I am trying to prove an inequality which goes as

For $\alpha\geq 1$,$\int_{t}^\infty u^{\alpha}e^{-u}du\geq (t+1) \int_{t}^\infty u^{\alpha-1}e^{-u}du$ for any $t\geq 0$. Using the notation of incomplete gaussian function $\Gamma(\alpha+1,t)\geq (t+1)\Gamma(\alpha,t),\forall t\geq 0$.

For example:

enter image description here


We have $$\int_t^\infty u^\alpha e^{-u} du = t^\alpha e^{-t} + \alpha \int_t^\infty u^{\alpha -1} e^{-u} du.$$ The inequality is now equivalent to $$t^\alpha e^{-t} \geq (t+ 1-\alpha) \int_t^\infty u^{\alpha -1} e^{-u} du.$$ If $t\leq \alpha -1$, the inequality is trivial. We consider $t > \alpha -1$. Define the function $$f(t) = \frac{t^\alpha}{t+1 -\alpha} e^{-t} - \int_t^\infty u^{\alpha -1} e^{-u} du,\quad t > \alpha -1.$$ Differentiating $f$, we get $$f'(t) = -\frac{(\alpha -1) t^{\alpha -1}}{(t+ 1-\alpha)^2} e^{-t} < 0.$$ Therefore $$f(t) > \lim_{t\to \infty} f(t) = 0,\quad \forall \alpha -1 < t < \infty.$$

  • $\begingroup$ nice and effective ! (+1) $\endgroup$ – G Cab Apr 19 '18 at 21:29

I don't have the rep points to comment, so I've put this in an answer box, but have you tried showing the inequality holds at some convenient $t$ (say $t=0$), then differentiating both sides with respect to $t$ (using product rule on the right). If the derivative on the left is greater than that on the right for all $t$ and $\alpha$ under your conditions, then you've got it.

[update] You can consider this comparing two curves. One is the curve on the left of the inequality, the other on the right. If the curve on the left starts out greater than the one on the right in the range of interest ($t \in [0,\inf)$), and the slope of the curve on the left is always greater than the one on the right, then the function on the left must always be greater. [another update--I wrote this, but I think it's wrong, sorry; I see your point--] [wrong part] This can be true even though the difference between the two curves gets smaller and smaller as $t$ goes to $\inf$. [end wrong part]

  • $\begingroup$ The difference is not increasing as t goes to infinity. Actually, it is a very tight bound when t goes to infinity. $\endgroup$ – Ethanabc Apr 19 '18 at 20:04

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.