]\gt ; Dispersion of Light by Vapour .
By Prof. P. .
BEVAN , Royal Holloway College .
( Received June 7 , \mdash ; Read June 23 , 1910 .
) [ PLATE 2 .
] Anomalous dispersion in the region of the red lines of potassium was first observed by Ebert*in 1904 .
The method adopted for Ebert 's experiments was a modified form of the crossed method used by Wood in the ation of the corresponding phenomena in the case of sodium .
Potassium was heated in a tube through which two currents of hydrogen passed from each end to a central outlet .
The cool kept the potassium vapour in a prismatic form , so that light passing along the length of the tube suffered deyiation and dispersion by the potassium vapour prism .
The author of the present communication has shown that there is no need for the hydrogen streams .
If the tube be kept cool on its upper surface the metallic vapour takes of itself a prismatic form or is in layers of decreasing density , and so behaves in a similar way to a prism of homogeneous vapour .
The present communication deals with quantitative results from the measurement of dispersions at different wave-lengths , and it appears that the deviation due to potassium vapour is observable over the whole of the visible spectrum and for a distance in the ultra-violet .
Strong absorption takes place at the lines of the principal series and for wave-lengths near these series lines we have " " anomalous\ldquo ; dispersion .
This phenomenon has been observed at seven of the pairs forming the principal series lines for potassium\mdash ; as the pairs of lines in this series get closer and closer together with diminishing wave-length , the dispersion effects after the first two pairs are only observable outside the lines forming a pair , but there appears a lack of symmetry in the observed dispersion curves corresponding to the different intensities of the lines forming the pair .
The dispersion to be observed may then be regarded as that corresponding to the principal series absorption lines ; no other absorption region seems to affect the dispersion\mdash ; at any rate at low densities of vapour .
The method of experiment was the well-known method of Wood for similar experiments in sodium vapour .
Some metallic potassium was placed in a steel tube about 70 cm .
in length and 3 cm .
in diameter .
The ends of the tube were jacketed by concentric brass vessels through which water * H. Ebert , ' Boltzmann-Festschrift , ' 1904 .
' Phil. Mag 1909 , vol. 18 , p. 407 .
Prof. P. V. Bevan .
[ June 7 , could be passed to keep them cool .
The tubs was closed at each end with a quartz plate .
A side tube was soldered to the main tube , which could thus be connected with an air PumP or a supply of hydrogen .
An open brass trough with its bottom formed to fit the steel tube could be placed on the top of the steel tube , and cold water could be kept continuously running through this .
The under side of the tube could be heated by a row of bunsens or other suitable burner .
Wood has put forward the view that the dispersion is due to the layers of vapour decreasing in density from the bottom upwards , and not to the vapour forming a nearly homogeneous prismatic mass about a piece of metal .
I am inclined to think that the * latter is , however , the true view .
The best results are obtained with a low pressure of gas in the tube , and it is difficult to see how there can be much difference in density of vapour in the tube unless there is very rapid evaporation of the metal going on .
But under the conditions when the dispersion is observable the metal is evaporating extremely slowly\mdash ; a tube can be used for many hours with practically no observable distillation .
It seems more probable that there is in the metallic vapour itself very little temperature gradient , and that we a surface at which condensation of the vapoul takes place , and small drops of liquid metal are formed which fall back into the original metal .
This view is supported by the fact that I have been able to obtain the sharpest and steadiest dispersion curves in the spectroscope when only a small portion of the tube is heated .
I find that heating a larger portion adds little to the deviation but considerably to the absorption .
Whatever the actual state may be , the vapour behaves like a prism and deviates light sent through it .
Light from a horizontal slit illuminated by an arc lamp was sent through the vapour tube and then focussed by a lens on the vertical slit of a spectrograph .
The lens used for focussing this image of the slit was , I believe , a new kind , and proved very useful , so that it may merit a short description .
An ordinary quartz lens is troublesome because of the difficulty of focussing for light of different .
Achromatic combinations of quartz and fluor spar have been made , but their expense makes them not available for ordinary use .
The lens used for these expeliments was a combination of quartz and water .
Distilled water is transparent for ultra-violet light to wave-length 2000 Angstrom units to nearly the same extent quartz .
A layer of water 1 cm .
thick absorbs per cent. of the light at wave-length 2200 , per cent. at 2100 , and per cent. at 2000.* A plate of quartz of the same thickness absorbs per cent. at 2220 , 8 per cent. at 2140 , and * Kreusler , ' Annalen der Physik , ' 1901 , vol. 6 , p. 412 .
1910 .
] Dispersion of Light by Potassium Vapovr .
211 per cent. at 2030.* After standing for some time in glass .s vessels , water loses a good deal of its transparency , but it is still very transparent until a of 2100 is reached .
An advantage that a water lens has is that it may be made as thin as one chooses near its centre , so that the proportion of absorbed can be made very small .
The scheme adopted for the lens was to make a thin cell of a parallel plate of quartz and a quartz lens with surfaces of appropriate curvature .
These two quartz cell walls were fixed together with shellac to a very short wide brass tube so that the convex face of the lens was nearly in contact with the quartz plate .
We thus have a plano-concave water lens when the vessel is filled , which with the convergent quartz lens can give an achromatic combination by suitably choosing the radius of curvature of the outer surface of the quartz lens .
A better combination would be to have the vessel made with walls of two similar quartz lenses , one dextro-rotatory , the other laevo-rotatory , to eliminate the double refraction effect introduced by the rotatory power of the quartz .
It is easily shown that for such a combination to be achromatic for wave-lengths for which the indices of refraction are , for quartz , and , for water , the radii of curvature of the surfaces , must be related by the equation and then diyided by the focal length of the combination is for a for which the refractive indices are for quartz and for water .
If we choose the ratio so that the combination is romatic for wave-length 2570 and 3940 , we find this ratio is for this ratio , we do not introduce much error , and have a number more convenient for the actual grinding of the lens .
The lens is thus a concavoconvex lens , and the convex side has to be in contact with the water .
For the combination of two quartz lenses , two with the same radii of curvature will , of course , give a combination of double the power of the single one .
With the value for , we find for being the focal strength of the combination , the following numbers:\mdash ; Wave-length 2190 2310 2570 3400 3940 4670 5890 0 .
0 .
From 2570 to 5890 the focal length is thus very nearly constant , and the change ill focal length down to 2190 is not very large .
We have thus a * Pfliiger , ' Physik .
Zeitschr 1904 , vol. 5 , p. 215 .
the time the tube is heated .
As the dispersion is increased , absorption also increases , spreading chiefly from the yellow and the red at the extreme end of the visible spectrum inwards towards the two red lines .
The two actual absorption lines at 7699 and 7665 also broaden , but not so rapidly as the absorption regions in the orange and extreme red , so that the part of the spectrum seen in this region , when the vapour gets dense enough to cut off neal .
lie all the light , is two red portions of the dispersion curve on opposite sides of the potassium lines , and deviated on opposite sides of the position the undeviated spectrum .
With denser vapour still , the curve on the red :. .
: 1910 .
] Dispersion of Light by Potassium Vapour .
side disappears , and we get a short piece of the red on the orange side of the lines deviated a considerable distance from its original position .
These effects have not been photographed with success , as the intensity of the light becomes too little , and the conditions cannot be kept steady enough for an exposure of some hours , which would be necessary .
The shape of dispersion curves does not seem to be affected by the absorption regions in the orangered and the extreme red .
The curves observed seem to keep their character , and only lose the intensity of the light .
This fact points to a difference of mechanism in the vibrating systems corresponding to the lines in the region of which dispersion takes place and those regions where absorption takes place without appreciable deviation , except that corresponding to the lines of the other type .
These absorption regions are seen to consist of a very large number of fine lines , if sufficient resolving power be used , as has been shown by Carter , who has pointed out that the orange-red absorption region for potassium is very similar to the corresponding region investigated by Wood for sodium .
* When a very considerable amount of dispersion appears at the red rays , the dispersion begins to show itself at the violet rays 4047 , 4044 .
It is difficult to follow the effects near these lines visually , as they are near the end of the visible spectrum , but similar phenomena can be observed , and it is easier to obtain photographs showing similar curves to those at the red pair .
For the round these lines , and also about other lines of the principal series in the ultra-violet , the quartz spectrograph used for work described in a former paper communicated to the Royal Society was employed and I take this opportunity of again thanking the Government Grant Committee for enabling me to purchase this instrument .
The arrangement of apparatus was practically the same as for the visible spectrum , the quartz lens already described being used .
Fig. 4 shows an enlargement of the dispersion curves found in the region of the lines 4047 , 4044 , fig. 3 , being the same photograph without enlargement .
Here the region between the lines is clearly visible , though now the difference of wave-length is only 3 ngstrom units .
Thus , with considerable density of vapour\mdash ; at this stage the red end of the spectrum has disappeared nearly down to the position of the lines\mdash ; giving a considerable deviation of light near the lines 4047 , 4044 , there is very little absorption in between these lines ; ' Physical Review , ' vol. 27 , 1908 .
I was not aware of this work by Carter when I communicated a paper to the Physical Society of London on the absorption spectrum of potassium ( ' Phil. Mag January , 1910 ) .
The similarity there discussed between the spectra of potassium and sodium had been observed and commented on by Carter before that paper was written .
'Roy .
Soc. Proc 1910 , , vol. 83 , p. 421 .
VOL. LXXXIV.\mdash ; A. first seven lines of the series , or , rather , pairs of lines .
Figs. 5 and 6 show the curves at the lines 4047 , 3447 , 3217\mdash ; the effect is very small at the last of these , and smaller still at the next three lines .
It is , however , certainly observable in certain negatives when examined carefully under a microscope .
Measurements have been made on the amount of the deviation , with a view to finding the relation between the constants for the Sellmeier dispersion formula which is applicable to the case .
In this way there seems some hope of obtaining information as to the vibrating systems to which these lines are due , and hence as to the constitution of the atom .
We shall first consider the measurements of deviation obtained from photographs of the type of fig. 5 , deferring the consideration of the more detailed measurements about the lines 7699 and 7665 , and 4047 , 4044 .
To obtain photographs of this type , dense vapour has to be used and the red end of the spectrum is completely absorbed , and also the light between 4047 and 4044 disappears .
The curves obtained can then be regarded as corresponding to a set of single lines in the mean positions of actual pairs of the principal series .
The quantities to be measured were the displacements of points due to light of particular yths owing to the deyiation produced by the vapour prism from their positions in the undeviabed spectrum .
This would be simply proportional to the deviation of the rays if the effective distance of the photographic plate from the slit of the spectrograph were the same for all colours .
But with the quartz spectrograph a considerable tilt has to be given to the plate in order that various parts of the spectra should be in focus together .
We therefore have the images of the slit of continuously varying length along the spectrum , increasing from the violet to the red .
This change of length of S the image of the slit is considerable .
If the length of the image of the slit for light of wave-length 2400 be 1 that for light of wave-length 6000 is about Measurements , therefore , of length perpendicular to the spectrum have to be corrected owing to this circumstance .
The relation between the length of 1910 .
] Dispersion of Light by .
215 image of the slit and the distance from some origin on the plate is not exactly 3S a linear one , as the plate is bent slightly in the holder , the surface of good focus not being exactly a plane , ut this made a small effect which could be : : .
neglected , and as far as the present measurements are concerned the continuous spectrum due to any point the slit could be regarded as a straight line in the negative .
This approximation was useful , as measurements of the displacement could be made from the image of a speck of dust on the slit .
The actual method adopted was to rule on the negative with a needle point a fine line parallel to the image of the speck of dust and make measurements from this line .
The reason for not using the dust line directly that it was not continuous enough .
Two sets of corrections were made to these observed lengths\mdash ; firstly , owing to the fact that the ruled line was parallel to the dust line , and hence , owing to the tilt effect , could not be parallel to the undeviated spectrum ; and secondly , to the tilt effect on the actual measurements .
Curves of corrections were drawn from measurements on the breadth of a continuous spectrum\mdash ; the dispersion increasing in the violet makes the relation between length of image of the slit and wavelength not linear\mdash ; these correction curves gave a quantity to be added to and a factor to multiply each measurement of deviation .
It was no use attempting to photograph the undeviated spectrum on the same plate as the deyiated and measure from this directly , as the deviations measured in the smaller wave-lengths were so small that comparison of the two spectra rendered this method impossible .
The actual readings were made with a micrometer microscope reading directly to mm. Measurements on negatives of the type represented in fig. 4 could not be made with any accuracy for regions corresponding to wave-lengths gleater than about 4500 \mdash ; the line on the red side of this position being very much broadened out losing all sharpness .
Measurements close to the positions occupied by the absorption lines at wave-lengths 3100 and 3034 are also of not much value .
The lines become blurred close to the absorption region , and the measure- ments tabulated can only be taken to mean that the anomalous dispersion eHect is really observable .
Anomalous dispersion could be observed at the next line , 2992 , but no quantitative estimate could be made .
The measuuements finally corrected were not from the undeviated line , but from an arbitrary base line .
The position of the undeviated line was first obtained from the negative , and then a corrected value found for its position relatively to the arbitrary line by taking an arbitrary set of points and applying the dispersion formula to be discussed later .
The positions of the undeviated spectrum determined in this way agreed to within cm .
For measurements quite close to absorption lines a set of was ruled Prof. P. V. Bevan .
[ June 7 , on the negative parallel to the slit images , and the wave-lengths corresponding to these lines determined in the ordinary way .
In this manner readings close to the absorption line could be obtained .
At other places known lines in the arc spectrum were used , and also specially ruled lines .
On the negatives used for measurement there was a sufficiency of lines extending across the spectrum , due to diffused light from the arc , to determine with sufficient accuracy the wave-length corresponding to any position on the dispersion curves .
It was found that the deviations on any one negative were , at different wave-lengths , proportional to those on any other .
The following table gives the deviations for the wave-lengths in the first column for the best negative:\mdash ; The actual measurements were made to one more decimal place than is recorded ; but as the line on which the cross wire was set was not very definite , this last figure has been omitted , and the nearest number to the mean of the observed values tabulated .
The error in any of the numbers is not more than as far as observational errors are concerned .
In the curve ( fig. 7 ) these values are plotted , and this represents the complete dispersion curve for the vapour from the region about 5000 to the ultra-violet 2400 .
The observed values of the deviations are represented by dots about the curve .
The deviation being equivalent to that due to a thin prism is propor1910 .
] Dispersion of Light by Potassium Vapour .
tional to , where is the refractive index for light of a particular wavelength .
The Sellmeier dispersion formula may be written where is the wave-length.of the light corresponding to the refractive index , and , etc. , are the -lengths corresponding to absorption lines , , etc. , being constants .
For a first approximation for our case we can take the dispersion as that corresponding to two absorption lines at wave-lengths 7682 and 4045 , corresponding to the means of the two first pairs of the principal series .
It is clear from the curve of fig. 7 that the effect of the absorption lines of smaller wave-length is only noticeable in their immediate neighbourhood , so that for all except such wave-lengths the assumption of the two constants for our formula will be near enough to the actual circumstances of the case .
We afterwards obtain an estimate of the value of the constants corresponding to the higher lines of the series .
Now as is nearly equal to 1 , may be taken as equal to 2 , so that our curve should represent values proportional to , ( 1 ) except for values of quite near and We may therefore take , for applying the Sellmeier theory to our case , S , the deviation at any wave-length , as equal to the expression just written down .
The observed deviation was in the first instance measured from an arbitrary base line parallel to the undeviated position of the linear Prof P. V. BevalL [ June 7 , spectrum .
A little uncertainty remained as to the actual position of the undeviated spectrum , so that calling the observed value of the deviation measured from the roughly estimated position of the undeviated spectrum we have when is a small quantity to be determined which is constant for all wavelengths .
The ratio was found by taking points on the curve at which the deviations are equal .
These points must , of course , be on opposite sides of the line 4045 .
Such a pair of points are at and .
In this way , taking several points , we obtain was then found from pairs of values of .
These gave for a mean value of , so that .
We can then find and so the corrected position of the undeviated spectrum .
With these values of the constants we obtain for , by substituting in the following values at the specified wave-lengths:\mdash ; All the deviations are ative , that is , the light in the whole of the region ated travels with a velocity greater than that in vacuo .
These points are marked on the curve of fig. 5 by small crosses .
It is seen that these values for , give very fair agreement between the observed and calculated values long as we are not quite near the other absorption lines .
If we add the term to and take deviations at wave-lengths near 3447 , we can obtain a value of .
From the observed deviations at wave- lengths 3453 , 3466 , 3476 , 3444 , 3441 , we obtain .
For wave-lengths further removed from the absorption line than these we 1910 .
] Dispersion of Laght by Potassiscm Vapour .
219 can get no observational errors are too then can be taken to be 11 For the next line we have two observations which just give an estimate onstant fengths 3they gThese numbers cegarded aiving trder oagnitude oThese observations .
For the next line we cannot even get an estimate of the nitude of the constant .
We have then finally the values and the complete dispersion curve represented by .
( 2 ) We may now return to the more detailed examination of the dispersion in the region of the red and violet lines .
Fig. 1 shows a series of photographs with increasing density of vapour of the dispersion curves in the red .
Fig. 2 represents enlargements of the series of fig. 1 .
For representing these curves by the Sellmeier formula we must now use two constants for corresponding to the two 7699 and 7665 .
Calling these and and the corresponding constants the approximate value of for this should be , ( 3 ) the terms in ( 2 ) depending on , m3 , etc. , are too small to be taken into account in this region .
We could find the ratio of : by finding the point at which the fine dispersion live between the two absorption lines cuts the undeviated portion of the spectrum , for at this point , so ( 3 ) gives : ? ?
directly ; but the difficulty of obtaining any one wave-length accurately from photographs in this region makes this method uncertain .
It may be mentioned here that point is invariably the same whatever be the density of the vapour , as of course follows if the Sellmeier formula represents the facts .
This may easily be observed by fixing the cross wires of the telescope on this point for a low density of vapour , and then dually heating the tube to obtain denser vapour .
The dispersion curve spreads upwards and downwards , but the fine bright line between the absorption lines always passes through the same point on the undeviated position of the spectrum .
1910 .
] Dispersion of Light by Potassium Vapour .
scale than those corresponding to denser vapour , but it shows the dispersed light nearer to the absorption lines .
By drawing curves for each photograph , the relation of scale for the series can be found , and thus a factor obtained for each to reduce all to the same scale .
The factors were 25 , , 1 , that is , the measured deviations in the photograph with least dispersion were to be multiplied by 25 , and so on , to bring all to the same scale as the one with greatest dispersion .
These factors were determined from several points on parts of the curves represented on two photographs , and the numbers obtained reed as well as could be expected .
For example , the six numbers for the factor to multiply deviations on the second photograph to make it comparable with the third were , with a mean of .
The curves for the first and second did not agree very well in the region with shorter wave-length than 7666 , but in this region , on all the Taphs , there was considerable spreading of the line , so that it was difficult to estimate the exact position of the intersection of the ruled line and the deviated spectrum .
The four sets of values thus obtained are plotted in fig. 8 , and the dispersion curve drawn by means of them .
Certain points are not represented on the curve , as they would require the rest of the curve to be on too small a scale .
The following table gives the reduced deviations corresponding to the wave-lengths in the second column .
The number at the side of the wave-length refers to the number of the photograph to which it belongs in the series:\mdash ; Prof P. V. Bevan .
[ June 7 , There is a good deal of uncertainty in the measurements of the wavelengths , the distance apart of the two absorption lines at 7699 and 7666 being almost exactly mm. , so the determination of a particular wavelength may have a considerable error .
The numbers representing the deviations were measured in centimetres and divided by 3 , for convenience in plotting .
They are only expressed to two decimal places , as the errors in wave-length determinations do not warrant further accuracy .
The decimal point is omitted on the curve .
A similar curve is shown in fig. 9 for the dispersion in the neighbourhood of the lines at 4047 , 4044 .
This curve is from a single photograph ; composite curve was not constructed , as in this case the two absorption lines are so close together that the instrumental appliances would not admit of sufficient accuracy in the wave-length determinations to combine valious curves .
The deviations shown in this curve are not measured from the undeviated position , but they represent the deviation due to the two lines 4047 , 4044 .
In other words , they represent the quantity and for 4047 and 4044 , and for the quantities corresponding to The obseryations for the two curves representing the dispersion in the region of the red and violet lines are not accurate enough to calculate with any certainty the constants for the appropriate Sellmeier expressions .
1910 .
] Dispersion of Light by Potassium .
223 There appears to be an equality , indicated in the tios m : and : but at present the uncertainty in the numbers makes this not very definite .
The curves are of the type represented by the equation , but as a slight change in wave-length affecting all the observations would make a considerable change in the constants , I have decided to leave for the present the further investigation of these curves until measurements under greater dispersion can be obtained .
There is a deal of interest in the l.atios : and I hope to be able to obtain more reliable curves and similar ones for other metals of the alkali group , so as to make a comparative study of the lines and the dispersion in their region for the whole group of metals .
At present , the interest lies in the conclusions we can draw from the general dispersion curve .
On Drude 's representation of dispersion , if we assume that the vibrating systems , which give the natural periods of the vapour , are simple negative electrons vibrating about an equilibrium position , the quantities of the general dispersion expression will be proportional to when is the number of electrons per cubic centimetre of the type givin rise to of wave-length .
If we suppose each molecule of the vapour to be capable of absorbing or light of , then the number of special corpuscles per molecule is proportional to .
This requires far too many electrons per molecule , and we are driven to the conclusion that it is not every molecule which takes part in the absorption of light in the same way .
This is on the assumption that the ] ecGrons are vibrating about equilibrium positions , but it is clear that if the vibrations are about a steady state of nJotion , a similar conclusion will follow .
The quantities will depend on the number of electrons which can have a certain definite oscillation about the steady state , and the enormously greater value of for the first pair of the principal series lines and its rapidly diminishing value through the series show that it is impossible that in each molecule of the vapour there can be sufficient electrons to provide all the natural periods .
We can obtain an estimate of the number per molecule on Drude 's theory for the case of sodium , for which Wood has determined absolute values of the refractive indices at a definite temperature of the vapour by an interferometer method .
We do not know the actual density of the vapour at this temperature , but we can obtain an estimate of it .
Jewett*has measured the vapour pressure of sodium for temperatures up to 42 C. The boiling point of sodium is estimated to be 80 C. , so that by a very rough interpolation we find the vapour density to be at Wood 's temperature , 64 C. With this value and the value of , as determined by Wood , , we find on 'Phil .
Mag 1902 , vol. 4 , p. 546 .
Prof. P. V. Bevan .
[ June 7 , Drude 's theory*that the number of electrons per molecule is .
This , of course , gives us only the order of che number , as the numbers we have used are really only guesses , and as we have assumed that the atom of sodium is identical with the molecule .
However , the number per molecule corresponding to the absorption line is at any rate small .
A similar state of things may be expected with potassium , so that with our ratio roughly , only about 1 molecule in about 2000 can be in operation in producing absorption at the lines 4047 , 4044 .
Fewer still will be taking part in the absorption of lines lower down in the series ; and although it appears that the ratio of successive 's dimimshes , stillvery few , comparatively , can be taking part in the absorption corresponding to numbers near the limit of the series .
We are thus led to look for the explanation of the natural periods corresponding to the series lines , not to the atom itself , but to the atom as differentiated possibly for each period of the series .
The identity of structure of the atom must on this view be given up , as far , at any rate , as its properties as regards light are concerned .
A estion of Sir J. J. Ihomson is interesting in this connection .
He has pub forward the view that lines in the spectrum of an element which show the Zeeman effect are due not to vibrations of corpuscles in the normal atom , but to vibrations of corpuscles : which form semi-permanent combinations with the normal .
He regards these corpuscles as forming swarms round the atom when the element is emitting light in an arc or in a flame , and so providing plenty of opportunity for temporary combinations with the normal atom .
It is clear that there could be many types of temporary combination , and thus in a metallic vapour we could have normal atoms with varying numbers and arrangements of corpuscles attached to them .
It will be to these arrangements , which will occur with relatively only a few of the atoms , that we may have to look for the explanation of the lines in spectra .
Another view that leads to somewhat similar conclusions is that of Wood , who supposes that the principal series lines are due to atoms differing in that they have lost different numbers of negative electrons .
The argument for this is that there seems to be independence between the principal series lines of sodium in the phenomena of On Drude 's theory the constant of the Sellmeier formula is , where is the nuluber of electrons per unit volume giving rise to a Particular period and is a constant connected with the wave-length and constants of the electron by the relation We hence obtain , since are known\mdash ; assuming that a negative gives rise to the vibration\mdash ; N , the number of corpuscles per unit volume .
This is , for the case of sodium and the lines , .
The number of atoms of sodium per unit volume in the vapour at the temperature at which Wood 's determination was made is approximately .
Hence we would have about twelve atoms to electron emitting the line .
'Phil .
Mag 1906 , vol. 11 , p. 774 .
* 1910 .
] Dispersion of Lighf by Potassium Vapour .
225 $ fluorescence .
It is difficult on this view to see how an atom could lose sufficient corpuscles to provide enough different systems for all the lines of a series .
But in any case it appears that all the atoms cannot be in operation , and that we have a real difference of physical properties indicated .
One would expect that on either of these views an increase in the ionisation of the vapour would produce an effect on the number of atoms exercising a particular type of absorption , and so an effect on the dispersion of the vapour .
But in any case , in the vapour of these metals there is a large amount of ionisation , and the increase due to , say , ultra-violet light may not be sufficient to produce any observable effect .
I have tried the effect of cutting off most ultra-violet light from the vapour by a glass plate interposed in the path of the train of passing through the absorption tube , and no visible effect could be observed in the dispersion in the visible part of the spectrum .
This view that the lines , at any rate of the principal series , in spectra are due to special atoms simplifies the picture we must make of the atom itself .
We no longer need to provide such an enormous number of degrees of freedom for each individual atom , and we may be able in this way to simplify some of the difficulties in the theory of the partition of energy .

