[ Back to EurekAlert! ]

PUBLIC RELEASE DATE:
8-Jun-2007

[ | E-mail ] Share Share

Contact: Mike Breen
paoffice@ams.org
401-455-4000
American Mathematical Society
@amermathsoc

Solving sudokus -- Coloring by numbers

Providence, RI -- Have you ever been trying to solve a Sudoku puzzle and been gripped by a sinking feeling that maybe you were stuck with a lemon" That maybe the puzzle you are struggling with actually has no solution at all" And, if you do find a solution, how can you be sure it's the only one" What if half an hour ago you had written 5 instead of 3---would you then have gone down a path to a completely different solution"

These questions and others are explored in the article "Sudoku Squares and Chromatic Polynomials", by Agnes M. Herzberg and M. Ram Murty, which appears in the June/July 2007 issue of the Notices of the AMS. The authors use tools from the branch of mathematics called graph theory to systematically analyze Sudoku puzzles. They also find that analyzing Sudokus leads to some unsolved problems in graph theory.

In this context, a "graph" is a collection of nodes connected by line segments (this object is also sometimes called a "network"). We can think of the 81 squares in a Sudoku puzzle as 81 nodes in a graph. We will also represent each of the numbers one through nine by a different color. In a Sudoku graph, two nodes are connected by a line segment if the two squares they represent are in the same row, column, or 3 by 3 subgrid. Since no row, column, or 3 by 3 subgrid can contain more than one instance of each number, the graph will have no connected nodes of the same color. (For example, suppose we represent 1 with red. Two red nodes connected with a line segment would mean a row, column, or 3 by 3 subgrid had two 1's in it, which is forbidden in Sudoku.)

In the language of graph theory, a colored graph with no connections between same-colored nodes is a called a "proper coloring." What Sudoku solvers do every day is try to extend a partially-colored graph (the original puzzle with open squares means the graph representing it has yet-to-be-colored nodes) to a proper coloring.

With this analogy between Sudoku puzzles and graphs in place, Herzberg and Murty are able to use tools from graph theory to prove theorems about Sudokus. For example, they prove that the number of ways of extending a partial coloring of a graph is given by a polynomial. If the value of this polynomial is zero for a given Sudoku puzzle, then the puzzle has no solution; if the value is 1, then the puzzle has only one solution; and so forth. They also prove that, in order for any Sudoku puzzle to have only one solution, at least 8 of the 9 numbers must appear as given entries in the puzzle; if only 7 numbers are given, then the puzzle has at least two solutions. And this brings up an unsolved mathematical question: "It would be extremely interesting to determine under what conditions a partial coloring can be extended to a unique [proper] coloring," Herzberg and Murty write.

Some Sudokus are harder than others, with the really difficult ones containing very few given entries. What is the minimum number of given entries needed to ensure that a puzzle has a unique solution" Herzberg and Murty give an example of a Sudoku puzzle with just 17 given entries that has only one solution. So the minimum number is at least 17. But could it be 16---or something even smaller" No one knows. One might think that a puzzle with many given entries is likelier to have a unique solution, but this is not necessarily the case. The article gives an example of a puzzle with 29 given entries that actually has two different solutions.

And if you have been wondering when your newspaper will run out of Sudoku puzzles, the authors argue that the number of distinct Sudoku puzzles is somewhere around 5.5 billion---enough to keep Sudoku afficianados busy for a long time to come.

###

The article "Sudoku Squares and Chromatic Polynomials" is available on the web site of the AMS Notices at the URL http://www.ams.org/notices/200706/tx070600708p.pdf.

Founded in 1888 to further mathematical research and scholarship, the more than 30,000-member American Mathematical Society fulfills its mission through programs and services that promote mathematical research and its uses, strengthen mathematical education, and foster awareness and appreciation of mathematics and its connections to other disciplines and to everyday life.

American Mathematical Society
201 Charles Street
Providence, RI 02904
401-455-4000



[ Back to EurekAlert! ] [ | E-mail Share Share ]

 


AAAS and EurekAlert! are not responsible for the accuracy of news releases posted to EurekAlert! by contributing institutions or for the use of any information through the EurekAlert! system.