betaFeedback

WorksPhysics and Astronomy

Morphology-resolved scrambling in a chaotic quantum billiard

Abstract

Chaotic quantum systems can retain spatial memory through scarred eigenstates, but whether these static structures control scrambling remains unclear. This work establishes a morphology-resolved connection between scarred eigenstates and eigenstate-resolved OTOCs in a peanut-shaped quantum billiard. Scalar localisation diagnostics, including differential entropy and continuum participation ratios, detect anomalous concentration but discard spatial architecture. A scale-normalised density overlap, in contrast, directly compares probability density profiles, revealing families of orthogonal eigenstates with nearly identical spatial morphology. Comparing the complete OTOC time traces of these orthogonal eigenstates reveals that morphological recurrence has dynamical content: moderate density overlap yields no universal prediction, whereas strongly recurring morphologies exhibit nearly identical OTOC growth and saturation. Thus, scarred structures act as spatial templates for operator growth, not merely static violations of ergodicity. This morphology-resolved framework turns eigenstate shape into a quantitative predictor of scrambling and provides a scale-controlled diagnostic of weak ergodicity breaking in quantum chaos.

Introduction

Chaotic quantum systems do not always erase memory of their initial conditions; instead, a sparse set of nonthermal eigenstates can emerge throughout the spectrum. A central challenge is therefore not only to identify the origin of these nonthermal structures, but also to classify the recurring morphologies and determine whether these static structures leave measurable signatures in the subsequent dynamics.

The Eigenstate Thermalisation Hypothesis (ETH) suggests that individual eigenstates of nonintegrable many-body systems should behave thermally[1, 2]. This implies that local observables will equilibrate and lose memory of their initial conditions. In single-particle chaotic billiards, the analogous expectation appears through quantum ergodicity and random-wave statistics[3, 4]. However, significant deviations from this prediction are evident. Quantum scars are a prime example of such deviations. These are anomalous nonthermal eigenstates whose probability densities remain enhanced near slowly diverging classical trajectories[5, 6, 7], leading to weak ergodicity breaking across a predominantly chaotic spectrum.

Scarring, however, is not just a question of localisation strength. Quantifying their existence, recurrence across the spectrum, and determining their dynamical consequences remains nontrivial. Standard scar diagnostics, including overlaps with selected classical trajectories, Husimi projections, participation ratios, entropy measures, and semiclassical constructions, have been highly successful. However, they often rely on prior knowledge of the relevant orbit, phase-space representation, or system geometry. On top of these dependencies, they do not determine whether different eigenstates share the same spatial morphology, nor whether such morphological similarity leads to similar scrambling dynamics. The Shannon differential entropy[8, 9] and the Inverse Participation Ratio[10] (IPR), both members of the Rényi entropy hierarchy, are widely used to quantify how anomalously localised a continuum PDF ρ(x)=∣ψ(x)∣2\rho(x)=|\psi(x)|^2 is[8, 9, 10]. These quantities must nevertheless be interpreted with care, since differential entropy depends on the choice of coordinates and length units, whereas the continuum IPR is dimensional and sensitive to normalisation, domain size, and spatial resolution. These measures are therefore useful indicators of concentration alone.

To overcome these limitations, we use scale-normalised and relative diagnostics. In particular, we compare the probability density of each eigenstate with a carefully constructed uniform probability density distribution through relative measures such as the Kullback-Leibler (KL) divergence and the normalised Partition Ratio (nPR). Then we introduce a pair-wise normalised density overlap that directly compares the full probability density profiles of distinct eigenstates. This allows eigenstates to be grouped into morphology classes (G\mathcal{G}) without assuming in advance which classical orbit, if any, is responsible for the structure.

To test whether these static classes have dynamical significance, we compute eigenstate-resolved Out-of-Time-Order Correlators (OTOCs)[5, 11, 12, 13, 14, 15], which quantify the growth of operator noncommutativity and the scrambling of initially local information. For each pair of eigenstates, we compare the complete OTOC time traces through a normalised curve distance and its associated similarity. This directly asks whether near-identical density morphologies lead to near-identical scrambling responses.

We apply this framework to the peanut-shaped quantum billiard[5], whose boundary (Ωpeanut\Omega_{\mathrm{peanut}}) is defined by (x2+y2)2−2(x2−y2)+1−1.3752=0(x^2+y^2)^2-2(x^2-y^2)+1-1.375^2=0. Its mixed-curvature geometry supports unstable periodic orbits embedded in a predominantly chaotic phase space, making it a minimal two-dimensional fertile ground for exploring scar formation. Representative scarred eigenstates and their associated slowly diverging classical trajectories are shown in Figure 1. In this article, we show that morphology-resolved density overlap identifies recurring scar families across the spectrum and that these families exhibit closely related OTOC dynamics. Thus, scar morphology is not merely a static feature of anomalous eigenfunction; it provides a predictive link between nonthermal eigenstate structure and quantum information scrambling.

Classical trajectories and eigenstate intensities in the peanut-shaped billiard

Figure 1. Classical trajectories and eigenstate intensities in the peanut-shaped billiard. Top: coloured curves mark selected slowly diverging orbits, while the white dashed curve indicates chaotic motion. Bottom: eigenstate intensities ∣ψ(x)∣2|\psi(\mathbf{x})|^2 show enhanced probability density along the corresponding classical structures. The colour coding connects each eigenstate morphology to the corresponding classical structure in the upper panel.

Relative entropy— For a normalised eigenstate ψn(x)\psi_n(\mathbf{x}) on the billiard domain ωpeanut\omega_{\mathrm{peanut}}, the spatial spread of ρn(x)\rho_n(\mathbf{x}) is provided by the Rényi entropy, a one-parameter family of measures, defined as

Hα(ρn(x)):=log⁡(∫Ωρn(x)α dx)1−α,α∈(0,1)∪(1,∞)(1)H_{\alpha}\left(\rho_n(\mathbf{x})\right) := \frac{\log\left(\int_{\Omega} \rho_n(\mathbf{x})^{\alpha}\,\mathrm{d}x\right)}{1-\alpha}, \qquad\alpha\in(0,1)\cup(1,\infty) \tag*{(1)}

The parameter α\alpha controls which part of the density is emphasised: α>1\alpha> 1 gives stronger weight to high-density regions, while α<1\alpha< 1 is more sensitive to low-density support. In the limit α→1\alpha\to1, Eq. (1) gives the differential Shannon entropy:

HS=−∫Ωρn(x)log⁡(ρn(x)) dx(2)H_S = -\int_{\Omega} \rho_n(\mathbf{x})\log\left(\rho_n(\mathbf{x})\right)\,\mathrm{d}x \tag*{(2)}

Large HSH_S indicates a spatially extended state, whereas a reduced value signals a localised state.

A related localisation measure is the IPR[10],

IPR=∫Ωρn(x)2 dx(3)\mathrm{IPR} = \int_{\Omega} \rho_n(\mathbf{x})^2\,\mathrm{d}x \tag*{(3)}

It is the second moment of the probability density and corresponds to the Rényi entropy at α=2\alpha= 2 in Eq. (1) through H2=−log⁡(IPR)H_2=-\log(\mathrm{IPR}). Thus, sharply localised eigenstates have large IPR, whereas extended states have small IPR that scales inversely with the effective volume (or area) of the system. Intuitively, the inverse of the IPR can serve as a natural measure of delocalisation.

Both HSH_S and IPR are logical extensions of their discrete counterparts, given that the probability distributions are continuous. Unlike Wehrl entropy[16, 17], which relies on the Husimi QQ function to ensure non-negativity, differential entropy directly employs the probability density function, with no built-in smoothing, making it sensitive to the units and coordinate system used and hence is not invariant under transformations[18]. In particular, HSH_S can assume both positive or negative values[19, 20, 21, 22] because continuous PDF (ρ(x)\rho(\mathbf{x})) is not restricted to the values less than unity[23, 24]. In a similar vein, for continuous probability distributions, if ψ(x)\psi(\mathbf{x}) is sharply peaked, then ∣ψ(x)∣4|\psi(\mathbf{x})|^4 becomes large, and the resulting IPR value may surpass unity[25]. These limitations motivate the use of normalised or relative measures that remove dependence on units and allow meaningful comparisons across different states, systems, and geometries.

One such relative measure is the Kullback-Leibler (KL) divergence[26] between ρn(x)\rho_n(\mathbf{x}) and a reference density ρ0(x)\rho_0(\mathbf{x}),

DKL(ρn∥ρ0):=∫Ωρn(x)log⁡(ρn(x)ρ0(x)) dx=∫Ωρn(x)log⁡(A×ρn(x)) dx(4)D_{\mathrm{KL}}(\rho_n\Vert\rho_0) := \int_{\Omega}\rho_n(\mathbf{x})\log\left(\frac{\rho_n(\mathbf{x})}{\rho_0(\mathbf{x})}\right)\,\mathrm{d}x = \int_{\Omega}\rho_n(\mathbf{x})\log\left(\mathcal{A}\times\rho_n(\mathbf{x})\right)\,\mathrm{d}x \tag*{(4)}

where ρ0(=1A)\rho_0\left(=\frac{1}{\mathcal{A}}\right) is the referenced uniform density distribution over a bounded two-dimensional domain Ω\Omega of area A(=∫Ωdx)\mathcal{A}\left(=\int_{\Omega}\mathrm{d}x\right). Thus, DKL(ρn∥ρ0)D_{\mathrm{KL}}(\rho_n\Vert\rho_0) measures the deviation of an eigenstate from the uniform distribution. It vanishes for a perfectly uniform density and increases as the probability weight becomes more spatially concentrated in a sub-region.

We also use the normalised Participation Ratio (nPR),

nPR=ρ0IPR=1A×IPR∈[0,1](5)\mathrm{nPR}=\frac{\rho_0}{\mathrm{IPR}}=\frac{1}{\mathcal{A}\times\mathrm{IPR}}\in[0,1] \tag*{(5)}

Here, nPR is the exponential of Rényi entropy of order 2, normalised by the domain size A\mathcal{A}. The value nPR =1=1 corresponds to a uniform density over the entire billiard domain, while smaller values indicate that the eigenstate effectively occupies only a fraction of the available area. Unlike the raw IPR, nPR is dimensionless and directly comparable across different system sizes.

FIG. 2 summarises these diagnostics across the spectrum. The raw quantities HSH_S and IPR, in Figure 2(a & b), identify states with enhanced spatial concentration, including scarred or partially localised eigenstates. However, their raw absolute values are scale-dependent. For example, for a strongly concentrated eigenstate such as for n=483n = 483, HSH_S acquires a negative value, and IPR exceeds unity for the same state. These features do not invalidate the localisation trend, but they make absolute comparisons across states less transparent and create interpretational difficulties.

Eigenstate morphology and scar diagnostics across the spectrum

Figure 2. Top: representative intensity profiles ∣ψ∣2|\psi|^2, with colours matching the highlighted data points. Bottom: spectral variation of (a) Differential Shannon entropy HSH_S, (b) IPR, (c) DKLD_{\mathrm{KL}}, and (d) nPR versus eigenstate index nn. Insets display a few selected eigenstate probability densities. While these scalar diagnostics reveal anomalous concentrations and deviations from ergodic spatial filling, they collapse each eigenstate to a single number and therefore cannot resolve whether distinct eigenstates share the same spatial morphology.

Figure 2(c & Figure 2d) present dimensionless diagnostics that remove this ambiguity by using dimensionless, scale-independent measures. The KL divergence highlights deviations from the uniform density, so localised or scarred eigenstates appear as large positive outliers. The nPR rescales the participation ratio by the billiard area and therefore measures the fraction of the domain effectively occupied by the eigenstate. Scarred or localised states have comparatively large HSH_S and small nPR, while extended states show the opposite trend. Taken together, these measures separate genuine spatial concentration from unit-dependent artefacts.

Pairwise density overlap and morphology classes—Localisation strength alone, however, does not determine whether two scarred states have the same spatial structure. To compare morphology directly, we introduce the L2L_2-normalised2 density overlap

S(n,m)=∫Ω(ρn(x)×ρm(x)) dx(∫Ωρn2(x) dx)×(∫Ωρm2(x) dx)(6)\mathcal{S}(n,m)=\frac{\int_{\Omega}\left(\rho_n(\mathbf{x})\times\rho_m(\mathbf{x})\right)\,\mathrm{d}x}{\sqrt{\left(\int_{\Omega}\rho_n^2(\mathbf{x})\,\mathrm{d}x\right)\times\left(\int_{\Omega}\rho_m^2(\mathbf{x})\,\mathrm{d}x\right)}} \tag*{(6)}

By construction, 0≤S(n,m)≤10\leq\mathcal{S}(n,m)\leq1, with S(n,n)=1\mathcal{S}(n,n)=1 (diagonal elements in the matrix plot reflect the trivial self-similarity as shown in Figure 3(a)). Thus, S(n,m)≃1\mathcal{S}(n,m)\simeq1 implies

Density-overlap similarity between eigenstates

Figure 3. Density-overlap similarity between eigenstates. (a) Thresholded matrix S(n,m)≥0.61\mathcal{S}(n,m) \ge0.61, where off-diagonal elements near unity identify distinct eigenstates with nearly identical probability densities. (b) Network of spatially similar eigenstates, with edges drawn for S(n,m)≥0.67\mathcal{S}(n,m) \ge0.67; edge colour and thickness encode S(n,m)\mathcal{S}(n,m). The balloon layout is only visual. (c) Eigenstates most similar to the reference state n=71n=71; the red line marks Sth=0.61\mathcal{S}_{\mathrm{th}}=0.61.

∥ρ~n−ρ~m∥22=2(1−S(n,m))≪1.(7)\left\lVert\widetilde{\rho}_n-\widetilde{\rho}_m\right\rVert_2^2=2(1-\mathcal{S}(n,m))\ll1. \tag*{(7)}

That means the two eigenstates have almost identical spatial intensity fingerprints. They place their probability weight in nearly the same sub-regions of the billiard, whereas S(n,m)≪1\mathcal{S}(n,m)\ll1 indicates that the two states occupy different sub-regions or have strongly different intensity patterns.

The numerator in Eq. (6) is a cross-participation or an intensity-overlap integral. For n=mn=m, it reduces to IPR in Eq. (3). Therefore, S(n,m)\mathcal{S}(n,m) can be interpreted as a normalised cross-participation ratio: it measures the degree to which two eigenstates concentrate their probability weight in the same spatial regions, after removing the trivial effect of different individual localisation strengths. This is distinct from the Hilbert-space overlap ⟨n∣m⟩\langle n|m\rangle, which vanishes for different eigenstates of a Hermitian Hamiltonian. Two orthogonal eigenfunctions can therefore have ⟨n∣m⟩=0\langle n|m\rangle=0, but still display a large density overlap for S(n,m)≈1\mathcal{S}(n,m)\approx1.

A coarse-grained interpretation makes this connection transparent. Let

P^a=∫Ωa∣x⟩⟨x∣ dx,(8)\hat{P}_a=\int_{\Omega_a}|\mathbf{x}\rangle\langle\mathbf{x}|\,\mathrm{d}x, \tag*{(8)}
pna=⟨n∣P^a∣n⟩=∫Ωaρn(x) dx,(9)p_{na}=\langle n|\hat{P}_a|n\rangle=\int_{\Omega_a}\rho_n(\mathbf{x})\,\mathrm{d}x, \tag*{(9)}

where Ωa\Omega_a is a coarse-graining of the billiard. The vector pn=(pn1,pn2,…)\mathbf{p}_n=(p_{n1},p_{n2},\ldots) gives a discrete representation of the local spatial weight of the eigenstate nn such that ∑apna=1\sum_a p_{na}=1. In this representation, S(n,m)=pn⋅pm∥pn∥∥pm∥\mathcal{S}(n,m)=\frac{\mathbf{p}_n\cdot\mathbf{p}_m}{\lVert\mathbf{p}_n\rVert\lVert\mathbf{p}_m\rVert}. In the continuum limit, this reduces to Eq. (6). So, S(n,m)\mathcal{S}(n,m) is nothing but the cosine similarity2 between two density vectors. Hence, S(n,m)\mathcal{S}(n,m) does not compare the amplitudes of ψn\psi_n and ψm\psi_m directly; it compares their observable spatial probability distributions.

Hence, S(n,m)\mathcal{S}(n,m) is a configuration-space diagnostic. It is insensitive to the overall phase or sign of an eigenfunction, which is useful for comparing standing-wave patterns. It does not resolve momentum direction or phase-space structure.

Figure 3 shows that eigenstate similarity is not distributed randomly across the spectrum but organises into structured families. The thresholded density-overlap matrix, S(n,m)≥0.61\mathcal{S}(n,m)\geq0.61, in Figure 3(a) shows that similarity is not restricted to the diagonal (n=mn=m), but appears in structured off-diagonal bands and isolated high-overlap pairs, indicating distinct eigenstates with comparable spatial intensity profiles. In Figure 3(b), recasts these high-overlap pairs as a network, where vertices denote eigenstates and edges connect states with δth=0.67\delta_{th}=0.67, a stricter threshold chosen for graph clarity. The connected components identify families of eigenstates with similar density morphology, while edge colour and width encode the magnitude of S(n,m)\mathcal{S}(n,m). Finally, Figure 3(c) shows the similarity profile relative to n0=71n_0=71, identifying states whose densities closely resemble the reference state. Together, Figure 3 reveals recurring spatial structures embedded in the eigenstate sequence. This morphological information is dynamically relevant because local operators probe the regions where an eigenstate carries probability weight.

Scrambling dynamics and scarred multiplets—Comparing eigenstate densities with eigenstate-resolved (microncanonical) OTOCs provides a direct test of whether static morphology has dynamical consequences[27]. For local Hermitian operators W^\hat{W} and V^\hat{V} the squared commutator

Cn(t):=⟨n∣[W^(t),V^]†[W^(t),V^]∣n⟩(10)C_n(t):=\langle n|[\hat{W}(t),\hat{V}]^\dagger[\hat{W}(t),\hat{V}]|n\rangle \tag*{(10)}

can be written as a combination of time-ordered and out-of-time-ordered 4-point correlators, which depend on products of local matrix elements. In the Heisenberg picture,

⟨a∣W^(t)∣b⟩=eitℏ(Ea−Eb)W^ab,W^ab=⟨a∣W^∣b⟩,(11)\langle a|\hat{W}(t)|b\rangle=e^{\frac{it}{\hbar}(\mathcal{E}_a-\mathcal{E}_b)}\hat{W}_{ab},\qquad\hat{W}_{ab}=\langle a|\hat{W}|b\rangle, \tag*{(11)}

the 4-point OTOC has the energy-basis representation

Cn(t)=∑a,b,ceitℏ(En−Ea+Eb−Ec)W^naV^abW^bcV^cn(12)C_n(t)=\sum_{a,b,c}e^{\frac{it}{\hbar}(\mathcal{E}_n-\mathcal{E}_a+\mathcal{E}_b-\mathcal{E}_c)}\hat{W}_{na}\hat{V}_{ab}\hat{W}_{bc}\hat{V}_{cn} \tag*{(12)}

This expression shows that eigenstate-resolved OTOCs are governed by three ingredients: energy gaps, which generate oscillatory phases; operator matrix elements, which determine the couplings induced by W^\hat{W} and V^\hat{V}; and eigenfunction structure, which controls the spatial organisation of those matrix elements.

For the Hermitian pair {x^(t),p^}\{\hat{x}(t),\hat{p}\}, the rapid growth of Cn(t)C_n(t) measures the increasing noncommutativity. In a semiclassical chaotic system, its early-time growth reflects local phase-space stretching and can scale as Cn(t)∼e2λLtC_n(t)\sim e^{2\lambda_L t} before finite-size saturation, where λL\lambda_L is the quantum Lyapunov exponent. Thus, eigenstates in the same high-S(n,m)\mathcal{S}(n,m) morphology class are sampled by x^(t)\hat{x}(t) and p^\hat{p} over nearly the same regions of the billiard. Their local matrix-element products are therefore correlated, producing similar short-to-intermediate-time OTOC dynamics. In this sense, recurring spatial morphology constrains how local information is scrambled.

This provides the physical basis for morphology-resolved scrambling. We first identify morphology classes Gg\mathcal{G}_g as groups of eigenstates connected by large S(n,m)\mathcal{S}(n,m). Its representative scrambling profile is quantified by the class-averaged OTOC Cgavg(t)(=1∣Gg∣∑n∈GgCn(t))C_g^{\mathrm{avg}}(t)\left(=\frac{1}{|\mathcal{G}_g|}\sum_{n\in\mathcal{G}_g}C_n(t)\right). As shown in Figure 4, classes with similar intensity morphology display comparable OTOC growth and saturation behaviour. Thus recurring spatial structures are not merely static features of the eigenfunctions, but also organise the dynamical response.

Morphology-resolved OTOC dynamics

Figure 4. Morphology-resolved OTOC dynamics. (a-e) show eigenstate groups Gg\mathcal{G}_g with similar density morphology, identified by the overlap S(n,m)\mathcal{S}(n,m). Grey curves are individual OTOCs and coloured curves are group averages Cgavg(t)C_g^{\mathrm{avg}}(t). Density thumbnails show representative members from the corresponding morphology class. (f) compares averaged curves across all classes, demonstrating that recurring eigenstate morphologies yield distinct yet internally consistent OTOC growth and saturation profiles.

To quantify this connection pair-wise, we compare the complete OTOC time traces through the normalised

OTOC distance between two eigenstates nn and mm:

dc(n,m)=∫0t∣Cn(t)−Cm(t)∣2 dt∫0t∣Cn(t)∣2 dt+∫0t∣Cm(t)∣2 dt(13)d_c(n,m)=\frac{\sqrt{\int_0^t\left|C_n(t)-C_m(t)\right|^2\,\mathrm{d}t}}{\sqrt{\int_0^t\left|C_n(t)\right|^2\,\mathrm{d}t}+\sqrt{\int_0^t\left|C_m(t)\right|^2\,\mathrm{d}t}} \tag*{(13)}

For uniformly sampled data, the integrals are replaced by sums with time step Δt\Delta t. This distance compares the full-time traces directly, without fitting an exponential growth rate or extracting a single characteristic time. Therefore, it is sensitive to differences in early growth, oscillations, saturation, and revivals.

By the triangle inequality, 0≤dc(n,m)≤10 \le d_c(n,m) \le1, provided the denominator is nonzero. Thus, dc(n,m)≃0d_c(n,m) \simeq0 implies Cn(t)≃Cm(t)C_n(t) \simeq C_m(t), i.e. nearly identical OTOC profiles, while dc≃1d_c \simeq1 indicates strongly different profiles. It is then useful to define an OTOC-curve similarity: Qc(n,m)=1−dc(n,m)Q_c(n,m) = 1 - d_c(n,m), so that Qc(n,m)≈1Q_c(n,m) \approx1 corresponds to nearly identical scrambling dynamics. In contrast, a small value corresponds to dissimilar dynamics.

The scatter and binned analyses in Figure 5 show that OTOC similarity is threshold-like rather than a linear function of density overlap. While S(n,m)\mathcal{S}(n,m) compares the static intensity profiles of PDFs, Qc(n,m)Q_c(n,m) compares the complete time-dependent response of OTOC. In Figure 5(a), pairs close to the density-overlap cut-off S(n,m)>0.61\mathcal{S}(n,m) > 0.61 display a broad range of QcQ_c, showing that marginal spatial similarity alone is not sufficient to determine the OTOC response. In contrast, the largest-S\mathcal{S} pairs are strongly biased toward high QcQ_c and low dcd_c. The binned median in Figure 5(b) confirms this trend: increasing density-overlap similarity systematically enhances the median OTOC similarity and suppresses the median OTOC distance. Therefore, strongly overlapping eigenstate morphologies scramble information in nearly the same manner. In comparison, the residual spread arises from spectral phases, phase-space structure, symmetry constraints, and local matrix-element interference not captured by S(n,m)\mathcal{S}(n,m) alone.

Pair-wise relation between spatial morphology and OTOC

Figure 5. Pair-wise relation between spatial morphology and OTOC. (a) OTOC-curve similarity Qc(n,m)Q_c(n,m) as a function of S(n,m)\mathcal{S}(n,m) for high-overlap pairs satisfying S(n,m)≥0.61\mathcal{S}(n,m) \ge0.61. The inset shows the complementary dc(n,m)d_c(n,m) vs S(n,m)\mathcal{S}(n,m). Pairs near the cut-off show broad dynamical variation, whereas the largest-S\mathcal{S} pairs are biased toward high QcQ_c and low dcd_c. (b) Binned median OTOC similarity as a function of S(n,m)\mathcal{S}(n,m). The systematic increase of median QcQ_c, together with the decrease of median dcd_c shown in the inset, demonstrates that strongly overlapping spatial-density profiles lead to nearly identical OTOC time traces.

Conclusion— In summary, we have shown that recurring eigenstate morphology provides a direct bridge between static nonthermal structure and dynamical scrambling in a chaotic quantum billiard. While entropy, participation ratios, and related scalar diagnostics identify anomalous localisation, they do not determine whether distinct eigenstates share the same spatial architecture. The normalised density overlap (S(n,m)\mathcal{S}(n,m)) resolves this missing structural information by grouping orthogonal eigenstates into morphology classes with a common intensity pattern. Comparing these classes with eigenstate-resolved OTOCs reveals that morphology has measurable dynamical consequences: strongly overlapping density profiles exhibit enhanced OTOC similarity (Qc(n,m)Q_c(n,m)) and suppressed curve distance (dc(n,m)d_c(n,m)). The relation is threshold-like rather than purely linear, indicating that moderate spatial overlap is insufficient, whereas near-identical morphology strongly constrains short-to-intermediate-time operator growth.

Our results show that scars are not merely static anomalies of eigenfunction intensity; when their morphologies recur, they also influence how quantum information scrambles. In conclusion, spatial morphology provides a structural constraint on scrambling.

Acknowledgments— Numerical calculations were performed using n2-highmem-32 virtual-machine instances on Google Cloud Platform, supported by Google Cloud research credits.

Data availability— The data supporting this study are available from the corresponding author upon reasonable request and are intended to be deposited in a public repository upon publication.

References

References

  1. [1]M. Rigol, V. Dunjko, and M. Olshanii, Thermalization and its mechanism for generic isolated quantum systems, Nature 452, 854 (2008).
  2. [2]L. D’ALESSIO, Y. Kafri, A. Polkovnikov, and M. Rigol, From quantum chaos and eigenstate thermalization to statistical mechanics and thermodynamics, Advances in Physics 65, 239 (2016).
  3. [3]E. J. Heller, Bound-state eigenfunctions of classically chaotic hamiltonian systems: Scars of periodic orbits, Phys. Rev. Lett. 53, 1515 (1984).DOI
  4. [4]C. J. Turner, A. A. Michailidis, D. A. Abanin, M. Serbyn, and Z. Papić, Weak ergodicity breaking from quantum many-body scars, Nature Physics 14, 745 (2018).DOI
  5. [5]P. Das, T. Patra, and B. Ganguli, Classical and quantum chaos in bean- and peanut-shaped billiards (2026), arXiv:2501.08839 [nlin.CD].arxiv.org/abs/2501.08839
  6. [6]A. Pizzi, Quantum trails and memory effects in the phase space of chaotic quantum systems, Phys. Rev. Lett. 134, 140402 (2025).DOI
  7. [7]A. M. Graf, J. Keski-Rahkonen, M. Xiao, S. Atwood, Z. Lu, S. Chen, and E. J. Heller, Birthmarks: Ergodicity breaking beyond quantum scars (2024), arXiv:2412.02982 [quant-ph].
  8. [8]J. Michalowicz, J. Nichols, and F. Bucholtz, Handbook of Differential Entropy (CRC Press, 2013).DOI
  9. [9]A. Lazo and P. Rathie, On the entropy of continuous probability distributions (corresp.), IEEE Transactions on Information Theory 24, 120 (1978).
  10. [10]A. Daniel, A. Hallam, J.-Y. Desaules, A. Hudomal, G.-X. Su, J. C. Halimeh, and Z. Papić, Bridging quantum criticality via many-body scarring, Phys. Rev. B 107, 235108 (2023).DOI
  11. [11]A. I. Larkin and Y. N. Ovchinnikov, Quasiclassical Method in the Theory of Superconductivity, Soviet Journal of Experimental and Theoretical Physics 28, 1200 (1969).
  12. [12]K. Hashimoto, K.-B. Huh, K.-Y. Kim, and R. Watanabe, Exponential growth of out-of-time-order correlator without chaos: inverted harmonic oscillator, Journal of High Energy Physics 2020, 68 (2020).DOI
  13. [13]I. García-Mata, R. A. Jalabert, and D. A. Wisniacki, Out-of-time-order correlations and quantum chaos, Scholarpedia 18, 55237 (2023).
  14. [14]P. P. Das and B. Ganguli, Out-of-time-order correlation in perturbed quantum wells, The European Physical Journal D 79, 74 (2025).DOI
  15. [15]G. Styliaris, N. Anand, and P. Zanardi, Information scrambling over bipartitions: Equilibration, entropy production, and typicality, Phys. Rev. Lett. 126, 030601 (2021).DOI
  16. [16]K. Życzkowski, Localization of eigenstates and mean wehrl entropy, Physica E: Low-dimensional Systems and Nanostructures 9, 583 (2001), proceedings of an International Workshop and Seminar on the Dynamics of Complex Systems.
  17. [17]C. T. Lee, Wehrl’s entropy of spin states and lieb’s conjecture, Journal of Physics A: Mathematical and General 21, 3749 (1988).DOI
  18. [18]M. J. Schroeder, An alternative to entropy in the measurement of information, Entropy 6, 388 (2004).DOI
  19. [19]J. Soch, Differential entropy can be negative — statproofbook.github.io, https://statproofbook.github.io/P/dent-neg, [Accessed 02-09-2025].
  20. [20]T. Cover and J. Thomas, Elements of Information Theory (Wiley, 2012).
  21. [21]C. E. Shannon, A mathematical theory of communication, The Bell system technical journal 27, 379 (1948).
  22. [22]Z. Kala, Global sensitivity analysis based on entropy: From differential entropy to alternative measures, Entropy 23, 10.3390/e23060778 (2021).DOI
  23. [23]D. Griffiths and D. Schroeter, Introduction to Quantum Mechanics (Cambridge University Press, 2018).
  24. [24]G. Casella and R. Berger, Statistical Inference, Chapman & Hall/CRC Texts in Statistical Science (CRC Press, 2024).
  25. [25]T. B. P. Clark and A. Del Maestro, Moments of the inverse participation ratio for the laplacian on finite regular graphs, Journal of Physics A: Mathematical and Theoretical 51, 495003 (2018).
  26. [26]J. M. Joyce, Kullback-leibler divergence, in International Encyclopedia of Statistical Science, edited by M. Lovric (Springer Berlin Heidelberg, Berlin, Heidelberg, 2025) pp. 1307–1309.
  27. [27]J. Maldacena, S. H. Shenker, and D. Stanford, A bound on chaos, Journal of High Energy Physics 2016, 1 (2016).DOI

Paper details

vlnc:W_mst6qk52
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents