Systematic Analysis of Ferroelectric Domain Dynamics in Periodically Poled Lithium Niobate Waveguides Using Two-Photon Microscopy and Digital Imaging Processing

Abstract

We present a characterization and analysis methodology suitable for volume production for characterizing and optimizing x-cut thin-film periodically poled lithium niobate (PPLN) devices using two-photon (2P) microscopy with quantitative image processing. This method enables direct extraction of key structural parameters, such as duty cycle, phase-matching behavior, and domain uniformity, across large device sets in a non-destructive manner. By correlating 2P microscopy-derived structural metrics with systematic variations in poling conditions, we establish a scalable, image-driven approach for evaluating and improving PPLN fabrication. The resulting workflow supports wafer-level process development and accelerates the manufacturing and packaging of lithium niobate photonic integrated circuits (PICs).

Figures (7)

• Figure 1: Schematic of reflection non-interference two-photon (2P) microscopy for imaging periodically poled lithium niobate (PPLN) waveguides. (a) In this configuration, two near-infrared excitation beams are tightly focused onto the PPLN waveguides on the chip, generating second-harmonic (SH) signals through localized two-photon absorption within both inverted and uninverted ferroelectric domains. (b) The SH waves generated in oppositely poled domains possess a $\pi$ phase difference, creating interference-enhanced contrast at domain boundaries and revealing variations in inversion depth. (c) Representative 2P images of (i) an unpoled TFLN waveguide and (ii) a periodically poled waveguide. The strong contrast between adjacent domains enables high-fidelity visualization of the poling pattern.
• Figure 2: Workflow for SHG image processing and domain duty-cycle extraction, along with representative 2P microscopic images of under-poled, critically poled, and over-poled PPLN waveguides. (a) A representative SHG microscopic image of a PPLN waveguide region, with the waveguide direction indicated by an arrow. (b) Binarized inverted-domain map obtained by thresholding, where bright regions correspond to poled (inverted) domains and dark regions correspond to unpoled (uninverted) areas. (c) Edge detection and length analysis applied to the binarized image to identify the left (red dashed lines) and right (blue dashed lines) boundaries of each inverted domain. (d) Local duty-cycle distribution $D_i = w_{\mathrm{rev},i}/w_{\mathrm{per},i}$ as a function of period index, showing the spatial variation and mean duty cycle across the analyzed waveguide. (e) Images of under-poled PPLN waveguides. (i) Example showing an inverted-domain duty cycle smaller than 50$\%$, indicating insufficient domain broadening. (ii) Example of an under-poled PPLN region exhibiting incomplete domain inversion. (f) Images of critically poled PPLN waveguides. (i) and (ii) show representative examples of PPLN waveguides with small and large electrode gaps, respectively, both exhibiting near-50$\%$ domain duty cycles characteristic of optimal poling. (g) Images of over-poled PPLN waveguides. (i) An example in which the inverted-domain width significantly exceeds the uninverted-domain width, indicating an over-poled condition. (ii) If the poling conditions that promote domain inversion are further intensified, the PPLN becomes fully inverted across the entire waveguide, as illustrated here.
• Figure 3: Second-harmonic generation (SHG) microscopy images showing the evolution of ferroelectric domain structures in PPLN waveguides under different poling conditions.(a). PPLN waveguides poled at different temperatures, showing the evolution of domain morphology with increasing poling temperature. The SHG contrast highlights changes in domain uniformity and boundary sharpness as a function of the applied poling parameters. Image i. corresponds to $145^{\circ}C$, Image ii. to $160^{\circ}C$, and Image iii. to $180^{\circ}C$. All other poling parameters were kept constant: electric-field strength = 15.38 V/$\mu$m, number of pulses = 3, and electrode gap = 13 $\mu$m. As the poling temperature increases, the fraction of inverted domains gradually increases, with the average duty cycles measured as 0.4195, 0.6357, and 0.6484 for Images i-iii, respectively. (b). PPLN waveguides poled with different numbers of pulses, showing the evolution of domain morphology as the pulse count increases. Image i corresponds to 3 pulses, Image ii to 5 pulses, and Image iii to 9 pulses. All other poling parameters were held constant: electric-field strength = 15.38 V/$\mu$m, temperature = $160^{\circ}C$, and electrode gap = 13 $\mu$m. With increasing pulse number, the fraction of inverted domains increases progressively, and the average duty cycles are 0.6357, 0.6725, and 0.7503 for Images i-iii, respectively. (c). PPLN waveguides poled under different electric-field strengths, illustrating the effect of increasing field on domain morphology. Image i corresponds to a poling field of 15.38 V/$\mu$m, Image ii to 16.62 V/$\mu$m, and Image iii to 19.69 V/$\mu$m. All other poling parameters were kept constant: number of pulses = 3, temperature = $160^{\circ}C$, and electrode gap = 13 $\mu$m. As the poling field increases, the proportion of inverted domains grows gradually, with average duty cycles of 0.6145, 0.6357, and 0.7010 for Images i-iii, respectively. (d–e). Images comparing domain morphologies of PPLN waveguides fabricated with different electrode gaps under two poling temperatures. (d). Poling temperature = $145^{\circ}C$: Image i corresponds to a 13 $\mu$m electrode gap and Image ii to a 20 $\mu$m gap, with average duty cycles of 0.509 and 0.522, respectively. (e) Poling temperature = $160^{\circ}C$: Image i corresponds to a 13 $\mu$m gap and Image ii to a 20 $\mu$m gap, with average duty cycles of 0.700 and 0.569, respectively. All other poling parameters were identical: electric-field strength = 19.69 V/$\mu$m and number of pulses = 3. At the lower temperature ($145^{\circ}C$), a larger electrode gap results in a slightly higher average duty cycle, whereas at the higher temperature ($160^{\circ}C$), a smaller gap yields a higher duty cycle, indicating an interplay between temperature and gap size in determining domain inversion efficiency.
• Figure 4: Residuals vs Fitted - Extended Model. Residuals scatter randomly around zero across the fitted range with no visible funneling, indicating no strong heteroscedasticity or leverage-driven patterns.
• Figure 5: Q-Q plot of extended-model residuals. Points follow the 45$^\circ$ line closely with only minor tail deviations, consistent with approximate normality.