A fractional factorial design is a DOE approach that tests only a carefully selected subset of the full factorial combinations, dramatically reducing the number of experimental runs while still extracting the most important information about main effects and key interactions.

Why Fractional Factorial?

• A full factorial with 7 factors at 2 levels requires $2^7 = 128$ runs — impractical in semiconductor manufacturing where each run costs wafers and fab time.
• A half-fraction ($2^{7-1}$) requires only 64 runs. A quarter-fraction ($2^{7-2}$) needs only 32 runs. An eighth-fraction ($2^{7-3}$) needs just 16 runs.
• The tradeoff: fewer runs means some effects become aliased (confounded) — you can't distinguish between certain main effects and interactions.

How It Works

• In a $2^{k-p}$ fractional factorial, $k$ is the number of factors and $p$ is the number of fractions (each $p$ halves the runs).
• The arrangement is chosen using generators — mathematical relationships that define which combinations to include.
• Example: $2^{4-1}$ = 8 runs for 4 factors (instead of 16). Factor D is defined as $D = A \times B \times C$. This means the main effect of D is aliased with the 3-way interaction $ABC$.

Resolution

• Resolution III: Main effects are aliased with 2-factor interactions. Useful for screening many factors but risky if interactions are large.
• Resolution IV: Main effects are clear of 2-factor interactions, but 2-factor interactions are aliased with other 2-factor interactions.
• Resolution V: Main effects and 2-factor interactions are clear of each other. 2-factor interactions are aliased with 3-factor interactions (usually negligible).
• Higher resolution = better information but more runs.

Semiconductor Applications

• Screening DOEs: When 6–10+ factors need initial evaluation, use Resolution III or IV fractional factorials to identify the 3–4 most important factors.
• Follow-Up: After screening, run a full factorial or RSM on only the important factors identified in the screening step.
• Process Transfer: When transferring a process to a new tool or fab, screen for factors that need adjustment.

The Sparsity Principle

Fractional factorials work because of two empirical observations:

• Effect Sparsity: In most real systems, only a few factors (and even fewer interactions) are important.
• Effect Hierarchy: Main effects are generally larger than 2-factor interactions, which are larger than 3-factor interactions.
• These principles mean that the information lost through aliasing usually involves effects that are negligibly small.

Fractional factorial designs are the workhorse of screening experiments — they efficiently separate the vital few factors from the trivial many with minimal experimental cost.