Monitor and manage

Continuous monitoring of variables such as airborne particle contamination, airflow velocity and parallelism, which potentially could lead to environmental contamination, is one of the most significant aspects of cleanroom control. Software applications developed to support monitoring are consolidating over time. An interesting aspect is represented by a software application capable of monitoring the cleanroom layout with reference to filter measures. Monitoring is intended to provide the capability to file data on the variables, with the possibility of viewing a graphic layout showing the status of each filter contained in controlled contamination areas in detail. The software has the following advantages: 1. easy measure memorisation 2. viewing of filter status in graphic format 3. simple data consultation.

Particle monitoring of cleanrooms, in terms of contamination class, still represents state-of-the-art performance monitoring in controlled contamination environments. The need for observation of molecular compounds (AMC: Airborne Molecular Contamination) is arising in the new millennium; AMC is however a field for future development. In an open space cleanroom like the one in the Micron plant of Avezzano, contamination control must involve specific infrastructures such as filters in addition to the environment. This Class 1 cleanroom has an area of about 7,500m2 and is located between two areas defined as machine plenum and filter plenum. Fig. 1 is a diagrammatic view of the situation.

Focus on filters The filter plenum is the area where the air coming from the machine plenum is conveyed through powerful fans (VLF). The machine plenum area receives both the air from the cleanroom and the air coming from fresh air input and treatment systems (make-up air). The filtering infrastructure that separates the Class 1 cleanroom from the filter plenum is obtained with ULPA16 type filters. The need to set periodic testing of the filtering infrastructure may seem obvious but if we consider that the filtering infrastructure is composed of over 5,000 filters (production area), testing of particles (leak test), velocity and parallelism requires an organisational structure to support the activity and design of a software application to manage and process testing information. This paper is organised in compliance with PDCA (Plan Do Check Action) methodology. This highlights the improvements obtained by setting the periodical testing of the filtering infrastructure in terms of tightness (leak test), velocity and parallelism of the air flow. The following advantages can be considered: a) Electronic management of information produced by testing activities b) Rapid processing of velocity, parallelism and tightness data of the filtering system c) Improved management costs to maintain the efficiency of the filtering system d) Optimisation of infrastructure status visibility: all of the above in benefit of products in production bays e) Optimisation of maintenance activities on filtering system.

PDCA methodology Following is a summary of activities associated with PDCA. Plan: This stage was dedicated to the verification of specific measure standards for periodic monitoring of the filtering system. The ISO 14644-3 standard was used as reference for leak test activities. The VDI2083/3 was used for airflow velocity and parallelism testing. As regards the software application, the possibility of developing an AutoCad-interfaced software application environment was checked. Do: This stage was dedicated to realising testing procedures and complete measure activities on the entire infrastructure. Software application was implemented by the IT team. Check: This stage included processing of environmental contamination data coming from the on-line data acquisition system (Facility Net) as well as velocity and parallelism data processing. Action: This stage included the validation of testing activities with reaction procedures to findings, infrastructure maintenance activities and activities for the modification of extractor system settings to guarantee air flow balance after structural variations of the cleanroom layout.

Filter tightness monitoring For years it has been a common belief that the tightness of a cleanroom filtering system could guarantee absolute separation with the upper plenum (filter plenum). However, this was questioned when a particle contamination event of environmental type, recorded on the on-line testing system (see Fig. 2), was related to a particle leak produced by a microhole on a ULPA16 filter. In fact, the particle concentration value expressed in particles per cubic foot recorded on sampling point C06 results in environmental dilution because the concentration values measured on the micro hole a few centimetres under the filter with a mobile station (off-line test) were about 10Kpcs per cubic foot per minute (cfm). As shown by the trend in Fig. 2, the problem was solved by repairing the filter properly. This allowed us to decide the need to set periodical testing on all filters (~5,000) in the production area. The analysis of particle contamination data in the Wet area shown in Fig. 3 from October 2002 to January 2003 showed improvement in terms of number of warnings (blue line), number of alarms (red line) and number of occurrences higher than 10 0.1 um/cfm particles (yellow line). All of this after the combined action of maintenance activities on 28 ULPA16 filters (November to December 2002) and velocity reduction of airflow produced by filters (starting from middle of December 2002) through the reduction of the pressure differential (Dp) between the cleanroom and filter plenum. This last item is also discussed in the air balance section. This urged the company to set specific reaction models to control the environment not only in terms of specification levels, but also of baseline trend. The baseline deviation of sampling point C06 was induced by a filter at a distance of approximately 6m, and not by close filters along the vertical. This is direct evidence of how transversal flows can make it difficult to identify contamination sources. In conclusion, monitoring issues of airborne environmental particle contamination must be integrated with issues strictly related to airflow balance.

Software tool To support the activities of the Air Balance and Contamination Control teams in the IT department, we designed a software application that allows: File measure data on laminar flow and particle contamination on filters. Process measure data to determine the filter status based on specification limits. Generate and view the cleanroom layout with graphic representation of filter status. The decision to develop the software application in-house enabled us to design a tool capable of satisfying the different requirements of the teams. The project was realised by using the RUP (Rational Unified Process) methodology. In this specific case the RUP methodology contributed to integrate independent entities and products within Micron. The Generation and Viewing section illustrates how products like AutoCad2002 were integrated with the software application in order to improve existing business modelling. The main features of the software application are as follows: - Data filing - Data processing - Layout generation and viewing The software operation is based on the concept that each filter of the cleanroom is placed in a location that can be identified with Cartesian coordinates (x, y). N filters can transit through each position identified with (x, y). Each filter is identified by a serial number. This guarantees that: a) every filter is associated with a set of coordinates (x,y) that is passed through at the same time, mantaining the relative measurements and b) every coordinate (x,y) identifies a set of filters that have passed through it.

Data filing Data is filed through a specific graphic interface. The memorisation of data by the software application and the suitable structure of the database allow us to extract measurements and provide reports on filter measurements to the final user. Data processing is intended to track out-of-specification events according to velocity and particle measures on filters. The status of each filter (if installed) is signalled by the application based on the measures made on the filter. The output (filter status) is calculated by a suitable algebraic function to take account of different environmental parameters. The specification constraints can be modified with the application.

Generation and viewing The processing illustrated in the section above allows us to take a step forward towards monitoring. The status of each filter is constantly present in the application data base. And the graphic layout, which represents the C/R filter layout, is constantly updated. Fig. 4 shows the database (section A) with the status of each filter represented by an ACAD numerical code and the database (section B) with the geographical position of each filter in the cleanroom C representing the ACADA layout that gives a graphic representation of filter positions (x, y). The improvement allowed for interaction of A with C (dotted line). The result is section D, an AutoCad layout that includes also A, not only B. The codes contained in A allow us to colour each filter based on its status. The layout is a superimposition of multiple AutoCad layers. Each layer has a specific meaning. Each colour of the layer is dedicated to indicate the filter status. The layer contains green filters (specification filters), yellow filters (warning filters) and red filters (out-of-specification filters). The arrow indicates the value of the airflow direction after parallelism testing. The superimposition of multiple layers based on a suitable colour code gives the user an evidence of the filter status in relation to the combination of air balance and environmental contamination data. Considering the type of cleanroom in the Avezzano Micron plant, the analysis of measures of variables linked to air balance with the following corrective actions is not immediate, since it requires specific experience and technical knowledge. To make it easier to understand this paper, we followed a chronological approach that illustrates the main steps required to take appropriate decisions based on measured data and data processing.

PDCA process For the PDCA, the first step was to measure all air velocity values on all filters in the production areas. This led to statistical processing, which showed the existence of a quite pronounced gradient effect along the north-south direction of the cleanroom. As shown in Fig. 6, this showed the imbalance of the cleanroom system in relation to airflow velocity. We then evaluated pressure differentials between the filter plenum and production areas to check alignment. Measures confirmed non alignment for this parameter in the three sub-environments of the cleanroom. The Dp value was higher on the south of the cleanroom. This required adjustment of the amperage of all extractor systems (VLF) in the southern area of the machine plenum. The same was done in the central and northern areas. The pressure differential non-alignment with consequent diversification of some velocity values measured next to the filters not only depends on parameters related to the layout and other infrastructure factors but also on the fact that not all installed filters had the same characteristics in terms of pressure differential set by the manufacturer. This was one of the main reasons to optimise material incoming procedures during purchasing. The controlled reduction of amperage on VLF produced some benefits in terms of airflow velocity with a reduction of the gradient effect shown in Fig. 6. Further to this, the parallelism associated with the airflow produced by each filter in the production areas was evaluated. In total, about 10,000 velocity and parallelism measures were made, revealing problems in both aspects. In some cases, they showed a direct correlation with significant variation on the distribution of calculations of environmental particle concentration values. In particular, very high angular deviation of the airflow was calculated in the north-west area of the cleanroom, near two sampling points indicated as F01-F02 (77 degrees with respect to the vertical). The distribution of environmental particle concentration values (Fig. 7) on sampling points F01-F02 showed a clear difference from the others as they were completely immersed in the area affected by turbulence. In Fig. 5 of the software section a white arrow (set by the software according to measured data) indicates the airflow direction in production areas in view of the obtained parallelism measures.

Conclusions Within the general optics of environmental control, the integration of information supported by specific software tools coming from airborne environmental contamination control activities and airflow velocity and direction control activities allows optimisation of the cleanroom performance in order to obtain higher environmental control and cleanness for the production of new generation wafer devices.

Special thanks For their perseverance and effort our sincere thanks go to our colleagues in Contamination Control, Construction, Air Balance, and Drafting. Special thanks to Mario Gasperini, section manager of the Material Analysis Department of Micron Technology of Avezzano.