Take the 2-minute tour ×
Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

Given an $n \times n$ square grid graph, what is the maximum number of turns a Hamiltonian path can take?

share|improve this question
    
The same question would seem to be of interest for a Hamiltonian circuit. –  Joseph Malkevitch Aug 12 '11 at 14:37
add comment

1 Answer 1

up vote 2 down vote accepted

Answer to OP's question (proof follows):

For Hamitlonian paths: if $n$ is even max turns=$n^2-n$

if $n$ is odd max turns equals $n^2-n-1$

The minimum number of turns on an Hamiltonian path on an $n \times n$ lattice is $2n-2$ which can be constructed at least two non-isomorphic ways (rows or spiral) for each graph. The maximum number of turns Hamiltonian path onin a $2 \times n$ lattice is also $2n-2$

To maximize the number of turns, if $n$ is even then you can consider the graph to be a collection of $n \times 2$ graphs each of which can have a maximum of $2n-2$ turns in them starting from an outside corner on each side the $-2$ factor comes from the fact that in order to turn the corner you must have one vertex that does not turn, for an outside $2 \times n$ you lose one going into the corner and one from the starting/ending vertex. For the interior $2 \times n$'s you must have one "straight vertex coming in and one coming out.

$2n-2 \times (n/2)$seperate rectangles gives a formula of $n^2-n$ for even $n$.

If $n$ is odd, it is not so simple to construct a path because the rows do not divide evenly by $2$. The cases for constrcution of a maximal turning path are different if $n$ is 1 or 3 mod $4$.

If $n=1$ mod $4$ you begin again at a corner in the first row you again have $2n-2$ turning vertices in the first $2\times n$ rectangle. You then continue down the side of the graph (a $2\times (n-2)$ rectangle) and collect $2n-4-1$ more turning vertices. Repeat on the bottom of the graph. Continue towards the starting corner now you have a $2 \times (n-4)$ rectangle that yields $2n-8-1$ turning vertices. Continue on towards the center every two new directions you turn the rectangle get smaller and there are $4$ fewer vertices. When you reach the center (where the path ends) you gain only 1 turn for the final $3$ vertices.

Putting it all together there are $n-2$ turns in the first rectangle, $1$ turn for the center and the spiral which contains $n-7+4$ rectangles in sets of two beginning with $2n-5$ turns and adding $4$ less for every set. $2n-2+1+[(n-3)(\frac{2n+5-5}{2})]=n^2-n-1$

If $n=3$ mod $4$ then you can proceed as above but the actual center vertex of the graph is contained in the last rectangle before the end of the paththe last $3$ vertices and their $1$ turn remains the same but the path is off-centered in the graph. The formula for the vertices remains the same $2n-2$ for the first rectangle $n-5$ rectangles at an average of $n$ turns per spiral rectangle and one for the last turn of the path still sums to $n^2-n-1$ for all odd $n$.

Answer to question asked in comments. For Hamiltonian Cycles

If $n$ is odd there is no hamiltonian cycle on a $n \times n$ grid. This can be proved from the contrapositive of theorem 3.16 on page 214 of "Graphs and Digraphs" by Chartrand. By numbering the vertices of an odd $n \times n$ grid from left to right top to bottom from $1,2,3...n^2$ Select the set of vertices $S$ as all the even numbered vertices. Then $|S|=\frac{n^2-1}{2}$ and there are $\frac{n^2+1}{2}$ disconnected components so the graph cannot be hamiltonian.

If $n$ is even you can construct a "spiral" cycle like the spirals for the odd paths given above. I haven't found a closed form expression yet that deals with the center of the cycle. $(\frac{3n^2}{4}+\frac{3n}{2}-12)+C$ where $C$ is the number of turns at the center of the graph is a good description of the maximum number of turns in a hamiltonian cycle on a $n \times n$ grid.

share|improve this answer
    
Thanks for the response. I'm not sure I completely follow your $n$ odd case, but it seems similar to what I did. –  Michael Aug 12 '11 at 16:47
    
@Michael it is necessary to demonstrate that you have obtained the maximum number of turns for each segment of vertices in the graph. The case for $n$ odd is just a demonstration that by taking smaller and smaller connected $ 2\times n-4x$ rectangles, each rectangle having all vertices turn except the one that leads to the next rectangle you have achieved the maximum number of turns. –  user12998 Aug 12 '11 at 18:25
    
@Robert Please contact me at my first.lastname@gmail.com Please also see my comment here. –  Bill Dubuque Sep 8 '12 at 0:58
add comment

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

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