Unhappy vertices in artificial spin ice: new degeneracies from vertex frustration

In 1935, Pauling estimated the residual entropy of water ice with remarkable accuracy by considering the degeneracy of the ice rule solely at the vertex level. Indeed, his estimate works well for both the three-dimensional pyrochlore lattice and the two-dimensional six-vertex model, solved by Lieb in 1967. A similar estimate can be done for the honeycomb artificial spin. Indeed, its pseudo-ice rule, like the ice rule in Pauling and Lieb's systems, simply extends to the global ground state a degeneracy which is already present in the vertices. Unfortunately, the anisotropy of the magnetic interaction limits the design of inherently degenerate vertices in artificial spin ice, and the honeycomb is the only degenerate array produced so far. In this paper we show how to engineer artificial spin ice in a virtually infinite variety of degenerate geometries built out of non-degenerate vertices. In this new class of vertex models, the residual entropy follows not from a freedom of choice at the vertex level, but from the nontrivial relative arrangement of the vertices themselves. In such arrays not all of the vertices can be chosen in their lowest energy configuration. They are therefore vertex-frustrated and contain unhappy vertices. This can lead to residual entropy and to a variety of exotic states, such as sliding phases, smectic phases and emerging chirality. These new geometries will finally allow for the fabrication of many novel, extensively degenerate versions of artificial spin ice. (paper)