Kavita Joshi

Kavita Joshi

Physical and Materials Chemistry Division

About Me

Kavita did her MSc and PhD from the Department of Physics, University of Pune (Now known as Savitribai Phule Pune University). After a short stint at CEA-Grenoble as a post-doctoral fellow, she worked in various capacities at the Centre for Modeling and Simulation. In 2010, she joined NCL as a scientist fellow. She is a Principal Scientist at CSIR-NCL and leads an enthusiastic group of young researchers. Her research interests include computational heterogeneous catalysis, understanding reaction mechanisms in industrially important reactions like ethylene epoxidation, in silico catalyst design for CO2 to value-added products, Methane activation, and MeOH to DME. In the energy/storage materials field, her group is employing DFT-based models to understand the lithiation and delithiation processes. Developing ML models based on available DFT and experimental data to predict properties or design catalysts is also an active area of research in the group.

Professional Experience

  • 2019 --       Principal Scientist
  • 2014 -- 19  Sr. Scientist, NCL, Pune
  • 2010 -- 13  Scientist Fellow, NCL, Pune, India
  • 2005 -- 09  Research Associate, University of Pune, Pune , India.
  • 2004 -- 05  Postdoctoral Fellow at CEA-Grenoble France.

Selected Publications

  • Ashwini Verma, Nikhil Wilson And Kavita Joshi, Solid state hydrogen storage: decoding the path through machine learning, International Journal of Hydrogen Energy., 50, 1518 - 1528 (2024), DOI:https://doi.org/10.1016/j.ijhydene.2023.10.056.
    We present a machine learning (ML) framework HEART (HydrogEn storAge propeRty predicTor) for identifying suitable families of metal alloys for hydrogen storage under ambient conditions. Our framework includes two ML models that predict the hydrogen storage capacity (HYST) and the enthalpy of hydride formation (THOR) of multi-component metal alloys. We demonstrate that a chemically diverse set of features effectively describes the hydrogen storage properties of the alloys. In HYST, we use absorption temperature as a feature which improved H2wt% prediction significantly. For out-of-the-bag samples, HYST predicted H2wt% with R2 score of 0.81 and mean absolute error (MAE) of 0.45 wt% whereas R2 score is 0.89 and MAE is 4.53 kJ/molH2 for THOR. These models are further employed to predict H2wt% and H for 6.4 million multi-component metal alloys. We have identified 6480 compositions with superior storage properties (H2wt% 2.5 at room temperature and H 60 kJ/molH2). We have also discussed in detail the interesting trends picked up by these models like temperature dependent variation in the rate of hydrogenation and alloying effect on H2wt% and H in different families of alloys. Importantly certain elements like Al, Si, Sc, Cr, and Mn when mixed in small fractions with hydriding elements like Mg, Ti, V etc. systematically reduce H without significantly compromising the storage capacity. Further upon increasing the number of elements in the alloy i.e from binary to ternary to quaternary, the number of compositions with lower enthalpies also increases. From the 6.4 million compositions, we have reported new alloy families having potential for hydrogen storage at room temperature. Finally, we demonstrate that HEART has the potential to scan vast chemical spaces by narrowing down potential materials for hydrogen storage.
  • Kavita Thakkar And Kavita Joshi, Single-atom alloys of Cu(211) with earth-abundant metals for enhanced activity towards CO dissociation, Journal of Molecular Graphics and Modelling., 126, 108656 (2024), DOI:https://doi.org/10.1016/j.jmgm.2023.108656.
    CO2, a byproduct from various industrial reactions, must not be released into the atmosphere and should be managed through capture, conversion, and utilization. The first step in converting CO2 into valuable products is to break the C–O bond. This work focuses on designing Single Atom Catalysts (SACs) by doping Cu(211) surface with 13 different s, p, and d block elements with an aim to minimize the activation barrier for C–O bond cleavage. Our work demonstrates that SACs of Mg/Al/Pt@Cu(211) favour CO2 chemisorption compared to Cu(211) where CO2 physisorbs. The barrier for CO2 dissociation is lowest for Mg@Cu(211) and it increases in the order Mg@Cu(211), Al@Cu(211), Pt@Cu(211), Zn@Cu(211) Ga@Cu(211) Cu@Cu(211) Pd@Cu(211). These findings suggest that doping Cu(211) with earth-abundant metals like Mg can potentially be a viable catalyst for CO2 conversion, providing a promising solution to reduce carbon footprint and mitigate climate change.
  • Aathira Nair And Kavita Joshi, What leads to direct epoxidation? An exhaustive DFT investigation of electrophilic oxygen-mediated epoxidation of ethylene on Ag(100), Computational Materials Science., , (2024), DOI:Just accepted.
  • Rohit Modee, Ashwini Verma, Kavita Joshi And U Deva Priyakumar, MeGen - generation of gallium metal clusters using reinforcement learning, Machine Learning: Science and Technology., 4, 025032-1 - 025032-9 (2023), DOI:10.1088/2632-2153/acdc03.
    The generation of low-energy 3D structures of metal clusters depends on the efficiency of the search algorithm and the accuracy of inter-atomic interaction description. In this work, we formulate the search algorithm as a reinforcement learning (RL) problem. Concisely, we propose a novel actor-critic architecture that generates low-lying isomers of metal clusters at a fraction of computational cost than conventional methods. Our RL-based search algorithm uses a previously developed DART model as a reward function to describe the inter-atomic interactions to validate predicted structures. Using the DART model as a reward function incentivizes the RL model to generate low-energy structures and helps generate valid structures. We demonstrate the advantages of our approach over conventional methods for scanning local minima on potential energy surface. Our approach not only generates isomer of gallium clusters at a minimal computational cost but also predicts isomer families that were not discovered through previous density-functional theory (DFT)-based approaches.
  • Shweta Mehta And Kavita Joshi, Electronic fingerprints for diverse interactions of methanol with various Zn-based systems, Surface Science., 736, 122350 (2023), DOI:https://doi.org/10.1016/j.susc.2023.122350.
    We have investigated various Zn-based catalysts for their interaction with methanol (MeOH). MeOH is one of the most critical molecules being studied extensively, and Zn-based catalysts are widely used in many industrially relevant reactions involving MeOH. We note that the same element (Zn and O, in the present study) exhibits different catalytic activity in different environments. The changing environment is captured in the underlying electronic structure of the catalysts. In the present work, we compared the electronic structure of Zn-based systems, i.e., ZnAl2O4 and ZnO along with oxygen preadsorbed Zn (O-Zn) and metallic Zn. We demonstrate the one-to-one correlation between the pDOS of the bare facet and the outcome of that facet’s interaction (i.e. either adsorption or dissociation of MeOH) with MeOH. These findings would pave the way towards the in-silico design of catalysts.
  • Sheena Agarwal, Shweta Mehta And Kavita Joshi*, Understanding the ML black box with simple descriptors to predict cluster–adsorbate interaction energy, New Journal of Chemistry., 44, 8545 - 8553 (2020), DOI:10.1039/D0NJ00633E.
    Density functional theory (DFT) is currently one of the most accurate and yet practical theories used to gain insight into the properties of materials. Although successful, the computational cost required is still the main hurdle even today. In recent years, there has been a trend of combining DFT with Machine Learning (ML) to reduce the computational cost without compromising accuracy. Finding the right set of descriptors that are simple to understand in terms of giving insights about the problem at hand, lies at the heart of any ML problem. In this work, we demonstrate the use of nearest neighbor (NN) distances as descriptors to predict the interaction energy between the cluster and an adsorbate. The model is trained over a size range of 5 to 75 atom clusters. When the training and testing is carried out on mutually exclusive cluster sizes, the mean absolute error (MAE) in predicting the interaction energy is ? 0.24 eV. MAE reduces to 0.1 eV when testing and training sets include information from the complete range. Furthermore, when the same set of descriptors are tested over individual sizes, the MAE further reduces to ?0.05 eV. We bring out the correlation between dispersion in the nearest neighbor distances and variation in MAE for individual sizes. Our detailed and extensive DFT calculations provide a rationale as to why nearest neighbor distances work so well. Finally, we also demonstrate the transferability of the ML model by applying the same recipe of descriptors to systems of different elements like (Na10), bimetallic systems (Al6Ga6, Li4Sn6, and Au40Cu40) and also different adsorbates (N2, O2, and CO).
  • Sheena Agarwal, Shweta Mehta, Nivedita Kenge, Siva Prasad Mekala, Vipul Patil, T. Raja And Kavita Joshi*, Mixed metal oxide: A new class of catalyst for methanol activation, Applied Surface Science., , (2020).
    In this work, we propose a mixed metal oxide as a catalyst and demonstrate it's ability to not only activate the MeOH molecule upon adsorption but also dissociate O-H and one of it's C-H bonds. MeOH activation is compared on two prominent facets of \znalo~ viz. (220) and (311). While spontaneous O-H bond dissociation is observed on both facets, C-H bond dissociates only on the (311) surface. Multiple factors like atomic arrangement and steps on the surface, coordination of surface atoms, and their effective charges have a combined effect on MeOH activation. The (311) surface offers higher catalytic activity in comparison with (220) surface. Having a stepped surface, availability of multiple sites, and variation in the charge distribution are some of the reasons for better catalytic performance of (311) facet. Effect of orientation of MeOH with respect to the surface adds both, information and complexity to the problem. Observations pertinent to understanding this effect are also reported. A detailed analysis of atomic arrangement on the two surfaces provides a rationale as to why MeOH gets dissociated spontaneously on the mixed metal oxide. The promising results reported here opens up a new class of catalyst for research.
  • Nivedita Kenge, Sameer Pitale And Kavita Joshi, The Nature of Electrophilic Oxygen : Insights from Periodic Density Functional Theory Investigations, Surface Science., , (2018), DOI:https://doi.org/10.1016/j.susc.2018.09.009.
    Increasing demand of ethylene oxide and the cost of versatile chemical ethene has been a driving force for understanding mechanism of epoxidation to develop highly selective catalytic process. Direct epoxidation is a proposed mechanism which in theory provides 100% selectivity. A key aspect of this mechanism is an electrophilic oxygen (Oele) species forming on the Ag surface. In the past two and half decades, large number of theoretical and experimental investigations have tried to elucidate formation of Oele on Ag surface with little success. Equipped with this rich literature on Ag-O interactions, we investigate the same using periodic DFT calculations to further understand how silver surface and oxygen interact with each other from a chemical standpoint. Based on energetics, Löwdin charges, topologies and pdos data described in this study, we scrutinize the established notions of Oele. Our study provides no evidence in support of Oele being an atomic species nor a diatomic molecular species. In fact, a triatomic molecular species described in this work bears multiple signatures which are very convincing evidence for considering it as the most sought for electrophilic entity.
  • Vaibhav Kaware and Kavita Joshi*, Scaling up the shape: A novel growth pattern of gallium clusters, J. Chem. Phys., 141, 054308 (2014), DOI:http://dx.doi.org/10.1063/1.4891867.
    Putative global minima for Ga+N clusters with size “N” ranging from 49 to 70 are found by employing the Kohn-Sham formulation of the density functional theory, and their evolution is described and discussed in detail. We have discovered a unique growth pattern in these clusters, all of which are hollow core-shell structures. They evolve with size from one spherical core-shell to the next spherical core-shell structure mediated by prolate geometries, with an increase in overall diameter of the core, as well as the shell, without putting on new layers of atoms. We also present a complete picture of bonding in gallium clusters by critically analyzing the molecular orbitals, the electron localization function, and Bader charges. Bonding in these clusters is a mixture of metallic and covalent type that leans towards covalency, accompanied by marginal charge transfer from the surface to the core. Most molecular orbitals of Ga clusters are non-jellium type. Covalency of bonding is supported by a wide localization window of electron localization function, and joining of its basins along the bonds.
  • A. Susan, A. Kibey, V. Kaware And K. Joshi*, Correlation between the variation in observed melting temperatures and structural motifs of the global minima of gallium clusters: An ab initio study, Journal of Chemical Physics., 138, 014303 (2013), DOI:http://dx.doi.org/10.1063/1.4772470.
    We have investigated the correlation between the variation in the melting temperature and the growth pattern of small positively charged gallium clusters. Significant shift in the melting temperatures was observed for a change of only few atoms in the size of the cluster. Clusters with size between 31?42 atoms melt between 500–600 K whereas those with 46?48 atoms melt around 800 K. Density functional theory based first principles simulations have been carried out on Ga+n clusters with n = 31,?…, 48. At least 150 geometry optimizations have been performed towards the search for the global minima for each size resulting in about 3000 geometry optimizations. For gallium clusters in this size range, the emergence of spherical structures as the ground state leads to higher melting temperature. The well-separated core and surface shells in these clusters delay isomerization, which results in the enhanced stability of these clusters at elevated temperatures. The observed variation in the melting temperature of these clusters therefore has a structural origin.
  • Kavita Joshi, Sailaja Krishnamurty And D. G. Kanhere, “Magic Melters” Have Geometrical Origin, Phys. Rev. Lett., 96, 135703 (2006), DOI:http://dx.doi.org/10.1103/PhysRevLett.96.135703.
    Recent experimental reports bring out extreme size sensitivity in the heat capacities of gallium and aluminum clusters. In the present work we report results of our extensive ab initio molecular dynamical simulations on Ga30 and Ga31, the pair which has shown rather dramatic size sensitivity. We trace the origin of this size sensitive heat capacities to the relative order in their respective ground state geometries. Such an effect of nature of the ground state on the characteristics of heat capacity is also seen in case of small gallium and sodium clusters, indicating that the observed size sensitivity is a generic feature of small clusters.
  • S. Chacko, Kavita Joshi, D. G. Kanhere And S. A. Blundell, Why Do Gallium Clusters Have a Higher Melting Point than the Bulk?, Phys. Rev. Lett., 92, 135506 (2004), DOI:http://dx.doi.org/10.1103/PhysRevLett.92.135506.
    Density functional molecular dynamical simulations have been performed on Ga17 and Ga13 clusters to understand the recently observed higher-than-bulk melting temperatures in small gallium clusters [G.?A. Breaux et al., Phys. Rev. Lett. 91, 215508 (2003)]. The specific-heat curve, calculated with the multiple-histogram technique, shows the melting temperature to be well above the bulk melting point of 303 K, viz., around 650 and 1400 K for Ga17 and Ga13, respectively. The higher-than-bulk melting temperatures are attributed mainly to the covalent bonding in these clusters, in contrast with the covalent-metallic bonding in the bulk.

Research Interest

  • Theory AND Computational Science

Contact Details

Kavita Joshi

Office: Convergence, MSM building, Near East Gate
Physical and Materials Chemistry Division
CSIR National Chemical Laboratory
Dr. Homi Bhabha Road
Pune 411008, India
Phone   +91 20 2590 2476
E-mail k.joshi@ncl.res.in