Study On Property-design Of Carbon Dots Assisted By Machine Learning

As a new member of carbon nanomaterials,carbon dots(CDs)have shown promising applications in the fields of biomedicine,catalysis,physics,material science and environmental engineering due to their inherent multiple advantages(such as low toxicity,high biocompatibility,tunable luminescence wavelength and structural diversity).At present,the researches on CDs are developing rapidly;Meanwhile,CDs show diversity in synthesis,structures,properties and applications.Diversified synthesis routes,rich properties,complex structure and diversified research perspectives not only promote the rapid development of CDs,but also bring new challenges for understanding and analyzing a large number of complex experimental data.Recently,machine learning(ML)technology has developed rapidly and has evolved into a multi-disciplinary field.It has become one of the most effective methods to analysis and summarize the algorithm model based on datasets,and autonomously to exlore unknown fields,so as to predict and optimize their own conditions.With the assistance of ML,this paper carried out a series of studies on the preparation and property-design of CDs.The main work is focused on investigations of controllable synthesis,fluorescence and electroluminescent properties,third-order nonlinear optical(NLO)effect and C-H selective catalytic oxidation activity for CDs under the guidance of ML.In particular,under the optimization of ML,a series of CDs being used as active emission layer of light emitting diodes(LEDs)with excellent performances were fabricated.With accurate prediction and optimization of ML,CDs based on polyascorbic acid-like structure were synthesized as homogeneous catalysts to realize the direct synthesis of adipic acid(AA)by one-step high-efficiency catalytic oxidation of cyclohexane.Specific works can be included as follows:1.Violet,pure blue,green and orange fluorescent CDs were successfully synthesised by adjusting the experimental parameters of solvothermal method to precisely control the size and surface groups of CDs.These CDs exhibit strong fluorescence which are even more than fluoresecence of metal quantum dots.Then,under the ability to explore unknown experimental conditions of ML,we successfully modulated white fluorescent CDs with adjustable fluorescence correlated corlor temperature(PL CCT)by mixing pure blue,green and fluorescent CDs.2.Based on the ultra-strong fluorescenct properties of violet,pure blue,green and orange fluorescent CDs,LEDs based on CDs(CDs LEDs)with violet,pure blue,green and orange light-emitting were fabricated with them as the active emission layer,respectively.These devices not only show excellent electroluminescence performance,but also have excellent stability.At the same time,based on the white fluorescent CDs with adjustable PL CCT,the variation law of PL CCT and electroluminescence correlated color temperature(EL CCT)was explored by artificial neural network(ANN).And the controllable fabrication of white light-emitting CDs LEDs from cold EL CCT to warm EL CCT was realized.3.CDs synthesized by traditional electrochemical etching graphite rod method were used as precursors.A series of CDs with different surface oxygen-containing functional groups and high carbon crystalline core were synthesised by modifying their surface functional groups and carbon core crystalline to be catalysts for cyclohexane oxidation.These series of CDs showed high catalytic activity and achieved high-efficiency catalytic oxidation of cyclohexane to one-step synthesis of AA.Density functional theory(DFT)and correlated experiments show that the CDs with a large number of surface hydroxyl or higher carbon core crystalline are beneficial to the high-efficiency catalytic oxidation of cyclohexane to one-step synthesis of AA.4.Using ML to train a large number of experimental data of cyclohexane catalytic oxidation,the structure of metal-free catalyst that can efficiently catalyze the oxidation of cyclohexane was accurately predicted,and the CDs based on ascorbic acid-like polymer structure was designed and synthesized.It can be easily dissolved in most organic or inorganic solvents.Therefore,we dispersed it in acetone as a homogeneous catalyst to combine the advantages of homogeneous catalysts and heterogeneous catalysts so that it can achieve one-step efficiently selective oxidation of cyclohexane to AA in a homogeneous system.In addition,we also designed and built a continuously homogeneous catalytic reaction device which can realize the synergistic effect of homogeneous catalysis and phase transfer process,and it has been applied to most C-H bond activation reactions.5.A series of CDs with third-order NLO effects were synthesized by a hydrothermal method.ML was used to explore the variation between the synthesis conditions of CDs and the third-order NLO polarization coefficient(χ(3)).Also,the main conditions affecting the value of χ(3)were grasped,and they were the reaction time and reaction temperature.The ability of ML to adaptively explore the unknown reaction conditions was used to achieve CDs with adjustable χ(3)of controllable synthesis.At the same time,as the incident laser energy increasing,it is found that CDs has the third-order NLO switching behavior.