Simulating Circuit Layout for Distributed Quantum Computing
Extracted source text · arXiv 2512.21403
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% File: OPTICAmeetings.tex %
% Date: 31 July 2023 %
% %
% For preparing LaTeX manuscripts for submission %
% submission to Optica meetings and conferences %
% %
% (c) 2018-2022 Optica Publishing Group %
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\documentclass[letterpaper,10pt]{article}
%% if A4 paper needed, change letterpaper to A4
\usepackage{opticameet3} %% use version 3 for proper copyright statement
%% provide authormark
\newcommand\authormark[1]{\textsuperscript{#1}}
%% standard packages and arguments should be modified as needed
\usepackage{amsmath,amssymb}
\usepackage[colorlinks=true,bookmarks=false,citecolor=blue,urlcolor=blue]{hyperref} %pdflatex
%\usepackage[breaklinks,colorlinks=true,bookmarks=false,citecolor=blue,urlcolor=blue]{hyperref} %latex w/dvipdf
\usepackage{xcolor}
% \usepackage[colorinlistoftodos,textsize=small]{todonotes}
\newcommand{\todo}[1]{\textcolor{red}{[TODO: #1]}}
\setlength{\abovecaptionskip}{0.5pt}
\setlength{\belowcaptionskip}{0.5pt}
\setlength{\floatsep}{3pt}
\setlength{\textfloatsep}{3pt}
\setlength{\dbltextfloatsep}{8pt}
\setlength{\dblfloatsep}{0pt}
\setlength{\intextsep}{0pt}
\begin{document}
\title{Simulating Circuit Layout for Distributed Quantum Computing}
% \author{Author name(s)}
% \address{Author affiliation and full address}
% \email{e-mail address}
\vspace{-15pt}
\author{Sen Zhang\authormark{1}, Yipei Liu\authormark{1}, Brian Mark\authormark{1}, Weiwen Jiang\authormark{1}, Zebo Yang\authormark{2}, Lei Yang\authormark{1}}
\address{\authormark{1}George Mason University, Fairfax, VA 22030, USA \ \ \ \ \authormark{2}Florida Atlantic University, Boca Raton, FL 33431, USA}
% \\
% \authormark{2}ITS Department, George Mason University, Fairfax, VA 22030, USA\\
% \authormark{3}EECS Department, Florida Atlantic University, Boca Raton, FL 33431, USA}
% \author{Author One,\authormark{1} Author Two,\authormark{2,*} and Author Three\authormark{2,3}}
% \address{\authormark{1}Publications Department, Optica, 2010 Massachusetts Avenue NW, Washington, DC 20036, USA\\
% \authormark{2}Peer Review, Publications Department, Optica, 2010 Massachusetts Avenue NW, Washington, DC 20036, USA\\
% \authormark{3}Currently with the Meetings Department, Optica, 2010 Massachusetts Avenue NW, Washington, DC 20036, USA}
% \email{\authormark{*}cstech@optica.org} %% email address is required
% \email{\authormark{*}} %% email address is required
%% Do not add a copyright statement. Optica will add it.
\vspace{-10pt}
\begin{abstract}
The proposed framework represents the first tool to compile a quantum circuit across photonic-connected distributed quantum processors. Its design follows a divide-and-conquer paradigm for circuit partitioning, transpilation, and assembly, producing simulatable and implementable circuit layouts.
% SimDisQ-Layout is a fully automated layout simulator that performs heterogeneous compilation across distributed Quantum Processing Units (QPUs), integrating circuit scheduling and execution to generate implementable layouts for Distributed Quantum Computing (DQC). It serves as a core module within our all-in-one DQC simulator, SimDisQ.
%SimDisQ is a fully automated layout simulator that orchestrates the scheduling of circuits across QPUs prior to heterogeneous compilation, resulting in implementable layouts for Distributed Quantum Computing.
\vspace{-15pt}
\end{abstract}
\section{Introduction}
Current Quantum Processing Units (QPUs), such as those from IBM, Regitti, IonQ, Quantinuum, QuEra, and Infleqtion, lack the capacity to support the scale and complexity of quantum circuits (in particular, for the fault-tolerant quantum computing circuits) required for practical applications, challenging scalability and reliability \cite{cuomo2020towards}. Monolithically scaling a QPU is extremely difficult due to fabrication limits and errors. Distributed Quantum Computing (DQC) offers a promising solution by interconnecting multiple QPUs through quantum photonic links and optical switching networks to collaboratively solve a single problem at scale \cite{8910635}. The heterogeneity of distributed QPUs, across superconducting, trapped-ion, and photonic platforms, together with the complexity of optical links and quantum repeaters, introduces significant challenges for DQC system design and verification \cite{CALEFFI2024110672}.
Given the rapid progress in DQC, it is essential to initiate research efforts during the interim period while the hardware ecosystem continues to mature.
However, the field lacks a comprehensive validation platform to support the development and testing of DQC strategies, as well as the benchmarking quantum algorithms in distributed settings.
The cornerstone for quantum validation platforms or simulator is the quantum circuit layout \cite{tan2020optimal}, which specifies how logical qubits and gates in a circuit are mapped and arranged on given quantum hardware.
Unlike targeting a single QPU, the layout across distributed QPUs faces a set of new needs and challenges.
\textit{First}, the distributed quantum circuit layout requires a circuit-level modeling for the communication (more specifically, the two-qubit gates) between a pair of QPUs.
\textit{Second}, it requires QPU backend-specified compilation, as heterogeneous QPUs have different connectivity topologies and basis gates for implementation.
\textit{Moreover}, due to inter-QPU communication, the compilations among QPUs require synchronization to ensure the correctness of the function.
All the above challenges cannot be simply addressed by revising an existing quantum circuit layout for a single QPU.
% A quantum circuit layout specifies how logical qubits and gates in a circuit are mapped and arranged on a given quantum hardware topology.
To address this pressing issue, we introduce our framework, the first automated layout compilation tool that accounts for the heterogeneity of QPUs in DQC, providing an end-to-end automated compilation to generate the layout of a given quantum circuit across distributed QPUs, which can be both simulatable and implementable.
The design of it follows \textbf{the divide-and-conquer paradigm}.
% distributed quantum circuit layout and execution.
% which is the cornerstone for DQC circuit simulator design and validation platform build-up.
% It provides automated distributed execution of quantum circuits and generates preliminary benchmarking samples. SimDisQ-Layout serves as a core module of an upcoming all-in-one DQC simulator, SimDisQ, that enables a fully automated workflow for distributed circuit execution, spanning from network-layer strategies (e.g., circuit optimization, partitioning, and scheduling) to physical-layer processes such as compilation, transpilation, qubit assignment, and execution. SimDisQ provides a foundation for testing and validating designs ahead of future testbed implementations and hardware rollouts.
% Built on Qiskit \cite{qiskit2024}, SimDisQ-Layout provides end-to-end compilation for distributed quantum circuit layout and execution.
Specifically, in the divide step, it begins by partitioning a quantum circuit into subcircuits, which are then mapped to a predefined QPU network topology.
Then, in the conquer step, each subcircuit is independently compiled according to the hardware specifications of its assigned QPU, such as qubit connectivity and supported gate sets.
Finally, in the combine step, following the compiled subcircuits, an automated placement algorithm will be developed to assemble these subcircuits with additional remote quantum gates to generate a fully instantiated circuit, where
% the compiled subcircuits are assembled into a full quantum circuit by carefully adding remote quantum gates to ensure correct synchronization
% , and later integrated into a global layout as a unified compiled circuit.
% In addition,
Our framework seamlessly manages entanglement distribution and remote gate execution through quantum networking and TeleGate \cite{PhysRevA.62.052317,ferrari2021compiler}.
The generated quantum circuit layout can enable users to simulate a quantum circuit across multiple heterogeneous backends to assess circuit performance and correctness under realistic DQC conditions; meanwhile, it is also implementable if Einstein–Podolsky–Rosen (EPR) pairs can be established on physical quantum computers.
%As the number of qubits and the complexity of quantum operations continue to grow, single quantum processing units (QPUs) face increasing challenges in scalability and reliability\cite{cuomo2020towards}. Distributed Quantum Computing (DQC) offers a promising solution by interconnecting multiple QPUs to enable sharing of quantum states and cooperative computation through the quantum internet\cite{8910635}. However, the heterogeneity of distributed systems, combined with the complexity of entanglement generation and classical communication latency, poses significant challenges to system design and verification\cite{CALEFFI2024110672}. Consequently, developing a simulator that accurately models and evaluates distributed quantum architectures is essential to advance research in this field.
%To address this need, SimDisQ is developed as an automatic simulator for heterogeneous distributed quantum computing based on the IBM Qiskit\cite{qiskit2024} framework. It provides end-to-end automation of distributed circuit layout and execution. In SimDisQ, subcircuits are compiled independently and later integrated, while classical communication and entanglement resource generation and management are automatically handled by the framework through TeleGate\cite{PhysRevA.62.052317,ferrari2021compiler}—capabilities beyond the native support of Qiskit. This enables users to interconnect multiple heterogeneous backends and evaluate the performance and correctness of circuits under near-realistic heterogeneous DQC conditions.
\begin{figure}[t]
\centering
\includegraphics[width=1\textwidth]{PNG/OFC_A1.eps}
\caption{End-to-End Design Flow and Architecture of Our Framework.}
\label{fig:workflow}
\end{figure}
\begin{figure}[b!]
\centering
\includegraphics[width=1\textwidth]{PNG/GHZ_6.eps}
\caption{Example of Distributed Layout Compilation for GHZ6 Circuit using Our Framework.}
\label{fig:layoutCompilation}
\end{figure}
\section{Proposed Framework}
Figure~\ref{fig:workflow} illustrates the end-to-end architecture of our framework.
Given a set of QPUs connected via photonic channels, a monolithic quantum circuit, and a set of user-specified qubits partitions, as shown in the Input in the figure, our framework can automatically transform the quantum circuit into a fully simulatable and implementable distributed layout, as shown in the Output of the figure.
In addition, a set of associated metrics, such as different types of qubits and circuit depth, will be reported in the output.
% with user-specified qubits partitions
% by integrating backend-aware compilation, remote gate management, and performance evaluation.
As introduced, the workflow of our framework follows the divide-and-conquer paradigm, including four main stages, detailed below.
\textbf{\ding{172} Circuit Partitioning and Logical Mapping (pre-processing)}:
The given circuits can contain different types of multi-qubit gates, such as Toffoli and CZ gates. However, the TeleGate execution can only support a limited set of gates, such as a remote CNOT gate.
Therefore, the pre-processing step is to use the partitions given by the user to convert multi-qubit gates across partitions to supported remote gates, as shown in the red dashed box in Figure \ref{fig:workflow} \ding{172}.
Based on the remote gate conversion, this step further parses each partition sub-circuit into a logical group with additional communication qubits.
% multiple groups
% The input quantum circuit, along with the properties of available QPUs, is parsed and partitioned into logical subcircuits. Multi-qubit gates that span across QPU groups are identified and converted to CNOT gates (e.g., unitaries + CNOT) to facilitate TeleGate execution. The associated qubits are categorized into data qubits and communication qubits.
These groupings define the logical boundaries for distributed execution.
% and remote gate handling.
\textbf{\ding{173} Scheduling and Pre-Compilation (divide)}:
With the logical groups, this divide step aims to isolate the subcircuits onto different QPUs, such that they can be compiled independently on QPUs in the next step.
This is motivated by two facts: (1) existing compilation tools do not support remote gates; and (2) each QPU may have different capabilities (e.g., qubits, connectivity), which prefer independent compilation.
To achieve this, we first assign subcircuits to specific QPUs with consideration of resource limitations.
Then,
% Each QPU may have different capabilities (e.g., number of qubits, connectivity, noise model), and the scheduling module maps subcircuits accordingly.
remote TeleGates are explicitly scheduled to bridge operations between QPUs, including the establishment of EPR pairs between QPUs, coordinating classical communication, and resetting qubits post-execution.
Finally, to isolate different sub-circuits, we replace remote gates with placeholders, which ensures they will not be compiled with other local gates.
% in the next step but will be used to assemble subcircuits in the final step.
% To ensure function correctness, we also need to prevent placeholders from being involved in gate reordering for optimization, which is implemented by adding barrier gates around each placeholder.
% , enabling independent compilation of subcircuits across different backends. In subsequent stages, these subcircuits are compiled separately and later integrated for execution.
\textbf{\ding{174} Backend-Aware Compilation (conquer)}: This conquer step is to perform backend-aware transpilation of a subcircuit based on its assigned QPU's characteristics, including native gate sets and qubit topology. It ensures that the compiled circuit segments are executable on heterogeneous backends, including real devices (e.g., IBM, IonQ, or others) or simulated devices (e.g., FakeVigo).
% The pre-compilation step in the previous stage ensures that subcircuits can be compiled independently without compromising the correctness of the overall circuit.
\textbf{\ding{175} Subcircuit Assembly (combine)}: This is the key step to establish
% the
% final The transpiled subcircuits are integrated into
a fully instantiated distributed circuit.
This assembly process leverages the placeholders in step \ding{173} that
% to
% is guided by the global execution context established during the pre-compilation phase. Specifically, the placeholders that were previously inserted to
represent cross-QPU interactions, which are now resolved back into remote gate operations, i.e., TeleGate. These remote gates restore the logical dependencies across subcircuits, reconnecting the independently compiled subcircuits into a coherent global circuit.
%\subsection{Overview}
% All PDF submissions must contain the following items in order to be published:
% \begin{enumerate}
% \item Complete title
% \item Complete listing of all authors and their affiliations
% \item Self-contained 35-word abstract (indexers such as Google Scholar will not index papers that do not contain abstracts)
% \item Do not include a copyright statement following the abstract. Optica will add the copyright statement to the footer in the PDF file, e.g., This Article \textcopyright \number\year \hskip.05in The Author(s).
% \item Permission and attribution for any trademarked or copyright images. Note that images of people or images owned or trademarked by other entities (including well-known logo's or cartoon characters for example) will also require official written permission.
% \item Please include consecutive callouts to figures, equations, tables, and references in the text
% \item Two-page limit unless designated otherwise on conference website
% \end{enumerate}
%\subsection{Pre-Processing and Logical Distribution}
%\subsection{Remote Gate Establishment}
%\subsection{...}
% Margins and type size will be set by the Optica \LaTeX{}
% commands for title, author names and addresses, abstract,
% references, captions, and so on. The \texttt{opticameet3.sty} package
% references \texttt{mathptmx.sty} for Times text and math fonts.
% Authors who require Computer Modern font may modify the style file
% or, preferably, invoke the package \texttt{ae.sty} or similar for
% optimum output with Computer Modern.
% \subsection{Author Names and Affiliations}
% Author names should be given in full with first initials spelled out to assist with indexing.
% Affiliations should follow the format division, organization, and address---and complete postal information should be given.
% Abbreviations should not be used. United States addresses should end
% with ``, USA.''
% \subsection{Abstract} The abstract
% should be limited to no more than 35 words. It should be an
% explicit summary of the paper that states the problem, the methods
% used, and the major results and conclusions. If another publication author is referenced in the abstract, abbreviated information
% (e.g., journal, volume number, first page, year) must be
% given in the abstract itself, without a reference number. (The item referenced in the abstract should be the first
% cited reference in the body.)
% \subsection{Notation}
% \subsubsection{General Notation}
% Notation must be
% legible, clear, compact, and consistent with standard usage. In
% general, acronyms should be defined at first use.
% \subsubsection{Math Notation}
% Equations should use standard \LaTeX{} or AMS\LaTeX{} commands (sample from Krishnan \textit{et al.} \cite{krishnan00}).
Figure~\ref{fig:layoutCompilation} presents an example of compilation process applied to a 6-qubit GHZ state generation circuit. It demonstrates how a monolithic quantum circuit is transformed into a distributed execution layout across multiple QPUs, each with its own local qubit resources and communication channels. The left panel shows the inputs, including a standard 6-qubit GHZ circuit, the user-specified partition and a set of QPUs.
% which serves as a minimal but nontrivial benchmark for distributed compilation.
The right panel shows the resulting distributed quantum circuit layout over three QPUs labeled $Q_0$, $Q_1$, and $Q_2$, each with distinct data and communication qubits. Each QPU handles a partitioned subcircuit, with computation localized to available data qubits. Remote CNOTs based on Bell pairs ($\left|\Phi^+\right\rangle$ and helical wires) bridge operations across QPUs. Native gates (e.g., $R_z$, $\sqrt{X}$) are generated where appropriate.
% \textbf{Execution and Evaluation}: The global circuit is then executed, and the results are collected for fidelity analysis and performance evaluation. Users can analyze correctness, circuit characteristics, and scalability of DQC executions.
\section{Simulation and Performance Results}
\begin{table*}[t]
\centering
\scriptsize
\caption{Performance Evaluation of Distributed Quantum Circuit Benchmarks.}
\label{tab:ComparisonOfQuantumCircuitSets}
\renewcommand{\arraystretch}{1.08}
\setlength{\tabcolsep}{0.92pt}
\begin{tabular}{|c|ccc|ccc|c|c|c|ccc|c|c|c|c|c|c|}
\hline
\multirow{3}{*}{\textbf{Benchmark}}
& \multicolumn{6}{c|}{\textbf{Input}}
& \multicolumn{8}{c|}{\textbf{Circuit Metrics}}
& \multicolumn{4}{c|}{\textbf{Fidelity}} \\
\cline{2-19}
& \multicolumn{3}{c|}{Partition}
& \multicolumn{3}{c|}{Assignment}
& \multirow{2}{*}{\#QData}
& \multirow{2}{*}{\#QComm}
& \multirow{2}{*}{\#QTotal}
& \multicolumn{3}{c|}{Sub-Circ Depth}
& Layout
& Gate
& \multirow{2}{*}{State$^{*}$}
& \multirow{2}{*}{IProb}
& \multirow{2}{*}{EProb}
& \multirow{2}{*}{Error Rate} \\
\cline{2-4} \cline{5-7} \cline{11-13}
& P0 & P1 & P2
& P0 & P1 & P2
& & &
& Min & Max & Avg
& Depth & Count & & & & \\
\hline
GHZ-6 & q0-q1 & q2-q3 & q4-q5 & Q0 & Q1 & Q2 & 6 & 4 & 14 & 5 & 10 & 7.33 & 12 & 30 & $|0\rangle^{\otimes 6}$ & 0.5 & 0.50144 & $10^{-6}$ \\
GHZ-6 & q0-q1 & q2-q3 & q4-q5 & Q0 & Q2 & Q1 & 6 & 4 & 14 & 5 & 10 & 7.33 & 12 & 30 & $|0\rangle^{\otimes 6}$ & 0.5 & 0.49861 & $10^{-6}$ \\
GHZ-6 & q0 & q1-q2 & q3-q5 & Q0 & Q2 & Q1 & 6 & 4 & 14 & 5 & 10 & 7.67 & 13 & 30 & $|0\rangle^{\otimes 6}$ & 0.5 & 0.50106 & $10^{-6}$ \\
GHZ-12 & q0-q3 & q4-q7 & q8-q11 & Q0 & Q1 & Q2 & 12 & 4 & 16 & 7 & 11 & 9 & 17 & 42 & $|0\rangle^{\otimes 12}$ & 0.5 & 0.49863 & $10^{-6}$ \\
BitCode-3 & q0-q2 & q3-q4 & - & Q0 & Q1 & - & 5 & 4 & 9 & 9 & 15 & 12 & 14 & 40 & $|00100\rangle$ & 1 & 1 & 0 \\
BitCode-3 & q0 & q1-q4 & - & Q0 & Q1 & - & 5 & 4 & 9 & 9 & 17 & 13 & 16 & 40 & $|00100\rangle$ & 1 & 1 & 0 \\
TFIM Hamiltonian & q0 & q1-q2 & - & Q0 & Q1 & - & 3 & 12 & 15 & 28 & 43 & 35.5 & 43 & 102 & $|1\rangle^{\otimes 3}$ & 0.6 & 0.60458 & $10^{-5}$ \\
TFIM Hamiltonian & q0-q1 & q2 & - & Q0 & Q1 & - & 3 & 12 & 15 & 34 & 37 & 35 & 43 & 102 & $|1\rangle^{\otimes 3}$ & 0.6 & 0.60413 & $10^{-5}$ \\
Qaoa-4 & q0-q1 & q2-q3 & - & Q0 & Q1 & - & 4 & 16 & 20 & 36 & 44 & 40 & 45 & 102 & $|0001\rangle$ & 0.093612 & 0.094750 & $10^{-5}$ \\
Qaoa-6 & q0-q1 & q2-q3 & q4-q5 & Q0 & Q1 & Q2 & 6 & 48 & 54 & 67 & 83 & 75.33 & 122 & 279 & $|001101\rangle$ & 0.048330 & 0.048520 & $10^{-4}$ \\
Qaoa-8 & q0-q2 & q3-q5 & q6-q7 & Q0 & Q1 & Q2 & 6 & 84 & 90 & 123 & 140 & 130 & 205 & 486 & $|01110101\rangle$ & 0.028230 & 0.028270 & $10^{-3}$ \\
Qaoa-10 & q0-q3 & q4-q6 & q7-q9 & Q0 & Q1 & Q2 & 10 & 132 & 142 & 194 & 213 & 201.33 & 319 & 759 & $|0000001011\rangle$ & 0.007284 & 0.006950 & $10^{-3}$ \\
\hline
\end{tabular}
\vspace{2pt}
\noindent
\parbox{\linewidth}{%
\footnotesize ``*'': Due to space limitations, for each benchmark, we only present the result of the state with the highest probability in output distribution.}
\end{table*}
To validate the resultant layout, we implement our framework using IBM Qiskit \cite{qiskit2024}, and the layouts are simulated using IBM Aer simulator .
Table~\ref{tab:ComparisonOfQuantumCircuitSets} shows the results of different distributed quantum circuit benchmarks executed using our framework. We evaluate four representative quantum algorithms: the GHZ circuit for state preparation, BitCode circuit for syndrome measurement in bit-flip quantum error correction code, TFIM (Transverse Field Ising Models) circuit for Hamiltonian simulation, and the QAOA for optimization problems.
The circuits of BitCode, TFIM, and QAOA are obtained from the SupermarQ benchmark suite \cite{tomesh2022supermarq}.
% the BitCode with two auxiliary qubits, the 3-qubit 1D TFIM Hamiltonian, and the QAOA vanilla proxy algorithm with varying numbers of qubits.
In the table, the \textit{Input} columns list the user-defined partitions and backend assignments. Here, Q0, Q1, and Q2 correspond to the quantum backends FakeVigoV2, FakeAthensV2, and FakeLagosV2, respectively.
% The Circuit Metrics columns report several key indicators of the simulatable and implementable circuit layouts produced by SimDisQ-Layout.
The \textit{Circuit Metrics} columns summarize the key configurations of the circuit layouts.
% \textit{\#QData} represents the total number of data qubits.
\textit{\#QComm} denotes the number of communication qubits, which corresponds in practice to EPR pairs established through optical links.
% ; this reflects the quantum entanglement resources required under different partitions and highlights the importance of partitioning and QPU scheduling.
The \textit{Sub-Circ Depth} reports the depth of each subcircuit on the assigned QPU without remote gates, while the \textit{Layout Depth} is for the resultant layout.
% column shows the total depth of the simulatable (and implementable in practice) distributed layout, while
\textit{Gate Count} lists the number of quantum gates within the layout.
From the table, we observe that all circuit metrics grow proportionally, as the complexity of the quantum algorithm and the number of qubits increase. Notably, the total number of gates in the simulatable layouts generally exceeds that of the subcircuits, highlighting the additional overhead of distributed quantum computation.
For identical benchmarks, variations in partitioning strategies result in significant differences between the Sub-Circ Depth and the final Layout Depth. This behavior, observed in the BitCode, TFIM Hamiltonian, and GHZ circuits, underscores the critical role of circuit partitioning in the overall performance of distributed quantum computation.
% , and it will likely remain an essential direction for future research in this field.
% In the Sub-Circ Depth and Gate Count columns, we present the circuit depth and gate count after partitioning, assignment, and transpilation. For comparison, the total simulatable and implementable layouts are also shown. As the algorithm complexity and the number of participating qubits increase, all these metrics rise accordingly. Notably, the total number of gates in the simulatable and implementable layouts is generally higher than the sub-circuit depth, which demonstrates the advantage of synchronized distributed quantum computation. Furthermore, under certain identical benchmarks, varying the partition strategies leads to noticeable differences between the Sub-Circ Depth and the final Layout Circuit Depth. This phenomenon, as observed in experiments on BitCode, 1D TFIM Hamiltonian, and GHZ-6 circuits, demonstrates that circuit partitioning plays a crucial role in the performance of distributed quantum computation, and it will likely remain an essential direction for future research in this field.
The \textit{Fidelity} column reports the accuracy of the circuits generated by our framework.
The results show that after heterogeneous compilation, the layout fidelity remains nearly identical to the ideal state, validating the correctness of the our workflow.
% The State corresponds to the quantum state with highest probablity.
Here, \textit{IProb} denotes the ideal probability, while \textit{EProb} shows the experimentally obtained probability. The error rate is defined as $1 - f$, where $f$ is the Hellinger Fidelity, which quantifies the similarity between two probability distributions.
% , well within expected error margins.
For QAOA, the error rate increases with the number of qubits because larger circuits take a longer time to simulate with a larger number of measurement shots. With fewer shots, the output probabilities become less accurate, leading to higher measured error rates.
% However, if more experimental resources (such as additional shots or longer run times) are provided, SimDisQ-Layout can still achieve high fidelity even for larger QAOA circuits.
%For QAOA, the error rate increases with the number of qubits due to a reduction in the number of shots as the gate count grows in experiments; if the experimental constraints is addressed, the SimDisQ-Layout can still achieve high fidelity.
% the error rate decreases when the number of shots is increased.
\section{Conclusion}
Our framework achieves near-perfect results in the noiseless setting, demonstrating the effectiveness of distributed quantum execution across heterogeneous QPUs. As the first simulator designed for distributed quantum circuit layouts, it serves as a valuable development framework for DQC researchers to benchmark their own solutions and establishes a solid foundation for future research in this field.
%\section{References}
\bibliographystyle{opticajnl}
\bibliography{sample}
\end{document}