Quantum computers – beautiful promises, but more challenges

At the end of 2019, the world of science held its breath. The American giant Google has announced that its prototype quantum computer has surpassed even the best supercomputers. Some have argued that this is a real milestone known as quantum supremacy – the start of a new era in technology. Others said it was too early to celebrate. How is it really? What are the biggest challenges of quantum computers?

While full-blooded quantum computers will have to wait a little longer, scientists are putting more and more pressure on major companies to create their own designs. Microsoft, IBM, Google – these are just the most “media” brands. Even those who question the importance of Google’s quantum supremacy stress the importance of quantum error correction. in a Science An interesting work appeared on this topic.

There really is a difference between a $100 million quantum computer with 10,000 qubits, a random noise generator, and the most powerful computer in the world.

Chad Rigetti, physicist and co-founder of Rigetti Computing

A quantum computer is like a ship

Experts argue that the main goal of quantum computers should be able to spread information encoded into a single qubit, even when noise disrupts the underlying qubits. A good comparison here is the one used by Scott Aaronson, a computer scientist at the University of Texas at Austin:

We try to build a ship that will remain the same ship, even if every plank of it rots and has to be replaced.

It is an important milestone for the largest companies that manufacture quantum computers – Google, Rigetti Computing and IBM. Hartmut Nevin of quantum AI at Google and Jay Gambetta of IBM argue that in the coming years “we’ll see the results of this work”.

A simple reminder – a classical computer displays data in a series of bits written with the number 0 or 1. A quantum computer in turn uses qubits (quantum bits), which can be 0, 1, or both values ​​at the same time. It’s a bit like being in two places at the same time. Presented as ions, photons, or superconducting small circuits, these double states make quantum computers better than conventional computers. However, at the same time it makes them fragile – even the smallest interaction with the environment can distort the recorded data. To create functional quantum computers, you must learn how to correct these errors. Google experiments and reported quantitative superiority are the first step towards achieving this goal.

How does the noise penetrate?

53 qubits was enough to prove Google’s quantum superiority. To encrypt data in one qubit with sufficient accuracy, you must have about 1000 of them, why?

This goes back to the 1994 Shor algorithm, one of the foundations of quantum cryptography. Peter Shor of the Massachusetts Institute of Technology showed that a hypothetical quantum computer should be able to quickly compute large numbers, and thus perform a possible factorial decomposition of the natural number. The algorithm represents the . Through computing qubits, thanks to their dual nature. The waves interfere with each other in such a way that the wrong analyzes cancel each other out and only the correct ones appear. A machine running according to Shore’s algorithm can, among other things, crack the cryptographic systems that currently secure the Internet connection. In 2001, a group of computer scientists from IBM and Stanford University offered to run Shor’s algorithm on a 7-qubit quantum computer based on nuclear magnetic resonance. Then distribute the number 15.

Shore saw that each qubit would maintain its state, so quantum waves would be able to propagate for as long as necessary. Real qubits are much less stable. Google, IBM, and Rigetti use qubits made of superconducting metallic resonant circuits embedded in microchips that have so far proven easier to control and integrate into circuits than other types of qubits. Each circuit contains two different energy states, which can be 0 or 1. By manipulating the circuit with microwaves, researchers can put it into one or any combination of these states – eg 30% 0 and 70% 1. But these states The argument is blurry. (decoheres) in a split second. Even before this happens, noise can disrupt and alter the state, which can mess with the calculations.

It was this noise that almost drowned out the signal in the Google experience. The researchers arranged 53 qubits to encode all possible outcomes, which ranged from 0 to 253. Then they implemented a set of randomly selected interactions between the qubits, making some outcomes more likely than others. Given the complexity of the interactions, even the most powerful supercomputer could take thousands of years to calculate the formula. Google’s quantum computer did something no other machine on Earth could do. The problem was that the computed formula was slightly distinct from the random results due to qubit noise. This is why many experts say it wasn’t a show by the Mountain View giant, but rather a clever manipulation.

Confusion helps or is everything confusing?

In order for the results obtained in quantum computers to be meaningful, the qubits must be as reliable as the qubits in classical computers. This means that they must be immune to any external disturbance. Unfortunately, it is impossible to translate error-correcting codes from classical computers into quantum computers.

Moreover, quantum mechanics requires scientists to blindly look for errors. Although qubits can be 0 and 1, according to current theories, you can’t measure an exact value without reducing the qubit to 0 or 1. It’s just like Schrödinger’s cat.

The prohibition of cloning, and the claim that a copy of an unknown quantum state cannot be made, was formulated and proven by William Waters and Wojciech Żurek and independently by Dennis Dix in 1982 and is central to quantum mechanics and quantum computing. Although you can’t copy the unknown quantum state of a qubit, you can extend it to other qubits with a phenomenon called quantum entanglement.

Quantum entanglement is very subtle. When exposed to microwaves, the original qubit interacts with the other through a CNOT gate (2 qubits). It changes the state of the second qubit if the first qubit is 1 and leaves it unchanged if the first qubit is 0. Interestingly, this process does not measure or disturb the state of the first qubit at all. Two entangled qubits remain at 0 and 1 simultaneously.

If the original qubit is 30% at 0 and 70% in state 1, you can combine it with other qubits to form a string, for example three qubits in an entangled state where 30% of all three are 0 and 70% of all three are 1. This case is different The three copies of the original qubit. In fact, none of the three qubits entangled in the chain has its own uniquely defined quantum state, but the three qubits are completely interconnected. If we scale the first number and it changes to 1, then the other two should do that as well. This connection is the essence of entanglement. Thanks to quantum entanglement, scientists can easily detect errors in quantum computers by entangling additional qubits, called helper qubits. Although every scaling breaks it down, the basic bits of the encoding remain intact.

When will the quantum future come?

However, scientists face a difficult challenge. Manipulating individual qubits can lead to errors, and if this error rate does not fall below a certain value, then entanglement of subsequent qubits with the original will only increase the noise in the system. Auxiliary qubits and other error-correction mechanisms add more noise, and when these effects are taken into account, the necessary error threshold is lower. For the scheme to work, physicists must reduce the error rate to less than 1%. Recent experiments have resulted in an error rate of 3%. Error correction also requires qubits to be processed multiple times, which makes the whole process more demanding than quantum excel.

At the moment, quantum computers cannot provide definitive answers – instead, the answer they give is the most likely answer, but must be verified by a classical computer. For example, a quantum computer could give the most likely answer to a question about breaking an encryption algorithm, but it would need a classical computer to test the answer to determine if it was actually broken.

It will be decades for quantum computers to be fully functional and error-free, but we’re on our way to that reality.

Want to stay up to date with CHIP? Follow us on Google News