Title:Trivalent network model for d3 transition metal dichalcogenides in the 1T structure: Distortions from an effective boundary theory

Abstract:Dimer models are well known as prototypes for locally constrained physics. They describe systems where every site on a lattice must be attached to one dimer. Loop models are an extension of this idea, with the constraint that two dimers must touch at each site. Here, we present a further generalization where every site must have three dimers attached -- a trivalent network model. As concrete physical realizations, we discuss d$^3$ transition metal dichalcogenides in the 1T structure -- materials with the structural formula MX$_2$ (M = Tc, Re) or AM$'$X$_2$ (A = Li or Na; M$'$ = Mo, W), where X is a chalcogen atom. These materials have a triangular layer of transition metal atoms, each with three valence electrons in $t_{2g}$ orbitals. Each atom forms valence bonds with three of its nearest neighbours. The geometry of the 1T structure imbues each bond with sharp orbital character. We argue that this enforces a `bending constraint' where two dimers attached to the same site cannot be parallel. This leads to a highly structured space of configurations, with alternating bonds along each line of the underlying triangular lattice. There is no dynamics, as constraints forbid local rearrangements of dimers. We construct a phase diagram, identifying configurations that minimize potential energy. We find a rhombus-stripe phase in a wide region of parameter space that explains a distortion pattern seen across several materials. Remarkably, this model can be recast as an effective boundary theory, in terms of three Ising chains that are coupled by mutual long-range interactions. As a testable prediction, we propose that a single impurity will generate long-ranged domain walls.