Ring-Opening Reaction of 1-Phospha-2-Azanorbornenes via P-N Bond Cleavage and Reversibility Studies

Abstract

The reactive P-N bond in 1-phospha-2-azanorbornenes is readily cleaved by simple alcohols to afford P-chiral 2,3-dihydrophosphole derivatives as a racemic mixture. The isolation of the products was not possible due to the reversibility of the reaction, which could, however, be stopped by sulfurization of the phosphorus atom, thus efficiently blocking the lone pair of electrons, as exemplified for 6b yielding structurally characterized 8b. Additionally, the influence of the substituent in the α position to the phosphorus atom (H, Ph, 2-py, CN) on the reversibility of the reaction was studied. Extensive theoretical calculations for understanding the ring-closing mechanism suggested that a multi-step reaction with one or more intermediates was most probable.

1. Introduction

Phospholes are versatile starting materials for the production of phosphorus heterocyclic compounds [1,2,3,4,5,6]. The tautomerization of poorly aromatic 1H-phospholes leads to 2H-phospholes [7], which readily undergo hetero-Diels–Alder cycloaddition reactions with various dienophiles [8,9,10,11,12,13,14]. Thus, the stereoselective phospha-aza Diels–Alder reaction between 2H-phospholes and N-sulfonyl-α-iminoester affords 1-phospha-2-azanorbornenes (PANs) [9]. The cycloaddition reaction was successfully extended to a variety of previously reported P-substituted 2H-phospholes [15,16,17] to give a range of diastereomeric α-substituted PANs (endo-1a–d, Scheme 1) via a sigmatropic shift of the substituent at higher temperatures [7,18]. The general formation and properties of compounds containing phosphorus–nitrogen bonds have been extensively studied [19,20,21,22,23,24,25,26,27], such as cleavage reactions [28,29,30,31,32]. Thus, the P-N bond of PAN endo-1b can be cleaved by achiral nucleophiles, namely H₂O, H₂S, and EtMgBr, to give 2,3-dihydrophosphole derivatives (5a,b or 3, Scheme 1). Enantiopure lithium alkoxides have also been used as powerful nucleophiles to afford diastereomeric mixtures of 1-alkoxy-2,3-dihydrophospholes (4a–c, Scheme 1), as well as enantiopure compounds obtained via the crystallization of the diastereomers (4b,c, Scheme 1) [33]. Furthermore, the reduction of PAN endo-1b with lithium aluminum hydride yielded a seven-membered phosphorus heterocycle (2, Scheme 1) [34].

All the abovementioned reactions were conducted with endo-1b only, while the reactivity of further PANs has remained largely unexplored. Herein, we report the P-N bond cleavage of PANs endo-1a–d (Scheme 1) with ethanol, resulting in the reversible formation of 1-ethoxy-2,3-dihydrophospholes 6a–d, thus hampering the isolation of the final products. This behavior was already observed during the ring-opening reaction of endo-1b with H₂O and alkoxides [9,33]. Consequently, the nature of this reversibility and its possible mechanism were studied.

2. Results and Discussion

2.1. P-N Bond Cleavage of endo-1b with Alcohols

Test reactions with ethanol as a nucleophile in the P-N bond cleavage were conducted with endo-1b (Scheme 2). Depending on the amounts of EtOH used, no product (5 eq. EtOH, Supplementary Materials, Figure S11), small amounts of 6b (10 or 20 eq. EtOH), or a 1:1 ratio of endo-1b:6b (200 eq. EtOH, Scheme 2) were observed (determined using ³¹P{¹H} NMR spectroscopy in C₆D₆). We assume the reaction occurs in an S_N2 fashion, as shown in Scheme 3. A singlet at 141.7 ppm (cf. endo-1b 57 ppm), which is in the typical range for P^III-O compounds, is assigned to the product 6b, implying P-N bond cleavage and P-O bond formation [35]. When 20 or 200 eq. EtOH was employed, a second singlet appeared at 135.7 ppm, suggesting the formation of another diastereomer, as the small nucleophile can attack the P-N bond in endo-1b from different sides (ratio 4.5:1, according to ³¹P{¹H} NMR spectroscopy).

All attempts to isolate the product 6b failed. When the solvent was evaporated in vacuum, the ring-closing reaction took place, yielding predominantly endo-1b. A similar behavior was already observed for the ring-opening reaction of endo-1b with H₂O and alkoxides [9,33].

Similarly, the P-N bond cleavage reaction in endo-1b with 200 eq. methanol also afforded two isomers of product 7b in a 3:1 ratio, as confirmed by ³¹P{¹H} NMR spectroscopy (C₆D₆) (145.3 and 140 ppm). Interestingly the second isomer was only produced when small nucleophiles were used, corroborating the logical dependence of the nucleophile size and number of products. Thus, in P-N bond cleavage reactions with bulky lithium alkoxides, only one isomer was formed [33]. The P-N cleavage reaction of endo-1b with MeOH is also reversible and undergoes a ring-closing reaction to form endo-1b when the reaction mixture is concentrated in vacuo.

As the lone pair of electrons at phosphorus in 6b and 7b is involved in the ring-closing reaction, the sulfur protection of 6b and 7b stopped the reversibility (Scheme 2) and allowed for a facile isolation of the 8b and 9b products via crystallization. Single crystals of 8b and 9b suitable for X-ray crystallography were obtained by dissolving 8b or 9b in hot isopropanol, followed by cooling at −25 °C overnight.

In the structurally characterized compounds 8b and 9b (Figure 1), the phosphorus atom has a distorted tetrahedral environment. The P-O bond lengths are in the range from 158.8 to 159.5 pm, which corresponds to a P-O single bond. Consistent with the literature, the P=S double bonds in both compounds (191.5–193.8 pm) are in the expected ranges [36].

2.2. Reversibility Studies

In order to study the nature of the reversibility of the P-N bond cleavage reactions with ethanol, we included previously reported 1-phospha-2-azanorbornenes (PANs) endo-1a,c,d (Scheme 1), which have different substituents in the α position to the phosphorus atom. These PANs are also racemic mixtures resulting in racemic products. The enantiomers were never isolated. Firstly, endo-1c was reacted with 200 eq. of EtOH to afford 6c in a ratio of 1:3 (6c:endo-1c, Supplementary Materials, Figure S12). Concentrating the THF solution in vacuum at 20 °C resulted in ring closing mainly yielding the starting material. To verify if the size of the substituent in the α position to phosphorus in PANs correlates with the rate of conversion and reversibility of the reaction, we conducted a P-N bond cleavage reaction with endo-1a, the smallest PAN bearing a hydrogen atom in the α position, with 200 eq. of EtOH (Supplementary Materials, Figure S10) overnight at room temperature. The ³¹P{¹H} NMR spectrum of the reaction mixture showed a 1:1 ratio of endo-1a and 6a. Interestingly, the ring-closing reaction took place only to some extent when the solution was concentrated in vacuum at 20 °C, affording a ratio of 2:1 of endo-1a to product 6a. Finally, endo-1d containing a nitrile group in the α position was reacted with 200 eq. EtOH. The reaction proceeded with the highest conversion (1:8 for endo-1d:6d) compared to all the PANs, as identified using ³¹P{¹H} NMR spectroscopy. Surprisingly, the ring-closing reaction did not take place when concentrating the solution in vacuum, even at 60 °C, suggesting that there is no equilibrium between endo-1d and 6d (Supplementary Materials, Figure S13).

Density functional theory calculations were performed to obtain the relative free energies ΔG of the ring-closing reactions (

Table 1. ΔG values of the P-N bond cleavage reactions of endo-1a–d with EtOH.

Reaction	ΔG/kcal mol−1
endo-1a → 6a	−0.2
endo-1b → 6b	0.8
endo-1c → 6c	0.4
endo-1d → 6d	−1.1

).

These values support the experimental findings and show that the equilibrium is shifted towards the reactants for the reaction of endo-1b,c with EtOH, while in the reaction of endo-1a,d with EtOH, the equilibrium is shifted towards the products. This trend seems to be largely independent of the size of the α substituents of the PANs, as the ³¹P{¹H} NMR spectroscopic reversibility studies did not show any logical dependance.

The reversibility of these reactions could, however, be influenced by the electronic environment around the P atom governed by the substituents. Therefore, we considered the frontier orbitals of the products 6a–d being located either on P or N for a nucleophilic attack of HOMO to LUMO, thus facilitating the ring-closing reaction (for frontier orbitals calculations of 6a–d see Supplementary Materials, Figure S17). Most of the HOMO contributions in 6a–d come from P atomic orbitals, as well as the substituents in the α position. The LUMO contributions in 6a–d are mostly scattered throughout the molecule, mainly on the backbone of the sulfonamide moiety. The largest HOMO-LUMO energy gap was calculated for endo-1a and 6a (

Table 2. Calculated HOMO-LUMO gaps of the different reactants endo-1a–d and products 6a–d.

Compound	Gap/eV	Compound	Gap/eV
endo-1a	12.0	6a	11.3
endo-1b	10.8	6b	10.5
endo-1c	10.4	6c	11.1
endo-1d	11.8	6d	11.2

). However, the calculation of the frontier orbitals did not give a clear explanation for the abovementioned reactivity trends.

To understand the mechanism of the back-reaction, we searched for transition states and intermediates using density functional theory. However, we were not able to obtain a reasonable transition state between the reactants (endo-1a–d) and products (6a–d). This could indicate a multi-step reaction with one or more intermediates. Since EtOH was used in a large excess in the P-N bond cleavage reaction, an additional structure containing the product (6a–d) and a second ethanol molecule was calculated. However, this structure was found to be unstable, suggesting that the excess of EtOH does not play a role in the back-reaction. Furthermore, ab initio molecular dynamics (AIMD) simulations of the products 6b and 6d in ethanol were performed. Interesting developments in the simulations of the product structures 6b (fully reversible) and 6d (completely irreversible) were found. While product 6b looks as expected from the X-ray structure determination and DFT optimizations, the P···N distance in product 6d shows a significant elongation after some simulation time (Figure 2).

In order to analyze the different behaviors of 6b and 6d in ethanol, the radial distribution functions (RDFs) of the P···N distance were calculated (Figure 2, bottom). In the case of 6b, only one peak at about 310 pm with a small shoulder at longer distances can be seen, indicating that the P···N distance stays rather constant at shorter distances. The RDF of 6d shows two broad peaks, with the smaller one at short distances around 320 pm and the slightly bigger peak at around 415 pm. Thus, the P···N distance in 6d is, on average, elongated compared to the one in 6b. This could explain why a back-reaction would not occur in the case of product 6d.

3. Conclusions

In summary, we described the formation of racemic P-chiral 1-ethoxy-2,3-dihydrophosphole derivatives via the nucleophilic cleavage of the P-N bond in 1-phospha-2-azanorbornenes endo-1a–d with ethanol. The isolation of the products was not possible due to the reversibility of the reaction, which could, however, be stopped by sulfurization of the phosphorus atom, thus efficiently blocking the lone pair of electrons, as exemplified for 6b, yielding structurally characterized 8b. Additionally, the influence of the substituent in the α position to the phosphorus atom (endo-1a,c,d) was studied. Extensive theoretical calculations, conducted to understand the ring-closing mechanism, suggested that a multi-step reaction with one or more intermediates was most probable. Overall, the ring-closing reaction seemed to depend strongly on the electronic environment around the phosphorus atom, as a nitrile group in the α position hindered the reversibility. Although the exact mechanism of the ring-closing reaction is not understood, it could be shown that small and electron-withdrawing groups located at the α position to the phosphorus atom gave access to thermodynamically stable 1-ethoxy-dihydrophosphole derivatives. Further mechanistic studies are underway.

4. Materials and Methods

4.1. General Information

All air-sensitive reactions were carried out under dry high-purity nitrogen using the standard Schlenk technique. THF was degassed and distilled from potassium. EtOH and MeOH were degassed and kept over activated molecular sieves. The NMR spectra were recorded with a Bruker Avance DRX 400 spectrometer (¹H NMR 400.13 MHz, ¹³C NMR 100.63 MHz, and ³¹P NMR 161.98 MHz, Bruker, Rheinstetten, Germany). ¹³C{¹H} NMR spectra were recorded as APT spectra. The assignment of the chemical shifts and configurations was performed using the COSY and HSQC techniques. TMS was used as the internal standard in the ¹H NMR spectra and all the other nuclei spectra were referenced to TMS using the Ξ-scale [37]. The numbering scheme of 6d, 8b, and 9b is given in the Supplementary Materials. High-resolution mass spectra (HRMS; ESI) were measured using a Bruker Daltonics APEX II FT-ICR spectrometer (Billerica, MA, USA). IR spectra were obtained with an FTIR spectrometer (Nicolet iS5 FTIR by Thermo Scientific, Waltham, MA, USA) in the range of 400–4000 cm⁻¹ in KBr. The compounds endo-1a–d were prepared according to the literature [9].

4.2. Synthesis

General procedure for the preparation of 8b and 9b.

Alcohol (EtOH: 3.16 mL; MeOH: 2.18 mL; 200 eq., 54 mmol) was added at room temperature to a solution of 120 mg (0.27 mmol) of endo-1b in 5 mL of THF. The reaction mixture was stirred overnight followed by the addition of 13 mg (0.4 mmol) of elemental sulfur and 2 drops of NEt₃. After stirring for 6 h, the solvent was removed in vacuo and the crude product was isolated by dissolving in hot iPrOH, followed by cooling at −25 °C overnight to give 8b or 9b as a white powder.

8b: 103 mg (73%)

¹H NMR (400 MHz, CDCl₃) δ 7.74–7.66 (m, 2H, H-aryl), 7.49–7.23 (m, 6H, H-aryl), 6.14 (bs, 1H, N-H), 4.43–4.30 (m, 2H, H-5b), 4.08–3.92 (m, 2H, H-12), 3.23–3.1 (m, 1H, H-4), 2.54 (s, 3H, H-8a), 2.18 (m, 1H, H-4), 1.69 (s, 3H, H-3a/2a), 1.56 (s, 3H, H-2a/3a), 1.41 (t, J = 7.1 Hz, 3H, H-5c), and 1.17 (t, 3H, H-13) ppm; ¹³C{¹H} NMR (101 MHz, CDCl₃) δ 169.2 (s, C-5a), 156.7 (d, J = 30.1 Hz, C-1 or C-2), 144.1 (s, C-8), 139.0 (d, J = 98.8 Hz, C-1/2), 134.9 (s, C-quart. aryl), 134.0 (s, C-quart. aryl), 132.7 (d, J = 9.8 Hz, C-quart. aryl), 129.2 (d, J = 4.2 Hz, C-1b or C-1c), 128.5 (s, C-1b or C-1c), 128.0 (s, C-1d), 127.7 (s, C-7), 121.7 (s, C-10), 73.5 (d, J = 10.1 Hz, C-5), 64.5 (s, C-5b), 61.7 (d, J = 6.6 Hz, C-12), 42.9 (d, J = 70.9 Hz, C-4), 24.7 (s, C-2a/3a), 22.2 (s, C-8a), 16.5 (d, J = 6.5 Hz, C-13), 15.7 (d, J = 17.2 Hz, C-2a/3a), and 14.2 (s, C-5c) ppm; ³¹P{¹H} NMR (162 MHz, CDCl₃) δ 104.7 (s) ppm; ³¹P NMR (162 MHz, CDCl₃) δ 104.7 (m) ppm; HRMS (ESI+, MeCN), m/z: found: 520.1385; calc. for [M+H]⁺: 520.1376, found: 542.1223; calc. for [M+Na]⁺: 542.1195; found: 558.0936, calc. for [M+K]⁺: 558.0935; found: 1061.2525, calc. for [2M+Na]⁺: 1061.2498; and found: 1077.2222, calc. for [2M+K]⁺: 1077.2238; IR (KBr): ṽ = 3278 (m), 2978 (m), 1728 (s, C=O in ester), 1621 (w), 1596 (m), 1492 (w), 1473 (w), 1441 (w), 1409 (w), 1387 (m), 1292 (s), 1252 (s), 1230 (s), 1183 (s), 1148 (s), 1130 (s), 1105 (s), 1015 (s, O-C from P-O-C₂H₅ fragment), 964 (m), 949 (m), 922 (s, C-C from P-O-C₂H₅ fragment), 874 (s), 861 (s), 846 (m), 823 (s), 807 (s), 788 (m), 755 (s), 701 (s), 685 (s), 668 (s), 652 (s), 626 (s), 607 (s), 575 (s, possibly P=S), 548 (m), 514 (m), 495 (s), and 433 (s) cm⁻¹.

9b: 94 mg (68%)

¹H NMR (400 MHz, CDCl₃) δ 7.67–7.59 (m, 2H, H-aryl), 7.39 (d, J = 8.0 Hz, 1H, H-aryl), 7.32 (m, 2H, H-aryl), 7.26–7.14 (m, 3H, H-aryl), 6.03 (bs, 1H, N-H), 4.30 (m, 2H, H-5b), 3.53 (d, J = 14.1 Hz, 3H, H-12), 3.10 (m, 1H, H-4), 2.47 (s, 3H, H-8a), 2.18–2.06 (m, 1H, H-4), 1.62 (s, 3H, H-2a/3a), 1.48 (s, 3H, H-3a/2a), and 1.34 (t, J = 7.1 Hz, 3H, H-5c) ppm; ¹³C{¹H} NMR (101 MHz, CDCl₃) δ 169.1 (s, C-5a), 156.9 (d, J = 29.8 Hz, C-1/2), 144.1 (C-8), 138.7 (d, J = 98.8 Hz, C-1/2), 134.4 (d, J = 90.3 Hz, C-quart.), 132.6 (d, J = 9.4 Hz, C-quart), 131.9 (s), 129.1 (d, J = 4.3 Hz), 128.5, 127.9 (d, J = 1.9 Hz), 127.6, 73.4 (d, J = 10.3 Hz), 64.5 (s, C-5b), 52.1 (d, J = 6.7 Hz, C-12), 41.9 (d, J = 71.0 Hz, C-4), 24.6 (s, C-2a/3a), 22.1 (s, C-8), 15.6 (d, J = 17.1 Hz, C-2a/3a), and 14.1 (s, C-5c) ppm; ³¹P{¹H} NMR (162 MHz, CDCl₃) δ 107.3 (s) ppm; ³¹P NMR (162 MHz, CDCl₃) δ 107.3 (m) ppm; HRMS (ESI, MeCN/DCM), m/z: found: 520.1385, calc. for [M+H]⁺: 520.1376; found: 528.1059 calc. for [M+Na]⁺: 528.1039; found: 544.0793, calc. for [M+K]⁺: 544.0778; and found: 1033.2223, calc. for [2M+Na]⁺: 1033.2185; IR (KBr): ṽ = 3269 (m), 2978 (w), 2940 (w), 1715 (s, C=O in ester), 1616 (w), 1593 (m), 1487 (w), 1440 (m), 1409 (w), 1381 (m), 1365 (m), 1301 (s), 1249 (s, C-O in ester), 1184 (s), 1140 (s), 1114 (m), 1070 (s), 1033 (s, O-C from P-O-CH₃ fragment), 1011 (s), 965 (s), 908 (m), 877 (w), 859 (s), 846 (m), 846 (m), 822 (w), 792 (w), 765 (s), 747 (s), 726 (s), 698 (s), 646 (s), 615 (m), 600 (m), 582 (m), 567 (s, P=S possibly), 538 (s), 521 (m), 498 (m), 473 (m), and 435 (m) cm⁻¹.

4.3. Reversibility Studies

The reversibility experiments were conducted by adding 200 eq. of EtOH (20 eq. was used in the case of endo-1d) to a solution of the corresponding PAN in 3 mL of THF (50 mg, 0.137 mmol of endo-1a and 27 mmol, 1.6 mL of EtOH; 50 mg, 0.113 mmol of endo-1b and 23 mmol, 1.3 mL of EtOH; 50 mg, 0.113 mmol of endo-1c and 23 mmol, 1.3 mL of EtOH; 50 mg, 0.128 mmol of endo-1d and 2.6 mmol, 0.15 mL of EtOH) at room temperature. After stirring overnight, NMR samples were taken. The solutions were concentrated in vacuo followed by drying for one hour. Then, another NMR sample was taken to study the reversibility (Supplementary Materials, Figures S10–13).

4.4. Computational Details

The quantum chemical calculations were performed with the Orca program version 4.2.1. [38]. All structures were optimized using the B3LYP [39,40] functional with the D3 [41,42] dispersion correction with Becke-Johnson damping and the def2-TZVP [43] basis set. For the SCF cycle, a tight convergence criterium was chosen. To ensure that all optimized structures are minima on the potential energy surface, a subsequent frequency calculation was performed, where all the stationary points resulted in non-negative eigenvalues of the Hessians. The molecular orbitals were calculated at Hartree–Fock level of theory on the DFT-optimized geometries, employing the same level of theory as before.

Additional ab initio molecular dynamics simulations were performed. The initial configurations of the simulation boxes were set up with PACKMOL [44], each containing 128 molecules of ethanol and one molecule of 6b and 6d, respectively, using the experimental density of ethanol at 300 K [45]. The simulations were performed with CP2K version 5.1 [46] using periodic boundary conditions and applying the QUICKSTEP [47] module. The BLYP [40,48] functional and the corresponding BLYP Goedecker–Teter–Hutter pseudopotentials [49,50,51] for core electrons were applied for the DFT part in combination with the molecularly optimized double-zeta basis set (MOLOPT-DZVP-SR-GTH) [52]. To account for dispersion effects, the DFT-D3 correction was used. The simulation boxes were equilibrated for 5 ps at 350 K by applying a thermostat, and a subsequent production run was performed for 27.5 ps at 350 K. A density cutoff of 400 Ry was chosen, as well as a relative cutoff of 40. For the SCF, an accuracy threshold of 10⁻⁶ was applied. The simulations were performed in the canonical (NVT) ensemble using the Nosé-Hoover chain thermostats [53,54,55] and a time step of 0.5 fs. The resulting AIMD trajectories were analyzed using the TRAVIS program [56].

4.5. X-ray Crystallography Data

The data were collected on a Gemini diffractometer (Rigaku Oxford Diffraction) using Mo-Kα radiation and ω-scan rotation. Data reduction was performed with CrysAlisPro [57], including the program SCALE3 ABSPACK for empirical absorption correction. All structures were solved using dual-space methods with SHELXT [58] and the refinement was performed with SHELXL [59]. With the exception of hydrogen atoms at nitrogen, all H atoms were calculated on idealized positions using the riding model. The S=P-O-Me fragment of 9b was disordered with a ratio of 0.866 (6):0.134 (6). Structure figures were generated with DIAMOND-4 [60].

CCDC deposition numbers 2291772 for 8b and 2291773 for 9b contain the supplementary crystallographic data for this paper. These data can be obtained free of charge via https://www.ccdc.cam.ac.uk/structures (or from the Cambridge Crystallographic Data Centre, 12 Union Road, Cambridge CB2 1EZ, UK; fax: (+44)1223-336-033; or [email protected]).