2
$\begingroup$

How can I prove the function $f(t) = t\left(s-1\right)+1-s^t$ is strictly positive for $t \in (0, 1)$ when $s \geq 0, s \neq 1$? I tried using inequalities but could not find a satisfactory solution.

Here is a plot of the relevant graph (in green) on Desmos: https://www.desmos.com/calculator/wqttkdfinw.

$\endgroup$
  • $\begingroup$ presumably $s$ is more constrained than that? for example, I would think $s$ is positive $\endgroup$ – qbert Apr 20 '18 at 22:37
  • $\begingroup$ @qbert Correct. $\endgroup$ – user76284 Apr 20 '18 at 22:38
4
$\begingroup$

Hint:   the case $s=0$ is trivial, otherwise note that $\,f(0)=f(1)=0\,$ and $f$ is strictly concave for $s \gt 0, s \ne 1$ since $f''(t) = - \ln^2(s) \cdot s^t \lt 0\,$.

$\endgroup$
  • $\begingroup$ what is baffling with this solution is that $f''$ is such easy to study, while both $f$ and $f'$ are intractable. $\endgroup$ – zwim Apr 20 '18 at 23:32
  • $\begingroup$ @zwim The problem sort of obfuscates the simpler fact that $f(t)=0$ are the intersections between the (convex) exponential $s^t$ and the line $(s-1)t+1$, which so "happen" to be the endpoints of $(0,1)$. $\endgroup$ – dxiv Apr 20 '18 at 23:48
  • $\begingroup$ I'm probably being dense, but how does this show $f(t) > 0$? I can "see" why it's the case intuitively, especially when I plot the graph, but I'm having trouble formally proving it. What are the relevant theorems? Intermediate value theorem? $\endgroup$ – user76284 Apr 21 '18 at 0:02
  • 1
    $\begingroup$ @user76284 Jensen's inequality: $\;f(t) = f\big((1-t)\cdot 0 + t \cdot 1\big) \gt (1-t)\cdot f(0) + t \cdot f(1) = 0\,$. $\endgroup$ – dxiv Apr 21 '18 at 0:15

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.