Golden inequality

Algebra Level 5

For positive real numbers a a , b b , and c c , find the minimum value of

( a + b ) ( b + c ) ( a + b + c ) a b c \frac{(a+b)(b+c)(a+b+c)}{abc}


The answer is 11.090.

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.

4 solutions

Mark Hennings
Nov 10, 2020

Since the expression ( a + b ) ( b + c ) ( a + b + c ) a b c \frac{(a+b)(b+c)(a+b+c)}{abc} is homogeneous in a , b , c a,b,c , we can assume that a + b + c = 1 a+b+c = 1 . Thus we want to minimize ( 1 a ) ( 1 c ) a c ( 1 a c ) = 1 a c + 1 1 a c = F ( a , c ) \frac{(1-a)(1-c)}{ac(1-a-c)} \; = \; \frac{1}{ac} + \frac{1}{1-a-c} \; = \; F(a,c) over a , c > 0 a,c > 0 , a + c < 1 a+c < 1 . Since 1 2 ( a + c ) a c \tfrac12(a+c) \ge \sqrt{ac} . we deduce that F ( a , c ) F ( 1 2 ( a + c ) , 1 2 ( a + c ) ) F(a,c) \ge F\big(\tfrac12(a+c),\tfrac12(a+c)\big) and hence we will achieve the minimum value of F ( a , c ) F(a,c) when a = c a=c . Thus we want to minimize G ( a ) = F ( a , a ) = 1 a 2 + 1 1 2 a 0 < a < 1 2 G(a) \; = \; F(a,a) \; = \; \frac{1}{a^2} + \frac{1}{1-2a} \hspace{2cm} 0 < a < \tfrac12 Since G ( a ) = 2 ( 1 2 a ) 2 2 a 3 G'(a) = \frac{2}{(1-2a)^2} - \frac{2}{a^3} , we see that the minimum value of G G occurs when 0 = a 3 ( 2 a 1 ) 2 = a 3 4 a 2 + 4 a 1 = ( a 1 ) ( a 2 3 a + 1 ) 0 = a^3 - (2a-1)^2 = a^3 - 4a^2 + 4a - 1 = (a-1)(a^2 - 3a + 1) and hence when a = 1 2 ( 3 5 ) a = \tfrac12(3-\sqrt{5}) , at which point G ( a ) = 1 2 ( 11 + 5 5 ) G(a) = \boxed{\tfrac12(11+5\sqrt{5})} .

Your solution is similar to mine. Brilliant!

Isaac YIU Math Studio - 7 months ago

Beautiful usage on F ( a , c ) F ( 1 2 ( a + c ) , 1 2 ( a + c ) ) F(a,c) \geqslant F(\tfrac12(a+c), \tfrac12(a+c)) .

I did get F ( a , c ) = 1 a c + 1 1 a c F(a,c) = \frac1{ac} + \frac1{1-a-c} as well. But here's where I failed spectacularly:

Then using AM-HM on the pairs on the numbers, we F ( a , c ) = 1 a c + 1 1 a c = 2 A M ( 1 a c , 1 1 a c ) 2 H M ( 1 a c , 1 1 a c ) = 4 a c + ( 1 a c ) F(a,c) = \frac1{ac} + \frac1{1-a-c} = 2 \cdot AM \left(\frac1{ac}, \frac1{1-a-c} \right) \geqslant 2 \cdot HM\left(\frac1{ac}, \frac1{1-a-c} \right) = \frac4{ac + (1-a-c)} or F ( a , c ) 4 ( a 1 ) ( c 1 ) F(a,c) \geqslant \frac4{(a-1)(c-1)} so we want to maximize H ( a , c ) : = ( a 1 ) ( c 1 ) H(a,c) := (a-1)(c-1) where a , c a,c are both in the interval ( 0 , 1 ) (0,1) . And it's trivial that max ( H ( a , c ) ) \max( H(a,c) ) occurs when a = c = 1 2 a=c=\frac12 , but this means that b = 0 b=0 . Oh no!!! Where did I go wrong?!

Pi Han Goh - 7 months ago

Log in to reply

Another issue: "And it's trivial that max ( H ( a , c ) ) \max( H(a,c) ) occurs when a = c = 1 2 a = c= \frac12 ."

Actually, the closer a a and c c are to zero, the closer H ( a , c ) H(a, c) will be to 1 1 . eg. H ( 0.1 , 0.1 ) = 0.81 H( 0.1, 0.1 ) = 0.81 , which is greater than H ( 1 2 , 1 2 ) = 1 4 H( \frac12, \frac12 ) = \frac14 . So H H does not have a true maximum in the open interval. (But even if it did, the minimum of F F may occur at a different point, as Mark said.)

Matthew Feig - 7 months ago

Log in to reply

Yup, I just got that after Mark's explanation. I've made many blunders everywhere when I tried to use a no-calculus approach.

Pi Han Goh - 7 months ago

Log in to reply

@Pi Han Goh Yes, so far everyone has turned it over to calculus at some point. Perhaps after eliminating b b or arguing that a = c a=c .

Matthew Feig - 7 months ago

Log in to reply

@Matthew Feig I tried using the substitution b 5 + 1 2 b b \leftrightarrow \tfrac{\sqrt5 + 1}2 b so that the minimum can occur when a = b = c a=b=c , but no luck so far...

Pi Han Goh - 7 months ago

While the AM/HM inequality is true, it is not sharp, since the actual minimum value occurs when a = c a=c , but not when a c = 1 a c ac = 1-a-c ...

Mark Hennings - 7 months ago

Log in to reply

Superb. I spent an hour on this but I couldn't figure out why. It's amazing that I couldn't understand why I messed up on something so elementary.

Pi Han Goh - 7 months ago

How a+b+c=1? Can we assume any finite number?

Monu Kumar - 4 months, 2 weeks ago

Log in to reply

Yes, we can scale a , b , c a,b,c by any constant. In this case it helped to have a + b + c = 1 a+b+c=1 . Sometimes we might want a b c = 1 abc=1 , or a 2 + b 2 + c 2 = 1 a^2+b^2+c^2=1 or even a b + a c + b c = 1 ab+ac+bc=1 - it all depends on what we are trying to establish. The bottom line is that, when we are looking at a homogeneous expression, we can perform a scaling, reducing the degrees of freedom by one.

Mark Hennings - 4 months, 2 weeks ago

Log in to reply

Of course, the rescaling equation doesn't have to treat the variables symmetrically. b = 1 b = 1 is a pretty good option here.

Matthew Feig - 4 months, 2 weeks ago

Got it thanks

Monu Kumar - 4 months, 2 weeks ago

I also used a calculus approach. Let X = ( a + b ) ( b + c ) ( a + b + c ) a b c X = \dfrac {(a+b)(b+c)(a+b+c)}{abc} . Then consider the value of a a in respect to b b and c c , when X X is minimum.

X a = a ( ( b + c ) ( a + b + c ) + ( a + b ) ( b + c ) ( a + b ) ( b + c ) ( a + b + c ) a 2 b c \frac {\partial X}{\partial a} = \frac {a((b+c)(a+b+c)+(a+b)(b+c) - (a+b)(b+c)(a+b+c)}{a^2bc}

Putting X a = 0 \dfrac {\partial X}{\partial a} = 0 , we have a 2 = b 2 + b c a^2 = b^2 + bc . Similarly, X c = 0 \dfrac {\partial X}{\partial c} = 0 , c 2 = b 2 + a b \implies c^2 = b^2 + ab . Since a a and c c are interchangeable, we can assume a = c a=c when X X is minimum. Then

a 2 a b b 2 = 0 a = b + b 2 + 4 b 2 2 for a > 0 = 1 + 5 2 b = φ b where φ denotes the golden ratio. \begin{aligned} a^2 - ab - b^2 & = 0 \\ \implies a & = \frac {b+\sqrt{b^2 + 4b^2}}2 & \small \blue{\text{for }a > 0} \\ & = \frac {1+\sqrt 5}2 b = \varphi b & \small \blue{\text{where }\varphi \text{ denotes the golden ratio.}} \end{aligned}

Then minimum value of X X :

min ( X ) = ( a + b ) ( b + c ) ( a + b + c ) a b c = ( φ + 1 ) 2 b 2 ( 2 φ + 1 ) b φ 2 b 3 = φ 4 φ 3 φ 2 = φ 5 = 5 φ + 3 11.090 \begin{aligned} \min (X) & = \frac {(a+b)(b+c)(a+b+c)}{abc} \\ & = \frac {(\varphi+1)^2 b^2 (2\varphi +1)b}{\varphi^2b^3} \\ & = \frac {\varphi^4 \varphi^3}{\varphi^2} \\ & = \varphi^5 = 5\varphi + 3 \approx \boxed{11.090} \end{aligned}

Looks good!

Matthew Feig - 7 months ago

Log in to reply

Thanks. I will change it.

Chew-Seong Cheong - 7 months ago

I solved this the same way. What's interesting is if you try to divide the numerator and denominator of the objective function by a 3 a^3 to reduce the number of variables by 1, then (using the calculus approach) the equations because become harder to solve (but still doable).

James Wilson - 5 months ago

Log in to reply

Interesting

Chew-Seong Cheong - 5 months ago

Actually, I took a closer look at your solution, and my way required more work. But I think it might've been better justified.

James Wilson - 4 months, 4 weeks ago
Pi Han Goh
Nov 9, 2020

Here's a terrible solution. Because I have to resort to calculus, which I don't think is the intended approach.


Without the loss of generality, let a + b + c = 1 a + b + c = 1 , then the expression in question can be reduced to a 2-variable expression: f ( a , c ) : = ( 1 c ) ( 1 a ) a c ( 1 a c ) f(a,c) := \dfrac{(1-c)(1-a)}{ac(1-a-c)} where a a and c c are both in the interval ( 0 , 1 ) (0,1) . Near the boundary points, f ( a , c ) f(a,c) diverges to infinity, so f ( a , c ) f(a,c) has no maximum point.

At the minimum point, the partial derivatives of f ( a , c ) f(a,c) must be 0, { δ δ a f ( a , c ) = 0 0 δ δ c f ( a , c ) = 0 { ( c 1 ) ( ( a 1 ) 2 c ) a 2 c ( a c + 1 ) 2 = 0 0 ( a 1 ) ( ( a 1 ) 2 c ) a c 2 ( a c + 1 ) 2 = 0 { ( a 1 ) 2 c = 0 ( c 1 ) 2 a = 0 \begin{cases} \dfrac{\delta }{\delta a} f(a,c) = 0 \\ \phantom0 \\ \dfrac{\delta }{\delta c} f(a,c) = 0 \end{cases} \quad\Rightarrow \quad \begin{cases} \dfrac{(c-1)((a-1)^2 - c)}{a^2c (-a-c+1)^2} = 0 \\ \phantom 0 \\ \dfrac{(a-1)((a-1)^2 - c)}{ac^2 (-a-c+1)^2} = 0 \end{cases} \quad \implies \quad \begin{cases} (a-1)^2 - c = 0 \\ (c-1)^2 - a = 0 \end{cases} Solving the system of equations above gives a = c = 3 5 2 min [ f ( a , c ) ] = 11 + 5 5 2 11.0901699 a = c = \dfrac{3-\sqrt5}2 \quad\implies \quad \min[ f(a,c) ] = \boxed{\dfrac{11+5\sqrt5}2} \approx 11.0901699 Hence, expression in question has a minimum value of 11 + 5 5 2 , \frac{11+5\sqrt5}2 , when ( a , b , c ) (a,b,c) take the ratio a : b : c = 2 : ( 5 1 ) : 2. a:b:c = 2 : (\sqrt5 - 1) : 2 .


I wonder if there's a way to prove that the minimum value of f ( a , c ) f(a,c) is 11 + 5 5 2 \frac{11+5\sqrt5}2 without resorting to calculus. That is,

Prove that 2 ( 1 c ) ( 1 a ) ( 11 + 5 5 ) a c ( 1 a c ) 2(1-c)(1-a) -(11 + 5\sqrt5) \cdot ac \cdot (1-a-c) is strictly non-negative. And it equals 0 when a = c = 3 5 2 a=c= \frac{3-\sqrt5}2 .

I tried the substitution ( a , c ) ( 1 a , 1 c ) (a,c) \leftrightarrow (1-a,1-c) but I'm still stuck.

I have cut it down to one-dimensional calculus. Doubtless some devious use of the AM/GM inquality could finish it off, but never mind!

Mark Hennings - 7 months ago

Admittedly this is not a very good solution but it spared me a lot of tedious algebra and calculus. I made extensive use of the graphing calculator Desmos to solve this problem.

Since we have three variables that can change, we can try to find when the value of ( a + b ) ( b + c ) ( a + b + c ) a b c \frac{(a + b)(b + c)(a + b + c)}{abc} is minimized if we only allow one variable at a time to vary. Let c = x c = x and let f ( x ) = ( a + b ) ( b + x ) ( a + b + x ) a b x f(x) = \frac{(a + b)(b + x)(a + b + x)}{abx} . We can expand out f ( x ) f(x) as follows:

f ( x ) = ( a + b ) a b ( b + x ) ( a + b + x ) x \displaystyle f(x) = \frac{\left(a+b\right)}{ab}\frac{\left(b+x\right)\left(a+b+x\right)}{x} f ( x ) = ( a + b ) a b a b + b 2 + b x + a x + b x + x 2 x \displaystyle f(x) = \frac{\left(a+b\right)}{ab}\frac{ab+b^{2}+bx+ax+bx+x^{2}}{x} f ( x ) = ( a + b ) 2 a 1 x + ( a + b ) a b ( a + 2 b + x ) \displaystyle f(x) = \frac{\left(a+b\right)^{2}}{a}\frac{1}{x}+\frac{\left(a+b\right)}{ab}\left(a+2b+x\right)

Taking the derivative of f ( x ) f(x) and setting find the x x such that f ( x ) = 0 f'(x) = 0 gives the equation ( a + b ) 2 a 1 x 2 + ( a + b ) a b = 0 \displaystyle -\frac{\left(a+b\right)^{2}}{a}\frac{1}{x^{2}}+\frac{\left(a+b\right)}{ab}=0 or x = ( a + b ) b x=\sqrt{\left(a+b\right)b} .

Here is where desmos comes in. If we let c = ( a + b ) b c = \sqrt{\left(a+b\right)b} , we can graph g ( x ) = ( a + x ) ( x + c ) ( a + x + c ) a x c g(x) = \frac{(a + x)(x + c)(a + x + c)}{axc} against x x and vary the value of a a in the interval ( 0 , ) (0, \infty) . It just so happens that the minimum of this function always occurs at the same y-value independent of our choice for a a (so long as a a is positive and real). The minimum of g ( x ) g(x) is approximately 11.09017 \boxed{11.09017} .

0 pending reports

×

Problem Loading...

Note Loading...

Set Loading...