9

A Sudoku puzzle is a 9 by 9 matrix of blanks(which we can represent as 0), and elements of the set {1,2,3,4,5,6,7,8,9}. How many Sudoku puzzles are there with at least one solution. Yes, I am even counting grids with no blanks.

Rodrigo de Azevedo
  • 23,223
user107952
  • 23,768
  • This seems to be an open problem. Here's a significant result: http://www.nature.com/news/mathematician-claims-breakthrough-in-sudoku-puzzle-1.9751 – Edward ffitch Jul 04 '14 at 18:17
  • Upper bound is $2^{81} \cdot 6,670,903,752,021,072,936,960$ – NovaDenizen Jul 04 '14 at 18:19
  • Let $S$ be the number of solutions states. Then we can bound the answer to your question above by $S(\binom{81}{0}+\binom{81}{1}+\cdots+\binom{81}{81})=2^{81}S$. By Felgenhauer and Jarvis $S\approx 6.671\times 10^{21}$ (they do give an exact number). – Peter Woolfitt Jul 04 '14 at 18:20
  • Just a side note: the typical number of givens is 17, which I believe as a minimum is a necessary but not sufficient condition to ensure a unique solution. I haven't done many sudokus, but alas! I have done a couple with two possible solutions that were published in national periodicals. – Robert Soupe Jul 04 '14 at 19:40

2 Answers2

5

Due to Felgenhaur and Jarvis

(see their paper here: http://www.afjarvis.staff.shef.ac.uk/sudoku/sudoku.pdf)

and Fangying, Mengshi, and Aslasksen

(see their paper here: http://www.math.nus.edu.sg/aslaksen/projects/SHI-ZHANG_Sudoku.pdf)

we know that the total possible number of legal 9x9 Sudoku enumerations is $N_{\text{enumerable}}=6670903752021072936960$.

I will provide an upper bound to the number of puzzles with at least one solution solution, $N_{\text{solvable}}$.

Summing up the number of ways we can replace $n$ squares on a given enumerated grid with $0$'s, and then multiplying by the number of legal grids gives

$$N_{\text{solvable}}<N_{\text{enumerable}}\cdot\sum_{n=0}^{81}\binom{81}{n}=6670903752021072936960\cdot2^{81}=16129255571964761146444296776129424146741329920\approx1.62\times10^{46}$$

This is only an upper bound because it will certainly be an over-count (this is easy to see), so I welcome any refinements to this upper bound as edits below my post.

  • 1
    Sorry, I see that this was posted as a comment while I was typing this out. My apologies @Peter Woolfitt ! –  Jul 04 '14 at 18:25
0

Building off of user 155385's answer we can get a lower bound as well as tighten the upper bound. You need at least 17 clues to solve a sudoku puzzle, and it's believed (albeit not proven) that any more than 40 clues uniquely determine a solution (https://en.wikipedia.org/wiki/Mathematics_of_Sudoku#Maximum_number_of_givens).

If we trust this maximum value of 40, then a lower bound is $$N_{enumerable}*(\binom{81}{81} +\binom{81}{80} + ... + \binom{81}{41}) = N_{enumerable}*1208925819614629174706176$$

Where we are choosing grids with 81 clues filled in, then 80 clues filled in, and so on.

We can also sharpen the upper bound slightly to

$$N_{enumerable}*(\binom{81}{81} +\binom{81}{80} + ... + \binom{81}{17}) = N_{enumerable} * 2417851466759455344675448.$$ The ratio between lower and upper bounds here is almost exactly 1:2.

Alex Li
  • 548