Documente Academic
Documente Profesional
Documente Cultură
1.
INTRODUCTION
In [2] , Webb and P a r b e r r y study the divisibility properties of the Fibonacci polynomial
sequence {f (x)} defined by the recursion
f n + 2 (x) = xf n + 1 (x) + f n (x);
f0(x) = 0,
ft(x) = 1.
As one would expect, these polynomials possess many properties of the Fibonacci sequence
which, of course, is just the integral sequence {f (1)}. However, a most surprising result
is that f (x) is irreducible over the ring of integers if and only if p is a prime. In contrast,
for the Fibonacci sequence, the condition that n be a prime is necessary but not
for the primality of f (1) = F .
For instance,
sufficient
F 1 9 = 4181 = 37-113.
u 0 (x,y) = 0,
u t (x,y) = 1.
The first few t e r m s of the sequence are as shown in the following table:
n
u n (x,y)
x2 + y
x 3 + 2xy
x 4 + 3x2y + y2
x 5 + 4x 3 y + Sxy2
x 6 + 5x4y + 6x2y2 + y3
114
DIVISIBILITY PROPERTIES
The basic fact that we will need is that Z [ x , y ] ,
[April
and y a r e in
Z[x,y]
irreducible,
/
x +
a = a(x,y) x =
N/
x2 + 4y
g
*
and
o = j3(x,y)
ot
\ =
P
2
- \/x
~ + 4y
*
are
For n ^ 0,
n
Theorem 2.
a - jS
For m ^ 0 and n ^ 0,
, ,_. = u 1 1 u J 1 + y Ju u
m+n+1
m+1 n+1
m n
(u , u + - ) = 1 for
n ^ 0.
To obtain
For n > 0,
(y, u ) = 1.
The assertion is clearly true for n = 1 since \xt = 1. Assume that it is true
V i
x u
k+ y\-i
the assertion is also true for n = k + 1, and hence for all n ^ 1 as claimed.
We can now prove
1974]
115
For n ^ 0 ,
(u
N , u , - ) = 1.
n n+1'
Proof. Again the result is trivially true for n = 0 and n = 1 since u0 = 0, % = 1,
and u2 = x. Assume that it is true for n = k - 1 where k is any fixed integer, k ;> 2, and
let d(x,y) = (u,, u, - ).
Since
V i
this implies that d(x,y) | u. -y.
But then d(x,y) | 1 since
(u,
y\-i
claimed.
Lemma 5. For n ^ 0,
r(n-l)/2]
n-lJ/2]
X ^
u (x, y)
Proof.
/ n - i - 1 \ n-2i-l i
We define the empty sum to be zero, so the result holds for n = 0. For n = 1,
0)
x u y u = 1 = ut
Vi
Vi
[(k-l)/2]
[0^2)/2]
k2
= v ( -1 - * )x - y + 2
i=0
Then
[(k-l)^],
'
i=0
t 2
E ( -;- ) x - v
'
|V2l
|;(t-1i-1i),k-v
[k/2]
^ *
/ k - i\
i=0
k-2i i
'
Thus, the result holds for n = k + 1 and hence also for all n ^ 0 as claimed.
116
DIVISIBILITY PROPERTIES
3.
[April
Theorem 6. For m ^ 2, u
u if and only if m n.
J
;
' m ' n
'
Proof. ClearlyJ u
u . Now suppose that u
u.
where k ^ 1 is fixed.
m ' m
^
m ' km
using Theorem 2,
(k+l)m
Then,
km+m
= u. u M + Jyu.
-u
km m+1
km-1 m
But, since u
| u,
Thus, u
u if m n.
m ' n
'
Now suppose that m ^ 2 and that u
| u,, ^
u = u
n
mq+r
= u
,., u + Jyu
u
.
mq+1 r
mq r - 1n
Since u
u
by the first part of the Fproof, this implies
that u
u
, _. u . But, since
F
m i mq J
*
m ' mq+1 r
(u
, u
- ) = 1 by Theorem 4, this implies that u | u
and this is impossible, since
H
>
0 and s
<
0,
such that
(m,n) = r m + sn .
Thus, by Theorem 2,
rm
(m,n)+(-s)n
1/
vU
, -
(m,n) -sn+1
But then d
-, U
(m,n)-l -sn
and d u
by Theorem 6 and so d u,
,-. But, (d, u
,-, )=
xu
-sn
' rm J
' (m,n) -sn+1
-sn+1
1 by Theorem 4, and so d | u,
\ by Useful Property A from Section 1. Thus, d =
u,
as claimed.
(m,n)x
1
-r y u /
J
1974]
117
F.
Since
u (x, 1)
is
irreducible by Theorem 1 of [2], it follows that u (x,y2) and hence also u (x,y) is i r r e d u cible over the rational field and thus is irreducible over the integers.
4.
, then r | F
| F
. Also, if m
sequence of Fibonacci polynomials {f (x)} considered by Webb and P a r b e r r y and the generalized sequence {u (x,y)} considered here. In a sense, the first problem is solved by Webb
and P a r b e r r y for the sequence of Fibonacci polynomials, since they give explicitly the roots
of each such polynomial.
s e s s the derived property.
However, it is still not clear exactly which polynomials r(x) p o s On the other hand, it is immediate that the first result mentioned
can divide any f (x) since this would imply that f (c) = 0, and this is impossible
u (x,y)
and hence
, then r | u
By Theorem 6, if m j n ,
then u
if and only if m | n.
| u . Therefore, if r | u
and yet m | n .
Therefore, by Theorem 2,
= u
= u
Since r
mq+s
,., u + J yu
u _. .
mq+1 s
mq s-1
complete.
we have by
118
DIVISIBILITY PROPERTIES
Theorem 10.
For n ^ 2,
n1
'
u n (x,y) =
Proof.
[April
/
n i x -
kir \
2Wycos J
Vx.y2) = y n - \ f f. A = y n " lf n m
where f (x) is the n
Thus,
y(n_1)/2fn(x/^7)
u n (x,y) =
(-5_
- 2i cos i^L )
k=i v ^
This, with the preceding equation, immediately yields the desired result.
Corollary 10.
For n ^ 2, n even,
u (x,y) = x
(n-2)/2 /
\
k
II
I x 2 + 4y cos 2 I
n
k=i \
/
(n
u (x,y) =
n
Proof.
"l)/2 /
n
k=l
kTT \
( x2 + 4 y c o s 2 J
n
\
/
(n - k)77
= - cos
n
It is clear from the preceding theorems that there is a precise correspondence between
the polynomial factors of u (x,y) and those of u (x, 1) = f (x). Thus, it suffices to consider
1974]
119
only those of f (x). Also, it is clear that, except for the factor x, the only polynomial factors of f (x) with integral coefficients contain only even powers of x. While we a r e not able
to say in every case which even polynomials a r e factors of some f (x) we offer the following
partial results.
Theorem 11.
(i) x J f (x) if and only if n is even.
(ii) (x2 + 1) | f (x) if and only if 3 | n .
(iii) (x2 + 2) | f (x) if and only if 4 | n.
(iv) (x2 + 3) |f (x) if and only if 6 |n.
(v) (x2 + c)|f (x) if c / 1, 2, o r 3 and c is an integer.
Proof.
Since, except for x only, all polynomials with integral coefficients dividing
any f (x) must be even, the results (i) through (iv) all follow from Theorem 9 with y = 1.
One has only to observe that f2(x) is the first Fibonacci polynomial divisible by x,
that
f3(x) is the first Fibonacci polynomial divisible by x + 1, and so on. P a r t (v) follows from
the fact that 1 < 4 cos 2 a < 4 for an a in the interval (0, TT/2).
Theorem 12. Let m be a positive integer and let N(m) denote the number of even
polynomials of degree 2m and with integral coefficients which divide at least one (and hence
infinitely many) m e m b e r s of the sequence {f (x)}. Then
N(m) < n I r 1 4k
Proof.
Let f(x) be any polynomial counted by N(m). It follows from Corollary 10 with
y = 1 that
n/ x
2m ,
f(x) = x
+
_ix
2m-2 ,
,
" '
ix
2 .
is the k
0 < a
<
m-k
m
n (x2 + a.)
J
3=1
where a. = 4 cos 2 j3. with 0 < |3. < TT/2 for each j . Therefore,
Since a
m
( ty
120
DIVISIBILITY PROPERTIES
OF GENERALIZED FIBONACCI POLYNOMIALS
N(m) <
as claimed.
(")
n f r 1 4k
Of course, the estimate in Theorem 12 is exceedingly crude and can certainly be i m proved.
It is probably too much to expect that we will ever know the exact value of N(m)
for every m.
Our final theorem shows that with but one added condition the generalization to u (a,b)
of the first result mentioned in this section is valid.
Theorem 13. Let r be a positive integer with (r,b) = 1. Then there exists m such
that r
(a,b).
m
Proof. Consider the sequence
1
u (a,b)
modulo
r.
r2
r.
and t with
such that
u (a,b) = u,(a,b)
s
z
(mod r)
and
u
(a,b) = u t + 1 ( a , b )
(mod r)
But
bu s _ 3 (a,b) = u g + 1 ( a , b ) - au g (a,b)
and
bu,
1 (a,b)
= u, - (a,b) - au,(a,b)
(mod r)
1 (a,b)
= u t _ x (a,b)
(mod r) .
(a,b) = u, (a,b)
t-s
(mod r)
CS
REFERENCES
1. S. MacLane and G. Birkhoff, Algebra, The MacMillan Company, New York, N.Y.,1967.
2. W. A. Webb and E. A. P a r b e r r y , "Divisibility P r o p e r t i e s of Fibonacci Polynomials,"
Fibonacci Quarterly, Vol. 7, No. 5 (Dec. 1969), pp. 457-463.
This paper is modelled after an article by Hansen [l] dealing with identities for Fibonacci and Lucas triples.
F r e e use has been made of the methods of that article, and this
paper follows its format closely. It is hoped that seeing Fibonacci methods used in a slightly
different context will lead the reader to a deeper understanding of those methods, in addition
to the production of some new Pell identities.
The Pell sequence is closely akin to the Fibonacci sequence; it is defined by P 0 = 0,,
PA = 1, P n + 2 = P
985, ' .
+ 2Pn+r
R0 = 2,
B1 = 2,
It is simple
to show that, with these definitions, P ,- + P n = R .
F
l 0 = K + 2R ,.,.
n+2
n
n+1
n+1
n-1
n
Another useful result, easily proved by the usual Fibonacci methods, gives the Pell sequence
and its Lucas analogue as functions of their subscripts:
a^_-j
= an+
a - p
where
a = 1 + N/2*
and
j3 = 1 - si2 .
Note that a and j8 are roots of the equation x 2 - 2x - 1 = 0, and hence aj3 = - 1 and a +
(3=2.
Using the generating functions of
{P
and
{H
I 00
122
11=0
[April
n=0
m \X ""^ on n
D
I m X"^ nn nn
oHi
a
N a x - jS
> ]3 x
1
a - |3
(1)
1
/
a - p \
n=0
1
fim
1 - ax ~ P
1
\
1-jSx)
/ zm
\
0 niv
0/ m-1
0m-h
/ (a - P ) - ofifc
- p
)x \
I
(1 - ax)(l - j3x)
I
1
or - ]8
P
n=0
- P
x
m - n1
1 - 2x - x2
R
2
<>
> W
+ R
,x
m
-'
1 - 2x - x^
n=0
letting 1 - 2x - x2 = D.
*
n-2
-2
+ P
- 3 X _ -2 + 5x
D
~
D
'
n=0
n-2
-2
+ R
-3X
6-
14x
D
~D
n=0
x*
VlX
-l
+ P
- 2 X _ 1 - 2x
"~D
Y ^
'
n=0
2^
xn
n-l X
-l
+ R
- 2 X _ -2
-1
+ X
>
n =0
P v11 -
-1
6x
n=0
P v11 n=0
\ ^
/ ^
2X
1974]
123
oo
pn+1^ . i D^ . iD
1,v J
V.,/-"
-DD
,'
/ j n+1
n=0
n=0
n+2 X
_J
1_
2 +x
D
X
n
/ _ ^ i n+2 X
'
n=0
J2
1_
6 + 2x
D
n=0
n _
3+
2 X _ 5 + 2x
D
D
4+
n+3
.
'
n=0
T>
n+4 X
12 + 5x
D
'
n=0
\ ^ ,,
n _
/ j n+3
n=0
E".
V ^
n
/ J n+4
n=0
3+
4+
2X
D
3X
D
14 + 6x
D
34 + 14x
D
Using the fact that two s e r i e s a r e equal if and only if the corresponding coefficients are
equal, we now find several elementary identities.
Since
2 - 2x
D
_1_ 1 - 2x
D
D
it follows that
n=0
n=0
n=0
(P ^ + P n ) x n
n+1
n-1
n=0
and hence
(3)
R = P , - + P _. ;
n
n+1 n - 1
n a whole number .
n+1
P
= (-1) P
-n
n
and R
-n
124
[April
(-n)+l
+ P
(-n)-l
-(n-l)
-n(n+l)
= (-l)(n-1)+1P
,
n-1
(-l)(n+1)+1P
+1
n+1
= (-l) (P , + P _,J
'
n-1
n+1
(-l)nR
'
n
= R
-n
Hence Eq. (3) holds for all integers n.
=
We now proceed with some theorems necessary to the development of Pell triples.
Theorem 1. P R
+P , R
1 = R _ L 1 .
n m
n-1 m - 1
m+n-1
Proof. Let m be any fixed integer. Then
EP x
(PR
+ P nR
, )x n = R
,
n m
n-1 m - 1
m / J
n=0
Rm - 1, Y
ip
immaKd
n=
n=0
n=0
xn
n -n1
^ + R
m D
m
R x
m
+ R
_, - 2R
,x
R
,+R
Qx
m-1
m-1
_ m-1
m-2
D
D
OO
n+m-1
-IT-
n=0
m-1 /
n=0
P
n-1
n=0
. + P
1 - 2x
m D
m-1
D
P x + P
n - 2P
nx
m
m-1
m-1
D
n=0
n+m-1
D
_,
0x + P
m-2
m-1
D
1974]
n
(R R + R n R n )x
n m
n-1 m - 1 7
n=0
= R
E xn+R
n
m-1 / J
n=0
n=0
m
m-1
lX
n-1
-- 2R m x - 2R m - 1. + 6R m - 1n x
D
2R
m
2(R
V.B
- R
J
m-1
+ 2(3R
-, - R )x
m-1
m
D
R
+ R
+ (2R
m-20
, - 2R
)x
m-1
m-20
D
R
+ R
m
+ (R
m-20
+ R
m - n1
D
x
+ R
m - 1, x + Rm-2o + R m - 3o
D
R
n=0
+
9
n+m-2
xn
n=0
(R a.
n=0
)x
m -03
n+m
-,
o)* 11
n+m-2
and hence,
RR
+ R , R
, = R ^
n m
n-1 m-1
n+m
Now,
+R,
n+m-2
125
126
G E N E R A T I N G I D E N T I T I E S FOR P E L L T R I P L E S
R
n - 1n
+ R
n - n1
=(P_
L O
n+2
= 2P
= 4P
=
+ P )
n
+ (P
+ P
[April
)
n - 02
, - + 3P + P 0
n+1
n
n-2
+ 3P
+ 2P
- + P
n-1
n - 20
8P
n
D2
(pr + qt) + (pt + qr - 2qt)x
-
D2
OO
+P
- x R + R , x
m-1
s
s-1
D
"
D
oo
=V P
fx /
n=0
(4)
n+m
00
xn.VE
/
xn
n+s
n=0
n
k+m
n-k+s
n=0 k=0
but also, by Eq. (4),
P
+ P
D
_,x R + R ,x
-P
, R ,
m-1 t s
s-1 _
m-1 s-1
D
D
(P
R
m
-P
+ P
s
,R 1
m - 1 s-1
D
z -,
-D
m-1
"1
,R
- ) + (P R n + P
-R - 2 P
-.R
Jx
m - 1 s-1
m s-1
m-1 s
m-1 s-1
D2
R
^ , + (P
,R + P
Jx
QR
m+s-1
m-1 s
m-2 s-1
2
D
R ^ - + (P
-R + P
Jx
0R
v 1 , m+s-1
m-1 s
m-2 s-1
S-1
D
i)2
s _ 1
'
n + 1
n=0
oo
(-P
n=0
E
n=0
n+m+s-1
n=0
n=0
n
.^^ZE
R
)xll+
X1P
n+1
m - 1n s - n1
>
/
>
J /
n=0
+y^A
k=0
P a,R
.j. ^
xn
A l k+1
n - k + m + s -n1
k=0
k+lRn-k+m+s-l)x
1974]
127
Hence,
k+m
-P
n-k+s
k=0
P
R
+ 1 P
R
n+1 m - 1 s-1
/ i *k+l n-k+m+s-1
k=0
and
n + l P mi - l RR s - l
- j/ / l fJ
(P
k+lRn-k+m+s-l '
k+mRn-k+s)
k=0
Letting p = m - 1, q = n + 1, and r = s - 1, we obtain
Theorem 4.
q-1
P
"P
P 9
"D
r
^,
/p
_|_ J}
13
jLszJ
k=0
^+1 p+q+r-k
T>
p+k+1 q+r-k
Now we convolute
P
+ P
m
,x
m-1
.,,
with
P, + P, . x
t
t-1
D
Similarly, we convolute
R
m
+ R nx
m-1
D
to obtain
...
with
R, + R, n x
t
t-1
D
Theorem 6.
P-l
P
PVr
Yl
(8P
k=0
Now,
R R R
= (P _,, + P - ) R R
p q r
p+1
p-l
q r
= P_L1RR + P
,RR
p+1 q r
p-l q r
128
April 1974
(8P
q+r+k+l P p - k + l
q+k+l R p + r - k + l *
k=0
p-2
(8P
X j
q+r+k+lPp-k-l '
q+k+l R p + r - k - l *
k=0
P-l
8P
q+r+k+l
(P
+ P
p-k
p-k-l )
k=0
R
+
q+k+l(Rp+r-k+l
(8P
2 q+r+p -
p+r-k-
q+p r+2
A
'
+ (8P n P ^ ^ _,_. - R 1 R 1 )
1 p+q+r+1
q+p+1 r+1
p-2
8 (P
q+r+k+l R p - k
q+k+l P p + r - k )
k=0
+ 8(2P ^ ^ + P _^^ _ )
p+q+r
p+q+r+1
p+q
r+2
p+q+1
r+1
p-2
8 (p "q+r+k+l
= ^2 -^
k=0
R p
R
p-k
~ ^ - ^ ^
q+k+1 p+r-k
^ l j 0 - ( 2 R ^ R , n + R ^ R
+ R J ^ R
J
+ 8Pp+q+r+2
p+q r+1
p+q r
p+q+1 r - 1
and, by Theorem 3, we obtain
Theorem 70
'p-2
R R R =
p q r
2>
T?
"P
T?
p+r-k
q+k+1
k=0
p+q+r-1
2R
p+q
1
r+1
REFERENCES
1. Rodney T. Hansen, ''Generating Identities for Fibonacci and Lucas T r i p l e s , " Fibonacci
Quarterly, Vol. 10, No. 6 (December 1972), pp. 571-578.
2.
Carl E. Serkland,
Thesis,
Gauss, in his work on quadratic reciprocity, defined in [1] an analogue to the binomial
coefficients:
nri
(x11- DCx"-1- ! ) ( x n - k + 1 - l )
LkJ
(x k - l)(x k - X - ! ) (x - 1)
'
[!]'
for n = 0, 1, 2, , and
H-
for n < k.
It
ll
where
H-CO(0
is the usual binomial coefficient, then the above recursion formula becomes
130
functions.
This is easily seen from the recursion formula and mathematical induction.
[April
(See
[ 7 ] . ) The Gaussian binomial coefficients and their multinomial analogues have some interesting geometric interpretations and combinatorial applications in counting inversions and special partitions of the integers.
There is another well known analogue to the binomial coefficients, the so-called " F i b onomial coefficients:"
n I
k
n n-1
jF
n-k+1
F F
k k-r-- F i
Vh = \VF
= 1
(x
(x
F n
iw
- l)(x
k
- l)(x
-1
i\
n-k+l
...
- 1) (x
- 1)
k
" ! - 1) . . . (x*l - 1)
nl
_ In
= i
x-l,
ea - ( 4
Since
F
= F
n
(xFk+lFn-k
k F n - k - l _ i ) ( x F n - l _ -Q . . . ( x F n - k + l _ ^
F,
(x
t
(1)
F
+ F F
k+1 n-k
*k n - k - 1 '
(x
- l)(x
Fi -,
Fn
k 1
~ - 1) . . . (x X - 1)
F F + F F
FF
FF
k+1 n-k " V n - k - l
l n - k - 1 _, r k n - k - l
^
- x
+ x
- 1)
~
(x k - 1)
F F
F
F
/kVk-l(xFk+lFn-k _
(x
1}
- 1)
(x
F F
kFa-k-l _
x)
P Z IJF
fn - I"!
1974]
<F ,r
>(!>
-' -)[v] F
"FT
kFn-k- / ^
131
'F
n-k-1 ,
i
>
x
<F,^-DF.
V k
1 _
1=1
so that we have a recursion formula for the "Gaussian Fibonomial coefficients" and this, with
mathematical induction, implies the rather remarkable property of these functions: they are
polynomials rather than rational functions as they appear to be.
Furthermore if we let x
= F
(kj F
k+l^ n k j
+ F
n-k-l^k - i j j
the recursion formula for the Fibonomial coefficients. This is the recursion formula used in
[ 3 ] to prove that the Fibonomial coefficients are integers.
The more general sequence g n where g0 = 0, gt = 1, g f l + 2 = P - g n + 1 + q-gn> n >
p > 0, q > 0 ,
satisfies g Q = g k + 1 ' g n _ k
[]
as follows:
[]
a n d if w e
define
, sn-k+l
- nX1) (x
-nX 1)
Q'%k'%n-.k-l ^
- l1)W (x
^
n-1
(x " - l)(x
K l
See
t l)
- 1) . . . (x
- 1)
eL-H.-
(
+!
Y^
i=l
Again,
k+1
y^
(q g
x<Sk+1-i)'gn_k
" n-k + l- l) -g k
J fn - 11
j-n-1-
| _ k - 1.
0,
132
a r e polynomials.
April 1974
= f (t),
the
n+2(t) = 'W'*
+ f (t)
n >
Since the Pell sequence can be generated as a special case of the Fibonacci polynomials
(where t = 2), the above "coefficients" are polynomials also when defined in t e r m s of the
Pell sequence.
F u r t h e r m o r e , because of the direct analogy between the definitions of the Gaussian b i nomial coefficients and the related Fibonacci analogues defined above and the expression for
the binomial coefficients as ratios of factorials, the polynomials when arranged in a triangul a r a r r a y like P a s c a l ' s Triangle will have the beautiful hexagon property described byHoggatt
and Hansell in [4], that the product of the elements "surrounding" an element in the a r r a y is
a perfect square and the set of six elements can be broken down into two sets of three, the
products of the elements in each set being equal. In fact all the perfect square patterns of
Usiskin in [8] will appear in these new a r r a y s ; the proofs c a r r y over directly.
REFERENCES
1. L. Carlitz, "Sequences and I n v e r s i o n s , " Duke Math. J. , Vol. 37, No. 1 (Mar. 1970),
pp. 193-198.
2.
3. V. E. Hoggatt, J r . , "Fibonacci Numbers and Generalized Binomial Coefficients," Fibonacci Quarterly, Vol. 5, No. 4 (Nov. 1967), pp. 383-400.
4.
V. E. Hoggatt, J r . , and W. Hansell, "The Hidden Hexagon S q u a r e s , " Fibonacci Quart e r l y , " Vol. 9, No. 2 (April 1971), p. 120.
Z. Usiskin, "Perfect Square Patterns in the Pascal T r i a n g l e , " Math. Mag. 46 (1973),
pp. 203-208.
Recently at the 78
Montana (August 20-24, 1973), Professor P . Erdbs and Professor E. G. Straus proposed the
following classical problem to this author: Consider four digits where each digit can have a
value of 0, 1, 2, , 9. Divide the four digits into four sets where each set contains three
digits in the following way: Set A = 1st, 2nd, 3rd digits;
set C = 1st, 3rd, 4th digits; and set D = 2nd, 3rd, 4th digits.
tains the four digits 3742 then 374 would belong in set A,
000, 001,
002, , 999, without repetition? (It i s , of course, evident there will be four digits in each
and every cell of the 1000 cells.) This author has solved the above problem and we a r e able
to construct for the first time orthogonal triples in four
10 X 10 X 10 superimposed Latin
Cubes.
Note.
With the method of construction shown in this paper, we are also able to con-
superim-
7630
6861
3405
2793
1152
0796
2633
1972
4544
6321
5017
6971
5407
8639
0016
3795
4324
9408
8549
2013
1632
4284
7150
5547
0326
9978
7280
9868
8409
3155
9548
2153
1282
7860
6791
0976
3865
5327
3975
8019
5797
4404
4014
2323
0286
7540
6151
9638
1862
5287
4974
9328
7400
8869
3635
0156
1792
2543
6011
3545
1322
0866
9288
7010
2973
5637
6401
4154
8799
8159
7790
6281
3325
5977
0406
2863
4634
9018
1542
4864
3015
5157
8979
0546
9798
1402
7320
6631
2283
1012
9158
4794
5867
2403
6541
8329
3285
7970
0636
133
134
[April
Table 2
Square Number 1
8721
5386
6649
9850
1854
9720
4007
5006
0648
3722
2645
3212
9470
4727
7163
8931
9590
1164
8211
5936
2725
4387
0168
7003
2595
8641
3382
6729
6219
4597
1384
2165
8471
3932
8851
5166
6599
0008
7383
6479
0938
3002
4477
2935
7853
0388
4937
3162
7213
5596
1474
6859
7473
0218
1594
2005
0478
8161
7593
2215
2385
3642
6939
9930
4167
8381
5856
1004
6389
0598
6009
3472
0858
7643
1934
4857
9210
5476
9000
0728
5646
7933
3852
1644
9380
7723
2475
4217
1214
2855
4647
8591
5726
9160
9640
5216
3592
6169
8001
1724
Table 3
Square Number 2
5902
3244
1718
0139
9086
4650
8895
2371
6463
7527
6133
0909
9526
8375
3464
2891
5652
7247
4710
1088
3524
2711
4900
6893
1138
8464
7377
0089
9656
5242
7717
4370
0899
9906
8655
5082
3134
6523
1248
2461
0469
6653
5372
3084
7907
9246
1528
4890
2131
8715
2651
8525
7467
5712
4240
1908
6083
9136
0379
3894
1378
9466
6243
7657
5892
0529
2901
3714
8085
4130
4080
5132
3654
1468
2521
6713
0249
8905
7897
9376
8245
1898
2081
4520
6373
7137
9716
5462
3904
0659
9896
7087
8135
2241
0719
3374
4460
1658
5522
6903
2026
5948
1755
3181
8664
Table 4
Square Number 3
9873
4509
7232
6317
0490
3311
6877
0660
5758
4189
1945
9023
8504
2236
7492
4669
1235
2876
3941
7312
5188
8754
6497
0020
9503
8234
2756
6947
0870
5028
9493
4319
3661
7502
1185
6187
3021
9753
4499
8874
0500
7662
2946
1315
5238
1025
5668
8184
9233
2506
7872
3491
0310
6757
4949
7752
0180
3501
8024
9943
6667
1875
4239
5498
2316
2496
9313
4029
7182
1665
3231
6507
5878
8944
0750
5508
7942
1495
2666
3751
8314
0230
9183
4879
6027
0940
8494
5318
1505
6237
4759
2186
7022
9663
3871
1974]
FOUR 10 X 10 X 10 S U P E R I M P O S E D LATIN C U B E S
135
Table 5
Square Number 4
0064
2417
4551
8278
9272
8068
6785
3629
2787
7556
5063
9102
6345
5993
3109
7626
9832
1780
2837
7106
0994
4271
3839
1620
1410
5553
4341
8348
6995
0414
7836
1550
5623
8108
6065
3999
0344
2277
9782
4411
8838
9992
0624
2347
1060
6415
4781
5103
7276
3559
7996
3789
1830
0554
5413
4061
9342
6275
8628
2107
4621
6835
9412
1990
0104
8788
7066
2557
3349
5273
5343
0274
2997
4831
7786
9552
8418
3069
1100
6625
3419
4101
7346
5783
9622
1270
6555
0834
2067
8998
6105
1340
3279
7416
8558
2627
5833
4991
0784
9062
Table 6
Square Number 5
4255
1693
5880
3426
7574
6308
2762
9039
8917
0141
8427
3256
7144
2032
1913
9769
4305
0691
6888
5570
1143
9889
6258
8767
5420
2912
0031
3576
7304
4695
0881
6038
3766
7254
2302
4575
1423
8147
5690
9919
3916
8307
4035
1573
0251
7694
5140
6768
9429
2882
9309
2142
0911
4885
6698
5250
8577
7424
3036
1763
5030
7914
8697
0301
4765
3146
9259
1883
2572
6428
7761
7034
6578
4425
1303
5910
9149
8887
3696
2252
2692
5760
9579
6148
8037
0421
7884
4915
1253
3306
7764
0571
2422
9699
3886
1033
6918
5300
4145
8257
Table 7
Square Number 6
6446
0122
2364
7985
8613
1777
9531
3800
4058
5299
0052
3530
6776
5129
1367
2614
2984
9051
5809
7615
8773
6126
2124
3050
4988
7445
8293
9801
0292
3360
1447
4538
5369
1807
7535
8443
9771
6616
0982
4298
7055
4778
6806
0612
5449
8123
2294
1537
3980
9361
3770
6366
1127
2444
4618
8983
7805
0532
9291
5059
2804
8053
4128
5779
6536
7295
3440
0362
9611
1987
1617
6986
0772
2054
3290
4368
7125
9441
5539
8803
9121
2534
3610
1297
4808
5989
8363
6056
0442
7775
8533
5619
9981
3120
7365
0802
1057
2774
6296
4448
136
[April
Table 8
Square Number 7
3397
7738
0173
1041
2829
9514
6680
8962
5205
4456
5045
1391
2459
6960
7208
8682
3517
4736
9174
0823
7458
8172
9394
5685
0043
6200
4966
1821
2519
3737
4176
9964
1681
2399
6510
3827
7048
5455
0733
8202
1201
5515
3967
7828
4396
2739
0453
9684
8042
6170
8512
6450
4206
3177
9734
0393
5825
2049
1961
7688
0963
2209
5735
4516
3687
1451
8392
7178
6820
9044
9824
3047
7518
0203
8452
5175
1731
6390
4686
2969
6730
0683
8822
9454
5965
4046
2179
3207
7398
1511
2689
4826
6040
8732
1171
7968
9204
0513
3457
5395
Table 9
Square Number 8
2118
8950
9097
4562
5701
0845
1226
6484
7679
3333
7569
4112
5331
1486
8670
6224
2848
3953
0095
9707
8330
6094
0115
7229
9567
1676
3483
4702
5841
2958
3093
0485
4222
5111
1846
2708
8560
7339
9957
6674
4672
7849
2488
8700
3113
5951
9337
0225
6564
1096
6844
1336
3673
2098
0955
9117
7709
5561
4482
8220
9487
5671
7959
3843
2228
4332
6114
8090
1706
0565
0705
2568
8840
9677
6334
7099
4952
1116
3223
5481
1956
9227
6704
0335
7489
3563
5091
2678
8110
4842
5221
3703
1566
6954
4092
8480
0675
9847
2338
7119
Table 10
Square Number 9
1589
9075
8926
5604
3268
7431
0357
4193
2740
6812
2600
5584
3818
0197
9745
4353
1439
6072
7921
8266
9815
4923
7581
2350
8606
0747
6192
5264
3438
1079
6922
7191
5354
3588
0437
1269
9605
2810
8076
4743
5744
2430
1199
9265
6582
3078
8816
7351
4603
0927
4433
0817
6742
1929
7071
8586
2260
3608
5194
9355
6432
1359
2814
4583
9925
0267
7601
6352
3198
8196
3748
2070
7261
1609
9435
8746
4813
2920
5074
0587
0077
8356
4263
7811
2190
6602
3928
1749
9585
5434
3358
6262
0607
4073
5924
9195
7741
8436
1819
2580
1974]
137
Proof that Construction is Correct. Before going on with the proof, we will set down a
few definitions to facilitate our explanation of the proof.
Tables 1-10 a r e labeled Square 0 through 9. Then suppose we wish to find a certain number
of a certain cell we shall write S (row number, column number, square number) = number in cell. To find a row on a certain square, we write S (row number, * , square number),
and S (* , c, s) = column number on a certain square.
The ten columns in each square a r e considered to be numbered 0, 1, , 9 from left
to right; the ten rows on each square a r e considered to be numbered 0, 1, , 9 from top
to bottom.
For example: The number 7630 on Square Number 0 = S(0,0,0); o r the row on
7630 in is
S(* , 0, 0). Finally if we refer to a specific square, say square 0, we write S(* , *, 0); if
we refer to each and every one of the ten squares we write S(*, *, A); to refer to each and
every top row (say) in each and every one of the two squares we write S(0, * , A).
(1) We now consider the 2nd and 3rd digits in each cell of the S(0, * , A), and keeping
the cells in the same positions, we place S(0, * , 0), on top of S(0, * , 1), e ,
on top of
S(0, * , 9) it is easily verified that we have constructed the following 10 X 10 square which
was formed by superimposing two Latin Squares in such a way that the 100 two-digit numb e r s a r e mutually orthogonal.
(la)
63
86
40
.-
97
72
_
38
_
64
_
--
00
58
07
92
--
81
(lb) It should also be noticed that the 2nd and 3rd digits in each cell of the 8(0, * , A)
is repeated ten times in its own respective square.
3rd digits in S(0, * , 0) a r e 63 86 40 97, and it is seen that in the Square 0, the number
63 is repeated (as a 2nd and 3rd digit) ten times in a different row and a different column,
the number 86 is repeated (as a 2nd and 3rd digit) ten times in a different row and a different column, , the number 97 is repeated (as a 2nd and 3rd digit) ten times in a different
row and a different column.
(lc) Now it is easily verified; each and every one of the ten Squares is constructed in
the exact way we constructed the Square Number 0 in (lb).
(2) We now look at the first digit in each cell, where it is easily verified that the first
digit in each cell of the S(0, * , A) is repeated ten times in a different row and different column on its own respective square.
(2a) For example: the first digit 0 on Square 0 will be found in ten different cells
where each cell is in a different row and different column, and this exact arrangement of the
first digit 0 is constructed into each and every square 0 through and including Square 9.
It is also easily verified that each first digit 0 is on a different file.
[April
138
(2b) Now, each and every first digit (0, 1, , 9) in every cell is arranged in the
exact way we placed the 0f s in our example (2a).
Therefore, there are no two identical first digits in the same row, the same column,
o r the same file throughout the cube.
(Let the 100 numbers 000, 001, 002, . . , 099 = OQ ;
the 100 numbers 100, 101, 102, , 199 = a.t ;
the 100 numbers 900, 901, 902, , 999 = a 9 .)
Now, combining (1, a, b, c) with (2, a, b) leads to
(3) The first three digits in each cell in the cube that belongs to
a,
0,
then
file,
1, ,
then 9.
(3a) In (3), we have then satisfied the requirement that set A (set A = the 1st, 2nd,
and 3rd digit in each and every cell throughout the cube) will contain (in some way) every one
of the 1000 three-digit numbers 000, , 999, without repetition.
(3b) We now combine in each cell throughout the cubethe second and third digits with
the fourth digit and in the exact way we found (3a) we find that we have satisfied the r e quirement that set D (set D = the 2nd, 3rd, and 4th digit in each and every cell throughout
the cube) will contain (in some way) every one of the
000, ,
8.
It should
also be noticed that the ten first digits (say 1st digit = A) paired with ten fourth digits (say)
B to get the numbers AB in ten cells on a particular square shall never again have
this particular first and fourth digit combination repeated (that i s , the combination AB) on
any one of the nine remaining squares.
is
paired with the fourth digit 0, on Square 1 the first digit 7 is paired with the fourth digit
3, 'm , on Square 9, the first digit 7 is paired with the fourth digit
1.
This arrangement
for first and fourth digits is ridgidly enforced throughout the construction.
(5) Now, the first and second digits in each square (we consider one square at a time)
are mutually (pairwise) orthogonal.
10 X 10 Latin Squares.
(5a) The exact orthogonal properties of digits 1 and 2 in each of the ten squares (we
consider one square at a time) that we find to hold true in (5) also are easily verified to hold
true for the first and third digits,
That i s , the first and third digits in each and every one of
the ten squares (we consider one square at a time) are mutually (pairwise) orthogonal.
1974]
139
(6) Now, we combine (4) and (5), which leads us to the fact that set
(set B = 1st,
2nd, and 4th digits in each and every cell throughout the cube) will contain (in some way)
every one of the 1000 three-digit numbers 000, , 999, without repetition.
(6a) Finally, we combine (4) and (5a), which leads us to the fact that set C (set C =
1st, 3rd, and 4th digit in each and every cell throughout the cube) will contain (in some way)
every one of the 1000 three-digit numbers 000, , 999, without repetition.
Remark.
We used The Arkin-Hoggatt method (1] to get the 100 mutually orthogonal
numbers in (1).
Note.
methods with Bose, Shrikande and P a r k e r ' s work on mutually (pairwise) orthogonal numbers
[ 2 ] and after the proper extensions of their magnificent theorems it is easily shown that we
can obtain a solution of orthogonal triples in four
superim-
7540, 7320, 7970, that are found in Square Number 0 will be used as an illustrative example.
It is evident that in each of the ten numbers above, the first and faurth digits form the
two-digit number 70, and also the second and third digits in the above ten numbers are
mutually (pairwise) orthogonal.
(7a) Now, let us add a fifth digit to each of the ten four-digit numbers written above.
It is evident that it would be impossible to form orthogonal triples if any two of the ten fifth
digits we placed a r e identical.
7630 and
7280.
We then have
76300 and 72900 and it is evident that the 700 in 76300 and the 700 in 72800 a r e not in
a set of orthogonal triples.
Therefore, every one of the ten fifth digits we add to the ten
numbers in (7) above must be different and thus the fifth digit in (7) must include each number
in 0, 1, , 9. However, since the second and third digits in each of the ten numbers in (7)
a r e mutually (pairwise) orthogonal, it follows that the second, third, and fifth digits in the
above ten numbers in (7) a r e mutually (pairwise) orthogonal.
Then, using the exact method of our example in (7a) we extend our reasoning (step-bystep) to include the entire Square 0, and then Square 1, , and Square 9.
In this way,
(4m + 2) X
April 1974
It should be noted that the methods of construction of the cube in the above
paper are the same methods that were used to construct the cubes in the following two pap e r s (we mention the following two p a p e r s , since each paper stated that a method of construction was forthcoming).
See [ 3 ] and [ 4 ] ,
REFERENCES
J. Arkin and V. E. Hoggatt, J r . , "The Arkin-Hoggatt Game and the Solution of a C l a s s i cal P r o b l e m , " J. of Recreational Math. , 1973 Spring Edition.
R. C. Bose and S. S. Shrikande, "On the Falsity of E u l e r ' s Conjecture About the Nonexistence of Two Orthogonal Latin Squares of Order 4t + 2 , " P r o c . Nat'l. Acad. Sci. ,
45 (5), 5/1959, pp. 734-737.
E. T. P a r k e r , "Orthogonal Latin S q u a r e s , " P r o c . Nat'l. Acad. Sci. , 45 (6), 6/59, pp.
859-862.
J. Arkin, "The F i r s t Solution of the Classical Eulerian Magic Cube Problem of Order
T e n , " Fibonacci Quarterly, Vol. 11, No. 2 (April 1973), pp. 174-178.
J. Arkin, "A Solution to the Classical Problem of Finding Systems of Three Mutually
Orthogonal Numbers in a Cube Formed by Three Superimposed
1 0 X 1 0 X 10
Cubes,"
Math.
three
s u c h a r t i c l e s , appearing on p a g e s ^ g ^ 167
196
1.
INTEODUCTION
The object of this paper is to present a new combinatorial interpretation of the Fibon
acci numbers.
There a r e many known combinatorial interpretations of the Fibonacci numbers (e. g. ,
[9 ]); indeed, the original use of these numbers was that of solving the rabbit breeding problem of Fibonacci [10]. The appeal of this new interpretation lies in the fact that it provides
combinatorial proofs of several well known Fibonacci identities. Among them:
(:' 3 \JIF.3 - *^ 2 n
3=0 \
These results will be presented in Section 2. In Section 3, we shall describe further p o s s i bilities for exploration of Fibonacci numbers via combinatorics.
2.
FIBONACCI SETS
Fibonacci
*. {i}
*. (i}> {2}
3
4
<P,
0 , { l } 5 { 2 } , { 3 } , { 4 } , { 2 , 3 } , { 2 , 4 } , {3,4}
r-Fibonacci
1-Fibonacci
2-Fibonacci
0, {2}
(P
</>, {3}
0, {3}, {4}
142
[April
for n ^ r - 1.
Proof.
When n = r - 1 o r r ,
Since F 1 = F 2 = 1, we see
1 from each set in the second c l a s s , we see that we have established a one-to-one c o r r e s pondence between the elements of the second class and the
{L l , 2, , nUJ
0 ) , hence there a r e F
that there a r e
(r + 1)-Fibonacci subsets of
" E n 0 +2-r
F
This means
n 0 +2-(r+l)
(no+l)+2-r
F o r n c 0,
n+2
= 1
Proof.
(")
singleton Fibonacci subsets of { l , 2, , n } . The two-element Fibonacci subsets a r e just
the two-element subsets of { 2 , 3, , n} , and there a r e
(v)
of these. In general, the j - e l e m e n t Fibonacci subsets of { l , 2, , n} a r e just the j element subsets of { j , j + 1, * , n} and there a r e exactly
(-r)
of these.
1974
143
n+2
Theorem 3. For n a 0
n+1
2n+2
or
E(r)
Remark.
Proof.
n+l-j
2n+2 '
, 2n} 9 for if there were n + 1 elements then at least one element would be ^ n which is
impossible.
Let T. denote the number of Fibonacci subsets of ( l , 2, - , 2n}
j elements ^n.
Clearly
n
F
2n+2
3=0
j elements a j-Fibonacci
(T)
choices of the j elements from
1 v + 2 _.
^ ='
iVn
Therefore
n
F
Theorem 4.
2n+2
For n ^ 0,
n
T
3=0
2-1 V
X
j=0
)Fn+l-j
F
n+1,
j-Fibonacci
144
[April
1 + F<
^ .
1 + FL2 + + F
n = Fn+2
Proof.
the largest element is j . Let R0 = 1 in o r d e r to count the empty subset <p. Clearly for
j
>
1,
Therefore
n
1=1
J=l
3.
CONCLUSION
The genesis of this work lies in the close relationship between the Fibonacci numbers
and certain generating functions that are intimately connected with the Rogers-Ramanujan
identities.
then
Indeed if D
and D (q) = D
(q) + q n D 0 (q),
n
nJ.
nz
(3.1)
D n (q) =
xj > 0 y [}-']
where
r
~i
-4-1
(1 - q11"3
_1
) ( l - qJ)
for
0 < m < n5
= 0 ot
It is not difficult to see that D (q) is the generating function for partitions in which each part
is l a r g e r than the number of parts and <n.
onacci subsets of ( 1 , 2, " , n ) , and this is clear from (3.1) and Theorem 2 since
[m]
e( uals
fmj
Namely,
n+1
(3.2)
D2n(q) = X V [ n r ] D n - l - ^ .
]=0
and
n
(3.3)
Dn(q) = 1 +2^
q JD j_ 2 (c|)
j=l
While (3.3) is a trivial result (3.2) is somewhat tricky although a partition-theoretic analog of
Theorem 3 yields the result directly.
1974]
145
Since I>n(q) is also the generating function for partitions in which each part is ^ n and
each part differs from every other part by at least 2, we might have defined a Fibonacci set
in this way also; i. e. , a finite set of positive integers in which each element differs from
every other element by at least 2. Such a definition provides no new insights and only tends
to make the results we have obtained more cumbersome.
= F
n+m
[12; p. 7]
*-lF
F
+ F F
n-1 m
n m+1
E( 2 A ,y
first one to extensively develop such formulas [11] (see also [2], [7]).
REFERENCES
1. George E. Andrews, "Some Formulae for the Fibonacci Sequence with Generalizations,"
Fibonacci Quarterly, Vol. 7, No. 2 (April 1969), pp. 113-130.
2.
Vol. 8, No. 1
George E. Andrews, "A Polynomial Identity which Implies the Rogers-Ramanujan Ident i t i e s , " Scripta Math. , Vol. 28 (1970), pp. 297-305.
146
4.
April 1974
George E. Andrews, "Sieves for Theorems of Euler, Rogers and Ramanujan, from the
Theory of Arithmetic F u n c t i o n s / ' Lecture Notes in Mathematics, No. 251,
Springer,
(to appear).
Vol. 8,
W. B. Saunders,
Philadelphia, 1972.
11.
The formulas pertain to generalized Fibonacci numbers with given TA and T 2 and with
(1)
T ^ = T + T n
n+1
n
n-1
and with generalized Lucas numbers defined by
(2)
V = T _,_, + T n .
n
n+1
n-1
Starting with a finite difference relation such as
(3)
A(b
/a)kT2kT2k+2
"
(bk ak+1
>T2k+2(bT2k+4 "
aT
2k>
values of b and a are selected which lead to a single generalized Fibonacci or Lucas number for the term in parentheses.
3T .
(4)
Thus for
b = 2, a = 13,
E (2/l3)kT2kT2k+5 k=l
(l/3)[(2n+1/13n)T2n+5T2n+7-2T5T
L
INTRODUCTION
The Morgan-Voyce
<Ri
V = JR.
Now
V = IjRt = I2R2 = Ci + I 2 )B
Thus
V ~
y - R2
i'
so that
Ri
R2
'
1R = R +R
t
For r e s i s t o r s in s e r i e s
-AAAvr-i
A I
V = I(Ri + R 2 ) = IR
147
Let
148
[April
LADDER NETWORKS
x
- W ^
x
VW
x
r--Wv
1
Si
| i
S i
Bo
x
\A/v
L
n sections
x
*a * A A V -
\ i
"TTzo
-J
o D
Assume that the terminals A and B a r e open. We desire the resistance as measured a c r o s s
terminals C and D. Fc
F o r n ladder sections, let us assume that the resistance i s Z , and
n
consider the output Z .
0
x
-'Ws*
z
<
n f
* CT
f
i <
f
z a,
n+1
*
D'
Since x and Z a r e in s e r i e s ,
n
R = x +Z .
n
Now R and 1 are in parallel, so that
x + Z +1
Z ^
n+1
x +Z
n
n+1
+l -
x + Z
n
x + Z
n
x +Z +1
n
'
To see what this means, l e t Z = b (x)/B (x), where b (x) and B (x) a r e polynomials.
Wx)
x + b (x)/B (x)
n
n
xB (x) + b (x)
n
n
1974]
149
so that
|Vl(x)
(2a)
= -Bn(x)
b n (x)
Clearly,
Z a = 1,
so we set b0(x) = 1 and B0(x) = 1. This completely specifies the two sequences which we
call the Morgan-Voyce polynomials.
Without too much trouble, one can derive that both sequences
satisfy
(2.2)
This takes care of the resistance as seen from the output end of the ladder network.
We now
^V"v
*VVv
VW~j
Vi
o
1 , 1
Z + T1 9,
n
Again let Z
= P (x)/Q (x).
_
R = Z
^o r',
xZ
n
_
Z + 1
n
o D
n
+ 1
n
Z
7
=
n+1
' * * *
+ x + Z
n
n
Z
+ 1
n
Then,
P n (x)
Qn(x)
'
That i s ,
P n + 1 (x) = (x + l)P n (x) + xQ n (x) ,
Qn+l(x) =
n(x>
Qn ( x )
'
Simplifying,
P
n(x)
n+l(x) -
n(x)
Q2(x) = x + 2, so that
150
[April
-b-*W
z
Z
n ~
B (x)
n
'
b n (x)
B (x)
n
x + 1
x + 2
x 2 + 3x + 1
x'1 + 4 x + :3
x 3 + 5x2 + 6x + 1
3
4
X 3 + 6x + lOx + 4
x + 7x + 15x + lOx + 1
+ 8x 3 + 21x 2 + 2 Ox + 5
f (x)
n
X
2
x 3 + 2x
4
hA
A
h
AA
hA y
+ 1
3
4
+ 3x
y1
y 1
4
2
X 5 + 4X3 + 3x
+ 1
X6
5
31
X 7 + 6x + 10x + 4 x
+ 5x 4 + 6x 2 + 1
h'
Ay
6
7
/ l
.6
/ 4
10
10
15
20
15
21
35
35
21
1974
Wx> =
xB
{ B (x)},
and
-i ( x 2 )
f 2 n + 1 (x) = b n ( ^ )
Proof 1.
{b (x)},
151
By Generating Functions.
= y \ (x)xn
1 -A
1 - (x + 2)A + A2
JL*d n - 1
1 - (x + 2)A + A2
Hence,
X(1 A2)
2
1 - (x
~+ 2)A
+ A
A2x
2
- E*-*!b > 2 ) x 2 n + i
n
n=0
_ \ ^
2
1 - (x + 2)A + A
. ? X x2n
Z>n-1^2
n=0
1 _ 2A2 +A 4 - x2A2
1 - xA - A2
Efn(x)Xn
n=Q
where we recognized the generating function for the Fibonacci polynomials (f (x)}.
Proof 2.
By Binet F o r m s .
a =
^x2
+ 4
= x - N/X2 + 4
it can be shown by mathematical induction that the Fibonacci polynomials have the Binet form
fn(x) = (<*n - j3n)/(<* - P) .
152
[April
Similarly, from the recurrence relation for the Morgan-Voyce polynomials, we have the auxiliary equation
Y2 = (x + 2)Y - 1
with roots
+ 2 + ^ x 2 + 4X
x + 2 - N/X2 + 4x
3n
x2 + 2 + XN/X2 + 4 \ n
(X2) = f / x 2
/ - (/3 2 ) n ]/(a - p)
/x2
) "(
+ 2 - x ^x 2
2 H- ^ x 4 + 4 x 2 \ n _ / x 2
2 - ^x4
+4 \n1
An
) J / ^+ 4
4x2\n1 A C T
+ 4xz
N/X4 + 4X 2 = jx|\/x 2 + 4 ,
xB
,(x 2 ) = f_ (x) .
n-1
2n
= xB ^.(x 2 ) - xB (x2)
n+1 '
n '
=
2n+4 ( x ) - f 2n+2 ( x )
rf
2n+3W
'
leading to
bn+1
a.i( x2 ) = fo2n+3
xoW
bn (x2) = f02n+l
^(x)
or
bi(x) = x + 1,
f l( x)
f3(x) = x 2 + 1,
= 1,
n + 2
(x + 2 ) b n + l ( x ) " b n ( x )
Thus,
b 0 (x 2 ) = 1,
Now, the sequences \b
bi(x 2 ) = x2 + 1,
(x 2 )} and {f
b n + 2 (x 2 ) = (x2 + 2)b n + 1 (x 2 ) - b n (x 2 ) .
(x)} have both the same starting pair and the same
Bi(x) = x + 2 ,
f2(x) = x,
f4(x) = x 3 + 2x,
f 2 n+6 ( x )
n+2
(x)
=
=
Similarly,
(x + 2)B
n+l(x) "
(x2 + 2)f
.n ( x )
2n+4 ( x ) " f 2n ( x )
'
1974]
153
Next,
xB 0 (x 2 ) = x,
xBi(x 2 ) = x 3 + 2x s
- xB (x 2 ) ,
Several results follow immediately by applying known properties of the Fibonacci polynomials.
(See [ 3 ] , [ 6 ] , [7].)
Corollary 3.1.1.
b (1) = F Q _L1
n
2n+l
and
B ,(1)
= F0
n-lv
2n
prime.
4.
Let
x + 2
-1 \
OJ
Then
Q2
x + 2
1
- l \ = / x 2 + 4x + 3
0 ) - y
x + 2
-l\
/x + 2
0J- y 1
-(x + 2 ) \
-1
j
/B 3 (x)
-B2(x)\
\B 2 (x)
-Bt(x)J
'
n
Q
= (det Q)
/
\
n+l(x)
B
~VX> \
n(x)
-Vl
( x
,
B
n+l<x)Bn-l(x) -
nW
= "X
Thus, one can write much by virtue of having B (x) trapped in a matrix.
Let
R
/x + 2
y 2
-2
\
-(x + 2) J '
ROn_/
KW
n+l^
^ C n (x)
~Cn^\
-Cn_1(x)j '
so that
C n + 1 (x)C n _ 1 (x) - C^(x) = -(x 2 + 4x + 4) + 4 = -(x 2 + 4x)
2n + 1 is a
154
[April
Actually,
L (x) = f , - (x) + f - (x) ,
n
n+1
n-1
1, 3, 4, 7, 11, 18,
29,
Now, C ^ x ) = 2, C0(x) = 2, Citx) = x + 2. Thus, since
L 2 n + 4 (x) = (x* + 2)L 2 n + 2 (x) - L 2 n (x) ,
2
we have L 0 (x) = C . ( x ) ,
2n
n-1
so, since
L
2n(x)
and C
2n+l(x)+f2n-l(x)'
1 (x
and
2n+l(x)
Vx2)'
) implies that
C (x) = b (x) + b ^ ( x ) .
n
n
n+1
Also,
xB (x) = b ^ ( x ) - b (x) ,
n
n+1
n
'
so that
b n + 1 (x) = [Cn(x) + xB n (x)]/2 .
Finally, applying the divisibility properties of Lucas polynomials [6] , [8], [9], we
have the
Theorem.
(x) is irreducible.
2n
5. ATTENUATION RESULTS
The attenuation is the ratio of input voltage VT to output voltage V n . Since the system
is linear, we can assume that the output voltage is IV. Let us s t a r t with no resistive network.
There is no current (I
1 volt
Ij = 1 amp
So we see that
Vt = x + 1 volts
19^4]
v 0 = 1 = b_ 1 (x) ,
Ii = 1 = B0(x) ,
Vj = 1 = B0(x) .
155
and
Vn = b n - -l v(x) .
n+l (x) = V x )
Vl ( x )
unit resistance is
V ; hence, the
current is also V .
n
_ui
n+1
O
,7
^ A A
X
"
11+1
-i
" V N A
< 1Q
j
<
_J
Vs/s-
-t & * s
<% a $ s
= B
Jx)
V
and V
+l =
xB
T n = V +1
n+1
n
n
= b (x).
n(x)
V l = b n (x)
applying Lemma l to the expression for I
+ b
Then,
n(x) =
n+l(x)
V l ( x ) = B n (x)
+1,
'
'
We note that
V
B
n+1
n(x) = \
n+lW
Vl
X B
t n(x)
'
+ B
+
n-l(x)
Vo = b n (x)
''"
+
+ B
o(x)
b ^ W
b0(x)
156
April 1974
REFERENCES
1. A. M. Mo rgan-Voyce," Ladder Network Analysis Using Fibonacci Numbers," IRE T r a n s actions on Circuit Theory, Vol. C T - 6 , Sept. 1959, pp. 321-322.
2.
3. Richard A. Hayes, Fibonacci and Lucas Polynomials, M a s t e r ' s Thesis, San Jose State
College, San J o s e , Calif., 1965,
4.
M. N. S. Swamy, " F u r t h e r P r o p e r t i e s of Morgan-Voyce Polynomials," Fibonacci Quarterly, Vol. 6, No. 2, Apr. 1968, pp. 167-175.
There are 625 formulas in all arranged in categories according to the difference r e -
1.
INTRODUCTION
In a previous paper [4] I found that two theorems of Hermite concerning factors of b i nomial coefficients might be extended to generalized binomial coefficients [2] , however one
of my proofs imposed severe restrictions on the sequence {-A } used to define the generalized coefficients-
Also it was found that the Mann-Shanks primality criterion [6] follows
from one of the Hermite theorems and it appeared evident that the criterion also held in the
Fibonomial a r r a y , but the proof was not completed.
In the present paper I remove all these defects by proving the Hermite theorems in a
m o r e elegant manner so that very little needs to be assumed for the generalized a r r a y , and
the Mann-Shanks criterion is not only proved for the Fibonomial a r r a y but for the s-Fibonomial and q-binomial a r r a y s .
2.
<a i\
A A , A
'" n" n - 1
n-k+1
.,,
jn)
A,
= k
identically.
P r o p e r t i e s of the a r r a y and their history may be found in [2]. Our attention here is fixed on
the case when these coefficients are all integers. Arithmetic properties a r e then of p r i m a r y
concern. As usual,
and a | b
(2.2)
fn\
\
and
A
(2.3)
n-k+l
(A
{:
, A )|A
have (An ,A.K ) ' A n , so that (2.3) is only slightly l e s s general than (2.2).
In [4] I stated that (2.3) holds provided A 1 - A, = A
, or something close to
this.
158
[April
Proof of (2.2), By the Euclidean algorithm we know that there exist integers x and y
such that (A , A. ) = xA + JyA. .
n
k
n
k
Therefore
(vv{ n k J = *\{ k } + ^ k { k } =
for some integer E.
Since (A , A, )|A
Proof of (2.3).
XAJJ} + y A n { j : ; }
= An.E,
x and y,
(A
- , A, ) = xA
- + yA, ,
whence
= xA ^ . in * H + JyA ^ . (.
n+l-k (
n+l-k [ k - 1 f
= A _,, . -F ,
n+l-k
(2-4)
n+l-k I n+1' k \ kf
, A, )|A - , we obtain (2.3).
The proof I tried in [4] motivated by Hermite T s own argument ran as follows: We have
(A
n+l' V
^n+1
yAk
= x(
\+l
" V
+ (x + y ) A
k '
whence
(A
- - A,
^ A k {J}
A ^ - A,
n+1
A
A
^ n+1-k{k}+(x
+ y)A
n+1-
(x +
(A
- - A. = A
-, A, )|A
+1_k
n + l-k{k-l}
and we shall have proved (2.4), but as seen in our general proof none of
k{k(
\ \
k - 1/ '
+1F
n+l' V
= F(n+l,k) = F(n+l-k,k) = (Fn+l-k'
V
which means that (F - , F, ) |F + 1 _ k - In any event, our results are obtained more elegantly
by our present proofs.
1974]
159
is an integer.
(9 f\)
U b
' '
(a,b)(a + b - 1)!
~
alb!
is an integer.
This follows at once from Hermite's original form of (2.2), because by putting
H(n,k) = M
\l\
|kf
H(a
v
b, b) =
(a + b b )
'
a +b
a + b
l = (a ' b)(a , + h , b -
1)!
a! bt
I was reminded of these things by a l e t t e r from Gupta [5] who remarked that a nice
(2.7)
is an integer.
= F
eralized factorials by
(2.8)
[n]l
= A n A n - 1 . . . A2Alf
with
[0] !
= 1 .
,
M
H (n,k)
(A 5
n V
= j _ n
fn\k
is an integer, so also is
(A
,A ) ,
, x
(A , , A ) [ m + n - 1] !
J
TT/ ,
\
m+n' n '
m + n _
m+n? n / L
H ( m + n , n) = x
<
\
r ,
r ,
A
f
n
}
m+n
[m] ! [n] !
,<, v
(2.9)
an integer.
According to Dickson [1, p. 265] Cauchy also proved Schonemann's theorem for (2.5),
and Catalan (1874) proved that (2.6) is an integer in case (a,b) = 1.
Catalan, Segner, Euler, etc., found that (n + 1) | <
arguments. See my bibliography [3] for a l i s t of 243 items dealing with the Catalan numbers,
ballot numbers, and related m a t t e r s . A supplement of over 75 items is being prepared.
The fact that (n + 1) | \ > follows at once from (2.3) so that the number
160
(2.10)
c(Q.k) =
( k
n+l-k
is a natural generalization.
[April
'
= F
(3.1)
f ~
nrt}
l \ C - 2 R /
for every integer R such that C/3 < R < C/2,
R > 1.
Here R and C a r e the row and column numbers, respectively, in the original Mann-Shanks
shifted binomial a r r a y .
= F
in (2.1))
That i s , we have
9T>\
R > 1.
Note that the single difference between this and (3.1) is that the row number R must be r e placed by the corresponding Fibonacci number
FR.
When
C = prime,
the divisibility
(F
R'
C-2R)
(R,C-2R)
(R,C)
proof is just a slight modification of the proof givenby Mann-Shanks. Suppose C = 2k, with
k = 0, 2, 3, 4, ; then the unit < ft ? = 1 always occurs in the column, so divisibility cannot occur, and it is sufficient to consider odd composite C.
of C, and write C = p(2k + 1), with k >. 1.
R-row and C-column is {
1
F .
pk
Choose R = pk.
> , and
/kp)
\ p J
F ;
pk'Fpk-l'Fpk-2
F..F-F
,
pk p p - 1
pk-p-fi
F1
1974]
Cancel F , with F . .
The factors
F _ , F
(F , F
) = F,
v = F,
v = F- = 1
p
p-r
(p,p-r)
(p,r)
1
while on the other hand F
161
for all
1 < r < p - 1 ,
y
(P,PN)
(P,J)
for
' -
and so F , which is g r e a t e r than 1 for odd p r i m e s p, cannot divide into the numerator.
This means, equivalently, that the row number
F ,
>.
= F
s being any
(4.1)
R >
That i s ,
1.
SR
Mann-Shanks type a r r a y .
(F
sR'
~ F .
n
sn
is a factor of the coefficient in the R-C position of the
S ^ ZS-t\
sC-2sR)
(sR,sC-2sR)
(sR,sC)
since C = prime implies (R,C) = 1 for each C/3 < R < C/2,
F /F
as a factor.
C=2k,
s(R,C)
R > 1.
R.
s '
k = 0, 2, 3, 4, .
unit < Q > = i occurs in the column; so that it is sufficient to study the situation for odd composite C. Let p be an odd prime factor of C, and put C = p(2k + 1), k ^ 1. Choose as
before R = pk. Then the coefficient in the R-C spot is the s-Fibonomial coefficient
We find now that
<
>.
162
Cancel F
and F
[April
(F
, F
sp
) = F,
v = F,
x = F .
x = F
(sp, s p - s r )
(sp,sr)
s(p,r)
s
sp-sr
, F
sp
. .) = F ,
. .v = F ,
., = F . .v = F
spk-sj
(sp,spk-sj)
(sp,sj)
s(p,j)
s
Remove the common factor F
/F
i SP
\ Fx ss '
V
\
sp-sr i
xF
ss /)
1,
throughout.
for all
1 < r < p - 1 ,
for all
1 < j < p - 1.
and
/ F^
Forar
.\
Also,
complete.
It would appear that a Fibonacci-type property (a homomorphism)
<4-2)
( A
a'V
= A (a,b)
a
b
(a b)
They are polynomials in q. Since in fact (q - 1, q - 1) = q
- 1, it is not surprising
now that we can a s s e r t the Mann-Shanks criterion for the q-binomial a r r a y . The q-analogue
of (3.1) is motivated by Hermite's generalized theorem (2.2) for we now have that the coefficient in the R-C position is divisible by
g - i
/ R
.,
C-2R
.,\
(q - 1, q
- 1)
which reduces to
1974]
163
q - 1
when C is a prime and C/3 ^ R ^ C/2,
Theorem 4.
That i s :
R ^ 1.
I C-2 E |
C/2,
R > 1,
C = 35.
Up to this point a row number fails to divide an a r r a y number because of the presence of a
unit in the column.
The next such numbers are 49, 55, 65, 77, 85, 95, corresponding to those numbers of form
6j 1 which a r e composite.
The column entries for C = 25 in the ordinary Pascal case are 36, 252, 165, 12, with
corresponding row numbers
10 fails to divide
252,
t r i e s are 714, 136136, 83215, 144, with row numbers 34, 55, 89, 144. Here
55 fails to
(q - l)(q - 1) '
(q - l)(q
- l)(q
- l)(q
(q12 - 1)
q - 1
x)
- l)(q - 1)
and
It is again, of course, the second row number that fails to divide the coefficient in the column.
For a r r a y s of the type we are studying this behavior is typical.
The column entries for C = 35 in the Pascal a r r a y a r e 12, 715, 3432, 3003, 560, 17,
with row numbers 12, 13, 14, 15, 16, 17. Here 14/3432, and 15J3003. For the Fibonomial a r r a y the entries are 144, 27372840,
row numbers 144, 233, 377, 610, 987, 1597, and the row numbers
ones which fail to divide their corresponding column entries.
377 and
164
[April
By placing units in the (R, 2R) and (R, 3R) positions in their rectangular a r r a y and
carefully choosing the other entries (which turned out to be binomial coefficients) Mann and
Shanks developed a kind of sieve which tests numbers of the form 6j 1 for primality.
This
suggests that there may be ways to devise similar sieves based on other arithmetic p r o g r e s sions. After all, it is a very old theorem of Dirichlet that if (a,b) = 1 then there are infinitely many primes of the form a + bt, where t ranges over the integers.
We might expect
4j 1 for
example. Although I have not found amy simple formula for gene rating the entries in an a r r a y ,
I can suggest some obvious n e c e s s a r y properties of such an a r r a y , by analogy with the original Mann-Shanks a r r a y .
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21
0 1
1
2
3
4
5
6
7
8
9
10
1-1
1 2-2 1
1 3 _*
1 *
_ 3
- 4
1
- 4
_ 5
- *
_ *
_ *
1 *
_ 5
_ 6
- 7
1
_ 5
_ 6
_ 7
_ 8
1
_ *
_ *
_ *
*
1 *
Numbers listed above are the smallest factors which an entry musthave in order to be allowed, so that the row number will divide each entry in a prime column.
However, of the
marked by a s t a r (*) must be chosen so that at least one of the s t a r r e d numbers in each column will not be divisible by the row number.
25, 27, etc.
One may imagine that it would be desirable to have a symmetrical row, in an-
9, 15, 2 1 ,
It would be very
remarkable if we could determine simple formulas for generating such generalized MannShanks a r r a y s based on Dirichlet progressions.
In the outline a r r a y based on 4j 1, it is easy to see that the bottom s t a r in the special
columns will always occur for row number
(K - l ) / 2 ,
where K = 4j 1 ^ prime.
If we
choose an entry for that position which is not divisible by the row number and otherwise fill
open spots in the a r r a y by the row number in any given row, we shall obtain the following
a r r a y having the Mann-Shanks property:
1974]
165
c 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10
1 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 1
where 4/fa,
7/fb,
lo|c,
= . . . = 1 throughout.
Theorem 5.
12/[d, etc.
a = b = c = d
Let an a r r a y be defined by
A(n,0) = A(n, 2n) = 1,
A (n,k) = n,
A(n,l) = n,
A(n,l) = x,
2 = k < 2n - 1 ,
T*C 1
if
n ^ 0 ,
n f ^ ,
with
n/fx,
where K = 4j 1 f- p r i m e ,
if
K - 1
Then this a r r a y has the Mann-Shanks property when shifted in the way of the original MannShanks a r r a y .
Similarly, the binomial coefficients in the original Mann-Shanks a r r a y maybe replaced
by numbers chosen in the same way.
Theorem 6.
We have
Let an a r r a y be defined by
A(n,0) = A(n,n) = 1,
A(n,k) = n,
A(n,l) = n,
A(n,l) = x,
if
n ^ 0 ,
2 ^ k < n - 1,
n /
with
n/x,
, where K = 6j 1 ^ p r i m e ,
if
K - 1
Then this a r r a y , when shifted as prescribed by Mann-Shanks has the Mann-Shanks primality
criterion property.
These two examples are not what we mean by a ? simple formula' of course, because we
must prescribe and know the prime nature of a + bt in advance, whereas the beauty of the
166
April 1974
binomial coefficients, Fibonomial coefficients, or q-binomial coefficients is that they automatically take care of the situation.
(iLT1)
read
[S-^-lJ ;
p. 360, lines 6 and 8 from bottom, for " E r d o s " read " E r d b s " ; p. 372, in Ref. 2, for "Institute" read "Institution. "
REFERENCES
1. L. E. Dickson, History of the Theory of Numbers, Carnegie Institution, Washington, D.C.,
Vol. I, 1919. Reprinted by Chelsea Publ. Co. , New York, 1952.
2. H. W. Gould, "The Bracket Function and Fontene-Ward Generalized Binomial Coefficients with Application to Fibonomial Coefficients," J^bo^cjcTQu^rter^y, Vol. 7 (1969),
No. 1, pp. 23-40, 55.
3. H. W. Gould, "Research Bibliography of Two Special Number Sequences," Mathematica
Monongaliae, No. 12, May, 1971, Morgantown, W. Va. iv + 25 pp.
4. H. W. Gould, "A New Primality Criterion of Mann and Shanks and its Relation to a Theor e m of Hermite with Extension to Fibonomials," Fibonacci Quarterly, Vol. 10 (1972), No.
4, pp. 355-364, 372.
5. H. Gupta, Personal Correspondence, 19 March, 1972.
6. Henry B. Mann and Daniel Shanks, "A Necessary and Sufficient Condition for
and its S o u r c e , " J. Combinatorial Theory, Ser. A, 13 (1972), 131-134.
Primality,
(D
n+1
(2)
(3)
Tn = A i r f + A2r? + A3rf
THE SEQUENCE S
=r?
Since the individual elements of these sums are powers of the roots, the sums obey the
given recursion relation.
by means of
symmetric functions and thereafter generate additional t e r m s of the S sequence. Since this
sequence is basic to all the algorithms, its generation constitutes the first algorithm. (Note.
This use of the S sequence is exemplified in [ 1 ].)
ALGOEITHM FOE FINDING THE TERMS OF S
Three consecutive terms of the sequence are:
Si = ax
S2 = a2x - 2a2
Then use the recursion relation to obtain positive and negative subscript t e r m s of the sequence,
The algorithm will be illustrated for two recursion relations which will be used to check
other algorithms numerically.
167
168
[April
E X A M P L E 1: x - x
30
29
28
27
26
25
24
23
22
21
20
19
-14429
13223
-3253
-4459
5511
-2201
-1149
2161
-1189
-177
795
-571
-18
-17
-16
-15
-14
-13
12
-11
-10
-9
-8
-7
47
271
-253
65
83
-105
43
21
-41
23
3
-15
-6
-5
-4
-3
-2
-1
0
1
2
3
4
5
EXAMPLE 2:
X3
11
-1
-5
5
-1
-1
3
1
3
7
11
21
--
6
7
8
9
10
11
12
13
14
15
16
17
39
71
131
241
443
815
1499
2757
5071
9327
17155
31553
2
7x
7x^ + 5x + 4 = 0
18
19
20
21
22
23
24
25
26
27
28
29
30
58035
106743
196331
361109
664183
1221623
2246915
4132721
7601259
13980895
25714875
47297029
86992799
S n <-4)n
-23
-22
-21
2450995949 6004997927 85
2879858678 8067714806 5
3383761613 1827843249
-20
-19
-18
-17
-16
15
14
13
12
11
3975834906
4671506147
5488902409
6449322392
7577792077
8903714463
1046164399
1229215792
1444301540
1697004500
10
-9
-8
-7
-6
1993985121
234271601
27532161
3232913
380577
-5
-4
-3
-2
-1
44465
5313
593
81
5
620902593
59541201
1011041
180465
14561
1313
5681
433
49
9
0
1
2
3
4
5
3
7
39
226
1359
8227
6
7
8
9
10
49890
302659
1836255
11140930
67594599
11
12
13
14
15
410112523
2488250946
1509681561 1
9159600445 5
5557349493 46
16
17
18
19
20
3371777360
2045738276
1241198527
7530649458
4569025826
703
0371
21698
07219
000559
21
22
23
24
25
2772137664
1681922475
1020462746
6191392481
3756466464
2081026
81335511
554941211
409586818
6767059627
26
27
2279138391 3410171845 5
1382807980 7792383837 78
recursion relation for a spacing of k among the t e r m s , namely, for the sequence T .
Since
(5)
nk+a
, A nk+a
AAt rtnk+a +, AA 2 rnk+a
+
A3r3
2
1974]
169
and since there i s a change of r, from one t e r m to the next, the recursion relation i s that
k x
whose roots correspond to r. . Let the coefficients be given in the relation
x 3 - Btx2
+ B 2 x - B3 = 0 .
Then
-r,
B
v ^ k k
2 = Erirj
k - k
= a9E r i
k_
= a
3 s _k
B3 = 4
Hence the recursion relation is given by
x 3 - S k x 2 + a!f S_ k - a!f = 0 .
(6)
r2
next as the m f s change by 1. The second type obey the recursion relation for the quantities
r.r..
1 J
ALGORITHM FOR THE SECOND-DEGREE FUNCTIONS
The recursion relation governing the quantities
given by:
(7)
x 3 - S 2 x 2 + asS_ 2 x - 4 = 0 .
170
[April
For the second we need to find the symmetric functions of the roots r. r . .
B
X) r i r j
B 3 = r? r? r? = a3
(8)
- a3 = 0 .
(9)
- a|) = 0 .
mx
m2
m3
nii+mo m
i
j *
mi m ? mo
i
j
k
The first type corresponds to the recursion relation for r ? , the second to the recursion r e lation for r 2 r . 5 and the third to the recursion relation for ao. The first relation i s :
I
j'
x 3 - Sgx2 + a 3 S_ 3 x - a| = 0
(10)
The last relation is:
(11)
x - a3 = 0 .
For the second we have a relation of the sixth degree with coefficients symmetric functions
of the roots
Rx = r ^ r 2 ,
R2 = r f r ^
R3 = r f r 3 ,
R4 = r 3 r l 9
R5 = r 2 r 3 ,
R6 = r | r 2
1974]
< 21 )
"3a3
171
a 2 ai ,
Ba-SRjR.
B 2 = 6a 3 - 5a 3 a2a x + a 3 a 3 + a2
B3 = L R . R . R k
B 3 = - 7 a | + 6 a 3 a 2 a t - 2a 3 aj - 2a 3 a 2 + a 3 a 2 a 2
(12)
B 6 = ag
The product of (10), (11) and the polynomial whose coefficients are given by (12) is the r e quired recursion relation for the third degree.
APPLIED TO x
-x
-x-l
= 0,
we have
(13)
(211)
(14)
(31)
A t = (31) = -SLZSLt - 2a 2 + a 2 a 2
172
[April
A 3 = 2a3 - I 3 a | a 2 a 1 + a 3 a^ + a 3 a | + 10a 3 a 2 ai - 3a 3 a 2 a^
- S a g a ^ i + a 3 a 2 af
A4 = af(5 2 ) + a| (4) + af (31) + a| (1)
A4 = -afa-i + 5a 3 a 2 + 2a 3 a 2 a^ - 5a 3 a 2 a x + a 3 af + a 3 af
A 5 = aj(32) = - a | a 2 - 2afa21 +
44*1
A6 = a|
(22)
B x = (22) = -2a 3 ai + a2
(16)
B3 = 4
The product of the polynomials given by (13), (14), (15), and (16) gives the required recursion
relation for the fourth degree.
APPLICATION TO x 3 - x2 - x - 1 = 0.
(x3 - l l x 2 - 5x - l)(x 6 + 4x 5 + 15x4 - 24X3 + 7x2 + l)(x 3 + x 2 + 3x - 1)
X (x3 - x2 - x - 1) = 0
or
X15
7x 14
33xi3
while the
1974]
(17)
(ab) = S a S b - S a + b
(18)
(a 2 ) = (S2a - S 2 a ) / 2 .
173
IS
:3 - S 5 x 2 +
(311)
afs_6x
B i == a s (2)
8
B2 = a | ( 2 )
B3
Ci = a 3 a 2
(221)
c 2 = 4*1
c3 =
(41)
Di =
2
(41)
D 6 = aj
(32)
E t = (32)
E 2 = (622) + (55) + (532) + (433)
= a|(4) + (52) + a(31) + 4*1
E 3 = (852) + (654) + 2(555) + (753)
= af (63) + a 3 (21) + 2a| + ajj (42)
E 4 = (884) + (10,55) + (875) + (776)
= af(4 2 ) + a|(5) + a f(32) + a36a2
E 5 = (10,87) = a3'(31)
E 6 = aj
174
April 1974
APPLICATION TO x 3 - x2 - x - 1 = 0.
(x3 - 21x2 - x - l)(x 3 - 3x2 - x - iMx3 + x 2 + x - l)(x 6 + Ox5 + 7x4 - 24x3 + 15x2 + 4x + 1)
X (x6 + 10x5 + 75X4 + 28x3 - x 2 - 6x + 1) = 0 .
The product is
x 21 - 13x20 - 110x19 - 1374x18 2425x 17 + 543x 16 + 60340 x 15 - 3976x 14
- 43106 x 13 - 149310 x12 + 137592 x 11 + 88200x 10 + 63126 x 9 - 21742 x 8 - 13076 x 7
- 8932x 6 + 1041x 5 .- 37x 4 + 150x 3 - 10x2 + x - 1 = 0 .
CONCLUDING NOTES
1. That the symmetric functions of the roots can always be expressed in t e r m s of the
quantities S
2.
For the n
I '
This follows from the fact that the number of t e r m s involving the roots is equivalent to the
solution of x + y + z = n in positive integers and zero.
3.
For the sixth-degree relations, the coefficients AA and A 5 , A2 and A 4 , are com-
The coefficient A,
of weight kn where n is the degree being considered for the t e r m s of the original recursion
relation.
4X8.
5. If a3 = 1, all the factors for the n
degree are found for degree n + 3.
6. With some modifications on the symmetric functions involved, this approach could
be used to produce algorithms relating to recursion relations of higher o r d e r .
7. The algorithms were checked numerically by using a relation with roots 1, 2, and
4, finding the symmetric functions directly and comparing the result with that given by the
algorithms.
REFERENCES
1. Trudy Y. H. Tong, "Some Properties of the Tribonacci Sequence and the Special Lucas
Sequence," Master T s Thesis, San Jose State University, August 1970.
2. P . A. MacMahon, Combinatory Analysis, Cambridge University P r e s s , 1915, 1916. R e printed by Chelsea Publishing Company, 1960.
Editor's Note: There a r e an additional twelve pages on this subject, going through the tenth
degree. If you would like a Xerox copy of the additional material at four cents a page (which
includes postage, materials and labor), send your request to:
Brother Alfred Brousseau, St. Mary ! s College, Moraga, California 94575.
divides
A , Q = PA ,., - QA ,
n+2
n+1
^ n '
U n + 2 = P U n + 1 - QU n>
U = 0,
Ut = 1 .
The general t e r m
= _
p will denote a
prime not dividing Q. In this paper, we shall be working in the field of integers modulo
2.
Let R
p.
be the quotient U
- /U
XU = U ,-,
n
n+1
(mod p) .
*'
exists, unless p divides U , in which case the value of R will be denoted by o (All
J
n
n
n
quotients which have a zero divisor will be denoted e) If R exists and is nonzero, then
(3)
Rn+1 ,
Un+2 /Un+1 ,
P - QE;1
175
(mod p) ;
176
[April
if P | R then R ^ = ; if R = then p
= P (mod p).
F |u , so R
. ' n
n+1
n
' n
n+1
^
Theorem 1. (R ) is a f i r s t - o r d e r r e c u r r e n c e mod p and is periodic with primitive
period k(p).
Proof.
primitive period k(p) follows from the definition of k and the fact that R = 0 if and only
if P | u n + 1 .
The following theorem gives a criterion for determining whether p is a divisor of
t e r m s of (A ). It is known that if a number m divides some t e r m A
vides A
,,,
of (1), then m d i -
v for any integer t for which the subscript is nonnegative, and only those
terms.
Theorem 2.
(Divisibility criterion),
p is a divisor of A.,,
(for any
t for which
(mod p) .
if and only if
A
k(p)+1/Ak(p)
"
n '
Furthermore,
p|A
k(p)+1
/A
k(p)
Al
/ A o
It is well known that k(p) |p - (A/p). A proof is given in [4] for the Fibonacci s e r i e s ,
and it may be easily generalized to the recurrence (2).
many p r i m e s p such that k(p) = p - (A/p).
consider such p r i m e s .
The following result was proved in [l] and [2] for the Fibonacci s e r i e s .
Theorem 3. If
k(p) = p + 1
then p divides some t e r m s of
conversely.
and
1974]
177
c = R
(mod p).
Therefore,
there
is an n such that
AA /A 0 = R n
(mod p) ,
and the first p a r t follows by the criterion of Theorem 2. If k(p) is l e s s than p + 1 then not
every residue class is included in (R ), and the converse follows.
Theorem 4. p is a divisor of t e r m s of (A ) for any initial values A0 and AA,
ex-
Since
k(p) = p - 1 ,
we have
(A/p) = 1 ,
so a and b a r e in the field of Integers modulo p and p|A . By definition,
R n = (a
If R
- b
)/(a
- b ) .
Thus, R
k(p) = p - 1
and
At /A 0 ^ a or b
(mod p) ,
(mod p)
then pF divides no t e r m of (A ).
n
Proof. If
At /A 0 = a or b
then
(A/p) = 1
and pJA . If
178
= a (or b)
April 1974
(mod p)
then
/ n+1
,n+lw/ n
,riv
.
,x
(a
- b
)/(a - b ) = a (or b)
so that a = b and p | A ,
PzfA
giving a contradiction.
Thus,
a (or b) = A x / A 0 ,
and so
4.
Hall [3] has given a different criterion for whether a prime p divides some t e r m s of
(1). Bloom [2] has studied the related question of which composite numbers (as well as which
primes) a r e divisors of r e c u r r e n c e s of the form (1) with P = 1, Q = - 1 .
Ward [5] has pointed out that the question of whether or not there a r e infinitely many
p r i m e s for which k(p) = p + 1 or
p-1
For r e c u r r e n c e s
REFERENCES
1. Brother U. Alfred,
Marshall Hall,
FIBONACCI NOTES
2: MULTIPLE GENERATING FUNCTIONS
L CARLITZ
Duke University, Durham, North Carolina 27706
/Q\ z
2az-z 2
(a) . = e
n
n!
n=0
m n
H __, (a)H (b)H (c) x y
"m+n VM ' / "m w/ '"n v ~ / m! n!
2>
(1.1)
>
m,n=0
= (1 - 4x2 - 4y2)-2 exp j -4a 2 (x 2 + y 2 ) + 4a(bx + cy) - 4(bx + cy) 2 |
I
1 - 4x 2 - 4y*
4y2
)
OO
(1.2)
= (1 - 4x2 - 4y2
.m
m nn pp
^
H
_, (a)H
(a)H (b)H
(b)H (c)H
(c)H (d)
(d) X- . y f z .
M ^
>
H M
m+n+p
m
nn
ZsmJ
m+n+p
m
pp
mint p!
m,n,p=0
2
2
2
2
a ( b x + +C y + d z )
- l z 2 H c::p { -4a
' ^ ^(x ++ yy* ++ ^z ) ++ 4 4a(bx
"
2
2
I
1 - 4x2 - 4y -- 4z
_^
m n p
>
Ef
(a)H ^ (b)H ^ (c) x y z
x -^
n+p
p+m
m+n
m! nipt
m 5 n,p=Q
i
( 2
Sa 2 - 42a 2 x 2 - 4Sabz + 8Sabxy|
= d 2 e x p J 2 a^
-g
*I
(1.3)
4(bx
+ Cy +
dz)
where
d = 1 - 4x2 - 4y2 - 4z 2 + 16xyz
and Sa 2 ,
Sa 2 x 2 , Sabz,
The object of the present note is to prove formulas of a similar kind for the Fibonacci
and Lucas numbers.
2.
S = y
/, /
F F xmyn
m+n m n J
m,n=0
* Supported in part by NSF Grant GP-17031
179
180
FIBONACCI NOTES
[April
Since
F
" P ,
a - B
a + p = 1,
afi = - 1 ,
we get
c
S =
g-
X
>
_
F
/ m+n
m+n, m n
F (a
- 0]8
) x y
m,n=0
1
a
ayfixfiy
2
" ^ 1 - ax - A
1
#x
1 - ay - a V
fi2xy
Ay
2
tffiy)
I
2
_
xy_
^ ( - W l H x H H y ,
Thus
(2.1)
^
^ 0
M
n
F xmyn
n
2LzJ?
(1 + x)(l + y)(l - 3x + x)(l - 3y + y 2 )
^
T
-, m n
3 - 2(x + y) + 3xy
>
L , F F x y =
m+n
m
n
m,n=0
"
(1 + x)(l + y)(l - 3x + x 2 ) ( l - 3y + y 2 )
The sum
T
m n
T
T
, L L x y
J
m+n m n
m,n=0
is somewhat more complicated.
We get
(2.3)
V ^ L
L L xmyn
J
/ J
m+n m n
m,n=0
8 - 14 (x + y) - 2(x2 + y 2 ) + 27xy + 7xy(x + y) + 3x2 y2
(1 + x)(l + y)(l - 3x + x 2 ) (1 - 3y + y 2 )
1974]
FIBONACCI NOTES
181
F F + F
F
F + F
F F
+F
F
F
m+n m n
m+n-1 m - 1 n
m+n-1 m n - 1
m+n-2 m - 1 n - 1
= F
F
- F
F
2m 2n
2m-2 2n-2
F F + L
F . F + L
F F
+L
F
F
m+n m n
m+n-1 m - 1 n
m+n-1 m n - 1
m+n-2 m - 1 n - 1
= 3F
F
- 2F
F
- 2F
F
+ 3F
F
or
r
2m+2r2n+2
2 n i 2n+2
*2m+2 2n
^ 2 H I 2n '
It may be of interest to mention that the generating functions (2.1), (2.2), (2.3) can be
extended in various ways.
V ^
>
(2.6)
^
F _,
M
m+n+
F
m n
n+2Xy " F n X y ( x + ^ + Fn 2 x 2 y 2
F x m y n = -Elf
_E
.
Eli
n
(l+x)(l + y ) ( l - 3 x + x2)(l-3y+y2)
F ^ ^ F F + F ^ ^ - t F
- F + F F J + F . ^ ^ F
. F m+n+p m n
m+n+p-1 m - 1 n
m n-1
m+n+p-2 m - 1 n - 1
= F
Since L
F F - F ( F
F + F F
p+2 2m 2n
p l 2m-2 2n
2m
)+F
F
F
2n-2;
p-2 2m-2 2n-2
L ,
F x y
m
m,n=0
L , 0 - L xy(x + y) + L 0 x2 y2
Ei^
E_l
P^_i
(1 + x ) ( 1 +
y ) ( 1 - 3x + x*)(l - 3y + y*>
and
(2 9)
m+n+pFmFn
=
m+n+p-l
p+2F2mF2n "
(F
(F
m - l Fn
2m-2 F2n
+
F
mFn-l)
2m F2n-2 ]
+ L
+ L
respectively.
3. We consider next the triple sum
(3.1)
\
F ^ M F F F xmynzP
J
AmmJ
m+n+p m n p
m,n,p=0
m+n+p-2 F m - 1 F n - 1
p-2 F2m-2 F2n-2
'
182
FIBONACCI NOTES
& I (1 - a2x)(l
- a/3x)(l - a2y)(l
[April
- c*j3y)(l - a2z)(l
- apz)
]63xyz
_ _
2
f
2
\ ^
F _, _, F F F x m y n z P
J
/ j
m+n+p m n p
m,n,p=Q
._
2 - (x + y + z) + (yz + zx + xy) - 2xyz
(1 + x)(l + y)(l + z ) ( l - 3x + x 2 ) ( l - 3y + y 2 )(l - 3z + z 2 )
=
(1 - x ) ( l - y)(l .- z) + 1 - xyz
(1 + x)(l + y)(l + z)(l - 3x + x 2 ) ( l - 3y + y ^ U - 3z + z 2 )
V*
>
(3.3)
X ^
n
i ' * * ' n k = 0
F A__ A__ F
F
F ...... FFnn. ixf
xr11 .. .. .. x"k
xr K
n
n ^ - +n^ nx
k l
k
n (l + x.)- n (i - 3x. + x 2 )
J
j=l
where c. is the j
j=l
'
j=0
]=0
1974]
FIBONACCI NOTES
2j<k
183
2j>k
2j<k
2j<k
2j<k
In exactly the same way we can prove the more general result
>
F __,
__, ^ F F n i xf 1 xk
Z-#
n ^ . - - + n k + p n*
% i
k
n
i - e >nk=0
(3.4)
j=i
j=i
,
, ^ F Fn n i X-. 1 xv,
n ^ . *-+n k +p ni
k *
*
n ls . 5 n k =0
g^V^j
k
3=1
5=1
T-.
1
(l-a2x) (1+y) (1+z)
(a-|3)3
.^
1 _ y^
(a-j8)3
(a2 - ff2)x
(l-3x+x 2 )(l+y)(l+z)
184
[April
FIBONACCI NOTES
It follows that
oo
(4.1)
n+p F
m,n,p= =0
m n p
, x Jy z F
_, :_Fm+n
p+m
2x - 3 2 x y + 8xyz
1
5
_,
o-.
2 \ ,/-,
2\/i
iE5
2 \
"
(1 - 3x + x 2 ) ( l + y)(l + z)
,* 2)
q - 5r + 2pr + 2qr + r 2 - q2
_ _
13x'
<* =
Hxy>
x z
y -
(4.3)
\ ^
F ^ F ^
_ F M ^ xmynzP
J
x J
n+p+r p+m+r m+n+r
m,n,p=0
- F 0 - F 0 0 V x + F n . V x yJ - F 0 xyz
A 3r
3r-2 *-*
3 r - 4 Z^
3r-6 J
5
(1 - 3x + x 2 ) ( l - 3y + y 2 )(l - 3z + z 2 )
(-1)
5
r v
F - F 0x
^/^
r
r-2
Z ^ (i _ 3x + x 2 ) ( l + y)(l + z)
(4.4)
V "
xmynzP
J
x J
n+p+r p+m+r m+n+r
m,n,p=0
L^ - L 0 rt V*x + L 0 . V*xy - L ~xyz
3r
3r-2 *>
3r-4 ^ J
3r-6 J
(1 - 3x + x 2 ) ( l - 3y + f)(l
J-j
J_J
- 3z + z 2 )
X
L
^2
(1 - 3x + x 2 ) ( l + y)(l + z)
'(1
1974]
FIBONACCI NOTES
185
nf = p + m ,
pf = m + n
Then
I -m f + nf + p' = 2m
< m? - nf + pf = 2n
( mf + nf - p' = 2p
(4.5)
so that
mf + n ! + pT = 0
(4.6)
(mod 2)
and
m1 ^
(4.7)
n r + pf 5
nf ^
pf + m f ,
p! ^
mf + nf .
uv,
(4.3) becomes
(4-8)
JI
J^J
. F Fm
_ Ff +r
_ F ^n??+ r F p f + r U ^
vXi w^
- F 0 - F ^ y ^ u v + F 0 . uvw y ^ u - F u 2 v 2 w 2
1^ 3r
3r-2Z^
3r-4
Z^j
3r-6
(1 - 3vw + v 2 w 2 )(l - 3wu + w 2 u 2 ) ( l - 3uv + u 2 v 2 )
(-1)
r -"*%
\/^
- F 0 vw
r 2 2 r-2
(1 -' 3vw + v w )(l + wu)(l + uv)
REFERENCE
1.
A COMBINATORIAL IDENTITY
MARCSA ASCHER
Ithaca College Ithaca, New York 14850
Define
(1)
f(n,k)
i>'M(*y
where
n/2,
(n - l ) / 2 ,
n even
n odd
*- < - t ( i V , ) - ^ l t ( . , : i )
for
The usual induction procedure must be modified since the identity involves both n and k but
only restricted values of k associated with each n.
proceeds.
For the n and k shown, the identity is valid at the darkened grid points.
The
(n,k),
n even,
k /
0,
k ^
true for (n, k - 1), and using the result of A that (2) is true for (n + 1, k), then (2) is true
for (n + 2, k). P a r t C shows that (2) is true for (n,0) and P a r t
(3)
Starting with
a factor of ( - l ) 1 2 n "
April 1974
A COMBINATOBIAL IDENTITY
187
C
6 C
5 C
4 C
Tr
3 C
2 C
1C*
Figure 1
(4)
l,n
2,n
3,n
f(n + 1, k) - 2
4,n
l,n
S2
= f(n,k) - 2 n " 2 k ( - l ) k
(5)
S
"4,n
If (2) is true for (n,k),
35n
" f ( n " *
k) 2
f(n - 1, k - 1) + 2 n + 1 " 2 k ( - l ) k
(-)]/
f(n + 1, k) = (-1)
(6)
(V)
k/
n + 2
I n + 1 - 2k J
188
A COMBINATORIAL IDENTITY
April 1974
is introduced into each t e r m and each term of the equation is summed over
n/2.
i = k + 1,
The result is
l , n + 1 " W*)(-V
(4')
n+2 / n + 2
2 /
2
2,n+1
(l/2){ 1]
3.n+1 "
n+2 / n
2 I 2
{l/2){ 1]
4,n+1 "
n+2
2
n
2
k - 1.
If (2) is true for (n,k), (n, k - 1), and (n + 1, k), (4T) can be solved for f(n+2, k) and
f(n + 2, k) = (-1)J
(5')
d
n + 3 \
1 n + 2 - 2k 1
("j-')(-.-')*(?--})
(3")
(4")
l,n
2,n
where S- , S 0 , S 0
are as defined in (5).
l,n
2,n
3,n
Hence, if f(n,0) = n + 1 and f(n - 1, 0) = n,
3,n
if
f(n,0) = n + 1
and
f(n + 1, 0) = n + 2 for n even. Verifying by substitution into (1) that f(2,0) = 3 and f(l,0)
= 2 completes the case of k = 0.
D.
(n, n/2)
and
In [ l ] , we find three well known divisibility properties which exist between the Fibonacci and Lucas numbers.
They a r e
(1)
iff
IF
m = kn ;
n I m
(2)
L I F
n I m
(3)
iff
L I L
n ' m
iff
m = 2kn,
n > 1;
m = (2k - l)n,
n > 1.
o r n|L .
As a r e -
sult of this investigation, we will also illustrate several interesting congruence relationships
which exist between the elements of the sequences \ F } and (L } .
The first result, due to Hoggatt, is
Theorem 1. If n = 2 - 3 k , k > 1, then n|L .
Proof.
L
= a + rj3 , we have
T
L
3n
3n , 0 3n
+fi
, n , n w 2n
= \a
=
However,
+ jS )(a
-aft
V L 2n " (-1)n>
n 0 n , D2n^
+ j3
V L 2n " '
L 2 = L rt + 2 if n is even so that
n
2n
(4)
L 3 n = L n ( L ^ - 3) .
The theorem is true if k = 1 because n = 6 and L 6 = 18. The result now follows by
Curiosity leads one to ask if there are other sequences (n k ) such that njL n k . The
authors were unable to find other such sequences until they obtained the computer results of
Mr. Joseph Greener from which they were able to make several conjectures and establish
several r e s u l t s .
!90
j _
t = p
then p |L
k = 1.
[April
k_i > k ^. 1.
Assume
p L
^ -i and let
i np^"1
,.
(5)
nt p
+ dS nt )
n t ( ^
(6)
and (p + 1 - q)
^^q+l^nttp-qj^nttq-l)
(-1)(n+l)(p-l)/2
t e r m s , where q =J= (p + l ) / 2 ,
is
^p-q^ntfa-l^ntfo-q)
"lj
v(n+l)(q-l) T
nt(p-2q+l) '
nl ipp tt "
nl itt l1 ^ j
(_1) n
C|
nt(4k-2q+2)
q=l
L
k-1
nt(I^L4nt(k-q)
, q=0
(
(
^+1T
X / ' ^ ' ^ -J 2nt(2k-2q+l)
' o , a^ + 1
q=l
k-1
X )
q=0
k-1
[5F
2nt(k-q)
+ 2
X ) (-1)n+1[Lnt(2k-2q+l) "
q=l
^ " 1
X 5 F 2 n t ( k - q ) + S (-1)n+lLnt(2k-2q+l) + P
q=0
q=l
since L. = 5F 2 + 2, L2 = L 0 + 2 ( - l ) r , and t(2k - 2q + 1) is odd.
4r
2r
r
2r
^
Now pt|L , (2k - 2q + 1) is odd, and 2(k - q) is even so that by (2) and (3) one sees.
that p is a factor of the expression in the parentheses of (8). Hence, p 2 t|L
r e m is proved if we have p = 1 (mod 4).
1974]
191
k+1
(9)
k-1
q=0
we have
However,
2|L
3k
, if p = 3 (mod 4).
ki
l L 2,,3k-l
or
ki
|L2-3k
for
'
F u r t h e r m o r e , this p r o -
k zLl,
qu,
Q^
by
|
2-3kpt
t
By an argument similar to that of Theorem 3, it is easy to see that the following a r e true.
Corollary 1. If p and q a r e distinct odd primes such that p L
FM
and n a r e odd, then (pq)
and q |L
where m
and
Corollary 2* If p and q a r e distinct odd p r i m e s different from 3 such that ptL 9 Qk
k t r
i
and q|L 9 ok w n e r e k 1 a n ( i n ~ 2 ' 3 P Q t n e n n | L r t ^ 0 and r ^ 0.
Using F
= F L , we have
I
k i
then p
| F9
and
Corollary 4, If p and q a r e distinct odd p r i m e s such that p|L
m and n a r e odd integers then (pq)
| F
and q|L
where
then pt J F
the factorization of x
(10)
fc.
F
,.
- y , we have
= F^
J\nt(p-i)pnt(i-l)
i=l
192
(-1)
"
(p + 1 - q)
(11)
a nt(p-q)^nt(q-l)
^nt(q-l)^nt(p-q)
"lj
5F
(_1)
nt(p-2q+l)/2
[April
(q-D
and
= 5 F | + 2(-l) , is
^2nt(p-2q+l)/2
Z{ 1}
% t = Fnt I 2^
q=i
Using ptjF , and (1), we see that p is a factor of the expression in the parentheses of
(12) so that p 2 t I F
pJF
and
qJF
then
Another result of Theorem 4 which was already discovered by K r a m e r and Hoggatt and
occurs in [2] is
sk F
(13)
k'
for
k > 1
since F 5 = 5. Note that this result also gives us a sequence {n^} such that n, | F n , .
J u s t as the authors could find several sequences
also able to show that there a r e several other sequences {n^} such that n, [F n , .
they were
With this
Proof. By the discussion following Theorem 2 and Corollary 3, we have 3 \F. o m for
m > 1 . But 4 | F 6 so that 4 | F 4 < 3 m for m > 1. Since (4, 3 m ) = 1, we have 4 - 3 m | F ^ 3m
for m ^. 1 and the theorem is proved if r = 1.
Since
F
3n2r+2
^ n ^ r + l (5F|m2r
l
m r+21
and 2 | F 3 , we have by induction on r that 3 2
| F om.9r+2
Theorem 6. If
,
~
n = 2r+13m5k,
where r > 1, m > 1, and k >. 1 then n | F .
1974]
193
This result follows immediately from Theorem 5, (1), and (13) because
(5 k , 2 r + 1 3 m )
= 1.
Let n = 2
3 p
where k ^ 1, then n F .
1
n
and
Theorem 8.
1
and q F . Let n
For our next
Theorem 9.
r~t~l m
1
Let s = 2
3 . Let p and q be distinct odd p r i m e s such that p F
s
k t
1
= sp q where k > 0 and t > 0 then n F .
divisibility property, we establish
If k > 1 then 2 k + 2 1 F 9k
k-1
Proof.
9
X^JS
= F
L
OS
8s F , .
^^
Since F
and
OS
'6s
(5F2Q +2) and 21F 3 , the result follows by induction with the
= F
DS
OS
use of (1).
Throughout the remainder of this paper, we analyze the prime decomposition of
where n is odd and establish several congruence relations between the elements of { F }
and ( L } . With this in mind, we first establish
Lemma 1. If n is odd then L = 4 M where t = 0 o r 1 and M is odd.
n
Proof. Since n is odd, we have (1) L = L 0 ,., where m is even, (2) L = L n , _
n
3m+l
' n
3m+2
where m is odd, o r (3) L = L 0
where m is odd.
n
3m
If L = L 0 ^ and m = 2r then L = L - _ . Since 2 | F O , L c = 5F20 + 2 ( - l ) r ,
n
3m+l
n
6r+l
' 3r
6r
3r
'
and (L^ , L i n ) = 1, we have L 0 in is odd o r that L_ , - = 4M where M is odd.
6r.
6r+l
3m+l
3m+l
Suppose L = L
where m = 2r + 1. By an argument similar to that of Theorem
By a similar argument, it is easy to show that L
= 4M where M is odd.
Suppose L n = L 3 a
2, it is easy to show that
L
<">
if
L
r is even;
6r+3
if r is odd
Now 2 F n /
so that the t e r m s in the parentheses a r e odd and L = 4M where M
i 3(r-q)x
^
n
is odd.
The following theorem is due to Hoggatt while the proof is that of Brother
Alfred
Brousseau.
Theorem 10. The Lucas numbers L
prime divisor
p of
L .
n
= (-D n 5 5
However,
we have that
L , n = L + L - so that L ,- = L -. (mod pF ) .
n+1
n
n-1
n+1
n-1
194
Therefore,
.. = 5 (mod p) and
[April
primes having 5 as a quadratic residue are of the form 10m 1 , we are through.
Using Binet's formula, it can be shown that
(15)
12t+j =
5F
(l2t+j-l)/2F(12t+j+l)/2
(-{]~m>
odd
L 1 2 t + 1 = ION - 1 ,
(b)
L 1 2 t + 3 = 4(10N
1) ,
(0
L 1 2 t + 5 = ION f 1 ,
(d)
L 1 2 t + ? = ION - 1 ,
(e)
L 1 2 t + 9 = 4(10N - 1) ,
and
(f)
1 2 t + l l = WN - 1 .
Since the proof of Theorem 11 is trivial, it has been omitted. However, a word of caution about the results is essential.
lOn + 1 since
192|L12.i443 and
1992 |LI2.I82+5L
or
HowL
must be even.
Since l l 2 L ^ ^ + T *
21l|L l 2 e l +9
12t+j
n+6 -
n-2 =
+ L
n+4 =
n+2L2 "
25-l
(1?)
J2
i=0
n+4i
= F
n + 2i + 2. 2 " F n-2
1974]
195
J2
j
L
n+4ki
V(2J-l)2k "
1 1
i=0
2ik
'
(19)
n+2J+1 . n n V
n+2J +2 "
1=1
so that
F
(20)
n+23
+2
"
(mod
2i)
for
1 ^ i < j
and
(21)
Fn+2J+2
F n (mod L n + 2 j + 1 )
if
j t
0.
In papers to follow, the authors will generalize, where possible, the results of this paper to the generalized sequence of Fibonacci numbers as well as to several general linear
recurrences.
They will also investigate sums and products of the form occurring in (18).
REFERENCES
Brousseau,
L a r r y Lang, Fibonacci Quarterly, Elementary Problem No. B-247, Vol. 10, No. 5 (Dec.
1972).
i>-
2.
i=l
b
F(i) = 0
H F(i) = 1
i=a
5. Bj = l / 6 ,
B2 = - 1 / 3 0 , B 3 = 1/42,
?Q(xt,
x2, , x m ) =
for
for
etc.,
a > b
a> b
^[A(I;;:)]-(V-0
For example,
ao) = rMo5)
ad. s) = d - s r 1 ^ ) ^ )
ftd.3.4) = (1-3-4)-1 ( ; ) ( { ) ( J)
7.
d Q (x 1? x 2 , , x m )
= gq(Xi, x 2 , " , x m ) - n X l
Then
Q
(Q + 1)S^ = n Q + 1 + (Q + l ) n Q - 1 + n (1 - rx. )
^
i=l
where
196
April 1974
197
n (i
i=2
rx
, where xt<
r.,
m > 2,
we let
dQ(x l5 x 2 , , x m ) .
Theorem 2. Say Q > 1.
r x -rx1 ^x2
Then
(2Q + l ) B n - -Tt
Q
2Q
n (1 - r . )
x
i=2
where
-rt
2Q
n (1 - r . )
x
i=2
i ' V
x
rv
r.,
for m ^ 2, we let
r xv l -r.
"x2
3
&9rv
2Q i' 2 ' ' "' ' m ' *
Theorem 3. Say Q > 1.
Then
(S + 1) Q - s Q = (n + 1) Q - 1 ,
where S
+ n
n2
+ n4
So = - ^
+ na
etc. , where the entries in the determinants are binomial coefficients, z e r o s , and powers of
We now illustrate two more methods for finding S~.
198
E[ (i
For example,
+ i)2 (i i)2 =
- - ] Z'
"1=1
April 1974
4i
i=l
n
(n + l) 2 + n2 - 1 = J
4i
i=l
n
4 y
i = 2n2 + 2n .
i=l
n
-
n2 + n
_ n(n + 1)
i=l
Method 2. Lagrange interpolation.
1 in n, we now compute St.
Q+
Fl
^2
- t 2 ) ( n - <*>
I t T - taXti - %)
(n - 2)fa - 3)
(-l)(-2)
( n
- t i ) ( n - t3)
TtT- t t )(t 2 - tj)
(n - i)fa - 3)
d)Pi5
(n-^Hn-t^)
Tti" - tjHta - t 2 )
(n
1)(n
(SHU
2)
'
Editors 1 Note: This abstract qualifies for the Fibonacci Note Service. It is an abstract of a
paper which is fifty pages long.
cents a page (which includes postage, m a t e r i a l s and labor), send your request to:
Brother Alfred Brousseau, St. Mary's College, Moraga, California 94575.
1. In this article, a generalized form of E u l e r f s law concerning the sigma function will
tained and used to derive expressions for ^T%
be obtained
addition and multiplication.
which contain ju
just functions involving
(1)
to obtain equations with classes of solutions identical with the class of prime numbers.
2.
Let
F n =
Proposition 1. If
D Ftn) xn
n=l
converges on some interval about 0,
then
n
(2)
0 = nR(n) + V
F(a)E(n - a)
a=l
where
_^
(3)
2 ^ E(n)x n =
n=0
II (1 - x n )
f(n)/n
n=1
The proof mimics E u l e r ' s for the case f = identity, which is the recursive expression
for sum of divisors he obtained by describing E.
199
[l]
200
[April
Proof.
J^ f ( n ) x n / ( l - x n ) =
n=l
n
f(n)
2]
k
x nk
+ f (2)x6 + . . .
+ f (2)x*
+ f (3)x3
+ f (3)x6 + . . .
+ f (4) x 4 +
+ f(5)x5 +
+ f (6)x6 + . . .
OO
= X>
OO
f(d)
n=l
d n
F(n)x n .
n=l
That i s ,
J ] f ( n ) x n / ( l - x n ) = Yl
(4)
n=l
FWx11 .
n=l
Suppose
(5)
o < n(i -
f(n)/n
< >
on some interval about 0. We show that (2) holds under (5) and then that (5) holds when
n=l
converges on some interval about 0.
Let (5) hold. We have the identity:
00
log n (1 - x n )
n=l
f(n)/n
= J2
f(n)/nlog (1 - x n )
1974]
(i
sf? - -
f(n)/n
E-WVtt-^) =_L1
TCJ75T
n (1 - x11)
1
^Etm)^)/EBto)^
mRdnJx 1 1 1 " 1 / ^
2
0
R(m)x m .
0
Hence, by (4),
(6)
mR
" 2
(m>xm/
(m)xm
n ) x n / (1 - x n ) = J^
So,
I2
F(n)xJ1
J{ 2 R ( m ) x m J + 2
0=
is 0:
F(a R(n
> - a) + nR<n)
2
a=l
2F(n) *n
n=l
converges on some interval about 0.
E
By Eq. (6),
F(n)x
1
^w^
202
[April
where
n (l . x11)
1
P(x) =
f(n)/n
Therefore,
OO
P(x) = exp
(7)
f - E FdDx11"1 dx
J
x
I E
*
iff
F(n)x
E
1
Thus (5) holds iff
(l/n)F(n)x n
E
1
F(n)x n
<
oo
1
on the same interval.
Q.E.D.
Let
E
f(d) = F(n)
d|n
Then,
|EF(n)xn|
<
. Actually,
1974]
203
Ef(n)xn|
<
|5>(n)xn|
< - - | ^ f ( n ) x n / l - xn|
< - -*|^f(n)xn|
<
I E fWx11 | <
By the root test,
lim sup |f(n) |
That i s ,
sup L.
<
<
, where
L. = lim |f(a
k
a
)| ik
For each k,
lim
c,
f(c, )
< sup L
< oo,
sup lim
{ak}
f(c. )
k
<
sup L. <
'
That i s ,
sup lim
max |f(d)
dla,
< sup L. .
Then f
204
[April
Now
max
da,
max |f(d)|
dla,
f(d)
So:
sup L. <
That i s ,
lim sup max |f(d)|
d|n
<
Thus,
lim sup
T(n)
<
Ld|n
1
< lim sup T(n)
1
lim sup [ m a x
Thus,
Ld|n
|f(d)|1n
<
OO
Then,
|5>(n)xn|
<E
E |f(d)|x n
djn
W max | f ( d ) | x n < o
din
q. e.d.
We r e p a i r the gap in the proof of Proposition 2, the assertion without demonstration that
Er(n)xn
converges on some interval about 0, by comparing this sum with another.
obvious on comparing T(n) with the identity function:
The result is
1974]
205
11
IE** 1 <
on (-1,1).
One more proposition is needed to finish the background for a demonstration that P r o p osition 1 applies to 9/
Proposition 3:
J ] l / n F(n)xn
converges on some interval about 0 iff
^F(n)xn
converges on some interval about 0.
Proof.
< <*> .
|F(ak)(l/ak)|
converges, its limit is
a,
lim | F ( a k ) |
Also, it is clear that
a.
|F(ak)|
So
1
1
a,
lim sup j F(n) | n = sup lim | F(a fe ) |
= lim sup
F(n)(l/n)
1
a,
= sup lim | F (a- M l / a , ) |
206
So
X>(n) x
0.
[April
q. e.d.
Now we prove that Proposition 1 may be applied to
Z^.
Proposition 4.
E 1fai(n)\ x n
CZP
we express the coefficients of the power s e r i e s for P(x) as the coefficients of the expansion
of a finite product.
Proposition 5.
n
0 = nR(n) + ^ ( a ) R ( n - a) ,
a=l
where R(k) = coefficient of x
in
k
n (i n=l
n
)
)n
= SW* 11
n=0
Let
k
n*'1
n (i - x )
= E Rk(n)xn
1
(Definition).
Then
53 \ + 1 ( n ) x n
n
k+1
n
n (1 - x n )
n=l
E ] 7
r=0
\
[ k +
, - Lk+lJ
= (1 - x K + 1 )
r
r
l ] a _ 1
x
n
R k (n)x n
)(-l)rx(k+1)rxE\(n)xn
/
n
1974]
= E R k (n)x*
n
[ k +
^_1
r=l
( fk
>
I]"" 1
207
= E W)**
n
None of the t e r m s in the second summand have exponents ^ k . Thus
R k (D = R k + 1 (D
for all i ^ k Indeed, R*k(i) = R^i) for all i and 1 such that i < k < 1.
k
n
E s ( n ) x n = lim n (1 - x11)
k l
and R: (n) = S(n),
Thus
q.e.d.
, which is expressible in
t e r m s of just addition and multiplication, and which leads to the equation mentioned in the
title.
Define F (1) = 1 and, supposing F
defined on 1, 2, o e o , n - l f l e t F (n)
satisfy
n
0 = nR(n) + ] P F (a)R(n - a) ,
a=l
where R is defined as in the statement of Proposition 5. Then, by Proposition 5, F = ^ 9
a
and F satisfies 0 = F (n) - n - 1 just when n i s a prime number.
REFERENCE
1, Euler, Opera Omnia, Series 1, Vol. 2, pp6 241-253, n Discovery of a Most Extraordinary
Law of the Numbers Concerning the Sum of Their Divisors^"
AFtn) = E
F(d) .
din
Let
00
converge.
JjxnAf(n)
Let
n
n (1 - x n ) f(n)
= V"V(n)x n .
f
n=l
*-*
n=0
(2)
Then for all n:
n
(3)
0 = nR f (n) + V ^ A f ( a ) R f ( n - a)
induction) from (3) that R (n) = (-1) /nS and hence that
w(n)/n
^"% / i \ n
n=0
fl (1 - x )
n=l
off roots of 1.
REFERENCE
1. B a r r y Brent, "Functional Equations with P r i m e Roots from Arithmetical Expressions for
^ i " Fibonacci Quarterly, Vol. 12, No. 2 (April 1974), pp. 199-207.
208
bases three [ l ] , five [2], and nine [5]. Also, it has been shown [3] that, in a system of
numeration with base (2k + I) 2 , when k(k + l ) / 2 is annexed to n(n + l ) / 2 then
[n(n + l)/2](2k + I) 2 + k(k + l ) / 2 = [(2k + l ) n + k] [(2k + l)n + k + l ] / 2 ,
another triangular number.
k(k + l ) / 2 ,
so that each
breviated notation is employed, wherein a subscript in the decimal system following an expression indicates the number of times it is repeated in the integer containing it.
Thus, the
with a b a s e ,
b = 8m + 1,
where
itself is a
In general, if n
has
a palindromic A(n) = m 9 ,
2k
n = 4m, .
BASE NINE
The smallest base of the form 8m + 1 is nine, for m = 1. Hence n = 4,
the palindromic A(n) = 1 2 . ,
k = 1, 2, 3, .
generates
(2k + I) 2 .
The
1 , with p = 1, 2, 3,
210
[April
F u r t h e r m o r e , for k = 0, 1, 2, * , we have
n = 24 k 6 = 2(10 k + 1 ) + (10 k - l)(10)/2 + 6 = [ 5 ( l Q k + 1 ) + 3 ] / 2 .
It follows that
A(n) = [ 5 ( 1 0 k + 1 ) + 3 ] [ 5 ( 1 0 k + 1 ) + 5 ] / 8
= 5 2 ( 1 0 2 k + 2 ) / 8 + 8(5)(10 k + 1 )/8 + 3(5)/8
= 3 ( 1 0 2 k + 2 ) + 1 0 k + 2 ( 1 0 k - l ) / 8 + 6(10 k + 1 ) + 10(10 k - l ) / 8 + 3
= 31 k 61 k 3 .
Thus there are two infinite sequences of palindromic triangular numbers in base nine.
These do not include all the palindromic A(n) for n < 42161. Also, there a r e :
A(2) = 3, A (3) = 6, A(35) = 646, A(115) = 6226, A(177) = 16661,
A(353) = 64246
A(2662) = 3678763,
A(3525) = 6382836,
A(3535) = 6428246
then
A(n) = (10 2 k - l ) / 2 3 = ( 2 m 2 k ) / 2 3 .
Now, if
[2m(2m + 1) + 2 m ] / 2 3 = m(m + l ) / 2
< 2m + 1 ,
base seven, A(3 k ) = "06^. In base nine, A(4 k ) = TT^. In base eleven, A(5 k ) = T4 k .
1974]
211
In these
odd bases <nine, these triangular numbers a r e palindromic with 2k - 1 digits. In such odd
bases >nine, these triangular numbers consist of repeated p a i r s of unlike digits, so they
a r e not palindromic.
In bases of the form 8m + 1 (including nine), these triangular numbers are
repdigits
n < 11(10 4 )
a r e of the
A d j ) = 0T k type.
In base five, for n < 102140, the other palindromic triangular numbers are
A(l) = 1,
A(3) = 11,
A(1303) = 1130311,
A(12143) = 102121201,
A(13) = 121,
A(1331) = 1222221,
A(102) - 3003 ,
A(10232) = 30133103 ,
A(12243) = 103343301,
A(31301) = 1022442201 .
In base seven, for n < 54145, the other palindromic numbers are:
A(l) = 1,
A(2) = 3,
A(242) = 33033,
A(1611) = 1525251,
A(2442) = 3352533,
A(13065) = 102252201,
A(15) = 141,
A(24) = 333,
A(254) = 36363,
A(1301) = 1012101 ,
A(2414) = 3251523,
A(2522) = 3546453,
A(135) = 11211,
A(2424) = 3306033 ,
A(12665) = 100646001 ,
A(13531) = 112050211,
A(15415) = 142323241 ,
A(25412) = 363030363 .
Thus, in bases five, seven, and nine (but evidently not in base three) there are palindromic A(n)
palindromic.
BASE TWO
In base two, for k > 1, if n = 10 + 1 ,
then
= io 2 k - x + io k + io k - x + i = io 2 k - x + iido^1) + i
For n
<
sequence a r e :
101101, in the binary system, palindromic A(n) not contained in this infinite
212
A(10) = 11,
A(110) = 10101,
A(11001) = 101000101,
April 1974
A(10101) = 11100111,
A(101010) = 1110000111 .
No infinite sequence of palindromic triangular numbers has been found in base ten [4]
or in other even bases > two.
REFERENCES
1.
Charles W. Trigg and Bob Prielipp, "Solution to Problem 3413," School Science and
Mathematics, 72 (April 1972), p. 358.
2.
3.
4.
Charles W. Trigg, Mathematical Quickies, McGraw Hill Book Co. (1967), Q112 p. 127.
Charles W. Trigg, "Palindromic Triangular N u m b e r s , " Journal of Recreational Mathematics, 6 (Spring 1973), pp. 146-147.
5.
G. W. Wishard and Helen A. Merrill, "Solution to Problem 3480," American Mathematical Monthly, 39 (March 1932), p. 179.
It is a well known fact that 3 + 2 \l2 is the fundamental solution of the Fermat-Pellian
equation x2 - 2V2 = 1. Hence, if U + VN/2 is any other solution then there exists an integer
n such that u + v N/2 = (3 + 2\H)
a33 = 3, and a.. = 2 for all other values. It is interesting to observe that there exists a
relationship between the integral powers of T and 3 + 2N/"2. In fact, a necessary and sufficient condition for M = T
b 12 = b 2 i = m,
bn =
results.
Lock Haven,
(1) x, for i = 1, 2, 3, , or
(2) If A l 5 A2 are wff f s,
xl5 x2, , x
in that o r d e r with each letter occurring exactly once. There is one wff of o r d e r 1, namely
xA. There is one wff of order 2, namely Px t x 2 .
Px1Px2X3 and P P x j X ^ ,
Define a sequence
( G )
J
i=i
as follows:
g. is the number of distinct wff?s of o r d e r i.
a.
b.
G.
and
i=i
i=l
213
namely
214
H-235
Proposed
a.
by G. Wulczyn,
Bucknell
University,
[April
Lewisburg,
Pennsylvania.
n+1
n=0
b.
L . ., x
n+1
n=0
H-236
Proposed
by L. Carlitz,
Duke
University,
Durham,
North
Carolina.
Show that
(i)
>
(-lTx"
= >
n=0
n=0
.
n (i _ x k )
k=l
2 n+1
j2(-a*(n+i)2
(2)
2n
-jfj
"W2n
<*>
= y ^ ^^2 -n +L l k=l
n (i - x S ,
n=0
n=0
(x)0 = 1.
SOLUTIONS
TO COIN A THEOREM
H-199
Proposed
Carolina.
by L. Carlitz
and R.
Scoville,
Duke
University,
Durham,
North
= V o
n-3
n-2Q
+V
9,
for
- .
n-1
Show that any (finite) pocketful of coins is equal in value to a pocketful containing at most one
coin of each type.
1974]
215
- and c
9)
a r e present.
Call two
C , C
-, C
can be replaced by
+2
etc.
Assume for the moment, the following statement:
R: S is true for any canonical pocketful to which one extra coin has been added.
Then the general result follows by induction on the number of coins for any pocketful P: R e move a coin to get P ? , apply the induction hypothesis to P f to get a canonical pocketful Q,
return the removed coin and apply R.
Now to prove R, let us prove by induction on j the s e r i e s of statements
or a
. and a C
R.:
C ,
f
then if
then S
is true for P ! .
Assume R,
for all k < j (it is obvious if k < -3). Now let Q be canonical. We can sup-
C,
5. = 0 or 1,
5.5. -6.
Then
Q U C0
...
6_ 3> 5 _ 2 , 8 _ r
5 _ 3 + 1, S_ 2 ,
Again, if S1 = 0, by induction,
Q U C0 = . . .
S0
l,
5 2 , 83>
Then
S_r
0, 5 1 + 1, 5 2 ,
we a r e finished, so assume 5 1 = 1.
5 _ 3 + 1, 5_ 2 + 1, 5 _ 1 ,
0,
53,
Then
0, 6 2 + 1, 5 3
-..
Q U C0 U C t = . . .
5 _ 3 , 5 _ 2 , 5__ r 5 Q + 1, S + 1, 8 2 ,
5 3 .
[April
216
Then
Q U C0 U
Ct
S3
S_ 2
_3
-2
+ 2
_; 3
_2
+ 2
5
8
- 3
+ 1
- 2
5_1
-l
+ 2
-l
1
2
1
+ 1
8n
82
83
2
2
1 8
2 +
83
Proposed
by Guy A. R.
Guillotte,
Cowansville,
Quebec,
Canada.
Let M(n) be the number of primes (distinct) which divide the binomial coefficient,
.n _ / n \ *
Clearly, for 1 < n < 15, we have M(l) = 0, M(2) = M(3) = 1, M(4) = M(5) = 2, M(6) =
M(7) = M(8) = M(9) = 3,
etc.
M(10) = 4,
Show that
( m ( n ) } n =i
has an upper bound and find an asymptotic formula for M(n).
*Divide at least one C, . w h e r e 0 < k < n.
Solution
by D. Singmaster,
Instituto
Mathematica,
For a prime p, if
1(0
Pisa,
Italy.
M(13) = M(15) = 6,
1974]
for some k5
0 ^ k ^ n,
217
number of primes l e s s than or equal to n. We claim M(n) ~~ 77(n). To see this, we can use
the following result of B. Ram.
1; Chelsea, 1952; p. 274, item 98.) There is at most one prime p < n such that
Since Ram's paper is somewhat inaccessible, I will prove a slight sharpening of it, u s ing an accessible result.
-/CO
for 0 ^ k ^ n if and only if n = ap
two primes pA and p 2 with pt<
p 2 n + 1 and
*
for 0 ^ k ^ n.
n + 1 = 3iipi l
1 < a2 < p 2 .
a2P2 2
so
s2 = 0
p < n + 1
such that
#
for 0 < k ^ n and such a p exists if and only if n + 1 = ap
cussion related to this may be found in my survey paper: "Divisibility of Binomial and Multinomial Coefficients by P r i m e s and P r i m e P o w e r s , " (to appear).)
By carefully examining the role of
M(n).
n + 1,
218
=\ ^n + ^ i f
) 7T(n + 1) - 1
/ v _ , 7T(n)
^ 7T(n) - 1
+ 1
^ aPS
otherwise .
if n + 1 / ap
otherwise.
M W
with
with
[April
K
1 =s a < p ^ n
[(lognj/n]"1 .
Incidentally, M(13) = M(15) = 5, contrary to what was asserted in the statement of the
problem.
= 22.
The first place where M(n) > M(n + 1) is n = 83, where M(83) = 23 and M(84)
M(n),
we have
the following n e c e s s a r y conditions for such an n: n + 1 must have three distinct prime fact o r s and n + 2 must not be prime,
Proposer.
DISPLAY CASE
H-201
San Jose,
1
1
1 1
21
1 1
3
- (n = 0, 1, 2, ) .
(ii) Show that, if the columns are multiplied by 1, 2, 3, , sequentially to the right,
then the row sums a r e F 0 i n (n = 0, 1, 2, )
2n+2
(iii) Show that the rising diagonal sums ( / ) are F 2 - (n = 0, 1, 2, " e ) .
Solution
by P.
Bruckman,
(i) Let R
University
n-1
B
n = X+ ]
k=0
of Illinois
n
F2n 2k = X +
"
Y1
k=l
at Chicago,
row,
Chicago,
Illinois.
(n = 0, 1, 2,- ), with R0 = 1.
n
F
2k =
X+
X)
<F
k=l
!974]
219
as asserted.
(ii) Let S
row,
(n = 0,
1, 2, . . . ) , with S0 = 1. Then, if n ^ 1,
n
S
n-1
kF
1C
2n+2-2k +
(n + 1} =
k=0
n-1
= (n + 1] +
(n
k=l
n-1
i=0
as asserted.
(n + X) +
] C (F2n-2i+l - W = (n + D ^
k=0
i=0
n-1
2 3 F2k+2 =
f F 2k+2
k==0
n-i-1
]C
n-k- 1
i=0
rF
F.
4 2 i + l
i=l
= 1 + > (F
- F ) = 1+F
-F
= F
/ j K 2i+2
*2i;
2n+2
2
2n+2 *
i=l
This is also true for n = 0.
(iii) Let T
Then, if n ^ 2,
\2 (n+1)
211
Tn = / J F 4 k
+ ls
if n i s e v e n ;
k=l
V^
4k-2S
if n i s o d d ;
T0 = T t = 1
k=l
m
2m = Yi, F4k + X = ^
2m
k=l
k=l
i=0
also,
T
2m+1
m+1
m+1
7 * F4k-2
k=l
k=l
2m+l
(F
4k-l
4fc.3> = Z ,
( 1)i+lF
i=0
n^ - ^ m =Z^
i=0
<1+1 + 1 >
i=0
2i+i
prefer-
ably typed in double spacing, on a separate sheet or sheets, in the format used below. Solutions should be received within four months of the publication date.
DEFINITIONS
The Fibonacci numbers F
F 0 = 0,
Fj = 1
and
Ln+2
10
n+2
= F ,- +F ,
n+1
n
Proposed
by Maxey
Brooke,
Sweeney,
Texas.
Identify A, E, G, H, J, N, O, R, T, V as the ten distinct digits such that the following holds with the dots denoting some seven-digit number and *p representing zero:
VE RNE R
E
- Rcbcb cbcb J R
HOGGATT
B-281
Proposed
by Verner E. Hoggatt, Jr.,
San Jose State University,
San Jose,
California.
Let T = n(n + l ) / 2 . Find a positive integer b such that for all positive integers m,
n- i
11 1 where the subscript on the left side has m l T s as the digits in base
Proposed
by Herta
T. Freitag,
Roanoke,
Virginia.
2L , n L ,
n+1 n
and
2L 0 + L n _
2n
2n+l
April 1974
B-283
Proposed
Mexico.
by Phil
Manaf
University
of Mew
Mexico,
221
Albuquerque,
New
c <
1000, and c/a as close to 2 as possible. [This approximates the sides of a 30, 60,
90
B-284
Proposed
by Phil Mana,
University
of New Mexico,
Albuquerque,
New
Mexico.
Let z2 - xy - y = 0 and let k, m, and n be nonnegative integers. Prove that:
(a) z
= p ( x , y ) z + q (x,y), where p
and q
and t,
B-285
Proposed
Canada.
Show that
by Barry
Wolk,
University
of Manitoba,
Winnipeg,
Manitoba,
[n/2]
wi>^X>>r(k-1)(n;r)Lr2''
r=0
SOLUTIONS
A LUCAS PRODUCT
B-256
Proposed
by Herta
Show that L 9
Solution
by Gregory
T. Freitag,
- 3(-l)
Virginia.
Wulczyn,
L
Roanoke,
Bucknell
L+1 =
n - 1 n+1
n
University,
Lewisburg,
Pennsylvania.
**n+1+0n+1)e*n-1+0n-1)
= L 2 n + (a2 + / 3 2 ) ( - l ) n " 1
= L2n - 3(-l)n .
Also solved by Wray G. Brady, Paul S. Bruckman,
Juliana D. Chan, Warren Cheves,
Ralph Garfield,
Fe D. Parker,
C. B> A. Peck, M. N. S. Swamy,
Zeitlin,
and the
Proposer.
James
JohnE.
William
D. Bryant,
Tim
Homer,
Graham
E. Thomas,
Jr.,
Carroll,
Lord,,
David
222
[April
A FIBONACCI PRODUCT
B-257
Proposed
by Herta T. Freitag,
Roanoke,
Virginia.
Temple
Philadelphia,
Pennsylvania.
= ("-l . p i - l ^ n + l - , 3 n + 1 ) / 5
n-1
University,
n+1
= (a2n + j3 2 n - ( a / 3 ) n ' V + |32))/5
= (L2n + 3(-l)n)/5
Also solved by Wray G. Brady, Paul S. Bruckman, James D. Bryant, Tim Carroll,
Juliana D. Chan, Warren Cheves, Ralph Garfield, John E. Homer, F. D. Parker,
C. B. A. Peck, M..N. S. Swamy,
William E. Thomas, Jr.,
Gregory
Wulczyn,
David Zeitlin, and the Proposer.
GOLDEN RATIO FORMULA
B-258
Proposed
Illinois.
by Paul S. Bruckman,
University
of Illinois,
a = (1 + \[E)/29
and
Chicago
e
Circle,
= (1 + (-1) )/2.
nF ^, = [naF 1 + e
L
n+1
nJ
n
nF ^ = F ([naF ] + e ) + nF
-F .
nr+n
m L
nJ
n
m-1 n
(b)
Solution
by Graham
Since
n
|nb | < 1.
aF
= F
Lord,
Temple
- - b ,
University,
where
Philadelphia,
b = (1 - \/"5)/2,
Pennsylvania.
But
l'W"5 - l ) / 2 < .65 < 1
and
2.(^5" - l) 2 /4 < 2- (.65)2 < 1 .
The latter inequality verifies the case
n
n|b |<l.
Then
(n + l ) | b n + 1 | < (n + l ) ( . 6 5 ) n + 1 < (n + l)(.65)/n < 1 ,
for n > 2, which completes the induction and the proof of (a).
Equality (b) comes from substituting (a) in the known identity:
Y
= F
m+n
F
+ F
F .
m n+1
m-1 n
Proposer.
if n > 2 then
!974]
B-259
Proposed
223
OF BINOMIAL COEFFICIENTS
by L. Carlitz,
Duke
University,
Durham,
North
Carolina.
(m,r)
/ m - 2\
r 2
/ m - 2 \
r
V
/
'
V
'
/
'
V
/
a r e in arithmetic progression.
Solution
by Paul Smith,
University
of Victoria,
Victoria,
B.C.,
Canada.
(-;)(":,*)-(?:.)
A simple computation yields m = (m - 2r) 2 5 whence m = n2 and r = (m - N r m)/2; 2r
strictly less than m.
is
(n2 - n)/2)}
n>2
Tim Carroll,
Herta
T.
Freitag,
of
Georgia,
SUMS OF DIVISORS
B-260
Proposed
Athens,
by John
Georgia.
L.
Hunsucker
and Jack
Nebb,
University
Show that
by Paul S. Bruckman,
University
of Illinois,
Chicago
Circle,
Illinois.
We may write
r
m =
where the p,
n Pk
k=i
n pk ,
k=l
one has
n =
and f,
mn =
k+fk
n pk
k=l
a r e nonnegative integers.
r
e,
nn (1 + p k + p|. + e e . + p k k ;
k=l
Since
224
April 1974
e,
a(m)/m =
n
k=l
jL*i
3=0
Proposed
Mexico.
by Phil
Mana,
University
of New
Lord,
Mexico,
C.
B.
A.
Albuquerque,
Let d be a positive integer and let S be the set of all non-negative integers
that 2 - 1 is an integral multiple of d.
Peck,
New
n
such
by Tim Carroll,
Western
Michigan
University,
Kalamazoo,
Michigan.
0 E S since d |(2 - 1). Let n be the least positive integer in S when S ^ {o}. For
any positive integer k,
2 kn
(2n
1)(2n(k-l)
+ 2 n(k-2) +
. . . + - 2 n + 1) .
Suppose
m E S and
0 < r < n.
2m - 1 = 2qn2r - 1
= 2qn2r - 2qn + 2qn - 1
= 2 q n ( 2 r - 1) + ( 2 q n - 1) .
on
r
Since q E. S, m E S, and d does not divide 2M , d divides 2 - 1 .