Electronic Structure and Ligand Vibrations in FeNO, CoNO and FeOO Porphyrin Adducts

Abstract

The gaseous ligands, CO, NO and O₂ interact with the Fe ion in heme proteins largely via backbonding of Fe electrons to the π* orbitals of the XO (X = C, N, O) ligands. In these FeXO adducts, the Fe–X stretching frequency varies inversely with the X–O stretching frequency, since increased backbonding strengthens the Fe–X bond while weakening the X–O bond. Inverse frequency correlations have been observed for all three ligands, despite differing electronic and geometric structures, and despite variable composition of the `FeX' vibrational mode, in which Fe–X stretching and Fe–X–O coordinates are mixed for bent FeXO adducts. We report experimental data for 5-coordinate Co^II(NO) porphyrin adducts (isoelectronic with Fe^II(OO) adducts), and the results of DFT modeling for 5-coordinate Fe^II(NO), Co^II(NO) and Fe^II(OO) adducts. Inverse ν(MX)/ν(XO) correlations are obtained computationally, using model porphyrins with graded electron-donating and -withdrawing substituents to modulate the backbonding. Computed slopes agree satisfactorily with experiment, provided non-hybrid functionals are used, which avoid overemphasizing high-spin states. The BP86 functional gives correct ground states, a closed-shell singlet for Co^II(NO) and an open-shell singlet for the isoelectronic Fe^II(OO), as corroborated by structural data for Co^II(NO), and the ν(MX)/ν(XO) slope agreement with experiment for both adducts. However, for Fe^II(OO) adducts, the computed inverse ν(MX)/ν(XO) correlation applies only to porphyrins with electron-donating and withdrawing substituents of moderate strength. For substituents more donating than –CH₃, a direct correlation is obtained, the Fe–O and O–O bonds weakening in concert. This effect is ascribed to the dominance of σ bonding via the in-plane d_xz(+d_z2)-π* orbital, when electron-donating substituents raise the d orbital energies sufficiently to render backbonding (d_yz-π*) unimportant.

Introduction

The gaseous molecules CO, NO and O₂, all constituents of the atmosphere, are also important signaling molecules in biology. The signaling receptor is generally a heme protein, to which the XO molecule binds, and initiates a conformation change that induces further biochemical events.^1–3 The heme prosthetic group is poised to bind these diatomic ligands, because they all have low-lying π* orbitals, which can interact with d_π orbitals on the heme Fe. The resulting d_π-π* backbonding is a key determinant of the binding affinity.

The binding event is also controlled by secondary interactions with the protein, involving a proximal ligand to the heme Fe, generally a histidine or cysteine sidechain, and a variety of steric and electronic interactions with sidechains that are distal to the heme-bound XO molecule. These interactions are also responsible for transducing the XO binding into protein conformation change.

Vibrational spectroscopy provides a useful probe of heme-XO binding, since the Fe–X and X–O stretching frequencies reflect the bond strengths, as modulated by the protein.^4–6 In general, these frequencies correlate inversely, because backbonding strengthens the Fe–X bond while weakening the X–O bond.^7,8 Electrostatic effects that increase or decrease backbonding can be gauged from these correlations. Other interactions that alter the geometric or electronic structure can be evaluated by the resulting deviations from the inverse correlations.^7–10

Systematics for these effects have been worked out most thoroughly for FeCO porphyrin adducts,^8,9,11 which have the simplest FeXO electronic structure. CO has empty π* orbitals and the Fe^II(CO) adduct is linear, with backbonding interactions in both perpendicular directions. Computational modeling has been important for understanding the experimental data. DFT calculations on FeCO adducts of model porphyrins have reproduced the observed backbonding trends satisfactorily, and have illuminated the sources of deviations in secondary interactions.^9,12,13

Studies have also been extended to NO, which has an extra electron, resulting in bent Fe^II(NO) adducts.^8,13,14 The extra π* electron is involved in σ bonding, but backbonding remains available perpendicular to the bending plane (Figure S1). Also encountered, however, are Fe^III(NO) complexes, which are isoelectronic with Fe^II(CO), but have altered polarity, and quite different secondary interactions.^15–18 The important Fe^II(OO) adducts are less-well studied, primarily due to experimental difficulties,^19–21 but useful data are accumulating.^22–36 O₂ has two π* electrons, and the complex FeOO electronic structure has been much investigated recently.^37–45 The aim of the present work is to understand trends in the FeOO vibrational frequencies, extending the computational approach developed for FeCO and FeNO adducts. We have included Co^II(NO) adducts in the study, and provide new experimental data on these adducts. Co^II(NO) is isoelectronic with Fe^II(OO), but has properties closer to Fe^II(NO); it represents an intermediate case.

The present work is limited to 5-coordinate adducts. The effects of axial ligation, and of other secondary interactions will be considered in a subsequent study.

Materials and Methods

Resonance Raman Measurements

To obtain vibrational frequencies for 5-coordinate Co^II(NO) adducts, four cobalt(II) porphyrin complexes were obtained from Porphyrin Systems, (Germany): meso-Tetraphenylporphyrin (Co(II)TPP); 5,10,15,20-Tetrakis-(4-methoxyphenyl)-porphyrin (Co(II)TMeOPP); 5,10,15,20-Tetrakis(2,6-difluorophenyl)porphyrin (Co(II)TF₂PP); 5,10,15,20-Tetrakis-(2,3,4,5,6-pentafluorophenyl)-porphyrin (Co(II)TF₅PP), and 5,10,15,20-Tetrakis-(mesityl)-porphyrin (Co(II)TMP) was obtained from Midcentury Chemicals (Posen, IL). Co(II) porphyrin solutions in CH₂Cl₂ (~5 mM) were prepared by transferring, under Ar gas, deoxygenated CH₂Cl₂ into sealed, deoxygenated vials containing the Co(II) porphyrin samples. Portions of the solution were transferred, under Ar gas, to a sealed, previously deoxygenated NMR tube. ¹⁴NO gas, chemically generated by reduction of sodium nitrite with ascorbate, was added, via a gas tight syringe, to the Co(II) porphyrin solution in the NMR tube to form Co^II(¹⁴NO)P adduct. Samples of Co^II(¹⁵NO)P were prepared in a similar way using ¹⁵NO gas. RR spectra were obtained via backscattering geometry at room temperature in spinning NMR tubes. The excitation wavelengths of 406.7 nm and 413 nm were obtained by frequency doubling, via a lithium triborate crystal, a Ti:sapphire laser (Photonics International TU-UV), which was pumped by the second harmonic of a Q-switched Nd:YLF laser (Photonics Industries International, GM-30-527). These laser wavelengths were confirmed with an Ocean Optics UV-Vis spectrometer (Model USB2000+). The laser power at the sample was kept to a minimum (<1 mW) by using a cylindrical lens, to minimize the photodissociation of the NO adducts. Scattered light was collected and focused onto a triple spectrograph (Spex 1877) equipped with a CCD detector (Roper Scientific, Model 7375-0001) operating at −110 °C. Each spectrum was averaged over 15 min of acquisition. Spectra were calibrated with dimethylsulfoxide-d₆ and analyzed with Grams A/I (Thermo-Galactic).

DFT Calculations

Density functional theory (DFT) calculations on 5-coordinate model complexes were performed using Gaussian 03 or 09 programs.^46,47 The standard 6–31G* basis set was used for all the atoms except Fe and Co, which were described by the Ahlrichs' valence triple-ζ (VTZ) basis set.⁴⁸ Geometry optimization and frequency calculations were performed using tight convergence criteria and an ultrafine integration grid. C_s symmetry was assumed for the model porphyrins with eight substituents, except for –NO₂ substituents. For the latter, C_s symmetry constraint resulted in small imaginary frequencies corresponding to –NO₂ group rotations; a stable C₁ structure was found instead. The models having four porphyrin substituents and protoporphyrin models were optimized without symmetry constraints. Vibrational frequency values for the optimized structures were taken directly from the Gaussian program without scaling. Fe^II(NO) models were optimized with the unrestricted BP86 (UBP86) functional, assuming a doublet ground state. Co^II(NO) and Fe^II(OO) models were first optimized with the BP86 functional assuming a closed shell singlet ground state. The closed shell wavefunction was then tested for the presence of spin unrestricted, lower energy solutions using the STABLE=OPT key in Gaussian. When the RHF→UHF instability was found and an open shell singlet wavefunction was located, the geometry was reoptimized using the newly found open shell singlet wavefunction. In the case of the Co^II(NO) complexes, the closed singlet ground state was the lowest energy configuration with the BP86 functional. In the case of the 5-c Fe^II(OO) models, an open shell singlet state was the lowest electronic configuration. For the unsubstituted Fe^II(OO)P complex in the open shell singlet, Gaussian 09 experienced a problem with SCF convergence; thus, the results for this complex were obtained with the Gaussian 03 version, which apparently did not have the same issue. Other functions, both pure and hybrid, were also tested during geometry optimization of the simple porphine complex of each model. The results are given in Tables S1–S3 of Supporting Information. The choice of BP86 functional to calculate vibrational stretching frequencies correlation is justified in the results section.

Results

Mode assignments

The assignment of metal-ligand vibrational modes in bent FeXO adducts is complicated by the fact that mixing of stretching and bending coordinates is permitted by symmetry. The inherent frequency is similar for Fe–X stretching and Fe–X–O bending, and mixing of these two coordinates is therefore substantial. (XO stretching is much higher in frequency, so that mixing with other coordinates is inconsequential.) The result is that two normal modes in the 400–600 cm⁻¹ region have major contributions from Fe–X stretching and Fe–X–O bending coordinates. The percentage contributions vary from case to case. The resulting assignment ambiguity has been discussed extensively.^13,49–51

The Fe–X stretching and Fe–X–O bending coordinates have the same phase in the lower frequency mode and the opposite phase in the higher frequency mode. Experimentally, the higher frequency mode experiences a larger frequency shift upon isotopic substitution of the X atom. It also has higher intensity in RR spectra. Indeed the lower frequency mode is generally hard to detect^51,52. We have found¹³ that the higher frequency mode correlates with the X–O stretching mode in a manner similar to the correlation of ν(FeC) and ν(CO) in Fe^II(CO) adducts, for which mixing with Fe–C–O bending is precluded by symmetry. For simplicity, we label this RR-detectable, isotope-sensitive mode `νFeX', recognizing that it necessarily contains some contribution, possibly a large one, from Fe–X–O bending.

Co^II(NO) Spectra

Our laboratory has previously used RR spectroscopy to locate ν(CoN) and ν(NO) in several 5-c CoNO porphyrin adducts (Table 1), in order to establish that the NO adduct of Co-substituted sGC (soluble guanylate cyclase) also contains a 5-c Co^II(NO) protoporphyrin.⁵³ In the present study, we examine backbonding effects on the Co–NO and N–O frequencies, by employing a series of tetraphenylporphyrin adducts with a variety of electron-donating or withdrawing substituents on the phenyl rings, a method applied previously to Fe^II(CO) and Fe^II(NO) porphyrin adducts^13,22. Figure 1 shows RR and isotope difference spectra for the 5-c Co^II(NO) adducts of porphyrins with electron donating (tetra-mesityl) and withdrawing (tetra-difluorophenyl) substituents. The RR spectra are dominated by the expected porphyrin modes^5,54, but the ν(CoN) and ν(NO) modes are readily identified in the ^14/15NO difference spectra, as is the q(Co-N-O) mode. Spectra for several other Co^II(NO) porphyrins are displayed in Figures S2–S4, and the frequencies are collected in Table 1.

Table 1.

Co^II–NO and N–O Stretching Frequencies (cm⁻¹) of Five-Coordinate Co^II(NO) Porphyrin Adducts.

Complex	ν(Co–NO) (Δ14NO/15NO)	ν(N–O) (Δ14NO/15NO)	solvent	reference
Co(II)(TPP)(NO)	514 (22)	1692 (30)	CH2Cl2	This work
Co(II)(TMP)(NO)	517 (16)	1680 (30)	CH2Cl2	This work
Co(II)(TPP-p-OMe)(NO)	519 (21)	1682 (30)	CH2Cl2	This work
Co(II)(TF2PP)(NO)	513 (16)	1699 (30)	CH2Cl2	This work
Co(II)(TF5PP)(NO)	501 (17)	1709 (32)	CH2Cl2	This work
Co(II)(OEP)(NO)	521 (17)	1673 (30)	CH2Cl2	53
Co(II)(PPIXDME)(NO)	519	1678	CH2Cl2	53
Co(II)MO sGC1	523 (26)	1675 (37)	25 mM TEA buffer, pH 7.4	53

Figure 1.

Resonance Raman spectra and difference spectra (¹⁴NO–¹⁵NO) (low-frequency region, left; high frequency region, right) for Co(II)TMP(NO) and Co(II)(TPPF₂)(NO), in CH₂Cl₂ solution.

As expected from backdonation, the ν(NO) frequency systematically increases while ν(CoN) decreases when the substituents become increasingly electron-withdrawing. The newly obtained ν(CoN) and ν(NO) frequencies plotted together with available points from the literature give a linear negative correlation, just as other 5-c M^II(XO) porphyrins do (Figure 2).

Figure 2.

RR ν_MX/ν_XO correlation plots obtained for the five-coordinate Fe^II(XO) and Co^II(NO) porphyrin complexes. For the Fe^II(CO) adducts (black points): data from ref.[55]; for the Fe^II(NO) adducts (blue points): data from ref.[22]; for five-coordinate Fe^II(OO) (red points): data from refs.[23–25,56] (see Table S4 for details); data points for five-coordinate Co^II(NO) (green stars) are obtained in this work (Table 1).

When the Co^II(NO) backbonding correlation plot is compared to the correlation plots obtained experimentally for similar TPP substituents of the Fe^II(CO), Fe^II(NO) and Fe^II(O₂) series in Figure 2, it is apparent that for these four adducts, the backbonding correlations are quite similar, albeit with different slopes. The slope of the backbonding correlation lines reflects the sensitivity of the adducts to metal ${\mathrm{d}}_{\pi }\to {\pi }_{\text{XO}}^{\ast }$ backbonding mechanism. The Fe^II(CO), Fe^II(NO) and Co^II(NO) adducts are similarly responsive to the backbonding modulation, while Fe^II(O₂) adducts, having the highest slope, are least responsive to backbonding.

DFT Modeling

Appropriate choice of the DFT functional is essential for proper modeling of heme-ligand interactions. We have successfully modeled the vibrational frequency backbonding trend for Fe^II(CO) porphyrin adducts using the popular hybrid functional, B3LYP^13,22. Although the frequencies themselves are overestimated, as usual with DFT⁵⁷, negative ν(FeC)/ν(CO) correlations were obtained for both 5- and 6-coordinate adducts, with slopes that were close to the experimentally determined ones. However, B3LYP was less successful for Fe^II(NO) adducts¹³, because of its tendency, common to hybrid functionals, to overestimate the stability of high-spin states⁵⁸. Less spin contamination and more accurate bond distances were obtained with the pure functional, BLYP¹³ (Table S1).

The spin level problem is even more significant for Co^II(NO) and Fe^II(OO) adducts, for which open- and closed-shell singlet states are close in energy^58,59. We explored a number of functionals (Tables S2, S3) and found a closed-shell ground state for Co^II(NO) with pure functionals, and bond distances that are in reasonable accord with experiment (Table S2). However, the hybrid functionals gave a lower energy for the open-shell configuration, whose computed Co–NO bond distance is significantly longer than experimentally determined ones (Table S2). Similar CoNO porphyrin results were reported for B3LYP vs. BP86 or BLYP functionals by Jaworska^60,61, and studies of the related cobalamin molecule with alkyl or NO ligands likewise indicated superior performance for the non-hybrid functionals^59,62. B3LYP was found to underestimate the ligand binding energies and overestimate the Co–N distance to the axial ligand.

Many DFT calculations have been reported for FeOO porphyrins^{37–41,44,63,64}. There is general agreement that the ground state is an open-shell singlet^39–44. (No experimental structures are available for 5-coordinate FeOO adducts, so the computed structural parameters (Table S3) do not provide guidance.) Thus, although Co^II(NO) and Fe^II(OO) adducts are isoelectronic, they have different ground states.

The FeOO spin-density distribution, shown in Figure 3, reveals opposed but localized spins on the OO and Fe fragments in the open-shell singlet ground state. Shown for comparison is the spin–density distribution for Fe^II(NO) adducts, which have a doublet ground state, the lone unpaired electron being distributed over the NO and Fe orbitals.

Figure 3.

Spin density distribution in the Fe^II(NO) and Fe^II(O₂) models obtained from the DFT/UBP86 calculations. Spin-up and spin-down densities are shown as red and blue areas, respectively.

In view of its superior predictions of ground states and structures, we limit the subsequent discussion to computations with the BP86 functional (restricted BP86 for Co^II(NO), and unrestricted BP86 for the Fe^II(NO) and Fe^II(O₂) adducts), although results with other functionals can be found in the Supporting Information. The computed structures for Fe^II(NO), Co^II(NO) and Fe^II(OO) adducts of porphine are shown in Figure 4. In each case the optimization left the XO ligand oriented between adjacent pyrrole rings, presumably to minimize non-bonded repulsions.

Figure 4.

Optimized structures of 5-coordinate Fe^II(NO)P, Co^II(NO)P, and Fe^II(OO)P. Indicated are bond distances (Å, black) and Mulliken charges (blue).

Computed Vibrational Trends

As in previous studies,^13,15,22 we investigated changes in electron density at the central metal ion by attaching electron-donating or -withdrawing substituents to the porphine periphery (Figure 5). Structures and vibrational frequencies were computed in order to examine the ν(MX)/ν(XO) correlations.

Figure 5.

Structural diagram for FeP–X₈ and FeP–X₄ porphyrins used to model the backbonding trend. –X are indicated substituents, where –X₄ implies that –X and –H substituents alternate around the ring.

The substitutions were limited to the pyrrole C_β atoms, because direct substitutions at the methine bridges (C_meso) specifically influence the porphyrin a_2u molecular orbital, which in turn interacts with the d_z2 orbital of the central metal ion, altering its relative energy.¹⁵ This interaction can have consequences beyond the general inductive effect of the C_β substituents, which we employ to modulate backbonding. The trends compare well with the experimental data, even though these involved changing the X atom on the phenyl–X ring, which is itself attached to the C_meso atoms. The phenyl ring attenuates electronic communication with –X, so that only the inductive effect is expressed.

The `ν(FeX)' mode was chosen as the calculated mode in the 400–600 cm⁻¹ region having the largest Fe-X displacement.

Fe^II(NO) adducts

We have previously reported computations for Fe^II(NO) adducts using the UBLYP functional¹³. Satisfactory bond distances were obtained, and the computed ν(FeN) and ν(NO) frequencies correlated negatively for 6-coordinate adducts having imidazole axial ligands, with a slope similar to that observed experimentally. However, the results for 5-coordinate adducts were scattered¹³.

We have now re-examined the 5-coordinate correlation using the UBP86 functional and a larger set of porphyrin substituents (Table 2). A negative linear correlation ν(FeN)/ν(NO) was obtained (r = 0.96), with a slope of −0.26, in decent agreement with the experimental slope, −0.40 (Figure 6). The same set of substituted porphyrins calculated with the UBLYP functional gave more scattered frequencies data points resulting in a poorer ν(FeN)/ν(NO) correlation (slope = −0.19, r = 0.91) (Figure S5), though the bond length correlation d(Fe–NO)/d(N–O) was similar with the two functionals (Figure S6).

Table 2.

Optimized (DFT, UBP86) Bond Lengths (Å), Bond Angles (deg) and Vibrational Stretching Frequencies (cm⁻¹) Obtained for the Five-Coordinate Fe^II(NO) Porphine Adducts with Different β-Substituents, including those of protoporphyrin, FePP(PropH)₂.

Complex	Fe–NO	N–O	∠FeNO	Fe–Np	ν(Fe–NO)	ν(N–O)
5-c Fe(II)NO, UBP86 functional
FeP(NH2)8	1.696	1.194	143.6	2.005/2.036	622.4	1715
FeP(CH3)8	1.698	1.192	143.6	2.002/2.032	621.8	1721
FePP(PropH)2	1.699	1.190	143.9	2.006/2.031	620.3	1727
FePH	1.699	1.190	143.9	2.002/2.031	621.3	1731
FePF4	1.700	1.189	144.0	2.000/2.029	618.2	1736
FePF8	1.701	1.188	143.8	1.998/2.027	616.1	1738
FePCl8	1.702	1.186	144.2	2.006/2.035	616.8	1747
FeP(NO2)4	1.706	1.184	144.4	2.006/2.034	614.0	1755
FeP(NO2)8	1.710	1.180	144.6	2.003/2.031	607.2	1771

Figure 6.

ν_FeN/ν_NO correlation computed for the five-coordinate Fe^II(NO)PX₈ and Fe^II (NO)PX₄ complexes (black points) and experimentally obtained (red points). PP(PropH)₂ point corresponds to Fe^II(NO) protoporphyrin with protonated propionates. Linear correlations are shown for RR data (red line, r = 0.97), and for the theoretical models (blue line, r = 0.96).

Co^II(NO) adducts

In agreement with the available experimental structures (Table S2), the Co^II(NO) models are calculated to have a bent CoNO unit, with CoNO angle of ~ 120°. As expected from backbonding, the Co–NO bond length decreases and the N–O bond increases when substituents become increasingly electron-donating (Table 3 and Figure S7). The same methodology applied to CoNO adducts gives a set of ν(CoN)/ν(NO) points for a similar series of porphyrin substituents (Table 3), which again describes a negative linear correlation (Figure 7). The slope, −0.54, is slightly higher than that obtained experimentally for the CoTPP-X adducts, – 0.51. Computed and experimental slopes are both higher for Co^II(NO) than for Fe^II(NO) adducts.

Table 3.

Optimized (DFT, BP86) Bond Lengths (Å), Bond Angles (deg), and Vibrational Stretching Frequencies (cm⁻¹) Obtained for the Five-Coordinate Co^II(NO) Porphine Adducts with Different β-Substituents.

Complex	Co–NO	N–O	∠CoNO	Co–Np	ν(Co–NO)	ν(N–O)
5-c Co(II)NO, BP86 functional
CoP(NH2)8	1.799	1.191	121.2	1.986/2.010	594	1688
CoP(CH3)8	1.805	1.190	121.1	1.980/2.003	590	1695
CoPP(PropH)2	1.806	1.189	121.1	1.982/2.001	588	1700
CoPH	1.808	1.188	121.1	1.979/2.002	588	1704
CoPF4	1.809	1.187	121.1	1.981/2.004	584	1710
CoPF8	1.810	1.186	121.0	1.979/2.002	582	1714
CoPCl8	1.813	1.184	121.1	1.983/2.005	579(a)	1720
CoP(NO2)4	1.822	1.182	121.1	1.972/1.994	576(a)	1729
CoP(NO2)8	1.829	1.179	121.1	1.968/1.989	561(a)	1746

^(a)Averaged values over several modes with significant Co-NO displacement.

Figure 7.

ν_CoN/ν_NO correlation plots computed for the five-coordinate Co^II(NO)PX₈ complexes (black points) and experimentally obtained (red points). PP(PropH)₂ point corresponds to Co^II(NO) protoporphyrin with protonated propionates. Linear correlations are shown for RR data (red line, r = 0.94), and for the theoretical models (blue line, r = 0.98).

Fe^II(OO) adducts

The same methodology applied to Fe^II(OO) adducts (Table 4) gives a more complex result (Figure 8). For substituents more electron-withdrawing than –CH₃, a negative ν(FeO)/ν(OO) correlation is obtained, with a slope, −0.64, close to that given by the available experimental data, −0.79. However, the correlation levels out for the strongly electron. withdrawing –Cl₈ and –NO₂ substituents, for which the computed ν(Fe–O) stays above 500 cm⁻¹. This leveling effect is not seen, however, if computed distances, rather than frequencies are correlated. Figure 9 shows a linear variation of d(Fe–O) with d(O–O) for all electron withdrawing substituents. The leveling effect in the frequency correlation represents a breakdown in correlation of bond distance and vibrational mode frequency, possibly reflecting a shift in the computed mode composition.

Table 4.

Optimized (DFT, UBP86) Bond Lengths (Å), Bond Angles (deg) and Vibrational Stretching Frequencies (cm⁻¹) Obtained for the Five-Coordinate Fe^II(OO) Porphine Adducts with Different β-Substituents.

Complex	Fe–OO	O–O	∠FeOO	Fe–Np	ν(Fe–OO)	ν(O–O)
5-c Fe(II)OO, UBP86 functional (open shell singlet)
FeP(NH2)8	1.773	1.286	119.8	1.994/2.012	504.8	1230.5
FeP(OH)8	1.770	1.285	119.7	1.993/2.008	506.8	1232.7
FeP(NH2)4	1.768	1.284	120.1	1.995/2.012	510.7	1234.1
FeP(CH3)8	1.766	1.283	120.3	1.994/2.010	512.9	1235.9
FeP(CH3)4	1.766	1.282	120.3	1.994/2.010	509.7	1239.3
FePP(PropH)2	1.767	1.282	120.2	1.998/2.010	510.9	1240.2
FePH	1.766	1.281	120.3	1.994/2.010	509.1	1242.1
FePF4	1.768	1.280	120.1	1.992/2.007	505.4	1246.8
FePF8	1.770	1.279	119.9	1.990/2.005	504.0	1249.4
FePCl4	1.769	1.279	120.1	1.996/2.011	503.6	1250.8
FePCl8	1.771	1.277	120.0	1.998/2.013	505.9	1255.6
FeP(NO2)4	1.772	1.274	120.2	2.000/2.015	505.0	1263.8
FeP(NO2)8	1.777	1.270	120.1	1.994/2.011	501.3	1276.6

More striking is the slope reversal, in both ν(FeO)/ν(OO) and d(Fe–O)/d(O–O) plots, for porphyrin substituents more electron-donating than –CH₃. In this region of both plots, a strongly positive correlation is seen.

Figure 8.

ν_FeO/ν_OO correlation plots computed for the five-coordinate Fe^II(OO)PX₈ and Fe^II(OO)PX₄ complexes (black points) and experimentally obtained for selected matrix-isolated Fe^II(O₂) porphyrins (red points). Linear correlations are shown for RR data (red line, r = 0.95), and for the theoretical models (blue line, r = 0.99), but excluding the –Cl₈, –(NO₂)₄, –(NO₂)₈, –(NH₂)₈, –(OH)₈, –(NH₂)₄ data points.

Figure 9.

Fe–OO/O–O bond lengths correlation for the five-coordinate Fe^II(OO)PX₈ and Fe^II(OO)PX₄ models obtained from the spin-unrestricted DFT/UBP86 calculations.

We also computed BP86 frequencies and bond distances for the same set of Fe^II(OO) adducts in the closed shell singlet state, which is slightly higher in energy than the open shell singlet ground state (Table S6). The results, shown in Figure S8, are qualitatively similar but the slope of the negative part of the ν(FeO)/ν(OO) correlation, −0.24, is much lower than that obtained in the ground state, or seen in the experimental data.

Discussion

The similar negative ν(FeX)/ν(XO) correlations seen for all three XO molecules, CO, NO and OO, demonstrate the overriding importance of backbonding in all these adducts, despite differences in electronic and geometric structure. Fe^II(CO) is linear, while Fe^II(NO) and Fe^II(OO) are increasingly bent, with one and two additional electrons in the XO π* orbitals, respectively. Linear Fe^II(CO) has two backbonding interactions, the two CO π* orbitals overlapping with the Fe d_xz and d_yz orbitals, while the bent Fe^II(NO) and Fe^II(OO) adducts have only one backbonding interaction, perpendicular to the FeXO bending plane of FeNO and FeOO. However, the π* orbital energies are lower for NO and OO than for CO, and interact more effectively with the Fe orbitals. The present results show that Co^II(NO) adducts likewise fall into this pattern.

Moreover, the correlations are similar despite the fact that the compositions of the `ν(MX)' modes differ. They are purely M–X stretching for linear FeCO, but mixed M–X stretching and M–X–O bending for the bent MXO adducts, the extent of mixing varying with the bending angle and with the vibrational frequencies. Nevertheless, DFT does a good job of reproducing the observed backbonding trends in model adducts having varying electron-withdrawing and - donating porphyrin substituents.

However, accurate modeling requires the use of non-hybrid functionals that do not overemphasize the stability of high-spin states. Using BP86, we find that the ground state of Co^II(NO) adducts is a closed shell singlet, consistent with the experimental structure. The computed Co–NO distance agrees with the experimental distance, whereas the computed distance for the open-shell singlet state is significantly longer.

The computed ground state for the isoelectronic Fe^II(OO) adducts is the open-shell singlet, consistent with other studies^37–41,63. In this case there is no experimental structure of a 5-coordinate Fe^II(OO) adduct to provide guidance. However, DFT yields a ν(FeO)/ν(OO) backbonding slope that agrees well with experiment (leaving aside models with strongly electron-donating substituents), provided that the open-shell singlet ground state is used. Agreement is significantly poorer for the closed-shell singlet state. This connection to the vibrational frequency pattern provides support for the open-shell singlet indeed being the ground state.

A remarkable computational result is the reversal of the backbonding slope when electron-donating substituents are included in the correlation, when either frequencies (Figure 8) or bond distances (Figure 9) are plotted. For substituents more strongly donating than –CH₃, the Fe–O and O–O bonds become weaker together, contrary to backbonding expectations. This effect is not seen in the experimental data, which do not include adducts with electron-donating porphyrin substituents. (The O₂ adducts are unstable, quickly oxidizing in solution, and have only been examined via matrix isolation.)

How can one account for this reversal? As has often been pointed out, the open-shell singlet ground state implies a Fe(III)O₂⁻ electronic structure, consistent with the Mossbauer parameters, and with the O–O stretching frequency itself, which is essentially that of superoxide (Table S5). The main bonding interaction is σ donation from superoxide to Fe(III). This interaction is augmented by backbonding via overlap of the Fe d_yz and O₂ π_y* orbital (y being the direction perpendicular to the FeOO plane). As electron donation from the substituents increases, backbonding is expected to increase, strengthening the Fe-O bond while weakening the O-O bond. This effect accounts for the inverse correlation. However, electron donating substituents also raise the energy of the porphyrin-Fe orbitals relative to those of the O₂ orbitals so that the d_yz-π_y* overlap becomes smaller. This effect can be seen in the energy-level diagram in Figure S9. At the same time, the increasing electron density on Fe diminishes σ donation from the superoxide, lengthening both the Fe–O and O–O bonds. At some point, the σ effect becomes dominant, and the correlation becomes direct, rather than inverse. The computation indicates that this point corresponds to a model with –(CH₃)₈ substituents. For the three most electron-donating substituents, which define the direct branch of the correlation, the d_yz-π_y* overlap is minimal, appearing only in the α-spin manifold (Figure S9).

Conclusions

Five-coordinate FeXO porphyrin adducts (X = C, N, O) all show negative backbonding correlations with similar slopes between Fe–X and X–O mode frequencies, despite differing electronic and geometric structure, and despite the `ν(FeX)' mode being a mixture of Fe–X stretching and Fe–X–O bending coordinates when the FeXO geometry is bent. Co^II(NO) adducts are likewise found to follow such a correlation.

DFT computations on model porphyrin adducts in which backbonding is modulated by the inductive effect of electron-donating or -withdrawing substituents, capture these correlations with good agreement of calculated and experimental slopes. For adducts with low-lying open-shell states (NO, O₂), non-hybrid functionals are required for good results. Using the non-hybrid BP86 functional, the ground state is found to be closed shell singlet for Co^II(NO) adducts, but open shell singlet for the isoelectronic Fe^II(OO) adducts. Agreement of calculated and experimental frequency correlations support these assignments.

For Fe^II(OO) adducts, the ν(FeO)/ν(OO) correlation is predicted to be direct, rather than inverse, for substituents more electron donating than –CH₃. This reversal in slope is attributed to the dominance of superoxide-Fe(III) σ bonding when electron-donor substituents raise the porphyrin-Fe orbital energies sufficiently to make the backbonding π overlap ineffective.