With the continuous development of photonic crystal structure design

It is extremely difficult to establish an analytical model based on the basic principles of physics and to calculate and solve the appropriate structure to achieve specific photonic properties.Especially when the geometric shape and spatial distribution of the structure are very complex

it is almost impossible to start.The electromagnetic modeling based on numerical simulation method is adopted

Designers also tend to constantly fine-tune the geometry and repeatedly perform simulations to get closer to the target

Relies on past experience in designing templates and takes a long time.Therefore

the reverse design algorithm based on deep learning is used through data-driven modeling.The neural network is used to learn a large amount of data to obtain the corresponding law between the optical characteristics and the design parameters

It can help designers find non-intuitive and irregular photonic crystal structures more effectively.And is beneficial to that refine development of the photonic crystal design.The Artificial neural network model of two-dimensional GaAs photonic crystal point defect microcavity is established to predict the radius of the microcavity.And can provide a rapid design method for subsequent photonic crystal structure research.