Tripodal P₃^XFe–N₂ Complexes (X = B, Al, Ga): Effect of the Apical Atom on Bonding, Electronic Structure, and Catalytic N₂-to-NH₃ Conversion

Abstract

Terminal dinitrogen complexes of iron ligated by tripodal, tetradentate P₃^X ligands (X = B, C, Si) have previously been shown to mediate catalytic N₂-to-NH₃ conversion (N₂RR) with external proton and electron sources. From this set of compounds, the tris(phosphino)borane (P₃^B) system is most active under all conditions canvassed thus far. To further probe the effects of the apical Lewis acidic atom on structure, bonding, and N₂RR activity, Fe–N₂ complexes supported by analogous group 13 tris(phosphino)alane (P₃^Al) and tris(phosphino)gallane (P₃^Ga) ligands are synthesized. The series of P₃^XFe–N₂^[0/1−] compounds (X = B, Al, Ga) possess similar electronic structures, degrees of N₂ activation, and geometric flexibility as determined from spectroscopic, structural, electrochemical, and computational (DFT) studies. However, treatment of [Na(12-crown-4)₂][P₃^XFe–N₂] (X = Al, Ga) with excess acid/reductant in the form of HBAr^F₄/KC₈ generates only 2.5 ± 0.1 and 2.7 ± 0.2 equiv of NH₃ per Fe, respectively. Similarly, the use of [H₂NPh₂][OTf]/Cp*₂Co leads to the production of 4.1 ± 0.9 (X = Al) and 3.6 ± 0.3 (X = Ga) equiv of NH₃. Preliminary reactivity studies confirming P₃^XFe framework stability under pseudocatalytic conditions suggest that a greater selectivity for hydrogen evolution versus N₂RR may be responsible for the attenuated yields of NH₃ observed for P₃^AlFe and P₃^GaFe relative to P₃^BFe.

Graphical Abstract

INTRODUCTION

While it was not long ago that well-defined molecular N₂-to-NH₃ conversion (N₂RR) catalysts were scarce and limited to Fe- and Mo-based transition metal complexes,^1–3 significant progress in the development of N₂RR catalysts in recent years has resulted in their increase in number, diversity, and efficiency.⁴ These advancements have largely been made possible by a combination of systematic reactivity studies, in-depth mechanistic investigations, and structure–activity surveys. Notably, the latter have been instrumental in improving the performance of Fe and Mo N₂RR catalysts, as well as in the discovery of new catalytic systems based on other metals. For instance, building upon their first report on (PNP)Mo-mediated N₂RR,² Nishibayashi and co-workers have performed systematic pincer ligand variations to develop a number of related (PXP)Mo complexes (X = N, P, C) capable of catalyzing N₂RR with more mild proton and electron sources, higher turnover numbers, and/or greater selectivities.⁵ Similarly, metal center substitution of known Fe N₂-to-NH₃ conversion catalysts by V⁶ or the nonbiologically relevant metals Co,⁷ Ru,⁸ and Os⁸ has led to new catalytic systems supported by the same ligand platforms and featuring comparable N₂RR efficiencies. These systematic comparative studies (see Chart 1 for group 8 systems) are a critical component to gaining a better understanding of key N₂RR catalyst design principles.

Chart 1.

Group 8 Metallotrane Systems for Exploring Factors That Impact Catalytic N₂RR Efficacy

In this context, extensive investigations carried out by our group on the dinitrogen coordination chemistry and N₂RR catalysis by tripodal P₃^X-ligated Fe complexes (P₃^B = (o-ⁱPr₂PC₆H₄)₃B; P₃^C = [(o-ⁱPr₂PC₆H₄)₃C]⁻; P₃^si = [(o-ⁱPr₂PC₆H₄)₃Si]⁻) have established that the identity of the apical atom (X) influences the observed N₂RR versus HER (hydrogen evolution reaction) selectivity.^3,9–13 Notably, from this series, the P₃^BFe system is the most efficient N₂RR catalyst under all conditions canvassed thus far, highlighting the privileged role of the group 13 Lewis acidic borane. This enhanced efficacy is in part attributed to the electron-accepting nature and geometric flexibility of the Fe → B interaction, which helps stabilize both low-valent and Fe–N multiply bonded species proposed to be intermediates of a P₃^BFe-mediated N₂ fixation cycle.^14–20

Given the importance of the boron center, a common question that has surfaced concerns the effect of X(III) apical atom substitution in the P₃^XFe framework, with respect to bonding, electronic structure, and N₂RR activity (Chart 1). To address this issue, herein we report the synthesis and characterization of a series of P₃^XFe–Br and P₃^XFe–N₂^[0/1−] compounds supported by analogous group 13 tris(phosphino)alane (P₃^Al = (o-ⁱPr₂PC₆H₄)₃Al)²¹ and tris(phosphino)gallane (P₃^Ga = (o-ⁱPr₂PC₆H₄)₃Ga)²² ligands. Structural, spectroscopic, electrochemical, and computational studies reveal that all three P₃^XFe systems (X = B, Al, Ga) possess similar electronic structures, degrees of N₂ activation, and geometric flexibility. However, P₃^AlFe and P₃^GaFe display significantly lower N₂RR efficiencies than P₃^BFe. Preliminary reactivity studies show that all P₃^XFe frameworks are robust under the catalytic conditions employed, suggesting that a greater selectivity for competing HER may be responsible for the attenuated NH₃ yields observed for the former.

RESULTS AND DISCUSSION

Synthesis and Characterization of P₃^XFe–N₂^[0/1−] Complexes.

As an entry point into the desired P₃^XFe dinitrogen chemistry, we sought to prepare an Fe(I) halide precursor similar to our approach in the related P₃^BFe system. Whereas P₃^BFe–Br (2a) can readily be prepared by heating a mixture of P₃^B (1a), Fe(0), and FeBr₂ to 90 °C in THF,¹⁴ P₃^Al (1b) and P₃^Ga (1c) proved thermally unstable under these conditions. Instead, stirring 1b and 1c with FeBr₂ in benzene, followed by one-electron reduction with sodium amalgam (Na(Hg)), afforded the corresponding P₃^XFe–Br complexes 2b and 2c as bright green solids in moderate isolated yield (Scheme 1). Like P₃^BFe–Br, both P₃^XFe–Br derivatives feature paramagnetically shifted ¹H NMR resonances, solution magnetic moments of 4.3 μ_B (2b) and 4.2 μ_B (2c), and low-temperature rhombic EPR spectra (toluene glass, 10 K), consistent with an S = 3/2 ground state.

Scheme 1.

Synthesis of P₃^XFe–N₂^[0/1−] Complexes

The long Fe–P bond lengths in the solid-state structures of 2b (2.4603(3) Å) and 2c (2.4690(3) Å) are also consistent with this high-spin configuration assignment (Figure 1). It is noteworthy that the Fe–Br bond in 2b and 2c remains intact, as previous examples of P₃^Al and P₃^Ga coordination to CuCl and AuCl(SMe₂) have instead yielded the zwitterionic complexes P₃^Al–ClCu and P₃^X–ClAu (X = Al, Ga) resulting from halide abstraction by the apical X atom.^21,22 For comparison, reaction of P₃^B with these metal precursors instead yields P₃^BCu–Cl and P₃^BAu–Cl.²³ These differences in reactivity suggest that Al(III) and Ga(III) are stronger Lewis acids than B(III) within the P₃^X framework.

Figure 1.

Solid-state structures of 2b, 3b, and 4b with thermal ellipsoids set at 50% probability. Hydrogen atoms and cocrystallized solvents are omitted for clarity. The XRD structures of 2c, 3c, and 4c are analogous and are provided in the Supporting Information. See Table 1 for select structural data.

Reduction of 2b and 2c with Na(Hg) in benzene under a nitrogen atmosphere results in the formation of mononuclear dinitrogen complexes of the form P₃^XFe–N₂ (X = Al (3b), X = Ga (3c)) (Scheme 1). 3b and 3c display intense, diagnostic infrared (IR) absorption bands at ν(NN) = 2003 and 1997 cm⁻¹, respectively, where the end-on N₂ ligand is only slightly more activated than in the related P₃^BFe–N₂ (3a, 2011 cm⁻¹)¹⁴ and iron alumatrane (AltraPhos)Fe–N₂ (6, 2010 cm⁻¹, AltraPhos = Al[N(o-C₆H₄NCH₂PⁱPr₂)₃])²⁴ complexes and is moderately labile under vacuum.

The terminal nature of the N₂ moiety was further corroborated by X-ray diffraction (XRD) analysis of single crystals of 3b and 3c (Figure 1). The dinitrogen ligand in 3b and 3c exhibits a d(N–N) of 1.0616(17) and 1.118(6) Å, respectively, consistent with minimal N₂ activation. In the solid state, 3b and 3c feature two short (3b: 2.3781(4), 2.3843(4) Å; 3c: 2.3898(5), 2.3969(3) Å) and one comparable (3b: 2.4220(4) Å; 3c: 2.4342(5) Å) Fe–P contact compared to 2b and 2c. This contraction in the Fe–P bonds is in line with increased π-backdonation from the Fe center to the phosphines upon reduction, as well as the S = 1 ground state adopted by 3b (3.1 μ_B) and 3c (2.8 μ_B) in room temperature C₆D₆ solutions. XRD characterization of 3b and 3c allows for a structural comparison across the series of compounds 2–4 (vide infra) that span the formal oxidation states Fe(I) to Fe(−I). This contrasts with the P₃^BFe system, where the XRD structure of 3a could not be authenticated due to severe twinning.¹⁴

Treatment of P₃^XFe–Br or P₃^XFe–N₂ with excess Na(Hg) in the more polar solvent THF leads exclusively to the anionic S = 1/2 complex [Na(THF)₃][P₃^XFe–N₂] (X = Al (4b), X = Ga (4c)) (Scheme 1). Like [Na(THF)₃][P₃^BFe–N₂] (4a), 4b and 4c feature a phase-dependent Lewis acid–base interaction between N_β and the Na⁺ counterion that can be monitored by IR spectroscopy.

Whereas in the solid state 4b and 4c exhibit a single ν(NN) at 1883 and 1879 cm⁻¹, in THF two bands (4b: 1879, 1922 cm⁻¹; 4c: 1878, 1920 cm⁻¹) are observed as a result of solvation of the Na⁺ ion to yield some population of the free anion P₃^XFe–N₂⁻. Encapsulation of Na⁺ by treatment of 4b and 4c with two equivalents of 12-crown-4 (12-c-4) yields [Na(12-c-4)₂][P₃^AlFe–N₂] (5b) and [Na(12-c-4)₂][P₃^GaFe–N₂] (5c), respectively, which give rise to only the higher energy vibration. Compound 4a (1877, 1918 cm⁻¹),¹⁴ compound 5a (1918 cm⁻¹),¹⁴ and the reduced iron alumatrane, [K(18-crown-6)][(AltraPhos)Fe–N₂] (7, 1925 cm⁻¹),²⁵ display similar degrees of N₂ activation to 5b and 5c.

The intimate ion pair in salts 4b and 4c was also authenticated by their corresponding solid-state structure, where the capping Na⁺ is coordinated by three additional THF molecules (Figure 1). As was the case in going from the formally Fe(I) to Fe(0) complexes, the Fe–P bond lengths in 4b (2.2751(6) Å) and 4c (2.2824(5) Å) are significantly contracted from those of 3b and 3c, respectively. Increased π-backbonding into the N₂ unit also leads to shorter Fe–N (4b: 1.769(3) Å; 4c: 1.759(5) Å) and longer N–N (4b: 1.134(4) Å; 4c: 1.141(4) Å) distances in 4b and 4c. Complexes 4a (d(Fe–N) = 1.775 Å; d(N–N) = 1.149 Å) and 7 (d(Fe–N) = 1.783 Å; d(N–N) = 1.135 Å) possess comparable Fe–N₂ bond metrics, in line with their similar IR profiles.

Electronic Structure and Bonding in P₃^XFe Systems.

While isostructural alkyl (X = C),^9,10 silyl (X = Si),¹⁰ and borane (X = B)^3,10,11,13 P₃^XFe–N₂⁻ complexes have all demonstrated the capacity to facilitate catalytic N₂RR, the P₃^BFe system affords the highest efficiencies under all conditions canvassed. This can be attributed, in part, to the greater electronic and geometric flexibility of the dative Fe → B interaction, which allows access to both reduced Fe–N₂ and Fe–N multiply bonded species that may be traversed along an Fe-mediated nitrogen fixation cycle. Storage of additional electron density in the Fe–B manifold stabilizes the low-valent states, and shuttling of the iron center between trigonal bipyramidal and pseudotetrahedral geometries affords access to intermediates with different electronic structures.^14–20 Accordingly, the catalytically relevant species Fe–N₂^{[0/1−/2−]},^14,18 Fe═NN₂H₂^{[1+/0]17–19}, and Fe≡N^[1+]18 have proven accessible on the P₃^B scaffold. Given the significant role of the Fe → B interaction in this respect, we sought to determine whether the Fe→Al and Fe→Ga bonds would behave similarly.

The bonding and electronic structures of P₃^BFe–N₂^[0/1−] and (AltraPhos)Fe–N₂^[0/1−] have previously been elucidated through spectroscopic and computational methods.^14,24–26 In short, the bonding in these complexes is best described as involving an Fe → X(III) dative interaction where the apical Lewis acidic atom serves to lower the energy of the σ(Fe–X) orbital of Fe ${\text{d}}_{z^2}$ and X p_z parentage below that of the ${\text{d}}_{xy,x^2-y^2}$ set in a typical trigonal bipyramidal orbital scheme. Therefore, reduction of 3 to 4a/5a and 6 to 7 is primarily Fe centered, giving rise to formally Fe(−I) species with electronic configurations ${\left({\text{d}}_{xz,yz}\right)}^4{(\sigma (\text{Fe}-\text{X}))}^2{\left({\text{d}}_{xy,x^2-y^2}\right)}^3$ and a substantial Fe-to-X backbond.²⁷

A consequence of having three electrons in a set of degenerate orbitals of d_xy and ${\text{d}}_{x^2-y^2}$ parentage is a Jahn–Teller active state. Distortion from C₃ symmetry is clearly reflected by the asymmetry of the P–Fe–P angles in the solid-state structures of 4a (two independent molecules present in asymmetric unit; 107.3°, 110.3°, 134.6° and 113.4°, 114.1°, 124.7°), 5a (105.4°, 112.3°, 135.0°), and 7 (107.1°, 111.9°, 131.1°). In contrast, 4b and 4c crystallize in the trigonal space group R3, being 3-fold symmetric about the apical axis defined by the X–Fe–N₂ unit. While this may suggest that 4b and 4c possess an electronic structure that differs from that of 4a/5a, density functional theory (DFT) geometry optimization and single-point energy calculations on the anions P₃^AlFe–N₂⁻ and P₃^GaFe–N₂⁻ reveal a relaxation to a Jahn–Teller distorted configuration with computed P–Fe–P angles of 111°, 112°, 125° and 111°, 112°, 126°, respectively. Corresponding spin density plots show that this distortion also arises from localization of the unpaired electron in an Fe-based d-orbital of xy symmetry (Figure 2; Mulliken spin density on Fe: X = B = 1.08; X = Al = 1.09; X = Ga = 1.10). X-band EPR spectra of 4b and 4c collected in 2-methyltetrahydrofuran glass (77 K) display axial signals with no discernible hyperfine coupling to ²⁷Al or ^69/71Ga nuclei. Negligible spin leakage onto the B, Al, and Ga atoms is detected computationally. In conjunction with the measured solution magnetic susceptibilities and computed frontier molecular orbital schemes (Figure 2), these data indicate that P₃^XFe–N₂^[0/1−] (X = Al, Ga) possess electronic structures similar to those of P₃^BFe–N₂^[0/1−] and (AltraPhos)Fe–N₂^[0/1−].^14,24–26 The perfect trigonal symmetry exhibited by 4b and 4c in the solid state is likely the result of crystal packing effects.²⁸

Figure 2.

DFT-computed molecular orbital diagram (α-manifold, isovalue = 0.06 au) for P₃^AlFe–N₂⁻ (top) and spin density plots (isovalue = 0.003 au) for P₃^AlFe–N₂⁻ (bottom left) and P₃^GaFe–N₂⁻ (bottom right). Hydrogen atoms are omitted for clarity. The calculated frontier molecular orbitals for P₃^GaFe–N₂⁻ are analogous.

Having concluded that significant Fe → X σ-backdonation is operative in P₃^AlFe–N₂^[0/1−] and P₃^GaFe–N₂^[0/1−], we next sought to determine the strength of this interaction relative to that in P₃^BFe–N₂^[0/1−].²⁷ To do so, we explored the redox behavior of 3b and 3c by cyclic voltammetry. 3b and 3c display reversible Fe–N₂^[0/1−] redox couples centered at −1.97 V and −1.99 V vs Fc^[1+/0] (Fc = ferrocene), respectively (Figure 3). Under identical conditions, the analogous reduction event for 3a occurs at the slightly more negative potential of −2.19 V.¹⁴ The scan rate dependence of all three couples is consistent with a diffusion-controlled process. For comparison, the observed trend in the P₃^XFe–N₂^[0/1−] redox potentials is also in accordance with the theoretically^29,30 and experimentally³¹ determined acceptor ability of XPh₃ according to the order X = Al > Ga > B.

Figure 3.

Cyclic voltammograms depicting the P₃^XFe–N₂^[0/1−] redox couple of 3a (red), 3b (blue), and 3c (green) at a scan rate of 40 mV/s in THF with 0.1 M [Buⁿ₄N][PF₆] supporting electrolyte.

The ca. 200 mV anodic shift in the Fe–N₂^[0/1−] reduction potential for 3b and 3c relative to 3a suggests that exchange of the apical Lewis acidic element in the P₃^X scaffold results in only modest electronic perturbations at the iron center. This is in line with the ν(NN) vibrational data for P₃^XFe–N₂^[0/1−] complexes, which reveal similar degrees of N₂ activation, as well as with electronic structure calculations, which show that the Fe → X interaction directly lowers the energy of the Fe ${\text{d}}_{z^2}$ orbital but only indirectly influences the energy of the singly occupied molecular orbital (SOMO) through σ-inductive withdrawal of electron density from the Fe center. Interestingly, the reversible (AltraPhos)Fe–N₂^[0/1−] reduction wave occurs at −2.08 V vs Fc^[1+/0],²⁴ suggestive of a weaker Fe → Al interaction than that in 3b. We attribute this difference to competition between both Fe and N_apical for donation into the empty Al p_z orbital in the former.

To determine whether geometric flexibility is also conserved in the P₃^AlFe and P₃^GaFe platforms, we turned our attention to the XRD structures of 2–4. With the exception of the Fe → X interaction, all isostructural complexes exhibit comparable bond distances and angles, irrespective of the identity of the apical Lewis acidic atom (Table S3). However, inspection of the Fe–X distances reveals that all three systems possess similar degrees of flexibility (Table 1). The overall changes in these axial bond lengths upon reduction of P₃^XFe–Br to [Na(THF)₃][P₃^XFe–N₂] are 0.149 Å (X = B), 0.177 Å (X = Al), and 0.177 Å (X = Ga). Moreover, the value calculated for the ratio (r) of the Fe–X bond length to the sum of the respective covalent radii,^32,33 which accounts for the differing sizes of B, Al, and Ga, bares an overall net difference of Δr = 0.07 in each case, suggesting that the Fe–X bond is equally flexible in all three systems.

Table 1.

Select Structural Data for Complexes 2–4, 6, and 7

	Fe–Xa	∑C–X–Cb	∑P–Fe–Pb	rc
2ad	2.458	341.8	342.7	1.14h
3ae	2.417	332.8	345.8	1.12
4ad	2.309	330.1	352.3	1.07
2b	2.662	340.8	344.8	1.05h
3b	2.539	336.2	347.8	1.00
4b	2.485	334.8	349.6	0.98
2c	2.666	341.9	345.8	1.05h
3c	2.544	337.8	348.7	1.00
4c	2.489	335.8	350.4	0.98
6f	2.809	351.6	335.0	1.11
7g	2.574	344.4	350.1	1.02

^aUnits of angstroms (Å).

^bUnits of degrees (°).

^cRatio of the Fe–X bond length to the sum of the covalent radii (Fe: 1.32 Å; B: 0.84 Å; Al: 1.21 Å; Ga: 1.22 Å; from ref 32).

^dFrom ref 14. For 4a, values are an average of two independent molecules in the asymmetric unit.

^eValues obtained from a DFT geometry optimized structure.

^fFrom ref 24. In solution, the terminal dinitrogen species (AltraPhos)Fe–N₂ is believed to be in equilibrium with its dinitrogen-bridged analogue, [(AltraPhos)Fe]₂(μ-N₂). The structural data presented here is for the latter complex.

^gFrom ref 25.

^hSee footnote in ref 33.

The decreasing values of r in the order 2 > 3 > 4 are also indicative of a stronger Fe–X bond upon reduction.³⁴ As expected, such an increase in the dative interaction is accompanied by a greater pyramidalization at the X(III) apical atom and more trigonal planar geometry about Fe (Table 1). Interestingly, while the r values of P₃^AlFe and P₃^Ga Fe are consistently lower than those of P₃^BFe, alluding to a stronger Fe → Al and Fe → Ga interaction relative to Fe → B, the slightly more pyramidalized geometry about boron across the series would seem to imply it forms the strongest Fe–X bond. This discrepancy likely stems from the difference in atomic sizes between B, Al, and Ga, as well as steric constraints imposed by the cage structure. This has also been observed in related P₃^XPd compounds (X = B, In), where the larger size of In prevents pyramidalization that accurately reflects the strength of the donor–acceptor interaction. While computations and the smaller r = 0.93 value for P₃^InPd indicate it has a stronger Pd → X interaction than P₃^BPd (r = 1.01), the indium center resides in a less pyramidalized environment than boron (∑C–In–C = 354.9°; ∑C–B–C = 341.8°).^23,35 Therefore, in accord with the electrochemical data, we favor a bonding scheme where the Fe → X donor–acceptor interaction is stronger for X = Al, Ga than for X = B. This assignment is also validated computationally, where normalized Löwdin bond orders (with respect to the corresponding Fe–B bond) of 1.08, 1.05, 1.05, and 1.02 are calculated for 3b, 3c, P₃^AlFe–N₂⁻, and P₃^GaFe–N₂⁻, respectively.

N₂RR Activity.

Having concluded that the P₃^XFe platforms (X = B, Al, Ga) exhibit similar electronic structures, flexibility, and degrees of N₂ activation, we reasoned that 5b and 5c might function as competent catalysts for N₂RR. Thus, we subjected these compounds to the catalytic conditions that have proved most fruitful for the P₃^BFe system, namely, treatment with excess [H(OEt₂)₂][BAr^F₄] (HBAr^F₄, BAr^F₄⁻ = tetrakis(3,5-bis(trifluoromethyl)phenyl)borate)/KC₈^3,10 or [H₂NPh₂][OTf]/Cp*₂Co¹¹ as the acid/reductant source at −78 °C in diethyl ether (Et₂O). Under the former conditions, 5b and 5c generate 2.5 ± 0.1 (17 ± 1% selectivity based on H⁺) and 2.7 ± 0.2 (17 ± 1% efficiency) equiv of NH₃ per Fe, respectively (Table 2, entries 2 and 3). Moving to the [H₂NPh₂][OTf]/Cp*₂Co combination, which is in the activity regime with the highest N₂-to-NH₃ selectivity observed for P₃^BFe,¹³ results in only modest improvements, with 5b and 5c producing 4.1 ± 0.9 (27 ± 6% based on acid) and 3.6 ± 0.3 (24 ± 2% efficiency) equiv of NH₃ per Fe (Table 2, entries 5 and 6). Hydrazine is not detected under either set of conditions. These efficiencies are appreciably lower than those of P₃^BFe under similar conditions (Table 2, entries 1 and 4),^3,11 and suggest that, in addition to the extent of N₂ activation and flexibility of the Fe–X linkage, other factors associated with the identity of the apical atom in the P₃^X scaffold play an influential role in the nitrogen fixation process.

Table 2.

N₂RR Mediated by [Na(12-c-4)₂][P₃^XFe–N₂] Complexes^a

	cat	acid (equiv)	reductant (equiv)	NH3/Fe (equiv)	Yield NH3/H+ (%)
1b	5a	46d	50e	7.0 ± 1.0	46 ± 7
2	5b	46d	50e	2.5 ± 0.1	17 ± 1
3	5c	46d	50e	2.7 ± 0.2	17 ± 1
4c	-	54f	108g	12.8 ± 0.5	72 ± 3
5	5b	46f	50g	4.1 ± 0.9	27 ± 6
6	5c	46f	50g	3.6 ± 0.3	24 ± 2

^aCatalyst, acid, reductant, and Et₂O sealed in a Schlenk tube at −196 °C under an N₂ atmosphere, warmed to −78 °C, and stirred. For runs utilizing HBAr^F₄, reactions were stirred at −78 °C for 1 h, followed by stirring at room temperature for 45 min. For other runs, reactions were allowed to stir and warm to room temperature overnight. Yields are reported as an average of three runs.

^bFrom ref 3.

^cFrom ref 11. P₃^BFe⁺ was used as the precatalyst.

^dHBAr^F₄.

^eKC₈.

^f[H₂NPh₂][OTf].

^gCp*₂Co.

In order to discern whether the divergent N₂RR catalytic profiles of 5a–c are a consequence of catalyst deactivation/decomposition, the Fe speciation present after exposure of the complexes to 10 equiv of acid and 12 equiv of reductant was examined. P₃^BFe is known to be rather robust under both sets of catalytic conditions, with substantial active (pre)catalyst remaining at the conclusion of reactions as evidenced by Mössbauer spectroscopic studies and substrate reloading experiments.^10,11 Indeed, NMR and IR analysis of the postreaction of 5a with HBAr^F₄/KC₈ reveals [M(solv)_x][P₃^BFe–N₂] (solv = solvent) and (P₃^B)(μ-H)Fe(L)(H) (L = H₂, N₂) as the major Fe-containing products. Similarly, P₃^BFe–N₂ and (P₃^B)(μ-H)Fe(N₂)(H) are obtained when [H₂NPh₂][OTf]/Cp*₂Co are employed (see Supporting Information for full procedures). Free or decomposed P₃^B is not observed in either instance. It is worth noting that while the hydride-borohydride complex (P₃^B)(μ-H)Fe(L)(H) is an off-path species of P₃^BFe-mediated N₂RR catalysis, in the presence of protons and electrons it can revert back to P₃^BFe–N₃⁻ and serve as a competent precatalyst.¹⁰

Analogous experiments utilizing 5b and 5c yield slightly different results depending on the conditions. In the case of reaction with 10 equiv of HBAr^F₄ and 12 equiv of KC₈, [M(solv)_x][P₃^XFe–N₂] (X = Al, Ga) is the major terminal Fe-containing product, with very little ligand decomposition observed by ³¹P{¹H} NMR spectroscopy. On the other hand, use of [H₂NPh₂][OTf] and Cp*₂Co yields a mixture of compounds that includes P₃^XFe–N₂ (X = Al, Ga) and products of P₃^XFe decomposition. The latter is evidenced by the presence of an intense ³¹P{¹H} NMR signal consistent with uncoordinated phosphine (see Supporting Information).

While the observation of substantially more decomposition when 5b and 5c are reacted with [H₂NPh₂][OTf]/Cp*₂Co appears to be at odds with the milder nature of these reagents, and the higher NH₃ yields obtained relative to HBAr^F₄/KC₈ in catalytic reactions, this can be rationalized by the differing strength of the reductants and reactivity of P₃^XFe–N₂ species with H₂. Because of the highly reducing nature of KC₈ (E ≤ −3.0 V vs Fc^[1+/0]), and its slight excess in the reactions, after complete acid consumption, any residual KC₈ is anticipated to reduce P₃^XFe–N₂ to P₃^X Fe–N₂⁻, which is unreactive toward H₂. Additionally, conversion of (P₃^B)(μ-H)Fe(L)(H) to P₃^BFe–N₂⁻ is viable under the HBAr^F₄/KC₈ conditions.¹⁰ In contrast, the reduction potential of Cp*₂Co (E = −1.96 V vs Fc^[1+/0]) is very close to that of the P₃^XFe–N₂^[0/1−] couples (X = B, Al, Ga). While Cp*₂Co is capable of reducing P₃^BFe–N₂ to P₃^BFe–N₂⁻ at −78 °C,¹¹ reaction of P₃^BFe–N₂ with excess Cp*₂Co at room temperature does not produce P₃^BFe–N₃⁻ (as judged by IR spectroscopy). Such observations are consistent with a temperature-dependent redox equilibrium analogous to that reported for P₃^SiOs–N₂^[0/1−] (E = −1.97 V vs Fc^[1+/0]).⁸ As a consequence, upon allowing the catalyst speciation reactions to warm and stir at room temperature, P₃^XFe–N₂ (rather than [M(solv)_x][P₃^XFe–N₂]) is expected to be the dominant species in solution when Cp*₂Co is employed as the reductant. Reaction of P₃^XFe–N₂ with H₂ formed from background and/or catalyzed HER may therefore account for the remaining terminal products observed in the reaction of 5a–c with [H₂NPh₂][OTf]/Cp*₂Co. P₃^BFe–N₂ is known to react with H₂ to cleanly generate (P₃^B)(μ-H)Fe(L)(H) (L = H₂, N₂).³⁶ In turn, treatment of P₃^AlFe–N₂ and P₃^GaFe–N₂ with H₂ gives rise to product profiles that match those observed at the end of catalysis model reactions (see Supporting Information).

The above results suggest that P₃^AlFe and P₃^GaFe platforms are relatively robust under both sets of N₂RR catalytic conditions explored. Although minor catalyst decomposition/deactivation does occur (and is anticipated to be slower at −78 °C), this does not correlate well with the dramatically lower N₂RR efficiencies of P₃^AlFe and P₃^GaFe versus P₃^BFe, or with the higher NH₃ yields obtained with the [H₂NPh₂][OTf]/Cp*₂Co conditions. Instead, our findings are suggestive that P₃^AlFe and P₃^GaFe exhibit a greater selectivity for HER versus N₂RR. Bimolecular coupling of Fe═NNH₂^[1+/0] N₂RR intermediates featuring weak N–H bonds is predicted to be an operative unproductive HER pathway for the P₃^BFe system.¹² In this respect, it is worth noting that Fe-imido (Fe≡N—R; R = adamantyl,¹⁷ p-methoxyphenyl¹⁴) and Fe-disilylhydrazido (Fe═NNSi₂)^15,25 complexes that are accessible and stable on the P₃^BFe and (AltraPhos)Fe platforms could not be isolated on the P₃^AlFe and P₃^GaFe scaffolds despite our attempts to do so.

CONCLUSION

In summary, the series of P₃^AlFe and P₃^GaFe complexes 2–4 spanning the formal oxidation states Fe(I) to Fe(−I) have been synthesized, isolated, and thoroughly characterized. These complexes expand upon and complement known P₃^XFe (X = B, C, Si) N₂RR catalysts. While P₃^AlFe–N₂^[0/1−] and P₃^GaFe–N₂^[0/1−] compounds exhibit degrees of N₂ activation, flexibility in their Fe → X interactions, and overall electronic structures similar to those of the P₃^BFe system, 5b and 5c display significantly attenuated yields of ammonia under analogous N₂RR catalytic conditions. Preliminary reactivity studies confirm that P₃^AlFe and P₃^GaFe are relatively robust under these conditions, suggesting that the lower NH₃ turnover numbers observed are likely the result of greater HER versus N₂RR selectivity, as opposed to differences in core–framework stability.