124 Analyst, February, 1967, Vol. 92, pp. 124-131 Comparative Elemental Analyses of a Standard Plant Material BY H. J. M. BOWEN (Chemistry Defiartment, The U?ziversity, Reading, Bevks.) Results are reported by 29 laboratories for the analytical determination of 40 elements in standard kale powder. Consistent results were obtained by more than one laboratory for Au, €3, Ba, Br, Ca, C1, Co, Cr, Fe, Ga, I, Mn, Mo, N, P, Rb, S, Sc and W. Small differences between results obtained by different techniques were found for Cu, K, Mg, Na, P, Se, Sr and Zn. Of these, the most significant were: that flame photometry gave high results for Na; that activation analysis without chemical separation was unreliable for determining K and Mg ; and that atomic-absorption spectrometry gave high results for Cu and Sr.Gross discrepancies were found in the results reported for Al, As, Hg, Ni and Ti. Similar anomalies have been reported in analyses of mammalian blood. Where possible, the precisions of different analytical techniques were compared, INTER-LABORATORY comparison of a standard material is a good method for assessing the accuracy and, incidentally, the precision of analytical techniques. This paper describes the results of elemental analyses of a standard consisting of dried kale powder, whose production has already been described.1 Quantitative results have been obtained for 40 elements by submitting batches of the powder to 29 laboratories listed in an Appendix. A statistical review of the results obtained reveals numerous inconsistencies, but shows that certain techniques give consistently high or low results.The results received for 19 elements are consistent enough for the sample to be used as a standard for assessing the accuracy of analytical methods for these elements in biological materials. OUTLINE OF METHOD OF COMPARISON Each laboratory was asked to determine as many elements as possible in the kale powder, and to report at least four replicate determinations together with the method used. The results were set out in the form of a table for each element studied, and any sets of results differing greatly from the mean were noted. It soon became clear that when a few grossly aberrant results were omitted, the differences between techniques were usually larger than the differences between individual laboratories in which the same technique was used.Wherever sufficient results were available, means and standard deviations were calculated for each technique, and Student’s t-test was applied to find the probability that the difference between means was caused by chance. The t-test was also used to reject a few aberrant results and to justify the pooling of results from different laboratories. Results were rejected when the probability that their mean and the grand mean were the same was less than 1 per cent. A grand mean and standard deviation were usually calculated for each element, but in some cases differences between techniques were so large that it seemed better to quote two or even three different means, where these means differ significantly by the t-test.The standard deviation calculated from the results of several laboratories, obtained by using a given technique, is a measure of the precision of that technique. The accuracy of a technique can only be assessed by its degree of consistency with other techniques, and where several means are quoted the accuracy of all techniques concerned is still in doubt. It is hoped that further work on this material will clear up these inconsistencies. AS only four or fewer results were reported for each element by most laboratories it was not possible to make a reliable comparison of inter-laboratory and intra-laboratory precisions. DRY I NG- The kale powder is slightly hygroscopic and contained about 5 per cent. of water when despatched to the laboratories. Different workers used different drying conditions, which may have led to errors of up to 1 per cent.in their reported results. The effect of dryingBOWEN 125 kale powder at 60", 75", 90" and 100" C is shown in Fig. 1. In future work with this material, it is recommended that an aliquot of the material be dried for 20 hours at 90" C for the deter- mination of dry weight. Hours in oven Fig. 1. Loss in weight by kale powder as a function of drying time CONVENTIONS USED- symbols to individual laboratories will not be given here. Each laboratory has been assigned a symbol (A-2, AA, etc.), but the code relating these Each technique is referred to by a three-letter symbol, as follows- act = activation analysis ata = atomic-absorption spectroscopy cat = catalytic technique col = colorimetry fla = flame photometry vol = volumetric analysis flu = fluorescence analysis gra = gravimetry pol = polarography spe = spectrometry tur = turbidimetry Means ( M ) , standard deviations ( S ) , and the number of determinations made ( N ) are All means are given in p.p.m.of dry kale powder. The results are summarised under individual elements which are in alphabetical order given in the form M + S ( N ) . RESULTS by symbol. Rejected results are enclosed in square brackets. Element Result Laboratory <Om45 T 0-50 5 0.10 (3) M Ag Technique act act A1 Results inconsistent- 7.35 -J= 1.05 (4) A act (y-ray spectrometry) 78.5 2.5 (2) S act (y-ray spectrometry) 6.4 & 0.7 (3) 0 col * 35.5 -J= 7.3 (16) A, B, C, E col 80.1 -J= 11.9 (9) c, p SPe As Results inconsistent- * 0.127 f 0.029 (16) A, U, BB 1.80 -J= 0.16 (4) 0 Au 0-00222 & 0.00055 (7) Y, cc act col act col col] B 50.9 4.7 (35) A, C, E, J, K, L, R, AX [21.1 (8) B [41.2 (5) P spel Ba 4.38 & 0.54 (12) A act c, p SPe * "Best" mean pending further work.126 Element Br Ca Cd c1 C O Cr CS CU DY F Fe Ga Hg I I r BOWEN : COMPARATIVE ELEMENTAL ANALYSES [Analyst, Vol.92 Result Laboratory Technique A,D, E, L, M, R,T,V, X,AA 24.3 & 1.6 (18) A, M, s, u act 41,400 -J= 2230 (41) 39,350 f 958 (4) M, T act 42,350 f 492 (16) n, L, v, x ata 41,720 f 910 (13) R, AA fla Significant differences between techniques, e.g.- 40,950 f 1050 (2) E col 39,700 f 433 (6) A, E vol The following low results were reported, but have not been used in computing these means- [31,150 (4) C flal [33,320 (5) P SPel 1.0 & 0.1 (4) Q p01 3330 & 1060 (21) A, M, s, T act R vol [29,600 (5) W ata] [25,800 (1) Z ata] 0-0562 & 0.0077 (22) 0.0520 f 0.0082 (8) A, u act A, E, I, K, L, N, U No significant differences between techniques, e.g.- 0.0586 f 0.0066 (14) E, I, K, L, N col The following high results were reported, but have not been used in com- puting means- 10.075 (4) C spel L2.0 (5) P spel 11.6 (5) w ata] 0.331 & 0.155 (13) A, u act C SPe w ata 0.0688 & 0.0071 (6) U act 4-81 & 0.735 (88) Small but significant differences were found between techniques- 4-12 f 0.731 (12) A, G, u act 5.25 f 0-479 (18) V, W, X, 2, AA ata col 20 laboratories 4-65 f 0.518 (44) 5.34 f 0-536 (5) EJ Q p01 5.39 f 1.104 (9) c, P sPe 117 (3) 0 col] L9.4 (5) R col] < 0-024 T act 5.55 0.57 (4) Q vol 119.5 & 19.5 (79) 123.2 f 14.2 (8) M, u act 118.2 & 11-7 (17) V, W, X, AA ata col [59 (1) z ata] C, D, E, I, J, K, L, N, V, AA 17 laboratories No significant differences were found between techniques- 121.0 f 19.4 (45) 111.2 f 32.7 (9) c, p sPe B, C, D, E, J, K, L, 0, R, V 0.045 & 0.020 (6) A, u act 0.150 f 0.008 (9) U, BB act Results inconsistent- 0.0122 f 0.0024 (4) Q col 0.0800 3 0.0234 (12) I, K, N cat <0.021 T actFebruary, 19671 OF A STANDARD PLANT MATERIAL 127 Element Result Laboratory Technique K 24,630 1218 (53) 11 laboratories Three results obtained by y-ray spectrometry after activation were low and have been rejected; otherwise the differences between techniques were not significant- [18,890 & 3318 (11) fif, s, T, act (y-ray spectrometry)] 24,160 f 1377 ( 5 ) x, ata 24,570 f 1222 (43) A , C , D , E , L , R,V,AA fla 25,620 & 427 ( 5 ) P SPe La 0.0767 & 0.0103 U act 1604 & 119 (43) 10 laboratories Results obtained by y-ray spectrometry after activation and by spectro- metry were low, and results obtained by flame photometry were high and have been rejected; otherwise the differences between techniques were not significant- Mg [1350 f 188 (6) S act (y-ray spectrometry)] D, L, V, W, X, Z, AA 1611 f 72 (31) ata 1533 f 142 (8) B, c col [3461 (5) R flal [1140 f 155 (5) P SPel 1700 & 271 (4) E vol Mn 14-9 5 1.8 (83) 20 laboratories A, F, G, M, S , T The differences between techniques were not significant- 14.7 f 1.34 (24) act 15.5 & 1.80 (23) V, W, X, Z, AA ata col 14.6 f 1.92 (36) [29 (4) [25 (5) [lo (5) D, E, G, I, J, K, L, N, V C c01; R col] P SPe! Mo 2-33 & 0.47 (44) 9 laboratories Two spectrometric results were lower and have been rejected ; otherwise there was no significant difference between techniques- 2.36 f 0.856 (9) A, u act 2.33 f 0.320 (35) A , C , E , I, K, L,N,AA col [0-996 2 0.497 (9) c, p spel N 43,102 & 102 (35) 9 laboratories No significant differences between techniques- 43,500 & 806 (2) M act vol 42,460 f 583 (5) P SPe 43,188 f 120 (28) B, C, D, E, L, AA Na Results inconsistent- Grand mean 2594 5 617 (52) 9 laboratories 2168 5 310 (15) M, S, T act 2318 f 332 ( 5 ) x, ata 2837 f 324 (32) A, C, R, AA fla [965 (2) W ata- [1220 f 45 (5) P spe; The results from flame photometry are significantly higher than those: obtained with other techniques .Ni Results inconsistent- 10.98 f 0.85 (4) X ata 2.65 & 1.56 (4) C sPe <1 A col P 4524 & 158 (38) B,C,D,E, L, R,V,AA col [4020 f 84 (5) P SPel Pb 3-21 & 1.61 (21) B, C, H, Q 1.6 (4) B col 3.8 ( 5 ) Q col 5-4 (4) H p01 3.0 (4) Q p01 2.1 (4) C sPe Precision poor ; individual laboratories found- 52.8 6.25 A, u act Rb128 Element S Sb s c Se Si Sn Sr Ti iv Zn BOWEN : COMPARATIVE ELEMENTAL ANALYSES [Analyst, Vol. Result Laboratory Technique 16010 2648 (21) B, AA col R gra C vol [24900 (4) C tur] 0.0653 & 0.0125 (6) U act 0.00835 0.00074 (4) A, M act 0.148 & 0.0137 (20) - Differences between techniques significant at 1 per cent. level- 0.155 f 0.0143 (12) A, u act 0.139 4 0.0040 (8) C, AA flu 242 & 10 (5) M act B col 0-160 & 0.037 (4) C SPe 84-1 & 10.7 (20) A, M, w, x Differences between techniques significant at 1 per cent.level- 74.7 & 4.2 (6) A, M act 88.1 f 10.2 (14) w, x ata [149*6 3.4 (5) P spel 0.330 & 0.050 (4) B col 2.75 f 0.13 (4) C sPe 0.0605 & 0.00123 (8) A, u act 31-88 & 4.82 (77) 32.27 & 1.88 (13) A, G, u act 34-23 & 1.57 (28) ata col Results inconsistent- 19 laboratories Some differences between techniques were significant- D, L, V, W, X, Z, AA 30.71 f 5.49 (24) 24.57 8.00 (7) E, R p01 33-60 & 0.89 (5) P sPe C, E, J, K, N, 0 The polarographic results were significantly low at the 1 per cent. lei with respect to the activation results. TABLE I RESULTS FOR MANGANESE IN KALE Laboratory Ashing method Technique A - Activation analysis F - Activation analysis G - Activation analysis S - Activation analysis T - Activation analysis 31 __ Activation analysis v \v x Z AA C n E G I K L 1; R v J - _ _ - - - Wet Dry Wet Wet Wet Wet Wet 4.z Dry i Atomic absorption Atomic absorption Atomic absorption Atomic absorption Atomic absorption Colorimetry - Colorimetry - Colorimetry (formaldoxime) Colorimetry (permanganate) Colorimetry (permanganate) Colorimetry - Colorimetry (permanganate) Colorimetry (perrnanganate) Colorimetry (perrnanganate) Colorimetry (permanganate) Colorimetry (permanganate) P - Spectrometry Manganese, p.p.rn.* 12.6, 13.4, 13.7 13-2, 13.6, 13.8, 14.1, 14.2, 14.5, 14.6, 14 13.1, 16-1 16.0, 16.9, 16.9, 1’7.8 13.9, 14.0, 14.4, 14.9, 15.2, 15.2 16.0 15.9, 17.2, 19.0, 19.8 13.4, 13.6, 14.0, 14-7, 14.7, 16.2 13.5, 13.9, 14-0, 14-0 15.0 14, 15, 15, 15, 16, 17, 17, 18 26, 26, 30, 34t 13.5, 13.5, 13.8, 13.8, 14.1, 14.8 12.0, 12.0, 12.2, 12.5 14.4, 14.8 15, 17, 17, 19 11, 11, 11, 13 15, 16, 16, 16 15, 15, 15, 15 16, 16, 17, 17 25, 25, 25, 25, 251- 15, 15, 15, 17 9, 9, 10, 11, l l t * A11 results are corrected to p.p.m.of dry weight. As these results are significantly different from the mean of all other results by the t-test, they were rejected as being probably erroneous. $ Method not known.February, 19671 OF A STANDARD PLANT MATERIAL 129 RESULTS FOR THE ELEMENT MANGANESE- The results for the element manganese are given in Table I to illustrate the extent of inter-laboratory and intra-laboratory variation in a typical case.There are no significant differences between techniques, except for the single result by spectrometry. The consistency or accuracy of the results can be summed up as follows. (1) Consistent results were obtained by more than one laboratory with the same technique for Au, B, Br, Ga, I, Rb, Sc and W. (2) Consistent results were obtained by different laboratories with several different techniques for Ba, Ca, C1, Co, Cr, Fe, Mn, Mo, N, P and S. No change in nitrogen content with storage, as reported for citrus leaves by Steyn,2 was found. (3) Small but significant differences between results were obtained with different techniques for Cu, K, Mg, Na, P, Se, Sr and Zn. (4) Gross differences between results were obtained with different techniques for Al, As, Hg, Ni and Ti.The large number of elements in categories (1) and (2) should be a source of satisfaction to the analysts concerned in this programme. It is difficult to account for the small differences in category (3) without detailed knowledge of all the techniques used. The best defined difference was found for sodium, where flame photometry gave significantly higher results than activation analysis or atomic-absorption spectrometry. For copper, activation analysis gave significantly lower results, and atomic-absorption spectrometry significantly higher results, than colorimetry. The differences between techniques for the other five elements in this group, although statistically significant, are felt to be based on too few experimental results to be more than guides to future work with other materials.They include the following provisional findings. K Activation analysis (y-ray spectrometry only) gives lower results than other tech- niques. Mg Activation analysis (y-ray spectrometry only) and spectrometry give lower results than atomic absorption and colorimetry. P Spectrometry gives lower results than colorimetry. Se Fluorimetry gives lower results than activation analysis. Sr Atomic absorption gives higher results than activation analysis. Zn Polarography gives lower results than other techniques. SUMMARY OF RESULTS For category (4) elements, further analyses may resolve some of the discrepancies. It appears likely that high values for aluminium and titanium may have arisen from contamination by dust, and nickel contamination might arise from handling the material with a spatula. Arsenic and mercury are both volatile elements which could readily distil into, and out of, the sample during heat treatment; the activation analysis results are probably the more reliable.COMPARISON OF PRECISIONS The precisions of two techniques can be compared by computing their variance ratio and applying the F-test. This has been carried out for a few elements, for which plenty of experimental results obtained by using more than one technique were received, with the following results. Ca Atomic absorption was more precise than flame photometry (P < 0.02). Co The precisions of activation analysis and colorimetry did not differ significantly (P > 0.05).Cu The precisions of activation analysis, atomic absorption, colorimetry and polaro- graphy did not differ significantly (P > 0.05). Fe Atomic absorption was more precise than colorimetry (P = 0.01) ; activation analysis was intermediate in precision. Mn Activation analysis was more precise than either colorimetry (P > 0.02) or atomic absorption (P < 0.05). Mo Colorimetry was much more precise than activation analysis (P < 0.01).130 BOWEN : COMPARATIVE ELEMENTAL ANALYSES [Analyst, Vol. 92 Na The precisions of activation analysis, atomic absorption and flame photometrg did not differ significantly (P > 0.05). Zn Both activation analysis and atomic absorption were much more precise thar colorimetry or polarography (P < 0.01).DISCUSSION There are comparatively few comparisons of this type reported for biological materials Ward and Heeney3 reported a similar study for calcium, potassium and magnesium in driec plant powders. They found flame photometry the best method for determining potassium but it was unsatisfactory for calcium and magnesium, for which titrimetric techniques werc much more precise. The Agricultural Research Council has published a Report of a grouj on Comparison of Methods of Analysis of Mineral Elements in Plants, in June, 1963. A compilation of values for the elementary composition of mammalian blood4 confirm: several of the findings from the present work. Thus, gross differences were found in value reported for each of the elements aluminium, arsenic, mercury, nickel and titanium in wholc blood by different workers.Activation analysis was found to give lower results for coppe in whole blood than either colorimetry or spectrometry. Flame-photometric results fo sodium in whole blood, but not those in serum or plasma, were somewhat higher than thosc found by other techniques. For phosphorus in blood, spectrometry gave lower results thar either activation analysis or colorimetry. There was no indication of significant difference between techniques used to determine potassium, magnesium or zinc in whole blood, an( there were too few results to test such differences for selenium and strontium. Cook, Crespi and Minczewski5 have carried out a valuable survey of the accuracy an( precision of techniques used to measure chromium, copper, mercury and manganese in ai artificial standard.They obtained consistent results for all four elements from nine labora tories by using activation analysis, colorimetry, polarography and spectrometry. In general the precisions of the first three techniques did not differ significantly, but spectrometry wa significantly less precise. The mercury content of Cook, Crespi and Minczewski’s standarc was 15 p.p.m., so that it was easier to determine than in the kale standard used here. The precision of analytical determinations on standard rocks appears to be much lowe than that attainable for biological standards, judging by the results compiled by Webber. The accuracy and precision for aluminium and titanium, which are present in large amounts were good, but gross differences were found by different workers for some trace constituents such as arsenic and nickel.It appears that flame photometry gives higher results thar spectrometry for sodium in Syenite rock. Appendix Analysts and laboratories (in alphabetical order) who supplied the results that mad Beardsley, D. A., Briscoe, G. B., RtiiiEka, J., and Williams, M., College of Advancec Bowen, H. J. M., Wantage Research Laboratories (A.E.R.E.), Wantage, Berkshire, an( Bradfield, E. G., Long Ashton Research Station, Bristol. Carson, R. B., Department of Agriculture, Ottawa, Ontario, Canada. Cawse, P. A., Wantage Research Laboratories (A.E.R.E.), Wantage, Berkshire. Collier, R. E., National Agricultural Advisory Service, Shadlow Hall, Derby. Cook, G. B., International Atomic Energy Agency, Vienna, Austria.Cuypers, J., and Wainderdi, R. E., Texas A & M University, College Station, Texac David, D. J., C.S.I.R.O., Canberra, Australia. Davies, E. B., and McNaught, K. J., Department of Agriculture, Hamilton, New Zealanc Fricker, D. J., National Agricultural Advisory Service, Ashford, Kent. Girardi, F., Euratom, Ispra, Italy. Heinerth, K. E., 28 Truchsess Strasse, Dusseldorf, Germany. Jackson, E., National Agricultural Advisory Service, Leeds, Yorks. Jewell, E. J., National Agricultural Advisory Service, Starcross, Exeter, Devon. Jones, J. B., Ohio Agricultural Experiment Station, Wooster, Ohio, U.S.A. it possible to write this paper are- Technology, Birmingham. The University, Reading, Berkshire. U.S.A.February, 19671 OF A STANDARD PLANT MATERIAL 131 Lane, J. C., Johnston Castle Agricultural College, Wexford, Ireland. Miettinen, J. K., and Puumala, H., University of Helsinki, Finland. Miller, S. T., and Cotzias, G. C., Brookhaven National Laboratory, New York, U.S.A. Morris, D. F. C., and Gupte, J. C., Brunel College of Technology, London. Pickett, E. E., University of Missouri, Columbia, Missouri, U.S.A. Samsahl, K., Aktiebolaget Atomenergi, Studsvik, Nykoping, Sweden, Sjostrand, B., and Westermark, T., Royal Institute of Technology, Stockholm, Sweden. Teichman, T., Department of National Health and Welfare, Ottawa 3, Ontario, Canada. Tyler, J. F. C., Ministry of Technology, Laboratory of the Government Chemist, Cornwall Ward, G. M., Department of Agriculture, Harrow, Ontario, Canada. Williams, A. M., National Agricultural Advisory Service, Cardiff, Glamorgan. Williams, T. R., National Agricultural Advisory Service, Coley Park, Reading, Berkshire. Yule, H. P., General Atomic, P.O. Box 608, San Diego, California, U.S.A. House, Stamford Street, London, S.E. 1. REFERENCES 1. 2. 3. 4. 5. 6. Bowen, H. J. M., in Shallis, P. W., Editor, “Proceedings of the SAC Conference, Nottingham, Steyn, W. J. A., J . Agric. Fd Chem., 1959, 7, 344. Ward, G. M., and Heeney, H. B., Can. J . PI. Sci., 1960, 40, 589. Bowen, H. J . M., “The Elementary Composition of Mammalian Blood,” U.K. Atomic Energy Research Establishment Report AERE-R 4196, H.M. Stationery Office, London, 1963. Cook, G. B., Crespi, M. B. A, and Minczewski, J., Talanta, 1963, 10, 917. Webber, G. R., Geochim. Cosmochim. Acta, 1965, 29, 229. 1965,” W. Heffer & Sons Ltd., Cambridge, 1965, p. 25. Received February 18th, 1966