Crystal Originated Particle

A crystal-originated particle, or COP, is not contamination at all — it is a tiny octahedral void baked into the silicon itself as the ingot cools, formed when point defects called vacancies clump together before the crystal has fully solidified, and it only shows up looking like a surface particle once a wafer finishing step exposes it.

Every growing crystal carries two kinds of point defects, and their balance decides whether COPs form. As silicon crystallizes out of the melt, the lattice incorporates both vacancies — missing atoms — and self-interstitials — extra atoms squeezed between lattice sites — in concentrations that depend on how fast the crystal is pulled and how it cools afterward. Crystal growers like Shin-Etsu, SUMCO, and Siltronic tune the pull rate specifically to steer the ingot toward whichever defect regime they want, because a vacancy-rich crystal and an interstitial-rich crystal age into completely different defect populations.

COP: A Void Born Inside the Crystal vacancies agglomerate as the ingot cools, then surface a wafer finish exposes Lattice while hot missing atoms = vacancies free vacancies diffuse while silicon is still soft Agglomeration on cooling octahedral void tens to a few hundred nanometers across clustered vacancies form one empty-lattice cavity After wafer finish pit or particle signal on scan looks like surface dirt — isn't a gate dielectric straddling one COP sees local thinning and leaks early the defect is grown in before the wafer is even cut — polish and anneal manage it, they don't remove its cause

The void only becomes visible once a wafer-finishing step reveals it. As-grown, a COP sits entirely buried in the bulk, invisible to any optical inspection. A particle-counting laser scan or an SC1/SC2 wet clean can expose the top of a void sitting right at the polished surface as a small etch pit, and a scanning tool reads that pit back as a surface particle — hence the name, even though nothing ever landed on the wafer from outside. KLA's unpatterned-wafer inspection tools are built specifically to tell a real surface particle apart from a COP signature by the characteristic pit shape it leaves.

The real risk is a device straddling one, not the particle count itself. A COP sitting directly under a gate oxide locally thins the dielectric at that spot, and a thinned gate oxide breaks down at a lower voltage and fails time-dependent dielectric breakdown screening far earlier than the surrounding, defect-free oxide — so the yield impact of COPs is almost entirely about gate-oxide integrity on logic and memory product, not cosmetic particle counts on a bare-wafer spec sheet.

Fabs manage COP density with the crystal, not with cleaning. Pulling the crystal slower, adding a post-growth high-temperature anneal in argon or hydrogen to dissolve near-surface voids, or growing epitaxial silicon on top of the COP-bearing substrate to bury the defect population below the active device layer are the standard levers — all three change where or whether COPs end up near the active silicon, none of them chase the particle after the fact. Epitaxial wafers in particular trade a thin, COP-free epi layer grown on top of a cheaper COP-bearing substrate for most logic and memory manufacturing today.

Read COPs through the same lens as Dash necking: a crystal-growth decision made once, in the ingot, that silently sets a ceiling on gate-oxide yield two process generations and dozens of steps later — the fab's defect budget is being spent at the crucible long before the first mask is ever loaded.

Take crystal originated particle further

Ask the copilot about this term, or have our engineers assess it against your process.

Ask CFSGPT about this → Request a technical assessment → Open the CFS app