]\gt ; The Thermal Conductivity of Air .
By E. O. HERCUS , M.Sc .
, New Zealand Government Research Scholar , and T. H. LABY , M.A. , Professor of Natural Philosophy , University of Melbourne .
( Communicated by Sir Joseph Larmor , F.R.S. Received August 21 , 1918 .
) Introduction .
The thermal conductivity , , of a gas is connected with its specific heat , , and its viscosity , , by the relation .
The numerical constant , in this relation can be deduced theoretically , and various values , which are iven later , from to have been obtained according to the law of force assumed to act in a molecular collision .
Which of these laws of force holds is determined when , and are accurately known for a number of ases .
A large number of determinations of both the viscosity and specific heat of each of the more simply constituted gases have been made that are in reasonable eement with one another , but , compared with these , the independent values of the thermal conductivities are fewer , in poorer agreement , and the methods of determination are more open to criticism both in conception and performance .
Eucken has dete1mined relative to air the conductivity of a number of gases ; the accuracy of his experiments 1nade a new absolute determination of for air desirable .
The investigation to be described was an attempt to supply an absolute value of the thermal conductivity of air by a method which would not be open to the objections which can be brought against all the methods previously used , excepting that of one observer .
Heat in general is carried through a gas by conduction , radiation , and convection .
In all past determinations of the thermal conductivity of gases , except those of Todd , the flow of heat by convection was present to an uncertain amount ; in the experiments to ) described it was absent .
Previous Deternainations .
A number of determinations of the thermal conductivity of gases have been made .
* absolute experiments , with one exception , have been Stefan , 'Wien .
, vol. 65 , p. 45 ( 1872 ) : vol. 72 , p. 69 ( 1876 ) .
Kundt and Warburg , 'Pogg .
Ann vol. 156 , p. 177 ( 1875 ) .
Winkelmann , ' Pogg .
Ann vol. 153 , p. 481 ( 1874 ) ; vol. 156 , p. 497 ( 1876 ) ; 'Wied .
Ann vol. 14 , p. 534 ( 1881 ) ; vol. 29 , p. 68 ( 1886 ) ; vol. 44 , pp. 177 , 429 ( 1891 ) ; vol. 48 , p. 180 ( 1893 ) .
Christiansen , ' Wied .
Ann. ' vol. 14 , p. 23 ( 1881 ) ; vol. 19 , p. 257 ( 1883 ) .
Gratz , ' Wied .
Ann vol. 14 , pp. 232 and .
Schleiermacher , 'Wied .
Ann vol. 34 , p. 623 ( 1888 ) .
Muller , 'Wied .
The Thermal Conductivity of .
191 modifications of two principal methods , namely , the cooling thermometer method of Kundt and Warburg , or the electrically heated wire method of Andrews .
In the former method , a thermometer loses heat through the gas by conduction , convection , radiation , and also by conduction along the thermometer stem .
It has usually been assumed that , with he gas at the lowest obtainable pressure , the thermometer only loses heat by radiation and conduction along the stem , and that , at a few millimetres pressure , the loss of heat is by radiation , stem conduction , and gas conduction , but not by convection .
Insufficient attention has been given to the effect of convection currents .
Winkelmann , Gratz , Muller , Compan , and Eckerlein ( lo cit. ) ( Eckerlein ) to cal cm .
sec. found by Miiller .
Conduction of Heat through Gases .
Fundamental Assumption.\mdash ; It is worthy of notice that , in the measurement of the thermal conductivity of a , Fourier 's fundamental assumption , that the transfer of heat across a thin lamina of gas per unit area per second is prop.ortio.emperature gradient.lways badeA)panThat i the relation is assumed .
This has not been proved experimentally , nor is it a consequence of the kinetic theory of .
Experiment has shown that the conductivity of a gas depends upon its temperature , and Rendus .
varze , 'Ann .
phys vol 11 , pp. 303 and 1144 1903 ) .
Todd , ' Roy .
Soc. Proc , vol. 83 , p. 19 ( 1909 ) .
Pauli , ' Ann. .
Phys vol. 23 , p. 907 ( 1907 ) .
Stafford , 'Zeit .
Phys. Chem .
Eucken , ' Phys. Zeit vol. 12 , p. ; vol. 14 , p. 324(1913 ) .
Andrews , ' Proc. Roy .
Irish Acad vol. 1 , p. 465 ( 1840 ) .
192 Mr. E. O. Hercus and Prof T. H. Laby .
it is possible that part of this dependence is a spurious effect due to the failure of the above assumption .
In other words , putting , it has always been assumed that the temperature gradient enters linearly in , but this is difficult to prove experimentally , as it is in practice impossible to use a lamina , and a variation of involves in general a variation of vection Currents .
Convection currents must be present in any mass of gas through which heat is passing , if the gas is not in the form of a horizontal layer with its upper surface at the highest temperature .
Now , in the two principal methods of measuring gaseous conductivity , the cooling thermometer and hot-wire method , this condition is not satisfied , and convection currents must be present .
Such mass motions due to the action of gravity are reduced , however , at low pressures , and , in the application of the methods mention ed , the gases have usually been at less than one atmosphere pressure ; pressures such as 150 and 20 mm. have been used .
Tests have shown that the convective passage of heat is thereby very much reduced .
Oberbeck , *making certain simplifying assumptions , has given the theory of heat convection in a spherical shell of rarefied gas .
He that the heat transferred is proportional to the fourth power of the density of the gas .
The practice of eliminating convection by reduction of pressure is consequently to some degree justified , but it has limited most of the values of found to gases in a somewhat rarefied state .
Variation of onductivity with Pressure.\mdash ; Within limits , the conductivity of a gas is , according to the dynamic theory , independent of pressure .
The passage of heat from a warm layer to an adjacent colder layer is , according to this theory , cted by the passage of molecules possessing a largeamount of kinetic from the first layer to the second , the passage of molecules in the opposite direction not compensating this flow of energy .
A reduction .
pressure , while reducing the number of molecules per unit volume , increases their free path , and consequently the conductivity is independent of pressure .
There is no very exact experimental verification of this .
So soon as the molecular free path is comparable in magnitude with the thickness of the gas layer , conductivity depends on the pressure .
Soddy and and have shown that , at pressures within a certain range , the conductivity of a gas is proportional to the pressure .
Coeffic nt of Temperature.\mdash ; The variation of the thermal * Oberbeck , ' Wied .
Ann vol. 7 , p. 271 ( 1879 ) .
Soddy and Berry , ' Roy .
Soc. Proc , vol. 83 , p. 254 ( 1909 ) .
Knudsen , 'Ann . .
Phys vol. 34 , p. 593 ( 1911 ) ; see also Dunoyer , 'Les Idees Modernes sir la Constitution de la ' Gauthier-Villars , p. 253 ( 1913 ) .
The Conductivity of Air .
conductivity of gases with temperature has been studied , from both the experimental and theoretical aspect , by a number of observers .
The results obtained , previous to the experiments of Eucken , *appear untrustworthy .
Most of the earlier observers expressed their results for the change of with temperature by the relation .
The values found for differ rather widely , varying for air from to On the theoretical side , Maxwell at first found to be proportional to the absolute temperature , and Clausius that it varied as .
The former gives , the latter .
The existence of the relation shows that temperature may be expected to have the same effect on as on , since it is to be expected that and are independent of temperature .
The observed values of , and of the ratio of the specific heats , and consequently of , vary slightly with temperature .
Eucken has verified for five gases that is constant over the range to 37 K. These considerations allow us to apply to thermal conductivity Sutherland 's relation connecting viscosity and temperature , which is known to hold over a wide range of conditions .
We have , then , that Accepting for air , then for air between and Principle of the Bxperiment .
The ideal conditions of the experiment were essentially as follows .
The layer of gas under observation is confined between two horizontal circular copper discs and C. Heat passes through the gas by conduction and FIe .
1 .
radiation from the surface , at temperature , to the surface , at the temperature .
Convection currents are absent , since , at every point between and , the density of the gas is constant in a horizontalplane , and decreases upwards .
The disc is surrounded by the guard-ring which is at the same temperature as B. In order to prevent the passage of * Eucken , ' Phys. Zeit vol. 12 , p. ; vol. 14 , p. 324 ( 1913 ) .
194 Mr. E. O. Hercus and Prof. T. H. Laby .
heat to or from the upper surface of the disc , a third disc A was placed above and parallel to it .
A was kept at the same temperature as B. , and were heated electrically .
was cooled with water .
We have , then , that the electric energy supplied per second to the middle disc , is equal to the heat which it loses to per unit time by conduction , , and by radiation , , plus the heat , , absorbed per unit time by its temperature is not stationary , and the heat which to and 1 ) , , that is , where is the potential difference between the terminals of the " " heater its resistance , the conductivity of the gas , the temperature gradient nornd to in the gas layer , the area of , and its capacity for heat , its increase of temperature per unit time .
These quantities must be measured , of course , in terms of units belonging to a consistent system .
Actually the electrical measurements were a little different from this .
coil of resistance was put in series with the heater embedded in the central plate , and the drop of potential , , on this coil was measured .
We have , then , in the usual unlts , where is the distance between the plates and C. All temperature measurements were made by thermo-junctions formed by soldering constantan wire to the copper plates and guard-ring at suitable points .
These junctions were standardised by means of a platinum thermometer .
Apparatus .
The three plates , ( see fig. 2 ) of thickness 1 cm .
and of 30 cm .
diameter ) were mounted horizontally , and held together by three thin steel bolts passing through the plates and glass distance pieces .
The lower plate rested on a water circulator , and the top surface of the upper plate was insulated by a wooden covering filled with cotton wool .
Each plate was made up of two 12 cm .
copper sheets , screwed together after the insertion of the heating and thermo-couple wires .
The joints were made water-tight with a mixture of red lead and oil .
The central disc of the middle plate was supported from the guard-ring by three -shaped ivory buttons accurately turned to fit the plates .
The surfaces of and which faced each other were carefully ground ; all the surfaces were silver plated and highly polished .
For the temperature measurements silk-covered constantan wires were The Thermal Conductivity of Air .
led into the centre and bottom plates through grooves in the lower halves , and insulated from the plates by wax of high melting point .
In the case of the top plate the constantan wires were led in directly through ebonite standards packed with red lead and oil mixture , and were soldered to the lower half of the plate .
The heating wires , which were of No. 30 manganin , insulated with four layers of silk , were laid in spiral grooves cut in the lower halves of the discs , the grooves being then filled up with paraffin wax .
To neutralise the extra cooling at the edges , the spiral grooves were slightly concentrated near there .
The use of paraffin insulation imposed on the apparatus a upper temperature of C. The water circulator was made up , as represented in fig. 2 , from a zinc casting with a spiral projection to which was screwed the brass sheet forming the base .
The water supply was led in from the mains , and , in order to damp minor temperature fluctuations , passed through a coil in a thermally insulated water tank , thence through a lagged pipe to the apparatus .
The apparatus was made by a rin of stout rubber , bound to the upper plate and to the circulator by steel tapes tightened by bolts .
The wires from the two lower plates were led out through rubber tubes vulcanised to the rubber ring , and the tubes were tightly bouud with cord .
Leaks through the upper plate bolt-holes were stopped by a beeswax and resin mixture sucked in by diminished pressure .
Once the constants of the apparatus had been determined , the measurements required in the course of an experiment were all potential measurements .
As it was important that these readings should be taken quickly and without handling the wires , a switch was designed by which any of the Mr. E. O. Hercus and Prof. T. H. Laby .
desired connections to the potentiometer might be made .
The accompa.nying circuit plan shows the wiring ngements .
Points with ' FIG. 3.\mdash ; Diagram of thermo-junction circuits .
Points with the same letters are connected , e.g. , to .
Con constantalL Cucopper .
corresponding plain and suffixed letters were so attached to the switch that they could be simultaneously connected to the potentiometer terminals .
Measurements .
Accuracy Required.\mdash ; The accuracy required in measuring each quantity entering into the expression for in order that shall be correct to 1 in 500 , is readily found from the relation , The resistances and , and the electrical pressure , applied to are readily measured to 1 in 1000 .
The Thermal Conductivity of Air .
where and are the diameters of the guard-ring and middle plate .
To attain the accuracy mentioned , and are required to be known to cm .
The average distance between the middle and lower plates , , is required to cm .
The radiation forms 4 per cent. of the heat crossing the gas , and so it needs to be known to within 5 per cent. The terms and can be mad .
small by careful temperature adjustments , and need only be determined to 1 in 5 .
In the temperature measurements , which were made with thermo-couples calibrated with a platinum thermometer , the E.M.F. of the couples requires to be known to about 1 microvolt and the platinum temperatures to O. The main difficulty in attaining accuracy in , then , is to so control the temperatures as to realise the conditions assumed in the theory , and to measure with sufficient accuracy the temperatures so realised .
Distance between Plates.\mdash ; If the plates were perfect planes , their distance would be the common thickness of the glass pieces .
As the plates were not perfectly flat they were " " contoured\ldquo ; by replacing the glass by longer steel distance pieces and then measuring with an end micrometer the distance between the plates at a number of points .
The difference between the mean of these measurements and the mean length of the steel pieces was then applied as a colTection to the thickness of the glass pieces .
The glass pieces were carefully ground to the same thickness , and measured with three screw micrometers ; the mean thickness was found to be cm .
A simple deyice was hit upon which enabled the mean length of the steel pieces , under the conditions in which they were used , to be found with accuracy .
Two sheets of plate glass were set up with the steel cylinders separating them , and readings were taken with an end micrometer ( kept normal to the glass plates ) in a small circle round each distance piece .
The results of measurements made on three different days gave for the mean length of the three cylinders , taking a 3-inch spherical-ended rod as standard : ; mean of results , cm .
Contouring of Plates.\mdash ; To determine the contour correction , th.ree circles of radii 3 , 7 , and 11 cm .
were drawn on the plate intersecting eight radial lines at intervals , giving 24 points of intersection in all .
With the steel pieces between the plates , end-micrometer readings , taken at the 8 points on each circle , gave means for an annular strip , and these means were weighted proportionally to the area they represented .
Four contours of this character were made after the conductivity experiments were completed .
The mean distance between the plates when separated by the steel distance pieces was found to be : cm .
; weighted mean Mr. E. O. Hercus and Prof. T. H. Laby .
cm .
Hence msan distance between plates cm .
Diameter of Central Disc.\mdash ; The diameter of the central disc was carefully measured by Mr. H. S. Tily .
Brass pieces with engraved lines were fixed in the lathe , and the distance between the lines was measured by means of a travelling microscope , ( a ) with the ends of the in contact , ( b ) with the ends separated by a diameter of the disc .
The microscope readings were checked by reference to an invar scale .
Readings three diameters separated by 12 gave the following:\mdash ; ; mean , cm .
The effective area of the disc was taken as extending half across the airgap between the disc and the guard-ring .
The width of the air-gap was cm .
, consequently the effective area , , was sq .
cm .
jmperature Measurements .
Calibration of -couples.\mdash ; The temperature measurements , as already mentioned , were made by copper-constantan thermo-elements because constantan gives a higher temperature E.M.F. with copper than any other convenient material .
The constantan wire used was all from the same batch , nevertheless it was considered necessary to calibrate all the junctions independently .
The results obtained showed the junctions to be remarkably similar in their action .
The calibration was carried out by comparison with a platinum resistance thermometer , the copper plates and thermometer being immersed for this purpose in a thermostat , provided with heating coil and motor-driven stirrer .
The constantan wires were protected from the water by narrow rubber tubing .
Platinum Thermometry.\mdash ; The thermometer used was supplied by the ambridge Scientific Instrument Company , and had a fundamental interval of 2 ohms .
It was provided , in lieu of terminals , with fusible alloy cups , into which !
the leads were fixed .
These were of No. 19 copper ( 250 cm .
instead of the usual " " flexible\ldquo ; wire , in order to prevent variations due to broken strands .
The Callendar-Griffiths bridge was provided with manganin coils on the binary scale , viz. , , .
ohms .
The coils were provided with Collins ' mercury contacts , which are Snverted copper 's pressed by springs int mercury cups .
The bridge wire had an effective resistance of ohm per cm .
, and it could be read by a vernier to ohm , which was equivalent to C. The Cond.uctivity of Air .
A Callendar-Griffiths key and a reversing commutator in the circuit were used ; both of these were enclosed , to further eliminate thermo- electric effects , in a lead-lined box .
The coils and bridge wire were calibrated by Mr. Alexander , M.Sc .
, in terms of the -ohm coil , and a second calibration agreed with his .
The resistance of the platinum thermometer was thus found in terms of the -ohm coil at C. , the temperature coefficient of the nanganin coils having been found to be during the course of this calibration .
Callendar 's equation , for correction of platinum temperatures to the air scale was used .
in this equation is determinable by a reading at a third lixed point in addition to the ice and steam points , but the Cambridge Scientific Instrument Co. 's certificate gives the value for the batch of platinum wire of which the thermometer is made as with a variation probably not exceeding 1 per cent. As this variation would introduce at most an error of in our temperature measurements , the value quoted was taken as correct , and the ice and steam points were alone re-determined .
Ice points were determined by imbedding the thermometer in crushed distilled-water ice in a Dewar flask , and the extreme values obtained agreed to C. Steam-point obseryations did not agree so well , partly .
to uncertainty of the accuracy of the brass scale of the barometer available .
But since no observations were to be taken at a higher temperature than C. , an error of in the steam point would not introduce an error of more than C. in the actual observations , and the agreement of observations taken at different times was well within this limit , as was also the error due to the barometer .
Thernw-couple Measurements.\mdash ; These were carried out with a dial pattern potentiometer , by Wolff , of 15,000 ohms resistance , the coils in the last dial being ohm , so that the instrument without a multiplier read to 10 microvolts .
An alteration of ohm gave a galvanometer deflection of about 100 scale divisions , so that one-tenth of a microvolt could be estimated , but in order to prevent error due to lack of proportionality in scale deflections , the device was adopted of altering the external resistance of about 5000 ohms in series with the potentiometer until the deflection was reduced to zero , and then , 1etaining this external resistance constant , a new balance with the standard cell was taken .
Thus if the true value for the standard cell was volt at the temperature of the experiment , and the readings on the potentiometer of the thermal E.M.F. and of the standard cell were respectively Mr. E. O. Hercus.and Prof. T. H. Laby .
80 and 103035 , the thermal E.M.F. was 10/ 103035 or 786 microvolt .
To facilitate the plotting of results , the E.M.F. 's of the thermo-junctions were determined at definite temperatures , namely , at C. , C. , C. In this calibration accuracy of C. , or microvolt , was aimed at ; but it was found that differences in results obtained at different times amounted sometimes to 2 or 3 microvolts .
Variable brush contact resistance is a possible cause of this , for though the total brush resistance of ohm would be negligible for measurements to 10 microvolts , for which the potentiometer was constructed , this resistance introduces appreciable error when finer measurements are attempted .
A low resistance potentiometer was not available to us .
A second possible reason for the discrepancy is lack of uniformity of the thermostat temperature , in spite of the vigorous stirring applied .
It was found that on account of the smallness of the temperature range the .
temperature curve for the couples departed only slightly from the straight line which represented 40 microvolts per 1o C. The differences from the straight line values were plotted for each couple .
Resistance of Central Disc Heater.\mdash ; The resistance of the central disc heater was obtamed connecting it in series with a standard 10-ohm resistance , and comparing the potential drops on the heater and on the 10-ohm coil .
This gave the resistance of the heater ohms .
On allowing for the resistance , ohm , of that part of the leads which was outside the centre plate , and therefore ineffective in heating it , we obtain an effective heater resistance of ohms .
Radiation.\mdash ; The heat carried by radiation has generally been determined by evacuation , or by using two layers of gas of different thicknesses .
Neither of these methods could be used conveniently in our apparatus .
As preliminary to calculating the radiation from a silver surface,4 we may calculate the radiation , from the central to the lower plate , assuming both to be full radiators , and find the ratio of this heat stream to that transferred by conduction .
We have and where denotes the rate of loss of heat by conduction , the area , and the absolute temperature of the central plate , the absolute temperature of the lower plate .
o ergs cm .
sec. cal .
cm .
sec. Mean of nine published values .
The Thermal Conductivity of Air .
c sec. cm . .
abs .
, .
abs .
From which we have To the accuracy sought , and for the range of values of and used in this research , this ratio may be taken as independent of and .
It was found , as set out later , that the value of with silvered surfaces is only 1/ 33 of this value for a full radiator and receiver , consequently radiation forms of the heat the gas , and it is unnecessary for a final accuracy of 1 in 500 in to determine for silver more closely than to within 4 per cent. The radiation from a surface of silver was found by experiments carried out on Dewar flasks with walls of silvered glass .
For bodies which are not full radiators no rigid formula has been given for their total radiation as a function of their temperature .
In the case of platinum , it has been shown that its radiation is fairly represented by substituting for the fourth power of the temperature a somewhat higher power .
Assuming a similar radiation law for a body of mass , of specific heat with a polished silver surface of area , we have for the rate of cooling , , that where is the temperature of the body , and that of a silvered surface surrounding it , and is a constant for that surface .
A spherical Dewar flask was used filled with water .
Let be the radius of its inner spherical surface .
Since the density and specific heat of water may each be taken as 1 in the umits used , , and ecting for the moment the water equivalent of the silvered glass inner vessel , therefore : ( 1 ) It follows that for the same values of and different flasks should show that is inversely proportional to .
The quantity gives the efficiency of a spherical flask .
The best of four flasks was found by this test and used .
Its capacity was 2010 cub. cm .
To avoid convection and conduction up the neck of the flask , observations were taken with the inside wall at a lower temperature than the outside wall and surrounding bath .
When the inside of the flask was hotter than the outside , half of the heat lost was lost by convection currents in the neck of the flask .
Two sets of observations were taken , one with the outside at room temperature , the other with it at C. In equation ( l ) both and are unknown .
It was is calculated by equation ( 1 ) , assuming .
The approximate constancy of the numbers in the last column justifies this assumption .
We may take .
sec. cm .
which is 3 per cent. of for a full radiator .
We have then that , or that Thus of the heat lost by the central to the lower disc per cent. is rffiiated , and the remainder conducted through the gas .
Calculation of Results .
In the relation ( see p. 194 ) , the terms and , which express the heat used in changing the temperature of the central disc and the heaG lost by it to the guard-ring and upper disc , were made as small as possible in all our experiments .
These terms did not affect the value of to more than about 1 per cent. In the numerical calculation of from the observations , it is convenient to treat these terms as corrections .
Neglecting them for the moment , we have ohms , watt sec. per cal .
, ohms at C. , cm .
cm.2 .
between the terminals of the resistance : The small corrections to the above expression are as follows:\mdash ; -ring Correction.\mdash ; An allowance is necessary for the heat conducted across the air-gap from the guard-ring to the central disc .
If denote the mean excess of temperature of the guard-ring over the central disc , the heat , , conducted per second across the is yiven roughly by The heat developed in the central disc is calories per Putting the effect of in the form of a correction , to be added to , we have , therefore Top-plate Correction.\mdash ; In order that it should be possible to calculate the heat passing between the top and central plates , the top plate was kept slightly above the temperature of the central one : this conduction without convection .
Since the distances of the top and bottom plates from the centre one are the same to 1 in 30 , the correction to be dded for conduction from the top plate to the central disc is given by , where is the mean excess of temperature of the top plate over the centre plate .
rrution for Thermal apacity , \mdash ; If the temperature of the centre plate during an experiment , a correction must be applied for heat it thereby loses or gains .
From the volume , density , and specific heat of the central disc its heat capacity , , was calculated to be 340 calories per joules per degree .
This value for the thermal capacity was checked by heating the central plate electrically at such a rate that the heat . .
Mr. E. O. Hercus and Prof. T. H. Laby .
lost was small .
The observed rise of temperature and energy used gave the thermal capacity as 1480 joules per degree , which is in satisfactory agreement with the calculated value for the present purpose .
This correction to is Details of an Experiment .
The current in the central heater was adjusted to give a temperature difference between the centre and bottom plates of , and the guard-ringand top-plate heater circuits were adjusted by trial until these plates .
remained at the same temperature as the central plate .
The currents were left on for 12 hours before observations were taken , in order that the peratures should have reached a steady state .
The readings occupied minutes , in which time it was found possible to obtain six readings of each of the four thermo-junctions ; this involved some 160 instrumental readings .
The following is an abstract of the observations for one experi- meant : jriment 7 Table I. *This correction arises because the thermo.couples on calibration were not exactly 40 microvolts per degree .
The of Air .
The first six experiments showed a larger departure from the mean than to be expected .
In trying to discover the reason for the irregularity , it : was noticed that a brass peg in the top plate ( see fig. 3 ) and in a thermojunction circuit would cause error in the result if its ends were at appreciably different temperatures , as a small thermo-electric effect not allowed it for in the calculations would be introduced .
A direct thermo-couple I measurement showed a difference of temperature of as much as C. under : the conditions of the experiments .
A correction to for this of is made in these six experiments .
Two more observations were taken with a I$ short copper screw substituted for the brass , and the results so obtained are given double weight .
The following values of for dry air corrected to S C. were obtained .
The mean temperature of the air anged from to C. :\mdash ; Table II .
Mean 5 .
cm .
sec. , which gives , if we take the temperature coefficient of as .
cm .
The Most Probable Value of The value we obtain for the conductivity of air may be compal'ed with that obtained by other observers , and it will be convenient to deduce a value of from all the observations .
In deriving a mean value we have given greater weight to our own and Todd 's result , both because they are the only experiments in which convection currents are absent , and because the apparatus constants in these experiments.could be accurately determined .
The other observations are so weighted that cooling methods and the hot-wire methods have nearly equal weight in the final mean .
The mean of the 14 observations is .
sec. and the ayerage departure from the mean is , or 7 per cent. , showing that the agreement between different observers is very poor .
Value of Having obtained the absolute conductivity of air , we can use the relative determinations of the conductivity of gases recently made by Eucken , and the other data available , to test the relation , and find the absolute value of .
This is done in Table .
We are not acquainted with any paper in which all the observations available for this purpose have been used .
The values of , and givenin Landolt and Bornstein , Tabellen , in the ' Beibliitter 'An den Annalen der Physik ' ( 1904 to 1911 ) , in ' Donnees Numeriques , ' and in recent papers have been used .
As the different determinations are ot of equal value , weighted means are taken in certain cases .
The number of determinations used in deriving the mean is stated in the columns in Table following the values of each quantity .
Thermal Conductivity , \mdash ; The observations were weighted on the same The Conductivity of principle as is set out above for air .
is reduced to C. by use of the relation and recorded values of derived from viscosity experiments .
specific Heat .
Different observers of this constant in agreement and their results are equally weighted in deriving a mean .
\mdash ; Equal weight ; was given to the few determinations of that have been made .
The methods which have been used in measuring the ratio of specific heats are certainly not of equal accuracy .
A weighted mean was derived and it was found that ' different experimental methods yield the same value for .
It is the most accurately known of the constants we are now concerned with .
Vucosity , large number of observations of viscosity are available , but most of them unfortunately have been made by the capillary tube method , which yields discordant results .
The 31 values of for air were so weighted as to attach equal importance in the result to different methods .
A simple mean was taken in the case of other gases .
Table IV .
( a ) The number given after each constant is the number of values of ( b ) The theoretical value 5/ 3 is used in preference to the one obse and Geiger .
Todd 's value is not included as the accuracy of the reduotion doubtful .
Chapman* has recently concluded that for a gas constituted of spherically symmetrical molecules .
Maxwell 's result is a special case of Chapman , ' Phil. Trans , vol. 211 , p. 433(1911 ) . .
E. O. Hercus and Prof. T. H. Laby .
Chapman 's more general theorem , since the inverse fifth-power law gives , of course , spherically symmetricalmolecules .
The two facts that for a monatomic gas , but that the variation of viscosity with temperature is not in accord with Maxwell 's predictions , based on the inverse fifth-power law , show , if we accept Chapman 's conclusion , that the monatomic gases are spherically symmetrical , but in some way other than as a result of a repulsion between them varying as the inverse fifth power of the distance .
Chapman 's conclusion is contrary to those of Meyer and Clausius , for the former concluded that , for elastic spheres , , and the latter that The monatomic gases , helium , argon , and mercury*vapour , which we may reasonably assume have spherically symmetrical molecules , are consequently constituted of molecules such as are assumed in these theories .
In 1903 Schwarze found for helium and for argon , which as taken to imply that the law of force between the molecules varied as the inverse fifth power of the distance , but a large body of evidence on the effect of temperature on viscosity is contrary to the inverse fifth-power law .
The observed value of for the polyatomic gases throws no on the mechanism of molecular collision in the absence of a value of for such gases theoretically deduced .
The observed value for the diatomic gases ( excluding chlorine ) is 7/ 4 .
Chapman states that his theory does not apply to polyatomic gases , " " this is not to say that is not equal to 5/ 2 for any polyatomic gas that his ' theory leaves the question perfectly open gave a theory of conduction in gases , which was partly empirical , partly based on the inverse fifth-power law , and which took into account the internal energy of the molecule .
He found which gives for helium and argon , and when as it is for the diatomic gases , and for carbon dioxide , where The agreement with experiment is very poor .
Nevertheless when is plotted against , as is done in fig. 4 , the two quantities appear to be linearly related .
The straight line in fig. 4 is drawn through the point ( the theoretical value for argon and helium ) , and close to the point representing and for carbon dioxide , We have nct included mercury in Table IV , nor do we discuss it further , because the only observation of the viscosity of mercury ( by Koch ) appears to be inaccurate , as its variation with temperature is anomalous .
Ann. .
Physik , ' vol. 11 , p. 303 ( 1903 ) .
Boltzmann , ' Pogg .
Ann vol. 157 , p. 457 ( 1876 ) .
The Thermal Conductivity gas for which both are known with high of the line FIG. 4 .
Abscissae Ordinates It will be noticed that the points for the six gases , air , NO , and CO fall on this straight line , and fall near it .
is not near it .
But it is to be noticed that in the cass of and Dr. C. Chree .
Electric Potential Gradient and only determination of is by Eucken , that eight of his values of are low compared to those of other observers .
If the values of we have used for , and are correct , the direction of the displacement of the oints for these gases from the straight line would arise from using a low value of in deriving In concIusion , we wish to express our indebtedness to Mr. M. Alexander , M.Sc .
, who carried out part of the preliminary work , and calibrated some of the instruments used ; to Mr. H. S. Tily , who took the measurements of the discs ; and to Mr. H. E. Strawbridge , whose skilful execution of the mechanical work made the success of the experiments possible .
Grants for the apparatus were received from the New Zealand Government Research Scholarship Fund .
The experiments recorded were completed in 1914 .
Electric Potential Gradient Atmospheric Opacity at Observatory .
By C. CHREE , Sc. D. , LL. D. , F.R.S. ( Received October 19 , 1918 .
) S1 .
It has long been known at Kew Observatory that thick fog is generally accompanied by a high value of the electrical potential gradient .
During fog the oscillations of potential , which are common even in the electrically quietest of days , are sometimes very large ; but whether they are exceptionally large , when expressed as a percentage of the mean gradient for the hour , has not been investigated .
There is this peculiarity , however , that sometimes during fog the gradient oscillates between a positive and a negative value , a phenomenon which is quite exceptional except during rain or a thunderstorm .
It occurred to me to examine the records for a series of years to see whether any information could be derived as to the existence or formation of fog .
The appearance of the electrogram during heavy is usually so characteristic that from it alone one could say with a close approach to certainty that fog existed .
Thus electrograms may be of considerable value as confirmatory evidence of fog when direct evidence is lacking .
But the * Eucken 's observations were relative to air , for which he assumes We have reduced his observations to the basis of

