Complex frequency

Complex frequency is the variable $s=\sigma +j\omega$ used by the Laplace transform. The imaginary part $\omega$ is the ordinary radian frequency we know from the Fourier transform; the real part $\sigma$ is a new degree of freedom — the “damping rate” added to make the Laplace integral converge for signals the Fourier integral can’t handle.

The pair $(\sigma ,\omega )$ is exactly a point in the s-plane, a two-dimensional complex plane with $\sigma$ horizontal and $\omega$ vertical.

Why “complex”

A pure imaginary $s=j\omega$ corresponds to a pure oscillation $e^{j\omega t}$ — a constant-amplitude rotation in the complex plane. Adding a real part $\sigma$:

$e^{st}=e^{(\sigma +j\omega )t}=e^{\sigma t}\cdot e^{j\omega t}.$

The factor $e^{\sigma t}$ is an exponential envelope:

• $\sigma <0$: envelope decays. The signal is a damped oscillation.
• $\sigma =0$: envelope constant. Pure oscillation.
• $\sigma >0$: envelope grows. Unstable oscillation.

So $s=\sigma +j\omega$ generalizes “frequency” beyond pure rotation to include exponential damping or growth. Same idea as the complex sinusoid $e^{({\sigma }_0+j{\omega }_0)t}$ — just viewed as a transform variable.

What it gives us

By choosing $\sigma$ to compensate for a signal’s growth rate, the Laplace integral converges where the Fourier integral didn’t. For example, $e^{t}u(t)$ has no Fourier transform (the integral diverges at $\infty$), but its Laplace transform is $1/(s-1)$ with ROC $\sigma >1$ — any $s$ with real part greater than 1 makes the kernel $e^{-st}$ decay faster than $e^t$ grows, and the integral converges.

For stable signals (those whose Laplace transform converges on a strip containing the imaginary axis), the Fourier transform exists and is the Laplace transform restricted to the imaginary axis: $X(j\omega )=X(s){|}_{s=j\omega }$. So the Fourier transform is a slice of the Laplace transform along the line $\sigma =0$.

Why we use it

Three reasons:

1. Convergence: handles signals that grow or are bilateral.
2. Initial conditions: the unilateral form naturally absorbs initial conditions into the algebra.
3. Algebraic structure: transfer functions of LTI systems are rational functions in $s$ — easy to manipulate, factor, decompose, multiply. This is the practical reason Laplace dominates engineering analysis.

The cost: $s$ is a complex variable, and you have to keep track of the ROC for the bilateral form. For most engineering work, the unilateral form (where ROC is implicit) is what gets used.

A note on $j$ vs $s$

In some texts (especially older ones), the Laplace variable is written as $p$ or $z$. The lowercase $s$ is the modern convention. Don’t confuse it with the discrete-time $z$ of the $z$-transform — those are related but distinct objects, with $z=e^{sT}$ where $T$ is the sampling period.