Insights
Notes on semiconductor and display process engineering — process conditions, device characteristics and defect analysis.

CVD Film Uniformity: What a Radial Wafer Map Cannot Show You
A CVD wafer map built from a radial profile shows ring-shaped thickness spread, not left-right or angular patterns. What the two uniformity numbers mean and where the map stops.

ALD Temperature Window: What Happens on Each Side and How to Find It
The ALD temperature window is where both half-reactions saturate. Learn why GPC rises or falls below and above it, and how to measure and report your own window.

Etch Selectivity Formula: Rate Ratio or Mask Actually Consumed
Etch selectivity is a ratio of two etch rates, but on a patterned wafer the depth etched per mask consumed is usually lower. Learn both formulas and how to report them.

Semiconductor Device Cross Section: How to Draw One That Reads at a Glance
What a semiconductor device cross section must show — layer order, materials, thickness with units, dimension lines, a scale note — and when to switch from 2D to 3D.

ALD Purge Time: How Long Is Long Enough?
Learn how to set ALD purge time by finding the saturation plateau and why the right value changes between tools.

Film Thickness Uniformity Formula: Which One Your Number Came From
Three film thickness uniformity formulas give three different percentages from the same wafer. Learn which one your tool uses and how to report it.

Etch Non-Uniformity: How to Find Which Knob Is Driving It
Etch non-uniformity has many possible causes, but at any given operating point one variable usually dominates. Identify it through sensitivity analysis, then confirm on wafers.

Step Coverage in ALD: Why the Bottom of the Trench Thins First
In ALD, step coverage drops from the bottom up at high aspect ratio. Which half-cycle runs short, why PEALD falls off earlier, and what to adjust first.

Is That ALD GPC Number Measured or Predicted?
The unit is the same either way. Before you place an ALD GPC value next to your own, check whether it was measured on a wafer or produced by a model — and write that down next to the number.

How to Actually Reduce the Number of DOE Splits
Narrowing the range does not cut the run count. What reduces the number of DOE splits is dropping factors and cutting levels — and how to read a model-based screening pass before you spend wafers on it.
