3
$\begingroup$

Consider a $\mathbb{Z}^d$ lattice whose edges connect nearest-neighbor sites only, i.e. a $d$-dimensional hypercubic grid. Let $C$ be a closed curve along such edges. In general, for $d>2$ such curves are non-planar. If an elementary square is given area 1, is there any formula that computes the minimal area bounded by a generic lattice curve $C$? For example, the following curve has minimal area of 2:

Lattice curve

Edit: Precisely, what I would like to know is if it is possible to decide whether a generic curve $C$ bounds an even or odd area. For example, the curve above has area 2, which is even. I suspect that parity is a property of the curve and is preserved in non-minimal areas, but I don't know how to prove it.

$\endgroup$
4
  • 1
    $\begingroup$ What area would you give to the six edge curve that passes through all but two diametrically opposed points on a 3d cube (thus different from your example)? $\endgroup$ May 2, 2014 at 15:25
  • $\begingroup$ The minimal area of your example would be 3, which is the sum of the areas of 3 adjacent faces of the 3d cube. But my question is simpler (see edit above): I just care to know that the example you gave have an odd minimal (and probably not just minimal) area. $\endgroup$
    – FreeQuark
    May 2, 2014 at 15:30
  • 1
    $\begingroup$ If I am restricted to lattice planes for area, then I agree. (For me there is the issue of bringing up knotted curves, but I won't do that now.) If I am not so limited, I get a smaller area using a folded unit hexagon, and I can go smaller still with triangles. $\endgroup$ May 2, 2014 at 15:49
  • 1
    $\begingroup$ The elementary square has area 1. Take any planar curve with even area A, glue to it an adjacent square and you get a curve with area A+1. $\endgroup$
    – FreeQuark
    May 3, 2014 at 14:36

3 Answers 3

4
$\begingroup$

For any $\mathbb Z_2$-valued function $f$ on the set of segments $[v,v+e_i]$ for $v\in \mathbb Z^3$ and $i=1,2,3$, and any $S$ that bounds $C$ $$ \sum_{P\in S}\sum_{edge\in P} f(edge) $$ is independent of the choice of $S$ (it is equal to the sum of $f$ over the segments of $C$).

Thus to prove the independence of parity of the area from the choice of $S$, it suffices to find a function on the set of (directed) segments $[v,v+e_i]$ such that the sum of it over edges of any size one lattice square is $1\in \mathbb Z_2$.

One way of thinking about such $f$ is the following homological algebra construction. If you naively tensor three complexes, then the corresponding squares will commute rather than anticommute. However, it is possible to change some of the signs to make the total complex to be an actual complex. The edges that needed to be altered are the ones that will have $f=1$.

For a more elementary argument, we will declare $f([v,w])$ to be $1$ if and only if $[v,w]$ is equal modulo $2$ to one of the following: $$ [(0,0,0),(1,0,0)],~[(0,0,1),(0,1,1)],~[(0,1,1),(1,1,1)], $$ $$ [(1,0,0),(0,0,0)],~[(0,1,1),(0,0,1)],~[(1,1,1),(0,1,1)], $$ I will now argue that every lattice square contains exactly one such edge.

Note that the set above is preserved by translations by elements of $2\mathbb Z^3$. It is also preserved by symmetries across lattice planes, as these have an effect of switching $[v,w]$ and $[w,v]$ modulo $2$ or preserving $[v,w]$ modulo $2$. Every lattice square can be moved into one of the six facets of the standard cube $[0,1]^3$ by the above shifts and symmetries. It remains to observe that each of the faces of the standard cube contains exactly one edge from the first line and none from the second line.

$\endgroup$
2
  • $\begingroup$ Thank you for your answer! I am not versed in homological algebra, but I think I understood the latter argument. I think that in your list of special edges, the last ones should be $[(1,1,0),(1,1,1)]$ and $[(1,1,1),(1,1,0)]$? Otherwise one the faces of $[0,1]^3$ does not contain any of the edges above. Also, do you think this proof generalizes to higher dimensions (I am particularly interested in $\mathbb{Z}^4$). It seems to depend only on the choice of a list of segments like the ones you wrote. $\endgroup$
    – FreeQuark
    May 8, 2014 at 14:53
  • $\begingroup$ Yes, you are right. Thank you for the correction! $\endgroup$ May 8, 2014 at 14:56
3
$\begingroup$

This is not a full answer, but at least it can be proved that the parity is kept using few basic properties of homology of cubical complexes. (See also the last paragraph "After edit" discussing how to determine this parity.)

Let $S_1$ and $S_2$ be two surfaces bounded by $C$. The symmetric difference $S$ of $S_1$ and $S_2$ is a cycle in $\mathbb Z_2$-chain complex of the grid. [If you are not familiar with homology, think of $S$ as a multicomponent possibly self-intersecting $2$-surface with empty (!) boundary.] The fact that $S_1$ and $S_2$ have the same parity is equivalent with showing that the number of squares forming $S$ is even.

Since $S$ is a cycle and the homology of the cubical grid is trivial, it is also a boundary of some $3$-chain over $\mathbb Z_2$. That is, there are cubes $Q_1, \dots, Q_k$ such that the boundary of the union $Q$ of these cubes is $S$. Start removing these cubes from $Q$ one by one. (That is, make the symmetric differences.) In each single removal, the parity of the number of squares of the boundary does not change, since a single cube has an even number of squares. After removing all the cubes, the boundary is empty. Therefore the boundary of $Q$ has to consist of an even number squares.

After edit: This homology approach should essentially also answer whether a given curve $C$ bounds an even or an odd area. It is just necessary to find a 2-chain $S'$ such that the boundary of $S'$ is $C$, perhaps not a minimal one. (In some bounded subcomplex of the grid containing $C$ in order to make the computations finite.) This should be then just some matrix manipulation.

$\endgroup$
2
  • $\begingroup$ Thank you for your answer! I am not too familiar with homological algebra, but I understood your argument. Does it extend to higher dimensions, $d=4$ in particular? And for hypertoroidal lattices I think your argument would be valid for the set of contractible lattice curves too, wouldn't it? $\endgroup$
    – FreeQuark
    May 8, 2014 at 15:15
  • $\begingroup$ Yes, it works in higher dimensions as well. I did not emphasize it, but the proof is written for arbirtrary $d \geq 3$. I am not sure what you exactly mean with hypertoroidal lattice. But if I guess well, the contractibility of curves is not important. It is more important whether you can have a nontrivial $2$-cycle ($S$ in the answer) such that it has an odd number of squares. It seems to me to be possible in some hypertoroidal lattices of some appropriate dimensions such as $3 \times 3 \times 3$. $\endgroup$ May 8, 2014 at 15:38
2
$\begingroup$

You might look at this work by John Sullivan: "Discrete minimal surfaces in the cubic lattice," and his lecture on the topic here:
                Lattice Polyhedron
Incidentally, another version of the Schwartz $P$-surface is at Alan Schoen's link:
  Schoen Polyhedron

$\endgroup$
2
  • $\begingroup$ Hi! Thanks for the links! They seem quite interesting, but I'll need some time to sieve through them. In fact, I would like to know the answer of a (seemingly) simpler question, which I edited above. Do you have any knowledge on how to approach the problem of deciding whether areas bounded by a lattice curve are even or odd? $\endgroup$
    – FreeQuark
    May 2, 2014 at 15:28
  • $\begingroup$ @helvio: Thanks for sharpening your question to parity. That is an interesting conjecture, that all lattice surfaces bound by the same curve have the same area parity. I do not know if it is true or false; it is certainly worth investigation. $\endgroup$ May 2, 2014 at 15:53

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.