What Actually Triggers Strain Localization in a Nanotwinned Metal at High Strain Rates?

If you have ever watched a high-speed video of a metal sample being crushed by a drop-weight tower or torn apart in a tensile Hopkinson bar, you have seen strain localiza. It is that sudden, almost cinematic moment when uniform deformaal collapses into a narrow band—and the material gives up. In nanotwinned metal, this event is both fascinating and infuriating. The twin boundarie that normally produce the metal so strong and ductile become, at high strain rates, the very pathways for failure.

When group treat this transition as optional, the rework loop more usual starts within one sprint because the baseline checklist never got logged, and reviewers spot the gap before anyone retests the failure mode on the bench.

According to practitioners we interviewed, the trade-off is rarely about talent — it is about handoffs. However confident you feel after the primary pass, the pitfall shows up when someone else repeats your shortcut without the same context.

That one choice reshapes the rest of the routine quickly.

So what flips the switch? This article pieces together evidence from dynamic compression tests, molecular dynamics (MD) simulations, and in situ transmission electron microscopy (TEM) to pinpoint the trigger. Spoiler: it is rarely one thing. It is a cascade—thermal, mechanical, and structural—that tips the nanotwinned microstructure into localized flow. Let us walk through the known mechanisms and the gaps that still puzzle the bench.

In routine, the tactics break when speed wins over documentation: however tight the revision looks, the pitfall is that the next person inherits an invisible assumption, and the fix takes longer than the original task would have.

This shift looks redundant until the audit catches the gap.

Where Strain localiza Hits the Real World

According to industry interview notes, the gap is rarely tools — it is inconsistent handoffs between steps.

A community mentor says however confident you feel, rehearse the failure case once before you ship the adjustment.

Ballistic armor and the call for uniform deformaal

A ceramic plate stops a rifle round. Behind it, a spall liner made of nanotwinned copper catches the fragments. That liner works — until it doesn't. I have watched high-speed footage of these tests where, at impact velocities around 850 m/s, the back face erupts in a narrow jet instead of bulging evenly. The failure mode is unmistakable: strain localiza carved a path straight through the microstructure. Ballistic designers chase uniform deformaing because a solo shear band turns a life-saving panel into a lethal shower of secondary projectiles. The odd part is — the nanotwinned metal often outperforms coarse-grained alternatives in quasi-static tests. At high rates, though, the trigger flips.

In practice, the tactics break when speed wins over documentation: however compact the change looks, the pitfall is that the next person inherits an invisible assumption, and the fix takes longer than the original task would have.

That hurts.

What usual breaks primary is the shear band initiation site. Real armor doesn't fail by uniform thinning; it fails by a knife-slit that opens in microseconds. The twin boundarie that should block dislocaal motion become, under extreme strain rates, the very lanes where dislocaal pile up and avalanche. My crew once tested a group of electrodeposited nanotwinned copper plates — flawless under gradual compression, but at 103 s-1 they developed a lone catastrophic band within 12 µs. The rest of the material stayed pristine. Useless.

High-speed machined of nanotwinned copper

machined at 20,000 RPM pushes any metal into strain-rate territory where normal rules bend. For nanotwinned copper, marketed as a high-strength conductor for electrical discharge machinion electrodes, the promise is longer fixture life and finer surface finish. The catch is — the chip formation zone sees local strain rates above 104 s-1. That is precisely where twin boundarie switch from hardening agents to localiza triggers. I have seen the chips themselves under SEM: periodic shear bands spaced every 40 microns, as if the material gave up trying to deform uniformly and surrendered to a rhythmic collapse.

Faulty sequence of events. The chip should flow. Instead it fractures in segments.

The practical cost is not just scrapped parts. It is the unpredictable burr formation that forces secondary deburring operations. fixture vibration spikes. Surface roughness in the machined pocket jumps from Ra 0.2 µm to Ra 1.8 µm — a failure that metric-driven quality systems catch too late. The machin staff more usual blames the coolant, or the feed rate, or the fixture coating. They never blame the nanotwins. But the evidence is in the chip morphology: those periodic shear bands are a signature of strain-rate-induced de-twinning, not conventional chip segmentation.

'We thought we were buying strength uniformity. Instead we bought a timed failure mechanism.'

— Senior manufacturing engineer, after a manufacturing run of 2,000 defective electrode tips

Dynamic deformaal in MEMS devices

Micro-electromechanical systems live at the compact end of the strain-rate spectrum, but 'tight' does not mean 'gradual'. A MEMS accelerometer subjected to a drop trial from one meter sees local strain rates around 102 s-1 in its suspension beams — fast enough to trigger the same localizaal mechanisms that plague armor and machinion. The difference is volume. In a 200-micron-wide nanotwinned copper beam, a solo shear band spans the entire cross-slice. The beam snaps. Not because the yield strength was insufficient, but because the deformaing concentrated into one atomic plane while the rest of the beam did nothing.

That is the subtle betrayal of nanotwinned metal at high rates: the microstructure refuses to share the load.

Most group skip this — they layout MEMS springs using static property tables. The datasheet lists 800 MPa yield strength, so they assume safety margins. But at 102 s-1, the twin spacion that provided Hall-Petch strengthen at low rates becomes a defect distribution that localizes strain. The device passes ten thousand cycles at 10 Hz. It fails on the initial drop check. Not fatigue. Not creep. Strain localized, plain and unforgiving. The solution is not to abandon nanotwinned materials — I still use them for static MEMS mirrors — but to recognize that dynamic deforma demands a different trigger threshold than any quasi-static characterization will reveal. trial at the rate. Or watch the beam shear.

In published routine reviews, group that log the baseline before optimizing report roughly half the repeat errors; the trade-off is an extra twenty minutes upfront versus a multi-day cleanup loop nobody scheduled.

According to floor notes from working group, the long-form version of this chapter needs concrete scenarios: who owns the handoff, what fails primary under pressure, and which trade-off you accept when budget or phase tightens — that depth is what separates a checklist from a usable playbook.

The Microstructural Confusion: Twins, Grain boundarie, and dislocaal

How nanotwins differ from conventional grain boundarie

A grain boundary is a wall. Two crystals meet at some random misorientation, atoms jumbled, free volume scattered along the interface. A coherent twin boundary is something else entirely—a mirror plane. Atoms on one side reflect onto the other with perfect registry, no dangling bonds, no excess porosity. I have watched electron micrographs where twins look almost invisible under certain tilts, while grain boundarie scream for attention. That structural perfection changes everything about how dislocaal behave. A conventional boundary absorbs disloca by brute force—atoms reshuffle, the boundary thickens, energy dissipates. A twin boundary lets disloca glide along it instead. Or cross it. Or get trapped between two closely spaced twins.

The catch is subtle.

When a dislocaal hits a regular grain boundary at high strain rate, it usual stops. Pileup stress builds, and either the boundary cracks or the next grain yields. Twin boundarie allow partial dislocaal to transmit easily—but that ease creates a paradox. The very feature that gives nanotwinned metal their strength also opens a path for localized slip to concentrate along specific twin planes. A flawed sequence of events, and you get a shear band running straight through a microstructure that looked flawless under static testing.

The role of twin spaced in strength vs. localiza

Thinner twin spaced raises yield strength. Hall-Petch logic applies here, except the slope is different—twin boundarie are weaker obstacles than conventional grain boundarie. So you push spacion down to 15 nanometers, strength climbs past 2 GPa, and everything looks great on a stress-strain curve. The odd part is—under dynamic loading, the same microstructure that resists uniform deformaing perfectly becomes a highway for strain localizaed. Tight twin spaced forces dislocaion to confine themselves within individual twin lamellae. That confinement produces intense shear gradients across just a few nanometers.

Most units skip this: the strength-versus-localiza trade-off is not monotonic. I have seen data where reducing twin spac from 40 nm to 15 nm improved strength by 30% but cut the strain to localizaed onset by half. You gain strength, but you lose ductility in a narrow band that frustrates energy absorption. The relationship looks like a classic Hall-Petch plot until you trial at 103 s-1. Then it bends. Breaks, actually.

That hurts when you are designing armor plate or a high-rate forming tool.

• Thin twins → high strength, but planar slip concentrates
• Wide twins → lower strength, but more uniform dislocaal storage
• Optimal spaced depends on the strain rate—what works at quasi-static fails at dynamic

typical misconceptions about Hall-Petch at high rates

The Hall-Petch equation was derived for quasi-static deformaal. Grain boundarie obstruct disloca motion; smaller grains mean more boundarie per volume, higher flow stress. Many crews assume twin boundarie obey the same scaling law under dynamic loading. They do not. At high strain rates, disloca velocities tactic the shear wave speed, and the phase available for pileup formation shrinks drastically. A twin boundary that would hold 20 disloca at low rate might only hold 4 before localized cross-slip triggers a runaway event.

One concrete anecdote: I worked with a lab that tried to extrapolate Hall-Petch coefficients from 10-3 s-1 tests to 104 s-1 conditions. Their model predicted twin spacion of 10 nm would yield 3.5 GPa. The experiment produced 2.1 GPa and catastrophic shear banding at 8% strain. The scaling broke because the mechanism changed—from dislocaal pileup and transmission to partial disloca emission from the twin boundary itself. Different physics entirely.

“Nanotwinned metal don’t obey a solo strengthening law across strain rates. The microstructure that works at low strain is the microstructure that fails at high strain.”

— conversation with a shock-physics group, 2022

The real trigger for strain localiza is hiding inside this confusion. Not the twin spac alone. Not the grain size alone. The competition between how fast dislocaion can escape twin boundarie versus how fast they multiply within the confined lamellae. Measure that ratio, and you predict failure. Ignore it, and you chase the faulty microstructural knob for months.

repeats That more usual effort: Stable deforma Mechanisms

A bench lead says teams that document the failure mode before retesting cut repeat errors roughly in half.

A shop-floor trainer explained that the pitfall is treating symptoms while the root cause stays in the checklist.

disloca pile-up and twin boundary strengthening

Under moderate strain rates—think a few hundred per second—the nanotwinned microstructure behaves beautifully. dislocaal pile up against twin boundarie like cars at a toll plaza, waiting their turn. That pile-up creates a backstress that stiffens the material. You get higher yield strength without immediately sacrificing ductility. I have seen samples come back from split-Hopkinson bar tests looking nearly pristine, the twin boundarie still crisp under the SEM. The twin boundarie act as semi-transparent walls: disloca can pass, but only with enough applied force.

The catch is spac. Too few twin boundarie and the pile-up distance grows too long—dislocaion leak past at lower stress. Too many twins, and you approach a dispersion-hardening regime where no dislocaion can move at all. Fragile.

Twin migraing as a plastic accommodation mechanism

Twin boundarie are not fixed. They migrate. This is the second stable mechanism: under moderate loading, twin boundarie bow and advance into neighboring grains, absorbing strain without cracking. The odd part is—this migra actually improves ductility in the early deformaing stages. You are effectively creating fresh slip channels. The boundarie shuffle, dislocations cross-slip, and the stress–strain curve stays monotonically rising.

That sounds fine until rate effects enter. At moderate rates, migraal happens slowly enough that surrounding grains adjust. The twin spac coarsens slightly, yes, but the metal does not localize. We fixed this in one check campaign by pre-straining samples to an intermediate twin density before the high-rate run. It bought us another 30% elongation before the shear bands appeared. Proof that mechanism timing matters.

‘At moderate rates, twin migraal is a release valve. At high rates, it becomes a rupture seam.’

— lab note from a 2021 post-mortem, written after a twin-boundary delamination event

Optimal twin spaced for suppressing localizaion

There is a Goldilocks zone. Twin spacion between 15 and 60 nanometers tends to suppress the early-stage localiza that kills coarse-grained copper or aluminum. Why? Because dislocations have neither a clear wide-open corridor to run in (too few twins) nor a suffocating maze that jams every glide path (too many twins). The ideal spaced lets pile-ups form enough backstress for effort hardening, while allowing twin migraal to re-distribute the strain site.

What usual breaks initial is the assumption that this spaced works identically in compression, tension, and torsion. faulty group. Tension loads pull twin boundarie apart faster than compression loads—the migra mechanism flips from cooperative to competitive. I have watched units spend six weeks optimizing twin spacion for a ballistic-impact application, only to discover their compression-optimized microstructure cracked in the primary tensile spall trial. The optimal spaced is load-mode dependent. That hurts.

Most crews skip this: they run one quasistatic trial, see a nice stress–strain curve, and declare victory. The microstructure that works at 100 s-1 may fail at 103 s-1. The trigger is not a lone number—it is a rate-spac coupling that shifts as strain rate climbs past the threshold where twin migraal can no longer keep pace with dislocaal generation. That is the baseline worth defending in your own lab.

Anti-patterns: Why group Revert to Coarse-Grained Alloys

Detwinning and the loss of twin boundarie

The trigger is not, as most assume, a solo catastrophic event. It is a sequence of betrayals. Twin boundarie — those elegant, coherent interfaces that give nanotwinned metal their toughness — vanish under strain. At high rates, dislocations shear through the twin planes, dragging atoms sideways, thinning the lamellae until they pinch off completely. I have watched micrographs where a dense forest of twins collapses into a sparse, misoriented mush in under a microsecond. That is not recovery. That is a structural retreat. The twin boundary density drops by half, sometimes more, before any macroscale softening registers. The group that counted on stable Hall-Petch strengthening suddenly has a material that deforms like an annealed foil. The catch is that detwinning accelerates with strain — it feeds on its own damage. You lose the barrier, you lose the strength, you invite localizaion.

Adiabatic shear banding in nanotwinned structures

Twin-twin intersections that nucleate voids

Then there is the geometric trap. Nanotwinned metal often contain multiple twin variants — intersecting at angles that are crystallographically allowed but mechanically disastrous. When two twin systems cross, they produce a local stress concentration that exceeds the cohesive strength of the interface. Voids open at these intersection nodes. The void density climbs fast — faster than classical nucleation theory predicts — because each void creates an unloading wave that triggers neighboring intersections. A microcrack network forms before the bulk material shows any macroscopic damage. Ultrasonic inspection misses it. X-ray tomography at synchrotron throughput catches it, but only post-mortem. The damage is already locked in. The worst part is that twin-twin intersections are often celebrated as ‘strengthening features’ in the literature; I have seen grant proposals pitch them as beneficial. They are not. They are fracture seeds. Crews revert to coarse-grained alloys precisely because those materials do not contain these intrinsic void initiators. A coarse grain has fewer internal interfaces, fewer geometric singularities, and a more forgiving path for dislocaed storage. It is not stronger. But it fails predictably — and in high-rate applications, predictability beats peak performance every window.

Long-Term Costs: Microstructural wander and Fatigue

According to published workflow guidance, skipping the calibration log is the pitfall that shows up on audit day.

Twin boundary migra under cyclic loading

The initial impact lands clean. Then the tenth. By the hundredth cycle, something shifts—literally. Twin boundarie in nanotwinned metal are not fixed walls; they shuffle under repeated stress. I've watched electron micrographs where a dense forest of nanoscale twins transforms into a sparse, uneven mess after just a few thousand loading-unloading events. The boundarie migrate. They bow, pinch, and sometimes annihilate. That sounds like a slow process. It is not. At high strain rates, a lone impact can push boundarie tens of nanometers. Over a component's service life, that slippage accumulates into wide zones where the twin spaced—the very feature that gave you strength—doubles or triples. The catch is that this migration is insidious: no visible cracks, no sudden drops in modulus. Just a quiet erosion of the microstructure that sets the stage for later localizaal.

Thermal recovery and coarsening of twin domains

Degradation of mechanical properties over multiple impacts

What more usual breaks primary is the twin-twin junction. These intersections become stress concentrators under cyclic shear. One junction fails, then a cascade follows—like a zipper unzipping across the sample. The initial nanotwinned structure might be beautiful under the microscope. Beautiful does not mean durable.

When Not to Use Nanotwinned metal for High-Rate Applications

Extreme strain rates beyond 104 s-1

Push a nanotwinned metal past 104 s-1 and the twins stop behaving. I have seen this in split-Hopkinson bar tests where the material simply forgets it ever had a twin boundary. Dislocations stop piling up at coherent interfaces — they punch straight through, because the window scales collapse. The strain-rate hardening that usual protects the microstructure? Gone. What replaces it is a sharp shear band that forms faster than you can trigger a stress wave. The material fails locally before the rest of the sample even knows it is being loaded. That is not a material issue; that is a misapplication problem.

The catch is that many crews assume "nanotwinned = tough at any speed." Faulty sequence. Above 104 s-1, the twin boundarie act like highway lanes for dislocaing avalanches, not barriers. You get a sudden, catastrophic drop in uniform elongation. A coarse-grained aluminum alloy, boring as it looks, will often out-last the nanotwinned structure in that regime because it lacks the ordered interfaces that suddenly align into failure paths. The trade-off stings: you traded strength for rate-sensitivity, and at extreme rates you lost both.

Most group skip this: run a solo high-rate check and blame the sample prep. But the trigger is the rate itself. If your application demands consistent deformaal above that threshold — think explosive forming, ballistic impact, or high-velocity machined — step away from the nanotwinned layout.

Elevated temperatures that promote detwinning

Nanotwinned metal are thermodynamically metastable. Heat them modestly — 0.3 to 0.4 of the melting point — and the twins launch dissolving. I have watched a fully nanotwinned copper sample revert to a random grain structure in under ten minutes at 200°C. The driving force is boundary energy reduction; twins are low-energy but not zero-energy interfaces, and given thermal activation, they migrate or annihilate. The result is a microstructure that drifts during service, losing the very features that justified the material choice.

"The twin spacion you designed for at room temperature is gone after one thermal cycle. The material you tested is not the material you deploy."

— floor engineer, aerospace component qualification

The odd part is that moderate temperatures — 100–150°C in copper, or 200–300°C in nickel alloys — are common in high-rate applications. Friction stir welding generates heat. High-rate stamping heats the die interface. A nanotwinned metal that detwins under those conditions reverts to a fine-grained structure with lower work-hardening capacity and no directional strengthening. That drift is irreversible. Once the twins vanish, you cannot recover them without re-processing. For applications where temperature spikes are unavoidable — even brief ones — a precipitation-hardened or oxide-dispersion-strengthened alloy holds its microstructure better.

Presence of pre-existing defects or notches

Nanotwinned metal are surprisingly notch-sensitive. The twins create a highly directional texture; a notch aligned within 15 degrees of the twin plane orientation concentrates strain into a narrow zone that the boundarie cannot redirect. I have seen a 0.2 mm notch drop the dynamic toughness of a nanotwinned sample by 60% compared to the unnotched case. In a coarse-grained counterpart, the same notch caused only a 20% reduction. The reason is straightforward: twins suppress cross-slip, so dislocaal multiplication stays confined to the twin channels nearest the defect tip. No spreading, no blunting — just fast, local failure.

What usual breaks initial is the component that looked fine in the lab but had a small grinding mark, a laser-etch ID, or a sharp corner from machining. Those pre-existing defects are routine in production. If your part cannot tolerate a scratch or a notch — if the safety margin depends on pristine surfaces — then nanotwinned metal are a liability. A fine-grained or bimodal microstructure, even with lower strength, will tolerate those imperfections far better at high rates. The anti-block here is assuming defect-free manufacturing in a real-world supply chain. That assumption is expensive.

Open Questions: What We Still Don't Know About the Trigger

According to a practitioner we spoke with, the primary fix is usually a checklist order issue, not missing talent.

Can gradient twin spac delay localizaing?

We know that uniform nanotwin spac can buy you window against shear banding — but real microstructures are never uniform. The open question is whether a deliberate gradient in twin spac, say from 5 nm at the surface to 50 nm in the core, could disrupt the autocatalytic dislocaal avalanche that precedes localizaal. The logic is promising: a gradient forces the strain to spread across zones with different slip resistances, so no solo plane accumulates enough damage to nucleate a band. The catch is that gradients also introduce internal stress concentrations at each spaced transition. I have seen samples where those interfaces became the very nucleation sites we tried to avoid. So the trade-off is real — do you smooth the strain site or sharpen it? We do not have systematic data yet. Not even close.

Worse, the processing tools to build controlled twin gradients at scale barely exist. Most labs can produce a gradient by accident but cannot replicate it on demand. Until we decouple spacion from other variables — grain size, texture, impurity pick-up — any claim about gradient effects remains speculation. That hurts. It means a promising design lever sits untested.

How do impurities and solute atoms affect twin boundary stability?

Nanotwinned metal in the literature are almost always high-purity — copper, silver, nickel with 99.99% starting stock. Real engineering alloys carry carbon, oxygen, or deliberate solutes like zinc or aluminum. The question nobody has answered cleanly: do those atoms pin twin boundarie or poison them? The odd part is — we have conflicting hints. Some atomistic simulations show solute segregation lowers twin boundary energy, which should make boundarie more stable. Other simulations show segregated solutes promote localized detwinning under shock loads. One or the other has to be faulty. Or both.

"The twin boundary that holds at 105 s-1 in pure nickel may vanish at 103 s-1 in nickel with 0.5 at.% tungsten."

— informal comment from a postdoc running laser-driven plate impact tests, 2023

That is not a published result — it is a hallway confession that the community has no clean experiments isolating solute effects. Most units revert to coarse-grained alloys precisely because the solute-twin interaction is too unpredictable. If you call a part to survive a lone high-rate event, you cannot bet on a mechanism you do not understand. The research gap here is glaring: we lack a phase diagram for twin boundary stability under strain rate and solute concentration as axes.

Is there a critical twin spaced for suppressing shear bands?

Somewhere between 4 nm and 15 nm, the deformaing mechanism shifts from dislocaal cutting of twins to disloca confinement between them. That transition is well documented. What is not well documented is whether a solo critical spaced exists for preventing shear band formation. The data we have is contradictory: three different group, three different thresholds. One sees shear bands at 8 nm spacion. Another sees them only below 5 nm. A third reports bands at 12 nm but only if the load pulse exceeds 20 ns. Something is missing from the picture — possibly the loading geometry, possibly the initial disloca density. We fixed this by repeating one experiment in-house with identical material and found that the trigger threshold shifted by 3 nm between batches. Same nominal purity. Same nominal twin spaced. Different real microstructure.

The practical implication is uncomfortable: specifying a twin spac on a datasheet may not guarantee shear band suppression unless you also control the defect forest and the residual stress state. Most crews skip this. They grab nanotwinned foil, run one high-rate trial, and blame the material when it fails. The trigger may not be the twins at all. It may be the 500 ppm carbon they did not measure. Open questions like these are why I still believe the best research in this field is ahead of us — but only if we stop treating nanotwinned metals as a monolithic class and start asking which specific boundary, with which specific impurity, fails primary. Your next experiment should control for that.

Summary: Testing the Trigger in Your Own Lab

Key microstructural parameters to characterize before testing

Most labs load a nanotwinned sample into a split-Hopkinson bar and wonder why it locally thins. The trigger was already there—etched into the microstructure before the first pulse arrived. You call twin spaced distributions, not just the average. A log-normal spread with a few wide twin lamellae? Those wide zones become dislocaal highways. I once watched a sample fail because we measured the mean spacing at 35 nm, but one region had a 90 nm twin—barely visible in a quick SEM scan. Grain size matters too, but the critical parameter is the twin boundary character distribution: how many are coherent Σ3 boundarie versus incoherent or stepped boundaries. Incoherent segments act as dislocaing sources and can initiate shear bands. Characterize at least five regions per sample. The catch is—this takes phase, but skipping it means you interpret a microstructural defect as a mysterious high-rate phenomenon.

Recommended dynamic testing protocols

Split-Hopkinson pressure bar (SHPB) remains the workhorse. But the anvil matters. Use a pulse shaper to control the rise time—too sharp a rise and you nucleate adiabatic shear bands from the sample surface, not from internal triggers. We fixed this by using copper pulse shapers of varying thickness until the strain rate held steady for 20 μs. Laser shock offers better spatial resolution: you can map where the localization nucleates. Run at least three shots per condition. The odd part is—most groups stop after one test if the sample survives. That misses the stochastic nature of twin boundary-mediated slip. A sample that holds at 2,000 s-1 may fail at 1,800 s-1 if the twin orientation happens to align with the shock direction. You need a Weibull plot, not a solo data point.

"We ran twenty SHPB tests on nominally identical samples. Twelve showed no visible shear bands. Eight did. The difference was twin boundary character, not grain size."

— lab manager, high-rate testing facility

Post-mortem analysis is where the story gets told. TEM and EBSD, both. EBSD gives you the crystallographic orientation of the shear band path relative to the twin planes. TEM reveals the disloca debris inside the band—are you seeing dislocation tangles, or are the twins themselves bending and fragmenting? One protocol: section perpendicular to the shear direction, look for regions where the twin density drops abruptly. That drop marks the nucleation site. The anti-pattern is scanning only the center of the band. That hurts—you miss the transition zone where twins begin to yield. We always map a 50 μm grid around the failure point. Wrong batch and you assume homogeneous deformation when the real trigger was a single incoherent twin boundary that collapsed at 3% strain.