After decades of hard work without much hope of success, a hectic activity suddenly developed around quantum computing. Almost two years ago, IBM showed the world a quantum computer with five quantum bits (qubits), which they now (which sounds a little strange) call the IBM Q Experience. Back then, the device was more like a toy for researchers than a tool for serious data processing. However, 70 thousand users all over the world have registered with the project, and by now the number of qubits has quadrupled. A few months ago, IBM and Intel announced the creation of quantum computers with 50 and 49 qubits. It is also known that another computer is waiting in the wings within the walls of Google. "The community is full of energy and the recent breakthroughs are amazing."- says physicist Jens Eisert from the Free University of Berlin.
Currently, there is talk of impending "quantum supremacy": the time when a quantum computer can perform a task beyond the power of even the most powerful classical supercomputers. If we compare only numbers, then such a statement may seem ridiculous: 50 qubits versus billions of classical bits in any laptop. But the whole point of quantum computing is that the quantum bit is capable of much more than the classical one. For a long time, it was believed that 50 qubits would be enough to perform calculations that a conventional computer would perform indefinitely. In mid-2017, researchers at Google announced they were going to demonstrate quantum superiority by December. (To a recent request for new data, a company spokesperson replied: “We will announce results,as soon as they are sufficiently substantiated, but for now a thorough analysis of the existing developments is being carried out. ")
I'd like to conclude that all the main problems can be solved and the future, in which quantum computers are a ubiquitous phenomenon, is just a matter of technical equipment. But he will be wrong. The physical issues at the heart of quantum computing are still far from being solved.
Even if we soon step into an era of quantum supremacy, the next year or two could be decisive - will quantum computers really completely change the way we do computing? The stakes are still high and there is no guarantee that the target will be met.
Shut up and calculate
Both the benefits and challenges of quantum computing are integral to the physics that make it possible. The basics have already been said more than once, although it has not always been clarified what quantum mechanics requires. Classic computers store information and process it in binary code (0 or 1). In quantum computers, the situation is almost the same, only each bit is in the so-called superposition, that is, it can be both 0 and 1 at the same time. This means that the state of a qubit can only be determined with a certain degree of probability.
To perform a computation with a large number of qubits, all of them must be in interdependent superpositions - in a state of "quantum coherence", in which all qubits are considered entangled. In this case, the slightest change in one qubit can affect all the others. That is, computational operations using qubits have a higher performance than using classical bits. In a classical device, computational capabilities are simply dependent on the number of bits, but the addition of each new qubit increases the capabilities of a quantum computer by 2 times. This is why the difference between a 5-qubit device and a 50-qubit device is so significant.
Promotional video:
Note, I did not say, as is often done, that the advantage of a quantum computer over a classical one lies in the existence of superpositions, which greatly increase the number of possible states of encoded information. As I did not say, entanglement allows many calculations to be performed simultaneously. (Actually, a high degree of entanglement of qubits is not a prerequisite.) There is some truth in this, but none of the statements describe the essence of quantum computation.
Because of the complexity of understanding quantum mechanics, explaining why quantum computation is so powerful is a daunting task. The equations of quantum theory definitely show that it will work - at least with some kinds of computation: factoring or searching a database speeds up the process enormously. But how much exactly?
Perhaps the safest way to describe quantum computing is to say that quantum mechanics in some way creates "possibilities" for computation that are not available to classical devices. As physicist Daniel Gottesman of the Perimeter Institute for Theoretical Physics (Perimeter Institute) in Waterloo noted: "If enough quantum mechanics is available, then in a sense the process is accelerating, and if not, it is not."
Although some points are still clear. Quantum computing requires all qubits to be coherent, which is extremely difficult to implement. The interaction of the system of coherent qubits with the environment creates channels through which coherence quickly "leaks". This process is called decoherence. Scientists planning to build a quantum computer must prevent decoherence. Now they only manage to stop her for a split second. The situation becomes more complicated when the number of qubits, and, accordingly, the ability to interact with the environment increases. That is why, although the idea of quantum computers was first proposed by Richard Feynman back in 1982, and the theory was developed in the early 1990s, devices capable of performing real computation have only now been created.
Quantum errors
There is a second major reason why building a quantum computer is so difficult. Like any other process in the world, it makes noise. Random fluctuations, arising, say, due to the temperature of the qubits or due to the peculiarities of fundamental quantum mechanical processes, can change the direction or state of the qubit, which leads to inaccurate calculations. Such a threat exists in working with classical computers, but it is quite easy to solve. You just need to create two or more backups of each bit so that an accidentally flipped bit is not counted.
Scientists working on the creation of a quantum computer have developed several ways to solve the problem, but all strategies lead to the appearance of too many additional computational costs, since all the computing power is spent on correcting errors, and not on executing the given algorithms. “The current error rate significantly limits the length of time a computation can take,” explains Andrew Childs, co-director of the Joint Center for Quantum Information and Computational Sciences at the University of Maryland. "We need to significantly improve the results if we want to create something interesting."
Much research in fundamental quantum computing focuses on error correction techniques. Part of the complexity of the problem stems from another of the key properties of quantum systems: superpositions can only be maintained as long as you don't measure the value of a qubit. The measurement will destroy the superposition and lead to a certain value: 1 or 0. How can you tell if there was an error in the operation of a qubit if you do not know what state it was in?
One clever scheme suggests using indirect computation by combining a qubit with a second auxiliary qubit. The latter is not involved in the calculation, so its measurement does not affect the state of the main qubit. But it is rather difficult to implement it. This solution means that many physical qubits are needed to create a true "logical qubit" that is immune to errors.
How many? Quantum theorist Alan Aspuru-Guzik from Harvard University believes that it will take about ten thousand physical qubits to create one logical qubit, which is not currently possible. According to him, if all goes well, this number will decrease to several thousand or even hundreds. Aisert is not so pessimistic and believes that about eight hundred physical qubits will be enough, but admits that even in this situation, "the additional costs of computing power will still be great." You need to find a way to deal with mistakes.
There is an alternative to bug fixing. They can be avoided or prevented from occurring in what is called error mitigation. Researchers at IBM design circuits to mathematically calculate the likelihood of an error and then take that result as zero noise.
Some researchers believe that the problem of error correction will remain unresolved and prevent quantum computers from reaching their predicted heights. “Creating quantum error-correcting codes is much more difficult than demonstrating quantum superiority,” explains Hebrew University in Israel mathematician Gil Kalai. He also adds that "non-error correcting devices are very primitive in their calculations, and superiority cannot be based on primitiveness." In other words, quantum computers will not outperform classical computers if errors are not eliminated.
Other scientists believe the problem will eventually be solved. One of them is Jay Gambetta, a quantum computer scientist at the IBM Center for Quantum Computing. Thomas J. Watson. “Our recent experiments have demonstrated the basic elements of error correction in small devices, which in turn paves the way for larger devices that can reliably store quantum information for extended periods of time in the presence of noise,” he says. However, Gambetta also admits that, even with the current state of affairs, "there is still a long way to go to the creation of a universal, error-resistant quantum computer using logic qubits." Thanks to such research, Childs is optimistic. “I'm sure we'll see a demonstration of even more successful [bug fixing] experiments, but,it will likely take a long time before we start using quantum computers for real computing.”
Living with mistakes
In the near future, quantum computers will malfunction. The question arises: how to live with it? IBM scientists say that for the foreseeable future, the field of "approximate quantum computing" research will focus on finding ways to adapt to noise.
This requires the creation of such algorithms that will produce the correct result, ignoring errors. The process can be compared to the counting of election results, which does not take into account spoiled ballots. “Even if it does make some mistakes, a large enough high-quality quantum computation should be more efficient than [classical],” says Gambetta.
One of the more recent error-tolerant applications of the technology appears to be of greater value to scientists than to the world at large: modeling materials at the atomic level. (In fact, this was the motivation that led Feynman to come up with the idea of quantum computers.) The equations of quantum mechanics describe how stability or chemical reactivity is calculated (for example, in drug molecules). But these equations cannot be solved without using a lot of simplifications.
However, according to Childs, the quantum behavior of electrons and atoms "is relatively close to the natural behavior of a quantum computer." This means that an accurate computer model of the molecule could be built. “Many members of the scientific community, including myself, believe that the first successful application of a quantum computer will be associated with quantum chemistry and materials science,” says Aspuru-Guzik: he was one of the first who began to push quantum computing in this direction.
Quantum modeling is proving useful even on the smallest quantum computers available to us today. A team of researchers, which includes Aspuru-Guzik, developed an algorithm that they called the "Variational method for solving problems in quantum mechanics" (hereinafter - VMR). This algorithm allows you to find the least energy-consuming state of a molecule, even in noisy qubits. At the moment, it can only handle very small molecules with few electrons. Classic computers do this task well. But the power of quantum is constantly growing, as Gambetta and colleagues showed last September when they used a six-kbit device to calculate the electronic structure of molecules like lithium hydride and beryllium hydride. The work was "a significant breakthrough for the quantum sciences"as the chemical physicist Markus Reicher of the Swiss Higher Technical School of Zurich put it. “Using BMP to model small molecules is a great example of how short-term heuristic algorithms can be applied,” says Gambetta.
But, according to Aspuru-Guzik, logic qubits capable of correcting errors will be required even before quantum computers overtake classical ones. “I can't wait until error-correcting quantum computing becomes a reality,” he commented.
“If we had more than two hundred qubits, we could do really innovative things,” added Reicher. "And with 5,000 qubits, a quantum computer could have a major impact on science."
What is your volume?
These goals are incredibly difficult to achieve. Despite all the difficulties, quantum computers from five-qubit to 50-bit in just a year - this fact gives hope. However, don't get too hung up on these numbers, because they only tell a small part of the story. Now, it's not how many qubits you have more important, but how well they work and how efficient the algorithms you have developed are.
Any quantum computation ends with decoherence, which shuffles the qubits. Typically, the decoherence time for a group of qubits is several microseconds. The number of logical operations that can be performed in such a short time depends on the switching speed of the quantum gate. If the speed is too low, it doesn't matter how many qubits you have at your disposal. The number of operations required for a given computation is called computation depth: low-depth algorithms are more efficient than deep algorithms. However, it is not known for certain whether they are useful in calculations.
Moreover, not all qubits are equally noisy. It is theoretically possible to create low-noise qubits from materials that are in the so-called "topological electronic state": if particles in this state are used to encode binary information, it will be protected from random noise. In an attempt to find particles in a topological state, researchers at Microsoft are primarily studying exotic quantum materials. However, there is no guarantee that their research will be successful.
To denote the power of quantum computing on a particular device, researchers at IBM coined the term "quantum volume." This is a number that unites all important factors: the depth of the algorithm, the number and connectivity of qubits, as well as other indicators of the quality of quantum gates (for example, noise). In general, this "quantum volume" characterizes the power of quantum computing. According to Gambetta, it is now necessary to develop quantum computing equipment that will increase the available quantum volume.
This is one of the reasons why the vaunted quantum supremacy is a rather vague idea. The idea that a 50-qubit quantum computer will outperform modern supercomputers sounds attractive, but many unresolved questions remain. When solving exactly what problems does a quantum computer outperform supercomputers? How can one determine if a quantum computer has received the correct answer if it cannot be verified with a classical device? What if a classical computer is more efficient than a quantum computer if a better algorithm is found?
Thus, quantum supremacy is a concept that requires caution. Some researchers prefer to talk about the "quantum advantage", about a leap in the development of quantum technologies, rather than about the final victory of quantum computers over ordinary ones. Moreover, the majority try not to use the word "superiority" as it contains negative political and racist connotations.
Regardless of the name, if scientists can demonstrate that quantum computers can perform tasks that classical devices cannot do, then this will be an extremely important psychological moment for this field. “The demonstration of an undeniable quantum advantage will go down in history. This will prove that quantum computers can really expand our technological capabilities,”says Aizert.
Perhaps this will be a symbolic event, rather than a radical change in the field of computing. Nevertheless, this is worth paying attention to. If quantum computers outperform conventional computers, it won't be because IBM and Google suddenly launch them on the market. To achieve quantum supremacy, you need to establish an intricate system of interaction between developers and users. And the latter must be firmly convinced that the novelty is worth trying. In pursuit of this collaboration, IBM and Google are trying to provide users with their developments as quickly as possible. Previously, IBM offered all sign-ups to the site access to its 16-qubit IBM Q computer. Now the company has developed a 20-qubit version for corporate clients including JP Morgan Chase, Daimler, Honda, Samsung and the University of Oxford. Such collaboration not only helps clients find something useful and interesting, but also creates a quantum-literate community of programmers who will develop new functions and solve problems that cannot be solved within the framework of one company.
“For the field of quantum computing to actively develop, you need to give people the opportunity to use and study quantum computers, - says Gambetta. "The entire scientific and industrial world now needs to focus on one task - preparing for the era of quantum computers."
Project translation New
Philip Ball
