How can we determine the number of magic squares with magic constant 0?

  • 5
    $\begingroup$ What's the range, from which the square entries are taken? $\endgroup$
    – Gamow
    Mar 2, 2016 at 10:41
  • 4
    $\begingroup$ Can you specify the properties you want your squares to have? Must the numbers be consecutive, what size may the squares be? $\endgroup$ Mar 2, 2016 at 14:47
  • 1
    $\begingroup$ Any odd order $n$ magic square using entries $1,2,3,\dots n^2$ can be adapted by subtracting $(n^2+1)/2$ from all entries to make a sum of $0$. There are many of them of any size greater than $3 \times 3$, even if you remove ones related by reflection/rotation. Please think about your question. $\endgroup$ Mar 3, 2016 at 0:25

2 Answers 2


There are an infinite number of such squares

Take the example square below:

-3  2  1
 4  0 -4
-1 -2  3

To generate a new square, simply multiply each element of this square by any positive integer. As there are an infinite number of positive integers, there are an infinite number of possible squares.

  • 1
    $\begingroup$ The same is true about negative numbers since you'd be just switching parity. Because there is a requirement that magic squares be comprised of unique numbers, you can't use 0, but any non-zero integer would work. $\endgroup$ Mar 2, 2016 at 14:13
  • 1
    $\begingroup$ To my knowledge, a magic square has to use the numbers n,n+1,...,m-1,m. For some integers n < m. This answer only works if you allow non-sequential numbers. $\endgroup$
    – Lacklub
    Mar 2, 2016 at 14:29
  • $\begingroup$ @Lacklub good point, I've asked for clarification. $\endgroup$ Mar 2, 2016 at 14:48
  • $\begingroup$ "a square that is divided into smaller squares, each containing a number, such that the figures in each vertical, horizontal, and diagonal row add up to the same value." There is no stipulation in any definition that I have seen that the entries must be sequential. $\endgroup$ Mar 2, 2016 at 15:00
  • $\begingroup$ @IanMacDonald Each containing a distinct number, right? Otherwise it's trivial. $\endgroup$ Mar 2, 2016 at 22:54

If you add a number $k$ to every element of a magic square, you get another magic square whose magic constant is the original constant plus $3k$. In particular, if you set $k$ to $-\frac 13$ of the original constant, you get a magic square whose constant is $0$. Thus every magic square induces one whose constant is $0$.

Now, for a square with constant $0$, if we name the following elements, $$\begin{matrix}x & y & \cdot\\\cdot & z & \cdot\\\cdot & \cdot & \cdot\end{matrix}$$ we can start filling in the other elements from the requirement all columns, rows and diagonals sum to $0$.

$$\begin{matrix}x & y & -x-y\\\cdot & z & \cdot\\\color{red}{w} & -y-z & -x-z\end{matrix}$$ With $w$, we run into an impasse: to make the diagonal $0$, we must have $w = x + y - z$, and to make the bottom row $0$, we must have $w = x + y + 2z$. These can both be realized only if $z = 0$. With that change, we can complete the pattern: $$\begin{matrix}x & y & -x-y\\-2x -y & 0 & 2x+y\\x + y & -y & -x\end{matrix}$$ By the definition quoted by Ian MacDonald, any $x$ and $y$ would work. But I've always understood that the numbers in a magic square should also be distinct (Wikipedia agrees). If we add that condition then we have a list of restrictions on the values of $x$ and $y$: $$x \ne 0, y \ne 0, y \ne kx\text{ for }k \in \{1, -\frac 12, -1, -\frac 32, -2, -3\}$$

If we do not consider rotations/reflections as separate squares, then these restrictions allow only one magic square that consists of consecutive integers, the one that frodoskywalker has already produced: $$\begin{matrix}1 & 2 & -3\\-4 & 0 & 4\\3 & -2 & -1\end{matrix}$$

If you do consider rotations and reflections as separate squares, then there are 8 possible squares, depending on which of the 4 corners the $1$ is in, and whether the $2$ is clockwise or counter-clockwise from it.

If you allow non-consecutive integers, or non-integers, then there are infinitely many solutions, even if you don't allow multiples like frodoskywalker mentions. Just choose a non-zero $x$, then choose any value for $y$ other than the 7 that are restricted (and also relatively prime to $x$, if you want to avoid multiples of other squares).

Addendum: why there is only one magic square with consecutive integers.

Any pattern with $x < 0$ is a rotation of $180^\circ$ from one with positive $x$, so we can assume $x > 0$. Further, if $|y| < x$, then $-x < y$, so $x < 2x + y = |-2x - y|$. So by flipping the square if needed, we can also assume $|y| > x$.

Now, the sum of all the numbers in a magic square is the sum of the values of all three rows, and therefore is 3 times the magic number, which in this case is $0$. The only 9 consecutive integers that sum up to $0$ are those between $-4$ and $4$. So all the elements in our square need to be in this range. In particular $|2x + y| \le 4$. Consider the cases:

  • $x = 1$: $y \ne -2, -3$ by the restrictions, and $2x + y < 4$ gives $y \le 2$. So $y = 2$ and $y = -4$ are the only choices, and both give the same square.
  • $x = 2$: $y \ne -1, -2, -3, -4$ by the restrictions, and $2x + y < 4$ gives $y < 0$, so there is no solution.
  • $x = 3$: Since $|y| > x$, the only possibilities are $-4$ and $4$, but $2\cdot 1 + 4 > 4$, so it must be that $y = -4$, which gives the same pattern again.
  • $x = 4$: $|y| > x$ cannot be satisfied.

Which exhausts all possibilities.

  • $\begingroup$ Here is magic square with 25 fields and a sum of 0: -2 11 -6 7 -10 -9 -1 12 -5 3 4 -8 0 8 -4 -3 5 -12 1 9 10 -7 6 -11 2 But I agree with your calculation for 3x3 squares $\endgroup$
    – Etoplay
    Mar 3, 2016 at 13:01

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.