Equations in Integers

Arkadii Slinko
The following methods are often used. Of course, they are not completely independent and many problems can be solved by using dierent ones.

1. Prime factorization
This is a powerful method which we illustrate by the following example: Problem 1. (New-York, 1977) Solve the Diophantine equation 2x +1 = y 2 . Solution: We write 2x = y 2 1 = (y 1)(y + 1). Since the prime factorization is unique, we get y 1 = 2k and y + 1 = 2m with m > k > 0. But the 2m 2k = (y + 1) (y 1) = 2, whence 2k (2mk 1) = 2 and k = 1 and m = 2. Finally we get x = y = 3. Problem 2. (Moscow, 1945) Find all integer solutions of the equation xy + 3x 5y = 3. Solution: Let us express y as a function of x: y= 3 3x 15 3x 18 18 = = 3 . x5 x5 x5

Thus x 5 divides 18, hence x 5 can take any of the following values: 1, 2, 3, 6, 9, 18. Hence we will have 12 solutions shown in the following table: x5 x y -1 4 15 1 6 -21 -2 3 6 2 7 -12 -3 2 3 3 8 -9 -6 -1 0 6 11 -6 -9 -4 -1 9 14 -5 -18 -13 -2 18 23 -4

2. Proving that there are no large solutions

Problem 3. (Moscow, 1944) Find all integer solutions of x + y = x2 xy + y 2 . 1

Solution: We write this equation as 1 2 1 1 1 1 1 x xy + y 2 + x2 x + + y 2 y = 1. 2 2 2 2 2 2 or ((x y)2 + (x 1)2 + (y 1)2 = 2. Now it is clear that |x y| 1, (x 1) 1, (y 1) 1, and we nd all solutions by inspection: (0, 0), (1, 0), (0, 1), (2, 1), (1, 2), (2, 2). This method often involves using inequalities. Problem 4. Find all pairs positive integers (x, y) which satisfy the equation x3 y 3 = xy + 61. Solution 1: Let (x, y) be a solution. By AM-GM inequality x2 + y 2 2xy and hence x3 y 3 = (x y)(x2 + xy + y 2 ) 3(x y)xy. Hence 3(x y)xy xy + 61 or (3(x y) 1)xy 61. It is clear that x = y. Let us consider the case, when x y = 1. Then substituting x = y + 1 into the original equation we get 2y 2 + 2y + 1 = 61 from which we deduce that y = 5, x = 6. If x y = 2, we obtain the equation 5y 2 + 10y + 8 = 61, which has no integer solutions. Finally, let us suggest that x y 3. Then 3(x y) 1 8, hence xy 7. Since x y 3, the only possibility is x = 7 and y = 1 which do not satisfy the equation. Hence (x, y) = (6, 5) is the only solution. Solution 2: We may also prove that the dierence d = x y is small by substituting x = y + d into the equation. We will obtain (3d 1)y 2 + (3d2 d)y + d3 = 61 from which we deduce that d3 61 or d 3. Then we need to consider three cases: d = 1, d = 2, and d = 3. We leave this to the reader.

3. Squeeze principle
Squeezing an unknown between two consequtive squares is a method which is surprisingly often used. In the simplest form it is as follows:

Problem 5. Prove that the equation m2 = n 4 + n 2 + 1 does not have integer solutions. Solution: We note that (n2 )2 = n4 < m2 < n4 + 2n2 + 1 = (n2 + 1)2 which is impossible as the square m2 is squezed between two consequtive squares (n2 )2 and (n2 + 1)2 . This argument can be more elaborate as in the following problem. Problem 6. (GDR, 1973) Find all integer solutions to the equation x(x + 1)(x + 7)(x + 8) = y 2 . Solution: Let (x, y) be any solution. Then y 2 = x(x + 8)(x + 1)(x + 7) = (x2 + 8x)(x2 + 8x + 7) = z 2 + 7z, where z = x2 + 8x. If z > 9, then (z + 3)2 = z 2 + 6z + 9 < z 2 + 7z = y 2 < z 2 + 8z + 16 = (z + 4)2 , which is impossible. Hence x2 + 8x = z 9 which implies 9 x 1. Trying these numbers one by one, we nd all solutions: (9, 12), (8, 0), (7, 0), (4, 12), (1, 0), (0, 0), (1, 12). This is a very well-known argument used in proving that 2 is irrational. We reformulate irrationality of 2 in terms of the existence of solutions of a Diophantine equation. Problem 7. Prove that the equation x2 2y 2 = 0 does not have integer solutions dierent from (0, 0). Solution: Suppose that an integer solution (x, y) exits. Then x must be even, hence x = 2x1 and (2x1 )2 2y 2 = 0 which implies 2x2 y 2 = 0. 1 Now we see that y must be also even, hence y = 2y1 . Substituting this 2 expression into the equation we get 2x2 (2y1 )2 = 0 or x2 2y1 = 0. 1 1 Hence we obtained another solution (x1 , y1 ) of the original equation with |x1 | < |x|. What is wrong with that? Well, we can now repeat everything 3

4. Innite Descent

for the new solution and get a solution (x2 , y2 ) with |x2 | < |x1 |. And we can repeat it again. There fore we obtain an innite sequence of solutions (x1 , y1 ), (x2 , y2 ), . . . , (xn , yn ), . . . with |x1 | > |x2 | > . . . > |xn | > . . . . This is impossible since all numbers |xi | are positive.

5. Quadratic Residues and Residues of Higher Order

Let p be an odd prime. The nonzero squares modulo p are called the quadratic residues (mod p). If n p, then n = kp + r, where 0 r < p and n2 r2 (mod p), hence it is enough to consider the squares of the numbers from 1 to p 1. Moreover since (p a)2 (a)2 = a2 (mod p), it is enough to consider the squares 1 2 , 22 , . . . , p1 2
2

(mod p)

These are distinct since if i2 j 2 (mod p) for i > j and 1 i, j 1 (p 1), 2 then i2 j 2 = (ij)(i+j) is divisible by p, which is only possible, when i = j, which is a contradiction. Therefore, among the numbers 0, 1, . . . , p 1 there are 1 (p 1) quadratic residues (mod p). The 1 (p 1) remaining nonzero 2 2 numbers are called quadratic nonresidues. Zero i9s neither a residue nor a nonresidue. Example 1: If p = 11, the quadratic residues (mod 11) are 1, 3, 4, 5, 9 since 12 = 1, 22 = 4, 32 = 9, , 42 = 16 5, 52 = 25 3.

We can now apply this to some Diophantine equations. If p = 3, then 1 is a quadratic residue and 2 is a nonresidue. This means that the equation x2 = 2003 cannot have a solution because 2003 2 (mod 3) and 2 is a quadratic nonresidue. The next example is the equation x2 + y 2 = z 2 . We can easily prove that it does not have any solutions with odd x and y. In such case x2 1 (mod 3) and y 2 1 (mod 3) leading to the conclusion that z 2 2 (mod 3), which we know is impossible. Problem 8. Prove that there are no positive integers a, b such that a2 3b2 = 8. Solution: Let us consider this equation modulo 3. Then we get a2 2 (mod 3) which we know means no solution. Sometimes it is useful to consider residues for composite numbers as well. 4

Problem 9. (USA, 79) Find all integer solutions of the equation x4 + x4 + + x4 = 1599. 1 2 14 (1)

Solution We claim that x4 can have only two residues (mod 16), namely 0 and 1. Indeed, if x is even, then x4 0 (mod 16). Suppose x = 2k + 1, where k is an integer. Then x4 = (2k + 1)4 = (4k 2 + 4k + 1)2 = (8s + 1)2 = 64s2 + 16s + 1 1 (mod 16)

Here we denoted k 2 + k = 2s, because k 2 + k = k(k + 1) is even. Since 1599 15 (mod 16), the equation (1) does not have solutions at all.

Problems to solve
1. Prove that there does not exist pairwise distinct positive integers x, y, z, t such that xx + y y = z z + tt . 2. x, y and z are positive integers such that x2 + y 2 = z 2 . Prove that xy is divisible by 12. 3. (Croatia, 97) Let x, y, z, a, b, c be integers such that x2 + y 2 = a2 x2 + z 2 = b2 y 2 + z 2 = c2 . Prove that the number xyz is divisible by 55. 4. (Moscow, 1963) Find all integer solutions to xy xz yz + + = 3. z y x 5. Prove that the equations (a) 6x3 + 3 = y 6 , (b) x3 + y 3 + 4 = z 3 . does not have solutions for which x, y, z are positive integers. 5

6. (Romania, 81) Find all integer solutions of the equation x6 + x3 + 1 = y 4 . 7. (Moscow, 1963) Prove that for an odd n the equation xn + y n = z n does not have integer solutions (x, y) such that x + y is a prime. 8. (Hungary, 83) Prove that the equation x3 + 3y 3 + 9z 3 9xyz = 0 has the only rational solution x = y = z = 0. 9. (USA, 76) Find all integer solutions of the equation x2 + y 2 + z 2 = x2 y 2 . 10. (GDR, 81) Prove that the equation x2 + x2 + + x2 = y 2 1 2 n has a natural solution for every n N . 11. (Moscow, 1949) prove that the only integer solution of the equation x2 + y 2 + z 2 = 2xyz is x = y = z = 0. 12. (Bulgaria, 65) Prove that there exists only one triple of positive integers greater than 1 with the property that the product of any two of these integers plus 1 is divisible by the third integer.

Solutions
2. Hint: Consider rst residues modulo 3 and then residues modulo 4. 4. Let us rst nd all solutions where x, y, z are positive integers. Multiplying the original equation by xyz we will obtain the equation x2 y 2 + x2 z 2 + y 2 z 2 = 3xyz By AM-GM inequality we get 6xyz = 2(x2 y 2 + x2 z 2 + y 2 z 2 ) = (x2 y 2 + x2 z 2 ) + (x2 y 2 + y 2 z 2 ) + (x2 z 2 + y 2 z 2 ) 2x2 yz + 2xy 2 z + 2xyz 2 = 2xyz(x + y + z), whence x + y + z 3. Since x, y, z are positive integers, we get x = y = z = 1. We notice that all three numbers cannot be negative nor can be just one of them. Hence only two of them may be negative and if we change the signs of these two negative numbers, we will get solution with all numbers positive, i.e. x = y = z = 1. Therefore we have three more solutions: (1, 1, 1), (1, 1, 1), (1, 1, 1). 5. (a) Hint: calculate cubic residues modulo 7 (b) Hint: calculate cubic residues modulo 9 7. Let x + y = p is a prime. We know that for odd n the number xn + y n is always divisible by x + y. Hence, if (x, y, z) is a solution, then z n = xn + y n is divisibly by p and hence by pn . But this is impossible since z n = xn + y n < (x + y)n = pn . 8. Let us note rst that, if a triple (x, y, z) is a nonzero solution of the equation x3 + 3y 3 + 9z 3 9xyz = 0, (3) (2)

then for any k = 0 (not necessarily an integer) the triple (kx, ky, kz) will be also a nonzero solution. Suppose that (x, y, z) is a nonzero rational solution of (??). Then for a suitably chosen k the triple

(kx, ky, kz) will be an integer solution of (??). So it is enough to prove that the equation (??) does not have a nonzero integer solutions. Now we suppose that (x, y, z) is a nonzero integer solution. If the numbers x, y, z are not relatively prime, we may divide by their greatest common divisor and still be left with a nonzero solution. Since x3 = 3y 3 9z 3 + 9xyz x3 is divisible by 3, hence x is divisble by 3, say x = 3x0 . Substituting this into (??), after cancelling 3, we will get 9x3 + y 3 + 3z 3 9xyz = 0, 0 (4)

so now y 3 is divisible by 3, whence y = 3y0 . Substituting this into (??), after cancelling 3, we will get
3 3x3 + 9y0 + z 3 9xyz = 0. 0

(5)

From (??) we notice that z is also divisible by 3, which contradicts to the assumption that the number x, y, z were coprime. This is another form of the innite descent method. 9. This is a classical innite descent problem. Firstly, we may consider that all numbers x, y, z are nonnegative. (If we nd all such solutions, then all integer solutions will be easy to nd in an obvious way.) Suppose that (x, y, z) is a nonzero integer solution with x, y, z being nonnegative. Suppose that xy is odd. Then x and y are both odd which forces z to be odd as well. Hence x2 y 2 z 2 1 (mod 3). This lead to a contradiction because x2 + y 2 + z 2 3 (mod 3) while x2 y 2 1 (mod 3). Hence xy must be even. But then x, y, z all must be even! Indeed, it is clear that an odd number of them must be even but it is impossible that one of them is even and the other two is odd because in this case x2 + y 2 + z 2 2 (mod 4) while x2 y 2 0 (mod 4). Hence x = 2x1 , y = 2y1 and z = 2z1 . Substituting into the original equation, we get
2 2 2 4x2 + 4y1 + 4z1 = 16x2 y1 1 1

or
2 2 2 x2 + y1 + z1 = 4x2 y1 . 1 1

Repeating the same procedure we will nd a triple of nonnegative integers (x2 , y2 , z2 ) satisfying
2 2 2 x2 + y2 + z2 = 16x2 y2 . 2 2

and after n steps of the same procedure we will nd a triple of nonnegative integers (xn , yn , zn ) satisfying
2 2 2 x2 + yn + zn = 4n x2 yn . n n

Since the numbers decrease x > x1 > x2 . . . we see that it was impossible to have a nonzero solution in the rst place. 10. It is clear that for any nonzero solution we have x = 0, y = 0, and z = 0. We also may consider that x > 0, y > 0, and z > 0 (changing signs if necessary). By induction we will prove that if there exists a nonzero solution (x, y, z) to the equation x2 + y 2 + z 2 = 2n xyz, then there exists a solution (x1 , y1 , z1 ) to the equation
2 2 x2 + y1 + z1 = 2n+1 x1 y1 z1 , 1

(6)

such that x1 < x, y1 < y and z1 < z. Indeed, considering this equation modulo 2 we note that either all x, y, z are even or one of them. The latter case is impossible since then x2 + y 2 + z 2 2 (mod 4) while 2n xyz 0 (mod 4). Hence x = 2x1 , y = 2y1 , and z = 2z1 . Substituting into the equation (??) we obtain
2 2 4x2 + 4y1 + 4z1 = 2n+3 x1 y1 z1 , 1

or
2 2 x2 + y1 + z1 = 2n+1 x1 y1 z1 , 1

as required. There are no innite descending sequences of positive integers, hence the existence of a nonzero solution is impossible. 11. We have a|(bc + 1), b|(ca + 1), c|(ab + 1),

where x|y means that x is a divisor of y. It is easy to see that each two of these numbers are coprime. The crucial observation is that the 9

number n = ab + bc + ca + 1 is divisible by each of them, and since these numbers are coprime, n is divisible by their product abc and, in particular, n abc. It is clear now that the numbers a, b, c cannot be too big. Let us work on this idea further. Since these numbers are coprime, they are dierent and we may assume 2 a < b < c. Then b 4 and c 5, and n = abc 2 4 5 = 40. But on the other hand n = ab + bc + ca + 1 = abc abc abc + + +1 c a b

abc abc abc abc 40 + + + 1 = abc + 1 abc + 1 < abc. 5 2 4 20 20 This contradiction proves the statement. 12. Firsly, we will prove that, for any odd number y there exists a number z such that y 2 +z 2 is again a square of an odd number. Let z = (y 2 1)/2. Then y2 + z2 = y2 + y 4 2y 2 + 1 y 4 + 2y 2 + 1 = = 4 4 y2 + 1 2
2

Since y was odd, say y = 2k + 1, then (y 2 + 1)/2 = (4k 2 + 4k + 2)/2 = 2(k 2 + k) + 1 is also odd. Now starting from x1 = 3 we can construct x2 = (32 + 1)/2 = 5, x3 = (52 + 1)/2 = 13, etc. Because of what we proved above, for any n the sum x2 + . . . + x2 will be the square of an odd number, hence n 1 the equation x2 + . . . + x2 = y 2 1 n will have a solution.

10