THE FIBONACCI QUARTERLY

THE OFFICIAL JOURNAL OF

THE FIBONACCI ASSOCIATION
VOLUME 8

NUMBER 4

#%
CONTENTS

**

PART I ADVANCED

2K
rV>r

A Theorem Concerning Odd

\
^\P&*
Perfect Numbers . . . . . .>. Suryanar&yana m^^@ker Hagis, Jr. 337
An Addition Algorithm for Greatest Common Divisor / . D. E. Daykin 347
On Determinants Whose Elements
Are Products of Recursive Sequences. . . . . - . <, David Zeitlin 350
An Application of the Lucas Triangle . . . . . . Verner E. Hoggatt, Jr. 360
One-One Correspondences Between the Set N
of Positive Integers and the Sets N n and ^ N N
. Eugene A. Maier 365
The "Difference Series" of Madachy . . . . . . . R. G. Buschman 372
Infinitely Many Generalizations
of Abel ! s Partial Summation Identity .
The Smallest Number with Divisors
A Product of Distinct Primes . . .
Advanced Problems and Solutions

. Kenneth B. Stolarsky 375

K. U. Lu 380

Edited by Raymond E. Whitney 383

PART II ELEMENTARY
Recreational Mathematics . . . . . .
Edited by Joseph S. Madachy
Angle Multisection by Parallel Straightedges .
Janet Waterman Glaze 393
A Note on Fibonacci Functions
W. R. Spickerman 397
Complete Diophantine Solution
of the Pythagorean Triple. .
Fibonacci to the Rescue . . .

.
.

.
.

.
.

.
.

.
.

.
.

A P r i m e r for the Fibonacci Numbers Part VII

.
.

.
.

. Ernst M. Cohn 402

J A. H. Hunter 406

Marjorie Bicknell 407

An Extension of a Theorem of E u l e r .

. Joseph Arkin 421

Fibonacci Series in the Solar System*

. B. A. Read 428

A L u c a s Analogue .

Elementary Problems and Solutions .

. Edited by A. P. Hillman 443

OCTOBER

Brother Alfred Brousseau

439

1970

THE FIBONACCI QUARTERLY

THE OFFICIAL JOURNAL OF THE FIBONACCI ASSOCIATION
DEVOTED TO THE STUDY
OF INTEGERS WITH SPECIAL PROPERTIES
EDITORIAL BOARD
H. L. Alder
Marjorie Bicknell
John L. Brown j J r .
Brother A. Brousseau
L. Carlitz
Ho W0 Eves
H. Wo Gould
A. P. Hillman
Vo Eo Hoggatt, J r .

Donald L. Knuth
George Ledin, J r .
D. A. Lind
Co T. Long
Leo Moser
I. D Ruggles
M. N. S. Swamy
Do E. Thoro
WITH THE COOPERATION OF

P. Mo Anselone

Charles H. King
L. Ho Lange
James Maxwell
Sister M. DeSales McNabb
C. Do Olds
Do W. Robinson
Azriel Rosenfeld
John E Vinson
Lloyd Walker
Charles H. Wall
The California Mathematics Council
All subscription correspondence should be addressed toBro A. Brousseau, St.
Mary's College, Calif. All checks ($6.00 p e r year) should be made out to the
Fibonacci Association or the Fibonacci Quarterly. Manuscripts intended for
publication in the Quarterly should be sent to Verner E. Hoggatt, J r . , Mathematics Department, San Jose State College, San J o s e , Calif. All manuscripts
should be typed, double -spaced. Drawings should be made the same size as
they will appear in the Quarterly, and should be done in India ink on either
vellum or bond paper. Authors should keep a copy of the manuscript sent to
the editors.
T e r r y Brennan
Maxey Brooke
Paul F. Byrd
Calvin D. Crabill
John EL Halton
Richard Ao Hayes
A. F . Horadam
Dov Jarden
Stephen Jerbic
R. P . Kelisky

The Quarterly is entered as third-class mail at the St. Mary's College Post
Office, California, as an official publication of the Fibonacci Association.

A THEOREM CONCERNING ODD PERFECT NUMBERS

D. SURYANARAYANA
Andhra University, Waltair, India
and
PETER HAGIS, JR.
Temple University, Philadelphia, Pennsylvania

1, INTRODUCTION
Although the question of the existence of odd perfect numbers is still
open, many necessary conditions for an odd integer to be perfect have been
established. The oldest of these is due to Euler ( s e e p . 19 in [1]), who
proved that if n is an odd perfect number then n = pa k 2 where p is a
prime, k > 1, (p,k) = 1, and p = a = 1 (mod 4). In 1953, Touchard [6]
proved that if n is odd and perfect, then either n = 12t + 1 or n = 36t + 9.
More recently the first author [5] has established upper and lower bounds for

yl
p

LJ

where n is an odd perfect number. In fact, these bounds are improved ones
over those established in [3] and [4]. For convenience, we give in Table 1
the results of [5] correct to five decimal places.
Table 1
Lower Bound

Upper Bound

Range

(A)

.64412

.67841

.03429

(B)

.65696

.69315

.03619

(C)

.59595

.67377

.07782

(D)

.59993

.66172

.06179

Our objective in the present paper is to improve (some of) the results
of [5]. Our bounds for

p(n
337

338

A THEOREM CONCERNING ODD PERFECT NUMBERS

[Oct.

are given in Theorem 1 while the five decimal place approximations appear
in Table 2. In what follows, n denotes an odd perfect number, and p denotes a prime. The notation
11

2-< p
p=5
for example, will be used to represent the sum
1,1
5
7

1
11

Tsible 2
Lower Bound

Upper Bound

Range

(A)
(B)

.64738

.67804

.03066

.66745

.69315

.02570

(O
CD)

.59606

.67377

.07771

.60383

.65731

.05348

Theorem 1. If n is an odd perfect number, then

(A) if n = 12t+ 1 and 5|n ,
19
log {2 n (P - D / p }
19
y l + _ J 2 z 5
< y 1 < 1 +
~ P
23 log (23/22)
V P
&
p=5
'
p|n
(B) If n = 12t + 1 and 5^n,

59
^- p
p=7

59
log {2 II (p - l ) / p }
P=7
61 log (61/60)

(C) if n = 36t + 9 and 5 n ,

log

< i < log 2 ;

p|n

(50/31)

1970]

A THEOREM CONCERNING ODD P E R F E C T NUMBERS

I + 1 + X + log (256/255)
3 + 5 + 17 + 257 log (257/256)

^
2-

<

1 < !
p
3

+
+

T5

+ w
^

+ lo

339
/*/-n

(65/61)

P|n

(D) if n = 36t + 9 and 5^n ,

I ++ i ++ _L ++ log (80/77)
< ^ l < l
3
7
11
13 log (13/12)
2^ p
3
P|n

i
13

++

i.
17

++ 1l00 gf f

(37349/3094D
l J ' J 4 y /<*U941)

The u p p e r bounds in (B) and (C) a r e due to the f i r s t author

[5] .

The

r e s t of the t h e o r e m i s new.

2.

THE U P P E R BOUNDS

In this s e c t i o n , we shall e s t a b l i s h the u p p e r bounds for

Pjn

given in (A) and (D) of T h e o r e m 1. O u r a r g u m e n t p a r a l l e l s t h a t in [5]

According to E u l e r ? s t h e o r e m , we can w r i t e
a
0
n = pQ

a
2
p2 . -

P l

pk

w h e r e p 0 = a 0 = 1 (mod 4) and a. = 0 (mod 2) for

that Pi < p 2 <

<

Pk
k

(2.o)
so that

1 g j = k.

We a s s u m e

Since n i s p e r f e c t , we have i m m e d i a t e l y
.
a.+l

2 = n i - l/pj
3=0

k
-

n a - i/ P .)
j=0

340

(2.1)

A THEOREM CONCERNING ODD PERFECT NUMBERS

k
k
oo
__ ,
, _ _-^
log 2 =
i +
L
3=0 J
j=l i=l

/(i

[Oct

. .
+ 1
(a +l)i
DPj
" i/iPj 3

+ i/2p20 - i/pfo+1 + f ; i / e + DPS4"1 - i/ip^ ao+1)i .

i=2
Remark 1. Since a. > 2 for 1 ^ j ^ k each term is positive in the
second summation of (2,1).
Remark 2. Since a0 k 1 and i > 2, each term is positive in the last
summation of (2.1).
Remark 3. Since a0 2* 1, we have
l/2p20 - l / p ^ + 1 > -l/2pg .
Remark 4. Since a0 is odd (p0 + l)|o-(p0 )> and since n =cr(n)/2, it
follows that (p0 + l)/2jn and a fortiori that n is divisible by a prime p g
(Po + l)/2.
Remark 5. If p

is the prime mentioned in Remark 4 and p0

>

5,

then
W = l/2p| - l/p*s
For since 3 ^ p

^ (p0 + l)/2, a
s

+ l/2p20 = l / p ^ + 1 > 0 .
<^2, a0 ^ 1, we have

W ^ l/2p g - l/3p2s - l/2p20 ^ 2/3(p0 + l) 2 - l/2p20 > 0 .

We consider first the case n = 12t + 1 and 5Jn. Since 3^n,

we see

from Remark 4 that p0 f- 5. Therefore, pA = 5.

If (p0 + l)/2 ^ 5 i , then we can assume that the p
a

of Remark 4 is

not 5. Since aj ^ 2, it follows from (2.1) and Remarks 1 , 2 , 5

lo

2>

that

00

^ p|nZ j i=l
E ^ + 1)5i+1 - x/i5
i
i
= J j - j - l o g ( l pjn

1+1

l / U + J0

w,-3i

1/5) + log (1 - 1/53)

1970]

THEOREM CONCERNING ODD PERFECT NUMBERS

341

Therefore,
<+log<50/31).
pjn
Since the smallest prime such that (p + l)/2 = 5 m is p = 1249 = 254
- 1, we see that if (p0 + l ) / 2 = 5 m 5 then p 0 > 1249, so that -l/2p
-l/2(1249) 2 . Also, in this case, it follows from Remark 4 that aA ^ 4. From
(2.1) and Remarks 1, 2, 3, we have

log 2 >

p ~ i " log

(1

"

l / 5 ) + log(1

~ l / 5 5 ) " 1 /2(1249) 2 .

Therefore,

1 < 1 + 1/2(1249)2 + log (1250/781) < A + log (50/31) .

pjn
This completes the discussion of the upper bound for this case.

We

remark that the upper bound established in [5] for (A) exceeds ours by 1/2738.
Turning to the case n = 36t + 9 and 5|n we have p^ = 3.

We con-

sider four mutually exclusive and exhaustive possibilities.

F i r s t , suppose that aj = 2 and p 0 = 17. Since cr(32) = 13 and n =
a(n)/2, we see that 13jn. Let 13 = p . If a

= 2 then since o"(132) = 183

and since p 0 + 1 = 18, it would follow from Remark 4 that 3 3 |n. Since this
is impossible, we conclude that a ^ 4. Since a0 ^ 1, it follows from (2.1)
and Remark 1 that

log 2 > J^

i - | - log (1 - 1/3) + log (1 - 1/33) - ^

- log (1 - 1/13)

pjn
+ log (1 - 1/135)

1 - log (1 - 1/17) + log (1 - 172) .

342

[ct-

A THEOREM CONCERNING ODD PERFECT NUMBERS

Therefore,
<

S i
pjn

T5

T7

+ log

( 3 7 3 4 9 /30941) .

Second, suppose that aj = 2 and p 0 = 13. Then the p

of Remark 4
s

is 7, and it follows from (2.1) and Remark 1 that

i - J - log (1 - 1/3) + log (1 - l/3 3 ) - i - log (1 - 1/7)

log 2 > 2
pjn

+ log (1 - 1/73) - A _ log (l _ 1/13) + log (1 - 1/13 2 ).

Therefore,

Sp

<

p|n

T3+log

(21/19) <

Next, suppose that aj = 2 and p 0

T3 TT + log < 37349 / 30941 )

>

17.

As before, we have

13jn,

while p 0 ^ 37. For if p 0 = 29, it would follow from Remark 4 that

which is impossible.

5|n

From (2.1) and Remarks 1, 2, 3, we have

log 2 > ] T i - J - log (1 - 1/3) + log (1 - 1/33) -

- log (1 - 1/13) + log (1 - 1/133) - l/2(37) 2 .

Therefore,

I
p|n

<

T3

2758

+ l0g (78 61)

< 3

17

+ log

(37349/30941) .

1970]

343

A THEOREM CONCERNING ODD PERFECT NUMBERS

Finally, suppose that a4 > 4.

Since p 0 ^ 13, we have -l/2pj;

-1/338. From (2.1) and Remarks 1, 2, 3, it follows that

log 2 >

2 } - I " loS (1 - ^3)

+ l0

pjn

S t 1 " */ 35 > ~ 35s

Therefore,

2Z |

<

3 + ms

+ l0g

< 162 / 121 >

<

JS + T7 + l o S (37349/30941)

This completes the discussion of the upper bound for this case.

3.

THE LOWER BOUNDS

In this section, we change our notation and write simply

a
n

where Pi < p 2 <

<

a
p

'''

\
k

>

p^- We first establish two lemmas.

Lemma 1. If
a

n =

Pl

2
\
P2 Pk

is an odd perfect number and q is a prime such that p N < q ^

N

PM+I>

tnen

iog2< logjnp.A^ -i)} + q i o g 5 4 T ( 2 5 - S 5 : )

j=l
Proof. From (2.0), it follows that

\p|n

3=1

344

A THEOREM CONCERNING ODD PERFECT NUMBERS

2 < 0 (1 - 1 / p r 1

[Oct.

Taking logarithms, we have

N

log 2 < log

k oo

n P, /(P, - i)

+ 5 ] E 1 / i Pi

J=l

]=N+1 i = l

N
=

k oo

log

Pj/(P.

-i +

3=1

i/a^q 1 " 1 )

J=N+1 i = l

oo

= log ri P,' 3 /(P31 i ) + Ya i: S

3=1L

j=N+l

N
N

= log

q/iqi

i=l
/

PJApr i)

J=l

qiog-fT(x;j-E5:)
\ p | n j=l V

The necessary modifications in both the statement and proof of this

lemma in case q S p- o r p, < q are obvious and are therefore omitted.
Lemma 2. The function f(x) = x log x/(x - 1)
ing on the interval

is monotonic decreas-

[2,oo) .

Proof, We easily verify that

Since log (1 + z) < z if 0 < z g 1, we see immediately that f?(x) < 0 if

x > 2,
We are now prepared to prove the lower bounds for

T.JU p
pjn

1970]

A THEOREM CONCERNING ODD PERFECT NUMBERS

345

stated in Theorem 1. We shall defer the proof of (C) until last since it differs
in spirit from the others.
From Lemma 19 we have

log | 2 n (p. - D/p. J

E ^ Ej t JEJ :

p^P

qlog{q/(q- D }

while from Lemma 2 it follows easily that if s is a prime such that s

then

i + -^4(sy7

(3.i)

1)/s

<

<o.

q l o g | q / ( q - 1)}

If n = 12t + 1 and 51n9 then pA = 5. If r is the greatest prime less

than q then it follows from (3.0) and (3.1) that

S^EJf^
pn

+
P

log J 2 n (P - D/P
I P=5
qlog{q/(q - 1) }

An hour's work on a desk calculator shows that the right-hand member of

(3e2) is maximal for q = 23 9 r = 19,

This completes the proof for this

case. We remark that the lower bound for (A) established in [5] is (3.2) with
q = 11, r = 7.
If n = 12t + 1 and 5|n s

then p1 ^ 7.

With r defined as before 9 it

follows from "(3.0) and (3.1) that

<3.3,

log } 2 n (p - D/p

i > A _L_E2
P
P
<HS|V(q-W|
P|

346

A THEOREM CONCERNING ODD PERFECT NUMBERS

Oct. 1970

Some rather tedious calculations verify that the right-hand member of (3.3)
is maximal for q = 61, r = 59. The lower bound for (B) established in [5]
is (3.3) with q = 11, r = 7.
If n = 36t + 9 and 3J[n9 then pj = 3 and p 2 = 7.

With r

defined

as before, we have from (3.0) and (3.1),

r

<>

+
frjxk'Zk
t
^
p

*
2 n (P - i ) / p
! P=7

*y*wi*-\

where the asterisk indicates that the prime 5 is to be omitted from consideration. A few minutes of calculation verifies that the right-hand member of
(3.4) is maximal for q = 13, r = 11. The lower bound for (D) established
in [5] is (3.4) with q = 7, r = 3.
Now suppose that n = 36t + 9 and 5|n.

Then 7|n

by a result of

Kuhnel [2]. We consider three mutually exclusive and exhaustive possibilities.

If either 11 or 13 divides n, then

V i > i +1
Z-^p
3
5
p|n

i > i + I + l
13
3 5 1 7

log (256/255)
257 fog (257/256) "

If neither 11 nor 13 divides n but 17(n, then p 3 = 17 and either

(i) P4 < 251, or (ii) p 4

>

251. In case (i), we have

V i > i +i + i - + J L > + i
L*t p
p|n

17

251

.L

17

log (256/255)

257 log (257/256)

In case (ii), if we take q = 257 in (3.0), we have

Ep|n pl s
[Continued on p. 374. ]

1 , 1 ,
3
5

1 , log (256/255)
17
257 log (257/256)

A N ADDITION: ALGORITHM FOR GREATEST COMMON DIVISOR

D. E. DAYK1N
University of Reading, Reading, England

ABSTRACT
An elementary algorithm is presented ifor finding the greatest common
divisor of two numbers. It is trivial to programme and fast, even for large
numbers. Only addition is used, and the only storage space needed is enough
to hold the two numbers.
About three years ago, I discovered an algorithm which K Y. Choong,
C. R. Rathbone and I used to obtain the first 20,000 partial quotients of the
continued fraction of IT* I here show how an adaptation of the algorithm may
be used to find the greatest common divisor (g.c.d.) of any two positive integers. The complete process when the numbers are 1168 and 2847 is:
N

8832 = 10000 -- 1168

2847 =

1679

0511

9343

N
Z

9854

0219

0073 = g. c. d. = 73

9927

2847

0365

0000

I will now describe the general process. Let P , Q be the two given
positive integers, and suppose that k is the number of digits in the larger of
P and Q. From a finite sequence of numbers, each of not more than k
digits, according to the rules:
k
(a) The first two numbers are 10 - P and Q.
an N number and Q is a Z number.
347

k
Of these 10 - P is

348

AN ADDITION ALGORITHM FOR GRE ATEST COMMON DIVISOR

[Oct.

(b) At each subsequent stage, the next number is the last N number
plus the last Z number.

Carries beyond k digits are ignored.

If there is no such carry, this nextnumber is an N number; otherwise, it is a Z number.

(c) Stop when the next number would be zero.
Than the last Z number is the g. c. d. of P and Q.
We start with 10 - P and it might be argued that this requires a subk
traction. However, we can obtain 10 - P by applying the transformation
0 - 9, 1 - 8, ' , 9 - 0 to the digits of P ,

and then adding 1 to P.

Hence, I am justified in saying that the algorithm only uses addition.

It will be noticed that, as we move down the sequence, the N numbers
begin with more and more nines, while the Z numbers begin with more and
more zeros. Hence the designations N and Z. It is not necessary to evaluate the next number in order to determine whether it is an N or a Z. Suppose the last N and Z numbers, respectively, are
N
z

W i ' "

V k - i -

i
i

Then we look for the largest integer j such that the sum n. + z . of the j
digits is not 9. If there is no such j then the g. c.d. is 1. If there is such
a j , and n. + z.

<

9, we will get no carry and so, by the definition in rule

(b), the next number will be an N number. Moreover, the addition is worked
out step-by-step from the right, and so we can over-write the digits of the
last N number, step-by-step, with the digits of the new N number.
larly, if n. + n,

>

Simi-

9, the next number is written over the last Z number.

Hence, the space for the storage of the 2k digits of P and Q is sufficient
for the complete calculation. Once we have n, = p and z, = 0 we can ignore
these digits, and similarly for n, _-, z, - , etc. Thus the amount of work
required in the additions steadily diminishes, and this is indicated by the
dotted line in the example. The simplest flow diagram is also shown.
I will now prove that the algorithm does produce the g. c. d.
k

stage, we have N = 10 - p with 0 < p < 1 0

and we want the g.c.d.

At each

and Z = q with 0 < q < 10

(p,q). By rule (a), this is certainly the case initially.

1970] AN ADDITION ALGORITHM FOR GREATEST COMMON DIVISOR

349

Case 1. Suppose N + Z would give no carry; that i s , 10 - p + q < 10

- 1 o r q < p. Then the next number will be 10* - (p - q). It will be an N
number bigger than N but less than 10 . Since g. c.d.

(p - q,q) = g. c.d.

(p,q) the next stage will be of the correct form,

Case 2. Suppose N + Z would give a carry; that i s , 10 - p + q> 10
or q ^ p. This carry is ignored, so the next number will be (10 - p) + (q)
- 10

= q - p , and 0 < q - p < q0 Again the next stage will be of the correct

form.

Since the size of Z foes down at each stage, we will reach a stage

where the next Z would be zero; then p = q so that the g.c.d. (p,q) = q in
Z.
I have described all this in terms of arithmetic to the base 10* Clearly,
the algorithm works with any base, and in particular in binary. Sometimes it
is convenient to work to the base 10 but with several consecutive digits of the
numbers in a computer word. With a little more programming effort, one can
speed the process up as followsLet r and s be the largest integers such that n

^ 9 and z ^ 0
r
s
respectively, and suppose that r > s Then, instead of replacing N by N +
Z, it saves work to replace N by N + 10 " ~ Z.

For this operation, we

again only need addition with the appropriate shift* In most cases, we are in
fact able to improve this to replacing N by N + 1 0 r " s Z ,

it is not difficult

to distinguish the exceptional cases. As one would expect, the corresponding

situation obtains if r < s.
Read

,
k into k
10K - P into N list
Q into Z list

Flow Diagram for g, c@ d.

ON DETERMINANTS WHOSE ELEMENTS ARE PRODUCTS

OF RECURSIVE SEQUENCES
DAVID ZEITLIIM
Minneapolis, Minnesota
1.
Let

W0,

Wj,

INTRODUCTION

p ^ 0,

and

q f 0 be a r b i t r a r y r e a l n u m b e r s , and

define

(1.1)

U n = (A11 - B n ) / ( A - B)

(1.2)
(1.3)

p 2 - 4q ^ 0,

W n + 2 = p W n + 1 - qWn,

= An + B n ,

-n

= V

(n = 0, 1, ) ,

/qR,
n' ^ '

w h e r e A ^ B a r e r o o t s of y 2 - py + q = 0.

(n = 0, 1, . . . ) ,

(n = 0, 1, - . ) ,

C a r l i t z [ l , p . 132 (6)] , u s i n g

a well-known r e s u l t for l i n e a r t r a n s f o r m a t i o n s of a q u a d r a t i c f o r m , h a s given

a c l o s e d f o r m for the c l a s s of d e t e r m i n a n t s

W1^
n+r+s

(1.4)

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

As a f i r s t g e n e r a l i z a t i o n of (1.4), we will show that for m = 1 , 2 , ,

and n 0 = 0, 1, ,

(1.5)

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

m(n+r+s)+n 0
=

( _ 1 } (k+l)(k/2) e q ( m n + n 0 ) ( k + l ) ( k / 2 ) + ( m k / 3 ) ( k

k
-l)

3=0

(Wi

- PWQWJ +

qWo)

fc+l)k/2.

n2(k+l-i)

mi
i=l
350

/k\

ON DETERMINANTS WHOSE ELEMENTS ARE

6bl
PRODUCTS OF RECURSIVE SEQUENCES
For m = 1 and n 0 = 0, Eq. (1.5) gives the main result (1.4) of [ l ] . As in

lq70

[l] 5 our proof of (1.5) will require the following known result for quadratic
forms ( e . g . , see [2, pp. 127-128]):
Lemma 1. Let a quadratic form
n
E

n
z2 Q>..x.x.

(a.. = a..)

i] i J

ij

ji

1=1 j=l
be transformed by a linear transformation
n
k=l
to
n

E E c.Y.Y.

(a. = c.)

i=l j=l
Then

(1.6)

K-I = 1 ^ 1 - ^ 1

tt.J

= 1.2.....I1)

2. STATEMENT OF THEOREM 1
We note that (1.5) is a special case of Theorem 1*
Theorem 1. Let W , n = 0,
09 1, 8e%
, satisfy (1*1
(1.1), where A B ^ 0
are the roots of y2 - py + q = 0. Let m , k = 1, 2, , and define

(2.D

= n w
i=i

m n + n

a, = 0 . 1 , . . . ) .

ON DETERMINANTS WHOSE ELEMENTS ARE

352

[Oct.

where n., i = 1, 2, , k, are arbitrary integers or zero. Let N, = nj

+ n2 + + n , . Then, with. u + 1 as the row index and v + 1 as the column
index, we have

(2.2)

n wm(n+u+v)+n.

( u , v = 0, 1, * , k)

i=l

( - 1 ) f e + l ) k / 2 # q mn(k+l)(k/2)+(mk/3)(k

-l)

f]

r=0

. (wl - pwow,

2 (k+1)k 2
/
+ qw 0)

n u^f1"" ,
i=l

with C0 = A

N, -S(j,r)
(2.3)

mi

G/.

BS(J'r)

(r = 1, 2, - . . , k) ,

3=1

(2.4)

S(j,r) = n P + n2(i) + n3(j) + . . . + n j *

H = 1, 2, , Q j ,

where, for each j , S (j, r ) , as the sum of r integers, n/** , i = 1,2,

r , represents one of the ( J combinations obtained by choosing r

,
num-

bers from the k numbers, n l9 n 2 , n 3 , , % .

Remarks. If n. = n 0 , i = 1, 2, , k, then N^ = kn 0 , S(j,r) = rn 0 ,
and
Cr = /k\A(k-r)n0Brn0 ^

Since AB = q, we have

1970]

PRODUCTS OF RECURSIVE SEQUENCES

353

n c r = qn<k+1>k/2 n ft) ,
r=0

j=0 ^ '

and thus (2.2) gives (1.5) as a special case.

For the case rij = n2 = = n, - = d and n, ^ df it is readily seen
that
C

= fk " A A ( k - r ) d B ( r ~ l ) d + n k

/k - l\

A (k-r-l)d+n k B rd

As a footnote to Theorem 1, we have

Lemma 2. For r < k - r , r = 0, 1, , we have

(k)

<2'5)

C C

r k-r = Vrj q

i-

'

2S(j,r)-2S(i,r)

j=2 i=l
Thus,

(2.6)

k
(k-l)/2
II
C
=
II c rr = nn C r C k r
r=0

(2.7)

(k = 1, 3, 5, . . - ) ,

r=0

k
(k-2)/2
n Cr = C k / 2
n
CrCk_r
r=0

(k = 2, 4, 6, ) ,

r=0

where

(2-8)

Ck/2 =

j=l

q S ( J ' k / 2 ) - VN

2s(j?k/2)
k

(k = 2 , 4 , 6, . . . ) .

354

ON DETERMINANTS WHOSE ELEMENTS ARE

[Oct.

Proof of Lemma 2. Since AB = q, we obtain from (2.3),

G)
X

'

qn r\

N k -2S(j,r)

Cr = q S(j,r) A

Noting that a choice of r numbers from

choice of k - r

numbers, we have from (2.3)

(k)

W
Ck_r =

(2.9)

k numbers leaves a complement

(*)

ft

N.-S(j,k-r) S(j,k-r)
W
s(j,r) N.-S(j.r)
B
A *
= Z A
B k

qS*1'1' .

N.-2S(i,r)
k

i=l
In forming the product C C.

, we note that ( ) product pairs have equal

i and j indices and the same value q K

j,

we combine those pairs having the same values of i and j , noting that
S(],P)

For the cross products with i /

N k -2S(j,r)

S(i,r) N . - 2 S a , r )
S(i,r) N -2S(i,r)
B
+ q A

S(j,r) N -2S(j,r)
. q B

S(i,r)-S(j,r)+N^

=q

"V
" ' ^( j . r ^ S f t . r )

Set k = 2r in (2*3)w Since a choice of r numbers from a set of 2r

numbers leaves another set of r numbers, we may again pair off related
terms of the sum in (23)0 Since

N 9 -S(j,r) S(j,r)
A

S(j,r) N
+ A

-S(j,r)

S(j,r)
" *

N2r-2S(j,r)

1970]

PRODUCTS OF RECURSIVE SEQUENCES

and

355

(?) - ( 2 v *) ,

we obtain (2.8) from (2.3) with r = k/2.

3. PROOF OF THEOREM 1
Since A f B9 the general solution to (1.1) is
W n = aA n + bB n ,

n = 09 1, - - ,

where a and b are arbitrary constants whose values satisfy

W0 = a + b

and Wi = aA + bB. We readily find that (B - A)a = W0B - Wt

and (B - A)b

= Wi - AW0. Since A + B = p and AB = q9 we have that

(A - B)2ab = -(wf - pW0Wi + qW(>) -

(3.1)

We observe that
k
(3.2)

P n = n W ^ ^
i=i

where K., j = 09 1 9

=
i

o8e

9k9

K (Bm(k"3)AmJ)n

(n = 0, 1, ) ,

j=o
denote arbitrary constants independent of n,

The quadratic form

r,s=0
(3.3)

r,s=0

k
= K.(Bm(k-j)Amj)n
j=0

k.

j=0
AmJ(r

r,s=0

z
k

= i: K .(B m ( k - j ) A m j ) n x|
j=0

s)Bm(k-j)(r+s)y^

356

ON DETERMINANTS WHOSE ELEMENTS ARE

[Oct.

where

(3.4)

j = E (AmjBm(k-j))rYr
r=0

(j = 0, 1, - , k) .

Thus, by means of the linear transformation (3.4), we have reduced Q to a

diagonal form. If M denotes the determinant of the linear transformation
(3.4), it follows from Lemma 1 (see (1.6)), that
k

(3.5)

= M*. 11 K . C ^ ^ ^ A ^ )

n+r+s

n
J

= M q^(k+l)k/2.n

j=0

j=0

where

M = |(AmJBm(k-)r|

(3.6)

(J,r = 0, 1, . . . , k )

is a Vandermonde determinant.
We find now that
k-1
M=

(A

mr

m(k

r)

- A

mj

m(k

^) = 0

0j<rgk

AmjBm(k-r)(A-B)Um(r

j=0 r=j+l
k-1 k-j

= (A -

k(k+1)/2

. n

TI

AmiBm^'B)T3
ms

j=0 s=l
(3.7)
= (A - B ) k ( k + 1 ) / 2 .

k-1
k-1 k-i
H A mJfc-J) B m(k-J)0c-J-l)/2. n n
j=0

= (A -

B)

k(k+1) 2

i=0 s=l

k
. qiaktf-W/e . ^ 1 - 1
i=l

ms

1970]

PRODUCTS OF RECURSIVE SEQUENCES

357

We proceed now to evaluate

k

K,
3

3=o
of (3,5). From (3.2) we have

(3 8)

xnn + n. = B ^ Z K. ((A/B) 11 " 1 ) J ,

1
i=l
j=0

'

which is a polynomial in the variable (A/B)

. Since W = aA + bB , we

have
W

xnn + n

B ^ ^ V E )

VB*1 ) ,

and thus

k
(3 9)

'

k
W

mn+n

= ^

* ^1 (aA
i=l

i=l

= B

n,kn

WE)1

k
N
k
. A . n

+ bB* 1 )

((A/B) m n + (b/aXB/A)^) .

i=l
Recalling the definition of the elementary symmetric functions of the roots of
a polynomial, we conclude s after comparing (3.8) and (3.9), that (see (23))

r
r

n j>

(3.10) K r = a . A . ( - l ) . I ] ( - b / a ) H (B/A) ' S = a k _ r b r C r

j=l
s=l
(r = 0 , V ik) .

358

ON DETERMINANTS WHOSE ELEMENTS ARE

[Oct.

Using (3.1), we obtain from (3.10)

k

n K = (ab)

k(k+1)/2

nc

r=0

r=0

(3.11)

k
=

WW+I)/*^

_ Br k(k + l) (w 2 _ ^

+ qW 2 o) k(k + l)/2

r=0
Thus, (3.5), with the use of (3.7) and (3.11), gives the desired result, (2.2).
4.

THE CASE p 2 - 4q = 0

In [l] , Carlitzgave an alternate proof of (1.4) for the case p 2 - 4q = 0.

Although (1.4) was proved for the case p 2 - 4q ^ 0, the two results a r e
shown to be the same for the case p 2 - 4q = 0.
In the derivation of (2.2), we assumed that p 2 - 4q ^ 0. It can be
shown (by a repetition of the argument in [l] ) that (2.2) is also valid for the
case p 2 - 4q = 0, where now U = n(p/2) " , and W = (a + bn) (p/2) n ,
with a = W0 and pb = 2Wt = pW0, Since A = B = p / 2 , we obtain from
(2.3) that

cr - (iK*
Moreover, we have

i=l

ujf+l-i)

(mi)2(k+l-i)(p/2)2(mi-l)(k+l--i)

i=l
= mk(k+1)-

k
( p / 2 ) k(kH-l)(mk + 2m-3)/3.

r=0
and

(r|)2^

1970]

PRODUCTS OF RECURSIVE SEQUENCES

k

359

n (i)(n) = (k.) k+1 .

r=0
Thus, from Theorem 1, we obtain the simplified result
Theorem 2

II ( a + bn. + bm(n + r + s)) (p/2)

m(n+r+s)+n. {
1

( r , s = 0 , 1 , - . . ,k)

i=l

(4.1)

nr+iHr/9
(k+l)(k(mn+l)+(2/3)mk(k--l)+(k/3)(mk+2m--3)+N, )
k
= ( - l ) ( k + W i * (p/2)
,k(k+l)
, vk+1
v
u
/(bm)
' /T
(kl)

Remarks, If m = 1 and n. = 0,

i = 1, 2, * , k,

then N, = 0,

and thus (4.1) contains, as a special case, the second (and the last) principal
result, (7), of [ l ] ,
Additional simplifications of (4,1) are readily obtained,
REFERENCES
1.

L. Carlitz, "Some Determinants Containing Powers of Fibonacci Numb e r s , " Fibonacci Quarterly, 4 (1966), 129-134.

2. W. L. F e r r a r , Algebra, Oxford University P r e s s , London, 1941.

3. D. Zeitlin, "Power Identities for Sequences Defined by W
cW , " Fibonacci Quarterly, 3 (1965), 241-256.

+2

= dW

- -

A N APPLICATION OF THE LUCAS TRIANGLE

VERNER E. HOGGATT, JR.
San Jose State College, San Jose, California

1. INTRODUCTION
Consider the integer triangle whose entries are given by

A. Q 1 j

A. . = 2 9

j = l , 2, o, ;

A .- . = A . + A . ,
n+l,j
n,j
n,j-l

(0 < j < n, n > 1).

The first few lines of the triangle are listed left-justified below:

1-2
1 3

1 4

1 5

A:
1

14

16

20

30

25

11

One notes that the recurrence relation is the same as the one for
Pascal's triangle. Apart from no A0

term the array is really the sum of

two Pascal triangles.

The rising diagonal sums are the Lucas numbers,

LA = 1, L 2 = 3, L

= L

+2

- + L . The A 0>0 = 2 would also add L 0 = 2

to the rising diagonal sum sequence. The triangular array is now the Lucas
triangle of Mark Feinberg [1].

It is also closely related to a convolution

triangle [3].
Consider the new array obtained in a simple way from our first array
A by shifting the j

column down j places

umn on the left is the 0

column.
360

(j = 1, 2, 3, ).

The col-

Oct. 1970

AN APPLICATION OF THE LUCAS TRIANGLE

361

B:

14

20

16

The relationship is
B. . = A. . .

0 < j < fi/21 ,

where [x] is the greatest integer not exceeding x. The recurrence relation
for B. . is

B.

=1

for all i ,

1,0
B

i(j =

i-1,3

+ B

i-2.J-1'

along with other useful relations true for all j :

B. . = 2
B Q . ^ . = 2j + 1
B

2j+l,j+l

2, ANOTHER ARRAY
Harlan Umansky [2] laid out the following display of formulas for powe r s of Lucas numbers*

362

AN APPLICATION O F THE LUCAS TRIANGLE

L1
n
L2
n
L3
n
L4
n
L5
n
L6
n
L7
n
L8
n

C:

[Oct.

= L
n
= L9
2n
= LQ
3n
= L,
4n
= L.
5n
= Lc
6n
= I Jr7
7n
= LQ
8n

+ 2(-l)n
+ 3(-l)nL

n
+ 4(-l)nL2
n
+ 5(-l)nL3
n
+ 6(-l)nL4
n
+ 7(-l)nL5
n
+ 8(-l)nL6
n

- 2
- 5L

n
- 9L 2 + 2 ( - l ) n
. n
- 14L 3 + 7 ( - l ) n L
n
n
- 20L 4 + 1 6 ( - l ) n L 2 - 2
n
n

The display given in [2] contains 7 m i s s i n g p a i r s of parentheses., The above

displayed f o r m w a s suggested by E d g a r K a r s t who, along with B r o t h e r Alfred
B r o u s s e a u , noted the typing e r r o r s in [ 2 ] . S u r e l y , we note t h a t exclusive of
s i g n s , the coefficients in display C a r e p r e c i s e l y those of A r r a y B. We shall
p r o v e the t h e o r e m :
T h e o r e m 1.
[m/2]
m
L.J m
n

= L

~~mn

+ -d
Y

^4
3=1

Cm , i v.(-if^-V 1 1n-- 2 3
~m,j'

where

k,0 -

. = C
m,j

Proof.

'
- . + C
0 ,,
m-l,j
m-2,3-1'

1^3^
J

[m/2]
J

for

The proof shall p r o c e e d by induction,, F o r all n ,

i s t r u e for m = 1, the s u m being empty.

A s s u m e , for n > 1,

[k/2]
L

= L . +
nk

JL*

3=1

for k = 1, 2 , 3 , , m

along with

Cy . ( - l ) n J t J - l L k - 2 j
k,j
n

m > 2 ,
the t h e o r e m

1970]
C

AN

k , 0 = *'

APPLICATION OF THE LUCAS TRIANGLE

2k,k = 2>

2k+l,k =

2k + 1

'

**

363

2k+l,k+l = -

Therefore,

[m/2]
V^
Z-t
j=l

mn

( _ 1} nj+j-l
m,jv '

m-2j
n

'

and
Lm+1 = L L
+
n
n mn

[m/2]
Y
C
.(_D^ + J-l L mH-l-2j
Z-r
m,jv '
n

But,
L L
= L, _* + ( - 1 ) ^ ,
,,
n mn
(m+l)n
(m-l)n
Thus,
[m/2]
n

(m+l)n

(m-l)n

Z_/

m,j

Returning to the inductive assumption for k = m - 1 yields

(-DnL,

..

(m-l)n

= (-if^-1 + (-Dn+1
n
= (-DV11-1 +
n

Now let p = j + 1; then since

[Cm-D/2]

y
t^
j=l

A-ifw-1^-1-2)

m-l,j

[(m-l)/2]
V
C
(.^(j+D+O+D-lLm-l-^
Z-^
m-1,j
n
3=1
[(m - l)/2] + 1 = [Cm + l ) / 2 ] ,

[Oct.

AN APPLICATION OF THE LUCAS TRIANGLE

364

(-l)V

. , = (-ifl^-1
(m-l)n
n

[(m+D/2]
V
C ,
^ n p + p - l m+l-2p
Z_<
m-l,p-lv '
n
p=2

Therefore,
[Cm+D/2]
. m+1

(m+l)n

'

L-t
p=2

m-l,p-l

[m/2]
y
C
(.Dnp+p-lLm+l^p
-J
m,p
n
p=l
[(m+l)/2]
y
(C
v
-j
m,p
P=l

(m+l)n

,
.(.^np+p-^m+l^p
m-l,p-l
n

We examine the possible extra term added to the second summation. If m

is 2k, then [m/2] = [(m + l ) / 2 ] = k and C 2 k k = 2 and C 2 k _ x k _ 1 =
2k - 1; thus, C 2 k + 1 k = 2k + 1. If m = 2k + 1, then [m/2] + 1 **
[ (m + l)/2 ] = k + 1 and the term C 2 k+1 , + 1 = 0 and C 2 k
C02k+2,
i . o k+1
i ,i = 2 .
C

k - 1 , 0 = *

= 2; thus

Thus, if one defines

C

2 k , k " 2>

2k+l,k =

2k

'

= 0

P
u

2k+l,k+l

for k > 1, and

m+l,p

m,p

m-l,p-l

1 <

P * [-T-J-

m 1 ,

then

T m+1
L

_
-

(m+l)n

[Continued on p. 427. ]

+
+

[tm+l)/2]
V
p
f ^np-ip-i m+l-2p
L
L.
S n ^ p ^
n
P=l

ONE-ONE CORRESPONDENCES BETWEEN THE SET N OF POSITIVE

INTEGERS AND THE SETS N n AND U N n
EUGENE A. MAIER
University of Oregon, Eugene, Oregon

1. Let N be the set of positive integers and let ]ST be the set of all
n-tuples of positive integers., It is well known that there exist one-one correspondences between N

and N for all N, and between

pNN

and N.

In this paper j we give examples of such functions.

J
2. Theorem 1, Define fn : N n ~* N by

(i)
k=i

where
s

--.
i=l

for k ^ n and the combinatorial symbol (

J is defined to be 0 if m < k.

(?)
h

Then f is a one-one correspondence.

Proof. We begin by defining a relation < on N
Definition,
s

(xj, x\9 , x n )<(x l s x 2 ,

888

as follows?

x n ) if and only if s n

<

or s n = s n and there exists k ^ n such that x! < x, and x' = x for

k < i < n.
M

It is readily established t h a t - < well-orders N . For a E. N , let

= {j3 E N n | p-z< of} and let f (a) = #(M ) where #(M ) is the number

of elements in Ma . Since Ma is a finite set, it follows that f n is a oneone mapping from 1ST onto N. We prove by induction on n that f (a) is
given by (1).
If n.= 1, we have fi(xA) = #{j3 E N | /3 < x j = x4 which is the value
(1) gives for fiCxj). Assume (1) is valid for n. Observe that

(x

i>X2' " ' x n + l ) ( x i ' X 2 '

365

"'xn+l)

366

ONE-ONE CORRESPONDENCES

[Oct.

if and only if
(i)

;+l<Sn+l'

or
(ii)

; + l = 8 n+l

k+l

< X

n+1

or
(iii) sf. , n = s , - , xf , n = x , n+1
n+1' n+1
n+1

and

(x\, , xf ) :z< (x-, , x )

1
n
1
' n'

Thus if
a - (xv

x 2 , , x n + 1 ) ,

M^ may be expressed as the union of three disjoint sets A, B and C which

consist of those elements of N

satisfying, respectively, conditions (i),

(ii), and (iii). Thus,

'fn+1(<*) = tKMa) = #(A) + #(B) + #(C) .
We now compute #(A) +#(B) + #(C). We will have occasion to use the
combinatorial identity,
t+r

(2)

gC-l)-0:j)

j=t+l

(which may be established by induction on r) and the fact that the number of
n-tuples of positive integers which satisfy the equation xj + + x n = t is

(.': 0

1970]

ONE-ONE CORRESPONDENCES

367

(Think of placing t objects in a row and placing dividers into n - 1 of the

t - 1 spaces between the objects. Then xt is the number of objects before
the first divider, x 2 is the number between the first and second dividers,
etc.)
yi

Note that (3 = (yj, y 2 , e e o , Yn+l) i s a n element of A if and only if

+ y2 + . . . + y n + 1 = j where n + 1 j < s n + 1 . Thus,
s ^-1
n+1
#(A) =
i=n4-1
j=n+l

and hence, using (2),

"-(Vii 1 )Now p G B if and only if 1 < y n + 1 < x n + 1 - 1 and

yi

"*

+y

n+l

= X

+ 8e

"

+X

n+1

V l '

Thus /3 G B if and only if yt + + y n = j where SR + 1 j . s n + 1 - 1.

Hence,
S

n+1"1

#(B) =
n
Using (2), we have
n+1
#(B) =

!(;:!)-(;:;)
(-.')-(';)
j=n
"
j=n

Finally, j8 G C if and only if

368

ONE-ONE CORRESPONDENCES
y

n+l

n+1 '

+ Jyn = sn
and
^1' ' " ' ^

( x

l'

"'

The least such fi is the (n + l)-tuple

( s n - n + l , 1, 1, , ! , x n + 1 )
Thus j 3 6 C if and only if
(s n - n + 1, 1, -, 1) =^{yv

, y n ) =^ ( x ^ , x n )

Hence,
#(C) = f (x-, , x ) - f (s - n + 1, 1, , 1) + 1 .
n 1
n
n n
Therefore, using the induction hypothesis and (2), we have

-ft)-5f )-gr-""J
-1

n-1
k=l

%
x

'

n-1
k=s -n
n

*
x

x
n

'

-t(-r)-(\ ')(:;::)
Thus, since

1970]

ONE-ONE CORRESPONDENCES

369

(v-VM^- )-^)
and

(v)-ft::)-Wwe have

W X 1 ' "'xn+l)

and the t h e o r e m i s e s t a b l i s h e d .
3e

T h e o r e m 2.

gCx 1 , , x n )

Define g:

NH> N by

= 2
k=l

k=l

where
k
s

i=l
for k < n and ( ^ J i s defined to be 0 if m < k.

Then g i s a o n e - o n e

correspondence.
Proof.

Define a r e l a t i o n < on

Definition,

(x^, 8 , x j ^ ) <l ( x ^ , x n )

if and only if

sf < s ,
m
n

(i)
v
'
or
(ii)

U . N 1 1 as follows:

and

<

370

ONE-ONE CORRESPONDENCES

[Oct.

or
(iii)

n*

and

The relation < well-orders

* x l ' " ' Xn* ~< * x l ' " '

n* '

U^ N . For a E UN n , let

sa = {/3GNn \p

3a},

and let g(a) = #(S ). Then g is a one-one mapping from

We may express S

onto N.

as the union of three disjoint sets X, Y, and Z which

consist of those elements of

satisfying, respectively, conditions

(i), (ii), and (iii) in the definition of <.

Now p = (yl9 , ym") E X if and only if y 1 + + y m = j where
1

^ 3 ^ s - 1, The number of elements in

satisfying this equation

for fixed j is

mEN \

m=l >

'

Thus
s -1
n

2J X

#( = j ^ " =

2 n

-*

3=1
We have p

Y if and only if y 1 + + y
n-1
#(Y)

= s

where m < n. Thus

sfr:0m=l

Finally, p E Z if and only if p e N11 and j80=^j8 =^ tei* , x n )

where3 p0 is the n-tuple (s -- n + 1, 1, 1, , 1). Thus, using the result
of Theorem 1 and (2), we have

1970]

ONE-ONE CORRESPONDENCES

#(Z) = f n ( x 1 , . x n ) - f n ( s n - n
n-1 /

+ l,l,...,l)

r/

371
+ l =

n-1

W-ENO-IW-E^""^-1)]
i
n-1

k=l

/ J

k=l

s -1
n

.v

+ 1

ten-j--)
Therefore

g(xl9 . . . , x n )

= #(X) + #(Y) + #(Z) =

n-1 ,

,.

n-1

'-. .(t::)-(v)
k=l

k=l

SOME RESULTS IN TRIGONOMETRY

BROTHER L. RAPHAEL, F.S.C.
St. Mary's College, California
G r a p h s of the six c i r c u l a r functions in the f i r s t q u a d r a n t yield

some

p a r t i c u l a r l y elegant r e s u l t s involving the Golden Section.

L e t cp2 + cp = 1,

so that cp = (ST5 - l ) / 2 = 0.61803 and notice that:

a r c cos cp = a r c sin ^ 1 - <j#K= a r c sin ^

a r c sin cp = a r c c o s ^ l F u r t h e r , if tan x = c o s x ,

cp = a r c c o s ^
2

= 0.90459
= 0 o 66621

then sin x = cos x and sin 2 x + sin x = 1,

that

i s 9 x = a r c sin cp in which c a s e t a n a r c s i n < ^ = c o s a r c s i n ^ = cosarccosN/c^ = "fcp

[Continued on p. 392. ]

THE "DIFFERENCE SERIES" OF MADACHY

R. G. BUSCHMAN
University of Wyoming, Laramie, Wyoming

In a recent issue, Madachy [1] has raised two conjectures concerning

difference series which result at the various levels in the Sieve of Eratosthenes;
both conjectures are valid. In this modified sieve, the sieving prime is discarded along with its multiples.

The difference series associated with a

prime is the series of differences which occur between the members of the
remaining list after sievings through that particular prime.
Conjecture 1. If d (d > 2) is the number of terms in one period of
the difference series, then the series is symmetrical about the (d/2)th term.
Conjecture 2. The (d/2)

term (d > 2) will always be 4.

The validity of Conjecture 1 can be argued as follows.

Consider the

numbers which remain after we have sieved by 2, 3, 5, , p n and let

P = 2-3-5- p n . Then, as Madachy shows, there are P + 1 numbers in the
original set of consecutive integers from which one period of the difference
series is formed; we note that both endpoints must be included in order to
generate the differences.

Starting from P - 1 of these integers, sieving by

2 can be done by sieving each second number and beginning at either end,
since 2 divides P. The numbers 0 and P play like roles, as do the n u m bers 1 and P - 1. Similarly, since 3 divides P we can again count from
either end of the original set and sieve each third number.
this for each number through p .

Likewise we do

To illustrate, consider n = 3 so that

P - 1 = 29. We underline to indicate sieving.

29 28 27 26 25 24 23 22 21 20 19 18 17 16 15 14 13 ' '

Thus the difference series begins 6, 4, 2, 4, .

The last element of a

period of the difference series occurs from the pair P - 1, P + 1,

and is

always 2. Since we can apply the process forward or backward over the first
period (deleting this last difference),

the differences are symmetrically

arranged about the "middle difference," i. e . , about the (d/2)

372

term.

Oct. 1970

THE "DIFFERENCE SERIES" OF MADACHY

373

Conjecture 2 can be settled by considering how this "middle difference"

must be formed.

Since d > 2 9 the middle difference must span the number

P / 2 in the original set of consecutive integers 9 which is an odd number;

actually P / 2 = 3-5 p n . Since (P/2) + 1 and (P/2) - 1 are even, this
"middle difference" must be at least 4. From the results of Conjecture 1 9
symmetry then shows that if it is not 4, it must be at least 8, In this case,
both of the odd numbers (P/2) + 2 and (P/2) - 2 must have been sieved out
at some stage; i . e . , by some number 3 9 5 9 , p n . However, if we consider a decomposition of (P/2) 29 we see that this is not possible, for
(P/2) 2 = 3-5 . p n 2

and the remainder is 2 when it is divided by any of these primes, hence it

could not have been sieved out*
It is of interest to note that no sieving with any prime p greater than
3,

although the actual numbers which are sieved out are not regularly

arranged, a ratio of exactly l / p of them disappear,,

In terms of the dif-

ference series this means that within one period of that series formed after
sieving by 29 3 9 59 *e , p , exactly l / p of the pairs of members of tha p r e vious difference series are combined,,

An example for P = 30 illustrates

this.
4 2 4 2 4 2 4 2__4 2 4 2 4 2
6

4 2 4 2 4

4 2

From the 10 differences (of 5 periods) of the previous series we form 2

new differences, a ratio of 2/10 = 1/5 for the sieving prime 5.
A very difficult problem is to try to determine in general exactly which
pairs of differences are to be combined to form the next difference series.
In the above example, pairs numbered 1, 89 1 1 , 18, are combined. For
the next sequences we combine pairs numbered*
(7) 1, 13, 20, 24, 31, 35, 42, 54; 1 + 56,
(11) 1, 27, 32, 42, 47, 58, 739 77, 93, 1 0 3 ,
(13) 1, 35, 44, 51, 62, 77, 84, 99, 110, 1 1 5 ,
(17) 1, 56, 62, 75, 94, 100, 119, 132, 139, 0 0

374

THE " D I F F E R E N C E SERIES" OF MADACHY

No g e n e r a l f o r m u l a for the n

ih

Oct. 1970

difference s e r i e s s e e m s to e x i s t ,
REFERENCE

1. J o s e p h So Madachy, " R e c r e a t i o n a l M a t h e m a t i c s 'Difference

Series'

Resulting from Sieving P r i m e s , " Fibonacci Q u a r t e r l y , 7 (1969), p p . 3 1 5 318,

[Continued from p. 346. ]

If n i s not divisible by

11, 13,

or

17,

then p 2 < 19 ^ p 3 . Taking

17

q = 19 in (3.0), we have

l>

1 , 1 ,

log (16/15)

>

19 log (19/18)

log (256/255)
257 log (257/256) '

p|n
T h i s c o m p l e t e s the proof of the l o w e r bound for (C) and a l s o that of
the new p a r t s of T h e o r e m 1.
REFERENCES
1.

L . E . Dickson, H i s t o r y of the T h e o r y of N u m b e r s , Vol, 1, New Y o r k ,

1934,

2.

U. Kiihnel, " V e r s c h a r f u n g d e r notwendigen Bedingugen fur die E x i s t e n z

von u n g e r a d e n vollkommenen Z a h l e n , " Math. Zeit. 52 (1949), 2 0 2 - 2 1 1 .

3.

M. P e r i s a s t r y , "A Note on Odd P e r f e c t N u m b e r s , " Math. Student, 26

4.

D. S u r y a n a r a y a n a and N. V e n k a t e s w a r a R a o , "On Odd P e r f e c t N u m b e r s , "

(1958), 179-181.
Math. Student, 29 (1961), 133-137.
5.

D. S u r y a n a r a y a n a , "On Odd P e r f e c t N u m b e r s , I I , " P r o c . A m .

Math.

Soc. , 14 (1963), 896-904,

6. J . T o u c h a r d , "On P r i m e N u m b e r s and P e r f e c t N u m b e r s , " S c r i p t a Math. ,
19 (1953), 35-39.

INFINITELY MANY GENERALIZATIONS OF ABEL'S

PARTIAL SUMMATION IDENTITY
KENNETH B. STOLARSKY*
Institute for Advanced Study, Princeton, New Jersey

HI

It is well known that if 2, = 1 B, (x) is bounded independently of m and

x (say for all x in an interval I) and A, tends to zero mono tonic ally as
k oo, then 2,

1 A,

B, (x) is uniformly convergent on I. This follows from

a finite identity first used systematically by Abel, namely,

n
W

E
\
k~m

n-1
B

= snAn " s m - l A m - l

s (A
E
k k " W
k=m-l

'

where

k
\

"LB,
i=l

The purpose of this paper is to show that an infinite sequence of finite

identities involving summations (of which (1) is the simplest example) can be
deduced from the so-called

P. Hall commutator collecting process" which

is fundamental in the theory of finitely generated nilpotent groups.

Let G be the free, group on two generators a and b , {G } its lower
central series

(Gt = G,

Gn+1 = [Gn,G]),

and {^n}

the corresponding

natural homomorphisms, so <>. : G -*G/G . P. HalPs commutator collecting process yields for every g E G an integer r = r(n) such that

(2)

0n(g) = c ? 1 ^ 2 . . . c r r G n ,

* Supported in part by a WARF fellowship, and also by National Science

Foundation Grant GP-7952X.
375

376

INFINITELY MANY GENERALIZATIONS

[Oct.

where the c. E G are the so-called basic commutators (c* = a, c 2 = b ,

c 3 = [ b , a ] , c 4 = [ b , a , a ] , c 5 = [ b , a , b ] , c6 = [ b , a , a , a ] , c 7 = [ b , a , a , b ] ,
c 8 = [ b , a , b , b ] , ) and the e. are integers uniquely determined by g and
n.

A detailed explanation of these concepts can be found in Chapter 5 of

Magnus, K a r r a s s , and Solitar, Combinatorial Group Theory, Interscience

Publishers, John Wiley and Sons, Inc., New York, 1966.
Now let * denote the operator on G which turns words backwards;
e.g.,

(a3b2ab)* = bab 2 a 3 . If <p(g) is given by (2), define 0*(g) by

(3)

tf(g)

= c? c*2

. . . c* r G n .

Since g* can be formed by making the substitutions a a~

and b b~

in g" , it follows that

(4)

0(g) = 0 n (g*) .
T

Similarly, let
e.g.,

3 2

denote the operator on G which interchanges a and b;

(a b ab) = b 3 a 2 ba.

Then,
0^(g) = 0 n (g f ) .

(5)

Equations (4) and (5) provide infinitely many generalizations of (1).

To obtain specific identities from (4) and (5) write g in the form

(6)

g = b a b

b ^m

where the x. are integers. Then g* is obtained from (6) by replacing x.

with xQ ., and gT is similarly obtained by replacing m with m + 1 and
x. with y., where yA = y n , 0 = 0 and y. = x. - for 1 i <> 2m + 1.
l

2m+2

l-l

Tables I, II, and III of the appendix show how to calculate 05(g) from g,
where g has the form (6), and 4>$g)9 0|j(g) from </>5(g), where 05(g) has
the form (2).

1970]

OF ABEL'S PARTIAL SUMMATION IDENTITY

377

Example 1. By equating the exponents of c 3 in (5) s we obtain

r

m-l

/ m

m-l/m-1

X
L 2iJlE 2iJ- Hi I E
**
-j ~2j+lJ^2i
i=0

/ \i=0

(7)

m+1

i=0 \ j=4

m +

m / m

= E d 2 y2j+i) = Z ( S
i=0

\ j=i

i=l\H

By letting
t

u(t) = Y* 2i
X

i=0
and
t
v(t) =

i=0

2i+l

this may be expressed in the more familiar form

m-1
u(m - l)v(m - 1) - )
i=0

v(i

"

1)(u(i)

"

u(i

"

1))

(8)
m

Xu(i"1)(v(i"1}"v(i"2)) f

i=l
which is the discrete analogue of the familiar uv - J v d u = / u d v .
(1) is easily verified from (8).
Example 2. By equating the exponents of c 5 in (4), we obtain

Equation

378

[Oct.

INFINITELY MANY GENERALIZATIONS

m
X

/i-1 X / i - 1
x
X

i=l

\j=0

E 2i(E 2j|(E 2j+l

/\j=0

m-1

m /

/ m
X

1=0

\j=i+l
/m-1

i(2*
21 Z ^
j=0

2j+1 1

/\j=i
\ / m

r > (E
i=0 V ' \ j=i
m-1

X
2J+1II E
Z v ~2j)
2j
/\j=0 /

/m-1

"S 2i(Z

2J+1,
m

2j+i

i=0

J\j=o

It follows by equating the coefficients of Xj i n (9) t h a t

i-1

m
X

2il E X 2 j

(10)
1=1

1=0

V j=0

\j=0 / \ j = 0

^2j

Equation (10) i s a l s o d i r e c t l y o b v i o u s , and can be c o n s i d e r e d a d i s c r e t e a n a logue of

/ u d u = ^- u 2 .
It i s c l e a r that t h e s e i d e n t i t i e s p r o v i d e new t e s t s for the convergence of
infinite s e r i e s , but the author h a s n e i t h e r been able to u s e t h e m to decide
the convergence of any s e r i e s whose convergence i s p r e s e n t l y unknown, n o r
to show that t h e s e i d e n t i t i e s always have i n t e g r a l a n a l o g u e s .
APPENDIX
F o r a g given i n the f o r m (6), Table 1 gives t h e exponents e. of (2)
for

r(5) = 8 , and T a b l e s 2 and 3 give t h e exponents f. and h . of $%(g)

and ^ ( g ) ,

respectively*

If p i s a complicated e x p r e s s i o n ,

note t h e binomial coefficient

(by hand) to r(6) = 14.
t e r m s , one of which i s

(p) shall d e -

( ) T h e a u t h o r h a s extended t h e s e t a b l e s

T h e f o r m u l a f o r e j 4 i s a n unwieldy s u m of five

1970]

OF ABEL'S PARTIAL SUMMATION IDENTITY

m-1

379

SwJ E U
i=0

\j=i+l

where

m-1
p = x

2i E

2j+1

TABLE 1
m-1
ei =

HI

2i+1

i=0

/m-1
x

2i( E

i=0

s =E

2j+l

2i( E

\ /
x

2i E x<V 2 j + l | 2
\ j=i

2J+1H
2j
E
+i)( Lu

2j ! + ^ V 2 M Z ^ ~2j+l
i=0 X
\ j=i
/ \j=i+l

\ H

m-1

/m-1
x

i=0

/m-1
x

i=0

/m-1

e =E
i=0

2i(Ex2j+lJ3
\ j=L

m-1 .

m=l
x

V j=i

m-1
e

2i
2i

i=0

m-1
e3 =

. /m-1

m-1

/m-1

/ m

? = E ffjl E 2j+i)2 E J E 2j+i)J E

i=0

m-1/
e

' \ j=i

\ /m-1

i=0

m-1

\ j=i

\j=L+l

/m-1

x..,||
3 ) l ^ 2j 1 r) ^E(f)(E
= MEff)(E,
1+1 + ^ A ^ 2 lji

m-1
+

[Continued on p. 405. ]

/m-1

EX2i Ex2j+i
i=0

\J=L

x
2j

Z2 . xx
2j

j+

\ /

/\j=i+l

2j/2

THE SMALLEST NUMBER WITH DIVISORS

A PRODUCT OF DISTINCT PRIMES
K. U. LU
California State College, Long Beach, California

1.
L e t P . denote the i

INTRODUCTION

prime.

i s a n u m b e r k such that for k > K,

P

k+s
i

T h i s p a p e r contains a proof that t h e r e

the n u m b e r

-1 P
-1
P
-1
,/k-l+s
-,/k-r+l+s
?2
' p r
""

P
-1
T 1+s
P
k

i s the s m a l l e s t n u m b e r having
P

P
P
k+s k - l + s
1+s

d i v i s o r s , w h e r e & 0,. 1 ^

r k - 1.
2.

LEMMAS

The following L e m m a i s r e p e a t e d l y u s e d in the proof of L e m m a 2.

L e m m a 1.

T h e r e e x i s t positive c o n s t a n t s

9 log 2
and d such t h a t c r log r < P

< d r log r .

L e m m a 2 L e t P . denote the i
l a r g e enough such that for k > K,
p

prime.

we have

k-r+l>

for r = 1, 2 , , k - 1.
Proof.
(1)

See [2 p . 1 8 6 ] ,

F o r r = 1, we do have
2

>

Pk

380

There exists a number K

nnf 1Q70
UCL. aru

T H E

SMALLEST NUMBER WITH DIVISORS

O F D I S T I N C T PRIMES

A P R O D U C T

QI
dS1

for all k.
F o r r = 2 , by L e m m a 1, we have

k-l

3c(k-l)log(k~l)

>

T h e r e e x i s t s a k 2 such that for k > k 2 ,

we have

S ^ " 1 > a^-Dlogfc-l) > d k ^ g k

(2)

By L e m m a 1 and Eq. (2), t h e r e i s a constant k 2 such t h a t for k > k 2 s

we have
P,
(3)

> P,
k

S i m i l a r l y , for

3 < r ^

t h e r e i s a constant k
r

such that for k > k , we have

r

(4)

P^"

r + 1

>

Pk

F o r r = k - 1 , by L e m m a 1 , we have
P

k-1

>

(c(k

"

1)lo

(k "

1))2

= c 2 (k - l ) 2 log 2 OK - 1) .
Hence, there is a constant k ^
(5)

such that for k > k ^ ,

we have

P 2 . _ 1 > c 2 (k - l ) 2 log 2 Os - 1) > dk log k > P k .

S i m i l a r l y , for

o82

THE SMALLEST NUMBER WITH DIVISORS

A PRODUCT OF DISTINCT PRIMES

K^<
there is a constant k

UCX

"

r < k - 2,

such that
p
p
^ k-r+1 > / ,
v k-r+1
P' r
> (cr log r)
>
dk log k
>
\

for k > k .
r
Let K be the maximum of kl9 k 2 , " , k, ^
P

P
r

>

(r

Then for k > K, we have

= 1

2-*^-l) Q.E.P.

Immediately following from Lemma 2, we have

Lemma 3. There is a constant K such that for k > K, we have
p*k-r*Us

> pk >

P.

for r = l , 2, , k - 1, and s > 0, where r < i < k - 1.

Since we know p f B " 1 > p f _ 1 p f "X if A > 1, B > 1, and Pj A > P 2
([1, Lemma 1]), together with Lemma 3, we conclude the following theorem.
Theorem.

There Is a constant K such that for k > K,

P
-1 P
-1
o k+s - V k - l + s

P
-1
,/k-r+l+s

P
-1
-D 1+s

*"

is the smallest number such that it has P k + s P k _ 1 + s * *'

divisors.

k-r+l+s ' ' '

l+s

REFERENCES
J. M. E. Grost,

The Smallest Number with Given Number of Divisors,"

American Mathematical Monthly, Vol. 75, No. 7, 1968.

2. Ivan Niven and Herbert S. Zuckerman, An Introduction to the Theory of
Numbers, John Wiley and Sons, Inc. , Second Edition, 1966.

ADVANCED PROBLEMS AND SOLUTIONS

Edited by
RAYMOND E. WHITNEY
Lock Haven State College, Lock Haven, Pennsylvania

Send all communications concerning Advanced Problems and Solutions

to Raymond E. Whitney, Mathematics Department, Lock Haven State College,
Lock Haven, Pennsylvania, 17745,

This department especially welcomes

problems believed to be new or extending old results. Proposers should submit solutions or other information that will assist the editor.

To facilitate

their consideration, solutions should be submitted on separate signed sheets

within two months after publication of the problems.
H-172 Proposed by David Englund, Rockford College, Rockford, Illinois.
Prove or disprove the "identity, n

t=l
where F

and L denote the n

Fibonacci and Lucas numbers, respectn
n
ively, and [x] denotes the greatest integer function.
H-173 Proposed by George Ledin, Jr., Institute of Chemical Biology, University of
San Francisco, San Francisco, California.
Solve the Diophantine equation,
x2 + y2 + 1 = 3xy

H-174 Proposed by Daniel W. Burns, Chicago, Illinois.

Let k be any non-zero integer and {S }
be the sequence defined
n=l
by S n = nk .
Define the Burn f s Function, B(k), as follows: B(k) is the minimal
value of n for which each of the ten digits, 0, 1, , 9, have occurred
383

384

ADVANCED PROBLEMS AND SOLUTIONS

[Oct.

in at least one S where 1 ^ m < n. For example, B(l) = 10, B(2) =

45. Does B(k) exist for all k? If so, find an effective formula or algorithm for calculating it.
SOLUTIONS
OLDIES BUT GOODIES
The following problems are still lacking solutions:
H-22

H-46

H-74

H-86

H-94

H-IM

H-108

H-115

H-125

H-23
H-40
H-43

H-60
H-61
H-73

H-76
H-77
H-84

H-87
H-90
H-91

H-100
H-102
H-103

H-105 H-110
H-106 H-113
H-107 H-114

H-116
H-118
H-122

H-127
H-130

GENERATING FUNCTIONS
H-144 Proposed by L. Carlitz, Duke University, Durham, North Carolina.
A. Put

[(1 - x)(l - y)(l - ax)(l - b y ) ] " 1 =

] T

A^x3^11

m,n=0
Show that

n=0
B.

n n

'

xn

1 - abx2
( 1 - x)(l - ax)(l - bx)(l - abx)

Put

(1 - xr\l

- y ) ^ ( l - axyr A

y
JL~4

m,n=0
Show that

B
xmyn
J
m,n

1970]

ADVANCED PROBLEMS AND SOLUTIONS

^ B _ x
n,n
n=0
Solution

385

= (1 - x)"" 1 !! - ax)~ X

by the Proposer.

Solution, A.

We have

m
A

m,n

m+1 v M
, n+1 x
(1 - a
)(1 - b
)
(1 - a ) ( l - b)
M

i=0 j=0

so that

n nx

zL*
n=0

" 2-J

n+1
n+1,
(1 - a ^ H l - b " x )
(1 - a ) ( l - b)

n=0
b
(1 - a ) ( l - b) J l - x ~ l - a x ~ l - b x

ab

1-

abx

1
\
1
_
b
|
1 - b | ( 1 - x ) ( l - ax)
(1 - bx)(l - abx)\
1 - abx 2
(1 - x ) ( l - a x ) ( l - bx)(l - abx)

Solution, B .

We have

(1 - x p ^ l - y ) " 1 ! ! - a x y ) " A
00

r , s ,E
t=0

(A) t t r+t s+t

a x
y
~tT

where
(A)t =

(X - D(X - 2) (A - t + 1)

(t > 1)

and

(X)0 = 1,

386

ADVANCED PROBLEMS AND SOLUTIONS

[Oct.

so that
min(m,n)

1^1

,^

= L-d
V 1
t!
t=0

Hence
bo

oo

/x\

oo

t
n=0

n=0

t=0

oo / x \

oo

- E^ 'IX
t=0
n=0
= (1 - x)" 1 (l - ax)" A
Also solved by M. Yoder and D. Jaiswal.
FACTOR ANALYSIS
H-145 Proposed by Douglas Lind, University of Virginia, Charlottesville, Virginia.
If
e

n = PX P 2

P r

is the canonical factorization of n, let A(n) = e- + + e . Show that A(n)

r
th
+ 1 for
n
^ MFn)
^
> where F is the n
Fibonacci number.
Solution by the Proposer.
Clearly, A(mn) = A(m) + A(n), and if mjn then A(m) < A(n).
1 = A(p) ^ ^-( F D ) ^ o r

an

Y prime p.

A(F fc) for all k, except when p = k = 2, when A(4) = A(F4) + 1.

p

Also,

We show by induction that A(p ) S

The

j^

cases when p < 12 are checked directly. Assume the result is true for
k-1
k
p
. Then since p > 12, by CarmichaeFs theorem ("On the Numerical

1970]

ADVANCED PROBLEMS AND SOLUTIONS

387

Factors of the Arithmetical Forms an 0 n , " Annals of Math. (2 n d Ser.), 15,

pp. 30-70, Theorem XXIII) there is a prime dividing F , not dividing F , -.
I
p&
p^-JThen since F , -, F , , we have
pK-l|

pK

A(F . ) > 1 + X(F . - ) *

completing the induction.

k,

Hence A(p ) ^ X(F t ) except when p = k = 2.

In the factorization
e.
e
1
r
n = Px 80 P r >
we can assume pj = 2, and ej = 0 if necessary.
F

e,-

P?

,-', F e
r

pr

are pairwise relatively prime since p i , , p

divides F

Then

a r e , and since F

for each i, so their product

F

ei'-'
Pi

elF
r I

e#

*i
n

Hence,
X(F ) X(F e
V

) = X(F e -,) + + X(F e ) *

P
P
P/
l
r
(et - 1) + e 2 + + e r = X(n) - 1 ,
e

which completes the proof.

^4/so so/ve<2 fry M. 7o<ier.
CONVERGING FRACTIONS
H-147 Proposed by George Ledin, Jr., University of San Francisco, San Francisco,
California.
Find the following limits. F, is the k
Fibonacci number, L^ is
th
the k
Lucas number, 1T = 3.14159, a = (1 + V ) / 2 = 1 . 6 1 8 0 3 - - ,
m = 1, 2, 3, .

ADVANCED PROBLEMS AND SOLUTIONS

388

[Oct.

F^
Xi

n+1

lim
F

F
X2 =
&

X*6 =

lim

-* Q

m
m

F m

lim

n -> o

n
F

XA4

m
m-L
n
F

lim

n o

lim
n ->o

XR

-2

Solution by David Zeitlin, Minneapolis, Minnesota.

EDITORIAL NOTE: We have assumed Binet Extensions,

aX - f
x

a - jS '

'

in the calculations of x 2 , x 3 , , x 5 since we are concerned with neighborhoods of zero I

(1) As n-+oo, F Ja11-> (a - p)

Let p

= F ,-

and q

fa J%1 -> (a

- 0"

= F .

Then, as n->oo,

-+ (a - p)'1

and

- F?

Since aq Q - p _ o, we have xt = (a - /3) a _ 1 = 5 ( a _ 1 ) / 2 * 5 ' 3 0 9 ^ 1.644.

For x real, we define L = a + '(P and F = (<*x - f)/(a
- /3).
X

Let Y., i = 2, 3, 4, 5,
then X. = Y.
i

denote the limits without absolute value signs;

(2) Using L f Hospital 1 s rule, we have (since ap = -1) ,

1970]

ADVANCED PROBLEMS AND SOLUTIONS

Yo =

lim

/xm\ =

/ F

lQ

g <* -

lQ

g P

2 lQ

389
g <* - i*

where i 2 = - 1 , and log (-1) = i77, using principal values. Thus*

X2 = |Y 2 | = ^(4: log2 a -f

TT 2 )/5

(3) Using L f Hospital's rule, we have

ry

,.
x
x*>o F
x

a - p
log a - log p

a -
2 log a - irr

Thus,

mIK) =
(v/f)
- *2 -

3 =A l *

and so

x3 = ir, i = |T,|. iz,r = x 2 | Zs r = ( 4 i o ^ + ff2)"(m"1)/2

(4) We readily find that
Y4 =

lim

(vr **)
/x x l i

] = Y2 Z 3 = 1 ,

and so X4 = 1.
(5) Using L f Hospital's rule,-we have
Y5 =

lini (L
X

>0

- 2)/x = log a + log 0 = iff' ,

X

a n d s o X 5 = | Y 5 | = 7T ,

Also partially

solved by the Proposer,

and also solved by M. Yoder and D. Jaiswal

390

ADVANCED PROBLEMS AND SOLUTIONS

[Oct.

SHADES OF EULER
H-149 Proposed by Charles R. Wall, University of Tennessee, Knoxville, Tennessee.
For s = a + it , let
P(s) = 2 p ~ S ,
where the summation is over the primes.

Set

CO

^ a ( n ) r T s = [1 + P(s)]"1 ,
n=l
00

^ b ( n ) n " s = [1 - P(s)]"1 .
n=l
Determine the coefficients a(n) and b(n).
Solution by the Proposer.
For n = p* i* p^mm
We claim that

let p(n) = a-1 + + am and

, i = a ^I P ]a l . . . pM J =
a(n)
x
'

Mn)(a

Xn) = ( - l ) p ( n ) .

l + . .". ' + a m ) ;
1
m

and that b(n) = |a(n)|.

The proof is by induction on p(n), If p(n) = 1, n is prime and we
have a(n) + a(l) = 0 and the validity of the assertion is obvious. Since in
general, we have
a(n) + a(n/Pi) + * + a(n/p

) = 0 ,

the result follows by induction. A similar method works for b(n),

that here we have

except

1970]

ADVANCED PROBLEMS AND SOLUTIONS

391

b(n) - bdi/pt) - . . . - b ( n / p m ) = 0 .
Also solved by L. Carlitz, D. hind, D. Khmer, and M. Yoder.
TRIPLE THREAT
H-150 Proposed by M. N. S. Swamy, Nova Scotia Technical College, Halifax, Canada.
Show that
n-1

25

E E E F 2 r - l = F4n + < n / 3 >< 5n2 " 14 > '

p=l q=l r=l

where F n is the n

Fibonacci number.

Solution by the Proposer.

To establish this results we need the following identities which have
already been established earlier (Fibonacci Quarterly, December,
pp. 369-372):
5(F2i + F | + . . . + F 2 2 n - 1 ) = F 4 n + 2n
F4

F8

... + F 4 n = F2nF2n+2

5(F 2 F 4 + F 4 F 6 + . . . + F 2 n _ 2 F 2 n = F 4 n - 3n

Hence,

q
5

Or,

2r-1 = %

+ 2c

1966,

392

ADVANCED PROBLEMS AND SOLUTIONS

p

E
E 2r-l
q=l r = l

[Oct.

Z%
1

+2

L1 q = W2

*> + P

Hence,
n-1

25

ZE

n-1

p=l q=l r = l

2r-1

(L*F2pF2p+2

n-1

n-1

X> 2 + 5 I> =
1

= F 4 n - 3n + (5/6)n(n - l)(2n - 1) + (5/2)n(n - 1) =

= F 4 n . + (n/3)(5n 2 - 14) .

Also solved by C. Peck, M. Yoder, A. Shannon, S. Hamelin,

EDITORIAL NOTE.

C. B . A. P e c k , in h i s solution, obtained the identity

n

25

and D. Jaiswal.

EE

2r-l

4n+2

5n(n +

)."-3-

q=l r = l
[Continued from page 371. ]
and a l s o ctn a r c c o s <p = sin a r c c o s cp
below.

Wr

^v

iK
-^2/2 -

=s

Jqi.

The results a r e summarized

RECREATIONAL MATHEMATICS
Edited by
JOSEPH S. MADACHY
4761 Bigger Road, Kettering, Ohio

ANGLE MULTISECTION BY PARALLEL STRAIGHTEDGES

JANET WATERMAN GLAZE
La Sierra High School, Carmichael, California

Trisection of an arbitrary angle is* of course, impossible by means of

compass and unmarked straightedge alone, but the attempt to do this by other
means has fascinated mathematicians since the time of the ancient Greeks.
The simplest trisection is probably Archimedes 1 famous paper strip
construction.

This method involves using a straightedge with two m a r k s ,

known as an application of the "insertion principle. n It is illustrated in

Fig. 1.
The angle to be trisected here is 8. A circle of arbitrary radius r is
drawn whose center O is the vertex of the given angle.

The sides of the

angle intersect the circle at points A and B. BO is extended.

A segment

of length r is marked on a straightedge or paper strip. The edge is placed

so that it passes through A and so that one endpoint of the marked segment
intersects the circle at C and the other endpoint falls at D on BO, outside
the circle. Then m / C D O , here marked a, is one-third m0. The proof
is easily seen:^\ OCD, having two sides of length r , is isosceles, so that
m / C O D = m / C D O = a. By the exterior angle theorem, m/ACO = m/_CAO
= 2a, since / OAC is also isosceles. The given angle 6 forms an exterior
angle of ^OAD. Thus m6 = 2a + a = 3a.
A solution of the problem using parallel straightedges, and a generalization are given here.
In Fig. 2, let the angle to be trisected be 6; a circle of arbitrary radius r is drawn with center at vertex O.

A and B are the intersection

points of the sides of the angle and the circle. K and F represent the two
parallel straightedges; K passes through point A and F passes through
center O. C and E are points where each straightedge intersects the circle,,

Mark point D on straightedge K such that CD = r.

393

Adjust the

394

ANGLE MULTISECTION BY PARALLEL STRAIGHTEDGES

[Oct,

straightedges so that CEj[_BO. When this occurs, point D will fall upon
BO, for the reason that QOCDE is a rhombus. To prove it, let D! be the
point of intersection of K and BO.

AOPC and AOEP are = by H-L;

I CD P = [_ EOP by alt. int. angles; CP = EP by cpctc. Therefore,

= AD'CP bySAA. It then follows that OE = OC = CD',
OE, diagonals are _[_, therefore QOCD'E
Thus m / ODA = a = 1/3 m0.

AOEP

CD* parallel to

is a rhombus, and D = D'.

/_ DOE and the angle vertical to it have

measure a.
It is possible to broaden the scope of the previous method to certain
problems of multisection that of dividing an arbitrary angle into a given
number of equal parts combining the ideas used in the Archimedes t r i s e c tion with the properties of what might be thought of as a set of "collapsing
rhombuses. M Specifically, it makes possible division of a given arbitrary
angle into 2 + 1

equal p a r t s , where n may be any positive integer.

In

fact, the angle may be divided into any number of parts which is a divisor of
a number of the form 2 + 1 for example, if n = 5, it is possible to divide the angle into 33 parts or 11 parts by taking three of the parts each equal
to 1/33 of the angle. The method shown above for trisection represents the
o,

, it may be

seen that any given angle may be divided into 5, 9, 17,

equal p a r t s ,

case where n = 1; that i s ,

2 = 3 .

When n = 2, 3, 4,

respectively.
If n = 2 so that 2 + 1 = 5, we have a 5-sectLon as shown in Fig. 3.
In each case, incidentally, there appear n rhombuses we see two here.
As before the angle to be 5-sected is represented by 6 , the circle is drawn
with radius r and center O. An inserted length equal to r on straightedge
K has endpoints C and D, with C on the circle.
allel to K.

Straightedge F is par-

K and F pass through points A and O, respectively.

E is

the point of intersection of F and the circle. This time D OEDC is a rhombus with diagonals OD and CE (sides CD and OE are equal in length and
parallel). H is the point of intersection of diagonal OD and the circle. K
is adjusted so that HE J[_ BO.

Now using OD as radius,

O as center,

draw a circle concentric to the original. The intersection of F and this c i r cle is point M. Draw DM. It can be seen that DM J_ BO. ODGM is a
rhombus so that OD = DG. If m/DGO = a,

then m / D O G = a.

By the

1970]

ANGLE MULTISECTION BY PARALLEL STRAIGHTEDGES

395

exterior angle theorem m / CDO = m/_ COD = 2a. Now / AGO is an exterior angle to ACOG, so that m / AGO = m / COG + m / CGO = 3a + a = 4a =
m/OAC.
m

The given angle,

69

is an exterior angle to

AAOG,

so that

= 4a + a = 5a. By alt. interior angles, m / GOE = m/DGO = a, which

is the measure also of the angle vertical to /_ GOE.

Thus the angle 6

is

5-sected.
In each case, it may be seen, as in Fig. 4, which shows a 9-section
with 3 rhombuses, that a diagonal of each small rhombus becomes a side of
the next larger rhombus. Each succeeding case is similar to that outlined in
the 5-section, in principle. The properties of the rhombus, in particular its
equal sides and perpendicular diagonals, provide the means for this interesting method whereby the chain of rhombuses could be extended to infinity.

Fig. 1 Archimedes 1 Trisection

396

ANGLE MULTISECTION BY PARALLEL STRAIGHTEDGES

Oct. 1970

Fig. 2 Trisection Using Parallel Straightedges

I (Proved, not Given)

Fig. 4 Nine-Section

A NOTE ON FIBONACCI FUNCTIONS

W. R. SPICKERMAN
East Carolina University, Greenville, North Carolina

Recently, a number of authors [ 1 , 2, 3] have considered Fibonacci

functions continuous functions possessing properties related to Fibonacci
sequences.

In this note, some Fibonacci functions are derived and their

properties verified.

The derivation is based on the following definition,,

Definition: If f is an infinitely differentiable function and f satisfies

the recursion relation:
(1)

f(x + 2) = f(x) + f(x + 1),

then f is a Fibonacci function.

An immediate consequence of the definition i s :
Theorem 1. If f(x) is a Fibonacci function, then fT(x) and Jf(x)dx
are also.
The theorem is established by elementary calculus,
f (x + 2) = [f(x) + f(x + 1)]' =
= ff(x) + f'(x + 1)
/f(x + 2)dx = J[f(x) + f(x + l)]dx =
= /f(x) dx+Jf(x + l)dx.
Theorem 2. If f(x) and g(x) are Fibonacci functions, then their sum
is also.
Proof. Let F(x) = f(x) + g(x). Then
F(x + 2) = f(x + 2) + g(x + 2) = [f(x + 1) + g(x + 1)] + [f(x) + g(x)] =
= F(x + 1) + F(x) .

Theorem 3. If f (x) is a Fibonacci function and c is a real constant,

then cf (x) is a Fibonacci function.
Proof. Let F(x) = cf(x). Then
397

398

A NOTE ON FIBONACCI FUNCTIONS

[Oct.

F(x + 2) = cf(x + 2) = c[f(x + 1) + f(x)] = cf(x + 1) + cf(x) =

= F(x + 1) + F(x) .
Since the function e ^

where p is a real constant, k an integer,

and i = V ^ T is real for integer values of x, we look for Fibonacci functions

dx
of the form y = e
where d is complex. Substitution into the recursion
relation (1) yields
ed(x+2)'- ed(x+1) - e * * = 0

(3)
or,
tA\

dx,,2d

V4)

(d

- e

-v

- 1) = 0

Since 0 is omitted by the first factor of (4),

(5)

- e

-1 = 0

Solving (5) for e :

e

i =-f(i +

N/5)

= a ,

NT5)

= p .

and
ed2 = i ( l Let dj[ = aA + bji, then
a a i (cos bj + i sin bj) = a
Since a > 0, a4 = lno- = 0.48 and bj = 2k7T for k an integer.
if d2 = a2 + b 2 i, then
e 2 (cos b 2 + i sin b 2 ) = ft .

Similarly,

1970]

A NOTE ON FIBONACCI FUNCTIONS

Since p < 0, a2 = In |/31 and b2 = (2k + i ) for k an integer.

more,

399
Further-

1 = | ( e a i c o s 2k)(e a 2 cos (2n + l ) | = | e a i | | e a 2 |

and so a2 = -Ina = -0.48, or a2 = -a 4 . Thus, the subscript on a is not
necessary and two solutions of (1) are:
y(x) = e

cos 2k7JX

and
y(x) = e

cos (2k + 1)TTX

Applying Theorems 2 and 3, we have:

(6)

y(x) = cje

cos 2k7jx + c2e

cos (2n + l)7Tx,

where a = lna; k and n integers. Equation (6) may be written:

/r7,

(7)

, x
(a+2k7Ti)x ,
(-a+(2n+l)7T)x
y(x) = cje'
+ c2ev
,
Some interesting and useful relations between e

and e

rived by substituting the values of dj and d2 into Eq. (5).

e (a+2k7Ti)2 a+2k7ri
e
e2ae4k7ri_eae2k7ri_

2 a

-e

Also,

1 = 0 ,

or
(8)

- x n u

e2a = 1 + ea

can be de-

400

A NOTE ON FIBONACCI FUNCTIONS

e

(-a+(2k+l)7Ti)2

- e

- 2 a 2(2k-M)77i

(-a+(2k+l)iri)
- a 2k7Ti 771

- e e

[Oct.

- 1 = Au
-

-1

= 0n

-2a +, e -a - -1 - n0 .

or
e"2a

(9)

= 1 - e~a

Furthermore,
e a + e ~ a = |or| + |j81 =

(10)

The t r i g o n o m e t r i c identity

\/5 .

cos kn"(x + 2) = c o s kirx = - c o s k7T(x + 1),

r e l a t i o n s (8) and (9), and s o m e a l g e b r a verify t h a t (6) i s a solution to (1).

Since (6) i s a different!able function satisfying r e l a t i o n (1) in view of
T h e o r e m 1,
y ! (x) = (cje

(11)

cos (2n + l)7rx)T

c o s 2k7rx + e 2 e~

and
(12)

Jy(x)dx = ![cte

c o s 2k7rx + c 2 e

cos (2n + l)?Tx]dx

a r e also Fibonacci functions.

The values of ct

and c 2 for which Eq. (6) a s s u m e s the Fibonacci n u m -

b e r s for i n t e g e r x can be computed by applying the conditions y(0) = 0 and

y(l) = 1.

That i s ,

ci + c 2 = 0
(13)
a
c^e

a
- c2e

= 1 .

The solutions to the s y s t e m (13) a r e cj = 1/N/5 and c 2 = - I N / 5 .

T h u s , the

Fibonacci functions that a g r e e with the Fibonacci n u m b e r s for i n t e g e r x

are

1970]

A NOTE ON FIBONACCI FUNCTIONS

y(x) = ( e ^ cos 2k77x - e - a x cos (2n +

(14)

1)7IX)/N/"5

401
.

The function f(x) = (a - b cos 77x)/ N/5 [2] is a special case of (14), where
X

k = 0, n = 0, and a and b are not identified as exponentials base e.

The usual extension of the Fibonacci sequence to the negative integers
satisfies the relation F_

= (-1)

F . For integer values of x, the Fib-

onacci functions (14) have the same property*

Since
cos 2kn77 = (-1)

= 1 j

and
cos (2kn + n)7T = ( - l ) 2 k n ( - l ) n = (-l) n
we have
NH> y(-n) = e " a n cos 2k7r(-n) - e a n cos (2n + l)7T(-n) =
-an
e

, .,11 an
. -Ji+1, an
-an>
- (-1) e
= (-1)
(e
- e
) =
= (-l) n+1 y(n) N/5

REFERENCES
1. M. Elmore, "Fibonacci Functions, 11 Fibonacci Quarterly, 5 (1967), pp
371-382.
2. F. Parker, "A Fibonacci Function," Fibonacci Quarterly, 6 (1968), pp.
1-2.
3. A. Scott, "Continuous Extensions of Fibonacci Identities," Fibonacci
Quarterly, 6 (1968), pp. 245-250.

COMPLETE DIOPHANTINE SOLUTION OF THE PYTHAGOREAN TRIPLE

(a, b = a + l,c)
ERNST M. COHN
National Aeronautics and Space Administration, Washington, D. C.

In connection with problem B-123 (Fibonacci Quarterly 5 (1967), p. 238)

the question was raised whether Pell numbers provide the only possible Diophantine solutions for the Pythagorean triple (a,b = a + l , c ) . To prove that
this is indeed so, it is necessary and sufficient to show that the general solution for the Pythagorean triple, when modified for this special case, acquires
a form identical with relationships that are characteristic for Pell numbers,
P .
n
The proof is based on a property of a class of sequences, of which both
the Fibonacci and the Pell sequences are particular cases.

By applying the

recursion formula that is specific to the Pell sequence, an identity for the
general sequence is transformed into one that is valid only for the Pell
sequence. It is precisely this identity that must be satisfied by the special
Pythagorean triples.
We start with the single-membered, purely periodic, infinite continued
fraction
J_ " ' '
g+

= -g

\/g2 + 4
2

where g is a positive integer. The limiting value shown is the positive root,
x l 3 of x2 + gx - 1 = 0 The numerators and denominators of the convergents
are obtained by the well-known recursion formula
(1)

n + 2 = SG n + l

+ G

(G = 0, G, = 1) .

These numbers appear in the powers of, e. g. ,

, Vg 2
2

-x 2 =
402

+ 4

nnf

COMPLETE DIOPHANTINE SOLUTION

PYTHAGOREAN T R I P L E ( a , b = a + l , c )

1Q7n

Oct. y/u

OF THE

403

Now

(3)

(f

+ G

n+3

n +2 ) 2 = t

^3

SGn+3Gn+2

n+2

and

(4)

(Gn+3 ^ f l ) 2

= i G^+3

+ gGn+3Gn+2

G ^ G ^

By subtracting (4) from (3) and using (1), we obtain

<5)

-<Gn+l "

n+2Gn> =

n+2 "

n+3Gn+l

(the left side being simply a t r a n s f o r m a t i o n of the right),

Since (5) holds for all values of n ?
constant.

each side m u s t equal the s a m e

Upon substituting G 0 = 0 and Gj = 1 on the left (note that the

value of G2 = g i s not r e a l l y needed), we find - ( l 2 - g-0) = - 1 , and hence

<6>

n+2Gn "

When . g = 1,

n+1 = ^

* G n+1 G n "

n+1

the left side reduces to the well-known Fibonacci-number

identity
F ^F
- F 2 , . = N( - l ) n + 1
n+2 n
n+1
When g = 2,

the right side of (6) can be rewritten as the Pell-number

identity
(Pn + P n+1
^)

- 2P*n+1 = N( - l') n + 1 9,

404

COMPLETE DIOPHANTTNE SOLUTION

[Oct.

whence

(7)

Pn

P n + 1 = V2P^+1

(-l) n + 1

Also, since the recursion formula for Pell numbers is

P

n+2 "

2P

n+l

+ P

n ( P o = *

* =

from (1) above, we have

n+2

n+1

n+1

These equations have thus been shown to be characteristic for Pell numbers,
The Diophantine solution for the Pythagorean triple, with legs a and
b and hypotenuse c, is 2 p q = a or b, p2 - q2 = b or a, and p 2 + q2 =
c,

where p and q are positive integers of different parity and p > q.

When b = a + 1, p 2 .- q2 - 2pq = 1. Solving for p and q, and rearranging t e r m s ,

(9)

p + q = V 2 P 2 . 1

(10)

p - q = V2Q2 * 1

Since (7) and (8) are obviously equivalent to (9) and (10), p and q must be
Pell numbers. In fact, when q = P then p = P n + 1 The even leg of the
triangle is
2P , - P = a or b ,
n+1 n
n
n
the odd leg,
p2
- p2 = b or a ,
n+1
n
n
n
and the hypotenuse,

1970]

O F THE PYTHAGOREAN T R I P L E
p2

p2

n+1

(a,b = a + l , c )

^2n+l

405

Also, the smaller leg is

2n
Pm
_
S
m=l

= an or bn .

Except for the lowest nontrivial value 3, the values for both legs are obviously
composite numbers.

[Continued from p. 379. ]

TABLE 2
fi = ej

f2 = e2

f4 = e 3 - eie 2 - e ^ + e 4 + e2 (

) f 5 = -ete2 + e 3 - a%ez + e 5 + e J

f8 = e ^ - e 3 + 2ete 3 - 2e 4 - 2e 2 ^
f7 = eAe2 - e 3 + e t e 3 - e 4 - e 2 1

f3 = ete2 - e 3

J + e ^ - e6 - e 2 ^
+ e

j - e3(^

hA = -e K + e i (

J + e2e5 - e 8 - e 3 ^

h7 = - e 7 +

h2 = e t

-e*

/ ~e*\ 3 /

h3 = e ^ - e 3
+ e

( 2 )( 2 )

TABLE 3
hi = e2

2 e 3 "" e 5 "" e i ( 2 )

+ eie 5 - e 7 + e 2 e 4 - e j e ^
f8 = eje 2 - e 3 + 2e 2 e 3 - 2e 5 - 2eA

2( 2 J

_e8 + e i

\ 3 /

FIBONACCI TO THE RESCUE

J. A. H. HUNTER
Toronto, Ontario, Canada

Every year my PUZZLER problem in the December issue of the magazine Canadian Consulting Engineer is made the subject of a contest, the
prize being won by the first acceptable theoretical solution that is opened on
a specified date. This last December, in 1969, the problem was as follows:
Charlie was through with the paper,,

HeTd read the comics,

studied the sports pages, and even glanced at the headlines of world
news, And now he was doodling on the margin of the front page.
"It's funny about all our a g e s , " he said suddenly.

"Yours is in

the same proportion to mine as mine is to one less than our two ages
combined. "
Mary had been watching the two children playing with the boy f s
new scooter on the sidewalk outside.

"What of i t ? " she asked.

"A lot

of fractions."
"I mean the complete y e a r s . " Her husband smiled.

"But the

funny thing is that the same applies to the ages of the two k i d s . "
Do you know MaryT s age ?
Many acceptable solutions were received from readers.

But one of

these, not the first opened however, was most ingenious in its use of the
Fibonacci concept for dealing with the relevant diophantine equation. For this
reason, I give it in full, as received from the originator, Michael R. Buckley
of Toronto, Canada.
We notice Fibonacci lurking between the lines.
Charlie T s ages be x and y respectively.
i =
x

y y

Let Mary T s and

Then:

Obviously, the larger x and y a r e , the more golden becomes

their ratio. The simplest solution is x = y = 1: here x and y being
[Continued on p. 420. ]

A PRIMER FOR THE FIBONACCI NUMBERS: PART VII

MARJORIE BICKNELL
A. C. Wilcox High School, Santa Clara, California

AN INTRODUCTION TO FIBONACCI POLYNOMIALS

AND THEIR DIVISIBILITY PROPERTIES
An elementary study of the Fibonacci polynomials yields some general
divisibility theorems, not only for the Fibonacci polynomials, but also for
Fibonacci numbers and generalized Fibonacci numbers. This paper is intended also to be an introduction to the Fibonacci polynomials.
Fibonacci and Lucas polynomials are special cases of Chebyshev polynomials, and have been studied on a more advanced level by many mathematicians,, For our purposes, we define only Fibonacci and Lucas polynomials.
1. THE FIBONACCI POLYNOMIALS
The Fibonacci polynomials {F (x)J- are defined by
(1.1)

Fi(x) = 1,

F2(x) = x,

and

F n + 1 (x) = xF n (x) + F ^ ^ x )

Notice that, when x = 1, F (1) = F , the n

th

Fibonacci number. It is easy

to verify that the relation

(1.2)

F_ n (x) = (-l) nHLl F n (x)

extends the definition of Fibonacci polynomials to all integral subscripts. The

first ten Fibonacci polynomials are given below:
Fi(x) = 1
F2(x) = x
F3(x) = x2 + 1
F4(x) = x3 + 2x
F5(x) = x 4 + 3x2 + 1
F6(x) = x 5 + 4x3 + 3x
F7(x) = x 6 + 5x4 + 6x2 + 1
407

408

A PRIMER FOR THE FIBONACCI NUMBERS

[Oct.

F8(x) = x 7 + 6x5 + 10x3 + 4x

F9(x) = x8 + 7x6 + 15x4 + 10x2 + 1 F10(x) = x 9 + 8x7 + 21x5 + 20x3 + 5x .
It is important for Section 4, to notice that the degree of F (x) is | n| - 1 for
n ^ 0. Also, F0(x) = 0.
In Table 1, the coefficients of the Fibonacci polynomials are arranged in
ascending order.

The sum of the n

row is F , and the sum of the n

diagonal of slope 1, formed by beginning on the n

row, left column, and

going one up and one right to get the next term, is given by
2 (n-l)/2

2 t 2 (n-3)/2

when n is odd.
Table 1
Fibonacci Polynomial Coefficients Arranged in Ascending Order
n

1
0

0
2

1
0

1
0

7
8

1
0

4
5

X1

X2

X3

X4

X5

X6

X7

1
0

6
0

0
10

5
0

1
0

10

15

4
0

1
0

To compare with Pascal 1 s triangle, the sum of the n

the sum of the n

row there is 2 , and

diagonal of slope one is F . In fact, the (alternate) dia-

gonals of slope 1 in Table 1 produce Pascal 1 s triangle.

If the successive binomial expansions of (x + 1)
scending order,

are written in de-

19701

A P R I M E R FOR THE FIBONACCI NUMBERS

409

n = 0.
n = 1
n = 2
n = 3:
n = 4;

the s u m of the 4th diagonal of slope 1 i s F 4 (x) = x 4 + 3x 2 + 1,

the nth diagonal of slope 1 i s F (x),

and the s u m of

or5

[(n-D/2].

F (x)
n

(1.3)

j=0

for

[x] the g r e a t e s t i n t e g e r contained in x9

and binomial coefficient

[ .j .

a s given by Swamy [ l ] and o t h e r s .

20

LUCAS POLYNOMIALS AND GENERAL FIBONACCI POLYNOMIALS

The L u c a s polynomials

(2.1)

L 0 (x) = 2 5

| L (X)} a r e defined by

Li(x) = x 9

Ln+1(x) =xLn(x) +Ln^1(x)

A g a i n 9 w h e n x = 1, L (1) = L , the n

th

L u c a s n u m b e r . L u c a s polynomials

have the p r o p e r t i e s that

(2.2)

L n (x) = F n + 1 W
xL

n(x)

n+2(x) "

F^W
F

= xFn(x)

n-2(x)

and can be extended to negative s u b s c r i p t s by

(2.3)

-n

(x) = (-1)"L (x) .

n

2Fn_1(x)

410

A PRIMER FOR THE FIBONACCI NUMBERS

[Oct.

The degree of L (x) is Jn|, as can be observed in the following list of

the first ten Lucas polynomials:
L4(x)

= x

L2(x)

= x2 + 2

L3(x)

= x2 + 3x

L4(x)

= x 4 + 4x2 + 2

L5(x)

= x 5 + 5x3 + 5x

L6(x)

= x 8 + 6x4 + 9x2 + 2

L7(x)

= x 7 + 7x5 + 14x3 + 7x

L8(x)

= x 8 + 8x8 + 20x4 + 16x2 + 2

L9(x)

= x 9 + 9x7 + 27x5 + 30x3 + 9x

L10(x) = x10 + 10x8 + 35x6 + 50x4 + 25x2 + 2 .

If the Lucas polynomial coefficients are arranged in ascending order in a leftjustified triangle similar to that of Table 1, the sum of the n
and the sum of the n

row is L ,

diagonal of slope 1 is given by 3 - 2

for even

n, n ;> 2.
When general Fibonacci polynomials are defined by
(2.4)

Ht(x) = a,

H2(x) = bx,

HR(x) = x H ^ x ) + Hn_2<x) ,

then
(2.5)

H (x) = bxF

n '

-(x) + aF
n-1^

Q(x)

n-2N

If the coefficients of the | H (x)|, written in ascending order, are placed in a

left-justified triangle such as Table 1, then the sum of the n
slope 1 is

(a

b) 2< n " 3 >/ 2 = (a

b)-2t<n-2^

diagonal of

1970]

A PRIMER FOR THE FIBONACCI NUMBERS

for odd n, n > 3.

(Notice that, if a = 2, b = 1,

411

then H ^ ( x ) = L (x),
n+1
ir "

and if a = b = 1, H (x) = F (x).)

n
n

3. A MATRIX GENERATOR FOR FIBONACCI POLYNOMIALS

Since Fibonacci polynomials appear as the elements of the matrix defined below, many identities can be derived for Fibonacci polynomials using
matrix theory, as done by Hayes [2"] and others, and as done for Fibonacci
numbers by Basin and Hoggatt [3],
It is easily established by mathematical induction that the matrix

( ;

! )

when raised to the k

power, is given by

(3.1)

k
/ Fk+l(x)
K
Q = ' K l
~ F k (x)

k(x)

F k _ l ( x)

for any integer k, where Q is the identity matrix and Q~

k

inverse of Q . Since det Q = - 1 , det Q

F

(3.2)

k+l(x)Fk-l(x) -

= (det Q) = (-l)

k(x)

( 1)k

"

Since Q m Q n = Q m + n for all integers m and n,

Q

and Q

'

is the matrix
gives us

'
matrix multiplication of

gives

/Fm

^ n

A nv
Q
Q

( x )
W x ) + Fm(x)Fn(x)
+ l
F
m<x)Fn+l(x) + Fm-l<x)Fn<x)

(x)F
n<x) +
m+l
F
(x)F (x) + F

while
nm+n
y

Fm+n+l(x)
W
m+n

" I F

W x )

., .
m+n-lix),

m(x)Fn-l<x)

m-l<

x)F

n-l(xV

412

A PRIMER FOR THE FIBONACCI NUMBERS

[Oct.

Equating elements in the upper right corner gives

(3.3)
v
'

F _, (x) = F ^Ax)F (x) + F (x)F - (x) .

m+n '
m+1
n
m
n-1

Replacing n by (-n) and using the identity

F_ n (x) = (-l) n + 1 F n (x)
gives
F

m-n(x)

( 1)n

<-Fm+l(x)Fn(x)

+ F

m(x) W

'

Then,
F j . (x) + (-l) n F
(x) = F (x)F Ax) + F (X)F' - x(X)
m+n
m-n
m
n-1
m
n+l
= F

(x)L (x) .
n

If we replace n by k and m by m - k
(3.4)

FmW

= L k (x)F m _ k (x)

above, we can obtain

+

(-D k + 1 F m _ 2 k (x)

which results in the divisibility theorems of the next section

4. DIVISIBILITY PROPERTIES OF FIBONACCI AND LUCAS POLYNOMIALS
Lemma. The Fibonacci polynomials F

(x) satisfy

(2i+l)k-2im ( x ) '

, ,p(m-k)+m+l
(v
F
(-1)
(2p-l)m-2pk ( X )

for all integers m and k, and for p ;> 1.

Proof: If p = 1, the Lemma is just Equation (3.4).
call Q (x) the sum of Lucas polynomials in the Lemma.
the Lemma holds when p = j , or that

For convenience,
Then, assume that

1970]

A PRIMER FOR THE FIBONACCI NUMBERS

<A>

V x > = *m-*WQj

+ ( 1)3(m

413

"k)+m+lF(23-l)m-2jk

Substitute (2jk - (2j - l)m) for m in Equation (3.4), giving

F

2Jk-<2j-l)m W = L k . W P 2 J k - C l J - D m - k . W

Since we want to express

2'k-(2"-l) ^

^^'^Vl^-Dm^W

i nternis

_k^9

S6t

2jk - (2j - l)m - k ! = m - k

and solve for k f 9 yielding kf = (2j + l)k - 2jm, so that
F

2jk-(2j-l)m ( x )

= L

(2j+l)k-2jm ( x ) F m-k ( x )

+ (-1)

(2j+l)m-(2j+2)k (x) '

Substituting into (A) and using

F -n (x) = ( - l ) n + 1 Fn(x)
to simplify gives
F m (x) = (Q.(x)

(-D3(m-k)L(2j+1)k2jm(x))Fmk(x)
(j+l)(m-k)+m+l
+ { 1}
"
*(2j+l)m-(2j+2)k W '

which i s the Lemma when p = j + 1, completing a proof by mathematical

induction.
Notice that the Lemma yields an interesting identity for Fibonacci numb e r s , given below:

'

p 1

To establish (4.1), use algebra on the subscripts of the Lemma and then take
x = l.

414

A PRIMER FOR THE FIBONACCI NUMBERS

T h e o r e m 1: Whenever a Fibonacci polynomial F

Fibonacci polynomial F

, (x), m f k ,

[Oct.

(x) i s divided by a

of l e s s e r o r equal d e g r e e , the r e -

mj

m a i n d e r i s always a Fibonacci polynomial o r the negative of a Fibonacci p o l y n o m i a l , and the quotient i s a s u m of L u c a s polynomials w h e n e v e r the division
i s not exact. Explicitly, for p > 1,
(i) the r e m a i n d e r i s F / r > 1X
(x) when
01
(2p-l)m-2kpv
2plmj
2p + 1

o r , equivalently, if

,,.
IKI

m._2p(m.-k)

JH-

>

(x)

(2p - 2) M
2p - 1

for

lm?

<im - k I<
K

2p! + l ^ I

2p - 1

(ii) the quotient i s L,(x) when | k | < 2Jm!/3;

(iii) the quotient i s given b y

Q P - E M) i ( m - k , L ( 2 1 + 1 ) k . 2 t o - L <- i<m " k) ^., a+1 ,<,- k ,

i=0
for m , k ,

i=0

and p r e l a t e d a s in (i), and by Q (x) + ( - l ) p ( m " k )

_2PHL
2p + 1 '
(iv) the division i s e x a c t when

2 m
P
2p + 1

oo rr

Kk

(2p

" 1)m
2p

Proof: When

2ElmL>|k|

2p + 1

i i

(2p - 2)jm|
2p - 1

if

1970]

A PRIMER FOR THE FIBONACCI NUMBERS

415

and the degree of F m (x) is greater than that of F m _ k (x), we can show that
|m| > |m - k| > |(2p - l)m - 2pk|. Since the degree of F n is |n| - 1, we
can interpret the Lemma in terms of quotients and remainders for the r e s t r i c tions on m, k, and p above, establishing (i), (ii), and (iii). As for (iv), the
division is exact if

ir -

(2

P -

!)

for then
F/o

o , (X) = F 0 (X)
u

(2p-l)m-2pk
When

2pm
2p + 1

F/o
-n Q , (x) = F.k-m (x) = ( - l ) m ~ k + 1 F m-k
. (x)
(2p-l)m-2pk

- '"^X^w
because k is even. Referring to the Lemma, increasing the quotient by
, -.pdn-kj+m+l+m+l _ , -vp(m-k)
will make the division exact.
Corollary 1.1: F (x) divides F (x) if and only if q divides m.
Proof: If q divides m, then either m/2p = q or m/(2p + 1) = q,
Let q = m - k and apply Theorem 1.
If F (x) divides F (x), then let q = m - k and consider the remainder
q
m
of Theorem 1. Either
F

(2p-l)m-2pk ( x )

(X)

or
F

giving

(2p-l)m-2pk< x ) = F m - k ( x ) '

416

A PRIMER FOR THE FIBONACCI NUMBERS

k

(2p-l)m
2p

(2p - 2)m
2p 1

Qr

[Oct.

2pm
2p + 1

by equating subscripts. The possibilities give

q - m - k = m/2p o r q =
m/(2p - 1) or q = m/(2p + 1), so that q divides m.
Corollary 1.2: If the Fibonacci number F

is divided by F

,, m f

k, then the remainder of least absolute value is always a Fibonacci number or

its negative.

Further,

(i) the remainder is F 0 , , v when

m-2p(m-k)
T < | m - k | -<2p1
2pM
+ 1 - | - - - l

M.

m-M2,

and the quotient is the sum of Lucas numbers;

(ii) the quotient is L, when |k| < 2 | m | / 3 , for Lucas number L, 0
Proof: Let x = 1 throughout Theorem 1. Since the magnitudes of Fibonacci numbers are ordered by their subscripts, F

, v represents a

remainder (unless m - k = 2 since F 2 = FA = 1).

To illustrate Corollary 1.2, divide F 1 3 by F 7 :
233 = 17 . 13 + 12 = 18 13 + (-1) .
Now, 12 is the remainder in usual division, but we consider the positive and
negative remainders with absolute value less than that of the divisor, so that
(-1) = - F j is the remainder of least absolute value. Here, m = 13, k = 6
< 2 m / 3 , p = 1, and the quotient is L 6 = 18. The remainders found upon
dividing one Fibonacci number by another have been discussed by Taylor [ 4 ] ,
and Halton [5] k
Corollary 1.3: The Fibonacci number F

divides F

if and only if q

divides m, jq|' f 2.
Proof: If q divides m, let x = 1 in Corollary 1.1. If F divides
F m , let q
,v
J 1.2 becomes F
0 ,
^ = m - k. The remainder of Corollary
m-2p(m-k)
= F 0 = 0 or F __? , , v = F . . The algebra on the subscripts follows
the proof of Corollary 1.1, which will prove that q divides m, provided that
there a r e no cases of mistaken identity, such a s F

= F ,
s

|sl r |ql> and

1970]

A PRIMER FOR THE FIBONACCI NUMBERS

417

such that s does not divide m9 Thus, the restriction jqj f 2 since
F 2 = Fi = 1.
Unfortunately, as pointed out by E. A, Parberry 5 Corollary 1.3 cannot
be proved immediately from Corollary 1.1 by simply taking x = 1.
divides

does not imply that F (x) divides

That F

(x), just as that f(l)

divides g(l) does not imply that f(x) divides g(x) for arbitrary polynomials
f(x) and g(x). Also, Webb and Parberry [8] have proved that a Fibonacci
polynomial F

(x) is irreducible over the integers if and only if m is prime.

But, if m is prime, while F

ber F , I qj ^ 3, F

is not divisible by any other Fibonacci num-

is not necessarily a prime. How to determine all-val-

ues of m for which F

is prime when m is prime, is an unsolved problem,

Corollary 1.4: There exist an infinite number of sequences | s \ having

the division property that, when S

is divided by S

, , m / k,

the r e -

mainder of least absolute value is always a member of the sequence o r the

negative of a member of the sequence,,
Proof: We can let x be any integer in the Lemma and throughout Theorem 1. If x = 2, one such sequence is . . . , 0, 1, 2, 5, 12, 29, 70, 1 6 9 , " *
Theorem 2: Whenever a Lucas polynomial L (x) is divided by a Lucas
polynomial L

, (x), m ^ k,

of l e s s e r degree, a non-zero remainder is

always a Lucas polynomial or the negative of a Lucas polynomial.

(i) non-zero remainders have the form
2p|m|
,. ,
2p + 1 > ! K ' >
or, equivalents,

2p(mk)-m

2p + l

(x)

(2 D _i) i n _2Dk^

Explicitly,
wnen

(2p - 2)1 ml
2p - 1

for

- 2p - 1 '

(ii) if | k | < 2 f m | / 3 , the quotient is Lk(x);

(iii) the division is exact when k = 2pm/(2p + 1), p ^ 0.
Proof: Since the proof parallels that of the Lemma and Theorem 1 ? details are omitted.

Identity (3.4) is used to establish

418

A PRIMER FOR THE FIBONACCI NUMBERS

(4.2)

L <x) = L, (x)L . (x) + ( - l )

m
k
m-k '

Since L

-n

k+1

Q1

[Oct.

(x) .

m-2k

(x) = (-1) L (x), it can be proved that

n
L m N(x) = Q
L(2p-l)m-2pk
-v 0 , v(x)' ,'
x
/0
^ p N(x)L
' m - .k (x)

for I m | 2 > J m - k | j > |<2p - l)m - 2pk|.

Since the degree of

L (x) is | n | ,

the rest of the proof is similar to that of Theorem 1. However, notice that it
is necessary to both proofs that F_ (x) = +F (x) and L

(x) = L (x).

Corollary 2.1: The Lucas polynomial L (x) divides L (x) if and only
if m is an odd multiple of q.
Proof: If m = (2p + l)q, let q = m - k and Theorem 2 guarantees
that L (x)
divides L (x).
qw
nr
If L (x) divides L (x), then let q = m - k.
exact, the term L, 2

-*

, (x) must equal L

For the division to be

, (x) since it cannot be

the zero polynomial. Then, either k = 2pm/(2p + 1) or k = 2pm/(2p-1), so

q = m - k = m/(2p + 1) or q = m/(2p - 1). In either case,

m is an odd

multiple of q.
Corollary 2.2: If a Lucas number L

is divided by L

,.> then the

non-zero remainder of least absolute value is always a Lucas number o r its

negative with the form L2D/m._u-\

\m\

for

< im . ki<

2p + 1 ^ '-

K|

2p - 1 '

and the quotient is L, when | k | < 2|m|/3.

Proof: Let x = 1 throughout the development of Theorem 2.
Corollary 2.3: The Lucas number L
(2s + l)q for some integer s.

divides L

if and only if m =

(This result is due to Carlitz |j>3).

Proof: If m = (2s + l)q, let x = 1 in Corollary 2.1. If L

divides

, take q = m - k and examine the remainder LQ , . v

of Corollary
m
4p\mK;m
2.2 which must equal L , or L.
since it cannot be zero. The algebra
^
m-k
k-m
follows that given in Corollary 2.1. Since there are no Lucas numbers such
that L

= L

restrictions.

where | q | f\ s j , and since L

f 0 for any q, there are no

1970]

A PRIMER FOR THE FIBONACCI NUMBERS

Since the g e n e r a l i z e d Fibonacci polynomials

b x F ^ - (x) + a F ^

(4.3)

(x)

419
satisfy

(x) =

(x), we can show that

Hm(x) = Lk(x)Hm_k(x)

(-Dk+1Hm-2k(x) >

but s i n c e H m ( x ) f H _ m ( x ) , we have a m o r e l i m i t e d t h e o r e m .
T h e o r e m 3: Whenever a g e n e r a l i z e d Fibonacci polynomial

(x)

is

divided by H

, (x), 2 m / 3 > k > 0, any n o n - z e r o r e m a i n d e r i s always a n mvL

o t h e r g e n e r a l i z e d Fibonacci polynomial o r i t s n e g a t i v e , and the quotient i s
Lk(x).
As a consequence of T h e o r e m 3 , when a g e n e r a l i z e d Fibonacci n u m b e r
H

i s divided by H , a n o n - z e r o r e m a i n d e r of l e a s t absolute value i s g u a r -

anteed to be a n o t h e r g e n e r a l i z e d Fibonacci n u m b e r only when jm - q| < 2 m / 3 .

T a y l o r [4 J h a s p r o v e d t h a t , of all the g e n e r a l i z e d Fibonacci s e q u e n c e s
satisfying the r e c u r r e n c e H

= H

- +H

_2,

JH

the only s e q u e n c e s with the

division p r o p e r t y that the n o n - z e r o r e m a i n d e r s of l e a s t absolute value

are

always a m e m b e r of the sequence o r the negative of a m e m b e r of the s e q u e n c e ,

a r e the Fibonacci and L u c a s s e q u e n c e s .

F o r y o u r f u r t h e r r e a d i n g , Hoggatt

[ 7 ] gives a lucid d e s c r i p t i o n of divisibility p r o p e r t i e s of Fibonacci and L u c a s

numbers.

REFERENCES
1.

M. N. S. Swamy,

" F u r t h e r P r o p e r t i e s of M o r g a n - V o y c e P o l y n o m i a l s , "

The Fibonacci Q u a r t e r l y , Vol. 6, No. 2 , A p r i l , 1968, pp. 167-175.

2.

R i c h a r d A. H a y e s , Fibonacci and L u c a s P o l y n o m i a l s , M a s t e r 1 s T h e s i s ,
San J o s e State C o l l e g e , J a n u a r y , 1965, pp. 36-39.

3.

S. L . B a s i n and V e r n e r E. Hoggatt, J r . , "A P r i m e r on the Fibonacci

Sequence P a r t n , n Fibonacci Q u a r t e r l y , Vol. 1, No. 2,

April,

1963,

pp. 61-68.
4.

Laurence Taylor,

" R e s i d u e s of F i b o n a c c i - L i k e S e q u e n c e s , "

Fibonacci

Q u a r t e r l y , Vol. 5, No. 3 , O c t o b e r , 1967, pp. 298-304.

5.

J . H.

Halton,

"On Fibonacci R e s i d u e s , " Fibonacci Q u a r t e r l y , Vol. 2 ,

No. 3 , Oct. , 1964, p p . 217-218.

420

A PRIMER FOR THE FIBONACCI NUMBERS

Oct. 1970

6. Leonard Carlitz, "A Note, on Fibonacci Numbers, M Fibonacci Quarterly,

Vol. 2, No. 1, February, 1964, pp. 15-28.
7. Verner E. Hoggatt, J r . , Fibonacci and Lucas Numbers, Houghton-Mifflin
Mathematics Enrichment Series, Houghton-Mifflin, Boston, 1969, pp. 37-47.
8. W. A. Webb and E. A. Parberry, "Divisibility Properties of Fibonacci
Polynomials," Fibonacci Quarterly, Vol. 7, No. 5, D e c , 1969, pp. 457463.
[Continued from pe 406. ]

the first two terms in the Fibonacci series 0 Who could resist the
temptation to test the conjecture that y/x = F - / F ?
Now let x = kF , y = kF , - . Then,
n J
n+1
F

n + l / F n = !MFn + F n + 1 ) - l ] / k P n + 1 .

so
k(F

n+l " V W = "Fn

but
F

n+1 " FnFn+2 = ("1)U

hence n is odd, and we have k = F . So,

x = F | m _ 1 = 1, 4 , 25, 169, etc. ,
y = F2m-lF2m

= 1

'6'40'273'etc-

Hence, the children were 4 and 6 years old, Charlie 40, and Mary 25.

AN EXTENSION OF A THEOREM OF EULER

JOSEPH ARKSN
Spring Valley, New York

Leonhard Euler, the great mathematician of the 18th Century, wrote

some of the greatest works ever read by man. Among the numerous mathematical interests of this genius was the study of the problem of partitions.
A partition of an arbitrary positive integer, say n, is a representation
of n as the sum of any number of integral parts. For example, the number
6 has 11 partitions, since
6 = 5 + 1 = 4 + 2 - 4

+ 1 + 1 = 3 + 3 = 3 + 2 + 1 = 3 + 1 + 1 + 1

= 2+2 + 2 = 2 + 2 + 1 + 1 = 2 + 1 + 1+1+1 = 1+1+1+1+1+1 .

(Problem. Show that the number of partitions of 7 is equal to 15.)
It is apparent that to find the partition of some positive integer n using
the same methods we used to find the solution for n = 6 above is a clumsy
and difficult procedure.

To overcome this difficulty, Euler combined the

partition with a generating function which led to his discovering a powerful

recurrence formula with which to attack the study of partitions.
(For those readers who wish to discover how Euler found his recurrence
formula, the author recommends they read a chapter on partitions in any
good book on number theory. One such book is An Introduction to the Theory
of Numbers by Niven andZuckerman, published by John Wiley and Sons, Inc.)
EULER'S FAMOUS RECURRENCE FORMULA is usually written as
follows:
p(n) = p(n - 1) + p(n - 2) - p(n - 5) - p(n - 7) + . . .
+ (-l) k + 1 p(n - fk(3k - 1)) + (-l) k + 1 p(n - | k ( 3 k + ) + ,
where p(0) = 1 and n,k = 1, 2, 3, .
To understand the above theorem (1)$ let us go back to the beginning of
this paper where we have shown that the number n = 6 has 11 partitions.
421

422

AN EXTENSION OF A THEOREM OF EULER

[Oct.

Then, what the formula says in a very precise way is that p(6) = 11, or
more generally, p(n) is the number of ways an arbitrary positive integer,
say n, can be partitioned into equal or distinct parts.
Now that we know what p(n) means, let us find out how to use the
formula (1) and in so doing, it will become evident how greatly the genius of
Euler reduced the work required in order to find a solution for each p(n)
(n = 0, 1, 2, ) .
We define p(0) = 1 so that in (1) when n = 1, we have
p(l) = p(l - 1) = p(0) = 1 ,
and now that we have found a value for p(l), we are in a position to find a
value (or the number of partitions) for p(2) since for n = 2, we have
p(2) = p(2 - 1) + p(2 - 2) = p(l) + p(0) = 1 + 1 = 2 .
We continue in this exact way, using the information that p(0) = 1, p(l) = 1,
and p(2) = 2 to find a value for p(3) and then step-by-step values for p(4),
p(5), p(6), and so on.
In the following are examples of how to find values for the p(n) when
n = 1, 2, 3, , 7 by using Leonhard Euler 1 s very important recurrence
formula (in (1)): We define p(0) = 1, then

p ( l ) = P(1 - 1) = p(0) = 1,
p(2) = p(2 -- 1) + p(2 - 2) = p ( l ) +p(0) = 1 + 1 = 2 ,
P(3) = p ( 3 -- 1) + p(3 - 2) = p(2) + p ( l ) = 2 + 1 = 3 ,
p(4) = p ( 4 -- 1) + p(4 - 2) = p(3) +p(2) = 3 + 2 = 5,
p(5) = p ( 5 -- 1) + p(5 - 2) - p ( 5 - 5 ) = p(4) +p(3) - p ( 0 )
p(6) = p ( 6 -- 1) + p(6 - 2) - p(6 - 5) = p(5) + p(4) - p ( l )

(Compare this way of finding p(6) = 11 with the way we showed that the number n = 6 has 11 partitions at the beginning of the paper, and you will

1970]

AN EXTENSION OF A THEOREM OF EULER

423

realize the magnificence of Euler f s formula as a systematized labor-saving

device, especially for large n where his formula is really needed/)
For our last example, we find p(7):
p(7) = p(7 - 1) + p(7 - 2) - p(7 - 5) - p(7 - 7) =
= p(6) + p(5) - p(2) - p(0) = 11 + 7 - 2 - 1

= 15 .

Continuing in this exact way, we may, of course, step-by-step find values for
p(8), p(9), p(10), , p(n)9 , where n runs through the positive integers n = 8, 9, 10, , to infinity.
It is evident that Euler 1 s great recurrence formula (1) systematized the
study of partitions,, However, to determine the values of p(n) for still large
n required an enormous amount of work. (For example, to show that p(243)
= 133978259344888, we must first find a value for each p(n) (n = 0, 1, 2,
888

, 242) from p(0) through p (242) inclusive.) To this end, in what follows

of this paper, we show how to greatly reduce the work required in finding
values of the p(n) by applying a new theorem from a paper entitled "Recurrence Formulas," by Joseph Arkin and Richard Pollack (The Fibonacci Quarterly, Vol. 8, No. 1, February, 1970, pp. 4-5).
In fact, using formula (1) of "Recurrence Formulas" and applying the
method that has been found by this author to formula (1) so greatly reduces
the work involved that to find the value of, say, p(243), it would only be
necessary to know the value of each p(0) through p(122). The reduction in
work is evident, since in Euler 1 s recurrence formula, to find a value for
p(243), we must first find a value for each p(0) through p(242).
To explain the new method of determining the value of any partition
(p(n)), we shall, as examples, find the values of p(16) and p(17).
Then, to find the value of p(16), we set up the table on the next page.

EXPLANATION OF HOW TABLE WAS MADE

1. On row A we have placed the values of (-1)

^k(3k T 1) for k = 1,2,3,

2. We then take half of 16 to get 8 and so we write under column a the consecutive numbers from 8 through 16.
3. Now, next to the number 8 (under column a) we place under column b the
value p(16 - 8), next to the number 9 (under a) we place under column b
the value p(16 - 9) and so on to complete column b with p(16 - 16) = p(0).

424

AN EXTENSION OF A THEOREM OF EULER

A

1a

-7

12

15

b
p(6)

-p(3)

-p(D

p(7)

-p(4)

-p(2)

pL(f p(6)

-p(5)

-p(3)

11 P(5)

-p(6)

-p(4)

12 p(4)

-p(7)

-p(5)

p(0)

1 13 p(3)

-P(6)

P(D

14 p(2)

-p(7)

p(2)

8 p(8)

[Oct

p(7)

9 p(7)

15 p(l)

p(3)

p(0)

16 p(0)

p(4)

P(D 1

4. We fill in the rest of the table in the same way we plot a graph.

For

example, the p(7) under the column where A = 1 and on the row where
a = 8 is said to be in box (8,1) or more generally this p(7) is found in
box (a, A) where a = 8 and A = 1. Now, it will be observed that in each
box (a, A) we find the term p(a - A), or in box (a, -A) we find the term
-p(a - A) except that there are no terms =Fp(a - A) entered at all when
. 16 = 8 ^ a - A < 0.
Let us consider a few numerical examples of what was just said.
Written into the five boxes (8,1),

(8,2), (13,-7), (13,15), and (16,7) we

find respectively the following: p(8 - 1) = p(7), p(8 - 2) = p(6), -p(13 - 7) =

-p(6), no entry (since 13 - 15 = -2 < 0) andno entry (since 16 - 7 = 9 >8).
Now that we have filled in the table, we then multiply each partition
under the column b in the table together with the sum of the partitions directly to its right and on the same row.

We then have the following products

(row-by-row) i
p(8)[p(7) + p(6) - p(3) - p(l)]
p(7)[p(7) - p(4) - p(2)]
p(6)[-p(5) - p(3)]
p(5)[-p(6) - p(4)]
p(4)[-p(7) - p(5) + p(0)]
p(3)[-p(6) + p(l)]
p(2)[-p(7) +p(2)]

19701

AN EXTENSION OF A THEOREM OF EULER

425

p(l)[p(2) +p(0)]
p(0)[p(4) + p(l)]
and after replacing the p( ) with their numerical values, we find p(16) as
follows:
22(15 + 1 1 - 3 - 1 ) =
15(15 - 5 - 2) =

22-22 =

484

15-3

120

11 (-7 - 3) = -11-10 = -110

7 ( - l l - 5) =

-7-16 = -112

5(-15 - 7 + 1) =

- 5 - 2 1 = -105

3 ( - l l + 1) =

-3*10 =

-30

2 (-15 + 2) =

-2-13 =

-26

1(3 + 1) =

T4

1(5 + 1) =

1-6

= _ _6

231 = p(16)

Total

To find a numerical value for p(17),

we use the exact methods that

were used to find a numerical value for p(16). The important difference is
that since 17 is an odd number, we must then take half of 17 - 1 and then
complete the following table using the same methods that we used to complete a table for p(16) (that i s , we shall begin by writing under column a the
consecutive numbers from 8 through 17, e t c ) ,
To find the value of p(17), we erect the following table:
A
a
8
9
10
11
12
13
1 14
15

-5

p(7)

p(6)
p(7)

-p(3)
-P(4)
-P(5)
-P(6)
-p(7)

-7

12

15

p(9)
P(8)
p(7)
p(6)
p(5)
p(4)
p(3)
p(2)

1 16 p(l)

[17 p(0)

-p(l)
-p2)
-P(3)
-p(4)
-P(5)
-p(6)
-p(7)

p(0)
p(l)
p(2)
p(3)
p(4)

P(0)

PW
PTBF ~R(H

426

AN EXTENSION OF A THEOREM OF EULER

[Oct.

We now multiply each partition under column b in the table together

with the sum of the partitions directly to its right and on the same row to get
the following products:
p(9)|Vp(7) + p(6) - p(3) - p ( l ) ]
p(8)|; P ( 7 ) - p(4) - P(2)]
P(7), ; - P ( 5 ) - P(3)]
P(6)[; - P ( 6 ) - P(4)]
P(5), ; - P ( 7 ) - p(5) + p(0)]
p(4) ; - P ( 6 )

+ p(D]

P(3) ! -p(7) + P(2)]

p(2)|> ( 3 ) + P(0)]
P ( l ) |> ( 4 ) + P ( l ) ]
P(0)|> ( 5 ) + P(2)]

and after replacing the p( ) with their numerical values, we find p(17) as
follows:
30(15 +

1 1 - 3 - 1 ) =

22(15 - 5 - 2) =

30-22 =

660

22-8

176

15(-7 - 3) = -15-10 = -150

1 K - 1 1 - 5) = -11-16 = -176
7(-15 - 7

+ 1) =

5 (-11 + 1) =
3(-15 + 2) =

- 7 - 2 1 = -147
-5-10 =

-50

-3-13 =

-39

2(3 + 1) =

2:4

1(5 + 1) =

1-6

1(7 + 2) =

1-9

Total

297 = p(17) .

In conclusion, it may be interesting to mention that we could have used

smaller p( ) to find p(16) and p(17).

For example: Since p(0) through

p(8) will determine p(16) and by the methods used in this paper, it is evident that the numerical values of p(0) through p(4) will enable us to find
values for p(0) through p(8) then we need only have used the values of the

1970]

AN EXTENSION OF A THEOREM OF EULER

427

partitions p(0) through p(4) to find p(16). This reduction rule is applicable
in finding the value of any p(n), however it defeats the purpose of the method
to reduce too much.
Of course, applying the method of this paper to find small partitions'
like p(16) or p(17) does not show the method to its fullest but when used
to find a value for large partitions, like say, p(243) = 133978259344888, the
method shown in this paper very greatly reduces the work involved.

[Continued from p. 364.]

and the induction is complete. Thus the C-array is precisely the B-array.
Thus, !B
= C
, and further, the pattern observed by Umansky and
m,p
m,p
Karst persists for all n > 1 , m ^ 2. The case m = 1 was earlier verified.
Theorem 2 (Independent). If one ignores the signs of the coefficients in
Array C, then the sum across the m

row is L .

Proof. Interchange the first column on the right with the column on the
left and set n = 1. The left column is now -L

and all of the terms on the

right are negative. Equality still holds since Theorem 1 is true.

[m/2]
1 +

[m/2]

>
y> cC -.. == X
'
j=l
Z-/

Thus

m,j

j=o

C . = L
m,j
m

REFERENCES
1. Mark Feinberg, "Lucas Triangle," Fibonacci Quarterly, Vol. 5, No. 5,
D e c , 1967, pp. 486-490.
2a Harlan Umansky, "Letters to the Editor," Fibonacci Quarterly, Vol. 8,
No. 1, Feb. , 1970, p. 89.
3, V. E. Hoggatt, Jr , "Convolution Triangles for Generalized Fibonacci
Numbers," Fibonacci Quarterly, Vol. 8, No0 2, M a r . , 1970, pp. 158171.

FIBONACCI SERIES IN THE SOLAR SYSTEM

B. A. READ
Captain, Canadian Armed Forces
U. S. Army Satellite Communications Agency, Fort Monmouth, New Jersey

ABSTRACT
The Fibonacci Series is shown to predict the distances of the moons of
Jupiter, Saturn and Uranus from their respective primary.

The planets are

shown to have a trend which follows the Fibonacci Series with individual offsets
attributed to planetary densities.
1. INTRODUCTION
Many series exist where successive terms are a function of previous
terms* When this function is a linear combination, each term can be
expressed as
n
z

= V * a.z . + C
L< ] n - j
3=1

where {a.} is a weighting function acting upon each successive

{z _ . } , and

C is a constant.
The Fibonacci series {f } = {0, 1, 1, 2, 3, 5, 8, 13, } which is of
the form
(1)

= f
n

, + f o
n-1
n-2

is an excellent example of this with {a.} = {l, 1, 0, , 0} and C = 0.

Subsequently, however, a modified Fibonacci Series will be employed
having the form

{l, 2, 3, 5, } but with again Eq. (1) holding.

2. PLANET MOONS

The distances of moons from their parent planet have not been known to
follow any sequence. The mean distances z of the moons of Jupiter, Saturh,
428

Oct. 1970

FIBONACCI SERIES IN THE SOLAR SYSTEM

429

and Uranus, as shown in Figs* 1 through 5, respectively, can be assigned to

terms in the Fibonacci Series f
(2)

in the form

m f n + b

The following table shows the "best fit" values of m and b for Jupiter,
Saturn, and Uranus and the possible values for Mars and Neptune which, because they have only two moons each, cannot be fitted to a linear relationship.
Table: Values of m and b
(103 km)

Primary

No. of Moons

Mars

m"
m
(14

r-4.8

2.3

f 3.5

5.8

Jupiter

12

240

-50

Saturn

50

:L40

Uranus

70

60

Neptune

420

j--66

250

(:104

Equation (2) is expressible in the form

z n = zn-1- , + z n-20 - b ,
comparable to Eq. (1).
3.

PLANETS

Figure 6 shows the mean distance of the planets from the sun,

z ,

plotted against the Fibonacci Series f . As can be seen, the trend of the distances follow the series, although the individual values seem to be offset in a
sinusoidal manner from it. Figure 7, however, shows the quantity z /f

for

FIBONACCI SERIES IN THE SOLAR SYSTEM

430

[Oct.

20-

/
/*5

16

12

/ * 4
w
o
o

8 /

m
iH

= 240 f -- 50 (10 3 km)

n

/ l
0

- 4 nt

Fig. 1 First Moons of Jupiter

1970]

FIBONACCI SERIES IN THE SOLAR SYSTEM

431

400-

11,12:/

200 /6

100 70

/
40

20 /

z
10 -

= 240 f -- 50 (10 3 km)

n

7 -

4 -

2 1

i
13

\.
21

L-

34

Fig. 2 Moons of Jupiter

1
55

-J

89

l . -,.

144

432

FIBONACCI SERIES IN THE SOLAR SYSTEM

/
'2

= 50 f

+140

(10 3 km)

F i g . 3 F i r s t Moons of Saturn

[Oct.

1970]

FIBONACCI SERIES IN THE SOLAR SYSTEM

1 I1

_L_

13

21
f

433

= 50 f + 140
n

34

Fig* 4 Moons of Saturn

55

89

144

J
233

I
377

FIBONACCI SERIES IN THE SOLAR SYSTEM

434

64-

n4

o
o
(10 3 km)

CD
O

I
02

F i g . 5 Moons of U r a n u s

[Oct,

1970]

435

FIBONACCI SERIES IN THE SOLAR SYSTEM

Saturn
Fibonacci S e r i e s
'' J u p i t e r

/ A s t e r o i d s

Mars
/ * Earth

Mercury

J_

J_

13

21
n

F i g . 6 The P l a n e t s

34

55

-L.

JL

89

144

436

% li
n/ n
(A.U.)

FIBONACCI SERIES IN THE SOLAR SYSTEM

[Oct.

.55

10

.50

40.2

.45

10.4
(Earth Densities)

40

10.6

.35

0.8

.30 h

1.0
_1_
1

L
2

JL
3

L
4

1
5

Planet

i
6 7

j
9

_|
10

Number

F i g . 7 zzn /fftrt and 8

n/ n

for P l a n e t N u m b e r n

19^0]

FIBONACCI SERIES IN THE SOLAR SYSTEM

437

each planet. Superimposed upon this figure are the respective planet densities
$ n (in terms of earth densities)* Figure 8 is a plot of z /f

against 8

which yields
z /f = 0856 - 0.28 8 ,
n' n
n-1
(in astronomical units). The actual form of the density dependence of the offset is not readily determinable. It is conjectured from further study that the
offset of a particular planet is either due to the average density of that planet
and the planet previous, or is due to some weighting of previous densities with
terms of the Fibonacci Series*
4. CONCLUSIONS
The significance of the values of m and b for the moons of the planets
is left for subsequent interpretation.

The quantity m may be regarded as a

planetary scale length although neither m , b ,

or the ratio b / m

seems to

bear any relationship to such planetary parameters as density, m a s s , or

volume.
The fact that the distance of the moons follows the Fibonacci Series
means that a particular moon ! s position is dependent upon the positions of the
previous two moons closer to the primary. Also, the moons seem to reside
and, in the case of Jupiter, even congregate at "potential levels" predicted by
the series.
5. ADDENDUM
The offset of a planet from the Fibonacci Series f

being a function of

the density of the previous planet is found to hold also for the offset of the
moons of Jupiter and Saturn. The "normalized" offset for the moons of these
two planets and for the planets of the sun is of the form f
z . +C

= 1.0 - 0.09 8 n _ 1 (cgs)

mf n
although the reliability of the density data is poor.

438

FIBONACCI SERIES IN THE SOLAR SYSTEM

[Oct.

.55

.50
z

/f
n/ n
(A. U.)

56 - .28 8

>n-l

Fig. 8 "Offset" versus Previous Planet Density

[Continued on p. 448. ]

n-1

A LUCAS ANALOGUE
BROTHER ALFRED BROUSSEAU
St. Mary's College, California

The Lucas sequence is defined in terms of the roots r1 and r 2 of the

equation x2 = x + 1 by the formula:
n , n
L n = rt + r 2
For this simple quadratic equation* the roots can be calculated as:

- j + V5

_ i - yi>
2

It can be ascertained directly that r j + r 2 = 1 and rj + r 2 = 3.

r

n+1
i

n ,
i

n-1
l

and
n+1
n
n-1
r2
= r 2 + r2
it follows by mathematical induction that if:
J

n-1 ,

n-1

n-1

and
T

Ln

= ^

n , n
+ r2 ,

then

Ln+1
n+ i =

n+1 ,
*

by direct addition of the two equations,

439

n+1
2

Then s since

440

A LUCAS ANALOGUE

[Oct.

If we seek to extend this idea to a sequence in which the last three consecutive terms are added to get the next term s the corresponding equation to
be used i s : x3 = x2 + x + 1, which has three roots r l f r 29 r 3 .

These need

not be calculated. It suffices to know that:

3

iL* r i

= lj

i=i

XT"]

"1

and

r r r

l23 =1>

1,3=1

using the standard relations between roots and coefficients of an equation. The
analogous definition of a sequence in terms of the roots would be:
m
n , n ., n
T n = r 4 + r2 + r 3 .
The first three values calculated on the basis of symmetric function formulas
would be:
3

i = S ri =

i=l

3
T,2 = V

r? = (Sr.) 2 - 2Sr.r. = 3

i=l
T

3 =!>? = (S^)8 - SGr.XSr.r.) + 3rir2r3

T3 = 1 + 3 + 3 =

With these values defined, T 4 would be equal to

2r

on the basis of the relation

7.

1970]

A LUCAS ANALOGUE

441

r?1 = 1r? +1 r? 1+ r.
Carrying the procedure one more step, if the last four quantities in a
sequence are added to obtain the next quantity, the roots of the equation
X4 = X3 + X2 + X + 1

would be employed with the sequence defined by:

i=l
With the aid of symmetric functions, it can be shown that:

Qi = 1,

Q2 = 3,

Q3 = 7,

Q4 = 15.

From these preliminary investigations, two points emerge: (1) As we

extend the Lucas analogue, the basic starting quantities carry over from one
k
stage to the next; (2) All the initial quantities are of the form 2 - 1 , That
these relations remain generally true is not difficult to prove*
Consider the n
order recursion relation:
x n = x 11 " 1

xn~2

x n - 3 + ... +x + 1,

which corresponds to adding the last n quantities to obtain the next quantity
in a sequence. Expressed as a polynomial equated to zero, this becomes:
n
x

n-1
-x

n-2
-x

n-3
A
- x . - - x - 1 = 0 ,

so that regardless of the degree of the equation:

IT.

= 1,

IT.

r. = - 1 ,
j

IT. T.

r,

1 3 k

= 1, etc.

442

A LUCAS ANALOGUE

[Oct 1970]

Then, since

is expressible in terms of those summations which have j or less components, it follows that the sum must be the same for any two equations having a
degree greater than or equal to j . This accounts for the persistence of the
basic starting quantities from one stage to the next.
Now assume that for the equation of degree n - 1
Ir}

= 2^ - 1

By the equation for the n

(j = 1, 2, , n - 1) .
degree:

v n
n-1 , v n-2 , v n-3 ,
j_ v
, vi
Zr. = xz r .
+ Zr.
+ zr.
+ + z r . + Z l
l

= (211""1 - 1) + (2 n " 2 - 1) + (2n""3 - 1) . . . (2 - 1) + n

n-1

=j 2

= 2n-l,

i=0
This proves the second contention,
CONCLUSION
For an n

order recursion relation of the form

X
=T
+T
+T
+ + T
m+n+1
m+n
m+n+1
m+n-2
m+1
it is possible to define a Lucas-type sequence using the roots of the equation
n
n-1 , n-2 , n-2 ,
, , i -4.u
x = x
+x
+x
+ + x + 1 with
T
= lvm
.
m
I
With such a definition, the first n starting values would be given by:
Tk = 2k - 1

(k = 1, 2, , n) .

ELEMENTARY PROBLEMS AND SOLUTIONS

Edited by
A. P. HILLMAN
University of New Mexico, Albuquerque, New Mexico

Send all communications regarding Elementary Problems and Solutions

to Professor A. P. Hillman, Department of Mathematics and Statistics, University of New Mexico, Albuquerque, New Mexico, 87106o Each problem or
solution should be submitted in legible form, preferably typed in double spacing, on a separate sheet or sheets in the format used below. Solutions should
be received within three months of the publication d a t e
Contributors (in the United States) who desire acknowledgement of r e ceipt of their contributions are asked to enclose self-addressed

stamped

postcards.
B-190 A repeat of B-186 with a typographical error corrected.
Let L

be the n

Lucas number and show that

L^ / L = l[ L Q - 3 ( - l ) n ]J2 + (-l) n 25F 2 .

5n' n
2n
n
B-191 Proposed by Guy A. Guillottee, Montreal, Quebec, Canada.
In this alphametic, each letter represents a particular but different
digit, all ten digits being represented here. It must only be that well-known
mathematical teaser from Toronto, J 9 A. H. Hunter, but what is the value
of HUNTER?
MR
HUNTER
MADE
A_
TEASER
B-192 Proposed by Warren Cheves, Littleton, North Carolina.
Prove that F Q = L F 0 - (-l) n F .
3n
n 2n
n
443

444

ELEMENTARY PROBLEMS AND SOLUTIONS

[Oct,

B-193 Proposed by V. E. Hoggatt, Jr., San Jose State College, San Jose, California.
Show that L , L
i s 5F F o r L L depending& on the choice
n+p
n-p
p n
p n
^
of sign on w h e t h e r p i s even o r odd.

B-194 Proposed by Phil Mana, University of New Mexico, Albuquerque, New Mexico.
Show that L ^A. - L = 5F, L[ F ^
+ (-l)nF ^ ].
n+4k
n
k n+3k
'
n+k
B-195 Proposed by David Zeitlin, Minneapolis,
Let

,
r|

Minnesota.

denote L L
L
,- / L - L 0 L .
n n-1
n-r+1
12
r

Show that

Following i s a l i s t of s o l v e r s whose n a m e s w e r e i n a d v e r t e n t l y omitted

from l i s t s in r e c e n t i s s u e s :
B-144 Don Allen
B - 1 4 8 , B - 1 4 9 , B - 1 5 0 , B - 1 5 1 , B-153 - D. V. J a i s w a l
B - 1 6 0 , B - 1 6 1 , B - 1 6 3 H . V. K r i s h n a
B-166 Michael Yoder
B-167 T. J . Cullen, B r u c e W. King, R. W. Sielaff, Michael Yoder
B-168 Michael Yoder
B-169 W r a y G B r a d y , Michael Yoder
B - 1 7 0 , B-171 Michael Yoder
SOLUTIONS
A CUBIC IDENTITY
B-172 Proposed by Gloria C Padilla, Albuquerque
Mexico.
Let

F 0 = 0,

F t = 1,

and

Fn+2

High School, Albuquerque,

= Fn + Fn+1

for

Show that

?a
= F 3 + F*
+ 3F F 1 F
n+2
n
n+1
n n+1 n+2

New

n = 0, 1, - .

1970]

ELEMENTARY PROBLEMS AND SOLUTIONS

Solution

445

by C B. A. Peck, State College, Pennsylvania.

(x + y) 3 = x 3 + y 3 + 3xy(x + y)

We obtain the d e s i r e d r e s u l t f r o m
substituting F n = x and F

- = y and hence F

+2

= x + y.

on

Note that the

r e s u l t does not depend on the initial values

Also solved by W. C. Barley, Wray G. Brady, T. J. Cullen, Andrew Dias, David Englund,
Herta T. Freitag, Bernard G. Hoerbelt, John E. Homer, Jr., John Ivie, Bruce W. King,
Peter A. Lindstrom, Bruce Lynn, John W. Milsom, Klaus-Gunther Recke, Michael Rennie,
Gerald Satlow, A. G. Shannon, Richard W. Sielaff, Charles W. Trigg, John Wessner,
Gregory Wulczyn, Michael Yoder, and the Proposer.
ANOTHER CUBIC IDENTITY
B-173 Proposed by Gloria C Padilla, Albuquerque
Mexico.

High School, Albuquerque,

New

Show that
Y
r
3n
Solution

= F3
- F3
- 3F F
F
n+2
n-1
n n+1 n + 2 '

by T. J. Cullen, California State Polytechnic

College, Pomona,

California.

F r o m F o r m u l a XXI on p . 68 of Vol. 1, Fibonacci Q u a r t e r l y ;

*3n

n+1

n-1 '

H e n c e , by B - 1 7 2 ,
TT

^3n

= T?3

_ p3

n+2

n-1

- 3F

* n r n+1 * n+2

Also solved by W. C Barley, Wray G. Brady, Herta T. Freitag, Bernard G. Hoerbelt,

John E. Homer, Jr., John Ivie, Bruce W. King, Peter A. Lindstrom, John W. Milsom,
C B. A. Peck, Klaus-Gunther Recke, A. G. Shannon, Charles W. Trigg, Gregory Wulczyn,
Michael Yoder, David Zeitlin, and the Proposer.

MODULO 10
B-l 74 Proposed by Mel Most, Ridgefield Park, New Jersey.
L e t a be a non-negative i n t e g e r .

Show that in the sequence

446

ELEMENTARY PROBLEMS AND SOLUTIONS

2F

a+l'

22F

2 3 F

a+2'

[Oct.

a+3'"'

all differences between s u c c e s s i v e t e r m s m u s t end in the s a m e d i g i t

Solution by Michael Yoder, Student, Albuquerque Academy, Albuquerque, New Mexico.

T h e sequence s a t i s f i e s the r e c u r r e n c e
S l Q = 2S ,- + 4S
n+2
n+1
n

= 2nF

(S
n

, ) .
a+n

Thus

s i n c e all S

n+2 "

n+1 =

(S

n+l "

n>

5S

n+1 "

n <mod

10)

a r e even.

Also solved by T.J. Cullen, David Englund, Hertd T. Freitag, C B. A. Peck, Klaus-Gilnther
Recke, A. G. Shannon, Charles W. Trigg, John Wessner, David Zeitlin, and the proposer.

A GENERALIZED 2 - B Y - 2 DETERMINANT
B-175

Composed from the Solution

by David Zeitlin to B-155.

L e t r and q be c o n s t a n t s and l e t U 0 = 0,
qU .
^ n

Uj = 1, U

rU

n-M

"

Show that
U , U ^u - U _,_ ,, U = q n U IL .
n+a n+b
n+a+b n
^
a b

Solution
Mexico.

by Michael

F o r a = 0,

Yoder,

Student,

Albuquerque

Academy,

Albuquerque,

the identity i s obviously t r u e ; and if it i s t r u e for a = 1,

i t will be t r u e by induction for all a.

S i m i l a r l y , the identity need only be

p r o v e d for b = 1, and the p r o b l e m r e d u c e s to that of proving

= q .

F o r n = 0,

New

this i s t r u e ; and s i n c e

n+

i "un+9Un

1970]
U

ELEMENTARY PROBLEMS AND SOLUTIONS

n+2 "

n+3Un+l =

n + 2 <rUn+l " * u n > "

= <*(Un+l -

(rU

447

n+2 " ^ n + l ) U n + l

n + 2 U n>

the identity is verified for all a s b s and n

Also solved by Wray G. Brady, T. J. Cullen, Herta T. Freitag, C. B. A. Peck, Klaus-Giinther
Recke, A. G. Shannon, Gregory Wulczyn, and David Zeitlin.
CUBES IN TERMS OF FIBONOMIALS ON DIAGONAL 3
B-176 Proposed by Phil Mana, University of New Mexico, Albuquerque, New Mexico.
Let

[-]

denote the Fibonomial Coefficient

n n-1

n-r+l/12

Show that

n - [-;s] -'p1 r]-[i]

Solution by David Zeitlin, Minneapolis, Minnesota.
Let H n satisfy H n + 2 = H n + 1 + H n , and define
jn
= H H - H
^ , / H - K L
1r i
n n-1
n-r+1/ 1 2

where H. > 0 , i = 1 , 2 , . Since

<*>

2H

n =

n+2Hn+l "

2H

n+lHn-l "

multiplication of (1) by H n gives

The desired result is obtained from (2) for

n-lHn-2 '

448

ELEMENTARY PROBLEMS AND SOLUTIONS

H

= F

where

IrJ

|_ r J

Oct. 1970

Also solved by L. Carlitz, T J. Cullen, Herta T Freitag, John E. Homer, Jr., Peter A.
Lindstrom, John W. Milsom, C B. A. Peck, Klaus-Gunther Recke, A. G. Shannon,
Charles W. Trigg, Gregory Wulczyn, Michael Yoder, and the Proposer.

FOURTH POWERS IN TERMS O F FIBONOMIALS

B-177 Proposed by Phil Mana, University of New Mexico, Albuquerque, New Mexico.
Using the notation of B - 1 7 6 , show that

n-["rMi2]-*[%+1H?]for s o m e i n t e g e r a and find

Solution

a.

by R. M. Grassl, University of New Mexico, Albuquerque,

Letting n = 2 , we find that a would have to be 4.

4,

Hence it suffices to verify the formula for

Mexico.

Then letting a =

both s i d e s satisfy the s a m e f o u r t h - o r d e r (i. e 0 , f i v e - t e r m )

relation.

New

recurrence

n = 0, 1, 2, 3 and it

follows for all v a l u e s of n by induction,

Also solved by L. Carlitz, T. J. Cullen, Herta T. Freitag, John E. Homer, Jr., C. B. A.
Peck, Klaus-Gunther Recke, Charles W. Trigg, Gregory Wulczyn, Michael Yoder, David
Zeitlin, and the Proposer.
[Continued from p . 438. ]
It i s found, a l s o , that the slope m of the d i s t a n c e s ZR v e r s u s the Fibonacci
Series f

for each planet s y s t e m i s a p o w e r law function of the m a s s M and

r a d i u s R of the planet in the f o r m

m oc M 3 R" 7 .

SUSTAINING MEMBERS
*H. L. A l d e r
V. V. A l d e r m a n
G. L. A l e x a n d e r s o n
R. H. Anglin
*Joseph A r k i n
L. L. Badii
Don B a k e r
Col. R. S. B e a r d
B. C. B e l l a m y
Murray Berg
Leon B e r n s t e i n
*Marjorie Bicknell
J . H. Biggs
J. J. Bohnert
M. B. B o i s e n , J r .
T. A. B r e n n a n
C. A. B r i d g e r
Maxey Brooke
*Bro. A. B r o u s s e a u
*J. L. Brown, J r .
C. R. Burton
P . F . Byrd
N. S. C a m e r o n
L. C a r l i t z
P . B. C h a r l a n d
P. J. Cocussa
L e e Corbin
Nannette Cox
A. B. C u m m i n g s
D. E. Daykin
J. W. D e c e l i s
F r a n c i s DeKoven
J. E. Desmond
A. W. Dickinson
M. H. D i e m
N. A. D r a i m
* Charter Members

D. C. Duncan
M. H. E a s t m a n
Merritt Elmore
R. S. E r l i e n
H. W. E v e s
F . A. F a i r b a i r n
A. J. Faulconbridge
H. H. F e r n s
D. C. F i e l d e r
E. T. F r a n k e l
C. L. G a r d n e r
G. H. Glabe
*H. W. Gould
Nicholas G r a n t
G. B. G r e e n e
G. A. Guillotte
B. H. Gundlach
V. C. H a r r i s
W. R. H a r r i s , J r .
*A. P . Hillman
M r . and M r s . B.H. H o e l t e r
*V. E. Hoggatt, J r .
*A. F . H o r a d a m
J . A. H. Hunter
D. F . Howells
F . R. J a c h i m
A. S. J a c k s o n
*Dov J a r d e n
J . H. J o r d a n
R. P . K e l i s k y
* Kenneth K l o s s
D. E. Knuth
* Sidney K r a v i t z
George Ledin, J r .
P. F. Lehman
Eugene Levine

*D. A. Lind
* C . T. Long
A. F . Lopez
F . W. Ludecke
J . S. Madachy
* J . A. Maxwell
* S i s t e r M a r y d e S a l e s McNabb
W. A. Morin
Lucile Morton
Stephen Nytch
R. D. O f Connell
P . B. Onderdonk
F . J. O s s i a n d e r
Ann P a p e
R. J . P e g i s
J . W. P h i l l i p s
Alwin P o r t n e y
M. M. Risueno
* D . W. Robinson
F . G. Rothwell
A. M. Routh
B. B. Sharpe
L. R. Shenton
J . A. S h u m a k e r
D. C. Stone
M. N. S. Swamy
A. S y l w e s t e r
*D. E. T h o r o
H. L. U m a n s k y
M. E. Waddill
* L . A. W a l k e r
R. J. Weinshenk
R. A. White
R. E . Whitney
P . A. Willis
C h a r l e s Ziegenfus

ACADEMIC OR INSTITUTIONAL MEMBERS

DUKE UNIVERSITY
D u r h a m , No. C a r o l i n a

ST. MARY'S COLLEGE

St. M a r y ' s C o l l e g e , California

NORWICH UNIVERSITY
Northfield, V e r m o n t

V A L L E J O UNIFIED SCHOOL DISTRICT

V a l l e j o , California

SACRAMENTO STATE COLLEGE

S a c r a m e n t o , California

WAKE FOREST UNIVERSITY

Wins t o n - S a l e m , No. C a r o l i n a

SAN JOSE STATE COLLEGE

San J o s e , California

WASHINGTON STATE UNIVERSITY

P u l l m a n , Washington

THE CALIFORNIA MATHEMATICS COUNCIL

SONOMA STATE COLLEGE

Rohnert Park, Calif.

BINDERS NOW AVAI LABLE

The F i b o n a c c i A s s o c i a t i o n is making available a binder which
can be used to take c a r e of one volume of the p u b l i c a t i o n at a t i m e .
This binder is d e s c r i b e d as follows by the company p r o d u c i n g it:
" . . . . The binder is made of heavy weight v i r g i n vinyl,
e l e c t r o n i c a l l y sealed over r i g i d b o a r d equipped with
a c l e a r label holder extending 2 - 3 / 4 " high from the
bottom of the backbone, round c o r n e r e d , fitted w i t h
a 1 1/2' " multiple m e c h a n i s m and 4 heavy w i r e s . "
The n a m e , FIBONACCI QUARTERLY, is p r i n t e d in gold on the
front of the binder and the s p i n e . The color of the binder is d a r k
green.
T h e r e is a s m a l l pocket on the spine for holding a tab
giving y e a r and v o l u m e . These l a t t e r w i l l be supplied with e a c h
o r d e r if the volume or volumes to be bound a r e indicated.
The p r i c e p e r binder is $ 3 . 5 0 which includes postage ( r a n g i n g
f r o m 50 to 80 for one b i n d e r ) . The tabs will be sent w i t h the
r e c e i p t or invoice.
All o r d e r s should be sent to: B r o t h e r Alfred B r o u s s e a u ,
Managing E d i t o r , St. Mary 1 s College , Calif. 94575