ALP Seminar: Uncovering Electronic Structure in Extreme Environments

It takes only a cursory glance at the world around us to encounter a wealth of fascinating material behaviours. Conductivity, magnetism, bond strength, phase, transport, and opacity — these properties shape our universe. And yet, the majority are so fundamental to our lives that we likely never stop to consider that they share a common origin: the arrangement and interactions of a material's electrons. By understanding these configurations — often described in terms of a material's density of states — we can begin to predict, and perhaps even control, these properties.

However, electronic structures can be as complex as they are difficult to probe. Even well-established diagnostic approaches, such as optical and X-ray spectroscopies, struggle to recover bulk-sensitive, element- and orbital-resolved information. As a result, we are often left with the unenviable task of piecing together a picture of a material's electronic structure from indirect measurements — for example, tracking nuanced phenomena such as insulator-to-metal transitions or superconductivity from resistivity alone. Combined with the challenges associated with modelling these systems, for which DFT and its extensions are often inadequate, we can find ourselves stranded. 

X-ray free-electron lasers (XFELs), with their ultrabright and ultrafast pulses, offer a promising route to probing electronic structure. However, their intrinsically stochastic and broadband spectrum is not compatible with conventional spectroscopies like XANES or resonant inelastic x-ray scattering (RIXS), which typically require narrow-band excitation, sacrificing photon flux and limiting signal-to-noise ratio. 

In this talk, I will present an alternative approach to RIXS that instead leverages the stochastic nature of the XFEL spectrum to exploit the full brightness of the beam. From the resulting spectra, we reconstruct density of states information using a simple physics-informed neural network optimisation. This technique enables the extraction of high-resolution, low-uncertainty data from low signal-to-noise measurements, finally offering a glimpse behind the curtain at the electronic structures that dictate our universe.