NSREC 2024 – Understanding the Contribution of Terrestrial Radiation Sources to Quantum Devices Error Rate
Quantum Computing (QC), despite being a highly promising computational paradigm, suffers from an incredibly high radiation sensitivity. Recent discoveries highlighted that the impact of a particle in the quantum bit (qubit) is tens of thousands times more likely to induce a fault compared to traditional CMOS devices. Moreover, the deposited charge quickly diffuses in the substrate affecting multiple qubits, inducing faults that can persist for hundreds of seconds.
In this paper, we aim to better understand the effect of different radiation sources and mechanisms of energy propagation on quantum devices. We present data from the simulation of more than 18 billion particle interactions.
Through GEANT4 simulations, we compare the effect of neutrons, alpha particles, muons, and gamma rays in a quantum device. We combine non-equilibrium generation probability with natural flux to identify the most harmful radiation source for qubits. We found that muons are, by far, the more likely cause of faults in qubits. Moreover, through G4CMP simulations, we track the energy propagation within the substrate. We show that even particle hits far from the qubit can lead to energy transmission to the superconductor, also pointing out that this mechanism is 1,000 more likely than a direct energy deposition on the qubit. In addition, we show that the time persistency of secondary particles in the substrate is in the order of O(100 μs). Finally, we look at particle impacts on a four-qubit device to show that with a common layout, multiple-qubit are likely to be corrupted.
Download and read the full paper here:
Devices tested
SQUID layout. The aluminium superconducting loop (grey) is the sensitive volume of the experiments together with the Josephson junctions.
Qubit layout. In grey the substrate, in cyan the superconducting layers and in brown the copper frame.
SQUID analysis
Probability for 1 MeV (blue), 10 MeV (red), and 100 MeV (green) neutrons to deposit energy on the SQUID.
Energy deposition probability for six packets of 10^8 muons, at energies 100 MeV (blue), 1 GeV (red), and 10 GeV (green). The solid (μ+) and dashed line (μ−) overlaps.
Energy deposition probability for three packets of 10^8 photons, at energies 100 keV (blue), 1 MeV (red), and 2 GeV
(green).
Energy deposition probability for six packets of 10^8 alpha particles, at energies 1 MeV (cyan), 3 MeV (blue), 5 MeV (orange), 7 MeV (red) and 10 MeV (green).
Different particles contribution. The left axis shows the cross-section (blue) and the right one is the cross-section multiplied by the flux of that particle at sea level (red).
Superconductor analysis
Energy deposited by muons of different energies (100 MeV in blue, 1 GeV in orange and 10 GeV in green) by direct interaction with the substrate and the superconductor.
In the embedded figure a zoom-in of the deposition on the superconductor itself.
Energy deposition on the superconductor after phonon absorption over time. Phonons are generated from the interaction of 10 GeV muons with the Silicon substrate of the chip.
Energy absorbed by the superconductor with respect to the impact point. Top-down view of the Silicon substrate.
The qubit is centred at coordinates (0, 0).
Energy deposited by 100 MeV positive muons across four qubits on a Silicon substrate. From left to right, the muon generation coordinate changes from the centre of the substrate (equidistant from all qubits), to the midpoint between two qubits, and finally to be centred on one of the four qubits.
