Google will control the qubits of its quantum computer thanks to a cryogenic controller.
Traditional data processing and quantum data processing
In traditional computing, information is processed in transistors connected to each other within a microprocessor. This information is measured in "bits" (for "binary digit"). They perform calculations through logic gates. Logic gates are materialised by the transistors of the microprocessor. The two possible values of a bit are 0 or 1, which results, in a basic way, in the absence or the presence of electric current.
The traditional calculations (addition, subtraction, multiplication), are feasible from an assembly of logic gates.
For example, the logic gate "not" reverses the value of a bit. During a calculation, if the value of the bit is 1 at the input in the transistor, it will be 0 at the output, or vice versa. In traditional computing, if one wishes to carry out this calculation with, as successive doors, the entry 0, then 1, the computation will have to be completed twice.
Quantum computing allows both calculations to be performed at the same time. Indeed, in quantum computing, the unit of measure for information is the "quantum bit". It is characterized by its ability to simultaneously have a value of 0 and 1. This is the principle of quantum superposition.
In computer science, registers are spaces for storing information.
In traditional computing, a 4-bit register can generate 16 states: 0000, 0001, 0010, 0100 ... In quantum computing, a register of 4 qubits can superpose these 16 states. A register of 8 qubits would superpose 256 states, and thus go 256 times faster than a traditional computer. Nevertheless, quantum computers are not able to solve all the problems that conventional computers are working on.
But some calculations are out of reach of conventional computers. In quantum computing, quantum gates make these calculations possible, and allow them to be performed quickly. An example of the application of quantum computing is the factorisation of large whole numbers into prime numbers.
Simplified operation of a qubit
In a quantum processor, it is necessary to have a system that can superpose two states at the same time, and that is small enough to obey the laws of quantum mechanics. Various systems are thus conceivable to create qubits:
Since the early 2000s, large groups like IBM, Intel or Google have been developing quantum computers, but their capabilities are limited. The first calculation performed by a quantum computer was done by IBM in 2002. It was the factorisation of the number 15 (3x5). This computer contained 7 qubits.
Quantum supremacy
The ultimate goal of high-tech firms is to achieve "quantum supremacy". Quantum supremacy is the moment when the computing power of the quantum computer exceeds that of conventional supercomputers. It is considered that this supremacy could be reached at the threshold of 50 qubits.
In March 2018, Google introduced Bristlecone, a processor of 72 qubits, exceeding this theoretical threshold. This experimental system should make it possible to advance research on error rates in this type of system, as well as that on qubits. Google has not released results for Bristlecone trials yet, but the company is still improving its processor. For example, the company is currently developing technologies to improve control over the flow of qubits.
Bristlecone and its cryogenic controller
For now, Bristlecone operates with 72 logic qubits. To do so, it needs several hundreds of thousands of physical qubits, thanks to a superconducting system. Qubits are stored in a cryostat cooled to a temperature of 10 mK (-273.14°C). The very low temperatures make it possible to improve the operation of the qubits and limit the risks of quantum decoherence^{2}. One of the main goals of Google teams is to reduce error rates in calculations. It is therefore necessary to perform controls on the qubits.
Currently, control is done using individual coaxial cables that transmit signals to each physical qubit. The number of cables is extremely important. They connect the cryostat to server racks at room temperature. These cables generate 1 watt of lost power per qubit. The number of cables could be reduced by being directly integrated into the cryostat, but the heat generated would be too great for the cryostat.
In order to solve this problem, Google has presented its cryogenic single-bit CMOS controller ("Complementary Metal Oxide Semiconductor"), at the end of February 2019. This integrated circuit measures 1 mm on 1.6 mm and it consumes less than 2 milliwatts of power, according to the press release of the company4. As its name suggests, it can only work with a single bit, which limits its use for the moment.
The cryogenic controller was first tested at room temperature and then placed in a portion of the cryostat cooled to 3 K (-270.15°C). It was connected to a qubit in the cooled part of the cryostat at 10 mK. A series of experiments then established that the controller was operating as expected and did not cause overheating in the cryostat. According to Google, processor performance is similar to that of a system using coaxial cables.
The reduction of energy consumption of this new controller is promising, according to the firm.
^{1} https://www.lemagit.fr/definition/Qubit (in French)
^{2}Quantum decoherence is the problem of the disappearance of superposition quantum states at the macroscopic lebel. https://www.sciencesetavenir.fr/fondamental/la-decoherence-prise-sur-le-vif_22114 (in French)
^{3}https://ai.googleblog.com/2018/03/a-preview-of-bristlecone-googles-new.htmlEzrtr
^{4}https://ai.googleblog.com/2019/02/on-path-to-cryogenic-control-of-quantum.html