Transferring accuracy to the trace level and then to the field · 072/8 072/9 072/10 072/I1 072/12 072/13 072/14 072/15 Uranyl-nitrate ... 0.992 12 0.992 56 0.992 99 0.993 10 0.993

Volume 93, Number 3, May-June 1988

Journal of Research of the National Bureau of Standards

Accuracy in Trace Analysis

[5] Piccone, T. J., Butrymowicz, D. B., Newbury, D. E., Man-nig, J. R., and Cahn, J. W., Scripta Met., 839 (1982).

[6] Myklebust, R. L., Newbury, D. E., Marinenko, R. B., andBright, D. S., Background Correction in Electron Mi-croprobe Compositional Mapping with Wavelength-Disper-sive X-ray Spectrometry, Microbeam Analysis, SanFrancisco Press (1987) 25.

[7] Garruto, R. M., Swyt, C., Fiori, C. E., Yanagihara, R., andGajdusek, D. D., Lancet, 1353 (1985).

Transferring Accuracy to the TraceLevel and Then to the Field

Figure 2. Quantitative compositional maps of grain boundaryregions in polycrystalline silver-gold alloy: a) distribution of sil-ver, with contrast enhancement to show an enrichment of 5weight percent silver against a background of 65 weight percentsilver; b) corresponding gold image, showing the deficiency ofgold in the grain boundary region. Image field width =50micrometers.

References

[I] Duncumb, P., X-ray Microscopy and Microradiography, ed.by V. E. Cosslett, A. Engstrom, and H. H. Pattee, Aca-demic Press (1957) 435.

[2] Marinenko, R. B., Mykiebust, R. L., Bright, D. S., and New-bury, D. E., J. Microsc. 145, 207 (1987).

[3] Goldstein, J. I., Newbury, D. E., Echlin, P., Joy, D. C.,Fiori, C. E., and Lifshin, E., Scanning Electron Microscopyand X-ray Microanalysis, Plenum Press, New York (1981)305.

[4] Myklebust, R. L., Newbury, D. E., Marinenko, R. B., andBright, D. S., Defocus Modeling for Compositional Map-ping with Wavelength-Dispersive X-ray Spectrometry, Mi-crobeam Analysis, San Francisco Press (1986) 495.

Paul De Bievrel

Central Bureau of Nuclear MeasurementsCommission of the European Communities-JRC

B-2440 Geel, Belgium

The problem of arriving at an "accurate" deter-mination in trace analysis-determining a "true"small amount in a large amount-requires the as-sessment of the performance of the measurementmethod used to an appropriate detail.

In the case of "direct" measurement techniquesof trace concentrations, i.e., assaying directly asmall amount relative to the large matrix sample,the evaluation should include the assessment of thereproducibility of the linear, or non-linear responseof the measurement chain and of the high sensitiv-ity detector over many orders of magnitude downto small concentrations. (See example in fig. 1.) Toallow this assessment, a set of synthetic isotopemixtures can be prepared gravimetrically and usedto examine the measurement chain and detector.An example of such a set is given in table 1. Itspreparation is described elsewhere [1]. The built-inratio 23 5U/2.. Uz 1 (certified to 0.02%) can be usedto determine any systematic errors in the "excita-tion" or "ion" source and correct the observed233u/235U or 233U/23

'U ratios. Comparison of the lat-ter to the certified values (given to ±0.03%) thenprovides a means to determine systematic errors ofthe (sensitive) detectors used in the trace analysis(e.g., electron multipliers) and to determine possi-ble deviations from linearity of the (electronic)measurement equipment used. It is also possible todetermine the reproducibility of these systematic

' Professor at the University of Antwerp.

520




Table 1. Set of (isotopic) reference materials to test the linearity of a detector/amplifier measurementsystem

Atomic isotope ratios

Code 233U/35U 35u/3&unumber Description ±0.03% of value 0.000 30 Unit

CBNM-IRM-072/1072/2072/3072/4072/5072/6072/7072/8072/9

072/10072/I1072/12072/13072/14072/15

Uranyl-nitratesolution

1.000 330.699 670.499 850.299 870.100010.050 0910.019 9940.010 1650.005 000 00.002 001 20.000 968 920.000 500 880.000 101 820.000 019 9960.000 001 999 5

0.991030.991680.992 120.992 560.992 990.993 100.993 170.993 190.993 200.993 210.993 210.993 210.993 210.993 210.993 21

I mg U in I gsolution

errors and/or deviations allowing to correct forthem. It is important to point out that the uncer-tainties of such corrections must be carried on andcorrectly propagated until the final result in anymeasurement of unknowns.

We now have a means for specialized measure-ment laboratories to verify how their detector andmeasuring equipment is really performing at (very)low signals, i.e., at the trace element level and afirst example obtained is given in figure 2.

In the case of measurement techniques which as-say small amounts in weighed samples, the "jump"to the trace level is performed through weighing(by using small weights) and the problem becomesone of assaying a trace amount with a known,proven, or provable accuracy.

Every indication is there that this will have tocome from isotope-specific-or should I say nu-clide-specific?-methods on the basis of the factthat isotopes are very specific representatives of el-ements and specificity is an essential key to accu-racy.

We have at present a few nuclide-specific assaymethods in our array of analytical assay methods:

* Those which use the property of one nuclideas in a specific nuclear reaction ("activation analy-sis"); the degree of specificity will influence heav-ily the potential for accuracy; the degree ofquantitative knowledge of every parameter con-cerned will determine the size of the total "inaccu-racy."

* Those which modify an isotope (nuclide)abundance ratio R. in the sample ("isotope dilu-tion"), by the addition of a "spike" which is aknown number of atoms of the same element as theunknown but with another RY value for the abun-dance ratio of the same isotopes (an "enriched" sta-ble isotope- or "nuclide"); this spike acts as analmost ideal "internal standard" and the measure-ment is reduced to the determination of the newisotope abundance ratio RB in the blend resultingfrom the change induced by Ry in R.. Of course,proper isotopic homogenization and complete de-struction of the matrix (this makes the assay ma-trix-independent) in a closed system, is anabsolute prerequisite for this type of assay.

Isotope- or nuclide-specific methods do have in-herently more potential for accuracy since theymeasure isotopes and not elements. Isotopic atomsare measured on the basis of properties of the atomnucleus (e.g., difference in mass) which areshielded from chemical interferences before or dur-ing the measurement, by identical outer electronclouds. They can therefore potentially lead togreater accuracy (when the isotopic measurementmethods are correctly applied of course). It is to beexpected that isotope- or nuclide-specific methodswill more and more deliver reference values andmethods for elemental trace analysis in the future.

Thus the measurement of an element concentra-tion ratio (= conventional analytical chemistry) isreplaced by the measurement of isotope abundance

521




ratio. The essential step of isotopic homogenizationcan be guaranteed when properly carried out, be-cause it takes advantage of the fact that the outerelectron clouds (i.e., the chemical form) of bothsample and spike isotopic atoms are identical. Thisensures that both will behave in the same way andend up in the same chemical forms after chemicaldestruction of the sample. At the moment of fullisotopic homogenization, the end result is in fact"frozen in." Any chemical effect in a later stage ofthe process, e.g., losses in normal purification orseparation stages, will therefore affect equally theisotopes of sample and spike in the same way andleave their abundance ratio-the end result-un-changed.

The "accurate" values obtained as explainedabove for trace elements in unknown samples, cannow be carried to the field by Interlaboratory Mea-surement Evaluation Programmes (IMEPs) wheremeasurements on these samples are collected andgraphically displayed around the "referencevalue." One could also call such a programme an"external" measurement evaluation.

Current work at the Central Bureau for NuclearMeasurements (CBNM) aims at following theroute described above to establish reference valuesfor IMEPs. An example of what this yields is givenin figure 3, taken from the nuclear field: even aftersuitable calibration with available Reference Mate-rials, a picture develops showing considerablylarger spread than what laboratories believe to betheir uncertainty. It is a surprisingly classical pic-ture which can be produced for any assay measure-ment in any field at any concentration level. Onlythe ordinates and the title of the picture must bechanged, the general nature of the picture remainsthe same (see an example in fig. 4). In other words,the concept is expected also to be applicable to thetrace analysis field.

The question arises where such "referencevalue" or "baseline" takes its authority from andwhy. Since it is very difficult to unequivocally es-tablish these baseline values, this should be done byspecialized institutes or laboratories (the "stan-dards" laboratories in the world) along the follow-ing lines:

(a) It must be absolutely transparent how thevalue intended to be the "closest approximation ofthe true value" (i.e., intended to be accurate), isarrived at, based on a detailed explanation of eachstep in the characterization process which leads

from our bank SI units to the value pretending tobe able to serve as "reference," and

(b) It must be absolutely transparent how the"uncertainty" or "inaccuracy" of the value is ar-rived at, based on the establishment of a list of un-certainty contributors which are identifiedqualitatively and individually quantified by variousappropriate procedures so that their componentcontribution to the total "inaccuracy" is clearlyvisible; a work model for that is given in table 2.

Table 2. Proposed Work ModaL to ... btisba "Composition of Uncertainty

a) repeatablity of seasurementespressoed as 2s

(3 • 5 • 10)

b) repeatabililty of correction actorsexpressed as 2s

(6 n I • 10)

1st correction factor2nd correction facror

nth cotterctbar factor

subtotal = 4(1st)2

+ (2nd)' + (rbh)2

c) isrimate of possible (as yet) unknown5yst.. itic errors

on a 2s basis

lot

2nd

nth

subtotal (lst) I (2nd)' + (.I,)'TOTAL

Note that such IMEP programmes are result-oriented and not method- or procedure-oriented. Inthe graphs presenting the results, distinction isclearly made between:

* The state of the practice (SOP) which is ma-terialized in the spread of the results and in the(non-)deviation from the established referencevalues.

522




* Same form of the state of the art (SOA)which is materialized in the uncertainty level of the"reference values" which must be demonstrated tobe reliable enough and usable for such "external"QA programmes.

Conclusions

1. It has been shown how the response of mea-suring equipment can be evaluated for (non-)linearity over many orders of magnitude, this possi-ble error contributor being separated from the con-tributors located in the "source" or "excitation"

10T

onW

to-s

10-7

10tb

part of the measurement method.2. Isotope- or nuclide-specific methods have a

great potential to serve "reference" purposes sobadly needed in trace analysis.

3. IMEPs with carefully established baselinesshould carry references to the field.

4. "Standards-" or "Reference-" Institutesshould consider it as part of their basic mission todeliver regularly "baseline-values" for real lifesamples which are circulated as "blind" samplesamongst the laboratories wanting to assess theirtrue measurement capability through an "external"evaluation programme.

Figure 1. Example of dynamic range of trace concentrations: impuritiesin Pt metal (Spark Source Mass Spectrometry).

523




-10C

-1

-1

,-lo-'

-2

10 1 10o- 'V 2 io_I . . . I I . . i

I10- 10-5 10-5

233U/235u RATIO

Figure 2. Test of the response of a mass spectrometer as function of the ratio to be measured.

CERTIFIEO VRLUE- 7.193

T

.j- b a18 (PG PUJG SOL.)

e� +0~~ I

' O - Y 5X Te

2 .3 t 12 14 tO t 0 2 2 2 6 3

Lo_

60

- tI'

-2Cita

Figure 3. Inierlaboratory measurement evaluation programme on Pu assay amongst 31 laboratories(Reference Value: CBNM by IDMS).

524

-1

1 -

zjo-22

00 i

0=

LtO~ 1

2

Of

w-j

C

ixU-

n

a ¾?m-

00

,p n_IC -ILEI _a -zrw a.

E -

r ws

o rn -

7. 3800

7.27e0

7.240 a

'7.2 1a0

7. 5800

7. 1500

7. 1200

7.0s00

7.0600

7.6300

7 .ee2eLEIcAO.

o It" EsFI~errc sfo LA leS CODE NUMtERS OF LR80RRTORtES




1 70

LE,

II

I

5I

I

I

I

I

I

I

160 -

1.50-

1 40 -

REFERENCE VALUE (1I514 5 0 0201 trot 'L=' 1(25

(CNM- InIMS)

1.3n-

120 -

IlO.i

I.00 -

090-

*o

- 4-2

-2-4

K10LiI

-20

30

30 2

-401 2 3 4 5 6

LAB NR

Figure 4. Interlaboratory measurement evaluation programmeon Li assay amongst 61 laboratories (Reference Value: CBNMby IDMS).

References

(1] Lycke, W., De Bikvre, P., Verbruggen, A., Hendrickx, F.,and Rosman, K. J. R., "Isotopic Reference Material Set forTesting the Linearity of Isotope Mass Spectrometers,"CBNM IRM 072, 1987; Proc. EURO ANALYSIS VI, Sep-tember 1987, Paris (to be published).

525

>

oo

Documents

Transferring accuracy to the trace level and then to the field · 072/8 072/9 072/10 072/I1 072/12 072/13 072/14 072/15 Uranyl-nitrate ... 0.992 12 0.992 56 0.992 99 0.993 10 0.993