272 Hon. R. J. Strutt .
A Study of the Line the drops cannot form anything like 1/ 100,000 of the total volume .
For example , though the limiting velocity of drops a tenth of an inch in diameter has not been measured , ordinary observation of heavy rain will show that it is not less than 10 ft. per sec. , and thus with Q = 10-5 vj would be not less than 1-2 x 102x 10-1/ 104 in .
per sec. , 100 in .
per day .
There are no measures , as far as I know , * of rainfall per second , or for any short intervals , but it is not impossible that a rate of 10 in .
per day , or even more , might be reached occasionally .
It may safely be inferred , however , that even in the heaviest rain Q is rarely as great as 10-6 .
The case is rather different for fog and mist , where the velocity of the drops is very small , but even when a fog is so dense as to make a street-lamp only just visible at a distance of 10 ft. ( the densest white fog in my experience ) it will be found that either the drops must have been much less than a thousandth of an inch in diameter or Q much less than 10-5 .
Bakerian Lecture : A Study of the Spectrum of Sodium as Excited by Fluorescence , By the Hon. B. J. Strutt , F.R.S. , Imperial College , S. Kensington .
( Received June 17 , \#151 ; Lecture delivered June 19 , 1919 .
) [ Plate 5 .
] CONTENTS .
page S 1 .
Introductory ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... .
272 S 2 .
The Sodium Yapour Lamp ... ... ... ... ... ... ... ... ... ... ... ... . .
274 S 3 .
Radiation of the Lamp .
Minute Structure of the D Lines ... ... 276 S 4 .
Resonance with I ) Light ... ... ... ... ... ... ... ... ... ... ... ... .
277 S 5 .
Breadth of the Resonance D Lines ... ... ... ... ... ... ... ... ... . .
278 S 6 .
Resonance Radiation at the First Ultra-violet Line ... ... ... .
280 S 7 .
Resonance and the Zeeman Effect ... ... ... ... ... ... ... ... ... 281 S 8 .
Stimulation of D Light by Excitation at the First Ultra-violet Line 282 S 9 .
The Two D Lines Separately Considered ... ... ... ... ... ... ... . .
283 S 10 .
Polarisation Experiments ... ... ... ... ... ... ... ... ... ... ... ... .
284 S11 .
Connection between D Lines and the Band Spectrum ... ... ... ... .
285 S 12 .
Summary of hitherto unpublished Experimental Work ... ... ... ... 286 S 1 .
Introductory .
It was recognised in the early days of spectrum analysis by Stokes and Kirchhoff that the absorption of light by a sodium flame or* by sodium vapour * AViesner gives an account of the rainfall at Buitenzorg , and has measured the rate of rainfall over intervals of 5 to 15 minutes .
In one case the rate of rainfall of 3-27 metres per day ( nearly 46,000 inches per year ) was recorded as lasting over 15 minutes .
i Spectrum of Sodium as Excited by Fluorescence .
was conditioned by resonance of the vibrating sodium molecules to that period in the exciting light which was removed .
In more recent days our knowledge of these phenomena has been much extended by E. Wiedmann , and later , R. W. Wood , who showed that the absorption just mentioned was accompanied to some extent by lateral re-emission.* L. Dunoyer has also made important observations .
This is the phenomenon known as resonance radiation .
[ Since this paper was written I have to lament the death of my beloved father , Lord Rayleigh .
I found the following rough memorandum , dated September , 1897 , among his papers , not put away carefully , but in a pile of letters and pamphlets of no value .
It was written long previously to my own interest in the subject , and I think he had quite forgotten having considered the matter ; at all events , he never referred to it when I showed him some of my own experiments .
The memorandum runs thus:\#151 ; " A moderately fed soda flame stops light of D quality incident upon it .
This may be seen to some extent with sunlight , but better with electric arc , or perhaps with another and brighter soda flame .
What is the nature of this stopping ?
It is generally spoken of as absorption .
But this seems unlikely .
It must be a molecular operation .
If the molecule in the Bunsen shines on excitation , how can it fail to do so as the result of energy of the right sort already falling upon it , and certainly operative upon it ?
It seems more likely that the energy is re-radiated without absorption , i.e. , that the light is scattered .
If so , it would probably be scattered as from small particles of the ordinary sort with polarisation effect."\#151 ; July 21,1919 .
] The range of conditions for resonance radiation is much more restricted than for absorption .
Had it been otherwise , the discovery would not have been so long deferred .
Absorption is not interfered with\#151 ; indeed , the absorption lines are much broader and easier of observation , when the sodium vapour is enormously diluted with foreign gases , as , for instance , in a Bunsen flame .
The causes of this broadening are not yet clear , but it is certain that the lines become very much narrower when the sodium is , and it is under these conditions that the resonance radiation makes its appearance .
Evidently the same interaction with the surrounding gas which broadens the absorption line in some way conditions the disappearance of the resonance radiation .
* For an account of Wood 's investigations prior to 1912 , see his ' Physical Optics , ' 2nd ed. , and papers in the ' Philosophical Magazine ' since that date .
Dunoyer 's more important papers will be found in 'Le Radium , ' vol. 9 , May , 1912 , and 'Journal de Physique , ' January , 1914 .
My own earlier observations will be found in 'Roy .
Soc. Proc. , A , vol. 91 , pp. 389 and 511 ( 1915 ) .
They are summarised in the present paper , which also contains new material .
VOL. XCVI.\#151 ; A. U Hon. R. J. Strutt .
A Study of the Line Since the absorption lines are extremely narrow under the conditions which are favourable for observing resonance radiation , and since only the absorbed radiations can be effective , it results that for good efficiency the source employed to stimulate the resonance should be one giving extremely narrow bright lines .
Light belonging to any other part of the spectrum is worse than useless , for it is inevitably diffused to some extent by the walls of the vessel used , and tends to mask the effects to be looked for .
The salted Bunsen flame is a very poor source of radiation for these experiments , because of the breadth of the lines .
As Dunoyer has shown , it works to best advantage when sparingly supplied with salt , for when salt is freely supplied the lines become broader , and useless radiation outside the effective region of the spectrum is added ; on the other hand , the effective radiation from the centre of the line is little if at all increased , for the line becomes conspicuously reversed , the centre being the darkest part .
I have found very great advantage in substituting for the soda flame a sodium vacuum arc in quartz , similar in principle to the familiar quartz mercury arc .
The sodium lamp is , of course , considerably more troublesome than the mercury lamp , but of late I have succeeded in making improvements in this respect .
S 2 .
The Sodium Vapour Lamp .
The present construction of the lamp is shown in fig. 1 .
( a ) is the silica envelope .
The electrodes are fitted to it by indiarubber connections , which are sufficiently tight and mechanically better than cement .
The cathode ( ) is a pool of molten sodium , and connection is made to it by the wire ( which may be of iron or copper , and fits the silica tube pretty closely .
The anode ( d ) is of tungsten rod , 5 mm. in diameter .
This metal , though it gets red hot , seems to stand the action of sodium vapour indefinitely .
Iron anodes , on the other hand , after an hour or two of use , fuse and drop off , presumably owing to the formation of an alloy : for the temperature is far below the melting point of iron .
To economise tungsten , the actual electrode is ground down to a smaller diameter at one end , and fitted into a hole drilled in an iron rod ( e ) , which is pinched on to it .
The iron rod is carried by the brass fitting ( / ) , through which the tube can be exhausted by a Gaede mercury pump , kept continuously in action while the lamp is running .
If the lamp is to last long , the most important point is to run it with only a small current , so as not to heat and reduce the silica of the narrow connecting tube , which is the source of light .
The lamp can be run at a current of 5 or 6 amperes without the need for any extraneous heating to raise sodium vapour , except at the start , but so large a current is very hard ) Spectrum of Sodium as Excited by Fluorescence .
on the silica .
If it is attempted to reduce the current much by additional resistance , the lamp goes out .
After much trouble , it was found that the difficulty could be overcome by gently heating the lower bulb containing melted sodium over a small ring gas burner ( g ) , concentric with the lamp .
A test made on an empty tube showed that the gas burner alone , as used , kept the inside of the lower bulb at a temperature of about 180 ' C. , apart from any heating effect of the current , he latter heats the middle and upper part of the lamp more than adequately , but does not raise enough vapour from the sodium cathode without the help of the gas ring .
The working current was 2'5 amperes , and did not heat the middle tube of the lamp hot enough to char a piece of paper held against it for a few seconds .
With the self-induction of a large electromagnet in the circuit , and on the 200-volt supply , the lamp would run indefinitely without going out .
It has been run 24 hours or more , being left unattended all night .
The drop of potential over the lamp was very capricious , even after long running at a constant current .
It varied from 20 to 50 volts .
The sudden movements of the voltmeter seemed to correspond with a wandering of the point of departure of the arc from the sodium surface .
The lamp is started some minutes after lighting the ring burner by means of a small induction coil , one terminal connected to the sodium cathode , the other to a wire twisted round outside the central silica tube .
If any difficulty is experienced in starting , it is usually traceable to insufficient exhaustion .
I have not been able to start the lamp on a 100-volt supply , although this gives a good margin over the voltage dropped on the lamp while running .
After an aggregate run of 24 hours it is generally advisable to dismount and wash out the lamp , first with alcohol and water , then with weak hydrofluoric acid .
This method of removing the brown coating of silicon is better than the use of boiling caustic soda , or than applying the blowpipe .
U 2 j- Uj Fig. 1 ( 1 actual size ) .
Hon. R. J. Strutt .
A Study of the Line S 3 .
Radiation of the Lamp .
Minute Structure of the D Lines .
The D line is by far the strongest in the spectrum , and , in experimenting with it , it is generally unnecessary to use colour screens to cut out the other lines .
If the lamp is used to excite D-line resonance radiation in a bulb containing sodium vapour , most baffling caprice is observed in the intensity of the effect , which will sometimes be extremely brilliant , and at other times will disappear , without obvious change in the conditions .
A Fabry and Perot ttalon was set up , to examine the structure of the D lines from the lamp , and to test whether the caprice just mentioned could be connected with it .
The 4talon was about 5 mm. thickness , and had to be packed with paper ( not an ideal arrangement ) to get the distance right for separating the Di and D2 fringes .
It was then observed that each of the D lines was reversed , but that the breadth and darkness of the reversal varied with the same caprice that had been observed in the resonance radiation .
Sometimes one of the D lines ( I did not determine whether Di or D2 ) was much more conspicuously reversed than the other .
A weak electromagnet was then arranged to throw the column of discharge against the forward wall of the tube , and squeeze out the reversing layer of cool sodium vapour.* The reversal was thus made much less conspicuous in the interferometer , though it was never wholly got rid of .
Plate 5 , No. I , shows the change in the interference rings when the magnet current is reversed .
In the top half the discharge is thrown against the front wall of the tube , in the bottom against the back wall .
It will be noticed how enormously the reversal is developed by this change .
A strong field tends to concentrate the discharge so much to one side of the tube that the wall becomes unduly heated , with attendant bad effect on the silica .
This becomes brown , and may even be locally heated so strongly as to soften and be perforated .
In practice the field was not measured , but was increased by moving the magnet or increasing the oxciting current , as far as could be done without making the front of the tube hot enough to quickly char paper .
Even in this case the front of the tube deteriorates sooner than the back , which latter is often cool enough for sodium to condense upon it .
After a few hours ' run it is convenient to turn the tube round on its axis , the india-rubber connection to the pump being long enough to allow of this .
Any deposit of sodium then soon clears off , leaving a clear silica wall through which the light can emerge .
* This is in imitation of the device used by Kerchbaum , ' Electrician , ' vol. 72 , p. 1074 ( 1914 ) , with the mercury vapour lamp .
Spectrum of Sodium as Excited by Fluorescence .
2 S 4 .
Resonance with I ) Light .
With the sodium lamp as described , focussed upon the wall of an exhausted bulb containing sodium vapour , very brilliant resonance is obtained , bright enough to be readily shown in the largest lecture room .
This brightness makes possible various experiments which would be very difficult to carry out with the salted flame as a source .
At the lowest temperature and least vapour density where the phenomena are perceptible , the beam , as observed from the side , extends right across the bulb , without losing intensity as it proceeds .
With rising temperature , luminosity is not limited to those parts of the bulb traversed by the primary beam , but spreads beyond them .
This is due to a tertiary emission , the scattered radiation being itself absorbed and re-emitted .
At one stage the whole bulb appears to glow almost uniformly with yellow light .
With further increase of density , the luminosity begins to draw in to the front surface of the bulb , the toll taken from the primary beam being so great that it becomes rapidly impoverished in the effective constituents .
Finally , at temperatures in the neighbourhood of 300 ' C. , the emission observed is almost entirely from the front wall of the bulb , and is limited to a skin a fraction of a millimetre in thickness .
Closer examination shows , however , that there is always present as well , a much feebler D emission , which extends with little diminution right across the bulb .
Dunoyer* has observed something of this effect , which he regards as a transition stage between the phenomena of pure volume and pure surface resonance .
I have , however , been able to observe it up to the highest temperatures attainable without blackening in glass vessels , and have also observed a volume effect of the same nature in a special iron vessel heated to 500 ' C. As Dunoyer has remarked , it is natural to attribute the superficial resonance to excitation by the core of the D line , and the subsequent luminosity to excitation by less central portions of the line .
When the bulb is observed or photographed from in front , the superficial resonance is overpowering ( see Plate , No. III ) .
But , when observed from behind , the superficial resonance is invisible , owing to absorption by the vapour , and the luminous track across the bulb stands out strongly ( Plate , No. IV).f The light emitted along the track , being of slightly different refrangibility irom the superficial emission , is not so strongly absorbed , and can emerge through the layer of intervening vapour , and reach the observer 's eye .
The * 'Journal de Physique , ' January , 1914 .
t The patches of light marking the entrance and exit of the beam in the photograph are due to diffusion by the glass walls .
278 Hon. R. J. Strutt .
A Study of the Line annexed diagram ( fig. 2 ) will show clearly what is meant by observation from in front and from behind .
EXCITING BEAM Fig. 2 .
The different refrangibilities of the superficial and volume radiation as they emerge from the bulb can be shown by observing it through an independent layer of sodium vapour .
A thin steel tube containing sodium was closed at the ends by glass plates cemented on .
It could be evacuated and heated electrically to raise sodium vapour .
The ends were water-cooled to protect the sealing-wax .
The resonance bulb was photographed from in front through the tube , first cold ( Plate , No. IV ) , then heated to a temperature carefully adjusted by trial to get the right effect ( Plate , No. V ) .
It will be noticed that absorption has taken far more toll from the superficial luminosity , which is now so much diminished that it no longer masks the volume effect .
No. IV was not given so long an exposure as No. V. If it had been , the brilliant superficial resonance would have blurred the whole picture by over-exposure .
S 5 .
Breadth of the Resonance D .
The experiments just described show that the breadth of the resonance line is not a very definite quantity .
It increases with the thickness of vapour traversed by the incident beam .
It was , however , of interest to examine the superficial resonance under high spectroscopic resolving power .
A Lummer plate was used , mounted on a Hilger " wave-length " spectroscope .
The interference pattern is seen in Plate , No. VI .
The spectroscope did not separate the two D lines horizontally as widely as might be desired for the present purpose , and this is the more unfortunate in that the thickness of the Lummer plate did not happen to be right for fully separating the Di and D2 fringes vertically .
However , they can be seen to be separate vertically in the first few orders , from the bottom of the photograph , which , Spectrum of Sodium as Excited by Fluorescence .
279 of course , corresponds to grazing emergence from the Lummer plate .
A wide slit was used , thus Di and D2 are not separated horizontally , except at the right and left edges of the pattern .
The full observable breadth was determined on a measuring machine as 0192 of the distance between successive orders , and the distance between successive orders can be calculated from data for the Lummer plate supplied by the makers as 0T77 Angstrom .
This makes the full observable breadth 0-034 Angstrom .
There is a little uncertainty , owing to the inadequate separation of Di and D2 .
It was assumed that the edges of the Di and D2 fringes were just in contact .
In any case , however , a measurement of this kind only gives a rough notion of the true breadth of a line as defined by Buisson and Fabry.* Assuming a distribution of light of the form when x is the distance along the spectrum from the centre of the line , the half breadth is defined as the distance in which the intensity is reduced to half the maximum .
Nearly the whole of the light is contained in something like double this distance .
If it is assumed that this corresponds to what was measured on the photograph , the true breadth of the line would be 0-034/ 2 or 0-017 Angstrom .
The breadth of the line as determined by the Doppler effect , taking the radiator as the sodium atom , and the temperature as 250 ' C. , is given-by the formula of Buisson and Fabry as 0-0201 Angstrom , which agrees as well as can be expected with the value as roughly determined by the Lummer plate .
A more satisfactory determination of the breadth of the line could be made with a sliding Fabry and Perot interferometer , but such an instrument was not available .
I had , however , the use of a Fabry and Perot 4talon of about 3 cm .
distance between the plates .
A new distance piece was made for it , of such thickness as to bring the Di fringes midway between the D2 fringes .
A photograph of the resonance radiation taken with this showed the rings , though they were apparently not very far from the vanishing point .
Taking into account the complications introduced by the selective absorption of the vapour , it is not probable that the resonance lines have the exact intensity distribution which is postulated for definition of the breadth .
Thus it seems doubtful whether an attempt to determine the breadth more precisely would have much value .
The general conclusion is that the breadth of the lines is the smallest it could possibly be , if the radiators are monatomic molecules of sodium .
The thermal motion of the molecules necessarily gives rise to a broadening of the ideal infinitely narrow line , and this accounts for the whole observed breadth of the line .
In some cases , e.g. , the condensed discharge in hydrogen , other * 6 Societe FranSaise de Physique/ April110 , 1912 .
Hon. R. J. Strutt .
A Study of the Line causes enlarge the line much beyond the limit set by the Doppler effect .
But here such causes have no place .
S 6 .
Resonance Radiation at the First Ultra-violet Line .
In a salted Bunsen flame practically no lines due to sodium except the D line can be observed .
With electrical stimulation , for example , with the vacuum lamp described , many other bright lines due to sodium appear , some in the visual spectrum , and some in the ultra-violet .
These latter alone fall into the same spectrum series as the D line ; apart from the numerical relations of frequency , this is evidenced by the fact that they alone show the same type of Zeeman effect as the D line , and that they alone , like the D line , show as dark absorption lines , even when the vapour is at a moderate temperature and is not electrically stimulated .
My experiments on the ultra-violet lines of this series have been usually limited to the first one , at wave-length 3303 .
This light is not seriously weakened by passing through thin glass bulbs , and it can be very conveniently freed from D light by passing through cobalt glass , preferably the blue uviol glass made at Jena .
Blue light is removed by means of a dilute solution of nitrosodimethyl-aniline .
The radiation of the quartz lamp filtered in this way consists exclusively of the first ultra-violet line ; the second and higher ones cannot penetrate uviol glass.* Since resonance is obtained with the D line , we might expect it at 3303 .
My earlier attempts to observe this were unsuccessful , owing , as it now appears , to the small size and bad optical quality of the silica vessels used , which scattered so much light that they masked the effect to be looked for .
Owing to the circumstances of the time , the work had to be broken off somewhat hastily .
In taking up the subject again , a glass bulb 7*5 cm .
in diameter was used , charged with sodium according to Dunoyer 's methods , and exhausted as for experiments on the D emission .
Thin blown glass is not seriously opaque at X 3303 , and the surface is optically better than that of any silica bulbs that I have seen .
The bulb was heated to about 250 ' C. , and mounted so that the beam from a limited portion of the lamp , made parallel by a quartz lens , passed diametrically through it .
The bulb was * These latter lines are feeble unless the lamp is run at heavy currents , in which case its life is short .
I have on occasion run the lamp at 30 amperes or more , for a moment or two , heating it thereby till the silica walls collapsed under atmospheric pressure .
In this case the relative and actual intensity of the ultra-violet lines is much increased .
When the silica is very hot , the brown stain caused by moderate heating in contact with sodium vapour clears off , leaving the walls transparent .
It might be feasible to run the lamp for longer periods in this way if an external vacuum jacket were provided to prevent collapse .
Spectrum of Sodium as Excited by Fluorescence .
281 photographed laterally , with a quartz lens and blue uviol filter , stopping all D light .
The result is reproduced in Plate , No. II .
The track of the beam across the bulb is well seen .
To verify that this re-emission consisted of X 3303 , the bulb was photographed again with a quartz prismatic camera , set up at a distance .
Two small monochromatic images of the bulb with the beam stretching across it corresponding to D and X 3303 were obtained .
Using an ordinary plate ( not colour sensitive ) , X 3303 was somewhat more intense , but on orthoehromatic plates the D image was greatly over-exposed by comparison .
The subordinate series lines gave no re-emission .
They were represented only by dots corresponding to the point where the beam entered and was partially diffused by the glass wall .
The third line of the principal series would not penetrate even a thin glass bulb .
An attempt was made to observe its re-emission , using a quartz sodium bulb and the prismatic camera , but without success , probably from the relatively small intensity of this line in the source .
This latter point was not pursued to the utmost .
The success now attained in detecting 3303 resonance radiation in contrast to the former failure , is mainly due to the use of a large bulb and a temperature low enough for the radiation to be able to penetrate across the vessel instead of being confined to the neighbourhood of the surface .
In this way the resonance radiation is well separated from radiation scattered or reflected by the wall of the vessel .
It is true that the volume resonance is likely to be intrinsically faint compared with the superficial , but this is more than compensated by using a photographic plate instead of the fluorescent screen then employed .
The photographic method is usually much to be preferred in looking for faint ultra-violet radiation .
S 7 .
Resonance and the Zeeman Effect .
The resonance radiation so far described , whether at the D line or at X 3303 , is an effect apparently quite in harmony with the classical mechanics .
Nothing is more natural than that radiations of a certain frequency falling upon a resonator should set the resonator into vibration , and cause it to re-emit the same frequency .
From the same point of view , it is equally natural that a slight disturbance of the tuning should greatly diminish the effect obtained ; and so it is found .
It is well known from Zeeman 's discovery that we can slightly alter the frequency of a spectrum line by placing the radiating or absorbing gas which is responsible for it in a strong magnetic field .
It is not convenient to place the vacuum lamp in a strong magnetic field , * but the experiment * The weak magnetic field applied to the lamp , as already described for a different purpose , is not enough to produce an appreciable Zeeman effect .
Hon. R. J. Strutt .
A Study of the Line can be well carried out if a slightly salted Bunsen flame is used .
If either this or the sodium bulb is placed in a strong magnetic field the tuning is disturbed , and the resonance radiation much weakened .
S 8 .
Stimulation of D Light by Excitation at the First Ultra-violet Lim . .
We now come to another effect , which seems to have no analogy in the behaviour of the vibrating systems dealt with in the classical mechanics .
This is the excitation of the lower frequency ( D line ) by stimulation with the higher one ( X 3303 ) .
To detect this in experiment , the line 3303 from the sodium lamp must be isolated by ray filters , opaque to visual light , as already described .
These filters are set in the wall of a dark room , into which the ultra-violet beam enters , and converges on the walls of a glass or quartz bulb containing sodium vapour .
The conditions of this experiment are advantageous , in that there is no " false light " to be feared , and the D line to be looked for is situated in the visually brightest part of the spectrum .
Even so , however , it is advisable to carefully rest the eye in the dark for some minutes before beginning observations .
With these precautions , the emission of visual light when the ultra-violet beam enters the vapour is well seen , and visual examination with a spectroscope shows the D line only .
It is quenched when a piece of thick plate glass is held in the path of the exciting beam .
The general validity of Stokes ' law made it seem improbable that D radiation would excite X 3303 in fluorescence .
Experiment confirmed this anticipation .
Ultra-violet light was cut off from the sodium bulb by putting a piece of plate glass 1*2 cm .
thick , and in other tests a quinine cell , between it and the lamp .
Photographs through blue uviol glass showed no trace of a beam traversing the glass even with 24 hours ' exposure .
The outline of the bulb was itself faintly visible , owing to stray light from the burner used to keep it hot .
To return to the emission of I ) light by stimulation at 3303 .
The interest of this observation is that ^ it clearly establishes the existence of some mechanical connection between the system which absorbs X 3303 and that which emits the D line .
There are known cases of mechanical connection between two different periods in a vibrating system , for example , Melde 's experiment , in which a vibrating string is set in motion by a tuning-fork of half the period , applied so as periodically to alter the tension of the string .
The Goldschmidt alternator used in wireless telegraphy is another instance .
But in this class of cases the periods are commensurable .
I mean that they stand in the same ratio as two small whole numbers .
The periods of D and X 3303 I Specti'um of Sodium as Excited by Fluorescence .
283 certainly do not do this .
Here we encounter in a simple form the known difficulty of dealing with spectra on accepted mechanical principles .
Although our notions of the mechanism by which spectrum lines are emitted and absorbed are very vague , it seems certain that the quantum relations are concerned .
If one quantum of the radiation X3303 were absorbed , and one quantum of D light emitted , there would remain over one quantum of the radiation X , 7519 .
I have not been able to find this line in the spectrum of the vapour excited by X 3303.* But , even if the residual energy really goes into this radiation , we could scarcely hope to detect it , taking into account how little sensitive the eye is in this deep red region of the spectrum .
The vapour stimulated by X 3303 emits , as we have seen , both X 3303 and D light .
From the quantum point of view , however , it seems improbable that these two radiations are simultaneously emitted by the same molecule , for the absorption of one quantum of X 3303 would allow of emission of the same , but would leave no energy over for the emission of D light .
S 9 .
The Two D Lines Separately Considered .
In the observations mentioned so far , it has not been necessary to take into account that the sodium lines are double .
An interesting field of work was first opened up by the work of Wood and Dunoyer , who found that , if one of the D lines only was used to stimulate the vapour , then that line only was emitted .
Later experiments by Wood and Mohlerj* have shown that some qualification of this statement is required .
It is only valid if the vessel is perfectly free from hydrogen , a condition not easily realised , because of the extraordinary tenacity with which sodium retains hydrogen , even when distilled .
It is further necessary that the density of the vapour should be below a certain limit .
As we have seen , stimulation at wave-length 3303 causes emission of the D line .
But 3303 , like D , is a doublet .
This naturally suggested that the corresponding lines of each doublet might be coupled .
If so , we might expect to find that stimulation at the more refrangible component of 3303 would lead to emission of Di only , without D2 .
I was able to try this experiment by making use of the fortunate accident that there happens to be a line of the zinc spectrum exactly coincident with the more refrangible number of 3303 .
Stimulating the sodium vapour by * Nor does it ever seem to have been observed from sodium vapour in any case .
I could not detect it in the vacuum arc .
t ' Phil. Mag. , ' vol. 37 , p. 456 ( 1919 ) .
Hon. R J. Strutt .
A Study of the Line this source , it was found possible to get a luminescence bright enough for examination with a special spectroscope capable of separating the D lines .
The result was to show that both the D lines were emitted , without marked difference of intensity .
At the time this experiment was made , nothing was known of the effect of hydrogen , or excess of sodium vapour , in communicating energy from one D line to the other .
It should be repeated in the light of Wood and Mohler 's result .
For it is now seen to be possible that Di alone was directly stimulated by the ultra-violet source , D2 deriving its energy from Dx .
The possibility must also be borne in mind that some similar complication may be concerned in the transference of energy between X 3303 and the D lines .
S 10 .
Polarisation Experiments .
Dunoyer* has elaborately examined the polarisation of D resonance .
I have repeated some of his experiments in a slightly modified form , preliminary to examining the new ultra-violet resonance at 3303 .
For observing weak polarisation , a Savart polariscope is usually employed , but no suitable instrument of this kind was at hand .
I found , however , in the laboratory stock a pair of quartz plates , which , when superposed in the right way , gave straight fringes like those of Babinet 's compensator , localised in the plates , one of which was wedge-shaped .
This combination was set up near the source for examination .
Photographs were taken from about 1 metre distance through the quartz plates , and a polarising prism placed over the camera lens .
The camera pointed at right angles to the primary beam .
The experiments were made in large glass bulbs .
The quartzes were arranged so as to produce fringes perpendicular to the beam stretching across the bulb .
With sodium vapour at low density , giving volume resonance only , the fringes were distinct but not strong , and , since they did not appear on the darker background , but only on the bright beam , it seems safe to assume that they indicate gehuine though slight polarisation of the scattered light .
In experiments of this kind it is necessary to guard carefully against spurious polarisation effects due to the curved glass walls .
To minimise these , a bulb of 300 c.c. capacity was used , and observation was through an area of rather less than 1| inches diameter .
As the temperature was raised the resonance radiation began to assume the superficial character , but as before , a fainter beam stretched right across the bulb .
This was photographed with the polariscope arrangement , and the * 'Le Radium , ' vol. 9 , May , 1912 ; 'Journal de Physique , ' January , 1914 .
i Spectrum of Sodium as Excited by Fluorescence .
285 fringes appeared decidedly less distinct than they had done at lower temperatures .
The loss of distinctness in the fringes with a higher temperature , and the gain with a lower one , was also observed visually , and could be repeated as often as desired .
The newly discovered resonance at X 3303 could be examined in the same way , substituting a blue uviol glass screen for the yellow one , and using ordinary instead of orthochromatic plates .
No trace of the fringes could be detected in this case .
The polarisation , if any , must be very slight indeed .
The slight degree of polarisation in the case of D light , and its apparent absence with X 3303 , is in very remarkable contrast to the nearly complete polarisation observed when ordinary gases scatter white light.* Unfortunately it is impossible to examine the behaviour of sodium vapour far away from the region of selective absorption and resonance , because of technical difficulties in handling vapour of the necessary density .
I have attempted to trace the transition by working as far away from the core of the D lines as possible .
The broadened D lines from an oxyhydrogen sodium flame were used as source .
The sodium vapour was contained in a cross-shaped iron vessel similar to that used in my studies of the scattering of white light by gases .
The light passed through a column of sodium vapour before arriving at the scene of action , and in this way the core of the D lines was removed , leaving only the extreme edges of the lines .
Even in this case , as long as any resonance remained observable , it failed to show strong polarisation , so that the transition between ordinary scattering and resonance radiation has not been traced .
The corresponding problem for mercury'vapour may prove more tractable .
An effort will be made to overcome the difficulties .
S11 .
Connection between Line and Band Spectrum .
The fluorescence of sodium vapour presents puzzling problems even in those simple cases above discussed , when the line spectrum alone is apparently concerned .
I do not enter upon the complex phenomena of the band spectrum which Prof. Wood has investigated working with dense sodium vapour .
It is desirable , however , to refer to the relation which he has found between the line and the band spectrum in the dense vapour .
These are , briefly , that the D lines are excited when stimulation takes in the blue green region giving the channelled absorption spectrum .
Also that each line of the principal series is accompanied by a similar channelled absorption spectrum .
In the light of these facts it seems doubtful whether any theory * ' Roy .
Soc. Proc. , ' A , vol. 95 , p. 155 ( 1918 ) .
286 Line Spectrum of Sodium as Excited by of the line spectrum can be complete which does not take the band spectrum into account .
But the complexity of the latter is discouraging .
Dunoyer has advocated the view that hydrogen is necessary to the appearance of the fluorescent band spectrum , but this I find difficult to accept .
S12 .
Summary ( including the hitherto unpublished Experimental Work , without reference to the General Discussion of the whole subject ) .
1 .
The sodium vapour lamp has been improved , and the conditions have been studied for steady running , and for getting the right kind of radiation to excite brilliant resonance with D light .
2 .
It is found that excitation of sodium vapour by the second line of the principal series leads to the emission of both 3303 and the D line .
On the other hand , as might be anticipated , excitation of the D line leads to the emission of the D line only , with 3303 .
3 .
When D light falls on sodium vapour of appropriate density , it is known that an intense surface emission occurs from the front layer , and a weak one from succeeding layers .
Analysis by absorption in an independent layer of sodium vapour shows that the superficial emission is more absorbable , and therefore nearer the centre of the D lines .
4 .
The breadth of the D lines in superficial resonance has been estimated by interferometer methods .
It is found to correspond with the breadth conditioned by the Doppler effect , on the assumption that the luminous centre is the sodium atom .
5 .
Polarisation could not be detected in the ultra-violet resonance radiation , though in accordance with previous observers it was readily observed in D resonance .
DESCRIPTION OF PLATE .
I. Interference rings obtained with 5 mm. e talon from sodium lamp .
Upper halfr with discharge thrown against front silica wall by the magnet .
Lower half , magnet reversed , so as to throw discharge against back wall .
This greatly develops the reversal of each D line .
II .
Bulb containing sodium vapour , traversed by the light of sodium lamp .
Photographed through blue glass , showing emission of the second line of the principal series .
III .
Bulb containing sodium vapour , illuminated by a beam from the lamp .
Superficial D emission as seen from in front .
IV .
Ditto .
Volume D emission , as seen from back .
V. Bulb containing sodium vapour .
Photographed without sodium vapour filter .
VI .
Ditto .
Photographed through sodium vapour filter , with longer exposure .
Greater relative intensity of the volume effect .
The oval patches to the left are false light , due to glass walls .
VII .
Superficial D resonance , photographed with Lummer interferometer .
i S'utt .
Roy .
Soc. Proc. , A , vol. 9(5 , Plate 5 .

