Mathematics of 2-dimensional lattices (Q6566149)

Periodic lattices in the plane might seem to be rather simple: in one sense, they are all just linear images of the integer lattice \(\mathbb{Z}^2\). However, this simple description is problematic: every lattice is a linear image of \(\mathbb{Z}^2\) in infinitely many ways. (For instance, \(\{(1,0),(0,1)\}\) and \(\{(1,0),(1,1)\}\) generate the same lattice.) Things get even more complicated when we try to classify lattices up to isometry: the basis \(\{(\frac{1}{\sqrt{2}},\frac{1}{\sqrt{2}}),(\frac{1}{\sqrt{2}},\frac{-1}{\sqrt{2}})\}\) generates the same lattice rotated through \(45^\circ\). We call the quotient by isometry the ``Lattice Isometry Space'': the problem is to find a representation of this with good properties.\N\NVoronoi (``nearest-neighbour'') domains give a partial solution: two lattices are isometrically equivalent if and only if they have congruent Voronoi domains (in the case above, a unit square). However, this approach has a problem with continuity: arbitrarily small perturbations may change the combinatorial type of the Voronoi domain, making that square into an irregular hexagon. Other representations have more subtle problems relating to computability and reconstructing the set of bases. This paper presents an explicit space of complete invariants representing the Lattice Isometry Space, continuous in a metric computable from the bases in constant time, and permitting explicit reconstruction of those bases. This is a surprisingly major undertaking (see the ``roadmap'' on page 814).\N\NIt is unfortunate that a typesetting error has resulted in the inscrutable string ``bright2021proof'' appearing no less than fourteen times in the paper, and that this very visible mistake was missed during proofreading. Inspection suggests that this should be replaced, wherever it appears, by the word ``images''.