Zwitterionic amidinates as effective ligands for platinum nanoparticle hydrogenation catalysts

Zwitterionic amidinates as effective ligands for platinum nanoparticle hydrogenation catalysts† L. M. Mart´ ınez-Prieto,* a I. Cano, a A. M´ arquez, b E. A. Baquero, a S. Tricard, a L. Cusinato, a I. del Rosal, a R. Poteau, a Y. Coppel, c K. Philippot, c B. Chaudret, a J. C´ ampora* b and P. W. N. M. van Leeuwen* a Ligand control of metal nanoparticles (MNPs) is rapidly gaining importance as ligands can stabilize the MNPs and regulate their catalytic properties. Herein we report the first example of Pt NPs ligated by imidazoliumamidinate ligands that bind strongly through the amidinate anion to the platinum surface atoms. The binding was established by 15 N NMR spectroscopy, a precedent for nitrogen ligands on MNPs, and XPS. Both monodentate and bidentate coordination modes were found. DFT showed a high bonding energy of up to 48 kcal mol 1 for bidentate bonding to two adjacent metal atoms, which decreased to 28 4 kcal mol 1 for monodentate bonding in the absence of impediments by other ligands. While the surface is densely covered with ligands, both IR and 13 CMASNMRspectraprovedtheadsorptionofCOonthe surface and thus the availability of sites for catalysis. A particle size dependent Knight shift was observed in the 13 C MAS NMR spectra for the atoms that coordinate to the surface, but for small particles, 1.2 nm, it almost vanished, as theory for MNPs predicts; this had not been experimentally verified before. The Pt NPs were found to be catalysts for the hydrogenation of ketones and a notable ligand effect was observed in the hydrogenation of electron-poor carbonyl groups. The catalytic activity is influenced by remote electron donor/acceptor groups introduced in the aryl-N-substituents of the amidinates; p-anisyl groups on the ligand gave catalysts several times faster the ligand containing p-chlorophenyl groups. Introduction The interest in metal nanoparticles (MNPs) is growing fast in both the academic and industrial community thanks to their applications in multiple elds such as sensors, medicine, optoelectronics and catalysis. 1 In particular, MNPs have a high catalytic activity in some specic transformations like hydrogenation, polymerization, oxidation and C–C coupling reactions. 2 Their high potential in catalysis arises from their particular electronic conguration and large surface area, and the potential to combine the advantages of homogeneous and heterogeneous catalysts. The stability and activity of MNPs are strongly inuenced by the ligands used as stabilizers, which are able to modify their surface properties. 3 The exploration of new families of ligands capable to stabilize MNPs and modify their reactivity is always a challenge. In a recent publication, the betaine-type adduct of N-heterocyclic carbenes (NHCs) and carbodiimides (Scheme 1), in particular 1,3-dicyclo- hexylimidazolium-2-di-p-tolylcarbodiimide (ICy$ (p-tol) NCN), was identied as an effective ligand to produce ultra-small ruthenium nanoparticles (Ru NPs) with a size ca. 1nm. 4 Due to their zwitterionic structure, the nitrogen atoms present a large electrondonor capability and coordinate strongly to transition metals. The coordination chemistry of such imidazolium-amidinate ligands has been investigated a short while ago by some of us. 5 In analogy with amidinates, which frequently coordinate in am 2 -k 1 N, k 1 N0bridging mode, ICy$ (p-tol) NCN forms multi-bridged binuclear “paddlewheel”complexes of copper(I). Besides, small Scheme 1 Synthesis of zwitterionic imidazolium-amidinate ligands. a LPCNO, Laboratoire de Physique et Chimie des Nano-Objets, UMR5215 INSA-CNRS-UPS, Institut des Sciences Appliqu´ ees, 135, Avenue de Rangueil, F-31077 Toulouse, France. E-mail: vanleeu[email protected]; [email protected] b Instituto de Investigaciones Qu´ ımicas, CSIC-Universidad de Sevilla, C/Am´ erico Vespucio, 49, 41092 Sevilla, Spain. E-mail: [email protected] c CNRS, LCC (Laboratoire de Chimie de Coordination), Universit´ e de Toulouse, UPS, INPT, 205 route de Narbonne, BP 44099, F-31077-Toulouse Cedex 4, France †Electronic supplementary information (ESI) available. See DOI: 10.1039/c6sc05551f Cite this: Chem. Sci.,2017,8,2931 Received 19th December 2016 Accepted 31st January 2017 DOI: 10.1039/c6sc05551f rsc.li/chemical-science This journal is © The Royal Society of Chemistry 2017 Chem. Sci.,2017,8,2931–2941 | 2931 Chemical Science EDGE ARTICLE Open Access Article. Published on 01 February 2017. Downloaded on 6/26/2018 11:27:49 AM. This article is licensed under a Creative Commons Attribution 3.0 Unported Licence. View Article Online View Journal | View Issue changes on the N-substituents give us the possibility to modify their electronic properties. For example, introducing electron acceptor/donor moieties (–Cl or –OMe) in these pending groups enables us to modulate the surface properties of MNPs. By the use of spectroscopic techniques, such as NMR or FT- IR, the coordination and dynamics of surface ligands of MNPs can be properly investigated. Recently, the coordination of NHC ligands to ruthenium and platinum nanoparticles was investigated by solid-state NMR spectroscopy. 6 Moreover, NMR studies on Ru and Pt NPs stabilized by hexadecylamine (HDA) conrmed that Pt displays a Knight shi(due to the presence of free electrons on the surface) which is much stronger than that of Ru. 7 This effect was also observed in Pt NPs of 1.2 and 2.0 nm, for which the different electronic states resulted in unequal magnitudes of Knight shi. 8 However, ultra-small Pt NPs (<1 nm) behave as molecular species without free electrons and no Knight shioccurs. 9 Another important instrument to characterize the surface of the nanoparticles is the use of carbon monoxide (CO) as a probe molecule, since CO coordination can be especially useful to determine the active sites. 8,10 Theoretical calculations are of great importance in order to shed more light on experimental data, as for example to secure the assignment of experimental solid state NMR spectra or for the calculation of properties that cannot be measured experimentally. 11 In the context of the nanoparticle surface chemistry, the theoretical calculations provide detailed information not only on the coordination mode of the species bound to the nanoparticle surface, 12 but also on the chemical reactions that may take place on these surfaces. 13 Ligand-stabilized metal NPs are effective catalysts for hydrogenation of olen, carbonyl and nitro functionalities, as well as aromatics. 14 In this context, Pt NPs have shown important chemoselectivities in hydrogenation of olen 6b,15 and nitro groups. 16 Herein we present a series of Pt NPs ligated by imidazolium-amidinates (ICy$ (Ar) NCN) which exhibit an exciting ligand effect in the hydrogenation of carbonyl groups when electron donor/acceptor groups are introduced in the N- substituents. These Pt NPs were characterized by the state-of- the-art techniques (TEM, HRTEM, WAXS, TGA and EA) and studied their surface chemistry through the coordination of CO (FT-IR and solid state MAS NMR). Here a clear correlation between the Knight shiof adsorbed CO and the NPs size was observed. In addition, the coordination of the imidazoliumamidinate ligands was investigated in Pt and Ru NPs by XPS, solid state 15 N MAS NMR and DFT calculations. Results and discussion Synthesis and characterization of Pt NPs Zwitterionic adducts of N,N-dicyclohexylimidazolidene and diarylcarbodiimide (ICy$ (Ar) NCN; Ar ¼Ph, p-tol, p-anisyl, p-ClC 6 H 4 ) were prepared similarly to previously reported betaine-type adducts, 4,5 by reaction of N-heterocyclic carbene (ICy) and a suitable carbodiimide, namely bis(4-methylphenyl) methanediimine ( (p-tol) NCN), bis(4-methoxyphenyl)methanediimine ( (p-anisyl) NCN), bis(4-chlorophenyl)methanediimine ( (p-ClC 6 H 4 ) NCN) and 15 N-labelled bis-phenylmethanediimine ( (Ph) NC 15 N) (Scheme 1). Non-commercial carbodiimides were obtained in high yields by desulphurization of the corresponding thioureas with I 2 in the presence of NEt 3 , using literature procedures. 17 The symmetric thioureas (p-RC 6 H 4 ) 2 C]S(R¼OMe, Cl) were readily synthesized from thiophosgene and the corresponding aniline. 15 N-labeled diphenylthiourea used to prepare (Ph) NC 15 N was obtained from phenylisothiocyanate and aniline. 17a With this method we expected to form the singly 15 N-labeled product, but the ESI-MS spectrum of the thiourea showed a signal for the corresponding anion of which the isotopic pattern was consistent with a statistical 1 : 2 : 1 mixture of the normal ( 14 N 2 ), monolabeled ( 14 N 15 N) and doubly-labeled isotopologues ( 15 N 2 ). As shown in Scheme S1 (see ESI, Section S1†), scrambling of the 15 N label indicates that aniline addition to phenylisothiocyanate is reversible. Most of Pt NPs reported in this paper were conveniently obtained by reaction of tris(2-norbornene)platinum(0) [Pt(NBE) 3 ] in THF under 3 bar H 2 in the presence of the corresponding imidazolium-amidinate ligand (Scheme 2). Herein we report for the rst time the use of Pt(NBE) 3 as metallic precursor in the synthesis of Pt NPs by the organometallic approach. The advantages of Pt(NBE) 3 versus the typical Pt precursors [tris(dibenzylideneacetone)diplatinum (0); Pt 2 (DBA) 3 or dimethyl(1,5- cyclooctadiene)platinum(II); Pt(CH 3 ) 2 (COD)] are evident. 7,8,18 First, the purication of the obtained nanoparticles is simpli- ed as the norbornane formed during the decomposition can be rapidly eliminated under vacuum and makes it this complex a“clean”precursor. And, second, the highly reactive Pt(NBE) 3 requires less time for the formation of Pt NPs under H 2 . Thus, the use of Pt(NBE) 3 as precursor is advantageous both for the formation and purication of Pt NPs. Different amounts of ICy$ (p-tol) NCN were employed for the synthesis of Pt NPs, specically 0.1, 0.2 and 0.5 molar equiv. (Pt/ICy$ (p-tol) NCN 0.1 , Pt/ ICy$ (p-tol) NCN 0.2 and Pt/ICy$ (p-tol) NCN 0.5 ), with the aim to obtain different sizes and reactivities. Indeed, Transmission Electronic Microscopy (TEM) micrographs of Pt/ICy$ (p-tol) NCN 0.1 , Pt/ ICy$ (p-tol) NCN 0.2 and Pt/ICy$ (p-tol) NCN 0.5 exhibit Pt NPs with a mean diameter of 2.3 (0.3), 2.1 (0.2) and 1.9 (0.4) nm respectively (Fig. 1), and one observes a correlation between the quantity of ligand and the average size of the obtained NPs; more ligand leads to smaller particles. This behavior was already observed in our previous work with Ru/ICy$ (p-tol) NCN, 4 and also with other type of ligands, such as NHCs, 6a,19 aminosilanes 20 and sulfonated diphosphines. 21 Pt/ICy$ (p-tol) NCN 0.1 Scheme 2 Synthesis of platinum nanoparticles using imidazoliumamidinate ligands as stabilizers. 2932 |Chem. Sci.,2017,8,2931–2941 This journal is © The Royal Society of Chemistry 2017 Chemical Science Edge Article Open Access Article. Published on 01 February 2017. Downloaded on 6/26/2018 11:27:49 AM. This article is licensed under a Creative Commons Attribution 3.0 Unported Licence. View Article Online and Pt/ICy$ (p-tol) NCN 0.2 micrographs revealed NPs that are very well distributed and monodisperse in size, while the image for Pt/ICy$ (p-tol) NCN 0.5 showed a worse dispersion. For an unknown reason, the range of sizes of these NPs increases with the amount of ligand, while the monodispersity is lost. High resolution TEM (HRTEM) micrographs of Pt/ ICy$ (p-tol) NCN 0.2 show crystalline Pt NPs (Fig. 2), displaying the face centered cubic (fcc) structure typical of bulk Pt. Fast Fourier Transformation (FFT) applied to the image of Fig. 2 revealed the reections corresponding to the (111), (200) and (111) atomic planes. Wide-Angle X-ray Scattering (WAXS) analyses performed on solid samples of Pt/ICy$ (p-tol) NCN 0.1 ,Pt/ICy$ (p-tol) NCN 0.2 and Pt/ ICy$ (p-tol) NCN 0.5 conrmed the crystallinity of these Pt NPs, which retain the fcc structure (see ESI, Fig. S17–S19†). The coherence lengths indicated by the Radial Distribution Function (RDF) resultant from WAXS analysis are slightly higher than the mean diameters determined by TEM. WAXS analysis revealed a coherence length of 2.5 nm for Pt/ICy$ (p-tol) NCN 0.1 , 2.2 nm for Pt/ ICy$ (p-tol) NCN 0.2 ,and2.1nmforPt/ICy$ (p-tol) NCN 0.5 ,whilethe mean diameters observed by TEM were 2.3, 2.1 and 1.9 nm, respectively. The Pt NPs were obtained in acceptable yields (60–70%; based on Pt), and the metal content was determined by thermogravimetric analysis (TGA). The Pt content was 90.1% for Pt/ ICy$ (p-tol) NCN 0.1 , 77.3% for Pt/ICy$ (p-tol) NCN 0.2 , and 60.1% for Pt/ICy$ (p-tol) NCN 0.5 . Moreover, elemental analysis (EA) of the NPs revealed that the amounts of C, H and N are in good agreement with the ratio of these elements in the ligand (C, 79.25%; H, 8.42%; N, 12.32%), suggesting that the imidazolium-amidinate ligand remains intact at the platinum surface (see ESI, Table S1†). It is important to mention that the Pt : L ratio in Pt/ICy$ (p-tol) NCN 0.5 shows a large quantity of coordinated ligand at the platinum surface, as the number of ICy$ (p-tol) NCN molecules is approximately half of the surface atoms (Pt surface 50% of Pt total ). Surface studies The surface chemistry of Pt/ICy$ (p-tol) NCN NPs was studied by Fourier transform infrared (FT-IR) and magic angle spinning solid-state 13 C and 15 N NMR (MAS-NMR) with and without cross-polarization (CP). CO was used as a probe molecule as it is well known that coordination of CO allows the identication of different available surface sites on metal nanoparticles. 10,22 Commonly, CO is adsorbed onto the metal surface either in a bridging (COb) or in a terminal (COt) mode. Fig. 3 shows attenuated total reectance (ATR) FT-IR spectra of Pt/ ICy$ (p-tol) NCN NPs aer bubbling CO during 5 min in a THF solution. The FT-IR spectrum of Pt/ICy$ (p-tol) NCN 0.1 [Fig. 3(a)] presents two distinct bands, at 1845 cm 1 and 2038 cm 1 , which can be respectively attributed to CO coordinated on the Pt surface in a bridging (COb) and in a terminal mode (COt). The stretching band attributed to COb is very intense, evidencing that a large part of the CO is coordinated in a bridging mode. Analogous stretching vibrations for adsorbed CO were previously observed in Pt(111) crystal surfaces and Pt NPs. 23 The IR Fig. 1 TEM images and size histograms of (a) Pt/ICy$ (p-tol) NCN 0.1 , (b) Pt/ICy$ (p-tol) NCN 0.2 and (c) Pt/ICy$ (p-tol) NCN 0.5 . Fig. 2 HRTEM image of Pt/ICy$ (p-tol) NCN 0.2 (left, right bottom) and fast Fourier transformation analysis (right, top) with the planar reflections. This journal is © The Royal Society of Chemistry 2017 Chem. Sci.,2017,8,2931–2941 | 2933 Edge Article Chemical Science Open Access Article. Published on 01 February 2017. Downloaded on 6/26/2018 11:27:49 AM. This article is licensed under a Creative Commons Attribution 3.0 Unported Licence. View Article Online spectrum of Pt/ICy$ (p-tol) NCN 0.2 displays similar frequencies at 1816 cm 1 and 2032 cm 1 , but here the ratio COt/COb has increased [Fig. 3(b)] in comparison with the previous one. This trend persists for Pt/ICy$ (p-tol) NCN 0.5 , and the intensities of CO bands (1808 and 2032 cm 1 ) decrease in relation to the characteristic C]N absorption frequencies of the amidinate ligand at 1630 and 1598 cm 1 [Fig. 3(c)] Here we observe an evident relation between the COt/COb ratio and the amount of ligand used. We think that both the size and the presence of more ligand onthesurfaceincreasetheamountofCOtatthecostofCOb,such that 1.9 nm Pt/ICy$ (p-tol) NCN 0.5 NPscontainsmostlyCOt.Itis noteworthy that when we decrease the amount of ICy$ (p-tol) NCN from 0.5 to 0.1 equiv., we observe a slight shiof the CO bands to higher frequency (from 1808 to 1845 cm 1 for COb and from 2032 to 2038 cm 1 for COt). This behavior is in agreement with a higher surface coverage of CO in Pt/ICy$ (p-tol) NCN 0.1 ,whichremovesmore electron density from the Pt surface (see ESI, Fig. S25†). The displacement of the CO stretching frequency in relation with the quantity of CO has been previously described for surfaces and MNPs. 8,23b,24 Thanks to FT-IR we obtained the rst indication about the coordination mode of the imidazolium-amidinate ligand. Comparing the IR spectra of the ligand ICy$ (p-tol) NCN (see ESI, Fig. S20†) and the particle Pt/ICy$ (p-tol) NCN 0.5 (see ESI, Fig. S26†), we observed a clear 100 cm 1 displacement of the strong stretching C]N bands, from 1530 and 1495 cm 1 to 1630 and 1598 cm 1 , which suggests that the coordination of ICy$ (p-tol) NCN takes place through the NCN moiety. These observed frequencies are in agreement with reported values. 5,25 Coordination of CO on the Pt NPs was also studied by solidstate NMR. The 13 C CP-MAS NMR spectrum obtained for Pt/ICy$ (p-tol) NCN 0.5 before exposition to 13 CO displays most of the characteristic signals of the ICy$ (p-tol) NCN ligand (Fig. 4). The group of peaks at 125 ppm (a) corresponds to the aromatic ring of the p-tolyl moiety and the signal at 116 ppm (b) belongs to the aromatic imidazolium backbone. The distinctive peak at 58 ppm (c) is attributed to the methine group next to the nitrogen atom. The cyclohexyl methylene resonances are observed between 32 ppm and 25 ppm (d) overlapping with the p-tolyl methyl group resonance at 25 ppm (e). As was previously observed for Ru/ICy$ (p-tol) NCN, the signals for the C ipso of the ptolyl group and the imidazolium and carbodiimide quaternary NCN atoms are not clearly observable in the 13 C spectrum (usually in the 130–150 ppm range; see ESI, Fig. S33†) due to a line broadening caused by the coordination of ICy$ (p-tol) NCN at the metal surface. 4 When the Pt/ICy$ (p-tol) NCN 0.5 NPs (1.9 nm) were exposed to 1 bar of 13 CO during 20 h at r.t., the 13 C MAS NMR spectrum [Fig. 5(c)] showed a very broad peak centered at ca. 360 ppm which corresponds to coordinated CO. This broad and high frequency resonance can be assigned to a Knight-shied 13 CO signal. In the 13 C MAS spectra of Pt/ICy$ (p-tol) NCN 0.2 (2.1 nm) Fig. 3 ATR FT-IR spectra of (a) Pt/ICy$ (p-tol) NCN 0.1 , (b) Pt/ ICy$ (p-tol) NCN 0.2 and (c) Pt/ICy$ (p-tol) NCN 0.5 after CO adsorption. Fig. 4 13 C CP-MAS NMR spectrum of Pt/ICy$ (p-tol) NCN 0.5 NPs. Fig. 5 13 C MAS NMR spectrum of (a) Pt/ICy$ (p-tol) NCN 0.1 (2.3 nm), (b) Pt/ICy$ (p-tol) NCN 0.2 (2.1 nm), (c) Pt/ICy$ (p-tol) NCN 0.5 (1.9 nm) and (d) Pt/CO/ICy$ (p-tol) NCN 0.2 (1.2 nm) after exposure to 13 CO (1 bar, 20 h, at r.t.). 2934 |Chem. Sci.,2017,8,2931–2941 This journal is © The Royal Society of Chemistry 2017 Chemical Science Edge Article Open Access Article. Published on 01 February 2017. Downloaded on 6/26/2018 11:27:49 AM. This article is licensed under a Creative Commons Attribution 3.0 Unported Licence. View Article Online and Pt/ICy$ (p-tol) NCN 0.1 (2.3 nm) aer reaction with 13 CO we also observed the broad peak of CO at 390 and 400 ppm respectively [Fig. 5(a) and (b)]. Thus, when the size of the NPs decreases, the CO signal is displaced to lower frequency. Because of the fast relaxation associated with the presence of surface electrons (paramagnetic effect), the efficiency of the magnetization transfer in the CP experiments is dramatically reduced and no signal for the coordinated CO could be observed in the 13 C CPMAS spectra (see ESI, bottom of Fig. S28, S30 and S32†). As we observed an evident relationship between the size and the magnitude of the Knight shion the CO band, we decided to synthesize very small Pt/ICy$ (p-tol) NCN NPs with the intention to suppress the Knight shicompletely. Pt NPs of ca. 1.2 (0.3) nm (see ESI, Fig. S8†) were prepared using a well-known two-step procedure. 7,23a Pt 2 (DBA) 3 was decomposed under 1 bar of CO in THF, forming the colloid Pt x (CO) y (THF) z , and aer washing this colloid with pentane, 0.2 equiv. of ICy$ (p-tol) NCN were added to obtain Pt/CO-ICy$ (p-tol) NCN 0.2 NPs (Scheme 3). The 13 C MAS spectrum of Pt/CO/ICy$ (p-tol) NCN 0.2 [Fig. 5(d)] aer reaction with 13 CO showed a broad peak at 230 ppm due to bridging CO (COb) and two sharp peaks at 210 ppm and 194 ppm which correspond to terminal (COt) and multi-terminal CO (nCOt), respectively. These ones are the values expected for bridging and terminal CO in Pt NPs. 8,9,18b However, we still observe a remaining broad Knight-shied CO resonance at 350 ppm due to the CO coordinated on the larger NPs. 13 C Knight-shied resonances have already been observed in Pt NPs, 7–9,26 but this is the rst case which reports a direct dependence of the size of Pt NPs with the degree of the Knight shiobserved, perceiving a displacement of the CO band as the nanoparticle size decreases (2.3 nm–2.1 nm–1.9 nm), and at 1.2 nm the Knight shiis almost suppressed (Fig. 5). To understand the state and coordination mode of these amidinate ligands, we used the labelled adduct 1,3-dicyclohex- ylimidazolium-2-di-phenylcarbodiimide (ICy$ (Ph) NC 15 N) and studied its coordination on the NP metal surface by 15 N MAS NMR spectroscopy. ICy$ (Ph) NC 15 N was fully characterized by NMR (liquid and solid state) and DFT calculations (see ESI,† Section S1). Employing the aforementioned synthetic method, Pt/ICy$ (Ph) NC 15 N 0.5 NPs with an average size of 1.9 (0.2) nm (see ESI, Fig. S9†) were prepared (Scheme 1). These NPs were studied by 15 N MAS NMR but we could not observe any detectable signal in 15 N Hahn-echo NMR [see ESI, Fig. S36(a)†] not even in 15 N CP/Hahn-echo MAS NMR [see ESI, Fig. S36(b)†], which are usually employed to detect broad NMR signals. The lack of signals is possibly due to the Knight shias discussed above. To suppress the Knight shiwe synthesized Pt NPs of ca. 1.2 nm with labelled ICy$ (Ph) NC 15 N, Pt/CO/ICy$ (Ph) NC 15 N 0.2 (see ESI, Fig. S10†). The 15 N CP/Hahn-echo MAS NMR spectrum displays two sharp signals at 120 and 260 ppm [Fig. 6(a)]. The sharpness of the signals points to a non-coordinated nitrogen compound, as the signals of coordinated ligands are usually broad. The intense peak at 120 ppm is reminiscent of a protonated nitrogen ligand, therefore we prepared the singlyprotonated phenyl betaine adduct as its tetraphenylborate salt [ICy$ (Ph) NC 15 NH] + [BPh 4 ]  . In contrast with the strongly basic ligand Me 2 IiPr$ (iPr) NCN reported by Kuhn, 27 aryl imidazoliumamidinate ICy$ (Ar) NCN (e.g. Ar ¼Tol) does not react with water to a signicant extent, but they are immediately protonated by acetic acid (HOAc). [ICy$ (Ph) NC 15 NH] + [BPh 4 ]  was obtained reacting ICy$ (Ph) NC 15 N with HOAc in the presence of NaBPh 4 . Indeed the solid-state 15 N CP/Hahn-echo MAS spectrum of this compound exhibits two signals at 116 and 266 ppm, attributed to the protonated and non-protonated N atoms, respectively. The former is ca. four-fold as intense as the latter, because the CP effect is much more intense in the protonated nitrogen due to the presence of the strong N–H dipolar coupling [Fig. 6(b)]. Solution NMR spectra in CD 2 Cl 2 (see ESI,†Section S1) indicate that the cation exists as a 2 : 1 mixture of two isomers obtained by protonation of either one of the two N-atoms. DFT calculations conrm the experimental shis and corroborate that the two E/Zisomers of [ICy$ (Ph) NC 15 NH] + [BPh 4 ]  can coexist as evidenced by their relative energy difference of 1.1 kcal mol 1 . In the same way as for non-protonated ligands, the E/Zisomers are more stable than the Z/Zisomer by 3.4 kcal mol 1 and 2.3 kcal mol 1 (see ESI, Fig. S59†). Thus, the peaks at 120 and 260 ppm in the 15 N CP/Hahn-echo MAS NMR spectrum of Pt/ CO/ICy$ (Ph) NC 15 N 0.2 correspond to a minor part of the ligand that has been protonated during the synthesis and the signal of which is strongly enhanced thanks to the 15 N CP-MAS experiment. However, we were not able to detect 15 N resonances of coordinated ICy$ (Ph) NC 15 Nin 15 N Hahn-echo MAS NMR, most likely because of the deleterious effect of the NP surface electrons (see ESI, Fig. S38†). X-ray photoelectron spectroscopy (XPS) is an ideal technique to measure the elemental composition and chemical state of surface catalysts. Recently, this technique has been described as a useful tool to investigate the binding mode of NHC ligands on Scheme 3 Synthesis of Pt/CO/ICy$ (p-tol) NCN 0.2 NPs. Fig. 6 15 N CP/Hahn-echo MAS of (a) Pt/CO/ICy$ (Ph) NC 15 N 0.2 and (b) [H$ICy$ (Ph) NC 15 N] + $[BPh 4 ]  . This journal is © The Royal Society of Chemistry 2017 Chem. Sci.,2017,8,2931–2941 | 2935 Edge Article Chemical Science Open Access Article. Published on 01 February 2017. Downloaded on 6/26/2018 11:27:49 AM. This article is licensed under a Creative Commons Attribution 3.0 Unported Licence. View Article Online a metal surface. 28 Thus, we applied XPS to study the coordination of the imidazolium-amidinate ligands in our Pt NPs. The N 1s signals of ICy$ (p-tol) NCN present binding energies (BE) of 401.3 and 397.4 eV [Fig. 7(a), blue]. The highest BE peak at 401.3 eV can be assigned to the tightly bound electrons of the nitrogen atoms of the imidazolium fragment. The second peak at 397.4 eV derives from the N atoms of the amidinate group, which bear partial negative charges. 29 The coordination of ICy$ (p-tol) NCN through the latter N-atoms to the platinum surface involves a loss of electron density, which increases the binding energy (400.1 eV) compared to the N-atoms of free ICy$ (p-tol) NCN (397.4 eV). This leads to the overlap of the peaks of the two fragments, as can be seen in Fig. 7(a) (red). This signal is the result of convolutions of peaks corresponding to imidazolium and amidinate N atoms, coordinated or not. In order to improve our understanding of these binding energies, we theoretically compared the natural charges of the N atoms of the ICy$ (Ph) NCN ligand, free and coordinated in m 2 -k 1 N, k 1 N0 mode to a model Ru carbonyl cluster (see ESI, Fig. S60†). Interestingly, no change is observed in the natural charges of the N atoms of the imidazolium fragment by coordination of the ligand on the cluster surface. However, the anionic character of the N atoms of the amidinate moiety is partially reduced, as a result of the balance between s-donation and p-back- donation. Thus, the coordination of the ligand induces the decrease of the charge difference between the N atoms of both moieties leading to overlap of the peaks of the two fragments as observed in Fig. 7(a) (red). This result suggests that the coordination of the ligand to the nanoparticle is through the N atoms of the amidinate moiety. We deconvoluted the N 1s signals of Pt/ICy$ (p-tol) NCN 0.2 at 399.9 eV in three contributions with different binding energies, at 401.1, 399.7 and 398.2, corresponding to d  , neutral and d + N atoms, respectively [Fig. 7(c), center]. When the electron-withdrawing ligand ICy$ (p-ClC 6 H 4 )- NCN is used as stabilizer, the contribution of the peak at 401.1 (N d+ ) intensies, generating an increase of the average BE of the N 1s signal to 400.8 eV [Fig. 7(c), top]. On the other hand, the N(1s) signals of Pt/ICy$ (p-anisyl) NCN 0.2 containing the electron donor OMe substituent, display the opposite behavior, intensifying the peak at 398.2 (N d ) and decreasing the global BE to 398.7 eV [Fig. 7(c), bottom]. The binding energy of Pt 4f 7/2 in the XPS spectra of Pt/ICy$ (p-tol) NCN 0.2 NPs is 71.1 eV. The main contribution is located at 71.0 eV, which is characteristic of Pt(0), 30 and an additional contribution at 72.7 eV is attributed to the surface platinum atoms linked to the nitrogen atoms of ICy$ (p-tol) NCN (Pt(d+)) [Fig. 7(b)]. 31 Coordination of imidazolium-amidinates to metal surfaces Amidinates have a rich coordination chemistry 32 as they can coordinate to a single metal center either in terminal (k 1 ), chelating (k 2 ) modes, and, less oen they can even bind in h 3 mode, like an allyl group. The amidinate can also bridge two metal centers, i.e.,inm 2 -k 1 N, k 1 N0mode. On a metal surface one N-atom can bridge more than two metal atoms and thus a variety of coordination modes results. As the measurements on Pt NPs gave not satisfactory evidence for the bonding mode in 15 N NMR spectroscopy we turned to different metals for 15 N NMR spectral data. As ruthenium presents insignicant or no Knight shi, we synthesized Ru/ICy$ (Ph) NCN NPs as reported before 4 stabilized with 0.1, 0.2, 0.5 and 1 equiv. of the 15 N labelled ligand and studied the coordination of the ligand by solid state 15 N MAS NMR. The resulting Ru/ICy$ (Ph) NC 15 N NPs have sizes between 1.0 and 1.3 nm, they are very well dispersed and show a narrow size distribution (see ESI, Fig. S11–S14†). As for Pt NPs, we perceived a correlation between the size and the amount of ligand used during the synthesis; the size increases when less ligand is used. In the 15 N CP/Hahn-echo MAS NMR spectra of Ru/ ICy$ (Ph) NC 15 N NPs (see ESI, Fig. S43†) we can see clearly the peak around 120 ppm corresponding to the protonated amidinate ligand, as was already observed for Pt/CO/ICy$ (Ph) NC 15 N 0.2 . Furthermore, a new broad featureless signal is observed underneath the 15 N resonance of the protonated imidazoliumamidinate ligand. Note that for Ru NPs stabilized with 1 equiv. [see ESI, Fig. S43(a)†], we also observed a sharp peak at 212 ppm assigned to the excess of free ligand. 15 N Hahn-echo MAS NMR spectra (see ESI, Fig. S45, S48, S54 and S57†) allowed the amplication of this new broad 15 N resonance relative to the signal of the protonated free ligand. Unfortunately, the broadness and the poor signal to noise ratio of these experiments still preclude a clear characterization of the 15 N signals. In order to increase the sensitivity of the 15 N MAS experiments, we used the Carr–Purcell–Meiboom–Gill pulse sequence (CPMG). 33 It works well for NMR resonances that have long spin-state lifetimes but signicant inhomogeneous broadening. Fig. 8 displays the 15 N CPMG MAS NMR spectra of Ru/ICy$ (Ph) NC 15 N NPs stabilized with the different amounts of ICy$ (Ph) NC 15 N. With this series of experiments and DFT calculations on the chemical shiswe could establish the different states and coordination modes of Fig. 7 X-ray photoelectron spectroscopy (XPS) of (a) the N 1s signals of ICy$ (p-tol) NCN (blue) and Pt/ICy$ (p-tol) NCN 0.2 (red), (b) the Pt 4f 5/2 and 4f 7/2 signals of Pt/ICy$ (p-tol) NCN 0.2 and (c) the N 1s signals of Pt/ ICy$ (p-ClC 6 H 4 ) NCN 0.2 (top), Pt/ICy$ (p-tol) NCN 0.2 (center) and Pt/ ICy$ (p-anisyl) NCN 0.2 (bottom). 2936 |Chem. Sci.,2017,8,2931–2941 This journal is © The Royal Society of Chemistry 2017 Chemical Science Edge Article Open Access Article. Published on 01 February 2017. Downloaded on 6/26/2018 11:27:49 AM. This article is licensed under a Creative Commons Attribution 3.0 Unported Licence. View Article Online the amidinate at the Ru surface. The broad peak centered at ca. 250 (100) ppm corresponds to coordinated ICy$ (Ph) NC 15 N which can coordinate in different ways. Following a computational strategy successfully used for computing 1 H and 13 C chemical shis, a [Ru 6 ] carbonyl cluster was rst used as a model for Ru NPs as larger systems cannot be conveniently handled. 34 The chemical shis were calculated by DFT studies and compared with the observed ones (see ESI, Fig. S61†). When ICy$ (Ph) NC 15 N bridges two adjacent Ru atoms, the estimated deltas are 221 and 205 ppm [see ESI, Fig. S61(a)†](k 2 coordination to one metal is not a stable conguration, vide infra). For k 1 coordination the calculated chemical shis are 244 ppm for the non-coordinated nitrogen and 118 ppm for the bound one [see ESI, Fig. S61(b)†]. These values change to 262 and 134 ppm if there is p-stacking between the phenyl groups and the ruthenium surface [see ESI, Fig. S61(c)†]. In view of the broadness of the resonances centered at ca. 250 ppm (Dn 1/2 ¼100–300 ppm), the 15 N NMR spectrum suggests that all three coordination modes are present. Lastly, the broad peak observed between 315 and 380 ppm in the 15 N CPMG MAS NMR spectra of Ru/ICy$ (Ph) NC 15 N NPs (Fig. 8) is attributed to a species resulting from an insertion reaction of amidinate and carbon monoxide on the surface. This conclusion was corroborated by pressurization of Ru/ ICy$ (Ph) NC 15 N 0.2 NPs with 13 CO (1 bar, r.t., 20 h), which led to a signicant increase of these resonances (Fig. 9). 35 Unfortunately, the peak corresponding to CO ins , expected at 219 ppm in 13 C MAS NMR, cannot be observed due to overlap with the broad peak of COb centered at 230 ppm (see ESI, Fig. S51†). The calculated 15 N shis of monocoordinated ICy$ (Ph) NC 15 Naer insertion of CO are 151 ppm for the migrated N-atom and 308 ppm for the non-coordinated N-atom (Fig. 9). Aer a careful analysis of NPA charges on this cluster model, we conclude that this signicant unshielding of the latter is due to a delocalization of the negative charge on the CO–Ru fragment and the localization of the pbond between C and this 15 N-atom (see ESI, Fig. S62 and S63†). A charge analysis on its Ru 55 NP counterpart leads to the same conclusion (see ESI, Fig. S64†). This 308 ppm value is in the range of 15 N chemical shis observed for imines. 36 Subsequently, the energetics of ICy$ (Ph) NCN coordination onto the Ru surface was investigated by DFT calculations. The chosen model is a Ru nanoparticle of 55 atoms with a hydride coverage of 1.6 hydrides per ruthenium surface atom, which has a size of ca. 1 nm (Ru 55 H 70 ). This model, apart from having a size similar (1 nm) to our Ru/ICy$ (Ph) NCN NPs, has been already satisfactory employed by some of us 11 to determine the preferred surface composition of Ru NPs as a function of environmental conditions from an application of the ab initio thermodynamics method, 37 that provides a connection between the microscopic and macroscopic regimes. In the present case, the DFT calculations showed that the most stable coordination mode is through the two nitrogen atoms of the amidinate as the Z/Zconformer to two adjacent Ru-atoms (m 2 -k 1 N, k 1 N0), with a binding energy of 47.9 kcal mol 1 . The binding of the same conformer through only one N atom has a stability of 33.3 kcal mol 1 , while that of monocoordination of the E/Zconformer is 25 kcal mol 1 (Fig. 10). When the ligand was initially bound to a single Ru atom as bidentate (k 2 N, N0), geometry optimization led to a more favorable terminal k 1 N coordination mode, as observed for some recently reported Cu complexes. 5 Were the bonding prevailingly polar, such bidentate coordination mode could be favored, as observed for amidinate or carboxylate derivatives of the alkali and rare earth metals. 38 In summary, the adsorption energy per Ru–N bond of a single amidinate which is not sterically discomforted by its mutual interaction with other ICy$ (Ph) NCN ligands lies between 24 and 33 kcal mol 1 . It is interesting to note that the interaction between the ICy$ (Ph) NCN ligand and the Ru 55 nanoparticle does not involve a displacement of the surface hydrides, considering that the loss of four H 2 molecules is disfavored by +20.6 kcal mol 1 (see ESI, Fig. S65†). The maximum number of m 2 -k 1 N, k 1 N0coordi- nated ICy$ (Ph) NCN ligands that we can t on the Ru 55 surface is six (see ESI, Fig. S66†), which corresponds with Ru NPs Fig. 8 15 N CPMG MAS NMR of Ru/ICy$ (Ph) NC 15 N stabilized with (a) 1 equiv., (b) 0.5 equiv., (c) 0.2 equiv. and (d) 0.1 equiv. Fig. 9 15 N CPMG MAS NMR of Ru/ICy$ (Ph) NC 15 N 0.2 (a) before and (b) after exposure to 13 CO (1 bar, 20 h, at r.t.). (c) Calculated 15 NNMR displacements of ICy$ (Ph) NCN after CO insertion on a [Ru 6 ] cluster model. This journal is © The Royal Society of Chemistry 2017 Chem. Sci.,2017,8,2931–2941 | 2937 Edge Article Chemical Science Open Access Article. Published on 01 February 2017. Downloaded on 6/26/2018 11:27:49 AM. This article is licensed under a Creative Commons Attribution 3.0 Unported Licence. View Article Online stabilized with 0.1 equiv. of the ligand. The adsorption strength per betaine is lowered from 47.9 kcal mol 1 in the case of a single ligand to 29.1 kcal mol 1 for the maximum number of them (i.e., 6). For k 1 coordinated amidinates one could t perhaps a few more ligands on the surface, which is still energetically favorable. Thus, the Ru/ICy$ (Ph) NC 15 N NPs stabilized with 0.2, 0.5 and 1 equiv. have a second sphere of non-coordi- nated ICy$ (Ph) NC 15 N which is strongly bonded to the rst sphere by ionic and p–pstacking interactions. At least some part of the second sphere is present in the protonated form as indicated by 15 N NMR. We shall now evaluate the coordination energy of the species resulting from an insertion reaction of CO into the Ru–N bonds, which we shall call CO ins , responsible for the broad peak observed between 315 and 380 ppm in the 15 N NMR spectrum of carbonylated nanoparticles using the [Ru 55 ] cluster model. We initially considered a Ru 55 (CO) 66 cluster, with a coverage corresponding to 1.5 CO per surface Ru atom, as established in previous ab initio thermodynamics calculations. 12 The adsorption energies of m 2 -k 1 N, k 1 N0bicoordinated ICy$ (Ph) NCN, k 1 N monocoordinated Z/Zconformer and the CO ins compound are slightly endothermic by 3–5 kcal mol 1 whatever the ligand, probably owing to the crowding of carbon monoxides on the surface. This is why we considered the adsorption properties on aRu 55 (CO) 59 cluster. The adsorption energy of the CO ins compound becomes exothermic by 27.3 kcal mol 1 –a value similar to the Ru–N bond energy –and its graing mode is similar to what has been obtained on the small [Ru 6 ] cluster (see ESI, Fig. S67†). Although its adsorption is strongly exothermic, it is signicantly less strongly bound than bicoordinated ICy$ (Ph) NCN ligands (m 2 -k 1 N, k 1 N0and k 2 N, N0) (between 47 kcal mol 1 and 76 kcal mol 1 depending on the graing site, i.e. between 24 and 38 kcal mol 1 per Ru–N bond, close to the values calculated for the Ru 55 H 70 model). These results suggest that, according to thermodynamics, CO insertion will more likely occur on the monocoordinated amidinate ligands. In conclusion, the DFT calculations fully support the 15 NMR measurements in that the Ru NPs are covered with m 2 -k 1 N, k 1 N0 amidinates, k 1 coordinated amidinates, and free ligand in part in protonated form. The counter anion of the latter is not known as the synthesis of the MNPs does not involve salts. In the absence of conclusive 15 NMR measurements for Pt NPs we propose that the coordination mode is the same as that found for Ru. Catalytic studies To probe the catalytic activity and chemoselectivity of the new NPs we chose Pt/ICy$ (p-tol) NCN 0.2 as model system and tested it in the hydrogenation of several substrates containing various functional groups such as olenic bonds, carbonyl groups, and aromatic rings. In addition, we investigated the inuence of N- aryl groups with different electronic properties using Pt NPs ligated to functionalized imidazolium-amidinate ligands (Pt/ICy$ (p-anisyl) NCN 0.2 and Pt/ICy$ (p-ClC 6 H 4 ) NCN 0.2 ; see ESI Fig. S15 and S16†). The catalytic results gave interesting differences in terms of activity depending on the type of stabilizing ligand used. In general, all nanocatalysts hydrogenated olenic and C]Obonds,but,aswasexpectedforPtNPs,they were not capable of hydrogenating aromatic rings; e.g. the hydrogenation of styrene catalyzed by Pt/ICy$ (p-tol) NCN 0.2 ,gave selectively ethylbenzene with a maximum turn over frequency (TOF) of 75 300 h 1 (Table 1). This high TOF value demonstrates the high catalytic power of this system in hydrogenation of olenic double bonds. This is peculiar in view of the high coverage with strongly bound imidazolium-amidinate ligands and indicatesthattheseligandsplayafavorableroleaselectrondonors. Interestingly, 1.2 nm Pt/CO/ICy$ (Ph) NC 15 N 0.2 NPs prepared from Pt 2 (DBA) 3 and CO (Scheme 3) were totally inactive in the hydrogenation of styrene, due to their surface poisoned by CO. Indeed, as expected from previous works 7,23a the IR spectrum of such NPs, recorded immediately aer the synthesis, presented two characteristic bands for COt and COb at 2041 and 1859 cm 1 , respectively which remained essentially unchanged when these are exposed to CO (see ESI, Fig. S24†). This conrmed that the Pt surface of Pt/CO/ICy$ (Ph) NC 15 N 0.2 NPs was totally covered with CO due to their synthesis conditions, thus explaining why these NPs are completely inactive as hydrogenation catalyst. The catalytic behavior of Pt NPs Pt/ICy$ (p-anisyl) NCN 0.2 , Pt/ ICy$ (p-ClC 6 H 4 ) NCN 0.2 and Pt/ICy$ (p-tol) NCN 0.2 was studied in the hydrogenation of 4-phenyl-3-buten-2-one, 3-methyl-2-cyclo- hexenone and 4-nitrobenzaldehyde (Table 2, entries 1–9). All reactions were carried out with Pt NPs synthesized with 0.2 equiv. of the corresponding amidinate ligand as the results changed little for Pt NPs prepared with 0.1 or 0.5 equiv. (Table S2†). In all cases, the olens were smoothly hydrogenated (in less than 1 h), but the carbonyl groups reacted more slowly. Aer 20 h of reaction, ketone and aldehyde groups were partially Fig. 10 Coordination modes of ICy$ (Ph) NCN at the Ru 55 H 70 surface (a) k 1 N monocoordinated Z/Econformer (E ads :25.0 kcal mol 1 ); (b) k 1 N monocoordinated Z/Zconformer (E ads :33.3 kcal mol 1 ); (c) m 2 -k 1 N, k 1 N0bicoordinated Z/Zconformer (E ads :47.9 kcal mol 1 ). Table 1 Hydrogenation of styrene catalyzed by Pt/ICy$ (p-tol) NCN 0.2 Entry Substrate Products c Conv. c (%) TOF (h 1 ) 1 a >99 12 800 2 b 10 75 300 a Reaction conditions: substrate (32 mmol), Pt NPs (0.0025 mmol of Pt assuming 77.3% of Pt from TGA analysis), THF (3 mL), 1 h, T¼333 K, P¼5 bar. b Reaction conditions: substrate (320 mmol), Pt NPs (0.0025 mmol of Pt assuming 77.3% of Pt from TGA analysis), THF (10 mL), 10 min, T¼333 K, P¼5 bar. c Conversions and products identities were determined by 1 H NMR (average of two runs). 2938 |Chem. Sci.,2017,8,2931–2941 This journal is © The Royal Society of Chemistry 2017 Chemical Science Edge Article Open Access Article. Published on 01 February 2017. Downloaded on 6/26/2018 11:27:49 AM. This article is licensed under a Creative Commons Attribution 3.0 Unported Licence. View Article Online hydrogenated in different grades depending on the N-aryl group of the ligand. We observed a slight increase in activity with the electron-donor strength of these substituents. Thus, the NPs stabilized with the strongest donor ligand, Pt/ICy$ (p-anisyl) NCN 0.2 , were the most active systems in the hydrogenation of the carbonyl group. This trend was more pronounced for ethyl pyruvate (Table 2, entries 10–12). Signicant differences of reactivity between the three nanosystems were observed in the hydrogenation of this substrate. This ligand effect was conrmed in the hydrogenation of 2,2,2-triuoroacetophenone (Table 2, entries 13–15), for which also the Pt NPs ligated by the stronger donor ligand ICy$ (p-anisyl) NCN provided the most active nanocatalyst for this reaction. We conclude that more electron-rich Pt NPs yield faster catalysts, but as yet we cannot speculate on the mechanism. Conclusions We have successfully ligated zwitterionic amidinates to Pt NPs which leads to the peculiar property that the charges of the anions need no compensation by cationic metal ions as in clusters that contain thiolates on the surface for instance. Usually N-based ligands are weakly binding ligands to MNPs, but unexpectedly amidinates bind strongly to the Pt surface atoms. We have shown that 15 N NMR spectroscopy can be used for the study of the adsorption mode of the nitrogen ligand to the surface with the use of 15 N enriched ligands. The disturbing Knight shis on the chemical shis of the surface bound nuclei could be largely suppressed when small Pt NPs were used. Amidinates may well nd broader application in MNP and metal nanocluster (MNC) synthesis and catalysis. Even more so because subtle electronic modication of the N-aryl groups of the amidinates has an effect on the catalytic performance of the Pt NPs. Acknowledgements This work was supported by CNRS, UPS-Toulouse, INSA, and IDEX/Chaires d'Attractivit´ e de l'Universit´ eF ´ ed´ erale Toulouse Table 2 Hydrogenation reactions catalyzed by Pt/ICy$ (p-anisyl) NCN 0.2 , Pt/ICy$ (p-tol) NCN 0.2 and Pt/ICy$ (p-ClC 6 H 4 ) NCN 0.2 Entry/Pt NPs Substrate Products b Conv. c (%), select. c 1/Pt/ICy$ (p-anisyl) NCN 0.2a >99, 2/3 ¼79 : 21 2/Pt/ICy$ (p-tol) NCN 0.2a >99, 2/3 ¼81 : 19 3/Pt/ICy$ (p-ClC 6 H 4 ) NCN 0.2a >99, 2/3 ¼86 : 14 4/Pt/ICy$ (p-anisyl) NCN 0.2a >99, 5/6 ¼67 : 33 5/Pt/ICy$ (p-tol) NCN 0.2a >99, 5/6 ¼68 : 32 6/Pt/ICy$ (p-ClC 6 H 4 ) NCN 0.2a >99, 5/6 ¼78 : 22 7/Pt/ICy$ (p-anisyl) NCN 0.2a 98, 8/9/10 ¼52:2:44 8/Pt/ICy$ (p-tol) NCN 0.2a 98, 8/9/10 ¼53:1:44 9/Pt/ICy$ (p-ClC 6 H 4 ) NCN 0.2a 98, 8/9/10 ¼56:4:38 10/Pt/ICy$ (p-anisyl) NCN 0.2b 99 11/Pt/ICy$ (p-tol) NCN 0.2b 86 12/Pt/ICy$ (p-ClC 6 H 4 ) NCN 0.2b 24 13/Pt/ICy$ (p-anisyl) NCN 0.2b 96 14/Pt/ICy$ (p-tol) NCN 0.2b 47 15/Pt/ICy$ (p-ClC 6 H 4 ) NCN 0.2b 33 a Reaction conditions: substrate (0.5 mmol), Pt NPs (0.0025 mmol of Pt assuming 77.3% of Pt from TGA analysis), THF (0.75 mL), 20 h, r.t., 5 bar. b Reaction conditions: substrate (0.5 mmol), Pt NPs (0.0025 mmol of Pt assuming 77.3% of Pt from TGA analysis), THF (0.75 mL), 3 h, r.t., 5 bar. c Conversions, selectivities and products identities were determined by 1 H NMR (average of two runs). This journal is © The Royal Society of Chemistry 2017 Chem. Sci.,2017,8,2931–2941 | 2939 Edge Article Chemical Science Open Access Article. Published on 01 February 2017. Downloaded on 6/26/2018 11:27:49 AM. This article is licensed under a Creative Commons Attribution 3.0 Unported Licence. View Article Online