Synthesis, structure and gas-phase reactivity of the mixed silver hydride borohydride nanocluster [Ag₃(μ₃-H)(μ₃-BH₄)L^Ph₃]BF₄ (L^Ph = bis(diphenylphosphino)methane)

Abstract

Borohydrides react with silver salts to give products that span multiple scales ranging from discrete mononuclear compounds through to silver nanoparticles and colloids. The cluster cations [Ag₃(H)(BH₄)L₃]⁺ are observed upon electrospray ionization mass spectrometry of solutions containing sodium borohydride, silver(I) tetrafluoroborate and bis(dimethylphosphino)methane (L^Me) or bis(diphenylphosphino)methane (L^Ph). By adding NaBH₄ to an acetonitrile solution of AgBF₄ and L^Ph, cooled to ca. −10 °C, we have been able to isolate the first mixed silver hydride borohydride nanocluster, [Ag₃(μ₃-H)(μ₃-BH₄)L^Ph₃]BF₄, and structurally characterise it via X-ray crystallography. Combined gas-phase experiments (L^Me and L^Ph) and DFT calculations (L^Me) reveal how loss of a ligand from the cationic complexes [Ag₃(H)(BH₄)L₃]⁺ provides a change in geometry that facilitates subsequent loss of BH₃ to produce the dihydride clusters, [Ag₃(H)₂L_n]⁺ (n = 1 and 2). Together with the results of previous studies (Girod et al., Chem. – Eur. J., 2014, 20, 16626), this provides a direct link between mixed silver hydride/borohydride nanoclusters, silver hydride nanoclusters, and silver nanoclusters.

---

Introduction

The chemistry of alkali metal borohydrides reacting with silver salts has a rich history of over 60 years¹ and continues to yield a variety of silver containing products, including silver borohydrides,^2–4 silver hydrides^5–10 and reduced silver species.^11–16 These products span multiple scales ranging from discrete mononuclear compounds^2–4 through to ligand capped silver nanoclusters,^9–17 silver nanoparticles, and colloids capped with various species including BH₄⁻ (Scheme 1).^18–21,‡ In 1952 Wiberg and Henle reported that at −80 °C silver perchlorate reacts with lithium borohydride to yield AgBH₄ (eqn (1)), which decomposes when the temperature rises to −30 °C via hydrogen and BH₃ loss (eqn (2)).^2–4 Hydrogen can also be liberated via reaction with methanol (eqn (3)). Ab initio calculations suggested that the decomposition of AgBH₄ should yield silver hydride (eqn (4)).²³ Finally, there have been several reports that discrete mononuclear complexes of AgBH₄ could be isolated when phosphine ligands were employed,^2–4 ultimately leading to the X-ray crystallographic characterization of (Ph₂MeP)₃AgBH₄ (Scheme 1(a)).⁴

AgClO₄ + LiBH₄ → AgBH₄ + LiClO₄ (1)

AgBH₄ → Ag + 0.5H₂ + BH₃ (2)

AgBH₄ + 3MeOH → Ag + 3.5H₂ + B(OMe)₃ (3)

AgBH₄ → AgH + BH₃ (4)

Scheme 1 Reactions of BH₄⁻ with silver salts to give products spanning multiple scales. (a)–(f) represent discrete isolated species whose structures have been determined via X-ray crystallography. The scale indicates ionic interactions or the outermost diameter of each silver cluster not inclusive of ligands. Counterions and hydrogen atoms have been omitted for clarity.

In the last few years, several new silver nanoclusters have been isolated from reaction mixtures containing silver salts, sodium borohydride and either anionic ligands or neutral ligands and structurally characterised via X-ray crystallography. Structural reports include (i) silver hydride clusters [Ag₂(μ₂-H)L₂]BF₄ ⁸ (where L = 1,3-bis(2,6-diisopropylphenyl)imidazolin-2-ylidene (Scheme 1(b)), [Ag₃(μ₃-H)(μ₃-Cl)L^Ph₃]BF₄ ⁹ and [Ag₃(μ₃-H)L^Ph₃](BF₄)₂ ¹⁰ (where L^Ph = bis(diphenylphosphino)methane), [{Ag₇(μ₄-H)(E₂P(OR)₂}₆] (R = ⁱPr, E = Se),¹⁸ [Ag₈H(S₂P(OEt)₂)₆]⁺⁵, and [Ag₁₁(H)(S₂P(OEt)₂)₉]⁺ (Scheme 1(c)),^6,7 (ii) the silver clusters Ag₁₄(SC₆H₃F₂)₁₂(PPh₃)₈ (Scheme 1(d)),¹² Ag₁₆(dppe)₄(SC₆H₃F₂)₁₄, {Ag₃₂(dppe)₅(SC₆H₄CF₃)₂₄}²⁻ (dppe = 1,2-bis(diphenylphosphino) ethane, Scheme 1(e)),¹⁴ [Ag₄₄(SAr)₃₀]⁴⁻ (where ArS is an arylsulfide, Scheme 1(f)),¹³ [Ag₂₁{S₂P(OⁱPr)₂}₁₂]⁺,¹⁵ Ag₂₉(BDT)₁₂(PPh₃)₄ (BDT = 1,3-benzenedithiol)¹⁶ and [Ag₂₅(SR)₁₈]⁻.¹⁷ Importantly, Liu et al. have shown that [{Ag₇(μ₄-H)(E₂P(OR)₂}₆] (E = Se, S) are precursors to further growth into silver nanoparticles.²⁴

Here we report the “mass spectrometry directed synthesis” of the first mixed silver hydride/borohydride cluster [Ag₃(μ₃-H)(μ₃-BH₄)L^Ph₃]BF₄,^25–27 and its structural characterization by X-ray crystallography. DFT calculations indicate that the loss of a ligand (L) under collision induced dissociation (CID) conditions results in a change in cluster geometry that facilitates decomposition of the ligated BH₄⁻via loss of BH₃ (cf.eqn (2)).

Results and discussion

Electrospray ionization mass spectrometry (ESI/MS) was used to monitor the identity of cationic silver clusters formed upon mixing silver(I) tetrafluoroborate with diphosphine ligands in cooled acetonitrile solutions and subsequently treating with excess NaBH₄. The ligands bis(diphenylphosphino)methane (L^Ph) and bis(dimethylphosphino)methane (L^Me) were added to an acetonitrile solution of AgBF₄ (1.9 mg, 5 mM) with a ligand-to-metal ratio of 1 : 1 giving a clear solution. Addition of 15 equivalents of NaBH₄ to each of these solutions gave an immediate colour change from colourless to light yellow for both L^Phand L^Me. ESI/MS analysis 5 minutes after the addition of NaBH₄ are shown in Fig. 1. The bulkier L^Ph ligand yields the abundant peak [Ag₃(H)(BH₄)L^Ph₃]⁺m/z 1493 (Fig. 1a), confirmed by high resolution mass spectrometry experiments (ESI Fig. S1a and b†). In contrast the L^Me yields a mixture of silver clusters assigned as [Ag₃(H)L^Me₃]²⁺m/z 366, [Ag₃(H)(BH₄)L^Me₃]⁺m/z 746 (Fig. 1b, ESI Fig. S2†) and [Ag₃(BH₄)₂(MeCN)L^Me₃]⁺m/z 801. Since clusters formed from L^Ph were less prone to decomposition reactions in solution compared to those formed from L^Me,²⁸ the synthesis of crystalline material suitable for characterization was pursued for L^Ph.

Fig. 1 Full LTQ ESI/MS for solution phase synthesis of silver hydride clusters protected by diphosphine ligands: (a) L^Ph; (b) L^Me. Solutions containing condensed phase silver clusters were diluted to 50 μM in acetonitrile. Spectra were recorded 5 minutes after the addition of NaBH₄. The most intense peak in the cluster represents the m/z value.

Structural characterization of (1) by ESI/MS, NMR and IR spectroscopy and X-ray crystallography

The crystals of (1) formed in the bulk synthesis were first analysed via ESI/MS in both the positive and negative ion mode. The former gave an almost identical mass spectrum to that shown in Fig. 1a (data not shown), while the latter gave an abundant signal due to the BF₄⁻ counter ion (ESI Fig. S3†). IR spectroscopy confirmed the presence of both BF₄⁻ and BH₄⁻ (ESI Fig. S4†).

We next attempted to characterise 1via various NMR experiments (ESI Fig. S5–S16†). 1 was dissolved into cold CD₃CN to produce a saturated solution and this solution immediately introduced into the pre-cooled NMR probe at −15 °C. The ¹H NMR spectrum collected at −15 °C displayed a very broad multiplet centred at 0.5 ppm, attributed to coordinated BH₄, from which no fine structure could be resolved (ESI Fig. S5†). This may be due to (i) the fluxional nature of the binding of BH₄, (ii) the complex splitting patterns due to spin-active nuclei and isotopomers of silver (^107/109Ag) and boron (^10/11B) and (iii) the influence of the quadrupolar ¹⁰B nucleus. This signal collapsed into a broad singlet at 0.5 ppm under ¹¹B-decoupling (ESI Fig. S6 and S7†). There was no apparent change to the signals upon ³¹P-decoupling (ESI Fig. S8 and S9†). The ¹H NMR spectrum further displayed a broad multiplet centred at 4.5 ppm attributed to the coordinated hydride (ESI Fig. S5†), the peaks of which sharpened slightly under ³¹P-decoupling (ESI Fig. S8 and S9†). The ¹H-decoupled ³¹P NMR spectrum displayed a broad peak at 0.56 ppm, shifted downfield from the free ligand resonance at 23.1 ppm (ESI Fig. S10†). The ¹H-decoupled ¹¹B NMR spectrum displays two resonances at 0.57 and 41.49 ppm attributable to BF₄^– and BH₄, respectively (ESI Fig. S11†), as confirmed via the ¹⁹F-¹¹B and ¹H-¹¹B HSQC NMR experiments (ESI Fig. S12 and S13† respectively). The ¹H-decoupled ¹⁹F NMR spectrum displays a resonance at −150.16 ppm corresponding to BF₄^– (ESI Fig. S14†). The ¹H-decoupled ¹³C NMR spectrum displays resonances attributable to coordinated phosphine ligand (ESI Fig. S15†).

Heating the sample from −15 °C to +25 °C in the NMR probe enabled the collection of ¹H data at various temperatures. The most obvious change in the spectra with time at −15 °C and then upon heating was the increase in the intensity of the singlet at 4.56 ppm which is attributable to dissolved H₂ and the corresponding reduction in the intensity of the coordinated hydride signal (ESI Fig. S16†).

This series of experiments required the preparation of several different samples, as it was noted that 1 appears to undergo decomposition/reactions in these highly concentrated solutions, ultimately resulting in precipitation of a black material after approximately 3 hours at 25 °C. This rapid change in solution, even at −15 °C, precluded the overnight or longer acquisition of a ¹⁰⁹Ag NMR spectrum.

X-ray crystallography was used to determine the structure of [Ag₃(H)(BH₄)L^Ph₃]BF₄.§ The cation [Ag₃(H)(BH₄)L^Ph₃]⁺ which has crystallographic 3-fold symmetry, Fig. 2, consists of a trinuclear core with silver(I) ions occupying the vertices of an equilateral triangle.

Fig. 2 ORTEP-3 representations of: (a) the cation present in [Ag₃(μ₃-H)(μ₃-BH₄)L^Ph₃]BF₄ and, (b) the trinuclear silver hydride/borohydride ‘Ag₃(μ₃-H)(μ₃-BH₄)’ core, where phenyl rings are omitted for clarity. Displacement ellipsoids are set at the 50% probability level. Ag(1)–Ag(1) 2.9100(3), Ag(1)–P(1) 2.4483(5), Ag(1)–P(2) 2.4486(5), Ag(1)–H 1.93(3), Ag(1)–H(1a) 2.17(3), B(1)–H(1a) 1.10(3), B(1)–H(1b) 1.07(6).

The Ag(1)–Ag(1) distances that connect the edges of the triangle are 2.9100(3) Å. The hydride H which lies on a crystallographic 3-fold axis is 0.96 Å displaced from the plane defined by the triangular silver(I) core and is coordinated to all silver(I) ions as a μ₃-bridging ligand with a Ag(1)–H distance of 1.93(3) Å and Ag(1)–H–Ag(1) angle of 97.5(3)°. Relative to the μ₃-hydride H, the opposing face of the trinuclear core has a μ₃ borohydride with distorted tetrahedral geometry, the boron lies on the 3-fold axis of symmetry and three symmetry related hydrides (H1a) coordinate to the three silver atoms. Related μ₃-borohydride binding to the metal triangle of Fe(CO)₃ fragments has been reported for the trinuclear cluster [Fe₃(μ-H)(μ₃-BH₄)(CO)₉].²⁷ The boron–hydrogen bond lengths for BH₄⁻ are: B(1)–H(1a) 1.10(3) Å and B(1)–(H1b) 1.07(6) Å. There are two types of bond angles in the BH₄⁻ tetrahedron: (i) H(1a)–B(1)–H(1a) 110.1(16)° and (ii) H(1a)–B(1)–H(1b) 108.9(16)°. The tetrahedral face H(1a)–H(1a)–H(1a), is parallel to the larger Ag(1)–Ag(1)–Ag(1) plane. Each of the three hydrogen atoms of BH₄⁻, H(1a), are individually coordinated to one silver(I) ion where H(1a)–Ag(1) is 2.17(3) Å. The core is surrounded by three μ₂-bridging phosphine ligands which coordinate to two silver(I) ions through P(1) and P(2): Ag(1)–P(1) is 2.4483(5) Å and Ag(1)–P(2) is 2.4486(5) Å. The distance between P(1) and P(2) within the chelate ring is 3.041 Å and the P(1)–C(20)–P(2) angle is 110.8(1)°. P(1) is below the plane of the trinuclear core and both C(20) and P(2) are above the plane with C(20) above P(2).

The phenyl rings extending from the phosphorus atoms adopt one of three distinct geometrical conformations. Both P(1) and P(2) have one phenyl ring towards the μ₃-hydride. A pseudoequatorial phenyl ring in regards to the Ag₃ triagnular plane extends from P(1). One phenyl ring from each P(2) extends toward the μ₃-borohydride.

The structures of [Ag₃(μ₃-H)(μ₃-BH₄)L^Ph₃]⁺, [Ag₃(μ₃-D)L^Ph₃]²⁺ (ref. 10) and [Ag₃(μ₃-H)(μ₃-Cl)L^Ph₃]⁺ (ref. 9) are compared in Table 1. All structures consist of a trinuclear silver(I) core and maintain a triangular geometry. The dication [Ag₃(μ₃-D)L^Ph₃]²⁺ has the longest Ag–Ag and Ag–P interactions at 3.1193 and 2.4632 Å respectively. All hydrides coordinate as μ₃-bridging ligands where the longest Ag–H interaction exists in [Ag₃(μ₃-H)(μ₃-Cl)L^Ph₃]⁺ at 1.91(2) Å.

Table 1 A comparison of selected bond lengths (Å) and angles (°) with estimated standard deviations in parentheses for [Ag₃(μ₃-D)L^Ph₃](BF₄)₂,⁹ [Ag₃(μ₃-H)(μ₃-Cl)L^Ph₃]BF₄,¹⁰ and [Ag₃(μ₃-H)(μ₃-BH₄)L^Ph₃]BF₄

	[Ag3(μ3-D)L^Ph3]^2+	[Ag3(μ3-H)(μ3-Cl)L^Ph3]^+	[Ag3(μ3-H)(μ3-BH4)L^Ph3]^+
Ag–Ag	3.1193	2.8988(2)	2.9100(3)
Ag-(H/D)	1.83	1.91(2)	1.93(3)
Ag-(Cl/BH4)	NA	2.859(1)	2.17(3)
Ag–P	2.4632	2.4421(9)	2.4486(5)

---

The BF₄⁻ counterion is disordered over two crystallographic special positions, a −3 site with normal 1/6 occupancy and on a 3-fold axis with 50% the normal occupancy of 1/3, with the same position occupied the other 50% by a molecule of acetonitrile. The relative occupancy at these special positions were supported by SQUEEZE calculations²⁹ which were not used to remove these disordered components. The Ag bound hydride and BH₄ hydrides were located on Fourier difference maps and refined isotropically without restraint.

Unimolecular gas-phase chemistry of [Ag₃(H)(BH₄)L₃]⁺

Given that AgBH₄ is known to undergo thermal decomposition reactions that liberate BH₃ (eqn (2)),^1,23 we were interested in examining whether such reactions occur in the gas phase for isolated, stoichiometrically well defined cluster cations. Thus, CID was carried out in a 2D linear ion trap to probe the low-energy fragmentation pathways of [Ag₃(H)(BH₄)L₃]⁺, where the cluster identity has been confirmed by high resolution mass spectrometry (HRMS).

Mass selection and subsequent CID of [Ag₃(H)(BH₄)L^Ph₃]⁺ (m/z 1493, HRMS (ESI Fig. S1a and b†)) to ca. 50% relative intensity yields [Ag₃(H)(BH₄)L^Ph₂]⁺ (m/z 1109, HRMS (ESI Fig. S1c and d†)) via neutral ligand loss (eqn (5)) as the main fragmentation channel. [Ag₃(H)₂L^Ph₂]⁺ (m/z 1095, HRMS (ESI Fig. S17a and b†)), [Ag₃(H)(BH₄)L^Ph]⁺ (m/z 723, (ESI Fig. S1e and f†)) and [Ag₃(H)₂L^Ph]⁺ (m/z 709, (ESI Fig. S17c and d†)); [Ag₂(H)L^Ph]⁺ (m/z 601) and; [AgL^Ph]⁺ (m/z 491) are all also observed.

The main fragmentation channel of [Ag₃(H)(BH₄)L^Me₃]⁺ (m/z 747, HRMS (ESI Fig. S2†)) involves neutral ligand loss (eqn (5)) to form [Ag₃(H)(BH₄)L^Me₂]⁺ (m/z 610, Fig. 3b). Other ions observed include [Ag₃(H)₂L^Me₂]⁺ (m/z 596), [Ag₃(H)(BH₄)L^Me]⁺ (m/z 475) and [Ag₃(H)₂L^Me]⁺ (m/z 461).

Fig. 3 LTQ CID of (a) [Ag₃(H)(BH₄)L^Ph₃]⁺, m/z 1493; (b) [Ag₃(H)(BH₄)L^Me₃]⁺, m/z 747. The most intense peak in the cluster is represented by the m/z value. *Refers to the mass-selected precursor ion.

Energy-resolved CID (ERCID) was used in a 3D ion trap to determine whether the product ions in Fig. 3a are due to primary fragmentation pathways of [Ag₃(H)(BH₄)L^Ph₃]⁺ or secondary fragmentation of primary fragment ions (ESI Fig. S18†). The onset of ligand loss (eqn (5)) begins at ca. 0.6 V and continues to steadily increase up until 0.8 V (ESI Fig. S18†). An increase in the collision voltage beyond this point results in the consumption of [Ag₃(H)(BH₄)L^Ph₂]⁺ (m/z 1109) and the increase of [Ag₃(H)₂L^Ph₂]⁺ (m/z 1095). These results suggest that the primary product ion upon CID of [Ag₃(H)(BH₄)L^Ph₃]⁺ arises from ligand loss and that ions of lower m/z are from subsequent secondary fragmentation of [Ag₃(H)₂L^Ph₂]⁺ (m/z 1095).

[Ag₃(H)(BH₄)L₃]⁺ → [Ag₃(H)(BH₄)L₂]⁺ + L (5)

Unimolecular gas-phase chemistry of [Ag₃(H)(BH₄)L₂]⁺

The primary product ions formed via ligand loss (eqn (5), Fig. 3) were mass selected and subjected to CID in the 2D ion trap (Fig. 4). Mass selection and subsequent CID of [Ag₃(H)(BH₄)L^Ph₂]⁺ (m/z 1109, Fig. 4a) to ca. 30% relative intensity yields [Ag₃(H)₂L^Ph₂]⁺ (m/z 1095) via neutral borane loss (eqn (5)) as a minor fragmentation channel and [Ag₃(H)₂L^Ph]⁺ (m/z 709) as the main fragmentation channel; (ii) [Ag₂(H)L^Ph]⁺ (m/z 601) and; (iii) [AgL^Ph]⁺ (m/z 491) are all also observed. The main fragmentation channel upon CID of [Ag₃(H)₂L^Me₂]⁺ (m/z 596, Fig. 4b) involves neutral borane loss (eqn (6)). Other ions observed include [Ag₃(H)(BH₄)L^Me]⁺ (m/z 475), and [Ag₃(H)₂L^Me]⁺ (m/z 461).

Fig. 4 LTQ CID of: (a) [Ag₃(H)(BH₄)L^Ph₂]⁺, m/z 1109); (b) [Ag₃(H)(BH₄)L^Me₂]⁺, m/z 610. The most intense peak in the cluster is represented by the m/z value. *Refers to the mass-selected precursor ion. #Refers to background noise.

Once again, ERCID was used in a 3D ion trap to determine which of the product ions observed in Fig. 4a were primary (ESI Fig. S19†). [Ag₃(H)₂L^Ph₂]⁺ (m/z 1095) begins to appear at 0.5 V upon CID [Ag₃(H)(BH₄)L^Ph₂]⁺ (m/z 1109), which corresponds to BH₃ loss (eqn (6)). A minor primary fragmentation channel assigned to neutral ligand loss (eqn (7)) is observed at around 0.6 V. Although decomposition reactions of coordinated ligands in metal complexes and clusters have been well studied in the gas-phase,^30,31 this appears to be the first experimental report on the gas-phase decomposition of a coordinated BH₄⁻ ligand via BH₃ loss. This reaction is related to that described by Wiberg and Henle (eqn (2)).¹

[Ag₃(H)(BH₄)L₂]⁺ → [Ag₃(H)₂L₂]⁺ + BH₃ (6)

[Ag₃(H)(BH₄)L₂]⁺ → [Ag₃(H)(BH₄)L]⁺ + L (7)

Unimolecular gas-phase chemistry of [Ag₃(H)(BH₄)L]⁺

[Ag₃(H)(BH₄)L^Ph]⁺ (m/z 723), formed via sequential ligand losses (eqn (4) and (6)), was mass selected and allowed to undergo CID. The sole fragmentation channel observed is due to the loss of neutral BH₃ (eqn (8)) (ESI Fig. S20†).

[Ag₃(H)(BH₄)L]⁺ → [Ag₃(H)₂L]⁺ + BH₃ (8)

Unimolecular gas-phase chemistry of [Ag₃(H)₂L_n]⁺

The dihydride clusters [Ag₃(H)₂L^Ph_n]⁺ where n = 2, 1 were subjected to ERCID using a 3D ion trap (ESI Fig. S21 and S22†). The major primary fragmentation of [Ag₃(H)₂L^Ph₂]⁺ (m/z 1095) occurs via ligand loss (eqn (9)), with an onset requiring ca. 0.6 V. In contrast, The major primary fragmentation of [Ag₃(H)₂L^Ph]⁺ (m/z 708) involves AgH loss, as previously described,³² with the onset of fragmentation occurring at ca. 0.4 V.

[Ag₃(H)₂L₂]⁺ → [Ag₃(H)₂L]⁺ + L (9)

[Ag₃(H)₂L]⁺ → [Ag₂(H)L]⁺ + AgH (10)

Computational study of BH₄⁻ decomposition triggered via ligand loss in the clusters [Ag₃(H)(BH₄)L^Me_n]⁺ (n = 1–3)

To better understand how the number of diphosphine ligands, n, in the clusters, [Ag₃(H)(BH₄)L_n]⁺, influence the competition between decomposition of the ligated BH₄⁻ (eqn (11)) versus loss of a ligand (eqn (12)), we turned to DFT calculations to examine the structures and energetics of the reactants and products of eqn (11) and (12) for the case of clusters containing L^Me ligands. The initial geometry for [Ag₃(H)(BH₄)L^Me₃]⁺ was that related to the core structure from the X-ray structure for 1. Thus changing the phosphine substitutent from Ph to Me has little effect on the core structure. To calculate fragment ion structures, either BH₃ or L^Me was removed and the resultant fragment was allowed to fully optimise.¶||

[Ag₃(H)(BH₄)L_n]⁺ → [Ag₃(H)₂L_n]⁺ + BH₃ (11)

[Ag₃(H)(BH₄)L_n]⁺ → [Ag₃(H)(BH₄)L_n−1]⁺ + L (12)

To understand why BH₄⁻ in [Ag₃(H)(BH₄)L₃]⁺ bridges all three Ag atoms via three separate two-electron, two-centre (2e,2c) bonds, the structure and bonding of this cluster was analysed based on simple electron counting rules.^33–37 As expected, a transition metal centre with a d¹⁰ electron configuration has four empty orbitals. Two of these four are always available for ligand coordination to give linear complexes [ML₂]⁺. However, the availability of the other two orbitals depends on the identity of the L ligands. For example, if the L–M–L bond angle in [ML₂]⁺ using the bidentate ligands is forced to be bent, the two extra orbitals become available and as a result the tetrahedral complex [ML₄]⁺ is formed. Also, the monodentate L ligands with relatively weak σ-donor abilities increase the possibility of all four orbitals on the metal centre being available.

In [Ag₃(H)(BH₄)L₃]⁺, the presence of the three bidentate phosphine ligands render all the four empty orbitals on three Ag centres susceptible to coordination. In such a case, the cluster has 3 × 4 = 12 available orbitals. Six of these twelve orbitals are occupied by the phosphine ligands. Three of them are involved in interaction with the hydride ligand (a μ₃-bridging ligand) via a four-centre two-electrons bonding mode. Finally, the last three orbitals on Ag centres overlap with three filled B–H σ orbitals of BH₄⁻, leading to coordination of BH₄⁻ in μ₃-form. The 50 valence electron [Ag₃(H)(BH₄)L₃]⁺ cluster is not expected to have direct metal–metal interactions, consistent with other related M₃L₆ clusters where M has a d¹⁰ electron configuration.³⁵ This is highlighted by an examination of its HOMO, which suffers from the Ag–Ag anti-bonding interactions derived from the silver d_xz orbitals (Fig. 6). The short Ag–Ag bond distances in [Ag₃(H)(BH₄)L₃] (2.971–2.993 Å) can be mainly rationalised by the presence of the hydride ligand that creates the four-centre two-electron bonds with the Ag centres. A similar metal–metal bond distance was also observed by Harvey et al. in the [Pd₃(H₂PCH₂PH₂)₃(CO)(H)]⁻ cluster (2.932 Å), where the corresponding anti-bonding skeletal molecular orbitals are also fully occupied.³⁸

With regards to the unimolecular fragmentation chemistry, the calculations indicate that the first loss of the ligand, L (eqn (12)), from [Ag₃(H)(BH₄)L^Me₃]⁺ results in transferring the hydride to Ag² and causes this centre to adopt a mainly linear structure with a very weak interaction with Ag¹ (Fig. 5). The presence of the very strong σ donating hydride ligand on the Ag² centre makes the two empty orbitals on Ag² less available and thus does not allow BH₄⁻ to strongly interact with them. In this case, BH₄⁻ is only able to interact with the Ag¹ and Ag³ centres to give a μ²-coordination mode (Fig. 5). If we ignore the weak interactions between Ag centres in [Ag₃(H)(BH₄)L^Me₂]⁺, the Ag¹ and Ag³ centres can be considered as three coordinate centres. For [Ag₃(H)(BH₄)L^Me₂]⁺, the Ag²–H σ orbital interacts with one of the empty orbitals on Ag³ and creates a 3-centre 2-electron bond between two silver(I) ions via a μ₂-bridging hydride ligand.

Fig. 5 DFT calculated structures and energetics for the competition between BH₃ loss and L ligand loss. Hydrogen atoms on the L^Me ligands are omitted for clarity.

Fig. 6 Depiction of the DFT calculated HOMO of [Ag₃(H)(BH₄)L^Me₃]⁺.

From loss of a second L^Me ligand, bonding in the product [Ag₃(H)(BH₄)L^Me]⁺ can be viewed as interaction of the linear complex [BH₄–Ag³–H]⁻ with [Ag₂L^Me]²⁺ (Fig. 5). In this cluster, an Ag³–H σ orbital interacts with an empty orbital on Ag², and BH₄⁻ bridges Ag³ to Ag¹ through three of its B–H bonds. Therefore, each Ag(I) centre in [Ag₃(H)(BH₄)L^Me]⁺ experiences a linear two coordinated environment.

Our calculations show that, in excellent agreement with the CID data, the BH₃ loss becomes easier as the number of ligands, L, decreases. The loss of the ligand, L, increases the electron deficiency of the metal centres, leading to the stronger coordination of BH₄⁻ to the Ag centres, as evident from the shorter Ag–H(BH₄) and longer B–H bond distances in [Ag₃(H)(BH₄)L^Me]⁺ (Fig. 5). In other words, ligand loss enhances the acidity of Ag centres and makes them more prone to compete with BH₃ for hydride abstraction. The better the competition, the easier the BH₃ loss.

In contrast to BH₃ loss (eqn (11)) neutral ligand loss (eqn (12)) from [Ag₃(H)(BH₄)L^Me]⁺ is more difficult than that from [Ag₃(H)(BH₄)L^Me₃]⁺ and [Ag₃(H)(BH₄)L^Me₂]⁺, supported by the DFT calculations. This difference can be rationalised in terms of the molecular orbital approach.³⁹ In general, the HOMO of d¹⁰ complexes (ML_n) with a coordination number greater than two (n > 2) suffers from a slight anti-bonding interaction between L and M, leading to weakening of the M–L bonds. However, this anti-bonding interaction disappears in linear d¹⁰-ML₂ complexes, causing the M–L bonds in ML₂ to be much stronger than those in ML₃ and ML₄. As mentioned above, clusters [Ag₃(H)(BH₄)L^Me₃]⁺ and [Ag₃(H)(BH₄)L^Me₂]⁺ have the Ag centres not present as two-coordinate, and thus the loss of the ligand, L, from these clusters is relatively easy. By contrast, all the Ag centres in [Ag₃(H)(BH₄)L^Me]⁺ are mainly two-coordinate, thereby not having the relevant anti-bonding interaction, forming very strong M–L bonds.

Conclusions

The sodium borohydride induced reduction of silver(I) salts to form nanoparticles has been described as a “black-box” synthesis.¹⁹ While it is now well established that there are different growth stages,^18–20 the actual molecular species associated during growth to nanoparticles and the mechanisms for growth are not fully understood. By studying the formation and reactions of small ligand protected nanoclusters, we are able to better understand the fundamental interactions between silver salts and borohydride.

We have previously shown that BH₄⁻ is a source of hydride for [Ag₃(μ₃-H)L^Ph₃]²⁺ (ref. 10) and [Ag₃(μ₃-H)(μ₃-Cl)L^Ph₃]⁺ (ref. 9) at ambient conditions, however at −10 °C the decomposition of BH₄⁻ can be prevented and coordination to silver(I) ions can occur to yield [Ag₃(μ₃-H)(μ₃-BH₄)L^Ph₃]BF₄. This is the first silver nanocluster containing a “captured” borohydride anion, and may have relevance to binding of BH₄⁻ to silver nanoparticle surfaces,^20,21 or bulk silver metal surfaces.⁴⁰ Given that nanoclusters such as [{Ag₇(μ₄-H)(E₂P(OR)₂}₆] are precursors to further growth into silver nanoparticles,²¹ it will be interesting to establish whether the nanoclusters [Ag₃(μ₃-H)(μ₃-Cl)L^Ph₃]BF₄,⁹ [Ag₃(μ₃-H)L^Ph₃](BF₄)₂,¹⁰ and [Ag₃(μ₃-H)(μ₃-BH₄)L^Ph₃]BF₄ can further grow into silver nanoparticles.

The current and previous³² gas-phase experiments and DFT calculations on [Ag₃(H)_2−x(BH₄)_xL_n]⁺ clusters provide a direct link between mixed hydride/borohydride silver clusters (x = 1), dihydride silver clusters (x = 0)³² and silver clusters via discrete unimolecular reactions occurring for isolated clusters (Scheme 2). Thus CID triggers loss of the ligand, L (eqn (12)), resulting in a change in the binding mode(s) of the H and BH₄ ligands (Fig. 5). Perhaps related reactions occur at the surfaces of silver nanoparticles, which might drive the development of catalysts for hydrogen storage applications.⁴¹

Scheme 2 Direct link established between mixed hydride/borohydride clusters (Black), dihydride clusters (Blue) and “all metal” clusters (Red) based on gas-phase unimolecular fragmentation reactions of mass selected clusters reported here using CID and in ref. 32 using laser-induced dissociation (LID).

Experimental

Synthesis of solution phase silver clusters for MS analyses

Silver(I) tetrafluoroborate (1.9 mg, 0.010 mmol) and bis(diphenylphosphino)methane (3.8 mg, 0.010 mmol) in 20 mL acetonitrile were added to a 25 mL Quickfit Erlenmeyer flask equipped with a magnetic stirrer and stopper. The solution was cooled to ca.−10 °C by immersing the reaction flask in an ice/water bath above the solvent level. All reagents were kept in the dark and flasks covered in foil. Sodium borohydride (5.7 mg, 0.150 mmol) was added as a powder and the solution changed colour from clear to light yellow.

Synthesis of crystalline [Ag₃(H)(BH₄)(L^Ph)₃]BF₄ (1)

Silver(I) tetrafluoroborate (194 mg, 1.0 mmol) and bis(diphenylphosphino)methane (384 mg, 1.0 mmol) in 100 mL acetonitrile were added to a 250 mL Quickfit round bottomed flask equipped with a magnetic stirrer and stopper. The solution was cooled to −10 °C by immersing the reaction flask in an ice/water bath above the solvent level. All reagents were kept in the dark and flasks covered in foil. Sodium borohydride (57.0 mg, 1.50 mmol), was added as a powder and the solution changed colour from clear to light yellow over ca. 5 minutes. The solution was filtered after stirring for 3 hours and frozen solid by immersing the flask in liquid nitrogen. While frozen, 100 mL of diethylether was added and the flask moved to the fridge. After 72 hours crystalline material was formed and characterised by X-ray crystallography.

NMR spectroscopy experiments

The NMR experiments were performed on a Bruker Avance 400 NMR spectrometer (400.13 MHz ¹H frequency) equipped with a 5 mm triple resonance broadband probe (BB/²H–¹H/¹⁹F). Solutions for analysis by NMR were prepared by dissolving [Ag₃(μ₃-H)(μ₃-BH₄)L^Ph₃]BF₄ in 0.6 ml of deuteroacetonitrile. NMR experiments were performed with the sample held at temperatures between −15 °C and +25 °C (±0.1 °C). Chemical shifts for ¹H experiments are referenced to the residual protonated solvent signal (CD₂HCN, δ 1.94 ppm); ¹¹B externally referenced to BF₃·OEt₂ capillary in CD₃CN; ¹³C referenced to the solvent signal (CD₃CN, δ 1.39 ppm); ¹⁹F externally referenced to a CFCl₃ in CD₃CN (δ 0.00 ppm); ³¹P externally referenced to a 85% H₃PO₄ capillary in CDCl₃ (δ 0.00 ppm). One- and two-dimensional NMR experiments were acquired using standard Bruker library pulse sequences.

Mass spectrometry

Mass spectra were recorded using a Finnigan hybrid linear quadrupole (LTQ) Fourier transform ion cyclotron resonance (FTICR) mass spectrometer. The silver clusters prepared in the solution phase were diluted to 50 μM and introduced into the mass spectrometer via a syringe pump set at a flow rate of 5 μL min⁻¹ to the ESI capillary. The ESI conditions used, for optimum intensity of the target ions, typically were: spray voltage, 4.2–5.0 kV, capillary temperature, 250 °C, nitrogen sheath gas pressure, 5 (arbitrary units), capillary voltage 25 V, tube lens voltage 15 V. Selected ions were transferred to the FTICR cell for accurate mass measurement with the use of selected ion monitoring (SIM) and selected reaction monitoring (SRM) to obtain the most reliable results. The unimolecular fragmentation of silver clusters was examined via CID. The mass-selected precursor ion was depleted to 10–20% using a normalised collision energy typically between 20–25% and a mass selection window of 15 Th to isolate the full range of isotopes due to boron and silver isotopes.

Energy resolved CID experiments were carried out using a Finnigan 3D ion trap (LCQ) mass spectrometer. The method of Broadbelt was adapted.⁴² The silver clusters were diluted to 50 μM and introduced into the mass spectrometer via a syringe pump set at 5 μL min⁻¹ through a Finnigan ESI source. The source conditions used for optimum intensity of the target ions were: spray voltage 4.5–5.1 kV, capillary temperature 200 °C, nitrogen sheath gas pressure, 50 (arbitrary units), capillary voltage 30 V, tube lens voltage −55 V. The mass-selected precursor ion was isolated with a mass selection window of 15 Th. The normalised collision energy (NCE) was increased incrementally by 1.0% typically starting from a NCE where no fragmentation is observed, until reaching the NCE required for depleting the precursor ion to <5% relative intensity. The NCE was converted to an amplitude of the resonance excitation RF voltage (tick amp) as described in the ESI.† The relative intensity of precursor and product ions were plotted as a function of the increasing amplitude to determine: (i) the onset of precursor fragmentation and (ii) the assignment of product ions as primary or secondary fragments of the mass-selected silver cluster.

Crystallography

Intensity data for compound 1 was collected on an Oxford Diffraction SuperNova CCD diffractometer using Cu-Kα radiation, the temperature during data collection was maintained at 130.0(1) using an Oxford Cryostream cooling device. The structure was solved by direct methods and difference Fourier synthesis.⁴³ The thermal ellipsoid plot was generated using the program ORTEP-3 ⁴⁴ integrated within the WINGX⁴⁵suite of programs. The BF₄⁻ counterion was disordered over two crystallographic special positions, a −3 site with normal 1/6 occupancy and on a 3-fold axis with 50% the normal occupancy of 1/3, with the same position occupied the other 50% by a molecule of acetonitrile. The Ag bound hydride and BH₄ hydrides were located on Fourier difference maps and refined isotropically without restraint.

Crystal data for 1: C₇₅H₇₁B₂F₄P₆Ag₃·(0.25 CH₃CN) M = 1589.63, T = 130.0(2) K, λ = 1.54184 Å, cubic, space group Pa3̄, a = 24.1922(1) Å, V = 14158.79(18) ų, Z = 8, D_c = 1.491 mg M⁻³μ(Cu-Kα) 8.296 mm⁻¹, F(000) 6428, crystal size 0.17 × 0.16 × 0.09 mm. 67033 reflections measured, 4996 independent reflections (R_int = 0.0442), the final R was 0.0279 [I > 3(I) 4853 data] and wR(F) (all data) was 0.0735.

Density functional theory

Computational details.^46–53 Gaussian 09 ⁴⁶ was used to fully optimise all the structures reported in this paper at the M06 level of density functional theory.^47,48 The effective-core potential of Hay and Wadt with a double-ξ valence basis set (LANL2DZ) was chosen to describe Ag. The 6-31G(d) basis set was used for other atoms. Polarization functions were also added for Ag (ξf = 1.611). This basis set combination will be referred to as BS1. To further refine the energies obtained from the M06/BS1 calculations, we carried out single-point energy calculations for all of the structures with a larger basis set (BS2) at the M06 level of theory. BS2 utilises the def2-TZVP basis set on all atoms. Effective core potentials including scalar relativistic effects were used for silver atom. We have used the corrected potential energies obtained from the M06/BS2//M06/BS1 calculations throughout the paper unless otherwise stated.

Acknowledgements

We thank the ARC for financial support via grants DP1096134 (to GNK) and DP150101388 (to RAJO and AJC). The authors gratefully acknowledge the generous allocation of computing time from the University of Tasmania and the National Computing Infrastructure. We thank Assoc. Prof. Paul Donnelly for useful discussions.

Notes and references

1. E. Wiberg and W. Henle, Z. Naturforsch., B: J. Chem. Sci., 1952, 7, 575–576.
2. J. C. Bommer and K. W. Morse, Inorg. Chem., 1980, 19, 587–593.
3. F. Cariati and L. Naldini, Gazz. Chim. Ital., 1965, 95, 201–205.
4. E. B. Lobkovskii, M. Y. Antipin, A. P. Borisov, V. D. Makhaev, K. N. Semenenko and Y. T. Struchkov, Koord. Khim., 1981, 7, 307–310.
5. C. W. Liu, H.-W. Chang, C.-S. Fang, B. Sarkar and J.-C. Wang, Chem. Commun., 2010, 46, 4571–4573.
6. C. W. Liu, H.-W. Chang, B. Sarkar, J.-Y. Saillard, S. Kahlal and Y.-Y. Wu, Inorg. Chem., 2010, 49, 468–475.
7. C. W. Liu, P.-K. Liao, C.-S. Fang, J.-Y. Saillard, S. Kahlal and J.-C. Wang, Chem. Commun., 2011, 47, 5831–5833.
8. B. K. Tate, C. M. Wyss, J. Bacsa, K. Kluge, L. Gelbaum and J. P. Sadighi, Chem. Sci., 2013, 4, 3068–3074.
9. A. Zavras, G. N. Khairallah, T. U. Connell, J. M. White, A. J. Edwards, P. S. Donnelly and R. A. J. O'Hair, Angew. Chem., Int. Ed., 2013, 52, 8391–8394.
10. A. Zavras, G. N. Khairallah, T. U. Connell, J. M. White, A. J. Edwards, R. J. Mulder, P. S. Donnelly and R. A. J. O'Hair, Inorg. Chem., 2014, 53, 7429–7437.
11. A. Desireddy, B. E. Conn, J. Guo, B. Yoon, R. N. Barnett, B. M. Monahan, K. Kirschbaum, W. P. Griffith, R. L. Whetten, U. Landman and T. P. Bigioni, Nature, 2013, 501, 399–402.
12. H. Yang, J. Lei, B. Wu, Y. Wang, M. Zhou, A. Xia, L. Zheng and N. Zheng, Chem. Commun., 2013, 49, 300–302.
13. H. Yang, Y. Wang, H. Huang, L. Gell, L. Lehtovaara, S. Malola, H. Hakkinen and N. Zheng, Nat. Commun., 2013, 4, 2422.
14. H. Yang, Y. Wang and N. Zheng, Nanoscale, 2013, 5, 2674–2677.
15. R. S. Dhayal, J.-H. Liao, Y.-C. Liu, M.-H. Chiang, S. Kahlal, J.-Y. Saillard and C. W. Liu, Angew. Chem., Int. Ed., 2015, 54, 3702–3706.
16. L. G. AbdulHalim, M. S. Bootharaju, Q. Tang, S. del Gobbo, R. G. AbdulHalim, M. Eddaoudi, D.-e. Jiang and O. M. Bakr, J. Am. Chem. Soc., 2015, 137, 11970–11975.
17. C. P. Joshi, M. S. Bootharaju, M. J. Alhilaly and O. M. Bakr, J. Am. Chem. Soc., 2015, 137, 11578–11581.
18. J. Polte, X. Tuaev, M. Wuithschick, A. Fischer, A. F. Thuenemann, K. Rademann, R. Kraehnert and F. Emmerling, ACS Nano, 2012, 6, 5791–5802.
19. M. Wuithschick, B. Paul, R. Bienert, A. Sarfraz, U. Vainio, M. Sztucki, R. Kraehnert, P. Strasser, K. Rademann, F. Emmerling and J. Polte, Chem. Mater., 2013, 25, 4679–4689.
20. D. L. Van Hyning and C. F. Zukowski, Langmuir, 1998, 14, 7034–7046.
21. J.-S. Seo, D.-M. Son, H. Lee, J. Kim and Y. Kim, Bull. Korean Chem. Soc., 2009, 30, 2651–2654.
22. Q.-Q. Xu, X.-Y. Dong, R.-W. Huang, B. Li, S.-Q. Zang and T. C. W. Mak, Nanoscale, 2015, 7, 1650–1654.
23. D. G. Musaev and K. Morokuma, Organometallics, 1995, 14, 3327–3334.
24. C. W. Liu, Y.-R. Lin, C.-S. Fang, C. Latouche, S. Kahlal and J.-Y. Saillard, Inorg. Chem., 2013, 52, 2070–2077.
25. V. D. Makhaev, Russ. Chem. Rev., 2000, 69, 727–746.
26. M. Besora and A. Lledós, Struct. Bonding, 2008, 130, 149–202.
27. J. Vites, C. Eigenbrot and T. P. Fehlner, J. Am. Chem. Soc., 1984, 106, 4633–4635.
28. A. J. Clark, A. Zavras, G. N. Khairallah and R. A. J. O'Hair, Int. J. Mass Spectrom., 2015, 378, 86–94.
29. P. Van der Sluis and A. L. Spek, Acta Crystallogr., Sect. A: Found. Crystallogr., 1990, A46, 194–201.
30. R. A. J. O'Hair, Gas Phase Ligand Fragmentation to Unmask Reactive Metallic Species, in Reactive Intermediates. MS Investigations in Solution, ed. L. S. Santos, Wiley-VCH, Weinheim, 2010, ch. 6, pp.199–227, ISBN: 978-3-527-32351-7.
31. R. A. J. O'Hair and N. J. Rijs, Acc. Chem. Res., 2015, 48, 329–340.
32. M. Girod, M. Krstić, R. Antoine, L. MacAleese, J. Lemoine, A. Zavras, G. N. Khairallah, V. Bonačić-Koutecký, P. Dugourd and R. A. J. O'Hair, Chem. – Eur. J., 2014, 20, 16626–16633.
33. D. M. P. Mingos, T. Slee and Z. Lin, Chem. Rev., 1990, 90, 383–402.
34. D. M. P. Mingos and D. J. Wales, Introduction to Cluster Chemistry, Prentice Hall, 1990.
35. D. G. Evans and D. M. P. Mingos, J. Organomet. Chem., 1982, 240, 321–327.
36. D. M. P. Mingos, J. Organomet. Chem., 2004, 689, 4420–4436.
37. F. K. Sheong, W.-J. Chen and Z. Lin, J. Organomet. Chem., 2015, 792, 93–101.
38. C. Cugnet, D. Lucas, E. Collange, B. Hanquet, A. Vallat, Y. Mugnier, A. Soldera and P. D. Harvey, Chem. – Eur. J., 2007, 13, 5338–5346.
39. H. Batebi, F. Zarkoob, K. Daraei, B. F. Yates and A. Ariafard, J. Organomet. Chem., 2013, 748, 89–97.
40. M. C. Sison Escaño, E. Gyenge, R. Lacdao Arevalo and H. Kasai, J. Phys. Chem. C, 2011, 115, 19883–19889.
41. W. Grochala and P. P. Edwards, Chem. Rev., 2004, 104, 1283–1315.
42. A. Colorado and J. Brodbelt, J. Am. Soc. Mass Spectrom., 1996, 7, 1116–1125.
43. G. M. Sheldrick, Acta Crystallogr., Sect. A: Found. Crystallogr., 2008, 64, 112–122.
44. L. J. Farrugia, J. Appl. Crystallogr., 1997, 30, 565.
45. L. J. Farrugia, J. Appl. Crystallogr., 1999, 32, 837–838.
46. M. J. Frisch, et al., Gaussian 09, revision D.01, Gaussian, Inc., Wallingford, CT, 2009. For the complete reference see the ESI.†.
47. Y. Zhao and D. G. Truhlar, Acc. Chem. Res., 2008, 41, 157–167.
48. P. C. Hariharan and J. A. Pople, Theor. Chim. Acta, 1973, 28, 213–222.
49. A. W. Ehlers, M. Böhme, S. Dapprich, A. Gobbi, A. Höllwarth, V. Jonas, K. F. Köhler, R. Stegmann, A. Veldkamp and G. Frenking, Chem. Phys. Lett., 1993, 208, 111–114.
50. A. Höllwarth, M. Böhme, S. Dapprich, A. W. Ehlers, A. Gobbi, V. Jonas, K. F. Köhler, R. Stegmenn, A. Veldkamp, G. Frenking and R. Stegmann, Chem. Phys. Lett., 1993, 208, 237–240.
51. K. Fukui, J. Phys. Chem., 1970, 74, 4161–4163.
52. K. Fukui, Acc. Chem. Res., 1981, 14, 363–368.
53. S. Grimme, S. Ehrlich and L. Goerigk, J. Comput. Chem., 2011, 32, 1456–1465.

---

Footnotes

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c5nr05690j

‡ There are other classes of silver nanoclusters that are formed without the use of borohydride salts.²²

§ The crystallographic information file for [Ag₃(μ₃-H)(μ₃-BH₄)L^Ph₃]BF₄ has been deposited at the Cambridge Crystallographic Data Centre and assigned the code: CCDC 1419573.

¶ While there is no significant change in the DFT calculated gas-phase structure of [Ag₃(H)(BH₄)L^Ph₃]⁺ compared to the X-ray structure, detailed discussions of the calculations are limited to the L^Me systems, as the Ag–P bond energies in [Ag₃(H)(BH₄)L^Ph₃]⁺ are overestimated using the M06 functional. The same is also true when the B3LYP-D3BJ functional is applied. For example, the CID results show that loss of the phosphine ligand from [Ag₃(H)(BH₄)L^Ph₃]⁺ is preferred to the BH₃ loss while both the M06 and B3LYP-D3BJ functionals predict a reverse trend. This inconsistency is likely to arise from the overestimation of the magnitude of dispersive interactions in [Ag₃(H)(BH₄)L^Ph₃]⁺ due to the presence of the aromatic rings.

|| We have attempted to locate transition states for the BH₃ loss reactions. In all cases we have been unable to locate a transition state and attempts led to the loss of BH₃.

---