400

Thermostating the column is therefore necessary, if shifts in temperature cannot be excluded in the laboratory.

The composition of the eluent strongly influences the reten- tion times of the Cu-chelates. The retention times of all chelates increase with increasing methanol concentration [6]. This is shown in Figs. 4 and 5. Using methanol/water (75/25 vv), Cu-DTPA does not elute within 2.5 h, while the Cu-EDTA elutes approximately after 100 min. No separation of the chelates can be achieved with distilled water.

The linear range of the UV detector was from 0.5 to 2.5 mmol/L.

Conclusion

A single HPLC method with UV detection was evaluated to separate and quantitate five chelating agents (as copper chelates) in micronutrients for plants. Problems with the sepa- ration arise because some Cu-chelates exist in isomeric forms.

References

1. Mengel K (1984) Ern~ihrung und Stoffwechsel der Pflanze, 6th edn. Fischer, Stuttgart

2. Finck A (1992) Dtinger und Dtingung - Grundlagen und An- leitung zur Dtingung der Kulturpflanzen, 2rid edn. VCH Ver- lagsgesellschaft, Weinheim

3. Proceedings of the First International Symposium on Foliar Fertilization (1985/1986). Developments in Plant and Soil Sciences, vol 22. Nijhoff, Doordrecht

4. Kluge G, Embert G (1992) Das Diingemittelrecht mit fach- lichen Erl~iuterungen. Landwirtschaftsverlag, Mtinster

5. Barak PH, Chen Y (1987) Soil Sci Am J 51 : 893-896 6. G6ttlicher U (1993) Thesis at Fresenius Akademie, Wies-

baden 7. Frost EA et al (1958) J Amer Chem Soc 80 : 530-536

Letter to the editor

Fresenius J Anal Chem (1995) 352 : 400-402 - © Springer-Verlag 1995

Traceability of measurement and calibration in chemical analysis

Werner Hiisselbarth

Bundesanstalt ft~r Materialforschung und -prtifung, D-12200 Berlin, Germany

Received: 21 December 1993 / Revised: 28 December 1994 / Accepted: 6 January 1995

Dear Sir."

Due to the impact of the ISO 9000 and EN 45 000 series of standards, traceability has become a key item of quality man- agement in the fields of measurement and testing. Regardless of the application field, traceability has been universally de- fined through the existence of calibration chains up to the level of national or international measurement standards, realizing appropriate SI units. In analytical chemistry the requirement of traceability to SI units, in particular to the mole, has rather raised questions instead of answering questions. Throughout the past years our institute has put quite some effort into con- tributing to the international discussion on this subject. Re- cently this matter has finally cleared up considerably, and we are quite satisfied to see that the internationally agreed line of interpretation is essentially identical with the one advocated by our institute. In this letter I will restrict myself to a brief sum- mary of some key points, taking into account recent interna- tional developments.

First, I will say a few words about traceability of measure- ment in general, emphasizing the fact that traceability is no end in itself but only serves the purpose of accuracy assessment.

Secondly, I will touch the special field of chemical compo- sition measurements. Here the main conclusion is, that in al- most all cases of practical relevance certified reference materi-

als (CRM) are and will continue to be the appropriate reference standards of chemical composition.

Finally, some recommendations will be given concerning accuracy assessment in chemical analysis, in particular on a reference material (RM) basis.

General: Traceability of measurement

Let us begin with the universally accepted definition of trace- ability from the "International vocabulary of basic and general terms in metrology" [1], for short VIM. It is, in fact, not the lat- est one, since the second edition of the VIM has appeared re- cently. However, the definition of traceability has changed only slightly. So we can still use the customary one given here for the present purpose.

Traceability. The property of a measurement result whereby it can be related to appropriate standards, generally international or national standards, through an unbroken chain of compar- isons.

Note: The unbroken chain of comparisons is called a traceabil- ity chain.

The following indicates a typical implementation of this con- cept:

international kilogram prototype

OIML certified j mass pieces

sample mass J

Unfortunately the VIM definition severely fails in clearing up the purpose served by traceability requirements.

Traceability is not an end in itself. The only purpose of trace- ability is known accuracy.

This rating of traceability - traceability serves accuracy assess- ment - is already contained in the ISO 9000 documents. How- ever, until recently, traceability and accuracy/uncertainty were considered as quality targets of equal rank by many experts. Meanwhile the situation has changed, and, to my great personal relief, at the 2nd CITAC 1 workshop in March 1994 there was ahnost universal agreement between analytical chemists and physical metrologists, up to the highest authorities, that trace- ability serves accuracy assessment. In fact, the statement above is a quotation from a lecture given by Dr. Quinn, director of the BIPM 2.

For the reasons indicated above the recent draft ISO/DIS 14 111 [2] of an International Standard on traceability in natural gas analysis puts forth a reformulation of the VIM definition as follows.

Traceability. Ability to provide evidence of the overall accu- racy attributed to measurement results, through documented calibrations, using measurement standards of known accuracy, and comparison measurements of known performance.

Note 1: The purpose of traceability is known accuracy, but not necessarily high accuracy.

Note 2: Traceability refers to the overall accuracy of measure- ment. If major uncertainty components cannot be assessed ade- quately, traceability cannot be claimed for the complete mea- surement.

To make this important subject definitely clear, this section will be closed with a KISS 3 type summary of basic require- ments for traceability to the SI system, as follows:

A measurement result is traceable to the SI system if and only if

- it is expressed in the appropriate SI units, and - its total uncertainty is specified

Special: Traceabi l i ty in chemical composition measurement

Let us now turn to chemical analysis, more specifically, to chemical composition measurements. As mentioned in the in- troduction, the analytical community has largely been, and partly still is puzzled by the requirement that analytical results should be traceable to SI units, in particular to the mole as the SI unit of the amount-of-substance. These proposals have been discussed extensively, and I will not go into details of these dis- cussions. Here I will only sketch a few objections of quite dif- ferent type.

Principal (theoretical) objections

If you consider metrological traceability, i.e., traceability to ba- sic units and entities, you must of course consider all the basic units and entities involved. For chemical composition these are

- the SI unit of analyte concentration, - the chemical species of every analyte.

1CITAC stands for Co-operation on International Traceability in Analytical Chemistry, a recently founded international coun- terpart of EURACHEM

2 BIPM stands for Bureau International des Poids et M6sures

3 KISS stands for the demand: Keep it simple, stupid!

401

Therefore metrological traceability would have to refer to a pri- mary standard for the SI unit of concentration, and to primary reference materials for all the chemical species involved.

Practical objections

Traceability has to serve accuracy assessment. However, build- ing calibration chains up to the level of SI units and pure sub- stances is clearly inadequate for the purpose of accuracy as- sessment for the overall analytical procedure, in almost any real-life case. Metrological traceability at best applies to the fi- nal determination of isolated species. However, the main sources of uncertainty are sampling, calibration, chemical in- terferences and matrix effects. Metrological traceability is clearly unable to handle any of these.

The well-established method to handle these problems - ex- cept sampling - is validation using appropriate CRM, or in- house RM calibrated against CRM. That means: Measuring, parallel to the sample, a RM of established accuracy, whose composition (analyte concentration and matrix composition) is sufficiently similar to that of the sample, or measuring several such RM. This procedure, as described in ISO Guide 35, should be promoted, whereever applicable, in routine analyses. We will return to this point later.

Turning to the particular item of traceability to the mole it should be realised that determination of mass and conversion to amounts-of-substance using the known values of molecular and atomic masses is fully equivalent to the direct determination of amounts-of-substance, if the latter is possible at all. The use of molar/atomic masses admittedly imports an additional uncer- tainty, which is, however, negligible in every case of practical interest. So traceability to the kilogram is fully equivalent to traceability to the mole, whereever they apply.

C o n c l u s i o n s

Coming now to the end of my letter, there are a couple of points that I should like to emphasize once more.

1. For assuring the quality of measurement results it is of ut- most importance that

a) all relevant sources of uncertainty are identified, b) the uncertainty is assessed according to a procedure of vali-

dated performance, c) the assessment of uncertainty is fully documented.

For step b the metrological approach of tracing back to stan- dards for SI units constitutes a model of indisputable impor- tance. However, alternative/complementary models are indis- pensable in those fields of measurement and testing where the metrological approach is inadequate or inapplicable, as e.g. in chemical analysis.

2. In the field of chemical composition measurement, the ref- erence standards for the calibration and validation of analytical methods and measuring systems in the first place are and will continue to be certified reference materials (CRM) of complex composition, similar to that of the analysed sample.

For accuracy assessment in chemical analysis traceability to CRM of complex chemical composition should clearly be iden- tified as the main objective. To this end, a number of actions is necessary, on different levels, as indicated below. Among these I want to emphasize the use of CRM as reference standards for the calibration of in-house RM used as working standards.

3. Tasks for establishing traceability to CRM A. An international cooperation should be developed to - improve the comparability of CRM as standards of chemi-

cal composition, - harmonize CRM protocols, concerning quality relevant as-

pects such as evaluation and specification of uncertainty,

402

- if necessary, define different quality levels of CRM and elaborate corresponding certification and validation proto- cols.

B. For improving traceability of uncertainty on the routine level the following concepts and methods should be pro- moted:

- uncertainty monitoring of routine analyses by parallel mea- surement of quality control samples (CRM or in-house RM),

- traceable documentation of uncertainty monitoring, using e.g. control charts,

- use of CRM as reference standards, for establishing trace- ability of in-house reference materials used as working standards in routine analyses.

References

1. BIPM, IEC, ISO, OIML (1984) International vocabulary of basic and general terms in metrology, 1st ed. ISO, Geneva

2. ISO/DIS 14111 Natural gas - Guidelines to traceability in analysis, ISO/TC 193/SC 1, December 1994

We would be Very pleased to hear the opinions from any of our readers, who feel strongly about this matter. Please write to the Editorial Office.

Erratum

Fresenius J Anal Chem (1995) 352:402 - © Springer-Verlag 1995

On-line fractionation and characterization of aquatic humic substances by means of sequential-stage ultrafiltration

P. Burba ~, Valerij Shkinev 2, Boris Ya. Spivakov 2

l Institute for Spectrochemistry and Applied Spectroscopy, D-44139 Dortmund, Germany 2Vernadsky Institute of Geochemistry and Analytical Chemistry, Russian Academy of Sciences, Moscow, Russia

Published in Fresenius J. Anal. Chem. (1995) 351:74-82

Figures 4 to 6 of this paper have unfortunately not been repro- duced correctly. The shading was inadequate. Improved ver- sions are given below.

Distribution, % 12O

i 100! . . . . . ~L ..... ,{ . . . . . . . . . . . . . r . . . ~

60 I

40

20

0 4 5 6 7 8 9 10

pH value

Fig. 5. Influence of the pH-value on the molecular distribution of HS (HS: 1.0 mg/rnl BOC 3/9.5; sample volume: 10 ml, washing volume: 10 ml, HS fractions: F1 to F6)

. . . . HS distribution E3F 6 (<1 kD)

[ ] F 5 (1-5 kD)

.... ~ F 4 (5-10 kD)

~ F 3 (10-50 kD)

~ F 2 (50-100 kD)

--- EE]F1 (>100kD) .

120

100

80

60

40

20

0

Distribution, %

5 HS distribution I;~ F 6 (< 1 kD)

[~F 5 (1-5 kD)

I~lF 4 (5-10 kD)

E~ F 3 (10-50 kD)

F 2 (50-1 O0 kD

rTF 1 (>100 kD)

0 2 4 6 8 Humic substances, mg/ml

Fig.4. Molecular distribution pattern (FI: > 100 kD, F2: 50- 100 kD, F 3:10-50 kD, F4: 5-10 kD, F5:1-5 kD, F 6: < 1 kD) of an aquatic HS (BOC 3/9.5) attained by five-stage ultrafiltra- tion as a function of its concentration (pH 6.0)

Distribution, %

r HS distribution I~F 6 (<1 kD)

Eli F 5 (1-5 kD)

[ ] F 4 (5-10 kD)

[ ] F 3 (10-50 kD)

E~F 2 (50-100 kD)

C3F 1 (>100 kD)

VO 5 10 15 20 25 30 35 NaCI, g/I

Fig. 6. Salt influence (e.g., NaC1) on the molecular distribution of HS (HS: 1.0 mg/ml BOC 3/9.5, pH 6.0, other conditions as in Fig. 5)