Laurence R.
Doyle
^{a},
Daniel J.
Scott
^{a},
Peter J.
Hill
^{a},
Duncan A. X.
Fraser
^{a},
William K.
Myers
*^{b},
Andrew J. P.
White
^{a},
Jennifer C.
Green
^{b} and
Andrew E.
Ashley
*^{a}
^{a}Department of Chemistry, Imperial College London, London SW7 2AZ, UK. E-mail: a.ashley@imperial.ac.uk
^{b}Inorganic Chemistry Laboratory, University of Oxford, Oxford OX1 3QR, UK
First published on 18th July 2018
The synthesis and characterisation of the S = 1/2 Fe(I) complex [Fe(depe)_{2}]^{+}[BAr^{F}_{4}]^{−} ([1]^{+}[BAr^{F}_{4}]^{−}), and the facile reversible binding of N_{2} and H_{2} in both solution and the solid state to form the adducts [1·N_{2}]^{+} and [1·H_{2}]^{+}, are reported. Coordination of N_{2} in THF is thermodynamically favourable under ambient conditions (1 atm; ΔG_{298} = −4.9(1) kcal mol^{−1}), while heterogenous binding is more favourable for H_{2} than N_{2} by a factor of ∼300. [1·H_{2}]^{+}[BAr^{F}_{4}]^{−} represents a rare example of a well-defined, open-shell, non-classical dihydrogen complex, as corroborated by ESR spectroscopy. The rapid exchange between N_{2} and H_{2} coordination under ambient conditions is unique for a paramagnetic Fe complex.
(1) |
The bold symbols are 3 × 3 tensors (or matrices) and the vector of Cartesian spin operators are defined in appropriate Hilbert space eigenbasis. The third term of the spin Hamiltonian, hyperfine or is composed of an isotropic component, a_{iso}, transformed to a 3 × 3 matrix by, defined as the identity matrix, and an anisotropic component, T.
(2) |
It is the anisotropic tensor component that returns the structural relations of the spin system, based on the summation of dipolar interactions between the position vectors of the central iron atom and surrounding positions, approximated as ligand nuclei position vectors.^{9}
(3) |
Eqn (3) includes contribution of all centres, p_{j}, of spin density f_{j} at distance r_{ij} to the hyperfine interaction matrix for a single nucleus position, p_{i}. The principal axis elements in the limit of an axial tensor are [−T, −T, 2T] where the perpendicular component (eqn (4)) is used to estimate a distance in a single point-dipole approximation:
T = f_{Fe}g_{eff}μ_{B}g_{N}μ_{N}/r^{3} | (4) |
Analysis of the nuclear quadrupole term of the spin Hamiltonian of eqn (1) for an I = 1 nucleus (relevant to ^{2}H and ^{14}N in this work) can provide useful information on bonding and electronic structure,^{10} and is given by:
Q = K × diag[−(1 − η), −(1 + η), 2] | (5) |
ν_{dq} = 2[ν_{ef±}^{2} + K^{2}(3 + η^{2})]^{1/2} | (6) |
The [1]^{+} cation was found to be disordered and, while the structural model is in good agreement with the X-ray crystallographic data (with the molecular connectivity and absence of an N_{2} ligand being conclusive), the interatomic distances are approximate and will not be discussed in detail.^{14} The FeP_{4} unit is pseudo square planar and exhibits a tetrahedral distortion, as seen by a dihedral angle of 12.41(11)° between the two Fe(PP) [Fe(1)P(1)P(4) and Fe(1)P(1i)P(4i)] coordination planes, which is very similar to that observed in the approximately square-based pyramidal [1·N_{2}]^{+}[TfO]^{−} (15.39(9)°). Evidently, coordination of N_{2} results in minimal reorganisation of the [Fe(depe)_{2}]^{+} fragment. While two CH_{3} groups from ligand ethyl groups from each depe moiety are directed towards the vacant axial coordination sites of the Fe centre, large Fe⋯H separations suggest an absence of any agostic or anagostic interactions, which is supported by DFT results (vide infra).^{15}
Neither solid samples nor solutions of [1]^{+}[BAr^{F}_{4}]^{−} under Ar showed bands attributable to an ν_{NN} stretch in their IR or Raman spectra, further confirming the absence of N_{2} in the complex. [1]^{+}[BAr^{F}_{4}]^{−} is insoluble in alkanes, PhH and PhMe, yet highly soluble in THF and in the highly polar, non-coordinating 1,2-difluorobenzene (DFB). ^{31}P NMR spectra (DFB, Ar) of [1]^{+}[BAr^{F}_{4}]^{−} are silent, whereas very broad paramagnetically-shifted resonances for the [1]^{+} moiety feature in the ^{1}H NMR spectrum (see ESI†). The solution-phase magnetic moment (Evans NMR, DFB, 243–298 K, Ar) was found to be 1.75 μ_{B}, and the X-band ESR spectrum (PhMe/DFB glass, Ar, 40 K) revealed a rhombic signal (g_{1} = 2.483, g_{2} = 2.234, g_{3} = 1.985) with an isotropic g-value (g_{iso}) of 2.23, [g_{iso} = (g_{x} + g_{y} + g_{z})/3; Fig. 1(c)]. These data are consistent with a low-spin (S = 1/2, d^{7}) Fe(I) centre. The strong similarity between the X-band ESR spectrum of a powdered sample of [1]^{+}[BAr^{F}_{4}]^{−} (Ar, 1 bar, 40 K) and solution measurements implies that [1]^{+} is virtually isostructural in solution and the solid state.
Dissolution of [1]^{+}[BAr^{F}_{4}]^{−} in N_{2}-saturated solvents (1 atm; DFB or THF) afforded forest-green solutions at room temperature, which became pale yellow upon cooling to −30 °C; either heating these solutions above room temperature, or degassing with Ar, resulted in the rapid reappearance of the characteristic blue colour of [1]^{+}[BAr^{F}_{4}]^{−}. An IR active ν_{NN} stretch at 2067 cm^{−1} confirmed coordination of N_{2} to [1]^{+} to form [1·N_{2}]^{+}, which is intermediate in value between those seen for 1·N_{2} and the related Fe(II) [trans-Fe(H)(N_{2})(depe)_{2}]^{+} (1975 cm^{−1} and 2102 cm^{−1} respectively);^{16} this trend may be readily accounted for by decreased Fe → N_{2} π-backbonding as the oxidation state increases.
The thermodynamics of N_{2} coordination were obtained from variable temperature UV-vis spectroscopy by monitoring the concentrations of [1]^{+} and [1·N_{2}]^{+}via their absorption features (λ_{max} (nm) = 618 and 1018, respectively; Fig. 2).^{17} N_{2} association with [1]^{+} is accordingly found to be exoergic (ΔG_{298} = −4.9(1) kcal mol^{−1}) with, as expected, a favourable enthalpy (ΔH^{o} = −13.1(1) kcal mol^{−1}) and an unfavourable entropy (ΔS^{o} = −27.6(1) cal K^{−1} mol^{−1}) contribution; these values compare well with those for N_{2} binding by (P_{3}B)Co [P = o-(PiPr_{2})_{2}C_{6}H_{4}; ΔH^{o} = −13.9(7) kcal mol^{−1}, and ΔS^{o} = −32(5) cal K^{−1} mol^{−1}], which also produces an S = 1/2 dinitrogen complex.^{18,19}
The X-band ESR spectra of rapidly freeze-quenched N_{2}-saturated DFB or THF solutions (40 K) of [1]^{+}[BAr^{F}_{4}]^{−} revealed a pseudo-axial signal which differs markedly from the rhombic signal characteristic of the N_{2}-free complex (g_{iso} = 2.07; simulated g tensors = 2.0014, 2.0922, 2.125). Additionally, hyperfine coupling to the four ^{31}P nuclei was resolved for the THF glass giving A(^{31}P) = 62.7(1) MHz (Fig. S5†), which is very similar to A_{iso}(^{31}P) previously obtained for [1·N_{2}]^{+}[TfO]^{−} [66.2(2) MHz].^{8}
Similar spectroscopic observations have been described for the Fe(I) complex [Fe(DMeOPrPE)_{2}(N_{2})]^{+} (DMeOPrPE = R_{2}PCH_{2}CH2PR_{2}; R = CH_{2}CH_{2}CH_{2}OMe), which were attributed to an equilibrium between yellow [Fe(DMeOPrPE)_{2}(N_{2})]^{+} and a purple N_{2}-bridged bimetallic S = 1 species {[Fe(DMeOPrPE)_{2}]_{2}(μ-N_{2})}^{2+}; the latter is favoured at higher temperatures and/or low p(N_{2}).^{20} We noted that comparable optical absorptions were observed for [1]^{+} and [1·N_{2}]^{+} (see Table S1, ESI†), and accordingly 2D-ESR (HYSCORE; HYperfine Sub-level CORrElation) experiments were performed on equivalent ^{14}N_{2}- and ^{15}N_{2}-saturated PhMe/DFB (7:1) solutions of [1]^{+}[BAr^{F}_{4}]^{−}, to fully ascertain the solution-phase coordination mode of N_{2} to [1]^{+}. As shown in Fig. 3, two ^{14}N signals were clearly detected at the perpendicular field position, in panels a, b, e, f.
Fig. 3 X-band 2D ESEEM spectra of 2.5 mM [1·^{14}N_{2}]^{+}[BAr^{F}_{4}]^{−} in THF, with standard 4-pulse HYSCORE (panels (a–d)), and DONUT HYSCORE^{21} (panels (e–h)), at two fields (3471 G: (a, b, e and f); 3800 G: (c, d, g and h)) and two τ values (200 ns: (a, c, e and g); 132 ns: (b, d, f and h)) while τ_{2} = 800 ns for DONUT. The samples were frozen glasses at 20 K and measured with pulse lengths of π/2 = 8 ns and π = 12 ns, initial variable delays of 100 ns, and a time step of 20 ns over 200 points in both axes. The microwave frequency was 9.7449 GHz. Experimental data is in black, while separate simulations of two ^{14}N are overlaid in red and cyan. |
The proximal nitrogen, ^{14}N_{p}, gives intense double quantum dq, dq correlation peaks in the (−, +) quadrant corresponding to the ^{14}N directly bound to Fe(I) centre, leading to peaks at (∓16.2, ±12.2) MHz and, from eqn (6), A_{iso}(^{14}N_{p}) may correspondingly be calculated as 14.3 (±0.1) MHz.^{22} For the distal nitrogen, ^{14}N_{d}, field-dependent ^{15}N_{2} simulations of 4-pulse HYSCORE provide A_{iso}(^{14}N_{d}) = 4.4 MHz (seen in ESI, Fig. S7†). A clearly-resolved quadrupole interaction of ^{14}N_{p} reveals a very small asymmetry η ∼ 0, which compares well with sp-hybridized ^{14}N found in NN and [CN]^{−}.^{10a} For the ^{14}N_{d} (red in Fig. 3), it was found that K = 0.9 MHz. In the case of the ^{14}N_{p}, the data was insufficient for a more precise determination, and the quadrupole values of ^{14}N_{d} were used as an approximation. For ligand atoms directly bonded with covalent character, a point-dipole model can be inaccurate; for ^{14}N_{d}T = 1.2 MHz, and for ^{14}N_{p} the simulation value was T = 0.6 MHz. To confirm that both ^{14}N signals derive from the same molecule (considering that no large ^{15}N hyperfine coupling was observed) close examination of the dq, dq peaks at 3471 G reveals weak multinuclear combination frequency peaks such as {N_{p}^{α}(dq), N_{p}^{β}(dq), N_{d}^{β}(dq)} (seen in Fig. S7†), a signal class previously reported by Stich et al. for Mn_{2}(III/IV) Catalase.^{23} It is notable that no ‘half-field’ resonance (g ≈ 4) was observed in the CW-ESR experiments, which would be expected for the hypothetical triplet (S = 1) {[Fe(depe)_{2}]_{2}(μ-N_{2})}^{2+} due to a formally forbidden ΔM_{s} = 2 transition, as has been documented for other transition metal diradicals.^{24} Collectively alongside other spectroscopic data, these observations strongly support the solution-phase assignment of [1·N_{2}]^{+}[BAr^{F}_{4}]^{−} as [Fe(depe)_{2}(η^{1}-NN)]^{+}[BAr^{F}_{4}]^{−}.
Given that {TM(σ-N_{2})} and {TM(σ-H_{2})} fragments are related by the same VE count, numerous diamagnetic metal–ligand platforms have been shown to interconvert these species under N_{2} and H_{2} mixtures;^{25} however, analogous open-shell examples are extremely rare.^{26} Furthermore, only two thoroughly characterised paramagnetic dihydrogen ligand complexes have been reported to date: (P_{3}B)Co(H_{2})^{18} and (P_{3}Si)Fe(H_{2});^{7} both of these are idealised C_{3} symmetric, trigonal bipyramidal S = 1/2 complexes. In spite of the different coordination geometry and cationic charge, admission of H_{2} (1 atm.) to DFB or THF solutions of [1]^{+}[BAr^{F}_{4}]^{−} demonstrated clear reaction (Fig. 4(a)), as evidenced by an immediate colour change to pale green (λ_{max} = 850 nm, ε_{max} ≈ 19 m^{2} mol^{−1}) and the appearance of a new near-axial ESR signal (X-band, PhMe/DFB, 40 K; g = [2.000, 2.085, 2.160], g_{iso} = 2.08, Fig. S6†), which displays more pronounced hyperfine splitting for g_{2} and g_{3} than seen for [1·N_{2}]^{+}[BAr^{F}_{4}]^{−}.^{27} The most plausible identities of this species are the Fe(I) adduct [Fe(depe)_{2}(σ-H_{2})]^{+} ([1·H_{2}]^{+}) or the Fe(III) oxidative addition product [Fe(depe)_{2}(H)_{2}]^{+} ([1(H)_{2}]^{+}).^{28} Using the DFT-optimised structures of [1·H_{2}]^{+} and [1(H)_{2}]^{+} (vide infra), the angle (β) between the g_{1} principal axis and the Fe–H vectors was calculated to be 16° and 31°, respectively.
Orientation-selective ENDOR has previously been used to differentiate between Fe(H_{2}) and Fe(H)_{2} formulations using electron–^{1}H dipolar hyperfine coupling interactions, where only small β angles (<15°; see Fig. 4(b)) were able to reproduce the observed experimental lineshape.^{7,29} The 1:2:1 hyperfine seen at the five negative ^{31}P peaks of g_{min} in CW-ESR of Fig. S6,† was interrogated more closely with H_{2}/D_{2}-saturated solutions of [1]^{+}[BAr^{F}_{4}]^{−} using ^{1}H Davies ENDOR (Electron-Nuclear DOuble Resonance) and both ^{1}H & ^{2}H HYSCORE experiments (see Fig. 5). Fits to data from both techniques used two H hyperfine interaction values, for ^{1}H A(^{1}H) = [−17.99, −19.93, 26.58]/MHz, while the values were scaled by the nuclear g-factor ratio g_{n}(^{2}H)/g_{n}(^{1}H) = 0.1535 for the ^{2}H HYSCORE simulation, revealing a dipolar component of T = 15.2 MHz. The Fe–H distance, r_{FeH} (Å), can thus be obtained from r_{FeH} (Å) = ∛(ρ_{(Fe)}79.06/T), using a spin density of ρ_{(Fe)} ∼ 0.83 as remainder of the large isotropic ^{31}P hyperfine interactions;^{30,31} this coupling is consistent with an Fe–H bond distance of 1.64 Å. Considering the spin density at Fe and the coordinating ligand atoms and DFT coordinates (vide infra), eqn (3) was used to fit an angle of 2β = H–Fe–H, β = 5.5°, using the empirical principal dipolar values [−14.2, −16.2, 30.4] MHz (written as [−T(1 − δ), −T(1 + δ), 2T], with δ as the rhombicity parameter); this value is in excellent agreement with that found for (P_{3}Si)Fe(H_{2}) (β = 6°),^{7} which was interpreted as indicative of partial rotational averaging of the dipolar interaction. However, the angle β = 5.5° includes spin density on opposing ^{1,2}H, and Morris et al., has shown that rotational motion can reduce H–H dipolar interactions by up to a factor of four, implying that β = 5.5° should be considered an upper limit.^{32}
Simulations of ENDOR data for [1·H_{2}]^{+} in Fig. 6 are consistent with the DFT structure (vide infra), in having two classes of ^{31}P hyperfine interaction, a_{iso}(^{31}P) = 69.3 MHz of P in-plane orthogonal to the Fe(I)–H_{2} bond and a_{iso}(^{31}P) = 66.3 MHz for the P bent out-of-plane; furthermore the P–Fe–P depe ligand angle of 85.5° suggests that the symmetry lies closer to 2-fold than 4-fold. Correspondingly, these results provide strong evidence that [1·H_{2}]^{+} is best described as a non-classical σ-H_{2} adduct.
Freshly-prepared solutions of [1·H_{2}]^{+} are sufficiently stable for detailed in situ characterisation and, as with solutions under N_{2}, vacuum/Ar-degassing resulted in regeneration of spectroscopic signals attributed to [1]^{+}, demonstrating reversible coordination of H_{2}. Nevertheless, attempts to obtain thermochemical information for H_{2} binding were unfortunately frustrated by slow and irreversible formation of the Fe(II) trishydride trans-[Fe(depe)_{2}(H)(H_{2})]^{+}, [2]^{+}.^{33} Appreciating that ligation of N_{2} or H_{2} to [1]^{+} results in minimal deformation of the [Fe(depe)_{2}]^{+} core,^{34} we speculated that binding of these gases might also be reversible in the solid state, as in solution. Gratifyingly, admission of either N_{2} or H_{2} to powdered samples of [1]^{+}[BAr^{F}_{4}]^{−} led to comparable colour changes (deep blue to yellow or green, respectively), consistent with clean conversion to [1·(N_{2}/H_{2})]^{+}[BAr^{F}_{4}]_{(s)}^{−}. ESR spectra closely matched those obtained in solution, strongly suggesting that the geometry of the [Fe(depe)_{2}]^{+} core is preserved in both phases upon N_{2} and or H_{2} coordination. Crucially, removal of N_{2}/H_{2} from these samples under vacuum led to complete restoration of the original ESR signal of [1]^{+} (Fig. 7 and S2†), confirming that binding is fully reversible (over multiple cycles) in the solid state; importantly, no loss in signal intensity was observed (which would be expected from formation of diamagnetic [2]^{+}) under H_{2}. Furthermore, ready exchange of the N_{2} ligand for H_{2} is achieved via simple evacuation of N_{2} and replacing with H_{2}, and vice versa.
ESR spectra of [1]^{+}[BAr^{F}_{4}]^{−} obtained in the presence of a single equivalent of N_{2} or H_{2} allow for a quantitative comparison of the binding affinities of the two gases; the ratio [1·L]^{+}/[1]^{+} (L = N_{2} or H_{2}; determined by signal intensity at 303 K) is significantly larger for H_{2} (44) than N_{2} (0.15), revealing that binding of the former is almost 300 times more favourable at ambient temperature.^{35} For comparison, (P_{3}B)Co(L) (L = N_{2}, H_{2}) are in rapid dissociative equilibrium in solution under similar conditions, whereas (P_{3}Si)Fe(L) species require several days to interconvert (proposed to proceed via an associative mechanism involving partial dechelation of the P_{3}Si ligand);^{7,18} both demonstrate a preference for H_{2} binding, albeit to differing degrees (K_{H2}/K_{N2} ≈ 2 and 50 for Co and Fe, respectively). This difference in exchange kinetics was postulated to result from the poorer π-backbonding capability of Co vs. Fe which leads to weaker M–L interactions. Hence it is plausible that the ready reversibility of N_{2} and H_{2} exchange for [1]^{+} relative to (P_{3}Si)Fe could also be due (in part) to poorer π-donation from the former, by virtue of its cationic charge, which is also manifest in the higher ν_{NN} stretch value of the former (2067 vs. 2003 cm^{−1}).
Optimized geometries were ascertained as local minima via frequency calculations. Geometry optimisation of base-free [1]^{+} with S = 1/2 resulted in a D_{2} structure consistent with X-ray crystallographic data, with an angle of approximately 7° between the two iron-ligand Fe(PP) coordination planes (Fig. 8(a)); fixing the spin state to S = 3/2 showed the alternative high-spin structure to be some 0.77 eV (approximately 17.8 kcal mol^{−1}) higher in energy. Geometry optimisation of [1·N_{2}]^{+} results in a structure of C_{2} symmetry with a slightly increased angle between the two Fe(PP)_{2} planes of 19° (Fig. 8(b)). The coordination geometry around Fe is square-based pyramidal with P–Fe–N_{α} angles of 94.8° and 95.6°. The calculated stretching frequency for the bound N_{2} ligand was 2059 cm^{−1}, which is in good agreement with experiment (2067 cm^{−1}). Geometry optimisation of [1·H_{2}]^{+} (Fig. 8(c)) gave a σ-complex with a H–H bond length of 0.899 Å (cf. free H_{2}: 0.74 Å)^{6} and an average Fe–H distance of 1.61 Å, which correlates well with 2D-ESR data. An alternative [1·(H)_{2}]^{+} isomer (Fig. 8(d)), corresponding to the product of H_{2} oxidative addition and containing two well-separated hydride ligands (H⋯H = 1.57 Å and Fe–H distances of 1.51 Å), could also be located, albeit 12 kcal mol^{−1} higher in energy than [1·H_{2}]^{+} (see Table S2 in ESI for further details†).
Fig. 8 DFT-optimised structures of (a) [1]^{+}, (b) [1·N_{2}]^{+}, (c) [1·H_{2}]^{+}, (d) [1(H)_{2}]^{+}. |
Since all compounds had unpaired spins the DFT calculations were unrestricted, with different orbitals for α and β spins. The energies of the α spin electrons, of which there are more, tended to be lower than those of the β spin electrons in corresponding orbitals because of exchange stabilization (Fig. 9).
Fig. 9 Electronic energy levels and their occupancy for [1]^{+}, [1·N_{2}]^{+} and [1·H_{2}]^{+}. The principal Fe character is indicated. |
The unpaired electron in the three cases, [1]^{+}, [1·N_{2}]^{+}, and [1·H_{2}]^{+}, occupies an orbital of primarily d(z^{2}) character; in the case of [1·N_{2}]^{+} and [1·H_{2}]^{+} this is also hybridized with the Fe 4p(z) orbital (Fig. 10), and is antibonding with respect to the coordinated N_{2} or H_{2}, thus explaining the weak association of these ligands to the [Fe(depe)_{2}]^{+} core. Of particular interest is the virtual orbital (z) of [Fe(depe)_{2}]^{+}, the isosurface of which is shown in Fig. 10; its AO composition is predominantly Fe 4p(z) and P 3p.
Time-dependent DFT (TDDFT) was used to calculate the electronic absorption spectra of [1]^{+}, [1·N_{2}]^{+} and [1·H_{2}]^{+} (see Table S3 and Fig. S15 in the ESI†). The numerical agreement with experiment is only moderate, with the bands being calculated at higher energy than measured experimentally. All three compounds have as their lowest energy bands d–d transitions, with a β spin electron being excited into the hole in the d(z^{2}) orbital. However those bands of [1]^{+} have effectively zero oscillator strength on account of its high symmetry. The d–d transitions are of longer wavelength for [1]^{+} than for [1·N_{2}]^{+} and [1·H_{2}]^{+}, which fits with the smaller HOMO–LUMO gap in the former (see Fig. 9). The subsequent set of bands calculated at 587, 548 and 503 nm for [1]^{+} fit well with the features found at 500 and 618 nm experimentally; they are of significantly higher oscillator strength than the d–d bands of [1·N_{2}]^{+} and [1·H_{2}]^{+} and correspond to excitation from the occupied 3d orbitals into the Fe 4p_{z} orbital. In [1·N_{2}]^{+} and [1·H_{2}]^{+}, the corresponding band is absent as a consequence of ligand binding along the z axis, hence their respective absorption spectra have no analogous feature. Thus, in spite of the lack of numerical agreement for the d–d bands, many features of the calculated spectra give a good account of those observed.
Footnote |
† Electronic supplementary information (ESI) available: Full experimental details and characterisation data. CCDC 1451414. For ESI and crystallographic data in CIF or other electronic format see DOI: 10.1039/c8sc01841c |
This journal is © The Royal Society of Chemistry 2018 |