Arxiv:2008.00284V2 [Math.NT] 9 May 2021

arXiv:2008.00284v2 [math.NT] 9 May 2021 The Introduction 1 in ftehroi ubr,nml h eeaie harmo generalized the namely numbers, harmonic the of tions hyperha generalized as the defined generalization, unified a in rested ra se o xml [ example for (see, subjects areas other with numbers a harmonic numbers relating harmonic alizations, the for forms closed Therefore and representations sciences. computer theory, combinatori number as analysis, such cal applications and mathematics numbers in harmonic places The zero. be to understood conventionally with § † ‡ * eetKargın Levent [email protected] [email protected] [email protected] [email protected] mn ayohrgnrlztoso h amncnmes w numbers, harmonic the of generalizations other many Among n H us congruences. sums, n Stirling polynomials, poly-Bernoulli numbers, harmonic eosiette o h ye-usadsvrlcongruenc several and hyper-sums relationship the for This identities merous numbers. Bernoulli and harmonic between th from motivated polynomials, poly-Bernoulli the and bers hhroi ubri endby defined is number harmonic th k ( eeaie amncnmesvia numbers harmonic Generalized p, Keywords: 2010: MSC hyperhar generalized the between relationship a present We eateto ahmtc,AdnzUiest,Atla T Antalya, University, Akdeniz Mathematics, of Department 0) 1 = oyBrolipolynomials poly-Bernoulli /k p amncnmes yehroi ubr,generalized numbers, hyperharmonic numbers, Harmonic 17,1B8 17,11A07 11B73, 11B68, 11B75, se[ (see emtCenkci Mehmet , * 4 14 , 6 ) hs ubr xedtofmu generaliza- famous two extend numbers These ]). , H 10 n ( p,r , 14 ) = , Abstract 21 k X =1 n , 1 22 H H † ya Dil Ayhan , k n ( 32 p,r + 1 = , − 33 1) , , 35 1 2 + ]). ‡ n M and 1 3 + mes hyper- umbers, · · · r cieresearch active are connections e nrdcn new introducing , mncnumbers, rmonic es. + um ¨ s mathematics, dtergener- their nd oi num- monic ilsnu- yields urkey n 1 aual find naturally i numbers nic , where nCan un ¨ r inte- are e H 0 § is (p,1) (p) n p (1,r) (r) Hn = Hn = k=1 1/k and the hyperharmonic numbers Hn = hn = n h(r−1). k=1 k P There is an extensive literature on the harmonic and hyperharmonic num- Pbers. Among which we emphasize the formulas

n n +1 (−1)k B = n!H (1) k +1 k n+1 k X=0 (see [10] and see also [6, 21, 32]) and n n + r B = n!h(r−1) (2) k + r k n+1 k r X=0 (see [6, 21]). Here Bn is nth Bernoulli number defined by means of the generating function t ∞ tn = B , et − 1 n n! n=0 X n+r is the r-Stirling number of the first kind defined by k+r r n n + r (x + r)n = (x + r) (x + r + 1) ··· (x + r + n − 1) = xk k + r k r X=0 (see [5]), and n+1 = n+1 is the ordinary Stirling number of the first kind. k+1 1 k+1 Equations (1) and (2) represent harmonic and hyperharmonic numbers in terms of the Stirling and r-Stirling numbers of the first kind, and Bernoulli numbers. These then give rise to the natural question of representing generalized harmonic and generalized hyperharmonic numbers as similar formulas. An affirmative answer to this question is given by the following theorem, which represents the generalized hyperharmonic numbers in terms of the r- (p) Stirling numbers and poly-Bernoulli polynomials Bn (x) . The polynomials (p) Bn (x) are defined by

∞ tn Li (1 − e−t) B(p) (x) = p ext (3) n n! 1 − e−t n=0 X ∞ n p (see [2]), where Lip (t)= n=1 t /n stands for the polylogarithm function. Theorem 1 For all non-negativeP integers n and r, we have

n n + r B(p) (q)= n!H(p,q+r). (4) k + r k n+1 k r X=0 As a result, we deduce the following identity which answers the question of which type of the Bernoulli numbers are related to the generalized harmonic (p) numbers Hn .

2 Corollary 2 For any non-negative integer n

n n +1 B(p) = n!H(p) , (5) k +1 k n+1 k X=0 (p) (p) where Bk = Bk (0) is the kth poly-Bernoulli number (see [19]).

We also observe the following correspondence between the generalized (q) hyperharmonic numbers and the hyper-sums Sp (n):

(−p,q+1) (q) Hn = Sp (n) . (6)

The hyper-sum is introduced by Faulhaber as

n (r) (r−1) Sp (n)= Sp (k) , k X=1 (0) p p p with Sp (n) = Sp (n)=1 +2 + ··· + n (see [23]). The sums of powers of integers Sp (n) have been interested since the classical times, for details see [15, 16, 23, 26]. Some recent studies on the hyper-sums include explicit formulas, connection with the Bernoulli numbers, congruences, generating functions, and recurrence formulas ([8, 9, 18, 25]). The surprising correspondence (6) gives rise to numerous identities such as

p p q + n + j S(q) (n)= (−1)p+j n j!, p j q +1+ j j=1 X and congruences like

n + q +1 S(q) (n) ≡ (mod p), p q +2 for a prime number p. Apart from the identities (1) and (2), the harmonic numbers are related to the Bernoulli numbers and polynomials via

n n 2n! (−1)k−1 kB = H (7) k k−1 n +1 n k X=1 (see [22]) and

n n + m n + r + m − 1 kB (r)=n! (H − H )2 k + m k−1 r + m − 2 n+r+m−1 r+m−2 k m X=1 n (2) (2) −Hn+r+m−1 + Hr+m−2 (8) o

3 ′ (see [21]). Since Bk (x) = kBk−1 (x), the terms kBk−1 and kBk−1 (r) in the summands suggest whether n n + m dl B (x) k + m dxl k k=1 m x=r X can be written in terms of the harmonic numbers of any kind. We deal with this problem in Theorem 6. It is worthwhile to mention that in the proof of Theorem 6, the question raised in [13, 27] on the general form of higher order (x) derivatives of hn with respect to x is answered in a different sense. The organization of the paper is as follows. Section 2 is a preliminary section in which we give notation and basic deﬁnitions needed in the paper. In Section 3 we prove Theorems 1 and 6, and obtain some recurrence formulas for the hyperharmonic and generalized harmonic numbers. In Sec- (q) tion 4 utilizing (6) we derive numerous formulas for the hyper-sums Sp (n). Moreover, new results for the sums of powers of integers Sp (n) are presented. We conclude the paper with Section 5 where we present some congruences for the generalized hyperharmonic numbers, hyperharmonic numbers and hyper-sums.

2 Preliminaries

The r-Stirling numbers, which are natural generalizations of the ordinary Stir- ling numbers, may be defined either in a combinatorial or in an analytic way. Analytic way includes the generating functions. The r-Stirling numbers of the second kind, denoted by n+r , are defined by means of the exponential k+r r generating function ∞ n + r zn (ez − 1)k = erz (9) k + r n! k! n k r X= (see [5]). The r-Stirling numbers of the first and the second kind are related via the r-Stirling transform: n n n + r n + r b = a if and only if a = (−1)n−k b n k + r k n k + r k k r k r X=0 X=0 (see [5]). In particular n = n , n+1 = n+1 , where n is the ordinary k 0 k k+1 1 k+1 k Stirling number of the second kind, and n = n , n+1 = n+1 , where n k 0 k k+1 1 k+1 k is the ordinary Stirling number of the first kind. We list some of the basic facts about the Stirling numbers in the following. Lemma 3 We have 0 n 0 n n n(n−1) n (1) 0 =1, 0 = n =0, n> 0, 1 = (n − 1)!, n−1 = 2 , n =1. 0 n 0 n n n n(n−1) (2) 0 =1, 0 = n =0, n> 0, 1 = n =1, n−1 = 2 ([11]). 4 n+r n n+r (r) n+r n(n−1) n+r (3) = r , = n!hn , = + nr, = 1, and 0+r r 1+r r n−1+r r 2 n+r r n+n = δ , where δ stands for the Kronecker’s delta ([5]). m+n n m,n m,n (4) For any r, n+r ≡ n+r ≡ 0 (mod n) for a prime number n, provided k+r r k+r r that k =2, 3,...,n − 1 ([17]). (p,q) The generating function of the generalized hyperharmonic numbers Hn is given by ∞ Li (t) H(p,q)tn = p (10) n (1 − t)q n=0 X (see [14]), which reduces to

∞ ln (1 − t) h(q)tn = − (11) n (1 − t)q n=0 X (q) (see [12]), the generating function of the hyperharmonic numbers hn , that is related to the harmonic numbers by

n + q − 1 h(q) = (H − H ) . (12) n q − 1 n+q−1 q−1 (p) The poly-Bernoulli polynomials Bn (x), deﬁned in (3), can be expressed (p) (p) in terms of the poly-Bernoulli numbers Bn = Bn (0) as

n n B(p) (x)= B(p)xn−k. n k k k X=0 The poly-Bernoulli numbers can be also represented in terms of the Stirling numbers of the second kind by

n n (−1)k k! B(p) = (−1)n (13) n k (k + 1)p k X=0 (see [19]). Hence 1 1 B(p) =1, B(p) = , B(p) (x)=1, B(p) (x)= x + , (14) 0 1 2p 0 1 2p etc. The poly-Bernoulli polynomials and numbers, which are studied recently in different directions ([2, 7, 24, 34]), are generalizations of the Bernoulli poly- (1) nomials Bn (x) and the Bernoulli numbers Bn in that Bn (x − 1) = Bn (x) (1) (1) and Bn = Bn with B1 = −B1.

5 3 Generalized hyperharmonic numbers

We start this section by proving Theorem 1. Proof of Theorem 1. Using (3) and (10) we have ∞ tk Li (1 − e−t) B(p) (q +1 − r) = p e(q+1−r)t k k! 1 − e−t k X=0 ∞ (e−t − 1)n = (−1)n n!H(p,q+1) e−rt. (15) n+1 n! n=0 X Here by considering (9) we obtain

k ∞ tk ∞ k + r tk B(p) (q +1 − r) = (−1)k−n n!H(p,q+1) . k k! n + r n+1 k! k k n=0 r ! X=0 X=0 X tk Then, comparing the coefﬁcients of k! in the both sides of the above equation gives n n + r B(p) (q +1 − r)= (−1)n−k k!H(p,q+1). (16) n k + r k+1 k r X=0 Applying the r-Stirling transform to the above equation, we obtain the desired result (4). It worths to note recent works [29] and [30], which relate the Stirling numbers of the ﬁrst kind and poly-Bernoulli numbers in different manner. The following interesting alternating sum 1 −H + H − H + ···− H + H = H 1 2 3 2n−1 2n 2 n motivates the next result, which follows from the relation

1−p 2 Lip (−t) + Lip (t)=2 Lip t . Proposition 4 We have

n 2 q + k − 1 1 (−1)k H(p,q) = H(p,q). k 2n−k 2p n k X=0 In particular,

2n 2n 1 q + k − 1 1 (−1)k H(p) = H(p) and (−1)k h(q) = h(q). 2n−k 2p n k 2n−k 2 n k k X=0 X=0 Proposition 5 We have

n p − q H(p,q+1) = (−1)k H(p,p+1). n k n−k k X=0

6 In particular, n p H(p) = (−1)k H(p,p+1). n k n−k k X=0 Proof. We have

∞ (p,q+1) n Lip (t) p−q Hn t = p+1 (1 − t) n=0 (1 − t) X ∞ n p − q = H(p,p+1) (−1)n−k tn, k n − k n=0 k ! X X=0 from which the desired result follows. We now turn our attention to the sum

n n + r dl B (x) , k + r dxl k k=1 r x=r X which is a more general form of (1), (7) and (8). To evaluate this sum we ﬁrst recall that ∞ m + k (− ln (1 − t))r P (r, m + k,m) tk = (17) m m+1 k (1 − t) X=0 (see [35]), where

(1) (1) (2) (2) (r) (r) P (r, m + k,m)= Pr Hm+k − Hm ,Hm+k − Hm ,...,Hm+k − Hm , and the polynomial Pn (x1, x2,...,xn) is deﬁned by P0 =1 and

n Pn (x1, x2,...,xn) = (−1) Yn (−0!x1, −1!x2,..., − (n − 1)!xn) , where Yn is the exponential Bell polynomial [11]. A ﬁrst few of them may be 2 3 listed as P1 (x1)= x1, P2 (x1, x2)= x1 − x2, P3 (x1, x2, x3)= x1 − 3x1x2 +2x3.

Theorem 6 For nonnegative integers q, r, and n, we have

n n + r k (k − 1) ··· (k − l + 1) B (q) k + r k−l k l r X= n + q + r − 1 = n! P (l +1,n + q + r − 1, q + r − 2) . (18) q + r − 2 Proof. Differentiating both sides of (11) with respect to xl times gives

∞ dl (− ln (1 − t))l+1 h(x+1)tn = . dxl n x+1 n=0 (1 − t) X

7 Setting x = q in the above equation and using (17), we see that

dl q + n h(x+1) = P (l +1, q + n, q) . (19) dxl n q x=q

(1) For p =1 and q → x − 1, with the use of Bk (x − 1) = Bk (x) , (4) turns into

n 1 n + r B (x)= h(x+r−1). n! k + r k n+1 k r X=0 ′ Iterating Bk (x) = kBk−1 (x) l times on the left-hand side and observing (19) for the right-hand side yield the closed formula (18). It is seen that (18) reduces to (1) for r = q = 1 and l = 0, and to (8) for l =1. Another demonstration of (18) is the following example corresponding to l =2:

n 1 n + r k (k − 1) B (q) n! k + r k−2 k r X=2 n + q + r − 1 = (H − H )3 +2 H(3) − H(3) q + r − 2 n+q+r−1 q+r−2 n+q+r−1 q+r−2 n (2) (2) −3 (Hn+q+r−1 − Hq+r−2) Hn+q+r−1 − Hq+r−2 , o in particular for r = q =1

n 1 n +1 (−1)k k (k − 1) B = (H )3 − 3H H(2) +2H(3) . n! k +1 k−2 n+1 n+1 n+1 n+1 k X=2

Remark 1 We note that (19) gives an answer to the question raised in [13, 27] (x) on the general form of higher order derivatives of hn with respect to x.

4 Hyper-sums of powers

(q) The hyper-sums of powers of integers Sp (n) are deﬁned recursively by

n (q) (q−1) Sp (n)= Sp (k) , k X=1 (0) (q) with the initial condition Sp (n) = Sp (n). We note that Sp (n) satisﬁes the relation n n + q − k S(q) (n)= kp p q k X=1

8 (see [25]). We also observe that n n + q − k 1 H(p,q+1) = n q kp k X=1 (see [14, p. 1648]). These simply imply the relation

(−p,q+1) (q) Hn = Sp (n) . (20) Therefore, we can translate the results given for the generalized hyperhar- (p,q) (q) monic numbers Hn to the hyper-sums Sp (n). For instance,

2n 1 q +2n − k S(q) (n)= (−1)k S(q) (k) , p 2p 2n − k p k=0 n X n − k + p + q − 1 S(q) (n)= S(p) (k) , p n − k p k X=0 which follow from Proposition 4. In particular, we have

n n 1 2 n − k + p − 1 S (n)= (−1)k S (k) and S (n)= S(p) (k) . p 2p p p n − k p k k X=0 X=0 (−p) (−n) Moreover, using the duality theorem Bn = Bp for the poly-Bernoulli numbers [19, Theorem 2] in (5), we obtain an evaluation formula for sums of powers of integers:

n 1 n +1 B(−k) = S (n + 1) . n! k +1 p p k X=0 Our next result utilizes the following lemma, which is a polynomial extension of the Arakawa-Kaneko formula

min{n,p} p +1 n +1 B(−p) = (j!)2 n j +1 j +1 j=0 X (see [1, Theorem 2]).

Lemma 7 We have

min{n,p} p +1 n + q +1 B(−p) (q)= (j!)2 . (21) n j +1 j + q +1 j=0 q+1 X Proof. We can write (10) in the form

∞ ∞ p (q+2)t+y k y e (−1)k H(−p,q+1) e−t − 1 = 1 − e−t k p! 1 − (1 − et)(1 − ey) p=0 k=1 X X 9 by setting t → 1 − e−t. On the other hand, we equivalently have

∞ ∞ (e−t − 1)k yp ∞ (1 − et)j (1 − ey)j (−1)k H(−p,q+1) = (j!)2 e(q+2−r)t ey. k+1 ert p! j! j! p=0 k j=0 X X=0 X Last equality, (15) and (9) give the desired result. Now, taking −p in (4) and using (21) yield

n n p +1 n + r k + q +1 n!H(−p,q+r) = (j!)2 . n+1 j +1 k + r j + q +1 j=0 k j r q+1 X X= Utilizing the formula

n n + r k + s n! n + r + s − 1 = k + r j + s j! j + r + s − 1 k j r s X= (see [28, Theorems 3.7 and 3.11]), we obtain the following.

Theorem 8 For non-negative integers p, q, and n, we have

p p +1 n + q S(q) (n)= j! . (22) p j +1 j + q +1 j=0 X Remark 2 We note that a slightly different form of (22) can be found in [8, 9].

Equation (22) provides a natural extension of the expression

p p n +1 S (n)= j! p j j +1 j=0 X (q) (see [23, p. 285]) to Sp (n). (q) Next result is an alternative representation for Sp (n), which extends

p p n + j S (n)= (−1)p+j j! (23) p j j +1 j=1 X (see [23, p. 285]).

Theorem 9 We have

p p n + q + j S(q) (n)= (−1)p+j j!. (24) p j q + j +1 j=1 X

10 Proof. We set −p for p in (10). Then

∞ ∞ (−p,q) n 1 p k 1 t H t = k t = wp , n (1 − t)q q+1 1 − t n=0 k (1 − t) X X=1 where wn (x) is the nth geometric polynomial, deﬁned by

∞ 1 x kntk = w (25) 1 − x n 1 − x k X=1 n (see [3]). Using (25), the relation (1 + x) wn (x) = x (−1) wn (−x − 1) ,n> 0, (see [20, Eq. (22)]) and p p w (x)= j!xj p j j=0 X (see [3]), we reach at p ∞ ∞ p n + k + q − 1 H(−p,q)tn = tn (−1)j+p j!, n j n − 1 n=0 n=1 j=0 X X X which implies the result.

Remark 3 A slightly different form of (24) may be written as

p p +1 n + q + j +1 S(q) (n)= (−1)p+j j!, p j +1 q + j +1 j=0 X which can be obtained by utilizing

p p +1 −1 k+1 Li (t) = (−1)p+1 k! −p k +1 1 − t k X=0 in the generating function (10).

Taking into account (6), it can be seen from (16) and (24) that

p n p n +1 B(−p) (q)= j! (−1)p+j (−1)n−k (j + q + 2)k . n j k +1 j=1 k X X=0 This reduces to p p B(−p) (q)= (−1)p+j j! (j + q + 1)n (26) n j j=1 X utilizing the formula n + r (x − r)n = (−1)n−k xk k + r n k r X=

11 (see [5]). (26) stands for a polynomial extension of the Arakawa-Kaneko formula p p B(−p) = (−1)p+j j! (j + 1)n n j j=1 X (cf. [1]). We conclude this section by stating further results for hyper-sums and sums of powers, and some of their consequences.

Theorem 10 We have q 1 q + n +1 S(q) (n)= (−1)k S (n) , (27) p q! k + n +1 p+k k=0 n+1 q X q + n +1 k + n S (n)= (−1)k k!, (28) q k + n +1 k +1 k=0 n+1 Xq q + n +1 S (n)= (−1)k k!h(k+1). (29) q−1 k + n +1 n k n+1 X=0 (k) Proof. For (27), we appeal to the generating function of Sp (n) depending on the index k ∞ n (k) k 1 p j Sp (n) z = n+1 j (1 − z) k (1 − z) j=1 X=0 X (cf. [25]). We set z → 1 − e−t and see that

∞ k n (e−t − 1) (−1)k k!S(k) (n) e−(n+1)t = jpe−jt. p k! k j=1 X=0 X Hence, the proof follows from (9) and r-Stirling transform. (29) follows similarly by using the generating function

∞ n 1 (1 − z)j h(k+1)zk = n n+1 j k (1 − z) j=1 X=0 X (see [21]). Utilizing the identity

q q + n q j = nq−j k + n j k n j k X= (cf. [5]) and (23) we ﬁnd that q q + n +1 k + n (−1)k k! k + n +1 k +1 k n+1 X=0

12 q q = (−1)j (n + 1)q−j S (q) j j j=0 X q q q j q j −j = (n + 1) (−1) m (n + 1) = Sq (n) ,  j  m=0 j=0 X X   which is (28). Equations (22) and (27) yield the following interesting identity.

Proposition 11

q (−1)j n n + q − j n + q = (H − H − H ) . j j n q n+q q n j=1 X Proof. Taking p = −1 in (27) gives

q q + n +1 n + q q!h(q+1) = (−1)k S (n)+ q! H , n k + n +1 k−1 q n k n+1 X=1 by Lemma 3 (3). We now make use of (22) with q =0 and deduce that

q q + n +1 (−1)k S (n) k + n +1 k−1 k=1 n+1 X q q n q + n +1 k = (j − 1)! (−1)k . j k + n +1 j j=1 k j n+1 X X= Here we appeal to the equation

q q + r k + s q (−1)k−j = (r − s)q−j k + r j + s j k j r s X= (see [5, 28]). Hence,

q q q + n +1 n q (−1)k S (n)= (−1)j (j − 1)! (n + 1)q−j k + n +1 k−1 j j k n+1 j=1 X=1 X q (−1)j n n + q − j = q! . j j n j=1 X Thus, the proof follows from (12).

13 5 Congruences

In this section we present several congruences for the hyperharmonic numbers, the generalized hyperharmonic numbers, and the hyper-sums. The mo- tivation rises from the formulas established in Sections 3 and 4, where hyperharmonic numbers and hyper-sums are formulated in terms of the Stirling numbers. Throughout by a congruence x ≡ y (mod m), we mean that the rational a c numbers x = b and y = d satisfy x ≡ y (mod m) if and only if m| (ad − bc). We also call a rational number as an m-integer whenever its denominator is not divisible by m. We start by reviewing some basic facts about Bernoulli numbers and polynomials.

Lemma 12 We have (1) (von Staudt-Clausen) For n =1 and for any even integer n ≥ 2, we have 1 B = A − , n n p (pX−1)|n where An is an integer and p is a prime number. Equivalently, we have

−1 (mod p), if p − 1 divides n, pB ≡ n 0 (mod p), if p − 1 does not divide n.

Bn(q)−Bn (2) If q is an integer, then n is an integer ([31, p. 6]). The following result is about a divisibility property for generalized hyperharmonic numbers.

Theorem 13 Let n be an odd prime. For positive integers p and q, we have

p (p,q) n Hn+1 ≡ q (mod n).

Proof. Let n be an odd prime and p ≥ 2. Setting r =0 in (4), multiplying both sides by np−1 and separating out the terms with k = 0, k = 1, k = n − 1, and k = n, we have

n n np−1n!H(p,q) = np−1 B(p) (q)+ np−1 B(p) (q) n+1 0 0 1 1 n−2 n n n + np−1 B(p) (q)+ np−1 B(p) (q)+ B(p) (q) . n − 1 n−1 n n k k k X=2 We consider each of these terms separately. p−1 n (p) By Lemma 3 (1), we have n 0 B0 (q)=0.

14 By Lemma 3 (1) and (14), we have n 1 np−1 B(p) (q)= np−1 (n − 1)! q + . 1 1 2p We write n−2 n − 1 B(p) (q)= B(p) + B(p)qn−1−m. n−1 n−1 m m m=0 X (p) Now, from (13) we observe that Bm is an n-integer for m = 0, 1,...,n − 2. p (p) p (p) On the other hand, n Bn−1 is an n-integer satisfying n Bn−1 ≡ −1 (mod n). Thus, by Lemma 3 (1), we conclude that

n−2 n n − 1 n − 1 n − 1 np−1 B(p) (q)= npB(p) + np B(p)qn−1−m n − 1 n−1 n−1 2 2 m m m=0 X n − 1 ≡− (mod n). 2 Since n −2 n B(p) (q)= B(p) + nB(p) q + B(p)qn−m, n n n−1 m m m=0 X we have n n −2 n np−1 B(p) (q)= np−1B(p) + npB(p) q + np−1 B(p)qn−m. n n n n−1 m m m=0 X (p) Since p ≥ 2 and Bm is an n-integer for m =0, 1,...,n − 2, we obtain that

n −2 n np−1 B(p)qn−m ≡ 0 (mod n). m m m=0 X p (p) We also have n Bn−1q ≡−q (mod n). Now, in [1, Theorem 1] it has been also p−1 (p) proved that n Bn is an n-integer with 1 n n np−1B(p) ≡ − (mod n). n n n − 1 2p n n(n−1) p−1 (p) n−1 Since n−1 = 2 , wethenhave n Bn ≡ 2 (mod n). Thus, we arrive at n n − 1 np−1 B(p) (q) ≡ − q (mod n). n n 2 Finally we write

n n k −2 n −2 n k np−1 B(p) (q)= np−1 B(p)qk−m. k k k m m k k m=0 X=2 X=2 X

15 (p) By Lemma 3 (4) and the fact that Bm is an n-integer for m =0, 1,...,k, where k =2, 3,...,n − 2, we ﬁnd that

n −2 n np−1 B(p) (q) ≡ 0 (mod n). k k k X=2 Combining these results we conclude that

1 n − 1 n − 1 np−1n!H(p,q) ≡ np−1 (n − 1)! q + − + − q n+1 2p 2 2 ≡−q (mod n).

The result follows from Wilson’s theorem. (1) Now for an odd prime n, let p = 1 and r = 0 in (4). Since Bk (q) = 1 Bk (q + 1), with B1 (q +1)= q + 2 , we have n n n n!H(1,q) = B (q +1)+ B (q +1)+ B (q + 1) n+1 0 0 1 1 n − 1 n−1 n−2 n n + B (q +1)+ B (q + 1) n n k k k X=2 1 n (n − 1)2 B (q + 1) − B n − 1 = (n − 1)! q + + n−1 n−1 + nB 2 2 n − 1 2 n−1 n n B (q + 1) − B −2 n B (q + 1) − B −2 n + n n n + B + k k k + B . n n k k k k k k X=2 X=2 By Lemma 12 (2),

n (n − 1)2 B (q + 1) − B B (q + 1) − B n−1 n−1 ≡ n n n ≡ 0 (mod n), 2 n − 1 n and Bn = 0 since n is an odd prime. On the other hand, by Lemma 12 (1), nBn−1 ≡−1 (mod n), and Bk is an n-integer for k = 2, 3,...,n − 2. Then, by Lemma 3 (4), we have

1 n − 1 1 n − 1 n!H(1,q) ≡ (n − 1)! q + − ≡ (n − 1)! q + + (mod n) n+1 2 2 2 2 by Wilson’s theorem. Canceling (n − 1)! gives the desired result. Next we present a symmetric-type congruence for hyperharmonic numbers.

Theorem 14 For a prime q and an integer n with 1 ≤ n ≤ q − 1, we have

(q+1) (n+1) q hn + nhq ≡ n (mod q).

16 Proof. Let 1 ≤ n ≤ q − 1 and q be a prime. Applying the r-Stirling transform in (29) we ﬁnd that

q q + n +1 q!h(q+1) = (−1)k S (n) n k + n +1 k−1 k n+1 X=0 q + n +1 q + n +1 = H − S (n) + (−1)q S (n) 0+ n +1 n 1+ n +1 0 q−1 n+1 n+1 q−1 q + n +1 + (−1)k S (n) k + n +1 k−1 k n+1 X=2 (n+1) q ≡ (n + 1) ··· (n + q) Hn − nq!hq + (−1) Sq−1 (n) (mod q) by Lemma 3 (3) and (4). Now, 1 ≤ n ≤ q − 1 implies that

(n +1)(n + 2) ··· (n + q) Hn ≡ 0 (mod q) and q−1 q−1 q−1 Sq−1 (n)=1 +2 + ··· + n ≡ n (mod q). Thus, (q+1) (n+1) q q! hn + nhq ≡ (−1) n (mod q). If q =2, then n =1, and we have

5 2 h(3) + h(2) =2 1+ ≡ 1 (mod 2). 1 2 2 (q+1) (n+1) Otherwise, q! hn + nhq ≡ −n (mod q) and the result follows from Wilson’s theorem. (q) We conclude this section by stating two congruences about Sp (n). In (q) [25] congruences for Sp (n) when n is a prime number were given. Here we give results modulo a prime number p.

Theorem 15 For a prime number p, we have

n + q +1 S(q) (n) ≡ (mod p). p q +2 Proof. Let p be a prime in (24). Then

p n + q +1 p n + q + p S(q) (n) = (−1)p+1 + p! p 1 q +2 p q + p +1 p−1 p n + q + j + (−1)p+j j!. j q + j +1 j=2 X By Lemma 3 (2) and (4), we obtain the desired result.

17 Next result seems obvious, but we record it here as an application of the formula for the hyper-sums in terms of the poly-Bernoulli polynomials with negative index. For this we need the following result about the poly-Bernoulli polynomials, which follows from (26), Lemma 3 (2) and (4).

Lemma 16 Given a prime number p, a positive integer n, and a nonnegative integer q, we have (−p) n Bn (q) ≡ (q + 2) (mod p).

Proposition 17 Given a prime number p and a nonnegative integer q, we have

(q) pSp (p + 1) ≡ 0 (mod p).

Proof. With the use of (20) in Theorem 1 for r =0, we ﬁnd that

n n n!S(q) (n +1)= B(−p) (q) . p k k k X=0 Now let p be a prime and set n = p. By Lemma 3 (1) and (4),

p−1 p p p p p!S(q) (p +1)= B(−p) (q)+ B(−p) (q)+ B(−p) (q)+ B(−p) (q) p 0 0 1 1 p p k k k X=2 (−p) (−p) ≡ (p − 1)!B1 (q)+ Bp (q) (mod p).

By Lemma 16 and Wilson’s theorem, we conclude that

(q) p−1 pSp (n + 1) ≡− (q + 2) (q + 2) − 1 (mod p), h i and the result follows from Fermat’s little theorem.

References

[1] Arakawa T, Kaneko M. On poly-Bernoulli numbers, Comment. Math. Univ. St. Paul. 48 (2) (1999) 159–167. [2] Bayad A, Hamahata Y. Polylogarithms and poly-Bernoulli polynomials, Kyushu J. Math. 65 (2011) 15–24. [3] Boyadzhiev KN. A series transformation formula and related polynomials, Int. J. Math. Math. Sci. 23 (2005) 3849–3866. [4] Boyadzhiev KN. Notes on the Binomial Transform, World Scientiﬁc, Sin- gapore, 2018.

[5] Broder AZ. The r-Stirling numbers, Discrete Math. 49 (1984) 241–259.

18 [6] Can M, Daglı˘ MC. Extended Bernoulli and Stirling matrices and related combinatorial identities, Linear Algebra Appl. 444 (2014) 114–131. [7] Cenkci M, Young PT. Generalizations of poly-Bernoulli and poly-Cauchy numbers, European J. Math. 1 (2015) 799–828. [8] Cereceda JL. Generalized Akiyama-Tanigawa algorithm for hypersums of powers of integers, J. Integer Seq. 16 (2013) Article 13.3.2. [9] Cereceda JL. Iterative Procedure for hypersums of powers of integers, J. Integer Seq. 17 (2014) Article 14.5.3. [10] Cheon G-S, El-Mikkawy MEA. Generalized harmonic numbers with Ri- ordan arrays, J. Number Theory 128 (2) (2008) 413–425. [11] Comtet L. Advanced Combinatorics, Riedel, Dordrech, Boston, 1974. [12] Conway JH, Guy RK. The Book of Numbers. New York, USA: Springer- Verlag, 1996. [13] Dil A. On the hyperharmonic function, SDU J. Nat. Appl. Sci. 23 (2019) 187–193. [14] Dil A, Mezo˝ I, Cenkci M. Evaluation of Euler-like sums via Hurwitz zeta values, Turk. J. Math. 41 (2017) 1640–1655. [15] Graham RL, Knuth DE, Patashnik O. Concrete Mathematics, Addison- Wesley, Reading, MA, 1989. [16] Gould HW. Evaluation of sums of convolved powers using Stirling and Eulerian numbers, Fibonacci Quart. 16 (1978) 488–497. [17] Hsu LC, Shiue PJ-S. A uniﬁed approach to generalized Stirling numbers, Adv. Appl. Math. 20 (1998) 366–384. [18] Inaba Y. Hyper-sums of powers of integers and the Akiyama-Tanigawa matrix, J. Integer Seq. 8 (2005) Article 05.2.7. [19] Kaneko M. Poly-Bernoulli numbers, J. Theor.´ Nombres Bordeaux 9 (1997) 221–228. [20] Kargın L. Some formulae for products of geometric polynomials with applications, J. Integer Seq. 20 (2017) Article 17.4.4.

[21] Kargın L, Can M. Harmonic number identities via polynomials with r- Lah coefﬁcients, Comptes Rendus Mathematique, 358 (5) (2020) 535– 550. [22] Kellner BC. Identities between polynomials related to Stirling and harmonic numbers, Integers 14, #A54 (2014)

19 [23] Knuth DE. Johann Faulhaber and sums of powers, Math. Comp. 61 (203) (1993) 277–294. [24] Komatsu T, Liptai K, Mezo˝ I. Incomplete poly-Bernoulli numbers associ- ated with incomplete Stirling numbers, Publ. Math. Debr. 88 (2016) 357– 368. [25] Laissaoui D, Bounebirat F,Rahmani M. On the hyper-sums of powers of integers, Miskolc Math. Notes, 18 (2017) 307–314. [26] Merca M. An alternative to Faulhaber’s formula, Amer. Math. Monthly 122 (6) (2015) 599–601. [27] Mezo˝ I. Analytic extension of hyperharmonic numbers, Online J. Anal. Comb. 4 (2009).

[28] Nyul G, Racz´ G. The r-Lah numbers, Discrete Math. 338 (10) (2015) 1660– 1666. [29] Y. Ohno, Y. Sasaki, Recurrence formulas for poly-Bernoulli polynomials, Adv. Stud. Pure Math., 84 (2020), 353-360. [30] Y. Ohno, Y. Sasaki, Recursion formulas for poly-Bernoulli numbers and their applications, Int. J. Number Theory, https://doi.org/10.1142/S1793042121500081. [31] Rademacher H. Topics in Analytic Number Theory, Springer Verlag, New York, 1973. [32] Wang W. Riordan arrays and harmonic number identities, Comput. Math. Appl. 60 (5) (2010) 1494–1509. [33] Young PT. Symmetries of Bernoulli polynomial series and Arakawa– Kaneko zeta functions, J. Number Theory 143 (2014) 142–161. [34] Young PT. Bernoulli and poly-Bernoulli polynomial convolutions and identities of p-adic Arakawa–Kaneko zeta functions, Int. J. Number The- ory 12 (05) (2016) 1295–1309. [35] Zave DA. A series expansion involving the harmonic numbers, Inform. Process. Lett. 5 (3) (1976) 75–77.