Masters defense: Christian Bjørnholt

Title: Analyzing spin-qubit traces and dynamical quantum phase transitions using machine learning

Abstract: This thesis investigates how machine learning techniques can be applied in the study of physics. The first topic is about improving the readout speed of semiconductor qubits by employing principal component analysis. This results in a one microsecond measurement time compared to the currently used method of homodyne detection that requires several microseconds. The time can be lowered even more by systematically discarding signals that yield unclear results. The second topic is about dynamical quantum phase transitions, stemming from non-equilibrium phase transitions that occur when a quantum state is evolved through time. They are studied through Loschmidt rate functions by finding their non-analytic points. By making recurrent neural networks evaluate these functions, it is possible to tell if they contain dynamical quantum phase transitions or not. More complex variations can even specify where they occur.

Join Zoom Meeting

Meeting ID: 669 3579 3254
Passcode: 641232