Figure S1 (black lines) shows the experimental EXAFS data at the As K-edge for the As-MgAlFe HTLC and the moduli of the Fourier transform superimposed with the best fit (red curves) obtained using the structure of two different model compounds (Alarsite (AlAsO4): panels a and b and Scorodite (FeAsO4·2H2O): panels c and d). The fit of the two structural models was done by fixing the coordination numbers and the non-structural parameters S02 and E0 (refer to supplementary information for more details). When Alarsite was used as structural model, 4 best fits were done while keeping fixed the degeneracy of the paths As-O and As-O-O and setting different values from 1 to 4 for the degeneracy of the As-Al path. In the case where Scorodite was used as structural model, 4 best fits have been done while keeping fixed the degeneracy of the As-O and As-O-O paths and setting different values from 1 to 4 for the degeneracy of the As-Fe path. The structural parameters are listed in Table S2. Based on previous work by Robertson et al. 2017, the Alarsite model was used to verify if a bidentate-binuclear (BB) model would also fit in our case. In this case, CNAs−Al = 2 and RAs−Al = 3.14 Å was obtained. These values are in agreement with the bond length range typical of AsO4 bonded Al phases (RAs−Al = 3.11–3.21 Å)(Foster and Kim 2014) but the As-Al interatomic distances are shorter than previously reported (RAs−Al = 3.31 Å) by Robertson et al. 2017 These longer distances previously reported were suggested to arise from AsO4 bonding to the interlayers (outer-sphere) and/or edge-faces (inner-sphere), but no clear evidence for either was ever given. Since it has been previously reported (Wang et al., 2009; Wu et al., 2013; Foster and Kim 2014; Wang et al., 2018) that MgAl HTLCs are mainly outer-sphere bonded to AsO4 with no clear evidence of second neighboring scattering, we used a similar mode with a CNAs−Al = 1. Using the outer-sphere type model with a CNAs−Al = 1, the interatomic distance is the same as the BB model with CNAs−Al = 2. In addition, the goodness of fit values (χ2 and R-factor) did not change significantly. We then proceeded to test if varying the CN’s to a higher value would improve or make worst our fits to our As-MgAlFe HTLC data while keeping in mind that no 3C or higher coordination EXAFS has been reported for As-Al phase systems (Wang et al., 2018). Surprisingly, similar bond lengths were obtained (RAs−Al = 3.14–3.16 Å) as we increased the CNs, and the goodness of fits improved slightly but not enough to solidify a definite answer. Since our As-MgAlFe HTLC also contains 1/3 of Fe atoms per Al atom (Table S1) and since Fe has a greater affinity of As binding (Masue et al., 2006), the Scorodite model (Fig. S2) was used to fit an As-Fe second shell as previously done (Robertson et al., 2017). This model yielded worst results (Table S2-S3) and longer bond distances (RAs−Fe = 3.43 Å) than those reported in the literature (Foster and Kim 2014). Looking at some other works (Paikaray and Hendry 2013) where a MgFeSO4 HTLC was reported to fit a bidentate-mononuclear (BM) complex with RAs−Fe = 2.83 Å, an attempt was done to fit this BM (Table S2) to our data but yielded an RAs−Fe = 3.40 Å which is longer then expected for BM complexes (RAs−Fe = 2.80-3.00 Å)(Foster and Kim 2014). Similarly, to the As-Al shell case, we decided to investigate how varying coordination numbers of the As-Fe shells would affect our data fitting while keeping in mind that some are not existent (e.g. 3C) or reported (Foster and Kim 2014). Again, it was observed that, varying the degeneracy of the As-Fe bonds did not improve much the goodness of fit, while at the same time, the bond lengths remained largely similar (average of RAs−Fe ~ 3.41 Å). Therefore, the results of these two methods of analysis to include As-Al and/or As-Fe as second shells were insufficient to choose any binding mode with absolute confidence for our As-MgFeAl HTLC. But certainly, our results did not correlate with the BB model previously reported by Robertson et al., 2017 or other pure MgAl and/or MgFe HTLC varieties (Wang et al., 2009; Opiso et al., 2010; Paikaray and Hendry 2013; Burke et al., 2013). Yet, the fact that changing the CN numbers for either As-Al and/or As-Fe second shells has such a little effect on the goodness of fit and bond lengths indicates that there is little interaction of the AsO4 with our MgFeAl HTLC substrate. By comparison, if such a method is applied to a compound with a fixed CN (e.g. Scorodite where CNAs−Fe = 4 and RAs−Fe = 3.35 Å), an obvious worst fit is observed as the CN is varied. Therefore, this hints that our AsO4 is indeed forming an outer-sphere-like complex. As such, we decided to further look at the vibrational signatures (Figure S3 and Table S4) of our AsO4 and observe if the degeneracy of the υ3 mode was removed (Kloprogge and Wood 2017; Gomez et al., 2010) as it is expected for monodentate and bidentate complexes. However, upon inspection of the vibrational spectra, only a single band was observed, and no degenerate υ3 modes are lifted. This indicates that the AsO4 retains its free ion symmetry and is in agreement with what was observed in the As K-edge EXAFS. Something that further remained unanswered from previous works (Robertson et al., 2017) was whether the AsO4 occupied the interlayer spaces or was simply adsorbed on the surfaces of the MgAlFe HTLC. Hence, one way that has been traditionally used to determine this (Palmer et al., 2009; Palmer and Frost 2011) is by monitoring the changes in the basal (d003) and interlayer (d006) spacings via PXRD before and after adsorption, whereby changes ≥ 0.1 Å are considered to indicate intercalation. Based on our data (Fig. 1 and Table S5), we could observe that changes to the basal and interlayer spacings were < 0.1 Å and as such indicated surface adsorbed and not interlayer incorporation, in agreement with our previous work (Lu et al., 2020). A slightly higher background is noted here after the adsorption for the PXRD but is similar to what is observed for SeO4 and MoO4 adsorbed MgFeSO4 HTLC by Paikaray et al. 2013. Furthermore, an extra peak at ~ 19.55˚ after adsorption is noted and likely arises from cation ordering as previously observed for MgGa HTLCs (Belloto et al., 1996). Therefore, with the information now gathered, a DFT-optimized model (Fig. 2 and S4) to fit our EXAFS data (Table 1) using a single shell model was built that yielded reasonable results.
Table 1
The shell-fit results for the As, Se, and Mo K-edge EXAFS of our HTLCs. The DFT optimized adsorbed structures were used as the initial models for the As, Se, and Mo K-edge EXAFS fitting. CN refers to the coordination number. Rpath is the interatomic distance. σ2 is the Debye-Waller parameter. ΔE is the energy-shift parameter, and χ2red is the reduced chi-square. The R-factor is the mean-square misfit between the measured and the modeled data.
Sample | Path | CN | Rpath (Å) | σ2 (Å2) | ΔE (eV) | χ2red | R-factor |
As-HTLC(a) | As-O | 4 | 1.69(3) | 0.0011(3) | 7.4(9) | 242 | 0.020 |
As-O-O | 12 | 3.20(10) | 0.0015(58) | | | |
Se-HTLC(b) | Se-O | 4 | 1.65(2) | 0.0001(23) | 8.5(9) | 167 | 0.014 |
Se-O-O | 12 | 3.13(11) | 0.0032(71) | | | |
Mo-HTLC(c) | Mo-O | 4 | 1.76(2) | 0.0017(3) | 8.6(6) | 95 | 0.009 |
Mo-O-O | 12 | 3.20(4) | 0.0030(5) | | | |
(a), (b), and (c): the fitted k range was 2.8–13.3 Å−1 using the Kaiser-Bessel window, whereas the number of independent points (Nidp) and variables (Nvar) were (a) 16 and 5, (b) 14 and 5, (c) 15 and 4, respectively. S02 values were set at (a) 1, (b) 0.86, (c) 0.95, respectively (refer to supplementary information for more details). |
In the case of the Se-MgAlFe HTLC, the various model compounds used (Fig. S2 and Table S3) to fit our data (Fig. S5) and our obtained parameters are listed in Table S6. Again, we decided to first use a possible Al-based model compound [Al2(SeO4)3] to fit our data whereby a CNSe−Al of 4 and an RSe−Al bond distance of 3.43 Å was obtained with reasonable goodness of fit values (χ2 and R-factor). However, our obtained bond lengths were larger than those expected for an Al-Se coordinated compound such as Al2(SeO4)3 and thus was not deemed to be a reasonable solution. Next, an outer-sphere like model with a CNSe−Al of 1 was investigated as it has been previously reported (Opiso et al., 2016) that a MgAl based HTLC has this type of bonding mechanism with SeO4. In this case, the goodness of fit values was slightly better than the previous case (CNSe−Al = 4) and a bond length of RSe−Al = 3.62 Å was determined. Varying the CNAl−Se didn’t improve the goodness of fit values appreciably nor did the RSe−Al change significantly for the most part. Using the possibility of a Se-Fe second shell mode via the use of NaFe(SeO4)3 gave slightly shorter bond lengths (RSe−Fe = 3.34 Å) than expected for a CNSe−Fe of 3 and reasonably goodness of fit values. Again, varying the CNSe−Fe didn’t improve the goodness of fit values remarkably nor the variations in RSe−Fe. Therefore, as in the As-MgAlFe HTLC case, the EXAFS analysis of the Se-MgAlFe HTLC indicated that there was little interaction of the SeO4 with our substrate. Analysis of its vibrational spectra (Fig S3 and Table S4) and PXRD data (Fig. 1 and Table S5) showed only a single band with no degeneracy removal of the υ3 mode (i.e. outer-sphere bonding) (Su and Suarez 2000) and no intercalation of the SeO4 to the interlayers of the MgAlFe HTLC (i.e. only surface bound). Therefore, with this information in hand, we then proceeded to build a DFT-optimized structure (Fig. 2 and S4) to fit our EXAFS data (Table 1) using a single shell model that yielded reasonable results and agrees with what has been noted for other pure MgFe and MgAl HTLCs (Paikaray and Hendry 2013; Opiso et al., 2016) interactions with SeO4.
Finally, for the Mo-MgAlFe HTLC, a similar approach using Al-based [NaAl(MoO4)2] and Fe-based [Fe2(MoO4)3] model compounds to fit our data was undertaken (Fig. S6) and the obtained fitting results are reported in Table S7. In the case of the Mo-Al second shell with a CNMo−Al of 3, much shorter bond lengths than those expected for such a complex were obtained. Varying the CNs had little effect on the goodness of fit values and RMo−Al attained. Likewise, when this approach was applied to a Mo-Fe second shell using our model compound, a similar effect was observed. Thus, indicating that there was little interaction of the MoO4 with our MgAlFe HTLC. Vibrational spectroscopy again pointed to a single band component (Fig. S3 and Table S4) indicative of outer-sphere bonding while its PXRD only gave signs of surface adsorbed interactions. Hence, our final model built with a DFT optimized structure (Fig. 2 and S4) to fit our EXAFS data using a single shell model (Table 1) yielded reasonable fit results and agrees with what has been observed for MoO4 interactions with a pure MgFe HTLC (Paikaray ad Hendry 2013).