Unlocking Quantum Potential With High-Quality Qubits: How Quantinuum Achieved A Three-Year String Of Record-Breaking Quantum Measurements

Quantinuum recently announced that its first-generation System Model H1 trapped ion quantum computer has once again set a new quantum volume (QV) record of 32,768. This achievement is on track with Quantinuum’s initial 2020 roadmap that projected it would increase quantum volume by 10x annually for five years.

Early experimental versions of the Quantinuum H1-1 quantum computer demonstrated relatively small values of quantum volume. When the machine made its public debut in 2020, Quantinuum (Honeywell Quantum Solutions at that time) had a documented QV of 64. Since then, Quantinuum has steadily doubled the H1 quantum volume eight consecutive times to reach its current value of 32,768.

What is quantum volume?

There are many tests available for determining the performance of individual quantum components and systems. However, only a few tests can measure a quantum computer's overall performance. IBM filled this void for gate-based quantum computers in 2019 with the development of quantum volume (QV). QV has the added benefit of producing an easy-to-interpret, single-number measurement; the higher the QV number, the more powerful the quantum computer.

Quantum volume is holistic, which means it can’t be gamed by inflating one or two of the factors, such as adding many qubits without making any other adjustments. To raise QV, all parts of the system must be improved in an integrated fashion.

Quantum computers and speedometers

Simplistically, quantum volume can be compared to a speedometer in a high-performance race car. Both measurements reflect the total performance of a complex system. A speedometer measures speed created by powerful motors and an assortment of precision components and finely tuned systems. Racing scientists optimize the car’s aerodynamics, its fuel mixture, tires and the gearing of complex transmissions to obtain the highest speed.

However, there is another factor—friction—that works against speed. Most friction comes from air resistance, the road surface and mechanical parts within the car.

Similarly, quantum volume measures a quantum computer’s overall performance. QV is influenced directly and indirectly by the number of qubits, qubit connectivity, speed and fidelity of quantum gates, cross talk, circuit compiler efficiency and measurement errors. Just as friction negatively affects the performance of cars, errors reduce the performance of quantum computers. Quantum errors are also commonly enumerated as a fidelity percentage.

High-fidelity qubits translate into high quantum volume

Qubits are the basic units of information in quantum computers. Quantum computing’s awesome potential to exponentially out-compute classical supercomputers comes from the quantum properties of superposition and entanglement that allow qubits to interact simultaneously.

Qubit quality compared to qubit quantity is often misunderstood by people unfamiliar with quantum computing. Once, while participating in a quantum forum, I was asked if a large number of low-fidelity qubits could outperform a handful of high-quality qubits. For many reasons, the answer was no. Errors generated by qubits of poor quality will likely degrade the computer’s performance rather than improve it. By the same standard, low-fidelity qubits would also degrade quantum volume.

QCCD: The Secret Behind Quantinuum's High Performance

Quantinuum's ability to continually improve its quantum volume can be attributed to its trapped ion architecture, which is called a quantum charge coupled device (QCCD).

Quantinuum was the first company to implement and improve QCCD after it was proposed twenty years ago in a research paper by Dr. David Wineland (recipient of the 2012 Nobel Prize in physics) and his NIST group. Dr. Chris Monroe, co-founder and chief scientist for IonQ and professor of physics and electrical and computer engineering at Duke University, was one of the authors of that paper.

Quantinuum’s QCCD architecture consists of multiple dedicated zones, into which small numbers of ytterbium and barium ions can be transported to perform quantum computations. QCCD also supports an important feature called mid-circuit measurement and reset (MCMR). It allows an algorithm to be paused during its execution to measure qubits without affecting the outcome. MCMR is expected to play an important future role in quantum error correction. Because of its capability to reuse qubits, in some instances it can reduce the total number of qubits needed for an operation.

Quantinuum’s H1 generation H-Series quantum computer currently has 20 fully connected qubits spread across its five QCCD zones where qubit operations are performed. In the future, when Quantinuum chooses to scale up the number of qubits, it can add additional zones.

Increasing quantum volume

Quantinum’s QCCD architecture has helped its researchers increase quantum volume and system fidelity for the System Model H1 . Here are two examples of this:

1. Quantinuum’s continuous fidelity improvement for the past three years across five of the most important quantum computing performance metrics has helped boost its QV. Thesemetrics are single-qubit gates, two-qubit gates, state preparation and measurement (SPAM), memory errors and crosstalk.
2. Single-qubit gates and fully entangling two-qubit gates are routinely used to build quantum circuit operations. However, not all algorithms need fully entangling two-qubit gates.

QCCD creates two-qubit gates by moving both qubits into an isolation zone to reduce potential errors and crosstalk that could occur if the zone contained many qubits. With greater precision and control over a few qubits in a small zone, QCCD is not limited to creating only normal two-qubit full entangling gates; the architecture also allows creation of arbitrary angle partially entangling gates. These gates have the advantage of being able to run on many circuit types with fewer errors. For example, the quantum Fourier transform (QFT) is used in many quantum algorithms, the most famous being Shor’s algorithm. When arbitrary angle partially entangling gates are substituted for normal two-qubit entangling gates in the QFT, half as many gates are required, and errors are reduced by 2x.

It should be no surprise that Quantinuum used arbitrary angle partially entangling gates in the QV circuits that produced its latest quantum volume record of 32,768.

Performance insights

Quantum volume is not just a performance indicator. It can also be used to assist development efforts. Along with its quantum volume announcement, Quantinuum provided a few interesting insights into how its researchers solved several technical issues involving quantum volume, as well as a way to greatly improve circuit runtime:

• To continually improve quantum volume, researchers must understand their error budgets and how to reduce errors. This was apparent in Quantinuum's work where the increase in quantum volume was attributed to several improvements but mainly laser phase noise.
• Researchers in Quantinuum's Munich, Germany labs were working on a project where the answers given by the quantum computer are very sensitive to noise. Since quantum computing is probabilistic by its nature, circuits must be run many times (these are called “shots”) to find an answer with the most occurrences. The higher the noise the more shots are needed to say with confidence that you have extracted the right answer. More shots take more time, and extended runtimes can be very problematic for long circuits or when quantum hardware and resources are limited.

By benchmarking the Quantinuum H1’s performance on relevant circuits, the researchers found that a very slight change in gate fidelity reduced the algorithm’s runtime by 3x.

Wrapping up

Quantinuum has had many research achievements over the past 12 months that are important to its long-term success. One example is implementing fault-tolerant entangling gates on the five-qubit code and the color code, which are important to the future development of quantum error correction.

In another breakthrough, Quantinuum researchers eliminated a potential obstacle in its long-term roadmap by developing a method to simultaneously move two types of ions through junctions and make right turns in ion traps. Prior to this development it was believed each type of ion would have to be moved separately through the junctions and then recombined—at a high cost of time.

Everything considered, Quantinuum has made great progress over the past two years. Looking ahead, I expect Quantinuum will continue to focus on higher fidelities and expand on its real-time quantum error mitigation and quantum error correction research. It is also possible we could see a flip in its use of ytterbium ions as qubits and barium ions for cooling, which offers several advantages including increased gate fidelity.

Quantinuum has solid management, an excellent quantum technology platform and an aggressive roadmap. It will be interesting to see what changes, if any, its new CEO, Dr. Raj Hazra will make.

Analyst notes

1. Quantinuum and IonQ are the two major trapped-ion quantum computer companies. As the article highlights, Quantinuum's QCCD architecture consists of multiple small zones, each containing small numbers of qubits. IonQ's architecture is built around multi-core quantum processors containing larger numbers of qubits that will eventually be interconnected with photonic networking. Both companies have made significant contributions to the quantum ecosystem.
2. To avoid any purist criticism, I only used a speedometer for comparison with quantum volume because it was simple and something everyone should be familiar with. I also want to note that IBM also developed a speed benchmark for quantum computers called Circuit Layer Operations per Second (CLOPS) that measures how many QV circuits a quantum processor can execute per unit of time.
3. Quantinuum’s roadmap for its H1 series hardware covers its evolution from the first-generation H1 processor in 2020 to the H5 processor, which is expected to be released in 2030.
4. Although no formal announcement has been made about a release date for the System Model H2, my estimate is that it will probably be announced sometime in mid-2023.
5. Determining quantum volume involves the output probability obtained when a special quantum circuit protocol is run. Simplistically, specific quantum circuits must be optimized, then run numerous times to determine what percentage of the outcomes fall above a specified level.
6. According to Quantinuum, the average H1-1 single-qubit gate fidelity for the new QV milestones was 99.9955(8)%, the average two-qubit gate fidelity was 99.795(7)% with fully connected qubits and state preparation and measurement fidelity was 99.69(4)%. For both quantum volumes, the Quantinuum team ran 100 circuits with 200 shots each, using standard QV optimization techniques to yield an average of 219.02 arbitrary angle two-qubit gates per circuit on the 2¹⁴ test, and 244.26 arbitrary angle two-qubit gates per circuit on the 2¹⁵ test.
7. Quantinuum uses an agile development process for its quantum computers. The H-Series systems are continuously upgraded with improvements implemented directly on the commercially available H1-1 system. That gives users immediate access to all improvements to system performance.

Follow me on Twitter for current information on what’s happening in quantum computing and artificial intelligence