Formation of 1-aza-2-silacyclopentanes and unexpected products from the insertion of phenylisocyanate into 2,2-dimethyl-1-(trimethylsilyl)-1-aza-2-silacyclopentane

Synthesis of cyclic N-aminosilanes

All prepared cyclic N-aminosilanes are shown in Table 1. The starting materials can be heated at reflux and/or distilled without any evidence of the formation of cyclic products. Heating with sodium alcoholates (procedure E, similar to Ziche et al., 1996) is not as efficient as procedure D (see Table 3 in the Experimental section). According to the patent literature, the synthesis and isolation of unsubstituted cyclic N-aminosilanes is not possible (Pepe et al., 1994; Arkles et al., 2003). Nevertheless, Kirilin et al. (2009) reported the isolation of 2,2-diethoxy-1-aza-2-silacyclopentane with procedure B without mentioning the reaction time.

For the synthesis of the N-aminosilanes 2a–e in good yields and high purity, only procedure D is suitable. Heating without ammonium sulfate (procedures A and B) leaves the reactants mostly unchanged in the given time. Procedure C yields undefined polymers.

Ammonium sulfate is known as a catalyst for the transsilylation reaction of amines with HMDS. Therefore, it might be assumed that ammonium sulfate catalyses the silylation of the amine moiety. The lower boiling distillate was characterised to be the corresponding trimethylalkoxysilane that is formed by the intramolecular condensation of the N-trimethylsilylated 3-aminopropylsilane. The N-unsubstituted cyclic products could not be isolated. All attempts using procedure C yielded undefined polymers. Without a catalyst (procedure A), only small amounts have reacted in the given time, and no N-unsubstituted cyclic product could be found but N-trimethylsilylated ones. Also, no distillate is obtained by heating pure N-trimethylsilyl-3-aminopropyltrimethoxysilane with ammonium sulfate, but undefined polymers were found in the still pot showing similar ²⁹Si nuclear magnetic resonance (NMR) signals like in experiments using procedure C with 1a. Excess of HMDS is needed for the synthesis of cyclic N-aminosilanes, which reduces the tendency of polymerisation, most likely because of steric effects by silylation of the remaining N-H bond. The presence of the SiMe₃ moiety is confirmed by the NMR and infrared spectra, the elemental analysis (EA) of 2a, the determination of molar mass of 2a by ebullioscopy in cyclohexane, and the expected [M+H]⁺ peak in the mass spectrum of 2c. The N-trimethylsilylated products also tend to polymerise, forming high boiling macromolecules upon reflux but with much smaller reaction rates. Therefore, in all experiments performed with procedure D polymerisation, products can be found in the still pot after distillation lowering the yield of the pure cyclic product. On long time standing, the pure N-substituted cyclic N-aminosilane undergoes polymerisation, too. The polymerisation products can be converted into the starting materials by heating with excess of methanol or ethanol, respectively. The stabilisation of 1-aza-2-silacyclopentanes by the exocyclic SiMe₃ moiety is confirmed by heating of N,N-bis(trimethylsilyl)-3-aminopropyltrimethoxysilane with catalytic amounts of ammonium sulfate. In this case, the formation of 2c and trimethylmethoxysilane is observed. N-Methyl-3-aminopropyltrimethoxysilane is less reactive and does not form 2e with a similar yield like the primary amines.

Further derivatives of 2 are accessible starting from 2a. Treatment of 2a with EtMgBr solution in a stoichiometry of 1:1 leads to 2f. A higher amount of EtMgBr per mole 2a does not react with the N-aminosilane to form the twofold substituted product (Scheme 4).

In contrast to this observation, the reaction with ⁿBuLi leads to 2g and 2h depending on the stoichiometric ratio of the reactants. The removal of the byproducts by filtration and washing with n-pentane is not effective. Distillation should be avoided because of possible reactions at high temperature, i.e., polymerisation catalysed by the Grignard reagent or the corresponding ethanolate.

Scheme 4:

Substitution of the ethoxy group in 2a using EtMgBr-tetrahydrofuran solution.

Insertion reactions with PhNCO

The reaction of the cyclic N-aminosilanes with PhNCO results in a mixture of insertion and oligomerisation products of the isocyanate. Only in the reaction mixture obtained from 2d and PhNCO products 3 and 4 are crystallised and analysed (Scheme 5). In both molecules, the exocyclic trimethylsilane moiety is absent. There are two possibilities for this observation: (i) there were N-unsubstituted products in low concentration in the reaction mixture, not detected by NMR and having no influence on the other analyses; and (ii) partial hydrolysis took place. The Si-N bond is known to be moisture sensitive (see also section Alcoholysis reactions). Therefore, it is probable that this moiety is more prone to hydrolysis than the remaining groups in the cyclic N-aminosilanes 2. In view of 3 and 4, it is likely that the first reaction step is the insertion of PhNCO into the exocyclic Si-N bond. The urea moiety activates a second insertion into the endocyclic Si-N bond, probably due to steric reasons. The second molecule PhNCO inserts into the five-membered ring to form a seven-membered one. During this reaction step, the endocyclic Si-N bond is weakened and polymerisation as side reaction is more likely. Therefore, because of the evident partial hydrolysis, a mixture of different molecular species is obtained. Furthermore, isocyanates can undergo oligomerisation reactions, the products of which also are present in the final reaction mixture.

Scheme 5:

Reaction of 2d with PhNCO.

Two different products are crystallised from the reaction mixtures.

The observed insertion reaction could not be monitored by NMR spectroscopy upon melting solutions at −25°C of cyclic N-aminosilane and PhNCO in anhydrous nitromethane frozen separately in the NMR tube. Thus, after warming the mixture up to −25°C, the reaction was completed after seconds.

Alcoholysis reactions

In order to analyse the hydrolysis behaviour of 2a–d, we performed alcoholysis experiments with 1-pentanol and 2a. The results lead to the conclusion that the hydrolysis of both Si-N bonds is preferred compared with the hydrolysis of only one of them (Scheme 6). It is assumed that the first reaction step is the alcoholysis of the exocyclic Si-N bond, because of the steric hindrance of the endocyclic bond. Otherwise, the formation of 4 cannot be explained. The formed N-unsubstituted cyclic N-aminosilane undergoes ring-opening reaction with the alcohol but does not form polymers. In all alcoholysis experiments, 3-aminopropyldiethoxypentoxysilane was formed.

Scheme 6:

Alcoholysis of 2a with n-pentanol.

The cleavage of both Si-N bonds is preferred.

This is similar to the bis-insertion of CO₂ into diaminosilanes of the type (CH₃)₂Si(NHR)₂ to form bis-carbamates of the type (CH₃)₂Si(OCONHR)₂. Mono-insertion products can neither be isolated nor detected (Kraushaar et al., 2012). Similar isocyanate insertions have also been reported (Schöne et al., 2010). Considering these results, we cannot determine if and when the loss of the SiMe₃ moiety took place (via hydrolysis), leading to 3 and 4.

Single crystal X-ray diffraction analyses

Compound 3 crystallises in the orthorhombic space group Pbca with one crystallographically independent molecule in the asymmetric unit. Considering the outcome of the reaction leading to 3, one can state that one molecule of phenylisocyanate is inserted into the azasilacyclopentane ring of 2. This leads to a seven-membered ring with a 2,4-diazasilepane structure (Figure 1). Besides, a phenylcarbamic group replaces the trimethylsilyl group that was present in 2. Thus, a second insertion of an isocyanate molecule into the N-H bond has occurred.

Figure 1:

Molecular structure of 3 showing the atomic numbering scheme.

The thermal ellipsoids of the nonhydrogen atoms are drawn at the 50% probability level.

The ring pucker of the seven-membered ring (Si1-C1-C2-C3-N1-C13-N3) was analysed according to Evans and Boeyens (1989). The least-squares plane of these seven atoms in crystal coordinates is given by −8.404x−1.515y+14.142z=3.070. The out-of-plane displacements of the ring atoms in Å are 0.232 Si1, −0.595 C1, 0.130 C2, 0.602 C3, −0.391 N1, −0.278 C13, and 0.299 N3. These displacements lead to the conclusion that the seven-membered ring is in a boat conformation.

Besides, compound 3 contains a biuret unit (carbamylurea) given by the atoms N2-C6(=O1)-N1-C13(=O2)-N3. This unit is composed of two planar urea subunits with a dihedral angle between these subunits of 20.463(90)°. The phenyl ring C7-C12 is nearly coplanar with the urea subunit N2-C6(=O1)-N1 with a dihedral angle between these planes of 4.576(97)°. In contrast, the phenyl group C14-C19 shows a dihedral angle of 71.787(52)° with the urea subunit N1-C13(=O2)-N3. The angle between the phenyl units is 55.252(48)°. An intramolecular N2-H···O2 interaction (1.88Å, 142°) stabilises the molecular conformation.

The crystal packing is dominated by C-H···O interactions in the range of 2.4–2.7Å (Table 2). A bifurcated contact at O1 (C15-H15···O1, C5-H5···O1) leads to molecular chains along the crystallographic a axis. These chains are connected via C1-H1A···O2 along the crystallographic b axis. Further, C-H···π contacts of aliphatic and aromatic hydrogen atoms with both aryl moieties are in the range of 2.8–2.9Å.

Compound 4 crystallises in the monoclinic space group P2₁/c with two crystallographically independent molecules in the asymmetric unit (Figure 2). The reaction from 2 to 4 involves replacement of the trimethylsilyl group by a phenylcarbamic group. The central urea unit of 4 is bound to a phenyl ring via N2 and to the 1,2-azasilolidine via N1. Both crystallographically independent molecules differ in the conformation of the phenyl ring with respect to the central urea unit formed by N1-C6(=O1)-N2. The dihedral angle between these planes is in molecule A 39.535(82)°, but in molecule B only 18.173(96)°. Intramolecular C8-H8···O1 distances differ from molecules A to B and influences the dihedral angle between the urea unit and the phenyl moiety (A: 2.45Å, 108°; B: 2.25Å, 121°). Five-membered rings may adopt an envelope or half-chair (twisted) conformation (Fuchs, 1978). Ring puckering analysis of 4 shows that the five-membered ring in molecule A adopts a twisted conformation with the atoms C1A and C2A −0.229 and 0.283Å out of the plane of the five-membered ring. The least-squares plane of the five atoms in crystal coordinates is given by 9.646x−2.079y+3.340z=2.250. Molecule B features a disorder at C1 and C2. Both five-membered rings formed by this disorder are in twisted conformation with the atoms C1B/C2B (−0.249/0.299Å) and C1C/C2C (0.282/−0.323Å) alternating above and below the plane formed by the five-membered rings. The least-squares planes of the five-membered rings in crystal coordinates are given by 1.135x−1.756y+18.192z=6.812 for Si1B-N1B-C1B-C2B-C3B and 2.865x+0.622y+ 17.546z=8.763 for Si1B-N1B-C1C-C2C-C3B.

Figure 2:

Molecular structures of the two crystallographically independent molecules of 4 showing the atomic numbering scheme.

Molecule A is labelled with A, molecule B with B. The thermal ellipsoids of the nonhydrogen atoms are drawn at the 50% probability level. Molecule B shows a disorder at the five-membered ring with s.o.f. of 0.85.

In contrast to 3, the crystal packing of 4 is dominated by C-H···π interactions (Nishio, 2004) in the range of 2.56 to 2.71Å. However, the π-system of molecule A is not involved in C-H···π contacts. Instead, the oxygen O1 atom interacts with the NH and two CH groups in strong N-H···O [2.20Å, 159(2)°] and weak C-H···O hydrogen bonding (2.5Å). These interactions occur in the same direction as a possible C-H···π can be localised (C3B-H3D···Cg3, with Cg3 as the centre of gravity of C7A-C12A) and form intermolecular dimers with molecule B. Furthermore, via N-H···O (Table 2) bridges of molecule B, the crystal packing is stabilised.

Synthesis of 2a–e by cyclisation reactions

The 3-aminopropylsilanes 1a–e were refluxed for 2 h with the reactants listed in Table 3. The reaction mixture was distilled under reduced pressure to isolate 2a–e. The catalyst and the amount of HMDS were varied in order to find optimised reaction conditions.

Substitutions reactions with 2a

A solution of 2a and about 20 mL of n-pentane is cooled to −78°C by stirring in a bath of dry ice and isopropanol. Stoichiometric amounts of a solution of ethylmagnesiumbromide (1 m in tetrahydrofuran) or n-butyllithium (2.5 m in hexane) is added dropwise. After warming to room temperature, volatiles are removed in vacuo, and the residue is suspended in n-pentane. The suspension is filtered via a G4 frit, and the solvent is removed in vacuo. Some residues and byproducts remain in the liquid.

Insertion reaction of PhNCO with 2d

The cyclic N-aminosilane 2d (one part) in n-pentane is cooled to 0°C, the stoichiometric amount (two parts) of PhNCO is carefully added via syringe, and the solution is stirred overnight. In the first experiment, 3 crystallised from this mixture overnight. In the second experiment, the solvent is removed in vacuo and replaced with dry CDCl₃. After a few weeks, 4 crystallised from this solution.

Alcoholysis of 2a

Stoichiometric as well as sub-stoichiometric amounts of dry 1-pentanol were added dropwise to a solution of 2a in n-pentane in an ice bath. The ²⁹Si-NMR spectrum shows signals of 2a (δ=2.7, −17.0 ppm) and of the alcoholysis products (δ=16.7ppm (Me₃SiOⁿPent) and −45.4 ppm (3-aminopropyldiethoxypenthoxysilane)).

Single crystal X-ray diffraction analysis

Data collection on 3 and 4 was performed on a STOE IPDS-2T image plate diffractometer (Stoe & Cie. GmbH, Darmstadt, Germany) equipped with a low-temperature device with Mo-K_α radiation (λ=0.71073 Å) using ω and φ scans. Software for data collection: X-AREA, cell refinement: X-AREA and data reduction: X-RED (Stoe and Cie, 2009). Preliminary structure models were derived by direct methods (Sheldrick, 2008), and the structures were refined by full-matrix least-squares calculations based on F² for all reflections using SHELXL (Sheldrick, 2015). The hydrogen atoms at the amine nitrogen atoms N₂ in both structures were localised from the differences in the Fourier maps and refined without restraints. All other hydrogen atoms were included in the models in calculated positions and were refined as constrained to the bonding atoms. Further crystallographic and structure refinement data are listed in Table 4.