Yi-Hsuan
Wu
a,
Denis A.
Kuznetsov
*a,
Nicholas C.
Pflug
b,
Alexey
Fedorov
a and
Christoph R.
Müller
*a
aDepartment of Mechanical and Process Engineering, ETH Zürich, Leonhardstrasse 21, 8092 Zürich, Switzerland. E-mail: denisk@ethz.ch; muelchri@ethz.ch
bDepartment of Environmental Systems Science, ETH Zürich, Universitätstrasse 16, 8092 Zurich, Switzerland
First published on 4th February 2021
The electrochemical valorization of glycerol, a by-product from biodiesel production, has received significant attention, yet systems for the efficient reforming of glycerol that are based on non-precious metals have rarely been reported. Here, we introduce tungsten-doped bismuth vanadate (W:BiVO4) electrodes combined with an atomic-layer-deposited nickel (oxy)hydroxide (NiOx(OH)y) co-catalyst, as a promising photoanode material for the photoelectrochemical (PEC) oxidation of glycerol. To reveal trends in the reaction kinetics and selectivities, glycerol oxidation reaction (GOR) was investigated in varying electrolytes and at different applied biases. The photoanode developed in our study provides a rare example of the efficient production of the high value-added products, dihydroxyacetone (DHA), glyceraldehyde (GALD), and glycolaldehyde (GCALD), in the absence of precious metal catalysts. Under optimized conditions, W:BiVO4 with a NiOx(OH)y co-catalyst features oxidation currents and onset potentials for glycerol/water oxidation that are on par with state-of-the-art transition-metal-oxide photoanodes employed for the reforming of organic species, which marks an important step towards affordable solar-driven electrolyzers and direct alcohol fuel cells.
The use of semiconductor electrodes in a photoelectrochemical (PEC) setup allows to reduce considerably the reaction overpotentials, paving the way towards unbiased solar-driven photocatalysis.17,18 Bismuth vanadate (BiVO4) has been identified as one of the most promising photoanodes, exploited initially for water splitting.19 In the monoclinic phase, BiVO4 is the photoactive n-type semiconductor with a bandgap of 2.4 eV,20,21i.e. it absorbs in the visible light range. However, unmodified BiVO4 features low catalytic activity for photoelectrochemical water oxidation. To enhance kinetics, improve charge transport characteristics and reduce surface recombination of the photo-generated charge carriers in BiVO4, different strategies have been employed, i.e. aliovalent doping or deposition of co-catalysts for the oxygen evolution reaction.22 As a result, the range of application for BiVO4-based photocatalysts expanded from water oxidation20 to other photo(electro)catalytic transformations, such as the oxidation of 5-hydroxymethylfurfural, degradation of organic dyes or CO2 conversion.23–25 Despite these advances, BiVO4 remains underutilized for the reforming of abundant organic feedstock molecules to H2 and partially-oxidized value-added organic chemicals.
In the present study, we evaluate the performance of BiVO4-based photoanodes for the photoelectrocatalytic reforming of glycerol, C3H5(OH)3. Glycerol, a side product of biodiesel production, is a highly abundant and inexpensive chemical26 formed during the transesterification of triglycerides with methanol (one mole of glycerol is generated per three moles of biodiesel,27 which corresponds to ca. 100 kg of glycerol produced per ton of biodiesel). Oxidation, or the partial dehydrogenative reforming of glycerol, can in principle generate a plethora of the high value-added chemicals, e.g. 1,3-dihydroxyacetone, glyceraldehyde, tartronic acid, glycolic acid or hydroxypyruvic acid (Scheme 1).28 On the other hand, the complete dehydrogenative decomposition of one mole of glycerol can generate up to four moles of hydrogen at a relatively low energy input (ΔG = 3.9 kJ molglycerol−1).16 Therefore, glycerol has a potential to be used in both H2 generating electrolyzers and direct alcohol fuel cells.29,30
![]() | ||
Scheme 1 Products of electrochemical glycerol reforming identified in the literature. Products found in this study are shown in bold. |
Here, we present the study of the photoelectrochemical glycerol oxidation on BiVO4-based photoanodes in different pH environments. Our work demonstrates that tungsten doping in the BiVO4 structure, aided by the deposition of a nickel oxyhydroxide (denoted in the text as NiOx(OH)y, see discussion below) co-catalyst layer, notably improves the kinetics of glycerol oxidation on the photoanode. We focused on the development and optimization of the nickel-based coatings as the latter demonstrated promising electro- and photoelectrocatalytic activity and stability in a related reaction of water oxidation31,32 as well as in electrocatalytic glycerol oxidation.33 In addition, the pH of the reaction media substantially affects the distribution of products, e.g. formation of glyceraldehyde (GALD) is only observed in acidic conditions. Glycolaldehyde (GCALD), the simplest α-hydroxycarbonyl synthon for sugar formation, to the best of our knowledge, was for the first time observed in PEC glycerol oxidation. Our study clarifies the effects of different reaction variables (e.g. pH, applied bias, presence of co-catalysts) on the kinetics of PEC glycerol reforming. Thus, tungsten-doped bismuth vanadate (W:BiVO4) with an atomic-layer-deposited NiOx(OH)y co-catalyst achieved a photocurrent of 4.2 mA cm−2 at 1.2 VRHE (0.5 M Na2SO4), which is on par with the most active photocatalysts employed to date for the decomposition of organic substrates and biomass derivatives.34 Overall, our approach represents an effective, scalable strategy for producing solar hydrogen and offers a pathway to valorise abundant low cost glycerol.
X-ray diffraction (XRD) pattern of as-synthesized W:BiVO4 matches that of a monoclinic scheelite reference (no. 01-083-1699, ICDD database, Fig. 1a). W:BiVO4 has a porous structure according to scanning electron microscopy imaging (SEM, Fig. S2†). No additional peaks were observed in the diffraction pattern of W:BiVO4, after atomic-layer-deposition of nickel coating (Fig. 1a).
![]() | ||
Fig. 1 (a) XRD patterns of BiVO4 (ICDD no. 01-083-1699), W:BiVO4 and as-deposited NiOx(OH)y/W:BiVO4. The peaks corresponding to the FTO substrate are indicated by squares. (b) HRTEM, HAADF-STEM, and EDX elemental mapping images of electrochemically activated NiOx(OH)y/W:BiVO4 electrodes. Overlay of the (c) Ni 2p and (d) O 1s core level XPS signals of as-deposited and electrochemically activated NiOx(OH)y/W:BiVO4. The feature denoted as intersite (856.9 eV) is due to defects in the structure, as discussed elsewhere.39 |
Nickel-coated W:BiVO4 electrodes were subjected to five anodic scans, from −0.05 to 1.50 VRHE in 0.5 M potassium borate buffer solution (KBi, pH = 9.3) (Fig. S3a†), and then used for glycerol oxidation. Such electrochemical activation yields NiOx(OH)y species in the form of uniformly distributed crystalline nanoparticles on the surface of W:BiVO4, according to high-resolution transmission electron microscopy (HRTEM), high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) and energy dispersive X-ray spectroscopy (EDX) elemental mapping (Fig. 1b). XRD of anodically-activated NiOx(OH)y/W:BiVO4 electrodes is identical to that of the as prepared NiOx(OH)y/W:BiVO4 (Fig. S3b†). A negligible difference in the absorption spectra of NiOx(OH)y/W:BiVO4 and W:BiVO4 indicates that ALD-deposited NiOx(OH)y does not alter noticeably light absorption properties of the photoanode (Fig. S3c†).
The surface composition and valence states of as-deposited and anodically activated NiOx(OH)y/W:BiVO4 were probed by X-ray photoelectron spectroscopy (XPS). The Ni 2p core level XPS spectrum of as-deposited NiOx(OH)y/W:BiVO4 reveals features ascribed to metallic Ni, Ni2+ and Ni3+ at 852.9, 854.7 and 855.7 eV, respectively, while an additional feature at higher binding energy (denoted as intersite, 856.9 eV) can be ascribed to the presence of defects in the nickel (oxy)hydroxide structure (Fig. 1c).39 As expected, anodic activation increased the relative amount of oxidized nickel species (NiOx(OH)y), evidenced by the much enhanced intensity of the feature at 855.7 eV.
The O 1s core level spectrum (Fig. 1d) of as-deposited and activated NiOx(OH)y/W:BiVO4 films can be deconvoluted into four components: lattice oxygen (OL, 529.5 eV), lattice oxygen sites located in the vicinity of oxygen vacancies/defects (OV, 531 eV), surface oxygen species (Osurf, 532 eV), and adventitious species (weakly bound adsorbed water/CO2, Oadv).40,41 Comparison of the spectra of the as-deposited and activated films reveals a distinct increase in the concentration of surface oxygen groups for the electrochemically activated films, i.e. an increase from 2 to 42%, consistent with the formation of layered nickel oxyhydroxide species with a highly exposed surface (Table S1†).42
Fig. 2a and b show the linear sweep voltammetry (LSV) profiles under dark and illumination conditions for W:BiVO4 and NiOx(OH)y/W:BiVO4 in 0.5 M KBi and 0.5 M Na2SO4 electrolytes. In Na2SO4 solution, the PEC performance of W:BiVO4 is characterized by an onset potential of ca. 0.43 VRHE and a photocurrent for water oxidation of 2.3 mA cm−2 at 1.2 VRHE (in the absence of glycerol). The NiOx(OH)y co-catalyst enhances the photocurrent to 3.3 mA cm−2 (1.2 VRHE), leading to a ca. 50 mV cathodic shift of the onset potential. In the presence of glycerol (0.1 M), the photocurrent increases in the entire potential range, reaching 4.2 mA cm−2 at 1.2 VRHE.
![]() | ||
Fig. 2 Linear sweep voltammetry of W:BiVO4 and NiOx(OH)y/W:BiVO4 in (a) 0.5 M Na2SO4 and (b) 0.5 M KBi with and without the addition of 0.1 M glycerol under dark and AM 1.5 G, 100 mW cm−2 illumination; scan rate = 10 mV s−1. (c) Chopped chronoamperometry plots of W:BiVO4 and NiOx(OH)y/W:BiVO4 in 0.5 M KBi with (dotted lines) and without (solid lines) 0.1 M glycerol at 0.3 VRHE, 0.5 VRHE, and 0.8 VRHE. The color code is the same as on panels a and b. (d) Plots of ln![]() |
In mild-alkaline conditions (0.5 M KBi), W:BiVO4 and NiOx(OH)y/W:BiVO4 exhibit ca. 50 mV cathodic shift of the onset potential on the RHE scale relative to that in the Na2SO4 electrolyte (Fig. 2b). The photocurrent for glycerol oxidation on NiOx(OH)y/W:BiVO4 electrodes reaches 2.5 mA cm−2 (0.8 VRHE) and increases to 3.5 mA cm−2 at 1.2 VRHE, thus matching or exceeding the performance of some transition-metal-oxide-based PEC systems (e.g. TiO2) employed for the oxidation of organic alcohols (methanol, ethanol or glycerol).34 Control experiments show a high reproducibility of the data collected on W:BiVO4 and NiOx(OH)y/W:BiVO4 photoelectrodes (Fig. S3d†). Overall, the enhancement of the catalytic activity of NiOx(OH)y/W:BiVO4 for water and glycerol oxidation in mildly alkaline conditions relative to neutral conditions may originate from the pH-dependence of the electrocatalytic activity of the NiOx(OH)y co-catalyst layer, which is a typically observed phenomenon for nickel or cobalt-based oxides and (oxy)hydroxides.43,44
The charge-transfer dynamics at the semiconductor–electrolyte interface (SEI) was analysed by the transient photocurrent profile at a constant potential. Here, we used alkaline 0.5 M KBi conditions and relatively low biases, i.e. 0.3 VRHE, 0.5 VRHE, and 0.8 VRHE (Fig. 2c). The initial transient photocurrent, Iin, is caused by the separation of the photogenerated hole–electron pairs, where trapping or recombination of charge carriers takes place by surface states or reduced species in an electrolyte.45 The continuous decay of the photocurrent indicates ongoing recombination until a steady state, Ist, is reached. When the light is off, back-reaction of electrons at the conduction band with accumulated holes results in a negative current spike.
In the absence of glycerol, the steady-state photocurrent at 0.3 VRHE (i.e. close to the onset potential) is nearly zero for W:BiVO4 and NiOx(OH)y/W:BiVO4 electrodes, coinciding with the onset potential observed in the LSV profile. In the presence of glycerol, a distinct non-zero photocurrent is detected reaching 80 μA cm−2 (0.3 VRHE), 0.53 mA cm−2 (0.5 VRHE) and 1.22 mA cm−2 (0.8 VRHE) for the W:BiVO4 electrodes (Fig. 2c). In addition, the back-reaction current was significantly suppressed at 0.3 VRHE, likely indicating a reduced accumulation of holes at the material surface. In the presence of glycerol, the photocurrent of the NiOx(OH)y/W:BiVO4 electrodes increased to 0.2 mA cm−2 (0.3 VRHE), 1.2 mA cm−2 (0.5 VRHE) and 2.9 mA cm−2 (0.8 VRHE). Note that all the studied potentials are below the thermodynamic potential of the water oxidation, 1.23 VRHE.
The photocurrent relaxation dynamics were also analysed using the transient decay time τ, expressed by a logarithmic plot of the parameter D (eqn (1)) versus time, where It is the current at time t.45
![]() | (1) |
The transient decay time is defined as time at which lnD = −1. The value of τ reflects the electron lifetime within bulk semiconductor. BiVO4 is known to undergo facile recombination of charge carriers due to sluggish charge-transfer kinetics at the SEI.46 Since the accumulated holes at the surface facilitate electron–hole recombination leading to a reduced lifetime of the separated charges, a facilitated hole transfer to the reagents in the electrolyte will prolong the lifetime of the separated electrons. At 0.5 VRHE, W:BiVO4 suffers from severe surface recombination, resulting in τ of ca. 1.3 s. Upon addition of glycerol acting as a hole acceptor, an enhanced charge transport from the near-surface region to the electrolyte gives rise to a near three-fold enhancement of τ reaching 4.0 s (Fig. 2d). Overall, these observations of an increased photocurrent and a longer transient decay time imply faster kinetics of the photoelectrochemical glycerol oxidation reaction as compared to the oxygen evolution reaction (OER) in water.
The analysis of the hole-injection efficiency (ηinj, see ESI and Fig. S4 for details†) for glycerol oxidation on the studied photoanodes further corroborates this conclusion. The sulfite anion (SO32−) is a typically used hole scavenger, whose ηinj can be taken as a reference value (100%).47 The hole injection efficiency for water oxidation reaches ca. 55% and 60% on W:BiVO4 and NiOx(OH)y/W:BiVO4 electrodes, respectively, at 1.2 VRHE (Fig. 2e). In the presence of glycerol (0.1 M), NiOx(OH)y/W:BiVO4 photoelectrode exhibits a superior hole transfer to glycerol than W:BiVO4 throughout the whole potential window, reaching an efficiency of 72% at 1.2 VRHE, that confirms the enhancement of the kinetics of glycerol oxidation with respect to OER.
While the excellent stability of the W:BiVO4 electrodes in the presence of glycerol is observed at both low (0.8 VRHE) and high (1.2 VRHE) biases, the degradation of NiOx(OH)y/W:BiVO4 electrodes in 0.5 M KBi takes place, as the loss of 21% (0.8 VRHE) and 33% (1.2 VRHE) in the photocurrent after 10 h is observed (Fig. S6†). We attribute this decrease to a gradual dissolution of the nickel catalyst, however the poisoning of the surface by adsorbed reaction intermediates or by-products can also take place.49,50 Notably, in the absence of glycerol, the photoelectrochemical stability of NiOx(OH)y/W:BiVO4 in 0.5 M KBi is substantially higher (Fig. S7†). Inductively coupled plasma optical emission spectrometry (ICP-OES) was used to analyse the electrolyte composition after the reaction under various conditions (Fig. S8†). In the presence of glycerol, we observed a negligible dissolution of Bi and V into the electrolyte (≤0.6% of the total loading) from W:BiVO4 electrodes in 0.5 M KBi, while somewhat higher values (e.g. ≥6% loss of Bi and V after continuous GOR at 1.2 VRHE) are observed in 0.5 M Na2SO4. In contrast to bismuth vanadate, the NiOx(OH)y overlayer proved to be unstable under the reaction conditions, e.g. 10 h of continuous electrolysis in 0.5 M KBi resulted in a loss of >90% of the nickel content. However, in 0.5 M Na2SO4, the rate of nickel dissolution decreases significantly, i.e. to 35% (0.8 VRHE) and 24% (1.2 VRHE) after 10 h of electrolysis. XPS analysis on NiOx(OH)y/BiVO4 electrode revealed absence of the noticeable changes of the chemical states of nickel after GOR in Na2SO4 (Fig. S9†).
HRTEM imaging of the electrodes after 10 h of glycerol oxidation in 0.5 M KBi (Fig. S10†) confirms that BiVO4 maintains its original morphology. Consistent with ICP data, the used NiOx(OH)y/W:BiVO4 electrodes reveal no detectable amounts of nickel on the surface after the PEC test. Thus, the dissolution of the nickel co-catalyst layer coincides with a decrease of the photocurrent during constant potential electrolysis, serving as an indirect evidence for the crucial role of nickel surface species in catalysing glycerol oxidation on the photoanode (Fig. S6†).
To summarize, atomic-layer-deposited NiOx(OH)y enhances the OER kinetics of W:BiVO4 as reflected by an increased photocurrent and transient decay time. In the presence of glycerol, both W:BiVO4 and NiOx(OH)y/W:BiVO4 exhibit a high activity towards glycerol oxidation. NiOx(OH)y/W:BiVO4 electrodes demonstrate appreciable photoelectrochemical stability under PEC glycerol oxidation conditions, although the gradual dissolution of the nickel cocatalyst was observed in the presence of glycerol that resulted in a continuous decay of the catalytic currents for NiOx(OH)y/W:BiVO4 electrodes over time in alkaline condition. An increase of the current despite the dissolution of the NiOx(OH)y layer during photoelectrolysis in a Na2SO4 solution can be attributed to the gradual acidification of the medium, which can have a positive effect on the reaction kinetics.
1H and 13C spectra collected for the NiOx(OH)y/W:BiVO4 electrode at 1.2 VRHE in 0.5 M Na2SO4 exhibit complex patterns (δH ∼3.5–4.5 ppm and δC ∼60–100 ppm) (Fig. S11a and b†), in which new signatures in addition to the previously identified peaks of GCALD, GALD, and DHA are found. Presumably, these signatures are due to oxygenated species that we are currently unable to identify. Previous studies have shown that GCALD, GALD, and DHA form complex equilibria of monomers and oligomers in aqueous solutions.53–55 Analysis of the 1H NMR spectrum of the product mixture obtained for the NiOx(OH)y/W:BiVO4 photoanode at 1.2 VRHE assumes a presence of GCALD, GALD, and DHA (by matching with the signals of the respective reference chemicals, Fig. S14†), however 13C NMR of the same solutions (Fig. S11b†) showed that GCALD, GALD, and DHA do not account for all of the signals. A 1H–1H Correlation Spectroscopy (COSY) analysis for this mixture showed many overlapping signals making its interpretation difficult, and heteronuclear single quantum coherence spectroscopy (HSQC) (Fig. S15†) and heteronuclear multiple bond correlation (HMBC) analysis failed to show correlations for the 1H signals of interest. Thus, further work is necessary to identify the as yet undetermined specie(s).
The effect of the NiOx(OH)y co-catalyst layer and the applied bias on the production rate (R) and faradaic efficiency (FE) of the detected products of the glycerol oxidation is illustrated in Fig. 3, based on 1H NMR spectra and chromatography results. The distribution of the products of the electrolysis reaction depends on the applied potential, type of electrolyte and the electrode material (Table 1 and S19†). Specifically, in the 0.5 M Na2SO4, the average production rate of DHA, GALD, GCALD and FA increases with increasing potential when using the W:BiVO4 electrode (Fig. 3a). At 0.8 VRHE, FA is a primary product of glycerol oxidation on W:BiVO4 electrodes with a FE of over 70% and a production rate of ca. 54 mmol h−1 m−2, while at the higher potential of 1.2 VRHE, a four-fold and five-fold enhancement of the production rate of DHA (70 mmol h−1 m−2) and GCALD (ca. 60 mmol h−1 m−2), corresponding to FE's of 19% and 11%, is observed.
Electrolyte | Electrode | Applied bias | Identified productsa | Enhanced rate and FE of specific products observed with increased potential |
---|---|---|---|---|
a Reaction conditions: 0.1 M glycerol, electrode area 2.8 cm2, volume of anode part 33 ml, temperature at 25 °C (DHA: dihydroxyacetone, GALD: glyceraldehyde, GCALD: glycolaldehyde dimers, FA: formate, and FMALD: formaldehyde). b Product that is identified by 1H NMR only. c Product that is identified by 13C NMR only, hence the quantitative analysis is not available. d Product that is identified by 1H NMR; however, its chemical shifts overlap with GCALD. | ||||
Na2SO4 | W:BiVO4 | 0.8 VRHE | GALD, DHA, GCALD, FA, FMALD | R DHA, RGALD, RGCALD, RFA, FEDHA, FEGALD, FEGCALD |
1.2 VRHE | GALD, DHA, GCALD, FA, FMALD | |||
NiOx(OH)y/W:BiVO4 | 0.8 VRHE | GALD, DHA, GCALD, FA, FMALD | R DHA, RGCALD, RFA | |
1.2 VRHE | GALD, DHA, GCALD, FA, FMALD | |||
KBi | W:BiVO4 | 0.8 VRHE | GALDb,d, GCALD, FAb, FMALDc | R FA |
1.2 VRHE | GALDb,d, GCALD, FAb, FMALDc | |||
NiOx(OH)y/W:BiVO4 | 0.8 VRHE | GALDb,d, GCALD, DHAb, FAb, FMALDc | R DHA, RGCALD, RFA, FEDHA, FEFA | |
1.2 VRHE | GALDb,d, GCALD, DHAb, FAb, FMALDc |
For both 0.5 M Na2SO4 and 0.5 M KBi electrolytes, the deposition of a NiOx(OH)y co-catalyst layer results in an enhanced catalytic glycerol oxidation reflected in the higher photocurrents (higher total charge transfer) and product yields. For instance, at an electrolysis potential of 1.2 VRHE in 0.5 M Na2SO4, ca. two-fold increase of the photocurrent was achieved upon deposition of a NiOx(OH)y co-catalyst. In addition, an increase in the applied bias increases the production rate of the value-added DHA in the presence of NiOx(OH)y. Specifically, in 0.5 M KBi, a three-fold enhancement of production rate of DHA (138 mmol h−1 m−2) was achieved at 1.2 VRHE (compared to 0.8 VRHE) along with an improved FE of 19% (Fig. 3b and d).
Product analysis and identification enabled the proposal of reaction path(s) for the photoelectrocatalytic glycerol decomposition using W:BiVO4 and NiOx(OH)y/W:BiVO4 electrodes. It is proposed that DHA and GALD isomers are the first stable species to form upon 2e− oxidation of one of the hydroxyl moieties of glycerol.51,57 A higher yield of DHA compared to GALD (Fig. 3 and S19†) corroborates with DFT calculations that have identified DHA as the thermodynamically more stable product than GALD.58 GCALD is proposed to form from isomeric DHA/GALD species through C–C bond cleavage, releasing one equivalent of FA. Subsequent C–C bond cleavage to form one more equivalent of FA and FMALD from GCALD is proposed as one of the further steps. The presence of hydrated GCALD was confirmed by HPLC and NMR spectroscopy (identification of its exact structure(s) in the solution, i.e. hydrated monomers, dimers or oligomers,59 is beyond the scope of this work).
In 0.5 M Na2SO4, PEC glycerol oxidation using a W:BiVO4 electrode leads to higher amounts of GALD and GCALD at increased applied bias (1.2 VRHEvs. 0.8 VRHE). In contrast, NiOx(OH)y/W:BiVO4 electrodes preferentially produce GALD and GCALD at a lower potential, 0.8 VRHE, presumably due to their overoxidation at higher applied voltage. Although the deposition of a NiOx(OH)y co-catalyst does not appear to alter the general reaction pathway, the presence of a nickel co-catalyst strongly influences the product distribution (which is also affected by the applied bias and reaction media). For instance, NiOx(OH)y/W:BiVO4 electrodes show an increased selectivity toward DHA formation over GCALD at 1.2 VRHE in alkaline KBi buffer. In contrast to a recent report, in which it was argued that the selective formation of DHA from glycerol is feasible,60 our findings indicate that multiple products inevitably form upon glycerol photoelectrooxidation, emphasizing in turn the importance of the use of complementary analytical tools (e.g. HPLC and NMR spectroscopy) for an accurate quantitative analysis of the complex product mixture (see also Fig. S27†).
Footnote |
† Electronic supplementary information (ESI) available. See DOI: 10.1039/d0ta10480a |
This journal is © The Royal Society of Chemistry 2021 |