Homepage Solution manuals Ivan Niven An Introduction to the Theory of Numbers Exercise 3.7.6 (Representations of $p$ by forms of discriminant $-23$)

Exercise 3.7.6 (Representations of $p$ by forms of discriminant $-23$)

Show that any positive definite quadratic form of discriminant 23 is equivalent to exactly one of the forms f 0 ( x , y ) = x 2 + xy + 6 y 2 , f 1 ( x , y ) = 2 x 2 + xy + 3 y 2 or f 2 ( x , y ) = 2 x 2 xy + 3 y 2 . Show that if 23 p = 1 then p is not represented by any of these forms. Show that if 23 p = 1 then p has a total of 4 representations by these forms. Show that in the latter case either p has 4 representations by f 0 (*) or 2 representations apiece by f 1 and f 2 . Determine which of these cases applies when p = 139 .

Hint. Find h such that h 2 23 ( mod 4 139 ) , and then reduce the form 139 x 2 + hxy + k y 2 .

(*) Typo corrected (R.G.)

Answers

Proof.

(a)
Let f = ( a , b , c ) a reduced positive definite quadratic form of discriminant d = 23 . By Theorem 3.19, 0 < a d 3 = 23 3 , thus a = 1 or a = 2 .

If a = 1 , then 1 < b 1 , and b 2 4 c = 23 , thus b is odd, so b = 1 , and c = ( b 2 + 23 ) 4 = 6 , which gives f = f 0 = ( 1 , 1 , 6 ) .

If a = 2 , then 2 < b 2 , and b is odd, thus b = ± 1 , and c = ( b 2 + 23 ) 8 = 3 , so f = f 1 = ( 2 , 1 , 3 ) or f = f 2 = ( 2 , 1 , 3 ) . These three reduced forms are primitive, so

H ( 23 ) = h ( 23 ) = 3 .

By Theorem 3.25, these three forms are not equivalent. Since every form f of discriminant 23 is equivalent to a reduced form, f is primitive, and equivalent to exactly one of the forms f 0 , f 1 , f 2 .

(b)
Let p be a prime number such that ( 23 p ) = 1 . Then p 2 , p 23 , and d = 23 1 ( mod 4 ) satisfies p d and ( d p ) 1 . Then Corollary 3.14 shows that p is not represented by any form of discriminant 23 , so is not represented by f 0 , f 1 , f 2 .
(c)
The prime 2 is not represented by f 0 , and has two representations apiece by f 1 and f 2 . We suppose now that p is an odd prime.

Assume now that ( 23 p ) = 1 . By Corollary 3.14, there is a binary quadratic form f of discriminant 23 that represents p . Since f is properly equivalent to f 0 , f 1 or f 2 , p is represented by one of these three forms.

By the last formula of section 3.7, applied to n = p ,

f r f ( p ) = w N d ( p ) ,

where = { f 0 , f 1 , f 2 } , and by Theorem 3.26, w = w ( f 0 ) = w ( f 1 ) = w ( f 2 ) = 2 .

By definition,

N d ( p ) = Card { u [ [ 0 , 2 p [ [ u 2 23 ( mod 4 p ) } = 1 2 Card { u [ [ 0 , 4 p [ [ u 2 23 ( mod 4 p ) }

Since ( 23 p ) = 1 , the congruence x 2 23 ( mod p ) has two solutions, and the congruence x 2 23 1 ( mod 4 ) has also 2 solutions, so the congruence x 2 23 ( mod 4 p ) has 4 solutions, so N d ( p ) = 4 , therefore

i = 0 2 r f i ( p ) = 4 .

Since every representation of the prime p is proper, we can write this formula in the form

R f 0 ( p ) + R f 1 ( p ) + R f 2 ( p ) = 4 ,

which means that p has a total of 4 representations by the forms f 0 , f 1 , f 2 .

Since f 1 ( x , y ) = f 2 ( x , y ) , there are as many representations of p by f 1 and f 2 , so

R f 1 ( p ) = R f 2 ( p ) .

Put a = R f 0 ( p ) , b = R f 1 ( p ) = R f 2 ( p ) . Then a + 2 b = 4 . The only solutions ( a , b ) 2 of a + 2 b are ( 4 , 0 ) , ( 2 , 1 ) and ( 0 , 2 ) . Since f 1 ( x , y ) = f 1 ( x , y ) , R f 1 ( p ) = 1 is impossible, so the only possibilities are

R f 0 ( p ) = 4 , R f 1 ( p ) = R f 2 ( p ) = 0 , or R f 0 ( p ) = 0 , R f 1 ( p ) = R f 2 ( p ) = 2 .
(d)
Consider the example p = 139 . Following the hint, we search a solution of the congruence h 2 23 ( mod 4 p ) . The Tonelli-Shanks algorithm gives a square root of 23 modulo p : 3 3 2 23 ( mod 139 ) .

Moreover

1 2 23 ( mod 4 ) .

A solution of the system of congruence

{ h 33 ( mod 139 ) h 1 ( mod 4 )

is h = 33 (since 33 1 ( mod 4 ) ), thus

3 3 2 23 ( mod 4 139 ) .

( Check: 3 3 2 + 23 = 1112 = 2 ( 4 139 ) .) Now we take k such that the form g ( x , y ) = 139 x 2 + hxy + k y 2 satisfies d = discr ( g ) = 3 3 2 4 129 k = 23 , which gives k = 2 , so

g ( x , y ) = 139 x 2 33 xy + 2 y 2 .

Now we reduce g with the usual algorithm (see the program given in the notes of Problem 3.6.4.) We obtain

g = ( 139 , 33 , 2 ) , g S = ( 2 , 33 , 139 ) g ( S T 8 ) = ( 2 , 1 , 3 ) = f 1

Thus

g A = f 1 ,

where A = S T 8 = ( 0 1 1 8 ) .

Verification:

g ( y , x + 8 y ) = 139 y 2 33 y ( x + 8 y ) + 2 ( x + 8 y ) 2 = 2 x 2 + xy + 3 y 2 = f 1 ( x , y ) .

So g and f 1 are properly equivalent. Since p = 139 = g ( 1 , 0 ) is represented by g , we obtain that p is represented by the equivalent form f 1 . By part (c), we obtain that p has two representations apiece by f 1 and f 2 and no representation by f 0 .

Explicitly, p = g ( 1 , 0 ) = ( f 1 A 1 ) ( 1 , 0 ) = f 1 ( 8 , 1 ) = 2 8 2 + 1 8 + 3 1 2 , so

p = 139 = f 1 ( 8 , 1 ) = f 1 ( 8 , 1 ) , 139 = f 2 ( 8 , 1 ) = f 2 ( 8 , 1 ) .

User profile picture
2024-12-08 11:07
Comments