Find the largest integer n , such that if { x i } i = 1 n are non-negative real numbers such that
x 1 6 + x 2 6 + … + x n 6 = n ,
then we will have
x 1 2 + x 2 2 + … + x n 2 ≤ x 1 5 + x 2 5 + … + x n 5 .
Details and assumptions
The second condition is true for all sets which satisfy the first condition. You are not asked to find just 1 set of numbers which satisfies both conditions.
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.
Great solution , heads up to you man.
SOOOOOOO much more elegant than what I did. xD
Log in to reply
I thought the solution was too technical. What did you did (or in summary)?
Explanation, maybe?
Problem Loading...
Note Loading...
Set Loading...
Alternatively, we want to find the smallest n such that there exist some x i with ∑ i = 1 n x i 6 = n and ∑ i = 1 n x i 5 − ∑ i = 1 n x i 2 < 0 . So, first, I will find how a minimum solution for any n looks like, then I will solve the n that gets that minimum to less than 0.
Let f ( t ) = t 5 / 6 − t 1 / 3 (we only use t>0) then we want to minimize ∑ i = 1 n f ( y i ) , with y i = x i 6 and ∑ i = 1 n y i = n .
We will try to use Jensen (Or, rather, Karamata (http://en.wikipedia.org/wiki/Karamata's_inequality)). We should note that f ′ ′ ( t ) = 3 6 t 5 / 3 8 − 5 t 1 / 2 and that is positive if 0 < t < 2 5 6 4 and negative when 2 5 6 4 < t .
Karamata inequality, roughly, says that if f is convex then ∑ i = 1 n f ( α i ) . given that the sum of the α 's is constant, when all the α i are equal. When is concave, it minimizes when they are as further apart as they can. So let the solution that minimizes it be { y i } = { a i } ⋃ { b i } , with the a's less than 64/25, and the b 's greater than. Note that we can consider 64/25 's in any of the sets we want, but we will prefer to see them as a 's.
Because, in the domain ( 0 , 6 4 / 2 5 ] f is convex, then all the a's must be equal because if not, Karamata inequality would find a solution with lower sum of f .
Because in the domain [ 6 4 / 2 5 , ∞ ) f is concave, then if there are at least 2 b 's, one must be at the extreme, otherwise Karamata will give us a lower solution. But then we can consider it rather an a and not an. So there it is at most one b .
So the minimal solution is ( a , a , . . a , b ) with n − 1 a 's and one b , with a and b not necessarily different. If they are different, though, then b > 6 4 / 2 5 > a . So overall b ≥ a
Now we will proceed to find the n that works. For simplicity, I will suppose there are n+1 terms (So that I can use n a's and one b).
We will use the technique of homogenization(http://www.artofproblemsolving.com/Wiki/index.php/Homogenization). Because x i > 0 then the inequality is equivalent to squaring it and multiply it by (n+1) so it is equivalent to ( n + 1 ) ( ∑ i = 1 ( n + 1 ) x i 5 ) 2 ≥ ( n + 1 ) ( ∑ i = 1 ( n + 1 ) x i 2 ) 2 . Using the hypothesis on the right side, then ( n + 1 ) ( ∑ i = 1 ( n + 1 ) x i 5 ) 2 ≥ ( ∑ i = 1 ( n + 1 ) x i 6 ) ( ∑ i = 1 ( n + 1 ) x i 2 ) 2 . Now the inequality is homogeneous, that is, every term is of the same degree. So if we multiply every number by a positive constant k then both sides are multiplied by a positive constant C k , and it is equivalent so we can forget the hypothesis.
Remembering what we prove, it is true for every {x_i} if and only if is true for every ( a , a , . . . , a , b ) , for any a , b . So we will plug it in the inequality. So the inequality is equivalent to ( n a 2 + b 2 ) 2 ( n a 6 + b 6 ) ≤ ( n a 5 + b 5 ) ( n + 1 ) . If a=0 it is trivially true. Because homogeneity, we can make an assumption and retain generality. So we will assume a = 1 and now we know that b ≥ a = 1 (we can because a is not 0) and so the inequality is equivalent to ( n + b 2 ) 2 ( n + b 6 ) ≤ ( n + b 5 ) ( n + 1 ) . We pass the left side to the right side and factor it to ( b − 1 ) 2 ( b 2 + b + 1 ) ( b 6 + b 5 − b 4 − n ( b 2 − b − 1 ) ) . We know ( b − 1 ) 2 , b 2 + b + 1 are always positive, so the whole inequality is equivalent to b 6 + b 5 − b 4 − n ( b 2 − b − 1 ) ≥ 0 . If b 2 − b − 1 is not positive, then it is also trivially true because it would be positive plus positive. So we can asume is b 2 − b − 1 ≥ 0 then b > ϕ .
So the inequality is, finally, equivalent to n ≤ b 2 − b − 1 b 6 + b 5 − b 4 = g ( b ) for any b > ϕ . Because g it is well defined in ( ϕ , ∞ ) , has no singularities and the extremes tend to positive infinity, then the global minimum is also a local minimum. So we have to find the minimum of g . With a little of calculus and factorizing it, we see that g ′ ( t ) = 0 at t = − 1 , 0 , 1 / 2 , 2 . Of those, only 2 is greater than ϕ (also g ′ ′ ( 2 ) > 0 ). So that would be the minimum , and at that point g ( 2 ) = 8 0 . So we have that the inequality is true for n + 1 variables if and only if n ≤ g ( b ) for any b in ( ϕ , ∞ ) if and only if n ≤ min g ( b ) if and only if n ≤ 8 0 .
So the largest number of variables allowed is 8 0 + 1 = 8 1