Researchers in Australia say that they have created the first CMOS chip that can control the operation of multiple spin qubits at ultralow temperatures. Through an advanced approach to generating the voltage pulses needed to control the qubits, a team led by David Reilly at the University of Sydney showed that control circuits can be integrated with qubits in a heterogeneous chip architecture. The design is a promising step towards a scalable platform for quantum computing.
Before practical quantum computers can become a reality, scientists and engineers must work out how to integrate large numbers (potentially millions) of qubits together – while preserving the quantum information as it is processed and exchanged. This is currently very difficult because the quantum nature of qubits (called coherence) tends to be destroyed rapidly by heat and other environmental noise.
One potential candidate for integration are the silicon spin qubits, which have advantages that include their tiny size, their relatively long coherence times, and their compatibility with large-scale electronic control circuits.
To operate effectively, however, these systems need to be cooled to ultralow temperatures. “A decade or more ago we realized that developing cryogenic electronics would be essential to scaling-up quantum computers,” Reilly explains. “It has taken many design iterations and prototype chips to develop an approach to custom silicon that operates at 100 mK using only a few microwatts of power.”
Heat and noise
When integrating multiple spin qubits onto the same platform, each of them must be controlled and measured individually using integrated electronic circuits. These control systems not only generate heat, but also introduce electrical noise – both of which are especially destructive to quantum logic gates based on entanglement between pairs of qubits.
Recently, researchers have addressed this challenge by separating the hot, noisy control circuits from the delicate qubits they control. However, when the two systems are separated, long cables are needed to connect each qubit individually to the control system. This creates a dense network of interconnects that would prove extremely difficult and costly to scale up to connect millions of qubits.
For over a decade, Reilly’s team have worked towards a solution to this control problem. Now, they have shown that the voltage pulses needed to control spin qubits can be generated directly on a CMOS chip by moving small amounts of charge between closely spaced capacitors. This is effective at ultralow temperatures, allowing the on-board control of qubits.
CMOS chiplet
“We control spin qubits using a tightly integrated CMOS chiplet, addressing the interconnect bottleneck challenge that arises when the control is not integrated with qubits,” Reilly explains. “Via careful design, we show that the qubits hardly notice the switching of 100,000 transistors right next door.“
The result is a two-part chip architecture that, in principle, could host millions of silicon spin qubits. As a benchmark, Reilly’s created two-qubit entangling gates on their chip. When they cooled their chip to the millikelvin temperatures required by the qubits, its control circuits carried out the operation just as flawlessly as previous systems with separated control circuits.
While the architecture is still some way from integrating millions of qubits onto the same chip, the team believes that this goal is a step closer.
“This work now opens a path to scaling up spin qubits since control systems can now be tightly integrated,” Reilly says. “The complexity of the control platform has previously been a major barrier to reaching the scale where these machines can be used to solve interesting real-world problems.”
The research is described in Nature.
The post Spin-qubit control circuit stays cool appeared first on Physics World.
