Constructing an antiderivative of a function if the contour integral depends on initial and final point

74 Views Asked by At

I am working on the following problem:

Let $D \subset \mathbb C$ be a domain, $f: D \to \mathbb C$ a continuous function and $\gamma : [\alpha, \beta] \to D$ a contour. Assume that $\int_\gamma f$ only depends on the initial and final points of $\gamma$. Fix a point $a \in D$ and define a function $F_a : D \to \mathbb C$ by \begin{align} F_a(z) := \int_{\gamma_{a,z}} f, \end{align} where $\gamma_{a,z} : [\alpha, \beta] \to D$ is a smooth contour with initial point $a$ and final point $z$. Then $F_a'(z) = f(z)$ in $D$, that is $F_a$ is an antiderivative of $f$ in $D$.

I tried to prove this:

Proof: We have that \begin{align*} F_a'(z) &= \lim_{h \to 0} \frac{F_a(z+h) - F_a(z)}{h} = \lim_{h \to 0} \frac{1}{h} \left(\int_{\gamma_{a,z+h}} f(w) \, dw - \int_{\gamma_{a,z}} f(w) \, dw\right) \\ &= \lim_{h \to 0} \frac{1}{h} \left(\int_{\gamma_{a,z}} f(w) \, dw + \int_{[z,z+h]} f(w) \, dw - \int_{\gamma_{a,z}} f(w) \, dw\right) = \lim_{h \to 0} \frac{1}{h} \int_{[z,z+h]} f(w) \, dw \\ &= \lim_{h \to 0} \frac{1}{h} \int_0^1 f((1-t)z + t(z+h)) \cdot ((1-t)z + t(z+h))'\, dt \\ &= \lim_{h \to 0} \frac{1}{h} \int_0^1 f(z-tz+tz+th) \cdot h\, dt = \lim_{h \to 0} \int_0^1 f(z+th)\, dt. \end{align*} Hence for every $\epsilon > 0$ there exists $\delta > 0$ such that for all $h \in D$ with $|h| < \delta$ \begin{align*} \left|\int_0^1 f(z+th)\, dt - F_a'(z)\right| < \frac{\epsilon}{2}. \end{align*} Since $f$ is continuous in $z \in D$, there is $\delta' > 0$ such that for any $h \in D$ with \begin{align*} t|h| = |th| = |(z+th)-z| < \delta' \Rightarrow |h| < \frac{\delta'}{t}, \quad (t > 0) \end{align*} we have \begin{align*} |f(z+th)-f(z)| < \frac{\epsilon}{2}. \end{align*} Therefore for all $h \in D$ with $|h| < \min\left\{\delta, \frac{\delta'}{t}\right\}$ \begin{align*} |F'_a(z) - f(z)| &\le \left|F'_a(z) - \int_0^1 f(z+th) \, dt\right| + \left|\int_0^1 f(z+th) \, dt - f(z)\right| \\ &< \frac{\epsilon}{2} + \left|\int_0^1 f(z+th) - f(z) \, dt\right| \le \frac{\epsilon}{2} + \int_0^1 |f(z+th) - f(z)| \, dt \\ &= \frac{\epsilon}{2}+ \int_0^1 \frac{\epsilon}{2} \, dt = \frac{\epsilon}{2} + \frac{\epsilon}{2} = \epsilon. \end{align*}

My problems: 1. I don't know if \begin{align*} \int_{\gamma_{a,z+h}} f(w) \, dw = \int_{\gamma_{a,z}} f(w) \, dw + \int_{[z,z+h]} f(w) \, dw \end{align*} is possible? 2. I don't know how to go on from $\lim_{h \to 0} \int_0^1 f(z+th)\, dt$.

Edit: Is this technical $\epsilon$-$\delta$-argument correct? Maybe some modifications are necessary.

2

There are 2 best solutions below

4
On BEST ANSWER

Maybe the whole approach is wrong.

No, it's exactly right. You can go different ways, but that is the nicest I know.

Concerning question 1., this is technically possible, the premise

Assume that $\int_\gamma f$ only depends on the initial and final points of $\gamma$.

gives you the possibility to choose whatever path you find convenient. You should however explicitly state that $h$ be small, e.g. $\lvert h\rvert < r$, where $r > 0$ is such that the disk with radius $r$ and centre $z$ is contained in $D$.

Concerning question 2., use the continuity of $f$ in $z$ to conclude that for all small enough $h$, you have

$$\left\lvert \int_0^1 f(z+th)\,dt - f(z)\right\rvert < \varepsilon.$$


Regarding the update, part of the task is to show that the limit

$$\lim_{h\to 0} \frac{F_a(z+h)-F_a(z)}{h}$$

exists at all, so you can't split into

$$\left\lvert F_a'(z) - \int_0^1 f(z+th)\,dt\right\rvert + \left\lvert \int_0^1 f(z+th)\,dt - f(z)\right\rvert.$$

However, as you have

$$\frac{F_a(z+h) - F_a(z)}{h} = \int_0^1 f(z+th)\,dt,$$

you show a) the existence of the limit and b) $F_a'(z) = f(z)$ when you show

$$\left\lvert \int_0^1 f(z+th)\,dt - f(z)\right\rvert \to 0.$$

The last part shows that, although there is a small problem with your use of $\frac{\delta'}{t}$, since your bound has to be independent of $t$. But you know that when $\lvert h\rvert < \delta'$, then for every $t \in [0,1]$ you have $\lvert th\rvert = t\lvert h\rvert < \delta'$, so all you need is $\lvert h\rvert < \delta'$ to conclude

$$\left\lvert \int_0^1 f(z+th)\,dt - f(z)\right\rvert = \left\lvert \int_0^1 f(z+th)-f(z)\,dt\right\rvert \leqslant \int_0^1 \lvert f(z+th)-f(z)\rvert\,dt < \int_0^1 \varepsilon\,dt = \varepsilon.$$

1
On

If the integral of the a continuous function depends only on the starting and endpoint, then the function has to be holomorphic. This is an implication of Morera's Theorem. Once you use this fact then undoubtedly your life becomes easier!