Enantioselective Synthesis of γ-Oxycarbonyl Motifs by Conjugate Addition of Photogenerated α-Alkoxyradicals

Abstract

Enantioselective catalytic Giese addition of photogenerated α-alkoxy radicals to acyl pyrazolidinones can be accomplished using a tandem Sc(III) Lewis acid/photoredox catalyst system. Surprisingly, the excited state oxidation potential was not the only important variable, and the optimal photocatalyst was not the strongest oxidant screened. Our results show that both the oxidation and reduction potentials of the photocatalyst can be important for the reaction outcome, highlighting the importance of holistic consideration in designing photochemical reactions.

Graphical Abstract

Retrosynthetic disconnections are often defined by the relative disposition of oxygen-containing functional groups.¹ The 1,3-dioxygenation pattern is perhaps the most iconic of these retrons, suggestive of aldol or Claisen disconnection, and the development of stereocontrolled aldol reactions has figured prominently in the history of asymmetric synthesis.² Similarly, 1,5- and 1,2-dioxygenation patterns, suggestive of Michael addition or Rubottom oxidation retrons, respectively, have also been targeted by enantioselective catalytic methods (Figure 1A).³ The enantiocontrolled catalytic synthesis of 1,4-dioxygenation patterns, however, has remained a conspicuous challenge with few highly effective solutions,⁴ despite their presence in medicinal lead compounds.⁵

Figure 1.

a) Retrosynthetic disconnections defined by dioxygenation patterns. b) Stereocontrolled access to 1,4-dioxygenation patterns via asymmetric photoredox conjugate addition.

Our laboratory has a long-standing interest in stereocontrolled reactions of photogenerated radicals. We demonstrated that Giese addition of α-silyl amines to Michael acceptors, a classical photoreaction studied in detail by Mariano,⁶ could be made highly enantioselective using a dual catalyst protocol.⁷ This method utilizes a Ru(II) photoredox catalyst that initiates the oxidative desilylation of an α-silylamine pronucleophile and a chiral Lewis acid co-catalyst that activates the Michael acceptor and dictates the facial selectivity of the radical addition (Figure 1B). We reasoned that the use of α-silyl ether pronucleophiles might offer an analogous strategy for the generation of chiral γ-alkoxycarbonyl adducts. While this modification is conceptually straightforward, it presents a significant tactical challenge because oxygen functionalities are substantially harder to oxidize their nitrogen counterparts (Figure 1B).⁸ The use of α-silyl ethers in racemic conjugate additions has previously been reported by Steckhan⁹ and Woo,¹⁰ but these reactions have been limited to highly electron-deficient Michael acceptors. Recently, Melchiorre described enantioselective Giese additions to enals using organophotoredox/chiral amine dual catalysis.¹¹ A single example using a photogenerated α-alkoxyradical was described, but the generality of this process was not studied. We hypothesized that a Lewis acid cocatalyst might expand the scope of this reaction to include less electrophilic acceptors and provide an opportunity for stereocontrol in the Giese addition of oxyfunctionalized organoradical intermediates.

Subjecting α-silyl ether 1a and pyrazolidinone 2a to conditions optimized for radical conjugate addition of α-silyl amines, however, gave only returned reactants (Table 1, entry 2), indicating that the excited-state photocatalyst Ru*(bpy)₃²⁺ is unable to oxidize 1a to produce the requisite radical intermediate. We hypothesized that a more oxidizing photocatalyst might be able to initiate the reaction and conducted a screen of several candidates commonly utilized in oxidative photoredox catalysis. Surprisingly, there was no correlation between the excited-state oxidation potential of the photocatalyst and the yield of 3a (Table 1, entries 3–5); the most oxidizing transition metal and organic photocatalysts in our screen afforded poor conversions after 12 h. The Fukuzumi acridinium salt, a common oxidizing photocatalyst,¹² produced somewhat better results (entry 5). The highest yield, however, was obtained using [Ir(dFCF₃ppy)₂(dtbbpy)]PF₆ (Ir), which possesses a comparatively modest excited-state oxidation potential.¹³ Lowering the reaction temperature improved the mass balance and enantioselectivity of the reaction (entry 6). Control experiments (entries 7–10) showed that no product is formed in the absence of light, photocatalyst, or Lewis acid. The addition of a soluble chloride salt (Bu₄NCl) increased the yield and enantiometric excess (e.e.) of the conjugate addition, but the effect was less pronounced than in the α-silylamine reaction (entry 10).⁷ Finally, no product was formed when insoluble ScCl₃ was used in place of Sc(OTf)₃ (entry 11).

Table 1.

Variations of reaction conditions.^a

entry	Deviation from standard conditions	E*ox vs SCE	yield	e.e.
1	None	+ 1.21 V	82%	95%
2	[Ru(bpy)3]Cl2 instead of Ir, rt	+ 0.77 V	0%	-
3	[Ru(bpz)3]Cl2 instead of Ir, rt	+ 1.45 V	0%	-
4	MesAcrMe+ instead of Ir, rt	+ 2.06 V	45%	83%
5	TPPT instead of Ir, rt	+ 2.30 V	16%	85%
6	rt instead of −25 °C	+ 1.21 V	71%	87%
7	no light	+ 1.21 V	0%	-
8	no photocatalyst	-	0%	-
9	no Sc(OTf)3	+ 1.21 V	0%	-
10	no Bu4NCl	+ 1.21 V	72%	90%
11	ScCl3 instead of Sc(OTf)3	+ 1.21 V	0%	-

^aReaction conditions: 1a (0.15 mmol, 1 equiv), 2a (0.225 mmol, 1.5 equiv), photocatalyst (3 × 10^–3 mmol, 2 mol%), Sc(OTf)₃ and Bu₄NCl (0.0225 mmol, 15 mol%), (S,S)-sBuPybox (0.03 mmol, 20 mol%), MeCN (3.0 mL, 0.05 M), 34 W blue LEDs. Yields determined by ¹H NMR analysis of crude reaction mixtures using mesitylene as internal standard.

Figure 2 summarizes studies exploring the scope of the enantioselective photocatalytic Giese addition. With respect to the α-silylether pronucleophiles, functional groups such as halogens and boronic acids are well-tolerated under the reaction conditions, providing handles for further synthetic manipulation. Steric hindrance led to a lower yield (3ia), while both mildly electron-donating groups and electron-withdrawing groups provided good yields and excellent e.e. However, highly electron-rich and electron-deficient pronuclephiles did not yield the desired conjugate addition products. In the case of highly electron-rich substrates, we observed decomposition, while for highly electron-deficient substrates, we could recover the starting pronuclephiles in good yield.

Figure 2.

Reaction scope studies. Standard conditions: 1 (0.50 mmol, 1 equiv), 2 (0.75 mmol, 1.5 equiv), Ir (0.01 mmol, 2 mol%), Sc(OTf)₃ and Bu₄NCl (0.075 mmol, 15 mol%), (S,S)-sBuPybox (0.10 mmol, 20 mol%), MeCN (10.0 mL, 0.05 M), 34 W blue LEDs, 24 h. ^a 18 h.

With respect to the Michael acceptor, we noticed a sensitivity toward the nature of the β-substituent, while α-substituents shut down reactivity. Increasing steric bulk leads to diminished yields. On the other hand, we were pleased to see that synthetic handles such as chlorine and terminal alkene remained intact during the reaction. Polar functional groups including Boc-protected amine, ether, and amino acid derivatives are also tolerated. Lewis-basic functionalities such as esters result in decreased e.e. Unexpectedly, aromatic substituents, which were well-tolerated in previously reported α-amino radical addition, did not lead to desired products. Rather, we observed no further reaction after enone E/Z isomerization, suggesting that access to the triplet enone is feasible under the reaction conditions. Nevertheless, the current system enabled functionalization of moderately electrophilic Michael acceptors, overcoming a major limitation of Steckhan’s and Woo’s previous reports.^9,109,10 The absolute configuration of 3a was confirmed by X-ray crystallographic analysis of the corresponding carboxylate salt (see SI).

The most intriguing feature of our optimization studies was that the yield of the reaction did not correlate with the excited-state oxidation potential of the photocatalyst. Moreover, the excited-state oxidation potential of the optimal Ir photocatalyst (+1.21 V) seems insufficient to directly oxidize the α-silyl ether nucleophile (+1.45 V), suggesting that the initiation mechanism of this process is different from that of the α-silyl amine reaction we reported previously.⁷ We thus elected to study the mechanism of this process in greater detail.

We first conducted a Stern–Volmer study to interrogate the ability of both reactants to quench the excited-state of the photocatalyst, either alone or in the presence of added Lewis acid (Figure 2).¹³ The photoluminescence of the photocatalyst does not change significantly with varying concentrations of the α-silyl ether either with or without added Sc(OTf)₃. These resuls indicate that one-electron photooxidation of the α-silyl ether does not initiate this photoreaction. In contrast, we observed a concentration-dependent increase in photoluminescence quenching in the presence of Michael acceptor 2a, and the degree of quenching is significantly greater in the presence of added Sc(OTf)₃ (Figure 2). Thus, the principal photocatalytic quenching process in this reaction is between the excited-state photocatalyst and the Sc-bound complex of Michael acceptor 2a.

Because Stern–Volmer analysis does not distinguish between energy and electron-transfer mechanisms, we considered both possibilities (Figure 3). First, we recently discovered that Lewis acid co-catalysts can accelerate energy transfer from excited-state Ru(II) and Ir(III) photocatalysts to enone acceptors, producing Lewis acid-bound triplet-state enones that undergo enantioselective [2+2] cycloadditions.¹⁴ Mariano¹⁵ and Melchiorre¹⁶ have reported desilylative conjugate additions initiated by excited-state Michael acceptors. We wondered whether this reaction might proceed via initial Lewis acid catalyzed energy transfer (Figure 3A). The key activation step would be the formation of the Lewis acid bound triplet-state 2a, which might oxidize α-silyl ether 1a to the requisite α-alkoxy radical. However, we quickly determined that this electron-transfer step would be endergonic. The excited-state redox potential of a compound can be estimated using the ground-state reduction potential and the energy of the triplet excited state. Cyclic voltammetry experiments using 2a gave irreversible reduction features with estimated E_p values of −1.1 V and −1.2 V vs SCE in the absence and presence of Sc(OTf)₃, respectively (see SI). We were not able, unfortunately, to ascertain the triplet excited state energies of 2a either experimentally or computationally. Because efficient energy transfer depends on the thermodynamics in the exchange event, however, the maximum value for the triplet energies cannot be substantially higher than the triplet energy of the Ir sensitizer (2.6 eV).¹² Given this limiting condition, the maximum excited state reduction potential available from the Sc-bound complex of 2a cannot be significantly more positive than +1.4 V. One-electron oxidation of the α-silyl ether pronucleophile by excited-state 2a is thus thermodynamically unfavorable by at least 370 mV, and potentially more.

Figure 3.

Possible energy-transfer (A) and photoredox (B) mechanisms for asymmetric photocatalysis.

We conclude, therefore, that product formation in this reaction is initiated by electron-transfer quenching of the photocatalyst excited state. A plausible mechanism consistent with the available data is depicted in Figure 3B. The excited-state photocatalyst is oxidatively quenched by the Lewis acid bound Michael acceptor 2a. While it does not seem reasonable to expect the radical anion of 2a to be capable of oxidizing α-silyl ether 1a, the photoinduced electron-transfer event also generates a ground state Ir(IV) species with a reported redox potential of +1.7 V vs SCE,¹² from which the oxidative desilylation of 1a should be thermodyanamically feasible. The addition of the resulting α-alkoxy radical to Michael acceptor 2a would be accelerated by the chiral Lewis acid, which also determines the facial selectivity. This proposal is consistent with experiments examining the effect of photocatalyst identity on the outcome of the desilylative conjugate addition (Table 2). Significant product formation was only observed using the optimal [Ir(dFCF₃ppy)₂(dtbbpy)]PF₆ photocatalyst and [Ir(dFppy)₂(bpy)]PF₆, a photocatalyst with similar excited state reduction and ground-state oxidation abilities¹⁷ (entries 1 and 4). Both of these electrochemical characteristics appear to be critical. Ir(ppy)₃ features an excited state easily able to reduce 2a but an Ir(IV) ground-state oxidation potential insufficient to oxidize 1a and gives negligible product (entry 3). [Ru(bpz)₃](PF₆)₂ has an oxidized Ru(III) ground state capable of oxidizing 1a but an excited state unable to reduce 2a¹⁸ and is similarly ineffective in this reaction. Finally, we observe no correlation between the excited state triplet energy of the photocatalyst and the yield of the conjugate addition product. For example, Ir(4-F,4’-tBuppy)₃ features a similar excited state energy¹⁷ to [Ir(dFCF₃ppy)₂(dtbbpy)]PF₆ but results in the formation of no conjugate addition product (entry 2), consistent with the conclusion that triplet sensitization of Michael acceptor 2a is not part of the product-forming reaction pathway.

Table 2.

Reaction outcomes with photocatalysts of similar triplet energies.^a

entry	photocatalyst	ET (kcal/mol)	E1/2 (M*/M+) V vs SCE	E1/2 (M+/M) V vs SCE	yield (ee)
1	[Ir(dFCF3ppy)2(dtbbpy)]PF6	60.1	−0.89	+1.69	82% (95%)
2	Ir(4-F,4’-tBuppy)3	59.4	−1.27	+0.98	0%
3	Ir(ppy)3	55.2	−1.97	+0.78	4%
4	[Ir(dFppy)2(bpy)]PF6	54.7	−0.80	+1.49	68% (80%)
5	[Ru(bpz)3](PF6)2	48.4	−0.26	+1.86	0%

^aReaction conditions: 1a (0.15 mmol, 1 equiv), 2a (0.225 mmol, 1.5 equiv), photocatalyst (3 × 10^–3 mmol, 2 mol%), Sc(OTf)₃ and Bu₄NCl (0.0225 mmol, 15 mol%), (S, S)-sBuPybox (0.03 mmol, 20 mol%), MeCN (3.0 mL, 0.05 M), 34 W blue LEDs. Yields determined by ¹H NMR analysis of crude reaction mixtures using mesitylene as internal standard.

In summary, we have developed a highly enantioselective method for the synthesis of γ-aryloxycarbonyl structures for which there are limited existing retrosynthetic disconnects. This transformation is accomplished using a dual chiral Lewis acid/photoredox catalyst system. An analysis of the energetics of the possible electron transfer steps suggests a mechanism in which oxidative quenching results in an Ir(IV) complex, initiating an enantioselective radical chain process. An important feature of this study is the observation that the most effective catalyst is not the species with the strongest oxidizing excited state, but rather one with a balanced electrochemical properties. Given the range of activation mechanisms available using photoredox catalysis and the widely differing photophysical properties of excited-state compounds in different classes, it seems reasonable to conclude that no single thermodynamic parameteris likely sufficient to predict the optimal photocatalyst for a given application. A complete understanding of photoredox mechanisms requires a holistic appreciation of many potentially relevant thermodynamic parameters.