Fractions of a second to detect all quantum computing errors, moving towards an era of scalable error correction

On November 11, at the UK's National Quantum Technology Showcase (NQTS), Riverlane demonstrated live the entire operational cycle of accurately detecting specific quantum data errors in fractions of a second on a simulated quantum computer [1]: Riverlane's quantum decoder is the world's first to support multiple quantum bits and will power the first generation of fault-tolerant quantum computers.

01Scalable decoder chip: a critical leap towards fault-tolerant quantum computing

This demonstration is a prototype of what will become a decoding chip that can be placed in all future quantum computers; at the same time, due to the scalability of the decoding hardware (supporting much larger quantum bits than currently possible), this breakthrough decoding technology is the first invention of its kind in the world, called Deltaflow.Decode.

Today's quantum computers have limited utility due to the inherent instability of the quantum bits: this instability leads to a large number of data errors that overwhelm all current quantum computers. We are now entering the first generation of error-correcting quantum computing, where such errors can be detected, diagnosed and corrected in real time. riverlane's scalable high-speed decoder is a key component in enabling the transition to this new era.

"To effectively solve currently unsolvable human problems in areas such as clean energy and new drug design, we need to transition to a new generation of error-correcting quantum computers that can perform millions of high-speed operations without interference. Today's quantum computers still can only perform about 100 operations before they fail. This transition will take time, but it should start now. Our decoder is a key component and a leap forward." Steve Brierley, founder and CEO of Riverlane, said of this.

02Deltaflow.Decode Solution: Tackling Quantum Error Correction (QEC)

As the quantum computations we have to perform take longer and longer, larger and larger error correction codes are required and more and more data is generated. A common-scale quantum computer is expected to generate several terabytes of error correction data per second, an amount of data equivalent to that generated by Netflix's global streaming data, or CERN's ALICE detector. All of this data needs to be processed quickly as it is generated, or else the computation could come to a standstill.

Riverlane therefore proposes a new approach to parallelize the data processing problem (processing syndrome data in real time), thus enabling quantum error correction (QEC) problem processing at almost arbitrary speed without sacrificing accuracy. The team developed Deltaflow.Decode, a complete solution designed to help hardware partners achieve the goal of fault-tolerant quantum computing.

Riverlane's latest decoder (RL) with Union-Find (UF) algorithm - one of the leading solutions for real-time decoding today. By modeling the decoder hardware, Riverlane's improved algorithm is 60-80% faster in a range of decoding tasks. In addition, a 70% memory footprint is achieved, significantly improving the power consumption and scalability of the device.

Riverlane decoder frequencies for measurement errors were calculated using rotating plane codes (RLv4.1). (a) Decoding frequencies for different code sizes at different circuit-level depolarization noise p. (b) Predicted online decoding frequency of the Riverlane decoder (dark green line) at 0.4% circuit-level noise for an architecture using multiple decoding cores compared to the conventional serial approach. The red dashed line sets a standard for the minimum decoding speed required for in-line decoding on superconducting devices.

The solutions proposed in the literature have not addressed the scalability issue, and the decoding problem will become more difficult as the code spacing is increased to further suppress errors; therefore, no matter how fast the decoder is implemented, there will always be a maximum distance: beyond this distance, the hardware will not be able to keep up with the data acquisition. riverlane solves this critical problem, allowing the decoder to decode without compromising accuracy can be efficiently distributed over many cores, with bandwidth limited mainly by the number of classical resources that can be dedicated to the task. The results show that the new decoder is able to decode the traditionally challenging distance11 encoding with a modest number of cores; moreover, the additional bandwidth that this solution can provide is crucial for solving the more complex problems that arise during the logical operations required to generate the entangled quantum bits.

Finally, while the speed of the decoder is a key metric, it is also important to ensure robust suppression of logical errors. In order to obtain good logic performance, the decoder must be adapted to the noise that occurs in the syndrome extraction circuit. In the figure below, the team simulated the decoder's logic error rate on a realistic circuit-level noise model; ultimately, it was found that there was a significant exponential suppression of errors as the distance increased.

Decoder accuracy comparison. (a) Logic error rate as a function of code distance for RLv4.1 (square and solid lines) and MWPM (triangle and dashed lines) decoders at different circuit-level noise levels. (b) Logical error rate as a function of the number of decoding rounds using MWPM and RLv4.1 decoders for data from a real quantum computer published by Google Quantum AI.

In addition, Riverlane's decoder was compared to Minimum Weight Perfect Match (MWPM), the industry standard for decoding surface codes: there was only a small increase in the logical error rate, which was easily compensated for by an increase in code distance. The team also tested the decoder on data from a real quantum computer released by Google Quantum AI, and as the graph shows, Riverlane's decoder performed well in the experiment, with the MWPM maintaining only a small advantage in the logical error rate.

As a result, Riverlane said, "We are confident that our decoder has the ability to solve real-world problems and help our partners achieve fault-tolerant quantum computing." Riverlane's decoder will set a new industry standard for decoding speed that can scale to solve complex decoding problems with comparable accuracy to slower MWPM decoders.

03Riverlane's path to quantum error correction: an early layout

Back in December 2021, Riverlane announced [2] that it had received £7.5 million in funding through the UK government's National Quantum Technologies Programme to build an error-correcting quantum processor. Quantum computer manufacturer Universal Quantum will use Riverlane's software and expertise to address quantum error correction for trapped ion quantum computers; Riverlane will also work with Rolls-Royce to explore how quantum computers can provide practical applications for developing more sustainable, efficient jet engines.

In January 2022, Riverlane announced [3] the appointment of Dr. Earl Campbell, a researcher and senior lecturer at the University of Sheffield and former senior scientist at Amazon Web Services (AWS), as head of architecture.

Earl Campbell

Earl has spent over 16 years developing how best to design fault-tolerant quantum computing architectures, including contributions to quantum error correction, fault-tolerant quantum logic and compilation, and quantum algorithms. During his time at Riverlane, Earl will work with leaders from Microsoft, ARM, Samsung, Intel and the White House, and will be responsible for leading the technical development and architectural design of software to support fault-tolerant quantum computing hardware.

Commenting on the appointment, Steve Brierley, CEO and founder of Riverlane, said, "Solving the error correction problem is the next defining challenge for quantum computing and will be key to unlocking quantum utility across a range of fundamental challenges, including clean energy, drug discovery, materials science and advanced chemistry. We are excited that Earl is bringing his world-class expertise on this challenge to the Riverlane team to accelerate our efforts and unlock the potential of this technology."

In June 2022, Riverlane and Rigetti, a pioneer in hybrid-classical computing, collaborated and announced [4], with the support of Innovate UK, a partnership to solve the syndrome extraction problem for superconducting quantum computers - a key step in quantum error correction.

Specifically, quantum mechanics prohibits the direct measurement of the primary quantum bits that perform the actual computation because it would destroy the information they carry; therefore, error correction techniques use additional quantum bits, called "syndrome quantum bits," to measure their state - The syndrome allows inferences to be made about the occurrence of errors in the main quantum bits.

Commenting on the company's long-standing layout for quantum error correction, Steve Brierley, founder and CEO of Riverlane, said, "Some problems are quantum mechanical in nature, so they can only be solved by a quantum computer. Thus, quantum computers offer an opportunity to advance many areas of science from the era of discovery through extensive trial and error to the era of design where all possibilities can be simulated. Error correction is one of the keys to unlocking this future; in the future, we will further delve into error correction technology and aspire to solve the error correction problem for the entire industry."

About Riverlane.

Founded in 2016 and based in Cambridge, UK, Riverlane is a quantum computing software company spun out of the University of Cambridge. Founded by Dr. Steve Brierley, Senior Research Fellow in Applied Mathematics at the University of Cambridge: Steve has over a decade of research experience in quantum information and computing, studying quantum systems theory, quantum computer architectures and quantum algorithms.

Since July 2019, Riverlane has been working on the development of Deltaflow.OS, the world's first quantum operating system. As the first crab in the field of quantum operating systems, Riverlane will break the barriers between quantum computing hardware and application software, and take the development of generalization of quantum computing one step further.

Reference links:

[1]https://www.riverlane.com/press-release/riverlane-unveils-breakthrough-in-quantum-error-detection

[2]https://www.riverlane.com/press-release/riverlane-joins-7-5-million-consortium-to-build-error-corrected-quantum-processor

[3]https://www.riverlane.com/press-release/riverlane-appoints-leading-scientist-dr-earl-campbell-to-accelerate-efforts-to-solve-quantum-error-correction

[4]https://www.riverlane.com/press-release/riverlane-and-rigetti-computing-launch-partnership-to-tackle-error-correction-on-superconducting-quantum-computers