Elizabeth M. Timmersabc,
P. Michel Fransend,
Álvaro González Garcíacef,
Sandra M. C. Schoenmakersbc,
Jose Rodrigo Maganaabc,
Joris W. Peetersd,
Ronald Tennebroekg,
Ilse van Castereng,
Remco Tuiniercef,
Henk M. Janssen‡
d and
Ilja K. Voets‡*ac
aLaboratory of Self-Organizing Soft Matter, Department of Chemical Engineering and Chemistry, Eindhoven University of Technology, P.O. Box 513, 5600 MB, Eindhoven, The Netherlands. E-mail: i.voets@tue.nl
bLaboratory of Macro-Organic Chemistry, Department of Chemical Engineering and Chemistry, Eindhoven University of Technology, P.O. Box 513, 5600 MB, Eindhoven, The Netherlands
cInstitute for Complex Molecular Systems, Eindhoven University of Technology, P.O. Box 513, 5600 MB, Eindhoven, The Netherlands
dSyMO-Chem B.V., Den Dolech 2, 5612 AZ, Eindhoven, The Netherlands
eLaboratory of Physical Chemistry, Department of Chemical Engineering and Chemistry, Eindhoven University of Technology, P.O. Box 513, 5600 MB, Eindhoven, The Netherlands
fVan ‘t Hoff Laboratory for Physical and Colloid Chemistry, Department of Chemistry and Debye Institute for Nanomaterials Science, Utrecht University, Padualaan 8, 3584 CH, Utrecht, The Netherlands
gDSM Resins and Functional Materials, Sluisweg 12, 5145 PE Waalwijk, The Netherlands
First published on 22nd April 2021
Industrial and household products, such as paints, inks and cosmetics usually consist of mixtures of macromolecules that are disperse in composition, in size and in monomer sequence. Identifying structure–function relationships for these systems is complicated, as particular macromolecular components cannot be investigated individually. For this study, we have addressed this issue, and have synthesized a series of five sequence-defined polyurethanes (PUs): one neutral-hydrophobic, one single-charged hydrophilic, one single-charged hydrophobic and two double-charged amphiphilic PUs (one symmetric and one asymmetric). These novel precision PUs – that were prepared by using stepwise coupling-deprotection synthetic protocols – have a defined composition, size and monomer sequence, where the chosen sequences were inspired by those that are abundantly formed in the production of industrial waterborne PU dispersions. By performing dynamic light scattering experiments (DLS), self-consistent field (SCF) computations and cryogenic transmission electron microscopy (cryo-TEM), we have elucidated the behavior in aqueous solution of the individual precision PUs, as well as of binary and ternary mixtures of the PU sequences. The double-charged PU sequences (‘hosts’) were sufficiently amphiphilic to yield single-component micellar solutions, whereas the two more hydrophobic sequences did not micellize on their own, and gave precipitates or ill-defined larger aggregates. Both the neutral-hydrophobic PU and the hydrophilic single-charged PU were successfully incorporated in the host micelles as guests, respectively increasing and reducing the micelle radius upon incorporation. SCF computations indicated that double-charged symmetric PUs stretch whilst double-charged asymmetric PUs are expelled from the core to accommodate hydrophobic PU guests within the micelles. For the ternary mixture of the double-charged symmetric and asymmetric hosts and the neutral-hydrophobic guest we have found an improved colloidal stability, as compared to those for binary mixtures of either host and hydrophobic guest. In another ternary mixture of precision PUs, with all three components not capable of forming micelles on their own, we see that the ensemble of molecules produces stable micellar solutions. Taken together, we find that the interplay between PU-molecules in aqueous dispersions promotes the formation of stable micellar hydrocolloids.
Most studies on the association in solution of synthetic (macro)molecules focus on well-defined (model) systems, usually involving compounds that are examined under controlled conditions.3,4 Particularly well-studied is the self-assembly of both low and high molar mass amphiphiles into core–shell micelles.3,4 Far less is known about the co-assembly of mixtures of synthetic (macro)molecular components, as for several reasons such systems have been studied less frequently. Firstly, synthetic polymers are not as well-defined molecularly as, for example, low molecular weight surfactant compounds or biopolymers: synthetic polymers are generally disperse in composition, in size and in monomer sequence. These aspects have important consequences for their assembly5 and co-assembly properties. Secondly, the in-depth characterization in solution of mixtures of species of comparable, nanometric size is demanding, as contributions of the individual species with small differences in dimensions and properties are challenging, if not impossible, to resolve.
An appealing strategy to advance the understanding of the behavior of multicomponent mixtures is to prepare precision polymers with as little dispersity in composition, sequence, and size as possible and use these to prepare well-defined mixtures for in-depth characterization. In the past decade, this approach has gained increasing attention, for example in a study on the impact of size dispersity on the assembly of block copolymers and oligomers in bulk and in solution.5 Two main strategies to prepare (macro)molecules low in dispersity, size and sequence are (i) stepwise synthetic approaches and (ii) controlled radical polymerization (CRP) methods.2 These procedures yield so-called ‘sequence-defined’ and ‘sequence-controlled’ polymers, respectively. Stepwise synthetic routes are akin to solid-phase peptide synthetic approaches that were introduced to prepare perfectly defined oligopeptides from canonical amino acids,6 and such routes can be used to produce polymers and oligomers from other synthetic monomer origin as well.7 Stepwise methods yield truly discrete compounds termed ‘sequence-defined’ polymers, such as sequence-defined polypeptoids,8 polyamides2 and oligocarbamates.7 Unfortunately, only lower molecular weight materials, essentially oligomers, are accessible, which is not surprising in view of the upper limit of ∼50 monomers in solid-phase peptide synthesis.7 By contrast, CRP allows the synthesis of high molecular weight polymers. However, CRP always yields slightly disperse polymer products which are therefore termed ‘sequence-controlled’. Overall, this research field has advanced tremendously, and truly sequence-defined polymers of 15 and 17 kDa have now been successfully produced,9,10 opening up opportunities for applications that require higher molecular weight precision materials.
Polyurethanes (PUs) constitute an important class of polymer materials which find use in a broad spectrum of applications ranging from coatings and paints to high-performance elastomers to tough, rigid plastics.11,12 Traditionally, PU dispersions contained significant amounts of volatile organic solvents to facilitate processing, but in recent years waterborne polyurethanes (WPUs) have attracted great interest as useful alternatives with reduced adverse health and environmental impact.11,12 WPUs are typically dispersions of polyurethane ionomers (PUIs) that are produced in a two-step synthesis. In the first step, isocyanate-terminated prepolymers are prepared by reacting polyols (typically diols, e.g. poly-THF or poly-tetrahydrofuran diols) and ionizable monomers (typically dimethylpropionic acid, i.e. DMPA, a diol) with di- or tri-isocyanates (typically di-isocyanates, e.g. IPDI or isophorone diisocyanate).12 The result is a mixture of isocyanate terminated intermediates, of which the exact composition is determined by statistics. In the second step, the prepolymers are dispersed in water at a high pH (accordingly, the DMPA-units become anionic), and are chain-extended using a polyamine (typically a di-amine). Prior to chain-extension, the prepolymer solution is a multi-component mixture of (macro)molecules with different amounts of ionizable groups, molecular weights, and hydrophobicities. The components may or may not be surface active, and may co-assemble or instead segregate into different types of hydrocolloids. Desired solubilization into stable hydrocolloids (before and after chain-extension), as opposed to unwanted precipitation of components, is often accomplished by performing empirical (‘trial and error’) studies, and this approach has led to great progress and knowledge in PU formulation, application and product design.
Given their synthesis, industrial WPU dispersions comprise a mixture of macromolecules with a variety in composition, monomer sequence and length. In addition to amphiphilic compounds with a tendency to associate into micelles, WPU mixtures may contain fairly soluble, hydrophilic species and rather insoluble, hydrophobic compounds. In contrast to conventional amphiphiles and surfactants where one hydrophobic and one hydrophilic segment are simply joined together head-to-tail, PUI macromolecules have hydrophilic (DMPA) and hydrophobic units (isophorone diamine (IPDA), poly-THF) randomly distributed along the polymer chain. In prior work, sequence-controlled dendritic PUs have been obtained via a selective reactivity approach; the products were non-ionic and can therefore not be dispersed in water.13 In this work, we have set out to prepare sequence-defined PUIs of variable composition, size and monomer sequence to reflect the breadth of the various species present in industrial WPU dispersions. This should allow for in depth studies on these PUI macromolecules.
Accordingly, and aiming to elucidate the functional role of the various polyurethane molecules in industrially relevant multicomponent WPU mixtures (prior to chain extension), we have created a small but representative series of five sequence-defined PUs (or precision PUs, see Fig. 1 and Scheme S1†). The novel precision PU(I)s have been synthesized through a stepwise coupling-deprotection protocol using poly-THF (2 kDa), IPDA and/or DMPA building blocks. Next, we have studied their association behavior in aqueous solution, either per PU sequence individually or in combination with one or two of the other PU sequences in mixtures of controlled composition. Previously, SCF calculations revealed that the specific surface active PUI-A2 and PUI-S2 sequences are relatively abundant in PU dispersions and presumably play a major role in particle formation and stabilization in WPUs.14 However, to date such materials had not been prepared and tested experimentally. Dynamic light scattering (DLS) experiments and self-consistent field (SCF) computations have been used to investigate individual PUs or multicomponent PU mixtures with exactly known composition. In addition, cryogenic transmission electron microscopy (cryo-TEM) has been used to visualize the hydrocolloids formed by the PUs. The employed combination of experimental and computational tools, previously used to study other surface active materials, allows to investigate dissolution, micellar dimensions, colloidal morphology, influences of used co-solvents, chain packing, and location of the different types of chains within mixed micelles.15,16 Our findings reveal whether individual PU sequences preferentially reside in the micellar interior (core) or exterior (corona). Furthermore, we find that micelles co-assembled from two or three PUs display an increased encapsulation efficiency of the non-soluble hydrophobic PU guest, as compared to hydrocolloids self-assembled from single PU components. Accordingly, we have seen that mixed micelles of PUs display improved colloidal stabilities.
(1) |
(2) |
The SCF follows a grand-canonical approach in which, provided a certain initial molecular configuration, the free energy of the lattice is minimized. The computed grand-potential Ω depends on the concentration of compounds. The aggregation number gk is given by the excess number of molecules of compound k in the self-assembly. By combining small system thermodynamics with the SCF computations it is possible to identify the thermodynamically preferred self-organized morphology in terms of the amount of a certain compound in the lattice at which Ω = 0, with ∂Ω/∂gk < 0. For every computation, the incompressibility constraint is enforced. At the most probable micelle condition, radial concentration profiles (φ) are collected. From these, the hydrodynamic micelle size can be calculated.19 The theoretical guest loading, fguest,SCF, was defined as
(3) |
In this work, we have synthesized a series of chemically distinct, sequence-defined PU(I)s (Fig. 1; Scheme S1†) to investigate the role of such molecules in the formation of stable hydrocolloids, more particularly micelles. We have used protective group chemistry applying an iterative coupling-deprotection strategy to produce the PU(I)s. The synthetic route to PUI-A1 and PUI-A2 is shown schematically in Fig. 2. In an analogous approach, and starting from bi-functional poly-THF, we have prepared PUI-S0 and PUI-S2. The smaller sPUI-S1 material was synthesized in five steps from dimethylpropionic acid (DMPA) and isophorone diamine (IPDA). As can be appreciated from the followed route, we have employed the reaction between amines and activated carbonates to create urethane linkages, and have avoided using the typical chemistry between alcohols and isocyanates to produce PU(I)s. Amines as well as activated carbonates are stable, so both these type of building blocks can easily be manipulated, purified and stored. In contrast, the relatively high reactivity of isocyanates towards water renders isocyanate intermediates effectively unstable, either during synthesis, work-up or storage, precluding their use in step-by-step preparations of precision materials.
The employed organic synthetic approach made it possible to isolate PU(I)s with a strictly defined order of components, i.e. with a sequence-defined microstructure. In the prepared materials, the number of IPDA, DMPA and poly-THF groups is precisely controlled, as well as the positioning of these groups within the macromolecular structure. Note, however, that the produced PU(I)s are not chemically pure, where this is due to (i) the used poly-THF component, introducing a distribution and dispersity in molecular weight that is common for almost all synthetic polymers, and (ii) the employed IPDA component, introducing regio- and stereo-isomeric diversity. The latter is illustrated for IPDA-Boc in Fig. 2, but is not shown for the other PU(I)s for reasons of brevity. The five PU(I)s were molecularly characterized using NMR, ATR-FT-IR, SEC, HPLC-MS and MALDI-TOF-MS (see the ESI for full details†), and found data were in line with the assigned structures. Particularly, HPLC-MS analysis of sPUI-S1 showed two dominant peaks for the isomers of this compound (Fig. 3A), while MALDI-TOF-MS analysis of PUI-A1, PUI-A2 and PUI-S2 showed single arrays of peaks corresponding to masses of individual oligomers for the respective PUIs (Fig. 3B–D), proving the integrity of the prepared materials. For all three polymers, the periodicities of the arrays were 72 Dalton, which is the molecular weight of a THF-monomer.
Fig. 4 Cryo-TEM micrographs of PUI-A2 (A) and PUI-S2 (B) after vitrification from aqueous 0.1 M TEA solutions. The grey spots are the micelles vitrified in the ice layer, while the smaller black spots are ice-crystal artefacts at the surface. Concentration: approximately 2.5 mg mL−1. Magnification: 48000. The ESI contains additional cryo-TEM micrographs with ESI.† |
Next, the remaining three sequences PU-S0, PUI-A1 and sPUI-S1 have been briefly studied individually. These materials carry either a single charge (sPUI-S1 and PUI-A1) or no charge at all (PU-S0). Unsurprisingly, the neutral and hydrophobic PU-S0 is not soluble in a 0.1 M TEA solution in water. Rehydration of the PUI-A1 thin film gives a turbid dispersion, that does not show precipitation. DLS measurements do not show micelle formation and only reveal ill-defined larger aggregates in the 100–1000 nm radius size range (intensity plot; data not shown). For sPUI-S1, that lacks a surfactant architecture, we find clear solutions in water that show large non-micellar aggregates, roughly 40 nm in radius (data not shown).
Complementary SCF computations were performed to obtain further insight into the micellization of the PUIs that contain both hydrophobic poly-THF and charged DMPA. Initial SCF computations predicted that molecules with two charges, i.e.PUI-A2 and PUI-S2, form micellar-type of assemblies, in line with our experimental findings (Fig. S6†). SCF computations on PUI-A1 showed that this PUI type does not spontaneously micellize in water (Fig. S6†). This computation is in line with the observed turbidity and large aggregate sizes observed for PUI-A1 in DLS. Apparently, the hydrophilic content of this sequence is too low for self-assembly into stable micelle hydrocolloids. SCF computations on sPUI-S1 yield molecularly dissolved species, presumably due to thermodynamic equilibrium constraints (data not shown). This SCF prediction is in contrast to the non-micellar aggregates observed in DLS.
We first studied the solubilization of the symmetric guest PU-S0 (Fig. 5A) in micelles of either the asymmetric PUI-A2 or the symmetric PUI-S2. We anticipated the neutral and hydrophobic guest to favor association, and to reside in the micellar core preferentially. Therefore, we expected that increasing the guest loading would increase the micellar size. Indeed, solubilization of the hydrophobic guest PU-S0 increased the micellar dimensions markedly from Rh ≈ 8.0 nm and Rh ≈ 6.0 nm at fguest = 0 for the pure host solutions up to Rh ≈ 12.7 nm and Rh ≈ 9.6 nm for mixed micelles of PUI-A2 or PUI-S2, respectively, at a PU-S0 loading fguest ≈ 0.2. Hence, both micelle types grow approximately 1.6-fold in radius (so approximately 4.0-fold in volume) upon loading the micelle with about 20 w/w% guest molecules. As the host PUI concentration remains unchanged upon adding guests, with the micelles growing considerably, we assume that at a guest loading of fguest ≈ 0.2 the number of micelles is reduced markedly (corresponding to an increased host aggregation number per micelle).
Complementary SCF computations were performed to obtain further insight into the micellization of the PU(I)s. SCF computations allow to determine the location of not only the individual components, but also of their building blocks, within mixed micelles. This is achieved by comparing the radial concentration profiles, φ, as a function of the distance from the center of the micelle, d, of the PUI hosts alone (referred to as ‘empty’) to the individual PU(I) sequences in the mixtures with PU-S0 (Fig. 6). Not surprisingly, we found that the charged carboxylic acid-group of DMPA is mainly located near the outer surface of the micelle. In the mixtures, the radius (Rh) is approximated by the d-value where φhost approaches zero. In both empty micelle systems, the concentration distributions are fairly homogeneous throughout the micelle. The distribution of PUI-S2 within the micelle remains similar upon adding PU-S0 guest and only shifts to a larger d. This seems to indicate that the poly-THF chains of the PUI-S2 host stretch to accommodate PU-S0. Interestingly, the shape of the PUI-A2 host profile upon addition of PU-S0 shows a stark change as compared to that of the host alone, with the PU-S0 concentration in the core even exceeding the concentration of poly-THF chains from the host. This pinpoints that PUI-A2 is partially expelled from the center of the micelle and substituted with PU-S0 guest. We attribute the difference in host and guest distribution along the micelle to the bolaamphiphilic structure of PUI-S2, which structure limits the degree of freedom it has to rearrange and accommodate guest compounds. In comparison, PUI-A2 has a dangling poly-THF chain that can more freely move and rearrange. In the computations, this difference is reflected in the theoretical maximum encapsulation efficiencies as well, as the maximum attainable fguest,SCF before divergence occurs equals 0.049 for PUI-S2 and 0.224 for PUI-A2. These calculations therefore indicate that the bolaamphiphile architecture of PUI-S2 restricts the amount of hydrophobic guests that can be accommodated, whereas PUI-A2 is able to encapsulate a higher loading of guest. This discrepancy between the encapsulation efficiencies of the two hosts as predicted with SCF was not observed experimentally, as both PUI-A2 and PUI-S2 accomodated PU-S0 comparably well.
Having established the efficient encapsulation of the hydrophobic guest PU-S0 in PUI micelles of either PUI-S2 or PUI-A2, we subsequently focused on the ternary system wherein the PUI-S2 and PUI-A2 hosts were mixed at an equimolar ratio. The size of the mixed micelles remained intermediate between that of either of the constituent PUI hosts up to fguest ≈ 0.2, which indicates molecular mixing of both hosts.20 The mixed micelles grew approximately 1.4-fold in size from an Rh of 7.0 nm in the absence of PU-S0 to Rh ≈ 9.6 nm at fguest ≈ 0.2, corresponding to about a 2.6-fold growth in volume. The increase in micellar size at the same PU-S0 loading is thus smaller for the mixed PUI micelle as compared to micelles composed of either the PUI-A2 or the PUI-S2 host. This may be due to the increased configurational freedom within the mixed core: the PUI-S2 chains might relax in this morphology compared to their energetically unfavorable stretched state in the single host micelle. This clarification is reminiscent and in line with recent findings on a binary diblock mixture differing in core block length.21 Interestingly, when comparing the size distributions by intensity for guest fractions fguest from 0 progressively to 0.2, we observe broadening for the individual PUI-A2 (Fig. 5B) and PUI-S2 (Fig. 5C) host micelles, whereas the equimolar host mixture (Fig. 5D) yields a narrow distribution for all tested fguest values. Additionally, at higher guest loading for the PUI-A2 host, we observe a broad shoulder in the size distribution (Fig. 5B), indicating the onset of formation of larger ill-defined (non-micellar) aggregates, which signals that these samples are approaching the stability boundary. The data suggest a somewhat enhanced encapsulation efficiency and an improved colloidal micellar stability for the ternary system as compared to the binary ones. This is in line with observations made for a number of mixed polymeric systems reviewed by Attia and coworkers,22 and for mixed surfactant micelles.23,24 In future studies, it would be of interest to further optimize the composition of PU(I) mixtures to increase the solid content; for example with regard to coating applications, thereby improving drying processes and decreasing costs.24
Since sPUI-S1 augments the repulsive double layer interactions which enhance colloidal stability at low ionic strength in the above binary systems, we next studied whether sPUI-S1 could prompt micellization of a PUI component that is incapable of forming micelles on its own, i.e.PUI-A1. We first added increasing amounts of sPUI-S1 to PUI-A1 and monitored the turbidity of the solutions, thereby aiming at formation of stable hydrocolloids with dimensions similar to those found for our other hosts (Fig. 8A). At fguest < 0.27, the mixtures were turbid dispersions containing aggregates with sizes much larger than 10 nm. Interestingly, at fguest > 0.27, the samples became transparent and the main size distributions correspond to Rh ≈ 5.8 nm (Fig. 8A), which is comparable to the micellar sizes obtained with PUI-S2 or PUI-A2. Hence, fguest ≈ 0.27 demarcates the threshold between two regimes: the large aggregate regime and the small mixed-micelle regime. Accordingly, we have found a synergy between sPUI-S1 (that has a high charge-to-weight ratio, but lacks a surfactant architecture) and PUI-A1 (that has a head-to-tail surfactant molecular structure, but that is lacking in charge), as only together they form stable micelles.
Fig. 8 DLS encapsulation experiments. (A) Addition of sPUI-S1 guest to PUI-A1 host that does not micellize on its own. At sPUI-S1fguest > 0.27, micelle formation is induced. (A and B) Main Rh from size distributions by volume. The (cumulative) host concentration is 2.5 mg mL−1 for all data points. (B and C) Encapsulation of PU-S0 guest into PUI-A1:sPUI-S1 (1:6 molar ratio host) co-micelles. (B) Rh as a function of PU-S0 guest loading. (C) Size distributions by intensity from DLS, arrow indicating increasing guest load, at guest loading 0, 0.041, 0.095, 0.15, 0.20 (low to high guest loading in orange gradient). (D) CryoTEM micrograph of PUI-A1:sPUI-S1 (1:6 – mol:mol) host with added PU-S0 guest fguest ≈ 0.2. Cumulative host concentration: approximately 2.5 mg mL−1. Magnification: 48000. The ESI contains additional cryo-TEM micrographs with ESI.† |
Further, and to assess the encapsulation efficiency of the prepared mixed micelles of PUI-A1 and sPUI-S1, we have studied their size and stability upon addition of PU-S0 guest. For this purpose, we have employed mixed micelle hosts containing PUI-A1 and sPUI-S1 in a 1:6 molar ratio (this corresponds to fguest = 0.62 in Fig. 8A). In analogy to the other PU(I) micelles described in this work, Rh increased with the PU-S0 guest concentration from 4.9 nm at fguest = 0 to Rh ≈ 7.9 nm at fguest ≈ 0.2, displaying a 1.6-fold increase in radius, corresponding to about a 4-fold increase in volume (Fig. 8B). Note that in this ternary system a minor fraction of the PU(I)s remains incorporated in large aggregates, indicated by the bimodal size distributions at all guest loadings (Fig. 8C). However, as the intensity scales with R3, the secondary small peak represents a very small fraction of scatterers. Hence, the micelles are in all cases the most abundant species. Again, this ternary system shows synergy as none of the three components is capable of forming micelles individually, but the ensemble of molecules produces stable micelles with hydrophobic guest encapsulated. Finally, cryo-TEM measurements confirmed increased dimensions for PUI-A1/sPUI-S1 micelle hosts upon incorporation of PU-S0 guest (Fig. 8D; Fig. S9 and S10†), showing radii of 4.0 and 7.1 nm for micelles without and with PU-S0 guest (fguest ≈ 0.2), respectively. As also seen for PUI-A2 and PUI-S2 micelles (Fig. 4 and 5), we observe a good agreement between radii assessed with DLS and cryo-TEM.
On a final note, the sPUI-S1 sequence can be viewed as quite a good model for IPDI-DMPA-IPDI molecules that are abundant in WPU formulations prior to the chain extension reaction (vide supra), provided of course that both isophorone di-isocyanate (IPDI) and DMPA building blocks are employed. The experimental data in this study show that sPUI-S1 behaves as an effective co-surfactant, promoting stable mixed-micelle formation. Presumably, therefore, the actual IPDI-DMPA-IPDI molecule in industrial WPUs plays a similar role. It stabilises the WPU dispersion prior to chain extension, but probably also during and after the chain extension reaction: it is likely positioned at the periphery of particles prior to chain extension, and it presumably remains there as it gets trapped during the chain-extension reaction that results in a diminished mobility of molecular chains. This role of sPUI-S1 is not apparent from our SCF computations, nor from those reported on in literature,14 highlighting the importance of the combination of computational and experimental research in this field. Indeed, experimental results may serve as an incentive to improve SCF models, for example such that the role of sPUI-S1 will become better captured by SCF computations.
Footnotes |
† Electronic supplementary information (ESI) available. See DOI: 10.1039/d1py00079a |
‡ These authors contributed equally. |
This journal is © The Royal Society of Chemistry 2021 |