Photolithography Oxide Masking 1957 Strip Regrow Oxide
# Strip and Regrow the Oxide: Resetting the Mask Without Resetting the Wafer
## 1. Why a Furnace Cycle Keeps Charging Interest on a Debt the Wafer Already Owes
This step strips the oxide that masked Step 13's diffusion — because its windows are now in the wrong places for the next masking level — and grows a fresh oxide over the entire wafer, including directly over the region just diffused, so the wafer can begin a second complete photolithography cycle from a clean, continuous barrier. The complication this step cannot avoid is that growing that fresh oxide is itself a high-temperature operation, and high temperature is precisely what drives the dopant Step 13 just placed. Every subsequent thermal step the wafer experiences keeps deepening and spreading a junction this process already called finished:
where the sum runs over every high-temperature step the wafer experiences after a dopant is introduced, including this regrowth and the second diffusion still to come. The consequence is specific and easy to get wrong: Step 13's furnace schedule cannot target the final junction depth directly, because this step and the ones after it will keep adding to $(Dt)_{\text{effective}}$ long after Step 13's furnace door has closed — so Step 13 has to aim short, on purpose, and trust this step and the ones after it to finish the job it deliberately left incomplete.
## 2. Real Diagram: The Pocket Grows While Nothing Visible Happens to It
Stripping the old oxide and growing a fresh one looks, from the outside, like resetting the wafer back to a clean surface ready for the next pattern. It is not a reset for the dopant already inside the silicon — the diffused pocket from Step 13 is still there through every panel of this sequence, and it is measurably deeper and wider by the time the fresh oxide finishes growing than it was the moment Step 13's furnace door opened.
## 3. Why No Earlier Series Ever Had to Ask What Would Happen to a Wafer Later
Nothing in the 1954 diffused-base or 1956 drift-transistor process needed a step like this, because neither process ever ran a second patterned operation on the same wafer — the 1956 mesa etch, for instance, happened once, after every diffusion in that process was already complete. The ability to run multiple patterned levels is what separates a *process* from a mere *sequence*, and this step is the first place that ability's real cost becomes visible: every recipe upstream of a second masking level now has to be designed around the thermal history the wafer has not experienced yet. That is a new kind of engineering discipline this project has not needed before, and it will stay necessary for the rest of this series.
Step 14 does not undo anything Step 13 did; it resets the mask the wafer wears while quietly continuing to finish the diffusion Step 13 deliberately left unfinished, on a schedule Step 16 will be the one to finally measure.