JPACSM 11 A CHEMICAL SENSOR BASED SCHEME TO EVALUATE PACKAGING CONTAINMENT EFFECTIVENESS PART 1: DEVELOPMENT AND OPTIMIZATION Gregory K. Webster Pfizer Global Research and Development Ann Arbor, MI Douglas A. Farrand Pfizer Global Research and Development Groton, CT Mark A. Litchman Quality Operations Pfizer Global Manufacturing Lee s Summit, MO KEYWORDS: SURFACE ACOUSTIC WAVE DETECTION, METAL OXIDE GAS SENSOR, CONTAINER ANALYSIS, LEAK TESTING ABSTRACT A novel testing scheme was developed as a diagnostic tool to ensure primary packaging integrity for liquid products. The previous method of testing for primary packaging integrity involved manually applying pressure to each filled unit. In the spirit of continual improvement, along with the desire to minimize the variation inherent in manual test methods, a new test methodology was developed that relied on the application of sensor technology. The chosen alternative to traditional volatile organic carbon (VOC) detection was the use of a surface acoustic wave (SAW) or metal oxide gas sensor (FIS) device. As a chemical sensor, the SAW/FIS has notable traits such as high accuracy, extreme sensitivity, small size, reliable and fast interrogation, digital output, and perhaps most importantly, is cost effective. The detector used was chosen not only because of its SAW/FIS device capabilities, but also because it provided a self-contained gas chromatograph system which used air for its mobile phase. Though the SAW/FIS polymer membrane provides a certain level of selectivity, flame ionization work demonstrated the need for a separation column to ensure the integrity of the VOC response. As configured, the SAW/FIS detector was not only safe, but also operated as an environmentally green instrument in that no mobile phase or detector gases needed to be supplied. INTRODUCTION The goal of any manufacturing operation is to safely produce and supply high quality products to the marketplace at a competitive price. When a customer receives a product not fulfilling one of these elements, the business relationship can suffer, and if not corrected, may affect the future viability of the business. One attribute that is particularly bothersome to a customer is the receipt of a product that has a breach in its packaging integrity. For this reason, in-process checks, both mechanical and analytical, are routinely incorporated into each manufacturing process to ensure that the process is in control and reliably producing high quality product. Literature in this regard includes general reviews 1-3 and well as specific references for helium 4 and pressure 5 testing. Applications are common in the plastics and automotive 6, semiconductor 7, and medical 8,9 industries. Helium testing is very effective because this gas can permeate through the smallest of openings. Such testing was more applicable in piloting our process rather than for use during manufacturing. Secondly, since moisture permeation over time and stress conditions are common for many plastics, there was a concern that helium would be overly challenging to the containers used. Pressure testing was also used as an effective tool in piloting and as an investigational tool, rather than for process control. The simplest way to check packaging integrity is to manually check the packaging for breaches in seal integrity. For liquid and paste products, simply squeezing each individual package is sufficient for applying the force expected in the marketplace. However, squeezing each individual package is tedious and costly, as is automation of this process. The goal of this study was to develop a diagnostic tool for inspecting packaging integrity as well as for checking in-process quality. Knowing that a stress needed to be placed on the product packaging for inspection, the squeezing mechanism was replaced by pulling a vacuum on a series of product containers. With this stress applied, the success of this scheme was based not only on the ability to design a test system capable of drawing product through suspect openings, but also on the ability to detect any escaping product and thus signal breaches in packaging integrity.
JPACSM 12 EXPERIMENTAL Gas Chromatographs: Hewlett Packard 5890 and 6890 using Chemstation Chromatography Manager (Palo Alto, CA) SAW/FIS Detector: MicroSensor MSI-301 (Bowling Green, KY) Solvent Gas Standard: Matheson Tri-Gas Certified Gas Mixture in a 6R gas bottle or equivalent. (Montgomeryville, PA) The operating conditions for the GC and the SAW/FIS device are given in Tables I and II, respectively. Table I GC Operating Conditions GC capillary column Oven temperature 55 C Inlet temperature 200 C Detector temperature 250 C Carrier flow, Helium Split flow Septum purge flow Air flow Hydrogen flow Make up flow, Helium Inlet Purge Run time J&W Scientific, DB Wax Mega 30M, 0.53 mm Megabore, 1µm film 20.1 ml/minute or 110 cm/sec linear velocity at 30 C 100 ml/minute 3.0 ml/minute 450 ml/minute 40 ml/minute 25 ml/minute On at 1 minute, Off at 2 minutes 10 minutes RESULTS AND DISCUSSION Preliminary Studies in Detector Selection Visual Inspection The development of a leak detection scheme began with the idea of spreading out product tubes on a series of vacuum chamber shelves lined with colored paper. The notion was to apply enough vacuum to entice the product out of any seal openings and onto the lining. The lining would then be inspected for spotting; a clean liner would be indicative of seal integrity. While the product tested for this study consisted of a volatile solvent, a significant portion of the matrix is not volatile and readily spots adsorbent paper. While this design could be a quick test, the nature of the test relied on actual surface contact of the escaping product with the shelf covering. Such a scheme was found to be neither sensitive nor empirical in nature. However, the experiments were successful in proving that a vacuum system could be used to draw the product matrix out of tubes where the seal integrity had been breached. Table II MSI-301 Operating Conditions Packed column GC duration Injection duration Sample duration Pre-heat duration Injector Purge Column heater Signal start Signal End Peak search start Peak 1 location Window width 10 Calculation Method Response factor 1 Vap1, Pt 1, concentration 15 Supelco, 3 7 x 1/8 ID column, packed with 0.10% SP-1000 on 80/100 Carbopack C with a 2 void at both ends of the column. 300 seconds 20 seconds 10 seconds 20 seconds 30 seconds 55 C 10 seconds 300 seconds 40 seconds 128 seconds Linear Vap1, Pt 1, height 45000 Calibration SAW/FIS Device Conc = 3.33333E-4 * signal(hz) Microsensor Systems #MS104012 Photoionization Detection The manufacturing solvent for this study can be classified as a VOC (volatile organic carbon) source. In today s industrial setting, VOCs are routinely monitored by photoionization detectors that are both inexpensive and available as hand held, portable instrumentation. Visual inspection of shelf linings could be made more quantitative using such a detector. With a photoionization detector, any compound that has an ionization potential lower than the lamp photons can generally be measured. Under vacuum, the solvent from any product matrix leak will vaporize into the vacuum chamber. After the chamber is re-equilibrated to ambient pressure, this volatilized product solvent can be analyzed simply by testing the chamber atmosphere. To expound on this scheme, initial sniffing studies used a portable photoionization detector. The unit used was a MiniRAE 2000 Portable VOC Monitor Model PGM-7600 equipped with a 10.6 ev lamp (RAE Systems Inc, Sunnyvale, CA). For our application, the manufacturer recommended the 10.6 ev lamp. Product tubes that were identified as having a significant seal leak by visual examination, or had a seal breach induced by analysts
JPACSM 13 using sharp objects, were placed in a one cubic foot vacuum chamber (Labconco part 55300-00, Kansas City, MO) and exposed to approximately 20-25 inches of Hg vacuum from a house line. After holding vacuum for a few minutes and subsequently bleeding the chamber through a needle valve back to ambient pressure, the chamber air was tested using the VOC monitor and a significant response (10-20 ppm) was recorded. The experimental data continued to support the premise that a vacuum chamber scheme was worth developing. The scheme eliminated the need for individual tube squeezing and allowed a batch processing design for inspection. The chamber could hold and process several tubes at once, while manual squeezing required individual processing of each production tube. However, when challenged with a pinhole leak in the product tube, the photoionization detector (PID) used in the VOC monitor did not yield a significant signal above background. While photoionization detectors are generally considered to be a sensitive detector; the sensitivity for our solvent was low due to this solvent s relatively high ionization potential. An additional concern was the efficiency and cost of consumable lamps used in this detector. Because of this low sensitivity and high consumable cost potential, further development studies with PID were abandoned at this point. A related concern to the sensitivity of the photoionization detector was that the preliminary results were found to be biased with reliance upon actual contact of the escaping product with the shelf lining in order to siphon an appreciable amount of product and ultimately volatilize the matrix into the chamber for detection. This discovery was later expanded upon in the optimization of the vacuum chamber conditions. This bias can be minimized with the understanding that the detector used must be sensitive for trace amounts of solvent; any siphoning effect would not be detected if the suspect tubes were oriented in a way that no product contact with the liner would be made. In this scenario, the detector must be able to solely discriminate volatized matrix from the chamber background at very low levels. The preliminary data from the photoionization detector lead us to believe this detector was not applicable to this scenario; a search for more sensitive detection design commenced. Draeger Tubes In a manufacturing operation, keeping processes simple generally results in a more acceptable process. A visual inspection for leaking seals would be ideal in that it requires less training in a manufacturing line. It was already shown that previous visual inspection of shelf lining was found to be biased and that the vacuum chamber scheme needed to induce the volatilized solvent in the chamber compartment. However, if a simple visual test for analyzing the volatilized matrix could be developed, it would keep the scheme relatively simple. For this reason, the use of colorimetric VOC detection tubes was investigated. Along with instrumental VOC monitors, there are numerous industrial applications for the use of colorimetric detector tubes, which are available from several vendors. Generally, the sensitivity range from vendor to vendor is relatively constant. Using the same system discussed with the photoionization detector, the use of a Draeger tube to monitor the chamber air was investigated to determine if an inspection could be made by viewing a color change. The tubes used for this study were Draeger Short-term Detection Tubes with a measuring range of 25-5,000 ppm for the components of interest (SKC Inc, Houston, TX). An advantage of using a Draeger tube is that the total chamber volume could be pumped through the device. This, in effect, concentrated the volatilized solvent. If successful, the higher analytical range for the solvent recovered by the Draeger tube would not be an issue. The Draeger tubes yielded an easily noticeable color change for the containers used to simulate a significant seal leak and for those leaving a spotted liner. Yet, in a fashion similar to the photoionization detector, the colorimetric tubes did not yield an intense color change when challenged with a pinhole leak in the product tube, nor was a significant color change produced in the absence of spotted liner. It was hard to distinguish any color change from the original background for these samples. Thus the test went from an obvious and empirical change to a best guess. Development studies with colorimetric tubes were abandoned at this point. From the Draeger tube experiments it was confirmed that the test scheme required either a detector with the ability to detect trace matrix levels or a detection scheme that incorporates a concentration step with sample collection. Flame Ionization Detection In order to attain greater sensitivity than achieved with the photoionization and colorimetric tube studies, a new detector was needed if chamber sampling was to be an effective and empirical measure of tube seals. The traditional and routine way to detect low-level organics is through the use of flame ionization detection. With flame ionization detection, the response of the signal is proportional to the carbon content of the eluent. Many models for VOC detection in process streams and portable flame ionization instruments were reviewed on the Internet. However, in order to avoid investment at this early stage in development, it was decided to first evaluate flame ionization detector capability using in-house instrumentation in both the capillary and packed column modes. One of the vacuum chamber outlets was configured with standard gas chromatography septa over a ¼ inch Swaglok fitting. This enabled the chamber atmosphere to be sampled by a gas syringe. The syringe was
JPACSM 14 then manually transported and the 500 µl sample was injected into the gas chromatography systems. Vacuum chamber sampling confirmed that the use of flame ionization would be an effective and sensitive detector for the leak detection scheme. However, with flame ionization detectors, the system has potential drawbacks: 1. The detector requires consumable gases, one of which (hydrogen) has safety concerns in a manufacturing environment; 2. Flame ionization detectors use a small flame which is a potential hazard in areas having flammable vapors, and; 3. The instrument needs plant personnel to continuously monitor that the detector flame is lit. However, because this detection scheme was still the only design to date that satisfied the chamber requirements, developmental studies using flame ionization detection continued in tandem throughout this project with the viewpoint that flame ionization may perhaps be the primary, back-up, or complimentary mode of signal response. Although FID detectors are often marketed as intrinsically safe due to their housing, and it is acknowledged that the actual flame in an FID is confined in a controlled environment, caution is still warranted. Throughout the preliminary stages of this project, there was concern about whether the use of a flame ionization detector was safe or even allowed in a flammable solvent area. Alternative detection schemes needed to be investigated. The flame ionization detector provided the necessary sensitivity for the vacuum chamber scheme, even in the absence of a spotted liner. Shelving and lining were no longer required for the chamber and were discarded. Removing the shelves would be required to be effective in a batch-processing mode so the vacuum chamber can process pails of product tubes, not portions of each pail. In addition, this investigation found that plasticizers in the liner of the product container could produce a positive bias on the VOC result. The use of a gas chromatograph system provided not only the required sensitivity, but also allowed for signal integrity in the matrix response. The plasticizer peaks were easily resolved from the matrix and eluted at the beginning of the chromatogram. Surface Acoustic Wave Detection The chosen alternative to flame ionization detection was to use a surface acoustic wave (SAW/FIS) device. Like flame ionization, SAW/FIS detection has proven effective for low-level detection in gas chromatography and VOC monitoring 10-13. SAW/FIS devices are more accurately defined as chemical sensors than as instrument detectors. As a chemical sensor, the SAW/FIS has such notable traits as high accuracy, extreme sensitivity, small size, reliable and fast interrogation, digital output, and perhaps most importantly, is cost effective. The SAW/FIS detector unit chosen for this study also included a self-contained gas chromatograph system with an air pump so that air could be used for its mobile phase. Though the SAW/FIS polymer member provides a certain level of selectivity, the flame ionization work demonstrated the need for a separation column to ensure the integrity of solvent response. As configured, the SAW/FIS unit eliminated the previous safety concerns and also operated as an environmentally green instrument, requiring no mobile phase or detector gases. The vacuum chamber system continued to be sampled via a syringe port as discussed in the flame ionization section. Studies confirmed that the use of SAW/FIS detection would be an effective and sensitive alternative to flame ionization for our leak detection scheme. Optimization and development studies with SAW/FIS detection continued with the viewpoint that this technology was to be the primary mode of isopropanol detection. The SAW/FIS device offered high sensitivity, used few consumables, and posed no safety concerns when operated in the potential presence of flammable solvents. Another advantage of the Microsensor system is that the design of the instrument allows for sample collection independent of the GC analysis. The air sample is first absorbed onto a Tenax bed. After sampling, this column is heated to release the collected product solvent and routed onto the GC column for further separation and analysis. This concentrates the product solvent and thus extends the dynamic range of the instrument. OPTIMIZATION OF THE CHAMBER ANALYSIS SCHEME Optimization of Vacuum Chamber Operation It was found during the early experiments that the solvent response for a pinhole leak in a product tube was related to the amount of time the tube was under vacuum. Conversely, as a process analytical tool, the efficiency of the method is inversely proportional to the analysis time. Thus, the goal was to find a maximum signal in a short residence time. Tubes were placed in the vacuum chamber and held at 25 inches of Hg for 10 minutes because this was found to be the amount of time needed for a pin hole leak to yield a significant signal above background. Because the laboratory system was used solely for a feasibility investigation, this residence time was not thoroughly optimized. The lab design was from the start known to be a precursor to a chamber that could hold a larger volume of product containers. A more intense look at the residence time was done with the product scale chamber.
JPACSM 15 Container Studies Detection of Pinhole Leaks Flame Ionization Detection The flame ionization detector was challenged with several analyses. For each analysis, a product container was pierced with a 5 µl syringe needle. This procedure combined two worst case scenarios: 1. The small volume container representative of the lowest fill volume per tube for this product, and; 2. The needle lancing is representative of the smallest hole expected to be created in the sealing process. In addition, the container was oriented to prevent product contact with the pin hole leak to further challenge the system. The representative container was placed in vacuum chamber and vacuum applied at 25 inches of Hg for 10 minutes. After 10 minutes the chamber was reequilibrated to ambient pressure and sampled with a 500 µl gas tight syringe and run on a gas chromatograph. Gas chromatography investigations produced strong signals for the solvent for each 500 µl sample presented. This clearly indicated that flame ionization detection possessed the necessary sensitivity to analyze the solvent present in the chamber from a pin hole leak (Figure 1). Figure 1. FID chromatogram of a simulated pin-hole leak. SAW/FIS Detection The Model 301 vapor sensor instrument is configured to sample an air stream, not an injection port as the flame ionization detection system was. As an initial investigation to determine if the SAW/FIS system could detect the product solvent as effectively as the flame ionization detector, a 2 L tedlar bag was filled with ambient air and spiked with 1 ml of saturated product solvent vapor. For the flame ionization run, a 500 µl sample from this bag was analyzed. For the SAW/FIS investigation, the tedlar bag was sampled directly. The flame ionization detector response was 119 counts, which was in the same magnitude range as could be expected for a pinhole leak produced from the chamber system. The SAW/FIS detector response was 23,650 counts. The counts from each system are not directly comparable or equivalent. The conclusion for this experiment is simply that, like the flame ionization detector, the SAW/FIS system possesses adequate sensitivity to detect pinhole leaks in product tubes (Figure 2). Figure 2. SAW/FIS Detector chromatogram of a simulated pin hole leak. SAW/FIS detection was directly challenged using the chamber scheme to confirm that the arrangement used in the SAW/FIS detector was as effective as the flame ionization detector was for detecting product solvent. Three different tubes were pierced using a 5 µl syringe and individually processed by placing the tube in a 1 L flask, held under 25 inches of Hg vacuum for 10 minutes, and a 500 µl sample of the one cubic foot chamber taken upon re-equilibration. For the flame ionization analysis, the 500 µl aliquot was injected directly into the gas chromatography system. For the SAW/FIS device, the 500 µl aliquot was inserted into a 6-inch piece of ⅛ inch copper tubing connected to injection port with a Swaglok fitting. As illustrated in Table III, the actual detector response, in counts, cannot be directly compared between the SAW/FIS and flame ionization response only their calibrated results against external standards. The correct conclusion for reviewing the data in this table is that the SAW/FIS detector again proved to be a viable alternate to the flame ionization detector in that it also demonstrated acceptable sensitivity for product solvent.
JPACSM 16 Table III Direct SAW/FIS Versus Flame Ionization Detection Product Tube Flame Ionization Detector Counts SAW/FIS Detector Counts Medium Range 45,532 158,737 Large Range 34,901 134,101 Small Range 119,617 172,326 Plant Feasibility Studies The use of a volatile organic carbon detection scheme with the chamber system posed two concerns in regards to signal integrity: 1. Whether there was a significant matrix background in the production pail during manufacturing to mask the response from a single leaking tube, and; 2. Whether any of the numerous sources of plasticizers interfere with the response of the matrix. To investigate these concerns, two pails were sampled using a 500 µl gastight syringe directly on-line during product manufacturing. The syringe was oriented to sample beneath the first one or two layers of product. The two syringes were analyzed using the flame ionization detector conditions and had significant solvent responses. The presence of background product matrix was confirmed and must be addressed in the chamber testing scheme. For the chamber system to be successful, any residual solvent in the pail from manufacturing must be evaporated prior to testing for suspect seals. In order to determine the feasibility of incorporating a clean out procedure into the chamber scheme, a product tube was dipped in solvent and placed inside the vacuum chamber. After holding the chamber at 25 inches of Hg for 10 minutes, after returning to ambient pressure the chamber sample yielded a signal response of 59,105 counts. The dipping procedure was repeated, except that this time the chamber was purged through the needle valve port for 10 minutes prior to being held at vacuum. The gas chromatography response for this sample was 595 counts; approximately 99% of the background was removed. Purging the chamber for an additional 10 minutes removed all remaining traces of solvent. To verify that a leak could be determined above background, a small product tube was pierced and placed in a 1 L beaker with 120 additional sealed product tubes. The chamber was purged for five minutes through the needle valve port and held at 25 inches of Hg for 10 minutes. After returning to ambient pressure, the system yielded a signal response of 954 counts. Repeating this analysis with 121 sealed product tubes yielded a signal of 388 counts. Thus, the leak was responsible for approximately 566 counts above a background signal. While not intending to be truly quantitative at this point, the investigation confirmed that we could have confidence that a leaking tube could be detected above the pail background in an optimized chamber system. Small peaks were eluted prior to the product solvent signal on these chromatograms. These peaks were attributed to background and possibly from the plastic used in the linings and pails. However, because these peaks were well resolved from the product solvent, the effect of this background upon the chamber test scheme was determined to be insignificant. SAW/FIS Detector Optimization The SAW/FIS detector was investigated for acceptable response over the intended range for leak detection applications. The unit exhibited acceptable response in the lower portion of the curve. In fact, the correlation coefficient for the first three standards was 0.9940. This data provided assurance the detector was operating in a manner sufficient for detecting leaking solvent at or near the preliminary 50 ppm cut-off limit for product solvent response. Selectivity of Product Solvent Peak To optimize the gas chromatography column used with the SAW/FIS detector, the column temperature, mobile phase, and flow conditions of the SAW/FIS detector were mimicked on the thermal conductivity detector of the gas chromatograph. The use of this detector was simply due to its compatibility with packed column conditions. Selectivity was evaluated by comparing the chromatograms of nitrogen as well as alternative solvent vapor versus air. The resulting data demonstrated that none of these compounds causes a significant positive or negative bias to solvent determination or selectivity. Safety Analysis The vacuum chamber system was designed to operate inside the manufacturing filling area for solvent formulations since the site safety group was greatly concerned about the potential danger of having a flame-based detector operating within an environment containing volatile solvents. While it is debatable whether this concern was actually valid, this issue was a significant hurdle for this project. Fortunately, the SAW/FIS-based detection scheme eliminated such any potential hazard with volatile solvents. Because it was operating as effectively as the flame ionization detector without the associated safety concerns, the SAW/FIS detector quickly became the detector of choice for matrix detection.
JPACSM 17 CONCLUSION The use of a surface acoustic wave (SAW/FIS) detector in concert with vacuum chamber testing was fully characterized. It was shown to be the most feasible system that can operate inside the manufacturing work area. This conclusion stems from two points: 1. Concerns regarding the use of a flame ionization detector (FID) in the area due to volatile solvents, and; 2. The SAW/FIS unit has significantly fewer consumable parts to worry about and is inexpensive to operate. The developmental investigations supported the goal of having the vacuum chamber system as a diagnostic tool for adequately inspecting the integrity of product containers without the need for individual inspection. ACKNOWLEGMENT The authors would like to thank Norman O. Davis of MicroSensor Systems (Bowling Green, KY) for his assistance with this project. REFERENCES 1. M. Larson, Quality, 38, 40-43, (1999). 2. H. Heine, Foundry Management and Technology, 126, 32-34, (1998). 3. R.C. McMaster, Nondestructive Testing Handbook: Volume 1, Leak Testing. (review), Pulp and Paper, 58, pp. 187. 4. V. Comello, R&D, Vol. 39, 57, (1997). 5. V.D.Cahill, Foundry Management and Technology, 129, 24, (2001). 6. J. Hoffman, Plastics Technology, 43, 62-64 (1997). 7. J.P. Deluca, R&D, 40, 18, (1998). 8. J.M. Berklan, Healthcare Purchasing News, 24, 23, (2000). 9. L.M. Sherman, Plastics Technology, 42, 21-22, (1996). 10. H. Wohltjen and D.S. Ballantine, Analytical Chemistry, 61, 704A. 11. H. Wohltjen, D.S. Ballantine, and N.L. Jarvis, "Vapor Detection with Surface Acoustic Wave Microsensors", Chemical Sensors and Microinstrumentation, American Chemical Society, 157-175, (1989). 12. H. Wohltjen, N.L. Jarvis, and J. Lint, "A New Approach for On-Site Monitoring of Organic Vapors at Low PPB Levels", Proc. 2nd International Symposium, Field Screening Methods for Hazardous Wastes and Toxic Chemicals, (1991). 13. H. Wohltjen, A. Snow, and D.S. Ballantine, "The Selective Detection of Vapors Using Surface Acoustic Wave Devices", Proceedings of the International Conference on Solid-State Sensors and Actuators Transducers, 66-70, (1985).