No Additional Condition Needed

Algebra Level 5

( a + b + c + 1 ) 2 ( a 2 + 1 ) ( b 2 + 1 ) ( c 2 + 1 ) \large \dfrac{(a+b+c+1)^2}{(a^2+1)(b^2+1)(c^2+1)}

Let a , b , c > 0 a, b, c>0 . Find the maximum value of the expression above.


The answer is 3.2.

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.

3 solutions

Otto Bretscher
Sep 28, 2015

The partial derivatives are zero when a 2 + 1 = a ( a + b + c + 1 ) a^2+1=a(a+b+c+1) , b 2 + 1 = b ( a + b + c + 1 ) b^2+1=b(a+b+c+1) and c 2 + 1 = c ( a + b + c + 1 ) c^2+1=c(a+b+c+1) . Equivalently, it is required that a + 1 a = b + 1 b = c + 1 c = a + b + c + 1 a+\frac{1}{a}=b+\frac{1}{b}=c+\frac{1}{c}=a+b+c+1 Up to a permutation of the variables, we have either a = b = c a=b=c or a = b = 1 c a=b=\frac{1}{c} . The second case implies a = 1 a=-1 , which we reject. In the first case we find that a = b = c = 1 2 a=b=c=\frac{1}{2} (again, we reject a = 1 a=-1 ) and the maximum value will be 16 5 = 3.2 \frac{16}{5}=\boxed{3.2} .

Don't you need to prove that it's maximum ?

Pi Han Goh - 5 years, 8 months ago

Log in to reply

But of course I do... Responding to your inquiries is almost becoming a full time job, Pi Han Goh ;) give me some time please. I know how to do it (and so do you, I'm sure), but writing these things down properly is a bit tricky.

Otto Bretscher - 5 years, 8 months ago

Let me give it a try and see whether my rough outline will convince you.

Certainly our continuous function will attain a (global) maximum on a (compact) cube like C = [ 0 , 2 ] 3 C=[0,2]^3 . It is easy to see that f / a < 0 \partial{f}/\partial{a}<0 for a > 1 a>1 , and likewise for the other variables, so that for every point ( a , b , c ) (a,b,c) in space there is a point in the cube C C with a higher value. The maximum attained in C C is in fact the maximum overall.

We need to check that the maximum is not attained on the boundary, where one or two of the variables are 0. Using another Lagrange multiplier, we see that the maximum on the boundary is 1 2 ( 3 + 5 ) < 3.2 , \frac{1}{2}(3+\sqrt{5})<3.2, so that the maximum is indeed attained at the only critical point ( 0.5 , 0.5 , 0.5 ) (0.5,0.5,0.5) as claimed.

Otto Bretscher - 5 years, 8 months ago

Log in to reply

Can't you just do the second derivative test?

Pi Han Goh - 5 years, 8 months ago

Log in to reply

@Pi Han Goh That only gives a local max.

Otto Bretscher - 5 years, 8 months ago

Log in to reply

@Otto Bretscher Awhhh man. Alright thanks. Let me peruse your comments.

(And again, I don't mean to ruin your fun but I doubt the author posted this question with a calculus solution in mind)

Pi Han Goh - 5 years, 8 months ago

Yup, I too cheated with Lagrange Multipliers, Sir Otto....but it's still a tried -n- true method for these multivariable optimization problems. Good solution!

tom engelsman - 1 year ago
Ajay Saju
Oct 1, 2015

I did It like this. Don't know if its flawed. tell me- let a=tan x, b=tan y,c=tan z so x,y,z>0, all taking only principle values. therefore, LHS= ( t a n x + t a n y + t a n z + 1 ) 2 ) / ( ( s e c x ) 2 ( s e c y ) 2 ( s e c z ) 2 ) ( tan x + tan y + tan z + 1)^{2})/((sec x)^{2}(sec y)^{2}(sec z)^{2})

= ( ( t a n x + t a n y + t a n z + 1 ) c o s x × c o s y × c o s z ) 2 ((tan x +tan y +tan z +1) cos x\times cos y\times cos z)^{2}

= ( s i n x × c o s y × c o s z + c o s x × s i n y × c o s z + c o s x × c o s y × s i n z + c o s x × c o s y × c o s z ) 2 (sin x\times cosy\times cosz+cosx \times siny\times cosz+cosx\times cosy\times sinz+cosx\times cos y\times cosz)^{2}

= ( s i n ( x + y + z ) + s i n x × s i n y × s i n z + c o s x × c o s y × c o s z ) 2 (sin(x+y+z) + sinx \times siny \times sinz + cosx \times cosy \times cosz)^{2}

Conditional on x + y + z = θ x + y + z = \theta a constant, because the second derivative of ln sin x \ln \sin x and ln cos x \ln \cos x are negative, it tells us that sin x \prod \sin x and cos x \prod \cos x are maximized when x = y = z x = y = z . Hence, we want to find the maximum of sin 3 x + sin 3 x + cos 3 x \sin 3x + \sin ^3 x + \cos ^3 x

It remains to verify (my approach was to use calculus) that the maximum occurs at x = 2 π 2 tan 1 ( 2 5 ) x = 2 \pi - 2 \tan^{-1} ( 2 - \sqrt{5} ) , and has the value of 4 5 \frac{4}{ \sqrt{5} } . Thus, the max of the expression is 16 5 \frac{16}{5} .

why must x=y=z=pi/6?

Pi Han Goh - 5 years, 8 months ago

That's a nice substitution effect that you have used.

However, I disagree with "the above expression is maximium clearly ..." It is not clear why we cannot reduce sin ( x + y + z ) \sin (x + y + z ) slightly, in order to increase sin x sin y sin z + cos x cos y cos z \sin x \sin y \sin z + \cos x \cos y \cos z . Note that max sin x sin y sin z + cos x cos y cos z = 1 \max \sin x \sin y \sin z + \cos x \cos y \cos z = 1 which occurs when ( x , y , z ) = ( 0 , 0 , 0 ) , ( π 2 , π 2 , π 2 ) (x, y, z) = (0, 0, 0) , ( \frac{\pi}{2}, \frac{\pi}{2}, \frac{\pi}{2}) .

Calvin Lin Staff - 5 years, 8 months ago

Log in to reply

So here's how I would fix the gap that I mentioned above.

Conditional on x + y + z = θ x + y + z = \theta a constant, because the second derivative of ln sin x \ln \sin x and ln cos x \ln \cos x are negative, it tells us that sin x \prod \sin x and cos x \prod \cos x are maximized when x = y = z x = y = z . Hence, we want to find the maximum of sin 3 x + sin 3 x + cos 3 x \sin 3x + \sin ^3 x + \cos ^3 x

It remains to verify (my approach was to use calculus) that the maximum occurs at x = 2 π 2 tan 1 ( 2 5 ) x = 2 \pi - 2 \tan^{-1} ( 2 - \sqrt{5} ) , and has the value of 4 5 \frac{4}{ \sqrt{5} } . Thus, the max of the expression is 16 5 \frac{16}{5} .

Calvin Lin Staff - 5 years, 8 months ago

Log in to reply

oh! thanks bro for the advice.

Ajay Saju - 5 years, 8 months ago
Rui de Sousa
Sep 28, 2015

Suppose a = b = c. Then f ( a ) = ( 3 a + 1 ) 2 / ( a 2 + 1 ) 3 f (a) = (3a+1)^{2}/(a^{2}+1)^{3} . And d ( f ( a ) ) / ( d a ) = [ 6 ( 3 a + 1 ) ( a 2 + 1 ) 3 6 ( 3 a + 1 ) 2 × ( a 2 + 1 ) 2 × a ] / ( a 2 + 1 ) 6 d (f (a))/(da) = [6(3a+1)(a^{2}+1)^{3}-6(3a+1)^{2} \times (a^{2}+1)^{2} \times a]/(a^{2}+1)^{6} = 6 [ ( 3 a + 1 ) ( a 2 + 1 ) ( 3 a + 1 ) 2 × a ] / ( a 2 + 1 ) 4 = 6 [(3a+1)(a^{2}+1) - (3a+1)^{2} \times a]/(a^{2}+1)^{4} . When we make d ( f ( a ) ) / ( d a ) = 0 d (f (a))/(da) = 0 then we get a 2 + 1 ( 3 a + 1 ) × a = 0 a^2+1 -(3a+1) \times a = 0 2 a 2 + a 1 = 0 2a^{2}+a-1=0 So a=1/2. Finally, f ( a = 1 / 2 ) = 16 / 5 f (a=1/2)=16/5 .

But why can we suppose that a = b = c a=b=c ? ;)

Otto Bretscher - 5 years, 8 months ago

0 pending reports

×

Problem Loading...

Note Loading...

Set Loading...