on the number of magic squares matthias beck moshe cohen
play

On the Number of Magic Squares Matthias Beck Moshe Cohen Jessica - PDF document

On the Number of Magic Squares Matthias Beck Moshe Cohen Jessica Cuomo Paul Gribelyuk SUNY Binghamton www.binghamton.edu/matthias Semimagic square: square matrix whose entries are nonnegative integers and whose row and column sums


  1. On the Number of “Magic Squares” Matthias Beck Moshe Cohen Jessica Cuomo Paul Gribelyuk SUNY Binghamton www.binghamton.edu/matthias

  2. Semi–magic square: square matrix whose entries are nonnegative integers and whose row and column sums are equal Magic square: semi–magic square whose main diagonals add up to the row–column sum Symmetric magic square: magic square which is symmetric Pandiagonal magic square: semi–magic square whose pandiagonals add up to the row–column sum. H n ( t ) := # semi–magic squares M n ( t ) := # magic squares S n ( t ) := # symmetric magic squares P n ( t ) := # pandiagonal magic squares for n × n squares with row–column–(pan)diagonal sum t 2

  3. Example: H 2 ( t ) = t + 1 � 1 if t is even, M 2 ( t ) = S 2 ( t ) = P 2 ( t ) = 0 if t is odd. Theorem (Macmahon, 1915) 8 t 4 + 3 4 t 3 + 15 8 t 2 + 9 H 3 ( t ) = 1 4 t + 1 � 2 9 t 2 + 2 3 t + 1 if 3 | t M 3 ( t ) = 0 otherwise Theorem (Stein–Stein, 1970) H n ( t ) is a polynomial in t of degree ( n − 1) 2 . Theorem (Ehrhart, Stanley, 1973) H n ( − n − t ) = ( − 1) deg( H n ) H n ( t ) H n ( − 1) = H n ( − 2) = · · · = H n ( − n + 1) = 0 3

  4. Conditions for magic 3 × 3 squares: x 1 ≥ 0 x 2 ≥ 0 . . . x 9 ≥ 0 x 1 + x 2 + x 3 = t x 4 + x 5 + x 6 = t x 7 + x 8 + x 9 = t x 1 + x 4 + x 7 = t x 2 + x 5 + x 8 = t x 3 + x 6 + x 9 = t x 1 + x 5 + x 9 = t x 3 + x 5 + x 7 = t This system describes a polytope in R 9 . We are interested in counting integer points in this polytope. Geometrically, the “magic variable” t is a dilation parameter. 4

  5. Dilate the d –dimensional rational polytope P by a positive integer t : t P := { tx : x ∈ P} and count the number of integer points (“lattice points”) in t P : � t P ∩ Z d � L P ( t ) := # Theorem (Ehrhart, 1960’s) L P ( t ) is a quasi- polynomial in t whose degree is the dimen- sion of P . The period of this quasipoly- nomial divides any common multiple of the denominators of the vertices of P . 5

  6. A quasipolynomial is an expression c d ( t ) t d + · · · + c 1 ( t ) t + c 0 ( t ) , where c 0 , . . . , c d are periodic functions in t Example:  0 for t = 1 , 4 , 7 , 10 , . . .  0 for t = 2 , 5 , 8 , 11 , . . . M 3 ( t ) = 9 t 2 + 2 2 3 t + 1 for t = 3 , 6 , 9 , 12 , . . .  Theorem M n ( t ) , S n ( t ) , P n ( t ) are quasipoly- nomials in t with degrees deg( M n ) = n 2 − 2 n − 1 2 n 2 − 1 deg( S n ) = 1 2 n − 2 deg( P n ) = n 2 − 3 n + 2 6

  7. Idea of proof: The conditions of the 3 × 3 magic square yield the linear system   x 1     1 0 0 1 0 0 1 0 0 t x 2   0 1 0 0 1 0 0 1 0 t       x 3       0 0 1 0 0 1 0 0 1 t       x 4       1 1 1 0 0 0 0 0 0 t       = x 5       0 0 0 1 1 1 0 0 0 t       x 6       0 0 0 0 0 0 1 1 1 t       x 7       1 0 0 0 1 0 0 0 1 t       x 8   0 0 1 0 1 0 1 0 0 t x 9 7

  8. Denote by M ⋆ n ( t ) , S ⋆ n ( t ) , P ⋆ n ( t ) the counting functions for magic squares, symmetric, and pandiagonal magic squares, respectively, as before, but now with the restriction that the entries are positive integers. For a d –dimensional rational polytope P , define � t P int ∩ Z d � L ⋆ P ( t ) := # Theorem (Ehrhart–Macdonald reciprocity law, 1971) If P is a d –dimensional rational poly- tope homeomorphic to a d –manifold then L P ( − t ) = ( − 1) d L ⋆ P ( t ) . ⇓ 8

  9. Proposition M n ( − t ) = ( − 1) n 2 − 2 n − 1 M ⋆ n ( t ) 1 2 n 2 − 1 2 n − 2 S ⋆ S n ( − t ) = ( − 1) n ( t ) P n ( − t ) = ( − 1) n 2 − 3 n +2 P ⋆ n ( t ) M ⋆ n ( t ) = M n ( t − n ) ⇓ etc . M ⋆ n (1) = · · · = M ⋆ n ( n − 1) = 0 Theorem M n ( − t ) = ( − 1) n 2 − 2 n − 1 M n ( t − n ) 1 2 n 2 − 1 2 n − 2 S n ( t − n ) S n ( − t ) = ( − 1) P n ( − t ) = ( − 1) n 2 − 3 n +2 P n ( t − n ) M n ( − 1) = · · · = M n ( − n + 1) = 0 S n ( − 1) = · · · = S n ( − n + 1) = 0 P n ( − 1) = · · · = P n ( − n + 1) = 0 9

  10. Find vertices by setting some x k = 0 .   x 1     1 0 0 1 0 0 1 0 0 1 x 2     0 1 0 0 1 0 0 1 0 1 x 3             0 0 1 0 0 1 0 0 1 1 x 4             0 0 0 1 1 1 0 0 0 = 1 x 5             0 0 0 0 0 0 1 1 1 1 x 6             1 0 0 0 1 0 0 0 1 1 x 7         0 0 1 0 1 0 1 0 0 1 x 8   x 9 Vertices: � � � � 3 , 0 , 1 2 3 , 0 , 1 3 , 2 3 , 1 3 , 2 1 3 , 2 3 , 0 , 0 , 1 3 , 2 3 , 2 3 , 0 , 1 3 , 0 , , 3 � � � � 0 , 2 3 , 1 3 , 2 3 , 1 3 , 0 , 1 3 , 0 , 2 1 3 , 0 , 2 3 , 2 3 , 1 3 , 0 , 0 , 2 3 , 1 , 3 3 10

  11. To interpolate a quasipolynomial of degree d and period p , we need to compute p ( d +1) values. Example: M 3 ( t ) has degree 2 and period 3 M 3 (1) = 0 M 3 (2) = 0 M 3 (3) = 5 M 3 (4) = 0 M 3 (5) = 0 M 3 (6) = 13 M 3 (7) = 0 M 3 (8) = 0 M 3 (9) = 25  0 for t = 1 , 4 , 7 , 10 , . . .  0 for t = 2 , 5 , 8 , 11 , . . . = ⇒ M 3 ( t ) = 9 t 2 + 2 2 3 t + 1 for t = 3 , 6 , 9 , 12 , . . .  11

  12. � 2 9 t 2 + 2 3 t + 1 if 3 | t M 3 ( t ) = 0 otherwise � 2 3 t + 1 if 3 | t S 3 ( t ) = 0 otherwise 2 t 2 + 2 P 3 ( t ) = 1 3 + 1  480 t 7 + 7 240 t 6 + 89 480 t 5 + 11 1 16 t 4   30 t 3 + 38 15 t 2 + 71 + 49  30 t + 1 if t is even   M 4 ( t ) = 480 t 7 + 7 240 t 6 + 89 480 t 5 + 11 1 16 t 4   480 t 3 + 593 240 t 2 + 1051  + 779 480 t − 3 16 if t is odd   128 t 4 + 5 16 t 3 + t 2 + 3 5  2 t + 1 if t ≡ 0 (4)    128 t 4 + 5 16 t 3 + t 2 + 3 5 2 t + 7 S 4 ( t ) = 8 if t ≡ 2 (4)   0 if t is odd  1440 t 6 + 7 120 t 5 + 23 72 t 4 + t 3 7   180 t 2 + 31 + 341 P 4 ( t ) = 15 t + 1 if t is even 0 if t is odd  12

  13. What about the vertices/periods? Conjecture For n ≥ 4 , M n , S n , and P n are not polynomials. Application/next step: integer solutions to transportation problem/polytope 13

  14. H d n ( t ) := # semi–magic hypercubes of dimension d , size n , and axial sums t H d n ( t ) is a quasipolynomial in t . H 2 = ( n − 1) 2 � � Stein–Stein: deg n � � H d = ( n − 1) d Theorem deg n M d n ( t ) := # magic hypercubes of dimension d , size n , and axial/diagonal sums t M d n ( t ) is a quasipolynomial in t . Earlier Theorem: = n 2 − 2 n − 1 = ( n − 1) 2 − 2 M 2 � � deg n � � = ( n − 1) d − 2 d − 1 M d Conjecture deg n Conjecture For n, d ≥ 3 , H d n and M d n are not polynomials. 14

Recommend


More recommend