Underdamped, Critically Damped, and Overdamped Regimes

A cheap screen door slams shut, bounces off the frame, and rattles twice more before it settles. A well-fitted fire door glides home in one silent sweep and kisses the latch. The hydraulic door of a bank vault takes a small eternity to ooze through its final centimetre. Mechanically, these three doors are the same system — a mass, a restoring force pulling it home, and a velocity-proportional drag resisting the trip — and all three obey the same one-line equation, $m\ddot{x}+b\dot{x}+kx=0$ . The only thing that differs between them is a single dimensionless knob: the ratio of the damping rate $\gamma=b/2m$ to the natural frequency $\omega_0=\sqrt{k/m}$ . That knob is this article's whole subject. Because the equation is linear with constant coefficients, its entire behavior is encoded in the two roots of a quadratic, and the sign of one discriminant, $\gamma^2-\omega_0^2$ , sorts every damped oscillator in the universe into three regimes: underdamped (it rings), overdamped (it creeps), and — exactly at the watershed — critically damped, the celebrated boundary case that returns to equilibrium as fast as possible without ever overshooting. These are not three different physical laws but three faces of one solution, continuously connected as $\gamma/\omega_0$ is dialed through unity. You already have the damped equation of motion and the complex-exponential method; this is where they pay off. The same trichotomy tunes car suspensions, RLC circuits, robot joints, and every control loop in every drone — and it hides a genuinely counterintuitive prize: past the critical point, adding more damping makes the system settle more slowly.

Key takeaways

Every linear damped oscillator reduces to $\ddot{x}+2\gamma\dot{x}+\omega_0^2x=0$ ; its fate is fixed by one dimensionless ratio, $\gamma/\omega_0$ (engineers call it the damping ratio $\zeta$ ; physicists often quote $Q=\omega_0/2\gamma$ ).
The exponential guess $x\propto e^{rt}$ gives the characteristic roots $r_\pm=-\gamma\pm\sqrt{\gamma^2-\omega_0^2}$ ; the sign of the discriminant $\gamma^2-\omega_0^2$ selects the regime.
Underdamped ( $\gamma<\omega_0$ ): oscillation at the reduced frequency $\omega_1=\sqrt{\omega_0^2-\gamma^2}$ inside a shrinking envelope $e^{-\gamma t}$ .
Overdamped ( $\gamma>\omega_0$ ): no oscillation — two real decay rates, and the slower one, $\gamma-\sqrt{\gamma^2-\omega_0^2}$ , controls a long sluggish tail that gets slower as damping grows.
Critically damped ( $\gamma=\omega_0$ ): the roots merge, the solution becomes $(A+Bt)e^{-\gamma t}$ , and the system returns to equilibrium fastest without overshoot — the engineering sweet spot for doors, meters, and suspensions.
Key idea: one dimensionless number, $\gamma/\omega_0$ , decides everything — less than 1 means oscillatory decay, greater than 1 means monotone creep, and exactly 1 is critical damping, the fastest non-overshooting return.

The core insight

One dimensionless number, $\gamma/\omega_0$ , decides everything: $<1$ oscillatory decay, $>1$ monotone creep, $=1$ critical damping with the fastest non-overshooting return.

Where this lives today

Open any drone flight controller's tuning screen and you are staring at this article. Every axis of a quadcopter's attitude control is, in closed loop, a damped oscillator whose effective $\gamma$ and $\omega_0$ are set by software gains, and the universal starting prescription — $\zeta\approx0.7$ — is a direct application of the settling theorem: fast enough to feel crisp, damped enough not to wobble. The same number governs the autofocus in a phone camera, robot-arm joints, hard-disk head positioning, and the cruise control of every car on the road. When a control engineer speaks of "pole placement," the poles being placed are precisely the characteristic roots $r_\pm$ of equation (2), pushed around the complex plane of our root-locus figure by feedback instead of by a physical dashpot. At the hardware end, the trichotomy is a product spec. MEMS accelerometers and gyroscopes — the motion sensors in phones, airbags, and game controllers — are micron-scale sprung masses deliberately gas-damped near $\zeta\sim0.7$ –1 so they report a step in acceleration once, promptly, without ringing. Skyscraper tuned mass dampers (the 660-tonne pendulum in Taipei 101), seismometer suspensions, MRI gradient coils, and the shock isolation under optical tables are all exercises in placing two roots wisely. And the highest-stakes example runs the other way: LIGO's mirror suspensions are engineered for the least possible damping — $Q\sim10^8$ and up — because eq. (8) links damping to thermal noise; reading gravitational waves means building an oscillator as far from critical as human engineering can push.

The physical picture

Think of two clocks racing inside every damped oscillator. The restoring force wants the system to tick at its natural period, roughly $1/\omega_0$ ; the drag wants to erase all motion on the decay timescale $1/\gamma$ . If the tick is quicker than the erasing ( $\gamma<\omega_0$ ), the system completes swings before dying — it rings. If the erasing wins ( $\gamma>\omega_0$ ), the system never gets through a single swing — it creeps home like a spoon pulled through honey. The knife-edge between them, $\gamma=\omega_0$ , is critical damping: the fastest possible clean return.

Figure

Three regimes from identical starts — the ratio γ/ω₀ picks the personality

The figure above is the whole article in one frame. All three curves start from the same displacement with the same (zero) velocity and obey the same equation; only the knob $\gamma/\omega_0$ differs. The blue curve overshoots and rings — quick to reach zero, slow to stay there. The amber critical curve dives home and stops. The red overdamped curve looks safest but pays for its caution with a long exponential tail: at $t=14/\omega_0$ it is still about $9\%$ from home, long after the critical curve has settled to nothing. There is an equally illuminating way to see the trichotomy: watch the two characteristic roots move in the complex plane as you turn up the damping. For weak damping they form a complex-conjugate pair $-\gamma\pm i\omega_1$ whose real part is the decay and whose imaginary part is the ringing frequency. A lovely and underadvertised fact: since $\gamma^2+\omega_1^2=\omega_0^2$ , the pair always sits on a circle of radius $\omega_0$ . Turning up $\gamma$ slides the roots along that circle toward the real axis; at $\gamma=\omega_0$ they collide head-on at $-\omega_0$ ; past that, they split and run along the real axis in opposite directions — and the one running toward zero is the one that controls your settling time.

Figure

Where the roots live — the complex-plane story of the three regimes

Energetically, the same ratio counts swings. The quality factor $Q=\omega_0/2\gamma$ measures roughly how many radians of oscillation fit inside one amplitude $e$ -folding: $Q>\tfrac12$ permits at least a fraction of a swing (underdamped), $Q<\tfrac12$ forbids even one (overdamped), and $Q=\tfrac12$ is critical. A concert-hall piano string has $Q$ in the thousands; a screen-door closer is engineered to sit within a few percent of $Q=\tfrac12$ .

One equation, three personalities

Start from the universal form. Divide the physical equation $m\ddot{x}+b\dot{x}+kx=0$ through by the mass and package the two physical inputs into rates:

\ddot{x}+2\gamma\dot{x}+\omega_0^2\,x=0 \qquad (1)

where: $x$ — displacement from equilibrium; $\gamma=b/2m$ — the damping rate (units of $\mathrm{s^{-1}}$ ; the factor 2 is a convention that will keep every later formula clean); $\omega_0=\sqrt{k/m}$ — the natural frequency the system would have with no damping. Equation (1) is worth a pause: a mass on a wet spring, a series RLC circuit ( $\omega_0=1/\sqrt{LC}$ , $\gamma=R/2L$ ), a torsion pendulum in oil, and the pitch loop of a quadcopter all arrive at exactly this form. Solve it once, and you have solved them all. Because (1) is linear with constant coefficients, exponentials are its natural language, and the whole classification falls out of one quadratic.

Derivation AFrom the exponential guess to the three regimes

We are solving equation (1) completely — not one regime at a time, but all of them at once — by converting the differential equation into algebra. The strategy is the standard one for any linear constant-coefficient ODE: guess $x=e^{rt}$ , which turns each time derivative into a multiplication by $r$ , and let the resulting quadratic hand us the two allowed values of $r$ . The discriminant of that quadratic then does all the physical work: its sign is the only thing that distinguishes ringing from creeping. The case to watch is the degenerate one, $\gamma=\omega_0$ , where the two roots coincide and the two-exponential basis collapses — a second, genuinely independent solution has to be manufactured, and we devote a separate derivation to it below.

Step 1. Substitute the trial solution $x=e^{rt}$ into (1). Each dot brings down one factor of $r$ , and the common factor $e^{rt}$ (never zero) cancels:

r^2+2\gamma r+\omega_0^2=0

Why this guess is legitimate: a second-order linear ODE has exactly two independent solutions, so if the exponential family supplies two, it supplies everything — the general solution is their superposition.

Step 2. Solve the quadratic:

r_\pm=-\gamma\pm\sqrt{\gamma^2-\omega_0^2} \qquad (2)

Why every solution decays: for $\gamma>0$ both roots have negative real part — either $-\gamma$ itself (complex case) or two negative real numbers, since $\sqrt{\gamma^2-\omega_0^2}<\gamma$ . Damping can never produce growth; the equilibrium is always approached.

Step 3. Read the discriminant. If $\gamma<\omega_0$ the square root is imaginary: write $\sqrt{\gamma^2-\omega_0^2}=i\omega_1$ with $\omega_1=\sqrt{\omega_0^2-\gamma^2}$ , so $r_\pm=-\gamma\pm i\omega_1$ . Superposing the conjugate pair and demanding a real $x(t)$ gives

x(t)=A\,e^{-\gamma t}\cos(\omega_1 t+\phi)

Why the frequency drops below $\omega_0$ : the drag steals a little momentum on every pass, so each swing takes slightly longer than the undamped clock would tick — the oscillation runs at $\omega_1<\omega_0$ , sliding along the root circle rather than sitting at its top.

Step 4. If $\gamma>\omega_0$ the square root is real: two distinct negative roots, and the general solution is a plain sum of decays,

x(t)=A\,e^{r_+t}+B\,e^{r_-t}

Why nothing oscillates: with both $r_\pm$ real, $x(t)$ is a sum of two monotone exponentials — it can cross zero at most once (setting $Ae^{r_+t}=-Be^{r_-t}$ gives one equation with at most one solution in $t$ ), so a released overdamped system never rings.

Step 5. If $\gamma=\omega_0$ exactly, $r_+=r_-=-\gamma$ and Steps 3–4 hand us only one solution, $e^{-\gamma t}$ — not enough for a second-order equation. The resolution, derived separately below, is the extra factor of $t$ :

x(t)=(A+Bt)\,e^{-\gamma t}

A quick dimensional check confirms every exponent $rt$ , $\gamma t$ , $\omega_1 t$ is dimensionless.

The three results deserve to stand on their own, because you will reach for them constantly. In the underdamped regime the motion is a cosine trapped inside a shrinking envelope:

x(t)=A\,e^{-\gamma t}\cos(\omega_1 t+\phi),\qquad \omega_1=\sqrt{\omega_0^2-\gamma^2} \qquad (3)

where: $A$ and $\phi$ — amplitude and phase fixed by the initial conditions; $\omega_1$ — the damped frequency, always below $\omega_0$ . For light damping the shift is tiny ( $\omega_1\approx\omega_0(1-\gamma^2/2\omega_0^2)$ , a second-order effect), which is why a guitar string decays audibly yet holds its pitch. At the watershed the solution changes its functional species entirely:

x(t)=(A+Bt)\,e^{-\gamma t}\qquad(\gamma=\omega_0) \qquad (4)

where: $A=x(0)$ and $B=\dot{x}(0)+\gamma x(0)$ — two constants, as a second-order equation demands. The polynomial-times-exponential form is the universal signature of a repeated root, and despite the growing factor $t$ the whole expression still dies: $te^{-\gamma t}$ peaks at $t=1/\gamma$ and then the exponential wins. Past the watershed, the motion is a superposition of a fast and a slow decay:

x(t)=A\,e^{r_+t}+B\,e^{r_-t},\qquad r_\pm=-\gamma\pm\sqrt{\gamma^2-\omega_0^2} \qquad (5)

where: $r_+$ — the slow rate (smaller magnitude), which owns the late-time tail; $r_-$ — the fast rate, which handles the first instants and then bows out. Releasing the system from rest weights the slow mode heavily, which is why overdamped systems feel sluggish from the start. Finally, the knob itself comes with two standard aliases, and you should be bilingual:

\zeta\equiv\frac{\gamma}{\omega_0}=\frac{b}{2\sqrt{mk}}=\frac{1}{2Q},\qquad Q\equiv\frac{\omega_0}{2\gamma} \qquad (6)

where: $\zeta$ — the damping ratio of every engineering text (regimes: $\zeta<1$ under, $=1$ critical, $>1$ over); $Q$ — the quality factor beloved of physicists (regimes: $Q>\tfrac12$ , $=\tfrac12$ , $<\tfrac12$ ). They carry identical information. A control engineer saying " $\zeta=0.7$ " and a physicist saying " $Q=0.71$ " are describing the same lightly-ringing response.

Which formula answers which question

you want	use	notes
to classify a system fast	sign of $\gamma^2-\omega_0^2$ , or $\zeta=\gamma/\omega_0$ vs 1	eq. (2), (6)
the ringing frequency	$\omega_1=\sqrt{\omega_0^2-\gamma^2}$	eq. (3); only exists for $\gamma<\omega_0$
response at the boundary	$(A+Bt)e^{-\gamma t}$	eq. (4); $B=\dot{x}(0)+\gamma x(0)$
the overdamped settling time	slow rate $\gamma-\sqrt{\gamma^2-\omega_0^2}$	eq. (7); $\approx\omega_0^2/2\gamma$ when $\gamma\gg\omega_0$
damping from a decay measurement	log decrement $\delta=\gamma T_1\approx\pi/Q$	eq. (8)
translation between communities	$\zeta=1/2Q$	eq. (6)

The watershed, taken seriously

Two things about critical damping deserve honest proof rather than folklore: that the odd-looking solution (4) really is the limit of its neighbors, and that "critical is fastest" is a theorem, not a slogan.

Go deeper · Derivation B · 1 minThe critical solution as a smooth limit of the underdamped one

Derivation BThe critical solution as a smooth limit of the underdamped one

The goal is the critically damped solution $(A+Bt)e^{-\gamma t}$ , obtained not by fiat but as the continuous $\omega_1\to0$ limit of the underdamped solution — showing that the three regimes are one analytic family and that nothing physically discontinuous happens at $\gamma=\omega_0$ . The strategy: write the underdamped solution for fixed initial conditions and then let $\omega_1\to0$ while holding those initial conditions fixed. The subtlety to watch is that the naive coefficient of the sine blows up as $1/\omega_1$ in the limit — the initial-condition bookkeeping is exactly what tames it into the linear-in- $t$ term.

Step 1. Write the underdamped solution in sine-cosine form and fit the initial data:

x(t)=e^{-\gamma t}\big(A\cos\omega_1 t+C\sin\omega_1 t\big)

with $A=x(0)$ and, differentiating once, $\omega_1C-\gamma A=\dot{x}(0)$ , so $C=\big(\dot{x}(0)+\gamma x(0)\big)/\omega_1$ . Why we fit the data first: the limit must be taken along a family of physically identical experiments — same release position, same release velocity — otherwise the divergent $C$ is an artifact of comparing different motions.

Step 2. Introduce $B\equiv\dot{x}(0)+\gamma x(0)$ , which stays fixed as the regime knob turns, and rewrite:

x(t)=e^{-\gamma t}\Big(A\cos\omega_1 t+B\,\frac{\sin\omega_1 t}{\omega_1}\Big)

Step 3. Send $\omega_1\to0$ using the small-angle behavior of the trig functions:

\cos\omega_1 t\to1,\qquad \frac{\sin\omega_1 t}{\omega_1}\to t

Why this is legitimate for all $t$ : at any fixed $t$ , $\sin\omega_1 t=\omega_1 t-(\omega_1 t)^3/6+\cdots$ , so $\sin\omega_1 t/\omega_1\to t$ pointwise — the sine's first wiggle has been pushed off to $t\sim\pi/\omega_1\to\infty$ , beyond any time we care about.

Step 4. The limit lands exactly on the repeated-root form:

x(t)\to(A+Bt)\,e^{-\gamma t}

The same expression also emerges from differentiating $e^{rt}$ with respect to the root, $\partial_r e^{rt}=te^{rt}$ — the "second solution" is the derivative of the exponential family across the collision. The constants $A=x(0)$ and $B=\dot{x}(0)+\gamma x(0)$ agree with the direct initial-condition fit of eq. (4).

Now the optimality claim. It hinges on a single inequality about the slow root.

Go deeper · Derivation C · 1 minCritical damping gives the fastest return without overshoot

Derivation CCritical damping gives the fastest return without overshoot

We are proving the design theorem that makes $\zeta=1$ famous: among all damping settings that never overshoot ( $\gamma\ge\omega_0$ , at fixed $\omega_0$ ), the critical setting drives the system to equilibrium fastest at late times. The strategy is to compare the slowest surviving exponent in each case, because after a short transient every solution is owned by its slowest decay mode. Watch one subtlety: the underdamped settings are excluded from the contest not because they are slow — they reach zero sooner — but because they cross it and ring, which is precisely what a door, a meter needle, or a suspension must not do.

Step 1. Identify the late-time master of each regime. Overdamped, from eq. (5): the tail is $x\sim e^{r_+t}$ with slow rate

|r_+|=\gamma-\sqrt{\gamma^2-\omega_0^2} \qquad (7)

Critical, from eq. (4): $x\sim t\,e^{-\gamma t}$ decays at the full rate $\gamma=\omega_0$ , the linear prefactor being only a logarithmic drag on the clock.

Step 2. Show the overdamped slow rate always loses. For any $\gamma>\omega_0$ ,

\gamma-\sqrt{\gamma^2-\omega_0^2}\;<\;\omega_0

Why: rearrange to $\gamma-\omega_0<\sqrt{\gamma^2-\omega_0^2}=\sqrt{(\gamma-\omega_0)(\gamma+\omega_0)}$ and square — the claim reduces to $\gamma-\omega_0<\gamma+\omega_0$ , which is trivially true. So every overdamped setting has a slower dominant mode than the critical one.

Step 3. See how badly it loses when you overdo it. Expanding (7) for $\gamma\gg\omega_0$ :

|r_+|=\gamma\Big(1-\sqrt{1-\omega_0^2/\gamma^2}\Big)\approx\frac{\omega_0^2}{2\gamma}

Why this is the honey limit: for enormous drag the inertia never matters — the spring force $kx$ simply balances the drag force $b\dot{x}$ , giving $\dot{x}\approx-(k/b)x$ and a creep rate $k/b=\omega_0^2/2\gamma$ . Doubling the damping halves the return rate.

Step 4. Conclude the design rule. The settling rate as a function of $\gamma$ rises like $\gamma$ itself below the watershed (the envelope $e^{-\gamma t}$ ), peaks at $\gamma=\omega_0$ , and falls off like $\omega_0^2/2\gamma$ beyond it. Critical damping sits exactly at the maximum — the fastest non-overshooting return. For example, with $\gamma=10\ \mathrm{s^{-1}}$ and $\omega_0=4\ \mathrm{s^{-1}}$ , the exact slow rate is $|r_+|=10-\sqrt{84}=0.835\ \mathrm{s^{-1}}$ , close to the honey estimate $\omega_0^2/2\gamma=0.8\ \mathrm{s^{-1}}$ .

This is why the damping dial of the engineered world clusters so tightly. Too far left and things bounce and ring; too far right and they take forever; the valuable real estate is a narrow band around $\zeta=1$ — or, when a small overshoot is acceptable in exchange for speed, the control engineer's habitual $\zeta\approx0.7$ , which reaches the 2%-settled band sooner than exact critical damping does.

Figure

The damping dial — one number places every oscillator in its regime

Measuring your regime

Hand someone a black-box oscillator and the regime is a measurement, not a guess. Poke it. If it rings, count the fading: the cleanest diagnostic is the ratio of successive swing peaks.

Go deeper · Derivation D · 1 minThe logarithmic decrement measures your distance from critical

Derivation DThe logarithmic decrement measures your distance from critical

The goal is a formula that converts something trivially measurable — how much smaller each successive swing is — into the damping parameters $\gamma$ and $Q$ , locating an underdamped system quantitatively on the regime dial. The strategy: successive maxima of the underdamped solution (3) are separated by one damped period $T_1=2\pi/\omega_1$ , over which the cosine repeats exactly, so only the envelope ratio survives. Note what this buys experimentally: no force sensor, no calibration of $A$ — a ruler and a stopwatch suffice.

Step 1. Take the ratio of the solution at two successive peaks, a time $T_1$ apart:

\frac{x(t)}{x(t+T_1)}=\frac{A\,e^{-\gamma t}}{A\,e^{-\gamma(t+T_1)}}=e^{\gamma T_1}

Why the cosines cancel: $\cos(\omega_1(t+T_1)+\phi)=\cos(\omega_1 t+\phi+2\pi)$ — one full period returns the oscillatory factor to its exact value, peak for peak.

Step 2. Define the logarithmic decrement and read off the damping:

\delta\equiv\ln\frac{x(t)}{x(t+T_1)}=\gamma T_1=\frac{2\pi\gamma}{\omega_1} \qquad (8)

Step 3. Translate to the dial. For light damping ( $\omega_1\approx\omega_0$ ):

\delta\approx\frac{\pi}{Q}=2\pi\zeta

Why this is a regime-meter: measure two peak heights, take a log, and $\pi/\delta$ is your $Q$ — if $\delta$ is heading toward $2\pi$ (i.e. $\zeta$ toward 1), your wiggles are about to disappear entirely. *The amplitude-per-cycle factor $e^{-\gamma T_1}$ is dimensionally clean: the exponent $\gamma T_1$ is the product of a rate and a time, hence dimensionless.*

Energy tracks the same envelope, twice as fast: since $E\propto A^2$ , an underdamped oscillator's energy decays as $E(t)\approx E_0e^{-2\gamma t}$ for $Q\gg1$ , so the energy $e$ -folding time is $1/2\gamma$ , half the amplitude's. If the box does not ring, watch the tail instead: an overdamped system betrays itself by moving briskly at first (the fast mode) and then crawling the last stretch at the slow rate (7).

Limiting cases

Case	Result	Meaning
Undamped, $\gamma=0$	$x=A\cos(\omega_0t+\phi)$ ; roots at $\pm i\omega_0$	pure SHM; the roots sit at the top and bottom of the $\omega_0$ circle
Light damping, $\gamma\ll\omega_0$	$\omega_1\approx\omega_0\big(1-\tfrac{\gamma^2}{2\omega_0^2}\big)$	pitch essentially unchanged; only the envelope reveals the damping
Underdamped, $\gamma<\omega_0$	$x=Ae^{-\gamma t}\cos(\omega_1t+\phi)$	decaying oscillation at reduced frequency $\omega_1$
Critical, $\gamma=\omega_0$	$x=(A+Bt)e^{-\gamma t}$	repeated root; fastest non-overshooting return; at most one zero crossing
Overdamped, $\gamma>\omega_0$	$x=Ae^{r_+t}+Be^{r_-t}$	two real decays; slow tail $e^{r_+t}$ rules late times
Honey limit, $\gamma\gg\omega_0$	$r_+\to-\dfrac{\omega_0^2}{2\gamma}$ , $r_-\to-2\gamma$	inertia irrelevant: spring-drag balance gives diffusion-like creep that slows as $b$ grows

Common trapMore damping always means a faster, safer return

It does not — past the watershed the trend reverses. The overdamped slow rate $\gamma-\sqrt{\gamma^2-\omega_0^2}\approx\omega_0^2/2\gamma$ decreases as you add damping: a door closer stiffened to twice critical takes roughly twice as long to shut as one tuned at critical. Two smaller traps ride along. First, critical damping does not make the system reach zero soonest — an underdamped system crosses zero earlier; critical is the fastest return without overshoot, and "reaching zero" is settled only asymptotically in every regime. Second, do not read the overdamped system's brisk initial motion as speed: the fast mode $e^{r_-t}$ handles the opening moments, then hands the motion to the slow mode's long crawl. Judge settling by the slowest surviving exponent, never by the first hundred milliseconds.

Worked examples

1 · Classify, solve, and time three dampers (confidence builder)

A $0.5\ \mathrm{kg}$ mass on a spring of stiffness $k=8\ \mathrm{N/m}$ is fitted, in three separate experiments, with dampers of $b=1$ , $4$ , and $10\ \mathrm{kg/s}$ . Classify each, write the solution released from rest at $x_0$ , and find how long each takes to settle within 1% of equilibrium.

Step 1. The natural frequency, common to all three: $\omega_0=\sqrt{k/m}=\sqrt{8/0.5}=4\ \mathrm{rad/s}$ . The damping rates: $\gamma=b/2m=1$ , $4$ , and $10\ \mathrm{s^{-1}}$ . So the three experiments are underdamped ( $\gamma<\omega_0$ ), critical ( $\gamma=\omega_0$ ), and overdamped ( $\gamma>\omega_0$ ).

Step 2. Underdamped, $\gamma=1\ \mathrm{s^{-1}}$ : ringing frequency $\omega_1=\sqrt{16-1}=\sqrt{15}=3.87\ \mathrm{rad/s}$ — barely below $\omega_0$ . Released from rest, $x=x_0(\omega_0/\omega_1)e^{-\gamma t}\cos(\omega_1t-\phi)$ with a small phase tilt; the envelope is $1.033\,x_0\,e^{-t}$ . Settling to 1% means $1.033\,e^{-t}=0.01$ , i.e. $t=\ln 103.3=\boxed{4.6\ \mathrm{s}}$ — with about $2.9$ full wiggles along the way.

Step 3. Critical, $\gamma=4\ \mathrm{s^{-1}}$ : from eq. (4) with $x(0)=x_0$ , $\dot{x}(0)=0$ : $A=x_0$ , $B=\gamma x_0$ , so $x=x_0(1+4t)e^{-4t}$ . Solving $(1+4t)e^{-4t}=0.01$ numerically gives $t=\boxed{1.66\ \mathrm{s}}$ — nearly three times faster than the underdamped setting, with zero overshoot.

Step 4. Overdamped, $\gamma=10\ \mathrm{s^{-1}}$ : roots $r_\pm=-10\pm\sqrt{100-16}=-10\pm9.17$ , i.e. a slow rate of $0.83\ \mathrm{s^{-1}}$ and a fast rate of $19.2\ \mathrm{s^{-1}}$ (the honey estimate $\omega_0^2/2\gamma=0.8\ \mathrm{s^{-1}}$ is already excellent). Released from rest, the slow mode carries almost all the weight, and the 1% time is $t=\boxed{5.6\ \mathrm{s}}$ — the most damped experiment is the slowest, and slower even than the ringing one.

2 · Tuning a car suspension (exam level)

A quarter-car model has sprung mass $m=350\ \mathrm{kg}$ on a spring $k=2.0\times10^4\ \mathrm{N/m}$ . Find the critical damping coefficient. The manufacturer tunes the damper to $\zeta=0.3$ for comfort; find the ride frequency and the first overshoot after a curb strike. Then explain what worn shocks ( $\zeta\approx0.1$ ) do to the car.

Step 1. Natural frequency: $\omega_0=\sqrt{k/m}=\sqrt{2.0\times10^4/350}=7.56\ \mathrm{rad/s}$ , i.e. $f_0=\omega_0/2\pi=1.20\ \mathrm{Hz}$ — right in the classic 1–1.5 Hz passenger-comfort band.

Step 2. Critical damping coefficient: $b_c=2m\omega_0=2\sqrt{km}=2\times350\times7.56=\boxed{5.29\times10^3\ \mathrm{kg/s}}$ . The comfort tune installs $b=\zeta b_c=0.3\times5292=1.59\times10^3\ \mathrm{kg/s}$ .

Step 3. Ride (damped) frequency: $\omega_1=\omega_0\sqrt{1-\zeta^2}=7.56\times\sqrt{1-0.09}=7.21\ \mathrm{rad/s}$ — a period of $0.87\ \mathrm{s}$ per bounce, only 4.6% slower than undamped.

Step 4. Overshoot after a displacement (released from rest, first extremum on the far side occurs half a damped period later): the overshoot fraction is $e^{-\pi\zeta/\sqrt{1-\zeta^2}}=e^{-\pi(0.3)/0.954}=e^{-0.988}=\boxed{0.37}$ . The body sails 37% of the initial deflection past equilibrium once, wiggles once more at the 14% level ( $0.37^2$ ), and is effectively settled — the accepted comfort compromise: faster than critical into the tolerance band, at the price of one soft bounce.

Step 5. Worn shocks, $\zeta\approx0.1$ : overshoot $e^{-\pi(0.1)/0.995}=0.73$ . Each rebound retains 73% of the last — after the same curb strike the car visibly porpoises four or five times, tire load fluctuates, and braking distance grows. That is exactly what the workshop "bounce test" checks: push the fender down, and a healthy ( $\zeta\sim0.3$ ) car recovers with about one overshoot, while a worn one keeps bobbing.

Origins · 1 minLord Rayleigh and how this was discovered

Lord Rayleigh

1842–1919 · Cambridge; author of The Theory of Sound

John William Strutt, the third Baron Rayleigh, made damping a subject rather than a nuisance. His two-volume Theory of Sound (1877–78) — much of it written on a houseboat on the Nile while he recovered from rheumatic fever — treats the vibrations of strings, bars, plates, and air with damping built into the formalism from the start. He introduced the dissipation function, a quadratic form in the velocities that slots friction neatly into Lagrangian mechanics and still bears his name, and he systematized the small-oscillations toolkit — normal coordinates, energy methods, perturbation of frequencies — that this whole article tier relies on. A meticulous experimentalist as well as a theorist, he isolated argon (with William Ramsay), explained why the sky is blue, and won the 1904 Nobel Prize in Physics; the damped-oscillator algebra he standardized now runs through every textbook chapter on vibrations, including this one.

History

1660

— Robert Hooke states the linear restoring law; combined with Newton's second law it yields the undamped oscillator, the

\gamma=0

backbone of the story.

1743

— Leonhard Euler solves linear constant-coefficient ODEs with the exponential ansatz and characteristic roots — the exact algebra that classifies the three regimes.

1851

— George Gabriel Stokes analyzes pendulums moving in viscous fluids, giving the velocity-proportional drag law that makes

b\dot{x}

a physical force rather than a mathematical convenience.

1868

— James Clerk Maxwell's paper On Governors founds stability analysis: engine speed governors misbehave exactly when their characteristic roots stray toward oscillation — regime thinking applied to machines.

1877

— Lord Rayleigh's Theory of Sound systematizes damped vibration, introducing the dissipation function and the energy methods behind

Q

Going further

References. Morin, Introduction to Classical Mechanics, ch. 4 (Oscillations) · Goldstein, Poole & Safko, Classical Mechanics, 3rd ed., §3 and small-oscillation chapter · Landau & Lifshitz, Mechanics, Vol. 1, §25 (Damped oscillations) · Taylor, Classical Mechanics, ch. 5. Key algebraic checkpoints: the characteristic roots $r_\pm=-\gamma\pm\sqrt{\gamma^2-\omega_0^2}$ follow directly from the quadratic; the $\omega_1\to0$ limit yields $(A+Bt)e^{-\gamma t}$ ; the overdamped slow exponent satisfies $\gamma-\sqrt{\gamma^2-\omega_0^2}<\gamma$ ; and $\sqrt{25-9}=4$ . Worked-example numbers ( $\omega_1=\sqrt{15}=3.87\ \mathrm{rad/s}$ , settling times $4.6/1.66/5.6\ \mathrm{s}$ ; suspension $\omega_0=7.56\ \mathrm{rad/s}$ , $b_c=5.29\times10^3\ \mathrm{kg/s}$ , overshoots $0.37$ and $0.73$ ) are illustrative arithmetic computed from the boxed formulas.

Builds on: Simple Harmonic Motion · Damped Harmonic Oscillator · Complex Exponential Representation of Oscillations · Small Oscillations About Equilibrium. Leads to: Driven Oscillations · Resonance · Transient and Steady-State Response · Energy in Simple Harmonic Motion.

Is critical damping the fastest way back to zero, full stop?

No — and the distinction matters on exams and in design reviews alike. An underdamped system reaches zero sooner: it comes in fast and crosses. What it cannot do is stay — it overshoots and rings. Critical damping is the fastest return that never overshoots, which is the relevant contest for doors, meter needles, and read/write heads. And if a small overshoot is tolerable, you can beat critical into a tolerance band: a $\zeta\approx0.7$ system enters and stays inside a ±2% band sooner than the critical one, at the price of one ~5% overshoot. That is why control engineers' default is 0.7, while "aim exactly critical" is the rule only when any overshoot is forbidden — a galvanometer needle must not swing past the true reading.

Why does MORE damping make an overdamped system return more SLOWLY? That feels backwards.

Because deep in the overdamped regime the drag is fighting the return itself. Picture the honey limit $\gamma\gg\omega_0$ : inertia is irrelevant, and the motion is a quasi-static balance of spring force against drag, $b\dot{x}\approx-kx$ . The creep rate is then $k/b$ — the drag sits in the denominator. Doubling $b$ halves the speed at which the spring can haul the mass home. Formally, the slow root $\gamma-\sqrt{\gamma^2-\omega_0^2}\approx\omega_0^2/2\gamma$ shrinks as $\gamma$ grows; on the root-locus picture, adding damping pushes the slow root toward zero, not away. Below the watershed damping speeds settling (envelope $e^{-\gamma t}$ ); above it, damping obstructs it. The optimum is exactly at the collision, $\gamma=\omega_0$ .

Where does the t in (A+Bt)e^(−γt) come from? Can a physical solution really grow linearly?

The $t$ is the fossil of two exponentials that merged. Just below critical, the solution is a combination of $e^{-\gamma t}\cos\omega_1t$ and $e^{-\gamma t}\sin(\omega_1 t)/\omega_1$ ; as $\omega_1\to0$ the second one tends pointwise to $t\,e^{-\gamma t}$ — the sine straightens into a ramp because its first bend is postponed to $t\sim\pi/\omega_1\to\infty$ . Equivalently, when roots collide, $\partial_r e^{rt}=t\,e^{rt}$ supplies the missing independent solution. And no, nothing grows: $te^{-\gamma t}$ rises only until $t=1/\gamma$ , peaking at $e^{-1}/\gamma$ , after which the exponential crushes it. That early rise is physical — release a critically damped system with an inward shove and you can watch it swell once before dying.

How do I tell the regime of a black-box system without knowing m, b, or k?

Poke it and watch. If it wiggles even once past equilibrium, it is underdamped — then count: measure two successive peak amplitudes, take $\delta=\ln(x_n/x_{n+1})$ , and $Q\approx\pi/\delta$ places it precisely on the dial. If it never crosses, distinguish critical-ish from heavily overdamped by the tail: a near-critical system's return looks like one clean motion, while a strongly overdamped one moves briskly for an instant (the fast mode $e^{-2\gamma t}$ ) and then crawls interminably (the slow mode $e^{-\omega_0^2t/2\gamma}$ ) — brisk start, endless finish is the overdamped signature. Timing the late tail against the initial move even gives you the ratio $\gamma/\omega_0$ . This is exactly the mechanic's bounce test for shocks, done with numbers.

Are ζ and Q the same thing? My vibrations book uses ζ everywhere and my circuits book uses Q.

Same knob, two dials: $\zeta=1/(2Q)$ , always. The regime boundaries translate as $\zeta=1\Leftrightarrow Q=\tfrac12$ ; underdamped means $\zeta<1$ or $Q>\tfrac12$ . The cultural split is real, though: control and mechanical engineers quote $\zeta$ because design specs (overshoot, settling time) are clean functions of it; resonance physicists and RF engineers quote $Q$ because it counts oscillations and sets resonance sharpness ( $Q=\omega_0/\Delta\omega$ ) and energy loss per radian. For a series RLC circuit, $\gamma=R/2L$ and $\omega_0=1/\sqrt{LC}$ , so $\zeta=(R/2)\sqrt{C/L}$ and critical damping means $R=2\sqrt{L/C}$ . Fluency in both languages is worth five minutes of memorization — interviewers in either field assume it.

Do real systems ever sit exactly at critical damping?

Never exactly — $\gamma=\omega_0$ is a measure-zero point, and component tolerances, temperature, and wear guarantee you miss it. Engineers therefore aim near it, on whichever side failure is cheaper: door closers slightly over (a slow door annoys; a slamming door injures), instrument needles at 0.9–1, control loops under at $\zeta\approx0.7$ . Temperature is the classic saboteur: damper oil thins when hot and thickens when cold, dragging $\gamma$ with it while the spring's $\omega_0$ barely moves — which is why a door closer tuned in summer can turn overdamped in January, creeping so slowly the latch never catches. Good dashpots ship with an adjustment screw for exactly this reason.

Which regime do I want for a clean voltage pulse in an RLC circuit?

Critical or a shade over. A logic edge or reset pulse driving an underdamped RLC ( $R<2\sqrt{L/C}$ ) rings — the voltage overshoots and oscillates at $\omega_1$ , potentially double-triggering downstream logic; on a scope the tell-tale is the decaying sinusoid after each edge. Overdamping suppresses the ring but rounds the edge and stretches the settling, throttling your data rate through the slow root. Pulse and probe designers therefore aim at $R\approx2\sqrt{L/C}$ , often nudged a few percent overdamped as safety margin against tolerance drift. The same reasoning tunes oscilloscope probes themselves: the compensation trimmer you adjust for a flat square wave is walking the probe's RC-L response through this exact trichotomy.

Underdamped, Critically Damped, and Overdamped Regimes

Where this lives today

The physical picture

One equation, three personalities

Which formula answers which question

The watershed, taken seriously

Measuring your regime

Limiting cases

Worked examples

1 · Classify, solve, and time three dampers (confidence builder)

2 · Tuning a car suspension (exam level)

Going further

Before this

Where this leads