Magnetic Entropy Change And Machine Learning Exploration Of Metallic Rare Earth Based Compounds

The magnetic refrigeration technology,based on the magnetocaloric effect,has the advantages of environmental protection,energy saving,high efficiency,stability and reliability compared with the traditional?conventional?gas compression?Vapor-compression?refrigeration technology,which has been believed to be the best technology in future refrigeration.The practical application of magnetic refrigeration technology depends on the design of an efficient thermodynamic cycle and the study of magnetic refrigeration materials with large magnetocaloric effect and wide refrigeration temperature region.Therefore,it has scientific,economic and social significance to explore magnetic refrigeration materials with excellent magnetocaloric properties.In this thesis,magnetic properties and magnetocaloric effects has been studied for RNiSi₂?R=Gd,Dy,Ho,Er,Tm?compounds with CeNiSi₂ type structure,La_1-xMM_xFe_11.5Si_1.5C_0.2?MM is industrial mischmetal?compounds with NaZn₁₃ type structure and R₅Ge₃?R=Tb,Ho,Er?compounds with Mn₅Si₃ type structure.Magnetic properties and magnetic domain structure evolution process of Tb₅Ge₃ compound has been discussed.Prediction models for heavy rare earth based alloy with R_iT_jX_k type?R=Gd-Tm,T=Mn-Cu,X=Al,Si,Ge,Ga?and La?FeSi/Al?₁₃-based materials has been built by using machine learning methods to predict magnetocaloric effect parameters such as phase transition temperature and magnetic entropy change peak.The main conclusions are as follows:1)RNiSi₂?R=Gd,Dy,Ho,Er,Tm?compounds are antiferromagnetic at low temperature,and the transition temperatures are 18.0 K,25.0 K,10.5 K,3.0 K and 3.5K corresponding to R=Gd,Dy,Ho,Er and Tm.All of the compounds have field-induced AFM-FM metamagnetic transition under the Neel temperatures.ErNiSi₂compound has a large value of magnetic entropy change in a relatively wide temperature range above the Neel temperature,and the maximum values of magnetic entropy change under 0-2 T and 0-5 T magnetic field are 17.0 J/kgK and 27.9 J/kgK,which are higher than most of those reported in the near transition temperature.Both DyNiSi₂ and HoNiSi₂ compounds have strong antiferromagnetic coupling which results in a positive magnetic entropy change below the phase transition temperatures.A rare phenomenon whose the maximum value of positive magnetic entropy change?4.7 J/kgK?is larger than negative magnetic entropy change?4.4 J/kgK?under 0-5 T magnetic field can been found in DyNiSi₂ compound.If a reversed cycle is applied in the region under the Neel temperature,the cooling effect can also be achieved.2)La_1-xMM_xFe_11.5Si_1.5C_0.2 compound?MM is industrial mischmetal.The main composition is La 28.6%,Ce 50.7%,Pr 5.2%,Nd 15.3%,other 0.2%?maintains a good single phase when the content of industrial mischmetal is less 0.6.As the content of industrial mischmetal increases,the Curie temperature decreases,the thermal and magnetic hysteresis increase and the first order transition enhanced.The maximum values of magnetic entropy change of compounds with industrial mischmetal contents of 0.3,0.35,0.4,0.45,and 0.5 under the 0-5 T magnetic field were 26.9 J/kgK,26.1 J/kgK,24.8 J/kgK,25.3 J/kgK and 27.2 J/kgK,respectively.The results are similar to those of compounds prepared from pure rare-earth,which provides reference for the preparation of La-Fe-Si-based magnetic refrigeration materials from industrial mischmetal instead of pure element.3)R₅Ge₃?R=Tb,Ho,Er?compounds are antiferromagnetic at low temperature,and Er₅Ge₃ compound shows antiferromagnetic to antiferromagnetic?AFM-AFM?phase transition at 11K,ferromagnetic to paramagnetic?AFM-PM?phase transition at36 K.The maximum values of magnetic entropy change corresponding to the two phase transitions under the 0-5 T magnetic field were 6.8 J/kgK and 5.6 J/kgK,respectively.The distance between the two phase transitions is relatively close which broaden the cooling temperature region.The Neel temperature of the Ho₅Ge₃compound is 20 K,and a field-induced metamagnetic transition occurs under the Neel temperature.The maximum value of magnetic entropy change of the compound under the 0-2 T magnetic field was 2.0 J/kgK,and the refrigeration capacity was 36.0 J/kg.The maximum values of magnetic entropy change under the 0-5 T magnetic field is9.1 J/kgK,and the refrigeration capacity is 225.2 J/kg.Tb₅Ge₃ compound is antiferromagnetic at low temperature and the Neel temperature is 85 K.Polycrystalline Tb₅Ge₃ has a little amounts of Tb phase and occurs skyrmion-like nanodomains down to 128 K.But no skyrmion-like nanodomains were found in the single crystal Tb₅Ge₃.The appearance of skyrmion-like nanodomains may be caused by the coupling between Tb phase and Tb₅Ge₃ phase and the research work is still in progress.4)Using a machine learning algorithm to build models for R_iT_j X_k-type heavy rare-earth alloys and La?FeSi/Al?₁₃ based materials.For R_i T_jX_k-type heavy rare-earth alloys,the models learn the relationship between chemical composition,paramagnetic transition temperatures and magnetic entropy change?under the 0-5 T magnetic field?.For the paramagnetic transition temperature model,the performance of fitting?coefficient of determination R²?of the test set is 0.86.For the magnetic entropy model,the test set R² is 0.55.For La?FeSi/Al?₁₃-based materials,the models learn the relationship between chemical composition,Curie temperatures and magnetic entropy change?under the 0-2 T and 0-5 T magnetic field?.The coefficient of determination R² of the three models are 0.96,0.87,and 0.91 respectively,and the average absolute error MAE is 9.81 K,1.14 J/kgK,and 2.53 J/kgK,respectively.The performance of the models is good enough,which preliminarily confirms the feasibility of using machine learning methods in analysis and prediction of magnetic refrigeration materials.