Researchers from Ames National Laboratory, William & Mary, Iowa State University, Pennsylvania State University, Saarland University, Rochester Institute of Technology, and the Flatiron Institute have demonstrated a new approach to modeling the complex behavior of correlated electrons. The team’s time-dependent Gutzwiller quantum-classical embedding framework, t-GQCE, simulates these interactions by mapping lattice dynamics onto both classical and quantum processors.
Remarkably, the embedding problem can be reduced to a single qubit, successfully simulated on IBM quantum hardware using just two physical qubits total including an ancilla for measuring the quantum metric. This work demonstrates a path toward efficient quantum computing applications for materials science, suppressing artificial oscillations found in conventional approximations.
Gutzwiller Quantum-Classical Embedding for Correlated Electron Dynamics
This approach circumvents the limitations of purely classical methods when simulating correlated electron systems, which demand nonperturbative treatment of local interactions over extended timescales. The framework’s design allows flexibility, remaining independent of the specific quantum dynamics solver employed for the embedding models. The t-GQCE method distinguishes itself by reducing the quantum computational burden to a remarkably small scale; without ghost orbitals, the embedding problem collapses to require only a single qubit.
The framework’s efficacy was demonstrated through adaptive variational quantum dynamics simulations of interaction quenches within the Hubbard model on the Bethe lattice. Increasing the number of ghost orbitals within the t-GQCE framework actively suppresses artificial undamped oscillations inherent in the conventional Gutzwiller approximation.
Results then converge toward those obtained using nonequilibrium dynamical mean-field theory, validating the method in both weak and strong quench regimes. This convergence confirms t-GQCE’s ability to accurately model the system’s evolution across a range of conditions.
Statevector simulations with 2, 4 and 6 ghost orbitals per spin show that the adaptive circuits follow the embedding states with infidelities below 5× . With error rates three times lower than those of a noise model benchmarked against trapped-ion hardware, the framework reaches a target accuracy of 0.01 for the single-particle density matrix and the double occupancy with shots per time step.
Time-Dependent Gutzwiller Approximation and Undamped Oscillations
The conventional Gutzwiller approximation, while computationally efficient due to its limited embedding problem of only one bath orbital per correlated orbital, produces artificial, undamped oscillations of local observables, a limitation addressed by the time-dependent ghost Gutzwiller approximation. Specifically, simulations with seven bath orbitals per spin yielded a double occupancy of 0.100, closely aligning with the 0.104 value from the nonequilibrium dynamical mean-field theory at a time of five units.
The Dirac-Frenkel variational principle underpinned the dynamics studies, yielding a Lagrange function within the time-dependent ghost Gutzwiller approximation that defines the evolution of embedding states. These embedding states, represented by coefficients encoding the variational matrix, are defined within the Fock space of physical and bath modes, and their propagation becomes the primary computational bottleneck as the number of ghost orbitals increases. The method’s efficiency stems from mapping lattice dynamics onto the coupled evolution of a noninteracting quasiparticle lattice, propagated on classical processors and finite interacting embedding models, propagated on quantum processors.
Mapping Lattice Dynamics with Quantum and Classical Processors
The time-dependent Gutzwiller quantum-classical embedding framework, or t-GQCE, maps lattice dynamics by propagating a noninteracting quasiparticle lattice on classical processors alongside interacting embedding models on quantum hardware. This division of labor minimizes the quantum computing resources needed to simulate correlated electron systems, a long-standing challenge in materials science. The framework exchanges only embedding Hamiltonian parameters and single-particle density matrices between the classical and quantum components at each time step, enabling a hybrid approach to modeling complex interactions.
This method reduces the quantum computational burden by treating the majority of the system, the noninteracting environment, classically. The framework’s adaptability extends to the choice of quantum dynamics solver. Near-term variational algorithms and fault-tolerant Hamiltonian simulations are both compatible, offering flexibility as quantum technology advances.
Without incorporating additional computational elements, the embedding problem itself collapses to a single qubit, a surprising result given the complexity of the underlying physics. This convergence validates the accuracy of the hybrid quantum-classical approach and demonstrates its potential to provide reliable insights into the behavior of correlated electron systems across both weak and strong quench regimes.
Adaptive Variational Quantum Dynamics for Interaction Quenches
Simulations of interaction quenches on the Hubbard model, a system used to understand correlated electron behavior, have been performed using only two physical qubits on IBM quantum hardware, including one serving as an ancilla to measure the quantum metric. For this demonstration, the adaptive variational quantum dynamics simulation (AVQDS) algorithm was employed to propagate the embedding states, constructing compact circuits tailored to the problem with controllable accuracy. Results demonstrate that the framework effectively suppresses artificial undamped oscillations often seen in conventional Gutzwiller approximations, bringing the simulations into closer agreement with those produced by nonequilibrium dynamical mean-field theory.
Simulations with varying numbers of ghost orbitals, which enhance the approximation, further improved convergence with the results of the more computationally intensive nonequilibrium dynamical mean-field theory. The team performed simulations quenching the interaction strength from 0.05 to target values, examining the system’s behavior under both weak and strong quenches.
Single Qubit Embedding for Hubbard Model on Bethe Lattice
The team used IBM quantum hardware to simulate the single-qubit dynamics, employing an additional qubit as an ancilla specifically for measuring the quantum metric. This configuration, requiring only two physical qubits total, yielded results closely mirroring those obtained in noiseless simulations. These orbitals suppress artificial undamped oscillations, a common issue in conventional Gutzwiller approximations, and improve convergence toward more accurate results. Circuit compression, measurements in rotated orbital bases with post-selection, and noise mitigation reduce the errors severalfold.
Embedding State Infidelity with 2, 4, and 6 Ghost Orbitals
This approach allows for a more detailed variational description of the embedding state, capturing quantum fluctuations beyond simpler approximations. The framework’s performance was assessed using calculations with two, four and six ghost orbitals, designated by B = 3, 5, and 7 respectively, demonstrating significant improvements in capturing the dynamics of the Hubbard model, particularly in suppressing artificial oscillations in double occupancy. The inclusion of these additional orbitals expands the Hilbert space of the embedding state, allowing for a richer representation of quantum behavior.
The complexity of simulating the embedding state |Φ(t)⟩ increases with each added ghost orbital, preventing the reduction to a single-qubit problem achieved with B = 1. This process yields a set of equations of motion governing the evolution of both the quasiparticle lattice on classical processors and the embedding states on quantum processors.
Error Mitigation Strategies for Embedding Observable Measurement
These results demonstrate that error mitigation strategies can significantly improve the accuracy of quantum simulations of correlated electron systems, even with limited quantum resources. The accuracy of the measurement is ultimately limited by a combination of shot noise and residual noise bias remaining after mitigation, with the latter being particularly sensitive to hardware error rates.
Researchers found that combining orbital-rotation measurements, post-selection, symmetry averaging and mirror-circuit mitigation reduced errors by a factor of approximately three. This suggests that improvements in hardware error rates will directly translate to increased simulation accuracy, potentially unlocking more complex and realistic material modeling.
👉 More information
🗞 Gutzwiller Quantum-Classical Embedding for Dynamics Simulations of Correlated Electron Systems
✍️ Joshua Aftergood, Feng Zhang, Thomas Iadecola, Peter P. Orth, Cai-Zhuang Wang and Yong-Xin Yao (Ames National Laboratory, U.S. Department of Energy, Ames, Iowa, USA); Daniele Guerci (Department of Physics, William & Mary, Williamsburg, Virginia, USA); Jacob Brunton (Department of Physics and Astronomy, Iowa State University, Ames, Iowa, USA); Nicola Lanatà (School of Physics and Astronomy, Rochester Institute of Technology, Rochester, New York, USA)
🧠 ArXiv: https://arxiv.org/abs/2611.24592v1




See today’s quantum computing news on Quantum Zeitgeist for the latest breakthroughs in qubits, hardware, algorithms, and industry deals.
