Challenge
Much the same as a ‘bit’ in classic computing, a ‘qubit’ is used to encode information in quantum computing. Manipulating these qubits in a circuit for computation can be achieved using nanometre-thin junctions, which are often constructed using aluminium oxide tunnel barriers.
Noise in these tunnel barriers can cause errors, which is a major hurdle to efficient quantum computing. This noise is believed to arise from the interaction of the qubit state with defects comprised of two-level systems, but the origins of this noise are still not widely understood. By examining conduction within these junctions at the atomic scale, researchers hope to better understand the movement of a charge, and therefore improve the efficacy of the junction.
While barrier physics is an old problem, it has traditionally been approached from a relatively simple perspective, whereby the shape of the barrier is assumed. These simplifications were required, as the computational power to resolve greater detail in the barrier was all but out of reach for material scientists.
Solution
World-class supercomputers such as Gadi allow scientists to perform incredibly complex research at the atomic level. For researchers studying aluminium oxide tunnel junctions, the computational performance of Gadi opened up the possibility of analysing the formation of these junctions atom-by-atom.
Researchers from RMIT first benchmarked the molecular dynamics software that would be used in tunnel simulations, to ensure that the interactions between atoms at scale would be accurate. At this atomic level, it is also important to accurately simulate the flow of heat within the system, so a thermostat was added to the simulation to maintain all simulated atoms at the same temperature.
The above requirements were developed into code that acted as a wrapper for the remainder of the molecular dynamics simulations, which would then be used to perform the atom-by-atom study of the tunnel junctions. Previous research by RMIT researchers was also leveraged to compare these new models against older ones as they grew.
Dr Martin Cyster from RMIT and the ARC Centre of Excellence in Exciton Science explained that access to NCI “was absolutely essential to complete this intensively computational project”.
“Adding each oxygen atom to the aluminium surface sequentially means that we are simulating a much longer time period than we would if all the oxygen was introduced at once. We therefore needed many hundreds and thousands of CPU hours to get the work done.”
Since the final atomic configuration is dependent on a randomised process, these simulations had to be performed a number of times to ensure the accuracy of the results. More than 512GB of memory was required to calculate the distribution of current in the three-dimensional barrier system.
Future
Where classical computing has transformed modern life, quantum computing would represent an exponential leap forward in our understanding of the universe. Cracking the secret to fabricating high-quality tunnel junctions for use in superconducting circuits would bring researchers a step closer to this important goal.
“Presenting on this topic at conferences and meetings always generated a lot of interest from experimental scientists working on improving the junction fabrication process”, explained Dr Cyster.
The researchers behind this work are already looking at ways to improve their calculations even further, such as by including the physics of superconductivity.
