⚛️ quantum physics

Quantum algorithm for one-quasiparticle excitations in the thermodynamic limit via cluster-additive block diagonalization

This paper proposes a hybrid quantum-classical algorithm that combines the Variational Quantum Eigensolver (VQE) with Numerical Linked-Cluster Expansions (NLCEs) and a novel postprocessing technique called Projective Cluster-Additive Transformation (PCAT) to accurately compute one-quasiparticle excitation energies in the thermodynamic limit, even in systems where parity symmetry is broken.

Original authors: Sumeet, M. Hörmann, K. P. Schmidt

Published 2026-07-15

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: Sumeet, M. Hörmann, K. P. Schmidt

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 understand the music of a giant, invisible orchestra playing inside a block of quantum material. You want to hear the solo notes—the "quasiparticles"—that pop up when you poke the system. The problem is, the orchestra is infinite (the "thermodynamic limit"), and you only have a tiny, noisy instrument (a current quantum computer) that can't play the whole symphony at once.

This paper proposes a clever new way to listen to those solo notes without needing a supercomputer that doesn't exist yet. The authors, Sumeet, M. H¨ormann, and K. P. Schmidt, have built a hybrid recipe that mixes the best of classical math with the power of today's quantum machines.

The Big Idea: The "Lego" Strategy

Think of the infinite material as a massive wall made of Lego bricks. You can't build the whole wall on your desk to study it, so you build small, separate Lego clusters instead. This is called a Numerical Linked-Cluster Expansion (NLCE). The trick is that if you build these small clusters correctly, you can mathematically "glue" them together to predict what the infinite wall sounds like.

For years, scientists have used this Lego method for the ground state (the quietest note), but trying to hear the excited notes (the quasiparticles) was like trying to tune a radio while someone is shouting static. The math gets messy because the "excited" notes can get tangled up with each other in a way that breaks the Lego glue.

The Quantum Solution: VQE as a Tuning Fork

The authors use a tool called the Variational Quantum Eigensolver (VQE). Imagine the VQE as a super-smart, adjustable tuning fork. You give it a small Lego cluster, and it tries to twist and turn the quantum state until it finds the perfect shape that separates the "quiet" notes from the "loud" notes.

However, there's a catch. If you just let the VQE do its thing, the resulting shape might look good for that one small cluster, but when you try to glue it to the next cluster, the math falls apart. It's like trying to connect two Lego bricks that have slightly different shapes; the wall becomes wobbly and the prediction fails.

The Secret Sauce: PCAT (The "Glue" Fix)

This is where the paper's main innovation shines. They introduce a post-processing step called PCAT (Projective Cluster-Additive Transformation).

Think of the VQE as a sculptor who carves a beautiful statue from a block of stone. The statue looks great on its own, but if you try to stack it with other statues, they don't fit perfectly. PCAT is like a magical chisel that goes in after the sculptor is done. It trims the statue just enough so that it fits perfectly with any other statue, no matter how far away it is in the infinite wall.

The paper explicitly argues that you cannot skip this step. Without PCAT, the math allows "ghost" connections where a particle on one Lego cluster magically hops to a disconnected cluster far away. This is physically impossible (like a ghost walking through a wall), and it ruins the calculation. The authors show that if you don't use PCAT, your results oscillate wildly and never settle down, no matter how many clusters you add.

What They Found (The Results)

The team tested their method on a famous model called the Transverse-Field Ising Model (TFIM), which is like a row of tiny magnets that can point up or down.

The Easy Mode (Pure TFIM): When the magnets are in a simple setup, the VQE tuning fork works beautifully. They found that using a specific number of layers in their quantum circuit (roughly half the number of spins, written as $\lceil N/2 \rceil$ ) was enough to get results that matched the exact, perfect math solutions. It was as if the quantum computer learned the secret code of the system instantly.
The Hard Mode (TFIM with a Longitudinal Field): Then they added a "longitudinal field" (a magnetic push from the side). This broke the symmetry of the system, making it much messier. Suddenly, the easy $\lceil N/2 \rceil$ $⌈ N /2 ⌉$ layers weren't quite enough. They had to use $N$ layers (twice as many) to get the same level of accuracy.
- The Trap: They also discovered a hidden trap in the math. If they started the quantum computer with a "warm start" (using the solution from the quiet state as a guess for the noisy state), the computer got stuck in a local minimum—a dead end where it thought it was done, but it was actually wrong.
- The Fix: They found that starting with a "cold start" (random guesses near zero) or using a different math formula (called a "trace-based" cost function) helped the computer escape the trap and find the right answer.

How Sure Are They?

The authors are very confident in their method for the specific models they simulated. They ran these experiments on powerful classical computers simulating quantum circuits (statevector simulations), not on actual physical quantum hardware yet.

Proven in Simulation: For the pure 1D and 2D models, their method matched the exact solutions perfectly as they added more Lego clusters.
Suggested for the Future: They suggest that this method could be a game-changer for "frustrated" magnets (systems where magnets can't agree on a direction), which are currently impossible for classical computers to solve. However, they admit that for these harder systems, the optimization might get even trickier.
Not Yet on Real Hardware: The paper notes that they are currently working on running this on real quantum chips to see how noise affects the "glue" (PCAT).

The Takeaway

This paper doesn't claim to have solved all of quantum physics. Instead, it offers a new, robust toolkit. It says: "If you want to hear the solo notes of an infinite quantum orchestra, don't just tune your instrument; you must also use this special 'glue' (PCAT) to make sure the pieces fit together."

They showed that with the right glue, even a noisy, small quantum computer can help us predict the behavior of infinite materials, provided we are careful about how we start the calculation and how many layers of "tuning" we apply. It's a promising step toward using quantum computers to solve problems that are currently impossible, but it's a step that requires patience and the right mathematical tools.

Technical Summary: Quantum Algorithm for One-Quasiparticle Excitations in the Thermodynamic Limit via Cluster-Additive Block Diagonalization

Problem Statement
The accurate computation of elementary excitations (quasiparticles) in quantum many-body systems within the thermodynamic limit remains a significant challenge. While Variational Quantum Eigensolvers (VQE) have shown promise for ground-state properties on noisy intermediate-scale quantum (NISQ) devices, extending these methods to excited states is non-trivial. Specifically, constructing effective Hamiltonians for excited states requires cluster additivity—a property where observables on disconnected clusters factorize, ensuring convergence when extrapolating to the thermodynamic limit via Numerical Linked-Cluster Expansions (NLCE). Standard VQE block-diagonalization unitaries do not automatically satisfy cluster additivity, particularly for degenerate subspaces or when symmetries (like parity) are broken. Consequently, a hybrid quantum-classical framework capable of generating cluster-additive effective Hamiltonians for excited states has been lacking.

Methodology
The authors propose a hybrid quantum-classical algorithm that integrates VQE with NLCE and a post-processing technique called Projective Cluster-Additive Transformation (PCAT). The workflow proceeds in six steps:

Model Definition: A gapped quantum lattice model (e.g., Transverse-Field Ising Model, TFIM) is defined.
Cluster Decomposition: The lattice is decomposed into topologically distinct, finite connected clusters using a rectangular graph expansion.
VQE Cluster Solving: For each finite cluster, a VQE is employed to find a unitary transformation ( $U_{VQE}$ $U_{V QE}$ ) that block-diagonalizes the cluster Hamiltonian. This decouples the ground state and the one-quasiparticle (1QP) subspace from higher-energy excitations. The authors utilize the Hamiltonian Variational Ansatz (HVA), constructed from the problem Hamiltonian's terms, to ensure expressibility. Two classes of cost functions are tested:
- Variance-based: Minimizes the energy variance of the 1QP subspace (and optionally the ground state).
- Trace-based: Minimizes the trace of the Hamiltonian over the combined ground-state and 1QP subspace.
PCAT Post-processing: The raw VQE unitary is generally not cluster additive. The authors apply PCAT to construct a corrected unitary ( $U_{PCAT}$ ). PCAT modifies the VQE-prepared states by subtracting projections onto lower-energy subspaces of the unperturbed Hamiltonian. This ensures that the resulting effective Hamiltonian satisfies the cluster-additivity condition required for NLCE convergence. Crucially, PCAT requires only low-energy eigenspace information (energy expectations and state overlaps) obtainable via $O(N^2)$ measurements, avoiding exponential state tomography.
NLCE Embedding: The reduced contributions of the effective 1QP Hamiltonian are computed using the inclusion-exclusion principle and embedded into the infinite lattice.
Dispersion Calculation: The effective 1QP Hamiltonian in the thermodynamic limit is diagonalized via Fourier transform to yield the quasiparticle dispersion $\omega(\vec{k})$ .

Key Contributions

Algorithmic Framework: The paper introduces the first quantum algorithm capable of constructing cluster-additive effective Hamiltonians for excited states, enabling thermodynamic-limit calculations via NLCE.
PCAT Integration: It demonstrates that PCAT can be successfully applied to VQE outputs. This bridges the gap between variational optimization (which finds low-energy subspaces) and the rigorous requirements of linked-cluster expansions (which require specific factorization properties).
Cost Function Analysis: The work systematically compares variance-based and trace-based cost functions within the HVA framework, analyzing their convergence properties and sensitivity to initialization strategies.
Benchmarking: The method is rigorously benchmarked against exact diagonalization (ED) and known analytical solutions for the TFIM.

Results
The method was tested on the TFIM in one dimension (1D), two dimensions (2D), and with a longitudinal field (TFIM+LF):

Pure TFIM (1D and 2D): For the pure TFIM, where parity symmetry is preserved, the method achieves excellent agreement with exact diagonalization. The authors find that $\lceil N/2 \rceil$ layers of the HVA are sufficient to converge the results to the exact solution for cluster sizes up to $N_{max}=16$ . Both variance and trace cost functions yield identical convergence.
TFIM with Longitudinal Field (TFIM+LF): In this case, parity symmetry is broken, coupling the ground state to excitations and rendering the model non-integrable. Here, PCAT is essential; without it, the 2b (Schrieffer-Wolff type) transformation fails to produce convergent NLCE results due to unphysical hopping between disconnected clusters.
- Layer Depth: While $\lceil N/2 \rceil$ layers converge, using $N$ layers provides improved accuracy for intermediate cluster sizes.
- Initialization Sensitivity: The study reveals a critical difference in optimization landscapes. Variance-based cost functions are prone to getting trapped in local minima when initialized with ground-state parameters, particularly for larger clusters ( $N \ge 10$ ). In contrast, trace-based cost functions remain robust to ground-state initialization. Near-zero initialization resolves the convergence issues for variance-based approaches but requires more iterations.
- Convergence: Despite the increased optimization difficulty, the NLCE+VQE approach successfully converges to the NLCE+ED results as cluster size increases, validating the method for non-integrable, symmetry-broken systems.

Significance and Claims
The authors claim that this work establishes PCAT as a robust framework for extending variational quantum algorithms to excited-state calculations in the thermodynamic limit. The significance lies in:

Overcoming the Cluster Additivity Barrier: The paper demonstrates that VQE, often criticized for lacking the structural properties needed for NLCE, can be made compatible through PCAT post-processing.
Algorithm Agnosticism: While demonstrated with VQE, the PCAT post-processing step relies only on low-energy eigenspace information (overlaps and energies). The authors state this makes the approach applicable to any quantum eigenstate preparation method, including future fault-tolerant protocols like Quantum Phase Estimation or Adiabatic State Preparation.
Scalability: By using VQE as a cluster solver, the method aims to access cluster sizes that are classically intractable for exact diagonalization, potentially offering a pathway to study frustrated systems where classical Monte Carlo methods fail due to sign problems.

The paper concludes that while the method is currently limited by optimization challenges in symmetry-broken regimes (requiring careful initialization and potentially deeper circuits), it provides a viable and necessary pathway for computing excitation spectra in the thermodynamic limit using near-term and future quantum hardware.