931 0.7317 0.1790 0.0 93 3.53 .358 .305
62 -280 1100 0.2102 0.6817 0.1081 0.7208 0.2443 0.0349 3.42 .358 .323
63 -280 1100 0.2ll4 0.76ll 0.0275 0.7181 0.2~27 0.0092 3.40 .358 .334
64 -280 1100 0.2ll6 0.7884 0 0.7175 0.2 25 0 3.39 .358 .338*
65 -280 1400 0.2399 0 0.7601 0.7739 0 0.2261 3.23 .511* .297
66 -280 1400 0.2527 0.0663 0.6810 0.7578 0.0~30 0.2092 3.00 .498 .307
-280 1400 0.2662 0.1634 0.5704 0.7357 0.0 13 0.1830
~~ -280 1400 0.2792 0.2965 0.4243 0.7090 0.1476 0.14~4
2.76
2.54
.498
.498
.321
.338
69 -280 1400 0.3001 0.6522 0.0477 0.6565 0.3247 0.01 8 2.19 .498 .393
70 -280 1400 0.3030 0.6970 0 0.6530 0.3470 0 2.15 .498 .443*
71
72
73
-240
-240
-240
500
500
500
0.0449
0.0426
0.0395
0
0.1486
0.3279
U6§t
0.6326
0.2260
0.1974
0.1660
0
0.1395
0.3090
g:ll~
0.5250
~:~~
4.23
.942
.810
.819
.942 .830
74 -240 500 0.0356 0.508~ 0.4561 0.1357 0.4791 0.3852 3.79 .942
75
76
-240
-240
500
500
0.03~2
0.02 8
0.608
0.8242
0.3580
0.1470
0.1204
0.0945
0.57~8
0.77 8
0.30 §8
0.12 7
3.63
3.27
.942
.942
:~~~
.876
77 -240 500 0.0261 0.9739 0 0.0821 0.9179 0 3.15 .942
7B -240 800 0.1382 0 0.8618 0.3349 0 0.66§1 2.42 .772
-240 800 0.1387 0.1330 0.3095 0.1l20 0.57 5 2.23 0.842
~6 -240 800 0.2256 0.3530 um**
.794
* at infinite dilution ** Crt tical
-292-
It is apparent from the K-x plots that the K-value of
nitrogen at constant temperature and pressure is fairly
independent of composition except in the low pressure-low
nitrogen composition region, while the carbon monoxide
K-value becomes more concentration dependent as the pressure
increases, and it always increases with decreasing carbon
monoxide concentration. The hydrogen K-value seems to be
fairly uniformly independent of composition in the region where
the carbon monoxide composition is low •
. Discussion
Question by M. P. Sweeney, Process Associates: What is the

value of the nearly constant relative volatility between
nitrogen and carbon monoxide?
Answer by author: 1. 05 to 1.10 is the value of the relative

volatility of nitrogen to carbon monoxide over most of the
range.
-293-
F-2
VAPOR-LIQUID EQUILIBRIUM STUDIES ON THE

SYSTEM ARGON -OXYGEN
D. I. J. Wang
Linde Company
Tonawanda, New York
Vapor-liquid equilibrium for tke system argon-oxygen

is of considerable importance in the low temperature separa.-
tion of air constituents. The system in question has been
studied by a number of authors, for example, Inglis 1 ,
Sagenkahn and Fink2 , Burbo and Ishkin 3 , Fastovskii and
Petrovskii 4 , Din 5 , and Latimer 6 . Neither the bubble point
nor the liquid vapor composition measurements by the various
investigators are in complete agreement. In this paper a set
of argon-oxygen equilibrium composition data, taken at approx-
imately one and two atmospheres in an equilibrium still of
novel design, together with a brief discussion on the thermo-
dynamic consistency and compatibility of the observed data,
is presented.
Equilibrium stills in general consist of a contacting

chamber in which thermodynamic equilibrium between phase s
is thought to persist, and means for collecting and analyzing
1 J. K. H. Inglis, Phil. Mag., g, 640 (1906).
2 M . L. Sagenkahn and H. L. Fink, ORSD No. 4493,

Contract OEM sr-685, Pennsylvania State College (1944).
3 p . Burbo and I. Ishkin, Phy. Zeit. der Sowj., 16,

271 (1936).
4V. G. Fastov skii and Y. V. Petrovskii, Zur. Fiz.

Khim., 29, No.7, 1311 (1956).
5
F. Din, Proceedings of the Meeting of the
International Institute of Refrigeration, September 1953.
6 R . E. Latimer, A.I. Ch. E. Journal, 3, No. 1,75
(1957).
-294-
samples of vapor and liquid. In studying the system argon-
oxygen which boils at approximately 90 0 K at one atmosphere,
several considerations lead one to conclude that a still of
somewhat different design might be desirable:
1) The very low boiling point of the system renders

it impractical to attempt analysis in the liquid
phase.
2) Reliable equilibrium data throughout the range of

compositions are sought. However, since the
relative volatility in the vicinity of pure argon is
quite low, in the order of 1.1, the accuracy of
the data attainable in this range in a conventional
still is limited by the analysis of vapor and liquid
samples, the compositions of which differ by but
a small amount.
Unpublished studies at the Linde Company on the

process of mass exchange between vapor and liquid for the
system nitrogen-oxygen indicate that when vapor bubbles,
emerging from a small, single orifice, contact a body of
liquid of but a few milliIneters in depth, equilibrium between
phases is practically established. Subsequently, Chu, et a1. 7
demonstrated that very close approach to phase equilibrium is
attainable on a distillation tray consisting of one bubble cap and
employing as work fluid the system acetone-water. For these
systems, in which the resistance to mass transfer occurs
predominantly in the gas phase, such close approaches to
equilibrium may be readily predicted from a few simple calcu-
lations. These observations suggest then the possibilities of a
still consisting of a number of sieve plates in which the
separation on each tray is identical to that of a theoretical
plate. If such a column is operated at total reflux, the vapor
compositions on successive trays would then constitute one
point on the vapor, y, versus liquid, x, equilibrium diagram
of the more volatile component.
7
J. C. Chu, et a1., Preprint No. 22, Annual Meeting
A. I. Ch. E., St. Louis, Missouri, December 1953.
-295-
In the final design of a still possessing the above
characteristics, the colwnn consists of six trays, one inch
in diameter, with vapor sample taps located beneath each
tray. The trays, furthermore, have a high liquid depth and
low vapor velocity to insure good vapor-liquid contact, a
relatively small nwnber of perforations on the tray to prevent
the coalescence of bubbles, and a very short horizontal path
of the liquid flow to eliminate the existence of a concentration
gradient in the liquid across the tray. Heat for vaporizing the
column liquid is supplied at the reboiler through an electric
heater; liquid oxygen is used for condensing the colwnn vapor.
Proper insulation is provided to minimize heat leak.
A single calibration point on the still has been obtained

by means of a set of trial runs on the oxygen-nitrogen system,
the equilibriwn relations of which are fairly well established,
at a pressure of 56 psia and at a liquid composition of 98 per
cent nitrogen. The agreement between the test result and
that of Dodge and Dunbar 8 was in the order of 1 per cent
of the logarithm of the relative volatility, 0., and was consid-
ered satisfactory.
During the trial runs, it was also discovered that

vapor compositions on the various trays were not at all
affected over a two and half-fold variation in the vapor rate,
serving further to confirm the asswnption that the point and
plate efficiencies are very close to 100 per cent.
Equilibrium data for the system ar gon-oxygen has been

determined between one and two atmospheres absolute over the
entire range of compositions. Over most of the composition
range equilibriwn points were obtained by measuring vapor
compositions of successive trays. In the region of 98-100 per
cent argon, where the relative volatility is in the neighborhood
of 1.1, equilibrium data were obtained by observing the cwnula-
tive concentration change across five trays and the relative
volatility at the average composition over the five trays
8
B. F. Dodge and A. K. Dunbar, J. Am. Ch. Soc.,
49, 591 (1927).
-296-
concerned was then calculated. Reproducibility of the data
was within ± 2 per cent of In a over the entire composition
ranfe. The accuracy of the experimental method is estimated
at - 3 per cent of In a. A small correction factor to the
relative volatility a, in the order of 0.01, is applied to each
observed point in order to bring them all to the arbitrarily
chosen temperature of either 90.7 0 or 95 0 K.
In Figure I and Table I are presented all the experi-

mental data taken on the system argon-oxygen over the
temperature range 90.6 to 95.9 0 K. In Figure 2 is a
comparison between the present experimental results and the
smoothed data of Din, of Sagenkahn and Fink and the correla-
tion devised by Latimer which continues to be successfully
employed in process design at the Linde Company, all at a
temperature of 90.7 0 K. In these figures, it will be noted
that a maximum in the relative volatility has been observed
in the region 0-15 per cent argon. In the following paragraphs,
the plausibility of this observed behavior will be considered on
the basis of its compatibility with one of the criteria of
thermodynamic equilibrium as well as with some other known
properties of this system.
Thermodynamic Consistency with Respect to the Integral

Form of the Gibbs-Duhem Equations
Redlich and Kister 9 have shown the following equation

to be a consequence of the Gibbs-Duhem criterion
In (~) dx=O (1)

"2 p, T
where y 1 and" 2 are the activity coefficients of the more
and less volatile components, respectively.
9 0 • Redlich and A. T. Kister, Ind. Eng. Chern.,

40, 341 (1948).
-297-
0·7
d.'"
0 1~·7~.
0.>
x "P"X.
1 d·f ~
~
1"'-........
N ~--~ 1/1°·7 0"".
~l
~ O.~
~ f:::::-<,..
~ 0.2
15 ';1( v -- ~
0./
o
~ t:=::..
--
o /0 20 ,30 1'0 sO GO 70 80 f" )00
X, /,,!OGc ?.:~ CeNT AJli:.GON' IN ~/~(//o ..
Figure 1. Argon-oxygen equilibrium experimental data
o
o 10 ,?O.$o 'f0 SO GO 7" 80 7'0 /0"
x:, MO~6 /'.:~ C.eNT A,,&l,<1aN /N ~/QGNO ..
Figure 2. Comparison of various argon-oxygen equilibrium

data at 90.7 o K
-298-
TABLE I
Observed Equilibrium Data for the System Argon-Oxygen
Pressure Temp. Per cent a a In a

mm Hg T oK Argon Observed Corrected 90.7 o K
abs. in liquid at T to 90.7 o K
890 91.6 2.4 1. 514 1. 525 0.422

1000 92.3 2.6 1.483 1.506 0.409
1005 92.3 3.2 1. 501 1.521 0.420
994 92.2 3.8 1.508 1.530 0.425
1002 92.2 4.7 1. 521 1. 541 0.433
991 92.0 7.9 1.559 1.574 0.454
993 92.0 9.3 1.533 1.549 0.438
990 91. 8 13.6 1.548 1.561 0.446
1020 92.2 14.0 1. 561 1.580 0.458
1022 92.0 20.3 1. 493 1.508 0.411
1039 92.0 22.1 1.463 1. 477 0.390
1036 91. 9 29.4 1.393 1.405 0.340
1033 91. 6 36.7 1. 381 1.388 0.329
1065 91. 8 46.4 1.331 1. 338 0.292
1062 91. 5 53.5 1. 293 1. 299 0.262
1059 91.3 59.8 1.264 1. 267 0.237
1056 91. 0 65.3 1.246 1.248 0.222
1088 91.2 72.5 1. 200 1.204 0.186
1085 91.1 76.0 1.180 1. 183 0.168
1082 91. 0 78.9 1.180 1. 183 0.168
1033 90.4 79.9 1.179 1.178 0.163
1077 91.0 82.8 1.158 1. 161 0.149
1065 90.7 96.0 1.114 1.114 0.108
1070 90.7 99.1 1. 097 1.097 0.093
1060 90.6 99.3 1.098 1. 098 0.093
1076 90.7 99.8 1. 094 1.094 0.090
a
Corrected In a
to 25. Oo~ 25.0 o K
1325 95.9 2.0 1.466 1. 476 0.390
1322 95.8 3.0 1. 477 1.485 0.396
1285 95.2 4.0 1. 473 1. 475 0.390
1319 95.7 4.3 1. 503 1. 510 0.4l3
1282 95.1 5.8 1.492 1.493 0.401
1224 94.4 11.5 1. 495 1.490 0.399
1219 94.1 18.5 1.433 1. 425 0.354
1216 93.9 24.5 1.400 1.390 0.329
1670 95.7 94.4 1.106 1.109 0.104
-299-
At moderate pressures, relaxation of the constant
pressure restriction introduces but a negligible error; thus
In (-'!....!..) dx =- 0 (2)
Y2 T
Satisfaction of equation (2) may therefore be con-

sidered a necessary but not sufficient test of thermodynamic
consistency.
o
Calculations on the system argon-oxygen at 90.7 K
have been made and the results are tabulated in Table II in
terms of the ratio
s: In (2 ) dx
Y2
, where lIn ....!...L I is the
1 '(2
'I 1
.2L I dx absolute value of In(--)
So lIn
'{2 Y2
TABLE II
Thermodynamic Consistency of Vapor-Liquid Equilibrium
Relations for the System Ar gon-Oxygen at 90. 7 0 K
Present Linde Company Sagenkahn Din

Expe riments Correlation and Fink
-0.009 -0.015 -0.189 -0.321
Here it is seen that the equation of Redlich and

Kister may be considered as met by the Linde correlation to
within 1-1/2 per cent and by the present experiments to within
1 per cent; this latter close agreement is believed to be some-
what fortuitous as the experimental data is thought to be no
more accurate than'±' 3 per cent in In a. The x-y
equilibrium data of Din and of Sagenkahn and Fink seem to be
low, causing an unbalance to the Redlich-Kister criterion to
the extent of 32 and 19 per cent, respectively. Not sufficient
experimental points are available in the present work at 95 0 K
to permit the locating of the In a versus x curve with the same
degree of confidence. Using tentatively the graphical repre-
sentation as shown in Figure 1, the "unbalance" at this temper-
ature is roughly -4 per cent. The Linde Company correlation
at 95 K, as at 90. 7 Of<, is in very close agreement with the
0
-300-
Redlich-Kister criterion, since this criterion had to be
satisfied in the formulation of that correlation. It is also
noted that the 4 per cent "unbalance" in the author's experi-
mental work at 95 0 K appears essentially to be all in one
direction and is, of course, roughly equal to the amount of the
deviation of the present work from the Linde Company
correlation at the temperature in question.
Relative Volatilities at Infinite Dilutions According

to Thermodynamically Exact Equations
One of the interesting points brought out by the

present work is the observed shape of the In a versus x curve
in the vicinity of pure oxygen. It would prove especially
enlightening to determine if the observed relative volatility in
this region is compatible with other known thermodynamic
properties of this system.
At infinite dilution, i. e., as x approaches zero, the

following equation may be rigorously derived from the general
relation of thermodynamic equilibrium for two-phase, two
component systems, 10
(a-I) 0 = ( .!..2'..) ]
x- a x p x-O ( 3)
where ~ H2 is the heat of vaporization of the less volatile

component.
Equation (3) enables one to calculate the relative

volatility at the oxygen extremity of the concentration range
provided that the bubble point data of the mixtures and the
latent heat of vaporization of pure oxygen are available. Such
calculations have been made at 90. 7 0 K and are shown in
Table III.
10H. S. Taylor and S. Glasstone, A Treatise on

Physical Chemistry, Vol. 2, D. Van Nostrand Company, New
York, New York (1951).
-301-
TABLE III
Calculated and Observed Relative Volatilities of the
System Argon-Oxygen at Infinite Dilution
(a.-I), Calculated from the (p, T, x) Data of

Per Cent
TOK Argon in Din Sagenkahn and Burbo and
Liquid Fink Ishkin
90.7 0 0.51 0.52 0.55
Per Cent (a.-I), Observed

TOK Argon in Present Linde Company Sagenkahn
Liquid Experiments Correlation an d F"In k Din
90.7 0 0.51 0.60 0.50 0.51
Here it is interesting to note that the observed ( a. -1)'s

at infinite dilution in the present work as well as of Din and of
Sagenkahn and Fink all seem to be in good agreement with
calculated values based on Equation (3), implying therefore that
these observed values are quite plausible. Moreover, if the
observed "end value" is correct, then In a. must undergo
some rather strange variations with x in order to satisfy the
integral form of the Gibbs-Duhem equation discussed earlier.
Conclusion
In summary, an equilibrium still consisting of a

multitude of sieve trays, each so designed as to yield the
separation of a theoretical tray, has been constructed and
employed in the procurement of vapor-liquid equilibrium data for
the system argon-oxygen. A number of advantages for an
equilibrium still of such a design are suggested.
1) Continuous withdrawal and analysis of samples

eliminate transient state uncertainties and other
difficulties often encountered in connection with the
batch sampling techniques of conventional stills.
-302-
2) For syste111s c0111paratively easy to separate,
equilibriu111 data over the entire range of
c0111positions 111ay be deter111ined si111ultaneously
by 111easuring the c0111position on successive
trays.
3) For syste111s difficult to separate, or at C0111pO-

sitions in the vicinity of pure c0111ponents where
the difference in concentration between the vapor
and liquid is s111all, accurate equilibriu111 data
111ay be obtained by observing the cU111ulative
separating effect of a nU111ber of trays.
It is believed that such a still should find application

in 111any low boiling syste111s in which the resistance to 111ass
exchange occurs predo111inantly in the gas phase. For syste111s
that are liquid phase controlling, however, phase equilibriu111
111ay not be practically attainable in such a device.
Ther111odyna111ically, the syste111 argon-oxygen, long

thought to be "regular" by various investigators, e. g., Burbo
and Ishkin, Din, Fastovskii and Petrovskii and Lati111er, see111S
to exhibit a s111all but detectable departure fro111 this regularity
near the oxygen end, while over the range 20-100 per cent
argon in the liquid, the agree111ent between the pre sent work
and the regular solution correlation of Lati111er is good. This
s0111ewhat unexpected 'behavior see111S, however, to be ther111o-
dyna111ically consistent and c0111patible with other observed
properties of the syste111.
Data gathered at Linde plants over the years were

used by Lati111er to establish the plausibility of his correlation.
Si111ilar calculations e111ploying these sa111e plant data and the
observed equilibriu111 c0111positions fro111 the present work
yielded slightly 1110re pronounced scatter in the calculated tray
efficiencies than with the Linde correlation. However, in spite
of these s111all discrepancies, the results of the present work
are c0111patible with Linde plant data. In a private c0111111uni-
cation, R. E. Lati111er indicates that he believes it i111probable
that a five-constant integration of the Gibbs-Duhe111 equation is
required over one -third of the c0111position range of the author's
experimental data, whereas one constant is adequate over the
remaining two-thirds.
-303-
Sufficient data have not been gathered in this work
for correlation over the entire range of pressures and
temperatures of interest. This paper has instead been pre-
sented chiefly to suggest the possibilities of the novel
equilibrium still employed in these studies and to bring to the
attention of investigators in the field this rather abnormal
behavior in the argon-oxygen system with the hope that more
extensive and elegant endeavors may thus be stimulated.
Discussion
Question by M. P. Sweeney, Process Associates: What happens

to the activity coefficients in the 0-20 per cent argon region?
Answer by author: Activity coefficients of the two components

were not individually measured in this work. Whenever a
maximum in the ln a ver sus x curve occur s, however, the
Gibbs -Duhem relation then dictates the existence of a maximum
in the activity coefficient Y 1 of the more volatile component
and a minimum in the activity coefficient y 2 of the less
volatile component at the same liquid composition x.
-304-
F-3
EPOXY RESINS AS CRYOGENIC STR UCTU RAL ADHESIV ES~'
R. M. McClintock and M. J. Hiza

Boulder, Colorado
Introduction
This investigation of copper to copper bonds using

filled epoxy resins as adhesives was undertaken to provide
design information for a liquid hydrogen cooled electromagnet.
It was proposed that the adjacent coils of the magnet be spaced
with an epoxy resin adhesive which would provide structural
rigidity as well as electrical insulation at the operating
temperature of 20 0 K (-253°C). Since the completed coil would
be acted upon by thermal stresses and electrical reaction
forces, the present program was undertaken to evaluate the
mechanical strength of several adhesives subjected to treat-
ment corresponding to use in the end application.
The use of filled epoxy resin adhesives was suggested

for the ma,net application on the basis of purely pragmatic
information obtained in an earlier search for an adhesive
useful at temperatures as low as 4 o K. When it was found
that these adhesives exhibited desirable low temperature
properties it was decided that their use in the above applica-
tion only required verification of the fact that properties of
the bond are not seriously harmed by some of the treatment
the completed assembly would receive. In short, this study
is more the evaluation of the effects of some engineering
variables upon the properties of a few commercially available
adhesives than it is an attempt to tailor an optimum cryogenic
adhesive.
'~This work was supported by the U. S. Atomic Energy

Commission.
IG. VI. Koehn, Adhesion and Adhesives Fundamentals

and Practice, John Wiley and Sons, Inc., New York, New
York (1954).
-305-
These adhe sives were two component systems
consisting of a liquid amine type curing agent and an epoxy
resin paste containing an inorganic filler material. The filler
was predominantly alumina and asbestos totalling approximately
55 per cent of the resin-filler mixture. The curing agent
used was diethylaminopropylamine, a relatively slow reacting
type recommended fOr elevated temperature curing.
The following groups of experiments were performed:

The impact strength of the cured adhesive in bulk was
measured at 300 0 and 76°K; the thermal expansion of the
cured adhesive between 76 0 and 300 0 K was measured; tensile
and shear strengths of copper to copper bonds were measured
at 300 0 , 76 0 , and ZOoK after rapid cooling to these tempera-
tures; and using room temperature tensile and shear strengthS
as indices, the effects of exposure to high temperatures,
exposure to low temperatures, and adhesive thickness were
determined.
Test Procedures
The tensile and shear specimens were fabricated from

commercial electrolytic tough pitch copper. (See Figure 1.)
With an adhesive thickness of O.OZO inch in the shear specimen,
the center line of the adhesive coincided with the axis of the
threaded ends. The tensile specimens were designed to give
a circular cros s section 0.750 inch in diameter with bonded
surfaces perpendicular to the axis of loading. The end threads
on both specimens were standard 3/4 inch -10, and were lathe
cut in all cases to insure concentricity with the axis of the
specimen.
The same method 0.£ preparing the copper surfaces

to receive the adhesive was used for all specimens. The
machined surface was fir st degreased with carbon tetrachloride.
Wet crocus cloth under running water was then used to remove
oxide and salt deposits and a rinse was performed in distilled
water. The surface was then etched for about one minute in a
room temperature solution composed of 15 cc FeC1 3 (4Z per
cent aqueous solution), 30 cc HN0 3 (15N), and 197 cc HZO. The
specimens were then given a final rinse in distilled water and
dabbed dry with tissue. If deposits could still be seen on the
surface, the process was repeated. The adhesive bonds were
made as soon as possible after this cleaning process was
completed to prevent the formation of oxide on the surface.
-306-
t. 0
,
V
/
~
-.,...."..\
, /'
V
,........
.mc..<I) -.cJutl.n "-
0
""'
Figure 1 • Figure 2. Impact strength of

a filled epoxy resin vs. temp.
0000
'[
--- ~IOU
"l
1\ 0000
. ,
'"
e-
~; I
/ """""""
"""" ¥
'1t
o
- ....~ f..--
.......---
V
Figure 3. Strength of a filled epoxy o
resin adhesive bond vs. exposure
temp. prior to testing at room temp.
Figure 4. Strength of a filled

epoxy resin adhesive bond vs.
temperature
-307-
Specimens subjected to elevated temperatures prior
to testing were maintained at the temperatures noted in the
results for eight hours in a helium atmosphere. Those
subjected to low temperatures prior to testing were alter-
nately cooled and warmed twice between 76 0 and 300 0 K by
immersion in liquid nitrogen and warming in ambient air to
room temperature. Cooling and heating rates were conse-
quently quite high.
Figure I also shows the specimen used for conducting

impact tests on the cured adhesive in bulk. The specimens
were cast over-size in aluminum foil molds, cured, and
machined to the proper dimensions. Te sts were performed
both on notched and unnotched specimens. Bubble entrainment
was minimized by vacuum curing, but some small bubbles
were still present in the cured specimens. The testing was
done in a standard plastics impact testing machine. Low
temperature tests were performed sufficiently fast after
removal from liquid nitrogen that the temperature rise was no
greater than 3 0 C.
As it turned out, the impact strength was so low that

the kinetic energy imparted to the broken pieces accounted for
the major portion of the indicated strength at the lower values.
This was eliminated from the results by replacing the broken
ends in the machine and noting the energy re uired to
accelerate them by the swinging pendulum. The impact values
shown ill Figure 2, therefore, are accurate to perhaps only
50 per cent, but they do show notch sensitivity of the material
as well as increased brittleness with decreasing temperature.
Results
Nearly all the tensile specimens failed near the center

of the glue line so that considerable adhesive remained on both
broken halves. The exceptions to this cohesive type of failure
were the specimens which had been subjected to temperatures
above the cure temperature prior to testing at room temperatur
As the exposure temperature increased, these specimens showel
a change in mode of failure from cohesion to what was apparent
a failure of the bond between metal and adhesive. (The only
inspection of the surfaces after testing was visual, however, so
that a very thin layer of remaining adhesive would not have
been detected.) Tensile specimens given a pretest exposure
at' 423 0 K (150°C) and subsequently tested at 300 0 K showed about
-308-
a 90 per cent cohesive failure with the remainder of the
failure occurring adhesively. The specimens which had been
o
subjected to 43S K (165 C) failed onl} about 50 per cent
0
cohesively.
The decrease in strength of tensile specimens

treated in this way shows that while elevated temperatures
cause chemical damage to the adhesive material itself, there
is a temperature above which more serious damage occurs
to the adhesive bond. This is apparently caused either from
chemical damage to the molecular bond between metal and
adhesive, thermal stresses at the high temperatures which
are great enough to cause local failures of the bond, or both.
(See Figure 3.)
Without exception, the shear specimens tested under

all conditions fractured in such a way as to leave the entire
thickness of adhesive on one-half of the specimen in any
given local area. The other half in this area contained
either no adhesive or such a thin film that it could not be
detected by visual examination. It was at first suspected that
the molecular bond between metal and adhesive was weaker in
shear than the adhesive itself, and so always allowed failure
at the metal-adhesive interface before the shear strength of
the adhesive itself was attained. But small metal fragments
often torn from the copper surfaces by the adhesive testify
to a bond between metal and adhesive strong enough to consider
another possibility. It may also be that the fracture actually
occurred in the body of the adhesive while always leaving at
least a monomolecular layer of adhesive on one metal surface.
If we consider the fact that the thermal stresses in the
adhesive are a maximum near the metal-adhesive interface and
decrease toward the center of the adhesive layer, we realize
that a superimposed mechanical shear should produce failure
in a plane very near the interface. Such thermal stress exists,
of course, even in specimens tested at room temperature, if
the adhesive has been cured at an elevated temperature. In
fact, cure shrinkage establishes some initial stress even in
bonds cured at room temperature. The obvious remedy for
such steady state thermal stresses is to match the thermal
expansion of the cured adhesive with that of the metal.
Figure 3 shows average values of tensile and shear

strengths at 300 0 K as a function of the pretest treatment. The
-309-
sharply diminished strengths caused by elevated temperatures
are clearly shown.
Subjection of the cured bonds to low temperatures

prior to testing at room temperature also decreased the
shear strength, but the effect was not as great on the tensile
strength (a cohesive property of the adhesive itself as shown
by the mode of failure) as it was on the shear strength (a
property of the adhesive bond).
Specimens tested at cryogenic temperatures without

previous treatment exhibited a considerable decrease in shear
strength at 76°K compared to the room temperature values,
but there was no further decrease as the test temperature was
lowered to 20 o K. (See Figure 4.) The tensile strength under
the same test conditions is also shown in Figure 4.
Thermal Stress
If an adhesive bond is prestressed in some way by its

thermal history, less of its native strength will be available to
absorb service stresses before failure of the bond. Such
thermal stresses may not be large when adhesive bonds are
used at normal temperatures, but cryogenic adhesives often
are subjected to several such stresses of considerable magni-
tude, the total effect of which is to impose serious limitations
on the service strength. To illustrate the fact that this effect
can be large, can in fact cause failure of the bond, some of
the stresses which adhesives normally encounter in cryogenic
use will be considered.
An adhesive bond to be used at, say, 20 o K, will be

pre-stressed at that temperature by shrinkage of the adhesive
during cure, and the difference in thermal contraction between
adhesive and adherend over the range from the cure tempera-
ture to 20 o K. The first of these, cure shrinkage, is probably
least among the rigid adhesives in the epoxy resins, and
partially explains why a previous effort to find an adhesive
suitable for cryogenic use disclosed the e~oxy resins as the
best performers. The work of de Bruyne indicates, however,
2
N. A. de Bruyne, J. Appl. Chern., 6, 303 (1956).
-310-
that this shrinkage is not negligible. Within the family of
epoxies, shrinkage can be minimized by controlling the degree
of cross-linking during cure and by adding fillers to the
resin. But minimizing cure shrinkage would be desirable, of
course, only if it did not involve a more serious compromise
with other properties of the cured adhesive.
The second stress mentioned, that attributable to a

difference in thermal contraction between adhesive and
adherend, is present as a shear stress at the adhesive-
adherend interface and as a tensile stress throughout the
adhesive itself. In this regard, its effects are the same as
those caused by cure shrinkage, but the inference is that a
match in thermal expansion between adhesive and adherend
would produce the strongest cryogenic bonds. Figure 5 shows
the measured expansion of unfilled and a particular filled epoxy
compared with some of the cryogenic metals and with Pyrex
glass. The amount and type of filler as well as ~article size
are known to affect the properties of the filled resin, however.
For example, using glass filled epoxies in an attempt to
approach more closely the expansion of copper, we have
recently made bonds which are even stronger at 76°K than at
room temperature. This is in spite of the fact that the presence
of particles of a filler material of low expansivity reduces the
expansion of the mixture at the expense of stresses induced
in the resin component as it attempts to contract around a
particle of filler. The net effect is to distribute thermal
stresses throughout the adhesive instead of allowing them to
concentrate at the adhesive -adherend interface.
Extending this observation to the range between cure

and room temperature helps determine an optimum cure
temperature. If the same degree of polymerization of a resin
is effected by curing at two different temperatures, then the
lower temperature should be chosen if maximum bond strength
is desired. Choosing the cure temperature in this way mini-
mizes the effect of a mismatch in adhesive and adherend
expansions in cooling to ambient conditions from the cure. The
polymerization of the resin and its resulting properties, how-
ever, are often quite sensitive to cure temperature, so that the
lowest cure temperature may be chosen only when it does not
involve a compromise with the properties of the adhesive itself.
-311-
UU.-JLL[.n
r.00)()I Qt."IIt-J
(~!> ,. AL.U M '1J"1-7-r~"T/""7-r-r?,""""/"'7"""71
r'L L.ED
A"" A"'B~">TO'5)
~t>OXY Rt';'''1
hLUUIUUU
copp~o
'\04. ...T4IULr....,,>
">Tr.r.L
o 2. 4 to e 10 12. " 10·'

TOTAL (.XPAU'SIOU rQOU 7<.,° TO ~OOoK
Figure 5. Thermal expansion of

some adhesives and adherends.
8000
I
T~tJC;IOt.J
coooo
rr
Cl
-
~
1- 4000
I.!J
::2
LJ
CJ
tr ., 1.1I~.h.12
<!OOO
o 10 I"; e.o z.o;

ADIKC)IVt;: TI.lICKtJC"le;· UILe;
Figure 6. Strength of a filled epoxy

resin adhesive bond vs. thickness
-312-
The above argument based on thermal stress is
probably insufficient to account for the entire amount by
which the observed shear strength is reduced at low temper-
atures. Neither does this argument account for the lower
room temperature values observed after a thermal cycle to
76°K. It is not expected that any chemical or other irrevers-
ible changes are produced in the metal or adhesive by the
temperature cycling, so the most logical explanation of these
effects seems to lie in the consideration of the transient
stresses produced by the cooling rate. Even if adhesive and
adherend possess similar expansions between the temperature
extremes, the adhesive will lag the metal in most changes
of temperature because of its lower thermal conductivity and
the contact resistance presented at the interface. The stress
produced in cooling and warming may thus be sufficient to
cause local failures of the bond. This would explain the
observed residual effect on shear strength. (See again Figure 3.)
Furthermore, since the adhesive has the greater

expansivity, one would expect its temperature lag to help
prevent such local failures in cooling. But in the warm-up
process the opposite is true: the temperature lag of the adhe-
sive serves to increase the .transient thermal shear stress. A
system bonded with an adhesive of different expansivity and
thermal conductivity than the adherend, then, should be brought
to temperature rather slowly in order to prevent thermal
stresses from becoming high enough to cause local failures in
the bond.
The residual effect on tensile strength may be caused
in the same way, except that the reduction observed is due to
local failures within the adhesive caused by the fact that the
adhesive is a heterogeneous mixture of two components of
diff e ring expan s i vity •
Some other remedies to the situation suggest themselves

immediately. Filling the adhesive with material which would
improve its thermal properties, e. g., metal powder or filaments,
would decrease the transient thermal stress for a given cooling
or heating rate. The electrical properties, however, may be
correspondingly changed by improving the thermal properties
and this must be considered if high dielectric strength is a
requirement. Likewise, thinner glue lines are preferable to
thick ones when a choice is possible.
-313-
Epstein 3 mentions that the strength of adhesives which
do not liberate gase s during cure show little dependence upon
glue-line thickness. He states, however, that thickness
should not exceed ten mils for optimum strength. Figure 6
shows strength versus thickness at 300 0 K of bonds which
received no treatment other than the usual curing cycle prior
to test. The slight curvatures are probably better attributed
to experimental error than to any real effect.
Conclusions
The following conclusions have been noted in the

course of this investigation:
1) The impact strength of the bulk adhesive is low

at room temperature and decreases to a very low
value at 76°K. It displays increasing brittleness
and notch sensitivity with decreasing temperature.
2) Temperatures above the cure temperature cause

damage to epoxy resin adhesive bonds and result
in decreased tensile and shear strengths. In
particular, epoxy resins polymerized with
diethylaminopropylamine are seriously harmed by
temperatures of 125 0 C and higher.
3) Rapid thermal cycling between 76 0 K and room

temperature prior to testing at room temperature
results in decreased shear strength of these
adhesive bonds while the tensile strength is
affected only slightly.
4) Bonds tested at 20 0 and 76°K (again cooled

rapidly) displayed roughly the same tensile strengtl
as at room temperature, but showed a considerabll
decrease in shear strength at the low temperatures
There is no significant change in shear strength
between 76 0 and 20 0 K.
3C • Epstein, Adhesive Bonding of Metals, Reinhold

Publishing Corporation. New York, New York (1954).
-314-
5) Given a strong adhesive bond at room tempera-
ture, the match between thermal expansion of
adhesive and adherend is a good criterion of its
strength at low temperatures.
6) Thickness of the adhesive bond was found to have

no effect on strength between 10 and 30 mils.
Discussion
Question by W. H. Schweikert, General Electric Company:

The subject of cleaning surfaces was covered adequately.
However, were the surfaces prepared by sand blasting or was
there a rough machined surface used at the resin joint?
Answer by author: Prior to cleaning, the metal surfaces were

in a condition produced by fine machining operations, that is,
about a 32 micro-inch finish.
Question by J. Keeley, Martin Company: How were tensile

strengths affected by addition of Pyrex glass filler to the
unfilled adhesives?
Answer by author: We tested only shear strengths of glass

filled bonds, and even then not to determine the effect of
filler, but only the effect of temperature as stated.
Question by J. D. Sudbury, Continental Oil Company: Did you

determine tensile strengths of epoxy-aluminum or epoxy-
stainless steel bonds as well as epoxy-copper?
Answer by author: No.
Question by J. D. Sudbury, Continental Oil Company: Have you

determined fatigue strengths of epoxy-metal bonds?
-315-
F-4
THE IMPACT TESTING OF VARIOUS ALLOYS

AT LOW TEMPERATURES
R. P. Mikesell and R. P. Reed

Boulder, Colorado
Introduction
In recent years there has been an increasing demand

for materials suitable for fabrication of equipment utilized at
cryogenic temperatures. In order for the designer to properly
choose such materials, it has been found necessary to acquire
reasonably accurate data on mechanical properties at low
temperatures. One of the most important properties with
which the designer is concerned, is the notch toughness or
the impact strength of notched specimens. It has been found
that the face centered cubic type of metals are tough at the
very low temperatures when subjected to an impact type of
loading. Perhaps the most important of these metals are
copper, aluminum, and nickel. It is expected that the alloys
of these elements are also tough at cryogenic temperatures.
This paper presents results of the testing from 296 0 to 240 K
of notched impact specimens of copper, aluminum, and Haynes
(nickel) alloy~. The majority of these alloys exhibited only a
moderate change in toughness with a large change in tempera-
ture but a few were found to be brittle at the low tempera-
tures.
Specimens
All of the material was machined to the ASTM speci-

fications for Charpy impact specimens' l Different types of
notches were machined; a keyh01e-notcli for the oxygen free
high conductivity (OFHC) copper, aU-notch slightly modified
from ASTM specifications 2 for all but two of the Haynes alloys
and the V -notch 3 for the remainder.
1,2,3 A • S • T • M ., D eSlgna
' t'Ion E 23 - 47T •
-316-
With the exception of the OFHC copper, the notch of
each of the copper alloys was milled with a one tooth cu~ter.
For the OFHC copper, a special jig was devised to drill the
hole of the keyhole accurately. The slot of the keyhole-notch
was cut with a saw. The notch of each of the aluminum
alloys was milled with the one tooth cutter.
The Haynes alloys were found to be extremely diffi-

cult to machine. Several carbide-tipped saws were broken
in an attempt to cut the specimens to the approximate dimen-
sions. Also, it was almost impossible for a high speed steel
drill to drill the notch of a specimen to any depth even at a
low drill speed. Cobalt drills proved to be more successful;
that is, a cobalt drill could go entirely through the specimen,
but a somewhat inaccurate and rough hole resulted. Therefore,
it was desirable to develop a method for fabricating another
type of notch. At the suggestion of several members of the
machine shop, a modified U -notch was devised that could be
machined with a grinding wheel. The notch was designated
a modified U -notch because it was relieved at an angle of 50
on each side. The bottom of the notch was the same radius
as the hole of the keyhole-notch. A very smooth and accurate
notch resulted from these efforts. Furthermore, the grinding
process did not produce a work-hardened region at the notch.
Tests were run at room temperature and it was found that the
U-notch specimen exhibited the same impact strength as that
of a keyhole specimen. The notch of each of the alloys C and
X was milled with the one tooth cutter and ground to the
correct dimension.
The specimens of all three types of alloys were

longitudinally cut from the stock.
Experimental Procedure
The specimens were loaded as simple beams in a

standard pendulum type of impact machine having a 30 pound
hammer. The hammer was dropped from a height of 4 feet
and had a velocity of 16 feet per second at the point of impact •
. Testing near 1940 K is accomplished by quickly trans-

ferring the specimen from a mixture of dry ice and methyl
-317-
cyclohexane. Rapid transfer of the specimen from a liquid
nitrogen bath makes testing possible near 76°K.
Testing near 20 0 K is also accomplished by rapid

transfer of the specimen from a liquid hydrogen bath to the
machine. Too rapid a temperature rise is prevented by means
of two items. The first is a small paper boat glued to the
specimen which retains liquid hydrogen for about 3 seconds.
A top added to the paper boat serves to minimize the conden-
sation of air into the liquid hydrogen. The other is a special
tool which also contains some liquid and in addition serves
as a transfer and centering device for the specimen. With
this method and adhesive tape on the supports of the machine,
the temperature rise with aluminum, the metal with the
highest expected temperature rise at 20 0 K, is not mOre than
5 0 C if the specimen can be broken within 4 seconds after
removal from the liquid. Neither the adhesive tape nor the
paper container have any measurable effect on the impact
strength of a metal specimen.
Accuracy
At anyone temperature, the deviation was about 5 per

cent from the average impact strength.
Results
The results of the testing of copper, aluminum, and

Haynes alloys at temperatures of 296 0 , 196 0 , 80 0 , and 240 K
are presented in Figures 1, 2, and 3. These impact strengths
are reported for a particular set of conditions and would be
expected to change if the conditions were changed. The most
important of the conditions to be considered are heat treatment,
grain orientation, strain rate, and specimen geometry.
The description in the following section of the fracture

appearance of each specimen is the result of a macroscopic
examination.
-318-
COPPER ALLOYS
OL-~--~---L--'=OO~~--~'OO~~--~~~~--~~~--L-~XO
I
TEMPERATURE - oK
Figure 1.
0
P---....
100
/ r-- ~-t. ...-? ~

/ 1-;>k ./
0
...rl
'" ~ V
-;rs_f
:-- r--
~
#
0
ALUMINUM ALLOYS, ALL V - NOTCH
0
24S-'T4
7!5S-T6
0 00 100 150 200 250 XO
TEMPERATURE - °K
Figure 2.
100
HAYNES ALLOYS
0 1,-
trANi, U·N01CI1 ::::::l'
~_,,01C> -
V ;:::.-0 t-;;:l.LO'l~' ,
-
- - " ........-: f-"" HAST. C, V-NOTER:
.--0 '25. U."01CI'

0 V i - - I--
B. U·NOTCH
0 100 150 200 250 XO

TEMPERATURE -·K
Figure 3.
-319-
Copper Alloys
All of the copper alloys were machined to ASTM

specifications for a Charpy V -notch with the exception of the
OFHC copper. The OFHC copper was machined to the Charpy-
keyhole because the OFHC specimen cut to the V -notch pre-
sented two difficulties: The ductility of annealed copper is so
great that impact specimens are merely deformed, not broken.
With the large cross section of the V -notch specimen, the
high amount of energy required to push the deformed speci-
men through the supports is a larger fraction of the total
energy than the fraction required to push the unbroken keyhole
specimen through the supports. This extremely high energy
of the V -notch specimen is near machine capacity and so
causes deformation of the striking edge.
The OFHC copper was in the annealed condition; the

Everdur 1010, a silicon bronze, and Phosphor Bronze Grade
D, a tin-copper alloy, were in the hard condition; the Berylco
25, a beryllium-cobalt alloy, and Berylco 10 were in the half-
hard condition.
All of the copper alloys decreased in strength with a

decrease in temperature.
However, the strength of OFHC copper increased to a

maximum at 80 0 K. None of the specimens were torn, only
bent. There was a considerable amount of plastic deformation
at the notch, even at a temperature of 240 K. In addition, the
ends of each specimen were smeared as a result of the speci-
men being pushed through the supports of the machine. The
energy expended in this manner, of course, added to the
impact value. Because of this, the values of the OFHC
copper cannot be compared quantitatively with the strengths of
the other alloys.
The Everdur 1010 changed from a moderately fibrous

fracture at a temperature of 296°K to a moderately granular
fracture at 24 0 K.
The Berylco 25, in spite of having the same strengths

as the Everdur, exhibited a moderately granular fracture
throughout the entire temperature range. Both the B'erylco 25
and Everdur did not break completely even at a temperature of
24 0 K.
-320-
The Berylco 10 exhibited a very granular fracture
and broke completely at all temperatures.
It may appear from the curve in Figure 1 that Phos-

phor Bronze may have undergone an abrupt transition in the
296 0 to 196°K range. However, examination of the fractured
pieces shows that there is still a region of shear failure
at the low temperatures indicating that the decrease in
impact strength is probably more gradual. Consequently,
none of the copper alloys exhibited an abrupt ductile to
brittle transition.
Aluminum Alloys
All of the aluminum alloys were machined to ASTM

specifications for the Charpy V -notch.
The following alloys were tested: 75S-T6, 24S-T4,

2S-H14, 54S-0, 5Z:S-F. With the exception of 2S-H14, all
of the aluminum alloys decreased in strength with decreasing
temperature. The strength of 2S-H14 increased to a maximum
at 80 K.
0 A drop in the recorded impact strength between the
temperatures of 80 0 and 24 0 K was accompanied by a change in
the fracture appearance. That is, the fracture changed from a
very fibrous appearance at a temperature of 80 0 K to a partially
fibrous appearance at 24o K. Also, the plastic deformation
decreased by about 50 per cent, but was still quite significant.
Since the specimen broke completely at 24 o K, its strength
cannot be compared quantitatively with the strengths of the
specimens broken at the higher temperatures since these speci-
mens did not break through completely.
Both the 54S-0 and 52S-F exhibited a very fibrous

fracture at room temperature. Despite the large decrease in
impact strength both the 54S-0 and 52S-F still showed some
ductility at a temperature of 24 0 K. Also, the 54S-0 and 52S-F
did not break completely at 240 K.
Both 24S-T4 and 75S-T6 exhibited granular failures

and broke completely even at room temperature.
-321-
Haynes Alloys
The main alloying elements of each of the Haynes or

nickel alloys are as follows: Alloy 25, cobalt, chromium,
nickel, tungsten; Hastelloy B, nickel, molybdenum; Hastelloy
C, nickel,' chromium, molybdenum; Alloy X, nickel, chro-
mium; Multimet, iron, chromium, nickel. cobalt.
The alloys C and X were machined to the Charpy

V -notch while the alloys 25, B, and Multimet were machined
to the modified U -notch.
These alloys were in the hot rolled condition and

stress relieved at 2200 0 F for 10 minutes.
All of the Haynes alloys decreased in strength with a

decreasing temperature. However, the values of the alloys C
and X cannot be compared quantitatively with the other alloys.
The toughest of these alloys were 25, B, and Multimet. Alloys B
and 25 were, in fact, found to be too tough for the machine at
room temperature in the form of a Charpy V -notch.
The Multimet gradually changed from an almost

entirely fibrous appearance at a temperature of 296 0 K with a
comparatively large degree of tearing just below the notch to
about a one-half fibrous appearance at 240 K.
The fracture of Alloy 25 was moderately fibrous in

appearance at a temperature of 296°K and changed very little
when the temperature was decreased to 24 0 K.
Hastelloy B behaved nearly the same as Alloy 25.
The fracture of Alloy C was moderately fibrous at

296°K and changed very little in appearance when the tempera-
ture was decreased to 240 K.
The appearance of Alloy X changed from an almost

entirely fibrous break at 296°K to a moderately granular break
at 240 K.
Each of the Haynes alloys broke completely at all

temperatures.
-322-
Conclusions
Of the alloys tested, 25, B, Multimet, 2S-H14, and

OFHC copper were found to remain exceptionally tough to the
lowest temperature. None of the alloys underwent an abrupt
ductile to brittle transition.
As is always the case with the standard impact test,

the data as presented here are absolute only for the given
conditions and hence cannot be extrapolated to a material in a
different size or a different notch. These data, however, can
be used by the designer as a guide for the intelligent selection
of cryogenic materials which may contain areas of stres s
concentration. That is, if a certain material in the form of a
notched impact specimen exhibits a notch toughness at a
particular temperature, then the material will probably be
tough at the same temperature in a larger size. On the other
hand, a material in the form of a notched impact specimen
which fractures in a brittle manner at a particular tempera-
ture should almost certainly be rejected as a fabricating
material at that temperature unless adequate safety factors
are introduced.
Discussion
Question by V. J. Johnson, CEL National Bureau of Standards:

How is the ft-lbs of impact measured on the testing machine?
Answer by author: The units of energy are marked on an

attached scale, which is calibrated on the basis that the differ-
ence in the height of the pendulum hammer before and after
the impact is a function of the total amount of energy absorbed.
Question by D. R. Cheyney, Kaiser Aluminum: Since the

aluminum alloys do not break clean in the impact test, how do
you interpret this from the standpoint of the suitability of such
a te st for aluminum alloys? For this reason, the aluminum
industry has not recognized this test as suitable for aluminum
alloys.
Answer by author: It is true that the energy expended in push-

ing the unbroken specimen through the machine supports is not
-323-
strictly a part of the impact strength. On the other hand,
one might consider that the Charpy test is a good test for
aluminum alloys in that it enhances their property of
ductility.

How were the specimens cooled and maintained at low temper-
ature during the test?
Answer by author: Testing near 1940 K was accomplished by

quickly transferring the specimen from a mixture of dry ice
and methyl cyc1ohexane. Rapid transfer of the specimen from
a liquid nitrogen bath made testing possible near 76°K.
Testing the specimen near 20 0 K was accomplished by a rapid
transfer from a liquid hydrogen bath to the machine. Too
rapid a temperature rise near 20 0 K was prevented by means
of two items. The first was a small paper boat glued to the
specimen which retained liquid hydrogen for about 3 seconds.
A top added to the paper boat served to minimize the conden-
sation of air into the liquid hydrogen. The other is a special
tool which also contained some liquid and in addition served
as a transfer and centering device for the specimen.
Question by R. S. Hampton, Lox Equipment Company: Have

you made any impact tests at liquid helium temperature?
Question by W. H. Schweikert, General Electric Company:

Were the specimens stress relieved after machining the notch
or keyhole on the Haynes alloys? Experience in our labora-
tories indicate residual surface stresses (tensile) to be as high
as 100,000 psi on the surface. These were determined by
x-ray diffraction technique and have been found to reduce to
zero as much as 0.003 inch from the surface. These high
residual stresses can overshadow test results.
Answer by author: It was not necessary to stress relieve the

specimens after machining the notch. The method of machining
the notch, i. e., grinding with a profuse flow of cooling oil,
resulted in virtually no work hardening at the notch as indicated
by a microscopic examination.
-324-
F-5
PERFORMANCE PREDICTION OF LOW TEMPERATURE

AIR-COOLING EXTENDED SURFACE BY SEVERAL
COMPARATIVE METHODS
C. F. Kayan and R. G. Gates

Columbia University
New York, New York
Introduction
In a paper by one of the authors some time ago on

fin heat transfer l , use was made of the geometrical electrical
analogy in studying temperature patterns of extended surface,
particularly those located internally within the element
structures. One of the fins studied was of thin section, shown
in Figure 1; its length beyond the primary surface was twenty
times its own thickness, such as might be found with a
metallic fin as often used in low temperature air heat transfer
equipment.
Figure 2 shows the general layout of the model section

as studied by the geometrical electrical analogy method, using
a conductive sheet. Considering the fin itself, it will be seen
that the effect of the fluid thermal resistance in the heat flow
pattern was explored via twenty conducting sheet strips leading
from the main electrode to the fin; provision was included to
allow for the heat flow between the fluid and the tip-surface
(edge) of the uniformly thick and longitudinally extensive fin.
In the pre sent study, dealing particular 1y with fin

surfaces and projections as found associated conventionally
with gas processing equipment, it is proposed to further explore
the performance of the extended surface by several comparative
methods, including proviSions for varying conductivity.
In the earlier paper, use was made, through the elec-

trical model results, of the potential difference ratio c, defined
as the ratio of the partial measured potential difference D.e x
for a given point within the fin structure, to the total or overall
potential difference D. e t applied to the model circuit
IC. F. Kayan, Ind. Eng. Chem. 40, 1044 (1948).
-325-
,,~
J1...o,[
Figure 1.
Figure 2 .
..
r' ..
'\ \\
..
, , ~
.~.
---
Figure 3.
-326-
c = Ae / Ae ( 1)
x t
From the Analogger values of c (regarded as negative or

positive-depending on the relative position of the fluid
temperature 1£, i. e., either higher or lower than the fin
temperature), equivalent temperatures throughout the fin
assembly were determined,
tx = tf +
- cAt = tf ± c9 (2)
t
where t f was the fluid temperature, and At and 9 t' the

temperature span between the fin base and fluid contacting the
extended surface.
In the present paper, use will again be made of a

temperature difference ratio C for the fin, to apply to the
overall temperature difference between the base end of the fin
and the temperature of the fluid enveloping the fin. The
analysis by the different methods will again be for steady state,
and it will be presumed that the temperature across the section
of the fin material is uniform. For most of the studies, the
fin will be considered as of constant cross section, of constant
thermal conductivity, with the fluid of constant temperature and
constant thermal conductance, and with no internal heat sources;
these are, of course, the conventional assumptions of extended
surface heat transfer analysis.
The order of treatment is as follows: mathematical

relationships; computative numerical method; electrical-
analogy methods.
Mathematical Relationships
Consider zero heat flow at tip edge surface (insulated).
Base-Outward Approach: Figure 3a shows a rectangular

cross section, longitudinally extensive fin, in heat transfer
relationship with a fluid of constant temperature and of constant
surface conductance value, and attached to its base section.
Figure 3b shows the temperature pattern as a function of
distance x outwards from the base to the full length L. Figure
3b, as a matter of convenience, is transformed into Figure 3c,
with temperature difference a shown as a function of distance x
-327-
from the fin base outwards to the tip, where 8 L defines the
temperature difference there, as compared with 8 at the base.
The usual differential equation may be set up to ctover this
case, following the application of the general as sumptions
specified previously, and leading to the conventional expres-
sions,
( 3)
This equation has a solution of the form
a = ( 4)
l/l
where n = [(P/Ac)(h/ k)] (5)
The solution here is the well recognized expression

(enx + en(lL-x»
= 8t lnL ( 6)
(e + 1)
Following the treatment in the earlier paper, as expressed in

general by equation (1) and for convenience in use, the result
for the fin conditions may be resolved into the fin potential
ratio Cf ,
(e nx + en(lL-x) )
= ( 7)
(e 2nL + 1)
In addition, the value for the mean temperature difference 8 m

between the constant temperature fluid and the variable
temperature fin may be shown to be,
1 S L
(8dx)
__ 8t (e 2nL - 1)
- ( 8)
L 0 nL (e lnL + 1)
8 1 (e lnL _ 1)
C - 2!!. :: (9)
f, m - at nL (e 2nL + 1)
-3l8-
C is often known as .. fin effectiveness." From equation
f,m
(8), the heat flow between the constant temperature fluid and
the variable temperature fin (as controlled by its properties,
etc.) may be calculated,
= (10)
Tip-Inward Approach: Whereas the above treatment

was in terms of a base-outwards approach, another method of
attack, based on the identical assumptions as for the
preceding case, equations (3-10), may be employed. Figure 3
(right) herein is the counterpart of Figure 3 (left), with x
varying from 0 at the tip to full value L at the base. Using
the general solution form of equation (4), the result may be
shown in terms of the fin base temperature difference between
fluid and base, 9 t . Here likewise the fin edge surface is
assumed to be insulated, that is, no heat flows at point x = 0,
or more specifically, here d9/dx O. =
Thus,
9x = ( 90 I 2)( e nx + e
-nx
) ( 11)
and (12)
It is to be pointed out that at x = L, 9L = 9 t'
This equation, by manipulation, could be resolved into

the form of equation (7), but the present form and viewpoint are
convenient and serve a worthy purpose, as noted subsequently.
The fin dimension ratio PIA included within the value

of n takes on specific values for ext~nded surface elements of
different shapes:
PIA = rrDI (rrD 2/4) = 4D, for a round rod of

c
diameter D ( 13)
PIA c = 4/a, for a square rod of side a ( 14)
-329-
PIA = 2(a + b) I (ab), for a rectangle a x b (15 )
c
PIA = 2/a, for a longitudinally long rectangular

c
section rib of thickness a (16)
Table I, Column A, records C f values for the

rectangular 20:1, insulated end and no heat flow at the edge
fin, in terms of equation (12), calculated with the following
properties: h = 36 BTU/(hr sq ft OF), k = 30 BTU/(hr sq.
ft of/ft) , L = 4 in., a = 0.20 in., n = 12, nL = 4. The
fin may be considered, as in the original case and z"S shown
in Figure 2, as divided into 20 parts or strips, with values
stipulated for each midpoint, and as shown in Table I.
Consider with heat flow at tip edge surface (non-

insulated) .
Base-Outward Approach: Continuing the analysis with

the same type of fin, but considering the tip edge surface to
be in heat transfer relationship with the fluid, via the tip edge
area A c ' with tip edge fluid conductance he and the tip edge
temperature difference 8 L , for the case represented by Figure
3 (left),
(2h Ikn)(en(L-x)_ en(L+x)) ]

8t e
[nx n(2L-x)
8
x = 2nL
e + e +
e +1 nL -nL
(l+h e .I kn)e + (l-h
e
I kn)e
(17)
8x 1 [nx n(2L-x) +
( 2h e ikn)(e n (L-x) - en(L+x)) 1
Cf = = --- e +e
(l+he/kn)e
nL
+ (l-he/kn)e
-nL
8t e 2nL +l
(18)
- 330-
It is to be noted that with tip edge h = 0, equiva-
e
lent to an insulated tip, equation (18) reduces to equation (7).
Tip-Inward Approach: Following the attack presented

by Figure 3 (right), and allowing for tip edge surface heat
flow, i.e. at x = 0,
-nx
+ (l-he/kn)e ] (19)
nx -nx
(1 + hel kn)e + (1 - h e /kn)e
(20)
nL -nL
(1 + he/kn)e + (1- he /kn)e
Again, considering tip edge ~ = 0, equation (20) reduces to

equation (12).
Column F of Table I represents C values for the

Figure 3 (right) type of approach, with f
he / kn = 36 / (30 x 1 2) = O. 1 0 ,
for the illustrative fin, calculated via equation (20).
Table I, it is to be noted, shows series of compara-

tive values for the temperature difference fraction C f as
referred to in the earlier part of the paper, i. e. equation (2).
The present treatment is specifically directed, it should be
emphasized, to the determination of Cf by various means and
under different conditions. The specification of fin tempera-
ture and fin heat transfer follows from the evaluation of C f
and C f , m'
Computative Numerical Method
As suggested by the tip-inward approach typified by

equation (12), and also by an earlier comparative study in
another but related realm 2 the fin problem may be tackled via
a computative procedure, one which appears to have attractive
aspects of flexibility.
2C. F. Kayan,Trans. A.S.M.E., p. 9, January 1949.
-331-
Figure 4 shows. in simplified form. the division of a
typical fin into uniform s trips A x wide. The ener gy flow
process involves both fluid surface heat transfer
[ A% = (P A x)h 0] and fin conduction heat transfer
= (A k/ Ax)~). where 0 represents the fluid versus

c
surface temperature difference. and A 0 represents the
temperature difference impelling the heat flow across Ax along
the fin. Starting at the tip end strip and proceeding progress-
ively inward towards the remotely located base (as far distant
as desired), it is recognized that the energy conducted base-
ward. qc,is that surface received qh' Thus. for a strip
element,
(21)
Hence. in increments starting initially at the tip strip. the con-

ducted heat, Aq • to the next strip. is that previously received
c
as the total of surface heat transfer A qh; this accounts for a
temperature change (i. e •• a change in 0). amounting to AO at the
next inward strip. Thus for this change in 0 due to each strip
effect,
AO = (Ax/Ac k ) ~h = (Ax/Ack)(P AxhO) = (P/Ac)(h/k)(Axl
(22)
The general objective of evaluating Cf is readily

achieved by this computative procedure. Formalizing, and
starting at strip I with some arbitrarily assumed initial value
of 0 such as 1.00 (unity) at the tip end, and moving on to
strip 2. the step wise procedure is,
where F = (P / A c )(h/ k)(Ax2 ) (24)
-332-
This step wise computation is, of course, readily repeated
in the traverse from left to right towards the base. The
base can be any number of strips distant from the t~p, with
the successive strip computations via equation (23), i. e. ,
in effect, 8 = 8 2 + 1(~8)2 + 2(~8)3' etc., up to N in
num ber. It3 is to be noted that prosecution of the calculation
procedure, leading to the ultimate evaluation of C f , may
equally well be carried out graphically.
Based on the same rectangular fin as outlined above,

a 20:1 ratio, same properties etc., and where
F = (2/ a)(h/k)(~~h = 0.04, then for the start:
1 (~ 8) 2 = O. 04 (8 1 ) = O. 04 (1. 00) = O. 04; 8 2 = 1. 00 + O. 04 =
1.04.
Corresponding values calculated for the cited example,

in terms of C f , are shown in Column B of Table I, and thus
are the counterpart of the values shown in Column A as

calculated by equation (12). Here the strip width ~ = 0.20/12
= 0.0167 ft , 2/a = 2/(0.20/12) = 120, and h/k = 36/30
= 1.2, in homogenous units obviously. Good agreement is
indicated, as inspection of the values reveals. The value of

the h/k ratio depends, of course, on the value of the tempera-
ture dependent value k.
Electrical Analogy Methods
Consider first uniform and constant fin property

characteris tics.
Figures 2 and 4 lead directly to the electrical network

model development as represented in Figure 5, with its various
accessory aspects and with previous network procedures in
retrospect.
Resistance wise, following equations (21) and (22) in

principle, for a given strip the resistance to fluid heat transfer
may be represented by Rh =
1/ (P ~ xh); correspondingly, the
resistance to heat conduction for this strip element is
Rc =A x/(Ack). Thus,
-333-
Figure 4.
Figure 5.
10
~
V
/~
/; /
6
-- - --
~
::::V
4
/0 /
~ ~
2
- - I---""
~
00 4 6 10 12 14 16 18 20
POSITION ON FIN
Figure 6.
-334-
Rh/Rc :=(Ack)/ [(PDoX)(hDoX)] := 1/[(h/k)(P/A C )(.:;\x2)] :=
S/m := l/F (25)
where m := h/k; (26)
Thus, for the rectangular fin studied above and reported

on in Columns A and B of Table I, the pertinent fin data
include: m:= 36/30:= 1.2; P/A c := 2/a := 2/(0.20/12) := 120,
and Dox:= 0.20/12 := 0.0167; S := 30, and Rh/Rc := 30/1.2 := 25.
These results, including Rh/R c ' directly represent the
thermal data.
Zero Heat Flow at Tip Edge Surface: On the basis

of the above, ~lectrical rui~e network was built, in
fixed scale relationship to thermal resistances, and circuited
as indicated in Figure 5. It thus represents, with no end
flow provision, the identical problem evaluated by equation (12)
(Column A), also by computative numerical method (Column
B); the analogy results for C f (obtained directly from the
circuit) are reported as Column C in Table 1. It will be
noted that the agreement between the three procedures (i, e. ,
values A, B, C) is quite acceptable.
Columns D and E represent electrical analogy results

for two additional caSes on the same fin structure, namely
m := h/k := 0.60 and 0.40, respectively. For comparative
purposes, the values of Columns C, D, and E are plotted in
Figure 6, and show the variation of C f versus strip positions
along the fin.
Heat Flow at Tip Edge Surface (Considered non-

insulated): It willbe 7ecalled that in accordance with the
outlined procedure, C f was evaluated by equation (20) for the
comparative case of m := h/k := 1.20 and allowing for the
tip edge heat flow. Values are reported as Column F in
Table 1. Similarly, results were obtained for the same case
using the electrical simulation model and reported in Column
G (m := 1.20). Circuit results were also obtained for m := 0.60,
and shown in Column H. Thus the results for cases without and
with tip heat flow may be compared as Columns C and D versus
F and G, for values of m := 1. 20 and 0.60, respectively. The
-335-
values of m serve to directly modify the corresponding
values of the resistance ratio Rh/Rc'
Comparing the analogue (ColUlnn G) with the equa-

tion value (Column F), it may be noted that the agreement
is within experimental error. In further comparison,
Columns G and H versus C and D (all obtained by electrical
analogy), the slight impact of included tip edge heat transfer
is most noticeable at the regions near the tip end.
Consider now the effect of variable physical proper-

tie s a long the fin.
From examination of equation (25), it is apparent that

the physical properties involved in the fin heat transfer have a
very direct bearing on the analogue resistance ratio Rh/Rc
employed (in proper scale relationship) in the construction of
the model. Among the important factors are the value of h,
of k as varying with temperature primarily, and the value of
P and Ac or their ratio PI Ac representing structural shape
primarily.
Thus, whereas in all the other cases treated here, the

ratio Rh/Rc was constant as a function of the fin length x, it
might be visualized as varying, as shown in Figure 7. The
inverse ratio Rc l Rh is likewise shown here, for convenience.
Obviously, if R h , representing fluid surface heat transfer
resistance, is regarded as constant over the length of the fin,
then any variation of R due to variation either of cross section
c
area Ac or to thermal conductivity k, will have a direct bearing
on the Rh/Rc ratio. This is suggested in Figure 7. The
constant value of Rh/Rc is also shown here.
Suffice it to say, there are cases where Rh/R is not

constant over the length of the fin: typical of such a c~se •
representing the impact of a varying value of Rc by varying Ac
is the wedge -shaped fin. Column I of Table I represents this
case, with A varying regularly from the tip to the base, and
c
hence, by equation (25), Rh/Rc varying accordingly. Figure 8
depicts the variation of the resistance ratio Rh/Rc along the
fin length for the wedge-shaped fin, with appropriate scale
values. The constant Rhl Rc value of 25 for the rectangular fin,
- 336-
FIN POSITION
Figure 7.
Figure 8.
Figure 9.
-337-
illustrated in the different case treatments, is likewise shown.
Figure 9 shows the comparative results of Columns I and G,
i. e., wedge versus rectangular shape.
In the actual development of the analogue model

construction for the wedge-type fin, R h , representing the
fluid heat transfer resistance, was regarded( as required)
constant in value; hence the conduction Rc resistance was
appropriately adjusted. Actually, as already noted, the only
variable here entering into the deviation from constancy of the
Rh/R ratio has been that of a changing A , while all the
other cfactors covered in equation (25), of fi. k, P and li x,
are constant.
Further reference to variable Rh/Rc due to varying

thermal conductivity k will be found in the conclusion below.
Conclusion
As mentioned in previous discussions, the step-wise

treatment of the fin problem permits adjustment of the working
values in terms of the particular thermal conductivity at a
given point. Accordingly, on the basis of an initially assumed
value of k applicable to the range in question, position values
for C f over the fin length may be determined. Equivalent
values of fin temperature over the length may then be estab-
lished by equation (2). Then, in terms of the available k
versus temperature curve for the particular material and
temperature range (i. e., low temperature range in the present
application), applicable values of thermal conductivity at each
fin position may then be set up. The evaluation process is then
repeated. Thus the procedure is one of successive approxima-
tions; two sets of evaluation should in general be sufficient, and
are readily accomplished by either procedure.
As indicated for one of the early cases, the mean value

of the temperature difference serves for the computation of the
overall heat flow qf for the fin, as presented through equations
(9) and (10). Thus, qf = A f h8 m , where 8 m = C f , m 8 t. 8 m may
be established, as an average,either directly from a plot of 8 x

along the fin-such as represented by Figures 6 and 9-or by
-338-
arithmetic average of the C f values at the different fin
positions, as obtained by any of the different comparative
methods cited. em may also be further determined from the
electrical model.
If R t represents the total measured resistance of the

analogue model across which the electrical source potential is
imposed, and if Rh t represents the net fluid side resistance
of the multi-paralle'led Rh values, Cf,m will be defined from
(27)
as (28)
Examination of Figure 5 for the simulation circuit,

shown in simple form, discloses that the 1Ifluid temperature"
bus circuit can readily be modified to include variable resist-
ances between the successive junctions, with a continuing
connection at its end to the IIground" side of the system. In
this way, adjustment of the fluid "potential" in the circuit along
the "fin length" may be introduced, to permit investigation
of cases where the fluid temperature is not constant over the
fin length.
It is of course obvious, furthermore, that in the

different cases studied in the present paper, uniform symmetri-
cal conditions on both sides of the fin have been presumed.
Provisions may be made similarly as covered in the original
geometrical Analogger study, involving a building structure
problem of a floor as a fin, for asymmetrical conditions of
fluid temperature as well as of surface conductance.
It is also to be pointed out, that the resistive net work

model may readily be developed in further detail within each
section to show up internal variations of fin temperature in the
transverse directions, similarly as explored in the Analogger
study mentioned above.
To reiterate, both the computative numerical method and
the electrical analogy procedure provide useful tools for investi-
gation of fin structures with variable circumstances of structure
and physical properties, in which the use of orthodox mathemati-
cal approaches begets cumbersome problems.
-339-
TABLE I
Values of C f at Different Positions Along Fin

For Different Cases
Fin Poai tion Data

Btrip-lbint-Va1ue _With Zero Tip Heat Flow With Tip Heat Flow
No. (Mid) nx ABC D E F G H I
1 0.5 0.10 0.037 0.037 0.037 0.119 0.180 0.034 0.033 0.1l2 0.036
2 1.5 0.30 0.038 0.038 0.038 0.121 0.185 0.036 0.036 0.116 0.037
3 2.5 0.50 0.041 0.041 0.041 0.125 0.192 0.039 0.039 0.121 0.050
4 3.5 0.70 0.046 0.046 0.046 0.133 0.201 0.044 0.044 0.129 0.064
5 4.5 0.90 0.052 0.052 0.053 0.143 0.213 0.051 0.051 0.140 0.081
6 5.5 1.10 0.061 0.061 0.061 0.155 0.228 0.060 0.060 0.154 0.101
7 6.5 1.30 0.072 0.072 0.072 0.171 0.247 0.071 0.072 0.170 0.124
6 7.5 1.50 0.086 0.085 0.086 0.191 0.269 0.085 0.086 0.190 0.150
9 6.5 1.70 0.103 0.102 0.104 0.214 0.294 0.102 0.103 0.214 0.181
10 9.5 1.90 0.124 0.124 0.126 0.242 0.322 0.124 0.126 0.242 0.216
11 10.5 2.10 0.151 0.153 0.153 0.274 0.335 0.150 0.153 0.275 0.256
12 11.5 2.30 0.184 0.166 0.186 0.311 0.394 0.183 0.186 0.3l2 0.302
13 12.5 2.50 0.224 0.226 0.226 0.355 0.437 0.224 0.226 0.356 0.355
14 13.5 2.70 0.271 0.274 0.275 0.405 0.485 0.272 0.275 0.407 0.412
15 14.5 2.90 0.331 0.334 0.335 0.464 0.547 0.332 0.334 0.467 0.477
16 15.5 3.10 0.403 0.407 0.410 0.531 0.608 0.402 0.408 0.535 0.550
17 16.5 3.30 0.494 0.495 0.490 0.613 0.678 0.494 0.497 0.615 0.631
16 17.5 3.50 0.602 0.605 0.608 0.704 0.755 0.602 0.606 0.708 0.722
19 18.5 3.70 0.741 0.748 0.742 0.809 0.849 0.740 0.740 0.813 0.823
20 19.5 3.90 0.902 0.900 0.905 0.933 0.946 0.902 0.903 0.933 0.937
Base 20.0 4.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00
Acknowledgment
Acknowledgment is hereby gratefully made to the

author's associates for their continued advice and cooperation,
specifically to J. A. Balmford on electrical details, and to
G. F. Boeker on mathematical details.
This paper reports the results of another phase of

the investigations supported in part by a grant-in-aid from
The Refrigeration Research Foundation.
-340-
Nomenclature
L = Fin length, ft
t = Temperature, of
Cf = Fin temperature difference ratio, fraction of
overall value e t
e = Temperature difference, of 1/2
n = Fin characteristic = [(PI AcHh/k)]
P = Fin perimeter, ft.
Ac = Fin cross sectional area, sq ft
= Total temperature difference, overall, between

fluid and fin-base
h = Fluid surface heat transfer conductance, BTU
per (hr sq ft 0 F)
k = Thermal conductivity of fin material, BTU per
(hr sq ft OF 1ft)
q = Heat flow rate, BTUlhr
R = Thermal resistance, OF I(BTU/hr)
Subscripts are introduced where required, and

along with auxiliary symbols, are locally defined.
Discussion
Question by S. E. Colucci, Stratos-Division of Fairchild Engine

and Airplane Corporation: Is the method outlined in your paper
applicable to the condition where the fin is immersed in the
fluid being cooled and the coolant is flowing inside the base?
Answer by author: Yes it is. The methods outlined cover the

performance of the fin itself, from the tip to the base, and
as shown in equation (28), provide the means for computing
the thermal resistance for the fin between the "immersion
fluid" and the base. The thickness of the base support to
which the fin is connected may be considered to contribute
some additional resistance commensurate with its thickness
and its thermal conductivity-and possibly the joint itself may
add a little, thus reflecting the extent of poor connection. In
this way the total resistance of the assembly may be evaluated,
as effective between the "immersion fluid" and "coolant"
temperatures, i. e., these set up the temperature difference
actively responsible for the heat flow.
-341-
F-6
TEMPERATURE-PRESSURE-TIME RELATIONSHIPS
IN A CLOSED CRYOGENIC CONTAINER
S. C. Huntley
National Advisory Committee for Aeronautics
Lewis Flight Propulsion Laboratory
Cleveland, Ohio
Introduction
One of the problems in the storage of cryogenic

liquids is the increase in pressure in the closed container.
A considerable quantity of heat may be transferred into the
container because of the large temperature difference between
the ambient temperature outside the container and the temper-
ature of the cryogenic liquid. The increase in pressure in a
closed container due to this heat leak may be calculated from
the increase in liquid temperature, obtained from the heat-
leak rate, and the vapor pressure-temperature equilibrium
relation. Experience has shown, however, that the container
pressure may be considerably greater than the calculated
pressure based on an average liquid temperature. The
difference between the actual pressure and that calculated
from an average liquid temperature has been attributed to
the development of temperature gradients in the liquid.
O. A. Hansen,lLinde Air Products, Inc., and M. D.
Andonian,2 Cambridge Corporation, referred to the presence
of liquid-temperature strata in their papers presented at the
1956 Cryogenic Engineering Conference.
As part of the low temperature work at NACA, an

experimental study was made to verify the development of
temperature strata as a contributing factor to the increase in
pressure and to investigate the effect on pressure by altering
the temperature strata. Both the liquid and the vapor
temperatures of liquid nitrogen were measured in a closed
10. A. Hansen, Proceedings of the 1956 Cryogenic

Engineering Conference, p. 75, Boulder, Colorado, Septem-
ber 1956.
2M. D. Andonian, Proceedings of the 1956 Cryogenic
Engineering Conference, p. 83, Boulder, Colorado, Septem-
ber 1956.
-342-
cryogenic container. The container was provided with
means of altering either the liquid- or the vapor-temperature
gradients. Different heat-leak rates into the container and
different quantities of liquid were used in the investigation.
Procedure
The liquid nitrogen was contained in a 30 by 93 inch

metal dewar, shown in Figure 1. The dewar consisted of a
0.037 inch thick, 304 stainless steel inner shell surrounded
on the sides and bottom by a vacuum jacket. Radiation to
the inner shell was reduced by an aluminum foil wrap. The
lower section was further protected by a copper radiation
shield attached to the inner shell and also wrapped in alumi-
num foil. The dewar was closed by an uninsulated steel lid.
Temperatures were measured by two copper-constantan
thermocouple rakes. One rake measured temperatures at 8
inch increments along the entire height of the dewar; the
other rake measured temperatures at 1/4 inch increments
near the surface of the liquid.
The liquid-temperature gradient was altered by

stirring the liquid with a hand cranked propeller. An
electric fan was used to alter the vapor -temperature gradient.
The pressure history and the temperature history for each
increment of height were obtained for an established heat-
leak rate and a given quantity of liquid nitrogen in the dewar.
The heat-leak rate was changed by changing the pressure in
the vacuum jacket.
The method used to determine the heat-leak rate and

to calculate the increase in pressure is shown in Figure 2.
The vapor pressure-temperature equilibrium relation of
liquid nitrogen, obtained from NBS literature, 3 is shown,
with vapor pressure as a function of temperature.
The reference temperature used in this investigation is the
boiling point of liquid nitrogen at atmospheric pressure. The
measured increase in average liquid temperature was divided
by the time interval that the container had been closed to
obtain the rate of increase in average liquid temperature.
This rate of increase in average liquid temperature was used
as an indication of the heat-leak rate into the container for
the purpose of this investigation. The proportionate increase
3NBS Circular 564, p. 360.

-343-
OBTAINED FROM NBS LITERATURE
,-TO MANOMETER
r.==~d'I=60~ATMOS VE,T
INCREASE IN AVERAGE lI0UID TEMP _
INTERVAL OF TIME WITH CONTAINER CLOSED -
(M, A'v,
80 ~,oF;HR ~ HEAT LEAK RATE
6T~ A~
60 --r;:-)X 61. - PeAL AT .61 •
• O.H" 304 5.5. INNER SHELL
VAPOR
WITH ALUM-FOIL WRAP
PRESSURE,
HAND - CRANKED PROPELLER IN. Hg GA~5
COPPER RADIATION
SHIELD WITH
ALUM-FOIL WRAP- -_--.............
'-VACUUM JACKET
20
PRESS'>
o 5 10 15 20 25
TEMP .. OF ABOVE REFERENCE
Figure 1. Liquid nitrogen Figure 2. Vapor pressure-

dewar temperature equilibrium
relation of liquid nitrogen
HEAT LEAK RATE. (t::.T"AV ;(1) = 9.2° F/HR

HEAT LEAK RATE. (6T1 •AV /6.) -9.2' F/HR
LIQUID CONTENT, 30 'Yo ; AFTER 29 MIN
L10UID CONTENT, 30% 100 ROOM TEMP."",!
I
20 I
80 I
EXPERIMENTAL
28
60
" " 40
""
HEIGHT,
""
IN.
,-
20
.,.," "'--CALCULATED (REF. LID. N2 AT
,,'" '" ATMOS. PRESS.I
o 10 20 30 40 o 100 200 300 400

TEMP, OF ABOVE REFERENCE
TIME, MIN
Figure 3. Typical Figure 4. Typical temp.

pressure history gradient in closed container
-344-
in average liquid temperature at any time after the container
was closed was used with the curve of Figure 2 to get the
calculated pressure for the same time interval.
Typical Pressure History and Temperature Gradient
A comparison of a typical pressure history and the

calculated increase in pressure with time is shown in Figure
3. Pressure is shown as a function of time. The experi-
mental pressure obtained without stirring the liquid or mixing
the vapor and the calculated pressure are shown for a heat-
leak rate that resulted in a rate of increase in the average
liquid temperature of 9.2 0 F per hour with the dewar 30 per
cent full. The experimental pressure is observed to exceed
the calculated pressure from the instant the container was
closed, and the difference between the two pressures
continued to increase with time. After 30 minutes, the
experimental pressure was about 1-1/2 times greater than
the calculated pressure. Conversely, the calculated
pressure was experienced in about 1/2 the anticipated time.
A temperature gradient measured after the container

had been closed for 29 minutes is shown in Figure 4.
Temperature is shown along the abscissa for different heights
in the container. Container height is also represented by a
partial view, shown on the right. Temperatures measured at
8 inch increments varied from near reference temperature in
the liquid at the bottom of the container to near room
temperature in the vapor at the top. A sudden increase in
temperature occurred near the surface of the liquid.
Temperatures measured at 1/4 inch increments near

the surface are shown to an enlarged scale in the upper left
corner of Figure 4. The temperature at the surface had
increased 6.6 0 F above reference, whereas the average
liquid temperature was slightly less than 4. SO F above
reference. This average liquid temperature had been used
to obtain the pressure calculated from the vapor pressure-
temperature equilibrium relation (previously presented in Figure
2). It was found that using the surface temperature and the
same equilibrium relation resulted in a vapor pressure equal
to the experimental pressure. In other words, the difference
between the experimental and the calculated pressures was
shown .to be due to the development of the liquid-temperature
gradient.
-34S-
Effect of Stir ring Liquid on Pres sure
Because the difference between the actual and the

calculated pressures was due to the difference between the
surface temperature and the average liquid temperature,
altering the liquid-temperature gradient should have an effect
on container pressure. The effect on pressure due to
altering the liquid-temperature gradient was demonstrated by
stirring the liquid to equalize the liquid temperatures.
Figure 5 presents typical pressure and liquid-

temperature histories obtained with stirring. Pressure and
temperature are shown as functions of time including the time
interval when the liquid was being stirred. The experimental
and calculated pressures, the temperatures at the surface and
at the bottom, and the average liquid temperature are shown.
The change in experimental pressure is observed to follow
the change in surface temperature at all times. Stirring the
liquid decreased the surface temperature and slightly
increased the temperature at the bottom until the liquid
temperatures were equalized, as shown by the average liquid
temperature. The minimum pressure with. stirring was
obtained when the liquid temperatures were equalized, and
this pressure corresponded to that calculated from the average
liquid temperature. These results show, therefore, that the
average liquid temperature can be used to anticipate the
container pres sure, provided the liquid is stirred to equalize
the liquid temperatures.
Effect of Heat-leak Rate and

Liquid Content on Pressure
If the liquid is not stirred to equalize temperatures,

further consideration of the factors that caused the liquid-
temperature gradient to develop may be of benefit to the
designer of cryogenic container s. Referring to the container
used in this investigation (Figure 1), we see that an increase
in liquid temperature at the surface was probably due to two
factor s:
1) Conduction of heat through the inner shell from

the vacuum jacket, which with natural convection
caused the heated, less-dense liquid to accumu-
late at the surface.
-346-
HEAT LEAK RATE. (6T",AV 161).9.2- F/HR
lIaUID CONTENT. 30~_
20
PRESSURE, 10
IN. HQ. GAGE
30 40
~ I-STIRRING LIQUID
Figure 5. Typical pressure and

liquid temperature histories
20 VACUUM JACKET PRESSURE, < 5 MICRONS
LIQUID CONTENT,
10 ".
,,---:::;:::::.:::--~
~=-~--~~--~'---~80~--'IOOkn---7.'20
PRESSURE.
IN. Hg GAGE
TIME. MIN
Figure 6. Pressure histories for several cases
VACUUM JACKET PRESSURE, < 5 MICRONS
LIQUID CONTENT, 70% LlaUID CONTENT. 30%

HEIGHT Of f AN ABOVE HEIGHT Of FAN ABOVE
SURf ACE, 7 IN. SURF ACE. 410. IN.
60
40
PRESSURE.
-II- UI-
IN. HQ. GAGE
20
ELECTRIC e lECTRIC
FAN ON ....., FAN ON
I MIN 9 MIN
1IME-
Figure 7. Pressure histories including

time with vapor mixing
-347-
2) Heat conduction downward from the uninsulated
steel lid along the inner shell and also through
the quiescent vapor, which brought heat directly
to the surface.
The relative influence of heat conduction through the vacuum

jacket to the liquid or down from the lid directly to the
surface may be demonstrated with this container by changing
the quantity of liquid. With the same pressure in the vacuum
jacket, heat conduction through the inner shell per pound of
liquid would remain about the same; but with a greater liquid
content, the liquid surface is brought closer to the uninsulated
lid. The shorter heat path from the lid to the surface, both
down the inner shell and through the vapor, should result in
greater relative heat flow to the surface. The effect of this
increase in relative heat flow to the surface would be to
increase the surface temperature and consequently increase
the pres sure.
This higher pres sure was experienced, as shown by

the results presented in Figure 6. Pressure is shown as a
function of time for a pressure of less than 5 microns in the
vacuum jacket and the container either 30 or 60 per cent full
A higher pressure was obtained with the greater liquid content
at any equal interval of time. Increasing the pressure in the
vacuum jacket to about 200 microns resulted in an increase in
pressure for the same time interval because of an increase
in heat conduction per pound of liquid through the inner shell;
but a further increase in pressure was obtained when the
liquid content was increased from 30 to 60 per cent. There-
fore, in order to minimize the pressure developed without
stirring the liquid, consideration should be given not only
to reducing the heat-leak rate but also to minimizing the
direct heat flow to the surface. For instance, with this
container the direct heat flow to the surface could have been
decreased by insulating the top of the container.
Effect of Vapor Mixing on Pressure
Another factor that may affect the flow of heat to the

surface is the vapor -temperature gradient. (See Figure 4,
for example.) Referring to Figure I again, we see that
-348-
natural convection in the vapor causes the heated, less dense
vapor to accumulate at the top of the container. The heated
vapor, being farthest from the surface, has a beneficial
effect as long as the heat remains there, because less
conduction takes place between the cooler vapor near the
surface and the liquid layer at the surface. Altering the
vapor-temperature gradient by forced convection may result
in more heat getting to the surface, with a resultant increase
in surface temperature and a subsequent increase in pressure.
The effect on pressure of altering the vapor-temperature
gradient was demonstrated by using an electric fan to mix the
vapor. The fan was installed in a fixed position, which
required that the liquid content be changed to vary the height
of the fan above the surface.
The pressures obtained with vapor mixing for two

liquid contents are shown in Figure 7. Pressure is shown
as a function of time for a pressure in the vacuum jacket of
less than 5 microns. With this low pressure in the vacuum
jacket, the heat-leak rate was about the same for the two
liquid contents. The normal increase in pressure with time
without vapor mixing is shown for the container 70 per cent
full. A substantial increase in pressure occurred during the
one minute interval when the fan was operating. Stopping the
fan caused a slight decrease in pres sure, but the pressure
then remained steady for about an 8 minute interval, with no
indication of a further decrease. The fan was 7 inches above
the surface in this case. Increasing the height of the fan
above the surface to 44 inches by decreasing the liquid content
to 30 per cent full resulted in a rate of increase in pressure
with mixing about four times that without mixing. The
difference in the pressures resulting from the two cases of
vapor mixing is attributed to the difference in movement of
the vapor above the surface. These results show that, in
order to maintain minimum pressure in a closed cryogenic
container, even a momentary increase in direct heat flow to
the stable liquid layer at the surface should be avoided.
Summary
This experimental study has verified the existence

of liquid-temperature gradients in a closed cryogenic container
with liquid nitrogen. The gradients became more severe with
time and accounted for the increase in pressure greater than
that calculated from the vapor pressure-temperature
-349-
equilibrium relation based on the average liquid temperature.
Stirring the liquid to equalize liquid temperatures caused a
reduction in pressure, which would extend the time required
to reach a limiting pressure. If the design of a closed
container will not accommodate stirring the liquid, a lower
pressure can be achieved with a given heat-leak rate by
minimizing the direct heat flow to the surface. Even a
momentary increase in heat flow dir ectly to the stable liquid
layer at the surface should be avoided to achieve minimum
pressure in a closed container.
Discussion
Question by R. L. Hauser, The Martin Company: How do

you calculate the average temperature of the quiescent liquid?
Answer by author: The average liquid temperature for any

given time was calculated from the measured liquid tempera-
ture after stirring the liquid to equalize the temperature and
the assumption of a linear increase in average temperature
with time. The average temperature calculated in this
manner was found to be in agreement with a. mass-averaged
temperature calculated from the measured temperature
distribution in the quiescent liquid.
Question by D. B. Crawford, Air Products, Inc.: Is there

any explanation for the very rapid rise in pressure occurring
immediately after closing the dewar and starting the test?
Answer by author: In the vented dewar evaporative cooling

was experienced at the surface with the normal flow of heat
to the surface. Closing the dewar decreased the rate of
evaporization. The combined effect of a decrease in
evaporative cooling and the normal flow of heat to the surface
is believed to account for the rapid pressure rise which
occurred immediately after closing the dewar.
Question by J. F. DiStefano, Pesco Products, Division of

Borg-Warner: Commenting on the effect of rapid pressure
rise when the cover was placed on the NACA dewar during
stratification tests it should be noted that the slope of
pressure rise versus time immediately after placing lid on
was similar to the effect of convection as evidenced by motor
-350-
fan effects. Doesn't this then indicate that convection is the
major contributing factor in this phenomenon?
Answer by author: Convection was certainly a major factor

in transporting heat from the liquid to the surface and, by
forced means, in transporting additional heat from the vapor
to the surface. However, the decrease in mass rate of
evaporization is believed to be the major contributing factor
to the rapid pres sure rise which occurred immediately after
closing the dewar.
Question by L. Garwin, Kerr-McGee Oil Industries: Would

you venture an opinion on what results you would have
obtained had there been no heat leak into the system from
the top?
Answer by author: Similar results would be expected with no

heat leak into the system from the top, except that the
increase in surface temperature with time would be lower
because of less direct heat flow to the surface.
Question by G. H. Caine, Pesco Products, Division of

Borg- Warner: Did you find a higher temperature of liquid
at the bottom of the dewar with the liquid vented to the
atmosphere, i. e., a reverse temperature gradient, than with
the liquid not vented?
Answer by author: Yes, the liquid below the surface was

warmer than the surface when the dewar was vented to the
atmosphere. For instance, with the dewar vented, for the
conditions of Figure 4, the liquid below the surface was
O.3 0 F warmer than the surface. The change in temperature
occurred very near to the surface.
Question by L. R. Stoecker, Beech Aircraft Corporation:

Have you ever entertained the idea of taking the vapor and
bubbling it up through the liquid?
Answer by author: Yes. Nitrogen gas from an external supply

has been bubbled up through liquid nitrogen in a closed con-
tainer. An initial pressure drop was experienced, but the pres-
sure soon increased due to the addition of mass to the system.
-351-
Question by W, Blackman, United Aircraft Corporation:
Would you expect the same type of temperature distribution
within the dewar if the dewar pres sure were raised above
the critical pressure of nitrogen, i. e" would the steep
gradient exist at what was formerly the liquid surface?
Answer by author: If dewar pressure were allowed to

increase above initial pressure, I would expect the contents
to approach the characteristics of a gaseous phase with no
steep temperature gradients,
Question by T. F. Kiefer, Allison Division, General Motor s

Corporation: Has a vertical conductor (copper) in the liquid
been considered to equalize the temperature in place of the
propeller?
Answer by author: Copper conductors were not considered

in the present investigation. However, copper strips were
used in a high heat-leak container. This was not successful
because of insufficient conductive area. In this high heat-
leak container application, the required conductive area would
have resulted in an intolerable weight.
Question by G. E. McIntosh, Beech Aircraft Corporation:

Were tests performed with poor enough vacuums to induce
vigorous boiling? If so, was there a decrease in the
temperature stratification effect?
Answer by author: No tests have been performed under

conditions of vigorous boiling.
-352-
G-l
CALCULATION OF GASEOUS HEAT CONDUCTION IN DEWARS
R. J. Corruccini
Boulder, Colorado
Introduction
The use of high vacuum as an element of thermal

insulation is familiar through the example of the cryogenic
vessel. In such devices it is not always practical to achieve
such a high vacuum as to make the gaseous conduction of
heat negligible compared to radiation. Indeed, when both
boundaries are at a low temperature -as in a vessel for
liquid hydrogen or helium when surrounded by liquid nitrogen
or liquid air-the gaseous conduction may predominate.
Consequently the calculation of such conduction is a necessary
part of the engineering design of vacuum insulations.
The conduction of heat by gases is commonly

separated into two distinct cases, namely, L « d and
L »d, where L is the mean free path in the gas and d is
the distance separating the boundaries between which the gas
is transporting heat. The first of these is found at ordinary
pressures and involves the thermal conductivity coefficient of
the gas. In this region of pressure the transport of heat by
the gas depends on the molecular collision rate and the mean
energy transfer per collision. An increase in pressure
increases the former but decreases the latter, inasmuch as
more collisions then are involved in transporting unit energy
from the hot boundary to the cold boundary. These effects
nearly cancel and hence the heat flow and conductivity are
both nearly independent of pressure.
The second case, free-molecule conduction, is found

at low pressures. Here intermolecular collisions are
infrequent compared to collisions with boundaries. The flow
of heat is smaller and depends on the number of carriers
as before. However, the mean energy transfer per collision
does not depend on pressure in this case since each molecule
conveys energy directly from one boundary to the other
-353-
without intervening collisions. Hence the heat flow is propor-
tional to the pressure. It is not related to the thermal
conductivity coefficient since the latter property is meaningful
only for the gas in bulk. Free-molecule conduction usually
prevails in high vacuum insulation, though occasionally the
vacuum will be sufficiently poor that an intermediate condition
prevails.
The formulas for calculating free-molecule heat

conduction are well known. However, the boundary tempera-
tures are usually widely different in cryogenic insulations.
This condition favors violation of some of the assumptions
involved in these formulas, namely, that the distribution of
molecular velocities following collision with a given boundary
is Maxwellian and that the specific heat of the gas is constant.
Because of these departures the conduction formulas

will be re-examined and practical procedures for applying
them in engineering calculations will be pointed out.
The Heat Conduction Equation
In textbooks, the derivation of the equation for free-

molecule heat conduction is often carried out only for the
simplified case of a surface exchanging energy with an
equilibrium gas at a different temperature. Such a result is
applicable, for example, to a heated wire in a bulb of such
large diameter relative to the wire that the gas molecules
come to effective equilibrium with the bulb through repeated
collisions with it before each collision with the wire. How-
ever, in vacuum insulations, the two principal surfaces
between which the gas is transporting energy are of
comparable area and hence it is not possible for the gas to
come to full equilibrium with either surface by repeated
collisions. A satisfactory derivation of an equation for this
more general case has been given by Kennard l and will be
used as a basis for this discussion. The result can be put
in a form that is applicable to concentric spheres, coaxial
cylinders, or parallel planes and is as follows,
1
E. H. Kennard, Kinetic Theory of Gases, McGraw-
Hill Book Company, Inc., New York, New York (1938).
-354-
H ::: (1)
a :::
(2)
where H is net energy transfer per unit time per unit

area of inner surface
'( is C p / C v ' the specific heat ratio of the gas,

assumed constant
R is the molar gas constant
p is the pressure
M is the molecular weight of the gas
T is the absolute temperature
a is the overall accommodation coefficient
A is the area
and subscripts land 2 refer to the inner and outer surfaces.

respectively. The constant, (R/8rr) 1 /2. has the value 0.2426
for H in watt/ cm 2 • p in mm Hg. and T in oK.
Meaning of Pressure and Temperature
In the literature. T in the denominator of equation

(1) is sometimes left undefined and is sometimes incorrectly
taken to be the average temperature. In the authoritative
2
treatise by Dushman it is identified in one place as the
temperature of the colder boundary and in another as the
temperature of the outer boundary (for cylindrical configura-
tion). These definitions are both incorrect in addition to
being sometimes in conflict. The derivation of equation (1)
makes it clear that T is an effective temperature of the
2S. Dushman. Vacuum Technique. Wiley-Chapman

and Hall. New York, New York (1949).
-355-
non-equilibrium gas which has a value intermediate between
T 1 and T Z but is not the average. The meaning of p also
needs to be examined, since, under the conditions for free-
molecule heat conduction, the thermal transpiration effect
operates and the local pressure varies with local temperatYJ;e
throughout the evacuated space according to the law, p/(T) 1!:..
= constant. In the derivation of equation (1), p is as so cia ted
with T and since p/(T)l/ Z appears identically as a factor in
equation (1), it is only necessary that values of p and T
appropriate to the same point in the vacuum system are
inserted into the formula. In principle p could be measured
by a suitable gage located between the walls where the
effective gas temperature is indeed T. In practice, however,
a value of p appropriate to an external gage at ambient
temperature is used. If the condition, L» d, also applies
to the passage connecting the gage to the vacuum space
(where d is now the passage width), then the temperature at
the pres sure gage should be inserted for T making it
unnecessary to evaluate the effective temperature in the
region of heat flow. This will be true even when the gage
temperature lies outside the limits T1 and T Z '
Variable Specific Heat Ratio of the Gas
Over the large differences of temperature that are

frequently involved in cryogenic insulations, '( for some gases
may vary appreciably. For example, for hydrogen it varies
from 1. 63 to 1. 40 between 80 0 and 300 0 K.
From the derivation of equation (1) it is clear that

an average value of '( given by C IC should be used
p v
where C is defined as ST Z C pdT I (T Z - T 1) etc. This is
p
T1
not identical with Sr

T
Z '( dT/(TZ-T 1)' though it is nearly the
1
same for any practical case.
Mixtures of Gases
Since by definition of the free-molecule mode of

conduction the gas molecules interact predominantly with the
walls rather than with each other, the conduction by a
-356-
mixture is the sum of the conductions calculated for each
component by equation (1) using for p the respective partial
pressures.
Transition Region of Pressure
Unfortunately there is no rigorous formula for heat

conduction that is valid at all pressures. The calculation
may be extended a small way from the region where the
ordinary thermal conductivity prevails toward lower pressures
by introduction of the temperature discontinuity coefficient.
but the entrance of the intermediate pressure region from
the low pressure side where free-molecule conduction pre-
vails has not been analyzed. Two apparent exceptions have
been noted in the literature:
The first is an equation due to Kistler 3 in a form

which appears to bridge the transition region and which has
been used to calculate heat transfer in vacuum-insulated
vessels. 4. 5 This equation can be written as follows
k = d
] (3)
)L
where d and L are in cm •• T is in oK. p is in mm Hg. and

k the apparent thermal conductivity is in watt I cmoK or
call sec cm oK according to whether C v is expressed in
1/2
jl g oK or call g oK. The factor .058 is II (21T R) • The
quantity QO is the accommodation coefficient. assumed to be
the same for both walls. (Q O = Q. = a. 2 ). Kistler derived
3S • S. Kistler. J. Phys. Chern. 39. 79 (1935).
4 J . F. White. Chern. Eng. Prog. 44. 647 (1948).
5 B • W. Birmingham. E. H. Brown. C. R. Class.

and A. F. Schmidt. J. Res. NBS. 58. 243 (1957).
-357-
this formula for a random porous structure having mean pore
size d and applied it to the flow of heat in silica aerogel.
Use of the formula for calculating conduction acros s vacuum
insulation has been justified on the basis that the formula
should still hold for a uniform structure characterized by a
boundary spacing equal to the aforementioned parameter d.
Accepting this assumption, we can test the validity of the
formula by seeing if it reduces to accepted forms at high and
low pressures.
At high pressures (L « d) the bracketed factor is

unity and the equation should correspond to the kinetic theory
expression for the ordinary thermal conductivity,
( 4)
in which £ is a numerical factor, and 1'] is the viscosity.

Equation (3) can be put in the form of equation (4) by using
well known relations as follows,
.058 = 1/(2'lTR)1/2 (5a)
L = 1'] /0. 499 pum (5b)
p = Mp/RT ( 5c)
u = (8RT/'lTM)1/2 (5d)
m
where P is the density and u is the mean velocity. The

m
result is
k = 1']C /(4x 0.499) ( 6)

v
It is smaller than equation (4) by the factor 0.499 (4£) which

for ordinary gases (£ ;; 2) amounts to about four.
Similarly we can see if Kistler's equation reduces to

equation (1) at low pressures (L »d). Parallel configuration
is assumed and the following relations are used:
-358-
H = 1< (T 2 -T 1) /d ( 7)
cv = R/M ('{ - 1) (8a)
Q (8b)
The result is then seen to have the same form as equation

(1) except that it is smaller by a factor ('{+ 1)/ 2. This
discrepancy is readily traceable to the fact that Kistler in
effect takes C v T as a measure of the heat content per gram
of gas molecules effusing from an equilibrium body of gas
at temperature T whereas the correct expression can be
shown by Kennard to be (C v + R/2M) T.
Thus equation (3) does not reduce to accepted

formulas in the limits of high or low pres sures. Inspection
shows no reason to suppose that it is accurate at intermediate
pressures in the present application.
Dushrnan, on the other hand, has given equations for

the transition region utilizing the temperature discontinuity
coefficient symbolized by g. This discontinuity appears at
the boundary of a body of gas across which a temperature
gradient is impressed as the pressure is lowered from the
region of ordinary conduction into the transition region. The
coefficient, g. is a length of the order of magnitude of L.
The formulas for ordinary conduction are modified for this
effect by imagining each boundary to be moved back by the
distance, g (Figure 1). Thus for parallel configuration
H = (9a)
and for cylindrical configuration
k(T2 -T1)
H = = (9b)
Provided that g 1 < < r l' g2 < < r 2' (and the above equation
is meaningless without this restriction) then for cylinders
-359-
BOUNDARY BOUNDARY
#2 #1
T2 ......
............
...... ...
)'~
~A 4f»~ DISTANCE
~ ~,q,., >-
.-.- (In Direction
«
0::
u~
~ of Heat Flow)
lLJ
Q.
::E
lLJ
I-
",,,
,,
TI
I
~g2 >1< d >I<gl~
Figure 1. Representation of the temp-

erature discontinuity coefficients
-360-
H
N k(TZ - T 1)
r (10)
rl (In -=-z... + ~ + ~
rl rl rZ
Dushman shows that if gl = gz' equation (9a) reduces
to the Knudsen formula in the low pres sure limit. However,

the as sumption of equality of conditions at the two boundarie s
limits his result to cases where T Z - T 1 is small. Equation
(10) would not be expected to apply in the low pressure limit
and in fact does not reduce to equation (1). Duslunan's
success in fitting the formula to data by Knudsen covering a
wide range of pressures should therefore be regarded as a
fortunate coincidence due in part to the fact that the data
cover only a narrow temperature range. The expression
yields an incorrect value of a l'
In his derivation of an expression for g, Kennard

makes it clear that the concept of the temperature discon-
tinuity coefficient is applicable only near the high pressure
end of the transition region. Nevertheless expressions
having the form of equations (9) or (10) but regarded as
empirical, may be useful for approximate calculations in the
transition region. For example, we may insert in equation
(9a) Kennard's expression for g, which with the aid of
equations (4), (5c), (5d), and (8a) we will put in the
following form,
(11)
i = I, Z
Now if k 1 ,
1
and T
'I
1
*
are taken to be suitably defined
averages of these properties in the region of discontinuity

next to boundary 1 and similarly for the quantities in the
analogous expres sion for g~, then equation (9a) becomes
identical with equation (1) III the low pressure limit where
the "slip" regions merge. The expression thus is of
reasonable form though its accuracy in the transition region
will be que stionable. If T Z and T 1 are greatly different, the
values of k. and 'I. will depend on T ."~ and all three must be
1 1 1
found by trial.
-361-
The representation of the transition region by equation
(9a) is equivalent to multiplying k by the factor d/ (d + gl + gZ)
which for diatomic gases, small temperature difference, and
unit accommodation coefficient is approximately d/ (d + 4L).
By contrast, Kistler's equation applied to the same conditions
is equivalent to multiplying a value of K that is four times too
small by a transition factor d/ (d + L).
Mean Free Paths
To facilitate determination of the mode of conduction,

selected values of mean free path are given in the following
table.
TABLE I
Mean Free Path

L in cm at a Eressure of 1 micron
40 K 20 0 K 76°K 300 0 K
Air 0.87 5.1
Hydrogen 0.30 1.8 9.5
Helium 0.11 0.67 3.2 15.
The mean free path is conveniently obtained from tabulated

values of viscosity6 using the formula,
T )1/2
L = 8.6 --'!L ( - ( 12)
p M
where L is in cm. T is in deg K, T] is in poise sand p is
in mm Hg.
6 NBS Circular 564 (1955).
- 362-
Experimental Values of Accommodation Coefficients
The definition of accommodation coefficient has been

a subject of controversy. For engineering purposes it is
sufficient to consider it as an empirical factor needed to
bring conduction as computed by equation (1) and (2) into
agreement with experimental values. It is determined using
configurations such that Al / A2 < < 1 and hence u ~ ul which
permits isolation of the value for the inner of the two boundary
surfaces. Accommodation coefficients for light gases are of
special interest as they can have rather low values. Experi-
mental studies have shown two types of behavior.
For metallic surfaces that have been flashed in

vacuum to remove adsorbed films. the initial accommodation
coefficient can be strikingly small-only a few per cent for
helium on tungsten 7 -and it decreases with decrease in
temperature, provided the temperature is not so low that an
adsorbed film of the gas or a condensed layer forms imme-
diately. The interaction between a gas molecule and a clean
solid surface has been treat ed quantum mechanically. S
On the other hand, surfaces that have received no

special treatment to remove films show larger values that
increase toward unity as the temperature is lowered.
It will not escape notice that the first type of

behavior is ideally suited to cryogenic insulation since the
insulation would then be better and would improve (up to a
point) in the temperature direction in which the cost of
refrigeration increases. However, this behavior has not yet
been realized with other than filaments, a.nd the second type
of behavior must be regarded as typical of the unflashed
surfaces of large dewars. The suggested values in the
following table were derived from the meager low temperature
7J. K. Roberts, Proc. Roy. Soc. ~, 195 (1932).
SH. S. W. Massey and E. H. S. Burhop, Electronic

and Ionic Impact Phenomena, Oxford Press, England (1952).
-363-
data of the second type. 9, 10, 11, 12, 13
It is interesting to note that the low values for the

light gases at temperatures above 80 0 K counteract the effect
of the factor 1 /(M)1/2 in equation (1) and thus result in
conduction of the same magnitude as by air at the same
pressures. This is in marked contrast to the case at
atmospheric pressure where the conductivities, k, of
hydrogen and helium are about an order of magnitude greater
than that of air.
As Knudsen showed, 0.1 is increased by surface

roughness, due to the increased probability of multiple
collisions. Rowley and Bonhoeffer 13 found the interesting
result in the region around 150 0 K that a 1 for parahydrogen
is 10 to 15 per cent smaller than for normal hydrogen. This
was attributed to the accommodation coefficient for rotational
energy (which is a larger, part of the total energy in the para
form) being less than the accommodation coefficient for
translational energy.
TABLE II
Accomodation Coefficient a 1
T, oK Helium Hydrogen Air
300 0.3 0.3 0.8-0.9

80 0.4 0.5 1
20 0.6 1
4 (1)
9 W • H. Keeson and G. Schmidt, Physica~, 590,

1085 (1936); 4, 828 (1937).
10 M . Knudsen, Ann. d. Phys. 34, 593 (1911).

I1F. Soddy and A. J. Berry, Proc, Roy. Soc., A84
576 (1911).
12 B • Raines, Phys. Rev. ~, 691 (1939).
13H. H. Rowley and K. F. Bonhoeffer, Z. Phys.
Chern., B2l , 84 (1933).
-364-
The experimental values have nearly all been
obtained at gas pressures of 10 microns or more, since, in
working with filaments, free-molecule conduction predominates
when the mean free path exceeds the filament diameter, and
the latter is quite small. On the other hand, the pressures
required in vacuum insulation on an ordinary scale are usually
0.1 micron or less. The pressure variation of 0.1 for various
gases on platinum near ambient temperature has been
measured by.Alndur, Jones, and Pearlman 14 for pressures
in the range 10 to 500 microns. The values are comparable
to those in the above table but show small increases with
increasing pressure. .Alndur l5 has shown that this variation
can be accounted for on the assumption that the accomodation
coefficient is a linear function of the fraction of surface
covered by adsorbed gas, the latter being given by Langmuir's
adsorption isotherm. If his expression is used to extrapolate
the accommodation coefficient to pressures of 0.1 micron or
less, values of only a few per cent are obtained, approaching
the values mentioned previously for flashed surfaces. How-
ever, unpublished work at this laboratory at pressures down
to 0.1 micron does not show such a decrease but rather
indicates values conforming approximately to those in the
table.
Working Formulas
We can put equation (1) in the following form
H = (constant) 0. p(T 2 -T 1 ) (13)
where the constant is (~) ( R )1/2 and is listed

'Y - 1 8u MT
below for various gases using T =
300 0 K as the assumed
temperature at the pressure gage, H in watts/cm 2 of inner
surface, p in mm Hg, and temperature in oK.
141• .Alndur, M. M. Jones, and H. Pearlman, J.

Chern. Phys. g, 159 (1944).
15
1. .Alndur, J. Chern. Phys. 14, 339 (1946).
-365-
TABLE III
Gas T Z and T 1 in oK Constant
NZ ~400 .0159
°z ~ 300 .0149
HZ 300 and 77 or 300 and 90 .05Z8
HZ 77 and ZO .0398
He any .OZ80
In a longer form of this paper which will appear

elsewhere, several topics which are not now important to the
practicing engineer are discussed for the sake of completeness.
These are the definition of the effective gas temperature, the
definition of accommodation coefficient, the possibility of
observing the effect of non-Maxwellian velocity distribution,
and the effect on heat-flow calculations should the accomoda-
tion coefficient be found to be a function of the energy of the
incident gas mole cules.
Discussion
Question by R. S. Hampton, Lox Equipment Company: What

would be an average per cent of heat leak by gas conduction
as compared with radiation in the type of dewar being
discussed?
Answer by author: Considerable variation is found inasmuch

as the emissivity and residual gas pressure can both vary
widely. The following compuisons are believed to be repre-
sentative:
In a dewar for oxygen, nitrogen, or air having surface
emissivities of .03 and a gas pressure of .03 micron, the
gaseous heat conduction is about 1/10 of the radiation.
In a dewar for hydrogen or helium having a liquid nitrogen
cooled shield, emissivities of .03, and a pressure of .03 micron
of hydrogen or helium, the gaseous heat conduction from shield to
inner container is roughly ten times the radiation.
-366-
G-2
MEASUREMENT OF THERMAL STRESSES AT

-320 o F BY STRAIN GAGES
J. P. Hunsaker and A. Pastuhov

Introduction
In connection with the theoretical evaluation of an

aluminum tank for liquid nitrogen storage, it was necessary
to experimentally confirm the expected low thermal stres s
level. For an adequate investigation of these stresses 50
gaging locations were required. To our knowledge strain
gages had never been used on such a scale at this low temp-
erature level, and of even greater concern was the complete
lack of experience with strain gages to measure thermal
stresses at low temperatures.
The basic problems to be solved were:
1) To compensate for the inherent temperature

sensitivity of electric strain gages
2) To obtain the same null reading at low

temperatures as was obtained at room temper-
ature, since otherwise a zero shift could not
be differentiated from a real thermal strain
3) To arrange the strain gages in such a manner

that meaningful results are obtained whether or
not the gages are immersed in liquid nitrogen.
A search of the literature showed that the work

reported by Day1 and by Majors 2 is applicable to this problem.
Both papers indicate that conventional SR-4 type strain gages
as supplied by the Baldwin-Lima-Hamilton Corporation can be
1 E. E. Day and A. H. Sevand, Proceedings of the

Society for Experimental Stress Analysis, Vol. VIII, No.1 (1950).
2H. Majors, Proceedings of the Society for Experimen-
tal Stress Analysis, Vol. IX, No. 1 (1951).
-367-
glued satisfactorily to aluminum for use at low temperature,
that the gage output is repeatable, and that a zero shift at
low temperatures may occur. Day indicates that the gage
factor changes with temperature; Majors shows that it does
not. Majors shows that at _320 0 F the gages are extremely
temperature sensitive, indicating a false 400 psi stress for
10F temperature change in active gage when at -320 0 F
approximately, and with the compensating gage at room
temperature.
The National Bureau of Standards Cryogenic Labor-

atory' Boulder, Colorado, advised that they have found that
the gage output for Bakelite-base SR-4 gages is less subject
to variation from gage to gage at low temperature than paper
gages, that the gage factor does not appreciably change down
to 76°K, that Armstrong epoxy adhesive is still good at this
low temperature, and that the gages do experience a zero
shift at low temperature because of the inherent temperature
sensitivity of the gage.
Preliminary Experimental Work
The characteristics of various gage types and

arrangements were investigated on the simple apparatus shown
in Figure 1. Gages were applied to a 1-1/2 x 1/4 inch
aluminum bar (248) on both the upper (tension) surface and on
the lower (compression) surface. The gaging portion was
subjected to a uniform moment and was immersed in liquid
nitrogen. Each gage on the top of the bar was measured
against a gage on the lower surface (gage output twice the
true strain). With this arrangement no external compensating
gage was required, and we were certain that all gages were
at the same temperature.
The suitability of 120 ohm, A-3 paper-base gages was

first explored. Typical results are shown on Figure 2. We
concluded from this work that:
1) The gage factor does not change significantly

at -320 0 F, which confirms the work of others
2) Paper gages adhere at -320 0 F when 8R-4
nitrocelluous cement is used
-368-
ElflNrGMi£S
'Il#NSION
" COMP~S$IOAJ
Fli.i. WlrHNITR06£N
5R4· STJIlllfN 6106£

INPIC.IITOR
Figure 1. Arrangement for tests on bar s
1000
BOD
ROOM fEMRaIRy£
av/£}EPBY J./e
(12 % CNAN6E IN
I: AT -32o-F) .1
600
.~ 400
-~
"~
~
'3
~
~' 0
,2
~
t;
"~ -.
J>'/,
\;
~ ~
~
/
/ ~
-zoo
I- ~
/ / ~ -4000
/ / <!'
-~. ..,~
~
-'000
~ -IJOOO
-/0000
Figure 2, Bar test on 4 pairs of

A-3 gages in liquid nitrogen -J~~O~~~----7r~--~~-----r------r
-.300 -zoo -/00 0 100
7£MPERATVNE OF ACTIVE tiME, ~
Figure 3, Temperature
characteristics of ABD-l gages
-369-
3) The low temperature zero shift of the A-3
gage is large and unpredictable.
Next the characteristics of 120 ohm, AB-3 Bakelite-

base gages were explored. Various techniques for applying
the gages were tried. We found that Armstrong A-2 epoxy
resin adhesive thinned with allyl-glycidyl-ether for a thin glue
line worked as well as Bakelite resin baked in place and was
a practical technique to use on the vessel to be te sted.
These gages gave results similar to the A-3 paper gages with
zero shift of still as much as i 400 microinches. Connecting
four gages together in a bridge arrangement doubled the
sensitivity, reduced the zero shift due to averaging between
more gages, and eliminated the effect of variations in lead
wire lengths and thermoelectric effects. When a 20 inch
length of only one lead wire was immersed in liquid nitrogen,
the false indicated stress for the bridge arrangement was 80
psi; it was 600 psi for a two-gage arrangement. When 40 feet
of #28 gauge wire was added to each of the four leads of the
bridge arrangement, only a very small change in the meter
reading was observed (less than 20 microinches) either at
room temperature or at -320 o F.
The experiments with 120 ohm A-3 and AB-3 gages

led us to conclude that 350 ohm ABD-l gages should be used
because of their smaller sensitivity to wire lead resistances
and that the low temperature zero shift could be further
minimized by a nondestructive pre calibration or selection
technique by which those gages with a like zero shift as
compared with a standard would be used together in a bridge
of four. Our first attempts at grading ABD-l gages were
devoted to various methods of clamping 10 gages at a time
to an aluminum block, cycling the assembly from room
temperature to a lower temperature, and of course, keeping
track of the readings. All trials suffered from a wandering
error on return of the assembly to room temperature because
of creep and squirming of the parts. The final method of
preselection of all gages used was to stick one gage at a
time to a 2 x 1 x 1/8 inch coupon of aluminum with DeKatinsky
wax. The coupon was heated to about 300 o F, lightly coated
with wax, and the gage firmly pressed down. Ten gages and
coupons were then taped to a 1 inch block of aluminum, leads
were soldered to the gage wires, and finally the whole
assembly was placed in a polyethylene bag filled with silica
-370-
gel and securely taped at the opening. In this way, diffi-
culties with moisture were eliminated.
Repeated cold and warm readings were taken of the

gages versus a standard dummy and also versus gage #1
of the 10 as a check. At least three cool downs were necessary
before an intelligent selection could be made. The lowest
temperature used was about _30 o F. At lower temperatures,
the wax cracked, freeing the gage. Three readings could be
taken per day (warm, cold, and warmup). We readily
removed the gages from the wax by soaking them in acetone.
Typical results versus gage #1 are shown in Table 1.
To evaluate the success of the grading system we

cemented three bridges of four matched gages each, to
aluminum and tested them at -320 0 F. The zero shifts in
indicated stress were found to be 250, 250, and 800 psi.
We wanted to determine the temperature character-

istics of the ABD-l gages in order to correct the gage output
for those cases that would occur in the tank measurements
in which the compensating gage metal would not be at the
same temperature as the active gage metal. To obtain this
information, we ran two sets of tests. In the first, the
dummy gage (cemented to aluminum) was held at room
temperature I while the active gage was inserted in steps
down a wide mouth dewar one-quarter full of liquid nitrogen.
Thermocouples were attached to the gage blocks. The
second set of tests were devised to determine the gage
indication when both gages were cold, but with a small
temperature difference existing. To do this, we fixed the
dummy gage by a clamp and wooden stick a few inches above
the liquid level, while the active gage was moved from the
warmer region above the dummy, past it, and into the
colder region below the dummy. The results are shown on
Figures 3 and 4. The slope at -320 o F of 60 microinches/oF
of Figure 3 is about the same as that reported by Majors on
AB-14 gages.
Matched gage groups were applied to the tank under

test in the arrangement shown on Figure 5. The compen-
sating gage block, of the same material as the tank, was
-371-
JODO
~ 600
't
~"
~ 600
~ ~
\i 400
~
f
'l
ZOO
..,~ 0
+20 "'40
<:. ZOO TEhlP. OF t:/CTIVE 61/6,0 MINUS
~ TEMP. OF DUMMY 64GE, o f
~ 400
~
~ ~
~ 600
~
800
1000
Figure 4. Effect of variation of dummy

gage temperature for ABD-l gages
STRI/IN GI/GE
SELECTION
SWITCH
SR-+ STRRIN
INDICI/TOR
Figure 5. Diagram of typical gage installation
- 372-
secured to the tank shell by a single screw in order that
thermal strains in the compensating block would be eliminated.
Grease was liberally applied under the compensating block
for better heat conduction. Copper-constantan thermocouples,
connected in differential, were located near the active gages
and on the compensating block so that a correction to the
indicated strain due to temperature difference could be made.
Results of Tank Tests
All 50 of the gages, of course, did not work

perfectly. Two gages exhibited a drastic zero shift which
remained even after the tank had warmed up to room temper-
ature. However, the shift was so obvious, that a reasonably
reliable correction could be made.
The tests showed that the gages worked satisfactorily

when immersed in liquid nitrogen.
False strain indications occurred when the liquid level

passed through a gage station located on the inside of the tank.
The false indications, in the order of 3000 psi, were due to
the temperature difference between the active gage metal and
the compensating gage metal. This temperature difference
was measured by the differential thermocouples and a
correction to indicated stres s deduced. But the correction is
only approximate since the exact temperature under each gage
was surely different from that at the thermocouples.
When the tank was being filled, those gages located

on the inside of the vessel gave very erratic performance.
The meter needle fluctuated widely and the indicated stresses
were unbelievably high. No doubt this behavior was due to
drops of cold liquid hitting a gage.
In general the steady state stress measurements were

quite satisfactory. For the case of maximum stress when the
vessel is nearly empty, we are confident that, with suitable
corrections and averaging, the maximum stresses were
measured within ± 1000 psi.
- 373-
TABLE I
Typical Matching Results

Microinches per Inch
22 0 F
Gage Co1d- Warmup- Warm- Co1d- Warmup- Warm- Co1d- Warmup-
No. Warm Warm Warmup Warm Warm Warmup Warm Warm
2 +20 -5 -5 +20 0 0 +30 0
\
3 +70 -5 +25 +60 0 0 +60 -10
4 +75 -10 -15 +65 -5 0 +50 -5
5 +85 -5 +5 +80 0 0 +80 0
6 +100 -5 -20 +95 -5 0 +25 -5
7 +85 -5 -25 +115 -5 0 +185 -10
8 +80 -5 -70 +120 -5 0 +230 -10
9 +75 -5 -65 +115 -5 0 +270 -10
10 +95 0 -115 -115 -15 0 +300 -30
Acknowledgment
The advice and encouragement of W. W. Murray of

the Massachusetts Institute of Technology, A. C. Ruge of
Ruge-DeForest,Inc., F. Heinz of the Baldwin-Lima-Hamilton
Corporation, and G. Y. Robinson, Jr. of Arthur D. Little,
Inc. is gratefully acknowledged.
-374-
G-3
HEAT TRANSFER TO BOILING LIQUID NITROGEN AND

HYDROGEN FLOWING AXIALLY THROUGH
NARROW ANNULAR PASSAGES
R. J. Richards, R. F. Robbins, R. B. Jacobs,

and D. C. Holten
Boulder, Colorado
Introduction
The purpose of the investigations described here was to

obtain inforITlation concerning the ITlaxiITluITl heat dissipation of
boiling nitrogen and hydrogen flowing through narrow channels.
The sponsoring agency had two uses for this inforITlation:
1) To provide a check for the theory of Sydoriak and

Roberts. 1 In a paper by theITl, a theoretical
approach to the probleITl of heat transfer in a
frictionless flow evaporator was presented. The
resulting equation related Q (the power input to
the boiling liquid) to various properties in the
following ITlanner:
1/2
Q = A A (p LPG gf Z) ( 1)
where
A = cross-sectional area of passage
A = latent heat of boiling liquid
PL = density of liquid
PG = density of vapor boil-off
g = gravitational acceleration
f = ITlass fraction vapor leaving evaporator
Z = pressure drop across evaporator
IS. G. Sydoriak and J. R. Roberts, J. Appl. Phys.,

p. 143, February 1957.
-375-
Experimental data taken by Sydoriak and Roberts
agreed with this equation within 30 per cent for
hydrogen and 5 per cent for nitrogen.
2) To provide a check on the LASL experimental data.

The studies by Sydoriak and Roberts were not
extensive; it was felt that a thorough investigation
might provide more reliable predictions of maxi-
mum heat dissipation by boiling cryogenic liquids.
Considering practical application of this work, the only

concern here was the prediction of the maximum safe operatiq!;
power of an electromagnet cooled to 20 0 K by boiling hydrogen.
This magnet, described in the Sydoriak and Roberts' paper, has
narrow channels (the order of 0.010 inch gap width) through
which boiling hydrogen would flow. The studies here were
conducted using a test apparatus simulating an electromagnet.
Description of Apparatus
Figure 1 is a sketch of the test section used in the
heat transfer apparatus. To simulate the de sign of the magnet.
a 7 inch stainless steel plug was accurately centered in a 5
inch stainless steel tube.
This stainless steel tubing was used as the heater; its
other dimensions were 1/2 inch O. D. and 0.010 inch wall thick-
ness. The tubing was used as a resistance to current forced
through it by 12 six volt batteries connected in parallel. The
heated part of the tubing is between the bras s end pieces which
served as connections for the battery leads and also as
supports for the Styrofoam insulation on the outside of the
heater. The insulation made certain that essentially all of the
heat was transferred to the liquid in the annular gap.
Various sizes of plugs were made to fit inside the

tubing to form different size annular spaces or gaps. They were
held in position by centering pins, one stationary, the other
spring loaded to allow for contraction during cooldown. Holes
were drilled in the brass end pieces so the pressure drop across
the heated section could be measured. Thermocouples were
placed along the tubing wall and connected differentially to
thermocouples in the liquid to measure the temperature differ-
ences and temperature distribution along the heater.
-376-
.... ......
TO"
PlO P1(Cl
TO
I/OU -
WT"
~_JJ
IOn...
COfTV."..
PIN
_AT
TO "TTon'
TMItU
Figure 2. Water-
cooled rheostat
Figure 1.
Figure 3. Test section Figure 4. Plugs
-377-
A water cooled rheostat (Figure 2) was used to control
the voltage to the heater. The rheostat is made of stainless
steel tubing, one pair of 0.020 inch wall thickness (coarse
adjustment) and the other pair 0.006 inch wall (fine adjustment).
The two rectangular pieces at the bottom of the tubes contained
woven copper wire between two pieces of aluminum bolted
together. As these pieces were moved upward the copper wire
shorted across the two stainless tubes, thereby increasing the
power input to the heater.
Figure 3 is a photograph of the test section with the

insulation and thermocouples removed. The plug and spring
loaded centering pin can be seen taped half way out of the heater.
The wire in the spring was used to lift the plug out. The
insulating piece of micarta can be seen on top of the plug.
Figure 4 shows the micarta plug which was used first

and the stainless steel plug with its insulating ends. These
will be discus sed later.
Discussion of Experiments
Forced Flow. Data were taken while varying the

volume flow rate through the heater using both liquid hydrogen and
liquid nitrogen as the coolant. The hydrogen runs were of
greater importance, since magnet design was considered along
with the data comparisons and theory correlations.
The data were taken as follows: Liquid was transferred

from a 50 liter storage dewar to the test section through a
vacuum insulated transfer line. The flow was generated by
pressurizing the storage dewar and was calculated from the
pressure drop across an orifice plate (see Appendix B). After a
stable high flow through the te st section was established, power
was added to the heater; the flow rate was then decreased until the
temperature difference between the liquid and the heater rose to
lOoK. At this point temperature distributions, pressures, and
pressure drops as well as power input and flow rate, were
recorded. The lOOK ~T was arbitrarily established by Sydoriak
and Roberts and will be called "breakaway" hereafter. This point
was not necessarily the point at which film boiling began; it was,
however, regarded as a limit to which the temperature of a magnet
could rise and still maintain its low resistivity properties.
-378-
Natural Convection. The nearest agreement between
previous work and the experiments here was obtained when the
test section was immersed in a pool of nitrogen or hydrogen. In
these tests, called natural convection (no forced flow through
the heater), the power input was increased until the temperature
of the heater began to rise. In most cases the AT could not be
held stable at lOoK; it would instead rise sharply until the power
was lessened. The film boiling phenomena observed by others 2 , 3
seems to explain the sudden rise in temperature. Since there
was not a primary interest in film bOiling, no data were taken in
the film boiling region.
During natural convection runs a flow through the test

section was induced by bubble departure from the heater wall.
This flow was difficult to determine; hence, the natural convec-
tion data could not be compared with Sydoriak and Roberts l
theory.
Zero Power Te sts. During the preliminary tests using

nitrogen as the coolant, a marked variation in maximum power
input at a given flow rate was observed. Since the liquid entering
the test section was very close to its boiling point, one
pos sible explanation for the inconsistency was flashing due to the
pressure drop across the heater. To check this possibility,
pressure drop data were taken with zero power input to the
heater. Results of these tests showed that vapor formation
could occur unless the system was pressurized, essentially
sub-cooling the entering liquid. Hence all hydrogen heat
transfer data were recorded while pressurizing the test dewar
to at least 4 inches Hg gage.
Results
The Equation. Sydoriak and Roberts l theoretical

1/2
equation Q = A X. (p L PG gfZ) was found to be, at best, of
the order of magnitude of the data reported here. To calculate Q
2 N . R. Leroy and J. A. Roberts, UCRL-1894, Univer-

sity of California, Berkeley, California (1952).
3L. A. Bromley, Chem. Eng. Prog., 46, No.5, 221

(1950).
-379-
by this equation it was nece ssary to correlate f (the vapor
fraction) and Z (the hydrostatic head equivalent) to the experi-
mental data. In the derivation of the equation Sydoriak made
the as sumption that all the heat input was used to vaporize
liquid; hence, f was calculated from the equation
vapor produced, cclsec

f = (Za)
total flow, cclsec
Q
or f = (2b)
where V is the volume flow rate, ccl sec, and Q/PL >... is the
vapor production rate, cc of liquid

sec
The head equivalent Z was calculated from the pressure

drop, D.p, by converting the actual measurement (usually inches
of Hg) to inches of hydrogen or nitrogen. The equation used
for calculations now took the form,
Q ~p 1/2
observed
Qcalculated = C(------) ( 3)
V
where C is an accumulation of the other terms in the original

equation, all of which are constant for a given gap width.
Results of the correlations of this equation with experi-

mental data are shown in Tables I and II, Appendix A. Although
the Qobs l Q ca1c ratio is always between land 2, there is no
real agreement shown by these studies. With some gap widths
the Qobs I Q calc ratio stays fairly constant; however, this is not
always true.
The uncertainty here indicates that from the results of

the NBS tests one cannot verify Sydoriak and Roberts' equa-
tion to within 5 per cent for nitrogen and 30 per cent for hydrogen.
Forced Flow. Results of the breakaway runs, in which

power input was measured in conjunction with flow rate at a ~ T
of lOoK, are shown in Figure 5. This illustrates that the forced
flow data do not agree with the data taken by Sydoriak and Roberts.
-380-
88
80
H2 10 MIL GAP
72
64
HYDROGEN AND NITROGEN

56 HEAT TRANSFER
FLOW RATE AT BREAKAWAY (AT. looK)
VS.
48 POWER INPUT
W
I-
C2 40
FIGURE 5
3:
0
-'
u..
32
LOS ALAMOS
2' H2 22 MIL GAP\ .•...••... ·
.......................
N2 10 MIL GAP
16
~~0----~10-0----1-4-0----,-80-----22-0----2-W-----~----~3-W-----38-0-
POWER INPUT, WATTS
Figure 5.
NATURAL CONVECTION BOILING

LOS ALAMOS AND NBS DATA
FIGURE 6
'"
l-
I-
<t
3:
10'
,.:
::J
"-
;;:
a:
w
3:
0
"-
,~L,--------------------~-----,O--------------------------~,'O
GAP CROSS SECTIONAL AREA, CM 2
Figure 6,
-381-
There is no obvious explanation for this discrepancy since the
measurements have been repeated and the instrumentation
seems reliable.
One difference between the NBS and LASL studies

which may account for the difference in the re sults was in the
method of flow measurement. NBS flows were computed from
the pressure drop across an orifice which had been calibrated
and tested to insure accurate measurement (see Appendix B).
Sydoriak and Roberts measured flow rate by the rate of
liquid buildup in the test dewar.
Another difference lies in the type plug used to form

the annular gap in the test section. In the initial NBS tests
a micarta plug was used with spacer s to form the gap, as did
Sydoriak and Robe rts. However, it was found that the
micarta would shrink enough at the ends to cause the plug to
become loose. Therefore it was decided to insert stainless
plugs centered and spring loaded for greater accuracy in gap
control. No appreciable difference between maximum power
input data with the two types of plug was noticed; consequently,
this does not seem to be the reason for the differences.
Natural Convection. Results of the natural convection

tests are shown in Figure 6. The agreement between the
NBS and the LASL data is much better than that obtained in
the forced flow experiments, although the range of values is
somewhat different. This agreement indicates that determina-
tion of the breakaway point is satisfactory.
Conclusions and Acknowledgements
The experiments here indicate that a prediction of the

maximum operating power of an electromagnet by theoretical
analysis is not reliable at this time. The help of H. L.
Laquer, E. F. Hammel, S. G. Sydoriak, and J. R. Roberts,
all of LASL, is gratefully acknowledged.
-382-
Appendix A
The Sydoriak and Roberts theoretical equation values

and dimensions for hydrogen heat transfer
= latent heat = 446 watts-sec

gram
= density of liquid = 0.0709 gram
cc
= density of vapor = 2.6 x 10- 4 gram

cc
g = gravitational acceleration = 980 cm

sec 2
Q
f = mass fraction vapor =-.....;,....- = .0316 Q
VPL A. V
(dimensionless)
Q in watts, V in cc/sec
Z = hydrostatic head equivalent = 488 ~ p cm
~ p in inches of Hg
A = cross-sectional area of the annular passage
A (7 mil gap) = 0.0664 cm 2
A (10 mil gap) = 0.0963 cm 2
2
A (20 mil gap) = 0.1885 cm
1/2
Q calculated = 162 A ( (~p) (Qobserved)
V
2
Q in watts, A in cm , ~p In inches Hg, V in cc/sec
-383-
TABLE I
Correlation Between Experimental Data

and Theory for Hydrogen
Gap ~p V Qobs Q cal Qobs/Qcal

inches Hg cc/sec watts
7 2 7.4 73 69.8 1.045

7 3.5 13.2 94 71. 0 1. 325
7 5.3 18 106 87.7 1. 210
7 7.2 23.2 116 94.3 1.230
7 1 3.2 52 144 95 1.516
7 1 5.3 59 146 96.9 1. 505
7 21 76 159 104 1. 530
10 1.2 14 82 60.2 1.360

10 2.0 18 93 73.3 1. 269
10 2.8 33 114 70.8 1. 610
10 4.7 43 124 84 1.476
10 6.0 59 133 83.8 1.588
10 8.2 88 15.0 84.2 1. 782
20 0.6 25 122 76 1.606

20 0.8 32 144 84.2 1.701
20 1.4 41 166 107 1. 550
20 2.0 56 198 122.5 1. 616
20 3.0 70 218 136 1. 603
20 3.7 88 246 144 1. 708
-384-
TABLE II
Correlation Between Experimental Data

and Theory for Nitrogen
Q obs
Gap ~p V Qcal Qobs/Qcal
inches Hg cc/sec watts
7 9.0 4.2 180 143 1. 26

7 10.5 4.8 190 153 1. 25
7 12 5.4 200 153 1. 31
7 13.5 6.2 210 156 1. 34
7 15 7.2 220 155 2.41
7 16 8.25 230 153 1. 50
7 17 9.5 240 151 1. 59
7 18 10.9 250 148 1. 69
7 19 12.6 260 126 2.06
10 5.78 7.0 182 130 1. 398

10 8.54 9.0 216 151 1.430
10 8.38 10.2 234 147 1. 592
10 12.50 15.5 250 150 1.703
20 1.6 6.6 230 154 1. 49

20 2.4 8.7 260 174 1. 49
20 3.2 10.7 290 193 1. 49
20 3.6 13.3 320 193 1. 66
20 4.2 16.2 350 197 1. 78
20 4.8 17.3 380 212 1. 79
-385-
Appendix B
A Correlation of Orifice Flow Theory and Experill1ental

Data Using Cryogenic Liquids
According to elell1entary orifice flow theory, orifice

flow versus pressure drop data taken with a given fluid can be
used to calibrate the orifice for use with any fluid, sill1ply by
using the following proportion, 1
1/2
V (p )
A B
(1)
where V, p and ll.p are volUIne flow rate, density, and

orifice pressure drop, respectively.
During our recent heat transfer work the following

question arose: Is equation (1) applicable if an orifice
calibrated with water is used with liquids such as nitrogen and
hydrogen near their boiling points? In the ensuing investigation,
in which we actually calibrated the orifice with nitrogen and
hydrogen by transferring through the orifice to a receiver
dewar, the following inforll1ation, shown in Figure 7, was
established.
If the difference between the storage pressure and the

liquid vapor pressure is greater than the orifice pressure drop,
the calculated flow curves ll1ay be used. If not, the curves ll1ay
be inaccurate, since vapor ll1ay be formed due to the orifice
pressure drop alone.
Figure 8 illustrates the orifice used in this experiment.

The orifice is mounted on the end of a vacuum jacketed transfer
line; Figure 8 gives all the vital dimensions. Since the upstream
side of the orifice is open to the supply, it is not necessary to keep
the pres sure tap close to the orifice plate. The screen pro-
tects our heat transfer apparatus from solid parti.cles (ice,
rubber tubing, etc.) which may accumulate in the bottom of
low temperature storage dewars.
1 J. H. Perry, Chemical Engineers Handbook,

McGraw-Hill Book Company, New York, New York (1950).
-386-
Figure 7. Pressure history through an orifice
TRANSFER LINE,
114'00 ""'
t FLOW
.....-VACUUM
/ JACKET
//~:~~UDRE
/'" TAP LINES
188"10-
~.~ I~I~
5rtARP-
105"
",
EDGED
ORIFICE
PLATE
,,, :,, "
,,
SCREEN~'
FILTER I I
STORAGE
DEWAR --
+8~<tSH : I
, :
l .......L..l
Figure 8. Orifice assembly
':f:::= CALIBRATION OF .105' ORIFICE. ~ '.556. WITH ",0.

r-- AND CALCULATED H2 ANO N2 LINES.
I; ~ POINTS REPRESEN;
1
.~:~~~IMENTAL Nr AND H DATA
'0 '1--+-~I++~~~-4-T~~
v
.
I
~
'0
,; ~
~
~
// ~
~
N, ,
~ '.06.0r--··
~ 1---- , :
g'OV I ;
1
'0 ....t::
~t::-, '
,o~ilil
01 Ql 04 Q6 011 10
AP ACROSS ORIFICE, Inc;I'I" of HZO
Figure 9.
-387-
Our orifice was first calibrated by measuring the
time required for a given weight of water to transfer through
the line with a measured pressure drop across the orifice.
Figure 9 shows the water line drawn through points ranging
from 6.pl s of 1. 5 inche s water to 80 inches water. These
points are omitted to simplify the graph.
Then similar calibration runs were performed using

nitrogen and hydrogen as the transferring liquids. The
receiver dewar pressure was maintained with helium gas or
hydrogen gas and was measured during all runs. It was
found that if the receiver dewar was open to atmosphere many
of the points were considerably higher or lower than the lines
calculated from our water data (see Figure 9).
One possible explanation for this inconsistency at low

receiver pressures is that vapor is formed due to the orifice
pressure drop. Figure 7 shows how the pressure may vary from
the storage dewar to the receiver dewar. This figure shows
that if PI - P v is less than 6.p, the orifice pressure drop,
vapor may be formed. If PI - P v is greater than 6.p there is
no vapor formed since the liquid always stays at pressures
above its vapor pressure. It should be noted that if there is
much heat leak P v may rise, and consequently PI and
Preceiver must rise also. This was not the case in our
vacuum insulated system.
For our purposes 80 inches water, or 2.9 psig, was

a maximum orifice pressure drop. Therefore a 3 psig minimum
receiver pressure was enough to eliminate all vapor formation.
This experiment was a short sidelight to our primary

research and is far from complete; however, it does show that
care must be taken when flow measurement is attempted by
an orifice if the liquid being transferred is near its boiling
point.
-388-
Discussion
Question by H. Hoadley, United Aircraft Corporation: Was any

attempt made to use a nicely shaped inlet to the test chamber?
Also were these inlets the same for the two tests?
Answer by author: The inlets in the two tests were the same;
they were so small in comparison with the space immediately
before the plug that we were sure of no irregularities due to
entrance effects. The plug also extended one inch past the
test section to assure uniform entrance.
Question by J. T. Selldorff, A. D. Little, Inc.: Did you

establish the peak in the a T versus Q/ A curve for liquid
nitrogen or hydrogen?
Answer by author: Our investigations were pointed toward Q
versus V data at a given aT of lOoK; hence the variation of
aT with Q was not thoroughly studied. However, a peak was
noticed, usually in the range of 8 to 9 0 K, when our readings
at lOoK were close to the film boiling region. The Q at this
point, of course, varied with flow rate.
Question by D. White, Cooper-Bessemer Corporation: Were

the surfaces through which the heat was transmitted identical
to that of the Los Alamos tests?
Answer by author: The surfaces were of the same material,

but not identical; this could possibly cause a difference in
nucleation rate and only future investigations will prove whether
this has any significance.
Question by D. White, Cooper-Bessemer Corporation: Was

the purity of the hydrogen the same? Since the major portion
of the heat is dissipated by bubbling, the rate of bubble
function would be governed by the ratio of nucleation which
depends on the above two factor s.
Answer by author: The hydrogen was produced at the CEL

National Bureau of Standards for both tests. Purity was close
enough to assume that the rate of nucleation was not changed
by this factor.
-389-
G-4
FRICTION STUDIES OF VARIOUS MATERIALS

IN LIQUID NITROGEN*
D. W. Wisander, W. F. Hady, and R. L. Johnson

National Advisory Committee for Aeronautics
Lewis Flight Propulsion Laboratory
Cleveland, Ohio
Introduction
Friction, wear, and surface-failure properties of
materials sliding together in cryogenic liquid are vitally
important to the development of efficient pumps. Bearings and
shaft seals in particular have critical sliding surfaces and are
used in all types of pumps. Positive displacement pumps may
include additional parts such as vanes, pistons, or gears with
surfaces in sliding contact.
The physical and chemical properties of most

cryogenic liquids of interest are such that they might be
expected to have very poor lubricating properties. Some
available data suggest that conventional bearings and seals may
operate without difficulty in liquefied nitrogen and hydrogen.
Temperature might be expected to playa vital role in the sliding
properties of metals. Complete surface failure and high wear
of bearing metals are usually characterized by surface welding.
The generally accepted adhesion concept of sliding

friction considers that cold welding occurs between sliding
metals in intimate contact and that frictional force is the product
of the cross-sectional area of the welded junction and the
shear strength of the metal in the plane of shear.
Surface films of oxides have a primary influence on

the friction, wear J and surface-failure properties of
*This is a condensation of a longer paper. For

further details please contact the authors directly.
-390-
metals in sliding contact. The literature contains numerous
instances where formation of oxide films during sliding pre-
vented surface welding of mutually soluble metal couples.
Many cryogenic liquids, including nitrogen and hydrogen,
prevent the reformation or repair of surface oxide s, once
these oxide films are broken or disrupted. The probability of
metallic welding could therefore be greater when the original
oxide layer is worn away during sliding. This suggests that
the selection of slider materials in cryogenic engineering can
be more critical than for applications in room atmosphere.
Nonsoluble metal couples and couples including a nonmetallic
material might be of particular interest for cryogenic sliding
surfaces.
The present investigation was an exploratory study

to determine the friction, wear, and surface-failure properties
of various types of slider materials operating in liquid nitrogen.
Apparatus and Procedure
The apparatus used for this investigation is shown

schematically in Figure 1. The basic elements of the apparatus
are a rotating-disk specimen and a hemisphere-tipped rider
specimen. The rider specimen slides in a circumferential
path on the lower flat surface of the rotating disk. Unless
otherwise specified, the specimens were run submerged in
liquid nitrogen.
The disk specimen was driven by a hydraulic motor

assembly. A rotative speed of 5,000 rpm was used for the
data reported here. Two sets of helium pur ged shaft seals
prevented leakage from the test chamber. The lower seal was
a double bellows type face seal with carbon seal rings sliding
on both faces of a shoulder that was an integral part of the
shaft. The shaft was made of 304 stainless steel as were
most internal parts of the apparatus. The upper seal was a
double controlled gap seal that had two closely fitted carbon
rings sliding on the rotating shaft.
The rider specimen was supported by an arm

assembly that allowed both application of load and measurement of
friction force outside the test chamber. The vertical shaft of this
assembly was pivoted through gimbal bearings mounted in the top
-391-
HELIUM SUPPl Y
ce:. eUAS T DlSI!;

~W"CUUH JACJl[T
~AOT.H£H:R
-HELlUH
- 1£51 'LUID
-VACUUf1
-VEtH
OElECtilUC S.fETY VALYE"
.PRESSURE A£GULATOA
&FlOW CONTROL
GPA[SSURI[ GAG(
.SHUT-o,r, GAS OPEfU.TEO
8PIESSURI[ RElIIE:' VALViE
"LOSE "AlWf-eIELlOWS SEAL
""£e,; VALVE
Figure 1. Cryogenic Figure 2. Cryogenic friction

friction apparatus apparatus. fluid transfer system
Figure 3.
~O. DIS 'Z 100 DIS
Figure 4A.
-392-
housing. The shaft was sealed at the top and bottom of the
gimbal mounting by flexible metallic bellows and, in operation,
the enclosed volume was pressurized with helium. Friction
force was measured by a strain-gage dynamometer ring
assembly. The deflection necessary for friction measurement
was so small «0.003 inch) that the restraining spring action
of the bellows was negligible. The spring propertie s of the
bellows were important, however, where extreme wear of
rider specimens neces sitated considerable (>0.040 inch)
displacement of the arm toward the center of the rotating speci-
men. This deflection of the bellows reduced the true load on
the wearing surface. Deflection of the top of the arm in the
direction of loading was measured with a fixed micrometer.
The runs were terminated when excessive wear caused unload-
ing of the spe cimens. Previous calibration made it pos sible
to determine true load with any amount of deflection. Load
was applied by a pulley and dead-weight system.
The top housing of the apparatus was insulated by

foamed-in-place Urethane. The lower housing and the transfer
line were insulated by vacuum jackets. The static heat leak in
the apparatus resulted in vaporization of less than one-half
liter of liquid nitrogen per hour. In operation, however, the
vaporization was primarily a result of friction heat from the
test specimens and the shaft seals.
Pressurized liquid nitrogen was transferred (see

Figure 2) through the vacuum-jacketed stainless steel tubing to
the test chamber. About 2 liters of liquid was required in the
apparatus for proper operation. The liquid level was deter-
mined by carbon resistors at different levels in the test
chamber. Although the remotely operated valves in the transfer
line were either fully open or closed, it was possible to main-
tain the desired flow rate for the proper liquid level by
controlling the dewar pressure. A safety circuit made it
impossible to close both valves in the transfer line at the same
time.
Stainless steel O-rings were originally designed into
the apparatus. Where possible, thin sections were used in the
apparatus to minimize heat leak. This type of design made the
apparatus too flexible to allow the high clamping pressures that
are necessary with stainless steel O-rings. Hence, all the
-393-
static seals used were made of either Teflon or Kel-F. No
trouble except cold flow has been experienced with Teflon and
Kel-F static seals. Two concentric Kel-F O-rings were used
to seal between the lower and the upper housings when they
were bolted together before each experiment. These rings
were reused 45 times or more before replacement was neces-
sary. The experimental materials used in these studies were
selected for various reasons that will be mentioned as the results
are presented. Specimen finishing was varied somewhat in the
experiments because of the great difference in materials.
After installing the experimental specimens in the

apparatus, the lower housing was securely bolted in place. Tre
seals were pressurized with helium and the transfer tube, the
test chamber, and the vent were purged with gaseous helium or
nitrogen. The storage dewar was pressurized with gaseous
helium or nitrogen. It required at least 20 minutes to cool the
transfer lines and apparatus and obtain about 2 liters of liquid
to submerge the specimens in the test chamber. Gaseous
nitrogen was vented through a pressure relief valve.
With the liquid level stabilized, the drive apparatus

was turned on and a predetermined rotative speed obtained.
The 1000 gram load was then applied and the experiment
begun. Frictional force was measured using a recording poten-
tiometer as a strain indicator. Dewar pressure was adjusted to
maintain a flow that would keep the liquid level .3 inches above
the test surface.
Displacement of the vertical arm was checked with

a vernier micrometer and the run was terminated if wear was
sufficient to cause partial (greater than 20 per cent) unloading
of the specimens. Specimens that gave acceptable wear were
usually run for periods of 1 to 5 hours. The occurrence of
welding and surface failure was indicated by high and unstable
frictionj such runs were terminated after a few minutes of
operation.
The wear of the specimens was determined by measur-

ing the diameter of the wear area and then calculating wear
volume, or by weight los s using an analytical balance.
Wear was also checked during the run by the micrometer
-394-
displacement method. The specimens were carefully studied
and photographed at low magnification after each run. Unles s
noted, all photographs were made using oblique illumination.
Results
Metals and Metal Compounds. Results obtained from

sliding austenitic stainless steel (type 304) on itself in liquid
nitrogen are given in Figure 3. The appearance of the
specimens indicated that surface welding was characteristic of
this slider couple. High friction ( fJ. = 0.78), rapid wear
(6.2 x 10- 3 cu inch/hr), and catastrophic metal transfer
occurred. Other runs at room temperature both in dry air and
dry gaseous nitrogen produced similar friction results, but
surface failure was more severe than in liquid nitrogen.
The results of sliding mild steel (SAE 1020) against

chrome steel (SAE 52100) are given in Figure 4A. Friction
coefficient (0.59) and wear (1.18 x 10- 3 cu inch/hr)
Two specimens of hardened chrome steel (SAE 52100)

were rubbed together for a very short time in liquid nitrogen
to obtain the data of Figure 4B. The friction coefficient was
0.52, which is about the same as that obtained at room
temperature. It might be expected that, in a run of such short
duration (0.2 min), the normal oxide film initially present on
the metal specimens would be capable of preventing surface
welding. The disk surface, however, shows metal transfer;
also the slight feather on the trailing edge of the rider specimen
is indicative of plastic flow at the interface.
-395-
52100 ~IDE~ S210D DIS
Figure 4B.
A9 RIDER
l04 DI~K
Figure SA.
Figure SB.
VERTICAL ILLU HIN ATIO N
OBLIOUE ILLU H I AlION

KI6ZB RIDER
Figure 6A.
-396-
Surface welding and transfer with the metal couples
previously mentioned indicate that the low temperature and any
boundary lubricating influence the liquid nitrogen might have
are not sufficient to make the use of these combinations
practicable under the conditions reported herein. It was sug-
gested earlier that with a surface oxide film worn away, the
mutual solubility of metal couples should be of increased
importance to the occurrence of surface welding. Silver has
little or no solubility in the primary alloy constituents of type
304 stainless steel. Therefore, runs were made with cast
electrolytic silver riders sliding on type 304 stainless steel.
The results of sliding these metals together are given in
Figure SA. Severe plastic flow of the silver was indicated
by the extensive feather edge. This detritus was separated
from the rider specimen as a result of handling after the run.
The initial friction coefficient was 0.80 and the wear was
0.63 x 10- 3 cubic inch per hour.
Surface welding in the usual sense did not occur but

a thin plastically smeared film of metallic silver was formed
on the steel disk. There is no doubt, however, that the true
sliding surfaces during much of the run were solid silver
against a transferred silver film. Chemical spot tests were
used to verify that the film was silver and to etch it away
from the base metal. The underlying metal suffered little or
no welding deformation. The metal transfer experienced in
this case might be due partly to mechanical interlocking of
metal surface asperities or roughness.
Data obtained with a composition containing 8S per

cent silver, S per cent copper, and 10 per cent molybdenum
disulfide given in Figure SB can be compared with those shown
in Figure SA for solid silver. There was a marked reduction
in the amount of silver smeared on the mating surface J and
no evidence of welding deformation of the base metal could be
detected. The friction coefficient (0.34) was more constant
and significantly lower than with solid silver. Wear (0.66 x
10- 3 cu inch/hr) was essentially the same as for the cast elec-
trolytic silver. It would appear that the improved lubrication
by the MOS Z -containing material was only sufficient to compen-
sate for the poorer physical strength. The cast silver would
have greater strength than the hot pres sed body weakened by
the solid lubricant.
-397-
Observation of the plastic behavior (ductility) of the
various metal combinations during sliding in liquid nitrogen
indicates that materials normally considered brittle at room
temperature might continue to be useful in cryogenic appli-
cations. In unreported NACA experiments a titanium carbide
base cermet has been found to have low wear and to resist
surface failure when run against tool steel with temperatures
from 80 0 up to 700 0 F. Experiments were performed with the
cermet (K162B) sliding against hardened 440-C stainless steel.
High hardness (54 Rockwell C) of the mating disk was utilized
in order to obtain the minimum real area of contact and
sustain high contact stresses which should provide a continuing
severe test of a brittle material. The data shown in Figure
6A indicate that the test was too severe with the 1000 gram
normal load.
Subsurface brittle fracture appeared to have occurred

in the cermet rider material. =
Both high friction (jJ. 0.63)
and high Hertz surface stress ( >150,000 psi) probably contrib-
uted to the surface deterioration. The nominal wear rate
based on volume calculated from the projected diameter of the
wear area indicates that wear was relatively low (0.05 x 10- 3
cu inch/hr). The surface of the rider showed several furrows
where local fracture occurred. The considerable depth of the
furrows makes the wear-rate calculation of questionable signifi-
cance. The rider ir regularities cut grooves in the wear track on
the disk. There was no evidence of surface welding; however,
this could have been masked by the surface fracture.
An additional run with the same combination but with

lower surface load (800 grams) Figure 6B shows that at this
condition friction remained high, wear was low (0.014 x 10- 3
cu inch/hr), and the severity of surface fracture was reduced.
The disk showed some evidence of pickup, which may have
been embedded fragments from the rider.
Nonmetals. The application of nonmetals to cryogenic

bearing and seal problems would seem well advised because, as
a group, they are considered inherently nonwelding to metal
surfaces.
During the preliminary operation of the apparatus,

an experimental anomaly was observed with a carbon material.
-398-
Figure 6B.
~H(NOL IHPQEGNA 1 EO
CARBON qlOfR
Figure 7A,
PHE~OlIHPRtGNATED
CAQaON RIDER
Figure 7B.
LA~ NA TED PHENOL RIDER
Figure 8.
-399-
Runs made with a phenol impregnated carbon (Purebon P5J)
sliding on type 304 stainless steel using an open dewar gave
very low wear (0.009 x 10- 3 cu inch/hr) as indicated in Figure
7 A. There was no surface disturbance on the disk and only
slight evidence of solid contact. When the same material
combination was run with the closed system, entirely different
results were observed (Figure 7B). The runs gave extremely
high and erratic wear values in seven duplicate runs from
34 x 10- 3 to 91 x 10- 3 cubic inch per hour. Frozen adsorbed
moisture films in the case of the open dewar is the apparent
explanation for the difference in wear rate.
The same basic grade of carbon was also run with

no impregnation (Purebon P5), with a double phenol impregna-
tion (Purebon 652), and with a metal-organo complex impreg-
nant (Purebon P5HT). In addition, data were obtained with
another phenol-impregnated carbon base material (Graphitar
39) that is comparable with Purebon P5J. The surfaces of
these carbon materials after test were very similar to that
shown in Figure 7B; the wear rates were all high (from 10 x
10- 3 to 30 x 10- 3 cu inch/hr); and the initial friction
coefficients were 0.84 for the nonimpregnated carbon, 0.53 to
0.62 for the phenol impregnated carbon, and 1.08 for the metal-
organo complex impregnated material.
Service experience with carbon brushes of aircraft

engine generators has shown that rapid wear can occur at high
altitudes. Laboratory experience indicated that this wear
phenomenon, called "dusting," occurred in the absence of
moisture and oxygen. In some instance s dusting was elimi-
nated by the use of a metal haloid impregnant that was capable
of forming a protective surface film. A carbon material having
such a metal haloid impregnant (Purebon P5N) was run a~ainst
type 304 stainless steel. It gave lower wear (1.04 x 10- cu
inch/hr) than, and friction (IJ. = 0.59) comparable with,phenol-
impregnated carbon. The P5N was the best carbon run for
use in liquid nitrogen; however, it is not as good as other
nonmetals to be discussed later.
A number of laminated phenolics were run against

type 304 stainless steel. It was felt that varied fineness of the
fabric and additions of solid lubricants to the materials might
-400-
influence wear and friction propertie s. The lowes t initial
friction obtained with a cotton-fabric -laminated phenolic
sliding against type 304 stainless steel in liquid nitrogen was
0.49. The addition of graphite and molybdenum disulfide
neither reduced wear nor lowered friction. For example,
Formica LN gave a wear rate of 0.11 x 10- 3 cubic inch per
hour, and the addition of graphite or MoS 2 resulted in wear
rates from 0.28 x 10- 3 to 0.42 x 10- 3 cubic inch per hour.
The additives also had an adverse effect on friction, resulting
in initial friction coefficients from 0.81 to 1. 40.
A fiberglass-filled melamine resin (Formica FF55)

was reported by the manufacturer to have good strength prop-
erties at cryogenic temperatures and was therefore considered
as a slider material. The results indicated extremely high
wear (10.6x 10- 3 cu inch/hr), high friction(1l = 1.04), and
abrasion of the metal surface by the glass particles.
The phenolic materials all had good appearing surfaces

after sliding contact. (See Figure 8, e. g.) The mating metal
surface usually had a slightly burnished transfer film. As
mentioned, high friction was noted for the phenolic materials.
The burnished appearance and high friction suggest that heat
dis sipation from the interface may be a problem in spite of
the surrounding low-temperature fluid (liquid nitrogen).
Polytetrafluoroethylene (Teflon) and polytrifluoro-

chloroethylene (Kel-F) have found considerable application in
cryogenic static seals. These materials also have excellent
friction properties under slow-speed sliding conditions. The
limiting factors in the use of Teflon for high-speed sliding
applications appear to be its poor ability to conduct heat from
the sliding interface and its tendency to cold-flow under load.
Kel-F has somewhat less tendency to cold-flow than Teflon.
Filler s, such as graphite or metals, can be added to Teflon
to decrease cold-flow tendencies and increase thermal
conductivity.
Figures 9A and 9B show surfaces that were typical

of runs in liquid nitrogen with Teflon and filled Teflon compo-
sitions sliding on type 304 stainless steel. The rider specimens
-401-
1 [fLON ~IOt R lOC DIS
Figure 9A.
Figure 9B.
FRICTION IN LIQUID NITROGEN

WEAR IN LIQUID NITROGEN (2300 'PM. 1000 GI
(2300 fPM. 1000 G. S MIN TO S HRSI
I.'
;;;
...c 1. 2 :;
.
~
z
g
o
~
.... .
"e.
o
..
1.0
::! ~
u
z
.." .,..o
.8
~ 0
~
~
~ ~ .6
v
.. . •
u
ii ~
e 0
u
...
~
c .2
~
~~ ...,
z
~
"'- II c
" 0
HETAlS NON-HET ' ~ S ON 304
Figure lOA. Figure lOB.
-402-
were always very smooth, usually highly reflective to light,
and showed no evidence of any adverse sliding conditions. The
highly reflective character of the wear surfaces made it
pos sible to detect elastic recovery that was at a maximum in
the center of the contact area. There was no opportunity for
a quantitative evaluation of this elastic behavior. The disk
specimens had very smooth highly polished surfaces in most
cases. There were, however, thin uniform transfer films of
compositions visible on the disks after operation. Etching of
the disk surfaces that had been run with unfilled Teflon showed
that the transfer film on the disk was highly protective to the
base metal.
As Figure 10 shows, the wear and friction for these

Teflon compositions was, in general, lower than for any other
type of material considered. The behavior was somewhat
erratic in different experiments with one of the best composi-
tions. Wear of the graphite-filled material in one run
(0.011 x 10- 3 cu inch/hr) was nearly ten times greater than in
another essentially identical experiment. Friction data also
had periodic fluctuations within the runs. Close examination
of the specimens showed a lack of homogeneity that was the
probable explanation of the above observations.
Glass -filled Teflon had relatively good characteristics

but did not give as low a wear or friction as a graphite -filled
composition (Figure 10). It did not abrade the ma ting metal
surface, as was observed with the glass-filled melamine resin.
Conclusions
An exploratory study was made of the friction, wear,

and surface-failure properties of various metallic and non-
metallic materials sliding together in liquid nitrogen. Data
were obtained at a sliding velocity of 2300 feet per minute with
an initial load of 1000 grams on a hemisphere-tipped rider
specimen sliding on the flat surface of a rotating disk.
Filled Teflon compositions such as one containing a

nominal 10 per cent graphite gave low friction (jJ. =
O. 15) and
wear. They may have usefulness as slider and bearing surfaces
-403-
in cryogenic liquids. Failure of Teflon in slider applications
has previously been associated with thermal degradation
accentuated by poor thermal conductivity. Cooling by the
liquid nitrogen probably played a vital role in the good results
obtained with Teflon compositions.
Friction and surface-failure properties of metal

combinations in liquid nitrogen were basically the same as at
room temperature. Metal transfer either by welding or
mechanical adhesion occurred with both soluble (304 on 304)
and nonsoluble (silver on 304) metal combinations. Considerable
plastic deformation was observed and interface ductility was
greater than at room temperature. Except for a silver compo-
sition containing MOS Z( I-" = 0.34), all metal combinations
including SAE 10ZO on SAE 5Zl00 gave friction coefficients
greater than 0.5Z. A cermet suffered subsurface brittle
fracture; this was reduced considerably by decreasing the load
to 800 grams.
Molded carbon materials that are commonly used in

sliding seals wore very rapidly in liquid nitrogen. Wear of
the carbons was decreased significantly by a metal haloid
impregnant of the type used to inhibit "dusting" of aircraft
generator brushe s at high altitude.
Laminated phenolic compositions had high friction,

developed a burnished transfer film on the mating surface, and
wore considerably more than Teflon compositions. Additive
fillers such as graphite and MoS Z did not improve the
performance of the phenolic materials.
Discussion
Question by D. A. Burgeson, Pesco Products, Division of

Borg Warner: Do you have comparative data for other
materials such as hard chromium?
Answer by author: We have not obtained data on chromium

slider surfaces in liquid nitrogen. The NACA research
programs in this field are relatively new and the present
paper is our first publication on slider surfaces for cryogenic
applications. No data other than that reported here have been
released by NAGA.
-404-
Question by J. M. Novajosky, Allison Division, General
Motors Corporation: Why did you choose to use a round
nosed rider rather than the more easily fabricated flat rider?
Answer by author: The hemisphere tipped specimen was

chosen because it allowed us to define the exact contact
conditions, provided an accurate means of measuring wear,
and eliminated misalignment as a primary experimental
variable. In the case of a flat rider, any elastic deformation
in the restraining system will cause edge loading and there-
fore contact conditions cannot be defined accurately until the
specimen is worn to full face contact. It might be suggested
that the changing unit loads, as the rider specimen wore,
could be an uncontrolled variable.
Question by J. M. Novajosky, Allison Division, General Motors

Corporation: What new would you expect from materials in
other cryogenic fluids such as liquid oxygen and hydrogen?
Answer by author: We have no data to report on work with

cryogenic fluids other than liquid nitrogen. It is well known,
however, that the slider characteristics of many materials are
dependent on surface reaction films such as oxides. With
friction energy providing flash temperatures at contacting
asperities that can be sufficient to initiate surface reactions,
the manner in which the surrounding fluid (e. g., oxygen or
hydrogen) influences the reaction can be of primary importance.
Therefore, one might expect considerable difference in the
behavior of various materials in the other cryogenic fluids.
Question by W. B. Gurney, Chandler-Evans: At what point

during the tests was the friction load measured?
Answer by author: A continuous record of friction was

obtained during these experiments. The data reported were
obtained during the initial part of the runs. Materials that
gave low wear usually provided constant friction during the runs.
-405-
Question by R. Gilman, Precision Equipment and Company,
Inc.: Do you have any comparative data for heat treated
Be-Cu on stainless?
Answer by author: We have no data for beryllium-copper

in liquid nitrogen.
Question by C. R. Allison, General Electric Company: Did

you investigate the lubricating effect of the liquid nitrogen
itself?
Answer by author: Friction coefficients for the various

materials in liquid nitrogen were usually similar to those
obtained in air at room temperature. This would suggest
that liquid nitrogen has little if any lubricating ability. Wear
of several materials was much lower in liquid nitrogen,
however, than in gaseous nitrogen or air. This result was
probably a reflection of the cooling ability of the liquid
nitrogen.
-406-
G-5
PROPERTIES OF 9 PER CENT NICKEL

LOW TEMPERATURE STEEL*
J. G. Althouse
Lukens Steel Company
Coatesville, Pennsylvania
*Neither abstract nor paper available at time of

printing.
-407-
G-6
THERMAL CONDUCTIVITIES OF COMMON COMMERCIAL

ALUMINUM ALLOYS
W. J. Hall, R. L. Powell, and H. M. Roder

Boulder, Colorado
Introduction
Aluminum is increasing in the number and kinds of

uses for cryogenic equipment. Because of this increase, an
experimental study of the low temperature thermal conductivi-
ties and electrical resistivities of some of the common
aluminum alloys was conducted. To further facilitate cryogenic
design, a table of compositions, grain sizes, hardnesses, and
uses is also presented.
Apparatus and Procedure
The method of determining thermal conductivity by

axial heat flow through a long cylindrical sample is used. A
pictorial diagram of the apparatus is shown in Figure 1. The
sample rod is clamped to the heat sink at the top; the sample
heater is attached to the bottom of this rod. The temperature
distribution along the sample is measured by means of eight
thermocouples attached to thermocouple holders positioned
along the rod. Significant loss from gas convection and con-
duction is eliminated by evacuating the region surrounding
the sample. Errors resulting from radiation and lead-wire
conduction are reduced by enclosing the sample rod within
a symmetric cylindrical thermal shield maintained at approx-
imately the same temperature and axial gradient as the sample.
The temperature gradient along the sample is determined by
measurement of the temperature at each of the eight thermo-
couple positions. The thermal conductivity can then be
calculated from the power input, the dimensions of the sample,
and this temperature gradient.
-408-
SHIELD HEAT
RING SINK
PLATINUM REFERENCE
RESISTANCE JUNCTIQH
TlIERMOIIIETER POST
VACUUII
SAMPLE CONTAINER
THERIID·
COUPLE
THERIIAL HOlDERS
SHIELD
SHIELD
HEATER
GLASS
DEWAR
SAMPLE
HEATER AND
SHIELD
DEWAR BOIL·OfF
SUPPORT HEATER
Figure 2.
Figure 1.
20 2. 0
15
<,y
5 .,.. . .•.
..
.... ."
~ ",. '
0:':
E
10
- G , 0:': I 0
7" ~
, '
'
8.0 E ·8 7 "7
~ o ~
~ 6.0
:;- '- 6
#Y
-
\ !? 1.1. ~ ,/
•3
VI
:: 4 .0 "0
/ ~ -~ / 2 I' 51 40
~ 3.0
,~~ ~ ~ ~ ~~ .. .,- ~
>-- I ~, ~v
>- 2.0
-Or
~
:;: 2 II' ..
~~ V ,~~
~ 110
:;: !-' .•. - '-'fo··
0: ~ V I--' 0:
o
g 1.0
v.; .....
::>
o .1 " II'
o ~
r
.8 ~ .0 THERMAL
o~
/. / I THERMAL VII' l CONOUCTIVITY
--
.~
.6 CONOUCTIVITY 0 .0 6
#/ I ALUMINUMS
.0 4~
1/
II' I ALUMINUM ALLOYS
.3 / I I 3
I
.6810 20 4060 100 200 JOO .0 4 6 8 10 20 40 60 100 200 JOO
TEMPERATURE, 'K TEMPERATURE, 'K
Figure 3, Figure 4.
-409-
The electrical resistivity is determined from the
dimensions and resistance of the sample. To measure the
resistance, a current of 50 to 60 milliamperes is passed
through the sample and the potential drop between the bottom
and top thermocouples is measured on a microvolt potentiome-
ter.
The heat sink is maintained at a constant low temper-

ature by a refrigerant, usually a liquefied gas. This refriger-
ant, poured into the glass dewar from a central fill tube,
surrounds the heat sink and vacuum can. The glass dewar is
supported by a metal container that is in turn fastened leak-
tight to a top plate. The dewar support container is further
surrounded by a metal dewar containing liquid nitrogen,
reducing heat conduction into the colder refrigerant. The
associated vacuum, electrical control, and measuring equip-
ment is located on nearby racks and benches. Figure 2 is a
photograph of the assembly.
Thermal Conductivities and Electrical Resistivities
Aluminum, like most metals, in its pure state and in

dilute alloys, exhibits a maximum in thermal conductivity as
shown in Figure 3. Others, for example, F. A. Andrews 1 , et
aI, at the Naval Research Laboratory in Washington, D. C.
have made measurements of better conducting single crystals.
Our particular sample was primarily measured for the sake
of comparison.
Two samples of 1100 (2S) aluminum were measured,

one in the annealed state, "0", and the other in the lias
fabricated" state, II F • II The allowable variation of chemical
composition could probably overshadow the effect of annealing.
For the two samples which were measured, photomicrographs
showed large voids in the annealed sample which were not
present in the lias fabricated" specimen. The combination of
voids and composition variation probably caused the unexpected
result that the thermal conductivity of the annealed specimen
IF. A. Andrews, R. T. Webber, and D. A. Spohr,

Phys. Rev., 6, No.5, 994 (1951).
-410-
was lower than that of the "as fabricated" specimen.
The only other alloy measured which was "dilute"

enough to show a maximum was the 6063 (63S) aluminum. This
specimen had been precipitation heat treated (TS). Although
its tensile and yield strength 2 and its hardness are comparable
to 5052-0 aluminum, its thermal conductivity is quite high.
For example our values show the ratio of the conductivity of
6063-TS to that of 5052-0 at 4 0 K is about 7; at 40 0 K it is
about 5.
The thermal conductivity of the 3003-F aluminum in

Figure 3 is for comparison with the conductivities in Figure 4.
The aluminum alloys in Figure 4 show no maximum. As the
amount of impurities increases, the thermal conductivity
decreases. As expected, the lowest conductivity measured was
that of a structural aluminum, 2024 (245), which had been
solution heat treated (T4) and contained nearly 7 per cent
alloying materials.
The heat treatment of the alloys does not seem to

affect the thermal and electrical conductivity greatly, but
other properties, such as yield strength, tensile strength,
and hardness can be improved by heat treabnent. This is
exemplified by the 1100-0 and 1100-F alloys as well as
6063-TS and 5052-0.
A comparison of Figure 5, which shows the electrical

resistivities of the measured samples, with the graphs of
thermal conductivity indicates the close relation between
electrical and thermal resistance. The electrical resistance
is nearly proportional to the product of the temperature and
the thermal resistance.
2Metals Handbook, American Society for Metals, 1948

edition and 1954 supplement.
-411-
6
4 ,,-.-
. . .,,- .
.
,,- -.-
- -
, .'"
20( 4-T4
L
*:~
515 q.-O ~
E 2
50~ 2-0
. .--
0 ./"I .
I ELECTRICAL ,
/,'/
/ .
E 10- 6 t--- RESISTIVITY // .'
.s=-
O -- /. I
I .
. I
.. 6 r---
ALUMINUMS ~
>-
t-
>
4 W. I
.,- ~
~
6063 -T5 I
- V
r--
t-
en V I
~ J
en 2
w 1100 0
Q:
.-J 10-7 J
<l:
u
-
Q:
6
V
t-
U
w 4
V
.-J
W \,Ii "T~IA r- ..-'
/
,....
10 50 100 300
TEMPERATURE, oK
Figure 5.
-412-
TABLE I
Composition and Properties of Aluminum Alloys
5ample Temper Per Cent Type of Grain 5ize Diamond

Composition 5ection* in mm Point
Hardness
1100-25 F 99AI L .024 x .008 45

T .012 44
1100-25 0 99AI L .040 x .032 36

T .036 34
6063-635 T5 0.45i, 0.7 Mg, L .052 x .048 71

98.5 Al T .052 64
3003-35 F 1.2 Mn, L .016 x .008 57

98.5 Al T .012 56
5052-525 0 0.25Cr,2.5Mg,L .056 x .032 62

97Al T .040 56
5154-545 0 0.25 Cr, 3. 5 Mg, L .036 x .028 68

96 Al T .032 64
2024-245 T4 0.6 Mn,1.5Mg, L .080 x .052 157

4.5 Cu, 93 Al T .048 145
*L--Longitudinal
T --Transverse
-413-
TABLE II
Uses of Aluminum Alloys
Sample Typical Uses Comments
1100 Food and chemical handling Good formability,

equipment, tubing for corrosion resistant,
liquid lines, liquid systems low structural
and outer shells of dewars. strength.
6063 Extrusions, tubing, irriga- Relatively high con-

tion pipe, "T" sections, ductivity, fair struc-
structural reinforcement tural strength.
for dewar cylinders.
3003 Primarily the same uses as Same properties as

1100 alloy. 1100 alloy, but with
greater strength and
better weldability.
5052 Aircraft fuel and oil lines, Good workability,

fuel tanks, transportation high corrosion
and marine equipment, resistance, high
inner and outer cylinder fatigue strength, fair
construction. static strength.
5154 Primarily the same as Same properties as

5052 alloy but not as 5052 alloy, but with
commonly used. greater strength.
2024 Aircraft structures, rivets, High strength, low

hardware structural appli- c ondu c ti vity •
cations, baffles for the inner
cylinder of dewars.
-414-
Composition, Grain Size, and Hardness
The compositions listed in the following table are the

approximate compositions for each kind of alloy as stated by
the Alumi.num Company of America. Our own chemical
analysis showed each sample to be within the composition
limits specified by the company.
The hardne s s and grain size of each sample was

measured on both a longitudinal and a transverse section.
Summary
From the measurements that were made, there is an

indication that the heat treatment of an alloy has very little
effect on the thermal and electrical conductivities compared
to the effect of chemical composition. In the design of
cryogenic equipment, the thermal and electrical conductivities
are often as important as structural strength. Since aluminum
is corrosion resistant and light, its proper use in cryogenic
equipment is becoming more and more important.
Discussion
Question by D. Rahm, Brookhaven National Laboratory: What

about the corrosion resistance and cost of aluminum alloys
relative to stainless steel?
Answer by author: The corrosion resistance of alloys of

aluminum depends upon the alloying element, the percentage
of alloying, and, in some cases, the heat treatment. We have
made no measurements of corrosion resistance, but suggest
Metals Handbook (1948 edition), published by the American
Society for Metals, as a reference. We have no figures on
the relative cost.
-415-
G-7
A COMPARISON OF INSULATING MATERIALS*
J. E. Schrodt, M. M. Fulk, W. A. MCReynolds,

and W. S. Flanery
Boulder, Colorado
The increased use of various solids as insulating

material for low temperature equipment has prompted the
study of factors influencing heat transport through them.
A comparison of the apparent thermal conductivities

for organic foams, glass fibers, powders, gases, hi-vacuum,
etc., will be given.
The method of measurement will be illustrated

schematically with special attention given to the flat-plate
calorimeter.
~'Abstract of paper given. More complete information

will be published at a later date.
-416-
AUTHOR INDEX
Author Author Paper Index Page
Althouse, J. G. G-5 407

Akers, W. W. F-l 275
Aronson, D. A-3 19
Arrick, C. D. D-8 218
Aton, W. W. C-6 160
Ball, R. H. D-5 196
Birmingham, B. W. C-3 125
Blewitt, T. H. D-2 173
Brickwedde, F. G. B-5 99
Brown, R. E. C-2 114
Bunch, M. D. E-6 269
Chelton, D. B. A-I 1
Class, C. R. B-1 64
Coltman, R. R. D-2 173
Corruccini, R. J. G-l 353
Dean, J. A-I 1
Dearborn, F. K. E-l 226
Draper, J. W. A-6 58
B-2 73
D-4 191
Eeles, E. G. E-4 248
Eichstaedt, H. A. E-3 238
Eubanks, L. S. F-l 275
Flanery. W. S. G-7 416
Flynn, T. M. A-6 58
Fogelberg. C. V. E-2 232
Fulk, M. M. G-7 416
Gardner. J. B. A-4 32
Gates. R. G. F-5 325
Gomez, P. D-l 172
Goodwin. R. D. E-5 254
Greenfield. S. C-4 136
Hady. W. F. G-4 390
Hall. W. J. G-6 408
Hardy. R. J. D-6 209
Hiza. R. M. F-3 305
Holben. C. D. E-2 232
Holten. D. C. G-3 375
Howe, J. T. D-2 173
Hunsaker. J. P. G-2 307
-417-
Huntley, S. C. F-6 342
Isakson, V. E. E-2 232
Jacobs, R. B. D-6 209
G-3 375
Johnson, R. L. G-4 390
Johnson, V. J. A-2 11
Kayan, C. F. F-5 325
Macinko, J. A-I 1
Mann, D. B. C-3 125
Martin, K. B. D-6 209
McClintock, R. M. F-3 305
McIntosh, G. E. B-1 64
McReynolds, W. A. G-7 416
Meyers, N. D-l 172
Mikesell, R. P. F-4 316
Pastuhov, A. G-2 367
Powell, R. L. E-6 269
Purcell, J. R. D-4 191
Reed, R. P. F-4 316
Richards, R. J. G-3 375
Rind, H. D-l 172
Robbins, R. F. G-3 375
Robinson, G. Y. C-l 106
Roder, H. M. G-6 408
Schaffers, T. W. B-4 92
Scharle, W. J. C-5 149
Schrodt, J. E. G-7 416
Scofield. H. M. A-5 47
Seavey, D. W. D-7 217
Selldorf, J. T. C-l 106
Smith, H. J. D-3 179
Smith, K. C. A-4 32
Spero. R. P. B-1 64
Timmerhaus, K. D. A-6 58
Tracey, J. E. E-l 226
Vander Arend, P. C. C-3 125
Van Valin, C. C. B-2 73
Wang, D. 1. J. F-2 294
Weitzel, D. H. A-6
B-2
. 58
73
B-3 85
D-4 191
Wilson, W. A. B-3 85
Wisander, D. W. G-4 390
-418-

Advances in Cryogenic Engineering

Uploaded by

Document Information

Original Title

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

Advances in Cryogenic Engineering

Uploaded by

Copyright:

Available Formats

A Cryogenic Engineering Conference Publication

Proceedings of the 1957

By popular request, the National Bureau of Standards

To make the Proceedings more interesting an attempt

To obtain individual reprints of anyone particular

1957 CRYOGENIC ENGINEERING CONFERENCE COMMITTEE

B. W. Birmingham National Bureau of Standards

S. C. Collins Massachusetts Institute of

W. T. Ziegler Georgia Institute of Technology

The 1957 Cryogenic Engineering Conference Committee

L' Air Liquide

1957 CRYOGENIC ENGINEERING CONFERENCE

A-I Hydrogen Liquefaction Cycles . • . . . • . • 1

A-2 Removal of Nitrogen from Hydrogen

A-3 Preliminary Design Studies of Low

A-4 Power Consumption and Thermodynamic

A-5 The Reversible Separation of Multi-

A-6 Separation of Hydrogen Isotopes by

Cryogenic Processes (continued)

B-1 Efficient Utilization of Ortho-Para

B-2 Design Data for Ortho-Parahydrogen

B-3 Vapor Phase Ortho-Para Conversion

B-4 Design and Application of a Gas

B-5 The Thermodynamic Theory of a

C-l Design and Performance Data on a

C-2 Economic and Other Aspects of the

C-3 Technical Aspects of Large Scale

C-4 Dilution of Cryogenic Liquid Rocket

C-5 Liquid Oxygen for Missile Testing 149

C-6 Unattended Oxygen Plants. .. .... 160

D-l On the Design of a Cryogenic

J. T. Howe, R. R. Coltman, and

D-3 Integral Probe Capacitance Gaging of

D-4 A Unique Thermal Conductivity Gas

D-S Piston Type High Pressure Aircraft

D-6 Operation of Bearings and Pumps at

D-7 1000 Gallons Per Minute Liquid Oxygen

D-8 Contaminants in Liquid Oxygen as Related

E-l A Cryostat for Liquid Nitrogen Cool-

E-2 Beech Aircraft 6000 Liter Liquid

E-3 Vacuum Insulated Dewars for Liquid

E-4 A Precision Low Temperature X-Ray

E-5 Design of Simple DC Resistance

E-6 Calibration of Thermocouples at

F-l Vapor-Liquid Equilibria in the System

W. W. Akers and L. S. Eubanks

F-2 Vapor-Liquid Equilibrium Studies on

F-3 Epoxy Resins as Cryogenic

F-4 The Impact Testing of Various Alloys

F-5 Performance Prediction of Low

F -6 Temperatur e -Pre s sure - Time Relationships

G-l Calculation of Gaseous Heat Conduction

G-2 Measurement of Thermal Stresses at

G-3 Heat Transfer to Boiling Liquid Nitrogen

G-4 Friction Studies of Various Materials

G-5 Properties of 9 Per Cent Nickel Low

G-6 Thermal Conductivities of Common

G-7 A Comparison of Insulating Materials. 416

HYDROGEN LIQUEFACTION CYCLES

J. Macinko, D. B. Chelton, and J. Dean

Because of increased activities in cryogenic engineering,

This is a brief summary of the results of the cycle

In the thermodynamic analysis of cycles, certain