Bulletin of Mathematical Analysis and Applications ISSN: 1821-1291, URL: http://www.bmathaa.org Volume 5 Issue 1(2013), Pages 71-79. FINITE NUMBER SUMS IN HIGHER ORDER POWERS OF HARMONIC NUMBERS (COMMUNICATED BY TOUFIK MANSOUR) ANTHONY SOFO Abstract. We develop a set of identities for finite sums of products of harmonic numbers in higher order and reciprocal binomial coefficients. The results are analogous to some identities of Euler type. 1. Introduction and preliminaries Let, as usual, Hn = γ + ψ (n + 1) = n X 1 r=1 r = Z∞ 1 − tn dt 1−t 0 be the nth harmonic number and γ denotes the Euler–Mascheroni constant. Let, also, R, C and N denote, respectively thesets of real, complex and natural numbers. w A generalized binomial coefficient may be defined by z Γ (w + 1) w := ; w, z ∈ C z Γ (z + 1) Γ (w − z + 1) and in the special case when z = n, n ∈ N ∪ {0} , N := {1, 2, 3...} we have (−1)n (−w)n w (w − 1) ... (w − n + 1) w = . := n n! n! The familiar gamma function R∞ −y w−1 e y dy, ℜ (y) > 0, 0 Γ (w) = Γ(w+n) , w ∈ C\ {0, −1, −2, −3, ...} , n ∈ N n−1 Q (w+j) j=0 2000 Mathematics Subject Classification. 05A10, 05A19. Key words and phrases. Harmonic numbers, Binomial coefficients and Gamma function, PolyGamma function, Combinatorial series identities and summation formulas, Partial fraction approach. c 2013 Universiteti i Prishtinës, Prishtinë, Kosovë. Submitted August 20, 2012. Published February 28, 2013. 71 72 ANTHONY SOFO and (w)λ := 1, λ = 0; w ∈ C\ {0} Γ (w + λ) = Γ (w) w (w + 1) ... (w + n − 1) , w ∈ C, n ∈ N , with (0)0 := 1 is known as the Pochhammer symbol. The generalized nth harmonic number in (r) power r, Hn , is defined for positive integers n and r as n X 1 (−1)r−1 (r−1) (r−1) Hn(r) := ψ (n + 1) − ψ (1) , = mr (r − 1)! m =1 where m+1 ψ (m) (z) = (−1) m! ∞ X 1 r =0 (z + r) m+1 m+1 = (−1) m! ζ (m + 1, z) are the polygamma functions of order m which are defined by ψ (0) (z) ≡ ψ(z) and ψ (m) (z) := dm ψ(z)/dz m , m ∈ N and z 6= 0, −1, −2, . . .. Here ψ(z) is the psi, or digamma, function, given as the logarithmic derivative of the well-known gamma function Γ(z), i.e. ψ(z) := d log Γ(z)/dz. ζ (α, z) denotes the Hurwitz zeta function. There also exists the very useful recurrence relation ψ (m) (z + 1) = (−1)m m! (m) (z). We shall provide, in this paper, identities for the general finite z m+1 + ψ sums m m m m (p) (p) (p) (p) X X X X Hn Hn H Hn and n . , , n (n + a) (n + 1) (n + a) n+k n+k n =1 n =1 n =1 n n =1 k k Analogous results of Euler type for infinite series have been developed by many authors, see for example [3], [4], [17] and references therein. Many finite versions of higher order harmonic number sum identities also exist in the literature, for example [15] h p i X 6 p n+1 (−1) (Hn )3 + 3Hn Hn(2) + 2Hn(3) = 3 n p n=0 and by the inversion formula p n+1 X 6 (−1) p 3 = (Hp ) + 3Hp Hp(2) + 2Hp(3) . n n3 n=1 Also, after a minor modification, [11] 3 k 2 X k (1) (1) (2) (2) r+1 2 = 2. 1 + 3r Hk−r + Hr−1 + 3r Hk−r − Hr−1 − 1 (−1) r r=1 Other finite sum identities can be seen in [1], [18] or [19] where they obtain results like 3 2m X (−1)m (3m)! n (−1)k Hk = (Hm + 2H2m − H3m ) . 3 k 2 (m!) k=0 Further work in the summation of harmonic numbers and binomial coefficients has also been done by Sofo [14]. The works of, [2], [3], [4], [5], [6], [7], [8], [9], [10], [11], [12], [13],[16], [20], and references therein, also investigate various representations FINITE NUMBER SUMS 73 of binomial sums and zeta functions in simpler form by the use of the Beta function and by means of certain summation theorems for hypergeometric series, [17]. The following results will be useful for later analysis. Lemma 1.1. Let m and p be a positive integers and a > 0 then m (p) X a Hn n (n + a) n =1 p+1 (p+1) (p) = Hm + Hm (Ha−1 + Hm − Hm+a ) + (−1) + a−1 X p+1 (−1) jp j =1 (Hj − Hm+j ) + p−1 X p−s (−1) (p) Hm Ha−1 (p−s+1) (s) Hm Ha−1 (1.1) . s=2 Proof. Consider m X n (p) X Hn 1 = p n (n + a) n =1 k n (n + a) n =1 m X k =1 since these sums are absolutely convergent we can write n m X X n =1 k =1 = m X k =1 Now 1 p k n (n + a) = m m X X k =1 n =k kp 1 n (n + a) 1 [ψ(a + k) − ψ(k) − (ψ(m + a + 1) − ψ(m + 1))] . a kp m X 1 [ψ(a + k) − ψ(k) − (ψ(m + a + 1) − ψ(m + 1))] a kp k =1 m a−1 a−1 X X 1 1 X 1 = − ak p k + j m + 1 + j j =0 j =0 k =1 m a−1 X X 1 1 1 1 1 − + − ak p k m + 1 j =1 k + j m+1+j = k =1 (p+1) Hm a = − k =1 (p+1) = (p+1) Hm = a Hm a a−1 m (p) Hm 1XX 1 1 + − a (m + 1) a j =1 k p (k + j) k p (m + 1 + j) − a−1 a−1 m (p) (p) Hm Hm 1X 1XX 1 − + a (m + 1) a j =1 (m + 1 + j) a j =1 k p (k + j) k =1 a−1 m (p) (p) Hm Hm 1XX − − (Hm+a − Hm+1 )+ a (m + 1) a a j =1 k =1 p+1 p p−s X (−1) (−1) + p−1 j k (k + j) s=2 k s j p−s+1 ! 74 ANTHONY SOFO by partial fraction decomposition, hence (p) (p+1) Hm a = − a−1 1X + a j =1 (p) Hm Hm − (Hm+a − Hm+1 ) a (m + 1) a p+1 (−1) jp (Hj + Hm − Hm+j ) + p (s) p−s X (−1) Hm j p−s+1 s=2 ! . Therefore m (p) X a Hn n (n + a) n =1 = (p+1) Hm − (p) Hm 1 + Hm+a − Hm+1 − Ha−1 m+1 p−1 a−1 p+1 X X (−1) p−s (s) (p−s+1) (−1) Hm Ha−1 (H − H ) + + j m+j p j s=2 j =1 p+1 (p) Hm Ha−1 = (p+1) (p) Hm + Hm (Ha−1 + Hm − Hm+a ) + (−1) + (−1) p+1 + a−1 X p+1 (−1) jp j =1 (Hj − Hm+j ) + p−1 X p−s (−1) (p) Hm Ha−1 (p−s+1) (s) Hm Ha−1 s=2 and the result (1.1) follows. (α) Remark. When a = 1, m and p are positive integers, and we define H0 then m (p) (p) X Hn Hm (p+1) = Hm − n (n + 1) m+1 n =1 = 0, (1.2) and for a = 2 (p) p−1 m (p) p+1 X X m 2 + m − 1 Hm 2 Hn m (−1) p−s (s) (p+1) (−1) Hm . = Hm + + + n (n + 2) (m + 1) (m + 2) m + 1 n =1 s=2 A related Lemma which will be useful later is the following. Lemma 1.2. Let m and p be a positive integers and a > 0 then m (p) X (a − 1) Hn (n + 1) (n + a) n =1 p+1 (p) = Hm (Ha−1 + Hm+1 − Hm+a ) + (−1) a−1 X p+1 (−1) + jp j =1 (Hj − Hm+j ) + p−1 X s=2 (p) Hm Ha−1 (p−s+1) (s) (−1)p−s Hm Ha−1(1.3). FINITE NUMBER SUMS 75 Proof. Consider m X n (p) X 1 Hn = p (n + 1) (n + a) n =1 k (n + 1) (n + a) n =1 m X k =1 since these sums are absolutely convergent we can write m X m m X n X X 1 1 = p (n + 1) (n + a) p (n + 1) (n + a) k k n =1 k =1 n =k k =1 = m X 1 [ψ(a + k) − ψ(k + 1) − (ψ(m + a + 1) − ψ(m + 2))] . (a − 1) k p k =1 Now m X = 1 [ψ(a + k) − ψ(k + 1) − (ψ(m + a + 1) − ψ(m + 2))] (a − 1) k p k =1 m a−1 a−1 X X X 1 1 1 − (a − 1) k p k + j m + 1 + j j =1 j =1 k =1 = m X k =1 = a−1 X 1 1 1 − (a − 1) k p k+j m+1+j j =1 a−1 m XX 1 1 1 − (a − 1) j =1 k p (k + j) k p (m + 1 + j) k =1 = − a−1 a−1 m (p) X XX 1 Hm 1 1 + (a − 1) j =1 (m + 1 + j) (a − 1) j =1 k p (k + j) k =1 a−1 X m (p) X Hm 1 =− (Hm+a − Hm+1 ) + (a − 1) (a − 1) j =1 k =1 by partial fraction decomposition, hence p+1 p p−s X (−1) (−1) + j p−1 k (k + j) s=2 k s j p−s+1 ! (p) = − Hm (Hm+a − Hm+1 ) (a − 1) a−1 X 1 + (a − 1) j =1 p+1 (−1) jp (Hj + Hm − Hm+j ) + p (s) p−s X (−1) Hm s=2 j p−s+1 ! . Therefore m (p) X (a − 1) Hn (n + 1) (n + a) n =1 p+1 + (−1) (p) Hm Ha−1 + (p) = −Hm (Hm+a − Hm+1 − Ha−1 ) p−1 a−1 p+1 X X (−1) p−s (s) (p−s+1) (−1) Hm Ha−1 (H − H ) + j m+j p j s=2 j =1 76 ANTHONY SOFO (p) p+1 (p) Hm (Ha−1 + Hm+1 − Hm+a ) + (−1) = + a−1 X p+1 (−1) jp j =1 (Hj − Hm+j ) + p−1 X Hm Ha−1 p−s (−1) (p−s+1) (s) Hm Ha−1 s=2 and the result (1.3) follows. Some special cases are noted in the remark. Remark. When a = 2 and p and m positive integers, then m X (p) (p) p+1 (m + 1) Hm m (−1) Hn = + (n + 1) (n + 2) (m + 2) m+1 n =1 + p−1 X p−s (−1) (s) Hm . (1.4) s=2 2. Two theorems We now prove the following two theorems. Theorem 2.1. Let p be a positive integer and k be a positive integer greater than 1, then m X n =1 + k X r (−1) r r=2 k r (p) Hn (p) = kHm Hm+1 n+k k (p) p+1 −Hm Hm+r + (−1) + (−1)p+1 j =1 jp Pr−1 (Hj − Hm+j ) + Proof. Consider the following expansion: m X n =1 (2.1) (p) Hm Hr−1 + Pp p−s s=2 (−1) (s) (p−s+1) Hm Hr−1 . m m (p) (p) (p) X X Hn k! Hn k! Hn = = k Q (n + 1) (n + 2)k+1 n+k n =1 (n + 1) (n + r) n =1 k r=2 where (α)r is Pochhammer’s symbol given by (α)r = α (α + 1) (α + 2) ... (α + r − 1) , r > 0, (α)0 = 1. Now m X n =1 k m (p) (p) X Hn k! Hn X Ar = (n + 1) r =2 (n + r) n+k n =1 k where r Ar = lim n→−r 2 (−1) n+r = k Q k! (n + r) r =2 k r r 2 (2.2) . FINITE NUMBER SUMS 77 Now from (2.2) and using Lemma 1.2 X k k m m (p) (p) X X Hn k! Hn X Ar k r = (−1) r (r − 1) r (n + 1) r =2 (n + r) r =2 (n + 1) (n + r) n =1 n =1 k X = r (−1) r r =2 k r (p) = kHm Hm+1 + k X (p) + Pr−1 r + r (−1) r r =2 m X = n =1 (p) Hn n+k k j =1 (−1) r r =2 k X p+1 Hm (Hm+1 − Hm+r ) + (−1) (−1)p+1 jp k r (Hj − Hm+j ) + (p) Pp (s) s=2 p+1 −Hm Hm+r + (−1) (p) Hm Hr−1 (p−s+1) (−1)p−s Hm Hr−1 (p) Hm Hr−1 X p r−1 p+1 X (−1) k p−s (s) (p−s+1) (−1) Hm Hr−1 (Hj − Hm+j ) + p r j s=2 j =1 which is the result (2.1). Now we consider the following extension of Theorem 2.1. Theorem 2.2. Let k, m and p be positive integers, then m X n =1 n h k (p) i X Hn k (p) r p (p+1) (p) (p) = Hm + Hm Hm + (−1) Hm Hm+r + (−1) Hm Hr−1 r n+k r=1 k (2.3) − k X r=1 r (−1) X p r−1 p+1 X (−1) k p−s (s) (p−s+1) (−1) Hm Hr−1 . (Hj − Hm+j ) + p r j s=2 j =1 Proof. Consider the following expansion: m X n =1 n m m (p) (p) (p) X X k! Hn k! Hn Hn = = k Q n (n + 1)k+1 n+k n =1 n (n + r) n =1 k r=1 78 ANTHONY SOFO Now m X n =1 where m k (p) (p) X k! Hn X Br Hn = n (n + r) n+k n =1 r =1 k Br = lim n+r n→−r k Q (n + r) r+1 (−1) = k! r k r . r =1 Now from (2.3) and using Lemma 1.1 X k k m m (p) (p) X X Hn k! Hn X Br k r+1 = (−1) r r n (n + r) r =1 n (n + r) r =1 n =1 n =1 = k X r+1 (−1) r =1 k r (p+1) Hm + Pr−1 j =1 (p) p+1 + Hm (Hm − Hm+r ) + (−1) p+1 (−1) jp (Hj − Hm+j ) + Pp s=2 (p) Hm Hr−1 p−s (−1) (s) (p−s+1) Hm Hr−1 (p+1) (p) = Hm + Hm Hm + k X r (−1) r =1 = m X n =1 (p) n k r Hn n+k k (p) (p) p Hm Hm+r + (−1) Hm Hr−1 − (−1)p+1 j =1 jp Pr−1 (Hj − Hm+j ) − which is the result (2.3). Pp s=2 p−s (−1) (s) (p−s+1) Hm Hr−1 References [1] X. Chen and W. Chu. Dixon’s 3 F2 (1) series and identities involving harmonic numbers and the Riemann zeta function. Discrete Math. 310 (2010), 83-91. [2] G. S. Cheon and M. E. A. El-Mikkawy. Generalized harmonic numbers with Riordan arrays. J. Number Theory 128 (2008), 413-425. [3] J. Choi. Finite summation formulas involving binomial coefficients, harmonic numbers and generalized harmonic numbers. J. Inequalities Appl. 49 (2013), 11pp. [4] M. W. Coffey and N. Lubbers. On generalized harmonic number sums. Appl. Math. Comput. 217 (2010), 689-698. [5] T. X. He, L. C. Hu and D. Yin. A pair of summation formulas and their applications. Comp. Math. Appl. 58 (2009), 1340-1348. [6] M. J. Kronenburg. Some combinatorial identities some of which involving harmonic numbers. http://arXiv.1103.1268V1, March (2011). [7] H. Liu and W. Wang. Harmonic number identities via hypergeomeric series Bell polynomials. Integral Transforms Spec. Funct. 23 (2012), 49-68. [8] T. Mansour. Identities on harmonic and q-harmonic number sums. Afr. Mat. 23 (2012), 135-143. FINITE NUMBER SUMS 79 [9] T. Mansour, M. Shattuck and C. Song. q-Analogs of identities involving harmonic numbers and binomial coefficients. Applications and Applied Mathematics: An International Journal 7:1 (2012), 22-36. [10] T. Mansour. Identities for sums q-analogue of polylogarithm functions. Letters in Mathematical Physics 87:1-2 (2009), 1-18. [11] A. Sofo and H. M. Srivastava. Identities for the harmonic numbers and binomial coefficients. Ramanujan J. 25 (2011), 93-113. [12] A. Sofo. Computational Techniques for the Summation of Series, Kluwer Academic/Plenum Publishers, New York, 2003. [13] A. Sofo. Sums of derivatives of binomial coefficients. Advances Appl. Math. 42 (2009), 123134. [14] A. Sofo. Summation formula involving harmonic numbers. Analysis Mathematica 37 (2011), 51-64. [15] A. Sofo. Integral forms of sums associated with harmonic numbers. Appl. Math. Comput. 207 (2009), 365-372. [16] A. Sofo. New classes of harmonic number identities. J. Integer Seq. 15 (2012), Article 12.7.4. [17] R. Srivastava, Some families of combinatorial and other series identities and their applications. Appl.Math. Comput. 218 (2011), 1077-1083. [18] W. Wang. Riordan arrays and harmonic number identities. Comp. Math. Appl. 60 (2010), 1494-1509. [19] X. Wang and M. Li. Dixon’s formula and identities involving harmonic numbers. J. Integer Seq. 14 (2011), Article 11.1.3. [20] D. Y. Zheng. Further summation formulae related to generalized harmonic numbers. J. Math. Anal. Appl. 335 (2007), 692-706. Victoria University PO Box 14428, Melbourne City VIC 8001, Australia E-mail address: anthony.sofo@vu.edu.au