A Caged Neutral 17-Valence-Electron Iron(I) Radical [Fe(CO)2(Cl)(P((CH2)10)3P)]•: Synthetic, Structural, Spectroscopic, Redox, and Computational Studies

UV irradiation of yellow CH2Cl2 solutions of trans-Fe(CO)3(P((CH2)10)3P) (2a) and PMe3 (10 equiv) gives, in addition to the previously reported dibridgehead diphosphine P((CH2)10)3P (46%), a green paramagnetic complex that crystallography shows to be the trigonal-bipyramidal iron(I) radical trans-[Fe(CO)2(Cl)(P((CH2)10)3P)]• (1a•; 31% after workup). This is a rare example of an isolable species of the formula [Fe(CO)4–n(L)n(X)]• (n = 0–3, L = two-electron-donor ligand; X = one-electron-donor ligand). Analogous precursors with longer P(CH2)nP segments (n = 12, 14, 16, 18) give only the demetalated diphosphines, and a rationale is proposed. The magnetic susceptibility of 1a•, assayed by Evans’ method and SQUID measurements, indicates a spin (S) of 1/2. Cyclic voltammetry shows that 1a• undergoes a partially reversible one-electron oxidation, but no facile reduction. The UV–visible, EPR, and 57Fe Mössbauer spectra are analyzed in detail. Complex 2a is similarly studied, and, despite the extra valence electron, exhibits a comparable oxidation potential (ΔE1/2 ≤ 0.04 V). The crystal structure shows a cage conformation, solvation level, disorder motif, and unit cell parameters essentially identical to those of 1a•. DFT calculations provide much insight regarding the structural, redox, and spectroscopic properties.


■ INTRODUCTION
For a variety of reasons, the synthesis, isolation, and study of organometallic radicals has lagged behind that of diamagnetic species. 1 This dichotomy also extends to computational investigations. 2,3Among the many classes of interest, considerable attention has been given to 17-valence-electron pentacoordinate iron(I) carbonyl radicals (Figure 1).Teams spearheaded by Connelley, Baird, and Krossing have reported the synthesis and extensive spectroscopic and structural characterization of the cationic complexes trans-[Fe-(CO) 3 (PPh 3 ) 2 ] •+ PF 6  − and [Fe(CO) 5 ] •+ Al(OC(CF 3 ) 3 ) 4 − . 4,5−8 Additional isolable neutral pentacoordinate iron carbonyl radicals are illustrated in Figure 1. 9 Given the innate reactivity of most types of radicals, chemists often seek to sterically shield them.In this context, the P(OiPr) 3substituted radicals in Figure 1 are much more stable than the P(OMe) 3 homologues, 6 in accord with the greater phosphite ligand cone angle.Similar relationships have been established for numerous pairs of organometallic radicals. 6,10Some recent dramatic examples involve bulky meta-terphenyl isocyanide ligands that contain multiple isopropyl or trifluoromethyl substituents, which enable the crystallization of the pentacoordinate Group 7 complexes [M(CO)(CNAr) 4 ] • (M = Mn, Tc, Re). 10 However, comparably enveloping steric environments have not yet been applied to pentacoordinate Group 8 radicals.
In this paper, we report the (1) serendipitous isolation, (2) structural, spectroscopic, and electrochemical characterization, and (3) computational analysis of the neutral pentacoordinate iron dicarbonyl chloride radical trans-[Fe(CO) 2 (Cl)(P-((CH 2 ) 10 ) 3 P)] • (1a • ).The stability is attributed to a new motif of steric protection afforded by a cage-like, triply transspanning diphosphine ligand.Furthermore, the mechanistic sequences involved are believed to involve topologically unusual steps tantamount to turning a molecule inside out.The many unique physical properties of 1a • are thoroughly interpreted, often with the aid of the DFT calculations.
■ RESULTS Synthesis of the Title Complex.The iron tricarbonyl complexes 2 in Scheme 1 have been under study in the group of one author for some time. 11They feature trans-spanning diphosphines with three P(CH 2 ) n P tethers, the lengths of which can be varied.It was recently found that irradiation with a Hanovia mercury lamp (450 W) in the presence of excess PMe 3 in hexanes or CH 2 Cl 2 afforded the corresponding free dibridgehead diphosphines 3. 12,13 These could be isolated as white solids in 46−77% yields.All are capable of homeomorphic isomerization, a topological process that turns the molecules inside out, equilibrating in,in and out,out isomers without the need for pyramidal inversion of the phosphorus bridgeheads.This is illustrated in Scheme 2, and a video is available in the Supporting Information of a recent publication. 14s further elaborated on in the Discussion section, the Fe(CO) 3 moiety is believed to escape by an initial iron− phosphorus bond cleavage, followed by homeomorphic isomerization of the diphosphine to give an adduct of out,out-3 and Fe(CO) 3 .Subsequent attack of PMe 3 would displace the η 1diphosphine from iron.Although a small amount of trans-Fe(CO) 3 (PMe 3 ) 2 can be detected during photolyses, the conversion never exceeds 5%, presumably because this complex is itself photoactive under these conditions.
When 2b−e were photolyzed in hexane or CH 2 Cl 2 , the yellow solutions became cloudy orange. 15However, CH 2 Cl 2 solutions of 2a turned green. 16Workup afforded a paramagnetic green material in 31% yield based upon the structure established below (3a was also produced).The IR spectrum exhibited two strong ν CO bands (1944 and 1861 cm −1 ) at higher frequencies than those of the precursor (1853 and 1841 cm −1 ), and microanalysis supported the formulation trans-[Fe(CO) 2 (Cl)(P-((CH 2 ) 10 ) 3 P)] • (1a • ).A mass spectrum showed a strong ion corresponding to [M − 2(CO)] + .As illustrated in Figure s12, solid 1a • decomposed over the course of 20 h in air.Solutions decomposed over the course of 1.5 h.
Crystallography.Single crystals of a hexane hemisolvate of 1a • could be grown, and the X-ray structure was solved as described in Table 1 and the Experimental Section.The (CH 2 ) 10 segments exhibited two conformations, which could be modeled by a 59:41 occupancy ratio.Thermal ellipsoid plots of the dominant isomer are provided in Figure 2, and bond lengths and angles about iron are summarized in Table 2 Complex 2a, which differs from 1a • only by a CO/Cl replacement (identical tether lengths), was viewed as an especially valuable reference compound.Accordingly, the crystal structure of a hexane solvate of 2a was solved.Again, the (CH 2 ) 10 segments exhibited two conformations, which in this case was best modeled by a 54:46 occupancy ratio.The thermal ellipsoid plots are presented side-by-side with those of 1a • in Figure 2. As can be seen in Table 1, the space group, crystal system, and Z value of 2a and 1a • are identical (P2 1 /c, monoclinic, 4), and the unit cell dimensions vary by less than 0.4%, showing the lattices to be essentially isostructural. 17s would be intuitively expected from the reduced number of valence electrons and presumably attenuated back-bonding in 1a • , the Fe−C bonds [1.793(5) and 1.802(5) Å] are longer than those of 2a [1.764(2), 1.763(2), and 1.768(2) Å], 2c [1.761(3) and 2 × 1.764(2) Å], and other closely related 18-valenceelectron Fe(CO) 3 adducts.11a The CO bond lengths [1.074(5) and 1.095(5) Å] are in turn shorter [2a, 1.157(2), 1.160(2), and 1.162(2) Å; 2c, 1.162(3) and 2 × 1.164(3) Å].It merits note in passing that complexes of the formula Fe(CO) 3 (PPh 2 R) 2 can additionally adopt square pyramidal geometries, 18 and that isomerization can take place upon oxidation. 19dditional Physical Characterization.UV−visible spectra of 1a • and 2a were recorded in CH 2 Cl 2 .As depicted in Figure 3, 1a • exhibits a moderately intense band at 382 nm (ε 1230 M −1 cm −1 ) superimposed on an absorption tail, and a broader and weaker band at 692 nm (ε 490 M −1 cm −1 ).The latter is of course primarily responsible for the green color.The precursor 2a displays a shoulder of modest intensity at 359 nm (ε 510 M −1 cm −1 ), and did not absorb above 450 nm.The underlying Next, the magnetic susceptibility (χ) of 1a • was determined by Evans' method.The value, μ eff 1.67 μ B , is typical of complexes with a spin (S) of 1 / 2 .Two VT SQUID measurements were made with powdered samples.Both gave identical results, as depicted in Figure 4 (top), indicating a magnetic moment (μ B ) ranging from 1.76 (300 K) to 1.63 (2 K).
As shown in Figure 4 (bottom), CW X-band EPR spectra of 1a • were recorded under several conditions.An isotropic spectrum in liquid CH 2 Cl 2 at 293 K (black trace) exhibited a triplet due to 31 P coupling with a g iso value of 2.05 and a superhyperfine coupling constant (A iso ) of 2.48 mT (23.7 × 10 −4 cm −1 ).A spectrum of the corresponding frozen glass (95 K) also gave a signal with a g iso value of 2.05, but without coupling (red trace).
Next, zero-field 57 Fe Mossbauer spectra of solid samples of 1a • and 2a were recorded at 77 K.As shown in Figure 5 (top), 1a • exhibits a positive isomer shift (IS or δ) of 0.14 mm s −1 , a conspicuously small quadrupole splitting (QS) of 0.29 mm s −1 , and a line width (Γ) of 0.34 mm s −1 .In contrast, 2a (bottom) displays a negative isomer shift of −0.12 mm s −1 , a larger quadrupole splitting of 2.38 mm s −1 , and a line width of 0.27 mm s −1 .These properties are summarized in Tables 3 and 4, together with the experimental data from the literature 20−22 and DFT computational results 23 described and interpreted below.
Cyclic voltammograms of 1a • and 2a are depicted in Figure 6.These were recorded in CH 2 Cl 2 under the standard conditions summarized in the caption.Each gave a partially reversible oxidation with i c /i a values of 0.42 and 0.75, respectively.The E 1/2 values (vs Fc 0/+ ) show that the oxidation of 1a • , which would give the 16-valence-electron species 1a + , is only slightly less thermodynamically favorable (with respect to any arbitrary oxidizing agent) than that of 2a to give 2a •+ (0.02 vs −0.01 V; ΔE 1/2 = 0.03 V).A replicate determination on another apparatus but at 200 instead of 100 mV s −1 gave i c /i a values of 0.70 and 0.88 and a ΔE 1/2 of 0.04 V (Figure s10).Surprisingly, neither complex underwent reduction when scans were extended to −2.0 V. Computational insight is provided below.
As illustrated in Figure 7, gas phase energy minimization afforded structures of 1a′ • and 2a′ that were very close to those obtained crystallographically. Bond lengths and angles associated with the central iron atoms and carbonyl ligands are incorporated into Table 2, and the excellent agreement is apparent.The conformations along the P−Fe−P axes (Figure 2   right) were also nicely reproduced.To help understand the Fe−Cl bond length trends, the spin densities of 7′ • and 1a′ • were calculated.As can be derived from the former (Figure 9A), the electron lost from 7′ − comes from the iron 3d xy orbital, which is antibonding with respect to the chlorine 3p x orbital (x-axis perpendicular to the FeP 2 Cl plane).The spin density motif calculated for the full molecule 1a′ • (Figure 9B) is, as expected, very close to that of 7′ • .Parts C and D of Figure 9 show the corresponding plot for the HOMO.As expected, the spin density is dominated by the unpaired electron in this orbital, but the spin density shows spin polarization with the opposite spin perpendicular to the majority spin.
A reviewer noted that the experimental Fe−P bonds in 1a • were longer than those in 2a, although one might have expected a shortening due to the oxidized iron having a shorter bond radius in 1a • , and asked if this lengthening was due to decreased π-bonding.While parallel trends have been observed with other complexes of iron and phosphorus donor ligands where the corresponding radical cations have been crystallographically characterized 25 a more thorough examination of this issue was undertaken as this same trend is reproduced by the calculations for both the actual complexes, 1a′ • and 2a′, and the simplified model complexes 7′ • and 8′.The Quantum Theory of Atoms in Molecules (QTAIM) 26 provides a quantum mechanically accurate method to calculate atomic charges.The QTAIM results for 8′ and 7′ • (Table s1) show that, in spite of 7′ • having a higher formal oxidation state than 8′, the atomic charges differ     by less than 0.10 electrons.This small change occurs because the increased donation from Cl, a strong σ and π donor, offsets the expected electron loss from the increase in formal oxidation state.Thus, in spite of the increase in formal oxidation state, the bond radius of the iron should be similar.One can gain some insight into the evolution of the π-bonding by examining the bond lengths.Replacing a CO in 8′ with Cl − to produce 7′ − causes a substantial increase in the back-bonding of the other ligands and both the Fe−C and Fe−P bond lengths decrease substantially and the C−O and P−C bond lengths increase substantially, a characteristic signature of increased backbonding.When 7′ − is oxidized to produce 7′ • the bond length changes described above are reversed, an indication that the back-bonding in 7′ − is reduced in 7′ • .Relative to 8′, 7′ • appears to have somewhat less backbonding.Table 5 compiles the upper valence molecular orbitals (MOs) for the three model complexes, and Figure 10 illustrates how the energies of the MOs evolve from the 18-electron tricarbonyl 8′ to 7′ − and then 7′ • .It is helpful to include 7′ − in this sequence as 7′ • is an unrestricted calculation with different MOs for α and β spins, and 7′ − shows how the chloride ligand splits the iron 3d orbitals before they are split further by the unrestricted calculation in 7′ • .
As depicted in Figure 10, 8′ displays the typical symmetry and orbital distribution for a 3d 8 trigonal-bipyramidal complex. 27he two most stable iron 3d MOs, HOMO−2 and HOMO−3, are only π bonding with the carbonyl ligands.The two higher energy iron MOs, HOMO and HOMO−1, are π bonding with some of the carbonyls and σ antibonding with others.When one carbonyl ligand is replaced with a chloride ligand, the degeneracy of both of these pairs is broken.Comparison of HOMOs of 8′ and 7′ − demonstrates that the strong destabilization of two MOs in 7′ − is due to the π antibonding character with the chloride ligand.When a β electron is removed from the HOMO of 7′ − , the MO splits and the energy of the β MO increases, becoming the LUMO of 7′ • β.All of the MOs in 7′ • β are higher in energy due to the loss of an exchange integral, but they follow the same order as 7′ • α.
Attention was next turned to the cyclic voltammetry data in Figures 5 and s10.First, consider the oxidation of 2a by 1a + as expressed in eq 1: The sum of the E 1/2 values for the half-reactions (+0.03 to +0.04 V) predict that it should be exergonic by −0.69 to −0.92 kcal mol −1 , as calculated from the Nernst equation.Thus, the energies of each of the four species (i.e., 2a′/1a′ + /2a′ •+ /1a′ • )   were calculated as described in the Experimental Section.As summarized in Table 6, when the geometries were optimized with solvent corrections, the reaction was computed to be exergonic in CH 2 Cl 2 by −0.68 kcal mol −1 in accord with the experimental data.However, the process was predicted to be slightly endergonic in the gas phase and to have mild solvent dependency.In spite of 1a′ + having only 16 valence electrons, it has a singlet ground state because the strong π donation from the Cl destabilizes the HOMO of 1a′ • such that removal of its remaining electron offsets the alternative of removing the electron from d x 2 −y 2 orbital that is strongly stabilized by π-backbonding of the two CO ligands.Next, attention was focused on modeling the UV−visible spectra in Figure 3 with TD-DFT.The spectra computed for 1a′ • and 2a′ are depicted in Figure 11.No intensity is observed for 2a′ in the visible region, consistent with the white to pale yellow color of 2a and its experimental UV−visible spectrum (Figure 3).However, 1a′ • exhibits four moderately strong absorptions in the blue and orange/red regions, consistent with the green color of 1a • and its UV−visible spectrum.As expected, the PMe 3 -substituted model complexes 7′ • and 8′ show very similar spectra, as illustrated in Figure s11.
■ DISCUSSION Mechanism of Formation of 1a • .Most of the pentacoordinate 17-valence-electron iron(I) carbonyl radicals in Figure 1 were accessed by simple oxidative or reductive pathways.In contrast, the title complex 1a • is prepared by the photolysis of an iron carbonyl phosphine complex (2a) in the absence of conventional oxidizing or reducing agents (Scheme 1).Such photolyses can labilize either the carbonyl or phosphine ligands, 28 and the dichloromethane solvent is the only possible source of the chlorine atom in 1a • .Over the last 40 years, the coordination chemistry of dichloromethane has been extensively developed, evolving from conjecture 29 to spectroscopically characterized intermediates 30 to isolable adducts. 31Thus, the postulation of an initial adduct 4a (Schemes 1 and 3) seems plausible.
This would be followed by inner sphere chlorine atom transfer by one of several mechanistic variants, 32 generating the ironcentered radical 1a • and the carbon-centered radical • CH 2 Cl.Indeed, many dichloromethane complexes undergo carbon− chlorine bond cleavage, but with the next observable being an addition product of the type L y M(Cl)(CH 2 Cl).30b However, this might be suppressed by the cage-like diphosphine ligand.An experimental and computational study involving the transient 18-valence-electron complex (η 5 -C 5 H 5 )Re(CO) 2 (ClCH 2 Cl) implicated subsequent thermal conversion to the 17-valenceelectron species [(η 5 -C 5 H 5 )Re(CO) 2 (Cl)] • . 33one of the other dibridgehead diphosphine complexes 2b-e (Scheme 1) give an analogous iron(I) chloride complex upon photolysis.As shown in Scheme 3, the mechanism of demetalation of 2a-e to the free dibridgehead diphosphines 3a-e is believed to involve initial iron−phosphorus bond cleavage and homeomorphic isomerization to yield the η 1diphosphine adducts 5a-e.Since 2a has the smallest diphosphine cage, its conversion to 5a would be expected to be slower.Indeed, the barriers to isomerization from out,out-3a,b,c to in,in-3a,b,c decrease from 27.6 to 14.9 to 13.7 kcal mol −1 (ΔG ⧧ 353 K ). 12 Thus, it would be no surprise if CO/CH 2 Cl 2 photosubstitution of 2a to give the dichloromethane complex 4a and then 1a • were to become competitive (Scheme 3).
A less direct pathway to 1a • would involve a CO/CH 2 Cl 2 substitution of 5a, followed by intermediates of the types 6a and 7a (Scheme 3).If this sequence were operative, it would seem likely that the 5b-e generated from 2b-e would undergo analogous substitutions, and ultimately form some 1b-e • as byproducts.However, there has been no evidence for the generation of significant quantities of these species.
Redox Properties of 1a • .As illustrated by the cyclic voltammograms in Figures 6 and s10, 1a • can undergo a partially reversible oxidation to the 16-valence-electron cation 1a + in CH 2 Cl 2 .There is seemingly the possibility for dichloromethane coordination to 1a + as well.The octahedral coordination geometry that would result has abundant precedent.11a,34,35 Interestingly, no evidence was seen for reduction of 1a • , including cathodic scans out to −2.0 V, close to the solventimposed limit.Reduction would presumably yield 1a − , an 18valence-electron species analogous to computationally characterized 7′ − .
The E 1/2 value for the couple 1a • /1a + (0.020 V) is also a puzzle.This is only slightly greater than that for the tricarbonyl complexes 2a/2a + (−0.010V), each of which have one additional valence electron.Nonetheless, the oxidation of 1a • is only slightly less thermodynamically favorable than 2a, consistent with the DFT data in Table 6.As shown in Figure 10, when the strong π acceptor, CO, is replaced by the strong π donor, Cl − , 36 two of the iron d orbitals are markedly destabilized.The HOMO is so strongly destabilized that the 17-electron neutral species becomes more stable than the 18electron anion.Loss of an electron from the HOMO of the 18electron anion stabilizes the remaining electron in this MO; now its orbital energy is nearly identical to that of the HOMO of the 18-electron tricarbonyl system.Accordingly, the E 1/2 value for the 1a • /1a + couple is only slightly different from that of the 2a/ 2a + couple.However, the E 1/2 values for couples involving higher homologues of the tricarbonyl complexes, 2b/2b + and 2c/2c + , indicate much more thermodynamically favorable oxidations (−0.136 and −0.146 V). 11a,37 Thus, there is a significant cage size effect.
The Mossbauer data for 1a • and 2a in Figure 5 can be compared to those of the 18-valence-electron complexes trans-Fe(CO) 3 (L) 2 (L = CO, PPh 3 , PCy 3 ), as summarized in Tables 3  and 4.However, among the iron-based radicals in Figure 1, only [Fe(CO) 5 ] •+ Al(OC(CF 3 ) 3 ) 4 − has been characterized by Mossbauer spectroscopy.In all cases, the DFT calculations give quadrupole splittings (QS, Table 3) and isomer shifts (IS, Table 4) that are in excellent agreement.Also, the computational data for the PMe 3 model compounds are very close to those of their experimental counterparts.
The most interesting Mossbauer comparisons involve the data for 1a • and 2a versus those of [Fe(CO) 5 ] •+ Al(OC(CF 3 ) 3 ) 4 − Table 5. MOs of Model trans PMe 3 Complexes 8′, 7′ − , and 7′ • with Energies in Atomic Units Inorganic Chemistry and Fe(CO) 5 .In both pairs of complexes, the paramagnetic species exhibits a dramatically lower quadrupole splitting (0.29 vs 2.38 mm s −1 and 0.53 vs 2.55 mm s −1 ), with that of 1a • being conspicuously small.The larger splitting of the diamagnetic iron(0) complexes arises from the d 8 configurations, which, as compared to a spherical d 10 atom, have less electron density along the P−Fe−P axis.When another electron is lost from the d orbital in the (CO) 2 FeCl plane to give iron(I), the electron density becomes more spherical, reducing the quadrupole splitting.
Also, the isomer shifts of the paramagnetic iron(I) complexes are greater than those of their diamagnetic iron(0) counterparts (0.14 vs −0.12 mm s −1 and 0.17 vs −0.08 mm s −1 , respectively).This is opposite to some literature generalizations regarding oxidation state trends [e.g., Fe(0) > Fe(I)].22b,23 However, fuller treatments highlight the independent roles of ligand fields and covalency, coordination numbers, and bond lengths.22b,23a Furthermore, Peters has reported an iron carbonyl triphosphine system that can be isolated in three oxidation states, and his isomer shift trend parallels ours [Fe(II) > Fe(I) > Fe(0)].9c The doublet character and (in the case of 1a • ) the strong donation from the chloride ligand would be expected to enhance isomer shifts.So to sum, pairs of closely related pentacoordinate paramagnetic iron(I) and diamagnetic iron(0) complexes appear to give diagnostically different isomer shifts and quadrupole splittings.
Overview and Conclusion.The steric stabilization afforded by the dibridgehead diphosphine cage of the radical 1a • can also be visualized with space-filling representations, as shown in Figure 12.The equatorial CO and Cl ligands are visible only in a "peekaboo" mode, and the tightly fitting methylene chains strongly shield the metal.Despite these favorable factors, at least two potential Achille's heels remain.One is direct air oxidation (Figure s12), presumably by some outer sphere    process.Another would be homeomorphic isomerization, a pathway we consider operative for isosteric 2a as outlined in Scheme 3.This would expose the iron atom, facilitating a variety of possible degradation reactions.For this reason, related complexes with less conformationally flexible phosphorus− phosphorus linkages have been a long-standing synthetic goal of one author. 39n conclusion, this work has established a new strategy for the stabilization of organometallic radicals based upon the steric shielding provided by cage-like trans-spanning dibridgehead diphosphine ligands.Importantly, these ligands can also accommodate square planar 35,40 and octahedral 11a,34,35 coordination geometries, so this approach could have considerable generality.However, such will be facilitated by the development of other synthetic routes, as that used for 1a • (Schemes 1 and 3) is only feasible for one cage size.Nonetheless, the different reaction modes of 2a-e as a function of cage size provide valuable insight regarding homeomorphic isomerization in coordination chemistry.It is also easy to envision substituted dibridgehead diphosphines that provide even more steric shielding for the L y M core, and the analogous dibridgehead diarsines and distibines are also available. 13SAFETY STATEMENT Caution!The 450 W photochemical lamp emits considerable heat during use, requiring an external cooling well. 12,13Interruption of the water f low in the quartz cooling well can lead to glass failure, solvent ignition, and other potential hazards.Attached hosing should be checked thoroughly before operation of the lamp.Ultraviolet light produced by the lamp is damaging to biological tissues.When in use, proper PPE should be worn (e.g., UV protective goggles/glasses, lab coat, gloves) to ensure that exposure is minimized.Care should also be taken when using liquid nitrogen for Schlenk line traps to avoid condensation of liquid oxygen f rom air.
■ EXPERIMENTAL SECTION General Procedures.Reactions and workups were conducted under inert atmospheres.Chemicals were treated as follows: hexanes, CH 2 Cl 2 , and toluene, dried and degassed using a Glass Contour solvent purification system; PMe 3 (Strem, 98%) used as received; silica gel (40−63 μm mesh, Silicycle) flame-dried and left under vacuum for 1 day before use.
trans-Fe(CO) 2 (Cl)(P((CH 2 ) 10 ) 3 P) (1a • ).A flame-dried Schlenk flask was charged with trans-Fe(CO) 3 (P((CH 2 ) 10 ) 3 P) (2a; 0.278 g, 0.446 mmol), PMe 3 (0.69 mL, 6.69 mmol), and CH 2 Cl 2 (10 mL), and placed in front of a water-cooled quartz immersion well of a Hanovia 450 W lamp.As illustrated with photographs elsewhere, 12,13 the sample was irradiated overnight with stirring.The solvent was removed by an oil pump vacuum.The residue was dissolved in hexanes and applied to a small pipet column of silica gel.The column was rinsed with hexanes (eluting 3a) and then CH 2 Cl 2 .The dark green fractions were collected, and the solvent was removed by an oil pump vacuum.The residue was washed several times with hexanes and dried under vacuum to give 1a  41 Geometry optimizations and frequency calculations in the gas phase used the B3LYP 42 functional and the 6-311G(d) 43 basis set for all atoms.Tight convergence criteria were used for the optimizations.Wave function stability calculations were performed to confirm the of lower-energy numerical solutions for all computed structures.Ultrafine grids (99,590 points per atom) were used as implemented in the Gaussian software.Calculations involving redox phenomena were performed with TPSS 44 functional with the 6-311G(d) basis set for all atoms.The TPSS functional has performed well for predicting structures and electrochemical properties of other iron and nickel complexes. 45Initially, dichloromethane and acetonitrile solvent corrections with SMD 46 were applied to the gas-phaseoptimized geometries.The solvent dependency suggested that optimizing the structure in CH 2 Cl 2 would bring its predicted value closer to the experimental one.Mossbauer calculations were done in the gas phase with B3LYP/def2-TZVP 47 method/basis set combination in the Orca 5.0.3 software package, 48

as recommended. 23a
Crystallography. A. A small quantity of 1a • was suspended in hexanes, and CH 2 Cl 2 was added dropwise until the mixture was homogeneous.The sample was allowed to slowly concentrate under argon at −38 °C.After 5 d, green prisms were collected and data obtained per Table 1.Cell parameters were obtained from 90 data frames taken at widths of 1°and refined with 70838 reflections.Integrated intensity information for each reflection was obtained by reduction of the data frames with the program APEX3. 49Lorentz, polarization, and absorption corrections were applied, the last using the program SADABS. 50The space group was determined from systematic reflection conditions and statistical tests.The structure, a hexane hemisolvate, was solved using XT/XS in APEX3 49,51 and refined (weighted least-squares refinement on F 2 ) to convergence. 51,52longated ellipsoids and nearby residual electron density peaks on the methylene chains (C1−C10, C11−C20, and C21−C30) suggested disorder, which was successfully modeled.Further nearby residual electron density peaks, and the larger thermal ellipsoids of C21−C30, indicated additional disorder, but no efforts were made to model this.The occupancy ratios of disordered atoms were first refined individually.Since they were close, they were grouped together, giving a final 59:41 ratio.Appropriate restraints and constraints were added to keep the bond distances, angles, and thermal ellipsoids meaningful.All non-hydrogen atoms were refined anisotropically.Hydrogen atom positions were calculated and refined using a riding model.

B.
A hexane/toluene (2:1 v/v) solution of 2a was kept at −20 °C.After 3 d, a yellow block-shaped crystal was collected, which was cut as it appeared to be a multi twin or cracked.Data were obtained as per Table 1.Crystal screening, unit cell determination, and data collection were carried out using a XtaLAB Synergy, Dualflex, HyPix diffractometer.The diffraction pattern was indexed and the total number of runs and images was based on the strategy calculation from CrysAlisPro, 53 which was used throughout.The unit cell was refined using 45684 reflections.Integrated intensity information for each reflection was obtained by reduction of data frames within the same software suite, and Gaussian and numerical absorption corrections were similarly applied.A hexane molecule was found, with the C−C midpoint coincident with an inversion center.The occupancy was refined to 0.90, corresponding to 0.45 molecules of solvated hexane per iron atom.A residual electron density peak near C2s suggested disorder of the hexane, which was modeled between two positions with an occupancy ratio of 81:19.Also, elongated or abnormal thermal ellipsoids and/or residual electron density peaks was noted near all the carbon atoms in three hydrocarbon chains except C28.This disorder was modeled between two positions with an occupancy ratio of 54:46.Appropriate restraints and constraints were added to keep the bond distances and thermal ellipsoids meaningful.Systematic reflection conditions and statistical tests afforded the space group (Table 1), as confirmed by ShelXT 2018/ 2 54 using dual methods and Olex2−1.5. 52The structure was refined by full matrix least-squares minimization on F 2 using version 2019/1 of XL. 51 All non-hydrogen atoms were refined anisotropically.Hydrogen atom positions were calculated and refined using a riding model.

Figure 5 .
Figure 5. Zero field 57 Fe Mossbauer spectrum of solid 1a • (top) and 2a (bottom) at 77 K and referenced to α-iron.Collected data are represented by black circles, and additional data are in Tables3 and 4.

Figure 10 .
Figure 10.MO diagrams for model trans PMe 3 complexes 8′, 7′ − , and 7′ • .The orbitals are depicted in Table 5.The orbitals energies of 7′ • α, 7′ − , 8′ are aligned at HOMO−3 level, as this is the Fe 3d xz MO that does not interact directly with any Cl orbitals.In spite of the reduction in symmetry, this figure maintains the z axis as the P−Fe−P axis and the x axis as perpendicular to the FeP 2 Cl plane.

Table 1 .
, Summary of Crystallographic Data

Table 3 .
Computed and Experimental 57 Fe Mossbauer Quadrupole Splittings (QS) of Selected Complexes
■ AUTHOR INFORMATION