The Sycamore quantum computer is programmable and can run general-purpose quantum algorithms. The Google team has been working on near-term applications such as quantum physics simulation, quantum chemistry and generative machine learning.
The Sycamore processor is a 54-qubit processor comprised of a two-dimensional grid where each qubit is connected to four other qubits. 1 qubit failed, leaving 53 qubits for computation. The qubit states interact throughout the entire processor such that the overall state is impossible to emulate efficiently with a classical computer. The Sycamore Quantum Computer was used to demonstrate so-called quantum supremacy in a computational task which was reported in Nature in October 2019. The study was accidentally leaked online in the previous month. The goal of Google’s quantum supremacy experiment was to perform a contrived calculation involving 53 qubits that would likely take 253 or 9 quadrillion steps to simulate with a classical computer. Quantum supremacy is the crossover point where the smallest computational task is prohibitively hard for classical computers but not for quantum computers. The term “quantum supremacy” was coined by physicist John Preskill in 2012.
The Nature report was led by John M. Marinis of Google AI Quantum and University of California, Santa Barbara. The task to demonstrate quantum supremacy was sampling of the output of a pseudo-random quantum circuit. Sycamore performed the target computation in 200 seconds while the world’s fastest computer called Summit, a supercomputer built by IBM for the Department of Energy, would be expected to take 10,000 years to produce a similar output. IBM rebutted Google’s claim by announcing that by tweaking the way Summit approaches the task that it can complete it in 2.5 days and stated that the threshold for quantum supremacy had not been met. There are variations in interpretation for quantum supremacy and others have stated that the significance of the Sycamore quantum computer is that it performs in fast polynomial time whereas classical computers perform in slow exponential time. This means that when a quantum computer gets to 70 qubits, a classical supercomputer would need to occupy the area of a city to keep up for particular calculations.
For the quantum supremacy experiment, Sycamore was challenged to describe the likelihood of different outcomes from a quantum version of a random-number generator. This was done by running a circuit that passes 53 qubits through a series of random operations and generates a 53-digit string of 1s and 0s with 253 possible combinations. Owing to interference between qubits some strings of numbers are more likely to occur than others, similar to rolling a loaded die. Sycamore calculates the probability distribution by sampling the circuit, running it one million times and measuring the observed output strings. The method is described as similar to rolling a loaded die to reveal its bias. The solutions were verified from simulations of smaller and simpler versions of the circuits done by classical computers including the Summit supercomputer at Oak Ridge National Laboratory. It was by extrapolating from these examples that the Google team estimated that simulating the full circuit would take 10,000 years on a computer with one million processing units.
The Sycamore microchip is comprised of 53 loops of wire around which current can flow at two different energies to represent 0 or 1. The chip is placed into a closet-sized dilution refrigerator that cools the wires to a hundredth of a degree above absolute zero, so that the wires superconduct. This cooling causes the energy levels to briefly behave as quantum bits or qubits that are in superpositions of the 0 and 1 states. The Sycamore qubits are laid out in a roughly rectangular grid and are able to interact with neighboring qubits. The programming of Sycamore involves classical computers sending wires to the quantum computer to tell each qubit how to behave and which neighbors to interact with and when.
