C1‐4 Alkylation of Aryl Bromides with Light Alkanes enabled by Metallaphotocatalysis in Flow

Reaction Optimization

Our optimization efforts commenced by reacting aryl bromide 1 with propane, employing a synergistic combination of Ni complex I and tetrabutylammonium decatungstate (TBADT) as the Hydrogen Atom Transfer photocatalyst (Table 1). CD₃CN was selected as the solvent due to TBADT′s well‐known propensity to abstract hydrogen atoms from CH₃CN under the demanding conditions necessary to activate gaseous reagents. ^[25] A gas‐to‐liquid ratio (G : L) of 20 : 1 (8 equiv.) was employed, with 2,6‐lutidine used to quench the hydrobromic acid formed during the reaction. The solution was introduced into a continuous‐flow microreactor (ID=0.5 mm; 5 mL volume) and irradiated with UV‐A light (Chip‐on‐board LED, λ=365 nm; 144 W optical power) (See Supporting Information, Section 2.3).

By utilizing a back pressure regulator, we were able to elevate the reaction pressure to 32 bar, effectively liquefying the gaseous alkanes and thereby enhancing the probability of their activation through photocatalytic HAT. Under these conditions, a residence time of only 15 minutes (entry 1) was found to be sufficient for detecting the formation of the desired product 2 in a synthetically useful yield with a regioisomeric mixture of 1.2 : 1 (2 a : 2 b). Extending the residence time to 30 minutes (entry 2) did not have a positive impact on the chemical efficiency. However, increasing the G : L to 40 : 1 (16 equiv.), and 60 : 1 (24 equiv.), (entries 3 and 4, respectively) reflected into an increase in yield to 40 %. It is important to note that an excess of gas reagent should not pose any problem, as it can be easily recovered and recycled after the reaction by simply reducing pressure to atmospheric pressure, thus ensuring phase separation between the gas and liquid phases. Increasing the TBADT loading to 2.5 mol % (entry 5) led to a further rise in the chemical efficiency.

At this stage of the optimization, several ligands for the Ni catalyst were screened, with 4,4′‐di‐tert‐butyl‐2,2′‐dipyridyl confirming its superior activity (See Supporting Information, Section 4). An assessment of several reaction additives and solvent mixtures highlighted the positive impact of LiBr and tBuOH (entries 6 and 7). Although extending the residence time to 30 minutes (entry 8) and implementing a recirculation system to increase reaction times even further to 50 minutes (entry 9) resulted in only a minor increase in chemical efficiency, the introduction of a fresh solution of TBADT before the recirculation step (entry 10, See Supporting Information, Section 4.6) furnished the desired compound 2 in a 60 % yield with a regioisomeric mixture of 2 : 1 (2 a : 2 b).

Substrate Scope

After establishing the general reaction conditions and optimizing the design of the photochemical reactor, we explored the scope and generality of the C(sp³)−H (hetero)arylation of light alkanes with aryl bromides in our flow reactor platform (Figure 2). Notably, the utilization of flow technology facilitates the seamless execution of these reactions by harnessing flow energy in combination with back pressure regulators to apply pressure, while also enabling a continuous feed into the reactor to generate substantial quantities efficiently without having to reoptimize the reaction conditions (Figure 2). Additionally, due to the straightforward setup of the reactor platform, expansion of the substrate scope can be swiftly accomplished.

Diverse gaseous alkanes, including butane, propane, ethane, and methane, could be coupled with various aryl bromides efficiently. Notably, pharmaceutically relevant heteroaryl bromides, such as pyridine, pyrimidine, benzothiazole, and azaindole, with different electronic properties and ring sizes, could be alkylated in synthetically useful yields. Also, a menthol ester derivative, bearing many C(sp³)−H bonds with lower bond dissociation energy compared to gaseous alkanes, could be alkylated in synthetically useful yields (15 & 24, 32–50 % yield). Overall, our results demonstrate that n‐butane and propane yielded the desired alkylated analogues in moderate to good yields. In contrast, ethane and methane, which possess significantly higher bond dissociation energies (101 and 105 kcal/mol, respectively), produced lower yields. However, these yields are still considered synthetically valuable. One consideration of this methodology is that aryl bromides with strong electron‐donating groups (e.g., Ar−Br with OMe or NR₂ substituents) did not undergo the desired reactivity.

Mechanistic Investigation

While synthetically useful amounts were obtained for every single gaseous alkane, we observed reduced efficiency in the coupling reaction as the bond dissociation energies (BDEs) for the corresponding C−H scission increased.[ ²⁶ , ²⁷ ] Additionally, for propane and butane as coupling partners, we noted a selectivity erosion upon radical capture by nickel compared to the direct reaction of nucleophilic alkyl radicals with Michael acceptors (Figure 2).[ ²⁵ , ²⁸ , ²⁹ ] To explain these observed trends and gain further insights into the mechanistic rationale of the metallaphotocatalytic cross‐coupling of aryl bromides with gaseous alkanes, we conducted a detailed DFT calculation study and carried out some additional mechanistic experiments to support the computational observations.

Specifically, we conducted a parallel kinetic isotope effect (KIE) study using cyclohexane as a substrate (Figure 3A). We observed a negligible reaction rate difference between cyclohexane and its deuterated substrate, indicating that the HAT process by a photoexcited decatungstate is not involved in the rate‐determining step (RDS) in our reaction conditions. With these kinetic data in hand and based on recent computational work of Gutierrez and Molander on a related reaction system with tertiary alkyl radicals, ^[30] we investigated the reactivity trends with various alkyl radicals from gaseous alkanes using state‐of‐the‐art DFT calculations (Figure 3B and Figure S8 Supporting Information).[ ³¹ , ³² ] The reaction begins with the generation of an alkyl radical by the HAT process between the gaseous alkane and a photoexcited decatungstate anion (*[W₁₀O₃₂]⁴⁻), and subsequent radical coordination to Ni(0) species Int1 to furnish a stable Ni(I) intermediate Int2‐R. In turn, Ni(I) undergoes oxidative addition to the aryl halide through the transition state TS1‐R that leads to the formation of a pentacoordinate Ni(III) intermediate, which is then followed by an alkyl radical dissociation to furnish a square planar Ni(II) intermediate Int4. Upon the radical re‐coordination, the ligand orientation of the Ni(III) intermediate is primed for the formation of a C(sp²)−C(sp³) bond via the reductive elimination through transition state TS2‐R. The final Ni(I)−Br intermediate Int6 is then regenerated as Int2‐R by protonation and alkyl radical addition. Throughout the catalytic cycle, we found that the RDS is the oxidative addition, and more interestingly, the activation energy barrier for this step increases with smaller alkyl radicals, whereas the reductive elimination follows the opposite trend.

To gain deeper insight into the origin of reactivity differences observed for the various radicals, we turned to the activation strain model (ASM) and our energy decomposition analysis (EDA). We focused on the analysis of the rate‐determining oxidative addition step (Figure 3C). The ASM involves the decomposition of the electronic energy ΔE into the strain energy ΔE _strain associated with the structural deformation of the Ni(I)‐alkyl complex and the aryl halide 1 from their equilibrium geometry and the interaction energy ΔE _int between these deformed reactants [Eq. 1].[ ³³ , ³⁴ , ³⁵ , ³⁶ ] Our matching EDA separates the interaction energy (ΔE _int) into the following four chemically meaningful energy terms: classical electrostatic interaction (ΔV _elstat), destabilizing steric Pauli repulsion interactions (ΔE _Pauli), stabilizing orbital interactions that account, among others, for HOMO−LUMO interactions (ΔE _oi), and the dispersion interaction (ΔE _disp) between the deformed Ni(I)‐alkyl complex and aryl halide 1 [Eq. 2].[ ³⁷ , ³⁸ , ³⁹ ]

(1)

(2)

The oxidative addition with the methyl radical goes with the highest reaction barrier (ΔG ^≠=14.3 kcal mol⁻¹), and the energy decreases as the number of substituents of the radical increases (ΔG ^≠=11.6 kcal mol⁻¹ [R=Et], and 10.9 kcal mol⁻¹ [R= ⁱ Pr]).

We performed our activation strain analysis (ASA) at the same point on the reaction coordinate near all transition states with a C(aryl)−Br stretch of 2.13 Å to account for the shift in the transition state position which has been previously shown to skew results. ^[40] The trend in Gibbs free energy activation barriers is the same as for the electronic activation energy barriers (ΔE*). The differences in reactivity can be traced to the differences in the interaction energy, while the strain energies are nearly the same. The higher activation barrier for the methyl radical originates from a less stabilizing interaction energy (ΔE _int=−25.2 kcal mol⁻¹), whereas the lower barrier for the ⁱ Pr radical benefits from a much more stabilizing interaction energy (ΔE _int=−31.4 kcal mol⁻¹). Using our EDA method, we traced this trend back to differences in both the electrostatic interactions (ΔV _elstat) and orbital interactions (ΔE _oi). The origin of the less stabilizing ΔV _elstat in TS1‐Me compared to TS1‐ ⁱ Pr is that the less electron‐donating methyl radical does not effectively increase the electronegativity at the nickel center during the oxidative addition transition state, resulting in the increasing the activation energy barrier (Figure 3D). In addition, the HOMO orbital of Ni(I) species is strongly affected by the contribution from the 2p orbital of the adjacent alkyl substituent, where the HOMO−LUMO gap Δϵ with aryl bromide is larger for the less electron‐donating methyl group (Δϵ=1.8 eV) and the gap is much smaller and favorable for the bulkier and more electron donating ⁱ Pr group (Δϵ=1.5 eV) (Figure 3E). These observations, along with the potential catalyst deactivation pathway with smaller alkyl radicals (See Supporting Information, Figure S9), shed light on the origin of decreased reactivity in the coupling reaction of aryl bromides with methane.

As described above, we observed selectivity erosion upon radical capture for the coupling reaction with propane and butane, and the selectivity is influenced by the aryl bromide coupling partners (Figures 4A, and Figures S10, and S11, Supporting Information). When an electron‐poor aryl bromide was used, the branched alkylation was a major pathway, whereas the selectivity was inverted with an electron‐rich aryl bromide, and then the linear pathway became dominant. To elucidate the origin of this regioselectivity, we again turned to DFT calculations to investigate the alkyl radical isomerization mechanism using propyl radical as a model system (Figure 4B). Initially, Int2‐ ⁱ Pr is formed by the HAT process with a photoexcited decatungstate anion and a radical coordination to the nickel center. This intermediate could undergo β‐hydride elimination to generate Ni(I)−H intermediate Int7, and upon further migratory insertion of

propene leads to a more thermodynamically stable Ni(I) species Int2‐ ⁿ Pr. Interestingly, the energy barriers for this isomerization process are energetically feasible and similar to that of the oxidative addition. This suggests that the oxidative addition step not only determines the reaction rate but also has a major influence on the selectivity. Electron‐poor aryl bromides can undergo facile oxidative addition before radical isomerization thus, the branched selectivity dominates the reaction. On the other hand, electron‐rich aryl bromides are generally less reactive because of the increased energy barrier at the RDS, and radical scrambling is more likely to occur to furnish linear selectivity.