Given strictly concave function $f(t)$ that satisfies $f'(t)>0$, $f'(t)=o(1)$ (i.e. $\lim\limits_{t\to\infty}f'(t)=0$) , and $f'(t)=\omega\left(t^{-1}\right)$ (i.e. $\lim\limits_{t\to\infty}tf'(t)=\infty$) , we denote $F=\exp\left(f\left(t\right)\right)$. Let $H(t)$ be a solution of $$ \dot{H}=F $$ Can we approximate H by F?

Specifically, I want to show that $$ \lim_{s\to\infty}\frac{H^{-1}(s)}{F^{-1}(s)}=1\,, \ \ \ \ (a) $$ where $F^{-1}$ and $H^{-1}$ are the inverse functions of $F$ and $H$, respectively.

If necessary I can add additional assumptions on $f(t)$, though I would like to keep it as general as possible.

* In this question, $(a)$ was proved for $f'(t)=\Omega(1)$ (without the assumption that $f(t)$ is strictly concave). I'm interested in the case that $f'(t)=o(1)$.

  • $\begingroup$ By $\dot{H}$ you mean the derivative of $H$, or something else ? $\endgroup$ – N. de Rancourt Oct 18 '18 at 19:19
  • $\begingroup$ Yes, the derivative of $H$. $\endgroup$ – Mor Oct 18 '18 at 19:22

Not a full answer, but a hint at a negative answer: take $f(t)=\log t$, which doesn't satisfy $\lim_{t\to\infty}tf'(t)=\infty$, critically, but is nearer to the proven case (where $f'(t)=\omega(1)$ ).

In this case you have $F(t)=t$, $H(t)=t^2/2$, $F^{-1}(s)=s$, $H^{-1}(s)=\sqrt{2s}$.

  • 2
    $\begingroup$ Thank you for your comment. I agree that eq. (a) doesn't hold for $f(t)=\log(t)$ which doesn't satisfy the condition $f'(t)=\omega(t^{-1})$. However, I think that with this condition the equation should be true. We have that $\dot{H}(t)=F(t)=\exp(f(t))$. If $\exp(f(t))\approx\exp(f(t))f'(t)=\frac{d}{dt}\exp(f(t))$ then I think we can approximate $H\approx \exp(f) = F$. For this to hold we need $\log(f’(t))=o (f (t))$ (i.e. $\lim\limits_{t\to\infty}\frac{\log(f’(t))}{f(t)}=0$). Note that $f(t)=\log(t)$ doesn’t satisfy this requirement, but $f(t)=\log^\epsilon(t)$ for $\epsilon>1$ does. $\endgroup$ – Mor Oct 20 '18 at 11:46
  • $\begingroup$ Yes, it's more subtle than I thought... $\endgroup$ – Jean Duchon Oct 24 '18 at 16:16

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.