IBM launched the world's largest refrigerator, may disrupt the quantum computing industry landscape

Last year IBM announced their "Goldeneye project": an unprecedented mega-dilute cooler (commonly known as a refrigerator, scientists also call it a vat) for quantum computers - containing an experimental volume of 1.7 cubic meters. can cool a volume larger than three home kitchen refrigerators to temperatures colder than outer space, compared to previous coolers in the 0.4-0.7 cubic meter range. This is the largest experimental volume of dilution refrigerator in the world.

Today, IBM finally announced that it has successfully cooled the GoldenEye to operating temperature (~25 mK) and connected a quantum processor inside [1]. The Golden Eye will soon be transferred to the IBM Quantum Computing Center in New York, which is exploring large cryogenic systems to best meet the cooling needs of future quantum data centers, such as the Bluefors Kide platform being developed for IBM System Two. ease of use, will inspire the next generation of vacuum and cryogenic cooling technologies."

Inside the "Golden Eye" dilution chiller

01How to build the world's largest and coldest refrigerator?

A dilution chiller is an experimental cryogenic device that uses a mixture of two helium isotopes (helium-3 and helium-4) to cool the volume of space to the milli-kelvin (mK) range, or a few thousandths of a degree above absolute zero (-273.15°C).

Dilution chillers perform the cooling by removing heat from the helium isotope mixture using multiple steps, and then use a vacuum pump to circulate and dilute the helium-3 into the helium-3/helium-4 mixture until the target temperature is reached. In the early stages, all dilution chillers were "wet" systems, requiring already cold substances like liquid nitrogen and other cryogenic liquids to start the cooling process. Today's chillers are more commonly known as "dry", i.e., liquid helium-free dilution chillers: a mechanical component called a cryocooler is used to provide an initial temperature of 50 K and 4 K to pre-cool the helium mixture.

These are advanced machines, so they are difficult to scale up. "The GoldenEye project uses a new frame and cryostat structure - the main barrel component responsible for cooling - to maximize the experimental volume while reducing noise and to achieve the temperatures needed to cool the experimental quantum hardware. The design is modular, which makes prototyping, assembly and disassembly easier for a team of four IBM engineers. Other large dilution coolers may require larger cranes and a dozen or more technicians to assemble and disassemble.

Cryostats also feature a clamshell design that allows the external vacuum chamber to open sideways, giving access to the internal hardware without removing the entire housing. Most dilution chillers today require a team of operators to work properly, but GoldenEye's fully automated system includes a specially designed cantilever crane that will allow even one person to operate the refrigerator in the future - this can be monitored remotely with the help of an open source visualization platform.

Panoramic view of the "GoldenEye" dilution chiller

The interior of the cryostat is capable of housing a set of 10 internal plates for mounting components in its upper and lower halves: five "conventional" units at the top and five inverted units at the bottom. It can also accommodate up to six individual dilution chiller units, capable of achieving cooling power of nearly 10 mW at 100 mK and over 24 W at 4 K. Finally, the weight of the entire system (6.7 metric tons) also helps suppress vibration, reducing the need for other commonly used suppression technologies.

Most importantly, IBM says, it works. After just three years from the start of the project to the current 25 mK milestone, it was possible to put a quantum bit chip into it.

This test completed measurements of quantum bit frequencies and coherence times, demonstrating the performance of the GoldenEye by looking at a single quantum bit. The coherence time of 450 microseconds is equivalent to that measured on other commercial dilution cooling systems. No degradation in quantum bit performance was also observed, despite the different internal environment and larger experimental volume.

IBM says that despite its size, it is very efficient. Accommodating an equivalent amount of quantum hardware, it requires less space than today's large dilution chillers. Deploying equivalent hardware in today's state-of-the-art chillers requires 10 times more lab space.

IBM also said, "We don't know if the coolers used to cool future quantum computers will really be that big. But GoldenEye could allow us to consider many different ways to scale our quantum processors even beyond 2025, and will help us further conceptualize the cryogenic infrastructure for future quantum data centers. The bottom line is that every time we push a technological boundary, we learn something new. By simply building the GoldenEye, our team tested the limits of cryogenic science while showing us what kind of large-scale quantum experiments might soon be feasible. As we continue to map the quantum roadmap for the world, the lessons learned from this project may be used for future IBM Quantum System 3."

02What's the takeaway?

When I heard the news today, my first reaction was shock and my second reaction was concern. Can we build such a large dilution cooler? As far as I can see, the answer is no. 25 mK small dilution chillers, the domestic still can not do the degree of commercialization (mainly through Bluefors, Oxford Instruments and other companies imported), let alone such a huge chiller.

Uncle Box suggests that all quantum computing industry insiders should pay attention to this event. Uncle Box also envisions several questions, the answers to which may even turn the quantum computing industry landscape upside down.

First, does quantum bit scaling really require such a large refrigerator? Is there an alternative method of scaling that does not require such a large refrigerator? Or is there a new way of cooling (which might require a new kind of physics)?

Second, do other companies have the ability to build such a large refrigerator? If not, can they be purchased from IBM? For domestic companies, will they be banned from purchasing them, like the AMSL lithography?

Third, do millions of quantum bits require a larger refrigerator? The Golden Eye is planned to hold 4000 quantum bits, so how many times will the refrigerator to hold millions of quantum bits be expanded in the future? This actually runs counter to the trend of miniaturization of computers.

Fourth, if we cannot solve the problem of dilute refrigerators getting bigger and bigger, should we develop room temperature quantum computers? For example, in optical quantum computers, the computational process is performed at room temperature except for the single photon detectors which currently require refrigeration.

These are all open questions that need to be explored by the quantum computing industry.