Riddhi Kumari Riddhi‡
a,
Francesc Penas-Hidalgo‡
b,
Hongmei Chen
b,
Elsje Alessandra Quadrelli
a,
Jérôme Canivet
*a,
Caroline Mellot-Draznieks
*b and
Albert Solé-Daura
*cd
aIRCELYON, UMR 5256, Université LYON 1, 2 avenue Albert Einstein, 69626 Villeurbanne Cedex, France
bLaboratoire de Chimie des Processus Biologiques, CNRS UMR 8229, Collège de France, PSL Research University, Sorbonne Université, 75231 Paris Cedex 05, France. E-mail: caroline.mellot-draznieks@college-de-france.fr
cDepartment de Química Física i Inorgànica, Universitat Rovira i Virgili, Marcel·lí Domingo 1, Tarragona 43007, Spain
dInstitute of Chemical Research of Catalonia (ICIQ-CERCA), The Barcelona Institute of Science and Technology, Avgda. Països Catalans, 16, 43007 Tarragona, Spain. E-mail: asole@iciq.es
First published on 30th August 2024
Catalysis plays a crucial role in advancing sustainability. The unique reactivity of frustrated Lewis pairs (FLPs) is driving an ever-growing interest in the transition metal-free transformation of small molecules like CO2 into valuable products. In this area, there is a recent growing incentive to heterogenize molecular FLPs into porous solids, merging the benefits of homogeneous and heterogeneous catalysis – high activity, selectivity, and recyclability. Despite the progress, challenges remain in preventing deactivation, poisoning, and simplifying catalyst-product separation. This review explores the expanding field of FLPs in catalysis, covering existing molecular FLPs for CO2 hydrogenation and recent efforts to design heterogeneous porous systems from both experimental and theoretical perspectives. Section 2 discusses experimental examples of CO2 hydrogenation by molecular FLPs, starting with stoichiometric reactions and advancing to catalytic ones. It then examines attempts to immobilize FLPs in porous matrices, including siliceous solids, metal–organic frameworks (MOFs), covalent organic frameworks, and disordered polymers, highlighting current limitations and challenges. Section 3 then reviews computational studies on the mechanistic details of CO2 hydrogenation, focusing on H2 splitting and hydride/proton transfer steps, summarizing efforts to establish structure–activity relationships. It also covers the computational aspects on grafting FLPs inside MOFs. Finally, Section 4 summarizes the main design principles established so far, while addressing the complexities of translating computational approaches into the experimental realm, particularly in heterogeneous systems. This section underscores the need to strengthen the dialogue between theoretical and experimental approaches in this field.
The advent of molecular frustrated Lewis pairs (FLPs), introduced by Stephan et al. in 2006,4 showcases reactivity stemming from steric hindrance between Lewis acid (LA) and base (LB) partners, precluding the formation of the LA–LB adduct.5–7 FLPs, often referred to as a “catalysis Holy Grail”, have ushered in a new era of sustainable green chemistry. In particular, they allow the transition metal-free activation of small molecules, demonstrating significant potential in hydrogen activation and double bond reduction in fine chemicals, including CO2 hydrogenation.8–11
While the field of homogeneous FLPs has gathered substantial knowledge in the past decade,12,13 the transfer of FLPs’ chemistry to heterogeneous catalysis is rather recent and garnering increasing interest. Besides addressing recyclability issues, the heterogenization of molecular FLPs allows the isolation of active sites within a controlled chemical environment, while preventing undesired interactions and catalyst deactivation. This isolation alters substrate activation interactions on the heterogeneous catalyst, leading to unprecedented reactivities explicable through molecular chemistry mechanisms. The development of heterogeneous FLPs is currently challenging and mainly limited to doped/defective oxide and metal surfaces.14,15
Incentives for discovering heterogeneous FLPs encompass aspects ranging from molecular-scale control over the distance between FLP's partners, influencing reactivity, to their recyclability, shaping, and scale-up production. Quantum mechanical calculations serve as a powerful tool for the in silico design of molecular architectures, having matured in molecular catalysis for predicting efficient systems, remarkably for asymmetric synthesis.16–18 However, computational methodologies are at an earlier stage concerning heterogeneous catalysis.19–22
In addition to the perspectives on FLPs’ chemistry by Stephan et al.,14,23 recent reviews in the last five years have covered distinct advancements in FLPs’ chemistry by considering either the realm of homogeneous catalysis24–26 or strategies for their heterogenization into solid materials.15,27,28 A recent review covered the specific area of the computational chemistry of FLPs, focusing exclusively on molecular systems.29 Therefore, there remains a critical need for a comprehensive analysis of the recent progress and existing challenges in the discovery of FLP-based catalytic systems for CO2 hydrogenation, encompassing both molecular and solid-state FLPs and integrating recent experimental and theoretical facets.
This review delves into fifteen years of FLPs chemistry development, exploring the chemistry of FLPs based on main group elements from experimental discovery to computational mechanistic dissection. The discussion encompasses the recent experimental advances of FLPs in catalytic processes, tools developed for their heterogenization, and key aspects of structure–activity relationships for CO2 hydrogenation as determined in silico.
FLP formula | LA/LB pair | Inter/intramolecular | Solvent | T, P, time | Product (yield) | Ref. |
---|---|---|---|---|---|---|
B(C6F5)3/TMP | B/N | Inter | Toluene | 160 °C, 1 eq. CO2, excess H2, 6 days | MeOB(C6F5)2 but CH3OH after distillation (17–25%) | 31 |
B(C6F5)3/lutidine | B/N | Inter | Toluene | RT, 4 atm H2 and CO2 | [LutH] [HC(O)OB(C6F5)3] | 34 |
B(C6F5)3/amine | B/N | Inter | Bromobenzene | RT, 4 atm H2 | [PhNMe2H][HB(C6F5)3] | 35 |
B(C6F5)3/tBu3P | B/P | Inter | Bromobenzene | 25 °C, 1 bar CO2, immediate | tBu3P(CO2)B(C6F5)3 (87%) | 36 |
(Me3C6H2)2PCH2CH2B(C6F5) | B/P | Intra | Pentane | 25 °C, 2 bar CO2, 15 min | Cyclo-(Me3C6H2)2PCH2CH2–B(C6F5)2–(CO2) (79%) | |
(C6F4(C6F5))3B/P(tBu3) | B/P | Inter | Toluene | 145 °C, 1 atm CO2, 1 atm H2, 24 h | [HPt-Bu3][(C6F4(C6F5))3BO2CH] | 37 |
AlX3/Mes3P | Al/P | Inter | Bromobenzene | 25 °C, 1 atm CO2, 2 h | Methanol (37–51%) | 38 |
AlX3/Mes3P | Al/P | Inter | Bromobenzene | RT, 2 atm CO2, 16 h | CO, Mes3P(C(OAlI2)2O)(AlI3) and [MesPX][AlX4] | 39 |
[MeC(CH2PPh2)3Cu(NCMe)][PF6]/C10H16N2 (DBU) | Cu/N | Inter | Acetonitrile | 80–140 °C, 40 atm H2/CO2, 6–20 h | Formate salt[C10H17N2][HCO2] | 40 |
1-BR2-2-NMe2-C6H4 | B/N | Intra | Benzene | 80 °C, 1 atm CO2, 4 atm H2, 9 days | Formates, acetals, methoxides | 32 |
tBu2P–O–Al | Al/P | Intra | n-Hexane | −196 °C then RT, 2 eq. CO2, overnight | tBu2P(H)–O–Al(CO2H) | 41 |
K2[(BCF)2–CO3] | B/CO32− | Inter | THF | 160 °C, 40 bars H2, 20 bars CO2, 48 h | BCF-HCO2K | 42 |
Tris(bromo)tridurylborane/DBU | B/N | Inter | Acetonitrile | 120 °C, 130 bar, 48 h | Formate | 43 |
In 2009, Mömming, Stephan and coworkers described for the first time a novel approach for the metal-free binding of CO2 by FLPs based on the combination of BCF and phosphine.36 They demonstrated that the tBu3P/B(C6F5)3 FLP pair dissolved in bromobenzene can react with CO2 under ambient pressure at room temperature to yield a white solid product tBu3P(CO2)B(C6F5)3 in high yield (Fig. 1a). Specifically, the LB component (tBu3P) interacts with the carbon atom, while the LA component (BCF) stabilizes an oxygen atom of CO2 via coordination. They also showed that a pentane solution of an intramolecular FLP, (Me3C6H2)2PCH2CH2B(C6F5)2, could bind to CO2 at room temperature to form a white solid cyclo-(Me3C6H2)2PCH2CH2B(C6F5)2–(CO2) with 79% yield (Fig. 1b). The reversible nature of the CO2 binding was demonstrated upon heating the FLP–CO2 adducts under vacuum to 80 °C for 5 hours. Under these conditions, about half of the carbon dioxide was released and the initial FLP mixture was regenerated. These results were further supported by DFT calculations. This pioneering work offered a new metal-free approach to the activation of CO2, which avoids the drawbacks associated with traditional metal-based catalysts. Also, it provided unprecedented insights into the fundamental chemistry of CO2 binding by FLPs, which inspired further research in this area and led to potential applications in CO2 capture and recycling.
In the same year 2009, Ashley et al. illustrated the first homogenous selective hydrogenation of CO2 to CH3OH using an FLP-based non-metal-mediated procedure at low pressures (Fig. 2).31 The FLP mixture used here was a 1:1 mixture of 2,2,6,6- tetramethylpiperidine (TMP, Me4C5NH) and BCF which was already proven to cleave H2 heterolytically giving a salt [TMPH][HB(C6F5)3] at low pressure (1–2 atm).49 They discovered that this FLP mixture could not only cleave H2 but also subsequently insert CO2 in the B–H bond.
Other research groups have also employed a similar two-step mechanism, as described above, in their studies.50,51 The vacuum distillation of the solvated methoxyborate species resulted in the release of CH3OH (17–25% yield) as the only selective C1 product, along with C6F5H and TMP. The only source of labile protons in the decomposition reactions is provided by [TMPH]+, which triggers ion recombination to form TMP and CH3OH·B(C6F5)3. Dissociation of this hydroxy adduct could lead to the FLP's regeneration in a catalytic system. However, at 160 °C, the H+ attack on the ipso-C proceeds faster, resulting in Lewis acid breakdown. As a result, inert boroxine CH3OB(C6F5)2 is finally formed. The identification of these key intermediates along the reaction pathway was made possible through mechanistic investigations and multinuclear solution NMR spectroscopy.
Geier and Stephan reported that the FLP based on 2,6-lutidine (Lut = 2,6-dimethylpyridine) and B(C6F5)3 could heterolytically cleave H2 to produce the borohydride salt, [LutH][HB(C6F5)3], as a white solid (Fig. 3).52 Later, Mayer et al. explored the interaction of CO2 with the same salt.34 By treating it with 4 atm of CO2 in toluene-d,8 they observed the formation of an air-stable white solid, [LutH][HC(O)OB(C6F5)3]. The reduction of CO2 by [LutH][HB(C6F5)3] occurs much faster than the corresponding reaction with TMP, despite computational results by Pápai et al. suggesting that TMP/B(C6F5)3 is a more reactive FLP.53 The calculations showed that heterolytic cleavage of H2 is thermodynamically less favorable for Lut/B(C6F5)3 than for TMP/B(C6F5)3 by approximately 10 kcal mol−1. It was suggested that the higher reactivity of Lut/B(C6F5)3 for CO2 reduction may be due to a mechanism involving both hydride donation from [HB(C6F5)3]− and activation of CO2 by free B(C6F5)3. However, preliminary studies have shown that adding excess B(C6F5)3 did not significantly accelerate the reaction, and adding excess lutidine did not cause any substantial change in rate.
Later, Voss and coworkers examined how different amines (N-dimethylaniline, N-isopropylaniline, 1,4-C6H4(CH2NHtBu)2, and benzyldimethylamine) reacted with BCF and explored their FLP behavior.35 They illustrated how both steric and electronic factors may influence the formation of LP adducts. Although these factors may hinder adduct formation, they allow access to free FLPs that can be used for further reactions. Additionally, the borane's hydridophilic nature promotes the formation of minor side products that are derived from hydride abstraction from the α-C to produce iminium salts. They further studied the reaction of H2 and CO2 with these pairs. The Lewis pair of N-dimethylaniline–B(C6F5)3 reacted quickly with 4 atm H2 at 25 °C, even though the LB component had low basicity. By the heterolytic cleavage of H2, it produced the ammonium–hydridoborate salt [PhNMe2H][HB(C6F5)3]. The CO2 adducts of such salts have been investigated by Erker et al. as discussed above and were found to be thermally unstable, releasing CO2 quickly in solution at temperatures ranging from −20 to +80 °C.36,54
The following year, Travis et al. focused specifically on FLPs based on the Lewis acids tris(perchloroaryl)borane (BArCl) and tris(2,2′,2′′-perfluorobiphenyl)borane (PBB) with selected phosphines, which were extensively studied for small molecules activation.37,55,56 The reactivity was found mainly driven by the activation of H2 before a concerted addition of CO2. The Lewis acidity of the catalyst is crucial in this process, as it needs to be strong enough to split H2 but not too strong to allow hydride transfer to CO2. They designed Lewis acid assuming that the steric bulkiness at the ortho position could decrease the B–O bond strength of the methoxyborate formed and could consequently facilitate its cleavage to evolve methanol and regenerate the FLP catalyst. Such sterically congested LA was found less effective in activating H2 despite having higher Lewis acidity. This was mainly ascribed to the steric hindrance around the boron center, hindering the approach of H2 and increasing the energy barrier for the reaction.
Among all the phosphines tested with BArCl, only P(tBu)3 and P(Cy)3 could cleave H2 without thermal decomposition and were found to be the most active ones, completing the reaction in a long time of 56 and 40 hours at 90 °C, respectively. The formatoborate [PBB-OC(O)H][H-P(tBu)3] was prepared by exposing the [PBB-H][H-P(tBu)3] salt to one atmosphere of CO2. Still, the reaction did not occur at room temperature but required heating at 140 °C for several days.
The same group demonstrated that P/Al/CO2 adducts can also be used as a means to reduce CO2 to CO.39 Berke et al. used transition metal-based FLPs in the hydrogenation of CO2 in 2013.61 The authors showed that 0.5 mol% of ReHBr(NO)(PR3)2 and B(C6F5)3 in the presence of C5H6Me4NH catalyzed the formation of formate salt [C5H6Me4NH2][HCO2]. Similarly, in 2015, Zall et al. reported the CO2 hydrogenation catalyst based on a well-defined copper complex with triphos ligand 1,1,1-tris-(diphenylphosphinomethyl)ethane.40 The [MeC(CH2PPh2)3Cu (NCMe)][PF6] complex acted as a Lewis acid in the presence of C10H16N2 (DBU) base, H2, and CO2 to generate the salt [C10H17N2][HCO2].
More recently, Fernández, Breher and co-workers reported a “hidden FLP” based on a five-membered ring phosphorus ylide containing aluminium or gallium as a LA site.66 This system was shown to reduce CO2 to methanol in the presence of borane as a reducing agent and upon the addition of water.
Later in 2018, Zhao et al. described a similar “semi-catalytic” CO2 hydrogenation to formate using a catalytic amount of BCF as LA and an excess of carbonate base M2CO3 (M = Na, K, and Cs).42 This system resulted in the formation of a Lewis pair (K2–[(BCF)2–CO3]) that could react with both H2 and CO2 to produce the BCF–HCO2M (Fig. 7). Using a 10000-fold excess of base compared to BCF, under CO2:H2 = 20:40 bar, in THF for 48 hours at 160 °C, they could reach a TON of 4000 based on the moles of formate produced per mole of BCF.
Fig. 7 Proposed catalytic process for the hydrogenation of CO2 using K2-[(BCF)2–CO3]. Reproduced from ref. 42. Copyright 2019 Wiley-VCH. |
More recently, in 2022, Dyson and Corminbœuf used linear scaling relationships to explore FLPs combinations for the direct hydrogenation of CO2.43 Thousands of FLPs combinations were computationally screened based on the acidity and basicity of the individual components. Their computational approach and results are discussed in more details in Section 3.2.2. In short, the authors showed that balancing the cumulative strength is key to catalytic performance. From the wide set of FLPs explored, those consisting of tris(p-bromo)tridurylborane (tbtb)/DBU with a weak LA and B-Me2F2/pyridine with a weak LB were found to be the most effective ones. Their predicted high activity results from their ability to provide a suitable balance between the limiting steps, i.e. activation of H2 and release of the product. Notably, the tbtb/DBU-based FLP was found to experimentally catalyze the direct hydrogenation of CO2 to formate. It occurs in the presence of 100 equivalents of the base compared to the acid at a very high pressure of 150 bar (25 bar CO2 and 125 bar H2) at 120 °C after 48 hours (Fig. 8). TON of 24 were obtained, and, noteworthy, the product of the reaction was the formate salt of the Lewis base. They also explained the reason why previous attempts to use molecular FLPs for CO2 hydrogenation combining strong LAs like BCF with various nitrogen-containing bases (such as TMP or lutidine) resulted in boroformate salts with no product release. The authors concluded that the formate release step would always be the limiting factor and prevent catalytic turnover, no matter which base is chosen when such a strong LA is used.
Fig. 8 Catalytic hydrogenation of CO2 to formate using tbtb/DBU system. Adapted from ref. 43. Copyright 2022 Wiley-VCH. |
FLP type | LA/LB pair | Inter/Intramolecular | Solvent | T, P, time | Product (yield) | Ref. |
---|---|---|---|---|---|---|
HB(C6F5)2/PPh3@silica | B/P | Inter | Pentane | 80 °C, 40 bar H2, 4 h | [(SiO)2Al(OEt)(OC6H4PH(C6H5)2)][HB(C6F5)2] | 69 |
BCF@silica/tBu3P | B/P | Inter | Toluene | 65 °C, 2 bar H2, 72 h | [SiOB(H)(C6F5)2][tBu3PH] | 70 |
PR3@silica/BCFPR3/BCF@silica | B/P | Inter | Pentane, toluene | RT to 60 °C, 2 bar H2, 2 bar CO2, 48 h | Formic acid | 71 |
B(C6F5)2(Mes)/DABCO@MOF | B/N | Inter | Toluene | RT, 10 bar H2, 48 h | α,β-unsaturated amine | 72 |
B(C6F5)3/DABCO@MOF | B/N | Inter | Toluene | 80 °C, 60 bar H2, 24 h | Alkylidene malonate (88–95%) | 73 |
B-MOF/amine | B/N | Inter | Acetonitrile | 120 °C, 10 bar CO2, 24h | Benzimidazoles | 74 |
B(C6F5)3/MOF(porphyrin) | B/N | Inter | Toluene | RT to 100 °C, 20 bars H2 and CO2, 21 h | [(MeO)B(C6F5)3] then MeOH after hydrolysis | 75 |
B(C6F5)3/diamine@MOF | B/N | Inter | Toluene | RT, 20 bar H2, 48 h | Amine | 76 |
BCF/P-POP | B/P | Inter | Cyclohexane | RT, 6 bar H2 | [(Ar)3PH][HB(C6F5)3] | 77 |
BCF/N-POP | B/N | Inter | Toluene | 80 °C, 60 bar H2, 24 h | Diethyl-2-benzylmalonate | 78 |
P-POP/B-POP | B/P | Inter | Toluene | 60 °C, 1 atm CO2 | Formamide | 79 |
Fig. 9 Formation of silica-supported phosphine Lewis base combined with solubilized Lewis acids. Reprinted from ref. 69. |
Fig. 10 CO2 capture by silica-supported LA and dissolved LB mixture. Reprinted from ref. 71. Copyright 2020 American Chemical Society. |
Since 2015, several computational studies have proposed the possibility of creating MOF-based FLP catalysts.88–93 However, only in 2018 the Ma's group achieved experimentally the introduction of FLPs into MOF in a semi-immobilized manner.73 The Cr-MIL-101 was selected as an adequate MOF platform due to its large open porosity and the presence of Cr open metal sites.94 The selected classical FLP comprises BCF as the LA and 1,4-diazabicycl[2.2.2]octane (DABCO) as a potent LB. One of the N atoms in DABCO is coordinated to the open metal site of the MOF and the second N atom is paired to BCF simultaneously as illustrated in Fig. 11. Although the obtained Cr-MIL-101-LP was not reported for CO2 hydrogenation, the solid efficiently catalyzed the reduction of CN imine bonds with up to 100% yield using HBpin as a reducing agent, as well as the direct hydrogenation of alkylidene malonates under 60 bars H2 for 24 hours at 80 °C giving up to 91% yield. This catalyst was recyclable for up to seven runs for the reduction of imines.
Fig. 11 Synthesis of Cr-MIL-101-LP as a catalyst for CC and CN reduction reactions. Adapted from ref. 73. Copyright 2019, with permission of Elsevier. |
The following year, in 2019, the same group reported the chemoselective catalytic hydrogenation of α,β-unsaturated imine compounds under 10 bar H2 at room temperature in high yields up to 100% using the same Cr-MIL-101 MOF.72 The FLP anchored in the MOF was B(C6F5)2(Mes)/DABCO following the same strategy.73 Interestingly, this immobilized FLP could activate H2 to form MIL-101(Cr)-FLP-H2 at room temperature and hence, form an ammonium hydridoborate salt, [CH(CH2CH2)3NH] [HB(C6F5)2(Mes)]. Powder XRD was used to confirm the retention of the structural integrity of MIL-101(Cr)-FLP-H2, its surface area being of 1120 m2g−1 as obtained by N2 sorption studies. The MIL-101(Cr)-FLP displayed the same productivity for the imine reduction for up to five runs.
Dyson and Stylianou and co-workers reported a water-stable MOF named SION-105 which incorporated a bulky LA-functionalized ligand.74 MOF SION-105 was synthesized by the combination of Eu-(NO3)3 and tris(p-carboxylic acid)tridurylborane (H3tctb) in a 2:1 mixture (Fig. 12). SION-105 allowed for the in situ formation of FLP by employing Lewis basic diamine substrates which could be efficiently transformed to benzimidazoles in the presence of excess silane and 10 bar CO2 as a C1 source at 120 °C for 24 hours, with isolated yields up to 90%.
Fig. 12 Synthesis of SION-105 and the reaction of aromatic o-diamines with CO2 in the presence of SION-105. Reprinted with permission from ref. 74. Copyright 2020 Wiley-VCH. |
In the same year, the same group conducted another study on stoichiometric CO2 fixation using an inverse approach.75 In the MOF-545 hydride framework, which contains Zr6O8 clusters linked by tetrakis(4-carboxyphenyl)porphyrin (TCPP) ligands, nitrogen atoms in porphyrin act as LB and formed an in situ FLP when combined with the BCF LA introduced as a guest molecule (Fig. 13).
The BCF@MOF-545 was subjected in toluene to 20 bars of H2 at room temperature for 1 hour and then to 20 bars of CO2 at 100 °C for approximately 20 hours. Following this procedure, the formation of methoxyborate, [(MeO)B(C6F5)3]− was observed along with the side products methoxyborate and the hydroxyborate, [(HO)B(C6F5)3]−, which were confirmed by mass spectrometry and NMR. During the experiment, the decomposition of the products into [(MeO)B(C6F5)2] and C6F5H was observed. These products could potentially give methanol upon hydrolysis. Notably, the MOF host remained intact even after three runs. Although the NMR analysis showed the presence of the hydrogenated products, these were not quantified.
With a similar approach as in 2019, Ma's group developed a molecular chiral catalyst by incorporating chiral FLPs (CFLPs) into achiral MOFs for asymmetric hydrogenation of imines.76 Guided by a computational screening, the 2,5-dihydro-3,6-dimethoxy-2-isopropylpyrazine bifunctional basic amines were selected to coordinate the Cr(III) open site with one N atom and paired with BCF. The presence of the FLP was confirmed by elemental analysis, FTIR, XPS, EDS and 11B NMR comparison of CFLP@MIL-101(Cr)-H2 and BCF. When applied to the asymmetric hydrogenation of imines (20 bars of H2, 48 hours at room temperature) this catalyst gave yields above 95% in most of the different imines with enantiomeric excess up to 85%.
More recently, in 2022, a covalent organic framework (COF) functionalized with FLP was reported by Hu's group.95 Inspired by the catalytic activity of triaryl phosphine and BCF pair, they developed COF@FLP by a post-synthetic modification of the crystalline brominated COF with triaryl phosphine unit as Lewis base by cross-coupling reaction, followed by the encapsulation of BCF as Lewis acid. These COFs were used for the stereoselective hydrogenation of alkynes into Z-alkene with H2 and effectively recycled for up to ten runs.
Fig. 14 BCF-impregnated phosphine-based polymer for H2 cleavage. Reprinted with permission of ref. 77. Copyright 2017, American Chemical Society. |
Similarly, in 2018, Rose et al. reported a polyamine organic framework with tertiary amine as LB sites, which forms in situ FLP when exposed to BCF in solution.78 The impregnation of BCF was studied by ATR IR spectroscopy showing a new IR band at 1276 cm−1 corresponding to the B–N interaction between BCF and polyamine. As a semi-heterogenized FLP, this polymeric amine in combination with BCF was shown to cleave H2 heterolytically, as confirmed by 11B NMR, and applied in the catalytic hydrogenation of electron-poor diethyl benzylidenemalonate.
Following a strategy based on fully covalently immobilized FLP, Yan and coworkers reported the physical mixture of two linear polymers functionalized with either BCF derivative (LA) or triarylphosphine (LB) as illustrated in Fig. 15.79 Upon CO2 exposure under 1 atm. in toluene, the intermolecular polymeric FLP is assembled to form micelles with bridging CO2 between polymer chains. These micelles were shown to be nanocatalysts for the secondary amine formylation with TON up to 15000 in the presence of phenylsilane as a reducing agent, with efficient recycling for up to eight catalytic runs.
Fig. 15 Self-assembled FLP–polymer micelles for CO2 activation. Reprinted with permission from ref. 79. Copyright 2018, WILEY-VCH. |
From siliceous materials to MOFs, COFs and POPs, the above-reviewed studies evidence that the vast majority of heterogeneous FLPs are actually semi-heterogenized FLPs. As such, they consist of solids made around or functionalized with one FLP partner, while the second partner is non-covalently impregnated within the solid. Such semi-heterogenization strategies might lead to the leaching of one of the partners thus limiting the durability, and the subsequent productivity, of these however promising systems. The fully covalently heterogenized FLPs were based either on phosphine-functionalized metallated silica96 or on the physical mixture of two polymers functionalized respectively with one FLP partner,79 with however ill-defined and randomly distributed active sites as well as questionable utilization of the bulk FLP-functionalized solids. Furthermore, although the reported FLP-containing solids were demonstrated to efficiently activate either H2 or CO2, none of them were shown to perform direct CO2 hydrogenation so far.
Fig. 16 Overview of the main reaction mechanisms characterized by DFT calculations to govern the H2 splitting by (a) intramolecular and (b) intermolecular FLPs. 3D-structures on the right show representative transition-state geometries for: H2 activation by (c) intramolecular FLPs through an intramolecular path. Reprinted with permission from ref. 97. Copyright 2016, Wiley-VCH; (d) intramolecular FLPs through an intermolecular path. Adapted from ref. 98 with permission from the PCCP Owner Societies; (e) an intermolecular FLP. Reprinted with permission from ref. 99. Copyright 2008, Wiley-VCH. |
On the one hand, intermolecular FLPs proceed via the formation of a non-covalent adduct between the LA and LB partners also referred as an “encounter complex”. This is followed by the insertion of H2 into the reactive pocket and its splitting to form a hydrogenated [LB–H]+⋯[LA–H]− ion pair.
On the other hand, intramolecular FLPs may either activate H2 by themselves or dimerize to operate as an intermolecular FLP. The competition between intra- and intermolecular pathways in intramolecular FLPs has not been systematically investigated for a whole range of systems. Thus, setting clear conclusions about the factors that determine their relative likelihood remains challenging. Still, the prevalence of each of these paths might depend on the equilibrium LA⋯LB distance within the intramolecular FLP and on the ability of FLPs to dimerize into stable non-covalent adducts. Up to now, this mechanistic knowledge has been successfully applied to rationalize the H2 splitting activity of a wide variety of FLPs, including mostly P/B98–116 (known as thermally-induced FLPs53,117) N/B pairs,29,97,113,115,118–123 and less often carbene/B124 and N/TM (TM = Ni, Pt) pairs,125 or even that of heterogeneous systems such as hydroxylated indium oxide surfaces.126–128
Computational efforts have been devoted to identify the key factors that govern the H2 splitting, aiming at establishing structure–activity relationships and clear design rules. The primary structural requirement of an active FLP is to offer a suitable LA⋯LB distance. For instance, Vankova and co-workers determined that optimal P⋯B distances for H2 splitting lie within a range of 3 to 5 Å by performing constrained potential-energy surface scans along the P⋯B distance of a P(tBu)3/B(C6F5)3 FLP.129 As for B/N FLPs, Corminbœuf and co-workers identified that the optimal B⋯N distance for H2 splitting is ca. 2.9 Å. Still, the range of distances in which the energy of the TS varies within an energy range of 5 kcal mol−1 spans from ca. 2.6–3.5 Å.68 Thus, when targeting the heterogenization of FLP systems, it is essential to ensure that the positioning of the LA and LB partners within the material does not impose a suboptimal LA⋯LB separation.
Another key factor in FLPs’ reactivity is the electronic properties of the substituents of both LB and LA centers, affecting their basicity and acidity, respectively. A first attempt to set relationships between such molecular features of FLPs and the H2 splitting free energy (indicating its thermodynamic feasibility) was proposed by Pápai and co-workers.53 They carried out a comprehensive computational investigation using a series of intra- and intermolecular FLPs, which led to several conclusions, which were further supported by other works (see ESI† for details):97,114,123,129
(i) The thermodynamics of the H2 splitting correlates with the cumulative acid–base strength of the FLP partners, which can be quantified from proton and hydride attachment energies;
(ii) Linked or intramolecular FLPs benefit from a smaller entropic penalty along the reaction coordinate than intermolecular FLPs and thus, require smaller cumulative acid–base strengths to be active;
(iii) Reaction free-energies for H2 splitting by the analyzed set of intramolecular FLPs was correlated inversely with the LA⋯LB distance. In other words, the shorter the donor–acceptor distance is, the more exergonic the H2 splitting is.
Ye and Johnson further proposed an original approach to explore computationally the immobilization of intramolecular FLPs for CO2 hydrogenation in UiO-66, a porous MOF matrix.89 They aimed to set the very first structure–activity relationships between the structure of molecular FLPs and their ability to activate H2 and hydrogenate CO2 in the H2/CO2 gas stream. Here, we focus on their findings on the initial H2 splitting step, while CO2 hydrogenation will be discussed in Section 3.3. The authors selected a set of 8 B/N FLPs, which on paper, can be covalently grafted on the terephthalate linkers of UiO-66 (Fig. 17a). These consist of a pending pyrazole functionalized with a BR2 moiety that leads to a series of B/N FLPs of increasing acidity by varying R. Computing electronic energy profiles for all systems, the authors tried to find Brønsted–Evans–Polanyi (BEP) relationships130,131 between energetic parameters (reaction energies and barriers) and a set of molecular descriptors. The latter include the acidity of the LA, the basicity of the LB, the electronegativity, chemical hardness and softness of the FLP sites, atomic charges and structural parameters of both the bare FLP and the zwitterionic hydrogenated intermediate.
Fig. 17 (a) Octahedral cage of the UiO-66 metal–organic framework bearing a linker functionalized with a Lewis Pair (X). (b) Adsorption energy of H2 in UiO-66-X plotted against the free-energy of hydride attachment to the B center of the Lewis pair. Reprinted with permission from ref. 89. Copyright 2015, Wiley-VCH. |
The H2 dissociative adsorption energy (i.e. heterolytic splitting of H2) was found to correlate linearly with the hydride attachment energy (Fig. 17b).89 As the LA strength increases, so does the H2 dissociative adsorption energy. Notably, no correlation was found with the proton attachment energy, in line with the experimental findings of Berke and co-workers.132 Still, we can note this computational work only addressed the impact of the substituents at boron (LA) center, resulting in a rather narrow range of proton attachment energies. The analyzed FLPs do not significantly differ in terms of LB basicity, leaving the impact of the latter uncharted.
Another linear BEP relationship was identified between the H2 splitting energy barrier and the LA chemical hardness.89 Increasing the hardness of the LA decreases the barrier for H2 splitting in accordance with Pearson's theory. We wish to stress here that, unlike the H2 adsorption energy, this barrier did not correlate with the acidity of the LA. The BEP principle thus does not fully apply to the H2 splitting process, given the absence of a linear correlation between reaction energies (thermodynamics) and barriers (kinetics). The other BEP relationship found between the H2 splitting barrier and the bond angle formed by Nb, Na and B centers in the bare FLP was ascribed to the fact that larger angles induce strain in the FLP, which lowers the barrier. It is also relevant for the present review that the MOF environment was found to marginally affect the adsorption energy of H2 on the FLP, as similar reaction energies were obtained from both isolated and in-MOF FLP models.
Two years later, Ye et al. revisited the factors that influence the thermodynamics and kinetics of H2 splitting on FLPs grafted in UiO-66 by enlarging their study to four families of intramolecular FLPs including B/N and B/P pairs attached to structurally distinct molecular scaffolds.91 Although quantitative linear correlations between H2 splitting energies and hydride attachment-free energies were found again for each FLP family (with r2 > 0.95), as shown above, they did not apply for all FLP families taken together (r2 = 0.834). This evidenced that besides hydride attachment-free energies, other factors influence the binding energy of H2. An in-depth screening of molecular descriptors revealed that it is also proportional to the LA⋯LB distance in the bare FLP and to the variation of the angles concerning the LA/LB sites and the two consecutive atoms of the scaffold in the direction of the FLP partner upon H2 binding.
Overall, despite the complexity of FLP's reactivity, computational methods have significantly contributed over the years to the generation of valuable knowledge that can be used to guide the experimental design of FLPs to conduct H2 splitting. Still, several points lack a quantitative unambiguous answer. For instance, the exact structural and electronic parameters that determine whether an intermolecular FLP operates through an intra- or an intermolecular mechanism are yet to be understood. Other aspects of the H2 splitting by FLPs that remain controversial or unclear nowadays are the relative weights of the impact of electric field polarization and orbital overlap on H2 splitting, as well as those of LA and LB strengths and how they relate to the FLP's nature and structure (see Fig. S6 and related text in ESI†). Also, although a set of molecular parameters of FLPs have been recognized to influence H2 splitting kinetics and thermodynamics, generic and quantitative structure–activity relationships are still to be established. Most likely, this is because, besides eventual exceptions, the series of FLPs analyzed in each work are constrained to a single family of FLPs and quite often, account for a rather limited number of structures.
More importantly, within the present context of using FLPs not only for H2 splitting but also for subsequent hydrogenation of CO2, balancing the energetics of H2 splitting with those of the subsequent hydrogenation step is crucial. Specifically, FLPs should bind H2 strongly enough to make the H2 splitting step feasible, but not too strong so as to preclude subsequent hydride/proton transfer to an incoming CO2 molecule. In other words, a too strong Lewis base/acid would produce a too weak Brønsted acid/hydride donor, hampering CO2 hydrogenation. This adds an extra layer of complexity to the design challenge: the goal is drifted from minimizing energy barriers while favoring thermodynamics for each single elementary step process to identifying the molecular descriptors and the sometimes-overlooked experimental conditions that simultaneously optimize the performance of two distinct processes.
However, CO2 can be also captured by FLPs as shown in Section 1 and hence, compete with H2 for FLP sites.13 Notably, FLPs often bind CO2 more strongly than H2, which may hamper H2 splitting as the first step of CO2 hydrogenation because of an additional energy penalty associated to CO2 release. Still, such FLP–CO2 adducts are not necessarily detrimental for CO2 hydrogenation, as they could allow distinct mechanistic scenarios, as discussed in detail below.
At this point, computations had already granted atomically-resolved mechanistic details of both elementary steps underlying CO2 hydrogenation (i.e. H2 splitting and H−/H+ transfer). However, it was not until 2015 that Fontaine, Stephan and co-workers combined them both to study the overall hydrogenation mechanism of CO2 with H2 catalyzed by the intramolecular B/N FLPs, 1-BR2-2-NMe2-C6H4 (R = 2,4,6-Me3C6H2 and 2,4,5-Me3C6H2), as represented in Fig. 18a.32 These FLPs were experimentally found to afford the formation of boron-bound formates, acetals and methoxides, the product ratio being sensitive to the experimental conditions of H2/CO2 pressures. Traces of methane were also detected at low CO2 pressure.
The authors then applied DFT calculations to investigate the initial hydrogenation steps.32 They include the hydrogenation of the bare FLP, followed by the formation of formic acid via concomitant H− and H+ transfers to an incoming CO2 molecule. In line with previous calculations by Zimmerman et al.135 and Wen et al.134 the hydrogen transfer step was found to take place with a concerted TS. Thereby, the B-bound hydride is transferred to the electrophilic C atom of CO2, while one of the oxygen atoms of CO2 abstracts the acidic proton from the N center of the FLP. Interestingly, the analyzed FLPs (Fig. 18a) were observed to undergo protodeborylation after H2 splitting (Fig. 19a), giving access to 1-BHR-2-NMe2-C6H4 compounds. After a second H2 splitting step, the latter led to hydrogenated species analogous to that represented in Fig. 19b, which were found to hydrogenate CO2 through affordable free-energy barriers. Fig. 19b shows a representative TS structure for CO2 hydrogenation.32
Fig. 19 (a) First protodeborylation process in hydrogenated forms of 1-BRH2-2-NHMe2-C6H4 (R = 2,4,6-Me3C6H2 and 2,4,5-Me3C6H2). (b) Optimized geometry of the TS for CO2 hydrogenation by 1-BH2(2,4,5- Me3C6H2)-2-NHMe2-C6H4, obtained at the ωB97X-D/6-31 + +G(d,p) level of theory, including benzene as solvent through the SMD continuum solvent model. Adapted from ref. 32. Copyright 2015, Royal Society of Chemistry. |
In 2018, Jiang et al. studied computationally the hydrogenation of CO2 promoted by a series of B/P intramolecular FLPs represented in Fig. 18b–d.136 In addition to the previously established mechanism (H2 splitting followed by the concerted H−/H+ transfer to CO2, see path (i), green arrows in Fig. 20), the authors explored a novel mechanism that starts with the activation of CO2 on the bare FLP (path (ii), purple arrows in Fig. 20). Then, metathesis of H2 with the LB–C bond leads to an intermediate bearing an LA–bound formate and a protonated LB center, which further evolves to release formic acid regenerating the FLP. In most cases, path (i) was claimed to be more favorable than path (ii), except for FLPs shown in Fig. 18b, structures (i) and (iv). Despite the possible contribution of the mechanism associated with path (ii) (Fig. 20), path (i) which involves sequential H2 splitting followed by hydride/proton transfer to CO2 has been consistently adopted in subsequent computational studies on FLP-catalyzed CO2 hydrogenation, including both inter-43,50,137 and intramolecular91,138,139 FLPs as well as classical Lewis pairs.42 There is an exception, though. For the intramolecular B/N FLP displayed in Fig. 18e, Ghara and Chattaraj explored alternative mechanisms whereby both H2 and CO2 are concomitantly activated by the FLP sites.138,140 However, these were found to involve rather high free-energy barriers, ranging from 37 to 50 kcal mol−1. Besides, since the formation of non-covalent FLP⋯H2 and FLP⋯CO2 adducts is generally endergonic, the trimolecular nature of the proposed TSs is expected to make them kinetically unlikely. Such features render these concerted H2 + CO2 activation mechanisms not able to compete with stepwise processes represented in Fig. 20, path (i), as later proved by the same authors.138
In 2016, Liu et al. carried out a computational investigation aimed at finding relationships between the electronic structure of FLPs and the kinetics of CO2 hydrogenation into formic acid.50 In particular, their work begs important and timely questions. Is there any balance or relationship between the H2 splitting and the hydride/proton transfer to CO2 steps? Would a stronger FLP combination favor both steps or exclusively the H2 splitting step as detailed above, thus being detrimental to the kinetics of the hydride/proton transfer step? Seeking to answer these questions, the authors carried out DFT calculations to analyze the free-energy profiles for CO2 hydrogenation by five intermolecular B/N FLPs, based on combinations of B(C6F5)3 and B(p-C6F2F3)3 LAs with TMP, btam and Lut LBs (see Fig. 18f). In line with the conclusions inferred from Section 3.1, the kinetic analysis of the H2 splitting step indicates that the stronger the base or the acid, the lower the free-energy barrier. However, no systematic trend was found between the FLP strength and the barriers for hydride/proton transfer for CO2 hydrogenation. Still, stronger FLPs generally exhibit higher energy barriers for the hydride/proton transfer step assorted with later TS, thus following the opposite trend than that of the H2 splitting step. This was ascribed to the fact that stronger FLPs are more prone to “catch and hold” protons and hydrides, hampering in turn the subsequent hydrogenation of CO2. Later on, this phenomenon was further rationalized based on natural bond orbital and molecular electrostatic potential analyses as well as methods that are based on the application of electron localization functions.139 In short, these showed that lower occupancies of the p(B) orbital result in stronger donor–acceptor interactions upon H2 dissociation, thus easing H2 splitting but hampering CO2 hydrogenation. From a fundamental perspective, this can be interpreted as a strong (weak) Lewis base yields a weak (strong) Brønsted acid after H2 splitting. Similarly, a strong (weak) Lewis acid partner will provide a weak (strong) hydride donor. It is apparent that the factors favouring H2 splitting (strong LA and/or LB) might hinder the H+/H− transfer to CO2.
The results described above paved the way for setting structure–activity relationships. Optimal FLPs for CO2 hydrogenation must balance their ability to split H2 with that of transferring a hydride and a proton from the zwitterionic LBH+/LAH− intermediate to CO2, leading to a scenario where both steps take place through accessible free-energy barriers. Otherwise, too weak or too strong FLPs could suffer from too slow kinetics for H2 splitting (Fig. 21, purple arrow) or hydride/proton transfer to CO2 (Fig. 21, red arrow), respectively.
Corminbœuf and co-workers recently made substantial achievements in this context. In 2022, they reported a comprehensive computational exploration of intermolecular B/N and B/P FLPs for CO2 hydrogenation into formate.43 A wide spectrum of chemical space was covered with a library of 60 N- and P-based LBs and 64 triaryl borane LAs partners (Fig. 22a). Upon mutual combinations, a total of 3840 FLP candidates were computationally screened for CO2 hydrogenation. The relative stabilities of the intermediates and TSs shown in the catalytic cycle of Fig. 22b were computed, considering the LA in catalytic concentrations and LB in excess to provide the driving force for product formation. From the above data set, two sub-pools were extracted fixing one of the Lewis partners and varying the complementary one. These sub-sets allowed exploring the individual impact of the properties of both Lewis partners. To this end, linear free-energy scaling relationships (LFESRs) were established to map the performance of the FLP on CO2 hydrogenation as a function of two chemical descriptors, i.e. the free energy of proton attachment (FEPA) to the LBs, and the free energy of hydride attachment (FEHA) to the LAs, which were estimated as the free energies of the following reactions (eqn (1) and (2), respectively):
LB + NH4+ → [LBH]+ + NH3; ΔGr = FEPA | (1) |
LA + H− → [LAH]−; ΔGr = FEHA | (2) |
Fig. 22 (a) Library of B-based LAs and N- and P-based LBs used to create a pool of intermolecular FLPs, whose potential for CO2 hydrogenation was computationally explored by Corminbœuf and co-workers. (b) Catalytic cycle considered for the generation of DFT-derived TOFs. Reprinted with permission from ref. 43. Copyright 2022, Wiley-VCH GmbH. |
Analysis of the LFESRs over the two sub-sets defined in Fig. 22a yielded volcano plots relating the FEHA and FEPA descriptors to the calculated TOFs. These plots allowed the identification of different regions as a function of the nature of the rate-determining step as illustrated in Fig. 23b (i.e. the TDI/TDTS combination designating the TOF). On the one hand, using the rather weak B-Me4 LA partner, most of the screened FLPs face difficulties in splitting H2 through TS1 as defined in Fig. 22b. Strong amidine bases like DBU provide the optimal kinetic balance among the steps of the cycle, making this combination the most active. Still, the screening of the LAs using the weak pyridine LB as fixed partner revealed an over-stabilization of the boroformate complex 5 (as defined in Fig. 22b) when strong LAs with electron-withdrawing substituents are employed. Such over-stabilization is due to too strong B–O interactions. This adds energy penalty for splitting H2 reflected in the 5/TS1 identity of the TDS/TDTS combination, as the dissociation of 5 to regenerate the bare FLP is an uphill process.
Fig. 23 (a) Map correlating the TOF (log scale) for CO2 hydrogenation into formate to FEPA and FEHA descriptors for the 1664 B/N intermolecular FLPs. (b) Map analogous to that in panel (a) in which each grid point is coloured according to the identity of the TDI and TDTS. (c) Chemical compositions of the FLP combinations (A)–(C). Their corresponding FEPA and FEHA descriptor values are shown below. (D) corresponds to a “mismatched combination” between the LB from (A) and LA from (B). (E) is a poorly active combination with strong LA and LB components. Reprinted with permission from ref. 43. Copyright 2022, Wiley-VCH. |
Graphing both descriptors, FEHA and FEPA, against the calculated TOF revealed a dark-red colored region where the CO2 hydrogenation activity is maximized (Fig. 23a). Suboptimal performances in going from red to blue-colored regions arise either from over stabilization of intermediate 5 or from high-energy lying TSs, as shown in Fig. 23b. Importantly, this work revealed that the acidity and basicity of the components have to be appropriately balanced to attain efficient FLPs for CO2 hydrogenation, creating unprecedented mapping of FLPs chemical composition and linear scaling relationships.43
In the sub-field of intramolecular FLPs, the nature of the molecular scaffold or backbone that supports both LA and LB centers is an additional parameter that affects the energetics of CO2 hydrogenation. This does not only influence the electronic properties of FLP partners but also determines their interatomic distance, allowing for strategic control of the geometrical features of the FLP pocket, which have been demonstrated to govern its reactivity.68,129
Using a series of intramolecular “bridged” B/P FLPs (Fig. 18b–d), Jiang et al. investigated how the length and the chemical nature of the bridge between P and B centers influence their CO2 hydrogenation reactivity.136 For alkyl-bridged FLPs, decreasing the electron-donating nature of groups in the R′ position (near the B center) was found to significantly favor H2 splitting. Still, this goes along with the formation of too stable zwitterionic intermediates that hinder hydride/proton transfer kinetics for CO2 hydrogenation. Similarly, for intramolecular B/N FLPs shown in Fig. 18g, Ghara et al.138 reported that the free-energy barrier for H2 splitting decreases upon the incorporation of an amino group in the para position of the pyridine ring, while the barriers also decrease in the order X = NH > O > S. However, S-containing systems have more difficulties than the others in reducing CO2. Elongating the bridge length in intramolecular B/P FLPs (Fig. 18c) was found to have a detrimental impact on CO2 hydrogenation, as Jiang et al. reported a systematic increase of the barriers for both H2 splitting and CO2 hydrogenation steps due to an increase of the strain in the TSs.136 Vinyl bridges (Fig. 18d), were not identified as promising candidates neither because they yield very stable zwitterionic intermediates after H2 splitting, hindering hydride/proton transfer steps.136
The flexibility of the backbone, which impacts LA⋯LB separation, has been also identified as a relevant parameter to tune the reactivity of intramolecular B/P FLPs towards H2 and CO2 to produce formic acid.137 Inspired by the application of dimethylxanthene backbones to prevent the self-quenching of the FLP components,141 Delarmelina et al. analyzed in detail a series of xanthene-inspired scaffolds which display various P⋯B distances (Fig. 18h–l and Fig. 24).137 Compound H (Fig. 24) is found active for H2 splitting in line with experimental observations,141 but exhibits a very height barrier for the CO2 hydrogenation step. Compared to H, both compounds I and J (Fig. 24) show slightly higher but still affordable barriers for H2 splitting. However, the impact of their backbone on their CO2 hydrogenation step was found to be very different: Catalyst I, which displays a shorter P⋯B distance than J, has a very high barrier for the second step; conversely, in J, the more rigid backbone and longer P⋯B distance led to a substantial decrease of the free-energy barrier for hydride/proton transfer to CO2 when compared to H and I, which translates into a reduction in the activation energy for the whole process. Further tuning of the FLPs H–J via modification of the electronic properties of the substituents at the B and P groups was found to play a secondary role compared to the stiffness of the backbone. Of note, Delarmelina et al. also reported that more polar solvents stabilize both the zwitterionic intermediate and the TS of the second step. Still, such stabilization is very similar in both cases, having an overall negligible impact on the height of the CO2 hydrogenation barrier.137
Owing to the well-known ability of FLPs to activate CO2-forming LB–C(O)O–LA adducts,54,142–144 the competition between H2 and CO2 for the binding sites of FLPs is intrinsic to the conditions required for catalytic CO2 hydrogenation, in which the FLPs are exposed to a mixture of both reagents. Thus, depending on the thermodynamics and kinetics of CO2 binding, the latter may be regarded as a potential poisoning route, precluding CO2-hydrogenation reactivity. If the formation of a covalent FLP–CO2 complex is exergonic and faster than H2 splitting (Fig. 25, light-blue dashed lines), the H2 splitting process will be slowed down. This is because the overall barrier that needs to be overcome corresponds to the free-energy difference between the FLP–CO2 complex and the TS1 for H2 splitting, as exemplified in Fig. 25 (ΔG‡overall,1, light blue arrow). In the worst-case scenario, this can even inhibit H2 splitting (and subsequent CO2 hydrogenation) if CO2 binding is irreversible or if the resulting overall barrier is too high to be overcome at the experimental conditions. Still, it should be noted that for systems that can operate through path (ii) (Fig. 20) thanks to a low FLP–CO2 → TS1′ barrier, the preferred binding of CO2 with respect to H2 splitting might not lead to a poisoning route but rather to an in-cycle species. However, the factors that determine the preference for one path or another remain elusive, making it difficult to predict whether a given FLP would be able to bypass the limitations underlying competitive CO2 binding processes by proceeding through this alternative mechanism.
Fig. 25 Schematic free energy profile for the early steps of CO2 hydrogenation, highlighting the impact of possible stabilizing interactions between the bare FLP and CO2 (light-blue dashed lines) or formic acid (red dashed lines) on the kinetics of H2 splitting. Black, light-blue and red arrows represent the individual free-energy barrier for the H2 splitting step (ΔG‡), and overall barriers for the same process measured from stable FLP–CO2 (ΔG‡overall,1) and FLP–HCOOH (ΔG‡overall,2) complexes, respectively. TS1’ accounts for possible hydrogenation of CO2 through path (ii) of Fig. 20. |
Assuming path (i) in Fig. 20 (sequential H2 splitting followed by H+/H− transfer to CO2) as the mechanism at work, computational studies aimed to identify the factors that determine the relative affinity of FLPs toward H2 and CO2,91,121,145 hence providing guidance to favor H2 over CO2 binding. Changing the electronic properties of FLP sites via tuning of LA/LB substituents of structures shown in Fig. 18m–q was only found to either strengthen or weaken the adsorption of both H2 and CO2. Conversely, shortening the LA⋯LB distance can simultaneously strengthen the H2 adsorption and weaken the CO2 adsorption.91 Also, thermodynamics of H2 and CO2 binding can be controlled by playing with the electron-conjugation properties of FLPs. In particular, gains in aromaticity of the FLP structure along CO2 binding were found to result in lower activation energies and more enhanced exergonicity due to the stabilization of TS and product structures.121 On the other hand, reduction of anti-aromaticity along H2 splitting is also beneficial in terms of kinetics and thermodynamics.145 Thus, although not straightforward to control, these properties may be also tapped to achieve tailored H2/CO2 binding selectivity.
Akin CO2 or H2, the formic acid product resulting from CO2 hydrogenation can also form strong covalent adducts with bare FLPs, as illustrated in Fig. 25 (red dashed lines), which could hamper the subsequent H2 splitting process to initiate a new cycle. One of the rare examples of computational works that considered the participation of FLP–HCOOH complexes as in-cycle species for CO2 hydrogenation (Fig. 22b) revealed that these complexes act as the resting state of the catalyst in a wide range of LA/LB combinations (Fig. 23).43 Comparable occurrences may be also expected for further reduced species such as methanol. The latter has been reported to strongly bind boron LA centers of FLPs forming B-methoxy species, both computationally134 and experimentally.62–65 To circumvent these issues, as well as to provide the thermodynamic driving force for the per se endergonic CO2 hydrogenation reaction, bulky bases such as DBU may be used to “trap” the HCOOH product as [Hbase][HCOO] salts, while leaving unaltered LA sites of FLPs by avoiding quenching through steric repulsion (see Table 1).
Finally, as the vast majority of FLPs rely of boron LA centers, common FLPs are highly sensitive to moisture due to the oxophilicity of boron atoms. Moreover, they can also suffer from other degradation or inhibition pathways, such as protodeborylation, as shown in Section 3.2.1 (Fig. 19a).32 conspicuously, this could either promote the complete degradation of the FLP or lead to partially protodeborylated species with enhanced activity. DFT calculations suggested that the ability of FLPs to undergo protodeborylation may be modulated by tuning the steric hinderance around the B–C bonds, so that crowded environments hamper this process, allowing control over the electronic and steric properties around boron LA sites during the course of the reaction. In general, this process is commonly overlooked in the literature, although its assessment appears to be necessary to evaluate or predict the activity of an FLP towards CO2 hydrogenation. Otherwise, FLPs that may be initially predicted to be inactive could become active upon protodeborylation and vice versa.
Most of the knowledge gained from the computational study of molecular FLPs may apply to the design of heterogenized FLPs inside porous matrices for the sake of avoiding the limitations inherent to homogeneous systems. However, such a heterogenization strategy poses a series of additional challenges, related to the stability of the functionalized materials, the dynamics and accessibility of FLPs within the pores of the material, the selectivity of the material towards CO2/H2 adsorption, or the impact of confined environments on the electronic and structural properties of FLPs, as discussed in the following section.
A couple of computational studies have considered the possibility that Zr-based MOF may possess built-in frustration between LA and LB centers. Along that line, Slater et al. have conducted extensive ab initio molecular dynamics (AIMD) simulations on the defective Zr-based UiO-66.148 They show that missing terephthalate linkers are charge-balanced by hydroxide anions bonded to under-coordinated Zr sites, where rapid proton shuttling may be involved with physisorbed atmospheric water molecules. At high activation temperatures, AIMD simulations further reveal that FLP sites may form consisting of an undercoordinated Zr site (Lewis acid) adjacent to a hydroxide bonded to a proximal undercoordinated Zr atom (Lewis base). The authors suggest that similar defects may exist in a wide range of MOFs, whereby increased catalytic activity and tailoring of the functional behavior could be targeted.
More recently, a detailed DFT study considered the mechanistic details of CO2 hydrogenation to methanol in the linker-defective UiO-66, exploring various pathways and free-energy profiles on the resulting metal-based FLPs.149 The defective model of UiO-66 is created by removing an organic carboxylate linker from a Zr6-oxocluster node, leaving behind two adjacent undercoordinated Zr sites. One of them is proposed to act as a Lewis acid, while the –OH group added to the other Zr site for charge balance purposes acts as a Lewis base. The hydrogenation consists of a three-stage transformation: (i) CO2 is hydrogenated into formic acid (HCOOH); (ii) HCOOH is converted to formaldehyde (HCHO) via hydrogenation and dehydration; (iii) HCHO is hydrogenated into CH3OH. For CO2 hydrogenation to HCOOH, three possible pathways were investigated (Fig. 26). Typically, in pathway III, H2 is initially located above the open Zr site (LA) and the hydroxide (LB) and dissociates, leading to a Zr–H hydridic site and a vicinal Zr–H2O water molecule. CO2 is then hydrogenated into formic acid (FA) via a slightly endergonic concerted mechanism overcoming an energy barrier of 10.2 kcal mol−1. The cis-FA desorbs to the gas phase and then transforms into the trans-FA. HCOOH is then converted at the FLP in a stepwise fashion into formaldehyde HCHO and CH3OH via facile H2 dissociation and the concerted H+/H− transfer to HCOOH and HCHO. Interestingly, the authors explored the impact of a possible confinement effect in UiO-66 on the catalytic activity by considering two sizes (small/large) of clusters. It was found the energy barriers obtained with the two types of clusters differ only by 2 kcal mol−1, suggesting that the confinement effect in UiO-66 is negligible.
Fig. 26 Proposed pathways for CO2 hydrogenation to HCOOH in the linker-defective UiO-66. Reprinted with permission from ref. 149 from the Royal Society of Chemistry. |
Leaving apart these specific cases where the pristine MOF was studied for its intrinsic “FLP-like” catalytic properties for CO2 reduction,149 we are not aware of computational studies of functionalized FLP@MOF systems where the MOF itself plays a catalytic role as a Lewis partner for CO2 conversion.
Johnson and Ye pioneered the design of FLP@MOF by covalently grafting an intramolecular FLP to the terephthalate linker of UiO-6688,89 or embedding the FLP in UiO-6790 to study catalytic CO2 hydrogenation mechanisms. They aimed to functionalize UiO-66 with 1-(difluoroboranyl)-4-methyl-1H-pyrazole (P-BF2), developed to mimic 1-[bis(pentafluorophenyl)boryl]-3,5-ditert-butyl-1H-pyrazole FLP, which is known to cleave H2 heterolytically and fixe CO2.150,151 The less bulky P-BF2 was modelled by removing the tert-butyl groups and replacing C6F5 moieties by F atoms. Constructing the UiO-66-P-BF2 in silico with one P-BF2 group per primitive cell, static-DFT calculations in the gas phase provided relative energy profiles for CO2 chemisorption, H2 dissociation, and CO2 hydrogenation to formic acid, without accounting for entropic contributions.88 Two pathways were compared: CO2 activation followed by H2 cleavage or concerted H2 cleavage and CO2 hydrogenation. The first pathway involving the chemisorption of CO2 in UiO-66-P-BF2 lead to the formation of a very stable adduct and a potential poisoning of the FLP. The second pathway was found energetically more favourable than the first with a much lower activation barrier, making it the desired one. The MOF environment slightly reduced the barriers for H2 dissociation and CO2 hydrogenation compared to the isolated P-BF2 catalyst, though the absence of entropic contributions suggests caution.
Expanding on their initial proof-of-concept, the authors explored UiO-66-P-BR2 variants for CO2 hydrogenation,89 testing various electron-donating and -withdrawing substituents on the acidic partner while keeping the same diazole basic partner (see Fig. 17a). The process in the gas phase involved H2 heterolytic splitting (as presented in Section 3.1), CO2 chemisorption, and concerted hydride and proton transfer to form formic acid. They again identified BEP relationships for rapid screening of functional acid groups, finding that energy barriers for CO2 hydrogenation correlated linearly with H2 dissociation energy in the UiO-66-P-BR2 series (Fig. 27a). Weaker H atom bonds consistently facilitated H−/H+ transfers, easing CO2 hydrogenation. H2 splitting energies served as a first estimate for CO2 hydrogenation barriers. A Sabatier activity map combined BEP relationships for H2 dissociation barriers (i.e. B's hardness) and CO2 hydrogenation barriers (i.e. H2 dissociation energy), revealing optimal functional groups with hardness above 3.8 and H2 binding energies between −0.45 and −0.1 eV (red area in Fig. 27b).
Fig. 27 (a) BEP relationship between CO2 hydrogenation barriers and dissociation energies of H2 in UiO-66-P-BR2. (b) Sabatier activity map for overall CO2 hydrogenation at 298 K. The CO2 hydrogenation barriers are given by a BEP relationship with H2 dissociation (adsorption) energies (x-axis) and H2 dissociation barriers have a BEP scaling with the hardness of the LPs. Reprinted with permission from ref. 89. Copyright 2015, American Chemical Society. |
So far, we are not aware of any experimental report on the functionalization of UiO-66 with P-BF2. The functionalization of this well-known MOF with FLPs is yet to be reported and may be hindered by practical synthetic limitations, discussed above in Section 2.2. Notably, Johnson et al. were the first to have screened the reactivity of FLPs anchored in MOF and attempted to find descriptors for their catalytic activity in CO2 hydrogenation. Still, descriptors related to the MOF's impact as a “secondary coordination sphere” of the FLP on the CO2 hydrogenation-free energy profile are yet to be identified.
Along this line, the same authors have compared the impact of pore size and topologies of MOFs on the dissociation of H2, the chemisorption of CO2 and its reduction into formate, examining MOFs with increasing pore sizes: MIL-140B (∼8 Å) < UiO-66 (∼5 Å and ∼9 Å) < MIL-140C (∼10 Å) < UiO-67 (∼12 Å and ∼16 Å).92 The Lewis pairs are comprised of N and BF2 moieties linked via a C atom, embedded in the organic linker, except UiO-67 where the LP is part of a side chain pointing towards the pore center, generating minimal steric interaction with the hosting framework. The calculated dissociation energies of H2 are found similar in all systems, suggesting that pore size and confinement effects have no significant impact because of the small size of H2. In contrast, CO2 chemisorption energies exhibit significant differences among the four MOFs. They depend on the orientation of the bound CO2 and the proximity of carboxylate O atoms of nearby linkers, which may result in strong repulsions with CO2. The strongest chemisorption of CO2 observed in UiO-66-P-BF2 is allowed by the minimal steric interactions between the LP and this MOF. By contrast, the lower strengths of CO2 chemisorption in the three other systems are a subtle convolution of the flexibility of the MOF's linker and specific confinement effects around CO2 imposed by the number of linkers per secondary building units. Turning to the CO2 hydrogenation process, the authors claim that confinement and steric hindrance may allow the selection of specific MOF topologies to (i) favor the dissociation of H2 over the chemisorption of CO2, thus increasing the resistance of the MOF to CO2 poisoning, and (ii) improve the catalytic performances by pre-activating CO2 via a bent physisorbed configuration.
Johnson et al. also considered UiO-67 (Fig. 28a) – due to its larger pore volume and surface area than that of the isostructural UiO-66 – for hosting an intramolecular FLP, namely UiO-67-(NBF2)4 (Fig. 28b).90 Notably, the adsorption energy of CO2 was found to be much lower than that of H2 (−0.22 eV versus −0.50 eV). This allows a much easier desorption of chemisorbed CO2 and avoids the poisoning of the FLP by CO2 that was previously observed in UiO-66-P-BF2. Overall, UiO-67-(NBF2)4 provides multiple LP catalytic sites for H2 dissociation, making facile reduction of CO2 to methanol possible without requiring diffusion of the intermediate reactants or products into adjacent pores. The reaction pathways to methanol involve a series of concerted two-hydrogen transfer steps where the following pathway, CO2 → cis-HCOOH → CH2(OH)2 → CH2O → CH3OH, has the lowest potential energy surface, HCOOH acting as a proton shuttle.
Fig. 28 (a) Representation of the native UiO-67 and its biphenyl linker and (b) the modified UiO-67 incorporating a ligand with a B/N intramolecular FLP. |
Heshmat reinvestigated in depth the above UiO-66-P-BF2 catalyst using ab initio molecular dynamics (AIMD) simulations, computing free energy surfaces (FESs) using metadynamics.93 Here, the thermal and entropic contributions to the free energy profile were taken into account thanks to the motion of H2/CO2 molecules in the gas phase within the LP-functionalized MOF. The mechanism of CO2 hydrogenation towards formate was shown to be more eventful than initially proposed by Johnson et al.,89 with the possibility of a stepwise mechanism in addition to the concerted one, while providing unique features about the dynamics of the reaction.
The free-energy barrier for H2 splitting was found to be affordable (18.8 kcal mol−1), being ca. 6 kcal mol−1 higher than the reported electronic energy barrier reported in Johnson's work,89 due to the incorporation of entropic effects. Also, simulations revealed that the formation of the zwitterionic intermediate (hydrogenated FLP) is thermodynamically favorable by ca. 6 kcal mol−1. Importantly, although the free-energy barrier for CO2 activation by the bare FLP was estimated to be 10 kcal mol−1, being lower than that for H2 splitting, the reverse free-energy barrier for CO2 desorption was computed to be affordable as well, accounting for 21.1 kcal mol−1.93 This suggests that even though CO2 binds stronger to the FLP than H2, the formation of FLP–CO2 adducts does not necessarily imply poisoning of the catalyst. Still, in the presence of a CO2/H2 mixture, H2 splitting would be precluded by an overall free-energy barrier close to 30 kcal mol−1 (H2 splitting barrier + energy span associated to CO2 desorption from the FLP). On these grounds, it is postulated that suitable CO2 hydrogenation conditions might involve exposing the MOF sequentially to separate streams of H2 (to activate it on the FLP) and CO2 (to hydrogenate it), rather than to a mixture of CO2 and H2, whereby the binding of CO2 to FLP sites would hamper H2 splitting.
The FESs for the next CO2 hydrogenation step were calculated using cis and trans conformations of dissociated H2. The cis conformation reveals an asynchronous concerted mechanism, while the trans conformation shows a stepwise mechanism with H− is transferred to C(CO2) first, followed by H+ to O(CO2) after 400 fs. Both mechanisms are kinetically feasible and endergonic. The hydrogenation of CO2 may be thus fully described using the two H−⋯B and H−⋯C(CO2) distances as collective variables. Atomic charges analysis indicates a more polarized B–H bond vs. the N–H one, causing an earlier hydride than proton donation. This is also reflected in the 1 eV higher HOMO–LUMO energy gap between the lone pair of O(CO2) and N–H σ* than that between the B–H σ orbital and π* of activated CO2. MD simulations reveal strong interactions between the HCOOH product and Zr-clusters of the MOF, in line with known observations in Zr-based MOFs,152 implying the need for continuous HCOOH removal for high conversion rates.
Another recent computational study investigated the Zr-based UiO-66 MOF functionalized with alanine boronic acid for CO2 hydrogenation.153 The energy profile for CO2 hydrogenation in the gas phase of the immobilized FLP in UiO-66 was compared to that of the isolated molecular FLP. These calculations suggested an overall stabilization along the whole energy profile as the reaction proceeds in the FLP@UiO-66 solid, although entropic contributions were not considered. A notable decrease of 4.6 kcal mol−1 in the H2 splitting barrier was estimated upon immobilization of the FLP. Also, the comparison of the various components of the energy decomposition analysis (EDA)154 (electrostatic, Pauli repulsion, dispersion and orbital interactions) revealed a greater stabilization of reaction intermediates and transition states in the FLP@MOF system when compared to that obtained with the isolated FLP, pointing towards a beneficial impact of the confinement of the reaction within the MOF's pores.
However, there are key issues that need to be addressed in order to fully benefit from the aforementioned advantages of the immobilization of intermolecular FLPs. Firstly, it is crucial to control the synthetic or post-synthetic modifications of the MOF to achieve the desired relative positioning of the LA and LB centers within the required distance. Secondly, it is required to preserve the steric accessibility and chemical environment of both the LA and LB partners once anchored, in order to avoid any detrimental interactions between the LA or LB partners and the MOF. This raises yet-to-be-answered questions regarding the dynamics of both the MOF and the LA/LB partners. Lastly, it requires identifying the most suitable MOF structures from the vast number of existing ones that can meet the aforementioned constraints, while also considering the practical issues related to the compatibility of their synthesis protocols, thermal and chemical stability with those of the FLPs to be immobilized. In that respect, computational chemistry approaches have a key role to play, not only for exploring the potential catalytic performances of LA⋯LB pairs once immobilized within the MOF, but also for predicting the overall impact of their immobilization in the MOF pores or channels. The latter includes considerations on LA and LB's spatial partitioning, steric, chemical or electronic features and dynamical behavior within the MOF, to ultimately identify design principles.
In that direction, Corminbœuf et al. explored the constrained spatial arrangements of the prototypical BF3/pyridine Lewis pair.68 Using the B⋯N distance (d) and the angle (Φ) formed by the B's empty orbital and the N's lone pair as descriptors, the complete free energy profile for CO2 hydrogenation for intermediates and transition states as defined in Fig. 22b was evaluated at each increment of the B⋯N distance (0.1 Å steps), while allowing Φ to adapt freely (Fig. 29a). Fitting it to a Morse potential, a scaling relationship was created that predicts the reaction's rate or TOF (turn over frequency) based solely on the B⋯N distance. This allowed them to estimate the resulting gain or loss of catalytic performance of the « immobilized » FLP to the unconstrained one (ΔTOF), thus providing direct insight into how restricting the FLP's geometry affects its activity. The complete activity profile was constructed as depicted in Fig. 29b.
Fig. 29 FLP-catalyzed hydrogenation of CO2 following the catalytic cycle described in Fig. 22b. (a) Energy–distance scaling relationship showing the variation of free-energies of each intermediate and transition state with the d descriptor (d = LA⋯LB separation in Å), fitted with Morse potentials (dashed lines); (b) Activity map describing the relative TOF of constrained FLPs with respect to unconstrained FLPs as a function of the key descriptors, d and Φ; (c) Intramolecular FLPs that are mapped in (b) and the geometric descriptor values extracted from the corresponding intermediate 2. Copyright 2022 from ref. 68. Angewandte Chemie International Edition published by Wiley-VCH GmbH under the terms of the Creative Commons license. |
Remarkably, these calculations reveal that the catalytic activity of an N/B FLP may be boosted by constraining its geometry. The peak performance (dark red area in Fig. 29b) is estimated to be eleven orders of magnitude higher than that of the unconstrained FLP (ΔTOF = 0). High efficiency can be achieved over a wide range of distances as long as the N⋯B orientation is suitable. The optimal orientation for H2 activation is in the range of Φ = 90°–135°, which suggests that promoting this step through immobilization could enhance the catalyst's efficiency. Notably, the predictive capability of the activity map was successfully validated against reported intramolecular N/B FLPs and known experimental trends from Fontaine et al. on amine-boranes32 for CO2 hydrogenation.
This pioneering work of Courminbœuf et al. presents a comprehensive and generic conceptual approach to immobilizing FLPs in any type of solid-state system, regardless of structure and chemical environment provided by the host, as the explored geometric constraints closely resemble those that arise from FLP's immobilization. Although the specific influence of the host on the FLPs’ chemistry is not addressed, these findings convincingly demonstrate that constrained geometry correlates with a distinct catalytic activity. The strategic confinement of the FLP within specific distances and orientations can unexpectedly trigger enhanced catalytic activity when compared to the unconstrained FLP. Overall, this proposed approach has the potential to inspire future investigations for screening the catalytic activities of other chemical LA–LB partners in CO2 hydrogenation and identifying optimal constrained geometries to be further incorporated into MOFs or porous polymers for instance.
In this regard, the development of computationally driven strategies to identify sweet spots where the energetics of both steps are balanced to achieve optimal CO2 hydrogenation activity represents a matter of current research, and some guidance and rules of thumb have been recently reported. Specifically, optimal CO2 hydrogenation activity has been found to require a moderate cumulative acid/base strength, which can be estimated as the sum of proton and hydride attachment energies to the LA/LB partners (FEPA and FEHA, respectively). Thus far, systematic explorations have been only carried out with intermolecular B/N FLPs, which revealed that the TOF is maximized for LA/LB combinations with FEPA/FEHA ratios ranging linearly from ca. −35/−10 to +25/−35 kcal mol−1.43 Moving away from these ranges leads to too strong pairs, which form too stable zwitterionic hydrogenated FLP intermediates or bind products too strongly, or to too weak pairs, which have difficulties splitting H2.
Also, intramolecular FLPs allow for some “tricks” to move orthogonally away from these trends, by controlling either the LA⋯LB distances or the rigidity of the molecular scaffolds, as exemplified by the work by Delarmelina et al.137 More rigid backbones and longer LA⋯LB distances (of ∼4.4 Å) were found to substantially reduce the free-energy barrier for hydride/proton transfer to CO2, compared to more flexible scaffolds with LA⋯LB distances of <4 Å. While for H2 splitting, optimal B⋯N distances were found to lie in a range of 2.6–3.5 Å,68 not a single study has claimed yet what is exactly the optimal LA⋯LB distance to promote CO2 hydrogenation and how it varies as function of the nature of the FLP sites. Still, it is reasonable to assume that optimal performances would be found at the overlap of distance ranges that are suitable for both H2 splitting and H+/H− transfer elementary steps.
In addition, potential poisoning pathways such as competitive binding of species on FLP sites or protodeborylation, are analyzed independently in the literature, but rarely (if ever) in a joint manner to grant a full mechanistic picture of the reaction. Regarding the latter routes, some trends have been found for specific sets of FLPs. These include: shortening of LA⋯LB distances to favor the activation of H2 by FLPs over that of CO2; or increasing the steric hinderance around B sites to prevent undesired protodeborylation events.
Moreover, beyond the determination of the adequate energetics balance and the inclusion of poisoning in computational approaches, the calculated data should be translated into experimental parameters such as partial gas pressure and temperature, which remain crucial towards applicable CO2 hydrogenation systems. For instance, a pioneering system by Corminbœuf and Dyson allows for the FLP-driven CO2 reduction at 120 °C under 130 bars, specific conditions that were not anticipated from the calculated data but must have required intensive experimental screening.43
Two experimental aspects have to be considered when turning computational results into bench-scale trials: the synthetic accessibility of the FLP synthons and the conditions to be applied to reach the targeted FLP's reactivity. Considering the reactivity of computationally designed FLPs for the CO2 hydrogenation, the conditions reported in the two recent examples by Zhao42 and Corminbœuf,43 both claiming for a catalytic FLP system, were 160 °C under 60 bars of CO2/H2 and 120 °C under 130 bars, respectively. Far from being trivial, the findings of such harsh conditions to assess the efficiency of carefully designed Lewis pairs can only rely on a large screening of pressures and temperatures, including a range of values not often investigated for molecular catalysis. Thus, the quest for novel experimental FLP systems for CO2 hydrogenation, whether predicted by preliminary computational studies or not, must encompass from its beginnings these specific ranges of temperatures and pressures, especially when compared to transition-metal-based counterpart systems operating under smoother conditions.155
Besides the deployment of dedicated organic synthesis of the predicted FLPs, limitations that are related to their heterogenization may emanate from the reactivity/stability of the porous solids (MOF, POP, COF) themselves upon their functionalization with the targeted FLPs. Additional drastic limitations may arise from the catalytic conditions themselves (pressure, temperature, solvent…). In the proposed two-step synthetic pathway devised by Ye and Johnson90 for targeting the MOF-heterogenized FLP UiO-67-P-BF2, the initial step hinges upon a palladium-catalyzed Negishi cross-coupling reaction involving the commercially available 3-iodo-1H-pyrazole and a benzylzinc bromide derivative, which has been reported to yield up to 82%.156 However, it is worth noting that the synthesis of the 2,4,5-trimethylphenylzinc bromide coupling partner, though commercially procurable, is protected by patents and thus not readily accessible for research groups. Moving on to the second potential step, it entails the oxidation of p-xylene to terephthalic acid following the Amoco process.157 It is noteworthy that this oxidation process employs acetic acid as a solvent, compressed oxygen at approximately 200 °C, and a combination of cobalt, manganese, and bromide catalysts. These rigorous synthetic conditions raise concerns regarding the potential undesired oxidation of the benzylic position within the linker precursor, consequently leading to the decomposition of the intended motif.158 Lastly, the borylation of the pyrazole moiety has been reported utilizing a two-step methodology, which involves the quantitative utilization of an additional BCF:carbene adduct.150 Furthermore, it is important to acknowledge that this step relies on heterogeneous solid–liquid reactivity occurring at the interface of the MOF. It also requires to simultaneously ensure the stability of the HBF2 reactant against the reactivity of the Zr–OH groups present on the MOF's surface, as well as any pending carboxylic acid defects. This presents a significant challenge due to the potential risk of hydrolysis159,160 and the subsequent release of highly toxic hydrofluoric acid and fluoroboric acid, particularly given the potential diffusion limitations within the pores of the MOF.
Anticipating these synthetic limitations by computational approaches is extremely challenging. Consequently, the discovery of heterogenized FLP systems currently relies on the “trial-and-error” explorations of porous matrices from a broad range of potential candidates. However, the knowledge gained from both homogenous systems and theoretically derived structure–activity relationships can be used to narrow the range of FLP candidates to be immobilized, thus saving time consuming experimental efforts. Future work may explore potential synergies between the chemistry of organic linkers and their functionalization, and that of metal nodes in MOFs72,73 towards the construction of heterogeneous FLPs, which are still rather scarcely covered. Current developments in the field of hybrid porous materials hold the potential to make it possible to achieve unequaled FLP-based heterogeneous catalysts with the only limit being the synthetic capacity to prepare their elementary building blocks.
The mapping of crossed activity is a significant and achievable milestone. The next frontier is the computational exploration of functionalized MOFs, including the partitioning of multiple LA and LB grafts, as well as their dynamics within porous matrices. With the advanced capabilities of numerous force fields and QM/MM techniques, studying the realistic implantation of such LA/LB pairs is well within reach, promising groundbreaking developments in this field.
These challenges become even more complex when transitioning to porous solids through the heterogenization of FLPs. This heterogenization should enable precise control over the FLP's geometry, reactivity, and synergistic effects such as enhancing reactant concentrations locally. Recent pioneering examples already installed a dialogue between computations and experimentations but the implementation of computationally designed in experimentally constructed FLP still require intensive efforts for their in catalytic applications. The development of computational methodologies to go even deeper into the predictions of the synthetic and catalytic conditions combined with the proliferation of synthesis methods based on almost infinite variations of compositions and structures of porous hybrid solids is aimed to push the limits of current systems. Given the evolving nature of this research field, the collaborative efforts of synthetic and computational chemistry are poised to yield sustainable FLP-based catalytic systems for direct CO2 hydrogenation in the foreseeable future.
Footnotes |
† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d3cs00267e |
‡ These authors contributed equally to this work. |
This journal is © The Royal Society of Chemistry 2024 |