RESOURCES · SOLUTIONS BLOG

We Found the Note — Then We Shattered the Glass

RS Rand Simulation · Applications Engineering AI · June 2026 · 7 min read

Every wine glass has a note. Flick the rim and you hear it — a clean, ringing tone that is the bowl flexing back and forth between an oval and a circle. In an earlier demo we went looking for that note with a modal analysis. This time we did the reckless thing: we played the note back at the glass, loud and exactly on pitch, and watched the rim tear itself apart in an explicit LS-DYNA fracture run. Spoiler — it shatters. And the control that proves why it shatters is the best part.

The marketing-grade hero render of the break: the same resonance-driven shatter, rendered with physically-based glass shading and a pinned camera at the moment the bowl flies apart.

The bowl driven at its (2,0) ovalling resonance, 709.5 Hz. The rim flexes into an oval and back, swinging harder and harder cycle by cycle until it shatters. Color = motion amplitude (how far each point has moved, in mm) — you watch the ovalling grow. The break itself is tension-driven: as the rim flexes wider, the tensile stress at the antinodes climbs, and when it crosses the strength of glass (~50 MPa) the elements there erode — a crack initiates right at the ovalling antinodes and the bowl flies apart.

The physics: resonance is a patient thief

The glass-shattering trick is not about brute force. It is about timing. A wine-glass bowl is a very lightly damped structure — once it starts ringing, it keeps ringing. Push it with a pressure that oscillates at its natural frequency, perfectly in step with the motion, and every push adds a little more amplitude than the last. The rim swings wider and wider. That run-away growth is resonance, and it is how a modest, sustained input becomes a catastrophic output.

The specific mode that matters here is the (2,0) ovalling mode: two nodal diameters, so the circular rim pulses into an oval and back, four times per cycle. The bending is most severe at the tips of the oval — the antinodes — and that is where the tensile stress concentrates. Glass is strong in compression but brittle in tension; it does not yield, it cracks. So the question is simply: can the antinode tension outrun the ~50 MPa tensile strength of glass before the energy bleeds away? On resonance, yes.

It isn't the loudness that breaks the glass. It's the loudness arriving in perfect time, seven hundred times a second, exactly on the note.

Inside the model

Geometry and mesh

We modeled the bowl alone — a thin shell of revolution reusing the outer profile from our earlier wine-glass modal demo, offset to the mid-surface. The result is a clean shell mesh of 3,384 nodes and 3,312 fully-integrated shell elements (ELFORM=16, 5 integration points through the thickness), wall thickness 1.5 mm, sweeping from the base ring at y = 82 mm up to a rim radius of 41.2 mm at y = 178 mm. The base ring is fully clamped; the rim is free to ring.

Material and failure

Glass is *MAT_ELASTIC with E = 70 GPa, density 2500 kg/m³, and ν = 0.22. Brittle fracture is handled with *MAT_ADD_EROSION set to a maximum-principal-stress threshold of SIGP1 = 50 MPa: when the tensile stress in an element crosses that line, the element erodes and a crack is seeded. It is a deliberately simple surrogate for glass failure — more on that in the scope box — but it captures exactly the thing we care about: tension-driven cracking, where the tension is highest.

**Where it cracks.** The deformed bowl colored by **maximum principal (tensile) stress** at the moment of failure. The animation up top is colored by motion amplitude — it shows the flexing; this shows the *tension that actually breaks the glass*, concentrated right at the tips of the oval (the (2,0) antinodes) where the rim bends hardest.

The load and the solver

This is the elegant bit. Rather than model a roaring loudspeaker, we applied a pressure on the bowl shaped like the mode itself: P(θ,t) = P₀·cos(2θ)·sin(2πf·t). The cos(2θ) spatial shape is the (2,0) ovalling pattern, so the forcing pours energy into only that one mode — a clean, surgical drive. We ran it explicitly in LS-DYNA (lsdyna_dp, R16). To find the drive frequency in the first place, Stage A was an implicit eigenvalue (modal) analysis that pulled the ovalling pair out at f* = 709.5 Hz. Stage B then drove the bowl at that exact frequency.

Two decisions made the run honest and fast. First, drive amplitude: an undamped resonance grows roughly linearly, so we calibrated P₀ to fracture in a handful of cycles. Our first guess (0.02 MPa) shattered the glass in ~4 ms — too violent to be convincing — so we backed it down to P₀ = 0.005 MPa. Second, speed: we applied selective mass scaling to the small clamped base elements only. The rim and ovalling region were left untouched, which preserves f* exactly while buying roughly a 3× speedup on the time step.

The result: driven at its (2,0) ovalling resonance of 709.5 Hz, the bowl's rim tensile stress crosses the 50 MPa strength of glass and the first elements erode at t = 5.5 ms (~4 cycles in) — the crack initiates exactly at the ovalling antinodes (rim, θ ≈ 90°/270°), peak principal stress 44–50 MPa, and the glass shatters. The identical drive applied off-resonance at 500 Hz produces zero eroded energy and the glass stays intact.

Is it right?

A single shattering animation looks great but proves little on its own — maybe we just pushed hard enough to break anything. So we ran the experiment that scientists run: a control. Same glass, same mesh, same pressure amplitude, only the frequency changed — from the on-resonance 709.5 Hz down to an off-resonance 500 Hz.

**On-resonance (709.5 Hz)**: rim stress climbs past 50 MPa and eroded energy jumps at ~5.5 ms — fracture. **Off-resonance (500 Hz)**, with the *identical* pressure amplitude: the stress never reaches the threshold, zero elements erode, the glass survives. The only variable that changed is the frequency.

The on/off comparison is decisive on three counts. The location is correct: the crack starts at the (2,0) antinodes, where bending theory says tension is greatest — not somewhere random. The mechanism is correct: on resonance the stress accumulates and crosses strength; off resonance the same force merely sloshes the rim and the stress stays far below fracture (peaks ~18 MPa vs. 44+ MPa). And the frequency is grounded — it came straight out of the implicit modal analysis, not a number we picked. Together they isolate the cause: it is the resonance that breaks the glass, not the push.

The real-world connection

This is one of the most beloved "is it real?" myths in pop culture, and the answer is a satisfying yes. The legend goes back to the 1971 Memorex cassette commercials — Ella Fitzgerald hits a note, a wine glass shatters, and the tagline asks "Is it live, or is it Memorex?" The whole pitch was that the recording of her voice could break the glass just like the live note — resonance doesn't care whether the source is a larynx or a tape head, only whether the frequency is right.

Decades later, MythBusters put it to a real test. Adam Savage first broke a glass with an amplified voice, and then rock vocalist Jaime Vendera shattered a wine glass with his unaided voice — no microphone, no speakers — confirming the full myth. The key, exactly as our simulation shows, was finding the glass's own resonant note and holding it. Watch the iconic moment:

If you want the single unaided-voice break that sealed the episode, here is Jaime Vendera's glass-shattering moment. A human can do in one held note what we just did with a cos(2θ) pressure field and an erosion threshold: deliver energy on the glass's own note until the antinodes give way.

Honest scope. The bowl is modeled as a shell of revolution. Brittle fracture uses *MAT_ELASTIC + *MAT_ADD_EROSION at a 50 MPa max-principal-stress threshold — a simple surrogate for glass failure, not a fracture-mechanics flaw or crack-growth model, so it tells you where and when cracking initiates rather than tracing a physically exact crack path. Selective mass scaling was applied only to the clamped base elements (the rim/ovalling region is untouched, so f* is preserved) to keep the explicit run tractable. The drive pressure amplitude was tuned to fracture in a handful of cycles, not derived from a real sound pressure level — so this demonstrates the mechanism (resonance at the ovalling mode concentrates tension at the antinodes and breaks the glass), proven by the on/off-resonance control, rather than predicting the exact decibel level a singer would need.

Want resonance answers for hardware that can't shatter on stage? The same modal-then-transient workflow — find the mode, drive it, watch the stress concentrate — is the kind of workflow that helps teams keep brackets, enclosures, and rotating equipment off their resonances long before a prototype ever rings. That's innovation through insight.