Research Projects | Shambhu Bhandari Sharma

Ab-initio Phase Diagram of Binary Alloys at High Pressure

Melting curve calculations up to extreme pressures are essential for predicting materials that sustain nuclear fusion reactors, aerospace, and military applications. We introduce controlled impurities to pristine metals to predict stable alloys and study their structural, electronic, and thermo-mechanical properties using DFT and AIMD.

VASP LAMMPS Phase Diagrams

PCFC-Automation

Automates melting points & phase equilibria via TI and MD.

1. Ab initio melting curve of elements (JCP Article) 2. Phase diagram and melting under extreme conditions (arXiv Preprint)

Agentic DFT Workflows & Literature Reproducibility Audit

An autonomous, fully local neurosymbolic pipeline evaluating 4B open-weights LLMs (Qwen3:4b via Ollama) for materials science literature audits. It extracts first-principles methodology from published papers on 2D monolayers, enforces factual grounding with deterministic validation gates, and drives multi-agent crews (Translator, Convergence, Diagnostician, Critic) to autonomously set up and re-run verification calculations in SIESTA.

Local LLM (Qwen3) Ollama SIESTA DFT Multi-Agent Systems Literature Audit

Agentic-Workflow

Autonomous local LLM pipeline evaluating reproducibility and re-running first-principles calculations.

1. Evaluating a 4B open-weights local LLM for agentic DFT workflows: a literature reproducibility audit (Manuscript)

Hybrid Graph Neural Networks & Data Leakage in Materials Informatics

An independent reproduction and audit of a recent hybrid graph-neural-network (CGCNN, MEGNet, SchNet) framework for band-gap prediction. We revealed that the originally reported high accuracies were artificially inflated due to severe train-test contamination and target-derived feature leakage. Controlled ablations isolate the genuine lift of GNNs over tabular baselines, correcting the literature record.

Graph Neural Networks Data Audit / Leakage Band Gap

Hybrid-GNN

Data leakage analysis and predictive pipeline audit for J. Chem. Inf. Model. 2026.

1. View Manuscript

High-throughput Discovery & Properties Prediction

High-throughput materials screening grid

Accelerating materials discovery by combining high-throughput computations with machine learning and uncertainty sampling. This framework was successfully deployed to discover and predict the properties of novel 2D materials, including penta-SiCN and penta-PBN, and to uncover rare auxetic behaviours such as a negative Poisson's ratio.

High-Throughput penta-SiCN & PBN Auxetic Properties

PDC

Phase diagram construction based on uncertainty sampling.

1. Penta-SiCN: A Highly Auxetic Monolayer 2. Large Negative Poisson's Ratio and Anisotropic Mechanics in New Penta-PBN Monolayer 3. First-Principles Prediction of New 2D p-SiPN: A Wide Bandgap Semiconductor

Photocatalysis for Energy Conversion

Computational design of highly tunable 2D photocatalysts for clean energy applications. My foundational work investigated Cu-doped TiO₂ (112) surfaces for enhanced visible-light absorption, while my recent research demonstrates that strain-engineered pentagonal PdTe₂ (penta-PdTe₂) enables spontaneous, sustainable solar-driven water splitting.

Photocatalysis Cu-doped TiO₂ Water Splitting penta-PdTe₂

1. First-principles DFT study of Cu-doped TiO₂ (112) surface 2. Pentagonal PdTe₂ Monolayer for Sustainable Solar-driven Hydrogen Production

Low-dimensional (0D · 1D · 2D) Material Properties

We predict and characterise novel low-dimensional materials using DFT combined with non-equilibrium Green's function (NEGF) methods and ab-initio MD. Focus areas include quantum transport, nanomechanics, optoelectronics, and magnetism under varied conditions of doping, strain, and chemical passivation.

DFT + NEGF Quantum Transport 2D Materials

Phonopy-scripts

Phonopy-VASP interface to compute phonon & thermodynamic properties.

sisl-TBtrans Tutorials

Tutorials for the sisl-TBtrans-TranSiesta suite for quantum transport.

1. Transport properties of one-dimensional monoatomic ultrathin gold nanowire: A DFT-NEGF approach 2. Enhanced mechanical, electronic, magnetic and optical properties of 2D CrX₂ 3. Structural, electronic, magnetic and mechanical properties of V-doped BN monolayer

Novel Electrides — DFT & AIMD Exploration

Electrides are ionic compounds where trapped electrons serve as anions. We use DFT and ab-initio MD to explore novel electrides from 0D to 3D dimensionality, studying their bulk and facet-dependent surface properties, doping effects, and slab/cleaving models to uncover functional behaviour.

Electrides DFT Surface Properties Slab Models

magnetism-prediction

DFT-aided Machine Learning Search for Magnetism in Bimetallic Chalcogenides.

1. Signature of low-dimensional quasi-F centers in zirconium-rich electrides

🔬 Research Projects