Doping a Mott insulator with excitons in a moiré bilayer: Fractional superfluid, neutral Fermi surface, and Mott transition

Abstract

In this paper we explore possible phases arising from doping neutral excitons into a Mott insulator in the context of moir\'e bilayers. We consider two moir\'e layers coupled together only through inter-layer repulsion and there is a U(2)$\times$ U(2) symmetry. The densities of the two layers can be tuned to be $n_t=x,n_b=1-x$ with $n_t+n_b=1$. $x=0$ limit is a layer polarized Mott insulator and small $x$ regime can be reached by doping inter-layer excitons at density $x$. Charge gap can remain finite at small $x$, as is demonstrated experimentally in the WSe$_2$-hBN-WSe$_2$/WS$_2$ system. To capture the intertwinement of the spin and exciton degree of freedom, we propose a four-flavor spin model. In addition to the obvious possibility of single exciton condensation phase, we also identified more exotic physics with fractionalization: (I) We define spin gap $\Delta_t, \Delta_b$ for the two layers respectively. As long as the spin gap at either layer is finite, single exciton condensation is impossible and we can only have paired exciton condensation. If both spin gaps are finite, it can be a fractional exciton superfluid with paired exciton condensation coexisting with $Z_2$ spin liquid. Numerical evidences for such a phase will be provided. (II) If the layer polarized Mott insulator at $x=0$ is in a U(1) spin liquid with spinon Fermi surface, the natural phase at $x>0$ hosts neutral Fermi surface formed by fermionic excitons. There are metallic counter-flow transport and also Friedel oscillations in layer polarization in this exotic phase. Our numerical simulation in one dimension observes an analog of this neutral Fermi surface phase. (III) There could be a metal-insulator transition (MIT) by tuning $x$ in the universality class of a bandwidth tuned MIT. We provide one theory for such a continuous Mott transition and predicted a universal drag resistivity.