Anais Möller

Type Ia Supernovae

I work with a particular type of stellar explosions called type Ia supernovae (singular ‘supernova’). I am interested on them because they can be used for measuring distances. Supernovae Ia tend to all behave in a similar way, shining very brightly to a known luminosity and then fading away. They are known as ‘standard candles’ because we know how bright we expect them to be, and because they are very bright, they can be detected by telescopes on Earth.

My main research revolves around finding these type Ia supernovae, getting information from them (through imaging and spectroscopy) and using them for studying the expansion of the Universe (by measuring "distance" and "velocities" for a large sample of SNe Ia).

Current challenges

    1. get more precise measurements on the expansion of the Universe
    2. a better "close by" type Ia SN sample (low redshift)
    3. information about progenitors and environments of supernovae
    4. bigger samples of SNe Ia further back in time (or farther away).
    5. reducing selection biases of SN Ia samples for cosmology.

High precision measurements have been the goal of surveys such as the SuperNova Legacy Survey (SNLS) and the current Dark Energy Survey where our goal is to obtain the largest sample of supernovae Ia to date and do state-of-the-art cosmological analysis. OzDES plays a key role in this by observing host-galaxies and classifying supernovae.

Surveys

I am working in type Ia supernova cosmology in the upcoming LSST Dark Energy Science Collaboration (DESC) and current Dark Energy Survey (DES). I led the SkyMapper Transient Survey for 3 years. Using the SkyMapper telescope we searched for "close-by" supernovae and other electromagnetic counterparts to events such as Gravitational Waves and Fast radio Bursts . I also belong to OzDES, an spectroscopic survey at the Anglo-Australian Telescope that targets mostly galaxies with supernovae detected by the Dark Energy Survey (DES). This brings me to DES where we are working to do the state-of-the-art measurement of the expansion of the Universe with these SNe Ia.

Preparing for LSST: Fink

Large Survey of Space Time (LSST) will image during the next decade the sky deeper and faster than previous surveys. LSST will enable the discovery of unprecedentedly large numbers of astrophysical objects that exhibit temporal brightness variability, ranging from seconds to years. But only if we can identify interesting candidates! I am a PI of the team building Fink , a new broker designed to face the big data challenges linked to the LSST alert volume. We are building a robust broker for as many science cases we can munster (please contact us if you want to the team!), with a strong emphasis of state-of-the-art technology and classification algorithms.

Photometric classification, machine learning and new techniques

One of my main interests is increasing type Ia SNe samples for cosmology by using photometric classification and reducing selection biases. Until now, supernovae are typed (there are many types of supernovae) by using spectroscopy. This technique, although very useful, requires a lot of telescope time. I just released the framework SuperNNova , an open-source photometric supernova classifier that uses Deep Learning. The framework is based in Recurrent Neural Networks and only requires photometry (flux aver time in different band-passes) to classify supernovae accurately in subtypes. With SuperNNova we make a first implementation of Bayesian Neural Networks in supernova classification. We make emphasis on reproducibility, the cope is in GitHub and statistical robustness of our methods. In particular, I am exploring what do classification uncertainties represent when using BNNs. I have also done the first survey application to machine learning classifiers to real data with supervised learning in SNLS .

Finding SNe can be challenging, so in France I worked on improving detection of transient events and classification of supernovae using only photometry in the deferred pipeline of SNLS. I implemented Morphological Component Analysis (MCA) algorithms for signal extraction on subtracted images and developed a new extraction method for transient events detection.