Competition between charge-density-wave and superconducting orders on eight-leg square Hubbard cylinders
Hong-Chen Jiang, Thomas P. Devereaux, Steven A. Kivelson
TL;DR
This study addresses whether $d$-wave superconductivity emerges in the square-lattice Hubbard model at intermediate coupling by performing large-scale, $SU(2)$-symmetric DMRG on eight-leg square cylinders with varying $t'$ and hole doping $\delta$. The authors find a CDW-dominated ground state for $t'\le 0$, with exponentially decaying SC correlations, while $t'>0$ produces strong boundary-condition dependence that can enhance SC in some cases but does not robustly realize long-range $d$-wave order in this geometry. The work highlights intertwined competing orders and demonstrates that accurate, symmetry-respecting DMRG with very large bond dimensions is essential to distinguish CDW, SDW, and SC tendencies on quasi-2D lattices. The results suggest that achieving robust high-temperature SC may require additional model ingredients or parameter regimes beyond the studied eight-leg cylinder, guiding future explorations of cuprate-like physics in Hubbard-type models.
Abstract
The issue of whether $d$-wave superconductivity (SC) occurs in the square-lattice Hubbard model with $U$ of order of the bandwidth has been one of the most debated issues to emerge from the study of high temperature SC. Here, we report variational results on eight-leg cylinders with next-nearest-neighbor hopping in the range $-0.5 t \leq t'\leq 0.25 t$ with $U = 8t$ and $12t$ and doped hole concentrations $δ=1/12$ and $1/8$. For $t'\leq 0$, the ground-state appears to be a charge-density wave (CDW) of one sort or another with SC correlations that are extremely short-ranged. In contrast, in some cases, the local magnetic order has a correlation length greater than half the cylinder width - suggestive that magnetic order might also arise in the 2D limit. For $t'>0$, our results depend more strongly on boundary conditions (periodic vs antiperiodic), making it still harder to correctly guess whether SC or CDW correlations dominate in the 2D limit. These results were obtained employing matrix-product states with bond dimensions large enough that energy differences as small as $10^{-3}t$ per site can be resolved.
