1
$\begingroup$

I am trying to prove following theorem . Is there any flaw in below proof

Assume that $f$ is integrable on $[a,b]$ and has a jump discontinuity at $c \in (a,b)$ this means that both one sided limits exist as $x$ approaches $c$ from the left and right but that $\lim_{x \rightarrow c^{-}} f(x) \neq \lim_{x \rightarrow c^{+}} f(x) $ then show that function $F(x) = \int_a^x f(t) \, dt$ is not differentiable at $x=c$.

Following is the proof strategy

  • prove that $\lim_{x \rightarrow c^{-}} \frac{F(x) - F(c)}{x-c} = \lim_{x \rightarrow c^{-}} f(x) $
  • prove that $\lim_{x \rightarrow c^{+}} \frac{F(x) - F(c)}{ x- c} = \lim_{x \rightarrow c^{+}} f (x) $
  • from given hypothesis we immediately have that $\lim_{x \rightarrow c^{-}} \frac{F(x) - F(c)}{x-c} \neq \lim_{x \rightarrow c^{+}} \frac{F(x) - F(c)}{x-c}$
  • conclude that function $F$ is not differentiable at $c$.

first part can proved as follows

we will show that $\lim_{x \rightarrow c^{-}} \frac{F(c) - F(x)}{c-x} - f(x) = 0$ consider arbitrary $\epsilon > 0$

$|\frac{F(x) - F(c)}{x-c} - f(x) | \leq |\frac{F(x) - F(c)}{x-c} - f(c)| + |f(c) - f(x)| < \epsilon/2 + \epsilon/2 = \epsilon$

sine we know that $f$ is left continuous we have that $F$ is left differetiable from Fundamental Theorem of Calculus then there exists $\delta_1, \forall x$ s.t $ c-x < \delta \implies |\frac{F(x) - F(c)}{x-c} -f(c)| < \epsilon/2$ and again since $f$ is left continuous we have that there exists $\delta_2, \forall x$ s.t $\forall c-x < \delta \implies |f(x) - f(c)| < \epsilon/2$ required $\delta = \min\{\delta_1 , \delta_2\}$ and similarly we can show for right limits and we are done

$\endgroup$
6
  • $\begingroup$ Looks good to me although I wonder if you should write there exists $\delta_1, \forall x$ s.t $ c-x < \delta_1 \implies |\frac{F(x) - F(c)}{x-c} -f| < \varepsilon/2$ and similarly for $\delta_2$? Your picking of a $\delta$ which is the minimum of these two then holds up the proof, $\endgroup$
    – user284001
    Jan 22 '20 at 9:43
  • 1
    $\begingroup$ Are you assuming that $f$ is continuous everywhere except possibly at $c$? You have used this but have not mentioned it in the body, and you cannot use FTC in parts where you haven't been given it, right? $\endgroup$ Jan 22 '20 at 9:54
  • 1
    $\begingroup$ Are you assuming that $f$ is continuous everywhere except possibly at $c$? You have used this but have not mentioned it in the body. $\endgroup$ Jan 22 '20 at 9:56
  • 1
    $\begingroup$ I would like to think that it does not hold, but I can't find a counterexample. $\endgroup$ Jan 22 '20 at 12:16
  • 1
    $\begingroup$ I fixed a few typos. Hope you don't mind. $\endgroup$
    – Paramanand Singh
    Jan 22 '20 at 12:38
4
$\begingroup$

You can't assume left or right continuity of $f$. Use the same approach as in proof of FTC and show the following.

Lemma: If $f$ is Riemann integrable on $[a, b] $ and if for some $c\in[a, b) $ the limit $\lim_{x\to c^{+}}f(x) $ exists then the function $F$ defined by $$F(x) =\int_{a} ^{x} f(t) \, dt$$ has right derivative at $c$ and we have $D^{+} F(c)=\lim_{x\to c^{+}} f(x) $.

Let $f(x) \to L$ as $x\to c^{+} $ and $\epsilon >0$ be arbitrary. Then there is a $\delta >0$ such that $$|f(t) - L|<\epsilon $$ whenever $0<t-c<\delta$. Thus we have $$L-\epsilon <f(t) <L+\epsilon$$ whenever $0<t-c<\delta$. Integrating the above inequality in interval $[c, c+h] $ where $0<h<\delta$ we get $$h(L-\epsilon) <\int_{c} ^{c+h} f(t) \, dt<h(L+\epsilon) $$ or $$L-\epsilon<\frac{F(c+h) - F(c)} {h} <L+\epsilon$$ whenever $0<h<\delta$. Thus $$D^{+} F(c) =\lim_{h\to 0^{+}}\frac {F(c+h) - F(c)} {h} =L=\lim_{x\to c^{+}} f(x) $$ A similar lemma can be proved in same manner for left limit and left derivative. Going further if $f(x) \to L$ as $x\to c$ then $F$ is differentiable at $c$ with $F'(c) =L$ and if $f$ is continuous at $c$ then $L=f(c) $ so that $F'(c) =f(c) $ which brings us to the traditional form of FTC.

It is now obvious that if $f$ has a jump discontinuity at some point then $F$ is not differentiable at that point. However if $f$ has essential discontinuity at some point then $F$ may or may not be differentiable at that point.

$\endgroup$
6
  • $\begingroup$ But if $f$ has removable discontinuity then $F$ is differentiable at that point right ? $\endgroup$
    – manifold
    Jan 22 '20 at 12:40
  • 1
    $\begingroup$ @viru: I added that also in my answer. See second last paragraph. $\endgroup$
    – Paramanand Singh
    Jan 22 '20 at 12:47
  • $\begingroup$ Yes . That is what made me ask about remaining case of discontinuity which is removable discontinuity.I was thinking removable discontinuity is different from Essential Discontinuity. Removable discontinuity is when $\Big[ \lim_{x \rightarrow c^{+}} f = \lim_{x \rightarrow c^{-}}f \Big] \neq f(c)$ which readily implies from your proof that it is differentiable at c . right ? or is there anything that I am missing $\endgroup$
    – manifold
    Jan 22 '20 at 14:16
  • $\begingroup$ @viru: you are correct. I have stated that in that case $F'(c) =L=\lim_{x\to c} f(x) $. $\endgroup$
    – Paramanand Singh
    Jan 22 '20 at 15:32
  • 1
    $\begingroup$ @viru: essential discontinuity is when one or both of $\lim_{x\to c^{+}} f(x), \lim_{x\to c^{-}} f(x) $ don't exist. As an example let $f(x) =\cos(1/x),x\neq 0,f(0)=0$. Then $f$ has essential discontinuity at $0$ but $F'(0)$ exists and equals $0$. $\endgroup$
    – Paramanand Singh
    Jan 22 '20 at 15:34

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.