WQ Lunch Seminar Series

Venue: PS1.28 (Physical Sciences Building)

Suroj Dey (Physics)

• Title: Introduction to Classical-Quantum Hybrid Dynamics
• Abstract: In this talk, Suroj will address the following questions:
  • Why do we need classical-quantum hybrid dynamics?
  • What are they?
  • How can they be formulated?
  • Can they be made relativistic?

Natalie McHugh (Computer Science)

• Title: Tensor networks, MPS and "Entanglement Rerouting"
• Abstract: In this talk, Natalie will discuss how tensor networks provide a powerful framework for describing quantum systems. She will begin by introducing tensors, tensor networks, and matrix product states (MPS). She will then present "Entanglement Rerouting", a technique for simplifying the structure of a quantum state at the cost of increasing the number of parameters. Finally, Natalie will explore examples illustrating how this method can be applied.

Venue: CS1.04 (Department of Computer Science)

Zsuzsanna Koczor-Benda (Chemistry)

• Title: Property-driven computational molecular design for nanotechnology
• Abstract: When molecules are placed in electromagnetic hot-spots—such as those created at nanostructured metal surfaces—their interaction with light is dramatically amplified. This amplification can boost normally weak spectroscopic signals by orders of magnitude, opening the door to powerful nanoscale technologies, from ultra-sensitive medical imaging to single-molecule devices. To make such devices work at their best, however, we need molecules with very specific quantum-mechanical properties. Identifying these is not straightforward: small structural changes can drastically alter how a molecule interacts with light.

  In this talk, I show how combining quantum mechanical modelling with machine learning and generative AI can accelerate the search for optimal molecules. As a case study, I present a terahertz radiation detection strategy [1] that relies on converting low-frequency radiation into visible signals via molecular vibrations. Machine learning-based high-throughput screening [2] of commercially available molecules allows us to pinpoint promising candidates and extract design rules.

  However, existing databases cover only a tiny fraction of possible chemistry. Generative AI allows us to explore vast, previously uncharted chemical spaces and uncover deeper structure–property relationships. Nanotechnology applications impose strict physical and chemical constraints—stability near metal surfaces, compatibility with device architectures, and specific optical responses. I discuss our recent advances in constrained generative models [3] and property-driven design strategies [4] that enable the targeted creation of functional molecules tailored for nanotechnological applications.

  [1] Xomalis, A.; Zheng, X.; Chikkaraddy, R.; Koczor-Benda, Z.; Miele, E.; Rosta, E.; Vandenbosch, G.A.; Martínez, A.; Baumberg, J.J. Science 2021, 374 (6572), 1268-1271.

  [2] Koczor-Benda, Z.; Boehmke, A.L.; Xomalis, A.; Arul, R.; Readman, C.; Baumberg, J.J.; Rosta, E. Phys. Rev. X 2021, 11(4), 041035.

  [3] Koczor-Benda, Z.; Gilkes, J.; Bartucca, F.; Al-Fekaiki, A.; Maurer R. J. J. Chem. Inf. Model. 2025, 65 (13), 6644-6654.

  [4] Koczor-Benda, Z.; Chaudhuri, S.; Gilkes, J.; Bartucca, F.; Li, L.; Maurer, R.J. Digit. Discovery, 2025, 4, 2852-2863.

Hirak Ghosh (Physics)

• Title: Property Testing for Quantum Objects
• Abstract: Property testing is a field of theoretical computer science that involves building superfast algorithms to reliably certify whether an object has a desired property, or is far from having it. What are these objects anyway, and what constitutes a property, and by what notion of distance an object is considered “far”, are questions that can be answered in different ways depending on the setting and the audience. From a quantum information perspective, these “objects” of interest are quantum or quantum-induced: a catch-all that includes quantum states, quantum channels (and, by extension, gates, circuits, and quantum processes), as well as the classical distributions you obtain from measuring them. I’ll show that, under appropriate distance metrics, testing identicality — a particular property of these objects — can outperform traditional learning approaches involving complete characterization, often yielding exponential speedups. With some physically motivated promises about the system, such testing can enable even faster certification. I will also explore the potential utility of property testing for decision problems relevant to the broader audience at WQ, emphasizing its importance as an abstract tool for reliable, resource-efficient certification.

Venue: CS1.04 (Department of Computer Science)

Alex Newman (Physics)

• Title: Diamond Magnetometry with Nitrogen-Vacancy Centres
• Abstract: Solid-state defects can function as highly sensitive quantum sensors. One example is the nitrogen-vacancy centre in diamond, which can be used as a magnetic field sensor. This talk will give a simple introduction to how nitrogen-vacancy centres are used for magnetometry and highlight selected applications explored during my PhD.

Yilun Gao (Physics)

• Title: Putting machine learning to the test in a quantum many-body system
• Abstract: Quantum many-body systems pose a formidable computational challenge due to the exponential growth of their Hilbert space. While machine learning (ML) has shown promise as an alternative paradigm, most applications remain at the proof-of-concept stage, focusing narrowly on energy estimation at the lower end of the spectrum. In this work, we test an improved neural-network approach, HubbardNet, for the Bose–Hubbard model. By refining the training method and incorporating physics-informed features, we greatly improve accuracy, achieving very small energy errors and wave-function fidelities above 99%. We also show that the model correctly captures key physical behaviors such as localization, delocalization, and multifractality for both ground and excited states in one and two dimensions, and across a wide range of interaction strengths. These results indicate that machine learning can reliably predict qualitative properties of complex quantum systems, complementing established numerical methods.