Conduction of Heat through Rarefied Gases .
By Frederick Soddy , M.A. , and Arthur John Berry , B.A. , Physical Chemistry Laboratory , University of Glasgow .
( Communicated by Prof. Sir J. Larmor , Sec. R.S. Received November 16 , \#151 ; Read December 9 , 1909 .
) In a previous paper* we used a thermal method of indicating the degree of vacuum by measuring the rate of evaporation of liquid air in a Dewar flask exhausted in various ways .
In the present work we have attempted to-obtain information of the conduction of heat through twelve gases at pressures so low that the actual path of the molecule is comparable with its mean free path .
It is to be expected that this condition will hold good over a range of pressure the greater the smaller the diameter of the containing vessel , and for this reason we worked in a long narrow tube .
Previous investigations of this character have been carried out by Sir William Crookesf and C. F. Brush , J by measuring the rate of cooling of heated mercury thermometers placed inside globes exhausted by the Sprengel pump .
The observations of the latter bear most closely upon the present work , and partly anticipate them .
He points out that in the five gases he examined , at pressures up to a few millionths of an atmosphere , the heat-transmitting power of the gas varies directly as the pressure .
This is to be expected from the kinetic theory , as pointed out by Smoluchowski de Smolan .
S The principle of the method we have employed was to maintain a bright platinum strip , stretched through a long narrow tube containing the rarefied gas , at a constant temperature , found subsequently to be 61 ' , by means of a varying current .
The strip formed one arm of a Wheatstone 's bridge , the other arms of which consisted of three similar unvarying resistances made of stout Eureka resistance wire .
The heating current was adjusted until balance was obtained .
By means of potential leads fixed to the strip , the fall of potential across the strip was then immediately measured by means of a potentiometer .
The resistance of the strip being constant , the heat dissipated by the portion of the strip between the potential leads is proportional to the square of the difference of potential at the potential leads .
* 'Roy .
Soc. Proc.,5 A , vol. 78 , 1907 , p. 429 .
t 'Roy .
Soc. Proc.,5 A , vol. 31 , 1880 , p. 239 .
t 'Phil .
Mag.,5 vol. 45 , 1898 , p. 31 .
S 'Wiener Sitzungsberichte,5 vol. 107 , abth .
iia , 1898 , p. 304 .
Conduction of Heat through Rarefied Gases .
255 The method of electric heating had previously been employed by Schleier-macher for work on thermal conductivity of gases , * and more recently by Schwarze.f It had also been employed by Bottomley for radiation experiments .
] ; In our apparatus for pressures up to about 0'05 mm. for all the gases , and considerably higher for helium and hydrogen , the curves connecting the rate of dissipation of energy with the pressure were found to be straight lines , of varying slope for the various gases , which all converge on the axis of zero pressure at the same point .
This point , which gives the radiation loss , was also directly determined in the highest vacuum that could be obtained by the use of calcium , and the value found was identical with that deduced from the curves .
The heat lost by radiation was small , being about 4*5 per cent , of the heat conducted by carbon dioxide , the worst conductor examined , or 0*5 per cent , for hydrogen , the best conductor , at pressures such that the conductivity of the gas was not further increased by increase of pressure .
An interesting result followed when the slopes of the curves at low pressure were compared together .
It was found that this bore little relation to the ordinary limiting conductivity of the gas as determined at higher pressures .
Indeed at these low pressures hydrogen and helium no longer appear as the best conductors .
Methane acetylene and cyanogen are slightly better conductors than hydrogen , while helium is only slightly superior to carbon dioxide , which at ordinary pressures was the worst conductor of all the gases examined .
The experiments were carried out two years ago , but only recently have the results been completely reduced .
The apparatus employed has been partly dismantled , but it is proposed to set it up again and to re-examine some of these points , with better arrangements for keeping the temperature of the water in the water-jacket constant .
On account of fluctuations in the temperature of the water the results here given are not so accurate as the method is capable of , particularly in the case of the better conducting gases and at the higher pressures , where the energy dissipated is considerable .
But there is no reason to question the general character of the results , and a brief description of the work already done may be of interest .
Description of the Apparatus employed .
A diagram of the apparatus is shown in fig. 1 .
AB is the platinum strip , with bright polished surface , 50 cm .
long and 0*244 cm .
wide , stretched by * ' Annalen der Physik,5 vol. 34 , 1888 , p. 623 .
t Diss .
, Halle , 1902 ; 'Ann .
der Physik,5 1903 , p. 303 .
t 'Phil .
Trans.,5 A , vol. 184 , 1893 , pp. 591\#151 ; 616 ; 'Phil .
Mag.,5 June , 1900 , p. 543 , and Nov. , 1902 , p. 560 .
Messrs. F. Soddy and A. J. Berry .
[ Nov. 16 , means of a weight C along the axis of the narrow thin-walled tube DE , which is water-jacketed .
The internal diameter of DE was about 1 cm .
Current is supplied to the strip through the leads a , a , and b , 6,..are the Fig. 1 .
1909 .
] Conduction of Heat through Rarefied Gases .
257 potential leads of fine platinum wire 41 cm .
apart .
The tube containing the strip is connected through the U-tube F to the uppermost of three tubes , G , H , I of the measuring gauge JK .
The latter is similar to that already-described.* The tube H communicates with the arrangements for admitting and withdrawing gas , and the tube I with the calcium furnace L. By allowing mercury to stand at suitable height in the gauge the connection between the parts of the apparatus on the left of the gauge could be sealed off by mercury from the parts of the apparatus on the right of the gauge and all leakage prevented .
The arrangements made for introducing known amounts of gas , too small in quantity to be measured by the gauge , alone call for further description .
The volume between the taps X and Y was accurately determined to be 0479 c.c. The volume of the bulb M was 634 c.c. The ratio between the two volumes is 0-00282 .
Initially this bulb and the rest of the apparatus are exhausted to the highest possible degree by the calcium furnace , which was then shut off by allowing the mercury in the gauge to rise above I. The space between the taps is then filled with the gas at atmospheric pressure and , by reversing the three-way tap Y , the quantity taken is shared with the bulb .
Now , 0'00282 x 0479 c.c. = 0'0005 c.c. remains between the taps .
This is admitted to the apparatus .
One " complete dose " was found to increase the pressure in the apparatus as measured by the gauge by 0423 mm. Hence one " divided dose " increases it by 0-00119 mm. If the series of operations is carried out as described n times the pressure of gas in the apparatus after n divided doses have been admitted is 1/ 2 n(n + 1 ) x 0"00119 mm. The pressures calculated by means of this formula agreed perfectly with the gauge readings at pressures high enough to be measured by means of the latter .
The electrical arrangements are shown in diagram in fig. 2 .
AB is the platinum strip , C , D , and E the three similar resistances forming the bridge .
The galvanometer G was a sensitive d'Arsonval instrument of low resistance suitably protected by shunts , not shown in the figure , which were cut out when approximate balance was obtained in the bridge .
Owing to the heavy currents and sensitive galvanometer employed balance could be obtained with great accuracy .
The resistances for varying the heating current , shown diagrammatically at F , consisted of Eureka wires in tubes which could be filled or emptied of mercury , in parallel with a one ohm Paul-Harris test wire .
As soon as balance was obtained the connections shown by the dotted lines * Soddy and Mackenzie , ' Roy .
Soc. Proc. , ' A , vol. 80 , 1908 , p. 98 , fig. 1 .
258 Messrs. F. Soddy and A. J. Berry .
[ Nov. 16 , at a , by dy e were broken and those at a } c } d , f simultaneously made by a commutator , for the measurement of the fall of potential between the potential leads of the strip AB .
The potentiometer , shown in diagram at P , Q , K , consisted of 50 one-ohm coils in series with a one-ohm Paul-Harris Fig. 2 .
test wire .
The 50 one-ohm coils , any number of which could be inserted , were part of an old set of B.A. resistances , which had been calibrated in modern units by the National Physical Laboratory .
The potential of the cell used with the potentiometer was determined after each measurement by means of a good Weston voltmeter V , Determination of the Working Temperature of the Strip .
This was done at the close of the measurements as alterations in the connections were required .
The resistance of the strip between the potential leads was first found at the balancing temperature to be 0*600 ohm .
The resistance was then determined in ice ( 0 ' ) , in water at 14'*4 , in vapour of ethyl alcohol ( 78'#4 ) , and of benzene ( 80'*4 ) .
These resistances plotted 1909 .
] Conduction of Heat through Rarefied Gases .
259 against the temperature gave a straight line indicating that the working temperature was 61 ' .
This was confirmed by again determining the resistance in the vapour of chloroform ( 61 ' ) , at which it wTas found to be identical with that found at the balancing temperature .
The rate of dissipation of energy by the portion of the strip between the potential leads is given in watts by E2/ R\gt ; where E is the potential measured across the potential leads after balance has been attained and R = 0*600 ohm .
Throughout , the rate of dissipation is expressed in watts .
Determination of the Irreducible Radiation Loss .
The apparatus was exhausted as completely as possible by the mercury pump , oxygen was admitted and pumped out in the usual way in order to remove the last traces of atmospheric argon , and the residual gas was absorbed by calcium .
The lowest recorded reading of the potentiometer corresponded to 0*049 watt .
The U-tube F ( fig. 1 ) was then cooled in liquid air to condense mercury vapour and prevent it diffusing back from the gauge into the apparatus .
Equilibrium was at once disturbed and the strip tended to rise in temperature .
When balance was regained the new value was 0*0465 watt .
This may be taken as the irreducible radiation loss .
Under working conditions , in which liquid air was not used , 0*05 watt may be taken as nearer the actual value .
Results .
It was quickly found that the heat conducted through the gas still continued to be proportional to the pressure in the apparatus used at pressures large enough to be measured accurately by the gauge , and for most gases the \#166 ; " divided dose " method of measuring pressure was therefore not used .
But for two gases , argon and carbon dioxide , the curve was investigated at the lowest pressures that could be obtained by the divided dose method , and both curves were found to be nearly straight lines .
They need not be separately .given , as the results over a greater range of pressure naturally give the conductivity more accurately .
It is convenient first to arrange the twelve gases investigated in order of their conductivity at ordinary pressures .
With the apparatus employed , it was found that the conductivity of the gas did not sensibly increase with increasing pressure beyond 1*5 mm. for all the gases examined , except helium and hydrogen .
For the latter constancy was reached at about 20 mm. , the increase after 10 mm. being small .
In the following table the limiting value of the energy dissipated at these pressures is given , the gases being arranged in order of their conductivity .
The irreducible radiation loss ( 0*05 wratt ) , has in each case been subtracted:\#151 ; Messrs. F. Soddy and A. J. Berry .
[ Nov. 16 " Hydrogen Watts .
8-75 Carbon monoxide ... .
W atts . . .
1-37 Helium 7-30 Acetylene Methane 2-81 Argon 1-07 Neon 2-35 Cyanogen ... 0-97 Oxygen 1-55 Nitrous oxide . .
0-97 Nitrogen 1-44 Carbon dioxide . .
0-95 The order of conductivity agrees with those published in 0 .
E. Meyer 's " Kinetic Theory of Gases , " so far as they have been before determined .
The results obtained for the twelve gases at lower pressure , up to 0*38 mm.* have been plotted as curves in fig. 3 .
The ordinates represent watts , but to separate the curves different values have been added to the ordinates for different gases .
On the right-hand side of the diagram the symbol of the gas has been placed opposite to the value which has been added to the ordinate* and which must therefore be subtracted from the values shown in reading the curve .
The gases have been arranged in order of their ordinary conductivity as given in the preceding table .
In each case , 0*05 watt has been subtracted from the actual readings to correct for radiation .
The points on the curve have not been corrected in any way , except for the radiation loss , and represent the actual observations .
The departures-from the regular curves are due to fluctuations in the temperature of the water-jacket .
In most of the experiments the temperature of the latter was observed after each reading , and in the calculations of the next section these temperatures have been taken into account .
Over the whole series of experiments , the extreme variation of the temperature of the water-jacket was 50,8\#151 ; from 13 ' to 18'*8\#151 ; which is about 13 per cent , of the difference of temperature between the strip and the water-jacket .
In hydrogen the readings were less steady than in other gases .
Observations at higher pressures than can be included in the figure showed that the curve continued essentially straight up to a pressure of 0T8 mm. with the slope as drawn .
The energy dissipated at this pressure was 21*5 per cent , of the maximum .
From the slope of the curve , departure from the straight line relation would appear to begin at about 0*3 mm. , but the exact point was not determined .
It will be seen , firstly , that the initial part of all the curves are practically straight lines of very different slopes in the different gases , and , secondly* that the slope of the straight part of the curve , which is a measure of its conductivity at low pressure , bears no relation to the ordinary conductivity of the gas at higher pressures .
1909 .
] Conduction of Heat through Rarefied Gases .
.02 04 .
-06 08 -IO -12 14 16 i8 -20 -22 24 -26 -28 30 32 -34 -36 -38 Fig. 3 .
VOL. LXXXIII.\#151 ; A. T Messrs. F. Soddy and A. J. Berry .
[ Nov. 16 , Discussion of the Results .
It is convenient to define the conductivity of a gas at low pressure as the number of calories dissipated by the gas per second per 0*01 mm. pressure per square centimetre of hot surface , per 1 ' difference of temperature between the hot surface and its surroundings , and to express the observed conductivity as deduced from the experimental results by the symbol K , and the calculated conductivity as deduced from the kinetic theory by Q. Thus , for argon it was found when the temperature of the water-jacket was 17 ' , and the difference of temperature between this and the strip therefore 44 ' , that 0T92 watt was dissipated at 0*04 mm. pressure .
The surface of the strip was 20 cm.2 .
Hence 0192x0*24 4 x 20 x 44 1*30 x 10 5 calorie per sec. In the following table the gases are arranged in order of the value of K so calculated .
It is to be understood that no great accuracy is to be attached to the values at present , and that it is probable redeterminations will result in slight changes in the order given 1 .
Argon K( x 10~5 ) . .
1*30 7 .
Nitrous oxide . .
K ( x 10-5 ) .
211 2 .
Neon .
176 8 .
Nitrogen ... .
2-21 3 .
Carbon dioxide .
1-89 9 .
Hydrogen ... .
2-29 4 .
Oxygen . .
191 10 .
Cyanogen 2-35 5 .
Helium .
1*94 11 .
Methane ... .
2-70 6 .
Carbon monoxide. . .
1-96 12 .
Acetylene ... .
2-75 According to the kinetic theory , if the heat interchange between a molecule and the surface it impinges upon is perfect , so that after one impact the molecule leaves on the average with an amount of energy corresponding to the temperature of the surface , the conductivity Q of the gas , as already defined , can readily be approximately calculated .
The assumption will be made that the molecules before impact have a temperature corresponding to that of the water-jacket , which may be taken as 15 ' , since the surface of the enclosing wall is large compared to that of the strip .
The number of impacts per square centimetre per second is 1/ 6 % G , where n is the number of molecules per cubic centimetre of gas at the pressure and temperature considered , namely , 0*01 mm. and 15 ' ; and G is the mean molecular velocity in centimetres per second .
The average amount of heat required to raise one molecule 1 ' is H/ N where H is the molecular 1909 .
] Conduction of Heat through Rarefied Gases .
263 heat of the gas at constant volume in calories , and N the number of molecule in one gramme molecular weight of the gas .
Therefore Q = 1/ 6|HG .
For argon H = 3 , and G ( at 15 ' ) = 4 25 x 104 .
ST is the ratio between the amount of gas in 1 c.c. at 0'01 mm. and 15 ' , and that in 22'22 litres at 0 ' and 760 mm. , and is therefore 5-6 x 10-10 .
Therefore Q for argon = T20 x 10-5 .
The value experimentally found was 1-30 x 10-5 , so that this gas shows a gratifying agreement between theory and experiment , and it may be concluded that the heat interchange in this case is practically perfect .
The values of Q for other gases may be deduced from that of argon , being directly proportional to the molecular heat at constant volume , and inversely proportional to the square root of the molecular weight .
In the following table the values of Q for the gases experimented upon are given in the second column , K being given in the first , and the ratio of K to Q in the last .
The gases are arranged in descending order of the ratio of K to Q. K ( x 10-5 ) .
Q ( x 10~5 ) .
K/ Q. 1 .
Argon 1 *30 1 *20 1 09 2 .
Neon 1-76 1 *70 1 *04 3 .
Nitrogen 2-21 2*35 0*94 4 .
Oxygen 1 -91 2-23 0*86 5 .
Carbon monoxide 1 '96 2 -38 0*82 6 .
Nitrous oxide 2*11 2- 75 0*77 7 .
Acetylene 2 75 3-82 0*72 8 .
Carbon dioxide 1 *89 2*64 0*72 9 .
Methane 2*70 3*95 0-68 10 .
Helium 1*94 3*80 0*51 11 .
Hydrogen 2*29 8*95 0 25 Cyanogen is omitted as we have not found data from which to calculate the molecular heat .
It will be seen from the last column that all the gases , except argon and neon , have conductivities less than the calculated .
The results indicate that whereas for the denser monatomic gases the interchange of energy on impact is perfect , for the lighter gases , helium and hydrogen , this is far from being the case .
The polyatomic gases with great internal energy appear to be relatively less perfectly heated than the monatomic .
The order of the gases in the table is not greatly different from that of the value of Q , indicating generally that the greater the heat interchange between the surface and the molecules impinging upon it , the less perfectly is it effected by the single impact .
This , perhaps , is what is to be expected Conduction of Heat through Rarefied Gases .
if the view , that the heat interchange is imperfect in the case of the majority of gases , is well founded .
These conclusions are at present put forward tentatively .
It is clear that a study of the thermal conductivities of gases at the lowest pressures may be expected to throw light upon the nature of the single impacts of the molecules on a surface .
The view that the heat interchange is imperfect in the majority of cases can readily be tested in a number of ways .
The conductivity should approach more nearly to the theoretical the lower the temperature .
Helium in liquid air , for example , should possess a conductivity approaching the theoretical , since at this temperature its molecular velocity would be not much greater than that of neon at ordinary temperature for which the theoretical conductivity obtains .
Again , if any occlusion , or tendency towards combination , existed between the gas and the surface , the conductivity for heat should be increased .
Thus hydrogen may be expected to give a higher conductivity for heat , when tested with a surface of palladium than with one of platinum .
It is proposed to re-examine the question in the light of these predictions .

