Homepage Solution manuals Ivan Niven An Introduction to the Theory of Numbers Exercise 6.1.8 ($\sum\limits_{(b,b') \in A} (bb')^{-1} = 1$ )

Exercise 6.1.8 ($\sum\limits_{(b,b') \in A} (bb')^{-1} = 1$ )

Show that if n is a positive integer then ( b b ) 1 = 1 , where the sum is over all pairs ( b , b ) of integers for which 1 b n , 1 b n , g . c . d ( b , b ) = 1 , and b + b > n .

Answers

Proof. Consider the set

A = { ( b , b ) × 1 b n , 1 b n , b b = 1 , b + b > n } .

We must prove that ( b , b ) A ( b b ) 1 = 1 .

By section 6.1, all the terms b j of the Farey sequence of order n satisfy

1 b j n , 1 b j + 1 n , b j b j + 1 = 1 , b j + b j + 1 > n .

This shows that all ordered pairs ( b j , b j + 1 ) are in A , so we may consider the map

φ { [ [ 1 , k 1 ] ] A j ( b j , b j + 1 ) .

  • We prove that φ is surjective (onto). Let ( b , b ) A , so that

    1 b n , 1 b n , b b = 1 , b + b > n .

    • Consider first the particular case b = 1 . Then b + b > n implies b = n . By Problem 5, a k 1 b k 1 = ( n 1 ) n , and a k b k = 1 1 , therefore ( b , b ) = ( n , 1 ) = ( b k 1 , b k ) = φ ( k 1 ) .
    • We suppose now that b > 1 .

      Since b b = 1 , there are integers u , v such that bv u b = 1 . Then for all integers λ ,

      b ( v λ b ) ( u λb ) b = 1 .

      We take for λ the quotient of the Euclidean division of v by b , so that

      v = λ b + a , 0 a < b .

      Then a = v λ b and 0 a < b . Put a = u λb . Then

      b a a b = 1 , a b < a b < 1 .

      It remains to show that a 0 . If a < 0 then 1 = b a a b a b = | a | b > 1 (because b > 1 ). This is a contradiction, thus a 0 , so

      0 a b < a b < 1 , where  b a a b = 1 , 1 b n , 1 b n .

      Then a b = 1 and a b = 1 , so a b and a b are both fractions of the Farey sequence of order n . Moreover b + b > n . As in Problem 5, consider any fraction x y ( y > 0 ) between a b and a b , so that

      a b < x y < a b . (1)

      Since b a a b = 1 ,

      1 b b = a b a b = ( a b x y ) + ( x y a b ) = a y b x b y + bx ay by .

      By (1), a y b x > 0 and bx ay > 0 , therefore

      1 b b 1 b y + 1 by = b + b b b y ,

      where b b y > 0 , thus

      y b + b > n ,

      This shows that no fraction x y between a b and a b is in the Farey sequence of order n . Hence a b < a b are adjacent fractions in this sequence. By definition of the sequence ( b j ) , this means that there is some j [ [ 1 , k 1 ] ] such that ( b , b ) = ( b j , b j + 1 ) = φ ( j ) .

    In either case, there is some j [ [ 0 , k 1 ] ] such that ( b , b ) = φ ( j ) , so φ is surjective.

  • φ is injective (one-to-one):

    Suppose that φ ( j ) = φ ( i ) , 1 i k 1 , 1 j k 1 . Then ( b j , b j + 1 ) = ( b i , b i + 1 ) .

    We write for simplicity

    b = b j = b i , d = b j + 1 = b i + 1 , a = a j , a = a i c = a j + 1 , c = a i + 1 .

    We may assume without loss of generality that a j a i , so

    a a . (2)

    Since a j b j = a b and a j + 1 b j + 1 = c d are adjacent in the Farey sequence of order n (and also a i b j = a b and a i + 1 b i + 1 = c d ), we have

    0 a b < c d 1 , cb ad = 1 , b + d > n , (3) 0 a b < c d 1 , c b a d = 1 . (4)

    Then b ( c c ) = d ( a a ) , so b d ( a a ) , where b d = 1 , thus b a a , so a a = λb for some λ , thus b ( c c ) = dλb , where b 0 , so c c = λd . This gives

    a = a + λb , c = c + λd ,

    and by (2), λ 0 , so λ .

    Therefore

    a b = a b + λ , c d = c d + λ .

    • If c d = 1 , then by (4), c d = ( c d ) + λ = 1 + λ 1 , where λ 0 , thus λ = 0 .
    • If c d < 1 , then by (3) and (4), 0 < c d < 1 and 0 < ( c d ) + λ 1 , thus

      λ < c d + λ < 1 + λ , 0 < c d + λ 1 ,

      which gives by transitivity

      λ < 1 , 0 < 1 + λ ,

      therefore λ = 0 .

    In either case, λ = 0 . Hence a = a , c = c , so a j b j = a i b i . Since the sequence ( a j b j ) is strictly increasing, we obtain i = j . This shows that φ is injective.

We have proved that

φ { [ [ 1 , k 1 ] ] A j ( b j , b j + 1 ) .

is bijective, i.e. for every pair ( b , b ) A , there is a unique j , 1 j k 1 , such that b = b j and b = b j + 1 . The change of indices ( b , b ) = φ ( j ) gives, using the result of Problem 7,

1 = j = 1 k 1 ( b j b j + 1 ) 1 = ( b , b ) A ( b b ) 1

So ( b b ) 1 = 1 , where the sum is over all ordered pairs ( b , b ) of integers for which 1 b n , 1 b n , g . c . d ( b , b ) = 1 , and b + b > n . □

User profile picture
2025-07-05 09:46
Comments