]\gt ; Dr. Clay .
On the Application of ( 11 .
) Braus , H. " " Beitrage sir der Muskulatul ' und des peripheren der Selachier 'Morphol .
Jahrb 1899 .
( 12 .
) lnnett , B. C. , " " On the Formation of the Pelvic , \amp ; c. , in the Genus telus , \ldquo ; 'Phil .
, vol. 192 , 1900 .
: On the Application of Maxwell 's urves to Three-colour Work , with Especial Reference to the Nature of the Inks to be employed , and to the Determination of the Suitable Lightfilters By INALD S CLAY , , D.Sc .
Communicated by W. DE W. , K.C.B. , .
Received April 20- , June 20 , 1901 .
I.\mdash ; THEORETICAL .
1 .
Three-colonr Projection .
Maxwell showed that any colour in the spectrum could be matched a suitable mixture of three monochromatic lights , red , green , and iolet , and his oelebrated curves give at each parb of the spectrum the intensity of these lights necessary to imitate the colour at that point in hue and luminosity .
Other experimenters have since repeated his measurements with improved apparatus ; and throughou6 this essay I shall employ the curves found by Sir Wm. Abney as being probably the most accurate .
Thus it is possible to photograph a spectrum in three colours Three llegatives must ) taken through " " filters\ldquo ; which allow the colours to pass respectively in mounts determined by the above curves .
One filter must allow light to pass according to the red curve , the extreme red is just transmitted , and the bright red fully passed .
Then the filter absorl ) the slightly , and the absorption eases along the spectrum , until in the blue and violst it is nearly c.omplete .
From this negative transparency is made , projected with red ( monochromatic ) light 011 a white SCl'een .
So for the greell etnd violet .
Assuming this done , have a spectrum illuminated everywhere with the three monochromatic lights in the proportions indicated by the curves .
Thus the spectrum ] ) reproduced exactly .
( See Notes 1 and 2 .
) If , instead of illuminating the transparencies lvith monochromatic lights , we use red , gree1l , and violet lights produced by passing white light through coloured glasses ( providing that these lights are of the same dominant hue as the primary colour sensations ) , shall still obtain a spectrum that appears almost correct .
The light tained by a coloured glass is not monochromatic , but , if the hue is " " Colour Sensations in of Luminosity ' Phil. 1899 .
; to -colonr GOl.rect , it may be matched by a mixture of the monochromatic light white .
Thus the colours on the screen will not be trtle spectrum colours , but specrrum colours diluted with white .
The amount of white , of course , depends upon the glasses used , and in matching soms parts of the spectrum there will be more white than in others .
Yet the result is very good ; indeed , without direct comparison the white would hnrdly be noticed .
The beautiful pictures produced exhibited with such coloured glasses by Mr. Ives is very good proof of the fact that an admixture of white is of small consequence .
Not.\mdash ; The tbove is not strictly accurate , as it is not possible to match some parts of the spectrum even when using pure spectrum colours .
The match can only be made after white has been added .
The for this can be seen from the cnrves .
Owing to the overlapping of the se1tsation curvss , the green sensation is nowhere excited lone , but is vays accompanied ) a small excitation of the red and iolct .
Thnb the spectrltm green\mdash ; although really pure colour\mdash ; xcites e three sets of sensations in the eye , i.e. , it produces the tion g white .
Now , to match the yellow , we require to oxcite only red , and this we cannot do using spectrtlm red , for the green always excites to some extent the violet sen sation .
This violet combines with proper proportions of green and 1 ed to produce white , so that we can imitate the yellow when it is mixed with little white , but not the pure yellow alone .
The same is of the other end of the spectrum where the red sensation should only ) slightly exrited .
what follows neglect this neccssity the addition of hitc in of colour matches as it only the questio1l without affecting the resuIts .
Vhere my tenlents equire mtion in consequence I it ) a star ) ) No .\mdash ; In Abney 's above referred to , he shows the third not violct , but ) is the ) line in the spectrnm .
As I sh ] ] be using the word ' ' blue\ldquo ; in of the doul ) colonr ink , to confusion , I throughontlefer to the third colour sensatiolt as " " vioIet I shall , , use ) the ) iolet\ldquo ; shall me his \ldquo ; to printing , not so sintple .
's first impulse onld 1 ) to select those parts of the picture in which light occurs , or which excite the red sensation , print them with ink , so with the green and violet ; but ( little ) this vould not be right .
For yellow )ject ffccts 1 onr rod grcen to ] Dr. .
S. Clay .
On tlw ) lication of curves , and thus we should have to print it with the red and green inks .
Now the original yellow was equivalent to the addition of a red iUld green light , so that it ought to be brighter when both colours are there than when only one .
The reverse would be the case if printed as above , and , of course , as a white object reflects all the colours , it would be printed with all three inks .
Now , though red , green , and violet lights when added give white , red , green , and violet inks superposed on white paper certainly will not make white .
In fact , inks produce colour by absorption .
Thus , instead of successive inks adding to the light , they each reduce it .
Suppose , then , we proceed the other way , and of looking for those parts of the picture where red is reflected and printing with red ink , we print those parts where no red is reflected with an ink which tbsorbs red , but reflects all the other colours .
This will ) a bright -green ink .
In the same way we will those parts which reflect no green with an ink which absorbs thaIi colour ; this } } ' be a bright magenta or almost a pink .
And those parts which reflect no violet we will print with a yellow ink .
Now yellolv object reflects red and green but no violet .
Hence it will have to ) printed with the violet absorption ink only\mdash ; that is , the ctnd neither of the others .
A red object reflects neither green nor liolet , and lvill be printed with the pink and yellow inks , which will 1 only .
As a yellow object is printed with only the one ink , ] ) red one with two inks , the yellow will ) the brighter , this is as it be .
As the inks are to , green , violet respectively , they will , speaking , be complementary to those colours , that is , when added to them either } } a double-image prism , or by reflection from a clear glass surface , they should give white .
Still , comple- mentary is a very vague description of a colour , and it mnst be aim presently to define it more precisely .
The above theory is due to Mr. Ives .
3 .
Application of C'urves .
As we said at first , these curves were derived addiug three lights , and are only strictly true of snch colours .
If , however , ws use lights compounded of\mdash ; or , at least , capable of being matched by\mdash ; one of these monochromatic colours and white , we shall still obtain a resuIt very nearly as perfect as with the pure colours .
It call easily be seen , however , that if any other colour than white were c.ombined with the pure colours the would be spoilt .
All colours , including white , can ) tched both in luminosity ) combining the three monochromatic colours , reell , itnd violet , in proportions given by Maxwell 's curves .
It is sy to Note 1 , p. 27 .
eproduce them by adding lights ; is it possible to do so inting with inks It is , at least , theoretically .
result , however , lvotlld a dark picture of no use in practice , as it would require to be looked at either in full sunIight or some equally powerful light .
Suppose , then , we take a card and first print it all over ink that absorbs all the spectrum except a narrow band in the red , another rrow band in the green , and a third in the violet ( whether such ) ink could be found is immaterial as ou ] ' aim is pureIy theoretical ) .
If the vidth and position of these bands are correct , , if the colours left match the three sensation oolours in hue , and their luminosities properly proportioned , the card will now appear a nentral grey .
In a powerful light it will appear white .
This is to be our white .
The must each absorb one of these bands of colour lear the other two ) ands unaffected .
The yellow ink will absorl ) the violet .
, pink ink the green , aJld the ) ink the red band .
Now if we print with these inks in amounts which are the complementaries of the amounts indicated by Maxwell 's curves , we shall be transmitting lights in amounts lvhich the same as those given by Maxwell 's curves , and thus we shall exactly match all the colours .
picture so obtained will be correct both in hue and luminosity , but it will have to be examined in a very powerfuI light .
The three bands of colour left , referred to above , ought all to very narrow so that the light may be approximately monochromatic .
But now suppose these bands not to be infinitely narrow .
The light will no longer be monochromatic and a single band will excite more than one colour-sensation in our eyes .
For instance , a band the red\mdash ; unless in the very extreme red\mdash ; excites both the red and reen sensations ; a band in the green excites all three ( the red and violet will be nearly in the proportion to form white , and , if desired , the colour may be made equivalent to green and white by a proper selection of the width and position of the band ) ; a band in the unless in the extreme will excite also red and curves eferred to above are thoso in equal heights of , e.g. , Abney , loc. , p. 284 , fig. 8 .
* See Note 1 , p. 27 .
Dr. Ii .
S. Clay .
On the Application of with generally the green in excess of that required to coml ) with the red to form white .
with such bands for our primary colours , we should have not the true sensation colours , but those colours with an addition of either green or white .
The white not matter much , ) the green will , of course , spoil the eifect .
With these bands more light is used , and both the grey produced by the original printing and the final colours will bs much brighter ; thus the picture could be examined in a less powerful light .
Lastly , imagine the bands still wider ; in fact , allow them just to touch , so that they divide up the whole spectrum between them ; and suppose each ink to exactly absorb one of these bands .
The first printing in neutral colour is no longer necessary , and shall see the full white of the card .
Also the picture will no require strong illumination .
But we shall now have our colours mixed with large amounts of other colours .
The red ) mixed with green , the violet will be diluted with green and red ( see Abney 's ourve , loc. , and the green with both red and reen in portions which may or may not form white .
Note.\mdash ; It is possible to arrange that the red and violet in the midcUe shall make white , but it is not possible to so arrange that the red shall be diluted with white , for there is no violet at the red end of the spectrum .
Thus the blue ink cannot be made the complementary of the red sensation if it is to leave a red ) of any reasonable width , unless it has an absorption at the violet end of the spectrum , and this introduces a very serious fault , as same colon ] would then be absorbed by more than one ink .
( See 6 .
) It will now be obvious that if we print these inks in amounts the complementaries of the Maxwell 's curves ( as in the theoretioal above ) the colours we produce will be far from correot .
The red wilI be diluted with green and so on .
If the red could be diluted lvith white instead of the pictn.e would be much improved ; in faot , it old be almost as as if the colours were pure .
And mately this is quite ) , not however by modifying the inloe but by { dtering the proportio1ls in which they are printed , i.e. , by varying the curves to suit them .
Suppose , for instance , we want to print a colour to nlatch the red sensation .
If we print with the green and the violet ) inks \mdash ; the pink and yellow ones\mdash ; we shall leave the red end of the spectrum , whioh the curves show to excite the red and greel ] tions .
We want the red only , or\mdash ; if we cannot obtain that\mdash ; we mnst be content with red and white .
This we can obtain with onr inks by printing the violet absorption ink\mdash ; the \mdash ; rather less fully .
The violet and green then , with a part of the red will form hite , and leave red only .
So with any colour , if we cannot obtain it pure , adways match it when diluted with white .
amount of whitc to Thrcc-colonr so formed be incrensed in any part of the spectrum iuting all three inks less fully , and the best resnlt lvill be ) tained n the added white is everywhere proportional through the spectrum to the minosity of the colour .
The three inks we have been cnssing i snpposed ) sorb the spectrum thus : The curves to be used in priuting these mnst be adjusted so that the spectrlml may be matched in luminosity as well as ne , the added white being itself ( roughly ) proporlional to the .
To adjust the osity of the colours , it will be observed that the colours hich are relatively too bright can be reduced by moro with all thrcc inks .
The curves are roughly 4 .
of the Addition There are several ways in which this can be to bc important when the amount of white is not very great .
the first place , we are accustomed to use diluted with white .
According to Abney the light reflected from vermilion contains per cent. of white , from cobalt ) per cent. , from rench rine blue 61 per cent. , and from chrome yellolv 26 per cent. of whitc ( p. 166 , 'Colour Measurement ' ) ; whilst in the ) way the light nsmitted by a ruby glass contains 2 per cent. of hite , canary glass 26 per cent. of white , a green glass 31 to 61 per cent. , by a cobalt glass 42 per cent. of white .
That is to say , the trit1tsmitted light from a cobalt glass can be matched by 68 per ce1lt .
of if spectrum colour with 42 per cent. of white .
TheIt again about 3 per cent. of whitc can ) added to rather less to green , so along the ] ) per cent. .
R. S. Clay .
On th Application of the blue-green can be added without being perceived even when the diluted colour is compared directly with the pure ( p. 132 , 'Colour Measurement It can thus easily be believed that much more can addsd without appreciable sffect when no direct comparison is made with the pure colours .
But probably the best example of the unimportance of the addition of white is given by Mr. Ives 's resuIts , where the colours used as primary colours are produced by transmitting white through a red , reen , and violet glass respectively ; when , as in the examples above , there is transmitted a light which is equivalent to a pure colour with a percentage of white , which in some cases is quite considerable .
I confess my surprise at the excellence of the results obtained , which show how exceedingly bad a judge the eye must be of an addition of white to a colour .
Provided the quantity that will be left in the printing is not greater than the quantities he obtained by the transmission through his glasses\mdash ; and I see no reason why inks should not produced to secure this\mdash ; the pictures obtainable by the three-colour processes ought not to be inferior to those he obtains by the super- osition of the three-coloured transparencies .
We realise the effect of this addition of white by supposing a oloured picture projected on a screen in a room that is not quite dark .
The screen will then reflect , in addition to the colours projected , certain amount of the diffused light of the room .
As this coloured picture would not consist of pure colours undiluted with white , the of further white make the proportion greater than would ccur in printing .
On the other hand , the eye is a very good judge of , a very small variation in the proportion of the colours ( other than white being easily detected .
Much may be done , by training , to educate the eye to appreciate the relative luminosity of the colour , or , as it is usually termed , the value\ldquo ; of the colour , and many artists are able to recognise variations in this almost as easily as in the hue itself .
But the average person is not a good judge of this , and it is of much less importan than the hue .
For instance , many people fail to ses any improvement in a photograph of a landscape or a bunch of flowers taken through an orthochromatic screen , ower a photograph of ) same subject taken in the usual way .
But circumstances combine to make it difficult to recognise the additiou of white even in a picture : the irregularities in the varnishing ttnd the dust upon it are bound to add a of white to the light reflected , and indeed it is very seldom that a picture is so hung that the varnish does not reflect a amount of white , which we often do not notice until we attempt to take a photogrftph of it , and then very special lighting is found absolutely necessary .
's Curves to TVork , aec .
5 .
Printi)Black .
As we shall obtain all our colours diluted with white , one is tempted to ask whether printing with a black iIlk will not enable us to get them pure ; that is , as we can copy amy colour , except for the addition of white , by the three printings in yellow , ] ) , and pink inks , could we not , by the addition of black , match the colours exactly ?
Suppose it is the red we are matching .
We shall get a red certain percentage of white .
For instance , for 100 parts of red light reflected there may be 10 green and 10 violet .
If we now print 10 per cent. black what will be the result ?
If the black is a good one , i.e. , if it absorbs uniformly all along the spectrum , we shall now reflected 90 parts red , 9 green , and 9 violet .
In other words , merely darken the whole , without in the least altering the proportion of white .
The addition of black will then not improve this purity of the colours ; it will only make them dirty 6 .
Production of Colonr by successire Abso ption of Lighi .
There is a very fundamental difference between this and the production of colour by successive addition , as in Ives 's triple projection .
In the latter case , if any spectrum colour is thrown on the screen by one light\mdash ; say , the green\mdash ; and the same spectrum colour is transmitted also by a second light\mdash ; red , for instance\mdash ; the amount of that particnlar colour is doubled , and the result is an arithmetical addition of luminosity .
But in printing , if a spectrum colour is completely absorbed by one ink\mdash ; say , the pink\mdash ; and also absorbed by another\mdash ; perhaps the blue\mdash ; the total light of that colour absorbed by the superposition of inks is not twice that removed by a single one .
The absorptions do not successively subtract light .
If two inks each would separately transmit one-tenth of the light of a given wave-length , the two together would transmit not one-twentieth , but only one-hundredth , of the light of that colour .
The law is a geometrical and not arithmetical one .
In " " process\ldquo ; three-colour printing this is of the utmost importance , for there the inks are always printed full strength , and the tint is regulated by the size of the dots , that is , by the percentage area of the paper which is covered by the ink .
The dots are produced by a ruled screen in front of the negative while it is being exposed , the dots at any , when developed and etched , have areas very nearly proportional to the intensity of the light which fell upon part .
Each colour is produced by similar dots , and when the throe inks are printed these dots partly overlap one another .
As the ruling is very fine and the dots are very closely spaced , it is VOL LXIX .
1 ) 34 S. Clay .
On the of commercial printing to control the extent to which they overlap ; that is to say , the successive impressions will , owing to ltte differences in " " registering\ldquo ; the paper , have the dots of the different colours more or less laced in relation to each other .
If , for instance , the blue dots and the pink dots fiappened each to exactly cover half the area , in one impression they might be stlperlmposed , and in another they might hardly overlap at all .
In the former case half the area will be printed with both pink and blue , and half will be white .
In the latter case half the area is printedwith half with pink , and there is no white left .
If these impressions are to be equally good , the resulting absorptions through the should be the same in the two cases .
Let the adjoining curves represent the percentage of the inks .
At A both inks are perfectly transparent , and as the colour is not ) sorbed by either , it can of course make no difference to this colollr if the dots are superimposed or not .
At each colour absorbs half the light and reflccts hulf the the dots are printed adjacent to one another , half the total that colour be reflected .
}yhen they coincide , half the paper i$ left hitc : this reflects the light it receives , is the The on the other half absorbs half the it Maxwell 's Curves to Worl receives of this colour ) , and the pink absorbs half of what is left , so that finally a quarter of the light fadling on that half of the paper\mdash ; or one-eighth of the total light of this colour\mdash ; is reflected .
Adding this to the white reflecfed from the white half , we see that fiv-eighths of the total light of this colour is now reflected , instead of the on -half which is reflected when they are adjacent .
The colour at is entirely absorbed by one ink and is unaffected by the other .
The pink ink absorbs the light which falls on its half , and as the blue ink does not affect this colouu it will not matter hither it is on the pink or adjacent to it .
Also at , where one ink partly transparent and the other completely so , it cannot matter whether the perfectly transparent one is above or adjacent to the other .
If one or other of the is transparent at , the eotent which the dots overlap is immaterial .
If the absorptions of the inks overlap one another , as in Diagrttm 2 at , the lits are worse .
This colour is entirely absorbed by each ink .
If then the dots ars itdjacent , the colour is absorbed everywhere and none reflected .
But lvhen they coincide half the pape is white , and therefore half the total light of that colour is reflected .
Thus the reflected light of that part of the spectrum where the absorptions overl will vary from to half the total light .
In the case of the pink and blue inks this overlapping would occur at the , where their absorption terminates , the very brightest part of the spectrum .
According to Abney 's curves , the luminosity of the yellow from to is about 50 per ccllt .
of the whole whits light .
Thus the addition of one-half of this of light wonld ) snfficient 1 ) Dr. S. Clay .
On of to entirely spoil the result , this addition would be produced by a displacement of the dots of ) of an inch .
It is obvious , then , that in process work the absorptions of the inks must not overlap .
On the other hand , if we are to be able to produce a black , the absorptions must met .
Thus in process work ( in the case of the blue and pink inks at least ) the absorptions must just meet without overlapping , and therefore they must It will not matter so much in the case of the yellow and pink inks , for the part of the spectrum where their absorptions meet is of far inferior luminosity .
In collotype printing this difficulty would not arise in the same way .
Here the tint is produced by the strength of the colour , the lighter shades being produced by covering the whole surface with a thin layer of ink .
In this case there would be no uncertainty due to imperfect registering .
But at the same time the ink curves for a thick layer of ink are very different to those for a thin layer , as is shown by the measurements I have added of the curves of different depths of the same ink .
Thus the proportions in which the three inks should be combined to match a certain spectrum colour will vary with the luminosity of that colour .
To illustrate this .
Suppose an object all one colour , such as a cast , to be .
illuminated by a monochromatic light such as a sodium flame , and a photograph taken , ths result ought to be a picture in shades of yellow .
But if the curves of the sams ink when printed to different depths are different so that the proportions of the colour sensations that it reflects are different , the proportion in which it is combined with the other two inks to match a bright yellow will not be the same as that required to match a dull yellow of the same wave-length .
Thus if the filters are adjusted to give the exposures on the three plates that will reproduce that hue in , say , the high lights , it will be correct in the half-tones , and will be still worse in the shadows .
The ideal ink for this process would be one which would absorb the colours in the sams relativs proportions , whatever the depth to which it was printed .
The curve would be one of perfect transparency up to the absorption band , then a sudden drop ( for the full colour ) , say , to about 2 per cent. , to at this height till the end .
The curve should have no rounded corners .
It will be seen that the sharpness of the drop depends on the depth to which the colour is ; thus for " " process " " work , lvhere it is always printed to the same strength , the absorption can be made abrupt by using a very full tint .
But if this is to bs done , the ink must be a very transparent one to part of the spectrum it is not supposed to absorb , for , we have seen , it is very important that the same part of the spectrum should not be tbsorbed by more than one colour .
to Threc-colonr Work , 7 .
Best Limits Assuming now that the absorptions should be abrupt , the next question is , what limits should they have to give the best effects ?
If ome is at A and the other at in the accompanying diagram\mdash ; in which equal heights of each colour would give white\mdash ; the yellow and inks will be complementary to the colour sensations , but they will very pale , for they reflect large quantities of all three colours .
The pink ink will be very dark , it will reflect only a very small amount of the red sensation , and as the violet reflected is not equal to the red , it not be complementary to the green sensation .
It could be made so shifting A further to the violet , so as to reduce the violet reflecte the pink ink\mdash ; in fact , until the areas of the rsd and violet are equal .
It is obvious that this would give absurd inks ; the yellow and blue old be so pale as to be mere tints , and the pink so dark as to appear purple .
If the absorptions are supposed to end at and , the spectrum be much more evenly divided between the three colours , each will be slightly pale , but now neither lvill be the complementary of the colour sensation .
For instance , if the yellow ink is to be conplementto the violet sensation , it should absorb either violet or equal amounts of red and green , measured by equal areas on the curves .
But at ( or at any other line between A and B ) the green sensation is evidently in excess .
In the same way if the sorption of the blue ink extends from the red end of the spectrum beyond , green is absorbed as well as red , and the ink cannot be complementary to the red sensation .
To make this ink complementary , it must have another absorption band somewhere in the violet , and the difficulties we have been considering above will be introduced , for there will then be a part of the absorbed by more than one ink , namely , by the yellow and the blue , and in places where these inks are printed side by side the absorption of this colour will be twice as great as in those tvhcre they are on top of one another .
38 Dr. B. S. Clay .
On the -Application of Thus the blue ink should not be the complementary of the red sensa- .
And in fact the ink will be far from complementary to the red .
I find by taking the areas of Abney 's curves up to planimeter , that it should very nearly match the green at The yellow ink would be made complementary to the violet sensation by ending the absorption at , but it would be very pale , for it would transmit a large amount of white , as there is a large proportion of the violet beyond A. The result will be an unnecessary amount of white in the final picture .
The result will be better if the reaches to .
Thus this ink also should not be to the violet sensation .
The pink ink at the red end of the spectrum transmits red and green , and at the other end chiefly violet and green .
If the violet is equal to the red it be complementary to the green sensation .
This can be achieved by moving A towards the violet eIld , only at the of the yellow , which would be rendered very pale .
So this ink also should not compl mentary .
I have gone into this rather full .
as it has so frequently been stated that the inks ought to be complementary to the colour sensatio1ls .
a very rough and general way this is true , but there is no advantage in such inks ; on ths contrary , they would , even if obtainable with abrupt absorptions , not be so good as others that are not complementary .
The only possible advantage in using complementary would arise in the case in which the absorptions were abrupt , for then the filters would be complementary to Maxwell 's curves .
But the advantage here is only one of theory , the printed result being , as we have seen on the preceding page , a picture with a large excess of white .
As in practice the filters would be adjusted by trial , even advantage is illusory .
To summarise .
The inks could only bs complementary\mdash ; 1 .
If they had abrupt absorptions with limits near the ends of the spectrum , when the colours would be in some cases merc ints or 2 .
If they have overlapping absorptions , or if some colours absorbed by more than one ink , both of which are bad , the resultant absorption will follow a gsometrical and not an arithmetical law .
The exact positions of the limits lvill , no doubt , finally depend on the fact that there will bs very few pigments which sufficiently fulfil the conditions , and the blue and pink ones with abrupt near the yellow will not be numerous .
Still it is theoretically interesting to determine the best positions , apart from the difficulty of finding inks to suit .
This lvill be best done by calculating the curves for different absorption limits estimating the and distribnto -eotour tion of the added white , and the luminosity of the resultant colonrs in each .
I have worked out the curves for two positiolts of the absorption limits .
8 .
Clwice of Inks .
The mere appearance of an ink , even when a on is taken , is not much guide , as the eye is not suited for selecting the by their hue ; for they are double-colour inks , and we know that the eye cannot at all judge of the composition of a yellow or bluc .
Yellow , for instance , might be spectrum yellow , or a mixture of any part of the red end of the spectrum with almost any part of the green .
In other ords , the eye cannot distinguish between yellows of a great variety of compositions .
But the eye is a fairly good judge of the three primary colours\mdash ; at least , of their that if the inks be )rinted in pairs a fair estimate can be made .
The violet produced by the pink and blue inks should be almost pure , contain mg only a small percentage of white , as the red and green are almost equal .
The green should closely match the green sensation .
But the red will contain some green , and should abont match the hue of But the only true test is the spctroscopic one .
mined spoctroscopically\mdash ; 1 .
The inloe shouId each absorb one band of the in the red , green , and violet respectively .
Just transmit the remainder unaffected .
.3 .
Must absorb the whole between them .
4 .
For half-tone work the absorptions must cnd uptly , in the case of the pink and blue inks . .
The limits of the absorptions should be so chosen that the added white may be distributed throngh the spectrum as possible proportionally to the luminosity of the spectrtlm , and that 6 .
The luminosity of the resuIting colours in matching the spectrnm should be as high as possible .
7 .
The will not be complementary to the colour se1lsations .
the absorptions must meet , for ( 2 ) they must not ) and ( 3 ) lead to ( 4 ) .
The above conditions are , I believe , sufficient .
I ve cd number of inks supplied by differe1lt firms to sce how .
they itre realised by those at present in use .
-ls the ange , yellow , green , and blue-green parts of the spectrum are far brighter than the rest , it is here that any conditions down must be specially observed .
In the bluc and violet : the decper red it is not so important .
40 Dr. B. S. Clav .
On the Application of Thus , if the pink ink is only partly transparent in the blue and has not an abrupt absorption there it will not much matter , provided it is perfectly transparent in the red and orange , and has an absorption commencing suddenly in the yellow .
Such an ink is Fleming 's " " Theoretical Red No. 1303 .
The yellow must be perfectly transparent in the red , orange , yellow , and green with an absorption commencing in the blue ; but it does not matter if it is not very sudden .
The blue ink must be perfectly transparent in the green , blue-green , and blue ; there must be an abrupt absorption in the yellow .
The violet does not matter so much .
So far I have found no ink to fulfil these conditions .
All the ordinary inks are too opaque in the green blue-green .
9 .
as well as the Hue of the Spectrum must Iatched where .
It must not be forgotten that most natural colours are composed of a large range of spectrum colour , and the resultant hue will depend on the proportion in which these colours are compounded .
Now although the eye is a bad judge of theluminosity of the colour , it is a very good judge of the resultant hue of a compound colour ( and thus indirectly it is of course ( , able to compare the relative luminosity of its components as far as this luminosity is due to colour as distinct from white ) .
It follows that the component colours must be rendered in their correct proportions .
When a negative is exposed through its light-filter the light that will reach it from a object will be the sum of all the colours that the object reflects , endered according to the curve of the filter .
Suppose for instance that the light from the orange to the green is reflected by the in the first figure .
Then if the second figure represents the of the light-filter , the of the shaded portion gives the amotlnt of transmitted Jfaxwell 's Curves to Threc-cotouJ ' and ( supposing for the moment the plate to be equally sensitive to all colours ) will be proportional to the final opacity of the plate .
If the filter curves have been arranged so as to match the in lnminosity as well as hue , then the resulting print lvill reflect the sum of light which is equivalent in ray composition to the sum of this band of spectrum colour , and will therefore match it in dominant hue .
But if the spectrum has been matched in hue only , the resultant colour will be the sum of the colours in this band mixed in the wrong proportions , and will not have the same dominant hue as the original .
10 .
The ht-filters .
If the inks are neither perfectly transparent or perfectly opaque to each colour , so that there are no parts of the spectrum to which they are only partially transparent , and if also the absorptions do not erlap , the resultant colour will be that left after successive ) traction by the three inks , and will be the same as would result from the successive addition of the three complementary lights in triple projection ; and then the filters can be found by determining the dominant hue of the complementary colour .
The curves for the filters can also be calcuIated in this case by finding what percentage area the inks should cover so as to transmit light whose composition , in sums of the colour sensations , is everywhere that given by Maxwell 's curves .
This is how I calculated the curves for the hypothetical inks .
But when we deal with practical inks these conditions arc far from being fulfilled , and the filters cannot be found by dealing with the complementary lights .
For , as have seen , since the same colour is sorbed , at least to some extent , by more than one ink , the arithmetical law of absorption will not hold , and the colour resnlting from the printing with the inks is not that produced by the combination of the complementary lights .
So that in practice we shall have to adopt some other method , which be jrimental , mn } of 11 .
Depth of Colonr One of the greatest practical difficuIties in with only three colours , and especially with such strong colours as red and blue , is to ensure that the quantity of ink shall not vary from impressiolt to impression .
Usually , very slight variation in the of ink is accompanied by a change the whole tone of the picturc .
This is ecause with most inks the curve of ) sorption is different for different of colour , and thus the proportion of the three ) Abney , ' Photographic Journal , ' , 1900 , p. 121 .
Dr. .
S. Clay .
of colours left in the reflected light also will vary with the depth .
his will not matter much when a large number of colours\mdash ; twelve or fourteen\mdash ; are used , as the average amount of red and blue put on by the sucoessive inks will be nearly constant ; but with only one ink of colour the effect is very marked .
This effect is greatest when the curve for the ink is a gradual one , for as the quantity of ink is increased in an arithmetical progression , the absorption of any given colour will increase in a geometrical progression , assuming , of course , that the ink is transparent .
Thus , the distributing apparatus of the machine is very perfect indeed , it will be impossible lvith such an ink to obtain uniform impressions .
, Were the ink one with abrupt absorption , that is , one which was very transparent up to a certain colour , and then nearly opaque , slight variations in the quantity of ink old have very little effect , and one of the greatest difficulties would at once disappear .
12 .
on thoe Purity of the Colour of It might perhaps be supposed that inks which reflected most light in the parts of the spectrum for which they are to be transparent , and gradually reflect less up to the part they are to absorb , would be approximations to monoohromatic inks than those whose absorptions begin abruptly .
That this is not the case can be easily seen if we compare the two red-absorption inks opposite .
In the second we have replaced the reflected light ABD by BEC .
Now Maxwell 's curves show that the proportion of red to green in BEC is much greater than it is in ABD , so that the absorption of the red is not so complete in II as it is in I , whilc the transparency to green is less .
Thus II Yould not bc so good an ink as I. When , for instance , this and the vioIet ink are prinled , lvhich ought to leave green only , there would be some left , and this in grcater amolUlt than the I. This onId to Thrcc-colocuWork , mean that to match green more violet would have to be left to form white with the excess red , and the dilution of the green with hite would be greater .
13 .
Inks with Abrnpt tion . .
The colour will not be affected by slight variations of register .
ii .
Variations of the amount of ink in the several impressions will not be so important .
iii .
The colours will be much purer , being mixed with less white . .
The curves for the filters can be found if desired by determining the complementary light , and working as Abney has described in his paper in the ' Photographic Journal ' already mentioned .
This last will , however , only be possible if the curves are nearly 100 per cent. up to the absorption in each case , and so far I have found no set of inks , nor even a single ink for tvhich this is the case .
PART II .
14 .
lcnlntion of for By theoretical inks I merely mean inks which are perfcctly transparent until the absorption commences , and then are perfectly opaque .
With such inks we can calculate the percentage area to be covered to match any given colour by considering the absorptions they produce .
I drew on a large scale the curves found by Abney for the ray composition of the normal spectrnm , in which equal ordinates form white .
I drew vertical lines where the ink absorptions should terminate , and with a planimeter I found the areas of the curves up to these lines .
Thus I obtained the ray composition of the inks .
Thus , taking the vertical lines at wave-lengths 49 and 59 , I found the following areas .
From the violet end of the spectrnm up to rrhis will be the part of the spectrum absorbed by the yellolv ink .
From to areas were\mdash ; 44- S. Clay .
On the This will be the part of the spectrum absorbed by the pink ink .
From to the end of the spectrum they were\mdash ; yiolet , .
Green , .
Red , These add up to\mdash ; Violet , .
Green , Thus the ray compositions of the inks are\mdash ; The blue ink matches and white , and the yellow will match white .
I now took the ray composition of some spectrum colour , and found successive approximation the area to be covered by the inks to match it .
Thus has a ray composition of\mdash ; Then if 66 per cent. of the area is free from yellow ink , 66 per cenf , .
of the light the yellow would absorb will be transmitted , .
namely\mdash ; So if 5 per cent. of the area is left free from blue it will reflect in addition\mdash ; The pink ink must be printed all over .
Then shall have left\mdash ; matches the spectrum colour except for excess of white .
worked out the percentage areas to be left free colonr to match each of the wave-lengths , 40 , \amp ; c. , up to , and have found in each case the excess of white .
to Three-oolour Work , \amp ; c. I selected the above positions partly because they divide the spec- trum about evenly , but also because most of the pink inks have their absorptions at about , and the yellows at about .
As the blue inks are all bad , it is no use considering where their absorptions are .
Another interesting place for the absorptions to end is at the )oints of intersection of the ray composition curves , namely , at , where the green and violet curves cross , and at , where the green and red cross .
It is useful to compare the excess of white and its distribution in the two cases .
The results of these calculations are given in the ving tables:\mdash ; The sum of the ray compositions for the same -lengths is 1120 .
Thus the excess of white is about 9 per cent. It will be seen that the excess is chiefly in the brighter part of the spectrum , though seems rather in the green than the yellow .
These calculations entirely agree with the arguments advanced in the earlier part of this paper to show that the light filters for threecolour printing should be entirely different from those used for threccolour projection .
S. Clay .
On the of luminosity curve of the normal spectrum .
It is obvious from them that inks whose absorptions terminate at and will be better than those terminating at the points of intersection of the ray composition curves . .
Th Throughout the following experiments we used a copy of Abney 's colour-patch apparatus .
The only difference in the main instrument was the use of a right-angle prism in front of the collimator slit to reflect the light from the lantern down the tube .
This enabled us to set the apparatus up much more nearly in a direct line , and therefore to arrange it all on an ordinary long table .
The collimator was about 13 inches long , with a -inch achromatic lens .
The slit was one of Elliott Brothers ' .
The prisms were equilateral prisms cut from a very fine specimen of white glass especially selected for this purpose .
The face of each was about 2inches .
The projecting lens was an achromatic one , inches diameter and about 30 inches focus .
It produced a visible spectrum 8 .
long , which was very bright and sharp .
The slit was mounted in a brass frame arranged to slide horizontally across the spectrum in its focal plane .
It was about inches high , and usuaily about 5 mm. open .
A small 1-inch right-angle prism , carried by the same frame , was arranged to reflect the light from the top half of the slit to one side on to a second and larger prism , also carried on this frame , which reflected it forward .
bining Lenses.\mdash ; These were two 6-inch lenses of about 20 inches focus , one of which received the light from the lower half of the slit foeussed it on the patch .
This is the direct } ) .
The other lens focussed the reflected beam on the same patch .
This patch is the image of the aperture of the double-image prism mentioned below .
The distances of the lenses vere adjusted until the horizontal edges of the patch were sharp , and were then rotated slightly round a vertical axis until the patch remained stationary as the slit was moved across the spectrum .
The patches were formed by cutting a small square lvindow in a piece of black velvet pasted on a card .
To avoid stray light the placed at the far end of a box , 2 feet long , blacked all over .
The mouth of the box was covered with brown paper , and hoJes were cut in this just large enough to admit the two beams .
part of one side was removed to enable the patch to be observed .
A blacke1led pillar was placed , as in the Rumford photometer , to a sharp tlividing line , one-half of the patch being lighted by 's Curves to Thre-colour Work , \amp ; c. the direct beam and the other half by the reflected beam .
The two patches were adjusted ( before the pillar was inserted ) to exactly coincide , so that any want of uniformity in the illumination might affect them equally ; this we found most important .
The patches were slightly larger than the window , so that it was lighted up to its edges .
The pigment to be compared was always placed in the reflected beam , the white beam falling on a piece of the same paper that the pigment had besn printed upon .
The sector was one made by Hilger upon the lines laid down by Abney .
It closes entirely , and opens to one-half of the full beam .
The aperture is divided into 100 parts , so that percentage aperture can be read directly .
It has to run at a high speed to obtain good resuIts .
I found that there was a lot of backlash in the sleeve , amounting to about two divisions .
To avoid the error due to this , it must be stopped always in the same way .
I found by exactly closing the sector while it is running\mdash ; which can be done very accurately by looking through it to the light and adjusting it until the light is entil'ely extinguished\mdash ; and then stopping it by the sleeve , the reading is nearly always zero .
If in addition the handle has last been moved in the direction to dose the akerture , the reading was always found to be corrsct .
These experiments took a time .
The sector was placed in the direct beam just in front of the box spoken of above .
As the direct beam is cut down to one-half by the sector , even when the latter is fnlly open , it is necessary to reduce the reflected beam also to enable a balance to be obtained .
This was done by attaching a Nicol 's prism to the frame carrying the slit and the two right-angled in the beam between the two latter .
As the light has already been polarised by the double-image prism , it is possible by rotating this Nicol to reduce its luminosity to any desired extent .
The double-image prism was about 1 inch square aperture .
It was at first mounted against the collimator lens , but finally was placed against the projecting lens .
Its adjustment is of very great importance , and it was impossible to set it with sufficient accuracy by hand .
It was therefore mounted in a brass frame which could be given a very slow rotation by means of a screw , and after the prism had been adjusted as nearly as possible by hand , it was finally corrected by this screw .
Its rotation causes the two spectra , into which it divides the original one , to move relatively across one another , and therefore varies the colour which passes through the slit .
The prism is rotated until , with the slit in the yellow , the two beams are exactly the same colour .
As the lightest alteration here is very easily perceived , adjustment can be effected very accurately , and will then be correct throughout the spectrum .
A glass Zeiss millimetre scale was temporarily attached to the top of the camera , and a on the frame carrying the slib moved VOL. LXIX .
Dr. R. S. Clay .
On the Apptijation alon the scale and indicated the position .
The line was at 50 on this scale .
The temporary scale was afterwards replaced by a permanent one on which the line was at , and to avoid confusion , the early scale rsadings have been rsduced to this scale by subtracting from each .
16 .
Scaling .
For this , the slits were removed and the spectrum was received on a strip of celluloid placed between a glass plate and a glass Zeiss millimetre scale .
Various substances were then placed on the arc , such as Li , , Na , , and the bright lines read .
A -tube was also placed at the collimator slit .
The results are plotted in Curve No. 1:\mdash ; Maxwell 's C'urves to Three-colour Work , \amp ; c. The line on the new scale was at , and as it was also a milli .
metre scale , the new scale readings will all be less than the previous ones .
To test this the following readings were taken:\mdash ; 17 .
Determination of the Curves of Practical Inks .
January 1 , 1900.\mdash ; We set up the apparatus with the two right-angle prisms and obtained the patches .
We found some difficulty due to stray light , which we nearly eliminated , so far as the outside of the apparatus was concerned , by enclosing it with screens .
A cloth was placed over the arc and over the collimator and prisms , and a blackened canvas screen was hung up to divide the end of the room where the observations were made from the end containing the light But when matching pigments in a part of the spectrum in which they were nearly black , we were still able to perceive a small difference of colour in the two patches , which was due to internal reflections on the * Probably an error in reading .
Dr. R. S. Clay .
On the pplication o collimator and camera bodies and at the surface of the double-image prism .
We reduced this as far as possible by the liberal use of stops , but could not entirely avoid it .
This is not wonderful when we consider the small luminosity of the ends of the spectrum , and remember that we were often only using 2 or 3 per cent. of this feeble light .
It is obvious that a very small amount of reflected light would be easily notioed .
On some occasions the patches were observed through coloured gelatines to help to cut this off .
For these adjustments I am greatly indebted to my friend , Mr. Alex .
A. Talent , who worked most enthusiastically in getting the apparatus together , and who has assisted me throughout in the experiments .
The relative intensity of the direct and reflected beams was not constant through the spectrum , and it necessary to measure it .
For as the light strikes the second refracting prism at an angle very near the polarising an , it is thus partly polarised when it reaches the double-image prism , and therefore the two beams into which this splits it will not be of equal intensity , and this inequality is more marked for some colours than for others .
March measured a yellow and red ink by .
Mander , of which I had made rubbings .
Probably due to small deposits of moisture which form on bhe prisms and lens\mdash ; owing apparentIy to the glass being more than usually hydroscopic\mdash ; the inequality above mentioned varies from day to day .
The moisture probably reduces the polarisation by reflection at the surface of the prism , and thus alters the relative luminosity of the beams .
It is not advisable to clean the surfaces very often for fear of spoiling ths adjustments .
It is therefore necessary to frequsntly re-determine the relative brightness of the patches .
We found a great ovement in the accuracy of our readings as we became more experienced , and also as the stray light was more perfectly eliminated .
The chief difficulty remaining was caused by the unequal density of the inks the patch , which is far more obvious when measuring them in a monochromatic light than it is when casually observed in ordinary light .
18 .
of the Curves .
With this apparatus I measured the proportions of the light of each colour throughout the spectrum that was reflected by several of the inks sold commercially for three-colour printing .
I measured such a set of inks sent by Messrs. FIeming and Co. , and described as " " Theoretical Inks\ldquo ; for three-colour work .
They are not permanent , but that would not be important for book illustration , apparatus was entirely re-adjusted after No. 2 was measured .
llfaxwell 's Curves to Three-colour Work , \amp ; c. though it would be fatal for advertisements or pictures to be exhibited .
Jan. 2.\mdash ; The Pink Ink No. 1303 ( Curve No. 3 ) seems very satisfactory , as its absorption begins very rapidly in the orange , whilst the curve rises again fairly high in the blue .
The Blue Ink No. 1265 ( Curve No. is far too opaque in the yellow and green .
The absorption commences very gradually , and there is also an absorption at the violet end .
The ink thus leaves much to be desired .
It should rise at about 18 ( scale ) to 100 per cent. , and stay at that to the end , and this it is very far from doing .
The Yellow Ink No. 1271 ( Curve No. 5 ) is very transparent well into the blue , and will probably do well .
Messrs. Fleming also sent me a set of rubbings of the inks they ordinarily supply for photochrome work .
These inks are permanent or nearly Photochrome Red No. 1197 ( Curve No. 6 ) has an abrupt absorption rather fu1ther in the yellow than the ' Theoretical Red and is so far better , but is far less transparent in the blue , rising at the blue lithium line only to about 5 per cent. as against the 35 per cent. of the " " Theoretical Red Photochrome Blue No. 1199 ( Curve No. 7 ) rises fairly rapidly , but much too Iate , namely , in the blue-green instead of the yellow .
It will not be possible to make bright greens with this ink .
Phobochrome Yellow No. 1198 ( Curve No. ) \mdash ; The absorption is more abrupt than that of the previous yellow , but it is earlier , that is to say , it is not sufficiently transparent to the blue .
It ought to occur at about 46 or 50 instead of 32 ( scale ) .
I measured a set of inks which were printed for me by Mr. Gamble , of the Bolt Court School of Photo-lithography .
The pink ink ( Curve In the above table the column headed ' ' Average aperture for yellow ink\ldquo ; gives , for each spectrum colour , the aperture of the sector which made the light reflected from the white patch equally bright with that reflected from the patch printed with the yellow ink .
The next , headed ' ' White gives the aperture of the sector which made the two patches equal when both were white .
This column would have been constant but for the polarisation by reflection at the surface of the refracting prism already referred to .
From these two columns the proportion of the light reflected by the ink as compared with that reflected by white paper can be oalculated .
The result appears in the column headed ' ' Per ceIlt In the same way the percentage light reflected by the pink and the red patches were found , and are entered in the fifth and seventh columns .
In ths sixth column I have calculated the light which should be reflected from a paper printed both the yellow and the pink inks .
The agreement is very fair .
The actual red reflects more in all the darker parts than calculated .
This may be due to the yellow not being quite so heavily printed , but it is more probably because the pink ink , which was uppermost , was not quite transparent , and some Iight was reflected directly from it instead of from the white paper beneath .
Such light would not sufler by the absorption of the yellow ink .
The " " red\ldquo ; resulting from these intings has a very orange hue , as was to be expected .
A yellow ink by Mander , No. 226 ( Curve No. 25 ) , was less heavily printed than those previously examimed , and therefore seems better in the green and blue green , bnt the absorption in the violet suffers in consequence .
A pink ink by Mander , No. 0227 ( Curve No. 26 ) , comes up fairly in the violet , but is not opaque enough in the green .
's Curvss to Three-colour Work , \amp ; c. As the worst ink was always the Blue , I attempted to find a colour which should be more transparent in the green .
I obtained a bright green rubbing from Fleming , but I found it was green not because it was more transparent in the green and yellow-green , but because it was more opaque in the blue and violet .
There seems to be no ink in the market at present that is sufficiently transparent in the yellowgreen .
19 .
Attempt to the Effect of a Diference in the of the The intensity of the reflected beam is given ( for any one colour ) by I ' , where is the depth and I is the original intensity of the beam .
For another depth of tint Thus the ratio of the ought to be a constant through the spectrum .
Neglecting the readings in which the aperture was too small to bs taken aecurately , to the backlash of the sector ) , and in which also the scattered light is of importance , the following results show a very agreement with theory .
The inks were only rubbed on with the finger , so that the patches were not exactly uniform , especially the lightest one ( 1 ) .
In this case the readings were ehecked by drawing a smooth curve between the points found , and the numbers taken from the curve , where they were Dr. R. S. Clay .
On the Application of obviously wrong .
Otherwise the actual means were used so that the error might be the more readily estimated .
Had I been able to obtain printed uniform patches I have no doubt the ] would have been better .
But even these can leave no doubt of the approximate truth of the law .
60 Application of Maxwelt 's Curves to Three-colour Work , \amp ; c. These ratios omitting the ones in brackets where the lack of light made the readings too uncertain ) do not differ from the mean by one in eight , and omitting the ones marked * , the difference is not more than one in fourteen .
20 .
Experiment to see whether the Colour produced by Blue and side by side matched that when they were printed on one another .
Mr. Gamble prepared for me a card printed with and one printed with yellow .
Then also a third printed with both the blue and the yellow ; this , of course .e , made green .
I cut the blue and yellow into discs about 8 inches in diameter , and the green into one about 4 inches in diameter .
Also I cut a white\mdash ; the white on which they had all been printed\mdash ; into a disc 4 inches in diameter .
Then I set them on a motor , as in Maxwell 's discs , with equal areas of blue and yellow , and of green and white .
These ought when rotated to match if the inks are to be used for process work .
I found there was a very great difference in both hue and luminosity .
To make a match it was necessary to reduce the yellow to only and increase the blue to , and by making the white and the green the match correct in luminosity as well as hue .
Thus , as would be expected since the inks had not abrupt absorptions , it made a very great difference if the colours were superposed or were side by side .

