⚛️ quantum physics

Reuse-Aware Compilation for Zoned Quantum Architectures Based on Neutral Atoms

This paper introduces ZAC, a scalable compiler with qubit reuse and novel scheduling strategies for neutral-atom zoned architectures that significantly improves circuit fidelity by minimizing data movement overhead compared to monolithic designs.

Original authors: Wan-Hsuan Lin, Daniel Bochen Tan, Jason Cong

Published 2026-02-27

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Wan-Hsuan Lin, Daniel Bochen Tan, Jason Cong

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are trying to organize a massive, high-stakes dance competition in a giant warehouse. The dancers are qubits (the basic units of quantum computers), and the dance moves they need to perform are quantum gates.

In the world of neutral atom quantum computers, these dancers are trapped in invisible "cages" made of light. To make them dance together (perform a calculation), you have to bring two specific dancers close enough to touch.

The Problem: The "Flashlight" Nightmare

In the old way of doing things (called Monolithic Architecture), imagine the whole warehouse is lit up by a single, giant, blinding flashlight.

The Issue: When two dancers need to dance together, the flashlight turns on. But because the light covers the entire room, every single other dancer who is just standing there waiting (idling) gets blinded by the light.
The Result: This "blinding" causes the waiting dancers to get dizzy and make mistakes (errors). The more complex the dance routine, the more mistakes happen, and the whole performance falls apart.

The Solution: The "Zoned" Warehouse

The paper introduces a new way to build the warehouse, called a Zoned Architecture.

The Setup: Instead of one big room, the warehouse is split into three distinct zones:
1. The Dance Floor (Entanglement Zone): A small, controlled area where the flashlight is allowed to shine. Only the dancers who need to dance are here.
2. The Green Room (Storage Zone): A dark, quiet area where the other dancers wait. They are shielded from the blinding light, so they stay fresh and don't make mistakes.
3. The Exit (Readout Zone): Where the dancers leave to show the results.

This is a huge improvement! But it creates a new problem: Logistics.
Now, you have to constantly move dancers from the Green Room to the Dance Floor, make them dance, and then move them back. If you move them too much or too slowly, they get tired (decoherence) or drop out of the cage (atom loss).

Enter ZAC: The Super-Organized Choreographer

The paper presents ZAC, a new "compiler" (a smart software brain) that acts as the ultimate choreographer for this zoned warehouse. It solves the logistics nightmare with three clever tricks:

1. The "Stay Put" Strategy (Qubit Reuse)

In the old systems, if a dancer finished a move, they were immediately sent back to the Green Room, only to be dragged back out for the next move.

ZAC's Trick: ZAC looks ahead at the choreography. If a dancer is going to dance again immediately in the next step, ZAC says, "Don't send them back! Just keep them on the Dance Floor."
The Analogy: Imagine a relay race. Instead of running back to the starting line after every lap, the runner stays on the track if they are running the next leg immediately. This saves massive amounts of time and energy.

2. The "Smart Map" (Placement)

ZAC doesn't just throw dancers into random spots. It uses a super-smart algorithm (like a GPS for quantum dancers) to figure out exactly where to place them in the Green Room before the race starts.

The Goal: It places them so that when they need to run to the Dance Floor, they have the shortest possible path. It even plans for the future steps, ensuring that when they return, they are already close to where they need to be next.

3. The "Traffic Cop" (Load Balancing)

Imagine you have multiple moving trucks (called AODs) to transport the dancers.

The Problem: If you send all the trucks to move the same group of dancers at once, they might crash into each other. If you send them one by one, it takes forever.
ZAC's Trick: ZAC acts like a traffic cop. It breaks the moves into small "jobs" and assigns them to different trucks so they can work in parallel without crashing. It ensures the trucks are always busy but never blocking each other.

The Results: A 22x Improvement

The paper tested ZAC against the old "blinding flashlight" method and other new methods.

The Score: ZAC made the quantum computer 22 times more accurate (higher fidelity) than the old single-room method.
Why? Because it kept the "dizzy" errors (from the blinding light) to almost zero by keeping dancers in the dark Green Room, and it minimized the "tiredness" errors by moving them efficiently.
Near Perfect: ZAC is so good that it is only 10% away from the "theoretical perfect" solution, which is incredibly hard to achieve in the real world.

Why Does This Matter?

Quantum computers are currently very fragile; they make mistakes easily. This paper shows that by changing how we organize the hardware (the zones) and how we manage the software (the ZAC compiler), we can build quantum computers that are much larger, much more reliable, and capable of solving real-world problems like drug discovery or climate modeling.

In short: ZAC turns a chaotic, error-prone dance floor into a well-oiled, high-speed assembly line, allowing quantum computers to finally dance to their full potential.

1. Problem Statement

Neutral atom quantum computing offers scalability and long coherence times, but current compilation strategies face significant fidelity bottlenecks:

Monolithic Architecture Limitations: In traditional monolithic architectures, a global Rydberg laser illuminates the entire chip to perform two-qubit (2Q) gates. This causes "side-effect noise," where idle qubits are inadvertently excited, leading to significant decoherence and fidelity loss, especially in deep circuits.
Zoned Architecture Challenges: While "zoned" architectures (separating storage, entanglement, and readout zones) mitigate side-effect noise by shielding idle qubits, they introduce new compilation challenges:
- Data Movement Overhead: Qubits must be physically moved between storage and entanglement zones, incurring errors from atom transfer and decoherence during transit.
- Inefficient Reuse: Existing compilers (e.g., NALAC) either fail to fully utilize qubit reuse (keeping qubits in the entanglement zone for subsequent gates) or do so at the cost of increased movement distances and exposure to errors.
- Scalability: Current tools struggle with advanced configurations involving multiple Acousto-Optic Deflector (AOD) arrays and complex zone layouts.

2. Methodology: The ZAC Compiler

The authors propose ZAC (Zoned Architecture Compiler), a scalable compilation framework designed specifically for zoned neutral atom architectures. The workflow consists of three main stages:

A. Preprocessing

Resynthesizes the input circuit into the hardware-supported gate set ( $\{CZ, U3\}$ ).
Optimizes single-qubit (1Q) gates and schedules 2Q gates into "Rydberg stages" (layers where 2Q gates are executed in parallel).

B. Reuse-Aware Placement

This is the core innovation of ZAC, aiming to minimize movement overhead while maximizing qubit reuse.

Qubit Reuse Strategy: If a qubit is involved in a gate in the next Rydberg stage, ZAC attempts to keep it in the current Rydberg site rather than moving it back to storage.
- Conflict Resolution: Uses a bipartite graph and the Hopcroft-Karp algorithm to find the maximum matching of reusable qubits between consecutive stages, resolving conflicts where multiple gates compete for the same site.
Dynamic Placement & Cost Function:
- Initial Placement: Uses Simulated Annealing (SA) to find an optimal initial mapping of qubits to storage traps. The cost function estimates movement duration based on Euclidean distance, accounting for parallel movement capabilities (e.g., if qubits are in the same row, they move simultaneously).
- Lookahead Cost: The cost function includes a "lookahead" term to minimize movement for future stages, not just the immediate next one.
- Gate & Non-Reuse Placement: For each stage, ZAC generates two solutions (one with reuse, one without) and selects the one with the lower total movement cost. It uses minimum weight full matching (Jonker-Volgenant algorithm) to assign gates to Rydberg sites and non-reused qubits back to storage traps.

C. Load-Balancing Scheduling

Job Generation: Converts placement changes into "rearrangement jobs" (moving qubits between zones).
Dependency Handling: Identifies Trap Dependencies (a trap must be vacated before another qubit enters) and Qubit Dependencies (a qubit cannot be moved while being operated on).
Multi-AOD Scheduling: If the architecture has multiple AODs, ZAC employs a load-balancing algorithm. It assigns the longest rearrangement jobs to the earliest available AOD to minimize total execution time and prevent bottlenecks.

3. Key Contributions

ZAC Compiler: A novel compiler supporting advanced zoned architectures with multiple AODs and flexible zone configurations.
Reuse-Aware Placement: A strategy that keeps qubits in the entanglement zone if they are immediately needed, significantly reducing atom transfer errors without increasing excitation errors.
Optimized Cost Functions: A movement cost model that accounts for parallel execution and lookahead optimization for future circuit stages.
ZAIR (Intermediate Representation): A new IR that abstracts complex rearrangement operations into "jobs," facilitating the management of multiple AODs and bridging high-level placement with low-level machine instructions.
Architecture Specification: A flexible specification language defining AOD arrays, SLM arrays, and zones, enabling the design of scalable and customized layouts.
FTQC Support: Demonstrated capability to compile circuits for Fault-Tolerant Quantum Computing (FTQC), specifically handling transversal gates on logical code blocks (e.g., [[8,3,2]] codes).

4. Evaluation Results

The authors evaluated ZAC against state-of-the-art compilers (NALAC, Enola, Atomique) and superconducting baselines (IBM Heron, Google Sycamore) using benchmarks from QASMBench.

Fidelity Improvement:
- vs. Monolithic: ZAC achieves a 22 $\times$ improvement in fidelity over monolithic architectures (e.g., Enola) by eliminating side-effect noise.
- vs. Superconducting: ZAC shows a 1.56 $\times$ to 2.33 $\times$ fidelity improvement over superconducting platforms due to the superior coherence time of neutral atoms, despite longer execution times.
- vs. Zoned Baselines: ZAC delivers a 4 $\times$ fidelity improvement over NALAC by optimizing qubit reuse and reducing unnecessary movements.
Optimality Gap: ZAC operates within a 10% fidelity gap compared to an idealized "perfect reuse" solution, demonstrating near-optimal performance.
Efficiency:
- ZAC reduces circuit duration by 10% compared to Atomique and 55% compared to NALAC.
- Compilation time is comparable to or faster than existing tools, with a 63 $\times$ speedup over NALAC when initial mapping optimization is disabled.
Scalability & Flexibility:
- Multi-AOD: Utilizing two AODs improves fidelity by ~10% through increased parallelism.
- Multi-Zone: ZAC successfully supports architectures with multiple entanglement zones, showing a 15% fidelity gain for highly parallel circuits (e.g., Ising model) by reducing movement distances.

5. Significance

This work represents a critical step toward scalable, high-fidelity neutral atom quantum computing.

Bridging the Gap: It solves the fundamental trade-off in zoned architectures between shielding idle qubits (fidelity) and moving them (overhead) by introducing intelligent reuse strategies.
Enabling FTQC: By supporting transversal gates and logical block movements, ZAC provides the necessary compilation infrastructure for future fault-tolerant quantum algorithms.
Open Source: The release of ZAC and its specification language (ZAIR) allows the community to explore diverse architectural designs, accelerating the development of next-generation quantum processors.

In summary, ZAC transforms zoned neutral atom architectures from a theoretical concept into a practical, high-performance computing platform, achieving near-optimal fidelity and enabling complex quantum algorithms that were previously infeasible due to noise and compilation inefficiencies.