Efficiency and bandwidth improvement of integrated acousto-optic modulators using dielectric acoustic reflectors Open Access

With the development of advanced photonic integrated circuits (PICs),^1–6 integrated acousto-optic devices as an emerging branch have been accordingly demonstrated using opto-mechanical–electrical properties of the planar waveguide media.^7,8 Representative acousto-optic modulators (AOMs) exhibit strong manipulation capabilities for both photons and phonons, resulting in integrated acousto-optic frequency shifters (AOFSs), deflectors,^9,10 switches, isolators,^11,12 filters,¹³ converters,¹⁴ and so on. Surface acoustic waves (SAWs) excited by an interdigital transducer (IDT) are utilized to accomplish the acousto-optic interactions in AOMs. To control and enhance acousto-optic interactions, the engineering of acoustic wave propagation as a key technique is continually concerned combining with the physical properties of piezoelectric materials and waveguide configurations.

Advancements in thin-film piezoelectric material fabrication technologies have led to the rise of integrated AOMs using various waveguide platforms, such as lithium niobate (LN),^14,15 aluminum nitride (AlN),^16,17 aluminum scandium nitride (AlScN),¹⁸ and gallium nitride (GaN).¹⁹ Thin-film lithium niobate (TFLN) as an attractive medium has been proven to be an ideal waveguide platform for fabricating high-efficiency, low-loss AOMs in qubit transduction, microwave-to-optical conversion, and frequency manipulation due to its favorable piezoelectric and photoelastic advantages.^10,20–24 Limited by the low optomechanical and electromechanical coupling coefficients, weak acousto-optic modulation efficiency and narrow-bandwidth modulation characteristics have become both critical bottlenecks for pushing integrated TFLN AOMs into practical stage. To address the aforementioned problems, some acoustic cavities are commonly adopted to increase acousto-optic interaction in previous studies.^25,26 A conventional acoustic cavity is composed of metal reflectors or suspended membrane.^11,27 However, the precise fabrication of narrow-linewidth metal reflectors needs stringent process control, bringing great challenges for the large-scale applications of the metal reflectors. Similarly, the introduction of the suspended acoustic resonator also faces the complex fabrication recipe, meanwhile the modulation bandwidth of the device could be seriously decreased. Therefore, how to increase the modulation bandwidth under a single-finger IDT configuration while maintaining appropriate modulation efficiency on a non-suspended waveguide platform is worth pursuing for improving the stability and practicability of integrated TFLN AOMs.

Giving that the chalcogenide glass (ChG) has high mechanical reflectivity, large photoelastic coefficient, and low acoustic velocity, ChG waveguides combining with the hybrid-integration TFLN are expected to show excellent acoustic characteristics.^28,29 In this paper, we propose and experimentally demonstrate efficient and wideband Mach–Zehnder interferometer (MZI) AOMs based on the dielectric acoustic reflectors and non-suspended TFLN–ChG hybrid waveguides. The loss and reflection characteristics of the Rayleigh SAW over the ChG-LN hybrid waveguide platform are investigated. Periodic ChG acoustic gratings are carefully designed to reflect specific acoustic waves within the acoustic band along the different propagation orientations to engineer modulation performance of the single-arm MZI AOMs. Compared with counterparts without acoustic reflectors, the modulation efficiency of one device with acoustic reflectors is enhanced under the X–Z orientation, while the modulation bandwidth of the other X–30°Y device is broadened. Demonstration of ChG acoustic reflectors thus provides new solutions for the flexible control and manipulation of acoustic wave propagation in integrated acousto-optics.

The schematic diagram of the proposed dielectric acoustic reflector-engineered MZI AOM is shown in Fig. 1(a). The AOM consists of a hybrid MZI, a pair of ChG acoustic reflectors, and a single-finger IDT on non-suspended TFLN–ChG hybrid waveguide platform. Acoustic reflectors are located on both sides of an IDT to confine acoustic waves. An 850 nm-thick (H) Ge₂₅Sb₁₀S₆₅ (one of the ChGs) film is deposited on the X-cut TFLN with a thickness of 400 nm to form the hybrid MZI. The hybrid MZI ridge waveguide is made of the ChG rectangular waveguide and a TFLN slab. The Rayleigh SAWs are excited by an IDT at an angle θ relative to the crystal Y axis to explore the propagation behaviors of acoustic waves. The width of the ChG rectangular waveguide is set to be 1.6 μm (W), which matches the half acoustic wave's wavelength (Λ = 3.2 μm) for maximum acousto-optic overlap. The refractive indices of amorphous ChG and anisotropic LN are set to be n = 2.273 and n_e = 2.13 (n_o = 2.2) at 1550 nm, respectively. The stresses resulting from the SAWs will deform the hybrid waveguide, leading to a change in the refractive index of the supported fundamental transverse electric mode (TE00), as shown in Fig. 1(b).

To rigorously verify the effects of the dielectric acoustic reflectors in the engineering of single-arm MZI AOMs, both IDT configurations with and without acoustic reflectors are integrated on the same MZI waveguide, as shown in Fig. 2(a). The primary purpose of the configuration is to contrast their modulation performance. The detailed fabrication processes are illustrated in our previous study.²⁹ The integrated AOM is characterized using a heterodyne detection setup, as shown in Fig. 2(b). In this setup, light emitted from a tunable laser (Santec, TSL-570) is split into 99:1 with a coupler to generate a signal light and a reference light. The signal light is coupled into the device under test via a lens fiber with a mode field diameter of 3 μm. A fiber polarization controller is employed to adjust the fundamental TE mode into the waveguide. The vector network analyzer (Keysight, N5231B) interfaces with the IDT via port 1 to excite the SAW. The output light generated from the device is converted into an electrical signal through a photodiode, thereafter being detected by port 2 of the vector network analyzer to observe the RF transmittance spectrum (S₂₁). The reference light is coupled into an acousto-optic frequency shifter with a 120 MHz offset to enable the heterodyne detection via a photodiode with a responsivity of 450 V/W. The RF sideband spectra are measured using an electrical signal analyzer (Agilent, N9010A).

Before conducting the measurements of the AOMs, we design experiments to explore the propagation behaviors of the Rayleigh SAW excited using an IDT, and the results are shown in Fig. 3. By measuring reflection (S₁₁) and transmittance (S₂₁) spectra of the SAWs over the X–Z TFLN, as presented in Fig. 3(a), it could be seen that there are two main acoustic modes with resonance frequencies of 0.857 and 0.899 GHz in the range of 1 GHz, corresponding to the Rayleigh and shear modes, respectively. By performing the inverse Fourier transform for the S₂₁ spectrum, the system response |h| of the IDT group with a spacing of 500 μm is obtained, as shown in Fig. 3(b). Through linear fitting, the propagation loss of the Rayleigh SAW is determined to be 3.04 dB/mm. The propagation velocity of the Rayleigh SAW is calculated to be 2857 m/s; the detailed method can be obtained in Ref. 31.

To quantitatively estimate the reflection characteristics of the ChG waveguides for the SAWs, a rectangular ChG waveguide with a width of 1.6 μm is first placed between both IDTs, and the S₂₁ spectrum is characterized based on the same TFLN platform and IDT configurations, as shown in Fig. 3(c). The introduction of the ChG modulation waveguide alleviates the vibrations in the S₂₁ spectrum resulting from the round trip SAW interference in Fig. 3(a). Subsequently, the ChG acoustic gratings with 15 periods are put on one side of the waveguide to influence SAWs through observing the variation of the S₂₁ spectrum, as shown in Fig. 3(d). It could be clearly seen that the variation of S₂₁ is small for an improper ChG acoustic reflector (w = 0.9 μm, gap = 1 μm). For a proper configuration (w = 0.9 μm, gap = 1.1 μm), serious reduction in S₂₁ is observed. Finally, through the optimization, the S₂₁ of SAWs significantly decreased by 20 dB because the corresponding Rayleigh SAWs are obviously reflected in comparison with the case of Fig. 3(c). Of course, a small part of acoustic energy is attenuated. However, the shear mode at 0.899 GHz outside the bandgap is nearly unaffected. The difference between the two S₂₁ lies in whether they pass through the reflectors, and the average loss induced by a unit acoustic grating is calculated to be 1.33 dB. Then the reflection coefficient of the ChG acoustic grating is nearly calculated to be 0.34, as described in the  Appendix, approaching the reflection coefficients of some common metal gratings (e.g., aluminum ∼0.32 and gold ∼0.46).^32–34 Therefore, the configuration of the dielectric acoustic grating is helpful to adjust and control the performance of an integrated MZI AOM.

To demonstrate the effect of the dielectric acoustic reflectors on the modulation efficiency, we measure the modulation characteristics of both single-arm X–Z MZI AOMs (device A) with and without acoustic reflectors, which are presented in Fig. 4. Compared with the configuration without acoustic reflectors [see Fig. 4(a)], the device featuring reflectors manifests a nearly 10 dB improvement in S₂₁ (from −65.1 to −55.5 dB) at an acoustic frequency of 0.841 GHz under the RF power of 0 dBm [see Fig. 4(b)]. At the acoustic frequency of 0.845 GHz, the value of S₂₁ has slight reduction due to energy distribution changes caused by acoustic reflectors (as shown in S₁₁ spectra). The bias point of device A is chosen at 1560.58 nm (−21.27 dBm), and the modulation length is designed to be 120 μm. Utilizing these data and formula in Ref. 29, the half-wave-voltage V_π is thus extracted to be 6.3 V (18.9 V) for the device A with acoustic reflectors (without acoustic reflectors), and the corresponding half-wave-voltage-length product V_πL is calculated to be 0.075 V cm (0.22 V cm). Repeatable measurement results are shown in Table I. These results indicate that the configuration with reflectors obtains a twofold enhancement in modulation efficiency. The enhancement can be attributed to the strong reflection of acoustic waves within the bandgap and standing wave oscillation in the acoustic cavity.

To further reveal the difference of modulation characteristics of both configurations, we observe the evolutions of RF sidebands of the device using a heterodyne detection setup, as shown in Fig. 4(c). When the RF power is 15 dBm, up to second-order RF sidebands could be acquired for device A with acoustic reflectors. For device A without acoustic reflectors, only one-order RF sidebands could be obtained. We also explore the variations of intensities of the generated one-order RF sideband (anti-stokes signal) with RF power increased from −5 to 15 dBm for both configurations. The almost linear dependence indicates that the magnitude of the RF sideband is proportional to the amplitude of the acoustic wave. Crucially, the configuration with acoustic reflectors exhibits a superior modulation efficiency compared to that without reflectors as the RF power increases.

The introduction of dielectric acoustic reflectors can not only enhance the modulation efficiency of a MZI AOM but can also expand the modulation bandwidth of a device. To verify this point, we design a ChG acoustic grating-loaded single-arm MZI AOM (device B) with X–30°Y orientation aiming to explore the evolution of modulation bandwidth. Figures 5(a) and 5(b) show the acoustic S₁₁ and opto-acoustic S₂₁ spectra of device B under the configurations without and with acoustic reflectors, respectively. Different from the acoustic S₁₁ of device A, device B presents many more acoustic modes within the range from 0.822 to 0.9 GHz, corresponding to the diverse peaks in opto-acoustic S₂₁. Through estimating the profile of peaks in S₂₁ spectra, the 3 dB bandwidth of device B is increased from 11.8 to 20.5 MHz (fractional BW of 2.45%) with the help of the acoustic reflectors, as shown in Figs. 5(a1) and 5(b1). Herein, the 3 dB bandwidth is defined as the frequency interval at half maximum, referring to Ref. 35. It could be observed that modulation abilities of the partial acoustic modes around the center resonance frequency (0.838 GHz) in S₁₁ are enhanced, while some new acoustic modes are excited by the acoustic cavity to participate in the modulation. To clearly clarify the modulation mechanism, we analyze the evolutions of admittances corresponding to the acoustic S₁₁ spectra of device B under both configurations, as shown in Fig. 5(c). Obvious fluctuations in admittance Y₁₁ spectra verify the variations of parasitic modes within the acoustic bandgap that is induced using the acoustic reflectors. The presence of acoustic reflectors thus expands the bandwidth of device B, which is also demonstrated by the comparable RF sidebands measured at the four different S₂₁ peaks under the RF power of 15 dBm [see Fig. 5(d)].

Table II presents a comparative analysis of our acoustic reflector-engineered single-arm MZI AOMs with single-finger IDTs based on the non-suspended TFLN–ChG hybrid waveguides, against existing mainstream TFLN MZI AOMs. Compared with the MZI AOMs with metal reflectors,^26,27 device A exhibits superior V_πL up to 0.07 V cm, which is attributed to the creation of acoustic standing wave and increased amplitude of strain around the modulation waveguide. Although the introduction of a suspended acoustic cavity could significantly improve the V_πL of the device in Ref. 7, the bandwidth (BW) of the device is distinctly compressed, as demonstrated in our previous study.²⁹ Based on the choice of ChG acoustic reflectors, the BW of device B is broadened to 20.5 MHz in our work, while the V_πL of device B is maintained at a suitable level. Given that there is a trade-off between the BW and V_πL, BW divided by V_πL as a figure-of-merit (FOM) is defined to estimate the modulation performance of a device. As shown in Table II, the FOMs of our proposed device A and device B with acoustic reflectors are calculated to be 41.3 and 89.1, respectively, verifying the good modulation characteristics of the proposed TFLN MZI AOMs under the single-arm and single-finger configurations. Of course, to further increase the BW of a device, a split IDT or a chirped IDT is preferred to manipulate the propagation of acoustic waves in combination with the engineering of acoustic reflectors. In addition, narrow-linewidth IDT is also pursued to increase the modulation rate of a device. Therefore, the introduction of the dielectric acoustic reflectors provides a new pathway to the development of efficient and wideband AOMs integrated on chip.

In conclusion, we propose and demonstrate dielectric acoustic reflector-engineered MZI AOMs with single-arm configurations based on the non-suspended TFLN–ChG hybrid waveguide platform, which efficiently reveals the reflection mechanism and control capability of ChG acoustic gratings for the SAWs excited by the single-finger IDTs. Through the engineering of ChG acoustic reflectors, the V_πL of device A is improved from 0.22 to 0.075 V cm, while the BW of device B is increased from 11.8 to 20.5 MHz. The propagation loss of traveling Rayleigh SAW near 0.85 GHz is estimated to be to be 3.04 dB/mm, and the propagation velocity of the Rayleigh SAW in the experiments is confirmed to be 2857 m/s. The reflection coefficient of the ChG acoustic grating is nearly demonstrated to be 0.34. These experimental results render the current devices appealing for future work in the development of high-performance acousto-optic devices, such as efficient and wideband AOMs, non-reciprocal photonic devices based on the stimulated Brillouin scattering, and qubit transduction devices.

This work was supported by the National Natural Science Foundation of China (Grant Nos. 62175095, 62335014, and 12134009), the Science Fund for Distinguished Young Scholars of Ningxia (Grant No. 2024AAC04003), the Natural Science Foundation of Ningxia (Grant No. 2022AAC03239), the Key Research and Development Program of Ningxia (Grant No. 2024BEH04093), the Open Project of State Key Laboratory of Photonics and Communications in Shanghai Jiao Tong University (Grant No. 2024GZKF002).

The authors have no conflicts to disclose.

Huilong Liu and Jiying Huang contributed equally to this work.

Huilong Liu: Investigation (equal). Jiying Huang: Investigation (equal); Writing – original draft (equal). Jiantao Jiang: Methodology (supporting). Hongyi Huang: Methodology (supporting). Huipeng Chen: Methodology (supporting). Chengyu Chen: Methodology (supporting). Haiwei La: Methodology (supporting). Huan Li: Data curation (supporting). Xiaochun Xu: Data curation (supporting). Yuping Chen: Supervision (equal). Zhaohui Li: Supervision (equal). Lei Wan: Project administration (equal); Supervision (equal); Writing – review & editing (equal).

The data that support the findings of this study are available from the corresponding authors upon reasonable request.