Homepage Solution manuals Ivan Niven An Introduction to the Theory of Numbers Exercise 4.1.23* (When $\lfloor n \alpha \rfloor, \lfloor m \beta \rfloor$ fill $\mathbb{N}^*$?)

Exercise 4.1.23* (When $\lfloor n \alpha \rfloor, \lfloor m \beta \rfloor$ fill $\mathbb{N}^*$?)

Let 𝒮 be the set of integers given by ax and bx for x = 1 , 2 , . Prove that 𝒮 consists of every positive integer, each appearing exactly once, if and only if α and β are positive irrational numbers such that 1 α + 1 β = 1 .

Answers

Proof. Suppose first that α and β are positive irrational numbers such that 1 α + 1 β = 1 . Then 1 α = 1 1 β < 1 , thus α > 1 . We can write α = 1 + λ , where λ is a positive irrational number.

Then 1 β = 1 1 α = ( α 1 ) α = λ ( λ + 1 ) , so β = 1 + λ 1 . Therefore

𝒮 = { αx , x } { βx , x } = { x + λx , x } { x + λ 1 x , x } ,

where λ is a positive irrational number.

By Problem 22, we know that 𝒮 = , and that every positive integer appears exactly once among the terms αx = x + λx together with βx = x + λ 1 x , where x .

Conversely, suppose that 𝒮 consists of every positive integer, each appearing exactly once.

If α < 1 , then α 1 = 0 , in contradiction with the hypothesis. Therefore α > 1 , and similarly β > 1 , so we can write α = 1 + λ , β = 1 + μ , where λ > 0 , μ > 0 .

For all positive integers n , m , we define

u n = αn = n + λn = n + λn , v m = βm = m + μm = m + μm .

As in part (a) of the solution of Problem 22, the sequences u = ( u n ) n , v = ( v m ) m are strictly increasing.

If λ 1 and μ 1 , then u 1 2 and v 1 2 thus the positive integer 1 is not represented by any term of the sequences u , v . Therefore λ < 1 or μ < 1 .

If λ < 1 and μ < 1 , then u 1 = v 1 = 1 , so 1 is not appearing exactly once. Since λ and μ are irrational, λ 1 and μ 1 . Therefore the only two possibilities are λ < 1 and μ > 1 , or λ > 1 and μ < 1 . Since α and β play a symmetric role, these two possibilities are treated in the same way, so we assume from now that

λ > 1 , 0 < μ < 1 .

It remains to prove that μ = λ 1 (and then 1 α + 1 β = 1 ). As in part (b) of the solution of Problem 22, we prove that

v m + 1 v m { 1 , 2 } , (1) v m + 1 v m = 2 k , m = k μ 1 . (2)

Indeed,

v m + 1 v m = ( m + 1 ) μ + 1 , 0 < μ < 1 , < + 1 , < ( m + 1 ) μ < + 1 + μ < + 2 ,

Therefore ( m + 1 ) μ = or ( m + 1 ) μ = , which gives (1):

v m + 1 v m { 1 , 2 } .

Moreover, if v m + 1 v m = 2 , then

( m + 1 ) μ = + 1 ( = k ) ,

thus < + 1 = k , so

k ( m + 1 ) μ < k + μ , k μ 1 m + 1 < k μ 1 + 1 , m < k μ 1 < m + 1 , ( since  μ )

so m = k μ 1 .

Conversely, if m = k μ 1 for some k , then, using μ , we obtain m < k μ 1 < m + 1 , thus k μ 1 < m + 1 < k μ 1 + 1 , so k < ( m + 1 ) μ < k + μ < k + 1 . Therefore ( m + 1 ) μ = k . Moreover m < k μ 1 < m + 1 implies < k < ( m + 1 ) μ , thus k 1 < k μ < < k , so k = 1 + . This proves ( m + 1 ) μ = 1 + . Thus

v m + 1 v m = ( m + 1 ) α 1 m α 1 + 1 = 2 .

The equivalence (2) is proven.

These two items (1) and (2) mean that the only integers not represented by the sequence v are the integers v k μ 1 + 1 , where k . The least such integer is v μ 1 + 1 . This integer must be represented by the least term of the sequence u , so u 1 = v μ 1 + 1 .

If we suppose that u k = v k μ 1 + 1 for all positive integers k < n , then v n μ 1 + 1 must be represented by the least term of the sequence u greater than u n 1 . Therefore u n = v n μ 1 + 1 . This induction shows that, for all n ,

u n = v n μ 1 + 1 .

This is equivalent to

n + λn = n μ 1 + n μ 1 μ .

We know that λn λn in the neighborhood of + , if λ > 0 . Therefore

( 1 + λ ) n n + λn = n μ 1 + n μ 1 μ ( 1 + μ 1 ) n .

Then ( 1 + λ ) n ( 1 + μ 1 ) n shows that 1 + λ = 1 + μ 1 , thus λ = μ 1 , μ = λ 1 . Since α = 1 + λ , β = 1 + μ , we obtain

1 α + 1 β = 1 .

User profile picture
2024-12-21 11:22
Comments