[资源分享] XRF vs. ICP分析金属[e文]

Theresa recently asked which results tend to be more accurate for metals in concreate: XRF or ICP. She explained that she uses ICP and ICP-MS and has often compared her values to other analysts' XRF values and they never agreed. She noted that her laboratory usually does a borate powder press when they do XRF and then performs a standardless analysis.

EDXRF and/or Wavelength XRF suffer from matrix and interelement interference issues, as does ICP (OES and MS). With XRF, the standard and sample must be matrix matched with respect to both the matrix and the particle size. In addition, I have seen that different chemical forms can cause problems such as alkyl Si versus Si oxides in an organic matrix, i.e. the elemental composition is similar, yet results for the methyl silicon are much lower than oxygenated silicon due to scattering. Furthermore, the analyte emission intensity is enhanced by matrix elements lower in atomic number and suppressed by analytes higher in atomic number.

    With respect to Cements by XRF, it is possible to fuse them into a glass using lithium borate/carbonate. This eliminates the physical interferences and leaves the analyst to make corrections on the interelement effects - calculations are involved and not perfect. The standardless XRF determinations were started 20+ years ago and I have tried it. The common term then was 'MARS' (Matrix Analysis Routine using Scatter). It is an approach that I found to give reasonable results provided that major components are of prime interest and agreements to around 10 to 15 % relative are acceptable. I have used MARS for screening and semi-quantitative analysis (I love math and physics and this technique attracted me for those reasons, plus it was new at the time) but the ICP-OES approach has the potential for giving very accurate and precise results. Using the lithium borate fusion-dissolution in dilute nitric acid (matrix matched at this point) and using lines that do not spectrally interfere, the ICP-OES technique can do this while XRF cannot. With XRF, the analyst is stuck making corrections that contain a level of uncertainty.

    In short, I would choose ICP-OES over ICP-MS and XRF because the systematic errors are much easier to eliminate and the random measurement error can be as low as 0.5 %. (XRF can do better with respect to precision, but who need a very precise "wrong" answer?)