The invention relates to a
hafnium zirconium oxide all-component ferroelectric characteristic and fatigue optimization method based on
machine learning potential. The construction method comprises the following steps: step 1, acquiring atomic configuration, energy, stress and
electric dipole moment data of HfxZr1-xO2 under different components x (0 < = x < = 1) as an
original data set; 2, training a multi-task
machine learning potential model by using a
deep learning algorithm, wherein the model comprises an energy-force prediction
branch and an
electric dipole moment prediction
branch; 3, performing
molecular dynamics evolution under different temperatures and external electric fields to obtain a change curve of the polarization intensity along with the cycle period; and step 4, screening an optimal material component according to the nonlinear corresponding relationship between the fatigue rate and the Hf content obtained by
simulation. Compared with the prior art, the method has the advantages that the training
data set covering all-component space is constructed, the
machine learning potential model capable of synchronously describing energy, force and
electric dipole moment is obtained through deep neural network training, and the regulation rule of the Hf
doping proportion on the
crystal phase stability is disclosed through large-scale
molecular dynamics simulation; and the relationship between the domain wall movement rate and the fatigue rate is quantified, and theoretical guidance of atomic scale is provided for component design of the FeRAM device.