Scribd
Upload
16 views4 pages

Ostrowski Q

The document discusses Ostrowski's theorem, which states that every nontrivial absolute value on rational numbers Q is either the ordinary absolute value or a p-adic absolute value for some prime p. The proof uses properties of absolute values and considers two cases depending on whether the absolute value is greater than 1 for some integer. It is shown that in each case the absolute value must be a power of the ordinary or p-adic absolute values.

Uploaded by

Horus Ion
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
16 views4 pages

Ostrowski Q

The document discusses Ostrowski's theorem, which states that every nontrivial absolute value on rational numbers Q is either the ordinary absolute value or a p-adic absolute value for some prime p. The proof uses properties of absolute values and considers two cases depending on whether the absolute value is greater than 1 for some integer. It is shown that in each case the absolute value must be a power of the ordinary or p-adic absolute values.

Uploaded by

Horus Ion
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 4

OSTROWSKI’S THEOREM FOR Q

KEITH CONRAD

1. Introduction
Hensel created the p-adic numbers towards the end of the 19th century, and it wasn’t
until about 20 years later that Ostrowski [1] proved a fundamental theorem that explained
in retrospect why Hensel’s idea was natural: every nontrivial absolute value on Q is a power
of the ordinary (archimedean) absolute value or a power of a p-adic absolute value for some
prime number p, so every completion of Q with respect to a nontrivial absolute value is
either R or some Qp .
Theorem 1 (Ostrowski, 1916). If | · | is a nontrivial absolute value on Q then there is t > 0
such that either | · | = | · |t∞ or | · | = | · |tp for a prime p.
Proof. An absolute value on Q is determined by its values on the positive integers, so it
suffices to show there is a t > 0 such that |n| = nt for all n in Z+ or |n| = |n|tp for some
prime p and all n in Z+ .
Since | · | is nontrivial, |n| =
6 1 for some positive integer n. We consider two cases: |n| > 1
for some n ≥ 2 or |n| ≤ 1 for all n ≥ 2. We will show in the first case that | · | is a power of
the ordinary absolute value on Q and in the second case that | · | is a power of some p-adic
absolute value.
Case 1: |n| > 1 for some n ≥ 2.
First we prove that |n| > 1 for all n ≥ 2 by proving the contrapositive: if |n0 | ≤ 1 for
some n0 ≥ 2 then |n| ≤ 1 for all n ≥ 2. Write n in base n0 :
n = a0 + a1 n0 + · · · + ad nd0
where 0 ≤ ai ≤ n0 − 1 and ad 6= 0, so nd0 ≤ n < nd+1
0 . We have |ai | ≤ |1 + 1 + · · · + 1| ≤
|1| + |1| + · · · + |1| = ai < n0 , so
(1) |n| ≤ |a0 | + |a1 ||n0 | + · · · + |ad ||n0 |d < n0 + n0 |n0 | + · · · + n0 |n0 |d .
From |n0 | ≤ 1, (1) implies |n| ≤ n0 (d + 1) ≤ m(logn0 (n) + 1). Replace n by nk in this
inequality to get |n|k ≤ n0 (k logn0 (n) + 1), so
q
(2) |n| ≤ k n0 (k logn0 (n) + 1).

We have logn0 (n) > 0 since n0 > 1 and n > 1, so letting k → ∞ in (2) shows us that
|n| ≤ 1, and n was arbitrary.
The replacement of n with nk is an idea we will use again. Let’s call it the “power trick.”
For any integers m and n that are greater than 2, |m| > 1 and |n| > 1. Picking d ≥ 0
so that md ≤ n < md+1 , writing n in base m implies (in the same way that we proved (1)
above)
|n| ≤ m(1 + |m| + · · · + |m|d ).
1
2 KEITH CONRAD

Since |m| > 1, summing up the finite geometric series on the right gives us

|m|d+1 − 1 |m|d+1 m|m|


|n| ≤ m(1 + |m| + · · · + |m|d ) = m <m = |m|d .
|m| − 1 |m| − 1 |m| − 1
Since d ≤ logm (n),
m|m|
|n| < |m|logm (n) .
|m| − 1
Now it’s time for the power trick. Replacing n with nk ,
m|m|
|n|k < |m|k logm (n) .
|m| − 1
Taking kth roots,
s
m|m|
|n| < k
|m|logm (n) ,
|m| − 1
and letting k → ∞,

(3) |n| ≤ |m|logm (n) .

Writing |m| = ms and |n| = nt where s > 0 and t > 0, we get from (3) that nt ≤ ms logm (n) =
ns , so t ≤ s. The roles of m and n in this calculation are symmetric, so by switching their
roles we get s ≤ t and thus |m| = mt and |n| = nt .
Case 2: |n| ≤ 1 for all n ≥ 2.
For some n ≥ 2 we have |n| = 6 1, so 0 < |n| < 1. Let p be the smallest such positive
integer. Since 0 < |p| < 1 and also 0 < 1/p < 1, we can write |p| = (1/p)t for some t > 0.
We will prove |n| = |n|tp for all n ≥ 1.
The number p is prime, by contradiction: if p = ab where a and b are positive integers
that are both smaller than p then |a| = 1 and |b| = 1, so |p| = |a||b| = 1, which is false.
Next we show each positive integer m not divisible by p has |m| = 1. If |m| 6= 1 then
|m| < 1. We are going to use the power trick again: let’s look at pk and mk . Since |p and
|m| are both between 0 and 1, for a large k we have |p|k < 1/2 and |n|k < 1/2. Since pk
and mk are relatively prime, there are xk and yk ∈ Z such that 1 = pk xk + mk yk . Take the
absolute value of both sides:
1 1
1 = |pk xk + mk yk | ≤ |pk ||xk | + |mk ||yk | ≤ |p|k + |m|k < + = 1,
2 2
which is a contradiction.
For all integers n ≥ 2 pull out the largest power of p: n = pe n0 where e ≥ 0 and n0 is not
divisible by p. Then |n0 | = 1, so |n| = |pe n0 | = |p|e |n0 | = |p|e = (1/p)et . Also |n|p = (1/p)e ,
so |n| = |n|tp . 

Here is a second proof that an absolute value | · | on Q such that |n| > 1 for some positive
integer n ≥ 2 must be a power of the ordinary absolute value on Q.
First we show |2| > 1 by an argument very close to that used already in Case 1, but
we repeat it here to keep our argument self-contained. Assuming |2| ≤ 1 we will get a
contradiction.
OSTROWSKI’S THEOREM FOR Q 3

Write each integer n ≥ 2 in base 2: n = a0 + a1 · 2 + · · · + ad 2d where ai is 0 or 1 and


ad = 1, so 2d ≤ n < 2d+1 . Thus |ai | is 0 or 1, so by the triangle inequality
d
X d
X
|n| ≤ |ai ||2|i ≤ 1 = d + 1 ≤ log2 (n) + 1 ≤ 2 log2 (n).
i=0 i=0

This holds for all n ≥ 2, so if we replace n throughout with nk for k ≥ 1 then


|nk | ≤ 2 log2 (nk ) = 2k log2 (n),
so
|n|k ≤ 2k log2 (n).
Taking kth roots of both sides, p
k
|n| ≤ 2k log2 (n).
Letting k → ∞, this inequality becomes |n| ≤ 1. We have proved this for all n ≥ 2, but
that contradicts the assumption |n| > 1 for some n ≥ 2, so in fact we must have |2| > 1.
Since |2| and 2 are both greater than 1, we can write |2| = 2t for some t > 0. We will
prove |n| = nt for all n ≥ 2 by proving |n| ≤ nt (easier) and |n| ≥ nt (trickier).
As we have already done, write an integer n ≥ 1 in base 2: n = a0 + a1 · 2 + · · · + ad 2d
with ai equal to 0 or 1 and ad = 1, so 2d ≤ n < 2d+1 . An upper bound on n follows easily
from the triangle inequality:
|2|d+1 − 1
|n| ≤ |a0 | + |a1 ||2| + · · · + |ad ||2|d ≤ 1 + |2| + · · · + |2|d = .
|2| − 1
Writing |2| as 2t ,
2t(d+1) − 1 2t(d+1) 2t td 2t
|n| ≤ < = 2 ≤ nt .
2t − 1 2t − 1 2t − 1 2t − 1
It’s time to use the power trick again: replacing n in this inequality by nk with k ≥ 1,
2t
|n|k < nkt .
2t − 1
Taking kth roots of both sides implies
r
2t
nt .
k
|n| ≤
2t − 1
Letting k → ∞ (keeping n fixed), we get
(4) |n| ≤ nt
for all n ∈ Z+ (it is obvious at n = 1).
To prove the reverse inequality |n| ≥ nt for n ≥ 2, once again write n in base 2: n =
a0 + a1 · 2 + · · · + ad 2d with ai = 0 or 1 and ad = 1, so 2d ≤ n < 2d+1 . Once again we use
the triangle inequality, but in a less obvious way:
|2d+1 | = |2d+1 − n + n| ≤ |2d+1 − n| + |n|.
On the left side, |2d+1 | = |2|d+1 = 2t(d+1) . On the right side, since 2d+1 − n is a positive
integer we get |2d+1 − n| ≤ (2d+1 − n)t by (4), so
2t(d+1) ≤ (2d+1 − n)t + |n|.
4 KEITH CONRAD

From this we obtain a lower bound on |n|:


|n| ≥ 2t(d+1) − (2d+1 − n)t .
To decrease this lower bound we can increase 2d+1 − n: since n is between 2d and 2d+1 , we
have 2d+1 − n ≤ 2d+1 − 2d = 2d , so
|n| ≥ 2t(d+1) − 2td = (2t − 1)2td ≤ (2t − 1)nt .
One more time we will use the power trick: replace n by nk and take kth roots to get
p
|n| ≥ 2t − 1 nt .
k

Letting k → ∞, we get |n| ≥ nt . Since we already showed |n| ≤ nt , we have shown |n| = nt
for all n ∈ Z+ .

References
[1] A. Ostrowski, Über einige Lösungen der Funktionalgleichung, Acta Arith. 41 (1916), 271–284.

You might also like

pFad - Phonifier reborn

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


Alternative Proxies:

Alternative Proxy

pFad Proxy

pFad v3 Proxy

pFad v4 Proxy