ConiQ: Enabling Concatenated Quantum Error Correction on Neutral Atom Arrays
Pengyu Liu, Mingkuan Xu, Hengyun Zhou, Hanrui Wang, Umut A. Acar, Yunong Shi · 2025
Read ConiQ: Enabling Concatenated Quantum Error Correction on Neutral Atom Arrays by Pengyu Liu, Mingkuan Xu, Hengyun Zhou, Hanrui Wang, Umut A. Acar, Yunong Shi (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2508-05779.txt.
Introduction
Pengyu Liu and co-authors present ConiQ, a compiler for quantum error correction on neutral atom arrays, in a 2025 paper. Layered codes, in which small codes nest inside larger ones, are very space efficient. Yet they lacked cheap logical gates, and they need heavy parallelism and long range links that strain most hardware.
Neutral atom arrays supply both traits. The compiler adds a gate method that uses code symmetries to cut overhead sharply against distillation based routes. A virtual atom layer lets it optimize level by level. Tests show huge cuts in space time cost and compile time against leading compilers. Layered codes look more practical for near term fault tolerant machines.
Catalog record
| CREATOR | Pengyu Liu, Mingkuan Xu, Hengyun Zhou, Hanrui Wang, Umut A. Acar, Yunong Shi |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 49 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2508-05779
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
ConiQ: Enabling Concatenated Quantum Error Correction on Neutral Atom Arrays Pengyu Liu, Mingkuan Xu, Hengyun Zhou, Hanrui Wang, Umut A. Acar, Yunong Shi Abstract Recent progress on concatenated codes, especially many-hypercube codes, achieves unprecedented space efficiency. Yet two critical challenges persist in practice. First, these codes lack efficient implementations of addressable logical gates. Second, the required high degree of parallelism and long-range interactions pose significant challenges for current hardware platforms. In this paper, we propose an efficient compilation approach for concatenated codes, specifically many-hypercube codes, targeted at neutral atom arrays, which provide the necessary parallelism and long-range interactions. Our approach builds on two key innovations. First, we introduce Automorphism-assisted Hierarchical Addressing (AHA) logical CNOT gates that significantly reduce spacetime overhead compared to conventional distillation-based methods. Second, we develop Virtual Atom Intermediate Representation (VAIR) that enables level-wise optimization and legalization. We implement these innovations in ConiQ, a hardware-aware quantum compiler designed to compile fault-tolerant quantum circuits for neutral atom arrays using many-hypercube codes. Our evaluation demonstrates that ConiQ achieves up to 2000 × 2000\times reduction in spacetime overhead and up to 10 6 × 10^{6}\times reduction in compilation time compared to state-of-the-art compilers, with our AHA gates providing an additional overhead reduction of up to 20 × 20\times . These results establish concatenated codes as a promising approach for fault-tolerant quantum computing in the near future. I Introduction Quantum computing inherently suffers from noise, making quantum error correction (QEC) indispensable for practical quantum computation. Surface codes have long been the leading candidate due to their high error threshold, but their low encoding rate imposes substantial qubit overhead. For instance, achieving a 10^{-9} logical error rate at a physical error rate of 10^{-3} requires a surface code of distance 15, incurring 225\times space overhead [ 12 ] . Quantum Low-Density Parity Check (qLDPC) codes [ 26 , 9 , 18 ] , despite promising asymptotic properties, face significant near-term limitations, including poor hardware adaptability and insufficient understanding of practical logical gate implementations [ 16 ] . Fig. 1: Overview of ConiQ and its key innovations: (1) The Automorphism-assisted Hierarchical Addressing (AHA) scheme for efficient individually addressable logical gates. (2) The Virtual Atom Intermediate Representation (VAIR) for efficient optimization and legalization across concatenation levels. Recently, concatenated QEC codes have emerged as compelling alternatives due to their impressive encoding rates. For example, many-hypercube codes [ 14 ] represent a family of QEC codes constructed by recursively concatenating quantum error-detecting codes with minimal sizes and high qubit efficiency, such as the [[6,4,2]] codes (encoding 4 logical qubits into 6 physical qubits with distance 2). This approach enables a substantial reduction in physical qubit overhead. Four levels of concatenation yield an encoding rate of (4/6)^{4}\approx 20\% , allowing many-hypercube codes to encode 1 logical qubit using approximately 5 physical qubits on average at a code distance of 16, significantly outperforming both surface codes and qLDPC codes. However, two critical obstacles hinder the practical deployment of concatenated codes. First, current proposals rely on distillation for individually addressable logical gates, particularly CNOT gates, which incur considerable spacetime overhead [ 14 ] . Second, the requirements for long-range interaction and high parallelism pose substantial challenges for current hardware and compilers. These limitations have thus far restricted concatenated codes to theoretical study, impeding their application in real-world quantum systems. The second challenge can potentially be addressed by leveraging neutral atom arrays as the hardware platform. Neutral atom arrays offer numerous advantages for fault-tolerant quantum computing: recent experiments demonstrate their long coherence times, impressive scalability, and low-error operations [ 24 , 6 , 17 , 10 ] . Most importantly, the unique capability of neutral atom arrays to dynamically reposition qubits during computation provides long-range interactions and extensive parallelism—properties particularly well-suited for concatenated codes. However, the long-range interactions and parallelism are in a restricted form, and require special compilation techniques to exploit the capabilities provided by neutral atom arrays. Although compilers for neutral atom arrays such as Atomique [ 31 ] and Enola [ 30 ] exist, they focus on the compilation of general quantum circuits, and result in prohibitive overheads in the context of concatenated codes. To harness the advantages of concatenated QEC codes and neutral atom arrays, we introduce ConiQ, a novel compiler framework specifically designed for efficiently implementing concatenated QEC codes on neutral atom arrays. ConiQ builds upon two key innovations with an overview shown in Fig. 1 : (1) By leveraging code symmetry and concatenation, we develop an Automorphism-assisted Hierarchical Addressing (AHA) logical CNOT gate scheme that is individually addressable, requiring only a few error correction cycles of overhead, dramatically improving efficiency compared to prior distillation-based schemes. (2) We introduce a Virtual Atom Intermediate Representation (VAIR) that virtualizes logical registers as physical atoms across concatenation levels. Crucially, we prove that VAIR preserves consistent program states, instruction sets, and hardware constraints at each concatenation level, enabling efficient level-wise legalization and optimization. Compared to the state-of-the-art compilers Atomique [ 31 ] and Enola [ 30 ] with prior distillation-based CNOT gates, ConiQ achieves over 10^{4}\times improvement in spacetime overhead and 10^{6}\times reduction in compilation time. Thus, ConiQ realizes the potential of high-encoding-rate quantum error correction with dramatically reduced overhead, enabling practical deployment of concatenated QEC codes on near-term neutral atom quantum computers. II Background II-A Neutral Atom Arrays Neutral atom arrays possess the unique capability to move qubits during computation, providing effective long-range connectivity and massive parallelism. Here we formally describe the neutral atom array architecture, program states, instruction set, and constraints at the physical level. In Section VI , we extend these concepts to the logical level and design the Virtual Atom Intermediate Representation (VAIR) for efficiently compiling concatenated codes. Architecture. Neutral atom arrays utilize individually-focused laser spots to trap and manipulate atoms. Two types of traps are typically employed: fixed spatial light modulators (SLM) and movable acousto-optic deflectors (AOD). Different lasers are typically used for SLM and AOD traps, so different type of traps won’t interfere with each other. Each type of trap forms a two-dimensional square grid. Atoms can reside in either type of laser trap. AOD atoms can be dynamically repositioned by adjusting the corresponding lasers. Conceptually, this configuration creates two planar atom layers: a stationary SLM layer and a movable AOD layer, slightly offset along the z axis. Fig. 2: Neutral atom array architecture, instructions, and constraints: (a) An allowed move operation (left) and a forbidden move operation that violates the order-preserving constraint (right). (b) An allowed transfer operation affecting entire rows and columns (left) and a forbidden operation that violates the row-column constraint (right). Program State Representation. We represent the collective system state, particularly atom positions, as a four-tuple (I,J,A,S) : • I : Ordered x-coordinates of AOD traps. I[i] denotes the x coordinate of the i -th column, where I[i]<I[i^{\prime}] for i<i^{\prime} . • J : Ordered y-coordinates of AOD traps. J[j] denotes the y coordinate of the j -th row, where J[j]<J[j^{\prime}] for j<j^{\prime} . • A : Positions of AOD atoms, where A[i][j] identifies the atom at the i -th column and j -th row, located at (I[i],J[j]) or is empty (\emptyset) when no atom is present. • S : Positions of SLM atoms on an integer grid, where S[i][j]