Half-Filled Hubbard Ladders

• Half-filled Hubbard ladders are quasi-one-dimensional systems with one electron per site, leading to Mott insulator behavior, spin gaps, and rich competing orders.
• Research employs numerical simulations and field-theoretic approaches to reveal phase-string effects and non-BCS pairing mechanisms when carriers are doped into the ladder.
• Various deformations—including checkerboard, asymmetric, and triangular geometries—demonstrate how rung topology and frustration transform traditional antiferromagnetic and Mott phases into ferrimagnetic and chiral regimes.

Half-filled Hubbard ladders are quasi-one-dimensional Hubbard systems in which the average density is one electron per site. In the canonical two-leg case, the lattice interpolates between a chain and a two-dimensional plaquette geometry, so half-filling becomes a controlled setting for Mott localization, short-range antiferromagnetism, spin gaps, competing broken-symmetry states, nonlocal order parameters, and, in deformed geometries, ferrimagnetism, band insulation, or flux-induced chiral spin physics. The literature on uniform, doped, asymmetric, modulated, and geometrically decorated ladders shows that the phrase “half-filled Hubbard ladder” refers less to a single phase than to a family of closely related strongly correlated problems whose infrared structure depends sensitively on rung topology, sublattice balance, interleg hopping, and interaction scale (Zhu et al., 2015, Boschi et al., 2015, Essalah et al., 2021).

1. Canonical models and the meaning of half-filling

A standard half-filled two-leg Hubbard ladder is defined on $N_x$ rungs and two legs $j=1,2$, with Hamiltonian

$$\begin{aligned} H=& - \alpha t \sum_{i=1}^{N_x -1}\sum_{j=1,2} \big[c^{\dagger}_{\sigma}(i+1, j) c^{}_{\sigma} (i, j) + \text{h.c.}\big] \ & - t\sum_{i=1}^{N_x} \big[ c^{\dagger}_{\sigma}(i, 1) c^{}_{\sigma} (i, 2) +\text{h.c.}\big] \ & + U\sum_{i=1}^{N_x}\sum_{j=1,2} n_{\uparrow}(i,j) n_{\downarrow}(i,j) , \end{aligned}$$

where $\alpha t$ is the leg hopping, $t$ the rung hopping, and $U$ the on-site repulsion. Half-filling means

$$\langle n_{\uparrow}(i,j) + n_{\downarrow}(i,j) \rangle = 1$$

on average per site, so the total particle number equals the number of sites. For sufficiently large $U/t$, this realizes a Mott insulator with short-range antiferromagnetic correlations and, in the two-leg geometry, a spin gap with predominantly rung singlets (Zhu et al., 2015).

This canonical ladder admits several deformations that preserve the half-filled setting while changing the low-energy theory. A checkerboard ladder replaces uniform leg hopping by a period-2 modulation and yields a one-dimensional array of square plaquettes; an asymmetric ladder couples a Hubbard chain to a noninteracting one-dimensional electron gas; alternating ladders such as the 3–2 and 3–3–2–2 geometries vary the number of sites per rung; triangular ladders introduce geometric frustration and flux; and trimer ladders place three sites in each unit cell and can host a flat middle band. In each case, half-filling remains the reference density, but the resulting phases differ sharply because the symmetry content, single-particle structure, and effective strong-coupling spin models differ (Karakonstantakis et al., 2010, Abdelwahab et al., 2014, Essalah et al., 2021, Garuchava et al., 2024, Saha et al., 4 Mar 2025).

2. Uniform two-leg ladders: Mott insulator, Luther-Emery structure, and the small-$U$ issue

In the standard description of the repulsive uniform two-leg ladder, half-filling produces a spin-gapped Mott insulator with a finite charge gap and spin gap, while doping away from $n=1$ closes the charge gap but leaves a finite spin gap, yielding a Luther-Emery liquid with a single gapless charge mode (Karakonstantakis et al., 2010). This framework underlies much of the subsequent ladder literature, including studies of pairing scales, rung-singlet physics, and the crossover between weak and strong coupling.

A re-examination of the small-$j=1,2$0 limit gives a more delicate picture. In that treatment, the zero-temperature ground state for all small $j=1,2$1 in the regime where both bands are partially filled is argued to be a C1S0 Luther-Emery phase rather than a stable C2S1 phase; the apparent C2S1 regime is interpreted as an intermediate stage in a two-step renormalization-group flow. For $j=1,2$2, the gapped modes are characterized by a single emergent correlation length with $j=1,2$3. For $j=1,2$4, there is a hierarchy of scales: a primary gap $j=1,2$5, and secondary gaps

$j=1,2$6

so the system first looks C2S1-like and only at longer scales crosses to C1S0 (Gannot et al., 2020).

This juxtaposition is one of the central subtleties of the subject. The standard ladder picture emphasizes a half-filled spin-gapped Mott state, whereas the small-$j=1,2$7 analysis emphasizes a Luther-Emery fixed point with one gapless total charge mode in the genuine weak-coupling limit, while allowing that Umklapp can gap that mode at larger $j=1,2$8. This suggests that the infrared classification of the half-filled two-leg ladder is especially sensitive to whether one takes the moderate-coupling Mott regime or the asymptotically weak-coupling continuum limit as primary (Karakonstantakis et al., 2010, Gannot et al., 2020).

3. Competing insulating orders and nonlocal characterization

A broad field-theoretic classification of half-filled two-leg ladders emerges from the generalized Hund chain model. That analysis identifies eight possible Mott insulating phases and establishes a one-to-one correspondence with the phases of the two-leg Hubbard ladder with interchain hopping. In the ladder language, the phases are S-Mott, D-Mott, S′-Mott, D′-Mott, $j=1,2$9, PDW, SF, and FDW; in the generalized Hund-chain language they correspond respectively to SP, CDW, ODW, $$\begin{aligned} H=& - \alpha t \sum_{i=1}^{N_x -1}\sum_{j=1,2} \big[c^{\dagger}_{\sigma}(i+1, j) c^{}_{\sigma} (i, j) + \text{h.c.}\big] \ & - t\sum_{i=1}^{N_x} \big[ c^{\dagger}_{\sigma}(i, 1) c^{}_{\sigma} (i, 2) +\text{h.c.}\big] \ & + U\sum_{i=1}^{N_x}\sum_{j=1,2} n_{\uparrow}(i,j) n_{\downarrow}(i,j) , \end{aligned}$$0, RS, HC, HO, and RT. Four of these phases are two-fold-degenerate states with broken lattice symmetry, while RS, RT, HC, and HO are non-degenerate Haldane-type phases distinguished by string order and edge states rather than by conventional local order parameters (Nonne et al., 2010).

The same work shows that half-filled ladder physics is naturally organized by duality transformations acting on the Majorana representation of the low-energy theory. These dualities exchange density-wave and Mott sectors and map trivial gapped phases onto nontrivial Haldane sectors. In particular, RT is the spin Haldane phase with spin string order on rungs, HC is a charge Haldane phase with hidden order in the charge pseudospin sector, HO is an orbital Haldane phase, and RS is the trivial rung-singlet phase. This classification is important because it places familiar ladder states such as D-Mott and rung-singlet phases inside a larger symmetry-based taxonomy rather than treating them as isolated cases (Nonne et al., 2010).

A complementary nonlocal characterization uses parity “brane” correlators. For an $$\begin{aligned} H=& - \alpha t \sum_{i=1}^{N_x -1}\sum_{j=1,2} \big[c^{\dagger}_{\sigma}(i+1, j) c^{}_{\sigma} (i, j) + \text{h.c.}\big] \ & - t\sum_{i=1}^{N_x} \big[ c^{\dagger}_{\sigma}(i, 1) c^{}_{\sigma} (i, 2) +\text{h.c.}\big] \ & + U\sum_{i=1}^{N_x}\sum_{j=1,2} n_{\uparrow}(i,j) n_{\downarrow}(i,j) , \end{aligned}$$1-leg ladder, the fractional charge and spin parity operators are

$$\begin{aligned} H=& - \alpha t \sum_{i=1}^{N_x -1}\sum_{j=1,2} \big[c^{\dagger}_{\sigma}(i+1, j) c^{}_{\sigma} (i, j) + \text{h.c.}\big] \ & - t\sum_{i=1}^{N_x} \big[ c^{\dagger}_{\sigma}(i, 1) c^{}_{\sigma} (i, 2) +\text{h.c.}\big] \ & + U\sum_{i=1}^{N_x}\sum_{j=1,2} n_{\uparrow}(i,j) n_{\downarrow}(i,j) , \end{aligned}$$2

with corresponding brane correlators

$$\begin{aligned} H=& - \alpha t \sum_{i=1}^{N_x -1}\sum_{j=1,2} \big[c^{\dagger}_{\sigma}(i+1, j) c^{}_{\sigma} (i, j) + \text{h.c.}\big] \ & - t\sum_{i=1}^{N_x} \big[ c^{\dagger}_{\sigma}(i, 1) c^{}_{\sigma} (i, 2) +\text{h.c.}\big] \ & + U\sum_{i=1}^{N_x}\sum_{j=1,2} n_{\uparrow}(i,j) n_{\downarrow}(i,j) , \end{aligned}$$3

In this language, the half-filled Mott state is viewed as a background of singly occupied sites dressed by holon-doublon fluctuations localized in pairs. The charge parity brane remains nonzero at any repulsive $$\begin{aligned} H=& - \alpha t \sum_{i=1}^{N_x -1}\sum_{j=1,2} \big[c^{\dagger}_{\sigma}(i+1, j) c^{}_{\sigma} (i, j) + \text{h.c.}\big] \ & - t\sum_{i=1}^{N_x} \big[ c^{\dagger}_{\sigma}(i, 1) c^{}_{\sigma} (i, 2) +\text{h.c.}\big] \ & + U\sum_{i=1}^{N_x}\sum_{j=1,2} n_{\uparrow}(i,j) n_{\downarrow}(i,j) , \end{aligned}$$4, while the spin parity brane becomes nonvanishing only in even-leg ladders, where a spin gap opens in the D-Mott phase; it remains absent in odd-leg ladders, where a gapless spin mode survives (Boschi et al., 2015).

4. Doping a half-filled ladder: single-hole physics and phase strings

Removing a single fermion from an otherwise half-filled two-leg ladder produces an especially sharp probe of the Mott background. In the isotropic ladder $$\begin{aligned} H=& - \alpha t \sum_{i=1}^{N_x -1}\sum_{j=1,2} \big[c^{\dagger}_{\sigma}(i+1, j) c^{}_{\sigma} (i, j) + \text{h.c.}\big] \ & - t\sum_{i=1}^{N_x} \big[ c^{\dagger}_{\sigma}(i, 1) c^{}_{\sigma} (i, 2) +\text{h.c.}\big] \ & + U\sum_{i=1}^{N_x}\sum_{j=1,2} n_{\uparrow}(i,j) n_{\downarrow}(i,j) , \end{aligned}$$5 at large but finite $$\begin{aligned} H=& - \alpha t \sum_{i=1}^{N_x -1}\sum_{j=1,2} \big[c^{\dagger}_{\sigma}(i+1, j) c^{}_{\sigma} (i, j) + \text{h.c.}\big] \ & - t\sum_{i=1}^{N_x} \big[ c^{\dagger}_{\sigma}(i, 1) c^{}_{\sigma} (i, 2) +\text{h.c.}\big] \ & + U\sum_{i=1}^{N_x}\sum_{j=1,2} n_{\uparrow}(i,j) n_{\downarrow}(i,j) , \end{aligned}$$6, DMRG finds a pronounced oscillatory modulation in both the hole density $$\begin{aligned} H=& - \alpha t \sum_{i=1}^{N_x -1}\sum_{j=1,2} \big[c^{\dagger}_{\sigma}(i+1, j) c^{}_{\sigma} (i, j) + \text{h.c.}\big] \ & - t\sum_{i=1}^{N_x} \big[ c^{\dagger}_{\sigma}(i, 1) c^{}_{\sigma} (i, 2) +\text{h.c.}\big] \ & + U\sum_{i=1}^{N_x}\sum_{j=1,2} n_{\uparrow}(i,j) n_{\downarrow}(i,j) , \end{aligned}$$7 and the spin density $$\begin{aligned} H=& - \alpha t \sum_{i=1}^{N_x -1}\sum_{j=1,2} \big[c^{\dagger}_{\sigma}(i+1, j) c^{}_{\sigma} (i, j) + \text{h.c.}\big] \ & - t\sum_{i=1}^{N_x} \big[ c^{\dagger}_{\sigma}(i, 1) c^{}_{\sigma} (i, 2) +\text{h.c.}\big] \ & + U\sum_{i=1}^{N_x}\sum_{j=1,2} n_{\uparrow}(i,j) n_{\downarrow}(i,j) , \end{aligned}$$8, with an incommensurate period roughly equal to two lattice sites. Near the ladder center, the modulations in $$\begin{aligned} H=& - \alpha t \sum_{i=1}^{N_x -1}\sum_{j=1,2} \big[c^{\dagger}_{\sigma}(i+1, j) c^{}_{\sigma} (i, j) + \text{h.c.}\big] \ & - t\sum_{i=1}^{N_x} \big[ c^{\dagger}_{\sigma}(i, 1) c^{}_{\sigma} (i, 2) +\text{h.c.}\big] \ & + U\sum_{i=1}^{N_x}\sum_{j=1,2} n_{\uparrow}(i,j) n_{\downarrow}(i,j) , \end{aligned}$$9 and $\alpha t$0 can reach $\alpha t$1 of the average density, and they remain visible for $\alpha t$2, becoming weak only for ladders shorter than $\alpha t$3. The doped object is not a tightly localized quasiparticle but a loosely bound composite of charge and a spatially extended spin-$\alpha t$4 texture (Zhu et al., 2015).

The proposed mechanism is the phase-string Berry phase. In the $\alpha t$5-$\alpha t$6 description, a closed hole trajectory $\alpha t$7 carries

$\alpha t$8

where $\alpha t$9 counts exchanges between the hole and down spins. In the Hubbard model the exact sign structure is more complicated,

$t$0

but near half-filling and large $t$1 the holon-doublon factors become ineffective and the sign structure reduces to the same phase string. Interference between paths with different $t$2 then produces the real-space spin and charge modulations. The same interpretation is supported by the fact that the modulations disappear when the system is spin-polarized, when $t$3 is pushed toward the Nagaoka regime, when a second hole is added, or when strong rung asymmetry suppresses nontrivial loop motion (Zhu et al., 2015).

The two-hole result is especially significant. At $t$4 and $t$5, the charge modulation disappears upon adding a second hole, and the two holes form a bound pair whose coherent motion cancels the nontrivial phase strings. In the terminology of that work, this is evidence for non-BCS pairing driven by phase-string cancellation rather than by a simple attractive potential (Zhu et al., 2015).

5. Inhomogeneous and asymmetric ladders near half-filling

A distinct route away from the uniform ladder is mesoscale hopping modulation. In the checkerboard ladder, the leg hoppings alternate as

$t$6

so the unit cell is a $t$7 plaquette. At exact half-filling the standard two-leg ladder is taken as a spin-gapped Mott insulator; close to half-filling the checkerboard ladder remains in the Luther-Emery class but with substantially enhanced pairing. At $t$8 and $t$9, the spin gap reaches

$U$0

near $U$1, roughly four times the spin gap of the uniform ladder at the same $U$2 and $U$3. Combining the spin-gap and pair-binding data gives an optimal region

$U$4

with

$U$5

In the same regime $U$6 and the ratio $U$7 is close to unity, indicating that the enhanced pairing scale is not offset by a strongly suppressed phase stiffness (Karakonstantakis et al., 2010).

An asymmetric half-filled ladder realizes a different deformation: one leg is a Hubbard chain and the other a noninteracting one-dimensional electron gas. In that model, four phases appear as functions of $U$8 and rung hopping $U$9: a Luttinger liquid at very weak $$\langle n_{\uparrow}(i,j) + n_{\downarrow}(i,j) \rangle = 1$$0; a Kondo-Mott insulator at moderate $$\langle n_{\uparrow}(i,j) + n_{\downarrow}(i,j) \rangle = 1$$1 or strong $$\langle n_{\uparrow}(i,j) + n_{\downarrow}(i,j) \rangle = 1$$2; a spin-gapped paramagnetic Mott insulator with incommensurate excitations and pairing of doped charges at intermediate $$\langle n_{\uparrow}(i,j) + n_{\downarrow}(i,j) \rangle = 1$$3 and $$\langle n_{\uparrow}(i,j) + n_{\downarrow}(i,j) \rangle = 1$$4; and a correlated band insulator at large $$\langle n_{\uparrow}(i,j) + n_{\downarrow}(i,j) \rangle = 1$$5. The three gapped phases are distinguished by the momenta of their lowest single-particle excitations: $$\langle n_{\uparrow}(i,j) + n_{\downarrow}(i,j) \rangle = 1$$6 in the Kondo-Mott phase, incommensurate $$\langle n_{\uparrow}(i,j) + n_{\downarrow}(i,j) \rangle = 1$$7 and $$\langle n_{\uparrow}(i,j) + n_{\downarrow}(i,j) \rangle = 1$$8 with $$\langle n_{\uparrow}(i,j) + n_{\downarrow}(i,j) \rangle = 1$$9 in the spin-gapped Mott phase, and $U/t$0 in the band insulator. This makes clear that leg asymmetry can qualitatively reorganize half-filled ladder physics even when the lattice still contains only two legs (Abdelwahab et al., 2014).

6. Other half-filled ladder geometries: ferrimagnetism, flat bands, and flux-generated spin models

Alternating ladders show how half-filled behavior depends on sublattice topology. In the 3–2 ladder, the lattice is bipartite with

$U/t$1

so Lieb’s theorem gives

$U/t$2

DMRG confirms an exactly degenerate ground-state multiplet for $U/t$3, a finite charge gap, and vanishing spin gap in the thermodynamic limit, together with ferrimagnetic long-range order whose magnetization is concentrated mainly on one leg. In the 3–3–2–2 ladder, by contrast, $U/t$4, the half-filled ground state is a singlet, the spin sector is gapped, and the pair-binding energy is finite; strong singlet bond order along one leg leaves an effective two-leg ladder on the remaining legs (Essalah et al., 2021).

Geometric frustration adds a different layer of structure. For a half-filled triangular ladder with spin-dependent hopping and spin-dependent flux, a Schrieffer-Wolff expansion to third order in $U/t$5 gives an effective spin Hamiltonian of the form

$U/t$6

that is, an anisotropic $U/t$7 Heisenberg ladder with Dzyaloshinskii-Moriya interaction, an extended magnetic field, and an unconventional three-spin correlated-exchange term. In the spin-symmetric-flux limit it reduces to Heisenberg exchange plus scalar spin chirality,

$U/t$8

For attractive $U/t$9, the same structure reappears in pseudospin variables through a one-spin-component particle-hole transformation (Garuchava et al., 2024).

A trimer ladder gives yet another half-filled scenario. There the unit cell contains three sites, the noninteracting spectrum can host a nearly flat middle band, and exact diagonalization, DMRG, and perturbation theory find five phases in the $U$0 plane: ferrimagnetic insulator, insulating cell spin-density wave, metallic Tomonaga-Luttinger liquid I, metallic Tomonaga-Luttinger liquid II, and variable spin magnetic insulator. The nearly flat middle band localizes charge at small $U$1, while moderate $U$2 and $U$3 produce metallic TLL behavior; this shows that half-filling plus a three-site motif can stabilize either localized ferrimagnetic physics or metallic TLL behavior, depending on how flat-band localization competes with intercell dispersion (Saha et al., 4 Mar 2025).

Taken together, these results show that half-filled Hubbard ladders are a unifying framework rather than a single model class. Uniform two-leg ladders emphasize Mott and rung-singlet physics; generalized-Hund and parity-brane formulations expose the full taxonomy of competing insulating orders; single-hole studies reveal phase-string dynamics; and decorated, alternating, asymmetric, triangular, and trimer ladders demonstrate that sublattice imbalance, frustration, and flat bands can convert the half-filled problem into ferrimagnetic, chiral, or metallic regimes without abandoning the Hubbard ladder setting.