Looks so pretty, but it is too pretty

Algebra Level 4

Consider all positive reals a a and b b such that

a b ( a + b ) = 2000. ab (a+b) = 2000.

What is the minimum value of

1 a + 1 b + 1 a + b ? \frac{1}{a} + \frac{1}{b} + \frac{1}{a+b} ?

3 20 \frac{3}{20} 1 4 \frac{1}{4} 3 4 3 20 \frac{ 3 \sqrt[3]{4}}{20} 2 3 10 \frac{ \sqrt[3]{2} } { 10}

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.

11 solutions

Calvin Lin Staff
Aug 14, 2014

As a discussion, first consider why the "obvious" answers are wrong.
Yes, the choices were specially selected.


Wrong Approach 1:

Applying AM-GM, we get that

1 a + 1 b + 1 a + b 3 1 a b ( a + b ) 3 = 3 1 2000 3 = 3 4 3 20 . \frac{1}{a} + \frac{1}{b} + \frac{1}{a+b} \geq 3 \sqrt[3]{ \frac{1}{ab(a+b)} } = 3 \sqrt[3]{ \frac{1}{2000}} = \frac{ 3 \sqrt[3]{4} } { 20 }.

What is wrong with this approach?


Wrong Approach 2:

1 a + 1 b + 1 a + b = a + b a b + 1 a + b = ( a + b ) 2 2000 + 1 a + b . \frac{1}{a} + \frac{1}{b} + \frac{1}{a+b} = \frac{ a+b} { ab} + \frac{1}{a+b} = \frac{ (a+b)^2} { 2000} + \frac{1}{a+b}.

Letting x = a + b x = a+b , we want to minimize x 2 2000 + 1 x \frac{ x^2}{2000 } + \frac{1}{x} .
Using Calculus, or AM-GM, the minimum occurs when x = 10 x = 10 and has a value of 3 20 \frac{3}{20} .

What is wrong with this approach?


Correct solution:

Let a + b = x a + b = x , and then a b = 2000 x ab = \frac{2000}{x} . a a and b b are the real roots of the quadratic equation

X 2 x X + 2000 x = 0 X^2 - xX + \frac{2000}{x} = 0

This has a non-negative discriminant, so x 2 4 × 2000 x 0 x^2 - \frac{4 \times 2000} { x} \geq 0 . This gives us x 3 8000 x^3 \geq 8000 , or that x 20 x \geq 20 .

Now, we want to minimize f ( x ) = x 2 2000 + 1 x f(x) = \frac{ x ^2 } { 2000} + \frac{1}{x} SUBJECT to x 20 x \geq 20 . Using calculus, it is clear that for x 20 x \geq 20 , we have f ( x ) > 0 f'(x) > 0 . Hence, the minimum occurs when x = 20 x = 20 .

In this scenario, we have a + b = 20 , a b = 100 a + b = 20, ab = 100 . This is satisfied when a = 10 , b = 10 a = 10, b = 10 . Hence, the minimum can be achieved. This minimum is 1 10 + 1 10 + 1 20 = 5 20 = 1 4 \frac{1}{10} + \frac{1}{10} + \frac{1}{20} = \frac{5}{20} = \frac{1}{4} .

Wrong approach 1 is no good because that lower bound is never attained (equality cannot hold unless a = b = 0 a = b = 0 ).

Wrong approach 2 is pretty, but the problem seems to be that a + b = 10 a+b = 10 doesn't correspond to a real solution, since in this case a b = 200 ab = 200 , and this is impossible for real a , b a,b .

If you restrict the domain to values of x x that correspond to real solutions, you're looking at x 20 x \ge 20 (easy exercise), and the function is increasing on that domain by calculus (or whatever), so x = 20 x = 20 is where the minimum is attained.

That is a much nicer way to solve the problem than cranking out Lagrange multipliers, which is what I did.

Patrick Corn - 6 years, 10 months ago

Log in to reply

Yes, the main issue is that the solutions did not check that the minimum can indeed be achieved. All that they found was a lower bound. They did not show that is it the greatest lower bound.

Calvin Lin Staff - 6 years, 10 months ago

I used wring approach number 2. T.T

Gil Deon Basa - 6 years, 10 months ago

YES! Sorry I'm just happy to have gotten this right. It's probably because I don't like AM-GM. Which is pretty funny. :P

Finn Hulse - 6 years, 10 months ago

I didn't understand why the first approach is wrong. Please help.

Aaron Jerry Ninan - 5 years ago

Log in to reply

It is because equality will only hold if a=b=0. which can't be true.

Shishir Shahi - 3 years, 10 months ago

Amazing!, just trying to get this. I used the AM-GM inequality

Carlos David Nexans - 6 years, 10 months ago

it can be solved more easily..

nibedan mukherjee - 6 years, 10 months ago

Log in to reply

Certainly. Jon posed a straightforward solution.

My original intention was to get a discussion going, about what some of the "obvious choices" were false. I then tried to create a solution which was similar to the approaches used before, to show how one can "fix" them.

Calvin Lin Staff - 6 years, 10 months ago

really pretty got it right

Gautam Sharma - 6 years, 10 months ago

if you take a function f(x)=(x^2/2000) +(1/x). then using calculus if we differentiate f(x) w.r.t x then we get (x/1000)+ log(x) . it will not be zero for x=10 or x=20 or x=1. it may be zero only when x<1. so how can you say that at x=20 or x=10 will be an extreme point.

Varun Narayan - 6 years, 9 months ago

Log in to reply

The derivative of 1/x is not log(x). :-)

Durward McDonell - 5 years, 3 months ago

Log in to reply

It's -1/x^2

Anubhav Mahapatra - 3 years, 7 months ago

integrate cosx/x from npi to (n+1)pi

Neeraj Gupta - 6 years, 9 months ago

I do not know calculus, is there any other way to solve this?

Tisya Rawat - 5 months, 1 week ago

Log in to reply

There are several different ways. E.g. Jon's solution below deals with it nicely.

They key step is to realize that there is a restriction of a + b 20 a + b \geq 20 ,

Calvin Lin Staff - 5 months ago

ok, got it, thanks :)

Tisya Rawat - 5 months ago
Jon Haussmann
Aug 15, 2014

By applying AM-GM 2000 = a b ( a + b ) ( a + b 2 ) 2 ( a + b ) , 2000 = ab(a + b) \le \left( \frac{a + b}{2} \right)^2 (a + b), which implies a + b 20 a + b \ge 20 .

Again by the AM-GM inequality, 1 a + 1 b + 1 a + b = 1 2 a + 1 2 a + 1 2 b + 1 2 b + 1 a + b 5 1 ( 2 a ) 2 1 ( 2 b ) 2 1 a + b 5 = 5 1 16 a 2 b 2 ( a + b ) 5 = 5 a + b 16 a 2 b 2 ( a + b ) 2 5 5 20 16 200 0 2 5 = 1 4 . \begin{aligned} \frac{1}{a} + \frac{1}{b} + \frac{1}{a + b} &= \frac{1}{2a} + \frac{1}{2a} + \frac{1}{2b} + \frac{1}{2b} + \frac{1}{a + b} \\ &\ge 5 \sqrt[5]{\frac{1}{(2a)^2} \cdot \frac{1}{(2b)^2} \cdot \frac{1}{a + b}} \\ &= 5 \sqrt[5]{\frac{1}{16a^2 b^2 (a + b)}} \\ &= 5 \sqrt[5]{\frac{a + b}{16 a^2 b^2 (a + b)^2}} \\ &\ge 5 \sqrt[5]{\frac{20}{16 \cdot 2000^2}} \\ &= \frac{1}{4}. \end{aligned}

Equality occurs when a = b = 10 a = b = 10 , so the minimum value is 1/4.

1 a + 1 b + 1 a + b = ( a + b ) 2 + a b a b ( a + b ) \frac{1}{a}+\frac{1}{b}+\frac{1}{a+b}=\frac{(a+b)^2+ab}{ab(a+b)}

Plugging in the first equation, ( a + b ) 2 + a b 2000 \Rightarrow \frac{(a+b)^2+ab}{2000}

Now, by A.M. G.M. inequality, the minimum value of ( a + b ) 2 = 4 a b (a+b)^2=4ab , when a = b a=b

Now, 5 a b 2000 = a b 400 = a 2 400 \frac{5ab}{2000}=\frac{ab}{400}=\frac{a^2}{400}

In the first equation, plugging in a = b a=b

2 a 3 = 2000 a 3 = 1000 a = b = 10 2a^3=2000\rightarrow a^3=1000 \rightarrow a=b=10

Therefore, a 2 400 = 100 400 = 1 4 \frac{a^2}{400}=\frac{100}{400}=\boxed{\frac{1}{4}}

Interesting approach, but flawed. You cannot apply the condition that the minimum in the first inequality occurs when a = b a=b .

All that you have is that

( a + b ) 2 + a b 2000 a b 400 \frac{ (a+b)^2 + ab } { 2000} \geq \frac{ab}{400}

You have not shown that over ALL positive reals which satisfy a b ( a + b ) = 2000 ab(a+b)=2000 , we have a b 400 1 4 \frac{ab}{400} \geq \frac{1}{4} .

In fact, this statement is not true. Since 2000 = a b ( a + b ) a b × 2 a b 10 a b 2000 = ab (a+b) \geq ab \times 2 \sqrt{ab} \Rightarrow 10 \geq ab , thus we actually have a b 400 1 4 \frac{ab}{400} \leq \frac{1}{4} , which is not useful.

As it turns out, you have "over relaxed" the inequality with your first step.

Calvin Lin Staff - 6 years, 10 months ago

Log in to reply

I started to do the same thing but then I got to the part where I would have had to apply the "If it looks like a = b a=b will be the solution then a = b a=b will be the solution" principle.

Trevor B. - 6 years, 9 months ago

Log in to reply

The uniqueness thereom thing right???

Tanya Gupta - 6 years, 9 months ago

/((a+b)^2 = (a-b)^2 + 4ab but since, the expression is a sum of both positive terms so, if any one of the terms vanish the result will be always be smaller than the expression itself which needs no further logical explanation i think. And here in this expression, the only term that may vanish is the first term (at a=b) so what is obtained afterward should always be the smallest possible value!!

Shahzad Karim - 6 years, 9 months ago

You can verify your result using Calvin's result (ab<=100) and the fact that under the constraint, there is only one intersection point between (a+b)^2+ab and 5ab. They are equal iff a = b, and this only occurs at a = b = 10 because of the constraint. Therefore, their intersection point must be the minimum of (a+b)^2 + ab (because of continuity as well).

James Wilson - 3 years, 9 months ago

Log in to reply

Actually, I just realized that that's not sufficient.

James Wilson - 3 years, 9 months ago

Log in to reply

So, my apologies. I guess this gap can't easily be filled quite so easily.

James Wilson - 3 years, 9 months ago
Christian Howard
Aug 17, 2014

I approached this problem using Lagrange Multipliers. I essentially defined the following cost function, J J , to be minimized with respect to the variables a a , b b , and the Lagrange Multiplier λ \lambda :

F 2 ( a , b ) = a b ( a + b ) α = 0 F_{2} (a,b) = ab(a+b) - \alpha = 0

F 1 ( a , b ) = 1 a + 1 b + 1 a + b = ( a 2 + 2 a b + b 2 ) ( 1 α ) F_{1} (a,b) = \frac{1}{a} + \frac{1}{b} + \frac{1}{a+b} = (a^{2} + 2ab + b^{2})(\frac{1}{\alpha})

J = F 1 ( a , b ) + λ F 2 ( a , b ) J = F_{1}(a,b) + \lambda F_{2}(a,b) , where λ \lambda is unknown

Taking the partial derivatives of J J with respect to a, b, and λ \lambda (since we want to minimize with respect to all variables) leads to the following equations:

(1) ( 2 a + 3 b ) = α λ ( 2 a b + b 2 ) (2a+3b) = \alpha \lambda (2ab + b^{2})

(2) ( 2 b + 3 a ) = α λ ( 2 a b + a 2 ) (2b + 3a) = \alpha \lambda (2ab + a^{2} )

(3) F 2 = 0 F_{2} = 0

Now, solving for α λ \alpha \lambda in both (1) and (2) and setting them equal gives:

(4) 2 a + 3 b 2 a b + b 2 = 2 b + 3 a 2 b a + a 2 \frac{2a+3b}{2ab+b^{2}} = \frac{2b+3a}{2ba+a^{2}}

Simplifying (4) such that there are no fractions and multiplying through will lead to:

(5) 2 a 3 + a 2 b b 2 a 2 b 3 = 0 2a^{3} + a^{2}b - b^{2}a - 2b^{3} = 0

The following are some steps of algebraic simplification I made:

a b ( a b ) + 2 ( a 3 b 3 ) = 0 ab(a-b) + 2(a^{3}-b^{3}) = 0

a b ( a b ) + 2 ( a b ) ( a 2 + a b + b 2 ) = 0 ab(a-b) + 2(a-b) (a^{2}+ab+b^{2}) = 0

( a b ) [ a b + 2 ( a 2 + a b + b 2 ) ] = 0 (a-b) [ ab + 2 (a^{2}+ab+b^{2}) ] = 0

(6) ( a b ) [ 2 a 2 + 3 a b + 2 b 2 ] = 0 (a-b)[2a^{2}+3ab+2b^{2}] = 0

Obviously, based on (6), a = b a^{*} = b^{*} , where ( a , b ) (a^{*},b^{*}) are the optimal pair of values. Additionally, you can find that the other term can only be simplified if you allow complex numbers. Since I am under the assumption we are looking for a pure real solution, I ignore that term.

Now plugging a = b a^{*} = b^{*} into (3), we obtain that:

a = b = α 2 3 a^{*} = b^{*} = \sqrt[3]{\frac{\alpha}{2}}

Given that α = 2000 \alpha = 2000 for our case, we can find that

a = b = 10 a^{*} = b^{*} = 10

Plugging this into F 1 F_{1} gives:

F 1 ( a , b ) = 1 10 + 1 10 + 1 20 = 1 4 F_{1}(a^{*},b^{*}) = \frac{1}{10} + \frac{1}{10} + \frac{1}{20} = \boxed{\frac{1}{4}}

U Z
Sep 12, 2014

ab(a+b)=10x10x20 by comparing a=10 and b=10 i ticked 1/4 as the answer by this but i am not able to understand why 10 is the maximum value

Prateek Gupta
Jan 17, 2015

since the function has one a kind of symmetry we can say that the max or min value (whichever exists) is achieved when a=b .

solving it we get a = b = 10 substituting the values we get that answer as 1/4

No, that claim is not true.

What is the minimum of ( a 2 ) 2 ( b 7 ) 2 (a-2)^2 ( b-7)^2 subject to a + b = 10 a + b = 10 ? At what points does the minimum exist?

Calvin Lin Staff - 6 years, 4 months ago

If you could prove that the values of a and b that obtain the minimum are unique, then, in this case, you could conclude a = b. That seems to be beyond me at the moment though.

James Wilson - 3 years, 9 months ago

The expression can have minimum value, when a=b

Therefore, =>ab(a+b)=2000 =>a^2(2a)=2000 =>2a^3=2000 =>a=10

Simplifying the latter expression, =>1/a + 1/b + 1/a+b =>1/a + 1/a + 1/2a =>5/2a

Substituting a, we get =>5/20 = 1/4

Suresh Bala
Sep 4, 2014

x = 1 a + 1 b + 1 ( a + b ) x\quad =\quad \frac { 1 }{ a } +\frac { 1 }{ b } +\frac { 1 }{ (a+b) } x = ( a + b ) 2 + a b a b ( a + b ) \Rightarrow x\quad =\quad \frac { (a+b)^{ 2 }+ab }{ ab(a+b) }

From AM-GM, we have ( a + b ) 2 4 a b (a+b)^{ 2 }\geq 4ab

It is evident that for x to be minimum, a = b a = b a = 10 \Rightarrow a=10

L e a s t v a l u e o f x = 5 a 2 2000 = 1 4 \therefore \quad Least\quad value\quad of\quad x=\frac { 5a^{ 2 } }{ 2000 } =\quad \frac { 1 }{ 4 }

Faraz Khan
Aug 18, 2014

I was told by a tutor in my school days that whenever a problem with symmetrical variables is given ,consider all of them equal for getting a maximum ora minimum....and it works always...put a=b and get to the answer in secomds orally..

That's a good rule of thumb, but it isn't always literally true. For instance, try to find the maximum and/or minimums of f(x,y) = x^4 + y^4 subject to the constraints x^2 + y^2 = 1. Both of these are symmetric in x and y, and yet the solutions are NOT when x=y.

It happened to be true in the given problem above, but this requires additional proof.

Charlie Cunningham - 6 years, 9 months ago
Chirag Bharadwaj
Aug 17, 2014

I was wondering if you could use Lagrange Multipliers? I see that we have a function and a constraint in 2D, so applying the multipliers could definitely work? I ended up getting a symmetric equation in which the left side and the right side were the same except with a and b playing reversed roles.

As a result, I concluded that ONE solution is a = b (for sure, as if you reverse the roles of a and b and still have the same expression, surely a = b). From this I proved that a = b gives the minimum value of the sum, which turned out to be 1/4.

My question is if this is a valid approach? If not, can someone point out where my logic is flawed? Thank you so much!

Yes you can. I just posted a solution using this.

Christian Howard - 6 years, 9 months ago

Log in to reply

Just saw it. Thanks for the confirmation!

Chirag Bharadwaj - 6 years, 9 months ago

These are cyclic expressions. By putting a=b=10 we get extremum. With these values, the required expression = .25. let us assume a=1. Solving we get the value of the required expression > .25. So, .25 is the minimum.

OR
f ( a , b ) = 1 a + 1 b + 1 a + b = ( a + b ) 2 + a b a b ( a + b ) = ( a + b ) 2 + a b 2000 . F o r e x t r e m u m f a = 0 a n d f b = 0. 2 ( a + b ) 1 + b 2000 = 0 a n d 2 ( a + b ) 1 + b 2000 = 0. a = b , s u b s t i t u t i n g i n g i v e n e q u a t i o n a 2 2 a = 2000 , a = b = 10. S o f ( 10 , 10 ) = 1 10 + 1 10 + 1 20 = 1 4 . C h e c k i n g b y 2 n d d e r i v a t i v e t h i s i s t h e m i n i m u m . f(a,b)=\dfrac 1a+\dfrac 1b+\dfrac 1{a+b}=\dfrac {(a+b)^2+ab}{ab(a+b)}=\dfrac {(a+b)^2+ab}{2000}.\\ For ~extremum~ f_a'=0~ and ~f_b'=0.\\ \therefore~~~\dfrac{2(a+b)*1+b}{2000}=0~~~~~and~~~~~~\dfrac{2(a+b)*1+b}{2000}=0.\\ \implies~a=b,~~substituting~in~given ~equation~~~~~~~a^2*2a=2000,~~~~~~\implies~a=b=10.\\ So~f(10,10)=\dfrac1{10}+\dfrac1{10}+\dfrac1{20}=\dfrac 14.\\ Checking~by~2nd~derivative~this~ is~ the~ minimum.

0 pending reports

×

Problem Loading...

Note Loading...

Set Loading...