中文 English

Progress In Quantum Computing

Commercial viability appears complicated, but feasible; timing still unknown.

popularity

A recent wave of quantum computing investment has given rise to claims of a quantum computing bubble, based on overly optimistic technological claims in a field area that experts say has yet to demonstrate any real utility. But executives on the industry’s front lines say quantum computing is indeed a commercially viable technology, albeit one that is at least several years away from overcoming extremely challenging technical hurdles.

The entire tech world will change in just a few years if those more optimistic estimates are correct. Scientists are counting on quantum computing’s potential to solve highly complex problems, like cracking unbreakable codes and creating life-saving medicines. Last month, search engine company Baidu reported it had developed a quantum computer, joining the ranks of quantum developers like Google, IBM, and Intel. The past few years saw investors pouring money into the industry at a breakneck rate. According to a McKinsey report, funding for quantum computing startups jumped to $1.4 billion last year, up from $700 million the year before.

However, the McKinsey report also says that momentum may be slowing, in part because the demand for quantum computing experts is outpacing the supply of qualified talent. Others warn that the industry is over-hyped, and predictions of near-term quantum technology being able to give rise to more advanced quantum computers are without basis. Naysayers also point to a paper published last month in which researchers used traditional computing methods to challenge Google’s claims of reaching quantum supremacy in 2019.

Quantum insiders say the truth is a complicated mixture of all those data points. For one thing, the public’s focus on the race for quantum supremacy— creating a computer that can outperform its less advanced counterparts— may be missing the point. “A headline that includes some variation of ‘Quantum Supremacy Achieved’ is almost irresistible to print, but it will inevitably mislead the general public,” IBM researchers wrote in 2019. “Quantum computers will never reign ‘supreme’ over classical computers, but will rather work in concert with them, since each have their unique strengths.”

IBM’s most recent development roadmap predicts the company will scale its quantum technology to more than 1,000 qubits and operate parallelized quantum processors next year. The following year introduces error suppression and mitigation techniques, which IBM says lays the groundwork for quantum error correction: one of the major keys to unlocking quantum computing’s full potential.

Focusing on an arrival date for quantum computing could even be dangerous, both from the standpoint of missed business opportunities and potential security threats. Michael Osborne, who leads security and privacy activities at IBM’s research center in Switzerland, says the public “cannot afford to risk not being ready” for quantum advances.

“You can’t wait until such a machine is here, you have to be preparing way ahead for when these things advance,” said Osborne. “For organizations where quantum computers are going to have an impact, whether good or bad, that journey starts now because we can’t predict when there’s going to be a major breakthrough in certain forms of hybrid memory, which is really going to advance how quickly large machines are available.”

Furthermore, these executives say, it will take years for researchers to tackle multiple challenges that will allow quantum technology to become commercially viable. The industry has yet to reach consensus on even basic questions about quantum computing, such as which materials are best suited for quantum chips. Imec, for example, relies on silicon in its CMOS-compatible fabrication technique, which limits defects that can cause qubits to lose energy. Researchers at the U.S. Department of Energy-affiliated Fermilab, on the other hand, say silicon decreases the lifespan of qubits through a process known as quantum decoherence, and thus may be less appropriate for quantum chips than sapphire or another material.

“When you compare the current state of quantum computing to the computing we know — to the standard digital computing — people very often say we are roughly in the 1940s, perhaps 1950s of the last century,” said Sebastian Luber, senior director of technology and innovation at Infineon Technologies. “Beyond the proof of concept, beyond the first examples, the current quantum computers are not able to do something useful, especially not something relevant for the industry or society. There’s still a long way to go.”

The road to usefulness
At the heart of many of quantum’s technical challenges are qubits, the information unit in a quantum mechanical computing system. The most important development work yet to be done involves increasing the reliability of the qubits and preserving their quantum properties for a longer amount of time.

“The coherence times have to be improved significantly for at least part of the qubits,” said Luber. “We also need much more of them because currently we’re still very much limited. We have a few dozen qubits, but we need hundreds, thousands, or perhaps even more.”

Another critical challenge is error correction, which becomes a bigger issue as more qubits are placed together. “After a few cycles of regulation, you can’t distinguish the result anymore from pure noise. One way out of that is quantum error correction. You bond a lot of physical qubits to logical qubits which are then safe,” he said.

Manufacturing quantum computers presents another set of issues. “These systems are very susceptible to the environment,” Luber said. “Whenever you have changes in the environment, you introduce noise to disturb the system. One way to protect them is to cool them down. There are other approaches as well, and in the future it may be possible to have room temperature quantum computers.”

Even when quantum computing’s science is established, there will be challenges like familiarizing the industry with a computing system that functions in a fundamentally different way than any other computer in history.

Eric Holland, business development manager for Keysight’s Quantum Engineering Solutions Group, said quantum computers won’t look like the standard laptop or desktop so familiar to today’s consumers. “Their appearance varies based on the quantum technology. However, all have familiar elements to an RF engineer — arbitrary waveform generators, digitizers, and digital input/output cards. There are more exotic elements, such as high-powered visible lasers or cryogenics. The resemblance is closer to a hybrid between a server farm and a science project.”

Improving and monitoring quantum systems looks different, as well. “This could be from improvements in the materials stack, design improvements, or improvements through control and the environment,” Holland said. “Measuring these improvements is challenging as an improvement corresponds to a decrease in errors. The smaller the error, the more difficult it is to measure it accurately and in a timely manner. Additionally, since all operations have a certain associated error, it is challenging to accurately represent the system error budget in a robust and reliable manner.”

Case study: GlobalFoundries and PsiQuantum
Manufacturing quantum computers is another matter entirely, and one fraught with unknowns.

“It’s either the curse or the perk of my job,” said Anthony Yu, vice president of silicon photonics product management at GlobalFoundries, about harnessing traditional manufacturing methods and altering them for use in quantum computing fabrication. Last year, GlobalFoundries and its partner PsiQuantum announced they had begun manufacturing the foundational components of a full-scale commercial quantum computer using silicon photonic and electronic chips. “We believe we can control the photons and be able to make them behave with the classic quantum properties of entanglement and superposition, and allow these very complicated algorithms to be performed on these photonic qubits,” Yu said.

The technical capabilities required of the project are both significant and varied. “You need to be operating at very cold temperatures — either milli-kelvin, or some of our customers are able to operate at 4 Kelvin, which is still pretty darn cold,” Yu said. “You also need the ability to do control electronics, even though you’re using quantum states in order to be able to perform the calculations and apply the algorithms. Third, you need very, very fast conductivity. Fourth, which is the art of all this, you need to be able to manufacture these qubits. There are various methods to do this — superconducting technology, quantum dots.”

As with any quantum technology, even the slightest flaw can create an unusable product.

“You have to be careful to create the perfect qubits,” said Yu. “To reach these quantum calculations, the parts basically have to be undisturbed by the environment. You have to do generation of photons, manipulation of photons, and the detection of photons — all in an undisturbed way so you can take advantage of the quantum computing algorithms to solve these very complicated problems.”

There are signs of progress. “We can actually control the photons on the chip, and we’re employing the photonic technology that we have in other commercial areas for things like data communications and data centers,” he said. “We’re using the same types of features, like filters, detectors, switches, and controlling the delay of the photons as they travel. We’re using the same manufacturing techniques and the same manufacturing structures, but you’re doing it at very, very cold temperatures to make sure these are perfectly-controlled photons.”

At the same time, there must be extremely low loss as the photons travel on the chip but the chip where the qubits are controlled is simply a silicon-based chip manufactured in the same type of fab. “We’re using different materials than we would use in normal CMOS processing because we’re dealing with these photons in a very different way.

“We use immersion lithography to create these waveguides, which is where the photons travel,” Yu said. “It gives them extremely smooth sidewalls because the perfect photon will be disturbed when it scatters along the sides of the walls of the waveguides. It’s sort of the same methods used to control that for our logic chips and photon chips for data centers.”

While other parts of the quantum computing system like control electronics tend to be produced at very small nodes, Yu says the quantum chips do not strain Moore’s Law. “We’re doing a lot of stuff at 45nm, which a lot of people would consider trailing technology.”

Of course, not all silicon-based methods transfer over to the field of quantum computing, particularly as it relates to waveguide loss or single photon detection. Yu says that SPD requires photons to travel along a material that “has to be near superconducting.” He would not disclose the material the companies used in this step of the process, but described it as “new” and one that “we wouldn’t ever use in any other conventional CMOS technology.”

Conclusion
The sum of all of these techniques and processes remains promising. “A lot of people in the industry see the world’s first useful quantum computer being available in the 2027 to 2028 time frame,” said Yu. “I see it happening. Otherwise, I wouldn’t be as invested in this process as I am.”



Leave a Reply


(Note: This name will be displayed publicly)