Particle physics experiments have seen puzzling signals in short‑baseline neutrino experiments. These include results from LSND, MiniBooNE, and gallium detectors. One way to explain those signals is to add a fourth kind of neutrino that does not interact in the usual way, called a “sterile” neutrino. The simplest version of that idea, known as the 3+1 model, is already hard to reconcile with other experiments and with cosmology. This work tests a non‑minimal variant in which the heavy fourth state (called ν4) is unstable and decays into two invisible particles.

The author used 10.67 years of high‑energy atmospheric neutrino data collected by the IceCube Neutrino Observatory to look for signs of that unstable sterile neutrino. IceCube is a large detector that observes neutrinos that pass through the Earth. The search compared the data to predictions from the unstable 3+1 hypothesis and to the standard no‑decay 3+1 case. The unstable model would change how many neutrinos of different flavors and energies arrive at the detector, because the heavy state would disappear by decaying.

The analysis found no preference for the unstable‑decay model over the no‑decay 3+1 model. It excludes most of the parameter space that global fits to the short‑baseline anomalies had favored, at the 90% confidence level. In plain terms, this is a strong constraint on this particular non‑minimal explanation for the anomalies. It does not claim to rule out all possible sterile‑neutrino ideas, but it removes a large portion of the region that was compatible with those earlier anomalous results.

Beyond the direct search, the thesis presents technical work that prepares IceCube for future sterile‑neutrino studies. The author produced a comprehensive calculation of neutrino–nucleon and neutrino–nucleus deep‑inelastic scattering cross sections — the basic probabilities for high‑energy neutrinos to interact with matter. New machine‑learning reconstruction methods were developed to better infer particle properties from the detector signals. An improved event selection was also created, yielding about twice the signal efficiency of the previous analysis. Together these tools allow a statistical separation of neutrinos and antineutrinos and set up a future search that targets resonant disappearance of antineutrinos.