Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Let $a,b>0$, $$f(x)=\dfrac{xa+(1-x)b}{a^xb^{1-x}},\quad x\in [0, 1].$$ What is the maximum of $f(x)$?

I tried to find the derivative of $f(x)$, which is $$f'(x)=\left(\frac{a}{b}\right)^{1-x}\left(1-x\ln \frac{a}{b}\right)+\left(\frac{b}{a}\right)^{x}\left((1-x)\ln \frac{b}{a}-1\right).$$ It seems difficult to find the critical points of $f(x)$.

share|cite|improve this question
The expression $(a/b)^xf'(x)$ being an affine function of $x$, to find the location of the maximum seems straightforward. – Did Jul 24 '11 at 22:23
up vote 3 down vote accepted

It is natural to take the logarithm, but we do not have to. Here is an informative first step. Let $b=ca$. If we substitute $ca$ for $b$ in our expression, after a small amount of calculation we obtain $$\frac{cx+(1-x)}{c^x}.$$ This is useful, and not only as a simplifying device: It tells us that only the ratio $b/a$ matters.

Differentiation is straightforward: We get $$\frac{c^x(c-1)-(cx+ 1-x)(\ln c)c^x}{c^{2x}}.$$ The sign of the derivative is determined by the sign of $$(c-1)-(cx+1-x)\ln c.$$

Added: Suppose that $a \ne b$. We show that the maximum does not occur at an endpoint. By interchanging the roles of $a$ and $b$ if necessary, we can assume without loss of generality that $c>1$.

Let $g(x)=(c-1)-(cx+1-x)\ln c$. For $x$ positive but very close to $0$, $g(x)\approx -1-\ln c$. But from the Taylor expansion of $\ln(1+t)$, or otherwise, it is easy to see that $\ln(c)<c-1$ when $c-1$ is positive, so $g(x)$ is positive when $x$ is close enough to $0$. A similar argument shows that $g(x)$ is negative when $x$ is close enough to $1$. Thus the maximum occurs in the open interval $(0,1)$. In particular, the unique $x$ at which $(c-1)-(cx+1-x)\ln c=0$ is in the interior of $[0,1]$.

share|cite|improve this answer
Thanks for a different argument. Let $0<A\le a,b\le B$, is it true $f(x)\le \frac{A+B}{2\sqrt{AB}}$? I now find it hard to prove this in terms of finding the maximum of $f(x)$. – Sunni Jul 25 '11 at 13:12
@Sunni: A bit of experimentation seems to show it is plausible. No proof yet! – André Nicolas Jul 25 '11 at 23:12

HINT: (Answering the original question: Where are the critical points of the function) It is enough to maximize $g(t)$, where $$g(t) := \ln f(t) = \ln(at + (1-t)b) - t \ln a - (1-t) \ln b$$


$$ g'(t) = \frac{a-b}{at + (1-t)b} - \ln a + \ln b $$

and hence, $g'(t) = 0$, when

$$\begin{align} \frac{a-b}{at + (1-t)b} = \ln (a/b) \end{align}$$

(and so on...)

share|cite|improve this answer
A typo... I am asking the maximum of $f(x)$...Now I am sure that $1/2$ is not a critical point. – Sunni Jul 24 '11 at 22:40
DJC, I went ahead and corrected a minor typo in the last equation; hope that's ok :) Added: Ah, I don't have edit privileges, so the typo will persist till someone approves it... – Srivatsan Jul 24 '11 at 22:59
@Srivatsan: That's great with me! – JavaMan Jul 24 '11 at 23:00

Your Answer


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.