Hybrid Barium Titanate Waveguide Designs For Efficient Nonlinear Frequency Conversion

Barium titanate (BaTiO$_3$) is emerging as a powerful integrated photonic material, combining strong $X^{(2)}$ and electro-optic nonlinearities with rapidly improving thin-film waveguide quality. Recent demonstrations of low-loss BaTiO$_3$ waveguides and high-Q resonators have established BaTiO$_3$-on-insulator as a promising platform for next-generation frequency-conversion and quantum photonic technologies. However, while BaTiO$_3$ electro-optic modulators are now well developed, nonlinear BaTiO$_3$ waveguide engineering remains comparatively immature. Techniques widely used in lithium niobate, such as periodic poling for quasi-phase-matching, are poorly suited to BaTiO$_3$ because epitaxial thin films exhibit high coercive fields, strong strain-clamping effects, multivariant domain structures, and slow, complex switching dynamics. These factors make accurate periodic poling challenging and hinder the development of efficient $X^{(2)}$ frequency converters. Here, we introduce a fabrication-robust alternative based on linear-nonlinear hybrid waveguides, where TiO$_2$ is selectively incorporated into BaTiO$_3$ ridge waveguides to enhance nonlinear mode overlap while relying solely on modal phase-matching. Using coupled-mode-theory simulations, we identify phase-matched geometries and show that the hybrid design achieves a 2.75x increase in normalized second harmonic generation efficiency over monolithic BaTiO$_3$ waveguides. The uniform, lithographically defined cross-section makes the approach highly scalable. These results position hybrid BaTiO$_3$-TiO$_2$ waveguides as a practical route to CMOS-compatible, high-efficiency $X^{(2)}$ devices for integrated quantum photonics.

💡 Research Summary

This paper addresses the challenge of achieving efficient second‑order nonlinear frequency conversion in thin‑film barium titanate (BaTiO₃) photonic platforms, where traditional quasi‑phase‑matching via periodic poling is hindered by high coercive fields, strain‑clamping, multivariant domain structures, and slow switching dynamics. To circumvent these limitations, the authors propose a fabrication‑robust, modal‑phase‑matching approach combined with a linear‑nonlinear hybrid waveguide architecture.

First, they perform a systematic sweep of monolithic BaTiO₃ ridge waveguides (varying width and height) to locate geometries where the fundamental TM₀₀ mode at 1550 nm naturally phase‑matches a higher‑order TE₀₂ mode at the second‑harmonic wavelength (775 nm). The optimal monolithic geometry is identified as 0.56 µm wide and 0.78 µm high, where the TE₀₂ mode exhibits three vertically stacked lobes of alternating sign. Because the nonlinear polarization generated by χ^(2) integrates over both positive and negative lobes, the effective mode‑overlap κ is reduced.

To enhance κ, the authors introduce a thin TiO₂ layer—chosen for its refractive index lying between the ordinary and extraordinary indices of BaTiO₃ and its centrosymmetric nature (χ^(2)=0). By sandwiching the BaTiO₃ core between two TiO₂ claddings, the nonlinear interaction is confined to the central lobe of the TE₀₂ mode, while the outer lobes reside in the linear TiO₂. This selective placement eliminates sign cancellation, increasing κ by a factor of 1.6 despite the BaTiO₃ volume being only 28 % of the monolithic case.

Extensive coupled‑mode‑theory (CMT) simulations of the hybrid stack (total core height 0.80 µm, width 0.46 µm, BaTiO₃ thickness 0.30 µm, TiO₂ layers 0.25 µm each) reveal a normalized second‑harmonic‑generation (SHG) efficiency of 1404.5 %·W⁻¹·cm⁻², a 2.75× improvement over the best monolithic BaTiO₃ waveguide (510.5 %·W⁻¹·cm⁻²). This performance rivals or exceeds that of established platforms such as GaP (6.1 %·W⁻¹·cm⁻²), AlN‑SiN hybrids (12 %·W⁻¹·cm⁻²), and even approaches LiNbO₃ periodic‑poling efficiencies (≈3000 %·W⁻¹·cm⁻²).

From a manufacturing perspective, the proposed process separates material growth from device patterning: crystalline BaTiO₃ films are grown epitaxially on a suitable substrate, TiO₂ layers are deposited by sputtering or ALD, and the two stacks are bonded via low‑temperature wafer bonding. Subsequent lithography and dry‑etching define the waveguide geometry. This “grow‑bond‑pattern” flow relaxes lattice‑mismatch constraints, allows independent thickness tuning (tolerances of 30–40 nm for overall width/height and ~200 nm for BaTiO₃ thickness), and is fully compatible with CMOS‑compatible fabrication lines.

In conclusion, the study demonstrates that modal phase‑matching combined with a TiO₂‑BaTiO₃‑TiO₂ hybrid waveguide can deliver high‑efficiency χ^(2) nonlinear interactions without the need for periodic poling. The approach offers a scalable, reproducible pathway to integrate efficient second‑order nonlinear optics with the already mature BaTiO₃ electro‑optic modulators, paving the way for compact, broadband, and quantum‑compatible photonic circuits on a single material platform.