Homepage Solution manuals Ivan Niven An Introduction to the Theory of Numbers Exercise 3.4.9 (Forms whose discriminant is a perfect square)

Exercise 3.4.9 (Forms whose discriminant is a perfect square)

Let f ( x , y ) = a x 2 + bxy + c y 2 be a quadratic form with integral coefficients whose discriminant d is a perfect square, possibly 0 . Show that there are integers h 1 , h 2 , k 1 , and k 2 such that f ( x , y ) = ( h 1 x + k 1 y ) ( h 2 x + k 2 y ) .

Hint. Suppose that a 0 . Show that there are rational numbers r 1 and r 2 such that f ( x , y ) = a ( x r 1 y ) ( x r 2 y ) . Argue that a r 1 r 2 , and hence there exist integers h 1 and h 2 such that h 1 h 2 = a , h 1 r 1 , h 2 r 2 . Treat the case a = 0 separately.

Answers

I didn’t understand the “hence” in the hint, so I’m going to do it a little differently.

Proof.

  • Suppose first that a 0 . Consider here x , y as (indeterminate) variables in the field ( x , y ) of rational functions of two variables x , y . Since d is a perfect square, Problem 7 shows that

    a u 2 + bu + c = a ( u r 1 ) ( u r 2 ) ,

    where r 1 , r 2 are rational numbers. For u = x y , we obtain

    f ( x , y ) = y 2 ( a ( x y ) 2 + b x y + c ) ) = a y 2 ( x y r 1 ) ( x y r 2 ) = a ( x r 1 y ) ( x r 2 y ) .

    Then

    a x 2 + bxy + c = a ( x r 1 y ) ( x r 2 y ) = a x 2 a ( r 1 + r 2 ) xy + a r 1 r 2 y 2 .

    Therefore a ( r 1 + r 2 ) = b , a r 1 r 2 = c .

    Write r 1 = s 1 t 1 , r 2 = s 2 t 2 , where s 1 t 1 = 1 , s 2 t 2 = 1 . Put

    α 1 = a r 1 , α 2 = a r 2 .

    Then α 1 , α 2 are the roots of the polynomial

    p ( u ) = ( u a r 1 ) ( u a r 2 ) = u 2 + bu + ac .

    Moreover α 1 , α 2 are rational numbers. If we write α 1 = p q , p q = 1 , q > 0 , then p 2 + bpq + q 2 = 0 , thus q p 2 , where p q = 1 , therefore q 1 (and q > 0 ), thus q = 1 . Hence α 1 , and similarly α 2 are integers (more generally, in algebraic terms, an algebraic integer which is in is a rational integer). This shows that t 1 a s 1 , where t 1 s 1 = 1 , thus t 1 a , and similarly t 2 a , so a = q 1 t 1 = q 2 t 2 , where q 1 , q 2 are integers.

    Since c = a r 1 r 2 = a s 1 s 2 t 1 t 2 , then

    f ( x , y ) = a ( x + s 1 t 1 y ) ( x + s 2 t 2 y ) = a t 1 t 2 ( t 1 x + s 1 y ) ( t 2 x + s 2 y ) = ( q 2 x + c s 2 y ) ( t 2 x + s 2 y ) ,

    where all coefficients are integers, except perhaps c s 2 . But b = q 2 s 2 + c t 2 s 2 (coefficient of xy ), thus c t 2 s 2 is an integer, so s 2 c t 2 , where s 2 t 2 = 1 , therefore s 2 c .

    Put

    h 1 = q 2 = a t 2 , k 1 = c s 2 , h 2 = t 2 , k 2 = s 2 .

    Then h 1 , k 1 , h 2 , k 2 are integers, and

    f ( x , y ) = ( h 1 x + k 1 y ) ( h 2 x + k 2 y ) .

  • Suppose now that a = 0 . Then

    f ( x , y ) = y ( bx + cy ) = ( h 1 x + k 1 y ) ( h 2 x + k 2 y ) ,

    where h 1 = 0 , k 1 = 1 , h 2 = b , k 2 = c are integers.

If f ( x , y ) is a quadratic form with integral coefficients whose discriminant d is a perfect square, there are integers h 1 , h 2 , k 1 , k 2 such that f ( x , y ) = ( h 1 x + k 1 y ) ( h 2 x + k 2 y ) .

User profile picture
2024-11-14 10:46
Comments

I propose a more symmetric proof.

Second proof.

Lemma Let a , b , c be integers. If a bc , there are integers a 1 , a 2 such that

a = a 1 a 2 , a 1 b , a 2 c .

Proof. Put a 1 = a b , and a 2 = a a 1 = a a b . Then a = a 1 a 2 and a 1 a . Moreover a bc , therefore

a a b b a b c ,

where a a b and b a b are relatively prime. Hence a 2 = a a b c . □

Now we prove the sentence.

Proof. Suppose first that a = 0 . Then

f ( x , y ) = y ( bx + cy ) = ( h 1 x + k 1 y ) ( h 2 x + k 2 y ) ,

where h 1 = 0 , k 1 = 1 , h 2 = b , k 2 = c are integers.

Now we suppose that a 0 . Since the discriminant d is a perfect square, there is an integer δ such that d = δ 2 . By equation (3.3),

4 af ( x , y ) = ( 2 ax + by ) 2 δ 2 y 2 = ( 2 ax + by δy ) ( 2 ax + by + δy ) .

Therefore

f ( x , y ) = 1 a ( ax + b δ 2 y ) ( ax + b + δ 2 y ) , (1)

where b δ 2 , b + δ 2 are integers. Indeed, b and d = b 2 4 ac have same parity. If b is even, 4 d , so δ is even, and if b is odd, d and δ are odd.

Moreover

b δ 2 b + δ 2 = b 2 d 4 = b 2 ( b 2 4 ac ) 4 = ac .

Thus a b δ 2 b + δ 2 . By the Lemma, there are integers a 1 , a 2 such that

a = a 1 a 2 , a 1 b δ 2 , a 2 b + δ 2 .

By equation (1),

f ( x , y ) = ( a a 1 x + 1 a 1 b δ 2 ) ( a a 2 x + 1 a 2 b + δ 2 ) = ( h 1 x + k 1 y ) ( h 2 x + k 2 y ) ,

where

h 1 = a a 1 , k 1 = 1 a 1 b δ 2 , h 2 = a a 2 , k 2 = 1 a 2 b + δ 2

are integers.

If f ( x , y ) is a quadratic form with integral coefficients whose discriminant d is a perfect square, there are integers h 1 , h 2 , k 1 , k 2 such that f ( x , y ) = ( h 1 x + k 1 y ) ( h 2 x + k 2 y ) .   □

User profile picture
2024-11-16 09:22
Comments