Laplace transform — definition, region of convergence: Visual walkthrough

We assume nothing except: you can add up areas under a curve (that is all an integral is), and you know $2^{3} = 8$ style powers. Every symbol is earned before use.

Step 1 — What "a function of time" even means

WHAT we are looking at: a plain height-versus-time graph.

WHY we start here: the whole Laplace idea is going to reshape this curve, so we must first know what curve we are reshaping. Our running example will be $f (t) = e^{0.5 t}$ — a curve that grows as time goes on. Grab that word "grows"; it is the villain of the story, and the number $0.5$ sitting in the exponent is exactly the growth-rate we will later christen $a$ .

PICTURE — the blue curve climbs forever. That climbing is the problem: if we tried to sum its area all the way to infinity, we would get an infinite answer.

Figure — Laplace transform — definition, region of convergence

Step 2 — The fade-out filter $e^{- σ t}$

WHAT we did: introduced a second curve, the orange decaying exponential $e^{- σ t}$ .

WHY this tool and not another? We need something that (a) starts at $1$ so it doesn't distort the early signal, (b) shrinks toward zero so it can tame growth, and (c) has the property $\frac{d}{d t} e^{- σ t} = - σ e^{- σ t}$ — differentiating just multiplies by $- σ$ . No polynomial or sine does all three. The exponential is the unique tool that turns calculus into algebra, which is the entire payoff (see Laplace Transform of Derivatives).

PICTURE — orange starts at height $1$ and slides down. A bigger $σ$ means a steeper orange slide.

Step 3 — Multiply them: the tug-of-war

WHAT we did: combined the two exponentials by adding their powers, $- σ t + 0.5 t = (0.5 - σ) t$ .

WHY we combine: a single exponent tells us the net behaviour at a glance. The sign of $(0.5 - σ)$ decides everything:

If $σ > 0.5$ → exponent is negative → product decays → area is finite. ✓
If $σ < 0.5$ → exponent is positive → product blows up → area is infinite. ✗
If $σ = 0.5$ → exponent is $0$ → product is the constant $1$ forever → area still infinite. ✗ (the razor's edge)

PICTURE — three products drawn together: green ( $σ$ large, decays), gray (borderline, flat), red ( $σ$ small, explodes). This single picture is the ROC condition.

Step 4 — The integral is the area under the winning curve

WHAT we did: shaded the running area under the product from Step 3 and drew where its total heads as the right edge $T$ marches off to infinity.

WHY an integral answers our question: we want one number that summarises the faded signal for a given fade-speed $σ$ . Adding all the faded heights (the area) and taking the $T \to \infty$ limit is exactly that number.

PICTURE — the green (converging) case: the running area up to $T$ rises but flattens toward a finite ceiling as $T$ grows. The red case: the running area keeps climbing without bound — no limit, no answer.

Step 5 — Compute one honestly: $f (t) = e^{a t}$

Here the growth-rate parameter earns its name: $a$ is the number in the signal's exponent, so $f (t) = e^{a t}$ grows at rate $a$ (for our running example $a = 0.5$ ).

Term by term:

$e^{- σ t} e^{a t} = e^{- (σ - a) t}$ — combine exponents ( $- σ t + a t = - (σ - a) t$ ), same trick as Step 3.
$\int_{0}^{T} e^{- (σ - a) t} d t = [\frac{- 1}{σ - a} e^{- (σ - a) t}]_{0}^{T} = \frac{1}{σ - a} (1 - e^{- (σ - a) T})$ — the antiderivative of an exponential is itself over its rate, evaluated at the two ends.
Take the limit $T \to \infty$ : everything hinges on $e^{- (σ - a) T}$ .
- If $σ - a > 0$ the exponent $- (σ - a) T \to - \infty$ , so $e^{- (σ - a) T} \to 0$ , and the bracket $\to 1$ . Finite. ✓
- If $σ - a < 0$ the exponent $\to + \infty$ , so $e^{- (σ - a) T} \to \infty$ — no limit, diverges. ✗
- If $σ - a = 0$ the integrand is the constant $1$ , so $\int_{0}^{T} 1 d t = T \to \infty$ — diverges. ✗
Therefore, valid exactly when $σ > a$ : $F (σ) = \frac{1}{σ - a} (1 - 0) = \frac{1}{σ - a} .$

WHAT we did: evaluated the improper integral as a genuine limit, watching the tail term $e^{- (σ - a) T}$ decide the outcome.

WHY it matters: the growth rate $a$ literally becomes the wall of the ROC. Set $a = 0$ and you recover $L {1} = \frac{1}{σ}$ with ROC $σ > 0$ . (Once we allow complex $s$ in Step 7, this same wall becomes the line $Re (s) > a$ — but the arithmetic here needs nothing complex.)

PICTURE — the number line of $σ$ with a vertical wall at $σ = a$ ; everything to the right is "allowed."

Step 6 — The general bound: why it is always a half-line (soon: half-plane)

Term by term:

$∣ F (σ) ∣ \leq \int e^{- σ t} ∣ f (t) ∣ d t$ — the size of a sum is at most the sum of sizes, and here the fade $e^{- σ t}$ is a positive real number.
$∣ f (t) ∣ \leq M e^{a t}$ — the exponential-order cap from the definition just above; $a$ is the growth rate, $M$ the constant cap.
$\int_{0}^{\infty} e^{- (σ - a) t} d t = \frac{1}{σ - a}$ — same limit computation as Step 5.
$\frac{M}{σ - a}$ is a finite number exactly when $σ > a$ .

WHAT we did: bounded any well-behaved $f$ , not just exponentials — still using only the real knob $σ$ .

WHY: this proves the allowed set is the half-line $σ > a$ for every signal of exponential order, no case-by-case work needed. The single denominator $σ - a$ carries the whole conclusion. In Step 7 we let $s$ go complex and this half-line fattens into a half-plane.

PICTURE — the ceiling curve $M e^{a t}$ sitting above a wiggly $f (t)$ ; below it, the faded area is guaranteed finite once $σ > a$ .

Step 7 — Now let $s$ go complex: only $Re (s)$ matters

Everything so far treated $s$ as the real number $σ$ . The real definition of the Laplace transform allows $s$ to be complex, so let us upgrade it now and check the earlier steps still hold.

WHAT we did: peeled $s$ into a real part $σ$ and an imaginary part $ω$ , and split the kernel to match.

WHY: because $∣ e^{- iω t} ∣ = 1$ , the spinning part never changes the area's size — it just rotates. Convergence is decided entirely by $σ = Re (s)$ — the very number our Steps 5–6 inequalities were about. So those steps were never "wrong to use $σ$ ": $σ$ is $Re (s)$ . The condition $σ > a$ simply reads $Re (s) > a$ , and since $ω$ is free, the half-line on the number line becomes a half-plane in the complex $s$ -plane.

PICTURE — the complex $s$ -plane: horizontal axis $σ$ , vertical axis $ω$ ; a vertical line at $σ = a$ , green shading (converges) to its right, red (diverges) to its left. The imaginary $ω$ direction is "free."

Definition Which "convergence" we demand — absolute, not merely conditional

Two notions live here. Absolute convergence means $\int_{0}^{\infty} ∣ e^{- s t} f (t) ∣ d t = \int_{0}^{\infty} e^{- σ t} ∣ f (t) ∣ d t$ is finite — the sizes add up. Conditional convergence means the signed/oscillating integral $\int_{0}^{\infty} e^{- iω t} f (t) e^{- σ t} d t$ might settle to a limit thanks to cancellation from the spin $e^{- iω t}$ , even when the sizes do not add up. On the exact boundary $Re (s) = a$ the size-integral $\frac{M}{σ - a}$ blows up, so absolute convergence fails there. A cleverly oscillating $f$ could sometimes make the integral converge conditionally on that boundary line — but the Laplace transform is defined via absolute convergence, because that is what guarantees the algebra (differentiation-under-the-integral, swapping sums and integrals) used everywhere later is legal. That is exactly why we insist on the strict inequality $Re (s) > a$ and treat $Re (s) = a$ as outside the ROC.

Step 8 — The degenerate case that has no transform

WHAT we did: raced $t^{2}$ against $σ t$ .

WHY it matters: no matter how large you crank the fade $σ$ , once $t > σ$ the exponent $t^{2} - σ t$ is positive and climbing. The product explodes, the area is infinite, the ROC is empty — $e^{t^{2}}$ has no Laplace transform. This is the boundary of the whole theory: exponential order is not a technicality, it is the entry ticket.

PICTURE — even the steepest orange fade cannot flatten the $e^{t^{2}}$ curve; the product still curls upward.

Step 9 — The opposite extreme: fast decay converges everywhere useful

WHAT we did: checked what happens when $f$ decays super-fast on its own.

WHY it matters: it nails down that the ROC rule $σ > a$ holds even when $a < 0$ . For a rapidly decaying signal the wall $a$ slides far to the left, so the ROC covers the whole plane you would ever care about — you can even fade backwards ( $σ < 0$ ) and still get a finite area, because the signal was already collapsing to zero. Contrast Step 8 (wall at $+ \infty$ , ROC empty) with this (wall at $- \infty$ , ROC everywhere). The single rule $Re (s) > a$ spans both extremes.

PICTURE — the blue $e^{- t^{2}}$ bump already hugs zero; multiplying by any fade (even a mild backwards-growing one) still leaves a tiny finite area.

The one-picture summary

Everything above compressed: a signal is capped by an exponential ceiling $M e^{a t}$ ; the fade $e^{- σ t}$ fights that ceiling; the fight is won (finite area, transform exists) exactly in the right half-plane $Re (s) > a$ , and lost to the left — while the vertical $ω$ direction is irrelevant. The wall $a$ can sit anywhere: at $+ \infty$ (no transform, like $e^{t^{2}}$ ), at a finite value (like $e^{a t}$ ), or at $- \infty$ (converges everywhere, like $e^{- t^{2}}$ ).

Recall Feynman retelling — the whole walkthrough in plain words

Picture a signal as a sound that might get louder over time. You put on magic headphones with a "fade knob" $σ$ — a plain real number: the bigger you turn it, the faster every sound fades. You then add up all the faded sound into a single loudness number — but "add up forever" really means: add up to some stopping-time $T$ , then push $T$ out and see if the running total settles. If the original sound doesn't grow too wildly (it stays under some exponential ceiling $M e^{a t}$ ), there's a fade speed fast enough to make the total settle: any $σ$ bigger than the sound's own growth rate $a$ works. Turn the knob too gently ( $σ < a$ ) and the sound outruns the fade — the total never settles. There's also a second knob $ω$ that only makes the sound spin, never louder, so it never affects whether the total is finite — that's why the answer is a right half-plane, not just a half-line, and we insist the sizes add up (absolute convergence), which is why the razor's-edge line $Re (s) = a$ is left out. Two extremes pin it down: a sound growing insanely fast like $e^{t^{2}}$ can't be caught by any fade at all (no transform), while a sound that's already collapsing like $e^{- t^{2}}$ works for every fade speed, even negative ones.

Recall Rapid self-check

Wall of the ROC for $L {e^{3 t}}$ ? ::: $Re (s) > 3$ . Which part of $s = σ + iω$ controls convergence? ::: only $σ = Re (s)$ . Why does $e^{- iω t}$ not affect the ROC? ::: its magnitude is always $1$ ; it only spins. What does $\int_{0}^{\infty}$ actually mean? ::: $lim_{T \to \infty} \int_{0}^{T}$ — the running area's limit. Why strict $Re (s) > a$ and not $\geq$ ? ::: we demand absolute convergence, which fails on the boundary $Re (s) = a$ . Why does $e^{t^{2}}$ have no transform? ::: $t^{2}$ beats every $σ t$ , so the product diverges for all $σ$ . What is the ROC of $e^{- t^{2}}$ ? ::: all $s$ — its growth rate is $a = - \infty$ , so $Re (s) > a$ is always true. Value of $L {e^{a t}}$ ? ::: $\frac{1}{s - a}$ , valid for $Re (s) > a$ .

Connections

Improper Integrals — the $\int_{0}^{\infty} = lim_{T \to \infty} \int_{0}^{T}$ machinery whose convergence is the ROC.
Exponential Order and Growth Rates — the ceiling $M e^{a t}$ that defines the wall $a$ .
Laplace Transform of Derivatives — why the $\frac{d}{d t} e^{- s t} = - s e^{- s t}$ property is the real prize.
Inverse Laplace Transform — undoing $F (s)$ ; the ROC restores uniqueness.
Solving ODEs with Laplace Transforms — the payoff.
Fourier Transform — the same picture read along the vertical line $Re (s) = 0$ .

#flashcard