Photoinduced bond oscillations in ironpentacarbonyl give delayed synchronous bursts of carbonmonoxide release

The study investigates the ultrafast, sub-picosecond photodissociation dynamics of Fe(CO)5, a prototypical transition metal carbonyl. While experiments have established sequential CO loss and a singlet dissociation pathway, limited temporal resolution has left the early-time mechanism unclear. The central research question is how initial MLCT excitation channels energy into nuclear motion and non-adiabatic electronic transitions to drive selective CO release. The authors aim to resolve whether Fe–C bond oscillations and state-to-state population transfer govern preferential axial versus equatorial CO dissociation and whether dissociation occurs in temporally structured events. Using explicit excited state molecular dynamics they probe how MLCT-induced Fe–C oscillations, state crossings, and angular distortions (toward C4v via pseudorotation) control timing and selectivity of CO ejection.

Prior ultrafast studies (UV pump–probe, photoionization, ultrafast electron diffraction, time-resolved photoelectron spectroscopy, ultrafast IR) established sequential CO loss from Fe(CO)5 to Fe(CO)4 and Fe(CO)3, predominantly via a singlet pathway, with initial CO dissociation within ~100 fs and intersystem crossing occurring on ~15 ps timescales in gas phase. Core-level and valence photoelectron studies corroborated the singlet nature and sequential kinetics. In solution, femtosecond RIXS revealed competing pathways including ISC and solvent coordination. Theoretical works have treated excited states of Fe(CO)4 and ground-state dissociation pathways of Fe(CO)5 → Fe(CO)4 + CO with high-level methods (CASSCF/CASPT2/NEVPT2). However, detailed early-time (sub-ps) dynamics after MLCT excitation remain insufficiently described, motivating ab initio excited-state MD approaches. Related ESMD studies on Cr(CO)6 and photosensitizers demonstrate the utility of TDDFT-driven dynamics for transition metal photochemistry.

• Electronic structure evaluation: TDDFT (CAM-B3LYP/def2-TZVP) with Tamm–Dancoff approximation in ORCA 4.2.0, RIJCOSX acceleration. Comparisons performed with multireference NEVPT2/CASPT2/CASSCF to validate spectra and PES shapes; TDDFT underestimates absolute excitation energies but reproduces qualitative features.
• Ground-state reference and sampling: Started from CASPT2(12,12)/TZVP optimized D3h geometry; vibrational analysis at B3LYP/cc-pVDZ provided modes for Wigner sampling. Generated 300 Wigner initial conditions.
• UV–vis simulations: Computed spectra for equilibrium and Wigner ensembles; identified bright MLCT (A2′) excitation near 267 nm consistent with experiment.
• Excited-state molecular dynamics: SHARC 2.1 surface-hopping dynamics within singlet manifold. From 300 Wigner conditions, 116 trajectories were excited into S1 with MLCT character. Dynamics propagated including 10 singlet states (S0–S9), timestep 0.5 fs, up to 600 fs, with local diabatization (wavefunction overlap), standard SHARC hopping probabilities, energy-difference-based decoherence correction, and velocity rescaling for kinetic energy adjustment. Triplet states were excluded based on experimental evidence for a singlet pathway in gas phase.
• Dissociation criterion: For each trajectory, defined dissociation time Tdissoc by the running-averaged Fe–C distance of the departing CO crossing 2.5 Å once, suppressing fast oscillations.
• PES analyses: Rigid 1D/2D scans along axial Fe–C distances R1(ax), R2(ax) for lowest 10 singlet states to classify bound MLCT (S5–S9) vs dissociative MC (S1–S4) surfaces; overlaid trajectory segments to relate dynamics to PES topologies.
• Population analysis and kinetic modeling: Aggregated populations of bound (S5–S9) and dissociative (S1–S4) adiabatic states over 0–300 fs; examined state-to-state hopping matrices and lifetimes; constructed a kinetic model with predominant adjacent-state transitions (Si→Si±1) to rationalize population flow.
• Geometric descriptors: Monitored axial Fe–C pair breathing coordinate R12(ax) (distance between axial C atoms) for synchronous oscillations; defined angular distortion parameter Θ from the difference between the two largest C–Fe–C angles to quantify motion from D3h toward near C4v (pseudorotation).

• Photodissociation outcomes: Of 116 MLCT-initiated trajectories, 110 (94.8%) showed single Fe–C bond dissociation to Fe(CO)4 + CO; 4 remained intact after 600 fs. Among single-dissociation events: 94 axial CO releases (85%) and 16 equatorial releases (15%).
• Synchronous Fe–C breathing and bursty dissociation: The MLCT excitation drives synchronous oscillations of the two axial Fe–C bonds with a period ~90 fs (amplitude ~0.3 Å in R12(ax)). Dissociation times cluster into periodic bursts with peaks at ~50, ~140, and ~225 fs, reflecting periodic non-adiabatic access to dissociative surfaces.
• State character and mechanism: Bound MLCT states (S5–S9) and dissociative MC states (S1–S4) were identified. Dissociation proceeds via non-adiabatic transfer from MLCT to MC states when axial Fe–C elongations bring states into crossing regions.
• Orbital control of selectivity: In MC states, the electron occupies a metal-centered orbital (14a1 in D3h symmetry) with σ*(Fe–C) antibonding character. At D3h, antibonding overlap localizes along axial COs, favoring axial dissociation. Upon angular distortion toward near C4v (Jahn–Teller-like mode/pseudorotation), the antibonding character delocalizes over four CO ligands, making axial and equatorial dissociation more equally probable.
• Timing–geometry correlation: The minimum value of the Θ angular parameter attained before Tdissoc correlates with longer dissociation times: later events show greater distortion toward near C4v; initial Θ distribution does not determine Tdissoc.
• Population dynamics: Aggregated populations show decay of MLCT (S5–S9) and concomitant rise of MC (S1–S4) populations aligned with the burst times. Trajectories dissociating in the second burst display a lagged population transfer relative to R12(ax) crests (ballistic-like behavior).
• Vibrational frequencies: Symmetric Fe–C stretching modes at ~400 cm−1 (period ~80–90 fs) at both D3h and near C4v geometries match the observed oscillation period.

The simulations resolve the early-time photodissociation mechanism of Fe(CO)5, showing that MLCT excitation triggers coherent axial Fe–C breathing motion that periodically brings the system to regions of strong non-adiabatic coupling to dissociative MC states. This produces delayed, synchronous bursts of CO release with a dominant axial channel early on. As low-frequency pseudorotation is activated in the bound MLCT manifold, geometrical distortion toward near C4v redistributes antibonding character, increasing the likelihood of equatorial dissociation at later times. The findings reconcile observed singlet pathways and sequential dissociation with a detailed state and orbital mechanism, link nuclear motion to electronic population transfer, and predict experimentally observable oscillations in relative MLCT/ground-state energies and vibrational signatures. The orbital overlap perspective provides a unifying explanation and connects to familiar ground-state reactivity analogies (S_N2-like motion on dissociative surfaces). The mechanism should generalize across metal carbonyls with varying MLCT–MC energy gaps and couplings, implying tunable dissociation dynamics by ligand/metal choice.

This work uncovers that Fe(CO)5, upon MLCT excitation, undergoes synchronous axial Fe–C oscillations that gate non-adiabatic transfer to dissociative MC states, yielding periodic bursts of predominantly axial CO release (~90 fs spacing). A unified orbital mechanism based on σ*(Fe–C) antibonding character in the receiving MC orbital explains early axial preference and delayed equatorial dissociation as pseudorotation distorts the geometry toward near C4v. The study quantifies state populations, dissociation timing, and geometry correlations, and provides experimentally testable predictions (e.g., oscillatory signatures and axial vs equatorial selectivity). Future work should: (i) experimentally resolve which ligand dissociates and detect oscillations/bursts with higher time resolution; (ii) extend simulations to include broader initial state distributions, triplet manifolds where relevant, and solvent effects; (iii) explore other metal carbonyls to establish general design rules via MLCT–MC coupling engineering.

• Electronic structure approximations: TDDFT (CAM-B3LYP, TDA) underestimates excitation energies and may miss multireference effects; however, comparisons to NEVPT2/CASPT2 show qualitative agreement in PES shapes. Differences in MLCT–MC crossing positions between methods could affect absolute timescales.
• State manifold restriction: Dynamics were confined to singlet states, excluding triplets based on gas-phase experimental guidance; potential ISC contributions at later times or in other environments were not treated.
• Trajectory termination: Some trajectories terminated shortly after dissociation due to SCF/gradient convergence issues and limitations in describing Fe(CO)4 near-degeneracy with TDDFT; post-dissociation dynamics were not followed.
• Initial excitation simplification: Trajectories were initiated primarily in a bright MLCT state; realistic pump bandwidth populating multiple states and geometry-dependent state mixing were not fully sampled, limiting direct comparison to experimental time constants.
• Gas-phase focus and finite sampling: Solvent effects were not included; 116 initiated trajectories provide statistical but finite sampling of initial conditions.