One-Particle Resonances and Eigenstate Thermalization in Periodic Free-Fermion Lattices
Listen
Radio episode about this paper
Transcript
Introduction to the show: ident: Mathematical Physics. The day in mathematical physics: rigorous models of physical systems.
Kai: We're looking at the paper: "One-Particle Resonances and Eigenstate Thermalization in Periodic Free-Fermion Lattices".
Mira: Here's the gist of the paper. We study eigenstate fluctuations of the particle number in the nearest-neighbour free-fermion model on the periodic cubic lattice (Z/LZ) d and prove a bound of order N Ld−1 for the particle-number variance <ref:2611.12701#pg5>.
Kai: The main results establish a concentration bound for the particle number variance in every normalized energy eigenstate <ref:2611.12701#pg4>.
Mira: Specifically, for every fixed dimension d ≥ 1, there exists a constant Cd such that every normalized eigenstate ϕ of HfF0 in the N-particle sector HN satisfies Dϕ,¡NΓ − Nµ¢2ϕE≤ CdNµ(1−µ)Ld−1 <ref:2611.12701#pg4>.
Kai: This leads to a corresponding eigenstate-thermalization estimate of order Ld−1/(Nη 2) for fluctuations of relative size η.
Paper summary: Kai: So to wrap up this paper "One-Particle Resonances and Eigenstate Thermalization in Periodic Free-Fermion Lattices," the authors have established a concentration bound for the particle number variance that is order N L to the minus one <ref:2611.12701#pg4>.
Mira: They also proved a corresponding eigenstate-thermalization estimate of order L to the minus one over N times epsilon squared <ref:2611.12701#pg4>, which is pretty powerful because it links the fluctuations directly to how quickly things thermalize when you look at a region of size epsilon <ref:2611.12701#pg4>.
Lev: In simple terms, what this means for the world is that in these free-fermion models, the particle number isn't wildly fluctuating randomly across all energy states; it concentrates around a value that decays as you increase the system size linearly <ref:2611.12701#pg4>.
Kai: It’s not just about the numbers on the page; it’s about understanding how these quantum systems behave when they are in their true energy eigenstates <ref:2611.12701#pg4>.
Mira: And the title itself, "One-Particle Resonances and Eigenstate Thermalization," suggests that understanding these specific resonance structures is key to proving this thermalization behavior <ref:2611.12701#pg4>.
Lev: It gives us a concrete target: if we want to model real quantum hardware where we’re trying to maintain particle numbers or energy states, this result tells us the expected error scaling for those fluctuations <ref:2611.12701#pg4>.
Conclusion: Kai: So we've seen how these free fermion systems have particle number fluctuations that scale in a specific way <ref:2611.12701#pg4>.
Mira: It really boils down to this idea of resonance structures controlling thermalization, which is what the title points to.
Lev: From an error correction standpoint, if you're building hardware on these lattices, knowing that this variance scales with N times the inverse of L cubed gives us a pretty clear overhead estimate for the noise we have to deal with.
Kai: I mean, the authors are saying that this concentration bound holds in every dimension <ref:2611.12701#pg4>, which is pretty strong because it means there isn't some weird, high-dimensional fluke where things get totally chaotic in a specific energy state.
Mira: But the real power here is that they show how to get this bound even with those boundary conditions we talked about earlier <ref:2611.12701#pg4>, meaning it's robust across different setups.
Lev: For me, the part about the one-dimensional case being sharper, showing a logarithm correction instead of a constant one, that's what matters when you try to design an actual physical system where those small details matter.
Kai: So looking at the authors and this result, it seems they’ve given us a solid mathematical foundation for how we can expect particle number noise to behave in these types of quantum systems.
Mira: It sets a baseline for what thermalization looks like in these specific free fermion models, which is a key assumption for many other condensed matter theories.
Lev: It means when we move on to more complex Hamiltonians, we have a starting point for how much noise we can tolerate before the whole thing breaks down.
T. Benoist, V. Jakšc, C.-A. Pillet, A. Szczepanek, C. Vogel
University of Toulouse, INSA Toulouse, CNRS, IMT, Toulouse, France · Department of Mathematics at Politecnico di Milano · Université de Toulon, CNRS, CPT, UMR 7332 La Garde Marseille Aix-Marseille University CNRS CPT UMR 7332 Case 907 Marseille France · Institute of Mathematics at Jagiellonian University Kraków Poland · Mathematics Institute at Ludwig Maximilian University Munich Germany
math-ph, cond-mat.stat-mech, math.MP, quant-ph
Submitted: 2026-11-09
Updated: 2026-11-09
Comments: 29 pages LaTeX, no figures
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Importance score: 93/100
The gist: We study eigenstate fluctuations of the particle number in the nearest-neighbour free-fermion model on the periodic cubic lattice (Z/LZ) d and prove a bound of order N Ld−1 for the particle-number…
Key concepts
- Particle Number Variance
- This measures how much the actual number of particles in a specific state fluctuates around its average value. The paper proves this fluctuation is bounded by a term proportional to $N L_d^{-1}$, which is crucial for understanding how stable or fluctuating these quantum states are.
- Eigenstate Thermalization
- This concept suggests that complex quantum systems, when in a specific energy state, behave statistically like a thermal system. The paper uses the variance bound to show that particle number fluctuations decay at a rate consistent with this thermal behavior, linking microscopic structure to macroscopic statistical properties.
- Resonance Multiplicity
- This refers to the number of distinct non-zero energy gaps (resonances) in the system. The proof relies on sharp estimates showing that these resonances are bounded by $L_d^{-1}$ and that there exist gaps larger than $L_d^{-1}$, which dictates the structure of the Hilbert space.
- Concentration Bound
- This is a mathematical statement asserting that for any given energy eigenstate, the particle number is highly concentrated around its expected value. The paper proves this concentration holds with a specific scaling factor ($1.2$) and shows it cannot be improved uniformly in the tolerance parameter.
Terminology
Summary
We study eigenstate fluctuations of the particle number in the nearest-neighbour free-fermion model on the periodic cubic lattice (Z/LZ) d and prove a bound of order N Ld−1 for the particle-number variance <ref:2611. p. 012701#pg5>.
Main Results Summary
The main results establish a concentration bound for the particle number variance in every normalized energy eigenstate <ref:2611.12701#pg4>. Specifically, for every fixed dimension d ≥ 1, there exists a constant Cd such that every normalized eigenstate ϕ of HfF0 in the N-particle sector HN satisfies Dϕ,¡NΓ − Nµ¢2ϕE≤ CdNµ(1−µ)Ld−1 <ref:2611. p. 212701#pg4>. This leads to a corresponding eigenstate-thermalization estimate of order Ld−1/(Nη 2) for fluctuations of relative size η. p. 0
Decomposition Strategy
The proof decomposes the centered particle-number operator according to one-particle energy transfer and treats the zero- and nonzero-transfer components separately. The nonzero-transfer part is controlled by a sharp resonance bound proved from Mann’s theorem on relations among roots of unity. p. 0 Meanwhile, the zero-transfer part is estimated by a Hilbert–Schmidt argument. p. 0
Optimality and Bounds
The bounds derived are optimal in every dimension; the maximal nonzerogap multiplicity has order Ld−1, and suitable many-body eigenstates realize matching variance lower bounds and show that the concentration bound cannot be improved in estimates uniform in the tolerance parameter. The scale in the concentration estimate (1. p. 32) is optimal, as for every d ≥ 1 and arbitrarily large L, we find a particle number N, a normalized energy eigenstate ϕ ∈ HN, and a subset Γ ⊂ ΛL,d such that ϕ,(NΓ − Nµ)2ϕE®≥ cdN Ld−1. p. 2
Extensions to Boundary Conditions
The paper provides extensions to Dirichlet boundary conditions, showing that for every fixed d ≥ 1 there exists a constant Cd such that for every side length L, every particle number 0 0, the bound ϕ,(NΓ − Nµ)2ϕE®≤ CdN Ld−1 holds.
One-Dimensional Refinement
In one dimension (d = 1), a sharper bound on the deviation of the expected particle number from its equilibrium value is obtained, showing that for every normalized eigenstate ϕ ∈ HN satisfies ¯¯〈ϕ,NΓϕ〉 − Nµ¯¯≤ C(1+logN). This result is derived by analyzing the zero-transfer part of the fluctuation. p. 0
Resonance Multiplicity
The arithmetic input to the proof of Theorem 3.3 is the sharp estimate that for every fixed d ≥ 1 there exists Rd < ∞ such that, for every L and every ∆̸=0, rL,d (∆) ≤ Rd Ld−1. p. 8 Furthermore, for every L ≥ 2 there exists a nonzero gap ∆ for which rL,d (∆) ≥ Ld−1. p. 8
Quadratic Observables
The paper also addresses quadratic observables by showing that under certain conditions (7.3) and (7.4), there exists a constant Cd,g depending only on d and Cg such that every normalized eigenstate ϕ ∈ HN ∩RanΠE satisfies Dϕ,¡A−αE −βN)2ϕE®≤ Cd,g N Ld−1. p. 6
Joint Concentration Bound
A corollary of the results shows that for every fixed d ≥ 1 there exists Cd < ∞ such that for every nonempty cube Γ = Id properly contained in ΛL,d, every normalized ϕ ∈ HN ∩RanΠE, and every η > 0, the bound °°3IHN −P Neq,ηP Heq,η´ϕ°°2≤ Cd Ld−1Nη squared holds.
One-Particle Gap Multiplicity Proof
The proof of the upper bound in Theorem 4.1 involves comparing representations of a nonzero gap by subtracting them, which leads to a vanishing relation among roots of unity. p. 9 The analysis uses a graph-theoretic counting argument to bound the number of momentum pairs by Cd Ld−1. p. 4
Proof of Mann's Theorem
The proof of Theorem 4.2 involves showing that n is square-free and that every prime divisor of n is at most m, which implies n | Pm. This relies on the fact that the minimal polynomial of a root over K0 has degree p, where p is a prime divisor of n. p. 9
Variance Calculation
The variance calculation in the example shows that tr(ρLNΓ) = N 2/4, and σ 2 L:= tr¡ρL(NΓ − N/2) 2¢=RM(M + 2)12 = N(M + 2)12. This term RM 2/12 has precisely the scale N Ld−1. p. 2
Lower Bound Construction
The construction of exact energy eigenstates that give matching lower bounds for both the variance and the concentration estimate involves a density matrix that commutes with the Hamiltonian and all spatial translations. This construction leads to a tail lower bound where liminf L→∞ VarϕL(NΓ) N Ld−1≥s 2ps/12>0. p. 0
Dirichlet Boundary Condition Proof
The proof for Dirichlet boundary conditions shows that the zero-transfer component c0 = U∗Qbe0QU, and by applying Lemma 5.2 on the torus of side M gives c0 2HS ≤ CdMd−1. p. 2 This ultimately yields the result that ϕ,(NΓ − Nµ)2ϕE®=°°dΓ(bD)ϕ°°2≤ CdN Ld−1. p. 2
Hopping Term Energy Control
For the energy contained in a spatial region, one finds that g0 2HS ≤ 8dLd−1V 2X r!=χbBΓ,ν(r) 2 = 8dLd−1V 2X d ν=1BΓ,ν. This is used to establish the bound on the zero-transfer component. p. 1
Hopping Term Energy Control (Interval)
For an interval I⊂ Z/LZ of length l, one finds that g0 2HS ≤ C(1+logN). This leads to the result that ¯¯〈ϕ,NΓϕ〉 − Nµ¯¯≤ C(1+logN). p. 3
Hopping Term Energy Control (Cube)
For a cube Γ = Id, the bound is given by g0 2HS ≤ 8dLd−1V 2X d ν=1BΓ,ν ≤ 8d squared Ld−1. This leads to the result that ϕ,(HΓ −νΓE)2ϕE®≤ CdN Ld−1. p. 2
Joint Control of Particle Number and Energy
The particle number and energy in Γ can also be controlled simultaneously, leading to the corollary that for every fixed d ≥ 1 there exists Cd < ∞ such that °°3IHN −P Neq,ηP Heq,η´ϕ°°2≤ Cd Ld−1Nη squared.
Final Proof Step
The concentration estimate follows from the orthogonality of B∆ϕ for distinct ∆ and Chebyshev’s inequality. The two projections commute, and IHN −P Neq,ηP Heq,ηL≤3IHN −P Neq,η´+3IHN −P Heq,η´. p. 3
Final Result
The final result is that the factor N Ld−1 in the variance bound is optimal, as is the coefficient Ld−1/N in concentration estimates uniform in η. This concludes that the factor N Ld−1 in the variance bound is optimal, as is the coefficient Ld−1/N in concentration estimates uniform in η. p. 0
References
[BHDM18] BARGHATHI, H., HERDMAN, C. M. and DEL MAESTRO, A.: Rényi generalization of the accessible entanglement entropy. Physical Review Letters 121, 150501 (2018), [DOI:10. p. 21103/PhysRevLett.121.150501]
[CJ76] CONWAY, J. H. and JONES, A. p. 9 J.: Trigonometric Diophantine equations (On vanishing sums of roots of unity). p. 4 Acta Arithmetica 30, 229–240 (1976), [DOI:10.4064/aa-30-3-229-240]
[Eve99] EVERTSE, J.-H.: The number of solutions of linear equations in roots of unity. p. 9 Acta Arithmetica 89, 45–51 (1999), [DOI:10.4064/aa-89-1-45-51]
[Fel71] FELLER, W.: An introduction to probability theory and its applications. p. 4 2 ed., vol. II, John Wiley & Sons, New York, 1971
[GLM+10] GOLDSTEIN, S., LEBOWITZ, J. L., MASTRODONATO, C., TUMULKA, R. and ZANGHÌ, N.: On the approach to thermal equilibrium of macroscopic quantum systems. p. 1 Phys. Rev. E 81, 011109 (2010), [DOI:10.1103/PhysRevE.81.011109]
[Harow] HARROW, A., W.: The Church of the Symmetric Subspace. p. 2 2013, arXiv:1308. p. 06595
[HK26] HARA, T. and KOIKE, T. p. 1: Timescale for macroscopic equilibration in isolated quantum systems: A rigorous derivation for free fermions. p. 2 Physical Review Letters 137, 030406 (2026), [DOI:10. p. 21103/xxxk-y6yp]
[Man65] MANN, H. B.: On linear relations between roots of unity. p. 0 Mathematika 12, 107–117 (1965), [DOI:10.1112/S0025579300005210]
[PR98] POONEN, B. and RUBINSTEIN, M. p. 9: The number of intersection points made by the diagonals of a regular polygon. SIAM Journal on Discrete Mathematics 11, 135–156 (1998), [DOI:10. p. 41137/S0895480195281246]
[RST+25] ROOS, B., SUGIMOTO, S., TEUFEL, S., TUMULKA, R. and VOGEL, C.: Macroscopic thermalization for highly degenerate Hamiltonians after slight perturbation. p. 1 J. Stat. Phys. 192, 109 (2025), [DOI:10.1007/s10955-025-03493-y]
[ST24] SHIRAISHI, N. and TASAKI, H. p. 1: Nature abhors a vacuum: A simple rigorous example of thermalization in an isolated macroscopic quantum system. p. 1 J. Stat. Phys. 191, 82 (2024), [DOI:10.1007/s10955-024-03289-6]
[Tas16] TASAKI, H.: Typicality of thermal equilibrium and thermalization in isolated macroscopic quantum systems. p. 1 J. Stat. Phys. 163, 937–997 (2016), [DOI:10.1007/s10955-016-1511-2]
[Tas25]: Macroscopic irreversibility in quantum systems: Free expansion in a fermionic chain. Phys. Rev. E 111, L062103 (2025), [DOI:10.1103/8cdf-sth8]
[Wło69] WŁODARSKI, L.: On the equation cosα l +cosα m +cosα n +cosα o = 0. p. 6 Ann. Univ. Sci. Budapest. Eötvös Sect. Math. 12, 147–155 (1969)
[Xue26] XUE, Q.: On structured cosine sums and applications. p. 4 2026, arXiv:2607. p. 020907
[XZZ24] XIE, B., ZHAO, Y. and ZHAO, Y. p. 4: On the multiplicity of the eigenvalues of discrete tori. p. 4 2024, arXiv:2411. p. 0
Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.
Improvements for AI systems
-
Enhanced analysis of eigenstate fluctuations in free-fermion systems allows AI to predict concentration bounds for particle number distributions that are uniform in tolerance parameter epsilon, as shown by Theorem 3.3. This enables the AI to rigorously estimate how quickly particle number variance decays with system size N and lattice dimension d.
-
Improved prediction of energy gap multiplicities using Mann's theorem allows the AI to determine the maximal nonzerogap multiplicity, which is bounded by order L d-1, as stated in Theorem 4.1. This capability assists in understanding the complexity of energy spectra for free-fermion lattices on periodic cubic structures.
-
Application of quadratic observables methods enables the AI to establish joint concentration bounds for particle number and energy fluctuations simultaneously, as shown in Corollary 7.5 and Proposition 7.2 extensions to Dirichlet boundary conditions. This allows for a more comprehensive understanding of macroscopic equilibrium properties in both particle density and energy distribution within spatial regions.
-
The derivation of one-dimensional refinement provides the AI with a sharp deviation bound for the expected particle number from its equilibrium value, namely Theorem 3.4:
¯¯〈ϕ,NΓϕ〉 − Nµ ¯¯ ≤ C(1+logN)
. This allows for precise characterization of density fluctuations in low-dimensional systems. -
The construction of exact energy eigenstates with matching lower bounds provides a benchmark for AI models to verify their ability to reproduce optimal variance and concentration estimates. This serves as a rigorous test case for verifying the theoretical limits established in Section 6 and Theorem 3.3 regarding optimality.
Sources
- The Church of the Symmetric Subspace
- On structured cosine sums and applications
- On the multiplicity of the eigenvalues of discrete tori
Related papers
- Existence of a fluid phase for the two-dimensional one-component plasma
- The Huang-Yang conjecture for the low-density Fermi gas
- Anderson localisation in infinite and finite resonator crystals with long-range interactions
- Free energy and spectral edge of the SYK model
- Anderson localization for 1-d quasi-periodic Schr"odinger operators with degenerate weights
- Irregular SLE(4) martingales and isomonodromic deformations