Researchers “Hide” Ions to Reduce Quantum Errors
A new technique developed by Honeywell researchers reduced the amount of measured “crosstalk” errors by an order of magnitude
Cambridge Researchers at Honeywell Quantum Solutions have turned problematic micromotion that jostles trapped ion qubits out of position into a plus.
The team recently demonstrated a technique that uses micromotion to shield nearby ions from stray photons released during mid-circuit measurement, a procedure in which lasers are used to check the quantum state of certain qubits and then reset them.
Mid-circuit measurement is a key capability in today’s early-stage quantum computers. Because the qubit’s state can be checked and then re-used, researchers can run more complex algorithms – such as the holoQUADS algorithm – with fewer qubits.
By “hiding” ions behind micromotion, Honeywell researchers significantly reduced the amount of “crosstalk” – errors caused by photons hitting neighboring qubits – that occurred when measuring qubits during an operation. (Details are available in a pre-print publication available on the arXiv.)
“We were able to reduce crosstalk by an order of magnitude,” said Dr. John Gaebler, Chief Scientist of Commercial Products at Honeywell Quantum Solutions, and lead author of the paper. “It is a significant reduction in crosstalk errors. Much more so than other methods we’ve used.”
The new technique represents another step toward reducing errors that occur in today’s trapped-ion quantum computers, which is necessary if the technology is to solve problems too complex for classical systems.
“For quantum computers to scale, we need to reduce errors throughout the system,” said Tony Uttley, President of Honeywell Quantum Solutions. “The new technique the Honeywell team developed will help us get there.”
Eliminating errors
Today’s quantum computing technologies are still in the early stage and are prone to “noise” - or interference - caused by qubits interacting with their environment and one another.
This noise causes errors to accumulate, corrupts information stored in and between physical qubits, and disrupts the quantum state in which qubits must exist to run calculations. (Scientists call this decoherence.)
Researchers are trying to eliminate or suppress as many of these errors as possible while also creating logical qubits, a collection of entangled physical qubits on which quantum information is distributed, stored, and protected.
By creating logical qubits, scientists can apply mathematical codes to detect and correct errors and eliminate noise as calculations are running. This multi-step process is known as quantum error correction (QEC). Honeywell researchers recently demonstrated they can detect and correct errors in real-time by applying multiple rounds of full cycles of quantum error correction.
Logical qubits and QEC are important elements to improving the accuracy and precision of quantum computers. But, Gaebler said, those methods are not enough on their own.
“Everything has to be working at a certain level before QEC can take you the rest of the way,” he said. “The more we can suppress or eliminate errors in the overall system, the more effective QEC will be and the fewer qubits we need to run complex calculations.”
Cutting out crosstalk
In classical computing, bit flip errors occur when a binary digit, or bit, inadvertently switches from a zero to one or vice versa. Quantum computers experience a similar bit flip error as well as phase flip errors. Both errors cause qubits to lose their quantum state – or to decohere. In trapped ion quantum computing, one source of errors comes from the lasers used to implement gate operations and qubit measurements.
Though these lasers are highly controlled, unruly photons (small packets of light) still escape and bounce into neighboring ions causing “crosstalk” and decoherence.
Researchers use a variety of methods to protect these ions from crosstalk, especially during mid-circuit measurement where only a single qubit or a small subset of qubits is meant to be measured. With its quantum charged-coupled device (QCCD) architecture, the Honeywell team takes the approach of moving neighboring ions away from the qubit being fluoresced by a laser. But there is limited space along the device, which becomes even more compact as more qubits are added.
“Even when we move them more than 100 microns away, we still get more crosstalk than we prefer,” said Dr. Charlie Baldwin, a senior advanced physicist and co-author of the paper. “There is still some scattered light from the detection laser.”
The team hit on hiding neighboring ions from stray photons using micromotion potentials, which are caused by the oscillating electric fields used to “trap” these charged atoms. Micromotion is typically thought of as a nuisance with ion trapping, causing the ions to rapidly oscillate back and forth, and occurs when the ions are pushed out of the center of the trap by additional electric fields.
“Usually, we are trying to eliminate micromotion but in this case, we were able to use it to our benefit,” said Dr. Patty Lee, chief scientist at Honeywell Quantum Solutions.
The team’s goal is to reduce by 10 million the probability of a neighboring ion absorbing photons at 110 microns away. By moving neighboring ions and hiding them behind micromotion the Honeywell team is approaching that mark.
How and why the technique works
In their paper, Honeywell researchers delved into how and why hiding ions with micromotion works, including the ideal frequency of the oscillations. They also identified and characterized errors. (The basic physics behind the concept of hiding ions was first explored by the ion storage group at the National Institute of Standards and Technology.)
“Mid-circuit operations are a new feature in commercial quantum computing hardware, so we had to invent a new way to validate that the micromotion hiding technique was achieving the low level of crosstalk errors that we predicted,” said Dr. Charlie Baldwin.
Though the new method resulted in a significant reduction of crosstalk errors, the Honeywell team acknowledged there is further to go.
“Crosstalk is one of those scary errors for scaling,” Gaebler said. “It has to be controlled because it becomes more of a problem as you scale and add qubits. This is another tool that will help us scale and help us compact our systems and pack in as many qubits as we can.”
About Quantinuum
Quantinuum, the world’s largest integrated quantum company, pioneers powerful quantum computers and advanced software solutions. Quantinuum’s technology drives breakthroughs in materials discovery, cybersecurity, and next-gen quantum AI. With over 500 employees, including 370+ scientists and engineers, Quantinuum leads the quantum computing revolution across continents.
While it sounds like a gadget from Star Trek, teleportation is real – and it is happening at Quantinuum. In a new paper published in Science, our researchers moved a quantum state from one place to another without physically moving it through space - and they accomplished this feat with fault-tolerance and excellent fidelity. This is an important milestone for the whole quantum computing community and the latest example of Quantinuum achieving critical milestones years ahead of expectations.
While it seems exotic, teleportation is a critical piece of technology needed for full scale fault-tolerant quantum computing, and it is used widely in algorithm and architecture design. In addition to being essential on its own, teleportation has historically been used to demonstrate a high level of system maturity. The protocol requires multiple qubits, high-fidelity state-preparation, single-qubit operations, entangling operations, mid-circuit measurement, and conditional operations, making it an excellent system-level benchmark.
Our team was motivated to do this work by the US Government Intelligence Advance Research Projects Activity (IARPA), who set a challenge to perform high fidelity teleportation with the goal of advancing the state of science in universal fault-tolerant quantum computing. IARPA further specified that the entanglement and teleportation protocols must also maintain fault-tolerance, a key property that keeps errors local and correctable.
These ambitious goals required developing highly complex systems, protocols, and other infrastructure to enable exquisite control and operation of quantum-mechanical hardware. We are proud to have accomplished these goals ahead of schedule, demonstrating the flexibility, performance, and power of Quantinuum’s Quantum Charge Coupled Device (QCCD) architecture.
Quantinuum’s demonstration marks the first time that an arbitrary quantum state has been teleported at the logical level (using a quantum error correcting code). This means that instead of teleporting the quantum state of a single physical qubit we have teleported the quantum information encoded in an entangled set of physical qubits, known as a logical qubit. In other words, the collective state of a bunch of qubits is teleported from one set of physical qubits to another set of physical qubits. This is, in a sense, a lot closer to what you see in Star Trek – they teleport the state of a big collection of atoms at once. Except for the small detail of coming up with a pile of matter with which to reconstruct a human body...
This is also the first demonstration of a fully fault-tolerant version of the state teleportation circuit using real-time quantum error correction (QEC), decoding mid-circuit measurement of syndromes and implementing corrections during the protocol. It is critical for computers to be able to catch and correct any errors that happen along the way, and this is not something other groups have managed to do in any robust sense. In addition, our team achieved the result with high fidelity (97.5%±0.2%), providing a powerful demonstration of the quality of our H2 quantum processor, Powered by Honeywell.
Our team also tried several variations of logical teleportation circuits, using both transversal gates and lattice surgery protocols, thanks to the flexibility of our QCCD architecture. This marks the first demonstration of lattice surgery performed on a QEC code.
Lattice surgery is a strategy for implementing logical gates that requires only 2D nearest-neighbor interactions, making it especially useful for architectures whose qubit locations are fixed, such as superconducting architectures. QCCD and other technologies that do not have fixed qubit positioning might employ this method, another method, or some mixture. We are fortunate that our QCCD architecture allows us to explore the use of different logical gating options so that we can optimize our choices for experimental realities.
While the teleportation demonstration is the big result, sometimes it is the behind-the-scenes technology advancements that make the big differences. The experiments in this paper were designed at the logical level using an internally developed logical-level programming language dubbed Simple Logical Representation (SLR). This is yet another marker of our system’s maturity – we are no longer programming at the physical level but have instead moved up one “layer of abstraction”. Someday, all quantum algorithms will need to be run on the logical level with rounds of quantum error correction. This is a markedly different state than most present experiments, which are run on the physical level without quantum error correction. It is also worth noting that these results were generated using the software stack available to any user of Quantinuum’s H-Series quantum computers, and these experiments were run alongside customer jobs – underlining that these results are commercial performance, not hero data on a bespoke system.
Ironically, a key element in this work is our ability to move our qubits through space the “normal” way - this capacity gives us all-to-all connectivity, which was essential for some of the QEC protocols used in the complex task of fault-tolerant logical teleportation. We recently demonstrated solutions to the sorting problem and wiring problem in a new 2D grid trap, which will be essential as we scale up our devices.
The central question that pre-occupies our team has been:
“How can quantum structures and quantum computers contribute to the effectiveness of AI?”
In previous work we have made notable advances in answering this question, and this article is based on our most recent work in the new papers [arXiv:2406.17583, arXiv:2408.06061], and most notably the experiment in [arXiv:2409.08777].
This article is one of a series that we will be publishing alongside further advances – advances that are accelerated by access to the most powerful quantum computers available.
Large language Models (LLMs) such as ChatGPT are having an impact on society across many walks of life. However, as users have become more familiar with this new technology, they have also become increasingly aware of deep-seated and systemic problems that come with AI systems built around LLM’s.
The primary problem with LLMs is that nobody knows how they work - as inscrutable “black boxes” they aren’t “interpretable”, meaning we can’t reliably or efficiently control or predict their behavior. This is unacceptable in many situations. In addition, Modern LLMs are incredibly expensive to build and run, costing serious – and potentially unsustainable –amounts of power to train and use. This is why more and more organizations, governments, and regulators are insisting on solutions.
But how can we find these solutions, when we don’t fully understand what we are dealing with now?1
At Quantinuum, we have been working on natural language processing (NLP) using quantum computers for some time now. We are excited to have recently carried out experiments [arXiv: 2409.08777] which demonstrate not only how it is possible to train a model for a quantum computer in a scalable manner, but also how to do this in a way that is interpretable for us. Moreover, we have promising theoretical indications of the usefulness of quantum computers for interpretable NLP [arXiv:2408.06061].
In order to better understand why this could be the case, one needs to understand the ways in which meanings compose together throughout a story or narrative. Our work towards capturing them in a new model of language, which we call DisCoCirc, is reported on extensively in this previous blog post from 2023.
In new work referred to in this article, we embrace “compositional interpretability” as proposed in [arXiv:2406.17583] as a solution to the problems that plague current AI. In brief, compositional interpretability boils down to being able to assign a human friendly meaning, such as natural language, to the components of a model, and then being able to understand how they fit together2.
A problem currently inherent to quantum machine learning is that of being able to train at scale. We avoid this by making use of “compositional generalization”. This means we train small, on classical computers, and then at test time evaluate much larger examples on a quantum computer. There now exist quantum computers which are impossible to simulate classically. To train models for such computers, it seems that compositional generalization currently provides the only credible path.
1. Text as circuits
DisCoCirc is a circuit-based model for natural language that turns arbitrary text into “text circuits” [arXiv:1904.03478, arXiv:2301.10595, arXiv:2311.17892]. When we say that arbitrary text becomes ‘text-circuits’ we are converting the lines of text, which live in one dimension, into text-circuits which live in two-dimensions. These dimensions are the entities of the text versus the events in time.
To see how that works, consider the following story. In the beginning there is Alex and Beau. Alex meets Beau. Later, Chris shows up, and Beau marries Chris. Alex then kicks Beau.
The content of this story can be represented as the following circuit:
2. From text circuits to quantum circuits
Such a text circuit represents how the ‘actors’ in it interact with each other, and how their states evolve by doing so. Initially, we know nothing about Alex and Beau. Once Alex meets Beau, we know something about Alex and Beau’s interaction, then Beau marries Chris, and then Alex kicks Beau, so we know quite a bit more about all three, and in particular, how they relate to each other.
Let’s now take those circuits to be quantum circuits.
In the last section we will elaborate more why this could be a very good choice. For now it’s ok to understand that we simply follow the current paradigm of using vectors for meanings, in exactly the same way that this works in LLMs. Moreover, if we then also want to faithfully represent the compositional structure in language3, we can rely on theorem 5.49 from our book Picturing Quantum Processes, which informally can be stated as follows:
If the manner in which meanings of words (represented by vectors) compose obeys linguistic structure, then those vectors compose in exactly the same way as quantum systems compose.4
In short, a quantum implementation enables us to embrace compositional interpretability, as defined in our recent paper [arXiv:2406.17583].
3. Text circuits on our quantum computer
So, what have we done? And what does it mean?
We implemented a “question-answering” experiment on our Quantinuum quantum computers, for text circuits as described above. We know from our new paper [arXiv:2408.06061] that this is very hard to do on a classical computer due to the fact that as the size of the texts get bigger they very quickly become unrealistic to even try to do this on a classical computer, however powerful it might be. This is worth emphasizing. The experiment we have completed would scale exponentially using classical computers – to the point where the approach becomes intractable.
The experiment consisted of teaching (or training) the quantum computer to answer a question about a story, where both the story and question are presented as text-circuits. To test our model, we created longer stories in the same style as those used in training and questioned these. In our experiment, our stories were about people moving around, and we questioned the quantum computer about who was moving in the same direction at the end of the stories. A harder alternative one could imagine, would be having a murder mystery story and then asking the computer who was the murderer.
And remember - the training in our experiment constitutes the assigning of quantum states and gates to words that occur in the text.
4. Compositional generalization
The major reason for our excitement is that the training of our circuits enjoys compositional generalization. That is, we can do the training on small-scale ordinary computers, and do the testing, or asking the important questions, on quantum computers that can operate in ways not possible classically. Figure 4 shows how, despite only being trained on stories with up to 8 actors, the test accuracy remains high, even for much longer stories involving up to 30 actors.
Training large circuits directly in quantum machine learning, leads to difficulties which in many cases undo the potential advantage. Critically - compositional generalization allows us to bypass these issues.
5. Real-world comparison: ChatGPT
We can compare the results of our experiment on a quantum computer, to the success of a classical LLM ChatGPT (GPT-4) when asked the same questions.
What we are considering here is a story about a collection of characters that walk in a number of different directions, and sometimes follow each other. These are just some initial test examples, but it does show that this kind of reasoning is not particularly easy for LLMs.
The input to ChatGPT was:
What we got from ChatGPT:
Can you see where ChatGPT went wrong?
ChatGPT’s score (in terms of accuracy) oscillated around 50% (equivalent to random guessing). Our text circuits consistently outperformed ChatGPT on these tasks. Future work in this area would involve looking at prompt engineering – for example how the phrasing of the instructions can affect the output, and therefore the overall score.
Of course, we note that ChatGPT and other LLM’s will issue new versions that may or may not be marginally better with ‘question-answering’ tasks, and we also note that our own work may become far more effective as quantum computers rapidly become more powerful.
6. What’s next?
We have now turned our attention to work that will show that using vectors to represent meaning and requiring compositional interpretability for natural language takes us mathematically natively into the quantum formalism. This does not mean that there doesn't exist an efficient classical method for solving specific tasks, and it may be hard to prove traditional hardness results whenever there is some machine learning involved. This could be something we might have to come to terms with, just as in classical machine learning.
At Quantinuum we possess the most powerful quantum computers currently available. Our recently published roadmap is going to deliver more computationally powerful quantum computers in the short and medium term, as we extend our lead and push towards universal, fault tolerant quantum computers by the end of the decade. We expect to show even better (and larger scale) results when implementing our work on those machines. In short, we foresee a period of rapid innovation as powerful quantum computers that cannot be classically simulated become more readily available. This will likely be disruptive, as more and more use cases, including ones that we might not be currently thinking about, come into play.
Interestingly and intriguingly, we are also pioneering the use of powerful quantum computers in a hybrid system that has been described as a ‘quantum supercomputer’ where quantum computers, HPC and AI work together in an integrated fashion and look forward to using these systems to advance our work in language processing that can help solve the problem with LLM’s that we highlighted at the start of this article.
1 And where do we go next, when we don’t even understand what we are dealing with now? On previous occasions in the history of science and technology, when efficient models without a clear interpretation have been developed, such as the Babylonian lunar theory or Ptolemy’s model of epicycles, these initially highly successful technologies vanished, making way for something else.
2 Note that our conception of compositionality is more general than the usual one adopted in linguistics, which is due to Frege. A discussion can be found in [arXiv: 2110.05327].
3 For example, using pregroups here as linguistic structure, which are the cups and caps of PQP.
4 That is, using the tensor product of the corresponding vector spaces.