Photocatalytic Arylation of P₄ and PH₃: Reaction Development Through Mechanistic Insight

Previous studies and open mechanistic questions

As part of our initial report of photoredox-catalysed P₄ arylation we described a series of preliminary mechanistic studies, on the basis of which we were able to propose an approximate outline mechanism for the catalytic phenylation of P₄ to Ph₄P⁺.^9a This is reproduced in Scheme 1 a, and involves a fairly typical photoredox catalytic cycle for the generation of phenyl radicals (Ph^. for Ar=Ph), in which the photoredox catalyst [1]PF₆ (1=Ir(dtbbpy)(ppy)₂, dtbbpy=4,4′-di-tert-butyl-2,2′-bipyridine, ppy=2-(2-pyridyl)phenyl; structure shown in Scheme 1) undergoes photoexcitation by blue light, followed by reductive quenching by Et₃N (which acts as terminal reductant in this reaction), and oxidative regeneration by the substrate PhI. This last redox step prompts mesolytic fragmentation of the aryl halide substrate, generating radicals Ph^.. These can then add to P₄ to generate in sequence PhPH₂, Ph₂PH, Ph₃P and Ph₄P⁺, with the formation of these intermediates being observable by ³¹P{¹H} NMR spectroscopy. The H atoms required for formation of PhPH₂ and Ph₂PH were proposed to derive from the reductant Et₃N, presumably via direct abstraction of H^. (HAT) or H⁺ following oxidation to Et₃N^.+. As part of the study described herein it was possible to confirm this hypothesis through use of deuterium-labelled Et₃N-d₁₅, which led to clear deuterium incorporation into these intermediates (see Section 3.6 of the Supporting Information for further details). Subsequent loss of these H atoms was proposed to proceed by formation of ArH (observed as a byproduct by ¹H and ²H NMR spectroscopy).¹⁴

Each of the individual arylation steps was previously confirmed to be photocatalytic. Indeed, in a recent follow-up study, we reported that the same methodology could be optimised to use PhPH₂ and Ph₂PH as starting materials for the generation of unsymmetrically substituted products (e.g. PhPAr₃⁺, Ph₂PAr₂⁺; Scheme 1 b).^9b This report also showed that the previously used precious metal catalyst [1]PF₆ could be replaced by the inexpensive and easily accessible organic catalyst 3DPAFIPN (2; structure shown in Scheme 1) for the arylation of both P₄ and PhPH₂/Ph₂PH.¹⁵

Despite the superficial simplicity of the catalytic cycle shown in Scheme 1 a it is clear that the overall mechanism is highly complex (we have emphasised previously that exhaustive phenylation of one equivalent of P₄ requires the cleavage of six separate P-P bonds, formation of 16 new C-P bonds, and consumption of at least 16 equivalents each of PhI and Et₃N) and that several important questions remain to be answered. From among these, at the outset of this study we identified three that are of particular significance:

1) Is it possible to identify any other intermediate(s) formed prior to PhPH₂? Although PhPH₂ is the first P-containing intermediate that it has been possible to observe during the phenylation of P₄ thus far, it clearly cannot be the first such intermediate. Notably, by identifying PhPH₂ and Ph₂PH as intermediates it has been possible to develop new functionalisation reactions of these substrates,^9b which might also be true for any earlier intermediates.

2) Is it possible to account for missing ³¹P intensity in these reactions? Preliminary in situ ³¹P{¹H} NMR monitoring has shown that initial consumption of P₄ is far faster than the formation of the first observable P-containing intermediate (PhPH₂). Similarly, standard ³¹P{¹H} spectroscopic measurements of the final reaction mixtures suggest that formation of the target products (e.g. Ar₃P, Ar₄P⁺) is typically very clean (no other P-containing species present, including P₄), yet conversion to these products is consistently less than quantitative based on the initial amount of P₄.

3) Can discrepancies in performance between catalysts [1]PF₆ and 2 be explained? Although their photoredox properties are very similar, [1]PF₆ has been found to give consistently higher product conversions for P₄ arylation than 2. Conversely, 2 has been found to provide better conversions than [1]PF₆ for analogous arylations starting from PhPH₂/Ph₂PH.

Identifying new intermediates and side-products during the arylation of P₄

Having previously shown ³¹P{¹H} NMR monitoring to be a valuable tool for the investigation of the P₄ arylation reaction,^9a we reasoned that answers to our question set might be found through more detailed NMR spectroscopic analysis. For consistency with previous studies, arylation using PhI was again chosen as the simplest model system. To begin, the phenylation of P₄ using organic photocatalyst 2 was monitored over time, to provide a comparison with the analogous experiment already performed using [1]PF₆ as catalyst.^{9a, 16} As expected, a similar overall reaction profile was observed (Figure 2 a), with rapid consumption of P₄ followed by intermediate formation of PhPH₂, Ph₂PH and Ph₃P, and ultimately formation of Ph₄P⁺ as the major reaction product. Observation of a photo-CIDNP effect in the corresponding ¹H NMR spectra was also consistent with previous observations using [1]PF₆ (see Section 3.8 of the Supporting Information for more details).^{9a, 17} Nevertheless, some qualitative differences between the two catalysts could clearly be resolved. In particular, formation of the early intermediates PhPH₂ and Ph₂PH appears to peak significantly more quickly using 2, while the magnitude of these peaks is reduced (c.f. Figure 2 a,b).

More dramatically, in addition to the expected intermediates Ph_xPH_3−x (x=1–3), ³¹P{¹H} NMR monitoring clearly showed the formation of a significant number of other minor peaks during the reaction catalysed by 2 (for representative example spectra, see Figure 3 a). Some of these (designated Int1–Int4) were identified as intermediate species (either for productive or unproductive processes) based upon their concentration profile over time, while others (designated Side1–Side5) were identified as side-products and were persistent until the end of the reaction (Figure 2 a). The same side-products (Side1–Side5) could also subsequently be detected in the reaction catalysed by [1]PF₆, but at appreciably lower concentrations (Figure 2 b). Nevertheless, the combined intensity of these signals accounts for a significant proportion of the overall ³¹P intensity in the final reaction spectra, and it seems feasible that there may be further unresolved signals of lower intensity, which could sum to a further significant value (vide infra).

These observations suggest a clear explanation for less-than-quantitative target product formation highlighted above, with maximum formation of the final target products being limited by the formation of a collection of minor side products in amounts that are individually insignificant, but collectively substantial. Moreover, the fact that far greater amounts of these side-products are observed when using the organocatalyst 2 provides a rationale for the observation that 2 furnishes consistently poorer final product conversions for the arylation of P₄, despite being a more productive catalyst for the optimised synthetic arylations of the “downstream” intermediates PhPH₂ and Ph₂PH (for a discussion of catalyst performance in these latter reactions see Section 3.9.3 of the Supporting Information).

Plausible structures of the species Int1–Int3 and Side1–Side5 are shown in Figure 3 b and are based upon detailed in situ NMR studies including 2D techniques and proton-coupled ³¹P spectra which are presented in the Supporting Information (Sections 3.10 and 3.11), alongside discussion of possible mechanisms of formation. As an illustrative example, the product Side1 could be assigned based on a combination of its ³¹P chemical shift (suggestive of a Ph₂PNR₂ species), long range ¹H-³¹P correlation experiments (showing correlations between the ³¹P resonance and ¹H signals at chemical shifts consistent with Ph and CH₂Me peaks), and comparison with literature data.

Interestingly, the majority of these species appear to be derived from non-innocent behaviour of the Et₃N terminal reductant (with the notable exception of Int3, Ph₄P₂, which was also observed in our previous investigation of the 2-catalysed reaction of Ph₂PH with cyclohexyl iodide, and was proposed to arise through dimerization of transiently-formed radicals Ph₂P^.).^9b Given that Et₃N is widely used as a “privileged” terminal reductant in photoredox transformations, and that [1]PF₆ and 2 have also been used to catalyse many other photoredox reactions, these results suggest that similar patterns of non-selective side-reactivity could be an underappreciated but highly significant limiting factor in many other photoredox transformations (indeed, related non-innocent side reactions have previously been highlighted in some other catalytic reactions).¹⁸ In the absence of the convenient ³¹P NMR handle, these would likely be extremely challenging to detect. Minimisation of these side reactions is a clear avenue for further reaction optimisation.¹⁹

To try and identify further reaction intermediates—particularly those that may form in the early stages of the reaction—additional monitoring experiments were carried out on reactions performed at lower temperatures and using higher catalyst loadings. It was anticipated that transient, early intermediates might be better able to accumulate under these conditions and, indeed, several further ³¹P{¹H} resonances could be observed (see Figure 4 a; designated Int5–Int8; again, these may be intermediates for either productive or unproductive processes), and plausible corresponding chemical structures assigned are depicted in Figure 4 b (see also Section 3.11.1 of the Supporting Information). The proposed structures of Int5–Int7 further emphasise the non-innocent behaviour of the Et₃N reductant.

PH₃ as a substrate for photocatalytic arylation

The specific observation of PH₃ as one of the above intermediate signals (Int8; readily assigned on the basis of its characteristic chemical shift and binomial quartet splitting in the proton-coupled ³¹P spectrum, see Figure 4 a) was particularly intriguing.^3b Although our previous mechanistic proposals (e.g. Scheme 1 a) had assumed PhPH₂ to be the first P₁ intermediate formed, this result—combined with our previous observation that PH₃ can be formed upon irradiation of a combination of P₄, Et₃N and [1]PF₆ in the absence of ArI—suggested an alternative synthetic possibility. Specifically, it was queried whether PH₃ could in fact be a key intermediate formed prior to arylation: that is, whether arylation of P₄ could proceed via initial Et₃N-mediated reduction to PH₃,²⁰ with arylation only subsequently occurring. Notably, such a possibility would have broad implications beyond the immediate mechanistic significance. PH₃ is a major synthetic intermediate, employed industrially for the preparation of many important phosphine and phosphonium derivatives. However, these syntheses proceed exclusively through the hydrophosphination of unsaturated substrates, which limits the scope of accessible products and thereby excludes the formation of aryl-substituted derivatives (Ar₃P, Ar₄P⁺). Indeed, efficient, direct arylation of PH₃ remains a long-unmet synthetic goal that, though related, is distinct from the target of direct P₄ functionalisation.^{3c, 4, 21}

To this end, it was decided to investigate the deliberate use of PH₃ as a substrate for our photoredox-catalysed arylation methodology. For reasons of practical simplicity, PH₃ was not employed directly, but rather generated in situ using a two-chambered apparatus, inspired by a recent report by Ball et al. (Figure 5).²² Addition of aqueous HCl to solid Zn₃P₂ in one chamber liberates gaseous PH₃, which can diffuse into the second chamber. This chamber contains the remaining components of the photocatalytic reaction mixture and is stirred under continuous blue LED irradiation (with concomitant water cooling to maintain approximately ambient temperature).

Gratifyingly, initial experiments using catalytic [1]PF₆ and reaction conditions identical to those used previously in the optimised phenylation of P₄ unambiguously confirmed the formation of Ph₄P⁺ in modest yields (Table 1, entry 1). This result clearly illustrates the feasibility of PH₃ as an intermediate en route to arylated products and to our knowledge is the first example of the successful catalytic arylation of PH₃. Control experiments confirmed that all reaction components ([1]PF₆, Et₃N, light) were necessary to achieve appreciable product formation (Table S1, SI). Catalytic formation of Ph₄P⁺ could also be achieved using organocatalyst 2 instead of [1]PF₆ (Table 1, entry 2), illustrating that this reactivity is not dependent on the use of a precious metal, and is mechanistically feasible for an organocatalysed reaction.

Table 1. Photocatalytic phenylation of PH₃.

Entry[a]	PH3 source (mmol)	PC (mol %)	t [h]	Solvent	Conversion to Ph4P+ [%]
1	Zn3P2 (0.05)	[1]PF6 (2)	18	MeCN/PhH (3:1)	17
2	Zn3P2 (0.15)	2 (10)	48	MeCN/PhH (3:1)	23
3	Zn3P2 (0.15)	[1]PF6 (2)	48	MeCN	30
4[b]	Zn3P2 (0.15)	[1]PF6 (2)	48	MeCN	35
5[b]	NaPH2 (0.1)	[1]PF6 (2)	24	MeCN	41
6[c]	NaPH2 (0.1)	[1]PF6 (2)	24	MeCN	29
7[c,d]	NaPH2 (0.1)	[1]PF6 (2)	24	MeCN	68

• [a] Reactions performed using a 20 mL two-chambered apparatus described in Figure 5, using 12.5 equiv HCl, 11 equiv PhI, 14.4 equiv Et₃N in 2.0 mL solvent under blue LED irradiation and N₂ atmosphere. [b] A 10 mL two-chambered apparatus was used. [c] A 10 mL single-chambered apparatus was used. [d] No HCl was added. Equivalents of reagent (equiv) and catalyst loadings (in mol %) are defined per P atom.

Conversion to the target Ph₄P⁺ could be significantly increased through modification of the reaction time, solvent, and concentration (Table 1, entry 3 and Table S2, SI). Although these conversions are still appreciably lower than those obtained previously when starting from P₄, this may be attributable at least in part to differences in experimental setup. In particular, the need for gaseous PH₃ to diffuse between reaction chambers before dissolving in the reaction solution is expected to significantly impact the rate of reaction. This conclusion was supported by reactions performed using a two-chamber apparatus of reduced volume, which led to a further improvement in performance (Table 1, entry 4; for full details of the apparatus see Section 2.1 of the Supporting Information). An additional improvement was observed upon replacement of Zn₃P₂ with NaPH₂, as an alternative source of PH₃ upon acidification (Table 1, entry 5).²³

To further mitigate problems associated with gas transfer of PH₃, the use of a single-chambered experimental apparatus was pursued, within which it was hoped that addition of acid would trigger release of PH₃ directly into the photocatalytic reaction solution (for full details see Section 2.2.1 of the Supporting Information). Disappointingly, use of this apparatus in combination with NaPH₂/HCl as a source of PH₃ did not lead to an improvement in reaction outcome, although clear catalytic formation of Ph₄P⁺ was once again observed (Table 1, entry 6). Gratifyingly, however, when this reaction was repeated in the absence of HCl significantly improved results were observed, with very good conversion to the target product achieved following slight further optimisation (Table 1, entry 7 and Table S3, SI). Thus, the conjugate base of PH₃ appears to be a highly effective and convenient synthetic surrogate for PH₃ in these reactions, allowing catalytic arylation to be performed in a practically convenient manner while achieving synthetically relevant conversion and selectivity.

Having identified an optimised set of reaction conditions for the [1]PF₆-catalysed phenylation of PH₃, the use of other, substituted aryl iodides was investigated (Table 2). Satisfyingly, a range of other substrates ArI could be employed successfully, including examples bearing both electron-donating and electron-withdrawing groups and with varying steric bulk. These reactions yielded either phosphonium salts Ar₄P⁺ or tertiary phosphines Ar₃P, with the latter being favoured for more sterically hindered substrates (notably 2-iodotoluene, which provided good conversion to the important ligand (o-tol)₃P with good conversion and excellent selectivity; o-tol=2-methylphenyl; Table 2, entry 2),⁵ as well as electron-deficient substrates (such as methyl 4-iodobenzoate; Table 2, entry 7). Notably, the particularly bulky substrate MesI (Mes=2,4,6-trimethylphenyl) failed to reach even the tertiary phosphine, instead producing the primary and secondary phosphines MesPH₂ and Mes₂PH as the only products detected by standard ³¹P{¹H} NMR spectroscopic measurements (Table 2, entry 15). The heteroatomic substrate Ph₃SnCl could also successfully be employed, providing reasonable, selective conversion to the interesting “P³⁻” synthon (Ph₃Sn)₃P without any further reaction optimisation (Table 2, entry 16). In all cases the observed trends in reactivity and selectivity are in excellent agreement with those noted previously for the photocatalytic arylation of P₄.⁹

Table 2. Photoredox arylation of NaPH₂ (as a surrogate for PH₃) catalysed by [1]PF₆.

Entry[a]	R	Conversion to Ar4P+ [%]	Conversion to Ar3P [%]	Conversion to Ar2PH [%]	Conversion to ArPH2 [%]
1	H	77	–	–	–
2	2-Me	–	63	–	–
3	3-Me	61	6	–	–
4	4-Me	64	<5	–	–
5	2-CO2Me	–	10	–	–
6	3-CO2Me	29	<5	–	–
7	4-CO2Me	–	32	–	–
8	2-SMe	–	13	–	–
9	2-OMe	–	42	–	–
10	3-OMe	53	<5	–	–
11	4-OMe	39	6	–	–
12	2-CF3	–	–	–	–
13	3-CF3	12	7	–	–
14	4-CF3	7	19	–	–
15	2,4,6-Me3	–	–	11	17
16	Ph3SnCl[b]	–	55[c]	–	–

• [a] Reactions performed using 0.1 mmol NaPH₂, 2 mol % [1]PF₆, 11 equiv ArI, 15 equiv Et₃N in 2.0 mL MeCN under blue LED irradiation and N₂ atmosphere for 24 h. [b] Ph₃SnCl instead of ArI. [c] (Ph₃Sn)₃P formed instead of Ar₃P.

Having established the scope of the [1]PF₆-catalysed reaction, the use of the less expensive and more abundantly available, sustainable organic photoredox catalyst 2 was also investigated. Gratifyingly, after only minor further reaction optimisation (Table S4, SI), good conversions could also be achieved using this more practical, precious metal-free catalyst, under similar conditions (Table 3). In general, both catalysts achieve similar conversions, and the same trends in product selectivity are observed with respect to the substrates’ steric and electronic properties. Nevertheless, in several cases an appreciably superior preference is observed for either the metallo- or organo-catalyst (for example: compare entries 2 and 9 in both Table 2 and Table 3), indicating that the choice of catalyst can be tailored to the specific target product.

Table 3. Photoredox arylation of NaPH₂ (as a surrogate for PH₃) catalysed by organocatalyst 2.

Entry[a]	R	Conversion to Ar4P+ [%]	Conversion to Ar3P [%]	Conversion to Ar2PH [%]	Conversion to ArPH2 [%]
1	H	72	–	–	–
2	2-Me	–	51	–	–
3	3-Me	51	<5	–	–
4	4-Me	56	–	–	–
5	2-CO2Me	–	7	<5	–
6	3-CO2Me	20	<5	–	–
7	4-CO2Me	–	9	–	–
8	2-SMe	–	29	–	–
9	2-OMe	–	52	–	–
10	3-OMe	72	7	–	–
11	4-OMe	34	14	–	–
12	2-CF3	–	–	–	–
13	3-CF3	30	6	–	–
14	4-CF3	<5	18	–	–
15	2,4,6-Me3	–	–	7	19
16	Ph3SnCl[b]	–	41[c]	–	–

• [a] Reactions performed using 0.1 mmol NaPH₂, 10 mol % 2, 13 equiv ArI, 16 equiv Et₃N in 2.0 mL MeCN under blue LED irradiation and N₂ atmosphere for 24 h. [b] Ph₃SnCl used instead of ArI. [c] (Ph₃Sn)₃P formed instead of Ar₃P.

Mechanisms of photocatalytic arylation

Based on our prior studies of photocatalytic P₄, PhPH₂ and Ph₂PH arylation, as well as the studies described herein, it is possible to propose complete mechanistic courses for both the photocatalytic formation of PH₃ from P₄ and Et₃N, and the photocatalytic arylation of PH₃ (Scheme 2). For the former, the photoexcited state of the photocatalyst PC (PC=[1]⁺ or 2) undergoes reductive quenching by Et₃N (confirmed previously by fluorescence quenching experiments) to generate Et₃N^+. and the reduced catalyst PC^.−. Thus generated, PC^.− can effect single electron reduction of one of the P-P bonds present in P₄ (confirmed previously by UV/Vis spectroelectrochemistry) to generate a phosphorus-centred radical and anion. Abstraction of H^. and H⁺ respectively by these species from Et₃N^+. (Et₃N having been established as the ultimate H atom source for this chemistry; vide supra; for the corresponding in situ NMR study applying fully deuterated Et₃N see Section 3.6 in the Supporting Information) then generates new P−H bonds and ultimately PH₃.

Arylation of PH₃ is proposed to occur via an analogous mechanism to that already established for P-H arylation of the “downstream” intermediates PhPH₂ and Ph₂PH. Reduction of ArI by PC^.− (generated as outlined above) generates aryl radicals Ar^. that can abstract H^. to generate transient ^.PH₂ which can then combine with a second equivalent of Ar^. to produce ArPH₂ (most likely via intermediate formation of the dimer P₂H₄; c.f. our previous observation of Ph₄P₂ during P-H functionalisation of Ph₂PH).^9b Analogous, subsequent arylation steps then generate in turn Ar₂PH and Ar₃P (and, for certain Ar, Ar₄P⁺), as established in our previous investigations. Formation of this sequence of intermediates has been further confirmed by NMR monitoring of the “single chamber” arylation of NaPH₂ (Figure 6).

Notably, the reaction profile in Figure 6 shows that only small amounts of NaPH₂ are dissolved at any specific time (PH₂⁻), and also that the sequential reaction steps leading to PPh₄⁺ proceed very cleanly with nearly no side product formation (c.f. Figure 2 for analogous, less selective reactions starting from P₄). Based on this apparent correlation it was speculated that using lower concentrations of other monophosphorus starting materials might also provide cleaner reactions, and thus higher eventual yields. And indeed, NMR investigations of the light-driven arylation of H₂PPh (applied in different concentrations) confirmed this hypothesis by showing that the total yield of side products increases with increasing concentration of starting material, alongside a concomitant decrease in yield of the final product (for a more comprehensive discussion of side product formation during the arylation of PhPH₂, see Section 3.9.4 in the Supporting Information).

Competing mechanisms of P₄ arylation

Taken together, the elementary steps outlined above (reduction of P₄ to PH₃ and its subsequent arylation) provide a plausible alternative mechanistic proposal for our previously reported catalytic arylation of P₄. Nevertheless, the mechanistic complexity of these reactions should again be acknowledged, and several specific observations suggest that these are unlikely to be the only kinetically viable elementary mechanistic steps. For example, it was noted in our original report that upon irradiation in the presence of [1]PF₆ and Et₃N, P₄ is consumed more quickly in the presence of PhI than in its absence, with the reverse being observed for consumption of PhI in the presence/absence of P₄. While the latter trend is easily explained based upon the above mechanism (since PhI must compete with P₄ to react with PC^.−), the former is more difficult to account for. Moreover, attempts to quantify the PH₃ formed during the photocatalytic reaction between P₄ and Et₃N in the absence of ArI suggest this to be a rather inefficient process that produces PH₃ in only very low yield (see Section 3.12 of the Supporting Information for full details). It is therefore likely that reaction with Ph^. can provide an alternative means by which the P-P bonds of P₄ are broken (Scheme 3), in line with our initial mechanistic hypothesis.²⁴ We thus propose that initial degradation of the P₄ tetrahedron to P₁ species can proceed through a combination of both pathways, to generate a mixture of PH₃ and ArPH₂.²⁵ The relative rates, and hence importance, of these competing mechanistic routes is likely to depend strongly on the identity of the arene substrate (for example, more hindered aryl radicals may add more slowly to P₄, and so render this pathway less competitive).