We view the universe at large through the particles created by astronomical objects. Charged particles (called cosmic rays) must traverse the weak but enormous magnetic fields that span interstellar space. These fields bend the trajectory of charged particles, mixing them up. The result is that it is impossible to tell where these particles come from.
In order to do astronomy, we must be able to trace particles back to their source. This can be done by looking for uncharged particles, which ignore the magnetic fields of the galaxy. One such obvious candidate is the photon; the particle of light. We have been doing astronomy for thousands of years by gazing at the night sky.
The neutrino is also uncharged, and so it is possible to do astronomy using it as well. The IceCube experiment was built to find these astrophysical neutrinos. They are the products of pulsars, galactic nuclei, and other intense and exotic objects in our universe.
IceCube is a one billion ton particle detector, the largest in the world. It is located at the Amundsen-Scott south pole station, below the Antarctic ice sheet. Around 2km below the surface, the immense pressure squeezes the bubbles out of the ice. This process creates the clearest ice on earth, exactly the kind of material needed to make a giant particle detector.
To do neutrino astronomy, we need to be able to see individual objects emitting neutrinos. These “point sources” can be correlated with existing catalogs to provide a better understanding of our cosmos.
Along with Nick Rodd and Ben Safdi of MIT, I am applied the Non-Poissonian Template Fitting (NPTF) statistical test to data collected by IceCube. Previously, this test has been applied by Ben Safdi and others to the galactic center gamma-ray excess in FermiLAT data. The gamma-ray excess has been interpreted as a possible signal from dark matter annihilating in the center of our galaxy. The test showed that the data fit unresolved point sources below Fermi’s detection threshold, rather than a diffuse flux. A diffuse flux would have been expected if this signal originated from dark matter, the point-like nature of the sources suggests that the gamma-rays were generated by something else.
IceCube has seen evidence of high-energy neutrinos that come from outside our solar-system. However, these neutrinos are consistent with an isotropic distribution — we don’t see individual objects. We brought the NPTF test to IceCube, as it can be more sensitive than traditional methods in certain situations, such as sources of neutrinos within our own galaxy.
When cosmic rays strike the atmosphere of Earth, they initiate a nuclear reaction that spews out many exotic particles that quickly decay into neutrinos. While these atmospheric neutrinos form a background for astrophysical studies, they can be used for particle physics.
IceCube has performed a search for atmospheric muon neutrino disappearance — a signal for the existence of sterile neutrinos. This study has set a world leading limit on muon neutrino disappearance, and has had significant impact on global fits for sterile neutrino models. I created the atmospheric neutrino models using the MCEq software package and weather satellite data. These models were important for understanding the effect that seasonal temperature variations have on the production of muon neutrinos.
Watch a video on the sterile neutrino search
Read about the result on MIT news
With the support of a American-Australian association Bechtel fellowship, I investigated the application of deep learning to event classification in IceCube. IceCube is sensitive to astrophysical tau neutrinos, but common events are nearly indistinguishable from electron neutrinos. Convolutional neural networks may be able to improve on previous searches by learning correlations between detector modules. The positive identification of even a single tau neutrino would allow IceCube to put new constraints on astrophysical models. I showed that a deep neural network can be competitive with contemporary algorithms.
