Abstract
We compute the ground-state phase diagram of the three-band Hubbard model in various published parametrizations using density matrix embedding theory (DMET). We study the atomic-scale nature of the antiferromagnetic (AF) and superconducting (SC) orders, explicitly including the oxygen degrees of freedom. While all parametrizations display AF and SC phases, the decay of AF order with doping is significantly slower than observed experimentally, and coexistence of these orders occurs across all parameter sets. Our work highlights the importance of oxygen degrees of freedom and the need for re-evaluating current model parametrizations.
Introduction
The mechanism behind high-temperature superconductivity in cuprates remains a central challenge in condensed matter physics. A key debate centers on whether the simplified one-band Hubbard model is sufficient or if the three-band Hubbard model (Emery model), which explicitly includes oxygen $p$-orbitals, is necessary to capture the essential physics of the $CuO_2$ planes. The three-band model is more realistic but presents a vast and difficult-to-solve parameter space.
Methodology: Density Matrix Embedding Theory
To address this, we utilize Density Matrix Embedding Theory (DMET), a powerful numerical method that embeds a small, highly correlated cluster (here, a $2 \times 2$ unit of $CuO_2$) within a self-consistently determined environment. The local physics of the cluster is solved using the Density Matrix Renormalization Group (DMRG), allowing for a precise treatment of both copper and oxygen degrees of freedom. This approach enables the simultaneous investigation of antiferromagnetic (AFM) and superconducting (SC) orders.
The Ground-State Phase Diagram
Our calculations across various standard parameter sets reveal a consistent but problematic phase diagram. In all cases, AFM and SC orders coexist over a broad range of doping.
As shown in the figure above, the AFM order parameter $m_{AFM}$ decays linearly with doping but persists to much higher concentrations ($x \approx 0.3$ in minimal models) than seen in experiments, where magnetism typically vanishes by $x \approx 0.05-0.10$. Similarly, the SC order parameter $m_{SC}$ peaks at higher doping levels (~30-35%) than the experimental optimal doping of ~16%.
Sensitivity to Model Parameters
A significant insight from this study is the sensitivity of the magnetic phase to specific Hamiltonian parameters. We find that the AFM order is highly dependent on the charge-transfer gap ($\Delta_{pd}$), the energy difference between the oxygen $p$ and copper $d$ levels.
The 3D visualization demonstrates that increasing $\Delta_{pd}$ significantly stabilizes the AFM phase, confirming that these systems behave as charge-transfer insulators. Interestingly, the magnetic order is relatively insensitive to the on-site repulsion $U_d$, suggesting that the charge-transfer physics is the primary driver of the phase boundaries. Furthermore, the asymmetry between hole-doping ($x > 0$) and electron-doping ($x < 0$) is evident, with electron-doping showing a more complex magnetic landscape.
Atomic-Scale Insights and Conclusions
Beyond the phase boundaries, DMET reveals "hidden" atomic-scale symmetries. We observed a local $p_{x(y)}$-symmetry modulation of the pair density on the Cu-O bonds within the larger $d_{x^2-y^2}$-wave SC order. This underscores the active role of oxygen in the superconducting state.
However, the quantitative mismatch between our theoretical phase diagrams and experimental data suggests that current standard parametrizations of the three-band Hubbard model may be incomplete. Future work must explore larger clusters to account for long-range orders like stripes and consider re-parameterizing the model to better reflect the experimental reality of cuprate superconductors.
Summary updated