MANUSCRIPT Designing Effective Frustrated Lewis Pair Hydrogenation Catalysts.Pdf

Please do not adjust margins

Chem Soc Rev

TUTORIAL REVIEW

Designing Effective ‘Frustrated Lewis Pair’ Hydrogenation Catalysts

Received 00th January 20xx, Accepted 00th January 20xx Daniel J. Scott,* Matthew J. Fuchter and Andrew E. Ashley*

DOI: 10.1039/x0xx00000x The past decade has seen the subject of transition metal-free catalytic hydrogenation develop incredibly rapidly, transforming from a largely hypothetical possibility to a well-established field that can be applied to the reduction of a www.rsc.org/ diverse variety of functional groups under mild conditions. This remarkable change is principally attributable to the development of so-called ‘frustrated Lewis pairs’: unquenched combinations of bulky Lewis acids and bases whose dual reactivity can be exploited for the facile activation of otherwise inert chemical bonds. While a number of comprehensive reviews into frustrated Lewis pair chemistry have been published in recent years, this tutorial review aims to provide a focused guide to the development of efficient FLP hydrogenation catalysts, through identification and consideration of the key factors that govern their effectiveness. Following discussion of these factors, their importance will be illustrated using a case study from our own research, namely the development of FLP protocols for successful hydrogenation of aldehydes and ketones, and for related moisture-tolerant hydrogenation.

Key learning points 1. Rational FLP design must be based on an understanding of the relevant key mechanistic steps. 2. H+ and H– affinities are crucial parameters and must be balanced relative to both the substrate and each one another. 3. Reactivity can be inhibited by either exceedingly high or low steric bulk, and the ideal profile will be substrate-dependent. 4. Intramolecular FLPs offer the possibility of improved reactivity, but at the cost of more challenging catalyst development. 5. Successful FLP design requires an understanding of inhibition/decomposition mechanisms, which are often LA-related.

some examples do exist (notably the various well-known

Introduction to FLP chemistry carbenes and related group IV R2E species) and in recent years there has been great interest in the isolation and study of such Since the earliest days of the field, the study of homogeneous compounds in the hope that they may demonstrate a similar catalysis has been all but synonymous with the study of potential for catalysis, and ultimately provide counterparts or transition metal (TM) catalysis, particularly in the activation of alternatives to TMs (many of which suffer from high toxicity, 1 relatively inert small molecules or of strong chemical bonds. high cost, or low abundance). Indeed, some such compounds The privileged reactivity demonstrated by TM compounds can have been shown to readily undergo a variety of ‘TM-like’ be attributed to their characteristic electronic structures, with reactions. For example, Bertrand et al. were able to partially occupied sets of d-orbitals leading to the demonstrate activation of inert E—H bonds (E = N, H) by simultaneous presence of both nucleophilic/Lewis basic and addition to singlet carbenes, with the observed reactivity electrophilic/Lewis acidic frontier orbitals located on the same attributed to simultaneous interaction of the substrate with 2 atom. It is the ability of both types of orbital to interact the electrophilic 2p and nucleophilic sp orbitals on the synergistically with a substrate that allows for the activation of reactive carbon centre, in a manner clearly reminiscent of TMs 2 functional groups that would normally be kinetically inert, (Fig. 1b). even where these groups would be unreactive towards a Lewis Nevertheless, the adaptation of stoichiometric bond activation chemistry by main group compounds into useful acidic or Lewis basic site on its own (illustrated for H2 in Fig. 1a). Comparable electronic structures are uncommon for catalytic cycles has proven highly challenging, and only very stable main group compounds, which explains their general few such examples have been reported. This can broadly be inability to mediate similar catalytic reactions. Nevertheless, attributed to the typical low stability of unsaturated p-block compounds, which leads to difficulties in catalyst regeneration (and thus prevents closure of the catalytic cycle) and tendency towards decomposition, as well as more general difficulties in Department of Chemistry, Imperial College London, SW7 2AZ, UK. E-mail: [email protected] initial isolation and handling.

This journal is © The Royal Society of Chemistry 2017 Chem. Soc. Rev., 2017, 00, 1-3 | 1

Please do not adjust margins Please do not adjust margins

ARTICLE Journal Name

CO2 and other p-block oxides; alkenes and alkynes; acidic and hydridic E—H bonds).6 The TM-like reactivity of FLPs has again been attributed to

the simultaneous action on the H2 molecule of energetically- accessible Lewis acidic and basic orbitals (Fig. 1c).7 However, In 2006 Stephan and co-workers described results that unlike the other examples discussed so far, in FLPs these have led to an alternative and much simpler approach for orbitals are spatially separated from one another, and 3 obtaining TM-like reactivity using main group compounds. localised on different functional groups. As a consequence, it

The authors observed that under an atmosphere of H2 a is typically relatively easy to fine-tune the properties of one solution of an intramolecular phosphine-borane was converted (e.g. sterics or electronics) without having a significant impact to into the zwitterionic phosphonium borohydride, via on the other. Given also that FLPs are readily constructed from activation and cleavage of the homopolar H—H bond (Fig. 2a). robust, well-understood functional groups (usually an amine or It was quickly realised that this reactivity could be generalised phosphine combined with a strong fluoroaryl-substituted to simple intermolecular phosphine/borane combinations, borane), this means that FLPs are uniquely well-suited among provided that their steric bulk was sufficient to prevent adduct unsaturated p-block compounds for the development of 4 formation between the Lewis acid (LA) and base (LB). catalytic applications. Indeed, within two years of the first

Because of their sterically-induced inability to quench one report of FLP H2 activation, the same authors also described another, such systems have come to be known as ‘frustrated’ the first example of FLP-catalysed hydrogenation; the 5 Lewis pairs (FLPs). Subsequent work by many groups has conversion of simple imines to amines (Fig. 2b).8 Subsequent shown that FLP reactivity can be observed with a much wider rapid progress has expanded the scope of FLP-catalysed variety of both inter- and intramolecular LA and LB hydrogenations to include substrates ranging from alkenes and combinations [e.g. boranes, boreniums, alanes, carbocations, aromatics to aldehydes and ketones.6 FLPs have thus provided silyliums and stannyliums; phosphines, amines and N- the first general methodology for catalytic hydrogenation that heterocyclic carbenes (NHCs)], and can lead to activation of a does not require the use of a TM. great many other small molecules and chemical bonds (e.g.

Daniel Scott is an EPSRC doctoral Dr Matthew Fuchter is a Reader in prize fellow currently working in the Chemistry at Imperial College. The group of Dr Andrew E. Ashley at Fuchter group has a wide-ranging Imperial College, where he had track record in the design, synthesis previously obtained his PhD studying and application of organic molecules the development of FLP catalysis. His in chemistry, medicine and materials. current research focuses on the Representative examples include the development of Fe-based catalysts design and development of novel

for homogeneous N2 fixation. bioactive probes, the study of novel chiral semiconducting molecules, and the development of novel FLP catalysts.

2 | J. Name., 2012, 00, 1-3 This journal is © The Royal Society of Chemistry 20xx

Please do not adjust margins Please do not adjust margins

Journal Name ARTICLE

A note on FLPs and other branches of chemistry As with any chemical transformation, the rational development of effective FLP hydrogenation catalysis is above Given the simplicity of the FLP concept, it is perhaps not all dependent upon a basic understanding of the key steps surprising that it has begun to be invoked in relation to quite a underlying the reaction mechanism. Understanding the broad range of chemical processes. This includes discussion of mechanism by which FLPs are able to activate H2 is thus clearly newly developed or discovered reactions, but also of many central for considerations of hydrogenation catalysis. As that pre-date the FLP formalism. Notable examples in the already discussed, H—H cleavage is believed to occur through latter category include Piers-type hydrosilylation,9 metal-ligand simultaneous interaction with both the LA and LB (Fig. 1c). cooperative catalysis,10 and the chemistry of solid surfaces,11 While for intramolecular FLPs this leads to a feasible among many others. In one particularly dramatic example, bimolecular reaction, for intermolecular systems it implies the KOR (R = alkyl) was reported to catalyse ketone hydrogenation need for an entropically unfavourable termolecular step. This in turn indicates that some transient interaction must form under very forcing conditions, with H2 activated by ‘the joint between two of the reaction components prior to the action of a […] base and a Lewis-acid […] on the H2 molecule’, clearly foreshadowing the development of FLP-catalysed involvement of the third, in order to render the bond cleavage hydrogenation.12 An analogy can also be drawn between FLP step kinetically accessible. At first it was supposed that this was likely to be either a weak LA←H or LB→H interaction. H2 activation and the chemistry of TM·(H2) complexes, where 2 2 However, after initial attempts to find either experimental or binding to a Lewis acidic TM creates a Brønsted acidic H2 moiety that can be deprotonated by LBs.13 As a consequence, computational evidence for such interactions were it can sometimes be unclear where the formal boundaries of unsuccessful, further theoretical studies instead suggested the ‘FLP catalysis’ field should be placed (for example, where it formation of so-called ‘encounter complexes’ in which the LA may overlap with LA catalysis, especially in reactions that do and LB are held together by weak intermolecular interactions not involve a co-catalytic LB). Ultimately, it is up to the in such a way that they are pre-organised for subsequent H2 14 individual chemist to decide whether invocation of the FLP activation (Fig. 3). Subsequent experimental work has concept is helpful in understanding the system in question, as confirmed the existence of intramolecular interactions for two is the case for many other descriptive models of chemistry PR3/B(C6F5)3 FLPs (R = tBu, or Mes; 2,4,6-trimethylphenyl), (e.g. valence bond versus molecular orbital theory). through NMR spectroscopic techniques (e.g. observation of 1 19 intermolecular H/ F correlations via 2D HOESY; Fig. 3), and as such this is believed to be the general mechanism by which H Tutorial review aims and scope 2 activation is effected by a diverse range of FLPs.15 Nevertheless, it should be emphasised that these In this tutorial review we will outline the key factors that can determine the outcome of FLP-catalysed hydrogenation reactions, and illustrate how these principles can be used in the design of effective catalysts. Note that while they will not be discussed here explicitly, most of these principles will also be directly applicable to the development of various other FLP- catalysed reactions.

Key aspects of FLP catalyst design

Understanding the reaction mechanism

Fig. 3 H2 activation by intramolecular and intermolecular FLPs. For the latter, the termolecular reaction step is facilitated by formation of a weakly-bound ‘encounter complex’ between the LA and LB,

into which H2 can add.

This journal is © The Royal Society of Chemistry 20xx J. Name., 2013, 00, 1-3 | 3

Please do not adjust margins Please do not adjust margins

ARTICLE Journal Name investigations have largely been limited to ‘typical’ are used. These may involve hydride transfer prior to amine/borane or phosphine/borane FLPs, and alternative protonation (if the substrate is sufficiently electrophilic and mechanisms cannot conclusively be ruled out in other can stabilise the resulting negative charge), direct intermolecular systems. hydroelementation of the substrate, or activation of the + Following H2 activation, catalytic hydrogenation requires substrate through coordination of the LA rather than H (Fig. transfer of the resulting H+ and H– fragments to the substrate 4b). In some cases it is also possible that hydrogenation can in order to close the catalytic cycle. In almost all examples, this proceed without the need to add an auxiliary LB catalyst (‘LA- is believed to involve initial protonation of the substrate in only catalysis’), if the substrate is sufficiently basic and can act order to activate it towards subsequent hydride transfer (Fig. as the basic component to activate H2 directly in combination 4a). This can be attributed to the ubiquitous use in FLP with the LA (Fig. 4c; imines, for example, are commonly hydrogenation catalysts of boranes incorporating very strongly hydrogenated by this mechanism). When attempting to electron-withdrawing fluoroaryl-substituents as LAs; following rationally design or optimise a reaction it is crucial to – H2 activation these form relatively stable [Ar3BH] anions that determine if one of these alternative mechanisms might be are not sufficiently powerful hydride donors to reduce the operative. This can usually be achieved fairly simply through unactivated substrate. Only after protonation (or, at a stoichiometric reaction of the substrate with pre-formed + – + – minimum, hydrogen-bonding to [LB·H] ) does the substrate [LA·H] reagents (e.g. [Bu4N] [LA·H] salts, which contain an become sufficiently electrophilic for further reaction to occur. inert countercation) in the presence and absence of possible Nevertheless, some hydrogenations can proceed via activators (such as additional LA or ‘H+[WCA]–’, where [WCA]– alternative mechanisms, particularly where less ‘typical’ LAs is a weakly-coordinating anion; Fig. 5). Likewise, computational studies may be used to provide additional insight.

Tailoring LA and LB strength

The ‘strength’ of the LA and LB components used to construct an FLP are of crucial importance to the success of FLP-

catalysed hydrogenation reactions. FLP H2 activation has been

reported using LBs that vary in strength by over 20 pKa units, and LAs whose calculated hydride ion affinities (G for LA+H– →[LA·H]–) vary by more than 140 kcal/mol. In an important study, Pápai and co-workers analysed the thermodynamics by

FLP H2 activation by considering it as the sum of five separate conceptual steps (Fig. 6):16

+ – Heterolytic cleavage of H2 into H and H The size of this term will depend on factors such as the solvent (with more polar solvents making ionisation more favourable), but is independent of the FLP used.

4 | J. Name., 2012, 00, 1-3 This journal is © The Royal Society of Chemistry 20xx

Please do not adjust margins Please do not adjust margins

Journal Name ARTICLE

such as solvent polarity. In particular, apolar solvents may lead to precipitation of [LB·H]+[LA·H]– salts; in these cases the ion pairing term becomes effectively very large, and may provide

the main thermodynamic driving force for H2 activation.

In general, of the five terms outlined above, only two are expected to vary very significantly upon variation of the FLP: H+ attachment and H– attachment. Thus, the thermodynamic

ability of any FLP to activate H2 will critically depend on the magnitude of these terms, and hence with the combined ‘strength’ of the LB and LA; a conclusion that has been found to compare very well with experimental results. In particular, it can be seen that when designing FLP hydrogenation catalysts, the key measures by which the ‘strength’ of the LB and LA should be judged are their proton affinity (PA) and hydride ion affinity (HA), respectively. Experimentally-determined proxies

for PA have been extensively tabulated in the form of pKa values.17 Importantly, these are often available for a variety of different solvents (solvation having a significant effect not just

on PA and pKa but also HA; in general, more polar solvents are expected to render FLP activation of H2 by neutral LA/LB combinations more favourable, by stabilising the ionic products relative to the neutral reactants). Experimental values for HA are unfortunately far less abundant and so alternative measures of LA strength are often used as alternatives to aid FLP design (for example the commonly- employed Gutmann-Beckett method, which uses changes in 31P chemical shift to probe the strength of binding between 18 LAs and Et3PO). Nevertheless, the relationships between HA and other measures of abstract ‘Lewis acidity’ or Separation of any LA←LB adduct into the uncoordinated Lewis ‘electrophilicity’ are not always trivial and may show very pair different sensitivity to other relevant parameters such as steric For an FLP this term must, by definition, be close to zero. In bulk (other chemical probes will have different steric profiles some instances FLP reactivity can be observed for Lewis pairs than H–).19 As such, these values are often most reliable (and that do form a weak adduct (discussed in the next section); so most useful) as guidelines when comparing structurally- however, even in these cases this term would not typically be similar LAs that show variation in positions distant from the expected to be too large. acidic centre, rather than for comparison of more diverse LAs from different ‘families’. It is also often useful to consider such + – Attachment of H and H to the LB and LA, respectively values in conjunction with calculated estimates of HA; These terms are highly variable and depend above all on the fortunately, values for a fairly diverse collection of FLP- choice of LA and LB. relevant main-group LAs were recently reported by Heiden and Latham.20 – Any stabilising interaction between the resulting [LA·H] and Taken together, PA and HA values afford an invaluable + [LB·H] moieties (e.g. ion pairing, dihydrogen bonding) predictive tool for the design of FLP hydrogenation catalysts. Computational studies have suggested that this term, while By comparing the values for prospective FLPs with those of not negligible in magnitude, does not vary significantly across a systems already reported in the literature, it is possible to selection of intermolecular FLPs (only neutral LAs and LBs were anticipate their likely degree of reactivity towards H2. If the considered; it is not clear to what extent this conclusion will combined PA and HA are very low, then H activation will be 16 2 hold for charged species). For intramolecular systems the highly disfavoured, and successful hydrogenation catalysis is – + [LA·H] /[LB·H] pairing term is typically larger, as enthalpically- likely to be infeasible. Conversely, for effective catalysis the favourable ion pairing can be achieved without such a combined PA and HA of the FLP should also not be excessively significant entropic penalty. H2 activation in these systems is high, as this will lead to a very stable and hence unreactive thus generally more favourable than in intermolecular systems + – + [LB·H] [LA·H] H2 cleavage product. In such cases either H or of equivalent LA/LB strength. The magnitude of this additional H– transfer (or both) to the substrate will be unfavourable, and stabilisation can be highly variable, and depends on the linker turnover will again be limited. If the reaction mechanism used (often unpredictably; vide infra). Again, the size of this involves LA activation of the substrate (Fig. 4b) this also term can also be expected to depend appreciably on factors requires that H2 activation be sufficiently reversible to ensure

This journal is © The Royal Society of Chemistry 20xx J. Name., 2013, 00, 1-3 | 5

Please do not adjust margins Please do not adjust margins

ARTICLE Journal Name the presence of some ‘free’ LA (assuming 1:1 LA:LB The basicity of the substrate is of even more importance in stoichiometry). It should be noted that a number of ‘weak’ LA-only catalyst systems, where it is directly involved in H2 FLPs have actually been observed to effect successful catalytic activation (Fig. 4c). Typically, this reaction pathway is feasible hydrogenation despite not activating H2 to a sufficient extent for more strongly basic substrates, while less basic analogues for it to be observed by standard NMR spectroscopic benefit strongly from addition of a stronger auxiliary LB which techniques. In these cases the ability of a FLP to effect can speed H2 activation and act as an intermediate ‘proton transient, reversible H2 cleavage can often be demonstrated by shuttle’. admitting HD gas or a mixture of H2 and D2, and observing In some cases it may also be important to consider the isotopic scrambling to form the statistical 2:1:1 mixture of HD, basicity of the intended reaction product. In a particularly

H2 and D2 (Fig. 7). Thus, B(C6F5)3 has been found to effect extreme example, Stephan et al. reported that while B(C6F5)3 is successful catalytic hydrogenation in combination with LBs as capable of mediating the stoichiometric hydrogenation of 21 weak as simple ethers (aqueous pKaH < 0). In contrast, FLPs simple anilines to cyclohexylamines, catalytic turnover is consisting of the same LA and very powerful LBs such as N- prohibited due to the high basicity of the product (amine pKaH heterocyclic carbenes (NHCs; aqueous pKaH ~ 20-25) have not ~ 10 in H2O), which prevents protonation of the much less yet found use in hydrogenation catalysis, even though they basic aromatic substrate (aniline pKaH ~ 5 in H2O; note that 22 readily activate H2. In this context, it is noteworthy that the while this value relates to protonation at nitrogen, the initial pKa of H2 itself has been experimentally estimated to be site of protonation required for reduction is actually at carbon, approximately 35 in THF.23 In principle, the LA-free which will be less basic). In effect, the Brønsted acidity is 25 deprotonation of H2 could be thought of as a conceptual ‘levelled’ to the weak cyclohexylammonium species (Fig. 8a). limiting case for FLP H2 activation, where a very large PA term Conversely, Paradies et al. have described a degree of is necessary in order to compensate for negligible HA. autocatalysis in certain borane-catalysed imine In addition to combined PA and HA, it is important that the hydrogenations.26 This is attributed to the increased basicity of

PA affinity alone (and, in principle, HA alone) is tailored to the product amines, which means that rate-limiting H2 those of the substrate and product (analogous to the activation becomes more favourable as the reaction proceeds electronic fine-tuning typically required of TM catalysts). As (Fig. 8b; both of these examples involve LA-only catalysis). shown in Fig. 4a, the standard mechanism for FLP-catalysed + hydrogenation requires that the [LB·H] intermediate be a Balancing steric bulk sufficiently strong Brønsted acid to protonate the substrate (or else activate it appreciably through hydrogen-bonding), which Perhaps the most obvious variables in the design of FLP places an upper limit on the strength of the LB that can be catalysts are the steric bulk of the acidic and basic centres. The used. These principles were elegantly illustrated by Paradies fundamental FLP concept clearly requires that the LA and LB and co-workers during the development of phosphine/B(C F ) 6 5 3 possess sufficient combined bulk to prevent formation of a FLPs for alkene hydrogenation, where it was found that sub- optimal rates were obtained when using phosphine LBs that were either too weak (where disfavourable H2 activation is rate-limiting) or too strong (where substrate protonation to form an intermediate carbocation becomes rate-limiting instead).24 Notably, different optimum LB strengths were found when the basicity of the substrate was changed, highlighting the need to tailor LB strength to the specific substrate under investigation.

6 | J. Name., 2012, 00, 1-3 This journal is © The Royal Society of Chemistry 20xx

Please do not adjust margins Please do not adjust margins

Journal Name ARTICLE strong classical adduct. If these functional groups are too small particular, the decomposition of highly reactive FLPs may be then mutual quenching will eliminate the ability of the FLP to suppressed by allowing them to form a reversible adduct. For engage in the H—H cleavage step that is crucial to catalysis example, while investigating the use of NHC LBs in FLP

(even if PA/HA and other factors would otherwise be chemistry, Tamm et al. found that, while an NHC/B(C6F5)3 FLP favourable, the unfavourable separation term in Fig. 6 will undergoes rapid rearrangement to form an unreactive become insurmountably large). Steric bulk, particularly around ‘abnormal NHC’ adduct over the course of a few hours at room the LA, is also a key factor in determining substrate scope and temperature, the equivalent less hindered NHC/B(FXyl)3 pair functional group tolerance for FLP hydrogenation catalysts. [FXyl = 3,5-bis(trifluoromethyl)phenyl)] forms a reversible Soós and co-workers have emphasised the value of the ‘size- normal adduct that is stable up to significantly elevated exclusion principle’ in FLP design, where the use of especially temperatures, while still retaining its FLP-type reactivity 22 bulky LAs can allow for the effective hydrogenation of less towards H2 (Fig. 10a). bulky imine substrates, for example, by limiting unproductive Taking the concept of thermally-induced frustration to its adduct formation between the LA and product amines.27 Use extreme, it can actually be possible for a classical adduct to of a bulkier LA also means an FLP can be formed using a less display FLP-like reactivity without ever achieving complete bulky LB (or vice versa). separation of the Lewis pair. Ashley et al. have previously + – Nevertheless, is it not simply the case that using bulkier described how the adduct [iPr3Si←PtBu3] [B(C6F5)4] , which components will lead to a more active FLP catalyst. Extremely bulky LAs will lead to similarly bulky [LA·H]– reductants, whose size can lead to low kinetic reactivity, particularly if the hydride needs to be transferred to a relatively bulky substrate. In particularly extreme cases steric bulk may even inhibit H2 activation. For example, while investigating the very hindered

LA MesB(C6F5)2, Soós et al. found that H2 activation was much slower when using LBs that were also very bulky (e.g. 2,2,6,6- tetramethylpiperidine), compared to when less hindered LBs 27 of comparable pKa were employed (e.g. quinuclidine). As a result, and perhaps counterintuitively, it can sometimes be productive to pursue the use of less bulky LAs, even if this appears to lead to the formation of a classical LA←LB adduct. Provided that any such dative interaction is reversible, transient cleavage can generate the active FLP in situ. In particular, Lewis pairs that appear to form a strong adduct at room temperature may dissociate significantly at elevated temperatures: this has come to be known as ‘thermally- induced frustration’.28 While the need to separate the LA and LB provides an additional energetic barrier that must be overcome prior to H2 activation (the LA←LB separation term in Fig. 6 in no longer negligible), this is potentially outweighed by the kinetic advantages of generating a less bulky and more reactive [LA·H]– reductant (Fig. 9). Reversible adduct formation can also have significant implications for the stability of the catalytic Lewis pair. In

Fig. 10 Reversible adduct formation can stabilise FLPs against decomposition, for example in the systems shown in (a). FLP-like Fig. 9 A qualitative summary of how catalytic activity can typically reactivity can also sometimes be observed by Lewis pairs without vary as a function of FLP steric bulk, assuming combined HA/PA even transient separation (b). This can be compared with direct and other parameters are suitable for catalysis. The ideal catalyst addition of H2 across certain polar chemical bonds (c; this is must be neither too large nor too small, but the ideal mid-point another strategy that has recently been used to achieve TM-free will be substrate-dependent. catalytic hydrogenation).

This journal is © The Royal Society of Chemistry 20xx J. Name., 2013, 00, 1-3 | 7

Please do not adjust margins Please do not adjust margins

ARTICLE Journal Name

+ – does not dissociate to form PtBu3 + [iPr3Si] [B(C6F5)4] even at high temperature, is nevertheless capable of activating H2 in an FLP-like manner.29 This is attributed to transient lengthening and weakening of the Si←P interaction, which generates a structure analogous to an FLP encounter complex that is capable of inserting H2. From here, it is possible to see an analogy between FLP H2 activation and the direct hydrogenolysis of certain weak or polar chemical bonds, which has also recently begun to be exploited to develop mild protocols for TM-free catalytic hydrogenation of alkenes (Fig. 10b,c).30

Choosing between intramolecular, intermolecular and ‘LA-only’ FLPs Further complicating the design of intramolecular FLPs, the As noted previously, tethering the LA and LB together to form choice of linker will also directly impact both the steric and an intramolecular FLP transforms H2 activation from a formally electronic properties of the attached LA and LB centres, which termolecular into a bimolecular reaction step, which is can lead to difficulties in their rational fine-tuning. expected to have a much lower entropic barrier. As such, Intramolecular FLPs also typically require longer, more optimised intramolecular FLPs might intuitively be expected to demanding synthetic routes for their preparation than be superior catalysts than their intermolecular counterparts unlinked LAs and LBs (of which many of the most commonly (cf. TM catalysts, where the acidic and basic functionalities are used are commercially available). Thus, the expectation that by necessity also localised on a single molecule). Indeed, many intramolecular FLPs might ultimately provide superior catalysis investigations of intramolecular FLPs have appeared to support must be balanced against the much greater ease and speed this conclusion. For example, in 2008 Erker et al. reported one with which intermolecular catalysts can be developed (note of the first such systems: an ethylene-linked P/B system that that translating a successful intermolecular catalyst into an showed far greater activity as an imine hydrogenation catalyst intramolecular analogue is also not necessarily trivial, for 31 than previous intermolecular P/B FLPs. similar reasons). In particular, intermolecular FLPs lend Nevertheless, it is important to acknowledge that the themselves very well to screening efforts that can rapidly development of intramolecular FLP catalysts suffers from identify promising catalytic leads (especially valuable when the some significant drawbacks. In particular, the activity of such optimum catalyst is highly substrate-dependent). By contrast, systems is typically highly sensitive to the nature of the linker even a rather modest screen of five acidic and five basic used to connect the LA and LB, in a manner that is not easily moieties would require the synthesis of 50 separate predictable. As such Erker et al., expanding on their earlier intramolecular FLPs, even if only two possible linkers were work, have shown that P/B FLPs with a variety of simple alkyl investigated. Mechanistic investigations are also often simpler 32 linkers show dramatically different reactivities towards H2. using intermolecular systems, for similar reasons, as it is For example, in the series of oligo(methylene) linked FLPs easier to synthesise possible intermediates, or to exclude Mes2B(CH2)nP(C6F5)2, the members with n = 2 and n = 4 are either the acidic or basic component (cf. Fig 5; for an both active hydrogenation catalysts, while the intermediate intramolecular system these investigations would require the member with n = 3 is unreactive towards H2. Similarly, Aldridge further synthesis of separate monofunctional model and co-workers have reported that while a dimethylxanthene- compounds). linked P/B FLP readily activates H2 under mild conditions (room The advantages enjoyed by intermolecular FLPs are even temperature, 1 bar H2), no such activation is observed using an more apparent in the ‘LA-only’ catalytic systems discussed 33 otherwise identical dibenzofuran-linked system (Fig. 11). previously, where only one reaction component needs to be Such differences in reactivity can be attributed to varied and the overall reaction mixture is significantly thermodynamic changes that arise when using different linkers simplified. For example, the groups of Stephan and Crudden + – (for example if the resulting ‘H ’ and ‘H ’ moieties are held too have shown how simple screening studies can be used to far apart for effective ionic stabilisation) or, in other cases, to rapidly optimise the structure of borenium LAs for imine kinetic factors that relate to the ease with which different hydrogenation catalysis.35,36 Again, however, this advantage intramolecular FLPs are able to preorganise themselves into a must be weighed against the expectation that, with fewer conformation suitable for H2 activation. In extreme cases, if no variables available for optimisation, the best LA-only system such conformation is energetically accessible, then formally may be less effective than the best inter- or intramolecular intramolecular FLPs may in fact only activate H2 in an equivalent. As an example, it has been reported that, while intermolecular fashion, as is the case for the original, rigid hydrogenation of weakly-basic imines such as PhCH=NSO2Ph systems reported by Stephan et al. (see Fig. 2a).34 can proceed using B(C6F5)3 as the sole catalyst, greatly

Please do not adjust margins Please do not adjust margins

Journal Name ARTICLE

improved rates can be achieved upon addition of P(Mes)3 as a decomposition of the ubiquitous B(C6F5)3 in the presence of 37 co-catalytic LB. simple alcohols (or H2O) which is believed to occur via initial coordination to form ROH→B(C6F5)3 adducts in which the O—H Understanding and controlling inhibition and decomposition has been dramatically acidified, prompting intramolecular pathways protodeborylation (Fig. 12b; note that the decomposition products ROBAr have much lower HAs, so are considerably 2 less likely to activate H than the initial BAr ).38,39 In general, Just as important as designing an FLP that will be able to 2 3 aryl group transfer of this type should be particularly facile perform all the component steps of a catalytic cycle (H 2 from anionic, 4-coordinate borate intermediates. Any catalytic activation; H+ and H– transfer) is ensuring that the potential protocol that involves a build-up of such intermediates in the catalyst can avoid any irreversible inhibition or decomposition presence of a suitable electrophile (which may simply be more steps that might hinder catalysis. Several general of the initial borane LA, leading to sequestration of a second considerations have already been mentioned in earlier equivalent of borane in the form of a BAr – anion) is therefore sections with respect to catalyst inhibition. For example, steric 4 likely to be particularly susceptible to decomposition, tuning can be used to avoid adduct formation between the particularly if the reaction is required to run at significantly LA/LB and any other basic or acidic functional groups present, elevated temperatures. while reversible adduct formation can be exploited to help stabilise overly reactive FLPs. A large number of chemical bonds aside from H—H are Case study: C=O hydrogenation and moisture known to be susceptible to cleavage by FLPs, which can lead to tolerance quenching of the LA and LB. As such, if any of these functionalities are present in the substrate and/or product then efforts should be made to minimise the reactivity of the Over the past several years we have focused on the FLP towards them. In practice, given the overwhelming development of some of the first FLP systems capable of reliance of FLP hydrogenation catalysis on boron-based LAs catalysing the hydrogenation of aldehydes and ketones, and of and early p-block LBs, reducing reactivity towards other moisture-tolerant FLP hydrogenation catalysis. In this section functional groups without a concomitant reduction in we will discuss the rational development of these systems in order to illustrate the use of the principles that have been reactivity towards H2 is not always easy, and may require the use of less typical FLP components. An example of this will be outlined so far (and which are summarised in the learning discussed in the case study at the end of this review. points listed at the start of this review). Inevitably, relevant decomposition routes will be highly The first goal of our work was to develop a protocol for dependent on the precise system under study. Nevertheless, FLP-catalysed hydrogenation of aldehydes and ketones to because most of the FLP hydrogenation catalysts reported to alcohols, which had not previously been reported. Based on date are based on bulky, electrophilic fluoroaryl-substituted the principles outlined above it was possible to identify the boranes it is worth specifically considering the decomposition following key factors: of these compounds, which typically involves loss of an aryl  The standard mechanism for FLP-catalysed hydrogenation – group to an electrophile (Fig. 12a). A specific example is the requires protonation of the substrate prior to H transfer. Because organic carbonyls are very poor bases (aqueous pKa < 0), this suggests the LB employed in the FLP must also be very weak. This in turn means a LA with fairly high HA will

be necessary so that H2 activation is feasible.  Unlike isoelectronic imines and amines, organic carbonyls and alcohols lack steric bulk around their Lewis basic oxygen atoms, which suggests that relatively bulky LAs might be desirable to avoid inhibitory product→LA and substrate→LA adduct formation. Conversely, however, these neutral oxygen bases are much weaker donors than their nitrogen-based counterparts, which could mitigate this issue, and overly bulky [LA·H]– anions might lead to slow hydride transfer. It therefore seemed sensible to investigate LAs with a range of steric profiles.  The use of ‘LA-only’ systems for stoichiometric carbonyl

hydrogenation had been reported previously, using B(C6F5)3 as the catalyst.38 These investigations had confirmed that

direct H2 activation using the substrate as LB is feasible; however, decomposition of the borane under these highly acidic conditions meant that turnover could not be

Please do not adjust margins Please do not adjust margins

ARTICLE Journal Name

observed. It seemed likely that a stronger auxiliary LB might cheaply commercially available they are well-suited towards

lead both to improved H2 activation kinetics, and to more catalyst screening efforts, and because they can be used as stable Brønsted acidic intermediates. It was also speculated solvents the entropic penalty faced by intermolecular FLPs

that further improvements in stability might be observed when activating H2 can be minimised. for a system with ‘thermally induced frustration’. Because The system chosen for initial investigation was the

the desired reactivity lacks precedent, the initial goal was to B(C6Cl5)(C6F5)2/THF Lewis pair; previously studies had shown obtain a system that is catalytically competent, prior to full that adduct formation in this system is highly reversible, optimisation. It therefore seemed sensible to investigate leading to effective catalysis in the hydrogenation of other intermolecular rather than intramolecular systems. substrates.41 As hoped, this system was also able to achieve  Though some FLPs are known to activate C=O bonds, it was successful catalysis in the hydrogenation of a model ketone judged that the most significant inhibitory side-reaction was (acetone), but turnover in the reaction was unexpectedly likely to be activation of the product O—H bonds, as limited due to side-reactions of the product alcohol. previous borane-based FLPs had been reported to be very Fortunately, subsequent screening (made simple by the sensitive to inhibition by hydroxylic species, including decision to pursue an intermolecular system) was able to 37 alcohols. While this had typically been attributed in rapidly identify commercially-available B(C6F5)3 and 1,4- general terms to the oxophilicity of boron, a closer analysis dioxane solvent as a superior protocol that avoids this issue.42 of the literature suggested that the more specific problem Having confirmed that catalytic systems of this type are relates to irreversible deprotonation of the highly Brønsted capable of tolerating simple alcohols, it was realised that 39 acidic borane-alcohol adduct (aqueous pKa < 0; Fig. 13). similar systems might also be tolerant of another significant hydroxylic inhibitor of FLP catalysis: H2O (low air and moisture Developing boron-based FLP catalysts40 tolerance have typically been severe limitations of early FLP catalysis). This was quickly confirmed, although reactions were

observed to be significantly slower than under strictly Based on the above considerations, and given the well- anhydrous conditions, which necessitated the use of increased established utility of highly electrophilic triarylborane LAs in H pressures.43 Indeed, closer inspection of both the ‘wet’ and FLP hydrogenation catalysis, it was decided to begin 2 anhydrous reactions suggested that, despite their apparent investigations by examining the boranes B(C Cl ) (C F ) (n = 6 5 n 6 5 3-n tolerance, activation of O—H bonds remains a significant 0-3) whose chemistry we had investigated previously, and with limiting factor in the rates of these reactions (as indicated, for which we were therefore familiar.41 These also satisfied the requirement of having large HAs and a range of different steric bulk. To begin with, preliminary mechanistic investigations + – were carried out in which pre-formed [Bu4N] [HBAr3] salts were reacted with a model substrate (acetone). These – confirmed that direct reactions of [HBAr3] with the substrate, either alone or in the presence of additional BAr3, were unlikely to be feasible mechanistic steps (cf. Fig. 5). Thus, catalytic hydrogenation would have to proceed via the standard mechanism (Fig. 4a), which involves protonation of the substrate. It was therefore decided to combine the initial borane LAs with simple ethers as LBs (aqueous pKaH < 0; Fig. 14). The use of such weak bases should ensure that substrate activation through protonation or hydrogen-bonding is feasible, while also preventing irreversible deprotonation of any ROH→BAr3 adducts (cf. Fig. 13). Reversible coordination between the ethers and boranes might also reduce any tendency towards decomposition (e.g. via alcoholysis; cf. Fig. 12b). Furthermore, combinations of this type had previously been shown to be effective in the hydrogenation of other weakly-basic substrates.41 Finally, because these LBs are

Please do not adjust margins Please do not adjust margins

Journal Name ARTICLE example, by 11B NMR spectroscopy), even in the absence of any very strong bases. Note that increased temperatures could not be used to increase reaction rate as significant hydrolysis/alcoholysis of the borane was observed in these cases, consistent with the previously-anticipated decomposition route.

The switch to Sn-based FLP catalysis44

Development of the borane-based systems described above had emphasised the importance of O—H cleavage as a mechanism for inhibition of the FLP catalyst. Because boron is particularly oxophilic, even moderately strong LBs will cause this reaction to be irreversible when using borane LAs. This severely limits the scope of the ‘OH-tolerant’ reactions that had been developed, as neither the substrates nor products can contain any appreciably basic functional groups (Fig. 15). The use of very weak LBs also leads to the formation of very + powerful [LB·H] Brønsted acids after H2 cleavage, which means that highly acid-sensitive functional groups also have to be avoided. moisture, despite the presence of fairly strong Brønsted In order to overcome these limitations and develop bases. protocols that include and tolerate stronger Lewis bases the following further key factors were identified:  To minimise the problem of unproductive RO—H activation, Conclusions and future outlook LAs should be incorporated that bind more weakly to –OR moieties. However, in order to maintain the ability of the Despite the enormous advances that have been made in field of FLP catalysis in the decade since it was established, there FLP to activate H2, the HA of the LA must be preserved. In other words, a LA is needed that is relatively less oxophilic, remains significant scope for further development. It is our but still hydridophilic. hope that the principles outlined in this review may provide a  The presence of stronger LBs will prevent protonation of useful framework for this ongoing work. much more weakly basic substrates such as aldehydes and While the systems reported to date have provided a ketones. Hydrogenation of such compounds is therefore dramatic proof-of-principle for TM-free catalytic likely to require an atypical reaction mechanism, such as hydrogenation, it would be hard to argue that any of these activation of the substrate by the LA rather than H+ (Fig. reactions yet constitutes a truly attractive synthetic tool. If 4b). FLPs are eventually to take their place alongside TM complexes Based on this analysis, subsequent work was directed towards as practical hydrogenation catalysts then future work must increasingly be focused on ensuring broad functional group investigating the use of previously-unexplored R3SnOTf LAs (R + tolerance, ‘open bench’ stability and well-defined = alkyl; these are surrogates for [R3Sn] cations, which are chemoselectivity, alongside optimised reactivity. In this vein, it valence isoelectronic with BAr3), in combination with typical nitrogen LBs. Inspection of the literature indicated that these has been encouraging to see recent reports focusing on the development of catalytic protocols that can achieve high acids should be significantly less oxophilic than boranes of 45 35 20 enantioselectivity, that use very low catalyst loadings, or similar HA (such as B(C6F5)3), as indicated by the aqueous that catalyse reactions for which there is no existing TM- pKa values of the respective LA·OH2 adducts (< 0 for B(C6F5)3 46 + catalysed alternative. versus ~6 for [Bu3Sn] ; note that this is a specific example where it is important to consider a measure of LA strength Based on our own experiences with Sn-based systems, we would suggest that achieving the above goals may require a other than just HA). In addition, R3SnOTf-catalysed addition of willingness to investigate a broader range of LAs and LBs than R3SnH (which is generated following H2 activation by R3SnOTf- 47 based FLPs) to aldehydes and ketones is well established, have currently been investigated. In particular, s-block, late which suggested that the necessary atypical reaction p-block, and cheap d-block LAs all remain relatively unexplored mechanism should be feasible (simple mechanistic for applications in catalytic FLP chemistry. Detailed investigations were again performed to confirm this). experimental investigations into the kinetics and mechanisms of most FLP-catalysed hydrogenation reactions are also still Gratifyingly, an optimised FLP consisting of iPr3SnOTf and 2,4,6-collidine was found to be capable not only of activating needed, to provide the theoretical basis for rational future development. Finally, it may be possible that FLP H2 activation H2, but also of both hydrogenating aldehydes and ketones and performing hydrogenation catalysis in the presence of could find application in areas other than chemical hydrogenation catalysis. Plausible examples include reversible

Please do not adjust margins Please do not adjust margins

ARTICLE Journal Name hydrogen storage using very low molecular-weight FLPs, and 22 See E. L. Kolychev, T. Bannenberg, M. Freytag, C. G. Daniliuc, P. G. Jones and M. Tamm, Chem. Eur. J., 2012, 18, 16938, electrocatalytic H2 oxidation reactions for use in hydrogen fuel 48 and references therein. cell applications. 23 See E. Buncel and B. Menon, J. Am. Chem. Soc., 1977, 99, 4457, and references therein. 24 See L. Greb, S. Tussing, B. Schirmer, P. Ona-Burgos, K. Acknowledgements Kaupmees, M. Lokov, I. Leito, S. Grimme and J. Paradies, Chem. Sci., 2013, 4, 2788, and references therein. We would like to thank GreenCatEng, Eli Lilly (Pharmacat 25 T. Mahdi, Z. M. Heiden, S. Grimme and D. W. Stephan, J. Am. consortium) and the EPSRC for providing funding (D. J. S.) and Chem. Soc., 2012, 134, 4088. the Royal Society for a University Research Fellowship (A. E. 26 See S. Tussing, K. Kaupmees and J. Paradies, Chem. Eur. J., A.). 2016, 22, 7422, and references therein. 27 See T. Soós, Pure Appl. Chem., 2011, 83, 667, and references therein. 28 T. A. Rokob, A. Hamza, A. Stirling and I. Pápai, J. Am. Chem. Notes and references Soc., 2009, 131, 2029. 1 For an excellent summary of this area, see P. P. Power, 29 T. J. Herrington, B. J. Ward, L. R. Doyle, J. McDermott, A. J. P. Nature, 2010, 464, 171. See also cited and citing references. White, P. A. Hunt and A. E. Ashley, Chem. Commun., 2014, 2 G. D. Frey, V. Lavallo, B. Donnadieu, W. W. Schoeller and G. 50, 12753. Bertrand, Science, 2007, 316, 439. 30 See, for example, Y. Wang, W. Chen, Z. Lu, Z. H. Li and H. 3 G. C. Welch, R. R. S. Juan, J. D. Masuda and D. W. Stephan, Wang, Angew. Chem. Int. Ed., 2013, 52, 7496. Science, 2006, 314, 1124. 31 P. Spies, S. Schwendemann, S. Lange, G. Kehr, R. Fröhlich and 4 G. C. Welch and D. W. Stephan, J. Am. Chem. Soc., 2007, 129, G. Erker, Angew. Chem. Int. Ed., 2008, 47, 7543. 1880. 32 See T. Özgün, K.-Y. Ye, C. G. Daniliuc, B. Wibbeling, L. Liu, S. 5 G. C. Welch, L. Cabrera, P. A. Chase, E. Hollink, J. D. Masuda, Grimme, G. Kehr and G, Erker, Chem. Eur. J., 2016, 22, 5988, P. Wei and D. W. Stephan, Dalton Trans., 2007, 3407. and references therein. 6 The field of FLP chemistry has seen a number of broad 33 Z. Mo, E. L. Kolychev, A. Rit, J. Campos, H. Niu and S. summaries in recent years. See, for example, D. W. Stephan, Aldridge, J. Am. Chem. Soc., 2015, 137, 12227. Science, 2016, 354, 1248, and references therein. 34 Y. Guo and S. Li, Inorg. Chem., 2008, 47, 6212. 7 An alternative, electric-field-based model was also proposed, 35 J. M. Farrell, R. T. Posaratnanathan and D. W. Stephan, but has largely fallen out of favour and is not regularly Chem. Sci., 2015, 6, 2010. invoked. See, for example, T. A. Rokob, I. Bakó, A. Stirling, A. 36 P. Eisenberger, B. P. Bestvater, E. C. Keske and C. M. Hamza and I. Pápai, J. Am. Chem. Soc., 2013, 135, 4425. Crudden, Angew. Chem. Int. Ed., 2015, 54, 2467. 8 P. A. Chase and D. W. Stephan, Angew. Chem. Int. Ed., 2008, 37 See, for example, D. W. Stephan, S. Greenberg, T. W. 47, 7433. Graham, P. Chase, J. J. Hastie, S. J. Geier, J. M. Farrell, C. C. 9 See W. E. Piers, A. J. V. Marwitz and L. G. Mercier, Inorg. Brown, Z. M. Heiden, G. C. Welch and M. Ullrich, Inorg. Chem., 2011, 50, 12252, and references therein. Chem., 2011, 50, 12338, and references therein. 10 J. R. Khusnutdinova and D. Milstein, Angew. Chem. Int. Ed., 38 See L. E. Longobardi, C. Tang and D. W. Stephan, Dalton 2015, 54, 12236. Trans., 2014, 43, 15723, and references therein. 11 K. K. Ghuman, L. B. Hoch, T. E. Wood, C. Mims, C. V. Singh 39 C. Bergquist, B. M. Bridgewater, C. J. Harlan, J. R. Norton, R. and G. A. Ozin, ACS Catal., 2016, 6, 5764. A. Friesner and . Parkin, J. Am. Chem. Soc., 2000, 122, 10581. 12 See A. Berkessel, T. J. S. Schubert and T. N. Müller, J. Am. 40 In parallel with our own work described in this case study, Chem. Soc., 2002, 124, 8693, and references therein. very similar boron-based systems that address the same 13 For a recent review, see R. H. Morris, Chem. Rev., 2016, 116, issues were independently developed by the groups of 8588. Stephan and Soós, and published near-simultaneously. See T. 14 T. A. Rokob, A. Hamza, A. Stirling, T. Soós and I. Pápai, Mahdi and D. W. Stephan, J. Am. Chem. Soc., 2014, 136, Angew. Chem. Int. Ed., 2008, 47, 2435. 15809; and Á. Gyömöre, M. Bakos, T. Földes, I. Pápai, A. 15 L. Rocchigiani, G. Ciancaleoni, C. Zuccaccia and A. Macchioni, Domján and T. Soós, ACS Catal., 2015, 5, 5366. J. Am. Chem. Soc., 2014, 136, 112. 41 See D. J. Scott, M. J. Fuchter and A. E. Ashley, Angew. Chem. 16 T. A. Rokob, A. Hamza and I. Pápai, J. Am. Chem. Soc., 2009, Int. Ed., 2014, 53, 10218, and references therein. 131, 10701. 42 D. J. Scott, M. J. Fuchter and A. E. Ashley, J. Am. Chem. Soc., 17 See, for example: F. G. Bordwell, Acc. Chem. Res., 1988, 21, 2014, 136, 15813. 456 and cited and citing references. Note that where pKa 43 D. J. Scott, T. R. Simmons, E. J. Lawrence, G. G. Wildgoose, M. values are given in this review, they have typically been J. Fuchter and A. E. Ashley, ACS Catal., 2015, 5, 5540. taken from the related cited work (or references therein). 44 D. J. Scott, N. A. Phillips, J. S. Sapsford, A. C. Deacy, M. J. 18 M. A. Beckett, G. C. Strickland, J. R. Holland and K. Sukumar Fuchter and A. E. Ashley, Angew. Chem. Int. Ed., 2016, 55, Varma, Polymer, 2996, 37, 4629. 14738. 19 See, for example: A. E. Ashley, T. J. Herrington, G. G. 45 For a recent review of enantioselective FLP-catalysed Wildgoose, H. Zaher, A. L. Thompson, N. H. Rees, T. Krämer hydrogenation, see J. Paradies, Chiral Borane-Based Lewis and D. O’Hare, J. Am. Chem. Soc., 2011, 133, 14727; or É. Acids for Metal Free Hydrogenations, Top. Curr. Chem., Dorkó, B. Kótai, T. Földes, Á. Gyömöre, I. Pápai and T. Soós, J. Springer GmbH, Berlin, 2017, pp 1-24. Organomet. Chem., 2017, doi: 46 S. Wei and H. Du, J. Am. Chem. Soc., 2014, 136, 12261. 10.1016/j.jorganchem.2017.04.031. 47 For a review of less standard main-group LAs in FLP 20 Z. M. Heiden and A. P. Latham, Organometallics, 2015, 34, chemistry, see S. A. Weicker and D. W. Stephan, Bull. Chem. 1818. Soc. Jpn., 2015, 88, 1003. 21 L. J. Hounjet, C. Bannwarth, C. N. Garon, C. B. Caputo, S. 48 See, for example, E. J. Lawrence, V. S. Oganesyan, D. L. Grimme and D. W. Stephan, Angew. Chem. Int. Ed., 2013, 52, Hughes, A. E. Ashley and G. G. Wildgoose, J. Am. Chem. Soc., 7492. 2014, 136, 6031.

Please do not adjust margins