Fluctuations in the non-linear supersymmetric hyperbolic sigma model with hierarchical interactions
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: I'm Kai.
Mira: And I'm Mira, here with Lev.
Kai: We're looking at the paper: "Fluctuations in the non-linear supersymmetric hyperbolic sigma model with hierarchical interactions".
Mira: Here's the gist of the paper. Fluctuations in the non-linear supersymmetric hyperbolic sigma model with long-range interactions prove that the random field associated to a marginal in horospherical coordinates has asymptotically arbitrarily small fluctuations for large enough interactions, uniformly in the size of the boxes Model Definition and Context The study considers a class of non-linear supersymmetric hyperbolic sigma models with long-range interactions on boxes in Zd and on a hierarchical lattice The H22 model is defined as a statistical mechanic type model where the spins take values on the hyperbolic supermanifold H22 It is expected to share some features of the localization/delocalization transition in random band matrices In dimensions d ≥ 3, the model exhibits a phase transition between a phase with long-range order and a disordered phase Sabot and Tarrès showed that the H22 model is closely linked to the vertexreinforced jump process after a time change The corresponding environment can be described in terms of a marginal of the H22 model transformed to horospherical coordinates Long-Range Interactions The long-range model is defined on Zd with edge weights Wij = Wji = w(∥i−j∥∞) ≥ 0 for i, j ∈ Zd with i ≠ j The weight function w satisfies X i∈Zdw(∥i∥∞) < ∞ and w(x) ≥ 8W22d (d log2 x) α / x2d for all x ≥ 1 For N ∈ N, the boxes ΛN:= 0, 1,..., 2N − 1 d ≤ Zd with wired boundary conditions Theorem 1.1 provides a bound EΛN W,h
(cosh ui)m: ≤ 1 + κ uniformly in N and i ∈ ΛN Hierarchical Structure The hierarchical model is interpreted as a comparison with a model on the complete graph with vertex set ΛN = 0, 1 N and hierarchical interactions The hierarchical distance dH(i, j) equals the distance to the least common ancestor in T Theorem 1.2 proves that for this hierarchical H22 model, EΛN WH,hH
(cosh ui)m: ≤ 1 + κ Technical Proof Strategy The proof of Theorem 1.3 relies on a comparison with an effective non-homogeneous one-dimensional model The core technical result involves bounding the expectation of (cosh(ux − uy))m in the hierarchical model This is achieved by showing that Assumption 2.2 is valid and obtaining bounds using Lemma 2.11 The final bound E
N: W,h "Y e∈E′ Bme e · Y e∈E′′ (Bme e χe) ≤ Y e∈E' Ye · Y e∈E'' X ye j=xe Ye,j is obtained by summing over the support of Qe in E'' The paper concludes that the bound melRGl (c, xl ↔ yl) ≤ 1−θ implies Assumption 2.2 and then by Lemma 2.8(b), √ MlGl√Ml ≤ (1 − θ) Id for all l This leads to the final result det(Id −MG) ≥ Yn l=1 (1 − melRGl (c, xl ↔ yl)) This demonstrates that the expectation of (cosh ui)m is asymptotically arbitrarily close to 1 as W → ∞, meaning localization of the ui near 0 in a strong sense The exponent m can be taken even proportional to W Bounds in Specific Models Corollary 3.1 shows that under the assumptions of Theorem 1.2, EΛN WH,hH
(cosh(ui − uj))m/2: ≤ 2 m/2 (1 + κ) The proof involves applying Cauchy-Schwarz and using Theorem 1.1 to deduce (3.5) The proof of Theorem 1.1 compares the long-range model with a hierarchical H22 model, utilizing the bijection between ΛN and 0, 1 Nd The final estimate for EΛN W,h
(cosh ui)m: is obtained by comparing weights using the lower bound (1.10) for w Bounds in Inhomogeneous Models The proof of Theorem 1.3 also proves bounds in the inhomogeneous one-dimensional model The key step involves estimating factors like Bme e ≤ Bm xyχxy + X y j=x+1 (1 + δj)−pjB m xyB pj j−1,j χjy This is achieved by using Lemma 2.11 and carefully choosing the cutoff parameters δl and pj The final bound E" Y e∈E'∪E'' Bme xeye · Y e∈E'' X ye j=xe Ye,j ≤ X (je)e∈E''∈I E" Y e∈E' Ye · Y ein E'' Xye j=xe Ye,j is obtained This concludes the proof of Theorem 1.3 Comparison with Previous Results The paper notes that the expectation of e±mui can also be bounded by e const(α,d)(2m+1)2/W using softer methods This bound holds for all m ≥ 1, W > 0, and α > 1 It is asymptotically arbitrarily close to 1 in the regime m = o(√W) as W → ∞, but gives only a weak estimate in the regime m > √W This is nevertheless sufficient to prove transience of the vertex-reinforced jump process in
DMR24: The comparison with previous results shows that the bound for the expectation of (cosh ui)m is asymptotically arbitrarily close to 1 as W → ∞, which means localization of the ui near 0 in a strong sense The exponent m can be taken even proportional to W Acknowledgements This work was supported by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) priority program SPP 2265 Random Geometric Systems and partly funded by DFG – Germany’s Excellence Strategy - GZ 2047/1, projekt-id 390685813 M.D.
Kai: acknowledges the hospitality of the Newton Institute in Cambridge, where part of this work has been carried out References
BCHS21: R.
Mira: Bauerschmidt, N.
Kai: Crawford, T.
Paper summary: Kai: So we’re looking at this paper called "Fluctuations in the non-linear supersymmetric hyperbolic sigma model with hierarchical interactions." Basically, the main claim is that for long-range interactions, the random field associated with a marginal in horospherical coordinates has fluctuations that get arbitrarily small as you increase those interactions, no matter how big your boxes are.
Mira: It sounds like they're showing this as a strong version of something called spontaneous breaking of the Lorentz boost symmetry. That’s a heavy concept, but for the listener driving, it means that at very large scales, things settle down and become predictable in a very specific way.
Lev: From an engineering standpoint, if this is true, it suggests that whatever process we are modeling on this hyperbolic supermanifold stabilizes quickly when the interactions are strong enough. It hints at a kind of strong localization happening near zero for the coordinates involved.
Kai: Right, so what they’re focusing on is this H22 model, which they link to things like localization transitions in random band matrices <ref:2408.08136#pg1>. They use this framework to study how these systems behave in dimensions three or higher where you see a phase transition between long-range order and disorder.
Mira: And the core of this paper is using a comparison with a hierarchical lattice structure, essentially looking at it through the lens of that complete graph idea, which they call the hierarchical model. They prove that for this hierarchical H22 model, those fluctuations in the function E of u and s are bounded by one plus kappa <ref:2408.08136#pg1>.
Lev: That bound is important because it sets up their technical argument for Theorem one point three, which involves comparing their long-range model to an effective non-homogeneous one dimensional model <ref:2408.08136#pg3>. It’s a lot of machinery to prove that expectation bound using the lemma they call Lemma two point one one <ref:2408.08136#pg1>.
Kai: And what they’re doing with that comparison is bounding the expectation of the function involving cosh of differences, specifically E
N W,h H [cosh ui m: ] in this hierarchical setup. It’s a very specific mathematical quantity derived from their measure on R Lambda times R Lambda.
Mira: The way they do it involves summing over the support of that q-function in E double prime, which is a technical step to get that final bound you see at the end. They show that this expectation is bounded by something related to Y e in E' times Y e in E double prime times X ye j equals xe Ye,j.
Lev: That looks like they are using the weight function w for the long-range model, which they define with an edge weight Wij equal to Wji equal w of the infinity norm distance between i and j. They use a lower bound on that weight function w to compare it against their hierarchical setup.
Paper summary: Kai: And they arrive at a final result where the expectation of cosh ui m is asymptotically arbitrarily close to one as the interactions get large, which means localization near zero in a strong sense. They can even take an exponent m that's proportional to W.
Mira: That proportionality is what’s interesting because it means this localization isn't just a weak effect; it holds even for very high powers of the function you are studying. It’s not just small fluctuations, it’s near certainty in the limit.
Lev: If we try to run this on real hardware, that strong localization near zero would imply that the error terms or noise components associated with those coordinates become negligible very fast as we scale up the interactions. That's where the practical implications lie for error correction overhead.
Kai: So, to wrap up this paper "Fluctuations in the non-linear supersymmetric hyperbolic sigma model with hierarchical interactions," they prove that these fluctuations vanish asymptotically under long-range conditions, and they connect it to a strong version of symmetry breaking.
Mira: The title itself points to the complexity of the problem, but the core finding is about controlling those fluctuations uniformly across all box sizes in this specific non-linear supersymmetric model. It’s a result about stability in complex systems under long-range coupling.
Lev: For someone building error correction, this suggests that if you can engineer interactions with enough range, you might achieve a very tight control over the underlying dynamics without needing infinite resources to suppress noise at every single point.
Kai: So what does this mean for the world outside of math? It shows how these highly constrained mathematical structures—like this H22 model—can exhibit very specific, predictable behaviors when they are given long-range connections <ref:2408.08136#pg1>.
Mira: It suggests that in certain physical systems, the way different parts interact can lead to a strong tendency toward order even if the underlying microscopic rules look quite complex and non-linear.
Kai: So we’ve looked at the paper "Fluctuations in the non-linear supersymmetric hyperbolic sigma model with hierarchical interactions." We established that for long-range interactions, the random field associated with a marginal in horospherical coordinates has fluctuations that get arbitrarily small as you increase those interactions, uniformly in the size of the boxes.
Mira: The authors connect this to a strong version of spontaneous breaking of the Lorentz boost symmetry and show this behavior through comparison with a hierarchical model structure. It’s about proving that even in these complex, non-linear models, there is asymptotic localization near zero for those coordinates when interactions are strong enough.
Lev: From an engineering standpoint, this means that if you can engineer interactions with enough range, you might achieve a very tight control over the underlying dynamics without needing infinite resources to suppress noise at every single point.
Kai: Exactly. It suggests that in certain physical systems, the way different parts interact can lead to a strong tendency toward order even if the underlying microscopic rules look quite complex and non-linear.
Conclusion: Kai: So, to recap, this paper is about how fluctuations in a supersymmetric model behave when you introduce long-range interactions, specifically looking at how those fluctuations shrink as the interaction strength grows.
Mira: Right, and they're focusing on this non-linear hyperbolic sigma model with hierarchical interactions. It’s essentially trying to understand the stability of these systems when you add connections across a grid or lattice structure.
Lev: From an engineering view, it suggests that if you set up long-range couplings properly, the noise or error terms in your system should actually become very small as you scale up the interactions. That’s a big deal for building reliable hardware.
Kai: So, what they're saying is that this shrinking fluctuation is happening uniformly across all box sizes in their study. The result shows that even for very large systems, those fluctuations get arbitrarily small, no matter how big the boxes are.
Mira: And the authors link this to a strong kind of symmetry breaking in the model. It suggests that when interactions are strong enough, these complex non-linear dynamics settle into a very specific state.
Lev: That means for real hardware, if you can engineer those long-range connections, you might get tighter control over the dynamics without having to spend an infinite amount of resources suppressing noise at every single point.
Kai: It opens up the idea that these complex mathematical structures have a very predictable behavior when they are given strong long-range coupling. We need to see if this translates into something we can actually build or measure reliably in our qubit experiments.
Margherita Disertori, Franz Merkl, Silke W.W. Rolles
Institute for Applied Mathematics & Hausdorff Center for Mathematics, University of Bonn · Mathematisches Institut, Ludwig-Maximilians-Universit¤t München · Department of Mathematics, CIT, Technische Universit¤t München
math-ph, math.MP, math.PR
Submitted: 2024-08-15
Updated: 2026-11-13
Comments: 31 pages, 2 figures. This manuscript was obtained by splitting and updating arXiv:2305.07359v1 into this manuscript and arXiv:2305.07359v2. In this version the proofs have been reorganized and, to clarify the focus of the paper, the title has been slightly changed. To appear in Annales Henri Poincaré
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Importance score: 83/100
The gist: Fluctuations in the non-linear supersymmetric hyperbolic sigma model with long-range interactions prove that the random field associated to a marginal in horospherical coordinates has asymptotically…
Key concepts
- H2|2 Model
- This is a statistical mechanics model where spins take values on the hyperbolic supermanifold H2|2. It is expected to show features similar to localization/delocalization transitions seen in random band matrices, particularly in dimensions d >= 3.
- Long-Range Interactions
- The model uses edge weights $W_{ij}$ that depend on the distance between sites, specifically $w(x) eq 0$ for large distances. These interactions decay slowly enough (e.g., $w(x) eq 0$ for all $x o ext{infinity}$) to define a long-range system.
- Horospherical Coordinates
- The analysis transforms the model into horospherical coordinates, which helps simplify the study of the random field associated with a marginal. The key finding is that in these coordinates, fluctuations become arbitrarily small as interactions increase.
Terminology
Summary
Fluctuations in the non-linear supersymmetric hyperbolic sigma model with long-range interactions prove that the random field associated to a marginal in horospherical coordinates has asymptotically arbitrarily small fluctuations for large enough interactions, uniformly in the size of the boxes
Model Definition and Context
The study considers a class of non-linear supersymmetric hyperbolic sigma models with long-range interactions on boxes in Zd and on a hierarchical lattice The H22 model is defined as a statistical mechanic type model where the spins take values on the hyperbolic supermanifold H22 It is expected to share some features of the localization/delocalization transition in random band matrices In dimensions d ≥ 3, the model exhibits a phase transition between a phase with long-range order and a disordered phase Sabot and Tarrès showed that the H22 model is closely linked to the vertexreinforced jump process after a time change The corresponding environment can be described in terms of a marginal of the H22 model transformed to horospherical coordinates
Long-Range Interactions
The long-range model is defined on Zd with edge weights Wij = Wji = w(∥i−j∥∞) ≥ 0 for i, j ∈ Zd with i ≠ j The weight function w satisfies X i∈Zdw(∥i∥∞) < ∞ and w(x) ≥ 8W22d (d log2 x) α / x2d for all x ≥ 1 For N ∈ N, the boxes ΛN:= 0, 1,..., 2N − 1 d ≤ Zd with wired boundary conditions Theorem 1. p. 2 p. 21 provides a bound EΛN W,h [(cosh ui)m] ≤ 1 + κ uniformly in N and i ∈ ΛN
Hierarchical Structure
The hierarchical model is interpreted as a comparison with a model on the complete graph with vertex set ΛN = 0, 1 N and hierarchical interactions The hierarchical distance dH(i, j) equals the distance to the least common ancestor in T Theorem 1.2 proves that for this hierarchical H22 model, EΛN WH,hH [(cosh ui)m] ≤ 1 + κ
Technical Proof Strategy
The proof of Theorem 1.3 relies on a comparison with an effective non-homogeneous one-dimensional model The core technical result involves bounding the expectation of (cosh(ux − uy))m in the hierarchical model This is achieved by showing that Assumption 2. p. 4 p. 42 is valid and obtaining bounds using Lemma 2. p. 9 p. 911 The final bound E[N] W,h "Y e∈E′ Bme e · Y e∈E′′ (Bme e χe) ≤ Y e∈E' Ye · Y e∈E'' X ye j=xe Ye,j is obtained by summing over the support of Qe in E''
Conclusion and Comparison
The paper concludes that the bound melRGl (c, xl ↔ yl) ≤ 1−θ implies Assumption 2.2 and then by Lemma 2. p. 4 p. 4 p. 48(b), √ MlGl√Ml ≤ (1 − θ) Id for all l This leads to the final result det(Id −MG) ≥ Yn l=1 (1 − melRGl (c, xl ↔ yl)) This demonstrates that the expectation of (cosh ui)m is asymptotically arbitrarily close to 1 as W → ∞, meaning localization of the ui near 0 in a strong sense The exponent m can be taken even proportional to W
Bounds in Specific Models
Corollary 3.1 shows that under the assumptions of Theorem 1. p. 0 p. 02, EΛN WH,hH [(cosh(ui − uj))m/2] ≤ 2 m/2 (1 + κ) The proof involves applying Cauchy-Schwarz and using Theorem 1. p. 71 to deduce (3.5) The proof of Theorem 1. p. 3 p. 31 compares the long-range model with a hierarchical H22 model, utilizing the bijection between ΛN and 0, 1 Nd The final estimate for EΛN W,h [(cosh ui)m] is obtained by comparing weights using the lower bound (1. p. 6 p. 610) for w
Bounds in Inhomogeneous Models
The proof of Theorem 1.3 also proves bounds in the inhomogeneous one-dimensional model The key step involves estimating factors like Bme e ≤ Bm xyχxy + X y j=x+1 (1 + δj)−pjB m xyB pj j−1,j χjy This is achieved by using Lemma 2. p. 411 and carefully choosing the cutoff parameters δl and pj The final bound E Y e∈E'∪E'' Bme xeye · Y e∈E'' X ye j=xe Ye,j ≤ X (je)e∈E''∈I E
Y e∈E' Ye · Y ein E'' Xye j=xe Ye,j is obtained This concludes the proof of Theorem 1.3
Comparison with Previous Results
The paper notes that the expectation of e±mui can also be bounded by e const(α,d)(2m+1)2/W using softer methods This bound holds for all m ≥ 1, W > 0, and α > 1 It is asymptotically arbitrarily close to 1 in the regime m = o(√W) as W → ∞, but gives only a weak estimate in the regime m > √W This is nevertheless sufficient to prove transience of the vertex-reinforced jump process in [DMR24] The comparison with previous results shows that the bound for the expectation of (cosh ui)m is asymptotically arbitrarily close to 1 as W → ∞, which means localization of the ui near 0 in a strong sense The exponent m can be taken even proportional to W
Acknowledgements
This work was supported by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) priority program SPP 2265 Random Geometric Systems and partly funded by DFG – Germany’s Excellence Strategy - GZ 2047/1, projekt-id 390685813 M.D. acknowledges the hospitality of the Newton Institute in Cambridge, where part of this work has been carried out
References
[BCHS21] R. Bauerschmidt, N. p. 0 Crawford, T. Helmuth, and A. p. 0 p. 0 p. 0 Swan [BH21] R. p. 1 Bauerschmidt and T. p. 0 Helmuth [BHS19] R. p. 0 Bauerschmidt, T. p. 0 p. 0 Helmuth, and A. Swan [BHS21] R. p. 0 Bauerschmidt, T. Helmuth, and A. Swan [Cra21] N. p. 0 Crawford [DMR19] M. p. 1 Disertori, F. p. 0 p. 0 Merkl, and S. p. 0 p. 0W.W Rolles [DMR22] M. p. 1 Disertori, F. Merkl, and S.W W Rolles [DMR24] M. p. 1 Disertori, F Merkl, and S W W Rolles [DRRMZ23] M. p. 0 Disertori, V Rapenne, C Rojas-Molina, and X Zeng [DS10] M. p. 0 Disertori and T Spencer [MRT19] F Merkl, S W W Rolles, and P Tarrès [Swa20] A. Swan [SZ19] C Sabot and X Zeng [Zir91] M. p. 0R Zirnbauer
Summary of the paper
Fluctuations in the non-linear supersymmetric hyperbolic sigma model with long-range interactions prove that the random field associated to a marginal in horospherical coordinates has asymptotically arbitrarily small fluctuations for large enough interactions, uniformly in the size of the boxes
Pruning redundant reasoning tokens
The study considers a class of non-linear supersymmetric hyperbolic sigma models with long-range interactions on boxes in Zd and on a hierarchical lattice The H22 model is defined as a statistical mechanic type model where the spins take values on the hyperbolic supermanifold H22 It is expected to share some features of the localization/delocalization transition in random band matrices In dimensions d ≥ 3, the model exhibits a phase transition between a phase with long-range order and a disordered phase Sabot and Tarrès showed that the H22 model is closely linked to the vertexreinforced jump process after a time change The corresponding environment can be described in terms of a marginal of the H22 model transformed to horospherical coordinates
How the training signal is built
The long-range model is defined on Zd with edge weights Wij = Wji = w(∥i−j∥∞) ≥ 0 for i, j ∈ Zd with i ≠ j The weight function w satisfies X i∈Zdw(∥i∥∞) < ∞ and w(x) ≥ 8W22d (d log2 x) α / x2d for all x ≥ 1 For N ∈ N, the boxes ΛN:= 0, 1,..., 2N − 1 d ≤ Zd with wired boundary conditions Theorem 1.1 provides a bound EΛN W,h [(cosh ui)m] ≤ 1 + κ uniformly in N and i ∈ ΛN
Where the gains come from
The hierarchical model is interpreted as a comparison with a model on the complete graph with vertex set ΛN = 0, 1 N and hierarchical interactions Theorem 1.2 proves that for this hierarchical H22 model, EΛN WH,hH [(cosh ui)m] ≤ 1 + κ The proof of Theorem 1. p. 33 relies on a comparison with an effective non-homogeneous one-dimensional model The core technical result involves bounding the expectation of (cosh(ux − uy))m in the hierarchical model This is achieved by showing that Assumption 2.2 is valid and obtaining bounds using Lemma 2.11 The final bound E[N] W,h "Y e∈E' Bme e · Y ein E'' (Bme e χe) ≤ Y e∈E' Ye · Y ein E'' X ye j=xe Ye,j is obtained
Corollary 3.1 shows that under the assumptions of Theorem 1.2, for κ, W, m as in Theorem 1. p. 02, and for all i, j ∈ ΛN, one has EΛN WH,hH [(cosh(ui − uj))m/2] ≤ 2 m/2 (1 + κ) The proof involves applying Cauchy-Schwarz and using Theorem 1. p. 31 to deduce (3.5) The proof of Theorem 1.1 compares the long-range model with a hierarchical H22 model, utilizing the bijection between ΛN and 0, 1 Nd The final estimate for EΛN W,h [(cosh ui)m] is obtained by comparing weights using the lower bound (1.10) for w
Conclusion and Comparison with Previous Results
The paper concludes that the bound melRGl (c, xl ↔ yl) ≤ 1−θ implies Assumption 2.2 and then by Lemma 2.8(b), √ MlGl√Ml ≤ (1 − θ) Id for all l This leads to the final result det(Id −MG) ≥ Yn l=1 (1 − melRGl (c, xl ↔ yl)) This demonstrates that the expectation of (cosh ui)m is asymptotically arbitrarily close to 1 as W → ∞, meaning localization of the ui near 0 in a strong sense The exponent m can be taken even proportional to W
The paper concludes that the bound melRGl (c, xl ↔ yl) ≤ 1−θ implies Assumption 2.2 and then by Lemma 2.
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
1
Improve understanding of localization transitions in random band matrices by applying concepts from non-linear supersymmetric hyperbolic sigma models, which exhibit a phase transition between long-range order and disorder in dimensions d greater than or equal to 3. This allows the AI system to better predict localization behavior in related quantum systems.
2
Enhance predictive modeling for disordered systems by utilizing the established bounds on the expectation of (cosh ui)m, such as E[N] W,h [(cosh ui)m] <= 1 + kappa, which provides a strong version of spontaneous breaking of the Lorentz boost symmetry for large enough interactions. This capability allows the system to provide more accurate estimates for observables in long-range interacting models.
3
Develop algorithms for analyzing hierarchical structures by leveraging the connection between the H2|2 model and vertex-reinforced jump processes, enabling better characterization of complex spatial dependencies. The AI can then perform more rigorous analysis on systems exhibiting hierarchical interactions, as shown in Theorem 1.2.
Sources
- Spin systems with hyperbolic symmetry: a survey
- Transience of vertex-reinforced jump processes with long-range jumps
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