Tuesday, 11 July 2017

real analysis - Prove a "simple" inequality




I came across the following inequality and I would like to know if it is true
(xxyy)1yx(xy)2x+y,
which is the same as proving
exlogxylogyxy(xy)2x+y.



When I do the simulations it seems it really holds, for any x,y0 , (xy).



I tried to use geometric-arithmetic mean inequality, i.e.,




(xxyy)1yx=((1x)x(1y)y)1yxx1x+y1yyx=0x+y,
but obviously this is a wrong approach ("weights" x, y, can not be negative).



Do you have any idea how to prove it? Thanks.


Answer



Note that your formula show that your inequality is equivalent to
2log|xy|log(x+y)+xlogxylogyxy
Wlog, we may suppose that x>y>0, put x=y+u.



We have

xlogx=(y+u)log(y+u)=ylogy+ylog(1+u/y)+ulogu+ulog(1+y/u)
and



log(x+y)=log(2y+u)=logu+log(1+2y/u)



Hence we have to show:
2logulogu+log(1+2y/u)+(y/u)log(1+u/y)+logu+log(1+y/u)
or



0log(1+2y/u)+(y/u)log(1+u/y)+log(1+y/u)

and this is clear.


No comments:

Post a Comment

real analysis - How to find limhrightarrow0fracsin(ha)h

How to find lim without lhopital rule? I know when I use lhopital I easy get $$ \lim_{h\rightarrow 0}...