WorksPhysics and Astronomy
Microscopic Mechanism and Control of Reentrant Localization
Abstract
Increasing quasiperiodic modulation can restore extended states after localization, raising the question of how the same modulation drives both localization and recovery. We identify a microscopic mechanism in quasiperiodic dimer chains by resolving scattering between Fourier channels. Specifically, scattering into neighboring channels and back shifts the spectral edge relative to a localization boundary determined by a neutral auxiliary resonance. Their successive crossings drive the same spectral packet with fixed integrated-density-of-states (IDS) labels into localization and back into an extended regime. This recovery persists across finite parameter ranges in both gap-opening families, including relative-phase detuning. Furthermore, an exact sublattice mapping enables compensated control of the recovery window and predicts recovery thresholds for individual IDS labels.
Anderson localization shows that disorder can suppress quantum transport [1, 2]. Localization can also arise in quasiperiodic lattices without randomness [3]. For example, the Aubry–André–Harper model has a single modulation threshold [4, 5]; generalized onsite and hopping modulations allow mobility edges and coexistence of extended and localized spectra [6, 7, 9, 10, 11] [8]. Transport therefore depends on which states remain extended and on their fraction of the spectrum [12, 13]. Atomic and photonic platforms probe these distinctions through expansion and state-resolved measurements [14, 15, 16, 17, 18].
Yet the same modulation that localizes can, beyond a further threshold, restore extended states. Roy et al. established reentrance in a hopping-dimerized chain with staggered quasiperiodic modulation [19]. Related recovery phenomena occur in other quasiperiodic lattices [20, 21] and Aubry–André–Fibonacci interpolations [22, 23]. Further work examines critical behavior [24] and longer-range hopping effects [25, 26, 27]. Experiments have also observed reentrant localization in photonic quasicrystals [18] and photonic SSH lattices with random-dimer disorder [28]. Complementary duality-based approaches locate mobility edges and phase boundaries in quasiperiodic chains [29, 30]. Our earlier work identified multiple reentrant transitions in an onsite-staggered chain with a composite potential [31]. This onsite-staggered chain and Roy et al.’s hopping-dimerized chain provide distinct gap-opening routes. Numerical studies of both models emphasized competition between the two-site structure and quasiperiodic modulation. For these dimer chains, however, the remaining questions are why recovery windows open and close, which spectral states recover, and how their recovery can be controlled.
We connect a microscopic explanation of reentrance to an exact control protocol. Fourier-channel scattering shifts the central spectral edge relative to a zero-growth branch set by a neutral auxiliary resonance. Their crossings produce capture and release of a common IDS packet. We map recovery regions as hopping dimerization, onsite staggering, and relative phase vary. An exact mapping between equal and unequal sublattice modulations provides a compensation path that shifts release at fixed recapture and locates recovery thresholds for selected IDS labels.
Model and three-branch growth law. — We consider a quasiperiodic dimer chain with sites and in each unit cell and intracell and intercell hoppings Figure 1(a). Its Hamiltonian is given by
Figure 1. Reentrance at the central-gap edge. (a) Dimer chain with hoppings and relative phase . (b) Gray: 3194-site spectrum; blue/orange: growth-based extended/localized samples. Gray curves: zero-growth contour ; purple dashed: edge ; vertical dashed: recovery ceiling . Markers (i–iv): , with (critical release). (c) Three-branch LE envelopes; dashed: individual branches. (d) IPR scaling over ten sizes ; dashed: . Methods: SM Sec. S4.
where () creates (annihilates) a particle on sublattice in unit cell . The onsite potential combines a sublattice offset with a quasiperiodic modulation,
with amplitude and Diophantine frequency . Here is the initial phase, and the sublattice phase offsets define the relative phase . Energies and modulation strengths are measured in units of . The hopping model has and , opening the central gap through hopping dimerization. By contrast, the onsite model has and , with the gap opened by onsite staggering. The reference examples use and for the hopping model, and and for the onsite model. Both parameter sets use . Derivations and additional numerical checks are given in the Supplemental Material (SM) [32], Secs. S1 and S4.
For classification, the complex-phase dependence distinguishes subcritical and critical regimes that both have vanishing real-phase Lyapunov exponents (LEs). Under the phase continuation , measures exponential transfer growth at energy per two-site cell, with in radians. For this dimer chain, we obtain the three-branch form by combining transfer duality, determinant identities, and analytic cocycle theory [33, 34, 35, 36] (SM Sec. S1). For and real , it reads
where and . The affine branch has extrapolated intercept , which can be negative; its active interval may have zero length. The zero-slope plateau has half-width . Fourier expansion maps the model to a dual chain of coupled momentum channels . Let denote its two largest Lyapunov exponents. The leading exponent then fixes ; the transfer matrix and exact normalization are derived in SM Sec. S1.
At a spectral energy , Avila’s classification [37] identifies the physical transfer cocycle as subcritical when and , critical when , and supercritical when . These regimes are illustrated by the LE profiles in Figure 1(c). For the central-edge representatives in Figure 1(b), finite-size scaling of the inverse participation ratio (IPR) provides complementary real-space evidence [2, 38]. Here for a normalized state on physical sites. The IPR of the recovered state scales as , whereas those of the neighboring localized controls approach nonzero values [Fig. 1(d)]. At the critical contact, the IPR decreases with system size and its finite-size exponents vary across the sampled sizes; the critical classification follows from the LE profile.
Spectral mechanism of reentrant localization.— On the bulk spectrum , the Jacobi spectral criterion excludes a positive LE with a plateau around [36, 37, 39]. Equation (3) therefore gives on the spectrum, and the subcritical population is exactly
Combined with , Avila’s classes reduce to , , and for subcritical, critical, and supercritical spectral energies, respectively.
The outer branch gives an upper modulation bound, : at every real energy for . A fully supercritical spectrum at some , followed by a subcritical spectral energy at , therefore establishes global reentrance of a population with positive IDS weight. Continuity and full support of the density of states make the strict growth signs persist over modulation intervals. The global sufficient criterion is given in SM Sec. S1.3; numerical checks of central-edge recovery are reported in SM Sec. S4.
Both the hopping-dimerized and onsite-staggered [31] reference models realize this sequence at the central-gap edge. We track its upper edge against the selected energy branch of the zero-growth contour that meets the edge at capture and release [Fig. 1(b)]; SM Sec. S4.
Microscopic origin of capture and release.— In the hopping reference, Fourier scattering shifts the edge at fixed IDS label, whereas a neutral dual mode fixes the zero-growth branch . The corresponding onsite auxiliary resonances and contacts are checked in SM Sec. S4.2. Let be the IDS per site [40] and its energy quantile. Spectral symmetry gives , with the right-hand limit.
At , the upper central-gap edge has energy , with state at . Finite modulation couples momentum channels . Each channel carries an amplitude pair on which a Hamiltonian block acts; neighboring channels couple with magnitude . Define the physical Fourier cutoff by retaining , giving a matrix of dimension . Thus keeps only the bare edge channel. For , scattering into the channels on each side and back produces a return self-energy. Combining the two tails by reflection symmetry gives
where is the Green matrix at the first tail channel, obtained by recursive block elimination through [41]. Following the root from gives the estimates in Figure 2(a). Independent variational spectral-distance estimates [42], , decrease monotonically to and provide a reference for (SM Sec. S2).

Figure 2. Capture and release in the hopping model. (a) Edge estimates with physical Fourier channels per side. (b) Resonance estimates with auxiliary dimers per side; : independent bulk references in (a,b). (c) Return mechanism: , with linear overview (upper) and asinh detail (lower). The first channel-return correction to , , already yields capture and release. Circles: refined contacts; upper dashed line: recovery ceiling . (d) Growth-rate check: mode estimate (black), direct dual transfer (dots), and ordered controls (diamonds). Dotted: zero; dashed: channel-isolation threshold . (e) Fixed-packet scaling ; lines/bands: two-phase mean/range. Gray guides in lower (c) and (d,e): bulk contacts. Methods: SM Secs. S4.1 and S4.4.
For , multiplying the dual transfer by shifts its leading growth to . At , an isolated leading channel (, equivalently ) admits an analytic neutral mode. Fourier expansion of this neutral mode yields an auxiliary-chain eigenproblem whose selected resonance determines (SM Sec. S2). The auxiliary sites are Fourier components of the dual mode. Their chain [Figure 2(b)] has alternating hoppings and conjugate onsite terms about its center. The central pair, with onsite terms , , and coupling , gives the bare resonance . Define the auxiliary cutoff as the number of outer strong-bond dimers retained on each side: keeps this central pair, while adds return self-energies and gives in a chain of sites. Within the contact range, the central pair’s decreasing mean energy and increasing real splitting lower the bare resonance and permit capture.
The resulting relative motion of the edge and resonance produces capture and release, measured by [Figure 2(c)]. Here counts physical momentum channels per side, whereas counts dimers in the dual-mode Fourier chain. Below , the bare construction has only one crossing. Including the nearest Fourier-channel returns, , produces a negative edge self-energy and lowers back through , bringing release below already at . By contrast, adding only the first outer auxiliary dimers, , retains a single crossing below this ceiling; these complete crossing sequences are shown in the SM. Further returns at and bring both contacts toward the independent bulk values. The minimal construction explains the crossing sequence; the deeper curves remain finite-return approximations. Recursions, contact values, and onsite-model checks are given in SM Secs. S2 and S4.
The crossing mechanism predicts capture and release at , where the edge growth changes sign. Figure 2(d) tests this connection numerically through
Wherever the mode branch is resolved, the mode estimate agrees with direct dual-transfer products and reproduces the sign changes across the bulk contacts. Independent calculations locate capture and release near [guides in Figure 2(c–Figure 2e)]. Positive/negative identifies supercritical/subcritical edge states; a later return to marks recapture. At , , respectively, confirming a supercritical edge followed by recovery at the two ordered controls.
More generally, for fixed , a positive-weight interval of IDS labels immediately above recovers if and
The lower bound keeps the leading dual mode isolated; the negative sign identifies recovery. Continuity extends the criterion to a common interval of IDS labels , , specifying the same spectral packet at both controls.
To test this bulk prediction in finite chains, Figure 2(e) follows participation scaling in the fixed IDS interval as increases. All states with IDS-midpoint labels in this interval are retained at each size, so the number of sampled states grows with . The exponent compares successive size doublings, where and is the packet average at fixed phase. At the two ordered controls, stays near zero and one over the tested sizes, respectively, consistent with localized and recovered extended states. Wider-packet tests show the same behavior in both models (SM Sec. S4.4).
Reentrant recovery persists over finite parameter ranges for both gap-opening mechanisms, including nonzero relative-phase detuning [Figure 3]. The maps vary (a) hopping dimerization at and (b) onsite staggering at , each against . The regions predicted by Eq. (7) (blue) have boundaries close to finite-chain participation boundaries (dark) for the prescribed packet . On the shared grid, separate real-space diagonalization and participation analysis detect recovery at 98.1% and 96.7% of coordinates satisfying the growth criterion in the hopping and onsite models, respectively. These conditional rates depend on the specified packet, chain sizes, control searches, and participation thresholds. Some discrepancies remain unresolved (SM Sec. S4.3).
Figure 3. Regions of reentrant recovery in (a) hopping and (b) onsite models. Blue: theory prediction from Eq. (7); dark lines: numerical recovery boundaries from a separate real-space diagonalization and participation protocol for the fixed IDS packet; gray: no ordered pair satisfying Eq. (7) detected by the stated search. Here . Participation criteria and finite-size checks: SM Sec. S4.
Exact control of spectral crossings.— The recovery window can be tuned by pinning recapture and shifting release through the relative motion of the spectral edge and the zero-growth boundary . In the onsite-staggered reference, compensated sublattice modulation implements this control, transporting the selected contacts with their IDS labels and zero-growth conditions intact. The hopping counterpart, including window control and IDS-resolved thresholds, is given in SM Sec. S4.5. Specifically, we follow the central edge at , using for compensation and sublattice amplitudes :
where controls the amplitude contrast and recovers the equal-amplitude reference. The fixed amplitude product permits an exact pointwise map to the equal-amplitude chain with the same – hoppings. The map holds at fixed ; we set , hence , in the reference example.
Let denote the chain with potentials Eq. (8). With , the rescaling gives:
Here is the equal-amplitude source, evaluated separately at each test energy because depends on . The map is a hyperbolic rotation in , preserving . At matched parameters, the congruence preserves spectral membership and IDS, and transfer-matrix similarity preserves and . Together, these relations transport the spectral contacts underlying the crossing mechanism from source to target. In particular, fixing a source recapture contact on branch defines the compensation path
Along this path, the central edge and the recapture branch satisfy
The contact therefore stays at while it remains regular, with its energy moving along [Figure 4(a,Figure 4c)]. Control of release follows from the differential motion of and along this compensation path. For the controls in Figure 4(a), increasing raises both energies on the capture–release branch, with rising more strongly. The positive lobe of shrinks, delaying capture and advancing release, while recapture stays pinned. To calculate the moving release endpoint at fixed IDS, let be the source energy of label . The inverse map uniquely determines the matched source staggering from
Figure 4. Compensated recovery in the onsite model. (a) Edge (solid) and neutral boundaries (dashed) for . Circles: contacts; vertical guide: pinned . (b) Corresponding edge growth on an asinh scale. At , only recapture remains. (c) Blue/orange: . Dashed: ; diamond: release intersection. (d) Four-phase meant participation exponents for 8362 16724 sites versus and IDS label . Line/circles: predicted/direct target thresholds. Here , . Methods and endpoint-side tests: SM Secs. S3, S4.5.
At , growth invariance locates release by transporting its source contact. At a simple contact, , with the corresponding branch satisfying and the compensation included in the derivative (SM Sec. S3). For the width , the source has . The full map gives 57.3% widening at and 26.0% narrowing at ; direct contact root finding gives the curves in Figure 4(c). Near , capture and release merge as the positive lobe disappears. Beyond this merger, the recovered band connects to the weak-modulation subcritical regime, with still marking recapture.
The same zero-growth condition predicts recovery for other IDS labels, with the selected spectral energy taking the place of the central edge. At fixed probe modulation , we follow labels in , between the central and next principal gaps. Write for the source growth intercept. For each label , find a source contact satisfying . Equations (10) and (12) then give the recovery threshold :
A finite threshold exists when this ratio is defined and lies in . If each label has one positive-to-negative crossing as increases, . For example, at , all 305 packet labels of the 8362-site periodic approximant cross once over . The threshold distribution spans ; three sampled recovered fractions are reported in SM Sec. S4.5.
Independent real-space diagonalization tests the predicted recovery order and window control. At , we retain all 305 labels of the unequal-amplitude target and scan 81 contrasts with . Fixed IDS groups recover in the predicted order [colors in Figure 4(d)]. For each group, we use the same participation exponent as in Figure 2(e). The overlaid source thresholds and 13 directly located target contacts provide the growth-based reference. Direct target diagonalization also tests both sides of release and recapture at . For the fixed narrow packet , the corresponding changes from localized to extended scaling across release and reverses across recapture (SM Sec. S4.5). These tests use offsets from the predicted endpoints. Full-packet size scaling and endpoint checks up to 262144 sites are given in SM Sec. S4.5.
Conclusion.—Spectral-edge motion relative to a neutral auxiliary resonance explains capture and release of a common IDS packet in quasiperiodic dimer chains. Compensated sublattice modulation controls the recovery window by pinning recapture and shifting release. The exact sublattice mapping also predicts recovery thresholds for individual IDS labels from the zero-growth condition. Direct real-space scaling of the target Hamiltonian tests the predicted recovery order and endpoint
control.
Acknowledgments
Acknowledgments.—This work was supported by the National Natural Science Foundation of China under Grants No. 11974205, No. 61727801, and No. 62131002, and by the Key Research and Development Program of Guangdong Province (2018B030325002).
Data availability. The data supporting this study are available from the authors upon reasonable request.
References
- [1]P. W. Anderson, Absence of diffusion in certain random lattices, Phys. Rev. 109, 1492 (1958).
- [2]F. Evers and A. D. Mirlin, Anderson transitions, Rev. Mod. Phys. 80, 1355 (2008).
- [3]J. B. Sokoloff, Unusual band structure, wave functions and electrical conductance in crystals with incommensurate periodic potentials, Phys. Rep. 126, 189 (1985).DOI
- [4]S. Aubry and G. André, Analyticity breaking and Anderson localization in incommensurate lattices, Ann. Israel Phys. Soc. 3, 133 (1980).
- [5]S. Y. Jitomirskaya, Metal-insulator transition for the almost Mathieu operator, Ann. Math. 150, 1159 (1999).arxiv.org/abs/math/9911265
- [6]S. Das Sarma, S. He, and X. C. Xie, Mobility edge in a model one-dimensional potential, Phys. Rev. Lett. 61, 2144 (1988).pubmed.ncbi.nlm.nih.gov/10038995
- [7]J. Biddle, B. Wang, D. J. Priour, Jr., and S. Das Sarma, Localization in one-dimensional incommensurate lattices beyond the Aubry-André model, Phys. Rev. A J. Biddle and S. Das Sarma, Predicted mobility edges in one-dimensional incommensurate optical lattices: An exactly solvable model of Anderson localization, Phys. Rev. Lett. 104, 070601 (2010).
- [9]S. Ganeshan, J. H. Pixley, and S. Das Sarma, Nearest neighbor tight binding models with an exact mobility edge in one dimension, Phys. Rev. Lett. 114, 146601 (2015).pubmed.ncbi.nlm.nih.gov/25910146
- [10]X. Li, X. Li, and S. Das Sarma, Mobility edges in one-dimensional bichromatic incommensurate potentials, Phys. Rev. B 96, 085119 (2017).DOI
- [11]Y. Wang, X. Xia, L. Zhang, H. Yao, S. Chen, J. You, Q. Zhou, and X.-J. Liu, One-dimensional quasiperiodic mosaic lattice with exact mobility edges, Phys. Rev. Lett. 125, 196604 (2020).DOI
- [12]A. Purkayastha, A. Dhar, and M. Kulkarni, Nonequilibrium phase diagram of a one-dimensional quasiperiodic system with a single-particle mobility edge, Phys. Rev. B 96, 180204 (2017).DOI
- [13]N. Roy and A. Sharma, Fraction of delocalized eigenstates in the long-range Aubry-André-Harper model, Phys. Rev. B 103, 075124 (2021).DOI
- [14]G. Roati, C. D’Errico, L. Fallani, M. Fattori, C. Fort, M. Zaccanti, G. Modugno, M. Modugno, and M. Inguscio, Anderson localization of a non-interacting Bose–Einstein condensate, Nature 453, 895 (2008).
- [15]Y. Lahini, R. Pugatch, F. Pozzi, M. Sorel, R. Morandotti, N. Davidson, and Y. Silberberg, Observation of a localization transition in quasiperiodic photonic lattices, Phys. Rev. Lett. 103, 013901 (2009).pubmed.ncbi.nlm.nih.gov/19659147
- [16]H. P. Lüschen, S. Scherg, T. Kohlert, M. Schreiber, P. Bordia, X. Li, S. Das Sarma, and I. Bloch, Single-particle mobility edge in a one-dimensional quasiperiodic optical lattice, Phys. Rev. Lett. 120, 160404 (2018).
- [17]F. A. An, K. Padavic, E. J. Meier, S. Hegde, S. Ganeshan, J. H. Pixley, S. Vishveshwara, and B. Gadway, Interactions and mobility edges: Observing the generalized Aubry–André model, Phys. Rev. Lett. 126, 040603 (2021).DOI
- [18]S. Vaidya, C. Jörg, K. Linn, M. Goh, and M. C. Rechtsman, Reentrant delocalization transition in one-dimensional photonic quasicrystals, Phys. Rev. Research 5, 033170 (2023).
- [19]S. Roy, T. Mishra, B. Tanatar, and S. Basu, Reentrant localization transition in a quasiperiodic chain, Phys. Rev. Lett. 126, 106803 (2021).DOI
- [20]A. Padhan, M. K. Giri, S. Mondal, and T. Mishra, Emergence of multiple localization transitions in a one-dimensional quasiperiodic lattice, Phys. Rev. B 105, L220201 (2022).DOI
- [21]X. Wei, Reentrant localization transition induced by a composite potential, Phys. Rev. B 111, 134204 (2025).DOI
- [22]V. Goblot, A. Štrkalj, N. Pernet, J. L. Lado, C. Dorow, A. Lemaître, L. Le Gratiet, A. Harouri, I. Sagnes, S. Ravets, A. Amo, J. Bloch, and O. Zilberberg, Emergence of criticality through a cascade of delocalization transitions in quasiperiodic chains, Nat. Phys. 16, 832 (2020).DOI
- [23]H. Tabanelli, C. Castelnovo, and A. Štrkalj, Reentrant localization transitions and anomalous spectral properties in off-diagonal quasiperiodic systems, Phys. Rev. B 110, 184208 (2024).arxiv.org/abs/2406.14193
- [24]S. Roy, S. Chattopadhyay, T. Mishra, and S. Basu, Critical analysis of the reentrant localization transition in a one-dimensional dimerized quasiperiodic lattice, Phys. Rev. B 105, 214203 (2022).DOI
- [25]H. Wang, X. Zheng, J. Chen, L. Xiao, S. Jia, and L. Zhang, Fate of the reentrant localization phenomenon in the one-dimensional dimerized quasiperiodic chain with long-range hopping, Phys. Rev. B 107, 075128 (2023).DOI
- [26]P.-J. Chang, Q.-B. Zeng, J. Pi, D. Ruan, and G.-L. Long, Investigation of reentrant localization transition in one-dimensional quasi-periodic lattice with long-range hopping, New J. Phys. 27, 053501 (2025).
- [27]T. F. Allard and G. Weick, Reentrant localization transition in a dimerized quasiperiodic dipolar chain, J. Phys.: Condens. Matter 38, 225401 (2026).DOI
- [28]Z.-S. Xu, J. Gao, A. Iovan, I. M. Khaymovich, V. Zwiller, and A. W. Elshaari, Observation of reentrant metal-insulator transition in a random-dimer disordered SSH lattice, npj Nanophotonics 1, 8 (2024).DOI
- [29]M. Gonçalves, B. Amorim, E. V. Castro, and P. Ribeiro, Hidden dualities in 1D quasiperiodic lattice models, SciPost Phys. 13, 046 (2022).DOI
- [30]M. Gonçalves, B. Amorim, E. V. Castro, and P. Ribeiro, Critical phase dualities in 1D exactly solvable quasiperiodic models, Phys. Rev. Lett. 131, 186303 (2023).DOI
- [31]P.-J. Chang, M.-X. Zheng, Q.-F. Lu, Z.-Y. Ma, D. Ruan, and G.-L. Long, Dimer-driven multiple reentrant localization with composite potential, Phys. Rev. A 113, 012206 (2026).arxiv.org/abs/2509.21338
- [32]See Supplemental Material for mathematical derivations and additional numerical validation.
- [33]L. G. Molinari, Spectral duality and distribution of exponents for transfer matrices of block tridiagonal Hamiltonians, J. Phys. A: Math. Gen. 36, 4081 (2003).DOI
- [34]C. Deninger, Determinants on von Neumann algebras, Mahler measures and Ljapunov exponents, J. Reine Angew. Math. 651, 165 (2011).arxiv.org/abs/0712.0667
- [35]L. Ge, S. Jitomirskaya, J. You, and Q. Zhou, Multiple-live Jensen’s formula and quantitative global theory of one-frequency Schrödinger operators, Forum Math. Pi 14, e1 (2026).arxiv.org/abs/2306.16387
- [36]A. Avila, S. Jitomirskaya, and C. Sadel, Complex one-frequency cocycles, J. Eur. Math. Soc. 16, 1915 (2014).DOI
- [37]A. Avila, Global theory of one-frequency Schrödinger operators, Acta Math. 215, 1 (2015).DOI
- [38]H. Hiramoto and M. Kohmoto, Electronic spectral and wavefunction properties of one-dimensional quasiperiodic systems: A scaling approach, Int. J. Mod. Phys. B 6, 281 (1992).DOI
- [39]C. A. Marx, Dominated splittings and the spectrum of quasi-periodic Jacobi operators, Nonlinearity 27, 3059 (2014).DOI
- [40]J. Avron and B. Simon, Almost periodic Schrödinger operators II. the integrated density of states, Duke Math. J. 50, 369 (1983).DOI
- [41]M. P. Lopez Sancho, J. M. Lopez Sancho, J. M. L. Sancho, and J. Rubio, Highly convergent schemes for the calculation of bulk and surface Green functions, J. Phys. F: Met. Phys. 15, 851 (1985).
- [42]E. B. Davies and M. Plum, Spectral pollution, IMA Journal of Numerical Analysis 24, 417 (2004).


