Atypical and Asymmetric 1,3‐P,N Ligands: Synthesis, Coordination and Catalytic Performance of Cycloiminophosphanes

Abstract

Novel seven‐membered cyclic imine‐based 1,3‐P,N ligands were obtained by capturing a Beckmann nitrilium ion intermediate generated in situ from cyclohexanone with benzotriazole, and then displacing it by a secondary phosphane under triflic acid promotion. These “cycloiminophosphanes” possess flexible non‐isomerizable tetrahydroazepine rings with a high basicity; this sets them apart from previously reported iminophophanes. The donor strength of the ligands was investigated by using their P‐κ¹‐ and P,N‐κ²‐tungsten(0) carbonyl complexes, by determining the IR frequency of the trans‐CO ligands. Complexes with [RhCp*Cl₂]₂ demonstrated the hemilability of the ligands, giving a dynamic equilibrium of κ¹ and κ² species; treatment with AgOTf gives full conversion to the κ² complex. The potential for catalysis was shown in the Ru^II‐catalyzed, solvent‐free hydration of benzonitrile and the Ru^II‐ and Ir^I‐catalyzed transfer hydrogenation of cyclohexanone in isopropanol. Finally, to enable access to asymmetric catalysts, chiral cycloiminophosphanes were prepared from l‐menthone, as well as their P,N‐κ²‐Rh^III and a P‐κ¹‐Ru^II complexes.

Complexes, chirality and catalysts: 1,3‐P,N hybrid ligands are broadly applied in cooperative catalysis, but show limited electronic variation. We report novel analogues with a high N‐basicity, based on 7‐membered cyclic imines. Their transition metal complexes demonstrated their donor strength (W), hemilabile coordination (Rh), and catalytic activity for nitrile hydration (Ru) and transfer hydrogenation (Ru, Ir). Chiral derivatives were also obtained, synthesized from l‐menthone.

Introduction

Ligand‐based reactivity can enhance the activity of transition metal catalysts, ^[1] as is the case for hybrid ligands, ^[2] which combine the properties of different heteroatoms to enable hemilability, redox non‐innocence, proton shuttling, and substrate coordination.[ ¹ , ² ] 1,3‐P,N ligands are particularly subject to diverse binding modes (N‐κ¹, P‐κ¹, P‐κ¹η², κ², and μ; Figure 1) and their complexes have found application in homo/heterogeneous catalysis, bio‐inorganic chemistry, and photoluminescence.[ ³ , ⁴ , ⁵ , ⁶ , ⁷ ] Most of these 1,3‐P,N complexes are based on pyridyl‐ and imidazolyl‐based ligands A and B, which have structural limitations that are inherent to their syntheses.[ ^4c , ^5a , ^5b ] Recently we reported on the highly tunable iminophosphanes C and their tautomers phosphaamidines D that can be independently substituted on the P, C, and N atoms.[ ⁶ , ⁷ ] These 1,3‐P,N ligands are readily accessible from (base‐stabilized) nitrilium triflate precursors and even though they are obtained as (dynamic) E/Z isomer mixtures, the equilibrium shifts to the desired Z conformer on coordination to metals (Figure 2). ^[6] The hapticity in κ¹/κ²‐[(P,N)RhCp*Cl₂]‐complexes and the favorable performance in (κ¹‐P,N)‐Ru^II‐catalyzed nitrile hydration correlated with the electronic properties of the ligands ^[6d] and the basicity of the nitrogen donor (Figure 3).[ ³ , ⁴ ]

Figure 1.

1,3‐P,N ligands and their diverse transition metal complexes.

Figure 2.

Synthesis of iminophosphanes and hemilabile metal coordination.

Figure 3.

Catalyzed nitrile hydration.

Iminophosphanes C are accessed from nitrilium ion precursors, which are known both as reactive synthons and intermediates.[ ⁵ , ⁸ ] Illustrative is the Beckmann rearrangement of cyclohexanone oxime to the ring‐expanded seven‐membered nitrilium intermediate (with an iminium resonance form), which hydrolyzes to the ϵ‐caprolactam that is used as the building block for the commercial production of nylon‐6 (Figure 4, top).[ ⁹ , ¹⁰ ] The intermediate can also be trapped by nucleophiles, for instance by benzotriazole, ^[11] which under Lewis acid promotion can be displaced by other nucleophiles. ^[11a] This reactivity mirrors our reported nitrilium triflate approach for the synthesis of C and might be suitable to access novel 7‐membered cyclic iminophosphanes (Figure 4 E). Such “cycloiminophosphanes” would be conformationally locked, that is, unable to undergo E/Z isomerization, which could boost their efficiency (Figure 5). Moreover, the ring might carry chiral groups as the required chiral cyclohexanone precursors are readily available from terpenoids used in, for instance, the flavoring and perfume industry. ^[12]

Figure 4.

Cycloiminophosphane synthesis from the Beckmann nitrilium intermediate.

Figure 5.

Non‐isomerizable cycloiminophosphanes.

In this study, we report both the synthesis of these novel cyclic 1,3‐P,N ligands and their surprising electronic properties that set them apart from noncyclic iminophosphanes. We explore their coordination to early and late transition metals in κ¹/κ² complexes and assess their performance in catalytic nitrile hydration and transfer hydrogenation. Crystal structures are provided for the (a)symmetric tetrahydroazepine synthons, ligands, and a W complex.

Results and Discussion

The synthesis of the cyclic ligands is discussed first, followed by an analysis of their donor capacity using IR‐ and ³¹P NMR spectroscopy on W⁰ carbonyl complexes and dynamic Rh^III complexes, respectively. Next, Ru^II species are examined to evaluate the ligands’ performance in homogeneous catalysis. We also discuss analogous asymmetric ligands with a natural product‐derived backbone.

Precursor synthesis

Our synthetic strategy is based on reacting phosphanes with a seven‐membered ring nitrilium ion, that is, the 3,4,5,6‐tetrahydro‐azepinium ion. ^[10] Because this ion could not be accessed by our established methodology in which amides are reacted to imidoyl halides with subsequent halide to triflate exchange, ^[6] as the activating agents converted ϵ‐caprolactam to thick intractable mixtures of presumably protonated imidoyl chloride, dimers and/or nylon like polymers,[ ¹³ , ¹⁴ ] we decided to generate the desired nitrilium ion in situ using the Beckmann rearrangement and trap it with benzotriazole. ^[11]

Treating neat cyclohexanone with hydroxylamine ⋅ HCl salt by grinding them together in a mortar, while slowly adding NaOH, yielded the corresponding pure oxime conveniently, even on large scale (120 mmol, 82 %; Scheme 1). ^[15] Next, under an atmosphere of nitrogen, the oxime was activated in situ as the corresponding methylsulfonate with methylsulfonyl chloride and triethylamine in MeCN at 0 °C.[ ^11c , ¹⁶ ] Benzotriazole was added and the mixture was heated to reflux for 2 h to facilitate the ring expansion and trap the nitrilium ion. The reported work‐up ^[11c] was significantly simplified by adding water to the crude mixture to precipitate pure 4 as a white solid in good yield (60 %); alternatively, evaporation, extraction into Et₂O and filtration over neutral alumina also provides 4 (58 %). Single crystals suitable for X‐ray diffraction analysis were obtained from Et₂O and revealed a remarkably flat conformation [N1−C1−N2−N3=179.54(8); C6−C1−N2−N3=−0.37(12)], presumably due to N1⋅⋅⋅H11 and N3⋅⋅⋅H6A hydrogen bonding (2.50 and 2.29 Å, respectively; Figure 6). The C1−N1 bond length of 1.2643(13) Å is typical for an imine bond and the C1−N2 bond of 1.4334(12) Å is similar to those of other N‐heterocycle‐stabilized imines. ^[6c]

Scheme 1.

Synthesis of benzotriazolyl‐tetrahydroazepine.

Figure 6.

Displacement ellipsoid plot of benzotriazolyl tetrahydroazepine 4 at the 50 % probability level. Hydrogen atoms are omitted for clarity, with the exception of H6A and H11. Selected bond lengths [Å] and angles [°]: C1−N1=1.2643(13), C1−N2=1.4334(12), N1−C2=1.4693(12), C1−C6=1.5046(13), N2−N3=1.3746(11), N3−N4=1.2947(12), N2−C12=1.3780(12), N3⋅⋅⋅H6A=2.29, N1⋅⋅⋅H11=2.50, N1−C1−N2=114.81(9), N1−C1−C6=129.01(9), N2−C1−C6=116.18(8), C1−N1−C2=118.88(9), N1−C1−N2−N3=179.54(8), C6−C1−N2−N3=−0.37(12).

Phosphane introduction

In analogy to the formation of 4, we tried to capture the in situ generated 3,4,5,6‐tetrahydro‐azepinium ion directly with a phosphane to obtain the desired cycloiminophosphane ligand E, but to no avail. Instead, the benzotriazole group of 4 could be replaced for diphenylphosphane using Lewis acid promotion, which we examined under a variety of conditions (Scheme 2, Table 1). We started with a microwave reaction employed in related displacements by N nucleophiles, ^[11a] but this reaction using 10 mol % AlCl₃ (entry 1) proved to be less effective than regular heating under reflux (entries 2 and 3). The still modest conversion, probably due to Lewis pair interaction with the phosphane, ^[17] could be enhanced by increasing the amount of AlCl₃ (entries 4 and 5). Using equimolar amounts, AlCl₃ proved to be more selective than SnCl₂, SnCl₄, and BF₃ (entries 5–10) and resulted in 99 % conversion to an Al adduct of the desired ligand, which on treatment with water gave the protonated ligand and a mixture of oxy aluminum anions. However, the by far most effective and convenient manner to obtain the protonated ligand was found to be the direct activation of 4 through protonation with triflic acid (entry 11).

Scheme 2.

The activation of 4 for Ph₂PH introduction.

Table 1.

Lewis acid induced exchange of the benzotriazolyl group of 4 for PPh₂.^[a]

	Promotor	Loading [mol %]	Solvent	t [min.]	Conversion to E [%]	Selectivity for E [%][b]
1[c]	AlCl3	10 %	CHCl3	10	25	100
2	AlCl3	10 %	CHCl3	90	40	95
3	AlCl3	10 %	CHCl3	300	57	88
4	AlCl3	25 %	CHCl3	90	64	93
5	AlCl3	100 %	CHCl3	90	99	99[d]
6	SnCl2	100 %	CHCl3	90	70	95
7	SnCl2	100 %	CHCl3	300	72	97
8	SnCl2	100 %	Toluene	300	36	63
9	SnCl4	100 %	CHCl3	300	50	50
10	BF3 .OEt2	100 %	CHCl3	300	76	92
11[e]	HOTf	100 %	CHCl3	10	100	100

[a] Conditions: 1 equiv. Ph₂PH, solvent (0.35 M), reflux. [b] Determined by ³¹P NMR spectroscopy. [c] T=80 °C, μW conditions, 0.25 M. [d] Conversion to the AlCl₃ adduct of 6. [e] RT, 0.5 M.

Treatment of 4 with 1 equiv. of triflic acid at 0 °C resulted instantly in a suspension from which its iminium form 5 could be isolated by filtration (Scheme 3). Whereas 5 is subject to decomposition over time, both in solution and as an isolated solid, it reacted cleanly upon immediate resuspension with phosphanes to give 6 within minutes. After work‐up, diphenyl derivative 6 a was obtained in 77 % as an air‐stable solid. The aliphatic derivatives 6 b and 6 c (conversion 85 % (nBu) and 66 % (Cy)) could not be purified satisfactorily until after the subsequent deprotonation step (see below). Bulky substituents may hinder the formation of 6, as suggested by the lower selectivity found for 6 c. Crystals suitable for X‐ray diffraction could be obtained for 6 a and 6 c by slow diffusion of pentane into a THF solution. The molecular structures show a chair conformation for the tetrahydroazepine rings, with the P‐lone pair facing away from the imine (Figure 7). Generally, the structures are comparable to those reported for noncyclic iminophosphane ligands.[ ^6a , ^6b ] The imine bond lengths are in the expected range [6 a: N1−C1=1.2858(16) Å; 6 c: N1−C1=1.2892(19) Å], as are the P−C bonds [6 a: P1−C1=1.8269(13) Å; 6 c: P1−C1=1.8310(16) Å]. ^[18]

Scheme 3.

Acid‐facilitated activation of 4 to give ligands 6 and 7.

Figure 7.

Displacement ellipsoid plot of phosphaazepinium triflates 6 a (left) and 6 c (right) at the 50 % probability level. C−H hydrogen atoms and the triflate anions are omitted for clarity. Selected bond lengths [Å] and angles [°] of 6 a: P1−C1=1.8269(13), P1−C7=1.8249(13), P1−C13=1.8223(13), C1−N1=1.2858(16), N1−C2=1.4839(16), N1−H1=0.869(13), N1−C1−P1=123.06(10) N1−C1−C6=119.70(11), C1−N1−C2=125.03(11); Selected bond lengths (Å) and angles (°) of 6 c: P1−C1=1.8310(16), P1−C7=1.8478(15), P1−C13=1.8697(16), C1−N1=1.2892(19), N1−C2=1.4779(19), N1−C1−P1=124.44(12), N1−C1−C6=120.25(14), C1−N1−C2=125.72(13).

Compounds 6 a–c could be readily deprotonated in THF by using NaHMDS as base at −78 °C and then extracted into pentane to provide the desired novel ligands (7 a: 92 %; 7 b: 83 %; 7 c: 76 % yield). Surprisingly, and in contrast to 6, the products show sets of two distinct ³¹P NMR signals [7 a: δ 6.0 (14 %), 5.4 (86 %); 7 b: δ −10.7 (95 %), −17.5 (5 %); 7 c: δ 17.5 (88 %), 13.5 (12 %)]. Conformational flips of the aliphatic rings, known as flippamers, are known to be observable by NMR spectroscopy, ^[19] and this may also be the case for the tetrahydroazepine ring in 7 (Figure 8, with the presumed major conformer in the center).

Figure 8.

Possible tetrahydroazepine‐ring conformations (flippamers) for 7 a–c. Two axial protons are marked for clarity.

A striking property of 7 is its relative basicity. For example, dissolving 7 in CDCl₃ led to its instant decomposition, presumably due to protonation (chloroform pK _a>16). Compared to known iminophosphanes, the seemingly higher basicity of 7 could be attributed to its C‐ and N‐alkyl substituents. For comparison, we synthesized noncyclic C,N‐dimethyl iminophosphane 9 from the nitrilium ion obtained by methylation of acetonitrile (8, 91 %) ^[20] and diphenylphosphane (Scheme 4). Also in this case, a strong base (nBuLi) was needed to generate the 1,3‐P,N product (93 %; δ(³¹P)=6.7 (E), −13.2 (Z)), whereas NEt₃ (pK _a≈11) sufficed for previously reported ligands. ^[6] The unexpected high basicity of 7 seems akin to that of structurally related 1,3‐N,N bases such as DBU. ^[21]

Scheme 4.

Synthesis of 8, the noncyclic analogue of 7 a.

Coordination chemistry and catalysis

The donor strength of P,N ligands affects their efficiency in cooperative reactions.[ ³ , ⁴ , ⁶ ] To examine the influence of 7 in transition metal complexes, we synthesized the tungsten carbonyl complexes 10 and 11 as the IR frequency of their trans‐CO ligands reflect the ligand's P,N donor strength. ^[22] Treating ligand 7 a with either [W(CO)₅(MeCN)] or [(COD)W(CO)₄] provided, respectively, κ¹‐complex 10 (82 %) and κ²‐complex 11 (83 %; Scheme 5). IR spectroscopic analysis of the trans‐CO ligands indicated increasing electron donation to the metal for 11 (2069, 2008 (trans), 1869, 1825 cm⁻¹) with respect to 10 (2069 (trans), 1904, 1873 cm⁻¹). Compared to the analogous W complexes of the widely applied Ph₂PPy ligand (Figure 1, A; κ¹: 2050, 1980, 1920 cm⁻¹; κ²: 2017, 1890, 1870, 1826 cm⁻¹) ^[22a] 7 a appears to be a far stronger N donor. Of note is that the CO stretches for κ²‐complex 11 are weaker than those of κ¹‐complex 10, which illustrates that the additional coordination of the strong N donor provides a more electron‐rich metal center with stronger W→(CO) backdonation. Single crystals of 10 were obtained by cooling a toluene solution. The molecular structure (Figure 9) shows a tight bond of the metal center with the trans‐CO [W1−C19=2.0058(18) Å; W−C_avg=2.046 Å] and a slightly weakened CO triple bond [C19−O1=1.142(2) Å; C−O_avg=1.139 Å]. The W1−P1 distance [2.5333(4) Å] is comparable to the one in analogous PPh₃ and Ph₂PPy complexes.[ ^22a , ²³ ] Further parameters of the ligand are similar to those for 6 a.

Scheme 5.

Analysis of the donor capacity of 7 a by using W(CO) _n complexes.

Figure 9.

Displacement ellipsoid plot of the W carbonyl complex 10 at the 50 % probability level. Hydrogen atoms are omitted for clarity. Selected bond lengths [Å] and angles [°]: W1−P1=2.5333(4), W1−C19=2.0058(18), C19−O1=1.142(2), P1−C13=1.8216(15), P1−C7=1.8234(16), P1−C1=1.8617(16), C1−N1=1.265(2), C19−W1−P1=175.60(5), O1−C19−W1=177.95(15), N1−C1−P1=117.22(12), C1−P1−W1−C21=−15.71(8), C13−P1−W1−C22=14.32(7).

The influence of P substituents ^[5c] was apparent in [(7)Rh^III(Cp*)Cl₂] complexes, which can equilibrate between P‐κ¹ and P,N‐κ² forms with characteristic ³¹P NMR chemical shifts and coupling constants, as has been shown for related iminophosphanes (generally: ¹ J _P,Rh ^κ1≈146 Hz; ¹ J _P,Rh ^κ2≈114 Hz; for example [(Me‐NC(Ph)P(3‐Me‐Ph)₂)RhCp*Cl₂] (P‐κ¹: δ 34.7 ppm, ¹ J _P,Rh=144.6 Hz; P,N‐κ²: (δ −12.1 ppm, ¹ J _P,Rh=114.7 Hz); see also Figure 2).[ ^6b , ^6d ] To obtain the complexes, 7 was reacted with 0.5 equiv. [Rh^IIICp*Cl₂]₂ in CH₂Cl₂ (Scheme 6) to give exclusively P‐κ¹ complexes for the n‐butyl‐substituted ligand (³¹P NMR: 12 b: δ 29.7 ppm, ¹ J _P,Rh=134.2 Hz, major (98 %); 21.3 ppm, ¹ J _P,Rh=136.2 Hz, minor (2 %)) and mainly the P,N‐κ² complex for the dicyclohexyl ligand (12 c: δ −1.5 ppm, ¹ J _P,Rh=104.5 Hz, major (66 %); 41.0 ppm, ¹ J _P,Rh=133.4 Hz, minor (18 %); 31.2 ppm, ¹ J _P,Rh=130.6 Hz, minor (16 %)). The observation of two P‐κ¹ resonances suggests the presence of flippamers or rotamers; the absence of ³¹P,³¹P couplings rules out bridged complexes (μ coordination).

Scheme 6.

The coordination of ligands 7 to Rh^III and Ru^II.

Surprisingly, Rh complexation of phenyl ligand 7 a did not give a clean P‐κ¹/κ² mixture. The ³¹P NMR spectrum of complex 12 a showed a mixture of two broad signals (δ 36.9 ppm, d, ¹ J _P,Rh=142.5 Hz, (42 %); δ 30.1 ppm, br s, (58 %)) that resolved into three doublets on lowering the temperature to −94 °C (Figure 10). Their coupling constants suggests a mixture of two P‐κ¹complexes (δ 38.0 ppm, ¹ J _P,Rh=139.3 Hz (31 %); δ 36.5 ppm, ¹ J _P,Rh=137.7 Hz (54 %)) and possibly an N‐κ¹ complex, as it has a strikingly different P−Rh coupling and other coordination modes are unlikely due to the lack of additional P and/or Rh couplings. (δ 27.5 ppm, J _P,Rh=153.9 Hz (15 %)). The N‐monodentate coordination mode of 1,3‐P,N ligands has been reported only for Mn^II and Fe^II,[ ³ , ²⁴ ] and not for rhodium. Apparently, the two P‐κ¹ and one N‐κ¹ bonding modes interchange rapidly at room temperature.

Figure 10.

³¹P NMR spectra at room temperature and at −94 °C for complex 12 a.

The competing P‐κ¹/N‐κ1 coordination modes for 12 a can be attributed to the C,N‐dialkyl‐P‐phenyl substitution pattern of ligand 7 a. Its aryl groups reduce the donating property of the phosphane group as compared to 7 b and 7 c, and the cyclic alkyl chain makes its imine a stronger donor than in reported iminophosphanes. ^[6]

For comparison, we synthesized the Rh complexes of noncyclic ligand 9 (Scheme 4), which is similarly substituted as 7 a. The low temperature ³¹P NMR spectrum showed the P‐κ¹ complex (δ 39.1 ppm, ¹ J _P,Rh=144.2 Hz (55 %); δ 27.7 ppm, ¹ J _P,Rh=132.8 Hz (29 %)) together with small amounts of both the N‐κ¹ complex (δ 33.9 ppm, ¹ J _P,Rh=152.3 Hz (4 %)) and the P,N‐κ² complex (δ −15.0 ppm, ¹ J _P,Rh=115.0 Hz (6 %)) before AgOTf converted it fully to the κ² complex (72 % yield, δ(³¹P) −15.5 (d, ¹ J _P,Rh=114.6 Hz). Even though the amount of observed N‐coordination is lower for 9 than for 7 a, these results highlight the influence of the C,N,P substituents of 7 on the P,N coordination mode.

All Rh‐complexes 12 a–c could be fully converted to the bidentate complex 13 upon chloride abstraction with AgOTf (Scheme 6; 13 a: 82 % yield, δ(³¹P) −16.8 ppm, ¹ J _P,Rh=113.9 Hz; 13 b: 83 % yield, δ(³¹P) −19.7 ppm, ¹ J _P,Rh=110.0 Hz; 13 c: 76 % yield, δ(³¹P) −2.6 ppm, ¹ J _P,Rh=106.8 Hz).

Next, we explored the coordination to Ru^II (Scheme 6) and the catalytic activity ^[4a] of the resulting complexes. Reacting 7 a,b with [Ru(p‐cym)Cl₂]₂ (p‐cym=p‐cymene) provided the P‐κ ¹ complex 14 b (66 %, δ(³¹P) 27.3 ppm). Complex 14 a could not be isolated from the reaction mixture that showed the presence of two products (δ(³¹P) 31.2 (27 %), 23.4 (73 %) ppm), which we tentatively assign to the N‐κ ¹ and P‐κ ¹ complexes, respectively, in analogy to Rh^III complex 12 a (see above).

Next, three Ru^II complexes of ligands 7 a–c were preliminarily tested for their effectiveness as catalysts in the solvent‐free, closed‐vessel hydration of benzonitrile ^[25] at 180 °C for 3 h (Table 2). Surprisingly, 14 a generated in situ with the phenyl substituted ligand 7 a proved to be quite an active catalyst, yielding 79 % product. In situ generated 14 b with the n‐butyl substituted ligand 7 b afforded a somewhat lower yield of 59 %, which could be enhanced to 70 % by preforming the catalyst (entries 2 and 3, respectively). The least effective catalyst was the Ru^II complex of ligand 7 c, giving a hydration yield of only 15 % that may have its origin in the more limiting steric factors. Even though the catalytic conditions were not optimized in this brief screen, it is rewarding that a hydration yield as high as 79 % was obtained for P‐κ¹‐Ru^II complex 14 a, which resembles the highest yield of 82 % found for the comparable Ru catalyst with an acyclic iminophosphane. ^[25] Both perform much better than the analogous Ru complex of the established Ph₂PPy ligand, which gives a hydration yield of 6 %. ^[6d]

Table 2.

Solvent‐free catalyzed hydration of benzonitrile with precatalyst [Ru(p‐cym)Cl₂].^[a]

	Ligand (L)	T [°C]	t [h]	Yield[b] [%]
1	7 a	180	3	79
2	7 b	180	3	59
3	7 b [c]	180	3	70
4	7 c	180	3	15
5	PyPPh2	180	3	6 [6d]

[a] Reaction conditions: Ph−C≡N (3.6 mmol), H₂O (7.2 mmol), 1.4 mol % [Ru(p‐cym)Cl₂] and ligand 7. [b] Determined by GC. [c] Preformed catalyst (14 b).

As complex 14 b performed only modestly in the hydration of benzonitrile, we chose to further screen its potential by preliminarily exploring the transfer hydrogenation of cyclohexanone in iPrOH, under conditions adapted from Jalón et al., who used the analogous complex of 2‐PPh₂‐1‐methylimidazole to obtain a hydrogenation yield of 21 % on using a KOH/catalyst ratio of 333 : 1 and a substrate/catalyst ratio of 2000 : 1.[ ²⁶ , ²⁷ ] Table 3 summarizes the effect of changes in catalyst loading, reaction time, and the addition of KOH. After in situ generation of the catalyst, at 3 mol % catalyst loading the conversions were slow (up to 20 h; entries 1–3), but similar to the catalyst of Jalón et al., ^[26] the catalyst was substantially more active in presence of KOH (entries 4 and 5). Even the corresponding κ²‐complex of 14 b, obtained by ion exchange with NaBF₄, was active under these conditions (entry 7). With 0.5 mol % catalyst and 2.5 mol % KOH, a reaction time of 2 h still resulted in the quantitative hydrogenation of cyclohexanone (entry 8). Last, as iridium(I) complexes are generally very active hydrogenation catalysts, ^[28] we also explored the in situ generation of [(7 b)Ir(COD)Cl], which showed a similar trend as the Ru^II complexes (entries 9–11).

Table 3.

Catalytic transfer hydrogenation of cyclohexanone.

	Precatalyst [M]	L	Cat. Loading [mol %]	t [h]	Yield[a] [%]	Additives
1	[Ru(p‐cym)Cl2]2	7 b	3	2	3	–
2	[Ru(p‐cym)Cl2]2	7 b	3	3	20	–
3	[Ru(p‐cym)Cl2]2	7 b	3	20	quant.	–
4	[Ru(p‐cym)Cl2]2	7 b	1	20	0	–
5	[Ru(p‐cym)Cl2]2	7 b	1	4	quant.	5 mol % KOH
6	[Ru(p‐cym)Cl2]2	7 b	1	4	0	1 mol % NaBF4
7	[Ru(p‐cym)Cl2]2	7 b	1	4	quant.	5 mol % KOH, 1 mol % NaBF4
8	[Ru(p‐cym)Cl2]2	7 b	0.5	2	quant.	2.5 mol % KOH
9	[Ir(COD)Cl]2	7 b	1	4	2	–
10	[Ir(COD)Cl]2	7 b	1	4	quant.	5 mol % KOH
11	[Ir(COD)Cl]2	7 b	1	4	84	5 mol % KOH, 1 mol % NaBF4

Reaction conditions: cyclohexanone in i‐PrOH (2 M), reflux. [a] Determined by GC.

These preliminary screenings demonstrate that the conveniently in situ generated κ¹ and κ² complexes of Ru^II and Ir^I with cyclic 1,3‐P,N ligand 7 are active catalysts that warrant further scrutiny.

Chirality

As, to the best of our knowledge, no asymmetric 1,3‐P,N‐ligand‐based catalysts have been reported, ^[4] the synthesis of such ligands may be valuable. With a synthetic route toward cyclic iminophosphanes at hand, we pursued substituting the backbone with a chiral group by using an inexpensive terpenoid as asymmetric starting point. The readily available terpenoid l‐menthone ^[12] is well‐suited for this purpose, since its sizeable (2S)‐iPr group is expected to be favorable for asymmetric induction. ^[29] Following the synthesis of the chiral ligands, we report their Ru and Rh complexes and briefly reflect on their catalytic potential.

The asymmetric derivatives of 7 were pursued in analogy to the parent compound, albeit that the solventless oxime synthesis was not effective, but reacting l‐menthone with the hydroxylamine ⋅ HCl salt in an EtOH/H₂O mixture did provide oxime 16 as a colorless liquid after purification by crystallization at 5 °C (73 %; Scheme 7). ^[30] Subsequent treatment with MsCl, NEt₃ and benzotriazole induced the Beckmann rearrangement via 17 to the desired benzotriazolyl azepine adduct 19, which was isolated as an orange liquid (90 %).

Scheme 7.

Chiral benzotriazolyl‐tetrahydroazepine synthesis.

Whereas the ring expansion could have generated either or both regioisomers 18 and 19 (Scheme 7, pathways A and B), the NMR spectra showed only a single set of signals for the Me and iPr CH₃ groups (δ(¹H) 1.11, 1.08, 1.06 ppm; δ(¹³C) 24.2, 20.1, 17.7 ppm), indicating the formation of a single isomer. Based on the reported selectivity of related asymmetric cyclohexanone substrates In the Beckmann rearrangement, ^[31] we expected the formation of 19 to be favored. Whereas crystals of 19, grown in a MeCN solution at −20 °C, were too temperature sensitive to isolate for X‐ray crystallography, protonation with HOTf in CHCl₃ gave a thermally stable salt (20, 79 %) in sharp contrast to the highly unstable unsubstituted azepinium triflate 5. Crystals suitable for X‐ray analysis were obtained by slow diffusion of pentane into a CH₂Cl₂ solution of 20. Its molecular structure (Figure 11) concurs with the anticipated (2S,5R)‐2‐iPr‐5‐Me regioisomer 19 with both alkyl groups in equatorial positions. Compared to the non‐protonated 4 (Figure 6), the benzotriazolyl group of 20 is rotated by 155° and tilted with respect to the imine group [20: N1−C1−N2−N3=−25.9(3); 4: N1−C1−N2−N3=179.54(8)]. Clearly, protonation of the imine group prevents the intramolecular H‐bonding that facilitated the planar conformation of 4; the N1−H1 hydrogen interacts only with the triflate anion [H1⋅⋅⋅O1=1.97 (2) Å]. The positively charged N1 assumedly causes the slightly tighter benzotriazolyl bonding [20: C1−N2=1.384(3), N2−N3=1.403(3) Å; 4: C1−N2=1.4334(12), N2−N3=1.3746(11) Å]. The parameters of the iminium group [C1−N1=1.280(3) Å, N1−C1−C6=122.68(19)°, C1−N1−C2=125.01(19)°] are comparable to those for 6 a and 6 c (Figure 6).

Figure 11.

Displacement ellipsoid plot of 1‐benzotriazolyl‐(2S,5R)‐2‐isopropyl‐5‐methyl tetrahydroazepinium triflate 20 at the 50 % probability level. The triflate anion and C−H hydrogen atoms are omitted for clarity, with exception of H2 and H5. Selected bond lengths [Å] and angles [°]: C1−N1=1.280(3), C1−N2=1.384(3), N1−C2=1.487(3), N1−H1=0.89(2), C1−C6=1.490(3), N2−N3=1.403(3), N3−N4=1.278(3), N2−C11=1.390(3), N1−C1−N2=117.2(2), N1−C1−C6=122.68(19), N2−C1−C6=120.1(2), C1−N1−C2=125.01(19), N1−C1−N2−N3=−25.9(3), C6−C1−N2−N3=153.6(2).

The introduction of the phosphane group on the chiral ring could not be achieved in analogy with the synthesis of the non‐chiral ligands 7 (Scheme 3): surprisingly, treatment of the protonated precursor 20 with diphenylphosphane yielded its dehydrocoupling product tetraphenyldiphosphane.[ ³² , ³³ , ³⁴ ] Instead, the phosphane group was introduced by treating 19 with lithium phosphides LiPR₂ (R=Ph, nBu) in THF to give the desired chiral cycloiminophosphane 21 in 53 %, after purification by an acid/base work‐up involving salt 22 (Scheme 8). The ³¹P NMR spectrum showed a single resonance at δ 6.6, thus indicating the absence of flippamers, which was attributed to the reduced flexibility of the ring on which the iPr and Me substituents favor equatorial positions.

Scheme 8.

Access to asymmetric ligands 21 from chiral benzotriazolyltetrahydroazepine 19.

Single crystals of 22 suitable for an X‐ray structure determination were obtained by slow diffusion of pentane into a saturated CH₂Cl₂ solution. The molecular structure (Figure 12) shows a tetrahydroazepine chair similar to the one in 20 with the (2S)‐iPr and (5R)‐Me indeed in equatorial positions and confirms that the stereochemical information of the l‐menthone is retained over the synthesis. The conformation and bonding parameters of 22 compare closely to that of the achiral, unsubstituted 6 a (Figure 7) [22: C1−N1=1.289(2) Å, P1−C1=1.8252(18) Å, N1−C1−P1=121.79(14)°; 6 a: N1−C1=1.2858(16) Å, P1−C1=1.8269(13) Å, N1−C1−P1=123.06(10)°].

Figure 12.

Displacement ellipsoid plot of (2S,5R)‐phosphaazepinium triflate 22 at the 50 % probability level. The triflate anion and C−H hydrogen atoms are omitted for clarity, with exception of H2 and H5. Selected bond lengths [Å] and angles [°]: P1−C1=1.8252(18), P1−C11=1.8217(19), P1−C17=1.8259(18), C1−N1=1.289(2), N1−C2=1.490(2), N1−C1−P1=121.79(14), N1−C1−C6=120.82 (16), C1−N1−C2=124.57(16).

Coordination of chiral ligand 21 to Ru^II gave the corresponding P‐κ¹ complexes 23 (Scheme 9; 19 %, δ(³¹P) 24.8 ppm). Likewise, coordination of 21 to Rh^III afforded κ¹‐Rh^III complex 24 that showed, akin to complex 12 a (see above), a broad ³¹P NMR signal at room temperature, which resolved at −90 °C into a series of doublets with two major P‐κ¹ resonances (δ  34.3 ppm, ¹ J _P,Rh=140.9 Hz, 38 %; δ 22.8 ppm, ¹ J _P,Rh=142.5 Hz, 42 %; Figure 13) and four minor ones with couplings indicative of P‐κ¹ and κ² bonding (δ 29.9 ppm, ¹ J _P,Rh=140.9 Hz, 7 %; δ 26.5 ppm, ¹ J _P,Rh=137.7 Hz, 7 %; δ 24.7 ppm, ¹ J _P,Rh=137.7 Hz, 4 %; δ 20.7 ppm, ¹ J _P,Rh=115.0 Hz, 2 %). The P‐κ¹ signals likely reflect different rotamers, as the absence of ³¹P,³¹P couplings rules out μ‐complexation. In contrast to 12 a, no N‐κ¹ signal was detected for 24, presumably because the adjacent iPr group discourages coordination at this site. Chloride abstraction converted 24 and its isomers to κ²‐25 (δ(³¹P, CH₂Cl₂) −11.3 ppm, ¹ J _P,Rh=106.6 Hz), which was calculated to be energetically favored by 3.1 kcal mol⁻¹ over its epimer 25* (ωB97XD/6‐31+G(d,p), Def2‐TZVP for Rh).[ ^6b , ^6d , ³⁵ , ³⁶ , ³⁷ , ³⁸ , ³⁹ ] The obtained chiral transition metal complexes might be useful for asymmetric catalytic reactions, but such investigations were outside the scope of the present study. Based on the performance of ligands 7 (see above), asymmetric transfer hydrogenation seems a promising starting point.

Scheme 9.

The coordination of ligand 21 to Ru^II and Rh^III.

Figure 13.

³¹P NMR spectra of 24 at room temperature and −90 °C.

Conclusion

This study reports the synthesis of cyclic 2‐phospha‐tetrahydroazepines as novel 1,3‐P,N ligands with the intent of opening new opportunities in coordination chemistry and catalysis. This class of cycloiminophosphanes contains a seven‐membered aliphatic imine ring with which it complements other classes of 1,3‐P,N ligands, including the aromatic 2‐pyridyl‐ and 2‐imidazolyl‐phosphanes, as well as the recently reported acyclic iminophosphanes and phosphaamidines. The ligands were readily obtained in a one‐pot process through a Beckmann rearrangement of cyclohexanones to reactive nitrilium ion intermediates, which were trapped with benzotriazole. The benzotriazole was then quantitatively replaced with a secondary phosphane (R=Ph, nBu, Cy), facilitated by triflic acid activation. With respect to other 1,3,‐P,N ligands, these cycloiminophosphanes distinguish themselves by their high N‐basicity and their flexible backbone, as ³¹P NMR spectroscopy of the neutral ligands reveals the presence of flippamers, indicative of dynamic conformational behavior of the tetrahydroazepine ring. The ligands coordinate in both a P‐κ¹ and a P,N‐κ² fashion to W(carbonyl) complexes, which were analyzed by IR spectroscopy to quantify the ligands′ donor strength. Coordination to [RhCp*Cl₂]₂ gave a dynamic mixture of κ¹ and κ² complexes that, on treatment with silver triflate, lead only to the κ² complexes. Treatment with [Ru(p‐cym)Cl₂]₂ selectively provides P‐κ¹ complexes, which were also effective catalysts for the hydration of benzonitrile (1.4 mol %, 180 °C, 3 h, up to 79 %) and the transfer hydrogenation of cyclohexanone (0.5 mol %, 83 °C, 2 h, quant.); for the latter reaction iridium(I) could also be used (1 mol %, 83 °C, 4 h, quant.). Finally, as a preamble to asymmetric catalysis, a chiral cycloiminophosphane could be accessed from the natural precursor l‐menthone in a selective Beckmann rearrangement. It was characterized by X‐ray crystallography, and used to access Rh^III and Ru^II complexes. These chiral ligands form promising candidates for the future study of asymmetric 1,3‐P,N catalysis.

Experimental Section

Preparation of compounds: The syntheses and full characterization of 4–7, 9–14, 19–25, the Lewis acid catalyst screening of 4, and the P‐κ ¹ and κ ² Rh^III complexes of 9 are described in full detail (14 pages) in the Supporting Information, which also contains their ¹H, ¹³C{¹H}, ¹⁹F{¹H}, and ³¹P NMR spectra (37 pages).

Computational procedure: Density functional calculations were performed at the ωB97X‐D ^[36] level of theory using Gaussian09, revision A.02. ^[37] Geometry optimizations were performed using the 6‐31+G(d,p) ^[38] basis set (Def2‐TZVP for Rh) ^[39] and the nature of each stationary point (see the Supporting Information) was confirmed by frequency calculations.

X‐ray crystallography: Deposition Numbers 2084404 (for 4), 2084405 (for 6 a), 2084406 (for 6 c), 2084407 (for 10), 2084408 (for 20), and 2084409 (for 22) contain the supplementary crystallographic data for this paper. These data are provided free of charge by the joint Cambridge Crystallographic Data Centre and Fachinformationszentrum Karlsruhe Access Structures service.

Conflict of interest

The authors declare no conflict of interest.

Supporting information

As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer reviewed and may be re‐organized for online delivery, but are not copy‐edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors.

Supporting Information

M. K. Rong, F. Holtrop, E. O. Bobylev, M. Nieger, A. W. Ehlers, J. C. Slootweg, K. Lammertsma, Chem. Eur. J. 2021, 27, 14007.