Lithium niobate (LiNbO₃) has played an essential role in the fields of electro-optic modulation and frequency conversion. In its early stages, it was primarily used as laser frequency-doubling devices and electro-optic modulators in optical communication systems. With the advancement of Ti-diffusion and proton exchange techniques, LiNbO₃ devices were no longer confined to bulk configurations, facilitating the development of a range of integrated optical waveguide components. However, challenges in nanofabrication and constraints on integration density still persist. In the 21st century, with the explosive growth of information and data, traditional electronic computing is gradually approaching its physical limits. Emerging applications, including big data, artificial intelligence, high-speed networks, and virtual reality, are placing increasingly demanding requirements on computing devices in terms of performance, size, and energy efficiency.

Owing to its exceptional electro-optic and nonlinear optical characteristics, LiNbO₃ remains pivotal in traditional communication technologies, while also demonstrating irreplaceable value in emerging fields such as optical computing, artificial intelligence acceleration, and quantum information processing. In particular, the rise of the thin-film lithium niobate (TFLN) platform marks a new era for LiNbO₃ as a platform for optical computing chips, which may have a tremendous impact on the future of information technology.

TFLN, when combined with advanced nanofabrication techniques, has facilitated the development of compact, low-loss, high-speed photonic devices. TFLN has become a critical material platform for the development of high-speed electro-optic modulators, nonlinear photonic chips, and quantum photonic circuits. TFLN has gained increasing attention in the field of optical computing primarily due to the following reasons: Firstly, its strong electro-optic effect fulfills the high-speed modulation requirements for optical computing. Moreover, the low optical loss and high bandwidth of TFLN enable optical signal transmission over a broad frequency range with minimal attenuation. Finally, TFLN supports ultra-high integration density, making it suitable for constructing compact and miniaturized photonic integrated circuits (PICs).

In recent years, the demand for computing processing speed has been approaching the performance bottleneck of electronic computers. Optical computing, as a computing system that relies on photons to process data, has advantages such as low energy consumption, large bandwidth, and fast response, making it one of the most promising disruptive computing architectures. Optical computing is of strategic importance to safeguarding information security and computational sovereignty. Lithium niobate materials and photonics have long been a significant research focus in China, and a series of significant breakthroughs have already been achieved in recent years. Therefore, a systematic review of the progress and prospects of LiNbO₃-based optical computing is essential to guide future research and technological development.

Figure 1 shows the microscopy image of the fabricated chip, which consists of three Mach?Zehnder modulators (MZMs) with various microwave signal line widths and device lengths. Figure 2 presents the schematic of the cross-section of the hybrid waveguide, as well as the multifunctional photonic integrated chip and its characterization system. An integrated lithium niobate (LN) modulator with segmented electrodes (designed to reduce the modulation voltage) has significantly enhanced the modulation bandwidth (Fig. 3). Figure 4 depicts the schematic of an MZM with a periodic CLTW (capacitively loaded traveling-wave) electrode. Figure 5 illustrates the architecture of a typical hybrid photonic neural network (PNN). Figure 6 shows a schematic of the proposed integrated photonic tensor core (IPTC), which consists of four physical components: lasers, two TFLN MZMs, and charge-integration photoreceivers. Figure 7 presents the conceptual schematic of a fully integrated optical convolutional neural network (OCNN), which uses an integrated photonic convolution accelerator (IPCA) fabricated on the lithium niobate-on-insulator (LNOI) platform and a micro-ring resonator (MRR) filter. Figure 8 shows an LN-based microwave photonics (MWP) processing engine, which consists of a high-speed electro-optic modulation block and a low-loss, multipurpose photonic processing section. The implementation of the photon ray-tracing core (PRTC) on the TFLN platform is shown in Fig. 9; the PRTC comprises four high-speed push-pull MZMs for parameter encoding, followed by coherent optical processing and detection components for binary result generation. A polarized TFLN waveguide was fabricated, and non-destructive high-resolution in-situ imaging technology was used for characterization (Fig. 10). Figure 11 provides the experimental setup for characterizing the photon pairs generated from the LNOI waveguide. Furthermore, a photonic chip with dimensions of 50 mm×5 mm×0.5 mm (capable of generating and manipulating entangled photon pairs) has been reported (Fig. 12). Finally, the experimental setup for measuring on-chip quantum interference is also showcased (Fig. 13); this setup can realize multiple photonic information processing functions, including on-chip quantum interference and photon demultiplexing.

The emergence of TFLN has driven revolutionary advances in multiple fields, including optical modulation, nonlinear optical devices, optical computing, and quantum optics, thereby propelling the development of LiNbO₃-based optical computing chips. Notably, critical performance metrics such as operational bandwidth, processing speed, and energy efficiency have undergone remarkable improvements. Importantly, some LiNbO₃ optical computing chips fabricated in laboratories have already achieved performance levels suitable for market applications. However, the industrialization of LiNbO₃ optical computing is still constrained by challenges in material fabrication, nanofabrication techniques, integration density, and system complexity. Bridging the gap from laboratory prototypes to commercial products will require coordinated efforts across materials science, optical engineering, and computing architecture.