According to Claude, the example complex I went through on 2026-05-14 cannot be embedded in . I repeated the run on a modified complex that should be a triangulation of where is the Mobius strip.
Iteration 1-19: Phase 1 searching for initial basis
Iteration 20:

Iteration 21:

Iteration 22:

The remaining 7 iterations are the same solution but with equivalent bases.
The final basis matrix:

Which has Smith normal form
where
Exploration of SNF
At any feasible basis, is of the form
where each is a standard basis vector. Since order does not matter for basic variables, we assume WLOG that
Perform the following row swap operations
and
at this point
where and where
and is the permutation matrix corresponding to
We then cancel with column operations, affecting which we are not interested in. This leaves us with
The remaining operations only affect, . Let and such that . Then we get
Note here that where is the matrices to the left of . Note that and only affect the first columns. So
where and are the top left submatrices of and , respectively.
Therefore, our conditions only depend on , the submatrix of .