Research On Key Technologies And Applications Of Intelligent Mode-Locked Lasers

Optical pulses with shorter durations and higher peak power are the consistent pursuit of various applications ranging from high-precision measurement,material processing,and signal processing.Mode-locking technology,as the main method of generating femtosecond pulses,grows fast in the past few decades.Owing to the simple structure and superior performance,nonlinear polarization evolution(NPE)relying on polarization control and Kerr nonlinearities has become an important technique for realizing mode-locked lasers.Through precise polarization tuning,NPE-based modelocked lasers can operate under various pulsation regimes.However,due to the sensitivity to the polarization of NPE-based mode-locked lasers,it can be difficult to locate the desired pulsation regime via manual polarization control in a short time.Further,continuous environmental disturbances induced polarization drift,which increases the difficulty of keeping NPE-based mode-locked lasers operating under the desired pulsation regime for a long time by manual polarization control.The environmental disturbances induced detachments(i.e.,the laser loses the desired pulsation regime)can cause various hazards,and one of them is the Q-switched instability during the detachments may cause irreversible damage to the cascaded equipment at the backend of the laser.To resolve the problems of initial mode-locking and frequent detachments of NPE mode-locked lasers,automatic mode-locking(AML)technology,as a new research field of ultrafast lasers,has received many concerns since its emergence.However,previous AML lasers cannot automatically search various pulsation regimes,and it takes a long time for these lasers to achieve initial lock or recovery from a detachment.In this dissertation,we propose the concept of intelligent mode-locking based on AML and investigate its key technologies.The goal is to realize an intelligent modelocked laser automatically searches various pulsation regimes with excellent timeconsuming performance.Further,we study the applications of intelligent mode-locked lasers.The intelligent mode-locked lasers can quickly switch among different regimes,thereby providing a new perspective for the study of spectral nonlinear dynamics inside mode-locked lasers.Specifically,this dissertation includes:1.Intelligent mode-locked laser enabled by a human-like algorithm(HLA)In this dissertation,we propose the concept of intelligent mode-locking.Inspired by the manual polarization tuning process,we design and implement an intelligent modelocked laser enabled by HLA according to human logic.HLA is a set of self-developed closed-loop algorithms specifically for intelligent mode-locking.Under the guidance of HLA,the intelligent mode-locked laser can automatically search various pulsation regimes.Benefiting from the real-time feedback control system and the high efficiency of HLA,the intelligent mode-locked laser can achieve rapid initial lock and recovery.For the fundamental mode-locking regime,the intelligent mode-locked laser requires only 0.22 s for the fastest initial lock and 14.8 ms for the fastest recovery from a detachment,refreshing the corresponding records in the AML fields by 400 times and2000 times respectively.2.Intelligent spectrum control enabled by an intelligent mode-locked laser based on time-stretch dispersive Fourier transform(TS-DFT)In this dissertation,we propose a scheme of introducing TS-DFT into the real-time feedback loop of intelligent mode-locking,and a linear mapping between the spectrum and the temporal pulse is established via a large dispersive media.Then,low-speed temporal sampling is used to identify the instantaneous spectra of femtosecond pulses to obtain the spectral width and spectral shape.Combining with the intelligent control algorithm,the spectral width and spectral shape can be programmed with a resolution of 1.47 nm.For the first time,high-precision real-time programmable control over the spectral width and spectral shape of femtosecond mode-locked pulses is achieved in this field,thereby addressing the problem of precise control over femtosecond modelocked pulses.3.Observation of transition dynamics inside intelligent mode-locked lasersCombining the function of fast switching among different regimes of the intelligent mode-locked laser with TS-DFT,transition dynamics among various regimes inside intelligent mode-locked lasers are observed.In this dissertation,the transition dynamics from a narrow-spectrum mode-locking regime and a triangular-spectrum regime to the same wide-spectrum mode-locking regime are unveiled.Both transitions exhibit complex and rich spectral dynamics.The experimental results show that if the terminal polarization states during switching are identical,the observed transition dynamics are analogous.Therefore,an inference can be drawn is that in NPE-based mode-locked lasers,the terminal polarization state determines the transition dynamics.4.Deep learning enabled spectral interferometryIn this dissertation,we propose the use of deep learning for magnitude and phase retrieval in spectral interferometry,where a broad-spectrum mode-locked laser is used as the optical source.Through both simulations and experiments,we prove that the artificial intelligence(AI)model is significantly better than the traditional Hilbert transform algorithm in terms of recovery accuracy,and AI is more resilient to quantization noise and different sampling rates.Deep learning enabled spectral interferometry has the potential to become a standard technique in spectral interferometry.