State diagrams for harmonically trapped bosons in optical lattices

We use quantum Monte Carlo simulations to obtain zero-temperature state diagrams for strongly correlated lattice bosons in one and two dimensions under the influence of a harmonic confining potential. Since harmonic traps generate a coexistence of superfluid and Mott insulating domains, we use local quantities such as the quantum fluctuations of the density and a local compressibility to identify the phases present in the inhomogeneous density profiles. We emphasize the use of the “characteristic density” to produce a state diagram that is relevant to experimental optical lattice systems, regardless of the number of bosons or trap curvature and of the validity of the local-density approximation. We show that the critical value of U/t at which Mott insulating domains appear in the trap depends on the filling in the system, and it is in general greater than the value in the homogeneous system. Recent experimental results by Spielman et al. [Phys. Rev. Lett. 100, 120402 (2008)] are analyzed in the context of our two-dimensional state diagram, and shown to exhibit a value for the critical point in good agreement with simulations. We also study the effects of finite, but low (T<t/2), temperatures. We find that in two dimensions they have little influence on our zero-temperature results, while their effect is more pronounced in one dimension.

💡 Research Summary

In this work the authors present a comprehensive quantum‑Monte‑Carlo (QMC) study of strongly interacting bosons confined in optical lattices with an additional harmonic trapping potential. By focusing on the zero‑temperature limit they map out the full “state diagram” for both one‑dimensional (1D) and two‑dimensional (2D) systems, showing where superfluid (SF) and Mott‑insulating (MI) domains appear within the inhomogeneous density profiles generated by the trap.

The model is the standard Bose‑Hubbard Hamiltonian with tunnelling amplitude t, on‑site repulsion U, and a site‑dependent harmonic term V_i = V_t r_i^2. The QMC simulations employ the worm‑algorithm (or a related “wild‑card” technique) to obtain unbiased ground‑state observables for large particle numbers N and a range of trap curvatures V_t. To distinguish phases locally the authors compute three key quantities at each lattice site i: the average occupation ⟨n_i⟩, the density fluctuation ⟨(Δn_i)^2⟩, and the local compressibility κ_i = ∂⟨n_i⟩/∂μ_i. Regions where fluctuations are strongly suppressed and κ_i≈0 are identified as MI, while finite fluctuations and compressibility signal SF behavior.

A central conceptual advance is the introduction of the “characteristic density” ρ̃ = N (V_t/t)^{d/2}, where d is the dimensionality. This dimensionless combination simultaneously incorporates the total atom number and the trap strength, rendering the phase diagram universal: systems with different N and V_t but the same ρ̃ collapse onto a single curve in the (ρ̃, U/t) plane. Consequently, experimentalists can locate their parameters on the same diagram without needing to perform separate calculations for each trap configuration.

The QMC results reveal that the critical interaction strength for the appearance of a MI core in a trapped system is systematically larger than the homogeneous critical value (U_c/t≈3.3 in 1D, ≈16.7 in 2D). The shift arises because the harmonic potential produces a spatially varying chemical potential; only when the local density at the trap centre reaches exactly one particle per site does a true MI plateau form. In 2D, the first MI core emerges at ρ̃≈13, while in 1D it appears around ρ̃≈5.

Finite‑temperature effects were also examined by running simulations at low but non‑zero temperatures (T < t/2). In 2D the temperature‑induced changes in the local observables are negligible, confirming that the zero‑temperature state diagram remains accurate for typical experimental conditions. In 1D, however, thermal fluctuations modestly increase density variance and smooth the MI–SF boundaries, reflecting the enhanced sensitivity of one‑dimensional systems to thermal excitations.

To validate the theoretical framework, the authors compare their 2D state diagram with the experimental data of Spielman et al. (Phys. Rev. Lett. 100, 120402, 2008). The experiment reports the onset of a Mott plateau at U/t≈20 for a system with a known trap curvature and atom number. When plotted on the authors’ (ρ̃, U/t) diagram, this point lies precisely at the predicted MI boundary, demonstrating excellent quantitative agreement.

In summary, the paper delivers a robust, experimentally relevant phase diagram for harmonically trapped bosons in optical lattices. By employing local density fluctuations and compressibility as diagnostics, and by introducing the characteristic density scaling, the authors bridge the gap between idealized homogeneous theory and real trapped‑atom experiments. Their findings clarify how trap‑induced inhomogeneity shifts the MI transition, quantify the modest role of low temperatures (especially in 2D), and provide a practical tool for designing and interpreting future cold‑atom studies of strongly correlated bosonic matter.