Awesome Inequality

Algebra Level 5

( x y + y z + z x ) ( y x + z y + x z ) \large \left(\dfrac{x}{y}+\dfrac{y}{z}+\dfrac{z}{x}\right)\left(\dfrac{y}{x}+\dfrac{z}{y}+\dfrac{x}{z}\right)

Let S S be the minimum value of the above expression for positive reals x , y x,y and z z satisfying ( x y + y z + z x ) + ( y x + z y + x z ) = 8 . \left(\dfrac{x}{y}+\dfrac{y}{z}+\dfrac{z}{x}\right)+\left(\dfrac{y}{x}+\dfrac{z}{y}+\dfrac{x}{z}\right)=8\; .

Given that S S can be expressed as a b + c a\sqrt{b}+c where a , b a,b and c c are integers and b b square-free, find a + b + c a+b+c .


The answer is -24.

This section requires Javascript.
You are seeing this because something didn't load right. We suggest you, (a) try refreshing the page, (b) enabling javascript if it is disabled on your browser and, finally, (c) loading the non-javascript version of this page . We're sorry about the hassle.

2 solutions

Let P = ( x y + y z + z x ) ( y x + z y + x z ) P=\left(\dfrac{x}{y}+\dfrac{y}{z}+\dfrac{z}{x}\right)\left(\dfrac{y}{x}+\dfrac{z}{y}+\dfrac{x}{z}\right)

We have:

2 P = ( x y + y x ) 2 ( x y ) 2 ( y x ) 2 = 64 ( x 2 y 2 + 2 x y ) + ( y 2 x 2 + 2 y x ) = 48 ( x 2 y 2 + y 2 x 2 ) = 51 ( x 2 ) ( 1 x 2 ) \begin{aligned} 2P&=\left(\sum\dfrac{x}{y}+\sum\dfrac{y}{x}\right)^2-\left(\sum\dfrac{x}{y}\right)^2-\left(\sum\dfrac{y}{x}\right)^2\\ &=64-\left(\sum\dfrac{x^2}{y^2}+2\sum\dfrac{x}{y}\right)+\left(\sum\dfrac{y^2}{x^2}+2\sum\dfrac{y}{x}\right)\\ &=48-\left(\sum\dfrac{x^2}{y^2}+\sum\dfrac{y^2}{x^2}\right)\\ &=51-\left(\sum x^2\right)\left(\sum\dfrac{1}{x^2}\right) \end{aligned}

From the hypothesis, we get: ( x ) ( 1 x ) = ( x y ) ( 1 x y ) = 11 \displaystyle\left(\sum x\right)\left(\sum\dfrac{1}{x}\right)=\left(\sum xy\right)\left(\sum\dfrac{1}{xy}\right)=11

Applying the Cauchy-Schwarz inequality we have:

( 2 x y ) ( 2 1 x y ) + ( x 2 ) ( 1 x 2 ) ( 2 x y + x 2 ) ( 2 1 x y + 1 x 2 ) = ( x ) ( 1 x ) \begin{aligned} &\sqrt{\left(2\sum xy\right)\left(2\sum\dfrac{1}{xy}\right)}+\sqrt{\left(\sum x^2\right)\left(\sum\dfrac{1}{x^2}\right)}\\ \le&\sqrt{\left(2\sum xy+\sum x^2\right)\left(2\sum\dfrac{1}{xy}+\sum\dfrac{1}{x^2}\right)}\\ =&\left(\sum x\right)\left(\sum\dfrac{1}{x}\right) \end{aligned}

Hence, ( x 2 ) ( 1 x 2 ) ( 11 2 11 ) 2 \displaystyle \left(\sum x^2\right)\left(\sum\dfrac{1}{x^2}\right)\le \left(11-2\sqrt{11}\right)^2

So, P 51 ( 11 2 11 ) 2 2 = 22 11 57 \displaystyle P\ge\dfrac{51- \left(11-2\sqrt{11}\right)^2}{2}=22\sqrt{11}-57

Thus, a + b + c = 24 a+b+c=\boxed{-24} .

But is it possible to obtain this minimum value for some x,y,z? I got lower bound as 16 for x = 1 , y = ( 3 + 5 ) / 2 , z = ( 3 5 ) / 2 x = 1, y = (3 + \sqrt{5})/2, z = (3 - \sqrt{5})/2 But I couldn't find a lower bound than 16.

Siva Bathula - 5 years, 2 months ago

Log in to reply

The equality can hold when z = 1 ; y = 1 2 ( 1 + 11 + 8 2 11 ) ; x = 5 11 + 35 10 11 z=1; y=\dfrac{1}{2}\left(-1+\sqrt{11}+\sqrt{8-2\sqrt{11}}\right); x=5-\sqrt{11}+\sqrt{35-10\sqrt{11}}

Khang Nguyen Thanh - 5 years, 2 months ago

Log in to reply

Seems right. Thanks.

Siva Bathula - 5 years, 2 months ago

Please add the equality condition into your solution, and explain how to calculate it.

Great question btw :)

Calvin Lin Staff - 5 years, 2 months ago

How is that hypothesis equal to 11?

Saurabh Chaturvedi - 5 years, 2 months ago

Log in to reply

( x ) ( 1 x ) = 3 + x y + y x = 11 \displaystyle \left(\sum x\right)\left(\sum\dfrac{1}{x}\right)=3+\sum\dfrac{x}{y}+\sum\dfrac{y}{x}=11

Khang Nguyen Thanh - 5 years, 2 months ago

Log in to reply

How is (sum of x).(sum of 1/x) = (sum of xy).(sum of 1/xy)?

Saurabh Chaturvedi - 5 years, 2 months ago

Log in to reply

@Saurabh Chaturvedi There are 2 ways of seeing it:

  1. Expand all of the terms and verify that they are equal
  2. Multiply and divide creatively ( x y z xyz ).

Calvin Lin Staff - 5 years, 2 months ago

can you explain how you applied cauchy schwartz? i can't seem to see how the equation under the line "applying the cauchy-schwarz inequality we have" looks anything like the cauchy-schwartz inequality. Thanks!

Willia Chang - 5 years, 1 month ago

Log in to reply

(One version of) cauchy-schwarz inequality states that A C + B D ( A + B ) ( C + D ) \sqrt{AC} + \sqrt{BD} \leq \sqrt{(A+B)(C+D)} .

That is what he applied, where A = 2 ( x y ) A = 2( \sum xy ) .

Calvin Lin Staff - 5 years, 1 month ago

Log in to reply

oh...ic, how come there is a D on the right side of the inequality, but not one on the left? also, is the "other version" of the cauchy schwarz inequlity the vector form on the wiki? i can't see how the vector form is like the inequality you proposed. thanks!

Willia Chang - 5 years, 1 month ago

Log in to reply

@Willia Chang Sorry fixed the typo.

For the CS of 2 terms, you are most probably used to seeing ( a 2 + b 2 ) ( c 2 + d 2 ) ( a c + b d ) 2 (a^2 + b^2) ( c^2 + d^2) \geq (ac+bd)^2 .
Now, substitute a = A a = \sqrt{A} etc, and take square roots on both sides, you will get ( A + B ) ( C + D ) A C + B D \sqrt{ (A+B)(C+D) } \geq \sqrt{AC} + \sqrt{BD} .

Calvin Lin Staff - 5 years, 1 month ago
Mark Hennings
Apr 12, 2016

Write a = x + y + z a = x + y + z , b = x y + x z = y z b = xy + xz = yz and c = x y z c = xyz . Then 8 = ( x y + y z + z x ) + ( x z + z y + y x ) = x 2 y + x 2 z + x y 2 + y 2 z + x z 2 + y z 2 x y z = a b 3 c c 8 \; = \; \left(\frac{x}{y} + \frac{y}{z} + \frac{z}{x}\right) + \left(\frac{x}{z} + \frac{z}{y} + \frac{y}{x}\right) \; = \; \frac{x^2y + x^2z + xy^2 + y^2z + xz^2 + yz^2}{xyz} \; = \; \frac{ab - 3c}{c} and hence a b = 11 c ab \; = \; 11c \; and so x , y , z x,y,z are the roots of the cubic f ( X ) = X 3 a X 2 + b X 1 11 a b = 0 . f(X) \; = \; X^3 - aX^2 + bX - \tfrac{1}{11}ab \; = \; 0 \;. Playing with elementary symmetric polynomials gives Q = ( x y + y z + z x ) ( x z + z y + y x ) = 3 + ( x 2 y z + y 2 x z + z 2 x y ) + ( y z x 2 + x z y 2 + x y z 2 ) = 3 + x 3 + y 3 + z 3 c + x 3 y 3 + x 3 z 3 + y 3 z 3 c 2 = 3 + a 3 3 a b + 3 c c + b 3 3 a b c + 3 c 2 c 2 = 9 + a 3 c + b 3 c 2 6 a b c = 9 + 11 a 2 b + 121 b a 2 66 = 11 u + 121 u 1 57 \begin{array}{rcl} \mathcal{Q} & = & \left(\frac{x}{y} + \frac{y}{z} + \frac{z}{x}\right)\left(\frac{x}{z} + \frac{z}{y} + \frac{y}{x}\right) \\ & = & 3 + \left(\frac{x^2}{yz} + \frac{y^2}{xz} + \frac{z^2}{xy}\right) + \left(\frac{yz}{x^2} + \frac{xz}{y^2} + \frac{xy}{z^2}\right) \\ & = & 3 + \frac{x^3 + y^3 + z^3}{c} + \frac{x^3y^3 + x^3z^3 + y^3z^3}{c^2} \\ & = & 3 + \frac{a^3 - 3ab + 3c}{c} + \frac{b^3 - 3abc + 3c^2}{c^2} \\ & = & 9 + \frac{a^3}{c} + \frac{b^3}{c^2} - \frac{6ab}{c} \\ & = & 9 + \frac{11a^2}{b} + \frac{121b}{a^2} - 66 \\ & = & 11u + 121u^{-1} - 57 \end{array} where u = a 2 b 1 u \,=\, a^2b^{-1} .

Since Q ( u ) = 11 121 u 2 \mathcal{Q}'(u) \,=\, 11 - 121u^{-2} , we see that the minimum of Q \mathcal{Q} over all positive u u occurs when u = 11 u = \sqrt{11} , giving a value of Q = 22 11 57 \mathcal{Q} = 22\sqrt{11} - 57 .

It remains to show that these values of u u and Q \mathcal{Q} can be achieved with x , y , z x,y,z all positive. With a = b = 11 a = b = \sqrt{11} , we have the cubic f ( X ) = X 3 11 X 2 + 11 X 1 = ( X 1 ) ( X 2 + ( 1 11 ) X + 1 ) , f(X) \; = \; X^3 - \sqrt{11} X^2 + \sqrt{11} X - 1 \; = \; (X - 1)\big(X^2 + (1 - \sqrt{11})X + 1\big) \;, which has positive roots 1 1 , 1 2 [ 11 1 ± 8 2 11 ] \tfrac12\big[\sqrt{11}-1 \pm \sqrt{8 - 2\sqrt{11}}\big] . Thus this value of Q \mathcal{Q} can be achieved, and so the desired answer is 22 + 11 57 = 24 22 + 11 - 57 =\boxed{-24} .

How did you know that when u = 11 u = \sqrt{11} , then a = b = 11 a=b=\sqrt{11} must be true? Why can't other pairs of ( a , b ) (a,b) be true?

By the way, nice solution. I tried solving it via cubic discriminant and I got stuck.

Pi Han Goh - 5 years, 1 month ago

Log in to reply

I was looking for a simple solution to a 2 = b 11 a^2 = b\sqrt{11} , and found that the case I picked gives three positive real roots for f ( X ) f(X) . I am sure there are other solutions.

The other approach would be to consider the inequalities forced on a a and b b by the requirement that f ( X ) f(X) have three positive real roots, and solve those. Work those inequalities out! You need f ( X ) f(X) to have positive turning points 0 < u < v 0<u<v , and f ( u ) 0 f ( v ) f(u) \ge 0 \ge f(v) for a , b > 0 a,b > 0 .

Mark Hennings - 5 years, 1 month ago

Log in to reply

Hmmm, I can't seem to solve it by hand.


Here's my feeble approach:

f ( X ) = X 3 a X 2 + b X 1 11 a b = X 3 a x 2 + 11 a 2 X 1 a 11 f(X) = X^3 - aX^2 + bX - \dfrac1{11} ab = X^3 - ax^2 + \dfrac{\sqrt{11}}{a^2} X - \dfrac1{a\sqrt{11}} .

By cubic discriminant, ( B C ) 2 4 A C 3 4 B 3 D 27 ( A D ) 2 + 18 A B C D 0 (BC)^2 - 4AC^3 - 4B^3 D - 27(AD)^2 + 18ABCD \geq 0 , where A = 1 , B = a , C = 11 a 2 , D = 1 a 11 A = 1, B = -a, C = \dfrac{\sqrt{11}}{a^2}, D = - \dfrac1{a\sqrt{11}} . Simplifying it gives this monstrosity:

11 a 2 44 11 a 6 4 a 2 11 27 11 a 2 + 18 a 2 0 . \dfrac{11}{a^2} - \dfrac{44\sqrt{11}}{a^6} - \dfrac{4a^2}{\sqrt{11}} - \dfrac{27}{11a^2} + \dfrac{18}{a^2} \geq 0 \; .

Solving it by WolframAlpha tells us that a a satisfy the condition,

1.80726 x 1.83516 1.80726\ldots \leq x \leq 1.83516 \ldots

But this contradicts the fact that a = 11 > 1.83516 a = \sqrt{11} > 1.83516\ldots is a possible solution.

I'm so confused right now.

Either way, thanks for your response. I think your a = b = 11 a = b=\sqrt{11} is the simplest approach. I interpreted that your solution implies that it's the only pair for the solution ( a , b ) (a,b) , when in fact, it isn't.


On an unrelated topic, do you have any improvement for my solution here ? I've been told that Jensen's inequality is a viable approach, but I don't see how yet.

Pi Han Goh - 5 years, 1 month ago

Log in to reply

@Pi Han Goh We want to know the condition on a , b > 0 a,b> 0 which ensures that the equation f ( X ) = X 3 a X 2 + b X 1 11 a b = 0 f(X) \; = \; X^3 - aX^2 + bX - \tfrac{1}{11}ab \; = \; 0 has positive real roots. It turns out the the condition can be expressed solely in terms of u = a 2 b 1 u \,=\, a^2b^{-1} . The algebra proving this is slightly neater in terms of v = u 1 = a 2 b v = u^{-1} = a^{-2}b . Note that f ( X ) = X 3 a X 2 + a 2 v X 1 11 a 3 v f ( a X ) = a 3 [ X 3 X 2 + v X 1 11 v ] \begin{array}{rcl} f(X) & = & X^3 - aX^2 + a^2vX - \tfrac{1}{11}a^3v \\ f(aX) & = & a^3\Big[X^3 - X^2 + vX - \tfrac{1}{11}v\Big] \end{array} and so f ( X ) f(X) will have three positive real roots precisely when g ( X ) = X 3 X 2 + v X 1 11 v g(X) = X^3 - X^2 + vX - \tfrac{1}{11}v does. Looking for roots of g ( X ) g'(X) , we see that g ( X ) g(X) has turning points at X = λ ± X = \lambda_\pm , where λ ± = 1 3 [ 1 ± 1 3 v ] \lambda_\pm \; = \; \tfrac13\big[1\pm \sqrt{1 - 3v}\big] provided that v < 1 3 v < \tfrac13 (both of these numbers are positive), and we calculate g ( λ ± ) = 2 297 [ 36 v 11 11 ( 1 3 v ) 3 2 ] g(\lambda_\pm) \; = \; \tfrac{2}{297}\Big[36v - 11 \mp 11(1 - 3v)^{\frac32}\Big] The polynomial g ( X ) g(X) will have three positive roots precisely when g ( λ ) 0 g ( λ + ) g(\lambda_-) \ge 0 \ge g(\lambda_+) , and this will occur precisely when ( 36 v 11 ) 2 121 ( 1 3 v ) 3 121 ( 3 v 1 ) 3 + ( 36 v 11 ) 2 0 27 v ( 121 v 2 73 v + 11 ) 0 \begin{array}{rcl} (36v - 11)^2 & \le & 121(1 - 3v)^3 \\ 121(3v-1)^3 + (36v - 11)^2 & \le & 0 \\ 27v(121v^2 - 73v + 11) & \le & 0 \end{array} Thus, given that v v must be positive, we deduce that f ( X ) f(X) will have three positive roots precisely when 73 5 242 v 73 + 5 242 73 5 22 u 73 + 5 22 \begin{array}{rcccl} \tfrac{73-\sqrt{5}}{242} & \le & v & \le & \tfrac{73 + \sqrt{5}}{242} \\ \tfrac{73-\sqrt{5}}{22} & \le & u & \le & \tfrac{73 + \sqrt{5}}{22} \end{array} Note that, since 73 + 5 242 < 1 3 \tfrac{73 + \sqrt{5}}{242} < \tfrac13 , this final condition includes the previous condition v < 1 3 v < \tfrac13 for the existence of turning points.

To finish off, we note that u = 11 u=\sqrt{11} satisfies this inequality.

Mark Hennings - 5 years, 1 month ago

Log in to reply

@Mark Hennings Thank you for your detailed writeup! I've never seen this technique being used before. I wondering whether there is a name for this technique or not.

I still don't know why my cubic discriminant working fails though... Oh well...

Pi Han Goh - 5 years, 1 month ago

0 pending reports

×

Problem Loading...

Note Loading...

Set Loading...