Solid state frustrated Lewis pair chemistry

In solution the PCy3/B(C6F5)3 pair is rapidly deactivated by nucleophilic aromatic substitution. In the solid state (or in a fluorous liquid), however, it serves as an active frustrated Lewis pair to effectively split dihydrogen.


Introduction
Lewis acids and bases when brought together in solution typically undergo rapid formation of strong adducts. Similar to the neutralization reaction of Brønsted acids and bases this leads to an annihilation of the typical Lewis acid and Lewis base properties. 1,2 This situation can be changed if one effectively hinders the Lewis pair from the neutralizing adduct formation, e.g. by electronic means 3 or, more commonly, by attaching very bulky substituents at the core atoms of the pair. [4][5][6] This invariably leads to situations where active Lewis acids and active Lewis bases are present in a solution at the same time, opening possibilities for cooperative reactions with added substrates. Such "frustrated Lewis pairs" (FLPs) 7 have very successfully been used within the last decade for small molecule binding and activation, most notably among this the metal-free splitting and activation of dihydrogen. 8,9 A number of new reaction types have been found in this way 10,11 and a great variety of different FLPs have been devised, characterized and their reactions reported. [12][13][14][15][16][17][18][19][20][21][22][23] The vast majority of them relies on avoiding the effective neutralizing Lewis acid/Lewis base adduct formation by steric hindrance, and this has led to the discovery and development of a great variety of interesting reactions in solution. [24][25][26] We thought that we should not conne ourselves to searching for FLP reactions in the liquid phase. One may safely assume that large molecules in the solid state are more or less rigidly conned to their positions in the crystal lattice. Therefore, Lewis acids and Lewis bases should effectively be hindered from adduct formation or other deactivating reaction pathways even in the absence of efficient steric hindrance by their substituents as long as we keep them in the solid state. This may actually set the scene for possibly nding new frustrated Lewis pairs and, consequently, new FLP reactions, by exposing such Lewis pairs to suitable reagents in the solid state. We have tried this principle and found that FLP chemistry can be done in this way.
There had been some reports about frustrated Lewis pair related behavior at certain heterogeneous catalysts. In these cases the active sites were part of the catalytic solid. Activation of small molecules at such systems had been achieved either thermally or by photolysis. [27][28][29] There have also been a few reports about heterogenized Lewis acids, bonded to suitable supports, that have been employed in FLP type reactions. 30 Our here reported case is distinctly different: we have employed solid physical mixtures of phosphane Lewis bases with the strong B(C 6 F 5 ) 3 boron Lewis acid and reacted them under suitable conditions with selected small molecules. Activation occurred and the phosphane/borane pair became consumed with selective formation of the FLP reaction products. First examples of this FLP development will be described below in this account.

Results and discussion
Tricyclohexylphosphane (1a) and tris(pentauorophenyl)borane (2) represent a Lewis base/Lewis acid pair that is known to react rapidly in solution. It undergoes a nucleophilic aromatic substitution by the phosphane at the para-position of a C 6 F 5 group of the borane 4,31 to eventually yield the reaction product 3a. 7,32 This favorable S N Ar reaction eliminates any FLP reactivity of the 1a/2 pair very effectively in solution. Even in the presence of 50 bar H 2 no signicant hydrogen splitting is observed, the formation of 3a even prevails under these conditions. We isolated this compound in 86% yield and conrmed its formation by X-ray diffraction and by NMR spectroscopy [Scheme 1a (H 2 ); for details see the ESI †]. 32 The situation is drastically different in the solid state: we mixed equimolar quantities of PCy 3 (1a) and B(C 6 F 5 ) 3 (2) and exposed it in a glass vial inside a steel autoclave to 50 bar of dihydrogen for a total of 10 days with constant agitation by a Teon coated magnetic stirring bar. Aer this time a sample was taken and dissolved in D 2  Crystallization from dichloromethane/pentane gave single crystals which were used to conrm the formation of the FLP H 2 splitting product under these special conditions by X-ray diffraction (for details see the ESI †). 33 The salt 4a is an active reducing agent. Its reaction with the bulky N-phenyl-4-methylacetophenonimine gave the respective secondary amine reduction product (24 h at 70 C, 86% conversion, for details see the ESI †).
We carried out an ample characterization of the solid product material directly (i.e. without ever dissolving it) by solid state NMR spectroscopy. Evidence for the solid state hydrogenation of the PCy 3 /B(C 6 F 5 ) 3 mixture comes from the 31 P and 11 B MAS NMR data ( Fig. 1). In the 31 P NMR spectrum we recognize the starting material at 7.1 ppm, whereas the phosphonium ion gives a broad signal at 30.1 ppm. In the 11 B MAS NMR spectrum the signal of the free B(C 6 F 5 ) 3 gives rise to the previously documented second-order quadrupolar lineshape 34 whereas aer the hydrogenation a much narrower signal appears at the isotropic chemical shi of À24.9 ppm aer applying the correction for the second-order quadrupolar shi (see ESI †). In addition, the spectrum reveals the presence of a minor amount of the substitution product at À2.4 ppm. The latter is the only product formed when the reaction is carried out in solution. The complete NMR characterization of the substitution product both in solution and the solid state is given in the ESI. † In a control experiment, the formation of only a minor amount of the substitution product 3a was observed in the solid state NMR spectra of a PCy 3 /B(C 6 F 5 ) 3 mixture subjected to identical reaction conditions in the absence of H 2 .   signals from the MAS probe and the MAS rotor cap. 35 Under these conditions, the strong 1 H-1 H dipolar couplings are sufficiently suppressed, even though residual line broadening is still detected owing to higher order terms in the Hamiltonian which are not fully eliminated even at 60.0 kHz. 36 A distinct doublet (J( 1 H-31 P) $ 430 Hz) can be identied at 5.5 ppm which is assigned to P-bound hydrogen, whereas the singlet at 4.0 ppm is assigned to the B-bound hydrogen (the expected multiplet is not resolved in this case). This assignment is supported by 11 37 In the present material, no dephasing was observed over a dipolar mixing time of $5 ms. Comparing these experimental data with corresponding two-spin simulations we can conclude that the boron-phosphorus distance must be larger than 600 pm. Thus, all the experimental data are consistent with well-separated phosphonium and borate ions. The reaction of the phosphane PPhCy 2 (1b) with the Lewis acid B(C 6 F 5 ) 3 (2) in the solid state proceeds similarly. The reaction was carried out analogously as the one described above (r.t., 50 bar H 2 , 3 days). Our analysis of a product sample dissolved in CD 2 Cl 2 revealed almost exclusive formation of the dihydrogen FLP splitting product 4b [Scheme 1b (H 2 )]. It shows a characteristic 1  Keeping the PPhCy 2 /B(C 6 F 5 ) 3 mixture (1b/2) in CD 2 Cl 2 solution for 12 hours under H 2 (50 bar) gave a different result. The NMR analysis showed the formation of a ca. 1 : 6 mixture of the salt 4b with the substitution product 3b. [Scheme 1a (H 2 )]. The zwitterionic phosphonium/uoroborate product 3b was isolated from a separate experiment as a white solid in 82% yield. It shows a 11 B NMR doublet ( 1 J BF $70 Hz) at d À0.6 and a sharp 31 P NMR signal at d 33.6. The 19 F NMR spectrum shows the [B]F resonance at d À192.9, two signals of the bridging C 6 F 4 group and three resonances (o, p, m) of the remaining B(C 6 F 5 ) 2 unit with a Dd 19 F m,p shi difference of 4.9 ppm, which is typical for a borate type structure (for details and the depicted NMR spectra see the ESI †).
In this case the nucleophilic aromatic substitution by the markedly less nucleophilic phosphane PPhCy 2 (1b) relative to PCy 3 (1a) seems to allow the FLP reaction to compete as a minor pathway in solution. This becomes continued in the reaction of the much less nucleophilic phosphane PPh 2 ( t Bu) (1c) which in solution together with the B(C 6 F 5 ) 3 Lewis acid gave a ratio of the S N Ar and FLP products of 3c : 4c isolated from the workup procedure involving recrystallization from CH 2 Cl 2 /pentane in 60% yield (details of the characterization of the compounds 3c and 4c see the ESI †).
Quantum chemical simulations were used to investigate the mechanistic details of the different FLP reactivities in all three states of matter, i.e. gas, liquid, and solid phase. The FLP/H 2 thermochemistry is for the rst time investigated here in the solid state using relatively high-level periodic quantum chemistry methods. We employed a hierarchy of theoretical methods, ranging from semi-empirical tight-binding Hamiltonians to accurate London dispersion corrected hybrid density functionals. [38][39][40][41][42][43][44][45] More discussion of methodological points and computational details can be found in our previous benchmark study 46 and in the ESI. † The main representative results of the PCy 3 /B(C 6 F 5 ) 3 FLP are shown in Fig. 4.
All shown reactions are exergonic, and both the solvent and the crystal phase stabilize the products 4a and 3a by 10 to 15 kcal mol À1 compared to the gas phase. In this regard the FLP reaction to 4a is in fact thermodynamically feasible. However, the competing product 3a is in solution signicantly preferred over 4a (À50.7 kcal mol À1 vs. À15.5 kcal mol À1 ) and due to the expected high mobility it can readily react and prevent the desired FLP reaction. In contrast, the crystal eld provides much more pronounced energy barriers, which kinetically stabilizes the reactant and enables the targeted reaction to product 4a. These higher energy barriers can be rationalized by a simple geometric comparison (inset of Fig. 4). Apparently, the solid state reaction to 4a requires substantially less rearrangements of the crystal compared to 3a. Thus, we can identify two key roles that drive the reaction in the solid phase: (1) the crystal environment can adopt the solvent role in enhancing the FLP reactivity, typically explained by both the electrostatic screening and undirectional London dispersion stabilization of the FLP products. (2) The static crystal eld can selectively suppress certain undesired reaction routes, which is not possible in a liquid or gas environment with high molecular mobility.
While point (1) makes the heterogeneous formulation of typical FLP reactions possible, (2) goes beyond it and opens possibilities for new FLP systems as compellingly demonstrated for the here discussed compound. An additional important prerequisite for the discussed reaction is the possible diffusion of H 2 gas through the reactant crystal. Our molecular dynamics (MD) simulation (for details see the ESI †) conrm that H 2 can actually move more or less freely through the channels of the crystal thereby generating the correct conditions for the H 2 activation to take place. Moreover, we have conducted MD simulations for a model of the interface between the Lewis acid and base as it may occur experimentally in a mixture of the solid particles. According to these results which are shown in Fig. 5, at the interface the components (in particular the Lewis base) are spatially not constrained, possibly due to a mismatch of the molecular surfaces (a kind of interface strain). This leads partially to a "liquid phase" behavior of the FLPs in the solid state. It is seen that the Lewis acid and base components could move rather freely at the contact surface and adopt molecular FLP conformations enabling hydrogen activation as in solution (for details see the ESI †).
The new solid state phosphane/borane FLP reactions are not limited to the splitting of dihydrogen. We exposed the 1 : 1 mixture of PCy 3 (1a) and B(C 6 F 5 ) 3 (2) for 4 hours at r.t. in the solid state to SO 2 gas (1.5 bar). 47 A sample was dissolved in CD 2 Cl 2 and its NMR spectra revealed the almost quantitative formation of the P/B FLP SO 2 addition product 5. [Scheme 1b (SO 2 )] the product was also directly identied from the solid obtained by solid state NMR spectroscopy (see the ESI †), indicating essentially quantitative conversion. The distorted fourcoordinate boron environment in 5 is characterized by d iso ¼ À0.6 ppm, and a nearly axially symmetric electric eld gradient, with C Q ¼ 1.54 MHz and h Q ¼ 0.15. The 31 P MAS-NMR spectrum shows a single sharp signal at 51.5 ppm. In this case, 11 B{ 31 P} REDOR and 31 P{ 11 B} REAPDOR experiments consistently point towards a B/P internuclear distance of 450 pm, which is in good agreement with the distance of 434 pm from the crystal structure.
We performed the reaction on a preparative scale and isolated the product 5 as a white solid aer recrystallization from CH 2 Cl 2 /pentane in 84% yield. The product shows the typical 11 B (d À0.3) and 19 F NMR features (three resonances, Dd 19 F m,p ¼ 6.4 ppm) of the borate section of the molecule and a phosphonium 31 P NMR signal at d 50.0 (for further details see the ESI †).
Compound 5 was characterized by X-ray diffraction (Fig. 6). The X-ray crystal structure analysis shows the newly formed P1-S1 and O2-B1 bonds. Both the phosphorus and the boron atoms show pseudo-tetrahedral coordination geometries. The sulfur coordination geometry is distorted trigonal-pyramidal. We also exposed the PCy 3 /B(C 6 F 5 ) 3 pair to SO 2 in solution but   only observed the formation of the substitution product 3a [Scheme 1a (SO 2 ); see the ESI † for details].
Our study has shown so far that an agitated mixture of particles of the phosphanes 1a-c with particles of the B(C 6 F 5 ) 3 Lewis acid 2 did very effectively evade the deactivating S N Ar reaction that these pairs rapidly undergo in solution. Instead, they retained their frustrated Lewis pair character and, consequently, showed the ability to split dihydrogen heterolytically. While this observation is probably of a far-reaching principal interest, the rather harsh conditions of the solid state FLP H 2splitting (50 bars of dihydrogen, 3 to 10 d reaction time) made this far from a conveniently applicable procedure.
Fluorous liquids show some extraordinary properties. 48-55 They do not mix with a variety of common organic solvents; they show an enhanced solubility of many gases in them, among them dihydrogen. 48,[55][56][57] Moreover, many organic and element-organic compounds, among them the phosphanes 1a-c and B(C 6 F 5 ) 3 (2) are insoluble in them. Therefore, we decided to carry out the solid state FLP dihydrogen splitting reaction in per-uoromethylcyclohexane (F 11 C 6 -CF 3 ). In a typical experiment (see the ESI † for details) we suspended an equimolar mixture of PCy 3 (1a) and B(C 6 F 5 ) 3 in peruoromethylcyclohexane and stirred the suspension for 10 h in a dihydrogen atmosphere at near to ambient conditions (r.t., 1.5 bar H 2 ). Workup was simply done by evaporation of the volatiles. A sample of the obtained white powdery solid was then subjected to NMR analysis in D 2dichloromethane solution. It showed that a ca. 60% conversion to the hydrogen splitting product HPCy 3 + /HB(C 6 F 5 ) 3 À (4a) had been achieved. The remaining starting material had become converted to the S N Ar reaction product 3a under the conditions of the NMR analysis in solution. The solid state NMR spectra of the products obtained aer the suspension reaction showed the formation of 4a. The reaction of the PPhCy 2 (1b)/B(C 6 F 5 ) 3 pair with dihydrogen proceeded at least equally well in this uorous liquid. Under analogous conditions a ca. 95% conversion to the H 2splitting product 4b was achieved within the 10 h reaction time. The PPh 2 t Bu (1c)/B(C 6 F 5 ) 3 system even slightly surpassed this result. We obtained a near to quantitative conversion to the HPPh 2 t Bu + /HB(C 6 F 5 ) 3 À salt within 10 h at near to ambient conditions in the inert peruoromethylcyclohexane liquid (Table 1, see the ESI † for details).

Conclusions
Our study has shown that frustrated Lewis pair behavior can be achieved by other means than the usual electronic 3,19 or steric modications 4-9 of Lewis acids and bases. In the cases reported here we have deviated from the ubiquitous method of using sterically bulky substituents at the phosphorus Lewis base in order to prevent neutralizing adduct formation or other deactivating reactions with the strong B(C 6 F 5 ) 3 boron Lewis acid. On the contrary, the phosphanes used in this study are rather nucleophilic and in solution they undergo rapid nucleophilic aromatic substitution at one C 6 F 5 ring of the Lewis acid replacing uoride which consequently then becomes bonded to the boron atom 31 with annihilation of its Lewis acidic features. In all the cases looked at in our study this FLP deactivating reaction is very efficiently suppressed by localizing the individual Lewis acidic and Lewis basic molecules inside their solid state lattices. This prevents them effectively from undergoing the bimolecular deactivation reaction. From the solid state NMR and the DFT analysis in conjunction with general principles about molecular diffusion in the solid state, we assume that the individual Lewis acid and base components do not easily mix on a molecular level in our experiments, but that initially we are dealing with separate solid state phosphane and borane particles. This lets us assume that the FLP dihydrogen splitting reaction must take place at the surface, respectively the interface between phosphane and borane solid state entities; the ensuing reaction is, however, probably facilitated by the easy permeability of the respective crystal lattices by dihydrogen (and other gases). The dihydrogen splitting reaction may then have formed the phosphonium/ hydridoborate salt initially at the surface, but it may be assumed that accumulation of that species creates a local situation resembling an ionic liquid, which might facilitate diffusion and mixing since eventually we obtained homogeneous solid samples of the respective dihydrogen splitting products. The MD simulations which support this view are currently due to the large system size too approximate (short) to draw any quantitative conclusions nor are we able to simulate further parts of the solid state reaction dynamically. Nevertheless, the MD and static theoretical results clearly support the above described picture of partially "molten" material at the interface with sufficient molecular exibility for activation of almost freely diffusing molecular hydrogen.
Although the mechanistic aspects of our solid state FLP reactions must remain somewhat speculative at this time, we have greatly improved its practical applicability by using the uorous liquid effect. [55][56][57] This has made dihydrogen splitting reactions readily available from frustrated Lewis pair combinations which cannot be kept active in other ways. 33 We have further found rst indications that the resulting [P]H + /[B]H À

Conflicts of interest
There are no conicts to declare.