Artificial intelligence (AI) is one of the most active research fields at present. The AI models, represented by artificial neural networks, are computational models that mimic the neural synaptic networks in the brain and have been widely used in areas such as computer vision, speech recognition, and automatic driving. In the last decade, the AI technologies have experienced an explosive growth and the global computational volume has increased dramatically. The urgent need to process massive data in a fast and efficient way has placed an urgent demand on the computing hardware in terms of computing capacity and energy efficiency. Restricted by the inherent limits of electronic devices and the von Neumann architecture, traditional electronic computing has encountered a bottleneck in terms of speed and energy efficiency, which is difficult to break through. Optoelectronic intelligent computing uses photons instead of electrons to perform computation, hence optoelectronic intelligent computing can significantly improve computing speed and energy efficiency by overcoming the inherent limits of electrons. Compared with electronic computing, optoelectronic intelligent computing fully combines the unique advantages of multi-dimensional multiplexing, large bandwidth, and low energy consumption of optics and the fine-grained and flexible control of electronics, which is a more practical and competitive solution for accelerating AI computing.

Optoelectronic intelligent computing is particularly suitable for implementing large-scale neural networks containing a large number of neurons and synapses. Restricted by interconnection density and Joule heat, the processing speed of current neuromorphic electronic chips is basically limited to the MHz range, and the energy consumption per multiply accumulate (MAC) operation requires several picojoules. However, the neuromorphic computing hardware built from basic photonic devices, such as Mach-Zehnder interferometer (MZI) mesh and micro-ring resonator (MRR) array, requires only tens of femtojoules per MAC operation. This metric is two orders of magnitude smaller than that of the state-of-the-art complementary metal oxide semiconductor (CMOS) computing hardware, indicating that optical neural networks are far superior to electronic neural networks in terms of energy efficiency while achieving ultra-high-speed computing. As a result, optoelectronic intelligent computing naturally has significant advantages in the application scenarios such as automatic driving and drones which require large bandwidth and high real-time performance, as well as in the highly concurrent, high-throughput, computationally intensive supercomputing platforms and data centers. In fact, the optical interconnect technique has already been widely used in data centers and significantly reduces its time and energy consumption for large-scale interconnects.

Generally speaking, AI algorithms can be divided into two parts: training and inference. Since photons are difficult to be stored and the state of photons cannot be directly obtained, most of the current optoelectronic intelligent computing systems use the method of "training in the electronic domain and inference in the optical domain" to implement the AI algorithms. In other words, the simulation model of the neural network is first trained on the electronic computer, and then the parameters of the trained model are loaded onto the photonic chip for inference. However, this offline training method obviously has the difference between the the numerical simulation model and the actual physical model, and more importantly, if the neural network implementation of the photonic chip always needs to rely on the training in the electronic domain, then the performance advantages of the photonic chip over the microelectronic chip including low latency and high energy efficiency cannot be fully exploited. Therefore, developing hardware-friendly online training algorithms for optoelectronic intelligent computing is a key challenge.

Here, a comprehensive review of the research progress and challenges in optoelectronic intelligent computing is presented. There are three mainstream types of optical matrix-vector multiplication (MVM), which are plane light conversion (PLC)-based matrix computing, MZI-based matrix computing, and wavelength division multiplexing (WDM)-based matrix computing (Fig. 1). The applications of optoelectronic intelligent computing mainly consist of optical signal processing and optical neural networks (Fig. 2). Hardware-friendly online training algorithms for optoelectronic intelligent computing mainly include online training of optical neural networks through the back propagation (BP) algorithm (Fig. 3) and online training of optoelectronic intelligent computing chips through the stochastic gradient descent (SGD) algorithm (Fig. 4). Computing capacity and energy efficiency are important metrics to evaluate the performance of optoelectronic intelligent computing. Table 1 shows the comparison of computing capacity and energy efficiency of various microelectronic chips and optoelectronic chips. As for three typical optoelectronic intelligent computing architectures (Fig. 5), their computing capacity and energy efficiency are summarized (Table 2). Finally, according to the aforementioned evaluation methods, the ways to further improve computing capacity and reduce energy consumption can be explored in terms of improving parallelism, baud rate, operation scale, and optical-electrical/electrical-optical conversion efficiency and reducing energy consumption in the optical and electrical layers (Fig. 6).

Optoelectronic intelligent computing fully combines the advantages of multi-dimensional multiplexing, large bandwidth, and low energy consumption of optics and the fine-grained and flexible control of electronics. It is a category of the dedicated computing hardware architectures for AI computing, which breaks the bottleneck of traditional von Neumann architectures. Here, the research progress of optoelectronic intelligent computing is reviewed, the challenges in online training algorithms as well as computing capacity and energy efficiency improvement of the current mainstream computing architectures for optical signal processing and optical neural networks are discussed, and the perspectives are presented. Advanced optoelectronic integration and fabrication techniques enable the production of large-scale and low-cost optoelectronic computing chips. Through optoelectronic monolithic integration, electronic and photonic devices can be integrated on the same substrate, which eliminates on-chip and off-chip optoelectronic interconnections and builds on-chip optoelectronic hybrid computing architectures. In addition, the performance of optoelectronic intelligent computing can be further improved by in-depth combination between novel materials with excellent performance and customized design with hardware-software synergy.

Ultrasound, an acoustic wave with a frequency exceeding 20 kHz, is widely utilized in nondestructive testing, medical imaging, medical treatment, and other sectors owing to its considerable penetration depth, excellent resolution, good directivity, and minimal radiation. As the key component of ultrasound equipment, the ultrasound transducer is used for bidirectional conversion between ultrasound waves and electrical or visual signals. When the ultrasound transducer is used as a transmitter, the transducer converts the input excitation signal into the mechanical structure vibrations of the transducer element, which then releases ultrasound waves into the medium. Conversely, when the transducer is used as a sensor, the incident ultrasonic wave causes the structure deformation of the transducer element, causing the corresponding change in the electric, magnetic, or optical field inside the transducer.

Currently, piezoelectric ultrasound transducers represent the state of the art but have many drawbacks, such as enormous size, narrow bandwidth, and electromagnetic sensitivity. An optical fiber transducer offers an alternative to challenge the present piezoelectric hegemony. A fiber-optic ultrasound transducer, a newly created transducer, is used to convert optical or ultrasound signals to one another. Compared with standard piezoelectric ultrasound transducers, fiber-optic ultrasound transducers have various advantages. For example, most fiber-optic ultrasound transducers are the same size as the normal single-mode fiber (125 μm), making them suitable for minimally invasive detection. Furthermore, ultrasonic detection devices based on fiber-optic ultrasound transducers usually have remarkable resolutions owing to broad bandwidth properties. Easy multiplexing is another benefit of fiber-optic ultrasound transducers, which fits the need for rapid ultrasound imaging without mechanical scanning. To date, ultrasound transducers based on optical fiber have been widely used in photoacoustic/ultrasound imaging, nondestructive testing, structure safety monitoring, and other applications. Although these new technologies or applications exhibit numerous benefits and great development potential, several obstacles and challenges still exist. Therefore, the current development status of fiber-optic ultrasound transducers must be examined and described, providing a development path for further optimizing their performance and expanding their applications.

A fiber-optic ultrasound transducer comprises two parts: a transmitter and a sensor. The research development of two types of fiber-optic ultrasound transducers is summarized below, and the application of fiber-optic ultrasound transducer is reviewed.

As a type of ultrasound transducer, an ultrasound transmitter is used to convert a visual signal into an ultrasound signal. Several optical-fiber-based ultrasonic transducers have been proposed, exhibiting apparent performance differences, such as emission pressure, emission bandwidth, and conversion efficiency, owing to their diverse constructions and materials (Table 1). The materials with a high thermal expansion coefficient, low specific heat capacity, and high light absorption coefficient bring higher excitation efficiency. Currently, ultrasound transducer materials have progressed from single-component materials and are heading for the era of composite materials with improved performance. Several composite materials have been demonstrated, including polydimethylsiloxane (PDMS) + gold nanoparticles, PDMS + carbon black, and PDMS + carbon nanotubes. Among them, the fiber-optic ultrasonic transmitter based on PDMS + gold nanoparticles exhibits the highest excitation efficiency up to 0.073 MPa·mJ^-1·cm² and emission pressure of 0.64 MPa with a measured distance of 1 mm (Fig. 3).

A fiber-optic ultrasound sensor, as another type of ultrasound transducer, is also systematically explained (Table 2). To meet the requirements of high-precision ultrasound imaging, fiber-optic ultrasound sensors with high sensitivity, broad bandwidth, and small size are required. To achieve these goals, various fiber-optic ultrasound sensors based on diverse sensing concepts have been proposed, including the types of intensity, phase, and wavelength modulations. Among them, the phase-modulated Fabry-Perot interferometer (FPI) ultrasound sensor exhibits high sensitivities and wide detection bandwidth. The noise equivalent pressure as low as 1.79 mPa/Hz^1/2 and -6 dB detection bandwidth of 34 MHz have been established [Fig. 6(c)].

Another interesting study is the integrated fiber-optic ultrasound transducer. The first to be examined is the discrete device-based ultrasound transducer, which comprises an ultrasound transmitter and an ultrasound sensor, and has been used to image biological tissue. In the pursuit of smaller size, fiber-optic ultrasound transducers with high integration have been developed by combining the ultrasound transmitter and ultrasound sensor onto a single optical fiber. Currently, a fiber-optic ultrasound transducer with a diameter of 125 μm has been developed [Fig. 9(b)], which exhibits an emission pressure up to 846 kPa and a noise equivalent pressure as low as 1.7 kPa.

Fiber-optic ultrasound transducers have been used in industrial and medical fields, such as nondestructive testing, medical diagnosis, and therapy. In medical diagnosis, several imaging techniques based on fiber-optic ultrasound transducers have been developed, including photoacoustic microscopy, photoacoustic tomography, endoscopic imaging, and ultrasound imaging. Particularly, fiber-optic ultrasound transducers have considerable potential in endoscopic imaging owing to their tiny size of only 3.2 mm in diameter and excellent spatial resolution of 46 μm [Fig. 12(a)]. In the realm of nondestructive testing, fiber-optic ultrasonic transducers are progressing toward large-scale array multiplexing. Currently, a nondestructive testing system based on fiber Bragg grating sensors has been able to detect and locate faults. Moreover, partial discharge location with a positioning accuracy of 0.3 m has been realized using fiber-optic ultrasonic transducers [Fig. 15(e)].

To summarize, although the fiber-optic ultrasound transducer technology has achieved remarkable advances, such as high-pressure ultrasonic emission and broadband ultrasound detection, certain flaws still exist, such as low excitation efficiency, low detection sensitivity, and inadequate multiple abilities. Therefore, new materials must be developed and the structure of transducer elements must be optimized to obtain the high efficiency of ultrasound emission. Further, the ultrasound detection sensitivity needs to be further enhanced using resonant structures and new types of sensing materials. Parallel multichannel ultrasound detection should be researched to improve detection speed, which necessitates developing the demodulation technology. In the future, fiber-optic ultrasonic transducers will exhibit great potential in industrial, medical, and other industries.

Diabetes is a kind of non-infectious human metabolic disease caused by the inability to secrete insulin or secrete insulin normally. In recent years, the prevalence of diabetes has become much higher than before. In order to improve the life of patients and save the medical resources, it is necessary to further optimize the detection methods of diabetes. Human breath contains numerous information related to health. The breath of healthy people contains hundreds of volatile organic compounds (VOCs), whose composition and concentration can reflect the metabolism of a human body. The concentration of glucose in the blood of diabetic patients can be reflected by detecting the acetone concentration in breath. At present, the acetone detection methods mainly include gas chromatography, mass spectrometry, and laser spectroscopy. The required equipment is usually expensive and complex, which limits their application as a daily breath detection method for diabetic patients. As an ultra-high-precision 3D printing technology, femtosecond-laser induced two-photon polymerization microfabrication has been widely used in the fields of photonics, micromachines, microfluidics, and biomedicine, due to its high manufacturing precision and flexibility. Combining the two-photon polymerization technology with fiber optical sensing, the developed integrated device has the characteristics of high precision and flexibility, which provides a new solution for gas sensing. In this study, we report an acetone gas sensing method based on the polymer micropillar structure fabricated on the end face of a fiber. The absorption of acetone molecules by the photoresist micropillars polymerized on the end face of the fiber leads to the change of refractive index of the polymer-acetone mixed material, and subsequently makes the reflection interference spectrum from the Fabry-Perot interferometer (composed of the photoresist micropillars and the fiber end face) shift.

Two-photon polymerization is used to prepare the polymerized micropillars on the end face of a single-mode fiber. First, the single-mode fiber is cut flatly and placed on a glass slide. Then the cut end face is immersed in the negative photoresist. The sample is fixed on a three-dimensional air-floating displacement platform and polymerized by a femtosecond laser. After the laser polymerization process is complete, the sample is developed in a mixed solution of acetone and isopropanol. After the development, the micropillar structure is firmly adhered on the end face of the optical fiber. The morphology of the fabricated device is characterized by scanning electron microscope. The reflection spectrum of the polymerized micropillar sensor is measured by a broadband light source and a spectrum analyzer. The reflection spectra and free spectral range (FSR) values of the micropillar at different heights are compared. The spectral response of the sensor to acetone concentration is also tested. The temperature response of the sensor and the influence of temperature change on acetone concentration in the daily breath detection are studied.

In this study, the micropillar with a designed height of 20 μm is first characterized by scanning electron microscopy. The measured height of the micropillar is about 19.6 μm after polymerization (Fig. 3). The reflection spectra of micropillars with different heights and their FSR values at ~1550 nm are tested, showing that the heights of the micropillar devices are inversely proportional to their FSR values (Fig. 4). The spectral response of the sensor with a micropillar height of 30 μm to acetone concentration is measured. In the concentration range of 1×10^-9 to 1×10^-3, the reflection spectrum shifts by 5 nm. The mechanism is that after the polymerized micropillar absorbs acetone gas molecules, the material refractive index increases (Fig. 5). The trough wavelength decreases with the increase of acetone concentration in the environment. When the acetone gas concentration is less than 1×10^-7, the decreasing trend of trough wavelength becomes slow. This is because when the acetone gas concentration is very low, the acetone molecules absorbed by the polymerized material are limited, which is close to the lower detection limit of the sensor. Thus, the detection limit of the sensor is determined to be 1×10^-9. The sensor is repeatedly tested using acetone gas with the concentration range of 1×10^-9 to 1×10^-3. The blue shift of the sensor near 1550 nm in the reflection spectrum and the red shift of the sensor near 1550 nm are very similar, and both are with a close standard deviation of wavelength shift (less than 5%) (Fig. 6). The temperature response of the sensor from 25 ℃ to 55 ℃ is also studied. The reflection spectrum of the sensor shows a red shift with the increase of temperature. The reason is that with the increase of ambient temperature, the thermal expansion of polymerized micropillars leads to the increase of the FPI cavity length. During the detection, the temperature change introduces a large influence on the detection results of acetone concentration. It is necessary to strictly control the temperature of the gas to be tested and ensure the detection stability of acetone concentration (Fig. 7).

In this paper, the method based on two-photon polymerization for polymerizing micropillars on end face of a fiber and constructing an FPI acetone gas sensor is proposed. The sensor has a detection range of 1×10^-9 to 1×10^-3 for acetone concentration and a detection limit of 1×10^-9. This work provides a new, high-sensitivity, high-integration, and simple-to-use method for the detection of acetone in breath. If a new type of photoresist material with specific absorption capacity for acetone is further applied in sensor preparation, the sensor proposed in this paper will be more useful in practical clinical applications. Through the non-invasive breath detection, it will become a new method for the detection of blood glucose concentration in diabetic patients.

Optical parameters such as the attenuation coefficient are important environmental indicators of water bodies, including ocean, lakes, and rivers. The accurate measurement of the optical parameters of water bodies is crucial for ocean color remote sensing, ocean carbon cycle research, and ocean primary productivity assessment. Recently, the lidar technology has been gradually employed in the measurements of the marine chlorophyll concentration, scattering layer, optical attenuation coefficient, etc. However, the currently available lidar technique is mainly based on the time-of-flight principle, putting high demands on the laser energy, laser pulse width, and signal sampling rate to achieve large-depth and high-spatial-resolution measurements. In this work, we have designed and developed a Scheimpflug lidar (SLidar) system based on the Scheimpflug imaging principle (Fig. 1) for attenuation coefficient measurements in water bodies. The SLidar technique features a high range resolution, low cost, low maintenance, and wide wavelength selectivity, which can provide a novel approach for the quantitative detection of the water-body attenuation coefficient.

The water-body SLidar system (Fig. 2) mainly includes three parts: transmitting, receiving, and control units. The transmitting unit comprises a 450 nm multimode laser diode and a collimating lens. The 15-mm collimated laser beam is transmitted into a water tank. The backscattered light is collected using a receiving lens and then detected using a 45° tilted CMOS camera with an interference filter for ambient light rejection. The distance between the transmitting and receiving units is ~100 mm to satisfy the Scheimpflug principle. During measurements, the laser diode is on-off modulated and synchronized with the exposure of the CMOS camera, which alternately acquires the laser beam image and the background image. Thus, the dynamic subtraction of the background signal is achieved. In addition, the pixel-distance relation (Fig. 3) in water-body measurements is evaluated by considering the refraction effect at air-glass-water interfaces. The SLidar system is utilized to measure the attenuation coefficient of water bodies in a laboratory with different mass concentrations of fat emulsion (Intralipid). The slope method and Klett method are used to quantitatively retrieve the attenuation coefficient of the water body.

A water-body experiment is conducted under laboratory conditions. The transmitted laser beam in tap water is imaged using a CMOS camera. After pixel binning, background subtraction, and pixel-distance transformation, the range-resolved lidar profile is obtained (Fig. 4). Experiments with different mass concentrations of the fat emulsion are performed to examine the performance of the SLidar system. The fat emulsion solution with a mass concentration of 0.2 g/mL is diluted to 4 g/L, which is then added to the tap water to simulate different optical properties of the water body. During the measurement, the lidar signal is continuously recorded (Fig. 5), while the diluted fat emulsion (4 mL in total) is subsequently added to the water tank four times. The mass concentrations of the fat emulsion in the mixed water are then calculated, i.e., 0.17, 0.35, 0.52, and 0.69 mg/L. As shown in Fig. 6, the water body becomes inhomogeneous after adding the fat emulsion. Moreover, as the mass concentration of fat emulsion increases, the intensity of the lidar signal at a close distance first increases but then decreases rapidly. The attenuation coefficient profile is obtained using the Klett method (Fig. 7). The attenuation coefficient fluctuates significantly after the addition of the fat emulsion. As the fat emulsion and water body are fully mixed, the water body is almost homogeneous and the attenuation coefficients obtained using the Klett method at different distances are nearly the same. Furthermore, the attenuation coefficient of the water body generally increases with increasing mass concentration of fat emulsion (Fig. 8). The mean value of the attenuation coefficient under homogeneous conditions is evaluated. With an increase in the mass concentration of fat emulsion, the attenuation coefficients inverted using the Klett method and slope method increase correspondingly (Fig. 9). In addition, the correlation coefficient between the attenuation coefficient and mass concentration of fat emulsion reaches up to 0.997, successfully proving the reliability of the measurement result.

In this work, we design and develop a SLidar technique based on the Scheimpflug imaging principle for attenuation coefficient measurements in water bodies by utilizing a 450 nm laser diode as the light source and a CMOS image sensor as the detector. In addition, the pixel-distance relation in water-body measurements is calibrated by considering the refraction effect at air-glass-water interfaces. The SLidar system is used for water-body investigations in a laboratory with different mass concentrations of fat emulsion, and the slope method and Klett method are employed to retrieve the attenuation coefficient of the water body. Experimental results show that the SLidar technique can capture the dynamic changes in the water body as well as retrieve the water attenuation coefficients, which are consistent with the concentrations of the added fat emulsion. These promising results successfully demonstrate the great feasibility of using the SLidar technique for quantitative water body measurements, paving a way for open-water measurements.

Microwave photonic (MWP) systems generate, manipulate, transmit, and measure high-speed radio-frequency (RF) signals in the optical domain. Converting RF signals in the optical domain improves the signal-processing bandwidth and speed and reduces power consumption by complex electronic systems. Optical modulation is the most important step in converting microwave signals into optical signals in all MWP systems. This usually determines the performance of the whole system, including its bandwidth, system loss, linearity, and dynamic range. Nonlinear distortion is introduced mainly by the nonlinearity of the Mach-Zehnder electro-optical modulator (MZM) used in MWP systems. The modulation curve of a typical MZM takes the form of a cosine function. To achieve the approximate linearity of the modulation, the working point of the modulator is fixed at an orthogonal offset point, where it cannot fulfill the requirements of a microwave photonic link. A stable and efficient MWP system requires the modulator to exhibit a low noise figure, i.e., less than 10 dB, and a high spurious free dynamic range (SFDR) exceeding 120 dB·Hz^2/3. The SFDRs of a typical silicon MZM and a microring modulator are approximately 97 dB·Hz^2/3 and 84 dB·Hz^2/3, respectively. Therefore, a high-linearity electro-optical modulator is urgently needed.

Two options exist for improving the linearity of an electro-optical modulator: the electrical and optical domains. In the electrical domain, one method is to use electronic predistortion. Introducing arcsine predistortion into the RF signal compensates for the cosine modulation. Another method is to employ electronic post compensation. This method uses digital sampling at the output to remove the distortion term produced by the modulator. These methods in the electrical domain do not fundamentally improve the linearity of the modulator, and they require an accurate control of the introduced distorted signal and additional high-speed electronic devices. With increasing working time, the MZM itself experiences unstable factors such as temperature drift. This requires electronic compensation to adapt dynamically to the change in the modulator, which leads to a complex system with limited performance.

In the optical domain, various methods can be used to improve the linearity of the modulator and achieve improved performance. These methods commonly include the dual-polarization control, MZM series/parallel, and microring-assisted MZM (RAMZM) methods. The basic idea of the dual-polarization method is to control the third-order distortion power of TE and TM light at the same strength but in opposite directions to cancel each other. This is usually achieved by adding a polarization controller at the input or output of the MZM modulator, and it produces limited enhancement. The dual-polarization control method can be combined with the MZM parallel method, which is called the polarization-multiplexing MZM parallel method. By combining three linearization methods—power-weighting control, polarization multiplexing, and bias control—it can enhance the third-order SFDR from 95.4 dB·Hz^2/3 to 112.3 dB·Hz^2/3 compared with a conventional MWP link, and the second-order SFDR is 94.6 dB·Hz^2/3(Fig. 9). The basic idea of the MZM series/parallel method is to use one MZM to compensate for the third-order distortion caused by another MZM. The two MZMs can be connected either in series or in parallel. The double-parallel MZM method utilizes two MZMs connected so that the third-order intermodulation distortions generated by the upper and lower MZMs cancel each other. This method has a wide optical bandwidth, high manufacturing tolerance, and temperature-variation tolerance. Specific control is provided by the driving RF and bias voltage. The electrical-power distribution ratio between the two modulators and DC bias angle of the RF signal in the two submodulators can be controlled appropriately to construct two nonlinear distortion signals with opposite phases, so as to cancel the IMD3 intermodulation distortion in the link. Using bias-voltage control, the nonlinear distortion term can be cancelled by controlling the phases of the RF electrical and input optical signals. The principle of the MZM series method is basically similar to that of the parallel method. The MZM series/parallel methods need to adjust the bias voltage and power or phase of the driving RF signals accurately. Because the modulator is sensitive to temperature, an additional control circuit is needed to stabilize the bias voltage and temperature. The microring-assisted MZM (RAMZM) method utilizes the superlinear phase modulation of the microring to compensate for the nonlinear cosine modulation function of the MZM. It has a simple structure and can achieve high linearity. The key point of RAMZM is to control the coupling coefficient between the microring and MZM arm. The fabrication tolerance of the directional coupler between the microring and modulation arm of the RAMZM is small, and the losses in the microring also affect the linearity. Lithium niobate exhibits the characteristics of high bandwidth, good modulation, and low loss. Further, it exhibits a high-linearity electro-optical effect that silicon does not. Recently, the thin films of lithium niobate on an insulator (LNOI) have become promising platforms for photonic integration. Using materials with high linearity, we can expect to improve the linearity of an electro-optical modulator further and make it practical.

Although linearization methods in the electrical domain are traditional and widely applicated, the required electronic control equipment is complex. These methods are used for linearization-compensation for traditional modulators with poor linearity. Conversely, optical-domain linearization is committed to eliminating nonlinear components through optical methods without introducing redundant electronics or compensation equipment. With the continuous development of integrated optics and on-chip integrated optoelectronic devices, the optical-domain linearization method has gradually become a research hot spot. Reported methods include the optical-polarization, MZM series/parallel, and microring-assisted MZM methods. These methods can be used in combination with each other. With the fast development of thin-film lithium niobate platforms, the prospect of fabricating high-performance electro-optical modulators with high linearity on LNOI appears promising.

Big data services have yielded explosive growth of capacity in short-reach optical networks. The intensity modulation direct-detection (IMDD) systems with simple schemes and power-efficient digital signal processing (DSP) are typically preferred in short-reach scenes. However, they are unable to satisfy the demand for continually increasing interface speed. The Ethernet interface data rate is approaching 800 Gbit/s and 1.6 Tbit/s. Hence, the conventional IMDD will suffer from serious technical challenges, including dispersion-induced power fading, rapidly increasing cost, and limited sensitivity.

As an alternative, coherent technology can provide high spectral efficiency, high sensitivity, and good tolerance to the chromatic and polarization-mode dispersion. However, for short-reach application, this technology is considered overly costly and power-consuming. These drawbacks originate from two main challenges. On one hand, power-consuming DSP is required for solving various impairments on the received signal in traditional coherent schemes. Moreover, with the fading of the Moore’s law, the node gain brought by new footprints of application specific integrated circuits (ASICs) tends to be marginal. Using only advanced DSP in the development of coherent technology for short-reach high-speed interconnections is quite difficult. On the other hand, traditional coherent technology is associated with complex hardware structures, especially the adoption of narrow-linewidth, frequency-stable, and tunable lasers, such as external cavity laser and integrable tunable laser assembly (ILTA). Consequently, coherent technology is still inapplicable to short-reach networks.

Apart from conventional IMDD and coherent technology, many self-coherent schemes have been proposed with certain tolerance to laser linewidths and less implementation complexity (than that of the conventional systems). The Kramers-Kronig receiver (KKR) and Stokes vector receiver are two typical schemes, each receiving considerable attention. In terms of the product of analog to digital converter ADC bandwidth and number of ADCs, these schemes are more expensive while achieving the same capacity as that of the conventional coherent technology. The self-homodyne coherent (SHC) scheme has been proposed as another "coherent-lite" scheme, including polarization division multiplexing and space division multiplexing as the categories. The key feature of this scheme is that the signal lights transmit simultaneously along with their pilots in links. At receiver, coherent detection will be conduced though the remotely delivered pilot to achieve optical domain phase recovery of signal. Thus, the cost and the power consumption are reduced, and the use of low-cost and uncooled distributed feedback (DFB) laser and baud-rate-sampling receivers is realized. The advantages of the coherent technology are therefore inherited and the scheme is simplified, and hence this technology is considered one of the most promising technologies for future short-reach optical networks. Despite the excellent characteristics of SHC schemes, many key implementational issues must be solved prior to deployment.

In space division systems, the relative time delay will induce a unique phase noise and degrade the system performance, which may prevent use of the low-cost DFB laser in the SHC scheme. Fortunately, the derivative of such phase noise is a colored frequency modulation noise. Utilizing this characteristic, we proposed and demonstrated an in-service high-precision and large-dynamic-range estimation method of relative time delay (RTD), contributing to the realization of the SHC technique. Besides, the random birefringence in the optical fiber will lead to changes in the state of polarization (SOP) during delivery of the pilot and signal. Another implementational issue is that automatic compensations of such randomly changed SOP is required for real fiber links. By leveraging our in-house-developed adaptive polarization controller (APC), we solved both problems of polarization demultiplexing in polarization division multiplexed (PDM) SHC systems and pilot SOP locking in space division multiplexed (SDM) SHC systems. The APC technique allows further simplification of the DSP algorithms. Utilizing only one APC device and its symmetry property, we also demonstrated the first multi-input and multi-output (MIMO)-free SDM-SHC transmission and PDM-SHC transmission in bidirectional (BiDi) scenes. The APC technique paves the way for low-cost, power-efficient, high-speed BiDi optical interconnections. We present an overview of the progress that our group has realized for SHC systems, including the APC techniques, the in-service RTD estimation techniques, and the simplest BiDi SHC transmission architectures based on APC techniques.

The SHC scheme capable of optical-domain equalization, high spectrum efficiency, and compatibility with current ASIC architecture has been demonstrated. This scheme provides a promising method for future low-cost short-to-medium-reach optical interconnections. Moreover, the SHC technology will generate new demands from other technical areas, including photonic integration, special optical fibers, and DSP. This technology will promote and accelerate innovations in multiple fields.

Nonlinear Fourier transform (NFT) can convert a signal into a nonlinear spectrum, including continuous spectrum and discrete spectrum, where the eigenvalues are located in the upper half of the complex plane. With the approach of NFT, information is encoded into the nonlinear spectrum of a signal, which can effectively address the nonlinear transmission impairments arising in standard single-mode fiber. At the same time, as a new effective signal processing tool, NFT can be used to analyze pulses in fiber lasers. For a pure soliton, its nonlinear spectrum only contains the discrete spectrum. The eigenvalues in the discrete spectrum can then be used to characterize soliton pulses, wherein the real and imaginary parts refer to the frequency and amplitude of soliton pulses, respectively. This methodology provides a new perspective for the study of laser dynamics. The soliton and continuous wave background can be separated based on their different eigenvalue distributions after the obtainment of full-field information of pulses. We summarize the principle of NFT and its applications in the field of optical communications and fiber lasers, specifically the "soliton distillation" technology based on NFT.

In optical fiber communication systems, there are transmission impairments such as loss, dispersion, and nonlinearity. Loss and dispersion can be compensated by optical amplification and dispersion compensation technology. Pulse broadening and distortion caused by nonlinear effects related to optical pulse intensity have, however, become the main factors limiting the system communication capacity improvement. As a powerful mathematical tool, NFT can effectively solve the problem of lightwave propagation in a nonlinear medium such as an optical fiber. Recently, a new framework of optical fiber communication systems based on NFT has begun receiving extensive attention. NFT can decompose a signal into a continuous spectrum (nonsoliton component) and discrete spectrum (soliton component), which are considered nonlinear spectra. With this method, information can be encoded into the nonlinear spectrum of the signal, and the technique of doing so is known as nonlinear frequency division multiplexing (NFDM). Compatible with the nonlinear response characteristics of optical fibers, NFDM can effectively address the dispersion and nonlinear impairments arising in standard single-mode fiber (SMF) transmission.

Optical soliton is a special light field that does not change during transmission under the dispersion and nonlinear effects, and it can be generated and spread in optical fiber systems. Periodically repeating stable pulses generated in fiber lasers can also be considered as solitons, more specifically known as dissipative solitons. The output signal can be transformed into a nonlinear spectrum (including continuous spectrum, discrete spectrum, and corresponding eigenvalues) through NFT. In the nonlinear Fourier domain, nonlinear phenomena in optical fiber systems are investigated, such as rouge wave, optical frequency combs, and cavity solitons. Lately, NFT has been applied to laser pulse analysis. For pure soliton, its nonlinear spectrum only contains a discrete spectrum and the eigenvalues in the discrete spectrum correspond to the characteristics of the soliton. For example, the real and imaginary parts of an eigenvalue correspond to the frequency and amplitude of the soliton, respectively. When the dynamic characteristics of the pulse are dominated by solitons, that is, when the discrete spectrum cumulates most of the pulse energy, the eigenvalue distribution can reflect the pulse feature, as shown in Fig. 13. At the same time, pure soliton and the resonant continuous-wave background can be separated according to different eigenvalue distributions, to realize soliton distillation. These findings show how NFT provides new insights into ultrafast transient dynamics in fiber systems.

In an NFDM system (Fig. 3), after the nonlinear spectrum of the signal is recovered at the receiving end, the effects of dispersion and nonlinearity can be eliminated by simple phase compensation, and the system performance can be improved. According to the different modulation spectrum, the nonlinear spectrum can be divided into discrete spectrum communication and continuous spectrum communication. Research on discrete spectrum communication has focused on the number of multiplexed eigenvalues and advanced modulation formats for discrete spectrum. As for continuous spectrum communication, compared to the traditional OFDM systems, NFDM systems have higher quality factors under the optimal condition of fiber launching power. However, the NFT-based optical communication system still suffers from random noise. Meanwhile, problems, such as channel integrability and algorithm complexity, still exist, which greatly restrict the performance of the system.

To perform NFT analysis experimentally in the field of fiber lasers, full-field information, including the amplitude and phase of the pulse, must first be obtained. The current acquisition methods mainly include density functional theory and temporal lensing technology, combined with Gerchberg-Saxton phase recovery (Fig. 9) and coherent homodyne detection technologies (Fig. 11). After the full-field information of pulse is obtained, the pulse features can be projected onto the eigenvalue distribution by NFT. NFT can be used as an analysis tool to reduce the complexity of describing pulse evolution, whether for nonstationary (Fig. 5) or stationary pulses (Fig. 6). At the same time, according to the distribution of eigenvalues and their corresponding discrete spectra, the temporal evolution process of a pulse can be reconstructed using inverse nonlinear Fourier transform (INFT), which also shows that the NFT method can effectively characterize the laser pulse. INFT is not only effective for a single pulse (Fig. 7) but also achieves a good reconstruction effect for multiple pulses (Fig. 8). Further, the sideband eigenvalues can be removed and only the soliton eigenvalues are retained in the nonlinear spectrum. Through INFT, pure soliton can be recovered in the time domain (Fig 14). NFT can perform pure soliton distillation and reconstruction on various pulses generated in fiber lasers, including single pulse, single pulse in period-doubling, different double pulses (Fig. 15), and multiple pulses. The transient nonlinear dynamic analysis based on NFT can deepen the knowledge on soliton formation and its interaction process, and also clarify the transient working mechanism of fiber laser.

As an emerging signal processing tool, NFT provides new system design solutions in the field of optical communication, which is fully compatible with optical fiber. Transient nonlinear dynamics analysis based on NFT can describe laser pulses theoretically and completely, providing a basic overview on ultrafast nonlinear dynamics, and its application in ultra high-power fiber lasers.

Recently, there has been a rapid evolution in optical communication systems, leading to the establishment and development of various subdivisions, such as metropolitan area networks, access networks, and data center optical interconnection. However, current optical network architectures available on the ground are becoming insufficient to meet the growing demands of the society. Therefore, some new application scenarios, such as satellite communication, marine communication, and communication in some areas where optical fibers are difficult to arrange, e.g., mountains, forests, and lakes, are attracting widespread attention. Based on current optical communication network architectures, a three-dimensional, spatial, and multimodal optical network system is emerging. In this system, the free-space optical (FSO) communication featuring unlicensed bandwidth, high capacity, strong confidentiality, and easy setup plays an important role. Therefore, for practical in-field application of FSO communication, studies on embedded real-time FSO systems are necessary.

To investigate the real-time applications of FSO transmission, in this study, we experimentally demonstrate a real-time multicarrier FSO communication system with a self-designed electrical board, including a field-programmable gate array (FPGA), four-channel transmitter supporting 2.5 GBaud signals, and 4×5 GSa/s analog-to-digital converter (ADC). The 10 Gb/s polarization digital multiplexing quadrantile phase-shift keying (PDM-QPSK) signals were generated using a dual-polarization IQ (DP-IQ) modulator and loaded onto eight optical carriers spaced at 12.5 GHz. All optical carriers can be recovered with an error-free bit error ratio (BER) performance.

The experiment setup of the real-time 8×10 Gb/s PDM-QPSK coherent transmission over a 1 m FSO link and the experiment platform are shown in Figs. 1 and 5 (a), respectively. At the transmitter side of the system, eight external cavity lasers with ~100 kHz linewidth spaced at 12.5 GHz were used as the sources. All the channels were coupled using an optical coupler and fed into a DP-IQ modulator, where the 10 Gb/s pseudo-random-bit-sequences-23 were modulated. The signal spectrum of the transmitted signal is shown in Fig. 2. Then, the optical signal was amplified using erbium-doped fiber amplifiers (EDFA) to adjust the power and transmit it into the free-space link. At the receiver side, the signal was detected using an integrated coherent receiver and ADC. Subsequently, different channels were selected by tuning the wavelength of a local oscillator without any optical filter. All the sampled signals were processed using the FPGA with real-time digital signal processing (DSP) algorithms, including ADC synchronization, clock recovery, constant modulus algorithm (CMA), frequency offset recovery, phase offset recovery, and symbol decision, as shown in Fig. 1. The results of the experiment were shown in Figs. 6 and 7. Figure 6 (a) shows the receiver sensitivity of the real-time integrated coherent receiver in a back-to-back case. By adjusting the optical attenuator, we reduced the receiver power from -45 to -51 dBm and obtained the BER using the signal-to-noise ratio. When the receiver power is -50 dBm, BER is 2.9×10^-3, which is under the 7% FEC limit (BER is 3.8×10^-3). Figure 6 (b) shows the received spectrum and constellation diagram from channel 8 in the multicarrier FSO experiment. The results of each stage of the real-time algorithm processing are shown in Fig. 7. Figures 7 (a)-(d) represent the sampled data of ADC without DSP algorithms, results after CMA, results after frequency offset recovery, and results after phase compensation, respectively.

In this study, we propose an 8×10 Gb/s PDM-QPSK real-time digital coherent communication system via experiments using a free-space link. Based on an FPGA chip, we have completed the programming of the real-time DSP algorithms and conducted the corresponding performance test. The experimental results show that after using EDFA, under the 7% FEC threshold, the receiving sensitivity of the digital coherent module is as low as -50 dBm. Furthermore, all the channels of the system with 12.5 GHz frequency intervals achieve a real-time error-free transmission through a 1 m free-space link.

The analog radio-over-fiber (A-RoF) technique can directly transmit radio-frequency (RF) signals between the baseband unit (BBU) and remote antenna unit (RAU) and offers the advantages of high spectral efficiency, ultralow latency, and a simple structure. In addition, millimeter-wave (mm-wave) mobile communication can utilize wide spectral resources to transmit high-rate signals. Therefore, the mm-wave over fiber based on the A-RoF technique and mm-wave mobile communication is considered the most potential solution for beyond-fifth generation (B5G) fronthaul. However, the A-RoF technique is sensitive to linear and nonlinear distortions and the generation of mm-wave signals requires high-bandwidth photonic and electronic devices. In our previous work, four-independent mm-wave signals were modulated on two orthogonal polarization states of a single wavelength based on a dual-polarization IQ modulator (DP-IQMZM) using the dual single-sideband (SSB) modulation and polarization division multiplexing (PDM) technique. Furthermore, a novel carrier polarization rotation module based on the self-polarization stabilization technique was proposed; thus, the four-independent mm-wave signals could be detected via self-coherent detection. Experimental results showed that the measured error vector magnitude (EVM) value of 800 MBaud 16-ary quadrature amplitude modulation (16-QAM) signals at 28 GHz over 50 km standard single-mode fiber (SSMF) transmission was 12.99% without digital signal processing (DSP). However, photonic frequency upconversion was not realized in our previous work. The bandwidth requirement of photonic and electronic components at the transmitter is high. In this study, we propose a scheme for upconverting four independent low-frequency RF signals to high-frequency mm-wave signals using photonic frequency upconversion. Moreover, no digital signal algorithm is used at the receiver, which is helpful for constructing a DSP-free RAU.

By paralleling one DP-IQMZM and one single-drive Mach-Zehnder modulator (MZM), which is used to generate second-order optical subcarriers, four low-frequency RF signals can be upconverted to high-frequency mm-wave signals using the self-heterodyne detection technique. In this way, the bandwidth requirement and sampling rate of photonic and electronic components at the transmitter can be reduced considerably. In addition, frequency offset compensation and carrier phase recovery are avoided. Furthermore, we analyze the causes of crosstalk between symmetric sidebands and propose a method for crosstalk elimination by accurately matching the phase and amplitude of the in-phase (I) and quadrature (Q) components of the employed DP-IQMZM.

The sideband suppression ratio can be increased from less than 20 dB to more than 30 dB using our proposed crosstalk elimination method between symmetric sidebands (Fig. 10). Moreover, the transmission performance of dual-SSB signals is very close to that of SSB signals [Fig. 11(a)], verifying the effectiveness of our proposed method. Experimental results show that four independent 1.6 GBaud 16-QAM mm-wave signals with a carrier frequency of 30 GHz could be generated at the receiver using four independent 1.6 GBaud 16-QAM signals with a carrier frequency of 10 GHz and a single-tone signal with a carrier frequency of 20 GHz at the transmitter. As shown in Fig. 12(b), the measured EVM value of 25.6 Gbit/s 16-QAM signals at 30 GHz over 50 km SSMF transmission are all below the threshold of 12.5% without the use of any DSP. Moreover, the minimum sampling rate of the DAC at the transmitter is 24 GSa/s [Fig. 13(a)].

A coherent radio-over-fiber transmission system based on the self-heterodyne detection technique is proposed, which can realize photonic frequency upconversion. In the proposed system, four low-frequency RF signals are upconverted to high-frequency mm-wave signals and no DSP algorithms are required in the RAU. The causes of crosstalk between symmetric sidebands are analyzed, and a crosstalk elimination method at the transmitter is proposed. Experimental results show that four independent 1.6 GBaud 16-QAM mm-wave signals with a carrier frequency of 30 GHz can be generated at the receiver using four independent 1.6 GBaud 16-QAM signals with a carrier frequency of 10 GHz and a single-tone signal with a carrier frequency of 20 GHz at the transmitter. Using the crosswalk elimination method between symmetric sidebands, the proposed system can support the transmission of four independent 1.6 GBaud 16-QAM mm-wave signals with a carrier frequency of 30 GHz over a 50 km SSMF without any DSP at the receiver. Moreover, the minimum sampling rate of the DAC at the transmitter is 24 GSa/s, effectively reducing the cost and complexity of B5G fronthaul systems. This research provides a potential solution for the mobile fronthaul network in the B5G mobile communication.

Lidar has been widely used in wind ranging, automatic drive and sensing mapping. The reflected light signal is obtained through first emitting a Gaussian beam from the laser source and then reflecting after reaching the surface of the object. After the computer analysis, the information of the object such as orientation, attitude and distance can be obtained. However, as for a lidar system, its laser source is an important unit influencing the performance of the whole system. A fiber laser has become the best choice of the light source for a lidar system, because of its good beam quality, high pulse energy and high repetition rate. At the same time, an erbium-ytterbium co-doped fiber has attracted the attention of many researches due to its advantages such as "eye-safe" and low atmospheric transmission loss. Therefore, as the most important gain medium for the laser lidar, polarization-maintaining erbium-ytterbium co-doped fiber has important research significance. In this paper, a 10 μm/128 μm polarization-maintaining erbium-ytterbium co-doped fiber is successfully fabricated by the modified chemical vapor deposition (MCVD) technique combined with the solution doping technology (SDT). The structural parameters and optical properties of this polarization-maintaining erbium-ytterbium co-doped fiber are measured. And its laser performance is also studied.

MCVD combined with SDT is used to fabricate the erbium-ytterbium co-doped fiber. The content (mole fraction) of P₂O₅ in the core is increased by more than 10% with reverse phosphorus doping and gas phase compensation. In order to avoid the defect of the core bursting during drilling, the fiber prefabricated rod is first annealed due to the high stress of its core. Through the Sagnac interferometer and the optical spectrum analyzer (OSA), the birefringence value is measured. The measurement structure is shown in Fig. 3. And the measurement structure of polarization extinction ratio is also shown in Fig. 5. In order to analyze the laser performance, the structure of an erbium-ytterbium co-doped fiber laser is shown in Fig. 6. The seed source has a power of 20 mW and a central wavelength of 1551 nm. An isolator (ISO) connected to the seed is used to protect the seed source. The isolator is followed by a (2+ 1)×1 forward pump combiner (PC), and one of its pump fiber is used to monitor the backward power and observe the backward spectrum. The 940 nm light generated by the laser diode (LD) is coupled to the active fiber through the pump fiber of the (2+ 1)×1 backward PC, and the pump power is 16.5 W. The coiling diameter of the active fiber is 10 cm. The cladding pump stripper (CPS) is implemented by coating a high refractive index adhesive to filter cladding light from the fiber. Finally, an isolator is fused at the end to prevent reflection.

The dimension of the fiber is shown in the inset of Fig. 2(b). The diameters of core and cladding are measured to be 10.19 μm and 128.69 μm, respectively. The diameter of the boron rod is measured to be 32.59 μm. Figure 2(b) shows the refractive index profile of the prefabricated rod. A numerical aperture of 0.24 is finally achieved. The absorption coefficient measured by the truncation method is 2.42 dB/m at 940 nm. The interference image at 1500-1600 nm is observed in the OSA (Fig. 4). The beat length at 1550 nm is calculated to be 9 mm with a birefringence coefficient of 1.29×10^-4. At the same time, a polarization extinction ratio of 24 dB at 1310 nm is measured through the erbium-ytterbium co-doped fiber with a length of 4 m. As for the laser performance, due to the inherent loss, the final seed power coupling into the active fiber is 17.5 mW. Figure 7(a) shows the slope efficiency under different fiber lengths and pump powers. It can be seen from this figure that the optimal length is 7.5 m. When the pump power is 16.5 W, the output power and the slope efficiency reach the maximum, which are 5.8 W and 36%, respectively. The polarization extinction ratio is measured to be 21 dB. In addition, as shown in Fig. 7(b), the optical-to-optical efficiency tends to be saturated with the increase of pump power at different lengths. After reaching the saturation state, the optical-to-optical efficiency is more than 33% without a downward trend, which indicates that the laser power can be further increased at this time. The spectrum at 7.5 m is shown in Fig. 8. It can be observed from this output spectrum that the amplified spontaneous emission (ASE) power increases gradually with the increase of pump power, but the signal-to-noise ratio remains above 50 dB. From the backward spectrum, one can observe that the remaining pump light intensity is stable, which may be caused by the fact that some spiral pump light in the polarization-maintaining fiber is not absorbed by the fiber and the CPS is not added. Meanwhile, there is no parasitic oscillation at 1 μm. It shows that the polarization-maintaining erbium-ytterbium co-doped fiber prepared in this paper has good laser performances.

In this paper, a polarization-maintaining erbium-ytterbium co-doped fiber for lidar is successfully fabricated by MCVD combined with SDT. The performance of this polarization-maintaining fiber is measured. A birefringence coefficient of 1.29×10^-4 and a polarization extinction ratio of 24 dB@4 m at 1310 nm are achieved. In addition, a polarization-maintaining all-fiber erbium-ytterbium co-doped fiber laser system is built, and the slope efficiency reaches 36%. Above all, the highest efficiency of the polarization-maintaining erbium-ytterbium co-doped fiber is achieved, which provides the possibility for exact localization of a military lidar.

A vortex beam is a special light field possessing a spatial structure, including phase vortex and polarization vortex, and possessing a phase singularity and a polarization singularity correspondingly. A phase vortex laser carrying orbital angular momentum (OAM) has a helical phase front of exp(ilφ), where l is the topological charge value and φ refers to the azimuthal angle. The topological charge represents the twisting rate of the helical phase which is an unlimited value in principle. In addition, phase vortices with different topological charges are orthogonal to each other. A polarization vortex laser is a light beam with spatially variant polarization. Due to the phase singularity and polarization singularity, the intensity of a vortex laser beam at the center is canceled leading to a ring-shape intensity profile. Vortex beams have been widely used in astronomy, optical manipulation, microscopy, imaging, sensing, quantum science, and optical communications owing to their distinct advantages, such as inherent orthogonality and unbounded states in principle.

With the arrival ofa big data era, the dramatic increase of global internet traffic has attracted increasing research efforts for sustainable expansion of capacity. Beyond various advanced modulation formats and multiplexing techniques using the physical dimensions of photons including frequency, amplitude, phase, and time, space-division multiplexing (SDM) is recognized as an alternative technique to increase transmission capacity by exploring the spatial structure of photons. Optical vortices can be regard as a mode set to multiplex data information owing to the inherent orthogonality and unbounded states.

As one of the four major inventions in the 20th century, the laser is the basic part of an optical communication system and plays a very important role. Up to now, on one hand, a vortex beam is generated mainly through the conversion occurring outside the laser cavity, and the various applications based on vortex beam, such as diffractive optical elements, transform optics, spiral phase plate, fiber based devices, photonic integrated devices, metasurfaces, have obvious limitations in some aspects including imperfect spiral wave front phases. On the other hand, a vortex beam can be directly generated from a vortex laser, which can avoid some disadvantages of the conversions occurring outside the laser cavity, such as low conversion efficiency, poor beam quality after conversion, power limitation, and additional converter devices. Therefore, vortex lasers deserve more extensive and sufficient researches.

The ways to output a vortex beam directly from the laser cavity can be mainly divided into three types. 1) Inserting some elements, such as a spiral phase plate, lens, and diaphragm, into the cavity to generate a vortex beam from the laser cavity. 2) Using a special cavity mirror to select the oscillation mode in laser cavity. 3) Using a ring-shaped pumping to generate a vortex beam. By converting the pump beam into a ring-shaped one similar to the shape of the intensity profile of a vortex beam, the oscillation mode in the laser cavity can match the pump beam to the generated vortex beam.

The vortex laser based on discrete components is mainly composed of discrete single optical elements in free space. This laser is simple to construct, stable, and has large number of output modes. In 2005, Kozawa et al. demonstrated a polarization vortex laser by inserting a Brewster prism into the laser cavity based on Nd∶YAG [Fig. 3(a)]. A laser which can generate all states on the higher-order Poincaré sphere was demonstrated in 2016. By exploiting the geometric phase control inside the laser cavity to map polarization to OAM, the OAM degeneracy of a standard laser cavity may be broken to produce pure OAM beams, and the generalized vector vortex beams may be created with high purity at the source. A fiber laser is one with a doped fiber as gain medium which has better stability and better beam quality comparing with the semiconductor lasers. A fiber laser is more compatible to an optical fiber communication system, which can effectively reduce the system complexity and cost. A fiber laser based on a linear resonator and a ring resonator outputs different modes, which are demonstrated in Fig. 11. In recent years, the photonic integration technique has developed rapidly, and the miniaturization of optical devices is also a development trend. In addition, integrated devices are flexible, tunable, and reconfigurable comparing with the traditional discrete components. Therefore, a vortex laser based on integrated devices is of great research value. Vertical-cavity surface-emitting vortex lasers based on lead bromide perovskite and InGaAsP/InP platform are demonstrated (Fig. 15).

Here, we provide an overview of the recent progress of vortex optical lasers. We comprehensively review different types of vortex lasers, including vortex lasers based on discrete components, vortex lasers based on fiber, and vortex lasers based on integrated devices. Meanwhile, the future development trend of vortex lasers is analyzed and the prospect is discussed. Vortex lasers are expected to further promote the wide application of vortex beam in many fields.

High power, high energy ultrafast thin-disk laser oscillators have many important applications in the fields of scientific research, industry, biomedicine, and defense. Kerr-lens mode locking (KLM) is one of the most commonly used methods to generate ultrashort pulses directly from the thin disk oscillators. In this work, we review and discuss about the development of the KLM thin disk oscillators. The principle of KLM is introduced firstly, following by a presentation of the state-of-the-art of the ultrafast thin disk oscillators with respect to high average power, high repetition rate, short pulse duration, and new wavelengths. The applications and development prospect of the KLM thin disk oscillators are finally discussed.

Semiconductor saturable absorber mirror (SESAM) and KLM are two main mechanisms to realize ultrashort pulses from the thin disk oscillators. Compared with SESAM, KLM shows more advantages in modulation depth, damage threshold, and so on, and thus it enables higher power pulsed laser generation with a shorter pulse duration in the ambient air. In terms of high average power, Poetzlberger et al. introduced the active multi-pass cell (AMC) into the KLM thin-disk oscillator, increasing the laser gain within one roundtrip by the multiple passes of the laser pulse through the thin disk medium. The increased gain enables a higher output coupling rate and results in a higher output power. Under an output coupling rate of 50%, 150-W pulses with a pulse duration of 290 fs can be achieved (Fig. 2). For the shorter pulse duration, limited by the narrow emission bandwidth of the gain medium, it is difficult to achieve laser pulses with durations below 100 fs directly from a Yb∶YAG thin-disk oscillator. To solve this problem, on the one hand, one can utilize new materials with wider emission bandwidths as thin-disk gain media, such as Yb∶Lu₂O₃ and Yb∶CALGO. On the other hand, a broader spectrum can be generated by improving the mode locking technique with a stronger modulation depth. In 2021, Zhang et al. invented a distributed Kerr-lens mode-locking (DKLM) technique comprising of multiple Kerr-lenses in a Yb∶YAG thin-disk oscillator. It significantly increases the self-amplitude modulation (SAM) coefficient for KLM. The resulting spectral width exceeds the emission bandwidth of the Yb∶YAG gain medium by a factor of approximately four, leading to 47-fs pulse generation directly from the Yb∶YAG thin-disk oscillator (Figs. 3 and 4). Very importantly, the new concept is also applicable to other types of gain media, which may lead to new records in the generation of ultrashort pulses. With the same concept, Drs et al. realized 27-fs pulse generation from a Kerr-lens mode locked Yb∶YAG thin-disk oscillator by increasing the modulation depth, which is the shortest pulse duration ever obtained from a Yb∶YAG thin-disk oscillator. For a higher repetition rate, a strongly asymmetric configuration comprising of two concave mirrors with different radii of curvature is used to ensure a large beam size on the thin disk in a short oscillator cavity, and the 75-W, 260-fs pulse with a repetition rate of up to 260 MHz is obtained in this scheme, which is the highest repetition rate ever realized in a thin-disk oscillator (Fig. 5). Besides, for the generation of a new wavelength directly from the thin disk oscillator, Zhang et al. successfully realized the first femtosecond Ho∶YAG KLM thin-disk oscillator with an average power of up to 25 W, providing a solid foundation for the further development of 2 μm ultrafast thin-disk lasers. Compared with 1 μm laser, 2 μm femtosecond laser shows more advantages in driving the nonlinear frequency conversion for the generation of a mid-infrared laser due to the low multiphoton absorption and high conversion efficiency. Besides, more crystals can be used for the nonlinear conversion process with the 2 μm driving source, which is not possible with 1 μm driving source. With the 2 μm femtosecond thin-disk oscillator, Zhang et al. generated a broadband mid-infrared laser via intra-pulse difference-frequency generation (IDFG). The 2 μm femtosecond pulses delivered from the oscillator are first coupled into a photonic crystal fiber (PCF) for a further spectral broadening and temporal compression, and then focused onto the nonlinear crystal for the IDFG process. Nonlinear crystals such as GaSe and ZnS/ZnSe are used, and mid-infrared laser pulses with spectra ranging from ~3 μm to ~20 μm are generated finally.

Great progress has been made in 1 μm KLM thin-disk oscillators in terms of average power, pulse energy, pulse duration, and repetition rate during the last decades, which is beneficial for a lot of applications in the fields of fundamental research, industry, and medical science. At the same time, the realization of 2 μm KLM thin disk oscillators also makes itself more attractive for mid-infrared laser generation. In the future, more and more new materials will be explored as the thin-disk gain media, and pulses with shorter pulse duration, kW-level average power, and mJ-level energy can be expected from thin-disk oscillators.

Femtosecond laser technology has developed considerably in the past decades, promoting the progress of scientific areas such as ultrafast optics, strong field physics, super-resolution imaging, and precision measurement. In a typical femtosecond pump-probe experiment, the evolution dynamics of the microscopic structures of matter is excited and observed with micrometer to nanometer spatial resolutions and femtosecond to attosecond temporal resolutions. For a pump laser pulse, complete characterization of its spatiotemporal field distribution information such as pulse duration, dispersion, and wavefront distortion allows researchers to accurately control the laser-matter interaction process. For a probe pulse, the evolution history of the excited matter’s optical properties is encoded in the spatiotemporal amplitude and phase modulations of the probe pulse. Therefore, it is necessary to develop techniques for the spatiotemporal characterization of femtosecond laser fields in three dimensions, along two transverse spatial directions and one longitudinal temporal direction.

Conventionally, a series of femtosecond laser-pulse characterization techniques have been developed and most of them focus on measuring the "longitudinal" temporal profile of the laser field. For example, autocorrelation and frequency-resolved optical gating have matured and are widely applied worldwide in ultrafast optics laboratories. Moreover, commercial devices based on these techniques are available in the market. For obtaining the transverse spatial distribution information of a laser pulse, multiple techniques and devices are developed. Laser beam profilers are widely applied by academic and industrial users for transverse spatial intensity profiles of lasers. Phase information or wavefront distribution is obtained using wavefront sensors such as Shack-Hartmann devices. To obtain a three-dimensional spatiotemporal profile of a femtosecond laser field, the results of the longitudinal temporal profile characterization and the beam profile measurements are usually combined.

However, the conventional approach of combining the temporal characterization and spatial measurement results is not appropriate for complicated optical fields. First, with an increase in the peak power of femtosecond laser pulses to the petawatt level, the large-aperture optical elements in the petawatt laser facilities introduce severe spatially and temporally dependent optical field distortions. Such distortions are called the spatiotemporal coupling effect, and they have to be characterized simultaneously in the spatial and temporal domains. Second, complicated optical fields with optical vortices or structured polarization have attracted increasing attention of researchers in recent years and require advanced spatiotemporal characterization techniques. Third, some laser-driven short-wavelength light sources such as high harmonic radiations have spatiotemporal coupling nature. A complete characterization of their spatiotemporal profile enables studies on strong field physics in attosecond time scales.

Therefore, this paper aims at reviewing the recent progress in the spatiotemporal femtosecond laser-field characterization techniques, especially emphasizing their capability of simultaneously revolving two-dimensional spatial information and one-dimensional temporal information.

This paper is organized as follows. After a brief introduction in the first section, the second section reviews the techniques used to measure femtosecond laser pulses in the longitudinal time domain, as the traditional time-domain measurement techniques form the fundamentals of spatiotemporally resolved measurement techniques developed in recent years. According to the mechanism of the interaction between the pulse to be measured and the reference pulse in the time domain, the time-domain femtosecond laser-pulse measurement techniques are divided into three categories: intensity autocorrelation-based measurement (Fig. 1), frequency-domain interferometry-based measurement (Fig. 2), and phase modulation-based measurement (Fig. 3).

The third section is the most important part of this paper, which reviews the development of the spatiotemporally resolved femtosecond laser-field characterization techniques. We first discuss the applications of imaging spectrometers in different femtosecond laser spatiotemporal characterization techniques (Fig. 4). As the time-domain information is obtained from the spectral measurement, imaging spectrometers provide additional spatially resolved information simultaneously along the entrance slit direction. However, the entrance slit also blocks the light distributed along the direction perpendicular to it; therefore, almost all pulse spatiotemporal measurement techniques based on imaging spectrometers provide only one-dimensional lateral spatial information or two-dimensional spatial information by scanning the entrance position of the laser field on the slit in an unstable time-consuming way. Specific technologies include SEA-SPIDER, SRSI-ETE, and CROAK.

Alternative methods other than imaging spectrometers have been explored to measure the spectrum of the incident laser field. The related techniques are divided into two categories: multispectral and hyperspectral imaging methods. In multispectral imaging methods such as STRIPED-FISH and HAMSTER, a few spectral components of the incident femtosecond laser pulse are imaged and measured and the full spectrum is obtained through an interpolation scheme if the laser-pulse spectrum is simple. Figure 5 shows the principles and the experimental setup of the STRIPED-FISH technique as a typical multispectral imaging technique, and previous experimental results are shown. When the incident laser field has a complex spectral structure, hyperspectral imaging methods should be applied to resolve fine structures of the spectrum. Figure 6 shows the details of the typical hyperspectral imaging techniques, TERMITES and its derivative INSIGHTS, which adopt the principles of spatially resolved Fourier transform spectroscopy and measure the three-dimensional spatiotemporal amplitude and phase distributions of a femtosecond laser field in multiple shots. However, single-shot three-dimensional characterization of an optical field is still challenging.

Finally, we have briefly reviewed some techniques to determine the spatiotemporal coupling effect of a femtosecond laser pulse without obtaining the three-dimensional optical field distribution. These techniques are effective and simple in optical configurations, which are especially useful for pulse characterization of petawatt laser pulses in large laser facilities.

We expect to solve three problems in the near future. First, although the TERMITES and INSIGHT techniques can resolve the three-dimensional optical field information of a femtosecond laser pulse, single-shot measurement is yet to be achieved. Because of the development of the compressive sensing principle, an optical compressed imaging technique called CASSI (Fig. 7) provides a new idea of hyperspectrally measuring the spatial distribution of each spectral component in a single shot. However, CASSI can only obtain the intensity information of a laser pulse, leaving the phase distribution unknown. Based on CASSI, we have also proposed two global three-dimensional phase retrieval schemes, and we expect that the three-dimensional optical field distribution of an arbitrary femtosecond laser field should be determined in a single shot. Preliminary experimental results have justified our proposals.

Second, the application of spatiotemporal pulse characterization in ultrafast pump-probe experiments has been proposed. The evolution information of the optical properties of the pump laser-excited matter is encoded in the amplitude and phase modulations of the probe laser pulse; thus, a thorough characterization of the probe pulse reveals the dynamics of the femtosecond laser-matter interaction. So far, most of the probe characterization techniques are the multispectral imaging techniques, such as the STAMP and FINOPA techniques (Fig. 8), realizing the temporal resolving capability by linearly mapping the chirped probe spectral components to different time delays.

Finally, we have extended the femtosecond laser characterization techniques to the measurement of high harmonic generation. High harmonic generation is a type of laser-like coherent light source in the extreme ultraviolet or soft X-ray spectral range with small divergence, excellent spatial and temporal coherence, high brightness, and femtosecond or attosecond pulse duration. After reviewing the spatially integrated time-domain measurement techniques such as autocorrelation, RABBITT, and CRAB (or attosecond streaking camera), we have discussed the in situ measurement techniques for obtaining the spatiotemporal information of high harmonic radiation or isolated attosecond pulses (Fig. 9).

Femtosecond optical parametric oscillators (OPOs) are efficient and flexible wavelength-conversion devices capable of generating ultra-short optical pulses at wave regions not directly addressable by conventional laser gain medium. Over the past few decades, substantial progress has been made in femtosecond OPOs for generating emission with high output power, high pulse energy, high repetition rate, and wide wavelength-tuning range. Ultrashort-pulsed lasers with a broad bandwidth have unique advantages and are important for many emerging applications, including multicomponent trace gas detection, optical coherence tomography, and multispectral imaging. Many technical schemes focus on producing ultrashort optical pulses with a broad bandwidth, for instance, via laser mode locking, supercontinuum generation, or differential frequency generation. In contrast, femtosecond OPOs, which use nonlinear media as the gain materials, have inherent advantages in wavelength tuning and conversion efficiency. Moreover, extremely wide gain bandwidth of a nonlinear gain medium can be achieved through artificial tailoring with phase-matching techniques; therefore, femtosecond OPOs exhibit significant potential for generating ultrashort optical pulses with a broad bandwidth.

However, a femtosecond OPO is usually pumped by an ultrashort pulse train. To achieve an efficient parametric transfer in an OPO, the duration of the oscillating OPO pulses should be similar to that of the pump pulses. Indeed, in a synchronously pumped OPO, the duration of the pump pulse imposes a lower limit on the pulse width of the oscillating signal wave. Therefore, in traditional femtosecond OPOs, a pump source with very short pulse width is required for generating signal pulses with a broad bandwidth. The possibility of generating optical pulses with considerably wider bandwidths than the pump bandwidth in femtosecond OPOs is quite intriguing.

Intracavity spectral combinations can be used to enhance the instantaneous bandwidth of the output idler light in synchronous-pumped OPOs. Idler light with an instantaneous bandwidth that is significantly larger than that of the pump bandwidth can potentially be obtained using a nonlinear crystal with a wide phase-matching bandwidth and by implementing a multichannel configuration (Fig. 1). This scheme transforms conventional synchronously pumped OPOs into devices capable of generating idler light with a very broad instantaneous bandwidth.

Chirped-pulse optical parametric oscillators (CPOPOs) allow the generation of ultrashort optical pulses with an instantaneous bandwidth that is considerably wider than the pump bandwidth. Chirped-pulse formation can be achieved by inserting a material with a large nonlinear index coefficient into an OPO cavity or using an aperiodic QPM crystal. By properly managing the pulse dynamics by optimizing the intracavity dispersion and spectral broadening, optical pulses with an octave-spanning instantaneous bandwidth can be obtained.

Moreover, spectral broadening in CPOPOs can be enhanced to generate spectrum with an instantaneous bandwidth significantly wider than the parametric gain bandwidth of nonlinear crystals. Our study shows that a relatively high residual second-order-dispersion inside the OPO cavity is required for achieving the maximum signal bandwidth that exceeds the parametric gain bandwidth from a CPOPO (Figs. 2 and 3). In addition, the parametric deff2/n₂ of nonlinear crystals (n₂ and d_eff are the nonlinear refractive index and effective second-order nonlinear coefficient of the crystals, respectively) will play an important role in determining the amount of spectral broadening that can be achieved (Fig. 4). We demonstrated a CPOPO system that generated optical pulses with a bandwidth that was approximately 12 times of the gain bandwidth of the nonlinear gain medium (Fig. 5).

Pulse-compression in cavity length-detuned OPOs can be used to generate high-quality, soliton-like optical pulses with a considerably shorter duration than the pump pulse duration. This can be achieved in a femtosecond OPO via positive cavity-length detuning (Fig. 6). At a relatively high pump rate, the resonating signal wave can evolve into multiple pulses. However, we show that single-pulse operations can be recovered by simply increasing the level of cavity-length detuning.

In this article, we review recent progress on the generation of optical pulses with very broad bandwidths from femtosecond OPOs. We summarize the progress of our work performed at the Huazhong University of Science and Technology. Furthermore, we show that spectral broadening in OPOs can be achieved via intracavity spectral beam combination, by configuring a chirped pulsed OPO, or by simply detuning the cavity length. Ultrashort optical pulses with a broad instantaneous bandwidth may be beneficial for many applications, including optical coherence tomography, ultrashort pulse synthesis, and spectroscopy.

Around the world, high-power CO₂ laser has been the main light source for laser cutting, welding, and surface treatment, including cross-flow CO₂ laser, axial fast-flow CO₂ laser, and radio-frequency (RF) slab CO₂ laser. RF-excited diffusion-cooled slab CO₂ laser has compact structure and high beam quality, which once could not be achieved by all gas lasers above kilowatt level. It has an important application in the field of plate cutting and welding and represents the development direction of CO₂ gas laser. However, with the rapid development of fiber laser, the CO₂ laser metal processing market has almost been replaced by fiber laser. Yet, in recent years, RF slab CO₂ laser has been the only one used for laser annealing of VLSI wafers, with no other laser to replace it.

Today, two types of annealing equipment are mainly used in the annealing of integrated circuit wafers. One is the traditional rapid thermal processing (RTP) equipment. The RTP equipment uses halogen tungsten lamps to heat a single wafer to 300 ℃-1050 ℃ within 1-30 s. In addition, it has been adopted by semiconductor manufacturing industry for more than 20 years. Another novel heat treatment method is millisecond annealing (MSA), which can heat the wafer to 1100 ℃-1350 ℃ (just below the melting point of silicon) in a few hundred microseconds to a few milliseconds.

MSA can be realized using two different methods: laser spike annealing (LSA) or flash laser annealing. LSA uses long-wavelength CO₂ laser beam to irradiate the semiconductor wafer at grazing angle to form a "line beam" on the wafer surface to scan the wafer back and forth. For the formation of USJ (ultra shallow junction) and nickel silicide, each method has its own challenges.

With the reduction of device size, the formation of nickel silicide is a new challenge in wafer manufacturing. Conventional formation of nickel silicide includes two low-temperature RTP steps and an optional etching step between the two steps. When the device size is reduced to 28 nm or even smaller, the nickel diffusion in the second RTP processing step will result in junction leakage and reduce the yield. A feasible solution is to replace the second RTP treatment step with MSA because the MSA treatment time is short, which can reduce the nickel diffusion. The special laser annealing device adopts short-wavelength diode laser, high-power RF CO₂ laser, and two linear spots, and the incident light is close to the vertical incidence. To remove the lattice defects and obtain a silicon wafer with perfect electrical characteristics, the temperature should be controlled to avoid overheating during the laser annealing and the temperature uniformity should be ensured, which sets extremely strict requirements for laser stability.

This paper introduced the development of RF slab CO₂ laser. It mainly focused on key technologies such as laser structure principle, area amplification, diffusion cooling, strip electrode, thermal effect, power extraction, output beam characteristics, unstable-waveguide hybrid cavity, beam shaping, RF transmission and impedance matching, RF discharge plasma, and uniformity of gas discharge.

For example, in the aspect of electrode cooling and diffusion, the electrode deformation seriously affected the discharge power injection and laser power extraction due to the large discharge electrode area and small discharge spacing. When the laser worked, the heat of the gas diffused to the electrode and was taken away by the circulating cooling water in the polar plate. The electrode length, width, and height of the 2.5 kW laser were about 1000, 200, and 40 mm, respectively. Matching inductors were inserted at the outer ends of the upper plate in the length direction to realize impedance matching and ensure uniform and stable discharge. According to the heat transfer theory, the finite element model of the electrode was established by using the creative serpentine water-cooling channel design, and its temperature field distribution cloud chart was obtained by loading and solving. The highest temperature of the whole electrode was 31.994 ℃, and the lowest temperature was 23.333 ℃, which achieved good diffusion cooling effect(Fig. 6).

The goals of developing high-power RF CO₂ laser were as follows: (1) to solve the problem of localization of high-power, high beam quality, and high-stability laser that restricts the laser annealing equipment of VLSI wafers in China; (2) to break through the key technologies such as RF slab electrode, diffusion cooling, optical resonator, RF discharge excitation, beam shaping, etc.; (3) to realize the engineering of high-power diffusion cooling slab CO₂ laser; (4) to improve the localization level of high-end industrial gas lasers in China; and (5) to reverse the situation that the RF slab laser needed for laser annealing of integrated circuits in China is completely dependent on imports.

Currently, the growing bandwidth demands of cloud services are accelerating the hyperscale expansion of data centers. 850 nm vertical-cavity surface-emitting lasers (VCSELs) are widely used in data centers owing to their high-cost performance and high power. However, current commercial VCSELs operate in the multitransverse mode, and their applications are limited to the short connections of a few hundred meters. Owing to reduced mode partition noise and modal dispersion, single transverse-mode VCSELs can transmit over long distances in multimode fiber. Furthermore, in multitransverse-mode VCSELs, high-order modes consume a large fraction of the injection current, resulting in low relaxation oscillation frequencies. Therefore, 850 nm single-mode surface-emitting lasers are an urgent requirement in emerging hyperscale data centers to provide cost-effective, high-bandwidth, and long-distance optical communication systems. This paper presents an 850 nm single-mode surface-emission distributed feedback (SEDFB) laser with a rectangular oxide aperture and a shallow etched surface grating. A threshold current of 1.8 mA and a side-mode rejection ratio (SMSR) of 47 dB are achieved.

A first-order grating is used to provide adequate optical feedback for the laser. A second-order grating is used to achieve upward diffracted light. A λ/4 phase-shift structure is used to achieve stable single longitudinal mode lasing. Moreover, the simulation calculation is used to optimize the material and thickness of each layer of the waveguide structure and the coupling coefficient of the grating while ensuring the light confinement factor in the active region. Large-area rectangular oxide apertures are designed to confine current injection. Furthermore, the laser chip is epitaxially grown using metal-organic chemical vapor deposition (MOCVD). Wet etching with a citric acid-hydrogen peroxide solution is used to etch surface GaAs. The reactive coupled plasma (ICP) is used to etch the waveguides and gratings. Using a high-temperature wet oxidation method, a rectangular oxidation aperture is formed.

A low threshold current of 1.8 mA is obtained for the SEDFB laser with an active region length of 50 μm (Fig. 11). Its differential resistance is 46 Ω, which is smaller than that of VCSELs with the same oxidized aperture area. The laser exhibits good single-mode characteristics, resulting in an SMSR of approximately 47 dB (Fig. 11). The laser temperature increases from 20 ℃ to 60 ℃, and the SMSR remains above 40 dB (Fig. 12). The laser’s calculated thermal resistance is approximately 0.74 ℃/mW, which is smaller than the thermal resistance of a VCSEL with the same oxidized aperture area. The full width at half maximum of far-field pattern divergence angle is approximately 21°×26° (Fig. 13), and the quasicircular spot output is achieved. Simultaneously, the laser exhibits a relaxation oscillation frequency of 17 GHz at five times the threshold current (Fig. 14). An SEDFB laser with an active region length of 150 μm exhibits a 3 dB modulation bandwidth of 15 GHz.

In summary, this paper presents an 850 nm single-mode surface-emitting distributed feedback laser. It exhibits a threshold current of 1.8 mA, and the low threshold current shows our device’s application potential in energy-saving systems. To considerably reduce series resistance, we use a large-area rectangular oxide aperture. The SEDFB laser has a single transverse mode that is confined by the ridge waveguide structure and a single longitudinal mode that is selected by the surface grating. Therefore, the SEDFB laser exhibits a high SMSR of 47 dB. The relaxation oscillation frequency of the SEDFB laser is approximately 17 GHz.

Nanomaterials with small features and large surface-to-volume ratios have drawn tremendous research attention in various fields including energy devices, microelectronics, and biomedicine. By far, researchers have realized high-quality fabrication of various nanomaterials through solid-phase, liquid-phase, or vapor-phase method. However, the fabrication of nanomaterial-based functional devices usually requires subsequent material transfer and assembly processes. Therefore, to effectively realize the integration of nanomaterials and make full use of their unique properties, the transfer-free growth of patterned nanomaterials is very important.

Although methods have been developed to realize the in-situ transfer-free patterned growth of nanomaterials, such as ultraviolet lithography, electron beam lithography, solution-based direct-patterning technique, and continuous wave/long pulsed laser selective induction, it is still difficult to meet the demands of customized patterning, precise processing, and in-situ heterogeneous integration of nanomaterials on thermal-sensitive, flexible, and curved substrates. The UV lithography and electron beam lithography techniques are cumbersome, time-consuming, and usually need a vacuum chamber. Besides, they are difficult to apply to curved substrates. The solution-based direct-patterning technique requires the subsequent high-temperature annealing process, which is difficult to apply to thermal sensitive substrates. The CW/long pulsed laser selective induction method is difficult to achieve high precision and highly localized growth due to the diffraction limit effect and the sizeable heat-affected zone.

Due to these drawbacks of the existing methods, researchers have attempted to use a femtosecond laser to realize the direct patterned growth of nanomaterials. As a "cold processing" method with a high peak power, the femtosecond laser direct writing is a promising tool to achieve the direct patterned growth of nanomaterials. The focus of a femtosecond laser can be regarded as a flexible, controllable and highly localized micro-reactor, which can realize the fixed-point growth of nanomaterials. At the same time, according to the pre-designed patterns, the laser focus position can be changed by the galvanometer, displacement stage or other equipment to realize the transfer-free patterned growth of nanomaterials. Compared with the current commonly used CW or long pulsed laser, a femtosecond laser has unique advantages in the transfer-free patterned growth of nanomaterials. First, due to its small heat-affected zone, it can be applied to thermal-sensitive substrates. Second, the ultra-high energy density of a femtosecond laser can induce nonlinear multi-photon absorption of precursors, which can realize the direct absorption of laser energy. Therefore, the femtosecond laser induced direct patterned growth of nanomaterials can be applied to transparent substrates without heat-absorbing layers. Third, the threshold effect of nonlinear absorption and the small heat-affected zone of a femtosecond laser can realize the high-precision growth of nanomaterials. Thus, the femtosecond laser induced patterned growth of nanomaterials has unique advantages and excellent prospects.

In this review, we first summarize the commonly used patterned synthesis methods of nanomaterials and their problems, including UV/electron beam lithography, solution-based direct patterning, and CW/long pulsed laser induced growth of nanomaterials. Then we discuss the unique advantages of the femtosecond laser-induced patterned growth method of nanomaterials, including high precision, highly localized growth, and high processing compatibility with thermal sensitive and transparent substrates. Next, the recent progress of the femtosecond laser induced direct patterned growth of nanomaterials and their applications are reviewed, including metal, metal oxide, metal sulfide, and carbon-based nanomaterials. For metal materials, researchers realized silver and gold patterned micro-nano structures with high conductivity [Fig.4(a)], which are comparable to the bulk materials. To grow more high-precision products, researchers realized silver nanostructures with a minimum feature size of only 180 nm with the help of surfactant [Fig.4(c)]. Researchers realized a stable 3D connection between two pairs of metal electrodes. As for metal oxides, researchers realized the patterned SnO₂structure with the line width of about 150 nm through femtosecond laser direct writing (FLDW) and subsequent annealing process (Fig. 5). Our group realized the patterned growth of ZnO and SnO₂ through femtosecond laser direct writing without subsequent annealing (Figs. 8 and 9). The minimum linewidth is about 800 nm. For metal sulfide, our group realized the patterned growth of MoS₂ through femtosecond laser induced photochemical reaction (Fig. 10). For carbon-based nanomaterials, researchers realized the patterned growth of graphene through femtosecond laser induced reduction of graphene oxide [Fig.12(b)]. Researchers realized the patterned growth of graphene through FLDW of co-sputtering Ni/C films. The sheet resistivity of the products is about 205 Ω/sq [Fig.12(a)].

Compared with traditional methods, the femtosecond laser induced direct patterned growth technique has many unique advantages. Due to the extremely small heat-affected zone and the nonlinear multi-photon absorption effect of a femtosecond laser, the femtosecond laser induced direct patterned growth technique can realize the high precision, highly localized patterned growth of nanomaterials and has high processing compatibility with thermal sensitive and transparent substrates. Besides, the femtosecond laser induced direct patterned growth technique does not need a vacuum chamber or the high-temperature annealing process. Thus, it has drawn tremendous research attention around the world. Although the femtosecond laser induced direct patterned growth technique has made some progresses, several problems remain to be resolved. First, the products need to be expanded and the precursor needs to be optimized to reduce the required laser energy and take full use of the advantages of a femtosecond laser. Second, in term of the processing system, a Gaussian beam can be converted into a flat-top beam by beam shaping, thereby improving the uniformity of the products. The processing efficiency can be improved by employing scanning devices with high scanning frequency or adopting parallel processing strategies including multi-point scanning, line scanning , and plane projection. Finally, the application of this method needs to be explored, such as MEMS, soft electronics, metasurfaces, energy and catalytic devices.

Gas turbine is the most promising device for power generation and ship power in this century due to its high efficiency and low emission. The hollow turbine blade with a complex structure is the key part of a gas turbine. The working temperature of the turbine blade is very high, which requires high metal quality and elaborate structures of turbine blades. High gas temperature means high working efficiency. With the further increase of industrial demand, the gas temperature of a turbine reaches more than 1700 ℃. Therefore, the turbine blades with more delicate and complex hollow structures should be fabricated to improve the cooling efficiency. The ceramic cores and shells are important components for casting of superalloy turbine blades due to their high temperature capabilities and chemical inertness. The traditional methods to fabricate ceramic shells and cores are the investment casting method which is time-consuming, high cost, low yield, and not sufficient for the update-requirement of turbine blades. In order to overcome these problems, additive manufacturing (AM) has been gradually developed. The AM technology has been widely used in different fields such as medicine, engineering, and aerospace to fabricate delicate parts without any molds. It can save materials, accelerate the research of new products, and meet customized needs, thus greatly reducing fabrication costs. In the past few decades, dozens of AM technologies have been developed and each of them has its unique application fields. As for the fabrication of ceramic cores and shells, the most suitable laser-based AM technologies are selective laser sintering (SLS) and stereolithography apparatus (SLA).

Different from other AM technologies, the most significant advantage of SLS is that it doesn’t need any support during the fabrication process. Large-scale ceramic parts with precise structures can be fabricated by SLS. The preparation of ceramic cores and shells by the SLS process is shown in Fig. 3. First, the green bodies of ceramics are first built by the SLS equipment, and then the green bodies are infiltrated with silica sol or other materials which could fill the pores of ceramics to improve the density of green bodies. The organic binder in the green bodies can be removed through thermal decomposition at 600 ℃ and the green bodies are sintered at 1200-1600 ℃ to obtain the densified ceramic cores and shells. SLA has been known for several years as an AM technology to fabricate the polymer-based parts. This technology utilizes liquid photo-curable resin which can be cured under ultraviolet light or laser irradiation. At present, there are two main processes for preparing ceramic cores and shells by the SLA technology (Fig. 6). The first process combines SLA and gel-casting. A resin mold is first made by SLA and then the ceramic slurry is injected into the mold cavity to get the green body of ceramic core and shell. In the other process, the ceramic slurry and photo-curable resin are mixed and the ceramic green bodies are directly formed by SLA. Both of ceramic green bodies made by these two technologies have enough density so that the infiltration process is no longer required and the rest of the post-processing process including debinding and sintering are the same as that of SLS. Characteristics of different laser-based AM processes used for ceramic core and shell manufacturing are shown in Table 2. It can be concluded that both of two AM technologies have their own strengths and weaknesses. SLS has an advantage in manufacturing large size ceramic cores and shells, while the parts made by SLA have higher resolution.

In this review, the development of these two AM processes in the fabrication of ceramic cores and shells are introduced in terms of material preparation, green body fabrication, and post-processing. The advantages, weaknesses and future development of both two methods have been discussed.

The physical and chemical properties of materials are determined by their element compositions and contents. How to obtain the compositions and content information of materials quickly, accurately, and at low cost has always been the research direction of scholars. The existing methods for the analysis of elements in the materials can be divided into the chemical methods and the instrumental methods. Based on the law of chemical reactions, the chemical methods carry out the qualitative and quantitative systematic analysis on the chemical compositions of samples, including the gravimetric method, the volumetric method, and the colorimetric method. In contrast, the instrumental methods directly obtain the physical and chemical information of the unknown samples through the analytical instruments, such as the inductively coupled plasma mass spectrometry, the Raman spectrum, the near-infrared spectrum, the X-ray spectrometry, and the atomic absorption spectrometry. The above-mentioned methods can obtain the categories and composition information of the samples with high sensitivity and accuracy, but they present complex operation, high cost, and low efficiency. However, with the continuous expansion of application fields, there is a high demand for the analysis technologies. Looking for a newer, faster, and more adaptive detection technology has become a research hotspot.

Laser induced breakdown spectroscopy (LIBS) is an element analysis technology, which uses a laser as the excitation source to ablate the sample and produce plasma. The emission spectrum of the plasma is then detected by a spectrometer to obtain the element category and the content information of the sample to be measured. Compared with other analytical technologies, the LIBS technology has the unique advantages of simultaneous detection of multiple elements, simple structure, fast detection speed, and being not affected by sample morphology. It shows great application prospects in many fields. Based on this, the mechanism, device types, basic research progress, and applications of LIBS are summarized.

The plasma characteristics and self-absorption effect of the LIBS instruments raise the most concerns. By studying the relevant characteristics of plasma, it is helpful to understand the generation mechanism of laser-induced plasma and solve the relevant problems encountered in the LIBS analysis. The self-absorption infulences the linear relationship between the original plasma emission spectral intensity and the concentration of related elements, and thus it reduces the accuracy of a quantitative analysis. To satisfy different analytical requirements, variable types of LIBS instruments are developed, including an LIBS in the laboratory (Fig. 3), a stand-off LIBS (Fig. 4), an on-line LIBS (Fig. 5), and a portable LIBS (Fig. 6). The LIBS in the laboratory has higher sensitivity and reproducibility, which is often used in the study of mechanism and exploratory applications. The stand-off LIBS can realize an in-situ detection of dangerous samples under harsh conditions on the premise of ensuring personnel safety. With the unique advantages of in-situ detection, real-time, fast, and no complex sample pretreatment, the on-line LIBS can quickly process numerous samples on the production line. The portable LIBS has the advantages of small volume, light weight, and convenient use, which has better applicability in industrial fields with harsh conditions.

To improve the analytical performance of the LIBS technology, the signal enhancement methods and the methods for the qualitative and quantitative analysis have become the focus study. The signal enhancement methods mainly contain surface enhancement methods (Fig. 7), inert-gas protection enhancement methods (Fig. 9), confinement enhancement methods (Fig. 10), and double-pulse enhancement methods (Fig. 11). The surface enhancement method ablate the substrate and the sample to be measured at the same time. The high-temperature plasma generated by the substrate heats the sample which can improve the temperature and electron number density of the sample plasma. Using inert gas as ambient gas can prolong the life of luminous atoms in plasma and avoid the light signal from being absorbed by air. Confinement enhancement uses the confinement cavity or magnetic field to affect the external and internal conditions of the plasma and confine the plasma to achieve signal enhancement with the advantages of simplicity, economy, and high feasibility. The double-pulse technology uses the second laser pulse to excite and heat the plasma again, which can greatly increase the temperature of the plasma and enhance the spectral intensity. Various methods are carried out for the qualitative and quantitative analysis, including material identification, element detection, and quantitative analysis.

With the specific advantages of the LIBS technology and the development of above-mentioned methods, the LIBS technology has been successfully used in various fields, including space exploration (Fig. 14), geological prospection (Fig. 15), pollution monitoring (Fig. 16), food safety (Fig. 17), industrial metallurgy (Fig. 18), and biomedicine (Fig. 19). The rapid identification of sample category is the focus of current research, and good analytical results have been obtained. However, due to the change of experimental environments, surface dirt of samples, and the diversity of manufacturing processes and additives, the prediction accuracy of the LIBS technology for real samples is still low. Outlier screening, variable selection, scale transformation, and other spectral preprocessing methods, as well as the improvement and integration of algorithms, are effective ways to solve this problem. Due to the matrix effect, laser energy fluctuation, spectrometer resolution difference, detection environment limitation, and other reasons, the LIBS technology has a large deviation in the prediction of element contents in materials. Optimizing the LIBS instrument platform, studying the signal enhancement methods, and improving the analysis methods are the effective methods to improve the prediction accuracy of the quantitative analysis.

The matrix effect is the most critical problem that limits the wide application of the LIBS technology. With the continuous development of the LIBS instruments and the corresponding components, this problem can be effectively solved, but it will take a long time. The improvement of the analytical chemistry method will be an effective way to improve the application performance of the LIBS technology. In order to realize the rapid and sensitive detection of massive materials, the development of an on-line LIBS device will be the development trend in the future.

The automobile industry is a "machine to change the world" and an important pillar industry for promoting national economic development. The automobile industry is heavily invested in high-tech and high-end equipment, which reflects the national manufacturing technological level. Every automobile is a crystallization of modern high technologies. The automobile industry is the largest user of robots, computer numerical control machine tools, and automatic production lines. Modern automobiles also make extensive use of novel materials, processes, equipment, and electronic technologies. The popularity of automobiles has met the people’s enormous demand for travel, so the quality of automobiles has a direct impact on the safety of use. In the automobile manufacturing process, the value of car body accounts for approximately 1/5 of that of the whole car, and the weight of car body accounts for 1/3 of that of the whole care. As a result, the manufacturing quality of car body is directly related to the overall safety and comfort of the vehicle. At the same time, the choice of car body materials is critical in the lightweight development of a car. Stamping, welding, painting, and final assembly are all parts of the automobile body manufacturing process. The welding process, for example, is used to create a body-in-white by welding stamping sheet metal parts together. Poor welding quality can cause deformation and cracking of body sheet metal and abnormal noise, and even endanger the personal safety of passengers.

Steel has been traditionally used as the body material of automobiles. Because the steel plates that make up the body are generally thin, the welding of the body is primarily resistance spot welding, which is widely used in the welding of underbody, side wall, frame, roof, door, and body assembly with as many as 4000-6000 welding spots. In addition, CO₂ gas shielded welding, stud welding, arc welding, brazing, and other technological methods are used in the automobile body manufacturing process. Traditional welding technologies for automobile body can essentially meet the quality requirements of an automobile body after welding. As the quantity of produced automobiles grows, the demands for high automobile body manufacturing efficiency increase. Different automobile body parts and welding joint forms have higher requirements for the flexible automobile body manufacturing. Special welding requirements, such as lightweight material welding and dissimilar material welding, have higher requirements for the welding process. It can be seen that high-quality and efficient welding of automobile body is the development trend, and the traditional welding process struggles to meet this demand.

Laser welding technology, as an advanced forming technologyin opto-mechatronics, has many advantages including high energy density, fast welding speed, low welding deformation, and good flexibility. It has become increasingly popular in the welding of automobile body in recent years. Laser welding is a fast and precise welding method that uses a high energy density laser beam as a heat source. The use of laser welding for automobile body has obvious advantages. For example, there is no mechanical contact between the welding device and the weldment, which reduces pollution to the workpiece. Because the heat energy of the laser beam is concentrated, the heat affected zone is small and the thermal deformation and damage are weak. Laser welding produces a beautiful welding seam with excellent mechanical properties. Welding robots and numerical control systems allow for a precise control of energy output, fast welding speed, and high production efficiency. As a result, the laser welding technology can not only improve the precision and efficiency in the car body process, but also improve the rigidity and strength of the car body, and thus the vehicle driving comfort and safety are improved.

The laser welding technology for automobile body is divided into two categories: the laser welding technology and the laser welding intelligent technology. Aiming at the laser welding process of automobile body, the structures of the commonly used materials are summarized (Table 1). The characteristics of the most commonly used laser welding processes for automobile body as well as information on welding parts, welding forms, and welding materials applicable to automobile body are also summarized (Table 2). Then, the commonly used laser deep penetration welding, laser filler welding, laser brazing, and laser-arc hybrid welding processes for automobile body are introduced, with a focus on the principles of these four laser welding processes. In conjunction with the welding characteristics of automobile body, the research progress of various welding processes for automobile body is expounded and summarized. In addition, the new laser welding process is described. These processes primarily include laser spot welding, laser wobble welding, multi-laser beam welding, and remote laser welding. This paper discusses the intelligent laser welding technology for automobile body from two perspectives: welding seam tracking and defect online detection. The welding seam tracking technology employs an advanced vision sensor to identify and track the welding seam of an automobile body before and during welding, and it corrects the movement path of the welding robot in real time to ensure welding stability. For example, the welding seam tracking system developed by Scansonic Company in Germany can achieve a dynamic and accurate identification of welding seams (Fig. 13). The key to detecting weld defects is to create a correlation model between defects and monitoring signals. Ma et al. have developed a correlation model between multivariable signals and porosity defects using visual sensing and a keyhole depth measuring device, and accurately identified the blowhole defects (Fig. 14). Finally, this paper summarizes the current problems in the laser welding process and the intelligent technology, and the future development trend is prospected.

The laser welding technology has become an indispensable and important technology in the development of lightweight automobile body. The laser welding process for lightweight body materials and dissimilar materials still requires extensive research and exploration. At the same time, an intelligent welding system with weld tracking, weld defect detection, and closed-loop control of welding parameters is required.