Quantum Computing "Noise"
Originally posted Sept. 8, 2022; Revised, updated and reposted Jan. 30, 2025
We all know what “noise” is. And we all appreciate that it is usually an unwelcomed invasion of our peace and quiet. Screaming babies on airplanes, jackhammers in the street, leaf blowers outside your window - all can ruin an otherwise tranquil setting. “Noise” in computer lingo represents a similar disconcerting situation.
In Quantum Computing (QC), you likely have come across the concept of noise as a major obstacle to QC’s achieving their potential. In fact, John Preskill, a professor of theoretical physics at Caltech and one of the pioneers of QC, coined the acronym “NISQ”, standing for Noisy Intermediate-Scale Quantum Computers which is used to describe today’s QC stage. There are several significant challenges facing QC makers today, and “noise” is one of the most difficult to overcome.
Quantum Computing Noise
There are many causes for the underlying noise in QCs and here are a few of the core sources:
● The Environment: Qubits are exquisitely sensitive to any changes in their environment. Small changes in temperature or stray electrical or magnetics fields can disturb qubits and cause a degradation of the information. Even weak galactic space radiation can push qubits and thereby degrade them.
● Crosstalk: Quantum Computers are powered by qubits acting together. Generally, individual qubits are manipulated by lasers or microwaves. However, sometimes the laser or microwave signal can impact nearby qubits as well as the target qubit, an issue known as crosstalk.
● Quantum Decoherence: A qubit’s quantum state deteriorates rapidly, often even after just fractions of a second, requiring QCs to initiate and complete their algorithms before quantum states collapse.
● Implementation Errors: The commands or gates of a quantum algorithm apply various rotations to the qubit, which are implemented by laser or microwave pulses which can also be somewhat imprecise. For example, an X-Gate, which is analogous to a NOT gate in a classical computer, essentially “flips” the qubit rotating it by 180 degrees. If the pulse command to do this only leads to a 179-degree rotation, the subsequent calculations will be off by a potentially meaningful amount.
You may be familiar with the term “five 9’s” which has often been used in the context of super-high performance. It generally means a system with 99.999% accuracy, or only one error per 100,000 instances. For service level agreements with, say your cloud provider, five nines would mean less than 5.26 minutes of downtime per year. It’s a high standard, recognizing the reality that certain systems suffer from various unknown or unpredictable challenges. While Quantum Computer makers continue to improve upon the fidelities of their qubits (the underlying physical components which process quantum gates and algorithms), very few have been able to achieve greater than 99.9% two-gate fidelities. While that may sound high and would likely have been an acceptable grade on your physics final, it is not enough to enable Quantum Computers to perform the complex algorithms necessary for QCs to outperform existing classical computers.
The non-technical takeaway: Quantum Computations are run via qubits which are very difficult to control, are vulnerable to the tiniest environmental changes and have a natural tendency to move, leading to a degradation of the information, also known as noise.
Noise is one of the most impactful challenges facing QC advancement today, and there are many ways that companies are addressing this issue.
How to Overcome Noise Constraints in Quantum Computing
In the 19th century, ships typically carried clocks set to the time in Greenwich in combination with the sun’s position in the sky for determining longitude during long trips. However, an incorrect clock could lead to dangerous navigational errors, so ships often carried three clocks. Two clocks showing differing times would detect a fault in one, but three were needed to identify which one was faulty (if two matched the third one was off). This is an example of a repetition code, where information is encoded redundantly in multiple devices, enabling detection and correction of a fault. In QCs, because measurement fundamentally disturbs quantum information, we can’t do interim measurements to identify errors because that would terminate the process, so data is shared among multiple qubits, often referred to as ‘ancillary’ qubits, ‘syndrome’ qubits or ‘helper’ qubits. A series of gates entangles these helper qubits with the original qubits, which effectively transfers noise from the system to multiple helpers. We can then measure the helpers via parity check, which, like those redundant clocks, can reveal errors without touching or measuring the original system. However, the trade-off is the requirement for many physical qubits to act as helpers, adding enormous overhead to QCs.
Also, since each step of a quantum algorithm is an opportunity for noise to be introduced, efforts to quicken the runtime or reduce the number of steps (i.e., gates) are intended to minimize the opportunity for noise to corrupt the output. In addition to repetition code methods of finding and correcting errors, and overall efforts to minimize circuit depth, there are a few other tools being used to tackle quantum noise. A high-level view of the quantum computing software “stack” should help provide some context for these added methods:
The graphic above is generally referred to as the “full stack” and there are opportunities at each level of the stack to help compensate for or minimize noise. Here are a few methods being deployed:
Quantum Control: At the qubit level, often referred to as the “metal”, engineers continue to optimize the pulses and control signals focused on the qubits as well as create modalities with increasing coherence times. Various ways that the qubits are aligned and/or inter-connected affect this level and advances are being continually announced.
Hardware Aware Optimization: At the Machine Instruction level, focus on transpiler efficiencies can reduce errors and minimize noise impacts. Again, various qubit configurations as well as the specific modalities utilized (superconducting vs optical vs ion vs cold atom, etc.) have an impact on the performance of the algorithms and attention to this level of the stack provides another opportunity for noise reduction.
Compiler Efficiency: Circuit optimization is a target of many players in the QC space. Tools that re-write algorithms to focus on this level of the stack is a growing and important part of the ecosystem. For example, efficient usage of ancillary qubits and/or resetting them quickly to be re-utilized, requires less run-time and less steps, which means less opportunity for noise to impact the programs.
Algorithm efficiency: There are many ways to write quantum algorithms so ensuring that the code is as efficient as possible (i.e., eliminating redundant steps or minimizing needs to reset or recalibrate qubits) is another opportunity to minimize noise. The more efficient the code, the quicker it can run, or the shorter its circuit depth needs to be.
Many Shots: A final tool which is a standard procedure in quantum algorithms, is to run the algorithm many times. Each run is referred to as a “shot” and typical algorithms are run with 1000’s of shots. By averaging the output of these shots, a “regression to the mean” is often realized, meaning the averaging of the results helps various noise impacts cancel each other out. [The fact that quantum algorithms are probabilistic and not deterministic is a major reason for the redundant shots, but this redundancy is also a tool to help overcome noise].
The non-technical takeaway: Noise is a major problem impacting the ability of Quantum Computers to achieve their potential. Until fault-tolerant hardware can be developed, quantum engineers are deploying several creative ways to overcome noise in current QCs.
Quantum Companies Addressing Quantum Noise
There are a number of players focused on noise reduction and deploying inventive solutions to optimize the performance of today’s quantum machines. Some of these methodologies can achieve performance improvements of orders of magnitude, so these methodologies are yielding significant improvements. As the quantum hardware players release ever-larger quantum machines (for example, IBM has recently released its Condor processer with 1,121 qubits and Atom Computing has a machine with 1,180 qubits) these error correcting strategies will greatly accelerate the ability of QCs to achieve quantum advantage, with many prognosticators (including yours truly) expecting such achievement over the next 12-18 months (at least for certain types of problems). The following is a brief overview of some of the players that offer various quantum noise-reduction solutions:
Google: While Google has made a number of somewhat controversial quantum computing claims around quantum supremacy, their recent (12/9/24) achievement on their Willow quantum processor was the first time any quantum company had “below-threshold” error correction. Since extra qubits are required for current error correction schemes, most players have seen their error correction actually lead to higher error rates (because of the higher number of qubits needed). However, Google was the first company to actually decrease the error rate as it added error-correcting qubits.
Microsoft/Quantinuum: Through a series of collaborations and advancements including using Quantinuum’s high-fidelity System Model H2 and Microsoft’s innovative qubit-virtualization system and tesseract code design, Quantinuum was able to improve error rates 800-fold. The system successfully ran over 14,000 quantum circuits without a single error.
Q-CTRL: has established itself as a leader in quantum control techniques, offering comprehensive solutions to combat quantum noise and improve overall system performance. Their approach encompasses both quantum algorithm and hardware design optimizations, with a focus on reducing the adverse effects of noise on quantum systems. Q-CTRL has demonstrated substantial improvements in error rates and circuit performance through strategic partnerships with major players in the quantum computing field, such as IBM. Their tools and methodologies are designed to enhance the stability and reliability of quantum operations, pushing the boundaries of what's possible in quantum computing.
Classiq: is at the forefront of quantum algorithm design, focusing on reducing quantum noise through innovative circuit optimization. Their platform employs advanced compression techniques to minimize the complexity of quantum circuits, which in turn helps to mitigate the impact of noise on quantum computations. By streamlining quantum algorithms, Classiq's approach not only enhances the efficiency of quantum operations but also potentially increases the resilience of quantum systems against environmental disturbances that can lead to errors.
Parity QC: focuses on radically reducing control complexity. This allows them to provide a fully programmable, parallelizable (no SWAP gates), and scalable architecture which is independent from the problem. Due to its ability to parallelize gates, the ParityQC Architecture introduces algorithms based on global gates. In each step, a pattern of gates are executed at the same time. This removes the need to implement a control signal for each individual gate and only requires ONE single control signal for all gates instead. This provides a huge advantage for the hardware design and a route to mitigate crosstalk errors during qubit design.
Riverlane: addresses quantum error correction technology with a whole stack focus, aiming to squeeze out every bit of efficiency. Their Deltaflow.OS® operating system is compatible with all current quantum hardware platforms including both gate-based and annealing methods. Riverlane's approach involves converting large numbers of unreliable qubits into smaller sets of more reliable 'logical' qubits, enabling real-time error decoding and logical operations. The company has made significant strides, including the creation of the world's first QEC chip, DD1, which achieves a remarkable logical error rate of one in a trillion. Riverlane's goal is to reach "million quantum operations" (MegaQuOp) by 2026, potentially revolutionizing industries such as healthcare and manufacturing.
Super.tech: now part of ColdQuanta, has been focusing on developing software to optimize and accelerate quantum computing applications. Their SuperstaQ quantum software platform is optimized across the entire quantum stack and includes a library of sophisticated error mitigation techniques, including dynamical decoupling, excited state promotion, and zero noise extrapolation. SuperstaQ automatically optimizes quantum programs based on the target hardware’s pulse-level native gates.
Xanadu: Their PennyLane software is leading programming tool leveraging a cross-platform Python library which enables quantum differentiable programming — that enables seamless integration with machine learning tools. PennyLane also supports a comprehensive set of features, simulators, hardware, and community-led resources that enable users of all levels to easily build, optimize and deploy quantum-classical applications.
Zapata: a quantum computing software company that develops solutions for a wide range of industries. Zapata’s Orquestra™ platform allows users to compose quantum-enabled workflows and orchestrate their execution across classical and quantum technologies. Orquestra combines a powerful software platform, quantum algorithm libraries, and example workflows across machine learning, simulation and optimization. Orquestra automatically scales up and exploits task parallelization opportunities to run quantum algorithms faster.
When I began following Quantum Computing, it was thought that we would need thousands (or even tens of thousands) of ancillary qubits to create one logical qubit. However, recent advances in quantum noise reduction and clever error correction schemes have dramatically reduced that ratio. At the same time, qubit makers continue to improve their fidelities. Change is happening quite rapidly, and I expect we will see many impressive achievements announced this year. I look forward to continuing to track and report on progress.
Disclosure: The author has no beneficial positions in stocks discussed in this review, nor does he have any business relationship with any company mentioned in this post. The views expressed herein are solely the views of the author and are not necessarily the views of Corporate Fuel Partners or any of its affiliates. Views are not intended to provide, and should not be relied upon for, investment advice.
References:
Ackermann, Thomas J., “Noisy Intermediate Scale Quantum (NISQ) Technology,” BGP4.com, October 17, 2018.
Bertels, Sarac, Sarkar and Ashraf, “Quantum Computing – From NISQ to PISQ”, arXiv:2106.11840v3[quant-ph], April 15, 2022.
Fellner, Messinger, Ender and Lechner, “Universal Parity Quantum Computing,” arXiv:2205.09505v1[quant-ph], May 19, 2022.
Lee, Chris, “No sugarcoating: Donut math yields way to make qubits last longer,” Science, July 2022.
Nazario, Zaira, “How to Fix Quantum Computer Bugs,” Scientific American, May 1, 2022
Russell, John, “New IBM Blog Details Path to Quantum Advantage in 2023,” HPCWire, July 19, 2022.
Shaw, David, “Quantum Software Outlook 2022”, Fact Based Insight, January 17, 2022.
Smith-Goodson, Paul, “Quantinuum Makes a Significant Quantum Computing Breakthrough by Connecting the Dots of its Previous Research,” Forbes, 8/4/22.
“Zapata Computing Publishes New Research on using Orquestra Platform to Implement Fundamental Subroutine for Quantum Algorithms,” Zapatacomputing.com, Oct. 21, 2021.
Stace, Thomas M. and Biercuk, Michael J., “Quantum Error Correction: Time to Make it Work,” Spectrum.IEEE.org, June 26, 2022.
Photo by Ketut Subiyanto: https://www.pexels.com/photo/man-in-blue-button-up-shirt-smiling-4584184/
Graphic from Dunning, Alexander & Gregory, Rachel & Bateman, James & Cooper, Nathan & Himsworth, Matthew & Jones, Jonathan & Freegarde, Tim. Composite pulses for interferometry in a thermal cold atom cloud. Physical Review A. 90. 033608. 10.1103/PhysRevA.90.033608. (2014).