The paper will appear in Discrete Mathematics.

CONGRUENCES INVOLVING BERNOULLI POLYNOMIALS

Zhi-Hong Sun Department of Mathematics, Huaiyin Teachers College, Huaian, Jiangsu 223001, P.R. China E-mail: hyzhsun@public.hy.js.cn Homepage: http://www.hytc.edu.cn/xsjl/szh
Let {Bn (x)} be the Bernoulli polynomials. In the paper we establish some congruences for Bj (x) (mod pn ), where p is an odd prime and x is a rational p-integer. Such congruences are concerned with the properties of p-regular functions, the congru2 p ences for h(sp) (mod p) (s = 3, 5, 8, 12) and the sum , where h(d) is k the class number of the quadratic eld ( d) of discriminant d and p-regular functions are those functions f such that f (k) (k = 0, 1, . . . ) are rational p-integers and 2n n k n k=0 k (1) f (k ) 0 (mod p ) for n = 1, 2, 3, . . . We also establish many congruences for Euler numbers. MSC: Primary 11B68, Secondary 11A07, 11R29. Keywords: Congruence, Bernoulli polynomial, p-regular function, class number, Euler number

)>IJH=?J

kr (mod m)

1. Introduction. The Bernoulli numbers {Bn } and Bernoulli polynomials {Bn (x)} are dened by
n1

B0 = 1,
k=0

n Bk = 0 (n 2) k

and Bn (x) =
k=0

n Bk xnk (n 0). k

The Euler numbers {En } and Euler polynomials {En (x)} are dened by 2et tn = E (|t| < ) n 2 t e + 1 n=0 n! 2 which are equivalent to (see [MOS])
n

and

2ext tn = E ( x ) (|t| < ), n et + 1 n=0 n!

E0 = 1, E2n1 = 0,
r =0

2n E2r = 0 (n 1) 2r

and
n

(1.1)

En (x) +
r =0

n Er (x) = 2xn (n 0). r

1

It is well known that([MOS]) En (x) = (1.2) 1 2n

n r =0

n (2x 1)nr Er r .

2 x = Bn+1 (x) 2n+1 Bn+1 n+1 2

Let Z and N be the set of integers and the set of positive integers respectively. Let [x] be the integral part of x and {x} be the fractional part of x. If m, s N and p is an odd prime not dividing m, in Section 2 we show that (1)s m p
p1

k=1 ksp(mod m) (s1)p m

(s1)p [ m ]

p k Bp1
sp m (s1)p m

Bp1
1 2

(mod p) (1)
[ sp m]

if 2 | m, Ep2
sp m

(1)

Ep2

(mod p)

if 2 m.

For a discriminant d let h(d) be the class number of the quadratic eld Q( d) (Q is the set of rational numbers). If p > 3 is a prime of the form 4m + 3, it is well known that (cf. [IR]) (1.3) h(p) 2B p+1 (mod p).
2

If p is a prime of the form 4m + 1, according to [Er] we have (1.4) 2h(4p) E p1 (mod p).
2

a Let ( n ) be the Kronecker symbol. For odd primes p, in Section 3 we establish the following congruences:

h(8p) E p1
2

p B p+1 2 3 p h(12p) 8 B p+1 2 3 1 h(5p) 8B p+1 2 5 h(3p) 4

1 4

(mod p); 1 (mod p) for p 1 (mod 4); 3 1 (mod p) for p 7, 11, 23 (mod 24); 12 (mod p) for p 11, 19 (mod 20).

For m N let Zm be the set of rational numbers whose denominator is coprime to m. For a prime p, in [S5] the author introduced the notion of p-regular functions. n k n If f (k ) Zp for any nonnegative integers k and k=0 n k (1) f (k ) 0 (mod p )
2

for all n N, then f is called a p-regular function. If f is a p-regular function and k, m, n, t N, in Section 4 we show that
n1

(1.5)

f (ktp

m 1

)
r =0

(1)n1r

k1r n1r

k f (rtpm1 ) (mod pmn ), r

which was annouced by the author in [S5, (2.4)]. We also show that (1.6) f (kpm1 ) (1 kpm1 )f (0) + kpm1 f (1) (mod pm+1 ) for p > 2.

Let p be a prime, x Zp and let b be a nonnegative integer. Let t p be the least nonnegative residue of t modulo p and x = (x + x p )/p. From [S4, Theorem 3.1] we know that f (k ) = p(pBk(p1)+b (x) pk(p1)+b Bk(p1)+b (x )) is a pregular function. If p 1 b, in [S5] the author showed that f (k ) = (Bk(p1)+b (x) pk(p1)+b1 Bk(p1)+b (x ))/(k (p 1) + b) is also a pregular function. Using such results in [S4, S5] and (1.5), in Section 5 we obtain general congruences for pBk(ps )+b (x), pBk(ps )+b, (mod psn ), where k, n, s N, is Eulers totient function and is a Dirichlet character modulo a positive integer. As a consequence of (1.6), if 2 | b and p 1 b, we have Bk(ps )+b Bb Bp1+b (1 kps1 )(1 pb1 ) + kps1 (mod ps+1 ). s k(p ) + b b p1+b
k In Section 6 we establish some congruences for k=0 n k (1) pBk(p1)+b (x) modulo pn+1 , where p is an odd prime, n N, x Zp and b is a nonnegative integer. Let p be an odd prime and b {0, 2, 4, . . . }. In Section 7 we show that f (k ) = p1 (1 (1) 2 pk(p1)+b )Ek(p1)+b is a pregular function. Using this and (1.5) we give congruences for Ek(pm )+b (mod pmn ), where k, m N. By (1.6) we have n

Ek(pm )+b (1 kpm1 )(1 (1)

p1 2

pb )Eb + kpm1 Ep1+b (mod pm+1 ).

We also show that f (k ) = E2k+b is a 2regular function and

n1

2m kt+b

r =0

(1)n1r

k1r n1r

k E2m rt+b (mod 2mn+n ), r

where k, m, n, t N and N is given by 21 n < 2 .

1)p sp 2. Congruences for Bk ({ (s m }) Bk ({ m }) (mod p). We begin with two useful identities concerning Bernoulli and Euler polynomials. In the case m = 1 the result is well known. See [MOS].

Theorem 2.1. Let p, m N and k, r Z with k 0. Then

p1

xk =
x=0 xr (mod m)

mk p Bk+1 + k+1 m

rp m

Bk+1

r m

and
p1

(1)
x=0 xr (mod m)

xr m

r p mk p x = (1)[ m ] Ek + 2 m

rp m

(1)[ m ] Ek

r m

Proof. For any real number t and nonnegative integer n it is well known that (cf. [MOS]) (2.1) Bn (t + 1) Bn (t) = ntn1 (n = 0) and En (t + 1) + En (t) = 2tn .

Hence, for x Z we have Bk+1 = Thus Bk+1

p1

x+1 rx1 + m m r x x+1 Bk+1 m + m Bk+1

x+1 m

Bk+1
1 m x m

x rx + m m x r x Bk+1 m + m = (k + 1) r m Bk+1
x k m

=0

if m x r, if m | x r.

m1 m

Bk+1

p + m Bk+1

rp m

Bk+1

=
x=0

x+1 + m xk .

rx1 m

x + m

rx m

k+1 = mk

p1 x=0 xr (mod m)

Similarly, if x Z, by (2.1) we have

r x1 r x x+1 rx1 x rx (1)[ m ] Ek + (1)[ m ] Ek + m m m m x x+1 r x 1 x r x ] [ rm ( 1) E + } E + = 0 k m k m m m m if m x r, = r x r x r x 1 +1 x x k (1) m 1 Ek xm + mm (1) m Ek m = (1) m 2( m ) if m | x r.

Thus (1)[ =

r p m ]

Ek

p + m
r x1 ] m

rp m Ek

(1)[ m ] Ek rx1 m

r m

p1

(1)[
x=0

x+1 + m x + m
xr m

(1)[ 2 = k m This completes the proof.

r x m ]

Ek

rx m

p1

(1)
x=0 xr (mod m)

xk .

Corollary 2.1. Let p be an odd prime and k {0, 1, . . . , p 2}. Let r Z and m N with p m. Then
p1

xk
x=0 xr (mod m)

mk Bk+1 k+1

rp m

Bk+1

r m

(mod p)

and
p1

(1)
x=0 xr (mod m)

xr m

xk rp m r m

r p mk (1)[ m ] Ek 2

(1)[ m ] Ek

(mod p).

Proof. If x1 , x2 Zp and x1 x2 (mod p), by [S5, Lemma 3.1] and [S3, Lemma 3.3] we have (2.2) and (2.3) Ek (x1 ) Ek (x2 ) (mod p). Bk+1 (x1 ) Bk+1 (x2 ) x1 x2 pBk 0 (mod p) k+1 p

Thus the result follows from Theorem 2.1. Remark 2.1 Putting k = p 2 in Corollary 2.1 and then applying Fermats little theorem we see that if p is an odd prime not dividing m, then
p1

(2.4)
x=1 xr (mod m)

1 1 Bp1 x m

rp m
5

Bp1

r m

(mod p)

and
p1

(1) (2.5)
x=1 xr (mod m)

xr m

1 x rp m (1)[ m ] Ep2
r

r p 1 (1)[ m ] Ep2 2m

r m

(mod p).

Here (2.4) and (2.5) are due to my brother Z.W. Sun. See [Su2, Theorem 2.1]. Inspired by his work, the author established Theorem 2.1 and Corollary 2.1 in 1991. Corollary 2.2. Let p be an odd prime. Let k {0, 1, . . . , p 2} and m, s N with p m. Then (1)k Bk+1 k+1 and (1)[
(s1)p ] m

(s 1)p m

Bk+1

sp m

(s1)p <r sp m m

rk (mod p)

Ek

(s 1)p m
(s1)p <r sp m m

(1)[ m ] Ek (1)r rk (mod p).

sp

sp m

2(1)k1

Proof. It is clear that (see [S3, Lemma 3.1, Corollaries 3.1 and 3.3])
p1

xk = (2.6)
x=0 xsp(mod m) r Z 0sprm<p

(sp rm)k =
(s1)p <r sp m m

(sp rm)k

(m)k
(s1)p <r sp m m

rk (mod p)

and
p1

(1)
x=0 xsp(mod m)

xsp m

xk =
r Z 0sprm<p

(1)r (sp rm)k (1)r (sp rm)k

(s1)p <r sp m m

(2.7)

= (m)k

(1)r rk (mod p)
(s1)p <r sp m m

Thus applying Corollary 2.1 we obtain the result. Remark 2.2 In the case s = 1, the rst part of Corollary 2.2 is due to Lehmer ([L, p. 351]). In the case k = p 2, the rst part of Corollary 2.2 can be deduced from [GS, p. 126].
6

Corollary 2.3. Let p be a prime.

(p3)/4

(i) (Karpinski[K, UW]) If p 3 (mod 8), then

x=1 [p/6]

x p

= 0.

(ii) (Karpinski[K, UW]) If p 5 (mod 8), then (iii) (Berndt[B, UW]) If p 5 (mod 24), then
x=1

x=1 (p5)/12

x p

= 0.
x p

= 0.

Proof. By Corollary 2.2 and the known fact B2n+1 = 0, for m N with p m we have
[p/m]

(2.8)

x=1

x p

[p/m]

x
x=1

p1 2

(1)
p m

p1 2

p+1 2

B p+1 B p+1
2 2

p m if p 1 (mod 4),

2B p+1
2 2

(mod p)
2

2B p+1 + 2B p+1

p m

(mod p) if p 3 (mod 4).

3 2n1 It is well known that B2n ( 4 ) = B2n ( 1 )B2n /24n1 . Thus, if p 4 ) = (1 2 3 (mod 8), by (2.8) we see that
p3 4

x=1

p1 x 3 1 2B p+1 + 2B p+1 = p1 1 2 2 B p+1 2B p+1 2 2 2 2 p 4 2

1
p3

2 2 B p+1 = 0 (mod p). 2 p

(p3)/4

3 p3 x 4 (x As p x=1 p 4 , we must have x=1 4 p ) = 0. This proves (i). Now we consider (ii). For n {0, 1, 2, . . . } and m N it is well known that (cf. [IR], [MOS]) m1

(2.9) Thus

Bn (1 x) = (1)n Bn (x) and

k=0

Bn x +

k = m1n Bn (mx). m

B p+1
2

p1 1 1 1 1 + B p+1 + = 2 2 B p+1 2 2 2n 2n 2 n

and so (2.10) B p+1

2 p+1 1 2 1 n1 B p+1 (1) 2 B p+1 2 2 2n p n 2n

(mod p).

Since p 5 (mod 8), taking n = 3 in (2.10) we nd (2.11) B p+1

2

1 1 1 B p+1 + B p+1 = 0 (mod p). 2 2 6 3 3

7

This together with (2.8) and (2.9) yields

[p/6]

x=1 [p/6] x=1

x 2B p+1 2 p

p 6
[p/6]

= 2

p 1 B p+1 0 (mod p). 2 3 6

p x x As | x=1 p = 0. This proves (ii). p | [ 6 ] we have Finally we consider (iii). Assume p 5 (mod 24). By (2.10) and (2.11) we have

B p+1
2

1 2 1 5 5 B p+1 + B p+1 B p+1 2 2 2 12 p 6 12 12

(mod p).

On the other hand, by (2.9) we have B p+1

2 p1 1 5 1 9 + B p+1 = 3 2 B p+1 B p+1 2 2 2 12 12 4 12 p+1 3 1 1 B p+1 (1) 2 B p+1 2 2 p 4 4 = 0 (mod p).

Thus B p+1
2

1 12

B p+1
2

5 12

0 (mod p). Now applying (2.8) we see that p 12 = 2B p+1

2

[p/12]

x=1

x 2B p+1 2 p

5 0 (mod p). 12

This yields (iii) and so the corollary is proved. Corollary 2.4. Suppose p, q, m N, n Z, gcd(p, m) = 1 and q m. For r Z let Ar (m, p) be the least positive solution of the congruence px r (mod m). Then n pq + n r : Ar (m, p) q, r Z, n r p 1 n = . m m Proof. Using Theorem 2.1 we see that r : Ar (m, p) q, r Z, n r p 1 n
q p1n q p1

=
x=1 q r =n r px (mod m)

1=
x=1 s=0 spx+n (mod m)

=
x=1 q

B1

p + m

px + n p m =

B1 px + n m

px + n m

=
x=1

p + m

p(x 1) + n m

pq n pq + n + m m m n pq + n = . m m This proves the corollary. =

pq + n m

pq + n m

n m

n m

Theorem 2.2. Let m, s N and let p be an odd prime not dividing m. Then (1)
sm p1

k=1 ksp(mod m)

p k

(s1)p <k< sp m m

(1)km k Bp1 Ep2

sp m (s1)p m

Bp1
1 2

(s1)p m
(s1)p ] m

(mod p) (1)[ m ] Ep2

sp

if 2 | m,
sp m

(1)[

(mod p)

if 2 m.

Proof. Let r Z. Since that 1 p

p1

p1 j p1

(1)j (mod p) for j {0, 1, . . . , p 1} we see 1 p1 k k1

p1

k=1 kr (mod m)

p k

=
k=1 kr (mod m)

k=1 kr (mod m) 1 k

(1)k1 k if 2 | m,

p1 r 1 ( 1) k=1 kr (mod (1)r1

p1

(mod p)
kr m

m)

(1)
k=1 kr (mod m)

1 k

(mod p) if 2 m.

Putting this together with (2.4) and (2.5) we see that 1 p

p1

k=1 kr (mod m) (1)r m (1)r 2m

p k
r p m
r p m ]

Bp1 (1)[

Bp1
r p m

r m

(mod p)
r

if 2 | m,
r m

Ep2

(1)[ m ] Ep2

(mod p)

if 2 m.

Taking r = sp we obtain (1)

sm p1

k=1 ksp(mod m) (s1)p m

(s1)p [ m ]

p k Bp1
sp m (s1)p m

Bp1
1 2

(mod p) (1) sp m
9
[ sp m]

if 2 | m, Ep2
sp m

(1)

Ep2

(mod p) 1 (mod p) r

if 2 m.

On the other hand, putting k = p 2 in Corollary 2.2 we see that Bp1 (s 1)p m Bp1
(s1)p <r< sp m m

and (1)[
(s1)p ] m

Ep2

2
(s1)p <r< sp m m

sp (s 1)p (1)[ m ] Ep2 m (1)r (mod p). r

sp m

Now combining the above we prove the theorem. Corollary 2.5. Let m, n N and let p be an odd prime not dividing m. (i) If 2 | m, then Bp1 np m Bp1 m p
n p1

(1)
s=1

s 1 k=1 ksp(mod m)

p k

(mod p).

(ii) If 2 m, then (1)

[ np m ]

Ep2

np m

2p 2 2m + p p

p1

(1)
s=1

s1 k=1 ksp(mod m)

p k

(mod p).

Proof. It is well known that pBp1 p 1 (mod p). Thus, by (1.2) we have n Ep2 (0) = 2(1 2p1 )Bp1 /(p 1) (2p 2)/p (mod p). Note that s=1 (f (s) f (s 1)) = f (n) f (0). Then the result follows from Theorem 2.2 and the above immediately. p Combining Theorem 2.2, Corollary 2.5 with the formulae for k in the
kr (mod m)

cases m = 3, 4, 5, 6, 8, 9, 10, 12 (see [S1,S2,S3,SS,Su1]) we may deduce many useful results, which had been given in [GS] and [S3]. 3. Some congruences for h(3p), h(5p), h(8p), h(12p) (mod p). Let {Sn } be dened by
n1

(3.1)

S0 = 1

and

Sn = 1
k=0

n 2n2k1 2 Sk k

(n 1).

Then clearly Sn Z. The rst few Sn are shown below: S1 = 1, S2 = 3, S3 = 11, S4 = 57, S5 = 361, S6 = 2763. Theorem 3.1. Let p be an odd prime. Then h(8p) E p1
2

1 S p1 (mod p). 2 4
10

Proof. From [UW, p. 58] we know that

p1

(3.2)

h(8p) = 2
a=1 a1(mod 4)

8p . a
p1 2
p1 2

Thus applying Corollary 2.1 in the case r = 1, m = 4 and k =

p1

we see that

h(8p) = 2
a=0 a1(mod 4)
p1 2

2 a
1p 4 ]

a 2 p E p1
2

p1

(1)
a=0 a1(mod 4)

a1 4

a 1 4

4 Since E2n (0) = E p1

2

(1)[

1p 4

E p1
2

(mod p).

2 2n+1 (B2n+1

22n+1 B2n+1 ) = 0 by (1.2), we see that E2n (0) = 0 E2n1 ( 1 2)

p1 2

1p 4

if p = 4n + 1,
12n

=2

E2n1 = 0

if p = 4n 1.

Thus h(8p) 4 E p1
2

1 1 E p1 2 4 4

(mod p).

Let Sn = 4n En ( 1 4 ). Now we show that Sn = Sn for n 0. By (1.1) we have

n

4n Sn +
k=0

n k 4 Sk = 2 4n k
n1 n k=0 k

and so

Sn +
k=0

n nk 4 Sk = 2. k

That is, Sn = 1 That is, (3.3) Hence S p1 = 4

2 p1 2

22n2k1 Sk . Since S0 = S0 = 1 we see that Sn = Sn . Sn = 4n En 1 . 4

E p1 ( 1 4 ) h(8p) (mod p). This proves the theorem.

2 2

Corollary 3.1. Let p be an odd prime. Then p S p1 . Proof. From (3.2) we have 1 < h(8p) < p. Thus the result follows from Theorem 3.1. Remark 3.1 Since Sn = 4n En ( 1 4 ), by (1.2) and the binomial inversion formula we have
n

(3.4)

Sn =
r =0

n (1)nr 2r Er r
11

and
r =0

n Sr = 2n En . r

Theorem 3.2. Let p be a prime greater than 3. (i) If p 1 (mod 4), then h(3p) 4B p+1
2

1 3

(mod p) (mod p)

if p 1 (mod 12), if p 5 (mod 12).

4B p+1
2

1 3

(ii) If p 3 (mod 4), then 1 8B p+1 12 (mod p) 2 1 h(12p) 8B p+1 12 (mod p) 2 8B p+1 1 + 8B p+1 (mod p) 12
2 2

if p 7 (mod 24), if p 11 (mod 12), if p 19 (mod 24)

and h(5p) .

8B p+1 ( 1 5 ) (mod p)
2 p+1 (mod p) 8B p+1 ( 1 5 ) + 4B 2 2

if p 11, 19 (mod 20), if p 3, 7 (mod 20).

Proof. We rst assume p 1 (mod 4). From [UW, p. 40] or [B] we have
[p/3]

h(3p) = 2
x=1

p . x

Thus applying (2.8), (2.9) and the quadratic reciprocity law we see that
[p/3]

h(3p) = 2
x=1

x 4B p+1 2 p

p 3

= 4

1 p B p+1 2 3 3

(mod p).

This proves (i). Now let us consider (ii). Assume p 3 (mod 4). From [UW, pp. 3-5] we have x if p 7, 11, 23 (mod 24), 4 p <x< 2p p 12 12 h(12p) = x 4 if p 19 (mod 24). p 4p 5p
12 <x< 12

By Corollary 2.2 and the fact B2n (x) = B2n (1 x) we nd x p x

p 2p 12 <x 12 p1 2

2 B p+1
2

p 2p 12 <x< 12

p 12

B p+1
2

1 6

(mod p)

and x p x
5p 4p 12 <x 12 p1 2

2 B p+1
2

5p 4p 12 <x< 12

1 B p+1 2 3

5p 12

(mod p).

12

Thus 8 B p+1 2 8 B p+1 h(12p) 2 8 B p+1

2

5 12 1 12 1 3

B p+1
2

1 6 1 6 1 12

(mod p) (mod p) (mod p)

if p 7 (mod 24), if p 11 (mod 12), if p 19 (mod 24).

B p+1
2

B p+1
2

By (2.10) we have B p+1

2

1 2 1 5 B p+1 B p+1 2 2 12 p 6 12
1 6

(mod p). B p+1

2

Thus, if p 7 (mod 24), then h(12p) 8(B p+1 2 (mod p). It is well known that ([GS]) B2n Thus B p+1
2

5 12

) 8B p+1
2

1 12

1 312n 1 = B 2n 3 2

and B2n

1 (212n 1)(312n 1) = B2n . 6 2 3 1 B p+1 (mod p) 2 p

1 p1 1 1 = 3 2 1 B p+1 2 3 2 2

and B p+1
2

(2 1 = 6

p1 2

1)(3 2
3 p

p1 2

1)

B p+1
2

1 2
2

2 1 p
1 6

3 1 B p+1 (mod p). 2 p

If p 11 (mod 12), then 8B p+1

2

= 1 and so B p+1

0 (mod p). Hence h(12p)

3 p

1 12

(mod p). If p 19 (mod 24), then

= 1 and so B p+1
2

1 3

1 B p+1 (mod p). Thus h(12p) 8(B p+1 12 + B p+1 ) (mod p). 2 2 2 Finally we consider h(5p) (mod p). From [UW, p. 40] or [B] we have

h(5p) = 2
p 2p 5 <a< 5

p . a

p Observe that = a a p by the quadratic reciprocity law. Thus applying Corollary 2.2 and (2.9) we obtain

h(5p) = 2
p 2p 5 <a< 5

a 2 p p

p1 2

2p 5 <a< 5

(1) 2 p 2p 2 B p+1 B p+1 2 2 (p + 1)/2 5 5 p 1 2 B p+1 B p+1 (mod p). 4 2 2 5 5 5

13

p1

From (2.9) we see that B and so B p+1

2

p+1 2

+ 2B

p+1 2

1 2 + 2B p+1 = 2 5 5

B p+1
k=0
2

k 5

= 5

p1 2

B p+1
2

1 2 1 + B p+1 2 5 5 2

p 1 B p+1 (mod p). 2 5

Thus h(5p) 4 = p 1 1 p 2B p+1 + 1 B p+1 2 2 5 5 2 5 8B p+1 1 (mod p ) if p 11, 19 (mod 20), 5

2 2

8B p+1

1 5

+ 4B p+1 (mod p) if p 3, 7 (mod 20).

2

The proof is now complete. When d is a negative discriminant, it is known that 1 h(d) < p. Thus, from Theorem 3.2 we deduce the following result. Corollary 3.2. Let p be a prime. (i) If p 1 (mod 4), then B p+1 ( 1 3 ) 0 (mod p). 2 1 (ii) If p 7, 11, 23 (mod 24), then B p+1 ( 12 ) 0 (mod p). 2 1 (iii) If p 11, 19 (mod 20), then B p+1 ( 5 ) 0 (mod p).
2

Remark 3.2 For n = 0, 1, . . . it is well known that k=0 1 ), then this we deduce that if m N and an = mn Bn ( m

n 1 n k nk+1 Bk (x) = x . From n n+1 nk ak = n + 1. k=0 k m

4. p-regular functions. For a prime p, in [S5] the author introduced the notion of p-regular functions. If f (k ) is a complex number congruent to an algebraic integer modulo p for any given n k n nonnegative integer k and k=0 n k (1) f (k ) 0 (mod p ) for all n N, then f is called a p-regular function. If f and g are p-regular functions, in [S5] the author showed that f g is also a p-regular function. Thus we see that p-regular functions form a ring. In the section we discuss further properties of p-regular functions. Suppose n N and k {0, 1, . . . , n}. Let s(n, k ) be the unsigned Stirling number of the rst kind and S (n, k ) be the Stirling number of the second kind dened by
n

x(x 1) (x n + 1) = and xn =
k=0

(1)nk s(n, k )xk

k=0

S (n, k )x(x 1) (x k + 1).

14

For our convenience we also dene s(n, k ) = S (n, k ) = 0 for k > n. For m N it is well known that
n

(4.1)
r =0

n (1)nr rm = n!S (m, n) r

In particular, taking m = n we have the following Eulers identity

n

(4.2)
r =0

n (1)nr rn = n! . r

Lemma 4.1. Let x, d be variables, m, n N and i Z with i 0. Then

n r =0

n rx + d i (1)nr r r m
m m k=j

n! m! j =ni

k (1)mk s(m, k )dkj S (i + j, n)xj . j

In particular we have
n r =0

n rx i n! (1)nr r = (1)mj s(m, j )S (i + j, n)xj . r m m! j =ni

Proof. Since m! rx + d m = (rx + d)(rx + d 1) (rx + d m + 1)

m

=
k=0 m

(1)mk s(m, k )(rx + d)k

k

=
k=0 m

(1)
m

mk

s(m, k )
j =0

k (rx)j dkj j

=
j =0 k =j

k (1)mk s(m, k )dkj rj xj , j

we have
n r =0

n rx + d i (1)nr r r m
m j =0 m k =j

1 m!

k n (1)mk s(m, k )dkj xj (1)nr ri+j . j r r =0

15

Now applying (4.1) we obtain the result.

Lemma 4.2. Let p be a prime and m, n N. Then m!s(n, m) nm p Zp n! Moreover, if m < n, we have m!s(n, m) nm m!S (n, m) nm p p 0 (mod p) n! n! and m!s(n, m) nm 2 n! m nm (mod 2). for p > 2 and m!S (n, m) nm p Zp . n!

Proof. It is well known that (ex 1)m xn = S (n, m) . m! n! n=m Thus, applying the multinomial theorem we see that

(ex 1)m =
k=1

xk k!

=
n=m k1 +k2 ++kn =m k1 +2k2 ++nkn =n

1 m! xn k1 !k2 ! kn ! r=1 r!kr

and so (4.3) S (n, m) =

k1 +k2 ++kn =m k1 +2k2 ++nkn =n

n! 1!k1 k1 !2!k2 k2 ! n!kn kn !

Hence m!S (n, m) nm p = n! pr1 (k1 + k2 + + kn )! k1 !k2 ! kn ! r! r =1

n kr

k1 +k2 ++kn =m k1 +2k2 ++nkn =n

From [S5, pp. 196-197] we also have (4.4) s(n, m) =

k1 +k2 ++kn =m k1 +2k2 ++nkn =n

n! 1k1 k1 !2k2 k2 ! nkn kn !

and m!s(n, m) nm p = n! (k1 + k2 + + kn )! pr1 k1 !k2 ! kn ! r r =1

n kr

k1 +k2 ++kn =m k1 +2k2 ++nkn =n

16

It is known that (k1 + +kn )!/(k1 ! kn !) Z. For r N we know that if p r!(that r r is p | r! but p+1 r!), then = i=1 p p . Thus pr1 /r, pr1 /r! Zp . For i p > 2 we see that pr1 /r pr1 /r! 0 (mod p) for r > 1. Hence the result follows from the above. For p = 2 we see that 2r1 /r 0 (mod 2) for r > 2. Thus m!s(n, m) nm 2 n! (k1 + k2 )! = k1 !k2 ! m nm (mod 2).

k1 +k2 =m k1 +2k2 =n

Summarizing the above we prove the lemma. From Lemma 4.1 we have the following identities, which are generalizations of Eulers identity. Theorem 4.1. Let x, d be variables and m, n N. (i) If m n, then
n r =0

n rx + d nm n! m (1)nr r x . = r m m!
n r =0

In particular, when m = n we have n rx + d (1)nr r n = xn .

(ii) If m n + 1, then
n r =0

n rx + d n+1m n! n(n + 1) m m(m 1 2d) m1 (1)nr r = x x . r m m! 2 2

n r =0

In particular, when m = n + 1 we have n rx + d (1)nr r n+1 = d+ n(x 1) n x . 2

Proof. Observe that s(m, m) = 1 and S (n, n) = 1. Putting i = n m in Lemma 4.1 we obtain (i). By (4.3) and (4.4) we have s(n, n 1) = S (n, n 1) = n(n 1)/2 for Thus applying Lemma 4.1 we see that if m n + 1, then
n r =0

n = 2 , 3, 4, . . .

n rx + d n+1m (1)nr r r m
m m k =j

n! = m! j =m1

k (1)mk s(m, k )dkj S (n + 1 m + j, n)xj j

m k=m1

n! = S (n + 1, n)xm + m! =

k (1)mk s(m, k )dk(m1) xm1 m1

m(m 1) m1 n! n(n + 1) m x + dm x . m! 2 2 This yields (ii) and so the theorem is proved.

17

Corollary 4.1. Let p be an odd prime, m Z and d {0, 1, . . . , p 1}. Then mp m (mod p) and mp m p
p p1

k=1

1 km + d +m k p

d k=1

1 (mod p). k

Proof. From Theorem 4.1(i) we have m =

k=0 p

km + d p (1)pk k p
p1

= As
p1 1 k=1 k

mp + d + p

k=1

km + d p (1)pk . k p

0 (mod p), we see that (mp + d)(mp + d 1) (mp + d p + 1) p! mp (mp + 1) (mp + d)((m 1)p + d + 1) ((m 1)p + p 1) = p (p 1)! =
d

mp + d p

m 1 + mp
k=1 d

1 + (m 1)p k 1 (m 1)p k (mod p2 ).

p1

k=d+1 d k=1

1 k

m 1 + mp
k=1 d

1 k

=m 1+p
k=1

1 k

Let rk be the least nonnegative residue of km + d modulo p. For k {1, 2, . . . , p 1} we see that p p(p 1) (p k + 1) (1)k1 = p (mod p2 ). k k! k Thus, p1 p km + d (1)pk k p
k=1 p1

k=1

p (km + d)(km + d 1) (km + d p + 1) k p! 1 km + d rk 1 k p (p 1)! 1 km + d rk =p k p

18
p1 p1

p1

=p
k=1 p1

(km + d i)
i=0 i=rk

p
k=1

k=1

1 km + d (mod p2 ). k p

Now putting all the above together we obtain the result. Remark 4.1 In the case d = 0, Corollary 4.1 was rst found by Lerch [Ler]. For a dierent proof of Lerchs result, see [S5]. Theorem 4.2. Let p be a prime. Let f be a p-regular function. Suppose m, n N and d, t Z with d, t 0. Then
n r =0

n (1)r f (pm1 rt + d) 0 (mod pmn ). r

k k r =0 r

Moreover, if Ak = pk
n r =0

(1)r f (r), then

n (1)r f (pm1 rt + d) r pmn tn An (mod pmn+1 ) 2

mn n

if p > 2 or m = 1,
mn+1

n n r =0 r

Ar+n (mod 2

if p = 2 and m 2.

Proof. Since f is a p-regular function, we have Ak Zp for k 0. Set

n

a0 = A0

and

ai = (1)

i r =i

s(r, i)

pr Ar r!

for i = 1, 2, . . . , n.

As pr /r! Zp and Ar Zp we have a0 , . . . , an Zp . From [S5, p. 197] we have

n

f (k )
i=0

ai k i (mod pn+1 )

for

k = 0, 1, 2, . . . .

Thus applying (4.1) and (4.2) we see that

n r =0

n n (1)r f (rt + d) (1)r r r r =0

n

ai (rt + d)i
i=0

=
r =0

n (1)r (an tn rn + bn1 rn1 + + b1 r + b0 ) r pn An (t)n n! n!

= an (t)n n! = (1)n s(n, n) = pn tn An (mod pn+1 ),

where b0 , b1 , . . . , bn1 Zp . Thus the result is true for m = 1.

19

Now assume m 2. By the binomial inversion formula we have f (k ) = (p)s As . Thus

n r =0 n

k k s=0 s

n (1)r f (pm1 rt) r n (1)r r

pm1 rt

=
r =0

k=0 n r =0

pm1 rt (p)k Ak k pm1 rt n (1)r r k n! (1)kj s(k, j )S (j, n) pm1 t k ! j =n

k k j

pm1 nt

=
k=0 p
m1

(p)k Ak
nt

=
k=n p
m1

(p)k Ak (1)n
nt

(by Lemma 4.1)

j

=
k=n

(p) (1) Ak
pm1 nt

(1)kj
j =n

s(k, j )j ! kj S (j, n)n! j n p p pm1 t k! j! (1)kn s(k, n)n! kn (m1)n n p p t k!

= An t p +

n mn

+
k=n+1

(p)n (1)k Ak

(1)kj s(k, j )j ! kj S (j, n)n! j n p p (pm1 t)j . ! ! k j j =n+1 s(k, j )j ! kj p Zp k! S (j, n)n! j n p Zp . j!

By Lemma 4.2, for j, k, n N we have and

Hence, by the above, Lemma 4.2 and the fact (m 1)(n + 1) + n mn + 1 we obtain
n r =0

n (1)r f (pm1 rt) r

pm1 nt

pmn tn An +
k=n+1

s(k, n)n! kn p Ak k! if p > 2,

n r =0 n r

mn n mn+1 ) p t An (mod p 2mn tn

2
m1

nt

k =n

n kn

Ak = 2mn tn

Ar+n (mod 2mn+1 )

if p = 2.

Thus the result holds for d = 0. Now suppose g (r) = f (r + d). By the previous argument,
n r =0

n (1)r g (r) pn An (mod pn+1 ). r

20

Thus g is also a p-regular function. Note that

n r =0

n n (1)r f (pm1 rt + d) = (1)r g (pm1 rt). r r r =0

By the above we see that the result is also true for d > 0. The proof is now complete. Theorem 4.3. Let p be a prime, k, m, n, t N and d {0, 1, 2, . . . }. Let f be a p-regular function. Then
n1

f (ktp

m1

+ d)
r =0

(1)n1r
s s r =0 r

k1r n1r

k f (rtpm1 + d) (mod pmn ). r

Moreover, setting As = ps
n1

(1)r f (r) we then have k1r n1r k f (rtpm1 + d) r if p > 2 or m = 1, if p = 2 and m 2. k F (r) r

f (ktpm1 + d)
r =0

(1)n1r
n

k pmn n k 2mn n

(t) An (mod pmn+1 ) (t)n

n n r =0 r

Ar+n (mod 2mn+1 ) k1r n1r

r

Proof. From [S4, Lemma 2.1] we know that for any function F ,
n1

F (k ) = (4.5)
r =0

(1)n1r
k

+
r =n

k r (1)r (1)s F (s), r s s=0

where the second sum vanishes when k < n. Now taking F (k ) = f (ktpm1 + d) we obtain
n1

f (ktp

m1

+ d) =
r =0

(1)n1r
k

k1r n1r
r

k f (rtpm1 + d) r

+
r =n

k r (1)r (1)s f (stpm1 + d). r s s=0

By Theorem 4.2 we have

k r =n

k r (1)r (1)s f (stpm1 + d) r s s=0

n

(1)

k n

n s=0

n (1)s f (stpm1 + d) s if p > 2 or m = 1, if p = 2 and m 2.

n n r =0 r

k n k n

pmn (t)n An (mod pmn+1 ) 2mn (t)n Ar+n (mod 2mn+1 )

Now combining the above we prove the theorem. Putting n = 1, 2, 3 and d = 0 in Theorem 4.3 we deduce the following result.
21

Corollary 4.2. Let p be a prime, k, m, t N. Let f be a p-regular function. Then (i) ([S5, Corollary 2.1]) f (kpm1 ) f (0) (mod pm ). (ii) f (ktpm1 ) kf (tpm1 ) (k 1)f (0) (mod p2m ). (iii) We have f (ktpm1 ) k (k 1) f (2tpm1 ) k (k 2)f (tpm1 ) 2 (k 1)(k 2) + f (0) (mod p3m ). 2

(iv) We have f (kpm1 ) f (0) k (f (0) f (1))pm1 (mod pm+1 ) f (0) 2m2 k (f (2) 4f (1) + 3f (0)) (mod 2m+1 ) if p > 2 or m = 1, if p = 2 and m 2.

Theorem 4.4. Let p be a prime and let f be a p-regular function. Let n N. (i) For d, x Zp and m {0, 1, . . . , n 1} we have
n k=0

n kx + d (1)k f (k ) 0 (mod pnm ). m k

(ii) We have
n k=1

n (1)k f (k 1) f (pn1 1) (mod pn ). k

Proof. From [S5, Theorem 2.1] we know that there are a0 , a1 , . . . , anm1 Z such that f (k ) anm1 k nm1 + + a1 k + a0 (mod pnm ) for Thus applying Lemma 4.1 and (4.1) we have
n k=0 n

k = 0 , 1, 2, . . .

n kx + d (1)k f (k ) k m n kx + d (1)k k m
n nm1

k=0

ai k i
i=0

nm1

=
i=0

ai
k=0

n kx + d i (1)k k = 0 (mod pnm ). k m

This proves (i).

22

Now we consider (ii). By [S5, Theorem 2.1] there are a0 , a1 , . . . , an1 Zp such that s!as /ps Zp (s = 0, 1, . . . , n 1) and f (k ) an1 k n1 + + a1 k + a0 (mod pn ) for k = 0, 1, 2, . . .

Note that ps1 /s! Zp for s N. We then have a1 an1 0 (mod p). Let an1 (k 1)n1 + + a1 (k 1) + a0 = bn1 k n1 + + b1 k + b0 . Then clearly b1 bn1 0 (mod p) and f (k 1) bn1 k n1 + + b1 k + b0 (mod pn ) for Thus k = 1, 2, 3, . . .

f (pn1 1) bn1 (pn1 )n1 + + b1 pn1 + b0 b0 (mod pn ).

Hence, applying (4.1) we have

n k=1

n (1)k f (k 1) k =

n k=1 n1

n (1)k (bn1 k n1 + + b1 k + b0 ) k
n

bi
i=1 k=0

n (1)k k i + b0 k
n1

n k=1 n

n (1)k k

= b0 f (p So the theorem is proved.

1) (mod p ).

5. Congruences for pBk(pm )+b (x) and pBk(pm )+b, (mod pmn ). For given prime p and t Zp we recall that t p denotes the least nonnegative residue of t modulo p. Theorem 5.1. Let p be a prime, and k, m, n, t, b Z with m, n 1 and k, b, t 0. Let x Zp and x = (x + x p )/p. Then pBkt(pm )+b (x) pkt(p
n1
m

)+b

Bkt(pm )+b (x ) k r pBrt(pm )+b (x) prt(p

m

r =0

(1)n1r

k1r n1r

)+b

Brt(pm )+b (x )

where

k (b, n, p) n (t)n pmn1 (mod pmn ) 0 (mod 2mn )

if p > 2 or m = 1, if p = 2 and m 2,

1 (b, n, p) = 0

if p = 2 and n {1, 2, 4, 6, . . . } or if p > 2, p 1 | b and p 1 | n, otherwise.

23

Proof. From [S4, Theorem 3.1] we know that

n k=0

n (1)k pBk(p1)+b (x) pk(p1)+b Bk(p1)+b (x ) pn1 (b, n, p) (mod pn ). k

n n k=0 k

Set f (k ) = p pBk(p1)+b (x) pk(p1)+b Bk(p1)+b (x ) . Then

(1)k f (k )

(b, n, p)pn (mod pn+1 ). Thus f is a p-regular function. Hence appealing to Theorem 4.3 we have
n1

f (ktp

m1

)
r =0

(1)n1r

k1r n1r

k f (rtpm1 ) r if p > 2 or m = 1,

k pmn n k 2mn n

(t)n (b, n, p) (mod pmn+1 ) (t)n

n n r =0 r

(b, n + r, 2) (mod 2mn+1 ) if p = 2 and m 2. 1 0 if n + r {1, 2, 4, 6, . . . }, if n + r {3, 5, 7, . . . }.

Note that (b, n + r, 2) = We then have

n r =0

n (b, n + r, 2) r

(b, 1, 2) + (b, 2, 2) = 1 + 1 0 (mod 2) if n = 1, n n = = 2n1 0 (mod 2) if n > 1. r=0 r

2|n+r

Thus

f (ktpm1 ) k1r (1)n1r p n1r r =0 pmn1

k n

n1

k f (rtpm1 ) r p if p = 2 and m 2.

(t)n (b, n, p) (mod pmn ) if p > 2 or m = 1,

0 (mod 2mn )

This is the result. Corollary 5.1. Let p be a prime, and k, m, b Z with k, m 1 and b 0. Let x Zp and x = (x + x p )/p. Suppose p > 2 or m > 1. Then pBk(pm )+b (x) 3 (mod 4) pBb (x) pb Bb (x ) (mod pm ) if p = m = 2, k = 1 and b = 0, otherwise.

Proof. Putting n = t = 1 in Theorem 5.1 we see that pBk(pm )+b (x) pk(p
m

)+b

Bk(pm )+b (x ) pBb (x) pb Bb (x ) (mod pm ).

24

If p = m = 2, k = 1 and b = 0, then pBk(pm )+b (x) = 2B2 (x) = 2(x2 x + 1 6) m k(pm )+b 3 (mod 4). Otherwise, we have k(p ) + b m + 1 and so p Bk(pm )+b (x ) m 0 (mod p ). Thus the result follows from the above. In the case p > 2, Corollary 5.1 has been proved by the author in [S4]. Let be a primitive Dirichlet character of conductor m. The generalized Bernoulli number Bn, is dened by
m r =1

tn (r)tert B = . n, emt 1 n ! n=0

m

Let 0 be the trivial character. It is well known that (see [W]) B1,0 = 1 , Bn,0 = Bn (n = 1) 2 and Bn, = mn1
r =1 m r =1

(r)Bn

r . m

If is nontrivial and n N, then clearly

m

(r) = 0 and so
m

r Bn ( m ) Bn Bn Bn, = mn1 (r) + n n n r =1

=m

n1 r =1

(r)

r Bn ( m ) Bn . n

r When p is a prime with p m, by [S4, Lemma 2.3] we have (Bn ( m ) Bn )/n Zp . Thus Bn, /n is congruent to an algebraic integer modulo p.

Lemma 5.1. Let p be a prime and let b be a nonnegative integer. (i) ([S5, Theorem 3.2], [Y2]) If p1 b, x Zp and x = (x+ x p )/p, then f (k ) = (Bk(p1)+b (x) pk(p1)+b1 Bk(p1)+b (x ))/(k (p 1) + b) is a pregular function. (ii) ([S5, (3.1), Theorem 3.1 and Remark 3.1]) If a, b N and p a, then f (k ) = (1 pk(p1)+b1 )(ak(p1)+b 1)Bk(p1)+b /(k (p 1) + b) is a p-regular function. (iii) ([Y3, Theorem 4.2], [Y1, p. 216], [F], [S5, Lemma 8.1(a)]) If b, m N, p m and is a nontrivial primitive Dirichlet character of conductor m, then f (k ) = (1 (p) pk(p1)+b1 )Bk(p1)+b, /(k (p 1) + b) is a pregular function. (iv) ([S5, Lemma 8.1(b)]) If m N, p m and is a nontrivial Dirichlet character of conductor m, then f (k ) = (1 (p)pk(p1)+b1 )pBk(p1)+b, is a pregular function. From Lemma 5.1 and Theorem 4.3 we deduce the following theorem. Theorem 5.2. Let p be a prime, k, n, s, t N and b {0, 1, 2, . . . }. (i) If p 1 b, x Zp and x = (x + x p )/p, then Bktps1 (p1)+b (x) pktp (p1)+b1 Bktps1 (p1)+b (x ) ktps1 (p 1) + b
n1
s 1

r =0

(1)n1r

k1r n1r

k r
s1

Brtps1 (p1)+b (x) prtp (p1)+b1 Brtps1 (p1)+b (x ) (mod psn ). rtps1 (p 1) + b
25

(ii) If a, b N and p a, then 1 pktp

n1
s1

(p1)+b1

aktp

s 1

(p1)+b

Bktps1 (p1)+b ktps1 (p 1) + b

s 1

r =0

(1)n1r artp
s 1

k1r n1r 1

k r

1 prtp

(p1)+b1

(p1)+b

Brtps1 (p1)+b (mod psn ). rtps1 (p 1) + b

(iii) If b, m N, p conductor m, then

m and is a nontrivial primitive Dirichlet character of

s 1

(1 (p)pktp (p1)+b1 )Bktps1 (p1)+b, ktps1 (p 1) + b

n1

r =0

(1)n1r
s1

k1r n1r

k r

(1 (p)prtp (p1)+b1 )Brtps1 (p1)+b, (mod psn ). s 1 rtp (p 1) + b (iv) If m N, p m and is a nontrivial Dirichlet character of conductor m, then 1 (p)pktp
n1
s1

(p1)+b1

pBktps1 (p1)+b, k r pBrtps1 (p1)+b, (mod psn ).

r =0

(1)n1r 1 (p)prtp

k1r n1r
s1

(p1)+b1

Remark 5.1 Theorem 5.2 can be viewed as generalizations of some congruences in [S5]. In the case n = 1, Theorem 5.2(i) was given by Eie and Ong [EO], and independently by the author in [S5, p. 204]. In the case s = t = 1, Theorem 5.2(i) was announced by the author in [S4] and proved in [S5], and Theorem 5.2(iii) (in the case p 1 b) and Theorem 5.2(iv) were also given in [S5]. When n = 1, Theorem 5.2(iii) was given in [W, p. 141]. Combining Lemma 5.1 and Corollary 4.2(iv) we obtain the following result. Theorem 5.3. Let p be an odd prime, k, s N and b {0, 1, 2, . . . }. (i) If p 1 b, x Zp and x = (x + x p )/p, then
b1 Bk(ps )+b (x) Bb (x ) s1 Bb (x) p s1 Bp1+b (x) (1 kp ) + kp (mod ps+1 ). k(ps ) + b b p1+b

26

(ii) If b, m N, p conductor m, then

m and is a nontrivial primitive Dirichlet character of

Bk(ps )+b, Bb, Bp1+b , (1 kps1 ) 1 (p)pb1 + kps1 (mod ps+1 ). s k(p ) + b b p1+b (iii) If m N, p m and is a nontrivial Dirichlet character of conductor m, then (1 (p)pk(p
s

)+b1

)pBk(ps )+b,

(1 kps1 ) 1 (p)pb1 pBb, + kps1 1 (p)pp2+b pBp1+b, (mod ps+1 ). Corollary 5.2. Let p be an odd prime and k, s, b N with 2 | b and p 1 b. Then Bk(ps )+b Bp1+b Bb (1 kps1 )(1 pb1 ) + kps1 (mod ps+1 ). s k(p ) + b b p1+b Theorem 5.4. Let p be a prime, a, n N and p a. (i) There are integers b0 , b1 , , bn1 such that 1 pk(p1)1 ak(p1) 1 Bk(p1) k (p 1) for k = 1, 2, 3, . . .

bn1 k n1 + + b1 k + b0 (mod pn ) (ii) If p > 2 or n > 2, then

n k=1

Bk(p1) 1 a(p n (1)k (1 pk(p1)1 )(ak(p1) 1) k (p 1) pn k

(mod pn ).

Proof. Suppose b N. From Lemma 5.1(ii) we know that f (k ) = 1 pk(p1)+b1 ak(p1)+b 1 Bk(p1)+b k (p 1) + b

is a p-regular function. Hence taking b = p 1 and applying [S5, Theorem 2.1] we know that there exist integers a0 , a1 , . . . , an1 such that 1 p(k+1)(p1)1 a(k+1)(p1) 1 B(k+1)(p1) (k + 1)(p 1) for k = 0 , 1, 2, . . .

an1 k n1 + + a1 k + a0 (mod pn ) That is, 1 pk(p1)1 ak(p1) 1 Bk(p1) k (p 1)

an1 (k 1)n1 + + a1 (k 1) + a0 (mod pn ) for

27

k = 1 , 2, 3, . . .

On setting an1 (k 1)n1 + + a1 (k 1) + a0 = bn1 k n1 + + b1 k + b0 we obtain (i). Now we consider (ii). Suppose p > 2 or n > 2. Since f (k ) is a p-regular function, by Theorem 4.4(ii) we have
n k=1

B(k1)(p1)+b n (1)k (1 p(k1)(p1)+b1 )(a(k1)(p1)+b 1) k (k 1)(p 1) + b

n 1

(1 p(p

1)(p1)+b1

)(a(p

n1

1)(p1)+b

1)

B(pn1 1)(p1)+b n (p 1 1)(p 1) +

(mod pn ).

Substituting b by p 1 + b we see that for b 0,

n

(5.1)

k=1

Bk(p1)+b n (1)k (1 pk(p1)+b1 )(ak(p1)+b 1) k k (p 1) + b

n

(1 p(p

)+b1

)(a(p

)+b

1)

B(pn )+b (mod pn ). (pn ) + b

By Corollary 5.1 we have pB(pn ) p 1 (mod pn ). Thus taking b = 0 in (5.1) and noting that (pn ) n + 1 we obtain
n k=1

Bk(p1) n (1)k (1 pk(p1)1 )(ak(p1) 1) k (p 1) k

n

B(pn ) (pn ) n (pn ) 1 pB(pn ) a(p ) 1 (pn )1 a = (1 p ) (mod pn ). pn p1 pn (1 p(p

)1

)(a(p

1)

This completes the proof of the theorem.

k n+1 ). 6. Congruences for k=0 n k (1) pBk(p1)+b (x) (mod p For a N and b Z we dene (a | b) = 1 or 0 according as a | b or a b. n

Lemma 6.1. Let p be an odd prime and n N. Then

n s=1 sn+1 (mod p1)

n s

(p 1 | n) (mod p).

Proof. Let n0 {1, 2, . . . , p 1} be such that n n0 (mod p 1). Since Glaisher (see [D]) it is well known that
n s=0 sr (mod p1)

n s

n0

s=0 sr (mod p1)

n0 s

(mod p) for

r Z.

28

From [S1] we know that

n s=0 sr (mod p1)

n s

=
s=0 snr (mod p1)

n . s

Thus
n s=0 sn+1 (mod p1)

n s

=
s=0 s1 (mod p1)

n s

n0

s=0 sp2 (mod p1)

n0 s

p 1 1 (mod p) if n0 = p 1, = 1 (mod p) if n0 = p 2, 0 (mod p) if n0 < p 2.

n

Hence
n s=1 sn+1 (mod p1)

n s

=
s=0 sn+1 (mod p1)

n (p 1 | n +1) (p 1 | n) (mod p). s

This proves the lemma. Proposition 6.1. Let p be an odd prime, n N and x Zp . Let b be a nonnegative integer. Then
n k=0

n x + x (1)k pBk(p1)+b (x) pk(p1)+b Bk(p1)+b k p

p1

j =0 j = x
p

(x + j )bn pn Bn

(x + j )p (x + j ) + pn (b, n, p) (mod pn+1 ), p(p 1)

where

(n b)T n (n b)T (b, n, p) = bn 0

p1

if p 1 | b and p 1 | n, if p 1 b and p 1 | n, if p 1 | b and p 1 | n + 1, otherwise

and T =

j =0 j = x

(x + j )p1+b (x + j )b . p
p

Proof. Let
n

Sn =
k=0

n x + x (1)k pBk(p1)+b (x) pk(p1)+b Bk(p1)+b k p

29

From [S4, p.157] we know that

n(p1)+b p1 n k=0
p

Sn =
r =0

pr Br
j =0 j = x

n k (p 1) + b (1)k (x + j )k(p1)+br . k r

By [S5, p.199] we know that for any functions f and g we have

n k=0 n

n (1)k f (k )g (k ) k n s
ns i=0

(6.1)

=
s=0

ns (1)i f (i + s) i

s j =0

s (1)j g (j ). j

Now taking f (k ) =
n k=0 n

k(p1)+b r

and g (k ) = ak(p1)+br (a = 0) in (6.1) we obtain

k (p 1) + b k(p1)+br n (1)k a k r n s
ns i=0

=
s=0 n

ns (i + s)(p 1) + b (1)i i r
ns i=0

s j =0

s (1)j aj (p1)+br j

=
s=0

n br a (1 ap1 )s s

ns i(p 1) + s(p 1) + b (1)i . i r

Thus applying the above and Lemma 4.1 we have

n(p1)+b p1 n s=0
p

Sn =
r =0 ns

p Br
j =0 j = x

n (x + j )br 1 (x + j )p1 s

i=0 p1 n

ns i(p 1) + s(p 1) + b (1)i i r n s 1 (x + j )p1 p

s n(p1)+b

=
j =0 j = x s=0
p

pr+s Br (x + j )br
r =ns

ns

i=0

ns i(p 1) + s(p 1) + b (1)i . i r

Since pBr Zp and so pr+s Br 0 (mod pn+1 ) for r n s + 2, by Theorem 4.1 we

30

have
n(p1)+b ns br r +s

(x + j )
r =ns

Br
i=0 ns

ns i(p 1) + s(p 1) + b (1)i i r ns i(p 1) + s(p 1) + b (1)i i ns

ns

(x + j )b(ns) pn Bns
i=0

+ (x + j ) = (x + j )

b(ns+1) n+1

Bns+1
i=0 ns

ns i(p 1) + s(p 1) + b (1)i i ns+1

b(ns) n

p Bns (1 p)

+ (x + j )b(ns+1) pn+1 Bns+1

(s(p 1) + b + (n s)(p 2)/2)(1 p)ns (x + j )b(ns) (1 p)ns pn Bns + (x + j )b(ns+1) (b n)pn+1 Bns+1 (mod pn+1 ). Thus,
p1 n s=0
p

Sn
j =0 j = x

n s

1 (x + j )p1 p

(x + j )bn+s (1 p)ns pn Bns

+ (x + j )bn+s1 (b n)pn+1 Bns+1

p1 n

=
j =0 j = x
p

(x + j )
p1 n

bn

(1 p) p

n n s=0

n s

1 (x + j )p1 x + j p 1p
s

Bns

+
j =0 j = x p1 s=0
p

n s

1 (x + j )p1 p

(x + j )bn+s1 (b n)pn+1 Bns+1

p1 n s=0 p1|ns+1

j =0 j = x
p

(x + j )

bn

(1 p) p Bn (xj ) +
j =0 j = x s
p

n n

n s

where

1 (x + j )p1 p

(x + j )bn+s1 (n b)pn (mod pn+1 ), (x + j )p (x + j ) . p(p 1)

xj = In the last step we use the facts

n

Bn (t) =
s=0

n s t Bns s

and

pBk (p 1 | k ) (mod p) (k 1).

31

For a Z, using Lemma 6.1 and Fermats little theorem we see that
n s=0 sn+1 (mod p1)

n s a = s

n s=1 sn+1 (mod p1) n n+1

n s a + (p 1 | n + 1) s n + (p 1 | n + 1) s

s=1 sn+1 (mod p1)

(p 1 | n)an+1 + (p 1 | n + 1) n+1 a (mod p) if p 1 | n, a = 1 (mod p) if p 1 | n + 1, 0 (mod p) if p 1 n and p 1 n + 1. We also note that (see [S5, (5.1)])
p1 p1

(6.2)
j =0 j = x
p

(x + j )
r =1

rb (p 1 | b) (mod p).

Thus
p1 j =0 j = x n n s=0 p1|ns+1 p1

n s

1 (x + j )p1 p
n

(x + j )bn+s1 (n b)pn

p (n b)
j =0 j = x
p

(x + j )

b s=0 sn+1 (mod p1)

n s

1 (x + j )p1 p

p1 n p ( n b ) (x + j )b ((x + j )p1 1)/p (mod pn+1 ) j =0 j = x p if p 1 | n, pn (n b) (x + j )b (p 1 | b)(n b)pn (mod pn+1 ) j =0 j = x p if p 1 | n + 1, n+1 0 (mod p ) if p 1 n and p 1 n + 1.
p1

On the other hand, for t Zp we have Bn (t) Bn Zp (cf. [S4, Lemma 2.3]) and so (np)pn Bn (xj ) npn+1 Bn
32

npn (mod pn+1 ) if p 1 | n, 0 (mod pn+1 ) if p 1 n.

Thus applying (6.2) we get

p1

(x + j )bn (np)pn Bn (xj )

j =0 j = x

p1 j =0 j = x

(x + j )b npn npn (p 1 | b) (mod pn+1 ) if p 1 | n,

p

0 (mod pn+1 )

if p 1 n.

Hence, by the above and the fact (1 p)n 1 np (mod p2 ) we obtain

p1 p1

(x + j )
j =0 j = x
p

bn

(1 p) p Bn (xj )
j =0 j = x
p

n n

(x + j )bn pn Bn (xj )

p1

j =0 j = x
p

(x + j )bn (np)pn Bn (xj ) npn (mod pn+1 ) if p 1 | b and p 1 | n, 0 (mod pn+1 ) if p 1 b or p 1 n.

Now combining the above we see that

p1

Sn
j =0 j = x
p

(x + j )bn pn Bn (xj )

npn + (n b)pn T (mod pn+1 ) if p 1 | b and p 1 | n, pn (n b)T (mod pn+1 ) if p 1 b and p 1 | n, n n +1 p (b n) (mod p ) if p 1 | b and p 1 | n + 1, n+1 0 (mod p ) otherwise. This is the result. Remark 6.1 When p = 2, b 1 and n 2, setting (b, n, p) = b n we can show that the result of Proposition 6.1 is also true. Theorem 6.1. Let p be a prime greater than 3, x Zp , n N, n 0, 1 (mod p 1) and b {0, 1, 2, . . . }. Let n0 be given by n n0 (mod p 1) and n0 {2, 3, . . . , p 2}. Set
n

Sn =
k=0

n x + x (1)k pBk(p1)+b (x) pk(p1)+b Bk(p1)+b k p

33

Then Sn
n n0 n n0

Sn0 (n+2)b pn0 + 2 Sn 0 n pn0 p (mod

pn (mod pn+1 ) pn+1 )

if p 1 | b and p 1 | n + 1, if p 1 b or p 1 n + 1.

Proof. Since p 1 n we know that Bn /n Zp . For t Zp , by [S4, Lemma 2.3] we have (Bn (t) Bn )/n Zp . Thus Bn (t) Bn (t) Bn Bn = + Zp . n n n As n 0, 1 (mod p 1), by [S5, Corollary 3.1] we have Bn0 (t) pn0 1 Bn0 (t + t p )/p Bn0 (t) Bn (t) (mod p). n n0 n0 Set xj = ((x + j )p (x + j ))/(p(p 1)). Then xj Zp . Thus Bn (xj )/n Zp and Bn (xj )/n Bn0 (xj )/n0 (mod p). From Proposition 6.1 and the above we see that Sn pn
p1

(x + j )bn Bn (xj ) + (b n)(p 1 | b)(p 1 | n + 1)

j =0 j = x
p

p1

n
j =0 j = x
p

(x + j )bn0

Bn0 (xj ) + (b n)(p 1 | b)(p 1 | n + 1) (mod p) n0

and so Sn0 n0 pn0 Thus Sn n Sn0 (b n0 )(p 1 | b)(p 1 | n + 1) n p n0 pn0 + (b n)(p 1 | b)(p 1 | n + 1) n Sn0 n = n0 + b 1 (p 1 | b)(p 1 | n + 1) n0 p n0 n Sn0 n n0 + b 1 + (p 1 | b)(p 1 | n + 1) (mod p). n0 p 2 This proves the theorem.
34
p1

(x + j )bn0
j =0 j = x
p

Bn0 (xj ) + (b n0 )(p 1 | b)(p 1 | n + 1) (mod p). n0

Theorem 6.2. Let p be an odd prime, x Zp , b, n Z with n 1 and b 0. If p | n and p 1 n, then

n k=0

n x + x (1)k pBk(p1)+b (x) pk(p1)+b Bk(p1)+b k p bpn (mod pn+1 ) 0 (mod p

n+1

if p 1 | b and p 1 | n + 1, if p 1 b or p 1 n + 1.

Proof. As p 1 n and p | n, for t Zp we see that Bn (t)/n Zp and so Bn (t) = nBn (t)/n 0 (mod p). Thus the result follows from Proposition 6.1. Theorem 6.3. Let p be an odd prime, n N and b {0, 2, 4, . . . }. If p(p 1) | n, then n n (1)k (1 pk(p1)+b1 )pBk(p1)+b k
k=0

pn1 2pn (mod pn+1 ) 0 (mod p

n+1

if p 1 | b, if p 1 b.

Proof. From Proposition 6.1 we see that

n k=0

n (1)k (1 pk(p1)+b1 )pBk(p1)+b k j bn pn Bn jp j bT pn (mod pn+1 ), p(p 1)

p1

p1

j =1

where T =

j =1

j p1+b j b . p

For p > 3 and m N, from [S5, (5.1)] we have

p1

j m pBm +
j =1

p2 p3 mBm1 + m(m 1)Bm2 (mod p3 ). 2 6

If m 4 is even, then Bm1 = 0 and pBm2 Zp . Thus

p1

(6.3)
j =1

j m pBm (mod p2 ) for

m = 2, 4, 6, . . .

Hence

pBp1+b pBb p pBp1 (p1) p 2

2+b

(mod p) (mod p)
35
3B2 2 3

if p > 3 and b > 0, if p > 3 and b = 0, (mod 3) if p = 3.

2 3

= 2b (1)b = 1

If p > 3 and b = k (p 1) for some k N, by [S4, Corollary 4.2] we have (6.4) and pBp1+b = pB(k+1)(p1) (k + 1)pBp1 k (p 1) (mod p2 ). Thus T pBp1+b pBb pBp1 (p 1) (mod p). p p pBb = pBk(p1) kpBp1 (k 1)(p 1) (mod p2 )

If p > 3 and p 1 b, by Kummers congruences we have Bb Bp1+b (mod p) and so p1+b b Thus T Bp1+b (b 1) Bb (mod p). b

pBp1+b pBb b1 Bb Bb Bb = (mod p). p b b

Summarizing the above we have (6.5) T

pBp1 (p1) (mod p b B b (mod p)

p) if p 1 | b, if p 1 b.

As p(p 1) | n, from Corollary 5.1 we have pBn (x) p 1 (mod p2 ) for x Zp . Note that j n 1 (mod p2 ) for j = 1, 2, . . . , p 1. Combining the above we obtain
n k=0

n (1)k (1 pk(p1)+b1 )pBk(p1)+b k j bn pn1 pBn jp j bT pn p(p 1)

p1

j =1 p1

j =1

j b pn1 (p 1) bT pn (mod pn+1 ).

From (6.3) and (6.4) we see that b b 2 pBb p 1 pBp1 ( p1 1)(p 1) (mod p ) if p > 3, b > 0 and p 1 | b, p1 b 2 j pBb (mod p ) if p > 3 and p 1 b, 2 j =1 p 1 (mod p ) if p > 3 and b = 0, b 3 b 1 + (1 + 3) 2 2 + 2 2 + 6b (mod 9) if p = 3.
36

That is,
p1 b p1 (pBp1

jb
j =1

(p 1)) + p 1 (mod p2 )

if p 1 | b, if p 1 b.

pBb (mod p2 )

Hence
n k=0

n (1)k (1 pk(p1)+b1 )pBk(p1)+b k

p1 n1

(p 1)
j =1

j b bT pn

n1 (b(pBp1 (p 1)) + (p 1)2 ) pn1 b(pBp1 (p 1)) p = pn1 (p 1)2 pn1 2pn (mod pn+1 ) if p 1 | b, n1 Bb n+1 n+1 n Bb 0 (mod p ) if p 1 b. p (p 1) pBb bp ( b ) = p This completes the proof. Theorem 6.4. Let p be a prime greater than 3, x Zp , n N, n 0, 1 (mod p 1) and b {0, 1, 2, . . . }. Let n0 be given by n n0 (mod p 1) and n0 {2, 3, . . . , p 2}. Let f (k ) = pBk(p1)+b (x) pk(p1)+b Bk(p1)+b Then for k = 0, 1, 2, . . . we have
n1

x + x p

f (k )
r =0

(1)

n1r

k1r n1r

k n f (r) + r n0

n0 n0 s=0 s

(1)s f (s) k (p)n pn0 n (p)n (mod pn+1 ).

+ (p 1 | n + 1)(p 1 | b)

(n + 2)b k k 2 n n+1

Proof. From [S4, Theorem 3.1] we have

m k=0

m (1)k f (k ) pm1 (p 1 | m)(p 1 | b) (mod pm ) k

37

for

m N.

Thus applying [S4, Lemma 2.1], Theorem 6.1 and the above we see that
n1

f (k )
r =0 k

(1)n1r
r

k1r n1r

k f (r) r

=
r =n

k r (1)r (1)s f (s) r s s=0

n n+1

k n+1 n k (1)s f (s) (1)n (1)s f (s) + (1)n+1 s s n+1 n s=0 s=0
n0 0 (1)s f (s) (n + 2)b n k (1)n pn s=0 s n0 (p 1 | n + 1)(p 1 | b) + n n0 p 2 k + (1)n+1 pn (p 1 | n + 1)(p 1 | b) (mod pn+1 ). n+1 n

This yields the result. Corollary 6.1. Let k, n N. (i) If n 2 (mod 4), then
n1

(5 5 )B4k
r =0

4k

(1)n1r + 3n

k1r n1r

k (5 54r )B4r r

k n 5 (mod 5n+1 ) n k1r n1r k (5 54r+2 )B4r+2 r

and
n1

(5 5

4k+2

)B4k+2
r =0

(1)n1r n

k n 5 (mod 5n+1 ). n

(ii) If n 3 (mod 4), then

n1

(5 54k )B4k
r =0

(1)n1r +

k1r n1r

k (5 54r )B4r r

k 5n (mod 5n+1 ) n+1 k1r n1r k (5 54r+2 )B4r+2 r

and
n1

(5 5

4k+2

)B4k+2
r =0

(1)n1r +n

k n 5 (mod 5n+1 ). n
38

7. Congruences for Euler numbers. We recall that the Euler numbers {En } are given by
n

E0 = 1, E2n1 = 0

and
r =0

2n E2r = 0 (n 1). 2r

The rst few Euler numbers are shown below: E0 = 1, E2 = 1, E4 = 5, E6 = 61, E8 = 1385, E10 = 50521, E12 = 2702765, E14 = 199360981, E16 = 19391512145. By (1.2) and (2.9) we have E2n = 22n E2n 1 22n+1 3 1 = 2 2n B2n+1 B2n+1 2 2n + 1 4 4 4n+1 2 1 1 = B2n+1 B2n+1 . 2n + 1 4 4
1 2n+1 B2n+1 ( 4 )

That is, (7.1) . 2n + 1 Lemma 7.1. Let p be an odd prime and b {0, 2, 4, . . . }. Then f (k ) = (1 p1 (1) 2 pk(p1)+b )Ek(p1)+b is a pregular function. Proof. As p > 2 and 2 | b we see that p 1 b + 1. For x Zp , from Lemma 5.1(i) we know that F (k ) = (Bk(p1)+b+1 (x) pk(p1)+b Bk(p1)+b+1 (x ))/(k (p 1) + b + 1) is a pregular function, where x = (x + x p )/p. It is clear that
1 4

E2n = 4

+ 1 4 p
1 4

1 1 p(4 1 1 p(4

+ +

p1 1 4 )= 4 3p1 3 4 )= 4

if p 1 (mod 4), if p 3 (mod 4). 1 . 4

Thus, using (2.9) we see that Bk(p1)+b+1 Hence g (k ) = 1 (1)

p1 2

+ 1 4 p

= Bk(p1)+b+1

p 4

= (1)

p1 2

Bk(p1)+b+1

k(p1)+b

Bk(p1)+b+1 ( 1 4) k (p 1) + b + 1
p1 2

= 4(k(p1)+b+1) 1 (1) is a pregular function. For n N we see that

n k=0

pk(p1)+b Ek(p1)+b

n (1)k 4k(p1)+b+1 = 4b+1 (1 4p1 )n 0 (mod pn ). k

Namely, 4k(p1)+b+1 is a pregular function. Hence, using [S5, Theorem 2.3] we see that f (k ) = 4k(p1)+b+1 g (k ) is also a pregular function. This proves the lemma. From Lemma 7.1 and Theorem 4.3 we have:
39

Theorem 7.1. Let p be an odd prime, k, m, n, t N and b {0, 2, 4, . . . }. Then 1 (1)

n1
p1 2

pktp

m1

(p1)+b

Ektpm1 (p1)+b k r 1 (1)

p1 2

r =0

(1)n1r

k1r n1r

prtp

m1

(p1)+b

Ertpm1 (p1)+b (mod pmn ). Putting n = 1, 2, 3 and t = 1 in Theorem 7.1 we obtain the following result. Corollary 7.1. Let p be an odd prime, k, m N and b {0, 2, 4, . . . }. Then p1 (i) ([C, p. 131]) Ek(pm )+b 1 (1) 2 pb Eb (mod pm ). p1 (ii) Ek(pm )+b kE(pm )+b (k 1) 1 (1) 2 pb Eb (mod p2m ). (iii) We have Ek(pm )+b
p1 m k (k 1) E2(pm )+b k (k 2) 1 (1) 2 p(p )+b E(pm )+b 2 p1 (k 1)(k 2) + 1 (1) 2 pb Eb (mod p3m ). 2

From Lemma 7.1 and Corollary 4.2(iv) we have: Theorem 7.2. Let p be an odd prime, k, m N and b {0, 2, 4, . . . }. Then Ek(pm )+b (1 kpm1 )(1 (1)
p1 2

pb )Eb + kpm1 Ep1+b (mod pm+1 ).

Corollary 7.2. Let p be an odd prime and k, m N. Then Ek(pm ) kpm1 Ep1 (mod pm+1 ) 2 + kpm1 (Ep1 2) (mod pm+1 ) if p 1 (mod 4), if p 3 (mod 4).

From [S5, Theorem 2.1] and Lemma 7.1 we have: Theorem 7.3. Let p be an odd prime, n N and b {0, 2, 4, . . . }. Then there are integers a0 , a1 , . . . , an1 such that (1 (1)
p1 2

pk(p1)+b )Ek(p1)+b an1 k n1 + + a1 k + a0 (mod pn )

for every k = 0, 1, 2, . . . Moreover, if p n, then a0 , a1 , . . . , an1 (mod pn ) are uniquely determined. As examples, we have (7.2) (7.3) (7.4) (1 + 32k )E2k 12k + 2 (mod 33 ), (1 54k )E4k 750k 3 + 1375k 2 620k (mod 55 ), (1 54k+2 )E4k+2 1000k 3 + 1500k 2 + 540k + 24 (mod 55 ).
40

Theorem 7.4. Let n N and b {0, 2, 4, . . . }. Suppose n N and 2n 1 n < 2n . Then n n (1)k E2k+b 0 (mod 22nn ). k
k=0

Proof. We rst prove the result in the case b = 0. Taking x = 0 in (1.2) we nd

n r =0

n 2n+1 (1)nr Er = Bn+1 2n+1 Bn+1 . r n+1

Thus applying the binomial inversion formula we have

n

En =
m=0

n 2m+1 (1 2m+1 ) Bm+1 . m m+1

Using this we see that

n k=0

n (1)nk E2k = k =

2k

k=0 m=0 2n

n 2k 2m+1 (1 2m+1 ) Bm+1 (1)nk m+1 k m n 2k (1)nk k m

2m+1 (1 2m+1 ) Bm+1 m + 1 m=0 2m+1 (1 2m+1 ) Bm+1 m + 1 m=1

2n

m 2 k n

n k=0

= By Lemma 4.1 we have

n k=0

n 2k (1)nk . k m

n 2k (1)nk k m

n! (1)mj s(m, j )S (j, n) 2j = m! j =n

m

= Thus,
n k=0

(1)mj
j =n

j !s(m, j ) mj n!S (j, n) j n j +nm 2 2 2 . m! j!

n (1)nk E2k k
2n m

2m+1 (1 2m+1 ) j !s(m, j ) mj n!S (j, n) j n j +nm Bm+1 (1)mj 2 2 2 = m + 1 m ! j ! m=1 j =n j !s(m, j ) mj n!S (j, n) j n j +nm 2m+1 (1 2m+1 ) Bm+1 (1)mj 2 2 2 . = m+1 m! j! m=n j =n
41
2n m

It is well known that 2Bk Z2 . Suppose 2ord2 (m+1) 1 2mord2 (m+1)

m + 1. We then have

2m+1 Bm+1 2Bm+1 = ord (m+1) Z2 . 2 m+1 2 (m + 1)

j !s(m,j ) mj 2 m!

On the other hand, by Lemma 4.2 we have Hence, if n j m 2n, then

Z2 and

n!S (j,n) j n 2 j!

Z2 .

2m+1 (1 2m+1 ) j !s(m, j ) mj n!S (j, n) j n j +nm Bm+1 (1)mj 2 2 2 m+1 m! j! 0 (mod 2j +nord2 (m+1) ). When n j m 2n, we also have m + 1 < 2(n + 1) 2n +1 and so ord2 (m + 1) n , thus j + n ord2 (m + 1) j + n n 2n n . Therefore, by the above we n k 2nn ). So the result holds for b = 0. obtain k=0 n k (1) E2k 0 (mod 2 From [S5, (2.5)] we know that for any function f ,
n

(7.5)
k=0

n (1)k f (k + m) = k

m k=0

m (1)k k

k+n r =0

k+n (1)r f (r). r

Thus,
n

(7.6)
k=0

n (1)k E2k+b = k

b/2

b 2

k+n

k=0

(1)

k r =0

k+n (1)r E2r . r

As s+1 = s or s +1, we see that 2(s+1)s+1 2ss and hence 2r r 2ss for r s. As the result holds for b = 0 we have
k+n r =0

k+n (1)r E2r 0 (mod 22(k+n)k+n ). r

k+n r =0 k +n r

Since 2(k + n) k+n 2n n , we must have Hence applying (7.6) we obtain

n k=0

(1)r E2r 0 (mod 22nn ).

n (1)k E2k+b 0 (mod 22nn ). k

This proves the theorem.

42

Corollary 7.3. Let n N and b {0, 2, 4, . . . }. Then

n k=0

n (1)k E2k+b k

2 (mod 4) 0 (mod 2
n+1

if n = 1, ) if n > 1

and thus f (k ) = E2k+b is a 2regular function. Proof. Suppose n N and 2n 1 n < 2n . By Theorem 7.4 we have
n k=0

n (1)k E2k+b 0 (mod 22nn ). k

If n n, then 2n1 2n 1 n. For n 3 we have 2n1 > n, thus n < n n k and hence 2n n n + 1. Therefore, for n 3 we have k=0 n k (1) E2k+b n+1 0 (mod 2 ). As E0 E2 = 1 (1) = 2 and E0 2E2 + E4 = 1 2(1) + 5 = 8, applying (7.6) and the above we see that Eb Eb+2 E0 E2 = 2 (mod 8) and Eb 2Eb+2 + Eb+4 0 (mod 8). So the result follows. Theorem 7.5. Suppose k, m, n, t N and b {0, 2, 4, . . . }. For s N let s N be s s given by 2s 1 s < 2s and let es = 2s r=0 r (1)r E2r . Then
n1

2m kt+b

r =0

(1)n1r + 2mn

k1r n1r

k E2m rt+b r

k (t)n en (mod 2mn+n+1n+1 ). n

Moreover, for m 2 we have

n1

E2m kt+b
r =0

(1)n1r + 2mn

k1r n1r

k E2m rt+b r (mod 2mn+n+2n+1 ).

k n(n 1) (t)n en + nen+1 + en+2 2 n

s

s (1)r E2r+b . Since s s, by Theorem 7.4 Proof. For s N set As = 2s r=0 r we have As Z2 and 2ss | As . As s+1 s + 1 we have s + 1 s+1 s s and hence r r s s for r s. Therefore 2s s | Ar for r s. As 1 + n+1 n+3 we see that n + 3 n+3 n + 2 n+1 and thus 2n+2n+1 | Ar for r n + 3. By (7.6) we have b/2

An =
k=0

b 2

(1)k 2k ek+n .

Since 2n+2n+1 | er for r n + 3, 2n+2n+1 | 2en+1 and 2n+2n+1 | 22 en+2 , we see that An en (mod 2n+2n+1 ).
43

From Corollary 7.3 and the proof of Theorem 4.2 we know that
n r =0

n (1)r E22m1 rt+b r

2m1 nt n mn

= An t 2
r

+
r =n+1

(2) (1) Ar

(1)rn s(r, n)n! rn (m1)n n 2 2 t r!

(1)rj s(r, j )j ! rj S (j, n)n! j n 2 2 (2m1 t)j . r! j! j =n+1

By Lemma 4.2, for n + 1 j r we have s(r, j )j ! rj S (j, n)n! j n 2 , 2 Z2 r! j! and s(r, n)n! rn 2 r! n rn (mod 2).

As 2n+1n+1 | Ar for r n + 1, by the above we obtain

n

(7.7)
r =0

n (1)r E2m rt+b 2mn An tn 2mn tn en (mod 2mn+n+1n+1 ) r

and so
n

(7.8)
r =0

n (1)r E2m rt+b 0 (mod 2mn+nn ). r

For r n + 1 we have mr + r r m(n + 1) + n + 1 n+1 mn + n + 2 n+1 . Thus, if r n + 1, by (7.8) we have

r

(7.9)
s=0

r (1)s E2m st+b 0 (mod 2mn+n+2n+1 ). s

By (4.5) we have
n1

2m kt+b

=
r =0

(1)n1r
k

k1r n1r
r

k E2m rt+b r

+
r =n

k r (1)r (1)s E2m st+b . r s s=0

Hence, applying (7.9) we obtain

n1

E2m kt+b (7.10)

r =0

(1)n1r
n

k1r n1r

k E2m rt+b r

k n (1)n (1)s E2m st+b (mod 2mn+n+2n+1 ). n s s=0

44

In view of (7.7), we get

n1

2m kt+b

r =0

(1)n1r +

k1r n1r

k E2m rt+b r

k (1)n 2mn tn en (mod 2mn+n+1n+1 ). n

Now assume m 2. Then (m 1)(n + 1) + n mn + 1. From the above we see that

n r =0

n (1)r E2m rt+b r

2m1 nt mn

An t +
r =n+1

(2)n (1)r Ar
2m1 nt

(1)rn s(r, n)n! rn (m1)n n 2 2 t r!

n+2

mn n

An +
r =n+1

n n Ar 2mn tn Ar rn rn r =n n en+2 (mod 2mn+n+2n+1 ). 2

2mn tn en + nen+1 +

This together with (7.10) yields the remaining result. Hence the proof is complete. As 2nn | en and n + 1 n+1 n n , by Theorem 7.5 we have: Corollary 7.4. Let k, m, n, t N and b {0, 2, 4, . . . }. Let N be given by 21 n < 2 . Then
n1

E2m kt+b
r =0

(1)n1r

k1r n1r

k E2m rt+b (mod 2mn+n ). r

Corollary 7.5. Let k, m N and b {0, 2, 4, . . . }. Then E2m k+b 2m k + Eb (mod 2m+1 ). Proof. Observe that e1 = 1 and e2 = 2. For m 2, taking n = t = 1 in Theorem 7.5 we obtain E2m k+b Eb + 2m (k )(e1 + e2 ) 2m k + Eb (mod 2m+1 ). So the result holds for m 2. Now taking m = 2 and b = 0, 2 in the congruence we see that E4k 1+4k (mod 8) and E4k+2 1+4k (mod 8). Hence E2k (1)k (mod 4) and so E2k+b (1)k+b/2 (1)b/2 + 2k Eb + 2k (mod 4). So the result is also true for m = 1. This completes the proof. Remark 7.1 Corollary 7.5 is equivalent to the following Sterns result (see [St]): 2m En1 En2 2m n1 n2 . Putting n = 2, t = 1 in Theorem 7.5 and noting that e2 = 2, e3 = 10, e4 = 104 we obtain the following result.
45

Corollary 7.6. Let k, m N, m 2 and b {0, 2, 4, . . . }. Then E2m k+b kE2m +b (k 1)Eb + 22m k (k 1) (mod 22m+2 ). Taking m = 2 and b = 0, 2 in Corollary 7.6 we get: Corollary 7.7. For k N we have E4k and E4k+2 4k + 1 (mod 64) 4k + 33 (mod 64) 4k 1 (mod 64) 4k 33 (mod 64) if k 0, 1 (mod 4), if k 2, 3 (mod 4) if k 0, 1 (mod 4), if k 2, 3 (mod 4).

Corollary 7.8. Let k, m N, m 2 and b {0, 2, 4, . . . }. Let k = 0 or 1 according as 4 k 3 or 4 | k 3. Then E2m k+b k k1 E2m+1 +b k (k 2)E2m +b + Eb + 23m+1 k (mod 23m+2 ). 2 2

Proof. Observe that e3 = 10, e4 = 104, e5 = 1816 and k 3 k (mod 2). Taking n = 3 and t = 1 in Theorem 7.5 we obtain the result. Taking m = 2, b = 0, 2 in Corollary 7.8 and noting that E8 105 (mod 256), E10 89 (mod 256) we deduce: Corollary 7.9. Let k N and k = 0 or 1 according as 4 k 3 or 4 | k 3. Then E4k 48k 2 44k +1+128k (mod 256) and E4k+2 16k 2 76k 1+128k (mod 256). Remark 7.2 Let {Sn } be given by (3.1). From Remark 3.1 we know that (1)k Sk is a 2-regular function and hence f (k ) = (1)k+b Sk+b is also a 2-regular function, where b {0, 1, 2, . . . }. Thus, by Corollary 4.2, for m 2, k 1 and b 0 we have S2m1 k+b Sb (mod 2m ) and S2m1 k+b Sb 2m2 k (Sb+2 +4Sb+1 +3Sb ) (mod 2m+1 ). References
[B] [C] [D] [EO] [Er] [F] [GS] B.C. Berndt, Classical theorems on quadratic residues, Enseign. Math. 22 (1976), 261-304. K.W. Chen, Congruences for Euler numbers, Fibonacci Quart. 42 (2004), 128-140. L.E. Dickson, History of the Theory of Numbers, Vol.I, Chelsea Publ., 1999, p. 271. M. Eie and Y.L. Ong, A generalization of Kummers congruences, Abh. Math. Sem. Univ. Hamburg 67 (1997), 149-157. R. Ernvall, A congruence on Euler numbers, Amer. Math. Monthly 89 (1982), 431. G.J. Fox, Kummer congruences for expressions involving generalized Bernoulli polynomials, J. Th eor. Nombres Bordeaux 14 (2002), 187-204. A. Granville and Z.W. Sun, Values of Bernoulli polynomials, Pacic J. Math. 172 (1996), 117-137. 46

[IR] [K] [L] [Ler]

K. Ireland and M. Rosen, A Classical Introduction to Modern Number Theory (2nd edition), Springer, New York, 1990, pp. 238,248. ber die Verteilung der quadratischen Reste, J. Reine Angew. Math. 127 L.C. Karpinski, U (1904), 1-19. E. Lehmer, On congruences involving Bernoulli numbers and the quotients of Fermat and Wilson, Ann. Math. 39 (1938), 350-360.
p1

M. Lerch, Zur Theorie des Fermatschen Quotienten a p 1 = q (a), Math. Ann. 60 (1905), 471-490. [MOS] W. Magnus, F. Oberhettinger and R.P. Soni, Formulas and Theorems for the Special Functions of Mathematical Physics (3rd edition), Springer, New York, 1966, pp. 25-32. [St] M.A. Stern, Zur Theorie der Eulerschen Zahlen, J. Reine Angew. Math. 79 (1875), 67-98. n 2 n [S1] Z.H. Sun, Combinatorial sum and its applications in number theory I, J. k
k=0 kr (mod m)

[S2]

Nanjing Univ. Math. Biquarterly 9 (1992), 227-240. n 2 n Z.H. Sun, Combinatorial sum and its applications in number theory II, J. k
k=0 kr (mod m)

[S3] [S4] [S5] [SS] [Su1] [Su2] [UW] [W] [Y1] [Y2] [Y3]

Nanjing Univ. Math. Biquarterly 10 (1993), 105-118. 2 n Z.H. Sun, Combinatorial sum and its applications in number theory III, J. k
kr (mod m)

Nanjing Univ. Math. Biquarterly 12 (1995), 90-102. Z.H. Sun, Congruences for Bernoulli numbers and Bernoulli polynomials, Discrete Math. 163 (1997), 153-163. Z.H. Sun, Congruences concerning Bernoulli numbers and Bernoulli polynomials, Discrete Appl. Math. 105 (2000), 193-223. Z.H. Sun and Z.W. Sun, Fibonacci numbers and Fermats last theorem, Acta Arith. 60 (1992), 371-388. 2 n and related congruences, Israel J. Math. 128 (2002), Z.W. Sun, On the sum k
kr (mod m)

135-156. Z.W. Sun, Binomial coecients and quadratic elds, Proc. Amer. Math. Soc. 134 (2006), 2213-2222. J. Urbanowicz and K.S. Williams, Congruences for L-Functions, Kluwer Academic Publishers, Dordrecht, Boston, London, 2000, pp. 3-8, 28, 40, 55. L.C. Washington, Introduction to Cyclotomic Fields, Springer, New York, 1982, pp. 30-31, 141. P.T. Young, Congruences for Bernoulli, Euler, and Stirling numbers, J. Number Theory 78 (1999), 204-227. P.T. Young, Kummer congruences for values of Bernoulli and Euler polynomials, Acta Arith. 99 (2001), 277-288. P.T. Young, Degenerate and n-adic versions of Kummers congruences for values of Bernoulli polynomials, Discrete Math. 285 (2004), 289-296.

47