Study On Thermal Conductivity Of Amorphous GeTe Based On Molecular Dynamics Simulation

In recent years,phase change memory stands out in some emerging nonvolatile memories with its advantages of nonvolatile,fast switching speed,long storage time,large storage capacity and good scalability,and is expected to become one of the cores of the next generation semiconductor memory.In which,GeTe has attracted extensive attention because of its simple composition,easy preparation and reversible transformation between crystalline and amorphous phases in tens of nanoseconds.The performance of phase change storage of GeTe is subject to the thermal properties of the material.It is found that the crystalline GeTe often deviates from the stoichiometric ratio and has defects such as Ge vacancy.It has been reported that vacancy can significantly reduce the thermal conductivity of crystalline GeTe.However,the effect of off-stoichiometry with vacancy on the thermal conductivity of amorphous GeTe is not clear.In this paper,the thermal conductivity of amorphous GeTe materials deviating from stoichiometric ratio at room temperature is studied by using non-equilibrium molecular dynamics method through traditional molecular dynamics and machine learning molecular dynamics.Results of the research mainly include:1.By using the modified Tersoff empirical potential function,the thermal conductivity of amorphous GeTe is simulated based on the traditional molecular dynamics method.In order to consider the influence of Ge vacancy,we use the energy-volume equation of state to optimize the corresponding amorphous phase density.The calculation results show that the thermal conductivity decreases with the increase of the deviation from the stoichiometric ratio.The thermal conductivity of the material is about 0.17 ± 0.01 W /(m · K),which is within the range of experimental values.Combined with ab-initio molecular dynamics,the coordination number and bond angle distribution are obtained,which shows that the existence of off-stoichiometric ratio will not significantly affect the rapid phase transition of GeTe.2.The thermal conductivity of amorphous GeTe is analyzed by machine learning molecular dynamics method.Firstly,we use the ab-initio molecular dynamics to sample different states of GeTe,and then it is combined with the machine learning method to fit the neural network potential function,which is applied to the calculation of the thermal conductivity of amorphous GeTe.The results show that when the thickness reaches 12 nm,the value of neural network potential function is consistent with the experimental values,which is slightly larger than that of empirical potential.It shows that the neural network potential function is effective in the calculation of thermal conductivity.