Take the 2-minute tour ×
Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

Let there be given a set of $n$ real numbers, $\{r_i\} \subset (0,1)$; is it possible to find some conditions satisfied by a real number $m \in \mathbb{R}$ to ensure that:

$\hspace{6cm}r_1^m+r_2^m+r_3^m+...+r_n^m=1$

For example, it might be possible to define such an $m$ by an infinite series.

share|improve this question
1  
Firstly, you should probably require $r_i \in (0,1)$ instead of $[0,1]$. If $r_i = 0$ then you throw it out, and it does not change the answer. If $r_i = 1$, then all other terms $r_j$ with $i \neq j$ must necessarily be $0$, and then it matters not what power of $m$ you have. Secondly, I am not sure of what you are truly asking. Are you asking, "Given a set of $n$ integers $r_1 , \dots r_n$, how does one find a real number $m$ such that $r_1^m + \dots + r_n^m = 1$"? Or are you asking, "For what values of $m$ can one find $r_1, \dots , r_n$ with $r_1^m + \dots + r_n^m = 1$"? –  JavaMan Feb 28 '11 at 16:09
1  
After re-reading your entry, you may be asking a third question. Namely, "Is there any number $m$ such that for any sequence $\{r_i\}_{i=1}^n$, we have $r_1^m + \dots + r_n^m = 1$? –  JavaMan Feb 28 '11 at 16:20
    
@DJC: edited to make the interval open instead of closed. Thx. In response to your second comment, I am asking that question implicitly as well. –  Matt Calhoun Feb 28 '11 at 20:40

2 Answers 2

up vote 6 down vote accepted

A hint: For given $r_i\in ]0,1[$ consider the function $f(x):=\sum_{i=1}^n r_i^x$ and see what happens for $x\to 0$ resp. $x\to\infty$.

My response to the comments:

One has $f(0)=n>1$ and $\lim_{x\to\infty} f(x)=0$; furthermore $f$ is monotonically decreasing, so there is exactly one $x>0$ with $f(x)=1$. Unless all $r_i$ are equal (and maybe in some other special cases) it is not possible to solve the equation $f(x)=1$ explicitly. For a numerical solution assume $r_1\leq r_2\leq\ldots\leq r_n$ and put $$a:={\log n\over \log(1/r_1)},\qquad b:={\log n\over \log(1/r_n)}.$$ Then $f(a)\geq 1$ and $f(b)\leq 1$. Now use binary search.

share|improve this answer
2  
so it has at least one solution... how does that answer the question? –  yoyo Feb 28 '11 at 16:57
1  
@yoyo: It actually has a unique solution, since $f(x)$ is monotone. –  Yuval Filmus Feb 28 '11 at 19:19
    
I'm wondering since you phrased this as a hint, do you know the solution to this problem, or were you trying to show that the problem would or wouldn't have a solution? –  Matt Calhoun Mar 1 '11 at 1:03
    
@Matt, he meant to show that there is a solution, not to construct it. I don't think there is a general rule to construct a "closed formula" for the solution. But you could use some numerical method like Newton-Raphson once you establish there is a unique solution. If we know more about the $r_i$ though, we can make more informed guesses as to what the solution is. –  Raskolnikov Mar 1 '11 at 9:15
    
@Raskolnikov: I actually don't have any particular set of $r_i$ in mind! @Christian: I'm afraid you might be correct that only numerical methods will work for this problem, but I don't see why in a clear way. Would you mind linking me or explaining a short proof of this fact? Thx! –  Matt Calhoun Mar 1 '11 at 14:41

It is possible to tighten the range of possible values for $m$ to the interval

$$-\frac{\ln n }{\ln \min_i r_i} \leq m \leq -\frac{\ln n }{\ln \max_i r_i}$$

which is not so good if the maximum and the minimum of the $r_i$ are very close to 1 and 0 respecively.

share|improve this answer

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.