The Diffusion of Gases through Metals. II.

Diffusion of Hydrogen through Aluminium

Author(s): C. J. Smithells and C. E. Ransley
Source: Proceedings of the Royal Society of London. Series A, Mathematical and Physical
Sciences, Vol. 152, No. 877 (Nov. 15, 1935), pp. 706-713
Published by: The Royal Society
Stable URL: http://www.jstor.org/stable/96543 .
Accessed: 04/05/2014 09:10

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp

.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.

The Royal Society is collaborating with JSTOR to digitize, preserve and extend access to Proceedings of the
Royal Society of London. Series A, Mathematical and Physical Sciences.

http://www.jstor.org

This content downloaded from 194.29.185.92 on Sun, 4 May 2014 09:10:21 AM

All use subject to JSTOR Terms and Conditions
 706

The Diffusion of Gases through Metals

JI-Diffusion of Hydrogen through Aluminium
By C. J. SMITHELLS,D.Sc., and C. E. RANSLEY, B.Sc. (Research Staff of
the General Electric Co., Ltd., Wembley)

(Communicatedby R. H. Fowler, F.R.S.-Received June 21, 1935)

The only direct attempt to determine the rate of diffusion of hydrogen

through solid aluminium, of which we are aware, is described in a paper
by Deming and Hendricks,* in which they conclude that aluminium is
impermeable to hydrogen at temperatures up to 550 C. Very slow
rates of diffusion would not have been detected by the method employed.
The solubility of hydrogen in aluminium has been measured by Iwase,t
Rdntgen and Braun,j and Rontgen and Moller.? No solubility was
detected in solid aluminium, but molten aluminium dissolves about 04 1
to 0 2 cc of hydrogen per 100 gm of metal at the melting-point, and the
solubility increases with temperature.
It is well known that molten aluminium in contact with the ordinary
atmosphere decomposes water vapour with the formation of aluminium
oxide, and that much of the hydrogen liberated is absorbed by the metal.
On freezing,the greater part of this hydrogen is liberated in the form of
bubbles, some of which remain entrapped in the solid metal. As a
result of this, commercial aluminium usually contains an appreciable
amount of gas, 80% of which is hydrogen. An average analysis shows
about 14 cc of gas per 100 gm of metal.lI Hanson and Slater? concluded
from indirect evidence that during atmospheric corrosion of aluminium
the nascent hydrogen produced is absorbed by the metal, and if this is
true the gas must be soluble in the solid metal.
In an earlier paper** we have shown that the diffusion of a gas into
a solid metal depends on adsorption on the surface and may be markedly
influenced by the presence of an oxide film. Pilling and Bedworthtt
* ' J. Amer. Chem. Soc.,' vol. 45, p. 2857 (1923).
t ' Sci. Rep. T6hoku Imp. Univ.,' vol. 15, p. 531 (1926).
1 ' Metallwirtschaft,' vol. 11, p. 459 (1932).
? ' Metallwirtschaft,' vol. 11, pp. 685, 697 (1932).
1IR6ntgen and Braun (loc. cit.).
?T'J. Inst. Metals,' vol. 46, p. 187 (1931).
** ' Proc. Roy. Soc.,' A, vol. 150, p. 172 (1935).
tt 'J. Inst. Metals,' vol. 29, p. 573 (1923).

This content downloaded from 194.29.185.92 on Sun, 4 May 2014 09:10:21 AM

All use subject to JSTOR Terms and Conditions
 Diffusion of Gases through Metals 707
haveshownthata filmof aluminiumoxide0 *00002cm thickon aluminium
is imperviousto oxygen. Since aluminiumin contact with the atmo-
sphereis always coveredby a film of oxide it seemedpossible that the
failureto establisheitherdiffusionor solid solubilityof hydrogenmight
be due to the action of this film in preventingdiffusionof the gas into,
or out of, the metal.
We have thereforeexaminedthe diffusionof hydrogenthroughalumin-
ium in the normally tarnishedcondition, after anodic oxidation, and
with the oxide film removedby abrasionin vacuo. Preliminaryresults
already published* led us to the conclusion that hydrogen diffuses
throughaluminium. These resultsare confirmed,and it is shown that
the rate of diffusionis dependenton the state of the surface.

APPARATUS AND METHOD

The apparatusused was substantiallythat describedin the earlier

paper except for modificationsintroducedso that both the inner and
outer surfaces of the aluminiumtube could be vigorously scraped in
vacuowithout removingthe tube from the diffusionapparatus. Only
the modificationsto the apparatusneed be described.
Material-Commercialaluminiumbar 1 inch in diameterwas drilled
and drawninto tubinghavingan externaldiameterof 3 2 mm and a wall
thicknessof 0 45 mm. The metal had the followinganalysis:
0/

Copper .................... 0-03

Iron ...................... 0*90
Manganese ................ 0-03
Silicon .................... 0i4
Zinc ...................... 0d17
A piece of tube 30 cm long was closed at one end by welding,and the
otherend was solderedto a copper-to-glass joint leadingto the analytical
system.
Surface Treatment-Whenaluminiumis exposed to the air a trans-
parentbut protectivefilm of oxide forms rapidlyon the surface. The
tube when introducedinto the diffusionapparatushad been exposedfor
some days and was thereforein this condition. The only practical
method of removingthis film is by mechanicalabrasion,and arrange-
* 'Nature,' vol. 135, p. 548 (1935).
3 B 2

This content downloaded from 194.29.185.92 on Sun, 4 May 2014 09:10:21 AM

All use subject to JSTOR Terms and Conditions
708 C. J. Smithells and C. E. Ransley
ments were made for scraping both the inside and outside surfaces of
the tube in vacuo in the diffusion apparatus. Scratch brushes of fine
steel wire bearing with considerable pressure on both the inner and outer
surfaces were attached by nickel tubes to blocks of iron which could be
moved horizontally by means of a powerful solenoid, so that either surface
could be scratched independently. This method of controlling mechani-
cal movement in a vacuum was described in the previous paper with
respect to the exploring thermocouple. The brushes were found to be
very effective, producing pronounced scratches over the surface of the
tube, and a considerable quantity of abraded matter collected in the
bottom of the diffusion bulb.
TemperatureControl-The presence of the scraper made it difficult to
attach currentleads as before, and the tube was therefore heated indirectly
by means of a furnace, which consisted of a tungsten helix between two
concentric quartz tubes. The temperature distribution along the tube,
measured as before with an exploring thermocouple, is less satisfactory
with this method of heating. The equivalent length of tube at constant
temperaturewas estimated graphically, using the temperature coefficient
of diffusion determined as described later in this paper. The effective
hot length of tube varied from 3 to 4 cm in different experiments.
Preliminary Out-gassing-Before carrying out diffusion measurements
the tube was thoroughly out-gassed at 605? C, and the glass work baked
out with both sides of the apparatus connected to the vacuum pumps.
The out-gassing was usually complete after 12 hours, and was repeated
each time the tube was exposed to the atmosphere.

EXPERIMENTALRESULTS

Effect of Surface Treatment-A series of diffusion measurementswere

made on the same tube subjectedto various surface treatments. All these
experiments were made with the tube at 580? C (853? K) and a hydrogen
pressure of 29 cm, under which conditions sufficient hydrogen diffused
during half an hour to enable an accurate estimate of the rate to be
made in that time. The details of these separate runs are given below
and the results are summarizedin fig. 1. As before, the rate of diffusion
D is expressed as the volume of gas in cubic centimetres at N.T.P.
diffusing per second through 1 sq. cm of surface of 1 mm thickness.
Run 1-The tube was run in hydrogen for 3 hours before any observa-
tions were made. The rate was then 2-0 x 10-7 but appeared to be
diminishing slightly with time.

This content downloaded from 194.29.185.92 on Sun, 4 May 2014 09:10:21 AM

All use subject to JSTOR Terms and Conditions
 Difusion of Gases through Metals 709
Run 2 The tube was allowed to cool and, without altering any other
condition, the outer surfacewas vigorously scratched. The tube was again
heated to 580? C, and at the end of the first hour the rate (3 *8 x 10-7)
was appreciably higher than before, but was definitely decaying.
Run 3-The outer surface was scratched again and the average rate of
diffusion determined during successive 2-hour periods. The initial rate
was now 4 2 x 10-7, but fell during 5 hours to a fairly constant value
at 1i4 x 10-7.
Run 4-The outer surface was scratched again, producing an initial
rate of 4 0 x 10-7, which fell in 7 hoursto 0 9 x 10-7.
7

6
5

i _4

O l 3 4 5 6 7
Hours
FIG. 1-Effect of surface treatment on rate of diffusion. Run 1, exposed to air;
run 2, outer surface scratched; run 3, outer surface scratched again; run 4,
outer surfacescratchedagain; run 5, inner and outer surfacesscratched;run 6,
outside anodicallyoxidized.
Run5-Both the inner and outer surfaces of the tube were scratched.
The average rate during the first half-hour was now 6 3 x 1O-', and
from the rapid rate of decay it would seem that the rate during the first
few minutes must have been nearly 10 0 x 1O-.
From these experiments it appeared that hydrogen diffuses quite
readily through aluminium at 580? C, the rate being of the same order
as we have found for copper at this temperature. It is apparent, how-
ever, that the rate may vary by a factor of at least 10 to 1 according to
the condition of the surface. Whilst the initial rates in the different runs
vary considerably, the curves of fig. 1 indicate that after about 10 hours
the surfacereturnsto a definitestate, for which the rate is about 0 *8 x 10-.
It seems almost certain that this condition corresponds to a definite

This content downloaded from 194.29.185.92 on Sun, 4 May 2014 09:10:21 AM

All use subject to JSTOR Terms and Conditions
710 C. J. Smithells and C. E. Ransley
degree of oxidationof the surface. Although the hydrogenwas pure
and dry, being introducedby diffusionthrougha palladiumtube, some
of it wouldbe convertedby contactwiththe hottubeinto atomichydrogen,
whichcan reducethe oxidizedpartsof nickelwires,etc., evenin the cold.
The water vapour so formedwould oxidize the aluminium. It seemed
of interest,therefore,to determinethe rate of diffusionthroughaluminium
coveredwith a definitelyprotectiveoxide film such as that producedby
anodic oxidation.
Run 6-The tube was removedfrom the apparatus,carefullypolished
andanodicallyoxidizedin a 3%ochromicacid solution. Afterthe usual
out-gassingthe initialrate of diffusionwas found to be 1 33 x 10-7 and

0-7

0 2 4 6 8 10 12 14 16 l8mm

FIG. 2-Effect of pressureon rate of diffusionat 8310K.

this fell during a period of 22 hours to about 0*9 x 1O7. These rates
are of the same order as the final values obtained in the previousruns, and
support the conclusion that the decay is due to oxidation of the surface.

Effect of Pressure-The effect of pressure on the rate of diffusion at

558? C was determined for a tube in the fully decayed condition (D-=
0 8 X 10-7 at 580? C and 29 cm pressure). The results, which are given
in Table I and plotted in fig. 2, are representedby the usual equation

D = k VP

within the limits of experimental error. The rate of diffusion was too
slow to enable accurate measurements to be made at very low pressures,

This content downloaded from 194.29.185.92 on Sun, 4 May 2014 09:10:21 AM

All use subject to JSTOR Terms and Conditions
 Diffusion of Gases through Metals 711
and it is not possibleto say whetherthe usualdeviation*fromrrthe square
root law at low pressurestakesplace.
Effectof Temperature-Theeffectof temperature on the rateof diffusion
was determinedfor a tube havingthe high rate of diffusioncharacteristic
of a freshlyscrapedsurface,and also for the anodicallyoxidizedtube.
The high rate was obtainedwith the same tube used for the previous
measurements,carefullypolished before assembling,but not otherwise
specially treated. On this occasion the tube maintaineda high rate
TABLE I-EFFECT OF PRESSURE ON RATE OF DIFFUSION AT
5580C (831? K)
Pressure Rate
mm D
32.5 0 18 x 10-7
95 0-32
150 0.42
290 0.59

(5 0 x 1O-' at 580? C and 29 cm pressure) for 7 hours without showing

any markeddecay, althoughthe experimentalconditionswere normal.
The resultsare given in Table II, and in fig. 3, log D is plotted against
1/T. It is clearthat the usual exponentialequation
D ke-bT
representsthe effect of temperatureon the rate of diffusion.The value
of b given by the slope of the line in fig. 3 is 15,600. This is in close
agreementwith the value of 15,400foundpreviouslytfor a differenttube.
The temperaturecoefficientis thereforemuch higherthan for any other
metal-hydrogensystem for which data are available.
It was difficultto obtainconsistentresultswith the anodicallyoxidized
tube owingto the verylow rateof diffusionat any but the highesttempera-
tures.
The observationsplotted in fig. 3, and markedin the orderthey were
made, indicatea slightlyhighertemperaturecoefficientfor the anodically
oxidizedsurface. Since the actualrate of diffusionof hydrogenthrough
aluminiumvaries so much with the condition of the surface,it is not
possible to give an absolute value to the diffusioncoefficientk in the
equation
D - \/Peb/T
d
* Smithellsand Ransley, ' Proc. Roy. Soc.,' A, vol. 150, p. 172 (1935).
t Smithellsand Ransley, 'Nature,' vol. 135, p. 548 (1935). In this note the value
was given as 14,000,owing to an arithmeticalerror,insteadof 15,400.

This content downloaded from 194.29.185.92 on Sun, 4 May 2014 09:10:21 AM

All use subject to JSTOR Terms and Conditions
712 C. J. Smithells and C. E. Ransley
TABLE II-EFFECT OF TEMPERATUREON THE RATE OF DIFFUSION
AT 29 CM PRESSURE
Temperatutre Rate
?K D
782 1 *09 x 10-7
793 1-17
805 1*75
818 2 33
842 4 09
865 5 *92
878 9 70

Z0

112 114 116 118 120 122 124 126 128

\. ~~100,000
T
FIG. 3- Effect of temperature on the rate of diffusion at 29 cm. after exposure to
air; - - - after anodic oxidation.

The value found in our experimentsvaries from 3 3 for a freshly scraped

s-urfaceto 0 42 for an anodically -oxidi'zedsurface.

These experimentsshow that hydrogen diffuses-throughsolid aluminium.

The usual diffusion e,quation represents the effects of temperature and
pressure on the rate of diffusion, and we therefore conclude that the
hydrogen is dissociated at the metal surface. Even under the best con-

This content downloaded from 194.29.185.92 on Sun, 4 May 2014 09:10:21 AM

All use subject to JSTOR Terms and Conditions
 Diffusion of Gases through Metals 713
ditions it is difficult, if not impossible, to avoid contamination of the
surface, probably by oxygen. This reduces the rate of diffusion both
into and out of the metal. The very high temperature coefficient of
diffusion, in conjunction with this poisoning effect, results in extremely
slow rates of diffusion at temperaturesmore than 1000 below the melting-
point of the metal. The fact that hydrogen does diffuse through alumin-
ium shows that the gas must have some solubility in the solid metal.
The failure of previous authors to detect this solubility is probably due
to the slow rate of diffusion under ordinary experimentalconditions.

SUMMARY

It is found that hydrogen diffuses through aluminium at a measurable

rate at temperatures above 4000 C. The rate of diffusion depends on
the state of the surface. The highest rate is obtained with a surface
freshly scraped in hydrogen, but it falls off rapidly and after some hours
reaches a steady value which is only about one-tenth of the initial rate.
This, is attributed to contamination of the surface by oxygen, since this
low rate is obtained with an anodically oxidized surface.
The effects of temperature and pressure are satisfactorily represented
by Richardson's equation
D -kkVPe-'T

The value of b for the freshly scraped surface is 15,600 and for the
anodically oxidized surface about 21,500. The value of k varies from
3 *3 to 0 *42 for different states of the surface.

This content downloaded from 194.29.185.92 on Sun, 4 May 2014 09:10:21 AM

All use subject to JSTOR Terms and Conditions