ADVANCED PROBLEMS AND SOLUTIONS

EDITED BY
FLORIAN LUCA

Please send all communications concerning ADVANCED PROBLEMS AND SOLUTIONS

to FLORIAN LUCA, SCHOOL OF MATHEMATICS, UNIVERSITY OF THE WITWA-
TERSRAND, PRIVATE BAG X3, WITS 2050, JOHANNESBURG, SOUTH AFRICA or by
e-mail at the address [email protected] as files of the type tex, dvi, ps, doc, html, pdf,
etc. This department especially welcomes problems believed to be new or extending old results.
Proposers should submit solutions or other information that will assist the editor. To facilitate
their consideration, all solutions sent by regular mail should be submitted on separate signed
sheets within two months after publication of the problems.

PROBLEMS PROPOSED IN THIS ISSUE

H-787 Proposed by Hideyuki Ohtsuka, Saitama, Japan.

Prove that √
∞
X 1 7−3 5
= .
n=1
α2Fn+1 F2Fn F2Fn+2 2

H-788 Proposed by Hideyuki Ohtsuka, Saitama, Japan.

Given c > 0 determine
v v
u u s
u u r q
u t
cF22 cF42 cF82 · · · + cF22n .
t
lim + + +
n→∞

H-789 Proposed by D. M. Bătineţu-Giurgiu, Bucharest and Neculai Stanciu,

Buzău, Romania.
q
2 2
For any real numbers x, y we denote B(x, y) = x +xy+y 3 . Prove that for n ≥ 1, we have
 2
Ln+2 − 3 1 X Ln Ln+1 − 2
(i) ≤ B 2 (L) ≤ ;
n n n
n cyclic
 
Fn+2 − 1 2 1 X Fn Fn+1
(ii) ≤ B 2 (F ) ≤ ,
n n n
n cyclic
where for a sequence X := {Xm }m≥1 we use
X
B 2 (X) = B 2 (X1 , X2 ) + B 2 (X2 , X3 ) + · · · + B 2 (Xn−1 , Xn ) + B 2 (Xn , X1 ).
n cyclic

MAY 2016 185

THE FIBONACCI QUARTERLY

H-790 Proposed by Ovidiu Furdui, Cluj-Napoca, Romania.

Calculate
∞  
X ζ(2) ζ(3) ζ(n)
Hn − γ − − − ··· − ,
n=2
2 3 n
1 1
where ζ denotes the Riemann zeta function and Hn = 1 + + · · · + is the nth harmonic
2 n
number.

H-791 Proposed by Hideyuki Ohtsuka, Saitama, Japan.

For an integer n ≥ 0 find a closed form expression for the sum
n k
X (−1)2
.
F3k+1 (L3k L3k+1 · · · L3n )2
k=0

SOLUTIONS

The Number of Solutions of a Family of Boolean Equations

H-756 Proposed by Russell J. Hendel, Towson University.

(Vol. 52, No. 3, August 2014)
We seek to generalize a known problem that states that
#{hx1 , . . . , xn+1 i : x1 x2 ∨ x2 x3 ∨ · · · ∨ xn xn+1 = 0} = Fn+3 (1)
where xi are Boolean variables for i = 1, . . . , n. To generalize the above formula, we
(i) fix integers d, i with d > i ≥ 1;
(ii) let Dj be products of d Boolean variables xk with consecutive indices such that Dj
and Dj+1 have i variables in common;
(iii) let m be the total number of variables occurring in D1 , . . . , Dn and
(iv) let
Sn = #{hx1 , . . . , xm i : D1 ∨ D2 ∨ · · · ∨ Dn = 0}.
Determine the coefficients of the minimal recursion satisfied by the {Sn }n≥1 .

Solution by the proposer.

Let us start with some examples.
Examples: To illustrate the notation we use (1):
d = 2, i = 1, Dj = xj xj+1 and Sn = Fn+3 .
The (minimal) recursion satisfied by the {Sn }n≥1 is Sn+2 = Sn+1 + Sn .
For certain special cases of d and i, it is easy to discover the minimal recursions satisfied by
{Sn }n≥1 . Some illustrative examples are as follows:
 
Sn = 2d−1 − 1 Sn−1 + 2d−2 Sn−2 , for i = 1 and arbitrary d.



Sn = 2i − 1 Sn−1 + 2i − 1 Sn−2 , for d = 2i and arbitrary i.
d
X
Sn = Sn−j , for i = d − 1 and arbitrary d.
j=1

186 VOLUME 54, NUMBER 2

 ADVANCED PROBLEMS AND SOLUTIONS

Now we are ready to proceed to the solution. With notations as presented in the proposed
problem, we have the following theorem.
Theorem 1. (i) Define k by d = 2i + k. If k ≥ 0, then
   
Sn = 2i+k − 1 Sn−1 + 2i+k − 2k Sn−2 . (2)
(ii) Define g, p, k by g = d − i and d = pg + k, 1 ≤ k ≤ g. If p ≥ 1,
Xp  
Sn = (2g − 1) Sn−u + 2g − 2g−k Sn−p−1 . (3)
u=1

Since, the proofs of (2) and (3) are similar, it suffices to present the proof of (3). Prior to
presenting the proof, we present some simple examples illustrating our notations.
Example 2. In (3), let d = 5, i = 3. Then g = 2, p = 2 and k = 1.
The equation D1 ∨ D2 ∨ · · · ∨ Dn = 0 becomes
x1 x2 x3 x4 x5 ∨ x3 x4 x5 x6 x7 ∨ · · · ∨ x2(n−1)+1 · · · x2(n−1)+5 = 0.
Equation (3) asserts that the solutions satisfy the recursion Sn = 3Sn−1 + 3Sn−2 + 2Sn−3 for
all n ≥ 4.
Further numerical examples may be found in Table 1.

Table 1. The table presents numerical examples illustrating the main theo-
rem. The top row lists the degree d while the left-most column lists i. For
example row i = 1 and column d = 2 corresponds to the family of Boolean
equations x1 x2 ∨ x2 x3 ∨ · · · ∨ xn xn+1 = 0 where all disjuncts have degree d = 2
and every two consecutive disjuncts have i = 1 variables in common. Row i = 1
and column d = 2 contain the coefficients of the minimal recursion satisfied by
the {Sn }n≥1 , that is, Sn = Sn−1 + Sn−2 .

d=2 3 4 5 6 7 8

i=1 h1, 1i h3, 2i h7, 4i h15, 8i h31, 16i h63, 32i h127, 64i

2 h1, 1, 1i h3, 3i h7, 6i h15, 12i h31, 24i h63, 48i

3 h1, 1, 1, 1i h3, 3, 2i h7, 7i h15, 14i h31, 28i

4 h1, 1, 1, 1, 1i h3, 3, 3i h7, 7, 4i h15, 15i

5 h1, 1, 1, 1, 1, 1i h3, 3, 3, 2i h7, 7, 6i

To prove (3), it will be convenient to only treat the case k < g as the proof for the case
k = g is similar and omitted. Using this assumption it is easy to verify that
$ % * +
d d
p= , k= , (4)
g g
where bxc is the greatest integer not exceeding x and hd/gi is the remainder of the division of
d by g.

MAY 2016 187

THE FIBONACCI QUARTERLY

In the proof, we will use word terminology from semigroups. More specifically, we will speak
about the prefix, factor or suffix of an elementary conjunction; for example, if discussing the
elementary conjunction x2 x3 x4 x5 , x2 is a prefix, x5 is a suffix, and x3 x4 is a factor. We will
also interchange word and vector notation: e.g. we will say hx3 , x4 i is a factor of x2 x3 x4 x5 .
We use boldface 1 to indicate the vector of all 1’s, so that e.g. hxj+1 , . . . , xj+g i =
6 1 means
that not all g variables xj+1 , . . . , xj+g are identically 1.
Figure 1, which facilitates the presentation of the proof, compactly summarizes the rela-
tionship between the indices of the Boolean variables and the disjuncts Dj .
The proof uses an induction argument. The two propositions below correspond to the base
case and induction step.
Proposition 3.
n
6 1} = (2g − 1)Sn−1 .
#{hx1 , . . . , xm i : ∨ Dj = 0, with hxm−g+1 , . . . , xm i =
j=1
n
Proof. The requirement that ∨ Dj = 0, implies Dn = 0, which in turn requires that some
j=1
variable occurring in Dn has value 0. There are (2g − 1) ways for hxm−g+1 , . . . , xm i =
6 1. By
definition of the Sn , there are Sn−1 ways for the remaining m − g variables to be solutions to
n−1
∨ Dj = 0. 
j=1

length g suffix of Dn−(b d c+1) length (g − h gd i) factor of Dn−b d c

g g
z }| {z }| {
xm−(b d c+2)g+1 · · · xm−(b d c+1)g xm−(b d c+1)g+1 · · · xm−(b d c+1)g+(g−h d i)
g g g g g
| {z }| {z }
last g literals of Dn−(b d c+1) (g − h dg i) consecutive literals of Dn−b d c
g g

length h dg i prefix of Dn
z }| {
xm−b d cg−h d i+1 · · · xm−b d cg
g g g
| {z }
length h dg i factor of Dn−q , 0 ≤ q ≤ b dg c
Suffix of Dn−q Suffix of Dn−(q−1) Suffix of Dn
z }| {z }| { z }| {
· · · xm−(q+1)g+1 · · · xm−qg xm−qg+1 · · · xm−(q−1)g · · · xm−g+1 · · · xm
| {z }| {z } | {z }
length g factor of Dn g consecutive literals in Dn last g literals in Dn

Figure 1. Illustration of the relation of the variables xk to the Dj in (3).

Proposition 4. Suppose that for some integer q ≥ 1, we have

#{hx1 , . . . , xm i : nj=1 ∨ Dj = 0, with hxm−qg+1 , . . . , xm−(q−1)g i =
6 1 and
q
X
xm−u have arbitrary values for 0 ≤ u ≤ (q − 1)g − 1} = (2g − 1)Sn−u . (5)
u=1
If additionally,
(q + 1)g < d, (6)
then (5) holds with q replaced by q + 1.
Proof. Replacing q with q + 1 in the left-hand side of (5) necessitates assuming
hxm−(q+1)g+1 , . . . , xm−qg i =
6 1, (7)

188 VOLUME 54, NUMBER 2

 ADVANCED PROBLEMS AND SOLUTIONS

and
xm−u have arbitrary values for 0 ≤ u ≤ qg − 1. (8)
By (6), xm−(q+1)g+1 · · · xm−qg is a factor of Dn and hence (7) implies that Dn = 0.
There are 2g −1 ways that (7) can take place. By the definition of the Sn , there are Sn−(q+1)
n−(q+1)
ways for the first m−(q +1)g variables, x1 , x2 , . . . , xm−(q+1)g , to be solutions to ∨ Dj = 0.
j=1
n
Using an induction assumption, we conclude that the total number of solutions of ∨ Dj = 0
j=1
with (8) and (7) holding is
q
X q+1
X
(2g − 1)Sn−(q+1) + (2g − 1)Sn−u = (2g − 1)Sn−u .
u=1 u=1

Corollary 5. Using the notation in (4), we have
n
#{hx1 , . . . , xm i : ∨ Dj = 0, with hxm−pg+1 , . . . , xm−(p−1)g i =
6 1 and
j=1
p
X
xm−u have arbitrary values 0 ≤ u ≤ (p − 1)g − 1} = (2g − 1)Sn−u . (9)
u=1

Proof. A routine induction argument with Proposition 3.1 as the base case and Proposition
3.2 as the induction step. 
Completion of the Proof of (3).
Proof. Assume $ %
d
xm−u = 1, 0≤u≤ g − 1. (10)
g
Since we require Dn = 0, this assumption requires that at least one of the h dg i variables,
xm−b d cg−h d i+1 , . . . , xm−b d cg equals 0; that is, (10) implies
g g g

hxm−b d cg−h d i+1 , . . . , xm−b d cg i =

6 1. (11)
g g g

hdi
There are 2 g − 1 ways (11) can take place. Note that (11) also implies that Dn−q = 0, 1 ≤
q ≤ b dg c, because the word on the left side of (11) is also a factor of Dn−q , 1 ≤ q ≤ b dg c.
Consequently, we are indifferent to the values of the g − h dg i variables
xm−(b d c+1)g+1 , . . . , xm−(b d c+1)g+(g−b d c) .
g g g

g−h dg i
There are 2 ways this can happen. By the definition of the Sn , there are Sn−(b d +1c) ways
g
n−(b gd c+1)
for the remaining m − (b dg c + 1)g variables to be solutions to ∨ Dj = 0.
j=1
h dg i g−h dg i
Hence, by (4), there are a total (2 − 1)2 Sn−(b d +1c) = (2k − 1)2g−k Sn−(p+1) solutions
g
n
to ∨ Dj = 0 with (10) and (11) holding.
j=1
The proof of (3) is completed by combining this case with (9). 

MAY 2016 189

THE FIBONACCI QUARTERLY

The Lucas Factorial of a Prime

H-757 Proposed by H. Ohtsuka, Saitama, Japan.

(Vol. 52, No. 2, May 2014)
For an odd prime p prove that
p 
Y 2(−1)(p+1)/4 (mod Fp ) if p ≡ −1 (mod 4),
Lk ≡
(−1)(p−1)/4 Fp−3 (mod Fp ) if p ≡ 1 (mod 4).
k=1

Solution by the proposer.

 
n
Let denote the Fibonomial coefficient. For 1 ≤ k ≤ p − 1, we have
k F
gcd(Fk , Fp ) = Fgcd(k,p) = F1 = 1. (1)
From Theorem 1 of [1], we have
n n  
Y X 2n + 1
L2k = .
k F
k=1 k=0
Letting p = 2n + 1 in the above identity, by (1), we have
(p−1)/2 (p−1)/2   (p−1)/2
Y X p X Fp Fp−1 · · · Fp−k+1
L2k = =1+ ≡ 1 (mod Fp ). (2)
k F Fk Fk−1 · · · F1
k=1 k=0 k=1
From Theorem 3 of [1], we have
n 2n  
Y X
n−k 2n √
L2k−1 = i , where i= −1.
k F
k=1 k=0
Letting p = 2n − 1 in the above identity, by (1), we have
(p+1)/2 p+1  
(p+1)/2−k p + 1
Y X
L2k−1 = i
k F
k=1 k=0
(p+1)/2
=i + i−(p+1)/2 + i(p−1)/2 Fp+1 + i−(p−1)/2 Fp+1
p−1
X Fp+1 Fp · · · Fp−k+2
+ i(p+1)/2−k ×
Fk Fk−1 · · · F1
k=2
(p+1)/4
= (−1) (1 + (−1)(p+1)/2 ) + (−1)(p−1)/4 (1 + (−1)(p−1)/2 )Fp−1 (mod Fp ).
So from the above, we immediately conclude that if p ≡ −1 (mod 4), then
(p+1)/2
Y
L2k−1 ≡ 2(−1)(p+1)/4 (mod Fp ), (3)
k=1
while if p ≡ 1 (mod 4), then
(p+1)/2
Y
L2k−1 ≡ (−1)(p−1)/4 2Fp−1 ≡ (−1)(p−1)/4 (Fp + Fp−3 ) ≡ (−1)(p−1)/4 Fp−3 (mod Fp ).
k=1
(4)

190 VOLUME 54, NUMBER 2

 ADVANCED PROBLEMS AND SOLUTIONS

Since
p (p−1)/2 (p+1)/2
Y Y Y
Lk = L2k L2k−1 ,
k=1 k=1 k=1
the desired congruence follows from (2), (3) and (4).

References
[1] E. Kılıç, I. Akkuş, and H. Ohtsuka, Some generalized Fibonomial sums related with the Gaussian q-binomial
sums, Bull. Math. Soc. Sci. Math. Roumanie Tome, 55.103 (2012), 51–61.

Limits of Factorials and Tangents at Exponents Involving Fibonacci Numbers

H-758 Proposed by D. M. Bătineţu-Giurgiu, Bucharest and Neculai Stanciu,

Buzău, Romania.
Compute:
  √  Fm+2 !!
√
n Fm p
n Fm+1 π(n + 1) n+1 n + 1
lim n! (2n − 1)!! tan √ −1 .
n→∞ 4n n n

Solution by Ángel Plaza.

Since Fm+2 = Fm+1 + Fm+1 the proposed limit may be obtained by the product of the
following two limits:
   √  Fm
√
n π(n + 1) n+1 n + 1
lim n! tan √ −1 , and
n→∞ 4n n n
   √  Fm+1
p
n π(n + 1) n+1 n + 1
lim (2n − 1)!! tan √ −1 .
n→∞ 4n n n
 π Fm  π Fm+1
These limits are respectively equal to and from where the result follows.
2e e
Let us show that   √  
√
n π(n + 1) n+1 n + 1 π
lim n! tan √ −1 = .
n→∞ 4n nn
2e
√
n π


n! 1 tan 4 x − 1 π
By Stirling’s formula lim = , and lim = . Then
n→∞ n e x→1 x−1 2
  √    √ 
√
n π(n + 1) n+1 n + 1 π (n + 1) n+1 n + 1
lim n! tan √ − 1 = lim n √ − 1
n→∞ 4n n n 2e n→∞ nnn
π √ √

= lim (n + 1) n+1 n + 1 − n n n
2e n→∞ √
π (n + 1) n+1 n + 1
= lim (Stolz-Cezaro)
2e n→∞ n
π
= .
2e
(2n)!
The second limit may be obtained similarly taking into account that (2n − 1)!! = n ,
p 2 n!
n
(2n − 1)!! 2
and by Stirling’s formula that lim = .
n→∞ n e
MAY 2016 191
THE FIBONACCI QUARTERLY

Also solved by Dmitry Fleischman, Hideyuki Ohtsuka, and the proposers.

Late Acknowledgement. Kenneth B. Davenport solved H-755.

√
35 − 15 5
Errata. In H-783 (iii), the right-hand side of the equality to be proved is instead
√ 18
35 − 15 3
of .
18

192 VOLUME 54, NUMBER 2