Portable X-ray Fluorescence Analysis of Organic Amendments: A Review

Chronologically, the following research on the subject is that of McWhirt et al. ^[3] in 2012. The researchers calculated LOD in compost samples for the elements As, Ca, Cd, Cr, Cu, Fe, K, Mn, P, and Zn and stated substantial improvements for all elements compared to the previous research by Weindorf ^[4] attributable to improvements in technology. Their conclusion was that pXRF was able to acceptably quantify Ca, Cr, Cu, Fe, K, Mn, P, and Zn in dry samples but the divergences were important for As and Co.

Healy et al. ^[5] characterized metals in Irish sludges that had undergone treatment by thermal drying, lime stabilization, or anaerobic digestion. Correlation coefficients between pXRF and ICP-MS (aqua regia digestion) results indicated the suitability and satisfactory use of the pXRF technique for the quantification of Fe, Cu, Se, Zn, and Pb (r > 0.90). However, it should be noted that the regulated levels in sludge are higher than for other amendments.

Shand and Wendler ^[6] used pXRF for soil analysis but, being Scottish peat soils with very high OM contents, it is really equivalent to the matrices. Shand and Wendler determined the LOD in the organic soils and found values below 5 mg kg⁻¹ for Cr, Ni, Zn, As, and Sr, values between 5–10 mg kg⁻¹ for Cu, and values between 10–15 mg kg⁻¹ for Mn and Pb. These LOD indicated better sensitivity for Cr, Ni, and Zn and worse for Mn and Pb (and also Fe and Ca) compared to the previous McWhirt study ^[3]. The researchers compared the concentrations of K, Ca, Ti, Cr, Mn, Fe, Ni, Cu, Zn, As, Sr, and Pb determined by pXRF with pseudototal concentrations determined by aqua regia extraction and ICP-OES. Their results indicated that the in-built calibrations of the instrument were not adequate for these organic soils and should be modified. The researchers indicated that mineral or peat soils were not completely dissolved by the aqua regia reflux procedure they used, leaving behind undissolved refractory minerals such as quartz and feldspars, so comparison of aqua regia extraction with pXRF analysis is not straightforward. Nevertheless, the researchers attribute the difference between the pXRF vs. aqua regia extractable analysis to matrix effects in the pXRF and not to low recovery by aqua regia/ICP analysis.

Although it does not appear in the database searches undertaken, it is necessary to cite the research by Tighe and Forster ^[7]. The researchers applied pXRF to litter analysis and found good results only for the major elements calcium and potassium.

In 2018, two studies by Weindorf et al. ^[8] and Li et al. ^[9] adapted methods developed for soils and used elemental data as a proxy for the prediction of derived properties, such as compost electrical conductivity (EC) and cation exchange capacity (CEC), respectively. To calculate the elemental contents, the recovery percentages based upon NIST certified reference material 2711a were used to establish correction factors for each element.

Havukainen et al. ^[10] compared the concentrations obtained by pXRF with those by ICP-MS (following nitric-hydrochloric acids microwave digestion) in several types of waste: fine fraction reject from solid recovered fuel production, fly ash, biowaste, and compost. The results obtained by the two methods were unacceptably divergent. Regression analysis showed a linear correlation only with Ca and Zn values and, thus, linear correction to transform pXRF results to be comparable to ICP-MS results could be possible for these elements. Nevertheless, pXRF is reported to be best suited for samples such as ash and compost because of their following physical properties: not too moist, quite small particle size, and not too heterogeneous. The researchers concluded that various factors such as sample moisture content, physical and chemical matrix effects (e.g., inter-element effects, particle size, and homogeneity), instrument resolution, and inconsistent sample positioning affected the reliability of the measurement.

Thomas et al. ^[11] used pXRF to measure the total elemental content of soils, crops, and organic fertilizer samples (anaerobic digestate, compost, farmyard manure, and straw). The researchers measure several nutrients in the amendments but it is not indicated if another comparison method was used. In the case of soils, it is indicated that a CRM sample was used and that the soil calibration does not predict potassium with accuracy, and underpredicts 2–3-fold.

López-Núñez et al. ^[12] used the technique for organic samples from an interlaboratory program that determined certified aqua regia contents. They found good linear correlations between measured and certified values of 16 elements when corrections were applied to pXRF elemental readings by measurements of other main elements, with silicon content being one of the most influential fitting terms in eight of the modelling equations. These results indicate that the low recovery of the aqua regia extraction, due to its inability to dissolve silicates, could be responsible for the poor fit of pXRF results when not compared to total matrix digestion.

In the study by Sapkota et al. ^[13], the correlations among pXRF and ICP-OES (nitric acid digestion) were strong (R² > 0.90) for all elements (P, K, Ca, Mg, Fe) except Mg (R² = 0.59) in dried samples (<10% moisture).

Analogous to the previous research, Horf et al. ^[14] compared pXRF and ICP-OES and found R² above 0.8 for the elements P, K, Mg, Ca, S, Mn, Fe, Cu, and Zn when analyzed in dry manure samples. The researchers found a clear tendency of improving regression models due to drying, but the R² was satisfactory (only) for K in fresh and liquid samples. Despite the goodness of the regression coefficients, the slopes of several elements (Ca, Cu, Zn, Fe, K) were greater than unity (greater concentration by pXRF than by ICP-OES). Although the researchers indicate that the digestion they applied did not yield total concentrations, they considered this to be a small effect that cannot explain the high slopes and considered the bias to be due to element-specific matrix-dependent effects.

The LOQs of pXRF or their respective LODs are poorer than those of large laboratory-based XRF instruments or of ICP-related techniques.

Several of the previous studies ^[3][6][10] also evaluated the influence of other factors that will be outlined in the following sections.

The above studies highlight the limited use of pXRF for the analysis of organic amendments. Despite the fact that the evolution of the technique has led to better detection limits that would have allowed the measurement of almost all the elements of interest in these matrices, the results of the various studies are contradictory. In general, they show important divergences between the results obtained by pXRF and the reference analytical procedure used in each study. There may be different reasons to account for these divergences. One explanation could be the different calibration strategies they have used. In a majority of studies, the pre-calibrated method called “Soil” has been used in the instruments, but in some studies other pre-calibrated procedures have been used. More information about pre-calibrated analytical modes in different pXRF instruments and their relevance and applicability to geochemical investigations can be found in a recent research by Laperche and Lemière ^[2]. On some occasions, calibration corrections based on measurements made in reference samples have been used. Measurement results should have been validated by the analysis of a suitable reference material but this procedure has not always been followed. On many occasions, the CRMs used are soils or sediments in the absence of similar organic materials certified for XRF. It is possible that the different pre-calibrations, by making use of different spectral peaks for the same element and different correction formulas, may give rise to different deviations due to the sample matrix, causing a good fit of some elements in some matrices and not in others.

In most of the studies discussed, the comparison of methods (pXRF vs. wet chemistry method) was elucidated by calculating the correlation (r) or the regression (r or R² and slope) between both procedures. Especially for the case of comparing a simple or approximate method (as is the case of the pXRF measurement of the elements in concentrations closer to the LOQ) with a more precise one, these statistical parameters are inadequate to determine the similarity of methods, as Altman and Bland revealed in their classic work ^[15]. It would be necessary to use other methods for the statistical comparison of the procedures.

However, the probable reason for the lack of adjustment between the measurements with pXRF and the chosen reference procedure lies in the different recovery that can be obtained with the different extractant mixtures used (aqua regia, nitric acid, with or without hydrofluoric acid (HF)) and even the different extraction conditions (microwave, open digestions). García-Delgado et al. ^[16] tried different digestion procedures on amendments and found that microwave HF acid mixtures obtained the highest trace element recoveries for all tested metals except Al. They verified that Cr, Fe, Ni, and Pb were mainly associated with the residual fraction of the BCR sequential extraction procedure, suggesting that these metals were primarily contained in silicates and other resistant minerals. García-Delgado et al. ^[16] found the recoveries for Cr, Ni, Mn, Fe, and Pb were between 11% and 23% greater than for non-HF extractions. The recoveries for Cu, Zn, and Cd were about 10% higher when HF was used. The Si recovery was negligible if HF was not used, while that of Al hardly reached 50%. Similar results were obtained for Cu, Fe, and Al recoveries in a similar work by Sandroni and Smith ^[17] using a sewage sludge CRM. The acid mixtures with HF more efficiently attacks aluminosilicates and the metals extracted with HF were absorbed into aluminosilicate compounds, so aqua regia and nitric acid were not able to extract them completely.

Surprisingly, there are no studies evaluating the effect that organic matter content may have on pXRF readings in organic amendment matrices. This effect in soils has been recently research by Silva et al. ^[18] when evaluating the application of pXRF in tropical soils. There is common agreement that soil organic matter may cause a scattering and attenuation of signals when a soil sample is irradiated by X-rays and that the influence and interference magnitude of soil organic matter are still poorly understood. Consequently, the result for a target element could be modified.

Probably the most detailed study in this regard is that of Ravansari and Lemke ^[19]. In their study, three organic matter substitutes (cellulose, graphite powder, and confectioner’s sugar) were added to a soil standard reference material and the concentrations of 13 elements (As, Cr, Cu, Fe, Mn, Pb, Rb, Sr, Th, Ti, V, Zn, and, Zr) were measured as a function of variable organic matter fractions. Their results demonstrate the attenuation of the pXRF signal with the increase in the organic matter fraction of the sample, although the deviations depend on each element. For example, pXRF-Pb concentrations were not affected, pXRF-As concentrations were underestimated, and Mn showed a more complex behavior, both overestimating and underestimating depending on its concentration. The correction factors for the effect of organic matter were generally less than ±8% for organic matter contents up to 30%.

Costa et al. ^[20] verified the effect of organic matter removal on oxides determination by pXRF in soil samples (<2 mm) and their clay fraction. The soil organic matter presence led to underestimation of the pXRF results for the lightest oxides (Si and Al) compared to the heaviest oxides (Fe, Ti, and Mn) which were less affected. Soil organic matter in these samples was below 4.4% (44 g kg⁻¹), which is a content one order of magnitude lower than that found in organic amendments. The lightest elements have lower XRF photon energy values and, as a consequence, the effect of organic matter coating the soil particles can be more pronounced. The reduction of the SiO₂ content reached 20–22.5% with respect to the content of the mineral soil (3–5% as SiO₂). The reduction of contents of the light oxides (SiO₂ and Al₂O₃) can be considered relevant. On the contrary, Fe₂O₃, TiO_2, and MnO content reductions obtained by pXRF can be considered irrelevant with variation before and after organic matter combustion lesser than 1%. The researchers indicate that the possible effect of organic matter may be related to the scattering of primary and secondary X-rays, underestimating the results, in the same way that has been observed for soil with a high moisture content.

Shand C.A. and Wendler R. ^[6] evaluated pXRF analysis of peat and organic soils. They found in an ombrotrophic peat, satisfactory results for Cu (certified content 5.28 ± 1.04 mg kg⁻¹) and Pb (certified content 174 ± 8 mg kg⁻¹) but the concentrations of Ca, Ti, Cr, Ni, and Zn were overestimated by 2–3 times, and Fe by 5 times. In the same study, they analyzed the concentrations of K, Ca, Ti, Cr, Mn, Fe, Ni, Cu, Zn, As, Sr, and Pb of 183 Scottish topsoils that had a wide range of organic carbon (OC) concentrations (1.23 to 48.8%). They concluded that modifications to manufacturer calibrations are needed and that empirical calibration of XRF instruments requires unavailable certified reference peats and peat soils, certified for total element concentration across a wide range of elements.

In a recent study, Sut-Lohmann et al. ^[21] compared the feasibility of common laboratory methods (microwave plasma-AES and pXRF) to evaluate element concentrations in soil highly influenced by sewage waste. The lowest correlation between both methods was obtained for Ni (R² = 0.70, stronger points scattering), although the correlations were satisfactory for the other elements (Pb, Cu, Zn and Cr). The researchers also approached another strategy for the analysis of organic samples with pXRF: the elimination by ignition of the organic matter to then carry out the readings in the calcined samples. This same sample calcination strategy was already used by Vodyanitskii et al. ^[22] to measure the metals Ba, Sr, V, Cr, Ni, Cu, Zn, Zr, Pb, Y, La, and Ce, although in this case, a nonportable laboratory instrument was used. Sut-Lohmann et al. ^[21] found that OM removal in the soil sample influences its density and effective volume, which interferes with pXRF elemental analysis. However, if the data were corrected for organic matter, the influence of the matrix on the pXRF measurements was reduced. Their conclusion was that results validate a usage of pXRF as an alternative method for the conventional wet-chemical-based AES analysis for soil samples highly influenced by sewage wastes. However, the researchers indicate that the prior calcination invalidates the use of the technique in the field.