Skip to content

Calculus for Teachers: Complex Differentiation

By David Bressoud @dbressoud


As of 2024, new Launchings columns appear on the third Tuesday of the month.

It was Cauchy in 1821 who first began studying complex-valued functions defined on the complex plane. He was actually sitting on almost 70 years of what amounted to research into these functions. The input defines a point in a two-dimensional plane. The output can be thought of as a two-dimensional vector describing laminar or two-dimensional flow. Complex functions are most usefully thought of as describing two-dimensional flow across a plane. Jean Le Rond d'Alembert in 1752 was the first person to seriously study such flow, actually using real and imaginary numbers to locate points in the plane and to label the two orthogonal directions of flow. The step he was not able to make was to unite the real and imaginary values into a single complex number. That would have to wait for Cauchy. I will leap directly into the use of complex numbers.

Jean le Rond d'Alembert 1717--1783.
$$

Let $z = x+iy$ denote the input into a complex-valued function $f$, and let $f(z) = u(z) + iv(z)$ denote the output. Letting $\delta$ denote a small change in the real direction (positive or negative), then the derivative of $f$ at $z_0 = x_0 + iy_0$ must satisfy \begin{eqnarray*} f'(z_0) & = & \lim_{\delta \to 0} \frac{u(x_0+\delta + iy_0) +iv(x_0+\delta + iy_0)- u(x_0+iy_0) - iv(x_0+iy_0)}{\delta} \\ & = & \lim_{\delta \to 0} \frac{u(x_0+\delta + iy_0) - u(x_0+iy_0) }{\delta} + \lim_{\delta \to 0} i \frac{v(x_0+\delta + iy_0)- v(x_0+iy_0)}{\delta} \\ & = & \left.\frac{\partial u}{\partial x}\right|_{x_0 + i y_0} + i \left.\frac{\partial v}{\partial x}\right|_{x_0+iy_0} \\ & & {\rm or\ more \ simply}\\ \frac{df}{dz} & = & \frac{\partial u}{\partial x} + i \frac{\partial v}{\partial x}. \end{eqnarray*}

$$
$$

But we need to get the same result if we approach $z_0 = x_0 + iy_0$ from the imaginary direction.

$$
$$

\begin{eqnarray*} f'(z_0) & = & \lim_{\delta \to 0} \frac{u(x_0+\ iy_0 + i\delta) +iv(x_0 + iy_0)- u(x_0+iy_0+ i\delta) - iv(x_0+iy_0)}{i\delta} \\ & = & \lim_{\delta \to 0} \frac{u(x_0+ iy_0 + i\delta) - u(x_0+iy_0) }{i\delta} + \lim_{\delta \to 0} i \frac{v(x_0 + iy_0+ i\delta)- v(x_0+iy_0)}{i\delta} \\ & = & \frac{1}{i}\left.\frac{\partial u}{\partial y}\right|_{x_0 + i y_0} + \left.\frac{\partial v}{\partial y}\right|_{x_0+iy_0} \\ & & {\rm or\ more \ simply}\\ \frac{df}{dz} & = & -i\frac{\partial u}{\partial y} + \frac{\partial v}{\partial y}. \end{eqnarray*}

$$
$$

This leads us to what are known as the Cauchy-Riemann equations, first stated by d'Alembert in 1752, long before Cauchy or Riemann were born. Cauchy observed these conditions as a desirable property of a complex-valued function of a complex variable, but it was Riemann in his doctoral thesis of 1851 who built the study of complex analysis around the requirement of differentiability.

$$
$$

In order for the derivative of $f$ to exist at a point $z$, it is necessary that $$ \frac{\partial u}{\partial x} + i \frac{\partial v}{\partial x} = \frac{\partial v}{\partial y} - i \frac{\partial u}{\partial y} $$ which implies that $$ \frac{\partial u}{\partial x} = \frac{\partial v}{\partial y} \quad {\rm and} \quad \frac{\partial v}{\partial x} = -\frac{\partial u}{\partial y} . $$ More simply, this condition can be stated as $$ i\frac{\partial f}{\partial x} = \frac{\partial f}{\partial y}. $$

$$
$$

This is a necessary condition. It is not quite sufficient because we must guarantee that we get the same limit no matter from which direction we approach $z_0 = x_0 + iy_0$. I will spare you the details, but it is enough that the partial derivatives of $u$ and $v$ exist and are continuous in some open set containing $z_0$. (Recall that a set is open if every element of the set is the center of an open disc---which could be extremely small but must have positive radius---contained within the set.)

$$

In the complex plane, differentiability guarantees that the flow is smooth. For functions on the real line, a function is differentiable if the relationship between input and output is essentially linear at very small scale. The same thing is true for differentiable complex functions, but now we are looking at two-dimensional change. At very small scale, doubling the change in the location will essentially double the change in the direction and speed of the flow. A function that is differentiable at all points in an open set is said to be analytic in that set.

$$

So what functions are analytic? We start with $z^n$ for a positive integer $n$, $$ f(z) = z^n = (x+iy)^n = \sum_{j=0}^n \frac{n!}{ j!\,(n-j)!} x^j i^{n-j} y^{n-j}. $$ The two partial derivatives we need to examine are \begin{eqnarray*} i\frac{\partial f}{\partial x} & = & i \sum_{j=0}^n \frac{n!}{ j!\,(n-j)!} \, j \, x^{j-1} i^{n-j} y^{n-j}\\ & = & \sum_{j=1}^n \frac{n!}{ (j-1)!\,(n-j)!}\, x^{j-1} i^{n-j+1} y^{n-j} \\ & = & \sum_{j=0}^{n-1} \frac{n!}{ j!\,(n-j-1)!}\, x^j i^{n-j} y^{n-j-1}, \end{eqnarray*} where we have replaced $j$ by $j+1$ in the last line, and \begin{eqnarray*} \frac{\partial f}{\partial y} & = & \sum_{j=0}^n \frac{n!}{ j!\,(n-j)!}\, x^j\, i^{n-j} (n-j) y^{n-j-1}\\ & = & \sum_{j=0}^{n-1} \frac{n!}{ j!\,(n-j-1)!}\, x^j\, i^{n-j} y^{n-j-1}. \end{eqnarray*}

$$
$$

Since the partial derivatives are clearly continuous, the function defined by $f(z) = z^n$ is differentiable. Furthermore, since the partial derivatives satisfied the required equation, we can find the derivative by taking the partial derivative with respect to $x$: \begin{eqnarray*} \frac{df}{dz}\ = \ \frac{\partial f}{\partial x} & = & \sum_{j=0}^{n-1} \frac{n!}{ j!\,(n-j-1)!} \,x^j\, i^{n-j-1} y^{n-j-1}\\ & = & n \sum_{j=0}^{n-1} \frac{(n-1)!}{ j!\,(n-j-1)!} \,x^j\, i^{n-j-1} y^{n-j-1} \\ & = & n(x+iy)^{n-1}\ =\ nz^{n-1}. \end{eqnarray*} Whew. We would have been in trouble if it had been anything else.

$$

Now, the same arguments that worked for real-valued functions on the real line show that the derivative of a sum is the sum of the derivatives and the derivative of a constant times f is that constant times the derivative of f, where now the constant can be any complex number. This tells us that all polynomials in z with complex coefficients are analytic and the derivative is what we would expect. Furthermore, with just a bit more work, it can be shown that any power series in z that converges on an open set is differentiable, and both the power series and its derivative have the same radius of convergence. In particular, the exponential, sine, and cosine functions can be defined by their power series and so are differentiable over the entire complex plane. But now magic happens. Polynomials and power series are the only analytic functions on the complex plane. Demonstrating why that is true will take a lot more effort and will be saved for future columns.


David Bressoud is DeWitt Wallace Professor Emeritus at Macalester College and former Director of the Conference Board of the Mathematical Sciences. Information about him and his publications can be found at davidbressoud.org

Download a Word file containing the list of all past Launchings columns, dating back to 2005, with links to each column.