A Smart Mode-Locking System Based on SESAM Mode-Locking Solid-State Laser

Abstract

The purpose of this invention is to provide a SESAM mode-locked solid-state laser system that achieves fully automated light extraction and stable mode-locking. The system collects the fluorescence emitted by the laser before it emits light, uses a neural network structure to classify and determine the position of the cavity mirror, and combines machine learning algorithms to initially adjust the position and angle of the cavity mirror to achieve the laser extraction process. Then, intelligent algorithms are used to analyze and determine the speckle pattern after the laser passes through the scattering medium, further adjusting and optimizing the resonant cavity to achieve stable mode-locking of the laser. This invention applies intelligent algorithms to traditional solid-state laser tuning methods, achieving low cost and high efficiency. It solves the problems of long adjustment time, low accuracy, and high requirements for researchers' experience in traditional mode-locked solid-state lasers, and has strong practicality and broad application prospects.

Description

Technical Field

[0001] This invention relates to a technology for intelligent beam modulation and mode locking based on a SESAM mode-locked laser, belonging to the field of solid-state laser technology. Background Technology

[0002] Ultrafast lasers refer to laser pulses with extremely short pulse widths (typically on the order of femtoseconds or picoseconds), possessing advantages such as high peak power and a wide wavelength range, playing a crucial role in scientific research. Mode-locking technology is a common method for generating ultrafast lasers; by adjusting the parameters of internal components of the laser, the phases of multiple longitudinal modes within the laser cavity can be locked. Mode-locked solid-state lasers based on semiconductor saturable absorber mirrors (SESAMs) are a common structure for generating ultrashort pulses, offering advantages such as high power, long lifetime, and good scalability. However, in this structure, due to the involvement of multiple optical components, such as laser crystals, plane mirrors, concave mirrors, SESAMs, and output coupling mirrors, the angular alignment and positional positioning of these components significantly affect the laser's performance. Traditional design and optimization of solid-state lasers often require researchers with extensive experimental experience, which limits the widespread application of solid-state lasers.

[0003] Many intelligent algorithms, such as neural networks, genetic algorithms, and particle swarm optimization algorithms, excel at solving multi-parameter adjustment problems in complex experiments, demonstrating a clear advantage over humans. Therefore, they have wide applications and significant advantages in solid-state laser tuning. They can improve tuning efficiency, stability, and output performance, and possess rapid adaptation and learning capabilities. Traditional tuning methods typically require manual adjustment of laser parameters, a time-consuming, labor-intensive, and error-prone process. Intelligent algorithms, however, can automatically search for the optimal combination of tuning parameters based on a set objective function. Through intelligent algorithms, tuning time can be shortened, and tuning accuracy can be improved. Furthermore, by monitoring and analyzing the laser's input and output, intelligent algorithms can optimize laser parameters to achieve higher power, narrower linewidth, and other performance indicators. Currently, research on stabilizing laser output characteristics or achieving automatic mode-locking from laser emission is relatively mature. For laser cavities that have already achieved laser emission, the laser output characteristics can comprehensively reflect the cavity's state, allowing for convenient and efficient feedback control through laser output monitoring. However, for resonant cavities that have not yet achieved laser emission, how to achieve stable mode-locking using fully automatic intelligent control remains a challenge. Summary of the Invention

[0004] The purpose of this invention is to provide a SESAM mode-locked solid-state laser system that achieves fully automated light extraction and stable mode-locking. First, the fluorescence emitted by the laser before it emits light is collected. A network structure is used to classify and determine the position of the cavity mirror. Combined with a machine learning algorithm, the position and angle of the cavity mirror are initially adjusted to achieve the laser's light extraction process. Then, an intelligent algorithm is used to analyze and determine the speckle pattern of the laser after passing through the scattering medium, further adjusting and optimizing the resonant cavity to achieve stable mode-locking of the laser.

[0005] The technical solution adopted by this invention to achieve its objective is: an intelligent light modulation and mode-locking system based on a SESAM mode-locked solid-state laser, comprising a pump source, a pump coupling system, a laser crystal, three reflecting plane mirrors, three concave reflecting mirrors, a SESAM, an output coupling mirror, two non-polarizing beam splitters, a frosted glass, a converging lens, two optical switches, two CCD cameras, and a control module. Its structural features are: the two CCD cameras are black and white CCDs with different exposure times, used to collect fluorescence information and speckle information with significant intensity differences; the control module includes a neural network unit and an enhanced random search algorithm control unit; the neural network unit is used to classify and analyze the fluorescence and speckle images, and the enhanced random search algorithm control unit is used to generate feedback adjustment signals based on the fluorescence intensity information; The automatic light extraction process first requires classification and judgment using a neural network. Before being used to classify and judge the fluorescence process, the neural network needs to be trained to establish a training set of correspondences between fluorescence images and concave mirrors within the cavity. The process of establishing each correspondence in the training set is as follows: the pump source is set below the laser's threshold to pump the laser crystal. A first CCD is used to collect fluorescence patterns in real time at the output coupling mirror. The fluorescence patterns correspond to different position states of the concave mirror. The concave mirror states corresponding to different fluorescence patterns are preprocessed and then input into the neural network. The neural network is used to establish the correspondence between fluorescence images and the positions of the concave mirrors. The process of using a neural network to adjust the position of a concave mirror to achieve light emission is as follows: Based on the classification results of the neural network, the current position of the concave mirror to be adjusted is obtained. The feedback control driver adjusts the position of the mirror with an appropriate step size. During the adjustment process, the fluorescence image acquired by the first CCD changes with the position of the mirror. The trained neural network identifies and classifies the acquired fluorescence image in real time. Once the fluorescence image is classified into the region corresponding to laser emission, the movement stops. The machine learning algorithm used to adjust the cavity mirror angle to achieve light emission is as follows: As the output coupling mirror and SESAM adjustment range gradually approach the light emission position, the intensity of the fluorescence image detected on the first CCD will oscillate and increase. To avoid getting trapped in local optima and failing to reach the region of strongest fluorescence, an enhanced random search algorithm is used. Based on the collected fluorescence information, the SESAM and output coupling mirror are controlled to always move in the direction of increasing fluorescence intensity. Furthermore, by pre-setting a maximum consecutive search failure value parameter, optimization will immediately terminate when the threshold is exceeded, and a new search will begin. This algorithm has a unique exit mechanism to avoid local optima. The machine learning algorithm used to achieve mode-locking by adjusting the SESAM and output coupling mirror angle is as follows: A second CCD is used to collect the speckle pattern of the laser beam after passing through the scattering medium. When the mode-locked oscillator is in the stable region, cavity alignment is achieved by adjusting the output coupling mirror angle and the SESAM to obtain laser output. A neural network is used to classify the speckle pattern, and the spectral information in the speckle is used to determine whether mode-locking has occurred. Then, a sliding window method is used to update the variance of the intensity and contrast of the speckle pattern within the window in real time to determine whether the laser is currently stably mode-locked.

[0006] Compared with the prior art, the beneficial effects of the present invention are: This paper presents an intelligent mode-locking and mode-adjustment system based on a SESAM-locked solid-state laser. This optical system can automatically transition the laser from a non-emitting state to a stable mode-locked state using intelligent algorithms, achieving low cost and high efficiency. It solves the problems of long adjustment time, low precision, and high requirements for researchers' experience in traditional mode-locked solid-state lasers, demonstrating strong practicality and broad application prospects.

[0007] Furthermore, the laser resonator type described in this invention is a Z-shaped cavity.

[0008] The Z-cavity is a common resonant cavity structure in solid-state lasers. The combination of a folding mirror and a confocal lens effectively suppresses laser drift and oscillation during operation, improving the stability of the laser output. The relatively long cavity length of the Z-cavity allows the laser to generate a narrower beam, thereby improving the quality and focusing ability of the laser beam.

[0009] Furthermore, when constructing the neural network training set, the total number of fluorescent pattern samples collected is N, where N < 2000.

[0010] To simplify the training process, reduce neural network training time, and lower the difficulty of data collection, it is possible to achieve good training results without using excessive datasets, thus saving resources and costs and reducing the risk of overfitting.

[0011] Furthermore, when the pump light source is set to different pump power, the optical switch in the system is turned off or on to protect the first and second CCDs with different exposure times, while improving the speed and accuracy of image acquisition.

[0012] The present invention will be further described in detail below with reference to specific embodiments and accompanying drawings, but this does not imply any limitation on the scope of protection of the present invention. Attached Figure Description

[0013] Figure 1 This is a schematic diagram of the overall structure of the mode-locked laser according to Embodiment 1 of the present invention.

[0014] Figure 2 This is a schematic diagram of the fully automatic intelligent light output and mode locking process in Embodiment 1 of the present invention.

[0015] Figure 3 This is a diagram showing the neural network classification results during the light emission process in Embodiment 1 of the present invention.

[0016] Figure 4 These are fluorescence patterns corresponding to different positions of the concave mirror during the light emission process in Embodiment 1 of the present invention. Detailed Implementation

[0017] Example 1 A smart modulated beam and mode-locking system based on a SESAM-locked solid-state laser includes a pump source, a pump coupling system, a laser crystal, a reflecting plane mirror, a concave reflecting mirror, a SESAM, an output coupling mirror, a non-polarizing beam splitter, a scattering medium, a converging lens, an optical switch, a CCD camera, and a control module. Figure 1This is a schematic diagram of the overall structure of the SESAM mode-locked solid-state laser system in this embodiment. A 980nm fiber-coupled laser diode is used as the pump source. The pump light is focused into a 2×8×2.5mm³ glass doped with 10% Yb:phosphate (Yb:QX) using a pump coupling system. The Yb:QX is placed at a Brewster angle between two concave mirrors with a radius of curvature of 100mm. One end mirror of the laser resonator is replaced by the SESAM, whose nearest concave mirror also has a radius of curvature of 100mm. These concave mirrors are mounted on three piezoelectric inertial drive platforms (Thorlabs, PD1 / M). These displacement platforms can achieve linear movement in 1μm steps and have a travel range of 20mm and open-loop operation. The SESAM and output coupling mirror are mounted in two piezoelectric inertial mirror supports (Thorlabs, PIM1 / M). They provide an angular resolution of approximately 0.5μrad per cycle through piezoelectric adjustment. The total angular range is ±2°, and the angular velocity is 0.05 rad / min. When driven by an inertial piezoelectric controller (Thorlabs, KIM101), the piezoelectric inertial drive platform and piezoelectric inertial mirror support allow for continuous long-term stepping. The reflectivity of the output coupler is 98.4%. In this embodiment, a schematic diagram of the fully automated intelligent light output and mode-locking process is shown below. Figure 2 As shown in the diagram. First, the pump power was set below the laser threshold (200mW). A first CCD camera (CCD1) was used to acquire fluorescence patterns behind the output coupling mirror. These fluorescence patterns corresponded to different positional states of the concave mirror. After preprocessing the concave mirror states corresponding to different fluorescence patterns, the data was input into a neural network. The neural network was used to establish the correspondence between the fluorescence images and the positions of the concave mirrors. The final classification accuracy was above 99%. In this example, the training process used a dataset of N=1200. Finally, the trained neural network was used to classify the fluorescence images into three categories: laser emission region mirror position, excessive leftward shift, and excessive rightward shift. The classification results for concave mirror 1 and concave mirror 2 are shown in the diagram. Figure 3 As shown in (a) and 3(b), depending on the category of the fluorescence pattern, the piezoelectric inertial drive platform on the concave mirror is moved to the right and left respectively by feedback control to achieve the optimal position for laser emission. Typical fluorescence patterns corresponding to concave mirrors at different positions are shown in Figure 3. Figure 4 As shown. When the collected fluorescence enters the light-emitting region, it indicates that the reflector has been optimized. In this embodiment, the angles of the SESAM and the output coupling mirror also need to be adjusted during the light extraction process. To avoid getting trapped in local optima and failing to reach the region of strongest fluorescence, an enhanced random search algorithm is used. Based on the collected fluorescence information, the SESAM and the output coupling mirror are controlled to always move in the direction of increasing fluorescence intensity. Furthermore, by pre-setting a maximum value parameter for consecutive search failures, optimization is immediately terminated and a new search begins when the threshold is exceeded. When the intensity variance of the continuously collected fluorescence patterns is less than the set threshold, it can be determined that the maximum intensity has been found, and the light extraction process is complete. In this embodiment, after laser emission, a secondary adjustment is required to achieve mode locking of the cavity. The pump power is increased (1200mW), and a second CCD camera (CCD2) is used to collect the speckle pattern after the laser passes through the scattering medium. The concave mirror in front of the SESAM is moved in feedback, and finally, a neural network is used to classify the speckle pattern in real time. The spectral information in the speckle is used to determine whether the laser is currently in a mode-locked state. A sliding window method is used to update the variance of the intensity and contrast of the speckle pattern within the window in real time to determine whether the laser is stably mode-locked.

Claims

1. A smart modulated beam and mode-locking system based on a SESAM-locked solid-state laser, comprising a pump source, a pump coupling system, a laser crystal, a reflecting plane mirror, a concave reflecting mirror, a SESAM, an output coupling mirror, a non-polarizing beam splitter, a scattering medium, a converging lens, an optical switch, a CCD camera, and a control module, characterized in that: The CCD camera includes a first CCD camera (CCD1) and a second CCD camera (CCD2). The first CCD camera is used to collect fluorescence emitted by the laser crystal, and the second CCD camera is used to collect speckle images after the laser passes through the scattering medium. The control module includes a neural network unit and an enhanced random search algorithm control unit. The neural network unit is used to classify and analyze the fluorescence images collected by the first CCD camera and the speckle images collected by the second CCD camera. The enhanced random search algorithm control unit is used to generate feedback adjustment signals based on fluorescence intensity information. The concave mirror is mounted on a piezoelectric inertial drive platform, and the SESAM and output coupling mirror are mounted on a piezoelectric inertial mirror support. The piezoelectric inertial drive platform and the piezoelectric inertial mirror support are both electrically connected to the control module and are used to receive feedback adjustment signals from the control module and perform step adjustment. The pump coupling system is used to focus the pump light onto the laser crystal placed between the two concave mirrors at a Brewster angle. One end mirror of the laser resonator is the SESAM, and the mirror adjacent to the SESAM is a concave mirror.

2. The intelligent mode-locking and mode-elimination system based on a SESAM mode-locked solid-state laser according to claim 1, characterized in that: The pump source is configured to pump the laser crystal at a power lower than the laser threshold; the neural network unit is pre-trained with a training set that corresponds to the position of the fluorescent image and the concave mirror before being used for fluorescence image classification; and the first CCD camera is used to acquire fluorescence patterns in real time at the output coupling mirror.

3. The intelligent mode-locking and mode-locking system based on a SESAM mode-locked solid-state laser according to claim 2, characterized in that: The neural network unit is configured to classify the fluorescence image into three categories and establish a correspondence between them and the position of the concave mirror in the cavity. Based on the accuracy of the fluorescence classification results, the real-time position of the concave mirror to be adjusted is determined, and a feedback adjustment signal is generated.

4. The intelligent modulation and mode-locking system based on a SESAM mode-locked solid-state laser according to claim 1, characterized in that: The enhanced random search algorithm control unit is configured to control the piezoelectric inertial mirror support of SESAM and output coupling mirror to always move in the direction of increasing fluorescence intensity based on the collected fluorescence intensity information, so as to avoid getting trapped in local optima until the laser emits light.

5. The intelligent modulation and mode-locking system based on a SESAM mode-locked solid-state laser according to claim 2, characterized in that: The training set consists of N sets of fluorescence images and samples corresponding to the positions of the concave mirrors, where N < 2000.

6. The intelligent modulation and mode-locking system based on a SESAM mode-locked solid-state laser according to claim 4, characterized in that: The second CCD camera is used to acquire speckle images after the laser passes through the scattering medium; the neural network unit is used to classify the speckle images to determine whether the laser has entered the mode-locked state; the control module is also configured to use the sliding window method to update the intensity variance and contrast variance of the speckle pattern in the window in real time to determine whether the laser has reached stable mode-locking.

7. The intelligent modulation and mode-locking system based on a SESAM mode-locked solid-state laser according to claim 6, characterized in that: The optical switches are respectively set in the optical paths of the first CCD camera and the second CCD camera; the control module is configured to control the corresponding optical switches to be turned off or on according to the different power of the pump light source, so as to protect the first CCD camera and the second CCD camera with different exposure times, while improving the speed and accuracy of image acquisition.

Images