On a pinning model in correlated Gaussian random environments

Basic Information

• Paper ID: 2501.00445
• Title: On a pinning model in correlated Gaussian random environments
• Authors: Zian Li, Jian Song, Ran Wei, Hang Zhang
• Classification: math.PR (Probability Theory)
• Publication Date: December 31, 2024
• Paper Link: https://arxiv.org/abs/2501.00445

Abstract

This paper investigates the pinning model in correlated Gaussian random environments. For models in the disorder-relevant regime, the authors study the moderate disorder region and prove that the rescaled partition function converges to a non-trivial continuous limit in both the Skorohod and Stratonovich settings. The results partially confirm the predictions of Weinrib and Halperin regarding disorder relevance/irrelevance.

Research Background and Motivation

Problem Background

1. Importance of the pinning model: The disordered pinning model is an important class of random polymer models defined through Gibbs transformations of renewal process distributions. These models are valuable for studying phase transitions and critical phenomena, with widespread applications in physics, chemistry, and biology.
2. Disorder correlation problem: When disorder is i.i.d., the Harris criterion shows that disorder is irrelevant when α < 1/2 and relevant when α > 1/2. However, when disorder exhibits correlation, the situation becomes more complex.
3. Weinrib-Halperin prediction: When disorder correlation takes the form n^(-p) (p > 0), Weinrib and Halperin predicted that disorder should be relevant when α > min(p,1)/2 and irrelevant when α < min(p,1)/2.

Research Motivation

1. Theoretical gap: Berger has confirmed the Weinrib-Halperin prediction for p > 1, but the case p < 1 exhibits different behavioral characteristics where traditional methods no longer apply.
2. Need for new perspectives: When p < 1, the disordered system lacks a localization/delocalization phase transition because the quenched free energy is always strictly positive, and the system remains in a localized state.
3. Scaling limit investigation: Studying weak limits of rescaled partition functions provides a new perspective for understanding disorder correlation, particularly for the p < 1 case.

Core Contributions

1. Theoretical completion: Supplements Berger's results for the case where environmental correlation takes the form n^(-p) (p < 1), obtaining scaling limits of rescaled partition functions in the disorder-relevant region.
2. Prediction verification: Partially confirms the correctness of the Weinrib-Halperin prediction in cases different from the Harris criterion.
3. New phenomena discovery: Proves that rescaled partition functions are not L² bounded throughout the disorder-relevant region, a novel characteristic compared to other disordered models.
4. Methodological innovation:
   • Develops decoupling methods for proving uniform integrability
   • Provides sharp upper bounds for trace norms
   • Gives sharp upper bounds for second moments of multiple Stratonovich integrals

Methodology Details

Task Definition

Investigating the scaling behavior of the pinning model in correlated Gaussian random environments, specifically including:

• Input: Gaussian random environments with power-law decaying correlations
• Output: Weak convergence limits of rescaled partition functions
• Constraints: Hurst parameter H ∈ (1/2, 1), different ranges of renewal process parameter α

Model Architecture

1. Basic Model Setup

Renewal process: Let τ = {τᵢ, i ∈ ℕ₀} be a continuous renewal process with inter-arrival distribution:

P_τ(τ₁ = n) = L(n)/n^(1+α), n ≥ 1

where α > 0 and L(·) is a slowly varying function.

Random environment: ω = {ωₙ, n ∈ ℕ} is a family of Gaussian random variables independent of τ, satisfying:

E_ω[ωₙ] = 0, E_ω[ωₙωₙ'] = γ(n-n')

The covariance function satisfies: γ(n) ≲ |n|^(2H-2) ∧ 1

2. Partition Function Definition

Original partition function:

Z^(ω,h)_(N,β) = E_τ[exp{∑_(x∈Ω_N) (βωₙₓ + h)δₓ}]

Wick-ordered partition function:

Z̃^(ω,h)_(N,β) = E_τ[exp{∑_(x∈Ω_N) (βωₙₓ + h)δₓ - (1/2)Var_ω(∑_(x∈Ω_N) βωₙₓδₓ)}]

3. Scaling Parameters

Corresponding scalings are defined according to different cases of the renewal process:

For 0 < α < 1:

β_N = β̂ L(N)/(N^(α+H-1)), h_N = ĥ L(N)/N^α

For E_ττ₁ < ∞:

β_N = β̂/N^H, h_N = ĥ/N

Technical Innovation Points

1. Wick Expansion Technique

Since disorder exhibits correlation, polynomial chaos expansion cannot be directly applied. Correlation is handled through Wick ordering using Wick exponential expansion:

Z̃_(N,β)^(ω,h) = E_τ[exp{h∑δₓ} exp_◦{β∑ωₙₓδₓ}]

2. U-Statistic Method

U-statistics are defined to handle the transition from discrete to continuous:

I_m^(N)(f) = N^(-mH) ∑_(n∈[N]^m) A_N(f)(n/N) ω_n^◦

3. Trace Estimation Technique

For the Stratonovich integral case, precise trace estimates are required. These are obtained through kernel correction procedures and domain partitioning techniques:

‖Tr_j ψ_(m,k)‖²_(|H|⊗(m-2j)) ≤ ‖Tr_j φ_(m+k)‖²_(|H|⊗(m+k-2j))

Experimental Setup

Theoretical Verification Framework

This work is primarily theoretical, with results verified through rigorous mathematical proofs, including:

1. Convergence verification: Proving weak convergence of each chaos term
2. Consistency verification: Proving uniform convergence of truncated approximations
3. Integrability verification: Proving L² integrability of limits

Key Assumptions

Assumption (A1): H ∈ (1/2, 1) and α + H > 1 Assumption (A2): H ∈ (1/2, 1) and α + 2H > 2 Assumption (A3): H ∈ (1/2, 1) and α + H > 3/2

Main Results

Theorem 1.1 (Skorohod Case)

Under assumptions H ∈ (1/2, 1) and α > 1/2:

Z̃_(N,β_N)^(ω,h_N) →^d Z̃^(β̂,ĥ)

where the limit is given by an L² convergent series:

Z̃^(β̂,ĥ) = 1 + ∑_(k=1)^∞ (1/k!) ∫_([0,1]^k) φ_k(t₁,...,t_k) ∏ᵢ(β̂◦W(dtᵢ) + ĥdtᵢ)

Theorem 1.2 (Stratonovich Case)

Under assumptions H ∈ (1/2, 1) and α + H > 3/2:

Z_(N,β_N)^(ω,h_N) →^d Z^(β̂,ĥ)

where the limit is:

Z^(β̂,ĥ) = ∑_(k=0)^∞ (1/k!) ∫_([0,1]^k) φ_k(t₁,...,t_k) ∏ᵢ(β̂W(dtᵢ) + ĥdtᵢ)

Key Findings

1. Criticality: α = 1/2 is indeed the critical point; when α < 1/2, {Z̃_N} is not uniformly L² bounded.
2. Uniform integrability: Under assumption (A2), the family of rescaled partition functions is uniformly integrable.
3. Conditional boundary cases: Results hold for conditional partition functions with appropriately modified integral functions.

Related Work

Historical Development

1. Harris criterion: When disorder is i.i.d., it determines α = 1/2 as the critical point for relevance/irrelevance.
2. Weinrib-Halperin prediction: Generalizes the Harris criterion to the correlated disorder case.
3. Berger's work: Confirms the prediction for p > 1 and discovers new phenomena for p < 1.

Related Models

1. Directed polymers: Work by Caravenna, Sun, and Zygouras provides new perspectives for this research.
2. Spatiotemporal correlation models: Related work covers various cases of spatial correlation, temporal correlation, and spatiotemporal correlation.

Conclusions and Discussion

Main Conclusions

1. Partially confirms the correctness of the Weinrib-Halperin prediction for the p < 1 case
2. Discovers new characteristics of the pinning model in correlated environments: L² non-boundedness
3. Establishes scaling limit theory from discrete to continuous settings

Limitations

1. Condition restrictions: Requires α > 1/2 to ensure L² convergence, while the Weinrib-Halperin prediction suggests α + H - 1 > 0 suffices
2. Technical assumptions: Proof of uniform integrability requires stronger assumption (A2)
3. Incompleteness: Only partially confirms the prediction; the case α ≤ 1/2 requires further investigation

Future Directions

1. Weakening conditions: Investigating weak convergence under conditions α + H - 1 > 0, α ≤ 1/2
2. Other models: Extending methods to other disordered models in correlated environments
3. Numerical verification: Verifying theoretical predictions through numerical simulations

In-Depth Evaluation

Strengths

1. Theoretical rigor: Complete and rigorous mathematical proofs with refined technical handling
2. Methodological innovation: Develops new techniques for handling correlated disorder, particularly decoupling methods and trace estimation
3. Important results: Partially resolves important open problems and discovers new phenomena
4. Clear presentation: Well-structured with appropriate technical detail handling

Weaknesses

1. Condition restrictions: Main results require stronger conditions than expected
2. Incompleteness: Does not fully confirm the Weinrib-Halperin prediction
3. Limited applications: Primarily theoretical results with limited practical application value to be further explored

Impact

1. Theoretical contribution: Makes important contributions to disorder model theory in correlated environments
2. Methodological value: Developed techniques have reference value for related research
3. Open problems: Points directions for subsequent research

Applicable Scenarios

This research is applicable to:

1. Study of disordered systems in statistical physics
2. Theoretical analysis of random polymer models
3. Limit theory of correlated Gaussian fields
4. Related problems in stochastic partial differential equations

References

The paper cites 26 important references, primarily including:

• Berger's foundational work on pinning models in correlated environments
• Giacomin's monograph on pinning models
• Work by Caravenna, Sun, and Zygouras on new perspectives on disorder correlation
• Nualart's classical textbook on Malliavin operators
• Original prediction papers by Weinrib and Halperin