Electro‐mediated PhotoRedox Catalysis for Selective C(sp3)–O Cleavages of Phosphinated Alcohols to Carbanions

Abstract We report a novel example of electro‐mediated photoredox catalysis (e‐PRC) in the reductive cleavage of C(sp3)−O bonds of phosphinated alcohols to alkyl carbanions. As well as deoxygenations, olefinations are reported which are E‐selective and can be made Z‐selective in a tandem reduction/photosensitization process where both steps are photoelectrochemically promoted. Spectroscopy, computation, and catalyst structural variations reveal that our new naphthalene monoimide‐type catalyst allows for an intimate dispersive precomplexation of its radical anion form with the phosphinate substrate, facilitating a reactivity‐determining C(sp3)−O cleavage. Surprisingly and in contrast to previously reported photoexcited radical anion chemistries, our conditions tolerate aryl chlorides/bromides and do not give rise to Birch‐type reductions.


Introduction
Synthetic methodologies involving single electron transfer (SET) are increasingly popular for the facile synthesis or modifications of important organic compounds.PhotoRedox Catalysis (PRC) [1] and Synthetic Organic Electrochemistry (SOE) [2] lead to easy SET processes,providing notable redox power for various organic transformations under mild conditions.G enerally,v isible-light PRC generates radical intermediates with good functional group tolerance in am ild manner.H owever,s ynthetic applications of PRC in terms of transformations needing highly oxidizing or reducing potentials are limited by the energetic limitations of visible light photons.O ne solution is to generate photoexcitable radical ions by multi-photon processes. [3] Such photoexcited radical ions are highly oxidizing [3a,b] or reducing species, [3c-h] leading to as ignificantly expanded redox "window" for activating inert substrates.S acrificial redox additives (e.g. DIPEA) are employed in stoichiometric excesses in consecutive Photoinduced Electron Tr ansfer (conPET) processes to prime catalysts prior to excitation. Their excesses and organic byproducts can plague purification steps.I nc ontrast, SOE allows direct access to high, user-controlled redox energy without involving photocatalysts or sacrificial redox additives, offering advantages to net-oxidative/reductive reactions. However,t he applied constant current or voltage can cause uncontrollable over-reductions/oxidations to afford by-products.T oa ddress the aforementioned limitations in PRC and SOE, organic chemists have recently explored their combination (Scheme 1). [4] Merging the advantages of these two important techniques has made photoelectrochemistry at ool for greener, more challenging and more selective molecular activations. [5] Pioneering reports by Xu, [5b-c,m] Lambert, [5g,h,i,k] Lin [5h,j] and Scheme 1. Previous reductive e-PRC reports involving C(sp 2 )ÀXc leavages to afford aryl radicals vs. this work involving C(sp 3  Wickens [5f] have shown that introducing applied potential in photoredox catalysis is not only beneficial for accessing challenging redox reactions,b ut is also ag reen replacement for sacrificial redox additives. Among the various strategies for combining photocatalysis and electrochemistry [4a] the sub-category coined electrochemically-mediated photoredox catalysis (e-PRC) is highly attractive.I na ddition to turning over "spent" closed-shell photocatalysts,e -PRC can also involve electrochemical generation of open-shell (radical ion) photocatalysts,followed by their photoexcitation to species with ultra-high redox potentials.Aseminal report from the Lambert group demonstrated this strategy for super-oxidations of highly electron-poor arenes. [5k] In the reductive direction, photoexcited radical anions of dicyanoanthracene (DCA) [5h] and of 2,6-diisopropylphenyl-containing naphthalenemonoimide (NpMI) [5f] are highly reducing species (E8 8 red < À3.0 Vv s. SCE) that reduce challenging aryl chlorides to their aryl radicals.E ven pchloroanisole was reduced, beyond reach of the photon energy limit of monophotonic PRC and where SOE inevitably leads to dehalogenation via subsequent aryl radical reduction (Scheme 1A). [6] Despite these elegant advances, reductive e-PRC and biphotonic strategies [3] are still heavily focused on the reductions of aryl halides/pseudohalides through C(sp 2 )ÀXb ond cleavages to generate aryl C(sp 2 ) radicals in an overall dehalogenation or functionalization with excesses of radical trapping agents. [5f,h] Inspired by previous reports, [5] we envisioned that phosphinates of aliphatic alcohols (E p red = À2.2 ! À2.6 Vvs. SCE) could undergo e-PRC reduction to give carbanions (Scheme 1B). Thereby,a ne lectroactivated-PhotoRedox Catalyst (e-PRCat) undergoes cathodic activation and photoexcitation to afford apotent reductant. SET reduction of 1 to its radical anion followed by C(sp 3 ) À Ob ond cleavage delivers benzyl radical 1' '.I ts further reduction [7d] to carbanion intermediate 1" would enable either an olefination (X = Cl, Br) or ad eoxygenation (X = H) process by am echanism that does not depend on hydrogen atom transfer agents or decarboxylation. [7] Herein, we report the e-PRC reduction of alkyl phosphinates to alkyl(sp 3 )c arbanions for olefination and deoxygenation reactions that i) proceeds under exceedingly mild conditions,i i) tolerates aryl halides/pseudohalides with similar or more accessible redox potentials than the target alkyl phosphinate moiety.

Results and Discussion
To assess the viability of our proposed e-PRC alkyl phosphinate reduction, we employed 2-chloro-1,2-diphenylphosphinate 1a as am odel substrate for the olefination reaction (Table 1). By using DCA as an e-PRCat and Zn(+ +)/ RVC(À)a st he electrodes in ad ivided H-cell, we examined the reduction of 1a under blue light irradiation and with different applied constant potentials.Ahigh constant voltage (U cell = À3.2 V) as used previously [5h] for electron-priming DCA to its radical anion for photoexcitation gave notable decomposition, desired product E-stilbene (E-2a)inonly 7% yield and a2 5% yield of diphenylethane 3a [8] (Table 1, entry 1). Al ower potential (U cell = À1.6 V) led to ar emarkable improvement in the reaction profile and yield of E-2a to 70 % ( Table 1, entry 2). Theo ptimal yield of E-2a was obtained at an even lower potential (U cell = À1.0 V). Cyclic phosphate ester 4a was also asuitable substrate for preparing product E-2a (entry 4), offering an attractive Corey-Wintertype olefination that avoids explosive/toxic trimethylphosphite,harsh activating reagents or high temperature.Control reactions omitting light, constant potential or e-PRCat confirmed the photoelectrochemical nature of the olefination reaction (entries 5-7). In contrast to DCA, NpMI as catalyst delivered higher amounts of Z-2a (entry 8). [9] Allowing the reaction to proceed for 48 h( entry 9) increased the E-/Zratio to 1/10 (71 %o fZ-2a). Detailed investigations (see Supporting Information (SI)) revealed that light, constant potential and NpMI are all advantageous to the isomerization, representing an ovel photoelectroisomerism of alkenes.
Here we opted to use Fe instead of RVCasacheaper,robust cathode material. [10] However, it was quickly identified that DCA and NpMI were ineffective e-PRCats for the majority of phosphinates.F or example,c yclic substrate 1d underwent no reaction with these catalysts (entries [10][11]. We synthesized n BuO-NpMI as an ovel e-PRCat which afforded the desired product 2d in very good yield (entry 12). Control reactions confirmed operation of e-PRC (entries [13][14][15], while cathode materials greatly impacted the reaction (for detailed optimizations,s ee SI). [11] Optimal conditions were examined for ar ange of olefination reactions (Scheme 2). Unsymmetrical Z-stilbenes 2b, 2c were prepared in high yields from the tandem e-PRC reduction/photoelectroisomerism process.C yclic olefins 2d-2h,r arely synthesized by the Wittig reaction due to the inconvenience of substrate preparations,w ere prepared in good to excellent (69-83 %) yields.T erminal olefin 2i could not be prepared in high selectivity by dehydration of its corresponding tertiary alcohol as such am ethod inevitably leads to the most substituted olefin, [12] in this case,atetrasubstituted instead of aterminal olefin.
After the successful preparations of as eries of E-styrene derivatives (exclusive isomers) bearing divergent substituents including -Ph (2j), -OBz(2k), -OMe(2l)and -CF 3 (2n)attheir arene rings,w eq uestioned whether halogen substituents could be tolerated by our reaction. This is ahighly challenging issue,s ince the reductions of aryl chlorides and bromides by photoexcited radical anions (either e-PRC or conPET-type) are highly efficient and heavily reported as discussed earlier (Scheme 1). [3c-g, 5f,h] With this aim, we tested phosphinates bearing either achloro-or bromo-substituent on their arene. To our delight, aryl chlorides 1o-1q and aryl bromide 1r underwent olefination in moderate to good (39-69 %) yields with high or exclusive selectivities for their E-orZ-isomers; only traces of dehalogenated styrenes were observed (> 10:1 in favor of olefination for 2p). Compared with products 2o-2p, p-chlorostilbene 2q has am ore conjugated p-system and is easier to reduce,y et still gave only traces of dechlorinated product 2a.S ubstrate 1s,b earing both an alkyl and aryl phosphinate, [13] selectively underwent e-PRC reduction of the alkyl phosphinate leading only to C(sp 3 ) À Ocleavage to afford 2s in good yield. Our method retains reductively labile C(sp 2 )ÀOf unctionality,p roviding complementary selectivity to arecent report involving aphenothiazine photocatalyst. [13] Styrene-forming substrates containing longer-chain aliphatic groups or ab enzyl group retained high E-isomer selectivity,a ffording 2t-2v in good to high (62-79 %) yields and high selectivities (> 10:1 in favor of their E-isomers). Olefin geometry is not impacted by the diastereomeric ratio of phosphinate precursors,but by the reaction conditions.For example,a lthough the diastereomeric ratios of phosphinate precursors to 2r, 2t and 2v were all > 30:1, the E-/Z-r atios were 4:1, 10:1 and 20:1 respectively.Hindered olefins derived from carbocycles 1w-1x were formed in high (83-87 %) yields.I nt he synthesis of 2x,o ur conditions offer an alternative to i) n BuLi or Grignard chemistry with expensive bromocyclobutane and ii)e xpensive Wittig reagents/cyclobutanone,i nstead starting from commercial, inexpensive cyclobutyl phenyl ketone.Our e-PRC phosphinate reduction offers complementary selectivity to Birch-type photochemical reports involving SET, [14] or E n T. [15] Naphthalene-based substrate 1ywas well-tolerated, affording 2yin good (62 %) yield without Birch-type reduction products.A mide 1z was also well-tolerated, in spite of its free proton and labile heterocycle that would react with strong bases.A lthough an alkyl phosphinate derived from anon-benzylic alcohol 1aa did not react, alkyl phosphinates derived from allylic alcohols were feasible.A llylic substrates 1ab, 1ac derived from naturallyoccurring terpenes were found to be sluggish, but afforded dienes 2ab, 2ac in satisfactory (30-33 %) yields in ac omplementary fashion to previous reports that require strong bases [16] or transition metal catalysis. [17] Demonstrating the utility of our base-free approach, products 2ad-2ag were synthesized from their alkyl pacetylbenzoate precursors.G iven the properties of Geraniol and Nootkatone as fragrance oils and cholesteryl benzoate as al iquid crystal, our reaction is au seful entry to terpeneloaded monomers for the synthesis of functional polymers. [18] Strategies involving strong base (for example i) Wittig reaction of an aldehyde or ii)ketone reduction, mesylation and E 2 -elimination) lead to hydrolysis or E 2 elimination of the benzoate, [19] while direct esterification suffers from the caveats that 4-vinylbenzoic is thermally sensitive and formulated with BHT stabilizer.F urther exemplifying utility, substrate 1ah,r eadily prepared from its a-dichloroketone, underwent selective reduction to its unsymmetrical stilbene 2ah in good yield while leaving the olefinic Cl atom untouched (Scheme 3). This demonstrates the value of our method which retains reductively labile halides for further functionalizations.T he method provides alternative access to unsymmetrical halogenated stilbenes that does not rely on transition metal catalysis. [20] While conPET photocatalysis and e-PRC are complementary approaches in the reductions of aryl halides/pseudohalides, [3d,g] conPET conditions did not effect the net-reductive transformation herein (Scheme 4).
Concerning the first question, measured reduction potentials (E p red )ofthe alkyl phosphinates (in good agreement with those calculated by DFT) did not correlate with reactivity (Table 2). Instead, comparison of the C(sp 3 ) À Ob ond-dissociation free energies (BDFEs) of phosphinate radical anions correlated well with reactivity.T his corroborated C(sp 3 )ÀO cleavage as the rate-limiting step and rationalized i) the unique tolerance of our conditions to aryl halides due to their less exergonic C À XBDFEs (entries 4,5;6,7) and ii)the lack of reactivity of phosphinates derived from non-benzylic/allylic alcohols that require higher temperatures [22] to assist C(sp 3 )À Oc leavage (entries 9,10).
As to the second question, NpMI and n BuO-NpMI had identical redox potentials (E 1/2 = À1.3 Vv s. SCE, Figure 1, left) by cyclic voltammetry.T heir radical anions are electrogenerated with equal efficiency,w hich is entirely consistent with the spin densities of their radical anions (Figure 1, right) being localized on the naphthalene and being unaffected by substitution on the N-aniline.S pectroelectrochemistry of both e-PRCats gave identical UV-vis bands for their radical anions ( Figure 2, left and see SI). Taken together, these results indicate that their excited radical anions are equally potent reductants.T op robe further,w ee lectrochemically generated NpMIC À and n BuO-NpMIC À under inert conditions for analysis by EPR (Figure 2, right). [23] In both cases,apentet was observed whose intensity was unchanged upon irradiation with blue LEDs.I nb oth cases,i nt he presence of 1d (10 equiv.), the EPR signal was identical in the dark (see SI), but upon irradiation by blue LEDs the EPR signal quenched, corroborating successful SET from the doublet states (D n )o fb oth catalysts 2 [NpMIC À *] and 2 [ n BuO-NpMIC À *] to 1d.G iven that the reaction of 1d is only successful with n BuO-NpMIC À and taken together with the discussion of E p red s and BDFEs in Table 2, this confirms SET is not the determining factor for the success of n BuO-NpMIC À .
Neutral and electroreduced forms of NpMI and n BuO-NpMI were probed by luminescence spectroscopy (Table 3). Forn eutral e-PRCats,a bsorbance and emission (fluorescence) spectra corresponded with the literature. [24] Measured lifetimes were t % 3.0 ns in both cases.A lthough some N-arylnaphthalimide derivatives have ultrashort-lived singlet states,d ue to rapid intersystem crossing to triplet states, [24] phosphorescence does not occur for the N-aryl-1,8-naphthalimides where N-aryl rotation becomes considerably hindered. [24] Electroreduction for 1hand selective excitation of the radical anions at 452 nm led to anew emission band (l max ca. 540 nm) and alonger-lived species with biexponential decay (t 1 % 7ns and t 2 % 20 ns) for both NpMIC À and n BuO-NpMIC À .T he doublet (D 1 )s tates of similar radical anions (naphthalene diimide radical anions,p erylene diimide radical anions) are picosecond-lived and do not luminesce, [25] and we confirmed by excitation spectra (see SI) that this emission was not deriving from the initially-formed excited state 2 [ n BuO-NpMIC À *] (Figure 2, left), but from alower-lying,longer lived excited state,t ermed "ES 1 ". Intersection of the longest wavelength excitation and shortest wavelength emission bands allows an estimation of E 0-0 for photoexcited states. [26] Fort hese emitting excited states,e stimated E 0-0 values (E ES ) for both [NpMIC À *] and [ n BuO-NpMIC À *] were (E ES = 56.6 kcal mol À1 )a lmost identical to the triplet energies (E T ) of *Ir III photosensitizers used in olefin photoisomerisms. [9a-c] It is therefore reasonable to propose E-/Z-p hotoisomerism occurs via energy transfer (E n T) from ES 1 .E n Tw ould be exergonic to E-stilbene and less so to Zstilbene (E T = 51.0 vs. E T = 55.5 kcal mol À1 , respectively), rationalizing high Z-stilbene selectivity. [9b,c,27] However,the lifetime of ES 1 was unchanged in the presence of 1d (10 equiv.), confirming its catalytic inactivity in the initial SET step. In their study of photoexcited benzo-[ghi]perylenemonoamide (BPI) radical anions for Birch reductions,M iyake and coworkers made similar observations. [14] They assigned the long-lived excited state as the lowest-lying quartet excited state ( 4 BPIC À *) arising from intersystem crossing (ISC) from the doublet state ( 2 BPIC À *). Therefore,t he lowest-lying quartet state 4 [ n BuO-NpMIC À *]   is ac andidate for ES 1 ,t hat allows E n Tt obe spin-conserved. We calculated the vertical excitation energy of this lowest quartet state with CASSCF (see SI) and found ar easonable agreement with the observed l max of luminescence.I ti s energetically close to the doublet states underlying the 415 nm absorption band so that ISC is plausible. Miyake similarly found that the putative 4 BPIC À *w as not catalytically active in the Birch SET step.T hey hypothesized SET from ahigher lying excited doublet state 2 BPIC À *(D n )in an anti-Kasha fashion. Consistent with previously reported anti-Kasha photochemistry of doublet excited state photocatalysts, [5a, 14] excitation of the broad absorption of 2 [ n BuO-NpMIC À *] between 650-900 nm (D 0 !D 1 )w ith 740 nm or 850 nm LEDs gave only traces of 2d. [28] Ruling out participation of the first excited state (D 1 ), "effective minimum" potentials (E 0 1/2 )ofNpMIC À * (D n )a tÀ3.7 Vv s. SCE and n BuO-NpMIC À *(D n )a tÀ3.8 Vv s. SCE can be calculated by previously described methods, [29] easily reaching E p red of all phosphinates herein as well as aryl halides. [30,31] Participation of ad oublet excited state in SET is consistent with aforementioned quenching of the EPR signal (Figure 2).
High-level DFT/MRCI calculations were carried out for n BuO-NpMIC À to characterize this D n state.T he computed spectrum (Figure 3, top) is in excellent agreement with the experimental absorption spectrum, especially at the band with l max = 415 nm comprising two bright p-p*s tates (D 0 ! D n and D 0 !D n+1 ). Contrary to the D 0 !D 1 transition around 870 nm, both these excitations transfer electron density from the naphthalene to the N-aniline unit of n BuO-NpMIC À (Figure 3, bottom). Preassembly of ground state radical anion and substrate could explain (i)p hotochemistry of ultrashortlived doublet states [25] and (ii)faster than rates of diffusion. [5a] Preassembly of n BuO-NpMIC À with 1d being more favorable than that of NpMIC À may explain the reactivity differences of the e-PRCats in effecting C(sp 3 ) À Oc leavage following SET, and may rationalize profound shift in the molecular site of reduction compared to previous reports. [32] However, like Miyake and co-workers,wewere unable to find spectroscopic evidence of preassembly by UV-vis or EPR (see SI). While the absence of spectroscopic perturbations does not rule out apreassociation, [33] preassembly could occur at the N-aniline that is spin-disconnected from the naphthalene where the radical anion spin density is localized (Figure 1, right). Spin densities of favorable candidate preassemblies at the Naniline unit of n BuO-NpMIC À found by computational geometry optimizations do not differ from that of n BuO-NpMIC À alone,w hile af avorable candidate preassembly at the naphthalene unit of n BuO-NpMIC À does differ (see SI). A preassembly at the N-aniline could also rationalize anti-Kasha photochemistry,since charge transfer to the N-aniline in the D n/n+1 states is proximal to the bound substrate and promotes intermolecular SET upon photoexcitation (Figure 3). In contrast, the charge density of the lowest excited doublet state D 1 remains localized on the naphthalene and is not close to the substrate.
Where spectroscopy offers little insight, at op-down approach varying catalyst structure and examining product yields has proven useful in investigating the mechanisms of reactions involving in situ-formed organic electron donors. [34] To probe the importance of ap reassembly of 1d at the Naniline of the e-PRCat, we explored the influence of aseries of e-PRCats with varying electronics and steric bulk (5a-f, Scheme 6). Compared to NpMI,c atalysts with electron donating alkoxy or p-anisole substituents on the naphthalene unit (5a, 5b)g ave no reaction. Compared to n BuO-NpMI, acatalyst with additional alkoxy substituents on the N-aniline  (5c)gave alower (41 %) yield of 2d.The yield of 2dincreased with decreasing steric hindrance at the ortho-positions of the N-aniline (NpMI ! 5d< 5e). [35] Ad ecrease in "steric bulk" likely promotes preassociation of radical anion e-PRCat and 1d.I no ur computational investigations we found multiple stable ground state preassemblies.G eometry optimizations (see SI) converged to pincer-like conformations for all candidates,w here two of the substratesa ryl groups coordinate to the N-aniline of the e-PRCat in aT Àp and pÀp orientation, respectively.The thermodynamics and kinetics of their formations (see SI) mirror reactivity trends in Scheme 6, corroborating apreassembly between e-PRCat and substrate before photoexcitation.

Conclusion
We report an electro-mediated photoredox catalytic reductions of phosphinates derived from a-chloroketones toward selective olefinations and deoxygenations.T his study reports reductive formation of alkyl carbanions via photoexcited radical anions as super-reductants.T he selective reduction of C(sp 3 ) À Ob onds in the presence of C(sp 2 ) À X bonds was achieved. Reactivity differences of various radical anion photocatalysts and anti-Kasha photochemistry,backed by computational insights,s uggest the importance of ac lose catalyst-substrate interaction for an effective,s elective reaction. In this context, our calculations indicate that intramolecular charge transfer in the catalyst radical anion upon photoexcitation promotes SET to the substrate.P hotocatalyst-substrate preassemblies such as EDAcomplexes, [36] noncovalent interactions, [5a, 37] hydrogen bonding [38] and ordering of solvent [39] are receiving increasing attention to unveil the next generation of photocatalytic transformations and offer new frontiers in selectivity and efficiency. Further studies into the nature of interactions and structure of preassemblies,a s well as catalyst stability, [40] are ongoing.