In various applications, the ability to precisely and swiftly alter the direction of light is crucial^[1,2]. Among these technologies, lidar, for instance, plays a crucial role in future autonomous navigation systems by emitting lasers and analyzing reflected data to determine an object’s position, velocity, and other characteristics^[3–11]. Unlike traditional radars such as microwave and millimeter wave radar, lidar offers significantly enhanced precision and extended scanning range^[12,13]. It also boasts several advantages, such as reduced sensitivity to environmental lighting and resistance to electromagnetic interference. However, achieving efficient laser beam redirection becomes increasingly difficult, as commercial applications demand faster, smaller, and more efficient lidar systems^[14]. Typically, lidar achieves angular resolution equal to or less than 0.1 mrad, range resolution down to 0.1 m, and velocity resolution up to 10 m/s. These capabilities enable lidar to generate precise three-dimensional (3D) point-cloud images through scanning. With their high coherence and directivity, lasers serve as ideal lidar sources, facilitating long-haul detection with extensive applications across autonomous driving, laser display, cartographic mapping of terrain and oceans, and free-space optical communication. In recent years, researchers have primarily focused on improving and expanding lidar’s scanning field of view (FOV), scanning resolution, and detection distance, predominantly by pursuing miniaturization and integration^[15–19].

Lidar systems can be broadly categorized into three types: mechanical, hybrid solid-state, and solid-state^[20–25]. Mechanical lidar uses expensive optics and a rotating or galvanometer system with mirrors or prisms attached to a mechanical actuator to provide a wide FOV, typically 360°^[26]. Mechanical lidar structures can achieve large signal-to-noise ratios (SNRs) over a wide FOV but have several drawbacks, including large size, bulky design, high cost, reliability and maintenance issues, and sensitivity to vibration. Mechanical lidars almost always use pulsed laser sources. In addition, due to the inertia of the rotating components, it is characterized by high power consumption and is limited to a frequency of about 100 Hz. This has driven research interest towards solid-state beam-steering methods^[27].

Unlike mechanical lidars, solid-state lidars have no moving mechanical parts and therefore have a smaller FOV. However, using multiple sensors and fusing their data, these systems can increase the FOV to a level comparable to that of mechanical lidar. Solid-state lidars fall into two main categories: microelectromechanical systems (MEMS) scanners, which use electromagnetic or piezoelectric actuators to drive micromirrors to scan the FOV, usually supported by extended optics; and optical phased arrays (OPAs), which point beams based on the principle of multibeam interference with optical antenna arrays. MEMS lidars are lightweight, compact in design, and consume low power, and therefore are increasingly used in automotive, transportation, robotics, and space exploration. OPA lidars, like the MEMS structure, do not use moving components but are based on steering the laser beam using optical phase modulators and multiple microstructured waveguides. The OPAs are highly compact and rugged, achieve high measurement speeds above 100 kHz over a wide FOV, and can be realized in a single chip. This type of lidar system can provide very stable, precise, and rapid beam steering, which has emerged as a promising approach^[28–34]. We compare the key parameters of the lidar in Table 1, including scanning range, angular resolution, scanning frequency, and detection distance.

Conventional beam-scanning methods are usually based on mechanical or liquid crystals. Mechanical rotating lidars have the advantage of being able to scan the surroundings with a 360° horizontal FOV, whereas single semi-solid lidars or solid-state lidars tend to be limited to a maximum of 180° horizontal FOV. However, block mechanical systems are limited by their size and therefore speed, with typical response time for liquid crystals being only a few milliseconds. The OPA can quickly and precisely steer light in a nonmechanical way, thus representing a new enabling technology for compact solid-state two-dimensional (2D) beam steering as an alternative to traditional mechanical beam steering including MEMS. While a single OPA lidar has a limited scanning range, a specific number of lidars can be integrated into the vehicle to accomplish 360° coverage simultaneously. By applying OPA technology, the laser beam is guided and scanned within the sensor’s FOV without requiring any rotating parts.

Compared to competing technologies such as liquid crystals and MEMS, one prominent feature of OPAs based on silicon photon platforms is their high operation speed. The scanning speed of 2D OPA is affected by many factors, such as optical device performance, scanning drive mode, control system performance, incident light source, and system structure layout. Since there are no mechanical components in an OPA system, which instead utilizes the radiation field of each antenna unit in the array to interfere in the far field to achieve fast and precise beam control, it is an entirely electronically controlled form of scanning. Therefore, the scanning speed of OPAs is generally limited only by the response time of the phase shifter. In the studies reported so far, optical phase shifters based on thermo-optic and electro-optic effects are capable of achieving scanning rates from megahertz to gigahertz^[35]. When using the OPA as a scanner, the waveguide loss in the silicon-on-insulator (SOI) platform is less than 3 dB/cm, while the bending loss is 0.013 dB/bend. This loss is not bigger than 1 dB, since the total length of the waveguide is of the order of millimeter^[36]. The average efficiency of a grating coupler etched at 70 nm once on a 220 nm silicon top layer is about $-5.2\mathrm{dB}$ using a silicon overlayer, which improves the efficiency to $-1.6\mathrm{dB}$^[37]. This is the efficiency when the grating coupler is used to direct light between the fiber and the photonic integrated circuit. Releasing the space is more efficient than releasing the fiber. Therefore, the total loss from the waveguide input to free space can be limited to less than 3 dB. In silicon-based OPA lidars, lasers, beam splitters, phase modulators, and other electronic devices are integrated onto a single miniaturized platform, leading to significant size reduction^[35,38–43]. The phase modulators can utilize the thermo-optic^[35,57] or electro-optic effect^[41,43] to tilt the far-field wavefront, which can obtain the steered beam. A wide variety of OPAs benchmark from 1–100 kHz, since they employ thermo-optic phase modulation, with response time limited by the thermal diffusivity of the device. Even faster is electro-optic modulation, which enables switching speeds of $>100\mathrm{GHz}$. Although the modulation rate of the thermo-optical phase shifter is not as fast as the electro-optical effect, its process is simple and the loss is small, and it can be used in some application scenarios with low-speed requirements. The relatively weak electro-optical effect of silicon causes the modulator to take up a large space. Additionally, compatibility with mature CMOS fabrication processes offers substantial cost advantages for silicon-based lidar systems^[44,45]. Despite the progress, increasing the FOV and angular resolution remain key challenges in current integrated silicon-based OPA lidars, as the number of components and array size continues to grow.

Similar to radio detection and ranging, target range can be measured by pulsed signals and pulsed time of flight (TOF). However, TOF is one way of performing lidar measurement; the other option is to use the frequency-modulated continuous wave (FMCW) technique, based on heterodyne detection. Comparing with TOF, the FMCW method has the following advantage: the sensing system will not suffer from interference from nearby lidar systems due to the coherent detection nature of FMCW. In addition to distance information, it can obtain the velocity based on Doppler’s effect. Also, the FMCW method can achieve higher depth accuracy compared with TOF. Lastly, FMCW requires relatively lower optical peak power compared with the pulsed TOF method, where a strong optical pulse is required.

Detection range is also one of the key properties of lidar. The lidar’s detection range refers to the maximum distance at which the system can detect a target. This parameter is generally determined by the furthest detection distance of lidar for 10% low reflectivity targets. It is not absolute in practical applications due to variations in the environment and target surface conditions. The OPA-based TOF-type lidar mainly realizes the lidar ranging function through the direct TOF method, and most of the achievable detection distances are within 100 m. At the same time, TOF-type lidar mainly uses a pulsed laser, and when designing for OPA, it is often necessary to pay extra attention to the maximum power that the chip can withstand, and in the field of autonomous driving, it is also essential to take into account the safe power limit of the human eye. In the field of autonomous driving, the speeds of vehicles and pedestrians are critical, and being able to directly measure the speed information is crucial. In this regard, LuminWave, Aeva, and other companies began to study high-performance FMCW lidar. Such lidar has a wide range of measuring, strong anti-interference ability, and can directly measure the speed according to the Doppler effect. The FMCW lidar utilizes a coherent detection method for a very high SNR. Thanks to this, the detection distance of FMCW can realize 150 m or even more under the safety condition of human eyes. Coherent detection technology also brings excellent anti-interference performance, which can resist the interference of daylight, scattered light, stray light, and the mutual interference of other lidars. However, the laser light source plays a decisive role in the overall performance of the FMCW lidar. Long-range detection requires a high-quality light source that meets the requirements of narrow linewidth, wide tuning range, no mode-hopping, fast and high linearity sweep, stable output power, and simple control.

In this paper, we provide a review of recent advancements in integrated solid-state lidar systems, which utilize orthogonal polarizations and counterpropagation to enhance beam steering range and angular resolution. In Sec. 2, we introduce the principle of beam steering realized by silicon-based lidar. Section 3 delves into the strategies and advancements aimed at expanding the scanning range and enhancing spatial resolution along the longitudinal scanning dimensions of silicon-based lidar. This is accomplished using the counterpropagating beams and control over the polarization state of the beam. In Sec. 4, we offer a comprehensive summary and outlook for integrated silicon-based OPA lidar systems using orthogonal polarizations and counterpropagation and evaluate their potential prospects.

The silicon-based OPA lidar technology is rapidly advancing, offering efficient generation of focused beams in specific directions. These devices operate through a series of common steps. First, the incident light is coupled into an integrated waveguide bus and is then evenly distributed among waveguide splitters. A phase shifter is utilized to manipulate the light phase in each channel. Subsequently, the light traverses the emission array and is emitted into free space via a grating coupler (GC) emitting array. The direction of the scanning beam is determined by the design of the emitting components, the wavelength of the incoming light, and the relative phase between adjacent channels. Figure 1 illustrates a fundamental beam-steering apparatus in a seminal work by Acoleyen et al., offering insights into the operational principles of silicon-based OPA^[46]. In this work, a near-infrared laser beam was coupled into a waveguide via a lensed fiber. Energy was uniformly delivered to an array of 16 waveguides, spaced 2 µm apart, via a tree of multimode interferometers (MMIs). The waveguide modes were transmitted through TiN thermo-optic phase modulators, with modulator lengths increasing linearly across the array, inducing a linear phase gradient. The light was coupled into free space by a GC etched in each waveguide, featuring a subwavelength period of 630 nm. As shown in the inset of Fig. 1, each point corresponds to an individual direction of off-chip emission and can be defined by the parameters ($\sin \theta$, $\sin \psi$), with $\theta$ and $\psi$ denoting the outcoupling angles with respect to the normal of the sample surface in the longitudinal plane (i.e., along the waveguide axis) and transverse plane (i.e., perpendicular to the waveguide axis). 2D beam steering is possible with such a basic system^[47–52]. The dispersive grating diffraction guides the $\theta$ direction, whereas the OPA mechanism directs the $\psi$ direction. By tuning the wavelength within the range of 1500 to 1600 nm, the beam angle along the waveguides’ longitudinal direction can be adjusted by a span of 14.1°. The device offers a total FOV of 2.3° horizontally and 14.1° vertically.

The beam-steering angle $\theta$ is governed by the grating equation and can be expressed by $$n_{\mathrm{eff}}-n_c\sin \theta =\frac{\lambda }{\mathrm{\Lambda }},$$(1)where $n_{\mathrm{eff}}$ and $n_c$ are the effective refractive index of the guided mode and the refractive index of the cladding, respectively. $\mathrm{\Lambda }$ is the period of the grating and $\lambda$ is the wavelength of the incident light. Given that $n_{\mathrm{eff}}$ is dependent on both temperature and wavelength, beam guidance along the $\theta$ angle can be achieved through either temperature tuning or wavelength tuning. In this study, wavelength tuning is utilized. However, the GC’s transmission characteristics exhibit significant wavelength dependency, while the achievable beam steering angle remains limited. Effectively increasing the steering angle presents a central challenge. Among the variety of mechanisms for extending the scanning range, polarization multiplexing and the direction switching of light incidence stand out for their simplicity and effectiveness. Therefore, in this review, we present a dual-polarized bidirectional integrated solid-state lidar system capable of achieving a large beam-steering angle. Solid-state lidar systems can employ either a tunable laser source (TLS) or a wavelength division multiplexing (WDM) laser array^[53]. With a TLS, its output can be directed into the SOI waveguide followed by the GCs, offering a straightforward and advantageous setup. By adjusting the TLS wavelength, the beam’s output angle, diffracted from the GC into free space, can be readily controlled. On the other hand, lidar systems using WDM laser arrays as the light sources offer cost-effectiveness and rapid data processing and measurement speed. Similar to current multiline lidars, it can be utilized in several areas such as intelligent driving assistance, terrain mapping, security protection, and other related disciplines. In these systems, output beams from multiple lasers are combined using a WDM multiplexer, and the output are then diffracted using a GC array, where the output angles are determined by their respective wavelengths.

Figure 2 shows an overview of both the traditional single-polarized unidirectional lidar and the dual-polarized bidirectional integrated solid-state lidar. Compared to the traditional lidar setup, the integrated solid-state lidar employing orthogonal polarizations and counterpropagating beams can alter the direction of light incidence and the polarization state to extend the steering range and/or improve angular resolution. The method for altering the polarization mode and the propagation direction may be implemented independently or concurrently, as illustrated in Fig. 3. Both these methods can double the FOV or improve the angular resolution by a factor of 2. By producing multiple diffracted beams with varying propagation directions or polarization states using the same device, and manipulating the positions at which these beams diffract from the device for different wavelengths, it is possible to combine these diffracted beams in various spatial configurations and achieve the following objectives: $${\theta }_p\ge 2{\theta }_T,$$(2)$$\mathrm{\Delta }{\theta }_p\le \mathrm{\Delta }{\theta }_T,$$(3)$$\frac{{\theta }_P}{\mathrm{\Delta }\theta p}\approx 4\frac{{\theta }_T}{\mathrm{\Delta }{\theta }_T},$$(4)where ${\theta }_p$ is the steering angle of the dual-polarized bidirectional integrated lidar, ${\theta }_T$ is the steering angle of traditional lidar, $\mathrm{\Delta }{\theta }_p$ is the angular resolution of the dual-polarized bidirectional integrated lidar, and $\mathrm{\Delta }{\theta }_T$ is the angular resolution of traditional lidar.

This facilitates the selective expansion of the beam-steering angle or the enhancement of angular resolution, as shown in Fig. 2. Through simultaneous utilization of both approaches, the design can achieve both a twofold increase in FOV and a twofold improvement in angular resolution concurrently. As displayed in Fig. 3, solely prioritizing the enhancement of the angular resolution or expansion of the beam-steering range, this strategy can yield a fourfold improvement in each of these aspects.

Steering along the $\psi$ angle (see Fig. 1) relies on the phased array principle. The lateral outcoupling angle $\psi$ of a phased array with spacing $d$ of the GCs is given by $$\sin \psi =\frac{\lambda \mathrm{\Delta }\phi }{2\pi d},$$(5)where $\lambda$ is the operating wavelength of the phased array, $\mathrm{\Delta }\phi$ is the phase difference between the GCs, and $d$ is the distance between the elements in the array. In accordance with the fundamental theory of periodic phased array, the scanning angle $\psi$ is mainly affected by the operating wavelength of the OPA, the phase difference, and the distance between the GCs in the array. To increase this angle, one might decrease the gap between the individual GCs.

The power transmission of the GC in the frequency domain can be computed using the following equation: $$T(f)=\frac{\frac{1}{2}{\int }_{\text{monitor}}\mathrm{Re}[P(f)]·\mathrm{d}s}{\text{source power}(f)},$$(6)where $T(f)$ is the normalized frequency-dependent transmission, $P(f)$ represents the Poynting vector, and $\mathrm{d}s$ represents the surface norm. The outcome of the normalized transmission $T(f)$ is unaffected by the continuous wave’s normalization. The nonlinear effects of silicon greatly limit the power that can be transmitted by silicon waveguides, and therefore there is a limit to the peak power and energy that can be withstood by a solid-state OPA.

Over prolonged use, temperature drift due to high temperatures will affect the pointing accuracy of the scanner. On the one hand, due to the direct relation between the refractive index $n$ and the optical phase $\phi$, and owing to the large thermo-optic coefficient of silicon ($\partial n_{\mathrm{Si}}/\partial T\approx 1.84·{10}^{-4}1/\mathrm{K}$), phased sensitive silicon photonic components are sensitive to temperature. On the other hand, the emission wavelength of semiconductor lasers also depends on the temperature. Since in most cases, laser sources are separated from the photonic integrated circuits^[54], and their temperature is individually stabilized, Temperature drift has little effect on the scanning accuracy. Nevertheless, it must be noted that in systems such as^[31,55], in which laser sources are integrated within the photonic integrated circuits, the influence of temperature on the emission wavelength must be also considered. In order to determine the influence of temperature offsets on the beam-steering angles, Yepez et al. heated up the OPA system. It is expected that heat generated by the silicon heater spreads through the metallic stage and the printed circuit board, leading to homogeneous temperature offsets on the OPA. To obtain the relation between the angular offset and the temperature offset, the temperature and the measured angles are averaged over the time period where the temperature has stabilized. The experimental results show that $\mathrm{\Delta }\theta$ is about 0.25° and $\mathrm{\Delta }\psi$ is approximately 0.08° at a temperature drift of 20°^[56].

This section carefully summarizes the technical solutions for improving the beam quality of OPA in angular resolution and beam steering. In Sec. 3.1, we review the solutions to improve the quality of beam scanning and angular resolution by manipulating the polarization state of the input light. In Sec. 3.2, the solutions with single polarized beams employing counterdirections in recent years are reviewed. In Sec. 3.3, we introduce the integrated silicon-based OPA lidar employing orthogonal polarizations and counterdirections simultaneously, resulting in superior performance.

According to Sec. 2, the longitudinal scanning angle $\theta$ of the OPA is affected by the structure of the GCs and the wavelength of input light. The angular resolution $\mathrm{\Delta }\theta$ was determined by the effective emission length of the GCs. The steering angle $\theta$ for GCs with a fixed structure exhibited a nearly linear relationship with the working wavelength. Consequently, the longitudinal scanning range was restricted by the limited working bandwidth of the light source. A tunable laser with a 100-nm tuning range can generally achieve a beam steering angle of $\sim 14°$. There was also longitudinal beam steering by thermo-optical tuning of the grating emitter area, with a smaller steering angle of $\sim 10°$^[57]. The dual polarization OPA can double the FOV or improve the angular resolution by a factor of 2 when only single-polarized light is used, as depicted in Fig. 2.

The working principle of dual-polarization lidar is as follows. The SOI waveguide was highly birefringent. The effective refractive index of the ${\mathrm{TE}}_0$ mode was significantly different from the effective refractive index of the ${\mathrm{TM}}_0$ mode. When an optical signal was diffracted from a GC, the angle of the ${\mathrm{TE}}_0$ beam would be different from that of the ${\mathrm{TM}}_0$ beam. By carefully designing GC, one can seamlessly stitch the ${\mathrm{TE}}_0$ beam and ${\mathrm{TM}}_0$ beam together, which leads to a doubling of the FOV.

In recent research by Yan et al., an OPA expanding the scanning range by polarization multiplexing was designed in 2021^[58]. Figure 4 displays the diagram of the proposed polarization-multiplexed OPA, consisting of two main components: the polarization-decision part and the OPA part. The polarization-decision section consisted of a Mach–Zehnder interferometer (MZI) switch and a polarization splitter-rotator (PSR). This part was designed to convert the input light between the ${\mathrm{TE}}_0$- and ${\mathrm{TM}}_0$-polarized modes. The OPA part consisted of a $1\times N$ power splitters array, a phase shifters array, and the GC emitters. The grating was made up of a 340-nm thick silicon waveguide and was etched to a depth of 70 nm. By appropriately determining the cross-sectional dimensions of the waveguide, the disparity in effective refractive indices between the ${\mathrm{TE}}_0$ and ${\mathrm{TM}}_0$ polarized modes was decreased. This resulted in a continuous scanning range of 28.2° over the wavelength range of 1500–1600 nm. The calculated far-field distribution and beam-steering range of the GC emitter between two fundamental modes are shown in Fig. 5. Initially, the OPA was equipped with an MZI switch for polarization selection. Next, thermo-optical phase shifters were used to regulate the phase for lateral beam steering $\psi$, and a tunable laser was used to control the wavelength for longitudinal beam steering $\theta$. By employing this configuration, the scanning angle along the longitudinal axis was nearly double that of a device operating in a single polarization state. The calculation results indicated that it could be accomplished to reach an expanded steering angle of $77.4°\times 28.2°$.

To achieve large lateral and longitudinal FOV, broadband tunable optical source and nonuniform spacing waveguides are widely used, leading to significant expense and a complicated structure. In 2022, a polarization-multiplexed OPA with a large FOV based on a 220-nm SOI platform was demonstrated, as shown in Fig. 6(a)^[59]. The steering range can be greatly increased by utilizing customized optical antennas that are assisted by polarization switching. The enhanced optical antenna operated on the principle of radiation following interference, with the beam emanating from an individual grating. The device consisted of five primary components: an edge coupler, a power splitter, polarization converters, phase shifters, and the improved optical antenna. Two stages of the multimode interference (MMI) coupler were designed to divide the fundamental TE mode into four coherent channels. To achieve polarization switching of the input light, a polarization converter was implemented in each channel. This was accomplished by combining an MZI with a bilevel mode converter. By adding heat to the upper branch of each MZI, a $\pi$ phase shift can be induced, resulting in the conversion of the ${\mathrm{TE}}_0$ mode to the fundamental TM mode. Four more independent TiN microheaters were utilized to control the phase difference of the channel, allowing for adjustment of the lateral steering angle, and the beam was going to send to the antenna following phase control. Unlike classic antennas based on grating arrays, this design introduced an interference region (IR) to divide the waveguide array and the single grating, as shown in Fig. 6(b). To explain the operation of the antenna, the interference pattern was generated by feeding the ${\mathrm{TE}}_0$ mode or the ${\mathrm{TM}}_0$ mode, which possesses a particular phase, into the IR region. Subsequently, the entire pattern was transmitted to the grating and ultimately emitted into free space.

By employing this design, the cross talk between adjacent channels was reduced. As a result, the spacing of the waveguide array could potentially be reduced, allowing for a wider lateral beam steering range. On the other hand, in the case of longitudinal steering, polarization switching was used to reduce the reliance on the tunable laser source. This was achieved by taking use of the varying effective refractive indices and resulting diffraction angles caused by distinct polarization states. In comparison to the single polarization instance, the longitudinal steering range had the potential to be increased twofold. The suggested OPA offered an alternate method for achieving a broader 2D FOV.

Figure 6(c) illustrates that the longitudinal steering ranges of the ${\mathrm{TE}}_0$ mode and the ${\mathrm{TM}}_0$ mode are 15.62° (ranging from 22.71° to 38.33°) and 16.08° (ranging from $-16.04°$ to $-32.12°$), respectively. These ranges were achieved by adjusting the wavelength of the tunable laser from 1510 to 1630 nm. The polarization switching OPA offered double FOV and significantly increases the longitudinal steering range without compromising the lateral steering angle, when compared to standard methods. As a proof-of-concept demonstration, the device was equipped with a mere four channels; increasing the number of channels would likely result in a more precise light point. Furthermore, they conducted measurements on the side-mode suppression ratio (SMSR). The SMSR values for the ${\mathrm{TE}}_0$ and ${\mathrm{TM}}_0$ modes at 1550 nm were 12.08 and 13.54 dB, respectively. The primary beam powers of the two modes were $-8.4$ and $-8.8\mathrm{dBm}$, respectively, when the optical power pumped into the device is 0 dBm. The loss encompassed the approximately 7-dB antenna transmission loss. The enhanced antenna minimized the interference between channels that commonly occurred in conventional 2D steering OPA systems utilizing GCs and achieved a wide lateral steering range in 2D beam steering. For the ${\mathrm{TE}}_0$ and ${\mathrm{TM}}_0$ modes, the observed steering ranges were $99.24°\times 15.62°$ and $96.48°\times 16.08°$, respectively. The proposed scheme provided a promising approach to realize a larger FOV.

In conventional single polarization OPAs, the efficiency of wavelength tuning was normally around 0.14°/nm. Enhancing the efficiency of tuning the wavelength was a viable approach to fulfill the practical requirement of lidar. The overall system of the multiline OPA enhanced the wavelength-tuning efficiency. However, this resulted in a corresponding increase in complexity and footprint. Without sacrificing complexity, polarization-multiplexing OPA using multiple polarized states was a viable approach^[60]. In 2023, a novel and universal polarization-multiplexing OPA scheme in which ${\mathrm{TE}}_0$ and ${\mathrm{TM}}_0$ modes were transmitted in two staggered arrays^[61]. The blind area could be eliminated by releasing the polarization dispersion limitation on the grating antennas. Polarization multiplexing was achieved for the splitter tree and phase shifter array using a polarization switch. Following coupling to the OPA, the light was converted to the ${\mathrm{TE}}_0$ mode or the ${\mathrm{TM}}_0$ mode via the optical switch. The PSR array divided two polarized beams into superlattice antennas composed of a pair of grating antennas placed alternately. These antennas were specifically designed to eliminate the blind spot in the vertical direction and enhance the efficiency of wavelength tuning. Both sets of grating antennas applied the identical power splitter tree and phase shifter array. An array was formed by combining two waveguide gratings of varying widths in a superlattice structure. Light traveled through power splitters after entering the ${\mathrm{TE}}_0$ or the ${\mathrm{TM}}_0$ mode. It then traveled through the PSR and into the corresponding array, obtaining a 28° scanning range. Polarization multiplexing had significant promise in the field of OPA due to the enormous and mature design of PSR^[52,53]. The addition of a single-voltage channel doubled the wavelength tuning efficiency when compared to the conventional single-polarization OPA. The wavelength tuning efficiency, as tested, was 0.31°/nm, while the FOV was $24.8°\times 60°$.

Figure 7(a) displays the diagram of the proposed polarized multiplexing OPA. It comprised a polarization switch, a power splitter tree, a phase shifter array, a PSR array, and superlattice grating antennas. This work considered an SOI platform with three layers: a 1.8-µm cladding oxide layer, a 2-µm buried oxide layer, and a 220-nm silicon core layer. The ${\mathrm{TE}}_0$ mode light was coupled into the waveguide by a TE-polarized GC. The MZI-based optical switch and the reversed PSR made up the polarization switch. The two-port outputs of the optical switch were linked to the two outputs of the PSR. The superlattice grating antenna array allowed the propagation of light with two different polarized states through a PSR. By improving the characteristics of the waveguide grating, it was possible to achieve a twofold increase in the efficiency of wavelength tuning.

In this work, the technique has been demonstrated on the widely used 220-nm SOI substrate. The experimental results demonstrated that a vertical scanning range of 24.8° could be achieved with a high efficiency of wavelength tuning, approximately 0.31°/nm. The observed FOV was $24.8°\times 60°$. With Wgrt2 ranging between 0.36 and 0.46 µm, the calculated vertical beam steering range of Grt2 was represented by the green area in Fig. 7(b). An increase in Wgrt2 resulted in a progressive upward movement of the beam steering range. The red area depicts the vertical beam steering range of Grt1, based on the parameters mentioned above. This study presented the first verification by experiment of polarization-multiplexed OPA with a high efficiency in tuning the wavelength and without any blind area in the FOV. This OPA technique demonstrated adaptability among several platforms, indicating a promising potential for implementation in all-solid-state and tiny lidar systems.

The system employs two single-polarized beams that travel in opposite directions, and the resulting diffracted output beams can be smoothly merged to increase the beam steering angle or angular resolution by 2 times.

More recently, Ito et al. utilized a particularly shallow-etched diffraction grating cut directly into the photonic crystal surface to preferentially couple light in the upward direction and achieve the wide range 2D beam steering^[62]. The shallow-etched grating on the silicon layer decreased the loss of emitted light in the downward direction. Additionally, it reduced the loss caused by internal reflection and collimation by limiting the angular spread, denoted as $\delta \phi$. The prism lens eliminated the $\theta$ dependency of the collimation requirement and transforms angle $\theta$ into ${\theta }^{\prime }$, which encompasses 0°. To increase the steering range, they implemented an adjustment in the direction of light incidence on the lattice-shifted photonic crystal waveguide (LSPCW).

Figure 8(a) shows an overview of a single beam-steering device. An SOI comprising an upper cladding layer, a 2-µm silicon dioxide (${\mathrm{SiO}}_2$) BOX layer, and a 210-nm thick silicon layer was assumed. Between arrays of triangular lattice holes, a single-line defect waveguide was formed in the silicon layer and covered by ${\mathrm{SiO}}_2$ in the LSPCW. The lattice constant, hole diameter, and third-row lattice shift were set at S = 394, 192, and 95 nm, respectively, for operation at wavelengths $\lambda \approx 1.55\mathrm{\mu m}$. The photonic band construction can control the value of the group index $n_g$ for the slow-light mode. $n_g$ had an average value of 15–25, which was 4 to 6 times that of silicon wire waveguides. Within the range of $\mathrm{\Delta }\lambda =15–20\mathrm{nm}$, the lattice shift $s_3$ efficiently flattened the $n_b$ spectrum. Typically, the formula for $n_g$, which accounted for the vacuum light velocity $c$, the wavenumber $k$, and the frequency $\omega$, was $(1/c)(\mathrm{d}k/\mathrm{d}\omega )=-({\lambda }^2/2\pi c)(\mathrm{d}k/\mathrm{d}\lambda )$. Hence, a substantial $n_g$ was equal to a significant first-order dispersion in the LSPCW as well as a significant angular dispersion when the mode was transformed into a free-space beam.

The 2D beam steering procedure is demonstrated in Fig. 8(b). One LSPCW was launched by light coming from the array. The beam was guided in the $\phi$ direction after passing through the lens because of the change in relative position against the collimator lens. For this collimation, the authors developed two surfaces of a prism. The tilt angles of the two surfaces with respect to the LSPCW plane were calculated in a manner that converted the minimal value of $\theta$ to ${\theta }^{\prime }=0°$ while ensuring that the median value of $\theta$ within the steering range met the minimum deviation criteria. The elimination of the $\theta$ dependence for the collimation requirement can be achieved by determining the prism thickness and lens curvature of a given structure. The effective steering range, denoted as $\theta =10°–30°$, was assumed to be unaffected by the band edge. This range was subsequently translated to ${\theta }^{\prime }=0°–20°$. Considering $\delta \varphi$ and the length of the emission aperture, the author’s employed ray tracing to build a prism lens with aspherical lens shapes. The values of $\delta {\theta }^{\prime }$ and $\delta \varphi$ were found to be below 0.1° after traversing the lens for nearly all values of $\theta$ and $|\varphi |\le 6$ except for $\delta \theta =0.12°$, which was seen only at ${\theta }^{\prime }\approx 0°$ and $|\varphi |\approx 6°$.

Figure 8(b) further demonstrates how the direction of light incidence can be switched to change ${\theta }^{\prime }$ from positive to negative. To accomplish continuous collimation and guiding within the double-width range, the prism lens was intentionally constructed to possess a symmetrical shape. The silicon shallow-etched grating in this study enhanced the efficiency of emitting upward light by a factor of 2 to 8. Additionally, the prism lens ensured that the collimation condition was preserved for the desired steering range and adjusted the beam angle to double the continual steering range by changing the direction of light incidence. The number of resolution points $N=266\times 16=4256$ and the 2D beam-steering range of $\mathrm{\Delta }\theta =40°(\mathrm{continuous})\times \mathrm{\Delta }\phi =4.4°(\mathrm{discrete})$ were calculated. By implementing techniques that lengthen the aperture length, increase the number of LSPCWs, and eliminate manufacturing faults in the prism lens, this device exhibited the potential to attain a $N$ value of 345,600. Additionally, they have showcased the capability of adjustable 2D steering, a task that proves challenging when employing traditional mechanical systems. This finding suggested that the potential applications of this device extend beyond lidar and encompass security systems, free-space communications, and other related domains.

The far-field patterns of OPAs were determined by multiplying the far field of each individual antenna and an array factor. To attain a broad steering angle of OPAs, two distinct attempts should be accomplished. Initially, it was necessary to suppress grating lobes using aperiodic arrays^[35]. Furthermore, it was recommended to enhance the steering angle by augmenting the FOV of each antenna. Nevertheless, this task presented a significant challenge due to the difficulty in achieving a wide FOV of around 180°, as previously stated. Furthermore, the elimination of downward radiation was a challenge in achieving high emission efficiencies. As a result, achieving both goals simultaneously had proven to be extremely challenging.

To build advanced solid-state integrated lidar, a wide-angle waveguide grating antenna was proposed and demonstrated by Guo et al.^[63]. Instead of focusing on reducing the downward radiation to increase efficiency, the authors’ approach was to utilize downward radiation and achieve equalization of radiated powers and FOV in both directions. This strategy aimed to double the range of beam steering. Both upward and downward emission exhibited FOV values over 90° at the full width at half-maximum (FWHM) of radiation intensity in the polar direction. Consequently, the proposed bidirectional antenna had a total FOV exceeding 180°, indicating its potential for wide-angle OPAs. In conjunction with expanded FOVs, the directional guided beams were generated by a shared array of power splitters, phase shifters, and antennas. This configuration significantly reduced the chip complexity and power consumption, especially in the context of large OPAs.

Figure 9(a) displays an array of bidirectional wide-angle WGAs. Incident light $S_0$ with a wavelength of 1.55 µm was directed as the two inputs to the proposed WGA, $S_1$ and $S_2$. To accomplish this, it passed through 3-dB splitters and was subsequently turned by strongly bent waveguides^[36] or Bragg mirrors. The bidirectional wide-angle WGA is depicted in Figs. 9(b) and 9(c). Adjacent to the silicon substrate, there were two layers: a layer of silicon nitride (${\mathrm{Si}}_3{\mathrm{N}}_4$) and a layer of ${\mathrm{SiO}}_2$. The thickness of the ${\mathrm{Si}}_3{\mathrm{N}}_4$ layer was half the wavelength, while the ${\mathrm{SiO}}_2$ layer was a film with a quarter-wavelength thickness. The ${\mathrm{Si}}_3{\mathrm{N}}_4$ and ${\mathrm{SiO}}_2$ layers served as a hybrid antireflection coating (AC) to enhance the downward transmissivity.

Figure 10 illustrates the relationship between normalized far-field radiation intensity and radiation angle in 2D design. The data revealed a consistently high value ($>0.9$) and minimal variance ($<10\%$) throughout the range of $-39°$ to 39° for upward radiation and $-42°$ to 42° for downward radiation. The bidirectional WGA can accommodate a total beam-steering angle of 180°.

Later, in 2023, the on-chip lidar architecture was found to be much simplified by employing counterpropagating beams with the same polarization in this proposed structure that was intended to double the beam tuning angle^[64]. The configuration of the on-chip lidar system is shown in Fig. 11(a). On a single substrate, a GC and an optical switch were integrated. Linearly polarized signals were often generated by light sources and coupled into the SOI waveguide with the elementary TE mode. The TE mode was selected due to its higher effective refractive index and lower bending loss. A TLS was selected to be the light source in Fig. 11(a). The identical operational principle is applied when utilizing a WDM laser array as the light source, as depicted in Fig. 11(b). The optical switch can completely manipulate the beam and send it towards one of the two opposite directions of propagation^[24]. By utilizing a counterpropagating TE polarized beam, it was possible to create two output beams from a single GC. The output beams from the GC will propagate in opposite directions. The two beams can be smoothly combined to double the beam steering angle.

To reduce expenses, one can implement a strategy of utilizing a single TLS for several integrated lidars. By employing an MMI, TLS had the capability to divide its light output into many light paths. Using a discrete-wavelength light source, each GC also can generate counterpropagating beams. Utilizing this strategy can substantially decrease the expenses in comparison to employing TLS for a lidar system.

A 3D finite-difference time-domain (FDTD) approach was employed to obtain the simulation results. The simulation results and theoretical results exhibited strong concordance, as depicted in Fig. 12. The resulting angles showed that an even distribution of eight beams is attained. The transmittance values of eight TE output beams were plotted against the right $Y$ axis in Fig. 12. The transmittance values observed were between 0.42 and 0.35. Utilizing a WDM laser array in the O-band can effectively minimize the fluctuations in transmittance. By optimizing the thickness of both the ${\mathrm{SiO}}_2$ upper cladding layer and the ${\mathrm{SiO}}_2$ lower cladding layer, the radiation intensity can be reduced. According to the study provided, this design can utilize a CWDM laser array with four distinct wavelengths to reach a greater scale, lower cost, and higher efficiency.

Orthogonal polarization and counterpropagating beams can be used simultaneously and attain a double increase in FOV and an approximately twofold enhancement in angular resolution. This method can produce 4 times the improvement in either of the two main aspects—angular resolution and beam-steering range—if one only concentrates on these areas.

In 2021, Han et al. proposed a dual-polarized bidirectional integrated solid-state lidar utilizing the polarization and to essentially double the beam-steering angle^[65]. In the proposed mechanism, two orthogonally polarized modes, the TE mode and TM mode, existed in the SOI waveguide. Two sets of GCs were designed, one for each polarized mode. The beam-steering angle can be doubled by combining the outputs from these two GCs into a continuous beam by carefully selecting the required parameters. Figure 13 depicts the suggested configuration of orthogonally polarized grating couplers (OPGCs), where the TE and TM modes are propagated in opposite directions to each other. There was no crossing between waveguides carrying different polarized modes. When it came to wavelength tuning, the TE mode and TM mode diffracted at the same angle at one end of the range (${\lambda }_1$) and at two different angles at the other end (${\lambda }_2$). This design enabled the creation of a continuous beam by combining two beams with perpendicular polarizations, effectively increasing the beam-steering angle twofold.

Figure 14 displays the angle $\theta$ at which the beams diffract for both the TE mode and TM mode. The wavelength range of TLS spanned 1.2 to 1.3 µm. The $\theta$ value was derived by locating the maximum position of the electrical field in the far field. In general, the theory accurately predicted the output angle under various wavelengths. There was an excellent agreement in the TE mode between the theory and the simulation. The slight discrepancy in the TM mode can be ascribed to the fact that the effective mode theory (EMT) was less suitable when the period of the GC was near the beam’s wavelength.

When focusing on the solid-state lidar constructed with a CWDM4 laser array and photodiode array, large channel spacing in CWDM enabled an easier layout of the WDM multiplexer/demultiplexer. A bigger volume of CWDM optical transceivers would result in a reduced cost. The 80-nm wavelength range provided a substantial FOV of around 12°, which was sufficient for most applications. Nevertheless, the CWDM4 scheme’s limited four-channel configuration resulted in an angular resolution of around 4, which fell below the desired standard for several lidar applications. The widespread use of this suggested solid-state lidar may be hampered by this restriction.

An approach involved augmenting the quantity of channels from four to eight while diminishing the channel separation from 20 to 10 nm. Nevertheless, implementing this channel plan would result in higher expenses, and there was currently no industrial progress in implementing it for the short-reach optical transceiver. As shown in Fig. 15, a fresh strategy called interleaved orthogonal polarization GC (iOPGC) was presented to solve this issue^[66].

There were two output GCs where the TE mode was propagating in the first GC, and the TM mode was propagating in the second GC. The TE mode and the TM mode were propagating in opposite directions. The outputs of two GCs can be alternately combined. This resulted in a single TE beam positioned centrally between two TM beams and vice versa. Consequently, this unique arrangement can enhance the angular resolution by a twofold factor. Figure 15(a) illustrates the architectural layout of iOPGC, with the TE mode and TM mode propagating in opposite directions within SOI waveguides.

Figure 16(a) displays the results of the numerical simulation and theoretical analysis conducted using 3D-FDTD. The TE GC had a period of 450 nm and a filling factor of 0.35, whereas the TM GC had a period of 740 nm and a filling factor of 0.6. The $\theta$ value was determined by identifying the position of the biggest electrical field emitted from the GCs. It was evident that the simulation aligns closely with the theoretical predictions for the TE mode. However, there was a discrepancy between the theory and simulation for the TM mode, because EMT is more suitable for TE modes than for TM modes. Figure 16(b) displays a linear regression analysis of output angles for eight channels. The $R^2$ was 0.993, indicating an extremely high level of accuracy in the fit. It was evident that two channels diverged from the desired output angle indicated by the linear curve.

Following this initial work, Zhao et al. improved the design and proposed a polarization multiplexed OPA with a bidirectional shared grating emitter array, which can further improve the wavelength tuning efficiency and longitudinal steering angle^[67]. Two identical OPAs were positioned on either side of the emitter array. Two-stage MZIs were employed to determine the directions and polarization states of the light that was introduced into the emitter array. By adjusting the width of the waveguide and the period of the grating, it was possible to enhance the performance of dual polarization multiplexed and dual-direction OPAs to cover blind spots within the scanning range. The design allowed for a total longitudinal scanning angle of 54.5°, with the wavelength scanning range spanning 1500 to 1600 nm. The lateral steering range was approximately 77.8° achieved through phase tuning with a waveguide spacing of 1.2 µm.

Figure 17(a) displays the schematic diagram of the proposed dual polarization multiplexed OPA. It consisted of two MZIs, a PSR, and two OPAs. By combining the four states, the proposed bidirectional and dual-polarization OPA can accomplish beam steering in the longitudinal dimension with a 54.5° FOV, as shown in Fig. 17(b).

Recently, an improved structure using a CWDM laser array as the light source for solid-state photonics-based lidar was designed to simultaneously increase the viewing angle and enhance the angular resolution^[68]. The CWDM laser source working in the O-band has been suitable for the short-reach optical communication modules, leading to a significant cost advantage. When a CWDM4 laser array was utilized as the signal source, a single GC diffracts 16 output beams, 4 of which were produced at each wavelength. There was a notable difference in the effective index of the TE mode and the TM mode, causing the output angles of the TE beams to be distinct from those of the TM beams. By employing optimal design techniques, it was possible to position a TE beam precisely between two TM beams, and vice versa. This resulted in a twofold improvement in the angular resolution. In the GC, the signal can also travel in both directions, forward and backward. The output beams of the forward and backward signals were mirror-symmetric. By optimizing the design of a single GC’s design, 16 diffracted beams were uniformly distributed, with 4 beams produced from each wavelength. Sixteen beams can be easily integrated, resulting in a total viewing angle that was more than doubled. By employing this method, it was feasible to attain a combination of superior angular accuracy and expansive viewing perspectives.

Figure 18 displays the schematic image of the proposed configuration, illustrating the simultaneous transmission of the forward-propagating beam and the backward-propagating beam within the SOI waveguides. In the same substrate, the authors combined an optical switch, a shallow-etched GC, and two polarization elements (PEs). Four lasers were configured in the TE mode and then merged using a WDM multiplexer before being coupled into the SOI waveguide. Subsequently, an optical switch was employed to ascertain the direction of light transmission, enabling the output signal to be selectively routed to one of the two channels as required^[32]. Figure 18 illustrates that each path is fitted with a PE, comprising an optical switch enclosed within the PE, a polarization rotator, and a polarization beam combiner. The layered optical switch guided the signal to the bottom path while TE polarization was required. When the TM polarization was needed, the nested optical switch routed the signal to the upper path. The TE polarization is converted to the TM polarization by the polarization rotator and TM polarizations. By employing this approach, it was possible to manipulate both the direction of propagation and the mode of polarization for the CWDM4 laser signal. Figure 19(a) displays the numerical simulation results of the output angles obtained by 3D-FDTD. At a wavelength of $\lambda /4$, the TM mode produced the largest output angles, which were roughly 16.9°. Two adjacent beams were spaced about 2° apart. When light entered the GC from different directions, the output angle was evenly distributed in a symmetrical manner along the vertical axis. By merging all the emission beams, a FOV of approximately 33.8° can be achieved, as shown in Fig. 19(b). This novel arrangement optimized the utilization of the diffracted beam in spatial dimensions and effectively mitigated a significant decrease in external power.

With the advantages of small size, high precision, and fast scanning speed, the OPA-based solid-state lidar is expected to become an important configuration in the spatial information sensing system in the future and has broad application prospects in the unmanned domain and other fields.

At present, limited by the high requirements of the processing technology, the OPA cannot yet meet the application requirements of the actual engineering, and the application of OPA to the all-solid-state lidar system requires not only the research on the high-performance OPA technology but also in-depth study on the entire lidar system. Nevertheless, with the development of the integration process, it has already become possible not only to develop a low-cost lidar detection system based on OPAs. At the same time, it also shows broad application prospects in terms of excellent performance such as high-precision target capture and large-range flexible scanning.

For the whole lidar system, at present, in addition to the scanning system, some of the remaining components such as the laser source, modulation module, amplifier, and photodetector, have also been chip-based results. However, the discrete chip devices still need to be coupled to each other through optical fiber or free space, and problems such as power consumption, volume, and stability still exist. However, with the development of a heterogeneous integration platform by researchers, it is foreseeable that the future silicon-based chip will be able to integrate the various active and passive components required in the lidar system so that lidar can be fully realized on a chip, thus improving the stability of the system, simplifying the manufacturing and installation process, greatly reducing its size and production costs, and ultimately enhancing the competitiveness of lidar in the field of unmanned vehicles and other areas.

In terms of product development, thanks to the development of silicon-based photonics technology for many years, AMF from Singapore, SilTerra from Malaysia, Shanghai Industrial Technology Research Institute and Chongqing CUMEC from China, and other companies can be fully capable of OPA flow of custom services. With the design of lidar based on OPA becoming mature and the system gradually improving, Samsung Electronics from Republic of Korea, Analog Photonics from the United States, and other companies have launched a full-solid-state lidar based on the OPA chip prototype, and gradually promote the product development process.

In summary, the CMOS-compatible integrated OPA is a highly promising approach for advancing all-solid-state and compact lidar technology. Crucial elements of the OPA, including laser, beam splitters, phase modulator arrays, GC antennas, and photodetectors, can all be consolidated onto a single chip. Recent years have witnessed significant improvements in the development of compact all-solid-state lidar systems, featuring impressive performance metrics like a wide beam steering range and high spatial resolution. Utilizing orthogonal polarizations and counterpropagating beams, integrated solid-state lidar systems can further expand the steering range and enhance the angular resolution. The ongoing progress in silicon monolithic integration and superior laser/detector technology will facilitate the evolution of OPA lidar systems. It is foreseeable that fully developed lidar technology will find practical applications in the realms of autonomous driving, robotics, artificial intelligence, and optical communication.