Is anyone able to provide a(n) example(s) of strictly positive, increasing, concave function where the second derivative becomes larger (in absolute value) as $x$ increases?

Or, if this is not possible, can someone provide some intuition or a proof/link to explain?


  • $\begingroup$ I don't think you can have strictly positive and $|f''|$ increasing since that would mean that $f''$ is decreasing ($f$ is concave so $f'' \le 0$) and therefore that $f'$ becomes negative at some point on any interval containing a positive ray. $\endgroup$ Feb 7 '20 at 0:16

Not on all of $(-\infty,\infty)$, no. In fact, if $f$ is increasing and concave on $(-\infty,\infty)$ then it must satisfy $\lim_{x\to-\infty}f(x)=-\infty$.

To see this, note that since $f$ is increasing, its derivative exists at some $x_0\in(-\infty,\infty)$. Since $f'(x_0)>0$, the tangent line $T(x)$ to $f$ at $x_0$ has positive slope, so that $T(x)\to-\infty$ as $x\to-\infty$. However due to concavity of $f$ we have $f(x)\leq T(x)$ for all $x\in(-\infty,\infty)$, and hence $f(x)\to -\infty$ as $x\to-\infty$ as claimed.

It's also impossible to find an example on a positive ray, although for different reasons. Without loss of generality we may suppose $f:(0,\infty)\to(-\infty,\infty)$ is concave and that $f''$ exists with $|f''(x)|$ increasing. Since $f''(x)<0$, that means $f''$ is decreasing. In particular, there are $a,\delta>0$ such that $f'(x)<-\delta$ for all $x\geq a$. Note that there is $C\in(-\infty,\infty)$ such that $f'(x)=C+\int_a^xf''(t)\;dt\leq C+\delta a-\delta x\to -\infty$ as $x\to\infty$. Hence $f$ is eventually negative.

If you want $f$ defined on an interval, that means the interval must be bounded. This is quite easy to do. For instance let $f:(0,1)\to(0,\infty)$ be defined by $f(x)=3x-x^3$.

  • $\begingroup$ I think your answer also generalizes to show that it cannot happen on a half line like $(0,\infty)$ or $(-\infty, 0)$ either. $\endgroup$
    – angryavian
    Feb 7 '20 at 0:06
  • $\begingroup$ @angryavian You need a different argument for that (see my edits). But if all you want is $f$ increasing, concave, and positive, then you can take $f(x)=\log(x+1)$ restricted to $(0,\infty)$. $\endgroup$
    – Ben W
    Feb 7 '20 at 0:58
  • $\begingroup$ Thanks. This is a nice answer. Two quick follow up questions if you have a moment: 1) For the paragraph on the positive ray, is the WLOG referring to the domain of $f$ being $(0,\infty)$ (as opposed to say $(5,\infty)$? Because I don't think we need WLOG for the rest of that sentence (concave, increasing absolute value and second derivative exists), as those are our assumptions. 2). $f'(x) = C +\int f''(t)dt)$ -- is this trivial? i.e., just define C to be the difference between the integral and the first derivative. $\endgroup$
    – user106860
    Feb 7 '20 at 4:31

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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