The defects, bubbles, and dirty spots in the optical elements of a laser interferometer introduce coherent noise similar to the "bull's eye ring" into the interferogram. To solve this problem, a coherent noise suppression method based on an extended light source of multimode fibers is proposed. The proposed method uses the extended light source to suppress coherent noise and selects the optimal multimode fiber core to ensure a favorable interference fringe contrast. Then, a composite speckle suppression technique is employed, and the speckle contrast is reduced by introducing the multimode fibers and rotating ground glass to suppress the speckle noise generated by the mode interference of the multimode fibers. Simulation analysis and experimental verification are carried out in an interferometer with a diameter of 300 mm. The following observations can be made from the results. Under the conditions of a vertical planar Fizeau interferometer with a diameter of 300 mm, a cavity length of 500 mm, and a camera exposure time of 5 ms, an interference fringe contrast higher than 0.75 can be ensured by keeping the core diameter of the multimode fibers in the proposed light source system between 0.52 mm and 1.70 mm. The speckle contrast of the final image obtained by the interferometer is maintained at about 0.04, with a minimum of 0.044. The extended light source of multimode fibers further improves the uniformity of the light source, contributing to a more uniform imaging background in the interferometer. The experiment proves that the proposed method can effectively suppress the coherence noise in the interferogram.

A novel plasmonic D-shaped optical fiber sensor with a graphene-metal nanowire array structure was designed. The refractive index sensitivity was improved on the core by using the localized surface plasmon resonance (LSPR) produced by graphene and metal nanowires. The full-vector finite-element method was utilized to investigate the dispersion relationship between the plasmonic mode and the core mode of the structure. The effects of the thickness of the metal sensing layer and the diameter of the metal nanowires on the sensor's performance were also examined. Finally, the sensitivity of three types of sensors with different structures (gold film structure, graphene-gold film-graphene structure, and graphene-metal nanowire array structure) was compared. The results reveal that adopting the optimized structural parameters considerably improves the sensitivity of the developed sensor based on the graphene-metal nanowire array structure. Specifically, when the refractive index is between 1.33 and 1.40, the design structure obtains a maximum sensitivity of 7383.79 nm/RIU and an average sensitivity of 4136.00 nm/RIU. The findings of this study serve as a theoretical foundation for the development of next-generation plasmonic optical fiber sensors.

A methane gas detection device based on anti-resonant hollow-core fibers is presented. A distributed feedback laser with a center wavelength of 1653 nm is combined with anti-resonant hollow-core fibers with a length of 1.8 m. The methane gas is detected in real time with a technique integrating tunable diode laser absorption spectroscopy with wavelength-modulation spectroscopy. A fiber coupling device of a ceramic ferrule and a ceramic sleeve is used to achieve stable butt-coupling between hollow-core fibers and single-mode fibers. Then, a lensless all-fiber gas detection device is built, and experiments are carried out on methane gas with different concentration gradients. The results show that there is a good linear relationship between the peak-to-peak value of the second harmonic signal and the gas concentration, and the linear correlation coefficient is R2=0.997. The accuracy of the volume fraction of the methane gas obtained by the wavelength modulation technique through inversion is 0.918×10-6, and the relative accuracy is 1.3%. The stability of the system is evaluated by the Allan variance, and the lowest detection sensitivity of the device is 13×10-9 when the averaging time is 66.8 s.

By using the two technologies of frequency division multiplexing and in-phase/quadrature (I/Q) receiving, this paper proposes a distributed optical fiber vibration sensing demodulation scheme with high spatial resolution and large bandwidth. Theoretical analysis and numerical simulation are carried out. This distributed vibration sensing structure is composed of a Mach-Zehnder interferometer (MZI) and a time-gated digital optical frequency-domain reflectometry (TGD-OFDR). Specifically, the phase of the MZI output signal is demodulated by the homodyne method to detect the frequency and amplitude of the vibration signal, whereas TGD-OFDR enables the positioning of the vibration signal through heterodyne detection. The numerical simulation results show that the upper limit of the vibration frequency which the proposed system can detect reaches the order of magnitude of MHz, and the spatial resolution on the 4-km sensing optical fiber is up to 0.5 m. This distributed vibration sensing system, with the advantages of high spatial resolution and large detection bandwidth, has application potential in civil infrastructure health monitoring and oil and gas pipeline leakage monitoring.

Upon the analysis of the existing generation methods for the optical spectrally efficient frequency-division multiplexing (O-SEFDM) system, a generation method based on the modified universal inverse fractional Fourier transform (uIFrFT) is proposed. On the basis of uIFrFT with P symmetry, Hermitian symmetry can be applied to generate real-valued signals. Therefore, the up-conversion in the intensity modulation/direct detection (IM/DD) O-SEFDM system is avoided, which simplifies the system. Meanwhile, a suitable spherical part detector (SPD) is designed in this scheme. The simulations demonstrate that the generation method based on uIFrFT outperforms the existing generation methods in flexibility and performance for the IM/DD O-SEFDM system. Furthermore, SPD has improved the bit error rate (BER) performance and reduced complexity by about 51% compared to the conventional fixed spherical detector (FSD) in fourth-order quadrature amplitude modulation (4QAM) and 16QAM transmission systems.

This paper proposes a deep convolutional neural network with an encoder-to-decoder structure and constrains the network's in-depth learning from the monocular image at both two-dimensional (2D) and three-dimensional (3D) levels. At the 2D image level, an attention mechanism of channels is introduced to connect encoder features with decoder features with weights at the same scale, so as to balance the shallow detail features and deep semantic features extracted by the network. In addition, a scale-invariant loss and a multi-scale edge loss based on image pyramids are designed to obtain a depth map with rich edge detail information. At the 3D geometric level, a global geometric constraint loss and a local geometric constraint loss of depth are designed based on the local and global geometric relationships of coordinate points in space, in a bid to enhance the geometric consistency between point clouds. Furthermore, the results obtained through the proposed method are quantitatively and qualitatively compared with that obtained through other methods from the NYU Depth-v2 dataset, and it is shown that the proposed method can estimate indoor scene depth with higher accuracy and detail representation, obtaining accurate and smooth 3D reconstruction results on a single image.

The on-board head-up display (HUD) system produces virtual images through the reflection of the windshield. Due to the aberration of the optical system, actual virtual images are not distributed in an ideal planar pattern. Existing measurement methods for HUD virtual images treat such images as planar ones perpendicular to the optical axis, which is obviously inconsistent with the actual situation. To understand the characteristics of HUD virtual images, this paper proposes a method of determining the three-dimensional (3D) coordinates of the HUD virtual image points within the eyebox range by performing ray tracing, so as to solve the position of the minimum circle of confusion. HUD virtual image shapes observed at different positions are thereby obtained. Then, an HUD virtual image measurement model based on binocular vision is further developed, and the effects of binocular baseline and camera parameters on the measurement error are analyzed. Finally, experiments are carried out to measure the 3D coordinates of actual HUD virtual images. The measured virtual image shapes are basically consistent with the simulation results, with an overall average relative error of 2.5%. The proposed method can be further applied to the 3D coordinate measurement of virtual images produced by HUD systems such as augmented-reality HUD systems.

In the iterative reconstruction method for optical transfer function (OTF) measurement, several images are captured and taken into update computation, which offers this method strong anti-noise robustness and anti-aliasing. However, in image acquisition, the inevitable vibration will cause a small relative translation between multi-frame images, which results in an increase in the number of iterations and a decrease in the measurement accuracy of the modulation transfer function (MTF). Moreover, the images captured by this method are randomly distributed point spread functions (PSFs), and the relative translation between multi-frame images cannot be calculated by the phase correlation method. In view of this, a calibration method for relative translation between multi-frame images is proposed, which is based on the phase transfer function (PTF). The PTF calculated by a single iteration of the first image is subtracted from the PTF calculated by a single iteration of each image, and the relative translation between multi-frame images is calibrated according to the difference. The numerical simulation and experimental results reveal that the calculation accuracy of the proposed method is independent of the relative translation between images, and the proposed method can reduce the number of iterations by up to 80% and the mean square error of MTF measurement by one to two orders of magnitude.

This paper established an eccentric error mathematics model based on analysis of the effect of concentric circle eccentricity error caused by perspective projection, and put forward a high-precision camera calibration method based on concentric eccentric error compensation iteration. The method used optimization algorithm based on geometric constraint to constantly update the coordinates of the center projection of concentric circles, so as to achieve high precision camera calibration. Experiments and simulations verify the effectiveness of the proposed method in visual applications. The results show that the proposed method can provide higher precision calibration results for computer vision and three-dimensional reconstruction tasks compared with traditional methods.

In the three-dimensional measurement methods based on fringe projection, circular sinusoidal fringes have higher phase sensitivity than linear fringes. Therefore, a coaxial three-dimensional measurement method of circular fringe projection is proposed, which is based on the geometric relationship of divergent projection rays. Specifically, the fringes on the object surface are changed by the movement of the projector along the optical axis, the charge-coupled device (CCD) is employed to collect the corresponding fringe patterns, and different phase fields are calculated from these fringe patterns. The constraints between phase information and height of the object are established by the use of the change in projection rays caused by the movement of the projector. In this way, the height information of the object surface can be reconstructed without obtaining complex system parameters, and the problems of shadow and occlusion can be reduced in the measurement. Simulations and experimental results verify the effectiveness and practicability of the proposed method.

Movement parallelism is one of the key parameters of motorized stages, and it can directly affect the performance of the stages. A spot image method for accurately measuring the movement parallelism is proposed. In addition, based on theoretical research and analysis, a measurement system with a precision better than 50 nm is set up. Utilizing this measurement system, movement parallelism of the motorized stage is measured, and the movement parallelism error is 11.66 μm. After optimizing the motorized stage according to the above results, the optimal movement parallelism error can reach 6.22 μm. In order to verify the feasibility of the proposed method, a displacement sensor is applied to re-measure the movement parallelism. The root-mean-square error between the measurement results of the displacement sensor and that of spot image method is lower than 248 nm. In other words, the two measurement results are basically the same.

A simultaneous particle size and position prediction model based on Faster-RCNN and VGG16 convolutional neural networks (CNNs) is constructed for the defocused images of particles obtained by a dual-camera imaging system. Nine different dots with diameters ranging from 50 to 350 μm are taken in a depth range of 75 to 95 mm (about 9 to 10 times the depth of field of the imaging system) for the training of the proposed model, and the proposed model is compared with the processing method based on the depth from defocus (DFD) model. The measurement results show that compared with the processing method based on the DFD model, the particle depth measurement range of the CNN model is improved, the diameter measurement error is reduced, and the depth measurement error is increased. The standard particles with a particle size of 120 μm flowing in a circulating sample cell are further photographed by a dual-camera system, and the images are processed by applying the proposed CNN model. The relative error of the particle size prediction results ranges from -8% to 8%.

In the laser direct writing lithography system, the focus locking performance is one of the key factors affecting the exposure uniformity of the system. The dynamic response characteristics of a piezoelectric ceramic as the focus locking actuator of the lithography system are studied, and their influences on the exposure uniformity of the polar coordinate lithography system are explored. In the process of exploration, the nonlinear relationship between the dynamic response of piezoelectric ceramics to general anharmonic signals and that to characteristic harmonic signals is found, and the correctness of the relationship is verified by sample exposure experiments. Based on the relationship, the fastest working speed of the rotary table that can meet the focus locking requirements is quantitatively calculated. At the same time, the guidance for the piezoelectric ceramic manufacturers to optimize the dynamic response characteristics of the piezoelectric ceramics suitable for the polar coordinate lithography system is provided. The focus locking ability of the optimized piezoelectric ceramics is increased by 125.2%.

Stimulated Brillouin scattering is an effective way to achieve a laser with low background noise due to its narrowband gain. Based on a high Q fiber ring resonator, a Brillouin laser with low noise is studied. The pump light is locked into a single-mode ring resonator with a length of 8 m by the Pound-Drever-Hall (PDH) locking technique, and a backward Stokes light with a frequency difference of 10.81 GHz from the pump light is obtained. The frequency noise of the Stokes light is measured by the correlative delay self-heterodyne method. The experimental results show that the threshold of the Brillouin laser based on the fiber ring resonator is 5.3 mW. In the high-frequency part (the frequencyis greater than 10 kHz), the backward Stokes light suppresses the frequency noise of the pump light up to 30 dB, which is close to the theoretical suppression limit (34 dB).

An all-solid-state single-longitudinal-mode passively Q-switched laser in a ring cavity is demonstrated. The laser diode (LD) side-pumped ring cavity is adopted to eliminate the spatial hole burning effect to stabilize the number of longitudinal modes in the resonator, and the etalon is used to control the net gain difference between adjacent longitudinal modes of the laser to realize the operation of a high single-longitudinal-mode ratio laser. The laser operates at a repetition rate of 10 Hz, producing pulses with a pulse width of 23.6 ns and a pulse energy of 6.1 mJ. The laser has strong stability in output energy and single-longitudinal-mode ratio, a relative standard deviation of 1.56% is demonstrated, and a single-longitudinal-mode ratio of 100% is achieved when 10000 pulses are recorded continuously.

Quasi-periodic nanostructures induced by ion bombardment (IB) on solid surfaces are characterized by small periods (10-100 nm) and large areas. Quasi-periodic nanoripple structures with the transverse feature size of around 100 nm and the gradually significant transverse periodicity and longitudinal continuity were fabricated on antireflection coatings by Argon-IB. To improve the characterization area, the morphological characteristics of the self-organized nanoripples were characterized by using extreme ultraviolet (EUV) scatterometry. The results show that in terms of samples, their transverse and longitudinal morphological features obtained by the in-plane and conical mode of the EUV scatterometry are in agreement with those obtained by atomic force microscope. These results demonstrate that the proposed method is feasible to characterize the basic morphological characteristics of quasi-periodic nanoripples and can provide a basis for subsequent quantitative analysis. In addition, the characterization area of self-organized nanoripple structures has reached an order of the mm2 by EUV synchrotron radiation, and the characterization range of the Metrology Beamline of Hefei Light Source is extended to self-organized nanostructures, which can provide a reference for future studies on the scattering characterization of EUV lithography masks.

The existing three-dimensional (3D) human hand pose estimation algorithms do not fully exploit the characteristics of fingers and the key features. To solve this problem, a finger-point reinforcement (FPR) strategy and a multi-layer fusion squeeze and excitation (MFSE) block are proposed. The FPR strategy highlights the role of the finger position points in the human hand point cloud, strengthens the attention of network feature extraction layers to the finger position points in the point cloud, and improves the regression accuracy of the finger joint points. The MFSE block improves the ability of the layered network to extract and express local features. This module realizes the fusion and weight distribution of different levels of features between the layered networks, thereby enhancing the robustness of the model and the accuracy of human hand pose estimation. Experiments on two public benchmark datasets, MSRA and ICVL, verify that the proposed algorithm can achieve high-precision 3D human hand pose estimation.

This paper presents a dynamically tunable narrow/broad band switchable THz absorber. The absorber is composed of tunable materials, i.e., graphene and VO2, and gold and cycloolefin copolymers. When VO2 is thermally induced to achieve insulator-to-metal transition, the absorber switches between broad bands and narrow bands. Its switching amplitude can reach 98.9%. When VO2 is in metallic phase, the absorber has a wide absorption bandwidth. At that time, the bandwidth of the absorber can be adjusted by changing the Fermi level of graphene. The absorber also has characteristic of wide-angle absorption for transverse electric (TE) and transverse magnetic (TM) waves. When VO2 is in insulating state, the device switches to a narrowband absorber with a sensitivity of 439 GHz/RIU, which can be used in sensor. The THz absorber has many potential applications in multi-functional devices such as modulation, sensing, and electromagnetic cloaking.

Phase change materials have good photothermal stability and are rewritable. As their optical properties change significantly after phase change, and the crystallization states of phase change materials can be precisely controlled by applying laser excitation with different parameters. Therefore, on the basis of different optical parameters of phase change materials in different crystallization states, a thin-film flat lens with a multi-order refractive index design is proposed. The typical phase change material Ge2Sb2Te5 is used as the optical modulation medium, and the crystallization state of the dielectric film is controlled discretely in multi-order and multi-region according to the phase constraints required by the focusing lens. Flat lenses with different numerical apertures are designed, and the focusing parameters of lenses are simulated by the finite-difference time-domain method and Zemax separately to verify the imaging performance of the lenses.

A new-type color filter has been designed using an all-dielectric ring-nanorod structure. The interaction of incident light with the all-dielectric ring-nanorod array structure is used to activate the Mie resonance, which has high reflection capabilities in the visible light band, allowing for a wide gamut of distinct colors. Using the finite-difference time-domain (FDTD) method, we compare the reflectance spectra and color display laws of three filters with other structures (nanorods, nanorings, and silicon ring-nanorods). Simultaneously, the impact of key parameters such as nanorod diameter, ring diameter, height, and period on the reflection spectrum and color characteristics is examined. According to the findings, the developed all-dielectric ring-nanorod structure filter considerably improves the reflection properties of the specific light waveband; the optimum color filter has a visible light reflectivity of more than 70%. The color gamut area of the nanorods can reach 0.115, and the diameter of the nanorods can be modified to achieve wide gamut color filtering properties. This study provides a theoretical foundation for the development of next-generation color filters.

A directional coupler (DC)-based polarization-independent demultiplexer filled with SiNx is designed to separate the 1310-nm and 1550-nm optical signals. The plasma-enhanced chemical vapor deposition (PECVD) method is used to adjust the refractive index of the SiNx material filled in the gap between the DC waveguides. As a result, the coupling length of the transverse-electric (TE) polarization mode equals that of the transverse-magnetic (TM) polarization mode at the same wavelength, and the polarization-independent function of the device is thereby fulfilled. The ratio of the coupling lengths corresponding to the two optical signals with different wavelengths is adjusted by optimizing the gap between the waveguides. When a proper value of the coupling length ratio is chosen, the two optical signals can be output from two ports, respectively, to achieve the wavelength separation function. Modeling and simulation are conducted by the three-dimensional finite-difference time-domain method to optimize the parameters of the device and analyze its performance. The results show that the proposed demultiplexer achieves a coupling region as short as 22.8 μm and an insertion loss and a crosstalk (CT) as low as 0.05 dB and -21.58 dB, respectively. Besides, the bandwidth corresponding to a CT smaller than -10 dB reaches 79 nm, and the device offers favorable tolerance on the whole. The device designed has application potential in future integrated optical circuit systems.

In order to further improve the luminous efficiency of GaN-based light-emitting diodes (LEDs), a novel electrode structure with an interdigitated electrode and electrode holes etched under the P/N electrode is designed and fabricated, with the improvement of the electrode structure as the research point. In this structure, the metal electrodes are in direct contact with the ITO and the N-GaN layers at the P/N electrode holes respectively, so as to improve the current spreading capacity and luminous efficiency of the device. In order to obtain better current blocking layer (CBL) structure, electrode hole size and electrode hole spacing, seven different devices are designed, and their photoelectric properties are tested. The test results show that under the working current of 150 mA, the discontinuous CBL structure cannot effectively improve the luminescence performance of LEDs. The size of the P electrode hole has little effect on the properties of the device. When the spacing of the P electrode hole increases from 20 μm to 30 μm, the external quantum efficiency (EQE) and wall-plug efficiency (WPE) increase by about 5.0% and 3.8%, respectively. When the size of the N electrode hole reduces from 17 μm×5 μm to 10 μm×5 μm, the EQE and WPE increase by about 6.5% and 3.0%, respectively. When the spacing of the N electrode hole reduces from 45 μm to 40 μm, the luminescence performance of the device is not effectively improved.

Ultra-high quality factor whispering gallery mode crystal microresonators have promising applications in nonlinear optics, coherent optical communication, and microwave photonics. Improving the surface roughness to reduce the scattering loss is an effective way to improve the quality factor of magnesium fluoride (MaF2) crystal microdisk resonators. In this paper, we fabricated ultra-smooth MaF2 crystal microdisk resonators with sub-nanometer surface roughness based on ultra-precision machining method, and the quality factor Q of MaF2 crystal microdisk resonators in 1550 nm band is up to 1.2×109 by using the tapered fiber coupling combined with cavity ringdown method, which helps to promote the research development in the field of ultra-high quality whispering gallery mode crystal microresonators in China.

As Moore's law slows down and the scale of electronic transistors approaches the physical limit, the computing speed is hard to be improved. In order to solve this problem, this paper proposes a VGG16-based diffractive optical neural network (VGG16-DONN). The structure uses the optical diffractive layer as the optical front end of VGG16 and thus replaces the first electronic convolutional layer which consumes the most time during the computing in VGG16. In addition, the CelebA dataset and cat and dog datasets are classified by the proposed structure, with high accuracy of 86.34% and 88.53%, respectively, which are equivalent to that of the electronic neural network. Furthermore, based on the proposed structure, the paper constructs a VGG16-DONN method for context-dependent processing (CDP) and classifies the CelebA dataset, with an average accuracy of 83.10%, which is equivalent to that of the electronic neural network. It should be noted that the VGG16-DONN and its combination with the CDP module can address the slow computation speed of the electronic neural network by taking advantage of the fast optical computation, and they can obtain a similarly high accuracy compared with the electronic neural network, which is of great significance to image processing, medical treatment, communication, and other fields.

As randomized singular value decomposition (RSVD) is widely used in data compression, signal processing and image denoising, the increasing matrix scale puts forward higher requirements for the traditional computing platform. Therefore, a scheme of RSVD based on the spatial optical computation is proposed. The dimensions of a matrix are reduced by the inherent properties of the complex media, and there is no need to generate and store random Gaussian matrices. In this way, the computing overhead of RSVD can be effectively reduced. The experiment proves that the proposed scheme can achieve RSVD for a 80×80 matrix with a relative error of less than 0.1 when 220 mesh ground glass is used as a complex medium, the sampling rate is 0.2, and the dimension of macropixel block is 10×10. Compared with the traditional method, it effectively reduces the time complexity and space complexity of RSVD. Finally, the effect of the scheme is verified through image compression, which provides a basis for further research on large-scale image matrix algorithms.

In order to improve the light-field coupling efficiency of a polarized light emitting diode (LED), a composite device with an integrated all-dielectric nanostructure is designed based on a one-dimensional photonic crystal and a subwavelength dielectric grating. The all-dielectric nanostructure model for collimating polarized beams is established by using the finite-difference time-domain (FDTD) method. The influence of structure parameters of the photonic crystal and the dielectric grating on the light-field regulation of the polarized LED is studied systematically. The physical mechanism of collimation characteristics affected by photonic crystal thickness, photonic crystal period, dielectric grating period, dielectric grating height and dielectric grating linewidth is analyzed. The optimized nanostructures are composed of TiO2 nano-gratings with a period of 550 nm, a linewidth of 160 nm and a depth of 120 nm, and a photonic crystal structure composed of two pairs of Al2O3/SiO2 films (the thickness of each layer is 80 nm). Calculation results show that the designed all-dielectric nanostructure can control the divergence angle of the polarized LED within the range of -6o-6o in the green light band, and achieve the far-field collimation of the light radiation, with the light extraction efficiency greater than 77%. The far-field radiation intensity of the designed structure is 6.6 times higher than that of the bare LED in the vertical direction (with a divergence angle of 0o).

To provide theoretical support for modeling and designing perovskite quantum dot devices pumped by electron beams, this paper explores the microscopic luminescence process of perovskite quantum dot films pumped by electron beams, which reveals the energy conversion model and luminescence mechanism. Firstly, the paper analyzes the microscopic luminescence and lasing processes of perovskite quantum dots pumped by electron beams theoretically. Furthermore, the lasing and luminescence thresholds of quantum dots pumped by electron beams are found to be the macroscopic physical constraint boundaries that should be detected to build the energy conversion model. Then, the effective detection methods of lasing and luminescence thresholds are analyzed and discussed, and specific evaluation experiments have been carried out repeatedly. Finally, the relationship among luminescence energy conversion efficiency, polarization distribution of nanocrystals, and electron beam pumping intensity in perovskite quantum dot films is simulated and constructed by combining with the detection of physical constraint boundaries.

Photoacoustic spectroscopy plays an important role in the field of gas detection due to its advantages of high sensitivity and good selectivity. To improve the photoacoustic detection performance without increasing the volume of devices and the reflective mirrors, a T-type photoacoustic cell on the basis of the resonant photoacoustic technology is developed, which is composed of an absorption cell with gold-plated inner walls and an acoustic resonance tube. The gold-plated boards are used to replace the traditional optical windows at both sides of the absorption cell. At the same time, the optical fiber collimator is fixed on the board to make the light beam reflect multiple times in the absorption cell, which can increase the equivalent absorption path of the gas samples. The improvement in detection performance by the proposed method is verified by mathematical modeling, theoretical derivation, finite element simulation, and experimental analysis. The signal-to-noise ratio (SNR) of the proposed method is 14.6 times higher than that of the traditional light beam excitation mode when the phase-locked integration time is 1 s, and the single detection time is 5 s. On this basis, a CO2 photoacoustic detection device is built. The experimental results reveal that the lowest limit of detection for CO2 samples is 15×10-6, and the normalized noise equivalent absorption coefficient is 11.9×10-9 cm-1·W·Hz-1/2.