The Study Of Advanced Optimization Algorithms For High-gain Free Electron Lasers

As the world's brightest accelerator-based light sources,the X-ray Free Electron Lasers(XFELs)have given us for the first time the possibility to explore the mysteries of natural phenomena at Angstrom level spatial scales and femtosecond level tempo-ral scales.They can generate photon pulses with high intensity(typically higher than 1 GW),short pulse duration(typically shorter than 1 ps),high peak brightness(about 1032 photons/s/mrad2/mm2/0 1%BW)and fully spatial coherence.These novel capa-bilities fuel progress in many disparate areas of research,from quantum materials and condensed matter science,to atomic molecular and optical(AMO)science,crystallog-raphy,photochemistry,structural biology,medicine and other research disciplines.The FEL works by converting a relatively small fraction of the power of a relativis-tic electron beam into high power radiation.Specifically,the FEL extracts energy from the relativistic electron beam and converts it to high power radiation by the resonant in-teraction of the electron beam with an electromagnetic field in a periodic array of dipole magnets which is known as undulator.However,this conversion efficiency from the electron beam to radiation power(??gradiation/Pe-beam)is quite low,typically about 0.1%of the total electron beam power for most of the FEL facilities.On the other hand,the FELs in particular are some of the largest,most data-intensive and most complex scientific systems in existence.The interrelations between subsystems are often compli-cated and non-linear,and the system dynamics involve large parameter spaces,making it difficult to optimize the entire machine manually.Due to these intrinsic mechanisms and external difficulties,the optimization of high-gain FEL facilities increasingly relies on advanced optimization algorithms.In such a context,the central subject of this dissertation is the use of various ad-vanced optimization algorithms to solve the corresponding FEL problems.The work begins with theoretical analysis and experimental demonstrations of the alignment of electron beam orbits,which is the heart of efficient coupling of relativistic electron beams with the radiation field.In terms of theoretical research,the beam-based alignment(BBA)simulations are systematically carried out with the relevant parame-ters of SXFEL,and the influence of the BPM resolution on the FEL power is analyzed.A new BBA method is proposed with the help of genetic algorithms(GA)which over-comes the theoretical deficiencies of the conventional BBA technique.The simulation results show that the proposed technique can solve the problems encountered in the BBA experiments of soft X-ray FELs driven by low energy linacs.This method pro-vides a new option for the BBA experiment of soft X-ray FEL facilities.In terms of experimental research,the preliminary BBA experiments on SXFEL are introduced.The experimental results show that the orbit of the electron beam can be improved to a certain extent.Additionally,the dissertation investigates the online optimization algorithm of high-gain FEL facilities,which is the basis of machine tuning and parameters optimization.A high-level application software is developed for the SXFEL undulator system which can meet the daily needs of machine tuning.It operates reliably and has laid the foun-dation for the commissioning of the SXFEL facility.Another work concerning online optimization algorithms is the development of a super-fast FEL simulation code.It does not only improves the calculation efficiency by two orders of magnitude but also greatly expands the applicability with the help of the diffraction factors and special designed modulators.The code is benchmarked against GENESIS based on SASE and EEHG operation mode.A specific emphasis is placed on on-line prediction of FEL properties.The work also includes a GA-based online automatic tuning experiment and the seed laser transverse position feedback system.The last but not the least,the dissertation also demonstrates an online method to retrieve FEL pulses on a shot-to-shot basis at SXFEL.The initial beam central energy and energy spread for the lasing part of electron bunches are evaluated by the locally weighted polynomial regression since the region of the lased electron bunch is much smaller than the entire electron bunch.The proposed method can provide a reliable characterization of FEL pulses shot-by-shot,especially during the machine tuning.This method helps researchers directly observe the evolution of FEL pulse profiles during the FEL amplification and analyze the correlation of the FEL properties in real time,which helped us to realize the lasing and parameters optimization of the SXFEL.Based on this method,the experiments at SXFEL also indicate the feasibility to construct a timing diagnostic and feedback system for the electron beam and an external laser source with a resolution better than 3 fs.This has been proved to be very useful to stabilize the seeded FEL output.It should be noted that although this method is designed for seeded FELs,it is still applicable to most SASE-dependent ultrafast schemes.At present,advanced-algorithm-based approaches are technologically mature enough to be brought to bear on a wide variety of problems in FEL domains.Those advanced optimization algorithms will certainly become increasingly valuable tools to meet new demands for high-gain FEL facilities.