Suppose that $g,k: [0,a] \to \mathbb R$ are continuous, $a >0 $, $\,k(t) \ge 0$,$\ c(t) \in C^1([0,a])$, $\, \dot c(t) \ge 0 $ (i.e. $c(t)$ is non decreasing) and $g(t)$ satisfies $$g(t) \le c(t) + \int^{t}_{0} k(s) g(s)ds$$ for all $0 \le t \le a$.

I want to show that for all $t \in [0,a]$, $$ g(t) \le c(t)e^{\int^{t}_{0} k(s) ds} $$

I have noticed that there is a proof for a more general case in wikipedia. However, I do not quite understand the proof and since the above is a less general case, I would guess that there is a simpler way to prove it.

Let $G = c(t) + \int^{t}_{0} k(s) g(s)ds $. Then $G \in C^1([0,a])$. Taking the derivative, $$\dot G = \dot c + k(t)g(t) \\ \dot c = \dot G - k(t)g(t) \ge \dot G - k(t)G(t)$$.

I am stuck here and not sure if I am going in the right direction. I suppose the goal is to reach $$ \frac{d}{dt} (G(t)e^{-\int^{t}_{0} k(s)ds}) \le\frac{d}{dt} c(t) $$ and integrate both sides. Any hints would be appreciated.


Almost as in the answer of user161825, but with correct estimation.

Since $k(t) \geq 0$, we have $\int_0^t k(s) ds \geq 0$. Hence, $$ 0 < e^{-\int_0^t k(s) ds} \leq 1. $$ Therefore, since $c' \geq 0$, we get $$ \left(G'(t) - k(t) G(t) \right) e^{-\int_0^t k(s) ds} \leq c' e^{-\int_0^t k(s) ds} \leq c'. $$ However, $$ \frac{d}{dt}\left( G(t) e^{-\int_0^t k(s) ds} \right) = G'(t) e^{-\int_0^t k(s) ds} - k(t) G(t) e^{-\int_0^t k(s) ds} = \left(G'(t) - k(t) G(t) \right) e^{-\int_0^t k(s) ds}. $$ And thus, $$ \frac{d}{dt}\left( G(t) e^{-\int_0^t k(s) ds} \right) \leq \frac{d}{dt} c. $$

  • 1
    $\begingroup$ As far as I can tell, the argument I give is correct except for mentioning that we may assume a certain function to be positive. $\endgroup$ Dec 3 '14 at 14:47

You are almost there. Observe that \begin{align*} \dot c = \dot G - k(t)g(t) &\geq \dot G - k(t)G(t)\\ &= \exp\left(\int_0^t k(s)ds\right)\frac{d}{dt}\left\{G(t)\exp\left(-\int_0^t k(s)ds\right)\right\}\\ &\geq\frac{d}{dt}\left\{G(t)\exp\left(-\int_0^t k(s)ds\right)\right\}. \end{align*}

  • 2
    $\begingroup$ How do you know that the expression in the last line is non-negative? $\endgroup$
    – saz
    Nov 26 '14 at 13:37
  • $\begingroup$ @saz in case it is negative, the inequality is trivial! $\endgroup$ Dec 3 '14 at 14:46
  • $\begingroup$ Why? If $x <0$, then $x \leq \exp\left(\int_0^t k(s) \, ds \right) x$ does not hold. $\endgroup$
    – saz
    Dec 3 '14 at 15:46
  • $\begingroup$ I don't understand the relevance. If $$\phi(t)=\frac{d}{dt}\left\{G(t)\exp\left(-\int_0^t k(s)ds\right)\right\}$$ is negative, then it is trivial that $\dot c\geq 0 \geq \phi$. $\endgroup$ Dec 3 '14 at 15:57

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.