🔬 materials science

Strong Coupling Quantum Impurity Solver on the Real and Imaginary Axes

This paper introduces a bold hybridization-expansion quantum Monte Carlo solver that efficiently implements strong-coupling expansions with closed-form real-axis Feynman rules to achieve rapid convergence and high-accuracy spectroscopy for quantum impurity problems on both real and imaginary frequency axes.

Original authors: Yueyi Wang, Kristjan Haule

Published 2026-09-22

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: Yueyi Wang, Kristjan Haule

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

In the microscopic world of materials science, understanding how electrons move through solids is a quest that has driven physicists for decades. When electrons interact with one another, they create a complex, crowded environment where their behavior cannot be predicted by looking at them in isolation. To make sense of this, scientists often use a powerful theoretical tool called dynamical mean-field theory. This approach simplifies the problem by focusing on a single, representative atom and treating the rest of the material as a surrounding "bath" that constantly exchanges energy and particles with it. The challenge lies in solving the math for this single atom when the interactions are strong. For years, researchers have relied on methods that work well when interactions are weak, but these tools struggle when the electrons are tightly bound or when scientists need to know exactly how the material responds to light or electricity at specific frequencies. Without accurate answers for these real-world frequencies, predicting the properties of new materials remains a guessing game.

A team of researchers at Rutgers University has now developed a new computational method that cuts through this difficulty, allowing them to solve these strong-interaction problems directly on the real frequency axis. Instead of working with abstract, imaginary numbers that require a difficult and often unstable mathematical translation to get back to reality, their new approach calculates the results directly in the language of real frequencies. This is a significant shift because previous attempts to do this directly were so computationally expensive that they were impractical for all but the simplest cases. By using a technique called the strong-coupling expansion, which treats the interaction between the atom and its surroundings as the starting point rather than a small disturbance, the researchers found that the math simplifies dramatically. They discovered that the complex web of interactions, which usually explodes in complexity as you try to calculate higher levels of detail, actually collapses into a manageable form. This allows them to simulate the system with high precision without needing to perform the unstable mathematical translation that has long plagued the field.

The researchers tested their new solver, which they named the bold hybridization-expansion quantum Monte Carlo method, against the most accurate existing techniques for the Hubbard model, a standard theoretical framework for describing electrons in solids. They focused on a scenario known as the Mott transition, where a material can switch from being a conductor to an insulator as the electron interactions grow stronger. In these tests, their method produced results that matched the gold-standard numerical renormalization group calculations almost perfectly across the entire range of frequencies and interaction strengths. The agreement was so close that it validated the new approach as a reliable way to see the detailed "spectrum" of the material, revealing how electrons scatter and lose energy. This level of accuracy is crucial because it allows scientists to predict how a material will behave under different temperatures and how it will respond to external probes, such as light or magnetic fields, without the errors introduced by older approximation methods.

One of the most striking findings of this work is how the new method handles the transition between metallic and insulating states. At a specific interaction strength, the material undergoes a phase change where the electrons stop flowing freely. The researchers were able to map out this transition with high confidence, showing that their method correctly identifies the point where the material becomes an insulator. They also observed that the scattering rate of the electrons, which determines how easily they move, follows a predictable pattern at low temperatures, confirming that the material behaves like a Fermi liquid, a standard state of matter for interacting electrons. This precision is not just a theoretical victory; it provides a solid benchmark for future studies of more complex materials, such as high-temperature superconductors, where the interactions are even more intense and difficult to model.

Beyond static properties, the researchers used their new tool to investigate the electrical resistance of the material, a property that is notoriously difficult to calculate accurately. In previous attempts, simulations often failed to match experimental data from cold-atom experiments, which measure the resistance of similar systems in a controlled environment. The discrepancy was suspected to be due to missing "vertex corrections," which are subtle adjustments to how electrons carry current that are often ignored in simpler models. By applying their new solver to a cluster version of the theory that includes these corrections, the researchers found that the calculated resistance matched the experimental data much more closely, especially at higher temperatures. This suggests that the missing piece in earlier theories was indeed the way electrons interact with each other to transport charge, and that their new method can capture these interactions with the necessary fidelity.

The implications of this work extend to the design of future materials. Because the method works efficiently on both imaginary and real frequency axes, it can be used to calculate not only the static properties of materials, like their total energy, but also their dynamic responses, such as how they absorb light or conduct electricity. This dual capability means that scientists can now simulate realistic materials with a level of detail that was previously out of reach, bridging the gap between abstract theory and experimental observation. The researchers note that their approach is particularly well-suited for problems where other methods fail due to severe computational bottlenecks, opening the door to studying complex systems like the p-d model of cuprates, which are central to the mystery of high-temperature superconductivity.

In the end, this paper represents a significant step forward in our ability to simulate the quantum world. By finding a way to bypass the mathematical hurdles that have limited progress for years, the authors have provided a new, robust tool for exploring the behavior of electrons in strongly interacting materials. Their work demonstrates that with the right mathematical perspective, the most difficult problems in quantum many-body physics can be solved with surprising efficiency. The results are not just a theoretical curiosity; they offer a clear path to understanding and predicting the properties of the next generation of electronic materials, from superconductors to advanced semiconductors, with a reliability that brings us closer to truly mastering the quantum realm.

Technical Summary: Strong Coupling Quantum Impurity Solver on the Real and Imaginary Axes

Problem Statement
Quantum many-body problems are frequently addressed by expanding the action in powers of a control parameter and summing the resulting perturbative series. While diagrammatic Monte Carlo (DiagMC) has been highly successful for weak-coupling expansions across various systems (e.g., uniform electron gas, unitary Fermi gas), these methods typically rely on a weak-coupling limit where Wick's theorem applies straightforwardly. Strong-coupling expansions, conversely, are less common due to the absence of Wick's theorem and the complexity of their diagrammatic rules.

A critical bottleneck in existing strong-coupling solvers, such as the hybridization-expansion continuous-time quantum Monte Carlo (CT-HYB) method, is their formulation in imaginary time. To obtain real-frequency observables (essential for spectroscopy and transport), these methods require numerical analytic continuation, an ill-posed and unstable procedure. While real-time formulations for weak-coupling expansions have been developed to bypass this, they suffer from an exponential growth in computational cost with perturbation order, as a single imaginary-axis term generates at least $2^n$ real-frequency contributions at order $n$ . The challenge remains to develop a strong-coupling expansion that is efficient on the real axis without incurring this exponential penalty.

Methodology
The authors present a bold strong-coupling expansion implemented via a diagrammatic Monte Carlo (DiagMC) algorithm, termed Bold Hybridization-Expansion Quantum Monte Carlo (BHQMC). The method is applied to the quantum impurity problem, a central component of Dynamical Mean-Field Theory (DMFT).

Pseudo-particle Formulation: The atomic Hamiltonian is diagonalized, and the action is expressed in terms of pseudo-particles ( $a_m$ ) that create atomic eigenstates. This formulation allows Wick's theorem to hold in the grand-canonical space, simplifying the diagrammatic expansion.
Projection to Physical Hilbert Space: The enlarged Hilbert space of pseudo-particles includes unphysical sectors. The authors enforce the constraint of exactly one pseudo-particle per site ( $Q=1$ $Q = 1$ ) by introducing a pseudo-particle chemical potential $\lambda$ $λ$ and taking the limit $\lambda \to \infty$ $λ \to \infty$ .
- On the real-frequency axis, this projection shifts all pseudo-particle propagators as $G_p(\omega) \to G_p(\omega + \lambda)$ , while leaving the hybridization function $\Delta(\nu)$ unshifted.
- Crucially, this projection suppresses contributions from branch cuts of the pseudo-particle propagators, leaving only the branch cut of the hybridization function.
Simplification of Frequency Summation:
- In conventional real-time weak-coupling expansions, summing over $n$ hybridization frequencies generates $2^n$ terms.
- In the proposed strong-coupling formulation, the projection leads to a drastic simplification: the summation over each hybridization frequency $i\nu_m$ reduces to a single real-axis integral involving the hybridization spectral function $A_c$ and the Fermi function.
- Consequently, the pseudo-particle self-energy reduces to a single $n$ -dimensional integral. The impurity Green's function requires a summation over the backbone frequency $i\omega$ , generating at most $2n$ projected terms. This is negligible compared to the $2^n$ terms in weak-coupling real-time formulations.
Algorithm: The authors sample the Luttinger-Ward functional of the strong-coupling expansion, retaining only skeleton diagrams. Updates include changes in time/frequency, insertion/removal of hybridization lines, and vertex exchanges. The pseudo-particle propagators are fully dressed by the self-energy derived from the Luttinger-Ward functional. Physical electron quantities are computed via functional derivatives, utilizing Bulla's trick for the self-energy to improve numerical precision.

Key Contributions

Closed-Form Real-Axis Feynman Rules: The authors derive explicit rules for evaluating diagrams of arbitrary order directly on the real axis. These rules replace Matsubara sums with integrals over real variables, leveraging the projection to eliminate the exponential explosion of terms.
Efficient Real-Axis Implementation: The method demonstrates that strong-coupling expansions can be carried out nearly as efficiently on the real axis as on the imaginary axis, a significant departure from the limitations of weak-coupling real-time methods.
Bold Expansion Convergence: By sampling the Luttinger-Ward functional (proportional to $\log Z$ ) rather than the partition function $Z$ itself, the bold expansion converges substantially faster than the bare expansion (CT-HYB), resulting in lower average expansion orders.

Results

Convergence: Benchmarks on the 2D square-lattice Hubbard model show that the bold expansion converges much faster than the bare expansion. Higher-order contributions to the spectral function fall off rapidly and concentrate near zero frequency.
Spectral Functions: The method produces highly accurate real-axis spectra for the DMFT Mott transition of the Hubbard model. Results show excellent agreement with state-of-the-art Numerical Renormalization Group (NRG) data across the entire frequency range and various interaction strengths.
- At finite temperature ( $T=0.04t$ ), the method correctly identifies the Mott transition around $U \approx 10t$ , consistent with finite-temperature expectations, whereas zero-temperature NRG results show a transition at a higher $U_c$ .
- For the Bethe lattice, the method resolves sharp structures at the inner edges of Hubbard bands with accuracy comparable to finite-size scaling DMRG results.
Transport Properties: The method was applied to calculate the resistivity of the 2D Hubbard model.
- Single-site DMFT results agreed with previous estimates at low temperatures but deviated at high temperatures.
- By incorporating longitudinal vertex corrections via the Ward identity within a 2x2 Dynamical Cluster Approximation (DCA), the authors recovered the correct high-temperature trend, aligning with cold-atom experimental data and finite-temperature Lanczos calculations. This highlights the necessity of vertex corrections in two dimensions.
Multi-orbital Benchmarks: The method was tested on a two-band $e_g$ Kanamori model, showing quantitative agreement with CT-HYB at expansion order $n=5$ . The real-axis implementation required significantly fewer core-hours (approx. 330) compared to imaginary-axis bold sampling (approx. 750) or inchworm Monte Carlo (approx. 1500).

Significance and Claims
The paper claims that strong-coupling expansions offer two key advantages in the DMFT context: rapid convergence due to hybridization being a small parameter, and an efficient real-axis implementation enabled by closed-form Feynman rules. By implementing a bold solver on both axes, the authors demonstrate that finite-temperature spectra can be computed accurately within DMFT without the need for numerical analytic continuation.

The method provides reliable spectroscopy and benchmark transport results. The authors note that while the current implementation is robust, it opens the door for several extensions:

Combining DFT and DMFT for realistic materials applications.
Addressing regimes where CT-HYB suffers from severe fermionic sign problems (e.g., the $p$ - $d$ model of cuprates).
Solving steady-state non-equilibrium problems on the real axis.

The work establishes BHQMC as a precise tool for computing frequency-dependent scattering rates and transport properties, serving as a benchmark for future quantum impurity solver developments.

More like this