Definition. Square faces on a finite graph [fdip-006M]

A finite cubical 2-complex, in the square-cell convention used here, consists of a finite incidence presentation \((V,E,s,h)\), a finite set \(\mathcal P\) of squares, and a specified four-edge attaching walk for each square. Choose one orientation for each edge and one boundary orientation and starting vertex for each square. Its walk is \[\partial p=(e_{p,1}^{\epsilon _{p,1}},\ldots , e_{p,4}^{\epsilon _{p,4}}),\qquad \epsilon _{p,j}\in \{1,-1\},\] where an exponent \(-1\) means traversal against the chosen edge orientation. Consecutive endpoints match, including the last to the first.

Starting from the one-dimensional realization, attach one copy of \([0,1]^2\) for each \(p\), mapping its 4 sides to the specified traversed edges, and use the quotient topology. Boundary vertices or sides may be identified in this presentation, as in periodic quotients; no regularity or nonpositive-curvature condition is imposed. No higher-dimensional cubes are supplied. The squares and their attaching maps are additional data, just as the filled face is additional data in the triangle comparison.

The main example is the square torus with periods \(N_1,N_2\ge 3\). Its vertices are \[V=(\mathbb Z/N_1\mathbb Z)\times (\mathbb Z/N_2\mathbb Z).\] There are horizontal edges \(a_{x,y}:(x,y)\to (x+1,y)\) and vertical edges \(b_{x,y}:(x,y)\to (x,y+1)\), with indices read modulo the periods. The square at \((x,y)\) has boundary walk \[a_{x,y}\,b_{x+1,y}\,a_{x,y+1}^{-1}\,b_{x,y}^{-1}.\] This written order is the order of traversal. Put \[n=|V|=N_1N_2,\qquad |E|=2n,\qquad P=|\mathcal P|=n.\] The realization is the usual square grid on a torus. The periods ensure that each square has 4 distinct vertices and sides, although periodic identifications occur globally.