生成式人工智能、實用化哈密頓量子模擬、氫硼聚變反應 | 本週物理講座
1
報告人：王磊，中國科學院物理研究所
時間：7月18日（週四）10:00
單位：中國科學院物理研究所
地點：M樓236會議室
摘要：
長期以來計算機最顯著的特點是算得快、算得準。生成式人工智能賦予了計算機從經驗中學習“直覺”的能力，甚至可以用來“創造”新的經驗。定量描述這種“直覺”和“創造”的數學工具和統計物理一樣，都是自然界中的概率分佈。我將從物理學視角解讀ChatGPT和Sora等生成式人工智能產品背後的關鍵技術：生成模型。生成模型學習數據背後的概率分佈，並據此隨機採樣產生新的數據樣本。基於和統計物理的深層聯繫，生成模型在原子尺度物質結構設計和變分自由能計算中具有特別的優勢。我將以晶體按需生成和稠密氫狀態方程爲例說明生成模型在物質科學研究中的應用前景。
報告人簡介：
王磊2006年本科畢業於南京大學，2011年在中國科學院物理研究所獲得博士學位，此後在蘇黎世聯邦理工學院從事計算量子物理的博士後研究，2016年加入中國科學院物理研究所工作。主要研究方向是機器學習與量子多體計算的交叉領域。
王磊
2
報告人：龍江，華中科技大學
時間：7月17、18、19日 10:00
單位：中國科學院理論物理所
地點：南樓6620
摘要：
Flat holography is stimulated by the success of AdS/CFT and has attracted much attention due to its potential applications in realistic physical processes. In this seminar, I will introduce Carrollian holography in which the dual field theory is assumed to live on the null boundary of asymptotically flat spacetime. At first, I will review basic elements on Carrollian manifold and introduce quantum flux operators which realize various geometric symmetry of Carrollian manifold. Second, I will explore the definition of Carrollian amplitude which is the central object that integrates geometry, quantum effects and holography. At last, I will discuss its potential applications to gravitational physics.
報告人簡介：
Dr. Jiang Long is currently an associate professor of physics at Huazhong University of Sicence and Technology (HUST). He received his PhD in Physics from Peking University in 2015. After that, he became a postdoctoral fellow at Universite libre de Bruxelles (ULB). He moved to the Asia-Pacific Center of Theoretical physics (APCTP) before he joined HUST. His current research focus on flat holography and gravitational physics.
3
報告人：李佳樂，倫敦大學學院
時間：7月17日（週三）14:00
單位：中國科學院理論物理研究所
地點：南樓6620
摘要：
The theory of general relativity has been the cornerstone of modern physics for over a century, and yet, many of its predictions remain untested. In recent years, the field of multi-messenger astronomy has emerged as a powerful tool for testing Einstein's theory, enabling us to study the universe through a combination of different cosmic messengers such as gravitational waves, electromagnetic waves, and cosmic rays. In this talk, we will explore how the upcoming Laser Interferometer Space Antenna (LISA) mission will revolutionise our understanding of gravity and astrophysics. LISA is a space-based gravitational wave detector that will be able to observe low-frequency gravitational waves, which are inaccessible to ground-based detectors. With its unprecedented sensitivity, LISA will enable us to probe a wide range of astrophysical phenomena, including the mergers of supermassive black holes, extreme-mass-ratio inspirals, and galactic binaries. I will also discuss how combining LISA observations with optical observations and radio pulsar timing, will allow us to test the relativistic dynamics in binary and triple systems, to constrain the maximum mass of neutron stars and to test gravito-magnetic effect.
報告人簡介：
Kaye Jiale Li is a senior PhD student from Mullard Space Science Laboratory, University College London, working with Dr Kinwah Wu and Dr Ziri Younsi. Her research interests centre on multi-messenger astrophysics, with a focus on testing general relativity through the combination of gravitational waves and pulsar timing.
4
報告人：趙琦，香港大學
時間：7月17日（週三）14:30
單位：中國科學院物理研究所
地點：B樓II段412會議室
摘要：
量子模擬，作爲量子計算機最具潛力的應用之一，可以幫助模擬複雜的量子系統演化，從而幫助解決凝聚態物理、高能物理、量子化學、生物製藥等領域中的諸多問題，加速相關的科學技術發展。然而量子模擬距離實際應用仍然有着很長的道路。哈密頓量子模擬算法是一種系統可靠的量子模擬算法。該類算法的誤差分析是該領域的重要問題，它決定了量子計算任務中的量子資源估計。主流的誤差分析往往只衡量了最壞情況下的初態和觀測量，這一分析大大高估了量子算法中的誤差，阻礙了量子模擬的實際應用。爲了解決這一問題,我們考慮了量子模擬任務中初態和測量對於誤差分析的影響，提出了隨機初態[PRL129(27),270502，QIP2022talk]、糾纏態[arXiv:2406.02379]和基於局域觀測量的量子模擬算法分析。這些結果極大減少了量子模擬任務中的資源需求，推動量子模擬算法的未來實現和應用。
報告人簡介：
趙琦，香港大學計算機系助理教授。他主要從事量子模擬、量子糾纏、量子計算、量子信息基礎理論等方面的研究，共發表期刊論文43篇，相關的一系列工作發表在國際頂級期刊 Nature, PRL, PRX, npj Quantum Inf., PNAS, IEEE Transactions on Information Theory; 成果也多次入選量子信息理論重要會議 QIP, AQIS, TQC, QCrypt 報告環節，並擔任AQIS程序委員會委員（PC member）。主持或參與了國自然-香港研資局聯合計劃、香港傑出青年學者計劃、香港影響力基金、國自然理論物理專款研究項目等。
5
報告人：周通，寧波東方理工大學
時間：7月18日（週四）10:00
單位：中國科學院物理研究所
地點：物理所A332
摘要：
Majorana zero modes (MZM) with their nonlocal degrees of freedom and non-Abelian statistics have been viewed as a promising candidate for fault-tolerant quantum computing. However, despite intensive efforts, experimental evidence for MZM remains indirect and does not probe their crucial non-Abelian statistics. In this talk, I will reveal how we demonstrate MZM non-Abelian statistics through fusion and braiding in planar Josephson Junctions (JJs) with gate, phase, or magnetic texture control. Our proposals to identify non-Abelian statistics are supported by dynamical simulations and experimental findings in the fabricated InAs/Al JJs, opening an important route towards topological quantum computing.
報告人簡介：
周通，寧波東方理工大學研究員、博導，上海交通大學、中國科學技術大學兼職博導，入選國家級高層次人才、寧波市特優人才等項目。2017年博士畢業於復旦大學，之後在美國紐約州立大學布法羅分校歷任博士後研究員、研究科學家、研究助理教授，2023年加入東方理工。長期從事拓撲物態、量子輸運與量子計算等研究，已發表論文40餘篇，其中一作/通訊作者論文10餘篇，包括近三年的Nat. Mater., Nat. Commun., 3篇Phys. Rev. Lett.等, 谷歌學術引用2800餘次，H-index 24；多次受邀在APS March Meeting, SPIE, WINDS等國際知名學術會議上做邀請報告。
6
報告人：Bhupal Dev，Washington University
時間：7月18日（週四）15:30
單位：中國科學院理論物理研究所
地點：南樓6620
摘要：
Neutrinos and dark matter remain two of the least known sectors of fundamental physics. What if they interact with each other? We will address this question in this talk. We implement a full catalog of constraints on the parameter space of the neutrino-dark matter/mediator couplings and masses, including bounds coming from cosmology, astrophysics, as well as new laboratory constraints from Z-boson and meson decays. As an application, we then consider galactic supernova neutrinos, identify new benchmark points that pass all the aforementioned constraints, compute the full set of cascade equations and sky maps for different dark matter density profiles in the Galaxy, and study their implications for the detection prospects in future large-volume neutrino experiments such as DUNE, Hyper-K and JUNO.
報告人簡介：
Bhupal Dev is an Associate Professor of Physics and a member of the McDonnell Center for the Space Sciences at Washington University in St. Louis, USA. He got his PhD from the University of Maryland in 2012, working with Prof. Rabi Mohapatra. He then held postdoctoral positions at the University of Manchester in the UK, Technical University of Munich and Max-Planck-Institute for Nuclear Physics, Heidelberg, Germany, before joining WashU in 2016. He is a recipient of the Fermilab Intensity Frontier Fellowship and the Universities Research Association Visiting Scholars Award.
7
報告人：Shangfeng Du，School of Chemical Engineering, University of Birmingham, UK
時間：7月19日（週五）10:00
單位：中國科學院物理研究所
地點：M255會議室
摘要：
The electrode structure plays a critical role in controlling the charge and mass transfer performance, thus determining whether the intrinsic catalytic activity of the electrocatalyst could be fully transported to practical electrodes or not during the operation of polymer electrolyte fuel cells. Electrochemical engineering research is required to further understand the influence mechanisms for reducing different overpotentials to bridge the gap between the highly active electrocatalysts and poor power performance devices. This talk will introduce our design of ordered electrodes based on Pt nanowire arrays for low temperature fuel cells to achieve large current density operation, discussing the lessons learnt during the thin film electrode fabrication, from nanostructure growing for tuning the electrode structure, the surface modification using functional Pd nanoseeds and plasma treatment, and alloying to boost the electrode activity and durability.
報告人簡介：
Shangfeng Du is Professor of Electrochemical Engineering, leading research for low temperature fuel cells and electrolysers in Birmingham Energy Institute at the University of Birmingham, UK. He graduated from Tsinghua University in Materials Science and Engineering, and obtained his PhD in Chemical Engineering from the Chinese Academy of Sciences. After working as a research associate at the Max Planck Institute for Metals Research, Germany, he joined the University of Birmingham as a Marie Curie Research Fellow. His research interests lie in electrochemical engineering research for energy application, with a focus on the development and evaluation of electrodes, composite membranes, MEAs and short stacks, as well as their application and degradation research. He has been leading a few EU, UK and industrial projects, (co-)authored several books and more than 70 research papers. He serves as President of the Chinese Society of Chemical Science and Technology in the UK (CSCST-UK).
8
報告人：孫小軍，廣西師範大學
時間：7月22日（下週一）15:00
單位：中國科學院理論物理研究所
地點：南樓6520
摘要：
簡要介紹氫硼聚變反應的背景及其潛在的應用前景；針對氫硼聚變反應所應用到的核反應理論基礎——輕核反應的統計理論( statistical theory of light nucleus reaction: STLN)及其相關核數據的初步研究進展；提出了今後關於氫硼聚變工程中還需要的核數據實驗測量和理論研究等方面的工作目標；最後簡介本團隊利用機器學習方法應用於中子共振核數據領域的初步進展。
報告人簡介：
孫小軍，廣西師範大學物理學院教授、西安交通大學兼職教授、博士生導師，《廣西物理》主編。先後在中國核數據中心攻讀博士學位、廈門大學從事博士後科研工作。先後主持4項國家自然科學基金項目、1項廣西重點研發項目(軍民兩用)、2項廣西自然科學基金重點項目以及多項橫向項目。主要從事輕核反應研究、機器學習在覈數據中的應用等領域的研究。
圖片來源於：https://new.qq.com/rain/a/20240403A095YK00
更多報告信息：
《物理》50年精選文章