Unbiased Krylov subspace method for the extraction of ground state from lattice correlators†
Date: 12/12, 14:00~15:00
Room: D410 at Institutes of Natural Sciences
Speaker: Ryutaro Tsuji (KEK)
Abstract: Ground-state energy and matrix element are reconstructed from correlators in lattice QCD by diagonalizing transfer matrix T within the Krylov subspace spanned by T|χ>, where |χ> is a state generated by an interpolating field on the lattice. This method is valid but a problem arises when the data have statistical noise, which appears as spurious eigenvalues. To circumvent this problem, we introduce a low-rank approximation based on a singular-value decomposition of a matrix made of the correlators. The bias associated with the approximation is eliminated by an extrapolation to the limit of vanishing variance of energy eigenvalues. In this seminar, I will introduce our method using a set of mock data and apply it to the real lattice data.
Deconfinement-Higgs continuity in SU(2) adjoint Higgs model at finite temperature†
Date: 11/28, 14:00~15:00
Room: CCS ワークショップ室 (1F)
Speaker: Masashi Kawahira (Kobe University)
Abstract: We study the finite-temperature phase structure of the four-dimensional SU(2) adjoint Higgs model, focusing on a possible deconfinement-Higgs continuity: the conjecture that the high-temperature deconfined phase of Yang-Mills theory and the finite-temperature Higgs phase form a single thermodynamic phase.
In this talk, we examine this conjecture using the following three approaches: (i) global symmetry analysis; (ii) a deformation analysis; and (iii) Hybrid Monte Carlo analysis on 16^3×8 and 12^3×6 lattices.
These results indicate that the Higgs and deconfined phases can be continuously connected, while the confined phase remains distinct.
The Worldvolume Hybrid Monte Carlo method: Its foundations and applications†
Date: 11/14, 14:00~15:00
Room: D413 at Institutes of Natural Sciences
Speaker: Masafumi Fukuma (Kyoto University)
Abstract: The sign problem has long been a major obstacle to first-principles calculations in a wide range of areas in physics. The Worldvolume Hybrid Monte Carlo (WV-HMC) method [Fukuma–Matsumoto, arXiv:2012.08468] is a reliable and versatile algorithm proposed to overcome this difficulty. While based on the Lefschetz-thimble approach, this method performs Hybrid Monte Carlo simulations on a continuous union of deformed integration surfaces in complex space (the worldvolume), thereby resolving the ergodicity problem inherent in conventional approaches at low computational cost. In this talk, I will first give an overview of the numerical sign problem and then explain the basic framework of the WV-HMC method. I will also present recent results on its applications to various lattice field theories. (Based on arXiv:2506.12002, 2507.23748, 2508.02659, and ongoing work.)
Abstract: For quantum field theories with global symmetry, we can study the behavior of the partition function with the background gauge field to diagnose different quantum phases. For the case of discrete symmetries, we find that the symmetry-twisted partition function works as an order parameter that discriminates spontaneous symmetry breaking (SSB), symmetry-protected topological (SPT) states, and symmetry-enriched topological (SET) states. We also study its behavior for the spontaneously broken U(1) symmetry and interpret the result from the viewpoint of the mixed anomaly with the emergent solitonic symmetry.
Large N and SO(1,3) in Euclidean IKKT Matrix Model†
Date: 9/24, 14:00~15:00
Room: D413 at Institutes of Natural Sciences
Speaker: Henry Liao (National Taiwan University)
Abstract: IKKT matrix model is a proposal toward non-perturbative formulation of superstring theory, which possesses various interesting features. For example, the dynamics of this zero-dimensional theory describes spacetime as an emerged quantity from matrix degrees of freedom. Naturally, researchers have been, in this framework, searching for emerged spacetime with proper symmetry group, which is SO(1,3), the Lorentz symmetry. In this talk, we start with toy models to demonstrate how spacetime can be dynamically determined (or emerged) from matrices. Then, we derive the form of IKKT matrix model from string theory and discuss the known classical saddles that are shared by both of its Euclidean and Lorentzian versions. From symmetry point of view, one expect to start from Lorentzian version, since it possesses SO(1,9) symmetry and SO(1,3) is a symmetry breaking pattern of it. However, with a simple and explicit calculation, we show that the Euclidean IKKT matrix model has SO(1,3)-symmetric saddle. This saddle allows us to extend a new class of solutions which is qualitatively different from any of the known solutions. We then argue that to faithfully emerge a metric from so(1,3) solution in Euclidean spacetime, a similar picture of Vilenkin to create universe through tunnelling emerges.
A General Method to Locate a Critical Point Using Ratios of Lee-Yang Zeros†
Date: 7/25, 14:00~15:00
Room: D413 at Institutes of Natural Sciences
Speaker: Tatsuya Wada (YITP)
Abstract: In recent years, research aimed at locating the QCD critical point using Lee-Yang zeros (LYZs) has been actively pursued. Lee-Yang zeros are the points in the complex parameter space at which the partition function equals to zero. In this talk, we propose a novel and general method, the Lee-Yang Zero Ratio (LYZR) method, to determine the location of the critical point. The LYZR is defined as the ratio of the imaginary parts of two Lee–Yang zeros located at different distances from the real axis. The method exploits the fact that LYZR values obtained at several spatial volumes intersect at a critical point. Because this crossing arises from the finite-size scaling of Lee–Yang zeros, the method is applicable to critical points in a wide variety of systems. Furthermore, we present results from the application of the LYZR method to the Ising model, the Potts model, and heavy-quark QCD systems, demonstrating its effectiveness in determining the location of the critical point.
Abstract: We investigate fermion systems on a square lattice with a mass term featuring a curved domain-wall. Similar to traditional flat domain-wall fermions, massless and chiral edge states emerge on the wall. In the cases of S^1 and S^2 domain walls embedded into flat hypercubic lattices, we find that these edge modes experience gravity through the induced spin or spin^c connections. The lattice results align well with the continuum predictions.
In this talk, we examine a fermion system with multiple domain-walls. As with the single domain-wall system, low-energy modes are localized at the junctions of these domain-walls. Since 4D Schwarzschild space can be embedded in 6D Euclidean space, our setup allows the simulation of a fermionic system in the vicinity of a black hole.
Bridging two semiclassical confinement mechanisms: monopole and center vortex†
Date: 6/27, 14:00~15:00
Room: D413 at Institutes of Natural Sciences
Speaker: Yui Hayashi (YITP)
Abstract: Two promising scenarios for quark confinement are based on monopole condensation and center-vortex proliferation. Recently, these two scenarios have been realized in the weakly-coupled semiclassical frameworks: one as a dilute monopole gas with S^1 compactification, and the other as a dilute center-vortex gas with T^2 compactification. In this talk, after reviewing these two approaches, we connect the two semiclassical frameworks, clarifying the relationship between monopoles and vortices. This result establishes the explicit, long-expected link between the two pictures in a semiclassically tractable regime. If time permits, we will also discuss further developments, such as the application to N=1 super-Yang-Mills theory.
Diffusion Models and Path Integrals: Interpreting Generative Models in the Language of Physics†
Date: 6/6, 14:00~15:00
Room: CCS 会議室B (1F)
Speaker: Yuji Hirono (Institute of Systems and Information Engineering, University of Tsukuba)
Abstract: In recent years, research at the interface of physics and machine learning has rapidly gained momentum. This emerging field includes two complementary directions: using machine learning to study and model physical phenomena, and using ideas from physics to better understand the behavior of machine learning models. In this talk, I will present a recent study that takes the latter perspective and reformulates diffusion models—widely used in image generation—in the language of path integrals. This formulation reveals a structural analogy between the sampling schemes in diffusion models and the relationship between quantum and classical dynamics, providing a framework for analyzing the model's behavior using tools from theoretical physics.
Does connected wedge imply distillable entanglement?†
Date: 5/9, 14:00~15:00
Room: D413 at Institutes of Natural Sciences
Speaker: Takato Mori (Rikkyo University)
Abstract: In holography, it is believed that entanglement implies spatial connectivity and vice versa. Indeed, entanglement wedge reconstruction implies that connected entanglement wedge between two boundary subsystems implies nontrivial entanglement between them. Nevertheless, its nature remains ellusive. We argue that there is no distillable entanglement, which is the amount of EPR pairs that can be converted from a state, via local operations. Even if we extend the possible operations to LOCC, including classical communication, we conjecture that there exists a regime where connected wedge does not imply any distillable entanglement. We present a concrete LOCC protocol in the holographic setup and the bulk dual formula for the distillable entanglement. It involves holographic measurements and one-way classical communication, which are made manifest in a random tensor network model. Time permitting, a measurement-based traversable wormhole, the connected wedge theorem, and more general holographic quantum correlations will be discussed.
Lattice QCD studies on Lambda(1405) in the flavor SU(3) limit†
Date: 4/25, 14:00~15:00
Room: CCS 会議室B (1F)
Speaker: Kotaro Murakami (Institute of Science Tokyo)
Abstract: We perform a numerical study in lattice QCD on $\Lambda(1405)$ in the flavor SU(3) limit. Previous studies based on the chiral symmetry have suggested that the spectrum corresponding to $\Lambda(1405)$ observed in experiments may be explained by a combination of two poles. To elucidate such property from lattice QCD, we first calculate meson-baryon potentials in the flavor SU(3) limit using the HAL QCD method. We find that the local potentials have singular behaviors, which prevent us from reliably extracting binding energies. To avoid such singular behaviors, we introduce a separable potential instead of the standard local approximation. Our results of separable potentials show attraction to produce bound states.
![[PukiWiki]](image/pukiwiki.png "[PukiWiki]")
