Helping quantum computers to scale with optical interfaces
We just published a report in Nature Nanotechnology detailing some exciting device performance we have realised at QphoX. We have put together this blog post to provide additional plain-language explanation and context for our results.
Quantum computers are expected to tackle challenging problems which are beyond the capabilities of even the most powerful classical supercomputers today. These include chemical simulations for the development of new medicines, information security protocols and research into new advanced materials. A major challenge in building useful quantum computers, however, is that even small error rates can interfere with calculations, and some of the most advanced quantum computing approaches therefore require ultra-low temperatures (close to absolute zero) to operate effectively. At the same time, full error correction for calculations will require tens of thousands or even millions of qubits. A key obstacle is finding a way to interface these qubits without overloading cryogenic systems. Typically bulky coaxial cables are used to connect qubits to the outside world, but this limits scalability.
Fortunately, classical computers and networks have already overcome similar challenges, and provide an outlook on how it will be possible to solve these difficulties for quantum computers. In a classical computer billions of transistors are fabricated onto chips. These chips are interconnected locally via electrical wiring, but over longer distances via optical fibers. The optical fibers lose less energy and have a much larger bandwidth for carrying information than their electrical counterparts. The key technology which makes supercomputers and the internet possible are converters between electrical and optical signals.
The advantage of using optical interconnects is even more critical for quantum computers. The signals used to carry quantum information within a quantum computer would be drowned out by thermal noise at room temperature. Optical frequency signals (100,000 times higher in frequency) remain noise-free at these elevated temperatures - allowing for quantum information to be brought out of the cryostat. As such multiple quantum computers could be connected together through an optical network. These interconnects are being developed for converting quantum signals, but none have yet reached the thresholds required to be helpful in scaling quantum computers.
What problems does this report address?
The main differences between classical converters and quantum converters (or transducers) are added noise and efficiency. Because quantum signals are so fragile and minute, the converter must simultaneously be very sensitive to the signal and very efficient. In recent years, a number of research groups have developed systems which meet these requirements. However, for practical application in a quantum computer, the transducer must have high operation rate, high bandwidth, and low thermal heat load. Scaling up also requires the ability to integrate many transducers in a small volume. Achieving all these practical requirements simultaneously is a major outstanding challenge.
What do we show?
In this work we explore a new materials platform: lithium niobate, a piezoelectric material, on silicon, the substrate used for the majority of electronics. We show that we can convert weak microwave signals to optical frequencies at levels down to the single photon level, by first converting to mechanical motion in the lithium niobate and silicon device layers. It is this mechanical step that enables the conversion process – reducing the amount of optical power we need to provide in order to convert the signal.
In the demonstrated transduction process we add only a few extra photons of noise and transfer signals with an efficiency of up to 1%. For comparison, a typical classical transducer adds billions of photons of noise with millions times lower efficiency. Because of the small excess noise, the current generation of devices cannot yet convert the smallest and most fragile quantum signals. Our device does, however, already meet the requirements for connecting the quantum computers to the outside world by carrying the input and output signals through fibers, replacing bulky and heat conducting electrical wires.
Microwave-to-optics transduction allows quantum computers at the base temperature of two dilution fridges to be connected together via optical fibers (red). Thanks to the long distance communication enabled by optical light the computers could additionally be connected into a large-scale quantum network, or internet.
QphoX's integrated microwave-to-optics transducer, captured with a scanning electron microscope. The whole image on the right is less than half the width of a human hair. The conversion is enabled by the small lithium niobate block (yellow).
Our transducer shows great potential for meeting the needs of a full-scale quantum computer. The transduction has a large bandwidth and operation rate, which match well with the expected signals needed to interface with a superconducting quantum computer. The area of the transducer is less than 0.15 mm2, which means that it is possible to fabricate over ten thousand transducers onto a standard 100 mm-wide silicon wafer. The minimal power required to allow for conversion allows us to pack many devices into the cold area of the cryostat without significantly heating it up. With this technology, we expect to be able to connect thousands of qubits to the outside world, using only tens of optical fibers due to the extra bandwidth provided by optical channels. In comparison, the largest quantum computers today are composed of 100-200 qubits. This means we can already start scaling up our design to meet the present day needs of quantum computers.
With our transducer performance we are ready to interface with contemporary quantum computers, such as those based on superconducting qubits. In the near future, our technology will be able to improve the input and output connections of these quantum computers. For example, one type of output is used to readout the state of a qubit, an important step for obtaining the results of a quantum computation and diagnosing errors. Typically this requires a wire and cryogenic amplifiers, but with converters many wires and amplifiers can be consolidated into a single fiber, allowing for more qubits to be cooled down simultaneously. In the coming years, we will develop the next generation of devices to convert quantum signals from the cryogenic electrical domain to the optical domain. These quantum signals can be passed between cryostats to create the first linkages between remote quantum computers, joining them into a larger distributed quantum computer. Distributed quantum computers such as these will be more powerful and flexible than their local counterparts.
In summary, we expect quantum electro-optic converters to be a foundational building block in quantum computers with tens of thousands to millions of qubits. These large scale quantum computers will explore new computational tasks which evade today's most powerful supercomputers.