416 ROBINSON AND OVENSTON : A SIMPLE FLAME PHOTOMETER [Vol. 76 A Simple Flame Photometer for Internal-Standard Operation and Notes on. Some New Liquid Spectrum Filters BY (THE LATE) A. M. ROBINSON AND T. C. J. OVENSTON (Presented at the meeting of the Physical Methods Group on Friday, October 6th, 1950) A flame photometer based on a Dutch design has been built for operation primarily by the internal-standard technique. Two beams, directed simul- taneously through “standard” and “sample” spectrum filters respectively, fall on barrier-layer photo-cells whose outputs are balanced potentiometrically by a null-point method. Special features in the design and operation of this instrument are described, and certain precautions are recommended. The choice of the internal-standard element is briefly discussed.The transmission characteristics of some new spectrum filters, made by combining layers of solutions of common substances, are given and in some instances compared with those of glass, gelatin and interference filters. FROM work already published it seems quite clear not only that sodium, potassium and related elements can be determined much more rapidly by means of a flame photometer than by the time-consuming gravimetric procedures, but also that the results so obtained for many materials are quite as reliable. It appears that useful results can be obtained with very simple apparatus in which the excited raldiation, after passage through a spectrum filter, falls on a selenium barrier-layer cell connected to a galvanometer. It was desired to explore possible applications of a simple flame photometer of this type, and one described by Boon1 in 1945 was taken as a model.Like that of Barnes, Richardson, Berry and Hood,, Boon’s instrument was designed for direct reading. From the subsequent work of Berry, Chappell and Barnes3 it appeared that the use of an internal-standard technique arid a dual optical system offered certain important advantages, so the design was modified to allow either method of measurement to be used. DESCRIPTION OF THE FLAME PHOTOMETER Fig. 1 is a general view of the flame photometer assembly. Cylinder gas is controlled by reducing valves to give a low-pressure supply, pressures being measured by a water manometer. Air from a small compressor is fed through a pressure stabiliser to the atomiser, final pressure being shown by a mercury manometer.The final pressure controls for both gas and air are glass taps that are mechanically connected to slow-motion dials of a type once common on radio sets. Attention is drawn to the cylindrical light shields, which are here shown slid back to reveal the optical assembly. In use these shields are slid forward against the vertical plates on the side limbs of the chimney so that all background radiation other than that from the flame itself is eliminated. The disposition of the principal parts and the electrical circuit used for the internal- standard method are shown in Fig. 2. The distances hetween the photo-cells, the lenses and the flame have been chosen as a satisfactory compromise between two opposing effects; greater distances tend to lower the over-all sensitivity, whereas the heat from the flame makes it undesirable to place the photo-cells any nearer. As it is, it is very important to keep the photo-cell shutters up except for the minimum time necessary to make a measurement.The electrical circuit, except for a minor modification, is that given by Berry, Chappell and Barnes.3 The filter F’ transmits only the radiation emitted by the internal-standard element, which has been included in known concentration in the sample solution, and the photo-cell C’ receiving this radiation generates a proportional amount of current, which is dispersed along resistance R,. At the same time the filter F transmits only the radiation emitted by the element being determined and the photo-cell C generates a proportional amount of current, which is dispersed along resistance R,.1 .( ;enera1 \.iew of the flame photonieter assembly SAMPLE BEAM .c-- R,= 500n t 1;ig. 2 . Optical and electrical layout of flame photometer for use with internal standardsFig. 3. The burner assembly Fig. 1. The atomising vesselJuly, 19511 AND NOTES OX LIQUID SPECTRUM FILTERS 41 7 The circuits are balanced by means of a galvanometer, used as a null-point meter, connected to tappings on R, and R,. A Cambridge spot galvanometer with a resistance of 450 ohms was found suitable. Normally, the tapping on R, is at the negative end of the resistance so that the whole of the current generated as a result of the emission of the sample element is balanced against the excess of internal-standard current, the range of the determina- tion depending on the latter.Range adjustments can sometimes be conveniently made, however, by moving the galvanometer tapping along R,, and the exact position can be found again on a subsequent occasion by resetting with the standard solutions. Once set for a given determination, this tapping must not be altered. Resistance R, is a precision potentio- meter, the galvanometer tapping on which is adjusted by means of the control knob to give no deflection of the galvanometer needle during measurement. The pointer attached to the control knob indicates in degrees the position of the tapping; this value can be related to known concentrations of a given element and hence a calibration graph can be constructed. The small resistance R, is useful for making minor adjustments in the course of a long series of determinations when the occasional introduction of standard solutions indicates that a slight drift has occurred.Cylinder gas is passed by way of the lower nozzle into the glass burner, in the body of which it is mixed with air containing the atomised solution passed in by way of the side tube. The base of the burner is pressed against a rubber washer by means of a spring held by an adjustable cross-bar. A short glass tube inserted in the rubber washer is placed over the gas inlet to protect it from the mist. The tip of the burner is a platinum collar 6 mm in diameter. The atomiser, which is based on the design of Rauterberg and Kni~penberg,~ is shown in Fig.4. Air is passed in at the bottom under pressure and is forced out of the vertical jet. The horizontal tube leading from the beaker containing the solution is so placed that the air jet, in passing the tip, causes the solution to be sucked into the atomiser, where it is blown against the baffle plate with considerable force. The only outlet for the air is through the side tube to the burner, so that the mist formed in this way is carried along with it. The un-atomised solution falls to the bottom of the vessel, where it syphons to waste. A con- siderable amount of adjustment is possible by rotation of the jets, both of which are carried on ground-glass joints. Both atomiser and burner are now obtainable commercially.* They have recently been fully described by Doming0 and K l ~ n e , ~ who have adapted Boon’s design to the deter- mination of sodium with a selenium photo-cell and of potassium with a caesium photo-cell by the direct method of measurement.The gas pressure required depends on the size of the nozzle leading into the base of the glass burner. The nozzle size is not critical, but the correct manometer reading for any given nozzle is constant and can be found by reducing the gas flame until it is as non-luminous as possible while at the same time maintaining the flame in a stable condition. The air pressure required depends on the size of the air jet in the atomiser, sufficient air having to be passed to maintain a steady flame. With the type of atomiser described here a pressure of about 40mm of mercury is usual, the exact value for a given model being found by experiment.The burner assembly is shown in Fig. 3. PRECAUTIONS IN OPEFLATION- To prevent minor explosions when lighting and extinguishing the burner for use with acetylene - air mixtures, a Y-tube was inserted in the acetylene lead just before it enters the burner. The free end of the Y-tube was permanently connected to a coal-gas supply tap. The coal gas was first turned on and lit. The air supply was then turned on and finally the acetylene. The coal gas was then turned off slowly, care being taken that sufficient acetylene was being supplied to support the flame. Final adjustments were then made in the usual way. When extinguishing the flame, the coal gas was first turned on and the acetylene then turned off.The coal gas was then turned off. This soot must be carefully removed before relighting, otherwise the flame may be so unsteady as to be uncontrollable. Even in normal usage of the burner, soot is slowly deposited near the tip; to ensure the best results, this deposit should frequently be removed. Lane End Road, Sands, High Wycombe, Bucks. Should a back-fire occur at any time, the burner quickly fills with soot. * The apparatus used by the authors was supplied by the Laboratory Glassblowers Co., Valley Works,418 cells to the flame. keep the shutters closed except for the half minute required for each measurement. in the bottom of the atomiser vessel was syphorting away. RELATIVE FLAME SENSITIVITIES- Barnes, Richardson, Berry and Hood2 preferred a flame with a relatively low temperature because less elements were excited and this made the task of isolating the required radiation simpler.Much sensitivity is lost in this way, however, and many workers prefer the hotter acetylene - air flame. In Fig. 5 are shown the relative sensitivities of acetylene - air, coal gas - air and butane - air flames as indicated by the galvanometer deflection given by various concentrations of sodium. It is clear that excitation in the acetylene - air flame is much greater than in the other two flames and that it should be used whenever sensitivity is of importance. SELENIUM CELL RESPONSE- barrier-layer photo-cells were employed. ROBINSON AND OVENSTON : A SIMPLE FLAME PHOTOMETER [Vol. 76 Another precaution that must be emphasisecl concerns the close proximity of the photo- It was found essential in order to prevent fatigue of the photo-cells to Finally, it was found desirable to avoid taking readings at the moments when the waste In the present apparatus standard Evans Electroselenium Ltd.(EEL) selenium These show good response over the whole of the Sodium, p.p.m. Curve A, acetylene - air; curve B, coal gas - air; curve C , Fig. 5. Relative sensitivities of burner gases butane - air visible spectrum and are particularly sensitive in the region near 589mp, the wavelength of the sodium doublet. Above 700mp, however, the response falls off sharply and small amounts of potassium cannot be determined. For instance, in a direct measurement circuit and with the liquid spectrum filters now recommended, a galvanometer deflection (one-tenth of the full scale) that was given by as little as 0-5 p.p.m.of sodium or 3.2 p.p.m. of lithium required 30 p.p.m. of potassium to reach the same value. For this reason it is usual to measure the potassium emission by means of a caesium photo-emissive cell. For internal-standard work, with two photo-cells in electrical balance, it is desirable to select a type of cell that covers the whole of the working range. Except for work on potassium (and on caesium and rubidium), the selenium cell described above is satisfactory and offers the advantage of cheapness and simplicity. To widen the range it has been the practice, as in the Perkin - Elmer model 52A, to use interchangeable red- and blue-sensitive photo-emissive cells in conjunction with a balanced electronic circuit and high amplification ; this, of course, involves considerable expense, and such an apparatus, combined as it is with a dual-prism monochromating.system, is possibly as, versatile an instrument as can be devised. Since the main body of this paper was written, however, attention has been drawn* to theJuly, 19511 AND NOTES Oh- LIQUID SPECTRUM FILTERS 419 recent production of selenium barrier-layer cells* having a response range extending to about 900mp. With these cells it is hoped to extend the range of the present simple assembly to cover all normal requirements. U S E OF INTERNAL STANDARDS ERRORS AND THE INTERNAL-STANDARD TECHNIQUE- Errors in flame photometry can be classified into: (a) those arising from flame fluctuations; (b) those arising from impurities affecting the viscosity or surface tension of the solution, i.e., urea, which depresses, and methanol, which enhances the emission intensity of the alkali metals3; (c) those arising from the depressing effect of acids and salts3,’; (d) those arising from inefficiency of the filters.In the direct (or absolute) method of measurement, flame fluctuations are controlled as much as possible, and it is good practice to alternate samples with standards to detect drift. To cope with the effects of impurities and of large concentrations of salt, a number of suggestions have been made. For example, Berry, Chappell and Barnes3 have recommended the making up of standards in solutions having the same composition as that containing the element to be determined.This technique is obviously limited to ranges of materials of fairly constant general composition. Shapiro and Hoagland* recommend further dilution of samples, by which means the relative effect of these interferences can be very greatly reduced. This dilution technique is certainly effective, but the sensitivity of the determination is, of course, correspondingly reduced. For water analysis, West, Folse and Montgomeryg have suggested the use of “radiation buffers,” that is, -concentrated solutions of selected salts added in fixed proportions to the sample solution to buffer any interference from small and variable amounts of these salts. By means of the internal-standard technique with a dual optical system, errors caused by flame fluctuations, by viscosity and surface tension effects and by the depressing effects of acids and salts are greatly reduced.In addition, any of the suggestions already made to cope with these effects in the direct method of measurement are equally applicable when using the internal-standard technique. Both methods are dependent on the efficiency of the light filters in cutting off unwanted radiation, but it should be mentioned that it is not necessary for the internal-standard filter to eliminate aZZ the radiation of the element being determined, or for the sample filter to eliminate all the radiation of the element used as the internal standard, as the calibration graph will take account of this. It can be concluded that the internal-standard technique can be used with advantage whenever it is appropriate.It is not appropriate unless an internal-standard element can be selected that not only provides a sufficiently strong emission at a convenient wavelength, but also does not occur in appreciable and variable quantities in the material to be analysed. POSSIBLE INTERNAL-STANDARD ELEMENTS- Until recently, lithium was the only metal that had been successfully used as an internal standard for flame photometry (excluding spectrographic applications). I t is known to be absent from many materials commonly subjected to analysis for sodium and potassium. Eubank and Bogue,lo however, prefer the direct method for the analysis of Portland cement because this material contains unknown and variable amounts of lithium.In addition to examining lithium, Berry, Chappell and Barnes,3 who used a type of photo-cell with a response similar to that of the standard Evans Electroselenium cell, examined the potentialities of rubidium, caesium and indium, but rejected them. The present authors have tried to use the weak emissions available in the blue end of the spectra of rubidium and caesium. Suitable filters were devised, but the concentrations of internal standard required to balance the common alkali metals were far too large to allow the develop- ment of practical or economical methods. However, these two metals are sufficiently rare in nature to make them valuable internal-standard elements if it were possible to utilise the radiation of their strong emissions in the near infra-red.It is hoped to give further study to this matter when the new infra-red-sensitive barrier-layer cells have been incorporated in the present apparatus. * These cells are manufactured by Megatron, Ltd., London.420 ROBINSON AND OVENSTON : -4 SIMPLE FLAME PHOTOMETER [Vol. 76 This gives a fairly strong emission a t 535 mp, which can be separated completely from the sodium doublet by means of the solution filter whose characteristics are shown in Fig. 9. No further experiments with thallium have yet been made. I t should be noted that this metal is toxic and should not be burnt in the flame photometer without an efficient flue. Finally, it should be mentioned that common elements can sometimes be used as internal standards. One of the authors (A.M. R.) has developed a method for the determination of lithium as a major constituent in magnesium - lithium alloys by using potassium as an internal standard. Sodium and other normal impurities or additions do not interfere and the amount of potassium added as standard is so much in excess of the maximum that could reasonably be expected to be present in these alloys that their true potassium content is of no account. The weak photo-cell response at 766 to 770 mp is then an advantage; in fact, with an infra- red-sensitive photo-cell it would be necessary to reduce the intensity of the standard beam by means of an iris diaphragm or similar device. This method will be described in a separate paper. FILTERS Another element worth mention is thallium. For filtering the radiation, glass or gelatin filters are undoubtedly very convenient to In addition, gelatin filters are apt to change use, but they are not particularly efficient.Wavelength, mp Fig. 6. . . . . . . . . .-._._ Cupric chloride component - - _ _ _ Transmission curves for sodium filters Complete liquid filter Sodium dichromate component Chance OY1 and Ilford 803 combined their transmission characteristics with use. It is considered preferable to use cells filled with solutions of known transmission characteristics. These are readily made up and can always be relied on. With two layers of solutions it is possible to devise combinations that completely stop all radiations outside a limited range of the spectrum. For example, potassium dichromate has a very sharp cut-off that can be varied over about 50 mp by changes of concentration. By using sodium dichromate, higher concentrat ions are possible and radiation shorter than that emitted by sodium can be readily cut off.The transmission curve of a l-cm layer of an aqueous 50 per cent. w/v solution of sodium dichromate dihydrate (Na,Cr,O,.ZH,O) is shown in Fig. 6, together with the corresponding curve for a 5.0 per cent. w/v solution of cupric chloride dihydrate (CuC1,.2H20) in 8 N hydrochloric acid. By combining these two layers in series they form an efficient filter for sodium radiation with transmission characteristics indicated by the continuous curve in Fig. 6. This filter cuts off all radiation in the photo-cell range of wavelengths from 553 mp downwards and from 671 mp upwards, and transmits 44 per cent.of the sodium radiation at 589 mp. Potassium, lithium (at 671 mp) and barium radiation are efficiently stopped. Calcium and strontium band emissions, which are most intense at about 620 mp and 605 mp, respectively, are partly transmitted. The weak lithium emission at 610mp is within theJuly, 19511 AND NOTES ON LIQUID SPECTRUM FILTERS 42 1 transmission range of this filter; fortunately, the effect of this is noticeable only when the concentration of lithium is high and, in any event, could not contribute an error when lithium is used as an internal standard, though it would increase the size of the blank measured when no sodium is present. The pecked curve in Fig. 6 shows the transmission spectrum of the combination of a Chance OY1 glass filter with an Ilford 803 gelatin filter, a system that has been recommended for the isolation of sodium radiation.The system transmits 18 per cent. of the sodium radiation and a correspondingly reduced proportion of the calcium and strontium radiation. Its use would, therefore, reduce the sensitivity for sodium to less than half of that obtained Fig. 7. . . . . . . . . ._._._ Potassium dichromate component _ _ _ - - Ilford 207 Transmission curves for potassium filters Complete liquid filter Aniline blue component (this coincides with that for the complete filter above 500mp) with the liquid filter combination, although the interference to be expected from calcium and strontium is no less. In addition, transmission down to about 490mp is appreciable, in which region the barium emission at 553 mp and various barium oxide bands at lower wavelengths occur.In Fig. 7 the transmission curves of a liquid filter for potassium and of Ilford 207 gelatin filter are compared. The liquid filter combination consists of 1-cm layers of a 2 per cent. w/v aqueous solution of potassium dichromate (K,Cr,O,) and a 0.02 per cent. w/v solution of aniline blue in 95 per cent. ethanol. The cut-off is due entirely to the aniline blue solution. The purpose of the dichromate layer is to prevent transmission below 500 mp, which would be permitted by a region of low absorption by the aniline blue solution. The concentration of the aniline blue solution has been selected so that all radiation a t 671 mp and below is stopped.At the same time, 96 per cent. of the potassium radiation at 766 mp and 770 mp is transmitted, as compared with 75 per cent. for the Ilford 207 filter. The Ilford 207 filter, however, is also very efficient in stopping unwanted radiation. Radiation from elements emitting at lower energy levels is cut off by the limiting response of the photo-cells in the infra-red. With photo-cells more sensitive to infra-red radiation it would be necessary to devise a filter to stop the lithium emission at 831 mp, the rubidium emissions at 780 mp and 795mp and the caesium emissions a t 852mp and 894mp, should these elements be present in appreciable quantities. The liquid filter combination for lithium, Fig. 8, consists of 1-cm layers of a 0-2 per cent. w/v solution of rhodamine B in water and a 1.0 per cent. w/v solution of cupric chloride dihydrate in 8 N hydrochloric acid.This combination has been designed to stop radiation at 620 mp and below and at 766 mp and above, and is thus highly efficient in isolating the lithium radiation, which is transmitted to the extent of 25 per cent. The liquid filter proposed for thallium, Fig. 9, consists of 1-cm layers of a 0.5 per cent. w/v aqueous solution of potassium dichromate and a 100 per cent. w/v solution of cupric nitrate trihydrate (Cu(N0,),.3H20) in 2 N nitric acid. The combination transmits 21 per[Vol. 76 cent. of the thallium radiation at 535 mp, while the presence of large amounts of sodium causes no appreciable increase in the total amount of transmitted radiation. Barium and barium oxide radiation and some of the calcium oxide radiation at 550 to 555 mp is partly t ransrnit t ed.The effect of increasing the optical thickness of a liquid filter is of interest. It is well known that an increase in the optical path of a homogeneous absorbing medium by a factor offwill have the effect of reducing the transmission by raising it to the power off. It follows that the efficiency of a liquid filter may be continuously increased by increasing its thickness, 422 ROBINSON AND OVENSTON : A SIMPLE FLAME PHOTOMETER Wavelength, mp Fig. 8. Transmission curves for the lithium filter ----- Rhodamine B component .-.-.- Cupric chloride component Complete liquid filter Wavelength, mp Fig. 9. Transmission curves fo:r the thallium filter Complete liquid filter ----- Potassium dichromate component ._._._ Cupric nitrate component for the smaller the transmission the greater will be the reduction in transmission.Hence, where a transmission of 50 per cent. would fall to 25 per ce:nt. on doubling the thickness of the filter, a transmission of 25 per cent. would fall to 6-25 per cent. and one of 10 per cent. to 1 per cent. In Fig. 10, curve A shows the transmission spectrum of the normal-thickness sodium filter already described and curve B shows that of the same filter made up of two 2-cm layers. The increase in efficiency is evident, for while the transmission of the sodium radiation has fallen froin 44 per cent. to 19.4 per cent., the transmission of the strontium radiation at 605 mp has been reduced more than threefold and that of the calcium radiation at 620 mp has beenJuly, 19511 AND NOTES ON LIQUID SPECTRUM FILTERS 423 reduced sevenfold. By continuing this process it is possible to obtain an almost perfect monochromatic filter.Unfortunately, with each reduction of maximum transmission the sensitivity of the determination is correspondingly decreased, and the problem resolves itself into choosing the most suitable compromise to meet any particular set of requirements. For comparison, the transmission spectrum of a typical interference filter of maximum transmission at 589 mp is shown in Fig. 10 as curve C. This curve was calculated from the Fig. 10. Transmission curves showing the effect of Wavelength, mp increasing the optical depth of the sodi<m liquid filter and a comparison with an interference filter.Curve A, normal sodium filter, 1 cm + 1 cm; curve B, double thickness sodium filter, 2 cm + 2 cm; curve C, typical interference filter (parallel beam) manufacturer’s data sheet and may be slightly optimistic, but there is no doubt that the efficiency of these somewhat costly filters is very good. It should be noted, however, that the maximum efficiency of an interference filter can only be attained when the transmitted beam consists of absolutely parallel radiation. For this reason the present optical lay-out is not considered suitable for use with these filters. By replacing the existing lenses by double lens systems designed so that the rays emanating from the centre of the flame are parallel when passing between the two parts of each lens system, it should be possible to use inter- ference filters in the present apparatus by inserting them in this part of the beam. The comparatively large flame area, however, makes it impossible to obtain a completely parallel beam. We acknowledge the helpful advice given by Dr. E. J. Bowen and Dr. L. Leyton, and also thank Dr. Bowen and Mr. L. G. Young for the loan of certain components of the first assembly. This paper is published with the approval of the Lords Commissioners of the Admiralty, but the responsibility for any statements of fact or opinions expressed rests solely with the authors.424 MILNER AND TOWNEND : THE DETERMINATION OF ALUMINIUM [Vol. 76 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. REEEREWES Boon, S. D., “Vlam-fatometrie,” D. B. Centen, Amsterdam, 1945. Barnes, R. B., Richardson, D., Berry, J. W., and Hood, R. L., Ind. Eng. Chem., Anal. Ed., 1945, Berry, J. W., Chappell, D. G., and Barnes, R. I3., Ibid., 1946, 18, 19. Rauterberg, E., and Knippenberg, E., Ernahr. Pflunze, 1941. 37, 73. Dorningo, W. R., and Klyne, W., Biochem. J., 1949, 45, 400. McGowan, G. K., private communication. Parks, T. D., Johnson, H. O., and Lykken, L., Anal. Chew., 1948, 20, 822. Shapiro, S., and Hoagland, H., J . Amer. Physiol. 1948, 153, 428. West, P. W., Folse, P., and Montgomery, D., Anal. Chem., 1950, 22, 667. Eubank, W. R., and Bogue, R. H., J . Res. Nut. Bur. Stand., 1949, 43, 173. 17, 605. ADMIRALTY MATERIALS LABORATORY HOLTON HEATH POOLE. DORSET