"Nine Chapters" and "Zu Chongzhi" Double Upgrade The future of quantum computing is coming

As a solid-state quantum computing scheme, superconducting quantum computing has the advantages of good scalability, long coherence time of qubits, fast operation speed, high fidelity, and mature processing technology, while optical systems have the advantages of easy manipulation of photons, small decoherence, operation at room temperature and can be used for long-distance communication, so they are all physical implementation platforms that have attracted much attention in the field of quantum information.

Recently, the team of Pan Jianwei of the University of Science and Technology of China (hereinafter referred to as the University of Science and Technology of China) developed a 66-bit programmable superconducting quantum computing prototype "Zu Chong Zhi 2.0", which achieved the superiority of quantum computing in the random line sampling task, and the difficulty of the task was 2-3 orders of magnitude higher than that of Google 'planewood' in 2019.

At the same time, Pan Jianwei's team upgraded version of "Nine Chapters 2.0" greatly improved the quantum advantage: for the Gaussian Bose sampling problem, the task that the "Nine Chapters" one-minute could complete 1 year ago, the world's most powerful supercomputer needed to take hundreds of millions of years; and the task completed in one minute of "Chapter Nine 2.0" would take another 10 billion times more time than the supercomputer. And "Chapter IX 2.0" also has some programmable capabilities.

The emergence of "Nine Chapters 2.0" and "Zu Chong Zhi 2.0" has made China the only country to achieve the superiority of quantum computing in two physical systems.

The main track to achieve the superiority of quantum computing

The powerful computing power of quantum computing will bring subversive changes to human society. However, quantum states are fragile and sensitive, extremely susceptible to ambient noise, and building a quantum computer with enough qubits and high control fidelity in an actual physical system is a great challenge.

In 2012, physicist John Preskill, a professor at the California Institute of Technology, proposed that two more milestones should be set before the long-term goal of universal quantum computing can be achieved, the first of which is the superiority of quantum computing.

The specific tasks that scientists initially use to demonstrate the superiority of quantum computing must be well-designed tasks that are well-suited to the computing potential of quantum computing devices. This task does not necessarily have practical value, but is mainly used to confirm the huge potential of quantum computing, while at the same time, in technology and theory, can pave the way for subsequent development.

Scientists are using the characteristics and advantages of different systems to carry out quantum computing research based on a variety of physical systems and approaches. Among them, superconducting quantum computing as a solid-state quantum computing scheme has many advantages such as good scalability, long qubit coherence time, fast operation speed, high fidelity, and mature processing technology, while optical systems have the advantages of easy operation of photons, small decoherence, operation at room temperature and can be used for long-distance communication, so they are all physical implementation platforms that have attracted much attention in the field of quantum information.

At this stage, the most likely problems to demonstrate the superiority of quantum computing include random quantum line sampling, Bose sampling, IQP lines, etc. Among them, the random line sampling task is very suitable for completion on a two-dimensional structure superconducting quantum computing chip.

Bose sampling and its "variant" Gaussian Bose sampling task, particularly suitable for optical systems. In fact, bose sampling experiments are an extremely challenging task, placing demanding requirements on photon sources, optical interferometers, and single-photon detectors.

Several major breakthroughs have made quantum computing faster and stronger

"'Chapter 2.0' has significantly improved the scale and complexity of computation over 'Chapter Nine', greatly improving the previous quantum advantage." Professor Chaoyang of Zhongke Continental said that compared with the "Nine Chapters", the "Nine Chapters 2.0" has focused on achieving three major breakthroughs.

First, the overall system efficiency of the "Nine Chapters" is low, about 30%, and one of the main losses comes from the light source. Inspired by the laser principle, the researchers developed a stimulated compression light source that can obtain a compression light source that meets both high compression, high purity and high collection efficiency.

Secondly, Gaussian Bose sampling has potential practical application value in many fields, which can be applied to quantum chemistry, machine learning, graph optimization, and preparation of quantum error correction codes. However, under the current technical conditions, there are still huge challenges in preparing programmable, low-loss, large-scale optical interferometers. In the Gaussian Bose sampling problem, the transformation matrix performing the operation is not only related to the interferometer, but also related to the compression parameters and phase of the compressed light. By controlling the phase of the light source, "Chapter 9 2.0" has some programmability. Phase-adjustable Gaussian boson sampling already has some potential application capabilities, and if the interferometer can be adjusted in the future, it will be useful in many practical fields. In addition, the scale of the interferometer of "Nine Chapters 2.0" has also increased from the previous 100 mode to the 144 mode.

In the end, "Nine Chapters 2.0" implements partial programmable Gaussian Bose sampling in 113 photons and 144 modes, which greatly increases the quantum superiority of the Gaussian Bose sampling problem from 1014 times of the classical supercomputing "Light of Taihu Lake" to 1024 times. At the same time, the dimensionality of the output state space of "Nine Chapters 2.0" reaches the order of 1043, which greatly increases the complexity of the problem and makes it more difficult to be simulated by the new classical algorithm.

In superconducting quantum systems, it is extremely difficult to build large-scale arrays of qubits and achieve coherent manipulation of each qubit with extremely high precision.

Zu Chong Zhi 2.0 achieves quantum computing superiority in random line sampling tasks by subtly regulating the 56 qubits on it. This is the largest published superconducting quantum system with the largest number of qubits, higher than the 62 qubits of the previous "Zu ChongZhi" and the 53 qubits of Google's "planewood" in 2019. Its important upgrade is first of all the introduction of an adjustable coupler, which greatly improves the single-bit gate fidelity and two-bit gate fidelity of the processor; secondly, the reverse solder packaging technology is used to solve the wiring problem on the two-dimensional arrangement quantum chip and greatly reduce signal crosstalk.

After upgrading, the comprehensive computing performance of the entire processor has reached the threshold of demonstrating quantum superiority. T1 life is an important indicator of qubit decoherence, and longer T1 life means that more coherence can be performed on qubits to complete more complex computational tasks. All components on the Zu Chong Zhi 2.0 chip work normally, and the average T1 life of the 66 bits reaches 31 microseconds, which is higher than the 16 microseconds of the "plane wood".

Five candidates are competing

A hot question around quantum computing is which technology path will ultimately win the race. Currently, there are five well-proven candidates competing: superconductivity, ion traps, light quanta, semiconductor quantum dots, and cold atoms. All of these schemes were developed in pioneering physical experiments and implementations in the 1990s.

The superconducting quantum computer solution is currently the fastest-advancing solution in the world, with the largest number of technology followers, IBM and Google with their deep technology accumulation and strong financial strength in the field of rapid development. Compared with foreign countries, in China's progress in various routes of quantum computing, the experiment of superconducting quantum computing started late, but it performed strongly. In the long run, this technology route is easier to achieve scale in the future.

The advantages of the ion trap technology route are good coherence, a large number of entangled qubits, and a high degree of logic gate fidelity. Ion trap systems are one of the two most funded quantum computing research directions by the U.S. government, the other being superconducting systems. In addition to quantum computers, it is also widely used in quantum simulation research in the fields of quantum chemistry, relativistic quantum mechanics, and quantum thermodynamics. Ion trap quantum computing has been developed for more than 20 years, which is comparable to the development of superconducting quantum computing. Internationally, Honeywell, IonQ and AQT have made rapid progress in the commercialization of ion trap quantum computers. However, the domestic experimental research on ion trap quantum computers has only been less than a decade.

China is at the international leading level in the research of optical quantum computing. Optical quantum is in addition to superconducting quantum and ion trap research progress of the rapid progress of the technical route, internationally, Xanadu and PsiQuantum are two well-developed optical quantum computer manufacturers.

Since the semiconductor quantum dot computer combines the current semiconductor industry technology, the future can be quickly industrialized, and because the semiconductor qubit volume is small, it is easier to achieve chipization than the superconducting technology route and the optical quantum technology route. However, the number of current semiconductor qubits is small and the coherence is weak. Internationally, Intel in the United States, Delft University of Technology and Qutech in the Netherlands, SQC in Australia, and RIKEN in Japan are engaged in the research and development of silicon spin qubits.

It is gratifying that the team of academician Guo Guangcan of the University of Science and Technology of China has made important progress in the research of silicon-based semiconductor germanium nanowire quantum chips. Genyuan Quantum, led by the team Professor Guo Guoping, has launched the second generation of silicon-based spin two-bit quantum chips, the Xuanwei XWS2-200.

The cold atom technology route has clear advantages in conducting quantum simulations. Internationally, the French PASQA research team began building programmable quantum simulators made of neutral atomic arrays in 2011. Although China has a layout in this regard, the overall number of participating units is small, and the research time is relatively short.

Theoretical research has proved that for some tasks, quantum computing can complete tasks faster and more efficiently than classical algorithms. The current general consensus in the physics community is that quantum computers cannot completely replace classical computers, but will replace classical computers on certain difficult problems.

Reporter Wu Changfeng

Source: Science and Technology Daily