Seed Dip and Pull Start
# Seed Dip & Pull Start: Setting the Czochralski Clock on an N-Type Melt
Every later event in the grown-junction process is scheduled relative to one moment that happens before any dopant switch, before any pellet drop, before the crystal is even a millimeter long: the instant the seed touches the melt and the puller starts winding it upward at a fixed rate. The melt itself is unremarkable at this point — plain n-type germanium, antimony-doped to N_D ≈ 10¹⁵ cm⁻³, held molten in a quartz crucible a few degrees above Ge's 938°C melting point. What makes this step irreversible is not the chemistry, it's the clock: once the seed crystal is dipped and pulling begins at constant rate v_p, every subsequent instruction in the recipe — when to drop the gallium pellet, when to drop the antimony pellet back in, how long to keep pulling before cutting the ingot — is defined as an elapsed pull time or pulled length *from this starting instant*, not as an independently adjustable event. Get the dip wrong — wet the seed too early, dwell too long and remelt it, or disturb the melt enough to nucleate stray crystallites — and there is no step downstream that can correct it; the whole ingot is scrapped before the first pellet ever gets dropped.
## 1. Why the Seed Interface, Not the Bulk Melt, Sets the Starting Condition
The melt is uniformly doped at N_D ≈ 10¹⁵ cm⁻³ before the seed ever goes in, but the solid that first freezes onto the seed does not inherit that concentration exactly. Antimony partitions between solid and liquid germanium with an equilibrium segregation coefficient k₀ < 1 — the growing solid always incorporates dopant less readily than the liquid it grew from, so the first material solidified at the seed tip is *slightly* less heavily doped than the melt it came from, with the rejected excess dopant piling up in the liquid just ahead of the interface:
As pulling continues and a larger fraction of the melt solidifies, that rejected dopant accumulates in the shrinking liquid, and the Scheil equation describes how the solid's doping rises through the length of the pull:
where $C_0$ is the melt's starting concentration and $X$ is the fraction of the melt already solidified. For germanium, $k_0^{\text{Sb}} \approx 0.003$ — antimony segregates strongly into the liquid — so the dopant concentration actually incorporated near the seed is lower than $C_0$, and it keeps *rising* slowly as the pull proceeds, well before anyone drops a gallium pellet on purpose. This is the baseline drift the whole process has to pull against: the "n-type" region of the finished sandwich is not uniformly doped by accident of the melt — it is a mild, predictable gradient set by $k_0$, and later pellet drops are timed and sized to overwhelm it decisively rather than compete with it.
## 2. The Dip Itself: Thermal Shock Is the Failure Mode, Not Contamination
A clean dip has to happen inside a narrow thermal window: the melt is held a few degrees above the 938°C germanium melting point (just enough superheat to keep it liquid without excess thermal gradient), and the seed is lowered until its tip just wets into the surface — not plunged deep enough to locally remelt its own crystal structure, not held back so gently that the melt freezes onto it with the wrong crystallographic orientation and nucleates unwanted grains. There is no chemical contamination risk being managed here; the single failure mode is thermal and mechanical — a dip that either erases the seed's single-crystal template or fails to wet cleanly enough to initiate ordered growth. Either failure is invisible until the finished ingot is pulled, by which point the entire pull — gallium drop, antimony drop, and all — has already happened on top of a flawed foundation.
## 3. Why This Step Is the Irreversible One in the Whole Six-Step Flow
Every later step in this process — the two pellet drops, ingot completion, slicing, lead attachment — operates on a crystal that already exists; a mistake there can sometimes be caught and the specific device discarded without destroying the whole ingot. A flawed seed dip corrupts the one thing nothing downstream can recreate: a single, continuous, correctly-oriented crystal lattice growing from a known zero-point in time. This is the same economic shape as every upstream step in a committed manufacturing sequence — the earliest irreversible decision carries the largest share of the total value at risk, because every later operation is a wager that the earlier ones already succeeded. The grown-junction process put that entire wager at the very first instant: one dip, one pull-rate setting, and a clock that starts ticking before the crystal has grown long enough to tell whether it worked.
## Seed Dip & Pull Start's Place in the Process Lineage
Seed dip and pull start is step one of the six-step grown-junction transistor device process flow — melt preparation and seeding, ahead of the first (gallium) and second (antimony) pellet drops that carve the n-p-n sandwich into the ingot as it grows, ahead of ingot completion, crosswise slicing, and final lead attachment. It is the step every later timing in the pull is measured against, and the only step with no corrective path if it fails.