So far, the use of quantum computers has been relatively limited, but researchers are trying to expand it. One way to build a fault-tolerant quantum computer architecture based on silicon qubits is to place a single phosphorus atom on a 2D grid. The calculation is then performed by controlling the logic gate of one or two qubits through nanoelectron lines. However, this method depends to a large extent on the order of magnitude of the phosphorus atom lattice dot, and the uncertainty of the atomic quantum dot is destructive to several orders of magnitude.
In this way, it will result in two quantum bit gates of operation error, the given calculation produced inaccurate results. In a large-scale quantum computing architecture, this effect is exponentially amplified.
To help solve this problem, in 2016, researchers at the University of Melbourne used a computer-scanning tunnel microscope (STM) image of the phosphorus atomic wave function to determine its spatial position on silicon.
This allows a single lattice to be highly accurately located, but the next challenge is how to extend this precise spatial positioning method to large-scale, fault-tolerant quantum computer architectures.
To develop this framework, the researchers now use deep learning tools to conduct computational training on 100,000 orders of magnitude OF STM image sets for convolutional neural networks (CNN) and then attempt to identify 176,000 test images.
It was found that, despite the fuzzy and asymmetric shapes common in the real environment, the convolutional neural network scored more than 98% of the classification accuracy of the test images.
Experiments have shown that this machine learning-based technique can process quantum bit measurementdata with high throughput, high precision and minimal human-computer interaction.
In addition, studies have shown that the technology has the potential to expand qubits consisting of multiple phosphorus atoms. At this setting, the number of potential image configurations can be multiplied.
The team says the machine learning-based technology, which can play a key role in developers of fault-tolerant universal quantum computers, is the ultimate goal of researchers around the world.
Details of the study have been published in the recently published journal Nature, originally titled: Framework for – Character leveling of quantum asty s arrays by machine learning.