Tunable Rh(I) Fischer carbene complexes for application in the hydroformylation of 1-octene Journal of Organometallic Chemistry

The preparation of a series of rhodium(I) complexes coordinated by various electronically tuneable Fischer carbene (FC) ligands, is reported. The Rh(I) metal complexes ’ electronic properties could readily be modulated by variation of a p - N,N- dimethylaniline moiety with a ruthenocenyl substituent, or alternatively, substituting the carbene O -heteroatom for an amino-group. The electronic properties of the complexes were evaluated, and it was determined from the Tolman electronic parameters that the donor-ability of the FC ligands are comparable to N -heterocyclic carbenes. Furthermore, the facile control of the electronic properties of the complexes was demonstrated by mild oxidation of a ferrocenyl ami- nocarbene rhodium(I) complex, yielding the corresponding ferrocenium rhodium(I) complex cation. Finally, the complexes were evaluated as catalyst precursors for the hydroformylation of 1-octene. ©


Introduction
The use of mono-heteroatom-stabilized Fischer carbene ligands in catalytic applications are still predominantly unexplored, although the exploitation of Fischer carbene complexes (FCCs) in photophysical, optical and sensing applications are on the rise [1]. This dearth in homogeneous catalysis application can mostly be ascribed to the challenges related to the preparation of FCCs of the late transition metals [2], whereby self-dimerisation of the carbene ligand to the corresponding olefin is observed during transmetallation reactions from the group 6 metal FCC precursors [3]. Examples of late transition metal FCCs isolated via this route are therefore not common [4]. This methodology has the benefit of bypassing the requirement for the presence of modifiable carbonyl or isonitrile ligands for nucleophilic attack in traditional Fischer carbene synthesis [5]. Even for the few isolated examples of rhodium(I) FCCs, thermal instability frequently prohibits the use of the complexes in applications requiring elevated temperatures [3]. We have previously found that the use of a donating ferrocenylcarbene ligand of the Fischer-type, was effective towards the preparation of Rh I FCCs, stable to both atmospheric conditions and high temperatures [4a]. More importantly, we could show that these complexes were efficient in catalyzing the hydroformylation of 1-octene. The catalyst selectivity could be modified by variation of both co-ligands (phosphines, phosphites, 1,5-cyclooctadiene (cod), carbonyls or arsines) and Fischer carbene heteroatom substituents (amino-vs alkoxy). The formation of the desired linear aldehyde products could hereby be favoured to yield n/iso-aldehyde ratios comparable to those mediated by known rhodium(I) Nheterocyclic carbene (NHC) complex catalysts [6]. In this work, we investigate the use of different (hetero)aryl Fischer carbene ligand substituents as an alternative route towards the tailoring of rhodium(I) FCC precatalysts. In particular, the use of metal-containing moieties as carbene substituents are employed as a tool to extend the carbene ligand reactivity beyond what is possible with classic organic (hetero)aryl ring substituents, where for example, an electron-rich ferrocenyl (Fc) or ruthenocenyl (Rc) group could act as a redox-noninnocent electron reservoir [2d].

Synthesis and characterization
Pentacarbonyl tungsten(0) ethoxycarbene complexes, containing the strongly electron-donating carbene substituent p-N,Ndimethylaniline (p-DMA) [7], or the organometallic ruthenocenyl (Rc) entity [8], were employed as the precursor group 6 FCCs for the preparation of the corresponding Rh I FCCs. [Rh(cod)Cl{C(OEt)p-DMA}] (1) and [Rh(cod)Cl{C(OEt)Rc}] (2) were synthesized (see Scheme 1), following a similar methodology as employed for the analogous complex with a ferrocenyl (Fc) carbene substituent, [Rh(cod)Cl{C(OEt)Fc}] (3) (cod ¼ 1,5-cyclooctadiene) [4a]. The carbene transfer reaction ensues with stirring of an equimolar mixture of the group 6 precursor FCC with the dimer [Rh(cod)Cl] 2 , in dichloromethane at room temperature, while the reaction progress was monitored with thin layer chromatography. Substitution of the cod co-ligand with two carbonyl ligands has previously proven to yield catalyst precursor complexes that are more efficient for the hydroformylation reaction, ascribed to the elimination of the first step of cod-substitution and catalyst activation, amongst other factors [4a]. Thus, complexes 1e3 were treated with carbon monoxide to yield the corresponding dicarbonyl complexes 4e6 (Scheme 2(a)). Similarly, a simple carbene ligand-modification can also tune the carbene ligand from the more electrophilic alkoxycarbenes to the stronger donor aminocarbenes. Aminolysis of the ethoxycarbenes 1 and 3 with n-propylamine yielded complexes 7 [Rh(cod)Cl{C(NH n Pr)p-DMA}] and 8 [Rh(CO) 2 Cl{C(NH n Pr)Fc}], which could be converted to the dicarbonyl analogues 9 [Rh(cod)Cl {C(NH n Pr)p-DMA}] and 10 [Rh(CO) 2 Cl{C(NH n Pr)Fc}], as described above (Scheme 2(a)). The complexes were characterized spectroscopically (see Experimental section 4.2), and a summary of the 13 C NMR and FT-IR data is given in Table 1. Upfield shifts, ranging between 19.9 and 21.2 ppm, are seen in the 13 C NMR spectra for the carbene carbon atom resonances of the dicarbonyl ethoxycarbene complexes 4e6 compared to the cod-precursors 1e3. Similarly, upfield shifts from the carbene carbon resonances of the codcomplexes 7 (252.2 ppm) and 8 (257.9 ppm) to the dicarbonyl aminocarbene-analogues 9 (233.8 ppm) and 10 (237.8 ppm), respectively, are observed (Table 1). As expected, a more marked shielding effect is observed when comparing the aminocarbene derivatives 7e10 with their ethoxycarbene precursor complexes 1, 3, 4 and 6, respectively. Upfield shifts ranging from 43.3 to 51.5 ppm are indicative of the increased electron-donation and stabilization from the N-heteroatoms towards the electrophilic carbene carbon compared to the ethoxycarbene O-heteroatoms [9].
Of interest in this work, however, is the more subtle effect of aryl/metallocenyl carbene substituent variation on the electronic properties of the carbene ligand. As mentioned before, the overall donating ability of the aryl carbene substituent is requisite for the prevention of carbene self-dimerisation. Using the 13 C NMR resonances of the different carbene carbon atoms as an indicative tool for the individual carbene ligand electrophilicity, a trend can be established in the order of p-DMA < Rc < Fc (compare complex series 1e3 and 4e6, Table 1). This is reflected again for the aminocarbene complex series, which excludes the ruthenocenyl substituent for the complexes 7e10 (Table 1), where the p-DMA carbene substituent consistently leads to higher field carbene carbon atom signals compared to the Fc-analogues. However, the differences are small, and the use of 13 C d (C carbene ) is not unambiguous.
Crystals suitable for single crystal X-ray diffraction were grown from layered hexane/CH 2 Cl 2 solutions, for complexes 1, 2, 4, 6, 7 and 9 (see Fig. 1 and Table 2 for selected bond lengths and angles). The preparation of complex 6 has been previously reported, but the characterization did not include the molecular structure [4a]. In all cases, a (pseudo)square planar geometry is observed for the ligands surrounding the central rhodium(I) ion. In general the OeC carbene eC Ar bond angle is smaller for the ethoxy-FCCs 1, 2, 4 and 6 (ranging between 110.23(15) e112.8 (2) ), compared to the NeC carbene eC Ar bond angles of 7 (116.40 (14) ) and 9 (118.42 (16) ). This is indicative of the greater sp 2 -character of the aminocarbene N-atoms, also reflected in the C carbene eN bond lengths of 1.470(2) Å and 1.304(2) Å, for 7 and 9, respectively, as well as in the markedly longer RheC carbene bond lengths for 7 and 9 (2.006(16) Å and 2.0608(18) Å), compared to the analogous ethoxy-FCCs where the RheC carbene bond lengths are 1.970(3) Å and 2.046(3) Å, for 1 and 4 respectively.

Catalytic studies
Complexes 6, 8 and 10 were previously screened as catalyst precursors for the hydroformylation reaction [4a], and bear the redox-active ferrocenyl unit as an access point to a carbene ligand for redox-switchable catalysts [12]. This strategy has become an increasingly attractive tool for catalyst tailoring in recent years, and we wanted to exploit this approach to fine-tune the chemo-and regioselectivity performance of the precatalysts. We have previously demonstrated the stability of the oxidized ferrocenium ethoxycarbene ligand, and could isolate the first examples of the radical cations of such ferrocenyl FCCs [13]. The ferrocenyl aminocarbene complex 10 was chemically oxidized using a stoichiometric amount of the oxidizing agent, acetylferrocenium hexafluorophosphate [FcOAc]PF 6 at room temperature in dichloromethane [14]. The suitability of the chemical oxidizing agent was previously determined through cyclic voltammetry experiments, to determine the required oxidation potential [4a].The ferrocenium carbene complex cation [10] þ was isolated as the hexafluorophosphate salt. Although recording of the NMR spectra was not possible for this paramagnetic compound (except for the 19 F NMR spectrum confirming presence of PF 6 counterion, Appendix A, Fig. S16), the IR stretching frequency of the carbonyl ligands could be measured (Appendix A, Fig. S1), allowing for the calculation of a TEP value of 2063 cm À1 . This shift of the absorption bands to higher energy absorption bands (D14 cm À1 is observed for the TEP of 10 vs [10] þ ), is consistent with the localization of the positive charge on the terminal ferrocenium group, and not on the Rh I centre bonded directly to the carbonyl ligands [12c,13a,13c]. However, the high- Fig. 1. Solid-state molecular structures of complexes 1, 2, 4, 6, 7 and 9. Thermal ellipsoids are shown at 50% probability. Hydrogen atoms are omitted for clarity and a partial atom numbering scheme included. frequency TEP of [10] þ unambiguously confirms the weaker donating ability of the oxidized FC-ligand, as expected by the electron-withdrawing nature of the ferrocenium group [13c,15].
The catalytic activity of complexes 1, 2, 4, 5, 7, 9 and [10] þ in the hydroformylation of 1-octene was evaluated, and compared with the results obtained previously for the ferrocenylcarbene complexes 3, 6, 8 and 10, under the optimized reaction conditions of 40 bar syngas (1:1 CO:H 2 ) and 75 C over 4 h (Table 3). Excellent conversions and chemoselectivity were observed for the ethoxycarbene complexes 1e3 and 5, 6, where the total aldehyde yield exceeded 98% in all cases with complete conversion, with the exception of 4 (79% conversion and 64% total aldehyde yield, respectively, as depicted in Table 3 (entries 1e6)). This is consistent with our previous finding that the more electrophilic ethoxycarbenes show improved chemoselectivity compared to their more stabilized aminocarbene counterparts 7e10 [4a]. Moreover, the TOFs reflect the same trend as a measure of the activity of the ethoxycarbene complexes 1e3 and 5, where the TOFs range from 418 to 615 h À1 . Correspondingly, the dicarbonyl ethoxycarbene 4 with p-DMA carbene substituent displayed the lowest turn-over frequency (TOF ¼ 318 h À1 ) with the lowest total conversion and aldehyde yield. Notably, comparison of the three aryl ethoxycarbene substituents employed, p-DMA, Rc and Fc, shows that both the ruthenocenylcarbene complexes 2 and 5 displayed the highest TOFs (608 h À1 and 615 h À1 ), where the possibility of a secondary, catalytically active Ru II metal centre acting in a synergistic manner with the Rh I catalyst, can not be excluded . However, comparison of the effect of the co-ligands (cod vs dicarbonyl) on either the catalyst activity (TOF and % conversion) or chemoselectivity for aldehyde formation (% total aldehyde yield) is unambiguous. In the case of Rc-substituted FCCs 2 and 5, replacement of the cod ligand in 2 with two carbonyl ligands in 5, is accompanied by a slight increase in the activity as the TOF increases from 608 h À1 to 615 h À1 . This is again consistent with our previous  (5) 86.02 (8) 85.78 (6) 89.23 (5) 87.35 (5) O/N-C carbene -C ipso 111.7(2) 110.23 (15) 112.8 (2) 111.87 (18) 116.40 (14) 118.42(16) (7).  (5 mL findings that an overall increase in the electrophilicity of the FCCs result in more active catalyst precursors. However, the opposite trend is observed for the p-DMA and Fc FCCs. In these cases, substitution of the cod ligands of 1 and 3 (TOF ¼ 543 and 418 h À1 , respectively) coincides with a decrease for the analogous dicarbonyl FCCs 4 and 6 (TOF ¼ 318 and 379 h À1 , respectively). In addition, the chemoselectivity for aldehyde formation over internal alkenes, remains virtually unaffected by this co-ligand substitution; except in the case of 4 where the total aldehyde yield significantly decreases from the 99% for its precursor 1, to 64% for 4. In the case of the aminocarbene complexes 7e10, complete conversions are observed (Table 3, entries 7e12), however lower total aldehyde product formation is observed for these catalyst precursors (82e90%), that is indicative of less chemoselective (pre) catalysts with an accompanying increase in regioselectivity (higher n/iso values ranging between 1.25 and 1.33). The n/iso ratio of the aldehydes range from 0.79 to 0.98 for the ethoxycarbenes 1e3, 5 and 6, while the least active precatalyst 4 demonstrates the highest overall regioselectivity (2.44, entry 4 Table 3). This again is in accordance with our previous finding that a decrease in chemoselectivity is accompanied by an increase in the regioselectivity for the desired linear aldehydes [4a]. Variation of the reaction time from that of the established optimum 4 h, to half the time (2 h) was done for complex 7 (compare entries 7 and 8, Table 3). As expected, longer reaction time yields a higher % total aldehyde yield (82% vs 70%), as well as an increased TOF (512 h À1 vs 386 h À1 ), but it is accompanied by a significantly lower regioselectivity for the linear aldehyde with an n/iso ratio of 1.32 compared to the shorter reaction time where 7 achieves n/iso ¼ 1.95 in 2 h.
These results are comparable with analogous rhodium(I) Nheterocyclic carbene complexes that displayed TOFs varying between 480 and 3540 h À1 , but were accompanied with n/iso values lower than 0.5 [16,17]. However, no clear trend could be established between the donating character of the carbene ligands (TEPs) and the TOFs, in isolation of the effect of the co-ligand substitution on the overall electrophilicity of the prepared (pre)catalysts.
The in situ chemical oxidation of 10 to [10] þ during the catalytic reaction was achieved by addition of a stoichiometric amount of the oxidant acetylferrocenium salt in the reaction vessel. The modification of the carbene ligand to a significantly more electrophilic carbene ligand (vide supra) yielded a notably increased regioselectivity result with an n/iso ratio of 2.25 (entry 13, Table 3) compared to 1.33 for the neutral precursor 10 (entry 12, Table 3). Again, the increased regioselectivity occurs at the expense of both the chemoselectivity (68% total aldehyde yield) and the activity of the complex (TOF 167 h À1 ).
A mercury drop-test was performed for 7 with no significant change in either the conversion or the product yields (entry 9, Table 3), compared to the mercury-free reaction (entry 7, Table 3), indicative of a homogeneous catalyst mode-of-action [18].

Conclusions
In summary, new examples of electronically tuneable and thermally stable Rh I FCCs were isolated, with the electronic properties modulated by the variation of electron-donating aryl carbene substituents, p-N,N-dimethylaniline and ruthenocenyl. Exploitation of the redox activity of the ferrocenyl FC ligand was achieved by chemical oxidation to modulate catalytic performance. The performance of these complexes as precatalysts for 1-octene hydroformylation was compared to the previously evaluated ferrocenylcarbene analogues. The carbene ligand modification yielded catalysts with higher TOFs in the case of the ruthenocenyl FCCs 2 and 5, while the highest regioselectivity was achieved for the p-DMA FCC 4.

Methods and materials
The preparation, purification and reactions of the complexes described were carried out under an atmosphere of dry, oxygenfree N 2 or Ar gas using standard Schlenk techniques. All reactions were mechanically stirred and monitored by IR spectroscopy where relevant. The precursors [W(CO) 5 {C(OEt)Ar}] [Ar ¼ p-DMA [7], Fc [4a], Rc [8]] and [Rh(cod)Cl] 2 were prepared according to literature procedures. Preparation of the rhodium carbene complexes 3, 6, 8, and 10 has previously been reported by our research group [4a]. Aluminum oxide 60 (particle size 0.05e0.15 mm) was used as resin for all column chromatography separations. Anhydrous tetrahydrofuran (THF), diethyl ether (Et 2 O), and n-hexane were distilled over sodium metal and dichloromethane (DCM) was distilled over CaH 2 . All other reagents are commercially available and were used as received.
Nuclear magnetic resonance (NMR) spectra were recorded on Bruker AvanceeIIIe400 and Bruker AvanceeIIIe500 spectrometers using CDCl 3 , CD 2 Cl 2 , and C 6 D 6 as solvents at 25 C. The NMR spectra were recorded for 1 H at 400.13 and 500.13 MHz, and for 13 C at 100.63 and 125.78 MHz. The 19 F NMR spectrum of complex [10] þ was recorded at 376.46 MHz. Infrared spectroscopy was performed on a PerkinElmer Spectrum FT-IR spectrophotometer over the range 3600e1600 cm À1 . Solution IR spectra were recorded in CH 2 Cl 2 using a NaCl cell with a path length of ca. 1.0 mm. Melting points were measured with a Stuart SMP10 melting point apparatus.
All crystals for single-crystal X-ray diffraction were grown by slow diffusion of n-hexane into a concentrated CH 2 Cl 2 solution of the carbene complex at 4 C. Single crystal X-ray diffraction data for complexes 1, 2 and 4 were collected at 173 K (1, 4, 6, 7 and 9) or at 150 K (2), on a Bruker Apex II CCD diffractometer (1, 6,7,6,7,9), while data for complexes 2 and 4 were collected using a Bruker Venture D8 Photon CMOS diffractometer, with a graphitemonochromated Mo-K a (l ¼ 0.71073 Å) radiation using an Oxford Cryostream 600 cooler. All data reductions were carried out using the program SAINTþ, version 6.02 [19] and empirical absorption corrections were made using SADABS [19]. Space group assignments were made using XPREP [19]. The structures were solved in the WinGX [20] Suite of programs, using intrinsic phasing through SHELXT [21] and refined using full-matrix least-squares/difference Fourier techniques on F 2 using SHELXL-2017 [21]. All C-bound H atoms were placed at idealized positions and refined as riding atoms with isotropic parameters 1.2 times those of their parent atoms. All diagrams and publication material were generated using OLEX2, ORTEP-3 [20] and PLATON [22]. Experimental details of the X-Ray analyses are provided in Table S1.
Mass spectral analyses were performed on a Bruker Compact Q-TOF mass spectrometer (Bruker Daltonics, Bremen, Germany) with a positive electron spray as the ionization technique by direct infusion at 0.3 mL min À1 . The m/z values were measured in the range of 50e1000 in acetonitrile. Prior to analysis, the instrument was calibrated with sodium formate (5 mM) in resolution mode. Elemental analyses were carried out using an Elementar vari-oELcube CHNSeO analyser.
The hydroformylation reactions were carried out in duplicates or triplicates, in 90 mL stainless steel pipe reactors. Each reactor was charged with the catalyst precursor (5.096 Â 10 À3 mmol), the substrate: 1-octene (0.805 g, 7.175 mmol), and the internal standard: n-decane (0.204 g, 1.435 mmol) all dissolved in toluene (5 mL). The reactor was sealed, purged three times with N 2 (g) and twice with syngas (1:1, CO/H 2 ) before heating to the desired temperature, under the desired syngas pressure. Once the reaction time was reached, the reactor was depressurized and the reaction mixture was allowed to cool to room temperature before analysing using gas chromatography. Authentic aldehyde and iso-octene standards were used for the confirmation of the products afforded.

Declaration of competing interest
Authors do not have any conflicts of interest to declare.