On the Absolute Measurement of Light .
the sodium test by the yellow flame , or even by the photography of the ultimate lines of sodium in the oxyhydrogen flame and spark .
Metallic sodium renders no lines with one spark when photographed , and with five sparks the lines X 3301*1 and X 3302*5 are both stronger than the yellow lines ( mean X 5893*2 ) .
These lines do not appear on any of the cadmium plates .
The reactions of lead , manganese , and magnesium in the spark are much more sensitive than those of sodium , calcium , or copper .
For instance , 0*00003 mgrm .
of manganese is volatilised by one spark discharge , and yields a spectrum with the following ultimate group of lines : 2949*3 , 2939*4 , and 2933*1 .
As no atmosphere is free from dust , and that of cities is particularly dusty , these mineral constituents must be regarded as possible reagents in cases where there is evidence that very minute quantities of basic substances can initiate chemical reactions and isodynamic changes , such as have generally been considered as spontaneous , and in all cases where a solution in contact with air is liable to be affected .
On the Absolute Measurement of Light : A Proposal for an Ultimate Light Standard .
By R. A. Houstoun , M.A. , Ph. D. , D.Sc .
, Lecturer in Physical Optics in the University of Glasgow .
( Communicated by Prof. Andrew Gray , F.R.S. Received April 5 , \#151 ; Read May 11 , 1911 .
) The measurement of the intensity of a source of light is , it is well known , a somewhat unsatisfactory process .
The eye cannot estimate light intensity ; it can only tell when the illumination of two adjacent surfaces is equal .
If , for example , we desire to measure the intensity of a metal filament lamp , we compare it with a Hefner lamp and say that the intensities are inversely as the squares of the distances from the photometer head , when equal illumination is obtained .
In strictness , however , this method is applicable only when the colours of the two sources , or more accurately when the distribution of energy in the spectra of the two sources , is exactly the same ; for the relative luminosity of the different colours of a spectrum varies with the intensity of that spectrum .
Abney has two well-known curves illustrating Dr. R. A. Houstoun .
[ Apr. 5 , this.* One , which represents the relative luminosity of the different colours of a spectrum at ordinary intensity , has a maximum in the orange ; the other which is for a spectrum with the same distribution of energy , but with an intensity of less than 1/ 100 candle-foot , has its maximum in the green .
If , therefore , we have an extremely long photometer bench , and an experimenter with normal colour vision compares the intensities of the metal filament lamp and the Hefner lamp , at first placing the Hefner lamp one foot from the photometer head and afterwards placing it more than 100 ft. from the latter , he should not obtain the same result both times .
In the first case , owing to the reddish tint of the Hefner lamp , the intensity of the metal filament lamp should appear less .
If , again , a second observer , whose colour vision is slightly abnormal , compares the lamps at the first distance , he gets a third result .
Of course the difficulty does not arise in practice , because the sources to be compared have usually the same colour and the illumination of the field of the photometer does not vary over a wide range .
Still , a standard unit of light should meet all conceivable cases , and we are at present unable to state satisfactorily in terms of our standards , once for all , the candle power of , for example , a mercury vapour lamp .
In order to be definite we must specify , first of all , normal colour vision on the part of the observer , and then we must state the illumination of the fields he compares .
It is , of course , the Purkinje effect , the change from rod to cone vision , that causes all this trouble .
And it is precisely within the range of illumination in common use , 1 to 100 metre-candles , that this change from rod to cone vision takes place .
I think it is now possible to place the photometry of different coloured lights on an exact footing , at least as far as ultimate measurements are concerned , by removing it out of the field of physiology altogether into the field of electricity , which is a much more exact science .
Why can we not measure candle power by means of a thermopile ?
Simply because the thermopile measures the total energy radiated , irrespective of wave-length .
The energy of every radiation receives the same value .
How infra-red and ultra-violet radiations produce no effect at all on the eye , and the light-producing effect of the same quantity of energy is much greater in the middle of the spectrum than at its ends .
If we place in front of the thermopile a light filter which has the property of stopping entirely all the infra-red and ultra-violet radiation and of cutting down the energy of the visible spectrum in inverse ratio to its light-producing effect , that is , if we weight each radiation according to its , then the deflections will be proportional to the light received .
This combination of filter and * ' Colour Vision , ' p. 103 .
On the Absolute Measurement of Light .
thermopile is then a kind of electric eye , which has a property that the human eye has not , namely , the property of registering the intensity of the light to which it is exposed.* The relative visibility of light of different wave-lengths for different intensities of that light has as yet been determined only for one eye , namely , that of Prof. A. Ivonig , and the results are given in a very convenient table in a paper by P. G. Nutting , f Prof. Konig 's colour vision was normal , and so in default of other data we can take this table as applying to the average human eye .
In choosing a filter we must fix on something capable of accurate reproduction .
This rules coloured glasses out of the question , unless they are made under standard conditions .
We are therefore restricted to liquids in glass cells .
Commercial dyes of ill-defined composition are subject to the same objection as coloured glasses .
Solutions of salts which can be varied in strength give us more latitude , and are therefore to be preferred to liquids of a fixed composition .
A series of researches on the absorption of light by aqueous solutions of inorganic salts in the ultra-violet , visible spectrum , and infra-red is now being carried out in the Physical Laboratory of the University of Glasgow , as a result of which I am in possession of a considerable amount of data not at present accessible to others , and from it I have no hesitation in concluding that the most suitable filter is a 3 cm .
thickness of copper sulphate in water of concentration 0'200 gramme-molecule per litre , followed by a 1 cm .
thickness of potassium bichromate of concentration 0-0025 gramme-molecule per litre .
I have examined eight * After writing this paper my attention was called to the fact that an attempt to put this method into practice has already been made by Ch. Fdry ( " Photometrie a lecture direct : Rendement optique de quelques luminaires , " 'Journ .
de Pliys .
, ' 1908 , vol. 4 , p. 638 ) .
He used a solution of copper acetate in water with a radiomicrometer , and has selected the same total thickness of water as myself , namely , 4 cm .
He has not , however , gone into the matter fully , and does not give the fraction transmitted by his filter throughout any part of the spectrum .
Copper acetate alone cannot give the proper absorption in the violet .
I have also used a radiomicrometer , and my experience is that it is inferior in sensitiveness to the apparatus described in this paper , although possessing the advantages of simplicity and cheapness .
J. E. Petavel , in a well-known paper ( " A Preliminary Study for a New Standard of Light , " ' Electrician , ' 1902-3 , vol. 50 , p. 1012 ) , proposes to let the radiation from incandescent platinum fall upon two thermopiles , one with a glass filter and the other with a filter of " black fluorspar " in front of it , and to use the ratio of the deflections as a means of determining the temperature of the platinum .
The brightness of the latter as measured by a photometer at a definite temperature is to be the standard .
His work is thus altogether on different lines .
Black fluorspar seems to me altogether too indefinite a substance for his purpose .
+ " The Luminous Equivalent of Radiation , " 'Bull .
Bureau of Standards , ' 1908 , vol. 5 , p. 279 .
VOL. LXXXV.\#151 ; A. X Dr. R. A. Houstoim .
[ Apr. 5 , different kinds of commercial coloured glass , but none of them is of the least use for the purpose , owing to not stopping the near infra-red .
The absorption of light in aqueous solutions of inorganic salts obeys the following equation , I = I0 .
10-^ , which also defines A , the molecular extinction coefficient of the salt for the wave-length considered .
I0 is the initial intensity of the light , I its intensity after passing through a thickness d of solution , d being measured in centimetres , and c is the concentration of the solution in gramme-molecules per litre .
We may take A as independent of c except for wide ranges and at particular points in some spectra .
The following tables give the values of A for the two salts in question in the ultra-violet and for the copper sulphate in the infra-red .
The values Infra-red .
A. A. CuS04,5H"0 .
\lt ; ?
= 0-03435 .
A. A. CuS04,5Ho0 .
c = 0-03435 .
M- fX .
0-684 6-36 0-910 9-1 0-720 9-8 0-980 7-1 0-750 11 -5 1-07 5 -8 0-794 11 -7 1-17 4-1 0-850 11 1 1 -27 3-0 Ultra-violet .
CuS04 ; 5H20 .
K2Cr207 .
c. A. A. c. A. A. n. M " 7160 o-oi 0-267 52 -0 0 *0001 0-241 0-03 0-278 34 -0 \#151 ; 0-269 11100 0-05 0-282 20 -2 \#151 ; 0-284 7440 0-287 13-0 \#151 ; 0-291 5070 o-io 0-291 10 -1 * \#151 ; 0-300 2720 0-292 8-6 0 *0005 0-308 2320 0-296 6-5 0 *0003 ?
0 -316 2080 0*30 0-299 3-37 0 -0005 0-325 2320 , j 0-309 1 -48 0 '0001 0-331 2720 0-313 0-78 \#151 ; 0-336 5070 ?
0 -322 0-08 \#151 ; 0-349 7440 ?
0 -375 10000 0-397 7440 0 '0003 0-410 3080 0 '001 0-416 1030 On the Absolute Measurement of Light .
1911 .
] for the visible spectrum have already been determined by Griinbaum* and by A. S. Bussell and the author.f The values for copper sulphate in the infra-red were obtained with a linear thermopile , the source of light being a Nernst glower .
The values for the ultra-violet were obtained by means of a quartz photometer of original design which has recently been worked out by John S. Anderson and the author , and a description of which is already in course of publication .
It is , of course , impossible to give values further into the infra-red owing to the absorption of water .
At the strength and thickness used in the filter potassium bichromate does not exercise any appreciable absorption in the infra-red .
The values of A cdfor the two filters have been calculated from the table ?
and are shown in the following diagram .
The values for copper sulphate art .
shown by \#169 ; 's and the values for potassium bichromate by x 's ; Griinbaum 's values are taken for the copper sulphate and our own for the potassium bichromate .
The smooth curve is the sum of the ordinates for each salt .
The logarithms of the columns D and E in Nutting 's table were then calculated , their sign was changed , and they were plotted on the diagram as the dotted curves .
In order to facilitate comparison with the smooth curve , the vertical scale for the dotted curves is displaced 008 up .
Of course , we can add any constant quantity to the ordinates of the smooth curve without altering the relative absorption of the different colours .
o o o o. * 'Ann .
d. Phys. , ( 4 ) 1903 , vol. 12 , p. 1004 .
t 'Roy .
Soc. Edin .
Proc. , ' 1908-9 , vol. 29 , p. 69 .
X Dr. R. A. Houstoun .
[ Apr. 5 , On the one side the smooth curve agrees with E , on the other side with D. D is for an illumination of 2 30 metre-candles and E for an illumination of 9'22 metre-candles .
We are therefore justified in assuming that our filters weight the radiation of the visible spectrum correctly according to its visibility to normal colour vision , the illumination of the field being about 6 metre-candles .
If we take as our standard another illumination of the field , it is easy to shift the minimum of the curve by diminishing the concentration of the copper sulphate and increasing the concentration of the potassium bichromate or vice versa .
The strength of the copper sulphate , however , cannot be weakened much unless an additional thickness of water or of , say , ferrous ammonium sulphate is used , otherwise there will not be a sufficient margin of safety in the infra-red .
Aqueous solutions of ferrous ammonium sulphate are , it is now known , very effective in stopping heat rays while allowing the light rays to pass.* They are much more efficient than alum , which in this respect is no better than water .
If we neglect the light absorbed in the glass , the fraction of light transmitted by our filter is equal to 10-'-6Ac x j[0~0'0025Ak x jc\gt ; where Ac is the molecular extinction coefficient of the copper sulphate , AK is the molecular extinction coefficient of the potassium bichromate , and the fraction that would be transmitted by 4 cm .
of pure water alone .
In order that the effect due to the different constituents of the filter may be seen , I have calculated 10-0'6Ac , lO-0'0025^ and k throughout the spectrum , using for k the values of E. Aschkinass.f The results are given in the second , third , and fourth columns of the following table .
The product of the three factors , that is , the transmissivity of the whole filter , is given in the fifth column .
Of course , in addition to the above , the absorption of the glass of the cells has to be considered .
Glass absorbs all radiations below ( \gt ; 330 ya and above 3-00 ya. The filter is weakest in the infra-red at E27 ya. In the normal energy spectra of our ordinary light sources the ordinates at 1*27 ya are about twenty times or thereabouts the ordinates at 05 ya. With the strength used , the stopping power of the filter is hence ample , but this point in the spectrum must be watched if the concentration of the copper sulphate is diminished much .
Having found by calculation that the radiation received by a thermopile * R. A. Houstoun and J. Logie , ' Phys. Zs .
, ' vol. 11 , p. 672 .
t ' Wied .
Ann. , ' 1895 , vol. 55 , p. 401 .
1911 .
] On the Absolute Measurement of Light .
A. 10-0-6AC .
10 -0-0025Ak .
0*227 \ \lt ; 6*3110~n \lt ; 1-3 10-1S 0-241 J 1 -3 10~18 0-284 6-3110"11 2 -5 10-19 0-300 1 -20 10-2 1 -6 10~7 0-308 0-107 1 -6 10~6 0 *325 0-89 1 -6 10-\#174 ; 0-336 1 -9 10-13 0-397 2 -5 10-19 0-416 2 -6 10~3 0-463 \ \lt ; 0-986 0-105 0-471 J 0-158 0-480 0-986 0-234 0-489 0-977 0 -355 0-499 0-968 0-501 0 -5086 0-951 0-661 0 -5461 0-792 0-982 ' 0 -5780 0-521 1 -oo 0-6004 0-292 \#151 ; 0-6239 0-108 \#151 ; 0 -6452 0-0336 \#151 ; 0-684 1 -54 10~4 0-720 1 -32 10~6 \#151 ; 0-750 1 -26 10_* \#151 ; 0-794 9 -55 10~8 \#151 ; 0-850 2 -2 10- 7 \#151 ; 0-910 3 -5 10-fi \#151 ; 0-980 5 -5 10~5 \#151 ; 1 -09 3 -3 10-4 \#151 ; 1-17 3 -5 10~3 \#151 ; 1 -27 1 -6 10~2 \#151 ; 1 -40 k. \ 10"0*6 Ac X 10-0*0025 Ak x k. 0-99 \ \lt ; 8 10-29 0-99 J 0-99 1 -6 10-'29 1 -oo 1 9 10~9 \#151 ; 1 -7 IO-7 \#151 ; 1 -5 10~7 \#151 ; \lt ; 1-9 10-13 \#151 ; \lt ; 2-5 10"19 \#151 ; \lt ; 2-6 lO"3 \lt ; 0-105 \#151 ; \lt ; 0-158 \#151 ; 0-229 \#151 ; 0-347 \#151 ; 0 -485 \#151 ; 0-63 \#151 ; 0-77 \#151 ; 0-52 0-99 0-29 0 -99 0-107 0-99 0 -034 0-99 1 -53 10-4 0-95 1 -25 KT6 0-91 1 -14 10"7 0-92 8 -8 10-8 0*86 1 9 10- ?
0-73 *2 -5 10-\#174 ; 0-20 1 -5 10~5 0-48 1 -6 10~4 0-054 1 -9 IQ"4 0-0085 1 -3 lO"4 2 10"41 \lt ; 2 10~41 Never \gt ; lO^18 all the way to 8 " 5 fx .
through this filter should be proportional to the light incident on the filter , I proceeded by direct experiment to determine if this actually was the case .
I had three thermopiles at my disposal , a Rubens linear one with iron-constantan couples and two older ones , of different types , but both with antimony-bismuth couples .
The Rubens thermopile consists of 20 couples on a length of 2 cm .
, the wires being soldered together with silver beads , which are flattened into discs of 1 mm. diameter .
Hence its receiving area is about 0*157 sq .
cm .
In the case of each of the other two , the receiving area was about 1 square inch , and the number of couples was greater , yet the Rubens thermopile was as sensitive , even without its reflectors , reached the steady state in a shorter time , and had a very much steadier zero .
Only the Rubens thermopile was used .
The galvanometer was a Du Rois Rubens ironclad one , of resistance 10 ohms , the property of the Carnegie Trust for the Universities of Scotland .
Dr. R. A. Houstoun .
[ Apr. 5 , In order to protect it from vibration , it was suspended from a bracket in the wall by three wires each about H metres long , and hung with its three levelling screws clearing the table by about 1 cm .
Between the table and levelling screws loose wads of cotton wool were placed for the purpose of damping any vibrations that might arise .
The lamp and scale were \\ metres from the mirror .
The resistance of the thermopile was 5 ohms .
At the sensitiveness used , the period was 3 seconds and ^ mm. on the scale indicated a current of 10~9 amperes .
It soon became evident that this high sensitiveness was to be fully utilised .
When the thermopile was set up at a distance of 33 cm .
from a 32 c.p. carbon glow lamp with the filters in front , the deflection was only 17 |-mm .
When this lamp is run at its marked voltage , 250 volts , about 2-6 per cent , of its total radiation is light , that is , lies between 0-400 [ x and 0760 / x , but only a fraction of this 2'6 per cent , gets through the filter .
Perhaps altogether about 1/ 1000 of the total radiation passes through the filter .
The thermopile was , then , set up at a distance of 33 cm .
from the 32 c.p. carbon glow lamp with the filters in front , the K2O2O7 filter being next the thermopile .
On the other side of the lamp was a photometer bench , at the other end of which there was a 125 volt tantalum lamp that was run off a storage battery .
The carbon lamp and the tantalum lamp were compared by a wedge photometer .
When this comparison was being made the tantalum lamp was always shunted by a current balance and constant resistance , the latter being chosen so that the open part of the scale of the balance was in use , and from the indications of the balance the voltage on the tantalum lamp could be kept constant to 1/ 800 by means of a rheostat .
At the voltage used the horizontal candle power of the tantalum lamp , measured by a Hefner lamp , was 7T3 British candles .
Headings were then taken for different voltages alternately with the thermopile and with the photometer .
One set of results is given in the following table and plotted in the following curve .
Test of a 32-candle power Glow Lamp .
I Voltage .
Galvanometer readings .
Photometer readings .
Throws .
Mean .
Individual settings .
Candle power .
207 -1 5-5 4-0 4-0 3-0 4-5 4 -2 117-5 114-5 112-5 114-5 6-15 222 *5 8-0 7-0 6-5 6-7 7-0 96 -5 98 -5 96 -5 98 -5 11 -2 251 -6 18 -0 16 -0 17 -5 18 -0 17 -4 77 -5 76 -5 76 -5 76 -5 24 -9 283 -7 38 -0 38 -0 36 .5 36 '5 37 -2 59 -5 57 -5 59 -5 58 5 53 -8 ' 304 *7 56-5 56-5 60-0 59 58-2 48-5 47*5 49-5 49 5 88 -1 ' ' \#166 ; : 1911 .
] On the Absolute Measurement of Light .
Every reading taken is given .
At each voltage two readings were first taken with the thermopile , then two settings of the photometer were made ; the remaining thermopile readings were then made , and , finally , the last two photometer readings were taken .
The numbers given in the table as photometer settings are the distances of the wedge in centimetres from the comparison lamp .
The total distance between the two lamps was 220 cm .
The curve should , of course , be a straight line .
The agreement is very good when we consider that the galvanometer readings are ^ mm. at 1^ metres distance from the mirror , and that for the last point the photometer was getting rather near the comparison lamp for the inverse square law to hold .
I have altered the strength of the K2C12O7 to 0'0125 and of the CuS04,5H20 to c = 0'350 , and the proportionality still holds .
A thermopile used with such a filter can therefore be employed for measuring candle power , more especially mean spherical candle power , because it is only necessary to set up equally sensitive thermopiles over the sphere and to connect them in series with the one galvanometer .
All difficulties connected with the integration thus vanish .
The method works also quite as well in broad daylight as in a darkened room , the deflections being the same in each case .
I think , however , that it is not suitable for commercial application , on account of the high sensitiveness of the galvanometer required .
Its importance lies in the fact that it can be used for defining our unit of light and for providing a satisfactory basis on which lights of different colour can be compared , irrespective of intensity and of idiosyncrasy on the part of the observer .
The amount of light lost by reflection at the glass surfaces and by 284 On the Absolute Measurement of Light .
absorption in the glass of the cells can be easily determined by filling the cells with water and using a spectrophotometer to determine the fraction of the incident light transmitted .
It is sensibly the same throughout the visible spectrum , and has the value 084 for both of the cells employed .
It is easy to determine what the galvanometer deflections would have been , had there been no reflection losses or absorption losses in the glass ; we have only to multiply by l/ ( 084)2 .
In determining the distance of the thermopile from the source of light , we have , of course , to multiply each thickness'of glass and liquid traversed by its index of refraction .
I propose therefore to define the unit of light intensity as follows :\#151 ; The unit of light intensity is that source the total intensity of radiation from which at an optical distance of 1 metre after passing through an ideal filter would be x ergs/ cm.2 sec. , the ideal filter to be one possessing the light absorbing properties of a 3 cm .
thick aqueous solution of CuS04,5H2(A of strength 0200 gramme-molecules per litre and a 1 cm .
thick aqueous solution of K2Cr207 of strengthO '0025 gramme-molecules per litre , but neither to reflect nor to absorb any light in any other way .
I think it better to eliminate the properties of the glass in this way .
Different cells have quite an appreciable difference in absorption and the transmission coefficient of a cell is easy to determine .
The definition , has the advantage of connecting up light closer with the C.G.S. system .
By means of his pyrheliometer K. Angstrom* has been able to measure the radiation from terrestrial sources to less than 1 per cent. , the chief difficulty being to allow for want of " absolute blackness " on the part of the receiving surface .
The radiation-receiving surface in my experiments was about 0157 sq .
cm .
, and the distance from the source 33 cms .
By increasing the light-receiving area and using Angstrom 's method it should be possible to determine x with sufficient accuracy .
For the standard candle in the units specified it is roughly 0*8 .
* 'jAstrophys .
Journ. , ' 1899 , vol. 9 , p. 332 ; 'Phys .
Rev. , ' 1893 , vol. 1 , p. 365 .

