Researchers traced the path from ionization of isolated D2O molecules to D-D bond formation using correlated photoelectron and ion imaging alongside trajectory simulations. The dominant D2+ channel was assigned to an indirect sequence rather than direct dissociation: removing an electron from the orbital labeled HOMO-1 populated a bound cationic A state, and two additional 400-nm photons then reached a dissociative B state. The trajectory analysis resolved three branches: direct, roaming and delayed dissociation.
A two-stage route
The reported energy picture was consistent with that assignment. The energy after HOMO-1 removal was 14.8 eV. Each additional 400-nm photon contributed 3.1 eV, bringing the total to about 21 eV, compared with a 20.5-eV D2+/O dissociation limit. The dominant route was distinguished from weak secondary side channels whose assignments remained tentative.
The experiment combined two kinds of measurement. Photoelectron-photoion measurements used 400-nm, 40-fs laser pulses at a 1-kHz repetition rate. Time-resolved ion-ion measurements used orthogonally polarized 800-nm, 6-fs pulses. Together, these measurements provided timing and fragment-channel information for following D2+ formation.
To connect the signals with molecular motion, the researchers used OpenMolcas CASSCF trajectory surface hopping, a calculation that follows nuclear motion while allowing electronic-state changes. It included nonadiabatic couplings calculated as the trajectories ran, Tully's fewest-switches hopping algorithm, time steps of about 0.5 fs, and Boltzmann-distributed starting velocities at 300 K. The main ensemble contained 500 trajectories; 69 met the D2+ selection criteria, or 13.8%. That fraction describes the computational selection, not an experimental participant or event count.
Timing separated the pathways
At zero delay, when the probe pulses overlap in time, both D2+ and HD+ yields were suppressed. The authors interpreted the pattern as evidence for dissociation from a non-Franck-Condon initial wave packet, meaning the nuclear motion did not begin in the simple equilibrium-like arrangement implied by a direct picture.
Error-function fits gave two rise centers for the D2+ signal. Experimentally they were 33.8 fs and 72 fs; in simulation they were 42.2 fs and 71.5 fs. The first rise for H2+ was earlier, at 25.4 fs. The first-rise isotope ratio was about 1.33, compared with about 1.41 for the square root of the mass ratio. Experiment and simulation shared the ordering of the two timescales, although their rise centers and widths were not identical.
Trajectory analysis gave the roaming branch a distinctive back-and-forth path. After a first hop at the conical intersection, the oxygen-to-D2 distance expanded to about 3.0 angstroms, then returned to roughly 1.5 angstroms while one deuterium orbited the other. A transient D-D bond appeared while an O-D bond remained. This nonmonotonic route was treated as a separate branch, although the assignment rests on trajectory and orbital analysis.
A bond forms before the breakup
The branching picture was tied to symmetry breaking in the model. At perfectly symmetric C2v geometry, A/B nonadiabatic coupling vanished. The only b2 nuclear coordinate available for that coupling was the asymmetric stretch, while bending changed the gap between the states. Near the conical intersection, asymmetric motion accompanied state mixing, and the analysis described the symmetry breaking as stochastic rather than fixed.
Orbital-density analysis filled in the electronic side of the picture. Near the conical intersection, asymmetric distortion mixed the 1b2 and 3a1 hole configurations. The singly occupied orbital changed from O-D bonding character to a shape that bonded the two deuteriums before the O-D2 structure had fully separated. In the described trajectories, electronic D-D bond formation preceded complete fragmentation.
What remains uncertain
One compact feature remains unsettled. Supplementary analysis examined a signal around plus or minus 8 fs and favored a pileup caused by point-charge inversion: converting kinetic energy into an apparent distance becomes unreliable at compact geometries. That interpretation did not require an enhanced dissociation probability, but double ionization or mixing involving a dicationic configuration could still contribute. Additional time-resolved photoelectron measurements were identified as a way to separate these possibilities.
The result is a molecular-scale picture for isolated D2O under strong-field laser excitation, not a demonstration about bulk water or an applied hydrogen-production system. The 69 selected trajectories should not be read as a measured branching ratio; they are a simulation-specific fraction from the 500-trajectory ensemble. The study's conclusions therefore describe how this isolated-molecule system behaves under the reported conditions.
Paper data and sources
Original title: Tracking molecular hydrogen formation from ionized water in real time
Authors: Chuan Cheng, Chi-Hong Yuen, Eleanor Weckwerth et al.
Journal/Repository: arXiv
Status: Preprint, not yet peer-reviewed
First online: 2026-08-24
DOI: Not available
Original paper · Full text