Green Light Promoted Iridium(III)/Copper(I)-Catalyzed Addition of Alkynes to Aziridinoquinoxalines Through the Intermediacy of Azomethine Ylides

Abstract

This manuscript describes the development of alkyne addition to the aziridine moiety of aziridinoquinoxalines using dual Ir(III)/Cu(I) catalytic system under green light-emitting diode (LED) photolysis (λ_max = 525 nm). This mild method features high levels of chemo- and regioselectivity and was used to generate 30 highly functionalized substituted dihydroquinoxalines in 36-98% yield. This transformation was also carried asymmetrically using phthalazinamine-based chiral ligand to provide 9 chiral addition products in 96:4 to 86:14 e.r. The experimental and quantum chemical explorations of this reaction suggest a mechanism that involves Ir(III)-catalyzed triplet energy transfer followed by a ring-opening reaction ultimately leading to the formation of azomethine ylide intermediates. These azomethine intermediates undergo sequential protonation/copper(I) acetylide addition to provide the products.

Graphical Abstract

Introduction

Due to their high occurrence in nature and unique chemical properties, nitrogen-containing heterocycles are of great importance to various areas of drug discovery, AgroSciences, and material science.^[1] Nitrogen-containing heterocycles are present in essential biomolecules such as DNA, RNA and proteins. Recent analysis of the US Food and Drug Administration (FDA) approved pharmaceuticals highlights the importance of the nitrogen containing heterocyclic molecules as 59% of unique drugs contained at least one nitrogen heterocycle, and the average number of nitrogen atoms per drug was found to be 2.3 N/drug.^[1b] The continuously increasing interest to nitrogen-containing heterocyclic compounds has fueled the development of many creative synthetic methodologies to access various nitrogen-containing scaffolds, particularly in their enantiopure forms.^[2] These studies have been enabled by the recent advances in asymmetric catalysis, photocatalysis, transition metal catalysis and organocatalysis as well as development of new mechanistic concepts for harnessing the reactivity of highly reactive intermediates.^[3] Among various heterocycles, significant efforts have been focused in achieving synthesis of substituted tetrahydroquinoxalines (Figure 1). These heterocyclic motifs are present in a variety of bioactive compounds such as molecules 1–3 (Figure 1A).^[4] The standard approaches to chiral tetrahydroquinoxalines involve hydrogenation or transfer hydrogenation of quinoxalines, which poses limitations to the types of products that could be accessed through these reductive methods.^[5] At the same time, recent advances in catalysis enabled asymmetric synthesis of complex tetrahydroquinoxalines that are not easily accessible through dearomative reduction of quinoxalines (Figures 1B-D).^[6]

Figure 1.

Summary of the recent approaches to chiral tetrahydroquinoxalines.

The Nagorny group has long-standing interests in developing new catalytic methods for the synthesis of chiral heterocyclic molecules of medicinal chemistry value.^[7] Our recent studies highlighted the use of both Ir(III)-based catalysts^[7c] and Chiral Phosphoric Acids (CPAs)^[7b,e] for the synthesis of chiral nitrogen-based heterocycles. Among various chiral heterocyclic systems that we have explored as substrates for the CPA-catalyzed transfer hydrogenation are aziridinoquinoxalines 6 (Figure 2B).^[7e]

Figure 2.

Summary of the prior work and proposed mechanism for the photochemical activation of aziridinoquinoxalines.

These unusual heterocycles could be generated in one step from o-phenylenediamines (4) by their condensation with α,β-dibromoketones (5).^[7e,8] Previously, we demonstrated that aziridinoquinoxalines (6a) may undergo chemoselective C=N addition upon exposure to nucleophiles (cf. Figure 2B and SI-II-5)^[7e] while the aziridine moiety stays intact. However, this manuscript summarizes our studies that demonstrates that this reactivity could be completely switched under the photochemical conditions and that the aziridine moiety could be selectively reacted with copper(I) acetylides^[9-11] in the presence of the imine (C=N) group to provide previously inaccessible dihydroquinoxalines. This reaction between alkynes and aziridinoquinoxalines (6) is promoted under mild photochemical conditions using green LED light (λ_max=525 nm) and is co-catalyzed by Ir(III) and Cu(I) catalysts (Figure 1E). In addition, the described studies demonstrate that this transformation could be carried asymmetrically using (R)-(+)-4-[2-(diphenylphosphino)-1-naphthalenyl]-N-[(S)-1-phenylethyl]-1-phthalazinamine ((S,R)-N-PINAP) as the ligand (L2).^[10d,12]

The presence of two electrophilic sites makes aziridinoquinoxalines to be versatile precursors for the synthesis of highly functionalized quinoxalines. While these species have not been extensively explored, prior work suggests that both the C=N imine moiety and the aziridine ring may participate in reactions (Figure 2A).^[8] The prior studies exploring the reactivity of 6a and related compounds suggest that these compounds may undergo thermal 4π-electrocylic ring-opening reaction to provide azomethine ylides that could be intercepted in various [3+2] cycloaddition reactions with alkenes, alkynes, aldehydes and diazo dicarboxylates.^[8a] It is noteworthy that 4π-electrocylic ring-opening of 6a requires significantly lower temperatures than the corresponding reaction of an inactivated aziridine ring (110-150 °C vs. 200-250 °C).^[13] At the same time, our studies demonstrate that under normal conditions aziridinoquinoxalines such as 6a may undergo polar reactions with nucleophiles at the C=N group, rather than aziridine, moiety. Thus, when 6a/6d was exposed to copper(I)-catalyzed allylation reaction with allyl silane, C=N addition products 7a/7d were obtained in excellent yield (cf. SI-II-5). Similarly, when exposed to standard transfer hydrogenations with Hantzsch ester and (R)-TRIP catalyst, (±)-6a underwent chemo- and enantioselective hydride addition to the C=N group to provide chiral diastereomeric products 8a and 9a (Figure 2B).^[7e]

One of the interesting physicochemical properties of aziridinoquinoxalines such as 6a is photochromism.^[14] Thus, it was noted that 6a may undergo reversible change in color from yellow to red upon exposure to sun light.^[8c] While this phenomenon was attributed to the reversible formation of strongly colored azomethine ylide in solid state, photochemical properties, and reactivity of 6 and related aziridinoquinoxalines under photochemical conditions have not been explored in details.^[15] Based on the prior work^[16] and our computational and experimental results (vide infra), we surmised that photoexcitation of 6 in its ground state (S₀) may produce its triplet state T1, which may undergo an aziridine ring fragmentation to intermediate T2 followed by a low-barrier Intersystem Crossing (ISC) step leading to the azomethine ylide formation (Figure 2C). The ylide, generated under mild conditions, may undergo polar reactions with electrophiles and nucleophiles to provide functionalized tetrahydroquinoxalines that are otherwise not accessible through other methods.^[17]

Results and Discussion

Based on the mechanistic considerations presented in Figure 2, we surmised that azomethine ylides derived from aziridinoquinoxalines (6) might be intercepted with copper(I) acetylides^[10j,18] through the catalytic cycle summarized in Scheme 1. The photoactivation of aziridinoquinoxalines is expected to provide azomethine ylide 10, which is expected to be protonated with ammonium salts generated during the formation of copper(I) acetylide. The resultant quinoxalinium ion 11 would be trapped with copper(I) acetylide to produce the desired product 13 and regenerate Cu(I) salt, which would reengage in the catalytic cycle.

Scheme 1.

Mechanistic cycle for the Cu(I)-catalysed photochemical addition of acetylenes 12 to aziridinoquinoxalines 6.

(1)

(2)

Our studies commenced with exploring the proposed in Scheme 1 catalytic cycle under the thermal activation conditions that was previously observed to convert substrate 6a into the corresponding ylides (Eq. 1 and 2). Initial reaction optimization involved selecting the optimal reaction temperature, stoichiometry, and catalyst loading (cf. Table SI-1). The efficient formation of product 13a (93% yield) was observed under the optimized conditions after heating the substrate for 18 h at 80 °C using 5 mol% of Cu(I) iodide as the catalyst and diisopropylethylamine (DIPEA) as the base. Further elevation of the temperature led to shorter reaction times; however, the formation of side-product and decomposition impurities was observed lowering the product yield. Unfortunately, our attempts to extend this method to other substrates such as the t-butyl-substituted aziridinoquinoxalines 6d, did not lead to the formation of the desired product, even at 110 °C. This, coupled with only moderate enantioselectivities observed for 13a with chiral ligands under thermal activation (cf. Tables SI-2 and SI-3) prompted us to pursue the photochemical activation of 6 next.

The possibility of photochemical pathway was evaluated by density-functional theory (DFT) computations (B3LYP/6-31+G** for TS search and structure optimization and SMD(Toluene)/wB97x-D3/cc-pVTZ for electronic energy calcualtion, cf. SI-IV-1) where the Growing String Method (GSM)^[19] was used to compute the ring-opening energy barriers (Table 1) in both the ground singlet state (S₀) as well as the excited triplet state (T₁). For aziridine 6a, the activation energy was found to be ΔG^ǂ = 32.5 kcal/mol for ring-opening in the S₀-state, which is in good agreement with the observed reactivity under the thermal activation conditions (Eq. 1). However, the computed energy barrier for ring-opening in T₁ state is only 4.7 kcal/mol, and the formation of ylide 6a (I₁-T₁) in triplet state is also more thermodynamically favored by 29.3 kcal/mol. The triplet state for compound 1a is photochemically accessible at 48.4 kcal/mol above S₀. Results from these simulations in Table 1 also explain the low reactivity of alkyl-substituted aziridines such as 6b in thermally activated ring-opening (Eq. 2). The predicted thermal activation energy barrier from the S₀ state is much higher for the alkyl-substituted substrates such as 14 and 16 (about 40.9 and 42.8 kcal/mol, correspondingly); however, the photochemically induced ring-opening of alkyl aziridines is more feasible from T₁ state due to significantly lower barriers (10.1 and 6.1 kcal/mol, correspondingly). It is noteworthy that changing the C₂-phenyl substituent in the aziridinoquinoxaline ring to a methyl group (models 15 and 16) results in significant increase of the T₁-state energies by ~9 kcal/mol due to the weaker biradical stabilization. Finally, replacing the imine (C=N) chromophore with the CH─NH moiety present in 8a renders the photochemical process infeasible due to the highly unstable triplet state 8a-T₁ (82.9 kcal/mol).

Table 1.

Computed Energy Barriers for Photoinduced Ring-Opening of aziridinoquinoxalines 6a-6d and 8a.^[a]

E (kcal/mol)	6a	14	15	16	8a
ΔGǂ(S0)	32.5	40.9	34.5	42.8	44.0
I1-S0	8.8	6.2	8.3	6.4	29.2
T1	48.4	48.3	57.3	57.5	82.9
ΔGǂ(T1)[b]	4.7 (53.7)	10.1 (58.4)	1.0 (63.6)	6.1 (63.6)	4.4 (87.3)
I1-T1	19.1	28.2	19.1	29.4	46.6

^[a]The provided energies were computed using DFT (B3LYP/6-31+G**: optimization, entropy corrections; SMD(Toluene)/wB97x/cc-pVTZ: electronic energy with solvation) and Growing String Method (GSM, DFT B3LYP/6-31+G**) (cf. SI-IV-1 for additional details).

^[b]The values in parenthesis represent the relative values of the computed transition states to the ground S₀ state of corresponding aziridinoquinoxalines.

Based on the modeling results of Table 1, our subsequent studies focused on exploring the direct excitation of aziridinoquinoxaline 6a with Blue LED (λ_max=440 nm) (Table 2). Achiral pyridine-bis-oxazoline (PyBox) ligand was used to solubilize copper(I) iodide. After 6 h, complete consumption of 6a was observed in both toluene and THF leading to the ylide decomposition product 2-phenylquinoxaline 17 (entries 1 and 2). Introducing Ir(III) photocatalysts (entries 3 and 4) didn’t significantly improve the reaction outcome, and only with Ir(4-CF₃ppy) the formation of minor amounts of 13a was observed along with 17 in 22:78 ratio.^[20]

Table 2.

Optimization of photochemical addition of alkyne 12a to aziridinoquinoxalines 6a.^[a]

entry	λ (nm)	PC	Solvent	time (h)	Conv. (%)[b]	13a:17
1	440	–	PhMe	6	>95	0:100
2	440	–	THF	6	>95	0:100
3	440	Ir(dF-ppy)3	THF	6	>95	0:100
4	440	Ir(4-CF3-ppy)3	THF	6	>95	22:78
5	525	–	THF	24	<5	–
6	525	Ir(4-CF3-ppy)3	THF	24	68	64:36
7	525	–	PhMe	24	26	84:16
8	525	Ir(4-CF3-ppy)3	PhMe	24	>95	93:7
9	525	Ir(ppy)3	PhMe	24	>95	84:16
10	525	Ir(ppy)2(4,4’-dtbbpy)	PhMe	24	11	65:35
11	525	[Ir(dF(CF3)ppy)2 (dtbbpy)]PF6	PhMe	24	21	–
12	525	Eosin Y	PhMe	24	0	–

^[a]The reaction was performed on 0.1 mmol scale using 0.05 M solution in the corresponding solvent as described in SI-II-2.

^[b]Determined by ¹H NMR analysis.

To understand the reasons behind the nonproductive substrate 6a decomposition and to identify better conditions for the photochemical activation, time-dependent DFT calculations (TD-DFT, SMD(Toluene)/wB97x-D3/cc-pVTZ (SG-2) used to optimize geometries and excited states analysis) were performed to characterize excited states of the compound 6 and their ylides (Scheme 2A and SI-IV-5). For compound 6a, the first excited state S₁ was found to have a vertical excitation energy of 92.1 kcal/mol (313 nm) and oscillator strength 0.35. Geometry optimization gives the lowest point of the S₁ PES at 79.7 kcal/mol (359 nm). This implies that direct excitation of 6a is possible with the Kessil Blue LED source (Figure SI-1). Computational results are also in good agreement with the experimental UV-VIS spectra of 6a (Figure SI-3) and fluorescence spectroscopy measurements that indicated emission at 337 nm and absorption at 305 nm for the S₀-S₁ transition (Scheme 2B, Figure SI-10).

Scheme 2.

Computational and spectroscopic studies to optimize the photochemical ring-opening of 6a and substrate scope exploration studies.

[a] Time-dependent DFT calculations (TD-DFT) were performed using SMD(Toluene)/wB97x-D3/cc-pVTZ(SG-2) (for both structure optimization and excited state analysis, cf. SI-IV-5 for additional details). [b] The fluorescence excitation spectroscopy (green – fluorescence spectrum excited at 305 nm, blue – absorption spectrum for fluorescence detected at 337 nm) of 50 μM solution of 6a in THF; [c] The fluorescence excitation spectroscopy (green – fluorescence spectrum excited at 397 nm, blue – absorption spectrum for fluorescence detected at 515 nm) of 50 μM solution of 6a in THF; [d] The reactions were carried on 0.1 mmol scale described in SI-II-2. Reactions leading to 13d-13z were performed without PyBox Ligand, and the formation of substrates 13a-13c, 13aa-13ab were carried with 6 mol% of PyBox.

Singlet-excited compound 6a (S₁) may undergo an intersystem crossing to its triplet state 6a (T₁), which has an energy of 50.4 kcal/mol (567 nm). The T₁ state is close to detected fluorescence emission peak at 515 nm (see Scheme 2C, Figure SI-11). Therefore, the T₁ state could be responsible for photoactivation of 6a under the tested conditions. In particular, aziridine ring opening transforms the initial triplet geometry of 6a (T₁) (50.4 kcal/mol) to a much lower energy structure (I₁-T₁) (12.1 kcal/mol), as shown in Scheme 2A. I₁-T₁ then can nonradiatively relax to the ring opened, ground I₁-S₀ state. From that point, TD-DFT analysis of the ylide intermediate 6a (I₁-S₀) indicated an S₀-S₁ excitation at 73.8 kcal/mol (441 nm) with a substantial oscillator strength of 0.92.^[21] The ylide’s S₀-S₁ energy gap is well within reach of the Blue LED, and its high oscillator strength makes the ylide intermediate 6a (I₁-S₀) photoactive. As a result, low yields of the desired product 13a can be explained by an undesired pathway that involves the excitation of intermediate 6a (I₁-S₀) by Blue LED at 440 nm, which leads to its decomposition.

In general, all other model substrates 14-16 (cf. SI-Scheme SI-11) produce photoactive ylides with accessible S₀-S₁ transitions. The photoexcited states are reachable due to the greater extent of conjugation in the open forms of aziridines. Methyl substituents at the C₁-position of aziridines 14 and 16 do not have any impact on the energies of the triplet states T₁, while making the corresponding ylides 14 (I₁-S₀) and 16 (I₁-S₀) less stable and photoactive due to the absence of the additional aromatic conjugation such as the one that is present in compound 6a. It is worth noting that the ylide derived from the C₂-phenyl containing substrate 14 absorbs shorter wavelength due to the alkyl substituent in the aziridine ring and should be more stable to decomposition under the photochemical conditions. In contrast, switching the phenyl substituent at the C₂-position of the quinoxalinium core to the methyl group (substrates 15 and 16) destabilizes both excited singlet (S₁) and triplet (T₁) states of 15 and 16. In summary, avoiding excitation of the ylide intermediates under the reaction conditions should eliminate decomposition pathways. This can be achieved by switching to light sources with a longer wavelength (>475 nm) and establishing energy triplet transfer (EnT) through a photocatalyst that selectively promotes the S₀-T₁ transitions.

Guided by the considerations above, we implemented a less energetic light source by switching to Green LED (525 nm, emission spectrum Figures SI-2) to avoid the excitation of the ylide intermediates (cf. Table 2, entries 5-12). As expected, we do not observe any reaction of 6a in THF when exposed to Green LED alone (entry 5) as the light with λ_max = 525 nm cannot effectively excite aziridinoquinoxalines such as 6a. Introducing Ir(III) photocatalyst Ir(4-CF₃-ppy)₃ allowed us to achieve higher conversion in THF (Entry 6) with much better 64:36 product 13a to 17 ratio. Reaction in toluene without photocatalyst was also not effective albeit provided some observable product (entry 7). However, the reaction progression was significantly improved when Ir(4-CF₃-ppy)₃ was used as the photocatalyst, and the reaction reached full conversion in 24 hours resulting in almost pure product 13a (Entry 8). Further photocatalyst screening did not lead to improvement (entries 9-12) although the observed results were in good agreement with the proposed EnT mechanism. Ir(ppy)₃ both matches the triplet state energy and has a relatively long T₁ lifetime, and this catalyst was found to have a similar catalytic activity to Ir(4-CF₃-ppy)₃ (entry 9). Ir(ppy)₂(4,4’-dtbbpy) photocatalyst in entry 10 has too low triplet energy (49.2 kcal/mol) while [Ir(dF(CF₃)ppy)₂(dtbbpy)]PF₆ catalyst in entry 11 is probably mismatched with the triplet gap of compound 6a. Eosin Y is similar to Ir(4-CF₃ppy)₃ in terms of its triplet state energy,^[20] but it has shorter excited state lifetime to promote EnT mechanism. Using organic dye, Eosin Y, (entry 12) with the absorption at 520 nm that matches well the employed light source with λ_max=525 nm neither lead to the formation of product nor to decomposition. This may be attributed to the low triplet state energy of Eosin Y or to low solubility of Eosin Y in toluene.

With the optimized protocol in hand, the substrate scope evaluation was performed next by exploring the variations in alkynes 12 and aziridinoquinoxalines 6 (Scheme 2D). As expected, the introduction of the electronwithdrawing substituents to the phenyl ring attached to the aziridine moiety had a minor effect on the reaction yield, and substrates 13b and 13c were obtained in 65% and 54%, correspondingly. Due to much lower energy barrier for ring-opening in T₁ triplet state, tBu-substituted aziridinoquinoxaline 6d, previously unreactive under the thermal activation conditions, was successfully converted to product 13d in 96% yield using the green LED/Ir(III) activation protocol. In general, compound 6d was significantly more stable under the photochemical conditions than 6a (that provided product 13a in only 70% yield) and other aryl-substituted aziridines such as 13b and 13c. Reaction proceeds with excellent yields (>90%) with a wide range of alkynes including various silyl-protected alkynes 12e-12g, phenylacetylene 12h, ethoxyacetylene 12i, alkyl substituted acetylene 12j and 12k, and cyclohexyl-substituted 13m. Alkynes with photolabile groups such as cyclohexene group (13n) and cyclopropyl group (13l) that can form radicals under photochemical excitation are well tolerated under the green LED radiation with λ_max=525 nm. Presence of the potentially nucleophilic groups such as amines and hydroxyls (13o–13r) did not lead to the formation of side products of competitive addition to quinoxalinium cations demonstrating high catalytic differentiation of Cu(I) catalysis towards alkyne addition. Similarly, the reaction proceeded smoothly with propargyl acetate to provide product 13s in 88% yield. The variation in the C₂-substituent of aziridinoquinoxalines was explored next (13t-13aa). The introduction of both electron withdrawing (13u-13x) and electron donating (13t) substituents at the para-position of the C₂- phenyl ring did not impact the yield and the corresponding adducts were obtained in 82-96% yield. Similarly, the C₂-naphthyl-substituted product 13y was formed in excellent 88% yield. At the same time, the 2-furanyl substituted product 13y was isolated in only 36% yield due to significant decomposition presumably due to the side-reactions of furan under the photochemical conditions. In the case of C₂-thiophen substituted substrate 6l, reaction proceeded without any signs of decomposition with 51% conversion by NMR resulting in 44% yield of 13z and recovered starting material. Low conversion can be attributed to Ir(III) photocatalyst poisoning by the sulfur heterocycle. In addition, pyridine-substituted substrate 13aa was also obtained in 54% yield without any signs of the decomposition of the starting material. The reaction rate retardation due to the copper(I) complexation with two molecules of the starting material 6d was noted in this case. Finally, the substrates containing the cyclohexyl- and ethyl-substituents at the aziridine ring, 6m and 6n were subjected to reaction with acetylene 12a to provide products 13ab and 13ac in 67% and 50% yield, correspondingly.

After establishing the optimal conditions for green LED-induced alkyne addition to aziridinoquinoxalines, our subsequent studies focused on developing the asymmetric variant of this reaction (Table 3 and SI-II-3).^[10,11]

Table 3.

Development of the enantioselective alkyne addition to substrate 6a.^[a]

entry	solvent	ligand	substrate	conversion /yield (%)[b]	13:17	e.r.
1	PhMe	L4	6a	>95	93:7	72:28
2	PhMe	L5	6a	>95	89:11	55:45
3	PhMe	L6	6a	>95	90:10	63:37
4	PhMe	L7	6a	>95	67:33	57:43
5	PhMe	L8	6a	>95	83:17	82:18
6	PhMe	L9	6a	>95	87:13	82:18
7[c]	PhMe	L9	6a	>95	78:22	86:14
8[c]	PhMe	L8	6a	>95/67	81:19	87:13
9	PhMe	L2	6a	>95/28	55:45	93:7
10	THF	L2	6a	>95/25	49:51	95:5
11	THF	L2	6d	>95/81	98:2	93:7

^[a]The reactions were performed on 0.1 mmol scale using 0.05 M solution in specified solvent as described in SI-II-2.

^[b]The conversion was determined by ¹H NMR analysis of the crude reaction mixture.

^[c]These reactions were performed with 5 mol% of copper(I) iodide and 2.5 mol% of L9 or L8.

Various PyBox ligands (L4-L9) demonstrated low-to-moderate enantioselectivities (entries 1-6), while the best enantiomeric ratio (e.r.) of 82:18 was obtained using either tertbutyl (L8) or adamantly (L9) derivative of Py-Box (entries 5-8). Further reaction optimization by reducing the catalytic loading of ligand L8 to 2.5 mol% (Cu/L = 2:1 ratio) helped to increase the e.r. to 87:13 for product 13a (entry 8). Excellent enantioselectivity (95:5) can be obtained using (S,R)-N-PINAP ligand L2^[12] (entries 9-10) albeit product 13a was isolated in only 25% yield due to significant decomposition to 2-phenylquinoxaline 17. As aziridinoquinoxaline 6d previously exhibited higher stability than 6a, we tested it with ligand L2 in THF and the resultant isolated product 13d was formed in 81% yield and 93:7 er. Further attempts to optimize the reaction conditions by evaluating copper(I) chloride, copper(I) or copper(II) trifluoromethanesulfonates as the catalysts did not lead to improved yields or selectivities (cf. Table SI-5). Therefore, the conditions for the formation of 13d with PINAP ligand L2 (entry 11) were selected for further evaluation (cf. Scheme 3). Using the optimized conditions with ligand L2 for the formation of chiral 13a and 13d, we explored the scope of this transformation using tert-butyl-substituted aziridinoquinoxalines next (Scheme 3). Changing trimethylsilyl- (TMS-) -substituted alkyne 12a to bulkier triethylsilyl- (TES-) protected alkyne 12b resulted in more enantioselective formation of the product 13e (94.5:5.5 e.r.). However, the use of phenylacetylene lowered both the selectivity and the yield for the formation of product 13h (68% yield, 87.5:12:5 e.r.). The subsequent reactions with TES-substituted alkyne 12b produced chiral substrates 13ad-13ah. In general, introduction of an electron withdrawing substituent at the para-position of the C₂-phenyl ring resulted in similar e.r. values, but lower yield. Thus, p-CF₃-substituted substrate 6u provided the corresponding product in 59% yield and 92.5:7.5 e.r., p-MeOCO-substituted substrate 6g led to product 13af in 58% yield and 94.5:5.5 e.r., and the use of p-CN-substituted 6h resulted in 13ag in 63% yield and 96:4 e.r. The configuration of 13af was determined via X-ray crystallographic analysis and was assigned as (R).^[22] At the same time, substitution with p-methoxy, and p-fluoro groups resulted in enantioselectivity erosion, and products 13ad and 13ah were obtained in 57% yield, 87:13 e.r. and 67% yield, 90:10 e.r., correspondingly.

Scheme 3.

Enantioselective alkyne addition with ligand L2.^[a]

[a] The reactions were performed on 0.1 mmol scale using 0.05 M solution in THF as described in SI-II-2.

With the aforementioned results in hands, our next studies were focused on getting further insights into the mechanism of these photochemical transformations. The entire transformation involves cooperative catalysis of Ir(III) and Cu(I) and consists of three stages i) photocatalysis with Ir(III) leading to the formation of the azomethine ylide intermediate; ii) formation of copper(I) acetylide; iii) reaction of copper(I) acetylide and azomethine ylide.

Our computational and fluorescence studies helped to shed some light onto the possible role of the initial stages of the photochemical cascade leading to the formation of azomethine ylide. Ir(III) photocatalysts usually are excited by Blue LEDs (400-440 nm) and we are not aware of using light with longer wavelengths such as Green LED (λ_max=525 nm) to promote either photoredox or energy transfer reactions with Ir(III). Recorded fluorescent spectrum of Ir(4-CF₃-ppy)₃ demonstrates wide absorption in 250-450 nm range that matches perfectly the spectrum of typical Blue LEDs; however, the absorption of Ir(III) and emission of Green LED spectra have minor overlap near 490-500 nm region (cf. Figure SI-9). The emission of the employed Green LED light source seems to be sufficient to perform EnT mechanism in the studied reaction. The emission of the photocatalyst also perfectly matches the energy of the computed and measured triplet state (Scheme 3A-C). While the photoexcited Ir(III) catalyst may facilitate 18-T1 formation via energy transfer mechanism (Scheme 4A), another possibility to consider is the photoredox mechanism proceeding through radical cation intermediates 20 and 21.^[23,29] Mechanism A has already been investigated computationally (Table 1, Scheme 2A, and SI-IV-1,5); however, to gain better understanding of Mechanism B, DFT computations (DFT, SMD/wB97x-D3/cc-pVTZ) were carried out (Scheme 4C and SI-IV-2,3). At first step, aziridinoquinoxaline 18-S₀ may undergo a single-electron transfer (SET) oxidation to form cation radical 20, for which the oxidation potential is estimated to be around E₁ = 2.0 V in toluene for both phenyl- and methyl-substituted aziridinoquinoxalines 6a and 14. In THF and acetonitrile this potential is lowered to 1.5 V and 1.3 V, respectively. The subsequent ring-opening of radical-cation 20 has low energy barrier for R₂ = Ph (8.7 kcal/mol) and slightly higher barrier for R₂ = Me (16.0 kcal/mol), but in both cases this step should be feasible. In addition, the ring-opening of radical-cation 20 is thermodynamically favored by approximately 20 kcal/mol (Gibbs free energy). Finally, the generation of ylide 21 should also proceed easily due to the low reduction potential of the open radical-cation intermediate 21 (E₂ = 0.7 to 0.9 V). For this alternative mechanism B, there are two key steps involving SET process, first is the oxidation of aziridinoquinoxalines, and second is the final reduction of the open form of radical-cation 21 to ylide 19. The reduction of 21 to ylide 19 should be possible by any Ir(II) photocatalyst present in the reaction mixture as for Ir(III)/Ir(II) redox reaction potentials are in range of −1.3 to −2.2 V (Scheme SI-2). However, the first oxidation step of 18-S₀ may have a potential mismatch with the oxidation potential of the excited Ir(III) photocatalyst (highest potential E = 1.21 V, Scheme SI-2). To confirm this, a cyclic voltammetry study was performed on aziridinoquinoxaline 6a in acetonitrile (Scheme 4B) demonstrating irreversible oxidation of 6a with the half wave at 1.49 V (vs SCE, corrected with ferrocene). The measured potential matches well computed 1.36 V in MeCN by the DFT studies. Based on this we can make a conclusion that the proposed SET mechanism B cannot be operational under the current photochemical conditions since Ir(4-CF₃-ppy)₃ is not a sufficiently strong oxidant to generate radical-cation 20.

Scheme 4.

Experimental and computational exploration of the reaction mechanism.

[a], [b] Calculations of redox potentials were performed using SMD(Toluene)/wB97x-D3/cc-pVTZ, transition state search and entropy analysis was performed using B3LYP/6-31+G**, electronic energy was reevaluated at SMD(Toluene)/wB97x-D3/cc-pVTZ level of theory.

To probe the last step of the mechanism involving the reaction of copper(I) acetylide and ylide, (cf. Schemes 1), the reaction of 6d and D-labeled alkyne 12j was investigated using the standard protocol (Scheme 4C). as the substrate was used to probe both the deprotonation step by DIPEA as well as the following H/D-transfer to ylide 10. If this mechanistic hypothesis is correct, then we should observe deuterium incorporation in the final product 13d. While the reaction with the D-labeled substrate 12k proceeded as usual, the resultant product 13d was isolated in 89% yield with no deuterium incorporation. As the deuterated toluene (d8-toluene) was used in the reaction, only DIPEA may serve as the source of hydrogen atoms that protonates the ylides instead of alkyne. This process may involve oxidation of the intermediate ylide 19 with Ir(III) to form intermediate 21 that may subsequently undergoe a HAT reaction with DIPEA to form a protonated quinoxalinium ion that undergoes further reaction with copper(I) acetylide to form 13d.

Following the studies above, we were seeking to explain consistently observed decomposition of aziridinoquinoxaline to 2-phenylquinxaline 17 (Scheme 4D). The product-to-decomposition ratio depended on the solvent used as well as chiral ligand and copper(I) salt. This led us to the hypothesis that there is a competitive decomposition mechanism(s) to product formation and the reaction conditions may regulate the favoring pathway. We envisioned that the generated quinoxalinium intermediates 22 should have similar properties to N-alkylpyridinium (Katrizky’s) salts,^[24] which are well-known to undergo radical extrusion after the SET reduction. Similarly, the reduction of quinoxalinium ion intermediates 22 to radical-cation 23 followed by the radical extrusion will lead to 2-phenylquinoxaline 17. Therefore, we calculated the reduction potential for intermediates 22 to radical 23 (Scheme 4D and Table SI-20). DFT calculations at ωB97X-D3/cc-pVTZ level of theory and SMD solvation model (for Toluene, Acetonitrile and THF) provided values ranging from −0.35 to −0.65 V (−0.45 V for toluene specifically). Therefore, the reduction of 22 can be accomplished with Ir(II) species generated by Ir(III) reduction with DIPEA. The subsequent step involving the C-N bond dissociation leading to 17 has high barriers (Scheme 4D), but phenyl-substituted aziridinoquinoxaline 6a (R₂ = Ph) decomposition proceeds with lower activation energy (29.1 kcal/mol) in comparison to the methyl (37.9 kcal/mol) and tert-butyl (42.4 kcal/mol) derivatives 14 and 6d. Such difference in activation energy can be explained by the formation of a more stable benzyl radical, and it is consistent with our prior observations of lower photochemical stability of phenyl-substituted substrates 6a-c in comparison to 6d and related compounds.

(3)

(4)

Finally, to validate the intermediacy of the quinoxalinium ions 19, we attempted a direct alkylation of 2-phenylquinoxaline with neopentyl and ethyl iodides (cf. Table SI-7). The exposure of 2-phenylquinoxaline to neopentyl iodide or ethyl iodide did not lead to the formation of the corresponding quinoxalinium salts even after heating it for 24 h at 150 °C. This is consistent with the low basicity and nucleophilicity previously observed for quinoxalines (pK_a of the protonated form in water = 0.56).^[25] These results also suggest that the described herein indirect method is unique for providing quinoxalinium ions and azomethine ylides for the substrates containing other basic nitrogen groups such as 13aa (Scheme 2).

The dihydroquinoxaline products 13 contain multiple functional groups that may enable their functionalization into more complex quinoxaline derivatives (Eq. 3 and 4). Thus, the C=N moiety of substrates 13k and 13m may undergo Cu(I)-catalyzed allylation with allyltrimethoxysilane to provide 24k and 24m in 72% and 64% yield as anti-diastereomers only (Eq. 3).^[26] Similarly, subjecting 13k to the reaction with stoichiometric Hantzsch ester resulted in syn-diastereomer 25k in 84% yield and >25:1 d.r. (Eq. 4). It is noteworthy that tetrahydroquinoxalines 24 and 25 cannot be easily derived by other methods such as hydrogenation of quinoxalines.

Conclusion

In summary, this manuscript describes a new method for accessing dihydroquinoxalines from readily available aziridinoquinoxalines. While aziridinoquinoxalines undergo C=N addition reaction with nucleophiles, we have discovered a photochemical method that enables alkyne addition to the aziridine, rather than imine, moiety of the aziridinoquinoxalines. While the attempts to accomplish this transformation using thermal activation or direct photolysis with Blue LED as the light source were not successful, the computational studies helped to discover a dual Ir(III)/Cu(I) catalytic system that promotes alkyne addition under green LED photolysis (λ_max = 525 nm). This mild method features high levels of chemo- and regioselectivity and was used to generate 30 highly functionalized substituted dihydroquinoxalines (13a-13ac) in 36-98% yield. This transformation was also carried asymmetrically using (S,R)-N-PINAP (L2) as the chiral ligand to provide 9 chiral addition products in 96:4 to 86:14 e.r. These dihydroquinoxaline products may serve as the substrates for further functionalization to provide highly functionalized tetrahydroquinoxalines that cannot be easily obtained using other methods. The experimental and theoretical explorations of this photochemical reaction suggest a mechanism that involves Ir(III)-catalyzed triplet energy transfer mechanism followed by a ring-opening reaction ultimately leading to the formation of azomethine ylide intermediates. These azomethine intermediates undergo sequential protonation to provide quinoxalinium ion intermediates^[27,28] that undergo copper(I) acetylide addition. To the best of our knowledge, this is the first example of successful generation and trapping of azomethine ylides with visible light under mild conditions,^[29] and further exploration of these transformation is the subject of the ongoing work in our laboratories.