Abstract:

Semiconductor manufacturing requirements continuously challenge the particle measurement supply base to provide instruments capable of detecting and sizing increasingly smaller particles. A number of manufacturers have or are introducing new liquid optical particle counters (OPC) with smaller particle size counting capabilities. It is important that the users of these instruments understand the particle counting efficiency of these instruments relative to particle size as they apply these improved instruments to their process measurements.

SEMI recently released C77-0912, TEST METHOD FOR DETERMINING THE COUNTING EFFICIENCY OF LIQUID-BORNE PARTICLE COUNTERS FOR WHICH THE MINIMUM DETECTABLE PARTICLE SIZE IS BETWEEN 30 nm AND 100 nm. This method specifies the use of mono-dispersed polystyrene latex (PSL) particles in suspension as the test material. One of the limitations recognized in the document is the reality that PSL particle size distributions are often not mono-dispersed. This is particularly true for PSL particles 50 nm and smaller where the particles have broad size distributions that are not normally distributed. It is important that the particle size distribution of the PSL be well understood and accounted for in the analysis of instrument size detection efficiency.

This paper will review the results of measuring the counting efficiency of several OPC’s (current and next generation) using multiple sizes of PSL and the method described in SEMI C77. The particle size distributions of the PSL used in this testing were measured using Liquid Nanoparticle Sizing (LNS) technology. This method converts the liquid into an aerosol the particles in which are then sized and counted using a scanning mobility particle sizing system. The PSL challenge solutions were prepared using the concentration data determined using the LNS measurements. In addition to the “mono-dispersed” PSL challenges, poly-dispersed challenges were also prepared with varying particle concentration vs. size slopes for measuring instrument response to a simulated, “real-world” distribution.