SpinPulse: Open‑Source Pulse‑Level Simulation of Spin Qubit Computers with Non‑Markovian Noise

Benoît Vermersch, Oscar Gravier, and Nathan Miscopein, together with colleagues from Université Grenoble Alpes and Quobly, introduce SpinPulse, a new open‑source Python package that simulates spin‑qubit computers at the pulse level while incorporating detailed modelling of non‑Markovian noise. This capability lets researchers move beyond idealised simulations, testing the resilience of quantum circuits against the imperfections inherent in physical qubits and accelerating progress in hardware development and error‑mitigation strategies. By bridging the gap between abstract quantum algorithms and the complexities of physical implementation, SpinPulse promises to be a significant asset to the quantum‑computing community.

SpinPulse uniquely simulates spin qubit computers at the pulse level, incorporating detailed modelling of non‑Markovian noise, crucial for understanding real‑world device limitations. This capability allows researchers to test the resilience of quantum circuits against imperfections inherent in physical qubits, accelerating progress in hardware development and error‑mitigation strategies. By bridging the gap between abstract algorithms and physical implementation, SpinPulse promises to be a significant asset to the quantum‑computing community.

---

Spin Qubit Simulation with Classical Noise Modelling

The research team developed SpinPulse, an open‑source Python package designed for simulating spin‑qubit‑based computers at the pulse level, addressing a critical need for realistic hardware development tools. This work pioneers a methodology that models the specific physics of spin qubits, crucially incorporating classical non‑Markovian noise to accurately represent experimental conditions.

A circuit, initially formulated in languages such as Pennylane or Cirq, is converted to OpenQASM using the qiskit.qasm3 module and then adapted to the model’s native gate set via a two‑stage transpilation process:

1. Gate transpilation – performed with parametrised Qiskit transpilation pass managers, followed by a specific decomposition of RZZ gates.

2. Pulse transpilation – translates gates into precisely defined pulse sequences.

The team engineered a hierarchical system comprising PulseLayer, PulseSequence, and PulseInstruction classes to manage this conversion, decomposing an instruction‑set‑architecture (ISA) circuit into layers where each qubit participates in at most one gate. Each layer is assembled into PulseSequence objects detailing single‑ and two‑qubit pulses, with PulseInstruction subclasses governing the Hamiltonian evolution at each time step.

The method from_circuit within the PulseCircuit class orchestrates these steps, culminating in a complete pulse‑level description of the circuit, visualised with the plot method. Numerical integration forms the final stage of the simulation pipeline, converting the PulseCircuit object back into a gate circuit to calculate the overall evolution operator (via the to_circuit method), generating a Qiskit circuit for simulation with tools such as qiskit_aer. This approach achieves high fidelity by approximating the SW adiabatic transformation and integrating the Heisenberg interactions described in the model, fostering the development of high‑fidelity quantum circuits.

---

Spin Qubit Simulation with Realistic Noise Models

SpinPulse introduces a method for simulating the specific physics of spin qubits, notably incorporating classical non‑Markovian noise to accurately represent experimental conditions. Experiments demonstrate that circuits are first transpiled into a native gate set, then converted into a pulse sequence, which is numerically integrated within a simulated noisy environment. This supports hardware development by enabling detailed simulations of native gates and circuits.

Key findings include:

• Optimised concatenated CPMG sequences maximise the number of (2nX) pulses within a finite time window, resulting in a negligible effective idle duration and simplifying the pulse‑level description.

• Fluctuations in coupling constants within the Heisenberg Hamiltonian are modelled as noisy fields with a coherence time (T^{*}_{J}), analogous to qubit frequency shifts.

• Gate‑exchange noise leads to a gate (R_{ZZ}(\theta + \Delta\theta)), where (\Delta\theta) is the integral of the noisy field over time, approximated by (\int \varepsilon_{i,j}(t'),dt').

• Measurements confirm the relationship between gate‑exchange noise and a dimensionless contrast quantity (C(t)), allowing parametrisation of time traces (\varepsilon_{i,j}(t)) using the coherence time (T^{*}_{J}).

The study establishes a modular API that realises each simulation step, enabling users to independently execute and visualise each stage. Hardware specifications (e.g., number of qubits, coupling constants) are encapsulated in a HardwareSpecs class. Gate transpilation from Qiskit circuits into the model’s native gate set employs parametrised Qiskit transpilation pass managers. This comprehensive framework paves the way for high‑fidelity pulse‑level simulations and improved mitigation strategies for quantum computing.

---

Realistic Spin Qubit Simulation with Non‑Markovian Noise

SpinPulse operates by first converting a quantum circuit into a native gate set, then translating these gates into a time‑dependent Hamiltonian pulse sequence that incorporates simulated noise. This approach enables realistic simulations of quantum computations, supporting hardware development and improved control strategies. By modelling non‑Markovian noise, SpinPulse distinguishes itself from existing pulse‑level simulators and provides a more accurate representation of spin‑qubit behaviour.

The package integrates with the quimb tensor‑network library, enabling large‑scale simulations and workflows that encompass transpilation, pulse‑level compilation, hardware benchmarking, and noise mitigation. While the authors acknowledge limitations inherent in simplifying complex physical systems, the modular design of SpinPulse facilitates future expansion and refinement. Planned extensions include additional physical models and features that will further enhance the package’s capabilities.

The researchers anticipate that SpinPulse will serve as a valuable tool for the quantum‑computing community, aiding in the design of high‑fidelity circuits and the development of effective noise‑mitigation techniques—representing a significant step toward bridging the gap between theoretical quantum computation and practical hardware implementation for spin qubits.

---

More information

• ArXiv: https://arxiv.org/abs/2601.10435