(Challenge) Validity of tiling methods. Show using reasoning that each of the three tiling methods we saw produces a tiling of the plane using the given 4-gon. You have to prove that when we place new copies using any of the three procedures, the entire plane is covered with no gaps and no overlaps. First, think carefully about what it means to prove this. Then find a proof for each procedure.
Quadrilaterals · Question 39 of 40
(Challenge) Validity of tiling methods.
What must be proved. A tiling means: (1) no gaps — every point of the plane lies in some copy; (2) no overlaps — no two copies share any inside point (they may share edges or corners). For a procedure that goes on forever, we must show that the copies it places are consistent (placing the same copy in two ways gives the same position), and that together they satisfy (1) and (2).
Setup. Let the 4-gon be SOME (taken convex here). Write for the shift from E to O (along diagonal EO, by its length) and for the shift from S to M. Let be SOME, and be SOME turned through about Y, the midpoint of OM. Let W, X, Z be the midpoints of ES, SO, ME. By the Midpoint Theorem, WXYZ (the Varignon parallelogram) has sides , and , .
Step A: all three methods place the same copies, namely the copies of shifted by and the copies of shifted by , for all whole numbers , .
-
Method 3 (Exercise 18) places the shifts of by directly, and its blank spaces are the copies of .
-
Method 1 turns copies through about edge midpoints. Turning about Y gives . Turning about the midpoint of its edge that is the image of SO gives back SOME shifted by ; using the images of ME and ES gives shifts by and by . (Two half-turns, about points K and L, always give a shift by twice the distance from K to L, in the direction from K to L; for example the centres Y and are apart.) So every copy Method 1 places is one of the copies above, and a copy reached in two ways is a shift by the same amount either way — this is why "both ways give the same result".
-
Method 2 places copies on the Varignon grid; these are the same copies, as Step B shows.
So it is enough to prove that these copies form a tiling.

Step B: the Varignon grid. Copies of WXYZ shifted by (, whole numbers) tile the plane — this is the parallelogram grid tiling from the introduction. Call the cell :
-
, both even: the Varignon parallelogram of a copy of ;
-
, both odd: the Varignon parallelogram of a copy of (the half-turn about Y carries WXYZ to the cell shifted by );
-
one even, one odd: a "gap cell".
Each copy of SOME is made of its Varignon parallelogram plus 4 corner triangles: WSX at S, XOY at O, YMZ at M, ZEW at E. The corner triangle at S stands on side WX, outside the parallelogram, so it lies in the neighbouring cell across WX, which is a gap cell. The same holds for every corner triangle of every copy.
Step C: each gap cell is filled exactly once. Take the gap cell G just below WXYZ (across WX). Exactly four copies have a side of G as a side of their Varignon parallelogram, and their corner triangles on those sides are:
-
the corner at S of (on WX), with apex S;
-
the corner at M of shifted by , with apex ;
-
the corner at O of shifted by : the half-turn sends O to M, so its apex is again S;
-
the corner at E of shifted by , whose apex is also S.
So these four triangles have the same apex S, their bases are the four sides of G, neighbouring triangles share an edge from S, and their angles at S are , , , , which add to . They fit round S with no gap and no overlap and fill G exactly. By shifting, every gap cell is filled in the same way.
Conclusion. Every cell of the Varignon grid is covered exactly once: a parallelogram cell by its own copy, a gap cell by four corner triangles. The cells tile the plane, so the copies cover the plane (no gaps) and no two copies share inside points (no overlaps). By Step A this proves that all three methods produce a tiling.
(We have proved this for a convex SOME. For a non-convex SOME, Step A is unchanged and the Varignon grid still exists (Exercise 13), but at the reflex corner the corner triangle lies inside the copy's own Varignon parallelogram: it is cut out of that cell instead of being added to a gap cell, so Step C has to be redone with a different count of which pieces cover each cell. Experiment (Exercise Set 12.4, Q2) shows that all three methods still tile the plane with a non-convex 4-gon, but we do not carry out the full proof for that case here.)
