Catalytic N₂-to-NH₃ (or -N₂H₄) Conversion by Well-Defined Molecular Coordination Complexes

Abstract

Nitrogen fixation, the six-electron/six-proton reduction of N₂, to give NH₃, is one of the most challenging and important chemical transformations. Notwithstanding the barriers associated with this reaction, significant progress has been made in developing molecular complexes that reduce N₂ into its bioavailable form, NH₃. This progress is driven by the dual aims of better understanding biological nitrogenases and improving upon industrial nitrogen fixation. In this review, we highlight both mechanistic understanding of nitrogen fixation that has been developed, as well as advances in yields, efficiencies, and rates that make molecular alternatives to nitrogen fixation increasingly appealing. We begin with a historical discussion of N₂ functionalization chemistry that traverses a timeline of events leading up to the discovery of the first bona fide molecular catalyst system and follow with a comprehensive overview of d-block compounds that have been targeted as catalysts up to and including 2019. We end with a summary of lessons learned from this significant research effort and last offer a discussion of key remaining challenges in the field.

Graphical Abstract

1. INTRODUCTION TO THE FIELD OF N₂ REDUCTION CATALYSIS

Catalytic nitrogen fixation is an essential chemical transformation in both biology and industry as it represents the primary means by which nitrogen (N₂) from the air becomes bioavailable. This review focuses on the development and study of synthetic molecular catalysts that mediate the catalytic conversion of nitrogen to ammonia (N₂-to-NH₃, often abbreviated as the nitrogen reduction reaction or N₂RR) in the presence of acid and reductant under moderate temperatures and pressures.

1.1. Motivation for New Ammonia Synthesis Catalyst Technologies

Conventional ammonia synthesis (i.e., Haber-Bosch) is among the most significant technological advances of the 20th century and has been critical to sustained global population growth.¹ However, with operating pressures of 150–250 bar and temperatures of 400–500 °C, it has high cost demands for infrastructure leading to centralization of the manufacturing process and thus requires a global distribution system.² This large-scale distribution and the necessary temperature and pressure to form NH₃ from N₂ and H₂ over a solid-state Fe catalyst necessitates significant fossil fuel input with related high carbon dioxide (CO₂) emissions. While estimates vary, approximately 1–2% of annual global energy consumption is accounted for by conventional ammonia synthesis, with some 4% of global methane (CH₄) and 60% of global hydrogen going into its production. The generation of the needed hydrogen via steam-reforming (providing for ~72% global ammonia production) accounts for nearly 0.5 gigatons of CO₂ released annually.³ In addition, other environmental consequences from fertilizer use are severe, including surface and groundwater pollution from runoff, eutrophication of freshwater systems, and massive killing of aquatic organisms in coastal regions that comprise so-called dead zones due to depleted oxygen.^4,5 These consequences could be mitigated if ammonia synthesis were electrified,⁶ and hence produced on scale, and on demand, in a distributed fashion. On demand distributed production of fertilizer could offset use (and hence production) of fertilizer that is sourced conventionally and could also be generated locally and at a rate that increases net absorption by crops (versus runoff), offering a possible environmental benefit to conventional practices in fertilizer acquisition and use.⁷

Although ammonia is a commodity chemical produced primarily for fertilizer (on a massive scale, ~150 million metric tons annually), it has also been identified as a promising alternative fuel. It is highly storable, easily liquefied, and has an energy density approaching half that of gasoline far exceeding that of compressed hydrogen. It can also be used as a fuel within an internal combustion engine (ICE) or via a solid-oxide fuel cell (SOFC). Moreover, globally substantial ammonia transport infrastructure and related safe-handling protocols already exist.⁸ For these reasons, ammonia synthesis is an extremely attractive target for electrification, especially via renewable energy technologies, requiring major advances in catalyst development. N₂RR electrification would enable surplus energy in the grid, at times of excess supply, to be converted to fertilizer and/or to a storable and transportable fuel, particularly desirable in areas where wind and solar resources are vast. The eventual realization of an “ammonia fuel economy” that can contribute to diverse future energy strategies, along with technologies for on-site and on-demand ammonia fertilizer generation, will require breakthrough research discoveries in catalysis.⁹

1.2. Inspiration for Organometallic and Inorganic Chemistry

In contrast to the forcing conditions required in the Haber–Bosch process, certain microorganisms can fix N₂ under ambient conditions, using extensive hydrolysis of adenosine triphosphate (ATP) to power the delivery of H⁺ and e⁻ equivalents to N₂. These enzymes may hold important clues as to how H⁺/e⁻ currency, potentially derived from photosynthetic water splitting, could be efficiently delivered to N₂ via an appropriate catalyst. Housed within any given nitrogen-fixing organism are conserved sets of proteins—the nitrogenase enzymes—that bind and convert N₂-to-NH₃. Nitrogenases appear to require iron as an essential transition metal and typically contain molybdenum (FeMo-nitrogenase, most common form), with either vanadium (VFe-nitrogenase) or Fe (FeFe-nitrogenase) being assembled (and functionally active) in the absence of Mo.^10,11 FeMo-nitrogenase was the first to be discovered and has been by far the most widely studied.^12,13 In addition to various exogenous cofactors required for its function, this enzyme consists of an Fe-protein that delivers reducing equivalents and a MoFe-protein.¹⁰ The latter contains two structurally unique clusters. The first is the P-cluster (Fe₈S₇), which serves as an electron relay to the Fe-protein. The second is the M-cluster (MoFe₇S₉C-homocitrate), an inorganic FeMo cofactor (FeMoco, Figure 1) that mediates the catalytic bond-breaking and making steps.¹⁴

Figure 1.

X-ray structure of the FeMo-cofactor active site in the MoFe protein (PDB: 1M1N); blue = N, red = O, yellow = S, brown = Fe, gray = C, purple = Mo.¹³

Inorganic chemists have long puzzled (and engaged in spirited debate!) over how this cofactor operates^15–19 and, in collaboration with biochemists, microbiologists, structural biologists, and spectroscopists, have pursued molecular model systems as a means of constraining hypotheses regarding viable inorganic mechanisms for catalytic N₂RR. This has proven to be a remarkably rich research area. The list of talented chemists who have significantly contributed to what is now an enormous body of knowledge reads like a “Who’s Who of Inorganic Chemistry.” Fortunately, much of this work has been reviewed previously.^15,20–29

The goal of this review is to be comprehensive only with respect to the comparatively recent body of literature pertaining to catalytic N₂RR that is mediated by nominally well-defined synthetic complexes, in the presence of H⁺/e⁻ sources. We acknowledge at the outset that much of the fascinating literature preceding such catalyst discoveries will not be detailed, except in cases where introducing background is needed to set the stage for the catalyst discoveries that will be covered. One such case is the early and pioneering work of Joseph Chatt and his co-workers.^15,30 Given how central the research and ideas he and his team espoused were to the development of the broader field of N₂RR catalysis, we felt it appropriate to briefly summarize some of this historical context here. The interested reader should consult a number of excellent reviews for a deeper dive into this and other early literature.^15,31–34

2. DISCUSSION OF THE CHATT CYCLE

Coordination chemists began to think seriously about alternative catalyst technologies to the Haber–Bosch ammonia synthesis in the early 1960s (Figure 2). Several factors had set the stage; key among these was that, in 1963, the British Agricultural Council, led by Secretary Sir Gordon Cox, himself a coordination chemist by training, appointed the British inorganic chemist Joseph Chatt to oversee a multidisciplinary research unit dedicated toward understanding the mechanism of biological nitrogen fixation.³⁵ It was by this time presumed that Fe and Mo were present within the active site of the single nitrogenase that was then known.³⁶ Refreshingly, the Secretary of Agriculture must have intuited that, at its heart, the mechanism of nitrogen fixation was an inorganic chemistry problem. This took imagination and foresight, as it was not until two years later (1965) that Allen and Senoff reported their landmark (and fortuitous) discovery that N₂ could coordinate as a ligand to a transition metal, via the isolation and characterization of [(NH₃)₅Ru(N₂)]²⁺ (Figure 2).^37,38 Hence, from the outset, the systematic study of inorganic and organometallic complexes for N₂RR was established as an area of bioinorganic model chemistry.

Figure 2.

Timeline (1930 to present) of selected advances in nitrogen fixation catalysis by synthetically well-defined complexes.^{18,37,48–56}

Fundamental studies toward understanding the binding, activation, and conversion of N₂ to protonated intermediates and/or products at well-defined transition metal centers were deemed essential to helping formulate and constrain hypotheses concerning the biological process. The Unit of Nitrogen Fixation, first located at the University of London but soon thereafter relocated to Sussex, was highly innovative in its approach and comprised not just chemists but also microbiologists, biochemists, and geneticists, working collectively.³⁹

Prior to his appointment leading the Unit of Nitrogen Fixation, Chatt (with Duncanson), contemporaneously with Dewar,⁴⁰ played a key role in defining and generalizing the bonding of olefins to transition metals, describing the interaction by both sigma donation (from the olefin π-electrons to the metal) and π-bonding (from the metal to the olefin π-antibonding orbital).⁴¹ This type of bonding is today often referred to as the Dewar–Chatt–Duncanson model.⁴² This proposal paralleled what had already been developed for metal carbonyls⁴³ and anticipated the type of bonding interactions in metal dinitrogen complexes.

Following Allen and Senoff’s isolation of a terminally bound Ru(N₂) complex,³⁸ it was not much of a stretch to postulate that synthetic metal complexes might be able to bind and catalyze N₂RR under suitable conditions. A surge of relevant research activity thus followed, not just in the UK but around the world, and within just a decade profound progress was made.^15,33,34 For instance, Shilov and his collaborators in the former Soviet Union reported the exciting discovery that certain transition metal mixtures containing, for example, molybdenum precursors mixed with Mg(OH)₂, could in fact catalyze N₂ reduction to N₂H₄ and NH₃ in alcohol/water mixtures in the presence of reducing agents such as sodium amalgam.^20,44–47 But it was the work of Chatt and his team at Sussex, via their careful, rigorous preparation and study of M(N₂) and M(N_xH_y) complexes, that most directly laid the groundwork for the well-defined synthetic N₂RR catalysts that have emerged thus far.^15,32

2.1. The Chatt Cycle As It Is Commonly Known

While M(N₂) complexes for a range of TMs (e.g., Ru, Re, Os, Co) were known by the end of the 1960s, much of the early biomimetic work focused on Mo(N₂) (and related W(N₂)) complexes.^32–34,57 Since the work of Bortels in 1930, it was long-held dogma that molybdenum was essential to nitrogen fixation⁴⁸ and this fact, combined with the practical reality that so many M(N_xH_y) complexes featuring Mo (and W) proved accessible, helped focus model chemistry research in this area. Relatedly, biochemical experiments had suggested that nitrogenase activity was highly sensitive to alteration at or near the molybdenum site, such as at the homocitrate ligand.¹⁰ Chatt’s team established N₂ binding and activation at Mo (and W), and showed that the coordinated N₂ ligand could be protonated to release NH₃ (and hydrazine) in variable yield (as high as 90% for W(PMe₂Ph)₄(N₂)₂ with H₂SO₄, assuming one N₂ equiv is released as gas)⁵⁸ depending on a range of factors (Figure 3). What is more, the group was able to identify a number of M(N_xH_y) complexes, including M(N),⁵⁹ M(NNH₂),^60–64 M(NHNH₂),⁶⁵ M(NNH),^60,66,67 M(NNH₃)^68,69 species (M = Mo or W), that likewise underwent protonation to release NH₃ in similar yields.

Figure 3.

Protonation of M(cis-N₂)₂(PMe₂Ph)₄ and M(N_xH_y); M = Mo or W.¹⁵

These findings led Chatt to propose a simple scenario in which triple functionalization at N_β leads to the release of NH₃ and a nitride intermediate. This scenario is now known as the Chatt cycle, and a simplified version is depicted below (Figure 4).³⁰ Worth noting is that while examples of the types of M(N_xH_y) species invoked in the Chatt cycle could be generated, most typically these species were not characterized in the same formal state of oxidation as invoked in the catalytic scheme.²⁶ It would not be until Schrock’s discoveries some 30 years later^50,70 that well-characterized examples of the proposed catalytic Mo(N_xH_y) intermediates, in an oxidation/protonation state matching the envisaged catalytic cycle, were discovered (Figure 11). Critically, this triamidoamine-ligated Mo system would also prove to be the first catalyst for nitrogen fixation, thereby cementing the importance of the Chatt cycle, sometimes formulated as the Schrock or “distal” cycle in the contemporary literature, in the field. Equally important, Chatt and his team underscored via their work the value of preparing and carefully studying well-defined M(N_xH_y) model complexes to test the validity of possible pathways for nitrogen fixation.

Figure 4.

Scheme demonstrating the distal (or Chatt) cycle for nitrogen fixation. In the modern literature, this cycle is also sometimes referred to as the Schrock cycle. Ligands on molybdenum are omitted for clarity.³⁰

Figure 11.

(Top) Protonation reactions of (HIPTN₃N)Mo(L) (L = CO or N₂). For the CO analogue, pulse EPR (ENDOR) data indicated protonation at an amido arm; for N₂, the site of protonation remains unknown;¹⁰² (middle) protonation/reduction reactions of the L = N₂ and N analogues gave rise to [Mo]-NNH and [Mo]-NH₃⁺; (bottom) reduction of [[Mo]-NH₃]⁺ under N₂ gives [Mo]-N₂ in ~10% yield; [Mo] = (HIPTN₃N)Mo.⁹⁹

2.2. Alternative Mechanistic Scenarios for NH₃ Formation

While it is the aforementioned cycle for which Chatt is most often credited, he and his co-workers also envisaged alternative scenarios that could account for NH₃ (and N₂H₄) production via intermediates not included on the so-called Chatt pathway. For example, protonation of an M(NHNH₂) or M(NH₂NH₂) intermediate can release hydrazine or ammonia (Figure 5).⁶⁵ It is perhaps ironic that, although a M(NHNH₂) species is not commonly referred to as a “Chatt-type” intermediate, it was his group that first prepared such a species (for W) and considered its direct relevance to NH₃ release.⁶¹

Figure 5.

An alternative to the Chatt pathway that can account for productive NH₃ formation.

Although Schrock’s best known contribution to N₂-fixation chemistry, the (HIPTN₃N)Mo system (HIPTN₃N = 3,5-(2,4,6-i-Pr3C₆H₂)₂C₆H₃NCH₂CH₂]₃N³⁻), appears to proceed via a distal reaction pathway, during the mid-1980s, Schrock and co-workers also reported important results regarding alternative mechanisms for NH₃ formation in Cp*-Group VI (Cp* = C₅Me₅⁻) compounds. During their study of the general reactivity of (Cp*)M (M = Mo, W) complexes with N_xH_y ligands, they found that protonation reactions of the N₂ complexes gave (primarily) NH₃, but studies suggested that the pathway for NH₃ release was unlikely to involve the terminal nitride, M(N), intermediate.⁷¹ Instead, protonation of the hydrazido(2−) complex, (Cp*)W(Me)₃(NNH₂), occurred at N_α to give [(Cp*)W(Me)₃(η²-NHNH₂)]OTf —a reversible process in the presence of base. This is an interesting transformation, as one can ask whether the kinetic site of protonation of (Cp*)W(Me)₃(NNH₂) is at N_α or N_β, and whether protonation occurs prior to, or following, isomerization of the nitrogenous ligand from an η¹- to η²-form (Figure 6). A subsequent study by Norton provided compelling evidence for initial protonation at N_β to generate [(Cp*)W(Me)₃(NNH₃)]⁺, followed by isomerization to give [(Cp*)W(Me)₃(η²-NHNH₂)]⁺.⁷² While the W(NNH₃)⁺ species in this system was not derived from N₂, the aforementioned transformations underscore that a so-called Chatt (or distal) intermediate can isomerize to an alternating intermediate W(η²-NHNH₂).⁷²

Figure 6.

Protonation of the hydrazido(2−) complex, (Cp*)W(Me)₃(NNH₂), can initially occur at N_α or N_β, but ultimately gives [(Cp*)W(Me)₃(η²-NHNH₂)]⁺—a reversible process in the presence of NEt₃. It remains unclear whether protonation occurs at the η¹- or η²-form of the hydrazido(2−) species.⁷²

With the (Cp*)M platform, the Schrock group could also synthesize hydrazine intermediates [(Cp*)M(Me)₃(NH₂NH₂)]⁺. Reduction of these intermediates with sodium amalgam led to the formation of NH₃ and (Cp*)M(NH). Indeed, reacting this platform with excess acid and reductant in the presence of hydrazine led to the catalytic formation of NH₃ in yields up to 95% (note that the maximum yield for hydrazine disproportionation is 67%).^73,74 This work, thus, conclusively demonstrated that NH₃ formation does not necessarily proceed via the Chatt cycle, but rather that formally “alternating” intermediates M(NHNH₂) and M(NH₂NH₂) can also lead to NH₃ formation. The potential of such intermediates to contribute to hydrazine formation will be discussed later in the context of catalysis with Fe complexes.

Much of the work from Chatt was pursued in the absence of exogenous reductant and thus necessarily focused on zerovalent Mo and W species that could themselves provide the electrons required for nitrogen fixation. Nonetheless, Chatt, as early as 1969, noted that “the reduction of complexed (di)nitrogen requires specific reagents or conditions, and these may be just as elusive as the first nitrogen complexes.”75

Schrock’s early work is not only important for its initial elucidation of hybrid mechanisms for NH₃ formation, but also for its preliminary efforts to perform protonolysis experiments in the presence of external reductants. Perhaps inspired by the success of Shilov and his collaborators in achieving catalytic N₂ reduction with transition metal mixtures and amalgam reductants,^44,76,77 Schrock and co-workers performed protonolysis experiments using [LutH]Cl (LutH = lutidinium) of μ-bridged Group VI dinitrogen complexes, such as [(Cp*)Mo(Me)₃]₂(μ-N₂), in the presence of ZnHg amalgam.⁷⁸ Although these experiments only gave 0.62–0.72 equiv of NH₃, this work provided a viable approach for future studies that would yield catalysis.

In fact, the study of nitrogen fixation has revealed that the choice of acid and reductant is critical to catalyst performance. In many ways, the reagents are as much a part of the nitrogen fixation story as the catalysts themselves, with certain systems demonstrating dramatically different results and, by implication, different mechanisms when distinct reagent cocktails are supplied.^{18,51,79–81} Combinations of compatible acid and reductant have moved the field forward and allowed for the discovery of new catalysts. Indeed, in recent years attention has focused how the choice of reagents also controls the overall thermochemistry of the N₂RR catalysis, providing a framework toward improvement of the thermal efficiency of the catalysis. In the next section, we will detail the two successive breakthroughs made by Schrock and co-workers. The first was the application of the (HIPTN₃N)Mo ligand to support a series of Chatt-cycle intermediates in the relevant oxidation states⁷⁰ and the second, the discovery that these intermediates could be turned over by application of an appropriate reductant and acid,⁵⁰ a combination that went on to find broad application in the field.^51,82–84

Another pathway for dinitrogen fixation concerns the formal scission of dinitrogen to two metal-bound nitrides followed by the transfer of proton and electron equivalents (via PCET or otherwise) to release ammonia (vide infra in Section 4.3 for an example from Nishibayashi and co-workers). This transformation has been shown feasible for a number of transition metal complexes, typically following formation of a μ₂-bimetallic N₂ adduct that undergoes N–N bond cleavage upon addition of a reducing agent.^85–90 A classic example is shown in Figure 7 and concerns a Mo tris(amido) complex, (N(R)Ar)₃Mo.^91,92 This species binds N₂ at room temperature reversibly to form (N(R)Ar)₃Mo(N₂), with the μ₂-N₂ bridged derivative, [(N-(R)Ar)₃Mo]₂(μ-N₂), being formed at −35 °C over a period of days. Warming this complex results in formation of two terminally bonded nitride complexes, (N(R)Ar)₃Mo(N).^92,93

Figure 7.

N₂ splitting by a Mo tris(anilide) complex; an alternative pathway to the formation of ammonia from μ-N₂ complexes.

3. SCHROCK’S TRIS(AMIDO)AMINE Mo: THE FIRST WELL-DEFINED CATALYST SYSTEM

3.1. Azatrane Mo Complexes Relevant to N₂ Fixation

With an eye toward nitrogen fixation, in 1994, Kol et al. showed that triamidoamine ligands [(RNCH₂CH₂)₃N]³⁻ where R = C₆F₅ could be used to gain access to molybdenum complexes of potential relevance to N₂RR (Figure 8).⁹⁴ A number of relevant molybdenum complexes were accessed, including a terminal nitride (((C₆F₅NCH₂CH₂)₃N)Mo(N)), methylimido cation ([((C₆F₅NCH₂CH₂)₃N)Mo(NCH₃)]⁺), as well as both bridged (((C₆F₅NCH₂CH₂)₃N)Mo]₂(μ-N₂) and terminal ([((C₆F₅NCH₂CH₂)₃N)Mo(N₂)]⁻ dinitrogen complexes. The latter of these was shown to be nucleophilic at N_β, undergoing facile silylation by ⁱPr₃SiCl to give [(C₆F₅NCH₂CH₂)₃N]Mo-N=N−Si(ⁱPr)₃. Silyl-substituted-triamidoamine ligands have also been studied (vide infra for-an example using Ti - Figure 69)—with reaction of MoCl₃(THF)₃ and [(Me₃SiNCH₂CH₂)₃N]Li₃ giving the paramagnetic μ-N₂ complex, [(Me₃SiNCH₂CH₂)₃N]Mo]₂(μ-N₂) in 10% yield. In 1997, O’Donoghue et al. showed that a related TMS anion, {[(TMSNCH₂CH₂)₃N]Mo-N=N}₂(Mg(THF)₂) reacted with FeCl₂ to give {[(TMSNCH₂CH₂)₃N]Mo-N=N}₃(μ₃-Fe).^95,96 These contributions laid the groundwork for further exploring azatrane-Mo complexes as competent for the binding and activation of dinitrogen, as well as the stabilization of potential N₂-fixation intermediates.

Figure 8.

(Top) Reactions of a Mo triamidoamine complex having C₆F₅ anilido groups provided access to complexes of potential relevance to N₂RR;⁹⁴ (bottom) reaction of an analogous Mo complex with TMS anilido groups with FeCl₂ afforded a μ₃-Fe bridging via N₂ to three triamidoamine Mo centers.⁹⁵

Figure 69.

Synthesis and characterization data of triamidoamine-Ti complexes relevant to catalytic nitrogen fixation.⁵⁵

Recognizing that TMS- and C₆F₅- substituted tren [(RNCH₂CH₂)₃N]³⁻ ligands underwent deleterious side-reactivity (due to Si−N bond cleavage and the sensitivity of C₆F₅ groups to nucleophilic reagents), in 2001 Greco and Schrock designed a new route to access aryl-substituted triamidoamine ligands by Buchwald—Hartwig amination.⁹⁷ A small library of ligand candidates was constructed including monoaryl (C₆H₅, 4–F-C₆H₄, 2,4,6-Me₃C₆H₃) and triaryl (3,5-Ph₂C₆H₃, 3,5-(4-^tBuC₆H₃)₂C₆H₃) derivatives that could be readily metalated with Mo (Figure 9). In a subsequent contribution, Greco and Schrock illustrated that a variety of different Mo^IIN₂ complexes could be synthesized in a reductant-dependent fashion. These aryl-substituted triamidoamine Mo complexes (^RPhN₃N, R = H; 4-F; 2,4,6-Me; 3,5-Ph; 3,5-(4-^tBu)Ph) all feature fairly activated N₂ complexes (1740 cm^–1 < ν_NN < 1815 cm⁻¹). Accordingly, these species could be silylated or alkylated to provide (^RPhN₃N)Mo(NNR) (R = SiMe₃ or Me) species. The alkyl diazenidos, (^RPhN₃N)Mo(NNCH₃), can be reacted further with alkylating agents (e.g., CH₃OTf, CH₃OTs) to form cationic hydrazidos [(^RPhN₃N)Mo(NN(CH₃)₂] ⁺ (Figure 9). Similarly, reaction of (^RPhN₃N)Mo(NNTMS) with alkylating reagents provides [(^RPhN₃N)Mo(NN(CH₃)(TMS)]⁺. Efforts to access neutral molybdenum hydrazido complexes were not successful, but reduction of [(^RPhN₃N)Mo(NN(CH₃)₂]⁺ with NaHg amalgam, provided the Mo(Vl) nitride complex, (^RPhN₃N)Mo(N) (and presumably a fixed nitrogen product).⁹⁸

Figure 9.

Synthesis of alkylated Mo diazenido and hydrazido complexes en route to a Mo nitride compound in a triamidoamine Mo complex with Ar = 3,5-Ph₂C₆H₃ or 3,5-(4-^tBuC₆H₄)₂C₆H₃ ligands.⁹⁸

To prevent undesired bimolecular pathways problematic to an N₂ reduction cycle, the highly encumbering 3,5-(2,4,6-ⁱPr₃C₆H₂)₂C₆H₃ (HIPT = HexaIsoPropylTerphenyl) derivative was prepared (Figure 10).⁷⁰ On this ligand, an impressive suite of N₂-derived monometallic molybdenum complexes were prepared (vide infra), including (HIPTN₃N)Mo(N₂), [(HIPTN₃N)Mo(N₂)] ⁻, (HIPTN₃N)Mo(NNH), and [(HIPTN₃N)Mo(NNH₂)]⁺. Additional, species potentially relevant to an N₂-fixation cycle could be accessed with azide as the N atom source, (HIPTN₃N)Mo(N) and (HIPTN₃NMo(NH). Lastly, [(HIPTN₃N)Mo(NH₃)]⁺ was accessed by treatment of (HIPTN₃N)Mo(Cl) with NH₃ in the presence of NaBAr^F₄.

Figure 10.

Synthesis of a series of monometallic molybdenum N_xH_y complexes on 3,5-(2,4,6-ⁱPr₃C₆H₂)2C₆H₃ (HIPT) ligated Mo; [Mo] = (HIPTN₃N)Mo.⁷⁰

With these species in hand, Schrock’s team explored the possibility of catalysis with soluble acid and reducing equivalents.⁹⁹ Treatment of (HIPTN₃N)Mo(N₂) with 1.0 equiv of [LutH]BAr^F₄ and 2.0 equiv of Cp₂Co (E_1/2 = −1.3 V vs Fc/Fc⁺) provided (HIPTN₃N)Mo(NNH) nearly quantitatively (Figure 10).¹⁰⁰ Intriguingly, (HIPTN₃N)Mo(N₂) reacted directly with neither [LutH]BAr^F₄or Cp₂Co, but rapidly reacted with the combination of these reagents. The Schrock group suggested that N-amido protonation could proceed reduction followed by proton transfer. This hypothesis has found indirect support from theoretical work¹⁰¹ and ENDOR studies. ENDOR revealed that treatment of (HIPTN₃N)Mo(CO) with [LutH]⁺ results in protonation of an N-amido arm, but similar behavior was not confirmed for (HIPTN₃N)Mo(N₂) (Figure 11).¹⁰² It has been hypothesized that this reactivity profile suggests a degree of concertedness in the transfer of the proton and electron to form the new N–H bond. In this review, we will refer to such reactions as proton-coupled electron transfer (PCET) reactions.^103,104 Two specific suggestions have been made. Chirik and co-workers suggested that reduction of pyridinium acids by the metallocenes results in pyridinyl radicals with weak N–H bonds that could therefore mediate the conversion of (HIPTN₃N)Mo(N₂) to (HIPTN₃N)Mo(NNH) (and potentially other N–H bond forming steps).^105,106 Although such a mechanism could be operative under catalytic conditions, it is known that triethylammonium is also efficacious for this reaction in a stoichiometric fashion. The Peters group has suggested with the support of computation that instead these acids protonate the metallocene species on the Cp(*) ring resulting in species with weak C–H bonds that can then mediate conversion of the N₂ species to the NNH species via a PCET reaction (vide infra).⁸⁰

If a larger quantity of [LutH]BAr^F₄ (7.0 equiv) and Cp₂Co (8.2 equiv) was reacted with (HIPTN₃N)Mo(N₂), then [(HIPTN₃N)Mo(NH₃)]⁺ was formed in 60% yield. Furthermore, the reaction of (HIPTN₃N)Mo(N) with [LutH]BAr^F₄ (3.5 equiv) and Cp₂Co (4.2 equiv) gave [(HIPTN₃N)Mo(NH₃)]⁺ in 80% yield, which could be liberated with Bu₄NCl and NEt₃ to afford 0.88(2) equiv of NH₃. Closing the catalytic cycle involved the reduction of [(HIPTN₃N)Mo- NH₃]+ to (HIPTN₃N)Mo(NH₃), the release of NH₃, and the binding of N₂. However, [(HIPTN₃N)Mo(NH₃)]+ is quite reducing, so the reaction with Cp₂Co only converted 10% of the starting material to the desired (HIPTN₃N)Mo(N₂) complex (Figure 11).⁹⁹

3.2. Description of the System and Key Findings

Building on this work, Yandulov and Schrock published the first example of N₂ reduction catalysis mediated by a well-defined, synthetic catalyst in 2003 (Figure 12).⁵⁰ In particular, they observed that using the more reducing Cp*₂Cr (E_1/2 = −1.47 V vs Fc^+/0) allowed for complete reduction of [(hIPTN₃N)Mo(NH₃)]⁺. Likewise, slow addition of reductant (6 h, syringe pump) and solvent selection (heptane) were needed to attenuate background H₂ formation. With these conditions, it was found that (HIPTN₃N)Mo(N₂), (HIPTN₃N)Mo(NNH), (HIPTN₃N)Mo(N), or [(HIPTN₃N)Mo(NH₃)]⁺, when combined with Cp*₂Cr (36 equiv) and [LutH]BAr^F₄ (48 equiv), were all competent (pre)catalysts for N₂ reduction, generating between 7.56 and 8.06 equiv of NH₃ (Figure 12). At the time, this selectivity (~66%) was second only to nitrogenase (~75%).⁵⁰

Figure 12.

Catalytic N₂RR cycle depicting (HIPTN₃N)Mo intermediates and their efficacy as (pre)catalysts for N₂RR using [LutH]BAr^F₄ (48 equiv) and Cp*₂Cr (36 equiv); [Mo] = (HIPTN₃N)Mo.^50,107 Characterized compounds are shown in purple and unobserved complexes in gray.

A series of experiments were performed in order to interrogate the catalytic reaction. (HIPTN₃N)Mo(NNH) was reacted with 1 equiv of [LutH]BAr^F₄ (Figure 13).¹⁰⁰ This was found to give a 44:56 equilibrium mixture of (HIPTN₃N)Mo(NNH) and [(HIPTN₃N)Mo(NNH₂)]⁺ (K_eq = 1.6). Similarly, protonation of (HIPTN₃N)Mo(N) using [LutH]BAr^F₄ gave a 3:1 mixture of (HIPTN₃N)Mo(N) and [(HIPTN₃N)Mo(NH)]+.¹⁰⁰ These results suggest that weaker acids would not be suitable for the catalysis.

Figure 13.

(Top) [Mo]-NNH and [LutH]⁺ are in equilibrium with [Mo]-NNH₂⁺ and Lut (K_eq = 1.6); (bottom) protonation of [Mo]N gives a mixture of the starting material and [Mo]NH⁺, which upon reduction, decayed to give [Mo]N and [Mo]-NH₂; [Mo] = (HIPTN₃N)Mo.⁹⁹

Consistent with these weak predicted bond strengths, reduction of [(HIPTN₃N)Mo(NNH₂)]⁺ by Cp₂Co did not yield the expected (HIPTN₃N)Mo(NNH₂), but rather the formation of NH₃, (HIPTN₃N)Mo(NNH), and (HIPTN₃N)Mo(N). Similarly, reduction of the cationic imide with Cp*₂Cr initially gave a species tentatively assigned as the neutral imide, (HIPTN₃N)Mo(NH), which then decayed to (HIPTN₃N)Mo(N) and (HIPTN₃N)Mo(NH₂) in solution. These results are consistent with calculations by Chirik and co-workers that suggest that these species (the neutral hydrazido and neutral imide) have the weakest N–H bonds on this platform.¹⁰⁶ In contrast, the alkylated congeners, (HIPTN₃N)Mo(NNEt₂) and (HIPTN₃N)Mo(NEt), are stable, but efforts to achieve the catalytic reduction of N₂ to NEt₃ with [Et₃O]BAr^F₄ and Cp*₂Cr were unsuccessful (Figure 14).¹⁰⁸ While many factors may of course be at play that impede catalysis, one possible conclusion is that the net exchange of H atoms between nitrogen fixation intermediates may be relevant to achieving overall catalysis in N₂RR. This is clearly speculative. Observations regarding the productive net transfer of H atoms between nitrogen fixation intermediate and the enhanced stability of alkylated congeners have been made on the silyl-anchored triphosphine system from Peters and co-workers (vide infra).¹⁰⁹

Figure 14.

Use of an alkylating agent provided access to neutral [Mo]-NNEt₂ and [Mo]-NEt; these species serve as structural analogues of the highly reactive R = H variants; [Mo] = (HIPTN₃N)Mo.^99,108

At the end of catalysis, [(HIPTN₃N)Mo(NH₃)]⁺ was the only Mo-containing species present, suggesting that the slow steps of the catalysis involve the reduction of this species and then exchange with N₂. Consistent with this observation, the reduction potential of this species (−1.51 V vs Fc^+/0 in THF) is slightly lower than that of the employed reductant, Cp*₂Cr (–1.47 V vs Fc^+/0 in THF).¹¹⁰ Exchange of N₂ for NH₃ in (HIPTN₃N)Mo(NH₃) was determined to undergo an associative mechanism on the basis of variable pressure N₂ experiments (k_obs = 2.4 fold faster at a 15 psi N₂ overpressure). Indeed, performing N₂RR at a pressure of 30 psi (instead of 15 psi) provides a small, but measurable, increase in ammonia yield (71% vs 63%) over a 6 h addition and 55% versus 45% over a 3 h addition time.¹¹⁰ Under catalytic conditions, dissociated NH₃ is anticipated to be trapped via protonation, which should drive the reaction toward the N₂ complex. Consistent with this, the rate of the substitution reaction was significantly enhanced by inclusion of triphenylborane as an NH₃ trap.¹¹⁰ Computational studies by Reiher and co-workers have found support for the involvement of a six-coordinate (HIPTN₃N)Mo(NH₃)(N₂) intermediate (as opposed to a five-coordinate species with a dissociated ligand arm) in the exchange process.^101,111 Accessing states where there is rapid exchange of NH₃ for N₂ remains a general challenge for N₂RR catalysts and accordingly is an ideal area for future catalyst design.

Schrock and co-workers performed X-ray diffraction analysis on (HIPTN₃N)Mo(NNH), [(HIPTN₃N)Mo(NNH₂)]⁺, [(HIPTN₃N)Mo(NH)]⁺, and [(HIPTN₃N)Mo(NH₃).¹⁰⁰ The accumulated structural data show that the Mo–N_amine distance varies as a function of the Mo axial ligand; as the N–N bond is weakened and cleaved on going from [HIPTN₃N]Mo(N₂) to [HIPTN₃N]Mo(N), a lengthening in the Mo–N_amine by ~0.19 Å is observed. An even more dramatic lengthening of the metal-axial ligand distance has been observed in the borane-anchored trisphosphine Fe system from Peters and co-workers.^112–114 This suggests that flexibility of this interaction is perhaps critical for allowing catalysts to stabilize both π-accepting intermediates and π-donating intermediates and/or low-valent and high-valent states.

Given the formation of H₂, presumably from Mo(H) species, during (HIPTN₃N)Mo catalyzed nitrogen fixation, the reactivity of Mo(H) and Mo(H₂) complexes was also of interest.^99,107 Intriguingly, (HIPTN₃N)Mo(H) could be prepared from the diazenido derivative, (HIPTN₃N) Mo(NNH), via an effective “β-H elimination” (k₁ = 2.2 × 10⁻⁶ s⁻¹) (Figure 15). This is a very unusual transformation that potentially warrants further study. The use of the hydride species as a (pre)catalyst produced 7.65(3) equiv of NH₃, suggesting that it could be readily returned to a catalytically on-path intermediate. Schrock and co-workers demonstrated that reaction of (HIPTN₃N)Mo(H) with [LutH]BAr^F₄ led to the formation of [(HIPTN₃N)Mo(Lut)]⁺, which can presumably be reduced and bind N₂ to rejoin the catalytic cycle. The performance of the (HIPTN₃N)Mo(H) species as a precatalyst contrasts with that of other M(H) (M = Fe, Co, Os) species. In those cases, the M(H) species have uniformly performed more poorly as precatalysts than the related N₂ complex (vide infra).^{53,54,115,116}

Figure 15.

β-H elimination from the diazenido derivative, [Mo]–NNH, generated [Mo]–H; [Mo] = (HIPTN₃N)Mo.⁹⁹

The H₂ complex, (HIPTN₃N)Mo(H₂), could be formed via the reaction of (HIPTN₃N)Mo(N₂) with H₂ over 2–3 days (Figure 16).¹⁰⁷ Exchange of N₂ for H₂ occurred at a rate of k = 3.4 × 10⁻⁶ s⁻¹ and was independent of the H₂ pressure, indicating a rate-limiting dissociation of N₂ prior to H₂ coordination, consistent with the proposed dissociative exchange of N₂ and ¹⁵N₂.⁹⁹ Experiments for the reverse reaction (i.e., formation of (HIPTN₃N)Mo(N₂) from (HIPTN₃N)Mo(H₂) found the rate to be 1.0 × 10⁻⁶ s⁻¹ (t_1/2 = 8 days)). Use of (HIPTN₃N)Mo(H₂) as a (pre)catalyst provided ammonia in 52% yield relative to 60–65% (from (HIPTN₃N)Mo(N₂)). Experiments carried out with dihydrogen injected (~32 equiv and 64 equiv of H₂ with respect to Mo) using (HIPTN₃N)Mo(N) as the catalyst, showed that only 1 equiv of NH₃ was formed, presumably from the nitride ligand.¹⁰⁷ Cumulatively, these data suggest that H₂ poisons the N₂ fixation catalysis. This presents a significant challenge to achieving higher turnover numbers with this catalyst. Indeed, the ability to manage H₂ formation is ultimately going to be an issue of concern in any catalytic nitrogen fixation system that is not perfectly selective for NH₃.

Figure 16.

Generation of [Mo]–H₂ and a plausible decay pathway involving intramolecular deprotonation to a terminal (HIPTN₃NH)Mo(H) species; [Mo] = (HIPTN₃N)Mo.¹⁰⁷

3.3. Use of EPR for the Study of (HIPTN₃N)Mo Complexes

Electron paramagnetic resonance (EPR) and electron nuclear double resonance (ENDOR) have enabled the characterization of a number of (HIPTN₃N)Mo complexes. These have included (HIPTN₃N)Mo(NH₃),¹¹⁷ (HIPTN₃N)Mo(N₂),^{102,117–119} (hIPTN₃N)Mo(CO),^102,117 the amido-protonated analogue, [(HIPTN₃N(H))Mo(CO)]⁺¹⁰² (vide supra) as well as the product of H₂ with (HIPTN₃N)Mo(N₂), the Mo(III) anion, [(HIPTN₃N)Mo(H)]⁻.¹¹⁹ More recently, Schrock, Hoffman, and Neese studied the electronic and geometric structures of [(HIPTN₃N)Mo(N)]^–-and (HIPTN₃N)Mo(NH) by irradiation of their oxidized congeners with γ-rays at 77 K.¹²⁰ Collectively, these studies serve to provide useful spectroscopic models to which more complicated [Mo]-containing clusters can be compared; several of these open-shell compounds are not characterizable by other means.

3.4. Evidence in Support of the Chatt/Schrock Cycle from Theory

Several studies that bear on the theoretical mechanism of N₂ fixation by the (HIPTN₃N)Mo system have been published (Figure 17),^{101,121–127} and in 2008, theoretical and experimental data for this system were compared by Schrock.¹²⁸ A brief summary is provided here. Early theoretical studies tended to simplify the catalyst structure to reduce computational expense. For example, Morokuma¹²⁹ and Cao¹²⁶ performed studies on truncated models that were not referenced to the catalytically relevant acid and reductant sources. In 2005, Tuczek and co-workers reinvestigated the complete cycle, modeling the HIPT portion of the ligand as hydrogen but using the relevant acid, [LutH]⁺ and reductant, Cp*₂Cr sources as energetic references.¹²¹ Independently, Reiher^101,122 and Neese & Tuczek¹²⁴ later performed a series of DFT calculations taking the entire HIPT ligand moiety into account, requiring models of ~280 atoms. In broad terms, these data corroborated the cycle reported above (Figure 12). The following key points are recapitulated with numerical values taken from Nees and Tuczek (B3LYP/def2-TZVP)124—this study also summarized and compared its findings with those from other studies.

Figure 17.

Reproduced with permission from ref 124. Copyright 2015 American Chemical Society. Red: Gibbs free enthalpy ΔG° scheme of the Schrock cycle calculated with the B3LYP functional and the def2-TZVP basis set including solvent correction. Gray: calculations by Studt et al.¹²¹

• Importantly, all on path intermediates were predicted to be neutral or positively charged; the formation of negatively charged intermediates are predicted to be highly unlikely.

• Protonation of (HIPTN₃N)Mo(N₂) was calculated to be the most endergonic reaction step of the reaction (ΔG = +16.6 kcal mol⁻¹) with protonation of an N-amido ligand arm being more favorable; this species has been characterized spectroscopically for (HIPTN₃N)Mo(CO).¹⁰² Subsequent reduction, followed by protonation, was proposed to yield [(HIPTN₃NH)Mo(NNH)]⁺. Later computational work from the Chirik group and Peters group have proposed PCET pathways to form the diazenido complex directly from (HIPTN₃NH)Mo(N₂) as a plausible first step in the catalysis.^80,105,106

• The most exergonic step of the reaction was N–N bond cleavage upon reduction of [(HIPTN₃N)Mo(NNH₃)]⁺ (ΔG = −68.4 kcal·mol⁻¹).¹²⁴

• Consistent with experiment, reduction of the ammine cation, [(HIPTN₃N)Mo(NH₃)]+ to the neutral species by Cp*₂Cr was essentially thermoneutral (ΔG_calc = +0.5 kcal·mol⁻¹, ΔG_exp = +0.9 kcal·mol⁻¹).¹²⁴

• The dissociation of NH₃ from (HIPTN₃N)Mo(NH₃) is endergonic (ΔG = +13.0 kcal·mol⁻¹). Subsequent N₂ coordination was exergonic (ΔG = −12.1 kcal·mol⁻¹), rendering the net exchange slightly uphill (ΔG = +0.9 kcal·mol⁻¹).¹²⁴ This calculation agrees with the observed K_eq value of 0.1 for this reaction, which corresponds to ΔG = +1.1 kcal·mol⁻¹.¹²⁴ An associative mechanism was also considered for this process, with calculation of the putative, six-coordinate (HIPTN₃N)Mo(NH₃)(N₂) species being +8.8 kcal·mol⁻¹ higher in energy than (HIPTN₃N)Mo(NH₃) and free N₂.¹²⁴

• A necessary shortcoming of these (and most) DFT studies is that they cannot consider all of the species/interactions that might be relevant in the catalytic mixture. Hence, their primary value lies in their ability to shed light on specific questions given certain assumptions. So as to explore secondary interactions that might be relevant to the Schrock system, Reiher et al. studied “one-pot” models (where acid (acid = [LutH]⁺) and base (base = a Mo compound) were initially placed approximately 8 Å from one another). This study predicted hydrogen bonded complexes between (HIPTN₃N)Mo(N) or (HIPTN₃N)Mo(N₂) and [LutH]⁺, indicating that the acid source is sufficiently small to approach the basic part of the molecule through voids made by the three HIPT ligands.¹²² A related hydrogen-bonded complex between [LutH]⁺ and a Mo(N) species in a different N₂-fixation relevant complex (vide infra) was crystallographically characterized by Schrock and co-workers in 2019 (Figure 18).¹³⁰ That complex underwent net PCET on treatment with reductant. These collective data support the possibility that a multisite PCET reaction could be relevant to diazenido formation in the (HIPTN₃N)Mo system. In this proposal, a hydrogen bonded [(HIPTN₃N)Mo(N₂)—LutH]⁺ complex would transfer the proton only upon electron transfer from Cp*₂Cr (Figure 17).

Figure 18.

(Left) A depiction of a hydrogen bonded complex between (HIPTN₃N)Mo(N₂) and [LutH]⁺ identified in a “one-pot” calculation;¹²² (right) a depiction of a crystallographically characterized hydrogen bonded complex between (Ar₂N₃)Mo(N) (Ar = 2,6-diisopropylphenyl) and [LutH]⁺.¹³⁰

3.5. Second Generation Schrock Systems

Despite extensive efforts invested in improving the TON of the original (HIPTN₃N)Mo system, reactivity (e.g., protonolysis) of the Mo-amido linkages and ligand exchange appears to plague it and related systems. These efforts are summarized here. By 2004, three new ligand variants appeared: two of which gauged sterics (^tBu and CH₃ in place of ⁱPr) and one that gauged electronics (Br in place of H at the para position of the middle terphenyl ring) (Figure 19). For each ligand, the Mo(N) complex was prepared and its efficacy for catalytic N₂RR was tested. The t-butyl analogue produced only 1.06 equiv of NH₃ using Cp*₂Cr (36 equiv) and [LutH]BAr^F₄ (48 equiv), while the methyl analogue gave only 1.49 equiv of NH₃. The p-Br analogue provided 6.4–7.0 equiv of NH₃, slightly less than the standard HIPT-based Mo catalyst system.¹³¹

Figure 19.

A survey of second generation complexes tested for N₂RR by Schrock and co-workers. Reported yields were found using the optimized conditions for the original catalytic result: [LutH]⁺ (48 equiv), Cp*₂Cr (36 equiv), heptane, room temperature.^131–134

In another iteration of ligand design, asymmetrical tren ligands were prepared in which one of the three HIPT groups was replaced by (1) a 3,5-disubstitutedbenzene-3,5-Me₂C₆H₃, 3,5-(CF₃)₂C₆H₃, or 3,5-(MeO)₂C₆H₃; (2) a 3,5-disubstituted-pyridine-3,5-Me₂NC₅H₃ or −3,5-Ph₂NC₅H₃ or (3) a 2,4,6-trisubstitutedbenzene-2,4,6-Me₃C₆H₂ or −2,4,6-ⁱPr₃C₆H₂ (Figure 19).¹³² For each ligand, the Mo(N) complex was prepared and probed for competency in catalyzing N₂RR. For these candidates, 0.2–2.0 equiv of NH₃ was found.

In 2010, Chin and Schrock outlined the synthesis of a system comprising pyrrolide rather than anilide donor groups, which were pursued in hopes of attenuating undesired aminolysis in the presence of acid (a limitation of the HIPT system) (Figure 19).¹³³ Diamido(2-mesityl)pyrrolyl complexes of the type ((ArN)₂Pyr)Mo(X) (Ar = C₆F₅, 3,5-Me₂C₆H₃, or 3,5-^tBu₂C₆H₃, X = Cl or NMe₂), having two amide and one pyrrole group were prepared. The degree of N₂ activation was attenuated in [((Ar^tBuN₂)Pyr)]Mo(N₂)]⁻ (ν_NN = 2012 cm⁻¹ vs 1990 cm⁻¹ [(HIPTN₃n)Mo(N₂)]⁻). Perhaps relatedly, ((Ar^tBuN₂)Pyr)Mo(N) proved to be a much poorer catalyst for N₂RR (1.02 ± 0.12 equiv of NH₃) with no evidence for N₂ uptake during catalysis.

In a subsequent study, Reithofer and Schrock developed a HIPT derivative incorporating pendant nitrogens (2,5-diiso-propylpyrrolyl, DPPN₃N) in the secondary coordination sphere (Figure 19).¹³⁴ The hypothesis was that if (HIPTN₃N)Mo undergoes decomposition by competitive protonation at the Mo–N_amide bond, then providing Brønsted basic groups in the secondary coordination would prevent that and allow a greater population of catalyst to remain on-cycle. In this vein, (DPPN₃N)Mo(Cl), (DPPN₃N)Mo(N₂), and [(DPPN₃N)Mo(N₂)]⁻ precursors were prepared similarly to those for the HIPT system (vide supra). For (DPPN₃N)Mo(N₂), υ_nn = 1993 cm⁻¹, indicating a similar degree ofactivation when compared to (HIPTN₃N)Mo(N₂). Of the Schrock Mo(N₂) compounds discussed in this review, this is the only one that undergoes associative N₂ substitution. In a standard experiment, treatment of (DPPN₃N)Mo(¹⁴N₂) with ¹⁵N₂ showed a first order dependence on [¹⁵N₂]. This conclusion is rationalized by virtue of having a smaller five-membered pyrrole-based ligand platform that allows for associative substitution, cf. a bulkier six-membered 3,5-substituted terphenyl-group. Other species, including (DPPN₃N)Mo(NNH), [(DPPN₃N)Mo(NNH₂)]⁺, (DPPN₃N)Mo(N), [(DPPN₃N)Mo(NH)]⁺, and [(DPPN₃N)Mo(NH₃)]⁺, were also prepared. Efforts toward N₂ reduction using (DPPN₃N)Mo(N) as the (pre)catalyst produced 2.53 ± 0.35 equiv of NH₃ making it formally catalytic.

In 2018, Wickramasinghe and Schrock again endeavored to maintain catalyst efficiency by limiting Mo–N_amide protonation or dissociation (Figure 20). In this effort, they synthesized a calix[6]azacryptand ligand (CAC(OMe)₃) that tethered the three HIPT triamidoamine ligand arms together.¹³⁵ The X-ray structure of the nitride complex, (CAC(OMe)₃)Mo(N), confirmed that the three-amido arms were locked in the desired configuration. Unfortunately, this system was not active for catalytic N₂ fixation, producing only 1.28 equiv of NH₃ using [Ph₂NH₂]OTf (96 equiv) and KC₈ (84 equiv). The yield under the standard conditions used for (HIPTN₃N)Mo were not reported.

Figure 20.

A calix[6]azacryptand ligand designed by the Schrock group aimed at protecting the reactive Mo active site for use in N₂RR.¹³⁵

3.6. Schrock Work on W, Cr, and V

The Schrock group explored the (HIPTN₃N)M platforms (M = W, Cr, and V) for catalytic N₂RR, but without success.^136–138 In the case of tungsten, the preparative chemistry was found to be very similar to that for Mo (vide supra), with many candidate intermediates of N₂RR being readily synthesized (Figure 21). However, attempts to demonstrate catalytic nitrogen fixation resulted in only stoichiometric yields of NH₃ (1.3–1.5 equiv). The authors attributed the inability of the (HIPTN₃N)W system to serve as an N₂RR catalyst to a problematic redox couple for the ammonia adduct: [(HIPTN₃N)W(NH₃)]^+/0.¹³⁶ It is much harder to reduce this species than its Mo analogue (−2.06 V vs Fc^+/0 compared to −1.63 V vs Fc^+/0).¹⁰⁰ Furthermore, chemical reduction does not lead to the neutral ammonia or N₂ adduct hinting at undesired side reactivity.⁸⁴

Figure 21.

Synthesis of candidate N_xH_y intermediates on (HIPTN₃N) W relevant to N₂ fixation. Note: this system produces only stoichiometric yields of NH₃ (1.3−1.5 equiv).¹³⁶

In the case of Cr, (HIPTN₃N) complexes of both Cr^III and Cr^II proved unable to bind N₂. This contrasts directly with the behavior of the Mo and W complexes, which readily form dinitrogen complexes at these redox states, but is consistent with the lack of known Cr^II or Cr^III dinitrogen complexes. The authors speculate that the relatively slow binding of CO to the HIPTN₃N-complexes of Cr^II and Cr^III compared to that observed for Mo and W could be due to the high-spin nature of the Cr centers. The nitride complex, (HIPTN₃N)Cr(N), was prepared via thermolysis of the corresponding azide complex (Figure 22). Reductive protonation of this species under their standard catalytic conditions led to the formation of 0.8 equiv of NH₃. This result suggests that downstream reduction steps are viable for (HIPTN₃N)Cr, but presumably the lack of N₂ uptake short-circuits viable catalysis.¹³⁷

Figure 22.

Synthesis of (HIPTN₃N)Cr(N) and (HIPTN₃N)V(NH) and their reductive protonation to give 0.8 and 0.76 equiv of NH₃, respectively. These results demonstrate that catalytic N₂RR is not accessible with these species, but downstream functionalization reactions are productive.^137,138

To compare V, the complex (HIPTN₃N)V(THF) was prepared. Dinitrogen binding was only observed upon reduction to the anionic state, [(HIPTN₃N)V(N₂)]⁻, which required the use of KC₈ (as opposed to Cp*₂Cr). Attempts to use the latter as a precatalyst for nitrogen fixation under the standard Schrock-type conditions were unsuccessful (0.2 equiv of NH₃). The (HIPTN₃N)V(NH) complex could be prepared either via reaction of the V(THF) species with methylaziridine (Figure 22) or via reaction of the V(NH₃) with base and oxidant. Its use as a precatalyst led to the formation of 0.78 equiv of NH₃ akin to results obtained with (HIPTN₃N)Cr(N) discussed above. This again points to the challenge of N₂-uptake, at least using the original catalytic conditions with Cp*₂Cr as the reductant.¹³⁸ A computational study of this system posited several reasons to explain its inefficiency. In particular, the study pointed to the lower basicity of the various N₂-fixation intermediates which might lead to increased preference for ligand protonation, the lesser thermodynamic driving force for N–N bond cleavage, and the persistent challenge of exchanging NH₃ for N₂.¹³⁹

3.7. Third Generation Schrock Systems (Tridentate Platforms)

Recently, Schrock and co-workers have exploited a geometrically distinct, diamido(pyridine) pincer framework for Mo catalyzed N₂ fixation (Figure 22).¹⁴⁰ In 2017, they reported that the diamido(pyridine) ligand, [2,6-(ArNCH₂)₂NC₅H₃]²⁻ (Ar₂N₃, Ar = 2,6-diisopropylphenyl), can be used to prepare (Ar₂N₃)Mo(N)(O^tBu). Reaction of this nitride with 108 equiv of [Ph₂NH₂]OTf and 54 equiv of Cp*₂Co (conditions discovered in the context of Fe-catalyzed N₂RR, (vide infra) gave 7.9 equiv of NH₃. By contrast, [Ar₂N₃]Mo(N)(OC₆F₅) was not competent for catalysis (1.3 equiv of NH₃ under same conditions). The authors suggested that the ^tBuO fragment in the more active species might be more readily converted to a Mo–OH or Mo(O) unit, which is the catalytically active species.

A series of stoichiometric experiments were performed to shed light on the mechanism by which (Ar₂N₃)Mo(N)(O^tBu) facilitates N₂RR (Figure 24). Reaction of this species with the relevant acid, [H₂NPh₂]OTf, was found to protonate a ligand Mo-N amido linkage. By contrast, use of the weaker/bulkier (and catalytically ineffective) acid, [LutH]OTf, was found to provide an adduct species via a hydrogen-bond to the nitride (vide supra).¹³⁰ Treatment of either of these species with Cp₂Co resulted in the reformation of (Ar₂N₃)Mo(N)(O^tBu), presumably via the loss of a half-equivalent of H₂, though this was not confirmed. The authors speculated that this reaction pointed to the possibility of PCET reactivity during N₂RR, but this has not been confirmed.

Figure 24.

Reaction of (Ar₂N₃)Mo(N)(O^tBu) with [H²NPh2]⁺ results in protonation of a Mo-N_anilido arm whose product is shown to the right and use of [LutH]⁺ results in hydrogen-bond formation with the nitride shown on the left. Following reduction, each of these complexes releases H2 to provide the starting material shown in the center.¹³⁰

4. NISHIBAYASHI’S LOW-VALENT Mo-PHOSPHINE SYSTEMS

4.1. Description of the Original System and Key Findings

In 2010, Nishibayashi and co-workers introduced a dinuclear molybdenum system as an N₂ fixation (pre)catalyst (Figure 25).⁵¹ Specifically, a PNP-based (PNP = 2,6-bis(di-tert-butyl-phosphinomethyl)pyridine) complex (PNP)MoCl₃ was prepared and reduced to the corresponding dinitrogen Mo(0) complex, [(PNP)Mo(N₂)₂]₂(μ-N₂), through reaction with NaHg amalgam (terminal υ_NN= 1936 cm⁻¹). Use of adamantly-or isopropyl-substitutents in lieu of the tert-butyl groups did not provide the analogous Mo(0) complexes, highlighting the critical role of P-substituent. In a trial experiment, N₂ reduction was assayed using [LutH]OTf (96 equiv) and Cp₂Co (72 equiv) as proton and electron sources, in toluene at room temperature. Notably, a solution of the reductant was added slowly over the course of the reaction by syringe pump to prevent rapid recombination of electron and proton sources. Altogether, this mixture produced 11.8 equiv of NH₃ (6.9 equiv of per Mo); using larger quantities of acid and reductant (266 equiv and 288 equiv, respectively) produced 23.2 equiv of NH₃ (11.6 equiv per Mo center). Similar activity was found using Cp*₂Cr, though use of less reducing Cp₂Cr produced no NH₃. The optimal acid was found to be [LutH]OTf, with Cl⁻ and BAr^F₄⁻ counteranions providing significantly less NH₃ (0.7 and 2.7 equiv, respectively). Other Mo complexes, such as cis-Mo(Me₂Ph)₄(N₂) and trans-Mo(N₂)(dppe)₂ (dppe = 1,2-bis(diphenylphosphino)-ethane), give only stoichiometric quantities of NH₃. To probe catalytic intermediates, [(PNP)Mo(N₂)₂]₂(μ-N₂) was treated with [HOEt₂]BF₄ in pyridine, providing the cationic hydrazido adduct, [(PNP)Mo(N-NH₂)(F)(pyr)]BF₄ (pyr = pyridine, Figure 25). Early on, the authors posited that this reaction proceeds through a mononuclear Mo complex, invoking a series of six stepwise proton and electron transfers, though this hypothesis was later subjected to refinement (vide infra). Notwithstanding, in 2011, this represented the most active catalytic method of accessing NH₃ directly from N₂ using soluble acid and reducing equivalents with a molecular species.

Figure 25.

Nishibayashi’s first generation, [(PNP)Mo(N₂)₂]₂}(μ-N₂) N₂RR catalyst. In these reactions, Cp₂Co was added via syringe pump and the acid counterion had a marked effect on the performance. Protonation of [(PNP)Mo(N₂)₂]₂}(μ-N₂) using [HOEt₂]BF₄ provided a hydrazido(2−) Mo complex.⁵¹

Further studies were performed using phosphine adducts of the aforementioned PNP-pincer complexes.¹⁴¹ In a follow-up contribution, Arashiba et al. showed that reduction of the corresponding (PNP)MoCl₃ precursors with NaHg amalgam in the presence of PMe₂R (R = Me or Ph) provided (PNP)Mo(N₂)₂(PMe₂R) (Figure 26). Initial protonation studies using excess sulfuric acid (H₂SO₄) revealed that these derivatives produced 1.38 and 0.85 equiv of NH₃, respectively. Under catalytic conditions, the PMe₂Ph derivative produced only a small amount of NH₃ (0.2 equiv).

Figure 26.

Protonation studies of six-coordinate (PNP)Mo(N₂)₂(PMe₂R) complexes gave 1.38 (R = Me) and 0.85 (R = Ph) equiv of NH₃.¹⁴¹

Examination of phosphine R group also provided information into the nature of PNP coordination mode (Figure 27).¹⁴¹ In this way, a variety of unsymmetrical PNP-based ligands were prepared, featuring R = ^tBu groups positioned on one phosphine arm and one of R′ = Ad, ^tBu, ⁱPr, or Cy on the other. In all cases, the corresponding (PNP)MoCl₃ precursors were accessed and reduced under standard NaHg amalgam conditions, giving binuclear [Mo] complexes having one of two geometries. In the first, the two terminal N₂ units remain characteristically trans-disposed (R = ^tBu, R′ = Ad or Ph) and in the second, the two terminal N₂ groups are mixed; both -trans and -cis geometries were found; both geometries feature a bridging N₂ unit. Subjecting the four of these catalysts to a solution of [LutH]OTf (96 equiv) and addition of Cp₂Co (72 equiv) via syringe pump over 1 h produces between 1 and 14 equiv of NH₃/Mo center—the greatest yields were observed for the catalyst containing R′ = ^tBu or Ad groups. Naturally, the phosphine R groups have a profound influence on downstream Mo(N_xH_y) adducts, causing either enhanced or decreased activity. This work highlights the subtle nuances of ligand design and shows that larger R groups are more adept at protecting the reactive Mo core.

Figure 27.

Modification of ligand phosphine substituent gave different Mo₂(μ-N₂) isomers. For R = ^tBu, R′ = Ad or Ph, trans-/trans- is observed, while for R = ^tBu, R′ = ⁱPr or Cy, trans-/cis-is observed. The R = ^tBu, R′ = Ad substituted phosphine provides the highest yield of NH₃ (14 equiv) during N₂RR.¹⁴¹

In an effort to develop catalytic N₂RR using N₂ and H₂ under mild reaction conditions, the Nishibayashi group studied the reactivity of high-valent molybdenum complexes with hydride reagents to generate on-cycle catalytic dinitrogen complexes (Figure 28).¹⁴² For this reason, Arashiba et al. showed that reduction of [PNP]MoCl₃ (R = ^tBu) with KC₈ produces a μ-N₂ bridged dimolybdenum precursor, {[PNP]Mo(Cl)₂]₂}(μ-N₂) (Figure 27). Treatment of this precursor with KHBEt₃ results in formation of the dinitrogen complex, {[PNP]Mo(N₂)₂]₂}(μ-N₂) as well as 2 equiv of H₂. Reaction of {[PNP]Mo(Cl)₂]₂}(μ-N₂) with excess H₂SO₄ provides no reduced nitrogen-based products (cf. 0.6 equiv of NH₃, 0.06 equiv of N₂H₄ for {[PNP]Mo(N₂)₂]₂}(μ-N₂)). This reaction demonstrates that high-valent [Mo]-halide bonds can be cleaved by way of hydride reagents, allowing for access to on-path [Mo]-N₂ complexes by way of H₂ elimination.

Figure 28.

On-cycle (PNP)Mo(N₂) complexes can be generated through H₂ evolution from their respective halide precursors.¹⁴²

In 2013, a modified ligand was reported where the ligand pincer phosphine donors were changed to arsine, a donor featuring a larger element of poorer σ-donating/π-accepting ability (Figure 29).¹⁴³ The ANA pincer ligand was prepared and coordinated to molybdenum; reduction of this precursor provided a cis-/trans-N₂ dimolybdenum complex having υ_NN = 1904 cm⁻¹ for the μ-N₂ ligand (cf. υ_NN = 1890 cm⁻¹ for the PNP analogue, {[PNP]Mo(N₂)₂]₂}(μ-N₂) R = ^tBu). In the presence of PMe₃, a phosphine complex, [ANA]Mo(N₂)₂PMe₃ was isolated. Despite producing similar amounts of NH₃ (cf. the PNP system) when exposed to excess H₂SO₄ (ca. 0.6 equiv of NH₃), catalysis was nonproductive using {[ANA]Mo(N₂)₂]₂}-(μ-N₂) R = ^tBu) and varying equivalents of Cp₂Co and [LutH]OTf, producing only 2 equiv of NH₃. This data point contrasts with the PNP-based system where 12 equiv of NH₃ was observed under similar conditions. The previously formed unsymmetrical phosphine complexes that contained trans-/cis-N₂ units were also ineffective for catalysis (R = ^tBu, R′ = ⁱPr (3 equiv of NH₃), Cy (1 equiv of NH₃)).

Figure 29.

Arsine analogues of the PNP class of ligands have been prepared, though Mo complexes of these ligands are not active for catalytic N₂RR under the conditions shown.¹⁴³

4.2. Mechanistic Conclusions That Can Be Drawn and Uncertainties

Next, the effect of substitution at the 4-position of such PNP ligands having P^tBu₂ groups was explored (Figure 30).⁸² The installation of both electron-donating and -withdrawing groups was carried out, with the hypothesis that electron-donating groups would increase N₂ activation and thus accelerate Nβ protonation. From this design perspective, five candidate [(^4-XPNP)Mo(N₂)₂]₂(μ-N₂) complexes were prepared having X = Ph, Me₃Si, ^tBu, Me, and OMe groups at the 4-position. Structurally, no distinct differences were observed from the parent (X = H) system. As judged by IR spectroscopy, substitution of more electron-donating groups provides a more activated N₂ entity (υ_NN = 1932 cm⁻¹ for X = OMe vs 1944 cm⁻¹ for X = H). All of these species are catalytically active for NH₃ formation, resulting in 28 and 31 equiv of NH₃ for the ^tBu- and Me-substituted PNP ligands, respectively, using 216 equiv of Cp₂Co and 288 equiv of [LutH]OTf. The greatest amount of NH₃ is formed when R = OMe (34 equiv) under the same conditions.

Figure 30.

Ligand modifications have been targeted by changing the ligand 4-X group. The reaction scheme depicts the first three steps relevant to catalytic N₂RR as calculated by DFT (1) N_β protonation, (2) N₂ dissociation, and (3) triflate coordination.⁸²

The study of N₂ reduction catalysis using density functional theory provided insight into the high barrier by which initial N_β protonation occurs (Figure 30). This process involves three steps, (1) protonation to give a Mo diazenido, Mo(NNH), followed by (2) displacement of the trans-N₂ ligand, and (3) attack of a triflate (OTF⁻) unit—the final two steps offsetting the high endothermicity of the first. In terms of phosphine R group, installation of ^tBu best stabilizes the protonated diazenido intermediate and was found to have the lowest energy pathway. Substitution of an electron-donating group at the PNP 4-position is also thought to accelerate Nβ protonation due to N₂ activation.

To test the importance of reductant concentration on the catalytic behavior of the 4-OMe catalyst, Cp₂Co was added over a longer period of time (5 h vs 1 h) and gave a greater amount of NH₃ (36 equiv), suggesting a longer catalyst lifetime. Increasing the amount of acid ([LutH]OTf: 480 equiv) and reductant (Cp₂Co: 360 equiv) provides 52 equiv of NH₃. In terms of reductant and acid optimization, it was found that Cp*₂Cr has a negligible effect on reaction outcome, while use of [^4-ClLutH]OTf (pK_a = 5.46 in H₂O) provides only a stoichiometric amount of NH₃. Use of a weaker acid, [^4-CH₃LutH]OTf (pK_a = 7.55 in H₂O) gives only 10 equiv of NH₃. Given the highly active nature of the 4-OMe catalyst, mechanistic studies comparing the ^4-OMePNP catalyst and the parent PNP catalyst were carried out. From these, it was determined that when X = H the reaction is first order in Mo, while when X = OMe the reaction is zeroth order in Mo, suggesting different rate-determining steps (the reaction order with respect to reductant was found to be first order in both cases). At present, it is unclear how such a mechanism can be zeroth order in Mo. Dihydrogen yields were also measured, showing that X = OMe not only produces the highest quantity of NH₃, but also the lowest quantity of H₂. Performing the reaction under a partial H₂ atmosphere also results in lower activity for NH₃ formation for the 4-OMe catalyst (96:4 N₂/H₂ — 23 equiv of NH₃), indicating that H₂ inhibits NH₃ formation—a conclusion that has been drawn for other synthetic nitrogenases (N₂ases).¹⁰⁷ Overall, substitution at the 4-position of the PNP ligand was found to have a marked effect on N₂ reduction selectivity, allowing for rate enhancement of the first protonation event and suppression of molecular hydrogen formation.

In 2014, Tanaka et al. explored the use of mononuclear molybdenum nitrides for catalytic N₂RR (Figure 31).⁸³ Treatment of Mo(Cl)₃(THF)₃ with Me₃SiN₃ at 50 °C for 1 h, followed by addition of the PNP ligand, gave (PNP)Mo(N)-(Cl)₂, which served as a synthetic precursor to related species including (PNP)Mo(N)(Cl), [(PNP)Mo(N)(Cl)]OTf, and [(PNP)Mo(NH)(pyr)(Cl)]OTf (pyr = pyridine). Of these compounds, only (PNP)Mo(N)(Cl) and the oxidized product, [(PNP)Mo(N)(Cl)]OTf, exhibited catalytic efficacy (6.6 and 7.1 equiv of NH₃).

Figure 31.

Mononuclear PNP-ligated molybdenum nitrides are active for catalytic N₂ fixation using [LutH]⁺ (48 equiv) and Cp₂Co (36 equiv).⁸³

In view of this result, DFT calculations were performed to assess the relevancy of [(PNP)Mo(N)]OTf or related complexes to the observed catalysis83 (Figure 32). The initial pathway was calculated to be similar to that outlined above (Figure 30), with a dinuclear molybdenum complex being protonated initially, and then breaking up upon a third (reductive) protonation step to liberate NH₃ and a terminal nitride product, [(PNP)Mo(N)]OTf. The latter is a competent precatalyst, and, following a series of reductive protonation steps, (PNP)Mo(NH₃) is generated and combines with (PNP)Mo(N₂)₃ to regenerate a μ-bridged N₂ complex; NH₃ can be replaced by N₂ in this species to restart the catalytic cycle. (PNP)Mo(N₂)₃ itself is not thought to be catalytically competent. Instead, the authors suggest that the terminal N₂ ligand that undergoes monoprotonation (the first step of the cycle) must be that of a dimolybdenum complex, with electron transfer from an adjacent Mo center preactivating the Mo(N₂) fragment. This is an interesting hypothesis that provides an opportunity for future experimental studies.

Figure 32.

Complete calculated N₂RR cycle for Nishibayashi’s dinuclear Mo₂ complexes. [Mo] = (PNP)Mo(N₂)₃.⁸³

Redox-active subunits have also been incorporated within the PNP ligand framework by the Nishibayashi group to canvass possible advantages of placing a redox relay close to the catalytic active site (Figure 33).¹⁴⁴ The direct incorporation of metallocene subunits such as ferrocene and ruthenocene, in the trans-/trans-N₂ (PNP)dimolybdenum framework, was found to have a negligible effect on the degree of N₂ activation as judged by IR spectroscopy; derivatives featuring ethylene or phenylene linkers to the metallocene subunits also showed similar activation. Cyclic voltammetry revealed marked shifts in the E_1/2 associated with the Fe(II/III) couple compared to free ferrocene, consistent with electronic communication between Fe and Mo; spectroelectrochemistry demonstrated low energy transitions in the near-IR at around 800–1800 nm, consistent with Fe-to-Mo charge transfer (MMCT). The degree of electronic communication was weaker when a linking (ethylene or phenylene) group was present.

Figure 33.

A survey of Nishibayashi’s Mo complexes for catalytic N₂RR Fc = ferrocene, EtFc = ethylferrocene.^{83,84,144–146}

In terms of catalysis, it was found that the ferrocene-appended catalyst gave 37 equiv of NH₃, compared with 23 equiv of NH₃ under the same conditions where X = H (Figure 33). Control experiments employing the X = H parent catalyst in the presence of exogenously added ferrocene gave only 21 equiv of NH₃, suggesting that inclusion of the redox-active subunit may have some benefit. From a DFT study, one step that was suggested to be dramatically affected by inclusion of a metallocene within the catalyst framework concerns reduction of the hydrazidium, [Mo(NNH₃)]⁺, complex, enabling release NH₃ and Mo(N). Azaferrocene-based PNP Mo complexes have also been employed to survey their efficacy for catalytic N₂ fixation, but they were not successful under the conditions explored.¹⁴⁵

PNN-based Mo complexes have also been studied by Nishibayashi and co-workers, especially for derivatives featuring the PNN ligand 2-(di-ferf-butylphosphinomethyl)-6-(diethylaminomethyl)pyridine (Figure 32).¹⁴⁶ (PNN)Mo(N₂)₂(PMe₂R) (R = Ph or Me) derivatives were accessed by reduction of the (PNN)MoCl₃ precursor and treatment with PMe₂R. The PNN ligand was observed to be a stronger donor ligand than the corresponding PNP ligand, as judged by N₂ stretching frequencies (1877 cm⁻¹ vs 1915 cm⁻¹ for the PNP ligand). However, treatment of these complexes, or the corresponding nitride derivatives, with excesses of reductants and acids (e.g., 36 equiv of Cp₂Co, 48 equiv of [LutH]OTf failed to reveal catalytic activity (Figure 33)). Other PNP ligands, such as N,N′-bis(di-ferf-butyl-phosphino)-2,6-diaminopyridine and 2,6-bis(diadamantyl-phosphinomethyl)pyridine, have also been used to access terminal molybdenum nitrides of the type [(PNP)Mo(N)(Cl)]BAr^F₄; however, these adducts produced little NH₃ (1.3 and 1.8 equiv., respectively) when exposed to 36 equiv of Cp₂Co and 48 equiv of [LutH]OTf.¹⁴⁷

PPP-pincer ligands were next targeted as promising candidates for catalysis (Figure 33).⁸⁴ Despite difficulties in synthesizing the N₂-bridged dimolybdenum PPP-pincer complex, the terminal nitrides, (PPP)Mo(N)(Cl) and [(PPP)Mo(N)(Cl)]BAr^F₄—akin to those prepared in the PNP systems83—were accessible. Exposure of the cationic Mo nitride to [LutH]OTf (48 equiv) and 36 equiv of Cp*₂Co produced 6.1 equiv of NH₃, whereas Cp*₂Cr and Cp₂Co were far less effective reductants. However, the use of [ColH]OTf (ColH = 2,4,6-trimethylpyridinium) as the acid produced 9.6 equiv of NH₃, and the addition of much higher amounts of reductant and acid (540 equiv of Cp*₂Co and 720 equiv of [ColH]OTf), produced as many as 63 equiv of NH₃.

4.3. Catalyst Redesign and Keys to Achieving Higher Turnover Numbers

In 2017, Nishibayashi and co-workers proposed that reactive terminal (PNP)Mo(N) could be formed via cleavage of a bridging dinitrogen unit, thus circumventing the formation of early protonation intermediates such diazenido, hydrazido, and hydrazidium (Figure 34).⁸¹ Use of (PNP)Mo(I)₃ as a precursor afforded access to (PNP)Mo(N)(I) following reduction by Cp*₂Co. By contrast, reduction of the chloride-containing species, (PNP)MoCl₃, with NaHg amalgam produced [(PNP)Mo(N₂)₂]₂(μ-N₂), thus highlighting the importance of the Mo-halide starting material and reductant. Using (PNP)Mo(I)₃ as a (pre)catalyst, treatment with standard loadings of 36 equiv of Cp*₂Co and 48 equiv of [ColH]OTf gave 10.9 equiv of NH₃, while (PNP)Mo(N)(I) gave 12.2 equiv of NH₃. Under similar conditions, [(PNP)Mo(N₂)₂]₂(μ-N₂) gave 7.1 equiv of NH₃. [LutH]OTf was not applicable as an acid source when using (PNP)Mo(I)₃ as the (pre)catalyst.

Figure 34.

Proposed pathway for reductive cleavage of N₂ to form a terminal nitride, (PNP)Mo(N)(I), of relevance to catalytic N₂RR. The observed reactivity profile is I (50.7 equiv of NH₃) > Br (40.5 equiv of NH₃) > Cl (24.4 equiv of NH₃).⁸¹

Much higher loadings established the viability of much higher, and truly impressive turnover numbers. For example, the use of 2160 equiv of Cp*₂Co and 2880 equiv of [ColH]OTf gave the highest amount of NH₃ (415 equiv), more than 35 times greater yield than achieved by [(PNP)Mo(N₂)₂]₂(μ-N₂) under similar acid and reductant loadings. A lower amount of ammonia was produced using (PNP)Mo(Br)₃ (40.5 equiv of NH₃) or (PNP)Mo(Cl)₃ (24.4 equiv of NH₃) when compared with (PNP)Mo(I)₃ (50.7 equiv of NH₃; [catalyst] = 0.002 mmol). In all cases, the catalysis showed little competing dihydrogen formation.

The proposed catalytic cycle, which warrants additional experimental support, invokes reduction of (PNP)Mo(X)₃ to give [(PNP)]Mo(X)]₂(μ-N₂), from which (PNP)Mo(N)(X) is then formed (Figure 34). Following a series of stepwise reductive protonation steps (PNP)Mo(NH₃)(X) is generated, which then loses NH₃ and (re)binds N₂ to restart the catalytic cycle. This cycle is interesting to consider by comparison to a distal (or alternating) pathway for N₂ reduction via discrete one-electron/one-proton steps, as it features fewer protonated intermediates. Moreover, this work points to the viability of halide precatalysts for generating on path intermediates.

The Nishibayashi group has more recently reported catalytic studies with a variety of molybdenum triiodide complexes containing nearly all of the PNP permutations discussed above, (^RPNP)Mo(I)₃, including different phosphine substituents (R = H, R′ = ^tBu, Ad, ⁱPr, or Ph) or different groups at the 4-position of the central ring of the ligand (X = Ph, Me, OMe, Fc, or Rc, R′ = ^tBu).¹⁴⁸ Interestingly, and in contrast to the Mo₂(μ-N₂) complexes discussed previously, the highest quantity of NH₃ was obtained when X = Ph. To rationalize this observation, it was proposed that (PNP)Mo(I)₃ and [(PNP)Mo(N₂)₂]₂(μ-N₂) operate via different rate-determining steps (RDS). In the molybdenum iodide system, electron-withdrawing groups are thought to promote an RDS involving reduction of a cationic “Mo(NH_y)” fragment, while in the Mo₂(μ-N₂) system, electron-donating groups are thought to promote an RDS involving N_β protonation.

PCP-type pincer ligands, comprised of both N-heterocyclic carbene and diphosphine donors, have also been explored by the Nishibayashi group in an effort to evaluate N₂RR catalyst performance (Figure 35).¹⁴⁹ Two such ligands: 1,3-bis(2-(di-tert-butylphosphino)methyl)benzimidazol-2-ylidene (^BImPCP) and 1,3-bis(2-(di-tert-butylphosphino)ethyl)imidazol-2-ylidene (^ImPCP) were prepared, from which (^rPCp)moCl₃ (R = Im or BIm) precursors were accessed. Reduction using NaHg amalgam provided the μ-N₂ molybdenum complexes [(^RPCP)Mo(N₂)₂]₂(μ-N₂), for which the IR spectrum provides signals at 1978 cm⁻¹ and 1911 cm⁻¹. The catalytic utility of these compounds was next interrogated using a mixture of [LutH]OTf (96 equiv) and metallocene (72 equiv); the metallocene was added in toluene via syringe pump. After optimization, it was found that pairing larger amounts of reductant (Cp*₂Cr: 1440 equiv) and acid ([LutH]OTf: 1920 equiv) with the 5,6-dimethylbenzimidazol-2-ylidene catalyst afforded the maximum (up to 115 equiv) amount of ammonia.

Figure 35.

PCP-type pincer ligands featuring N-heterocyclic carbene donors are useful for N₂RR in the presence of [ColH]OTf (96 equiv) and reductant (72 equiv); Cp*₂Cr was found to be most effective for both classes of catalyst.¹⁴⁹

In an effort to use H₂O as a proton source, Nishibayashi and co-workers have explored the combination of coupling wazter oxidation with nitrogen fixation (Figure 36).¹⁵⁰ In a model study, water oxidation was performed using ceric ammonium nitrate ([Ce(NO₃)₆](NH₄)₂) and a Ru complex, Ru(bda)-(isoq)₂ (bda = 2,2′-bipyridine-6,6′-dicarboxylate, isoq = isoquinoline) to generate O₂ and acid. Addition of lutidine formed [LutH]⁺, which upon combination with Cp₂Co in the presence of [(PNP)Mo(N₂)₂]₂(μ-N₂), generated 8.5 equiv of NH₃; this compares well with 9.7 equiv of NH₃ that was observed when purified [LutH]OTf was used. Using larger amounts of reagent resulted in up to 17.1 equiv of NH₃. Visible light-driven water oxidation was also studied using Na₂S₂O₈ as a sacrificial oxidizing reagent and [Ru(bpy)₃]OTf (bpy = 2,2′-bipyridine) as the photosensitizer. Using the same catalyst, 6.0 equiv of NH₃ was observed following sequestration of the solvated HOTf with lutidine; it was proposed that sulfate inhibits the formation of NH₃ from N₂.

Figure 36.

Efforts toward using H₂O as a proton source in N₂RR. Water oxidation was performed using [Ru(bpy)₃]OTf₂ as a photooxidant, peroxydisulfate as a sacrificial reductant, and a Ru complex, [Ru(bda)-(isoq)₂] (bda = 2,2′-bipyridine-6,6′-dicarboxylate, isoq = isoquinoline), as the water oxidation catalyst. The generated acid was trapped using lutidine to give [LutH]⁺, which enabled N₂RR catalysis by (PNP)Mo.¹⁵⁰

4.4. New Conditions Lead to Remarkable Rate and Efficiency

Very recently, Nishibayashi and co-workers have exploited SmI₂ as a reductant compatible with polar, protic H⁺ donors (e.g., alcohols or water) (Figure 37).⁷⁹ Related approaches have been gaining in popularity for the reduction of organic molecule substrates owing to the dramatic O–H bond weakening that occurs upon coordination of the alcohol or water to the Sm(II) center.^151–154 This is an exciting result that opens up a new approach for the study of N₂RR catalysts that shows great promise. Indeed, the selectivity for NH₃ and total turnover number demonstrated in this recent report are remarkable. Using the aforementioned (^BMimPCP)Mo(Cl)₃ precatalyst, reaction of N₂ with SmI₂ and ethylene glycol generates NH₃ at a turnover frequency (TOF) of 117 min⁻¹. Water can also be used in this reaction; treatment of N₂ with a [Mo]-precatalyst, SmI₂ (14 400 equiv), and H₂O (14 400 equiv) in THF at room temperature for 4 h gives 4350 equiv of NH₃ (91%).

Figure 37.

Remarkable efficiencies were achieved using a Sm/alcohol mixture for N₂RR. This reaction is proposed to occur via PCET. Use of H₂O (14400 equiv) and SmI₂ (14 400 equiv). gave ca. 4350 equiv of NH₃.⁷⁹

5. ACHIEVING NITROGEN FIXATION CATALYSIS AT Fe

For decades, Mo had played center stage in the modeling of biological nitrogenases, owing in part to its early discovery within the MoFe variant.⁴⁸ The initially unexpected presence of Mo, the sensitivity of nitrogenase performance to alterations near Mo, and the early success in developing model systems, led to a strong emphasis on this metal as being the site of N₂RR. Then, in 1986, it was discovered that there was a vanadium-dependent nitrogenase, VFe,¹⁵⁵ and subsequently in 1988, the existence of an all Fe biological nitrogenase was disclosed.¹⁵⁶ Subsequent study of the biological nitrogenases has led to the conclusion that they are likely structurally (and potentially mechanistically) similar.^{11,36,157–159} In particular, recent spectroscopic and crystallographic evidence has begun to emerge in both the MoFe and VFe nitrogenases that the site of N₂-binding and functionalization is likely to be at an Fe site. In particular, both the Rees and Einsle groups have obtained structures that demonstrate sulfide-loss and ligand binding between Fe₂ and Fe₆ (Figure 38).^160,161 Meanwhile, Hoffman, Seefeldt, and co-workers have established that reductive elimination of bridging iron hydrides may be essential to N₂ binding in all three biological nitrogenases.^158,159,162

Figure 38.

Structure of a recently reported, proposed intermediate of VFe-nitrogenase, featuring removal of a bridging sulfide as SH⁻ and the identification of a μ₂-bridging light atom in its place. The light atom (X) has been hypothesized to be the N atom of an imido ligand.¹⁶¹ Further studies are needed to validate this assignment.

The discovery of new nitrogenases expanded the focus not only from Mo to other metals, but also from NH₃ to another N₂ fixation product, N₂H₄. In 1991, Dilworth and co-workers discovered that the VFe-nitrogenase, unlike MoFe-nitrogenase, released N₂H₄ as a product of N₂ reduction, particularly at higher temperature.¹⁶³ In this initial publication, they proposed that VFe-nitrogenase can access two routes to fixed nitrogen products from a common hydrazido(2−) ligand. The first is a further N_β-functionalization (forming M(NNH₃)) and subsequent N-N bond cleavage to release NH₃, in a Chatt-like mechanism, while the other pathway would involve double N_α-functionalization leading to M(NHNH₂) then M(NH₂NH₂) and ultimately N₂H₄ release (Figure 39). This proposal disappeared relatively quickly from the literature (though it has been recently revisited by Peters and co-workers, and termed a “hybrid” mechanism, vide infra) and was replaced by a mechanistic scenario typically referred to as the alternating mechanism. In this latter mechanism, the second functionalization occurs at N_α rather than N_β, leading to a characteristic diazene (HNNH) intermediate. Indeed, this mechanism was popularized by the Hoffman group after the observation that diazene (and related derivatives) were substrates of FeMo-nitrogenase and led to the formation of a cofactor-bound intermediate that could be detected by ENDOR spectroscopy.^164,165 They hypothesized that further reduction of diazene to hydrazine would occur via alternating N_β- and N_α-functionalization. In a general sense, such a hydrazine-bound intermediate could then be released, protonated to release N₂H₅⁺, or undergo protolytic cleavage to form NH₃ and M(NH₂). Still, to date, no fixed N_xH_y intermediate has been definitively assigned in a biological nitrogenase and mechanistic uncertainty remains.

Figure 39.

(Left) Outer cycle depicts the distal or Chatt-cycle for N₂RR.³⁰ Central intermediate (highlighted in blue) represents a “hybrid” mechanism for N₂RR, which may be relevant in both biological¹⁶³ and synthetic^109,114 nitrogenases; (right) the alternating mechanism for the reduction of N₂ to NH₃ or N₂H₄.^164,165

In response to the growing evidence for the biological importance of Fe and mechanistic diversity in nitrogen fixation, inorganic chemists have articulated both multi-Fe^166,167 and single-Fe¹⁶⁸ hypotheses for nitrogen fixation, as well as approaches to interrogate mechanisms for nitrogen fixation that account for both NH₃ and N₂H₄ formation.^25,109 To support these hypotheses, researchers have developed a number of systems to achieve N₂-binding and functionalization, as well as the synthesis and interconversion of potential late-stage intermediates.^{26,29,169–171} However, the lack of early progress on both protonation reactions of Fe(N₂) complexes that led to identifiable intermediates or the development of catalytic systems for N₂-to-NH₃ conversion that featured Fe, left a notable gap for the field until relatively recently.¹⁷

Since 2013, there has been a surge in the discovery of catalytically active systems (Figure 40)^{18,80,115,116,172–175} and the identification of protonated intermediates from Fe(N₂) complexes.^{109,113,176,177} Indeed, many of the types of species and reactivity patterns of synthetic, homogeneous Mo model chemistry in the context of nitrogen fixation have now been demonstrated for Fe as well. This body of work adds chemical evidence to the growing weight of biological evidence that Fe is capable of N₂-to-NH₃ conversion when supplied with the appropriate H⁺/e⁻ source and unravels the historical bias that Mo is somehow unique in its ability to mediate N₂RR via cycles such as those shown in Figure 39.^16,17

Figure 40.

Homogeneous Fe complexes reported to mediate catalytic N₂-to-NH₃ conversion; Dipp = 2,6-diisopropylphenyl.^{18,115,116,172,173,175,178–180} **(depe)₂Fe(N₂) (depe = diethylphosphinoethane) is instead selective for N₂H₄.¹⁷⁴

5.1. Catalysis with a Trisphosphineborane Fe Species

In 2013, the Peters group reported that a tris-(isopropylphosphino(o-phenylene)borane iron complex, (P₃^B)Fe, was capable of catalytic nitrogen fixation.¹⁸ This atrane system was initially targeted for catalytic nitrogen fixation due to its ability to stabilize a variety of nitrogen fixation intermediates. The P₃^B ligand reacts with FeBr₂ in the presence of Fe powder to form (P₃^B)Fe(Br), a valuable synthon.¹¹² This complex could then be reduced by stoichiometric sodium naphthalenide to give the Fe(N₂) complex (ν_NN = 2011 cm⁻¹, Figure 41). Reaction with excess sodium naphthalenide led to formation of the anionic Fe(N₂) complex (ν_NN = 1877 cm⁻¹), and encapsulation of the sodium with 12-crown-4 afforded a separated ion pair (ν_NN = 1918 cm⁻¹). In addition to activating π-accepting ligands, (P₃^B)Fe was capable of stabilizing π-donating imido ligands (i.e., (P₃^B)Fe(NR)) formed via reaction of (P₃^B)Fe(N₂) with organoazides.^112,113

Figure 41.

Interconversion of different nitrogen fixation intermediates on the (P₃^B)Fe platform.^{112,113,176,182}

The functionalization of [(P₃^B)Fe(N₂)]⁻ with electrophiles also led to the formation of Fe–N multiply bonded species.¹⁸¹ In particular, reaction with TMSCl led to formation of the silyldiazenido, the solid-state structure of which reveals significant lengthening of the Fe–B interaction and the N–N bond and a contracted Fe–N bond. This flexibility of the Fe–B interaction in response to the apical ligand and redox state of the Fe has emerged as a key factor in stabilizing intermediates throughout the N₂ fixation process (vide infra). The anionic silyldiazenido could be readily synthesized via reduction of this complex with Na/Hg. Likewise, difunctionalization of the N₂ ligand could be achieved by reaction of (P₃^B)Fe(Br) with excess Na/Hg and the disilylating reagent, 1,2-bis(chlorodimethyl)-silylethane. DFT calculations performed on this disilylhydrazido(2−) complex suggested that it has structural features similar to (P₃^B)Fe(NR) species. As expected, calculations of the bond orders indicated that across the series: (P₃^B)Fe(N₂) → [(P₃^B)Fe(N₂)]⁻ → (P₃^B)Fe(NNTMS) → [(P₃^B)Fe(NNTMS)]⁻ → (P₃^B)Fe(NN(SiR₃)₂) the Fe–N bond order increased from ~0.5 to 2, with the N–N bond order concomitantly decreasing from ~2.6 to 1.2. Considering the silyl groups as proton surrogates, these results suggested that (P₃^B)Fe might support early nitrogen fixation intermediates.

Next, late nitrogen fixation intermediates were targeted for synthesis using the (P₃^B)Fe platform.¹⁸² Synthesis of the low-coordinate, cationic complex [(P₃^B)Fe]⁺ provided access to a number of species that would be candidate downstream intermediates of catalytic nitrogen fixation, including [(P₃^B)Fe(N₂H₄)]⁺, [(P₃^B)Fe(NH₃)]⁺, and (P₃^B)Fe(NH₂). Low temperature protonation of (P₃^B)Fe(NH₂) with HBAr^F₄ led to the quantitative formation of [(P₃^B)Fe(NH₃)]⁺. Reductive N₂-for-NH₃ substitution was then established by addition of KC₈ to [(P₃^B)Fe(NH₃)]⁺, which provided (P₃^B)Fe(N₂) quantitatively. These reactions demonstrated the final key steps of a hypothetical nitrogen-fixing cycle and suggested the possibility that HBAr^F₄ and KC₈ might be competent proton and electron sources to drive such a process, at least insofar as the productive cycle could compete with background hydrogen evolution.

In 2013, it was discovered that the reaction of [(P₃^B)Fe(N₂)]⁻ with 6 equiv of HBAr^F₄ at low temperature, followed by warming and then adding 6 equiv of 1,8-bis(dimethylamino)naphthalene (Proton Sponge) resulted in the generation of ~35% [(P₃^B)Fe(NH₃)]⁺, ~40% [(P₃^B)Fe]⁺, as well as several unidentified Fe-containing products.¹⁸ Low temperature reaction with acid followed by KC₈ led to the regeneration of [(P₃^B)Fe(N₂)]⁻, suggesting that the system might indeed be amenable to catalysis (Figure 41). After screening a variety of solvents, temperatures, reductants, and acids, it was found that catalytic yields of NH₃ (7.0 ± 1 equiv of NH₃ per Fe, 44 ± 6% based on H⁺ used) could be formed in diethyl ether at −78 °C using HBAr^F₄ and KC₈ as the proton and electron sources, respectively. This provided the first demonstration that Fe could catalyze N₂-to-NH₃ conversion. Worth underscoring is that it was surprising that HBAr^F₄ and KC₈ could provide an effective acid and reductant cocktail. It would have been reasonable to presume that essentially all of the acid would be reduced by KC₈ to produce H₂. While such a reaction does occur, at −78 °C in Et₂O it is comparatively slow and the [(P₃^B)Fe(N₂)]⁻ catalyst is sufficiently competent making N₂RR kinetically competitive.¹⁷³

Additional mechanistic studies of this catalytic system were undertaken using kinetics and freeze-quench Müssbauer spectroscopy.¹⁷³ The catalytic reaction was found to be first order in Fe but zeroth order in acid, suggesting that the rate-limiting step likely involved either a reduction step or a ligand exchange process. The initial rate of NH₃ formation was found to be 1.2 ± 0.1 mol of NH₃ mol of Fe¹⁻ min⁻¹, which was among the fastest rates for a synthetic system, despite the fact that the reaction was carried out at −78 °C.⁷⁹ Catalytic reactions with preadded NH₃ demonstrated that the reaction was product-inhibited. This is perhaps not surprising given the slow kinetics of NH₃ and N₂ exchange observed in (HIPTN₃N)Mo systems¹¹⁰ and points to a challenge for developing systems that can achieve substantially higher turnover numbers. Interestingly, (P₃^B)Fe was shown to catalyze both NH₃ and H₂ formation. This suggested that to achieve improved catalysis it would be necessary not only by limiting background reactivity between the acid and the reductant, but also to develop systems with improved intrinsic selectivity for NH₃.

Addition of further substrate (HBAr^F₄ and KC₈) following a standard catalytic run with (P₃^B)Fe furnished additional NH₃, suggesting that catalytically active “(P₃^B)Fe” material was still present. To explore this further, freeze-quench Müssbauer spectra of the catalytic reaction at various time points were used to evaluate the evolution of the catalyst speciation. Spectra at early times (~5 min) revealed the formation of a previously characterized hydride/borohydride complex, (P₃^μ-B-H)Fe(H)(L) (L = N₂ or H₂). These derivatives were originally accessed via the addition of H₂ to (P₃^B)Fe(N₂) (Figure 41).¹⁸³ At later reaction times (~25 min), the original [(P₃^B)Fe(N₂)⁻ (pre)catalyst had reformed, consistent with the use of a modest excess of reductant. This result was in apparent contrast with initial results that had indicated (P₃^μ-B-H)Fe(H)(N₂) was inactive for catalysis; indeed, it was presumed it might be a catalyst poison. Its in situ generation and subsequent consumption, regenerating a substantial quantity of the [(P₃^B)Fe(N₂)]⁻ (pre)catalyst, suggested it to instead be an off-path intermediate that ties-up the catalyst. Consistent with this idea, it was shown that the addition of a toluene cosolvent to help solubilize (P₃^μ-B-H)Fe(H)(N₂) in Et₂O at −78 °C significantly improved its catalytic activity (up to 5.6 ± 0.9 equiv of NH₃ per Fe). This result supported that (P₃^μ-B-H)Fe(H)(N₂) can be converted back into an on-cycle intermediate and is therefore an off-path resting state of the system (Figure 42).

Figure 42.

Reaction scheme showing synthetic conversions of the catalytic resting state, (P₃^μ-B-H)Fe(H)(N₂), including an α-hydride elimination step that leads to release of a hydrazine surrogate following acidic workup.^173,175,184

Understanding how the catalytic resting state, (P₃^μ-B-H)Fe(H)(N₂) returns to an on-cycle intermediate is of interest (Figure 42). One demonstrated pathway is that the sequential reaction of (P₃^μ-B-H)Fe(H)(N₂) with HBAr^F₄ and KC₈ in 3:1 Et₂O/toluene at low temperature led to the reformation of [(P₃^B)Fe(N₂)]⁻ (presumably accompanied by H₂ loss). It was also found that the reaction of (P₃^μ-B-H)Fe(H)(N₂) with KC₈ led to the formation of a dianionic species, [(P₃^μ-B-H)Fe(H)(N₂)]²⁻, in which one phosphine donor is dissociated from the iron center. The reaction of this dianionic species with silyl electrophiles at low temperature led to the formation of a disilylhydrazido(2−) species, (P₃^μ-B-H)Fe(H)(NNSi₂). Upon warming, the Fe–H group was observed to migrate to N_α giving a hydrazido(1−) ligand. The resulting hydrazido(1−) ligand could then be protolytically released as a hydrazine derivative by reaction with HBAr^F₄ in the presence of tert-butylisocyanide. This result,¹⁸⁴ along with those from other groups,¹⁸⁵ suggests that hydrides may be productively incorporated into fixed-N products, in addition to being lost as H₂.

It has recently been discovered that irradiation of (P₃^μ-B-H)Fe(H)(N₂) with a Hg lamp for 10 min at −78 °C results in formation of (P₃^B)Fe(N₂) along with (P₃^μ-B-H)Fe(H₂)(H). Such a process can improve overall catalysis by photo inducing an off-path species, (P₃^μ-B-H)Fe(H)(N₂) to return to an on-path species, (P₃^B)Fe(N₂). Irradiation of catalytic runs led to an increased total turnover number (~+150%) for a given loading of acid and reductant using [P₃^BFe(N₂)]⁻ as the (pre)catalyst.¹⁷⁵

In the original catalytic report,¹⁸ Peters and co-workers had noted that EPR spectra of the reaction of [(P₃^B)Fe(N₂)]⁻ with excess acid, in the absence of added reductant or base, led to a new and intense S = 1/2 EPR signature (~85% of the Fe present via spin-quantification). This new species that was tentatively ascribed to [(P₃^B)Fe(NNH₂)]⁺, or possibly an isomer such as [(P₃^B)Fe(NHNH)]⁺. To elucidate the chemical structure of this EPR-active intermediate, pulse EPR and XAS spectroscopic studies were undertaken.¹¹³ Consistent with retention of the (P₃^B)Fe-core, pulse EPR techniques revealed couplings to three phosphorus nuclei and a single boron nucleus. Additionally, protons (one more strongly and one more weakly coupled) and two coupled nitrogen nuclei were also identified. These data were consistent with formulation as either a hydrazido(2−) or an asymmetrically coordinated diazene (HNNH) ligand. X-ray absorption spectroscopy (XAS) revealed a short Fe–N bond (1.64 Å), highly consistent with bond lengths previously observed for disilylhydrazido(2−) and related imido species in the P₃^BFe-system.^112,181 Thus, this species was assigned as [P₃^BFe(NNH₂)]⁺.¹¹³ It was the first iron species to be characterized in which an N₂ ligand had been protonated to an Fe(N_xH_y) product, and the product is clearly an intermediate of a distal pathway.

The distinction between limiting distal and alternating pathways has already been discussed and is an issue that has been much discussed with respect to Fe- and Mo-mediated nitrogen fixation, both with respect to the enzymatic process and in synthetic catalysts. While interconversion between these limiting pathways may also be operative (vide infra),²² probing whether N–N bond cleavage occurred early (forming a nitride and ammonia) or late (forming an amide and ammonia) in this catalyst system was of interest. The ability to generate [(P₃^B)Fe(NNH₂)]⁺ led to studies to probe the viability of an early N–N cleavage step to generate [(P₃^B)Fe(N)]⁺, consistent with a distal pathway. The reaction of [(P₃^B)Fe(NNH₂)]⁺ with excess acid did not lead to N–N bond cleavage, likely because [(P₃^B)Fe(NNH₂)]⁺ is a very poor base. Thus, its one-electron reduced congener was targeted. To achieve this, [(P₃^B)Fe(N₂)]⁻ was further reduced by KC₈ to its dianionic derivative, [(P₃^B)Fe(N₂)]²⁻ (ν_NN = 1836 cm⁻¹). Reaction of this species at −135 °C in 2-MeTHF with excess HBAr^F₄ or HOTf for 15 min led to the clean formation of (P₃^B)Fe(NNH₂), characterized in situ via XAS (Fe–N bond length of 1.65(2) Å) and Mössbauer spectroscopy.¹⁷⁶ Its methylated analogue, (P₃^B)Fe(NNMe₂), was similarly characterized and proved sufficiently stable to also obtain its solid-state crystal structure.¹¹⁴ Protonation of [(P₃^B)Fe(N₂)]²⁻ with larger excesses of acid, and for longer reaction times, led to formation of a new species in the Mössbauer spectrum with ≥50% spectroscopic yield. The low isomer shift (−0.15 mms⁻¹) and the very large quadrupole splitting (6.2 mm s⁻¹) have previously been demonstrated to be diagnostic for tetrahedral, iron(IV) nitrides (Figure 43).^186–188 XAS characterization of [(P₃^B)Fe(N)]⁺, and comparison with its [(P₃^B)Fe(N₂)]²⁻, (P₃^B)Fe(NNH₂), and (P₃^B)Fe(NNMe₂) relatives, confirmed its expectedly short Fe–N bond distance (1.54(2) Å) owing to its Fe-to-N triple bond. It was additionally established that formation of [(P₃^B)Fe(N)]⁺ occurred concomitant with NH₃ formation (36.0 ± 0.5% isolated yield). The yield of NH₃ could be increased by sequential reaction of [(P₃^B)Fe(N₂)]⁻ with HOTf and Cp*₂Co (73 ± 7%). These results demonstrated, for the first time, the viability of an Fe-mediated distal pathway for N₂-to-NH₃ conversion, more than four decades after such a pathway was first by Chatt.¹⁵

Figure 43.

Representative tetrahedral Fe(IV) nitride complexes.^{176,187–189}

Regardless of which pathway(s) are viable for overall N₂-to-NH₃ conversion, a linchpin intermediate is the initial product of pronation, M(NNH). Parent diazenidos (sometimes referred to as imides) of this type are very rare and prior to recent work with iron had only been reliably characterized on select Mo and W systems.^{99,136,190,191} Germane to the iron systems being discussed here, efforts to isolate and crystallize a parent Fe(NNH) species have been unsuccessful owing to the presumed high reactivity of such species; the N–H bonds are predicted to be very weak for systems such as (P₃^B)Fe (vide infra)^80,192 and accordingly the addition of stoichiometric acid to [(P₃^B)Fe(N₂)]⁻, even at very low temperature, led to net oxidation of the iron complex to (P₃^B)Fe(N₂) with concomitant release of 0.5 equiv of H₂. In the presence of excess acid, the presumed (P₃^B)Fe(NNH) intermediate is trapped by H⁺ to produce ([(P₃^B)Fe(NNH₂)]⁺ faster than it can bimolecularly expel H₂.¹¹³ Silyl diazenidos have proven more stable and have been prepared and crystallographically characterized on (P₃^B)Fe¹⁸¹ (and other ligated Fe species, vide infra).^{172,193–196}

In an effort to slow the bimolecular decay of the (P₃^B)Fe(NNH), Peters and co-workers prepared a (P₃^B)Fe analogue with increased steric shielding by replacing the isopropyl substituents on the phosphines with bulky, electron-rich aryl substituents (3,5-diisopropyl,4-methoxyphenyl) (Figure 44). An anionic N₂ complex of this new ligand (^ArP₃^BFe) was prepared by analogy to the parent system. Although the N₂ ligand in [(^ArP₃^B)Fe(N₂)]⁻ is slightly less activated than the parent system (ν_NN = 1937 cm⁻¹ vs 1905 cm⁻¹), protonation with excess acid and warming led to the observation of NH₃ (~0.25 equiv of NH₃ per Fe), confirming the viability of N₂-to-NH₃ conversion with this platform. Reaction of [(^ArP₃^B)Fe(N₂)]⁻ with 1 equiv of HBAr^F₄, via mechanical mixing of a 2-MeTHF glass, afforded a new S = 1/2 species that was characterized via CW and pulse EPR spectroscopy. Using a range of isotopically labeled derivatives, and also via comparison with a more stable isoelectronic and isostructural silyldiazenido analogue, (^ArP₃^B)Fe(NNSiMe₃), a definitive assignment of the species as (^ArP₃^B)Fe(NNH) could be made. The distal proton showed strong coupling to the unpaired spin (a_iso(¹H) = 16.5 MHz). Consistent with prior results,¹¹³ annealing these samples to −78 °C led to decay of the signature for (^ArP₃^B)Fe(NNH) and (primarily) the formation of (^ArP₃^B)Fe(N₂), presumably via loss of H₂. However, annealing such a sample in the presence of excess acid showed its gradual evolution to [(^ArP₃^B)Fe(NNH₂)]⁺; the latter features similar hyperfine parameters to the previously characterized species [(P₃^B)Fe(NNH₂)]⁺.¹¹³

Figure 44.

Comparison of reactivity of (A) [(P₃^B)Fe(N₂)]⁻ and (B) [(^ArP₃^B)Fe(N₂)]⁻ with acid. The bottom figure was taken from ref 177 and shows the reaction progress of [(^ArP3^B)Fe(N₂)]⁻ with excess HBAr^F₄ in 2-Me-THF, as monitored by CW X-band EPR. Simulations for discrete species are shown in red. Spectra show the progression from (A) pure [(^ArP₃^B)Fe(N₂)]⁻. (B) Mixture of[(^ArP₃^B)Fe(N₂)]⁻ and (^ArP₃^B)Fe(NNH), on addition of 1 equiv of HBAr^F₄ for 15 min at 138 K. (C) Same as in spectrum B, but instead with 2.3 equiv of HBAr^F₄ present, showing full conversion to (^ArP₃^B)Fe(NNH). An identical spectrum was obtained on mixing for 30 min in the presence of 1 equiv of HBAr^F₄. (D) Reaction mixture from trace C, after warming to 195 K for 30 s and then rapidly freeze quenching in liquid N₂. Trace shows a mixture of (^ArP₃^B)Fe(NNH) and [(^ArP₃^B)Fe(NNH₂)]⁺. (E) Reaction mixture from trace D after warming to 195 K for 90 s, showing complete conversion to [(^ArP₃^B)Fe(NNH₂)]⁺. Reactions with up to 5 equiv of HBAr^F₄ provided identical spectra. Acquisition parameters for all spectra: temperature = 77 K; MW frequency = 9.44 GHz; MW power = 6.44 mW; modulation amplitude = 0.1 mT; conversion time = 5.12 ms·E.^113,177 EPR spectrum reproduced with permission from ref 176. Copyright 2017 American Chemical Society.

This study completed the characterization of early (and highly reactive) nitrogen fixation intermediates on tris(phosphine)borane Fe systems (Figure 45). Combined with the earlier studies on late-stage nitrogen fixation intermediates,¹⁸² many (P₃^B)Fe(N_xH_y) intermediates of a plausible N₂ fixation cycle are known (Figure 41). Work remains to generate and characterize a few intermediates in the middle of the cycle (shown in light gray; e.g., (P₃^B)Fe(NH)) and to study their respective interconversion to later stage species. Of special interest is to further explore the possibility of hybrid mechanisms wherein distal intermediates convert to alternating intermediates that can then undergo a later-stage N–N cleavage event, indicated by dashed line and discussed further below.

Figure 45.

A proposed N₂RR cycle for Peters’ (P₃^B)Fe platform. Characterized compounds are shown in green and unobserved complexes in gray. Solid arrows indicate a distal mechanism of nitrogen fixation, while dashed arrows indicate hybrid mechanisms of nitrogen fixation.^{109,112–114,176,182} Note that steps are shown as discrete e⁻ and H⁺ transfer steps, though concerted PCET steps may also be operative.

5.2. (P₃^B)Fe: More Efficient Catalysis at Lower Overpotential

In 2017, Peters and co-workers found that the reaction of [(P₃^B)Fe]⁺ at −78 °C in Et₂O under an N₂ atmosphere with excess Cp*₂Co and [Ph₂NH₂]OTf led to significantly enhanced efficiency for NH₃ production (up to 72 ± 3%, along with trace amounts of N₂H₄) (Figure 46).⁸⁰ This result was fascinating as it demonstrated the overall efficiency could be improved with reagents that provide significantly less net driving force for the overall transformation (ΔΔH_f= 77 kcal·mol⁻¹ NH₃ vs ≥156 kcal·mol⁻¹ NH₃ with KC₈ and HBAr^F₄). Also noteworthy, freeze-quench Müssbauer spectroscopy of the catalytic reaction under these new conditions failed to show any of the borohydride/hydride species, (P₃^μ-B-H)Fe(H)(L) (L = H₂ or N₂), which, as discussed above, dominated the speciation of the catalyst at early times when using KC₈ and HBAr^F₄. However, high-spin Fe^II species (presumably decomposition products) are formed, which were posited to be a consequence of the presence of a more coordinating triflate counteranion. Furthermore, the observation of [(P₃^B)Fe(N₂)]⁻ at early reaction times by freeze-quench Müssbauer and EPR spectroscopy suggested that protonation under these new conditions is slow, consistent with the highly insolubility of the acid under the catalytic conditions. However, employing a less coordinating and more soluble acid, [Ph₂NH₂]BAr^F₄, led to poorer catalytic performance (42 ± 6% efficiency). These data suggest that acid strength, concentration, and counteranion all play roles in determining the selectivity for NH₃ versus H₂.

Figure 46.

Comparison of performance for N₂RR by [(P₃^B)Fe]⁺ depending on the acid/reductant.^18,80,173 For discussion of overpotential see Section 8.3.

In a follow-up study, the effect of the pK_a on the selectivity for NH₃ versus H₂ was studied by using variably substituted anilinium triflate acids.¹⁹⁷ Notably, despite the known importance of pH on the selectivity of the MoFe protein,¹⁹⁸ the effect of pK_a on selectivity had not been previously addressed with a model system. The ease of pK_a tunability with anilinium acids, and their compatibility with N₂RR catalysis with [(P₃^B)Fe]⁺, made such a study possible. NH₃ selectivities greater than 70% were observed with acids that have pK_a’s in THF between 2 and 4. Moving to stronger acids decreased catalytic performance, while weaker acids showed a steady decline in performance, with only trace NH₃ being observed with [^4-MeOPhNH₃]OTf (pK_a = 8.8 in THF). This behavior is in fact quite similar to that observed for the MoFe protein itself.¹⁹⁸

Efforts to protonate [(P₃^B)Fe(N₂)]⁻ at low temperature with [^2,6-ClPhNH₃]OTf, which is very effective in catalysis, did not lead to productive N–H bond formation at −78 °C in Et₂O, contrasting with the aforementioned results with HBAr^F₄, where [(P₃^B)Fe(NNH₂)]⁺ is generated. Using the more soluble acid [^2,6-ClPhNH₃]BAr^F₄ enabled low temperature protonation to be observed (i.e., formation of [(P₃^B)Fe(NNH₂)]⁺) and the observation of NH₃ upon warming.¹¹³ These data suggest that the mechanism when soluble (BAr^F₄⁻) acids are employed (maximum efficiency of ~45%) is likely different than when insoluble (OTf⁻) acids (maximum efficiency of ~75%) are employed.

To explain the lack of productive N–H bond formation in the absence of reductant and insoluble acids, it was suggested that under these conditions, PCET mechanisms become relevant, whereas with HBAr^F₄ and KC₈ proton transfer-electron transfer (PT-ET) mechanisms likely dominate. Discussion of PCET mechanisms for N–H bond formation in the context of homogeneous nitrogen fixation catalysis had previously been limited to reactions mediated by pyridinyl radicals;^103,105,106 as such mechanisms were not relevant to this work, the authors explored the possibility of metallocene protonation to generate PCET reagents. Using DFT calculations, the authors demonstrated that even with pyridinium acids, reduction of the pyridinium rings is less favorable than proton transfer from the pyridinium to the metallocene reductant.⁸⁰ The relevant metallocenes (Cp₂Co, Cp*₂Cr, and Cp*₂Co) are all predicted by DFT to undergo ring-protonation to generate species that have homolytically very weak C–H bonds (BDE_C–H ≈ 30 kcal·mol⁻¹, Figure 47), suggesting they should behave as competent PCET donors.^80,197

Figure 47.

Bond dissociation enthalpies of the thermodynamically preferred protonation isomer for three commonly used reductants in N₂RR (Cp₂Co, Cp*₂Cr, and Cp*₂Co). The site of protonation and the C–H bond dissociation enthalpy of the resulting product has been confirmed experimentally for Cp*₂Co.^80,199

If protonation of the metallocene is critical for N–H bond formation under these types of conditions, then one would expect it to have a marked effect on NH₃ selectivity. As such, DFT was used to interrogate the ability of three different acids to protonate Cp*₂Co ([^2,6-ClPhNH₃]OTf, high experimental efficiency for N₂RR; [^2,6-MePhNH₃]OTf, moderate efficiency; [^4-OMePHNH₃]OTf, low efficiency). Although all of these reactions were predicted to occur with a low barrier, the reaction was found to be uphill with the least efficient acid, downhill with the most efficient acid, and thermoneutral for the moderately efficient acid. Thus, the experimental catalytic efficiency data correlate with the predicted availability of the protonated metallocene.¹⁹⁷

Pulse EPR spectroscopy was recently used to experimentally verify that Cp*₂Co undergoes both endo- and exoring protonation to form [(Cp*)(η⁴-C₅Me₅H)Co]⁺. Furthermore, thermochemical measurements confirmed that the exoisomer possessed a homolytically very weak C–H bond (BDFE_C–H < 29 kcal·mol⁻¹), consistent with earlier computational predictions (Figure 48).⁸⁰ This experimental work also unveiled that the protonated metallocene, [(Cp*)(η⁴-C₅Me₅H)Co]⁺, can be reduced by one electron at a very mild potential (−0.62 V vs Fc^+/0) to generate (Cp*)(η⁴-C₅Me₅H)Co; the latter features a very hydridic C–H bond (ΔG(H⁻) < 41 kcal·mol⁻¹). This result hints at the possibility that hydride (H⁻) transfer might also play a role in catalysis, an idea that warrants further study.

Figure 48.

Thermochemistry of protonated Cp*₂Co species that have been hypothesized to be relevant to PCET and hydride transfer in N₂RR.¹⁹⁹

Given the ubiquity of metallocenes in high efficiency N₂-fixation catalysis, such as that of the Mo systems discussed above,^50,51,84,149 it seems plausible that metallocene-mediated N–H bond formation, via PCET and/or hydride transfer processes, could be an important and previously overlooked pathway in N₂RR. The very recently reported catalytic results from the Nishibayashi group utilizing the known PCET donor mixture SmI₂ and H₂O/ethylene glycol offers added support to this suggestion.⁷⁹

5.3. A Catalyst with an Fe–C Bond

In 2014, the Peters group reported a related Fe catalyst for nitrogen fixation featuring a carbon-anchored tris(phosphine) atrane ligand, (P₃^C)Fe.¹¹⁵ This complex was of particular interest to be able to explore a model catalyst in which the N₂ substrate and a C atom are trans-disposed. Such an arrangement has been proposed upon initial binding of N₂ to either Fe₂ or Fe₆ of the FeMo-cofactor in MoFe-nitrogenase, which features an interstitial carbide.^13,200 The possibility that hemilability of the interstitial light atom could play a key role in N₂-binding at Fe in the cofactor was suggested by the Peters group as early as 2005¹⁶⁸ and was the subject of an in depth iron model study with a tris(phosphino)alkyl ligand, C^SiP₃Ph, in 2013 (Figure 49).¹⁹⁴ The 2013 study demonstrated that the Fe–C_apical in the model complex is highly sensitive to the overall charge (see the [(C^SiP₃Ph)Fe(CO)]ⁿ⁺ complexes (Fe–C = 2.303 Å in the anion, 2.236 Å in the neutral, and 2.138 Å in the cation). Indeed, the anionic charge state of both the CO and N₂ complexes feature unusually long Fe–C_apical bonds. Although the anionic N₂ complex is silylated to form (C^SiP₃Ph)Fe(NNSiⁱPr₃), the system was not catalytically active with the reagents tested at the time (KC₈ and HBAr^F₄).

Figure 49.

Reductive formation of [(C^SiP₃Ph)Fe(N₂)]− and subsequent silylation chemistry.¹⁹⁴

In contrast, the (P₃^C)Fe system showed the ability to bind N₂ across three oxidation states (Figure 50), akin to its isostructural analogue, (P₃^Si)Fe, featuring Si in place of C as the anchoring atom.¹⁹³ As with(C^SiP₃Ph)Fe(CO),¹⁹⁴ the Fe–C bond distance contracted upon oxidation, opposite to what is observed with (P₃^B)Fe complexes.^114,201 Nonetheless, (P₃^C)Fe also showed catalytic N₂RR when exposed to excess HBAr^F₄ and KC₈.¹¹⁵

Figure 50.

Trends in Fe–C bond distance for a redox series of (P₃^C)Fe(N₂) complexes.¹¹⁵ Related datafor [(P₃^B)Fe(N₂)]⁻ Figure 64.

Exposing [(P₃^C)Fe(N₂)]⁻ at −78 °C in Et₂O to excess HBAr^F₄ and KC₈ led to the formation of 4.6 ± 0.8 equiv of NH₃ per Fe. Under these conditions, a significant amount of the iron(II) hydride complex, (P₃^C)Fe(H)(N₂), was observed to form, which itself proved inactive as a precatalyst.¹⁷³ Thus, it was suggested that hydride formation limits the overall turnover. This contrasts with the aforementioned hydride-resting state, (P₃^μ-B-H)Fe(H)(L) (L = H₂ or N₂), of the (P₃^B)Fe-system, which can re-enter the catalytic cycle. These results (and those of other groups)^{50,53,107,116,173,178} highlight the importance of reversibly formed hydride intermediates in the context of the synthetic N₂-fixation catalyst design. Indeed, it may be that the fact that nitrogenase enzymes also display hydrogenase activity essential to their biological function; if hydride intermediates cannot liberate H₂, the active site can get stuck in an inactive hydride state and becomes effectively poisoned.¹⁷³

5.4. Catalysis and Mechanism of a Tris(phosphino)silyl Fe Species

The Peters group has extensively explored the chemistry of a silyl-anchored tris(phosphine)iron system, (P₃^Si)Fe (Figure 51). Indeed, the study of such species predated the discovery of Fe-mediated N₂RR catalysis in 2013.^18,193,202 In many regards, the N₂ activation and functionalization chemistry of the (P₃^Si)Fe-system are very similar to that of (P₃^B)Fe. This includes its demonstrated ability to support N₂ complexes in multiple oxidation states,^193,202 to form complexes with late-stage nitrogen fixation intermediates,¹⁹³ to undergo redox-induced expulsion of NH₃ with concomitant binding of N₂,¹⁹³ and also N₂ silylation¹⁹³ and protonation¹⁰⁹ at the distal N atom of [(P₃^Si)Fe(N₂)]⁻. However, in the presence of excess HBAr^F₄ and KC₈, the (P₃^Si)Fe-system is far more selective for the hydrogen evolution reaction (HER) than for N₂RR (max efficiency for NH₃ 5 ± 3% with HBAr^F₄/KC₈). Nevertheless, under a sufficiently high loading of acid (1500 equiv of HBAr^F₄) and reductant (1800 equiv of KC₈), [(P₃^Si)Fe(N₂)]⁻ can be confirmed to be a catalyst for N₂RR (3.8 ± 0.8 equiv NH₃ per Fe).¹⁷³ A similar lack of selectivity for N₂RR is also observed with [Ph₂NH₂]OTf/Cp*₂Co (1.2 ± 0.1 equiv of NH₃ per Fe, 6 ± 1% NH₃).⁸⁰

Figure 51.

Reactivity of (P₃^Si)Fe hydrazido(2−) complexes.¹⁰⁹

The difference in reactivity between [(P₃^Si)Fe(N₂)]⁻ and its [(P₃^B)Fe(N₂)]⁻ and [(P₃^C)Fe(N₂)]⁻ analogues poses interesting questions regarding the underlying cause of their divergent selectivities. [(P₃^Si)Fe(N₂)]⁻ and [(P₃^B)Fe(N₂)]⁻ have been studied in the most detail and allow for some insightful comparisons. Two key differences between these platforms emerge when interrogating their ability to stabilize Fe–N multiply bonded species. Whereas imido species of the type [(P₃^B)Fe(NR)]^0/+ are highly stable,^{112–114,181} related imides for (P₃^Si)Fe are highly reactive, have only been characterized at very low temperature, and display hydrogen atom abstraction reactivity and imide/nitrene coupling reactivity to generate azoarenes (ArN = NAr).²⁰³ This difference in stability of imido-like species might be pertinent to nitrogen fixation by considering the respective stabilities and reactivity profiles of imide, Fe(NH) and “imido-like” hydrazido(2−) intermediates, Fe(NNH₂). Surprisingly, the hydrazido(2−) complex, [(P₃^Si)Fe(NNH₂)]⁺, was accessible via double protonation of [(P₃^Si)Fe(N₂)]⁻ at low temperature and could be isolated and crystallographically characterized.¹⁰⁹ [(P₃^B)Fe(NNH₂)]⁺ has not proven stable enough to isolate.

One-electron reduction of [(P₃^Si)Fe(NNH₂)]⁺ by Cp*₂Co generated neutral, S = 1/2 (P₃^Si)Fe(NNH₂) as a highly reactive intermediate (Figure 51). Its characterization was aided by the generation of its far more stable (and crystallographically characterized) alkylated analogue, (P₃^Si)Fe(NNMe₂). Addition of 0.5 equiv of Cp*₂Co to [(P₃^Si)Fe(NNH₂)]⁺ led to the generation of 0.5 equiv of [(P₃^Si)Fe(N₂H₄)]⁺ and 0.5 equiv of (P₃^Si)Fe(N₂). These data are consistent with a scenario in which (P₃^Si)Fe(NNH₂) is generated in situ and reacts, via a net transfer of two H atoms, with remaining [(P₃^Si)Fe(NNH₂)]⁺. These data are moreover consistent with the observation that reaction of (P₃^Si)Fe(N₂) with excess Cp*₂Cr and [HNiPr₂Et]BF₄ led to the formation of N₂H₄ in 13% yield.²⁰⁴ One possibility for the reduced selectivity of (P₃^Si)Fe for N₂-to-NH₃ conversion is that the enhanced reactivity at N_α in (P₃^Si)Fe complexes, for example, via hydrogen atom transfer (HAT), can prevent the system from efficiently traversing a distal pathway that appears to be accessible for (P₃^B)Fe.¹⁷⁶

The question of selectivity for NH₃ versus N₂H₄ is of general interest with Fe catalysts. Every known Fe-catalyst (vide infra) has demonstrated an ability to form both NH₃ and N₂H₄, with the primary product varying based on catalyst and condition. However, prior to this report, the formation of an alternating pathway intermediate directly from N₂ was unprecedented for Fe. This led the authors to propose an alternative N₂-fixation mechanism, dubbed the hybrid mechanism because it begins in a distal fashion and then on the third functionalization switches to an alternating pattern, similar to the pathway proposed by Dilworth and co-workers in the context of N₂H₄ formation on VFe nitrogenase.¹⁶³ A recent study of methylhydrazido (2−) intermediates on (P₃^B)Fe may offer some insight into the question of NH₃ versus N₂H₄ selectivity. The enhanced stability of these methylated (vs protonated) species allowed for a redox series (P₃^BFe(NNMe₂)^+/0/−) to be studied with a number of spectroscopic methods (i.e., Müssbauer, XAS, nuclear resonance vibrational spectroscopy, UV–vis spectroscopy), which were complemented by theoretical methods (DFT and complete active calculations). These methods revealed that for [(P₃^B)Fe(NNMe₂)]⁺, the NNMe₂ unit is best described in one of its canonical forms, namely, isodiazene. However, in the neutral and anionic complexes, the ligand is best described as a hydrazyl radical(1−) antiferromagnetically coupled to the Fe center. A hydrazyl radical should be reactive at both N_β and N_α, potentially explaining the source of mixed selectivity (NH₃ and N₂H₄) observed upon (P₃^B)Fe mediated N₂RR in the presence of Cp*₂Co and [Ph₂NH₂]OTf. Spectroscopic measurements suggest that the isoelectronic (P₃^Si)Fe(NNH₂) has a similar electronic structure, consistent with the observed net transfer of protons and electrons reactivity to yield hydrazine.¹¹⁴ This relative distribution of electron density between N_α and N_β may be a significant and general factor controlling NH₃ versus N₂H₄ selectivity and warrants further study.

5.5. Theoretical Work on Atrane-Fe Systems

Phukan and co-workers performed a computational study comparing the potential for (P₃^B)Fe-and (P₃^Si)Fe for N₂RR.²⁰⁵ They correctly predicted a greater preference of the silyl system to undergo protonation at Fe, which would lead to unproductive H₂ formation. They also predicted smaller barriers for productive N₂ functionalization steps on the boratrane platform. However, they suggested that functionalization would occur in an alternating rather than distal pathway, which is inconsistent with current experimental observations.^109,113 Because it is a truncated ligand platform in which the phenylene linkers were replaced with ethylene linkers and the isopropyl substituents with methyl substituents, it is hard to draw reliable connections between this study and experimental observations. In addition to large conformational differences between the theoretically predicted structures and experimental observations due potentially to steric and electronic differences in the ligands, spin states are also systematically overestimated in the study.

In 2017, Peters and co-workers published a computational study evaluating the propensity of (P₃^E)Fe (E = B, C, Si) to release H₂ during catalysis (Figure 52).¹⁹² By contrast to most studies that posit metal hydrides as the source of catalyzed H₂ formation during nitrogen fixation, this study suggested that the source of H₂ might be early nitrogen fixation intermediates, owing to their weak (<50 kcal·mol⁻¹) N–H bonds. In support of this hypothesis, this study found that the N–H bond strengths in (P₃^E)Fe(NNH₂) are correlated with the overall selectivity for NH₃ versus H₂ formation, with complexes that build up and feature weaker N–H bonds giving greater yields of H₂. These bond strengths were correlated with the cumulative bond indices of the E–Fe–N–N unit. The more flexible the Fe–E interaction, the more Fe–N and N–N bonding is preserved as the N₂ is functionalized, and the stronger the resulting N–H bonds. Species with stronger N–H bonds are thus predicted to be less prone to bimolecular H₂ release.¹⁹² Important to note is that this study was almost exclusively based on thermodynamic arguments. Kinetic barriers for specific reaction steps, which can be challenging to reliably predict, especially in the absence of experimental data for calibration, were not computed. There is a need to directly measure some of the kinetic barriers for reaction steps in this Fe-mediated catalysis to better guide theoretical work. Such measurements are likely to be technically very challenging due to the highly reactive nature of the early stage (P₃^B)Fe(N_xH_y) intermediates predicted to influence overall selectivity in this study.

Figure 52.

Overview of predicted bimolecular HER and N₂RR pathways for (P₃^E)Fe(NNH_y) species and estimated BDFE_N–H values. Reproduced with permission from ref 192. Copyright 2018 American Chemical Society.

Li and co-workers have published two recent computational studies on N₂RR at (P₃^E)Fe-complexes (E = B, Si, N). In the first study, they used complete active space (CAS) calculations to interrogate the electronic structures of (P₃^B)Fe(N_xH_y) complexes in detail.²⁰⁶ They found that protonation from the [(P₃^B)Fe(N₂)]⁻ to the Fe(NNH) and then to the Fe(NNH₂)⁺ leads to successive lengthening and weakening of the Fe–B bond. This suggestion had been previously advanced by the Peters group to explain the ability of this platform to support both low-valent (e.g., [Fe(N₂)]²⁻) and high-valent (e.g., [Fe(N)]⁺) intermediates, and corroborated by a wealth of crystallographically characterized Fe–B distances.^{112–114,176,181} Li and co-workers, akin to a suggestion by Peters and co-workers,¹⁷⁶ suggested that this kind of buffering effect could also be relevant to the FeMo-cofactor in nitrogenase, or in heterogeneous catalysts that feature single-metal atoms on supports.

More recently, the Li group has evaluated the relative influence of having apical B, C, or N atoms on the catalytic system by interrogating how the Fe–E bond indices evolve over the catalytic cycle.²⁰⁷ In the case of both B and C, they found significant changes in the bond indices throughout the catalytic cycle. For the anionic C-based ligand, an increased bond order was observed upon protonation, and a decreased bond order was observed following reduction, consistent with experimentally observed bond lengthening in the [(P₃^C)Fe(N₂)]^+/0/− series. By contrast, the neutral borane-based ligand was predicted to show the opposite behavior, consistent with experimentally observed bone lengths.^114,181,201 Furthermore, this study predicted that the interaction between a Lewis basic N atom anchor and Fe would minimal throughout the catalytic cycle and is thereby unable to buffer the oxidation state changes experienced at Fe. However, the “peripheral” ligand, namely, the phosphine donors and the phenylene linkers, play a significant role in buffering the redox changes at Fe. From their studies, the authors suggested that catalysis with an N-anchored ligand might be achieved by installing electron-withdrawing −CF₃ groups at the para-position (relative to the apical N atom linchpin of the ligand), so as to enhance the electronic buffering capacity of the tris(phosphine)amine ligand.²⁰⁷

Visser and co-workers have investigated the thermodynamics and kinetics of [(P₃^B)Fe(N₂)]⁻ protonation by [(Et₂O)₂H]⁺.²⁰⁸ Consistent with spectroscopic results, double protonation at the distal nitrogen was found to be kinetically and thermodynamically more favorable than formation of a diimide-type intermediate (e.g., (P₃^B)Fe(η²-HNNH)). They then determined that protolytic N–N bond cleavage should be possible following either 0, 1, or 2 electron transfers, suggesting the thermodynamic plausibility of multiple mechanisms for NH₃ formation. Unfortunately, this report systematically overestimates the stability of high spin states (e.g., they predict that the S = 3/2 state of [(P₃^B)Fe(NNH₂)]⁺ to be more stable than the experimentally observed S =1/2 ground state by 20 kcal·mol⁻¹). This makes it difficult to draw firm conclusions from the work.

Oláh, Szilvási, and co-workers have interrogated the Gibbs free energy profile of nitrogen fixation catalysis on (P₃^E)Fe (E = B, Si).²⁰⁹ They identified that these systems could access both a distal and a hybrid distal-to-alternating mechanism, with the key intermediate being (P₃^E)Fe(NNH₂). Consistent with experimental results,^{109,114,176,204} the Si system was predicted to have a greater preference for the hybrid distal-to-alternating pathway. Furthermore, this theoretical study identified [(P₃^Si)Fe(N₂H₄)]⁺ as a thermodynamic sink that is difficult to escape once formed (i.e., unfavorable electron transfer, proton transfer, and Fe–N bond cleavage); however, they suggested that PCET to [(P₃^Si)Fe(N₂H₄)]⁺ to form NH₃ and [(P₃^Si)Fe(NH₂)]⁺ could rapidly consume such an intermediate. An identified challenge for these iron catalyst systems is that all of the free energy provided by the reagents is consumed as early as the formation of [(P₃^E)Fe(NH₃)]⁺, which precedes catalyst turnover. A similar problem has been identified computationally and experimentally for Schrock’s (HIPTN₃N)Mo system,¹²⁸ suggesting this maybe a general challenge in homogeneous N₂ fixation catalysts, offering a potential area for improvement via rational catalyst design.

5.6. Nishibayashi’s Fe Systems

An Fe catalyst was developed by the Nishibayashi group featuring a pyrrole-based PNP ligand.¹¹⁶ Although no fixed-N products were observed when they used conditions similar to those developed for their Mo catalysts (i.e., Cp₂Co, [LutH]OTf, room temperature),⁵¹ when they explored the original conditions reported for (P₃^B)Fe (i.e., KC₈, HBAr^F₄, −78 °C in Et₂O)¹⁸ they detected catalytic yields of fixed-N products. In contrast to the (P₃^B)Fe system, Nishibayashi and co-workers observed catalytic yields of not only NH₃ but also some N₂H₄. Furthermore, the selectivity for NH₃ versus N₂H₄ was affected by the solvent, with THF enhancing the yield of N₂H₄ relative to Et₂O. The electronic structure of the most efficient precatalyst, (PNP)Fe(N₂), was studied by Walter and co-workers via SQUID magnetometry, Mössbauer, and EPR spectroscopy and found to feature a well-isolated S = 1/2 ground state.²¹⁰

Notably, the Nishibayashi group found that two potential products of (PNP)Fe(N₂) protonation, (PNP)Fe(H) and a species resulting from protonation at the pyrrole-backbone (which DFT predicted to be the thermodynamic product of protonation), were also precatalysts for N₂RR, albeit with diminished efficiency (Figure 54). Again, this observation underscores the importance of forming hydridic species, in this case either Fe(H) or C–H of the dearomatized pyrrole, reversibly if N₂RR catalysis is to be productive, thereby avoiding an off-path sink.

Figure 54.

Catalytic N₂RR efficiencies for all of Nishibayashi’s (PNP)Fe systems. aThese complexes were only tested under higher loading conditions with KC₈ (200 equiv) and HBAr^F₄ (184 equiv).^29,116,178

Given the diminished efficiency of the pyrrole-protonated precatalyst, the Nishibayashi group synthesized two new catalysts that feature methyl and phenyl substitution at the 3- and 4-positions of pyrrole (Figure 54).¹⁷⁸ While phenyl substitution did not lead to improved catalysis, methyl substitution improved the catalyst lifetime (max fixed-N product: 26.1 ± 2 vs 17.9 ± 2; max efficiency: 42 ± 8% vs 37 ± 4%).^116,178 Markedly, reaction of these (^3,4-MePNP)Fe(N₂) complexes with HBAr^F₄ resulted in protonation at the 2-position of the pyrrole backbone rather than the 3-position. Thus, methyl substitution prevented protonation at the 3-and 4-position, but the electron-rich pyrrole ring remained the thermodynamically favored site of protonation.¹⁷⁸

Efforts to completely prevent pyrrole ligand protonation by replacing the pyrrole with a carbazole were unsuccessful (Figure 54).¹⁷⁹ Reduction of the carbazole-containing (^CarPNP)Fe(Cl) species led to bridged, rather than terminal dinitrogen complexes, which may retard protonation steps at N₂.²¹¹ Possibly as a consequence of slower kinetics, these N₂-bridged species were not catalysts for N₂RR.

5.7. A Non-Phosphine Fe System for Catalysis

The low coordination numbers of the iron centers in the FeMo-cofactor, especially prior to the reassignment of its structure with an interstitial light atom,¹³ when the iron centers were presumed to be three-coordinate has inspired inorganic chemists.^212,213 Holland and co-workers, in particular, have advanced the hypothesis that low-coordinate Fe centers would be particularly competent for activating N₂. Although Holland and co-workers have not yet demonstrated catalytic N₂RR with their β-diketiminate system, they have demonstrated important results regarding N–N bond cleavage and N–H bond formation on this platform.^{90,211,214–217}

However, to date, the only iron system featuring coordination numbers less than four reported to facilitate catalytic N₂-to-NH₃ conversion is the 2-coordinate complex (CAAC)₂Fe (CAAC = cyclic alkyl amino carbene, Figure 55), which goes three-coordinate upon N₂ binding.¹⁷² Interestingly, (CAAC)₂Fe is stable to reduction at room temperature, but entropy-controlled equilibrium N₂ binding (K_eq = 0.2 at 298 K) to (CAAC)₂Fe facilitated reduction at low temperature to [(CAAC)₂Fe(N₂)]⁻ (ν_NN = 1,850 cm⁻¹). The low N₂ stretching frequency correlates with strong N₂ activation. Accordingly, [(CAAC)₂Fe(N₂)]⁻ reacted productively with silyl electrophiles to yield the silyldiazenido complex (CAAC)₂Fe(NNSiMe₃). Consistent with temperature-controlled binding of N₂ (and hence reduction), the reaction temperature had a significant impact on catalytic efficiency when (CAAC)₂Fe was exposed to excess HBAr^F₄ and KC₈ (0.4 equiv of NH₃ at RT, 0.3 equiv of NH₃ at −50 °C, 0.9 equiv of NH₃ at −78 °C, 3.3 equiv of NH₃ at −95 °C; average values given).

Figure 55.

Catalytic conditions for (CAAC)₂Fe. Equilibrium binding of N₂ at −80 °C and access to a (CAAC)₂Fe(NNTMS) via reductive silylation; Dipp = 2,6-diisopropylpheyl.¹⁷²

5.8. Historic Work on Bis-Diphosphine Fe Complexes in Nitrogen Fixation

In 2016, the Ashley group demonstrated the catalytic formation of hydrazine (N₂H₄) using (depe)₂Fe(N₂) as the (pre)catalyst (depe = 1,2-bis(diethylphosphino)ethane).¹⁷⁴ The stage for this result was set by the rich history of bis-diphosphine–Fe-N₂ systems. We first describe early and mechanistically relevant details prior returning to the fascinating catalysis reported by Ashley and co-workers.

Leigh and co-workers first suggested that a dmpe-supported (dmpe = 1,2-bis(dimethylphosphino)ethane) Fe complex, (dmpe)₂Fe(N₂) (ν_NN = 1975 cm⁻¹), liberated fixed-N produts (up to 20%) upon treatment with H₂SO₄ (Figure 56). However, efforts to protonate the N₂/hydride species, [(dmpe)₂Fe(N₂)-(H)]⁺ (ν_NN = 2,074 cm⁻¹) at N₂, with H₂SO₄ were largely unsuccessful (<4% NH₃).²¹⁸ Employing HCl as the acid source instead allowed for recovery of the starting material trans-(dmpe)₂Fe(Cl)₂ in about 80% yield, hinting that such systems might ultimately be amenable to catalysis.²¹⁹ Nevertheless, these results were later called into question due to the lack of fixed-N products observed from protonolysis reactions with the structurally related complex (depe)₂Fe(N₂).^220,221 Later work definitively established that protonation of (dmpe)₂Fe(N₂) by HCl does not result in the formation NH₄⁺.²²² However, these results remain important as they provided motivation for the study of nitrogen fixation using these types of complexes. Furthermore, Leigh’s suggestion that complexes of the type: (diphosphine)₂Fe(N₂) could yield fixed-N products was ultimately proven correct by Ashley and co-workers, the key being replacing HCl with HOTf.²²³

Figure 56.

Reactivity of (dmpe)₂Fe (dmpe = 1,2-bis(dimethylphosphino)ethane) complexes.^{218,223–229}

Much of the early mechanistic work on (diphosphine)₂Fe-species relevant to N₂ fixation was performed with the ligands dmpe and DMeOPrPE (1,2-bis[di(methoxypropyl)-phosphino]ethane). Although much of this work was performed contemporaneously, the dmpe system will be discussed first, with distinct features of the DMeOPrPE system then being introduced.

The η²-N₂H₄ hydrazine adduct complex [(dmpe)₂Fe(N₂H₄)]²⁺ was synthesized by reaction of trans-(dmpe)₂Fe(Cl)₂ with excess N₂H₄ in THF (Figure 56).²²⁹ Reaction of trans-(dmpe)₂Fe(Cl)₂ with excess hydrazine in the presence of KC₈ led to the formation of an η²-bound diazene complex, (dmpe)₂Fe(η²-HNNH), in 17% yield. The latter complex is structurally (and electronically) unusual and was proposed to form via reduction of the aforementioned [(dmpe)₂Fe(N₂H₄)^]2+ complex to an (dmpe)₂Fe(N₂H₄) product that then loses H₂ to form the observed Fe-diazene. The observation of cis-(dmpe)₂Fe(H)₂ as a minor byproduct is at least consistent with the in situ formation of H₂. Although the diazene species can be isolated by crystallization, it ultimately decomposed in solution to afford [(dmpe)₂Fe]₂(μ-dmpe) with the loss of the diazene fragment. Consistent with the known decomposition pathways of free diazene, monitoring the decay of the ¹⁵N-labeled diazene complex by NMR spectroscopy confirmed the formation of ¹⁵NH₃ and ¹⁵N₂.

Additional work by Field and co-workers revealed that the diazene complex could be synthesized in improved yield (55%) via deprotonation of the hydrazine complex by KO^tBu (Figure 57).²²⁴ This reaction was reversible with [LutH]OTf able to protonate the diazene complex to the hydrazine complex. Reaction of the ¹⁵N-hydrazine complex with Schlosser’s base (KO^tBu and ^tBuLi) under a ¹⁴N₂ atmosphere led to the formation of (dmpe)₂Fe(¹⁵N₂), as confirmed by NMR and IR spectroscopy, suggesting that under certain conditions, the bound diazene could effectively lose H₂ without dissociating from the metal. DFT calculations performed on the diazene complex led the authors to propose an Fe⁰ oxidation state with backbonding to a neutral diazene fragment, as opposed to charge-transfer to form a hydrazido(2−) ligand bound to an Fe^II center; however, this assignment has been questioned.²³⁰

Figure 57.

(Top) Acid/base chemistry that interconverts hydrazine, hydrazido(1−), and diazene ligands on (DMeOPrPE)₂Fe;^232,233 (bottom) potential N₂RR intermediates of the form (dxpe)₂Fe(H)(L) species (x = m or e, L = H₂, N₂, NH₃, and N₂H₄) can be synthesized under aqueous conditions.^{224,231,234,235}

The potential relevance of Fe(H) intermediates to nitrogen fixation, both within this system and in biological nitrogen fixation, led Field and co-workers to also consider the coordination of nitrogen fixation intermediates (N_xH_y) to Fe(H) species. The reaction of trans-(dmpe)₂Fe(Cl)(H) with excess N₂H₄ in the presence of NaBPh₄ led to the isolation of trans-[(dmpe)₂Fe(N₂H₄)(H)]BPh₄ in moderate yield (36%). Reaction of trans-[(dmpe)₂Fe(¹⁵N₂H₄)(H)]BPh₄ with KO^tBu formed (dmpe)₂Fe(η²-H¹⁵N¹⁵NH) and trace amounts of (dmpe)₂Fe(¹⁵N₂) and (dmpe)₂Fe(H)₂, consistent with previous results.²²⁴ The hydrazine complex was unstable in solution, decomposing to trans-[(dmpe)₂Fe(NH₃)H]BPh₄, trans-[(dmpe)₂Fe(N₂)(H)]BPh₄, and free N₂. The amide complex, trans-(dmpe)₂Fe(NH₂)(H) was also accessible by reaction of trans-(dmpe)₂Fe(Cl)H with NaNH₂ (63% yield). It could be converted into the ammonia complex, trans-[(dmpe)₂Fe(NH₃)H]⁺, via protonation with H₂O or fluorene,²³¹ thus demonstrating that most of the potential intermediates of a N₂RR cycle can be synthesized in the presence of iron hydrides.

The reactivity of (dmpe)₂Fe(N₂) with electrophiles has been recently revisited (Figure 58). Field and co-workers found that reaction with methyl electrophiles (i.e., MeOTf, MeOTs) methylated the Fe center trans to the N₂ ligand. In contrast, silylation followed by quenching with HOTf led to the formation of NH₃ (17 ± 3%), suggesting N₂-based functionalization. Yields of NH₃ could be increased by the addition of KO^tBu and Cp*₂Co (51 ± 2% NH₃).²²² Ashley and co-workers also explored the reactivity of (dmpe)₂Fe(N₂) and found that under a vacuum this species dimerizes to form [(dmpe)₂Fe]₂(μ-N₂) (ν_NN = 1933 cm⁻¹). While acidification of (dmpe)₂Fe(N₂) with HCl formed only trace NH₃, fixed-N yields could be increased by using HOTf (up to 9.1% N₂H₄).²²³ The Ashley group also noted that by cyclic voltammetry (CV), (dmpe)₂Fe(N₂) decomposes upon oxidation to the putative [(dmpe)₂Fe(N₂)]⁺, but for (depe)₂Fe(N₂) this oxidation occurred reversibly at −2.0 V vs Fc^+/0.

Figure 58.

Treatment of (dmpe)₂Fe(N₂) with electrophiles.²²²

The mechanism of nitrogen fixation using (dmpe)₂Fe(N₂) has been studied computationally by the Tyler group.²³⁰ In these calculations, HOTf was the reference acid, and the electrons for N₂RR were provided by oxidation of (dmpe)₂Fe(N₂) to [(dmpe)₂Fe]²⁺. Protonation of (dmpe)₂Fe(N₂) was found to be most favorable at Fe (ΔE = −58 kcal·mol⁻¹), but it is known that the reaction of trans-(dmpe)₂Fe(N₂)(H) with protons did not lead to the formation of fixed-N products, so they considered only the further reactivity of the kinetic protonation product [(dmpe)₂Fe(NNH)]⁺ (ΔE = −18 kcal·mol⁻¹). Their key observation was that although protonation of [(dmpe)₂Fe(NNH)]⁺ to give the dicationic trans-diazene (HNNH) and isodiazene (NNH₂) species were similar in energy, ΔE = −2 and −6 kcal·mol⁻¹, respectively, the two-electron reduction of the isodiazene was predicted to be much more unfavorable than that of the trans-diazene (ΔE = 18 kcal·mol⁻¹ vs −7 kcal·mol⁻¹). Thus, a Chatt-type mechanism was considered unlikely; this is consistent with the ultimate discovery that (depe)₂Fe selectively catalyzes the formation of N₂H₄.¹⁷⁴

Although the core bis-phosphinoethane architecture is maintained between dmpe and DMeOPrPE, certain features are distinct to the alkyl-ether containing phosphine substituents that warrant comment. With the DMeOPrPE ligand, one can envision the possibility of secondary-sphere hydrogen bonding interactions between N_xH_y intermediates and the ethereal O-substituents of the ligands—[(DMeOPrPE)₂Fe(N₂H₄)]²⁺ demonstrates one such interaction (Figure 57). It is likely that similar interactions exist in other nitrogen fixation intermediates on this platform, but crystallographic evidence is limited. Reaction of this hydrazine complex with excess HOTf leads to N₂H₅⁺ (64%), NH₄⁺ (21%), and N₂. The authors attributed the formation of NH₄⁺ to hydrazine disproportionation.²³² It is possible that hydrogen bonding plays a key role in stabilizing intermediates such as the hydrazido(1−) on this platform, allowing for the reversible, stepwise interconversion of the hydrazine and diazene complexes.²³³

Another interesting feature of the DMeOPrPE system is that the alkyl ether substituents render the complexes water-soluble.^236–238 Unfortunately, although trans-[(DMeOPrPE)₂Fe(N₂)(H)]⁺ could be formed, the hydride was too basic to be deprotonated in aqueous solvent and thus access to the key nitrogen-fixing intermediate, (DMeOPrPE)₂Fe(N₂), was not available.²³⁹ These results are nonetheless of interest as being able to perform catalytic nitrogen fixation under aqueous conditions remains an important goal for the field; recent results from the Nishibayashi group have begun to address this issue.^79,150

Tyler and co-workers have also shown that with (DMeOPrPE)₂Fe, H₂ can serve as a source of electrons for nitrogen fixation. This was achieved by synthesizing trans-[(DMeOPrPE)₂Fe(H)(H₂)]⁺ from reaction of trans-(DMeOPrPE)₂Fe(Cl)₂ with H₂ in the presence of base. Substitution of the H₂ ligand for N₂ occurred upon exposure to an atmosphere of N₂, and deprotonation of the resulting trans-[(DMeOPrPE)₂Fe(H)(N₂)]⁺ by KO^tBu generated (DMeOPrPE)₂Fe(N₂) (Figure 57). Reaction of the latter species with acid provided fixed-N products, for which the reducing equivalents were derived from H₂.²⁴⁰ Despite exciting advances by a variety of groups aimed at the goal of utilizing H₂ as the source of both protons and electrons in catalytic nitrogen fixation (akin to Haber–Bosch),^{88,185,241,242} this overall target remains unrealized.

5.9. Achieving Catalysis with (depe)₂Fe(N₂)

Although much of the mechanistic work concerning bis-diphosphine systems has been performed with dmpe and DMeOPrPE, depe can also be used to form related N₂ complexes.^220,243 Given its enhanced selectivity for the formation of fixed-N products on protonation (26% yield vs 9% yield for dmpe) and the reversibility of its 0/1+ redox couple, (depe)₂Fe(N₂) was pursued as a (pre)catalyst for nitrogen fixation. The use of Cp*₂Co and [Ph₂NH₂]OTf in diethyl ether at −78 °C was found to be particularly effective for the catalytic fixation of N₂ (up to 24.5 ± 0.5 equiv of N₂H₄, Figure 59). The use of THF as the reaction solvent, [Ph₂NH₂]BAr^F₄ as the acid, or KC₈ as the reductant was all found to be deleterious.

Figure 59.

Catalytic N₂ fixation by (depe)₂Fe(N₂) using Cp*₂Co (270 equiv) and [Ph₂NH₂]OTf (360 equiv).¹⁷⁴

Ashley and co-workers made several observations during this work that led them to hypothesize that NH₃ was forming via a catalyst decomposition pathway. First, although some NH₃ forms irrespective of acid loading, it never exceeded 1 equiv (relative to catalyst). Second, the catalyst was not active for further nitrogen fixation upon reloading of substrate, but most of the quantified NH₃ (~ 80%) originated from an unidentified Fe-species that was in solution at the end of catalysis. Lastly, under the reaction conditions, N₂H₄ was rapidly protonated to form insoluble [N₂H₅]OTf, which control reactions demonstrated was stable under the relevant conditions. These data suggested that NH₃ cannot originate from the same pathway as N₂H₄. The question of whether there are different operative pathways to fixed-N products and what the consequence of that is for selectivity/turnover number is worthy of further interrogation.

Ashley and co-workers also suggested that [(depe)₂Fe(N₂H₄)](OTf)₂ cannot be an on-path intermediate of N₂RR due to its poor catalytic performance. However, (pre)catalyst insolubility¹⁷³ and the deleterious reactivity of certain intermediates at high concentration⁵⁴ have been previously postulated as reasons why even on-path intermediates can lead to anomalously low yields when employed as (pre)catalysts. Hence, caution in interpreting this observation is warranted.

Subsequently, the reactivity of (depe)₂Fe(N₂) with electrophiles has been studied (Figure 60). In 2017, Szymczak and co-workers interrogated a “push-pull” hypothesis for N₂ activation with this system. They found that (depe)₂Fe(N₂) reacted with a variety of Lewis acids (e.g., boranes and alkali metals) to form persistent adducts that significantly activated the N₂ ligand (Δν_nn ≈ − 50 to −170 cm⁻¹). Solid-state characterization of the tris-(pentafluorophenyl)borane adduct revealed an elongated N–N distance, a shorter Fe–N distance, and an N–N–B angle of 137.0(3)°, all of which suggested a significant interaction between the boron p-orbital and the N–N π* orbital (Figure 60). The ¹¹B NMR and ¹⁵N NMR spectroscopic data confirmed the interaction to be persistent in solution, with the latter also suggesting that there was increased electron density on N_β. The degree of activation of the N₂ ligand was found to correlate with the acceptor number^244,245 of the Lewis acid employed. Interestingly, the B(C₆F₅)₃-adduct was found to react selectively (~95%) with protons at N_β to produce [(depe)₂Fe(NNHBAr₃)]⁺ (BAr₃ = tris(pentafluorophenyl)borane), in direct contrast to the reaction of (depe)₂Fe(N₂) with protons which primarily led to the formation of Fe-hydrides. [(depe)₂Fe(NNHBAr₃)]⁺ has an N–N bond length of 1.252(8) Å and a ν_NN of 1519 cm⁻¹, suggesting it is best formulated as a hydrazido(2−) ligand (Figure 60).²⁴⁶

Figure 60.

Observed reactivity of (depe)₂Fe(N₂) with electrophiles occurred at N_β.^195,246,247 The relationship between this reactivity and the catalytic reactionz remains unclear.

It was also found that (depe)₂Fe(N₂) reacted with TMSCl in the presence of KBAr^F₄ to form a cationic silyldiazenido complex, [(depe)₂Fe(NNSiMe₃)]⁺ (Figure 60). DFT calculations suggested that this species is a good model of the first intermediate of catalytic nitrogen fixation, [(depe)₂Fe(NNH)]⁺.¹⁹⁵ It was subsequently found that one-electron reduction of [(depe)₂Fe(NNTMS)]⁺ caused a disproportionation reaction that resulted in formation of (depe)₂Fe(N₂) and (depe)₂Fe(NNTMS₂).²⁴⁷ These data are consistent with the previous suggestion by the Peters group that the N–E (E = H, SiR₃) bonds in early N₂ fixation intermediates are weak.^80,192 As with the report from the Szymczak group, difunctionalization of the coordinated N₂ ligand was observed at N_β rather than in an alternating fashion, which diverges from the observed formation of an η²-diazene ligand formed via deprotonation of the N₂H₄ complex and calculations.^224,230 Thus, whether protonation reactions uniquely favor an alternating pathway, if the hydrazido ligand (NNH₂) can rearrange to the diazene ligand (HNNH) or if hydrazine is formed via a hybrid mechanism, a possible scenario that has been considered by several groups,^109,163 remains an open question.

5.10. Triphos-Fe Systems

Peters and co-workers recently reported two triphosphine-Fe hydride species, (PP₂)Fe(H)₂ and [(PP₂)Fe(H)](μ-N₂) (PP₂ = bis(o-diisopropylphosphino-phenyl)phenylphosphine) that were (pre)catalysts for N₂-to-NH₃ conversion (Figure 61).¹⁷⁵ Despite only moderate efficiency for ammonia (up to 7.5 ± 0.8% versus HER), these examples highlighted that irradiation of the catalytic reaction mixture can significantly enhance performance (18.1 ± 0.8% efficiency). The enhancement was even more significant at higher loadings of substrate (3000 equiv of HBAr^F₄ and 3600 equiv of KC₈), with photolysis increasing the yields from 24.5 ± 1.2 equiv of per Fe to 66.7 ± 4.4 equiv of per Fe (272%). These results are of particular interest not only to improving catalyst importance, but also to understanding the dihydride/N₂ exchange process suggested to occur at the FeMo cofactor,¹⁹ which can also be triggered by irradiation.¹⁶²

Figure 61.

Pathways for HER, N₂RR, and hypothesized origins of the light enhancement for (PP₂)Fe.¹⁷⁵

Mechanistic experiments were employed to interrogate the origin of light enhancement. One possibility considered was that a mononuclear iron species is on-path for catalysis and that the dinuclear species undergoes photodissociation. However, freeze-quench EPR spectroscopy did not support this hypothesis. Interrogation of the catalytic reaction mixture after warming revealed that the main product (~90%) was (PP₂)Fe(N₂)(H)₂ (Figure 61). Experiments using this dihydride as the (pre)catalyst led to slightly poorer efficiency for NH₃ (5.2 ± 0.2% efficiency), whereas photolysis of such reactions recovered the same catalytic activity (8.9 ± 0.9 equiv per Fe, 17.8 ± 1.8% efficiency). This suggested that perhaps these species (i.e., dinuclear [(PP₂)Fe(H)](μ-N₂) and (PP₂)Fe(N₂)(H)₂ proceed via a unified mechanism. It had been previously demonstrated that photolysis of (dmpe)₂Fe(H)₂ leads to H₂ elimination and N₂ binding.^225–228 Similarly, low-temperature irradiation of (PP₂)Fe(N₂)(H)₂ also released H₂, leading to N₂ binding.¹⁹⁶

Efforts to detect protonated N₂ species using these Fe(H) species proved unsuccessful. Instead, it seemed that such species might be involved in a parallel, unproductive hydrogen evolution cycle. Furthermore, (PP₂)Fe(N₂)₂ was prepared and underwent protonation at Fe (and not N₂) to product iron-bound hydride species. However, (PP₂)Fe(N₂)₂ could be reduced by KC₈ to structurally unusual monoanionic ([(PP₂)Fe(N₂)]⁻) (Figure 62) and dianionic ([(PP₂)Fe(N₂)]²⁻)complexes. [(PP₂)Fe(N₂)]²⁻ underwent facile silylation to give the stable anionic diazenido complex, [(PP₂)Fe(NNSiMe₃)]⁻. However, oxidation of this latter species, or reaction of [(PP₂)Fe(N₂)]⁻ with trimethylsilyl chloride, formed an unstable, neutral diazenido complex (presumed to be (PP₂)Fe(NNSiMe₃)) that was characterized at low temperature by EPR spectroscopy. Over time, this complex decayed to form P₃Fe(N₂)(SiMe₃), via net silyl migration. This result highlights that migration steps that shift the electrophile to the metal can occur even when N₂ is the kinetic site of functionalization. A similar issue has been discussed for (HIPTN₃N)Mo.⁹⁹

Figure 62.

N₂ silylation reactions have been demonstrated with (PP₂)Fe. The migration of SiMe₃ from N_β to Fe is particularly noteworthy.¹⁹⁶

The nitrogen fixation chemistry of a related triphos-Fe system (cyclohexyl rather than isopropyl substituents on phosphorus) has been reported by Mézailles and co-workers (Figure 63). Notably, they found that an Fe^I hydride species was not stable (with isopropyl an N₂-bridged dinuclear Fe^I hydride species forms), but rather that it disproportionated to form an Fe^II dihydride and Fe⁰ decomposition products. This highlights the effect that subtle changes in the ligand can have on species of potential relevance to N₂RR. Nonetheless, they found similar efficiencies for nitrogen fixation as did the Peters group under analogous conditions (200 equiv of KC₈ and 200 equiv of HBAr^F₄ - up to 3.6 ± 0.3 equiv of NH₃ per Fe, 5.4 ± 0.4% efficiency) and also note that the dihydride complex showed slightly poorer activity (2.7 ± 0.3 equiv of NH₃ per Fe, 4.1 ± 0.4% efficiency).¹⁸⁰

Figure 63.

Nitrogen fixation using a cyclohexyl-substituted tris(phosphino) (PP₂)Fe complex.¹⁸⁰

6. BEYOND Mo-AND Fe-PROGRESS TOWARD N₂-to-NH₃ CATALYSIS USING OTHER METALS

The catalytic systems described thus far have featured either Mo or Fe, and it is these two metals, and the extensive studies that have been undertaken to explore the mechanisms by which they fix N₂, that have largely driven this field of catalysis forward. While Mo and Fe are known to be the biologically relevant metals in MoFe-nitrogenase, there is no reason a priori to anticipate that these should be the only metals able to mediate catalytic N₂-to-NH₃ conversion. Instead, catalytic studies focusing first on Mo and then Fe have played a dominant role in the advance of the field due to many inorganic chemists primarily rooting their explorations in the context of modeling biological nitrogenases.^{26,57,171,248,249} With the growing availability of different acid/reducing equivalent cocktails, conditions, and methods for analysis, it has become possible to systematically screen the efficacy of other transition metal systems. As a result, in the last five years exciting examples of N₂-to-NH₃ conversion catalysis by a number of metals, most of which are not biologically relevant (the exception being V), have been reported. As this occurs, design principles essential to this type of N₂-to-NH₃ conversion catalysis are beginning to emerge.

6.1. Evidence for Catalysis with a Tris(phosphino)borane Co System

As a complement to the Fe model work done with the P₃^B, P₃^C, and P₃^Si ligands, the N₂RR capacity of the corresponding Co complexes has been explored (Figure 64). In all three cases, Co(N₂) complexes could be prepared. An anionic [Co(N₂)]⁻ complex was only accessible for the (P₃^B)Co-system, which correlates with the extra valence electron provided by Co compared to Fe (i.e., [(P₃^B)Co(N₂)]⁻ is an 18-electron complex, whereas [(P₃^Si)Co(N₂)]⁻ and [(P₃^C)Co(N₂)]⁻ would be 19-electron complexes).^52,202,250 Despite the extra valence electron in [(P₃^B)Co(N₂)]⁻, the [(P₃^B)Co(N₂)]^0/− reduction potential was found to be anodically shifted by ~200 mV compared to [(P₃^B)Fe(N₂)]^0/− (Figure 64), suggesting the reduction potential is largely influenced by overall charge. Compared to the isoelectronic, [(P₃^C)Fe(N₂)]⁻ and [(P₃^Si)Fe(N₂)]⁻ (ν_NN for E = C, 1905 cm⁻¹; E = Si, 1920 cm⁻¹), the N₂ ligand in [(P₃^B)Co(N₂)]⁻ (ν_NN = 1978 cm⁻¹) is significantly less activated, consistent with poorer backbonding correlated with the greater electronegativity and less diffuse dorbitals for Co compared to Fe. When [(P₃^B)Co(N₂)]⁻ was subjected to excess HBAr^F₄ and KC₈ in Et₂O at −78 °C, greater than stoichiometric amounts of NH₃ (2.4 ± 0.3 equiv of NH₃ per Co) were produced. While inefficient, any NH₃ yield reliably greater than 2.0 per Co confirms that catalysis is operative; the NH₃ yields were carefully reproduced to ensure this was the case. Hence, this study established the first example of a synthetic catalytic N₂RR system featuring a nonbiologically relevant metal. The NH₃ yields using [(P₃^B)Co(N₂)]⁻ under these conditions fall between those observed when using [(P₃^C)Fe(N₂)]⁻ or [(P₃^Si)Fe(N₂)]⁻ (4.6 ± 0.8 equiv of NH₃per (p₃^c)Fe and 0.8 ± 0.5 equiv of NH₃ per (P₃^Si)Fe) under the same conditions. The neutral derivatives (P₃^C)Co(N₂) and (P₃^Si)Co(N₂) gave ≤0.1 equiv of NH₃ per Co under the same conditions, consistent with the general trend that access to an anionic [(P₃^e)M(N₂)]⁻ for these atrane scaffolds correlates with N₂RR catalysis. The catalytic efficiency of these same Co complexes has also been surveyed using excess Cp*₂Co and [Ph₂NH₂]OTf (Figure 64). With these reagents, [(P₃^b)Co(N₂)]⁻ yielded essentially the same amount of NH₃ as [(P₃^Si)Fe(N₂)]⁻ (1.1 ± 0.4 equiv of NH₃ per Co vs 1.2 ± 0.1 equiv of NH₃ per Fe). Once again, neutral (P₃^Si)Co(N₂) proved ineffective for mediating catalytic nitrogen fixation (0 equiv of NH₃) under these conditions.⁸⁰

Figure 64.

Electrochemical and spectroscopic comparison of [(P₃^B)Co(N₂)]⁻ and [(P₃^B)Fe(N₂)]⁻. Maximum N₂RR efficiencies are compared using KC₈/HBAr^F₄ and [Ph₂NH₂]OTf/Cp*₂Co.^52,80,173

These results, when considered collectively, suggest that the total electron count in the E–M linkage plays a key role in determining catalytic efficiency ({(P₃^B)Fe}⁸ ≫ {(P₃^C)Fe}⁹ ≈ {(P₃^B)Co}⁹ ≈ {(P₃^Si)Fe}⁹ ≫ {(P₃^Si)Co}¹⁰ = {(P₃^C)Co}¹⁰). This observation suggests the possibility that the more electron-deficient systems, such as {(P₃^Si)Mn}⁸ or {(P₃^C)Mn}⁸, could be promising targets for catalytic nitrogen fixation. Such systems have yet to be explored. Also, in accord with the observation that access to an overall anionic N₂ complex is critical for catalysis, DFT-calculated charge maps show that in [(P₃^B)Co(N₂)]⁻ significantly more negative partial charge (δ) was localized on N_β (δ = −0.51) than in its formally isoelectronic but neutral counterpart (P₃^C)Co(N₂) (δ = −0.39).⁵²

6.2. Improved Catalysis with (PNP)Co

Nishibayashi and co-workers have also studied Co-based N₂RR catalysis (Figure 65)⁵³ using the same bis-(phosphino)pyrrole platform they exploited previously for Fe (Figure 54).¹¹⁶ They found that stirring (PNP)Co with excess HBAr^F₄ and KC₈ at −78 °C in Et₂O led to N₂ fixation efficiencies similar to the Fe congener, both at lower (33 ± 1% NH₃ per Co vs 37 ± 4% NH₃ per Fe) and higher (max fixed-N Yield = 17.9 ± 1.0 for Co and 17.9 ± 0.8 for Fe) loadings of acid and reductant. The use of other solvents (e.g., methyl tert-butyl ether and THF) or reductants (e.g., K, Cp*₂Co) led to suboptimal results. The authors argued that the significant improvement in N₂RR by this (PNP)Co catalyst system as compared to [(P₃^B)Co(N₂)]⁻ points to an advantage of moving to a lower valence electron count system. (PNP)Co(N₂) is a 16-electron complex (excluding π-donation from the pyrrole ring), and hence access to its anion [(PNP)Co(N₂)]⁻, and possibly reduced downstream intermediates, may occur more readily than for the [(P₃^B)Co(N₂)]⁻ system. Clearly, other factors may also be at play.

Figure 65.

Transformation of (PNP)Co(H) to (PNP)Co(N₂) via protonation, and their respective competency as precatalysts for N₂ fixation using KC₈ (40 equiv) and HBAr^F₄ (38 equiv) in Et₂O at −78 °C.⁵³

As with (PNP)Fe(H),¹¹⁶ employing (PNP)Co(H) as the (pre)catalyst led to slightly decreased NH₃ yields, though it was established that, at least at room temperature, the reaction of (PNP)Co(H) with HBAr^F₄, and then KC₈, led to the release of H₂ and reformation of (PNP)Co(N₂). Efforts to identify Co speciation under catalytic conditions led only to the identification of(PNP)Co(N₂), suggesting the platform is fairly robust. Because of the observation that N₂H₄ forms in some reactions, (PNP)Co(N₂) was reacted with 4 equiv of N₂H₄ in the presence of excess HBAr^F₄ and KC₈ under an Ar atmosphere. This resulted in the generation of 0.8 equiv of NH₃ and recovery of 2.4 equiv of N₂H₄, with no observation of N₂ formation. These data suggest that the Co complex was able to mediate the reduction of N₂H₄ to NH₃, suggesting that N₂H₄ is a plausible intermediate and may be the source of some, or all, of the observed NH₃.

6.3. Os and Ru Catalysts with a Tris(phosphino)silyl Ligand

As discussed above for (P₃^E)M(N₂) complexes, accessing an anionic charge state appears to be important to N₂RR activity, at least under catalytic conditions surveyed thus far. Thus, for N₂RR to be achieved with the P₃^Si complexes of the heavier Group 8 metals, Ru and Os, electronically (and structurally) unusual zerovalent examples need to be accessible (Figure 66).²⁵¹ Such a study is attractive as it affords an isoelectronic and isostructural set of [(P₃^Si)M(N₂)]⁻ group 8 complexes to help delineate factors most important to catalysis when proceeding down a group. The anionic, zerovalent complexes [(P₃^Si)Ru(N₂)]⁻ and [(P₃^Si)Os(N₂)]⁻ were thus prepared and tested for their catalytic performance in N₂-to-NH₃ conversion using KC₈/HBAr^F₄ and Cp*₂Co/[Ph₂NH₂]OTf (Figure 66).⁵⁴ With KC₈ and HBAr^F₄, it was found that [(P₃^Si)Fe(N₂)]⁻, [(P₃^Si)Ru(N₂)]⁻, and [(P₃^Si)Os(N₂)]⁻ gave 0.8 ± 0.5, 4.3 ± 0.3, and 1.6 ± 0.3 equiv of NH₃ per M, respectively. By contrast, with Cp*₂Co and [Ph₂NH₂]OTf, it was found that [(P₃^Si)Fe(N₂)]⁻, [(P₃^Si)Ru(N₂)]⁻, and [(P₃^Si)Os(N₂)]⁻ gave 1.4 ± 0.3, 0.8 ± 0.5, and 7.1 ± 0.6 equiv of NH₃ per M, respectively. These results demonstrated the first examples of N₂RR catalysis by Ru and Os.

Figure 66.

Comparison of physical properties and catalytic N₂RR performance for known [(P₃^Si)M(N₂)]ⁿ complexes (M = Co, Fe, Ru, Os; n = 0 or −1).

Particular attention was devoted to the system that was most efficient using the milder [Ph₂NH₂]OTf and Cp*₂Co conditions, [(P₃^Si)Os(N₂)]⁻. Compared to the (P₃^B)Fe system, there was a lesser decrease in overall efficiency (46% to 24%) as the loading of the acid and reductant was increased (36-fold in this case), resulting in the demonstration of a higher maximum turnover number (120 ± 11 equiv of NH₃ per Os). This observation suggests potentially enhanced stability for the heavier Os analogue. Also, and in direct contrast to results described previously with (P₃^B)Fe,¹⁹⁷ the yield of NH₃ was found to be relatively invariant to the pK_a of the acid employed, suggesting that the overall mechanism, or some limiting step, might be distinct between the two systems.¹⁹⁷ Compared to the (P₃^B)Fe system, more soluble acid sources, such as HOTf, [Ph₂NH₂]BAr^F₄, and [H₃NPh]BAr^F₄, were more detrimental to the overall N₂RR performance of [(P₃^Si)Os(N₂)]⁻, in this case leading to substoichiometric yields of NH₃.

The poorer performance of the Os catalyst when using ether soluble acids may be accounted for by the rapid generation of hydrides. [(P₃^Si)Os(N₂)]⁻ reacted with HBAr^F₄ at −78 °C to form (P₃^Si)Os(N₂)(H) and (P₃^Si)Os(H)₃; these two hydride complexes were demonstrated to be inactive as (pre)catalysts for N₂RR. Potentially, hydride formation is diminished with the less soluble triflate-based acids, facilitating N₂RR by comparison to HER.

Performing protonation reactions of [(P₃^Si)Os(N₂)]⁻ at even lower temperature (−135 °C, Figure 67) formed the cationic hydrazido(2−) complex [(P₃^Si)Os(NNH₂)]⁺. This reactivity was similar to that observed for [(P₃^Si)Fe(N₂)]⁻ and [(P₃^B)Fe(N₂)]⁻.^109,113 Isolated [(P₃^Si)Os(NNH₂)]⁺ was reacted with 46 equiv of [Ph₂NH₂]OTf and 50 equiv of Cp*₂Co at −78 °C in Et₂O (i.e., the catalytic conditions), but only 2.6 equiv of NH₃ was generated. Under these conditions, [(P₃^Si)Os(N₂)]⁻ produced 7.1 ± 0.6 equiv of NH₃. The decreased catalytic performance by what is presumed to be an on-path intermediate could be explained by the high reactivity of [(P₃^Si)Os(NNH₂)]⁺, and also and its one electron reduced congener (P₃^Si)Os(NNH₂). In this scenario, [(P₃^Si)Os(NNH₂)]⁺ is present at much higher concentrations at early stages when used as the precatalyst, whereas it likely never builds up during catalysis when [(P₃^Si)Os(N₂)]⁻ is employed as the (pre)catalyst. The ability of [(P₃^Si)Os(NNH₂)]⁺ (or (P₃^Si)Os(NNH₂)) to undergo undesired reaction steps, for example, bimolecular H₂ release or disproportionation to hydride species is thus higher than under the standard catalytic conditions. Alternatively, the kinetic insolubility of [(P₃^Si)Os(NNH₂)]⁺ may hinder its performance; a similar explanation has been posited as an explanation for the poor catalytic performance of (P₃^B)Fe(N₂) and (P₃^B–H)Fe(H)(N₂).^18,173

Figure 67.

Comparison of the reactivity of [(P₃^Si)M(N₂)]⁻ (M = Fe or Os) complexes with excess acid at different temperatures.^54,109

Given the aforementioned observation that anionic [(P₃^E)M(N₂)]⁻ complexes are most effective for facilitating the first protonation step, the reduction of (P₃^Si)Os(N₂) was attempted with excess Cp*₂Co. Curiously, it was found that this reduction occurred reversibly, but only upon cooling; no reaction between (P₃^Si)Os(N₂) and Cp*₂Co occurred at room temperature. This fact suggests that the redox reaction is entropy-controlled. For Fe and Ru, the [(P₃^Si)M(N₂)]^0/− couple is more negative, suggesting that the key, anionic complexes, [(P₃^Si)M(N₂)]⁻ may be inaccessible with Cp*₂Co, potentially accounting for their poor performance under these conditions. This observation is also important to keep in mind with respect to the factors that lead to better N₂RR performance at low temperature by [(P₃^E)M(N₂)]⁻ and a variety of other systems.

6.4. Establishing Catalysis with V

Until recently, vanadium was the only metal ion known to be present in a functionally active nitrogenase enzyme (VFe-N₂ase),¹⁵⁵ which had not been successfully incorporated into a synthetic catalyst for N₂RR.²⁵² With the discovery of more successful conditions for N₂RR, it was possible to establish the efficacy of early metals such as V (and also Ti; vide infra). Specifically, Nishibayashi and co-workers reported in 2018 that using the same PNP ligand (2,5-bis(dialkylphosphinomethyl)-pyrrolide)) that they had previously exploited for N₂RR mediated by Fe (Figure 54)¹¹⁶ and Co (Figure 65)⁵³ could be exploited to achieve N₂RR by V in the presence of excess KC₈ and HBAr^F₄ at low temperature (Figure 68).⁵⁶ Inclusion of an aryloxide ligand, an approach also recently adopted by Schrock and co-workers in an diamidopyridine-pincer-ligated Mo system,¹⁴⁰ proved necessary for N₂RR.

Figure 68.

Synthesis of pyrrole-based (PNP)V precatalysts for catalytic N₂RR.⁵⁶

The maximum catalytic efficiency achieved by this V system (56 ± 3%) at low loadings was slightly superior to those reported for other catalysts using KC₈ and HBAr^F₄ at low temperature.^{18,53,116,173} However, increased turnover numbers (up to fixed-N yield of 16.0 ± 1.4) were only accessible with a significant attenuation in efficiency (23 ± 1.5%). The parent amide complex, (PNP)V(NH₂)(OAr), also proved an effective (pre)catalyst, consistent with (but not necessitating) its intermediacy along the catalytic cycle.

6.5. Nishibayashi’s Work on PNP with Other Metals

Related to the above study, Nishibayashi and co-workers recently reported the formation of PNP-ligated Ti and Zr dinitrogen complexes.²⁵³ It was found that the use of Cp as an additional ligand facilitated N₂ binding upon reduction of halide precursors. Formation of PNP- and Cp-ligated Ti and Zr–N₂ complexes has been reported previously.²⁵⁴ These bridging N₂ complexes were then tested for catalytic activity, both with KC₈/HBAr^F₄ and Cp*₂Co/[Ph₂NH₂]OTf; in all cases, less than 1.4 fixed-N atom equivalents were observed.

6.6. Triamidoamine Ligands on Ti

Triamidoamine-ligated Ti complexes have recently been reported to be competent for catalytic N₂-to-NH₃ conversion (Figure 69).⁵⁵ Like Schrock and co-workers previously,²⁵⁵ Liddle and co-workers found that the trimethylsilyl substituted triamidoamine ligand (TMSN₃N) gave rise to a dinuclear μ-N₂ complex [(TMSN₃N)Ti]₂(μ-N₂) (Figure 69). However, with Ti upon reduction the bridging motif was retained and a dianionic complex [[(Me₃SiN₃N)Ti]₂(μ-N₂)}²⁻ was formed, in which the N–N bond lengthened (1.121(6)−1.315(3) Å) and weakened dramatically (Δν_NN = −500 cm⁻¹). Even after encapsulation of the K⁺ ions with crown ether, the N₂ ligand was still highly activated. Indeed, DFT calculations suggested that the N₂ ligand is reduced to a hydrazido(2−) state in {[[Me₃SiN₃N]Ti]₂(N₂)]²⁻. Reaction of [[(Me₃SiN₃N)Ti]₂(μ-N₂)]²⁻ in pentane with ethereal HCl led to substantial fixed-N yields (0.88 equiv of N₂H₄ and 0.13 equiv of NH₃). In contrast, reaction of the neutral complex, [(Me₃SiN₃N)Ti]₂(μ-N₂), with ethereal HCl produced only 0.03 equiv of N₂H₄ and 0.01 equiv of NH₃.⁵⁵

It was found that by pairing KC₈ with phosphonium acids [R₃PH]⁺ (R = Cy, ⁿBu, ^tBu; pK_a = 9.7, 8.4, and 8.0 in CH₃NO₂, respectively)²⁵⁶ catalytic NH₃ formation could be observed.⁵⁵ The best efficiencies reported (17%) were achieved with I⁻ as the counteranion and Et₂O as the solvent. Of mechanistic interest was the suggestion by the authors that the end-on bridging N₂ mode was critical for functionalization. This mode of binding often leads to significant activation of the N₂ unit but is typically presumed to kinetically retard electrophilic functionalization.^211,257 The authors suggested that such a bridged state can be protonated in the present system to form a bridging hydrazine complex that is then released, with the final reduction to NH₃ mediated by the acid and reductant. The authors did suggest caution with respect to this hypothesis, noting it warrants further investigation to rule out the transient formation of a Ti(N) species. The recent suggestion by the Nishibayashi group that the inclusion of I⁻ in their catalytic reactions biases the mechanism toward cleavage of a bridging N₂ ligand to form two Mo(N) species⁸¹ underscores that such caution is warranted until/if more data become available.

If it ultimately proves true that μ-N₂ complexes in this Ti-system are active for nitrogen functionalization, then known (RN₃N)M–N₂–M(NN₃R) complexes (M = V, W, Mo)^{98,255,258,259} may warrant further study for applications in catalytic N₂RR. Furthermore, the pairing of KC₈ with the weak, insoluble phosphonium iodide acids suggests that known systems that rely on KC₈ should be reinvestigated for their activity with this acid source. Previously, all systems that used KC₈ relied on the use of the highly soluble and very strong acid HBAr^F₄,^{18,52–54,56,116} which can lead to rapid and undesired metal protonations, or background H₂ evolution.

7. ELECTROCATALYTIC NITROGEN FIXATION

The vast majority of studies of homogeneous molecular catalysts that mediate N₂RRhave utilized chemical reductants. However, replacement of chemical reductants with an electrode is an attractive route for developing new N₂RR technologies. Such an approach could enable distributed fertilizer production strategies as a complement to Haber–Bosch. Also, if one considers that ammonia is a high energy density liquid in its easily condensed form, it becomes a very attractive solar fuel target if its synthesis can be efficiently driven by renewable energy resources.^260–264 While there is a great deal of work ahead to move this realm of N₂RR catalysis forward, recent results have demonstrated that electrocatalytic nitrogen fixation with well-defined molecular catalysts is indeed possible. These studies will hopefully be complementary to the burgeoning field of heterogeneous electrocatalytic N₂RR^6,262,264 and the more recently demonstrated bioelectrocatalytic N₂RR.^265,266

7.1. Early Results on Mo and W

Early work on the electrosynthesis of ammonia from nitrogen using coordination complexes was motivated by the early work of Chatt and co-workers.¹⁵ In particular, Chatt and co-workers had demonstrated that the protonolysis of (dmpe)₂W(N₂)₂ (dmpe = dimethylphosphinoethane) in alcoholic solvents led to the formation of nearly 1 equiv of NH₃.⁵⁸ Thereafter, Pickett and co-workers demonstrated the first electrosynthesis of ammonia with a molecular complex using this same system.²⁶⁷ Protonation of (dmpe)₂W(N₂)₂ with p-toluenesulfonic acid (TsOH) led to quantitative formation of the hydrazido(2−) complex, [(dmpe)₂W(NNH₂)(OTs)]⁺ (Figure 70). Cyclic voltammograms of the cationic hydrazido complex featured an irreversible reduction at −2.37 V vs Fc^+/0.²⁶⁸ Exhaustive electrolysis of this species at −2.6 V vs Fc^+/0 at a mercury-pool electrode consumed 2 faradays. Product analysis showed the formation of 0.22−0.24 mol of NH₃ and 0.01−0.02 mol of N₂H₄ per mol of [(dmpe)₂W(NNH₂)(OTs)]⁺ (Figure 70). These results were confirmed both by ¹⁵N isotopic labeling and a variety of control reactions.

Figure 70.

Electrosynthetic cycle for the formation of NH₃ from N₂ by a (dppe)₂W complex.²⁶⁷

Post electrolysis, ³¹P NMR and IR spectroscopy and cyclic voltammetry measurements confirmed that the starting (dmpe)₂W(N₂)₂ complex was regenerated in 85−95% yield.²⁶⁸ The use of a platinum electrode led to similar fixed-N yields, but a lower recovery of the W(N₂)₂ complex. As no excess acid was added to these experiments and an aprotic solvent (THF) was used, the additional protons necessary to form NH₃ were likely derived from the starting complex. Consistent with this hypothesis, the cationic diazenido complex [(dmpe)₂W(NNH)]⁺ was observed by multinuclear NMR spectroscopy and cyclic voltammetry during the electrolysis. Thus, the maximum amount of NH₃ that could have been formed is 0.66 mol, so the overall electrolysis process was ~36% efficient. These data are consistent with a pattern, in which net H atom transfer between N_xH_y intermediates is critical to engendering productive N₂ fixation (see previous discussion of Schrock’s (HIPTN₃N)Mo(NNH₂)/(NH)¹⁰⁰ and Peters’ [(P₃^Si)Fe(NNH₂)]^0/+).¹⁰⁹ Cycling the process three times by first electrolyzing and then adding acid to the solution led to the formation of 0.73 mol of NH₃ in total per mole of starting [(dmpe)₂W(NNH₂)(OTs)]⁺.²⁶⁷

These systems were further investigated by Becker and co-workers.²⁶⁹ They studied related Mo and W phosphine complexes via cyclic voltammetry with the aim of evaluating their applicability to electrocatalytic N₂RR. Cyclic voltammograms of the dinitrogen complexes, (dppe)₂M(N₂)₂ (dppe = diphenylphosphinoethane, M = Mo, W), did not reveal any redox processes. However, as observed by Pickett and co-workers, the cationic hydrazido species, [(dppe)₂Mo(X)-(NNH₂)]⁺ (X = Br, F, HSO₄) and [(dppe)₂W(Br)(NNH₂)]⁺, featured irreversible reduction events. Cyclic voltammograms of the hydrazido species in the presence of an inorganic acid (3–5 equiv) led to an observable increase in current (10–20%), suggesting the possibility of an electrocatalytic process (i.e., HER or N₂RR). Therefore, controlled potential electrolysis (CPE) was performed in the presence of inorganic acids (3–5 equiv). However, yields of NH₃ were lower than those observed by Pickett and co-workers, with the observed NH₃ yields for OTs⁻ > HSO₄⁻ > Br⁻. While yields of NH₃ from CPE were never high enough to establish bona fide electrocatalysis, they were somewhat enhanced by running the experiment under N₂, but also under CO. These observations led the authors to suggest that while atmospheric N₂ was not being incorporated into NH₃, the electroreduction of one of the intermediates was enhanced by loss of the X-type ligand and binding of a π-acid.

Relatedly, Shilov and co-workers reported a number of studies looking at the electrocatalytic reduction of N₂ by Mo salts in methanolic solution at a Hg drop electrode.²⁰ In these studies, they primarily interrogated the activity of ill-defined metal salt mixtures that are not further described here. However, after discovering that phosphatidyl choline (a phosphine-capped fatty acid) enhanced the reactivity of their Mo salts,⁷⁶ they tested the effect of adding (PPh₂Me)₄Mo(N₂)₂ and (dppe)₂Mo(N₂)₂ to their CPE experiments.⁷⁷ The inclusion of these complexes did not lead to any enhancement in NH₃ formation.

The cumulative work by these groups suggests that while these bis-diphosphine Mo and W complexes are suitable for the stepwise electrosynthesis of NH₃, they are not capable of electrocatalytic NH₃ formation, at least under the conditions thus far studied.⁷⁷

7.2. Achieving True Electrocatalysis with (P₃^B)Fe

In 2016, Peters and co-workers studied the effect of acid addition, HBAr^F₄, to cyclic voltammograms of [(P₃^B)Fe]⁺ (conditions: −45 °C, 0.1 M [Na]BAr^F₄ in Et₂O).Upon addition of 5 equiv of HBAr^F₄, only minimal current enhancement was observed at the [(P₃^B)Fe]⁺/(P₃^B)Fe(N₂) couple (−1.5 V vs Fc^+/0), but a more substantial enhancement was observed at the [(P₃^B)Fe(N₂)]^0/−couple (−2.2 V vs Fc^+/0). Analysis of this current enhancement is likely complicated by competing catalytic hydrogen evolution and N₂RR. A CPE experiment with a vitreous carbon-working electrode at −2.6 V vs Fc^+/0 with 10 equiv of HBAr^F₄ led to the formation of NH₃ (0.5 equiv, FE = 18%) and H₂ (FE = 58%). A further CPE experiment at −2.3 V vs Fc^+/0 in the presence of 50 equiv of HBAr^F₄ formed 2.2 equiv of NH₃ per Fe (FE = 25%) and 6.6 equiv of H₂ per Fe (FE = 48%). This result placed this system on the precipice of electrocatalytic formation of NH₃ as any reliable NH₃ yield above 2.0 equiv per Fe requires a catalytic process.¹⁷³ Significant effort was made to improve the yield of NH₃, including changes in the acid (i.e., more acid, batchwise acid addition, constant acid addition) and the ratio of the electrode surface area to the working compartment solution volume (i.e., smaller cell geometries, different carbon morphologies).¹⁹⁷ These efforts did not result in substantial improvement to the overall turnover number (TON ≤ 2.6 equiv of NH₃ per Fe).

Following the discovery that chemical catalysis could be mediated at significantly higher efficiencies by Cp*₂Co and [Ph₂NH₂]OTf⁸⁰ than by KC₈ and HBAr^F₄^18,173 efforts to adapt these new conditions to electrocatalysis were undertaken. Cyclic voltammograms of [(P₃^B)Fe]BAr^F₄ at −35 °C in 0.1 M [Na]BAr^F₄ solution changed significantly upon addition of 10 equiv of [Ph₂NH₂]OTf. Notably, the [(P₃^B)Fe]⁺/(P₃^B)Fe(N₂) couple observed previously disappeared and was replaced by an irreversible feature at ~ −1.9 V vs Fc^+/0. Cyclic voltammograms of [(P₃^B)Fe]BAr^F₄ with added [TBA]OTf suggest this feature arose from the reduction of (P₃^B)Fe(OTf). Scanning the cyclic voltammogram further in the cathodic direction in the presence of [Ph₂NH₂]OTf revealed a substantial catalytic enhancement at the [(P₃^B)Fe(N₂)]^0/−couple. CPE experiments under these conditions (−35 °C, Et₂O, 0.1 M [Na]BAr^F₄, 50 equiv of [Ph₂NH₂]OTf) led to a similar yield of NH₃ (2.6 ± 0.3 equiv per Fe, 24 ± 5% FE) as observed with HBAr^F₄.

Given the evidence for the role of a protonated metallocene, [Cp*(exo/endo-η⁴-C₅Me₅H)Co]⁺, in N–H bond formation during the chemical catalysis,^80,197,199 it was explored whether the addition of cocatalytic [Cp*₂Co]BAr^F₄ additive might enhance the yield of NH₃. As expected, cyclic voltammograms of [Cp*₂Co]BAr^F₄ at −35 °C in 0.1 M [Na]BAr^F₄ solution revealed the reversible [Cp*₂Co]^+/0 couple at −2.0 V versus Fc^+/0. Addition of 10 equiv of [Ph₂NH₂]OTf led to the observation of a catalytic enhancement at this reduction event. CPE experiments with [Cp*₂Co][BAr^F₄] in the absence of [(P₃^B)Fe]BAr^F₄ resulted in catalytic H₂ formation (FE = 75%) but no NH₃ formation. Cyclic voltammograms with [(P₃^B)Fe]BAr^F₄, [Cp*₂Co]BAr^F₄, and [Ph₂NH₂]OTf also showed the onset of a catalytic wave at ~ −2.0 V versus Fc^+/0. CPE experiments (−2.1 V vs Fc^+/0) in the presence of [(P₃^b)Fe]BAr^F₄ and 1 equiv of [Cp*₂Co][BAr^F₄] provided unequivocal evidence for electrocatalytic nitrogen fixation (4.0 ± 0.6 equiv of NH₃ per Fe/Co, 28 ± 5% FE, Figure 71). The use of higher amounts of [Cp*₂Co]BAr^F₄ did not enhance the yields or improve the FE. However, it is important to note that in CPE experiments with 5 equiv of [Cp*₂Co]BAr^F₄, addition of further substrate after completion of the initial CPE experiment enhanced the yield of NH₃ from 4.0 ± 0.6 equiv of NH₃ per Fe to 5.5 ± 0.9 equiv of NH₃ per Fe. XPS experiments on the electrode at the end of the CPE indicated a modest degree of Fe decomposition but provided no evidence for Co decomposition. Furthermore, CPE experiments performed after rinsing the electrode did not lead to the formation of NH₃. In sum, this system represents the first example of electrocatalytic N₂RR by a well-defined coordination complex.

Figure 71.

First demonstration of molecular, electrocatalytic N₂RR enabled by the use of cocatalytic [Cp*₂Co]⁺. Conditions: glassy carbon, Et₂O, −35 °C, −2.1 V versus Fc^+/0, 0.1 M [Na]BAr^F₄, X equiv of [Cp*₂Co]BAr^F₄.¹⁹⁷

To compare the effect of using an electrode as compared to a chemical reductant, N₂ fixation experiments were performed with [(P₃^B)Fe]BAr^F₄ in the presence of 108 equiv of [Ph₂NH₂]OTf and 54 equiv of Cp*₂Co at −35 °C in 0.1 M [Na]BAr^F₄ Et₂O. Only 1.8 ± 0.7 equiv of NH₃ per Fe was formed (Figure 72). This result is much lower than the 12.8 ± 0.5 equiv of NH₃ per Fe observed under the standard chemical catalytic conditions,⁸⁰ potentially due to the higher temperature (−35 °C vs −78 °C) and/or higher polarity solvent (0.1 M [Na]BAr^F₄ Et₂O versus Et₂O). More significantly, the performance with the chemical reductant is poorer than that with the electrochemical reductant under analogous conditions (4.0 ± 0.6 equiv of NH₃ per Fe).¹⁹⁷ This observation suggests that under certain conditions electrocatalysis can be superior to chemical catalysis. However, challenges in developing better electrocatalytic systems remain, as the conditions that have thus far worked best (insoluble acid, low dielectric media, and low temperature) are not conducive to standard electrochemical techniques. Nevertheless, these results provide a springboard for future work in electrocatalytic N₂RR.

Figure 72.

Reproduced with permission from ref 197. Copyright 2018 American Chemical Society. (A) Cyclic voltammograms of 10 equiv of [Ph₂NH₂]OTf (gray trace), 1 equiv of [Cp*₂Co]BAr^F₄ (Cp*₂Co⁺) (yellow trace), 1 equiv of Cp*₂Co⁺ with 10 equiv of [Ph₂NH₂]OTf (green trace), and [(P₃^B)Fe]⁺ with 1 equiv of Cp*₂Co⁺ and 10 equiv of [Ph₂NH₂]OTf (red trace); (B) Cyclic voltammograms of 10 equiv of [Ph₂NH₂]OTf (gray trace), 1 equiv of [(P₃^B)Fe]⁺ (dark blue trace), 1 equiv of [(P₃^B)Fe]⁺ with 10 equiv of [Ph₂NH₂]OTf (light blue trace), and 1 equiv of [(P₃^B)Fe]⁺ with 1 equiv of [Cp*₂Co]⁺ and 10 equiv of [Ph₂NH₂]OTf (red trace). All voltammograms are collected in 0.1 M [Na]BAr^F₄ solution in Et₂O at −35 °C using a glassy carbon working electrode and externally referenced to the Fc^+/0 couple. Scan rate is 100 mV/s.

7.3. Work on (Cp)_xTi-Species

In one of the first reports regarding the fixation of N₂-to-NH₃ under ambient conditions by metal complexes, Shur and Vol’pin found that the use of Cp as a ligand dramatically enhanced the performance of Ti^IV complexes for N₂ fixation by comparing Ti(Cl)₄ with (Cp)₂Ti(Cl)₂.²⁷⁰ This report spawned substantial interest in (Cp)₂Ti(Cl)₂ for nitrogen fixation; however, the mechanism of N₂ activation and fixation remains disputed.^271–275 Nonetheless, (Cp)Ti-complexes are among the most highly investigated systems for electrosynthetic nitrogen fixation (Figure 73).

Figure 73.

Comparison of conditions and results for the electrosynthesis of NH₃ from N₂ by (Cp)₂Ti(Cl)₂.^276–278

The first report of (Cp)₂Ti(Cl)₂ in the context of electrocatalytic nitrogen fixation was by Becker and co-workers, who studied its ability to reduce N₂ in MeOH and THF at platinum and glassy carbon electrodes with catechol as the proton source. This system was found to be more efficient than with other titanium salts, but yields were nonetheless very poor, with the best results achieved in the presence of a Mg²⁺ additive (0.0145 equiv of NH₃ per Ti and 0.28% Faradaic efficiency (FE), Figure 73).²⁷⁶

A more recent report from Yoon and co-workers examined the electroreduction of (Cp)₂Ti(Cl)₂ in MeOH, H₂O, and wet THF with LiCl electrolyte (Figure 73). They found that the maximal rate of NH₃ formation (9.5 × 10⁻¹⁰ mol·cm⁻²·s⁻¹·M (Cp)₂Ti(Cl)₂−¹) occurred in water at −1 V vs RHE, and that the maximal FE for NH₃ formation (0.95%) occurred at −2 V vs RHE in THF. Yields with respect to Ti were not reported.²⁷⁷

Masuda and co-workers have looked at the effect of using ionic liquids as the solvent and electrolyte on the electrosynthesis of NH₃ by (Cp)₂Ti(Cl)₂. When (Cp)₂Ti(Cl)₂ was immobilized in a solid polymer electrolyte cell with an ionic liquid (Pyr₄FAP) as the supporting material and water as the proton source, they found that the introduction of N₂ led to significant current enhancement in the cyclic voltammogram at the reductive feature corresponding to (Cp)₂Ti(Cl)₂ reduction to (Cp)₂Ti(Cl) (Figure 73). However, they also noted that most of the current enhancement stems from hydrogen evolution mediated by a Ti(N₂) complex rather than nitrogen fixation. Nonetheless, CPE experiments at −1.5 V vs Fc^+/0 revealed a significant improvement in the yield of NH₃ with little change in the FE (0.27 equiv of NH₃ per Ti, 0.2% FE). This represented a notable improvement over the previous conditions, given the lower applied bias used along with a weaker acid source (H₂O vs catechol). The authors attributed the formation of NH₃ at lower applied bias to enhanced N₂ binding afforded by the ionic liquid environment. In a subsequent report, they demonstrated that in this ionic liquid, Ti^III can bind N₂, which has not been observed in traditional solvents (i.e., MeOH, THF, H₂O).²⁷⁹ These observations suggest that the use of ionic liquids may provide a promising strategy for electrocatalytic nitrogen fixation by helping to both polarize the N₂ molecule and to activate it for binding/reduction, and in limiting proton reduction. More recent work from Masuda and co-workers has demonstrated that the related dimeric species, {(Cp)₂Ti(μ-Cl)}₂, can provide higher yields and FEs for nitrogen fixation (up to 0.34 equiv of NH₃ per Ti and up to 1.44% FE).²⁷⁸

7.4. (Electro)synthesis of NH₃ with p-block Metals

In a series of papers, Furuya and co-workers examined the potential of different phtahlocyanines loaded on to gas-diffusion electrodes for nitrogen fixation. CPE experiments with Fe-substituted phthalocyanine at −0.6 V vs RHE in 1 M Na₂SO₄ led to an initial FE of 1.6%, but a sharp reduction to 0.1% was observed after 10 min.²⁸⁰ Interrogation of the effect of different electrolytes on the efficiency of the electrocatalysis led to the observation that potassium electrolytes were uniformly superior to sodium electrolytes, with 1 M KOH giving the best performance with retention of >0.1% efficiency for at least 30 min.²⁸¹ However, when a variety of phthalocyanines were tested (M = H, Ti, Fe, Co, Ni, Pd, Pt, Cu, Zn, Al, Ga, In, Sb, Sn), the best and most stable FE was observed with Sn (max 1.8%, 1.2% at 25 min).²⁸² However, more recent work on Sn-phthalocyanine has demonstrated that it is not an electrocatalyst for N₂RR, but rather that the detected NH₃ comes from the decomposition of contaminants in the starting materials.²⁸³ This and other recent work exploring the confounding sources of NH₃ in electrocatalytic N₂RR by materials²⁸⁴ illustrate the importance of careful control reactions for establishing the source of NH₃ in studies toward electrocatalytic N₂RR.

Recently, Berben and co-workers reported the electrosynthesis of NH₃ with an aluminum complex, notable as the first example of the formation of fixed-N products by a main group metal.²⁸⁵ Cyclic voltammograms of the pyridine-diimine (PDI) ligated Al complex (Figure 74) in the presence of 4-dimethylaminopyridinium and N₂ led to the observation of enhanced current. CPE at −1.16 V versus SCE yielded 21% NH₃ and 16% H₂ (the poor overall FE was not unexplained). The source of the NH₃ was confirmed to be N₂ via isotopic labeling. The lack of turnover was explained by product inhibition; the starting complex was demonstrated chemically to be unstable to NH₃. Interestingly, no N₂ complex is known for this platform, so more work is needed to begin to elucidate an overall mechanism (Figure 74) for this process. The authors asserted that the overall reduction reaction is occurring at a significantly lower overpotential than other known, catalytic systems. However, their analysis used pK_ip values as substitutes for pK_a values in nonpolar solvents. The arbitrary zero point of the pK_ip scale in these solvents creates a significant bias against those systems operating in nonpolar solvents (i.e., Schrock and Nishibayashi’s original Mo work).^50,51 Hence, caution in this interpretation is warranted. Regardless, this report, along with recent work from Braunschweig and co-workers describing reductive capture of N₂ by two borylene units and subsequent protonation to a hydrazine derivative, provides inspiration for future work on nitrogen fixation by p-block elements (Figure 74).²⁸⁶

Figure 74.

(Top) Preliminary, posited mechanism for the electrosynthesis of NH₃ by a (PDl)Al complex;²⁸⁵ (bottom) stoichiometric N₂ binding and protonation at a (CAAC)B species; Dipp = 2,6-diisopropylphenyl, Dur = 2,3,5,6-tetramethylphenyl.²⁸⁶

8. MECHANISTIC INSIGHTS FOR FUTURE CATALYST DESIGN

Having discussed aspects of synthetic nitrogenases that have been developed to date, we finish this review by summarizing some lessons we think will be useful to consider in designing future catalysts.

• 1Balancing sterics: Key to the design of efficient N₂ fixation catalysts is the preparation of molecular species with the appropriate sterics. Sterics can control the rate or nature of dimer formation. For example, see work from the Nishibayashi group on the effect of phosphine substituents on cis- vs trans-dimer formation (Figure 27),¹⁴¹ the control that wingtip bulk has over the formation of terminal vs bridged N₂ species in triamidoamine Mo chemistry from Schrock and co-workers.^98,255 The mechanism (and presumably rate) of reactant/product exchange is also dependent on sterics. For example, we have noted that among the Schrock systems that (DPPN₃N)Mo(N₂) undergoes associative rather than dissociative N₂ exchange.¹³⁴ Similarly the site (and presumably rate) of ligand protonation is controlled by sterics. One clear example comes from Nishibayashi and co-workers, in which they prevent protonation in the 3,4-positions of the pyrrole ligand in their (PNP)Fe ligand by introducing methyl substitution (Figure 54).¹⁷⁸ Lastly, sterics can be important for controlling the rate of bimolecular decomposition. This includes the rate of formation and degradation of intermediates. This idea is highlighted in ability to observe a diazenido intermediate, Fe(NNH), only when using a highly sterically encumbering ligand (^ArP₃^B) (Figure 44). Notably, while this system is better at stabilizing this highly reactive intermediate, it is a poorer catalyst than the parent system, (P₃^B)Fe.¹⁷⁷ Thus, a Goldilocks approach to sterics is necessary for achieving optimal functionality, as is the case for a variety of catalytic reactions.
• 2The design of innocent yet activating ligands: One factor that dictates the efficacy of N₂-to-NH₃ reduction catalysis is linked to the ease by which protonation occurs at N_β. Activation of N₂ proceeds by cooperative donation from a transition metal or main group compound and the given ligand set through inductive effects. The importance of N₂ activation is reflected in (i) the need to access [(HIPTN₃N)Mo(N₂)]⁻ for protonation at N_β to occur - (HIPTN₃N)Mo(N₂) is not protonated at N_β by catalytically relevant acids (Figure 10),¹⁰² (ii) the catalytic enhancement that is observed for the (PNP)Mo catalyst upon 4-OMe substitution (Figure 30),⁸² and (iii) in the necessity of accessing an anionic [(P₃^E)M(N₂)]⁻ state (M = Fe E = B, C, Si; M = Ru, Os E = Si) to observe N₂RR catalysis with (P₃^E)M species (Figure 66).^54,80 Again, protonation of N_β of neutral (P₃^E)M(N₂) species has not been observed. On the other hand, highly donating ligands that enhance the electron richness of the metal center also render it harder to reduce, thereby necessitating stronger reductants to achieve the low-valent states necessary to initiate catalysis. Furthermore, more donating ligands are often more Bronsted basic and are thus prone to protonolysis, thereby limiting turnover due to degradation of the catalyst. In this respect, neutral phosphine ligands appear to be far superior to amide ligands for enhancing overall stability and thus turnover.^{50,79,116,173} It is also worth highlighting the utility of adaptable ligand frameworks that help to stabilize a variety of formal metal oxidation states and [M]-N_xH_y bond orders—this concept is highlighted by Peters’ use of boratrane and silatrane-containing ligands, for example^{109,113,114,181,193}
• 3Tuning N₂ Reduction Overpotentials with different reagent combinations: Central to the reduction of N₂ are the reagents that are employed for catalysis. From this perspective, it is noteworthy that no single set of catalytic conditions can be applied to all of the catalytic systems that have been reported to date. Indeed, a number of electron sources (metallocene,⁵⁰ KC₈,¹⁸ electrode,¹⁹⁷ Sm(II)⁷⁹) and proton sources (pyridiniums,⁵⁰ protonated ether,¹⁸ aniliniums,¹⁷⁴ phosphoniums,⁵⁵ water/alcohols⁷⁹) have all been employed. The particular combination of acid and reductant, along with other relevant reaction conditions, defines the net overpotential for the catalytic reaction. It has long been known in heterogeneous catalysis that overpotential is correlated with reaction rate;²⁸⁷ more recently, it has been increasingly recognized in related fields (ORR, CO₂RR, HER, etc.) that similar constraints exist,^288–292 and some theoretical approaches, and physical intuition, suggest that a similar relationship exists in catalysts for homogeneous N₂RR.²⁹³

Notable in N₂RR is also that the acids and reductants typically only work in particular pairs. For example, metallocenes and pyridinium acids effectively promote N₂RR catalysis with Nishibayashi’s Mo systems (Figure 33),⁵¹ as do Sm(II) and water/alcohols (Figure 37), but combining Sm(II) and pyridinium acids has been shown to be ineffective.⁷⁹ Similarly, both KC₈/HBAr^F₄ and Cp*₂Co/[Ph₂NH₂]OTf are efficient N₂ fixation cocktails with the (P₃^B)Fe-system,^18,80 but the combination of Cp*₂Co and HBAr^F₄ proves ineffective. Indeed, changing these reagent pairs can have a larger impact on efficiency, TON, and TOF than does changing the catalyst itself (Figure 66). This is an important point. Whereas it may seem less exciting to a budding coordination chemist to survey reagents than to design an elegant new catalyst system supported by a new ligand, the development of reagent combinations for canvassing nitrogen fixation will drive the field forward at an accelerated pace.

In the past few years, defining the overpotential supplied by different reagent cocktails in catalytic N₂RR has become increasingly a point of focus. Two means of measuring the overpotential have been discussed in the literature: the first, more traditional approach, is to determine the thermodynamic potential of nitrogen fixation for a given solvent at a given pK_a (or pH).^285,294 The overpotential, or η, is then the difference between this value and the applied potential. This approach is useful in facilitating comparisons with heterogeneous N₂RR electrocatalysts and for defining the potential of the most reducing intermediate in the N₂RR cycle. However, developing accurate overpotential values in the apolar solvents that to date have been commonly used for chemical nitrogen fixation (i.e., heptane, toluene, Et₂O) is not feasible. These values are, however, well-defined for H₂O and acetonitrile,²⁹⁵ with extension to THF being seemingly possible.²⁹⁶ A second approach that has been developed as a measure of “overpotential,” typically noted as ΔΔG_f/ΔΔH_f, in the literature takes the approach of comparing the energetic input to the energy of formation for NH₃, in which N₂ is fixed using H₂ as the source of protons and electrons (Figure 75).^79,80,105 Thus, this metric provides for a ready comparison to the Haber–Bosch process, in which H₂ is employed to achieve N₂-to-NH₃ conversion. To define ΔΔG_f, one first needs to know the effective bond strength of the reagent combination used (i.e., Cp*₂Co and [Ph₂NH₂]OTf) as defined by eq 1 (where C_G is a solvent-dependent thermodynamic constant accounting for the energy of formation for H^• from H⁺ and e⁻ in solution).²⁹⁵ Comparing this effective BDFE with the energy of H^• formation derived from H₂, found by dividing the BDFE(H₂) by two, defines the extra energy being used in the system to generate each N–H bond. To get the overpotential (ΔΔG_f) per mol of NH₃, one then multiplies by three to account for the three N–H bonds in NH₃ (eq 2).

$${\text{BDFE}}_{\text{eff}}=1.37\times \text{p}{K}_{\text{a}}+23.06\times E^{\circ }+{\text{C}}_{\text{G}}$$ (1)

$$\Delta \Delta G_{\text{f}}=3\left(\left(\text{BDFE}\left({\text{H}}_2\right)/2\right)-{\text{BDFE}}_{\text{eff}}\right)$$ (2)

Figure 75.

N₂ overpotentials calculated as a function of catalyst reagent cocktail using eqs 1 and 2. All of the pK_a values and reduction potentials are literature values for these reagents in acetonitrile.^297–299 Consequently, the C_g and BDFE(H₂) in acetonitrile were also used.²⁹⁵ The exception is for Sm/H₂O for which aqueous values are used, and the values are approximated due to the inner-sphere nature of the Sm/H₂O interaction. The bottom portion of the figure provides a sample overpotential calculation.¹⁵⁴

Again, the necessity of a well-defined pK_a scale and a C_G constant also limits the accuracy of this approach to particular solvents (e.g., H₂O, MeCN, THF, etc.). This approach also does not account for the excess energy being supplied by the conversion of NH₃ to NH₄⁺, which readily occurs under most of the catalytic conditions discussed in this review. However, the appeal of this approach lies in the potential insight it provides into the N–H bond strengths of key N₂RR intermediates (i.e., [M]-NNH or [M]-NH).

• 4Reaction conditions: Nitrogen fixation catalysis is sensitive to solvent, temperature, and reagent concentration. In many cases, these factors control the interplay between catalyst-mediated nitrogen fixation, catalyst-mediated hydrogen evolution, and background hydrogen evolution. Different strategies have been employed in an effort to control these factors, including performing reactions at lower temperature,^18,116,174 the use of nonpolar solvents,^50,51 and the slow addition of reagents.^50,51 Further, in most cases discussed herein, at least one catalyst component (acid or reductant) has been less soluble than the other in an effort to attenuate the rate of reactions between them (e.g., HER). The importance of these factors can be highlighted by dramatic differences in efficiencies. For example, this is demonstrated by some of Nishibayash’s systems, for variable rates of reductant addition have led to improved catalysis (Figure 25),⁵¹ and also in the strong correlation of performance of the (CAAC)₂Fe-catalyst system with temperature (Figure 55).¹⁷² Nonetheless, many of these factors, and other potentially important factors (e.g., N₂ pressure, NH₃ concentration) remain unexplored for most of these platforms. Furthermore, continue to develop conditions that would be more highly compatible with electrocatalysis, namely, higher solvent polarity, higher reaction temperatures (especially for non-Mo catalysts), and conditions that do not require slow addition of the acid (especially for Mo catalysts).
• 5A role for PCET: Although the ability of PCET reactions to mediate stoichiometric N–H bond forming reactions in noncatalytic platforms has been extensively demonstrated,^{105,217,300–302} discussions on the mechanism for nitrogen fixation by homogeneous catalysts had traditionally focused on stepwise ET-PT or PT-ET pathways.^80,106,128 In the past few years, there has been increased discussion of PCET steps as a possible mechanism for the N–H bond forming steps in catalytic systems with a particular emphasis on its critical role in the formation of intermediates with weak N–H bonds (BDFE_N–H < 50 kcal·mo⁻¹). Proposed species of relevance to PCET in nitrogen fixation catalysis have included pyridinyl radicals,^103,105,106 protonated metallocenes (Figure 47);^80,197,199 encounter complexes of a reductant with an acid and a nitrogenous ligand that are hydrogen-bonded (MS-PCET) (Figure 18),^103,140 and alcoholic or aquo complexes of SmI₂ (Figure 37).⁷⁹ This recent work suggests that a fresh look at earlier mechanistic assumptions may be warranted, but the more important conclusion to draw is that PCET reactions may offer a path forward for developing catalyst systems that operate at lower net overpotentials and improved rates.

9. FUTURE OUTLOOK: REMAINING CHALLENGES

We next provide an overview of some of the important research opportunities that remain for N₂RR catalysis mediated by synthetic complexes.

• 6N₂ fixation with improved selectivity: Nishibayashi and co-workers have recently demonstrated two different cases, (PPP)Mo with Cp*₂Co/[ColH]OTf⁸¹ and (PCP)Mo with SmI₂/HOCH₂CH₂OH (Figure 37),⁷⁹ the ability to achieve >90% selectivity for NH₃. These results represent significant improvements on the maximum efficiency of these catalysts in the presence of the original Schrock conditions (Cp*₂Cr/[ColH]OTf) with which (PNP)Mo achieved ~50% selectivity for NH₃⁵¹ and (PCP)Mo achieved ~66% selectivity.¹⁴⁹ The improved selectivity of the new conditions comes at the cost of greater energetic input. Thus, the search for high selectivity systems under mild conditions continues.
• 7N₂ fixation conditions: Reaction temperature must also be considered. Although all Mo catalyst systems currently operate at room temperature under an atmosphere of N₂, all of the other N₂RR catalyst systems studied are far more efficient at low temperature (≤−78 °C) than at room temperature.^172,173 This represents an opportunity for improvement and will be particularly important with respect to electrocatalytic N₂RR, where low reaction temperatures can significantly attenuate accessible current densities. On the other hand, the potential benefits of elevated N₂ pressures in N₂RR catalysis have not been explored, though hints from the N₂-silylation literature suggest it could be a powerful tool for improving catalysis.³⁰³ Computational studies suggest that the low driving force for the exchange of NH₃ for N₂ is an issue in both (HIPTN₃N)Mo¹²⁴ and (P₃^E)Fe (E = Si, B).²⁰⁹ Additional studies exploring the possibility of using solvent media, including ionic liquids, to increase the availability or reactivity of N₂ is also of interest.²⁷⁹
• 8N₂ fixation with more biologically relevant platforms: Despite steady advances in nitrogenase model chemistry, current synthetic nitrogenases appear to involve key bond-making and breaking events at a single metal center, with the presumed exception of Nishibayashi’s pincer-ligated catalysts under specific conditions.³⁰⁴ None of these synthetic catalysts yet feature inorganic sulfide or carbide ligands as are present in the cofactors of nitrogenases, although progress toward this aim has been made.^249,305 As efforts to understand the nature of biological nitrogen fixation continue, the development of catalysts featuring more biologically faithful coordination environments will be of significant interest to further constrain and test the feasibility of mechanistic hypotheses. Such studies should also ultimately consider secondary sphere influences.^132,306
• 9N₂ fixation product selectivity: One of the fascinating areas of mechanistic diversity in nitrogen fixation catalysis is the differing selectivity for NH₃ and N₂H₄. Thus far, all Mo catalysts have been proposed to operate through a distal mechanism (Figures 12 and 32),^110,304 or an initial N≡N bond cleavage in a specific instance (Figure 34).^79,81 All Fe catalysts appear to form at least some amount of NH₃ and N₂H₄.^80,116,174 Despite this, mechanistic studies that demonstrate a strictly alternating pathway for an active catalyst are lacking. Mechanistic work has, by contrast, pointed to the viability of a hybrid distal-to-alternating pathway.¹⁰⁹ Therefore, increased attention toward identifying a system that traverses an alternating pathway, and/or further elucidating hybrid pathways that lead to N₂H₄ formation, are warranted. Given the utility of N₂H₄-based derivatives as fuels, catalysts that are selective for the direct reduction of N₂-to-N₂H₄ may find niche applications. Furthermore, biological nitrogenases display mechanistic diversity with respect to their selectivity for NH₃ and N₂H₄.¹⁶³ Thus, understanding which mechanisms are compatible from a synthetic catalyst perspective will help in considering hypotheses within a biological framework.
• 10N₂ fixation using light: Recent research from Peters and co-workers has demonstrated that N₂RR turnover can be enhanced via irradiation of the system (Figure 61).^175,196 This observation serves as a potentially important pathway for the removal of catalytically inactive hydride species, via H₂ loss, which in turn increases the availability of catalytically active species. Such studies may also have implications with respect to understanding the release of H₂ and binding of N₂ at biological nitrogenases, a process that can also be light-triggered.¹⁶² Furthermore, recent work from Nishibayashi and co-workers has begun to consider how known nitrogen fixation platforms can be integrated with known water photooxidation processes (i.e., H₂O → O₂ + 4 H⁺ + 4e⁻) (Figure 36).¹⁵⁰ However, there remains significant space for development in this area, which is critical to the future integration of N₂ fixation with renewable energy conversion technologies.
• 11N₂ fixation using electrocatalysis: The development of electrocatalytic nitrogen fixation systems is critical to the field’s future. In this arena, N₂ fixation lags significantly behind other multielectron reduction processes such as hydrogen evolution,^307,308 O₂ reduction,³⁰⁹ or carbon dioxide fixation.³¹⁰ Progress in the realm of N₂RR catalysis has been hindered by the significant kinetic requirement of biasing a system that is sufficiently reducing to catalyze N₂RR and does so in preference to HER. This challenge also confronts progress in heterogeneous N₂RR electrocatalyst studies.^6,262 Furthermore, N₂RR catalysis has typically been studied under conditions that are poorly suited to electrocatalysis (i.e., low-temperature, low dielectric solvent, heterogeneous acids). A recent study by Peters and co-workers shows that these challenges can be overcome to demonstrate electrocatalysis and comparatively high Faradaic efficiencies (Figure 72).¹⁹⁷ Indeed, this study underscores that electrode-mediated N₂RR catalysis can be superior to chemical reductant-mediated catalysis, when compared under identical conditions. Developing new electrocatalytic systems for N₂RR catalysis and comparing and contrasting their mechanisms with chemical catalysis systems presents an important area for exploration.
• 12N₂ fixation in water: A significant goal for the field is to drive N₂RR with the electrons and protons of water. Progress here is needed to advance the use of ammonia as a no-carbon solar fuel. Progress in this context has been limited for synthetic N₂RR catalysts. There have been some studies that offer a path for further development. For example, DMeOPrPE-ligated Fe complexes bind N₂ in water (Figure 57),²³⁶ and protons ultimately derived from water oxidation have been used for N₂-fixation catalysis (Figure 36).¹⁵⁰ Nevertheless, the remaining challenges are significant. Approaches that may lead to progress in this area will include the development of more water-soluble (and stable) catalysts, or the use of micelles that stabilize hydrophobic catalysts in water; the latter has been demonstrated by Shilov and co-workers.⁷⁶ The protic nature of H₂O cautions that its competitive reduction to H₂ is a serious consideration for aqueous N₂-fixation catalysis. In this regard, the development of catalysts that operate at lower net overpotentials will be essential.
• 13An assessment of best N₂RR practices: The accurate detection of NH₃ (or N₂H₄) is not without complication. In the context of electrocatalytic N₂RR, several papers have appeared recently addressing potential sources of NH₃ contamination.^284,311 Many of these factors, such as trace contamination of the N₂ gas and impure catalysts/reagents, are heightened in the set-ups used for heterogeneous electrocatalytic N₂RR, or cases where very small amounts of NH₃ are produced. Nonetheless, caution is warranted when analyzing homogeneous N₂RR catalysts as well. Of particular importance are both control experiments both without catalyst and without N₂ to detect potential sources of contamination, as well as isotopic labeling experiments to verify the source of the NH₃/N₂H₄. As always, it is essential that analytical rigor be employed when quantifying these products.

10. SUMMARY AND CONCLUSIONS

In this review, we have offered a lens into the field N₂-to-NH₃ catalysis (N₂RR) mediated by synthetic complexes. While we have focused almost exclusively on systems for which bona fide catalysis has been demonstrated, we began by discussing the pioneering studies of Chatt and co-workers as well as others who laid the early groundwork for this field (Figure 3). The groundbreaking work of Schrock and co-workers, who in 2003 demonstrated the first example of N₂RR catalysis via a well-defined tri(amido)amine Mo system (Figure 12), has been discussed in detail.⁵⁰ This system still represents one of the most important for the field because of the thorough characterization of many catalytic intermediates that was undertaken. The discussion of Schrock’s system was followed by a description of Nishibayashi’s low-valent Mo-phosphine systems. These systems are related to the original Chatt-type Mo systems in that they feature low valent molybdenum complexes with phosphine donors, and they have proven highly amenable to tuning via manipulations of the ligand (Figure 33). Mononuclear Mo nitrides in particular have proven to be effective (pre)catalysts for N₂ fixation in this context. Most recently, catalysis with unprecedented efficiency, turnover number, and turnover frequency has been realized via SmI₂ as a reductant in combination with protic solvents.⁷⁹ Many of the mechanistic details of the latter system remain to be explored.

We next focused on Fe-mediated N₂RR catalysis, which was first reported in 2013 by Peters and co-workers via a tris(phosphine)borane iron system (Figure 45). This represented a significant advance; prior to this report, the field had focused heavily on Mo-systems, which biased mechanistic proposals relevant to biological nitrogen fixation. Significant experimental and theoretical work for this system has been undertaken to place the mechanism/s by which it operates on firm footing. As for the Schrock system, many candidate intermediates and interconversions have been studied, with many of the early intermediates being highly unstable and requiring low temperatures for characterization. Of particular interest are comparative studies with related (P₃^C)Fe- and (P₃^Si)Fe-systems, which have helped to identify important principles that determine catalyst selectivity. With the optimal combination of reductant and acid at low temperature the (P₃^B)Fe system can catalyze N₂RR with selectivity for NH₃ approaching 80%. Also, studies of these iron systems have underscored a role for PCET in N₂RR catalysis (Figure 47); such processes may be operative in other systems, including Mo. A number of other Mo and Fe catalysts were also discussed, including a structurally unusual bis(carbene)-iron system, (CAAC)₂Fe (Figure 55), whose efficacy for N₂RR catalysis becomes apparent only at −95 °C, and also a bis-diphosphine-iron that is selective for N₂H₄ instead of NH₃ (Figure 59).

Recent studies have been undertaken with a host of other metals, most of which (except V) are not found in nitrogenases. These have included Co, Ru, Os, V, and Ti. Particular emphasis has been placed here on the mechanistic understanding that can be obtained via comparisons of these systems to the Mo or Fe catalysts discussed in the earlier sections. A discussion of historic and more recent efforts toward electrocatalytic nitrogen fixation with transition metal complexes has also been presented, including a discussion of the first (and still only) molecular system, (P₃^B)Fe⁺, that mediates bona fide N₂RR electrocatalysis (Figure 72). Other systems that show progress here, and especially those featuring main group (e.g., Al/B) rather than transition metal active sites, were also highlighted (Figure 74).

Lastly, in an effort to highlight some of the lessons drawn from this review, a bulleted discussion of specific insights important to future catalyst designs and, also, remaining challenges is presented. We hope these ideas help researchers devise new catalysts, and reagent/condition combinations, to continue the rapid progress now being made in this important realm of catalysis.

Figure 23.

A diamido(pyridine) pincer ligand used by the Schrock group for Mo-catalyzed N₂RR, Ar = 2,6-diisopropylphenyl.¹⁴⁰

Figure 53.

Reproduced with permission from ref 209. Copyright 2018 American Chemical Society. Gibbs free enthalpy, ΔG°, scheme of the (P₃^Si)Fe cycle calculated with the B3LYP functional and the def2-TZVP basis set.

Biographies

Dr. Matthew J. Chalkley received his B.S. from Yale University with exceptional distinction in Chemistry in 2013. There he undertook research with Prof. Nilay Hazari on Pd^I dimers as a Barry Goldwater Scholar. He then received a Fulbright Fellowship to perform research in the group of Prof. Karsten Meyer at the University of Erlangen-Nuremberg for one year. His Ph.D. was supervised by Prof. Jonas Peters at Caltech and supported by an NSF Graduate fellowship, a Caltech Environmental Microbial Interactions fellowship, and a Resnick Sustainability Institute fellowship. He is currently a postdoctoral scholar with Prof. William DeGrado at the University of California, San Francisco. His interests include catalysis and bioinorganic and organometallic chemistry.

Prof. Marcus W. Drover received his B.Sc. (Hons.) degree in 2012 from Memorial University of Newfoundland. In 2016, he earned his doctorate in chemistry from the University of British Columbia (UBC) under the cosupervision of Profs. Laurel L. Schafer and Jennifer A. Love. At UBC, Marcus was an NSERC Vanier scholar (2013–2016) and in 2016 was awarded the Chemical Institute of Canada’s Award for Graduate Research in Inorganic Chemistry. In 2014, Marcus was also awarded a Michael-Smith visiting fellowship for research abroad at the University of Oxford under the supervision of Prof. Andrew S. Weller. In January 2017, Marcus began work as a joint Resnick Prize Postdoctoral Fellow in Sustainability Science and NSERC Banting Fellow at Caltech under the guidance of Prof. Jonas C. Peters. Currently, Marcus is an assistant professor at the University of Windsor in Ontario, Canada. His interests include inorganic, organometallic, and main group chemistry.

Prof. Jonas C. Peters was born in 1971 in Chicago, Illinois. In 1993, he received his Bachelor of Science degree in chemistry at the University of Chicago, where doing undergraduate research with Prof. Gregory Hillhouse first piqued his interest in the transition metal chemistry of nitrogenous ligands such as HN=NH and HNO. Jonas then spent a year as a Marshall Scholar at the University of Nottingham, UK, working with Prof. James J. Turner, FRS, detecting short-lived transients by rapid time-resolved methods. In the fall of 1994, Jonas began his doctoral studies under the direction of Prof. Christopher C. Cummins at the Massachusetts Institute of Technology. Jonas’ research focused on the activation and functionalization of small molecules using low coordinate tris-amido molybdenum and titanium complexes. After receiving his Ph.D. in inorganic chemistry in 1998, Jonas was a Miller Fellow at the University of California, Berkeley under the guidance of Prof. T. Don Tilley, where he began to design and synthesize new phosphine ligands for metal-mediated bond activation reactions. Jonas began as Assistant Professor in the Division of Chemistry and Chemical Engineering at Caltech in August of 1999, was promoted to Associate Professor in 2004, and to Professor of Chemistry in 2006. In July of 2007, he relocated to the MIT Department of Chemistry as the W. M. Keck Professor of Energy, and then returned to Caltech in January 2010 as Bren Professor of Chemistry. He is also the Director of the Resnick Sustainability Institute at Caltech. His research focuses on the structure, bonding, and catalytic activity of inorganic compounds, particularly ones of relevance to global C, N, and O cycles and renewable energy applications.