SearcharxivSearch

arXiv subjects

Heejae Kim

Publications and source records attributed to Heejae Kim.

12 recordsLinked to original sources

The Thick Part of the $\mathrm{PSL}_n(\mathbb{R})$-Hitchin-Riemann Moduli Space has Infinite Volume

We prove that the thick part of the $\mathrm{PSL}_n(\mathbb{R})$-Hitchin-Riemann moduli space has infinite total Atiyah--Bott--Goldman volume for $n>2$. This result stands in contrast to Mumford's compactness criterion. To achieve this result, we employ Goldman flows and internal sequences to find an infinite series of subsets of identical volume, the images of which in the Hitchin-Riemann moduli space are all mutually disjoint and sit in the thick part.

math.GT

SCENIC: Stream Computation-Enhanced SmartNIC

Although modern, AI-centric datacenters heavily rely on SmartNICs, existing devices impose a hard trade-off. Commercial SmartNICs provide high bandwidth and easy software integration, but offer limited support for customization and data processing offload. In contrast, research SmartNICs often suffer from low bandwidth, limited functionality, and poor software compatibility -- to the point that many are not actual NICs in a technical sense. This gap can be closed by treating the NIC datapath as a first-class stream computation substrate with shared hardware/software abstractions for a tight co-design of infrastructure and applications. To demonstrate this, we introduce SCENIC, an open-source datacenter SmartNIC. SCENIC implements a 200G network datapath over offloaded TCP/IP and RDMA stacks, as well as a fallback path for processing arbitrary network traffic. On top of the network logic, SCENIC combines on-datapath Stream Compute Units (SCUs) for data processing and embedded ARM cores for flexible control path manipulation with direct access to GPUs and SSDs. SCENIC is fully integrated with the OS, exposing native Linux network and RDMA verb interfaces, making the programmable datapath transparent to existing applications while enabling control of, e.g., user-defined offloads and programmable congestion control. SCENIC's performance matches commercial platforms, and we show its versatility through several use cases such as offloaded collective communication and network-to-GPU hash-based data partitioning.

cs.AR

Generalized thermodynamic closure in ultrafast phonon dynamics

Driven-dissipative dynamics underlie a wide range of nonequilibrium phenomena in quantum materials, yet reduced descriptions beyond the quasi-equilibrium picture remain difficult to establish. Here, we experimentally demonstrate that a resonantly driven phonon mode admits a generalized thermodynamic description in which coherence and energy jointly organize the nonequilibrium evolution. Beyond a threshold driving field strength, we observe a delayed ultrafast response of a coherently driven phonon mode. Combined with experimentally constrained Lindblad dynamics, we show that this delay reflects the finite-time spreading of excitations across many phonon levels. At the same time, the full density-matrix trajectories for three driving conditions collapse onto a common surface defined by energy and coherence. Our results establish a coherence-extended thermodynamic regime for driven phonons and provide a framework for broader state engineering in driven-dissipative bosonic excitations.

cond-mat.mtrl-sci

Scalable Reactive Atomistic Dynamics with GAIA

Groundbreaking advances in materials and chemical research have been driven by the development of atomistic simulations. However, the broader applicability of atomistic simulations remains limited, as they inherently depend on energy models that are either approximate or computationally prohibitive for large-scale simulations. Machine learning interatomic potentials (MLIPs) have recently emerged as a promising class of energy models, but their deployment also remains challenging due to the scarcity of systematic protocols for generating training data spanning diverse structural regimes. Here we introduce GAIA, an end-to-end automated framework that streamlines dataset construction for the development of general-purpose reactive MLIPs. GAIA combines a metadynamics-based exploration scheme with closed-loop data expansion for the efficient sampling of a broad spectrum of atomic arrangements, thereby addressing the reliance on heuristics in conventional dataset generation. Using GAIA, we constructed Titan25, a benchmark-scale dataset, and trained an MLIP that closely matches both static and dynamic density functional theory results. The resulting model reproduces key experimental observations across distinct modes of reactivity, including detonation, coalescence, and catalytic processes. GAIA thus helps bridge the gap between simulation and experiment, paving the way toward scalable and general MLIPs capable of describing a wide range of materials and chemical processes.

physics.chem-ph

Temperature- and charge carrier density-dependent electronic response in methylammonium lead iodide

Understanding carrier dynamics in photoexcited metal-halide perovskites is key for optoelectronic devices such as solar cells (low carrier densities) and lasers (high carrier densities). Trapping processes at low carrier densities and many-body recombination at high densities can significantly alter the dynamics of photoexcited carriers. Combining optical-pump/THz probe and transient absorption spectroscopy we examine carrier responses over a wide density range (10^14-10^19 cm-3) and temperatures (78-315K) in the prototypical methylammonium lead iodide perovskite. At densities below ~10^15 cm-3 (room temperature, sunlight conditions), fast carrier trapping at shallow trap states occurs within a few picoseconds. As excited carrier densities increase, trapping saturates, and the carrier response stabilizes, lasting up to hundreds of picoseconds at densities around ~10^17 cm-3. Above 10^18 cm-3 a Mott transition sets in: overlapping polaron wavefunctions lead to ultrafast annihilation through an Auger recombination process occurring over a few picoseconds. We map out trap-dominated, direct recombination-dominated, and Mott-dominated density regimes from 78-315 K, ultimately enabling the construction of an electronic phase diagram. These findings clarify carrier behavior across operational conditions, aiding material optimization for optoelectronics operating in the low (e.g. photovoltaics) and high (e.g. laser) carrier density regimes.

cond-mat.mtrl-sci

Anisotropic electron-phonon interactions in 2D lead-halide perovskites

Two-dimensional hybrid organic-inorganic metal halide perovskites offer enhanced stability for perovskite-based applications. Their crystal structure's soft and ionic nature gives rise to strong interactions between charge carriers and ionic rearrangements. Here, we investigate the interaction of photo-generated electrons and ionic polarizations in single-crystal 2D perovskite butylammonium lead iodide, varying the inorganic lammelae thickness in the 2D single crystals. We determined the directionality of the transition dipole moments of the relevant phonon modes (in the 0.3-3 THz range) by angle-and-polarization dependent THz transmission measurements. We find a clear anisotropy of the in-plane photoconductivity, with a 10% reduction along the axis parallel with the transition dipole moment of the most strongly coupled phonon. Detailed calculations, based on Feynman polaron theory, indicate that the anisotropy originates from directional electron-phonon interactions.

cond-mat.mtrl-sci

Breaking MLPerf Training: A Case Study on Optimizing BERT

Speeding up the large-scale distributed training is challenging in that it requires improving various components of training including load balancing, communication, optimizers, etc. We present novel approaches for fast large-scale training of BERT model which individually ameliorates each component thereby leading to a new level of BERT training performance. Load balancing is imperative in distributed BERT training since its training datasets are characterized by samples with various lengths. Communication cost, which is proportional to the scale of distributed training, needs to be hidden by useful computation. In addition, the optimizers, e.g., ADAM, LAMB, etc., need to be carefully re-evaluated in the context of large-scale distributed training. We propose two new ideas, (1) local presorting based on dataset stratification for load balancing and (2) bucket-wise gradient clipping before allreduce which allows us to benefit from the overlap of gradient computation and synchronization as well as the fast training of gradient clipping before allreduce. We also re-evaluate existing optimizers via hyperparameter optimization and utilize ADAM, which also contributes to fast training via larger batches than existing methods. Our proposed methods, all combined, give the fastest MLPerf BERT training of 25.1 (22.3) seconds on 1,024 NVIDIA A100 GPUs, which is 1.33x (1.13x) and 1.57x faster than the other top two (one) submissions to MLPerf v1.1 (v2.0). Our implementation and evaluation results are available at MLPerf v1.1~v2.1.

cs.LG

Signatures of Mode-Resolved, Nonlocal Electron-Phonon Coupling in Two-Dimensional Spectroscopy

Electron-phonon coupling (EPC) is foundational in condensed matter physics, determining intriguing phenomena and properties in both conventional and quantum materials. In this manuscript, we propose and demonstrate a novel two dimensional (2D) EPC spectroscopy which allows for direct extraction of EPC matrix elements for specific phonon modes and different electron energies, simultaneously. Using this technique, we are able to measure the electron-energy dependence of the EPC strength for individual phonon modes. This capability allows us to identify unique signatures distinguishing nonlocal Su-SchriefferHeeger (SSH)-type couplings from local Holstein-type couplings. In application to a methylammonium lead iodide (MAPI) perovskite, we find that two pronounced phonon modes at room temperature exhibit highly distinctive EPC behaviors, concerning strength, anisotropy, and temperature-dependence across the structural phase transition. Our approach paves the way for unraveling the microscopic origin of EPC, the change of the phonon-mode-specific EPC with external conditions, and phonon-mediated ultrafast control of condensed materials.

cond-mat.mtrl-sci

Controlling the electro-optic response of a semiconducting perovskite coupled to a phonon-resonant cavity

Optical cavities, resonant with vibrational or electronic transitions of material within the cavity, enable control of light-matter interaction. Previous studies have reported cavity-induced modifications of chemical reactivity, fluorescence, phase behavior, and charge transport. Here, we explore the effect of resonant cavity-phonon coupling on the transient photoconductivity in a hybrid organic-inorganic perovskite. To this end, we measure the ultrafast photoconductivity response of perovskite in a tunable Fabry-Perot terahertz cavity, designed to be transparent for optical excitation. The terahertz-cavity field-phonon interaction causes apparent Rabi splitting between the perovskite phonon mode and the cavity mode. We explore whether the cavity-phonon interaction affects the material electron-phonon interaction by determining the charge carrier mobility through the photoconductivity. Despite the apparent hybridization of cavity and phonon modes, we show that the perovskite properties, in both ground (phonon response) and excited (photoconductive response) states, remain unaffected by the tunable light-matter interaction. Yet the response of the integral perovskite-terahertz optical cavity system depends critically on the interaction strength of the cavity with the phonon: the transient terahertz response to optical excitation can be increased up to 3-fold by tuning the cavity-perovskite interaction strength. These results enable tunable switches and frequency-controlled induced transparency devices.

physics.app-ph

Theoretical analysis of glide-Z_2 magnetic topological photonic crystals

Gapped systems with glide symmetry can be characterized by a Z_2 topological invariant. We study the magnetic photonic crystal with a gap between the second and third lowest bands, which is characterized by the nontrivial glide-Z_2 topological invariant that can be determined by symmetry-based indicators. We show that under the space group No. 230 (Ia-3d), the topological invariant is equal to half of the number of photonic bands below the gap, and therefore, the band gap between the second and third lowest bands is always topologically nontrivial, and to realize the topological phase, we need to open a gap for the Dirac point at the P point by breaking time-reversal symmetry. With staggered magnetization, the photonic bands are gapped, and the photonic crystal becomes topological, whereas with uniform magnetization, a gap does not open, which can be attributed to the minimal band connectivity exceeding two in this case. By introducing the notion of Wyckoff positions, we show how the topological characteristics are determined from the structure of the photonic crystals.

physics.optics

Glide-symmetric topological crystalline insulator phase in a nonprimitive lattice

We study the topological crystalline insulator phase protected by the glide symmetry, which is characterized by the Z2 topological number. In the present paper, we derive a formula for the Z2 topological invariant protected by glide symmetry in a nonprimitive lattice, from that in a primitive lattice. We establish a formula for the glide-Z2 invariant for the space group No. 9 with glide symmetry in the base-centered lattice, by folding the Brillouin zone into that of the primitive lattice where the formula for the glide-Z2 invariant is known. The formula is written in terms of integrals of the Berry curvatures and Berry phases in the k-space. We also derive a formula of the glide-Z2 invariantwhen the inversion symmetry is added, and the space group becomes No. 15. This reduces the formula into the Fu-Kane-like formula, expressed in terms of the irreducible representations at high-symmetry points in $k$ space. We also construct these topological invariants by the layer-construction approach, and the results completely agree with those from the k-space approach.

cond-mat.str-el

Glide-symmetric magnetic topological crystalline insulators with inversion symmetry

It is known that three-dimensional magnetic systems with glide symmetry can be characterized by a $Z_2$ topological invariant together with the Chern number associated with the normal vector of the glide plane, and they are expressed in terms of integrals of the Berry curvature. In the present paper, we study the fate of this topological invariant when inversion symmetry is added while time-reversal symmetry is not enforced. There are two ways to add inversion symmetry, leading to space groups No.~13 and No.~14. In space group No.~13, we find that the glide-$Z_2$ invariant is expressed solely from the irreducible representations at high-symmetry points in $k$-space. It constitutes the $\mathbb{Z}_2\times \mathbb{Z}_2$ symmetry-based indicator for this space group, together with another $\mathbb{Z}_2$ representing the Chern number modulo 2. In space group No.~14, we find that the symmetry-based indicator $\mathbb{Z}_2$ is given by a combination of the glide-$Z_2$ invariant and the Chern number. Thus, in space group No.~14, from the irreducible representations at high-symmetry points we can only know possible combinations of the glide-$Z_2$ invariant and the Chern number, but in order to know each value of these topological numbers, we should calculate integrals of the Berry curvature. Finally, we show that in both cases, the symmetry-based indicator $\mathbb{Z}_4$ for inversion symmetric systems leading to the higher-order topological insulators is directly related with the glide-$Z_2$ invariant and the Chern number. As an independent approach to these results, we also construct all invariants from the layer construction for these space groups, and we show complete agreement with the above results for the topological invariants constructed from $k$-space topology.

cond-mat.mes-hall