Next-Generation Data-Driven Probabilistic Modelling of Type Ia Supernova SEDs in the Optical to Near-Infrared for Robust Cosmological Inference

Type Ia supernovae (SNe Ia) are used as “standardiseable candles”: their peak luminosities can be inferred from their optical light curve shapes and colours, so their distances can be estimated from their apparent brightnesses.

SN Ia distances with high precision and small systematic error are essential to accurate constraints on the cosmic expansion history, local measurements of the Hubble constant, and the properties of the dark energy driving the acceleration, in particular, its equation-of-state parameter w. The current global sample used for cosmology has grown to over a thousand SNe Ia.

Future surveys will boost that number by orders of magnitude.

However, the constraints on dark energy with the current optical sample are already limited, not by statistical uncertainties from the numbers of SNe, but by systematic errors.

Near-infrared (NIR) observations of SN Ia are a route to more precise and accurate distances and significantly enhance their cosmological utility. SNe Ia are excellent standard candles in the NIR, and are less vulnerable to absorption by dust in the host galaxies.

These good NIR properties are not exploited by the conventional optical models currently used for cosmological SN Ia analysis.

Furthermore, the present useful sample of SN Ia with NIR data is relatively small compared to the growing nearby or distant optical samples.

In this Project, we will leverage our involvement in new SN surveys using the Hubble Space Telescope and ground-based observatories to build a ~10X larger sample of SNe Ia with high-quality optical and NIR data.

We will develop the next-generation probabilistic model for SN Ia spectral energy distributions (SEDs) in the optical-to-NIR, accounting properly for the variabilities and uncertainties inherent in the data by fusing advanced hierarchical Bayesian modelling and functional data analysis techniques. We will apply our state-of-the-art model to our new SN datasets and LSST to obtain robust cosmological inferences.