Accurate analysis - a standard OES calibration procedure

TWI Bulletin, July/August 1990

Sheila Stevens is a Senior Research Chemist in TWI's Chemical laboratory.

Her work at TWI has involved analysis of steel weld metal, fluxes, wire and wrought products, specialising in modern instrumental methods. This has extended to study of theoretical aspects of optical emission spectroscopy.

Sheila has also worked on projects involving sampling and analysis of welding fume and gases, including local extraction ventilation, determination of free nitrogen by hydrogen hot extraction and extraction and analysis of gases from pores in weld metal.

A current interest is the use of chemical and physical techniques for characterisation of structural properties in weld areas in thermoplastics.

Although optical emission spectrometry (OES) is probably the most widely used method for steel analysis, the accuracy of the results is only as good as the procedure used to calibrate the instruments. Surprisingly, there is no British Standard specifying OES calibration procedures. Sheila Stevens takes up the story and summarises a calibration method based on determining interference corrections using binary alloys.

Accurate and rapid analysis of steel is vital for production control during steelmaking, and OES is used extensively for this purpose. Similarly, the steel user is often required to undertake analysis for QC purposes, and OES has almost completely replaced conventional wet chemical analysis which is time consuming and therefore expensive.

However, OES is an instrumental rather than a primary technique involving standardised chemical methods. Reliable analyses depend on the calibration and operation procedures used for various commercial instruments available, and at present there is no British Standard giving guidance on these facets of OES analysis. Round-robin trials have shown that there is a risk of disagreement between laboratories using nominally the same analysis techniques.

In most situations, the reliability of OES results is perfectly acceptable, but the trials showed that the reported range of a given element may well be +10% of content at the 1% level, and as much as +60% for elements - such as sulphur - present at low levels. However, one of the trials showed that interlaboratory agreement could be improved by using a uniform calibration procedure.

A standard method is desirable, and a programme of work was carried out which demonstrated the feasibility of a calibration method incorporating interference correction factors determined using binary alloys, for the analysis of C, S, P, Si, Mn, Ni, Cr, Mo, V, Cu, Nb, Ti, Al, B, Sn, Co, As, Pb and Zr in steel by direct-reading OES. It also examined sample preparation and the suitability of a tungsten-argon arc remelting technique for inclusion in a standard method.

Interelement interferences

However, the analysis requires different sets of calibrations for each type and, while acceptable for steelmaking, this method is unsuitable for steel users, who may be required to analyse a wide range of compositions and are therefore unable to employ a narrow range of reference materials. The number of commercially available certified reference materials (CRMS) is too limited to allow wide-range calibrations.

An alternative is to extend the calibration range by using interelement interference correction factors, according to the alloys present in the material. Traditionally, two types of interference are recognised, namely 'translational' or 'line' (owing to overlapping spectral lines) which is directly proportional to the interfering element, and 'rotational' (owing to the matrix composition) which is proportional to the product of the concentrations of the interfered and interfering elements. The most reliable way to determine correction factors is to use binary alloys.

Calibration procedures

Sampling for OES analysis

Another useful aspect of remelting is that it is possible, by melting together known amounts of two or more CRMS, to produce synthetic CRMS which cover areas in the calibrations where no suitable CRMS exist; for example, low-alloy CRMS containing boron can be obtained by melting together appropriate amounts of a mild steel residual CRM containing boron with a stainless steel CRM. It follows that adequately-sized samples for analysis can be made from undersized samples by melting with pure iron or a CRM, and that alloy samples can be diluted to within the calibration range.

Preparing calibrations

Preliminary calibrations

A further improvement was obtained by using an 'iron-correction' method whereby a calibration graph was plotted of intensity ratio against concentration ratio. Concentration ratio is (% element x 100)/%Fe.

As there was no longer a direct relationship between the intensity ratios and the concentrations of the elements being measured, the internal standard concentration was determined by '100% normalisation' before the concentration ratios could be converted into concentrations. This uses a formula based on the fundamental rule that all the concentrations measured for all significant elements must add up to 100%, as illustrated below for stainless steels, where the element symbols represent their concentrations:

C + S + P + Si + Mn + Ni + Cr + Mo + Nb + Tt + Fe = 100%     [1]
Tt = total trace elements not measured.

Dividing equation [1] by the iron concentration gives:

C/Fe + S/Fe + P/Fe + Si/Fe + Mn/Fe + Ni/Fe + Cr/Fe + Mo/Fe + Nb/Fe + Tt/Fe + Fe/Fe = 100/Fe     [2]

This can be represented in general terms by:

Σ[C _i/C _s] + 1 = (100-Tt)/C _s     [3]

where

The internal standard concentration can be calculated from equation [3]:

C _s = (100-Tt)/(1+ Σ[C _i/C _s])     [4]

The concentration ratio method requires that all elements are measured if they are likely to occur in concentrations above 0.3%, as the effect on the internal standard calculation would be large enough to affect the concentrations of the other elements. If the total of the trace elements not measured is unlikely to exceed 0.5%, and if the total is not expected to vary by more than 0.1%, either Tt can be represented by the probable total as a constant, or it can be disregarded, without affecting any measurement significantly except the concentration of the internal standard. Once the concentration of the internal standard has been calculated, the other concentrations can be calculated by:

C _i = C _s (C _i/C _s)     [5]

The concentration ratio method is necessary for the analysis of stainless steels but is not normally used for low-alloy steels and mild steels, since it is more time consuming to set up the initial calibrations.

TWI uses the 'double-ratio' technique ( ie iron ratio and concentration ratio) for all samples, including carbon steels and low-alloy steels, because the diversity of samples analysed makes it impracticable to carry out 'type analysis' and the double ratio improves accuracy for low-alloy steels with a relatively high-alloy content.

Correction factors

The only accurate way to determine correction factors is to use binary alloys. For example, if element Y interferes with element X, binary alloys containing nominally 0, ½, 1, 1½, and 2% of Y are used.

A calibration graph was plotted using mild-steel residual CRMS containing little or no element Y. Using this, the intensity ratios obtained on the X channel of the instrument from the binary alloys were converted into concentration ratios. The 'apparent' concentration ratio of element X caused by element Y was calculated to give the 'error' which was plotted against the intensity ratio of the binary alloys obtained on the Y line. For linear graphs the equation of the line is y = mx + c; m is the correction factor. If the graph is a second-order quadratic, the equation is y = m ₁x + m ₂x ² + c with two correction factors, m ₁ and m ₂.

The Table gives the magnitude of the interferences which were determined, expressed as the apparent concentration which would be obtained if 1% of the interfering element were present, and the interferences for 10%Ni and 18%Cr; the Figure illustrates the importance of corrections, using Pb as an example.

Interferences, wt%, equivalent to 1% of interfering element

Final calibrations

For ferritic steels, common calibrations were obtained for carbon and low-alloy steels for all elements except Si, Mn and Cu, although whether the differences for these (0.02% for Si, 0.03% at 1.4% for Mn, and 0.03% at 0.4% for Cu) are significant would depend on the accuracy required.

Where sufficient CRMS are available for preparing separate ferritic steel and stainless steel calibrations, there is no real advantage to be gained from using a common calibration produced using both types of CRM. However, coincident calibrations are a very good check of the validity of the correction technique and, effectively, common calibrations were obtained, after correction, for S, P, Mn, V, Nb, Ti, Sn, Co and As. The corrections involved were large: for example, the calibration differences for P and As were reduced from 0.014 to 0.002, and from 0.033 to 0.001 wt % respectively.

Evaluating calibrations

The detection limits were measured using a pure-iron CRM, and the values checked with low-level CRMS. Instrumental precision was determined using two carbon steels, a low-alloy steel, and two stainless steels; the precision is level dependent but, at the detection limits, the precision is about ten times better than the detection limit.

The validity of the calibrations was checked using test samples from two sources to compare the accuracy of the results as follows:

1. By comparing OES results on solid and remelted CRMS (those which were not used to prepare the calibrations) with certificate values from independent analysis using standard wet analysis.
2. By comparing OES results from commercial materials (plates and wires) with results from conventional (although not necessarily standard) wet chemical analyses. The plates were analysed on the top and bottom surfaces, on the cross section, and after remelting full through-thickness samples into buttons. The wires were analysed in the remelted form.

Any decision as to whether or not the OES results were in agreement with the true value had to take into account the reproducibility of the standard method, the number of decimal places to which the results are normally reported, and the effects of rounding. On this basis, 384 OES results were compared with certificate/chemical figures and, of these, 351 were in agreement with the latter figures. However, this figure rises to 379, representing an acceptance rate for the OES results of 98.7%, if the following factors are taken into account:

1. Where the chemical figure was obtained by a non-standard method, the reproducibility of that method may be different from that quoted for the standard method.
2. Each certificate value is the mean of a number of values from individual laboratories and may encompass a considerable range of results.
3. The accuracy of the OES results cannot be better than the accuracy to which the CRMS used for the calibrations were certified.
4. Comparisons based on the SD (Standard deviation) of the standard method were made to assess instrument performance but the acceptance criteria for OES results may allow a larger SD.

The ability of OES to produce accurate analyses can be illustrated using the remelt results for C, S and P. The average deviation of the C results from the true value (certificate or chemical) was 0.002%, and for four of the ten samples the OES figure was within 0.001% of the true value. Five of the ten S results were within 0.001% of the true value and a further three were within 0.002%. The results for P showed that eight out of ten results were within 0.001% of the true value and the remaining two were within 0.002%.

Effects of remelting

An examination of the data revealed apparent losses for C, S, Mn, Ti, and Al, and apparent gains for P and Si in remelted CRMS. All other elements were unaffected by the remelting procedure. Chemical analysis carried out in the earlier study confirmed that the C and Mn losses (0.01% at 0.6%C and 1%Mn, respectively) were real, but not the changes for S, P and Si. The apparent loss for S was thought to be due to segregation in high-sulphur CRMS elevating the solid calibration graph with respect to the remelt graph, and the apparent gains for P and Si were probably due to differences in instrumental response to the two sample forms. No chemical analyses were carried out for Ti and Al, but these elements are deoxidants and losses on remelting have been reported.

Despite the homogeneous nature of the CRMS, segregation can occur for elements such as Mn and S, owing to MnS inclusions. This was found to be the case for S in two CRMS which plotted above the calibration line in the solid form, but could be incorporated into the calibrations after remelting. It is probably fair to say that the elemental changes on remelting would be considered insignificant in the majority of cases and that remelted samples could be analysed using solid calibrations and vice versa. However, it is always desirable to perform analyses using calibrations produced using CRMS in the same form as the sample.

Main conclusions

This work demonstrated the feasibility of a calibration method for OES using interference correction factors determined using binary alloys. In particular it showed that:

1. Interelement interferences are considerable, even for carbon steels, but wide-range instrument calibrations can be obtained enabling the successful analysis of carbon steels, low-alloy steels and stainless steels to be carried out.
2. Over 98% of the OES results, obtained using the calibration technique described, were in agreement with certificate or chemical values.
3. Remelting using a tungsten-argon arc remelter and Zr deoxidant is a highly reproducible procedure, and enabled materials for calibration and analysis to be prepared; any small elemental changes for S, P, Si, Mn, Ti and Al are calibrated using remelted CRMS.
4. While OES is not suitable for primary standardisation, it is the main method used by industry and it is therefore vital that procedures are standardised. Sample preparation, calibration techniques, and instrument performance assessment should be defined. The feasibility of this approach has been demonstrated.