⚛️ quantum physics

Improving Perturbation Theory with the Sum-of-squares II: Large Density-Density Terms

This paper proposes a method to extend the applicability of sum-of-squares perturbation theory for fermionic Hamiltonians by addressing its previous limitations in handling strong density-density interaction terms common in chemistry.

Original authors: Matthew B. Hastings

Published 2026-07-01

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Matthew B. Hastings

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

Imagine you are trying to find the lowest possible energy state of a complex quantum system, like a molecule. Think of this system as a giant, chaotic ballroom dance where thousands of fermions (the dancers) are interacting. Your goal is to prove a "floor" for how low the music can get (the ground state energy) without actually solving the impossible math of every single dancer's move.

For a long time, scientists have used a method called Sum-of-Squares (SOS) to build this floor. It's like trying to prove the dance floor is solid by showing that every possible wobble can be written as a perfect square (which is always positive).

The Previous Problem: The "Weak" Dance

In a previous paper (Ref [1]), the author showed a clever trick to build this floor faster and more accurately. They used a method called perturbation theory—basically, assuming the dancers are mostly dancing to their own rhythm, with only tiny, gentle nudges from each other.

Under these "gentle" conditions, the method worked beautifully. It was faster and more accurate than the old standard (called 2RDM). However, there was a catch: Real chemistry isn't gentle.

In real molecules, some dancers (atoms) push against each other with massive force. These are called density-density interactions. Imagine two dancers suddenly grabbing each other and refusing to let go, or pushing each other apart with the force of a sledgehammer. The old "gentle nudge" math broke down completely when faced with these strong pushes. The method couldn't handle the "strong terms" of the Hamiltonian (the energy equation).

The New Solution: Rewriting the Rules

This paper proposes a new way to handle those "sledgehammer" pushes. Instead of trying to treat the strong pushes as tiny nudges, the author changes the rules of the game to include the strong pushes right into the foundation of the dance floor.

Here is the simple breakdown of the new method:

1. The "Base" Dance Floor (H₀)
Previously, the "base" dance floor only included simple, independent movements. Now, the author says: "Let's build the base floor with the strong pushes included."

Analogy: Instead of assuming the floor is flat and the dancers just bump into each other lightly, we build the floor with specific ramps and bumps that match the strong pushes exactly.

2. The "Correction" Dancers (The τ terms)
The old method used a set of "correction dancers" (called $\tau$ ) to fix small errors. These corrections were calculated based on the idea that interactions were weak.

The Fix: The author recalculates these correction dancers. They change the "energy denominators" (the math that decides how much a correction matters).
Analogy: Imagine you are trying to balance a scale. If you add a heavy rock (the strong interaction), you can't just add a tiny pebble to balance it. You have to recalculate the entire counterweight system. The author updates the formula for these counterweights so they work even when the "rocks" are huge.

3. The "Closure" Trick
A major hurdle in these math problems is "closure." This means that when you do the math, you don't want to accidentally create new, weird types of interactions that you didn't start with (like creating a 12th-degree polynomial when you only wanted a 4th-degree one).

The Innovation: The author introduces a new type of term in the math equation that involves a "number operator" (counting how many dancers are in a spot) multiplied by other terms.
Analogy: Think of this as adding a specific "safety net" to the dance floor. If the strong pushes try to create a weird, unstable wobble, the safety net catches it and turns it back into a standard, manageable wobble. This ensures the math stays clean and solvable.

What This Achieves

The paper claims that with this new method:

It handles the "Sledgehammers": It can now accurately calculate energy bounds for systems with strong density-density interactions (the big pushes).
It keeps the speed: It doesn't require solving a massive, slow computer problem (a semi-definite program) for every step. It remains fast, similar to the old method.
It stays accurate: It still guarantees that the calculated energy is a true "lower bound" (the floor is real) and is very close to the true answer, even when the interactions are strong.

The Limitations

The author is honest about what this method doesn't do yet:

It handles "density-dependent hopping" (dancers moving only when others are present) but only in specific forms.
It doesn't yet have a special trick for "singlet hopping" (a specific type of paired dancing), though the author notes the old method actually worked okay for that one already.

The Bottom Line

This paper is like upgrading a bridge. The old bridge was great for light traffic (weak interactions) but would collapse under heavy trucks (strong interactions). The author has redesigned the bridge's supports to handle the heavy trucks directly, without making the bridge slower or more expensive to cross. This allows scientists to use this fast, accurate math method on real-world chemical problems that were previously too difficult to solve.

Technical Summary: Improving Perturbation Theory with the Sum-of-squares II: Large Density-Density Terms

Problem Statement
The paper addresses a significant limitation in applying self-consistent sum-of-squares (SOS) decompositions to quantum chemistry Hamiltonians. Previous work (Ref. [1]) demonstrated that a degree-six SOS method, utilizing a perturbative choice of cubic operators ( $\tau_i$ ), could outperform the standard degree-four SOS (2RDM) in both speed and accuracy for model Hamiltonians with weak, random quartic interactions. However, this method relies on the assumption that interaction terms are small (order $\epsilon$ ). Real-world quantum chemistry Hamiltonians often contain strong "density-density" interaction terms ( $U_{ij} n_i n_j$ ), as well as density-dependent hopping and spin-spin interactions. When these terms are large (order unity), the perturbative choice of cubic operators fails to accurately treat them, rendering the previous method ineffective for these applications. The core challenge is to extend the self-consistent SOS framework to handle these strong interaction terms while maintaining the method's computational efficiency and its ability to provide strict lower bounds on the ground state energy.

Methodology
The authors propose a modification to the definition of the cubic operators ( $\tau_i$ ) and the structure of the SOS decomposition to accommodate strong density-density terms.

Modified Hamiltonian Structure: The Hamiltonian is redefined as $H = H_0 + \epsilon V$ , where $H_0$ now explicitly includes the quadratic part ( $H_{quad}$ ) and the strong density-density interaction terms ( $\sum U_{ij} n_i n_j$ ). The term $\epsilon V$ contains the remaining weaker quartic interactions.
Renormalized Energy Denominators: In the previous perturbative approach, the coefficients of the cubic operators were determined by energy denominators of the form $1/(e_i + e_j + e_k + e_l)$ . To handle strong density-density terms, the authors redefine these denominators to include the interaction energy. Specifically, the denominator becomes the expectation value of $H_0$ in a state where modes $i, j, k, l$ are occupied and others are empty:
$e_{i,j,k,l} = e_i + e_j + e_k + e_l + \sum_{\text{pairs}} U_{pair}$
This ensures that the operator $\psi_i + \tau_i$ still annihilates the ground state up to an error of $O(\epsilon^2)$ , preserving the "error property" required for the method to yield an $O(\epsilon^4)$ accurate energy bound.
Handling Closure and Density-Dependent Terms: To maintain the "closure property" (ensuring the effective Hamiltonian remains within the same class of quartic Hamiltonians) and to handle density-dependent hopping terms ( $n_i \psi^\dagger_j \psi_k$ ), the authors introduce a more general decomposition structure. They expand the trial Hamiltonian to include terms of the form $\sum \epsilon n_i V_i$ .
The new decomposition takes the form $H = \sum_a c_a (|F_a|^2 + |\theta_a^\dagger|^2)$ , where:
$F_a = \psi_i + n_j \sum_{k \neq j} d_k \psi_k + \theta_a$
$\theta_a = \tau_i + \mu + \epsilon n_j \nu$
Here, $\mu$ and $\nu$ are additional cubic operators chosen to correct the error property and generate the necessary density-dependent terms, respectively. This structure allows the method to reproduce the target Hamiltonian $H$ via an iterative self-consistent procedure.

Key Contributions

Extension to Strong Interactions: The paper provides a method to generate SOS decompositions for Hamiltonians containing strong density-density interaction terms, which were previously a barrier to applying the self-consistent SOS method to quantum chemistry.
Generalized Operator Definitions: The authors introduce a modified definition for the cubic operators $\tau_i$ using renormalized energy denominators that account for interaction strengths, ensuring the "error property" holds even when interactions are not perturbatively small.
Closure Property Preservation: By introducing auxiliary terms involving number operators ( $n_i$ ) and specific cubic corrections ( $\mu, \nu$ ), the authors demonstrate how to maintain the closure property, ensuring the effective Hamiltonian remains a quartic operator suitable for the iterative solution.
Network Flow Formulation: The paper characterizes the range of Hamiltonians that can be decomposed using this method. It reduces the problem of constructing a valid decomposition for a Hamiltonian with negative interaction terms ( $U_{ij} < 0$ ) to a network flow problem. This provides a concrete criterion for determining if a given set of interaction parameters can be handled by the proposed decomposition.

Results and Claims
The paper claims that the proposed method:

Maintains Accuracy: It yields a lower bound on the ground state energy that is accurate to $O(\epsilon^4)$ relative to the true ground state energy, even when density-density terms are of order unity.
Preserves Efficiency: The computational effort required is not significantly higher than when all terms are small ( $O(\epsilon)$ ), as the method avoids solving full semi-definite programs (SDPs) of high degree, relying instead on matrix diagonalization and iterative updates.
Handles Specific Terms: The method successfully handles density-density interactions and, as a side effect, density-dependent hopping terms (of the form $n_i(\psi^\dagger_j \psi_k + h.c.)$ ) and spin-spin interactions (which can be mapped to sums of density-density terms).
Limitations: The authors note that the method currently handles density-dependent hopping only in a specific form (after particle-hole conjugation). Terms involving pair creation ( $n_i \psi^\dagger_j \psi^\dagger_k$ ) are not included, though the authors suggest these are typically smaller and can be treated as perturbative. Additionally, singlet hopping terms are not addressed by this specific extension, as they do not share the property of annihilating the quadratic ground state.

Significance
The significance of this work lies in bridging the gap between the theoretical success of self-consistent SOS methods on random model Hamiltonians and their practical application to real quantum chemistry problems. By overcoming the barrier of strong density-density interactions, the paper extends the range of applicability of a method that offers strict lower bounds on energy without the computational cost of full SDP solvers. The authors position this work as a necessary sequel to previous studies, enabling the treatment of physically relevant Hamiltonians that were previously out of reach for this specific non-perturbative, perturbatively-guided approach.