]\gt ; PROCEEDINGS OF THE ROYAL SOCIETY .
SECTION ]ATEEMA TAND PE YSICAL SCIE the Torqne produced by a Beam of Liqht Oblique Refraction through Glass By BAItLOW , D.Sc .
( Commumicated ) Prof. J. H. Pointing , F.R.S. Received March 27 , \mdash ; Read Apri125 , 1912 .
) In the Presidential Address to the Physical Society in 1905* Prof. Pointing developed the idea that a beam of light must be yarded as containing a stream of momentum , and he showed that this principle may be used to solve with great ease the various cases in which a beam of light is absorbed , reflected , or refracted at a surface .
This method is particularly powerful and convenient , for , once havin made the assumption that light carries with it a stream of momentum , the forces can be calculated by the ordinary mechanical laws without any further appeal to the theory of wave disturbance of which we suppose light to consist .
When a beam falls normally on a perfectly absorbing surface the force exerted is the momentum received per second , and it is therefore equal to the momentum in a length of the beam , where is the velocity of light .
But the pressure exerted is equal to the energy density of the beam\mdash ; the law first deduced by Maxwell from his electromagnetic theory and now well established by the experiments of Lebedew and Nichols and Hull .
It follows then that the density of the momentum in the beam must equal The particular case which concerns us in the present paper is that in which ' Phil. Mag April , 1905 .
VOL. LXXXVII.\mdash ; A. Dr. G. Barlow .
Torque produced by a of [ Mar. 27 , a parallel beam of light is displaced parallel to itself by means of some optical system .
Such a displacement may be produced by oblique of the beam through a glass plate .
If the reaction is on the matter through which the light is passing , the molnentum principle shows at once that the material system must experience a torque of magnitude , where is the per unit length of the beam , and is the lateral shift .
This case is analysed by Prof. Poynting , who shows that when a beam is refracted at a surface there will always be an outward pull ( i.e. from the denser medium ) along the , and it follows that when a beam of light passes obliquely through a parallel plate there is a normal pull outwards both at incidence and at emergence , and these two pulls constitute the torque .
The result must be slightly modified when we take into account the fact that some of the light will be reflected at the surfaces and will give rise to additional forces normal to the surfaces .
In the same paper a brief account was given of some experiments carried out by Prof. Poynting and the writer with the object of showing the outward pull at oblique refraction .
* Two glass prisms , each with refracting angle , were arranged 3 cm .
apart at the ends of a torsion arm suspended at its middle point from a quartz fibre .
The air pressure was reduced to about 15 mm. .
The lateral displacement of the beam of light sent t , hrough the system was about cm .
The observed torque agreed within a few per cent. with that calculated from the energ of the beam .
The experiment was repeated with a pair of smaller prisms with the hope of obtaining more accurate results .
Unfortunately this lighter system was not so satisfactory as we expected .
The results were not concordant , and in some cases the observed torque was 50 per cent. greater than the theoretical value .
Although it seemed impossible to doubt that these experiments indicated a real agreement with theory , it was felt that a fuller experimental investigation of this case of refraction would not be without value .
With this object the experiment described in the present paper was planned some years ago , but other work has prevented its being carried out until recently .
The arrangement in the earlier experiments appeared to possess several defects .
Thus , in spite of the presumably small absorption of heat by the prisms , the disturbing action of the gas was quite evident .
This action was probably mainly due to the somewhat complicated form of the suspended system , since experience gained later with various forms of pressure of apparatus has shown that simplicity and symmetry in the suspended system greatly reduce .
also " " The Pressure of Light by J. H. Poynting , ' Romance of Science ' Serie* 1912 .
] Light in Obliqne through Plate .
the gas action .
Another objection is that , owing to the considerable distance between the prisms , there must have bee , n appreciable dispersion in the beam during its passage .
Also a very large correction had to be applied to allow for the four reflections in the system , and this correction was calculated by assuming the Fresnel formulae for the reflection coefficients .
Finally , it is now known that much greater accuracy could be obtained by hydrogen instead of air as the residual gas in the case .
IIIethod .
The desirability of doing away with the internal air space between the two } prisms , and necessity for symmetry , suggested to the author that the best form to adopt would be a perfect cube .
Let ABCD ( fig. 1 ) be the cube seen FIG. 1 .
FIG. 2 .
in plan , and let a uniform and parallel beam of hght fall on the one half AE of the face AB at such an angle of incidence that after refraction through the cube it emerges from the half ] of the opposite face .
The cube should then experience a torque in the direction shown by the arrow .
It is evident that the portions of the cube , AEG , CFH , correspond to the two prisms used in the earlier experiments This oement has the following advantages:\mdash ; ( 1 ) There is perfect symmetry .
It is easy to make the faces all truly vertical .
( 3 ) The optical path in the system is short .
( 4 ) There are only two primary reflections .
This correction is ooreatly reduced and can .
be calculated with more certainty than before .
( 5 ) By rotation all four faces may be used for incidence .
Any errors due to slight want of verticality will then be eliminated .
Dr. G. Barlow .
Torque oduced by of [ Mar. 27 , ( 6 ) No is required , as the faces of the cube can be used for observing the deflections .
This is an , as the attachment of even a small mirror will introduce a certain amount of dissymmetry , and therefore increase the sensibility of the apparatus to gas disturbance .
Description of the Apparatus .
A cube of crown glass was obtained from Messrs. .
Its were all 1 cm .
with an error of less than 1/ 100 mm. Four of the faces were optically worked , but the top and bottom surfaces were left rough ground .
The method of suspension is shown in fig. 2 .
An aluminium disc , of 1 cm .
diameter , was to the end of the axial brass rod B. Both disc and rod were accurately turned by a watchmaker .
The cube was then cemented to the disc by a small amount of shellac .
The upper end of the suspension rod was filed down so that the quart fibre could be soldered to it axially .
The experimental case , a gunmetal casting cm .
see was provided with plate glass windows on front and back , and with smaller windows A and at the ends of two lateral tubes .
The beam of light , entered by one of these tubes and left by the other .
Owing to the considerable length ( 20 cm .
) of the tubes , it was expected that any gas action due to warming of the windows by the the beam would be very greatly reduced .
The vertical tube was closed by a ground stopper to which was fixed a suspension rod for the attachment of the quartz fibre .
The stopper also carried an index so that angles of rotation could be measured on the divided circle E. The source of was an Ediswan -volt focus lamp\ldquo ; which was fed from accumulators .
By means of adjustable resistances in series with the lamp the was maintained at 60 volts to vithin1/ 10 volt , the lamp then taking a current of ) amperes .
A lens of focal length 15 cm .
formed an image of the lamp filament on an achromatic lens which had a focal length of 19 cm .
In front of was placed a specially constructed rectangular provided with regulating screws so tlJat all its edges could be independently adjusted .
The lens formed an image of this in a plane through the centre of the cube .
As the depth of focus was considerable the cross-section of the beam could be fairly well regarded as uniform through a distance of 1 cm .
the thickness of the cube ) .
In the ideal case the beam of light should be strictly parallel , whereas in the experiment the light formed a htly converging cone .
A circular diaphragm ( radius cm .
) at limited the semi-angle of the cone to less than .
The errors introduced by this effect and also by the 1912 .
] Light in Oblique Refraction through dispersion at refraction are evidently very small and therefore have not been taken into account .
The experimental case was first set up on a pier built up from the floor .
Much inconvenience was caused by the slight pendulum motion of the cube produced by vibrations in the building .
For the " " critical measurements\ldquo ; ( referred to in Table ) the apparatus was reset up on a very massive brick pier built up directly from the ground and separated from the floor by a narrow trench .
The observing telescope was clamped to the pier , and the deflections were read on a millimetre scale at a distance of 250 cm .
The definition was so good that 1/ 10 scale division could be estimated with certainty .
In the experiments in which the cube was rotated slight adjustments of the telescope were sometimes required , but the scale was nsver moved .
Some metal screens , not represented in fig. 3 , protected the case extraneous radiation .
Dr. G. Barlow .
Torque produced by Beam of [ Mar. 27 , Optical ustments .
In adjusting the beam of light three conditions have to be fulfilled : ( 1 ) The image of the diaphragm formed by the lens L2 must be in the plane containing the axis of the cube .
( 2 ) The cube must be set so that the beam falls on it at the required angle of incidence .
( 3 ) The beam must be adjusted to cover exactly one half face at incidence .
These adjustments were made as follows:\mdash ; ( 1 ) An auxiliary achromatic lens ( fig. 3 ) was placed behind the exit tube , so that an enlarged image of the cube was projected on a screen at J. The lens was then moved until the image of the diaphragm opened fully so that it could be seen above and below the cube ) was also in focus at J. During this operation the cube was strongly illuminated .
( 2 ) The cube was first set so that one face was perpendicular to the axis of the incident beam .
To do this a vertical slit of width 2 mm. was placed over the diaphragm , and a half-silvered plate was set obliquely in the path of the beam .
When the face of the cube was perpendicular to the light , the reflected beam passed back again ohrough the slit and was easily observed by placing the eye at , and looking into the half-silvered plate .
When the cube was the torsion head was until the instants of the illumination of the slit occurred at equal intel.vals of time .
After allowing the cube to come to rest a more exact adjustment was effected if necessary .
The reading of the torsion head was then observed and the head was rotated through , the required of incidence .
( 3 ) The slides of lihe adjustable were now slowly closed in until the image of the diaphragm illuminated exactly one half the face of the cube .
This adjustment was carried out with great ease and accuracy by observing the image formed on the screen , and also the images formed by the lenses and on the screens and R The top and bottom slides were first adjusted by looking at the image or ) J. For the laGeral adjustment , which is particularly important , use was made of the fact that when the beam was too wide an internal reflection in the cube rise to a bnght patch on the screen , and refraction at another face gave rise to a second bright patch on the screen J. The slides were closed in until these two patches of illumination just vanished .
The screen then showed only a single rectangular patch of brilliant illumination\mdash ; the image of the half face of the cube through which the beam emerged .
The actual deflection of the cube produced by the was less than and the amplitude of oscillation was generally of this order .
Hence the beam could be regarded as being always incident at sensibly the same angle , and therefore the torque exerted by the light could be taken as constant .
1912 .
] Light in Oblique Refraction through In fact observations of the deflections for various amplitudes of swing showed that no appreciable error could be attributed to any variation of couple with displacement .
The Gas Action .
It is well known that the disturbance , or " " gas action in air is generally at a minimum for a pressure of about 1 or 2 cm .
Hg .
The existence of this favourable pressure region taken advantagefof by Nichols and Hull in their experiments .
Below this pressure the gas action rapidly increases , and is then due almost entirely to the radiometer effect .
But above this critical pressure it has been shown by Ghat convection plays the chief and he proved that the action then depended very greatly on the inclination of the receiving surface to the vertical .
In the writer has observed a similar minimum of the gas action , which occurs also at the pressure of 1 or 2 cm .
Hg .
But in this gas the disturbance due to convection is always very much less than in air .
This fact , which does not appear to be widely known , that in all experiments where use is made of a delicate torsion balance it will be found of great advantage to work in hydrogen at a reduced pressure .
The exceptional behavionr of this gas is probably due to its great thermal conductivity .
In the experiments the range of gas pressure extended from atmospheric pressure down to cm .
Hg .
As had been anticipated , the apparatus proved to be remarkably free from the effects of convection , especially when hydrogen was used .
In this gas only minute changes in the centre of swing took place during any set of observations at a favourable pressure .
It was not until the pressure exceeded about 40 cm .
Hg that any appreciable disturbances occurred , and even then they were not very serious .
The radiometer action being always a pressure back on the surface heated , it follows that in the present experiment ( and also in the earlier experiments with prisms ) any radiometer action will tend to deflect the system in the opposite direction to the torque due to the light .
This will be seen by reference to fig. 1 , since the radiometer pressures on the areas AE and which we must suppose warmed by the passage of the beam , will produce a torque on the cube in the direction opposite to the arrow .
The actual observations confirmed this supposition .
For example , in hydrogen he deflection was nearly constant for pressures from 76 cm .
down to about 10 cm .
Hg , and the value of the deflection agreed closely with that expected from the pressure of light .
But below 10 cm .
the 'Phys .
Review , ' 1905 , vol. 20 , p. 292 .
Dr. G. Barlow .
Torque oduced by of [ Mar. 27 , deflection fell off , at first slowly , and finally very rapidly , as the gas pressure was reduced , until at the lowest pressure available ( about cm .
Hg the deflection had fallen to 1/ 40 of its original value , .
the radiometer action had almost compensated the torque due to the light .
There seemed no doubt that a further reduction in the gas pressure would have resulted in a reversal of the deflection .
Unfortunately , the apparatus was not suitable for .
at lower pressures , so the reversal could not be obtained .
The general behaviour in air was similar to that in hydrogen , but the measurements were not nearly so consistent .
Hence hydrogen was used in most of the experiments .
A curve showing the variation of the observed deflection with gas pressure in the case of hydrogen is given in fig. 4 ( see also Table I ) .
It 1o so FIG. 4 .
will be noticed that even up to such high pressures as 20 or 30 cm .
Hg the radiometer action was still sensible , as evidenced by the cvht slope of the curve .
The slope continues at higher pressures but the observations are not reliable in this region on account of convection .
It , therefore , appeared a little difficult to determi1le what value of the deflection should be taken as representing the action of light pressure alone .
But an inspection of the curve suggested that the radiometer effect was inversely proportional to the gas pressure over the range used .
This same law is also given by theory .
* Hence the deflection and the pressure should be related by the equation where is the required deflection due to light pressure alone , and is some constant .
* Osborne Reynolds , " " On Certain Dimensional Properties of Matter in the Gaseous State ' Phil. Trans 1879 .
1912 .
] Light in Oblique Refraction a The above relation was tested by plotting the product as ordinate against as abscissa .
This should give a line ( since not passing through the origin , and the slope of the line determines .
The resulting graph for the observations made in ( see fig. 5 , which is obtained from the critical determinations given in Table II ) showed so little deviation from a straight line that we may regard the relation as satisfactorily established .
The value of is best obtained phically by the slope of the line , but it can be calculated from any pair of deflections observed at , respectively , by means of the expression Both these methods have been used .
Measurement of the Deflections .
The observations were made on the following plan : apparatus was first exhausted as completely as possible , and dry gas was then admitted Below 3 cm .
Hg the gas pressure was determined by a small mercury Dr. G. Barlow .
Torque produced by Beam of [ Mar. 27 , manometer inside the experimental case , but the greater pressures were measured on an external manometer immediately before closing the stopcock of the case .
The process of admitting or pumping out , in order to change the pressure , generally caused the cube to vibrate with a considerable amplitude .
After allowing the oscillations to die down to a convenient amplitude , the " " damping ratio wing ) : ( swing ) , was determined by observing about fifteen successive turning-points .
In hydrogen was about , in air about , but the value varied ) htly with the pressure , so that it was necessary to determine it for each pressure used .
Knowing , the centre of swing can be determined from any two consecutive turning-points , , by the relation .
The centre of swing was thus calculated for every pair of successive turning-points for both " " on\ldquo ; and " " light off This method is particularly convenient in such cases as the present , where the value of is known very accurately , and where it is of importance to estimate the centre of in as short a time as possible , so that any ressive change 1Jlay be detected .
At each pressure a set of observations was taken as follows:\mdash ; ( l ) Light off ; three -points observed , two values of the centre of swing .
( 2 ) on for three turning-points .
( 3 ) Light off for three turning- points , and so on for four or exposures .
The conditions were varied as much as possible .
The was put on ( or cut off ) sometimes just before a turning-point and sometimes just after one .
lso in some cases the exposures were made for only two , and in other cases for more than three , turning-points .
Care was taken to avoid up a large amplitude of swing .
The mean centre was taken for each of the groups ( 1 ) , ( 2 ) , ( 3 ) , etc. , and the deflection for each exposure then calculated as the difference between the mean centre for the light on , and the mean centre for the two groups , with light off , on either side .
A few observations were made with exposures lasting for seven or eight turning-points ( i.e. over six minutes ) , but in no case was there noticed any appreciable influence of time of exposure on the deflection .
Also a few deflections were calculated by observing the first swing of the cube from rest when the beam was suddenly put on .
These were in close agreement with the other values .
The turning-points were read with certainty to of a millimetre scale division .
The error in the calculated centre of swing was then probably not more than of a division , and in .
the mean of a number of observations the error should be still less .
On this account the final values of the deflections are given to of a scale division , but this last figure is uncertain .
An inspection of the tables appears to show that the 1912 .
] Light in Oblique Refraction through Glass uncertainty in this figure does not amount to more than one or two units for the determinations at favourable gas pres , sures .
Therefore the deflection , which is of the order of 5 mm. , can be regarded as measured to within 1 per cent. But differences greater than 1 per cent. were sometimes found when the observations were separated by several hours , and it seemed probable that slight changes in the intensity of the light may have affected the deflections .
Calculation of the Cube , ( iorrections for the Reflections .
Let edge of cube cm .
mean refractive index for white light angle of incidence angle of refraction The condition for the symmetrical transmission of the beam is that .
This ives r , and is then found from : The numerical values are given above .
Let total energy per unit of the incident beam measured in .
First , suppose that is transmitted without loss by reflections .
We have then a single torque on the system , and the arm of this torque is equal to the perpendicular distance of from the ray incident at the point ( see fig. 1 ) .
Hence Arm cm .
Torque When we take reflections into account , it is convenient to regard separately all the beams , incident , reflected , and transmitted , and to calculate the moment of each about the axis .
We have then ( see which shows the course of the central ) FIG. 6 .
Dr. G. Barlow .
Torque prodnced by a Beam of [ Mar. 27 , ( 1 ) The incident beam on the face .
The moment due to its reception is , where denotes half the arm given above .
( 2 ) A beam reflected from AE at incidence .
It can be shown that this will have arm , and gives a moment ( 3 ) A beam is internally reflected from .
It then falls on BC , is totally reflected at this face , and finally issues as a slightly diminished beam , , from the face .
Its direction is parallel , but opposite to that of the original beam , and it will have the same arm as that last considered .
Hence the moment due to it is ( 4 ) The transmitted beam of intensity .
This gives a moment Adding all these four effects together , we have the total torque on the cube given by The small fraction of the original beam is internally reflected at , but this may be left out of account .
A calculation showed that , even if this beam followed the theoretical path , it would not affect the total torque by more than per cent. Owing to imperfect parallelism in the beam , and the consequent scattering , its actual effect must be even less .
In the above calculation , it is assumed that there is no rption of by the cube .
In the experiment the ption was probably very small , since the beam had previously passed through a thickness of several centimetres of glass .
It may be noticed , however , that complete absorption would only reduce the torque to about half its value , so that , even if the absorption of energy were 1 per cent. of the whole , the error produced in the torque would be much less .
The coefficients , were calculated from Fresnel 's formulae for reflection .
The partial polarisation produced in the successive reflections was taken into account .
The following values were found:\mdash ; cm .
, , ; and subQtitution in the above expression gave Torque This is 10 per cent. less than the value given above for the ideal case of transmission without reflection .
In expressing the results of the experiments the method followed has been to compare the observed scale deflection ( corrected for radiometer action 1912 .
] Light in Oblique Refraction through Plate .
as explained above ) with the scale deflection which should theoretically be given by the beam .
The value of is calculated from the relation where scale distance , 2500 mm. dyne-cm .
, the torsion couple per radian for the fibre .
The constant was calculated in the usual way from the period , sec. , of the cube and the moment of inertia , .
cm.2 , of the system .
To obtain the last quantity the moment of inertia of the cube itself was calculated from its mass and dimensions , and to this was added the moment of inertia of the suspension rod , which was determined experimentally .
Determ of the EnerrJy of the The method is essentially the same as that adopted by Nichols and Hull and since used by Prof. Pointing and the writer in other experiments on the pressure of light .
The beam was allowed to fall on a blackened disc of pure silver diameter , thick ) , and the initial rate of rise of temperature of the disc was observed by means of a constantan-silver thermo-junction soldered to the back .
A Rubens " " Panzer\ldquo ; galvanotneter was used ; this was adjusted to have a period of about two seconds , and was then made dead beat .
The transit of each scale division across the spider line of the telescope was recorded on the drum of an electricallydriven chronograph .
A separate pen actuated by the laboratory clock simultaneously recorded seconds on the drum .
It was then possible by means of a graphical representation of the chronograph record to calculate , the rate of change of galvanometer deflection .
The number of microvolts per scale division of the galvanometer was determined with a cadmium cell and standard resistances .
The thermo-electric power of the couple was known from a separate investigation .
energy in unit length of the beam was then calculated from where is the thermal capacity of the disc , is the " " mechanical equivalent is the velocity of light , and is the fraction of the incident light not absorbed by the disc .
The optical system ( see was mounted firmly on a base which could be rotated in order to direct the beam on to the silver disc .
Measurements of the energy were made at the end of each complete set of readings Dr. G. Barlow .
Torque produced by a of [ Mar. 27 , of the cube deflections .
The mean of several consecutive energy determinations was always taken .
The silver disc was placed in a very massive cylinder of iron which served as a constant temperature enclosure .
The beam was admitted through a window of glass cut from the same sheet as the window of the experimental case .
Before the disc to the beam the rate of of galvanometer deflection was observed , but it was generally so small that it could be entirely neglected .
Some uncertainty in the determination of the energy is due to the absorption coefficient , , of the platinum black surface of the silver disc .
This coefficient has been assumed to have the value for the quality of the light used .
This is not likely to be far from the truth , but an error of 1 or 2 per cent. may possibly exist .
Experiments with an improved form of receiver are bein undertaken .
The Results .
A summary of the final results is given in Iable .
It will oe seen that there is a general tendency for the observed deflection to be slightly greater than the value calculated from the energy .
The difference is from 1 to 2 per cent. in the hydrogen experiments , and considering the difliculties in determining the absolute value of the energy this agreement may be considered satisfactory .
The experiments in ai show reater deviations , but the measurements were obviously much less reliable than those in hydrogen , so that no great weight must be attached to them .
Fuller details of some of the series of observations are given in Tables I , II , and III .
The values in Table I are also represented graphically in fig. 4 .
Table III contains a set of observations in which all four faces of the cube were used in turn for incidence , and it is seen that the mean deflection for any face does not differ by more than 1 per cent. from the mean for the four faces .
Except in the two series A and , the same face ( that denoted Table I.\mdash ; Variation of Deflection with Gas Pressure in .
( Series B. ) The value of is the mean of 3 to 6 determinations .
cm .
Hg .
cm .
Hg .
cm .
Hg .
cm .
Hg .
in scale divs .
3.44 0.55 to 0.1 From these results the raphical method gives 1912 .
] Light in Oblique Refraction through Glass by ( 1 ) in Table III ) was used for incidence in all experiments .
face , at least for the higher gas pressures , generally gave a deflection about 1 per cent. in excess of the mean .
On this account the values of for the and in Table should perhaps be reduced by about 1 per cent. , and this correction is in the right direction to bring about closer agreement .
Table II.\mdash ; Variation of Deflection with Gas Pressure in Hydrogen .
( Series E. ) From these results the graphical method.gives Torque produced by Light in Oblique Refraction .
Table III.\mdash ; Deflections in using the Four Faces of the Cube .
( Series F. ) From these two values of the formula ives ) ) ) .
Table cRange * This value has been corrected for radiometer action by applying a factor obtained from the later series .
" " Critical determinations\ldquo ; made after the apparatus had been re-set up .
We may therefore conclude that the oblique passage of a beam of light through a plate of refracting material produces on the matter of the plate a torque which has the magnitude deduced from the transfer of momentum in the beam .

