Cauchy
by math_explorer, Jun 21, 2012, 12:41 PM
![\[ \sum_{i=1}^n \frac{a_i^2}{b_i} \geq \frac{(\sum_{i=1}^n a_i)^2}{\sum_{i=1}^n b_i}\]](http://latex.artofproblemsolving.com/2/b/2/2b2ee04bd3aa0bd567df3373d9d9fd43b34b1e0d.png)
variously known as Engel form of C-S or Titu's lemma
This might be why I can't apply Cauchy often enough.
---


Solutions:
all
for a constant
, which clearly work.


Proof:
Obviously the solutions work. Let
denote the functional equation. Take
and
to get
... (1)
... (2)
Substituting both equations into the original becomes

which, by substituting
,
, is equivalent to the Cauchy functional equation
... (*).
In particular,
so
.
Next, setting
in (1), (2) and equating the two equations, we get
for
... (3)
But for
we already know both sides are 0, so (3) still holds.
Now, for any real
, put
in (3) to evaluate
, and then split into more terms with (*):
For a different evaluation, expand
, split into more terms with (*), and then apply (3):
And equating the last two equations above results in
for all real
, so we're done.





Substituting both equations into the original becomes

which, by substituting
![$x = \sqrt[3]{a}$](http://latex.artofproblemsolving.com/f/6/f/f6f4801d5f694440a3025b16def7878c8ec37129.png)
![$y = \sqrt[3]{b}$](http://latex.artofproblemsolving.com/6/5/e/65eef731baee6e4cfacd93c7deaf9f8b844f3737.png)

In particular,


Next, setting



But for

Now, for any real



\[ \begin{align*} f((y+1)^2) &= (y+1)f(y+1) \\ &= (y+1)(f(y)+f(1)) \\ &= yf(y) + f(y) + f(1)y + f(1) \]
For a different evaluation, expand

\[ \begin{align*} f((y+1)^2) &= f(y^2 + 2y + 1) \\ &= f(y^2) + 2f(y) + f(1) \\ &= yf(y) + f(y) + f(y) + f(1) \]
And equating the last two equations above results in


I spent way too long on this problem before solving it.
But then I spend way too long on geometry problems and still can't solve them.