応用物理学輪講 I
12月20日
[注意事項]
発表の10日前までに　office[at]ap.t.u-tokyo.ac.jp　宛てに「氏名」「指導教員」「発表題目（英語）」「要旨（英語）」「発表言語（英語または日本語）」を送付して下さい。
発表日
2024年12月20日（金）16：50～18：50

Aグループ

座長
平田　裕也
指導
教員名
齊藤　英治　教授
座長
町永　明海
指導
教員名
武田　俊太郎　准教授
発表者名 山下　涼介
指導教員名 森本　高裕　准教授
発表題目（英語） Tensor network simulation of nonlinear optical phenomena on multiferroic magnets
要旨（英語） The development of laser technology has led to recent progress in the theoretical and experimental research on nonlinear optical phenomena. In particular, the second-order DC response of noncentrosymmetric materials has attracted attention as the bulk photovoltaic effect (BPVE). While the BPVE has mainly been studied in free electronic systems, it has recently been proposed to be extended to multiferroic magnets via electromagnon excitation in strongly correlated electron systems.
In this talk, I will introduce the application of the infinite Time-Evolving Block Decimation (iTEBD), a numerical method for one-dimensional quantum systems using tensor network formalism, to the nonlinear optical response of 1D multiferroic magnets. Through non-perturbative calculations by iTEBD, we observe the non-perturbative effect of the ground state and excitation structures to the optical response, as well as the quantitative property of high harmonic generation.
発表言語 日本語
発表者名 陳　正力
指導教員名 中村　泰信　教授
発表題目（英語） Developing cos μφ elements in superconducting circuits
要旨（英語） In superconducting circuits, Josephson Junctions are widely used to construct superconducting qubits and various devices. Compared to conventional inductors, Josephson Junctions exhibit non-linear property, with their potential energy being proportional to the cosine of gauge-invariant phase (φ). Generalized Josephson Junctions (cos μφ elements), where μ is an integer, exhibit a 2π/μ periodic potential in the phase space. While cos 2φ elements have been realized in previous research, cos μφ elements with μ > 2 remain challenging and require further development. Cos μφ elements are expected to extend circuit design flexibility, enabling a wide range of application, such as the development of noise-protected 0-π qubits, GKP states in quantum information, quantum simulation of arbitrary Hamiltonian, and more.
発表言語 英語
発表者名 畠中　友也
指導教員名 森本　高裕　准教授
発表題目（英語） Denoising Diffusion Error Decoding for Topological Code
要旨（英語） We are investigating the potential of diffusion models, which have gained prominence in image generation, to function as effective decoders for topological error-correcting codes, such as surface codes.
Specifically, we conceptualize the generation of quantum errors as a forward diffusion process, where errors accumulate incrementally over time, akin to noise propagation. Decoding, in turn, is treated as a reverse diffusion process, aiming to reconstruct the original, error-free state by iteratively eliminating errors. Our objective is to rigorously evaluate whether this novel approach, leveraging the inherent strengths of diffusion models, can achieve higher accuracy and efficiency compared to traditional decoding algorithms commonly used in fault-tolerant quantum computing. By benchmarking its performance across various error rates and code distances, we hope to uncover new insights into its applicability and advantages.
発表言語 英語
発表者名 吉冨　成哉
指導教員名 求　幸年　教授
発表題目（英語）
要旨（英語）
発表言語

Bグループ

座長
三島　萌登
指導
教員名
小芦　雅斗　教授
座長
三宅　孝明
指導
教員名
沙川　貴大　教授
発表者名 山根　悠都
指導教員名 小芦　雅斗　教授
発表題目（英語） Twin-Field Quantum Key Distribution with Finite-Key analysis overcoming the Rate-Distance limit
要旨（英語） Quantum Key Distribution (QKD) provides information-theoretic security for cryptographic key sharing.
Conventional QKD protocols are limited by the O(η) scaling of the key generation rate due to channel losses.
TF-QKD, an innovative approach, achieves O(√η) scaling, overcoming these limitations.
However, existing security proofs rely heavily on the asymptotic regime, assuming infinite resources, which is impractical for real-world implementations.
In this paper, the key contribution is below:
Protocol Design:

Combines signal mode (key generation) and test mode (error estimation).
Utilizes the Operator Dominance Condition to approximate non-classical states and simplify phase error estimation.
Security Proof:

Replaces asymptotic assumptions with a finite-size security framework.
Constructs bounds for the phase error rate using statistical tools like Bernoulli sampling and Poisson distribution.
Numerical Results:

Demonstrates the protocol's ability to surpass the PLOB bound with realistic parameters (
N
=
1
0
12
pulses).
Confirms feasibility within practical timeframes (e.g., 20 minutes at a 1 GHz repetition rate).
発表言語 日本語
発表者名 渡邉　楓花
指導教員名 福谷　克之　教授
発表題目（英語） The mechanism and evaluation of a spin polarized atomic hydrogen beam
要旨（英語） A hydrogen atom consists of a proton and an electron, each of which is a Fermion with spin 1/2. The energy levels of a hydrogen atom have fine and hyperfine structures due to spin-orbit interaction. If an electron is in the 1 s orbital, a hydrogen atom does not have fine structure, and the total spin quantum number becomes 1 or 0 due to the hyperfine interaction.
Therefore, a hydrogen atom in its ground state is a compound boson. Because of this simple structure, hydrogen atoms play an important role in various fundamental physics studies such as atomic collisions.
In this study, we are developing and evaluating a spin-polarized atomic hydrogen beam. In previous studies, spin-selection by a hexagonal magnetic field and spatial spin-separation by a Stern-Gerlach magnet have been performed.
In this presentation, we will describe the mechanism of the spin-polarized hydrogen beam, the evaluation of the beam done in the previous study, and future prospects for the development of the beam.
発表言語 日本語
発表者名 尹　建
指導教員名 齊藤　英治　教授
発表題目（英語） Physical Deep Learning via Thermodynamic Correspondence: Free Energy Minimization in Stochastic Landau-Lifshitz-Gilbert Simulations
要旨（英語） In recent years, the application of physical systems to emulate the optimization processes of deep learning has gained significant attention. The foundational work by Hopfield, employing spin-glass models to define associative memory, laid the theoretical groundwork for this domain. While the fundamental principles of deep learning algorithms are rooted in physical systems, much of the progress in artificial intelligence has been achieved through abstractions above the Turing machine paradigm. This research aims to bridge the gap between AI and physical computation by leveraging the principles of thermodynamics and stochastic spin systems. Specifically, we investigate the use of free energy minimization within spin systems, modeled via the Stochastic Landau-Lifshitz-Gilbert equation, as a computational mechanism for deep learning optimization. In this presentation, we discuss the theoretical underpinnings of physical deep learning, provide a detailed correspondence between thermodynamic free energy landscapes and optimization in neural networks, and demonstrate proof-of-concept results from SLLG simulations.
発表言語 英語
応用物理学輪講 I一覧ページへ戻る