---
title: Creutz–Hubbard Ladder Insights
url: https://www.emergentmind.com/topics/creutz-hubbard-ladder
type: topic
---

# Creutz–Hubbard Ladder Insights

The Creutz–Hubbard ladder is an interacting two-leg lattice model obtained by augmenting the Creutz ladder with local Hubbard interactions. In the literature it appears in bosonic, spinless-fermion, spinful-fermion, and photonic forms, and it is used to study the interplay between flat-band dynamics, topology and interactions in a quasi-1D setting with synthetic magnetic flux, compact localized states, and symmetry-protected edge modes [1612.02996] [1909.00987] [2402.17494].

## 1. Canonical formulations and parameterizations

A common formulation considers a ladder of length \(L\), with sites \(\ket{j,\alpha}\), \(j=1,\dots,L\), \(\alpha=A,B\), and spinless particles which hop along legs, rungs, and diagonals in the presence of a synthetic magnetic flux \(\phi\) per plaquette. The Hamiltonian is written as
\[
H \;=\; H_{0} \;+\; H_{\rm int},
\]
with
\[
\begin{aligned}
H_{0} &= -\sum_{j=1}^{L-1} \sum_{\alpha\in\{A,B\}}
\Big[ J_\alpha\,c_{j+1,\alpha}^\dagger\,c_{j,\alpha}
+J\,c_{j+1,\alpha}^\dagger\,c_{j,\bar\alpha}
+\tfrac m2\,c_{j,\alpha}^\dagger\,c_{j,\bar\alpha}
+\mathrm{h.c.} \Big], \\
J_\alpha&=J\,e^{i\,\sigma_\alpha\,\phi/2},\quad \sigma_A=+1,\;\sigma_B=-1.
\end{aligned}
\]
For spinless fermions one uses
\[
H_{\rm int} =U\sum_{j}n_{j,A}\,n_{j,B},
\]
whereas for bosons one uses
\[
H_{\rm int} =\frac U2\sum_{j,\alpha}n_{j,\alpha}(n_{j,\alpha}-1).
\]
Here \(J\) is the diagonal hopping, \(m\) the vertical hopping, and \(\phi=2\pi\,\Phi/\Phi_0\) is the Peierls phase due to a flux \(\Phi\) per square plaquette [1909.00987].

A frequently studied spinless-fermion variant is the imbalanced Creutz–Hubbard Hamiltonian,
\[
H \;=\;
\sum_{i,\ell}\Bigl[\,-\widetilde t\,c^\dagger_{i+1,\ell}\,c_{i,\bar\ell}
+ s_\ell\,\widetilde t\,c^\dagger_{i+1,\ell}\,c_{i,\ell}
+ \text{H.c.}\Bigr]
+\frac{\Delta\epsilon}{2}\sum_{i,\ell}s_\ell\,n_{i,\ell}
+U\sum_{i}n_{i,u}n_{i,d},
\]
with \(s_u=+1\), \(s_d=-1\). In this notation \(\widetilde t\) sets the \(\pi\)-flux kinetic scale, \(\Delta\epsilon\) is a leg imbalance, and \(U>0\) is a rung repulsion [1612.02996].

Bosonic formulations also occur naturally. In the circuit-QED construction of the bosonic Creutz–Hubbard ladder, one introduces bosonic annihilation operators \(a_i,b_i\) on the upper and lower legs and writes a Bose–Hubbard Hamiltonian with vertical, diagonal, and leg hoppings \(\{t_\ell,t_d,t_v\}\), a Peierls phase \(\phi\), and on-site Kerr interaction \(U\) inherited from transmon nonlinearity [1805.12410].

These parameterizations are notationally distinct. A plausible implication is that the literature emphasizes different gauges, experimental constraints, and particle statistics rather than different underlying geometries.

## 2. Flat bands, Aharonov–Bohm caging, and topological structure

In the noninteracting sector the ladder is a paradigmatic flat-band topological system. For the \(J\)-\(m\)-\(\phi\) parameterization, the Bloch Hamiltonian can be written as
\[
H_0(k)=d_0(k)\,\mathbb I + d_x(k)\,\sigma_x+d_z(k)\,\sigma_z
\]
with
\[
d_0(k)=-2J\cos k\cos\frac\phi2,\quad
d_x(k)=-m-2J\cos k,\quad
d_z(k)=2J\sin k\sin\frac\phi2,
\]
and band energies
\[
E_\pm(k)=d_0(k)\pm\sqrt{d_x(k)^2+d_z(k)^2}.
\]
At the special point
\[
\phi=\pi,\quad m=0,
\]
one obtains
\[
E_\pm(k)=\pm 2J,
\]
namely two perfectly flat bands [1909.00987].

In the \(\theta\)-parameterization used for the bosonic Creutz ladder,
\[
H(k)=2J\bigl[\cos(k)\cos\theta\,\sigma_x+\sin(k)\sin\theta\,\sigma_z\bigr],
\]
with
\[
E_\pm(k)=\pm2J\sqrt{\cos^2(k)\cos^2\theta+\sin^2(k)\sin^2\theta}.
\]
The flat-band condition is
\[
\theta=\tfrac\pi2,
\]
which gives \(E_\pm=\pm2J\) and a lower-band Zak phase
\[
\gamma=\pi\quad (\bmod 2\pi),
\]
signalling a topologically non-trivial phase with protected edge modes [2402.17494].

The imbalanced synthetic Creutz–Hubbard model displays the same flat-band limit at \(\Delta\epsilon=0\), where the two bands are \(\epsilon_\pm(q)=\pm2\widetilde t\), the chiral symmetry satisfies \(\sigma_y H(q)\sigma_y=-H(q)\), and the winding number is \(W=1\), or equivalently the Zak phase is \(\phi_{\rm Zak}=\pi\). For open boundary conditions there are localized Aharonov–Bohm cage bulk orbitals and two exact zero-energy edge modes,
\[
l^\dagger|0\rangle=\frac1{\sqrt2}(c^\dagger_{1,u}+c^\dagger_{1,d})|0\rangle,\qquad
r^\dagger|0\rangle=\frac1{\sqrt2}(c^\dagger_{N,u}-c^\dagger_{N,d})|0\rangle.
\]
Turning on \(\Delta\epsilon\neq0\) warps the bands, and the Zak phase remains \(\pi\) for \(\Delta\epsilon<4\widetilde t\) and vanishes for \(\Delta\epsilon>4\widetilde t\), marking a non-interacting topological phase transition at \(\Delta\epsilon_c=4\widetilde t\) [1612.02996].

Aharonov–Bohm caging is one of the defining dynamical signatures. At \(\phi=\pi,m=0\), destructive interference cages any initially localized state into a small plaquette. One strictly localized eigenbasis is
\[
|\,j,\pm\rangle =\frac12\bigl[ |j,A\rangle + i|j,B\rangle \mp i|j+1,A\rangle \mp|j+1,B\rangle \bigr],
\]
and an initial single-site excitation undergoes perfectly periodic breathing with period
\[
T=\pi/J.
\]
A common source of confusion is that flatness, topology, and caging are related but distinct: flatness is realized at the special interference point, topology persists in a finite parameter region, and imbalance or rung hopping can drive the system into a trivial phase [1909.00987].

## 3. Repulsive interactions, correlated topology, and magnetic phases

Repulsive interactions convert the Creutz ladder into a genuine Creutz–Hubbard model. In the bosonic circuit-QED setting, turning on the transmon Kerr nonlinearity yields a bosonic Creutz–Hubbard model in which all Hubbard parameters \(\{t_{\ell},t_{d},t_{v},\phi,U\}\) are fully tunable. By tuning \(U/t_{\rm eff}\) one expects a superfluid–to–Mott-insulator transition, whose critical value depends on the ladder geometry and flux \(\phi\). No full many-body phase diagram is plotted in that work [1805.12410].

For spinless fermions at half filling, the synthetic Creutz–Hubbard model exhibits a much more detailed correlated phase structure. In the weak-coupling regime \(U\ll \widetilde t\), a particle–hole transformation plus Jordan–Wigner mapping gives two coupled transverse-field Ising chains, leading to a renormalized critical line between the topological insulator and the orbital paramagnet,
\[
\Delta\epsilon/\widetilde t \simeq 4 - (2/\pi)(U/\widetilde t) + O(U^2).
\]
In the strong-coupling regime \(\widetilde t\ll U\), second-order perturbation theory gives an orbital quantum Ising model on the rungs,
\[
H_{\rm eff} = J\sum_i T^y_iT^y_{i+1}+\Delta\epsilon\sum_i T^z_i,\qquad J=-8\widetilde t^2/U,
\]
with orbital ferromagnetic order for \(|2\Delta\epsilon/|J||<1\). The corresponding strong-coupling phase boundary is
\[
\Delta\epsilon/\widetilde t = 4\widetilde t/U.
\]
At \(\Delta\epsilon=0\), the balanced model has a topological-insulator to orbital-ferromagnet transition at
\[
U/\widetilde t=8.
\]
The full half-filled phase diagram contains a topological insulator, an orbital ferromagnet, and an orbital paramagnet. The topological-insulator to orbital-paramagnet line is a Dirac CFT with \(c=1\), whereas the other two lines are Majorana CFTs with \(c=\tfrac12\) [1612.02996].

The numerical characterization of these phases uses matrix-product-state DMRG and tracks the leg-imbalance susceptibility, single- and two-particle gaps, orbital magnetizations, entanglement entropy scaling, and the entanglement spectrum. In the topological-insulator phase all Schmidt levels occur in even degeneracy, while in the orbital-paramagnetic and orbital-ferromagnetic phases there is no systematic degeneracy [1612.02996].

A later spinful mean-field study identified an additional interaction-driven scenario: as the on-site Hubbard interaction grows, the ground state abruptly switches from an anti-ferromagnetic configuration to a ferromagnetic one, and this magnetic transition coincides with a topological transition marked by a quantized jump in the Zak phase from \(\pm\pi\) to \(0\). In that treatment, the Fubini–Study metric diverges at the gap-closing point, and above the critical interaction there coexist a lower-energy ferromagnetic trivial branch and a higher-energy anti-ferromagnetic topological metastable branch [2509.26320].

## 4. Attractive interactions and pair superfluidity

The attractive Creutz–Hubbard problem has been studied for spin-\(\tfrac12\) fermions with on-site interaction \(U<0\). In one standard form the Hamiltonian is
\[
\begin{aligned}
\mathcal{H} &=\;-\,\mathrm{i}\,t\sum_{j=1}^L\sum_{\sigma}
\Bigl(c^{A\dagger}_{j,\sigma}c^{A}_{j+1,\sigma}
-c^{B\dagger}_{j,\sigma}c^{B}_{j+1,\sigma}
+{\rm H.c.}\Bigr) \\
&\quad -\,t\sum_{j=1}^L\sum_{\sigma}
\Bigl(c^{A\dagger}_{j,\sigma}c^{B}_{j+1,\sigma}
+c^{B\dagger}_{j,\sigma}c^{A}_{j+1,\sigma}
+{\rm H.c.}\Bigr)
+U\sum_{j=1}^L\sum_{\alpha=A,B} n^\alpha_{j,\uparrow}n^\alpha_{j,\downarrow},
\end{aligned}
\]
and the noninteracting spectrum consists of two completely dispersionless bands,
\[
E_\pm(k)=\pm 2t.
\]
The single-particle eigenstates can be chosen as compact four-site plaquette states, so the flatness follows from exact destructive interference in the hopping network [1805.09359].

As soon as \(U<0\), two up- and down-fermions on the same site can bind into a local pair. In the weak-coupling regime \(|U|\ll t\), projection onto the lower flat band gives a flat-band gap equation implying \(\Delta\sim |U|\) for infinitesimal \(U\), and the superfluid weight is
\[
D_s =\pi\,|U|\,\rho\,(1-\rho),\qquad \rho=\tfrac{N}{2L}\quad(|U|\ll t).
\]
In the strong-coupling regime \(|U|\gg t\), tightly bound on-site pairs behave as hardcore bosons with effective Hamiltonian
\[
\mathcal H_{\rm eff} = -J\sum_{j,\alpha,\beta} \Bigl(b^{\alpha\dagger}_j b^{\beta}_{j+1}+{\rm H.c.}\Bigr)
+V\sum_{j,\alpha,\beta}n^\alpha_j\,n^\beta_{j+1},\qquad
J=V=\tfrac{2t^2}{|U|}.
\]
This yields a smooth BCS–BEC crossover in which pairs evolve from large weak-coupling objects to tightly bound strong-coupling bosons [1805.09359].

Mondaini et al. used exact diagonalization on rings up to \(2L=32\) sites and DMRG on open chains up to \(2L=384\) sites. For \(\rho<1\), the two-particle gap \(\delta_2\) vanishes in the thermodynamic limit, while the pair correlation function exhibits algebraic decay \(C(\ell)\sim \ell^{-\alpha}\). In the Creutz ladder, \(\alpha\) decreases from \(\alpha\approx1\) at weak coupling toward a saturated value \(\alpha\approx0.4\) for \(|U|\gg t\). In a standard two-leg nonflat ladder at the same filling, the corresponding exponent remains above \(\alpha\gtrsim1\). This sharper contrast is one reason the Creutz geometry is used as a reference model for flat-band pairing [1805.09359].

## 5. Doublons, trions, and composite topological quasiparticles

Repulsive Hubbard-type interactions can also generate bound states. In the photonic Creutz–Hubbard analysis of Zurita et al., a repulsive interaction term gives rise to repulsively bound pairs termed doublons. In the strong-coupling regime \(U\gg J,m\), a Schrieffer–Wolff projection onto the doubly occupied subspace yields an effective doublon Hamiltonian which is exactly a renormalized Creutz ladder with
\[
J\mapsto J^2/U,\qquad J_\alpha\mapsto J_\alpha^2/U,\qquad m\mapsto m^2/U.
\]
Accordingly, one recovers flat bands, topology, and Aharonov–Bohm caging, but now for doublons. At \(\phi=\pi/2,m=0\), the doublon bands flatten at
\[
E_\pm^D=\tfrac U2\pm2\,\frac{J^2}{U}.
\]
This produces a striking sector dependence: at \(\phi=\pi\), single-particle Aharonov–Bohm caging coexists with itinerant doublons, whereas at \(\phi=\pi/2\) one finds itinerant single-particles coexisting with caged doublons [1909.00987].

Pelegrí et al. extended the analysis to few interacting bosons in the flat-band Creutz ladder. In the flat-band limit \(\theta=\tfrac\pi2\), the two-boson sector contains on-site-plaquette doublons with dispersive bands and nearest-neighbour-plaquette doublons forming four strictly flat bands. On an open ladder there are two exact edge doublon states formed by bonding a single-particle edge mode at \(E=0\) with a neighbouring compact localized state, with energies
\[
E_{\pm}=\frac12\Bigl(U \pm\sqrt{\,U^2+16J^2\,}\Bigr).
\]
In the strong-\(U\) limit, the effective doublon and trion models are themselves Creutz ladders,
\[
J_{\rm eff}^{(2)}=\frac{2J^2}{U},\quad \theta\to2\theta,
\qquad
J_{\rm eff}^{(3)}=\frac{6J^3}{4U^2},\quad \theta\to3\theta,
\]
so that doublons and trions inherit flat-band and topological features from the original single-particle model. Trion flat bands occur at \(3\theta=\pi/2\) [2402.17494].

The same work also shows perfect two-body Aharonov–Bohm caging for arbitrary interaction strengths in the flat-band limit. For the initial state
\[
\bigl|\Psi_i\bigr\rangle
=\frac12\Bigl(\hat a_i^\dagger\hat a_i^\dagger
+\hat b_i^\dagger\hat b_i^\dagger\Bigr)|0\rangle,
\]
the time evolution remains perfectly confined within a two-plaquette cage for all \(U/J\). Matrix-product-state time evolution further provides numerical evidence that the main features of this effect are preserved in an interacting many-body scenario, resulting in many-body Aharonov–Bohm caging [2402.17494].

## 6. Disorder, flat-band many-body localization, and re-entrant thermalization

In the interacting flat-band Creutz ladder with disorder, the Hamiltonian is decomposed as
\[
\hat H=\hat H_{\rm CL}+\hat H_U+\hat H_{\rm dis},
\]
where \(\hat H_{\rm CL}\) is the flat-band Creutz-ladder kinetic term, \(\hat H_U=U\sum_j n^a_j n^b_j\), and
\[
\hat H_{\rm dis}=\sum_j\bigl[\varepsilon^a_j n^a_j+\varepsilon^b_j n^b_j\bigr],\qquad
\varepsilon^{a,b}_j\in[-W,+W].
\]
At the flat-band point \(t_0=t_1=t\), the single-particle spectrum is
\[
E_\pm(k)=\pm 2t,
\]
and there is an exact basis of compact localized states \(W_j^\pm\) satisfying
\[
\hat H_{\rm CL}\Big|_{t_0=t_1} =-2t\sum_jW_j^{+\dagger}W_j^+ +2t\sum_jW_j^{-\dagger}W_j^-.
\]
When \(U>0\) and \(W=0\), the occupations \(K_j^\pm=W_j^{\pm\dagger}W_j^\pm\) commute with the Hamiltonian, producing an extensive set of exact local integrals of motion and a stable flat-band many-body localized phase [2106.15385].

The disorder response is non-generic. Weak disorder first destroys the flat-band many-body localized phase and induces a thermal phase; further increase of disorder leads to the conventional many-body-localized phase. The resulting three-regime structure is
\[
\text{FMBL} \;\to\; \text{Thermal/ETH} \;\to\; \text{MBL},
\]
with approximate two-particle boundaries
\[
W_{c1}(U)\simeq W_0+\alpha U,\qquad
W_{c2}(U)\simeq W_1+\beta U,
\]
and typical numbers \(W_0\approx0.1\), \(W_1\approx4.0\) in units of \(t\) [2106.15385].

The dynamical diagnostics track the same structure. In the FMBL phase, the half-chain entanglement entropy remains bounded, the number entropy stays small, the fourth moment \(P_\psi(t)\) is approximately constant, and local densities show persistent oscillations associated with Aharonov–Bohm caging. In the thermal regime, entanglement grows rapidly to a volume law and the return probability decays to zero. In the conventional MBL regime, entanglement shows the familiar logarithmic growth \(S(t)\sim\ln t\) before saturation [2106.15385].

This phase sequence is often singled out because it reverses the naive expectation that added disorder should only strengthen localization.

## 7. Quantum-simulation platforms and experimental realizations

The Creutz–Hubbard ladder has been pursued in several synthetic platforms. In ultracold atoms, one proposal synthesizes the ladder in a one-dimensional spin-independent optical lattice by mapping the two legs onto two hyperfine states of fermionic atoms. Static tilting suppresses bare hopping, Raman-assisted inter-leg tunneling produces diagonal spin-flip links, and lattice shaking restores intra-leg hopping with Peierls phases so that the net flux per square is \(\pi\). The Raman detuning yields the leg imbalance \(\Delta\epsilon\), while the on-site \(s\)-wave interaction gives \(U\). Suggested probes include Ramsey interferometry with Bloch oscillations for the Zak phase, localized Bragg spectroscopy for edge states, spin-resolved imaging for leg imbalance, and spin-structure-factor measurements for orbital magnetization [1612.02996].

A related experimental step realized a topological Creutz ladder for ultracold fermionic atoms in a resonantly driven one-dimensional optical lattice. There the two legs are the two lowest orbital states \(s\) and \(p\), and the cross inter-leg links are generated via two-photon resonant coupling by periodic lattice shaking. The characteristic pseudo-spin winding was demonstrated using momentum-resolved Ramsey-type interferometric measurements. The same framework includes a two-tone driving method to extend the inter-leg link control and a topological charge pumping protocol with quantized pumped charge
\[
Q=1
\]
when the adiabatic path encircles a band-gap closing point [1902.10304].

In photonics, the full Creutz and Creutz–Hubbard ladders can be emulated by arrays of coupled optical waveguides or ring resonators, using either real and synthetic dimensions. One route uses a one-dimensional chain with a synthetic leg degree of freedom carried by transverse modes. Another uses a rhombus chain with detuned spine sites, where the effective diagonal coupling is
\[
J_{\rm eff}=J^2/\Delta.
\]
For the interacting two-particle sector, the standard mapping from two interacting particles in one dimension to a single particle on a two-dimensional lattice leads to a quasi-2D array or a 3D waveguide network with a synthetic particle dimension. Proposed protocols include direct observation of single-particle Aharonov–Bohm caging, edge-state trapping, and doublon caging or doublon edge modes. Typical parameters quoted for waveguide devices are \(J\sim0.5\)–\(2\,{\rm cm}^{-1}\), synthetic flux via \(\sim1\,{\rm GHz}\) modulation or \(\Delta\beta\sim10\,{\rm cm}^{-1}\), and effective \(U\sim10J\) in nonlinear ring-resonator networks [1909.00987].

Circuit QED provides a bosonic route. There, coupled transmons have their Josephson energies modulated harmonically in time,
\[
E_J(t)=E_J^{(0)}+\delta E_J\cos(\omega_M t+\Phi_0),
\]
which makes the transmon frequencies time dependent. In the rotating-frame Floquet description, the hopping acquires a time-periodic phase factor, and period averaging yields an effective complex hopping controlled by Bessel functions and by the site-dependent phase offsets \(\Phi_{0,i}\). These phase offsets act as an artificial gauge potential, and their difference controls the Peierls phase on each link. Because of the intrinsic non-linearity of the transmon qubits, such lattices are an ideal platform for simulating Bose–Hubbard Hamiltonians with non-trivial gauge fields, and the construction is directly realizable in a superconducting circuit and easily extendable to a scalable lattice [1805.12410].

Across these platforms, the Creutz–Hubbard ladder functions as a compact model in which flat bands, edge topology, synthetic gauge structure, and strong correlations can be tuned in a controlled way. Its unusually broad phenomenology spans topological insulators, orbital magnetic phases, pair superfluidity, doublon and trion quasiparticles, Aharonov–Bohm caging, flat-band many-body localization, and interaction-driven topological transitions.

Source: https://www.emergentmind.com/topics/creutz-hubbard-ladder