Colloquium Calendar

Title: The Legacies of Hubble, Webb,  and future NASA astrophysics flagship missions

Abstract: 

For the past four decades, the Space Telescope Science Institute has served as the bridge between NASA's flagship astrophysics missions and the scientific community through its role as the science operations center for Hubble, Webb, and Roman. The Hubble Space Telescope is celebrating its 35th year of operations,  and continues to lead innovative new discoveries with its ultraviolet-optical capabilities and decades of high precision data. In its third year of operations, the infrared James Webb Space Telescope is the most powerful telescope ever built and is rapidly transforming a broad range of astronomy fields, including galaxy evolution, cosmology,  star-formation, and exoplanet science. In two years,  the Nancy Grace Roman Space Telescope will launch and begin to survey the sky at speeds 1000 times that of Hubble.  Roman is designed to study cosmological questions about the nature of dark energy, the evolution of the expansion rate, and the growth of large-scale structure over cosmic time,  as well as detect tens of thousands of exoplanets and demonstrate cutting-edge technology to directly image exoplanets. Finally,  STScI is working to support the next generation NASA astrophysics flagship mission, the Habitable Worlds Observatory.  This mission will be a "super-Hubble" capable of detecting earth-like planets around sun-like stars, tracing the cosmic intergalactic medium, and spatially-resolving every known galaxy in the universe. These NASA flagship missions have immeasurable impact on the field of astrophysics and how astronomers around the world do science.  They enable STEM workforce development across the U.S. and are clear demonstrations of U.S. and NASA's leadership in science and technology.   I will discuss the challenges that cuts to NASA's science budget pose for the future of these missions.

Title: Assumptions of Physics: a new principled approach to the foundations of physics 

Abstract: 

We will give an overview and the main results of the Assumptions of Physics project, which aims to find minimal sets of physically meaningful starting points from which the different physical theories can be rederived. It consists of two main strategies. The first, Reverse Physics, starts from the known physical laws and studies principles and assumptions that are equivalent to the mathematical formulation. This allows one to create a clear dictionary between physical concepts and their mathematical representation, showing us, for example, what physical assumptions are equivalent to the principle of least action, or that geometric structures in physics are equivalent to entropic structures. The second approach, Physical Mathematics, aims to rederive all mathematical structures used in physics "from scratch," giving physics a more conceptually solid mathematical foundation. Ultimately, this will lead to a theory of physical theories, which can be used to better understand the current ones and to aid the search for new ones.

Title: Uncovering the Fundamental Interactions Guiding Thin Film Growth and Electronic Metal-Support Interactions

Abstract: There is a central relationship between material structure, composition, and function. For complex surface processes, such as thin film growth and heterogeneously catalyzed chemical transformations, representative systems interrogated with surface science techniques can reveal the atomic-scale mechanisms underpinning observed material activity. Such studies have broadened the understanding of, and are integral to, developing accurate models for many interfacial phenomena. In this presentation, I will discuss recent results investigating the nanoscale chemical, structural, and electronic processes guiding 1) binary and single-component self-assembled monolayer (SAM) formation and 2) the emergence of electronic metal-support interactions in oxide-supported metal nanoparticle catalysts.

Title: From Light to Learning: Predicting Excited-State Dynamics at the Spectroscopic Frontier

Abstract: 

Processes ranging from photosynthesis, to photocatalysis, to energy harvesting in photovoltaic cells all begin in the same way: the absorption of light creates an exciton—a correlated electron-hole pair that carries energy rather than charge. Exciton dynamics and coherences determine the efficiency of energy harvesting and transport, while excitonic manipulation enables the optical preparation and transduction of quantum states and offers the potential to integrate the fast speed of photons into electronics. However, quantitative predictions of exciton and other excited-state dynamics remain a significant challenge. The first-principles understanding of exciton dynamics requires a few basic building blocks: 1) The full exciton dispersion to capture the phase space of momentum and energy conserving scattering processes, 2) Interaction of excitons with external perturbations such as electromagnetic fields and lattice vibrations, and 3) An equation of motion describing the dynamical processes. In this talk, I will discuss some of my group’s developments in these directions. Firstly, we have recently measured the exciton dispersion in a 2D materials revealing for the first time the emergence of a massless excitons composed of massive electrons and holes [1]. Secondly, we will show how excitons play a surprising role in nonlinear optics beyond the perturbative regime. Using newly developed first principles techniques, we find that excitons enhance high-harmonic generation in monolayer semiconductors, in ways directly tied to the Berry curvature of the underlying bands. Finally, I will discuss how machine learning is opening new possibilities for simulating excited states far beyond the reach of conventional calculations [3,4].

References:

[1] L. Liu, S.Y. Woo, J. Wu, B. Hou, C. Su, D.Y. Qiu, “Direct Observation of Massless Excitons and Linear Exciton Dispersion,” arXiv:2502.20454v2 (2025).

[2] V. Chang Lee, L. Yue, M.B. Gaarde, Y.-H. Chan, D.Y. Qiu, “Many-body enhancement of high harmonic generation in monolayer MoS2,” Nature Comm. 15, 6228 (2024).

[3] B. Hou, J. Wu, D.Y. Qiu, "Unsupervised representation learning of Kohn–Sham states and consequences for downstream predictions of many-body effects," Nature Comm. 15, 9481 (2024).

[4] B. Hou, X. Xu, J. Wu, D.Y. Qiu, "MBFormer: A General Transformer-based Learning Paradigm for Many-body Interactions in Real Materials," arXiv:2507.05480 (2025).

Title: Resolving the Hubble Constant Discrepancy: Mitigating Biases in Cosmic Distance Ladders 

Abstract: 

The Hubble constant (H0), representing the present-day expansion rate of the universe, exhibits an approximately 8% discrepancy between direct measurements in the late universe and indirect measurements from the early universe. Accurate direct measurements are challenging because they require measuring cosmic distances far beyond our galaxy, necessitating the use of a distance ladder with multi-step calibrations of different distance indicators. While propagated measurement errors contribute only about 1% to the uncertainty in H0, biases in inference and priors may be at the same level as the observed discrepancy. I will discuss two important biases: Malmquist bias, stemming from observational selection and luminosity dispersion, and Eddington bias, resulting from the scatter and non-uniform distribution of the luminosity indicator. With simulated data, I will show that these biases can be mitigated by constructing the data likelihood function to account for their contributing factors. By applying these methods, future direct measurements of H0 can achieve unbiased precision, potentially resolving the current discrepancy.

Title: From simple to complex: learning pathways for solid-state quantum systems

Abstract:

Complex properties of solid-state quantum systems are entangled with a set of physical quantities and constraints, such as orbital interactions, Hamiltonian, and global/local atomic environment and symmetries. In recent years, artificial intelligence (AI) and machine learning (ML) have been widely applied to many aspects of physical sciences, while the use of ML for either complex solid-state systems or complex properties of solids remains challenging. In this talk, I will introduce recent developments of ML frameworks that enable the prediction of complex properties in solid-state quantum systems from simple graph representations and physical principles. The talk will focus on model developments for two types of complex properties that are associated with (i) electronic structures and (ii) configurations. 

Two routes for electronic structure learning will be discussed: end-to-end predictions and Hamiltonian learning. I will introduce a universal and end-to-end graph-Transformer-based model predicting the electronic band structure of any crystalline system based on atomic structure input only. The mean absolute error is ~0.3 eV for band energy predictions, which is believed to be essential for ML-accelerated discovery/design of functional quantum materials. In terms of Hamiltonian learning, I will discuss how existing ML force field models can be reused and combined with physical-principle-informed embeddings to build an efficient model for predicting electronic structures of large-scale twisted two-dimensional materials. Beyond basic electronic structures, I will introduce a model built on top of equivariant neural network to predict one of the most complicated electronic-structure-derived properties in solids: tensorial spectra. 

In the field of applying ML for material systems with tremendous numbers of configurations, I will introduce a general ML framework for modeling ensemble properties of atomically disordered materials such as multi-compositional alloys. I will discuss how equivariant neural network models can be combined with first-principles computational approaches for atomistic systems and statistical methods to enable fast and reliable predictions for both disorder-related properties (such as order-disorder phase transitions) and ensemble properties of these systems, which are not accessible by traditional computational approaches.

At the conclusion of the talk, I will highlight ongoing progress in modeling complex properties of solids, with particular attention to data challenges and the need for collective community efforts to address them.

Title: 

Abstract:

Title: 

Abstract: 

Title: 

Abstract: 

Title: 

Abstract: 

Title: 

Abstract: 

Title: 

Abstract: