]\gt ; INDEX SLIP .
J. U. M.\mdash ; Colours in Metal Glasses and in Metallic Phil. Trans. , , vol. 203 , 1904 , pp. S86-420 .
" " AlloLropio\ldquo ; tution , .
Origin of Colour of .
, J. C. M. Phil. Trans. , , vol. 20$ , 1904 , pp. $86-42( Coloration of Glass by Reduced Metals ; Action of Radium .
GARNBTT , J. C. M. Phil. Trans. , , vol. 203 , 1904 , pp. 886-420 .
Coloured Metallic Films\mdash ; Cause of Change on Annealing .
OhXNBTT , J. C. N. Phil. Trans. , , vol. 203 , 1904 , pp. Nascent Crystals of Metal , Sphericol Form .
GARNRTT , J. C. M. Phil. Trans. , , vol. 208 , 1904 , pp. 385-420 .
[ 385 ] XII .
Colours in Metal Glasses und in Metallic J. C. MAXWELL GARNETT , B.A. , Trinity College , Canvbridge .
Communicated Professor J. LARMOR , Sec. R. Received April 19 , \mdash ; Read JnIlC 2 , 1904 .
Introduction . .
THE present paper contains a discussion of some optical properties of a medium containing minute metal spheres .
The discussion is divided into two Parts : the first Part dealing with colours in metal glasses , in which the proportion of volume occupied by metal is small ; the second Part dealing with metal films , in which this proportion may have any value from zero to unity .
In Part I. the observations of SIEDENTOPF and ZSIGMONDY beyond the limit of microscopic vision ( ' Ann. der Phys. , ' January , 1903 ) are discussed .
It is shown that the particles seen in a gold ruby glass are particles of gold which , when their diameters are less than , are accurately spherical .
I have endeavoured to show that the presence of many of these minute spheres to a wave-length oflight in the glass will account for all the optical properties of\ldquo ; egular \ldquo ; gold ruby glass , and the ularities in colour nd in polarisation effects sometimes exhibited by gold glass are due to excessive distance between consecutive gold particles or to excessive size of such particles , the latter , however , involving the fonner .
It is also shown that the radiation from radium is pable of producing in gold glass the ruby colour which is generally produced by re-heating .
The method adopted enables us to predict from a knowledge of the metal present in metallic form in a glass what colour that glass will be in its " " regular\ldquo ; state .
In Part II .
the optical properties , and the changes in coloul on heating , of silver and gold films observed by Mr. G. T. Proc , and of the potassium and sodium films deposited on glass by Professor R. W. WOOD ( ' Phil. Mag. , ' p. 396 , 1902 are discussed , with a view to .
that they call be accounted for by supposing the films to be composed of minute metal spheres of varying sizes .
V0L .
CCIII .
PART I. S2 .
Consider the incidence of light of wave-length on a sphere of metal of radius .
Suppose the constants of the metal relative to the surrounding medium , which we may first suppose to be aether , are , the coefficient of refraction , and the coefficient of absorption .
Let us write . . . . .
( 1 ) , where , as usual , denotes We shall use the following notation to denote the electric vector:\mdash ; Incident light . . . . .
Transmitted reflected light .
Here being the velocity of light HERTZ ( ' Ausbreitung der electrischen Kraft , ' Leipzig , 1892 , p. 150 ) has shown that the electric and netic forces at any point due to an oscillating electric doublet of moment ' along the axis of are given by ( 2 ) , ( 3 ) , where for these expressions satisfy MAXWELL'S equations curl curl and and when .
is very small compared with the wave-length of the emitted waves the expression for reduces to , which is at any time the electric force which would be electrostatically due to the doublet if its moment remained constant and equal to its value at that time .
Lord RAYLEJGH ( ' Phil. ' XLIV .
, pp. 28-52 , 1897 , and 'Collected Papers , ' vol. 4 , } ) .
has extended this theorem to the case of a very small sphere .
In the region which the distance , , from the centre of a small sphere of radius excited COLOURS IN METAL GLASSES AND IN FILMS .
.387 by an field , is small compared with the , the electric force due to the sphere is .
By comparing this with HERTZ'S corresponding result for an oscillatin doublet of moment , as given above , it appears from ( 2 ) and ( 3 ) that the electric and magnetic forces at any point , due to waves emitted by the sphere , must be given by the equations , where now Replacing by , where is the quantity defined by equation ( 1 ) , we conclude that when a metal sphere is excited by a periodic electric force , it emits the waves which would be emitted by a Hertzian doublet which at time was of moment equal to The same result can be proved directly by adapting the analysis given by L. LORENZ Selsk .
Copenhagen , to the electromagrnetic theory .
The problem has also been treated by STOKES ( ' Camb .
Trans vol. , p. 1 , 1849 , and ' Papers , ' vol. , p. 245 , p. 262 ) .
At a great distance from the origin , , when / is great compared with equation ( 4 ) reduces to [ cf. .
cit. , equation ] ( 6 ) .
If we transform to spherical co-ordinates in the respective directions of increase of ( fig. 1 ) we obtain , at a great distance from the origin , .
( 7 ) .
It appears from equations ( 6 ) or ( 7 ) that such a small sphere , in common with any other minute system whose moment is proportional to the electric vector of the incident light , emits light with an intensity proportional to the inverse fourth power of the wave-length , provided that is independent of .
It is this property which , as Lord RAYLEIGH has shown , accounts for the blue colour of the light received from the sky .
Fig. 1 .
Fig. 2 .
S3 .
In the ' Annalen der Physik ' for January , 1903 , H. SIEDENTOPF and R. ZSIGMONDY publish some observations on the metal particles in gold ruby glasses .
By their method of illumination they were able to see particles whose dimensions were of the order of from 4 to 7 , where represents millim. The arrangement consisted of a system of lenses following a strongly illuminated and very narrow slit .
The system of ]enses , of which the last is a low power microscopic objective , serves as a condenser and forms a very narrow image of the slit inside the glass under observation .
This image of the slit may not be more than one or two wave-lengths thick .
The observation is made with a microscope having the tube perpendicular to the incident light , so that only the light emitted by the metallic particles travels up the tube .
This is the light the electric vector of which has been distinguished by the suffix unity in the preceding analysis .
The image of the slit , which is parallel to Ox in fig. 2 , comes directly under the microscope tube , which is in the direction Oy ; thus only the particles illuminated at the image of the slit send light up the tube .
The diffraction discs do not pile up on top of one another if the average distance between two metal particles is greater than the thickness of the image of the slit .
In this case , then , the number of particles per unit area can be counted .
On pp. 11 and 12 of the paper referred to , SIEDENTOPF and ZSIGMONDY discuss the appearances in the second focal plane of the microscope when the light incident in the glass is plane polarised .
The figs. 3-6 above are reproduced from their paper .
In fig. 3 the plane of polarisation of the incident light was that of incidence , the plane of incidence bein the plane containing the axis of the microscope and that of COLOUIffi IN GLASSES AND IN METALLIC FILMS .
the incident pencil of light ; in figs. 4 and.5 the plane of polarisation of the incident light was inclined at to the plane of incidence ; while in fig. 6 the two planes were perpendicular .
In the figs. 3-6 the upper represents the second focal plane of the microscope when the diameters of the particles of in the glass are less than , the small lines being parallel to the planes of polarisation of the emergent light in various parts of the field , the " " emergent \ldquo ; here meaning the light sent up the microscope tube by the metal particles in the glass under observation .
The Fig. 5 .
Fig. 6 .
Fig. 3 .
Fig. 4 .
lower diagrams in the same figures represent the appearances of a action disc for the same respective positions of plane of polarisation of the incident light .
It is to be noticed that the emitted in any particular direction comes to a focus at a corresponding in the second focal plane of the microscope .
Consequently a black spot that means that no light is in the ) onding direction .
lf all the particles are spheres sending up no light in some .
direction , there will thus be a black spot in the second focal plane , as well as in each diffraction disc , at the point corresponding to that direction .
Suppose now , as in S 2 , that the incident light travels in the direction Oz is polarised in plane , fig. 1 .
Instead of conceiving plane to alter as we consider the various cases of figs. 3-6 , we shall imagine the microscope tube to the plane Thus in fig. 3 the microscope is along Oy , in fig. 6 along Ox , , while in figs. 5 and 4 the tube lies in the intermediate positions , namely , respectively .
It will now be shown that the figs. 3-6 are completely accounted for if the particles are spheres small compared with a wave-length , i.e. , appreciably smaller than From equations ( 7 ) the character of the light emitted by such a sphere in the direction ( fig. 1 ) is determined by the electric force whose composition is : ( 8 ) , where Suppose first that , as in the case corresponding to fig. 8 , the microscope tube is along Oy ( fig. 1 ) , the centre of the field then corresponds to The fig. 7 represents the direction of , as deduced from equation ( 8 ) , for positions , the co-ordinates of which are , the centre of the diagram corresponding to Fig. 7 .
Fig. 8 .
, the axis of .
The same figure will therefore represent the directions of the electric vector in various parts of the second focal plane of the microscope .
From equation ( 8 ) it appears that when either or , we shall have , and therefore becomes and only has a component in the direction This is repl.esented by ) arrows for positions on the axes in fig. 7 .
In the middle of the quadrants the directions of the electric vector are no longer parallel to the axis of but are tilted as in the figure , being tilted in the same manner in opposite quadrants .
Now the planes of polarisation are perpendicular to the electric vector , and the small lines in fig. 3 are perpendicular to the arrows in fig. 7 .
When , therefore , the incident light is polarised in the plane of incidence , the appearances are accounted for if the particles are small spheres .
Next consider the case corresponding to fig. 6 , when the microscope tube is above Ox. The centre of the field is then .
The arrows represent the direction of in various parts of the field .
All these arrows point nearly towards the centre .
Along the two axes they point accurately towards the centre .
There is no force at the centre , for then both and vanish .
Consequently , a black spot should appear at the centre , if the particles were spheres .
Finallyj lines perpendicular to the arrows in fig. 8 are parallel to the lines in fig. 4 .
Consequently , in this case also the appearances are explained by supposing the particles to be spheres .
COLOURS IN METAL GLASSES AND IN METALLIC FILMS .
In this case , namely , when the incident light is polarised perpendiculal to the plane of incidence , it further appears that if an analysing nicol be introduced so as to polarise the emergent light in the plane of incidence , then the analysing nicol removes the component of and the vanishing of also , for causes a dark band to cross the field over the diffraction disc if there be only one particle sending light up the tube , the dark band lying along the axis of in fig. 8 , i.e. , in the plane of incidence , and this also was obseryed by SIEDENTOPF and for the particles in gold glass ( loc. cit. , p. 12 ) .
The discussion of the cases of and 5 presents no difficulty .
The phenomena , including the correct position of the black spot , are again explained , by means of the hypothesis that the small particles are spheres .
Thus all the phenomena , observed in the second focal ) lane of the microscope , due to particles smaller than , are exactly those which would be produced by spheres of metal of radius small compared with the length of a wave of light in the glass .
If now the particles were small spheroids , or in structure , then the position of the black spots , if indeed any existed , and the positions of the plane of polarisation of the light elnitted from the particles , would depend on the orientation of the particles .
Unless , therefore , the orientation of all the particles were the same , we should , if many particles were sending light up the tube , get no black spot in the focal plane , because the black spot , supposing to be one , due to one particle , would not coincide with that due to another .
And further , even if the orientation of all the particles were the same , and if every particle alone did send off no light in some particular direction , so that there were a black spot in the second focal plane , then , unless the common orientation were such that , for every plane of polarisation of the incident light , the black spot were in the same plane as if the particles were spheres , which is an impossibility , spheroidal or crystalline particles could not account for the effect observed .
These considerations show , therefore , that the small particles in gold ruby glass are really spheres of gold , so long as their dimensions are considerably smaller than ( centim. ) .
This result is of considerable interest in connection with the formation of crystals .
When a metal crystallises out of a vitreous solution , it appears that until the dimensions have increased beyond a certain limit , the forces of surface tension overcome the crystallic forces , and the particles of metal are spherical and not crystalline .
Mr. G. T. BEILBY has arriyed at the same conclusion from microscopic examination [ Note added 14th May , 1904.\mdash ; The presence of crystals , whether of silicates or of reduced , in many pottery glazes suggests that minute spheres of the same material as the crystals were before the formation of these crystads , and that some may -exist with the crystals .
The colours the may therefore be wholly or in part due to the )resence of these minute spheres , in the manner a gold ruby glnss depends for its colour on the presence of minute spheres of gold .
] of the films of metal deposited from solutions ( ' Proc. Roy .
Soc vol. , 1903 , p. 223 ) .
In the manufacture of gold and copper ruby glasses and of silver glass , the gold or copper or silver is mixed with the other ingredients of the glass before the first firing .
If , when the glass is formed in the furnace , the whole be quickly cooled , the glass with the metal in it is colourless and exactly resembles clear glass .
I have had in my possession several pieces of such clear gold glass , and some of clear silver glass .
One of the former was used in an experiment with the emanation from radium , to be described later .
In this clear glass the gold or silver is probably in solution in the glass .
But when the glass is re-heated the metal ' crystallises\ldquo ; out of solution , or , as we shall say , is " " excreted\ldquo ; from the glass and appears in the small particles obseryed by and ZSIGMONDY .
These particles of metal , as we shall show , account for the colour of the glass .
I have seen a copper glass which was allowed to cool down slowly in the glass pot along with the furnace , taking a week or more in the process , The glass formed a dark brown , nearly opaque , mass with minute crystals of bright , shining copper scattered throughout its substance , the crystals being large enough to be easily distinguishable with the naked eye , while the appearance of the whole mass somewhat resembled that of the well-known African stone , ayenturine .
It is suggested that the second heating , without melting the glass , confers sufficient freedom on the molecules of the glass to enable the forces of surface tension to exert themselves in bringing the molecules of the metal , which have been distributed amongst those of the glass , together into heaps , the phenomenon bein similar to that exhibited when a metal film is heated to 30 or 40 without being melted , when , as will be described later , the metal fornoe itself into minute granules , which , in the light of what we have proved for the particles in gold glass , must be spheres or spheroids with axes normal to the film .
The latter form is possible for the films of metal , though not for the metal in the gold glass , because a thin film , as opposed to a piece of glass , is not subjected to similar conditions in all directions .
S 4 .
We have thus to consider the problem oflight travsrsing a medium containing many small metal spheres to a wave-length oflight .
It has been seen ( S 2 ) that a small metal sphere produces in all surrounding space the same effect as would be produced by a Hertzian doublet placed at its centre .
We may therefore imagine the spheres replaced by such electric doublets and thus avoid considering their finite size .
Let the average ( for a large number of doublets ) moment of a doublet be , at time Then if there be spheres per unit volume , the polarisation of the medium will be .
If , due to the regular force together with forces due to the COLOURS IN METAL AND IN METALLIC doublets , be the force ' the polarisation , then we have proved ( S2 ) that Now by means of the vsis given by H. A. ( ' Wied .
Ann. , ' 9 , 1879 , p. 641 ) and by Phil. Trans , and which has been fully verified in own paper and by others , it can proved that ( see S 7 below ) ( 9 ) , provided the medium under consideration extends throughout a space of dimensions which in no direction are of an order of magnitude so small as a wave-length of light .
This provision is satisfied except in the case of very thin films .
When dealing with such films in a later portion ( S7 ) of this paper we shall to the consideration of this point .
From equation ( 9 ) obtain ' so that CLERK MAXWELL'S equations ] with units for this medium , now , therefore , are curl and curl where We have therefore proved that a medium consisting of small metal spheres distributed in , many to a of light , is optically equivalent to a medium of refractive index and ption given by where We shall throughout use the symbol to denote the volume of metal ) unit VOL. CCIII . .
3 volume of the medium ( except when is evidently used to denote the thousandth part of a millimetre ) .
Thus ) ) , and equation ( ) ecomes ( 10 ) .
If the metal ) be ) in glass of active index instead of this equation ecomes . . . .
The constants and of the ledium thus depend only on , the elative volume of metal , and not ] the radii of the individual spheres .
It is cleal that the spheres may now be posed to be of quite various adii , only that there be many to n wave-length of light in medium .
We have given the esult which holds for all values of , as we shall require it later .
But in the case of metal glasses , by which name we shall describe glasses in which a metal is present in metallic , the value of varies from about a silver ] down to about 10 a soda glass coloured by adium .
The lasc equation giving optical constant of the metal glass may be writlen , say , where is the optical constant of the metal and the index of refi.actiou of the glass by itself S5 .
Equation ( 12 ) may now be written .
Thus , equating real and imaginary parts , we find , after some reduction , .
( 13 ) .
th , 1904.\mdash ; This equation may be written COLOURS IN IETAL GLASSES AND IN TALLIC FILMS .
We have now to see whether by means of these equations ( 13 ) , and of the values of and for various metals , we shall be able to predict the colour of a glass which contains a number of small metal spheres , whose linear dimensions and distances apart small of visible light .
In the annexed table the refractive index of the glasses has been taken to be The values of and of for the metal are those by DRUDE Physikalische nuary , , for yellow centim and for red hght centim For the potassium-sodium amalgam , however , blue and yellow light were used instead of yellow and red .
Now let us suppose that , the quantity of metal per unit volume , is very small If , then , and represent the numbers in the penultimate and last columns of I. respectively , we have Hence Hence , neglecting higher powers of . . . . . . . . .
( 14 ) .
, suppose that light of wave-length A travels through this composite medium , hose constants are and .
The light being given by , in this medium it is given by so that measures the absorption .
In fact , the intensity of the light sinks to ( nearly ) of its original value in traversing a distance ( 15 ) of the medium .
We have now to apply the to the observations , order to test the validity of our analysis as ' the actual phenomena .
SIEDENTOPF give Physi k January , a table of ious gsses examined by them .
This table is reproduced in Table II .
IN IETAL C AND IN METALLIC FILMS .
If denote the volume of the gold particles , so that is given in the 8th or 9th column of the Table II .
, according as , the gold content , is taken from the 6th or 7th column , then the number of gold particles per cubic ( centim. ) is given by We said that for our analysis to be applicable there must be many spheres to a waye-length .
Since the spheres in glasses , Ca , ( Table II .
) can be separated by a eiss objective , they must be at a distance apart greater than ' half a wave-length of violet light .
We shall therefore not expect our analysis to apply to them .
It is further apparent that the particles are more widely separated Da thal ] Db .
If we take for the mean of the two numbers given , we have so that , since , and are larger than and , the glasses satisfy our condition best .
we are presented with a difficulty .
The wave-length of the yellow light , centim. , is in our glass only , centim. or .
Thus to find the number of gold particles in , we must multiply by .
We shall thus , even for glass , have less than one particle to a -length .
On the other hand , NTOPF warns us that the linear dimensions of the are only to be taken as upper limits and may be three times too large .
Suppose this is ] case , then the number of particles in a yellow -length in the glass is times the aboye numbers , , with of course a still greater value for red light .
On this hypothesis then the alone of the series satisfy our condition .
If , therefore , the theory is correct , it should explain the colour and other optical properties of these three glasses as set out in the first five columns of Table II .
Let us , for instance , consider the colour of glass G. From equation ( 15 ) we have as the distance in which the intensity of light of wave-length is reduced to of its original value given by ( 15 ) .
From the Table I. we have , supposing , , for centim. , for red light The value of found by colorimetry is 72 . .
Consequently we have Yellow , 48 centim. , nearly .
Red , centims. Since the latter number is greater than the former , it follows that this glass , and COLOURS IN METAL AND IN METALLIC FILMS .
indeed all ] asses for condition is satisfied , should be much more red than yellow .
] , therefore , they are still more yellow than green , and more green than expect the gold glasses )hich satisfy our condition , to be ils in fact The above vaJues for are of the ight order , but they may be somewhat too large .
If we ] taken the value of , called the Gold Content in Table the spoon ding values of old 1 been only half those gi above .
It is to be , in gold ] flashing\ldquo ; 011 to clear glass , use nor gold .
A collYlon yalue for the total gold content is means of equation I. we in this way whether a gold , silver , copper glass hich will t more red more yellow light , and whether sucl ] a glass containing small of " " potassium-sodium\ldquo ; ] nsluit more yellow more blue .
We thus find that ] } there metal sphel.es to a Silver glass nits yello ( yellow red ) , , , , , ( red yellow ) , Gold , , , , , , , Potassium-sodium lass tnsmits blue blue yellow ) .
From the values of on Table I. , , we see that for a silvel ' glass as much red light as a gold glass does would to ) , or , oughly , eight times as great for the glass for the gold .
And , since the values of yellow red light are more nearly equal for silver than ' gold , in order to produce the coloration there ) be eve more tlJan eight times as much silver volume ) gold .
1 am told put in ten times as much silver by ei ht into a silver glass as they put gold into a gold glass .
Again , the very large value of yellow light in a potassium-sodium glass shows that such a glass would absorb as much yellow light as a gold glass with only of the amount of metal excreted .
Thus a very slight excretion of the potassium-sodium metal would give a very strong blue or violet ation .
This explains the colouring of soda glass by radium , the radiation the excl.etion of the metal .
In order to test this hypothesis I asked Mr. F. SODDY , on 9th November , 1903 , at University College , London , to examine whether the emanation adium was capable of colouring quartz glass in which there could evidently be no possibility of the excretion of metal .
He stated that he and Professor RAMSAY had already made this expel.iment and had found no coloration .
At my request SoDDY then placed a small piece of colourless gold glass in a Appendix .
tube ) ) ' ) tion .
Within two days an unmistakable ruby tint in the Tt seerns ) that the violet coloration of soda glass bulbs used in the of rays may be due to the excretion of metal caused by the rays from the catbode .
observations of GEITEL , ' Wied .
Ann. , ' 5 p. 487 , 1896 , quoted by J. J. , ' Conduction of Electricity through Gases , ' p. ) , that slts the nred 1 cathode rays exhibit photo-electric tive of esence of aces of the metal , support this view to the exposed to the radiation from radium .
From equation ( 6 ) , as modified the case when the metal sphere is surl.ounded by glass of active index , it appears that the amplitude at any point of the light emitted from the is proportional to , where denotes the modulus , .
Using and as defined in equation ( 13 ) , have where and are to be found from the table on p. 396 , where .
Thus at any point the intensity of light emitted by a sphere of adius c is proportional to , say .
Measurin in millim./ 1000 , the Table 1 .
gives the following values of Silver .
Copper .
Gold .
Yellow these values of I it appears that white light falls on a small sphere the light eInitted is , Silver , llol'e than yellow , I Copper , , yellow , , red , Gold , , , , , , , , The presumption is that for the two latter the light may be more green than yellow .
In the table given by and ZSIGMONDY ( loc. cit of which a copy is given ( Table II .
, p. 397 ) , it is seen that of the five glasses Cc , , whose pai.ticles are small compal.ed with a wave-length of light in the glass , the four glasses Cc , contain particles which send out a green cone of light , and the glass contains some particles which send out and some which send out brown .
Thus far we have confine ed attention to glasses for which the condition of having , 1904.\mdash ; Sir WILLIAbI ANSAY has lately exposed some clear silver glass and some soda glass at the same time to the nation from radium .
After a fortnight 's exposure the silver glass had turned a faint yellow and the soda glass a deep blue-violet .
] COLOUR , IN AN1 IN TALLIC FILMS .
lnany metal particles to a wave-length is satisfied .
We have shown that ) the metal is gold such glasses should be pink ( cf. column.3 of Table Il .
) by itted light ; and that the gold send ) the microscope light which is -eminently yellow or .
columns4 ) ; and we have rked that for the same reason that the polarisation of sky , such small ) no directly up the microscope tube the electric vector of the incident is in that direction , so that in this case the cone of light examined with the power will be cut off , although the large numerical ] ) tnre of the Zeiss oil immersion lens will allow some to go the tube , but so as to leave a black } ) in the centre of the focal ) lane of the microscope as shown in fig. 6 .
All these deductions from onr analysis confirmed in every detail by the three glasses ( Table IF And it is these very glasses , of all the glasses in that table , for which , according to the numbers there given , the are both smallest and closest together . .
Let us now briefly notice the remaining glnsses of II .
these ) lasses the number of metal particles to a , measured by gold content ) of , as determined from the 6th and 8th or the 7th and 9th columns of that table , is smaller than for the glasses , which show the ' pink colour .
For the sses A to this Ylber is greatest for the glasses and , of which the former and parts of the latte show the regular coloul Even glasses which do llot satisf .
the condition of many particles to a wave-length , which consequently do not exhibit the " " regular\ldquo ; ( pink ) colour of gold have many of their properties co-ordinated by results have obtained for ular glasses .
Take , for instance , the lasses A and ( Table ) .
riso of the gold content with ) size of the obserced } ) clrticles shows that those ticles at any rate are so far apart not to satisfy our condition .
The that glass A is colourless shows that if there are also minute spheres present which escaped rvation , they also lie so far as not to be many to a wave-length .
On the other hand the pin colour of suggests the presence of minute unobseryed which are sufficiently close ether to satisfy onr condition , the absorptio1l of the glass to that small of the gold content is associated with the minute spheres .
In both glasses the large ticles reflect much more than is emitted by the minute spheres .
The coloul of this reflected light is the usual yellow-red metallic reflectiolt from gold .
Therefore the colour of the cone of light ] be gold- yellow ( i ) .
When the Nicol is introduced parallel to the plane of incideltc , presumably half the incident light is cut off .
Consequently the particles send only hadf the yellow-red light up the tube that previously sent .
, bowever , f'act .
CC111 .
' that the minute spheres send no light directly up the tube when the electric vector of the incident light is parallel to the microscope tube ( Nicol perpendicular to plane of incidence ) , less than half the green light from any small spheres will be cut off The cone of light will therefore have more green in proportion to the yellow-red than before the introduction of the Nicol .
Therefore the colour of the cone of light will ba more white than before ( ii ) .
the Nicol is perpendicular to the plane of incidence , the green light from the small spheres is cut off , so the colour of the cone of light will be more red than with no Nicol , and therefore the total qualltity of light sent up the tube will be rather lessened ( iii ) .
The conclusions ( i ) , ( ii ) , ( iii ) are in accordance with the phenomella tabulated in the 3rd , 4th , and 5th columns of Table II .
The glasses Ca , Cb , present no special difficulties .
We have seen ( S 3 ) that those metal particles in a gold glass whose diameters are less than ( centim. ) are sphel.ical , and ( S 5 ) that small gold spheres send green light up the microscope tube .
In the aboye-named glasses the figures in the 7th column of Table II .
show that the ) articles are so small as to approximate to the spherical form .
This is by the green cone of and its approximate extinction when the electric vector the incident light is in the direction of the microscope tube .
As here , too , the observed particles are far enough ) to be distinguished under the microscol ) , it is necessary to postulate additional minute spheres to explain the pink colours of these glasses .
In glasses and the blue and violet colours of the transmitted light present a culty w I have not yet been enabled completely to It is probable that the particles in this glass are not sufficiently thickly distributed to the condition of there being many particles to a wave-length of blue light .
When the incident light is blue , the absorption that we have investigated is therefore not present .
When , however , the incident light is red , there are sufficient particles to a wave-length for absorption to take place .
Thus , although if light of all wave-lengthy were absorl ) , the would be least absorbed ; yet here it is only the larger wave- that suffer the absorption whose nature we have investigated .
PART II .
S 7 .
With a view to examining whether these principles apply to the colour changes exhibited bvtrallslucent films of metal when heated , observed by Mr. G. T. BEILBY ( ' Roy .
Soc. Proc vol. 72 , 1903 , p. 226 ) and by Professor R. W. WOOD ( ' Phil. Mag. , ' vol. , 1902 , p. 396 , we proceed to consider the transmission of light by films of metal , the metal being in form of small spheres , many to a wave-length of light in the film * See Appendix added Sth for explanation of Blue and Violet Colonrs .
COLOURS IN AND IN MFTALLIC FILMS .
We shall first confine attention to very thin films , defining very thin films to be such that may be treated as small , being the thickness of the film , and the wave-length of light in the film .
It has already been noticed that equation ( 9 ) , p. 393 , does not hold for very thin fihns .
That equation is obtained by observing that the action of its neighbours on a particle is that due to a medium which is perfectly uniformly polarised in the neighbourhood of the particle , and whose external bounda1y is that of the actual medium , and whose internal boundary is a sphere of radius , equal to the smallest distance between the centres of two particles .
POISSON has shown that the effect of such a uniformly polarised medium is equivalent to that of surface distribution over its internal and external boundaries .
The medium actually present here can only be treated as uniformly ised throughout the region inside a sphere whose radius , , is small ) mpared with the of light in the medium .
When the outer boundary of the medium is in all directions many wave-lengths distant from the particle under consideration , the effect of the periodically varying polarisation outside can be allowed by neglecting the Poisson distribution on the outer boundary of the medium .
Consequently , in this case , the effect on any of the remaining rticles is that due to a Poisson distribution over the sphere , which leads to equation ( 9 ) .
When the external boundary of the is , in any directiolJ , at a very small distance the average , we are not justified in neglecting the P distribution over that boundary .
In the case of a thin film of the medium in the plane of it is , howevel , clear that when the electric force is to that plane , is no Poisson distribution over the surfaces of the film .
Consequently the fihn has ( complex ) dielectric constants in the direction of the axes of and , which are the same as for the medium in bulk .
omitting the accent in equation , this constant is given by ( 16 ) .
The dielectric constant , parallel , may be erent from ; if so , the film behaves optically like a uniaxial crystal whose three ( complex ) dielectric constants , the optic axis being normal to the film .
S 8 .
Putting in equations ( 12 ) and ( 13 ) , we have ( 12 ) , where \mdash ; , ( 13 ) .
We shall find it convenient to use to denote the constants of the sphel.es .
The constants of the metal itself will fheref ) denoted by it will , as is priori evident , that the latter are the of and when ince t , equation ( 1C ) gives us on substituting ( 17 ) , from which , by real and ( 18 ) , ( 1 whence ( 20 ) .
table gives the values of and found by means of rmula ( from the constants and of the solid metal as given by DRUDE loc. TAB In order to determine the values of and for various values of , the numerical values of the functions , where ( 21 ) COLOURffi IN ANY ) IN TALLIC FILIsIS .
were calculated and silver the of : Equations ( 1 nd ( 20 ) may be written . . . . . . . . . . .
( 19 . . . .
, whence the values ) For silver were calculated the above of .
The of thence ined checked )'ainst Lhose obtained by means of equation ( 18 ) , namely , . . . . . . . . .
( 8 In the of silver less than it was , , seen to be better to obtain as the quotient of the of ( IS the value of 21 got and , owing.to the probable errtlmined directly ( 19 ) and .
From equation ( 13 ) we find which the as tions ( .
Consequently , as should be the case , the medium of sphere is to the solid metal wherein the spheres are of such val.ied sizes that they fill the whole .
Anothel check on the tabulated llulnbers is afforded thel.efore by comparison of the calculated and qerved values of and I believe that nearly all the numbers here given for silver and for gold are subject to an error ofless than 1 per cent. The values of and of for the -sodium of table , ' Phys. ' January , 1900 , less cal.efully calculated . .
Consider now the incide1lce of polarised light a of this medium .
We shall first suppose ] plate to be very thin and therefore optically ystalline .
Suppose the two surfaces the film and , and of incidence .
COLOURS IN GLASSES AND IN TALLIC FILMS .
First let the incident light be polarised in the plane of incidence , so that the incident wave is The reflected wave is X Inside the film , i.e. , between and , Transmitted wave X [ . .
In these expressions we have and . . . . .
( a ) .
Since and are continuous at and ( b ) .
Since and are continuous at , if we replace by , and by we obtain , when the square of is neglected , the equations ( c ) .
From the last pair of equations we find , neglecting squares of , that ( d ) , on using ( a ) ; then eliminating from the equations ( b ) and ( d ) , we finally have .
On taking the modulus and substituting for from ( 17 ) we obtain . . . . .
( 22 ) .
Secondly , suppose that the incident light is polarised perpendicular to the plane of incidence , the gnetic fOrce .
The incident wave is The transmitted wave is The velocity V inside the film is connected with the angle of refraction by the equations final result after .
the two sets of boundary conditions is whence , using ( 17 ) , we obtain ( 2S ) .
When the light is normally incident , the Cl.ystalline character of the fihn does not manifest itself , and we have from ( or ( 23 ) . . .
( 24 ) .
The absorption of directly incident light by a thin film is therefore governed by wing to the difficulty of whether any particular metal film whose changes of colour have been obsel.ved , but whose ickness has not been recorded , for example , the films observed by ssor R. W. WOOD or by Mr. G. T. BEILBY ( loc. cit is to be regarded as very thin for the purpose of this section , formulae for thick films will now be found .
We consider here only the case of directly incident ised light , and to an equation esponding to ( 24 ) , reserving the full discussion of the behavioul of thick films uely incident light till later .
the axes sllown in , suppose that Incident wave is ; Ileflected wave , Bexp ; COLOURS IN GLASSES AND IN TALLIC FILMS .
Wave in film , i. e. , between Transmitted wave where .
The boundary conditions at give .
It follows from these equations that is of the order of ; if therefore shall be correct within 2 per cent. when we neglect .
Thus to the Table IV .
it appears that if a piece of gold leaf before annealing be so thick that or , then , so .
as yellow and red light are concelned , will be for all ulues of , if we suppose to yary inversely the number being the smallest value of for gold for values of from 5 up to unity .
Eliminating from the last two equations above , From the boundary condition we obtain ating A from the last two equations Taking the moduli , the ratio of the intensity of the smitted to that of the incident light is given by ( 26 ) .
It appears that when the thickness exceeds of the wave-lengtb , the absorption is governed by ; but , to the same order ) approximation , by ) when is less than VOL. CCIII.\mdash ; A. 3 The ( comparatively ) small effect of the coefficient the lonr - } ) ilted light will ) considered later .
For the present it is sufficient ) that becomes small when for any colour , and hence the variations of intensify the absorption bands , which will be shown to occur for silver neal order to illustrate the discussion of the colours exhibited by films of nefal f. various values of trraphs are given and for ] , for silver , and the potassium-sodiunl , the constants of which for were given by ( it .
) .
The raphs of , and ?
gold and for silver when the incident is or yellow are tted from the values given in the accompanying table , with } help of additional point to This last point is easy to plot , for we see from equations ( 18 ) and ( 19 ) that when so that and This point is ] very near the maximum of , owing to the smallness of in with , and is also , in the graph of , not far frol1l the maximum , and in the graph of not the lninimunl .
It will be shown that for each of these reasons there is in genel.al an ) tion band in the colour whose for when blue light is incident on gold is surmised ; 1 .
, it is constructed on the supposition that the constants ?
and for gold , when , are red through yellow to blue .
The curve for for gold under blue is made of the same shape as for yellow and red , the value of for bein ( plotted from the value of assumed in that graph .
The curves for for potassium-sodium are plotted , again with the help of the points for , the incident light being blue or yellow .
The graph of .
for red is surmised .
of ?
potassium-sodium are sbown by analogy with those for gold , the only points plotted for .
The red curve is constructedfrom that in the same lnanller as the blue curye for for gold was got from the assuuled curve for for gold .
COLOURS IN AND IN TALLIC F IN AND IN TALLIC FILM OURS IN AND IN S 11 .
In on " " olvents on Films of Metal ' Proc. , ' 7 ) ) , 1 ) .
) ) ives i some experiments on tures far below their ] .
poi11ts .
He ) ' rests that sufficient freedom is the ) .
to to behave the lnolecules of the liquid metal wonld to under the influe1lce of tensl either in -like forms .
We have eady shown , when dealing with ] the colours metal rrlasses , the small ] rticles ) themselves the into Mr. BEILBy lecords that of silver eased , annealing , from a few to 50 ) ohms ) to of of course , strongly the theory the breaks ) unde ] tension into minute .
Profess WOOD observed no conducti 'ity in his fihns as originally deposited .
Mr. BEILBY furtb e. states in one of the rold films tbere appeared to be a considerable depth of thules , ofessor WOOD ecords asence of conductivity in ( film in ) granules } ) in contact with ) top of one ( notber .
observations snpport our ) esis ( the structure of the films , ( the granules ) } those which effectiv in the colour enol ) we ] Let us see when ) ' onr ) thesi s to structure of illns is agreement with the effects by ofessol Woo First , then , consider a very thin fihn of gold .
Accordi n to the result given in equation { he ninution i light of wave-length the yellow red , it is for which , before is less \fnof ; g )1 yellow , this is true nearly down to leaf of gold should show the colour distinctive of leaf , but the red colour should ] ) edol n over the yellow .
arbitrary gr would , if , show that blue edolni t over ' yellow or The coloul 0 a film of gold leaf would , , be clneiiy blue , less red , least ) ] the in the nnest piece of leaf possessed .
It should be noticed that it been ) ) fihll will more than yellow light , that it , thel.efore , will not the of leaf .
Tt stated ) it seems ) it will lel blue than eit We now ) that when the fihn is annealed , and causes the gold to form into spherical drol ) , many to -length , but of quite varying , the volunle of metal in a unit volume the fihn , so as the metal is kept at a ture of about 40 Just ched begins to better than the red , but the ) of rapidly .
The value of for red becomes tely starts dilni n on.y is tion band in the red incident is yellow , to a near bas value , Between and bsorbe to the same large extent .
Tt seGms ) bl that blue not ) ) so eatly for this value of .
The filrn ) ) be ) hnds that a fihn turns blue purple chiefly of yellow ) in the earlier stage of nneali n , tbough , ) , the , for which ] tect was obseryed fall into our thick fibns .
turning blue will be again referred to when conlG to ' thick filnls of is JnucIl less linished in intensity than the and robably 1 is pink , remains pink down to ions of coloured ( rlasf which is ) of .
The thin film Mr. after ling ( p. 227 ) .
The nsparency o by FARADAY by .
BEILBY ) onds to the very small yalues of ' values of Consider next a thick film of gold .
The being ) , according to the result given on p. 409 , dependent upon the value of , we see from the table for or fronn the graph that , for the solid , yellow is than red .
The colour of thick gold leaf is , in fact , by nsmitted light .
As diminishes the absorption of both and red more rapidly .
Now when , there great absorption of red , according to the values of , which is intensified , since is valtle educed to 177 .
hold then more yellow than probably more either .
When is , the colour is much more red than yellow .
If assumed curve blue is correct , the colour of the film be blue between , and principally red from through all the range of values of gold glass down to .
( If our for blue is COl.rect , the shows that the film is red when the blue curve crosses the yellow .
) According to Mr. , a gold film , iginall green , turned blue-purple after .
Gold leaf annealing , pink with brown-green patches ; the latter , , corresponding to and the former to small yalues of rise ) ) rption diminish ) ] unity was noticed by Mr. .
COLOURS IN { GLASSES AND IN }TALLIC h It has therefore been shown that all the observed colour changes in gold fillns are in accordance with the theory and numerical results set forth in this ) The points corresponding to , referred to on p. 410 , which were plotted in for and yellow , were for for Let us now consider silver films .
The results for thin films are not of much in tel.est , as ) fihns served came into this class .
We may , however , notice that , according graphs of ?
, the thin film should start by being more yellow than red .
There absorption band in the ) ) valuo of There is great absorption in the yellow for when of these values of for the lnaxima of the absorption of yellow red ests t films of silver should be blue or else very opaque when is about8 .
The thin film should turn more red than yellow slightly remain to small values of , at least as far as Passing to thick films , which the ptio is measured by ) serve from the 5(fraphs that as diminishes from unity the ) at first creases rapidly .
This may be correlated with the increased conductivity sted 1 silver film in the early stages of annealing .
Shortly } ) the film ) ecoJne more red than yellow , and although by the tion has already become extremely small , the film remains more red than } Putting , we find the additional on aphs of for . . . .
, yellow . . .
, , The red colour of silver films for low values of is observed in those obtained by depositing silver on glass in the manmer described by ofessor WOOD Phil. Mag. , ' August , 1903 ) .
It is also often seen in fogged ) .
S12 .
We proceed to consider the potassium and sodium films discussed by Professor R. W. Wood , in the 'Phil .
Mag 1902 , p. ) wing , however , to the inavailability of the numerical values of the constants for potassium or sodium for more than one colour when , the numbel .
S used are those given by DRUDE ( loc. cit. ) for " " potassium-sodium , \ldquo ; .
blue and yellow light .
Consequentl the same degree of numerical accuracy as for gold and silver bas not been aimed at .
The yellow and blue curyes for are plotted from the nunnbers tabulated in Table , p. 406 .
The graph of has been constructed to pass through the points Yellow . . .
for for Blue . . . .
, , , , VOL. CCIII . .
3 The films made by Professor WOOD were obtained by the condensation of the yapotll of the evaporated metal on the insides of exhausted glass bulbs .
We should therefore expect the film its original form to be in drops , which , in accordance ith Part I. , S 3 , when very small , approximate to the spherical form .
The absence of conductivity in these films suppol.ts this view of their structure .
effect of heating up to and beyond the meltin point would be to fuse these metal , and although surface tension tends to a re-formation into spheres , it is probable that will generally be considerably increased by the fusion .
It , lrom our raphs of and , that thin or thic .
' potassium-sodium\ldquo ; films should transmit more than yellow light , provided that , there being a absorption of yellow for ( about ) in both cases .
It is to note that Professor WOOD always refers to the yellow absorp tion bands as particularly strong .
As increases , the absorption of yellow relative to blue increases in both thick and thin films .
If now we introduce our hypothetical curves and for for red light , that the film should be red .
Near the greatest absorption of yellow , red and blue should be equally absorbed and the film be purple .
As increases further , red sbould be most absorbed , and blue least , so the film should be blue .
Thus , eneral , the film should turn from to blue when heated , as the case with most of the films observed .
Professor WOOD ( loc. cit. , p. 40 further states that the particles which he observed were distinctly closer in the bhJe than in the -purple , thus again suggestin that a change from purple to accompanies an incl.ease in So far , then , as they go , our results are in good accordance with observation .
When , however , nUDlbers can be obtained for and for potassium and for sodium for , yellow , and red light , it may be possible to state with more certainty that ou ] lanation of the colours of the films and of the changes in colour due to ting is the one . .
By considering the oblique incidence of plane-polarised light on thick films of metal by the method adopted in S 9 in the case of thin films , it can be shown that equation ( 26 ) is replaced by :\mdash ; ( 1 .
) incident ] ight is ) in the plane of incidence ( 27 ) .
( 2 .
) When theincidenlt light is polarised perpendicular to the plane of incidence ( 28 ) , certain functions of such that when the angle of ncidence , , is zero , and COL URS IN GLASSES AND METALLIC FILMS .
It can there be iations with ) .
the coefficients are such that a change in ) to ) } ) the ) bands .
The complete analysis what 1 ; I therefore refrained reproducing it here .
This result , however , shows that in general the ) hold be when the incident light in the plane ence than when polarised perpendicular to ) .
And this WOOD in almost every film .
PART 1II .
S 14 .
Metallic media composed small spheres to a wave e have many interesting properties addition to those referred to .
colour effects which are exhibited according to the raphs given above for gold , silver and -sodium\ldquo ; when light verses such media , 11 consequence of the different ptions of rent c suggest enquiry metals in bulk have ever been tained giving )illiant c by transmitted and reflected light , such metals being ordinary metals less than unity .
instance , have any of the lletals we discussed been obtained states in ) the specific gravity was not ) rnal value that luetal( in which the ) loul .
changed with the specific !
I hope in the near future ) alnine C in detail with a to ding out whether silver is a mediu ] ) of the have sidered\mdash ; silver with less .
But the first papers ( ' of Science , ' 1889 ) } the following ) respondence b the properties he observed the ties should , .calculations for yellow and red light , be possessed by AREY ilvers ined from solution ; we have shown that gold , and therefore , , silvel ] of solutions into articles which spherical if they small .
Our silver is composed of minute UAREY LEA'S silver ) ] into normal We should expect to ) increased by ) The specific of the fOrnns 01 ( lotropic silver ppreciably less that of nol.mal silver .
From our graph of for silver ) see red yellow light are equally , and very lbsorbed ] ratio of the .
J. C. MAXWELL fflNETT ON IN GLASSES , specific gravities of AREY I-coloured silver , , and normal silver ; given by him to .
This strongly supports the theory that allotropic is of the nature of the media have discussed .
CAREY LEA'S silvel'S were very brittle , but could be toughened by heating .
Further , his gold-coloured silver could be transformed into normal silver by .
; and this transfor mation could be greatly impeded by the -coloured silver in cotton wool .
These properties suggest a discontinuous structure for allotropic silver .
If we might assume an absorption graph of ) , the fact that if is iquel reflected frolu a filnx of " " " " , then the yellow light is ised in the plane of incidence and the ) pendicular to that plane can , I think , be explained by our : but the is not yet complete .
The red colour exhibited by all the 1nore dilute forms of the allotropic silver is in accordance with the fact , exhibited the , that is smaller for red than for yello light for small values of LNDIX , the values the active index absorption coefficient of gold for red ( C ) , green ( E ) , blue light , as given by ( ' Wied .
Ann 1889 ) , the values of the quantity , which ( erns the sorption of the gold glass , have ) calculated:\mdash ; Gold Gold Gold Gold Gold Gold Gold Gold Gold Gold Gold Gold Gold Gold Gold Red ( C ) .
Red .
yello ( D ) .
) .
Bhle .
38 3 ] .
79 refractive index of the glass has been taken to be , as in Table II .
, from the values of for red and for yellow have been copied .
The colours , in the order of the degree in which they are transmitted by gold thelefore are Red , Yellow , Blue , The .
order for silver as obtained by calculation is Yellow , Red , reen , Blue .
The orders accord with observations on gold-ruby glasses and silver glasses respectively. .
It will be seen that large particles of gold ( diameter ) in a gold glass would , by reflecting out the red and yellow light , give the glass a blue colour by transmitted , and a brown turbidity by reflected in glasses of Table I. ]

