Researchers at QuTech in Delft, The Netherlands, have developed a new chip architecture that could make it easier to test and scale up quantum processors based on semiconductor spin qubits. The platform, called QARPET (Qubit-Array Research Platform for Engineering and Testing) and reported in Nature Electronics, allows hundreds of qubits to be characterized within the same test-chip under the same operating conditions used in quantum computing experiments. "With such a complex, tightly packed quantum chip, things really starts to resemble the traditional semiconductor industry." states researcher Giordano Scappucci.
When viewed under a microscope, the structure of the QARPET chip appears almost woven. Fabrication was in fact a stress test for engineering capabilities. "When I designed the first layouts, I honestly did not expect them to work," says Alberto Tosato, who did the engineering. "The number of crossing electrodes is extremely high. It pushes the limits of nanofabrication, we saw it as a test that would probably fail. So, seeing the device come alive at millikelvin temperatures, that was a very satisfying moment."
The device was engineered to tackle a key challenge for quantum technologies: how to efficiently evaluate large numbers of qubits, especially as devices with millions of quantum bits get developed. "Building a large-scale quantum processor is not just a matter of adding more qubits," says Giordano Scappucci, Associate Professor at TU Delft and lead researcher. "To make progress, we need to understand how qubits perform statistically, how uniform they are, how noisy they are, and how these properties vary across a chip. This really sets QARPET apart."
A tiled approach to qubit measurement
QARPET is designed to simplify the evaluation process. Instead of fabricating and testing each qubit device separately, the researchers created a grid of small, repeatable 'tiles'. Each tile contains two spin qubits and one charge sensor, forming a self-contained unit that can be individually measured.
These tiles are connected in a crossbar layout, where rows and columns share control lines, similar to the architecture used in computer memory. This design means that a single tile can be selected and probed without adding complex wiring or cryogenic electronics. As a result, the total number of control lines increases slowly as the array grows, making the approach scalable.
The first demonstration chip, made from a germanium/silicon-germanium (Ge/SiGe) semiconductor structure, includes 23 by 23 tiles – enough to potentially host up to 1,058 hole-spin qubits. Still, the entire array requires only 53 control lines, illustrating the efficiency of the crossbar scheme. "This device reaches a potential density of about two million qubits per square millimetre, which highlights how compact semiconductor spin qubits can be," says Scappucci. "In fact, with the right measurement infrastructure and automation, the chip we built could already allow us to probe more than a thousand qubits in a single cooldown."
From large-scale measurements to strategic insights
Using high-frequency electrical readout techniques, the team focused on the measurements of a subset of 40 tiles on the chip, demonstrating that nearly all could be addressed and tuned independently. From these measurements, key device parameters such as threshold voltages, charge noise levels, and variations in dot formation were extracted. The results revealed a high degree of consistency across the array, while also highlighting small variations that reflect differences in the underlying material and fabrication processes. Furthermore, the team demonstrated, as proof of principle, that the device could support spin qubits, with their properties not affected by the architecture.
These types of statistical insights are important for improving the reproducibility and reliability of future quantum devices. QARPET thus effectively provides a testing vehicle for evaluating new semiconductor materials and device designs under realistic conditions.
Toward scalable quantum technologies
Because the QARPET platform is modular and compatible with existing semiconductor fabrication techniques, it could be adapted to other material systems, including silicon-based qubits. The design also lends itself to automated or machine-learning-assisted tuning, which could further accelerate device optimization.
"Our platform brings together a realistic device architecture with the possibility to extract meaningful statistics from hundreds of qubits in a single experiment," Scappucci concludes. "Given the complexity and density of this chip, demonstrating that it actually works is an important milestone. It shows that we can already build and study the kind of large qubit arrays that future quantum processors will rely on."
