--- title: Local Caging Potential in Lattice Systems url: https://www.emergentmind.com/topics/local-caging-potential type: topic --- # Local Caging Potential in Lattice Systems Searching arXiv for the cited literature on local caging potential and closely related Aharonov–Bohm caging work. “Local caging potential” denotes an effective confinement mechanism that restricts dynamics to a finite spatial or configuration-space region without requiring a conventional scalar trapping well. In the literature surveyed here, the term appears in several technically distinct but conceptually related senses. In Aharonov–Bohm caging, confinement is generated by lattice geometry and gauge-induced destructive interference, so that compact localized states remain strictly supported on a finite cluster [1810.07641]. In interacting flat-band systems, local Kerr or Hubbard terms can either preserve, reconstruct, or perturb such cages, and can be interpreted as shaping the internal dynamics of a pre-existing interference-defined cage [1810.07641], [2409.05853], [2004.11880]. In colloidal and dense suspension settings, by contrast, the local caging potential is a microscopic structural measure of how strongly a particle is trapped by its nearest neighbors, derived from local pair structure and correlated with motif stability, strain amplification, and flow response [2508.15621], [2311.09112]. Across these uses, the common theme is locality of confinement together with a mechanism—interference, interaction, or geometry—that suppresses escape from a bounded neighborhood. ## 1. Interference-defined caging in translationally invariant lattices In the Aharonov–Bohm caging setting, the local caging potential is not an explicit on-site potential. Instead, localization arises in a translationally invariant lattice because multiple hopping paths interfere destructively. The canonical example is the rhombic (diamond) chain with three sites per unit cell, labeled \(A_n,B_n,C_n\), threaded by a flux \(\phi\). At \(\phi=\pi\), the noninteracting single-particle spectrum collapses into three perfectly flat bands with energies \[ \epsilon_\pm=\pm 2J,\qquad \epsilon_0=0, \] and the corresponding eigenstates are compact localized states supported on a finite cage [1810.07641]. This mechanism is distinct from Anderson localization and from confinement by a spatially varying scalar potential. In AB caging, all sites remain equivalent, and localization is produced by gauge phases and lattice connectivity alone. In the rhombic chain, the amplitudes along the two paths \(A_n\to B_n\to A_{n+1}\) and \(A_n\to C_n\to A_{n+1}\) cancel at \(\phi=\pi\), preventing propagation beyond a finite cluster. This is why several of the cited works describe the “caging potential” as effectively encoded in geometry and flux rather than in diagonal terms [2001.08551], [2208.02956]. A closely related formulation appears in multi-path lattices with \(N\) intermediate sites \(C_{n,i}\) between neighboring backbone sites \(A_n\) and \(A_{n+1}\). There, AB caging is realized when the phase set \(\{\phi_i\}\) satisfies \[ \sum_{i=1}^{N}\cos\phi_i=0,\qquad \sum_{i=1}^{N}\sin\phi_i=0, \] which makes the band energy \[ E_k=\sqrt{\sum_{i=1}^{N}2J^2\left(1+\cos\phi_i\cos k+\sin\phi_i\sin k\right)} \] independent of \(k\), yielding perfectly flat bands and compact localized states [2605.23546]. This suggests a useful general interpretation: a local caging potential may be realized through off-diagonal phase engineering that enforces exact cancellation of escape amplitudes. ## 2. Local nonlinearities and interaction-shaped cages in flat-band systems A central issue in flat-band physics is whether local interactions destroy caging by coupling compact localized states. In the rhombic chain with \(\pi\)-flux and local Kerr-type nonlinearities, the classical amplitudes obey \[ \begin{aligned} i\,\dot{a}_n &= - J \left( b_{n-1} + c_{n-1} + e^{i\phi} b_n + c_n \right) + U | a_n |^2 a_n, \\ i\,\dot{b}_n &= - J \left( e^{-i\phi} a_{n-1} + a_{n} \right) + U | b_n |^2 b_n, \\ i\,\dot{c}_n &= - J \left( a_{n-1} + a_{n} \right) + U | c_n |^2 c_n . \end{aligned} \] For \(\phi=\pi\), the main result is that caging survives under local nonlinearities: any state initially supported between two hub sites remains confined there at all times [1810.07641]. In that setting, the local nonlinear term does not create the cage boundary; the boundary remains an interference effect enforced by geometry and \(\pi\)-flux. What the local nonlinearity does is reshape the internal dynamics within the cage. This is why the paper’s discussion motivates the phrase “local caging potential” as an effective localized potential landscape inside a hard interference-defined box [1810.07641]. The distinction is important: local Kerr or on-site Hubbard terms modify amplitudes and phases within the cage, but do not open escape channels so long as the interaction remains local or very short-range. The same logic appears in quantum interacting AB cages. In the Bose–Hubbard rhombic chain, \[ \hat H = -J \sum_{\langle i,j \rangle} \left( e^{i\theta_{ij}} \hat \psi^\dagger_i \hat \psi_j + \text{H.c.} \right) + \frac{U}{2} \sum_i \hat n_i (\hat n_i - 1), \] on-site interactions generally hybridize localized states and promote spreading. However, the precise role of interactions depends on their structure. In one line of work, local Bose–Hubbard interactions on all-bands-flat lattices yield “quantum caging conditions” under which particles move only in pairs in the detangled basis, local number parity in each unit cell is conserved, and compact energy-renormalized many-body states persist for two and three particles, with an inductive conjecture for arbitrary finite \(M\) in one dimension [2004.11880]. In another rhombus-chain study, two-boson AB caging is restored when on-site and nearest-neighbor interactions are equal, \(U=V_1=V_2\equiv V\), so that the relevant two-boson bands become flat and localized two-particle dynamics re-emerge [2409.05853]. These results delimit a common misconception. It is not generally correct that interactions simply destroy AB caging. Local interactions can preserve caging [1810.07641], reconstruct it in an interacting few-body sector [2409.05853], or enforce parity-constrained transport of pairs while leaving compact many-body states intact [2004.11880]. A plausible implication is that “local caging potential” in flat-band systems is best understood as the joint effect of local interactions and an interference-defined compact support structure, rather than as either ingredient alone. ## 3. Effective low-dimensional descriptions inside a cage Once caging is preserved, the dynamics inside a finite cage can often be reduced to a lower-dimensional effective model. In the nonlinear rhombic chain, a five-site caged ansatz with \[ a_{n-1}(t)=a_{n+1}(t)=0,\qquad b_{n-1}(t)=c_{n-1}(t),\qquad b_n(t)=-c_n(t) \] reduces the full nonlinear lattice problem to an effective two-mode description [1810.07641]. The amplitudes are parametrized by a conserved particle number \(N\), a fractional imbalance \(z(t)\in[-1,1]\), and a phase difference \(\xi(t)\), leading to \[ \dot z(t) = -4J\,\sqrt{1-z(t)^2}\,\sin \xi(t), \] \[ \dot \xi(t) = 4J\,\frac{z(t)}{\sqrt{1-z(t)^2}}\cos \xi(t) + \frac{5}{8} g\,z(t) - \frac{3}{8} g, \] with \(g\equiv NU\), and an effective Hamiltonian \[ \mathcal{H}(z,\xi) = -4J\,\sqrt{1-z^2}\,\cos \xi + \frac{5}{16} g\, z^2 - \frac{3}{8} g\, z. \] The resulting caged solutions display periodic breathing of the intensity while remaining strictly confined [1810.07641]. This effective Hamiltonian is described as reminiscent of a bosonic Josephson junction. In that picture, the local caging potential is encoded in the nonlinear term of \(\mathcal H(z,\xi)\) together with the geometric factor \(\sqrt{1-z^2}\), while the cage boundaries remain fixed by AB interference. The internal population imbalance oscillates, but the support of the solution does not spread. A related reduction occurs in interacting all-bands-flat lattices. There, after transforming to a detangled basis of compact localized modes, the Hubbard interaction becomes a constrained pair-hopping and exchange problem [2004.11880]. The effective many-body dynamics separate into dispersive subsystems that describe mobile bound pairs and detangled subsystems that host compact few-body eigenstates with interaction-renormalized energies. This suggests that local caging potentials often admit a dual description: as finite-support structures in real space and as finite-dimensional effective Hamiltonians governing the trapped internal degrees of freedom. ## 4. Non-Abelian, non-Hermitian, and many-body generalizations The concept broadens considerably beyond Abelian \(\pi\)-flux rhombic lattices. In non-Abelian AB caging on a multi-component rhombic lattice, hopping is governed by matrix-valued link variables \(U_{ij}\in U(N)\), and the key object is the interference matrix \[ I=\frac{1}{2}\left(U_2U_1+U_4U_3\right). \] Non-Abelian AB caging is defined by nilpotency of \(I\), \[ \exists\, m\in\mathbb N\ \text{s.t.}\ I^{m-1}\neq 0,\quad I^m=0, \] which enforces a finite propagation range in units of cells [2001.08551]. In this formulation, the local caging potential is operator-valued: the nilpotent structure of \(I\) rather than a scalar onsite term determines which internal states can move and how far. The cage can depend on the initial spinor and on the Jordan structure of the interference matrix, which has no Abelian analogue [2001.08551]. A single-trapped-ion proposal later emphasizes the same principle and shows that caging can arise either from a nilpotent interference matrix or from a suitably chosen initial state in a non-Abelian \(U(2)\) rhombic lattice simulator [2505.00899]. Non-Hermitian generalizations introduce another route to tunable local caging potentials. In a generalized Creutz ladder with anti-\(\mathcal{PT}\)-symmetric imaginary couplings, the Bloch Hamiltonian remains flat with eigenvalues \[ E_k=\pm\sqrt{J^2-\Gamma^2}, \] independent of \(k\). The compact localized states remain finite-support eigenmodes, but the flat-band energy and oscillation period \[ T=\frac{\pi}{\sqrt{J^2-\Gamma^2}} \] become tunable through the imaginary intracell coupling \(\Gamma\) [2308.11968]. Here the local caging potential is best viewed as a non-Hermitian deformation of the decoupled AB-cage core, preserving confinement while changing spectral and dynamical properties. An even more strongly interaction-driven interpretation appears in non-Hermitian bosonic clustering. In a non-Hermitian SSH chain with asymmetric hopping and a single local density interaction, \[ H_{\text{int}}=\frac{U}{2}n_{x_0}^2, \] the two-boson configuration space \((x,x')\) acquires an L-shaped barrier along \(x=x_0\) and \(x'=x_0\), with a corner barrier at \((x_0,x_0)\). Together with non-Hermitian skin pumping, this creates an emergent cage \(\{x,x'\le x_0\}\) that produces ultra-strong bosonic clustering at the boundary [2410.01258]. In this case the local caging potential is explicitly interaction-induced in configuration space rather than interference-induced in real space. The many-body extension can be pushed further still. In kinetically constrained models, destructive interference in Fock space creates “Fock Space Cages,” finite induced subgraphs of the many-body basis graph that support exact localized zero modes [2504.20987]. There, the local caging potential is not a literal scalar potential but an effective graph-theoretic interference structure in Hilbert space. This suggests that the notion of local caging potential is portable across real-space lattices, synthetic dimensions, and configuration-space graphs so long as escape amplitudes can be locally canceled. ## 5. Breakdown, inverse transitions, and perturbations Local caging potentials are fragile to perturbations that spoil the symmetry or interference conditions responsible for confinement. In the nonlinear AB rhombic chain, caging persists for on-site and nearest-neighbor interactions but is destroyed by sufficiently long-range nonlinear couplings, such as next-nearest-neighbor terms, because they directly spoil the symmetric amplitude relations needed for destructive interference at the cage boundary [1810.07641]. The same pattern appears in multi-flux AB cages. Although phase-engineered couplings satisfying \[ \sum_i \cos\phi_i=0,\qquad \sum_i \sin\phi_i=0 \] yield exact caging, adding local onsite detuning terms such as \[ \Delta\,\hat C^\dagger_{n,1}\hat C_{n,1}-\Delta\,\hat C^\dagger_{n,5}\hat C_{n,5} \] breaks the path symmetry and destroys localization [2605.23546]. The paper quantifies this with the inverse participation number and its fluctuation \[ \overline{IPN} = \frac{1}{T}\int_0^T p^{-1}(t)\,dt,\qquad \sigma=\sqrt{\overline{(IPN)^2}-(\overline{IPN})^2}, \] finding maximal localization stability at \(\Delta=0\) and progressive decay as detuning increases [2605.23546]. A notable consequence is the inverse Anderson transition. In the experimentally realized rhombic momentum-space lattice of ultracold atoms, AB caging at \(\phi=\pi\) is robust in the clean system and under symmetric correlated disorder, but anti-symmetric correlated disorder, \[ \Delta_n^{(b)}=-\Delta_n^{(c)}, \] creates dispersive bands and ballistic transport, with a characteristic disorder scale \(\Delta_0=\sqrt{2}J\) in the infinite-system theory [2208.02956]. The same qualitative phenomenon was later shown for interacting bosons on the rhombus chain: antisymmetric correlated onsite disorder or a suitable potential gradient destroys interaction-restored AB caging and delocalizes the two-boson state [2409.05853]. In both cases, disorder does not merely add another localization mechanism; it can dismantle an interference-built local caging potential and thereby restore mobility. This resolves another common confusion. AB caging should not be conflated with generic localization. Because it is stronger than exponential Anderson localization and relies on exact compact support, perturbations can either weaken it, destroy it, or in some limits replace it with more conventional localization mechanisms such as Stark localization at large tilt [2409.05853]. ## 6. Structural and continuum meanings of local caging potential Outside flat-band lattice physics, “local caging potential” is used in a more literal structural sense. In dense colloidal suspensions, it is a particle-resolved mean-field quantity measuring how strongly a particle is trapped by its local environment: \[ \beta \Phi^i=\rho \int d\mathbf r\, C^i(r)\, g^i(r), \] with \(C^i(r)\approx g^i(r)-1\), so that \[ \beta \Phi^i \approx \rho \int d\mathbf r\, \big(g^i(r)-1\big)\, g^i(r). \] Here \(g^i(r)\) is a particle-level radial distribution function constructed from a mollified local RDF, and \(\beta\Phi^i\) serves as a structural order parameter linking local topology to dynamical stability [2508.15621]. In that framework, deeper caging potentials correspond to mechanically more stable environments. Icosahedral motifs \(13A\) in glasses have deeper caging potentials than crystalline motifs such as FCC and HCP, while defective five-membered-ring motifs \(10B\) and \(11E\) have intermediate depths [2508.15621]. Under shear, clusters of such motifs fragment, and particles leaving stable motif clusters exhibit shallower caging potentials together with larger non-affine displacements \[ D^2_{min,i}(\gamma,\Delta\gamma)= \frac{1}{N_i}\sum_{j=1}^{N_i} \left[ \mathbf r^{ij}(\gamma+\Delta\gamma)-\mathbf\Gamma_j(\gamma)\mathbf r^{ij}(\gamma) \right]^2, \] which ties the loss of mechanical stability to the topological evolution of the local cage network [2508.15621]. A related continuum-level but distinct use appears in dense bidisperse non-Brownian suspensions. There, caging is modeled geometrically through effective cells around particles, and small particles can be caged inside interstices between large ones when the size ratio exceeds \[ \lambda_{\text{int}}=(\sqrt{2}-1)^{-1}\simeq 2.41. \] The effective local strain amplification obeys \[ \frac{\dot{\varepsilon}_{\text{local}}}{\dot{\varepsilon}_{\text{macro}}} = \left(1-\frac{\phi}{\phi_c}\right)^{-1}, \] and the relative viscosity is \[ \eta_r(\phi)=\frac{1-\phi}{\left(1-\phi/\phi_c\right)^2}, \] with \(\phi_c\) interpreted as an effective packing fraction of cells [2311.09112]. In this context, the local caging potential is not an energy landscape in the quantum sense, but an effective geometric confinement encoded in the available free volume and in the amplified solvent deformation around particles. A more abstract but conceptually aligned picture is provided by the strong local potential limit in finite many-body systems with \[ H=T+V,\qquad V=\sum_{i