Applications of Waring’s formula to some identities of Chebyshev polynomials Jiang Zeng1,2 and Jin Zhou2 1 Institut Girard Desargues, Universit´e Claude Bernard (Lyon I) 5 69622 Villeurbanne Cedex, France 0 [email protected] 0 2 and n 2 Center for Combinatorics, LPMC, Nankai University a Tianjin 300071, People’s Republic of China J [email protected] 4 1 AMS Math Subject Classification Numbers: 11B39, 33C05, 05E05 ] O Abstract. Someidentities ofChebyshev polynomialsarededuced fromWar- C ing’sformulaonsymmetricfunctions. Inparticular,theseformulaegeneralize . h some recent results of Grabner and Prodinger. t a m [ 1 Introduction 1 v Given a set of variables X = {x ,x ,...}, the kth (k ≥ 0) elementary sym- 6 1 2 1 metric polynomial ek(X) is defined by e0(X) = 1, 2 1 e (X) = x ...x , for k > 1, 0 k i1 ik 5 i1<X...<ik 0 h/ and the kth (k ≥ 0)power sum symmetric polynomial pk(X) is defined by t p (X) = 1, a 0 m p (X) = xk, for k > 1. k i v: Xi Xi Let λ = 1m12m2 ...be a partition of n, i.e., m11+m22+...+mnn = n, where r mi ≥ 0 for i = 1,2,...n. Set l(λ) = m1 +m2 +...+mn. According to the a fundamental theorem of symmetric polynomials, any symmetric polynomial can be written uniquely as a polynomial of elementary symmetric polynomi- als e (X) (i ≥ 0). In particular, for the power sum p (x), the corresponding i k formula is usually attributed to Waring [1, 4] and reads as follows: k(l(λ)−1)! p (X) = (−1)k−l(λ) e (X)m1e (X)m2 ..., (1) k 1 2 m ! Xλ i i Q where the sum is over all the partitions λ = 1m12m2 ... of k. In a recent paper [3] Grabner and Prodinger proved some identities about Chebyshev polynomials using generating functions, the aim of this paper is 1 to show that Waring’s formula provides a natural generalization of such kind of identities. Let U and V be two sequences defined by the following recurrence rela- n n tions: U = pU −U , U = 0,U = 1, (2) n n−1 n−2 0 1 V = pV −V , V = 2,V = p. (3) n n−1 n−2 0 1 Hence U and V are rescaled versions of the first and second kind of Cheby- n n shev polynomials U (x) and T (x), respectively: n n 1 U (x) = U (2x), T (x) = T (x). n n+1 n n 2 Theorem 1 For integers m,n ≥ 0, let W = aU +bV and Ω = a2+4b2− n n n b2p2. Then the following identity holds k W2k +W2k = θ (m)Ωk−rWrWr , (4) n n+m k,r n n+m Xr=0 where k(k −j −1)! θ (m) = (−1)j Vr−2jU2k−2r. k,r j!(k −r)!(r−2j)! m m 06X2j6k NotethattheidentitiesofGrabnerandProdinger[3]correspondtothem = 1 and implicitly m = 2 cases of Theorem 1 (cf. Section 3). 2 Proof of Theorem 1 We first check the k = 1 case of (4): W2 +W2 = V W W +U2Ω. (5) n n+m m n n+m m Set α = (p+ p2 −4)/2 and β = (p− p2 −4)/2 then it is easy to see that p p αn −βn U = , V = αn +βn, n n α−β it follows that W = aU +bV = Aαn +Bβn, n n n where A = b+a/(α−β) and B = b−a/(α−β). Therefore V W W +U2Ω = (αm +βm)(Aαn +Bβn)(Aαn+m +Bβn+m) m n n+m m αm −βm 2 + (a2 +4b2 −b2p2), (cid:18) α−β (cid:19) 2 which is readily seen to be equal to W2 +W2 . n n+m Next we take the alphabet X = {W2,W2 }, then the left-hand side of n n+m (4) is the power sum p (X). On the other hand, since k e (X) = W2 +W2 , e (X) = W2W2 , e (X) = 0 if i > 3, 1 n n+m 2 n n+m i the summation at the right-hand side of (1) reduces to the partitions λ = (1k−2j2j), with j ≥ 0. Now, using (5) Waring’s formula (1) infers that W2k +W2k n n+m k(k −j −1)! = (−1)j (V W W +U2Ω)k−2j(W2W2 )j j!(k −2j)! m n n+m m n n+m 06X2j6k k−2j k(k −j −1)! = (−1)j Vk−2j−iU2iΩi(W W )k−i j!i!(k −2j −i)! m m n n+m 06X2j6k Xi=0 Setting k −i = r and exchanging the order of summations yields (4). 3 Some special cases When m = 1 or 2, as U = 1, V = p and U = p, V = p2 −2 the coefficient 1 1 2 2 θ (r) of Theorem 1 is much simpler. k,r Corollary 1 We have k(k −1−j)! θ (1) = (−1)j pr−2j, (6) k,r (k −r)!j!(r−2j)! 06X2j6r k(k −j −1)! θ (2) = (−1)j (p2 −2)r−2jp2k−2r. (7) k,r j!(k −r)!(r−2j)! 06X2j6k Wenoticethat(6)isexactlytheformulagivenbyGrabnerandProdinger[3] for θ (1), while for θ (2) they give a more involved formula than (7) as k,r k,r follows: Corollary 2 (Grabner and Prodinger [3]) There holds λ k(k −⌊λ⌋−1)!2⌈λ2⌉ ⌊2⌋−1 λ θ (2) = (−1)λp2k−2λ 2 (2k−2⌈ ⌉−1−2i). (8) k,r (k −r)!λ!(r−λ)! 2 06Xλ6k Yi=0 In order to identify (7) and (8), we need the following identity. 3 Lemma 2 We have j/2 (k −i−1)!2j−2i (−1)i (j −2i)!i! Xi=0 ⌊j/2⌋−1 (k −⌊j/2⌋−1)! = 2⌈j/2⌉ (2k −2⌈j/2⌉−1−2i). (9) j! Yi=0 Proof: For n ≥ 0 let (a) = a(a + 1)...(a + n − 1), then the Chu- n Vandermonde formula [2, p.212] reads: (−n) (a) (c−a) k k n F (−n,a;c;1) := = . (10) 2 1 (c) k! (c) Xk>0 k n Note that n! = (1) , so using the simple transformation formulae: n a a+1 a a+1 (a) = 22n, (a) = 22n+1, 2n 2n+1 2 (cid:18) 2 (cid:19) 2 (cid:18) 2 (cid:19) (cid:16) (cid:17)n n (cid:16) (cid:17)n+1 n and (a) (a) (a) = N = (−1)n N , N−n (a+N −n) (−a−N +1) n n we can rewrite the left-hand side of identity (9) as follows: (k−1)! F (−m,−m+ 1;−k +1;1) if j = 2m, (12)m(1)m 2 1 2 (k−1)! F (−m,−m− 1;−k +1;1) if j = 2m+1, (12)m+1(1)m 2 1 2 which is clearly equal to the right-hand side of (9) in view of (10). Now, expanding the right-hand side of (7) by binomial formula yields r−2j k(k −j −1)! r −2j (−1)j p2i(−2)r−2j−ip2k−2r. j!(k −r)!(r −2j)! (cid:18) i (cid:19) 06X2j6k Xi=0 Writing λ = r −i, so λ ≤ r ≤ k, and exchanging the order of summations, the above quantity becomes k (k −j −1)!2λ−2j (−1)λp2k−2λ (−1)j , (k −r)!(r −λ)! (λ−2j)!j! 06Xλ6k 06Xj6k/2 which yields (8) by applying Lemma 2. 4 References [1] William Y. C. Chen, Ko-Wei Lih, and Yeong-Nan Yeh: Cyclic Tableaux and Symmetric Functions,Studies inApplied Math., 94(1995),327-339. [2] Ronald L. Graham, Donald E. Knuth and Oren Patashnik: Concrete Mathematics, Addion-Wesley Pubilshing Co. 1989. [3] Peter J. Grabner and Helmut Prodinger: Some identities for Chebyshev polynomials, Portugalia Mathematicae 59 (2002), 311-314. [4] P. A. MacMahon: Combinatory analysis, Chelsea Publishing Co. New York, 1960. 5