Simulation of silicon ridge waveguide enhanced two-photon absorption from femtosecond pulses
Cael Warner, Ruoheng Zhang · 2025
Read Simulation of silicon ridge waveguide enhanced two-photon absorption from femtosecond pulses by Cael Warner, Ruoheng Zhang (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2506-18098.txt.
Introduction
Cael Warner and Ruoheng Zhang simulate how a silicon ridge waveguide boosts two-photon absorption from an ultra short light pulse. The pulse lasts 50 femtoseconds and falls downward onto the ridge. Its color centers near 800 nanometers, with a small spread. Their 2025 study runs in two dimensions with a standard FDTD solver.
The ridge has the standard size used in such work: 500 nanometers wide and 130 nanometers tall. It sits on a thin silicon layer above a thick oxide base. The team varies the pulse energy and tracks how much the guide absorbs. Tight trapping of the light mode inside the ridge drives the boost, they report.
Catalog record
| CREATOR | Cael Warner, Ruoheng Zhang |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 19 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2506-18098
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Simulation of silicon ridge waveguide enhanced two-photon absorption from femtosecond pulses Cael Warner, Ruoheng Zhang Abstract Field enhancement of two-photon absorption from a 50- fs \mathrm{f}\mathrm{s} pulse on a silicon ridge waveguide is simulated for a varying energy downward propagating 800 ± 9.42 800\pm 9.42 nm \mathrm{n}\mathrm{m} wavelength plane-wave orthonormal to a waveguide in 2D FDTD using ANSYS Lumerical FDTD. Energy absorbed by the waveguide is enhanced due to mode confinement within the standard 500 nm \mathrm{n}\mathrm{m} width, 130 nm \mathrm{n}\mathrm{m} height silicon ridge waveguide on 90 nm \mathrm{n}\mathrm{m} thick silicon and 3 µ \mathrm{\SIUnitSymbolMicro} silicon dioxide insulator. K eywords Two-photon absorption \cdot field enhancement \cdot silicon photonics. 1 Introduction Silicon photonic devices are finding rapid application as essential components in integrated electronics and data centers [ 1 ] . However, process variation in fabrication results in variation of passive device performance in terms of effective refractive index, n_{eff} . Production of multiple waveguides or use of micro-heaters, to modify the material for desired performance, is a wasteful process leading to significant energy usage and multi-billion dollar losses for data-center operation. Alternatively, free carriers can be excited via two-photon absorption (TPA) in silicon thereby causing defects in the crystal lattice which modify n_{eff} to achieve desired device performance. Changes in n_{eff} can be measured directly using planar thin-film silicon microring resonators on insulating silicate substrates. Quasi-neutral non-thermal melting retains geometric structure of the device but modifies the crystal lattice microstructure, and the device’s optical properties [ 2 ] . Non-thermal melting is attained near the band-gap energy of silicon, E_{\mathrm{g}}=1.12 \mathrm{e}\mathrm{V} , using \lambda_{0}=800 \mathrm{n}\mathrm{m} wavelength laser ablation to excite TPA. At an orthonormal angle of incidence with respect to the waveguide axis, \lambda_{0}=800 \mathrm{n}\mathrm{m} light couples strongly to the standard ridge waveguide dimensions (500 \mathrm{n}\mathrm{m} width and 130 \mathrm{n}\mathrm{m} height), such that the intensity and TPA of a 50- \mathrm{f}\mathrm{s} 300 \mathrm{n}\mathrm{J} pulse is enhanced. Therefore, with a low repetition rate, only the ridge wave-guide’s microstructure defects and n_{eff} are modified with negligible thermal melting or material strain [ 2 ] . This is in contrast to ordinary thermal ablation, where above threshold melting fluence causes craters in silicon [ 3 ] . Few studies simulate the initial two-dimensional field enhancement from two-photon absorption [ 4 ] , which occurs instantaneously compared with the free carrier absorption (over 6 \mathrm{p}\mathrm{s} ) [ 5 ] . Control of the interference mechanism may allow tailored in-situ micro-structural modification of silicon chips in future re-purposing or resonator tuning, enabling desired device performance without contributing additional electronic waste by fabricating multiple passive devices. 2 Two-photon absorption Model General nonlinear polarizability \mathbf{P}^{(n)}(z,\omega)=D\epsilon_{0}\mathbf{\chi}^{(n)}(\omega;\omega_{1},\omega_{2},...,\omega_{n})\mathbf{E}(z,\omega_{1})\mathbf{E}(z,\omega_{2})...\mathbf{E}(z,\omega_{n}), (1) Exhibits a degeneracy factor of D=n!/m!=3 for Raman absorption, compared with D=n!/m!=6 for TPA, where n represents the order of the polarizability and m represents the number of waves. If we assume frequency-domain two-wave degeneracy for TPA by an isotropic material, then the two-photon imaginary third-order susceptibility, \chi"^{(3)}_{\mathrm{TPA}} , is numerically equivalent to the Raman imaginary third-order susceptibility, \chi"^{(3)}_{\mathrm{R}} , apart from a constant factor of 2 from D . ANSYS Lumerical ® FDTD uses a built-in Chi3 Raman Kerr material plugin model based on earlier work by Blow & Wood [ 6 ] ,and Goorjian & Taflove [ 7 ] . In simplified terms, the ANSYS Lumerical model represents [ 8 ] \mathbf{P}(t)=\epsilon_{0}\chi^{(1)}\mathbf{E}(t)+\epsilon_{0}\eta\chi_{0}^{(3)}\mathbf{E}^{3}(t)+\epsilon_{0}(1-\eta)\mathbf{E}(t)(\chi^{(3)}_{R}(t)\ast\mathbf{E}^{2}(t)), (2) Where the susceptibility has a Lorentzian spectrum [ 7 ] \chi_{R}^{(3)}(\omega)=\frac{\chi_{0}^{(3)}\omega_{R}^{2}}{\omega_{R}^{2}-2j\omega\delta_{R}-\omega^{2}}, (3) Such that the time-domain response of \chi_{R}^{(3)}(t) is solved from a Fourier transform. We note that this form is from the ANSYS Lumerical documentation, and has some numerical coefficients which differ from Goorjian & Taflove’s [ 7 ] . With modification by a factor 2, \chi"^{(3)}_{\mathrm{R}}=2\chi"^{(3)}_{\mathrm{TPA}} , the Chi3 Raman Kerr material plugin can represent measured TPA coefficient \beta [ 9 ] , \chi_{0}^{(3)}(\omega)=\frac{\epsilon_{0}cn_{0}^{2}\beta[(\omega_{R}^{2}-\omega^{2})^{2}+4\omega^{2}\delta_{R}^{2}]}{24\omega^{2}\omega_{R}^{2}\delta_{R}}, (4) Where \omega_{R} is the Raman frequency, and \delta_{R} is the Raman spectrum full width at half maximum (FWHM), also experimentally measured [ 10 ] . Provided measurements of both Kerr nonlinear refractive index, n_{2} , and \beta , specific tuning of a dimensionless coefficient, \eta , represents proportion of real Raman and Kerr susceptibility added at a single wavelength such that \displaystyle\chi^{\prime(3)} \displaystyle=(1-\eta)\chi_{0}^{(3)}+\eta\chi^{\prime(3)}_{R} (5a) \displaystyle\eta \displaystyle=\frac{4\left\{\delta_{R}\left[(n_{2}\omega_{R}^{2}/c\beta-4\delta_{R})\omega^{2}\right]-(\omega_{R}^{2}-\omega^{2})^{2}\right\}}{\omega^{2}\left\{|\omega_{R}^{2}-\omega^{2}|+4\delta_{R}^{2}\right\}}, (5b) Where \eta\in(0,1) . Alternatively, given an impulse response for the Raman contribution, g_{R}(t)=\left(\frac{1}{\tau_{1}^{2}}+\frac{1}{\tau_{2}^{2}}\right)\tau_{1}\exp{\left(-\frac{t}{\tau_{2}}\right)}\sin{\left(\frac{t}{\tau_{1}}\right)}, (6) Where \tau_{1}=1/\omega_{R} and \tau_{2}=1/\delta_{R} , one may also solve for \eta as \eta=\frac{cn(\omega_{0})H_{R}(\Delta\omega)}{\omega_{0}\chi^{(3)}\mathrm{Im}\left(G_{R}(\Delta\omega)\right)}, (7) Where H_{R}(\Delta\omega) is the Raman-shifted spectrum [ 11 ] . For more details of the derivation, please refer to Supplementary material. Intensity dependent two-photon absorption occurs instantaneously compared with free carrier absorption at room temperature (6 \mathrm{p}\mathrm{s} ) [ 12 ] . Therefore, we initially neglect free carrier absorption for the first single 50- \mathrm{f}\mathrm{s} pulse (assuming it is a factor 100 lower frequency), and examine field enhancement from two-photon absorption alone. The analytical one-dimensional intensity is I=I_{0}\frac{\exp(-\alpha z)}{(\alpha+\beta I_{0})-\beta I_{0}\exp(-\alpha z)}, (8) Where \beta is the two-photon absorption coefficient and \alpha is the linear absorption coefficient. Absorbed energy density is described by u_{\mathrm{ABS}}=\int\beta I^{2}\mathrm{d}t . Each analytical field profile in one dimension is compared with the two-photon absorption profile in bulk silicon to verify the simulation accuracy of the material plugin model. 3 Two-photon absorption material model ANSYS Lumerical FDTD is used to model two-photon absorption in isotropic silicon with the Chi3 Raman Kerr material plugin. The Kerr refractive index and two-photon absorption coefficient [ 9 ] are fitted to the Raman spectrum for a linewidth typical of the femtosecond laser. This is achieved by equating the real and imaginary third-order susceptibility of the model to real and imaginary terms measured by Bristow, accounting for a degeneracy factor. Given this derivation (see the Appendix), the corresponding parameter of interest is \chi_{0}^{(3)}(\omega) from ( 4 ), where \beta , n_{2} , \omega_{R} , and \delta_{R} are used from previous measurements [ 9 , 10 ] . Table 1 includes a list of these parameters. The imparted intensity on the silicon is varied, I_{0}\in(113.2,11320) \mathrm{G}\mathrm{W}\mathrm{/}\mathrm{c}\mathrm{m}^{2} , to represent different pulse energies, U\in(80.56,4319.74) \mathrm{n}\mathrm{J} , although the temporal length of the pulse ( \tau=50 \mathrm{f}\mathrm{s} ) is the same for each case over a w(0)=13.4 \mathrm{\SIUnitSymbolMicro} spot size. The base material of the Chi3 Raman Kerr material plugin is Si (Palik) [ 13 ] , which is fitted using a simplex algorithm over the wavelength range \lambda_{0}\in(750,850) \mathrm{n}\mathrm{m} . Table 1: Model properties for Silicon Parameter name Symbol \lambda_{0}=800 \mathrm{n}\mathrm{m} Unit Linear absorption coefficient \alpha 1000 \mathrm{c}\mathrm{m}^{-1} Two photon absorption coefficient \beta 2.000 \mathrm{c}\mathrm{m}\mathrm{G}\mathrm{W}^{-1} Raman frequency \omega_{R} 15.56 \mathrm{T}\mathrm{H}\mathrm{z} Raman spectrum FWHM \delta\omega_{R} 2.91 \mathrm{T}\mathrm{H}\mathrm{z} Susceptibility model parameter \chi^{(3)}_{0} 1.4361(10^{-12}) \mathrm{c}\mathrm{m}^{2}\mathrm{V}^{-2} Kerr contribution \eta 1.1607(10^{-3}) - Figure 1: (a) SOI cross-section, (b) RW SOI cross-section, (c) SOI system, and (d) RW SOI system. In (a) and (b) the units are in \mathrm{\SIUnitSymbolMicro} . Each cross section in (a) and (b) is illustrated in (c) and (d), respectively. In (c) and (d), the red spot represents the laser, \mathrm{P_{z}}(t) is the Poynting vector, and \mathrm{E}(t) is the electric field. Number of spatial units in x , y and z Cartesian directions are N_{x} , N_{y} , and N_{z} . At the surface, the beam diameter is \omega(0) and the beam is weakly divergent and approximated as a