# How do I prove that $f'(z)=0$ implies $f$ is constant?

Let $V$ be an open connected subset of $\mathbb{C}$.

Let $f:V\rightarrow\mathbb{C}$ be a function whose derivative is $0$ on $V$.

How do I prove that $f$ is a constant on $V$?

I know that $V$ is path-connected, but I don't know whether this helps.

• Can you see the proof for a real-valued function on a real interval $(a,b)$, that if $f'(x)=0$ at every point in that interval, then $f$ is constant there? Apr 13, 2014 at 3:47
• @hardmath Yes sure. I know an argument by Fermat for that one.. Apr 13, 2014 at 3:57

Use the Cauchy-Riemann equations:

$$f(x + iy) = u(x, y) + iv(x, y)$$

Where $u$ and $v$ are real valued functions. We have the constraints:

$$\frac{\partial u}{\partial x} = \frac{\partial v }{\partial y} \\ \frac{\partial u}{\partial y} = -\frac{\partial v }{\partial x}$$

Giving:

$$f'(x + iy) = \frac{\partial u}{\partial x} + i\frac{\partial v}{\partial x}$$

If $f' = 0$ then we have that $\frac{\partial u}{\partial x} = \frac{\partial v}{\partial x} = 0$. This then entails that $\frac{\partial v}{\partial y} = \frac{\partial u}{\partial y} = 0$. If $\frac{\partial u}{\partial x} = 0$ then $u = \alpha(y)$. If you do all of them then you find that $u = \alpha(y) = \beta(x)$ and $v = \gamma(y) = \varphi(x)$. The only way that $\alpha(y) = \beta(x)$ for all values of $x$ and $y$ is if $\alpha$ and $\beta$ have no dependence on either $x$ or $y$ and thus are constants. The same follows for $v$. Therefore $u = A$ and $v = B$ and $f(z) = A + Bi$.

• What are g and h? Apr 13, 2014 at 3:46
• I think you mistakenly reused the name $f$. Apr 13, 2014 at 3:46
• Yes, I did, ran out of variables (I already didn't like using $j$). Apr 13, 2014 at 3:47
• @user140374 The $\alpha$, $\beta$, $\gamma$, and $\varphi$ are just some functions of single variables. Since the partial is $0$, the original multi-variable function must actually have been a function of a single variable. They must be equal so that $\alpha(y) = \beta(x)$. Let's choose a value of $x = x_0$ and $y = y_0$, then we must have $\alpha(y_0) = \beta(x_0)$. What if we change $y$? Then $\alpha(y_0 + h) \neq \alpha(y_0)$ unless $\alpha(y_0 + h) = \alpha(y_0) = A$, some constant. But if they aren't equal, then $\alpha(y_0 + h) \neq \beta(x_0)$, a contradiction. Apr 13, 2014 at 3:54
• @Jared I don't get it. Where your argument used 'connectedness'? Apr 13, 2014 at 4:01

Consider a point $z_0 \in V$. Now consider any other point $w \in V$. Since $V$ is connected, it is possible to find a curve starting at $z_0$ and ending at $w$. Let's call this curve $C_w$. We then have $$f(w) - f(z_0) = \int_{C_w} f'(z) dz = \int_{C_w} 0 \cdot dz = 0$$ Hence, we have $f(w) = f(z_0)$. This is true for any $w \in V$. Hence, $f(w) = f(z_0)$ and hence $f(w)$ is a constant in $V$.

• wait but, to clarify, you're assuming path connected instead of connected?
– BCLC
Oct 23, 2021 at 17:10
• wait never mind i remember now: open and connected subset of Rn implies path connected: math.stackexchange.com/questions/1088685/…
– BCLC
Oct 23, 2021 at 17:18
• Do we need simply connected domain instead of open, connected domain for the above integral to be well-defined? Feb 18, 2022 at 14:12

Hint: Choose a path $\gamma$ connecting two given points in $V$, and evaluate

$$\int_{\gamma} f'(z) dz$$

two different ways.

• I haven't learned line integral though.. Isn't there a way to prove this not using line integral? Apr 13, 2014 at 3:39
• @user140374: Have you learned the regular multivariable line integral? You can just treat the function as a pair of real-valued functions with two real-valued inputs. Apr 13, 2014 at 3:42
• @user2357112 No, i haven't. Apr 13, 2014 at 3:43
• @user140374 In that case, what have you learned? (Please note that adding additional details to the original question when asking will help a lot in getting answers that are actually relevant to you.)
– user61527
Apr 13, 2014 at 3:43
• @user140374: Huh. You're doing complex analysis before multivariable calculus? Weird. Apr 13, 2014 at 3:44

You can connect any point to a point $z_0$ by a polygonal line. Along each segment you can write $f(z) = f(z_i + t (z_f - z_i))$, and write this as two real functions of the real variable $t$, one for the real part and one for the imaginary part. Write the derivative in terms of them, you see both components are zero along the segment, so both components are constant and the function itself is constant on that segment, and thus over each segment of the path, and so over the full region.

• I succeed in proving the theorem in this way. Thank you. BTW, how do you define polygonal line precisely? I rather showed that given $\alpha,\beta$, there exist finite points $\alpha=z_0,\cdots,z_n=\beta$ such that $z_i-z_{i+1}$ is either purely imaginary or purely real, and the line segment connecting these two are in $V$, then apply Cauchy-Riemann equation. This formulation is precise but really messy when written in meta-language. Apr 13, 2014 at 6:02
• I know it is a mess to write up precisely, but it is easy to explain. BTW, straight lines between $z_i$ to $z_f$ are just $z=z_i + (z_f -z_i) t$ with real $t$ between 0 and 1. And my definition of a connected region if based on polygonals connecting any pairof points Apr 13, 2014 at 12:39