Photocatalytic Generation of Divalent Lanthanide Reducing Agents

Divalent lanthanide (Ln) compounds are excellent reducing agents with unique reactivity profiles. These reagents are typically used in superstoichiometric amounts, often in combination with harmful additives. Reactions catalytic in Ln(II) reagents that retain the reactivity and selectivity of the stoichiometric transformations are currently lacking due to the absence of effective and selective methods to form reactive Ln(II) species from stable precursors. Here, active Ln(II) is generated from a Ln(III) precursor through reduction by a photoexcited coumarin or carbostyril chromophore, which, in turn, is regenerated by a sacrificial reductant. The reductant can be metallic (Zn) or organic (amines) and can be used in strictly stoichiometric amounts. A broad range of reactions, including C–halogen, C=C, C=X (X = O, N), P=O, and N=N reductions, as well as C–C, C–X (X = N, S, P), and N–N couplings were readily carried out in yields and selectivities comparable to or better than those afforded by the analogous stoichiometric transformations. The reaction outcomes could be altered by changing the ligand or the lanthanide or through the addition of environmentally benign additives (e.g., water). EPR spectroscopy supported the formation of both Ln(II) and oxidized chromophore intermediates. Taken together, these results establish photochemical Ln(II) generation as a powerful strategy for rendering Ln(II)-mediated reactions catalytic.


■ INTRODUCTION
−12 SmI 2 often needs to be deployed in combination with superstoichiometric amounts of toxic additives, e.g., highly carcinogenic hexamethylphosphoramide (HMPA).Lanthanides, like many transition metals, are expensive, and their mining causes significant environmental damage. 13,14Thus, there is an urgent need to develop protocols that are catalytic in Ln(II). 15A handful of examples have demonstrated the possibility of Ln(II) catalysis by regenerating Ln(II) photochemically with rhodamine 6G and an amine, 16 with Zn, 17 or electrochemically; 18 sacrificial reductants were used in large excess. 18Crucially, existing examples have transformed only a narrow range of substrates lacking potentially sensitive functional groups and have thus not demonstrated the retention of the most attractive features of the stoichiometric reactions: high functional group tolerance, tunability of reducing power, and versatility.
−22 Catalysts are often based on organic dyes 19 or metal complexes, mainly transition metals; 23,24 they have short (ns) excited-state lifetimes, and their reducing power is limited to a narrow range.Ln(II)-based reductants have oxidation potentials that cover a broad range from −0.4 to −3.9 V; however, lanthanides have been underutilized as photocatalysts.Most Ln(III) ions can luminesce when sensitized by a light-harvesting chromophore (Figure 1, left cycle).Direct Ln(III) excitation is inefficient, as the 4f−4f transitions are Laporte-forbidden. 25The photoexcited chromophore is a powerful reductant.In the case of the most reducible Yb(III) 26 (E 1/2 (Yb(III)/Yb(II), YbCl 3 ) = −1.1 V vs NHE) and Eu(III) 27 (E 1/2 (Eu(III)/Eu(II), EuCl 3 ) = −0.42V vs NHE), Ln(III) reduction competes with Ln(III) luminescence sensitization via energy transfer (Figure 1, right cycle). 27,28Photochemical generation of Ln(II) from a Ln(III) precursor followed by reoxidation of the Ln(II) by a substrate and the reduction of the oxidized chromophore with a sacrificial reductant enables the use of catalytic amounts of lanthanide reagent.While such a lanthanide turnover has been demonstrated, the reported procedure relies on strongly reducing Dy(II), Nd(II), or Tm(II) (E 1/2 (Ln(III)/Ln(II) = −2.5, −2.6, − 2.3 V vs NHE, respectively 29 )); 16 the authors noted that using the much more accessible Ln(III) precursors instead of the Ln(II) ones afforded the product in lower yields. 16We hypothesized that a practical protocol for the in situ generation of Ln(II) from Ln(III) could be obtained by combining the chromophore with a multidentate ligand.
Here, we show that in situ-generated Ln(II) (Ln = Eu, Sm, Dy) species can reduce benzylic, allylic, and aryl halides efficiently and affect a broad range of functional group interconversions previously achieved only by (super)stoichiometric amounts of SmI 2 .Reductions were successfully coupled to a variety of C−C, C−N, C−P, C−S, and N−N bond formations, as well as to carbocycle and heterocycle syntheses.Taken together, these practical catalytic processes reproduce and even surpass stoichiometric Ln(II)-mediated reductions in scope, functional group tolerance, and versatility and are thus viable alternatives to the attractive, but resourceintensive stoichiometric transformations.

■ RESULTS AND DISCUSSION
Catalyst Design and Reaction Development.We designed three catalysts (LnL1−LnL3) (Ln = Eu, Sm, Gd, Dy) with different coordination environments (Chart 1).L1 and L2 were expected to strongly bind Ln(III) and Ln(II) and thus yield stable, long-lived catalysts.L3 relies on weak interactions between Eu(III) and 7-aminocarbostyril, and most lanthanide coordination sites are occupied by labile solvent molecules.L1−L3 incorporate a light-harvesting heterocycle, either a 6,7-oxycoumarin (L1, L2), or a 7-aminocarbostyril (L3), which allows for excitation with UV and visible light.Structurally L1 is related to the Allen cryptands. 17,30L2 is a flexible and synthetically more accessible open-chain analogue of L1.L1 and L2 were expected to stabilize Ln(II) compared to the solvated ion.The excited-state reduction potentials of the chromophores are < −1.88 V (vs Fc + /Fc, Page S78), which is sufficiently negative to reduce Eu(III) and Sm(III) and thus yield the reactive Ln(II) species.The catalyst design described here is highly modular.Catalyst reactivity can be tuned through the ligand that regulates Ln(II) stability and substrate access, through the chromophore excited state and the lanthanide.
Initial investigations focused on the mildest divalent lanthanide reductant Eu(II) to ensure a high functional group tolerance for the reaction and Zn as the sacrificial donor due to its ability to reduce cryptand-encapsulated Eu(III) to Eu(II) when used in large excess. 17Irradiation with 365 nm light or with a blue LED (λ em = 463 nm) of a solution of benzyl chloride (1a) containing 0.1 equiv of Eu(III)L (L = L1, L2, L3) yielded bibenzyl 1b using only 1 equiv of Zn (Condition A, Figure 2, Table S1).A range of nonmetallic sacrificial donors were then screened to reduce the amount of metal in the reactions using 1a as the substrate (Table S1).These studies allowed optimized conditions to be identified for the use of EuL1 and EuL2, as well as conditions for using either metallic (Zn) or organic sacrificial reductants.Conditions B (Eu(III)L1 (0.1 equiv), N,N-diisopropyl ethylamine (DIPEA, 1 equiv), HCO 2 H (0.5 equiv), 365 nm or blue LED) and C (Eu(III)L2 (0.1 equiv), DIPEA (10 equiv), LiCl (10 equiv), H 2 O (20%), blue LED).Conditions B and C are tailored to EuL1 and EuL2, respectively, and differ in the excitation wavelength and the amount of sacrificial donor.The shorter excitation wavelength is incompatible with several iodinated and brominated substrates that absorb in this region.Water addition (20%, condition C) shifts EuL2 absorption above 400 nm (ε(EuL2, 400 nm) = 3094 M −1 cm −1 with water vs 714 M −1 cm −1 without water), which allows for excitation with a blue LED, and thus increases the substrate scope and the selectivity toward bibenzyl formation.A catalyst loading of 10% was suitable for both small-and large-scale reactions and could be lowered to 5 mol % for reactions with >0.1 mmol substrate.EuL can be generated in situ from the ligand and an Eu(III) salt without impacting the reaction outcome (Table S4).Control experiments established that light was necessary for catalysis, as was the presence of EuL or SmL.The replacement of Eu/SmL either with the uncomplexed ligand (L1, L2, or L3) or with redox-inactive Gd(III)L gave only a trace amount of product.
Substrate Scope.Under the optimized conditions, either 1b or 1c could be obtained selectively from 1a in excellent yield (97 and 79%, respectively, first substrate in Figure 2a) using EuL1/L2 or EuL3, respectively.For comparison, the corresponding stoichiometric SmI 2 -mediated reaction affords 67% of 1b. 31 A range of benzylic halides could be reduced under conditions A, B, or C using EuL1 and EuL2 (Figure 2a).Benzyl bromides and chlorides were reduced selectively in the presence of a variety of redox-active functionalities, including ester (3a), ketone (11a), nitrile (2a), aryl bromide (7a), ether (4a, 11a, 18a), thioether (10a) and silyl ether (12a), and heteroaryl groups (13a−17a).Benzyl halide reduction could yield dehalogenated (1c− 17c) or bibenzyl (1b−18b) products, including biologically active Brittonin A (18b).While the reaction outcome was substrate-dependent, in several cases, either product could be selectively obtained by a simple change in the catalyst or reaction conditions.Such ligand-based selectivity is unprecedented for Ln(II)-mediated reductions.The product distribution likely depends on the rate at which a benzyl radical is generated and its stability, as is the case for Ir and Ru catalysts 32,33 (see entries 1, 3, and 11 in Figure 2a).Brittonin A synthesis could be scaled up to yield 807 mg of 18b in a single batch.
Aryl halides were successfully reduced by using catalytic amounts of lanthanide (Figure 2b).Electron-poor (hetero)aryl bromides and chlorides (19a, 22a−24a, 26a−27a) were efficiently dehalogenated, as were electron-rich bromides (e.g., 21a, 25a) and iodide (20a).The functional group tolerance of C(Ar)-X reductions was remarkable: aldehydes (21a) and esters (22a, 23a) were left intact.−36 Deuterium labeling experiments (24c) reveal the DMF solvent as a proton source under conditions A and C.However, no deuterium incorporation was seen under condition B when conducting reactions either in DMF-d 7 or with added formic acid-d 2 .These results are consistent with DIPEA being the proton source in condition B (Table S18).A range of catalytic functional group transformations that have previously been performed using superstoichiometric amounts of Ln(II) were then carried out under photocatalytic conditions (Figure 3).The reductions of 29a−37a proceed without side product formation in yields comparable to or better than that of the stoichiometric process while requiring up to 99% less lanthanide.Diazo compound 30a was reduced to hydrazine 30b, while azide in 32a yielded diazo compound 32b in 69% yield without over-reduction presumably due to the presence of the electron-donating p-amino groups.Pinacol coupling of 35a proceeded in yields comparable to that obtained previously using 2−10 equiv of SmI 2 . 37,38The reaction of 35a was catalyzed by EuL3 and not by EuL1 or EuL2.Substrate binding was ligand-and substrate-dependent, as indicated by the differences in the changes of Eu(III) luminescence spectral shapes and luminescence lifetimes (τ Eu ) upon the addition of various substrates to EuL1−3 (Figures S55−S67).The open Eu(III) coordination sphere of EuL3 is likely beneficial for binding a bulky substrate such as 35a.Additionally, the coordination of the H-donor (XH) to Ln(II) is more likely for EuL3 than for ligand-encapsulated EuL1 and EuL2.Such XH-coordination has been shown to generate a strong proton-coupled electron transfer reagent, capable of simultaneously transferring an electron and a proton to the substrates that are difficult to reduce. 39The reduction of pentavalent phosphorus compounds is an industrially important reaction, often requiring silyl or metal trapping reagents. 40Trimethyl phosphate was reduced in moderate yield to phosphite 34b using EuL2 as the catalyst and 1 equiv of Zn.
The reactivity of the catalytic system is readily tuned.Either 36b or 36c could be obtained selectively from 36a by performing the reaction in the presence of strictly 10 equiv.or a large excess of H 2 O (20% v/v), respectively.Catalyst reactivity was also lanthanide-dependent. Imine (31a), aldehyde (49a), and oxime (50a) were not efficiently reduced with EuL2, but the more powerful SmL2 afforded the products in yields comparable to what was observed using stoichiometric Sm(II) reagents. 41,42A similar improvement in the yield was seen upon reduction of 24a with SmL2 rather than EuL2, presumably also because of the higher reducing power of SmL2.Conversely, the more reducible substrates 1a and 2a underwent unselective reactions in the presence of SmL, while with the milder reductant EuL products were obtained with excellent selectivity (Figure 2a).The radical intermediate can be captured by acceptors such as α,β-unsaturated ketones (38a−41a), aldehydes (42a, 43a), heteroarenes (44a, 46a), and arenes (47a) enabling C−C bond formation in good yields (Figure 4a−c).Benzyl addition to α,β-unsaturated ketones (38a−41a) was possible in a 1,4-fashion to afford 38b−41b in good yields.Allylic bromides 42b and 43b reacted with aldehydes 42a and 43a, respectively; the 1,2-addition products were obtained in moderate yields.Quinoline (44a) alkylation produced a mixture of C2/C4 alkylated isomers (44b, 44c) (Figure 4c), while an analogous intramolecular reaction to a phenyl ring in 45a gave tricyclic fluorene (45b), albeit in low yield.A similar addition of 46b to isoquinoline 46a enabled the synthesis of a natural product, papaverine, in a good yield.The aryl radical formed from 24a could also be harnessed to enable C−S and C−P bond formations (Figure 4d).
Reactions with triethyl phosphite and DMSO gave the corresponding phosphate and methyl sulfide-substituted isoquinolines 47b and 47c, respectively.Such synthetic modifications of the isoquinoline pharmacophore 43 could open up new possibilities for medicinal chemists.
Some products are of particular interest.Pinacol cyclization of 49a to trans-1,2-diol 49b, a structural motif central to the pradimicin and benanomicin antibiotic classes, 44 was stereoselective.Preparations of formamide analogue 48b and tetrasubstituted pyrazines 50b from simple precursors (DMF and oxime, respectively) could open up new avenues for the synthesis of various drugs incorporating these important pharmacophores. 45,46imitations to the protocol remain.Figure S7 presents a range of aryl halides that were not reduced under conditions A, B, or C.These substrates may require more powerful reductants, e.g., Ln(II) species with more negative reduction potentials. 47Attempted cross-reactions between benzyl chloride and aryl, alkynyl, or carbonyl acceptors and unsuccessful pinacol coupling reactions are shown in Tables S48−S51 and Figure S7, respectively.The success of these cross-couplings likely requires an adjustment of the ligand.
Mechanistic Studies.LnL absorb at the excitation wavelengths used in the reactions; absorption is through the chromophore (Chr, Figure 5a).To investigate the fate of the excitation energy, the fluorescence quantum yields (Φ L ) and lifetimes (τ L ) of the chromophores (Chart 1) in GdL1 and GdL2 were compared to those of EuL1/SmL1 and EuL2/ SmL2, respectively.Gd(III) has a similar ionic radius and heavy atom effect to Eu(III) and Sm(III) but is not photoactive. 25The reduction potential required to access Gd(II) is much more negative than that required for accessing Eu(II). 29,48Thus, Gd(III)L provides a model for Eu(III)L that recreates the coordination and electronic properties of the complexes but does not allow for Ln(II) formation.The Gd complexes had higher Φ L and longer τ L , which is consistent with the presence of additional processes (energy or electron transfer) quenching the first singlet excited state of Chr in EuL and SmL.Electron transfer from Chr to Ln(III) (Ln = Eu, Sm) was calculated to be thermodynamically feasible (Tables S55  and S56).

Journal of the American Chemical Society
There are several conceivable mechanistic pathways for the catalytic processes to follow.Ln(III) reduction by the photoexcited chromophore yields Ln(II), 49,50 which may be intercepted by a substrate either before or after Chr regeneration by the sacrificial reductant (Figure 1).Control experiments established that it is the photoexcitation of Eu/ SmL that yields the active reagent (Table S2).Neither the uncomplexed ligand (L1, L2, or L3) nor Gd(III)L gave trace amounts of product.These experiments show that Ln(II) formed from the reducible Ln(III) rather than the excited chromophore reduces the substrate.An Eu(III) complex of the ligand lacking the sensitizing antenna (L2m, Chart 1) did not promote the reaction; therefore, the Eu(III) is reduced by the excited antenna and not directly by the sacrificial reductant.Irradiation of a mixture of 1 equiv of either EuL1 or EuL2 and 1a in the absence of a sacrificial reductant gave 1b in 78% and 42% GC yield, respectively, while the same reaction with GdL1 or GdL2 did not afford any product.Thus, Eu(III) can be reduced by the photoexcited Chr and in turn can reduce the substrate; the role of the sacrificial reductant is to regenerate Chr and enable the use of catalytic amounts of EuL.Eu(II) is a more powerful reductant in its excited state than in its ground state. 51Eu(II)L1 was inactive in the absence of light (Scheme S2).Thus, either the photoinduced electron transfer (PeT) yields Eu(II) in its excited state or ground-state Eu(II) is initially formed and gets excited.
In the presence of the radical quencher 2,2,6,6-tetramethylpiperidinyloxy (TEMPO), the reaction does not proceed (table S6).LCMS analysis of an irradiated sample containing EuL1 and TEMPO showed the protonated molecule ion of the L1-TEMPO adduct (m/z = 662, Figure S118).To gather further evidence of either Chr •+ or Eu 2+ formation, two EPR experiments were carried out.First, solutions of EuL/GdL containing equimolar amounts of the radical quenchers TEMPO or PBN (N-tert-butyl-α-phenylnitrone) were irradiated.TEMPO, but not PBN, is an EPR-active free radical.PBN can instead form stable radical species after reaction with an organic radical.EPR analysis of the irradiated TEMPOcontaining solutions indicated 20−50% reduction of the TEMPO signal in the presence of EuL (Figure 5b) but no quenching in the presence of GdL (Figure S69).The irradiation of a mixture of EuL2 and PBN showed the emergence of an EPR signal corresponding to a N-based radical from a PBN adduct (Figure S70). 52EPR analysis at cryo temperatures (10 K) of an irradiated (30 min, blue LED) solution of EuL2 showed features at g = 5.9, 3.13, 2.8 and a broad feature at g = 2, in addition to a sharp radical-like signal visible at g = 2.01 (Figure 5c, red).All of these features disappeared after the sample was thawed, kept at room temperature for 5 min, and frozen again (Figure 5c, blue).The sharp signal at g = 2.01 is an organic radical.The other features collectively indicated the formation of Eu 2+ species, 53,54 the identity of which was further supported by comparison to the EPR signals of Eu(II)L1 and isoelectronic GdL2 (Figure S73).These results are consistent with the formation of a radical cation and Eu 2+ in EuL under illumination by electron transfer from Chr.To the best of our knowledge, the Eu(II) EPR spectrum is the first spectroscopic evidence of such photochemical Ln(II) formation.

■ CONCLUSIONS
Strongly reducing Ln(II) species were accessed from Ln(III) precursors through photochemical reduction by an excitedstate organic chromophore and used as catalysts.Excitation could be performed with both UV and visible light.The Ln(II) could reduce a broad range of substrates with excellent functional group tolerance.The catalytic cycle was closed by reduction of the oxidized chromophore with Zn or nonmetallic sacrificial reductants.The selectivity of the reaction and the reducing power of the catalyst could be independently tuned by the ligand, the lanthanide, and additives (H 2 O).Ln(II) catalysis has been demonstrated on a broad substrate scope with the yield and selectivity comparable to or better than the corresponding stoichiometric processes on a synthetically useful scale.The  as they can be performed without toxic HMPA, and with up to 99% less lanthanide.

Figure 1 .
Figure 1.Proposed pathway for Ln(II) generation (right cycle) as an alternative to Ln(III) luminescence sensitization (left cycle).The order of the substrate reduction and chromophore (Chr) regeneration steps may be inverted.

Figure 2 .
Figure 2. Substrate scopes of benzylic (a) and aryl halide (b) reduction reactions.Reduced functionalities are shown in blue, and potentially sensitive functional groups are shown in red.

Figure 3 .
Figure 3. Catalytic C−S, N�N, C�C, and P�O reductions (left, center).C�O reduction followed by pinacol coupling and selective aldehyde and nitro group reductions.
scope of the Ln(II) catalysis includes synthetically important C−C, C−N, C−P, C−S, and N−N bond formation reactions, C−halogen, P−O, C−C, C−S, and N−N bond cleavage reactions, and the synthesis of biologically important carbocyclic and heterocyclic structural motifs.Eu(II) formation via the reduction of Eu(III) by a photoexcited nearby chromophore was supported by EPR spectroscopic findings.These photocatalytic reactions are substantially more benign than their stoichiometric analogues,

Figure 5 .
Figure 5. (a) Normalized absorption (gray), excitation (blue), and steady-state emission (red) spectra for EuL1 and EuL2 in DMF (λ exc = 339 and 348 nm, respectively).(b) Room-temperature EPR spectra of TEMPO and EuL1/EuL2/EuL3 containing TEMPO (1:1) after 12 h irradiation with a blue LED, microwave power 2 μW, and modulation amplitude 1 G.The integrated area of the radical signal (normalized to the TEMPOonly sample) is shown in the figure.(c) EPR spectra of EuL2 (1 mM in DMF) before illumination (black), directly after illumination (red), and after 5 min at room temperature (blue).EPR parameters: microwave power 2 mW, modulation amplitude 19.4 G, temperature: 10 K.The signal marked with * in the EPR spectra is a contamination from Fe(III), and the signals marked with ** are a small amount of Mn(II).The cavity signal has been subtracted from the spectra, and a baseline correction has been applied.