🔬 condensed matter

Solvable coupled cluster states in $d=\infty$ with prescribed momentum distribution

This paper establishes a connection between coupled cluster theory and dynamical mean-field theory in the infinite-dimensional limit, demonstrating that a specific family of coupled cluster states allows for the exact closed-form evaluation of static observables and the realization of any momentum distribution as a unique ground state, thereby isolating evaluation errors from expressivity limitations and revealing fundamental constraints of coupled cluster doubles in the strong-coupling Mott regime.

Original authors: Izak Snyman, Serge Florens

Published 2026-11-02

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Izak Snyman, Serge Florens

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 electrons moving through solids, scientists face a persistent dilemma: how to describe a crowd of particles that are constantly pushing and pulling on one another. When electrons interact strongly, they stop behaving like independent individuals and start acting as a complex, collective whole. To understand materials like high-temperature superconductors or the insulating states of certain metals, physicists need to build mathematical models that capture this collective behavior. For decades, two powerful but separate schools of thought have dominated this field. One approach, rooted in quantum chemistry, uses a method called coupled cluster theory to build highly accurate descriptions of electron clouds, but it often struggles to calculate the results for large systems without introducing errors. The other, known as dynamical mean-field theory, excels at handling strong interactions in solids by simplifying the problem to a single point in space, but it traditionally relies on approximations that can obscure the fine details of how electrons move. Bridging these two worlds has been a long-standing challenge, as researchers have struggled to find a way to calculate the properties of these complex electron states exactly, without losing the ability to describe the strong interactions that define them.

A team of researchers has now found a way to merge these approaches by taking a bold step into a theoretical extreme: they imagined a crystal lattice with an infinite number of dimensions. In this abstract setting, the usual complications of electron movement simplify in a surprising way. The researchers demonstrated that for a specific family of electron states, they could calculate the exact behavior of the system without any approximations or truncations. They focused on a class of states known as coupled cluster states, which are built by adding layers of complexity to a simple starting point, much like adding intricate details to a basic sketch. By restricting their study to a lattice where the number of dimensions is infinite, they showed that the influence of the entire rest of the crystal on any single point could be reduced to a simple, static mathematical kernel. This allowed them to solve the problem for a single site exactly, treating the rest of the universe as a static background rather than a dynamic, time-evolving field.

The most striking result of this work is the ability to freely design the momentum distribution of the electrons. In a metal, electrons occupy different energy levels, and the way they fill these levels determines how the material conducts electricity. The researchers showed that they could construct a state where the electrons fill these levels in any pattern they chose, provided the pattern was symmetric. They could create a state where the transition between filled and empty levels was sharp, like a cliff, or completely smooth, like a gentle slope. This flexibility is significant because many traditional models force the transition to be a sharp step, which fails to capture the subtle behavior of real, interacting electrons. The team proved that for any smooth distribution they desired, there exists a specific, explicit set of rules—a parent Hamiltonian—that would make that exact state the lowest energy, or ground, state of the system. This means they could not only describe a state but also write down the physical laws that would create it.

When the researchers applied their method to a standard model of interacting electrons, known as the Hubbard model, they found that their approach worked exceptionally well when the interactions between electrons were weak. In this regime, their new states recovered nearly all of the energy contributions predicted by established perturbation theory, far outperforming older, simpler models. However, the story changed when they turned up the strength of the interactions to simulate a Mott transition, a phenomenon where a material suddenly stops conducting electricity and becomes an insulator. Here, the method hit a hard wall. No matter how they adjusted the parameters, the model could not suppress the double occupancy of electrons on a single site below a certain limit. The electrons could not be forced to avoid each other as strongly as they do in a true insulator.

This failure is not a flaw in the calculation method, which was exact, but a fundamental limitation of the type of state they were using. The researchers concluded that while their coupled cluster states are excellent at describing the fluid-like behavior of electrons in a metal, they lack the specific structural property needed to fully project out the double-occupied sites required for a Mott insulator. The mathematical form of their state mimics the behavior of a metal well but cannot fully replicate the rigid exclusion of an insulator. By isolating this limitation in a setting where calculation errors were removed, the study provides a clear diagnostic tool: it shows that the difficulty in describing strong correlations is not just a matter of better math, but a structural constraint of the wave function itself. The work opens a new path for understanding exactly where these powerful theoretical tools succeed and where they fundamentally fall short, offering a precise laboratory for testing the limits of our ability to describe the quantum world.

Technical Summary: Solvable Coupled Cluster States in $d = \infty$ with Prescribed Momentum Distribution

Problem Statement
The paper addresses a fundamental tension in many-body theory: the trade-off between the flexibility of constructing correlated trial states and the ability to calculate observables exactly. While methods like the Gutzwiller ansatz allow for exact evaluation in the infinite-dimensional ( $d \to \infty$ ) limit, they often fail to capture the detailed single-particle momentum distribution ( $n_k$ ) of a correlated Fermi liquid, yielding a piece-wise constant profile. Conversely, standard Coupled Cluster (CC) theory is highly expressive and size-consistent but relies on projection methods that introduce evaluation errors (truncation of the hierarchy), making it difficult to disentangle the limitations of the ansatz from the errors of the approximation scheme. The authors ask whether it is possible to construct a class of coupled cluster states that can be evaluated exactly in the thermodynamic limit while retaining the ability to freely prescribe the single-particle momentum distribution.

Methodology
The authors establish a connection between the exponential coupled cluster ansatz and Dynamical Mean-Field Theory (DMFT) by exploiting the simplifications of the hypercubic lattice in the limit $d \to \infty$ .

Envelope Fermions and Low-Rank Decomposition: The trial state is defined as $|\Psi\rangle = \exp(-\sum_j T^\dagger_j)|F\rangle$ , where $|F\rangle$ is the free Fermi sea. The cluster operator $T^\dagger_j$ is constructed using "envelope fermions" ( $\psi^\dagger_{m\sigma\alpha\rho}$ ), which are linear combinations of momentum orbitals weighted by energy-dependent envelope functions $F_\alpha(\varepsilon_k)$ . This corresponds to a low-rank Tucker decomposition of the cluster amplitude tensor.
Static Cavity Construction: Borrowing the cavity argument from DMFT, the authors show that in $d \to \infty$ , the environment of a lattice site reduces to a static, quadratic Gaussian kernel (analogous to the Weiss field). Unlike DMFT, which involves path integrals over imaginary time, the "impurity problem" here reduces to a finite-dimensional integral over Grassmann variables associated with a single site.
Exact Evaluation: The expectation values of local operators are computed exactly by performing these Grassmann integrals. The self-consistency condition relates the interacting local Green's function to the cavity field and a site-diagonal "self-energy" (irreducible corrections), allowing for the exact determination of static observables without truncation errors.

Key Contributions and Results

Exact Closed-Form Evaluation: The paper demonstrates that a large class of coupled cluster states (specifically those with low-rank amplitude decomposition) can be evaluated exactly in the $d \to \infty$ limit. This separates the expressivity of the state from the errors of the evaluation method.
Arbitrary Momentum Distribution: For the simplest case (single-flavor doubles), the authors derive a closed-form relationship between the energy envelope $F(\varepsilon)$ and the momentum distribution $n(\varepsilon)$ . They show that any particle-hole symmetric momentum distribution can be realized. This includes distributions with a jump at the Fermi energy (finite quasiparticle weight $Z$ ) and completely smooth distributions ( $Z=0$ ).
Parent Hamiltonian: Using the "gossamer metal" construction, the authors prove that any such trial state is the unique, gapless ground state of an explicit parent Hamiltonian. This Hamiltonian features correlated hopping terms and a gapless single-particle tunnelling spectrum.
Application to the Hubbard Model:
- Weak Coupling: When applied to the half-filled Hubbard model, the single-flavor ansatz recovers 98.6% of the exact second-order correlation energy and 66% of the second-order correction to the quasiparticle weight. This significantly outperforms the Gutzwiller state (84% and 22%, respectively), demonstrating that controlling the momentum distribution improves the description of the Fermi liquid phase.
- Strong Coupling: In the strong coupling regime (near the Mott transition), the single-flavor ansatz fails to suppress double occupancy below $\approx 0.1283$ , whereas the Gutzwiller state can approach zero. The authors show that extending the ansatz to multiple flavors or including triples and quadruples (within the constraints of the two-flavor, flat-envelope limit explored) does not lower this bound.
Diagnosis of Coupled Cluster Limitations: The failure to capture the Mott transition (complete suppression of local charge fluctuations) is attributed to the structural nature of the coupled cluster exponential. Unlike the Gutzwiller factor, which is linear in the idempotent double-occupancy operator and can project out doubly occupied sites, the exponential of particle-hole creation operators lacks this projecting property. This provides an exact diagnosis of the limitations of coupled cluster doubles in strong correlation regimes, free from evaluation approximations.

Significance
The paper claims that its primary significance lies in providing a "laboratory" where the quality of a coupled cluster ansatz can be cleanly separated from the quality of the evaluation scheme. By establishing an exact evaluation framework in $d \to \infty$ , the authors demonstrate that the inability of coupled cluster doubles to describe the Mott transition is a fundamental limitation of the state's expressivity, not merely an artifact of truncated projection equations. Furthermore, the work offers a new variational tool that allows for the direct control of single-particle momentum distributions, bridging the gap between quantum chemistry methods and condensed matter embedding theories.

More like this