Professional Documents
Culture Documents
v1
Abstract
We derive new infinite series involving Fibonacci numbers and Riemann zeta num-
bers. The calculations are facilitated by evaluating linear combinations of polygamma
functions of the same order at certain arguments.
1 Introduction
The Fibonacci numbers, Fn , and the Lucas numbers, Ln , are defined, for all integers n by
the Binet formulas:
αn − β n
Fn = , Ln = α n + β n , (1.1)
α−β
where α and β are the zeros of the characteristic polynomial,√x2 − x − 1, of the Fibonacci
sequence. Thus α +√β = 1 and αβ = −1; so that α = (1 + 5)/2 (the golden ratio) and
β = −1/α = (1 − 5)/2. Koshy [8] and Vajda [10] have written excellent books dealing
with Fibonacci and Lucas numbers.
Our purpose in writing this paper is to employ the properties of the polygamma functions
to derive infinite series identities involving Fibonacci numbers, Lucas numbers and the
Riemann zeta numbers. We will obtain sums such as
∞
à √ ! à √ !
X (j + 1)(j + 2) 2π 3 π 5 π 5
2
j
ζ(j + 3)F2j = √ tan sec ,
j=1
3 5 6 6
and à √ !
∞
X
j−1 2 2π 5
(−1) jζ(j + 1)Lj−1 = π sec − 3,
j=1
2
and, in fact, more general series. Here ζ(n) is the Riemann zeta function.
The digamma function, ψ(z), is the logarithmic derivative of the Gamma function:
d Γ0 (z)
ψ(z) = log Γ(z) = ,
dz Γ(z)
where the Gamma function is defined for <(z) > 0 by
Z ∞ Z ∞
Γ(z) = −t z−1
e t dt = (log(1/t))z−1 dt ,
0 0
and is extended to the rest of the complex plane, excluding the non-positive integers,
by analytic continuation. The Gamma function has a simple pole at each of the points
z = · · · , −3, −2, −1, 0. The Gamma function extends the classical factorial function to the
complex plane: Γ(z) = (z − 1)!
The nth polygamma function ψ (n) (z) is the nth derivative of the digamma function:
dn+1 dn (n)
ψ (n) (z) = log Γ(z) = ψ (z), ψ (0) (z) = ψ(z) .
dz n+1 dz n
The polygamma functions satisfy the recurrence relation,
m!
ψ (m) (z + 1) = ψ (m) (z) + (−1)m , (1.2)
z m+1
and the reflection relation,
dm
(−1)m ψ (m) (1 − z) − ψ (m) (z) = π cot(πz) . (1.3)
dz m
The Taylor series for the polygamma functions is
∞
X (m + j)!
(−1)m+j+1 ζ(m + j + 1)z j = ψ (m) (z + 1), m ≥ 1, (1.4)
j=0
j!
∞
X
−γ+ (−1)j+1 ζ(j + 1)z j = ψ(z + 1) , (1.5)
j=1
where γ is the Euler-Mascheroni constant and ζ(n), n ∈ C, is the Riemann zeta function,
defined by
X ∞
1
ζ(n) = n
, <(n) > 1 ,
j=1
j
Further information on the polygamma functions can be found in the books by Erdé-
lyi et al. [3, §1.16] and Srivastava and Choi [9, p. 33]. The book by Edwards [2] is a good
treatise on the Riemann zeta function.
Infinite series involving Fibonacci numbers and Riemann zeta numbers were also derived
by Frontczak [4, 5, 6], Frontczak and Goy [7] and Adegoke [1].
2
Preprints (www.preprints.org) | NOT PEER-REVIEWED | Posted: 20 November 2020 doi:10.20944/preprints202011.0539.v1
2 Preliminary results
Here we derive more functional equations for the polygamma functions. We also evaluate
required linear combinations of the polygamma function at appropriate arguments.
m! m!
ψ (m) (−x) + ψ (m) (−y) = −(ψ (m) (x) + ψ (m) (y)) + m+1 + m+1
x y
¯ ¯ (2.3)
dm ¯ dm ¯
− π m cot(πz)¯¯ − π m cot(πz)¯¯ , m odd .
dz z=x dz z=y
In particular, if x + y = 1, then,
¯
d m ¯ m! m!
ψ (m) (−x) − ψ (m) (−y) = −π m cot(πz)¯¯ + m+1 − m+1 , m even , (2.4)
dz z=y x y
¯
dm ¯ m! m!
ψ (m)
(−x) + ψ (m)
(−y) = −π m cot(πz)¯¯ + m+1 + m+1 , m odd . (2.5)
dz z=y x y
Writing 1/2 − z for z in (1.3) gives
µ ¶ µ ¶ ¯
1 1 dm ¯
m (m)
(−1) ψ +z −ψ (m)
− z = π m cot(πx)¯¯ . (2.6)
2 2 dx x=1/2−z
3
Preprints (www.preprints.org) | NOT PEER-REVIEWED | Posted: 20 November 2020 doi:10.20944/preprints202011.0539.v1
¡ ¢
ψ (m) (1 + x) + ψ (m) (1 + y) = (−1)m ψ (m) (1 − x) + ψ (m) (1 − y)
dm dm
− π m cot(πx) − π m cot(πy)
dx dy (2.13)
(−1) m! (−1)m m!
m
+ + .
xm+1 y m+1
From the functional equation (2.12), it follows that if m is even and x + y = 1, then,
¯
dm ¯ m! m!
ψ (1 + x) − ψ (1 + y) = π m cot(πz)¯¯
(m) (m)
+ m+1 − m+1 . (2.14)
dz z=y x y
From the functional equation (2.13), it follows that if m is odd and x + y = 1, then,
¯
dm ¯ m! m!
ψ (1 + x) + ψ (1 + y) = π m cot(πz)¯¯
(m) (m)
− m+1 − m+1 . (2.15)
dz z=y x y
4
Preprints (www.preprints.org) | NOT PEER-REVIEWED | Posted: 20 November 2020 doi:10.20944/preprints202011.0539.v1
and ¯ ¯
dm ¯ d m ¯
−π m cot(πz)¯¯ + π m cot(πz)¯¯ = 0, m odd
dz z=y dz z=1−y
µ µ ¶ µ ¶¶ µ ¶ µ ¶
m (m) 1 (m) 1 (m) 1 (m) 1
(−1) ψ +x +ψ +y = −ψ −x −ψ −y
2 2 2 2
¯ ¯
dm ¯ dm ¯
+ π m cot(πz)¯¯ + π m cot(πz)¯¯ .
dz z=1/2−x dz z=1/2−y
(2.19)
Thus, if m is even and x + y = 1, then from (2.18) and the duplication formula, we have
µ ¶ µ ¶ ¯
1 1 dm ¯ m!
ψ (m)
+x −ψ (m)
+ y = −π m cot(πz)¯¯ + , (2.20)
2 2 dz z=1/2−y (x − 1/2)m+1
and if m is odd and x + y = 1, then from (2.19) and the duplication formula, we have
µ ¶ µ ¶ ¯
1 1 dm ¯ m!
ψ (m)
+x +ψ (m)
+ y = −π m cot(πz)¯¯ − . (2.21)
2 2 dz z=1/2−y (x − 1/2)m+1
¯
dm ¯ m! m! √
ψ (m) 3
(α ) − ψ (m)
(β ) = π m cot(πz)¯¯
3
+ √ + m+1 Fm+1 5, m even , (2.26)
dz z=2β ( 5)m+1 2
¯
d m ¯ m! m!
ψ (m) (α3 ) + ψ (m) (β 3 ) = −π m cot(πz)¯¯ − √ − m+1 Lm+1 , m odd , (2.27)
dz z=2β ( 5)m+1 2
5
Preprints (www.preprints.org) | NOT PEER-REVIEWED | Posted: 20 November 2020 doi:10.20944/preprints202011.0539.v1
¯
± ± dm ¯ m!2m+1
ψ (α 2) − ψ (β 2) = −π m cot(πz)¯¯ √ + √
(m) 3 (m) 3
, m even , (2.28)
dz z= 5/2 ( 5)m+1
¯
± ± dm ¯ m!2m+1
ψ (α 2) + ψ (β 2) = −π m cot(πz)¯¯ √ − √
(m) 3 (m) 3
, m odd , (2.29)
dz z= 5/2 ( 5)m+1
¯
dm ¯
ψ (α /Lr ) − ψ (β /Lr ) = −π m cot(πz)¯¯
(m) r (m) r
, m even , (2.30)
dz z=αr /Lr
¯
dm ¯
ψ (α /Lr ) + ψ (β /Lr ) = −π m cot(πz)¯¯
(m) r (m) r
, m odd , (2.31)
dz z=αr /Lr
µ r ¶ µ r ¶
(m) α (m) β √
ψ √ −ψ √ = (−1)rm+r+m m!Frm+1 (αr 5)m+1 , (2.32)
Fr 5 Fr 5
Proof. To prove (2.22) and (2.23), set x = α, y = β in (2.9). To prove (2.24) and (2.25),
set x = α, y = β in (2.14) and (2.15). Setting x = 2α, y = 2β in (2.16) and (2.17)
gives (2.26) and (2.27). To prove (2.28) and (2.29) set x = α, y = β in (2.20) and (2.21).
Use of x = αr /Lr and y = β √
r
/Lr in (2.9) gives√(2.30) and (2.31). Identity (2.32) is
obtained by setting x = α /(Fr 5) and y = β r /(Fr 5) in (2.8). Identities (2.33) and (2.34)
r
follow from (2.16) and (2.17), upon setting x = 2αr /Lr and y = 2β r /Lr . Identities (2.35)
and (2.36) are obtained from (2.4) and (2.5), with x = αr /Lr and y = β r /Lr .
6
Preprints (www.preprints.org) | NOT PEER-REVIEWED | Posted: 20 November 2020 doi:10.20944/preprints202011.0539.v1
3 Main results
Theorem. If r is an integer, then,
∞
X (m + j)!
(−1)j+1 ζ(m + j + 1)Frj z j
j!
j=1 (F)
1 © ª
= √ ψ (m) (1 + αr z) − ψ (m) (1 + β r z) , m even, m ≥ 0 ,
5
∞
X (m + j)!
(−1)j ζ(m + j + 1)Lrj z j
j=0
j! (L)
(m) r (m) r
=ψ (1 + α z) + ψ (1 + β z), m odd, m ≥ 1 .
Proof. Writing αr z for z in the Taylor series (1.4), we obtain
∞
X (m + j)!
(−1)m+j+1 ζ(m + j + 1)αrj z j = ψ (m) (αr z + 1) . (3.1)
j=0
j!
Similarly,
∞
X (m + j)!
(−1)m+j+1 ζ(m + j + 1)β rj z j = ψ (m) (β r z + 1) . (3.2)
j=0
j!
If m is even, then subtraction of (3.2) from (3.1) gives (F) while if m is odd, their addition
produces (L), on account of the Binet formulas.
Note that, in view of identities (2.10) and (2.11), the right hand side of (F) and (L) can be
expressed as:
1 © ª
√ ψ (m) (αr z + 1) − ψ (m) (β r z + 1)
5
(3.3)
1 © (m) r (m) r
ª (−1)r m!
= √ ψ (α z) − ψ (β z) − Fr(m+1) , m even ,
5 z m+1
∞
X (m + j)! ζ(m + j + 1)
(−1)j L3j
j=0
j! 2j
¯ (3.6)
dm ¯ m! 2m+1
= − π m cot(πz)¯¯ √ − √ − m! 2m+1 L3m+3 , m odd .
dz z= 5/2 ( 5) m+1
7
Preprints (www.preprints.org) | NOT PEER-REVIEWED | Posted: 20 November 2020 doi:10.20944/preprints202011.0539.v1
Proof. Set r = 3, z = 1/2 in (F) and (L), noting (3.3) and (3.4) and using (2.28) and (2.29).
Example 1. We have
∞
X (j + 1)(j + 2)
(−1)j+1 ζ(j + 3)F3j
j=1
2j
à √ ! à √ ! (3.7)
2π 3 π 5 π 5 13616
= − √ cot csc2 + ,
5 2 2 25
∞
à √ !
X (j + 1) π 5 364
(−1)j j
ζ(j + 2)L3j = π 2 csc2 − . (3.8)
j=0
2 2 5
∞
X (m + j)!
(−1)j ζ(m + j + 1)Lrj
j=0
j! Ljr
¯ (3.10)
dm ¯
= −π m cot(πz)¯¯ − m!Lm+1
r Lr(m+1) , m odd .
dz r
z=β /Lr
Proof. Set z = 1/Lr in (F) and (L), noting (3.3) and (3.4) and using (2.30) and (2.31).
Example 2. We have
∞
X (j + 1)(j + 2)
(−1)j+1 ζ(j + 3)F2j
j=1
3j
à √ ! à √ ! (3.11)
2π 3 π 5 π 5
= √ tan sec2 − 432 ,
5 6 6
∞
à √ !
X π 5
(−1)j−1 jζ(j + 1)Lj−1 = π 2 sec2 − 3. (3.12)
j=1
2
8
Preprints (www.preprints.org) | NOT PEER-REVIEWED | Posted: 20 November 2020 doi:10.20944/preprints202011.0539.v1
∞
X (m + j)! j
(−1)j 2 ζ(m + j + 1)Lrj
j=0
j!Ljr
¯ (3.14)
dm ¯ m!Lm+1 m!Lm+1 Lr(m+1)
= − π m cot(πz)¯¯ − √r − r
m+1
, m odd .
dz z=2αr /Lr (Fr 5)m+1 2
Proof. Set z = 2/Lr in (F) and (L) and use (2.33) and (2.34).
Example 3. We have
∞
X (j + 1)(j + 2)
(−1)j+1 ζ(j + 3)F3j
j=1
2j
à √ ! à √ ! (3.15)
13616 2π 3 π 5 π 5
= − √ cot csc2 ,
25 5 2 2
∞
à √ !
X (j + 1) π 5 364
(−1)j j
ζ(j + 2)L3j = π 2 csc2 − . (3.16)
j=0
2 2 5
∞ ¯
X (m + j)! dm ¯
ζ(m + j + 1)Lrj = −π m cot(πz)¯¯ , m odd . (3.18)
j=0
j! Ljr dz z=β r /Lr
Proof. Set z = −1/Lr in (F) and (L) and use (2.35) and (2.36).
Example 4. We have
∞
à √ ! à √ !
X (j + 1)(j + 2) 2π 3 π 5 π 5
ζ(j + 3)F2j = √ tan sec2 , (3.19)
j=1
3j 5 6 6
∞
à √ !
X (j + 1) 3π 5
ζ(j + 2)L4j = π 2 sec2 . (3.20)
j=0
7j 14
9
Preprints (www.preprints.org) | NOT PEER-REVIEWED | Posted: 20 November 2020 doi:10.20944/preprints202011.0539.v1
References
[1] K. Adegoke, Fibonacci series from power series, manuscript,
doi: 10.20944/preprints202011.0463.v1 (2020).
[4] R. Frontczak, Infinite series involving fibonacci numbers and the Riemann zeta func-
tion, Notes on Number Theory and Discrete Mathematics 26:2 (2020), 159–166.
[5] R. Frontczak, Problem B-1267, The Fibonacci Quarterly 58:2 (2020), to appear.
[6] R. Frontczak, Problem H-xxx, The Fibonacci Quarterly 58:2 (2020), to appear.
[7] R. Frontczak and T. Goy, General infinite series evaluations involving Fibonacci num-
bers and the Riemann zeta function, arXiv:2007.14618v1 [math.NT], (2020).
[8] T. Koshy, Fibonacci and Lucas Numbers with Applications, Wiley-Interscience, (2001).
[9] H. M. Srivastava and J. Choi, Zeta and q-Zeta Functions and Associated Series and
Integrals, Elsevier Inc., (2012).
[10] S. Vajda, Fibonacci and Lucas Numbers, and the Golden Section: Theory and Appli-
cations, Dover Press, (2008).
10