Dr Rajarshi Tiwari

Short Biography

I come from a city of Allahabad (now known as Prayagraj), in the state of Uttar Pradesh in India, where I got most of my education. I did my B.Sc. from University of Allahabad, India with Physics and Mathematics as major in 2005. Then I joined the Integrated Ph.D. (M.Sc. + Ph.D.) program at Harish-Chandra Research Institute (HRI), Allahabad, India. I finished my MSc in Physics from HRI in 2008, and PhD in Condensed Matter Physics in 2013 with my thesis on The Effect of Geometric Frustration on Some Correlated Electron Systems.

Thereafter, I joined Prof. Stefano Sanvito’s research group at Trinity College, Dublin in 2013 for a year as Research Assistant, and after defending my thesis in September 2014, I continued there as post-doctoral researcher till 2019. Later, I continued as Research Fellow in the School of Physics, and worked over a range of project that overlap material science, many-body theory, high-throughput DFT and machine learning. My interests in computing, machine learning and quantum science led me to Irish Centre for High End Computing (ICHEC) which has invested in the Quantum learning initiative for Ireland.

Research Interests

My interests range over high performance computation, computational material science, condensed matter Physics, application of Machine Learning and Quantum Computing in solving models of correlations. My interest in machine learning are a bit more open ended and geared towards looking at statistical side of data and functional side of neural networks, and other real life situations.

Correlated Systems

During my Ph.D. my research interest grew around strongly correlated electronic systems, where I primarily worked on models of correlation. I love to explore magnetism, transport, frustration, disorder and their interplay in correlated electronic systems, such as transition metal Oxides, magnetic perovskites, pyrochlore systems. The phenomenology of these systems is best explored with suitably simplified models, such Hubbard model, Kondo-Lattice Model, Holstein model, Heisenberg model, and their variations/combinations depending on whether the relevant degrees of freedom are (i) itinerant electrons, (ii) localized spins, or (iii) phonons. I explore solving and analysing appropriate models of correlations through real-space based techniques like Monte-Carlo methods and Exact-Diagonalization.

ML in Material Science

After joining Trinity College Dublin, I expanded my research interests over computational materials science along with condensed-matter physics, where I explore application of machine learning in (i) solving or learning features in correlated systems and (ii) high-throughput ab initio calculations. I learnt ab-initio simulations tools such as VASP/FHI-AIMS to compute energetics of real systems, organize and process the data for ML applications. In material science, I have done a few projects that involve running high-throughput DFT calculations, and processing corresponding data through statistical and ML tools to summarise, refine and predict scientific outcome. While I was at TCD, we also worked on developing a workflow to combine ab-initio and ML tools to build up force fields for simulating large, disordered systems. The ICHEC-Flagship project EuroCC-AF-3 was quite helpful in this direction.

Quantum Computing

Here at ICHEC, I am part of Quantum Computing initiative, where I explore tools for solving lattice models through Quantum computing Hardware. For this is relatively new, yet exciting area to explore and utilise the many-body knowledge. I am currently involved in developing skills and resources to launch quantum computing tasks in an HPC setting, and looking into possibility of deploying a hybrid classical-quantum workflow for HPC environment.

Here are few bullet points into my activity that span over the above-

• Exploring methods for structure property relations of materials with use of ML in High-Throughput ab initio data.
• Applying ML in models of many-body physics, such developing ML based lattice density functional theory for models.
• Exploring many-body physics models through Quantum computing, both circuit based, and simulation based.
• Exploring possibilities of DNNs as generative models for solving many-body problems in correlated systems.