26 Analyst, January, 1974, Vol. 99, pp. 26-37 Mass-spectrographic Analysis of Geological Samples Using the Low-voltage Discharge Source BY P. F. S. JACKSON AND A. STRASHEIM (National Physical Research Laboratory, South African Council for Scientific and Industrial ResearcP, P.O. Box, 395, Pretoria, South Africa) Problems associated with electrode performance were encountered in the application of the low-voltage discharge source to accurate quantitative analysis of non-conducting geological samples. By using compound elec- trodes of the sample and conducting powder, preferential loss of the con- ducting material from the electrode system led to imprecision and to eventual failure of the electrical discharge. The effects of discharge current and electrode composition on this denuding factor are discussed.Results are given for the analysis of three international standards: G2, BCR-1 and W-1, and a comparison is made between the results obtained by mass spectrography and by other analytical techniques used in the routine analysis of geological material. The analytical range for most elements is more than two orders of magnitude. Results reported indicate that analytical values with an accuracy of better than 10 per cent. can be obtained by using mass spectrography. THE first commercial spark-source mass spectrograph became available in the late 1950s. This apparatus was fitted with a radiofrequency spark source, which has been almost univer- sally used in this technique. Other sources, the laser and the low-voltage discharge sources, found limited applications but results from these sources were overshadowed by the volume of data produced by the use of the radiofrequency spark.A detailed description of the two types of spark source has been given by Fran2en.l Brown and \Volstenholme,2 in 1963, published mass-spectrographic survey analyses on several non-conducting matrices, including geological materials. By using the radiofrequency spark source and compound electrodes prepared with 50 per cent. m/m graphite, they obtained analytical results for almost fifty elements in each sample, some elements being determined down to well below 1 p.p.m. The analyses were described as quantitative, with the assumption that all elements had equivalent sensitivity in the spectrograph, which was soon proved to be erroneous. In the following years many author^^-^ incorporated internal reference standards and made use in their interpretative procedure of relative sensitivity factors determined from a known reference material.The use of other analytical matrices, such as silver and gold in the analysis of titanium(1V) oxide,s and silver in the analysis of rare earths (H. H. Whittaker, private communication, 1965), was made necessary by the relatively high abundance of major-element carbides formed in the radiofrequency spark when graphite was used as matrix. These carbides caused deconvolution problems that arose from the superimposed spectral lines, and made the determination of low-level impurities at the relevant mass positions impossible. Precisions and accuracies given in the literature varied from values close to 5 to about 300 per cent.in certain instances. Some of this imprecision stemmed from problems associ- ated with homogeneity and the very small amount of sample consumed during the analysis. Nicholls, Graham, Williams and Wood7 showed that by repeated fusion of the sample with rapid quenching and grinding, coefficients of variation of better than 5 per cent. could be obtained on known heterogeneous samples. Other authors6s9 preferred electronic gating of the beam, thus increasing sample consumption per unit exposure. Nearly all of these results were obtained by using the radiofrequency spark source. The low-voltage discharge spark was only very occasionallyl0>l1 used as a source of ions. Franzen, Schuy, and Maurerll developed an interpretative technique based on a comprehen- sive study of the photoplate and its reaction to ion bombardment.In this work, and in work on mass-spectrographic precision, they discussed the use of broad line profiles with homogeneous blackening in contrast to the distorted Gaussian profile normally obtained. @ SAC and the authors.JACKSON AND STRASHEIM 27 They proposed that the grain statistics of broad homogeneously blackened lines were naturally better than those otherwise obtained, and thus permitted a more accurate measurement for the photometric conversion of the analogue line intensity into the required digital counterpart. They indicated that with standard steel samples they were able to obtain a relative standard deviation of better than 1 per cent.for the isotopes of molybdenum, and analytical coefficients of variation of between 4 and 8 per cent. for some of the minor components of the steels. No successful analyses of silicate rocks by low-voltage discharge source mass spectro- graphy appear to have been reported and this paper describes an attempt to use such a source for the analysis of geological samples. DEVELOPMENT OF THE METHOD EXPERIMENTAL- A Varian-Mat SMIBF spark-source mass spectrograph was used throughout this work. In the course of developing the method the commercial apparatus was modified, in that a Pfeiffer 260 1 s-l turbomolecular pump was added to the source region instead of the Edwards 300 1 s-1 oil-diffusion pump and cold-trap. The butterfly valve leading from the pump to the source was replaced by a VAT gate-valve (Balzers Ltd.), and the optical baffles were removed except for the protective gauze at the entrance to the source.Minor modifications were also made to the electronic system so as to enable stabilised currents of as low as 1 A to be used. The graphite used was supplied by Ringsdorff (RW-A quality) and the silver powder by Cominco. Electrodes were prepared by admixing the dry powders by hand with a pestle in an agate mortar in order to obtain the best result. Mixtures produced by commercially available shakers and amalgamators proved unsatisfactory as in certain instances separation of the two phases could be seen even after mixing for 1 hour. The powder mixture was then placed in a heavy-duty polythene mould, so as to avoid contamination, and compressed to about 9 kbar.The press was supplied by Research and Industrial Instruments, London (Type 0025). The ion path was adjusted to give a broad, homogeneously blackened line-profile with a line width of about 0.1 mm. The electrodes, mounted on their respective holders, were positioned accurately within the source of the spectrograph in order to reduce variations in electrode position between sets of electrodes to a minimum. These variations had previously been shown12 to affect spectral quality and also precision. The positioning of the electrodes was made by using a calibrated Perspex spacer, the thickness of which controlled the position of the electrodes relative to the primary slit. The horizontal and vertical position of the electrode gap was set from the calibrations on this spacer, which itself is aligned with the source-gun.The discharge, under the conditions used, continues for approximately 100 ps after the ignition spark. In order to select material emanating only from the low-voltage discharge and not from the ignition spark, the gate, synchronised to the ignition spark, was kept closed for 20 ,US after ignition, then opened for approximately 80 ps. These conditions were maintained for the whole of the programme. In this manner any possible differentiation within the ion cluster produced by each discharge is avoided. OPTIMUM OPERATIKG CONDITIOKS- were studied; those investigated, including ranges, are given in Table I. To obtain optimum conditions, all parameters affecting the operation of the instrument TABLE I PARAMETERS STUDIED IN ORDER TO OBTAIN OPTIMUM EXCITATION CONDITIONS Discharge current ..1 t o 5 A Repetition rate . . . . 1 to 100 Hz lnitiator spark energy . . 0.125 to 0.625 J Silver to sample ratio 10 : 1 t o 5 : 1 Graphite to sample ratio 10 : 1 to 1 : 1 . . The criteria used to judge the performance of the system were precision and lifetime of Two factors were found to influence the performance of the system, vix., the The higher current settings also increased the These fragments, the electrode. electrode composition and current settings. number of flying fragments observed in the vicinity of the electrodes.28 JACKSON AND STRASHEIM : MASS-SPECTROGRAPHIC ANALYSIS OF [Analyst, VOl. 99 when the machine was used with the original pumping system, tended to cause the accelerating voltage to become unstable and many voltage breakthroughs were observed. The pressure in the source was found to vary, depending on the pulse repetition rate used.At 100 Hz, pressure peaks were observed to exceed loF5 torr, at which level the accelerating voltage could no longer be maintained. With the turbomolecular pump no pressure peaks were observed, and the vacuum, when sparking at 100 Hz, did not deteriorate beyond 5 x lo-' torr. The precision itself was not affected by the pulse repetition rate nor by the initiating spark energy and these were set to maximum repetition rate, for maximum ion output, and an intermediate value for the initiating spark energy of 0.375 J, which gave an adequately low resistance in the spark gap for the discharge to ignite. The influence of high current on the electrode surface is illustrated in Fig. 1.An electrode mix was prepared consisting of 9 parts of graphite and 1 part of sample, from which electrodes were prepared and exposed to 2, 4 and 6 A low-voltage currents for 1000 s. The local melting of the sample is increased at 6 A and the depth and size of the craters is also increased. The influence of low graphite to sample ratios on the performance of the electrode can be seen in Fig. 2. Electrodes consisting of 9 parts of graphite and 1 part of sample and also of 1 part of graphite and 1 part of sample were exposed to a current of 4 A for 500 s. Scanning electron micrographs of the resultant surfaces are shown in Fig.2. The structures are totally different with evidence of local fusing occurring in the case of the 1 : 1 system. The best precision was found for samples with a high conducting matrix content that were excited at low current values, From the evidence cited above, it was decided to use the more dilute sample elec- trodes, i.e., with a graphite to sample ratio of 6: 1, with a low discharge current of about 2 A. Initially, the results with this matrix were poor with relatively large coefficients of variation, notably for the alkaline earth elements, On pre-heating the original sample at 600 "C prior to preparation of the electrode, these coefficients of variation improved considerably. Also noticeable was the relatively poor sensitivity of the alkaline earth elements in this matrix.(The sensitivity was determined as the inverse of the correction factor required to convert the observed concentration, measured in terms of an internal standard, into the expected value, as defined by other techniques.) This phenomenon is much less prominent in graphite, and Table I1 gives an indication of the difference in sensitivity in the two matrices for six elements, all being determined relative to iron (the internal standard). Scanning electron micrographs of the sample surfaces can be seen in Fig. 1. Silver was also used as conducting matrix. TABLE I1 RELATIVE CORRECTION FACTORS USED IN THE ANALYSIS OF SILICATE MATRICES FOR SILVER AND GRAPHITE Relative correction factors - Silver Graphite Iron . . .. 1.0 1-0 Vanadium ..0.8 0.82 Chromium .. 0-7 0.69 Manganese .. 0.59 0.57 Calcium . . 3.1 0.4 Strontium . . 9-2 0.8 Barium .. 36 2.3 The final choice of electrode composition and source parameters is summarised in Table 111. STANDARD REFERENCE SAMPLE- The factors listed in Table I1 were determined from results obtained by using the inter- national rock standard G-1. Samples of the finely ground rock were mixed with silver and graphite and correction factors, relative to the iron concentration, were calculated for as many elements as possible. The comparison of these factors between the two conducting matrices0 + w [To face p. 28Fig. 2. Comparison of electrode surfaces after exposure t o a current of 4 A for 500 s. Electrode composition (ratio of graphite t o sample) : ( a ) , 9 : 1; and ( b ) , 1 : 1January, 19741 SAMPLES USING THE LOW-VOLTAGE DISCHARGE SOURCE 29 is in no way affected by the value adopted for the absolute concentration in the standard, as the same standard, from the same bottle, was used throughout these experiments.TABLE I11 SUMMARY OF SATISFACTORY ELECTRODE COMPOSITION AND DISCHARGE SETTINGS Discharge current . . 1.1 t o 2.2 A Pulse repetition rate . . 100 Hz Ignition energy . . . . 0.375 J Silver to sample ratio . . 9 : 1 m/m Graphite to sample ratio 6 : 1 m/m For calibration purposes, however, an absolute value had to be adopted. In view of the wide discrepancies between determinations in the compilation of Fleischer,13 it was decided to use a preferred value rather than a simple mean. The data of Fleischer were taken, to which were added any recent data available to us on the standard G-1.Care was taken to ensure that no systematic bias was included by over-emphasising any particular method, but due significance was given to those methods which, owing to the recent attention paid to them, have been proved to be precise and reliable, e.g., isotope-dilution mass spectrography for the determination of rubidium and strontium, the results for which are commonly quoted to three and four significant figures. A simple average was then taken and outliers were rejected by the normal statistical process suggested by the A.S.T.M. (E178-68 p. 156-165). The resulting data were then re-averaged and the residual data inspected for bias of the method. In general, little bias was found, most of the results displaying a random distribution about the weighted mean.This value was then adopted as a starting value. Table fV contains some of the preferred values adopted in this work for the purposes of comparison. TABLE IV VALUES ADOPTED FOR THE U.S. GEOGRAPHICAL SURVEY STANDARD G-1 I N THE CALCULATION OF RELATIVE CORRECTION FACTORS M a j o r elements ( a s oxides), p e r cent.- CaO 1.37 K,O 5-5 TiO, 0-26 MnO 0.024 P,O, 0.09 MgO 0.38 Trace elements, p.p.m.- Ba 1150 Ho 0.4 Sr Be 2.4 La 103 Tb Ce 160 Li 22 Tm co 2.2 Lu 0-12 v Cr 15 Nb 20 Y c u 15 Nd 55 Yb Dy 2-8 Ni 1.5 Zn Er 1.3 Pr 16 Zr Gd 5.0 Sm 8.0 Eu 1.2 Rb 220 250 0.6 0.16 15 10 0.80 42 200 DISCUSSION OF OBSERVATIONS DURING DEVELOPMENT OF METHOD As the transfer of material in a low-voltage discharge system has been shown to be almost entirely unidirectiona1,l a problem arises with compound electrode systems, in which the rates of consumption of the two intermixed, but discrete, phases are found to be dissimilar.The surfaces of three electrodes after sparking under different conditions have been shown (Fig. 1). With low currents (2 A) the depth of the craters is relatively shallow and siliceous protrusions do not dominate the sample surface. Under the influence of a higher current, and also with less graphite present (Fig. 2) in the electrode mix, a coral-like structure begins to appear to be dominant on the electrode surface, which is non-conducting and causes the spark discharge to become erratic and finally to be extinguished. At the extinction point, deep craters and large siliceous protrusions almost cover the sample surface.The formation of these structures has been described and discussed more fully elsewhere.l*30 JACKSON AND STRASHEIM : MASS-SPECTROGRAPHIC ANALYSIS OF [Analyst, Vol. 99 The remarkably high relative correction factors required to convert the observed mass- spectrographic values, which were obtained for the alkaline earth elements in a sample - silver mixture into quantitative data, indicate a very poor ion yield for these elements under the conditions selected. The ion yield for these elements is, however, dramatically improved by a small addition (2 per cent.) of graphite to the electrode mix. This observation was also noted by Scott, Strasheim and Jacksonlj when using the relatively low temperature plasma of normal-mode laser radiation to excite geological material for subsequent mass-spectro- graphic analysis.They proposed that the presence of carbon creates a reducing environment, which assists in the fracture of strong metal to oxygen bonds, as is found with refractory oxide type material. The same theory has been adopted in this paper to explain the remark- able change in the ion yield of the alkaline earth elements when graphite instead of silver is used as the conducting matrix. The oxides of these elements are very stable thermally and highly electrically insulating. Under the cool plasma conditions selected, these oxides would tend to remain behind in the cathode. The reducing nature of the graphite effects the release of these elements by means of a thermochemical process and thus accounts for their large change in sensitivity.An extension of this effect may also explain the poor precision obtained for the alkaline earth elements, prior to calcination of the original sample, when a silver matrix is used. It is suggested that residual carbon present in the original sample, which acts as such a releasing agent, was the cause of the erratic ion yields observed in these instances. Calcination of the original sample prior to electrode preparation removed this carbon and thus eliminated the cause of the extremely poor reproducibility. PRECISION- In the SMIBF mass-spectrographic system, provision is made for a total of thirty exposure positions on each photoplate, thus enabling the precision with which any single exposure may be recorded to be assessed. Replicate exposures were taken at the same exposure value, and a statistical analysis was made of the variations in line transmission, measured by micro- photometer, of the replicates for several mass lines, each of which had a different transmission, It was found that, between transmission values of 85 and 8 per cent., the standard devia- tion, which is defined as the positive square root of the variance, was almost independent of the isotope or element, but was strongly dependent on the value of the exposure selected.The precision of a single exposure of between 0.1 and 10 pC was found to be very poor, even for gating conditions so stringent that only about 0.1 per cent. of the total possible beam was transmitted to the photoplate.Exposures of such small magnitude were therefore considered too imprecise and their use was avoided, which restricted the upper limit of the concentration that could be determined to approximately 2000 p.p.m.a. for any specific isotopic species. Exposures in the range 10 to 10000 pC were found to be more precise. The standard deviation recorded in these instances indicated a certain dependence on the transmission value itself, with the smaller deviations being recorded at lower transmission. The trans- mission scale of the Optica Milano microphotometer used in this study is calibrated from zero (no light passing through the microphotometer) to 1000 (corresponding to clear glass on the photoplate). At a mean line transmission value of 120 scale units (12 per cent.transmission), the typical standard deviation was between 3 and 4 units. At a mean transmission value of 720 scale units (72 per cent. transmission), the typical standard deviation had deteriorated to close to 15 units, and at transmission values closer to the background value (clear emulsion at about 920 scale units) this value had deteriorated still further to approximately 28 scale units. Above 10 000 pC (below about 3 p.p.m.a.), the precision depended on the performance of the electrodes which, under the conditions proposed, did not change discernibly. EXPERIMENTAL- Three standards, six lunar basalts and eleven rocks of unknown composition were analysed by using graphite as the conducting matrix. A range of exposures covering the analytical range desired was run and the photoplate developed in the normal fashion.The transmission values of the line, read by microphotometer, and the exposure for each of the lines were fed into an IBM 360 computer, with which data processing was handled according to the Franzen, Scliuy and Maurer equations.ll APPLICATION OF METHODJanuary, 19741 SAMPLES USING THE LOW-VOLTAGE DISCHARGE SOURCE 31 With these equations, and provided sufficient data points are available for regression analysis, values can be obtained for the unknown parameters Ts, the saturation transmission of the photoplate due to primary blackening, and V , the parameter defining the slope of the function, for each ion species. When insufficient data points are available for this treatment, an estimate of T , can be obtained from the equation- where a, b and c are constants determined for each plate, and is the mass of the ion for which Tt9)(=, is required.The value of V in these cases is determined approximately by interpolation or extrapolation. (For each individual element, the values of Ts and V obtained over the whole series were almost constant.) The individual ion intensities could therefore be established, and, by comparison with the ion intensity for the iron, the concentration of which was determined by other methods, a value was calculated for the concentration of individual elements. The response function was developed from the measured responses of the isotopes iron-54, -56 and -57. The con- centration value for iron was taken to be the mean of the values obtained by four separate iron determinations, which were achieved by four dissimilar methods, vix., chemical analysis,ls X-ray powder analysis,17 X-ray fusion18 and atomic-absorption spectrophotometry.~~~~~ The coefficient of variation between these methods determined over ten test samples was less than 5 per cent., which is in agreement with that found by Strasheim and Jackson,21 who reported that the determination of iron in geological material gives highly reproducible results.This result could therefore be relied upon and contributed little or no systematic error to our results, even when the geological character of the sample changed considerably. RESULTS AND DISCUSSION The results obtained on the twenty rock samples have been subdivided into groups, and those for all of the elements determined are given only for the standard samples and are tabu- lated in Tables V, VI and VII.For the other seventeen samples, certain elements have been TABLE V T(s)(,, = a M(n) + b M(n)+ + c Iron was selected as internal standard in all silicate analyses. ANALYSIS OF G2 COMPARED WITH THE RECOMMENDED VALUES AND THE RANGE OF VALUES OBTAINED BY OTHER METHODS Major elements (as oxides), per cent.- Recom- This mended Oxide work value CaO 2.0 1.94 - 4.51 0.55 0.50 K O TiO, MnO 0-032 0-034 P,O, 0.14 0.14 MgO 0.80 0.76 Trace elements, p.p.m.- Recom- This mended Element work value Ba 2100 1870 Be 2.1 2.6 Ce 160 150 c o 5.4 5.5 Cr 15 7.0 c u 11 11.7 DY 2.42 2.6 Er 0-93 1.3 E U 1.7 1.5 Gd 6.0 5-0 Ho 0.34 0.4 La 87 96 Li 50 34.8 I,U 0.12 0.1 1 Range 1.8-2.3 4.3-5.1 0.42-0.5 7 0.02-0.04 0.1 1-0.23 0.34-1-08 Range 1500-3000 1-5-3 140-180 2-2 1 5-29 <2-17 2-5 0.8-2.6 1.3-3.2 3-7‘0 <0*3-0.7 76-250 25-63 0.10-0.2 This Element work Nb 15-5 Nd 58.1 Ni 5.7 Pr 16.8 Rb 240 Sm 9.05 Sr 500 Tb 0.47 Tm 0.12 V 37 Y 15 Yb 0.93 Zn 80 Zr 300 Recom- mended value 13.5 60 5.1 19 168 479 (0.3 35.4 12.0 85 300 7.3 0.54 0.88 Range 8-20 42-67 2-1 4 19-20 1 08-5 1 3 7-1 1 235-680 O .P l * O 0.3-0-5 26-60 8-17 0.5-1,o 42-138 250-40032 JACKSON AND STRASHEIM : MASS-SPECTROGRAPHIC ANALYSIS OF [AnaZySk, VOl. 99 TABLE VI ANALYSIS OF BCR-1 COMPARED WITH THE RECOMMENDED VALUES AND THE RANGE OF VALUES OBTAINED BY OTHER METHODS Major elements (as oxides), p e r cent.- Recommended Oxide This work CaO 7.0 1.55 2.11 TiO, MnO 0.18 0.39 Trace elements, p.p.m.- K2O p20.5 MgO - This Element work Ba 576 Be 1.59 Ce 49.0 co 36 Cr 14.5 c u 22 5.7 3.0 DY Er Eu 1.8 Gd 6.0 €30 1.06 La 23.8 Li 17 Lu 0.54 value 6.92 1.70 2.20 0.18 0.36 3.46 Recom- mended value 676 1.7 53.9 38 17.6 18.4 6.3 3.59 1-94 6.6 1.2 26.0 12.8 0.55 Range 6.14-8.3 1.49-1.82 1.83-2.45 0.1 0-0.20 0.2 8-0.4 7 24-3-7 Range 480-1 230 1-3 40-53 29-60 8-45 7-33 5.7-6.6 3.1-3.7 1.8-2.4 5-8.5 1-1.3 22-36 10-19 0.4-0.6 Element Nb Nd Ni Pr Rb Sm Sr Tb Tm V Y Yb Zn Zr This work 11 28.7 20 6.3 50 6.0 330 0.84 0.43 340 27 140 200 2.9 Recom- mended value 13.5 29.0 15.8 7 46.5 6.6 1.0 0.6 37.1 330 399 3.36 120 190 Range 10-97 22-34 8-30 5-7 40-150 5.5-7-5 244-525 0.7-1.2 <0.4-0*7 120-700 20-52 2.3-5.0 94-278 144-275 selected to demonstrate the mass-spectrographic value compared, in most instances, with the average value obtained by X-ray fluorescence, atomic-absorption spectrophotometry and chemical analysis. These results are plotted on log - log graphs (Figs.3 to 9). TABLE VII ANALYSIS OF W-1 COMPARED WITH THE RECOMMENDED VALUES AND THE VALUES OBTAINED BY MASS SPECTROGRAPHY Major elements (as oxides), per cent.- Oxide This work value CaQ 11 10.96 0.65 0.64 1.15 1-07 TiO, MnO 0-17 0.17 0-14 0.14 - 6-62 Recommended K2O p205 MgO Trace elements, p.fi.m.- Recom- This mended Element work value Ba Be Ce Co Cr CU DY Er Eu Gd Ho La Li Lu 165 0.59 18.8 46 99 105 3.7 2.0 0.96 3-8 0.62 9.4 0.30 11 160 23 47 114 110 4 2.4 1.1 1 4 0-69 9.8 14.5 0.35 0.8 N i c h o 11 s et al.’s value 145 - 17-7 42 98 110 3-89 2.08 1.2 3.82 0.63 11.9 - 0.20 Taylor’s value Element 200 Nb - Nd 18 Ni Pr - Rb - Sm 2.6 Sr 1.8 Tb 0.95 Tm 3.0 V 0.78 Y 14 Yb Zn - Zr - - This work 6-5 13-4 70 2.7 20 3.2 195 0.44 0.29 260 20 85 110 2.1 Recom- mended value 9.5 15 76 21 190 3.4 3.6 0.65 0.30 264 25 86 105 2.1 1 qicholls et al.’s value 12.5 76 4.0 21 3.3 1 175 0.49 0.28 - 260 25 75 90 - Taylor’s value 11.0 2.5 - - 19 190 - 0.66 0.31 - 29 1.6 - 95January, 19741 SAMPLES USING THE LOW-VOLTAGE DISCHARGE SOURCE 33 1 10 100 1000 Average value, p.p.m.Fig. 3. Comparison of mass-spectrographic values with average values obtained by X-ray fluorescence, atomic-absorption spectroscopy and chemical analysis for rubidium : 0, test samples ; a, lunar samples; and x, U.S.Geological Survey standards As between 20 and 30 individual determinations (the transmission - exposures pairs) are made in order to create the mathematical response function, the precision within a single plate with which the ion intensity of any component can be measured bears no direct relation- ship to the precision of the individual exposure. Because of the variability of the standard deviation with transmission, and the non-linearity of the mathematical model, the evaluation of this plate precision is too complex and has been omitted. The precision, determined by replicate analysis of a single sample, indicates that for elements at concentration levels be- tween 1 per cent. and 1 p.p.m., a coefficient of variation of between 5 and 7 per cent. can be obtained for most elements, one of the limitations being the determination of those elements which are mono-isotopic, when the upper limit of detection is exceeded.The accuracy of the method is demonstrated in Figs. 3 to 9 and in Tables V to VII. The results obtained by I I 100 1000 10 000 Average value, p.p.rn. Fig. 4. Comparison as for Fig. 3, for phosphorus34 JACKSON AND STRASHEIM MASS-SPECTROGRAPHIC ANALYSIS OF [A'HdySt, VOl. 99 0.01 ~ 0.1 1 10 100 Average value, per cent. Fig. 5. Comparison as for Fig. 3, for calcium mass spectrography are compared with the average value by other methods. (This average is occasionally weighted in favour of the popular result by discarding the few obviously erroneous results from the collection.) Tables V and VI contain comparative22 results for the internationally recognised standards G2 and BCR-1. In Table VII are shown the results obtained by this method compared with the recommended and with those of Nicholls et a1.' and Taylor3 (results obtained with a mass spectrograph fitted with a radio- frequency source).In most instances and certainly at higher concentration levels the results follow a distinct 1 : 1 relationship when compared with the "average value". The differences become more apparent as the level of concentration decreases, as the other techniques with which comparison is made approach their individual detection limits. In 1969, Graham and Nicholl~,~~ discussing their results on W-1, commented upon their absolute error for the deter- minations of six of the rare-earth elements.In these instances the error appeared to be larger than expected and they suggested a possible systematic error in their assumed R values. The values presented in this paper, however, vary considerably from their originals and as this disparity was not discussed the more popularly quoted values' have been used for comparison. 0.0 1 0.1 1 10 Average value, per cent. Fig. 6. Comparison as for Fig 3,. for titaniumJanuary, 19741 SAMPLES USING THE LOW-VOLTAGE DISCHARGE SOURCE 35 The values for titanium (Fig. 6), manganese (Fig. S), calcium (Fig. 5) and phosphorus (Fig. 4) are very good, with the average error less than 5 per cent. except for phosphorus, with which a trend towards a systematic error is established at the lower level. Agreement is also seen between the results for these elements and the values obtained for the international standards with which the error in the determination of these elements is very small indeed.The comparisons of the results for chromium (Fig. 7 ) , copper (Fig. 9) and rubidium (Fig. 3) show larger differences, which are particularly noticeable below 20 p.p.m. As already stated, some of this error may be due to errors in the “accepted average result,” as these techniques are approaching their limits of detection (see degree of scatter indicated by bar in US. Geological Survey results). I I I 10 100 1000 10000 Average value, p.p.m. Fig. 7. Comparison as for Fig. 3, for chromium No strong inter-element effect has been noted for any of the elements determined (about 35) in each of the twenty rock samples.The same relative correction factors were used throughout, no matter which sample type was being analysed, and a significant third-partner effect should have been easily recognised. Owing to the uncertainty attached to the “accepted Average value, p.p.m. Fig. 8. Comparison as for Fig. 3, for manganese36 JACKSON AND STRASHEIM : MASS-SPECTROGRAPHIC ANALYSIS OF [A?i?a&St, VOl. 99 average,’’ no final conclusion can be drawn as to the over-all accuracy of the mass-spectro- graphic technique. It is believed that if the accuracy obtained at higher levels of concentra- tion could be extrapolated over the one or two orders of magnitude towards the very low results, a value of 10 per cent. may not be unrealistic. Although tests have demonstrated that no electrode deterioration is manifest after the exposure periods required for the lower levels of detection, it is felt that additional results by other techniques are required in order to give a significant comparison with mass-spectrographic values at the single part per million level and below.Clearly, however, the results demonstrate that the mass-spectrographic values will be very reliable “estimates” of these low levels. V I 1 1 10 100 1000 Average value, p.p.rn. CONCLUSION The results obtained from the use of the low-voltage discharge source in the mass- spectrographic study of geological materials indicate that, with care, accurate figures can be obtained for a large number of elements. The effect of carbon, as the conducting matrix, on the ion-yield of alkaline earth elements appears to be of a similar kind to that shown in laser source mass spectrography.The analytical precision that has been attained in this work confirms that, provided the effects of the various parameters are controlled, precise results can be achieved with this technique. The analytical accuracy of the mass-spectrographic system is once again demon- strated to be better than 10 per cent. in most instances, a figure which, for most geological work, is considered to be satisfactory. Fig. 9. Comparison as for Fig. 3, for copper 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. REFERENCES Franzen, J., “Analysis by Mass Spectrometry,” Chapter 2, Academic Press, New York and London, Brown, R., and Wolstenholme, W. A., Paper presented to the E l 4 Committee of the A.S.T.M. a t Taylor, S. R., Nature, Lond., 1965, 205, 34. ---, Geochim. Cosmochim. Acta, 1965, 29, 1243. Whitehead, J., and Jackson, P. F. S., Analyst, 1966, 91, 418. Whitehead, J., Jackson, P. F. S., Vossen, P. G. T., and Brown, R., Analyt. Chem., 1967, 39, 141. Nicholls, G. D., Graham, A. L., Williams, E., and Wood, M., Ibid., 1967, 39, 584. Jackson, P. F. S., “Proceedings of the Fifth Annual MS7 Mass Spectrograph Users Meeting,” Vossen, P. G. T., Analyt. Chem., 1968, 40, 3. Riddoch, A., “Proceedings of the Fifth Annual MS7 Mass Spectrograph Users Meeting,” April, 1965. Franzen, J., Schuy, K. D., and Maurer, K., 2. analyt. Chem., 1967, 225, 295. Strasheim, A., and Jackson, P. F. S., XVIth Colloquium Spectroscopicum Internationale, Heidel- 1972. the Eleventh Annual Meeting, May, 1963, San Francisco, U.S.A. April, 1965. berg, 1971.January, 19741 SAMPLES USING THE LOW-VOLTAGE DISCHARGE SOURCE 37 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. Fleischer, M., Geochim. Cosmochirn. Acta, 1969, 33, 65. Jackson P. F. S., and Strasheim, A., “South African Institute of Physics Conference, D19,” July Scott, R. H., Strasheim, A., and Jackson, P. F. S., Nature, Lond., 1971, 232, 5313. Strelow, F. W. E., Liebenberg, C. J., and von S. Toerien. F., Analytica Chim. Acta, 1969, 47, 251. Gricius, A. J., Proc. X V I I Colloquium Spect. Int., Florence, 1973, 02, 475. Wybenga, F. T., X-ray Spectrom., in the press. Butler, L. P. R., and Kokot, M. L., “Modern Methods of Geochemical Analysis,” Chapter 8, 1971, Butler, L. R. P., “Flame Emission and Atomic Absorption Spectrometry,” Marcel Dekker, New York Strasheim, A., and Jackson, P. F. S., “South African Institute of Physics Conference, Dl,” July Flannagan, F. J., Geochim. Cosmochim. Acta, 1973, 37, 1189. Graham, A. L., and Nicholls, G. D., Ibib., 1969, 33, 555. 1973, Pretoria. Plenum Press, New York. and London, in the press. 1973, Pretoria. Received APril 30th, 1973 Accepted August 2nd, 1973