Oxidation Furnace
Oxidation furnace: grow an insulator by consuming silicon. An oxidation furnace holds silicon wafers in a clean, hot, controlled atmosphere so oxygen or water vapor can react with the surface to form silicon dioxide. Unlike depositing a film on top, thermal oxidation moves the silicon–oxide interface into the wafer. Temperature, ambient and time control the result, but cleanliness and uniformity decide whether that result is useful.
1. The batch must see the same process. In a horizontal furnace, a quartz boat holds wafers inside a heated quartz tube. Vertical furnaces arrange the batch differently; the requirement is the same: control temperature and oxidant delivery across the load. Heater zones compensate for heat loss near the ends. Gas flow alone does not guarantee equal growth on every wafer.
Dry and wet describe the oxidant, not the hardware. Dry oxidation uses O2: Si + O2 → SiO2. Wet oxidation uses water vapor: Si + 2H2O → SiO2 + 2H2. At comparable conditions, steam generally grows oxide faster, making it useful for thicker layers. Dry oxidation is often chosen for thin, electrically demanding oxides. Neither a universal wet/dry speed ratio nor “highest quality” describes every recipe: temperature, pressure, crystal orientation and subsequent treatment matter.
2. Oxide growth changes the surface position. For planar silicon oxidation, about 0.44 of the final oxide thickness corresponds to consumed silicon; about 0.56 extends above the original surface. A 100 nm thermal oxide therefore consumes approximately 44 nm of silicon. This is a geometry example, not a furnace recipe. It explains why oxidation matters to feature dimensions as well as insulation.
3. The growing film becomes a transport barrier. Oxidant must cross the existing oxide to reach the buried silicon interface. The classic Deal–Grove relation, x² + Ax = B(t + τ), captures a transition from approximately linear growth to approximately square-root-of-time growth. Here x is oxide thickness, A and B are fitted process constants, and τ accounts for starting oxide. A thicker film lengthens the diffusion path; twice the time does not generally give twice the thickness. Very thin oxides and nonplanar structures require more refined models.
The furnace setpoint is not the whole thermal budget. Loading, ramping, holding and cooling all contribute to the wafer's temperature history. Time at elevated temperature can redistribute dopants and change stress. A batch furnace spreads processing cost across many wafers, but its long thermal cycle may be incompatible with a device's later layers. Rapid thermal oxidation offers a different time–temperature profile; deposited oxides solve a different materials problem and do not reproduce the same grown interface automatically.
Qualify the oxide, not just the timer. Thickness maps check within-wafer variation; monitor wafers at different boat positions check batch variation. Ellipsometry measures optical thickness through a fitted film model, while electrical test structures assess leakage, breakdown and interface behavior. A correct average thickness can hide a bad wafer edge, a contaminated interface or a drifting tube position. Keep tube cleanliness, wafer history and measurement conditions traceable.
The practical takeaway: an oxidation furnace controls a moving chemical interface across an entire batch. A useful process must deliver the required thickness and electrical behavior without spending more of the device's thermal budget than it can tolerate.
Further reading: TU Wien, Thermal Oxidation and the Deal–Grove Model; NIST, Sandvik Oxidation Furnaces; Utah Nanofab, ProTemp Clean Ox.