Fault injection for Quantum Circuits

Quantum computing is an alternative type of computation that has the potential of outperforming classical computation in some specific problems. While shown to be theoretically possible, currently there are several technical challenges that prevent the construction of a quantum computer big enough to be useful. Among these, there is the fact that qubits, the fundamental elements of quantum computation, are highly susceptible to external sources of faults, such as ionizing radiation. These faults are also called transient faults, due to their temporary nature. The work done in this thesis aims to better understand how transient faults propagate and corrupt the execution of quantum circuits, through the simulation of faults impacting the quantum circuits. In order to simulate the fault, we use a fault injector built on the Qiskit framework, one of the most widely used frameworks for quantum computing. The injector simulates the fault by inserting into the tested quantum circuit a parametrized quantum gate, a U-gate, in a specific location in the circuit in order to arbitrarily change the state of the qubit at that point of the execution. By inserting arbitrary faults in arbitrary locations of the tested quantum circuit, we can study its different vulnerabilities. We perform three fault injection analyses: in the single fault injection, we inject one fault at a time in three different circuits; we learned that indeed different circuits can have different fault vulnerabilities: the same fault can be well tolerated by some circuits while being critical for others. In the double fault injection, we simulate a particle hitting two neighboring qubits instead of just one, by adding a second fault; as expected, we found that adding a second fault worsens the quality of the output in all cases. Lastly, in the injection on quantum circuit cuts, we apply our single fault injection analysis on a novel technique in the field of quantum computing research, which is quantum circuit cutting, a process that allows splitting a large quantum circuit into multiple, smaller subcircuits, which can then be independently executed in order to reconstruct the full output of the circuit. Here, by cutting a circuit and injecting faults in one of its subcircuits, we can study how faults propagate from a subcircuit to the final reconstructed output; we found that given a circuit, its subcircuits can have different vulnerabilities to faults, and some of them can be much more critical than the others. The work done in this thesis is an attempt to better understand fault propagation in quantum circuits. We believe some of these results can be used as first steps towards future work regarding, for example, the physical protection of quantum devices, by helping to concentrate the protection efforts to the more critical parts of the circuits.