Three air samples analyzed spectrophotometrically by the established diazotization-coupling method (8) contained 0.03, 0.01, and less than 0.01 ppm (by volume) toluene diisocyanate. A duplicate set of air samples taken simultaneously was analyzed by the thiotrithiazyl chloride method, and corresponding values of 0.03, 0.02, and less than 0.01 ppm, respectively, were obtained. Although the sensitivities of the two methods are about the same, the procedure employing (8) K. Marc3li, ANAL.CHEM., 29,552 (1957).
thiotrithiazyi chloride is somewhat simpler and is reiatively specific for meta-substituted diamines and diisocyanateo. Thus, an equimolar quantity of aniline does not interfere i * the determination of toluene-2,4-diisocyanatewith thi: trithiazyl chloride, whereas this interference in the diazotization-coupling method leads to a positive error of approximately 5 to 10 %. RECEIVED for review December 23, 1966. Accepted February 20,1967.
Precision and Accuracy in Trace Element Analysis of Geological Materials Using Solid Source Spark Mass Spectrography G . D. Nicholls, A. L. Graham, Elizabeth Williams, and Margaret Wood Department of Geology, University of Manchester, Manchester 13, England Investigations are described which have led to the development of procedures which yield a precision of better than +5% in solid source spark mass spectrographic analysis of geological and similar materials. The accuracy of the results obtained is thought to be within the limits of the precision, and analyses using this technique give results as reliable as other techniques of trace element analysis. The major change from the usually adopted procedures is in the method of electrode preparation, though other minor changes must also be made if maximum precision and accuracy are to be attained. In view of the very wide range of elements that can be determined and the high sensitivity of this technique, adoption of the procedures described makes solid source mass spectrography an extremely important method of trace element analysis.
GEOCHEMISTS have long been aware of the limitations imposed on their studies by the limits of sensitivity of many of those currently available analytical methods which permit a number of elements to be determined in one sample, e.g. optical emission spectrographic techniques, x-ray fluorescence analysis. ‘I’his had led, on the one hand, to the development of prearcing concentration techniques in optical emission spectrography (1,2) and on the other, to the adoption of more sensitive methods for a restricted number of elements, e.g. nuclear activation techniques. More recently, solid source spark mass Spectrography has been investigated as a possible analytical technique in geochemistry (3-5). This technique is highly sensitive and permits the simultaneous determination of a wide range of elements, including many whose geochemical distribution is still imperfectly known. Pioneer appiications of this technique to the analysis of geological materials led to suspicions that it lacked the accuracy and precision necessary for high quality geochemical investiga-
(1) D. M. Hint and G . D. Nicolls, J . Sediment. Petrol., 28, 4% (1 958). (2, R. K. Brooks, L. H. Ahrens, and S. R. Taylor, Geochim Cosmochim. Acta, 18, 162 11960) (3) K. Brown and W. A. Wolstenholrne, Nature (London), 201. 5Y1r (1 964). (4) S R. Taylor, Ibid., ZQ5, 34 (1965). ( 5 ) S. R. Taylor, Geochim. Cosmochim. Acta, 29, 1243 (1965).
584
e
ANALYTICAL CHEMISTRY
tions, and to expression of opinion that the technique is no: suited to such work. Such suspicions are quite unfounded and procedures have been evolved which yield an accu:a:y and precision comparable with, or better than, most of t h ~ other techniques available. These procedures are described in this paper, together with the reasons for adopting them, since if these reasons are fully understood there should be no difficulty in applying this generai technique for the analysis cf geological materials to more specific cases in allied analytical fields, e.g. analysis of semi-conductor materials. EXPERIMENTAL
The instrument used in this work was an Associated Elel.tricai Industries Limited M.S.7. mass spectrograph. General descriptions of this instrument nave been given by severci workers (5-7). Electrodes were prepared by compressip mixtures of the sample under analysis and RingsdorffwerKc RWA grade graphite in an electrode-forming die f8, unde: y. pressure of 7500 psi. Operating pressures in the anaiyzer to IO-’ torr region of the instrument were in the range and in the source region i X to 5 X torr. Fifteer graded exposures of the spectrum from a sample. were recorded polarographically on Ilford 4 2 plates, which were sub. sequendy processed according to the manufacturer instructions. In the approach adopted here to this method o i anaiys:: a basic equation can be written, viz, C E = CS
x
l%PS/hpE
x
IS/IE
>: 1/R
!I
where content of elemen: E in electrode analyzed, (in atomic parts per million) CL3 = content of a second element (9,e.g. an inrerna: standard element, (in atomic parts per millionj CE
=
( 6 ) R. D. Craig, G. A. Erroch, ana j. D. Waldron, Adwn. Musi Spectrometry, 1, 136 (19%; (n R. W. Brown, R. D. Crag, ana R. M. Elliot, Ibid., 2, l i l
(1962). (8) R Brown and W. A. Woistenhoime, Roc. Eleventh Con’* A.S.T.M. Comm. E14 (196%
fips
=
fipe
=
I, IE and R
= = =
exposure (in millimicrocoulombs) required to give a line of chosen density for a chosen isotope of eiement S on a photographic plate exposure in the same units required to give a line of the same chosen density for a chosen isotope of element E on the same photographic plate isotopic abundance of chosen isotope of S isotopic abundance of chosen isotope of E a factor (hereafter called the relative sensitivity factor) introduced as a measure of sensitivity of total recording procedure to line of element E used compared with sensitivity to line of element S used, which is arbitrarily assigned a value of unity.
Consider the various terms in this hasic equation. Is/IE can be obtained for almost all element pairs with an accuracy of 1 or better from published tables of isotopic abundances. ExpS/ExpEratios are obtained from measurements on the developed photographic plate. For each line the density of the line in several different exposures is measured on a JoyceL a b 1 recording :microdensitometer (fitted with a 0-1.5 density wedge) and, for each line, a plot of density us. exposure prepared, cxposure being plotted on a logarithmic scale-i.e., the plot is very similar to an H and D curve. The chosen density for determination of the exposure ratio is taken at a value falling on the straight portion of the line plot (density 0.8). The exposures required for the different lines used are then read from the plots. In the determinations of the values of R factors described subsequently, we foimd that correction for diffuse background on the photographic plates was negligible, but in the analysis o n samples it has, on occasion, proved necessary to make such a correction. This is done by recording the density of the background on either side of the line in question for each of the different exposures and preparing a plot of the mean background density in each exposure against exposure. From this plot ,and that of the internal standard line a value of ExpS/Exp~hksround is obtained and thus a content of element E in the electrode which would yield a line image of the same density as the background. This content of E equivalent to background is subtracted from the content of E determined from f i p ~ / E x pobtained ~ from the line density measurements to yield the true content of E in the electrode. Only rarely, and then in lines for which rnle is less than 60.have we found the correction to be significaxt. The need to o b a i n the exposure ratio as accurately as possible dictates certain features of the working conditions. In order that the hersections of the line plots with the exposure axis at the chosen density value shall be sharp, the photographic emulsion should be one for which a small change in exposure produces a considerable change in line density (in photographic terminology the emulsions should have high gamma characteristics). The Ilford Q2 emulsion is satisfactory in this respect. Working with such an emulsion necessitates making a series of exposures differing by only small amounts, otherwise too few exposures give densities within the measurable range and too few points are obtained for accurate location of the line plots. Two points falling on the straight portion of the line plot are not enough; three is thE minimum number; four or more, desirable. With only two points, misplacing one, because of photometry error, emulsion defect, etc., must affect the result. With three points the effect of misplacing one is less significant-with four or more the aberrant point is indicated. All 15 exposures on one plate are therefore assigned to a limited range exposure, e.g. between 10 X and loo0 x coulomb for t r a x element determination in the concentration range 5 ppm to 0.05 ppm atomic. In effect, by
sacrificing the range of concentration that can be accommodated on one plate, greater precision can be obtained over the range of concentration of interest. To extend the range of concentration covered, two or more plates of differing exposure ranges can be taken. This is a major departure from the practice frequently followed by users of this type of instrument. Cs. A consequence of the philosophy developed above is that the standardizing element S must also be present in the analyzed electrodes in such an amount that the isotope lines for this element will also have measurable densities over the same range of exposures. For trace element determination S must be present in trace amount. If an eiement already present in the analyzed material is chosen as S and its content independently determined by some other method of trace analysis, any errors in the second method automatically carry through into the mass spectrographic analysis. Consequently, S, the internal standard element, should be introduced. Taylor (5) lists 10 points to be considered in selecting a suitable internal standard element. These are not all equally significant. The important considerations are: The internal standard element must be one likely to be absent (or present at very low concentration level compared with the amount introduced) in the materials to be analyzed. It must be readily available in a high degree of purity either as the element (e.g. metal) or in a compound with other elements not sought in the analyses. The internal standard element should possess only isotopes of mass numbers indivisible by two or three, in order to diminish (virtually eliminate) the risk of interference in analysis lines of other elements by doubly or triply charged ions of the internal standard. The mass numbers of its isotopes should exceed 120 to avoid interference by molecular (bi- or polyatomic) ions of the internal standard. In the work reported in this paper, rhenium (isotopes I85 and 187) was used as an internal standard. Our restriction of exposure range recorded on one plate, together with out' approach to the question of determination of the more abundant trace elements (discussed later in this paper) led us to reject as necessary the condition that the internal standard element should possess at least two isotopes with s ratio of about 100 (5). Introduction of the internal standard element involves adequate mixing and homogenization of the charge used for electrode preparation. A very small amount of electrode is consumed in each analysis and 0.1 mg must be truly representative of the whole. With the need to introduce the internal standard element at ppm concentration level, reasonable doubt may be entertained regarding the possibility of achieving the necessary degree of homogenization. Fortunately it is possible to check the homogeneity of electrode charge mixes with the mass spectrographic method of analysis. First, consider a single element, e.g. Re, isotopes 185 and 187. From the basic equation given earlier: Cb
=
C RX ~ E X P i r n I E X P r a s X IimIIi,, X 1/R
(2;
Re-arranging and substituting for 1187 and Gss, E x P ~ J E x P ~=u62.93137.07R ~ = 1.6981R
(3)
In the work we have repeatedly found R = 1 for isotope pairs of the same element (1.e. fip1a5/Expim = 1.698 + 0.04). Variation in repeated determinations of E ~ p ~ ~ ~ j could Exp,~ be due to instrumental errors, line plotting errors, etc., bu! cannot be due to inadequate homogenization of the electrolz changes during mixing. We have found that precisior. VOL 39, NO. 6, MAY 1967
585
were prepared. The results were similar to those obtained for the first mixing method used. (4) By proceeding as in (3) as far as the mixture of Rebearing silicate base and finely ground sample and then subjecting the resulting powder to a fusion process. Halfg r a m lots of the powder were tightly wrapped (parcelled) in platinum foil, and the resulting capsule was heated to a temperature of approximately 1200" C in the ring heater of a radio-frequency heating unit for 3-4 minutes. After chilling, the capsule was unwrapped and the glass resulting from this treatment was ground to a very fine powder in an agate mortar. A considerable improvement in overall precision resulted from the use of this process when complete fusion was attained, but frequently the charge was not completely fused in the short time it was held at 1200" C and had to be abandoned. It is well known that silicate compositions corresponding to most natural rock compositions melt slowly Table 11. Composition. of Standard Base to a viscous liquid. However, it was deemed inadvisable to extend the period of heating on account of the dangers of Si02 48.1% CaCOa 17.9% (by weight) contamination of the charges with platinum and of alloying A1203 13.6% Na2C03 3.4% with the foil of the capsule. Likewise, it was deemed inFez08 10.1 % KiCOa 0.8% advisable to raise the temperature on account of the danger MgO 6.1% of blow-outs at weakened points in the folded foil of the a Composition chosen to correspond to approximately 90 parts capsule. of a typical basic igneous rock with 10 parts excess COIIintroduced A new composition for a fusion base was sought to meet in the carbonates. the following specifications. It should contain none of the elements likely to be sought as trace constituents of geological samples, thus eliminating boron-bearing fluxes, etc. It should melt at a temperature below 1000" C. of determination of this and other isotope exposure ratios It should produce a liquid of low viscosity on melting. (for isotopes of the same element) is +2-3 %. This liquid should be miscible with molten silicates corNow consider a second element E with isotopes E,, E2,etc. responding to rock compositions in all proportions. The precision obtained in repeated determinations of &pBJ General petrological knowledge immediately directs atfiplS5, f i p s , / f i p i m , ~ P d E X p i m fip~JEXpia5, , etc., reflects tention to the system FezSi04(fayalite)-NaAlSiO, (nepheline)the overall precision of the method and is influenced by mixing (Na,K) AISilOs (alkali felspar) for such a Composition. errors (failure to homogenize the electrode charges) as well We are indebted to W. S. MacKenzie for data on low-melting as instrumental errors, etc. If the precision obtained in compositions in this system. The composition chosen is repeated determinations of these ratios is significantly worse given in Table I, column 1. The alkali metals were added than that for fip1g6/Explmr failure to homogenize the charges as carbonates to give the final fusion base composition shown sufficiently during mixing is indicated. in Table I, column 2. The introduction of some carbonate Various methods of introducing Re into the charges were is highly advantageous. Small bubbles of carbon dioxide, tested using this approach. liberated on heating, move through the liquid in the sealed (1) By the method of successive dilution a mixture of capsule during the fusion and, by a stirring effect, assist the graphite and Re metal was prepared containing 20 ppm mixing process. Furthermore, the pumice-like nature of (atomic) Re and equal amounts of this Re-graphite mixture part of the resulting chilled glass makes subsequent crushing and the very finely ground sample mixed to produce the to a powder much easier. charges. This is the usual way of preparing charges (5). By the method of successive dilution Re (as Re metal) was Dry mixing was tried, by hand in an agate mortar and meadded to this low melting point base to yield a low melting chanically in a variety of mixing mills. The overall precision point flux containing 21.1 ppm (atomic) Re. (on determinations of inter-element ratios) was 20-25x: the precision on determinations of E X ~ ~ S was ~ / 2-3 E ~%.~ ~ S ~The method of introducing Re into the charges finally adopted was as follows. Equal amounts of finely ground Attempts to improve the overall precision by time-consuming sample and Re-bearing flux were mixed on a Turbula-Schatz ultra-fine grinding of the sample before mixirg did not type 2 pulsator mixing mill for 15 minutes and the resulting succeed in lowering the overall precision to below 15%. mixed powder was tightly packed into 3/4-inch X 3I4-inch Thus, this mixing technique fails to homogenize the charges platinum envelopes prepared from 0.001-inch Pt foil. Each sufficiently for this method of analysis. envelope was sealed, care k i n g taken to exclude air. Ex(2) By proceeding as above but converting the final mixed pansion of the charge, on fusion to a liquid, places the foil powder into a slurry with purified acetone and then mixing of the envelope under stress but serves to inhibit the loss of the slurry in a variety of mixing mills. The slurry was then volatile elements. The capsules (envelopes charges) were allowed to dry out thoroughly before the electrodes were heated in a radio-frequency heater unit for 1 minute and then formed from the dry powder. The overall precision obchilled. Air chilling would be adequate for many rock comtained was 12-20x: the precision on isotope ratios for positions but, in view of the known extremely rapid crystalisotopes of the same element was 2 - 3 x . Again, failure to lization of olivine from silicate melts of such Composition homogenize the charges is indicated. that this mineral might form, capsules were frequently chilled (3) By preparing from Specpure oxides and carbonates a by dropping them into liquid nitrogen. The chilled glass standard base to the approximate major element composition was removed from the envelope and ground to a fine powder. of the geochemical standard rock W.l. and adding Re metal This power was then tightly packed into a new envelope and to this base to produce, by the method of successive dilution, the fusion process repeated for a further minute. After a base containing 40 ppm (atomic) Re. Equal amounts of grinding the glass to a powder it was, yet again, packed into this Re-bearing silicate base and very finely ground sample a new envelope and subjected to a third fusion. Recently we were mixed, and then equal amounts of this mixture and have modified this triple-fusion procedure by adding to the graphite to produce the final mix from which the electrodes
Table I. Composition Data for Low Melting Point Flux (Figures in weight percentages) Composition 2 Composition 1 54.3% Si02 Si02 49.0% AllOa 16.6% 18.4% Also3 14.1 % NazCOa 21.7% NalO KD 1.7% KzCOa 2.3% Fda 11.5% Fez03 10.4% 1. Low melting composition in the system fayalite-nephelinealkali felspar. 2. Final composition selected for the low melting point flux. Rhenium added as metal to produce, by the method of successive dilution, a flux containing 21.1 ppm (atomic) Re.
+
586
ANALYTICAL CHEMISTRY
powdered glass 10% by weight Specpure Na2C03before the second and third fusions thus restoring the built-in stirrer lost during the grinding of the glass to a powder. The finely powdered glass from the third fusion was mixed with an equal weight of graphite on the pulsator mixing mill for 15 minutes, and the electrodes were prepared from the resulting mixed powder. The overall precision of repeated determinations of various inter-element pairs was better than f5 using elcctrodes prepared in this way. Precision of repeated determinations of E ~ p ~ 8 ~ / E x pwas ~ 8 7still 2-3 %. This procedure appears, therefore, to produce electrodes of satisfactory, though still not perfect, homogeneity. The term R. While there is considerable interest in relating values of the R factors to the physical parameters of the elements, emulsion response to particles of different mass and so on, it is unneci:ssary to do this for the evolution of a satisfactory analytical procedure. Though as used here they are undoubtedly composite in character, they can be empirically determined from standards subjected to the Same X analytical procedure, since R = C,/Cti X EXPS~EXPS Is/&. A standard base was prepared from Specpure chemicals to the composition given in Table 11, and a range of elements normally found in trace amounts of natural rocks added to this base as Specpure oxides, carbonates, etc. By the method of successive dilution, standards were prepared containing groups of these elements in the concentration range 1-10 ppm atomic, the content for each element being accurately known. One gram of each standard was well mixed with 1 gram of Re-bearing low melting point flux on the pulsator mixer and the ratio Cs/Cti calculated for the various element-Ra pairs in the standard. The resulting mixes were subjected to the fusion process described earlier. After mixing the powdered glass with graphite, electrodes were prepared in the usual way. These electrodes were sparked, the spectra being recorded under the following operating conditions. Spark voltage 25 kV; accelerating voltage 19.6 kV; pulse repetition rate 300/sec.; pulse length 200 pseconds; monitor range setting 100; integrator range setting 10 or 100 as appropriate; exposure range I X 10-9 to 100 X coulomb. Exps/ExpE was determined for each line pair of interest as described earlier and, by substitution in the ab,ove equation, values of R factors were determined for various elements bonded to oxygen against metallic bonded Re. Repeated standard runs were made to permit a reliable set of working R factors to be established (see also later section on Precision). These R factors, for 35 elements sought in rocks, are given in Table 111. They are considered to be accurate to f 1 in view of the number of repeat determinations made. Values for other elements not listed in Table 111 can be determined in the same way if desired. RESULTS
The precision and accuracy of the technique and procedures adopted were investigated in the following ways. Precision. In all determinations using this technique the measured parameters are Exps and ExpE, whether the analyzed material is a standard or an unknown. Thus the precision can be evaluated by considering the variation in values of the R factors for given element-Re pairs on a series of runs. In Table IV individual results for two typical pairs, Cu-Re and Ge-Re are given to illustrate the variation found in repeated determinations of the R factor for such pairs. The agreement between various determinations of the R factor for a given pair is good. For Cu-Re the maximur.1 error of any determination from the mean value is 5 % of that mean: for Ge-Re the corresponding maximum error is 7.4% of the mean. Too few repeat determinations have yet been made to satisfy the rigorous statistical requirements of standard devia-
Table III. Values for R Factors for Various Elements Bonded to Oxygen against Metallic Bonded Rea Sparkingconditions: Spark voltage ZkV, pulse repetition rate 300/ sec, pulse length 200 psec, accelerating voltage 19.6kV As 9.43 Ge 21.7 Sb 16.5 Ba 90.0 Hf 8.24 Sc 16.3 Ce 26.2 Ho 15.5 Sn 26.2 Co 20.5 In 34.9 Sr 57.8 Cr 10.5 La 18.0 Tb 15.5 Cs 87.1 Lu 16.4 Th 9.30 Cu 12.4 Nd 24.1 Tm 17.2 Dy 22.6 Ni 43.2 U 10.4 Er 14.0 Pb 10.8 V 15.4 Eu 28.0 Pr 15.5 Y 14.0 Ga 44.9 Rb 102 Zn 7.34 Gd 26.4 Re 1.0 Zr 29.7 (arbitrarily assigned) These values refer to the conditions stated--R factors for elements bonded to oxygen us. oxygen bonded Re are different.
Table Iv. Repeated Determinations of Value of R Factors for Oxygen-Bonded Cu and Ge us. Metallic Bonded Re under Sparking Conditions Listed in Table I11 cu Ge 12.8 12.7 12.6 12.6 12.4 12.4 12.0 11.8 Mean 12.4 Maximum error from mean f0 . 6 Rel. std. dev. 2.8
23.W 22.6 22.5 22.5 22. 4n 22.4 22.4 22.0 21.9 21.8 21.4a 20. 8a 20.6 20.64 20.3 20.1 Mean21.7 Maximum error from mean f 1.6 Rel. std. dev. 4.3
z
z
These values were determined on a standard for which the Re/Ge concentration ratio was 5 times that of the standard used to obtain the other values. a
tion calculations. This is true, however, of many standtird deviations quoted in the literature as indications of the precision of accepted analytical techniques. For comparative purposes, relative standard deviations have been calculated for both columns of data in Table IV, it being admitted that these cannot be regarded as real relative standard deviations for the method in view of the number of separate determinations. The relative standard deviation on the the Cu-Re R factor values is 2.8 %; that on Ge-Re 4.3 %. Even if the precision is judged on the maximum error figures given above, the precision obtained using the procedure described here is very much better than that usually attributed to solid source spark mass spectrographic analysis of semi- and nonconductors. This improvement is attributed largely to the precautions taken to ensure electrode homogeneity. Unless such precautions are taken, it is obvious that an instrument can earn a bad reputation by being too good, by seeing a real variation tacitly assumed by the analyst to be absent. VOL 39, NO. 6, MAY 1967
587
Accuracy. Assessment of the accuracy of a technique is more difficult than assessment of the precision and the usual approach adopted is the analysis of an accepted standard on which determinations have been made by other acceptable techniques. At best this merely establishes agreement between results obtained by different techniques rather than any absolute standard of accuracy. At worst, if the standard varies from batch to batch, no deductions can be made about accuracy at all. We prefer a different approach. Standards were prepared containing the same trace element assemblages, but having different values of the R e t r a c e element concentration ratios. Values for the R factor for the same element-Re pair were determined on each standard. If the method is accurate the same value for the R factor will be obtained from both standards (effectively the trace element content of the second standard has been accurately determined). The values of the R factor of Ge bonded to oxygen against metallic bonded Re given in Table IV were made on two standards, the values marked with an being determined on a standard for which the Re/Ge concentration ratio was 5 times that of the standard used to obtain the values not so marked. The mean of the values marked with an a is 21.6: that of the values not so marked is 21.7. This excellent agreement suggests that the technique is accurate within the limits of the precision. More conventionally, geochemists test analytical techniques on two standard rocks, G . l . and W.1:. The granite G.l. possibly varies from batch to batch in certain trace element contents and was not used in this work. W.l. was analyzed for the elements whose R factors are given in Table 111. The technique described earlier was modified to the extent that a wider range of exposures was covered on one plate (thus larger exposure intervals) than we consider generally desirable, since a wide range of concentration levels could be expected. Our results are given in Table V, together with comparative data from the literature. The agreement justified the belief that the technique gives results as reliable as other techniques of trace element analysis.
Table V. Results of Mass Spectrographic Analysis of Values given in Range of values quoted by Fleischer Eleand Other results, method used, ment This work Stevens (9) and source of data As 2.38 1.8-2.53 2.24 neutron activation a 2.38 neutron activation (10) J3a
145
127-225
134-225 optical spectrograph, range of values quoted by Fleischer (11). 180-magnitude value quoted ( I I ) , since agreement between results by various methods is poor
70
24.3 neutron activation (12) 15.1 neutron activation (13)
ce
17.7
co
42
35-55
43 colorimetric (14) 53 x-ray fluorescencea 25-50 optical spectrograph, range of values (11)
Cr
98
92-165
113 x-ray fluorescencea 116.3 neutron activation. 110-144 optical spectrograph, range of values (11)
cs cu
0.95 110
0.45-1.1 91-153
DY
3.89
no data
3.3 neutron activation' 4.38 neutron activation (13)
Er
2.08
no data
2.57 neutron activation (12)
Eu
1.20
no data
1.29 neutron activation (13) 1.09 neutron activation (12) 1.12 neutron activation~
Ga
16
12-19
DISCUSSION
Various other points have been considered in our investigations which have not led us to make significant modifications in technique or procedure. Since these might be suspected of influencing the precision and accuracy of this method of analysis, they will be discussed briefly here. Selective Volatilization of Elements from Electrodes. Selective volatilization has to be considered in optical emission spectrography and the possibility of it cannot be ignored in considering spark source m a ~ spectrographic s techniques. Taylor ( 5 ) suggests that the internal standard element shouid possess similar voiatility to the elements sought, ?hough he describes tests indicating that selective distillation effects are not serious. With the method of line plotting tiescribed earlier in this paper, the line plots for elements selectively volatilized in the early stages of a run should make angles with the exposure axis lower than those for elements not so affected; elements retarded by selective volatilization of the charge should g v e line plots making angles with the axis higher than the remainder. Thus selective volatilization effects in any single run should be discernible through variation in the angies between line plots and the exposure axis related in a systematic way to the relative volat'lity of the different elements. We have no evidence of this in the Ziectrodes we nave used under the sparking conditions given 588
ANALYTICAL CHEMISTRY
109 colorimetric (14) 118 atomic absorption spectrophotometry (15) 110 x-ray fluorescencea 110 neutron activation"
16.5 neutron activation" 18.3 neutron activation" 16 recommended value (11) 4 . 2 neutron activation (12)
Gd
3.82
no data
Ge
1.6
1.6
Hf
0.93
no data
HO
0.63
no data
1.35 neutron activation (12) 0.855 neutron activation (13)
In
0.068
0.064-0.094
Comment: 0.064 is by neutron activation, 0.094 is by bined chemical - spectrographic technique 4 . 3 neutron activation (13) 11.7 neutron activation (12) 0.325 neutron activation (12) 0.35 neutron activation (13)
1.6 chemical (Id) 1.45 spectrographic (17) !.7neutron activation (18)
La LU
11.9 0.20
Nd
12.5
Ni
76
27-30 no data 50 55-88
1 5 . 1 neutron activation (12) 20.2 neutron activation (13) 79 colorimetric (14)
6485 x-ray Fluorescence, range of values (11) 60-82 optical spectrograph, range of values (11) 78 recommended value (11)
Geochemical Standard Rock W.l. with Literature Data ppm by weight Range of values quoted by Fleischer Other results, method used, and El& and source of data ment Thii work Stevens (9) 9 colorimetric (19) 5-10 8.3 Pb 10 x-ray fluorescence" 7.55 polarographic (20) 8 suggested value (11) pr
Rb
Sb
sc Sn Sr
4.15 21
0.85 43
2.1 175
Tb
0.49
Th
2.2
no data 22-28
0.95-1.2 25-120
2-8.7 170-310
no data 2.3-2.4
3.51 neutron activation (13) 3.68 neutron activation (12) 20 flame photometer (21) 18-22 optical spectrograph, range of values (11) 18-50 x-ray fluorescence, range of values (11) 0.96 neutron activation" 1.03 neutron activation (IO) 36.3 neutron activation0 34-49 optical spectrograph, range of values (11)
2.7 spectrophotometric (22) 3 recommended value (11) 151-169 flame photometer, range of values (11) 155-227 optical spectrograph, range of values (11) 156210 x-ray fluorescence, range of values (11) 0.75 neutron activation (12) 0.81 neutron activation (13) 2.2 neutron activation (23) 2.2 gamma ray spectrometry (24)
2.5 spectrophotometrica 2.6 alpha count (25) 3 x-ray fluorescencea Tm
0.28
U
0.46
no data
0.28-0.6
0.33 neutron activation (13) 0.355 neutron activation (12) 0.3 recommended value (11) 0.53 gamma ray spectrometry
earlier in this paper. However, we have been careful to adjust the pulse repetition rate and pulse length so that the dead time exceeds the live time by a factor of at least 5 to minimize the likelihood of overheating of the electrode tips. If this precaution is not complied with, selective volatilization may result. Slight evidence of it occurs with a pulse repetition rate of 1000/sec and a pulse length of 200 psec and we can not recommend speeding up the analysis during long exposures by increasing the pulse repetition rate beyond the rate of 300/sec unless, of course, maximum precision and accuracy are not of prime importance in the investigation. It is an advantage of this technique and method of line plotting that selective volatilization is indicated if it has occurred, even if unexpected, and the appropriate remedy (reducing the pulse repetition rate) can be adopted in a repeat run. Matrix Effects. While matrix effects, such as are encountered in optical emission spectrography would hardly be expected in this excitation source and matrix differences are not considered a major effect (9,consideration had been given to an effect that may be regarded as a kind of matrix effect. It was recognized that the value of the R factor for an element pair might be influenced by the nature of the bonding by which the elements are held in the compounds in which they occur. This effect was sought by preparing some charges in which Re was introduced as KRe04instead of as Re metal and determining values of the R factors for oxygen bonded elements against oxygen bonded Re. These values of the R factors were considerably lower than those quoted in Table 111-e.g., for oxygen bonded Cs against oxygen bonded Re a value of 11.1 was obtained compared with 87.7 for oxygen bonded Cs against metallic bonded Re. Clearly the internal standard element added to the unknowns must be in the same form as that added to the standards used for determining the values of the R factors used. A further consequence of this finding appeared to be more serious. Though metallic bonding is rare in rock-forming minerals, sulfides with essentially co-valent bonding are frequently minor constituents of rocks. Such sulfides are often hosts for trace elements in the rocks and for some elements, e.g. Cu, a large fraction of the element present in the rock may be so located. We failed to obtain a co-valently bonded compound of Re in a sufficient degree of purity to repeat our experiments with Re introduced in this
(24)
0.55 neutron activation (23) 0 . 9 i 0.6 alpha count (25) '4
260
120-320
224-340 optical spectrograph, range of values (11)
Y
26
20-35
23.8 neutron activation (12) 28.0 neutron activation (13) 21 x-ray fluorescence. 25 recommended value (11)
Zn
75
60-85
78 flame photometry" 91 atomic absorption spectrophotometry (15) 66-78 colorimetric, range of values (11) 59-80 x-ray fluorescence, range of valses (11) 82 recommended value (11)
Zr
a
90
89.5-1fB optical spectrograph, range of values (11) 95-109 x-ray fluorescence, range of values (11) Personal communication to M. Fleischer, quoted in Reference
(11).
55-100
(9) M. Fleischer and R. E. Stevens, Geochim. Cosmociiim. Acta,
26,525 (1962). (10) J. Esson, R. H. Stevens, and E. A. Vincent, Mitiemi. Maa. . 35, 88 (1965). f11) M. Fleischer. Geochim. Cosmochim. Acta. 29. 1263 (1965). i12j L. Haskin and M. A. Gehl, J . Geophys. Res:, 68, 2037 (1963). (13) D. G. Towell, R. Volfovsky, and J. W. Winchester, Geochim. Cosmochim. Acta, 29, 569 (1965). (14) R. E. Stanton, A. J. Macdonald, and I. Carmichael. Aiialyst
87, 134 (1962). (15) C. R. Belt, Jr., Ecorz. Geol., 59,240(1964). (16) L. H. Ahrens and M. Fleischer, Report on Trace Constituents in Gran t e G.1. and Diabase W.l., U.S. Geol. Suro. Bull. 1113,
83 (1960). (17) E. Schroll and M. Weininger, Mikrochim. Acta, 1965 (2-), 378. (18) D. F. C. Morris and J. S. P. Batchelor, Genchim. Connochim. Acta, 30,737 (1966). (19) R. R. Marshall and D. C . Hess, ANAL.CHEW, 32, 960 (1960). (20) V. V. Zhirova, Geokhiniiya, 1962,542. (21) C. 0. Ingamells, Tulonfa,9,781 (1962). (22) A. J. Macdonald acd R. E. Stanton, Aiialysr, 87,599 (1952). (23) J. W. Morgan and J. F. Lovering, Anal. Chim. Acta, 28, 405 (1963). (24) K. S. Heier and J. J. W. Rogers, Geochim. Cosmochim. Acta, 27, 137 (1963). (25) R. D. Cherry, Geochim. Cosmochim. Acta, 27, 183 (1963). VOL 39, NO. 6, MAY 1967
589
Table VI. Comparison of Determinations of As Contents in Pyrite Samples by Neutron Activation and by Solid Source Spark Mass Spectrography As content (wt ppm) By neutron By solid source activation spark mass Source of sample (R. Mitchell) spectrography Pyrite from schist, Dalemyr, Norway 480 470 Pyrite from slate Penryn, Cornwall, England 13.2 15.0 Pyrite from shale, Llangranog, Wales 2235 2065 Hydrothermal pyrite, 48.2 52 Elba, Italy Hydrothermal pyrite, 73.0 72 Cornwall, England
form and were forced to adopt a different method of checking the effect of location of an element in a sulfide on the R factor for that element against Re. In collaboration with R. Mitchell, we analyzed FeSz (pyrite) samples on which he had determined As by neutron activation. In view of the well known immiscibility relationship of 50 :50 mixtures of liquid sulfide and silicate we were unable to use the fusion procedure. Instead the Re was introduced into the electrode charges in the first of the various ways described earlier and, in consequence, we could not expect agreement better than i20-25 %. As contents for the analyzed sulfide samples were calculated using the value of the R factor for oxygen bonded As against metallic Re. If the R factor for As in the sulfide lattice against metallic Re differs significantly from that for oxygen bonded As against metallic Re, our results would be expected to show a bias one way or the other from the neutron activation results. No such bias was found, our results falling within the range +13.6 to - 2 l z of the neutron activation results and being evenly distributed between high and low values. Illustrative examples are giveri in Table VI. From this correspondence of results within the expected agreement range, we conclude that the R factor for As in sulfides is virtually the same as that for oxygen bonded As. Further work on this topic is clearly desirable, but the evidence at present suggests that bonding variation such as is likely to occur in geological samples does not constitute a source of error in our technique. However, it is also clear that our values for R factors should not be used in analyses of metals in metallurgical investigations, and the analyst must always consider the nature of the material to be analyzed in selecting the values of the R factors to be used. Failure to observe this elementary precaution could lead to serious error Generation of Doubly or Triply Charged Ions. The possibility of interference by doubly or triply charged ions had been mentioned by most writers on this technique. It exists, but can be dealt with by selection of the appropriate isotope of the element being sought or by correction techniques (5). More serious, in our view, is the danger of error in the determination of elements producing the doubly or triply charged ions. If an element has produced doubly or triply charged ions in the source, a further term is required in our basic equation, viz. the reciprocal of the fraction of the element content that has produced singly charged ions. During our
590
ANALYTICAL CHEMISTRY
early work investigating the purity of the compounds used for standard preparation etc. and in subsequent work on isotope ratio determination on galena (PbS), we have observed that the ratio of singly charged to doubly charged ions of a given element is very sensitive to electrode positioning, and it is difficult to reproduce this ratio on successive runs, confirming the findings of Jackson (26). This leads to the conclusion that no satisfactory solution to the problem can & found in introducing the additional term in our basic equation. Elements in trace amount-below 20 ppm atomic, i.e. (20 X element atomic weight/average atomic weight of sample) ppm by weight-do not appear to produce detectable doubly or triply charged ions. This does not seem to be a question of detection limits, for in the case of Pb in PbS the ratio of singly to doubly charged ions averaged about 6, so that doubly charged ions of Pb should be easily detectable, if produced, at Pb contents of 1 ppm. We simply have not observed doubly charged ion lines for elements in trace amount. Our solution to this problem rests on this finding though it is also influenced by the restriction of the range of concentration level recorded on one plate imposed by our procedure. For the determination of the more abundant trace elements, the proportion of Re-bearing flux to sample should be increased (using a Re-bearing flux of half the normal Re content), thus effectively diluting the trace element contents in the final electrodes. This practice was not followed during the determinations of trace element contents in W.1. reported in Table V and the figures given in that table for the contents of the more abundant elements (Ba, Cr, Cu, Ni, Sr, V, and Zn) may be slightly low in consequence. There are, of course, limits to the amount of dilution which is practicable, but this practice should extend the upper concentration level for accurate determination by our procedure to 100-150 ppm atomic. For still higher contents of the elements sought (minor rather that trace amounts) other techniques and procedures (26) would appear appropriate. We have not, at present, extended our investigations of the technique to these higher conentration levels. We do not propose to duscuss here the effects of changing sparking parameters. It is self-evident that sparking conditions should be held as constant as possible and to change pulse repetition rates etc. in the course of an analysis seems such an obvious error in technique as hardly to require discussion. For very short exposures, if required, a beam-chopping device (comparable in function though not in design to the stepped sector of optical emission spectrography) is clearly preferable to tampering with sparking parameters. Finally, we assert that careful investigation of suitable procedures of solid source spark mass spectrography removes the foundations of the early suspicions of this technique and justifies the belief that it is one of the most valuable analytical techniques available to a geochemist. RECEIVED for review December 5, 1966. Accepted February 10, 1967. We gratefully acknowledge support of this research program by a special N.E.R.C. research grant BJSRI 242. (26) P. F. S. Jackson, Proc. Sixth MS7 Users ConJ, A.E.I. Lrd., Manchester, 28 (1966).
