Lithium niobate has long been regarded as the “silicon of photonics” due to its excellent optoelectronic properties. The recent development of wafer-bonding technology has enabled lithium niobate photonic integrated circuits (PICs), producing a wide range of applications [1–11]. The advantages of a lithium niobate PIC platform have been demonstrated in electro-optic modulators and frequency combs [4,12,13]. However, further investigation has revealed that lithium niobate has many disadvantages, including large photorefraction, high birefringence, and slow relaxation of its electro-optic response, which can limit the stability of on-chip electro-optic modulators. Compared with lithium niobate, lithium tantalate has electro-optic and piezoelectric coefficients at similar levels, but possesses a higher optical damage threshold ($1500{\mathrm{W}/\mathrm{cm}}^2$ at 514.5 nm) [14], a broader transparency window (0.28–5.5 μm) [15], lower birefringence (0.004) [16], lower microwave loss ($6.5\times {10}^{-4}$) [17], and more stable electro-optic response [18,19]. Meanwhile, its application in 5G filters has made it to the stage of high-volume production, making it a low-cost, ready-to-use material for large-scale integration [15]. Therefore, lithium tantalate can be an even more reliable, more preparation-friendly material for the development of next-generation PICs.

Here, we fabricated ultrahigh-$Q$ microring cavities, together with the coupling bus waveguide and grating couplers on thin-film lithium tantalate. We characterized the fabricated devices with scanning electron microscopy (SEM) and atomic force microscopy (AFM). We analyzed the mode families in the measured transmission spectra and identified the statistical relationship between the $Q$ factor and the resonator geometry. We measured the highest loaded optical quality factor of $>{10}^7$ from a nearly critically coupled microring cavity, which corresponds to the lowest propagation loss of $\sim 1.88\mathrm{dB}/\mathrm{m}$. We also conducted detailed experimental measurements of the photorefractive effect in these ultrahigh-$Q$ lithium tantalate microring cavities under various intracavity powers and observed a $<3.2\mathrm{GHz}$ frequency shift even under a high intracavity power of 320 mW, indicating $\sim 500$ times weaker photorefractive effect than their lithium niobate counterparts.

We fabricated the microring cavities on 400-nm-thick $x$-cut lithium tantalate on 2-μm silicon oxide on a silicon substrate (NanoLN). We fabricated the microring cavities with a top-down nanofabrication process. First, a thin layer of silicon oxide was deposited by plasma-enhanced chemical vapor deposition on the lithium-tantalate-on-insulator wafer as the hard mask layer. Second, the device patterns were defined by high-resolution electron-beam lithography and transferred to the hard mask layer by using plasma etching with a gas mixture of ${\mathrm{SF}}_6/{\mathrm{C}}_4{\mathrm{F}}_8$. Third, the device patterns were transferred further from the hard mask layer to the lithium tantalate device layer by argon plasma (${\mathrm{Ar}}^{+}$) milling in an ion-beam-etching system. Fourth, the remaining hard mask was removed by wet etching in a buffered oxide etchant, and then the device chip was soaked in a mixture of ${\mathrm{H}}_2{\mathrm{O}}_2$, ${\mathrm{NH}}_4\mathrm{OH}$, and ${\mathrm{H}}_2\mathrm{O}$ with a volume ratio of 2:2:1 at 85°C to remove the redeposition [20,21]. Lastly, the device chip was annealed at 500°C for 2 h to enhance the crystalline quality.

Figure 1(a) shows a top-view optical microscope image of a fabricated microring cavity (radius $r=80\mathrm{\mu m}$) with a coupling bus waveguide. The gap between the microring and the bus waveguide is 500 nm, and the coupling length is varied to achieve critical coupling of the microring. Figure 1(b) shows a zoomed-in SEM image of the straight part of the bus waveguide, indicating smooth etched sidewalls. Figure 1(c) shows the simulated cross-sectional optical modal profiles of the ${\mathrm{TE}}_0$ mode in the straight part of the bus waveguide with a waveguide width $w=1.5\mathrm{\mu m}$ and in the microring cavity with a waveguide width $w=4\mathrm{\mu m}$. A wider waveguide has a smaller propagation loss due to less scattering loss on the etched sidewalls. Figure 1(d) shows a cross-sectional SEM image of the straight part of the bus waveguide, which indicates an etched depth $h_{\mathrm{rib}}=200\mathrm{nm}$, a slab thickness $h_{\mathrm{slab}}=200\mathrm{nm}$, and a steep sidewall making an angle of $\sim 60°$ with respect to the chip surface. Figure 1(e) shows the measured AFM image of the coupling region between the microring and the bus waveguide. This measurement not only reconfirms the etched depth $h_{\mathrm{rib}}=200\mathrm{nm}$ but also shows that the gap between the microring and the bus waveguide is $\sim 500\mathrm{nm}$. Figure 1(f) shows the measured surface roughness profile of an etched region, with the root-mean-square (RMS) value as low as $\sim 0.481\mathrm{nm}$.

To facilitate fiber-to-chip coupling, we fabricated on the same device chip grating couplers with different grating pitches $p$ and etched trench widths $t$. Figure 2(a) shows an SEM image of a fabricated grating coupler. Figure 2(b) shows a close-up view of the region marked with the white rectangle in Fig. 2(a). We fixed $p-t=600\pm 50\mathrm{nm}$ and varied the ratio $t/p$ from 0.44 to 0.49. Figure 2(c) plots the measured transmission spectra for our fabricated grating couplers with three different $t/p$ values. We observed the center wavelength of 1520, 1560, and 1610 nm for $t/p=0.44$, 0.47, and 0.49, respectively. They all show a coupling efficiency of approximately $-7\mathrm{dB}$ and a 3-dB bandwidth of $\sim 100\mathrm{nm}$.

We characterized the fabricated microring cavities by sending light from a tunable semiconductor laser to the input of the bus waveguide and measuring the intensity of the transmitted light at the output of the bus waveguide. Figure 3(a) shows a broad-range normalized transmission spectrum of a microring cavity with a ring radius $r=100\mathrm{\mu m}$. This spectrum exhibits various resonance line shapes featuring different linewidths and extinction ratios, because the microring with a 4-μm width can support multiple transverse modes. The mode family for the resonances can be determined based on their line shapes, extinction ratios, and free spectral ranges (FSRs). The resonances that belong to the fundamental mode can be identified from its largest FSR ($\sim 1.8\mathrm{nm}$) due to its smallest group index. We measured the highest $Q$ factors from the fundamental mode among all the mode families, which is attributed to its smallest overlap with the waveguide sidewalls. By fitting the measured resonance dips with a Lorentzian line shape, we obtained the highest loaded quality factor $Q_{\mathrm{L}}=1.48\times {10}^7$, as shown in Fig. 3(c). The corresponding intrinsic quality factor is $Q_{\mathrm{int}}=2.33\times {10}^7$, as inferred from $Q_{\mathrm{int}}=2Q_{\mathrm{L}}/(1+\sqrt{T_0})$ under the under-coupled condition [22], where $T_0$ represents the normalized on-resonance transmission. Then, we obtained the lowest propagation loss $\alpha =1.88\mathrm{dB}/\mathrm{m}$ for the ${\mathrm{TE}}_0$ mode based on $\alpha =10{\log }_{10}({\mathrm{e}}^{-{\kappa }_{\mathrm{i}}{n}_{\mathrm{g}}L/c})$ [23], by using the intrinsic linewidth ${\kappa }_{\mathrm{i}}/2\pi =9.6\mathrm{MHz}$, the group index $n_{\mathrm{g}}=2.155$, the waveguide length $L=1\mathrm{m}$, and the speed of light $c=3\times {10}^8\mathrm{m}/\mathrm{s}$. Other resonances in the same mode family also exhibit ultrahigh $Q$ factors exceeding ${10}^7$. Figure 3(b) shows a resonance at the wavelength of 1571.597 nm, with $Q_{\mathrm{L}}=1.1\times {10}^7$ and the corresponding $Q_{\mathrm{int}}=2.0\times {10}^7$. Figure 3(d) shows a resonance at the wavelength of 1575.183 nm, with $Q_{\mathrm{L}}=1.22\times {10}^7$ and the corresponding $Q_{\mathrm{int}}=2.02\times {10}^7$.

In another device with a bend radius $r=80\mathrm{\mu m}$, by adjusting the coupling length between the microring and the bus waveguide, we could efficiently mitigate the excitation of higher-order modes in the microring cavity. Figure 4 plots the measured normalized transmission spectrum of a microring cavity in the wavelength range of 1595–1607 nm, where only the ${\mathrm{TE}}_0$ mode was excited with an FSR of $\sim 2.34\mathrm{nm}$. The insets show zoomed-in spectra of the resonances at the wavelengths of 1595.250, 1597.577, 1599.917, 1602.267, and 1604.629 nm. These resonances exhibit single or split dips, which may result from mode splitting due to the coupling between the clockwise and counterclockwise modes. Such modal coupling is usually weak and can only be observed in ultrahigh-$Q$ cavities where the resonance linewidth is even smaller than the modal coupling rate.

Next, we experimentally studied the photorefractive effect in these ultrahigh-$Q$ lithium tantalate microring cavities. This effect arises from light-induced intrinsic electric field, which leads to alteration in the material’s refractive index at different light powers. We chose a resonance of the ${\mathrm{TE}}_0$ mode at 1599.917 nm with $Q_{\mathrm{L}}=6.61\times {10}^6$ and $Q_{\mathrm{int}}=1.02\times {10}^7$ to investigate the photorefractive effect. These measured $Q$ factors are slightly lower than those of the 100-μm-radius device, which may be attributed to stronger light scattering on the sidewalls of a bent waveguide with a smaller bend radius. Figure 5(a) plots the measured normalized transmission spectra obtained by sweeping the laser wavelength in the short-to-long wavelength direction at a fixed sweep rate of 1 nm/s. We can observe that the resonance wavelength shifts slightly to shorter wavelengths as the waveguide power increases from $-21.08$ to $-8.13\mathrm{dBm}$, which is due to the photorefractive effect of lithium tantalate. Meanwhile, the resonance line shape also deviates from an ideal Lorentzian line shape at high incident powers, which is due to the thermo-optic effect. In addition, frequency shift or line shape deformation of the resonances can occur as a result of competition between the photorefractive and thermo-optic effects, and if so, nonlinear optical oscillation and pulsation can be observed during the wavelength sweeping [24]. However, we do not observe such phenomena in Fig. 5(a), which also confirms that the photorefractive effect is weak in the lithium tantalate microring cavities. Figure 5(b) plots the frequency shift of the resonance as a function of intracavity optical power based on the measured data in Fig. 5(a). The intracavity optical power $P_{\mathrm{cir}}$ can be inferred from the input power in the bus waveguide $P_{\mathrm{in}}$ based on the relationship $P_{\mathrm{cir}}=\frac{\mathrm{FSR}}{\pi \mathrm{\Delta }{f}_{\mathrm{FWHM}}}\frac{2Q_{\mathrm{L}}}{Q_{\mathrm{c}}}{P}_{\mathrm{in}}$ [22], where the coupling quality factor $Q_{\mathrm{c}}$ is $1.88\times {10}^7$ based on its definition $Q_{\mathrm{c}}=Q_{\mathrm{inc}}{Q}_{\mathrm{L}}/(Q_{\mathrm{inc}}-Q_{\mathrm{L}})$ and $\mathrm{\Delta }{f}_{\mathrm{FWHM}}$ is the FWHM linewidth of the resonance. It is clear that for compact microrings with a small radius (a large FSR), achievement of critical coupling and narrow resonance linewidth contributes to large resonant enhancement for intracavity power buildup. For our fabricated lithium tantalate microring cavities, the power enhancement factor $P_{\mathrm{cir}}/P_{\mathrm{in}}$ is $\sim 2167$ for the ${\mathrm{TE}}_0$ mode. As the intracavity power increases up to 320 mW, the resonance frequency shift is at most 3.2 GHz. This photorefractive effect is $\sim 500$ times weaker than that observed in lithium niobate microring cavities [25].

In conclusion, we have fabricated ultrahigh-$Q$ microring cavities on thin-film lithium tantalate and experimentally demonstrated its low photorefractive effect. We adopted SEM and AFM to characterize the surface morphology of the fabricated devices, confirming that the RMS roughness of the etched surface is as low as 0.481 nm. The measured highest loaded optical quality factor is $1.48\times {10}^7$ for the ${\mathrm{TE}}_0$ mode, from which an intrinsic quality factor of $2.33\times {10}^7$ and the corresponding waveguide propagation loss of $\sim 1.88\mathrm{dB}/\mathrm{m}$ can be inferred. We also investigated the photorefractive effect in the ultrahigh-$Q$ microring cavities by measuring the resonance frequency shift at different input powers. Even at the intracavity power of 320 mW, the resonance frequency shift is $<3.2\mathrm{GHz}$, indicating $\sim 500$ times weaker photorefractive effect than their lithium niobate counterparts. We believe that these ultrahigh-$Q$ lithium tantalate microring cavities will find wide applications, including telecommunication, sensing, nonlinear photonics, and quantum photonics.