Roughness-Limited Performance in Ultra-Low-Loss Lithium Niobate Cavities Journal: opticajournal Publication type: Research Article

Ali Khalatpour *, Luke Qi, Martin M. Fejer, and Amir H. Safavi-Naeini * · 2025

Read Roughness-Limited Performance in Ultra-Low-Loss Lithium Niobate Cavities Journal: opticajournal Publication type: Research Article by Ali Khalatpour *, Luke Qi, Martin M. Fejer, and Amir H. Safavi-Naeini * (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-2505-01913.txt.

Introduction

Ali Khalatpour, Luke Qi, Martin Fejer, and Amir Safavi-Naeini study loss in lithium niobate cavities in a 2025 paper. Thin-film lithium niobate is prized for quantum and nonlinear optics, but its cavities still lose more light than the material limit allows. Surface roughness is a prime suspect. The team separates sidewall roughness from interface roughness in racetrack cavities.

Very high quality factors are reached in both wide and narrow waveguides, they report. In wide guides, interface roughness dominates the loss. In narrow guides, sidewall roughness rules. Their model joins three-dimensional wave simulations with measured roughness and can apply to other photonic materials, they say.

Catalog record

CREATORAli Khalatpour *, Luke Qi, Martin M. Fejer, and Amir H. Safavi-Naeini *
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE49 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2505-01913

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Roughness-Limited Performance in Ultra-Low-Loss Lithium Niobate Cavities Journal: opticajournal Publication type: Research Article Ali Khalatpour *, Luke Qi, Martin M. Fejer, and Amir H. Safavi-Naeini * Abstract Achieving low optical loss is critical for scaling complex photonic systems. Thin-film lithium niobate (TFLN) offers strong electro-optic and nonlinear properties in a compact platform, making it ideal for quantum and nonlinear optics. While Q Q factors above 10 7 10^{7} have been achieved, they remain below the intrinsic material limit. We present a systematic study of scattering losses due to roughness in TFLN racetrack cavities, isolating contributions from sidewall and interface roughness. Quality factors up to 27 × 10 6 27\times 10^{6} are demonstrated in waveguides with widths of 2.2 ​ λ 2.2\lambda ( ∼ 3.5 ​ μ \sim 3.5\,\mu m), where interface roughness dominates, and up to 1.2 × 10 7 1.2\times 10^{7} in narrower waveguides 0.8 ​ λ 0.8\lambda wide ( ∼ 1.2 ​ μ \sim 1.2\,\mu m), where sidewall roughness is the primary limitation. Our modeling framework, based on 3D wave simulations informed by AFM-measured roughness, is material-independent and broadly applicable across integrated photonic platforms. \setcopyright 1 Introduction As integrated photonic systems continue to scale in size and complexity, the demand for material platforms that enable low-loss propagation while supporting both passive and active functionalities becomes increasingly critical. Thin-film lithium niobate (TFLN) presents a compelling solution, offering low optical loss over a broad spectral range—from the visible to the mid-infrared—alongside strong electro-optic and second-order ( \chi^{(2)} ) nonlinearities [ 1 ] . These attributes make TFLN a highly versatile and promising material system for next-generation integrated photonics [ 2 ] . While minimizing optical loss has always been central to system optimization, in emerging applications such as optical squeezing [ 3 , 4 ] , single-photon processing [ 5 ] , and quantum optical computing [ 6 ] , optical losses directly constrain system realization and yield. As a result, understanding the mechanisms of loss in TFLN photonics and achieving ultra-high quality factors has become essential for the feasibility of the next generation of photonic systems. The intrinsic absorption of TFLN is estimated to be approximately 0.2 dB/m, corresponding to an absorption-limited quality factor of \sim 160\times 10^{6} [ 7 , 8 ] . However, fabrication-induced scattering losses often dominate. Lithography processes introduce line-edge roughness which can contribute to optical loss. Physical etching using high-energy argon ions—currently the most effective method for TFLN—also induces sidewall roughness that increases scattering loss. Ion bombardment can also lead to re-deposition of amorphous lithium niobate on the sidewalls, requiring post-processing steps for removal. Standard cleaning agents such as HF (hydrofluoric acid) or SC-1 (a mixture of ammonium hydroxide and hydrogen peroxide) can further roughen the TFLN surface [ 9 ] . In addition, the material’s sensitivity to high-temperature processes poses further fabrication challenges. Fabricating high-quality factor devices across both low and high confinement regimes is essential, as performance in these asymptotic limits helps isolate the contributions of sidewall and interface scattering. This not only enables validation of modeling tools but also guides the identification of optimal geometries for complex photonic structures. However, achieving high- Q in the low-mode-volume limit—where optical fields are concentrated near rough surfaces—remains challenging due to extreme sensitivity to fabrication imperfections, despite the appeal of such devices for strong light-matter interactions and nonlinear effects [ 10 ] . In contrast, high-mode-volume devices are less sensitive to fabrication variations and are primarily limited by the intrinsic surface quality of TFLN. Although waveguides much wider than the optical wavelength have demonstrated intrinsic quality factors as high as 29\times 10^{6} [ 11 ] , their utility is limited by the onset of multimode behavior and reduced nonlinear interaction strength. While previous studies have examined the role of surface roughness in TFLN propagation loss [ 12 , 13 ] , a quantitative predictive tool is still lacking. A model that directly predicts propagation loss due to roughness for any given waveguide geometry and surface roughness would be valuable for optimizing TFLN-based devices. Previous approaches based on the Lacey-Payne model (using the equivalent current method [ 14 , 15 ] ), originally developed for low-index contrast waveguides, can produce order-of-magnitude errors for high-contrast waveguides [ 16 ] and are therefore not suitable for TFLN waveguides. In this work, we employ a full-wave simulation method to extract roughness-induced propagation loss from data obtained by atomic force microscopy for the waveguide geometries of interest. 2 Device Architecture Overview In order to probe the contribution of sidewall roughness to the quality factor, we fabricate racetrack cavities on both X- and Y-cut thin film lithium niobate on oxide wafers (LNOI). The waveguide widths are varied between devices, as wider waveguides are expected to exhibit reduced sidewall scattering loss. To minimize the impact of higher-order modes, we use a narrow and single-mode feedline with a width of 1.2\,\mu\text{m} that is single side-coupled to our cavities. The geometry of the racetracks and feedlines is shown in Figure 1 (a). Figure 1 : (a) Geometry of a racetrack and feedline, along with the cross-section of the racetrack resonator characterized by (w,H,h) , where w is the waveguide width, H is the waveguide height, h is the slab thickness, L is the length of the straight section of the racetrack, and L_{\text{t}} is the taper length ( 200~\mu m). (b) Schematic illustration of a waveguide with a rough boundary at x=-d/2 . In many applications, the thickness of TFLN needs to be adjusted from the thickness provided by the manufacturer (NanoLN in our case). Therefore, to better understand the impact of interface roughness, we prepared two sets of samples, one from X-cut wafers and the other from Y-cut wafers. The X-cut wafers undergo a pre-thinning step using Ar bombardment to remove 200\,\text{nm} of TFLN, whereas the Y-cut wafer surfaces remain at the as-manufactured quality, achieved via chemical mechanical polishing. Since both sets of waveguides have sidewalls oriented in the z -direction of the crystal, the impact of sidewall roughness is consistent across samples, providing an ideal platform to probe the influence of both top and bottom interface roughness. 3 Modeling Methodology 3.1 Roughness Modeling: Sidewalls In this section, we derive an estimate of the power lost by the guided mode due to boundary perturbations along the z -direction in a waveguide with its cross section in the (x,y) plane. The waveguide boundaries are at x=-d/2 and x=d/2 , and we assume that only the boundary at x=-d/2 is perturbed by roughness (see Figure 1 (b)). Suppose that the dielectric tensor takes the form \boldsymbol{\epsilon}(x,y,z)=\overline{\boldsymbol{\epsilon}}(x,y)+\delta\boldsymbol{\epsilon}(x,y,z) , where \overline{\boldsymbol{\epsilon}}(x,y) describes the ideal z -independent waveguide and \delta\boldsymbol{\epsilon}(x,y,z) represents a weak z -dependent perturbation. The electric field \mathbf{E}(x,y,z) in a waveguide with a dielectric tensor \boldsymbol{\epsilon}(x,y,z) satisfies Maxwell’s equation \nabla\times\nabla\times\mathbf{E}+\frac{\omega^{2}\mu}{c^{2}}\,\boldsymbol{\epsilon}\cdot\mathbf{E}=0. (1) We can rewrite ( 1 ) as: \nabla\times\nabla\times\mathbf{E}+\frac{\omega^{2}\mu}{c^{2}}\bigl(\overline{\boldsymbol{\epsilon}}\cdot\mathbf{E}+\delta\boldsymbol{\epsilon}\cdot\mathbf{E}\bigr)=0. (2) We write the total field as: \mathbf{E}(x,y,z)=\mathbf{E}_{0}(x,y)\,e^{i\beta z}+\mathbf{E}_{r}(x,y,z), where \mathbf{E}_{0} is a known modal field solution of the ideal waveguide and satisfies: \nabla\times\nabla\times\bigl(\mathbf{E}_{0}(x,y)\,e^{i\beta z}\bigr)+\frac{\omega^{2}\mu}{c^{2}}\,\overline{\boldsymbol{\epsilon}}\cdot\mathbf{E}_{0}(x,y)\,e^{i\beta z}=0. (3) Therefore, substituting \mathbf{E} into equation ( 2 ) and neglecting second-order correction terms (i.e., \delta\boldsymbol{\epsilon}\cdot\mathbf{E}_{r} ), we arrive at \nabla\times\nabla\times\mathbf{E}_{r}+\frac{\omega^{2}\mu}{c^{2}}\,\overline{\boldsymbol{\epsilon}}\cdot\mathbf{E}_{r}=-\,\frac{\omega^{2}\mu}{c^{2}}\,\delta\boldsymbol{\epsilon}\cdot\mathbf{E}_{0}\,e^{i\beta z}. (4) Equation ( 4 ) can be interpreted as an equivalent problem in which the scattered fields, \mathbf{E}_{r} ,

Read the complete text