Cryogenic Recharacterization of a Commercial 22 nm FD-SOI Foundry Standard-Cell Library

Quantum computers promise to solve problems that are currently impossible for even the most powerful supercomputers, but they come with a severe logistical challenge. These machines must operate at temperatures colder than deep space to function, yet the electronics that control them usually sit in warm rooms. To bridge this gap, engineers must run thousands of wires from the freezing interior of the quantum processor up to room-temperature equipment. This creates a massive bottleneck, limiting how much information can be sent and received. The solution is to move the control electronics inside the cold environment itself, placing digital logic right next to the quantum bits. However, standard computer chips are designed to work at room temperature, and their behavior changes drastically when frozen. The question facing researchers is whether these mass-produced chips can be made to work reliably in the deep cold without being completely redesigned from scratch.

A team of researchers set out to answer this by taking a standard, commercially available computer chip library and testing how it behaves at four Kelvin, a temperature just a few degrees above absolute zero. They focused on a specific type of modern chip technology known as fully depleted silicon-on-insulator, which is widely used in industry. Instead of inventing new circuits, they kept the existing designs exactly as the chip manufacturer created them, including the physical layout of the transistors and the logic functions. Their goal was to see if they could simply update the mathematical descriptions of how these chips work at low temperatures, allowing engineers to use standard design tools to build systems for the cold. The researchers found that while the chips slow down significantly when frozen, a technique called forward body bias can restore their speed. This method involves applying a specific electrical voltage to the back of the transistors, which counteracts the freezing effects and allows the chips to run at usable speeds without any changes to their physical structure.

The study revealed that without this special voltage adjustment, the chips become incredibly sluggish. When the researchers tested the library at zero bias, the time it took for a signal to travel through a basic logic gate increased by a factor of thirty-eight for rising signals and thirty-three for falling signals. In practical terms, a circuit that could switch states billions of times a second at room temperature would slow down to a crawl, operating at only a few million cycles per second. Furthermore, the signals became so slow that they violated the safety limits set by the chip manufacturer, potentially causing errors in more complex designs. The researchers also discovered that the chips behaved differently depending on the specific type of gate, meaning a single rule could not be applied to the entire library to predict performance. This variability meant that simply guessing how the chips would behave was not enough; a detailed, point-by-point re-evaluation was necessary.

To fix the speed issue, the team applied forward body bias, which essentially pushes the transistors to turn on more easily despite the cold. When they applied a voltage of plus or minus 1.8 volts, the performance improved dramatically. The delay penalties dropped from factors of thirty to factors of less than three, bringing the chips back into a range where they could be useful for digital control. At this level of bias, the chips could operate at speeds up to 132.9 megahertz, a massive improvement over the 1.3 megahertz achieved without the bias. This speed is significant because it exceeds the 50 megahertz limit of the input-output pads used in the test chip, proving that the internal logic is no longer the bottleneck. The researchers confirmed these findings by analyzing a fabricated 32-bit arithmetic logic unit, a small but complex calculator built into the chip. They kept the physical design exactly the same and only changed the library data to reflect the cold conditions, showing that the same physical chip could span a performance range of more than two orders of magnitude just by changing the operating voltage.

Another critical challenge addressed in the study was the difference in voltage levels between the chip's core and its input-output connections. The core of the chip operates at a very low 0.4 volts to save energy and reduce heat, but the connections to the outside world operate at 0.8 volts. Moving a signal from the low-voltage core to the higher-voltage outside world requires a translator. The researchers found that if they tried to skip this translator and drive the higher voltage directly from the low-voltage core, the chip would waste a tremendous amount of power. At full bias, this bypass method would consume 57.7 microwatts of static power for every single bit of data, whereas using the proper translator reduced that waste to less than one billionth of a watt. This finding highlights that while the chips can be made fast enough for control, the interfaces between different parts of the system must be carefully managed to avoid draining the limited cooling budget of the cryogenic environment.

The study also looked at the energy efficiency of these cold chips compared to their room-temperature counterparts. While the chips run slower at low temperatures without bias, they become surprisingly efficient when the forward body bias is applied. The researchers calculated that the energy required to perform a task while maintaining a certain speed actually improved at the cold temperatures when the bias was used. This suggests that placing control electronics inside the cryostat is not just feasible but potentially more efficient than running them at room temperature, provided the right voltage settings are used. The work demonstrates that the industry's existing chip designs are robust enough to handle the extreme cold, provided engineers use the correct modeling tools and voltage strategies.

Ultimately, this research provides a clear path forward for building scalable quantum computers. By showing that commercial chip libraries can be recharacterized for cryogenic use without redesigning the transistors, the team has removed a major barrier to entry. Engineers can now use the same design software and standard cells they use for room-temperature computers to build systems that operate near absolute zero. The key is to apply the right amount of forward body bias to recover the lost speed and to use voltage translators to manage the interface between the cold core and the warmer outside world. While the actual measurements of the silicon at these temperatures are still future work, the simulations and analysis presented here offer a high degree of confidence that the approach will work. This means that the complex wiring and bandwidth bottlenecks that currently limit quantum computing can be solved by placing the control logic right where it is needed, inside the cold, using the very same chips that power our everyday devices.