Coordination-induced weakening of ammonia, water, and hydrazine X–H bonds in a molybdenum complex

See allHide authors and affiliations

Science  11 Nov 2016:
Vol. 354, Issue 6313, pp. 730-733
DOI: 10.1126/science.aag0246

Coordinated scission of N–H or O–H bonds

Ammonia and water both have well-explored acid-base chemistry at room temperature, revolving around proton exchange. In contrast, radical chemistry involving H-atom exchange is comparatively rare in these molecules in the absence of a high-energy stimulus. Bezdek et al. now show that coordination of ammonia or water to a molybdenum complex substantially weakens the N–H or O–H bonds, so much so that heating to 60°C liberates hydrogen (see the Perspective by Hoover). Theoretical and electrochemical analyses reveal the underpinnings of the bond-weakening phenomenon.

Science, this issue p. 730; see also p. 707


Although scores of transition metal complexes incorporating ammonia or water ligands have been characterized over the past century, little is known about how coordination influences the strength of the nitrogen-hydrogen and oxygen-hydrogen bonds. Here we report the synthesis of a molybdenum ammonia complex supported by terpyridine and phosphine ligands that lowers the nitrogen-hydrogen bond dissociation free energy from 99.5 (gas phase) to an experimentally measured value of 45.8 kilocalories per mole (agreeing closely with a value of 45.1 kilocalories per mole calculated by density functional theory). This bond weakening enables spontaneous dihydrogen evolution upon gentle heating, as well as the hydrogenation of styrene. Analogous molybdenum complexes promote dihydrogen evolution from coordinated water and hydrazine. Electrochemical and theoretical studies elucidate the contributions of metal redox potential and ammonia acidity to this effect.

Ammonia and water are ubiquitous small molecules with strong bonds between hydrogen and the central atom (1). For over a century, transition metal–ammine (NH3) and –aquo (H2O) compounds have defined bonding paradigms in chemistry (2), found application in cancer therapy (3), and promoted important fundamental chemical reactions such as electron transfer that rely on the inertness of the N–H or O–H bonds in the supporting ligands (Fig. 1) (4).

Fig. 1 Comparison of classical coordination compounds of ammonia and nonclassical compounds that enable bond weakening by coordination, which in turn enables hydrogen evolution.

Classical compounds are from (24), with N–H BDFEs computed using DFT. Nonclassical coordination is defined as a ligand containing X–H bonds that are thermodynamically unstable to H2 loss. M, metal.

Common strategies for activation of ammonia and water include oxidative addition to a transition metal center (57), deprotonation by transition metal hydrides (8), reaction with bimetallic compounds (9), cooperative chemistry between a transition metal and a supporting ligand (1012), and element-hydrogen (X–H) bond cleavage through reaction with main group compounds (1316). Using most of these strategies, activation of the strong X–H bond is not typically coupled to H–H bond formation. One exception is the observation of H2 elimination following oxidative addition of ammonia to a tantalum(III) complex (17).

An alternative and less commonly explored strategy is homolytic cleavage of the X–H bond. Because of the high gas-phase bond dissociation free energies (BDFEs; 99.5 and 111.0 kcal/mol for NH3 and H2O, respectively) (1), interaction with a transition metal or other appropriate reagent is necessary to induce bond weakening. As shown in Fig. 1, most classical transition metal compounds with ammine (aquo) ligands have N–H (O–H) bond strengths that are only slightly perturbed from the gas-phase value. Because experimental data are lacking, we used density functional theory (DFT) to compute N–H BDFEs.

Coordination-induced bond weakening, whereby interaction of a ligand results in a considerable lowering of the X–H BDFE, has recently been identified or implicated in rare instances (1823) and has been applied by Knowles’s group (24) and others (2527) in reactions of organic molecules involving N–H and O–H bonds, respectively. Cuerva’s group (26, 27) and ours (28) have demonstrated the effectiveness of bis(cyclopentadienyl) titanium(III) complexes in weakening the O–H bonds of water and the N–H bonds of ammonia. However, this strategy has not yet been shown to be capable of weakening the N–H or O–H bond sufficiently to provide the thermodynamic driving force for H2 evolution (BDFEX–H < gas-phase free energy of H-atom formation = 48.6 kcal/mol; Fig. 1) (1). In (η5-C5Me4SiMe3)2TiCl(NH3), for example, the N–H BDFE was calculated to be 61 kcal/mol, too high to spontaneously form H2 (28). Here we describe a terpyridine bis(phosphine) molybdenum complex that, by virtue of its coordination environment and redox properties, enables H2 evolution from coordinated ammonia, water, and hydrazine. This effect is termed “nonclassical” coordination.

Chloride abstraction from (PhTpy)(PPh2Me)2Mo(Cl) (1-Cl; PhTpy, 4′-Ph-2,2′,6′,2′′-terpyridine; Ph, phenyl) (29) with NaBArF24 {ArF24, [C6H3-3,5-(CF3)2]4} in benzene solution at ambient temperature in the presence of 1 equivalent of ammonia resulted in isolation of a yellow-green crystalline solid identified as [(PhTpy)(PPh2Me)2Mo(NH3)](BArF24) [(1-NH3)+] in 77% yield (Fig. 2A). The formally Mo(I) ammonia complex has a spin S = 1/2 ground state [effective magnetic moment (μeff) = 1.7 Bohr magneton (μB), 23°C by magnetic susceptibility balance] and exhibits an isotropic electron paramagnetic resonance (EPR) signal (isotropic g value = 1.988) in fluid benzene solution (23°C) with hyperfine coupling to two 100%-abundant I = 1/2 phosphorus nuclei {isotropic hyperfine coupling constant [Aiso(31P)] = 33 MHz}, as well as to the two naturally occurring spin-active molybdenum nuclei, 95Mo and 97Mo [Aiso(95/97Mo) = 80 MHz; I = 5/2; 15.92% 95Mo and 9.55% 97Mo]. The solid-state structure was determined by x-ray diffraction and confirms formation of an octahedral molybdenum complex with the ammonia ligand trans to the central pyridine of the terpyridine chelate with a Mo–NH3 bond distance of 2.236(3) Å. The solid-state infrared spectrum (KBr) of (1-NH3)+ exhibits three isotopically sensitive low-energy vibrations at 2919, 2899, and 2847 cm−1, consistent with a coordinated ammine ligand. These vibrations are likely perturbed by hydrogen bonding, given that the solid-state structure of (1-NH3)+ revealed a close [2.395(3) Å] H-F interaction between the ammine hydrogens and the CF3 group of the (BArF24) counterion (Fig. 2B).

Fig. 2 N–H bond weakening in Mo-coordinated ammonia.

(A) Synthesis of (1-NH3)+ and hydrogen-atom abstraction by using substituted phenoxyl radical and chromium reagents. (B) Solid-state structure of (1-NH3)+ illustrated using 30% probability ellipsoids. Hydrogen atoms, except those connected to N4, have been omitted for clarity. rt, room temperature; h, hours; equiv., equivalent.

Experiments were conducted to establish an upper bound for the N–H bond strength in (1-NH3)+. Addition of 1 equivalent of 2,4,6-tri-tert-butylphenoxyl radical (tBu3C6H2O•) resulted in rapid H-atom abstraction from the ammonia ligand to quantitatively yield the olive-green diamagnetic molybdenum(II) amide complex, [(PhTpy)(PPh2Me)2Mo(NH2)](BArF24) [(1-NH2)+], setting the BDFE of the N–H bond as <77 kcal/mol (30) (Fig. 2A). The 1H nuclear magnetic resonance (NMR) spectrum of (1-NH2)+ in benzene-d6 shows the number of resonances expected for a C2v symmetric compound with a diagnostic downfield triplet centered at 10.02 parts per million (ppm) (J-coupling constant 3JP–H = 16.7 Hz) assigned to the terminal amide hydrogens. The 15N NMR spectrum of (1-15NH2)+ (prepared from 15NH3) in benzene-d6 features a triplet centered at 235.5 ppm (1JN–H = 68.5 Hz), as well as a doublet in the 31P NMR spectrum at 15.58 ppm (2JP–N = 4.1 Hz). The infrared spectrum (KBr) of (1-NH2)+ contains two peaks assignable to a -NH2 fragment at 3354 and 3287 cm−1 that shift to 2512 and 2426 cm−1, respectively, in the deuterated isotopolog, (1-ND2)+. The solid-state structure was determined by x-ray diffraction, establishing an octahedral molybdenum complex in analogy with (1-NH3)+. The Mo–NH2 bond distance of 1.994(3) Å is considerably contracted compared with the Mo–NH3 distance of 2.236(3) Å in (1-NH3)+, consistent with formation of an anionic ligand.

To more accurately define the N–H bond strength in (1-NH3)+, the ammonia complex was treated with [(η5-C5Me5)Cr(CO)3]2 (Fig. 2A). Immediate and quantitative formation of (1-NH2)+ and (η5-C5Me5)Cr(CO)3(H) established an N–H bond strength of <62 kcal/mol (31). DFT calculations for (1-NH3)+ gave a computed N–H bond strength of 45.1 kcal/mol, consistent with H-atom abstraction experiments.

The very low N–H bond strength of coordinated ammonia in (1-NH3)+ suggested that spontaneous H2 formation should be possible. Gently heating a benzene-d6 solution of (1-NH3)+ to 60°C for 6 hours resulted in clean and quantitative formation of (1-NH2)+ with concomitant H2 evolution, as confirmed by Toepler pump experiments (92% yield of H2; Fig. 3A). Carrying out the reaction in the presence of 1.5 equivalents of styrene (per Mo) furnished ethylbenzene in 25% yield, providing additional evidence for coordination-induced bond weakening and the application of ammonia as a hydrogen storage medium for the reduction of organic molecules. Dihydrogen evolution accompanied ethylbenzene formation and is likely the source of the relatively low yield. A maximum yield of 50% is possible based on reaction stoichiometry (2:1 molybdenum:styrene). The reduction of the olefin likely occurs through successive H-atom transfer steps, given that experiments in the presence of excess phosphine produced no inhibition. Performing the same procedure with (1-ND3)+ yielded 1′,2′-d2-ethylbenzene, 1′-d1-ethylbenzene, 2′-d1-ethylbenzene, and ethylbenzene in a 1:1:1:1 ratio, as detected by quantitative 13C-NMR spectroscopy (fig. S1), with a combined yield of 25%. Overall, a 1:1 ratio of deuterium incorporation into the methylene and methyl positions of styrene was observed by 2H NMR spectroscopy. These results are consistent with reversible H-atom transfer between (1-ND3)+ and styrene and provide direct chemical evidence for the DFT-computed N–H BDFE of 45.1 kcal/mol, given that the C-H bond strength adjacent to a benzylic radical is known to be 45 kcal/mol (32, 33).

Fig. 3 Spontaneous hydrogen evolution from molybdenum complexes.

(A) Mo-ammonia complex, (B) Mo-hydrazine complex, and (C) Mo-aquo complex.

Studies were also conducted to explore the pathway of H2 evolution from (1-NH3)+. A crossover experiment was performed whereby a 1:1 mixture of (1-NH3)+ and (1-ND3)+ was heated to 60°C for 6 hours and the evolved gas was collected and analyzed. Both H2 and D2 were detected by 1H and 2H NMR spectroscopy, respectively, together with a nonstatistical amount of HD gas (H2:HD was 5:1 by 1H NMR integration). In addition, a statistical mixture of the isotopologs (1-NH2)+, (1-NHD)+, and (1-ND2)+ was detected at the completion of the reaction by 31P NMR spectroscopy (fig. S2). The origin of the HD gas and the statistical mixture of molybdenum amides was investigated by a series of mixing experiments. Exchange among the products was evaluated by monitoring a 1:1 mixture of (1-NH2)+ and (1-ND2)+ in benzene-d6 by 31P NMR spectroscopy. A statistical distribution of the isotopologs (1-NH2)+, (1-NHD)+, and (1-ND2)+ was observed immediately after mixing at 23°C, demonstrating H-D exchange between products proceeding faster than hydrogen evolution. Isotopic exchange between the molybdenum ammine compound and the molybdenum amide product was also evaluated. Analysis of the 31P NMR spectrum of a benzene-d6 solution of a 1:1 mixture of (1-ND3)+ and (1-NH2)+ revealed immediate isotopic exchange into the molybdenum amide, demonstrating exchange between reactants and products. It is likely to be this pathway that accounts for the nonstatistical amount of HD gas observed, because dihydrogen evolution is slower than the exchange between reactants and products. Because the starting ammine complexes are paramagnetic, N-H and N-D exchange between (1-NH3)+ and (1-ND3)+ were not directly measured. On the basis of these results, a pathway involving bimetallic H2 evolution is disfavored; intermolecular chemistry with a large N-H or N-D kinetic isotope effect, however, cannot be rigorously eliminated.

The spontaneous liberation of H2 from coordinated ammonia prompted investigation of other small molecules to determine the generality of hydrogen evolution from coordination-induced bond weakening. DFT calculations for the putative molybdenum aquo [(1-OH2)+] and hydrazine {[1-(ĸ2-N2H4)]+} complexes established exceedingly weak O–H and N–H bonds of 33.7 and 34.6 kcal/mol, respectively. Accordingly, treatment of a benzene solution containing 1-Cl with 1 equivalent of NaBArF24 in the presence of hydrazine or water furnished the diamagnetic yellow-brown solids [(PhTpy)(PPh2Me)2Mo(ĸ2-NHNH2)](BArF24) {[1-(ĸ2-NHNH2)]+} and [(PhTpy)(PPh2Me)2Mo(OH)](BArF24) [(1-OH)+] in 68 and 61% yields, respectively. In each case, H2 gas evolution was confirmed by Toepler pump experiments (73 and 58% yield of H2, respectively; Fig. 3, B and C). Although the benzene-d6 1H, 13C, and 31P NMR spectra of (1-OH)+ are consistent with a C2v symmetric compound, the side-on bound hydrazido ligand in [1-(ĸ2-NHNH2)]+ lowers the overall symmetry of the molecule to Cs. The syntheses of the deuterated isotopologs [1-(ĸ2-NDND2)]+ and (1-OD)+ were carried out using N2D4 and D2O, respectively, and enabled assignment of the N–H/D and O–H/D peaks in the solid-state (KBr) infrared spectra. The N–H peaks of [1-(ĸ2-NHNH2)]+ appear at 3315, 3240, and 3186 cm−1 {2547, 2503, and 2434 cm−1 in [1-(ĸ2-NDND2)]+}; the sharp OH peak of (1-OH)+ appears at 3558 cm−1 [2627 cm−1 in (1-OD)+]. These results support the assignment of the hydrazido and hydroxo ligands in [1-(ĸ2-NHNH2)]+ and (1-OH)+, respectively. The intermediacies of the putative hydrazine and water complexes [1-(ĸ2-N2H4)]+ and (1-OH2)+ were probed by performing the syntheses of [1-(ĸ2-NHNH2)]+ and (1-OH)+ in the presence of 1 equivalent of tBu3C6H2O•. 1H NMR spectra revealed immediate and quantitative formation of [1-(ĸ2-NHNH2)]+ and (1-OH)+, along with the stoichiometric generation of tBu3C6H2OH. No H2 was evolved in these reactions, implying the intermediacy of metal-bound hydrazine and aquo complexes preceding H2 evolution in the absence of a radical abstracting reagent.

The solid-state structures of [1-(ĸ2-NHNH2)]+ and (1-OH)+ were determined by x-ray diffraction (figs. S14 and S15). With (1-OH)+, an octahedral coordination geometry around molybdenum was observed with trans phosphine ligands, similar to (1-NH3)+ and (1-NH2)+. The coordination environment of seven-coordinate [1-(ĸ2-NHNH2)]+ is best described as pentagonal bipyramidal, where the tridentate terpyridine chelate and the bidentate hydrazido ligand occupy the vertices of an equatorial pentagon, with apical phosphine ligands completing the coordination sphere of molybdenum. The solid-state structure of [1-(ĸ2-NHNH2)]+ revealed a rare example of a side-on bound ĸ2-hydrazido fragment, consistent with the Cs molecular symmetry observed by 1H and 13C NMR spectroscopy.

Having demonstrated the generality of the method, we sought to further elucidate the properties of the Mo complex underlying the bond-weakening phenomenon. The Bordwell equation (34, 35) expresses the BDFE of a given N–H bond in terms of the oxidation potential of the metal complex and acidity of the N–H bond upon oxidation (Fig. 4A). As such, the electrochemical behavior of (1-NH3)+ was examined, and the one-electron oxidized compound (1-NH32)+ was synthesized with the goal of experimentally determining the pKa (where Ka is the acid dissociation constant), thereby obtaining an experimental value for the N–H BDFE in (1-NH3)+ (Fig. 4A).

Fig. 4 Contributions of oxidation potential and pKa to the bond-weakening process.

(A) Thermochemical expression for N–H BDFEs. CG, solvent-specific H+/H• standard reduction potential. (B) Oxidation potentials and pKa measurements for a series of molybdenum complexes of varying oxidation states. Bolded values are experimentally measured, whereas italicized values are DFT-computed. *Values obtained from gas-phase DFT calculations. †Oxidation potentials reported relative to Fc/Fc+ in THF solution with 0.1 M [nBu4N][PF6] as the electrolyte. ‡Calculated value in THF solution from the Bordwell equation for the reaction [Mon+1-NH3](m+1)+ → [Mon+1-NH2]m+, assuming the corresponding DFT-calculated N–H BDFE value, the experimentally determined E°ox value, and a CG constant of 66 kcal/mol in THF solvent (40). The superscripts m and n are the integer overall charge of the complex and the integer oxidation state of the metal center, respectively.

The cyclic voltammogram in tetrahydrofuran (THF) solvent (fig. S16) of the formally Mo(I) complex (1-NH3)+ exhibits two reversible anodic waves, one at –1.09 V (relative to ferrocene/ferrocenium), which is assigned to one-electron oxidation to the dicationic complex [(PhTpy)(PPh2Me)2Mo(NH3)][BArF24]2 [(1-NH32)+]. The second wave, centered at –0.57 V, is assigned as the second oxidation event to furnish the two-electron oxidized compound (1-NH33)+. The cyclic voltammogram also exhibits a quasi-reversible cathodic wave at –2.56 V, likely corresponding to the reduction to (1-NH30).

With experimental oxidation potentials in hand for (1-NH32)+, (1-NH3)+, and (1-NH30), the isolation of the one-electron oxidized product (1-NH32)+ was targeted to determine the N–H pKa and hence measure BDFE for (1-NH3)+. Addition of [H(OEt2)2](BArF24) (Et, ethyl) to (1-NH2)+ yielded the NMR- and EPR-silent S = 1 product (μeff = 2.7 μB, 23°C by magnetic susceptibility balance) (1-NH32)+ in 78% yield. The N–H pKa of (1-NH32)+ was determined by measurement of the equilibrium concentration ratio with its conjugate base (1-NH2)+ in the presence of 2-methoxy pyridine (pKa = 2.6 in THF) (36). Using this method, the average of three equilibration experiments established the pKa of (1-NH32)+ as 3.6 in THF solution (table S1 and fig. S17). This value, coupled with the experimentally determined oxidation potential (ox) of (1-NH3)+ (–1.09 V), allowed for the determination of an experimental N–H BDFE of 45.8 kcal/mol for (1-NH3)+ in THF solution by using the Bordwell equation. This value is in excellent agreement with the DFT-computed gas-phase value of 45.1 kcal/mol. Both the acidity, likely arising from the overall cationic charge on the complex, and the reducing nature of the metal, a result of the formal Mo(I) oxidation state, contribute to the observed N–H bond weakening in (1-NH3)+.

To elucidate the individual contributions of metal reduction potential and N–H pKa on the phenomenon of bond weakening, computational (DFT) studies were carried out to determine the N–H BDFEs in the series of complexes (1-NH32)+, (1-NH3)+, and (1-NH30). The successive one-electron reduction from (1-NH32)+ to (1-NH3)+ to (1-NH30) results in a concomitant lowering of the N–H BDFE (Fig. 4B). Using the DFT-computed N–H BDFEs and experimentally determined oxidation potentials, application of the Bordwell equation allows evaluation of the pKa values for the members of the redox series for which an experimental determination is not possible. The pKa values shown in Fig. 4B correspond to the oxidized form of the compound presented, as shown in the square scheme in Fig. 4A.

Although calculated differences of ~11 kcal/mol in the N–H BDFE accompany each one-electron redox step, the dominant term in the Bordwell equation varies depending on the charge of the molybdenum complex. The most reduced member of the series, (1-NH30), has the least acidic pKa contribution to the strength of its N–H bond (pKa = 20.1) and demonstrates that the large negative potential (E°ox = –2.56) is the dominant component of the bond-weakening phenomenon. Attempts to synthesize this compound by chemical reduction of (1-NH3)+ have been unsuccessful, yielding a complex mixture of products upon treatment with potassium graphite. Although the origin of the decomposition is not definitive, isolation of a compound with a N–H BDFE of 34.5 kcal/mol would likely be challenging.

The corresponding molybdenum cation, (1-NH3)+, has a pKa contribution to the N–H bond strength that is 17 units lower than the value calculated for (1-NH30), consistent with known trends in metal-aquo complexes, where increased charge density and electropositivity increase O–H acidity (37). Further decrease in the pKa term of the N–H bond strength of (1-NH32)+ is minimal; only a slight decrease in pKa to 2.2 was observed. The leveling in acidity suggests that although both (1-NH32)+ and (1-NH3)+ have weak N–H bonds, the spontaneous H2 evolution in the latter is largely driven by the reduction potential of the formally Mo(I) complex.

Elucidation and delineation of the origin of coordination-induced bond weakening provide design principles for applications ranging from catalysis to bioinorganic chemistry to alternative energy. In cases where the function of ligands containing N–H and O–H is to stabilize metal complexes, often with various oxidation states throughout a catalytic cycle, combinations of oxidation potentials and pKa values that promote this effect should be suppressed. Alternatively, in applications such as small-molecule activation or the use of NH3 as a hydrogen storage medium (38, 39), these properties can be rationally tuned to favor weakening of X–H bonds. The kinetic stabilization of (1-NH3)+, a molecule with an unusually weak N–H bond, should inspire synthetic efforts to prepare additional examples of such coordination compounds.

Supplementary Materials

Materials and Methods

Figs. S1 to S17

Table S1

Crystallographic Data

References (4164)

References and Notes

Acknowledgments: Financial support was provided by the U.S. Department of Energy, Office of Science, Basic Energy Science (grant DE-SC0006498). M.J.B. thanks the Natural Sciences and Engineering Research Council of Canada for a predoctoral fellowship (PGS-D). Crystallographic parameters are available free of charge from the Cambridge Crystallographic Data Centre for the structures of (1-NH2)+ (CCDC 1477445), (1-OH)+ (CCDC 1477446), [1-(ĸ2-NHNH2)]+ (CCDC 1477447), and (1-NH3)+ (CCDC 1477448). Additional data supporting the conclusions are in the supplementary materials. The authors thank R. Knowles (Princeton University) for insightful discussions.
View Abstract


Navigate This Article