The Sound of the Śrī Yantra

§ 1 · A legend written in sand

Can a sound draw this figure?

There is a story that travels wherever sound meets sacred geometry. Sing the syllable Oṁ into a tonoscope — a drum-like membrane dusted with fine powder — and the grains, the story says, arrange themselves into the Śrī Yantra: nine interlocking triangles around a central point, the great diagram of the Śrīvidyā tradition. The tale is usually traced back to the cymatics experiments of Hans Jenny in the 1960s, and you will find it repeated in books, documentaries, and a thousand yoga-studio posters.

This essay takes the legend seriously enough to do the physics. Not to knock it down with a shrug, but to ask the precise question hiding underneath: what would a vibration actually have to be, for sand to settle into the true, classical Śrī Yantra? The answer turns out to be exact, surprising, and — in its own way — more beautiful than the legend.

You need only three ideas from a good physics or math class: what a sine wave is, how vectors point, and the fact that waves add together. Everything else is built on this page. The green boxes suggest exactly what to type into each interactive figure, and every "show the work" panel folds open to reveal the math behind a claim. Nothing here is an image: every yantra is re-computed live in your browser.

§ 2 · Reading a plate

How sand reads a vibrating surface

In 1787 Ernst Chladni drew a violin bow across the edge of a metal plate sprinkled with sand, and watched the grains leap away from the moving parts and gather along still curves. The explanation is standing waves. Drive a plate at one of its natural frequencies and it settles into a mode: a pattern in which some regions swing up and down with full force while other points — the nodal lines — do not move at all. Sand gets kicked around by the vibrating regions until, by pure chance, it lands somewhere quiet, and there it stays. The sand pattern is literally a photograph of the equation ψ(x, y) = 0, where ψ is the plate's up-and-down motion at each point.

A square plate makes this concrete. Its simplest modes are products of cosines — one factor counting half-waves across the plate, the other counting them down:

ψ(x, y) = cos(mπx) cos(nπy) ± cos(nπx) cos(mπy)x and y run from 0 to 1 across the plate; m and n are whole numbers counting the half-waves

The ± matters. At many frequencies the plate has two different modes with the exact same pitch — the pattern (m, n) and its mirror twin (n, m) — and it is free to vibrate in any blend of the two. The famous curvy Chladni figures are exactly such blends.

§ 3 · The control case

Why sound loves the Star of David

Now aim at triangles. A perfectly straight nodal line needs plane waves — waves whose crests are straight lines, all marching in one direction — and interlocking triangles need exactly three wave directions. So take three plane waves of the same wavelength, point them 120° apart, and give each a head start (a phase) of 2π/3:

ψ = sin(k e₁·r + 2π/3) + sin(k e₂·r + 2π/3) + sin(k e₃·r + 2π/3)e₁, e₂, e₃ are unit vectors 120° apart; r = (x, y) is position; the pitch k sets the wavelength λ = 2π/k

Something almost magical happens: this sum of three waves factors into a product of three sine sheets. A product is zero wherever any factor is zero, and each factor is zero along a family of straight, evenly spaced, parallel lines. Three families at 60° to one another: the sand draws an endless lattice of interlocking up-and-down triangles. Hexagram country.

This lattice is the figure vibration draws effortlessly, and it is what cymatics rigs actually produce: Star-of-David tilings, honeycombs, and their cousins. To nest triangles of different sizes you need different wavelengths — a chord of pitches — because triangle size is proportional to wavelength, which is proportional to 1/frequency. And the tones must be struck one at a time, each pattern photographed and the photos stacked. Struck together, the three fields interfere, and sand can only rest at the rare spots where all three are quiet at once: dots, not lines.

§ 4 · The strict figure

The Śrī Yantra as a system of equations

Everything Plate II can draw is equilateral, evenly spaced, and endlessly repeating. The Śrī Yantra refuses all three — and the refusal is precise, because the classical figure is not a freehand drawing but a system of constraints, studied in a small but serious mathematical literature from Bolton & Macleod in the 1970s through Kulaichev, Gérard Huet, and most recently Alessandro Chiodo's 2021 construction by straightedge and compass.

Here is the figure, stated as math. Nine triangles t₁ … t₉ share a vertical axis of symmetry, numbered by the height of their bases; t₁ through t₅ point down (the śakti triangles), t₆ through t₉ point up (the śiva triangles). Their interlocking must produce exactly 43 small triangles: the central trikoṇa holding the bindu, then rings of 8 (aṣṭakoṇa), 10 (antardaśāra), 10 (bahirdaśāra), and 14 (caturdaśāra). For those cells to close crisply, three families of rules must hold exactly:

(i) triangles t₃ and t₇ fit perfectly inside the same circle — apexes on its poles, base corners on its rim. (ii) Seven apexes must land exactly on other triangles' base lines: t₈'s apex on t₁'s base, t₆'s on t₂'s, t₉'s on t₃'s, t₁'s on t₆'s, t₅'s on t₇'s, t₄'s on t₈'s, and t₂'s on t₉'s. (iii) At twelve special points, three edges from three different triangles must pass through a single point — the triple concurrencies. The drawing tradition calls these points marma-sthāna — the same word Ayurveda uses for the body's vital junctions, where channels meet and injury cascades, and Vastu uses for the untouchable intersections of the vāstupuruṣamaṇḍala. The borrowing is exact: a marma is where the figure is most alive and most vulnerable, because an error at any one of the twelve spreads through the entire construction and breaks the cells.

Now count freedom. Nine triangles need 27 numbers to describe (each has a base height, a width, and an apex height). The rules above use up 23 of them. What is left is a family with just four free choices. The classical choice — Huet's — places the bases of t₃, t₆, t₇, t₉ at 0.332, 0.537, 0.602, 0.835 of the way down the vertical diameter. One last equation then remains, and it is equivalent to a problem Apollonius posed twenty-two centuries ago: find a circle tangent to a given circle and a given line, passing through a given point. That is why the strict yantra can be drawn with straightedge and compass — and why, on this page, your browser simply solves the last equation numerically every time you touch a slider.

Three things in this table matter enormously for the physics. First, every apex angle is different — nine triangles, nine shapes, no repeats (t₈ lands, by pure accident, within a third of a degree of equilateral — highlighted above). Second, the nine base heights share no common measure: they are messy algebraic numbers full of nested square roots, and the traditional draughtsman's grid of 48 equal divisions was only ever an approximation to them. Third, the bindu — computed as the incentre of the central triangle — sits 0.042 R above the circle's centre. Concurrency alone does not centre the figure, which is why "optimal" versions in the modern literature add two more demands: a centred bindu and an exactly equilateral central triangle. (The appendix at the end of this page solves those.)

§ 5 · Three prohibitions

What a nodal line is allowed to do

Now bring the solved figure to the plate. A vibration at a single frequency is a mix of plane waves that all share one wavelength; whatever directions and phases you choose, the resulting still-lines obey three hard rules. All three come from one idea: a straight nodal line acts as a mirror of odd symmetry — the plate's motion on one side is the upside-down mirror image of the other side — and symmetries, once present, spread.

Rule one: full chords only. If the plate is exactly still along a straight stretch, the mirror rule extends that stretch: it cannot simply stop at a point in the middle of the plate. Every straight nodal line runs boundary to boundary. But the yantra's edges must stop — every triangle side ends at a corner inside the plate. Whatever sound draws, it will draw each edge continued to the rim: the full chord, not the segment.

Rule two: parallels live on one ladder. Look across a family of parallel still-lines, perpendicular to them. Each wave component contributes a plain sine along that direction, and a sine crosses zero at perfectly equal steps. So within one tone, parallel nodal lines must sit on a single ladder of equal spacing — and every rung of the ladder prints, wanted or not. Lines at unrelated spacings cannot share a tone.

Rule three: crossings at 90°, or the equilateral trio. Two crossing nodal lines are two mirrors, and two mirrors meeting at an angle generate a rotation of twice that angle. Unless the angle divides 180° evenly, the copies never stop multiplying and the only field that can obey them all is zero everywhere. If the angle does divide evenly, the whole family of mirror copies prints — ghost spokes radiating through the crossing. Exactly two clean escapes exist: perpendicular crossings, where the field factors as sin(ax) sin(by) with the two spacings free; and the 120° trio of § 3, where the factorization forces equilateral triangles with all spacings equal.

Now hold the yantra's census against the rules. Its nine base heights share no common measure — rule two forbids them from sharing a tone. Its eighteen leg directions are all different, none at 90° to any other line of the figure, and no three of its directions form the equilateral trio — rule three forbids any two crossing edges from sharing a tone. Even the two legs of one triangle, mirror twins though they are, meet at an apex angle that is not a neat divisor of 180°, so pairing them under one tone would spray ghost spokes through the apex. The strict figure breaks all three prohibitions everywhere, at once. One tone — indeed any chord of tones on a uniform plate — cannot draw it.

§ 6 · The bill

One tone, twenty-seven bearings

Impossibility, made precise, becomes a price list. If no two edges may share a tone, give each edge its own exposure, just as the nested hexagrams demanded in § 3 — only now the trick runs in a different dimension. The recipe for one straight line is a pair of plane waves skimming along it at shallow angles ±β, phased so that their combined stillness lands exactly on the target. The pair's still-lines repeat, so choosing β small enough,

sin β < λ / 2Lλ = wavelength, L = width of the plate — the condition that pushes every ghost parallel off the plate

leaves exactly one line on the plate: the full chord through the target edge, positioned not by frequency but by phase.

That prices the figure precisely:

The last row is a refinement worth showing. Two parallel lines can share one ladder — if both of the ladder's next rungs fall off the plate. On a plate of radius 1, that means the pair (y₁, y₂) must satisfy 2y₁−y₂ ≤ −1 and 2y₂−y₁ ≥ +1, and exactly three pairs of the yantra's bases qualify, each joining a deep śiva-side line to a high śakti-side one: {−0.670, +0.208}, {−0.470, +0.336}, {−0.204, +0.774}. The true minimum is therefore twenty-four exposures and forty-eight waves; twenty-seven is the honest edge-by-edge prescription.

Forty-eight waves at the optimum, fifty-four in the full prescription — and here is the inversion that makes the strict yantra the mirror image of its stylized cousin. The nested hexagrams differed by size, so they were spread across frequencies: one direction set, a chord of pitches. The strict yantra's edges differ by direction and position — quantities carried by bearing and phase — so every exposure can use the same single frequency. Not a chord at all: one note, struck again and again from different compass bearings around the rim, the sand photographed between strikes and the photographs stacked.

§ 7 · Residues

What the sand finally says

Even the honest stack keeps residues worth staring at. First, every edge prints as its full chord — rule one is not negotiable — so what vibration renders is a wave-completed yantra: the classical figure plus the continuation of each of its twenty-seven edges, the threads of the loom left visible. Second, a nodal line has no width but a sand line does: its thickness is set by how quiet counts as quiet. And third — flip Plate IV to its all-at-once mode — the same fifty-four waves sounded together surrender most of the figure, because each pair’s silence fills with the others’ motion. The sequential strikes were never a convenience. They are what superposition demands.

So the legend receives a precise verdict. No syllable — no sound of any kind — sung over a uniform plate or membrane deposits sand into the strict Śrī Yantra; the incommensurate heights, the nineteen unrepeated directions, and the interior corners each forbid it on their own. What the tonoscope can produce, and what the cymatics films do show, is the equilateral lattice of § 3 — geometry's echo of the yantra, not the yantra. Engineered instruments can do better: plates cut to the figure's outline, thickness graded to bend nodal lines, or arrays of dozens of speakers doing acoustic holography can approach the strict figure to within a wavelength. But each of these is a way of building the answer into the instrument — which is, perhaps, the point.

Because the tradition never claimed the yantra was a recording. In the Śrīvidyā literature the figure is a vidhi — a prescription, a thing to be constructed, stroke by stroke, in copper or in the mind's eye, with the marma junctions closing exactly. The physics returns the compliment with unexpected grace: asked what it would take to sound the figure into being, the answer is itself a prescription of ritual clarity — one tone, struck twenty-seven times, from twenty-seven bearings, each strike placing one line by phase alone. Not a chord but a litany. The legend, read carefully, was never about what sand does on its own. It was about how much order you must already possess before the sand will agree with you.

Appendix · The optimal family

When the bindu must sit at the centre

Section 4 left the classical figure with two blemishes: a bindu riding 0.042 R above centre, and a central triangle of 63.0° rather than a clean 60°. The modern literature's "optimal" yantra demands both fixed — a centred bindu and an equilateral central trikoṇa. Those are two equations, and the figure has four free choices, so a two-parameter family of optimal yantras survives. Plate VI solves, live, the one nearest to Huet's classical figure — and then goes further, spending the remaining freedom on purpose: pinning t₈, and finally t₂, to 60° as well, ending in a figure with no freedom left at all. Gold marks the exactly equilateral triangles.

Two findings, one on each side of the ledger. The direction count can be pushed from nineteen down to fifteen — but only deliberately; the two optimality conditions by themselves move nothing, and even push t₈ away from the 60° it grazed by accident. The wave bill refuses to move at all: twenty-four exposures, forty-eight waves, across the whole family. Merged directions do create parallel legs, but they huddle within 0.37 R of the axis, and by rule two a shared tone would print ghost rungs right on the plate. Parallelism only pays across the plate's full breadth, and the only parallels that far apart are the three base pairs already counted.

The last variant carries a small theorem of its own. With every degree of freedom spent, t₂ and t₈ become exact equilateral triangles pointing opposite ways — a Star of David in the making — yet their sides differ by 6.7%, and nothing is left to close the gap. The hexagram of the cymatics films is not just a different figure from the Śrī Yantra; it lies provably outside the yantra's constraint system. Sound's favourite figure and the tradition's figure cannot be deformed into one another without breaking a marma.