<ph f="cmbx">Some binomial series obtained by the WZ-method</ph>

Jesús Guillera

Zaragoza (Spain) E-mail address : jguillera@able.es

1 The WZ-method

We recall [5that a discrete function A ( n , k )   is hypergeometric or closed form (CF) if the quotients A ( n + 1 , k ) A ( n , k ) a n d A ( n , k + 1 ) A ( n , k )   are both rational functions.
And a pair of functions F ( n , k )   , G ( n , k )   is said to be of Wilf and Zeilberger (WZ) if F   and G   are closed forms and besides F ( n + 1 , k ) F ( n , k ) = G ( n , k + 1 ) G ( n , k ) .   In this case H. S. Wilf and D. Zeilberger [4have proved that there exists a rational function C ( n , k )   such that G ( n , k ) = C ( n , k ) F ( n , k ) .   The rational function C ( n , k )   is the so-called certificate of the pair ( F , G )   .
We now define H ( n , k ) = F ( n + 1 , n + k ) + G ( n , n + k ) .   Zeilberger has proved that for every WZ pair F ( n , k )   , G ( n , k )   the following holds:
n = 0 G ( n , 0 ) = n = 0 H ( n , 0 ) .   In next sections we use WZ-pairs to get some Ramanujan's formulae and also some new Ramanujan-like ones.

2 First WZ-pair

We consider the following discrete function:
G ( n , k ) = ( 1 ) n ( 1 ) k 2 10 n 2 2 k ( 20 n + 2 k + 3 ) ( 2 k k ) 2 ( 2 n n ) 2 ( 4 n 2 k 2 n k ) ( 2 n k ) ( n + k n ) .   The package EKHAD [2allows to obtain the companion F ( n , k ) = 64 ( 1 ) n ( 1 ) k 2 10 n 2 2 k n 2 4 n 2 k 1 ( 2 k k ) 2 ( 2 n n ) 2 ( 4 n 2 k 2 n k ) ( 2 n k ) ( n + k n ) .   We derive the result n = 0 ( 1 ) n ( 4 n 2 n ) ( 2 n n ) 2 2 10 n ( 20 n + 3 ) = 1 2 n = 0 ( 2 n n ) 3 2 12 n ( 42 n + 5 ) .   We can extend the pair to have sense for every value of k   , not only integers, in the following way:
F ( n , k ) = 64 π 3 n 2 4 n 2 k 1 ( 1 ) n cos ( π k ) Γ ( 2 n k + 1 / 2 ) Γ ( n + 1 / 2 ) 3 Γ ( k + 1 / 2 ) 2 Γ ( n + k + 1 ) Γ ( 2 n + 1 ) 2 ,
G ( n , k ) = 1 π 3 ( 20 n + 2 k + 3 ) ( 1 ) n cos ( π k ) Γ ( 2 n k + 1 / 2 ) Γ ( n + 1 / 2 ) 3 Γ ( k + 1 / 2 ) 2 Γ ( n + k + 1 ) Γ ( 2 n + 1 ) 2 .
If k   is an integer, it is a routine to prove that n = 0 G ( n , k ) = n = 0 G ( n , k + 1 ) ,   and this implies applying Carlson's theorem [1that for every value of k   , even if k   is not an integer, n = 0 G ( n , k ) = A   , where A   is a constant. To determine the value of the constant, observe that lim t 1 / 2 n = 1 G ( n , t ) = 0 A = lim t 1 / 2 G ( 0 , t ) = 8 π .   And we have that, independently of the value of k   , n = 0 G ( n , k ) = 8 π .   But then we have also the sum of another family of infinite series because obviously we immediately get n = 0 H ( n , k ) = 8 π .   For k = 0   , we get the following results [3:
n = 0 G ( n , 0 ) = n = 0 ( 1 ) n ( 4 n 2 n ) ( 2 n n ) 2 2 10 n ( 20 n + 3 ) = 8 π ,
n = 0 H ( n , 0 ) = 1 2 n = 0 ( 2 n n ) 3 2 12 n ( 42 n + 5 ) = 8 π .
For other values of k   we obtain also interesting results. For example, for k = 1 / 4   we get
2 8 n = 0 ( 1 ) n ( 1 2 ) n ( 1 4 ) 2 n ( n ! ) 2 ( 1 4 ) n 2 4 n 40 n + 7 4 n + 1 = π Γ ( 3 4 ) 2 ,
3 2 8 n = 0 ( 1 2 ) 2 n 2 ( 1 2 ) n ( n ! ) 2 ( 1 4 ) 2 n ( 1 4 ) n 2 8 n 112 n 2 + 88 n + 11 ( 8 n + 1 ) ( 8 n + 5 ) = π Γ ( 3 4 ) 2 .

3 Second WZ-pair

We consider the following discrete function: G ( n , k ) = ( 1 ) k 2 16 n 2 4 k ( 120 n 2 + 84 n k + 34 n + 10 k + 3 ) ( 2 k k ) 3 ( 2 n n ) 4 ( 4 n 2 k 2 n k ) ( 2 n k ) ( n + k n ) 2 ;   the package EKHAD [2allows to get the companion F ( n , k ) = 512 ( 1 ) k 2 16 n 2 4 k n 3 4 n 2 k 1 ( 2 k k ) 3 ( 2 n n ) 4 ( 4 n 2 k 2 n k ) ( 2 n k ) ( n + k n ) 2 .   We have the following result: n = 0 ( 4 n 2 n ) ( 2 n n ) 4 2 16 n ( 120 n 2 + 34 n + 3 ) = 1 4 n = 0 ( 1 ) n ( 2 n n ) 5 2 20 n ( 820 n 2 + 180 n + 13 ) .   We can extend the pair to have sense for every value of k   , not only integers, in the following way:
F ( n , k ) = 512 π 5 n 3 4 n 2 k 1 × cos ( π k ) Γ ( 2 n k + 1 / 2 ) Γ ( n + 1 / 2 ) 6 Γ ( k + 1 / 2 ) 3 Γ ( n + k + 1 ) 2 Γ ( 2 n + 1 ) 3 ,  
G ( n , k ) = 1 π 5 ( 120 n 2 + 84 n k + 34 n + 10 k + 3 ) × cos ( π k ) Γ ( 2 n k + 1 / 2 ) Γ ( n + 1 / 2 ) 6 Γ ( k + 1 / 2 ) 3 Γ ( n + k + 1 ) 2 Γ ( 2 n + 1 ) 3 .  
If k   is an integer it is a routine to prove that n = 0 G ( n , k ) = n = 0 G ( n , k + 1 ) ,   and this implies applying Carlson's theorem [1that for every value of k   , even if k   is not an integer, n = 0 G ( n , k ) = A   , where A   is a constant. To determine the constant value A   observe that lim t 1 / 2 n = 1 G ( n , t ) = 0 A = lim t 1 / 2 G ( 0 , t ) = 32 π 2 .   And we have that, independently of the value of k   , n = 0 G ( n , k ) = 32 π 2 .   But then we have also the sum of another family of infinite series because obviously we immediately get n = 0 H ( n , k ) = 32 π 2 .   For k = 0   we obtain the following results:
n = 0 G ( n , 0 ) = n = 0 ( 4 n 2 n ) ( 2 n n ) 4 2 16 n ( 120 n 2 + 34 n + 3 ) = 32 π 2 ,
n = 0 H ( n , 0 ) = 1 4 n = 0 ( 1 ) n ( 2 n n ) 5 2 20 n ( 820 n 2 + 180 n + 13 ) = 32 π 2 .
For other values of k   we obtain also interesting results. For example, for k = 1 / 4   we get
1 8 n = 0 ( 1 2 ) n 3 ( 1 4 ) 2 n ( n ! ) 3 ( 1 4 ) n 2 2 6 n 240 n 2 + 110 n + 11 ( 4 n + 1 ) 2 = π Γ ( 3 4 ) 4 ,
1 8 n = 0 ( 1 ) n ( 1 2 ) 2 n 3 ( 1 2 ) n 3 ( n ! ) 3 ( 1 4 ) 2 n 2 ( 1 4 ) n 2 2 12 n 26240 n 4 + 41184 n 3 + 21448 n 2 + 4170 n + 279 ( 8 n + 1 ) 2 ( 8 n + 5 ) 2 = π Γ ( 3 4 ) 4 .
References

  1. W.N. Bailey, Generalized Hypergeometric Series, Cambridge Univ. Press, p. 39 (1935).
  2. M. Petkovs̆ek, H. S. Wilf, D. Zeilberger, A=B, Peters, (1996).
  3. S. Ramanujan, Modular equations an approximations to π   , Quart. J. Pure Appl. Math. 45 pp. 350-372 (1914).
  4. H.S. Wilf, D. Zeilberger, Rational functions certify combinatorial identities, J. Amer. Math. Soc. 3, (1990).
  5. D. Zeilberger, Closed-Form (pun intended!), Contemp. Math. 143 (1993).

Zaragoza (Spain) E-mail address : jguillera@able.es