Advances in Electro-Optical Devices Enabled by Waveguide-Based Thin-Film Lithium Niobate

Abstract

Lithium niobate (LN) materials have become a key platform for constructing core optoelectronic devices such as electro-optic (EO) modulators, optical frequency combs, and integrated optical waveguides, owing to their broad transparent window, mature waveguide processes, and excellent electro-optic effect. They demonstrate revolutionary application value in light source generation, signal transmission, and intensity modulation of optical communication systems, and are hailed as the “silicon of the photonics field,” attracting significant attention from both academia and industry. Especially with the commercialization of high-quality thin-film lithium niobate (TFLN) materials, the performance of thin-film optoelectronic devices based on waveguide structures has achieved leapfrog improvements, with their loss characteristics and modulation bandwidth far exceeding those of traditional bulk material devices. This paper systematically combs the photonic properties of LN materials, introduces in detail the electro-optic effect and electro-optic modulation principle of LN electro-optic modulators, reviews some recent research achievements of scholars, focuses on expounding the preparation processes of waveguide-based TFLN, the types of waveguide-based optoelectronic devices, and the research progress of these devices, and discusses and compares the advantages and development potential of different routes.

1. Introduction

LN is referred to as “the silicon of photonics” to emphasize its significance in photonics, analogous to the role of silicon in microelectronics. This is because LN possesses many properties required for photonic integrated circuits, including a wide optical transparency window, a large electro-optic effect, and a phase transition temperature, which are crucial for process compatibility and maintaining stable operation [1]. These superior physical and chemical properties have enabled commercially available off-the-shelf LN solutions to be widely adopted as the “workhorse of optoelectronic technology” for decades, and they continue to be extensively used today in numerous key fields such as integrated photonics [2], optical communications [3,4], computing [5], frequency metrology [6], and spectroscopy [7,8]. In recent years, the thin-film monolithic LN platform has emerged, combining low loss, high-quality photonic integration capabilities, and a strong Pockels effect, thus demonstrating excellent modulation performance. With its potential as a photonic integrated circuit and future photonic interconnection, thin-film monolithic LN holds broad prospects in advancing the development of electro-optic modulation and integrated photonics.

As a core device in photonics, LN waveguide modulators have undergone significant development, evolving from traditional bulk diffusion waveguides to modern nanophotonic thin-film platforms. Early conventional modulators relied on proton exchange or ion diffusion technologies [9]. While low driving voltages could be achieved through long metal electrodes, their integration into high-performance systems was hindered by limitations such as low refractive index contrast, high dielectric constant, and large bending radius. At the beginning of the 21st century, lithium niobate on insulator (LNOI) wafers were successfully fabricated with reference to the silicon-on-insulator (SOI) technology [10], laying the foundation for breakthroughs in integrated circuits. Combined with advancements in nanomanufacturing technologies such as optimized argon ion milling and reactive ion etching, this enabled the fabrication of ultra-low-loss (<0.3 dB/cm) waveguides with high confinement. Furthermore, hybrid approaches based on ridge loading or heterogeneous integration have further enhanced device performance [11]. This technological transformation not only overcomes the physical limitations of traditional platforms but also promotes comprehensive optimization of key modulator metrics (e.g., half-wave voltage, bandwidth, and loss), opening up new paradigms for the development of next-generation highly integrated photonic circuits. This paper aims to provide an overview of waveguide-based TFLN optoelectronic devices, covering their fundamental principles and latest advancements.

2. Photonic Properties of Thin-Film Lithium Niobate

LN features a broad transparency window ranging from 350 nm to 5 μm, covering the visible, near-infrared, and mid-infrared wavelength ranges. Its relatively large refractive index (approximately 2.2 at a wavelength of 1550 nm) enables the formation of high-refractive-index-contrast waveguides on most amorphous and crystalline substrates (such as SiO₂ or sapphire). Its high Curie temperature (~1210 °C) ensures the stability of its ferroelectric phase, making it compatible with a wide range of manufacturing processes and operating conditions. Unlike Si and SiN_x, LN is a non-centrosymmetric crystal with a large second-order nonlinear coefficient (d₃₃ = 27 pm/V). Moreover, it allows for ferroelectric domain engineering, which makes it a primary material for nonlinear optical frequency conversion and generation [12]. Crucially, LN’s exceptional Pockels coefficient (r₃₃ = 31 pm/V) has established LN-based modulators as core components in optical communication networks.

3. The Working Principle of Lithium Niobate Electro-Optic Modulator

3.1. Electro-Optic Effect

In an electrostatic field or low-frequency electric field environment, the optical properties of dielectric materials can be regulated through the electro-optic effect, and this change can be accurately characterized by the dynamic variation of the relative dielectric impermeability tensor. Specifically, when an external field acts on a dielectric, its internal charge distribution and polarization state will undergo corresponding adjustments, which in turn lead to changes in light transmission properties. As a key physical quantity describing the electromagnetic response of materials, the relative dielectric impermeability tensor has tensor elements whose changes can directly reflect the modulation process of optical properties induced by the electro-optic effect [13]:

$${\eta }_{ij}\left(E_0\right)={\eta }_{ij}+∆{\eta }_{ij}\left(E_0\right)={\eta }_{ij}+\sum _k{\gamma }_{ijk}{E}_{0k}+\sum _{k,l}{S}_{ijkl}{E}_{0k}{E}_{0l}+\cdots$$

(1)

Among them, η_ij is a quantity independent of the electric field; γ_ijk is the linear electro-optic coefficient, also known as the Pockels coefficient; and S_ijkl is the quadratic electro-optic coefficient, also referred to as the Kerr coefficient. The Pockels effect is a first-order electro-optic effect, which means the effect where η_ij(E₀) is linearly related to the applied electric field E₀ through the coefficient γ_ijk. The Kerr effect is a second-order electro-optic effect, referring to the effect that is quadratically related to the applied electric field E₀ through the coefficient S_ijkl. Optical modulators operating based on the Kerr effect are generally called all-optical modulators. It should be noted that both the Pockels effect and the Kerr effect are nonlinear optical effects.

The LN electro-optic modulator for optical communication is based on the Pockels effect. LN material has a negative uniaxial crystal structure with 3 m point group symmetry. When an external electric field E is applied to it, the change in the crystal’s refractive index can be expressed as:

$$∆{\displaystyle \frac{1}{n_i^2}}={\sum }_{j=1}^3{\gamma }_{ij}{E}_j$$

(2)

In the formula, γ_ij is a 6 × 3 Pockels coefficient matrix, where E₁ = E_x, E₂ = E_y, and E₃ = E_z. Among the elements of the Pockels coefficient matrix of LN, the largest one is γ₃₃, and the refractive index change is the most significant when E_z acts [13]. Therefore, to achieve efficient electro-optic modulation, the electric field is usually applied along the z-axis direction of the LN crystal to utilize the largest linear electro-optic coefficient. The crystal materials commonly used in LN photonic devices include two types: x-cut and z-cut [14]. The x-cut LN crystal devices are designed in the y-z plane, which can be fabricated by planar lithography. They utilize the transverse electric mode of light, and the microwave electrodes are in the same plane as the device layer, facilitating test and analysis. The z-cut crystal devices are designed in the x-y plane, utilizing the transverse magnetic mode, and require a structure with top and bottom electrodes to apply the electric field along the z-axis direction of the crystal.

3.2. Electro-Optic Modulation

The traditional nonlinear optical formalism for handling the EO process in LN [15,16] is based on the concept that an applied voltage introduces a phase shift into the optical field, which can be expressed as:

$$∆\varphi =\beta =\pi {\displaystyle \frac{V}{V_{\pi }}}$$

(3)

Here, V_π represents the voltage required for π-phase shift. β is the modulation index or modulation depth, which is used to quantify the amplitude of the phase shift and indicates how much of the π-phase shift is induced by the applied voltage. Applying a time-varying electric field, such as a sinusoidal microwave signal, to the electro-optic material causes the induced phase shift Δϕ to oscillate with the amplitude of the microwave signal. This results in the generation of new frequency components that are either higher or lower than the frequency components of the incident light, with the difference given by the frequency of the microwave signal [17,18]. This process is known as EO modulation and is the fundamental mechanism of electro-optics. Physically, EO modulation involves the interaction between microwave fields and optical fields, where the frequency of optical photons is altered by adding or subtracting microwave photons.

In terms of applications, EO modulation can bridge the microwave and optical domains, which is crucial for global optical communication networks [19] as shown in Figure 1. It can imprint analog microwave signals onto optical carriers and encode digital information into the amplitude and phase of optical signals. Conversely, optical signals of different frequencies can interact through electromagnetic effects to generate microwave signals that can be used to excite or control electronic systems. Typically, EO modulation is employed to rapidly modify the degrees of freedom of light, such as its temporal profile (e.g., by switching lights on and off) or spectral content (e.g., by generating sidebands).

4. Lithium Niobate Waveguide Fabrication Processes

4.1. Rib Load

As an important structure in LNOI integrated photonics, the rib-loaded waveguide effectively solves the problems of significant scattering and propagation losses caused by the difficulty in etching LN itself, the roughness of etched sidewalls, and the redeposition of chemical etching by-products [20,21]. Compared with other types of waveguides, LN rib-loaded waveguides can achieve lower propagation losses due to the smaller overlap between the optical field and the rib sidewalls [22]. Another advantage is that by appropriately selecting the material and geometric parameters of the rib, the mode size can be adjusted while maintaining a single mode, making it easier to couple with lensed fibers [23]. The specific scheme is to use materials with a refractive index close to that of LN for rib-loading of TFLN, see Figure 2a, and this method has been verified and applied in various materials, such as tantalum pentoxide (Ta₂O₅) [24,25], titanium dioxide (TiO₂) [26], chalcogenide glass (ChG) [27], Ge₂₃Sb₇S₇₀ [28,29,30], and silicon nitride (SiN) [31,32]. However, rib-loaded waveguides have polarization-dependent scattering losses, especially the inherent loss of the transverse magnetic mode caused by coupling with the transverse electric planar mode at the waveguide boundary [33]. This problem was later found to be eliminable by optimizing the waveguide cross-sectional parameters. In recent years, the loss characteristics of controlling rib-loaded waveguides have been interpreted as utilizing the principle of optically bound state in continuous media [34]. In addition, rib-loaded has also been applied to bulk LN crystals to form compact waveguides using single-crystal [35,36] or amorphous Si [37].

4.2. Dry Etch

Dry etching can be categorized into two types: F-based gas etching and Cl-based gas etching [38]. When F-based gas is used for dry etching, LiF by-products are generated. Due to its high melting point, LiF is difficult to remove and tends to accumulate on the sidewalls, thereby reducing the etching sidewall angle. It has been reported that when fabricating LN optical waveguide devices on x-cut substrates using a mixed gas of SF₆ and Ar, the sidewall angle of the resulting waveguides is 60–75° [39]. In contrast, the by-product generated during the etching of LN with Cl-based gas is LiCl, which has a lower melting point and is thus easy to clean without accumulating on the sidewalls, ultimately forming a superior etching morphology. Additionally, Cl-based gas can avoid the consumption of non-metallic masks and solve the compatibility issue between metal masks and traditional semiconductor processes. However, the etching rate of Cl-based gas is lower than that of F-based gas, and it has higher requirements for etching equipment, which to some extent limits its application in large-scale production and deep etching processes. For z-cut substrates, using a mixed gas of Ar, Cl₂, and BCl₃ [40] achieved an etching selectivity ratio of 1.45:1, with a sidewall angle of 83° and a smooth etched surface. In 2020, Shen et al. [41] also obtained an etching sidewall angle close to 80° using a SiO₂ mask and a mixed etching gas of Ar, Cl₂, and BCl₃. Li’s team proposed a process that combines plasma enhancement technology with dry etching [42], this process not only enables high-speed etching of TFLN but also avoids the redeposition of by-products. Low-loss channel waveguides with an etching depth of approximately 900 nm and excellent sidewall verticality were fabricated on x-cut TFLN substrates. These advancements have laid a solid foundation for the manufacturing of high-precision TFLN devices and the development of LN photonic integration technology. As shown in Figure 2b [43], dry-etched waveguide structures have been applied in TFLN electro-optic modulators (EOMs) [44] and nonlinear devices [45]. Compared with rib-loaded structures, this method can achieve stronger optical field confinement in LN, which in turn allows for a reduced electrode spacing in Mach-Zehnder devices, effectively lowering the half-wave voltage (V_π) of electro-optic modulators. However, when there are certain requirements for ultra-high-speed design, the rib-loading method is more advantageous because the dielectric constant of the rib is lower than that of LN.

4.3. Silicon on Insulator Bond

LN thin films and silicon-on-insulator (SOI) wafers achieve hybrid integration through precision bonding technology. This scheme effectively compensates for the inherent defect of silicon materials in terms of second-order optical nonlinearity, and can combine the excellent optical properties of LN materials with the scalability of the SOI platform, enabling the realization of TFLN devices with high performance and compatibility with complementary metal-oxide-semiconductor (CMOS) processes [46,47] (Figure 2c). The current mainstream bonding schemes are mainly divided into two categories: one is silica direct bonding technology, which has been successfully applied to silicon-lithium niobate [48] and silicon nitride-lithium niobate [49] modulators, achieving an ultra-large electro-optical bandwidth exceeding 110 GHz; the other is direct bonding schemes based on adhesive polymers such as benzocyclobutene (BCB) [50,51,52]. It is worth noting that although the BCB bonding process has high maturity, there are potential risks in its long-term stability and reliability. Therefore, the SiO₂-based direct bonding technology has become the preferred solution in this field due to its excellent long-term performance. In the preparation process of hybrid waveguides, by optimizing the design of silicon waveguide width, most of the energy of the optical mode can be confined in the TFLN region, thereby achieving efficient regulation of optical properties. However, the narrowing of the silicon waveguide also brings drawbacks: the interaction between the optical mode and the sidewalls of the silicon waveguide is significantly enhanced, which is very likely to introduce additional optical loss. In addition, although some studies have successfully explored the process steps for TFLN bonding and substrate removal [53], this method still faces many challenges, among which a key problem is to precisely control the number of layers of the thin bonding material between the SOI and TFLN regions. It is worth noting that after completing the bonding process, metal electrodes can be deposited and patterned on the top of the structure, laying a foundation for the preparation of EOMs and other optoelectronic devices [54].

Figure 2. Common waveguide structures of TFLN: (a) Rib load [22]. (b) Dry etch [43]. (c) Silicon On Insulator Bond [46]. (a) Reprinted with permission from Ref. [22]. Copyright 2016, Optica Publishing Group. (b) Reprinted with permission from Ref. [43]. Copyright 2018, Springer Nature. (c) Reprinted with permission from Ref. [46]. Copyright 2019, Springer Nature.

Figure 2. Common waveguide structures of TFLN: (a) Rib load [22]. (b) Dry etch [43]. (c) Silicon On Insulator Bond [46]. (a) Reprinted with permission from Ref. [22]. Copyright 2016, Optica Publishing Group. (b) Reprinted with permission from Ref. [43]. Copyright 2018, Springer Nature. (c) Reprinted with permission from Ref. [46]. Copyright 2019, Springer Nature.

5. Types of Waveguide-Based Electro-Optic Modulators

On the TFLN platform, EO devices based on waveguide structures have become the most widely used option due to their excellent performance. Typical devices include phase modulators, amplitude modulators, and in-phase and quadrature (IQ) modulators (Figure 3). These devices are versatile and powerful, capable of performing operations such as phase tuning, signal switching, optical frequency comb generation, and precise spectral control. They play a crucial role in fields like optical communication and optical signal processing. This chapter will delve into their basic structural designs, working mechanisms, as well as the cutting-edge research progress and latest achievements in related fields.

5.1. Basic Structure

EO phase shifters and modulators can be constructed by passing a waveguide structure through a pair of electrodes [55] (as shown in Figure 3a). Amplitude modulators can be further realized by embedding phase modulators in interferometers or resonators. Currently, the most widely used type of modulator is the traveling-wave Mach-Zehnder modulator (MZM) (as shown in Figure 3b), whose core structure consists of two arms of a Mach-Zehnder interferometer (MZI) passing through the two gaps of a coplanar waveguide electrode, respectively. It is worth noting that the V_π of a MZM is only half that of a phase modulator of the same length, which stems from the opposite phase shifts generated in the two arms of the interferometer. Combining two MZMs can form an IQ modulator (Figure 3c), which is an important component of coherent communication systems.

5.2. Phase Modulator

The working mechanism of a phase modulator is to apply voltage to the electrodes on both sides of the waveguide, which induces a change in the refractive index of the waveguide, thereby causing an additional phase shift in the transmitted light. When driven by a strong radio frequency signal, the phase modulator generates sidebands, and as the radio frequency driving power increases, higher-order sidebands are produced sequentially. This characteristic enables it to play a key role in fields such as frequency comb generation [56] and frequency shifting, and it is also an indispensable core component in high-precision optical gyroscope systems [57].

In 2019, the Ren team [58] developed a dual-channel phase modulator based on a Ground-Signal-Ground electrode structure. It achieves group velocity matching between light and radio frequency through a 50 Ω impedance-matched traveling-wave radio frequency transmission line and adopts a push-pull design to reduce the V_π. In the experiment, driven by a 30 GHz radio frequency signal with a power of about 3.1 W, it generated an optical frequency comb covering 10 nm in the telecommunication L-band with more than 40 sidebands. A central challenge in phase modulators is mitigating residual intensity modulation (RIM), where the driving voltage inadvertently modulates both the phase and intensity of the light wave. To address this, Shi’s team [59] developed a RIM correction model in 2020 specifically for LN phase modulators. Their model highlights distinct RIM characteristics based on laser linewidth: under wide-linewidth conditions, the electric field alters the optical mode field shape, leading to refractive index changes and loss, resulting in an approximately linear voltage-RIM relationship. In narrow-linewidth scenarios, interference from stray light reflected by the substrate induces a nonlinear RIM response. This model is a crucial step toward device optimization. To ensure optimal performance and stability, an adaptive bias control module was integrated, dynamically adjusting bias points through a pilot tone method. This resulted in a side-mode suppression ratio consistently exceeding 20 dB, peaking at 29.53 dB, with only a 1.79 dB fluctuation over an 8-h period, demonstrating the device’s excellent long-term stability. Alongside advances in RIM correction and stability, researchers are also focused on scaling down device sizes and optimizing structural integration of TFLN modulators to enhance overall functionality.

To improve the performance of TFLN modulators, researchers have also conducted a series of studies, such as on reducing device size and optimizing structural integration. In 2022, Yu’s team [60] built a single-channel time lens system on a TFLN, integrating components such as an amplitude modulator, a recycling phase modulator, and a chirped Bragg grating to develop an integrated femtosecond pulse generator (Figure 4a) with a chip size of only 25 mm × 4 mm. Its recycling phase modulator allows light waves to pass through the electrode gap twice, achieving a low V_π of 2–2.5 V in the frequency range of 4–39 GHz, and can generate 520 fs pulses with a repetition rate of 30 GHz, laying a foundation for the application of single-channel structures in efficient modulation. In 2024, the Du team further innovated and proposed a folded recycling phase modulator (Figure 4b) [61]. The folded structure reduces the device size to 0.6 × 12 mm², maintaining a low V_π of 2–3 V in the frequency range of 3.5–40 GHz, and expanding the low V_π bandwidth from the previous 1.5 GHz to 3 GHz. Meanwhile, the modulation efficiency is improved by extending the optical waveguide to optimize microwave-light speed matching, demonstrating progress in size reduction and V_π control. On this basis, the Zhu’s group [62] developed a folded V-shaped LN phase modulator with a V_π below 1 V (Figure 4c). For the ultra-long modulation length of 7.5 cm, through multiple foldings (including 180° bends) and the design of a transition region for the folded capacitively loaded traveling-wave electrode, it achieves a V_π of 0.52 V while maintaining a compact size, with a 3 dB bandwidth of 10 GHz. Moreover, it ensures refractive index matching between microwave and optical signals through path difference compensation. The above two groups both optimize the size with the folded structure as the core, further expanding the applicability of the folded design in scenarios with longer modulation lengths.

To further the goals of device miniaturization, enhanced modulation efficiency, and reduced V_π, researchers are actively pursuing advanced design strategies. In 2024, Cheng’s group [63] leveraged a photon wire-bonded TFLN platform to optimize the phase modulator’s size and performance through a recycling structure. This 3 cm-long modulator, equivalent to 6 cm in effective optical interaction length, utilizes a dual-arm modulation scheme that allows the light wave to traverse the electrode region twice. Combined with a single-electrode driving design that facilitates shared transmission lines with an amplitude modulator, it resulted in a compact overall chip size of 3 mm × 3.45 cm, while maintaining a low V_π of 1.45 V across the 10–30 GHz frequency range. This approach demonstrates the continuing benefits of the recycling structure for size reduction. Concurrent with this, the Gao team et al. [64] explored the concept of dual-arm modulation to enhance both performance and integration capabilities. Utilizing photolithography-assisted chemical mechanical etching, they developed a compact dual-arm TFLN electro-optical phase modulator. This design doubles the modulation effect for a chip length of about 1 cm, achieving a V_π as low as 3 V, half that of single-arm modulators. The modulator exhibited excellent performance in the 12–26 GHz frequency band, with an insertion loss of only 2.8 dB, effectively addressing issues of high V_π and large chip size. This contribution represents a significant advance in miniaturization and efficient modulation of TFLN modulators. In a separate development, Wang’s group [65] presented an ultra-compact electro-optical phase modulator based on a LN valley photonic crystal (VPC) structure (Figure 4d). The design exploits the VPC’s unique slow-light effect to achieve a powerful electro-optical modulation within an ultra-compact size of just 4 µm × 14 µm, with a V_π of 1.4 V. The spin-valley locking effect resulted in high transmittance, reaching 0.87 at 1068 nm. Furthermore, the design is compatible with existing nanomanufacturing technologies and is relatively insensitive to manufacturing defects.

Figure 4. (a) Schematic diagram of a phase modulator on a Si-LNOI platform [60]. (b) Schematic diagram of a foldable and recyclable phase modulator [61]. (c) Folded phase modulator with sub-1V half-wave voltage [62]. (d) Schematic diagram of an ultra-compact electro-optic phase modulator based on a lithium niobate valley photonic crystal structure [65]. (a) Reprinted with permission from Ref. [60]. Copyright 2022, Springer Nature. (b) Reprinted with permission from Ref. [61]. Copyright 2024, Wiley. (c) Reprinted with permission from Ref. [62]. Copyright 2025, IEEE. (d) Reprinted with permission from Ref. [65]. Copyright 2024, Optica Publishing Group.

Figure 4. (a) Schematic diagram of a phase modulator on a Si-LNOI platform [60]. (b) Schematic diagram of a foldable and recyclable phase modulator [61]. (c) Folded phase modulator with sub-1V half-wave voltage [62]. (d) Schematic diagram of an ultra-compact electro-optic phase modulator based on a lithium niobate valley photonic crystal structure [65]. (a) Reprinted with permission from Ref. [60]. Copyright 2022, Springer Nature. (b) Reprinted with permission from Ref. [61]. Copyright 2024, Wiley. (c) Reprinted with permission from Ref. [62]. Copyright 2025, IEEE. (d) Reprinted with permission from Ref. [65]. Copyright 2024, Optica Publishing Group.

5.3. Amplitude Modulator

The core principle of an amplitude modulator is based on the interference effect of light, among which the traveling-wave MZM is the most widely used. The fundamental operating principle is illustrated in Figure 3b. Upon entering the electrode structure, an incoming light beam is divided into two independent optical signals that propagate within the waveguide. With zero applied voltage, these two beams recombine with a phase difference of zero, resulting in the modulator’s “on” state. Applying a voltage induces opposing electric fields in the two optical paths, introducing phase shifts of π/2 and −π/2, respectively. This causes the recombined beams to destructively interfere, leading to zero output optical power and the modulator’s “off” state. By precisely controlling the phase difference between the optical signals in the two arms, the intensity of the combined beam can be effectively modulated.

Building on this principle, researchers have advanced amplitude modulators with significant breakthroughs. In 2018, Wang et al. developed an integrated electro-optical modulator compatible with CMOS on a thin-film LNOI platform (Figure 5a) [43]. Utilizing a traveling-wave MZI and leveraging the high refractive index contrast (>0.7) of LNOI, the electrode-waveguide gap was reduced to less than 2.5 μm. Optimized microwave and photonic circuits yielded high electro-optical efficiency, ultra-low optical loss, and group velocity matching. The device achieves a V_π of 1.4 V, a 3 dB bandwidth over 45 GHz for a 20 mm length, and supports data rates up to 210 Gbit/s with on-chip optical loss below 0.5 dB. This addresses issues of traditional LN modulators, such as large volume (>5 cm) and high driving voltage (>3.5 V). Building on this, Zhang’s team systematically studied the millimeter-wave performance of TFLN modulators (Figure 5b). They expanded the operational frequency from microwave and low-frequency bands to 30–300 GHz [66]. By shortening the modulation length to 5.8 mm, they increased the 3 dB electro-optical bandwidth to 170 GHz. Additional optimizations included using a 20 μm-wide transmission line to balance impedance and radio frequency loss, replacing gold with 800 nm-thick copper electrodes to reduce ohmic loss, and employing a 2 μm-thick buried oxide with 0.6 μm-thick cladding oxide to enhance microwave-optical velocity matching. These advancements facilitate the transition from traditional microwave communication to high-frequency applications such as millimeter-wave radar and ultra-high-speed wireless links.

Addressing the bandwidth limitations of high-Q ring modulators on TFLN platforms, Xue et al. proposed an electro-optical ring modulator based on a novel “ring resonator + MZI coupler” hybrid structure (Figure 5c) [67]. This design employed a pure coupling modulation mechanism, circumventing the modulation speed limitations imposed by photon lifetime, common in high-Q devices. This TFLN modulator achieved a 67 GHz 3 dB electro-optical bandwidth, significantly exceeding that of other high-Q designs, using a 2 mm modulation section with periodic capacitively loaded traveling-wave electrodes. The high-Q ring resonator (intrinsic Q = 7.7 × 10⁵) yielded a low V_π of 1.75 V (voltage-length product, 0.35 V·cm) within a compact 3.4 mm × 0.7 mm footprint. This modulator also supported 240 Gbps driver-free data transmission, paving the way for further bandwidth enhancements by overcoming the device’s inherent bottleneck, the resonator. Expanding on bandwidth limits, the Thomaschewski team introduced a plasmonic TFLN Mach-Zehnder modulator (Figure 5d) [68]. They utilized gold nanobars to create a plasmonic phase shifter on the LN surface, optimizing electro-optical field overlap via strong field confinement. The use of asymmetric gold nano waveguides (250 nm and 550 nm widths) constructed an asymmetric structure to passively ensure quadrature point operation and achieve modulation linearity. Push-pull phase modulation further reduced the voltage-length product to 0.23 V·cm. The device, with a 15 μm-long phase shifter and 150 nm isolation air gap, featured a sub-1 mm² area, a further size reduction from the Xue’s team ring modulator. Performance evaluations revealed a measured 3 dB bandwidth exceeding 10 GHz, but the ultra-fast response of the plasmonic mode suggests a potential bandwidth of 900 GHz. This work represents a shift from breaking the resonator bottleneck to surpassing the waveguide mode bandwidth limit, furthering the progress of TFLN modulation towards ultra-high frequency and miniaturization by virtue of strong field confinement.

In 2023, Li et al. developed a high-performance MZM on a TFLN platform [69]. Utilizing a coplanar waveguide electrode and a partially etched waveguide, the design optimized the optical and electric field overlap and velocity matching, yielding a low voltage-length product of 1.29 V·cm at 1550 nm and a 3 dB electro-optical bandwidth exceeding 40 GHz. While not surpassing the ultra-high bandwidth of plasmonic technology, this design offers advantages in low-voltage operation and multi-wavelength compatibility. The studies by Xue [67] and Thomaschewski [68] teams have achieved significant bandwidth improvements through modulation mechanism innovation and plasmonic enhancement respectively. However, in the pursuit of ultra-high bandwidth, the problem of increased V_π is often encountered. Consequently, maintaining low driving voltage while improving bandwidth is a critical challenge for the advancement of TFLN modulators. Addressing this, Farzaneh et al. proposed a novel TFLN optical modulator design in the same year (Figure 5e) [70]. By asymmetrically positioning the optical waveguide relative to the electrode and incorporating a dielectric buffer layer, the modulator achieved a voltage-length product of 2.2 V·cm with an 84 GHz 3 dB electro-optical bandwidth. A second demonstrated design extrapolated a 3 dB bandwidth of 170 GHz with a voltage-length product of 3.3 V·cm. Furthermore, the modulator successfully demonstrated 240 Gb/s eight-level pulse amplitude modulation data transmission, presenting a promising approach for low-voltage, ultra-high-frequency bandwidth TFLN electro-optical modulators in next-generation optical communication systems.

Figure 5. (a) Monolithic LNOI monolithic modulator based on Mach-Zehnder interferometer [43]. (b) Schematic diagram of a millimeter-wave photonic system and schematic diagram of a TFLN millimeter-wave-optical modulator [66]. (c) Schematic diagram of an electro-optic ring modulator [67]. (d) Plasmonic lithium niobate unbalanced MZM [68]. (e) Cross-sectional view of a high-speed modulator with asymmetric waveguides and microscopic image of the TFLN modulator [70]. (a) Reprinted with permission from Ref. [43]. Copyright 2018, Springer Nature. (b) Reprinted with permission from Ref. [66]. Copyright 2022, Chinese Laser Press. (c) Reprinted with permission from Ref. [67]. Copyright 2022, Optica Publishing Group. (d) Reprinted with permission from Ref. [68]. Copyright 2022, American Chemical Society. (e) Reprinted with permission from Ref. [70]. Copyright 2023, Wiley.

Figure 5. (a) Monolithic LNOI monolithic modulator based on Mach-Zehnder interferometer [43]. (b) Schematic diagram of a millimeter-wave photonic system and schematic diagram of a TFLN millimeter-wave-optical modulator [66]. (c) Schematic diagram of an electro-optic ring modulator [67]. (d) Plasmonic lithium niobate unbalanced MZM [68]. (e) Cross-sectional view of a high-speed modulator with asymmetric waveguides and microscopic image of the TFLN modulator [70]. (a) Reprinted with permission from Ref. [43]. Copyright 2018, Springer Nature. (b) Reprinted with permission from Ref. [66]. Copyright 2022, Chinese Laser Press. (c) Reprinted with permission from Ref. [67]. Copyright 2022, Optica Publishing Group. (d) Reprinted with permission from Ref. [68]. Copyright 2022, American Chemical Society. (e) Reprinted with permission from Ref. [70]. Copyright 2023, Wiley.

5.4. In-Phase and Quadrature Modulators

As data transmission capacity and speed continue to increase rapidly, optical interconnection technology faces an urgent need for further enhancements. In this context, IQ modulators based on coherent transmission have demonstrated significant advantages, enabling simultaneous encoding of information onto both the amplitude and phase of optical signals. By integrating advanced modulation formats with polarization division multiplexing technology, these modulators markedly improve spectral efficiency, playing a vital role in expanding the capacity of high-speed optical communication systems.

In 2020, the Xu team launched a high-performance coherent IQ modulator based on TFLN (Figure 6a) [71]. This design achieved a coordinated optimization between low loss, low driving voltage, and large bandwidth. Its core comprised a nested MZM, using a 1 × 2 multimode interference coupler for optical power splitting and combining, and incorporating a thermo-optic phase shifter to independently control the bias of the I/Q branches. This design addressed the issues of the large size (30–80 mm) and high loss inherent in traditional LN modulators. The modulator featured a waveguide with a top width of 4 µm and a ridge height of 300 nm, along with traveling-wave electrodes (900 nm thick, 7 µm gap). Balancing voltage and bandwidth, it achieved a V_π of 1.9–3.1 V, a 3 dB electro-optical bandwidth of 48–67 GHz, supported 320 Gbit/s 16-quadrature amplitude modulation transmission, and had an on-chip loss of only 1.45–1.8 dB. In 2022, on the basis of the results of 2020, Xu’s team [72] focused on the demand for higher transmission rates and polarization multiplexing, and developed a dual-polarization in-phase quadrature (DP-IQ) modulator. Structurally, it innovatively integrates four MZMs and a polarization rotation combiner (PRC). The PRC achieves a high polarization extinction ratio (≥20.5 dB) and low insertion loss (<0.3 dB). Furthermore, capacitively loaded traveling-wave electrode optimized microwave-optical velocity matching, overcoming the rate limitations of single-polarization structures. Compared to the 2020 results, this modulator reduced the V_π from 1.9–3.1 V to 1 V, while increasing the 3 dB bandwidth from 67 GHz to over 110 GHz. Through dual-polarization multiplexing, it achieved 1.96 Tb/s single-wavelength transmission for the first time, in a volume 30% smaller than the previous version. This design maintained low voltage advantages while breaking through limitations in data rate and polarization integration. In 2023, aiming at the volume and high-frequency loss issues faced by the first two generations of modulators in high-density integration, Xu et al. [73] proposed a folded structure DP-IQ modulator (Figure 6b). Based on the 2022 dual-polarization architecture, this design introduced an air-bridge structure and beveled electrodes at the U-shaped electrodes. The air-bridge suppresses radio frequency wavefront distortion caused by the coupled groove pattern, and the beveled design reduced the path difference between the inner and outer electrodes, resolving high-frequency performance degradation attributable to the folded structure. The design enabled a modulation length of 22.5 mm while compressing the physical size to 4 mm². Maintaining a 1 V V_π and a 67 GHz bandwidth, it supported 703 Gbit/s transmission with an energy consumption as low as 479 aJ/bit, achieving attojoule-level energy consumption in thousand-kilometer-level transmission for the first time. The use of a quartz substrate further reduced microwave loss, improved electro-optical bandwidth, and demonstrated good stability in high-frequency signal transmission, rendering it suitable for high-performance, long-distance communication scenarios.

Compared to silicon substrates, TFLN wafers based on quartz substrates offer superior electro-optic properties but are limited by higher costs, lower yields, and smaller sizes, restricting large-scale integrated applications. Conversely, silicon substrates are more cost-effective and facilitate easier heterogeneous bonding with distributed feedback lasers, making them advantageous for device integration. In 2024, the Larocque team proposed a photonic crystal cavity IQ modulator based on TFLN (Figure 6c) [74]. This device employs a Fabry–Perot-enhanced Michelson interferometer structure, integrating two photonic crystal cavities as phase shifters, with optical signals distributed and combined via directional couplers. Electrodes apply an electric field to tune the cavity resonance, achieving a tuning efficiency of about 1 GHz/V, a bandwidth of approximately 1.5 GHz, and supporting 1 GHz-rate 4-quadrature amplitude modulation with a 2 V peak-to-peak CMOS-compatible voltage. It features an extinction ratio exceeding 30 dB and a compact chip size of only 40 × 200 µm², offering solutions for compact coherent communication and quantum information processing. While this cavity-based design excels in compactness and low energy consumption, inherent characteristics of photonic crystal structures pose challenges to further improving bandwidth and loss performance.

Further optimization by Tang et al. on silicon substrates addressed velocity matching and microwave loss issues [75]. They proposed a TFLN IQ modulator with a 15 µm-thick silicon dioxide buried layer on a silicon substrate (Figure 6d). This modulator uses two MZMs for the I and Q branches, along with multimode interference (MMI) couplers and capacitively loaded traveling-wave electrodes. The thick silicon dioxide buffer layer enables microwave-optical velocity matching. Experimental results show an electro-optical bandwidth exceeding 67 GHz, fiber-to-fiber loss of as low as 4 dB, an extinction ratio above 30 dB, and a V_π of 3.5 V demonstrating support for high-speed, high-order modulation formats. These advances in structural design and performance lay a solid foundation for practical applications of high-performance TFLN modulators, especially in signal generation and sensing.

In 2024, Lei’s team applied the TFLN IQ modulator to an optical carrier suppression single-sideband signal generation system [76]. The setup features two parallel traveling-wave MZM for in-phase and quadrature components, using a 1 × 2 MMI splitter/combiner. Ground-Signal-Ground traveling-wave electrodes generate equal-amplitude, anti-phase signals for near-chirp-free modulation, complemented by three electro-optic phase shifters. To ensure optimal performance, an adaptive bias control module was implemented. This module dynamically adjusted bias points utilizing a pilot tone method, achieving a side-mode suppression ratio exceeding 20 dB, peaking at 29.53 dB, with only a 1.79 dB fluctuation over 8 h. This demonstrates the excellent long-term stability of the device. In a frequency-modulated continuous wave ranging experiment, they generated a 1 GHz bandwidth, 1 µs period chirp signal, measuring a standard deviation of ±2.60 mm within 1.5 m. This system provides critical technical support for chip-level applications such as LiDAR and microwave photonics.

6. Conclusions

This paper reviews the recent advancements in waveguide-based TFLN optoelectronic devices. Starting from the intrinsic photonic properties of TFLN and the working principles of LN electro-optic modulators, it details the fabrication processes—including rib loading, dry etching, and hybrid silicon integration—and introduces various types of waveguide-based devices such as phase modulators, amplitude modulators, and IQ modulators, along with their research progress. A comparative analysis of different technical approaches highlights their respective advantages and developmental prospects.

In summary, lithium niobate, renowned for its excellent photonic properties—and increasingly benefiting from the commercialization of high-quality TFLN—has significantly propelled the performance improvements of waveguide-based optoelectronic devices. These advances have led to numerous breakthroughs across various fields, providing robust technical support for optical communications and beyond. While challenges remain such as further reducing propagation losses, optimizing the trade-off between modulation bandwidth and V_π, and increasing integration density ongoing refinements in fabrication processes and innovative structural designs will further unlock their application potential. These developments are expected to energize next-generation optoelectronic technologies, fostering more efficient, low-cost solutions that will accelerate the growth of related industries.