Enhanced Multifractality Induced by Non-Hermitian Disorder in Quantum Percolation

arXiv:2609.04625 · cond-mat.dis-nn · Submitted 2026-09-04 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Condensed Matter. The day in condensed matter: materials, phases and the measurements under them.

Kai: We're looking at the paper: "Enhanced Multifractality Induced by Non-Hermitian Disorder in Quantum Percolation".

Paper summary: Kai: So we're looking at this paper, "Enhanced Multifractality Induced by Non-Hermitian Disorder in Quantum Percolation". Basically, it investigates how mixing geometric dilution and non-Hermitian disorder affects localization properties in a two dimensional quantum site percolation model.

Mira: The core claim is that this combination of disorder actually enhances the multifractal regime of the system. They use complex level spacing statistics, participation entropy, and multifractal analysis to characterize where the eigenstates are localized or delocalized as they tune these disorder and non-Hermiticity parameters.

Lev: What really matters for me is that they find this effect when you introduce purely imaginary on-site potentials, which represent random gain and loss in the system. They also found that this non-Hermitian disorder shifts the quantum percolation threshold, denoted as pq, toward higher occupation probabilities compared to a purely Hermitian version of equal strength <ref:2609.04625#pg1>.

Kai: That's interesting because it means that adding gain and loss doesn't just introduce noise; it actively changes where the system transitions from being localized to being delocalized in terms of occupation probability.

Mira: And they show that this shift is specifically enhanced by the imaginary nature of the disorder, which is a key point when you compare it to a Hermitian potential of the same strength <ref:2609.04625#pg2>.

Lev: From an error correction standpoint, if you try to run this on real hardware, you're looking at these localization lengths and exponents. They find that the localization length exponent nu remains the same relative to its Hermitian value, meaning reciprocity-preserving non-Hermiticity doesn't change the universality class of the transition <ref:2609.04625#pg3>.

Kai: So, even though they found a shift in where things delocalize, the fundamental way it happens—the critical exponents—is still governed by the same rules as standard Hermitian quantum percolation.

Mira: They pinpoint a genuine multifractal critical phase existing between the classical percolation threshold pc and this new quantum threshold pq, and this region actually gets wider as you increase the non-Hermiticity strength gamma/t <ref:2609.04625#pg3>.

Lev: That means for experimental setups, if you're looking at gain or loss effects on photonic waveguides, you might have a broader window where you can observe this complex critical behavior before things become fully delocalized or completely localized.

Kai: It suggests that the non-Hermiticity isn't just an extra complication; it's actively shaping the phase diagram in a way that broadens the interesting region for observation.

Mira: The structure of these states is also characterized by participation entropy, which scales with a generalized fractal dimension dq. In the localized phase, this dimension dq is zero and stays constant regardless of system size <ref:2609.04625#pg3>.

Lev: And in the extended phase, where things are delocalized, the participation entropy scales logarithmically with the system size L over N, which implies a fractal dimension dq of one <ref:2609.04625#pg3>.

Paper summary: Kai: So at that critical point between localization and extension, they see this multifractal behavior characterized by a dependence of dq on q where it's not just constant or one <ref:2609.04625#pg3>.

Mira: This dependence of dq on q is what confirms the region separating the localized and extended phases hosts this genuine multifractal critical phase, which they found is analogous to what you see in standard Hermitian cases <ref:2609.04625#pg3>.

Lev: The paper also compares their results with other studies, showing that the additional suppression of the delocalized phase reported in some literature isn't just a generic effect of adding diagonal disorder <ref:2609.04625#pg3>.

Kai: It was specifically enhanced by the non-Hermitian, gain-and-loss character, which means we need to be careful about how much disorder we introduce when modeling these systems experimentally.

Mira: And looking at the phase diagram they plot, for all occupation probabilities p less than or equal to pq with gamma being the disorder strength, all states are extended <ref:2609.04625#pg1>.

Lev: So, if you're designing a quantum system where you want delocalization, you need to ensure your occupation probability p is above this shifted threshold pq dictated by the non-Hermiticity <ref:2609.04625#pg1>.

Kai: The results from this paper confirm that random gain and loss actively push the system into a more complex, multifractal state in the intermediate region between classical and quantum percolation thresholds <ref:2609.04625#pg3>.

Mira: It really highlights how disorder isn't just about simple localization; it can fundamentally change the complexity of the states at the transition point through these non-Hermitian interactions <ref:2609.04625#pg3>.

Lev: This points toward needing more sophisticated models when you try to build actual devices, because you have to account for this interplay between geometry and gain loss <ref:2609.04625#pg1>.

Kai: So, to sum up the paper "Enhanced Multifractality Induced by Non-Hermitian Disorder in Quantum Percolation", the key is that non-Hermitian disorder, specifically random gain and loss, broadens the region where multifractal behavior is observed while keeping the fundamental universality class of quantum percolation intact.

Mira: It means that for photonic waveguide arrays or similar engineered systems, you can have a wider range of gain-loss strengths before you hit a fully extended phase <ref:2609.04625#pg3>.

Lev: The main limitation they flag is that this analysis relies on the Hamiltonian being complex symmetric, which only happens because the imaginary disorder and real hopping are combined in this specific way <ref:2609.04625#pg3>.

Kai: It's a necessary condition for their math to work out this way, so if your physical system doesn't naturally satisfy that symmetry, you might need a different approach to see these effects.

Mira: But the point is that even with those constraints, the results show that adding gain and loss specifically enhances this multifractal regime <ref:2609.04625#pg1>.

Lev: We'll be looking into how the scaling exponents nu change in real hardware implementations next, because you can't just assume they stay exactly the same when you introduce experimental noise <ref:2609.04625#pg3>.

Conclusion: Kai: So, this paper is about how mixing geometric dilution and non-Hermitian disorder affects localization in quantum percolation.

Mira: Yeah, they look at how random gain and loss changes where the system ends up being localized or extended.

Lev: They used a tightbinding Hamiltonian with complex symmetric properties, which is important for the math to work out this way.

Kai: The main thing here is that this combination of disorder actually enhances the multifractal regime.

Mira: Exactly, they found that non-Hermitian disorder shifts the quantum percolation threshold toward higher occupation probabilities.

Lev: And they showed that adding imaginary disorder, which is gain and loss, specifically enhances this suppression of the delocalized phase compared to a purely Hermitian potential of the same strength.

Kai: So what does this actually mean for us? It suggests that when you have random gain and loss in a system—like an engineered photonic waveguide array—it can actively change the complexity of the states at that transition point.

Mira: Right, because they found this multifractal critical phase exists between the classical percolation threshold and this new quantum one.

Lev: And they confirmed that even though things are shifting, the localization length exponent remains similar to what you'd see in standard Hermitian quantum percolation.

Kai: It's like the fundamental way things transition stays the same, but there’s a broader range of disorder where you can see this complex scaling behavior before it becomes fully localized or extended.

Mira: That means for experimental setups, if you're looking at gain and loss effects on photonic waveguides, you might have a wider window where you can observe this complex critical behavior.

Lev: We need to be careful about the fidelity of that measurement record because we’re dealing with these non-Hermitian terms that introduce gain and loss.

Kai: This points toward needing more sophisticated models when you try to build actual devices, because you have to account for this interplay between geometry and gain loss.

Mira: It's a good summary of what they found about how these specific non-Hermitian interactions shape the physics of the transition point.

Centro Brasileiro de Pesquisas Físicas · Departamento de Física y Astronomıa, Universidad Andres Bello · Departamento de Física, PUC-Rio

cond-mat.dis-nn

Submitted: 2026-09-04

Updated: 2026-11-16

License: http://creativecommons.org/licenses/by/4.0/

Importance score: 83/100

The gist: Enhanced Multifractality Induced by Non-Hermitian Disorder in Quantum Percolation investigates the interplay between geometric dilution and non-Hermitian disorder in the two-dimensional quantum…

Key concepts

Quantum Percolation
This model examines whether particles can form a connected path across a grid of sites with varying occupation probabilities (p). It determines the threshold where the system transitions from having localized, trapped states to having extended, delocalized states.
Non-Hermitian Disorder
This disorder introduces complex values into the system's energy landscape, simulating random gain and loss. Unlike standard Hermitian systems, this imaginary component specifically enhances localization effects and modifies the critical behavior of the percolation transition.
Multifractality
This describes a complex property of wave functions at a critical point where states are neither fully localized nor fully extended. It means that different parts of the state have different scaling properties, characterized by a non-trivial dependence on an exponent (Dq).
Participation Entropy (PE)
This metric measures how spread out or localized an eigenstate is. In this study, it scales with the generalized fractal dimension (Dq). A constant PE indicates localization, while scaling with system size indicates extended states.

Terminology

Summary

Enhanced Multifractality Induced by Non-Hermitian Disorder in Quantum Percolation investigates the interplay between geometric dilution and non-Hermitian disorder in the two-dimensional quantum site-percolation model to characterize localization properties as functions of disorder and non-Hermiticity strength

Model Formulation

The quantum percolation problem is formulated using a tightbinding Hamiltonian where sites are occupied with probability p, and non-Hermiticity is introduced via a purely imaginary on-site potential modeling random gain and loss The Hamiltonian is complex symmetric because the purely imaginary on-site disorder combined with real, reciprocal hopping results in this property The model incorporates geometrical dilution by setting hopping amplitudes to zero between randomly selected neighboring sites

Spectral Characterization

The localization-delocalized transition is probed by analyzing spectral correlations through the complex spacing ratio, defined as rα = s1 / s2, where s1 and s2 are the distances to the nearest and next-nearest neighbors in modulus For uncorrelated complex eigenvalues, the expected mean ratio is ⟨r⟩Poisson = 2/3 ≈ 0.667, while in the delocalized regime, it follows Wigner surmise values like ⟨r⟩GinUE ≈ 0. p. 1

Where the paper says it

For uncorrelated (Poissonian) complex eigenvalues, expected in the localized regime, ⟨r⟩Poisson = 2 3 ≈ 0. 667, while in the delocalized regime the eigenvalues repel according to the Wigner surmise value, with ⟨r⟩GinUE ≈ 0.

72 This ratio crosses over between these two values to signal the localization-delocalization transition as p is tuned

Structural Analysis of States

The structure and localization properties of eigenstates are characterized by the participation entropy (PE), defined using a biorthogonal formulation, which scales with the generalized fractal dimension Dq In the localized phase, Iq remains essentially independent of system size, corresponding to Dq = 0 and a constant participation entropy In the extended phase, pn(j) ∼ 1/N yielding Dq = 1 and Sq ∼ ln N At the critical point, eigenstates display multifractal behavior characterized by a nontrivial dependence of Dq on q with 0 < Dq < 1 [48]

Critical Behavior and Universality

The study finds that non-Hermitian disorder shifts the quantum percolation threshold pq toward larger occupation probabilities, and this suppression is specifically enhanced by the imaginary character of the disorder compared to a Hermitian potential of equal strength The localization-length exponent ν remains the same relative to its Hermitian value, suggesting that reciprocity-preserving non-Hermiticity keeps the localization universality class unchanged The intermediate region between the classical (pc) and quantum percolation thresholds presents a genuine multifractal critical phase, while the localization-length exponent ν remains the same relative to its Hermitian value

Conclusion and Relevance

The results demonstrate that random gain and loss enhance the multifractal regime while preserving the universality class of the quantum percolation transition This is relevant to photonic waveguide arrays where geometric dilution maps onto missing waveguides and random imaginary on-site potentials onto engineered gain or absorption The multifractal regime occupies a finite interval of p that broadens with gain–loss strength, reducing the need for fine-tuning

Table I Summary

Disorder Strength pq ν

|:---: |:---: |:---: |

Hermitian (ϵi) γ/t = 0.3 0.750(3) 2.58(5)

Non-Hermitian (iγi) γ/t = 0.3 0.776(7) 2.45(3)

Hermitian (ϵi) γ/t = 1.6 0.822(7) 2.56(5)

Non-Hermitian (iγi) γ/t = 1.6 0.861(4) 2.73(3)

Table II Comparison of pq and ν using real (Hermitian) and imaginary (non-Hermitian) on-site disorder of equal nominal strength The comparison indicates that the additional suppression of the delocalized phase reported in the main text is not merely a generic consequence of adding diagonal disorder of a given magnitude, but is specifically enhanced by its non-Hermitian, gain-and-loss character

Figure 7 Phase Diagram Summary

The phase diagram shows that for p pq(γ), all states are extended The intermediate region pc < p < pq(γ) is the multifractal critical phase, which increases with γ/t

Figure 8 Multifractal Analysis Summary

In the intermediate region between pc and pq, Dq decreases monotonically with q at fixed p — the hallmark of multifractal wave functions [48] The anomalous scaling exponent ∆q interpolates between these two limits and depends nonlinearly on q, in agreement with the behavior expected for genuinely multifractal eigenstates This confirms that the region separating the localized and extended phases hosts a multifractal critical phase, analogous to the one reported in the Hermitian case [45]

Appendix A Comparison Summary

The extracted threshold, pq = 0.750(3), coincides within error bars with the disorder-free value pq(γ = 0) ≈ 0. p. 8

Where the paper says it

The extracted threshold, pq = 0. 750(3), coincides within error bars with the disorder-free value pq(γ = 0) ≈ 0.

75 [44] A real on-site potential of this magnitude is therefore insufficient to displace the quantum percolation threshold, unlike the imaginary on-site term, which already produces a resolved shift, pq = 0. p. 8
Where the paper says it

750(3), coincides within error bars with the disorder-free value pq(γ = 0) ≈ 0. A real on-site potential of this magnitude is therefore insufficient to displace the quantum percolation threshold, unlike the imaginary on-site term, which already produces a resolved shift, pq = 0.

776(7), at γ/t = 0. p. 5
Where the paper says it

8 D2 pq = 0. 776 ν = 2.

p. 5
Where the paper says it

8 D2 pq = 0. 776 ν = 2.

3 A real on-site potential of this magnitude is therefore insufficient to displace the quantum percolation threshold, unlike the imaginary on-site term, which already produces a resolved shift, pq = 0. p. 8
Where the paper says it

750(3), coincides within error bars with the disorder-free value pq(γ = 0) ≈ 0. A real on-site potential of this magnitude is therefore insufficient to displace the quantum percolation threshold, unlike the imaginary on-site term, which already produces a resolved shift, pq = 0.

776(7), at γ/t = 0.3 Even so, this shift remains smaller than the one produced by imaginary disorder of the same magnitude, pq = 0. p. 8
Where the paper says it

822(7) with ν = 2. Even so, this shift remains smaller than the one produced by imaginary disorder of the same strength, pq = 0.

861(4) and ν = 2. p. 8
Where the paper says it

Even so, this shift remains smaller than the one produced by imaginary disorder of the same strength, pq = 0. 861(3) and ν = 2.

73(3) at γ/t = 1.6 The comparison indicates that the additional suppression of the delocalized phase reported in the main text is not merely a generic consequence of adding diagonal disorder of a given magnitude, but is specifically enhanced by its non-Hermitian, gain-and-loss character

Table II Comparison Summary

Table II summarizes this comparison at both disorder strengths The comparison indicates that the additional suppression of the delocalized phase reported in the main text is not merely a generic consequence of adding diagonal disorder of a given magnitude, but is specifically enhanced by its non-Hermitian, gain-and-loss character

Table I Critical occupation probability pq and critical exponent ν obtained using the finite-size-scaling analysis for each γ strenght The comparison indicates that the additional suppression of the delocalized phase reported in the main text is not merely a generic consequence of adding diagonal disorder of a given magnitude, but is specifically enhanced by its non-Hermitian, gain-and-loss character

Figure 9 Analysis Summary

The extracted threshold, pq = 0.776(7) and ν = 2. p. 5

Where the paper says it

8 D2 pq = 0. 776 ν = 2.

45(3) The comparison indicates that the additional suppression of the delocalized phase reported in the main text is not merely a generic consequence of adding diagonal disorder of a given magnitude, but is specifically enhanced by its non-Hermitian, gain-and-loss character

Figure 6 Evolution Summary

As γ/t increases, the crossing point shifts toward higher occupation probabilities, indicating that it requires a progressively more connected lattice to sustain extended states This behavior becomes particularly pronounced at the largest disorder strengths [Fig. 6 (d)], where the crossing region moves close to the clean limit p = 1 The results are summarized in the phase diagram presented in Fig. p. 6

Where the paper says it

Consequently, the critical region becomes compressed against the upper boundary of the physically accessible interval, 0 ≤ p ≤ 1, and the crossing between different lattice sizes becomes progressively more difficult to resolve, losing sensitivity to ν even after including nonlinear corrections. These results are summarized in the phase diagram presented in Fig.

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. Ability to predict quantum percolation thresholds for non-Hermitian systems based on gain and loss strength. The improved system can determine that non-Hermitian disorder shifts the quantum percolation threshold (pq) toward larger occupation probabilities.

  2. Capability to classify localization universality classes in two dimensions when reciprocity is preserved. The system can confirm that reciprocity preserving non-Hermiticity of this type is irrelevant under disorder averaging, so that the resulting universality class should coincide with the Hermitian counterpart.

  3. Functionality to distinguish between real and imaginary on-site potentials based on their effect on localization. The improved AI can identify that gain and loss suppress delocalization more effectively than an equally strong real potential.

  4. Capacity to characterize critical phases in quantum percolation transitions using multifractal analysis. The system can detect the genuine multifractal critical phase between the classical (pc) and quantum percolation thresholds, which is enhanced by non-Hermitian disorder.

  5. Skill to analyze complex spectral correlations for localization diagnostics. The system can monitor the mean level ratio r-langle as a diagnostic, observing its crossover from Poisson to Wigner-Dyson values signaling delocalization.

  6. Capability to quantify the generalized fractal dimension Dq for eigenstates across various occupation probabilities and moment orders q. The system can extract Dq values that depend nontrivially on q in the intermediate region between pc and pq, confirming multifractal wave functions.

Abstract

We study the two-dimensional quantum site-percolation model with reciprocal hopping and random imaginary on-site potentials, which describe spatially uncorrelated gain and loss. The localization properties of the eigenstates are characterized through the complex spacing ratio, the participation entropy, and the generalized fractal dimensions D q. Finite-size scaling yields an effective quantum percolation threshold p q that increases monotonically with γ, with a shift larger than that produced by a real on-site potential of equal strength. We show that the onset of extended behavior drifts toward larger occupation probabilities as the system size increases. The localization--delocalization change is therefore a crossover, consistent with the absence of a sharp transition for reciprocal non-Hermitian disorder in two dimensions. Accordingly, the interval between the classical threshold p c and p q constitutes an extended multifractal crossover regime and its width at accessible sizes grows with γ. Although all states are expected to localize in the thermodynamic limit, above p q the localization length exceeds the largest system size simulated, which is already larger than current photonic samples. The size-dependent threshold and the multifractal regime, which widens with the loss strength and can therefore be reached without fine-tuning, are thus directly accessible in photonic waveguide lattices.

Sources

Related papers