Epitaxial Process 1961 Autodoping Out Diffusion Substrate

# Confront Autodoping: Out-Diffusion From the Substrate Into the Epitaxial Layer During Growth

## 1. Why the Layer Step 2 Grew Was Never Quite as Clean as It Looked

The reactor that grows the epitaxial layer must hold the heavily-doped substrate at the same high temperature the growth itself requires, and at that temperature the substrate's own dopant does not stay put: some of it diffuses straight upward into the newly forming layer by ordinary solid-state diffusion, and some of it evaporates off the substrate's exposed surface into the surrounding gas stream and redeposits elsewhere on the growing crystal, a combined effect this project now has to name and budget for as autodoping. Step 3's thickness-and-resistivity check, averaged across the whole layer, can pass a wafer whose doping is contaminated in exactly this way, because the contamination is concentrated near the substrate interface and fades with distance from it — a graded transition hiding inside a measurement built to catch a uniform one. The epitaxial layer Step 2 grew was never quite as light, nor as uniform with depth, as the dopant gas flow rate alone would have predicted.

$$N(z) = N_{\text{intended}} + N_s\,\text{erfc}\!\left(\frac{z}{2\sqrt{Dt}}\right) + N_{\text{vapor}}(z)$$

where $N_{\text{intended}}$ is the doping the gas-phase dopant flow alone would produce, the erfc term is substrate dopant diffusing directly upward by the same solid-state mathematics this project has used since 1954, and $N_{\text{vapor}}(z)$ an additional term from substrate dopant that evaporated and redeposited through the gas phase — the erfc term this project already knows how to calculate; the vapor-phase term is why autodoping remains only partly predictable, and why real reactors are characterized empirically rather than from calculation alone.

Contamination That Fades With Distance From Its Source the inverse of a 1954-style diffusion profile, not a new shape of mathematics DOPING LEVEL VERSUS DEPTH INTO THE EPITAXIAL LAYER depth into epitaxial layer, z → doping, N(z) Nintended — flat, as Step 2 intended actual N(z), highest at the interface z = 0, substrate interface bulk of epitaxial layer N(z) = Nintended + Ns·erfc(z / 2√Dt) + Nvapor(z) two transport paths superposed on the intended flat profile

## 2. Real Diagram: Two Paths Carry the Substrate's Dopant Upward

The reactor cross-section below makes both contamination paths explicit rather than folding them into a single arrow: a short, direct arrow for solid-state out-diffusion straight up from the substrate into the layer above it, and a separate, longer arrow tracing substrate dopant that evaporates into the gas stream and redeposits elsewhere on the growing crystal surface.

Two Transport Paths, One Contaminated Layer solid-state out-diffusion is short and direct; vapor-phase redeposition is longer and less predictable gas stream above the wafer evaporate, redeposit elsewhere on the growing surface epitaxial layer, growing upward direct solid-state out-diffusion heavily-doped substrate — the dopant source for both paths the short arrow is calculable from 1954's own diffusion mathematics; the long arrow is characterized empirically

## 3. A Sibling to a Contamination Problem This Project Already Named

The 1957 series' own Step 14 discovered that a later thermal step — oxide regrowth, performed after a diffusion was already complete — continues to deepen that earlier junction simply because the wafer is hot again, a problem this project identified, named, and learned to budget for well before this series began. Autodoping is a sibling to that problem rather than a new kind of difficulty: contamination moving in an unwanted direction during a thermal process, governed by the same families of diffusion mathematics this project has relied on since its very first series. What differs is the source and the direction. 1957's problem moved an existing junction deeper because of a step that came after it; this series' problem moves a new layer's doping away from its intended value because of the growth step that creates the layer in the first place, with the contaminating dopant arriving from below by diffusion and, less predictably, from above by way of the gas phase.

Step 5 does not introduce a new category of process risk to this project; it finds this project's oldest kind of risk — unwanted dopant movement during a hot process step — arriving by a path, through the vapor phase, that no earlier series ever needed to characterize.

Take epitaxial process 1961 autodoping out diffusion substrate 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