The Evaluation of the Ultraviolet Radiation From the Sun and Other Sources.
[Read before Otago Branch, November 12, 1940; received by the Editor, November 19, 1940; issued separately, June, 1941.]
Two methods for the measurement of ultraviolet radiation have been developed for use in New Zealand, viz.:
the balanced (or differential) thermopile method, which is an absolute method provided a standard source of thermal radiation is available;
the photoelectric cell and direct-current amplifier method.
In both methods the principle of exclusion filter radiometry was used. The balanced thermopile method was applied to measure the amounts of radiant energy from artificial sources in the wavelength ranges 200–290, 290–313, and 313–380 millimicrons. By means of the photoelectric cell method measurements were made of the ultraviolet energy in the radiation from artificial sources and from the sun, and the spectral energy distribution of solar radiation in the region 290–340 millimicrons was investigated. A photographic method for determining relative spectral energy was devised and tested with mercury vapour lamps.
From the results deductions are made of the efficiency of certain ultraviolet lamps for radiation in different regions of the near ultraviolet portion of the spectrum (on which depends their suitability for medical applications), and of the solar ultraviolet radiation at Dunedin.
The properties of u.v.* radiation, some of which are of considerable importance, have been known for many years, but convenient and reliable physical methods for its measurement have proved difficult to find. Much of the experimental work on it has been of little value, because insufficient information has been recorded concerning the quality and quantity of the radiation used. The difficulties which have been encountered in the construction of reliable standards of u.v. radiation have only recently been overcome.
The increasing use of sources emitting radiation in the u.v. region of the spectrum, particularly by the medical profession, makes it important that some standard method for measuring the spectral distribution of this energy should be adopted. In 1932 Dr. W. W. Coblentz, of the United States Bureau of Standards, presented experimental evidence to the Copenhagen Meeting of the International Committee of Measurement and Standardization of Ultraviolet for
[Footnote] * The following abbreviations are used: u.v.—ultraviolet; p.e.—photoelectric; mμ.—millimicron; μV.—microvolt.
Use in Medicine. He recommended the evaluation of u.v. energy by means of radiometric instruments in absolute units, and that the correlation of physical measurements with biological and physiological reactions be left for the future. An accurate knowledge of the spectral distribution is generally unnecessary. Coblentz (1935) has shown that the information required for medical applications is the u.v. radiation comprised in the approximate wavelength ranges: 200–290, 290–313, and 313–380 mμ. These bands were selected for their recognized biological actions, viz.,
200–290 mμ.—the germicidal action.
290–313 mμ.—the erythemal and antirachitic actions.
313–380 mμ.—the effective healing of certain cutaneous diseases by the deeply penetrating radiation.
The principal methods which have been applied to this problem are spectroradiometry, photography, and filter radiometry, and the success met with depends chiefly on the type of source under consideration and the suitability of the apparatus available. After consideration it was decided that filter radiometry offered most advantages for application under New Zealand conditions. This comprises the use of filters with p.e. cells and with thermopiles. Thermopiles were first used to develop a reliable laboratory method which required inexpensive and easily obtainable apparatus. Later a specially designed p.e. cell was used as the basis of a method for measuring the u.v. energy and its spectral quality in radiation from the sun and from artificial sources. On account of the rapid development of p.e. cells in recent years and the improved stability of d.c. amplifier circuits, most workers, e.g., Coblentz and Stair (1934), Poole and Atkins (1935), now use the p.e. method of measurement.
Both methods were successfully employed with the available apparatus, but further work is desirable to improve the accuracy of the first method, and to redesign for increased portability the amplifier used in the second method. Fuller investigations of solar u.v. radiation could then be made conveniently with the portable p.e. apparatus. The investigations reported in Part I of this paper were completed in 1938; those in Part II were mainly performed in 1939, but were continued in 1940. Little work on u.v. radiation of a quantitative physical nature has previously been performed in New Zealand.
Part I.—The Balanced Thermopile Method.
1. Principles, Apparatus, and Procedure.
(a) The Exclusion Filter Method.
The principle of the balanced thermopile method of filter radiometry is the simultaneous exposure of two thermopiles of nearly equal sensitivity to a source providing uniform radiation at the thermopiles, which are connected differentially to a galvanometer. A filter which excludes all wavelengths in the spectral region to be evaluated is placed over one thermopile, while a quartz plate, which freely transmits those wavelengths, and which has approximately the same transmission losses as the exclusion filter throughout the remaining part of the spectrum, is placed over the other. The excess u.v. radiation transmitted by the quartz plate over that transmitted
by the exclusion filter generates a resultant e.m.f., which is measured. The quartz plate and filter are now interchanged, and the resultant e.m.f. again measured. The mean of these two e.m.f.'s is the e.m.f. generated by a “mean thermopile” (i.e., one whose sensitivity is the mean of the sensitivities of the two thermopiles used), when radiation equal to that excluded by the filter falls on it. Hence if the sensitivities of the thermopiles have been determined by means of a standard source of thermal radiation, then the energy of the u.v. radiation can be evaluated in absolute units.
There are two basic assumptions underlying the use of a thermopile in the absolute measurement of radiant energy. First, that the energy absorbed is a constant fraction of the incident energy over the range of wavelengths concerned, and secondly, that the response of the thermopile to the energy absorbed per second is linear and non-selective. If proper precautions are taken in the use of carefully constructed thermopiles these assumptions are found to be justified over a wide range within the limits of accuracy required for the energy measurements reported in this paper.
The method used to measure the steady e.m.f. of the thermopile correct to ±0·05 μV. was to balance the e.m.f. against the fall of potential across a standard 0·1 ohm resistance produced by a small adjustable current. This current passed through a standard 1000 ohm resistance producing a potential difference E volts, which was measured by a potentiometer. Then the e.m.f. generated by the thermopile = E × 10−4 volts. A Cambridge spot galvanometer (sensitivity 34 mm. per microamp.) fulfilled all requirements for the null instrument required to balance the e.m.f. of the thermopile. This method enabled the e.m.f.'s to be measured and reproduced to within 0·1 μV., so the effect of parasitic thermal e.m.f.'s was negligible.
Two Moll small surface thermopiles (No. 695 and 697) were used. The glass windows protecting the blackened elements were removed, as it was found that they had zero transmission for wavelengths less than 290 mμ. Two standards of thermal radiation (C 263 and C 279) supplied and certified by the National Bureau of Standards, Washington, were used to calibrate the thermopiles. These were specially aged carbon filament lamps. The work was done in a darkened room at steady temperature and all the precautions recommended by the Bureau of Standards were adopted. The sensitivities of the thermopiles were determined to an accuracy of 1 in 200, which was adequate for our purpose. The factors which limited this accuracy were the maintenance to within ±.05 milliamp. of the specified values of the current from a d.c. generator through the standard lamp, and the difficulty in aligning the thermopiles with the sighting marks on the lamp. Close agreement was obtained between the sensitivity values deduced from each of the two standards, and the final values were:—
Thermopile 695, 2·20 microwatts per cm.2 per μV.
Thermopile 697, 1·90 microwatts per cm.2 per μV.
Hence the mean sensitivity = 2·05. A portion of the observations is given in Table I to show the consistency attained.
|Current in milliamps.||E||e1||e2||e0||Thermopile sensitivity.|
E = radiant-flux density in microwatts per cm.2.
e1 = e.m.f. in μV. on exposure to standard lamp.
e2 = mean e.m.f. in μV. when shutter closed.
e0 = e1 − e2 = e.m.f. due to radiation from standard lamp alone.
The filters used were: (1) potassium hydrogen phthalate on cellophane, (2) barium-flint glass, (3) Ilford “Q” filter, (4) two Noviol A filters, (5) crystalline quartz plate, (6) two water cells with end plates of crystalline quartz. The properties of these filters are given in Table II, where the thickness is in mms., the wavelengths in mμ and the transmissions in percentages.
[The section below cannot be correctly rendered as it contains complex formatting. See the image of the page for a more accurate rendering.]
|6.||Water Cell||3.0 quartz 10.0 water||84||86||86||86||87||88||88||88||89|
From these figures smooth curves were drawn to represent the transmission characteristics of the filters. The filters Nos. 1, 2, and 3 are the exclusion filters selected as the best available for evaluating the radiant energy within the bands 200–290, 290–313, 313–380 mμ. They have relatively sharp cut-offs at about the required wavelengths and, with the exception of No. 1, they have stable properties. No. 1 was prepared from details given by Withrow (1931). Several such filters were prepared together, and each used a short time only, because they deteriorate slowly under the action of u.v. light. The transmissions of filters Nos. 2, 4, and 5 were determined at the Bureau of Standards; those for filters Nos. 1 and 3 were measured by the photographic method described later, and the transmission for the water
[Footnote] * Transmission less than 0.5 per cent.
cell was taken as equal to that determined by Coblentz, Stair, and Hogue (1931) for a cell with the same thickness of distilled water and quartz. A thickness of 1 cm. of water absorbs a large proportion of the infra-red energy from the source, so the proportion of u.v. energy in the radiation falling on the filters is considerably increased. The purpose of the two Noviol A filters is best explained by considering the determination of the radiant energy in the region 200–313 mμ. using the barium-flint as exclusion filter and the quartz as balance filter. For wavelengths greater than 400 mμ. the transmissions of the barium-flint and quartz are not exactly the same. The small correction required can be found by taking two additional measurements with a Noviol A filter added in front of each thermopile, and then with the barium-flint and quartz filters interchanged.
Filter holders consisting of blackened brass strip were made to fit firmly in position over the body of the thermopiles. Filters could be inserted in place and interchanged without fingering them by means of small tabs attached to their edges and a pair of tweezers. A camel hair brush was used to remove dust from the surfaces of each filter and before use they were cleaned by gently rubbing with cotton wool moistened with alcohol. The thermopiles were placed close together and thermally insulated by thick felt and cotton wool, and the temperature of their immediate surroundings was read on a thermometer. The thermopile assembly with water cells and filters in position is shown in Fig. 1, plate 8. Most of the indirect radiation from the source was excluded by large dull black screens and the direct radiation could be admitted by opening a shutter covering an opening 10 × 15 cms. The observer controlled the shutter from a distance by a thread to prevent any radiation from him falling on the thermopiles during observations.
[The section below cannot be correctly rendered as it contains complex formatting. See the image of the page for a more accurate rendering.]
Because it is not possible to construct a filter with a perfectly sharp cut-off curve, it is necessary to correct the observed radiant energy excluded in order to obtain the energy in the band of wavelengths totally absorbed by the filter. This is done by determining the filter factor (F), which is expressed mathematically by the equation F × ∫λ2λ0(I1−I0).dλ=∫λ1λ0I.dλ where λ0<λ1<λ2, and the filter completely absorbs all wavelengths up to λ1. In the wavelength range λ to λ + dλ, I is the intensity of the radiation used, I1 is the intensity of radiation incident on the filter considered, and I0 the intensity transmitted by the filter, respectively. I1 is slightly less than I on account of absorption in the water cell and in the quartz filter, which are accessory filters required in the experimental investigation.
(b) Photographic Method for Relative Spectral Energy.
An approximate knowledge of the spectral energy distribution of the source under investigation is required in order to obtain the filter factor, and in the case of a line spectrum the relative energy of each line has to be found. The most accurate method is to use a spectroradiometer, but as no such instrument was available and as a high accuracy in the spectral distribution was not necessary, a photographic method was devised for the purpose, which utilized only a
small Hilger quartz spectrometer and proved quite satisfactory for mercury vapour lamps. The method is a simplified comparative one and is similar to that recently described by Maddock (1940), which he claims gives results within 10 per cent. of the true values and for strong lines within 5 per cent. The accuracy of our method is of the same order, which was sufficient for the determination of the filter factors.
Schwarzschild's law states that the density of a photographic emulsion is proportional to I.tp, where I is the intensity of the incident light, t the time during which the incident light falls on the emulsion, and p is a constant. This relationship is the basis of the method used, although there is now sufficient evidence to prove that p is a function of the exposure (I.t), when the range of exposures is great. A determination of p for the Ilford Paget Half-tone plates used in this work was carried out for low density exposures in the u.v. region, and a value of 0·9 was obtained. This is in agreement with the value 0·86 used by Brock (1933) in his investigation using radiation in the near u.v. region. If progressively decreasing exposures of a photographic plate are made of the spectrum from the source produced by a quartz spectrograph, then on development under fixed conditions exposures required for minimum perceptible density (M.P.D.) for each wavelength can be found. This principle was used because a microphotometer for comparing densities of spectral lines was not available. From the series of exposures made on the same plate of the spectra of the two u.v. lamps under comparison, the exposure times t and ts to produce M.P.D. for a line of given wavelength is determined. The ratio of the intensities of these two lines is then found from the relation I.tp = Is.tsp, where the subscript s refers to the standard source. In this way the relative spectral energy distribution of the test source is found, that of the standard source being known. In a similar way the fraction of the energy of a given wavelength transmitted by a filter can be measured by comparing the two series of exposures made on the same plate with and without the filter between the spectrograph slit and the source.
A quartz lens was used to produce an image of the source on the slit and to illuminate uniformly the collimator lens. Preliminary tests were made to arrange the distances of the two lamps under comparison so that the intensities they produced at the plate were comparable. Then 16 exposures were made consecutively for each of the lamps, the plate being displaced 2 mms. after each exposure. The exposure times were increased from 1 second to several minutes in geometrical progression with common ratio = 1·5 in order that a linear density change would be obtained. The arrangement of the two spectra are shown in Figs. 2 and 3, Plates 8 and 9. Fig. 2 shows the comparison of the standard source with the “Osira” lamp. Fig. 3 shows the comparison of the standard source with the standard source and cellophane filter. It will be noticed that the spectra were arranged to bring the places of M.P.D. for each towards the middle of the plate. In this way the effect due to fogging of the plate which would be greatest at the edges was minimized, and further the development could take place as uniformly as possible over the areas enclosing the M.P.D.'s.
In the method described it is important that the conditions are such that sufficient contrast is obtained in the density of the images. By printing out the plates with ultra-contrast bromide paper, under conditions which were found experimentally, this was ensured. The plates and prints were carefully processed using a standard procedure to secure uniformity.
To illustrate the reduction of the data obtained by this method, the results of the comparison of the standard source with another mercury vapour lamp called the point source are shown in Table III.
[The section below cannot be correctly rendered as it contains complex formatting. See the image of the page for a more accurate rendering.]
|Wavelength length in mμ.||Number of exposures from datum to M.P.D.||Exposure times in secs. for M.P.D.||Ip/Is||Relative Intensity|
2. Results and Discussion.
Four sources of u.v. radiation were tested.
A large-surface quartz mercury vapour lamp, which was operated with a current of 2·5 amps. and a series resistance of 20 ohms. The power of the lamp alone was 175 watts. This lamp, supplied by the Thermal Syndicate, was similar to one that had been examined spectroradiometrically by Coblentz, Stair and Hogue (1931). Their recorded value for the distribution of energy was used for our lamp, which was adopted as the standard source.
A point-source mercury vapour lamp operated with a current of 3 amps. and a series resistance of 20 ohms. The power of this lamp was 40 watts, exclusive of the resistance.
An “Osira” electric discharge lamp. This is a commercial general service u.v. lamp supplied by 240 volt a.c. and used with a choke coil and 10 microfarad condenser. This lamp is of the high pressure type and the inner quartz tube is protected by a thin glass globe, so no wavelengths less than 285 mμ. are radiated. The power of the lamp alone was 125 watts.
A carbon arc of the regulation pattern used for street lighting and run from the a.c. supply. Impregnated carbon electrodes are used to increase the visible light. They are 10 mm. in diameter and were operated on 12·3 amps. and 30 volts across the arc.
The spectral quality of the point source and of the “Osira” lamp was obtained by comparison with the standard source by the photographic method. As the spectrum of the carbon arc consists of a large number of lines in the u.v., it could not be compared with
the standard source by this method. The transmission of the Ilford “Q” filter and of the cellophane filter were also determined by the photographic method. In order that the maximum transmissions of these filters could be found, each was tested with a thermopile and suitable filters whereby only transmissions for wavelengths longer than the cut-off wavelengths were determined. The results obtained by the photographic method are summarized in Table IV.
|Wavelength in mμ.||Standard||Relative Intensities Point Source||Osira||Transmissions (per cent.)
Cellophane Ilford “Q”
|Wavelength in mμ||Relative Intensity||Transmission Quartz-“Q” Filter||Transmission Water Cell||Radiation Excluded|
|a||b||c||d||e = b × c × d|
|29.66 (1)||24.68 (2)|
represents the total incident energy for wavelengths less than 380 mμ., this being the wavelength below which the “Q” filter completely absorbs the radiation.
represents the observed total energy excluded by the “Q” filter up to wavelength 436 mμ., above which the “Q” filter and quartz plate have practically equal transmissions and so compensate.
These spectral energy distributions are only approximate, partly on account of uncertainty in the distribution of the standard source and partly due to limitations of the method used for comparison. These results are, however, only required to determine the filter factors which fortunately are not greatly affected by small errors in the relative energy distribution. An example is given of the method used in determining the filter factor (F). The data for determining F for the Ilford “Q” filter within the region 200–380 mμ. using the standard source are shown in Table V. Column c gives the difference in transmission between the quartz plate and the “Q”
filter for wavelengths less than 440 mμ., and for wavelengths above this they have nearly the same transmission, the use of the Noviol A filters enabling a correction to be made for the slight inequality. Column d gives the spectral transmission of the water cells. Column e is the product of columns b, c, and d, so it represents the radiation excluded by the “Q” filter, and the sum of the intensities in this column generates the resultant e.m.f. in the balanced thermopiles. Hence in order to obtain the u.v. radiation within the region 200–380 mμ. this resultant e.m.f. (after correction by observations with the Noviol A filters) must be multiplied by the mean thermopile sensitivity and by the filter factor for the “Q” filter.
Hence filter factor for “Q” filter = 29·66/24·68 = 1·20.
In a similar way the filter factors for other filters, sources, and spectral regions were obtained. These are presented in Table VI. For the approximate spectral quality of the carbon are, which could not be found by the photographic method, the values obtained by Coblentz, Stair, and Hogue (1932) have been used to calculate the filter factors.
|Source||Spectral Region (mμ.)||Filter||Filter Factor (F)|
The method of evaluating the results from the observations on the resultant e.m.f. of the balanced thermopiles and the method of making the correction previously referred to with the Noviol A filters is illustrated in Table VII. The observations are for the standard source used at one metre distance from the thermopiles. The e.m.f.'s are in μV., and e1 = e.m.f. generated when thermopiles are exposed to direct radiation, e2 = e.m.f. generated when thermopiles are exposed to indirect radiation.
|Position of Filters||e1||e2||e1—e2|
|Thermopile 695||Thermopile 697||Shutter open||Shutter closed|
|+Noviol A||Noviol A|
|Noviol A||+Noviol A|
Hence, mean observed e.m.f. with barium-flint and quartz filters =½ (21·70 + 7·93) = 14·82 μV. Since the transmission of the Noviol
A filter for long wavelengths is 89·5 per cent., the mean correction to the observed e.m.f. =½ (2·77—2·06) × 100/89·5 = ·40 μV.
A summary of the results obtained for the u.v. radiation emitted by the four sources is given in Table VIII, the values recorded being the means of several independent sets of observations. An approximate check on the results was made by working at different distances and applying the inverse square law. In all cases the ratio of the length of source to the distance from source to thermopiles was less than 0·1, so the inverse square law should be correct to within 1 per cent. In Table VIII, e = mean observed e.m.f. obtained when the appropriate exclusion filter and the quartz plate are interchanged, and e0 = mean correction determined by Noviol A filters. The u.v. radiant-flux density in the specified spectral region is the product (e—e0) X filter factor X mean thermopile sensitivity. The thermopiles were calibrated twice during the course of these investigations and it was found that their sensitivities remained practically constant, the mean sensitivity being 2·05 microwatts per cm.2 per μV.
|Source||Spectral Region in mμ.||Distance cms.||e in μV.||e0 in μV.||e—e0||Filter Factor.||Energy Output at 1 m.|
The u.v. radiant-flux densities in the last column of Table VIII have been reduced to give the value in microwatts per cm.2 at 1 metre distance from the source. The maximum variation between these results is about 6 per cent., being greatest for the carbon are as was expected from its fluctuations in total output. A discrepancy of this order is reasonable on account of two principal errors: (1) the
currents through the lamps were controlled to ± ·02 amp., and this variation produced a change in the total radiation output of 2 to 5 per cent. depending on the source, (2) the amounts of radiation falling on the pair of thermopiles may not be equal in all tests. Measurements- of the radiation falling on each thermopile showed that the difference was as high as 4 per cent. As the error was smaller the greater the distance worked at, and it was found to reduce the observed output, this effect probably explains why the results in the last column of Table VIII generally increase at the greater distances.
Table IX provides a comparison for the sources investigated of the u.v. radiant-flux density in the three spectral regions examined, and of the u.v. efficiency. The total radiant-flux density in microwatts per cm.2 at 1 metre was measured for each source by each thermopile with the water cell removed.
|Source.||Total Radiation.||Radiation 200–290.||in Spectral 290–313.||Regions. 313–380.||U.V. Efficiency per cent.|
|“Osira” Lamp||610.0||12.6 *||29.6||7.0|
The standard source had the highest u.v. emission and 3 per cent. of its total radiation output was in the region 290–313 mμ., which produces the erythemal and antirachitic actions. A comparison of the energies in each region as determined by the balanced thermopile method (Table IX) can be made with that which was assumed for the spectral distribution of the standard source (see Table IV). From the data in Table IV the relative energy in the three regions is found to be 10·1: 24·8: 17·8, which is in fair agreement with the values for the standard source in Table IX, and hence supports the assumed spectral energy distribution for this source (which was the basis for the determination of filter factors). The energy measured by the thermopile for the region 200–290 mμ. exceeds that obtained by summing the energy of each line. This can be accounted for by the continuous radiation over an appreciable region which appears in the spectrum of the standard source shown in Fig. 4, Plate 9. Since the thermopile integrates the radiation it would yield the higher result.
The point source mercury vapour lamp emitted so little u.v. energy that no reliable measurements with the cellophane filter could be made. This source contained a tungsten electrode which glowed brightly during use, and its u.v. output dropped after some hours of running due to darkening of the quartz close to the discharge. On account of its glass globe the “Osira” lamp emitted negligible energy below 290 mμ. The spectrum of this lamp (see Fig. 4, Plate 9) shows that it emits some continuous radiation in the region 313–380 mμ. and this agrees with the relatively higher value obtained by the thermopile over that obtained by summing the principal emission lines in this region. Since the thermopile integrates all the
[Footnote] * Includes the region 200–290 mμ.
energy whether in the form of lines, bands, or continuous radiation, it has an advantage over the spectroradiometer. The results obtained with the carbon arc are the least reliable because its output fluctuated, and on account of the proportion of continuous radiation in its emission (see Fig. 4, Plate 9) it was impossible to measure filter factors for this source by the photographic method.
Atmospheric absorption had no appreciable influence on this work as it is negligible for 1 metre path length for wavelengths exceeding 230 mμ. The absolute values obtained in this work are dependent on the value of filter factors which are based on an arbitrarily chosen standard. The slight uncertainty due to this cause can be removed when a standard source of u.v. radiation is available. The balanced thermopile and exclusion filter method has been shown to be a reliable, simple, and fairly accurate method for obtaining the information usually required for medical and other applications about the emission of a source of u.v. radiation.
Part II.—The Photoelectric Cell and Amplifier Method.
1. Principle and Apparatus.
The principle used in Part I, that of exclusion filter radiometry, was again used, but in contrast to thermopiles, p.e. cells are selective receivers of radiation, and in general it is necessary to amplify their response in order to measure it reliably. For the measurements a titanium p.e. cell was selected which had an upper limit of response at 340 mμ. and was increasingly sensitive for the shorter wavelengths down to 254 mμ. This was used in conjunction with a d.c. amplifier bridge of conventional type. The p.e. cell and amplifier were calibrated by means of a standard of u.v. radiation. Four calibrated filters were used with this equipment to measure the u.v. energy in various spectral regions in the radiation from the “Osira” mercury discharge lamp and from the sun.
The chief disadvantage of d.c. amplifier circuits is their lack of stability. These difficulties have been minimized in bridge circuits in which two similar matched valves are used, so that the various drifts tend to compensate each other. Such improved bridge circuits have recently been described by Coblentz and Stair (1934), Stair (1939), and by Brentano and Ingleby (1939). The circuit shown in Fig. 5 is based on these. The two valves are type 32 screen-grid ones selected because of their high amplification factor, high impedance and low power consumption. They form two arms of a Wheatstone net, whose other arms consist of the 100,000 ohm resistors R1 and R2, one of which (R1) is adjustable. In general, the value of the grid leak R was 120 megohms, but smaller values 70, 20 and 10 megohms could be used to extend the range of currents measured. With the titanium p.e. cell (P.C.) it was found that a grid bias of −1·5 volts for each valve and a 120 megohm grid resistance S on the balancing valve greatly improved the stability and sensitivity of the amplifier. The stability is also improved by the common high potential supply to plates and screen-grids of the valves and the method of supplying the two filament currents from the same 2-volt accumulator. The detector (G) for measuring the out-of-balance current was either a
microammeter or a portable m.c. galvanometer of medium sensitivity. Grid control was provided by four potentiometers, R3 and R4 of 200 ohms, and R5 and R6 of 15,000 ohms. A Weston millivoltmeter (V) of 133 ohms per volt measures the voltage tapped off by R4. When no radiation falls on the p.e. cell, R4 is adjusted until zero reading is obtained on the voltmeter, then the bridge is balanced by means of R3 and R5. The response of the p.e. cell to radiation incident on its cathode can be measured either by reading the out-of-balance current on G, or else by reading the voltmeter after adjusting R4 to bring the circuit back to its initial state of balance as indicated by G. The latter method was adopted. The permanent grid bias ensured that the valves operated on the straight, steep part of their plate-current/grid-potential characteristic, so that the change of grid potential is proportional to the current through G. The potentiometer R6 has a scale affixed so that it can be set in a position to give suitable voltmeter readings. In order that these readings, which are proportional to the radiation-flux density, can all be reduced to a common basis corresponding to R6 set at scale number 40, the scale of R6 was calibrated.
Various tests were made as follows of the reliability and stability of the p.e. cell and bridge arrangement, and the results were found to be in agreement to an accuracy of about 1 per cent. (1) The p.e. currents were shown to be proportional to the value used for the grid leak (R). (2) The value of the incident radiation was shown to be proportional to readings of the voltmeter by changing the intensity of the radiation in a known ratio (a) by means of two lamps, and (b) by application of the inverse square law.
A type WL767 titanium p.e. cell was supplied by the Westing-house Electric and Manufacturing Company of New Jersey and its relative spectral response for equal energy in the region 250 to 340 mμ. was determined at the Bureau of Standards. These data were obtained by using a quartz mercury are lamp, a quartz cadmium are lamp, a quartz prism spectrometer and a vacuum thermopile, and were checked by using calibrated filters. The relative p.e. response for equal energy of this cell is given in Table X. The asterisked values are the standard ones; additional values were obtained by interpolation from the response curve. The values are relative, having been set at 15·0 for wavelength 313·2 mμ.
|Wavelength in mμ.||Relative Response.||Wavelength in mμ.||Relative Response.||Wavelength in mμ.||Relative Response.|
It is clear that the relative response of the titanium p.e. cell makes it especially suitable for measurements of u.v. energy in the region shorter than 315 mμ. and so for solar radiation measurements, in which case energy in the band 290–313 mμ. is most important. Currents generated in the cell during u.v. radiation measurements were of the order 10−9 amp., and these were amplified by a factor of about 20,000.
The p.e. cell was firmly mounted in a cylindrical metal can provided with a tube directly opposite the cathode of the cell and diaphragms which admitted a cone of rays of semi-angle 18° 26′ [= tan−1(⅓)] to the cathode. A shutter was arranged inside the can to eliminate incident radiation when required. The inside surfaces of can and tube, and the diaphragms, shutter, and filter holders were blackened with dull Dulux black. The tube supported a convenient form of filter holder. A mark was made on the can which exactly fitted the shadow of the filter holder when the optical axis of the p.e. cell was directed towards the sun. It was found that the grid lead to the cathode of the p.e. cell required special attention. A suitable form of lead was made by pulling good quality rubber
insulated wire through a length of rubber pressure tubing and enclosing this in a tight-fitting flexible conduit, which was connected to the can and to earth.
The filters used in this work are the four recommended by Coblentz and Stair (1936) for the evaluation of u.v. solar radiation. The spectral u.v. transmissions of the filters are given in Table XI. They were obtained radiometrically at the Bureau of Standards by using a quartz mercury are lamp, a quartz prism spectrometer and a vacuum thermopile.
From these data smooth curves were carefully drawn which were used to read off the transmissions at other wavelengths in this region.
|Type of Filter.||Thickness in mms.||Transmission per cent.|
2. Calibration of Photoelectric Amplifier Bridge.
In 1939 a standard mercury quartz are lamp (Hanovia Wehnelt electrode type 16100) was made available by Dr. E. R. Cooper, and with it calibration measurements were made on the p.e. amplifier bridge and on the balanced thermopiles. These measurements, which are reported herewith, illustrate the methods employed and the checks obtained, but, unfortunately, the operating conditions under which the standard lamp was calibrated by the Bureau of Standards were not reproduced. Therefore the absolute values recorded cannot be relied on until further tests are made on the lamp, or another standard lamp which has been calibrated on a 50 cycle a.c. supply is available. The calibration specified the total radiant-flux density for wavelengths 313·2 mμ. and shorter, and the relative intensities of the lines from 436 to 230 mμ. (See Table XII.) The supply was 60 cycle a.c., the 220-volt tap was used on the transformer and the primary voltage was 220·7. The lamp current was not specified, but was probably close to 2 amps. The conditions under which the lamp was operated in our work were 50 cycle a.c., 230-volt tap on transformer, primary voltage 220, primary current 4·05 amps., lamp current 2·80 amps. The lamp was allowed 30 minutes to attain equilibrium temperature before readings were made.
|Wavelength in mμ||436||405||365||334||313||302||297||289||280||275|
|Wavelength in mμ||270||265||258||254||248||240||238||236||230|
The absolute energy of wavelengths 313·2 mμ. and less at 61 cms. distant emitted by the standard lamp was measured by the balanced thermopile method. The filter factor for the barium-flint filter was found to be 1·21 by the method described in Part I, and a re-determination of the mean thermopile sensitivity gave the result 2·02
microwatts per cm.2 per μV., in close agreement with the value obtained a year earlier. The mean e.m.f. generated by the u.v. radiation excluded by the barium-flint filter, after correction with the Noviol A filter, was 41·6 μV. Hence the radiant-flux density at 61·0 cms. from the quartz envelope of the standard lamp = 41·6 × 1·21 × 2·02 = 101·7 microwatts per cm.2 in the wavelength region 313·2 mμ. and less. This figure was checked by repeating the measurements.
The procedure adopted in calibrating the p.e. amplifier bridge with the standard lamp was to eliminate wavelengths less than 275 mμ. by means of a calibrated Corex D filter placed over the p.e. cell. This reduced the uncertainty involved due to the presence of short waves with their high p.e. response and roughly limited the wavelength band to that required in solar radiation measurements. Then from the spectral energy distribution of the source and the transmission data of the Corex D filter, a conversion factor was obtained which enabled the absolute u.v. energy of wavelengths 313·2 mμ. and less falling on the cell to be calculated from the corresponding voltmeter readings. The data for the calculation of the conversion factor (P) are shown in Table XIII.
|Wavelength in mμ.||Relative Energy of Source.||Transmission of Corex D.||Relative Energy through Cx. D.||Absolute Energy through Cx. D.||Relative Response of p.e. cell.||Effect on p.e. cell.|
Column c is the product of columns a and b. Column d is obtained by multiplying column c by the factor 101·7/104·5, since 101·7 microwatts per cm.2 is the absolute radiant-flux density for wavelengths 313·2 mμ. and less received by the thermopile, while from the emission of the standard lamp (Table XII) the relative units of radiant flux density in this region are 104·5. Column f is the product of columns d and e. The total of column f (1058·3) represents the energy flux as weighted (or appraised) by the p.e. cell when the output from the 5·2 cms. of mercury are exposed by the opening in the shield of the standard lamp is incident on it. But the effective are length was reduced by a diaphragm to 1·95 cms., which is approximately equal to the diameter of the aperture to the p.e. cell, so the actual energy flux weighted by the p.e. cell when 61 cms. from the lamp = (1·95 × 1058·3)/5·2 = 396·9. Under these conditions the voltmeter read ·714 volts when R6 was set at scale number 40, hence the conversion factor = 396·9/.714 = 556. Hence, weighted radiant-flux density = 556 × voltmeter reading (reduced to scale number 40). The absolute radiant-flux density then follows by multiplying the above by another factor (G), whose evaluation is discussed in the next section.
3. Tests with the “Ositra” Lamp.
For these tests the outer glass globe was removed from the “Osira” lamp exposing the small quartz discharge tube. The lamp current was 1·26 amps. The radiant-flux density at 61 cms. from this source in a horizontal direction perpendicular to the vertical plane containing the two projecting side tubes was measured (a) by the p.e. cell method and (b) by the balanced thermopile method. Both these measurements refer to the wavelength region 313·2 mμ. and less, so agreement between these independent methods would check their reliability.
(a) To obtain the factor (G) for conversion of “weighted” to absolute units of radiant-flux density a knowledge of the approximate spectral energy distribution of the “Osira” source in the region 250–340 mμ. is necessary. On account of its close similarity to the Uviare 150 watt mercury are referred to by Johnson and Webster (1938) the spectral energy distribution for the “Osira” source has been taken as that for this Uviare lamp. The calculation of the factor G is given in Table XIV.
|Wavelength in mμ.||Relative Energy.||Relative Energy through Corex D.||Response of p.e. cell.||Column a x||Column b x|
|a||b||c||Column c.||Column c.|
Using no Corex D filter over the p.e. cell, value of G = 100·9/8377 = ·0120, and using the Corex D filter value of G = 42·1/1250 = ·0337. In one of the tests using the Corex D filter the voltmeter reading was 2·06 volts when reduced to scale number 40, hence the radiantflux density = 2·06 × 556 × ·0337 = 38·6 microwatts per cm.2 Repetitions yielded values 37·1, 38·8, and 39·1, giving a mean value by the p.e. cell method of 38·4.
(b) In the balanced thermopile method it was necessary to calculate the filter factor for the “Osira” lamp for wavelengths 313·2 mμ. and less, and also to measure the proportion of u.v. energy in this region which is transmitted by the Corex D filter. The mean value of three determinations by this method was 39·1 microwatts per cm.2, which is in good agreement with the mean value by the previous method.
4. Measurements of Solar Ultraviolet Radiation.
The same general procedure was adopted as that described by Coblentz and Stair (1936). Observations were made in September, October, November, December, and March, but satisfactory readings were only obtained on days when the sun was not obscured by cloud or thick haze. A simple device was constructed and calibrated which enabled the air mass to be read quickly and directly. The air mass is proportional to the length of the path of the solar radiation through the atmosphere, and its value = cosec A, where A is the altitude of the sun. Hence the air mass = 1·00 when the sun is directly overhead regardless of the elevation of the station. Each of the filters, clean and dust-free, was placed in turn in the filter holder, and voltmeter readings were taken when balance on the bridge had been restored. These readings were reduced to the value corresponding to scale number 40, and were then proportional to the response of the p.e. cell to the u.v. radiation transmitted by the filter. The order of readings was arranged to reduce as far as possible any effect of drift in the zero of the amplifier bridge or of progressive change in the solar radiation during the six minute interval required to complete a set of readings. A typical set of readings is shown in Table XV, and also the percentage transmissions through the filters deduced from these readings. Before and after taking a set of readings the shutter was closed and the bridge balanced, and this precaution was occasionally taken during a set.
|No Filter.||Corex D.||Nillite.||Ba.—Flint 1.||Ba.—Flint 3.||No Filter.|
The results of the observations taken during a typical clear day (5/10/39) are shown in Table XVI. The times are noted in New Zealand Summer Time, which is 30 minutes in advance of New Zealand Standard Time. The observed filter transmissions are stated as percentages of the total solar radiation less than 340 mμ., which is the maximum wavelength to which the p.e. cell responds. On clear days the voltmeter readings, which are proportional to the total u.v. radiation, reached a maximum when the air mass had its minimum value.
|Time.||Air Mass.||Volts.||Corex D.||Nillite.||Ba.—F. 1.||Ba.—F. 3.||Conditions.|
|11.50 a.m.||1.36||1.551||77.7||56.0||29.4||8.2||Light cloud|
|12 noon||1.35||1.550||77.6||59.3||30.4||9.1||Slight haze|
|12.30 p.m.||1.32||1.517||78.4||58.1||29.6||8.4||More haze|
In order to obtain in absolute units the radiant-flux density of the u.v. solar radiation in various regions, it is necessary to calculate the factor G. This involves, in addition to the transmissions of the four filters and the spectral response of the p.e. cell, an approximate knowledge of the spectral energy distribution of the solar radiation for a given air mass in the region 290–340 mμ. A solar curve is used showing relative intensity plotted against wavelength based on results obtained by previous workers (preferably for an equivalent air mass). If calculation shows that this curve yields about the same transmission values for the four filters as those observed, then it is identified as the solar energy distribution curve (Curve 1). To obtain factor G a second curve is constructed which shows the response of the p.e. cell, when the incident energy is distributed as in Curve 1, plotted against wavelength. By simple graphical integration or by use of a planimeter the area under Curve 1 up to wavelength 313·2 mμ. is measured, and also the area under Curve 2. Then factor G is the ratio of the first area to the second. Hence for a given air mass, the absolute energy in the u.v. region 290–313·2 mμ.=V × P × G microwatts per cm.2, where V is the voltmeter reading and P is the conversion factor (= 556, see Table XIII).
The above procedure was adopted. From the limited number of observations which have been taken over a comparatively short period, the observed transmissions have been calculated and representative values are plotted in Fig. 6 for air masses ranging from 1·08 to 2·10.
Considerable variations occur between individual points, which are only partially explained by seasonal change in the transparency of the atmosphere. The circles in Fig. 6 show the calculated transmissions at three air masses (or solar altitudes) and through them
straight lines are drawn with which to compare the distribution of the observed values. Many factors conspired against the consistency of the observed filter transmissions; for example, unsteadiness in atmospheric conditions due to cloud, haze, smoke, or wind, and unsteadiness in the measuring circuit produced by small disturbances of the long shielded grid lead. This lead from the cathode of the p.e. cell to the grid of the amplifier valve had to be several yards long in order that the cell could be exposed to the sun on the roof, while the measurements were made in the room below. The calculated transmissions at the three air masses were obtained from solar energy distribution curves which were finally selected after many trials. These curves are based on those adopted by Coblentz and Stair (p. 323, 1936) for Washington, air masses 2·35, 2·00, and 1·60 at Washington were found to correspond to 2·00, 1·65, and 1·25 respectively at Dunedin. Not only are the equivalent air masses at Washington greater than those at Dunedin (as was expected because of the greater latitude of Dunedin and consequent difference in atmospheric ozone absorption), but in addition the Washington curves required some adjustments which were made by a lengthy process of trial and error. Neither the precision attained in the observed filter transmissions, nor the agreement between the calculated and observed values are at present as good as has been achieved by more experienced investigators overseas, e.g. Coblentz and Stair (1935 and 1936), but additional measurements are contemplated using improved apparatus. Doubtless further adjustments to the adopted solar curves would improve the agreement, but this is not warranted until the precision of the observations is increased.
|Wavelength in mμ.||Relative Solar Energy.||Sensitivity of p.e. cell.||Response a × b.||Corex D.||c × d.||Nillite.||c × f.||Ba.—F.1.||c × h.||Ba.—F.3.||c × k|
The method of calculating the filter transmissions and the factor G is illustrated in Table XVII. These results are for air mass 1·65 at Dunedin, and the relative solar energy curve (column a) is that adopted by Coblentz and Stair for air mass 2·00 as Washington, with slight adjustments. Column c gives the relative response of the titanium p.e. cell. Under each of the four filters is listed its percentage transmission at the wavelength intervals differing by 2·5 mμ. The total noted under column a is for wavelengths less than 313·2 mμ. The values obtained by the method of calculation illustrated (which were performed on a slide rule) were checked by drawing curves and using a planimeter.
Hence factor G = 2·37/152·2 = ·0156. The percentage filter transmissions are: Corex D = (118·9 × 100)/152·2 = 78·0; Nillite = 59·2; Barium-flint 1 = 30·0; Barium-flint 3 = 10·2. In a similar manner the corresponding values were calculated for air masses 1·25 and 2·00. These results are summarized in Table XVIII.
|Transmission per cent.|
|Air Mass.||Factor G.||Corex D.||Nillite.||Ba.—Flint 1.||Ba.—Flint 3.|
Although the results reported are approximate and the absolute values may require some correction, it is of interest to estimate the intensity of the u.v. solar radiation of wavelengths shorter than 313·2 mμ. in the clear, mid-day sunlight at Dunedin and compare it with that obtained at Washington by Coblentz and Stair (1936). The maximum values during a clear day at Washington (latitude 39° N.) ranged from about 8 microwatts per cm.2 in midwinter to 75 in midsummer. The tentative values for Dunedin (latitude 46° S.). calculated from the observations by finding the product V × P × G (as previously explained) are: September, 14; October, 20; November, 26; December, 32; March, 16 microwatts per cm.2. No observations were made in late December, January, or February, so the maximum value in Dunedin could not be estimated, but it is unlikely to exceed 50 microwatts per cm.2, which is much smaller than at Washington. It was noted that the u.v. intensity in March exceeded that in September for a corresponding air mass. This increased atmospheric transparency in the autumn has been noted by other investigators. Our results (for example, see Table XVI) also confirmed the deduction previously made by Coblentz and Stair (1935), that while a little haze greatly reduces the intensity of solar u.v. radiation received at sea-level, its spectral quality is practically unchanged. This is ascribed to the relatively non-selective character of the transmission through water vapour.
The amplifier bridge has operated satisfactorily in its present form, but in order that u.v. measurements can be made in various localities, and the effects of local meteorological conditions can be examined, it should be constructed in a more convenient and portable form. This would be facilitated by employing high impedance valves requiring 1·5 volts filament supply in place of the two screen-grid valves now in use. All the necessary voltages could then be obtained from dry batteries. A somewhat greater sensitivity would also be an advantage, since the reading with the thicker barium-flint filter is always small with the present apparatus. Steadiness in operation of valves and resistors is essential. Good shielding, and a short, specially-constructed grid lead are also important.
The agreement between the results of the measurement of u.v. radiation from the “Osira” lamp by the balanced thermopile and p.e. cell methods is reasonably close in view of the uncertainties involved particularly in the assumed spectral energy distribution of the lamp.
The reliability of the solar u.v. measurements made by the p.e. cell method would be greatly improved if they had been confirmed by the independent balanced thermopile method. This has not been attempted yet, because in using the thermopiles out of doors it is necessary to protect the elements with a thin window transparent to u.v. radiation as short as 290 mμ., and the transmission curve for the window must be known. At present we have available no sufficiently accurate method for determining the transmission at a series of wavelengths in the u.v. region, but it may be possible to use for the windows a transparent material (like crystalline quartz), whose transmission curve is known.
Great reliability and accuracy is not claimed for the measurements of solar u.v. radiation. They are the results of a first attempt at what appears to be a difficult physical measurement, judging by the meagreness and lack of consistency in the results which have been reported by previous workers. A greater number of observations over a longer period will increase the reliability of the observed filter transmissions, and this in turn will enable the solar energy curves to be identified more precisely. In the present work most difficulty has been experienced in obtaining agreement between the observed and calculated transmission of the barium-flint 3 filter. This figure carries less weight than those for the other filters, first because the p.e. readings with this filter are the smallest and least consistent, and secondly because the transmission curve for this filter was plotted from data at only three points in the required region.
The amount of solar u.v. radiation which actually reaches the earth's surface is determined principally by the amount absorbed by the ozone in the stratosphere, and only to a minor extent by absorption in the lower atmosphere due to cloud, fog, haze, and atmospheric pollution. Nevertheless the variability of these latter factors has a profound effect on the intensity. Reliable information regarding the amount of u.v. energy received at a given locality entails a long series of systematic observations for which the p.e. cell method provides the most suitable means. The weather conditions at Dunedin
are not particularly favourable for such observations, but from the results obtained the following deductions were tentatively drawn: (a) the type of annual and diurnal variation in u.v. radiation; (b) the absolute value of the u.v. radiation below 314 mμ. for certain months, and an increased atmospheric transparency in the autumn as compared with the spring; (c) for all filters except the Corex D there is a distinct decrease in transmission with decrease in air mass (see Fig. 6). This indicates that there is an increase in the u.v. energy present in the region below 320 mμ as the air mass decreases; (d) the transmission of the Corex D filter remains sensibly constant for different air masses, which shows there is very little energy present of wavelengths in the region of the cut-off of this filter, i.e. 290–300 mμ. This indicates that during the spring months at least the limit to the solar u.v. radiation is about 300 mμ., which leaves a narrow band of only about 13 mμ. to produce the beneficial therapeutic effects; (e) the increase in filter transmissions on certain clear days compared with the values on neighbouring clear days indicates a decrease in u.v. energy of wavelengths less than 320 mμ., which can be accounted for by an increase in the ozone present in the upper atmosphere.
It is a pleasure to acknowledge our great debt to Dr. W. W. Coblentz, Head of the Radiometry Division, Bureau of Standards, Washington, D.C., for his valuable advice and assistance.
The writers wish to thank Professor R. Jack, University of Otago, in whose laboratory the work was carried out; Dr. E. R. Cooper, Office-in-Charge, Government Physical Testing Laboratory, for the loan of a standard mercury vapour lamp; and Mr. S. O. Hughes for assistance with the photographic and some other parts of the work.
Brentano, J. C. M., and Ingleby, P., 1939. J. Sci. Insts. 16, 3, 81.
Brock, G. C., 1933. Sci. Proo. Roy. Dublin Soc. 20, 563.
Coblentz, W. W., Stair, R., and Hogue, J. M., 1931. Bur. Stds. J. of Res. 7, 724.
— 1932. Bur. Stds. J. of Res. 8, 759.
Cobientz, W. W., and Stair, R., 1934. Bur. Stds. J. of Res. 12, 231.
— 1935. Bur. Stds. J. of Res. 15, 123.
— 1936. Bur. Stds. J. of Res. 16, 315.
Coblentz, W. W., 1935. Puerto Rico J. Public Health and Tropical Medicine 11, 1.
Johnson, L. B., and Webster, S. B., 1938. Rev. Sci. Insts. 9, 325.
Maddock, A. J., 1940. J. Sci. Insts. 17, 4, 89.
Poole, H. H., and Atkins, W. R. G., 1935. Philos. Trans. A. 235, 1.
Staib, R., 1939. Bur. Stds. J. of Res. 22, 295.
Withrow, R. B., 1931. Bull. Basic Sci. Res. May.