A New Expression for the Bernoulli Numbers and Its Applications Levent Kargın, Merve Mutluer Abstract This paper shows that a finite discrete convolution involving Stirling numbers of both kinds and harmonic numbers can be expressed in terms of the Bernoulli numbers. As applications of this expression, the linear recurrence relation for the Bernoulli numbers given by Agoh is reproved, and a new recurrence relation for the Bernoulli numbers is obtained. Furthermore, it is shown that a cumulative sum of the Bernoulli numbers can be written in terms of the Bernoulli and di-Bernoulli numbers. Finally, congruences for the sums of the Bernoulli and Euler numbers are established. Keywords: Bernoulli numbers, di-Bernoulli numbers, Euler numbers, harmonic numbers, Stirling numbers, congruences. MSC: 11A07, 11B68, 11B73. 1 Introduction The Bernoulli numbers, denoted by B_{n} , first arose in the context of computing sums of powers of positive integers in closed form, a topic that has been studied since antiquity. They may be defined by means of the generating function \frac{t}{e^{t}-1}=\sum_{n=0}^{\infty}B_{n}\frac{t^{n}}{n!},\text{\ \ \ \ }|t|<2\pi. It is easy to find the values B_{0}=1 , B_{1}=-1/2 , B_{2}=1/6 , B_{3}=0 , B_{4}=-1/30 , and B_{2n+1}=0 for all n\geq 1 . They also appear in the Taylor series expansions of the tangent and hyperbolic tangent functions, the Euler–Maclaurin summation formula, and certain representations of the Riemann zeta function. The Bernoulli numbers have consequently become one of the cornerstones of mathematical science, with numerous approaches available for their introduction, including generating functions, recurrence relations, integral representations, and explicit formulas. Among the numerous known explicit formulas, one of the most famous is Worpitzky’s identity \sum_{k=1}^{n}\left(-1\right)^{k}{n\brace k}\frac{k!}{k+1}=B_{n}, (1) which was later generalized to \sum_{k=j}^{n}\frac{\left(-1\right)^{k-j}}{k}{n\brace k}{k\brack j}=\binom{n-1}{j}\frac{B_{n-j}}{n-j}, (2) where 1\leq j\leq n (cf. [ 14 , p.217] ). Here \genfrac{[}{]}{0.0pt}{0}{n}{k} and \genfrac{}{}{0.0pt}{0}{n}{k} denote the Stirling numbers of the first and second kind, defined by \left(x\right)^{\bar{n}}=x\left(x+1\right)\cdots\left(x+n-1\right)=\sum_{k=0}^{n}{n\brack k}x^{k} (3) and x^{n}=\sum_{k=0}^{n}\left(-1\right)^{n-k}{n\brace k}\left(x\right)^{\bar{k}}, respectively (see, for example [ 14 , Chapters 9 and 12] ), and \dbinom{n}{j} is the binomial coefficient. In 2010, Mező [ 10 ] extended the identity ( 2 ) and derived a formula for the Bernoulli polynomials by using the r -Whitney numbers. Motivated by this approach, Merca further showed that a finite discrete convolution involving different generalizations of the Stirling numbers of both kinds can be expressed in terms of the Bernoulli polynomials [ 12 , 13 ] . In a recent paper [ 3 ] , new explicit expressions for the Bernoulli numbers, namely, \displaystyle\sum_{k=1}^{n}\left(-1\right)^{k-1}\genfrac{}{}{0.0pt}{0}{n}{k}\left(k-1\right)!H_{k} \displaystyle=B_{n-1}, (4) \displaystyle\sum_{k=2}^{n}\left(-1\right)^{k}\genfrac{}{}{0.0pt}{0}{n}{k}\left(k-1\right)!H_{k-1}H_{k} \displaystyle=\frac{n+1}{2}B_{n-2}, (5) were introduced. Here H_{n}=1+\frac{1}{2}+\frac{1}{3}+\cdots+\frac{1}{n} is the n th harmonic number. The identities \left(k-1\right)!=\genfrac{[}{]}{0.0pt}{0}{k}{1}\text{ and }\left(k-1\right)!H_{k-1}=\genfrac{[}{]}{0.0pt}{0}{k}{2} (6) motivate us to examine the closed form of the sum \sum_{k=j}^{n}\left(-1\right)^{k-j}\genfrac{}{}{0.0pt}{0}{n}{k}\genfrac{[}{]}{0.0pt}{0}{k}{j}H_{k}. As stated in the following theorem, this sum gives rise to a new formula for the Bernoulli numbers. Theorem 1 (Main Theorem) For integers n and j with 0\leq j\leq n , we have \sum_{k=j}^{n}\left(-1\right)^{k-j}\genfrac{}{}{0.0pt}{0}{n}{k}\genfrac{[}{]}{0.0pt}{0}{k}{j}H_{k}=\left(\dbinom{n}{j}-1\right)\frac{B_{n-j}}{n-j}. (7) We note that for j=1 and j=2 , the formula ( 7 ) reduces to ( 4 ) and ( 5 ), respectively. Moreover, using the following known relation \genfrac{[}{]}{0.0pt}{0}{k}{3}=\frac{1}{2!}\left(k-1\right)!\left[\left(H_{k-1}\right)^{2}-H_{k-1}^{\left(2\right)}\right], where H_{n}^{\left(m\right)}=1+\frac{1}{2^{m}}+\cdots+\frac{1}{n^{m}} is the generalized harmonic number, we reach at the following special case of ( 7 ): \sum_{k=3}^{n}\left(-1\right)^{k-1}\genfrac{}{}{0.0pt}{0}{n}{k}\left(k-1\right)!\left[\left(H_{k-1}\right)^{2}-H_{k-1}^{\left(2\right)}\right]H_{k}=\frac{n^{2}+2}{3}B_{n-3}. As applications of the main theorem, we reprove the linear recurrence relation for the Bernoulli numbers given by Agoh [ 1 ] and derive a new recurrence relation for the Bernoulli numbers. Moreover, we introduce a closed formula for cumulative sum of the Bernoulli numbers. This formula leads us to deduce some congruences for the sums of Bernoulli and Euler numbers modulo an odd prime and its square. These are presented in the next section. The proof of Theorem 1 is given in the final section. 2 Applications of the Main Theorem Multiplying both sides of ( 7 ) with x^{j} and summing over j from 1 to n and using ( 3 ) we obtain \sum_{j=1}^{n}\left(\dbinom{n}{j}-1\right)\frac{B_{j}}{j}x^{n-j}=\text{ }_{H}w_{n}^{\left(x\right)}-H_{n}x^{n}, (1) where n is a positive integer, x is a non-zero complex variable and {}_{H}w_{n}^{\left(x\right)}=\sum_{k=1}^{n}\genfrac{}{}{0.0pt}{0}{n}{k}\dbinom{x}{k}k!H_{k}. (2) In this section, we list some special cases of ( 1 ). First we have the following linear identity which was also proved by a different method in [ 1 ] . Corollary 1 For any positive integers n and m , we have \sum_{j=1}^{n}\left(\dbinom{n}{j}-1\right)\frac{B_{j}}{j}m^{n-j}=m^{n}\left(H_{m}-H_{n}\right)-\sum_{j=1}^{m}\frac{\left(m-j\right)^{n}}{j}. (3) Proof. x=m\geq 1 in ( 1 ) gives \sum_{j=1}^{n}\left(\dbinom{n}{j}-1\right)\frac{B_{j}}{j}m^{n-j}=\text{ }_{H}w_{n}^{\left(m\right)}-H_{n}m^{n}. (4) To examine {}_{H}w_{n}^{\left(m\right)} , we first use the generating function of the Stirling numbers of the second kind \sum_{n=k}^{n}\genfrac{}{}{0.0pt}{0}{n}{k}\frac{t^{n}}{n!}=\frac{\left(e^{t}-1\right)^{k}}{k!}, (5) and obtain \sum_{n=1}^{\infty}\text{ }_{H}w_{n}^{\left(m\right)}\frac{t^{n}}{n!}=\sum_{k=1}^{m}\dbinom{m}{k}H_{k}\left(e^{t}-1\right)^{k}. The equality \sum_{k=1}^{m}\dbinom{m}{k}H_{k}\left(z-1\right)^{k}=H_{m}z^{m}-\sum_{k=0}^{m-1}\frac{z^{k}}{m-k} (6) ( [ 2 , Eq. (2.4, iii)] ) gives \sum_{n=1}^{\infty}\text{ }_{H}w_{n}^{\left(m\right)}\frac{t^{n}}{n!}=H_{m}\sum_{n=0}^{\infty}m^{n}\frac{t^{n}}{n!}-\sum_{n=0}^{\infty}\left(\sum_{k=0}^{m-1}\frac{k^{n}}{m-k}\right)\frac{t^{n}}{n!}. Comparing the coefficients of t^{n}/n! yields to {}_{H}w_{n}^{\left(m\right)}=H_{m}m^{n}-\sum_{j=1}^{m}\frac{\left(m-j\right)^{n}}{j}. Inserting this equation in ( 4 ) completes the proof of ( 3 ). Note that ( 3 ) can also be written as \sum_{j=1}^{n}\left(-1\right)^{j}\left(\dbinom{n}{j}-1\right)\frac{B_{j}}{j}m^{n-j}=m^{n}\left(H_{m}-H_{n}+\frac{\left(n-1\right)}{m}\right)-\sum_{j=1}^{m}\frac{\left(m-j\right)^{n}}{j}, since \left(-1\right)^{n}B_{n}=B_{n} for n\geq 2 . In particular, we have \sum_{j=1}^{n}\left(-1\right)^{j}\left(\dbinom{n}{j}-1\right)\frac{B_{j}}{j}=n-H_{n}. Moreover, If we set m=1 and m=2 in ( 3 ) and combine the resulting formulas, we arrive at \sum_{j=1}^{n}\left(\dbinom{n}{j}-1\right)\frac{B_{j}}{j}\left(1-2^{-j}\right)=\frac{1-2^{n-1}}{2^{n}}. A slightly different form of the above equation can serve as a recurrence relation for the Bernoulli numbers. Corollary 2 For a positive integer n , we have \sum_{j=1}^{n}\left(\dbinom{n}{j}+1\right)\frac{B_{j}}{j}\left(1-2^{j}\right)=1. (7) Proof. We first set x=-1/2 in ( 1 ) to obtain \sum_{j=1}^{n}\left(-1\right)^{j}\left(\dbinom{n}{j}-1\right)\frac{B_{j}}{j}2^{j}=\left(-1\right)^{n}2^{n}\text{ }_{H}w_{n}^{\left(-1/2\right)}-H_{n}. Since \sum_{k=1}^{\infty}\binom{2k}{k}H_{k}t^{k}=\frac{2}{\sqrt{1-4t}}\ln\left(\frac{1+\sqrt{1-4t}}{2\sqrt{1-4t}}\right), ( [ 5 , Theorem 1, Eq. (4)] ) and \binom{-1/2}{k}=\frac{\left(-1\right)^{k}}{2^{2k}}\dbinom{2k}{k}, we obtain that \sum_{n=0}^{\infty}\text{ }_{H}w_{n}^{\left(-1/2\right)}\frac{\left(-2t\right)^{n}}{n!}=2e^{t}\ln\left(\frac{e^{t}+1}{2}\right). From the generating function of Euler polynomials [ 14 , Section 15.3] \sum_{n=0}^{\infty}E_{n}\left(x\right)\frac{t^{n}}{n!}=\frac{2}{e^{t}+1}e^{xt}, we deduce that \sum_{n=1}^{\infty}E_{n-1}\left(1\right)\frac{t^{n}}{n!}=2\ln\left(\frac{e^{t}+1}{2}\right). Hence, we reach at \left(-1\right)^{n}2^{n}\text{ }_{H}w_{n}^{\left(-1/2\right)}=\sum_{k=0}^{n-1}\dbinom{n}{k+1}E_{k}\left(1\right). We then use E_{k}\left(1\right)=-E_{k}\left(0\right)=2\left(2^{k+1}-1\right)\frac{B_{k+1}}{k+1} (8) ( [ 14 , Section 15.3] ) to obtain \sum_{j=1}^{n}\left(-1\right)^{j}\left(\dbinom{n}{j}-1\right)\frac{B_{j}}{j}2^{j}-2\sum_{j=1}^{n}\dbinom{n}{j}\frac{B_{j}}{j}\left(2^{j}-1\right)=2n-H_{n}. Since \left(-1\right)^{n}B_{n}=B_{n} for n\geq 2, we arrive at \sum_{j=1}^{n}\dbinom{n}{j}\frac{B_{j}}{j}\left(2^{j-1}-1\right)+\sum_{j=1}^{n}\frac{B_{j}}{j}2^{j-1}=\frac{1}{2}H_{n}-1. ( 7 ) now follows after some elementary manipulations. Before presenting the next result, we recall the poly-Bernoulli polynomials. The poly-Bernoulli polynomials are defined via the generating function \sum_{n=0}^{\infty}\mathbb{B}_{n}^{\left(p\right)}\left(x\right)\frac{t^{n}}{n!}=\frac{\mathrm{Li}_{p}\left(1-e^{-t}\right)}{1-e^{-t}}e^{xt}, (9) ( [ 4 ] ), where \mathrm{Li}_{p}\left(z\right)=\sum_{n=1}^{\infty}z^{n}/n^{p}, is the polylogarithm function. In particular, \mathbb{B}_{n}^{\left(p\right)}\left(0\right)=\mathbb{B}_{n}^{\left(p\right)} is the n th poly-Bernoulli number and when p=2, \mathbb{B}_{n}^{\left(2\right)} is called as di-Bernoulli number ( [ 9 ] ). We now state a cumulative sum of the Bernoulli numbers in terms of the Bernoulli numbers, di-Bernoulli numbers, and the values of the di-Bernoulli polynomials at x=1 . Corollary 3 For an integer n\geq 2 , we have \sum_{j=0}^{n}B_{j}=\mathbb{B}_{n}^{\left(2\right)}\left(1\right)+B_{n}-\mathbb{B}_{n}^{\left(2\right)}-1. Proof. The derivative of ( 1 ) with respect to x is \displaystyle n\sum_{j=1}^{n-1}\left(\dbinom{n}{j}-1\right)\frac{B_{j}}{j}x^{n-j-1} \displaystyle-\sum_{j=1}^{n-1}\left(\dbinom{n}{j}-1\right)B_{j}x^{n-j-1} \displaystyle=\frac{d}{dx}\text{ }_{H}w_{n}^{\left(x\right)}-nH_{n}x^{n-1}. (10) By ( 1 ), we have n\sum_{j=1}^{n}\left(\dbinom{n}{j}-1\right)\frac{B_{j}}{j}x^{n-j-1}=\frac{n}{x}\left(\text{ }_{H}w_{n}^{\left(x\right)}-H_{n}x^{n}\right). (11) The Bernoulli polynomials B_{n}\left(x\right) are defined by \sum_{j=0}^{n}\dbinom{n}{j}B_{j}x^{n-j}=B_{n}\left(x\right). Thus, we have \sum_{j=1}^{n-1}\left(\dbinom{n}{j}-1\right)B_{j}x^{n-j-1}=\frac{1}{x}B_{n}\left(x\right)-\frac{1}{x}\sum_{j=1}^{n-1}B_{j}x^{n-j}. (12) Now, using the obvious relation \frac{d}{dx}\dbinom{x}{k}=\dbinom{x}{k}\sum_{j=0}^{k-1}\frac{1}{x-j} in ( 2 ), we find that \frac{d}{dx}\text{ }_{H}w_{n}^{\left(x\right)}=\sum_{k=1}^{n}\genfrac{}{}{0.0pt}{0}{n}{k}\dbinom{x}{k}k!H_{k}\sum_{j=0}^{k-1}\frac{1}{x-j}. (13) Inserting ( 11 ), ( 12 ), and ( 13 ) in ( 10 ) gives \displaystyle\sum_{j=0}^{n}B_{j}x^{n-j} \displaystyle=x\sum_{k=1}^{n}\genfrac{}{}{0.0pt}{0}{n}{k}\dbinom{x}{k}k!H_{k}\sum_{j=0}^{k-1}\frac{1}{x-j}-nH_{n}x^{n} \displaystyle-n\sum_{j=1}^{n-1}\left(\dbinom{n}{j}-1\right)\frac{B_{j}}{j}x^{n-j}+B_{n}\left(x\right). Setting x=-1 and using the relation \left(-1\right)^{n}B_{n}\left(-x\right)=B_{n}\left(x\right)+nx^{n-1} in the above equation yields to \sum_{j=0}^{n}B_{j}=\sum_{k=1}^{n}\left(-1\right)^{n-k}\genfrac{}{}{0.0pt}{0}{n}{k}k!\left(H_{k}\right)^{2}+B_{n}\left(1\right)+n-n^{2}-1. (14) To evaluate the first sum on the right-hand side of ( 14 ), we use ( 5 ) and find that \sum_{n=1}^{\infty}\frac{t^{n}}{n!}\sum_{k=1}^{n}\left(-1\right)^{n-k}\genfrac{}{}{0.0pt}{0}{n}{k}k!\left(H_{k}\right)^{2}=\sum_{k=1}^{\infty}\left(H_{k}\right)^{2}\left(1-e^{-t}\right)^{k}. (15) With the use of the generating function of the harmonic numbers \sum_{k=1}^{\infty}H_{k}t^{k}=-\frac{\ln\left(1-x\right)}{1-x}, one can obtain that \sum_{k=1}^{\infty}\left(H_{k}\right)^{2}t^{k}=\frac{\mathrm{Li}_{2}\left(x\right)}{1-x}+\frac{\ln^{2}\left(1-x\right)}{1-x}. Using this relation with ( 9 ) in ( 15 ), we arrive at \sum_{k=1}^{n}\left(-1\right)^{n-k}\genfrac{}{}{0.0pt}{0}{n}{k}k!\left(H_{k}\right)^{2}=\mathbb{B}_{n}^{\left(2\right)}\left(1\right)-\mathbb{B}_{n}^{\left(2\right)}+n\left(n-1\right). Combining these results in ( 14 ) gives the desired formula. We conclude this section by presenting some congruences for cumulative sums of the Bernoulli and Euler numbers. Here we call E_{n}\left(0\right)=E_{n} is the n th Euler number. Corollary 4 For an odd prime p , we have \displaystyle\sum_{j=0}^{p}pB_{j} \displaystyle\equiv-1\text{ }\left(\operatorname{mod}p\right), (16) \displaystyle\sum_{j=0}^{p-3}B_{j} \displaystyle\equiv-1\text{ }\left(\operatorname{mod}p\right),\text{ \ \ }\left(p\geq 5\right), (17) \displaystyle\sum_{j=0}^{p}E_{j} \displaystyle\equiv\frac{3}{2}\text{ }\left(\operatorname{mod}p\right), (18) \displaystyle p\sum_{j=0}^{p}\frac{B_{j}}{\left(p-j+1\right)} \displaystyle\equiv-1\text{ }\left(\operatorname{mod}p\right). (19) Proof. Let n=p be an odd prime in ( 14 ). Since p\geq 3 , we have B_{p}\left(1\right)=B_{p}=0. Thus \displaystyle\sum_{j=0}^{p}B_{j} \displaystyle=\sum_{k=1}^{p}\left(-1\right)^{p-k}\genfrac{}{}{0.0pt}{0}{p}{k}k!\left(H_{k}\right)^{2}+B_{p}\left(1\right)+p-p^{2}-1 \displaystyle=p-p^{2}-1+\genfrac{}{}{0.0pt}{0}{p}{1}\left(H_{1}\right)^{2}+\genfrac{}{}{0.0pt}{0}{p}{p}p!\left(H_{p}\right)^{2}+\sum_{k=2}^{p-1}\left(-1\right)^{p-k}\genfrac{}{}{0.0pt}{0}{p}{k}k!\left(H_{k}\right)^{2}. Now, \genfrac{}{}{0.0pt}{0}{p}{1}=\genfrac{}{}{0.0pt}{0}{p}{p}=1 , H_{1}=1 , and p!\left(H_{p}\right)^{2}=p!\left(H_{p-1}\right)^{2}+2\left(p-1\right)!H_{p-1}+\frac{\left(p-1\right)!}{p}, so \sum_{j=0}^{p}B_{j}=p-p^{2}+p!\left(H_{p-1}\right)^{2}+2\left(p-1\right)!H_{p-1}+\frac{\left(p-1\right)!}{p}+\sum_{k=2}^{p-1}\left(-1\right)^{p-k}\genfrac{}{}{0.0pt}{0}{p}{k}k!\left(H_{k}\right)^{2}, or equivalently \displaystyle\sum_{j=0}^{p}pB_{j} \displaystyle=p^{2}-p^{3}+p\left(p!\right)\left(H_{p-1}\right)^{2}+2p!H_{p-1}+\left(p-1\right)! \displaystyle+p\sum_{k=2}^{p-1}\left(-1\right)^{p-k}\genfrac{}{}{0.0pt}{0}{p}{k}k!\left(H_{k}\right)^{2}. (20) Since \genfrac{}{}{0.0pt}{0}{p}{k}\equiv 0 \left(\operatorname{mod}p\right) for k=2,3,\cdots,p-1 , we find that \sum_{j=0}^{p}pB_{j}\equiv\left(p-1\right)!\equiv-1\text{ }\left(\operatorname{mod}p\right), which is the congruence given in ( 16 ). Now, Babbage’s theorem states that H_{p-1}\equiv 0 \left(\operatorname{mod}p\right) , so ( 20 ) can be written as \sum_{j=0}^{p}pB_{j}\equiv\left(p-1\right)!\text{ }\left(\operatorname{mod}p^{2}\right). A famous congruence by Glaisher [ 7 ] is \left(p-1\right)!\equiv-p+pB_{p-1}\text{ }\left(\operatorname{mod}p^{2}\right). Thus \displaystyle-p+pB_{p-1} \displaystyle\equiv\sum_{j=0}^{p}pB_{j}=pB_{0}+pB_{1}+pB_{p}+pB_{p-1}+pB_{p-2}+\sum_{j=2}^{p-3}pB_{j} \displaystyle=p-\frac{p}{2}+pB_{p-1}+\sum_{j=2}^{p-3}pB_{j}\left(\operatorname{mod}p^{2}\right) which implies that \sum_{j=2}^{p-3}pB_{j}\equiv-\frac{3p}{2}\text{ }\left(\operatorname{mod}p^{2}\right) for p\geq 3 , or equivalently \sum_{j=2}^{p-3}B_{j}\equiv-\frac{3}{2}\text{ }\left(\operatorname{mod}p\right) if p\geq 5. ( 17 ) now follows by using the values B_{0} and B_{1} . Inserting ( 8 ) into ( 7 ) yields to \sum_{j=1}^{n}\left(\binom{n}{j}+1\right)\frac{E_{j-1}}{2}=1, or \sum_{j=0}^{n-1}E_{j}=2-nE_{n}-E_{n-1}-\sum_{j=2}^{n-1}\binom{n}{j}E_{j-1}. Let n=p be an odd prime. Then, we have \sum_{j=0}^{p}E_{j}=2+E_{p}-p-\sum_{j=2}^{p-1}\binom{p}{j}E_{j-1}, since E_{0}=1 and E_{2m}=0 for m\geq 1. Moreover, 2^{m}E_{m}\in\mathbb{Z}, which means that E_{m} does not contain primes p in the denominator when p>2. Thus \binom{p}{j}E_{j-1}\equiv 0\text{ }\left(\operatorname{mod}p\right) for j=1,2,\cdots,p-1. Therefore, we have \sum_{j=0}^{p}E_{j}\equiv 2-\frac{1}{2}=\frac{3}{2}\text{ }\left(\operatorname{mod}p\right) (21) for p\geq 5. Note that \sum_{j=0}^{3}E_{j}=E_{0}+E_{1}+E_{2}+E_{3}=1-\frac{1}{2}+\frac{1}{4}=\frac{3}{4}, hence, ( 21 ) is also true when p=3 , that is, we have ( 18 ). To prove ( 19 ), we first integrate both sides of ( 1 ) with respect to x from 0 to 1 and use ( 3 ) and the well-known formula \sum_{j=0}^{n}\dbinom{n}{j}\frac{B_{j}}{n-j+1}=\int\limits_{0}^{1}B_{n}\left(x\right)dx=0,\text{ }n\geq 2. We then reach at \sum_{j=0}^{n}\frac{B_{j}}{\left(n-j+1\right)}=1-\left(n+1\right)\int\limits_{0}^{1}\text{ }_{H}w_{n}^{\left(x\right)}dx. From the definition of the Cauchy numbers of the first kind c_{n} (cf. [ 11 ] ) \frac{c_{n}}{n!}=\int\limits_{0}^{1}\dbinom{x}{n}dx we arrive at \sum_{j=0}^{n}\frac{B_{j}}{\left(n-j+1\right)}=1-\left(n+1\right)\sum_{k=1}^{n}\genfrac{}{}{0.0pt}{0}{n}{k}c_{k}H_{k}. We write n=p, where p is an odd prime in the equation above to obtain \sum_{j=0}^{p}\frac{B_{j}}{\left(p-j+1\right)}=1-\left(p+1\right)\sum_{k=1}^{p}\genfrac{}{}{0.0pt}{0}{p}{k}c_{k}H_{k}. (22) Employing \genfrac{}{}{0.0pt}{0}{p}{1}=\genfrac{}{}{0.0pt}{0}{p}{p}=1 and \genfrac{}{}{0.0pt}{0}{p}{p-1}=\dbinom{p}{2} and the explicit expression of Cauchy numbers of the first kind ( [ 11 ] ) c_{k}=\sum_{j=1}^{k}\genfrac{[}{]}{0.0pt}{0}{k}{j}\frac{\left(-1\right)^{k-j}}{j+1}, (23) we have \displaystyle\sum_{k=1}^{p}\genfrac{}{}{0.0pt}{0}{p}{k}c_{k}H_{k} \displaystyle=\frac{1}{2}+c_{p}H_{p-1}+\frac{c_{p}}{p}+\frac{p\left(p-1\right)}{2}c_{p-1}H_{p-1} \displaystyle+\sum_{k=2}^{p-2}\sum_{j=1}^{k}\genfrac{}{}{0.0pt}{0}{p}{k}\genfrac{[}{]}{0.0pt}{0}{k}{j}\frac{\left(-1\right)^{k-j}}{j+1}H_{k}, or equivalently \displaystyle p\sum_{k=1}^{p}\genfrac{}{}{0.0pt}{0}{p}{k}c_{k}H_{k} \displaystyle=\frac{p}{2}+pc_{p}H_{p-1}+c_{p}+\frac{p\left(p-1\right)}{2}pc_{p-1}H_{p-1} \displaystyle+p\sum_{k=2}^{p-2}\sum_{j=1}^{k}\genfrac{}{}{0.0pt}{0}{p}{k}\genfrac{[}{]}{0.0pt}{0}{k}{j}\frac{\left(-1\right)^{k-j}}{j+1}H_{k.} (24) Let A\left(p\right)=\sum_{k=2}^{p-2}\sum_{j=1}^{k}\genfrac{}{}{0.0pt}{0}{p}{k}\genfrac{[}{]}{0.0pt}{0}{k}{j}\frac{\left(-1\right)^{k-j}}{j+1}H_{k.}. The product \frac{1}{j+1}H_{k} does not involve p in the denominator, so A\left(p\right)\equiv 0\left(\operatorname{mod}p\right) since \genfrac{}{}{0.0pt}{0}{p}{k}\equiv 0\left(\operatorname{mod}p\right) for k=2,3,\cdots,p-2. Therefore, from ( 22 ) and ( 24 ), we conclude that \displaystyle p\sum_{j=0}^{p}\frac{B_{j}}{\left(p-j+1\right)} \displaystyle=p-\frac{p^{2}}{2}-\frac{p}{2}-p^{2}c_{p}H_{p-1}-pc_{p}H_{p-1}-pc_{p}-c_{p} \displaystyle-\frac{p^{2}\left(p-1\right)}{2}pc_{p-1}H_{p-1}-\frac{p\left(p-1\right)}{2}pc_{p-1}H_{p-1}-p^{2}A\left(p\right)-pA\left(p\right). (25) Now, in Corollary 5.2 of [ 6 ] , it has been recorded that if \lambda\equiv 0 \left(\operatorname{mod}p\right), then c_{p}\left(\lambda\right)\equiv 1-\lambda+\left(1-\lambda\right)\left(-\lambda\right)\cdots\left(2-\lambda-p\right)\text{ }\left(\operatorname{mod}p\right) and pc_{p-1}\left(\lambda\right)\equiv 1\text{ }\left(\operatorname{mod}p\right), where c_{k}\left(\lambda\right) is the degenerate Cauchy number. Since \underset{\lambda\rightarrow 0}{\lim}c_{k}\left(\lambda\right)=c_{k} , we conclude that c_{p}\equiv 1\text{ }\left(\operatorname{mod}p\right)\text{ and }pc_{p-1}=\underset{\lambda\rightarrow 0}{\lim}pc_{p-1}\left(\lambda\right)\equiv 1\text{ }\left(\operatorname{mod}p\right) Therefore, ( 25 ) implies the congruence ( 19 ). Remark 5 ( 16 ) also follows from the von Staudt-Clausen theorem for the Bernoulli numbers: pB_{2j}\equiv\left\{\begin{array}[c]{rl}0\left(\operatorname{mod}p\right),&\text{if }\left(p-1\right)\nmid 2j,\\ -1\left(\operatorname{mod}p\right),&\text{if }\left(p-1\right)|2j.\end{array}\right. Indeed, we first write \displaystyle\sum_{j=0}^{p}pB_{j} \displaystyle=pB_{0}+pB_{1}+pB_{p}+pB_{p-1}+pB_{p-2}+\sum_{j=1}^{\frac{p-3}{2}}pB_{2j} \displaystyle=p-\frac{p}{2}+pB_{p-1}+\sum_{j=1}^{\frac{p-3}{2}}pB_{2j}. As j ranges from 1 to \frac{p-3}{2}, \left(p-1\right) does not divide 2j. So pB_{2j}\equiv 0\left(\operatorname{mod}p\right) for j=1,2,\cdots,\frac{p-3}{2}. Moreover, pB_{p-1}\equiv-1\left(\operatorname{mod}p\right) and hence we have \sum_{j=0}^{p}pB_{j}\equiv-1\text{ }\left(\operatorname{mod}p\right) for an odd prime p. We also note that, from ( 25 ), the congruence ( 19 ) is equivalent to p\sum_{j=0}^{p}\frac{B_{j}}{\left(p-j+1\right)}=-\frac{p}{2}-c_{p}\text{ }\left(\operatorname{mod}p^{2}\right). 3 Proof of the Main Theorem Let \sum_{k=j}^{n}\left(-1\right)^{k-j}\genfrac{}{}{0.0pt}{0}{n}{k}\genfrac{[}{]}{0.0pt}{0}{k}{j}H_{k}=\mathcal{B}_{n,j}. Using the well known recurrence relation for the Stirling numbers of the second kind [ 14 , Eq. (9.1)] \genfrac{}{}{0.0pt}{0}{n+1}{k}=\genfrac{}{}{0.0pt}{0}{n}{k}k+\genfrac{}{}{0.0pt}{0}{n}{k-1}, we obtain that \displaystyle\mathcal{B}_{n+1,j} \displaystyle=\sum_{k=j}^{n+1}\left(-1\right)^{k-j}\genfrac{}{}{0.0pt}{0}{n}{k}\genfrac{[}{]}{0.0pt}{0}{k}{j}kH_{k} \displaystyle+\sum_{k=j}^{n+1}\left(-1\right)^{k-j}\genfrac{}{}{0.0pt}{0}{n}{k-1}\genfrac{[}{]}{0.0pt}{0}{k}{j}H_{k}. (1) Applying the well-known recurrence relation of the Stirling numbers of the first kind [ 14 , Eq. (12.3)] \genfrac{[}{]}{0.0pt}{0}{k+1}{j}=\genfrac{[}{]}{0.0pt}{0}{k}{j}k+\genfrac{[}{]}{0.0pt}{0}{k}{j-1}, and the fact that \genfrac{}{}{0.0pt}{0}{n}{k}=0 for k>n, ( 1 ) becomes \displaystyle\mathcal{B}_{n+1,j} \displaystyle=\sum_{k=j}^{n}\left(-1\right)^{k-j}\genfrac{}{}{0.0pt}{0}{n}{k}\genfrac{[}{]}{0.0pt}{0}{k+1}{j}H_{k}-\sum_{k=j}^{n}\left(-1\right)^{k-j}\genfrac{}{}{0.0pt}{0}{n}{k}\genfrac{[}{]}{0.0pt}{0}{k}{j-1}H_{k} \displaystyle-\sum_{k=j-1}^{n}\left(-1\right)^{k-j}\genfrac{}{}{0.0pt}{0}{n}{k}\genfrac{[}{]}{0.0pt}{0}{k+1}{j}H_{k}-\sum_{k=j-1}^{n}\left(-1\right)^{k-j}\genfrac{}{}{0.0pt}{0}{n}{k}\genfrac{[}{]}{0.0pt}{0}{k+1}{j}\frac{1}{k+1}. Employing the identity ( 2 ) in the form \sum_{k=j-1}^{n-1}\frac{\left(-1\right)^{k+1-j}}{k+1}\genfrac{}{}{0.0pt}{0}{n-1}{k}\genfrac{[}{]}{0.0pt}{0}{k+1}{j}=\dbinom{n-1}{j}\frac{B_{n-j}}{n-j}\,\ \left(1\leq j\leq n\right), we find that \mathcal{B}_{n+1,j}=\mathcal{B}_{n,j-1}+\dbinom{n}{j}\frac{B_{n+1-j}}{n+1-j}. From this reduction formula, we see that \mathcal{B}_{n+1,j}=\mathcal{B}_{n-1,j-2}+\left(\dbinom{n}{j}+\dbinom{n-1}{j-1}\right)\frac{B_{n+1-j}}{n+1-j}. Consecutive j-2 times applications yield to \mathcal{B}_{n+1,j}=\mathcal{B}_{n-j,0}+\frac{B_{n+1-j}}{n+1-j}\sum_{k=0}^{j-1}\dbinom{n-k}{j-k}. (2) Here, we have \sum_{k=0}^{j-1}\dbinom{n-k}{j-k}=\dbinom{n+1}{j}-1, (3) upon the use of [ 8 , p. 7, Eq. (1.49)] \sum_{k=0}^{n}\dbinom{x+k}{k}=\dbinom{x+n+1}{n}. Substituting ( 3 ) into ( 2 ), with the use of \mathcal{B}_{n-j,0}=0 , we arrive at \mathcal{B}_{n+1,j}=\left(\dbinom{n+1}{j}-1\right)\frac{B_{n+1-j}}{n+1-j}, which is the desired result.