Arxiv:1710.04779V1 [Math.NT]
Total Page:16
File Type:pdf, Size:1020Kb
BERNOULLI NUMBERS AND SUM OF POWERS OF INTEGERS OF HIGHER ORDER ANDREI K. SVININ AND SVETLANA V. SVININA Abstract. We give an expression of polynomials for higher sums of powers of integers via the higher order Bernoulli numbers. 1. Introduction As is known the sum of powers of integers [4], [9] n m Sm(n) := q q X=1 can be computed with the help of some appropriate polynomial Sˆm(n) for any m ≥ 0. Exponential generating function for the sums Sm(n) is given by n e(n+1)t − et S(n,t)= eqt = . (1.1) et − 1 q X=1 Expanding in series (1.1) yields an infinite set of polynomials {Sˆm(n) : m ≥ 0}, that is, tq S(n,t)= Sˆ (n) . q q! q X≥0 It is a classical result that these polynomials can be expressed as [7] m 1 m + 1 Sˆ (n)= (−1)q B nm+1−q, (1.2) m m + 1 q q q X=0 where Bq are the Bernoulli numbers that can be derived from the exponential generating function t tq = B . (1.3) et − 1 q q! q X≥0 arXiv:1710.04779v1 [math.NT] 13 Oct 2017 It follows from (1.3) that the Bernoulli numbers satisfy the recurrence relation m m + 1 B = δ . (1.4) q q 0,m q X=0 This relation is in fact the simplest one of many known recurrence relations involving the Bernoulli numbers (see, for example, [2] and references therein). One can derive, for example, an infinite number of recurrence relations of the form m m + k m + k S(m + k − q, k)B = S(m + k − 1, k − 1), ∀k ≥ 1. (1.5) q q k q X=0 Date: October 16, 2017. 1 2 ANDREIK.SVININANDSVETLANAV.SVININA In this paper we investigate a class of sums that correspond to a k-th power of generating function (1.1) for k ≥ 1. Our main result is a formula for polynomials allowing to calculate these sums. It turns out that these polynomials are expressed via the higher Bernoulli numbers. 2. The power sums of higher order Let us now consider a power of the generating function (1.1): tq (S(n,t))k := S(k)(n) . q q! q X≥0 We have k n kn k (S(n,t))k = eqt = eqt. (2.1) q q=1 n X Xq=k The coefficients k obviously generalizing the binomial coefficients originated from Abra- q n ham De Moivre and Leonard Euler works [10], [5] and extensively studied in the literature due to their applicability. From (2.1), we see that it is a generating function for the sums of the form k(n−1) k S(k)(n) := (k + q)m . (2.2) m q q n X=0 It is natural to call (2.2) the sums of powers of integers of higher order. Expanding k e(n+1)t − et tq = Sˆ(k)(n) , et − 1 q q! ! q X≥0 (k) we get an infinite number of polynomials Sˆm (n). Our goal in the paper is to prove that m 1 m + k Sˆ(k)(n)= (−1)q B(k)S(m + k − q, k)nm+k−q, (2.3) m m+k q q k q X=0 (k) where Bq are the higher order Bernoulli numbers defined as tk tq = B(k) . (2.4) (et − 1)k q q! q X≥0 The Bernoulli numbers of higher order appeared in [11] in connection with a theory of finite differences and then was investigated by many authors from different points of view (see, for example, [3]). These numbers are known to satisfy [11] k − n B(k+1) = B(k) − nB(k) . n k n n−1 (k) The number Bn with fixed n ≥ 0 turns out to be some polynomial in k. These kind of polynomials are known as N¨orlund polynomials. One can find a number of these polyno- mials in [11]. For convenience, we have written out several N¨orlund polynomials in the BERNOULLI NUMBERS AND SUMS OF POWERS OF INTEGERS OF HIGHER ORDER 3 Appendix. The numbers S(n, k) in (2.3) are the Stirling numbers of the second kind that satisfy recurrence relation S(n, k)= S(n − 1, k − 1) + kS(n − 1, k) (2.5) with appropriate boundary conditions [15], [4]. It is easy to prove that the higher order Bernoulli numbers satisfy the recurrence relation m m + k S(m + k − q, k)B(k) = δ . (2.6) q q 0,m q X=0 The most general relation involving (1.4), (1.5) and (2.6) as particular cases is m m + k m+k S(m + k − q, k)B(r) = k S(m + k − r, k − r), ∀k ≥ r. (2.7) q q m+k−r q=0 k−r X As is known S(m + k, k), for any fixed m ≥ 0, is expressed as a polynomial fm(k) of degree 2m, which satisfy the identity fm(k) − fm(k − 1) = kfm−1(k) following from the identity (2.5). Therefore we can replace S(m + k − q, k) by fm−q(k) in (2.3). In the literature the polynomials fm(k) are known as the Stirling polynomials [6], [8]. These are known to be expressed via the N¨orlund polynomials as (see, for example, [1]) m + k (−k) fm(k)= B . m m The following proposition also gives the relationship of the higher Bernoulli numbers with the Stirling numbers. Proposition 2.1. One has m (k) s(q + k, k) Bm = q+k S(m,q). (2.8) q=1 k X In (2.8), s(n, k) stands for the Stirling numbers of the first kind [14]. It is evident that in the case k = 1, (2.8) becomes m q! B = (−1)q S(m,q), m q + 1 q=1 X while in the case k = 2, it takes the following form: m (q + 1)!H B(2) = 2 (−1)q q+1 S(m,q), m (q + 1)(q + 2) q X=1 m where Hm are harmonic number defined by Hm := q=1 1/q. To prove (2.3), we need the following lemma: P 4 ANDREIK.SVININANDSVETLANAV.SVININA Lemma 2.2. By virtue of (2.7) we have m m + k R(k,r)(n) := (−1)q S(m + k − q, k)Sˆ(r)(n) (2.9) m q q q=0 X m 1 m + k = (−1)j S(m + k − r − j, k − r) k m + k − r − j r j X=0 × S(r + j, r)nr+j, ∀m ≥ 0, k ≥ r. (2.10) It should be remarked that in the case k = r, (2.10) becomes (k,k) m m+k Rm (n) = (−1) S(m + k, k)n . (2.11) Proof of lemma 2.2. We can rewrite (2.9) as (k,r) r+q−j Rm (n)= aqbq,jn , j q m 0≤ X≤ ≤ where m+k q aq := r+q S(m + k − q, k) r and r+ q b := (−1)q−j B(r)S(r + q − j, r). q,j j j Let ˜j = q − j and ˜ r + q b = (−1)j B(r) S(r + ˜j, r). q,˜j q − ˜j q−˜j In what follows, for simplicity, let us write ˜j without the tilde. Making use the identity r + q r + q q r+q = = r , q − j r + j j r+j r we get S(r + j, r) m + k q a b = (−1)j S(m + k − q, k)B(r) q q,j r+j q j q−j r and therefore (k,r) r+j Rm (n) = aqbq,jn j q m 0≤ X≤ ≤ S(r + j, r) m + k q = (−1)j nr+j S(m + k − q, k)B(r) . r+j q j q−j j m r j q m 0≤X≤ ≤X≤ In turn, making use the identity m + k q m + k m + k − j = , q j j q − j we get S(r + j, r) m + k R(k,r)(n) = (−1)j nr+j m r+j j j m k 0≤X≤ m +k −j (r) × S(m + k − q, k)B . q − j q−j j q m ≤X≤ BERNOULLI NUMBERS AND SUMS OF POWERS OF INTEGERS OF HIGHER ORDER 5 Finally, by virtue of (2.7), we get m + k − j S(m + k − q, k)B(r) q − j q−j j q m ≤X≤ m + k − j = S(m + k − j − q, k)B(r) q q q m j 0≤ X≤ − m+k−j k = m+k−j−r S(m + k − j − r, k − r) k−r and hence m+k m+k−j (k,r) j j k Rm (n) = (−1) r+j m+k−j−r S(m + k − j − r, k − r) 0≤j≤m r k−r X ×S(r + j, r)nr+j 1 m + k = (−1)j S(m + k − j − r, k − r). k m + k − j − r r j m 0≤X≤ × S(r + j, r)nr+j Therefore the lemma is proved. The recurrence relation, for example (2.11), uniquely determines an infinite set of poly- (k) nomials {Sˆm (n) : m ≥ 0}. We have written out some of them in the Appendix. For ˆ(k) k example, S0 (n)= n . On the other hand k(n−1) n−1 k (k) k q k S0 (n) := = t = n .