Callous Constraint

Algebra Level 5

Positive reals x x , y y , and z z are such that x 3 + y 3 + z 3 = ( 3 + 1 666 ) x y z \ x^3 + y^3 + z^3 = \left(3 + \dfrac{1}{666}\right)xyz .

Find the smallest integer value of k k satisfying the following inequality:

x y + y z + z x 3 + 1 k . \large \dfrac{x}{y}+\dfrac{y}{z}+\dfrac{z}{x} \ge 3+\dfrac{1}{k}.


The answer is 2016.

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.

1 solution

Mark Hennings
Aug 24, 2016

Relevant wiki: Classical Inequalities - Problem Solving - Advanced

Since the constraint a 3 + b 3 + c 3 = ( 3 + 1 666 ) a b c a^3 + b^3 + c^3 = \big(3 + \tfrac{1}{666}\big)abc and the target function a b + b c + c a \tfrac{a}{b} + \tfrac{b}{c} + \tfrac{c}{a} are both homogeneous (of degrees 3 3 and 0 0 respectively), we can apply the additional condition a + b + c = 3 a+b+c = 3 without loss of generality.

Suppose now that a , b , c a,b,c are positive reals such that a 3 + b 3 + c 3 = ( 3 + 27 α ) u a + b + c = 3 a b c = u a^3 + b^3 + c^3 \; = \; (3 + 27\alpha)u \qquad \qquad a + b + c \; = \; 3 \qquad \qquad abc \; = \; u for positive constants α , u \alpha,u . Then ( 3 + 27 α ) u = ( a + b + c ) 3 3 ( a 2 b + a 2 c + a b 2 + b 2 c + a c 2 + b c 2 ) 6 a b c = 27 3 ( a + b + c ) ( a b + a c + b c ) + 3 a b c (3 + 27\alpha)u \; = \; (a+b+c)^3 - 3(a^2b + a^2c + ab^2 + b^2c + ac^2 + bc^2) - 6abc \; = \; 27 - 3(a+b+c)(ab+ac+bc) + 3abc so that a b + a c + b c = 3 ( 1 α u ) ab + ac + bc \; = \; 3(1 - \alpha u) Since a , b , c a,b,c are positive this means that 0 < α u < 1 0 < \alpha u < 1 . Now a , b , c a,b,c are the roots of the cubic equation 0 = f ( X ) = X 3 3 X 2 + 3 ( 1 α u ) X u = ( X 1 ) 3 3 α u ( X 1 ) + 1 ( 1 + 3 α ) u 0 \; = \; f(X) \; = \; X^3 - 3X^2 + 3(1 - \alpha u)X - u \; = \; (X - 1)^3 - 3\alpha u(X - 1) + 1 - (1 + 3\alpha)u Since f ( X ) f(X) has three positive real roots, its derivative has two positive real roots. Since f ( X ) = 3 [ ( X 1 ) 2 α u ] f'(X) \; = \; 3\big[(X - 1)^2 - \alpha u\big] we see that the turning points of f f occur at λ ± = 1 ± α u \lambda_\pm = 1 \pm \sqrt{\alpha u} , which are indeed both positive, since 0 < α u < 1 0 < \alpha u < 1 . Note that f ( x ) < 0 f(x) < 0 for all x 0 x\le 0 , and hence that all real roots of f ( X ) f(X) are positive. Since f ( λ ± ) = 2 α u α u + 1 ( 1 + 3 α ) u f(\lambda_\pm) \; = \; \mp2\alpha u \sqrt{\alpha u} + 1 - (1 + 3\alpha)u and, in order that f ( X ) f(X) have three positive real roots, we require that f ( λ ) 0 f ( λ + ) f(\lambda_-) \ge 0 \ge f(\lambda_+) , it follows that ( 1 + 3 α ) u 1 2 α u α u \big|(1 + 3\alpha)u - 1\big| \; \le \; 2\alpha u \sqrt{\alpha u} In other words, we require that 4 α 3 u 3 ( ( 1 + 3 α ) u 1 ) 2 0 4\alpha^3 u^3 - \big((1 + 3\alpha)u-1\big)^2 \; \ge \; 0 In our case of α = 1 27 × 666 \alpha = \tfrac{1}{27\times666} , numerical calculations show that we must have 0.99983236525 u 0.99983402340 0.99983236525 \le u \le 0.99983402340 .

Consider the two quantities p = a b + b c + c a q = a c + c b + b a p \; = \; \tfrac{a}{b} + \tfrac{b}{c} + \tfrac{c}{a} \qquad \qquad q \; = \; \tfrac{a}{c} + \tfrac{c}{b} + \tfrac{b}{a} noting that the difference between p p and q q is simply a matter of the labelling of the roots of f ( X ) f(X) . Playing with Vieta's formulae, we can show that p + q = 9 u 3 ( 3 α + 1 ) p q = 9 + 54 α 27 u + 27 ( 1 α u ) 3 u 2 p+q \; = \; \tfrac{9}{u} - 3(3\alpha+1) \qquad \qquad pq \; = \; 9 + 54\alpha - \tfrac{27}{u} + \tfrac{27(1-\alpha u)^3}{u^2} and so p , q p,q are the roots of the quadratic equation h ( X ) = X 2 ( 9 u 3 ( 3 α + 1 ) ) X + 9 + 54 α 27 u + 27 ( 1 α u ) 3 u 2 = 0 h(X) \; = \; X^2 - \big( \tfrac{9}{u} - 3(3\alpha+1) \big)X + 9 + 54\alpha - \tfrac{27}{u} + \tfrac{27(1-\alpha u)^3}{u^2} \; = \; 0 Thus, for any particular values of α , u \alpha,u , the smallest possible value of a b + b c + c a \tfrac{a}{b} + \tfrac{b}{c} + \tfrac{c}{a} is the smaller of the two zeros of h ( X ) h(X) , call it L α ( u ) L_\alpha(u) . The explicit formula for L α ( u ) L_\alpha(u) is L α ( u ) = 3 2 u [ 3 ( 3 α + 1 ) u 3 [ 4 α 3 u 3 ( ( 3 α + 1 ) u 1 ) 2 ] ] L_\alpha(u) \; = \; \tfrac{3}{2u}\left[3 - (3\alpha+1)u - \sqrt{3\left[4\alpha^3u^3 - ((3\alpha+1)u-1)^2\right]}\right] which is reassuring, since the term inside the square root only gives a real value for L α ( u ) L_\alpha(u) when 4 α 3 u 3 ( ( 3 α + 1 ) u 1 ) 2 4\alpha^3u^3 \ge ((3\alpha+1)u-1)^2 , which is the condition we have already determined as being necessary.

In our particular case of α = 1 27 × 666 \alpha = \tfrac{1}{27 \times 666} , we want to minimize L α ( u ) L_\alpha(u) over the interval 0.99983236525 u 0.99983402340 0.99983236525 \le u \le 0.99983402340 . Numerical calculations (solving for the exact value of u u where L α ( u ) L_\alpha(u) has a turning point requires solving an order 6 6 polynomial) show that this occurs at u = 0.99983391233 u = 0.99983391233 , taking the least value 3.00049619108 = 3 + 1 2015.35 3.00049619108 = 3 + \tfrac{1}{2015.35} . Thus we have shown that if a , b , c a,b,c are positive real numbers such that a 3 + b 3 + c 3 = ( 3 + 1 666 ) a b c a^3 + b^3 + c^3 = \big(3 + \tfrac{1}{666}\big)abc , then a b + b c + c a 3 + 1 2015.35 \tfrac{a}{b} + \tfrac{b}{c} + \tfrac{c}{a} \; \ge \; 3 + \tfrac{1}{2015.35} Thus the smallest possible integer k k is 2016 \boxed{2016} .

I really hope that someone can come up with a neater solution than this!

Really a well written solution.

And thanks for writing so much. It would have taken so much time.

Priyanshu Mishra - 4 years, 9 months ago

Can you reconsider the solution using rearrangement inequality? I came up with some sort of k=1336 i.e (666*2)

Ariijit Dey - 3 years, 7 months ago

Log in to reply

which was also reassuring for that moment I must say :-)

Ariijit Dey - 3 years, 7 months ago

Log in to reply

Publish your solution, and we will see.

Mark Hennings - 3 years, 7 months ago

Log in to reply

@Mark Hennings I must say that I didn't come up with a tangible solution but will you help modifying it to a neater and (more precisely) an affordable solution than the above (you just posted) [Although it was an ingenious piece of work I must say;It requires unfathomable patience to numerically calculate all the values and also the process was FANTASTIC!!!]

Ariijit Dey - 3 years, 7 months ago

Log in to reply

@Ariijit Dey Well, since my answer does not have a particularly simple answer (involving the root of an order six polynomial) it is not likely that a really simple proof is to be found...

Mark Hennings - 3 years, 7 months ago

Log in to reply

@Mark Hennings By the way what do you do? And how do you publish such ingenious pieces of work?

Ariijit Dey - 3 years, 6 months ago

0 pending reports

×

Problem Loading...

Note Loading...

Set Loading...