Digital Micro/ Nanostructures Research Laboratory

In order to model material structure, properties and behaviour at multiscale, we are employing finite element method (including mesoscale phase field simulations), molecular dynamics (MD) and density functional theory (DFT) methods. The multivariate analysis, uncertainty quantification, generative design and optimization tasks are performed using machine learning technique. The graph network shows how the machine learning can help in integrating the computational methods to enhance the robustness of experimentally informed digital twin. At present the methodologies are being implemented in the design study of laser irradiated multicomponents microstructures. 

Using the first principles calculations, we compute the stiffness tensors for FCC Al, TETRAGONAL Al3Cu and ORTHORHOMBIC Al3Ni crystal systems. The calculation is performed not only at 0 K but also at higher temperatures for Al and Al3Ni systems, so as to convey temperature calibrated information to mesoscale phase field simulation. The DFT-computed elasticity coefficients are used by the multi-phase field model developed at T = 723.0 K to examine the elastochemical effects in the microstructural evolution of the three phases - namely, Al-rich FCC phase matrix with grains of AL2CU and AL3NI precipitates. Grains with larger stiffness coefficients are observed to have pronounced normal stress components as compared to the matrix with smaller elasticity value. It has been revealed that the model using constant (0 K based) elasticity coefficients deviates more and more with respect to thermally calibrated elasticity coefficients when the time increases during microstructural evolution. Hence, the computation of properties at different temperature (the hallmark aspect of the work of our laboratory) is very important for designing materials with high accuracy. 

(For more details on this research work , please contact:  S. Poudel, N. Moelans  and A. Kunwar )

AlloyManufacturingNet app is online. We have employed ensemble machine learning model to train the hardness and elongation datasets, and integrated the two target features using "scale invariant optimization" technique. The software app at present stage of development is established as 22 and 26 element framework for hardness and elongation features  respectively.  Four major manufacturing routes, namely, casting, sintering, annealing and wroughtt/others are categorized for feature testing. The following images  demonstrate the ternary contour diagrams  of  hardness (HV) and ductility (% elongation) features in VNbTaZrW multicomponent alloy for casting and sintering procedures as  predicted by  AlloyManufacturingNet. 

Through tensor computation, we have achieved a 28-fold reduction in the data size of initially computed microstructural information. Utilizing the tensor inpainting method for phase field-generated microstructural data images enables the use of sparse datasets in alloy design for many applications, including biomedical informatics. Highly elastic and corrosion-resistant microstructures are preferred when designing the Ti-Cr alloys for dental implant applications. Elastically and electrochemically favorable spatial regions within the binary Ti-Cr alloy microstructure are captured dynamically using scaled sigmoidal interpolation function of composition variable.  The detailed steps related to the integration of phase field method with tensor computation to create highly portable data of alloy microstructure, are described in Subedi et al. 2024 ( U. Subedi, N. Moelans, T. Tański, A. Kunwar, Rapid portabilization of elasto-chemical evolution data for dental Ti-Cr alloy microstructure through sparsification and tensor computation, Scripta Materialia, 244 [2024] 116027 ).  

Through knowledge engineering ( materials informatics  via natural language processing and quantified named entity recognition), we have been able to summarize the  state-of-the-art research summary of nanotwinned Cu in joining and engineering applications. The detailed methodology of this informatics approach is presented in  Zhang et al. 2025 ( Z. Zhang, Y. Hu, X. Li, F. Wang, Y. Li and A. Kunwar, Understanding the mechanothermally superior nanotwinned copper: Fabrication procedure, mechanistic models and technological applications, Advances in Colloid and Interface Science, XXX [2025] 103630 ).