Etch Chamber Wall Memory Budget
# Etch Chamber Wall Memory Budget
An etch chamber wall memory budget accounts for the deposit left on the reactor walls by one process and how it changes a later process run in the same chamber, and bounds the resulting variation. It treats the wall state as a hidden input that carries history between runs, steps and chemistries.
The inner walls of a plasma etch reactor accumulate films from the gases and etch products they are exposed to. Those films set how the walls recombine or release reactive species, so a process started against walls conditioned by a previous process is not the same as one started against reset walls. This budget does not restate the first-wafer or seasoning budgets; it focuses on the memory carried by the wall deposit across processes.
## See the walls as a carrier of history
During etching, reactive species reach the walls and are lost there at a rate set by the wall coating. A deposit from an earlier process therefore changes the neutral and ion balance that the next wafer sees, even with an identical recipe, because the walls participate in the chemistry.
The sketch shows a chamber whose walls carry a deposit that influences wall loss, with one run leaving a layer that the next run sees. It is not a deposition model and names no specific layer for a real tool; the composition and its effect depend on the chemistry and the history.
Reported work identifies these layers directly. Studies of gate-etch reactors describe silicon-oxychloride and fluorocarbon-type films building on the walls, and note that at low pressure, diffusion and recombination at the reactor wall can be a dominant loss mechanism for reactive species. The wall is an active surface, not a passive boundary.
## Recognize that the layer is process-specific
The wall layer is not one fixed film. Its composition shifts with the chemistry and conditions of the process that deposited it, so the memory a chamber carries depends on what ran before. A layer left by one etch can present a different surface to the next than a layer left by another.
Distinguish a wall-memory effect from a within-run drift or a hardware change. A result that shifts when the preceding process or clean changes points to the wall state; attributing it to the recipe alone misses the real input.
## Reset the walls to a known state
Because the walls carry memory, reproducibility comes from starting each process against a known wall condition. A reported conditioning strategy deposited a thin carbon-rich layer on the reactor walls before etching, and after the silicon gate etch cleared the oxychloride layer and removed the carbon layer with an oxygen plasma, resetting the walls to their initial state; the authors reported the process became reproducible because it always started under the same condition.
Define the reset or conditioning the budget relies on, and state where in the sequence it occurs. A clean or conditioning step that returns the walls to a repeatable state is what makes the recipe result comparable from run to run.
Separate the wall state from the recipe when recording a result. The same recipe against two wall states is effectively two processes, so the wall condition belongs in the process record alongside the recipe.
## Budget shared chambers and step order
In a chamber that runs more than one process or chemistry, the order of steps and the sharing of the tool matter. A step can leave a deposit that biases the next, so the budget should state the sequence assumed and any clean between steps, not just each step in isolation.
No single clean or conditioning is universal. A reset tuned for one chemistry may not fully clear another, and adding cleans costs throughput, so the approach needs qualification on the real sequence and tool.
## Close the budget with traceable evidence
Specify the assumed wall state, the preceding process or clean, the conditioning or reset step, and the results tied to that wall condition, at a named point in the sequence. Retain chemistry, step order and recipe revision so a later result can be compared with the earlier one.
A chamber wall memory budget ties a result to the wall state it ran against and the reset that defines that state. Wall condition, reset method and recipe result are separate questions, and each needs its own evidence.
## Sources and scope
- [Plasma-wall interaction / reactor wall conditioning for gate etch reproducibility (CNRS-LTM, AVS 2005, Paper PS-ThM3)](https://www2.avs.org/symposium2005/Papers/Paper_PS-ThM3.html): describes depositing a thin carbon layer on the walls before etching and, after gate etch, clearing the oxychloride layer and removing the carbon to reset the walls to their initial state for reproducibility.
- [Reactive-species loss at reactor walls coated with SiOCl or CFx polymers (AVS 2009, Paper PS2-ThA11)](https://www2.avs.org/symposium2009/Papers/Paper_PS2-ThA11.html): notes that at low pressure, diffusion and recombination at the reactor wall can be a dominant loss mechanism for reactive species, studied for chambers coated with SiOCl or fluorocarbon films.
Original diagrams and explanatory synthesis. No universal wall layer, reset recipe or qualified sequence is implied.