2001 IMO Shortlist Problems/A3: Difference between revisions
mNo edit summary |
No edit summary |
||
| (6 intermediate revisions by one other user not shown) | |||
| Line 4: | Line 4: | ||
<cmath> \frac{x_1}{1+x_1^2} + \frac{x_2}{1+x_1^2 + x_2^2} + \dotsb + \frac{x_n}{1+ x_1^2 + \dotsb + x_n^2} < \sqrt{n} . </cmath> | <cmath> \frac{x_1}{1+x_1^2} + \frac{x_2}{1+x_1^2 + x_2^2} + \dotsb + \frac{x_n}{1+ x_1^2 + \dotsb + x_n^2} < \sqrt{n} . </cmath> | ||
== Solution == | == Solution 1 == | ||
We prove the following general inequality, for arbitrary positive real <math>k</math>: | We prove the following general inequality, for arbitrary positive real <math>k</math>: | ||
| Line 25: | Line 25: | ||
<cmath> (\lvert \sin t \rvert + \sqrt{n} \cos t)/k \le (1 + n)^{1/2}(\sin^2 t + \cos^2 t)^{1/2}/k = \sqrt{n+1}/k, </cmath> | <cmath> (\lvert \sin t \rvert + \sqrt{n} \cos t)/k \le (1 + n)^{1/2}(\sin^2 t + \cos^2 t)^{1/2}/k = \sqrt{n+1}/k, </cmath> | ||
with equality only when <math>(\lvert \sin t \rvert, \cos t) = (1/\sqrt{n^2+1}, n/\sqrt{n^2+1}</math>. Since <math>\left\lvert n/\sqrt{n^2+1} \right\rvert < 1</math>, our equality cases never coincide, so we have the desired strict inequality for <math>n+1</math>. Thus our inequality is true by induction. The problem statement therefore follows from setting <math>k=1</math>. <math>\blacksquare</math> | with equality only when <math>(\lvert \sin t \rvert, \cos t) = (1/\sqrt{n^2+1}, n/\sqrt{n^2+1}</math>. Since <math>\left\lvert n/\sqrt{n^2+1} \right\rvert < 1</math>, our equality cases never coincide, so we have the desired strict inequality for <math>n+1</math>. Thus our inequality is true by induction. The problem statement therefore follows from setting <math>k=1</math>. <math>\blacksquare</math> | ||
== Solution 2 == | |||
By the [[Cauchy-Schwarz Inequality]] | |||
<math>a_1+a_2+.....+a_k \leq \sqrt{n}.\sqrt{a_1^2+a_2^2+................+a_k^2}</math> | |||
For all real numbers. <math>a_1,a_2,....</math> | |||
Hence it is only required to prove <math>a_1^2+a_2^2+................+a_k^2<1</math> where <math>a_k=\dfrac{x_k}{1+x_1^2+... x_k^2}</math> | |||
for <math>k \geq 2</math>, <math>a_k^2=(\dfrac{x_k}{1+x_1^2+... x_k^2})^2 < \dfrac{x_k^2}{(1+x_1^2+... x_{k-1}^2)(1+x_1^2+... x_k^2)} = \dfrac{1}{1+x_1^2+... x_{k-1}^2}-\dfrac{1}{1+x_1^2+... x_k^2}</math> | |||
For k=1, <math>a_1^2 = 1-\dfrac{1}{1+x_1^2}</math> | |||
Summing these inequalities, the right-hand side yields | |||
<math>\sum_{n=1}^{k}a_n^2\leq 1</math> | |||
Hence Proved by Maths1234RC | |||
P.S. This is my first solution on AOPS.<math>\blacksquare</math> | |||
== Solution 3 == | |||
Let <math>S_n = 1 + x^2 + \cdots + x_n^2</math> where <math>S_0 = 1</math> by convention. | |||
Rewriting the inequality, | |||
<cmath> | |||
\frac{x_1}{S_1} + \frac{x_2}{S_2} + \cdots + \frac{x_n}{S_n} < \sqrt{n} | |||
</cmath> | |||
must be proven. | |||
Note that by Chebyshev's inequality, the following could be driven. | |||
<cmath> | |||
\left( \sum_{i = 1}^n \frac{x_i}{S_i} \right)^2 \leq n \left( \sum_{i = 1}^n \frac{x_i^2}{S_i^2} \right) | |||
</cmath> | |||
Therefore, it suffices to prove that <math>\sum_{i = 1}^n \frac{x_i^2}{S_i^2} < 1</math>. Notice that <math>x_i^2 = S_i - S_{i - 1}</math>. | |||
<math>Lemma.</math> | |||
<cmath> | |||
\sum_{i = 1}^n \frac{S_i - S_{i - 1}}{S_i^2} < 1 | |||
</cmath> | |||
is true. | |||
<math>Proof.</math> | |||
Because <math>S_n</math> is a monotonically increasing sequence, <math>S_i^2 \geq S_i \cdot S_{i - 1}</math>. | |||
<cmath> | |||
\sum_{i = 1}^n \frac{S_i - S_{i - 1}}{S_i^2} | |||
\leq \sum_{i = 1}^n \frac{S_i - S_{i - 1}}{S_i \cdot S_{i - 1}} | |||
= \sum_{i = 1}^n \left( \frac{1}{S_{i - 1}} - \frac{1}{S_i} \right) | |||
= \frac{1}{S_0} - \frac{1}{S_n} | |||
= 1 - \frac{1}{S_n} < 1 | |||
</cmath> | |||
<math>\blacksquare</math> | |||
Hence, the inequality is true. | |||
~MaPhyCom (Thinking Tree) | |||
I also have a video explanation of this solution!! | |||
[https://youtu.be/mjUccETe65w Please consider subscribing :))] | |||
Latest revision as of 03:12, 25 September 2025
Problem
Let
be arbitrary real numbers. Prove the inequality
Solution 1
We prove the following general inequality, for arbitrary positive real
:
with equality only when
.
We proceed by induction on
. For
, we have trivial equality. Now, suppose our inequality holds for
. Then by inductive hypothesis,
If we let
, then we have
with equality only if
.
By the Cauchy-Schwarz Inequality,
with equality only when
. Since
, our equality cases never coincide, so we have the desired strict inequality for
. Thus our inequality is true by induction. The problem statement therefore follows from setting
.
Solution 2
By the Cauchy-Schwarz Inequality
For all real numbers.
Hence it is only required to prove
where
for
,
For k=1,
Summing these inequalities, the right-hand side yields
Hence Proved by Maths1234RC
P.S. This is my first solution on AOPS.
Solution 3
Let
where
by convention.
Rewriting the inequality,
must be proven.
Note that by Chebyshev's inequality, the following could be driven.
Therefore, it suffices to prove that
. Notice that
.
![]()
is true.
Because
is a monotonically increasing sequence,
.
Hence, the inequality is true.
~MaPhyCom (Thinking Tree)
I also have a video explanation of this solution!! Please consider subscribing :))
Alternate solutions are always welcome. If you have a different, elegant solution to this problem, please add it to this page.