Water Reduction and Dihydrogen Addition in Aqueous Conditions With ansa‐Phosphinoborane

Abstract Ortho‐phenylene‐bridged phosphinoborane (2,6‐Cl2Ph)2B‐C6H4‐PCy2 1 was synthesized in three steps from commercially available starting materials. 1 reacts with H2 or H2O under mild conditions to form corresponding zwitterionic phosphonium borates 1‐H2 or 1‐H2O. NMR studies revealed both reactions to be remarkably reversible. Thus, when exposed to H2, 1‐H2O partially converts to 1‐H2 even in the presence of multiple equivalents of water in the solution. The addition of parahydrogen to 1 leads to nuclear spin hyperpolarization both in dry and hydrous solvents, confirming the dissociation of 1‐H2O to free 1. These observations were supported by computational studies indicating that the formation of 1‐H2 and 1‐H2O from 1 are thermodynamically favored. Unexpectedly, 1‐H2O can release molecular hydrogen to form phosphine oxide 1‐O. Kinetic, mechanistic, and computational (DFT) studies were used to elucidate the unique “umpolung” water reduction mechanism.


Introduction
A cooperative reactivity of frustrated Lewis pairs (FLPs) opens up outstanding potential for small molecules activation. FLPs can split CÀ O, SÀ O, NÀ O, NÀ N, and CÀ H bonds, [1] but most notable is the activation and the catalytic transfer of molecular hydrogen to various organic substrates. [2] The FLP reactive centers can be represented by various atoms, including metals, [3] however, the majority of reported FLPs are based on highly Lewis acidic (LA) boranes and bulky phosphines or amines as bases (LB). Intrinsically high oxophilicity of boron leads to functional groups sensitivity, particularly, to OHcontaining molecules especially water, [4,12] limiting the utility of FLPs as catalysts.
As part of our general interest in extending the boundaries of the FLP chemistry, we have been focused on preparing FLPs that can tolerate water. Several successful attempts to develop FLP hydrogenation catalysts that could tolerate moisture were reported. [5,6] Our interest, however, lies in the preparation of stable, i. e., directly detectable, hydrogen adducts existing in the presence of over-stoichiometric amounts of water, ideally in aqueous solutions. Apart from purely fundamental significance, such FLPs might find practical applications as parahydrogeninduced polarization (PHIP) tags in biologically relevant media. Parahydrogen is an accessible source for creating nuclear spin hyperpolarization and we have shown previously that adducts of intramolecular FLPs and parahydrogen produce PHIP with orders of magnitude NMR signal enhancement. [7] Our previous works demonstrated that intramolecular FLPs bridged by o-phenylene (ansa-scaffold) exhibited enhanced reactivity in comparison to intermolecular FLPs. [8] In addition, fixed proximity of LA and LB sites in the ansa-FLPs offers flexible possibilities for their modification: ansa-FLP that was built even with the smallest LA group, -BH 2 , exhibited hydrogen splitting reactivity. [9] Herein we report the design of a highly sterically hindered activation pocket that can disfavor water binding by steric exclusion.
Previously we attempted preparation of water tolerant FLPs but they either strongly favored the formation of water adducts (Figure 1, Ia-c) [7b,c] or were reactive to neither water nor hydrogen (Figure 1, II). [10] In continuation of these efforts we designed new ansa-phosphinoborane 2-[(2,6-Cl 2 Ph) 2 B]-C 6 H 4 -PCy 2 1. Replacement of the CH 3 groups with Cl was expected to enhance the acidity of the boryl site while retaining necessary steric effect. This opens unique capability for the addition of H 2 and H 2 O, both in a reversible manner, and rather unexpected reduction of H 2 O to H 2 .

Synthesis of 1, splitting of H 2 and H 2 O
We developed a three-step synthesis of ansa-phosphinoborane 1 from commercially available 1,3-dichlorobenzene, 2-bromophenyl(dicyclohexyl)phosphine, and BCl 3 (Scheme 1). Recrystallization of 1 from a hexane-toluene mixture at À 20°C gave the pure product as yellow crystals with 48 % overall yield. 11 B and 31 P NMR displayed singlet signals at 67.03 and 0.51 ppm, respectively, revealing no dative PÀ B coordination in solution. [11] The structure determined from the X-ray diffraction analysis featured P···B separation 3.170(4) Å, thus excluding a dative PÀ B bond in solid 1 (Figure 2).
Having 1 in hands, we explored its reactivity towards H 2 and H 2 O (Scheme 2). Exposing solutions of 1 in CD 2 Cl 2 or C 6 D 6 to 10 bars of H 2 at room temperature led to the formation of zwitterionic adduct 1-H 2 identified by the characteristic PÀ H and BÀ H 1 H NMR signals (see Table 1). [12] Exposing the solution of 1-H 2 in C 6 D 6 to D 2 led to H/D exchange and isotopic scrambling, revealing that hydrogen activation is reversible (75 % conversion of the H 2 /D 2 mixture to HD after being kept at room temperature for 12 h under 5 bar pressure, see Supporting Information for details).
Upon release of H 2 pressure and exposure of the NMR sample to air, further monitoring by the NMR spectroscopy showed rapid conversion of 1-H 2 to water adduct 1-H 2 O. Alternatively, 1-H 2 O can be prepared directly from 1 and water. 1 H NMR spectroscopy exhibited two distinct signals corresponding to the PÀ H and BOÀ H groups (See Table 1). Recrystallization of 1-H 2 O followed by X-ray diffraction analysis revealed exoconfiguration of 1-H 2 O wherein the B-OH group is directed to the phosphorus atom while the PÀ H proton is directed outside of the FLP pocket ( Figure 3). This geometry agrees well with our computational studies (see below) and previous in silico studies of similar compounds derived from II. [10] In contrast, previously reported HX (X=OH, OR, F, Cl) adducts of ansa-FLPs adopted endo-configuration featuring intramolecular hydrogen bonding, which was thought to substantially stabilize these species. [13] Such endo-adducts feature slightly longer BÀ O bonds (1.52-1.53 Å) [13a,g] as compared to the experimentally observed exoform of 1-H 2 O (1.463(7) Å).
Quantum mechanical DFT calculations of water and hydrogen activation were in good agreement with experimental observations (Figure 4, see Supporting Information for details). Both processes were found to be thermodynamically favorable and kinetically feasible. Exo-adduct 1-H 2 O-exo was found to be by 1.4 kcal/mol more stable than 1-H 2 O-endo, whereas for 1-H 2 the endo-form was more stable. [12] In line with the much higher propensity of polar OÀ H bonds for the heterolytic splitting, the  addition of water to 1 is hindered by a very low, 7.5 kcal/mol, kinetic barrier whereas the hydrogen addition transition state 1-H 2 -TS lies much higher at 20.0 kcal/mol. Interestingly, computations predict hydrogen addition to be more preferred over the addition of water by almost 3 kcal/mol in DCM. It was also possible to computationally identify unstable classical Lewis adduct 1-H 2 O-LA.

Water reduction
When 1-H 2 O or 1 with various amounts of water in the 1 : 1 CD 2 Cl 2 :CD 3 CN solution were monitored by the NMR spectroscopy at 25°C or elevated temperatures, new signals were observed in 1 H NMR spectra along with additional singlets at 83.67 ppm in the 31 P and at 7.46 ppm in the 11 B NMR spectra.
The new species was isolated and, based on the results of the XRD analysis, identified as phosphinoborane oxide 1-O (Figure 5). Its structure features PÀ OÀ B fragment and can be described as an intramolecular Lewis adduct of the respective

Chemistry-A European Journal
Research Article doi.org/10.1002/chem.202201927 phosphine oxide. [14] Accompanied formation of hydrogen in this reaction was observed by 1 H NMR.
To demonstrate that reduction of water with 1 is stoichiometric and produces an equimolar amount of hydrogen, we performed hydrogenation of ethyl cinnamate in a commercial two-chamber reactor. [15] Placed in one of the chambers a mixture of 1 with 2 equivalents of H 2 O served as the source of H 2 . The other chamber contained equimolar to 1 amount of ethyl cinnamate in cyclopentyl methyl ether (CPME) with Pd/C as a catalyst. Nearly quantitative (94 %) reduction of ethyl cinnamate to ethyl 3-phenylpropanoate was observed after 24 h (Scheme 3).
Although formal oxidation of phosphines by water is a thermodynamically favourable process and is the main driving force for several synthetically important reactions such as Mitsunobu reaction [16] or reductive disulfide bond cleavage, [17] the direct reaction of water with phosphines under mild conditions is scarce. In this context, a bicyclic P(III) amidoester is remarkable for its ability to oxidatively add water and form P(V) derivative. [18] We note that while our article was in preparation, an alike deoxygenation of water with stoichiometric amount of ortho-phenylene linked bisborane-functionalized phosphine was reported, [14d] constituting the only other example of such chemistry to the best of our knowledge.
Considering other non-metals, two metal-free systems capable of H 2 O reduction to H 2 have been reported, namely intramolecular silylene-borane [19] and sp 3 -sp 3 diboron compounds. [20] Although the reports are lacking detailed mechanistic insights, heterolytic splitting of water is suggested as a key step in both cases.
We hypothesized that the bifunctional nature of 1 could trigger its reactivity with water. To verify the necessity of preorganized P/B sites for the observed unusual reactivity of 1, we probed the reaction with a combination of separated FLP components chemically and structurally comparable to our system, namely PCy 3 and B(2,6-Cl 2 C 6 H 3 ) 3 . Heating their equimolar mixture with 6 equivalents of H 2 O in a 1 : 1 CH 3 CN:CH 2 Cl 2 mixture at 70°C over night yielded no phosphine oxide (see Supporting Information for details).

Mechanistic studies of water reduction with 1
To elucidate mechanistic insights, we followed the kinetics of water reduction with 1 by 31 P NMR spectroscopy and supported it by DFT computations. Inverse gated proton decoupling pulse sequence ensured quantitative measurements whereas utilization of non-deuterated solvents prevented any isotopic exchange side effects. The samples were prepared in gas-tight NMR tubes by dissolving 1 in 1 : 1 CH 3 CN/CH 2 Cl 2 mixtures containing precise concentrations of water (Scheme 4). Kinetics of 1 reacting with variable H 2 O concentrations at 65°C obey second order in 1-H 2 O and reverse first order in H 2 O. However, similar experiments carried out at 25°C revealed first order in 1-H 2 O along with À 0.5 order in water (see Supporting Information for details).
In computational DFT studies we explored several monoand bimolecular (with respect to phosphinoborane) mechanisms ( Figure 6, see Supporting Information for details). In the interest of optimal utilization of computational resources, we modeled the water reduction mechanism using des-chloro compound 4. The possibility of unimolecular hydrogen release from 4-H 2 O via 4-centered transition state 4 c-TS can be ruled out due to the high kinetic barrier (42.5 kcal/mol, Figure S52). Guided by kinetic results, we considered alternative mechanisms in which free 1 facilitates water reduction. We found that the LA center of 4 can accept a water-derived hydrogen atom from the phosphine as a hydride. This "umpolung" of the protic PÀ H atom is accompanied by the migration of the OH group from the boron to the phosphorus atom inside the FLP pocket.  Figure S49). The respective alternative transition state was found to be higher in energy than the one starting from the exo-adduct by 15 kcal/mol (see Supporting Information for details).
The results of the DFT computations were consistent with the kinetic experiments carried out at 65°C. The most energetic transition state, the hydrogen atom transfer from 1-H 2 O to 1, should manifest in the first kinetic orders in each of these Scheme 3. In situ generation of H 2 via the stoichiometric reduction of H 2 O with 1 and its utilization as a reductant using a two-chamber reactor.
compounds. Provided free 1 exists in rapid equilibrium with 1-H 2 O and water, the expected kinetic orders 2 in 1-H 2 O and À 1 in H 2 O match the observed ones. The experimental data were fit to the above kinetic model via numeric kinetic simulations (see Supporting Information for details) with satisfactory accuracy and allowed for extracting of the kinetic parameters. We found the water reduction rate constant k = 0.183 mM À 1 h À 1 and the equilibrium constant for the water dissociation K = 2.86 mM or 3.9 kcal/mol, the latter was in good agreement with the value found by the DFT calculations. Notably, the reaction does not occur in pure CH 2 Cl 2 at ambient temperatures or upon heating. However, the addition of free Lewis acid (C 6 F 5 ) 3 B catalyzes the reaction at room temperature, which supports our mechanistic proposal. Water reduction by 1 is affected by a strong kinetic isotopic effect. Adduct of 1 with D 2 O, 1-D 2 O remains intact even upon prolonged heating in 1 : 1 CH 3 CN: CH 2 Cl 2 mixtures.
As we noted above, an alike reactivity, namely a stoichiometric reduction of water with phosphine giving molecular hydrogen and phosphine oxide, was reported recently for an ortho-phenylene linked bisborane-functionalized phosphine, while our manuscript was under preparation. [14d] A computational study presented in that study revealed conceptually the same mechanism as we proposed in our study: heterolityc splitting of water by phosphine and one Lewis acidic borane site followed by shuttling the borane-bound OH group to the phosphorus center and concurrent abstraction of the hydride from PÀ H group by the second Lewis acidic boron center. Thus, the reaction reported by Shang et al. [14d] represents an intramolecular version of the water reduction process providing a further support to our mechanistic proposal.

Hydrogen addition to 1 in aqueous mixtures
Since mechanistic studies indicated notable dissociation of 1-H 2 O, we further examined its reactivity in the 1 : 1 CD 2 Cl 2 :CD 3 CN mixture. In a gas-tight heavy wall NMR tube, the solution of 1 and 6 equivalents of H 2 O was exposed to 10 bar of H 2 . A control sample was prepared in the same way but omitting H 2 . After 1 h of heating at 80°C, both samples were analyzed by 1 H, 11 B, and 31 P NMR spectroscopy (Scheme 4). The 1 H NMR spectrum of the control sample (without H 2 ) featured a set of signals corresponding to 1-D 2 O and newly formed 1-O. 1 H NMR of the sample containing H 2 revealed the presence of partially deuterated 1-H 2 O, 1-O, but also 1-H 2 was detected at low concentrations by the appearance of characteristic signals corresponding to the BÀ H and PÀ H hydrogens. (Figure 7c, signals of 1-H 2 are indicated). A triplet at 4.47 ppm corresponding to HD revealed isotope scrabbling arising from facile proton transfer between phosphonium centers and the deuterated solvent. After another 2 h of heating, corresponding to 1-H 2 O and 1-H 2 1 H NMR signals of exchangeable BOÀ H, P-H, and BÀ H hydrogen atoms disappeared due to the complete deuteration. According to 1 H and 11 B NMR spectroscopies, the deuterated species 1-D 2 O, 1-D 2 , and 1-O were formed in the 3 : 1 : 1 ratio (Figure 7d, e).

Parahydrogen experiments
The above-described experiments revealed the formation of 1-H 2 in the presence of overstoichiometric amounts of water. Although 1-H 2 was not observed in the control experiment, the possibility of 1-H 2 to be an unstable intermediate of the

Chemistry-A European Journal
Research Article doi.org /10.1002/chem.202201927 background water reduction process during high pressure H 2 experiments could not be ruled out completely. To decouple the formation of 1-H 2 via direct addition of H 2 to 1 from the mediation of the water reduction mechanism, we studied the interaction of 1 with parahydrogen. Detection of hyperpolarized 1-H 2 in the presence of water would prove the direct H 2 activation mechanism because hyperpolarization is rapidly destroyed during secondary processes. We found that the interaction of ansa-phosphinoborane 1 with parahydrogen leads to nuclear spin hyperpolarization effects in both dry and aqueous solvents. In the first experiments, we bubbled parahydrogen through a 0.05 M solution of 1 in a dry CD 2 Cl 2 at 295 K under 3.2 bar pressure. The corresponding 1 H NMR spectra after the bubbling and after the relaxation to thermal equilibrium showed an enhanced signal for the PÀ H group with splitting equal to the corresponding spin-spin coupling constant (J PH = 478.0 Hz) (Figure 8a). The individual components of the doublet show opposite signs, revealing the created non-equilibrium nuclear spin order. The BÀ H group signal appearing as a 1 : 1 : 1 : 1 quartet under the thermal equilibrium due to 11 B nuclei (J BH = 79.5 Hz, spin 3/2) develops a slight distortion of the multiplet structure after the parahydrogen bubbling that also indicates the non-equilibrium state.
In the next experiments, 1 was dissolved in the aqueous solvent (3 equivalents H 2 O in 1 : 1 CH 3 CN:CH 2 Cl 2 ), letting the complete transformation of 1 into 1-H 2 O. Neither 1-H 2 nor hyperpolarization effects were observed while bubbling parahydrogen at 25°C through this solution. Heating to 80°C, however, unfroze the 1-H 2 O dissociation that enabled the formation of hyperpolarized 1-H 2 (Figure 8b), confirming the reversibility of interaction of 1 with both H 2 O and H 2 . As in the dry solvent, 1 H hyperpolarization was observed for the PÀ H and BÀ H hydrogens with some differing fine details such as the shapes of the BÀ H multiplet. In addition to 1 H, 31 P hyperpolarization was detected for 1-H 2 that manifested as an antiphase doublet in 31 P NMR (Figure 8b, right).
We note that these hyperpolarization effects differ from those commonly observed in high magnetic fields in PASADE-NA experiments. [7b,21] Normally, only 1 H NMR multiplets resulting from the homonuclear J-coupling between two parahydrogen nascent 1 H nuclei reveal the hyperpolarization. In contrast,  heteronuclear J-couplings ( 31 P-1 H and much less 11 B-1 H) reveal the hyperpolarization in our case. Practically, it means that 31 P (and slightly 11 B) nuclei are hyperpolarized in addition to 1 H. The mechanism of this effect must involve relaxation-driven transitions between different magnetization modes similarly as it was described for 1 H, 15 N, and 11 B hyperpolarization in ansaaminoboranes. [7c,d]

Conclusion
New ansa-phosphinoborane 1 features the ability for the reversible heterolytic splitting of H 2 and H 2 O. The zwitterionic water adduct 1-H 2 O can release H 2 through a multistep reaction pathway and form heterocyclic oxide 1-O. The hydrogen atom "umpolung" mechanism of the reaction was investigated experimentally and computationally, and dissociation of H 2 O from the adduct 1-H 2 O was shown to be vital for the observed reactivity. The nuclear spin hyperpolarization in parahydrogen experiments indicated that the addition of H 2 to ansaphosphinoborane 1 is pairwise, meaning that the H atoms do not lose each other in the H 2 activation process. It also confirmed the reversibility of H 2 O and H 2 additions, supporting the viability of dissociation steps in the proposed mechanism of the water reduction by the ansa-phosphinoborane. Moreover, this is the first time FLPs showing hyperpolarization effects in the presence of excess H 2 O, providing proof that metal-free activators for parahydrogen can be moisture tolerant.