| Nanoimprint lithography(NIL)technology,as the next generation lithography technology,has the advantages of low cost,high replication accuracy,and simple process steps.It has gotten widespread applied in a lot of fields.However,problems such as long heating cycles,severe degradation effects of stamps,and uneven residual layers limit the further development and application of NIL.In order to solve the pattern defects caused by incomplete filling of the cavity,it is necessary to deeply understand the physical mechanism of filling the stamp cavity with resist.In addition,the application of artificial intelligence research methods in the field of NIL can effectively improve research efficiency.Therefore,this thesis proposes a new fast global simulation method using Simprint Core,conducts a series of studies on the resist filling mechanism of thermal nanoimprint lithography(T-NIL)and establishes a prediction model for the replication quality of thermal nanoimprinted patterns using artificial neural network.The main work is as follows:1.Four different materials of stamps are designed and the specific mechanisms of pressure and temperature on the filling of stamp cavities with resists are studied.The results show that the residual layer after imprinting is thin but unevenly distributed for soft templates(PDMS,PU)and thick but very uniform for hard templates(Si,Ni).Increasing the pressure can reduce the residual layer thickness and promote the uniformity of residual layer distribution.The residual layer becomes thinner as the temperature increases,but the uniformity of its distribution decreases at first and then increases.Increasing the temperature and pressure can significantly improve the cavity filling proportion,but the improvement of imprint quality is very weak after the pressure reaches a certain value.An orthogonal experiment is designed with the initial thickness of the resist,imprint temperature,imprint time and average pressure as the study parameters,and the results show that the imprint temperature has the most significant effect on the imprint quality.2.A large number of stamps with different morphological features are designed,and different morphological elements of the stamps are studied,including the influence of stamp pattern density,pattern micro-nano ratio,and pattern aspect ratio on the filling mechanism of T-NIL resist.The results show that both adjacent non-uniform and non adjacent uniform pattern density stamps present a bimodal feature in the contact pressure at the beginning of imprint.The residual layer thickness increases with the increase of pattern density,and the residual layer thickness decreases after the stabilization period due to long time imprint,and the decrease is more obvious under high pattern density.The imprint process of stamp with adjacent uneven pattern density reveals the flow behavior of the residual resist.In a study of three micron-nano ratio(1:1,1:2 and 1:4)stamps,it is found that the overall condition of imprint becomes worse as the micro-nano ratio decreases.Adjacent mixed size features can lead to a negative impact on the pressure distribution and extent of filling during the filling process.Simulation results of stamps with different pattern aspect ratios prove that as the aspect ratio increases,it takes longer to fill the cavity and the residual layer thickness decreases monotonically with the increase of the aspect ratio.3.A three-layer BP neural network is established using the process parameters obtained from the simulation,with a minimum mean square error of 0.00070 and an overall correlation coefficient of 0.95143 for the model,with an average relative percentage error of 4.3667%.This shows that the model has good prediction performance,and the model can be applied to predict the replication accuracy of thermal nanoimprinted patterns,and the established"data+artificial intelligence" research model effectively improves the efficiency of the research. |
