Two real functions travelling together: limits, continuity, the Cauchy–Riemann equations, and harmonic functions
On completion of this section, you should be able to:
A complex function takes a point of the plane and returns another point of the plane. That means it is really two real functions travelling together — and the whole subject turns on what happens when those two are forced to cooperate.
A function $f$ assigns to each $z$ in a set $S$ a single complex number $w = f(z)$. Writing $z = x + iy$ and separating the output into real and imaginary parts:
$u$ and $v$ are real-valued functions of the two real variables $x$ and $y$.
Find $u$ and $v$ for $f(z) = z^2$.
So $u(x,y) = x^2 - y^2$ and $v(x,y) = 2xy$.
Find $u$ and $v$ for $f(z) = \dfrac{1}{z}$, $z \ne 0$.
Multiply above and below by the conjugate:
So $u = \dfrac{x}{x^2+y^2}$ and $v = \dfrac{-y}{x^2+y^2}$. Note the minus sign in $v$, and that neither is defined at the origin.
The graph of $w = f(z)$ would need four dimensions: two for the input, two for the output. So we never graph a complex function. Instead we think of it as a mapping: it picks up a region of one plane and lays it down, stretched and rotated, on another.
$\displaystyle\lim_{z \to z_0} f(z) = L$ means: $|f(z) - L|$ can be made as small as we like by taking $|z - z_0|$ small enough (but nonzero).
On the real line, $z$ can approach $z_0$ from two directions. In the plane it can approach from infinitely many — along any line, any spiral, any path at all. For the limit to exist, every single approach must give the same answer. This makes limits much harder to have, and it is the source of all the power later in the chapter.
Show that $\displaystyle\lim_{z \to 0} \frac{\bar{z}}{z}$ does not exist.
Approach along the real axis ($z = x$, so $\bar z = x$):
Approach along the imaginary axis ($z = iy$, so $\bar z = -iy$):
Two approaches, two different answers, so the limit does not exist. One counterexample path pair is enough — you never have to check them all.
$f$ is continuous at $z_0$ if $f(z_0)$ is defined, the limit exists, and $\displaystyle\lim_{z \to z_0} f(z) = f(z_0)$. Equivalently, $f$ is continuous exactly when both $u$ and $v$ are continuous as real functions of $(x,y)$.
Sums, products, quotients (where the denominator is nonzero) and compositions of continuous functions are continuous, exactly as in real calculus. Every polynomial in $z$ is continuous everywhere.
provided this limit exists — meaning it gives the same value however $\Delta z \to 0$.
That last clause is a severe demand. Let $\Delta z$ approach $0$ first horizontally and then vertically, and insist the two answers agree, and you are forced into a pair of partial differential equations.
If $f = u + iv$ is differentiable at $z_0$, then at that point
and the derivative is then $\;f'(z) = \dfrac{\partial u}{\partial x} + i\dfrac{\partial v}{\partial x}$.
Cauchy–Riemann is necessary, not by itself sufficient. Satisfying the equations at a point does not guarantee differentiability there. Add the hypothesis that the four partial derivatives are continuous in a neighbourhood, and then it does: that is the converse theorem, and it is the version you use in practice.
$f$ is analytic at $z_0$ if it is differentiable at every point of some open disc around $z_0$ — differentiable at the point is not enough, it must be differentiable near it.
$f$ is analytic in a domain $D$ if it is analytic at every point of $D$, and entire if it is analytic on the whole plane.
Show that $f(z) = z^2$ is entire, and find $f'(z)$.
$u = x^2 - y^2$, $v = 2xy$. The partials are
Check: $u_x = 2x = v_y$ ✓ and $u_y = -2y = -v_x$ ✓, at every point, and all four partials are continuous everywhere. So $f$ is entire, and
which is the answer real calculus would have guessed.
Show that $f(z) = \bar{z}$ is differentiable nowhere.
$\bar z = x - iy$, so $u = x$ and $v = -y$.
The first Cauchy–Riemann equation demands $u_x = v_y$, i.e. $1 = -1$, which is false everywhere. There is no point at which $\bar z$ is differentiable.
Yet $\bar z$ is perfectly continuous. In the complex world, continuity is cheap and differentiability is expensive.
Where is $f(z) = |z|^2$ differentiable? Where is it analytic?
$|z|^2 = x^2 + y^2$, so $u = x^2 + y^2$ and $v = 0$.
Both equations hold only at the single point $z = 0$, so that is the only place $f$ can be differentiable (and it is).
But analyticity needs differentiability on a whole disc around the point, and every disc around $0$ contains points where Cauchy–Riemann fails. So $f$ is differentiable at one point and analytic nowhere. This example is the reason the definition of "analytic" insists on a neighbourhood.
A real function $\phi(x,y)$ with continuous second partials is harmonic in a domain $D$ if it satisfies Laplace's equation there:
You met this equation in Chapter 6 as the steady-state heat equation: the temperature distribution that has stopped changing.
If $f = u + iv$ is analytic in a domain $D$, then both $u$ and $v$ are harmonic in $D$.
The proof is two lines of Cauchy–Riemann. Differentiate $u_x = v_y$ with respect to $x$, and $u_y = -v_x$ with respect to $y$:
Adding, and using the equality of mixed partials $v_{xy} = v_{yx}$:
If $u$ and $v$ are both harmonic in $D$ and satisfy the Cauchy–Riemann equations, then $v$ is called a harmonic conjugate of $u$, and $f = u + iv$ is analytic in $D$.
Show $u = x^3 - 3xy^2$ is harmonic, and find a harmonic conjugate.
Harmonic? $u_x = 3x^2 - 3y^2$, so $u_{xx} = 6x$. And $u_y = -6xy$, so $u_{yy} = -6x$.
Find $v$. The first Cauchy–Riemann equation gives
Integrate with respect to $y$, remembering the constant of integration is a function of $x$:
Now use the second equation, $v_x = -u_y = 6xy$:
And the function we have built is an old friend: $u + iv = x^3 - 3xy^2 + i(3x^2y - y^3) = z^3$.
Take $u = x^2 + y^2$. Then $u_{xx} + u_{yy} = 2 + 2 = 4 \ne 0$. It is smooth, symmetric and perfectly pleasant — and it is not harmonic, so it is not the real part of any analytic function. Compare $x^2 - y^2$, where the signs do cancel.
Click a problem to open the full solution.
Step 1: Substitute $z = x + iy$.
$$z^2 + 3z = \left(x^2 - y^2 + 2ixy\right) + 3(x + iy)$$Step 2: Collect real and imaginary parts.
$$= \left(x^2 - y^2 + 3x\right) + i\left(2xy + 3y\right)$$ $$\boxed{u = x^2 - y^2 + 3x, \qquad v = 2xy + 3y}$$Step 3: Check with Cauchy–Riemann. $u_x = 2x + 3$ and $v_y = 2x + 3$ ✓; $u_y = -2y$ and $-v_x = -2y$ ✓. The function is entire, with $f'(z) = (2x+3) + i(2y) = 2z + 3$.
Step 1: Read off $u$ and $v$. $u = e^{-y}\cos x$, $v = e^{-y}\sin x$.
Step 2: The four partials.
$$u_x = -e^{-y}\sin x, \qquad u_y = -e^{-y}\cos x$$ $$v_x = e^{-y}\cos x, \qquad v_y = -e^{-y}\sin x$$Step 3: Test. $u_x = -e^{-y}\sin x = v_y$ ✓ and $u_y = -e^{-y}\cos x = -v_x$ ✓, everywhere, with all partials continuous. So $f$ is entire.
Step 4: The derivative.
$$f'(z) = u_x + iv_x = -e^{-y}\sin x + ie^{-y}\cos x = i\,e^{-y}(\cos x + i \sin x) = i f(z)$$Along the real axis ($z = x$): $\dfrac{\operatorname{Re}(x)}{x} = \dfrac{x}{x} = 1$.
Along the imaginary axis ($z = iy$): $\operatorname{Re}(iy) = 0$, so the quotient is $\dfrac{0}{iy} = 0$.
$$\boxed{\text{The limit does not exist}}$$
These problems are similar in style and difficulty to past exam questions. Click each problem to reveal the step-by-step solution.
Step 1: Laplace's equation.
$$u_x = e^x\cos y, \quad u_{xx} = e^x\cos y, \quad u_y = -e^x\sin y, \quad u_{yy} = -e^x\cos y$$ $$u_{xx} + u_{yy} = e^x\cos y - e^x\cos y = 0 \;\checkmark$$Step 2: Build $v$ from the first CR equation. $v_y = u_x = e^x\cos y$, so integrating in $y$:
$$v = e^x \sin y + h(x)$$Step 3: Use the second CR equation. $v_x = -u_y = e^x\sin y$. But from Step 2, $v_x = e^x\sin y + h'(x)$, so $h'(x) = 0$ and $h$ is a constant.
Step 4: Assemble.
$$f = e^x\cos y + i\,e^x\sin y = e^x(\cos y + i \sin y) = e^x e^{iy}$$ $$\boxed{f(z) = e^{z}}$$Step 1: Second partials.
$$u_x = 2ax + 2y \;\Longrightarrow\; u_{xx} = 2a$$ $$u_y = -2y + 2x \;\Longrightarrow\; u_{yy} = -2$$Step 2: Impose Laplace.
$$u_{xx} + u_{yy} = 2a - 2 = 0 \;\Longrightarrow\; \boxed{a = 1}$$Step 3: Check what we get. With $a = 1$, $u = x^2 - y^2 + 2xy$. The $x^2 - y^2$ part is $\operatorname{Re}(z^2)$ and the $2xy$ part is $\operatorname{Im}(z^2)$, so
$$u = \operatorname{Re}\left(z^2\right) + \operatorname{Im}\left(z^2\right) = \operatorname{Re}\left((1 - i)z^2\right)$$which confirms it is the real part of an analytic function.
Step 1: Simplify first. $z\bar z = |z|^2 = x^2 + y^2$, which is entirely real, so $u = x^2 + y^2$ and $v = 0$.
Step 2: First CR equation.
$$u_x = 2x, \quad v_y = 0 \;\Longrightarrow\; 2x = 0 \;\Longrightarrow\; x = 0$$Step 3: Second CR equation.
$$u_y = 2y, \quad -v_x = 0 \;\Longrightarrow\; 2y = 0 \;\Longrightarrow\; y = 0$$Step 4: Conclude. Both hold only at $z = 0$. Since the partials are continuous there, $f$ is differentiable at the origin, with $f'(0) = u_x + iv_x = 0$.
$$\boxed{\text{Differentiable only at } z = 0, \text{ analytic nowhere}}$$Find $u$ and $v$ for $f(z) = z^3$.
$(x+iy)^3 = x^3 + 3ix^2y + 3i^2xy^2 + i^3y^3 = \left(x^3 - 3xy^2\right) + i\left(3x^2y - y^3\right)$.
Verify the Cauchy–Riemann equations for $f(z) = z^3$ and find $f'(z)$.
$u_x = 3x^2 - 3y^2$ and $v_y = 3x^2 - 3y^2$ ✓. $u_y = -6xy$ and $-v_x = -6xy$ ✓.
Is $v = 2xy$ a harmonic conjugate of $u = x^2 - y^2$?
Both are harmonic ($2 - 2 = 0$ and $0 + 0 = 0$), and $u_x = 2x = v_y$ ✓ with $u_y = -2y = -v_x$ ✓.
Yes — and $u + iv = z^2$. Note the order matters: $u$ is not a harmonic conjugate of $v$, because $v + iu$ is not analytic.
Test your understanding with these multiple-choice questions. Click a choice to see if you're correct.
Answer: (B). $(x+iy)^2 = x^2 - y^2 + 2ixy$, and the $i^2$ makes the $y^2$ term negative, which rules out (A). Option (C) is the imaginary part $v$, and (D) attaches the $3$ to the wrong variable: $3z = 3x + 3iy$ contributes $3x$ to $u$.
Answer: (D). Along the real axis the quotient is $x/x = 1$; along the imaginary axis it is $-iy/iy = -1$. Options (A) and (B) are each one of those routes mistaken for the whole answer. In the plane every approach must agree, and these two do not.
Answer: (B). With $u = x^2 + y^2$ and $v = 0$, the equations $u_x = v_y$ and $u_y = -v_x$ force $x = 0$ and $y = 0$. It is differentiable at that single point but analytic nowhere, since analyticity needs a whole disc — which is exactly why (C) is wrong even though the function looks smooth.
Answer: (C). Exactly one minus sign, and it sits on the second equation. Test any option against $f(z) = z^2$, where $u_x = 2x$, $u_y = -2y$, $v_x = 2y$, $v_y = 2x$: only (C) survives.
Answer: (B). $u_{xx} + u_{yy} = 2 + (-2) = 0$. For (A) the Laplacian is $2 + 2 = 4$; for (C) it is $2y^2 + 2x^2$; for (D) it is $2e^{x+y}$. Only in (B) do the two second derivatives cancel — the minus sign is doing all the work.
Answer: (A). Integrating $v_y = u_x = 3x^2 - 3y^2$ gives $v = 3x^2y - y^3 + h(x)$, and $v_x = -u_y = 6xy$ forces $h' = 0$. The resulting function is $u + iv = z^3$. Option (D) is the negative of the right answer, which satisfies the equations with the signs swapped — that would make $u$ a conjugate of $v$, not the other way round.
You now know what makes a complex function well behaved. The next section applies that machinery to the most important function in the subject, $e^z$ — and then tries to invert it, which turns out to be far more interesting than inverting $e^x$ ever was.