1T'-MoTe$_2$ as an integrated saturable absorber for photonic machine learning

TL;DR

A coupling efficiency of up to 20% between the 1T′-MoTe2 monolayer and the silicon nitride waveguide is demonstrated, with saturation achievable at input powers as low as a few microwatts.

Abstract

We investigate the saturable absorption behavior of a 1T'-MoTe$_2$ monolayer integrated with a silicon nitride waveguide for applications in photonic neural networks. Using experimental transmission measurements and theoretical modeling, we characterize the nonlinear response of the material. Our model, incorporating quasi-Fermi level separation and carrier dynamics, successfully explains these behaviors and predicts the material's absorption dependence on the carrier density. Furthermore, we demonstrate a coupling efficiency of up to 20% between the 1T'-MoTe$_2$ monolayer and the silicon nitride waveguide, with saturation achievable at input powers as low as a few uW. These results suggest that 1T'-MoTe$_2$ is a promising candidate for implementing nonlinear functions in integrated photonic neural networks.

Figures (6)

• Figure 1: (a) Optical microscopy image of a 1T'-MoTe$_2$ on top of tip of a fiber. (b) Integration of the Raman shift between 247cm^-1 and 272cm^-1, showing that all the fiber core is covered by MoTe$_2$. (c) Raman shift measured at the center of the fiber showing that the flake is a monolayer, based on the characteristic Raman peak position reported in luo20161t. The blue region indicates the integration region of (b).
• Figure 2: Transmission measurement of a monolayer of MoTe$_2$ on the tip of a fiber using a pulsed laser at 1560, plotted in a semilog scale. Inset: Focus on the region between 0.002MW/cm^2 and 0.02MW/cm^2 showing a double saturation process.
• Figure 3: (a) Quasi-Fermi levels calculation using the gap energy ($E_g=E_c-E_v$) $E_g=$-0.2eV huang2016controlling, $m=0.462m_0$keum2015bandgap, the thickness of a monolayer of 1T'-MoTe$_2$ of 1nm. The dashed line represents when the absorption is saturated because the separation of the quasi-Fermi levels is higher than the photon energy $E_p=$ 0.8eV ($\approx$ 1550nm). (b) Carrier density (blue) and laser pulse (red dashed line). For this calculations we used $\tau=$200ps, $\sigma=$ 2ps, $\alpha=$ 1e61/m, $I_0=$30MW/cm^2, $E_p=$ 0.8eV and $\eta=1$, these values are based on the experiment.
• Figure 4: (a) Calculated absorption coefficient as a function of the carrier density. (b) Experimental results in black, calculated transmission considering only the variation of the absorption coefficient in green and calculated transmission considering the variation of the absorption coefficient and the reflectance for different carrier densities. The fitted parameters for the absorption model were $m_{eff}=0.005$ and $\alpha=$1e101/m. Reprinted with permission from volpato2024analysis.
• Figure 5: (a) Cross section of the simulated structure with indications: the silicon nitride waveguide in purple, the SiO$_2$ substrate in gray the dipole on top of the waveguide in green. (b) Simulation of coupling the dipole radiation with the SiNx waveguide for different thicknesses. (c) Calculated transmission as a function of pump power for different device lengths. (a) and (c) reprinted with permission from volpato2024analysis.