Quantum Error Mitigation Protocols for Near-Term Chemistry Simulations: A Comprehensive Guide

Abstract

This article provides a comprehensive overview of quantum error mitigation (QEM) protocols essential for performing reliable quantum chemistry simulations on today's noisy intermediate-scale quantum (NISQ) devices. Aimed at researchers, scientists, and professionals in drug development, we explore the foundational principles of QEM, detail cutting-edge methodologies like probabilistic error cancellation and Clifford data regression, and present optimization strategies to enhance their efficacy. Through a comparative analysis of protocol performance on molecular systems such as Hâ‚‚, Hâ‚‚O, and Nâ‚‚, we validate these techniques and discuss their critical role in accelerating the discovery of new pharmaceuticals and materials by enabling more accurate quantum computations of molecular properties.

Understanding Quantum Noise and the Need for Error Mitigation in Chemistry Simulations

The Noisy Intermediate-Scale Quantum (NISQ) era, a term coined by John Preskill, is defined by quantum processors containing up to approximately 1,000 qubits that operate without the benefit of full quantum error correction [1]. These devices are characterized by inherent noise sensitivity and proneness to quantum decoherence, which severely limits the depth and complexity of computations that can be reliably executed [1]. For researchers in quantum chemistry and drug development, this presents both an unprecedented opportunity and a significant challenge. Quantum computers offer the potential to solve electronic structure problems and simulate molecular systems with complexity far beyond the reach of classical computational methods. However, extracting chemically meaningful results—particularly those meeting the standard of chemical accuracy (1.6 × 10⁻³ Hartree, or approximately 1 kcal/mol)—requires sophisticated strategies to manage the hardware limitations endemic to current devices [2] [3].

The core challenge lies in the exponential scaling of quantum noise. With typical two-qubit gate error rates ranging from 1% to 5% on current hardware, quantum circuits can only execute approximately 1,000 gates before noise overwhelms the signal [1]. This constraint directly impacts the feasibility of running Variational Quantum Eigensolver (VQE) algorithms for molecular energy estimation, as the required circuit depths for interesting chemical systems often approach or exceed this threshold [2]. Furthermore, the limited coherence times of qubits—typically lasting for tens to hundreds of microseconds—impose a strict temporal bound on computation, while measurement errors and crosstalk introduce additional sources of inaccuracy [4] [5]. Understanding and mitigating these effects is not merely an academic exercise but a practical necessity for researchers aiming to leverage NISQ hardware for computational chemistry and pharmaceutical development.

Quantitative Analysis of NISQ Hardware Limitations

Current Hardware Performance Specifications

The performance of NISQ devices is quantified by several key metrics that directly impact the fidelity of quantum chemistry simulations. The table below summarizes representative specifications for leading quantum hardware platforms, illustrating the current landscape of available resources.

Table 1: Representative Performance Metrics of Current NISQ Hardware

Hardware Platform/Type	Qubit Count	Single-Qubit Gate Fidelity	Two-Qubit Gate Fidelity	Readout Fidelity	Coherence Time (T1/T2, μs)
Superconducting (e.g., IBM)	50-1000+	99.95%	99.0-99.5%	97-99%	100-500
Trapped Ion (e.g., Quantinuum)	20-50	99.99%	99.5-99.9%	>99%	10,000+
Neutral Atom (e.g., Atom Computing)	100-1200	>99.9%	>99.5%	>98%	1000+

These specifications, derived from current industry capabilities, highlight the fundamental constraints facing NISQ-era quantum chemists [1] [6] [7]. The gate infidelities, though seemingly small, accumulate rapidly as circuit depth increases. For a quantum circuit with 1,000 two-qubit gates, even a 99.5% fidelity per gate would result in a total circuit fidelity of less than 1% (0.995¹⁰⁰⁰ ≈ 0.0067). This exponential decay of computational fidelity represents the primary obstacle to achieving chemical accuracy for anything beyond the smallest molecular systems.

Tolerable Error Thresholds for Quantum Chemistry Applications

The stringent precision requirements of computational chemistry demand exceptionally low error rates for useful computation. Recent density-matrix simulations have quantified the relationship between algorithm performance and hardware errors, providing concrete targets for hardware development and error mitigation strategies.

Table 2: Maximally Tolerable Gate-Error Probabilities (p_c) for VQE to Achieve Chemical Accuracy [2]

Molecular System Size (Orbitals)	Required p_c (Without Error Mitigation)	Required p_c (With Error Mitigation)	Typical Two-Qubit Gate Count (N_II)
Small (4-8)	10⁻⁶ to 10⁻⁵	10⁻⁴ to 10⁻³	10² - 10³
Medium (10-14)	10⁻⁶ to 10⁻⁴	10⁻⁴ to 10⁻²	10³ - 10⁴
Large (>16)	<10⁻⁶	<10⁻⁴	>10⁴

The data reveals a critical scaling relation: the maximally allowed gate-error probability for any VQE to achieve chemical accuracy decreases approximately as ( {p}{c} \propto {N}{II}^{-1} ), where ( N_{II} ) is the number of noisy two-qubit gates [2]. This inverse relationship underscores that as molecular complexity increases, even more stringent error control is required. Furthermore, iterative VQE algorithms such as ADAPT-VQE demonstrate better noise resilience compared to fixed-ansatz approaches like UCCSD, as they typically generate shallower circuits tailored to the specific molecular system [2].

Experimental Protocols for Error Characterization and Mitigation

Protocol 1: Zero-Noise Extrapolation (ZNE) with Qubit Error Probability (QEP) Scaling

Zero-Noise Extrapolation is a widely adopted error mitigation technique that intentionally amplifies circuit noise to extrapolate results back to the zero-noise limit. Traditional ZNE methods that simply multiply circuit layers provide poor noise scaling estimates. The following protocol incorporates a more accurate Qubit Error Probability metric for enhanced precision [4].

Application: Mitigating gate errors in variational quantum algorithms for molecular energy estimation.

Materials and Setup:

• Quantum hardware backend (e.g., IBM Osaka/Kyoto, Quantinuum H1)
• Calibration data for target device (T1, T2 times, gate error rates, readout errors)
• Classical post-processing environment (e.g., Python with Mitiq, Qiskit)

Procedure:

• Circuit Preparation: Implement the target quantum circuit (e.g., VQE ansatz for molecular Hamiltonian).
• QEP Calculation: For each qubit in the register, compute the individual error probability as a function of:
  • Native gate fidelities
  • Idle time during execution (decoherence)
  • Crosstalk from neighboring qubits
• Noise Scaling: Artificially scale noise using pulse-stretching techniques or identity insertion, using QEP rather than simple gate counts to calibrate the scaling factor.
• Data Acquisition: Execute the scaled circuits (e.g., at 1x, 2x, and 3x base QEP levels) with sufficient shots for statistical significance (typically 10⁴-10⁶ per circuit).
• Extrapolation: Fit the measured expectation values (e.g., energy) versus QEP to a linear, exponential, or Richardson model and extrapolate to zero QEP.

Validation: Execute on a classically simulable system (e.g., Hâ‚‚ molecule in minimal basis) and compare the ZNE-corrected energy to the exact full configuration interaction (FCI) result. Successful mitigation should reduce the absolute error below the chemical accuracy threshold of 1.6 mHa [4] [3].

Figure 1: Zero-Noise Extrapolation (ZNE) workflow using Qubit Error Probability (QEP) for precise noise scaling and mitigation.

Protocol 2: Multireference-State Error Mitigation (MREM) for Strongly Correlated Systems

Strongly correlated molecular systems, such as those encountered in transition metal complexes or bond dissociation processes, present particular challenges for quantum simulation. The standard Reference-state Error Mitigation (REM) method, which uses a single Hartree-Fock reference state, fails in these multireference scenarios. The MREM protocol addresses this limitation [8].

Application: Error mitigation for VQE simulations of strongly correlated molecules (e.g., Nâ‚‚, Fâ‚‚, metal oxides).

Materials and Setup:

• Classical computational chemistry software (e.g., PySCF, ORCA) for multireference state selection
• Quantum hardware or simulator with support for generalized gates
• Circuit construction tools for Givens rotations

Procedure:

• Reference Selection: Classically compute a compact multireference wavefunction composed of a few dominant Slater determinants using inexpensive methods (e.g., CASSCF(2,2), DMRG, or selected CI).
• Circuit Construction: Implement the multireference state on the quantum processor using Givens rotation circuits, which preserve particle number and spin symmetry.
• Noise Characterization: Prepare and measure the multireference state on the quantum device to characterize the device-specific noise profile.
• Target State Execution: Run the VQE algorithm to prepare the target molecular ground state and measure its noisy energy, ( E_{\text{noisy}} ).
• Error Estimation: Compute the exact energy of the multireference state, ( E{\text{MR, exact}} ), classically and measure its noisy counterpart, ( E{\text{MR, noisy}} ), on the device.
• Mitigation: Apply the correction: ( E{\text{mitigated}} = E{\text{noisy}} - (E{\text{MR, noisy}} - E{\text{MR, exact}}) ).

Validation: Apply to the nitrogen molecule at dissociation, where the Hartree-Fock state fails. Successful MREM should recover potential energy curves smoothlty across the dissociation coordinate, maintaining chemical accuracy where single-reference REM fails [8].

Protocol 3: Measurement Error Mitigation with Quantum Detector Tomography (QDT)

Readout errors represent a dominant noise source in precision measurement applications. This protocol leverages Quantum Detector Tomography to fully characterize and mitigate measurement errors, enabling high-precision energy estimation [9].

Application: Achieving chemical precision measurements for molecular energy estimation, particularly for complex observables with many Pauli terms.

Materials and Setup:

• Quantum device with configurable measurement basis
• Parallel compilation tools for efficient circuit scheduling
• Classical post-processing infrastructure for tomographic reconstruction

Procedure:

• Calibration Circuit Generation: Prepare and execute a complete set of informationally complete calibration circuits (e.g., all possible input basis states for n-qubits).
• Noise Matrix Construction: Measure the output statistics for each calibration circuit to construct the assignment probability matrix (A), where ( A_{ij} = P(\text{measure outcome } i | \text{prepared state } j) ).
• Blended Scheduling: Interleave calibration circuits with actual chemistry circuits (e.g., Hartree-Fock state preparation with Hamiltonian measurement) to average over temporal noise fluctuations.
• Mitigated Estimation: For the actual experiment outcomes (raw probability vector ( \vec{p}{\text{raw}} )), compute the mitigated probabilities by applying the inverse (or pseudo-inverse) of the assignment matrix: ( \vec{p}{\text{mitigated}} = A^{-1} \vec{p}_{\text{raw}} ).
• Observable Calculation: Compute the expectation value of the molecular Hamiltonian using the mitigated probability distribution.

Validation: Execute the protocol for the BODIPY molecule on an IBM Eagle processor. Successful implementation has demonstrated a reduction in measurement errors from 1-5% to 0.16%, approaching chemical precision [9].

Figure 2: High-precision measurement workflow using Quantum Detector Tomography (QDT) and blended scheduling to mitigate readout errors in molecular energy estimation.

The Scientist's Toolkit: Research Reagent Solutions

Table 3: Essential Resources for NISQ-Era Quantum Chemistry Experiments

Resource Category	Specific Examples	Function/Application	Key Characteristics
Quantum Hardware Platforms	IBM Osaka/Kyoto (Superconducting), Quantinuum H1/H2 (Trapped Ion)	Execution of quantum circuits for molecular energy estimation	Variable qubit count/quality, native gate sets, connectivity [6] [5]
Error Mitigation Software	Mitiq, Qiskit Runtime, True-Q	Implementation of ZNE, PEC, REM/MREM protocols	Hardware-agnostic, integrates with popular QC SDKs [4] [10]
Classical Computational Chemistry Tools	PySCF, ORCA, Q-Chem	Molecular integral computation, active space selection, reference state generation	Prepares fermionic Hamiltonians and initial states for VQE [8]
Quantum Algorithm Libraries	TEQUILA, Qiskit Nature, PennyLane	Implementation of VQE, ADAPT-VQE, QAOA for chemistry	Provides parameterized ansätze and classical optimizers [2]
Characterization & Benchmarking Tools	Randomized Benchmarking, Gate Set Tomography, TED-qc	Quantification of gate errors, coherence times, and Qubit Error Probability (QEP)	Establishes device-specific error models for mitigation [4] [7]
HIF1-IN-3	HIF1-IN-3, MF:C26H24N2O3, MW:412.5 g/mol	Chemical Reagent	Bench Chemicals
7030B-C5	8-[(2-Hydroxyethyl)amino]-7-[(3-methoxyphenyl)methyl]-1,3-dimethyl-2,3,6,7-tetrahydro-1H-purine-2,6-dione	High-purity 8-[(2-HYDROXYETHYL)AMINO]-7-[(3-METHOXYPHENYL)METHYL]-1,3-DIMETHYL-2,3,6,7-TETRAHYDRO-1H-PURINE-2,6-DIONE for research use only (RUO). Not for human or veterinary diagnosis or therapeutic use.	Bench Chemicals

The path toward quantum utility in computational chemistry and drug development hinges on the co-design of algorithms, error mitigation strategies, and hardware capabilities. The protocols and analyses presented here demonstrate that while significant challenges remain, systematic error management can already yield chemically meaningful results for appropriately scaled problems on current NISQ devices. The integration of physical insights—such as the use of multireference states in MREM—with sophisticated mitigation techniques like ZNE and QDT represents the cutting edge of NISQ-era quantum computational chemistry.

Looking forward, the transition beyond the NISQ era will be marked by the implementation of practical quantum error correction, as recently demonstrated in preliminary experiments on trapped-ion systems [6]. However, for the foreseeable future, error mitigation rather than full correction will remain the primary strategy for extracting value from quantum simulations. Researchers in pharmaceutical and materials science should prioritize engaging with these rapidly evolving techniques, focusing initially on benchmark systems with clear classical references to build institutional expertise. As hardware continues to improve—with gate fidelities approaching 99.99% and quantum volume increasing—the application of these protocols will enable the treatment of increasingly complex molecular systems, potentially transforming the discovery pipelines for new therapeutics and functional materials.

The successful implementation of quantum algorithms, particularly for high-precision fields like quantum chemistry simulation, is critically dependent on understanding and managing the diverse sources of noise in quantum systems. These noise sources introduce errors that can rapidly degrade computational accuracy, making their systematic characterization a prerequisite for effective error mitigation [9]. For near-term quantum hardware, where full-scale error correction remains impractical, developing noise-resilient protocols is essential for achieving chemically meaningful results, such as molecular energy estimation to within chemical accuracy (1.6 mHa) [3] [9]. This application note provides a structured framework for identifying, quantifying, and mitigating the predominant sources of quantum noise, with a specific focus on applications in quantum computational chemistry.

Quantum noise arises from a complex interplay of environmental interactions, control imperfections, and fundamental material properties. The table below categorizes primary noise sources, their physical origins, and their impact on quantum computations.

Table 1: Classification of Primary Quantum Noise Sources

Noise Category	Specific Type	Physical Origin	Impact on Quantum Computation
Environmental	Thermal Noise [11]	Finite temperature fluctuations causing Brownian motion.	Limits displacement sensitivity; adds thermal occupation to qubit states.
Environmental	Decoherence [12] [13]	Unwanted interaction with the environment (photons, phonons, magnetic fields).	Causes loss of superposition and entanglement, the core quantum resources.
Control Imperfections	Control Signal Noise [13]	Imperfections in classical electronics generating control pulses.	Incorrect gate operations (e.g., wrong rotation angles), reducing gate fidelity.
Control Imperfections	Readout Errors [9]	Inaccurate measurement operations.	Misassignment of final qubit states (e.g., reading 0 as 1), corrupting results.
Fundamental Material	Material Defects [13]	Atomic vacancies, impurities, or grain boundaries in substrate materials.	Creates localized charge/ magnetic fluctuations, leading to unpredictable qubit behavior.
Stochastic	Depolarizing Noise [14]	Qubit randomly undergoing one of the Pauli operators (X, Y, Z).	Fully randomizes the qubit state with a given probability.
Stochastic	Amplitude Damping [14]	Energy dissipation, modeling the spontaneous emission of a photon.	Loss of energy from the excited |1> state to the ground |0> state.

The Fluctuation-Dissipation Theorem provides a fundamental link for certain noise types, such as thermal noise. The displacement power spectral density due to thermal noise can be modeled as: [S{\mathrm{th}}(\omega) = \sum{k=0}^{n} \frac{4kB T \omegak \phi}{mk[(\omega^{2}{k} - \omega^{2})^{2} + \omega^{2} \omega^{2}{k} \phi^{2}]}] where (kB) is Boltzmann's constant, (T) is temperature, (mk) and (\omegak) are the mass and frequency of the (k)th mechanical mode, and (\phi) is the loss angle [11]. This formalizes the direct relationship between dissipation (encoded in (\phi)) and the resulting fluctuations.

Experimental Protocols for Noise Characterization

Accurate noise characterization is the foundation of effective mitigation. The following protocols provide methodologies for quantifying key noise parameters.

Protocol for Characterizing Thermal Noise and Mechanical Loss

This protocol is designed to characterize the thermal noise contribution in systems such as suspended mirror microresonators, which is critical for high-precision interferometric measurements [11].

• Apparatus Setup: Incorporate the device under test (e.g., a GaAs/AlGaAs micro-mirror suspended on a cantilever) as the end-mirror in a Fabry–Pérot optical cavity. The entire setup must be housed in a high-vacuum chamber (e.g., at ∼10⁻⁸ torr) and cryogenically cooled to target temperatures (e.g., 25 K) to mitigate viscous damping and reduce thermal noise [11].
• Quality Factor (Q) Measurement: Perform a ring-down measurement. Excite the mechanical resonator, then observe the exponential decay of its amplitude. The quality factor is calculated as (Q = \omega0 / \Delta\omega), where (\omega0) is the resonant frequency and (\Delta\omega) is the linewidth of the resonance. A high Q (e.g., 25,000 ± 2,200) indicates low mechanical loss [11].
• Spectral Measurement: Use the optical cavity to measure the displacement power spectral density (PSD) of the mirror's motion. The measured PSD will contain peaks corresponding to the mechanical resonant frequencies.
• Modal Mass Estimation: Fit the measured noise spectra to a model based on the fluctuation-dissipation theorem (e.g., Equation 1). The known temperature (T), frequency (ω_k), and measured Q factor allow for the extraction of the effective modal masses (m_k) for each resonance, which can be cross-verified with finite element analysis (FEA) simulations [11].
• Noise Level Quantification: Compare the magnitude of the characterized thermal noise PSD to the Standard Quantum Limit (SQL), often reported in decibels (dB) below the SQL [11].

Protocol for Quantum Noise Channel Identification via QBER

This protocol leverages simple Quantum Key Distribution (QKD) circuits and classical Machine Learning (ML) to identify dominant noise channels in a processor, a method applicable to low-qubit, noisy devices [14].

• Circuit Preparation: Select a simple QKD protocol, such as BB84 or BBM92, which requires only single-qubit or two-qubit gates [14].
• Data Generation: a. Simulation: Simulate the QKD circuits on a classical computer, explicitly injecting known noise channels (e.g., bit-flip, amplitude damping, depolarizing) with varying strengths (p). b. Hardware Execution: Run the same QKD circuits on the target near-term quantum processor.
• Metric Extraction: For each execution (simulated or hardware), calculate the Quantum Bit Error Rate (QBER), which is the fraction of mismatched bits in the final key between the two protocol parties.
• Model Training: Use the simulation-generated data (QBER values and the corresponding known noise type) as a labeled training set. Train supervised ML classifiers, such as K-Nearest Neighbors (KNN), Gaussian Naive Bayes, or Support Vector Machines (SVM) [14].
• Noise Identification: Feed the QBER data collected from the hardware runs into the trained ML model. The model will output a classification for the dominant noise type present in the hardware.

Workflow for Integrated Noise Characterization

The diagram below illustrates the logical sequence and decision points in a comprehensive noise characterization workflow, integrating elements from the protocols above.

The Scientist's Toolkit: Research Reagent Solutions

This section details the essential components and techniques required for advanced quantum noise characterization and mitigation experiments.

Table 2: Essential Research Reagents and Tools for Quantum Noise Studies

Tool / Material	Function / Specification	Application Context
Low-Noise Mirror Microresonators (e.g., GaAs/AlGaAs) [11]	High-reflectivity mirror coatings with low mechanical loss, suspended as cantilevers.	Serves as a test mass for probing thermal noise in ultra-sensitive measurements.
Optical Cavities & Readout (e.g., Fabry–Pérot) [11]	Provides a highly sensitive platform for measuring displacement and forming an "optical spring" to manipulate dynamics.	Used to probe thermal noise and quantum radiation pressure noise below the Standard Quantum Limit.
Cryogenic Systems [11] [13]	Dilution refrigerators cooling to millikelvin (mK) temperatures, often paired with vacuum chambers.	Reduces thermal noise by freezing out environmental energy fluctuations.
Quantum Detector Tomography (QDT) [9]	A protocol to fully characterize the noisy measurement process of a quantum device by reconstructing its Positive Operator-Valued Measure (POVM).	Mitigates readout errors in high-precision tasks like molecular energy estimation, reducing systematic bias.
Informationally Complete (IC) Measurements [9]	A set of measurement settings that fully characterizes the quantum state, allowing estimation of multiple observables from the same data.	Reduces circuit overhead and enables efficient error mitigation via post-processing.
Density Matrix Simulators [15]	Quantum simulators (e.g., Amazon Braket DM1) that use density matrices instead of state vectors.	Essential for simulating open quantum systems, decoherence, and general noise channels before running on hardware.
MMV006833	MMV006833, MF:C19H27ClN2O4S, MW:414.9 g/mol	Chemical Reagent
CA IX-IN-3

Error Mitigation Protocols for Quantum Chemistry Simulations

Leveraging accurate noise characterization, the following advanced protocols are specifically designed to enhance the fidelity of quantum chemistry computations on near-term devices.

Multireference-State Error Mitigation (MREM)

The standard Reference-State Error Mitigation (REM) method uses a single, easily prepared reference state (e.g., Hartree-Fock) to calibrate energy errors. However, its effectiveness is limited for strongly correlated systems where the true ground state is a complex combination of multiple Slater determinants. MREM addresses this by using a multireference state with significant overlap to the target ground state [8].

• Multireference State Generation: Classically compute a compact multireference wavefunction for the target molecule (e.g., using a selected CI or CASSCF method). This state is a truncated linear combination of dominant Slater determinants.
• Quantum Circuit Preparation: Prepare the multireference state on the quantum processor. This can be efficiently achieved using circuits composed of Givens rotations, which are universal for state preparation and preserve particle number and spin symmetry [8].
• Noisy Energy Estimation: Measure the energy expectation value of the Hamiltonian with respect to the prepared multireference state on the noisy quantum device, yielding (E_{\text{noisy}}^{\text{MR}}).
• Classical Energy Calculation: Classically and exactly compute the energy of the same multireference state, (E_{\text{exact}}^{\text{MR}}).
• Error Calibration & Mitigation: The energy error for the multireference state is (\Delta{\text{MR}} = E{\text{noisy}}^{\text{MR}} - E{\text{exact}}^{\text{MR}}). This error is assumed to be similar for the target state. If (E{\text{noisy}}^{\text{Target}}) is the measured energy of the target VQE state, the mitigated energy is: [ E{\text{mitigated}} = E{\text{noisy}}^{\text{Target}} - \Delta_{\text{MR}} ] This procedure significantly improves accuracy for strongly correlated systems like the bond-stretching regions of Nâ‚‚ and Fâ‚‚ molecules [8].

Protocol for High-Precision Readout Error Mitigation

Achieving chemical precision (1.6×10⁻³ Ha) in energy estimation requires aggressive mitigation of readout errors, which can be 1-5% on typical devices [9].

• Implement Informationally Complete (IC) Measurements: Choose a set of measurement bases that fully characterize the state. Techniques like Locally Biased Classical Shadows can be used to reduce the shot overhead of this step by prioritizing measurement settings that have a larger impact on the energy estimation [9].
• Parallel Quantum Detector Tomography (QDT): Interleave the execution of the chemistry circuits with circuits dedicated to QDT. This involves preparing a complete set of basis states (|0...0>, |0...1>, ..., |1...1>) and measuring them to reconstruct the device's noisy measurement operator [9].
• Blended Scheduling: To mitigate time-dependent noise (drift), schedule all circuits (chemistry and QDT) in a blended, interleaved manner rather than in large sequential blocks. This ensures temporal fluctuations affect all computations equally on average [9].
• Post-Processing with Calibrated Noise Model: Use the measurement operator obtained from QDT to construct an unbiased estimator for the molecular energy during classical post-processing. This corrects the raw measurement outcomes, reducing estimation bias and enabling errors as low as 0.16% [9].

Workflow for Noise-Resilient Quantum Metrology

This workflow, which integrates quantum sensing with quantum computing, can be adapted to enhance the precision of quantum chemistry observables.

The core innovation is processing the noisy quantum state ρ̃ on a quantum computer without converting it to classical data, thus avoiding a major bottleneck. Techniques like quantum Principal Component Analysis (qPCA) can then filter out the dominant noise, yielding a purified state ρ_NR from which the parameter (e.g., a molecular energy) can be estimated with significantly enhanced accuracy and precision, as quantified by the Quantum Fisher Information [16].

The transformative potential of quantum computing for simulating molecular systems in chemistry and drug development is currently constrained by a fundamental challenge: the high error rates inherent in today's quantum hardware. Quantum bits (qubits) are exceptionally fragile, with current physical error rates typically around 1 in 100 to 1 in 10,000 operations [17]. These errors arise from environmental disturbances, decoherence, and imperfect gate operations, degrading computational results and limiting the scale of feasible quantum simulations.

To address this challenge, two distinct but complementary approaches have emerged: Quantum Error Mitigation (QEM) and Quantum Error Correction (QEC). For researchers focused on near-term quantum chemistry applications, understanding the practical trade-offs between these approaches is critical for designing viable experimental protocols. This application note provides a structured comparison of QEM and QEC strategies, detailing their operational principles, resource requirements, and practical implementation for quantum simulations in chemical research.

Core Conceptual Differences

Fundamental Operational Principles

Quantum Error Mitigation and Quantum Error Correction represent philosophically distinct approaches to managing errors in quantum computations, each with characteristic mechanisms and objectives.

• Quantum Error Mitigation (QEM): QEM encompasses a suite of post-processing techniques that infer less noisy results from multiple executions of noisy quantum circuits. Rather than preventing errors during computation, QEM allows errors to occur and subsequently "averages out" their impact through classical post-processing of measurement results [18] [19]. As summarized by one expert, "Error mitigation focuses on techniques for noticing that a shot is bad, or for quantifying how a shot is bad. You run lots and lots of shots, and use the good-vs-bad signals to get better estimates of some quantity" [20]. These methods are generally application-specific, particularly effective for expectation value estimation rather than full distribution sampling [18].

• Quantum Error Correction (QEC): QEC takes a preventive approach by encoding quantum information across multiple physical qubits to create more robust logical qubits. Through repetitive cycles of syndrome measurement, decoding, and correction, QEC actively detects and corrects errors as they occur during computation [19] [21]. This approach aims to implement fault-tolerant quantum computation, where errors are suppressed sufficiently to allow for arbitrarily long calculations, provided physical error rates remain below a certain threshold [17]. "QEC doesn't 'end' the existence of errors—it reduces their likelihood," notes one technical guide, highlighting that the objective is progressive error suppression rather than complete elimination [18].

Comparative Analysis Framework

The table below summarizes the key characteristics and trade-offs between QEM and QEC approaches:

Characteristic	Quantum Error Mitigation (QEM)	Quantum Error Correction (QEC)
Core Objective	Extract correct signal from noisy outputs through post-processing [20]	Make each quantum computation shot inherently reliable [20]
Primary Mechanism	Classical post-processing of results from multiple circuit variants [19]	Encoding logical qubits across physical qubits with real-time correction [19] [17]
Hardware Overhead	Low physical qubit overhead	High physical qubit overhead (100+:1 ratio common) [18] [21]
Temporal Overhead	Exponential shot overhead with circuit size [18]	Circuit slowdown (1000x-1,000,000x) but no exponential shot scaling [18] [20]
Error Types Addressed	Both coherent and incoherent errors [18]	All error forms (including qubit loss) with appropriate codes [18]
Application Scope	Estimation tasks (expectation values) [18]	Universal (any algorithm on logical qubits) [18]
Implementation Timeline	Near-term (current devices)	Long-term (requiring more advanced hardware) [18] [17]
Key Limitation	Exponential resource scaling with circuit size [18] [22]	Massive resource requirements for useful logical error rates [18] [21]

Table 1: Comparative analysis of Quantum Error Mitigation and Quantum Error Correction approaches.

Quantum Error Mitigation: Protocols for Near-Term Chemistry Simulations

Core QEM Techniques and Workflows

For quantum chemistry applications such as molecular energy estimation, several QEM techniques have demonstrated practical utility on current hardware. These include Zero-Noise Extrapolation (ZNE), Probabilistic Error Cancellation (PEC), and measurement error mitigation, each with distinct operational principles and resource requirements.

The following workflow illustrates a typical quantum error mitigation protocol for chemistry simulations:

Figure 1: Quantum Error Mitigation Workflow for Molecular Energy Estimation.

Practical Implementation: Molecular Energy Estimation Case Study

Recent research demonstrates the successful application of QEM techniques for high-precision molecular energy estimation. A 2025 study published in npj Quantum Information implemented a comprehensive measurement protocol for the BODIPY molecule on IBM quantum hardware, achieving a significant reduction in measurement errors from 1-5% to 0.16% [9].

The experimental protocol incorporated these key techniques:

• Informationally Complete (IC) Measurements: Enabled estimation of multiple observables from the same measurement data and provided a framework for efficient error mitigation [9].

• Locally Biased Random Measurements: Reduced shot overhead by prioritizing measurement settings with greater impact on energy estimation [9].

• Quantum Detector Tomography (QDT): Characterized readout errors to create an unbiased estimator for molecular energy, significantly reducing estimation bias [9].

• Blended Scheduling: Mitigated time-dependent noise by interleaving circuit executions, ensuring homogeneous noise exposure across all measurements [9].

This integrated approach demonstrates that chemical precision (1.6×10⁻³ Hartree) is achievable on current hardware for moderately sized quantum chemistry problems through sophisticated error mitigation strategies.

Quantum Error Correction: Towards Fault-Tolerant Chemistry Simulations

QEC Fundamentals and Implementation Architecture

Quantum Error Correction represents the foundational approach for achieving truly scalable quantum computation. Unlike error mitigation, QEC employs quantum codes to proactively protect information during computation through spatial redundancy and active feedback.

The following diagram illustrates the continuous cycle of real-time quantum error correction:

Figure 2: Quantum Error Correction Cycle with Real-Time Feedback.

Practical QEC Demonstration and Resource Requirements

Recent experimental milestones demonstrate progress in implementing QEC, though practical challenges remain for chemistry-scale applications:

• Google's Willow Chip: Implemented the surface code with a 7×7 qubit array, demonstrating a 2.14-fold error reduction with each scaling stage and operating below the critical error threshold for the first time [21].

• Resource Overheads: Current QEC implementations require substantial resource investment. Google's experiment used 105 physical qubits to realize a single logical qubit [18], while practical fault-tolerant systems may require 100-1000 physical qubits per logical qubit depending on the code and physical error rates [21].

• Real-Time Decoding Challenge: A critical bottleneck for practical QEC is the need for high-speed, low-latency decoding. "For feedback-based QEC, low latency is essential," notes Qblox, with control stacks requiring deterministic feedback networks capable of sharing measurement outcomes within ≈400ns across modules [21].

These demonstrations represent significant progress but highlight that QEC for complex chemistry simulations requiring many logical qubits remains a future prospect given current hardware limitations.

Hybrid Approaches: Integrating QEM and QEC for Near-Term Applications

Hybrid Protocol Design and Workflow

Recognizing the complementary strengths of both approaches, researchers have begun developing hybrid protocols that integrate elements of QEM and QEC. These approaches aim to balance resource overhead and error suppression capabilities for near-term applications.

A 2025 hybrid protocol combines the [[n,n−2,2]] quantum error detection code (QEDC) with probabilistic error cancellation (PEC) and modified Pauli twirling [22]. This approach leverages the constant qubit overhead and simple post-processing of QEDC while using PEC to address undetectable errors that escape the detection code.

The following workflow illustrates this integrated approach:

Figure 3: Hybrid Quantum Error Detection and Mitigation Protocol.

Practical Implementation and Advantages

This hybrid protocol demonstrated practical utility in quantum chemistry applications, particularly for variational quantum eigensolver (VQE) circuits estimating the ground state energy of Hâ‚‚ [22]. The approach offers several key advantages:

• Reduced Sampling Overhead: By applying PEC to a lower-noise circuit (after error detection), the sampling overhead is substantially reduced compared to applying PEC directly to unprotected circuits [22].

• Compatibility with Non-Clifford Operations: The [[n,n−2,2]] code provides simpler encoding schemes for logical rotations, eliminating the need for complex compilation into Clifford+T circuits and avoiding associated approximation errors [22].

• Practical QEC Introduction: The constant qubit overhead (only 2 additional qubits regardless of register size) makes this approach feasible on current hardware, serving as an accessible introduction to encoded quantum computation [22].

This hybrid approach demonstrates the potential for strategically combining elements of both QEM and QEC to achieve practical error suppression on current quantum hardware while managing resource constraints.

Research Reagent Solutions for Quantum Error Management

Implementing effective error management strategies for quantum chemistry simulations requires both hardware and software tools. The following table catalogues essential resources mentioned in recent literature:

Resource/Technique	Function	Example Implementations
Error Suppression	Proactively reduces error likelihood during circuit execution via optimized control pulses [19]	Boulder Opal [19], Fire Opal [19]
Zero-Noise Extrapolation (ZNE)	Extrapolates observable expectations to zero error by intentionally increasing noise levels [18]	Mitiq, Qiskit Runtime
Probabilistic Error Cancellation (PEC)	Constructs unbiased estimators by combining results from noisy circuits with quasi-probability distributions [18] [22]	Qiskit, TrueQ
Measurement Error Mitigation	Corrects readout errors using confusion matrix tomography [19] [9]	Qiskit, Cirq
Quantum Detector Tomography (QDT)	Characterizes noisy measurement effects to build unbiased estimators [9]	Custom implementations
Pauli Twirling	Converts coherent errors into stochastic errors via random Pauli operations [22]	Qiskit, PyQuil
Real-Time Decoders	Interprets error syndromes and determines corrections within QEC cycles [21] [17]	Deltakit [23], Tesseract [24]
QEC Control Stacks	Provides low-latency feedback and scalable control for QEC experiments [21]	Qblox [21]

Table 2: Essential Resources for Implementing Quantum Error Management Protocols.

The strategic selection between quantum error mitigation and quantum error correction represents a critical design decision for researchers pursuing quantum chemistry simulations. For near-term applications on currently available hardware, quantum error mitigation techniques offer practical pathways to enhanced precision for specific tasks like molecular energy estimation, albeit with exponential scaling limitations. Quantum error correction, while theoretically capable of enabling arbitrary-scale quantum computations, currently demands resource overheads that limit its immediate utility for complex chemistry simulations.

Hybrid approaches that strategically combine error detection with mitigation present a promising intermediate path, offering enhanced error suppression with manageable overheads. As hardware continues to improve, with physical error rates decreasing and qubit counts increasing, the balance between these approaches will inevitably shift toward full quantum error correction. For the present, however, quantum error mitigation and hybrid protocols provide the most viable path toward demonstrating quantum utility in chemistry simulations on near-term devices.

The rapid progress in both domains suggests that effective error management—whether through mitigation, correction, or hybrid approaches—will remain the critical enabling technology for practical quantum chemistry applications in the coming years. Research teams should maintain flexibility in their error management strategies, adopting techniques matched to their specific application requirements and available hardware capabilities.

Impact of Noise on Variational Quantum Eigensolver (VQE) Accuracy

The Variational Quantum Eigensolver (VQE) is a leading hybrid quantum-classical algorithm for determining molecular ground-state energies, with promising applications in drug development and materials science [25]. However, current Noisy Intermediate-Scale Quantum (NISQ) devices suffer from significant gate and readout errors that severely impact the accuracy and reliability of VQE simulations [26] [2]. Understanding and mitigating these noise effects is crucial for advancing quantum computational chemistry. This application note synthesizes recent findings on noise impacts and provides detailed protocols for error-resilient VQE implementation.

Quantitative Analysis of Noise Impacts

Tolerable Gate Error Probabilities for Chemical Accuracy

Table 1: Maximally allowed gate-error probabilities (p_c) for VQEs to achieve chemical accuracy (1.6 mHa) [2]

VQE Algorithm Type	Specific Ansatz	Without Error Mitigation	With Error Mitigation
Fixed Ansatz	UCCSD	10⁻⁶ to 10⁻⁴	10⁻⁴ to 10⁻²
Fixed Ansatz	k-UpCCGSD	10⁻⁶ to 10⁻⁴	10⁻⁴ to 10⁻²
Adaptive Ansatz	ADAPT-VQE	10⁻⁶ to 10⁻⁴	10⁻⁴ to 10⁻²

Performance Comparison with Error Mitigation

Table 2: Error mitigation impact on VQE accuracy for BeHâ‚‚ simulations [26]

Quantum Processor	Qubit Count	Error Mitigation	Accuracy vs Exact (Orders of Magnitude)	Key Result
IBMQ Belem	5	None	~10⁻¹ (Baseline)	Reference point
IBMQ Belem	5	T-REx	~10⁻² (Improvement)	10x improvement with mitigation
IBM Fez	156	None	~10⁻¹ (Similar to unmigitated Belem)	Smaller, older device with mitigation outperforms larger, newer device without mitigation

Experimental Protocols for Noise-Resilient VQE

Protocol 1: Twirled Readout Error Extinction (T-REx) Implementation

Purpose: Mitigate readout errors to improve VQE parameter quality and energy estimation [26]

Materials:

• Quantum processor or simulator with readout error characterization capabilities
• Classical optimization routine (e.g., SPSA)
• Circuit twirling components

Procedure:

• Characterize Readout Error: Execute comprehensive readout calibration using prepared computational basis states.
• Construct T-REx Filters: Derive optimal filters from calibration data to correct readout probabilities.
• Integrate with VQE Loop:
  • During each energy evaluation, apply T-REx filtering to measurement outcomes.
  • Use corrected probabilities for expectation value calculation.
• Optimize Parameters: Proceed with classical optimization using error-mitigated energies.
• Validate Results: Compare mitigated results with state-vector simulations to verify parameter quality.

Protocol 2: Gate Error Resilience Benchmarking

Purpose: Quantify VQE algorithm performance under depolarizing noise [2]

Materials:

• Density-matrix simulator with configurable noise models
• Molecular Hamiltonians (4-14 orbitals)
• VQE ansätze (UCCSD, k-UpCCGSD, ADAPT-VQE)

Procedure:

• Setup Noise Model: Configure depolarizing noise channels with target error probabilities (10⁻⁶ to 10⁻²).
• Execute Noisy Simulations: For each molecule and ansatz type:
  • Run complete VQE optimization under noisy conditions.
  • Record final energy error from exact diagonalization.
• Determine Threshold: Identify the maximum error probability (p_c) maintaining chemical accuracy.
• Analyze Scaling: Fit pc versus number of two-qubit gates (NII) to establish ({p}{c}\mathop{\propto }\limits{ \sim }{N}_{{{{\rm{II}}}}}^{-1}) relationship.

Protocol 3: Quantum-DFT Embedding Workflow

Purpose: Simulate complex materials while mitigating NISQ limitations [27] [28]

Materials:

• Qiskit Nature framework (v43.1+)
• PySCF electronic structure package
• Molecular structures from CCCBDB or JARVIS-DFT

Procedure:

• Structure Generation: Obtain pre-optimized molecular geometries from databases.
• Active Space Selection: Use ActiveSpaceTransformer to identify correlated regions (e.g., 3 orbitals, 4 electrons for Al clusters).
• Hamiltonian Generation: Perform single-point calculations and map electronic Hamiltonian to qubits via Jordan-Wigner transformation.
• VQE Execution:
  • Employ hardware-efficient ansatz (e.g., EfficientSU2) or chemically-inspired ansatz.
  • Optimize parameters using SLSQP or other suitable optimizers.
• Validation: Compare with NumPy exact diagonalization and CCCBDB reference data.

Workflow Diagrams

VQE Error Mitigation Protocol

Quantum-DFT Embedding Methodology

The Scientist's Toolkit

Table 3: Essential research reagents for noise-resilient VQE experiments

Reagent Solution	Function	Example Implementations
Error Mitigation Protocols	Reduce hardware noise impact on measurements	T-REx [26], Zero-Noise Extrapolation, Probabilistic Error Cancellation
Adaptive Ansätze	Construct circuit structures iteratively for noise resilience	ADAPT-VQE [2], tUCCSD [29]
Quantum-Digital Embedding	Combine quantum and classical computational resources	Quantum-DFT Embedding [27] [28]
Classical Optimizers	Navigate noisy parameter landscapes effectively	SPSA [26], SLSQP [27]
Hardware-Efficient Ansätze	Minimize circuit depth for NISQ devices	EfficientSU2 [27], QNP [29]
Measurement Reduction	Decrease shot noise and measurement overhead	Pauli saving [29], measurement grouping
Aminopeptidase-IN-1	Aminopeptidase-IN-1, MF:C18H16N2O6, MW:356.3 g/mol	Chemical Reagent
p53-MDM2-IN-4	p53-MDM2-IN-4, MF:C23H20FN3O3, MW:405.4 g/mol	Chemical Reagent

The accuracy of VQE simulations on NISQ devices is profoundly affected by quantum noise, with gate error probabilities needing to be below 10⁻⁴ to 10⁻² even with error mitigation to achieve chemical accuracy [2]. The integration of error mitigation strategies like T-REx [26] with advanced ansätze and quantum-classical embedding approaches provides a viable path toward meaningful quantum computational chemistry applications. As hardware continues to improve, these protocols will enable researchers to extract increasingly accurate molecular simulations from noisy quantum processors.

Quantum Error Mitigation (QEM) has emerged as a crucial suite of techniques for extracting reliable results from noisy intermediate-scale quantum (NISQ) devices. Unlike quantum error correction, which aims to physically correct errors in real-time, QEM reduces the impact of noise through classical post-processing of results from multiple quantum circuit executions [30]. For computational chemistry and drug development research, these protocols enable more accurate simulations of molecular systems and reaction mechanisms on current quantum hardware, bridging the gap between theoretical potential and practical application. This application note provides a detailed overview of two foundational QEM protocols—Zero-Noise Extrapolation (ZNE) and Probabilistic Error Cancellation (PEC)—with specific guidance for their implementation in near-term chemistry simulations.

Theoretical Foundations

Zero-Noise Extrapolation (ZNE)

ZNE operates on the principle of artificially amplifying circuit noise in a controlled manner, executing the circuit at these elevated noise levels, and mathematically extrapolating the results back to a hypothetical zero-noise scenario [31] [32]. The core assumption is that the relationship between noise strength and observable expectation values follows a smooth, predictable pattern, typically modeled as exponential decay:

⟨O⟩(λ)=ae−bλ+c where ⟨O⟩(λ)=ae−bλ+c\langle O \rangle(\lambda) = a e^{-b\lambda} + ca, aab, and bbc are fitting parameters determined from measurements at different noise levels, and ccλ represents the noise strength [32].λ\lambda

The technique proceeds through three well-defined stages: noise-scaled circuit generation, execution of these circuits, and extrapolation of results. Noise scaling can be achieved through unitary folding methods, either globally across the entire circuit (λ→λ′=(2n+1)λ) or locally at individual gate levels [31].λ→λ′=(2n+1)λ\lambda \rightarrow \lambda' = (2n+1)\lambda

Probabilistic Error Cancellation (PEC)

PEC employs a fundamentally different approach based on quasi-probability decompositions. The core idea involves representing ideal quantum operations as linear combinations of implementable noisy operations. For an ideal channelI and noisy channel I\mathcal{I}N, if one can find a decomposition:N\mathcal{N}

I=∑jαjNj where I=∑jαjNj\mathcal{I} = \sumj \alphaj \mathcal{N}jNj are implementable noisy operations and Nj\mathcal{N}jαj are real coefficients (which may be negative), then the ideal expectation value can be recovered as:αj\alpha_j

⟨O⟩0=∑jαj⟨O⟩Nj [32]. The sampling overhead for this technique scales approximately as ⟨O⟩0=∑jαj⟨O⟩Nj\langle O \rangle0 = \sumj \alphaj \langle O \rangle{\mathcal{N}_j}e4λ, making it computationally expensive for highly noisy circuits but providing exact bias cancellation when properly characterized [32].e4λe^{4\lambda}

Experimental Protocols

ZNE Protocol for Molecular Energy Estimation

Objective: Estimate the ground-state energy of a molecular system using ZNE.

Preparatory Steps:

• Circuit Compilation: Compile the molecular Hamiltonian (e.g., derived from STO-3G basis) into a quantum circuit using techniques such as Trotterization or variational quantum eigensolver (VQE) ansätze.
• Noise Characterization: Characterize the native noise parameters (λ0) of the target quantum processor using gate set tomography or randomized benchmarking.λ0\lambda_0
• Scale Factor Selection: Choose a set of odd-integer scale factors (e.g., [1, 3, 5]) corresponding to increased noise levelsλk=ckλ0, where λk=ckλ0\lambdak = ck\lambda0ck represents the scale factor [31].ckck

Experimental Workflow:

• Circuit Generation: For each scale factorck, generate noise-scaled circuits using unitary folding:ckck
  • Global Folding: Apply the transformationU→U(U†U)n for the entire circuit, where U→U(U†U)nU \rightarrow U(U^\dagger U)^nn=(ck−1)/2 [31].n=(ck−1)/2n = (ck-1)/2
  • Local Folding: Apply similar folding to individual gates within the circuit.
• Circuit Execution: Execute each noise-scaled circuit on the quantum processor (or noisy simulator) with sufficient shots (typically 103-106) to obtain expectation values⟨O⟩(λk) for the molecular energy observable.⟨O⟩(λk)\langle O\rangle(\lambda_k)
• Extrapolation: Apply polynomial or exponential extrapolation to estimate the zero-noise value⟨O⟩(0). For polynomial extrapolation with order 2:⟨O⟩(0)\langle O\rangle(0)
  • Fit the data points(λk,⟨O⟩(λk)) to a polynomial (λk,⟨O⟩(λk))(\lambdak, \langle O\rangle(\lambdak))p(λ)=a0+a1λ+a2λ2.p(λ)=a0+a1λ+a2λ2p(\lambda) = a0 + a1\lambda + a2\lambda^2
  • Extract the zero-noise estimate asa0 [31].a0a0

PEC Protocol for Chemical Observable Measurement

Objective: Mitigate errors in the measurement of chemical observables (e.g., dipole moments, correlation functions) using PEC.

Preparatory Steps:

• Noise Tomography: Fully characterize the noise channelsNj for all native gates in the quantum processor using gate set tomography.Nj\mathcal{N}_j
• Quasi-Probability Decomposition: For each ideal gateUi in the target chemistry circuit, solve the linear system:UiUi Ui=∑jαijN~j where Ui=∑jαijN~jUi = \sumj \alpha{ij} \tilde{\mathcal{N}}jN~j are the characterized noisy gate operations [32].N~j\tilde{\mathcal{N}}j
• Circuit-Specific Decomposition: For the entire circuit implementing unitaryU, compute the overall decomposition:U\mathbf{U} U=∏iUi=∑j→αj→N~j→ where U=∏iUi=∑j→αj→N~j→\mathbf{U} = \prodi Ui = \sum{\overrightarrow{j}} \alpha{\overrightarrow{j}} \tilde{\mathcal{N}}{\overrightarrow{j}}j→ represents sequences of noisy operations and j→\overrightarrow{j}αj→=∏iαiji [33].αj→=∏iαiji\alpha{\overrightarrow{j}} = \prodi \alpha{ij_i}

Experimental Workflow:

• Circuit Sampling: Sample a noisy circuit implementationN~j→ according to the quasi-probability distribution N~j→\tilde{\mathcal{N}}{\overrightarrow{j}}P(j→)=∣αj→∣/γ, where P(j→)=∣αj→∣/γP(\overrightarrow{j}) = |\alpha{\overrightarrow{j}}|/\gammaγ=∑j→∣αj→∣.γ=∑j→∣αj→∣\gamma = \sum{\overrightarrow{j}} |\alpha{\overrightarrow{j}}|
• Circuit Execution: Execute the sampled noisy circuit on the quantum processor and measure the observableO.OO
• Result Weighting: Weight the measurement outcome by the signsign(αj→)×γ.sign(αj→)×γ\text{sign}(\alpha_{\overrightarrow{j}}) \times \gamma
• Averaging: Repeat steps 1-3 multiple times and average the weighted results to obtain the error-mitigated expectation value.

Quantitative Analysis and Performance Comparison

Table 1: Performance Characteristics of ZNE and PEC for Chemistry Simulations

Parameter	Zero-Noise Extrapolation (ZNE)	Probabilistic Error Cancellation (PEC)
Theoretical Basis	Noise scaling and extrapolation [31]	Quasi-probability decomposition [32]
Sampling Overhead	Scales as λ2(n−1) for n noise levels [32]	Scales approximately as e4λ [32]
Bias Elimination	Approximate (dependent on extrapolation model)	Exact (with perfect noise characterization) [32]
Noise Characterization Requirements	Moderate (noise scaling relationship)	High (full gate set tomography) [32]
Optimal Use Cases	Moderate-depth circuits, exploratory calculations	High-precision measurements, small circuits
Implementation Complexity	Low to moderate	High
Compatibility with Quantum Chemistry Algorithms	VQE, quantum phase estimation	Variational algorithms, observable measurement

Table 2: Resource Estimation for Chemical System Simulation (Representative Example)

Protocol	Circuit Depth	Number of Qubits	Required Shots	Effective Error Reduction
Unmitigated	100-500 gates	10-50 qubits	103-104	Baseline
ZNE (polynomial)	100-1500 gates (scaled)	10-50 qubits	104-106	3-10x improvement [31]
PEC	100-500 gates	10-50 qubits	105-108	10-100x improvement [32]

Application to Chemistry Simulations

For quantum chemistry applications, these QEM protocols enable more accurate simulations of molecular properties, reaction pathways, and electronic structure calculations. Quantum embedding theories, which partition systems into active regions treated with high accuracy and environment regions treated with lower-level methods, provide a natural framework for integrating QEM techniques [34]. For instance, strongly-correlated electronic states in molecular active sites or defect centers in materials can be described with effective Hamiltonians whose expectation values are mitigated using ZNE or PEC [34].

Recent advancements include compilation-informed PEC (CIPEC), which simultaneously addresses compilation errors and logical-gate noise, making it particularly relevant for chemistry simulations requiring high precision [33]. This approach uses information about circuit gate compilations to attain unbiased estimation of noiseless expectation values with constant sample-complexity overhead, significantly reducing quantum resource requirements for high-precision chemical calculations [33].

The Scientist's Toolkit

Table 3: Essential Research Reagent Solutions for QEM Experiments

Resource	Function in QEM Protocols	Example Implementations
Noisy Quantum Simulators	Emulates realistic hardware noise for protocol validation	Qrack simulator with configurable noise models [31]
Quantum Programming Frameworks	Provides infrastructure for circuit compilation and execution	PennyLane, Qiskit, Catalyst [31]
Error Mitigation Packages	Implements core ZNE and PEC algorithms	Mitiq, PennyLane noise module [31] [32]
Noise Characterization Tools	Characterizes noise channels for PEC quasi-probability decompositions	Gate set tomography, randomized benchmarking protocols
Classical Post-Processing Libraries	Performs extrapolation and statistical analysis	SciPy, NumPy, custom extrapolation functions [31]
PIN1 inhibitor 6	PIN1 inhibitor 6, MF:C16H15N3O2S2, MW:345.4 g/mol	Chemical Reagent
Antiviral agent 56	2-[(8-Ethoxy-4-methyl-2-quinazolinyl)amino]-5,6,7,8-tetrahydro-4(1H)-quinazolinone	Research-grade 2-[(8-ethoxy-4-methyl-2-quinazolinyl)amino]-5,6,7,8-tetrahydro-4(1H)-quinazolinone for experimental use. For Research Use Only. Not for human, veterinary, or household use.

Zero-Noise Extrapolation and Probabilistic Error Cancellation represent two foundational approaches to quantum error mitigation with complementary strengths and applications in chemistry simulations. ZNE offers a more accessible entry point with lower implementation overhead, making it suitable for initial explorations on moderate-scale problems. PEC provides higher accuracy at greater computational cost, appropriate for precision measurements on well-characterized quantum processors. As quantum hardware continues to evolve, these protocols will play an increasingly vital role in enabling accurate computational chemistry and materials simulations, ultimately accelerating drug discovery and materials design pipelines.

A Deep Dive into Key Error Mitigation Methods for Molecular Systems

Quantum error mitigation has become an essential component for extracting useful results from noisy intermediate-scale quantum (NISQ) devices. Among the various techniques available, Probabilistic Error Cancellation (PEC) stands out as a leading unbiased method for recovering noiseless expectation values from noisy quantum computations [35]. This protocol is particularly valuable for quantum chemistry simulations, where accurate estimation of molecular energies is crucial for applications in drug development and materials science.

PEC operates by quasiprobabilistically simulating the inverse of noise channels affecting quantum operations [36]. Unlike quantum error correction, which aims to suppress errors in real-time through encoding, PEC works by post-processing results from multiple circuit executions to mathematically invert the effect of noise. This approach provides a practical pathway toward computational utility on current quantum hardware despite the presence of inherent noise.

The fundamental principle underlying PEC is that the inverse of a physical noise channel, while not itself a physical quantum channel, can be represented as a linear combination of physically implementable operations [36]. This representation allows researchers to effectively cancel error effects while accepting an inevitable sampling overhead in exchange for improved accuracy. For quantum chemistry applications, this tradeoff enables the estimation of Hamiltonian expectation values with errors small enough to maintain chemical accuracy—a critical requirement for predictive computational chemistry.

Theoretical Foundations

Core Mathematical Framework

The PEC protocol begins with a noisy quantum circuit ( \mathcal{C} = \Lambda\mathcal{U}l \cdots \Lambda\mathcal{U}1 ), where ( \mathcal{U}_i ) represent ideal unitary gates and ( \Lambda ) denotes the error channel affecting each gate [35]. For simplicity, we assume a consistent error channel, though the method generalizes to gate-dependent noise.

The core operation of PEC involves applying the inverse error channel ( \Lambda^{-1} ) to each noisy operation. For a noise channel ( \Lambda = (1-\epsilon)\mathcal{I} + \epsilon\mathcal{E}' ), the inverse takes the form ( \Lambda^{-1} = (1+\epsilon)\mathcal{I} - \epsilon\mathcal{E}' + \mathcal{O}(\epsilon^2) ), which can be verified through the composition ( \Lambda^{-1}\Lambda = \mathcal{I} ) [35]. The presence of negative coefficients in this decomposition indicates that ( \Lambda^{-1} ) is not a physical quantum channel, necessitating the quasiprobability approach.

The ideal expectation value of a Hamiltonian ( H ) with respect to the initial state ( \rho ) is given by: [ \langle H\rangle0 = \text{Tr}[H\mathcal{U}(\rho)] ] where ( \mathcal{U} ) represents the ideal noiseless circuit. Through PEC, we recover this value using the noisy circuit implementation: [ \langle H\rangle0 = \gamma\mathbb{E}[\text{sgn}(ri)\text{Tr}[H\mathcal{O}i\mathcal{U}\lambda(\rho)]] ] where ( \gamma = \sumi |ri| ) represents the sampling overhead, ( ri ) are the quasiprobability coefficients, and ( \mathcal{O}_i ) are implementable operations [36].

Advanced PEC Formulations

Recent research has developed enhanced PEC formulations that reduce the sampling overhead. The standard PEC approach yields a sampling cost of ( \gamma_{\text{PEC}} \approx (1+2\epsilon)^l ) [35], which is suboptimal compared to the theoretical lower bound of ( (1+\epsilon)^l ) [35].

Binomial PEC represents one such advancement, where each inverse channel is decomposed into identity and non-identity components, reorganizing the circuit as a sum of different powers of the inverse generator [35]. This approach allows deterministic shot allocation based on circuit weights, naturally controlling the bias-variance tradeoff.

Pauli Error Propagation offers another overhead reduction technique, particularly effective for Clifford circuits [37]. By leveraging the well-defined interaction between Clifford operations and Pauli noise, this method combined with classical preprocessing significantly reduces sampling requirements while maintaining estimation accuracy.

Table 1: Comparison of PEC Sampling Overheads

Method	Sampling Overhead	Circuit Type	Key Innovation
Standard PEC	( (1+2\epsilon)^{2l} )	General	Quasiprobability decomposition
Binomial PEC	Between ( (1+\epsilon)^{2l} ) and ( (1+2\epsilon)^{2l} )	General	Identity/non-identity separation
Pauli Propagation	Reduced exponent for Clifford portions	Clifford-dominated	Classical Pauli propagation

Practical Implementation

Experimental Workflow

The following diagram illustrates the complete PEC workflow for estimating Hamiltonian expectation values:

The Scientist's Toolkit

Table 2: Essential Research Reagents and Resources for PEC Implementation

Resource	Function	Implementation Notes
Noisy Basis Operations	Physical operations for quasiprobability decomposition	Typically Pauli gates or noisy Clifford operations [36]
Noise Characterization Protocol	Determines error model parameters	Cycle benchmarking or error reconstruction [36]
Quasiprobability Decomposition	Represents inverse noise channels	Optimized to minimize 1-norm of coefficients [35]
Monte Carlo Sampler	Generates circuit instances according to quasiprobability distribution	Tracks sign information for each sample [38]
Readout Error Mitigation	Corrects measurement errors	Often implemented as separate pre-processing step [3]
CSRM617	CSRM617, CAS:1848237-07-9, MF:C10H13N3O5, MW:255.23 g/mol	Chemical Reagent
NSC260594	NSC260594, CAS:906718-66-9, MF:C29H24N6O3, MW:504.5 g/mol	Chemical Reagent

Implementation Protocols

Protocol 1: Noise Learning with Mirror Circuits

Mirror circuits provide an effective methodology for benchmarking quantum devices and characterizing noise parameters [38]:

• Circuit Generation: Generate mirror circuits with varying depths using the device's native gate set and connectivity. For each circuit, determine the correct output bitstring.
• Noisy Execution: Execute each mirror circuit on the target quantum processor, collecting measurement statistics.
• Fidelity Calculation: For each circuit, compute the probability of measuring the correct bitstring.
• Error Model Fitting: Analyze the decay of fidelity with circuit depth to extract average error rates per gate or per layer.

The Mitiq library provides implementations for generating and analyzing mirror circuits [38]:

Protocol 2: PEC Representation Learning

Learning optimal quasiprobability representations is crucial for minimizing sampling overhead:

• Select Basis Operations: Choose a set of physically implementable operations ( {\mathcal{O}_\alpha} ) (typically Pauli gates).
• Characterize Noise Channels: Use quantum process tomography or gate set tomography to reconstruct the actual noise channel ( \Lambda ) for each gate.
• Solve Optimization Problem: Find coefficients ( \eta{i,\alpha} ) that satisfy ( \mathcal{G}i = \sum\alpha \eta{i,\alpha} \mathcal{O}\alpha ) while minimizing ( \gammai = \sum\alpha |\eta{i,\alpha}| ) [38].
• Validate Representation: Verify the accuracy of the representation through randomized benchmarking.

For a depolarizing noise model with error probability ( p ), the inverse channel can be represented using the same gate with error probability ( p/(1-p) ) [35].

Protocol 3: PEC Execution and Estimation

The core PEC protocol combines noisy circuit executions according to the quasiprobability distribution:

• Sample Circuits: For each gate in the original circuit, sample a replacement operation ( \mathcal{O}\alpha ) with probability ( |\eta{i,\alpha}|/\gammai ), keeping track of the sign ( \text{sgn}(\eta{i,\alpha}) ).
• Execute Noisy Circuits: Run each sampled circuit on the quantum processor, measuring the Hamiltonian expectation value.
• Combine Results: Compute the unbiased estimate using the formula: [ \langle H\rangle{\text{PEC}} = \frac{\prodi \gammai}{M} \sum{m=1}^M \text{sgn}m \langle H\ranglem ] where ( M ) is the total number of samples, ( \text{sgn}m ) is the product of signs for the m-th sample, and ( \langle H\ranglem ) is the measurement outcome [38].

The sampling process can be visualized as follows:

Performance Analysis

Quantitative Performance Metrics

Table 3: PEC Performance Across Different Molecular Systems

Molecular System	Qubit Count	Circuit Depth	Unmitigated Error (mHa)	PEC Error (mHa)	Sampling Overhead
H₂ (unencoded)	4	10	>1.6	~1.6	~10² [3]
H₂ (encoded)	4+	15+	>1.6	<1.6	~10³ [3]
H₂O	8-12	50-100	~10	~1.6	~10⁴ [8]
N₂	12-16	100-200	~20	~1.6	~10⁵ [8]

Comparative Analysis with Other Methods

PEC provides theoretical guarantees of unbiased estimation but comes with significant sampling costs. Recent research has explored hybrid approaches that combine PEC with other error mitigation techniques to balance accuracy and overhead:

• PEC + ZNE: Uses zero-noise extrapolation to reduce the number of circuit inversions required while maintaining accuracy [35].
• PEC + CVaR: Applies conditional value at risk to obtain provable bounds on expectation values with lower sampling overhead [39].
• MREM: Multireference error mitigation extends reference-based approaches to strongly correlated systems where traditional PEC might be prohibitively expensive [8].

The binomial PEC approach offers a middle ground by systematically controlling the bias-variance tradeoff [35]. Rather than insisting on completely bias-free estimation, this method allocates shots to different noisy circuits based on their weights, enabling researchers to target biases smaller than the achievable statistical noise.

Application to Quantum Chemistry

Hamiltonian Expectation Values

For quantum chemistry applications, the target observable is typically the molecular Hamiltonian ( H ) expressed as a sum of Pauli operators after the Jordan-Wigner or Bravyi-Kitaev transformation: [ H = \sum{\alpha} h\alpha P\alpha ] where ( P\alpha ) are Pauli operators and ( h_\alpha ) are coefficients determined by the molecular integrals [8].

The PEC protocol estimates each term ( \langle P\alpha \rangle ) independently, though correlated measurement techniques can reduce the total number of measurements required. The final energy estimate is computed as: [ E = \sum{\alpha} h\alpha \langle P\alpha \rangle_{\text{PEC}} ]

Case Study: Water Molecule

Simulating the water molecule requires 8-12 qubits and circuit depths of 50-100 layers depending on the ansatz choice [8]. The multireference error mitigation (MREM) approach, which builds upon PEC principles, has demonstrated significant improvements for this system:

• Reference State Preparation: Prepare a multireference state using Givens rotations to capture strong electron correlations.
• Noise Characterization: Quantify the effect of noise on the reference state through measurement on the quantum device.
• Error Extrapolation: Use the reference state behavior to mitigate errors in the target state simulation.

This approach demonstrates how PEC principles can be adapted to domain-specific challenges in quantum chemistry, particularly for strongly correlated systems where single-reference methods fail [8].

Probabilistic Error Cancellation provides a mathematically rigorous framework for obtaining unbiased estimates of Hamiltonian expectation values on noisy quantum devices. While the sampling overhead presents significant challenges, recent advances in binomial expansion, Pauli error propagation, and hybrid methods have substantially reduced these costs.

For quantum chemistry applications targeting drug development and materials design, PEC enables the calculation of molecular energies with chemical accuracy—a crucial milestone on the path to quantum utility. As quantum hardware continues to improve, with gate errors decreasing and qubit counts increasing, PEC will remain an essential component in the quantum simulation toolbox, potentially bridging the gap between NISQ devices and fault-tolerant quantum computation.

The integration of PEC with problem-specific approaches like multireference error mitigation demonstrates how domain knowledge can be leveraged to enhance error mitigation efficacy. This synergy between general-purpose quantum error mitigation and application-specific optimizations will likely drive further improvements in computational accuracy for quantum chemistry simulations.

Clifford Data Regression (CDR) is a learning-based quantum error mitigation technique designed to enhance the accuracy of expectation values obtained from noisy quantum computations. It is particularly vital for the execution of variational quantum algorithms on Noisy Intermediate-Scale Quantum (NISQ) hardware, where gate and decoherence noise significantly corrupt results without the resource overhead of full quantum error correction [40] [41]. The core principle of CDR is to leverage the fact that quantum circuits composed predominantly of Clifford gates can be efficiently simulated on classical computers, as per the Gottesman-Knill theorem [40] [8]. CDR operates by training a regression model on a set of "near-Clifford" training circuits. For these circuits, both the ideal (noise-free) expectation values and their noisy counterparts from quantum hardware can be obtained. The model learns the functional relationship between noisy and ideal outputs, and this learned mapping is then applied to mitigate errors in the far more complex, non-Clifford target circuit of interest [40] [41].

The utility of CDR is acutely demonstrated in quantum chemistry simulations, such as those performed with the Variational Quantum Eigensolver (VQE) to find molecular ground state energies. For these problems, which are beyond exact classical simulation for large systems, CDR offers a pathway to more reliable results without prohibitive sampling overheads [40]. This protocol details the application of CDR and its enhanced variants within the context of near-term quantum chemistry research.

Theoretical Foundation and Protocol Enhancements

Core CDR Methodology

The CDR protocol begins with the identification of a target circuit, such as a VQE ansatz (e.g., the tiled Unitary Product State or tUPS) for a molecule like H₄, which contains a non-trivial number of non-Clifford gates [40]. The goal is to mitigate the noise in the expectation value of the molecular Hamiltonian, ⟨H⟩, measured from this circuit.

The foundational CDR workflow involves several key steps [40] [41]:

• Generate Training Circuits: Create a set of classically simulable training circuits that structurally resemble the target circuit. This is typically done by replacing most, but not all, of the non-Clifford gates in the target circuit with Clifford gates.
• Compute Ideal Values: Use efficient classical Clifford simulators to calculate the exact, noise-free expectation values for the observable of interest for each training circuit.
• Collect Noisy Data: Execute the same set of training circuits on the noisy quantum processor (or a high-fidelity noise model simulator) to obtain the corresponding noisy expectation values.
• Train Regression Model: Using the paired data (noisy value, ideal value), train a simple linear regression model. The model learns a mapping ( f ) such that ( \text{ideal} \approx f(\text{noisy}) ).
• Mitigate Target Circuit: Execute the target non-Clifford circuit on the quantum device to get a noisy expectation value. Apply the trained regression model to this value to obtain the error-mitigated estimate.

Enhanced CDR Protocols

Recent research has introduced significant improvements to the core CDR protocol, enhancing its accuracy and efficiency for chemistry simulations. Two notable enhancements are Energy Sampling and Non-Clifford Extrapolation [40].

• Energy Sampling (ES): This improvement addresses the quality of training data. Instead of using all generated training circuits, the ES protocol filters them based on their noiseless energy. Only the training circuits whose ideal energies are closest to the (estimated) ground state energy of the target molecule are selected for regression. This biases the learning process towards circuits whose quantum states are more physically relevant, improving the accuracy of the noise mapping for the target state [40].
• Non-Clifford Extrapolation (NCE): This enhancement improves the regression model itself. The standard CDR uses a single input feature—the noisy expectation value. The NCE protocol incorporates an additional input feature: the number of non-Clifford parameters in the training circuit. This allows the model to learn how the noisy-ideal relationship evolves as the circuit's structure approaches that of the target circuit, effectively enabling an extrapolation in "non-Cliffordness" [40].

Table 1: Key Enhancements to Clifford Data Regression

Protocol	Core Idea	Application in Chemistry Simulations	Key Benefit
Standard CDR [41]	Linear regression on noisy vs. ideal data from near-Clifford circuits.	Mitigating energy measurements in VQE for small molecules.	Reduces bias from noise without full error correction.
Energy Sampling (ES) [40]	Selects training circuits with energies near the target ground state.	Focusing mitigation on chemically relevant states in Hâ‚„ simulations.	Improves model accuracy by using physically meaningful training data.
Non-Clifford Extrapolation (NCE) [40]	Uses the number of non-Clifford gates as an additional regression feature.	Capturing how noise effects change with ansatz complexity in tUPS.	Enables better generalization from Clifford-dominated to target circuits.

Experimental Protocols and Workflows

Detailed Protocol for Enhanced CDR in VQE

This protocol outlines the steps for applying Energy Sampling and Non-Clifford Extrapolation to mitigate errors in a VQE energy calculation.

Objective: To compute the error-mitigated ground state energy of a molecule (e.g., Hâ‚„) using a specific ansatz (e.g., tUPS) and a noise model (e.g., ibm_torino).

Materials and Prerequisites:

• Molecular Hamiltonian (H) in qubit representation.
• Optimized parameters (θ*) for the target VQE ansatz circuit.
• A high-fidelity quantum device or a simulated noise model.

Procedure:

• Circuit Preparation:

  • Target Circuit: Construct the full VQE ansatz circuit ( U(\theta^*) ) with the optimized parameters.
  • Training Circuit Generation: Create a pool of ( N ) (e.g., 100) training circuits. For the tUPS ansatz, this can be achieved by generating new parameter sets ( {\theta_i} ) where a fraction of the parameters are set to zero (which can create Clifford gates) or other values that simplify the circuit, while a small number retain non-Clifford character [40].

• Classical Data Collection (Ideal Values):

  • For each training circuit ( i ) in the pool, use a classical Clifford simulator to compute the exact expectation value ( \langle H \rangle_i^{\text{ideal}} ).

• Energy Sampling (Filtering):

  • Sort the pool of training circuits based on their computed ideal energies ( \langle H \rangle_i^{\text{ideal}} ).
  • Select the ( n ) circuits (e.g., ( n=20 )) with the lowest energies to form the final, filtered training set. This set is biased towards states proximate to the true ground state.

• Noisy Data Collection:

  • For each of the ( n ) selected training circuits, execute it on the noisy quantum backend (or simulator) to obtain the noisy expectation value ( \langle H \rangle_i^{\text{noisy}} ).

• Feature Engineering for NCE:

  • For each of the ( n ) training circuits, calculate ( k_i ), the number of non-Clifford parameters (or gates) in the circuit.

• Model Training:

  • Train a linear regression model (e.g., ( y = a x + b )) using the ( n ) data points. For standard CDR, the input feature ( x ) is ( \langle H \ranglei^{\text{noisy}} ) and the target variable ( y ) is ( \langle H \ranglei^{\text{ideal}} ).
  • For NCE-enhanced CDR, use a multiple linear regression model (e.g., ( y = a1 x1 + a2 x2 + b )), where ( x1 ) is ( \langle H \ranglei^{\text{noisy}} ) and ( x2 ) is ( ki ).

• Target Execution and Mitigation:

  • Execute the full target circuit ( U(\theta^*) ) on the noisy backend to obtain ( \langle H \rangle_{\text{target}}^{\text{noisy}} ).
  • Let ( k_{\text{target}} ) be the number of non-Clifford parameters in the target circuit.
  • Apply the trained model to produce the final mitigated energy:
    • Standard CDR: ( \langle H \rangle{\text{mitigated}} = f(\langle H \rangle{\text{target}}^{\text{noisy}}) )
    • NCE-enhanced CDR: ( \langle H \rangle{\text{mitigated}} = f(\langle H \rangle{\text{target}}^{\text{noisy}}, k_{\text{target}}) )

The following workflow diagram illustrates this enhanced protocol, integrating both Energy Sampling and Non-Clifford Extrapolation.

Performance Data and Comparison

Numerical simulations on the Hâ‚„ molecule using the tUPS ansatz and the ibm_torino noise model demonstrate the superiority of the enhanced protocols. The following table summarizes typical performance data, showing the reduction in energy error achieved by each method compared to the unmitigated result [40].

Table 2: Performance Comparison of CDR Protocols on an Hâ‚„ Simulation

Mitigation Method	Energy Error (Absolute)	Relative Improvement over Unmitigated	Notes on Experimental Setup
Unmitigated	Reference Error	--	Baseline from noisy quantum device/simulator.
Standard CDR [40] [41]	~50% reduction	2x	Performance depends on training circuit selection.
CDR + Energy Sampling (ES) [40]	>50% reduction	>2x	More accurate than standard CDR by biasing training data.
CDR + Non-Clifford Extrapolation (NCE) [40]	Largest reduction	>2x	Outperforms standard CDR by learning noise vs. circuit complexity.

The Scientist's Toolkit

This section details the essential "research reagents" and tools required to implement CDR for quantum chemistry simulations.

Table 3: Essential Tools and Resources for CDR Experiments

Tool / Resource	Function / Purpose	Example Solutions
Classical Clifford Simulator	Computes exact, noise-free expectation values for the training circuits.	Modules in Qiskit, Cirq, or specialized high-performance simulators.
Quantum Backend	Provides noisy expectation values for both training and target circuits.	IBM Quantum processors (e.g., ibm_torino), QuEra's Gemini, or high-fidelity noise models in emulators [40] [42].
Circuit Generator	Creates the pool of near-Clifford training circuits from the target circuit.	Custom scripts that parameterize an ansatz and constrain parameters to Clifford values.
Regression Model	Learns the mapping from noisy to ideal expectation values.	Linear or multiple linear regression models from standard libraries (e.g., scikit-learn).
Molecular Ansatz	The parameterized quantum circuit for preparing the trial molecular wavefunction.	tUPS [40], UCCSD, or other hardware-efficient ansätze.
Noise Model	A software representation of device noise for simulation-based validation.	Built-in noise models in Qiskit Aer or custom models (e.g., QuEra's GeminiTwoZoneNoiseModel [42]).
(R)-VT104	N-[(1R)-1-(pyridin-2-yl)ethyl]-5-[4-(trifluoromethyl)phenyl]naphthalene-2-carboxamide	Get N-[(1R)-1-(pyridin-2-yl)ethyl]-5-[4-(trifluoromethyl)phenyl]naphthalene-2-carboxamide for your research. This small molecule is For Research Use Only (RUO), not for human or veterinary diagnosis or therapy.
Z-DEVD-R110	Z-DEVD-R110, MF:C72H78N10O27, MW:1515.4 g/mol	Chemical Reagent

Integration with Broader Research Context

CDR is one of several strategies being developed to combat errors on NISQ devices. Its learning-based approach contrasts with other methods like Zero-Noise Extrapolation (ZNE), which requires executing the same circuit at elevated noise strengths, and Probabilistic Error Cancellation (PEC), which necessitates detailed noise characterization [40] [43] [44]. A key advantage of CDR is that it is noise-model agnostic; it learns the effect of noise directly from data without requiring an a priori model [43] [41].

CDR can also be synergistically combined with other error mitigation techniques. For instance, measurement error mitigation is often applied as a pre-processing step before feeding data into the CDR regression model [45]. Furthermore, the principles of CDR have inspired other learning-based approaches, such as noise-agnostic neural models trained with data augmentation (DAEM) [43].

However, researchers must be aware of the fundamental challenges facing all error mitigation techniques, including CDR. Theoretical studies indicate that as quantum circuits grow in size and depth, the sampling overhead—the number of circuit repetitions required—can grow exponentially, potentially imposing a hard scalability limit [44]. Therefore, while CDR and its enhanced variants represent powerful tools for near-term quantum chemistry experiments, their practical utility for large-scale problems remains an active and critical area of research.

Reference-State and Multireference-State Error Mitigation for Strong Electron Correlation

Quantum error mitigation (QEM) has emerged as a crucial set of algorithmic techniques for improving the precision and reliability of quantum chemistry calculations on noisy intermediate-scale quantum (NISQ) devices. Without the extensive qubit overhead required for full quantum error correction, QEM strategies reduce noise-induced biases in expectation values through sophisticated post-processing of outputs from ensemble circuit runs [46]. For quantum chemistry applications, where calculating molecular ground state energies with chemical accuracy (approximately 1 kcal/mol) is often the goal, these techniques are particularly valuable [46].

Reference-state error mitigation (REM) represents a cost-effective, chemistry-inspired QEM approach that performs well for weakly correlated problems [8]. However, its effectiveness becomes limited when applied to strongly correlated systems, where the exact wavefunction often takes the form of a multireference state—a linear combination of multiple Slater determinants with similar weights [8]. This limitation arises because REM assumes that a single reference state (typically Hartree-Fock) provides a reasonable approximation of the target ground state, an assumption that fails in strongly correlated regimes like bond-dissociation regions [8].

This application note explores the extension of REM to multireference-state error mitigation (MREM), which systematically captures quantum hardware noise in strongly correlated ground states by utilizing multireference states. We provide detailed protocols for implementing these methods, along with performance data and visualization tools to guide researchers in applying these techniques to challenging quantum chemistry problems.

Theoretical Foundation

The Challenge of Strong Electron Correlation

Strong electron correlation presents significant challenges for both classical computational methods and quantum algorithms. In transition metal compounds, which exhibit phenomena such as high-temperature superconductivity, colossal magnetoresistance, and multiferroicity, the strongly correlated d- or f-electron shells require treatment beyond the mean-field approximation [47]. These effects cannot be described by finite orders of perturbation theory due to the macroscopically large degeneracy of the unperturbed state, necessitating summation to infinite order or non-perturbative methods [47].

The fundamental challenge is illustrated even in simple systems like the hydrogen molecule. In mean-field theory, electrons move independently, with equal probability of finding them on the same site or different sites—indicating absence of correlation in electron motion. Going beyond mean-field approximation requires accounting for residual electron-electron interactions, leading to Hubbard-type Hamiltonians that better capture strong correlation effects [47].

Reference-State Error Mitigation (REM)

REM leverages chemical insight to provide a low-complexity error mitigation approach. The core idea is to mitigate the energy error of a noisy target state measured on a quantum device by first quantifying the effect of noise on a close-lying reference state [8]. This reference state must be: (i) exactly solvable on a classical computer, and (ii) practical to prepare and measure on a quantum device.

The REM protocol can be summarized as:

• Classically compute the exact energy of the reference state, ( E_{\text{ref}}^{\text{exact}} )
• Prepare and measure the reference state on the quantum device to obtain its noisy energy, ( E_{\text{ref}}^{\text{noisy}} )
• Prepare and measure the target state (e.g., via VQE) to obtain its noisy energy, ( E_{\text{target}}^{\text{noisy}} )
• Compute the mitigated energy as: ( E{\text{mitigated}} = E{\text{target}}^{\text{noisy}} - (E{\text{ref}}^{\text{noisy}} - E{\text{ref}}^{\text{exact}}) )

For weakly correlated systems, the Hartree-Fock state often serves as an effective reference, maintaining sufficient overlap with the target ground state [8]. However, for strongly correlated systems, this single-determinant approximation fails, necessitating a multireference approach.

Multireference-State Error Mitigation (MREM)

MREM extends the REM framework by incorporating multiconfigurational states with better overlap to correlated target wavefunctions [8]. A pivotal aspect of MREM is using Givens rotations to efficiently construct quantum circuits that generate multireference states from a single reference configuration while preserving key symmetries such as particle number and spin projection [8].

To balance circuit expressivity and noise sensitivity, MREM employs compact wavefunctions composed of a few dominant Slater determinants. These truncated multireference states are engineered to exhibit substantial overlap with the target ground state, enhancing error mitigation in variational quantum eigensolver experiments for strongly correlated systems [48] [8].

Performance Comparison and Data

The following tables summarize key performance data for REM and MREM across different molecular systems, demonstrating the advantages of the multireference approach for strongly correlated cases.

Table 1: Error Mitigation Performance Across Molecular Systems

Molecule	Correlation Type	Method	Energy Error (Unmitigated)	Energy Error (Mitigated)	Improvement Factor
H₂O	Weak	REM	12.3 mHa	2.1 mHa	5.9×
H₂O	Weak	MREM	12.3 mHa	1.8 mHa	6.8×
N₂	Intermediate	REM	18.7 mHa	6.4 mHa	2.9×
N₂	Intermediate	MREM	18.7 mHa	3.2 mHa	5.8×
F₂	Strong	REM	42.5 mHa	28.9 mHa	1.5×
F₂	Strong	MREM	42.5 mHa	8.7 mHa	4.9×

Table 2: Computational Overhead Comparison

Method	Additional Classical Cost	Additional Quantum Cost	Measurement Overhead	Scalability
REM	Low (single-state energy)	Low (one additional state)	Minimal	Excellent
MREM	Moderate (MR state selection)	Moderate (Givens circuits)	Moderate	Good
Extrapolation	None	High (multiple noise levels)	High	Limited
PEC	High (noise characterization)	High (corrected circuits)	Very high	Poor

Table 3: Dynamical Correlation Treatment via NEVPT2

Method	Active Electrons	Active Orbitals	SC-NEVPT2 Correction	Total Energy Error
VQE-SCF	4	4	-0.152 Eh	8.3 mHa
VQE-SCF	8	8	-0.241 Eh	5.1 mHa
VQE-SCF	12	10	-0.385 Eh	3.2 mHa
FCI	8	8	-0.238 Eh	Reference

Experimental Protocols

Protocol 1: Standard REM Implementation

Objective: Mitigate errors in VQE calculations for weakly correlated systems using Hartree-Fock reference state.

Materials and Requirements:

• Quantum device or simulator with VQE capabilities
• Classical computer for Hartree-Fock calculation
• Quantum circuit ansatz for target state preparation

Procedure:

• Classical Hartree-Fock Calculation:
  • Perform restricted Hartree-Fock calculation for target molecule
  • Store exact HF energy: ( E_{\text{HF}}^{\text{exact}} )
  • Prepare HF state quantum circuit using Pauli-X gates

• Quantum Hardware Steps:

  • Prepare HF state on quantum device: ( |\psi{\text{HF}}\rangle = \prod{i} X_i |0\rangle )
  • Measure Hamiltonian expectation value: ( E{\text{HF}}^{\text{noisy}} = \langle \psi{\text{HF}} | \hat{H} | \psi_{\text{HF}} \rangle )
  • Run VQE optimization to prepare target state ( |\psi_{\text{target}}\rangle )
  • Measure noisy target energy: ( E_{\text{target}}^{\text{noisy}} )

• Error Mitigation:

  • Compute energy shift: ( \Delta E = E{\text{HF}}^{\text{noisy}} - E{\text{HF}}^{\text{exact}} )
  • Apply mitigation: ( E{\text{mitigated}} = E{\text{target}}^{\text{noisy}} - \Delta E )

Validation:

• Compare mitigated energy with classical reference (e.g., FCI, CCSD(T))
• Calculate improvement factor: ( \text{IF} = \frac{|E{\text{unmitigated}} - E{\text{exact}}|}{|E{\text{mitigated}} - E{\text{exact}}|} )

Protocol 2: MREM for Strongly Correlated Systems

Objective: Mitigate errors in VQE calculations for strongly correlated systems using multireference states.

Materials and Requirements:

• Quantum device or simulator with support for Givens rotation circuits
• Classical computer for multireference state selection (e.g., CASCI, DMRG)
• Quantum circuit ansatz for target state preparation

Procedure:

• Multireference State Selection:
  • Perform inexpensive classical multireference calculation (e.g., CASSCF with small active space)
  • Identify dominant Slater determinants with largest weights
  • Select truncated set of determinants (typically 3-10) balancing accuracy and circuit complexity

• Givens Rotation Circuit Construction:

  • For each selected determinant, compute unitary transformation from reference determinant
  • Decompose transformation into sequence of Givens rotations
  • Compile Givens rotations into quantum gates (typically two-qubit rotations)

• Quantum Hardware Steps:

  • Prepare each multireference state ( |\psi_{\text{MR}}^i\rangle ) using Givens circuits
  • Measure noisy energies: ( E{\text{MR}}^{i,\text{noisy}} = \langle \psi{\text{MR}}^i | \hat{H} | \psi_{\text{MR}}^i \rangle )
  • Compute exact classical energies for each multireference state: ( E_{\text{MR}}^{i,\text{exact}} )
  • Run VQE optimization for target state
  • Measure noisy target energy: ( E_{\text{target}}^{\text{noisy}} )

• Multireference Error Mitigation:

  • Compute average energy shift: ( \Delta E{\text{MR}} = \frac{1}{N} \sum{i=1}^N (E{\text{MR}}^{i,\text{noisy}} - E{\text{MR}}^{i,\text{exact}}) )
  • Apply mitigation: ( E{\text{mitigated}} = E{\text{target}}^{\text{noisy}} - \Delta E_{\text{MR}} )

Validation:

• Compare with full configuration interaction or DMRG reference
• Assess performance across bond dissociation curves
• Evaluate stability with respect to number of reference states

Protocol 3: Integrating Dynamical Correlation via SC-NEVPT2

Objective: Combine VQE with strongly-contracted N-electron valence state perturbation theory (SC-NEVPT2) to capture dynamical correlation.

Materials and Requirements:

• Quantum device with IC-POVM measurement capabilities
• Classical computer for NEVPT2 computation
• Pre-computed one- and two-electron integrals

Procedure:

• VQE-SCF Active Space Calculation:
  • Perform VQE self-consistent field calculation for active space
  • Obtain active space wavefunction and energy

• Informationally Complete Measurements:

  • Implement adaptive IC-POVM scheme for ground state measurement
  • "Recycle" measurement outcomes for higher-order reduced density matrices (RDMs)
  • Reconstruct three- and four-body RDMs through classical post-processing

• Classical NEVPT2 Correction:

  • Use higher-order RDMs to compute SC-NEVPT2 energy correction
  • Combine with active space energy: ( E{\text{total}} = E{\text{VQE-SCF}} + E_{\text{SC-NEVPT2}} )

Validation:

• Compare with conventional NEVPT2 results
• Assess measurement efficiency gains from IC-POVM approach
• Evaluate size consistency for multiple molecules

Workflow Visualization

The Scientist's Toolkit

Table 4: Essential Research Reagents and Computational Resources

Resource	Type	Function/Purpose	Example Implementations
Givens Rotation Circuits	Quantum subroutine	Construct multireference states from reference determinant	OpenFermion, Qiskit Nature
IC-POVM Measurements	Measurement protocol	Efficiently measure higher-order reduced density matrices	AIM, Google Quantum AI
VQE-SCF	Hybrid algorithm	Self-consistent field optimization using quantum resources	QEMIST Cloud, QCOR
SC-NEVPT2	Classical module	Compute dynamical correlation energy correction	PySCF, ORCA, BAGEL
Active Space Selection	Pre-processing tool	Identify strongly correlated orbitals	AVAS, DMRG-SCF, ASCI
Quantum Simulators	Development tool	Test and validate protocols without quantum hardware	Qiskit Aer, Cirq, Strawberry Fields
FITC-YVADAPK(Dnp)	FITC-YVADAPK(Dnp), MF:C62H67N11O20S, MW:1318.3 g/mol	Chemical Reagent	Bench Chemicals
Protonstatin-1	5-(2-Furylmethylene)-2-thioxo-1,3-thiazolidin-4-one		Bench Chemicals

The advancement from reference-state to multireference-state error mitigation represents a significant step forward in enabling accurate quantum chemistry simulations on NISQ devices. By systematically addressing the limitations of single-reference approaches in strongly correlated systems, MREM broadens the scope of tractable molecular problems while maintaining feasible resource requirements. The integration of these error mitigation strategies with perturbative treatments of dynamical correlation creates a comprehensive framework for tackling challenging electronic structure problems across chemical and materials science domains.

As quantum hardware continues to improve, these error mitigation protocols will play an increasingly important role in bridging the gap between algorithmic development and practical application, potentially enabling quantum advantage in computational chemistry for drug development and materials design.

Quantum error mitigation (QEM) has emerged as a crucial set of techniques for extracting reliable results from noisy intermediate-scale quantum (NISQ) devices. Unlike resource-intensive quantum error correction (QEC), QEM techniques reduce the impact of noise without requiring additional qubits for full fault tolerance, making them particularly suitable for the current era of quantum computing [49]. However, as quantum circuits grow in size and complexity, individual error mitigation strategies face significant limitations, including exponential sampling overhead and decreased effectiveness for larger systems [44] [50].

This application note explores a hybrid approach that integrates the [[n, n-2, 2]] quantum error detection code (QEDC) with probabilistic error cancellation (PEC) to overcome the limitations of individual techniques. By leveraging the complementary strengths of both methods, this protocol suppresses errors in near-term quantum simulations more effectively than either method alone, with particular relevance for quantum chemistry applications such as variational quantum eigensolver (VQE) circuits for molecular ground state energy estimation [22] [51].

The fundamental insight behind this hybrid protocol is that quantum error detection can filter out a significant portion of detectable errors through post-selection, thereby creating a effectively "cleaner" quantum channel. When PEC is subsequently applied to this post-selected output, its sampling overhead is substantially reduced because it only needs to mitigate the remaining undetectable errors, rather than the full noise of the unencoded circuit [22] [52]. This synergistic combination enables more accurate quantum computations while managing the resource overhead that typically plagues error mitigation techniques.

Theoretical Foundation

Quantum Error Detection Codes (QEDC)

The [[n, n-2, 2]] quantum error detection code is a stabilizer code that utilizes an even number of physical qubits (n) to encode a smaller number of logical qubits (n-2). This code is defined by two non-local Pauli stabilizers: X⊗n and Z⊗n. The codespace corresponds to the joint +1 eigenspace of these two operators [22].

The structure of the [[n, n-2, 2]] code offers several practical advantages for near-term implementation:

• Constant qubit overhead: The encoding requires only a constant overhead of two additional physical qubits regardless of the number of logical qubits.
• Simple syndrome measurement: Error detection requires measuring only the first two qubits after applying the decoding circuit.
• Efficient logical operations: The code enables straightforward implementation of Clifford gates without complex compilation [22].

A critical property of this code is its ability to detect any single-qubit error, though it cannot correct them. When errors are detected through syndrome measurement, the affected runs are discarded through post-selection, effectively filtering out a significant portion of the noise that would otherwise corrupt the computation.

Probabilistic Error Cancellation (PEC)

Probabilistic error cancellation is an error mitigation technique that reconstructs noiseless expectation values from a set of noisy quantum operations. The fundamental principle behind PEC is representing the ideal quantum operation as a linear combination of implementable noisy operations [22] [50].

For an ideal unitary operation [U], PEC finds a quasi-probability decomposition:

where ε_i are noisy operations that can be implemented on the actual quantum device, and q_i are real coefficients that form a quasi-probability distribution (Σi qi = 1, but individual q_i can be negative) [50].

The implementation of PEC involves:

• Noise characterization: Determining the actual noise channels affecting the quantum device through techniques like gate set tomography.
• Quasi-probability decomposition: Finding the appropriate set {q_i, ε_i} that approximates the ideal operation.
• Monte Carlo sampling: Sampling from the quasi-probability distribution to estimate the ideal expectation value.

The main drawback of PEC is its sampling overhead, which scales as (Σ_i |q_i|)^2, where Σ_i |q_i| is the negation factor representing how much the quasi-probability distribution deviates from a proper probability distribution [50].

Hybrid Protocol Methodology

Integrated Workflow

The hybrid protocol integrates QEDC and PEC through a structured workflow that leverages the strengths of both techniques while mitigating their individual limitations. The complete procedure consists of sequential stages that transform the original quantum circuit into an error-resilient form, execute it with appropriate error detection, and apply classical post-processing to mitigate remaining errors [22] [52].

Table 1: Key Stages of the Hybrid Protocol

Stage	Description	Technique Applied
Circuit Encoding	Encode logical qubits using [[n, n-2, 2]] code	Quantum Error Detection
Pauli Twirling	Apply custom partial twirling to simplify noise	Noise Characterization
Circuit Execution	Execute multiple shots of the encoded circuit	Quantum Processing
Post-selection	Discard runs where errors are detected	Classical Processing
Error Cancellation	Apply PEC to mitigate undetectable errors	Probabilistic Error Cancellation

Circuit Encoding and Decoding

The encoding process for the [[n, n-2, 2]] QEDC follows a specific structure. For a system with n physical qubits labeled {q_x, q_z, q_k, q_{k-1}, ..., q_2, q_1} where j represents the j-th logical qubit:

• Logical state preparation: The encoding circuit transforms the unencoded state of n-2 qubits into an encoded state of n qubits living in the codespace of the QEDC.
• Logical operations: Quantum gates are applied directly to the encoded state. Logical Pauli operators are defined as XÌ„_j = X_{q_z}X_{q_j}, ZÌ„_j = Z_{q_x}Z_{q_j}, and YÌ„_j = iXÌ„_jZÌ„_j.
• Decoding and error detection: The inverse of the encoding circuit is applied, followed by measurement of the first two qubits. Any measurement of -1 in these qubits indicates a detectable error, and the run is discarded [22].

This encoding scheme is particularly valuable for near-term devices because it provides error detection capability with minimal qubit overhead and straightforward implementation of logical operations.

Partial Pauli Twirling

A key innovation in the hybrid protocol is the use of partial Pauli twirling to reduce the sampling overhead associated with PEC. Traditional Pauli twirling converts general noise channels into Pauli noise channels by conjugating gates with random Pauli operators, but this requires a large number of Pauli operators in the twirling set [22].

The modified approach:

• Reduces the twirling set size by strategically selecting a subset of Pauli operators
• Simplifies the effective noise model while maintaining error suppression capabilities
• Decreases the sampling cost of subsequent PEC by reducing the negation factor

This partial twirling approach creates a noise channel that is more amenable to characterization and mitigation through PEC, while requiring fewer circuit variants than full Pauli twirling [22].

Error Detection and PEC Integration

The integration of error detection with PEC follows a specific sequence:

• Execute the encoded circuit multiple times on the quantum processor
• Perform post-selection by discarding runs where error detection flags indicate detectable errors
• Apply PEC to the remaining post-selected data to mitigate undetectable errors

This sequence is crucial because the error detection step removes a significant portion of the noise, resulting in a effectively "cleaner" quantum channel with reduced error rates. When PEC is applied to this improved channel, its sampling overhead is substantially lower because the negation factor Σ_i |q_i| is smaller for channels with lower error rates [52].

Experimental Implementation

VQE for Hâ‚‚ Ground State Energy Estimation

The hybrid protocol has been demonstrated experimentally using a variational quantum eigensolver (VQE) circuit that estimates the ground state energy of a hydrogen molecule (Hâ‚‚). This application is particularly relevant for quantum chemistry simulations, where accurate energy estimation is crucial for predicting molecular properties and reaction dynamics [22] [51].

Table 2: Experimental Parameters for Hâ‚‚ VQE Implementation

Parameter	Specification	Purpose
Quantum Processor	IBM ibm_brussels	Hardware execution
Simulator	Qiskit AerSimulator	Noise model verification
Encoding	[[4, 2, 2]] QEDC	Error detection
Logical Qubits	2	Molecular orbital representation
Physical Qubits	4	Hardware resource requirement
Circuit Type	Non-Clifford VQE	Quantum chemistry application

The experimental implementation involved comparing several scenarios:

• Unprotected circuit: Baseline VQE without error suppression
• QEDC only: VQE with error detection but no PEC
• PEC only: VQE with error cancellation but no error detection
• Hybrid protocol: Combined QEDC and PEC approach

Results demonstrated that the hybrid protocol achieved significantly improved accuracy in ground state energy estimation compared to either individual technique, with the combination providing mutual benefits that enhanced overall error suppression [22] [53].

The Scientist's Toolkit: Research Reagent Solutions

Table 3: Essential Components for Hybrid Protocol Implementation

Component	Function	Implementation Notes
[[n, n-2, 2]] QEDC	Encodes logical qubits with error detection capability	Requires n physical qubits for n-2 logical qubits
Pauli Twirling Set	Simplifies noise characteristics for PEC	Custom partial set reduces sampling overhead
PEC Quasi-Probability Decomposition	Represents ideal operations as noisy implementable ones	Characterized via gate set tomography
Syndrome Measurement Circuit	Detects errors in encoded states	Measured on first two qubits after decoding
Classical Post-processing	Implements PEC and error analysis	Computes error-mitigated expectation values
Leucodelphinidin	Leucodelphinidin, CAS:12764-74-8, MF:C15H14O8, MW:322.27 g/mol	Chemical Reagent

Performance Analysis and Scalability

Error Scaling Behavior

The performance advantage of the hybrid protocol becomes particularly evident when examining the scaling behavior of errors with increasing circuit size. Theoretical and experimental analyses reveal distinct scaling regimes for different error suppression techniques [50]:

• Unmitigated errors typically scale as O(εN), where ε is the single-gate error rate and N is the number of gates
• PEC alone can reduce this to O(ε'√N) under certain conditions, where ε' is a protocol-dependent factor
• The hybrid protocol demonstrates further improved scaling by combining the error reduction from both detection and mitigation

This improved scaling behavior is attributed to the error detection step removing a significant fraction of errors before PEC is applied, thereby reducing the burden on the error cancellation component of the protocol [22] [50].

Sampling Overhead Considerations

A critical practical consideration for near-term quantum applications is the sampling overhead required to achieve accurate results. The hybrid protocol provides advantages in this regard through several mechanisms:

• Reduced PEC overhead: By lowering the effective error rate through detection, the hybrid protocol reduces the negation factor Σ_i |q_i| in the PEC component, thereby decreasing its sampling cost [52].

• Balanced resource allocation: The protocol strategically allocates resources between error detection (which discards shots) and error mitigation (which requires more shots), finding an optimal balance for overall efficiency.

• Circuit-dependent advantages: For circuits with specific noise characteristics, the hybrid approach can achieve comparable accuracy to PEC alone with significantly fewer total shots, despite the shot discarding from error detection [22].

However, it is important to note that all error mitigation techniques face fundamental limitations, with worst-case sampling costs that can grow exponentially with circuit size for general quantum computations [44]. The hybrid protocol mitigates but does not eliminate these fundamental constraints.

Protocol Workflow Visualization

The hybrid protocol combining [[n, n-2, 2]] quantum error detection with probabilistic error cancellation represents a promising approach for enhancing the reliability of near-term quantum simulations. By leveraging the complementary strengths of both techniques, this method achieves improved error suppression while managing the sampling overhead that typically limits practical application of error mitigation [22] [52].

For researchers in quantum chemistry and drug development, this protocol offers a practical pathway for obtaining more accurate molecular simulations on current quantum hardware. The demonstrated application to Hâ‚‚ ground state energy estimation provides a template for extending this approach to more complex molecular systems, potentially enabling more reliable prediction of molecular properties and reaction mechanisms [22] [54].

Future developments in this area will likely focus on:

• Adaptive techniques that optimize the balance between error detection and mitigation based on circuit characteristics
• Integration with other error mitigation methods such as zero-noise extrapolation or virtual distillation
• Application to larger molecular systems beyond diatomic molecules
• Hardware-specific optimizations that account for the particular noise characteristics of different quantum computing platforms

As quantum hardware continues to improve, hybrid error suppression protocols of this type will play an increasingly important role in bridging the gap between current noisy devices and future fault-tolerant quantum computers, ultimately enabling practical quantum advantage in chemical simulation and other application domains.

Accurately estimating the ground-state energy of molecules is a fundamental challenge in quantum chemistry with significant implications for drug discovery and materials science. Classical computational methods often struggle with the exponential scaling required to simulate strongly correlated quantum systems. Quantum computers offer a promising path forward, particularly through near-term algorithms like the Variational Quantum Eigensolver (VQE). However, current noisy intermediate-scale quantum (NISQ) devices face significant limitations from quantum decoherence and gate errors, making error mitigation and optimized protocol design essential for obtaining chemically relevant results [49].

This article presents detailed application notes and protocols for ground-state energy estimation of three benchmark molecules: Hâ‚‚, Hâ‚‚O, and Nâ‚‚. These case studies exemplify the practical implementation of quantum computational chemistry on near-term quantum hardware, emphasizing error-aware experimental design. We provide consolidated quantitative data, step-by-step methodologies, and resource specifications to enable researchers to replicate and build upon these foundational experiments.

The following tables consolidate key experimental and computational results for the ground-state energy estimation of Hâ‚‚, Hâ‚‚O, and Nâ‚‚ from the surveyed literature, providing a reference for benchmarking quantum computational methods.

Table 1: Experimental Molecular Properties for Ground-State Energy Reference

Molecule	Bond Length (Å)	Vibrational Zero-Point Energy (cm⁻¹)	Dissociation Energy (eV)	Experimental Ground State	Reference
Hâ‚‚	0.741 [55]	-	~4.75 [55]	-1.13619 Ha (-30.917 eV) [55]	[55]
Hâ‚‚O	0.957 (O-H) [56]	-	-	-	[56]
N₂	1.098 [57]	1165.0 [57]	-	1Σg [57]	[57]

Table 2: Quantum Computational Results for Ground-State Energy Estimation

Molecule	Method	Hardware/Simulation	Estimated Energy	Accuracy/Target	Key Metrics	Reference
Hâ‚‚	VQE	NMR Quantum Processor	Full spectrum calculated	High accuracy with single qubit	1 qubit used	[58]
Hâ‚‚O	VQE (UCC ansatz)	Trapped-Ion QC	Near chemical accuracy	~1.6 mHa from exact	13 CNOTs for core circuit	[56]
H₂O	Quantum Computed Moments (QCM)	Superconducting Processor	Within 1.4±1.2 mHa of exact	Chemically relevant (c.0.1%)	8 qubits, 22 CNOTs	[59]
Nâ‚‚	Numerical MC SCF & CI	Classical (Reference)	-	Dissociation energy increased by 0.17 eV (MC SCF) & 0.08 eV (CI)	Classical benchmark	[60]

Experimental Protocols & Methodologies

Case Study 1: Hâ‚‚ Molecule using VQE on an NMR Quantum Simulator

Protocol Objective: To calculate the ground and excited state energies of the Hâ‚‚ molecule using a variational quantum eigensolver (VQE) algorithm on a nuclear magnetic resonance (NMR) quantum simulator.

Background: The Hâ‚‚ molecule serves as the primary benchmark for quantum chemistry algorithms due to its simplicity. The VQE approach hybridizes a quantum computer, which prepares and measures a parameterized trial wavefunction (ansatz), with a classical optimizer that adjusts parameters to minimize the expected energy [58].

Step-by-Step Protocol:

• Problem Mapping:

  • Map the electronic structure of Hâ‚‚, derived in a minimal basis set (e.g., STO-3G), to a qubit Hamiltonian using the Jordan-Wigner (JW) or Bravyi-Kitaev transformation. This results in a 2-qubit Hamiltonian after accounting for symmetries that allow the removal of two qubits [58].
  • For an even more resource-efficient simulation, further reduce the problem to a single qubit by exploiting the molecule's symmetry. This simplified approach has been demonstrated to calculate molecular energies to the desired accuracy [58].

• Ansatz Preparation:

  • Choose a parameterized quantum circuit (ansatz) capable of representing the ground state. A common choice is the Unitary Coupled Cluster with Singles and Doubles (UCCSD) ansatz, which for Hâ‚‚ can be simplified to a circuit with rotation gates and entangling operations.

• Quantum Execution:

  • Initialize the qubits, typically in the Hartree-Fock state (e.g., |01⟩).
  • Execute the parameterized ansatz circuit on the NMR quantum processor.
  • Measure the expectation values of the terms in the qubit Hamiltonian. This often requires measuring Pauli operators (X, Y, Z) on different qubits, which can be accomplished by applying appropriate single-qubit rotations before measurement in the computational basis.

• Classical Optimization:

  • A classical optimizer (e.g., gradient descent, SPSA) receives the computed energy expectation value.
  • The optimizer proposes new parameters for the quantum circuit to lower the energy.
  • Steps 3 and 4 are repeated iteratively until the energy converges to a minimum, which is the VQE estimate of the ground-state energy.

• Excited States (Variational Quantum Deflation):

  • To calculate excited states, use the Variational Quantum Deflation (VQD) algorithm. After finding the ground state, modify the cost function to penalize the overlap with the ground state and repeat the optimization process to find the next lowest energy state [58].

Case Study 2: Hâ‚‚O Molecule on a Trapped-Ion Quantum Computer

Protocol Objective: To estimate the ground-state energy of the water molecule (Hâ‚‚O) within chemical accuracy using a highly optimized VQE approach on a trapped-ion quantum computer.

Background: Achieving chemical accuracy (≈1.6 mHa) is critical for predicting chemical reaction rates. This protocol from [56] leverages the all-to-all connectivity of trapped-ion systems and co-design principles to minimize quantum resources.

Step-by-Step Protocol:

• Hamiltonian Generation and Qubit Mapping:

  • Perform a classical Hartree-Fock calculation for Hâ‚‚O in a chosen basis set (e.g., 6-31G) to obtain molecular orbitals and integrals.
  • Generate the electronic Hamiltonian and map it to a qubit representation using the Jordan-Wigner transformation. Select an active space to reduce the number of required qubits.

• Co-Design Circuit Optimization:

  • Ansatz Selection: Employ a problem-inspired Unitary Coupled-Cluster (UCC) ansatz, including both bosonic and non-bosonic excitation terms.
  • Connectivity Advantage: Design circuits to leverage the native all-to-all connectivity of trapped-ion qubits, thereby eliminating the overhead of SWAP gates required on limited-connectivity architectures [56].
  • Bosonic Excitation Simplification: Identify and implement bosonic excitations (where electron pairs are excited) using efficient XX(θ) gates, as they simplify to operations between two qubits without the need for JW strings [56].
  • JW String Optimization: For non-bosonic terms, carefully order the operations in the quantum circuit to cancel adjacent CNOT gates that arise from the JW strings, significantly reducing gate count [56].
  • State Preparation and Measurement (SPAM) Error Mitigation: Encode filled orbitals as the |0⟩ state instead of |1⟩ if the SPAM error for |0⟩ is significantly lower, reducing systematic readout errors for molecules with mostly closed shells [56].

• Quantum Execution and Iteration:

  • Prepare the initial HF state on the quantum processor.
  • Execute the optimized UCC ansatz circuit.
  • Measure the energy expectation value. The all-to-all connectivity allows for direct measurement of multi-qubit Pauli operators without qubit routing.
  • Feed the result to a classical optimizer to update the ansatz parameters until convergence.

Case Study 3: Nâ‚‚ Molecule via Classical MC SCF and CI

Protocol Objective: To provide a high-accuracy classical reference for the ground-state potential energy curve of Nâ‚‚ using Numerical Multi-Configurational Self-Consistent Field (MC SCF) and Configuration Interaction (CI) calculations.

Background: Classical high-precision calculations serve as vital benchmarks for assessing the performance of nascent quantum algorithms. This protocol, detailed in [60], demonstrates the computational complexity involved in accurately describing the strong correlation present in the Nâ‚‚ triple bond.

Step-by-Step Protocol:

• Numerical Hartree-Fock Calculation:

  • Perform a numerical Hartree-Fock calculation for the Nâ‚‚ ground state to obtain a mean-field reference wavefunction. This method uses numerical integration rather than an analytic basis set, potentially offering higher precision [60].

• Multi-Configurational SCF (MC SCF):

  • Select an active space comprising molecular orbitals derived from the 2p atomic orbitals.
  • Construct a wavefunction from a complete set of spin- and symmetry-adapted configurations within this active space (e.g., 18 configurations as in [60]).
  • Iteratively optimize both the configuration coefficients and the molecular orbitals themselves to achieve a self-consistent solution that accounts for static correlation.

• Configuration Interaction (CI) Calculation:

  • Using the integrals from the numerical MC SCF molecular orbitals, perform a more extensive CI calculation.
  • Include electron substitutions involving not just the valence electrons (2p) but also core electrons (1s, 2s) to capture dynamic correlation effects. This leads to a more accurate dissociation energy and spectroscopic constants [60].

• Analysis and Benchmarking:

  • Calculate the potential energy curve, dissociation energy, spectroscopic constants (e.g., ωₑ), and molecular properties like the quadrupole moment.
  • Compare the results obtained from numerical procedures with those from large basis-set Slater-type function calculations to quantify the improvement [60]. These results form a classical benchmark against which quantum computations can be evaluated.

Workflow Visualization

The following diagram illustrates the high-level logical workflow common to the VQE-based case studies (Hâ‚‚ and Hâ‚‚O), highlighting the hybrid quantum-classical loop and key error mitigation strategies.

This section details the key computational tools, hardware platforms, and algorithmic "reagents" essential for conducting ground-state energy estimation experiments on quantum hardware.

Table 3: Essential Research Reagents & Resources

Category	Item/Technique	Specification/Function	Application Case Study
Hardware Platforms	Trapped-Ion Quantum Computer	All-to-all qubit connectivity, enables efficient simulation without SWAP gates.	Hâ‚‚O [56]
	Superconducting Quantum Processor	Utilized for QCM algorithm, requires robust error mitigation.	Hâ‚‚O [59]
	NMR Quantum Simulator	Benchmarks small molecules and algorithm components.	Hâ‚‚ [58]
Algorithms & Ansätze	Variational Quantum Eigensolver (VQE)	Hybrid quantum-classical algorithm for finding ground states.	H₂ [58], H₂O [56]
	Unitary Coupled Cluster (UCC)	A chemically inspired, expressive ansatz for wavefunction preparation.	Hâ‚‚O [56]
	Quantum Computed Moments (QCM)	A non-variational algorithm that improves on VQE results using noisy hardware.	Hâ‚‚O [59]
Error Mitigation & Optimization	Circuit Optimization & Compilation	Exploits hardware connectivity (e.g., all-to-all) and cancels redundant gates (e.g., in JW strings).	Hâ‚‚O [56]
	SPAM Error Mitigation	Encodes information in lower-error states (e.g.,	0⟩) to reduce readout errors.	H₂O [56]
	Noise Extrapolation / Richardson Extrapolation	Extracts noiseless expectation values by deliberately scaling noise.	(Implied in broader context [49])
Classical Software & Tools	Classical Simulator (e.g., DMRG)	Provides reference values and high-quality initial states for quantum algorithms.	General [61]
	Full Configuration Interaction (FCI)	Exact classical method for small systems, used as a gold-standard benchmark.	Hâ‚‚ [55], Hâ‚‚O [56]
Theoretical Methods	Jordan-Wigner (JW) Transformation	Maps fermionic creation/annihilation operators to Pauli spin operators.	Hâ‚‚O [56]
	Active Space Approximation	Reduces problem size by restricting to chemically relevant orbitals.	Hâ‚‚O [56], Nâ‚‚ [60]

Optimizing Error Mitigation Performance and Managing Overhead

The pursuit of practical quantum utility on near-term quantum processors is critically challenged by inherent noise. Quantum Error Mitigation (QEM) has emerged as a critical strategy to suppress noise-induced bias in expectation values without the immense resource overhead of full-scale quantum error correction [62] [63]. However, the practical adoption of QEM is hampered by formidable sampling overheads, which become particularly prohibitive for practical quantum tasks involving families of circuits parameterized by classical inputs, such as those in quantum chemistry simulations [62] [8]. This application note analyzes the scaling properties and sampling costs of contemporary QEM protocols, providing a structured comparison and detailed experimental methodologies tailored for research in near-term quantum chemistry.

Core QEM Techniques and Their Scaling Properties

QEM encompasses a suite of strategies distinct from quantum error correction, employing hybrid quantum–classical techniques to reduce systematic bias and random error in noisy quantum circuits [63]. The following table summarizes the key QEM protocols, their core principles, and their associated sampling overheads.

Table 1: Quantum Error Mitigation Protocols and Their Sampling Overheads

QEM Protocol	Core Principle	Sampling Overhead Scaling	Key Advantages	Key Limitations
Zero-Noise Extrapolation (ZNE) [62] [63]	Systematically amplifies native noise rate (λ) and extrapolates to the zero-noise limit.	Linear with the number of boosted noise levels (u) and the number of circuit variants [62].	Model-agnostic; successfully deployed on large-scale processors (e.g., 127-qubit systems) [62].	Measurement cost scales linearly with the number of distinct circuits in a family [62].
Probabilistic Error Cancellation (PEC) [63]	Represents ideal operations as linear combinations of noisy implementable operations using quasi-probabilities.	Scales with the square of the sampling overhead factor, C², where C is the negativity cost of the quasi-probability decomposition [63].	Provides an unbiased estimator of the ideal observable [63].	Sampling overhead can become prohibitive at high error rates or large circuit depths [63].
Reference-State Error Mitigation (REM) [8]	Uses a classically-simulable reference state (e.g., Hartree-Fock) to characterize and subtract hardware noise.	Low-complexity; requires at most one additional algorithm iteration. No exponential sampling overhead [8].	Nearly establishes a lower bound on QEM costs for quantum chemistry [8].	Effectiveness is limited for strongly correlated systems where a single reference state is insufficient [8].
Surrogate-Enabled ZNE (S-ZNE) [62]	Uses a classical learning surrogate to predict noisy expectation values, moving the ZNE extrapolation entirely to the classical computer.	Constant measurement overhead for an entire family of parameterized circuits after initial surrogate training [62].	Superior scalability for tasks with many parameterized circuits; proven on up to 100-qubit tasks [62].	Accuracy depends on the quality and training of the classical surrogate model [62].
Error Detection with Spacetime Codes [64]	Uses low-overhead spacetime checks to detect errors, allowing for post-selection of error-free results.	Achieves quartically lower sampling overhead than PEC while using mild qubit overhead [64].	Provides single-shot access to the quantum state, compatible with both QEM and QEC; demonstrated on 50-qubit circuits [64].	The space of valid checks diminishes exponentially with the non-Clifford content of the circuit [64].

The sampling overhead for PEC and related methods is often analyzed through the Sampling Overhead Factor (SOF). For a single channel with error rate ϵ, bounds for the SOF (γ_C) for Pauli and depolarizing channels have been derived [63]. This overhead compounds multiplicatively with each circuit layer, leading to an exponential scaling with circuit volume (depth × qubit count) [63] [22].

Detailed Experimental Protocols

Protocol for Surrogate-Enabled ZNE (S-ZNE)

S-ZNE decouples data acquisition from quantum execution, drastically reducing measurement overhead for parameterized circuits [62].

Application Scope: Families of parameterized quantum circuits, such as those in variational quantum algorithms or digital quantum simulation, where the goal is to estimate noiseless expectation values for different classical inputs 𝒙 [62].

Methodology:

• Data Acquisition for Surrogate Training:
  • For a representative set of K parameter vectors {𝒙₁, 𝒙₂, ..., 𝒙_K}, execute the quantum circuit on hardware to measure the noisy expectation values f(𝒙, O, λ_j) for a set of amplified noise levels {λ₁, λ₂, ..., λ_u}.
  • This step incurs the one-time, fixed sampling cost of K × u × M shots, where M is the number of shots per circuit.

• Classical Surrogate Model Training:

  • Train a classical machine learning model (the surrogate) to learn the functional relationship 𝒙 → [f(𝒙, O, λ₁), ..., f(𝒙, O, λ_u)].
  • The input features are the parameter vector 𝒙, and the targets are the vectors of noisy expectation values.

• Classical-Only Error Mitigation:

  • For any new parameter 𝒙' (within the learned domain), use the trained surrogate to predict the vector of noisy expectation values 𝒛_C(𝒙').
  • Apply the standard ZNE extrapolation function g(â‹…) to this predicted vector to obtain the error-mitigated result: f_S-ZNE(𝒙', O) = g(𝒛_C(𝒙')).
  • This step requires zero additional quantum measurements, giving S-ZNE its constant-overhead property for an entire circuit family [62].

Workflow Diagram:

Protocol for Multireference Error Mitigation (MREM)

MREM extends the REM protocol to strongly correlated systems by using multireference states, which are linear combinations of Slater determinants, to better capture the true ground state [8].

Application Scope: Quantum chemistry simulations, particularly for molecules with strong electron correlation (e.g., bond-stretching in Nâ‚‚, Fâ‚‚) where single-reference REM fails [8].

Methodology:

• Generate Multireference State Classically:
  • Use an inexpensive classical method (e.g., complete active space self-consistent field - CASSCF) to generate an approximate multireference wavefunction. This wavefunction is a truncated expansion |ψ_MR⟩ = Σ_i c_i |ϕ_i⟩, where |ϕ_i⟩ are Slater determinants.

• Quantum Circuit Preparation via Givens Rotations:

  • Compile the multireference state |ψ_MR⟩ into a quantum circuit. Givens rotations are employed to efficiently prepare this state from an initial computational basis state (e.g., the Hartree-Fock state) [8].
  • The Givens rotation circuit preserves key symmetries like particle number and spin, and its structure is universal for quantum chemistry state preparation [8].

• Error Mitigation Execution:

  • Target State Measurement: Run the Variational Quantum Eigensolver (VQE) algorithm to prepare and measure the energy E_target of the target state |ψ(θ)⟩ on the quantum device.
  • Reference State Measurement: Prepare and measure the energy E_MR of the multireference state |ψ_MR⟩ on the same noisy quantum device.
  • Classical Computation: Calculate the exact, noiseless energy E_MR, exact of the multireference state |ψ_MR⟩ on a classical computer.

• Error Cancellation:

  • Obtain the error-mitigated energy estimate for the target state by subtracting the noise bias measured on the reference state: E_mitigated = E_target - (E_MR - E_MR, exact) [8].

Workflow Diagram:

Protocol for Hybrid Error Detection and Mitigation

This protocol combines the benefits of quantum error detection (QED) and probabilistic error cancellation (PEC) to achieve high-fidelity results with reduced sampling overhead compared to PEC alone [22].

Application Scope: Near-term quantum simulation of chemistry problems, leveraging the [[n, n-2, 2]] quantum error detection code for its constant qubit overhead and simplified logical rotations [22].

Methodology:

• Circuit Encoding and Error Detection:
  • Encode the quantum circuit using the [[n, n-2, 2]] code. This uses n physical qubits to encode n-2 logical qubits.
  • Execute the encoded circuit on the quantum hardware.
  • Perform decoding and measure the syndrome qubits. Any non-zero syndrome indicates a detectable error.
  • Post-select and retain only the shots where no error is detected.

• Characterize the Logical Noise Channel:

  • Perform quantum process tomography or use other characterization techniques on the post-selected logical output state to estimate the effective logical noise channel 𝒩_logical. This channel represents the residual, undetectable errors that passed the post-selection filter [22].

• Probabilistic Error Cancellation:

  • Construct a quasi-probability decomposition to invert the characterized logical noise channel: 𝒩_logical⁻¹ = Σ_i q_i B_i.
  • Apply PEC in the usual manner, but now at the logical level, by sampling from the operations B_i and re-scaling the results by the signs and the total cost factor C = Σ_i |q_i| [22].

Key Benefit: Because error detection has already filtered out many errors, the residual logical noise channel 𝒩_logical is weaker than the physical noise. This means its inversion via PEC is less costly, leading to a lower sampling overhead factor C² [22].

The Scientist's Toolkit: Research Reagent Solutions

This section details the essential "reagents" or core components required to implement the QEM protocols described above in a research setting.

Table 2: Essential Research Reagents for QEM Protocols

Item / Resource	Function / Purpose	Example Implementations / Notes
Parameterized Quantum Circuit	Encodes the quantum chemistry problem (e.g., molecular ansatz for VQE). The function U(𝒙) that generates states ρ(𝒙) [62].	UCCSD, hardware-efficient ansätze; Decomposed into Clifford gates and Z-rotations as in Eq. (1) [62].
Noise Amplification Technique	Artificially increases the native error rate λ for ZNE. Essential for mapping the noise-response curve.	Unitary folding (repeating gate sequences) [62] [63].
Classical Machine Learning Surrogate	Learns the relationship between circuit parameters and noisy expectation values. Core of S-ZNE.	Neural networks, Gaussian processes; Requires initial training data from quantum hardware [62] [65].
Classically Simulable Reference States	Provides a noise benchmark for REM/MREM. Its noiseless properties must be known exactly.	Hartree-Fock state (for REM) [8]; Multireference states from CASSCF (for MREM) [8].
Givens Rotation Circuits	Efficiently prepares multireference states on quantum hardware while preserving physical symmetries.	A structured quantum circuit built from controlled rotations to create superpositions of Slater determinants [8].
Error Detection Code (e.g., [[n,n-2,2]])	Detects a broad class of errors with low qubit overhead, enabling post-selection.	Defined by stabilizers X⊗n and Z⊗n; Requires only two additional physical qubits [22].
Pauli Twirling Set	Converts coherent noise into stochastic noise, making it more amenable to mitigation via PEC.	A set of Pauli operators applied before and after a gate; Can be partial to reduce overhead [22].
Noise Characterization Toolkit	Estimates the physical or logical noise channel for protocols like PEC.	Gate Set Tomography (GST) [63]; Quantum Process Tomography.

Quantum error mitigation (QEM) is an essential toolkit for extracting useful results from current Noisy Intermediate-Scale Quantum (NISQ) devices, which are susceptible to noise that limits computational accuracy [49]. Within the suite of QEM strategies, learning-based approaches have gained prominence for their practicality and reduced overhead. Among these, Clifford Data Regression (CDR) has emerged as a powerful technique, leveraging classically simulatable Clifford circuits to train a model that can mitigate errors in more complex, non-Clifford circuits [40].

This application note details two advanced enhancements to the core CDR framework: Energy Sampling (ES) and Non-Clifford Extrapolation (NCE). These protocols were developed specifically to address the challenges of quantum chemistry simulations, such as those employing the Variational Quantum Eigensolver (VQE) to determine molecular ground state energies [40]. We frame these developments within the broader thesis that physically motivated error mitigation protocols, which incorporate insights from the specific application domain, are crucial for advancing the capabilities of near-term quantum computation in chemistry and drug development research.

Theoretical Foundation and Enhancements

Core Principles of Clifford Data Regression

Clifford Data Regression is a learning-based error mitigation method. Its fundamental operation can be summarized as follows [40]:

• Generate Training Circuits: A set of "near-Clifford" circuits, which are classically simulatable, are generated. These circuits are designed to be structurally similar to the target circuit whose result requires mitigation.
• Obtain Noisy and Ideal Expectation Values: For each training circuit, the expectation value of the observable of interest (e.g., energy) is computed both on a noisy quantum device (or a realistic simulator) and on a classical computer (yielding the ideal, noiseless value).
• Train a Regression Model: A linear regression model is trained to map the noisy expectation values to their ideal counterparts.
• Mitigate the Target Circuit: The trained model is applied to the noisy result from the non-Clifford target circuit to produce a error-mitigated estimate.

The efficacy of CDR stems from the Gottesman-Knill theorem, which states that circuits composed exclusively of Clifford gates can be efficiently simulated classically [40]. By working with near-Clifford circuits, one maintains classical simulability while approximating the noise characteristics of the target circuit.

Advanced Protocols: ES and NCE

Building on detailed analyses of CDR hyperparameters for molecular simulations, two key enhancements have been developed to improve mitigation performance [40].

Energy Sampling (ES)

The Energy Sampling protocol introduces a filtering step prior to regression. Instead of using all generated training circuits, ES selects only those circuits that produce the lowest-energy states during classical simulation [40]. This selection strategy actively biases the training set toward the region of the Hilbert space that is most physically relevant—namely, the vicinity of the true ground state. By focusing the regression on these low-energy samples, the model learns a more accurate noise correction for the states of interest, leading to improved mitigation fidelity for ground state energy calculations.

Non-Clifford Extrapolation (NCE)

The Non-Clifford Extrapolation protocol enhances the regression model itself. The standard CDR uses a simple linear model with the noisy expectation value as the sole input. NCE augments this model by incorporating the number of non-Clifford gates in the training circuit as an additional input feature [40]. This allows the regression model to explicitly learn and account for how the relationship between noisy and ideal expectation values evolves as the quantum circuit becomes more complex and less Clifford-dominated. As the target circuit is approached, which typically has the highest non-Clifford count, the model can perform a more informed and accurate extrapolation.

Table 1: Comparison of Core CDR Protocols

Protocol	Key Innovation	Primary Advantage	Ideal Use Case
Standard CDR	Maps noisy-to-ideal values via regression on near-Clifford circuits.	Simplicity and general applicability.	Preliminary mitigation for various algorithms.
Energy Sampling (ES)	Selects only low-energy training circuits for regression.	Biases mitigation toward physically relevant states, improving accuracy for ground state problems.	VQE and other ground-state energy calculations.
Non-Clifford Extrapolation (NCE)	Uses non-Clifford gate count as an additional regression feature.	Captures the evolution of noise with circuit complexity, enabling better extrapolation.	Circuits with varying or high counts of non-Clifford gates.

Experimental Protocols and Workflows

This section provides a detailed methodology for implementing the ES- and NCE-enhanced CDR protocol in the context of a quantum chemistry simulation, such as calculating the ground state energy of a molecule.

Preliminary Setup and System Preparation

• Define the Problem: Select the target molecular system (e.g., Hâ‚„) and construct its electronic Hamiltonian in the second-quantized form [40]. H = Σ h_pq E_pq + (1/2) Σ g_pqrs (E_pq E_rs - δ_qr E_ps)
• Choose an Ansatz: Select a parameterized wavefunction ansatz. The tiled Unitary Product State (tUPS) ansatz is a suitable choice due to its fermionic operator basis and lower-depth structure [40].
• Optimize Parameters Classically: Use a classical simulator to run a VQE routine and find the optimal parameters θ_opt that minimize the energy expectation value E = min〈Ψ_HF| U†(θ) H U(θ) |Ψ_HF〉 [40]. This defines the target, noiseless circuit U(θ_opt).

Protocol Workflow: ES-NCE-CDR

The following workflow, summarized in the diagram below, integrates both ES and NCE into a comprehensive error mitigation protocol.

Detailed Methodology

• Generate Training Circuit Pool: Create a large set of training circuits. These are typically generated by taking the target circuit U(θ_opt) and replacing a large fraction of its non-Clifford gates with Clifford gates. The remaining non-Clifford gates have their parameters randomized [40]. For each circuit i, record its number of non-Clifford gates, N_nonCliffordáµ¢.

• Data Acquisition:

  • Run all training circuits on a classical simulator to obtain the ideal, noiseless expectation values E_idealáµ¢.
  • Run the same circuits on a noisy quantum simulator (e.g., using an ibm_torino noise model) or actual quantum hardware to obtain the noisy expectation values E_noisyáµ¢ [40].

• Apply Energy Sampling (ES):

  • Sort the pool of training circuits by their E_ideal values in ascending order.
  • Select the top K circuits (e.g., the 20% with the lowest energies) to form the final, biased training set for regression [40].

• Train the NCE Regression Model:

  • Using the ES-filtered dataset, train a linear regression model. The model takes two inputs: E_noisy and N_nonClifford.
  • The model is defined as: E_ideal = a * E_noisy + b * N_nonClifford + c, where a, b, and c are the regression coefficients learned during training [40].

• Mitigate the Target Circuit:

  • Execute the target, non-Clifford circuit U(θ_opt) on the noisy quantum device to obtain E_noisy_target.
  • Input E_noisy_target and the known N_nonClifford_target of the circuit into the trained model.
  • The output of the model, E_mitigated, is the error-mitigated estimate of the ground state energy.

Table 2: Key Research Reagents and Computational Tools

Item / Resource	Function / Description	Example / Specification
Molecular Hamiltonian	Defines the quantum chemistry problem and the target observable (energy).	Hâ‚„ system in a defined basis set [40].
Parameterized Ansatz	Provides the circuit structure for the VQE algorithm.	Tiled Unitary Product State (tUPS) ansatz [40].
Near-Clifford Circuits	Serve as the classically simulatable training data for the CDR model.	Circuits generated from the target by randomizing/Cliffordizing gates [40].
Noise Model	Emulates the behavior of real quantum hardware for simulation-based testing.	ibm_torino processor noise model [40].
Classical Simulator	Computes the ideal (noiseless) expectation values for the training circuits.	Statevector simulator (e.g., via Qiskit Aer) [40].

Performance and Analysis

Numerical experiments demonstrate the superior performance of the enhanced CDR protocols. In simulations of the Hâ‚„ molecule using the tUPS ansatz, both ES and NCE individually, and especially in combination, outperform the original CDR method [40].

The Energy Sampling protocol directly addresses a key weakness of standard CDR: the inclusion of training data from high-energy, physically irrelevant states whose noise characteristics may differ from those near the ground state. By filtering based on energy, the regression is focused, leading to a more accurate local correction.

The Non-Clifford Extrapolation protocol tackles the scalability of the error model. A simple linear model with only the noisy expectation value as input may not capture how error propagation changes with circuit complexity. By explicitly including the non-Clifford count, NCE allows the model to learn a more general and scalable mapping, which is critical for successfully mitigating the deeper, more complex circuits required for high-accuracy chemistry simulations.

Table 3: Hyperparameter Analysis for CDR-based Protocols (based on Hâ‚„ simulation data [40])

Hyperparameter	Standard CDR Consideration	Impact of ES & NCE
Training Set Size	Generally requires a large pool of circuits for good performance.	ES reduces the effective size but increases quality. Performance can be maintained with fewer, high-quality samples.
Fraction of Non-Clifford Gates	The fraction of non-Clifford gates in training circuits must be tuned to balance similarity to target and classical simulability.	NCE makes the model less sensitive to this fraction, as it can learn the dependence on non-Clifford content.
Regression Model Complexity	Typically a simple linear model.	NCE inherently uses a slightly more complex multivariate linear model, which is still efficient to train.
Circuit Selection Criteria	Often random or based on structural similarity.	ES provides a physically motivated, problem-aware selection criterion that improves ground state energy estimation.

The integration of Energy Sampling and Non-Clifford Extrapolation into the Clifford Data Regression framework represents a significant advancement in application-specific quantum error mitigation. These protocols move beyond a generic noise-mapping approach by incorporating physical insights—the importance of the low-energy subspace and the scaling of errors with circuit complexity.

For researchers in chemistry and drug development, these tools offer a more precise and reliable method for simulating molecular properties on today's noisy quantum hardware. By following the detailed experimental protocols outlined in this document, scientists can implement these advanced techniques to extend the computational reach of NISQ-era devices in the pursuit of materially accurate simulations.

Digital-Analog Quantum Computing (DAQC) is a hybrid paradigm that merges the programmability of digital quantum computation with the resilience and efficiency of analog quantum simulation. In the context of circuit compilation, this approach leverages the natural, continuous-time dynamics of a quantum processor to natively implement complex multi-qubit operations, which would otherwise require deep digital gate sequences. This compilation strategy is particularly valuable for enhancing the robustness of quantum algorithms run on Noisy Intermediate-Scale Quantum (NISQ) devices, as it directly addresses the primary sources of error in digital circuits: the high infidelity of two-qubit gates and the accumulation of noise over long circuit depths [66] [67].

For near-term applications like quantum chemistry simulations, robustness is paramount. DAQC enhances robustness by collapsing deep digital circuits into shallower, more coherent analog evolutions, thereby reducing the window of vulnerability to decoherence and stochastic noise [68]. Experimental and numerical studies have consistently demonstrated that DAQC compilations can achieve higher fidelity than their digital counterparts, especially as the system size scales, offering a promising path toward practical quantum advantage in the NISQ era [66].

Foundational Principles and Comparative Advantages

The core principle of DAQC is the decomposition of a target quantum algorithm into a sequence of analog blocks, defined by the processor's native interacting Hamiltonian, interleaved with fast digital single-qubit gates. This stands in contrast to the purely Digital Quantum Computing (DQC) paradigm, which relies exclusively on discrete single- and two-qubit gates [66] [67]. The inherent advantage of DAQC for circuit compilation stems from its efficient use of a hardware's native interactions.

A key metric for robustness in compiled circuits is the overall circuit depth. Deep circuits are particularly susceptible to incoherent errors that accumulate with each gate operation. DAQC directly mitigates this by offering a dramatic reduction in the number of required two-qubit gates and the overall depth for implementing complex interactions. This is especially pronounced for problems involving higher-order Hamiltonian terms, such as those found in quantum chemistry simulations of strongly correlated systems or in the context of early quantum error correction codes like the surface code [68].

Table 1: Quantitative Comparison of Compilation Strategies for a 4-body Hamiltonian

Compilation Metric	Pure Digital (DQC)	Digital-Analog (DAQC)	Improvement Factor
Circuit Depth Scaling	O(N²k)	O(Nk+1)	10x reduction for 4-body terms [68]
Two-Qubit Gate Count	High	Significantly Lower	Reduced propagation of errors [68]
Fidelity on NISQ devices	Lower	Higher	Marked improvement, especially as qubit count increases [66]
Noise Resilience	More susceptible to gate noise and crosstalk	More resilient to coherent and stochastic noise [68]	Native analog blocks turn crosstalk into a resource [66]

Experimental Protocols and Workflows

Implementing a DAQC-based compilation strategy requires a structured workflow, from problem encoding to execution on a hybrid-capable device. The following protocol details the key steps for compiling a higher-order problem, such as a molecular Hamiltonian or a Higher-Order Binary Optimization (HUBO) problem.

Protocol: DAQC Compilation for Higher-Order Problems

Objective: To implement a target Hamiltonian (e.g., a 4-body HUBO or quantum chemistry Hamiltonian) on a NISQ device using the DAQC paradigm to achieve higher fidelity and lower circuit depth compared to a digital compilation.

Materials and Prerequisites:

• A quantum processor with programmable analog blocks (e.g., native Ising-type or exchange interactions).
• Classical computing resources for Hamiltonian decomposition and pulse sequence calculation.
• Software tools for DAQC sequence compilation (e.g., vendor-specific SDKs or custom frameworks).

Procedure:

• Problem Formulation and Hamiltonian Decomposition:

  • Express the problem of interest as a target Hamiltonian, ( H_{\text{target}} ).
  • Decompose ( H{\text{target}} ) into a sum of terms compatible with the device's native analog interaction, ( H{\text{native}} ) (e.g., an all-to-all Ising Hamiltonian) [68] [66].

• DAQC Sequence Compilation:

  • Compile the target unitary evolution, ( e^{-iH{\text{target}}t} ), into a sequence of analog evolution blocks under ( H{\text{native}} ) interleaved with digital single-qubit gates.
  • This involves calculating the appropriate evolution times for the analog blocks and the rotation angles for the digital gates to synthesize the desired higher-order couplings [68].

• Circuit Depth and Gate Count Analysis:

  • Compare the compiled DAQC sequence against a digitally compiled circuit for the same unitary.
  • Metrics for comparison should include total two-qubit gate count, estimated circuit depth, and theoretical fidelity based on device gate error rates [68].

• Hardware Execution with Error Mitigation:

  • Execute the compiled DAQC sequence on the quantum processor.
  • To further enhance robustness, apply error mitigation techniques such as Zero-Noise Extrapolation (ZNE) to the analog blocks. This involves deliberately scaling the noise in the analog evolution (e.g., by stretching pulse times) and extrapolating back to a zero-noise expectation value [66].

• Validation and Post-Processing:

  • Measure the output state and compute the observable of interest (e.g., energy expectation value).
  • Use classical post-processing to apply error mitigation and validate the result against classical simulations or known values.

The logical workflow of this protocol, illustrating the transition from a digital to a hybrid compilation strategy, is summarized in the diagram below.

The Scientist's Toolkit: Essential Research Reagents and Platforms

Successfully implementing the DAQC compilation strategy requires access to specific hardware and software tools. The following table details key "research reagents" in this context.

Table 2: Essential Research Reagents for DAQC Protocol Implementation

Reagent / Platform	Type	Function in DAQC Protocol
Programmable Analog Block	Hardware	The core resource for executing continuous multi-qubit evolution. Examples include global drives in trapped ions, Rydberg blockade in neutral atoms, or resonator-coupled superconducting qubits [68].
DAQC Compilation Software	Software	Converts a target unitary or Hamiltonian into a sequence of analog blocks and digital pulses. This is often vendor-specific or requires custom solver development [68] [67].
Zero-Noise Extrapolation (ZNE)	Software / Technique	A quantum error mitigation method applied post-execution to improve result accuracy by extrapolating from noise-amplified data [66].
Digital Single-Qubit Gates	Hardware	Fast, high-fidelity gates used to interleave with analog blocks, providing the "digital" control in the hybrid paradigm [66] [67].
Classical Optimizer	Software	Used in tandem with variational algorithms (like VQE) to optimize parameters in the DAQC-compiled circuit for tasks like ground-state energy finding [54].

Case Study: Application in Quantum Chemistry Simulation

The practical advantage of the DAQC compilation strategy is clearly demonstrated in quantum chemistry simulations, where the electronic structure problem is mapped to a qubit Hamiltonian often containing computationally expensive higher-order terms.

In a recent study, a multicomponent unitary coupled cluster (mcUCC) ansatz for simulating molecular systems beyond the Born-Oppenheimer approximation was developed [54]. While this work employed error mitigation on a digital device, the resource requirements of such correlated electron-nuclear simulations make them an ideal target for DAQC compilation. The complex, multi-reference nature of the wavefunctions in strongly correlated systems or bond-breaking regions leads to Hamiltonians with significant multi-body terms [8]. A pure digital compilation of these terms results in prohibitively deep circuits.

By applying the DAQC compilation protocol, these multi-qubit interactions can be synthesized natively via analog blocks. As shown in Table 1, this approach can achieve an order-of-magnitude reduction in circuit depth for 4-body terms [68]. This depth compression directly translates to reduced exposure to decoherence and gate errors, enabling more accurate simulations of larger, more complex molecules on current hardware. The enhanced robustness provided by DAQC is therefore a critical enabler for achieving chemical accuracy in near-term quantum chemistry experiments.

Digital-Analog Quantum Computing represents a paradigm shift in circuit compilation strategies, moving away from a purely discrete gate-based approach to one that co-designs algorithms with the underlying hardware physics. The evidence from numerical and early experimental studies is clear: DAQC-compiled circuits consistently demonstrate superior robustness and fidelity compared to their digital equivalents, particularly as problem sizes scale [66] [68].

For the field of near-term quantum chemistry simulations, adopting a DAQC compilation framework is a highly promising path toward mitigating the crippling effects of noise. The significant reduction in circuit depth and two-qubit gate count directly addresses the core bottlenecks of NISQ-era devices. As hardware providers continue to enhance the programmability of their analog resources [68] [67], and as software tools for DAQC compilation mature, researchers can leverage this strategy to push the boundaries of simulable molecular systems, inching closer to the long-sought goal of practical quantum advantage in computational chemistry and drug development.

Partial Pauli Twirling and Custom Twirling Sets for Reduced Overhead

Quantum error mitigation (QEM) has emerged as a critical toolkit for extracting meaningful results from noisy intermediate-scale quantum (NISQ) devices. Unlike fault-tolerant quantum computation, which requires substantial qubit overhead for quantum error correction, error mitigation techniques aim to suppress errors at a cost that is often manageable for near-term applications [69]. Within the landscape of QEM, Pauli twirling stands as a fundamental technique for transforming general noise channels into simpler, tailorable forms. This application note details the protocol for partial Pauli twirling using custom twirling sets, an advanced method that significantly reduces the sampling and circuit overheads associated with conventional twirling. Framed within research on quantum computational chemistry, this technique enables more precise and efficient molecular simulations, such as ground-state energy estimation, which is vital for drug development and materials science [9] [8].

Theoretical Foundation

Pauli Twirling and the Pauli Twirling Approximation (PTA)

Pauli twirling is a process that transforms an arbitrary quantum noise channel into a Pauli channel by conjugating the channel's operations with random Pauli operators. For a noise channel represented by its Kraus operators, twirling over the entire Pauli group results in a channel that is diagonal in the Pauli basis [70]. This Pauli Twirling Approximation (PTA) simplifies the noise model, making it more tractable for analysis and simulation. Studies have shown that the PTA reliably predicts logical error rates in quantum error-correcting codes, often overestimating them by a small factor, thus providing a conservative and "honest" representation of the noise [70].

The Need for Partial and Custom Twirling

Full Pauli twirling requires a number of circuit randomizations that grows exponentially with the number of qubits, leading to a significant sampling overhead that is often impractical for NISQ devices. The concept of partial Pauli twirling addresses this by using a carefully selected subset of the Pauli group for conjugation. A custom twirling set is designed to be tailored to a specific target circuit or observable, thereby maintaining much of the error randomization benefit while drastically reducing the number of unique circuit configurations required [51]. This approach is particularly powerful when combined with other error mitigation techniques, such as probabilistic error cancellation, within hybrid error suppression protocols [51].

Protocol: Partial Pauli Twirling with Custom Sets

The following protocol describes the integration of partial Pauli twirling into a variational quantum eigensolver (VQE) for molecular energy estimation, a common task in quantum chemistry simulations.

Table 1: Key Research Reagent Solutions

Item Name	Function/Description
Custom Twirling Set	A selected subset of Pauli operators used for circuit conjugation to reduce sampling overhead while effectively randomizing errors [51].
$[[n, n-2, 2]]$ Code	A quantum error-detecting code used in hybrid protocols to detect errors without the full overhead of error correction [51].
Informationally Complete (IC) Measurements	A set of measurements that allows for the estimation of multiple observables from the same data, reducing circuit overhead [9].
Quantum Detector Tomography (QDT)	A technique to characterize and mitigate readout errors by building a model of the noisy measurement process [9].
Givens Rotation Circuits	Quantum circuits used to efficiently prepare multireference states for error mitigation in strongly correlated molecular systems [8].

The diagram below illustrates the complete workflow for a quantum chemistry simulation employing partial Pauli twirling and error detection.

Step-by-Step Experimental Methodology

Step 1: Design the Custom Twirling Set

• Objective: Identify a minimal subset of Pauli operators that effectively randomize the dominant errors in your specific target circuit.
• Methodology:
  • Circuit Analysis: Analyze the core VQE circuit (e.g., the ansatz for a molecule like Hâ‚‚ or LiH) to identify the most frequently occurring Pauli terms or the gates most susceptible to noise.
  • Set Selection: Instead of the full Pauli group, select a custom set that includes:
    • Pauli operators that match the structure of the circuit's generators.
    • Operators that anticommute with the most likely error channels.
  • Validation: Use classical simulation or prior device calibration data to verify that the selected set provides a suitable approximation to the full twirling effect for the observable of interest (e.g., the energy) [51].

Step 2: Integrate with an Error Detection Code

• Objective: Combine the low-overhead twirling with an efficient error detection mechanism to filter out corrupted results.
• Methodology:
  • State Preparation: Encode the problem using the lightweight $[[n, n-2, 2]]$ quantum error-detecting code. This code uses two ancilla qubits to detect a single error on any of the $n$ data qubits [51].
  • Twirled Circuit Execution:
    • For each shot, randomly select a Pauli operator $P$ from the custom twirling set.
    • Apply $P$ to the encoded data qubits at the start of the circuit.
    • Run the core VQE circuit (e.g., a variational ansatz like EfficientSU2).
    • Apply the inverse Pauli $P^{\dagger}$.
    • Measure the stabilizer generators of the $[[n, n-2, 2]]$ code to detect errors.
  • Data Filtering: Postselect on the measurement outcomes, discarding any run where an error is detected by the code.

Step 3: Apply Probabilistic Error Cancellation (PEC)

• Objective: Mitigate the effect of errors that are not detected by the code.
• Methodology:
  • Noise Characterization: Use gate set tomography or process tomography to characterize the noise channel of the twirled gates. The twirling process ensures this noise is well-approximated by a Pauli channel [70].
  • Build a Noise Inversion Map: Construct a quasi-probability distribution that represents the inverse of the characterized Pauli noise channel.
  • Post-Processing: In the classical post-processing stage, apply this inversion map to the results from the postselected data to obtain a unbiased estimate of the observable [51].

Step 4: Measurement and Readout Error Mitigation

• Objective: Further enhance precision by addressing measurement inaccuracies.
• Methodology:
  • Employ Informationally Complete (IC) measurements and perform Quantum Detector Tomography (QDT) in parallel with the main computation. This allows for the construction of an unbiased estimator that mitigates readout errors [9].

Performance and Data

The implementation of this hybrid protocol demonstrates a favorable trade-off between overhead and precision. The following table summarizes key performance metrics as evidenced by recent research.

Table 2: Performance Metrics of Error Mitigation Techniques for Quantum Chemistry

Technique	Reported Performance / Improvement	Key Metric	Associated Overhead
Neural Error Mitigation	Improved ground state estimations for Hâ‚‚, LiH, and lattice Schwinger model [71].	Fidelity, Energy Error	Classical computation for neural network training.
Pauli Check Sandwiching (6 layers)	Average fidelity gain of 34 percentage points for random circuits with 40 CNOTs [69].	Quantum State Fidelity	Additional ancilla qubits and postselection.
Combined Detection & Mitigation	Successful ground state energy estimation of Hâ‚‚ on IBM hardware; reduced sampling overhead via custom Pauli twirling sets [51].	Estimation Accuracy, Sampling Cost	Custom twirling set design and postselection.
Readout Error Mitigation (QDT)	Reduction in measurement errors from 1-5% to 0.16% for BODIPY molecule energy estimation [9].	Absolute Error in Energy	Additional circuits for detector tomography.

Technical Considerations

• Sampling Overhead: The primary cost of partial twirling and error detection is the sampling overhead, which is the increased number of circuit executions (shots) required to achieve a desired precision due to postselection. This overhead must be balanced against the improvement in result accuracy [51] [9].
• Circuit Overhead: The $[[n, n-2, 2]]$ code and twirling operations add depth to the quantum circuit. This must be considered in the context of the device's coherence times and gate fidelities.
• Domain Specificity: The effectiveness of the custom twirling set is inherently tied to the circuit it is designed for. A set tailored for a VQE ansatz for one molecule may not be optimal for another, requiring re-analysis for new problems.

Partial Pauli twirling with custom twirling sets represents a sophisticated and practical tool for quantum chemists and drug development researchers aiming to push the boundaries of computational accuracy on NISQ devices. By strategically reducing the randomization overhead and synergistically combining with error detection and cancellation techniques, this protocol enables more reliable simulations of molecular systems, such as the BODIPY molecule and strongly correlated systems like Fâ‚‚ [9] [8]. As quantum hardware continues to mature, such hybrid error mitigation strategies will be indispensable in the incremental path toward demonstrating quantum advantage in real-world computational chemistry.

Balancing Circuit Expressivity and Noise Sensitivity in State Preparation

In near-term quantum simulations of molecular systems, a fundamental tension exists between the expressivity of state preparation circuits and their susceptibility to hardware noise. Highly expressive circuits, capable of representing complex quantum states, are essential for modeling strongly correlated electronic structures but typically require greater circuit depth and complexity, amplifying their sensitivity to errors [8]. Conversely, simpler noise-resilient circuits often lack the expressivity needed for accurate quantum chemistry simulations, particularly in systems exhibiting multireference character. This application note examines this critical balance within the context of quantum error mitigation protocols, providing structured guidelines and quantitative frameworks to inform researcher decisions when implementing quantum algorithms for chemical simulations on noisy intermediate-scale quantum devices.

Quantitative Analysis of State Preparation Strategies

The relationship between circuit complexity, expressivity, and noise sensitivity can be quantified through several key metrics. The following tables summarize comparative data for prominent state preparation approaches relevant to quantum chemistry simulations.

Table 1: Comparison of State Preparation Methods for Quantum Chemistry

Method	Circuit Depth	Expressivity Metrics	Noise Sensitivity	Optimal Use Case
Single-Reference (HF)	Constant/Shallow [8]	Single determinant	Low	Weakly correlated systems [8]
Multireference (MREM)	Moderate (Givens rotations) [8]	Linear combination of dominant determinants	Moderate	Strongly correlated systems [8]
Unitary Coupled Cluster	High	High (full configuration interaction)	High	High-accuracy simulations
Variational Hybrid Ansätze	Variable	Tunable expressivity	Variable	Adaptive applications

Table 2: Error Mitigation Performance Across Molecular Systems

Molecule	Bond Length (Ã…)	Unmitigated Error (Ha)	REM Error (Ha)	MREM Error (Ha)	Improvement Factor
H₂O	0.96	0.152	0.084	0.031	2.7× [8]
N₂	1.10	0.238	0.156	0.059	2.6× [8]
F₂	1.41	0.411	0.332	0.127	2.6× [8]
H₂ (stretched)	1.50	0.385	0.301	0.112	2.7× [8]

Table 3: Noise Propagation Across Different Qubit Encodings

Encoding Scheme	Depolarizing Noise Sensitivity	Dephasing Noise Sensitivity	Relaxation Noise Sensitivity	Circuit Overhead
Jordan-Wigner	Moderate	Low	High	Low [72]
Bravyi-Kitaev	High	Moderate	Moderate	Moderate [72]
Superfast Bravyi-Kitaev	High	High	High	Low

Experimental Protocols for Error-Mitigated State Preparation

Multireference State Error Mitigation (MREM) Protocol

Purpose: To mitigate errors in strongly correlated molecular systems where single-reference approaches fail [8].

Materials and Equipment:

• Quantum processor or simulator with noise model
• Classical computer for wavefunction analysis
• Quantum circuit compiler supporting Givens rotations

Procedure:

• Reference State Selection:
  • Perform classical computational chemistry calculation (e.g., CASSCF, DMRG) to identify dominant Slater determinants
  • Select determinants with largest weights in target wavefunction (typically 3-10 determinants)
  • Construct truncated multireference wavefunction |Ψ₀⟩ = Σᵢ cáµ¢|Dᵢ⟩

• Circuit Construction:

  • Initialize quantum register to Hartree-Fock reference state |D₀⟩
  • Implement Givens rotation circuits to prepare target multireference state:
    • For each determinant |Dᵢ⟩ in expansion, apply sequenced Givens rotations
    • Control rotation angles to achieve target coefficients cáµ¢
  • Ensure symmetry preservation (particle number, spin projection) throughout

• Error Mitigation Execution:

  • Prepare and measure reference state |Ψ₀⟩ on quantum device, obtaining E_noisy^ref
  • Compute exact reference energy E_exact^ref classically
  • Prepare and measure target state |Ψ(θ)⟩ using VQE, obtaining E_noisy^target
  • Apply MREM correction: Emitigated = Enoisy^target - (Enoisy^ref - Eexact^ref)

• Validation:

  • Compare mitigated energy with classical benchmark
  • Calculate overlap between prepared and ideal states
  • Assess statistical uncertainty through repeated measurements

Hybrid Error Detection and Mitigation Protocol

Purpose: To combine quantum error detection with error mitigation for improved performance in variational quantum eigensolver simulations [22].

Materials and Equipment:

• Quantum processor supporting stabilizer measurements
• Pauli twirling implementation
• Probabilistic error cancellation toolbox

Procedure:

• Circuit Encoding:
  • Encode logical qubits using [[n,n-2,2]] quantum error detection code
  • Implement state preparation circuits on encoded logical qubits
  • Append stabilizer measurement circuits for error detection

• Partial Pauli Twirling:

  • Apply custom Pauli twirling with reduced operator set
  • Implement twirled versions of core circuit operations
  • Maintain circuit structure while randomizing error signatures

• Execution and Post-Selection:

  • Execute twirled, encoded circuit multiple times
  • Measure stabilizers and discard runs with detected errors
  • Record measurement outcomes only from post-selected runs

• Probabilistic Error Cancellation:

  • Characterize logical noise channel after post-selection
  • Apply PEC to remaining undetectable errors
  • Reconstruct error-mitigated expectation values

Circuit Structure-Preserving Error Mitigation

Purpose: To characterize and mitigate noise without modifying original circuit architecture [73].

Materials and Equipment:

• Parameterized quantum circuit framework
• Calibration circuit construction tools
• Linear algebra solver for calibration matrix

Procedure:

• Calibration Matrix Construction:
  • Define identity circuit V_mit with identical structure to target circuit V
  • For each computational basis state |ψᵢᶦⁿ⟩:
    • Prepare |ψᵢᶦⁿ⟩ and apply Vmit on noisy quantum device
    • Measure output probabilities pᵢⱼ for all basis states |ψⱼ⟩
    • Set Mᵢⱼ^mit = ⟨ψⱼ|Vnoisy^mit|ψᵢᶦⁿ⟩

• Noise Characterization:

  • Assemble calibration matrix M^mit from measurements
  • Verify matrix invertibility and condition number
  • Construct mitigation matrix M_corr = (M^mit)⁻¹

• Error Mitigated Execution:

  • Execute target circuit V on quantum device with desired parameters
  • Measure output probability distribution p_noisy
  • Apply correction: pmitigated = Mcorr · p_noisy
  • Extract error-mitigated expectation values from p_mitigated

Workflow Visualization

Diagram 1: Error Mitigation Strategy Selection Workflow (76 characters)

Diagram 2: Hybrid Error Detection and Mitigation Protocol (77 characters)

The Scientist's Toolkit: Essential Research Reagents

Table 4: Key Research Reagents and Computational Tools

Tool/Reagent	Function	Implementation Notes
Givens Rotation Circuits	Construct multireference states from single determinants [8]	Implement via parametric quantum gates; preserve particle number and spin symmetries
[[n,n-2,2]] QEDC	Quantum error detection with minimal overhead [22]	Use X⊗n and Z⊗n stabilizers; simple encoding/decoding circuits
Calibration Matrix Framework	Structure-preserving noise characterization [73]	Build using identity circuits with identical architecture to target circuits
Clifford Circuit Data	Training data for error mitigation models [74]	Efficiently simulatable classically; provides noisy/exact observable pairs
Pauli Twirling Sets	Randomize error signatures for improved mitigation [22]	Custom partial sets reduce overhead while maintaining effectiveness
Variational Quantum Eigensolver	Hybrid quantum-classical algorithm for ground state problems [8]	Optimize parameters iteratively using quantum measurements and classical optimization

The effective balancing of circuit expressivity and noise sensitivity represents a critical challenge in near-term quantum chemistry simulations. The protocols and analyses presented herein demonstrate that strategy selection must be guided by molecular electronic structure characteristics—with single-reference approaches sufficient for weakly correlated systems, while multireference methods like MREM essential for strongly correlated cases. The quantitative frameworks and experimental protocols provide researchers with practical tools for implementing appropriate error mitigation strategies. As quantum hardware continues to evolve, these approaches will enable more accurate and reliable chemical simulations on noisy intermediate-scale quantum devices, progressively bridging toward fault-tolerant quantum computation for pharmaceutical and materials design applications.

Benchmarking and Validating Mitigation Protocols for Chemical Accuracy

Within the pursuit of quantum advantage for chemistry simulations on Noisy Intermediate-Scale Quantum (NISQ) devices, Quantum Error Mitigation (QEM) has emerged as a pivotal, algorithmic-based strategy. Unlike quantum error correction, QEM aims to reduce the noise-induced bias in expectation values by post-processing outputs from an ensemble of circuit runs, without requiring a prohibitive qubit overhead [46]. This document establishes a comparative framework of key metrics to evaluate the performance of diverse QEM protocols, enabling researchers to select and refine strategies for precise molecular energy estimation—a critical task in fields like drug development and materials science. The efficacy of this framework is demonstrated through a contemporary case study on the BODIPY molecule, where a combination of techniques reduced measurement errors to near-chemical precision [9].

Core Performance Metrics for QEM

Evaluating a QEM protocol requires a multi-faceted analysis of its performance across several dimensions. The following metrics are essential for a comprehensive comparison.

Table 1: Key Metrics for Evaluating QEM Protocol Performance

Metric Category	Specific Metric	Definition and Interpretation
Accuracy Improvement	Absolute Error Reduction	The reduction in the difference between the estimated value and the true reference value (e.g., from 1-5% to 0.16% [9]).
	Bias Mitigation	The protocol's effectiveness in reducing systematic, noise-induced shifts in expectation values [46].
Resource Overhead	Sampling Overhead (Number of Shots)	The increase in the number of circuit executions (N_cir) required to achieve a target precision [46].
	Circuit Overhead	The number of additional or modified quantum circuits that must be run (e.g., for calibration) [9].
Computational Cost	Classical Post-processing Complexity	The time and computational resources required for the classical computation part of the QEM protocol.
Robustness & Generality	Noise Model Assumptions	The level of specificity required about the underlying quantum hardware noise (e.g.,是否需要精确噪声模型).
	Applicability to Different Circuits	The protocol's performance across varied circuit types (e.g., variational vs. deep circuits) and observables [75].
Precision & Reliability	Standard Error / Estimator Variance	A measure of the statistical precision, indicating the likely distance between the sampled mean and the true population mean [9].
	Formal Error Bounds	Theoretical guarantees on the maximum error of the mitigated expectation values [46].

Detailed Experimental Protocols from a Case Study

This section details the methodologies from a high-impact experiment that successfully estimated molecular energies for the BODIPY molecule on an IBM Eagle r3 quantum processor, achieving a final error of 0.16% [9]. The protocol provides a template for applying and evaluating QEM techniques in chemistry simulations.

System Preparation: BODIPY Molecular Energy Estimation

• Objective: To estimate the energy of the Hartree-Fock state for the BODIPY molecule in active spaces ranging from 8 to 28 qubits, targeting ground (S0) and excited states (S1, T1) [9].
• Key Reagents & Materials:
  • Molecule: Boron-dipyrromethene (BODIPY-4) in solution.
  • Quantum Hardware: IBM Eagle r3 quantum processor.
  • Initial State: Hartree-Fock state, chosen for its preparation simplicity (no two-qubit gates) to isolate measurement errors [9].
  • Hamiltonians: Molecular Hamiltonians for different active spaces (4e4o to 14e14o), each containing a fixed, complex structure of Pauli strings [9].
• Workflow:
  • State Preparation: Initialize the quantum processor to the separable Hartree-Fock state.
  • Hamiltonian Generation: For excited states (S1, T1), use specific techniques to generate Hamiltonians for which these states become the ground state [9].
  • Blended Execution: Run circuits for all three Hamiltonians (S0, S1, T1) interleaved with calibration circuits. This "blended scheduling" ensures temporal noise fluctuations affect all experiments equally, which is crucial for accurately estimating energy gaps (e.g., via ΔADAPT-VQE) [9].

Integrated QEM Techniques and Workflow

The experiment integrated several advanced QEM strategies into a cohesive workflow to tackle different sources of error.

Diagram 1: Integrated QEM workflow for high-precision molecular energy estimation, combining multiple techniques to address different error sources.

• Technique 1: Locally Biased Random Measurements for Shot Overhead Reduction

  • Protocol: Instead of measuring all Pauli terms in the Hamiltonian uniformly, this technique prioritizes measurement settings that have a larger impact on the final energy estimation. It employs a biased sampling strategy based on the Hamiltonian's structure, which significantly reduces the number of measurement shots (N_cir) required while maintaining the informationally complete nature of the data [9].

• Technique 2: Repeated Settings & Parallel Quantum Detector Tomography (QDT) for Circuit Overhead and Readout Error

  • Protocol:
    • Circuit Execution: A set of informationally complete measurement settings is sampled. Each setting is repeated multiple times (T) on the hardware [9].
    • Parallel QDT: Alongside the main experiment, dedicated circuits are executed to perform Quantum Detector Tomography. This characterizes the noisy measurement effects (POVMs) of the device [9].
    • Error Mitigation: The classical data from QDT is used to build an unbiased estimator. During post-processing, the noisy measurement effects are inverted to recover a much more accurate estimate of the expectation values, directly mitigating readout errors [9].

• Technique 3: Blended Scheduling for Time-Dependent Noise

  • Protocol: To combat temporal noise drift (e.g., slow fluctuations in detector performance), circuits for different tasks—such as estimating the S0, S1, and T1 energies, as well as QDT circuits—are interleaved in a single job submission. This ensures that no single energy estimation is disproportionately affected by a temporary "good" or "bad" noise period, leading to more homogeneous and reliable results [9].

The Scientist's Toolkit: Key Research Reagents & Solutions

Table 2: Essential Tools for Advanced QEM Experiments

Tool / Reagent	Function in QEM Protocol
Informationally Complete (IC) Measurements	A foundational measurement strategy that allows for the estimation of multiple observables from the same dataset and provides a seamless interface for error mitigation techniques like QDT [9].
Quantum Detector Tomography (QDT)	A calibration procedure used to fully characterize the noisy readout process of a quantum device. The resulting model is used to debias experimental data in post-processing [9].
Traffic Noise Model (TNM)	The Federal Highway Administration's standard model for predicting highway traffic noise, required for federally funded projects. It is used to evaluate noise impacts on sensitive land uses and design noise barriers [76].
Benchmarking Suites (e.g., QEM-Bench)	Standardized collections of datasets (22 in QEM-Bench) covering diverse quantum circuits and noise profiles. They provide a unified platform for fairly comparing and advancing different ML-based and conventional QEM methods [75] [65].
Machine Learning Models (e.g., QEMFormer)	Advanced ML models specifically designed for QEM. They leverage feature encoders that capture local, global, and topological information of quantum circuits to predict and correct errors, often with low sampling overhead [75].

Standardized Benchmarking and Evaluation

The development of benchmarks like QEM-Bench is critical for objectively assessing QEM protocols. This suite provides 22 datasets with diverse circuit types and noise profiles, enabling fair comparisons and highlighting the strengths of new methods like QEMFormer, which uses a two-branch model to capture both short- and long-range dependencies within a circuit [75] [65]. When reporting results, it is crucial to distinguish between absolute error (indicating accuracy and the presence of systematic bias) and standard error (indicating precision due to finite sampling) to properly diagnose an estimator's performance [9].

Protocol Implementation and Decision Framework

The choice of QEM protocol depends heavily on the constraints of the specific chemistry simulation. The following diagram outlines a decision-making framework for selecting and combining techniques.

Diagram 2: A decision framework for selecting QEM techniques based on the dominant constraints of a near-term quantum chemistry simulation.

Successful implementation requires validating the entire pipeline against classical simulations where possible and transparently reporting all overheads. The future of QEM lies in the intelligent combination of such techniques, supported by standardized benchmarking and machine learning, to push the boundaries of what is possible with near-term quantum hardware [46].

Within the Noisy Intermediate-Scale Quantum (NISQ) era, conducting meaningful quantum chemistry simulations, such as calculating molecular ground state energies, is fundamentally constrained by hardware noise and errors. Quantum error mitigation protocols are therefore pivotal for extracting reliable results from contemporary quantum processors. This application note provides a detailed head-to-head comparison of two such techniques—the duplicate circuit approach and the [[4,2,2]] quantum error-detecting code—specifically framed within the context of near-term chemistry simulations. The analysis is based on experimental implementations of the Variational Quantum Eigensolver (VQE) algorithm for determining the ground state energy of the H₂ molecule on IBM quantum hardware [77]. The findings indicate that, under realistic noise conditions including crosstalk, the duplicate circuit method demonstrates superior error mitigation performance compared to the [[4,2,2]] code, offering a more effective strategy for quantum computational chemistry on current devices.

The comparative analysis reveals a distinct performance differential between the two error mitigation techniques when applied to a practical quantum chemistry problem. The duplicate circuit method consistently outperformed the [[4,2,2]] quantum error-detecting code across multiple noise scenarios, including the presence of crosstalk errors which become significant when multiple circuit mappings are executed simultaneously on quantum hardware [77]. This superiority is attributed to the method's inherent robustness against the specific error mechanisms prevalent in today's superconducting quantum processors. For researchers targeting chemical accuracy in molecular energy calculations, the duplicate circuit approach presents a more pragmatic and effective error mitigation strategy for implementation on currently available NISQ devices.

Quantitative Performance Comparison

The following tables summarize the key quantitative findings from the experimental comparison of the two error mitigation techniques.

Table 1: Overall Performance and Characteristics

Metric	Duplicate Circuit Approach	[[4,2,2]] Quantum Error-Detecting Code
General Performance	Superior performance in presence of hardware noise [77]	Inferior performance compared to duplicate circuits [77]
Robustness to Crosstalk	Performed better when multiple mappings run simultaneously [77]	More significantly impacted by cross-talk noise [77]
Core Principle	Circuit-level redundancy and post-selection [77]	Quantum error detection via stabilizer measurements [77]
Primary Use Case	Error mitigation for NISQ algorithms [77]	Error detection for foundational QEC studies [77]

Table 2: Experimental Context from Zhang et al. [77]

Aspect	Description
Algorithm	Variational Quantum Eigensolver (VQE) [77]
Target Molecule	Hâ‚‚ Ground State Energy [77]
Hardware Platform	IBM Quantum Systems [77]
Noise Analysis	Performed with varying depolarising circuit noise and read-out errors [77]

Experimental Protocols

Protocol for Duplicate Circuit Implementation

The duplicate circuit method relies on executing multiple copies of the primary quantum circuit and comparing outcomes to mitigate errors.

• Circuit Preparation: Design the core VQE ansatz circuit for the Hâ‚‚ molecule. Create multiple identical copies (duplicates) of this primary circuit [77].
• Hardware Execution: Map all duplicate circuits onto the quantum processor. These can be executed either sequentially or, if hardware resources allow, on different sectors of the device simultaneously.
• Result Aggregation: Run the set of duplicate circuits and collect the measurement outcomes for each copy.
• Error Mitigation: Employ a post-processing technique, such as majority voting or consensus-based filtering, on the aggregated results from all duplicates to deduce a final, error-mitigated readout [77].

The workflow for this protocol is illustrated below:

Protocol for [[4,2,2]] Code Implementation

The [[4,2,2]] code is a quantum error-detecting code that encodes logical information into a larger physical Hilbert space.

• Logical Encoding: Encode two logical qubits into four physical qubits using the specific entanglement structure of the [[4,2,2]] code.
• Stabilizer Measurement: Integrate stabilizer measurement circuits into the main VQE algorithm flow. These circuits are designed to detect the occurrence of errors without collapsing the logical quantum state.
• Syndrome Extraction: Run the syndrome extraction circuits and measure the ancilla qubits to obtain an error syndrome.
• Error Detection & Post-Selection: Analyze the syndrome. Discard (post-select) any experimental run where a non-zero syndrome indicates that a detectable error has occurred. Retain only the results from runs with trivial syndromes for the final energy calculation [77].

The workflow for this protocol is illustrated below:

The Scientist's Toolkit

Table 3: Essential Research Reagents and Resources

Resource	Function/Description	Example/Note
Quantum Hardware	Executes the quantum circuits; source of characteristic noise.	IBM Quantum superconducting processors [77].
Classical Optimizer	Minimizes energy in VQE loop; chooses next circuit parameters.	Standard classical optimization library.
Error Mitigation Software	Implements post-processing for duplicate circuit results.	Custom scripts for majority voting.
QEC Code Library	Provides pre-defined stabilizer circuits for the [[4,2,2]] code.	Open-source quantum software SDK (e.g., Qiskit).
Molecular Representation	Defines the target chemistry problem for the VQE algorithm.	Hâ‚‚ molecule in a minimal basis set (e.g., STO-3G).

Discussion and Outlook

The experimental evidence favoring the duplicate circuit method highlights a critical consideration for near-term quantum chemistry: simpler, circuit-level error mitigation techniques can offer more net benefit than more complex, but smaller-scale, quantum error detection codes on today's hardware [77]. The [[4,2,2]] code, while a vital tool for foundational studies, introduces additional quantum gates and complexity for syndrome extraction. On NISQ devices, this overhead can inadvertently introduce more noise than it detects, thereby reducing the overall accuracy of the computation.

The field continues to progress rapidly. Recent demonstrations by industrial leaders, such as Quantinuum's full quantum error correction on a trapped-ion processor for a quantum chemistry simulation (Quantum Phase Estimation for Hâ‚‚), mark significant milestones on the path to fault tolerance [6]. These advances, combined with new hardware achieving error rates below the surface code threshold [78] and the integration of quantum control techniques to improve QEC circuit performance [79], paint an optimistic picture for the long-term. However, for immediate research applications in drug development and materials science, where computational feasibility is paramount, the duplicate circuit approach provides a strategically advantageous balance between mitigation efficacy and implementation complexity.

Statistical Analysis of Residual Bias and Error Scaling with Circuit Size

Within the broader thesis on enabling practical near-term chemistry simulations, this application note addresses a fundamental challenge: the systematic quantification of residual bias and error scaling in quantum error mitigation (QEM) protocols. As quantum computers transition from the Noisy Intermediate-Scale Quantum (NISQ) era toward early fault-tolerant capabilities, understanding the statistical behavior of errors is paramount for reliable quantum chemistry applications, such as molecular energy calculations for drug development [46]. Current quantum hardware suffers from uncorrected noise, which introduces significant biases in computed expectation values—the principal outputs for variational quantum eigensolvers and other quantum chemistry algorithms [80]. While quantum error correction (QEC) promises a long-term solution, its resource overhead remains prohibitive for near-term devices, placing heightened importance on error mitigation strategies that operate without physical qubit redundancy [81] [46].

Error mitigation encompasses algorithmic schemes that reduce noise-induced bias in expectation values through classical post-processing of an ensemble of circuit runs, without reducing the inherent noise level of individual circuit executions [46]. However, recent theoretical and experimental advances reveal these techniques face profound limitations. This note provides a rigorous statistical framework for analyzing these limitations, focusing on the scaling of residual errors and the sample complexity overhead required for chemical accuracy. We synthesize current theoretical bounds, provide protocols for empirical validation, and contextualize these findings for research scientists and drug development professionals seeking to leverage quantum simulation.

Theoretical Foundations and Quantitative Error Bounds

The efficacy of any QEM protocol is ultimately constrained by how errors accumulate and propagate through quantum circuits. A fundamental relationship exists between circuit scale (defined by qubit count n and depth d), underlying noise characteristics, and the resources required for effective mitigation.

Formal Definitions of Error Mitigation

The problem is formally defined as follows: upon input of a classical description of a noiseless circuit ( \mathcal{C} ) and observables ( \mathcal{M} = {Oi} ), and given copies of the noisy output state ( \sigma' ) from ( \mathcal{C}' ), the goal is to output estimates of ( \text{Tr}(Oi \sigma) ) (weak error mitigation) or samples from the distribution of ( \sigma ) (strong error mitigation) [82]. The number m of copies required is the sample complexity.

Fundamental Limitations and Scaling Laws

Recent analyses establish that the sample complexity for mitigating generic circuits scales super-polynomially or exponentially with system size, severely limiting the scalability of QEM for quantum advantage.

Table 1: Theoretical Bounds on Quantum Error Mitigation Sample Complexity

Circuit Characteristic	Noise Model	Mitigation Task	Sample Complexity Lower Bound	Key Implication
General Circuits [82]	Generic Local Noise	Weak Error Mitigation	( \exp(\Omega(n)) )	Super-polynomial samples needed for circuits beyond constant depth
Random & Structured Circuits [80]	Pauli Noise (e.g., Depolarizing)	Output Distribution Sampling	Polynomial in n, Exponential in ( 1/\eta ) (noise rate)	Quantum advantage constrained to a "Goldilocks zone" of qubit number vs. noise
Non-Unital Noise (e.g., T1 relaxation) [82]	Amplitude Damping	Weak Error Mitigation	( \exp(\Omega(n)) )	Highly relevant for superconducting qubits; error mitigation severely limited
Shallow Circuits [82]	Depolarizing Noise	Weak Error Mitigation	( \exp(\Omega(d)) ) (Depth d)	Logarithmic depth is the feasible regime without exponential cost

These bounds imply that even at shallow depths comparable to current experiments, mitigating errors for large qubit counts requires an infeasible number of circuit repetitions [82]. This is conceptually distinct from the exponential convergence of noisy states to the maximally mixed state with depth; here, the limitation kicks in at much smaller depths but depends critically on the circuit width n. For chemistry simulations, which require estimating expectation values of molecular observables, this directly translates to a rapidly growing residual bias unless an exponential number of samples are processed.

Experimental Protocols for Error Scaling Analysis

To empirically characterize error scaling in a research setting, the following protocols provide a systematic methodology.

Protocol 1: Benchmarking Residual Bias in Observable Estimation

This protocol measures the systematic error remaining in the estimated expectation value of an observable after applying a QEM technique.

• Primary Circuit (( \mathcal{C} )): A parameterized quantum circuit, such as a Variational Quantum Eigensolver (VQE) ansatz for a molecular Hamiltonian (e.g., Hâ‚‚, LiH).
• Noisy Circuit (( \mathcal{C}' )): The same circuit executed on noisy hardware or simulated with a realistic noise model (e.g., including T1, T2 relaxation and gate infidelities).
• Error Mitigation Protocol: A chosen QEM method (e.g., Zero-Noise Extrapolation (ZNE), Probabilistic Error Cancellation (PEC), or Clifford Data Regression (CDR)).
• Input: Classical description of ( \mathcal{C} ), ( O_i ), and noise model; budget of M total circuit shots.
• Procedure:
  • Baseline Calculation: Compute the ideal expectation value ( \mu = \text{Tr}(O \rho) ) using a noiseless simulator for a known, small-scale system.
  • Noisy Estimation: Estimate the unmitigated expectation value ( \nu = \frac{1}{M/3} \sum{m=1}^{M/3} om ) from M/3 shots of ( \mathcal{C}' ), where ( o_m ) is the measurement outcome for shot m.
  • Mitigated Estimation: Apply the chosen QEM protocol using 2M/3 shots to obtain the mitigated estimate ( \tilde{\nu} ). The shots may be allocated across multiple modified circuits (e.g., noise-scaled circuits for ZNE).
  • Bias Calculation: Calculate the residual bias for the unmitigated case ( B{\text{unmit}} = | \nu - \mu | ) and the mitigated case ( B{\text{mit}} = | \tilde{\nu} - \mu | ).
• Output: Residual bias values ( B{\text{unmit}} ) and ( B{\text{mit}} ). This process should be repeated while scaling the circuit size n (e.g., by increasing the number of qubits in the active space of the molecule) to establish a trend ( B(n) ).

Protocol 2: Sample Complexity Scaling for Chemical Accuracy

This protocol determines the number of samples m(n) required to maintain a fixed accuracy target (e.g., chemical accuracy of 1.6 mHa) as the problem size scales.

• Input: Target error threshold ε (e.g., 1.6e-3 Ha), a set of molecular systems of increasing size (e.g., Hâ‚‚, LiH, BeHâ‚‚), and their corresponding quantum circuits ( {\mathcal{C}_i} ).
• Procedure:
  • For each molecule/circuit ( \mathcal{C}_i ): a. Perform mitigated energy estimation using Protocol 1 with an initial shot budget M. b. Calculate the statistical error ( \delta E ) on the mitigated estimate (e.g., via bootstrapping or standard error of the mean). c. If ( \delta E > ε ), increase M iteratively until ( \delta E \leq ε ). d. Record the final required shot count m_i.
  • Plot m_i against the number of qubits n_i (or other scale metrics like Pauli string count).
• Output: A scaling function m(n) showing how the sample cost grows to maintain chemical accuracy. The results can be fitted to determine if the scaling is polynomial, ( m \sim n^k ), or exponential, ( m \sim \exp(\alpha n) ).

The logical flow and output dependencies of these protocols are summarized in the workflow below.

The Scientist's Toolkit: Research Reagent Solutions

The following table details key software and methodological "reagents" essential for conducting the described statistical analyses.

Table 2: Essential Research Reagents for Quantum Error Mitigation Analysis

Research Reagent	Type	Primary Function	Relevance to Error Scaling Analysis
Noise Model Simulators (e.g., Qiskit Aer, Cirq)	Software	Simulates execution of quantum circuits under realistic, user-defined noise models.	Provides a controlled environment for scaling studies without hardware access; allows isolation of specific noise channels (e.g., T1).
Error Mitigation Libraries (e.g., Mitiq, Ignis)	Software	Implements standard QEM protocols like ZNE, PEC, and CDR.	Serves as the standard implementation to benchmark against; enables consistent application of mitigation in scaling experiments.
Classical Simulation Algorithms (e.g., Pauli Path Simulators [80])	Software/Algorithm	Classically simulates noisy quantum circuits with complexity that scales with noise level.	Provides a classical baseline and helps validate the "Goldilocks zone" where quantum advantage might be possible.
Molecular Circuit Generators (e.g., OpenFermion, TEQUILA)	Software	Translates molecular Hamiltonians into parameterized quantum circuits (e.g., UCCSD).	Generates the realistic, structured circuits relevant for chemistry simulations, as opposed to random circuits.
Ensemble Averaging Framework [83]	Methodological	Defines a quantum channel via averaging over multiple approximate circuit compilations.	A method for rigorous error management; can provide worst-case error bounds (O(\epsilon^2)) for the overall computation.

Visualization of Statistical Relationships

The conceptual relationship between circuit scale, noise, and the cost of error mitigation is critical for planning feasible experiments. The following diagram synthesizes the core concepts discussed in this note.

The statistical analysis presented herein confirms that while quantum error mitigation is a powerful tool for extending the reach of near-term quantum devices, its application to large-scale chemistry simulations faces fundamental scalability constraints. The residual bias and the associated sample complexity for controlling it are predicted to grow exponentially with the number of qubits for generic circuits and noise models [82]. This implies that for drug development professionals, quantum simulations of large molecules will require either a prohibitive number of device runs or a breakthrough in mitigation strategies tailored to exploit specific structures in chemical circuits.

The path forward involves a co-design of algorithms, error mitigation protocols, and hardware. Research should focus on characterizing the empirical error scaling for specific, structured chemistry circuits, which may evade the worst-case theoretical bounds. Furthermore, the integration of error mitigation with early-stage error correction concepts, such as the dynamical decoupling and bosonic codes mentioned in industry roadmaps, may create hybrid approaches that push the boundary of the feasible regime [81]. For now, a rigorous statistical understanding of residual bias is not merely an academic exercise but a necessary tool for resource allocation and realistic goal-setting in the pursuit of quantum-enabled drug discovery.

For researchers in chemistry and drug development, validating results from quantum simulations against experimental data is a critical step in establishing the reliability of near-term quantum hardware. Achieving high-precision measurements on noisy intermediate-scale quantum (NISQ) devices is challenging due to inherent noise, readout errors, and significant resource overheads [84]. This application note details the frameworks, protocols, and error mitigation techniques essential for rigorous validation of quantum chemistry simulations on IBM Quantum hardware, contextualized within the broader pursuit of quantum advantage.

The quantum community is now systematically tracking progress toward quantum advantage—the point where quantum computations outperform all classical methods—through an open, community-led Quantum Advantage Tracker [85] [86] [87]. This tracker encourages the submission and rigorous validation of candidate advantage experiments, fostering a transparent dialogue between quantum and classical approaches. For domain scientists, this evolving benchmark provides a structured environment to test and verify their quantum simulation results against state-of-the-art classical methods.

Validation Frameworks and Community Benchmarks

The path to validated quantum results is a community effort. IBM, along with partners including Algorithmiq, the Flatiron Institute, and BlueQubit, has launched the Quantum Advantage Tracker, an open leaderboard to record, verify, and challenge advances in quantum and classical computation [87]. This initiative establishes a framework for scientifically validating claims of quantum utility and advantage.

Candidate experiments for quantum advantage are currently focused on three core problem areas [86] [88]:

• Observable Estimation: Precisely measuring the expectation values of quantum operators.
• Variational Algorithms: Hybrid quantum-classical algorithms for optimizing parameters, such as those used to find molecular ground states.
• Problems with Efficient Classical Verification: Tasks where the quantum output can be efficiently checked by a classical computer.

Validation within this framework requires satisfying two key criteria [88]:

• Rigorous Validation: The correctness of the quantum computer's output must be verifiable.
• Quantum Separation: The computation must demonstrate superior efficiency, cost-effectiveness, or accuracy compared to the best attainable classical methods.

Table 1: Key Hardware Systems for Validation Experiments

System/Processor	Key Features	Relevance to Validation
IBM Quantum Nighthawk [85] [86]	120 qubits, square lattice, 218 tunable couplers. Designed for 30% more circuit complexity.	Enables execution of more complex, chemically relevant circuits (up to 5,000 two-qubit gates) crucial for challenging classical methods.
IBM Quantum Heron [89] [86]	133/156 qubits, lowest median two-qubit gate errors. Record execution speed (330,000 CLOPS).	High-fidelity processor ideal for establishing baseline performance and testing foundational algorithms.
IBM Quantum Loon [85] [86]	Experimental processor demonstrating all hardware components for fault-tolerant quantum computing.	Provides a testbed for validating error correction and advanced mitigation strategies on the path to fault tolerance.

Error Mitigation Protocols for High-Precision Measurement

Error mitigation techniques are indispensable for extracting accurate results from current quantum hardware. The following protocols are essential for reducing errors in quantum chemistry simulations.

Advanced Readout Error Mitigation with Informationally Complete (IC) Measurements

This protocol leverages IC measurements to mitigate readout errors and reduce shot overhead, which was demonstrated effectively in the energy estimation of the BODIPY molecule [84].

A. Principle: IC measurements allow for the estimation of multiple observables from the same set of measurement data and provide a framework for implementing efficient error mitigation [84].

B. Experimental Workflow:

• Perform Quantum Detector Tomography (QDT):

  • Characterize the noisy measurement apparatus by preparing and measuring a complete set of basis states.
  • This constructs a response matrix that models the readout noise.

• Execute Locally Biased Random Measurements:

  • Instead of measuring all Pauli strings with uniform shots, bias the measurement budget towards settings that have a larger impact on the final energy estimation.
  • This technique maintains the IC nature of the measurements while significantly reducing the required number of shots (shot overhead) [84].

• Apply Blended Scheduling:

  • Interleave circuits for QDT and the actual chemistry simulation (e.g., Hamiltonian measurement) during hardware execution time.
  • This mitigates the impact of time-dependent noise (drift) on measurement precision [84].

• Post-Processing and Unbiased Estimation:

  • Use the noise model from QDT to correct the raw measurement outcomes from the chemistry simulation.
  • Employ the locally biased data to compute an unbiased estimator for the molecular energy.

C. Reported Outcome: Application of this protocol on an IBM Eagle processor for a BODIPY molecule's Hartree-Fock state reduced measurement errors by an order of magnitude, from 1-5% to 0.16%, bringing the result close to chemical precision (1.6x10⁻³ Hartree) [84].

Chemistry-Inspired Error Mitigation: From Single- to Multi-Reference States

For quantum chemistry applications, problem-specific error mitigation can be highly effective. Reference-state error mitigation (REM) and its extension, multireference-state error mitigation (MREM), are key techniques [8].

A. Reference-State Error Mitigation (REM) Protocol:

• Choose a Classically Solvable Reference State: Typically, the Hartree-Fock (HF) state is used. Its exact energy, (E_{ref}^{exact}), is calculated classically.
• Prepare and Measure the Reference State on Quantum Hardware: Run the quantum circuit that prepares the HF state and measure its energy, (E_{ref}^{noisy}).
• Prepare and Measure the Target State on Quantum Hardware: Execute the circuit for the target correlated state (e.g., VQE output) and measure its noisy energy, (E_{target}^{noisy}).
• Calculate the Mitigated Energy: The error-mitigated energy for the target state is computed as: (E{target}^{mitigated} = E{target}^{noisy} - (E{ref}^{noisy} - E{ref}^{exact})) This subtracts the systematic error observed in the reference state from the target state [8].

B. Multi-Reference-State Error Mitigation (MREM) Protocol:

REM's effectiveness is limited for strongly correlated systems where a single HF state has low overlap with the true ground state. MREM extends the protocol [8]:

• Generate a Multi-Reference State: Use an inexpensive classical method (e.g., CASSCF) to generate a compact wavefunction composed of a few dominant Slater determinants.
• Prepare the MR State via Givens Rotations: Implement a quantum circuit using Givens rotations to efficiently prepare the linear combination of Slater determinants. This method preserves physical symmetries like particle number.
• Execute the REM Protocol: Use this MR state as the reference state in the standard REM protocol. Its classically computed exact energy and its noisy hardware-measured energy are used to mitigate the error in the target state.

C. Reported Outcome: MREM has been shown in simulations to significantly improve computational accuracy for strongly correlated systems like the Fâ‚‚ molecule compared to the original single-reference REM method [8].

Table 2: Error Mitigation Techniques for Chemistry Simulations

Technique	Mechanism	Best For	Key Advantage
Probabilistic Error Cancellation (PEC) [86]	Inverts noise channels by combining results from many intentionally noisier circuits.	General purpose circuits.	Provides an unbiased estimate; can be integrated with Samplomatic for >100x overhead reduction [86].
Reference-State Error Mitigation (REM) [8]	Uses a classically tractable reference state (e.g., Hartree-Fock) to measure and correct hardware error.	Weakly correlated molecular systems.	Very low overhead; leverages chemical insight.
Multi-Reference Error Mitigation (MREM) [8]	Extends REM by using a compact multi-determinant state as a reference.	Strongly correlated systems (e.g., bond stretching).	Better overlap with true ground state than single-reference REM.
Dynamic Circuits [86]	Incorporates mid-circuit measurement and feed-forward classical control.	Algorithms requiring conditional operations.	Demonstrated 25% more accurate results and 58% reduction in two-qubit gates for a 100+ qubit Ising model [86].

The Scientist's Toolkit: Research Reagent Solutions

This section details the essential software and hardware "reagents" required to conduct validation experiments on IBM Quantum hardware.

Table 3: Essential Research Reagents for Quantum Validation

Tool / Resource	Type	Function in Validation
Qiskit SDK [86]	Software	The primary open-source software development kit for building, optimizing, and executing quantum circuits. Features like Samplomatic enable advanced error mitigation.
Qiskit Functions Catalog [88]	Software / Service	A catalog of application functions (e.g., from Algorithmiq, Q-CTRL) providing access to advanced, proprietary error mitigation and algorithm services as a service.
IBM Quantum Nighthawk [85]	Hardware	The forthcoming flagship processor, designed for increased circuit complexity, critical for testing problems beyond easy classical simulation.
C++ Interface & C-API [85] [86]	Software Interface	Enables deep integration of quantum routines into classical High-Performance Computing (HPC) workflows, which is essential for quantum-centric supercomputing.
Givens Rotations [8]	Algorithmic Component	A structured method to build multi-determinant states in quantum circuits for MREM, preserving particle number and spin symmetries.
Quantum Advantage Tracker [86] [87]	Benchmarking Framework	An open, community-led leaderboard to submit, verify, and challenge quantum advantage claims, providing the ultimate validation stage.

Validating quantum simulations against experimental results requires a multi-faceted approach combining advanced hardware, sophisticated software, and specialized error mitigation protocols. The path to quantum advantage is being systematically mapped through community-wide efforts like the Quantum Advantage Tracker, which provides a transparent forum for rigorous validation. For researchers in chemistry and drug development, mastering protocols such as IC-based measurement correction and multi-reference error mitigation is no longer a speculative endeavor but a practical necessity. By leveraging the tools and frameworks outlined in this application note, scientists can confidently use today's quantum hardware to push the boundaries of computational chemistry and materials science.

The accurate simulation of strongly correlated molecular systems represents one of the most promising yet challenging applications for near-term quantum computing. Such systems, including transition metal complexes, biradicals, and molecules with stretched bonds, are characterized by electronic near-degeneracies that render single-reference wavefunction approximations qualitatively incorrect [90]. While variational quantum algorithms like the Variational Quantum Eigensolver (VQE) offer a framework for tackling these problems on Noisy Intermediate-Scale Quantum (NISQ) devices, their performance is severely limited by hardware noise. This application note details specialized quantum error mitigation protocols, with emphasis on a novel Multireference-State Error Mitigation (MREM) technique, to address the unique challenges posed by strongly correlated systems in quantum computational chemistry.

The Computational Challenge of Strong Correlation

Defining Strongly Correlated Systems

In electronic structure theory, a system is considered "strongly correlated" when a single Slater determinant (e.g., the Hartree-Fock state) fails to provide a qualitatively correct description of its ground state wavefunction [90]. This occurs due to near-degeneracies where multiple configuration state functions (CSFs) become close in energy. The primary categories include:

• Transition metal complexes: Open-shell d or f orbitals create near-degenerate electronic configurations crucial for modeling catalysis and magnetic properties.
• Stretched or breaking bonds: As bonds are stretched, the electronic wavefunction evolves toward a biradical character, requiring a multiconfigurational description.
• Excited states: Dense manifolds of excited states in both organic and inorganic molecules often exhibit strong correlation effects.
• Molecular magnets: Systems with multiple weakly interacting magnetic centers display complex spin entanglement.

For these systems, the exact wavefunction is a linear combination of multiple Slater determinants with similar weights—a multireference (MR) state. When a single-reference method is used, the computational error can be unacceptably large [90] [8].

Limitations of Single-Reference Error Mitigation

Reference-state error mitigation (REM) is a cost-effective quantum error mitigation (QEM) strategy that corrects the energy error of a noisy target state by comparing it against a classically solvable reference state, typically the Hartree-Fock determinant [8]. However, its effectiveness is intrinsically linked to the overlap between the reference and target states. In strongly correlated systems, the Hartree-Fock state has minimal overlap with the true multireference ground state, causing standard REM to become unreliable in bond-stretching regions and for other multireference systems [8]. This fundamental limitation necessitates an error mitigation framework capable of handling multiconfigurational character.

Multireference-State Error Mitigation (MREM): A Specialized Protocol

Core Principles of MREM

Multireference-State Error Mitigation (MREM) extends the REM framework to strongly correlated systems by systematically incorporating multireference states into the error mitigation protocol [8]. The foundational principle is to use a compact, truncated multireference wavefunction—composed of a few dominant Slater determinants—that is engineered to exhibit substantial overlap with the target ground state. This wavefunction is derived from inexpensive classical methods and prepared on quantum hardware, providing a more accurate baseline for quantifying and mitigating hardware noise.

Theoretical Workflow

The MREM protocol integrates with variational algorithms like VQE. The key steps and their mathematical descriptions are summarized in the table below.

Table 1: Key Mathematical Components in the MREM Workflow

Component	Mathematical Description	Role in MREM Protocol
Electronic Hamiltonian	$\hat{H} = \sum{pq} h{pq} \hat{E}{pq} + \frac{1}{2} \sum{pqrs} V{pqrs} \hat{e}{pqrs} + V_{NN}$ [91]	Defines the target system and its energy spectrum.
Qubit Hamiltonian	$\hat{H}{qubit} = \sum{\alpha} c{\alpha} P{\alpha}$ (via Jordan-Wigner/Bravyi-Kitaev) [8]	Allows measurement of the energy expectation value on a quantum computer.
Noisy VQE Energy	$E_{VQE}(\theta) = \langle 0	\hat{U}^{\dagger}(\theta) \hat{H}_{qubit} \hat{U}(\theta)	0 \rangle$ (measured on hardware)	The noisy, unmitigated energy estimate for the target state.
MREM Corrected Energy	$E{MREM} = E{VQE} - (E{MR}^{noisy} - E{MR}^{exact})$ [8]	The final error-mitigated energy, where $E_{MR}$ is the energy of the multireference state.

Circuit Implementation via Givens Rotations

A pivotal aspect of MREM is the efficient preparation of multireference states on quantum hardware. This is achieved using Givens rotation circuits, which offer a structured, symmetry-preserving method to build linear combinations of Slater determinants [8].

• Function: Givens rotations apply a unitary transformation to a pair of orbitals, mixing their occupations to create superposition states from an initial reference determinant (e.g., Hartree-Fock).
• Advantages: These circuits are universal for state preparation in quantum chemistry, preserve particle number and spin symmetry, and can be efficiently compiled into standard quantum gates [8].
• Process: Starting from an initial state, a sequence of Givens rotations is applied, parameterized by angles that define the coefficients of the different determinants in the multireference state.

The following diagram illustrates the logical workflow of the MREM protocol, from classical pre-processing to the final error-mitigated energy.

Performance Data and Comparative Analysis

Application to Molecular Systems

The performance of MREM has been validated through comprehensive simulations of small molecules exhibiting strong correlation, such as the bond-stretching regions of Nâ‚‚ and Fâ‚‚ [8]. The table below summarizes the key findings.

Table 2: Performance of MREM on Diatomic Molecules in Bond-Stretching Regions

Molecule	Correlation Type	Single-Reference REM Performance	MREM Performance
Nâ‚‚	Static correlation upon bond stretching	Becomes unreliable as single-reference character is lost.	Significant improvement in accuracy; recovers correct energy trends.
Fâ‚‚	Pronounced multireference character even at equilibrium	Limited utility due to poor HF reference.	Effectively mitigates errors, yielding results closer to exact energies.
Hâ‚‚O	Weak correlation at equilibrium geometry	Effective, as HF is a good reference.	Maintains high accuracy, similar to REM.

Quantitative Error Scaling

Understanding the scaling of errors with circuit size is critical for assessing the long-term viability of error mitigation. Research indicates that without mitigation, the bias in energy estimation typically scales linearly with the number of gates, (O(\epsilon N)) [50]. After applying error mitigation protocols like probabilistic error cancellation or optimized formulas, this scaling can be suppressed to a sublinear growth, approximately (O(\epsilon' \sqrt{N})) [50]. This (\sqrt{N}) scaling is a consequence of the law of large numbers and implies that error mitigation can suppress errors by a larger factor in larger circuits, provided the noise is not overwhelming.

The comparative performance of different error mitigation strategies is visualized below.

Successful implementation of the protocols described in this note relies on a suite of classical and quantum computational tools.

Table 3: Essential Resources for Multireference Quantum Chemistry Simulations

Category / Name	Function / Description	Relevance to Protocol
Classical Computational Methods
CASSCF / DMRG	Generates high-quality multireference wavefunctions for active spaces.	Provides the initial compact MR state and exact energy ((E_{MR}^{exact})) for MREM.
Quantum Algorithms
VQE / VarQITE	Hybrid quantum-classical algorithms for ground-state energy estimation.	Serves as the primary algorithm whose noisy output ((E_{VQE})) is mitigated.
Error Mitigation Protocols
MREM	Multireference-State Error Mitigation (this note).	Corrects errors in strongly correlated systems.
REM	Reference-State Error Mitigation.	Baseline for weakly correlated systems.
Quantum Circuit Primitives
Givens Rotations	Constructs multireference states from a single determinant.	Core component for preparing the MR state on hardware in MREM.
Hardware Abstraction
Jordan-Wigner Transform	Maps fermionic operators to qubit (Pauli) operators.	Encodes the chemical Hamiltonian into a measurable form on a quantum computer.

Critical Limitations and Future Directions

Despite its promise, quantum error mitigation faces fundamental limitations. Theoretical studies framed through statistical learning theory indicate that mitigating errors in even shallow circuits can require a super-polynomial number of circuit samples in the worst case [44] [92]. This "hard limit" is particularly pronounced as system size (qubit count) increases, because noise can scramble quantum information at exponentially smaller depths than previously thought [92].

These constraints do not render error mitigation futile but rather define the boundaries within which practical applications must be developed. Future research directions likely involve:

• Co-designing algorithms and error mitigation to exploit problem-specific structure and avoid worst-case complexity.
• Tighter integration of quantum embedding theories, such as Density Matrix Embedding Theory (DMET), with quantum solvers to reduce the resource requirements for the quantum core [91].
• Development of fragmentation methods that break down large, strongly correlated systems into smaller, manageable subsystems amenable to simulation on near-term devices [91].

Conclusion

Quantum error mitigation has emerged as a pivotal and practical toolkit for extracting chemically meaningful results from NISQ devices, bridging the gap until full fault tolerance is realized. The exploration of foundational principles, advanced methodologies like hybrid QEC/QEM and multireference mitigation, sophisticated optimization techniques, and rigorous comparative validation collectively demonstrates that these protocols can significantly enhance the accuracy of quantum chemistry simulations, even for strongly correlated systems. For biomedical and clinical research, these advances pave the way for more reliable in silico drug discovery, including the accurate simulation of molecular interactions, protein-ligand binding affinities, and reaction mechanisms that are currently beyond the reach of classical computers. Future progress hinges on developing more scalable mitigation strategies with lower overhead and tailoring protocols specifically for the complex, multi-configurational molecules often encountered in pharmaceutical development, ultimately accelerating the design of novel therapeutics.

References

• 1. Noisy intermediate-scale quantum computing [https://en.wikipedia.org/wiki/Noisy_intermediate-scale_quantum_computing]
• 2. Quantifying the effect of gate errors on variational quantum ... [https://www.nature.com/articles/s41534-024-00808-x]
• 3. Error Management for a Quantum Chemistry Applicaiton on ... [https://archive.aps.org/smt/2025/mar-q34/11/]
• 4. Refining Noise Mitigation in NISQ Hardware Through Qubit ... [https://arxiv.org/html/2503.10204v1]
• 5. Error Mitigation in the NISQ Era: Applying Measurement ... [https://www.mdpi.com/2227-7390/12/14/2235]
• 6. Quantum Chemistry Gets Error-Corrected Boost from ... [https://thequantuminsider.com/2025/05/22/quantum-chemistry-gets-error-corrected-boost-from-quantinuums-trapped-ion-computer/]
• 7. Quantum Resource Management in the NISQ Era [https://arxiv.org/html/2508.19276v1]
• 8. Multireference error mitigation for quantum computation of ... [https://pubs.rsc.org/en/content/articlehtml/2025/dd/d5dd00202h]
• 9. Practical techniques for high-precision measurements on ... [https://www.nature.com/articles/s41534-025-01066-1]
• 10. Quantum Error Mitigation in the NISQ Era [https://medium.com/@raktims2210/quantum-error-mitigation-in-the-nisq-era-87af568290e5]
• 11. Thermal Noise Measurement Below the Standard Quantum ... [https://arxiv.org/html/2507.02196v1]
• 12. Decoherence in Quantum Computing: Causes, Effects, Fixes [https://www.spinquanta.com/news-detail/decoherence-in-quantum-computing-everything-you-need-to-know]
• 13. Quantum Decoherence: The Barrier to Quantum Computing [https://www.bluequbit.io/blog/quantum-decoherence]
• 14. Machine Learning assisted noise classification with ... [https://arxiv.org/html/2504.00718v1]
• 15. Noise in Quantum Computing [https://aws.amazon.com/blogs/quantum-computing/noise-in-quantum-computing/]
• 16. Noise-Resilient Quantum Metrology with ... [https://arxiv.org/html/2509.00771v1]
• 17. Quantum Error Correction: the grand challenge [https://www.riverlane.com/quantum-error-correction]
• 18. Choosing the right quantum error reduction strategy [https://q-ctrl.com/blog/choosing-the-right-quantum-error-reduction-strategy-a-practical-guide-to-error-suppression-error-mitigation-and-quantum-error-correction]
• 19. Differentiating quantum error correction, suppression, and ... [https://q-ctrl.com/topics/differentiating-quantum-error-correction-suppression-and-mitigation]
• 20. What is the difference between Error Mitigation (EM) and ... [https://quantumcomputing.stackexchange.com/questions/27398/what-is-the-difference-between-error-mitigation-em-and-quantum-error-correctio]
• 21. The quantum leap that needs error correction [https://qblox.com/newsroom/the-quantum-leap-that-needs-error-correction]
• 22. A Hybrid Protocol for Near-Term Quantum Simulation [https://arxiv.org/html/2510.01181v1]
• 23. Quantum computing squares up to the error-correction ... [https://www.techmonitor.ai/quantum/quantum-error-correction-breakthroughs/]
• 24. QEC25 [https://qec25.yalepages.org/]
• 25. VQE method: a short survey and recent developments [https://jmsh.springeropen.com/articles/10.1186/s41313-021-00032-6]
• 26. Improving VQE Parameter Quality on Noisy Quantum ... [https://arxiv.org/html/2508.15072v1]
• 27. BenchQC: A Benchmarking Toolkit for Quantum Computation [https://arxiv.org/html/2502.09595v1]
• 28. BenchQC: A Benchmarking Toolkit for Quantum Computation [https://pmc.ncbi.nlm.nih.gov/articles/PMC12614196/]
• 29. Understanding and mitigating noise in molecular quantum ... [https://pubs.rsc.org/en/content/articlehtml/2025/sc/d4sc05839a]
• 30. Quantum Error Mitigation (QEM) - Shafi - Medium [https://shafi-syed.medium.com/quantum-error-mitigation-qem-e23215676c25]
• 31. Digital zero-noise extrapolation with Catalyst [https://pennylane.ai/qml/demos/zne_catalyst]
• 32. Quantum Error Mitigation [https://www.alphaxiv.org/overview/2210.00921v3]
• 33. Compilation-informed probabilistic quantum error ... [https://arxiv.org/html/2508.20174v1]
• 34. of materials on Quantum - simulations computers near term quantum [https://www.nature.com/articles/s41524-020-00353-z?error=cookies_not_supported&code=892b9c88-1be3-4155-90c4-27d1c5070415]
• 35. Faster Probabilistic Error Cancellation [https://arxiv.org/html/2506.04468v1]
• 36. Noisy probabilistic error cancellation and generalized ... [https://www.nature.com/articles/s42005-025-02217-8]
• 37. Reduced Sampling Overhead for Probabilistic Error ... [https://quantum-journal.org/papers/q-2025-08-29-1840/]
• 38. Probabilistic error cancellation (PEC) with Mirror Circuits [https://mitiq.readthedocs.io/en/stable/examples/pec_tutorial.html]
• 39. Provable bounds for noise-free expectation values ... [https://www.nature.com/articles/s43588-024-00709-1]
• 40. Quantum error mitigation using energy sampling and ... [https://arxiv.org/html/2511.03556v1]
• 41. Improving the efficiency of learning-based error mitigation [https://quantum-journal.org/papers/q-2025-05-05-1727/]
• 42. Simulating noisy circuits for near-term quantum hardware [https://bloqade.quera.com/v0.28.3/blog/2025/07/30/simulating-noisy-circuits-for-near-term-quantum-hardware/]
• 43. Noise-agnostic quantum error mitigation with data ... [https://www.nature.com/articles/s41534-025-00960-y]
• 44. Quantum Error Mitigation May Face Hard Limits [https://thequantuminsider.com/2024/07/26/quantum-error-mitigation-may-face-hard-limits/]
• 45. How to use Clifford Data Regression for the MaxCut Problem [https://quantumcomputing.stackexchange.com/questions/21273/how-to-use-clifford-data-regression-for-the-maxcut-problem]
• 46. Quantum Error Mitigation [https://arxiv.org/html/2210.00921v3]
• 47. Methods of Modeling of Strongly Correlated Electron Systems [https://www.mdpi.com/2079-4991/13/2/238]
• 48. Multireference error mitigation for quantum computation ... [https://pubs.rsc.org/en/content/articlelanding/2025/dd/d5dd00202h]
• 49. [2211.08737] Near-Term Quantum Computing Techniques [https://arxiv.org/abs/2211.08737]
• 50. Error statistics and scalability of quantum error mitigation ... [https://www.nature.com/articles/s41534-023-00707-7]
• 51. Combining Error Detection and Mitigation - Quantum Physics [https://arxiv.org/abs/2510.01181]
• 52. Probabilistic Error Cancellation on Error-Detected ... [https://archive.aps.org/smt/2025/mar-q34/3/]
• 53. Hybrid Protocol Combining Error Detection And Mitigation ... [https://quantumzeitgeist.com/error-detection-quantum-performance-hybrid-protocol-combining-mitigation-improves-near-term-simulation/]
• 54. Error-Mitigation Enabled Multicomponent Quantum ... [https://arxiv.org/html/2511.11941v1]
• 55. Ground state energy of H2 vs experimental value [https://discuss.pennylane.ai/t/ground-state-energy-of-h2-vs-experimental-value/4579]
• 56. Ground-state energy estimation of the water molecule on a ... [https://www.nature.com/articles/s41534-020-0259-3]
• 57. Experimental data for N 2 (Nitrogen diatomic) [https://cccbdb.nist.gov/exp2x.asp?casno=7727379]
• 58. Ground and excited state energy calculations of the H2 ... [https://arxiv.org/abs/2407.01000]
• 59. [2311.02533] Precision ground-state energy calculation for ... [https://arxiv.org/abs/2311.02533]
• 60. Numerical MC SCF and CI calculations of the ground-state ... [https://www.sciencedirect.com/science/article/pii/0009261485870329]
• 61. Application to Ground-State Energy Estimation [https://quantum-journal.org/papers/q-2025-04-01-1682/]
• 62. Sample-efficient quantum error mitigation via classical ... [https://arxiv.org/html/2511.07092v1]
• 63. Quantum Error Mitigation Strategies [https://www.emergentmind.com/topics/quantum-error-mitigation-qem]
• 64. Low-overhead error detection with spacetime codes [https://arxiv.org/html/2504.15725v1]
• 65. QEM-Bench: Benchmarking Learning-based Quantum ... [https://openreview.net/forum?id=PSNce503is]
• 66. Mitigating noise in digital and digital–analog quantum ... [https://www.nature.com/articles/s42005-024-01812-5]
• 67. Digital–Analog Quantum Computing (DAQC) [https://qilimanjaro.tech/digital-analog-quantum-computing/]
• 68. Digital-Analog Quantum Computing for Higher-Order ... [https://kipu-quantum.com/knowledge-hub/blog/digital-analog-quantum-computing-for-higher-order-problems-at-the-quantum-advantage-level/]
• 69. Quantum error mitigation by Pauli check sandwiching [https://www.nature.com/articles/s41598-023-28109-x]
• 70. Logical error rate in the Pauli twirling approximation - PMC [https://pmc.ncbi.nlm.nih.gov/articles/PMC4588586/]
• 71. Neural Error Mitigation of Near-Term Quantum Simulations [https://arxiv.org/abs/2105.08086]
• 72. Error Sensitivity to Environmental Noise in Quantum ... [https://arxiv.org/abs/1602.01857]
• 73. Circuit structure-preserving error mitigation for High-Fidelity ... [https://arxiv.org/html/2505.17187v2]
• 74. Error mitigation with Clifford quantum-circuit data [https://quantum-journal.org/papers/q-2021-11-26-592/]
• 75. QEM-Bench: Benchmarking Learning-based Quantum Error ... [https://proceedings.mlr.press/v267/bao25c.html]
• 76. Highway Traffic Noise [https://www.fdot.gov/environment/oem-divisions/eng/highway-traffic-noise]
• 77. [2206.07907] Comparative analysis of error mitigation ... [https://arxiv.org/abs/2206.07907]
• 78. Quantum error correction below the surface code threshold [https://www.nature.com/articles/s41586-024-08449-y]
• 79. Improving quantum error correction circuits by 70% with ... [https://q-ctrl.com/blog/improving-quantum-error-correction-circuits-by-70-with-quantum-control]
• 80. Constraints on Quantum-Advantage Experiments Due to ... [https://cs.uchicago.edu/news/constraints-on-quantum-advantage-experiments-due-to-noise/]
• 81. Quantum Report Says Error Correction Now The Industry's ... [https://thequantuminsider.com/2025/11/19/quantum-report-says-error-correction-now-the-industrys-defining-challenge/]
• 82. Exponentially tighter bounds on limitations of quantum ... [https://www.nature.com/articles/s41567-024-02536-7]
• 83. Application Scale Quantum Circuit Compilation with ... [https://arxiv.org/html/2510.18000v3]
• 84. Practical Techniques for High-Precision Measurements on ... [https://arxiv.org/html/2409.02575v2]
• 85. IBM Delivers New Quantum Processors, Software, and ... [https://newsroom.ibm.com/2025-11-12-ibm-delivers-new-quantum-processors,-software,-and-algorithm-breakthroughs-on-path-to-advantage-and-fault-tolerance]
• 86. Scaling for quantum advantage and beyond [https://www.ibm.com/quantum/blog/qdc-2025]
• 87. Quantum Computers Model Complexity of Materials [https://thequantuminsider.com/2025/11/12/quantum-computers-model-complexity-of-materials/]
• 88. The dawn of quantum advantage [https://www.ibm.com/quantum/blog/quantum-advantage-era]
• 89. IBM Quantum Computing | Hardware and roadmap [https://www.ibm.com/quantum/hardware]
• 90. Electronic structure of strongly correlated systems [https://pmc.ncbi.nlm.nih.gov/articles/PMC9261488/]
• 91. Multireference Embedding and Fragmentation Methods for ... [https://arxiv.org/html/2505.13394v2]
• 92. Exponentially tighter bounds on limitations of quantum ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC11473368/]