- Email: [email protected]
The reporting of rock and mineral analyses by reference to analyses of standards
731
Notes phosphate crystals is obviously related to this problem. in progress ; the results will be reported elsewhere.
Such a study is presently
CoNcLUsIoNs
For the fifty phosphate samples from complex lithian-rich pegmatites analysed, rubidium concentrations vary from 7.5 ppb to 264 ppm and common strontium concentrations vary from 3.2 ppm to 4320 ppm. The Rb/Sr ratio varies from 4 x 10-s to 3.4. All pegmatites included in this study contain phosphates whose Srs7 content is anomalously high. Only two samples, both triphyllite with high Rb/Sr ratios, have less Sr*’ than would have been produced by in situ decay of Rb*‘. As the samples were randomly chosen from five continents and of widely varying geologic age, the phenomena noted here are considered to be characteristic of all zoned, lithium rich pegmatites. Further studies are needed to elucidate the mechanism which produced anomalous high SP/Sr8* ratios in phosphates from these environments. Acknowledg~elzte-This study was supportedby a National ScienceFoundation Grant #Ga-692.
E. F. (1966) Minor elements in igneous and metamorphic apatite. Geochks. Acta 30, 375-398.
&KUFT
Co.smochi?n.
Ceochlmicaet Coamoehimica Acta,1970,Vol. 34, pp. 731 Eo734. Pergamon Press. Printed in Northern Ireland
!Fhereporting of rock and mineral analyses by reference to analyses of stander PETER BLATTNER New Zealand D.S.I.R., Geological Survey, P.O. Box 30368, Lower Hutt, New Zealand (Received27 October 1969; accepted in revisedform 11 B’ebrwzq 1970) Abs~&c~-Rock and mineral analyses should always be reported together with data obtained for rock and mineral standarda analyzed with the unknowns. The problem of analytical aocuracy may then be eliminated from geochemioalcomparison. ‘Unlessthis is done, good analytical precision is lost from the point of view of such comparison.
IT IS an essential task of geochemistry
to recognize and to interpret historically the distribution of elements in space. Although many geochemical studies are concerned with relatively local problems, a greater si~cance of the results obtained may lie in the comparison of data from more than one locality. Unfortunately, such a comparison will normally be affected by analytical bias in each set of data, and the resulting relative bias can only be inferred from relative bias occurring in general. PRINZ
732
Notes
(1967) illustrates the procedure required, stating: “If interlaboratory variation is large the geochemistry of an element will be uncertain.” The common occurrence of bias in rock and mineral analysis has been demonstrated in the experiments with rock standards, initiated by FAIRBAIRN et al. (1951). Recent examples of relative bias between laboratories are given in FLAWAGAN(1967), and more may be drawn from FLANAGAN (1969), FLEISCHER (1969), SINE et al. (1969), and earlier summaries. Based on these data for rock standards, relative bias of 50 per cent of amounts reported at the 500 ppm level, or of 10 per cent of amounts reported at the 5 per cent level, may be anticipated for routine analyses. Table 1. Selected Rb data on rock standards (ppm) Analyst 1. 2. 3. 4. 5.
G-2
GSP-1
161 175 185 203 513
244 260 272 337 692
GOLDICHet al. (1967) CARMEHAELet al. (1968) TOMURAet al.* srrTxm* M.&m*
AGV-I
BCR-1
62 70 71 73 130
44 50 53 50 150
Method Conventional (recalc. from oxide) X-ray fluorescence Neutron activation Optical 1 spectrogrsphy
* (3)-(5) from FLANARAN (1969).
Examples of relative bias in reported concentrations are given for a trace and a major component in Tables 1 and 2. For Rb (Table 1) precise determinations suggest relative bias of less than 10 per cent of amounts reported for the laboratory pairs (l)-(2) and (2)-(3), w h ereas the difference between (5) and all other laboratories is of the order of 100 per cent, of the smaller values. For SiO, (Table 2) precise determinations suggest a small relative bias of O-5 per cent of amount reported between analysts (3) and (4), whereas a difference in bias of about 7 per cent appears between analysts (1) and (5). Comparisons between analysts, each involving four (Table 1) or six Table 2. Selected SiO, data on rock standards
(%)*
Analyst
G-2
GSP-1
AGV-1
BCR-1
PCC-1
IITS-
1. Sahores and Aubert 2. Lewis
67.41 68.3
64.73 65.0
57.17 67.2
52.00 63.8
38.20 41.1
;;
3. Lewis 4. Bouvior 5. Sine
69.0 69.45 70.12
66.9 67.45 67.36
58.7 68.84 59.99
54.5 64.80 55.12
41.5 41.77 43.47
* From
40.62 42.00
Method )X-ray
fluorescenoe
Photometric Gravimetric X-ray fluorescence
FLANAGAN(1969).
pairs of values, i.e. one for each standard, may be subjected to Student’s t-test. For 19 out of the 20 possible comparisons mean differences are significant at the 10 per cent level or less. The exception is the comparison (3)-(4) of Table 1. Differences in sensitivity seem to be a major source of relative bias, and bias is well recognizable against effects of replication error as well as of between-bottle variations. In general, CHAYES (1969) and FLANAGAN (1969) have found little evidence for between-bottle variations in the composition of rock standards, although further tests seem desirable. Whereas in some cases analytical bias may be traced to errors of technique and to uncontrolled “drift,” in others it may indicate unresolved analytical problems and may become the subject of prolonged argument. However, it is not surprising that we are unable to determine the composition of rocks accurately. The more significant
733
Notes
fact is that our precision is frequently better than our accuracy. In this situation rock standards, such as those issued by the U.S. Geological Survey, offer a way to record analytical bias for a field of physical and chemical rock properties. Independently of true magnitudes of bias, they allow dejhition of concentration scales for individual laboratories, procedures, analysts, etc. For some techniques it has been recognized early that unknowns must be matched to standards with similar properties (AHRENS, 1954; “golden rule”). However, analysis by reference also calls for reporting by reference. This principle has been strictly followed, for example, in the isotopic analysis of carbon and oxygen, with the Nier-McKinney type massspectrometer (CRAIG, 1957), but not in elemental analysis. Table 3. Reporting of data on standards and unknowns. example, ppm f standrtrd devietion 520 f 30
Ul
Unknowns
%’
. . . %I 51 Standards
For
. . . . . 580 f 30
s29 . %
.
. .
.
.
.
.
A growing awareness of the situation is reflected in the number of different ways in which rock and mineral analyses are reported in the current literature. Frequently, plain concentration values are still reported (e.g. CURTIS, 1969; stated precision 10 per cent). Sometimes it is merely said that certain rock standards have been used “for calibration” (SIEOERS et al., 1969; stated precision 5 per cent), or to “control the accuracy” (LSPET, 1969; stated precision 10 per cent). There is also an increasing number of reports that quote values obtained for rock standards, and MAXWELL (1968) has suggested that this be done. However, unless comparative data on standards are quoted routinely, much of the analytical precision, obtained at considerable expense, continues to be lost from a broader geochemical point of view. The case therefore needs to be re-stated. It is proposed here that all quantitative rock and mineral analyses be reported together with results obtained for rock and mineral standards, analyzed frequently enough to define a concentration scale within certain limits. A single array (Table 3) best expresses the tie between standards and unknowns in analysis. Standards referred to should obviously be internationally available, although in the actual analysis they may have to be replaced by well intercalibrated working standards. Using sl, s2, etc. from such arrays as fixed points, we can immediately compare data from different sources. Reporting in this or an equivalent form opens the way for evaluation of analyses in terms of accuracy, and at the same time allows a clear separation to be made between the problem of accuracy and that of geochemical When data on standards and unknowns are recalculated to correspond comparison. concentration scale, ambiguity may be avoided, if both are still to a “preferred” reported.
734
Notes
Ac~owl~~~~~~. L. GOGUEL,J. A. MAXWELL and J. A. RITCHIE are thanked for criticism leading to significant improvements of the text. Responsibility for remaining faults and shortcomings rests with the author. REEERENCES L. H. (1954) Quantitative Speotrochemicul Andy.& of Silicatea. Pergemon Press. CAREZICHAEL I. S. E., HAMPEL J. and JACK R. N. (1968) Analytioal data on the U.S.G.S. standard rocks. Chem. Qeol. 8, 59-64. &AYES F. (1969) A last look at G-l, W-l. Carrbegie In&. Wash. Yearb. Ann. Rep. Dir. Geophgs. Lab. 67,239~241. CRAIQ H. (1957) Isotopic standards for carbon and oxygen and correction factors for masssp~trometrie analysis of carbon dioxide. Geocham. Co~~och~rn. Acta 12, 133-149. CURTIS C. D. (1969) Trace element ~stribution in some British C~boniferous sediments. Geochim. Cosmochim. A&z 33, 519-523. FAIRBAIRN H, W., et al. (1951) A cooperative investigation of precision and accuracy in chemical, spectrochemical and modal analysis of silicate rocks. U.S. Geol. Survey Bull. 980, 1-71. FLANAGANF. J. (1967) U.S. Geological Survey silicate rook standards. Geochim. Cosnaochim. Acta 31, 289-308. FLANAGANF. J. (1969) U.S. Geological Survey standards-II. First compilation of data for the new U.S.G.S. rocks. Geoohim. Cosmochim. Acta 33, 01-120. FLEISCHER M. (1969) U.S. Geological Survey standards-I. Additional data on rocks G-1 and W-l, 1965-1967. Geochim. Cosmoehim. Acta 33, 65-79. GOLUICHS. S., IN~AMELLSC. O., SUHRN. H. and ANDERSOND. H. (1967) Analyses of silicate rock and mineral standards. Cam.J. Earth Sci. 4, 747-755. LSPET (LLJAR SAXFIB PBELIBfzNARY EXAllrINATlON TEAM) (1969) Preliminary examination of lunw samples from ApoIlo 11. S&ewe 165, 121l-1227. MAX~VELLJ. A. (1968) Rock ad Miraed Andgeis. Interscience. A Tmatise on PRXNZ M. (1967) Geochemistry of basaltic rooks: Trace e1ement.s. In Badts, Rocka of Basalt& Compodtion {editorsH. H. Hess and A. Poldervaart), Vol. 1, pp. 271-323. Interscience. SIEQERSA., PICHLERH. and ZEIL W. (1969) Trace element abundances in the “Andesite” formation of Northern Chile. Geochim. Coemochim. Acta 33, 882-887. SINE N. M., TAYLORW. O., WEBBERG. R. and LEWIBC. L. (1969) Third report of analytical data for CAAS sulphide ore and syenite rock standards. Geochim. Coemochim. Acta 83, 121131.
AHRENS
Notes phosphate crystals is obviously related to this problem. in progress ; the results will be reported elsewhere.
Such a study is presently
CoNcLUsIoNs
For the fifty phosphate samples from complex lithian-rich pegmatites analysed, rubidium concentrations vary from 7.5 ppb to 264 ppm and common strontium concentrations vary from 3.2 ppm to 4320 ppm. The Rb/Sr ratio varies from 4 x 10-s to 3.4. All pegmatites included in this study contain phosphates whose Srs7 content is anomalously high. Only two samples, both triphyllite with high Rb/Sr ratios, have less Sr*’ than would have been produced by in situ decay of Rb*‘. As the samples were randomly chosen from five continents and of widely varying geologic age, the phenomena noted here are considered to be characteristic of all zoned, lithium rich pegmatites. Further studies are needed to elucidate the mechanism which produced anomalous high SP/Sr8* ratios in phosphates from these environments. Acknowledg~elzte-This study was supportedby a National ScienceFoundation Grant #Ga-692.
E. F. (1966) Minor elements in igneous and metamorphic apatite. Geochks. Acta 30, 375-398.
&KUFT
Co.smochi?n.
Ceochlmicaet Coamoehimica Acta,1970,Vol. 34, pp. 731 Eo734. Pergamon Press. Printed in Northern Ireland
!Fhereporting of rock and mineral analyses by reference to analyses of stander PETER BLATTNER New Zealand D.S.I.R., Geological Survey, P.O. Box 30368, Lower Hutt, New Zealand (Received27 October 1969; accepted in revisedform 11 B’ebrwzq 1970) Abs~&c~-Rock and mineral analyses should always be reported together with data obtained for rock and mineral standarda analyzed with the unknowns. The problem of analytical aocuracy may then be eliminated from geochemioalcomparison. ‘Unlessthis is done, good analytical precision is lost from the point of view of such comparison.
IT IS an essential task of geochemistry
to recognize and to interpret historically the distribution of elements in space. Although many geochemical studies are concerned with relatively local problems, a greater si~cance of the results obtained may lie in the comparison of data from more than one locality. Unfortunately, such a comparison will normally be affected by analytical bias in each set of data, and the resulting relative bias can only be inferred from relative bias occurring in general. PRINZ
732
Notes
(1967) illustrates the procedure required, stating: “If interlaboratory variation is large the geochemistry of an element will be uncertain.” The common occurrence of bias in rock and mineral analysis has been demonstrated in the experiments with rock standards, initiated by FAIRBAIRN et al. (1951). Recent examples of relative bias between laboratories are given in FLAWAGAN(1967), and more may be drawn from FLANAGAN (1969), FLEISCHER (1969), SINE et al. (1969), and earlier summaries. Based on these data for rock standards, relative bias of 50 per cent of amounts reported at the 500 ppm level, or of 10 per cent of amounts reported at the 5 per cent level, may be anticipated for routine analyses. Table 1. Selected Rb data on rock standards (ppm) Analyst 1. 2. 3. 4. 5.
G-2
GSP-1
161 175 185 203 513
244 260 272 337 692
GOLDICHet al. (1967) CARMEHAELet al. (1968) TOMURAet al.* srrTxm* M.&m*
AGV-I
BCR-1
62 70 71 73 130
44 50 53 50 150
Method Conventional (recalc. from oxide) X-ray fluorescence Neutron activation Optical 1 spectrogrsphy
* (3)-(5) from FLANARAN (1969).
Examples of relative bias in reported concentrations are given for a trace and a major component in Tables 1 and 2. For Rb (Table 1) precise determinations suggest relative bias of less than 10 per cent of amounts reported for the laboratory pairs (l)-(2) and (2)-(3), w h ereas the difference between (5) and all other laboratories is of the order of 100 per cent, of the smaller values. For SiO, (Table 2) precise determinations suggest a small relative bias of O-5 per cent of amount reported between analysts (3) and (4), whereas a difference in bias of about 7 per cent appears between analysts (1) and (5). Comparisons between analysts, each involving four (Table 1) or six Table 2. Selected SiO, data on rock standards
(%)*
Analyst
G-2
GSP-1
AGV-1
BCR-1
PCC-1
IITS-
1. Sahores and Aubert 2. Lewis
67.41 68.3
64.73 65.0
57.17 67.2
52.00 63.8
38.20 41.1
;;
3. Lewis 4. Bouvior 5. Sine
69.0 69.45 70.12
66.9 67.45 67.36
58.7 68.84 59.99
54.5 64.80 55.12
41.5 41.77 43.47
* From
40.62 42.00
Method )X-ray
fluorescenoe
Photometric Gravimetric X-ray fluorescence
FLANAGAN(1969).
pairs of values, i.e. one for each standard, may be subjected to Student’s t-test. For 19 out of the 20 possible comparisons mean differences are significant at the 10 per cent level or less. The exception is the comparison (3)-(4) of Table 1. Differences in sensitivity seem to be a major source of relative bias, and bias is well recognizable against effects of replication error as well as of between-bottle variations. In general, CHAYES (1969) and FLANAGAN (1969) have found little evidence for between-bottle variations in the composition of rock standards, although further tests seem desirable. Whereas in some cases analytical bias may be traced to errors of technique and to uncontrolled “drift,” in others it may indicate unresolved analytical problems and may become the subject of prolonged argument. However, it is not surprising that we are unable to determine the composition of rocks accurately. The more significant
733
Notes
fact is that our precision is frequently better than our accuracy. In this situation rock standards, such as those issued by the U.S. Geological Survey, offer a way to record analytical bias for a field of physical and chemical rock properties. Independently of true magnitudes of bias, they allow dejhition of concentration scales for individual laboratories, procedures, analysts, etc. For some techniques it has been recognized early that unknowns must be matched to standards with similar properties (AHRENS, 1954; “golden rule”). However, analysis by reference also calls for reporting by reference. This principle has been strictly followed, for example, in the isotopic analysis of carbon and oxygen, with the Nier-McKinney type massspectrometer (CRAIG, 1957), but not in elemental analysis. Table 3. Reporting of data on standards and unknowns. example, ppm f standrtrd devietion 520 f 30
Ul
Unknowns
%’
. . . %I 51 Standards
For
. . . . . 580 f 30
s29 . %
.
. .
.
.
.
.
A growing awareness of the situation is reflected in the number of different ways in which rock and mineral analyses are reported in the current literature. Frequently, plain concentration values are still reported (e.g. CURTIS, 1969; stated precision 10 per cent). Sometimes it is merely said that certain rock standards have been used “for calibration” (SIEOERS et al., 1969; stated precision 5 per cent), or to “control the accuracy” (LSPET, 1969; stated precision 10 per cent). There is also an increasing number of reports that quote values obtained for rock standards, and MAXWELL (1968) has suggested that this be done. However, unless comparative data on standards are quoted routinely, much of the analytical precision, obtained at considerable expense, continues to be lost from a broader geochemical point of view. The case therefore needs to be re-stated. It is proposed here that all quantitative rock and mineral analyses be reported together with results obtained for rock and mineral standards, analyzed frequently enough to define a concentration scale within certain limits. A single array (Table 3) best expresses the tie between standards and unknowns in analysis. Standards referred to should obviously be internationally available, although in the actual analysis they may have to be replaced by well intercalibrated working standards. Using sl, s2, etc. from such arrays as fixed points, we can immediately compare data from different sources. Reporting in this or an equivalent form opens the way for evaluation of analyses in terms of accuracy, and at the same time allows a clear separation to be made between the problem of accuracy and that of geochemical When data on standards and unknowns are recalculated to correspond comparison. concentration scale, ambiguity may be avoided, if both are still to a “preferred” reported.
734
Notes
Ac~owl~~~~~~. L. GOGUEL,J. A. MAXWELL and J. A. RITCHIE are thanked for criticism leading to significant improvements of the text. Responsibility for remaining faults and shortcomings rests with the author. REEERENCES L. H. (1954) Quantitative Speotrochemicul Andy.& of Silicatea. Pergemon Press. CAREZICHAEL I. S. E., HAMPEL J. and JACK R. N. (1968) Analytioal data on the U.S.G.S. standard rocks. Chem. Qeol. 8, 59-64. &AYES F. (1969) A last look at G-l, W-l. Carrbegie In&. Wash. Yearb. Ann. Rep. Dir. Geophgs. Lab. 67,239~241. CRAIQ H. (1957) Isotopic standards for carbon and oxygen and correction factors for masssp~trometrie analysis of carbon dioxide. Geocham. Co~~och~rn. Acta 12, 133-149. CURTIS C. D. (1969) Trace element ~stribution in some British C~boniferous sediments. Geochim. Cosmochim. A&z 33, 519-523. FAIRBAIRN H, W., et al. (1951) A cooperative investigation of precision and accuracy in chemical, spectrochemical and modal analysis of silicate rocks. U.S. Geol. Survey Bull. 980, 1-71. FLANAGANF. J. (1967) U.S. Geological Survey silicate rook standards. Geochim. Cosnaochim. Acta 31, 289-308. FLANAGANF. J. (1969) U.S. Geological Survey standards-II. First compilation of data for the new U.S.G.S. rocks. Geoohim. Cosmochim. Acta 33, 01-120. FLEISCHER M. (1969) U.S. Geological Survey standards-I. Additional data on rocks G-1 and W-l, 1965-1967. Geochim. Cosmoehim. Acta 33, 65-79. GOLUICHS. S., IN~AMELLSC. O., SUHRN. H. and ANDERSOND. H. (1967) Analyses of silicate rock and mineral standards. Cam.J. Earth Sci. 4, 747-755. LSPET (LLJAR SAXFIB PBELIBfzNARY EXAllrINATlON TEAM) (1969) Preliminary examination of lunw samples from ApoIlo 11. S&ewe 165, 121l-1227. MAX~VELLJ. A. (1968) Rock ad Miraed Andgeis. Interscience. A Tmatise on PRXNZ M. (1967) Geochemistry of basaltic rooks: Trace e1ement.s. In Badts, Rocka of Basalt& Compodtion {editorsH. H. Hess and A. Poldervaart), Vol. 1, pp. 271-323. Interscience. SIEQERSA., PICHLERH. and ZEIL W. (1969) Trace element abundances in the “Andesite” formation of Northern Chile. Geochim. Coemochim. Acta 33, 882-887. SINE N. M., TAYLORW. O., WEBBERG. R. and LEWIBC. L. (1969) Third report of analytical data for CAAS sulphide ore and syenite rock standards. Geochim. Coemochim. Acta 83, 121131.
AHRENS