Seminar

Algebraic structures in QFT in the presence of a quantum reference frame

October 9 (Thu) 14:00 - 15:00, 2025

Kasia Rejzner (Professor, Department of Mathematics, University of York, UK)

In this talk I will show how operational description of measurement with the use of quantum reference frames (QRF) affects the algebraic structure of quantum field theory (QFT). I will focus on the example of a quantum clock coupled to a QFT on de Sitter spacetime, previously discussed by Chandrasekaran, Longo, Pennington and Witten. This talk is based on my recent work with Chris Fewster, Daan Janssen, Leon Loveridge and James Waldron.

Venue: Seminar Room #359 (Main Venue) / via Zoom

Event Official Language: English

Homo lupo lupus est: Man is a wolf to wolves.

October 9 (Thu) 14:00 - 15:00, 2025

Carlos Sarabia (Postdoctoral Researcher, Evolutionary Population Genetics Lab, Institute of Evolutionary Biology (IBE-CSIC), Spain)

The gray wolf (Canis lupus) is one of the most emblematic wild species in human history: revered as a symbol of strength and wildness, although unforgivably persecuted as a competitor and pest. Across Europe and much of Eurasia, wolves would still dominate as apex predators... were it not for millennia of human pressure. Today, their evolutionary trajectory is shaped not only by climate fluctuations and habitat loss, but also by a uniquely flexible species boundary. Due to their unique karyotype, canids can admix freely with other related species, a capacity that both threatens the genetic integrity of wild canids like wolves and enriches our understanding of hybridization as a driver of adaptation. In this talk, we will explore recent studies on wolf demography under human pressure and climatic change, with particular attention to admixture with domestic dogs and the consequences for their survival in increasingly anthropized environments. Finally, we will observe how the wolf's distinctive genomic architecture makes it a powerful model for testing population genetics theoretical frameworks and for applying state-of-the-art computational tools, offering new insights into the understanding of evolution as a force for change.

Venue: via Zoom

Event Official Language: English

Discovering and harnessing symmetry with machine learning

October 6 (Mon) 16:00 - 17:30, 2025

Escriche Santos Eduardo (Ph.D. Student, Department of Computer Science, Technical University of Munich, Germany)

Incorporating symmetry-inspired inductive biases into machine learning models has led to many significant advances in the field, especially for its application to scientific data. However, recently, a trend has emerged that favors implicitly learning relevant symmetries from data instead of designing constrained equivariant architectures. In this talk, I will first introduce these different modelling alternatives, together with their associated benefits and limitations. Then, I will describe some examples of automatic symmetry discovery methods as a way of mitigating some of those limitations. Finally, I will present our recent work that integrates symmetry discovery and the definition of an equivariant model into a joint learnable end-to-end approach, which further alleviates some of the limitations of current equivariant modelling approaches.

Venue: via Zoom

Event Official Language: English

Quantum tunneling in the curved spacetime

October 2 (Thu) 13:30 - 15:00, 2025

Masahide Yamaguchi (Director, Center for Theoretical Physics of the Universe, Institute for Basic Science, Republic of Korea)

False vacuum decay is theorized to have occurred frequently throughout the history of the universe, particularly during first-order phase transitions associated with spontaneous symmetry breaking. The decay rate of such a vacuum is governed by Euclidean bounce solutions, which can exhibit a wide range of configurations, even under fixed boundary conditions. In the absence of gravitational effects, it was established over four decades ago—under reasonable assumptions—that the most symmetric bounce solution, namely the O(4)-symmetric one, minimizes the Euclidean action. This renders it the dominant tunneling path in flat spacetime. However, when gravitational effects are taken into account—as is essential in cosmological settings—all prior studies have assumed, without rigorous proof, that the O(4)-symmetric bounce continues to minimize the action. This has remained a longstanding unresolved problem for more than forty years. In this work, we address this issue by employing the anti-de Sitter/conformal field theory (AdS/CFT) correspondence to determine the configuration with the lowest Euclidean action in a metastable AdS false vacuum. Within the Euclidean formalism of Callan and Coleman, we identify the most probable decay channel of the AdS vacuum. The AdS/CFT duality enables us to sidestep the technical challenges intrinsic to metastable gravitational systems. We demonstrate that the Fubini bounce in conformal field theory—which is dual to the Coleman–de Luccia (CdL) bounce in AdS—indeed minimizes the Euclidean action among all finite bounce solutions in a conformal scalar field theory. Consequently, under certain conditions, we establish that the CdL bounce yields the lowest action among all relevant configurations, including both large and thin-wall limits. Time permitting, we also discuss the prefactor of the decay rate, as obtained from one-loop quantum corrections.

Venue: Hybrid Format (3F #359 and Zoom), Seminar Room #359

Event Official Language: English

What constitutes a gravitational wave in an expanding universe?

October 1 (Wed) 16:00 - 17:30, 2025

Yi-Zen Chu (Professor, Department of Physics, National Central University, Taiwan)

Our understanding of gravitational waves produced by isolated astrophysical systems is primarily based on gravitational perturbation theory off a flat spacetime background. This leads to the common identification of gravitational radiation with massless spin-2 waves. In this talk, I will argue that gravitational waves may no longer be solely "spin-2" in character once the background spacetime is our expanding universe instead. As a result of the mixing between gravitational and other degrees of freedom, scalar "spin-0" gravitational waves may exist during the radiation-dominated epoch of our universe; as well as during its current accelerated expansion phase -- provided the main driver is not the cosmological constant, but some extra "Dark Energy" field. Moreover, during the radiation-dominated era, spin-0 Cherenkov gravitational waves may even be generated if its material source were traveling faster than 1/\sqrt{3}.

Venue: Hybrid Format (3F #359 and Zoom), Seminar Room #359

Event Official Language: English

From Data to Discovery: Chronobiology in Translation

October 1 (Wed) 13:00 - 14:00, 2025

Bharath Ananthasubramaniam (Professor, Institute for Theoretical Biology, Humboldt University of Berlin, Germany)

Disruption of circadian rhythms is increasingly linked to a range of pathologies. To harness circadian biology for disease prevention and treatment, we must first establish causal relationships between rhythm disruption and the underlying clock mechanisms. This requires both the ability to quantify the “clock state” and to define what constitutes “disruption.” While significant progress has been made in model organisms, translating these insights to humans presents distinct challenges for quantitative chronobiology. In this talk, I will highlight how we have leveraged novel computational methods and high-throughput molecular datasets to begin addressing these obstacles.

Venue: Seminar Room #359 (Main Venue) / via Zoom

Event Official Language: English

A Continuous Galactic Line Source of Axions: The Remarkable Case of 23Na

September 30 (Tue) 14:00 - 15:00, 2025

Wick C. Haxton (Professor, Department of Physics, University of California, Berkeley, USA)

While it is unusual for odd-A nuclear species to be abundant in massive stars, 23Na is an interesting exception. Typically 0.1 solar masses of 23Na is synthesized during the carbon burning phase of supernova and ONeMg white dwarf progenitors, then maintained at approximately 10^9 K for periods ranging up to 60,000 years. Under these conditions, 23Na can pump the thermal energy of the star into escaping axions: the mechanism is the Boltzmann occupation of and subsequent axion emission from the 440 keV level. We develop a galactic model to show that the resulting flux of line axions is continuous, arising from hundreds of contributing sources. As they travel through the intra-galactic magnetic field, some of these axions convert to detectable gamma rays. Consequently, future all-sky detectors like COSI will be able to set new limits on light axion-like particles. Other interesting aspects of these axions will be discussed.

Venue: Hybrid Format (3F #359 and Zoom), Seminar Room #359

Event Official Language: English

Spontaneous quasiparticle creation in an analogue preheating experiment

September 30 (Tue) 10:00 - 12:00, 2025

Amaury Micheli (Postdoctoral Researcher, Division of Fundamental Mathematical Science, RIKEN Center for Interdisciplinary Theoretical and Mathematical Sciences (iTHEMS))

Abstract: First, I will briefly outline the motivations and concepts that underpin the analogue gravity program. Next, I will provide a detailed description of a specific experiment designed to simulate various features of the cosmological reheating era. Finally, I will present our recent experimental results from this setup, where we demonstrated the parametric creation of quasiparticle pairs from the quantum vacuum, drawing an analogy with the preheating phase of reheating.

Venue: Hybrid Format (3F #359 and Zoom), Seminar Room #359

Event Official Language: English

The QCD phase diagram at finite densities

September 29 (Mon) 13:30 - 15:00, 2025

Franz Sattler (Postdoc, Faculty of Physics, Bielefeld University, Germany)

I discuss recent progress towards calculating the QCD phase diagram at finite density using the functional Renormalisation Group (fRG). After introducing the fRG as applied to QCD, I explain some of the challenges encountered in functional approaches to the QCD phase diagram. Many of these can be resolved by recent developments of new numerical methods. In particular, the application of numerical hydrodynamics to RG flows and resolution of momentum dependences allow us to make progress towards quantitative access to the region of the conjectured critical end-point (CEP) of the QCD phase diagram. An interesting result is the appearance of new phases characterised by spatial modulations (the moat regime) and inhomogeneous condensates at high densities from a self-consistent first-principles calculation. For the near future, a clear program emerges to further pinpoint the CEP and its possibly modified nature using the fRG.

Venue: Hybrid Format (3F #359 and Zoom), Seminar Room #359

Event Official Language: English

Confined Circumstellar Material as a Dust Formation Site in Type II Supernovae

September 26 (Fri) 14:00 - 15:15, 2025

Yuki Takei (Program-Specific Researcher, Yukawa Institute for Theoretical Physics, Kyoto University)

Some massive stars undergo episodic mass loss shortly before core-collapse, producing dense circumstellar material (CSM) in their immediate surroundings. If the supernova (SN) ejecta strongly interacts with such CSM, narrow emission lines appear in the spectrum, classifying the event as Type IIn. In these cases, efficient radiative cooling forms a cold, dense shell (CDS), providing ideal conditions for dust condensation. Infrared observations of several SNe IIn have indeed confirmed newly formed dust. Recent time-domain surveys, however, suggest that compact and dense CSM, often termed “confined CSM”, is also present around a broader class of Type II SN progenitors with hydrogen-rich envelopes, beyond the traditional Type IIn subclass. This raises the possibility that dust formation in dense CSM is more common among core-collapse SNe than previously thought. In this talk, I will demonstrate that CDS formation occurs robustly across a wide parameter space for confined CSM using numerical simulations based on the open-source code CHIPS. I will also discuss the resulting dust mass and infrared emission, as well as the potential contribution of this process to the galactic dust budget.

Venue: Hybrid Format (3F #359 and Zoom), Seminar Room #359

Event Official Language: English

The evolution of conditional dispersal promotes cooperation

September 25 (Thu) 13:00 - 14:00, 2025

Iris Prigent (Ph.D. Student, Department of Ecology and Evolution, University of Lausanne, Switzerland)

Kin selection is an important mechanism for the evolution of cooperative behaviours across multiple taxa. While limited dispersal can foster kin selection by generating a genetic correlation between cooperating individuals, it also increases competition among relatives, constraining the evolution of cooperation. Prior theory has explored the co-evolution of dispersal and cooperation but typically assumes dispersal is independent of social cues. Here, we use mathematical modelling to examine whether socially-mediated dispersal, whereby individuals adjust their dispersal based on social context, can mitigate kin competition and thus enhance cooperation. We model the joint evolution of: (i) the probability of cooperating within social groups; and (ii) the probability of dispersing conditional on the number of individuals that have cooperated within the group, leading to a reaction norm for dispersal. We show that when the probability of dispersal increases with the number of cooperators, cooperation is favoured because it increases the fitness of relatives. The joint evolution of the two traits can lead to the differentiation of two types of individuals, one that always cooperates and another that never does. Although both types evolve dispersal norms such that they disperse more often when there are more cooperators in the group, cooperators evolve a steeper norm, reflecting greater sensitivity to their social environment. Our study shows that dispersal responses to the environment can vary between individuals based on their own social tendency, which can help explain why dispersal proclivities may differ between genotypes and between environments within a single population.

Venue: Hybrid Format (3F #359 and Zoom), Seminar Room #359 (Main Venue) / via Zoom

Event Official Language: English

Data Assimilation for the Vicsek model

September 25 (Thu) 13:00 - 14:00, 2025

Tomoharu Takaki (Master's Student, Graduate School of Information Science and Technology, The University of Tokyo)

Venue: R311, Computational Science Research Building (Main Venue) / via Zoom

Event Official Language: English

Lecture 2 on "Modular Structures in N=4 supersymmetric Yang-Mills theory"

September 19 (Fri) 14:30 - 17:30, 2025

Daniele Dorigoni (Associate Professor, University of Durham, UK)

In this lecture series we present recent results in the study of exact correlation functions of half-BPS operators in N=4 supersymmetric Yang-Mills theory (SYM) averaged over the space-time insertion points. After presenting some basic properties of 1/2-BPS operators in N=4 SYM, we review how these integrated correlation functions can be obtained from a matrix model formulation of the N=4 path-integral. We then move to present two different integrated correlation functions of four superconformal primary operators of the stress-tensor multiplet which are holographically related to scattering amplitudes of 4-gravitons in type IIB superstring theory on an AdS_5 x S^5 background. We derive exact expressions both in the number of colours N, as well as in the complexified Yang-Mills coupling \tau. A key player in our discussion is electro-magnetic duality of N=4 SYM which provides strong constraints on the coupling dependence of such observables which, in particular, have to be real-analytic modular invariant functions of \tau. We then discuss the large-N fixed-\tau limit to show how these results can be interpreted on the dual stringy side. We also present some details on how these integrated correlator can be used to supplement the standard bootstrap approach leading to exciting coupling dependent bounds on the anomalous dimensions of non-protected operators in N=4 SYM, such as the Konishi operator. Lastly, we discuss an integrated correlation function involving two superconformal primary operators in the stress tensor multiplet in the presence of a half-BPS line defect operator, such as a Wilson line. Electro-magnetic duality is again fundamental in understanding the exact dependence from the coupling constant \tau. [OPTIONAL: If time and energy permit, I can also present some new results regarding integrated correlation functions of two light operators, dual to gravitons on the holographic side, and heavy giant graviton operators, dual to D3 branes extended on the background geometry]

Venue: Hybrid Format (3F #359 and Zoom), Seminar Room #359

Event Official Language: English

Steps between the Lorenz96 models and the real world (TBD)

September 19 (Fri) 13:00 - 14:00, 2025

Arata Amemiya (Research Scientist, Prediction Science Research Team, Division of Applied Mathematical Science, RIKEN Center for Interdisciplinary Theoretical and Mathematical Sciences (iTHEMS))

Venue: R511, Computational Science Research Building (Main Venue) / via Zoom

Event Official Language: English

Lecture 1 on "Modular Structures in N=4 supersymmetric Yang-Mills theory"

September 18 (Thu) 14:30 - 17:30, 2025

Daniele Dorigoni (Associate Professor, University of Durham, UK)

In this lecture series we present recent results in the study of exact correlation functions of half-BPS operators in N=4 supersymmetric Yang-Mills theory (SYM) averaged over the space-time insertion points. After presenting some basic properties of 1/2-BPS operators in N=4 SYM, we review how these integrated correlation functions can be obtained from a matrix model formulation of the N=4 path-integral. We then move to present two different integrated correlation functions of four superconformal primary operators of the stress-tensor multiplet which are holographically related to scattering amplitudes of 4-gravitons in type IIB superstring theory on an AdS_5 x S^5 background. We derive exact expressions both in the number of colours N, as well as in the complexified Yang-Mills coupling \tau. A key player in our discussion is electro-magnetic duality of N=4 SYM which provides strong constraints on the coupling dependence of such observables which, in particular, have to be real-analytic modular invariant functions of \tau. We then discuss the large-N fixed-\tau limit to show how these results can be interpreted on the dual stringy side. We also present some details on how these integrated correlator can be used to supplement the standard bootstrap approach leading to exciting coupling dependent bounds on the anomalous dimensions of non-protected operators in N=4 SYM, such as the Konishi operator. Lastly, we discuss an integrated correlation function involving two superconformal primary operators in the stress tensor multiplet in the presence of a half-BPS line defect operator, such as a Wilson line. Electro-magnetic duality is again fundamental in understanding the exact dependence from the coupling constant \tau. [OPTIONAL: If time and energy permit, I can also present some new results regarding integrated correlation functions of two light operators, dual to gravitons on the holographic side, and heavy giant graviton operators, dual to D3 branes extended on the background geometry]

Venue: Hybrid Format (3F #359 and Zoom), Seminar Room #359

Event Official Language: English

Strategies for tuning unsupervised learning hyperparameters in the context of dimensionality reduction for multimodal omics data

September 18 (Thu) 14:00 - 15:00, 2025

Dorothy Ellis (Postdoctoral Researcher, Laboratory for Integrative Genomics, RIKEN Center for Integrative Medical Sciences (IMS))

We are actively developing multi-omics winnowing in R (mowR), a non-negative matrix factorization (NMF)-based model that expands upon the functionality of joint graph-regularized single cell sparse non-negative matrix factorization (jrSiCKLSNMF) from Ellis et al. (2023). “Omics” data characterize the molecular components of a biological sample. Examples of omics modalities include transcriptomics (RNA), epigenomics (epigenetic modifications), metabolomics (metabolites), proteomics (proteins), and genomics (DNA). Multi-omics analysis involves the integration of two or more of these modalities, and omics data are often high-dimensional and sparse. Therefore, dimension reduction techniques are often required to extract interpretable information from these datasets. NMF, one such dimension reduction technique, finds a low-dimensional approximation of M omics features by N observations data matrix X via the product of an M × D loadings matrix W and D × N activations matrix H, where the number of latent factors D << min(M, N ). The jrSiCKLSNMF model extends the basic NMF model by fitting a shared H across v ∈ {1, ..., V } omics count modalities. It also incorporates ridge regularization on H, graph regularization on feature matrix Wv in modality v, and sum-to-one L2 norm constraints on the rows of H. We extend jrSiCKLSNMF to mowR by implementing mini-batch updates (Serizel et al., 2016), modality-specific loss functions (e.g. Poisson K-L divergence for count modalities and Frobenius norm for Gaussian modalities), modality-specific activation matrices Hv and weights ωv on H to allow constraints on Wv , loss weights, LASSO regularization on H, and L2 norm constraints on Wv . We also introduce a novel technique to tune hyperparameters for unsupervised data by combining the data thinning/count splitting techniques outlined in Neufeld et al. (2023, 2024) with Bayesian optimization as implemented in the R package ParBayesianOptimization from Wilson (2018). In this talk, we focus on mowR’s hyperparameter tuning strategy, highlighting its current limitations and strategies to overcome them.

Venue: via Zoom

Event Official Language: English

Cross-species transcriptome analysis using Gromov-Wasserstein optimal transport

September 18 (Thu) 13:00 - 14:00, 2025

Yuya Tokuta (Program-Specific Researcher, Institute for the Advanced Study of Human Biology (ASHBi), Kyoto University Institute for Advanced Study (KUIAS))

Sequence homology underpins cross-species analysis but cannot identify evolutionarily distinct genes that play analogous regulatory roles. Furthermore, ethical restrictions on human experiments necessitate analytical frameworks that translate insights from other animals to humans. To address these challenges, we developed Species-OT, a cross-species transcriptome analysis framework based on Gromov-Wasserstein optimal transport, which quantitatively compares the geometry of transcriptome distributions. Given a pair of bulk or single-cell RNA-sequencing datasets, Species-OT returns a gene-to-gene correspondence capturing probabilistic alignments of regulatory roles, and a transcriptomic distance quantifying overall divergence. Applied pairwise, Species-OT yields a transcriptomic discrepancy array and a hierarchical clustering tree analogous to a phylogenetic tree. We validated Species-OT using bulk RNA-seq data from human, mouse, and macaque germ cell specification as well as scRNA-seq data from pluripotent stem cells of six mammalian species. Species-OT identified evolutionarily related and distinct gene correspondences including biologically unexplored candidates, while transcriptomic discrepancies recapitulated expected species relationships. This is joint work with T. Nakamura, K. Fujiwara, M. Imamura, M. Nagano, M. Saitou, Y. Imoto, and Y. Hiraoka.

Venue: Hybrid Format (3F #359 and Zoom), Seminar Room #359

Event Official Language: English

Statistical Physics of In-Context Learning in Transformer

September 16 (Tue) 15:00 - 16:30, 2025

Haiping Huang (Professor, School of Physics, Sun Yat-sen University, China)

The pre-trained large model demonstrates the ability to learn from examples, that is, it can infer patterns and generalize from a small number of examples without retraining. How does this ability emerge? This report proposes a physical model mapping of the large model pre-training process, and finds that the training process corresponds to spin condensation, the unique energy ground state will determine the example generalization ability, and the diversity of training data is a key element in algorithm design. This study also reveals that the reasoning process of the large model may be fundamentally different from human thinking.

Venue: via Zoom

Event Official Language: English

Multi-strangeness matter from ab initio calculations

September 16 (Tue) 13:30 - 15:00, 2025

Hui Tong (Post-doctoral Fellow, Helmholtz-Institut für Strahlen- und Kernphysik, University of Bonn, Germany)

Hypernuclei and hypernuclear matter connect nuclear structure in the strangeness sector with the astrophysics of neutron stars, where hyperons are expected to emerge at high densities and affect key astrophysical observables. We present the first ab initio calculations that simultaneously describe single- and double- hypernuclei from the light to medium-mass range, the equation of state for -stable hypernuclear matter, and neutron star properties. Despite the formidable complexity of quantum Monte Carlo (QMC) simulations with multiple baryonic degrees of freedom, by combining nuclear lattice effective field theory with a newly developed auxiliary-field QMC algorithm we achieve the first sign-problem free ab initio QMC simulations of hypernuclear systems containing arbitrary number of neutrons, protons, and hyperons, including all relevant two- and three-body interactions. This eliminates reliance on the symmetry-energy approximation, long used to interpolate between symmetric nuclear matter and pure neutron matter. Our unified calculations reproduce hyperon separation energies, yield a neutron star maximum mass consistent with observations, predict tidal deformabilities compatible with gravitational-wave measurements, and give a trace anomaly in line with Bayesian constraints. By bridging the physics of finite hypernuclei and infinite hypernuclear matter within a single ab initio framework, this work establishes a direct microscopic link between hypernuclear structure, dense matter composition, and the astrophysical properties of neutron stars.

Venue: Seminar Room #359 (Main Venue) / via Zoom

Event Official Language: English

Ensemble transform Kalman filter (ETKF) extensions for near-bound variables: Results from simple aerosol data assimilation experiments

September 16 (Tue) 9:00 - 10:30, 2025

Jiang Richard Liang (Postdoctoral Researcher, Keio University)

Traditional data assimilation (DA) methods approximate the error distributions using Gaussian probability density functions (PDFs). However, the error distributions of some variables, such as clouds, precipitation, and aerosols, could be better approximated by gamma and inverse-gamma PDFs. For such bounded variables, the error standard deviation will likely increase with the distance of the unknown true value from its bound. To properly include these error distributions, a previous study by C. Bishop invented a method called the GIG filter, which is based on gamma and inverse-gamma distributions. We compared the performance of this new method and the traditional DA method with cycled DA experiments using a new tracer model based on the Lorenz-96 model. The GIG filter's performance is better for assimilating near-bound variables in our experiments.

Venue: Hybrid Format (RIKEN R-CCS room 107 and Zoom)

Event Official Language: English