What Is a Differential Equation and How Is It Used in Engineering?

How does a thermostat know exactly when to click the heater back on?

It doesn't know the room's future temperature. It doesn't have a formula that says "in ten minutes it'll be 19 degrees." All it has is a thermometer reading the temperature right now, and a rule: if it's too cold, turn on. That's it.

And yet, string that simple rule together moment after moment, and the room's whole temperature history unfolds from it, precisely, predictably, every single day.

That's what a differential equation is. Not a formula for where something ends up, but a rule for how it changes right now. Engineers reach for these constantly, because almost nothing they build hands them the future directly. A circuit only tells you the current flowing through it this instant. A car's suspension only feels the force acting on it this instant. You're always standing in the present, holding a rate, and your job is to figure out where that rate takes you.

A differential equation is a rule for change

Most equations you've solved before ask "what number makes this true?" Solve $2x + 3 = 7$ and you get one number: $x = 2$.

A differential equation asks something different: "what function makes this true, given a rule about its own rate of change?" Instead of one number, the answer is an entire curve, a function of time, that has to obey the rule at every instant along the way.

Here's the simplest one you'll meet. Say a quantity's rate of change is proportional to the quantity itself:

Read that sentence, not just the symbols: the bigger $y$ already is, the faster it changes. That single line describes a capacitor charging, a population growing, a hot cup of coffee cooling, and a radioactive sample decaying, all at once, just with different signs and starting values plugged in. We'll build up from there to see exactly how engineers turn a rule like this into something they can design around.

Every capacitor charging through a resistor obeys the same rule

Start with electronics. Wire a resistor $R$ and a capacitor $C$ in series with a battery of voltage $V_s$, and watch the voltage across the capacitor over time. It doesn't jump straight to $V_s$. It climbs, fast at first, then slower, easing into the final value.

Why? Because the rate the capacitor charges depends on how far it still has to go:

Big gap between $V$ and $V_s$, fast charging. Small gap, slow charging. That's a differential equation with the exact same shape as the population one above, just decaying toward a target instead of exploding away from zero. Solve it and you get:

$\tau$ is called the time constant, and it's the one number every electrical engineer memorizes: after one $\tau$, the capacitor is 63% charged. After three, it's essentially done.

Drag $\tau$ and watch the entire curve rescale around it. Every RC circuit on earth, in your phone charger, your car's turn signal flasher, a camera flash, traces this exact shape. Only $\tau$ changes.

One damping ratio decides whether a car wobbles or glides

Circuits are first order: one derivative, one rate. Mechanical systems usually need a second derivative, because force produces acceleration, not velocity. That's Newton's second law, $F = ma$, and $a$ is a rate of a rate.

Picture a car's suspension after it hits a bump. The spring pushes the body back toward level, and a shock absorber (the damper) resists however fast the body is moving, in either direction. Put both forces into Newton's law and you get:

Divide through and name the pieces $\omega$ (how fast it would oscillate with no damping at all) and $\zeta$ (the damping ratio, how much drag fights the motion), and the equation becomes:

That single number $\zeta$ is one of the most important design choices in mechanical engineering. Too low, and the car bounces several times after every pothole. Too high, and the ride feels stiff and slow to recover. Somewhere in between is the sweet spot engineers tune for.

Slide $\zeta$ up from near zero and watch the body overshoot and ring before settling. Push it past 1 and the wobble disappears entirely, the body just glides back to level, a little slower each time you increase it further. Real shock absorbers are tuned to sit just below that crossover, fast recovery with barely any bounce.

A thermostat clicking on and off is a differential equation with a switch in it

Back to that opening question. A room without a heater cools toward the outside temperature at a rate proportional to how far above it the room is, Newton's law of cooling:

Add a heater, and it either pushes heat in or it doesn't. There's no in-between setting on a basic thermostat, just on and off, switched by how close the temperature is to the setpoint:

Now here's the part that trips people up: this equation doesn't produce a smooth curve that settles at the setpoint. It can't. The moment the temperature reaches the setpoint, the heater has to make a decision, on or off, and whichever it picks pushes the temperature back past the setpoint from the other side. The result is a sawtooth, forever circling the target it can never quite sit still on.

Narrow the switching band and the heater fires constantly, holding a tight range but wearing out the relay. Widen it and the room drifts further before flipping, fewer switches, wider swings. That tradeoff, precision against switching frequency, shows up in every on-off controller you own: thermostats, mini-fridges, and old-school hard drive temperature fans included.

When there's no formula: why engineers simulate instead of solve

Every example so far had a clean closed-form solution, a formula you could type into a calculator. Most real engineering systems don't get that luxury. Take a pendulum, pulled back and released:

For small swings, $\sin\theta \approx \theta$, and the equation collapses into the same friendly linear shape as the spring: $\ddot\theta + \frac{g}{L}\theta = 0$. That has a tidy formula, $\theta(t) = \theta_0 \cos(\omega_0 t)$, with a period that doesn't even depend on the starting angle.

But that's an approximation, only valid because $\sin\theta \approx \theta$ breaks down as $\theta$ grows. The real equation keeps the actual $\sin\theta$, and nobody has ever found a formula for $\theta(t)$ built from ordinary functions. Not because no one's clever enough. It's been proven that one doesn't exist in that form.

At small angles the dashed formula and the solid simulated swing sit right on top of each other. Push the starting angle up and watch them peel apart, the real pendulum swings slower than the formula predicts, because gravity's pull weakens (in the tangential direction) the further the bob swings out. This is exactly why engineering software like MATLAB and Simulink exist: when the equation won't give up a formula, you step through it numerically instead, tiny time-slice by tiny time-slice, and read the answer straight off the simulation.

The short version

A differential equation is a rule connecting a quantity to its own rate of change, not a formula for the quantity itself. Engineers run into them everywhere because almost every real system only ever reveals its rate: a capacitor's charging speed, a car body's acceleration, a room's rate of heat loss, a pendulum's swing. Turning that rate into a prediction of the future is the actual work.

Some of these equations split apart into a clean formula, like the RC circuit's exponential curve or the small-angle pendulum's cosine. Some need a second derivative and a damping ratio to capture inertia, like a suspension settling after a bump. Some have switches built in, like a thermostat, that guarantee the answer is never a smooth curve at all. And some, like the full pendulum, have no formula whatsoever, so the only way forward is to simulate, step by tiny step, and trust the numbers.

Every circuit, structure, and control loop you'll ever design starts the exact same way: not with an answer, but with a rate, and a rule for what to do with it.

All visualizations are interactive React components running entirely in your browser, using SVG. The pendulum uses a fourth-order Runge-Kutta integrator, the same family of numerical method engineering simulation software runs under the hood. No libraries beyond React.