In simulations, stronger laser fields were accompanied by lower reflectance and higher absorption in modeled titanium dioxide, or TiO2, films. In bulk rutile, absorbed energy increased rapidly with field amplitude and followed a fourth-power pattern at high intensity, a result the authors describe as consistent with dominant two-photon absorption. The findings come from first-principles and multiscale simulations.
A model spanning electrons and films
The researchers combined real-time time-dependent density functional theory, or TDDFT, with a Maxwell-TDDFT multiscale method. TDDFT follows electron motion at microscopic scale, while the coupled calculation uses separate macroscopic electromagnetic and microscopic electronic grids. The simulations examined nonlinear optical response, estimated ablation thresholds, thin-film reflection and absorption, and differences among crystalline and amorphous TiO2 structures under strong laser fields.
The main rutile calculation used a rectangular unit cell measuring 4.5 Å by 4.5 Å by 2.8 Å. It modeled linearly polarized sin-squared pulses lasting from 12 to 50 femtoseconds, with field amplitudes from 0.027 to 2.74 V/Å. The photon energies were 1.55 eV at 800 nanometres and 1.17 eV at 1064 nanometres.
The low-field starting point
For rutile, the calculated band gap was 2.62 eV, below the cited experimental value of 3 eV. The calculated dielectric spectra were qualitatively similar to the cited experimental data, and absorption at the modeled 1.55 eV photon energy was negligible.
That low-field result contrasts with the behavior at stronger fields. In bulk rutile, the reported absorbed energy scaled with the fourth power of the pulse field in the high-intensity regime. The paper presents that scaling as consistent with two-photon absorption dominating the response there.
Threshold estimates depend on the pulse
Using absorbed-energy criteria, the paper estimated melting, cohesive and bond-breaking energy scales at about 0.5, 8.6 and 4.3 eV per atom, respectively. The modeled bond-breaking threshold was reached at 2.10 V/Å for a 12-femtosecond pulse and 1.27 V/Å for a 50-femtosecond pulse. The corresponding fluence estimates were 1.01 J/cm² for 12 femtoseconds and 1.58 J/cm² for 50 femtoseconds. The authors caution that the fluence estimate applies only in the low-intensity linear regime.
The film response shifted as intensity rose
In the reported rutile-film cases, increasing incident intensity was accompanied by decreasing reflectance and increasing absorptance, the share of incoming light absorbed by the film. The paper describes multiphoton excitation and optically excited free carriers as relevant high-field processes. Reflectance reached 90% and 50% of its linear value at different modeled intensities for the two pulse durations.
For 12-femtosecond pulses, the reported 90% and 50% reflectance benchmarks occurred at 9.9 × 10¹² W/cm² and 2.6 × 10¹⁴ W/cm². For 50-femtosecond pulses, they occurred at 4.7 × 10¹² W/cm² and 2.4 × 10¹³ W/cm². These are simulated thresholds for the reported rutile-film cases.
At low intensity, simulated reflectance agreed quantitatively with analytical thin-film Fresnel predictions for the selected configurations. At 800 nanometres, the predicted maximum and minimum occurred near film thicknesses of 70 and 150 nanometres. At 1064 nanometres, they occurred near 100 and 200 nanometres.
Crystal structure mattered most at weak fields
The comparison included rutile, anatase, brookite, TiO2-II, TiO2-B and an amorphous TiO2 supercell model. Absorption varied among these structures at low intensity but converged to similar characteristics at high intensity. The structural differences were therefore most visible in the weak-field regime.
The authors propose bulk crystalline TiO2 as a reasonable computational surrogate for industrially produced amorphous thin films when the analysis is restricted to high intensity. At low intensity, absorption still differed by structure.
What the calculations leave open
The results are model-based and concern selected rutile configurations, wavelengths, pulse durations and thicknesses. The fluence estimate is restricted to the low-intensity linear regime, while the high-intensity response is treated through the multiscale analysis.
A preprint study
The document is a preprint, arXiv:2608.25129v1 [physics.optics], dated 25 August 2026. The work was partially supported by MEXT through JSPS KAKENHI Grants 24K01224 and 24K06922. It also received support from the ILE Collaborative Research Program, No. 2026B2-047, and used computational facilities including ISSP, the University of Tsukuba MCRP/Miyabi system and RIKEN Fugaku. The authors state that supporting data are available from the corresponding author upon reasonable request.
Paper data and sources
Original title: First-Principles Prediction of Nonlinear Optical Response in $\text{TiO}_2$ for High-Power Dielectric Mirror Applications
Authors: Koya Shimaoka, Yusuke Kondo, Kazunori Shibata, Mitsuharu Uemoto
Journal/Repository: arXiv
Status: Preprint, not yet peer-reviewed
First online: 2026-08-25
DOI: Not available
Original paper · Full text