57 A Re-examination of the Light scattered by Gases in respect of Polarisation .
II.\#151 ; Experiments on Helium and * By Lord Rayleigh , F.R.S. , Professor of Physics , Imperial College of Science , South Kensington .
( Received July 9 , 1920 .
) S 1 .
Helium :Method of Keeping Pure .
This gas , owing to its small refractivity , scatters so little light that purity becomes of the first importance .
As is well known , the best method of purification is by cooled charcoal , according to the method of Dewar .
To use it to the best advantage , a convective circulation was maintained which delivered a stream of freshly purified gas into the region where the primary beam traversed it .
A charcoal vessel was used about 2 inches diameter and 8 inches long , containing 100 grm. of coarsely granulated charcoal , a perforated zinc fitting , as shown , prevented the charcoal from dropping out at the bottom .
The connections were as shown in the figure and fitted on to the brass cones of the vessel , fig. 6 of previous paper.f The liquid air was contained in a special vacuum vessel , which had an exit at the bottom as shown .
The vessel was made of silica , as a glass vessel of this shape could hardly be expected to stand the strains which would result from cooling .
The rubber cork at Z was in two halves to make assemblage easier , and a layer of vaseline , melted and poured on the top of it , held the liquid air satisfactorily .
J The weight of the vacuum vessel with the liquid air contained , was taken on the rubber cork , which kept it well pressed home .
As to the remaining supports of the vacuum vessel and the rubber cork , it will suffice to mention that they were carefully arranged to avoid unnecessary constraints .
The cold and heavy gas pouring down off the charcoal passed through a filter , T , of fine wire gauze about 2 inches square , before entering the space where it was traversed by the primary bean .
This was designed to break up and diffuse the stream , and prevent it being localised in a thin trickle .
I he glass work ended in ground female cones fitted to the brass male cones on the scattering vessel , to which they were cemented .
The glass tube was bent as shown to give elasticity .
Its internal diameter was \#166 ; S. inch .
Two 1 1 apei I of which this is a continuation is published in ' Roy .
Soc. Proc. , ' A , vol. 97 , p. 435 ( 1920 ) .
+ ' Roy .
Soc. Proc. , ' A. , vol. 97 , p. 145 ( 1920 ) .
f 1 owe this expedient to Sir Janies Dewar .
Lord Kayleigh .
A Re-examination of the platinum wires , 9 inches apart , were sealed in at W , V , as shown .
* These were to test the purity of the helium by the peculiar character of the electric discharge through this gas at atmospheric pressure .
The helium introduced1 in the first instance was as pure as was easily attainable , but the alternative spark in air was about 2 inches long , and the discharge showed a streaky form , with a rosy colour , probably due to nitrogen contamination .
After the charcoal had been in action for about an hour , the alternative spark had fallen to | inch\#151 ; the least value ever attained , however long the treatment continued .
At the same time the appearance had changed and had become very characteristic .
On the tip of each electrode there was a golden glow due to the D3 line , and between the electrodes there stretched a band of very faintly luminous discharge which swelled out to fill the entire diameter of the tube with a diffuse luminosity .
No trace of thread-like concentration along the axis remained .
'CHARCOAL -VASELINE Light scattered hy Gases in respect of Polarisation .
59 S 2 .
Comparison of Total Intensity with Helium and Air .
In the earlier paper* comparisons were given of the total intensity of the scattered light ( that is to say without distinction as to the direction of polarisation ) from various gases .
The expected relation with the refractivity ( intensity oc ( / x\#151 ; l)2 ) was confirmed .
The case of helium was not dealt with , however , and it appeared very desirable to examine this case , on account of the small refractivity of helium , even when compared with hydrogen .
The method adopted formerly was to use a mercury lamp as a constant source of light and to diaphragm the lens so as to equalise the intensities of scattered light when the apparatus was filled , first with , e.g. , hydrogen , then with air .
In the case of helium it was impracticable to work with the mercury lamp , which is not bright enough for the purpose : for the scattered light from helium may be expected to be only about one-seventieth of that from air .
It was necessary to fall back on the ordinary carbon arc as used for the polarisation experiments .
Unfortunately this source of light is subject to great and uncontrollable variations of brightness .
In preliminary tests with a photo-electric cell , it was found that the random and often sudden variations amounted to at least + 50 per cent , of the mean value , using the hand lamp , and working with given carbons .
As the carbons burn down and new ones have to be used , even greater variations may perhaps be introduced.f An automatic lamp was tried , but was certainly no steadier than the hand lamp .
Under these circumstances no great accuracy in the comparisons could be expected .
It was hoped , however , that the variations would average out to a considerable extent during an hour 's exposure , and that the results obtained might not be altogether without value .
The helium was kept purified throughout the experiments by cooled charcoal , with convective circulation as before .
Since it would have been very inconvenient to change over often from air to helium , a number of plates were exposed for one hour ( with lens at full aperture ) while the apparatus was filled with helium , and were put away undeveloped .
Air was then substituted , and then one hour exposures made with various small diaphragms in the lens , on each plate .
After development , the densities were measured on the photo-electric apparatus , and by interpolation the aperture was found , which would have given with air , the same intensity which the full aperture gave with helium .
* ' Roy .
Soc. Proc. , ' A , vol. 95 , p. 155 ( 1918 ) .
+ These variations are unimportant in the polarisation experiments , as they affect both images equally .
60 Lord Rayleigh .
A Re-examination of the The following table gives the results:\#151 ; Plate No. Helium , 31 '2 mm. aperture .
Air .
Intensity , Helium/ Air .
5 *1 mm. aperture .
4 '07 mm. aperture .
8 '18 mm. aperture .
i 74 65 70 -25 81 0 -0146 2 125 -5 88-5 114 143 0 -0144 3 75-8 66 -5 81 97 0 -0203 4 72 -5 59 75 -25 87 0-0186 The mean value for the intensity ratio helium/ air is 0-0170 .
The theoretical value obtained by comparison of the square of the refractivities is 0-0144 .
Although only one of the plates gives a value as low as this , the experimental evidence for any real discrepancy is not enough to go upon : especially if we take into account that the helium , being colder , and therefore denser , will tend to give too high a value .
This point was overlooked in designing the experiment .
I regard the determinations as approximately confirming the theoretical relation , which , it will be remembered , is itself only approximate , depending as it does on the assumption of spherical symmetry in the scattering particles .
S 3 .
Comparison of Polarisation Helium .
The result obtained before* for the comparative intensity of the two polarisations in helium was 0-42 , a much higher result than for any other gas , and approximating to the value 0-5 which would be appropriate to molecules which behaved as linear oscillators .
This result was wholly erroneous.f I think that the error arose from the previous use of organic vapours in the apparatus .
Although this danger was appreciated at the timej and every attempt was made to remove such vapours by prolonged washing out with a current of air , it seems probable that some traces lingered and gave rise , under the influence of light , to a very sparse but relatively coarse fog , which would scatter very imperfectly polarised light .
It must not be forgotten that the total scattered light from helium is excessively small , and that a fog which would be inappreciable in other cases would here be of importance .
The great advantage of maintaining a current of gas is that any incipient fog is continually swept away .
This is practically achieved by the method of convective circulation now used .
* ' Roy .
Soc. Proc. , 'A , vol. 95 , p. 155 ( 1918 ) .
t The correction was made first in 'Nature , ' December 29 , 1919 .
\ Loc .
tit .
, p. 163 .
Light scattered by Gases in respect of Polarisation .
61 Helium was first tested qualitatively , making use of the double image prism only ( nieol away ) : 4| hours exposure was given .
It was found that the scattered light was , to a first approximation , completely polarised .
A strong image was obtained from the vibrations perpendicular to the incident beam , while nothing was perceptible from the others .
The nicol , set at 277 ' , was then put in , and another exposure of 26 hours was taken .
The same image had still conspicuous intensity , the other being , of course , invisible as without the nicol .
A further exposure of 26 hours was given with the nicol at 271 ' .
The result was still the same , but the stronger image ( perpendicular vibrations ) was now only just strong enough to be-detected , either visually or with the photo-electric apparatus .
The result is then that equilibrium would correspond to a setting of the nicol less than 271 ' ( with the arbitrary zero used ) or , according to the-calibration experiments previously described , the weak image has certainly less than 6*5 per cent , of the intensity of the strong one .
It would be possible , by prolonging the exposure still more , to reduce this limit somewhat , but since no definite weak image has been obtained at all , the goal of obtaining a definite measurement rather than a superior limit is-not in sight .
The difficulty is , of course , due to the extreme faintness of the total scattering from helium , which makes the weak component ( if there is one ) too faint for detection .
It would be of great interest if it could be established that the weak component of helium was altogether absent , or rather , since no experiment could fully prove this , if its intensity could be shown to be less than , say , l/ 1000th of the intensity of the strong component .
But , even if the fact be so , I do not think the experimental proof would be feasible at present .
Even the present much more limited conclusion has not been reached without effort .
S 4 .
Argon .
The old experiments with this gas gave about the same value as air , but my subsequent experiments have led to quite a different result .
I think that , as with helium , in this case also a fog must have been developed by some impurity in the gas.* The argon used was obtained from the Osram lamp works , where it is used for * There is some confirmation for this view .
The argon formerly used , which was , , I believe , of French origin , had been put away in bottles over water .
In the course of the present experiments these bottles were emptied in order to store the new argon , which was known to be freer from nitrogen , and therefore better worth keeping .
A very definite smell of hydrogen sulphide was noticed in emptying the old bottles .
The presence of a trace of this gas would quite account for a fog giving misleading , results .
Lord Rayleigh .
A Re-examination the filling incandescent lamps .
The purity was the best that their plant ( which depends on rectification of liquid air ) could give , the nitrogen-content being stated as 2*5 per cent. The optical determination was made first with the gas as received , and the reading of the nicol for equalising the two polarisations was taken approximately , and was found to be 263'5 ' .
This is considerably lower than had been encountered with any of the common gases , and did not agree with the result of 1918 .
To see whether any condensable impurity was present , the gas was passed over cooled charcoal .
An appreciably lower reading still was now obtained , pointing to the removal of some impurity having an appreciable effect .
Further investigation showed that this was carbon dioxide , which was present to the extent of 1 per cent. The gas was accordingly bubbled slowly through three spiral tubes containing saturated caustic alkali , then dried and passed into the previously exhausted observation vessel .
The measurement of density of several photographs showed that the position of balance was 262*75 ' , the apparatus being in all respects as in the previous work on the common gases .
The lowest angle found in that work was hydrogen 268*47 ' .
It appears , therefore , that argon polarises much more perfectly than any of the common gases .
To deduce the ratio of polarisation corresponding to this setting , the method of calibration described in the previous paper* was not applicable .
The difficulty arises from false light , which with so near an approach to complete polarisation , involves a correction comparable with the whole quantity to be measured .
It appears best , therefore , to fall back on the assumption that the polarising prisms behave ideally , that is , that the double image prism introduces no bias , and that the nicol resolves the vibrations according to the cosine law , without any complication due to reflexion or absorption .
It is certain that the double image prism does introduce bias , as the two images from an unpolarised source are measurably unequal in intensity , but the error thus introduced is probably not very important in the present esse , when , owing to the extreme faintness of what may be called the abnormal component intensity , less accuracy can be looked for than in the common gases .
The position of extinction from an unpolarised source was determined usually as 258*4 ' , thus the angle measured from this , as a zero , which equalises the two images for argon , is 262*75'--258*4 ' = 4*35 ' .
Thus the ratio of intensities is tan2 4*35 ' == 0*00578 .
The weak image is a little over 1 per cent , of the intensity of the strong one .
* ' Roy .
Soc. Proc. , ' A , vol. 97 , p. 448 ( 1920 ) .
Light scattered hy Gases in respect of Polarisation .
63 It is necessary to correct this figure for the nitrogen present .
It would , of course , have been better theoretically to remove the nitrogen , but this would have made the work much more troublesome , and with doubtful advantage , in practice .
The estimate of the makers was 2'5 per cent , nitrogen .
A rather rough determination , which I made by sparking with oxygen , and measuring the loss of volume when the oxygen had been removed , gave 3'7 per cent. For the calculation , 3 per cent , is assumed .
Further , nitrogen is assumed to scatter more light than an equal volume of argon in the ratio 1T0 , the two polarisations for nitrogen being in the ratio 0'0406 .
On this basis , the corrected ratio of the two polarisations for argon becomes 0'0046 , less than half of 1 per cent. This figure is much lower than that found for any of the other gases , and shows that the monatomic argon molecule approximates more closely to behaving like a dielectric sphere .
Although the residual polarisation is very small , and its value seriously dependent on the amount of nitrogen impurity , which is not easy to measure , yet I do not think it can be explained away .
It is hoped that the results here given will be of value in the future for testing theories of molecular and atomic structure , when these are sufficiently developed for the test to be applied .
The lack of complete polarisation indicates that the molecules are not of spherical symmetry , but that they have certain preferential deviations of vibration .
In general , there will be three principal axes to the particle , with corresponding coefficients of radiation , but the experimental data do not afford material for comparing these coefficients in the general case .
If we assume the molecule to be symmetrical about an axis , there are then only two coefficients to be compared , and the ratio of these can be calculated from the ratio of intensities of of the two polarisations , subject to an ambiguity dependent upon whether the coefficient radiation along the axis of symmetry is taken as a maximum or a minimum\#151 ; whether we take the molecule as prolate or oblate .
The calculation in this case , as well as in the more general one , was given by my father.* It is probable , however , that the present data will be more advantageously used in the converse way , starting with some hypothesis as to the constitution of the atom and molecule which would allow a value of the polarisation to be calculated for comparison with the experimental result .
* * Phil. Mag. , ' vol. 35 , p. 373 ( 1918 ) .
64 Dr. W. F. Sheppard .
Tor convenience , the results of this paper and its predecessor are here recapitulated .
Gas .
Weak component polarisation ( strong component = 100 ) . .
Argon 0-46 Hydrogen 3*83 Nitrogen 4-06 Air 5 -0 Oxygen 9-4 Carbon dioxide 11 -7 Nitrous oxide 15 -4 Helium Not definitely determined , but \lt ; 6-5 The total intensities of light scattered by helium and by air have been* compared .
The ratio found was 0-0170 .
Considering the difficulty of the experiment , this is considered to agree within the limit of error with the ratio of the squares of refractivities , which is 0-0144 .
Reduction of Error by Linear Compounding .
By W. F. Sheppard , Sc. D. , LL. M. ( Communicated by Prof. E. T. Whittaker , F.R.S.\#151 ; Received April 23 , 1920 .
) ( Abstract .
) If U\ , u2 , us , ... are observed values of a single quantity U\#151 ; if , in other-words , there are observed values of U containing errors u\\#151 ; U , \#151 ; U , us\#151 ; U , ... \#151 ; , we may take as an approximate value of U any function of then 's which , if each of the us in it were replaced by U , would itself become identically U. Usually this function is a linear compound ( linear function ) of the u 's , and is called a weighted mean : and we regard as the best , weighted mean that in which the constants of the function are chosen so that it shall have the least possible mean square of error .
The principle-applies also when the w 's , instead of being observations of a single quantity U , are observations of a set of quantities Ui , Us , U3 , ... ; with the modification* that the effect of replacing ( say ) ua by a function of the s will be that , if the U8 in the function were put equal to the corresponding U 's , the function ; would not become identically equal to Ui , but only approximately equal to^

