Yield losses due to airborne molecular contamination (AMC) issues started to emerge in semiconductor manufacturing over 20 years ago.¹ At that time, contamination monitoring was typically conducted by manually pulling air samples through a bubbler or impinger. Then, the samples would be taken to a chemical analysis laboratory to determine the content. This process is labor intensive, and it can take days to obtain a result. Samples were usually collected as a reactive measure to determine the cause of yield problems or as part of a monthly or quarterly monitoring program.

In the 1990s, automated monitoring systems based on ion mobility spectrometry (IMS) and other technologies became available. These systems were designed to monitor continuously and required little labor to operate and maintain. The complexity and cost of these systems made it impractical to install one analyzer system for each monitoring location. The analyzer systems were almost always coupled with a sequential sampling manifold system. This allowed as many as 60 locations to be monitored daily by a single analyzer system and represented an improvement in measurement coverage over manual sampling techniques.

Manifold sampling systems cycle through multiple monitoring locations, providing a brief measurement of each location, without regard to the relative importance of monitoring AMC at each location. As clean manufacturing has advanced and increasingly cleaner environments have been required for certain manufacturing processes, it has become clear that not every location in the cleanroom has the same risk level regarding molecular contamination. High risk areas are those areas where sensitive products or processes are exposed or where there is a risk of molecular contamination generation that can lead to significant yield loss. Lower risk areas are non-critical process areas and background monitoring. The identification of high risk areas and the impact AMC can have, has led to an emerging need for continuous monitoring of these critical locations.

While the large manifold systems of the past provided better monitoring coverage than manual sampling techniques, they still left each location unmonitored for more than 99% of each day (Figure 1). The few minutes of monitoring time per day are not always representative of the entire day and missed contamination events made it difficult to diagnose molecular contamination issues which can be transient.