Camille
Bakkali-Hassani
*a,
Dimitri
Berne
a,
Pauline
Bron
ab,
Lourdes
Irusta
b,
Haritz
Sardon
b,
Vincent
Ladmiral
a and
Sylvain
Caillol
*a
aICGM, Univ Montpellier, CNRS, ENSCM, Montpellier, France. E-mail: camille.bakkali-hassani@umontpellier.fr; sylvain.caillol@enscm.fr
bPOLYMAT, Department of Polymers and Advanced Materials: Physics, Chemistry and Technology, University of the Basque Country UPV/EHU, Donostia-San Sebastian, Spain
First published on 24th July 2023
Various bases (DMAP, DBU, TBD, t-BuOK), acid (p-TSA), thiourea (TU) and organometallic Lewis acid (DBTDL) were investigated as potential catalysts for the preparation of polyhydroxyurethane covalent adaptable networks. Catalytic systems were first selected for their ability to promote cyclic carbonate aminolysis quantitatively (full conversion of cyclic carbonates) with few or no side reactions (urea formation). Selected PHU networks were extensively characterized using thermo-mechanical analysis (TGA, DSC, DMA and tensile test), rheology experiments (stress relaxation, frequency sweep), spectroscopy analysis (ATR-IR), swelling and reprocessing tests. Combining rheology, ATR-IR analysis and model molecular reactions, we suggest a catalyst-dependent exchange mechanism in which solely the organotin Lewis acid (DBTDL) was capable to promote transcarbamoylation in PHU efficiently with both secondary (major product of aminolysis) and primary alcohols and thus an efficient reprocessing.
Polyhydroxyurethane (PHU) chemistry and derivatives have been extensively studied to replace PU produced from isocyanate monomers in the so-called non-isocyanate polyurethane (NIPU) strategy.10,11 In classical PU chemistry, carbamates are generated by isocyanate alcoholysis whereas hydroxyurethane moieties result from the ring-opening of cyclic carbonate (CC) by primary amine. This approach brings several advantages such as the use of non-toxic and user-friendly monomers generated from CO2 and epoxides, high stability (isocyanate reacts with nucleophiles such as water, alcohols, etc.) and compatibility with green chemistry principles.12–15 However, despite their elegant and green synthesis, PHUs exhibit synthetic drawbacks which hinder their industrial application.16 The main limitation is the relatively low reactivity of bis-5-CC towards aminolysis, the reversibility of carbonate aminolysis, the occurrence of side reactions (mainly urea formation) and the formation of a dense hydrogen bond networks which were shown to limit the conversion and consequently the molar masses of such linear polymers.16 To increase the reactivity of cyclic carbonates towards aminolysis, different strategies have been explored such as increasing the CC ring size or installing electron-withdrawing substituent on the heterocycle.17–19
For thermoset application and especially for the design of dynamic covalent networks (CANs), two distinct exchange mechanisms are commonly admitted to occur in PU and PHU crosslinked networks.20 On the one hand, associative transcarbamoylation, also called transurethanisation, is the exchange reaction between carbamate linkages and free-hydroxyl groups (Scheme 1a). On the other hand, dissociative transcarbamoylation can occur through two mechanisms depending on the presence of free-hydroxyl groups at specific position along the polymer backbone. In PU-CANs, as no free-hydroxyl groups are available, it is the retro-formation of isocyanate and alcohol moieties which confer dynamic properties to PU network. In PHU-CANs, it is the retrocyclisation of hydroxyurethane moiety which yield to primary amine and cyclic carbonate that has been put forward to explain the dynamic exchange occurring in 5CC-PHU covalent adaptable networks (Scheme 1b).21 Often described as sluggish, transcarbamoylation (whether associative or dissociative) is known to occur in both PU and PHU networks at relatively high temperature (T ≥ 120 °C), with or without the addition of an exogenous catalyst.20 Due to a relatively poor control (side reaction observed at high temperature and slow processes), polymer chemists have first set aside those reactions to focus on introducing various dynamic covalent bonds into PHU crosslinked networks.22 Nonetheless, reprocessable PHUs based on transcarbamoylation via reversible cyclisation have been obtained from CC of different ring-size (from 5 to 8).23–25 In the particular case of 5-CC monomers, the retro-formation of primary amine and cyclic carbonate was first considered as a limitation, as side reactions leading to a decrease in crosslink density were detected.22 Torkelson and co-workers, by judiciously selecting non-volatile primary amine monomers and an organocatalyst (4-(dimethylamino) pyridine, DMAP), were able to produce PHU-CAN with good reprocessability at moderate temperature (120 °C).21,26–28 Apart from DMAP-based polyhydroxyurethane CANs (and two studies on PU network29,30), investigations on suitable catalysts, i.e. able to efficiently catalyse both aminolysis and transcarbamoylation with low level of or no side reactions, towards the control of covalent exchange properties (dissociative/associative mechanism) are still lacking.
Scheme 1 (a) Associative and (b) dissociative transcarbamoylation mechanism suggested for PHU covalent adaptable networks. |
Herein, we studied the influence of catalysis in PHU covalent adaptable networks to determine its effect on both network formation (kinetic and side reaction) and final material properties. A catalyst selection was first performed by following the aminolysis of cyclic carbonate and the occurrence of side reactions (urea formation) by ATR-IR spectroscopy during and after the curing process at 80 °C. Rheological analysis (frequency sweep and stress relaxation) combined with reprocessing test highlighted that only one candidate (a tin-based catalyst) was able to catalyse efficiently transcarbamoylation. 1H NMR kinetic experiments on model molecule combined with rheology experiments (stress relaxation and frequency sweep) and ATR-IR spectroscopy analysis suggested that the rate and the mechanism of exchange was influenced by both the catalyst and alcohol nature (primary vs. secondary).
Prior to study the influence of the catalyst in PHU formulations, the initial ratio of primary amine/5-CC ([NH2]0/[5-CC]0) was optimized in non-catalysed formulations to ensure complete conversion of cyclic carbonate (Table 1 and Fig. S1†). A slight excess of primary amine (1 mol%) was found necessary to reach full conversion of the cyclic carbonate. However, increasing this initial ratio leads to the formation of urea (by reaction of free primary amine with urethane moieties) easily discernible in ATR-IR spectra at 1650 cm−1 when [NH2]0/[5-CC]0 ≥ 1.05 (Fig. S1†). The initial ratio [NH2]0/[5-CC]0 = 1.01 was thus fixed for all the following experiments.
Catalyst | [NH2]0/[5-CC]0 | k Cat/kNCa | Urea?b | Swelling indexc (%) | Gel contentc (%) |
---|---|---|---|---|---|
a Relative polymerisation rate constant (normalized by non-catalysed polymerisation constant rate) and measured from ln([5-CC]0/[5-CC]t) vs. time plot presented in Fig. S2.† b Residual cyclic carbonate and urea bonds content were evaluated using ATR-IR spectroscopy. c Measured after immersion in DMF at 80 °C for 24 h. d DBTDL is used at 2 mol%; n.d. = not determined. | |||||
None | 1.00 | n.d. | Low | n.d. | n.d. |
None | 1.01 | 1.00 | Low | 218 ± 1 | 91 ± 0.5 |
None | 1.02 | n.d. | Medium | n.d. | n.d. |
None | 1.05 | n.d. | High | n.d. | n.d. |
p-TSA | 1.01 | 2.65 | Low | 230 ± 14 | 92 ± 1 |
DMAP | 1.01 | 0.82 | Low | 213 ± 5 | 90 ± 2 |
TU | 1.01 | 2.58 | Low | 245 ± 2 | 89 ± 3 |
DBU | 1.01 | 1.96 | Medium | 240 ± 10 | 85 ± 1 |
TBD | 1.01 | 3.19 | High | 198 ± 2 | 87 ± 1 |
t-BuOK | 1.01 | 0.78 | Low | 230 ± 15 | 93 ± 0.5 |
DBTDLd | 1.01 | 2.72 | Low | 226 ± 2 | 86 ± 0.5 |
The catalyst activity towards the aminolysis of cyclic carbonate was investigated by monitoring the curing process at 80 °C using ATR-IR spectroscopy. Conversions were calculated following the simultaneous disappearance of νCO carbonate vibration band at ∼1780 cm−1 and the appearance of urethane-carbonyl vibration band at ∼1680 cm−1 (Fig. 1b and c). The relative rate constants of catalysed (cat.) and non-catalysed (NC) polymerisation (kcat/kNC) were obtained by plotting ln([5-CC]0/[5-CC]t) vs. time (Fig. S2,† and Table 1) and allowed the distinction of two catalyst groups (assuming a first order kinetic). Only time points corresponding to the initial rate of the reactions were considered (conversion < 50%) for the determination of the rate constants. On one hand, in the presence of DMAP or t-BuOK, aminolysis rates were comparable to that of non-catalysed (NC) formulations (Fig. 1c), thus demonstrating a low catalytic activity. In the case of t-BuOK, the poor dispersion quality (catalyst solid particles were visible to the naked eyes) could also explain the low impact of this catalyst on the aminolysis rate. On the other hand, p-TSA, DBTDL, TBD, DBU and TU significantly accelerated the ring-opening reaction at the early stage of polymerisation (from 0 to 5 min). In any case, after 10 min at 80 °C, catalysed and non-catalysed curing exhibited roughly the same conversion profile meaning that the polymerisation process tends to be limited by the diffusion of reactive species (zero-order kinetic profile). All formulations were maintained at 80 °C for 24 h to ensure high conversion (>90%).
Apart from these kinetics considerations, discrepancies also arose between samples and in particular regarding the presence/absence of urea detected in ATR-IR spectra at 1650 cm−1 (Table 1 and Fig. S3†) in the fully cured material. Strong bases such as TBD or DBU (pKa = 26.0 and 24.3 respectively in acetonitrile31) appeared to favor the formation of urea probably because of the strong amine activation by such organocatalysts (Scheme 2c). This effect was particularly striking for TBD-based materials in which the intensity of the urethane carbonyl stretching observed in ATR-IR spectrum was even lower than the one from urea (Fig. S3†). With linear PHU, TBD was also reported to unexpectedly catalysed the formation of segmented polyurea-urethane32 and can be therefore considered as non-suitable for PHU-CAN preparation (starting from 5-CC monomers). Alternatively, TBD was also shown to activate carbamate groups through an acyl intermediate depicted in Scheme 2d.33 In contrast, mild bases such as DMAP or t-BuOK (pKa = 18.0 and 17.1 respectively34) did not lead to side reactions under these conditions. In the presence of electrophile activators such as TU, DBTDL or p-TSA, polymerisation rates were higher than for the non-catalysed process and no-or only small quantities of urea bond were detected at the end of the synthesis. Contrary to the base-catalysed mechanism (Scheme 2a), acid activates preferentially the carbonyl group of 5-CC (Scheme 2b) thus limiting undesired side reaction (e.g. formation of urea displayed in Scheme 2c). DBTDL due to its Lewis acid character also increased the aminolysis reaction rate following the same activation pathway on the carbonyl group (Scheme 2b). Following these results based on the reactivity towards aminolysis, the strong bases (TBD and DBU), despite their good activity, were put aside because of their tendency to promote urea formation. Catalyst-free PHU network as well as those containing DBTDL, TU, p-TSA, t-BuOK and DMAP were selected for the rest of the study.
Scheme 2 (a) Basic activation of amine and (b) electrophilic activation of carbonate by acids for cyclic carbonate aminolysis (c) proposed mechanism for TBD-catalysed urea formation from hydroxyurethane moieties32 and (d) from carbamate activation.33 |
The formation of a chemically crosslinked network was assessed via swelling experiments (in DMF at 80 °C for 24 h). Regardless of the catalyst employed, gel contents were ≥85% for all the samples. The swelling values also show no relationship neither with the conversion nor the urea formation. This is surprising knowing the fact that side reactions leading to urea formation were detected in DBU or TBD-based systems, the latter exhibiting slightly lower swelling index but not significant.
PHU | T g (°C) | T d 5%b (°C) | G′c (MPa) at 80 °C | Young modulusd (MPa) | Stress at breakd (MPa) | Strain at breakd (%) |
---|---|---|---|---|---|---|
a Measured by DSC (second heating curve). b Measured by TGA. c G′ (shear modulus) was measured by frequency sweep experiments at 80°. d Calculated from tensile test experiments. | ||||||
DBTDL | 10 | 267 | 0.44 | 2.0 ± 0.2 | 0.53 ± 0.10 | 35 ± 3 |
TU | 10 | 256 | 0.34 | 2.7 ± 0.5 | 0.52 ± 0.06 | 28 ± 2 |
DMAP | 11 | 285 | 0.26 | 3.8 ± 0.1 | 0.97 ± 0.08 | 33 ± 4 |
p-TSA | 9 | 276 | 0.40 | 3.3 ± 0.5 | 0.72 ± 0.08 | 30 ± 3 |
t-BuOK | 9 | 287 | 0.25 | 5.9 ± 0.3 | 0.96 ± 0.06 | 24 ± 1 |
NC | 8 | 283 | 0.26 | 3.5 ± 0.4 | 0.79 ± 0.06 | 30 ± 4 |
All networks demonstrated almost similar thermal stability (256 °C ≤ Td5% ≤ 286 °C) and only slight differences were noticeable between their respective thermograms (Fig. S4†). Differential scanning calorimetry (DSC) analyses showed, once again, that the catalysts did not affect the glass transition temperatures (Tg) which were for all samples in the 8–11 °C range. The catalysts employed in this study do not act as plasticizer and do not impact the thermal properties of the corresponding materials. The shear modulus in the rubbery state measured by frequency sweep analysis, showed only slight variation between materials with values between 0.26 MPa and 0.44 MPa (at T = 80 °C). These results combined with swelling data and from the rubber elasticity theory,35,36 confirmed that the network density of the PHU networks is not significantly affected by the catalysts or the occurrence of side reactions.
The mechanical properties of the PHU networks were then evaluated by tensile test experiments. As the experiments were performed at 25 °C, above the glass transition temperatures of the materials, the mechanical characteristics reported in Table 2 referred to the rubbery state of the materials. Hence, it is not surprising to obtain young modulus values with a magnitude order of the MPa. The slight differences observed can be related to slight difference of conversion or induced by urea formation already mentioned. In the case of PHU network prepared using t-BuOK as catalyst, a significant increase of mechanical properties was observed, the Young Modulus was roughly two times higher than that of the other samples. This behaviour was explained by the poor dispersion of the catalyst in the PHU matrix (and monomers) which can act as a solid charge in the crosslinked networks which improves the mechanical properties (see Fig. S5†). Overall and in agreement with the results observed via other characterization methods, there was no significant difference in thermoset properties between the different PHU samples. All analyses pointed out that the synthesis of cross-linked PHU network was effective and that the resulting properties were substantially independent of the catalyst employed during the preparation of the networks. The following dynamic evaluation is therefore not biased by the difference in structural or initial properties of the thermosets and is dictated only by the catalyst used.
Frequency sweep experiments performed at temperature between 80 and 160 °C, also highlighted different shear modulus evolution upon heating, depending on the catalyst used (Fig. 3). DBTDL-based networks shear modulus was stable with an increasing temperature (Fig. 3a). In contrast, for the material obtained with the other catalysts or in the NC-material (representative curve in Fig. 3b), the shear modulus tended to decrease as the temperature rose. These results suggest a more pronounced dissociative pathway for samples prepared with other catalysts than DBTDL.38 The ATR-IR spectra of DMAP-PHU networks after frequency sweep experiments (30 min at each temperature) presented in Fig. S8† showed the characteristic vibrations bands νCO of carbonate, urea and δN–H of primary amine with increasing intensity, thus confirming the retroformation of primary amine and cyclic carbonate and the occurrence of side reactions (urea bonds, see Scheme 2c) during the temperature treatment. ATR-IR analysis of DBTDL-PHU networks performed at 140 °C (1 spectrum per 10 min) show no evolution of characteristic carbonyl vibration bands in the same region (Fig. S9†). The dissociative mechanism cannot be ruled out for DBTDL-PHU networks as cyclic carbonate νCO was detected in ATR-IR after 3 reprocessing experiments (each of 8 h at 120 °C) but could be less predominant than for the other catalysts employed (Fig. S10d†).
Fig. 3 Frequency sweep experiments (parallel plate, d = 8 mm) of (a) PHU-CANs with DBTDL catalyst (2% mol) and (b) PHU-CANs prepared with TU catalysts (5 mol%) between 80 and 160 °C. |
The same tendency was observed in dissolution tests performed in 1,4 butanediol at 140 °C for one week (Table S1†). After only a few hours, DBTDL-based networks were solubilized while all the other materials tested remained insoluble (even after one week at 140 °C) and exhibited roughly the same swelling ratio and gel content between them (Table S1†). Finally, the DBTDL-PHU network was subjected to three consecutive reprocessing cycles (each of 8 h at 120 °C, 3 tons of pressure) and showed, despite a slight decrease of mechanical properties (measured by tensile test experiments), good reprocessability (Fig. S10†). The decrease of the mechanical properties recorded after each cycle was attributed to the formation of ureas and cyclic carbonates which reduce the crosslink density (Fig. S10d†). Swelling experiments of the three times-reprocessed sample (in DMF at 80 °C for 24 h) also showed a higher swelling index of 255 ± 5% and lower gel content 75% compare to the pristine PHU-DBTDL (swelling index = 226 ± 2 and 86% of gel content) thus confirming a reduced crosslink density. The typical pictures of the DBTDL-catalysed and NC materials, presented in Fig. S11† (after 24 h at 140 °C under 3 tons at the hot press), underlines the poor reprocessability which was observed for all other catalyst candidates under these conditions. Rheological properties, i.e. stress relaxation and frequency sweep, and ATR-IR analysis (at high temperature, Fig. S8 and S9†) also indicated that solely PHU containing the organometallic Lewis acid DBTDL effectively performed associative transcarbamoylation. The dynamic properties observed with the other catalysed PHU CANs are likely caused by dissociative transcarbamoylation exchange through the retro-formation of the cyclic carbonate and the amine functions, which is prone to irreversible urea formation. This combination of data reflects the low efficiency of catalysts, yet reported to be active in similar system, towards transcarbamoylation and prompted us to investigate different molecular model reactions.
Catalyst | I/II alcohola (%) | % conv. (I)b | % conv. (II)c | Product/urea (I)b | Product/urea (II)c |
---|---|---|---|---|---|
Determined by 1H NMR spectroscopy in CDCl3.a After 24 h at 80 °C in bulk.b After 72 h at 140 °C in bulk using hexanol (5 eq).c After 72 h at 140 °C in bulk using 2-heptanol (5 eq).d DBTDL was used at 2 mol% to [TMPTC]. n.d. = not determined. | |||||
NC | 33/67 | 3 | 1 | n.d. | n.d. |
t-BuOK | 28/72 | 100 | 65 | 81/17 | 30/35 |
TBD | 17/83 | 90 | 84 | 83/7 | 39/45 |
TU | 20/80 | ≤1% | ≤1% | n.d. | n.d. |
DBU | 26/74 | 80 | 57 | 47/33 | 17/40 |
DBTDLd | 21/79 | 85 | 73 | 77/6 | 73/6 |
DMAP | 16/84 | 80 | 17 | 64/16 | 6/11 |
p-TSA | 19/81 | 5 | 3 | n.d. | n.d. |
Following this result and to determine the influence of the alcohol hindrance on transcarbamoylation, hexanol and 2-heptanol were used as primary and secondary model alcohol respectively (Scheme 3b). 2-Hydroxyethyl butyl carbamate, the hydroxy carbamate model molecule, was prepared by reacting butyl amine and ethylene carbonate (60 °C for 24 h). The model reactions were carried out in bulk at 140 °C using 2-hydroxyethyl butyl carbamate, five equivalents of either primary or secondary alcohol and 5 mol% catalyst (as compared to carbamate groups, 2 mol% for DBTDL). 1H NMR analysis (representative kinetics presented in Fig. 4) allowed to measure both conversion (using ethylene glycol signal at 3.7 ppm, Fig. 5a) and the reactant/product/urea proportions (Fig. 5b). Model transcarbamoylation performed without catalyst or in the presence of p-TSA or TU, with either a primary or a secondary alcohol showed that no transcarbamoylation (associative or dissociative) occurred in these conditions (Fig. 5a, and Table 3). In contrast, DMAP, DBTDL and t-BuOK were efficient catalysts for transcarbamoylation induced by primary alcohol. High conversions (80–100%) after 72 h at 140 °C were measured by 1H NMR spectroscopy (Fig. 5a). In the case of DMAP, switching from a primary to a secondary alcohol drastically decreased the conversion (below 20% after 72 h at 140 °C). In contrast, DBTDL reach almost similar conversion with primary or secondary alcohol. Potassium tert-butoxide showed intermediate decrease in reactivity.
Fig. 4 1H NMR spectra (in CDCl3) of the transcarbamoylation of 2-hydroxyethyl butyl carbamate by 2-heptanol in bulk with DBTDL at 140 °C at different reaction time. |
The drop of reactivity observed in model transcarbamoylation catalysed by DMAP, i.e. using a secondary alcohol instead of a primary one, could explain the discrepancy of dynamic properties as compared to the previously reported PHU-DMAP covalent adaptable networks from 5-CC monomers.20,25 Kinetics results on model molecules demonstrated that transcarbamoylation in DMAP/secondary alcohols system was slow and unselective (urea formation). As the proportions between primary and secondary alcohols were not determined in other reported systems, one might suspect that primary alcohol content can vary depending on the selected amine/cyclic carbonate system, and were presumably higher than the one measured herein. Alternatively, the chemical structure of 5-CC monomer was also put forward to explain the poor reprocessability observed using carbonated sorbitol-based monomer to prepare PHU-CAN catalysed by DMAP.25 The authors suggested that the spatial proximity between cyclic carbonate moieties leads to the formation of a network with a higher crosslinked density and reduced chain mobility. A similar assumption can be made for TMPTC-based networks and explain the poor results (stress relaxation and reprocessing) observed for such system with DMAP catalyst.
According to model molecular reaction results, t-BuOK appears as a promising candidate to promote transcarbamoylation in PHU-networks. However, material characterizations (rheology) demonstrated only poor dynamic properties for the PHU-t-BuOK material. This behaviour was explained by the unsatisfactory dispersion of the catalyst in the PHU matrix (already mentioned above in the tensile test experiments) despite the dissolution/dispersion of this catalyst in DCM or THF prior to the materials synthesis. Looking at the relative proportions of reactant/product/urea monitored by 1H NMR spectroscopy (Fig. 5b), variations also appeared between the three most active catalysts (DMAP, DBTDL and t-BuOK). Relatively high urea content (16–20 mol%) were measured for the model reactions catalysed by DMAP and t-BuOK with 1-hexanol. When 2-heptanol was used, the urea content quantified was even higher than that of the desired transcarbamoylation product, highlighting the occurrence of side reactions. Urea formation is admitted to occur in a two-step process with: (1) retro-formation of primary amine and cyclic carbonate, and (2) nucleophilic attack of a primary amine on a urethane group. In contrast, employing hexanol leads to a rapid conversion of 2-hydroxyethyl butyl carbamate into hexyl butyl carbamate thus avoiding the retroformation of cyclic carbonate and urea formation. Surprisingly, only urea signals were detected in 1H NMR spectra whereas the signal of ethylene carbonate (singlet at 4.55 ppm) was either non-detected or of low intensity (although it should be equal to the amount of urea formed), thus suggesting that cyclic carbonate was consumed presumably by reaction with the alcohol in excess under these conditions (5 equivalents of alcohol, 140 °C). This was confirmed by a control experiment (ethylene carbonate (1 eq.), 2-heptanol (5 eq.), t-BuOK (5 mol%)) which showed the characteristic 1H NMR signal of ethylene glycol (Fig. S15†).
In the particular case of DBTDL, the reactivity was not affected by the alcohol nature and material properties (shear modulus stable at high temperature, dissolution test and faster relaxation) prompt us to suggest an alternative exchange mechanism. Organotin compounds are known to activate both alcohol and carbamate moieties through a coordinated transition state.42 The coordinated mechanism suggested in Scheme 4b, which involved a coordination of alcohol (regardless of the alcohol nature) and carbamate could explain the differences observed between DBTDL and the other catalysts used in this study.
1 H NMR (CDCl 3 , 298 K, 400 MHz) δ (ppm) = 5.03 (broad, 1H, NH(carbamate)); 4.17 (t, 2H, CH2–CH2–OH, J = 4.4 Hz); 3.77 (m, 2H, CH2–CH2–OH); 3.15 (q, 2H, CH2–NH, J = 7.2 Hz); 2.98 (t, 1H, –OH, J = 5.6 Hz); 1.46 (m, 2H, CH2); 1.33 (m, 2H, CH2); 0.90 (t, 3H, CH3, J = 7.5 Hz).
(1) |
(2) |
(3) |
Footnote |
† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d3py00579h |
This journal is © The Royal Society of Chemistry 2023 |