🈳 👩🏻‍🌾 👴🏿 New entrant in quantum computing with unique technology 👨🏿‍💻 🔽 💄

Honeywell will place quantum computers on captured ions in Microsoft's quantum cloud

For many years, scientists have been developing various systems on which quantum algorithms could be executed. Most of them have one or two advantages - ease of handling or the ability to hold their condition longer than others - however, there are no other positive qualities that prevent them from becoming practical solutions for computing. However, in recent years, some companies have figured out how to manipulate a significant number of solid-state qubits - the so-called transmons . Since the technology for producing transmons is similar to the production of conventional chips, many players in the emerging market — including Google, IBM, and Rigetti — have settled on transmons.

However, transmons are not perfect. They require extremely low temperatures, low variability from device to device, and they hold their condition well, but not perfectly. Many experts in this field believe that another technology still has a chance to surpass transmons.

And now, a company new to the quantum computing market is also betting on this. Honeywell, a company better known as a supplier to the Department of Defense, announced the creation of a quantum computer based on the alternative technology of "ion traps", and this year will give access to its computers through the Microsoft Azure cloud service. The company also claims that, according to some estimates, this is the most powerful quantum computer created to date, however, this statement should be taken very carefully.

Trapped

Transition qubits work when current is circulated in a loop of a superconducting wire connected to a resonator, which makes it possible to control and read the current state. However, since the superconducting wire and the resonator must be manufactured in production, this can lead to the appearance of small differences between the individual qubits. In addition, all this iron must be kept at an extremely low temperature, only slightly above absolute zero, in order to keep these relatively large objects within their main quantum state.

Caught ions provide an opportunity to get around some of these problems. The qubit itself is formed from a small number of atoms - in the case of Honeywell, from two. Honeywell's president of quantum solutions, Tony Attley, emphasizes that this fact eliminates production problems, since each device has the same properties as the atom used (in this case, ytterbium). “Each qubit begins with an ideal,” Attley told us. “Any mistake made comes from the surrounding infrastructure.”

Thanks to Honeywell’s experience in the manufacture and integration of this infrastructure, company engineers are ideally placed to minimize this noise. In addition, such small clusters of atoms can be cooled using lasers. Although the ambient temperature must be kept very low, it does not have to reach the extreme temperatures necessary for the transmon.

In the case of Honeywell, ytterbium atoms were not so easy to cool with lasers, so the company added a couple of barium atoms to the system and cooled them with a laser. A cluster of four ions is easy to cool and control, and it is enough to keep the environment at a temperature of 12 K. Although liquid helium is required for this, it does not require the sophisticated liquefaction cooling equipment that is required for iron from Google and IBM.

Since the ions are charged, they can be moved within the device by simply changing the local electric fields using 200 electrodes built into the device. The state of electrons in ions can be controlled using lasers with a certain wavelength, capable of placing electrons in a superposition with potential energy states. Entanglement and various logical operations can be carried out simply by moving two ions closer to each other, and using lasers working with both of them simultaneously. Reading is performed by stimulation of ions by another laser, forcing the ions to emit a photon, from which it is possible to judge their state.

Honeywell device

A device from Honeywell can be thought of as a line of individual devices. Ions come from one end, which then move along successive sections, where they can be delayed for storage or irradiated with lasers that manipulate qubits. Logical operations (the quantum equivalent of AND and NOT) can be performed simply by placing two ions side by side and performing an operation on them simultaneously. In addition, clusters of four ions (two ytterbium, two barium) can be divided into two parts, or two clusters of two ions can be combined.

The device Honeywell talks about today aligns four qubits on the same line of these storage and handling steps. However, the diagram of the device also shows two additional lines of the stages of storage and handling, going from both sides of the line used in the initial experiments. This is consistent with what Attley said: Honeywell believes that the device can be scaled quickly and that additional qubits can be added annually without changing the architecture at a fundamental level. So, although four qubits is not enough in comparison with what was obtained on devices with transmons, the company believes that it can quickly bridge this gap.

A lot of lasers are required to control the state of qubits

One of the interesting aspects of such a scheme, which, according to Attley, is lacking in other existing commercial systems, is the ability to measure qubits individually, without disturbing any other parts of the system. (Technically, this is carried out using the operation with the fantastic name CNOT - quantum teleportation fan). It allows the computer to perform the equivalent of IF branching, changing the algorithm based on the results of measuring a single qubit. After measuring, the qubit can also be restored to its original state and reused for further calculations.

The individual components of the system behave perfectly. One potential problem is “state preparation and measurement errors”, which has been abbreviated to SPAM [state preparation and measurement errors]. In this case, researchers from Honeywell found that SPAM is dominated by measurement errors, but they occur in less than 1% of cases. For single-qubit gates, errors occur an order of magnitude less often, for two-qubit gates, at a similar level. And all this is much lower than typical indicators of transmon.

About that performance indicator

Honeywell presents this as the “most powerful quantum computer in the world”, however, the correctness of the statement is highly dependent on the speed measurement schemes used. Honeywell uses a measure defined by IBM and called it "quantum volume." We will quote part of the analysis of the quantum volume made by Chris Lee, since he describes his connection with the computer from Honeywell well:

Since quantum gates can always give an error, there is a maximum number of operations that can be performed before it is unreasonable to consider the qubit state to be true. This amount, multiplied by the number of qubits, gives us the depth of the circuit. If used honestly, it pretty accurately describes what a quantum computer is capable of.

The problem with depth is that it is possible to keep the total number of qubits constant (and small), reducing the percentage of errors to very small values. As a result, you get a huge depth, but at the same time it turns out to carry out only calculations that fit into the number of qubits that you have. A two-qubit quantum computer of great depth will still be useless.

It turns out that the purpose of the assessment is to express the computational capabilities of the indicator, which includes the number of qubits and the depth of the circuit. For a specific volume of the algorithm and the problem, this will be the minimum number of qubits needed for the calculations. And depending on the connection of the qubits with each other, for the implementation of the algorithm, a certain number of operations will be required. Researchers express this number by comparing the maximum number of qubits involved in the calculations with the depth of the circuit, and squaring the minimum of these two indicators. So the maximum possible quantum volume will simply be the number of qubits squared.

As noted above, Honeywell reports an extremely low error rate, which means that every calculation running on four qubits of its computer will most likely not contain errors. And since ions can be moved inside the device at will, they can be arbitrarily connected to each other. It turns out that the quantum volume is equal to the number of qubits squared. This differs from the performance of the equipment used by Google and IBM, where 10 times more qubits, but much more errors, and qubits can be connected only with a small number of neighboring ones.

As a result, in order for a machine from Honeywell to catch up with the machines of its rivals in terms of quantum volume, it does not have to add too many additional qubits. The structure of the machine that it describes today definitely allows you to add qubits to it. As a result, the company claims a quantum volume of 64, which means eight qubits, and there is every reason to believe it.

However, if IBM has already introduced a computer containing almost 64 real qubits, and Google should soon follow suit, will the weather only make eight qubits? The answer, as usual, is ambiguous. Some algorithms will be highly dependent on the connectivity of qubits. And although they can be launched on larger machines with less connectivity, this will require that a larger number of qubits serve as connecting links, organizing equivalent connectivity, and each of them is potentially capable of introducing an error into the calculations. The high connectivity of the Honeywell machine can offset the need for additional operations, and operations are still not the main source of errors. Ion trap close-up; various zones for storing and manipulating qubits are visible

And there is also the problem of scaling. Attley said that the company should be able to increase its quantum volume by an order of magnitude annually over the next five years, which would require adding 3-4 qubits per year. This means that even after five years, the computer will have about 30 qubits - half the current performance of competitors. Meanwhile, Google and IBM are working to reduce errors and add a couple dozen qubits to their machines every few years.

If the plans of all companies come true, in a few years the situation will become very interesting. Honeywell will have a significant advantage in quantum volume, and its competitors with iron on transmons will have an order of magnitude more qubits. Meanwhile, teams using transmons intend to create quantum computers with error correction, which will require thousands of qubits - which means that researchers expect that from some point they will learn to add hundreds of qubits with each new generation of chips.

Since it is unclear when, according to companies, such a growth of qubits should begin, it is unclear how the Honeywell exit can change the competitive landscape.

What will we have so far

Honeywell, a company whose divisions do everything from sporting goods to contracting with the military, is definitely an unusual competitor in a market dominated by a mix of startups and traditional computing companies. However, the company told a consistent story of its entry into the market: as part of its manufacturing work, Honeywell has developed many computer components on ion traps - such as photonics, cryogenics, vacuum systems - for other purposes. And a group of scientists from the company said that the potential of this area is large enough to be worth trying to develop. And since Honeywell is a large company, it was able to recruit a fairly strong group of people passionate about the development of this project.

Like other companies in this field, Honeywell determined that most companies do not want to create their own infrastructure in which their liquid helium cooling system could operate. Therefore, Honeywell is going to provide access to its quantum computers through the cloud. She also agreed with Microsoft so that the system can be accessed through its Azure service.

To write the programs that are used in the current work, Honeywell researchers adapted Qiskitfrom IBM, an open source tool that allows you to describe quantum algorithms in a form that is not tied to a specific hardware, and then issue real commands to run the program on a specific hardware (something like a cross-platform compiler). The company thus hopes to take advantage of existing expert solutions. It may also mean that companies will be able to develop a set of quantum algorithms, and then run them on any systems with the properties they need - high connectivity or a large number of qubits - to achieve the necessary performance.

Friends and competitors

Perhaps due to the introduction of a completely new architecture, the company combined the announcement of its technology with two contributions to companies already developing quantum algorithms. She also announced that financial giant JPMorgan Chase will work with Honeywell to explore the possibilities of using her system to develop financial algorithms. This does not mean that the system is completely ready for use; We already talked with people from the JPMorgan Chase, and they said that they were trying to guarantee that the company was fully ready for practical quantum computers.

All this testifies in favor of the fact that Honeywell takes its developments seriously and hopes to become one of the main rivals in the space of quantum computing. And if her predictions for the future come true, then so it is likely to be.

The observer may be tempted to compare the situation with the rivalry of traditional computer architectures, where x86 and ARM are actively fighting today. However, these different architectures are made using the same methods and work with the same components. In the described case, two competing architectures are based on physically completely different systems, in which only some work rules coincide. This is a completely different set of conditions, and much more interesting.