Standing Waves and Resonance: Why Everything Has a Natural Frequency

Pluck a guitar string and you don't get a noise, you get a note — the same note whether you pluck it hard or softly, near the bridge or over the sound hole. Run a wet finger round the rim of a wine glass and the same stubbornness shows up: one pitch, and it is the glass's pitch, not yours. Two objects with nothing in common, both refusing to vibrate at anything except a frequency they chose for themselves.

That refusal is the whole subject. Anything with mass and springiness has a short list of frequencies it is willing to move at, and it more or less ignores everything else. Push it at one of those and the motion piles up cycle after cycle into something enormous; push it a few percent off and almost nothing happens. Below are four things to play with while that idea assembles itself: a string that only lights up on its harmonics, a metal plate where loose sand snaps into a sharp geometric figure and redraws it at every new note, a mass on a spring that shows why resonance is a question of timing more than effort, and a skyline where you choose the earthquake and watch it pick out which building to wreck.

What Is a Standing Wave and How Does It Form?

A standing wave isn't a special species of wave. It is what two perfectly ordinary waves look like when they pass through each other in opposite directions at the same frequency — and the easiest way to arrange that is to send one wave down a string and let the far end hand it back.

Fix both ends, wiggle the middle, and every wave you make runs to an end, reflects, and comes back through whatever is heading out. At every point on the string the two just add. At certain points they add to zero at every instant, forever: the outgoing wave lifts the string exactly as much as the returning one pulls it down, and the cancellation never lets up. Those permanently dead points are nodes. Halfway between them the two waves always agree instead, so those points swing hardest — antinodes. Apart from the nodes, nothing has stopped moving. It only looks that way because the pattern of where-the-motion-is has stopped sliding along the string. If the underlying $v = f\lambda$ bookkeeping is shaky, the post on wave speed, frequency and wavelength is the place to firm it up first.

The string below is a guitar-scale one, fixed at both ends, driven continuously at whatever frequency you dial in. There is no "pluck" here and nothing decays — a small shaker holds the drive going forever, and physics decides how much the string bothers to respond.

Why Only Certain Frequencies Fit

The fixed ends are doing all the selecting. A point clamped to a bridge cannot move, so whatever shape the string settles into has to be zero at both ends — no exceptions, no negotiation. That is a boundary condition — a rule the surroundings impose on the wave, and no part of the wave's own nature.

Only some wavelengths can satisfy it. The longest is the one with a single arch spanning the string, which needs half a wavelength to fit in the length $L$. Next is a whole wavelength, then one and a half, and so on: $L = n\lambda/2$ for whole-number $n$. Every other wavelength arrives back at the far end out of step with itself, so the reflections keep partially cancelling and the string never builds anything up. Since $v = f\lambda$ and $v$ is fixed by the string's tension and thickness, that list of allowed wavelengths becomes a list of allowed frequencies — the harmonics, evenly spaced at $f_n = n f_1$.

This is the same argument that quantises anything confined: a wave in a box, an electron in an atom, a sound in a pipe. Confine a wave and the shapes it can hold stop being a continuum and become a countable list. It is also why any complicated shape you can pluck into a string is really a mixture of those harmonics in different proportions — which is the entire premise of building waveforms out of stacked sine waves in Fourier synthesis.

How Do Chladni Patterns Form? Sand Maps the Silence

In 1787 Ernst Chladni published a trick that still stops people in their tracks. Scatter fine sand over a metal plate, draw a violin bow down its edge, and the sand skitters about for a second and then arranges itself into a sharp geometric figure — stars, grids, rosettes — that changes completely when you bow a different note [1].

The temptation is to read the pattern as a picture of where the plate is shaking hardest. It is the exact opposite. Sand sitting on a violently vibrating patch of metal gets thrown into the air and lands somewhere else, over and over, in a slow random walk that only ends when it stumbles onto a piece of plate that is holding still. So the sand accumulates on the nodal lines and nowhere else. What you are looking at is a photograph of the silence — the two-dimensional version of the dead points on the string above, joined up into curves.

The plate below is square and driven at its centre, which limits it to the modes that actually move at the centre. Nothing is marked on the slider to begin with: the eight resonances are yours to find. A tick appears for each one only after you have crossed it.

Every one of those figures is a standing wave with the same anatomy as the string's, given a second dimension to spread into. The nodal lines are where two-dimensional waves bouncing off the plate's edges cancel permanently; the blank regions between them are the antinodes, and those are the parts of the plate actually pushing air around and radiating the sound. And exactly as on the string, the plate holds a discrete list of shapes with nothing in between: the diamond belongs to 80 Hz and to almost nothing either side of it.

Watch: Chladni Plates, Standing Waves and Resonance Explained

This narrated video slices a vibrating Chladni plate to show a guitar string hiding inside every sand line, builds the harmonic series from two travelling waves, and follows resonance from a mass on a spring to earthquakes, base isolation, Taipei 101's tuned mass damper and the Millennium Bridge.

What Is Resonance? Everything Has a Natural Frequency

Now the general statement. A natural frequency is a rate at which an object will keep oscillating on its own once you disturb it and let go, set by nothing but its stiffness and its mass. Resonance is what happens when something outside drives it at that rate: each push arrives in step with the motion already there, so instead of fighting it, every push adds to it. Small effort, repeated in time, becomes a large motion.

It is not a niche phenomenon. Almost anything you can name has a natural frequency, and the range is absurd:

A few of those deserve a sentence each. The playground swing is the one everybody has already solved with their body: anyone who has pushed a friend on one learned to push as the swing starts moving away, in step with the motion, because a push timed against it does nothing. The quartz crystal is 32,768 Hz because that is $2^{15}$ — halve it fifteen times with fifteen flip-flops and you land on exactly one tick per second, which is how a digital watch turns a shivering sliver of quartz into a second hand [2]. The cesium figure got promoted: it is not measured against the second, it defines it [3], and how an atomic clock locks onto that 9.19 GHz transition is resonance done to the highest precision anyone has managed. At the other extreme, the Earth's gravest normal mode ₀S₂ has a period near 54 minutes, and after a great earthquake the whole planet rings in it for days.

The interactive below strips resonance down to one mass on one spring, hanging from a support that shakes at a frequency you choose. Watch the amplitude by all means, but the thing worth catching is the timing: how far behind the support the mass is running, shown both by the block and by the two trace lines under the scene.

Damping is the knob that decides whether resonance is a curiosity or a catastrophe. Its measure is the quality factor $Q = 1/(2\zeta)$: roughly how many cycles of driving it takes to build the motion up, and roughly how many times bigger than a steady push the result gets. A wine glass has a $Q$ in the thousands, which is why it rings for seconds after you tap it and why a loud enough note at exactly its pitch can shatter it. A mattress barely rings at all — it swallows its bounce in a single cycle and does nothing interesting. Engineers spend a lot of their lives adding damping to structures for exactly this reason.

Why Do Earthquakes Destroy Some Buildings and Not Others?

On 19 September 1985 a magnitude 8.0 earthquake struck off Mexico's Pacific coast, hundreds of kilometres from Mexico City. Towns near the epicentre were damaged. The capital, far inland, was devastated — and devastated with a strange selectivity. Buildings roughly 6 to 15 stories tall collapsed in large numbers [5]. Shorter buildings beside them, and taller buildings beside those, often came through with cracked plaster. The 44-story Torre Latinoamericana, the city's most famous tower and its tallest for a quarter-century, was undamaged.

The explanation is underneath the city, not in it. Mexico City sits on the soft sediments of a drained lake, and those sediments filtered the incoming shaking — damping out the quick jolts and amplifying motion with a period near 2 seconds [4]. Whatever swayed at that tempo was in trouble.

Working out which buildings those were takes a moment of care, because the quick estimate and the real thing disagree. Engineers reach first for $T_n \approx 0.1$ s per story [6], and that crude code-level figure, calibrated on modern stiff frames, lands a 2-second period on a 20-story tower. The blocks that actually came down were older and far softer: reinforced-concrete mid-rises of 6 to 15 stories whose fundamental periods already sat around 1 to 2 seconds, stretched slower still by the lakebed clay beneath them, since the ground a building stands on is part of the system that oscillates. So the 6-to-15-story damage band and the 2-second ground motion are one story. It is the 0.1·N shortcut, not the physics, that fails to join them.

The Torre Latinoamericana is where even the shortcut gets the answer right. At 44 stories it comes out near 4.4 s, more than twice the lakebed's period and comfortably clear of it. The ground was shaking at the wrong tempo for that tower. It was not spared for being strong; it was spared for not listening.

The skyline below takes the shortcut at face value: four buildings of 2, 5, 10 and 20 stories, each with its natural period pinned at exactly 0.1 s per story. Its 2-second resonance therefore lands on the 20-story model, not on a real 1985 mid-rise — read the buildings as idealised oscillators wearing round numbers. There is deliberately no strength control; the only thing you can change is the tempo of the ground.

If the danger is a match, the defence is to spoil the match. Base isolation puts a building on flexible bearings, deliberately lengthening its natural period until it sits well away from the periods that local ground motion actually delivers. Tuned mass dampers take the opposite approach: hang a large weight inside the top of the tower on springs, tune it to the building's own natural frequency, and let it swing out of step with the building so it drains the energy instead of the structure. Taipei 101's is a 660-tonne steel sphere hanging in an atrium near the top, visible to tourists [7].

Footbridges have the same problem in miniature. When London's Millennium Bridge opened in 2000 it swayed sideways alarmingly, because pedestrians on a slightly wobbling deck instinctively adjust their step to keep balance — and the adjustment synchronised the crowd's footfall near the bridge's own ≈1 Hz lateral mode, which fed the wobble that caused the synchronisation. The bridge closed after two days and reopened with dampers fitted [8][9].

The Math of Standing Waves and Resonance

Everything above can be written as equations, and none of them is long. This section works through each one in the order the companion video shows it: two travelling waves, the harmonic series, pipes, the Chladni plate, the driven oscillator, buildings and bridges, and finally clocks and scanners. Numbers in square brackets point to the sources listed at the end of the post.

The Standing Wave Equation: y = 2A sin kx cos ωt

Put the clamp at $x = 0$ and send a sinusoidal wave of amplitude $A$, wavenumber $k = 2\pi/\lambda$ and angular frequency $\omega = 2\pi f$ towards it. A wave moving in the $-x$ direction has the form

$y_\text{in}(x,t) = A\sin(kx + \omega t)$

The clamp imposes one condition: $y(0,t) = 0$ at every instant. The incoming wave alone gives $y_\text{in}(0,t) = A\sin\omega t$, which is not zero, so the end has to launch a reflected wave that cancels it exactly, $y_\text{r}(0,t) = -A\sin\omega t$. The wave moving in the $+x$ direction that has that value at $x = 0$ is

$y_\text{r}(x,t) = -A\sin(\omega t - kx) = A\sin(kx - \omega t)$

It is the same wave running the other way, upside down. Add the two with the identity $\sin a + \sin b = 2\sin\tfrac{a+b}{2}\cos\tfrac{a-b}{2}$:

$y(x,t) = A\sin(kx - \omega t) + A\sin(kx + \omega t) = 2A\sin(kx)\cos(\omega t)$

That is the standing wave equation [10], and it is worth reading slowly. Position and time have separated into two factors. $2A\sin kx$ is a fixed shape that never slides along the string; $\cos\omega t$ is a single clock that every point shares. Each point just oscillates up and down in place with its own amplitude $2A\lvert\sin kx\rvert$. All the points inside one loop move in step, and neighbouring loops move in opposite directions.

• Nodes are where the shape factor is zero at all times: $\sin kx = 0$, so $kx = n\pi$ and $x = n\lambda/2$. Neighbouring nodes are half a wavelength apart.
• Antinodes sit midway between them, at $kx = (n + \tfrac12)\pi$, or $x = (2n+1)\lambda/4$. There the amplitude reaches $2A$, twice that of either travelling wave.

Now add the second clamp at $x = L$. The equation already has a node at $x = 0$; it needs another at $x = L$, so $\sin kL = 0$, which means $kL = n\pi$. Substituting $k = 2\pi/\lambda$ gives exactly the fitting rule from earlier: $L = n\lambda/2$.

Why a Fixed End Inverts the Reflected Wave

The algebra forced that minus sign, and the physics behind it is Newton's third law. When an upward pulse reaches the clamp, the string pulls up on the support. The support cannot move, so it pulls back down on the string just as hard, and that downward pull launches an inverted pulse back along the string [10]. A free end, such as a ring that slides on a frictionless rod, is the opposite case: nothing pulls back, the end overshoots, and the pulse returns the right way up. So a free end is an antinode rather than a node. That one difference separates the open and closed ends of the pipes below, and it is also why the edge of a Chladni plate behaves differently from the end of a string.

Harmonic Frequencies of a String: fₙ = nv/2L

For a string of length $L$ fixed at both ends, the boundary conditions allow $L = n\lambda/2$, and with $v = f\lambda$ that gives the harmonic series directly:

$f_n = \frac{nv}{2L}, \qquad n = 1, 2, 3, \dots$

The wave speed comes from the string itself — tension $T$ and mass per unit length $\mu$ [10]:

$v = \sqrt{\frac{T}{\mu}}$

Tighten the string and every harmonic rises; use a thicker, heavier string and they all fall. That is the whole design of a guitar's six strings in one line.

Open and Closed Pipe Harmonics: Why a Clarinet Has Only Odd Harmonics

Air columns work the same way with different boundary conditions. A pipe open at both ends has an antinode at each end and gives every harmonic: $f_n = nv/2L$ with $n = 1, 2, 3, \dots$ as before. A pipe closed at one end has a node at the closed end and an antinode at the open one, which needs a quarter wavelength rather than a half, so it gives $f_n = nv/4L$ for odd $n$ only — a fundamental an octave lower than an open pipe of the same length, with the even harmonics missing. A real clarinet's even harmonics are weak rather than strictly absent, at least in its low register, and that missing-even-harmonic signature is why a clarinet (a closed pipe) sounds hollow next to a flute (an open one) of similar size [12][13].

Here is where those rules come from. In a pipe, the thing that waves is the air's back-and-forth displacement along the bore. A closed end stops the air moving, so it is a displacement node, the counterpart of the clamp. An open end lets the air move freely, so to a good approximation it is a displacement antinode [11]. An open–open pipe needs antinode-to-antinode fits, which are whole numbers of half wavelengths, just like the string. A closed–open pipe needs node-to-antinode fits: the shortest is a quarter wavelength, and each longer one adds another half wavelength:

$L = \frac{\lambda}{4},\ \frac{3\lambda}{4},\ \frac{5\lambda}{4},\ \dots \quad\Longrightarrow\quad f_n = \frac{nv}{4L}, \qquad n = 1, 3, 5, \dots$

Putting $n = 1$ into both formulas gives $f_1(\text{closed}) = v/4L = \tfrac12 f_1(\text{open})$, which is the octave drop. Real open ends behave as if the pipe were slightly longer than it is. This end correction is about 0.6 times the radius for an unflanged pipe [11], so measured pipes come out a little flat of $nv/2L$.

A clarinet counts as a closed pipe because its reed end acts as a closed end [12]. The reed's vibration contains even harmonics as well as odd ones. In the low register, though, only the odd harmonics line up with resonances of the bore and get reinforced, which is why the second and fourth harmonics come out weak but not missing [14].

The Chladni Plate Equation: Mode Shapes and Nodal Lines

The plate simulation on this page uses a deliberately simple model, and it is the same one the video draws. On a unit square, $0 \le x \le 1$ and $0 \le y \le 1$, each mode is labelled by two whole numbers $(m, n)$ and has the shape

$\phi_{mn}(x,y) = \cos(m\pi x)\cos(n\pi y) + \cos(n\pi x)\cos(m\pi y)$

That is a cosine across times a cosine down, plus the same pattern turned through 90°. On a square plate, $(m, n)$ and $(n, m)$ have the same frequency, so the two can appear together. Adding them is the trick Charles Wheatstone used in 1833 to explain Chladni's square-plate figures [15], and Rayleigh later treated it in The Theory of Sound [16]. Once it is vibrating, the plate's displacement is $z(x,y,t) = w(x,y)\cos\omega t$: a fixed shape times a shared clock, the same separation as $2A\sin kx\cos\omega t$ on the string.

Three details of the model matter:

• Only both-even modes respond. The plate is driven at its centre, and a driver there can only feed a mode that actually moves at the centre. At $(\tfrac12, \tfrac12)$ the mode shape is $\phi_{mn} = 2\cos(m\pi/2)\cos(n\pi/2)$, which is zero whenever $m$ or $n$ is odd. That leaves eight modes between 60 and 1100 Hz: (0, 2) at 80 Hz, (2, 2) at 160, (0, 4) at 320, (2, 4) at 400, (4, 4) at 640, (0, 6) at 720, (2, 6) at 800 and (4, 6) at 1040 Hz.
• Frequency grows with $m^2 + n^2$. Every one of those frequencies is $f_{mn} = 20(m^2 + n^2)$ Hz. That quadratic is the signature of a bending plate. Bending waves are dispersive: their angular frequency grows with the square of the wavenumber, $\omega = k^2\sqrt{D/\rho h}$, where $D$ is the plate's flexural rigidity, $\rho$ its density and $h$ its thickness [17]. On the unit square $k^2 = \pi^2(m^2 + n^2)$, so $f \propto m^2 + n^2$. A string has $\omega = vk$ instead, which is why its harmonics are evenly spaced and a plate's are not.
• The response mixes modes with Lorentzian weights. At drive frequency $f$ the shape is $w(x,y) = \sum_k w_k(f)\,\phi_k(x,y)$, where

$w_k(f) = \frac{1}{\sqrt{(f_k^2 - f^2)^2 + (f_k f/Q)^2}}, \qquad Q = 30$

This is the driven-oscillator response from the driven-oscillator subsection below, written in terms of $f$ with $2\zeta = 1/Q$. Near a resonance one term dwarfs the rest, and the plate takes that mode's shape. A $Q$ of 30 is what makes the simulation's patterns so picky: at 84 Hz, 5% above the (0, 2) mode, that mode's weight has already fallen to about 0.31 of its peak, the "about a third" quoted in the plate walkthrough above.

This is a simplified model, not the exact physics of a real plate. Cosines have zero slope at the edges, which copies one property of a free edge. A truly free edge must also carry no bending moment and no edge shear, and with those conditions the free square plate has no closed-form solution. Walther Ritz solved it in 1909, numerically, using the variational technique now called the Ritz method, and his computed nodal patterns reproduced Chladni's figures [18][17]. The $m^2 + n^2$ frequency ladder is borrowed from the simply supported plate, whose modes really are exact products of sines [17]. The model gets the look and the logic of Chladni figures right, but it will not predict the frequencies of a real brass plate.

The nodal lines are the curves where $w(x,y) = 0$: points that never move, whatever the clock $\cos\omega t$ is doing. For the diamond, mode (0, 2), the shape is $\phi_{02} = \cos(2\pi y) + \cos(2\pi x)$, since $\cos 0 = 1$. It vanishes where $\cos 2\pi x = -\cos 2\pi y$, which happens along four straight lines: $x + y = \tfrac12$, $x + y = \tfrac32$, $x - y = \tfrac12$ and $y - x = \tfrac12$. Together they make a square tilted by 45°, with its corners at the midpoints of the plate's edges. That is the diamond the sand draws at 80 Hz.

Slicing a Chladni Plate: Every Cut Is a Standing Wave

Fix $y = y_0$ and look along the cut. Now $\cos(n\pi y_0)$ and $\cos(m\pi y_0)$ are just numbers, so

$w(x, y_0) = \underbrace{\cos(n\pi y_0)}_{a}\,\cos(m\pi x) + \underbrace{\cos(m\pi y_0)}_{b}\,\cos(n\pi x)$

Each term is a standing wave along $x$ in its own right, because $\cos(m\pi x)\cos\omega t = \tfrac12[\cos(m\pi x - \omega t) + \cos(m\pi x + \omega t)]$ is two counter-travelling waves with wavenumber $m\pi$. Both terms share the same clock $\cos\omega t$, so the zeros of $a\cos(m\pi x) + b\cos(n\pi x)$ do not depend on time. The cut has fixed nodes, which is the defining property of a standing wave. Because it mixes two wavenumbers it is in general not a single clean sinusoid, so its nodes are not evenly spaced. As $y_0$ changes, $a$ and $b$ change, the shape changes, and the nodes slide. In the video the cut runs through mode (2, 4) at 400 Hz, where $a = \cos(4\pi y_0)$ and $b = \cos(2\pi y_0)$.

The diamond shows it most simply. Its cut is $w(x, y_0) = \cos(2\pi y_0) + \cos(2\pi x)$, a single cosine riding on a constant, and for $0 < y_0 < \tfrac12$ its nodes are at

$x = \tfrac12 - y_0 \qquad\text{and}\qquad x = \tfrac12 + y_0$

Slide the cut from the bottom edge towards the middle and the two nodes walk apart along straight lines. Join up their positions for every $y_0$ and you get the diamond. That is the precise sense in which a Chladni figure is the nodes of every slice, joined up.

The one real difference from the string is at the ends of the cut. Differentiate along it:

$\frac{\partial w}{\partial x} = -m\pi a\sin(m\pi x) - n\pi b\sin(n\pi x)$

This is zero at $x = 0$ and at $x = 1$ for every $y_0$, because $\sin 0 = \sin m\pi = \sin n\pi = 0$. The cut always meets the edge flat (zero slope), usually at a crest or trough of its shape. These are free ends, like the antinodes at the open end of a flute, not clamped ends like a guitar string's.

Why Sand Collects on Nodal Lines

A grain resting on the plate feels gravity and the plate's push, and the plate can push but never pull. If the point under the grain moves as $z = Z\cos\omega t$, its acceleration is $\ddot z = -\omega^2 Z\cos\omega t$. The contact force needed to keep the grain riding along is $m(g + \ddot z)$, and it can never go negative. Once the plate's downward acceleration exceeds $g$, the plate falls away faster than gravity can pull the grain after it, and the grain is launched. The condition for that is

$\omega^2 Z > g \quad\Longleftrightarrow\quad Z > \frac{g}{\omega^2}$

At 400 Hz, $\omega = 2\pi \times 400 \approx 2513$ rad/s and $g/\omega^2 \approx 1.6\ \mu\text{m}$. A vibration amplitude of under two thousandths of a millimetre is enough to throw sand. So almost the whole plate is a launch pad, and the only refuges are thin bands on either side of a nodal line, where $Z$ falls below that threshold. That is why the lines come out sharp.

Each bounce lands a grain somewhere slightly random, so on the moving parts of the plate the sand does a random walk. Near a node the hops shrink to nothing, and grains that arrive there stay. The simulation models this directly with the plate's intensity

$I(x,y) = \left(\frac{w(x,y)}{w_\text{max}}\right)^2$

which runs from 0 on a nodal line to 1 at the most violent point. Squaring throws away the sign, and that is deliberate: a grain does not care whether its patch is moving up while its neighbour moves down, only how hard it is moving. At every step, each grain takes a random hop that grows with $I$, plus a small drift down the slope of $I$, so the sand ends up where $I \approx 0$, on the nodal lines. This biased random walk stands in for the real bouncing. It does not simulate the grains' actual flight.

One historical footnote. In 1831 Michael Faraday showed that very light powders, such as lycopodium, do the opposite and gather at the antinodes. Air currents set up by the vibrating plate carry them there, while heavy sand ignores the air and goes to the nodes [19].

Driven Damped Oscillator: Amplitude, Phase and Q Factor

For the driven oscillator, a mass $m$ with natural frequency $\omega_0$ and damping ratio $\zeta$ pushed by a force of amplitude $F_0$, the equation of motion is $m\ddot x + 2\zeta m\omega_0\dot x + m\omega_0^2 x = F_0\cos\omega t$, and its steady-state amplitude is [20]

$A(\omega) = \frac{F_0/m}{\sqrt{(\omega_0^2 - \omega^2)^2 + (2\zeta\omega_0\omega)^2}}$

and the phase by which the response lags the drive is

$\tan\varphi = \frac{2\zeta\omega_0\omega}{\omega_0^2 - \omega^2}$

At $\omega = \omega_0$ the denominator of that fraction vanishes, so $\varphi = 90°$ exactly — the quarter-cycle lag from the simulation, independent of $\zeta$. The sharpness of the peak is the quality factor

$Q = \frac{1}{2\zeta}$

and the amplitude peak does not sit precisely at $\omega_0$ once damping matters. Differentiating $A(\omega)$ gives

$\omega_r = \omega_0\sqrt{1 - 2\zeta^2}$

which is real only for $\zeta < 1/\sqrt{2} \approx 0.707$. At or above that there is no peak at all: the response just falls away from its static value as the drive speeds up. For the lightly damped systems most exam questions care about, $\zeta \ll 1$ and $\omega_r \approx \omega_0$, which is why the distinction is usually swept under the carpet.

The same formula backs up the claim that $Q$ is roughly how many times bigger than a steady push the result gets. A steady force ($\omega = 0$) deflects the mass by $A(0) = F_0/(m\omega_0^2)$. At $\omega = \omega_0$ the first term under the square root vanishes, so only damping is left holding the motion back, and $A(\omega_0) = F_0/(2\zeta m\omega_0^2)$. The ratio is

$\frac{A(\omega_0)}{A(0)} = \frac{1}{2\zeta} = Q$

With the simulation's default $\zeta = 0.05$ that gives $Q = 10$, exactly the 10.0× gain it shows at 1.00 Hz. At $\zeta = 0.40$ it gives 1.25×, also matching the simulation.

Building Natural Period: The T ≈ 0.1N Rule and the Mexico City Lakebed

The $T_n \approx 0.1N$ seconds rule used in Example 2 below is not folklore. The US seismic loading standard ASCE/SEI 7-16 allows it as an approximate fundamental period (Section 12.8.2.1, Eq. 12.8-8), but only for concrete or steel moment-resisting frames of no more than 12 stories, with an average story height of at least 3 m [6]. A ten-story frame is inside that scope. The 44-story Torre Latinoamericana is far outside it, so its $0.1 \times 44 = 4.4$ s is an extrapolation of the rule. Read it as "several seconds, comfortably clear of 2 s", not as a measurement.

Why did the lakebed favour 2 seconds? For shear waves, a layer of soft soil of thickness $H$ on top of much stiffer ground behaves like the closed pipe above. The ground surface is free, so it is an antinode, and the stiff base barely moves, so it is close to a node. The fundamental mode fits a quarter wavelength into the layer, which gives a natural period of

$T_s = \frac{4H}{V_s}$

where $V_s$ is the shear-wave speed in the soil [21]. At the SCT strong-motion station, whose soil best matched the worst-hit zone, Seed and colleagues describe about 38 m of soil above the hard layer, mostly very soft clay. They inferred an average shear-wave speed in that clay of about 75 m/s from the site's natural period [4]. Putting those numbers back in gives $T_s \approx 4 \times 38 / 75 \approx 2.0$ s, which is consistent (as it should be, since the 75 m/s was itself inferred from the period). The recordings agree. The motion on firm ground at the National University already had a predominant period of about 2 s and a peak spectral acceleration (5% damping) of about 0.11 g. At SCT, the clay amplified that to about 1.0 g at a period of about 2 s, roughly ninefold [4][22]. The same study puts the major damage in the heavy-damage zone in structures of about 6 to 18 stories [4]. A survey by engineers from the National Autonomous University of Mexico (UNAM), reported by Beck and Hall, gives 6 to 15 stories [5], the range quoted earlier.

A tuned mass damper adds a second oscillator to the first. It is tuned so its natural frequency sits close to the building's, which means the building's sway drives it at resonance. It then swings out of step with the tower, so the force it exerts opposes the tower's motion and the energy goes into the damper's own dashpots instead of the structure. Taipei 101's pendulum hangs between the 87th and 92nd floors and has a mass of 660 tonnes [7].

The Millennium Bridge needed no earthquake. Walking pushes on the ground at the pacing rate, about 2 Hz vertically, but the body's side-to-side sway alternates with each foot, so the sideways force comes at half that rate, near 1 Hz. The engineers' investigation found that lateral modes below about 1.3 Hz were susceptible to that sideways footfall, and the bridge's own lateral modes sat well inside that range. Once the deck began to move, walkers adjusted their gait to its motion and so synchronised their sideways push with it [8]. Strogatz and colleagues later modelled the crowd as coupled oscillators and showed that this synchrony switches on suddenly once the crowd passes a critical size [9].

Resonant Frequencies in Watches, Radios, MRI Scanners and Atomic Clocks

• Quartz watch. $32\,768 = 2^{15}$, so a chain of fifteen divide-by-two stages turns the crystal's oscillation into exactly one pulse per second: $32\,768 / 2^{15} = 1$ Hz. That is why watch crystals are cut for 32,768 Hz [2].
• Radio. An inductor $L$ and a capacitor $C$ form an electrical oscillator with resonant frequency $f_0 = 1/(2\pi\sqrt{LC})$ [10]. With $L = 0.10\ \mu\text{H}$ and $C = 25$ pF, $f_0 \approx 101$ MHz, inside the FM broadcast band. Turning the tuning knob changes $C$, and with it the one station the circuit answers.
• MRI. A proton in a magnetic field $B$ precesses at the Larmor frequency $f = (\gamma_p/2\pi)B$, with $\gamma_p/2\pi = 42.577\,478\,461$ MHz/T (CODATA 2022) [23]. At 1.5 T that is $42.5775 \times 1.5 \approx 63.87$ MHz, the "about 64 MHz" in the video. Protons in tissue are shielded by their electrons, which shifts this by a few tens of parts per million. That matters for chemistry, not for this estimate.
• Caesium clock. The second has been defined by caesium since 1967. The current SI wording (2019) fixes the unperturbed ground-state hyperfine frequency of caesium-133 at exactly $\Delta\nu_\text{Cs} = 9\,192\,631\,770$ Hz [3]. One period is $1/\Delta\nu_\text{Cs} \approx 1.088 \times 10^{-10}$ s, about 109 ps, and one second is exactly 9,192,631,770 of them.

Real-World Applications of Resonance

Musical instruments are resonance made into a craft. A guitar body, a violin's f-holes, a saxophone's cone and a drum's shell all exist to have their own resonances at the right frequencies, so the tiny amount of energy in a vibrating string or reed is turned into sound instead of dying quietly.

MRI is resonance at the nuclear scale. Protons in a magnetic field precess at 42.58 MHz per tesla, so a 1.5 T scanner tickles them at 63.87 MHz [23] and nothing else in the body answers. Change the field slightly across the patient and the resonant frequency changes with it, which is how the machine turns a frequency into a position.

Radio tuning is a variable capacitor sweeping the resonant frequency of an LC circuit across the band. Every station's signal is present in the antenna at once; the circuit is simply built to respond enormously to one of them and negligibly to the rest.

Ultrasonic cleaning drives a tank of liquid at tens of kilohertz, forming and collapsing cavitation bubbles that scrub into crevices no brush reaches — jewellery, spectacle frames, surgical instruments, engine parts.

A shattered wine glass needs three things at once: a note at the glass's own frequency, enough amplitude, and a glass with a high enough $Q$ to build the motion up over many cycles. Volume alone will not do it; the pitch has to be right, which is why the demonstration usually starts with someone tapping the glass to find out what note to sing.

The Tacoma Narrows bridge is the example every textbook reaches for and the one to be most careful with. Its 1940 collapse was not simple forced resonance — the wind was steady, not oscillating at the bridge's natural frequency. What happened was aeroelastic flutter: the deck's own twisting motion shed vortices that pushed it further in the direction it was already going, so the airflow fed energy in continuously as a function of the motion itself. That is a self-excited instability rather than a matched drive, and it is why the bridge is a better story about aerodynamics than about resonance.

References and Further Reading

1. E. F. F. Chladni, Entdeckungen über die Theorie des Klanges (Leipzig: Weidmanns Erben und Reich, 1787).
2. M. A. Lombardi, "The Accuracy and Stability of Quartz Watches," Horological Journal, February 2008, pp. 57–59. NIST publication: tsapps.nist.gov/publication/get_pdf.cfm?pub_id=50647.
3. Bureau International des Poids et Mesures, The International System of Units (SI Brochure), 9th edition (BIPM, 2019); definition of the second: bipm.org/en/si-base-units/second.
4. H. B. Seed, M. P. Romo, J. Sun, A. Jaime and J. Lysmer, Relationships between Soil Conditions and Earthquake Ground Motions in Mexico City in the Earthquake of September 19, 1985, Report UCB/EERC-87/15 (Earthquake Engineering Research Center, University of California, Berkeley, October 1987). NIST NEHRP Clearinghouse PDF.
5. J. L. Beck and J. F. Hall, "Engineering Features of the Recent Mexican Earthquake," Engineering & Science (Caltech) 49(3), 2–9 (January 1986). calteches.library.caltech.edu/3478.
6. American Society of Civil Engineers, Minimum Design Loads and Associated Criteria for Buildings and Other Structures, ASCE/SEI 7-16 (Reston, VA: ASCE, 2017), Section 12.8.2.1. doi:10.1061/9780784414248.
7. Guinness World Records, "Largest tuned mass damper" (Taipei 101). guinnessworldrecords.com.
8. P. Dallard, A. J. Fitzpatrick, A. Flint, S. Le Bourva, A. Low, R. M. Ridsdill Smith and M. Willford, "The London Millennium Footbridge," The Structural Engineer 79(22), 17–33 (2001). See also the same team's "London Millennium Bridge: Pedestrian-Induced Lateral Vibration," Journal of Bridge Engineering 6(6), 412–417 (2001), doi:10.1061/(ASCE)1084-0702(2001)6:6(412).
9. S. H. Strogatz, D. M. Abrams, A. McRobie, B. Eckhardt and E. Ott, "Crowd synchrony on the Millennium Bridge," Nature 438, 43–44 (2005). doi:10.1038/438043a.
10. D. Halliday, R. Resnick and J. Walker, Fundamentals of Physics, 10th edition (Hoboken, NJ: Wiley, 2014). Chapters 16–17 cover travelling and standing waves, strings and pipes; Chapter 31 covers LC oscillations.
11. L. E. Kinsler, A. R. Frey, A. B. Coppens and J. V. Sanders, Fundamentals of Acoustics, 4th edition (New York: Wiley, 2000).
12. T. D. Rossing, F. R. Moore and P. A. Wheeler, The Science of Sound, 3rd edition (San Francisco: Addison Wesley, 2002).
13. N. H. Fletcher and T. D. Rossing, The Physics of Musical Instruments, 2nd edition (New York: Springer, 1998). doi:10.1007/978-0-387-21603-4.
14. J. Wolfe, "Clarinet acoustics: an introduction," Music Acoustics, University of New South Wales. newt.phys.unsw.edu.au/jw/clarinetacoustics.html.
15. C. Wheatstone, "On the figures obtained by strewing sand on vibrating surfaces, commonly called acoustic figures," Philosophical Transactions of the Royal Society of London 123, 593–633 (1833). doi:10.1098/rstl.1833.0027.
16. J. W. Strutt, Baron Rayleigh, The Theory of Sound, 2nd edition, Vol. I (London: Macmillan, 1894). Chapter X treats the vibrations of plates.
17. A. W. Leissa, Vibration of Plates, NASA SP-160 (Washington, DC: NASA, 1969). ntrs.nasa.gov/citations/19700009156.
18. W. Ritz, "Theorie der Transversalschwingungen einer quadratischen Platte mit freien Rändern," Annalen der Physik 333(4), 737–786 (1909). doi:10.1002/andp.19093330403.
19. M. Faraday, "On a peculiar class of acoustical figures; and on certain forms assumed by groups of particles upon vibrating elastic surfaces," Philosophical Transactions of the Royal Society of London 121, 299–340 (1831). doi:10.1098/rstl.1831.0018.
20. J. R. Taylor, Classical Mechanics (Sausalito, CA: University Science Books, 2005), Chapter 5 (driven damped oscillations and resonance).
21. S. L. Kramer, Geotechnical Earthquake Engineering (Upper Saddle River, NJ: Prentice Hall, 1996).
22. H. B. Seed, M. P. Romo, J. I. Sun, A. Jaime and J. Lysmer, "The Mexico Earthquake of September 19, 1985—Relationships between Soil Conditions and Earthquake Ground Motions," Earthquake Spectra 4(4), 687–729 (1988). doi:10.1193/1.1585498.
23. NIST, "CODATA Value: proton gyromagnetic ratio in MHz/T" (2022 CODATA recommended values). physics.nist.gov/cgi-bin/cuu/Value?gammapbar.

Standing Waves and Resonance Quick Reference