The Thermal Transport Properties Of Metal Halide Perovskites Phononic Crystals

Metal halide perovskites have attracted substantial attention in photovoltaic and thermoelectric applications due to their excellent energy conversion efficiency.However,it is difficult to further improve the thermoelectric figure of merits of metal halide perovskites via conventional methods due to their large band gaps and high thermal conductivity.As meta-structured materials with artificial periodicity,phononic crystals provide a new platform to control phonon dispersion and suppress lattice thermal conductivity to further improve thermoelectric properties.Machine learning is good at discovering the hidden laws of complex systems,which is a superior way to replace massive numerical calculations and reduce expensive experimental costs.Therefore,the energy conversion efficiency of metal halide perovskites can be effectively improved and estimated by constructing phononic crystals to reduce their thermal conductivity and using machine learning to predict the thermal transport properties of corresponding phononic crystals.In this context,the thesis was completed in the following logic flow:firstly,the elastic properties of lead-free halide double perovskites were systematically studied by means of first principles calculations and experimental characterizations;secondly,hybrid perovskite phononic crystals were fabricated and their thermal transport properties were studied by finite element method and lattice Boltzmann method;finally,the machine learning method was developed to efficiently evaluate the thermal transport properties of metal halide perovskite phononic crystals by reducing the cost of massive data calculations.The main discoveries and novelty of the thesis are listed as follows:（1）Elastic properties of lead-free halide double perovskite Cs₂AgBiBr₆The elastic constants of Cs₂AgBiBr₆ were calculated by the first principles method,and then the Young’s moduli,shear moduli,Poisson’s ratios and bulk modulus were obtained.The results show that the cubic Cs₂AgBiBr₆ perovskite has significant elastic anisotropy,which arises from the anisotropy of the crystal structure.The calculation results were confirmed by nano indentation and high-pressure synchrotron powder X-ray diffraction experiments.In addition,the elastic properties of Cs₂AgBiBr₆ are better than those of MAPbBr₃（MA=CH₃NH₃）,which is due to the stronger Ag-Br and Bi-Br bonds in the former than the Pb-Br bonds in the latter,hence leading to the enhancement of framework stiffness.（2）Thermal transport properties of MAPbI₃ phononic crystalsMAPbI₃based phononic crystals were developed and their thermal transport behaviours were fine-tuned by delicately optimizing the structure of the basal bodies（MAPbI₃）and the scatterers（vacuum）.The cross-scale simulation results show that MAPbI₃ phononic crystals with large size scatterers,low symmetry meta-structure,basal bodies and scatterers can exhibit low relative thermal conductance at low temperature.In particular,the relative thermal conductance of orthorhombic MAPbI₃ based phononic crystals,with square lattice and triangular holes,can be as low as 10.7%.In addition,it was discovered that the substantially increased disorder of MA⁺cations,from the tetragonal to cubic transition,in these Pn Cs significantly increases the phonon group velocities and reverses the dominant means of phonon transport from diffusive to quasi-ballistic,hence leading to anomalous decrease of relative thermal conductance.The proposed Pn Cs provide a new pathway for engineering thermal conductivity of metal halide perovskites and improving the performance of corresponding devices.（3）The thermal transport properties of metal halide perovskite phononic crystals predicted via machine learning.Due to the massive calculation cost of traditional numerical methods,it is impossible to fast predict the thermal transport properties of anisotropic phononic crystals.In this thesis,a machine learning method based on the elastic constants of materials was developed to rapidly predict the band structures and thermal conductance of halide perovskite-based anisotropic phononic crystals.In the prediction of band structures and thermal conductance,the prediction errors of artificial neural network are relatively small,and the peak positions and ranges of prediction results are largely consistent with the results of finite element method.Based on this method,MA₂CuBiI₆ with the strongest anisotropy of relative thermal conductance was successfully selected from 63 material data sets.The minimum and maximum relative thermal conductance of MA₂CuBiI₆ are 13.2%and 33.9%,respectively.This method successfully establishes and trains a mechanical-thermal model which can map the elastic constants of crystals and the dispersion/heat transport of phononic crystals.The proposed model can predict the thermal conductance of phononic crystals in a very reliable and rapid manner.