Reentrant Localization Length in a Quasiperiodic Thue-Morse Chain

summary

The gist

We investigate single-particle localization in a dimerized Su–Schrieffer–Heeger (SSH) chain with a quasiperiodic onsite potential masked by the deterministic Thue–Morse sequence The TM-masked…

In short

The study investigates localization in a dimerized SSH chain with a Thue-Morse sequence potential. It found that increasing potential strength causes mid-spectrum states to cycle through localized, multifractal, and localized regimes—a reentrant recovery of participation. This effect is driven by short-range sign anticorrelation aligned with strong hopping bonds, interpreted as an effective-dimer mechanism.

Key concepts

Correlation Dimension (Dq)
This metric measures the fractal nature of a wavefunction's probability distribution. By calculating Dq for different moment orders (q), researchers can characterize the complexity of the localized states, revealing a systematic hierarchy in how different moments probe distinct effective supports within the reentrant regime.
Reentrant Localization
This phenomenon describes a non-monotonic behavior where states initially localize under increasing disorder strength, then become partially delocalized (multifractal) at an intermediate strength, and finally localize again at even stronger disorder. This cycle is observed in the mid-spectrum of the chain.
Effective-Dimer Interpretation
This interpretation suggests that opposite signs in the quasiperiodic potential reduce the differential onsite detuning specifically across strong hopping bonds. This reduction favors hybridization between corresponding dimer states, providing a structural explanation for why reentrance occurs when short-range sign anticorrelation is present.
Thue-Morse Sequence Masking
The chain's onsite energies are modulated by the deterministic Thue-Morse sequence, which introduces a quasiperiodic potential. This specific sequence structure allows researchers to analytically track localization transitions with high resolution, providing a transparent setting to study complex quasiperiodic effects.

Terminology used across episodes

This episode discusses

The paper

Reentrant Localization Length in a Quasiperiodic Thue-Morse Chain · Read on arXiv

Taylan Yıldız, B. Tanatar

Department of Physics, Bilkent University

We investigate single-particle localization in a dimerized Su-Schrieffer-Heeger chain whose quasiperiodic onsite potential is masked by the deterministic Thue--Morse sequence. The behavior is characterized by the transfer-matrix Lyapunov exponent, and by the participation moments of exactly diagonalized chains. For intracell-dominated hopping the mid-spectrum Lyapunov exponent depends nonmonotonically on the potential strength, developing a pronounced minimum that enhances the localization length by an order of magnitude over an intermediate window, whereas an intercell-dominated chain responds monotonically. The exponent remains positive at every sampled nonzero potential strength and at every state of the spectral sector, and the second participation moment saturates over the full range of system sizes, so the intermediate window is an enhancement of a finite localization length and not a change of localization class. Finite-size generalized dimensions follow the same response in real space and in the Fourier-conjugate representation, while their moment-order separation decreases systematically with size. Controlled sign masks identify the origin: the response requires short-range sign anticorrelation aligned with the dominant bonds, and displacing the modulation by a single site, which destroys the exactness of that anticorrelation while leaving the sequence otherwise intact, removes the response entirely. An effective-dimer description relates the alignment to the division of the onsite modulation between the internal detuning of a bonded pair and its common-mode energy.

DOI: 10.1038/s41598-026-71207-9

Transcript

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

Kai: I'm Kai.

Mira: And I'm Mira, here with Lev.

Kai: We're looking at the paper: "Reentrant Localization Length in a Quasiperiodic Thue-Morse Chain".

Mira: Here's the gist of the paper. We investigate single-particle localization in a dimerized Su–Schrieffer–Heeger (SSH) chain with a quasiperiodic onsite potential masked by the deterministic Thue–Morse sequence The TM-masked SSH model provides an analytically transparent setting in which reentrant localization can be tracked with high resolution Localization Diagnostics and Scaling Metrics The localization landscape of the TM-masked SSH chain is mapped as a function of the quasiperiodic potential strength and hopping dimerization using diagnostics like the correlation dimension, inverse participation ratio (IPR), and normalized participation ratio (NPR) For appropriate hopping ratios, increasing the potential strength drives the mid-spectrum states through a localized–multifractal–localized sequence, producing a reentrant recovery of participation before localization is restored at stronger modulation Energy-resolved diagnostics show that this recovery is concentrated in the central spectral region rather than occurring uniformly throughout the spectrum Extrapolations of the generalized dimensions to the thermodynamic limit reveal a systematic moment-dependent hierarchy in the reentrant window, accompanied by a broadened thermodynamic-limit singularity spectrum Characterization of Multifractality The multifractal character of the intermediate reentrant regime is established through generalized dimensions, where the correlation dimension Dq depends systematically on the moment order q The generalized dimensions are first averaged over the complete sector at each finite size and are subsequently extrapolated according to Dq(L,δ) = D(∞)q(δ) + a q(δ)L In the reentrant window, the thermodynamic-limit dimensions exhibit a systematic hierarchy D(∞) 2 > D(∞) 3 > D(∞) 4 > D(∞) 5, showing that different moments probe distinct effective supports The singularity spectrum f(α) provides a complementary description of the multifractal scaling within the reentrant window Origin of Reentrant Localization The origin of reentrance is tested by comparing the TM modulation with unbiased random, globally balanced, pair-canceling, and block-permuted binary masks Short-range sign anticorrelation promotes reentrance when it is aligned with the dominant SSH hopping bonds This effect is interpreted as an effective-dimer interpretation of the reentrant response, relating it to the reduced differential onsite detuning of the corresponding strong-bond pairs The comparison shows that pair cancellation alone is not sufficient to produce reentrance independently of the hopping parameters Effective-Dimer Interpretation and Structural Origin The effective-dimer interpretation suggests that opposite signs reduce the differential onsite detuning across the strong bonds and favor hybridization of the corresponding dimer states The comparison between intracell- and intercell-aligned pair-canceling ensembles demonstrates that the effect of pair cancellation depends on its compatibility with the hopping hierarchy The TM recursion guarantees that the two binary signs within each intracell pair are opposite, isolating the shortest-scale component of the TM organization The results establish that bond-aligned short-range sign anticorrelation is a model-specific mechanism in the present TM masked SSH chain for reentrant localization The effective-dimer picture explains why opposite signs reduce the differential onsite detuning across the strong bonds and favor local hybridization over an intermediate modulation range Longer-range Thue–Morse correlations modify the position and strength of this response, while bond-aligned short-range anticorrelation provides a route to reentrant localization The sequence-comparison calculations clarify the structural origin of the reentrant response The results contribute to the emerging picture that quasiperiodic and aperiodic order can host localization phenomena qualitatively distinct from those in both periodic crystals and fully random media Page 1 --- Reentrant Localization Transition in a Quasiperiodic Thue-Morse Chain We investigate single-particle localization in a dimerized Su–Schrieffer–Heeger (SSH) chain with a quasiperiodic onsite potential masked by the deterministic Thue–Morse sequence Using exact diagonalization, we map the localization behavior as a function of the quasiperiodic potential strength and hopping dimerization through the correlation dimension, inverse participation ratio, and normalized participation ratio For appropriate hopping ratios, increasing the potential strength drives the mid-spectrum states through a localized–multifractal–localized sequence, producing a reentrant recovery of participation before localization is restored at stronger modulation Energy-resolved diagnostics show that this recovery is concentrated in the central spectral region rather than occurring uniformly throughout the spectrum Extrapolations of the generalized dimensions to the thermodynamic limit reveal a systematic moment-dependent hierarchy in the reentrant window, accompanied by a broadened thermodynamic-limit singularity spectrum Real- and momentum-space diagnostics provide complementary evidence for multifractal scaling in this regime, while two-size crossings yield finite-size estimates of the reentrant boundaries Comparisons with random, globally balanced, pair-canceling, and block-permuted masks show that short-range sign anticorrelation promotes reentrance when it is aligned with the dominant SSH hopping bonds The resulting reduction of the onsite mismatch across the strong bonds provides an effective-dimer interpretation of the reentrant response, while longer-range Thue–Morse correlations modify its location and strength Pruning redundant reasoning tokens The TM-masked SSH model provides an analytically transparent setting in which reentrant localization can be tracked with high resolution The correlation dimension, IPR, and NPR identify extended, localized, and intermediate regimes and reveal a reentrant window in which the mid-spectrum states recover participation after an initial localization tendency Full-spectrum and energy-resolved diagnostics further determine how this behavior is distributed across the spectrum and show that the reentrant response is concentrated primarily in its central region How the training signal is built The fractal dimension of the ith eigenstate D(i)q as: D(i)q = − lim L→∞ 1 q−1 lnP(i)q lnL (4) The second box moment is then P(i)2(l j;L) = NB(l j) ∑ k=1 h µ(i)k (l j) squared (10) The correlation dimension D(i)q is obtained from the slope of this linear relation over the selected box scales Where the gains come from The reentrant window is identified by a non-monotonic evolution of the mid-band eigenstate with δ: extended at δ = 0, sharply localized near δ = 0.6, partially delocalized (multifractal) around δ ≈ 1.1, and localized again by δ = 3.0 The thermodynamic-limit dimensions exhibit the systematic hierarchy D(∞) 2 > D(∞) 3 > D(∞) 4 > D(∞) 5 in the reentrant window Momentum-space generalized dimensions show that the enhancement of ∆Dk in the intermediate modulation range demonstrates a pronounced q dependence in this region The pair-canceling ensemble retains the reentrant response observed for the TM modulation at this hopping ratio Pruning redundant reasoning tokens The comparison between the mid-spectrum averaged correlation dimension and the composite participation-ratio metric in the (J2,δ) plane shows that the enhanced hD2 region in panel (a) qualitatively overlaps with the intermediate reentrant region identified by η in panel (b) The momentum-space analysis complements the real-space multifractal characterization by showing that states retain a structured distribution over momentum components and exhibit moment-dependent scaling The two-size crossing analysis provides finite-size estimates of the boundaries separating the localized and multifractal regimes Page 2 --- Model and Method A B A B A B We consider a one-dimensional quasiperiodic chain with two sublattices in each cell The tight-binding Hamiltonian follows as: H = −J1 N ∑ n=1 c† n,A cn,B +h.c−J2 N−1 ∑ n=1 c† n,B cn+1,A +h.c+ N ∑ n=1 εn,Ac† n,A cn,A +εnBc† n,B cn,B (1) This is an SSH Hamiltonian with N unit cells Onsite energies for two sublattice εn,A and εn,B read as: εn,A = (−1) S2n−2 δ cos(2πβ(2n−1) +φ1), εn,B = (−1) S2n−1 δ cos(2πβ(2n) +φ2) (3) where δ is the quasiperiodic potential strength, β is the irrational frequency, φ1, 2 are the potential phases and Sn is the nth element of the Thue-Morse sequence We choose β = (√ 5−1)/2, the inverse golden ratio, as the incommensurate frequency of the potential We fix the intracell hopping J1 = 1 and set φ1,2 = 0 without loss of generality Pruning redundant reasoning tokens The correlation dimension D(i)q as: D(i)q = − lim L→∞ 1 q−1 lnP(i)q lnL (4) The second box moment is then P(i)2(l j;L) = NB(l j) ∑ k=1 h µ(i)k (l j) squared (10) Where the gains come from For a chain of total length L, we select a family of box lengths l j Within the scaling region, the second moment follows P(i)2(l j;L) ∝ l j D(i)2/L (11) and therefore lnP(i)2(l j;L) = D(i)2ln l j/L + ci (12) Pruning redundant reasoning tokens The correlation dimension D(i)q is obtained from the slope of this linear relation over the selected box scales Page 3 --- Fractal Dimension Analysis In Fig.

Kai: 2 we find that the wavefunction exhibits a non-monotonic evolution of the mid-band eigenstate with δ: extended at δ = 0, sharply localized near δ = 0.6, partially delocalized (multifractal) around δ ≈ 1.1, and localized again by δ = 3.0 We will investigate this localized → delocalized → localized sequence with the correlation fractal dimension Where the gains come from The second box moment is then P(i)q (l j;L) = NB(l j) ∑ k=1 h µ(i)k iq (15) Within the scaling region, lnP(i)q (l j;L) = (q−1)D(i)q (L,δ)ln l j/L +const., so that D(i)q (L,δ) is obtained from the corresponding logarithmic slope Pruning redundant reasoning tokens The generalized dimensions are first averaged over the complete sector at each finite size and are subsequently extrapolated according to Dq(L,δ) = D(∞)q(δ) + a q(δ)L (17) The uncertainties are evaluated using the same contiguous-block jackknife procedure employed for D2 Page 4 --- Fractal Dimension Analysis In Fig.

Mira: 3 (a), we show the density plot of the correlation dimension D2 as a function of disorder strength δ for every eigenstate as a function of disorder strength δ We see that for very small values of δ, nearly all states have D2 ≈ 1, indicating Bloch-like delocalization similar to that of the clean SSH chain Slightly increasing the potential strength, we see that a tongue of reduced D2 (green) appears first in the mid-spectrum, indicating that these states are neither fully extended nor localized but exhibit nontrivial power-law scaling of their inverse participation ratios Near δ ≈ 0.8, immediately beyond the first critical window, there is a slight increase in D2 (lightening), signaling a transient delocalization before the final localization transition At large disorder, the entire spectrum collapses to D2 ≈ 0, as every eigenstate becomes exponentially localized by the strong quasiperiodic Thue–Morse potential Where the gains come from The finite-size curves and thermodynamic estimate are presented in Fig.

Kai: 3(b) All finite-N curves display a pronounced dip–peak (“reentrant”) feature near δ ≈ 0.7–1.2 before decaying to zero at strong disorder As N increases, the height of the reentrant peak diminishes, its maximum shifts slightly toward lower δ, and the N → ∞ estimate lies below all finite-size curves in this window Pruning redundant reasoning tokens The resulting exponent ratios γ/ν characterize the effective participation scaling of the corresponding spectral sectors, while additional large-system calculations test the stability of the nonmonotonic behavior The resulting exponent ratios, γ/ν ≃ 0.02–0.04, indicate a very weak dependence of the band-averaged participation ratio on system size Where the gains come from The small slopes indicate that PR depends only weakly on system size at the finite-size crossings and are broadly consistent with the small crossing ordinates obtained from the two-size analysis The resulting exponent ratios, γ/ν ≃ 0.02–0.04, indicate a very weak dependence of the band-averaged participation ratio on system size Pruning redundant reasoning tokens The small magnitude of the crossing ordinates accounts for the modest difference between the estimates obtained from the direct logarithmic slopes and from the crossing ordinates, since an almost size-independent participation ratio makes the crossing position more sensitive to finite-size drift Where the gains come from The resulting exponent ratios, γ/ν ≃ 0.02–0.

Title and authors: Kai: Now, just to set the stage for what we're hearing about in this paper, there's the title itself: "Reentrant Localization Length in a Quasiperiodic Thue-Morse Chain." It sounds very specific.

Mira: It’s specific because it zeroes in on that exact type of quasiperiodic potential—the Thue-Morse sequence—and it focuses on how the localization length changes as you dial up the potential strength.

Lev: What this means for us is that we're looking at a system where the structure itself isn't simple, but its complexity dictates whether particles stay stuck or spread out.

The paper's summary: Kai: So, what they’re doing here is using exact diagonalization on a dimerized Su–Schrieffer–Heeger chain with this Thue-Morse sequence masking the onsite potential. They map how localization behaves as you change the potential strength and the hopping dimerization using tools like correlation dimension, inverse participation ratio, and normalized participation ratio.

Mira: They find that when you increase the potential strength for certain hopping ratios, it drives those mid-spectrum states through a sequence: localized to multifractal to localized again. That's the reentrant recovery of participation they track.

Lev: They also look at energy-resolved diagnostics and find that this recovery isn't happening everywhere; it’s mostly concentrated in the central spectral region, which is a key detail for how we might try to engineer these states on real hardware.

The paper's improvements: Kai: The authors also look at how they can improve the analysis by going into the thermodynamic limit and looking at generalized dimensions. They show a systematic hierarchy in that reentrant window, with D2 greater than D3, D3 greater than D4, and so on.

Mira: That hierarchy is important because it tells us different moments are probing different parts of the structure within that intermediate regime where things recover participation.

Lev: And they use momentum-space diagnostics to complement the real-space analysis, showing a pronounced q dependence in that intermediate modulation range, which adds another layer to characterizing this multifractal behavior.

Conclusion: Kai: So, wrapping up on the main findings of "Reentrant Localization Length in a Quasiperiodic Thue-Morse Chain," it establishes that bond-aligned short-range sign anticorrelation is the mechanism causing this reentrant localization in this specific system.

Mira: The effective-dimer picture explains why those opposite signs reduce the onsite mismatch across the strong bonds and favor local hybridization over an intermediate modulation range.

Lev: This means longer-range Thue-Morse correlations change where that response happens, but it's that short-range sign anticorrelation across dominant bonds that drives the recovery.

Kai: What this suggests is that even in these quasiperiodic systems, we can see localization phenomena qualitatively different from what you’d expect in a clean crystal or a fully random system.

Mira: Exactly. This work contributes to the picture that quasiperiodic and aperiodic order can host localization behaviors that are distinct from both periodic crystals and fully random media.

Lev: From an error correction viewpoint, understanding this mechanism allows us to predict what kind of local structure we need to look for when trying to design error correcting codes on these complex lattices.

Kai: We'll see more work coming in on how these structural properties translate directly into observable physical parameters in the actual experiments.

More episodes

← Home