Professional Documents
Culture Documents
0. Introdu tion
Erdos [22 and Rigge [63 proved in 1939 that a produ
t of two or more
onse
utive positive integers is never a square. Further Erdos and Selfridge [27 proved in 1975 the analogous statement for
ubes and higher
powers. Therefore a produ
t of two or more
onse
utive positive integers
is never a power. The rst
ontribution in this dire
tion dates ba
k to 1724
when Goldba
h, in a letter to D. Bernoulli, showed that a produ
t of three
onse
utive positive integers is not a square. The theorem of Erdos - Rigge
is equivalent to saying that equation
(1)
n(n + 1) (n + k
has no solution. First we give a sket
h of the proof that (1) has no solution
whenever k ex
eeds a su
iently large absolute
onstant.
If n k, we refer to Bertrand's postulate to nd a prime p satisfying
n n+2 k p n + k 1 and dividing the left hand side of (1) to the
rst power. Therefore (1) implies that n > k. Now we apply a well-known
theorem of Sylvester [105 to nd a prime q ex
eeding k and dividing the
left hand side of (1). Consequently there is unique i with 0 i < k su
h
that q j (n + i). Furthermore we derive from (1) that n + i is divisible by q 2 .
3
Hen e
n + k > n + i q 2 (k + 1)2
n > k2 :
(2)
k > i j = aj (x2i
x2j )
= aj (xi xj )(xi + xj )
2aj xj 2(aj x2j )1=2 2n1=2
whi
h
ontradi
ts (2). The proof depends on
omparing a lower and an upper
bound for
4 =: aoa1 ak 1:
Sin
e ao ; a1 ; ; ak 1 are distin
t, it is
lear that
4 k!:
This
an be sharpened to
(4)
4 ( 32 )k k!;
k
p
ordp(4) [ + 1:
Thus
4 j k!
pk
p:
In parti ular
4 k!
pk
p k!3k :
Our intention is to nd an upper bound for 4 less than the lower bound
(4). The above upper bound is not su
ient for this purpose and we must
sharpen it. Let q 2 f2; 3g and we write
4 j k!
Y
pk
p 2g2
h2 3g3 h3 :
4 k!3k 2
3 1 2=3 1=4 k
2 3 ) 36k4
23
(n + d1 ) (n + dt ) = by 2 :
50
no solution other than the one given by
= 1402 . This in
ludes a result
3
of Erdos [23 of 1951 on (6) with t = k and b = k! i.e.
n+k
= y2:
k
n > k2 ; t k
C1
log k
implies that k C2 . The assumption n > k2 is satised if the left hand side
of (6) is divisible by a prime ex
eeding k. Saradha [67 showed that equation
0156)k
(6) with n > k2 ; t k [ (:log
k implies that k < 870. It is easy to see that
the above proof fails if (7) is repla
ed by
n > k2 ; t k
kf (k)
log k
with f (k) ! 1 as k ! 1. Therefore we should look for repla
ing the idea
of saving powers of 2 and 3 in the proof given above by a dierent argument
and it was found by Shorey [84 in 1986. For > 0, Shorey [87 applied in
1987 Baker's theorem on integral solutions of hyper-ellipti
equations and
Sieve - theoreti
arguments to show that (6) with
(8)
n > k2 ; t k
(1
) k
log log k
log k
n > k 2 ; t k
6
where
k = k(1
e1
(10)
0 + F (k )
< n k2 :
Then
P (n + di ) k for 1 i t:
For every prime p k, we
hoose an f (p) 2 fn + d1 ; ; n + dt g su
h that
p does not appear to a higher power in the fa
torisation of any other n + di
with 1 i t. Then, by an argument of Erdos [25, we have
nt (k)
Y
(n + di )
p[ p +[ p2 + kk
k
pk
Q
whi
h, by (10), implies that t < k . Here the produ
t is taken after
omitting all f (p) with p k. The above argument has turned out to be
basi
and we shall apply it again in se
tions 6, 7, 8 where we refer to it as
`a fundamental argument of Erdos'. Further we observe that the assumption
(9)
an be repla
ed by
n > e1
0 + F (k );
t k :
On the other hand, Balasubramanian and Shorey [7 showed that the pre
eding assumption is
lose to the best possible. More pre
isely, they [7 proved:
Let
> 0; k 3 and n e 1
0 F (k)
where
denotes the Euler's
onstant. For k k0 = k0 (), we
an nd
distin
t integers d1 ; ; dt 2 [0; k) with t k su
h that (n + d1 ) (n + dt )
is a square. Now we give a sket
h of the proof of this assertion. For a set T
of positive integers, we write ! (T ) for the number of distin
t prime divisors
of all elements of T . Then it is well-known that there exists a subset T 0 of T
with
j T 0 j j T j !(T )
7
j T j k + (k)
i.e.
j T j k k (k ):
1
It is easy to he k that
j T j
(n+k
=k
( (
1)
n+k
(
n 1
)):
The proof is
ompleted by applying deep results on dieren
e between
onse
utive primes for obtaining the desired estimate.
As promised, we turn to giving a sket
h of the proof that (6) with (8)
implies that k is bounded by a number depending only on . We may assume
that k ex
eeds a su
iently large number depending only on . By (6), we
have
(11)
n + di = ai x2i for 1 i t
a1 at
p[k=p+1 (3k)k :
pk
j S j k=2
1
and
(13)
ai k(log k)1
=2)
if ai 2 S1 :
xi k1=4 if ai 2 S1 :
8
k and ai is square
j S j k=4:
(15)
ai aj with ai 2 S3 ; aj 2 S3
are distin
t. This property is restri
tive; Erdos [25 showed by Sieve that
j S j 2G= log G
(16)
j S j k(log k)
3
=4
d1 ):
a = g
d (a1 ; a2 ; a3 ); bi = a 1 ai for 1 i 3
R = a 1 (d2
d1 ); R0 = a 1 (d3
d1 ):
We onsider (1) with the left hand side repla ed by two blo ks of onse utive integers. More pre isely, we onsider the equation
X (X + 1) (X + K
1)Y (Y + 1) (Y + K 0
1) = Z 2
x(x + 1) (x + k
1) = y (y + 1) (y + k + `
1)
Ax(x + 1) (x + k
1) = By (y + 1) (y + k + `
1):
All the solutions of (17) have been determined for several values of (A; B; k; `)
and we refer to [91, p.239 for an a
ount of these results. The rst result in
this dire
tion is due to Mordell [48 that (17) with A = B = 1; k = 2; ` = 1
implies that x = 2; y = 1 and x = 14; y = 5. For a given (A; B; k; `),
Beukers, Shorey and Tijdeman [11
onrmed the above
onje
ture of Erdos.
For the proof, they showed that the underlying
urve is irredu
ible and it
has positive genus. Then the assertion follows from a theorem of Siegel [104
on integral points on
urves but it provides no expli
it bound for the magnitude of the solutions. In fa
t they determined all (A; B; k; `) for whi
h the
urve has genus one and (17) has only nitely many rational solutions for all
10
x y C3 x2=3
for
ertain number C3 > 0 depending only on A and B . It is a di
ult
problem to
onrm the above
onje
ture in general. Even a very parti
ular
ase A = B = 1; k = 2 and y = 1 of (17) is an open problem. We
onsider
(17) with A = B = 1 and k + ` an integral multiple of k. For an integer
m 2; we re-write (17) in this
ase as
(18) x(x+1) (x+k 1) = y (y +1) (y +mk 1) in integers x > 0; y > 0; k 2:
Ma
Leod and Barrodale [40
onsidered (18) with m = 2. If m = 2, equation
(18) has a solution given by
8:9:10 = 6!
i.e.
(19)
x = 8; y = 1; k = 3:
Ma
Leod and Barrodale [40 observed that (19) is the only solution of (18)
with m = 2 and k 5. Saradha and Shorey [69 showed that (19) is the only
solution of (18) with m = 2. Further Saradha and Shorey [70 proved that
(18) with m = 3; 4 has no solution and Mignotte and Shorey [47 veried it
for m = 5; 6. In general Saradha and Shorey [71 showed that (18) implies
that
(20)
max(x; y; k) C4 (m)
(x + 1) (x + k) = (y + 1) (y + mk):
x 2k
k:
We write
(z + 1) (z + mk) =
(22)
mk
X
Aj (m; k)z mk j :
j =0
Bj = Bj (m; k) for 1 j m
su
h that
(23)
( z m + B1 z m
+ + Bm )k =
mk
X
Hj (m; k)z mk
j =0
satises
(24)
kB1
k
kB2 + 2 B12
A1 (m; k);
A2 (m; k);
side of (23). Therefore the k th root of the left hand side of (22) is
lose to
the k th root of the left hand side of (23). We use this observation with z
repla
ed by x and y , sin
e x and y are large as
ompared with k and m. The
k th root of the left hand side of (21) is
lose to x + k+1
and the k th root of
2
m
m
1
the right hand side of (21) is
lose to y + B1 y
+ + Bm . Consequently
k
+1
m
m
1
we derive that x + 2 is
lose to y + B1 y
+ + Bm . In fa
t, we show
that
j x (ym + B1 ym 1 + + Bm k +2 1 ) j<
where
= (2 l
m (den (B1 ); ; den (Bm ))) 1 :
Therefore
k+1
:
x = y m + B1 y m 1 + + Bm
2
We substitute this relation in (21) to get
k+1
2
Appendix and Mignotte and Shorey [47
he
ked the
onje
ture for m 12
and 13 m 20, respe
tively. Further Mignotte and Shorey [47 showed
that (21) with Hm+ (m; k) 6= Am+ (m; k) implies that
and
y 21=m 2k(1
=m
mk))=mk)
mk + 2
)
2
where log2 denotes the logarithm with respe
t to the base 2. Thus the above
estimates are valid for 7 m 20. We re
all that (21) with 2 m 6 has
no solution other than m = 2; x = 7; y = 0; k = 3.
1
(log 2 (
3. An equation of Goormaghtigh
We start this se
tion with the well-known a b
onje
ture : Let > 0
and a; b;
be relatively prime positive integers satisfying
a + b =
:
Then there exists a number depending only on su
h that
< G1+
where
G=
p:
pjab
Next we re
all that Nagell [51
onrmed a
onje
ture of Ramanujan [60
that the solutions of
(25)
are given by (x; n) = (1; 3); (3; 4); (5; 5); (11; 7); (181; 15). We
onsider an
equation of Goormaghtigh
xm 1 y n 1
=
in integers x > 1; y > 1; m > 2; n > 2 with x 6= y:
(26)
x 1
y 1
It has been
onje
tured that (26) has only nitely many solutions. Goormaghtigh [30 observed in 1917 that
(27)
31 =
213 1 903 1
25 1 53 1
=
; 8191 =
=
:
2 1
5 1
2 1
90 1
14
There is no loss of generality in assuming that x < y in (26) and then m > n.
Further we re-write (26) as
(28)
1)xm
(y
1)y n = y
(x
x:
We show that a b
onje
ture implies that (26) has only nitely many
solutions. By applying a b
onje
ture to (28) after dividing both the sides
by g = g
d ((x 1)y n; y x), we have
1)xm < 1 G1+
1
(y
xm
< 1 y 2+3
xm
1
2
> y (n
1)(
=m
2 ) (
1)
+ + x) + 1
and n = 3 has no solution other than the ones given by (27). Let n = 3 and
x = 2 in (26). Then y > 2 and we re-write (26) as
(2y + 1)2 + 7 = 2m+2 :
Now we apply the result of Nagell stated above in this se
tion to
on
lude
that (26) with x = 2 and n = 3 implies (27). Thus the numbers 31 and
8191 are given by Ramanujan - Nagell equation (25). Perhaps (26) has no
solution other than the ones given by (27). If x and y are xed, we read the
exponents mod 3 to write (28) as Thue equations with xed
oe
ients and
hen
e (26) has only nitely many solutions. Balasubramanian and Shorey [5
extended the pre
eding result by showing that (26) has only nitely many
solutions whenever x and y are
omposed of primes from a given nite set.
Further Shorey [83 showed that this is also the
ase if g
d (x; y ) = 1 and
either x; y x or y; y x are
omposed of primes from a given nite set. For
a given m and n; Davenport, Lewis and S
hinzel [20 proved that (26) has
only nitely many solutions. The proof depends on the well-known theorem
of Siegel [104 on integral solutions of polynomial equations in two variables
and therefore the proof of Davenport, Lewis and S
hinzel does not allow an
expli
it bound for the magnitude of the solutions. Further for a given m and
n with g
d (m 1; n 1) > 1, Davenport, Lewis and S
hinzel [20 gave an
ee
tive version of their result. Now the proof depends on an elementary
method of Runge [64. Further Nesterenko and Shorey [53 showed that (26)
with h = g
d (m 1; n 1) > 1 implies that max (x; y; m; n) is bounded by
a number depending only on (m 1)=h. This in
ludes the ee
tive version
of Davenport, Lewis and S
hinzel stated above. The proof depends on an
extension of Runge's method to exponential diophantine equations sket
hed
in the last se
tion. We
ounted the power of 2 on both the sides of (21) to
show that k is small as
ompared with x. Analogous assertion for (26) should
be that m is small as
ompared with y . This is a
hieved by the theory of
linear forms in logarithms. For a given pair (r; s) 6= (1; 1) of relatively prime
positive integers, let Sr;s be the set of all pairs (m; n) = (1 + dr; 1 + ds) with
d = 2; 3; . This is an innite set and the above result states that (26) has
only nitely many solutions when the exponents m and n are restri
ted to
the elements of Sr;s. This is the rst result on (26) of the type where the
exponents m and n are unbounded and there is no restri
tion on variables
x and y . We observe that (26) asks for integers with all the digits equal to
one with respe
t to two distin
t bases. Further we noti
e that 31 and 8191
are prime numbers su
h that ! (N 1) = 3 if N = 31 and ! (N 1) = 5 if
16
N = 8191. Shorey [90 showed that 31 and 8191 are the only prime numbers
N su
h that ! (N 1) 5 and N has all the digits equal to one with respe
t
xm 1
yn 1
=B
x 1
y 1
xm 1
= y q in integers x > 1; y > 1; m > 2; q 2:
x 1
This equation asks for perfe
t powers whose digits are all equal to one with
respe
t to base x. By writing y q = (y q=p)p ; there is no loss of generality in assuming that q is a prime number. We observe that (x; y; m; q ) =
(3; 11; 5; 2); (7; 20; 4; 2) and (18; 7; 3; 3) are solutions of (30). It has been
onje
tured that (30) has no other solution. Ljunggren [39
onrmed it whenever q = 2. Therefore we always suppose that q > 2 in (30). Further it follows from the results of Ljunggren [39 and Nagell [50 that (30) implies that
3 6 jm; 4 6 jm and if q = 3 then m 5 (mod 6) unless (x; y; m; q ) = (18; 7; 3; 3).
The proofs of these results are also available in Ribenboim [61.
We show that a b
onje
ture implies that (30) has only nitely many
solutions. We re-write (30) as
xm = (x 1)y q + 1:
17
Now we apply a b
onje
ture with = 1=8 to the pre
eding equation for
deriving that
xm
(31)
< 2 y 1+
where 2 is an absolute
onstant. On the other hand, we see from (30) that
y q < 2xm 1 whi
h implies that
1
2 2)=(m 1)
:
2
By the results of Ljunggren and Nagell stated above, we may suppose that
m 5. Therefore, sin
e = 1=8 and q 3; we see that the exponent of y
in the pre
eding inequality is at least 3=2. This
ontradi
ts (31) sin
e y is
su
iently large.
>From now onward in this se
tion, we
onne to a weaker version of
the
onje
ture that (30) has only nitely many solutions. Shorey and Tijdeman [94
onrmed this
onje
ture when x is xed. In parti
ular there
are only nitely many perfe
t powers with all the digits equal to one in
their de
imal expansions. This settles an old problem and the proof depends on the theory of linear forms in logarithms. Now we apply the above
result of Shorey and Tijdeman to show that it su
es to prove the
onje
ture when m is a prime power. Let m = P a m1 where a > 0 is an integer,
P = P (m); g
d (m1 ; P ) = 1 and we re-write (30) as
xm
> y q(m
xm 1 xP a 1
= yq:
xP a 1 x 1
Let p be a prime dividing ea
h of the fa
tors on the left hand side. Then p
divides m1 : Let be the least positive integer su
h that x 1 (mod p):
Then divides P a as well as p 1: Thus either = 1 or P < p P sin
e
P = P (m): Consequently = 1 whi
h implies that that p = P divides m1
ontradi
ting g
d (m1 ; P ) = 1: Therefore we
on
lude that the fa
tors on the
left hand side are relatively prime. Thus the se
ond fa
tor is a q th power.
Hen
e the assertion follows from the result of Shorey and Tijdeman stated
above. The above fa
torisation on equation (30) appears for the rst time
in [84, Lemma 7 and it has been very useful. Shorey [84, [85
onrmed the
onje
ture when ! (m) > q 2. For the proof of this result, Shorey [84, [85
showed that (30) has only nitely many solutions when either m 1 (mod q )
or x is a q th power. The proofs depend on the theory of linear forms in
18
ay n = 1
has no solution in integers x > 0 and y > 0 other than x = y = 1: A weaker
version of the pre
eding result was given by Bennett and de Weger [10. We
repla
e the above stated results of Shorey that (30) has only nitely many
solutions if x is a q -th power or m 1 (mod q ) by the results of Le and
Bennett that (30) has no solution if either x is a q -th power or m 1 (mod q )
in the proof of Shorey that (30) with ! (m) > q 2 has only nitely many
solutions. Then we
on
lude that (30) with ! (m) > q 2 has no solution.
Now we show that (30) with q j m implies that m is a power of q . We write
m = q e m0 where e 1 and g
d (m0 ; q ) = 1. Further we may suppose that m0
is not divisible by a prime 1 (mod q ) otherwise the assertion follows as in
[84 from the above fa
torisation on (30) and the result of Bennett . Now we
re-write (30) as
xm 1 xqe 1
= yq:
xqe 1 x 1
Let p be a prime dividing ea
h of the fa
tors on the left hand side. Then p
divides m0 : Let be the least positive integer su
h that x 1 (mod p):
Then divides q e and p 1: Therefore = 1 sin
e p 6 1 (mod q ): Now
we see that p = q divides m0
ontradi
ting g
d (m0 ; q ) = 1: Hen
e we derive
that the fa
tors on the left hand side are relatively prime. Therefore the rst
fa
tor on the left hand side is a q th power. Then we may suppose that
m0 = 1 or m0 = 2 sin
e xqe = (xqe 1 )q is a q th power. The latter possibility
is ruled out sin
e two
onse
utive positive integers
annot be q th powers
and the assertion is proved. On the other hand, a very parti
ular
ase of
the
onje
ture that (30) has only nitely many solutions when m is a power
of q is an open problem. An easier question than the
onje
ture that (30)
has only nitely many solutions is to repla
e ! (m) > q 2 by ! (m) 2 in
19
xn 1
= y1q ; xn + 1 = y2q
x 1
where y1 and y2 are relatively prime positive integers greater than one. The
se
ond is the equation of Catalan. Catalan [17
onje
tured that the only
solution of this equation is given by x = 2; n = 3; y2 = 3; q = 2: In other
words, the
onje
ture of Catalan states that 9 and 8 are the only perfe
t
powers that dier by one. This remains open but Tijdeman [108 proved
that the equation of Catalan has only nitely many solutions in integers in
x > 1; y2 > 1; n > 1; q > 1. Therfore we derive that (30) with m even
has only nitely many solutions. More generally, Shorey and Tijdeman [94
showed that (30) has only nitely many solutions if m is divisible by a xed
prime. It has not yet been possible to show that (30) has no solution whenever
m is even. Mignotte and Roy [46 proved that xn + 1 = y2q implies that
max (n; q ) > 106 ; min (n; q ) > 105 :
For positive integers a; b and non - zero integer k; the equation of Catalan is
a parti
ular
ase of the following equation of Pillai:
axm
Pillai [56
onje
tured that this equation has only nitely many solutions.
This
onje
ture has been
onrmed if at least one of the four variables in the
above equation is xed, see Shorey and Tijdeman [95, Chapter 12 whi
h we
also refer for a survey on exponential diophantine equations. This equation
is known as Pillai's equation.
The height of a rational number, in its redu
ed form, is the maximum
of the absolute values of its numerator and denominator. Let f(X) be a
polynomial of degree n with rational
oe
ients su
h that it has at least two
distin
t roots and f (0) 6= 0: We write
f (X ) = a0 X n + a1 X n 1 + + an
20
y m = f (x)
whi
h we suppose without referen
e in this paragraph. By a proper subsum
of y m f (x), we understand a proper subsum of y m a0 xn a1 xn 1 an :
S
hinzel and Tijdeman [79 proved that m is bounded by a number depending
only on f: We
onje
ture: There exists a number C5 depending only on L and
H su
h that either
m C5
f (X ) = (X 2
)=4; x = m + m ; y =
and
f (X ) = (x 1)(X m 1 + + X ) + x; y = x:
We observe that no proper subsum of y m f (x) vanishes in either of the
ases. Sin
e f (x) = ( )m in the rst
ase and f (x) = xm in the se
ond
ase, we see that the dependan
e of C5 on L as well as H is ne
essary in
21
Conje
ture 1. Further it is
lear that the assumption that f has at least two
distin
t roots su
h that f (0) 6= 0 is also ne
essary in Conje
ture 1.
Let us
onsider the equation
(32)
xm 1
= y q + 1 in integers x > 1; y > 1; m > 2; q 2:
x 1
xm 1 1
x
= yq :
x 1
We see that m 4 sin
e a produ
t of two
onse
utive integers is not a power.
Further we observe that the fa
tors on the left hand side are relatively prime.
Therefore
xm 1 1
= y4q
x = y3q ;
x 1
where y3 > 1 and y4 > 1 are integers. Then Shorey [89 applied his result stated above to
on
lude that (32) has only nitely many solutions.
Further Le [38
on
luded that (32) has no solution. This is an immediate
onsequen
e of his result that (30) has no solution whenever x is a q th
power. The rst result in the dire
tion of the pre
eding result is due to
Inkeri [33 that (30) with q = 3 has no solution whenever x is a
ube. Saradha
and Shorey [74 proved that (30) with x = z 2 implies that z 31 and
z 2= f2; 3; 4; 8; 9; 16; 25; 27g. The proof depends on an estimate of Laurent,
Mignotte and Nesterenko [36 on linear forms in two logarithms, irrationality
measures of Baker [1 of
ertain algebrai
numbers, p adi
method of Skolem
as given by Le [37 and
ongruen
e arguments due to Inkeri [33. Mignotte
[45 has re
ently improved the above referred estimate on linear forms in
two logarithms. By
ombining this estimate with estimates on solutions of
ertain Thue- Mahler equations and di
ult and elaborate
omputations,
Bugeaud, Mignotte, Roy and Shorey [15 showed that (30) with x = z 2 and
z 31; z 62 f2; 3; 4; 8; 9; 16; 25; 27g has no solution. The pre
eding result
was proved, independently and dierently, by Bennett [9. Hen
e (30) has no
solution whenever x is a square. The proof of Bennett depends on his result
on the equation (a + 1)xn axn = 1 stated above. Further Hirata - Kohno
and Shorey [54 proved that for a prime number 3, equation (30) with
x = z and q > 2( 1)(2 3) implies that max (x; y; m; q ) is bounded by
a number depending only on . If = 3, we may suppose that q 6= 3 in view
22
ah (g ) = 0:(g n1 )h (g n2 )h :
1
Mahler [41 proved that a10 (g ) is irrational for fni g1
i=1 = fi 1gi=1 . It is
now known that ah (g ) is irrational for any unbounded sequen
e fni g1
i=1 of
non-negative integers, see Sander [65. If an element o
urs in a sequen
e
innitely many times, it is
alled a limit point of the sequen
e. Let fni g1
i=1
be a bounded sequen
e of non-negative integers. If it has only one limit point,
it is ultimately periodi
and hen
e ah (g ) is rational. We always suppose now
onward that it has exa
tly two limit points N1 < N2 su
h that g N2 N1 6= h +1
whenever g N1 < h and it is not ultimately periodi
. Then the above result
23
htL 1
ht 1
for some integer L 2 and t given by ht 1 g N1 < ht : Bugeaud, Mignotte
g N2
N1
and Roy [14 improved the assertion of Saradha and Shorey by showing that
above in the result of Sander has to be even. Thus we are led to (30) with
m even and it has only nitely many solutions. Hen
e we
on
lude that, for
integers g 2 and h 2 su
h that g is even and h is odd, if ah (g ) is rational
then N2 is bounded by an absolute
onstant. For bounded sequen
es with
more than two limit points, Sander [65 obtained some partial results and
further investigations are desirable. For proving the above result on (30) applied to derive an irrationality
riterion, Saradha and Shorey [74 showed that
(30) implies that either max (x; y; m; q ) is bounded by an absolute
onstant
or there exists a prime p su
h that p j x and p6 j(y 1). M. Le (A
ta Arith. 69
(1995), 91-97) and L. Yu and M. Le (A
ta Arith. 73 (1995), 363-365)
laimed
stronger version of the pre
eding result but their proofs are not
orre
t. Corre
t proofs have re
ently been given by Bugeaud, Mignotte and Roy [14
who proved that (30) has no solution other than (x; y; m; q ) = (18; 7; 3; 3)
if every prime divisor of x divides y 1: This implies their result on irrationality of ah (g ) stated above. This also in
ludes a result of Bugeaud and
Mignotte [13 that (30) with 2 x 10 has no solution settling a problem
of Inkeri [33. Thus it is not possible to nd a perfe
t power greater than
one with digits identi
ally equal to one in its de
imal expansion. Further
the result of Bugeaud and Mignotte that (30) with x = 3 has no solution
nds appli
ation in group theory. We remark that p adi
linear forms in
logarithms with i0 s p adi
ally
lose to one appear in the work of Bugeaud,
Mignotte, Roy and sharp estimates of Bugeaud [12 for these p adi
linear
24
forms in logarithms are utilised. The work also involves very heavy
omputations depending on a result of Bennett that (30) with m 1 (mod q ) has
no solution.
Now we
onsider a more general equation than (30). For positive integers
A; B and prime number q with AB > 1, g
d (A; B ) = 1 and q free A, we
onsider the equation
(33)
xm 1
= By q in integers x > 1; y > 1; m > 2; q 2:
x 1
Oblath [55 showed that (33) with 1 < A < 10; B = 1 and x = 10 has no
solution. Inkeri [33 determined all solutions of (33) with 1 < A < x 10
and B = 1. Shorey and Tijdeman (see [94, [93) showed that (33) implies
that either max (x; y; m; q ) is bounded by a number depending only on A
and B or m = 2n with
xn 1
= y5q ; A(xn + 1) = By6q
x 1
where y5 > 1 and y6 > 1 are integers satisfying g
d (y5 ; y6 ) = 1 and y5 y6 = y .
Now it is
lear from the above relations that it is easier to deal with (33)
than (30). Therefore stronger results on (33) than (30) are available though
we have not been able to show that (33) has only nitely many solutions.
By
ombining the pre
eding result with the known results in the theory
of exponential diophantine equation, Shorey [93 observed that (33) implies
that max (x; y; m; q ) is bounded by a number depending only on A; B and
the greatest prime fa
tor of x. As already mentioned, the pre
eding assertion
remains unproved for (30). For integer x > 1 and prime number q , we write
U (A; B; x; q ) and U (x; q ) for the number of solutions of (33) and (30) in
integers y > 1 and m > 2, respe
tively. Shorey [85 showed that
U (x; q ) q + C6
where C6 is an absolute
onstant. Le [38 sharpened it to
U (x; q ) q:
Shorey [93 showed that
U (A; B; x; q ) C7
25
g d(m; AB'(AB )) = 1
where ' is the Euler totient fun
tion. For example, (33) with x = z 2 , (34)
and (33) with q = 2; (34) have no solution.
5. Equal produ
ts of integers in arithmeti
progression
8
>
< 14 for 2 k 7
(k) = > 50 for k = 8
: exp (k log k (1:25475)k
26
Then we onsider
(x
si + jd1 ) =
(mk
Y
Y 1)
i=1 j =0
(y
si + jd2 )
su
h that the fa
tors on the left hand side as well as the fa
tors on the right
hand side are non-zero and pairwise distin
t. Then Saradha [66 proved that
max (j x j; j y j; k) is bounded by a number depending only on d1 ; d2 ; `; m;
and s1 ; ; s whenever
m = 2 or = 2; 3; 4 or d2 = 1
27
unless
m = = 2; ` = k = d2 = 1; x = y 2 + y (1 s1
s2 ) + s1 s2 :
Further we observe that the above innitely many possibilities satisfy the
pre
eding equation.
Finally we turn to
onsidering (35) with m = 1:
(38)
x(x + d1 ) (x + (k
1)d1 ) = y (y + d2 ) (y + (k
1)d2 ):
This equation was suggested by Gabovi
h [29 in 1966. It is
lear that (38)
with k = 2 has innitely many solutions. Further Gabovi
h [29 gave an
innite
lass of solutions of (38) with k = 3; 4. Some innite
lasses of solutions of (38) with k = 5 were given by Szymi
zek [106 and Choudhry [16.
Choudhry [16 also provided an innite
lass of solutions of (38) with arbitrary k and unbounded d1 ; d2 . Finally we
onsider equation (38) with xed
d1 and d2 . If d1 = d2 , we observe that x = y . Therefore we suppose that d1
and d2 are distin
t. Further there is no loss of generality in assuming that
d1 < d2 and g
d (x; y; d1; d2 ) = 1. Then Saradha, Shorey and Tijdeman [76
proved that max(x; y; k) is bounded by a number depending only on d2 unless
(39)
x = k + 1; y = 2; d1 = 1; d2 = 4:
2) for k = 2; 3; ;
the possibilities (39)
an not be ex
luded. The proof depends on Prime Number Theorem for arithmeti
progressions. On the other hand, it is possible
to make the proof independent of Prime Number Theorem for
ertain values
of d1 and d2 . This led Saradha, Shorey and Tijdeman [77 to show that all
the solutions of (38) with d1 = 1 and d2 = 2; 3; 5; 6; 7; 9; 10 are given by
2.3=1.6, 7.8.9=4.9.14, 8.9=6.12, 5.6=3.10, 4.5.6=1.8.15, 15.16.17= 10.17.24,
9.10=6.15, 7.8=4.14, 24.25=20.30 and 32.33.34=24.34.44. This follows by
omputations from their result that (38) with d1 = 1 and d2 k + 1 implies
that y 2 (mod 4) and d2 = 2` for some integer ` 2.
6. The greatest prime fa
tor of integers in arithmeti
al
progression
28
P =: P (n(n + 1) (n + k
1)) k:
This leads us to
onsidering lower bound for the greatest prime fa
tor of
the produ
t of k
onse
utive positive integers. The rst result dates ba
k to
Sylvester [105 in 1892 that
P > k if n > k:
If n < k3=2 and k is su
eintly large, it follows from well-known results
on dieren
e between
onse
utive primes that there is prime p satisfying
n p n + k 1. Therefore we shall
onsider lower bound for P when
n k3=2 . Then Erdos [24 proved that
P > C8 k log k
(40)
(k)
nT
(k)
kk
and the assertion (40) follows from Prime Number theory. Rama
handra [57
applied Selberg's Sieve to prove (40) with C8 = (1 ) for > 0 and k
ex
eeding a number depending only on . Rama
handra [58 also obtained
partial results for (40) with C8 = (2 ). Combining these results with the
method of Roth and Halberstam on dieren
e between
onse
utive free
integers, Tijdeman [107 proved (40) with C8 = 2 . Further Rama
handra
and Shorey [59 sharpened (40) to
(41)
P > C9 k log k (
log log k
for log k > e
log log log k
where C10 > 0 is an absolute
onstant. The best possible estimate with
log A in pla
e of log A1 log An for linear forms in logarithms with i 's
29
lose to one are
ru
ial for the proofs of (41) and (42). Furthermore the proof
of (42) depends on estimates for exponential sums. We shall
ontinue with
the results on lower bound for P at the end of the next se
tion.
For an integer d > 1, we write
Pd = P (n(n + d) (n + (k
1)d)); = n + (k
1)d:
Let us re
all the result of Erdos and Selfridge mentioned in the beginning
of this paper. It states that a produ
t of two or more
onse
utive positive
integers is never a
ube or a higher power. This is equivalent to saying that
equation
(44)
n(n + 1) (n + k
(n + d1 ) (n + dt ) = by `
xn + y n = 2 z n :
Gyory derived his result from the theorems of Ribet [62 and Darmon and
Merel [19 that if x > 0; y > 0; z > 0; n > 2; 1 are integers satisfying
x; y; z relatively prime, n prime, < n and (46), then x = y = z = 1.
31
Erdos [25 observed that his elementary method sket
hed above yields
that for > 0, equation (45) with t k (1 )k logloglogk k implies that k is
bounded by a number depending only on . We put
1
2
` = (1 +
4`2 8` + 7
):
2(` 1)(2`2 5` + 4)
We observe that
45
47
; 4 = and ` < 2=3 for ` 5:
56
64
Shorey [84, [87 improved
onsiderably the result of Erdos by proving that
(45) with
3 =
t ` k
(47)
t k`
(48)
1 11
+ (k ) + 2
8 2
< 112`3 160`+29 if ` 1 (mod 2)
0
28`
76`+29
` = : 112
`2 160`+17 if ` 0 (mod 2):
3
28`
188`+129
0
and
(4
` (1
` (1
if ` 1 (mod 2)
` ) if ` 0 (mod 2)
`)
1
( 875)
1
(1 412)
0 :3664;
70 :4832; 80 :4556; 90 :3878; 10
0 :3243; 0 :3076; 0 :2787; 0 :2655:
11
12
13
14
32
if ` = 3; 4; 5; 6
00
` = `0 if ` 7:
`
We derive from the above results that there exists a prime p satisfying
00
p > (1 `
and
ordp (n(n + 1) (n + k
)k log k
1)) 6 0 (mod `)
Fermat (see [49, p.21) proved that there are no four squares in arithmeti
progression. As stated in the pre
eding se
tion, Darmon and Merel [19
proved that
xn + y n = 2z n
has no solution in positive integers n > 2; x; y and z other than the ones
given by x = y = z . Thus for ` > 2 there are no three ` th powers in
arithmeti
progression. For earlier results in this dire
tion, see [21, [96
and [100. Now we
onsider a more general situation. Let b; d and k 2 be
positive integers su
h that P (b) k. We
onsider
(49)
n(n+d) (n+(k 1)d) = by ` in integers n > 0; y > 0; k 2; ` 2 with g
d (n; d) = 1:
The rst result on (49) is due to Euler (see [49, p.21) that a produ
t of
four positive integers in arithmeti
progression is never a square. A related
question is that a produ
t of four positive integers in arithmeti
progression
an not be a produ
t of two
onse
utive positive integers. The answer to
this question is negative. Here the underlying equation is (35) with m =
2; d1 = 1; k = 2 whi
h, as pointed out in [77, has innitely many solutions
in positive integers x; y and d2 . Our survey of the
ase d = 1 of (49) is already
omplete in se
tion 7 and we assume that d 2 in this se
tion. Then we see
from (43) that P (y ) k if and only if (n; d; k) = (2; 7; 3). Thus the left hand
side of (49) is divisible by a prime ex
eeding k whenever (n; d; k) 6= (2; 7; 3).
There is no loss of generality in assuming that ` is a prime number in equation
(49) and the subsequent equations (50), (51). Erdos
onje
tured that (49)
implies that k is bounded by an absolute
onstant. We shall show at the end
of this se
tion that the
onje
ture of Erdos with l > 3 is a
onsequen
e of
a b
onje
ture. Saradha [67 proved that (49) with d 6 and k 3 has
no solution. If ` = 2, Saradha [68 showed that the assertion is valid even
when d 22 unless (n; d; k) = (2; 7; 3); (18; 7; 3); (64; 17; 3). Marszalek [43
onrmed the
onje
ture for xed d. If ` 3, Shorey [88 proved that (49)
implies that k is bounded by a number depending only on the greatest prime
fa
tor of d. Further Shorey and Tijdeman [97 proved that (49) implies that k
is bounded by a number depending only on ` and ! (d). It is an open problem
to bound k by a number depending only on ! (d) in the pre
eding result. If
` 7, Shorey [92 showed that k is bounded by a number depending only
34
on n whenever (49) is satised. The proof of the former result is elementary
whereas the proof of the latter depends on
ombining sharp estimates from
linear forms in logarithms with i 's
lose to one and irrationality measures
proved by hypergeometri
method.
Now we relax the assumption g
d (n; d) = 1 in the above stated results
on (49). For this, we
onsider
(50) n(n + d) (n + (k
(n + d1 d) (n + dt d) = by `
log log k
5
h(k) = log log log k ifif `` =
2; 3:
n0 = nG 1 ; d0 = dG 1 ; b0 = bG1 k :
We observe that g
d (n0 ; d0 ) = 1 and d0 2 sin
e d 6 jn. By dividing both the
sides of (50) by Gk , we see that
n0 (n0 + d0 ) (n0 + (k
1)d0 ) = b0 y `G2 k
where b0 and y `G2 k are positive integers su
h that P (b0 ) k. For a prime
p dividing G2 and y `G2 k , we observe that p > k and there is pre
isely one
integer F (p) in [0; k) su
h that p divides n0 + F (p)d0. We write fd1 ; ; dt g
for the set obtained by deleting all F (p) with p dividing g
d (G2 ; y `G2 k ). We
observe that
(52)
tk
! (d)
where b1 and y1 are positive integers su
h that P (b1 ) k and all the prime
divisors of y1 are greater than k. We show that the left hand side of (53) is
divisible by a prime ex
eeding k. We prove by
ontradi
tion. Let = 14 and
suppose that the greatest prime fa
tor of the left hand side of (53) does not
ex
eed k. Then we apply a fundamental argument of Erdos for deriving that
t Y
(k )
j =0
(n0 + jd0 ) kk
d kC12
and
log log
k;
n kC12
log log
if ` 7
where C12 > 0 is an absolute
onstant. Finally we apply the above estimate
for d to
on
lude that a b
onje
ture implies the
onje
ture of Erdos on
(49) with l > 3: We suppose (49) with l > 3 and k ex
eeding a su
iently
large absolute
onstant. Then we write
i.e.
Referen es
[11 F. Beukers, T.N. Shorey and R. Tijdeman, Irredu
ibility of polynomials and arithmeti
progressions with equal produ
ts of terms, Number
Theory in Progress : Pro
eedings of the International Conferen
e on
Number Theory in Honor of Andrej S
hinzel, held in Zakopane, Poland,
June 30 - July 9, 1997, Part I : Diophantine Problems and Polynomials,
Walter de Gruyter, Berlin, to appear.
38
[22 P. Erdos, Note on the produ
t of
onse
utive integers (1), Jour. London
Math. So
. 14 (1939), 194-198.
[23 P. Erdos, On a diophantine equation, Jour. London Math. So
. 26
(1951), 176-178.
[24 P. Erdos, On
onse
utive integers, Nieuw Ar
h. Voor Wiskunde 3
(1951), 124-128.
[25 P. Erdos, On the produ
t of
onse
utive integers III, Indag. Math. 17
(1955), 85-90.
39
[26 P. Erdos, Problems and results on number theoreti
properties of
onse
utive integers and related questions, Pro
. Fifth Manitoba Conf. Numeri
al Math., Utilitas Math. (1975), 25-44.
[27 P. Erdos and J.L. Selfridge, The produ
t of
onse
utive integers is never
a power, Illinois Jour. Math. 19 (1975), 292-301.
[28 G. Faltings, Endli
hkeitssatze fur abels
he Varietaten uber Zahlkorpern,
Invent. Math. 73 (1983), 349-366.
[29 Ya. Gabovi
h, On arithmeti
progressions with equal produ
ts of terms,
Colloq. Math. 15 (1966), 45-48.
[30 R. Goormaghtigh, L' Intermediaire des Mathemati
iens 24 (1917), 88.
[31 K. Gyory, On the diophantine equation
(1997), 289-295.
n = x` , A
ta Arith. 80
k
1)=(x
1) = bx` ,
1) = y m , A ta
[34 M. Jutila, On numbers with large prime fa
tor II, Jour. Indian Math.
So
. 38 (1974), 125-130.
[35 M. Langevin, Plus grand fa
teur premier d'entiers en progression
arithmetique, Sem. Delange - Pisot - Poitou, 18 e annee, 1976/77, No
3, 6pp.
[36 M. Laurent, M. Mignotte and Y. Nesterenko, Formes lineaires en deux
logarithmes et determinants d'interpolation, Jour. Number Theory 55
(1995), 285-321.
[37 M.H. Le, A note on the diophantine equation
64 (1993), 19-28.
xm
x
1
1
= y n, A ta Arith.
[38 M.H. Le, A note on the diophantine equation (xm 1)=(x 1) = y n +1,
Math. Pro
. Cambridge Philos. So
. 116 (1994), 385-389.
40
by n = , A ta Arith. 75
[57 K. Rama
handra, A note on numbers with a large prime fa
tor II, Jour.
Indian Math. So
. 34 (1970), 39-48.
[58 K. Rama
handra, A note on numbers with a large prime fa
tor III,
A
ta Arith. 19 (1971), 49-62.
[59 K. Rama
handra and T.N. Shorey, On gaps between numbers with a
large prime fa
tor, A
ta Arith. 24 (1973), 99-111.
[60 S. Ramanujan, Question 464, Jour. Indian Math. So
. 5 (1913), 120.
Colle
ted papers, Cambridge University Press, 1927, 327.
[61 P. Ribenboim, Catalan's Conje
ture, A
ademi
Press (1994).
[62 K.A. Ribet, On the equation ap + 2 bp +
p = 0, A
ta Arith. 79 (1997),
7-16.
ein diophantis
hes Problem, 9th Congress Math. S
and.
[63 O. Rigge, Uber
Helsingfors, 1938, Mer
ator, 1939, 155-160.
[64 C. Runge, Ueber ganzzahlige Losungen von Glei
hungen zwis
hen swei
Veranderli
hen, Jour. reine angew. Math. 100 (1887), 425-435.
[65 J.W. Sander, Irrationality Criteria for Mahler's numbers, Jour. Number Theory 52 (1995), 145-156.
[66 N. Saradha, On blo
ks of arithmeti
progressions with equal produ
ts,
Jour. de Theorie des Nombres de Bordeaux 9 (1997), 183-199.
42
1) = y q with
[80 T.N. Shorey, Linear forms in the logarithms of algebrai
numbers with
small
oe
ients I, Jour. Indian Math. So
. 38 (1974), 271-284.
[81 T.N. Shorey, On gaps between numbers with a large prime fa
tor II,
A
ta Arith. 25 (1974), 365-373.
[82 T.N. Shorey, On the ratio of values of a polynomial, Pro
. Indian A
ad.
S
i. (Math. S
i.) 93 (1984), 109-116.
[83 T.N. Shorey, On the equation a(xm 1)=(x
(II), Hardy Ramanujan 7 (1984), 1-10.
1) = b(y n
1)=(y
1)
1)=(x
[93 T.N. Shorey, The equation a(xm 1)=(x 1) = by q with ab > 1, Number
Theory in Progress : Pro
eedings of the International Conferen
e on
Number Theory in Honor of Andrej S
hinzel, held in Zakopane, Poland,
June 30 - July 9, 1997, Part I : Diophantine Problems and Polynomials,
Walter de Gruyter, Berlin, to appear.
[94 T.N. Shorey and R. Tijdeman, New appli
ations of Diophantine approximations to Diophantine equations, Math. S
and. 39 (1976), 5-18.
[95 T.N. Shorey and R. Tijdeman, Exponential Diophantine equations,
Cambridge Tra
ts in Mathemati
s 87 (1986), Cambridge University
Press, Cambridge.
[96 T.N. Shorey and R. Tijdeman, Perfe
t powers in arithmeti
al progression, Jour. Madras University (Se
tion B) 51 (1988), 173-180.
[97 T.N. Shorey and R. Tijdeman, Perfe
t powers in produ
ts of terms in
an arithmeti
al progression, Compositio Math. 75 (1990), 307-344.
[98 T.N. Shorey and R. Tijdeman, On the greatest prime fa
tor of an arithmeti
al progression, A Tribute to Paul Erdos, ed. by A. Baker, B. Bollobas and A. Hajnal, Cambridge University Press (1990), 385-389.
[99 T.N. Shorey and R. Tijdeman, On the greatest prime fa
tor of an arithmeti
al progression (II), A
ta Arith. 53 (1990), 499-504.
[100 T.N. Shorey and R. Tijdeman, Perfe
t powers in arithmeti
al progression II, Compositio Math. 82 (1992), 107-117.
[101 T.N. Shorey and R. Tijdeman, Perfe
t powers in produ
ts of terms in
an arithmeti
al progression (II), Compositio Math. 82 (1992), 119-136.
[102 T.N. Shorey and R. Tijdeman, Perfe
t powers in produ
ts of terms in
an arithmeti
al progression (III), A
ta Arith. 61 (1992), 391-398.
[103 T.N. Shorey and R. Tijdeman, On the greatest prime fa
tor of an arithmeti
al progression (III), Diophantine Approximation and Trans
endental Numbers, Luminy 1990, edited by Ph. Philippon, Walter de
Gruyter, Berlin (1992), 275-280.
45
[104 C.L. Siegel, Ueber einige Anwendungen diophantis
her Approximationen, Abh. Preuss. Akad. Wiss. Phys - math. K1., No.1, (1929), 70
pp.
[105 J.J. Sylvester, On arithmeti
al series, Messenger Math. 21 (1892), 1-19,
87-120.
[106 K. Szymi
zek, Note on arithmeti
al progressions with equal produ
ts of
ve terms, Elem. Math. 27 (1972), 11-12.
[107 R. Tijdeman, On the maximal distan
e of numbers with a large prime
fa
tor, Jour. London Math. So
. 5 (1972), 313-320.
[108 R. Tijdeman, On the equation of Catalan, A
ta Arith. 29 (1976), 197209.
[109 R. Tijdeman, Exponential diophantine equations 1986-1996, Number
Theory, Diophantine, Computational and Algebrai
Aspe
ts, Pro
eedings of the International Conferen
e in Eger, Hungary, 1996, ed. by
Gyory, Petho and Sos, Walter de Gruyter, Berlin (1998), 523-539.
[110 A. Wiles, Modular ellipti
urves and Fermat's Last Theorem, Annals
of Math. 141 (1995), 443-551.
[111 Kunrui Yu, Linear forms in p adi
logarithms, A
ta Arith. 53 (1989),
107-186.
[112 Kunrui Yu, Linear forms in p adi
logarithms II, Compositio Math.
74 (1990), 15-113.
S
hool of Mathemati
s
Tata Institute of Fundamental Resear
h
Homi Bhabha Road
Mumbai 400 005
INDIA.
46