Focus-Exposure Matrix
A focus-exposure matrix (FEM) is a lithographic experiment that deliberately exposes a test wafer with systematically varying focus and dose combinations across its fields — one specific focus setting and one specific exposure dose assigned to each individual field, turning the whole wafer into a structured, two-dimensional map of exactly how patterning quality responds across the entire parameter space at once. It's the standard, foundational method for qualifying any new lithography process or mask design, precisely because it generates the raw data every subsequent process-window analysis actually depends on.
Rows and columns in the matrix each correspond to a single, specific process variable held constant across every field in that row or column, which is exactly what lets the resulting data isolate each variable's independent effect. Every field in a given row shares the same focus setting (commonly ranging across something like -100 to +100 nanometers in 10-nanometer steps) while dose varies across that row's columns; every field in a given column shares the same exposure dose (often varied across roughly plus-or-minus 10 percent around nominal, in fine percentage steps) while focus varies down that column. This structured grid is what lets subsequent analysis cleanly separate focus's effect on patterning from dose's effect, rather than conflating the two.
Matrix size — typically ten to twenty focus settings crossed with ten to twenty dose settings — represents a real, deliberate tradeoff between how finely the parameter space gets sampled and how much wafer area and measurement time the whole experiment consumes. A finer grid, with more focus and dose steps, captures the true response surface with higher resolution, revealing subtler process behavior a coarser grid might miss entirely. But each additional row or column means more fields that need to be printed, developed, and individually measured — directly multiplying the time and cost of running the full experiment. This tradeoff is exactly why matrix size gets chosen deliberately based on what level of detail a given qualification effort actually needs, rather than defaulting to the largest grid possible every time.
Post-develop CD measurement at every single field is what actually turns the matrix from a printed pattern into usable data, converting each field's physical outcome into one data point in the two-dimensional focus-dose response map. After the wafer develops, measuring critical dimension at each individual field — using whatever CD metrology technique is appropriate — produces exactly one CD value per focus-dose combination tested. Plotting all of these measured CD values back against their corresponding focus and dose coordinates reconstructs the complete response surface showing exactly how CD actually varies across the full range of focus and dose conditions the experiment covered.
Bossung curves are the standard way this FEM data gets visualized and interpreted, plotting measured CD against dose for several different fixed focus values, each as its own curve on the same chart. This particular visualization makes focus and dose's combined effect on CD immediately visible: curves for different focus values that stay close together and relatively flat across a wide dose range indicate a process robust to both focus and dose variation; curves that diverge sharply or shift CD dramatically with small focus changes reveal a process with much less margin for real-world variation. Bossung curves are the direct bridge between raw FEM measurement data and the process-window boundaries that analysis ultimately extracts from it.
FEM's role as the standard qualification method for new lithography processes and mask designs reflects exactly what makes it uniquely valuable: it characterizes the complete process response across the full relevant parameter space in one coordinated experiment, rather than testing isolated single conditions one at a time. A new process or mask design genuinely needs to be understood across a realistic range of focus and dose conditions — the real variation a production line will actually experience — not just verified to work at one single nominal setting. Running a comprehensive FEM once, early in qualification, provides exactly the complete response-surface data needed to confidently determine process window boundaries, rather than requiring repeated piecemeal experiments to patch together an incomplete picture of how the process actually behaves.