Strongly Correlated Materials — a Korean matinée

Schedule

09:30am – Ara Go

Data-Driven Bath Fitting for Hamiltonian-Diagonalization DMFT

10:00am – Heung-Sik Kim

Correlation-induced flat bands in two- and three-dimensional transition metal compounds

10:30am – Choong Hyun Kim

From Jahn-Teller Distortions to Spin-Orbital Entanglement: Pathways to Mott Metal-Insulator

11:00am – Bongjae Kim

Strain-Engineered Ruthenates: Controlling Correlated Quantum Phases

Abstracts

Ara Go – Data-Driven Bath Fitting for Hamiltonian-Diagonalization DMFT

Department of Physics, Chonnam National University, Gwangju, Republic of Korea

We develop a machine-learning-based initialization strategy that alleviates a major practical bottleneck of Hamiltonian-diagonalization-based dynamical mean-field theory (HD-DMFT): nonlinear bath fitting. In HD-DMFT, the continuous hybridization function is approximated by a finite set of bath-site energies and hybridization amplitudes obtained from a highly non-convex multivariable optimization. As the number of bath sites increases, this optimization becomes increasingly sensitive to the initial guess and prone to trapping in suboptimal local minima, slowing or destabilizing the DMFT self-consistency loop.

We recast bath fitting as a supervised regression problem and train a kernel ridge regression model to predict near-optimal discrete bath parameters directly from the target hybridization function on the Matsubara axis. A key methodological element is a physically grounded data-generation protocol: rather than random sampling, we construct a diverse training set from tight-binding Hamiltonians of layered-perovskite-like ruthenate models across systematically deformed structures, with high-quality labels obtained from fully converged conventional bath fitting. Time-reversal symmetry is explicitly incorporated in both feature and target representations, reducing effective dimensionality and enforcing physical consistency.

Benchmarks in the non-interacting limit show that the learned initialization systematically reduces the initial fitting error, decreases the number of conjugate-gradient iterations, and improves robustness against local minima over a wide range of bath sizes. We further demonstrate transferability to an interacting HD-DMFT calculation for Sr2RuO4 solved with an adaptive-truncation impurity solver, where the ML initialization yields consistently faster convergence than a symmetry-preserving heuristic baseline while preserving the final fitted solution.

Heung-Sik Kim – Correlation-induced flat bands in two- and three-dimensional transition metal compounds

Department of Energy Engineering, Korea Institute of Energy Technology, Naju-si 58330, Republic of Korea

The exploration of electronic flat bands has recently attracted considerable attention due to their potential to host emergent quantum phenomena, including fractional Chern insulating phases, charge density waves, and superconductivity. Numerous theoretical proposals and candidate materials have been suggested; however, realizing flat bands near the Fermi level remains challenging due to the intricate requirements of lattice structure, chemical composition, and symmetry considerations. In this study, we propose an alternative pathway to achieve flat bands, employing electron correlations in quasi-two- and three-dimensional transition-metal-based compounds.

Firstly, we examine α-RuI3, demonstrating that the interplay between spin-orbit coupling and on-site Coulomb interactions stabilizes flat bands near the Fermi level, thereby underpinning the compound’s peculiar bad-metallic behavior[1,2]. Secondly, we discuss Sc3Mn3Al7SI5, illustrating that flat bands emerge due to the synergy between geometric frustration and Coulomb interactions; these correlation-driven flat bands are proposed as the source of intrinsic ferromagnetic fluctuations observed in this nominally nonmagnetic metal[3]. Lastly, we show our preliminary results on a spinel compound CuV2S4, employing our recent complex-time tensor-network impurity solver[4,5]. This work highlights the crucial role of electron correlations combined with structural and geometric factors in engineering and identifying flat-band systems.

[1] S. Samanta, D. Hong, and H.-S. Kim, Nanomaterials 14, 9 (2023).

[2] D. A. S. Kaib et al., npj Quant. Mater. 7, 75 (2022).

[3] S. Samanta et al., Nat. Commun. 15, 5376 (2024).

[4] J. Huang et al., Nat. Phys. 20, 603-609 (2024).

[5] J. Cha, H.-Y. Lee, H.-S. Kim, Sci. Rep. 15, 37490 (2025).  

Choong Hyun Kim  – From Jahn-Teller Distortions to Spin-Orbital Entanglement: Pathways to Mott Metal-Insulator

Department of Physics, Ajou University, Suwon, Republic of Korea

Mott metal-insulator transition (Mott MIT) is one of the most prominent emergent phenomena in strongly correlated systems. Unlike conventional band insulators, Mott MIT arises from the interplay of Coulomb interaction, spin-orbit coupling, and lattice distortion, leading to novel phase transitions and electronic states. In multi-orbital systems, the lifting of orbital degeneracy plays a crucial role in determining the MIT. This can be driven by Jahn-Teller distortion, spin-orbit coupling, or a complex interplay of Hund’s coupling and Coulomb U, which can either compete or cooperate in shaping the phase transition.

In this presentation, we will discuss how key factors such as Coulomb U, lattice distortion, Hund’s coupling, and spin-orbit coupling interact to produce intriguing phase diagrams in strongly correlated materials. Specifically, we will focus on two systems: the CuAl₂O₄ spinel, where strong spin-orbit coupling leads to a novel spin-orbital-entangled state, and monolayer SrRuO₃ thin films, where a new type of Jahn-Teller distortion drives the metal-insulator transition. These cases highlight the rich physics underlying Mott MIT and provide insights into the broader implications for correlated electron systems.

Bongjae Kim  – Strain-Engineered Ruthenates: Controlling Correlated Quantum Phases

Department of Physics, Kyungpook National University, Dague, Republic of Korea

Layered ruthenates provide a clean platform in which small changes of lattice geometry reshuffle the hierarchy among superconducting, magnetic, and metallic ground states. In this talk, we will discuss how uniaxial and epitaxial strain can be used as tuning parameters to manipulate correlated quantum phases in these systems. For Sr2RuO4, we will discuss the microscopic origin of the counterintuitive emergence of static magnetism under uniaxial strain, underlining the key role played by spin fluctuations. We will further show uniaxial strain can induce lattice softening driven by electronic instabilities and can stabilize distinct magnetic phases. We will then discuss the ruthenate heterostructures, where dimensional confinement and epitaxial strain provide an effective route to engineer competing electronic and magnetic phases that go beyond what can be reached in bulk layered ruthenates.