A drum skin, stretched square. Strike a note and the whole surface dances, except along certain secret curves that stay perfectly still. Scatter sand on the drum, and every dancing patch flings its grains away, while the still curves collect them like harbors. The sand is not drawing the sound; it is drawing the silence inside the sound, and every note keeps a different silence. Change the note. Blend two notes with the dial. Watch the sand hurry to redraw the map.
For the curious kid.
In 1787 Ernst Chladni (musician, physicist, and traveling showman) drew a violin bow across the edge of a brass plate dusted with sand, and the sand leapt into stars, crosses, and flowers. Crowds paid to watch. Napoleon paid too, and was so unsettled that he funded a prize for whoever could explain the patterns. Here is the secret Chladni knew: a surface can't vibrate everywhere. When it rings at one pure note, it swings up here and down there, and in between there must be lines that do not move at all, like the still point of a seesaw. Sand gets kicked off every moving patch and survives only on those still lines. So the pattern you see is a portrait of one single note, and each note wears a different face. Press The geometry to see the mathematics draw, in gold, the same lines the sand found by being bounced around blindly.
Deeper.
This drum is the mathematician's plate: a membrane pinned at its edges, whose vibration modes are exactly sin(nπx)·sin(mπy) with pitch proportional to √(n²+m²). Chladni's actual brass plates, free at the edges, stiff rather than stretched, are far crueler: no closed formula for their patterns exists at all. That is why Napoleon's prize stood unclaimed for years, until Sophie Germain, self-taught, barred from the academies for being a woman, submitting at first under a man's name, built the theory of elastic surfaces that won it in 1816. The curved figures come from a degeneracy: notes (n,m) and (m,n) have identical pitch, so the drum can sing any blend of the two (that is the dial) and each blend holds its own nodal curves. Notice, too, that the pitches √2, √5, √8, √10… are not multiples of one another: unlike a string, a drum's overtones are out of tune with each other, which is exactly why drums thud while strings sing.
Museum policy note: the sand is honest physics, not stagecraft. Each grain random-walks with step size proportional to the local vibration amplitude; diffusion of that kind provably concentrates grains where the amplitude vanishes (stationary density ∝ 1/|u|). Validated headless before this page shipped: the 1D version piles 63× more mass on the node than off it, and on this drum 96.8% of grains settle onto the 20% of surface where |u| < 0.1, a 4.7× enrichment, from nothing but being shaken. The gold lines are solved from u = 0 directly and are never fitted to the sand.
Every note the drum can sing, it also keeps a map of where it must stay silent. The sand only reads the map aloud.