Photoinduced copper-catalysed asymmetric amidation via ligand cooperativity

Abstract

The substitution of an alkyl electrophile by a nucleophile is a foundational reaction in organic chemistry that enables the efficient and convergent synthesis of organic molecules. Whereas substantial progress has been reported in recent years in exploiting transition-metal catalysis to dramatically expand the scope of nucleophilic substitution reactions using carbon nucleophiles^1–,2,3,4, there has been limited progress in corresponding reactions with nitrogen nucleophiles^5–6,7,8. Furthermore, for many substitution reactions, the bond construction itself is not the only challenge, as there is a need to control stereochemistry at the same time. Here, we describe a method for the enantioconvergent substitution of unactivated racemic alkyl electrophiles by a ubiquitous nitrogen-containing functional group, an amide, through the use of a photoinduced catalyst system based on copper, an earth-abundant metal. This process for asymmetric N-alkylation relies upon three distinct ligands (a bisphosphine, a phenoxide, and a chiral diamine) that assemble, in situ, a copper/bisphosphine/phenoxide complex that serves as a photocatalyst and a chiral copper/diamine complex that catalyzes enantioselective C–N bond formation. This study thus expands enantioselective N-substitution by alkyl electrophiles beyond activated electrophiles (those bearing at least one sp- or sp²-hybridized substituent on the carbon undergoing substitution)^{8–,9,10,11,12,13} to include unactivated electrophiles.

Amides, including chiral compounds of type A (Fig. 1a), are a ubiquitous functional group in bioactive molecules^14–,15,16, and, correspondingly, amide bond formation through the coupling of a carboxylic acid (or derivative) with an amine is perhaps the most widely used reaction in medicinal chemistry^17,18. The most common method for the synthesis of enantioenriched amides such as A is via the acylation of an enantioenriched amine (Fig. 1a: N-acylation). Although the N-alkylation of a primary amide by an enantioenriched electrophile would represent a complementary approach to chiral secondary amides of type A (Fig. 1a: N-alkylation), to the best of our knowledge such an S_N2 reaction has not been described, presumably due to the relatively poor reactivity of unactivated secondary alkyl electrophiles.

Fig. 1. Chiral secondary amides.

a, Strategies for their synthesis. b, This study. Ar, Ar¹ = aromatic substituents, R = carbon substituent, X = leaving/anionic group, t-Bu = tert-butyl, e.e. = enantiomeric excess, equiv. = equivalent, Et = ethyl, Me = methyl, Ph = phenyl, i-Pr = isopropyl.

Despite substantial limitations such as this, constructing C–N bonds via the S_N2 reaction of an alkyl electrophile with a nitrogen nucleophile is a widely exploited transformation in organic synthesis, reportedly the most frequently used method for achieving N-substitution¹⁹ and the fifth most-used reaction overall in the pharmaceutical industry¹⁷. Unfortunately, expanding the scope of C–N coupling reactions of alkyl electrophiles with nitrogen nucleophiles via classical substitution pathways (i.e., S_N2 or S_N1) is challenging, stymied by E2 elimination (S_N2) and by deactivation of the nucleophile by protonation or formation of a Lewis acid-base complex (S_N1); furthermore, neither of these classical pathways is readily amenable to enantioconvergent C–N bond formation beginning with racemic electrophiles (generally, inverted stereochemistry for S_N2 reactions and racemic products for S_N1 reactions).

Recognizing the potential impact of the development of powerful new strategies for N-alkylation, we and others have pursued the discovery of catalyzed processes, establishing that transition metals such as copper can enable bond constructions that had not previously been achieved, via new pathways^5,6,10,12,^{20–,21,22,23}. For example, in the case of photoinduced, copper-catalyzed reactions, the alkyl electrophile is transformed into an alkyl radical prior to C–N bond formation⁶.

With regard to our goal of developing a new, complementary approach to the synthesis of chiral secondary amides of type A (via N-alkylation, rather than N-acylation; Fig. 1a), a radical-based pathway opens the door to exploiting catalysis not only to increase the scope of suitable reaction partners, but also to control the stereochemistry of the product using a racemic electrophile. Thus, rather than needing to start with an enantiopure electrophile (Fig. 1a: failed alternative approach), a chiral catalyst would activate a readily available racemic electrophile and simultaneously achieve C–N bond formation and stereochemical control (Fig. 1b). It is important to note that, to obtain good enantioselectivity in the synthesis of chiral amide A, the chiral catalyst must effectively distinguish the two alkyl substituents of the unactivated secondary electrophile (alkyl vs alkyl¹ in Fig. 1); in asymmetric synthesis, differentiating between alkyl groups (e.g., ethyl vs n-propyl) is generally substantially more difficult than differentiating between an alkyl and an unsaturated group (e.g., ethyl vs phenyl)²⁴.

In this report, we describe the achievement of our proposed strategy for the enantioconvergent synthesis of chiral secondary amides via couplings of primary amides with unactivated racemic alkyl electrophiles through the use of a copper-based catalyst system that is activated by blue-LED irradiation (Fig. 1b). Key to this new method is the use of three ligands, a bisphosphine and a phenoxide that generate a copper photocatalyst that is highly reducing in its excited state and thereby capable of activating the unactivated electrophile, as well as a diamine that affords a chiral copper catalyst that simultaneously achieves C–N bond formation and control of stereochemistry.

As noted above, to achieve enantioconvergent N-alkylation in the case of an unactivated alkyl electrophile, the chiral catalyst must differentiate between two alkyl groups, which is often a difficult task in asymmetric catalysis. To obtain good enantioselectivity in challenging situations such as this, a common strategy is to exploit a Lewis-basic functional group that can provide a two-point interaction between the substrate and the chiral catalyst in the stereochemistry-determining step of the process, thereby enabling the differentiation of two similar substituents²⁵. Accordingly, in our initial studies, we examined photoinduced, copper-catalyzed enantioconvergent N-alkylations with electrophiles that include a phosphonyl group, since γ-aminophosphonic acids have been shown to be bioactive^26,27. Upon exploring a variety of reaction conditions, we determined that the desired catalytic asymmetric synthesis of secondary amides can be achieved in good yield and enantiomeric excess (e.e.) upon irradiating a copper-based catalyst system with blue-LED lamps (Fig. 2, 1; 95% yield and 93% e.e.). By comparison, under the same conditions, racemic 3-bromoheptane undergoes substitution to furnish the secondary amide in 59% yield and 8% e.e.

Fig. 2. An array of amides serve as nucleophilic coupling partners.

Couplings were generally conducted using 0.5 mmol of the amide. All data represent the average of two experiments. The percent yield represents purified product. Boc = tert-butoxycarbonyl, i-Bu = isobutyl, n-Bu = n-butyl, d.r. = diastereomer ratio.

The stereochemistry of the product (93% e.e. (S)) is the same regardless of whether (rac)-P, (R)–P, or (S)–P is used (see the Supplementary Information), establishing that the stereochemistry of diamine N1*, not that of bisphosphine P, determines the stereochemistry of the amide. Control experiments established that, in the absence of copper, light, or bisphosphine P, essentially no C–N coupling is observed (<1% yield; Fig. 2, 1). Furthermore, in the absence of diamine N1*, only a small amount of the coupling product is obtained (7% yield), pointing to cooperativity between diamine N1* and bisphosphine P in achieving C–N bond formation.

Under our standard conditions, the corresponding alkyl iodide undergoes C–N coupling with similar yield and e.e. (86% yield, 92% e.e.) as the bromide (Fig. 2, 1), whereas the corresponding chloride is essentially unreactive (<5% conversion). From a practical point of view, it is worthwhile to note that the reaction is not highly air- or moisture-sensitive, proceeding well in the presence of 0.1 mL of air or of one equivalent of water (>85% yield, 92% e.e.). On a gram scale, C–N coupling occurs with slightly diminished yield and similar enantioselectivity (85% yield and 92% e.e.). The combination of high yield and high e.e., while using only a small excess of the racemic electrophile (1.2 equiv.), establishes that a stereoconvergent coupling, not a simple kinetic resolution, of the electrophile is operative.

A diverse array of amides serve as useful nucleophiles in this photoinduced, copper-catalyzed enantioconvergent N-alkylation by unactivated electrophiles (Fig. 2). For example, a variety of aromatic amides, including para-, meta-, and ortho-substituted compounds, are suitable nucleophiles, providing secondary amides in good yield and enantioselectivity (1–8). Furthermore, heterocyclic amides (bearing a furan, a thiophene, or an indole; 9–11) and aliphatic amides, including sterically demanding nucleophiles (12–15), are effective coupling partners. Moreover, this asymmetric N-alkylation can be applied to more complex amides (16–19); for couplings of chiral primary amides, the stereochemistry of the chiral catalyst, rather than the stereochemistry of the chiral nucleophile, primarily determines the stereochemistry of the product (17–19). In contrast to the classical approach to the synthesis of such chiral secondary amides (N-acylation in Fig. 1a; exception: kinetic resolution/acylation of amines¹⁸), this method enables a new, convergent strategy wherein stereochemical control and C–N bond formation are accomplished in a single step.

The scope with respect to the phosphonyl-substituted unactivated secondary electrophile is also fairly broad. For example, the R substituent can vary in steric demand from Me to isobutyl (Fig. 3, 20–22). Furthermore, the method is compatible with an array of functional groups, including trifluoromethyl, an acetal, an alkene, an alkyne, and a ketone (23–28); moreover, through additive studies we have determined that the presence of an unactivated alkyl chloride, aniline, aryl triflate, benzofuran, benzothiophene, epoxide, N-methylindole, nitrile, tertiary amide, or thioether does not significantly impact the yield or the e.e. of the coupling to generate 1 in Fig. 2 (>80% yield and >90% e.e. of product, along with >95% recovery of the additive; see the Supplementary Information). Finally, the phosphonyl group can bear a range of substituents other than OEt (29–31).

Fig. 3. An array of alkyl bromides serve as electrophilic coupling partners.

Couplings were generally conducted using 0.5 mmol of the amide, and N1* was used as the diamine, unless otherwise noted. All data represent the average of two experiments. The percent yield represents purified product. a: N2* as the diamine, no Cs₂CO₃, 2-Me-THF as solvent. b: N2* as the diamine. c: electrophile (1.5 equiv.), Cu(CH₃CN)₄PF₆ (15 mol%), P (5 mol%), N1* (20 mol%), K₃PO₄•H₂O (1.5 equiv.; in place of Cs₂CO₃), 10 °C. Ar¹ = p-(F₃C)C₆H₄, DG = directing group, Ac = acetyl, Bn = benzyl, TIPS = triisopropylsilyl.

Furthermore, the same approach can be applied to photoinduced, copper-catalyzed enantioconvergent N-alkylations of amides by other classes of racemic unactivated secondary alkyl electrophiles. Thus, in addition to a phosphonyl group (Fig. 2 and 20–31 in Fig. 3), a diverse array of other Lewis-basic functional groups, including an amide, ester, ketone, sulfone, sulfonamide, and phosphine oxide, enable the often-challenging differentiation between two alkyl substituents to provide good enantioselectivity in the nucleophilic substitution reaction (32–53 in Fig. 3 and 54–64 in Extended Data), leading to derivatives of γ-amino acids, 1,2-diamines, and the like^28–,29,30. Whereas ligand-directed asymmetric catalysis is now well-established, the ability to utilize such a wide array of directing groups for a particular reaction is noteworthy³¹.

We have pursued mechanistic studies to elucidate the role that the various ligands play in this photoinduced, copper-catalyzed enantioconvergent N-alkylation process. Our working hypothesis is that, in contrast to all previous examples of enantioconvergent N-alkylations (all of which utilize activated electrophiles^{8–,9,10,11,12,13}), more than one catalytic cycle is involved in this new process.²¹ Specifically, we believe that one copper complex serves as a photocatalyst, cleaving the relatively strong (unactivated) C–Br bond, and a second copper complex effects enantioselective C–N bond formation (Fig. 4a).

Fig. 4. Mechanism.

a, Outline of a possible catalytic cycle. b, Redox potentials (vs Fc⁺/Fc). c, Stern–Volmer quenching of [PCu^I(OPh)]* by 2: irradiation at 350 nm, emission at 496 nm. d, EPR studies: X-band spectra (9.4 GHz, 77 K). e, Investigation of an organic radical (R•) as an intermediate: TEMPO trapping and radical-clock experiments. f, Reactions of enantioenriched electrophile. Ar = p-(F₃C)C₆H₄, TEMPO = 2,2,6,6-tetramethyl-1-piperidinyloxy.

With respect to the left-hand catalytic cycle (Fig. 4a), our observation that, for the N-alkylation illustrated in Fig. 2 (1), the electrophile is consumed in the absence of diamine N1* (>99% conversion of the electrophile; 7% yield of the product), but not in the absence of bisphosphine P (<2% conversion of the electrophile), is consistent with a P-bound copper complex serving as the photocatalyst. With respect to the right-hand catalytic cycle (Fig. 4a), our observation that the stereochemistry of diamine N1*, not that of bisphosphine P, determines the stereochemistry of the product is consistent with an N1*-bound copper complex being responsible for C–N bond construction.

Regarding the left-hand catalytic cycle, we determined that two representative unactivated electrophiles have reduction potentials of −2.3 to −2.4 V (Fig. 4b: 65 and 66; all redox potentials are referenced to Fc⁺/Fc), placing them beyond the reach of conventional ruthenium/iridium photocatalysts³² but within the reach of Cu^I/bisphosphine complexes³³. Two PCu^IX complexes that might plausibly serve as photocatalysts under our conditions are PCu^I(OPh) and PCu^I(amidate). We synthesized each complex via the treatment of CuCl with bisphosphine P and X⁻ (X = OPh, amidate; amidate = p-(F₃C)C₆H₄CONH) and then crystallographically characterized them (see the Supplementary Information).

Our observations point to PCu^I(OPh) complex as the more likely photocatalyst under our reaction conditions. Thus, during a coupling, PCu^I(OPh) accounts for ~35% of the copper-containing species that are present, whereas PCu^I(amidate) is not detected (see the Supplementary Information). Furthermore, the estimated excited-state reduction potential of PCu^I(OPh) (−2.8 V), but not of PCu^I(amidate) (−2.3 V), is lower than that of the electrophiles (−2.3 and −2.4 V). The excited state of PCu^I(OPh) has a relatively long lifetime (4.6 μs), and a Stern–Volmer study established that luminescence of the excited state of PCu^I(OPh) (emission at 496 nm) is readily quenched by electrophile 66 with a Stern−Volmer constant of 189 M⁻¹ and a quenching rate constant of 4.1 × 10⁷ M⁻¹s⁻¹ (Fig. 4c); in contrast, the excited state of PCu^I(amidate) is not effectively quenched by this electrophile (see the Supplementary Information). Alternatively, the reaction of the excited state of PCu^I(OPh) with the electrophile could proceed by an inner-sphere electron-transfer pathway (halogen-atom transfer).³⁴

The two catalytic cycles of the proposed coupling pathway intersect at the reaction of D and G to generate B and E (Fig. 4a). Two of the plausible mechanisms by which this process might occur are via electron transfer or via ligand exchange. To corroborate our intuition about the likely feasibility of electron transfer between complexes D and G, we carried out DFT calculations (BP86, def2-TZVP, SMD(THF) level of theory), which predict redox potentials (E⁰ for Cu^I/Cu^II) of 1.3 V and 1.2 V for PCu^II(OPh)₂ and PCu^II(OPh)Br, respectively, and −0.6 V for both (N2*)Cu^I(OPh) and (N2*)Cu^I(amidate) (Fig. 4b). Taken together, these computed redox potentials support the viability of electron transfer from (N2*)Cu^IX (G) to PCu^IIX₂ (D).

Alternatively, D and G could generate B and E via ligand exchange. To explore this possibility, we treated PCu^I(OPh) with Magic Blue (tris(4-bromophenyl)ammoniumyl hexachloroantimonate, a one-electron oxidant; 1.0 equiv.) in the presence of N2* (1.0 equiv.) and NaOPh (10 equiv.) in 2-Me-THF. An EPR spectrum taken after 2 minutes at room temperature (experiment 1 in Fig. 4d) revealed a spectrum consistent with (N2*)Cu^II(OPh)₂ (see below). This observation is consistent with rapid ligand exchange (P → N2*) upon oxidation of Cu^I to Cu^II, with the latter preferring the harder diamine ligand to the softer bisphosphine ligand. Thus, either or both of these pathways for the formation of B and E from D and G may be operative.

We carried out a series of EPR spectroscopic studies in order to gain insight into the Cu^II species that are present under our coupling conditions. Treatment of CuCl₂ with N2* (1.0 equiv.) and NaOPh (2.0 equiv.) in 2-Me-THF afforded a blue solution with the EPR spectrum illustrated in Fig. 4d (experiment 2), which we assign to (N2*)Cu^II(OPh)₂. Furthermore, reaction of CuBr₂ with N2* (1.0 equiv.) and NaOPh (1.0 equiv.) furnished a blue solution with the EPR spectrum shown in Fig. 4d (experiment 3), which we attribute to (N2*)Cu^II(OPh)Br. An EPR spectrum of a coupling reaction after four hours at −5 °C appeared to be an ~1:1 mixture of the species shown in experiments 2 and 3 (Fig. 4d, experiment 4; for a simulation, see the Supplementary Information); spin quantification of the EPR-active copper species indicated that ~40% of the copper in the reaction mixture is present as these two Cu^II complexes. ESI–MS analysis of a reaction mixture confirmed the presence of (N2*)Cu^II(OPh)₂ during the coupling process (see the Supplementary Information).

With regard to the proposed right-hand catalytic cycle (Fig. 4a), nucleophilic substitution of one or both of these Cu^II/N2* complexes (E) by the amidate anion leads to (N2*)Cu^IIX(amidate) (F), which reacts with organic radical R• in an out-of-cage process, via coordination of the directing group to Cu^II followed by C–N bond formation, to furnish the desired coupling product and (N2*)Cu^IX (G). Consistent with the hypothesis that C–N bond formation proceeds via an out-of-cage reaction of R•, we observe no C–N coupling (<1%) in the presence of TEMPO, a radical trap; instead, a TEMPO adduct of R• can be detected (20% yield; Fig. 4e; the sodium salt of TEMPO does not react with alkyl bromide 2 under these conditions (see the Supplementary Information)).

A radical-cyclization study (Fig. 4e) provides further support for the intermediacy of an organic radical and for the out-of-cage pathway for C–N bond formation that is depicted in Fig. 4a. Thus, under our standard reaction conditions, alkyl bromide 67 reacts with a primary amide to predominantly generate cyclized product 68 with 3:1 diastereoselectivity (<5% of the acyclic coupling product). The diastereoselectivity matches that observed in the n-Bu₃SnH-mediated reductive cyclization of the same electrophile, consistent with a common intermediate (a secondary organic radical) in the two processes. The rate constant for the cyclization of related radicals is ~1×10⁵ s⁻¹ at 25 °C³⁵, which is much slower than typical diffusion rates (generally >10⁸ s⁻¹)³⁶. These observations are thus consistent with the formation of an organic radical and its out-of-cage coupling to produce a C–N bond, as depicted in the mechanism outlined in Fig. 4a.

Our studies of couplings of enantiomerically enriched electrophiles have also been informative. Under our standard reaction conditions, the two enantiomers of electrophile 69 react at comparable rates in the presence of (R,R)–N1*, and the e.e. of the product is independent of the original stereochemistry of the electrophile (Fig. 4f, Entries 1 and 2). Slow racemization of the electrophile occurs during the photoinduced C–N coupling (Entries 1–3), but racemization ceases when irradiation ceases (Entry 3 vs Entry 4), suggesting that racemization is proceeding through a photoinduced pathway, e.g., reversible copper-mediated C–Br bond cleavage. Similar racemization occurs in the absence of N1*, consistent with Cu^II complex D, not E or F, serving as the Br donor (Entry 5).

In conclusion, by exploiting photoinduced copper catalysis, the first enantioconvergent substitution reactions of unactivated alkyl electrophiles by nitrogen nucleophiles (specifically, primary amides) have been accomplished, thereby enabling a new, convergent strategy for the asymmetric synthesis of secondary amides wherein C–N bond formation and control of stereochemistry are achieved in a single step. The method exploits three classes of ligands, which assemble the necessary catalysts in situ: a bidentate phosphine and a phenoxide bind to Cu^I to produce a photocatalyst with a sufficient excited-state lifetime and reduction potential to activate the comparatively strong C–Br bond of the unactivated electrophile, and a bidentate chiral diamine binds to Cu^II to effect enantioselective C–N bond formation. Efforts to exploit copper, an earth-abundant metal, to address other unsolved challenges in asymmetric catalysis are underway.

Extended Data

Extended Data Fig. 1 |. Continued scope of alkyl bromides serving as electrophilic coupling partners.

Couplings were generally conducted using 0.5 mmol of the amide, and N1* was used as the diamine, unless otherwise noted. All data represent the average of two experiments. The percent yield represents purified product. a: electrophile (1.5 equiv.), Cu(CH₃CN)₄PF₆ (15 mol%), P (5 mol%), N1* (20 mol%), K₃PO₄•H₂O (1.5 equiv.; in place of Cs₂CO₃), 10 °C. Ar¹ = p-(F₃C)C₆H₄, DG = directing group, Bn = benzyl.

Data availability

The data that support the findings of this study are available within the paper, its Supplementary Information (experimental procedures and characterization data) and from the Cambridge Crystallographic Data Centre (https://www.ccdc.cam.ac.uk/structures; crystallographic data are available free of charge under CCDC reference numbers CCDC 2055329-2055338.