Reduced Leak Calibration Uncertainties by the Outgassing Quantification Method

Transcription

1 University of Tennessee, Knoxville Trace: Tennessee Research and Creative Exchange Masters Theses Graduate School Reduced Leak Calibration Uncertainties by the Outgassing Quantification Method Philip Mason Carper Recommended Citation Carper, Philip Mason, "Reduced Leak Calibration Uncertainties by the Outgassing Quantification Method. " Master's Thesis, University of Tennessee, This Thesis is brought to you for free and open access by the Graduate School at Trace: Tennessee Research and Creative Exchange. It has been accepted for inclusion in Masters Theses by an authorized administrator of Trace: Tennessee Research and Creative Exchange. For more information, please contact

2 To the Graduate Council: I am submitting herewith a thesis written by Philip Mason Carper entitled "Reduced Leak Calibration Uncertainties by the Outgassing Quantification Method." I have examined the final electronic copy of this thesis for form and content and recommend that it be accepted in partial fulfillment of the requirements for the degree of Master of Science, with a major in Mechanical Engineering. We have read this thesis and recommend its acceptance: Spivey S. Douglass, H. Lee Martin (Original signatures are on file with official student records.) Madhu S. Madhukar, Major Professor Accepted for the Council: Dixie L. Thompson Vice Provost and Dean of the Graduate School

3 Reduced Leak Calibration Uncertainties by the Outgassing Quantification Method A Thesis Presented for the Master of Science Degree The University of Tennessee, Knoxville Philip Mason Carper May 2012

4 Copyright 2012 by Philip Mason Carper All rights reserved. ii

5 Abstract Standard leaks are used throughout industry for various applications, such as gas transfer, calibration of mass spectrometers, and calibration of leak detectors. The ability to provide a controlled delivery of gases at relatively low flow rates makes standard leaks a popular choice for calibration standards. This ability also allows for precise quantities of gas to accumulate over time. The accumulation of gas from standard leaks is used as a transfer standard for making in-process adjustments to measurement instrumentation. With the advent of tighter quality control constraints, the quantification and control of leak rates is becoming an increasingly important matter, placing increased demands on the understanding of leak behavior and on the accurate measurement of leakage rates. Because of stringent quality constraints imposed by industry and technological advancements, it is essential to develop a reliable and accurate method of calibrating standard leaks. The purpose of this research was to design, implement, and validate a leak calibration stand prototype that can detect and reduce systematic errors that lead to high calibration uncertainties of standard leaks. The main contributors to high calibration uncertainty were caused by internal leaks and the outgassing from the internal surfaces, fittings, and valves within the vacuum chamber. In order to quantify the rate at which the internal surfaces of the vacuum chamber outgassed, a custom vacuum chamber was constructed inside an oven to test the effects of outgassing at various pumpdown intervals and baking temperatures. Statistical methods were used to obtain a better insight on the selection of materials, pumpdown intervals, and baking temperatures needed to achieve low and repeatable outgassing rates. This research concluded that the quantification of leakage and outgassing will completely eliminate the once unknown uncertainty component of pressure measurements in vacuum. This discovery led to the realization that an additional isolated volume can be used to capture the accumulation of leakage and outgassing and to correct for pressure measurements during standard leak calibrations.

6 ii Chapter Table of Contents Page Chapter 1 Introduction Standard Leaks Types of Leaks The Vacuum Environment Outgassing Outgassing Rate Calibration Methods Current Calibration Method Development of a New Leak Calibration Stand... 7 Chapter 2 Literature Review Overview Vacuum Technology Vacuum System Design Vacuum Pumps Roughing Pumps Gas Flow Pumping Speed Vacuum Components Vacuum Flanges Valves Vacuum Gauges Barometers Capacitance Diaphragm Gauge Spinning Rotor Gauge Chapter 3 Methodology Overview Setup and Design Design Implementation Bakeout Temperature Determination Results Leakage and Outgassing Repeatability Study Results Leakage and Outgassing Predictability Study and Steady Outgassing Rate Study Leakage and Outgassing Predictability Study Steady Outgassing Rate Study Outgassing Quantification Study and Leak Calibration Testing Outgassing Quantification Study Results Leak Calibration Testing Development of the Prototype Chapter 4 Conclusion and Recommendations Conclusion Recommendations... 62

7 List of References Appendix A Appendix B Appendix C Appendix D Appendix D.1 Outgassing Predictability Study 1 st Order Polynomial R 2 Results Appendix D.2 Outgassing Predictability Study 2 nd Order Polynomial R 2 Results Appendix D.3 Outgassing Predictability Study 3 rd Order Polynomial R 2 Results Appendix E Appendix F Appendix F.1 Mounting Bracket: Assembly Exploded View Appendix F.2 Mounting Bracket: MB1 1/4 Laser Cut Nested Layout Appendix F.3 Mounting Bracket: MB1B 1/2 Laser Cut Nested Layout Appendix F.4 Mounting Bracket: MB1DN 1/4 Laser Cut Nested Layout Appendix F.5 Mounting Bracket: MB1 Plate A01 Machining Appendix F.6 Mounting Bracket: MB1 Plate B01 Machining Appendix F.7 Mounting Bracket: MB1 Plate B02 Machining Appendix F.8 Mounting Bracket: MB1 Plate B03 Machining Appendix F.9 Mounting Bracket: MB1 Plate B04 Machining Appendix F.10 Mounting Bracket: MB1B Assembly Weld Locations Appendix F.11 Mounting Bracket: MB1 Plate C01 Machining Appendix F.12 Mounting Bracket: MB1 Plate C02 Machining Appendix F.13 Mounting Bracket: MB1C Assembly Weld Locations Appendix F.14 Mounting Bracket: MB1 Plate D01 Machining Appendix F.15 Mounting Bracket: MB1D Assembly Weld Locations Appendix G Appendix G.1 Valve Layout: Parts Appendix G.2 Valve Layout: ORMC Provided Parts Appendix G.3 Valve Layout: Parts Model Appendix G.4 Valve Layout: Multiview Top Appendix G.5 Valve Layout: Multiview Left Hand Side Appendix G.6 Valve Layout: Multiview Front Vita iii

8 Figure List of Figures iv Page 1.1 Major Leak Rates for Common Leak Elements and Gases Leak Rates for Common Leak Elements and Gases Leakage and Outgassing in a Vacuum Chamber ORMC V P Method Uncertainties by Leak Rate Overview of Vacuum Pumps by Classification Gas Flow Regimes at Low Pressures Nitrogen Flow Regimes at Various Pressures and Pipe Diameters Hydrogen Flow Regimes at Various Pressures and Pipe Diameters ConFlat Flange Assembly KF Flange Assembly KF Metal Knife Edge Assembly VCR Fitting Assembly Measurement Ranges and Uncertainties of Typical Vacuum Gauges Experiment Design 1: SRG Ball/Flange Assembly and Measuring Head Experiment Design 1: CDG Porthole Experiment Design 1: Schematic Layout Experiment 1: SRG Pressure Stability Check Experiment 1: Bakeout Temperature Determination Study (ΔP vs. Time) Experiment 1: Bakeout Time and Temperature Relationship Contour Plot Experiment 1: Leakage and Outgassing vs. Time Experiment 2: Repeatability Study Results Average Error (ΔP vs. Time) Experiment 2: Leakage and Outgassing vs. Time st, 2nd, and 3rd Order Polynomial CDF Plots Predicted Intervals vs. Average R2 One-way Analysis Data Collection Start Time vs. Average R2 One-way Analysis Treatment Interactions Pareto Plot Change in Slope over 10 Minute Intervals Experiment Design 2: Design 1 Modified to Include Two Volumes Experiment Design 2: Schematic Layout Experiment Design 2: Test 1, Valve 1 Expansion (ΔP vs. Time) Experiment Design 2: Test 2, Valve 1 Expansion (ΔP vs. Time) Experiment Design 2: Test 1, Valve 2 Expansion (ΔP vs. Time) Experiment Design 2: Test 2, Valve 2 Expansion (ΔP vs. Time) Initial Leak Calibration Testing Design Experiment Design 3: Schematic Layout Experiment Design 3: Pretesting LO Repeatability Test Experiment Design 3: VOL 2 and VOL 1 Predictions Experiment Design 3: M 2 Predictions Experiment Design 3: Correlated M 1 +M 2 +VOL 1 +VOL 2 Projection Experiment Design 3: Standard Leak Corrected for LO Prototype Design: Evacuation Manifold Connections (Top View) Prototype Design: Multiview Prototype Design: Conceptual Rendering of the Leak Calibration Stand... 62

9 1 Chapter 1 Introduction 1.1 Standard Leaks Standard leaks are used throughout industry for various applications, such as gas transfer, calibration of mass spectrometers, and calibration of leak detectors. The ability to provide controlled delivery of gases at relatively low flow rates makes standard leaks a popular choice for calibration standards. This ability also allows for precise quantities of gas to accumulate over time. The accumulation of gas from standard leaks is used as a transfer standard for making in-process adjustments to measurement instrumentation. With the advent of tighter quality control constraints, the quantification and control of leak rates is becoming an increasingly important matter, placing increased demands on the understanding of leak behavior and on the accurate measurement of leakage rates [1]. Many available calibration techniques, in respect to overall cost, accuracy, and throughput, do not adequately reduce calibration uncertainties. Therefore, improvement to existing calibration methods or development of an alternative method is essential Types of Leaks Standard leaks, sometimes referred to as Leak Artifacts, Test Leaks, or Leak Rates, consists of three components; the leak element, a gas supply, and the gas outlet. A schematic diagram illustrating each of these components is shown in Figure 1.1. Figure 1.1 Major Components of a Standard Leak.

10 2 The gas supply or leak reservoir can take many forms. The simplest form is literally nothing, where atmosphere is allowed to leak through the leak element into the vacuum system [1]. The most common reservoirs used in calibrated leak applications are fixed volume reservoirs. This type of reservoir is often made of glass or metal and can be backfilled with high purity gases or mixtures at very high pressure to sub atmospheric pressures. The leak element is comprised of anything that limits the flow of gas from the reservoir side to the outlet side in a reproducible manner. Typically, the leak element consists of a stainless steel crimped capillary or a glass permeation membrane that will restrict the flow rate of gas from 1x10-1 atm cc/sec to 9.9x10-13 atm cc/sec. Common leak elements and the leak rates in which they are used are displayed in Figure 1.2 [2][3][4][5] (J. Alfrey, personal communication, January 5, 2012). The last component, the gas outlet, is a fitting or means for connecting the standard leak to the process in which it is used. Figure 1.2 Leak Rates for Common Leak Elements and Gases. 1.2 The Vacuum Environment Vacuum is a quantitative term used to describe the amount of matter that has been removed from a space and is inversely related to the quantitative term, pressure, which is a measure of force per unit area [6][7]. Since matter cannot be removed completely from a space, the phrase vacuum measurement is not possible because it is not a quantitative term, and instead pressure must be measured directly or indirectly [6]. For convenience, vacuum has been characterized into different vacuum regimes based on pressure values.

11 3 The different vacuum regimes are as follows; low vacuum [750 Torr to 25 Torr], medium vacuum [25 Torr to 7.5x10-3 Torr], high vacuum (HV) [7.5x10-3 Torr to 7.5x10-6 Torr], very high vacuum [7.5x10-6 Torr to 7.5x10-9 Torr], ultrahigh vacuum (UHV) [7.5x10-9 Torr to 7.5x10-11 Torr], and extreme high vacuum (XHV)[> 7.5x10-11 Torr]. A vacuum is created by a combination of pumps, valves, and pipes, producing a region of low pressure. Removal of air at atmospheric pressure is usually achieved by displacement pumps such as rotary vane and piston pumps whereas higher vacuum can be achieved by highly efficient turbomolecular pumps and ion pumps and combination thereof. Whether a direct or an indirect measurement takes place determines what device is used to measure pressure under a vacuum. A direct vacuum pressure measurement will produce a new measurement signal when the pressure being measured changes, and an indirect pressure measurement gauge monitors some other parameter of the system that proportionally responds as the pressure of the system changes [8]. The vacuum environment is a near perfect setting for the many applications in which standard leaks are used. At reduced pressures, molecules are less abundant and the mean free path or average distance the constituent molecule travels before colliding with other molecules is reduced proportionally with the system pressure. Increasing the mean free path is a key enabler for applications where standard leaks are calibrated particularly when changes in pressure and the composition of gases in the system are of interest. Most leak calibration techniques use either a direct measurement or an indirect measurement type gauge depending on the degree of vacuum in which the measurement takes place. Direct measurement devices such as the capacitance diaphragm gauge (CDG) are commonly used for leak measurements at medium and high vacuum because changes in gas composition do not affect the accuracy of the pressure readings. While indirect measurement devices such as the spinning rotor gauge (SRG) are used to measure changes in pressure at vacuum levels below the operating range of the CDG. The SRG is the most accurate gauge available today for measuring pressure between 1x10-2 Torr and 5x10-7 Torr, and like most indirect gauges, readings must be corrected to the gaseous species being measured [9].

12 4 1.3 Outgassing Outgassing is common to all vacuum chambers and is caused by the release of hydrogen that has been dissolved, absorbed, or trapped from the internal surfaces. Stainless steel and aluminum are commonly used in the fabrication of vacuum chambers because of their corrosion-resistant and low outgassing properties. Although at UHV, outgassing is not avoidable, and its affect is undesirable when calibrating standard leaks because it results in a condition of an initial pressure rise that yields a calculated value of leakage rate that is greater than the true value. When internal surfaces have been baked or cleaned prior to calibration, an initial influx of gas from the standard leak may absorb into the internal surfaces resulting in a condition of initial pressure rise that yields a calculated value less than the true value [1]. Pressure rise from system leaks and outgassing is schematically shown in Figure 1.3. Figure 1.3 Leakage and Outgassing in a Vacuum Chamber [10] Outgassing Rate The outgassing rate is the measurement of gas that has evolved from the interior surface per unit time per unit of exposed geometric surface area [11]. There are two general methods for measuring the outgassing rate in a vacuum chamber: the gas accumulation method and throughput method. The gas accumulation method measures the rise in pressure after a section of the system has been isolated at a lower pressure. The

13 5 initial rise in pressure caused by the leakage and outgassing often appears to be linear, but after time, the outgassing rate of the surfaces tends to reach a constant level and the slope of the line will become less positive and eventually stabilize at a linear rate. The throughput method makes use of two chambers, the chamber under test and the pumping chamber, which are separated by an orifice of known conductance. The outgassing rate of the chamber under test produces a drop in pressure while being pumped through the orifice. A differential pressure gauge is used to measure the drop in pressure across each chamber as a function of time. While both the gas accumulation and throughput methods are generally used for measuring outgassing, both setups are used in primary standard leak calibration systems. 1.4 Calibration Methods There are several techniques available for calibrating standard leaks. Most techniques for calibrating standard leaks use HV and very high vacuum chambers to accurately measure the rise in pressure of only the gaseous species of interest. The primary method for calibrating leakage rates depends on the measurements of pressure, volume, temperature, time, and the quantities of mass depleting from the standard leak. Several variations of the primary method exist. The pressure rise method, or method, is the simplest variation because it is the least costly in respect to the low uncertainties that are achievable when calibrating leak rates between 1x10-3 atm cc/sec to 9.9x10-6 atm cc/sec. However, since the method yields higher uncertainties when calibrating leak rates between 1x10-7 atm cc/sec to 9.9x10-10 atm cc/sec, special care must be taken to adequately quantify leaks in the system and outgassing from the internal surfaces in which the pressure rise is being measured. Calibration methods such as the accumulate dump and the direct comparison techniques can overcome the effects of outgassing by measuring only the elemental composition of interest. Both of these methods use mass spectrometers and high accuracy standard leaks by directly comparing the different leakage rates between the known standard and the Device Under Test (DUT). Although low uncertainties are achievable when either method is used, the repeatability and accuracy of the measurements taken

14 6 from the mass spectrometer can add additional uncertainties, particularly when calibrating standard leaks containing the same molecular content that is being outgassed by the system. 1.5 Current Calibration Method The Oak Ridge Metrology Center (ORMC) at Y-12 is currently using the method to accurately calibrate faster leaks in the range of 1x10-3 to 9.9x10-6 atm cc/sec. As ageing systems are being replaced with newer and more accurate instrumentation, there has been an increased demand for the calibration of slower leak rates in the range of 1x10-7 to 9.9x10-10 atm cc/sec with the same uncertainties as the faster leaks. At slower leak rates, the method produces uncertainties that become exponentially higher as the DUT approaches the same leak rate as the outgassing and internal leak rate of the system, as shown in Figure 1.4. At slower leak rates, the time required to fill a calibrated volume with the gas accumulated from the leak is exponentially longer than at faster leak rates. Longer test times lead to additional uncertainties in the leak rate measurement due to changes in ambient temperature, increased outgassing accumulation, and increased internal leakage accumulation. Uncertainties from 14% to 35% are typical for leak rates from 1x10-7 to 9.9x10-10 atm cc/sec. In respect to overall cost, accuracy, and turn-around time, the customer s needs are not being for the slower leak rates using the current calibration method.

15 7 Figure 1.4 ORMC V P Method Uncertainties by Leak Rate. 1.6 Development of a New Leak Calibration Stand Stringent quality constraints imposed by the customer necessitate a reliable and accurate method of calibrating standard leaks; rapid technological advancements have made it possible to develop such a method. The ORMC has allocated approximately 200,000 dollars for the design, implementation, and validation of a leak calibration stand prototype that will adequately detect and reduce systematic errors that lead to high calibration uncertainties of standard leaks. The main contributors to high calibration uncertainty were determined to be caused by internal leaks and outgassing from the internal surfaces, fittings, and valves within the vacuum chamber. In order to quantify the rate at which the internal surfaces of the vacuum chamber outgassed, a custom oven vacuum chamber was constructed to test the effects of outgassing at various pumpdown intervals and baking temperatures. Statistical methods were used to obtain a better insight on the selection of materials, pumpdown intervals, and baking temperatures needed to achieve repeatable outgassing rates. Based on the results of the experiments, modifications were made to the experimental setup to further reduce uncertainties of the

16 8 measurement instrumentation so that the effects of outgassing can be reduced even further. The prototype will take advantage of many of the lower uncertainty aspects from the techniques mentioned in Section 1.4 as well as the implementation of the Rate-of- Rise methodology. The Rate-of-Rise method measures and corrects for the nonlinear pressure readings caused by volume displacements from actuators, measurement errors from the instrumentation, and outgassing from internal surfaces. The prototype will also present a new concept that will allow for outgassing and systematic leaks to be measured during the calibration without introducing additional uncertainties. The concept will be referred to as the outgassing quantification method, because the prototype will make use of an additional volume, independent of the leak measurement volume, for quantifying the pressure accumulation caused by systematic leaks and outgassing. The outgassing quantification method, accuracies, and implementation challenges will be discussed in Chapter 3. Schematics of the prototype design and CAD model are shown in Appendix A and B, respectively.

17 9 Chapter 2 Literature Review 2.1 Overview Vacuum technology has led to the creation of a wide variety of vacuum components and equipment that have found extensive use in standard leak calibration applications where leakages and outgassing must be reduced. There is not an abundant amount of literature available on the components, equipment, and design concepts used for performing leak calibrations. Therefore, the literature review will be focused on the aspects of vacuum technology that pertain to this research. 2.2 Vacuum Technology Vacuum technology, sometimes referred to as vacuum science, is the application of sub-atmospheric pressures for performing physical measurements. Vacuum pumps, vacuum chambers, valves, fittings, and piping are used to remove enough molecules within a confined space to achieve a continuous pressure below atmospheric pressure. This application is fundamental to a range of scientific explorations and processes, from electron accelerators at XHV to production of incandescent light bulbs at low vacuum. Scientific research is commonly performed under vacuum to remove gases in the atmosphere that could produce a physical or chemical reaction; disturb an equilibrium condition existing at normal conditions; increase the distance molecules can travel before colliding (mean free path); or reduce the rate of molecular collision to lower the possibility of surface contamination [12]. 2.3 Vacuum System Design There are several factors that need to be considered when deciding which vacuum technology to use for a particular application. The goal of the process or the end product will determine the design of the system. Temperature requirements can significantly change the makeup and layout of a system, requiring modifications to accommodate ovens or other types of thermal heating elements. At higher vacuum levels, baking is

18 10 often used to remove moisture that has developed in the system. For processes that require the addition or removal of components while under vacuum, proper care must be taken to assure that pumps are not overworked by the abrupt exposure or cycling to atmospheric pressure. Another important aspect is the selection of proper constituents. In terms of complexity, a vacuum system is far greater than the sum of its parts [13]. Because all parts of a vacuum system interact with one another and each interaction is different, small componential errors can be more detrimental than a single big mistake. Therefore, it is imperative that the correct pumps, flanges, valves, and materials are chosen in order to achieve the desired pressure. Lastly, the monetary budget for the vacuum system must to be taken into account. The financial resources available will affect the entire design process, and limited funds can pose the greatest obstacle to overcome. Designing a vacuum system is a demanding process. Hundreds of decisions must be made, and one poor decision can hinder the system s overall performance. The level of effort that has been put into the design is directly proportional to the level of performance, which determines whether or not the goals of the vacuum system are achieved. The required vacuum level for the particular application is usually used for measuring the level of performance, since the system must pump down to this level for the process to run. In other words, anything less than the best possible performance is regarded as a failure [14]. A vacuum system is constructed of pumps, flanges, piping, and valves that can withstand forces of atmospheric pressure. Choosing appropriate materials is challenging since they vary depending on the application of use. Additionally, construction materials must provide as little gas load as possible, which is problematic due to the difficulty of assessing gas loads [13]. Provided in the next few sections is an in-depth discussion of pumps, flanges, valves, and other materials used in a vacuum system. 2.4 Vacuum Pumps Vacuum Pumps are used to generate vacuum by removing gas from a chamber that is isolated from the atmosphere. How they operate differs according to the

19 11 application and the level of vacuum in which they are used. Vacuum pumps are classified into two main types based on their general usage, either for gas transfer or for gas binding. Gas transfer pumps employ a mechanism that displaces or accelerates molecules from higher pressure to lower presser. Depending on the principle method of operation, gas transfer pump can be divided into two subclasses: positive displacement pumps and kinetic pumps. In positive displacement pumps, gas from sealed areas is displaced to the atmosphere or to a subsequent pump stage. In kinetic pumps, a mechanical drive system or an aligned vapor stream condensed at the end of the pumping section accelerates gas in the pumping direction [15]. Gas binding vacuum pumps operate by binding or capturing molecules in a solid or absorbed state. Figure 2.1 shows an overview of the different types of vacuum pumps by classification [15]. Although there are many types of vacuum pumps, most are impractical for leak calibrations. Figure 2.1 Overview of Vacuum Pumps by Classification.

20 Roughing Pumps Deciding which high vacuum pump is best-suited for a given application is usually time-consuming, requiring careful consideration of the many trade-offs between financial and quality costs. Following the selection of a high vacuum pump is that of the complementary roughing or backing pump, a device that performs two distinct operations, hence its two-part name [12]. First, the roughing pump allows the high vacuum pump to operate by rough pumping the chamber from atmospheric pressure to a pressure that is low enough. If a roughing pump was not used, a high vacuum would never reach its maximum performance Gas Flow The behavior of gas flow is an important aspect of vacuum systems. During the production of a vacuum, gas moves from the vacuum vessel to the atmosphere by flowing through pipelines and pumps. Gas flow in vacuum systems can typically be categorized into three main regimes: viscous flow, transitional flow, and molecular flow (See Figure 2.2).

21 13 Figure 2.2 Gas Flow Regimes at Low Pressures. The viscous flow regime occurs within a low vacuum setting, where the mean free path of the gas molecules are significantly smaller than the dimensions of the mechanism. This causes the gas to behave like a liquid, i.e., flow in a fluid manner. As vacuum increases, the gas moves through the transition flow regime. Transition flow is a highly complex condition occurring between viscous and molecular flow regimes, where the mean free path is just barely greater than the mechanism s dimensions. The gas flow under this condition has both viscous and molecular flow characteristics. Finally, when the gas is at high vacuum, it is in molecular flow. Under this condition, the mean free path exceeds the dimensions of the mechanism, and molecules behave independently. The gas regime can be determined by the nature of the gas and relative quantity of gas by calculating the Knudsen s number. The ratio between the average distance that a molecule can travel before colliding with another molecule (mean free path) and the diameter of the tube is known as the Knudsen number. The Knudsen number,, can

22 determine the various regimes of gas flow, viscous flow, transitional flow, or molecular flow by the equation: 14 { (2.1) where is the mean free path ( ) and is the diameter of the pipe ( ). The mean free path,, can be solved by the equation: (2.2) where is the Boltzmann constant ( ), is the temperature ( ), is the molecular diameter of the molecule ( ), and is the pressure ( ). For Example, Nitrogen ( ) at flowing through a tube ( ) at, the molecular flow regime in the tube can be determined by the mean free path,, and Knudsen number,. First, solving for the mean free path, ( ) ( ) ( ) ( ) [ ] [ ] [ ] Now, solving for the Knudsen number, and because 0.2 lies between, the flow is in the transitional regime. Nitrogen was considered for this example because vacuum systems are often vented with pure rather than of air to prevent the chance of water vapor from depositing on the chamber walls. However, hydrogen is the most predominant residual gas in high vacuum largely because of metals outgassing in the system and due to its smaller size. Now, solving for the Knudsen number from the example above using hydrogen;

23 15 ( ) ( ) ( ) ( ) [ ] [ ] [ ] Solving for and because the flow is in the molecular regime. This example demonstrates how drastically flow regimes change when different gaseous species are introduced at the same pressure and pipe diameter. Figures 2.3 and 2.4 further illustrate how the flow regimes change at different pressures and pipe diameters for nitrogen and hydrogen, respectively. Figure 2.3 Nitrogen Flow Regimes at Various Pressures and Pipe Diameters.

24 16 Figure 2.4 Hydrogen Flow Regimes at Various Pressures and Pipe Diameters Pumping Speed Pumping speed is the measurement of a pump s ability to remove gas from a chamber. More specifically, it is equivalent to the volume of gas that the pump removes from the system at any given pressure. This can be calculated by taking the ratio of the throughput of a particular gas to its partial pressure at a given point near the inlet port of the pump. Pumping speed is generally measured as liters per second (L/s), cubic feet per minute (cfm), or cubic meters per hour (m 3 /hr). When the pumping pressure is at a steady state, the effective pumping speed can determine the total volume evacuated per unit time (L/s) by the equation (2.3) and the amount of mass going through the pump (Torr L/s),, can be solved by (2.4)

25 17 where is the Pressure (Torr) and is the change in volume per unit time (L/s). The conductance of a tube describes its ability to convey gas per unit time. The shape, length, surface roughness, and the flow regime all affect the amount of conductance of a tube. Generally, conductance for in a tube can be solved by the equation ( ) (2.5) where, is the gas flow, is the input pressure of the tube, is the output pressure of the tube, and is the conductance. In the viscous flow regime, conductance of air flowing through a long tube with a uniform circular cross section is proportional to the average pressure entering and exiting the tube. The conductance can be calculated by (2.6) where is the diameter of the tube ( ), is the dynamic viscosity of the gas in poise, is the length of the tube ( ), is the inlet pressure ( ), and is the exit pressure ( ). In the molecular flow regime, conductance (liter/sec) through a long tube with a uniform circular cross section is independent of pressure and can be calculated by (2.7) where is the diameter of the tube (cm), is the length of the tube (cm), T is the temperature, and M is the atomic mass unit (amu). In the transitional flow regime, conductance (liter/sec) through a long tube with a uniform circular cross section can be calculated by adding conductance in viscous flow,, and conductance in molecular flow,, by ( ) (2.8) Like electrical conductance, when multiple pipes in vacuum systems are in parallel, the conductance is added

26 18 (2.9) and when in series, the conductance is added in reciprocal form (2.10) and solving for where the new term,, is the pumping speed specified by the manufacturer in liters per (2.11) second. In the viscous flow regime, the time (seconds) required to pump the system down from an initial pressure,, to a target pressure,, for a system with no gas load is where is the volume in the chamber. In the molecular flow regime, the time (hours) required to lower the pressure in the chamber to the desired working pressure, obtained from (2.12), can be ( ) [ ( ) ( ) ] (2.13) where is the initial pressure (torr), is the pressure in the chamber after t(s), and is the ultimate pressure in the chamber pressure (torr). The ultimate pressure that the vacuum chamber can reach is determined by the relationship between the sum of gas loads and the effective pumping speed by (2.14) the total outgassing is then obtained from the equation: (2.15) where is the gas load due to leaks, is the gas load due to outgassing, and is the gas load due to vapor pressure.

27 Vacuum Components The construction of vacuum systems has transformed dramatically over a short time span. Thirty years ago, the process was complex and cumbersome, and its results were usually crude and unreliable. Since then, there have been major advancements in vacuum technology that have simplified and improved vacuum system design. In the last twenty years, the aspect of vacuum technology that has undergone the most significant transformation is system hardware development. In comparison to earlier vacuum systems, today s system hardware is more functionally efficient and easier to use [17]. Vacuum hardware is the assortment of flanges, piping, and valves used to build the system Vacuum Flanges Flanges are available in a variety of shapes, sizes, and materials to fit the needs of almost every vacuum application. They are commonly made of metals, elastomers polymers, and ceramics. The selection of flanges is governed by the process requirements and the base pressures of operation. Although there are many vacuum flanges available, only the ConFlat flange, KF flange, KF metal knife edge seal, and VCR fitting will be discussed in sections A, B, C, and D respectively.

28 20 A. ConFlat Flange Figure 2.5 ConFlat Flange Assembly. The ConFlat flange or CF flange for short was introduced in 1961 by Varian Associates to provide the growing vacuum industry with an all-metal valve for use in UHV applications, particularly those where high temperatures are used. The CF flange is preferred over rubber gasket and O-ring type connections for use at higher temperatures because they do not evolve large amounts of gas, which can increase the pressure or contaminate the vacuum system. However, after prolonged exposures to high temperatures, CF gaskets are prone to deformation resulting in leakage from atmosphere to enter the system. Unlike most sealing methods used in vacuum, the CF flange seal is created by circular knife edge face that penetrates into a metal gasket. The flanges are sexless with the knife edge slightly recessed to protect the seal in case of mishandling. They are held together with bolts feeding through one

29 21 flange and anchored to the opposing flange with nuts, thus providing uniform compression and penetration of the knife edge into the gasket. A typical CF flange assembly can be seen in Figure 2.5. The flanges are available in a fixed and rotatable configuration to correct bolt pattern alignments [18]. Overall, CF flanges offer superior performance for pressures between atmosphere (760 Torr) and 10-8 Torr with leak rates around 10-9 Torr. B. KF Flange Figure 2.6 KF Flange Assembly. The KF quick release flange is produced by a variety of manufacturers and is commonly used in vacuum systems where components will be frequently reconfigured assembled, or disassembled. Most KF flanges make use of elastomeric materials to provide the vacuum seal. The letters, KF, were derived from the German word Kleinflansch meaning small flange [19]. The flange is also known by Quick Flange (QF), Klein Flange (KF) or NW, and sometimes DN [20]. KF flanges have metric interface dimensions defined by the International Standards Organizations (ISO),

30 22 ensuring a high degree of compatibility between flanges produced by other manufacturers. The ISO/KF family of vacuum components is an economical choice for use in vacuum systems because of their reusable interfacing (O-rings) which is uncommon among other vacuum flanges. The flange seal is made by compressing an elastomer O- ring between the two flat sexless mating surfaces. Metal hinged clamps are used to grasp the KF flange s chamfered back surfaces and hold the assembly in place This provides adequate compression to achieve a vacuum seal. An optional O-ring retainer or centering ring is used to assure proper placement of the O-ring. Practical base pressures for KF assemblies are from 10-4 Torr to 10-6 Torr [17]. A typical KF flange assembly is shown in Figure 2.6. C. KF Metal Knife Edge Seal Figure 2.7 KF Metal Knife Edge Assembly. The KF metal knife edge seal makes use of metal seals rather than O-rings that are typically used in KF flange assemblies. The seals are made of aluminum rings that rest on the outside edge of the flange rather than the inside centering ring groove like centering O-rings. The inside flange on the aluminum seal is equipped with knife-edges, which

31 23 are actually rounded edges that are compressed between the two flat sexless mating surfaces. Chain clamps or ISO claw clamps are used to grasp the KF flange s chamfered back surfaces. This holds the assembly in place and provides adequate compression to achieve a vacuum seal. Because these metal seals are a relatively new vacuum technology, non-biased base pressures have not been published by a liable source. A typical KF metal knife edge assembly is shown in Figure 2.7. D. VCR Fitting Figure 2.8 VCR Fitting Assembly. The Vacuum Coupling Radius (VCR) fitting is produced by the Swagelok Company for use in UHV and up to 351 psig (1/4 Tubing Specifications) [21]. (D. Lewis, personal communication, March 9, 2012). The seal is made by compressing a gasket between two stainless steel flanges known as glands. The glands have ribs that extrude out from the

32 24 face that penetrate into the gasket [23]. The glands are held together by a male and female coupling nut, thus providing uniform compression to each side of the gasket. Like CF gaskets, VCR gaskets are single-use and must be replaced with a new gasket once the seal is broken. The VCR assembly can be baked at elevated temperatures but is limited by the gas material. Common gasket materials are stainless steel, nickel, and copper; the temperatures at which they are 537 C, 315 C, and 204 C respectively. A typical VCR assembly is shown in Figure Valves Valves are used to isolate, direct, redirect, and control the flow of gaseous molecules to the vacuum pump or to conceal the flow of gas from a prearranged source. They can be activated manually, pneumatically, electronically, and remotely by controllers and data acquisition type interfaces. Isolation type valves are intended to completely seal a port and typically are either open or closed. Most valves operate by a movable plate, plunger, or rotational mechanism to seal the flow of gas. Seals can be made by a combination of different materials but are typically comprised of polymers or metals. In contrast, control valves are not intended to completely isolate, but rather to control the rate at which gas is allowed to pass upstream or downstream in the system. However, because many control valves operate on the same operating principle as an isolation valve, they can be used to for isolating purposes, although this would be considered as a misapplication for use as such [22]. Control valves are mostly used for one of two purposes, if not both; to regulate the admission of gas and to control the removal rate of gas by the vacuum pump [22]. A complete list of valves that will be used on the calibration stand and their technical specifications can be found in Appendix C. 2.6 Vacuum Gauges Vacuum gauges are instruments for measuring pressure under a vacuum. To tolerate a wide range of conditions, many variations are available. The specific type of vacuum gauge that should be used differs according to the range of the vacuum being measured, the accuracy of the measurement, and the response rate of the measurement.

33 25 Vacuum pressure measurements are either direct, in which a new measurement signal is produced when the pressure being measured changes; or indirect, in which the gauge monitors some other parameter of the system that responds proportionally as the system s pressure changes [24]. The readings obtained from a direct measurement are independent of the gas type and the pressure that is to be measured, whereas those from an indirect measurement are dependent on the type of gas. Most vacuum gauges are able to measure pressures when referenced to atmospheric and absolute pressures. The measurement ranges (in respect to pressure) of typical vacuum gauges can be found in Figure 2.9. A detailed review of the barometer, capacitance diaphragm gauge, and spinning rotor gauge is provided in the following sections. Figure 2.9 Measurement Ranges and Uncertainties of Typical Vacuum Gauges.

34 Barometers The barometer is the standard measuring device used for pressure changes at atmosphere. The term barometer literally means an instrument that measures bars and so pressure that is measured from this this device is commonly referred to as barometric pressure [25]. Since pressures do not deviate greatly from the barometric pressure at sea level ( ) and because one bar is a relatively large unit, we refer to changes in atmospheric pressure in millibars. Barometric pressure is computed in respect to changes in elevation from sea level using the equation [ ( ) ] (2.16) where is static pressure ( ), is standard temperature (K), is standard temperature lapse rate (K/ft), is height above sea level (ft), is height at bottom of layer b (ft), is the universal gas constant ( g 0 s 2 /(mol K)), is the standard gravity ( ft/s 2 ), and is the Molar mass of earth's air ( kg/mol). The term manometer is often used to describe the variety of pressure measuring devices for measurements in the vacuum and low-pressure ranges. Special manometers used as primary pressure standards such as the quartz helix manometer are capable of reading accuracies within in (0.001 mm) but will not be discussed Capacitance Diaphragm Gauge The Capacitance Diaphragm Gauge (CDG) uses a sensor to measure changes in electrical capacitance that results when pressure is applied to the diaphragm. The capacitance is referenced to a fixed capacitance depending on the range that the device is made to measure. The higher the process pressure, the farther the diaphragm is pulled away from the capacitance plates. Differential CDGs use the same design except that the diaphragm is allowed to move, while other differential manometers utilize a Wheatstone bridge to ensure a balanced condition is met. The greatest advantage of the devices is their ability to detect extremely small diaphragm movements. The thinnest diaphragms can measure into the 10-5 Torr range while most capacitance manometers measure from

35 Torr on up to 1000 Torr. Accuracies of 0.3% to 1% can be achieved at a laboratory level, but when used in the field, accuracies can be much lower. Disadvantages of the gauge include sensitivity to vibration and high sensitivity to over-ranging Spinning Rotor Gauge The Spinning Rotors Gauge (SRG) uses a precession controller that creates a magnetic field in order to levitate a spinning ball. When mid to high vacuum is applied, the ball accelerates and decelerates producing a molecular friction that is measurable. A driving coil spins the ball until steady speeds around 425r/s are reached before the drive is turned off, causing a sudden decrease in speed. The controller measures the speed until 405r/s can be seen, then the driving coil accelerates the ball once again. The process pressure is measured by the time it takes for the speed to decrease from 425 r/s to 405 r/s. The pressure can be determined from the equation [26]; ( ( )) (2.17) where is the diameter of the ball, is the mass-density of the ball (kg/m 3 ), / is the ball s relative deceleration rate (Hz), deceleration rate vs. rotation frequency linear fit of the SRG offset characterization, is the rotation frequency of the ball (Hz), is the intercept of the deceleration rate vs. rotation frequency linear fit of the offset characterization (Hz), is the accommodation coefficient of the SRG (dimensionless), is Boltzmann s Constant ( J/K mol), is the temperature of the ball (Kelvin), and is the molecular weight of the gas being measured (g/mol). The coefficients (,,,, and ) can be provided by a national standards laboratory such as the National Institute of Standards and Technology (NIST). Several advantages of the SRG includes the ability to perform measurements in corrosive environments, resistance to high temperatures (up to 400 C), and high accuracy of ±0.5% to ±1.5%. Disadvantages include sensitivity to vibration and sensitivity to temperature when not monitored.

36 28 Chapter 3 Methodology 3.1 Overview The objectives of the research were to determine the key parameters that influence the outgassing rate of stainless steel; isolate and characterize the conditions that cause high variability; and develop a prototype that can adequately quantify outgassing and leakages in a vacuum chamber. Based on the results of the experiments, changes were made to the prototype to further reduce measurement uncertainties. Valves and ports were also added to fully simulate a leak calibration through the outgassing quantification method. This chapter contains six key sections. Section 3.2 provides an overview of the vacuum system design. Section 3.3 defines the experimental parameters that were used to determine the length of bakeout, the system design, and results. Section 3.4 describes the leakage and outgassing repeatability study and the results. Section 3.5 outlines the predictability study and a method for determining where the outgassing recombination rate is stable. Section 3.6 contains the outgassing quantification study, leak calibration testing using the outgassing quantification method, and the results of each. Section 3.7 describes the prototype that is being built based on the results of this study. 3.2 Setup and Design As previously stated in the literature review on outgassing behavior, the outgassing rate of stainless steel is influenced by bakeout temperature, bakeout interval, prolonged pumping, and atmospheric exposure. This information provided the basis for establishing the construction requirements of the experimental setup and the parameters at which the testing will be performed Design Implementation During initial tests, prolonged pumping was demonstrated to be inadequate at reducing outgassing, especially after internal vacuum components were exposed to

37 29 atmospheric pressures. Exposure to atmosphere is frequent in leak calibrations, occurring whenever parts are added to or removed from the vacuum system. It became clear that outgassing could not be reduced in a timely manner without the requisite integration of an oven. Therefore, a convection oven was added to the experimental setup, and the results were favorable. Baking removed moisture much faster than prolonged pumping, and the oven provided a way to regulate the resistance of the heating element, thus creating a means for testing outgassing at various temperatures. Another cause of higher outgassing is the use of thick-walled tubing. To minimize the outgassing coming from the bulk material, the experimental setup was built out of thin (inch) walled stainless steel tubing. The SRG and CDG gauges were used to measure the rise of pressure in the experimental setup. These gauges were chosen because they are both able to take direct readings, be interfaced and controlled through Data Acquisition (DAQ) software, and provide high accuracy and long-term stability. The SRG was used to measure the initial outgassing rate from 5x10-7 to 1x10-3 Torr while the CDG was used to measure the rise of pressures from 1x10-3 to 1x10-1 Torr. The maximum temperatures at which the SRG measuring head and CDG can operate at 50 C and 40 C, respectively. The SRG measuring head can combat temperature limitations by being dismounted from the ball/flange assembly and removed from the oven during bakeout, as shown in Figure 3.1. The CDG, however, cannot be moved, and the decision of where to place it became a significant problem. The issue was resolved by modifying the chamber and adding an additional porthole so that piping could connect the CDG to the experimental setup from outside the oven, as shown in Figure 3.2.

38 30 Figure 3.1 Experiment Design 1: SRG Ball/Flange Assembly and Measuring Head. Figure 3.2 Experiment Design 1: CDG Porthole.

39 Bakeout Temperature Determination Initial testing of the setup produced high and unpredictable outgassing rates, particularly after the manifold was exposed to atmosphere. These results were most likely attributed to the absorption of water vapor into the stainless steel components while they were not in use. Removal of this initial H 2 O content from the interior surfaces would take days or weeks of pumping. Thereafter, 12 to 24 hours of pumping would be required each time the pressure was brought up to atmosphere. Even so, the outgassing rate would still be too high for calibrating leaks slower than 10-7 atm cc/sec. It was concluded that outgassing of absorbed H 2 O could not be removed quickly unless the chamber was baked. The question then was, to what temperature and for how long? The results of the first experiment were used to determine the bakeout time and temperature at which low repeatable outgassing rates would be produced. In order to simulate a leak calibration, the test setup was deliberately exposed to atmosphere and a VCR gasket was replaced prior to each test. Figure 3.3 displays a schematic of the experimental setup and test connection point where the system was exposed to atmosphere. Figure 3.3 Experiment Design 1: Schematic Layout.

40 32 Next, the setup was baked at 200 C, 150 C, and at 22 C for a time period of one, three, and five hours, for a total of nine unique tests. It should be noted that bakeouts at 22 C was determined by the average ambient temperature in the laboratory. Tests were performed in a completely random order to minimize bias between bake temperatures and to minimize random variation. Immediately following each bake, the oven was allowed to cool at ambient temperatures for a total of three hours. When 30 minutes of the cooldown remained, the spinning rotor gauge was placed onto the ball/flange assembly and measurements were taken to verify that the pressure was stable or that it was reaching stability. The pressure stability check, shown in Figure 3.4, depicts the 1 minute average changes between pressure readings over the period leading up to the start of test through the next 50 minutes of data collection. The pressure before testing was found to be very stable changing on average less than 1x10-8 Torr. Figure 3.4 Experiment 1: SRG Pressure Stability Check. At t=0, the test was started by actuating the diaphragm valve, isolating the SRG from the pump. The pressure was allowed to accumulate overnight and into the next day;

41 33 total accumulation time ranged from 12 to 18 hours. Because the SRG is highly sensitive to vibration and was secured by a 1/4 in VCR fitting, the actuation of the diaphragm valve at t=0 produced erroneous data for the first several readings. The data was ignored since initial pressure readings would not affect the overall test results Results In each experiment run, the rise in pressure was measured by the SRG and the data was collected by a LabVIEW application developed specifically for this task. The data recorded by the SRG was not offset, so the actual pressure at the start of each test was unknown. The freely spinning ball is typically zeroed at a pressure below its capable measurement range at approximately 10-8 Torr. Within this range, minor changes in gas loads have little to no effect on the residual drag of the ball, and an offset or zero pressure can be determined so that the actual pressure is displayed on the controller. Because only the rate of rise of pressure was of concern, applying an offset correction to determine the actual pressure at the start of test would have produced the same results. As discussed in Chapter 1, the two causes of the rise in pressure are the leakage of atmospheric pressure through seals and the outgassing of stainless steel walls. For simplicity, the rate at which pressure rises will be referred to as the LO (leakage and outgassing) rate. The results of the different baking treatments are graphed in Figure 3.5. The lowest LO rates were produced when the setup was baked for three hours at 22 C, three hours at 150 C, and one hour at 150 C. The highest LO rates were produced when the setup was baked for one hour at 22 C and when it was baked for any amount of time at 200 C. At 150 C, LO rates were reduced during the one hour and three hour bakes, but were much higher during the five hour bake.

42 34 Figure 3.5 Experiment 1: Bakeout Temperature Determination Study (ΔP vs. Time). Bakes at higher temperatures repeatedly produced high LO rates, suggesting that extended baking periods at 150 C will also produce higher LO rates. The 22 C bakeouts resulted in the most random LO rates, although there might be some correlation with the previous treatments bakeout temperature. The first run at one hour produced the highest LO rate most likely because there was not a prior bakeout while the three hour bake produced LO rates almost identical to the preceding bakeout at one hour and 150 C. The five hour bake at 22 C and the second run of the one hour bake at 22 C also produced consistently lower LO rates following a three hour bake at 200 C and a five hour bake at 150 C. The results of the bakeout temperature determination experiment were consistent with those of a previous research performed by Jousten. Jousten also found that lower

43 35 bakeout temperatures produced lower outgassing rates after a stainless steel chamber was exposed to atmosphere [27]. He showed that bakeout temperatures as low as 100 C would produce lower outgassing rates than 150 C. However, the bakeouts lasted 60 hours, an impractical duration for a leak calibration prototype. A bake at 100 C was not considered for this experiment. Baking at 150 produces the lowest repeatable LO rates and is commonly practiced throughout the industry. The bakeout time and temperature relationship obtained from the experiment is shown in Figure 3.6. Figure 3.6 Experiment 1: Bakeout Time and Temperature Relationship Contour Plot. To express leakage and outgassing in terms of the gas throughput, the leak rate equation was derived from the ideal gas law equation. The equation (shown below) can be used to determine the volumetric flow rate into atmosphere. [ ] [ ] (3.1)

44 36 where is the leak rate (atm cc/sec), is the reference temperature ( C), is atmospheric pressure reference (Torr), is the actual chamber pressure (Torr), is the actual temperature at the time of test ( C), is the known reference volume (cm 3 ), and is the time (sec). The measured leak rate is shown in Figure 3.7. The leak rate for each test was measured by the change in chamber pressure and time keeping the following terms constant;,,, and. As expected, (because the equation is linear in ) the results were the same as those of the pressure rise method shown in Figure 3.5. However, Figure 3.7 provides a different perspective that illustrates the initial pressure instabilities and the period at which the data becomes linear is better than that in Figure 3.5. Figure 3.7 Experiment 1: Leakage and Outgassing vs. Time.

45 Leakage and Outgassing Repeatability Study The leakage and outgassing repeatability study was used to test whether or not the previously obtained LO rates are repeatable when the experiment is carried out under the same conditions. Based on the result from the bakeout temperature determination study, it was found that prebaking at 150 C produced the lowest LO rates. Predominantly those prebaked for 3 hours and at 1 hour which resulted in LO rates of 1.4x10-10 and 1.3 x10-10 atm cc/sec, respectively. Because of the low LO rates achieved at the shorter 1 hour and 150 C bake interval, the same testing parameters were used for the repeatability study Results A total of seven tests were carried out using the same testing parameters as the bakeout temperature determination study. To assure that the tests were performed in a repeatable manner, all controllable testing parameters were identical. Unlike the previous study, the intervals between the repeatability tests were also started and ended at precisely the same time. The result of the repeatability tests is shown in Figure 3.8. The plot shows the percent error that each test deviated from the average change in pressure. All tests appear to agree with each other except that of the first replicate and the higher percent error could indicate leakage. The results show that after baking at 150 C for one hour, pressures are repeatable to approximately ±8% of reading (%RDG) at 60 minutes. Thereafter the repeatability continued to improve to approximately ±5%RDG after 600 minutes. The cause of error was found by plotting the results in terms of leakage rate (calculated from Eq. 3.1) versus time in Figure 3.9. The data appeared to be unstable for the first 20 to 40 minutes after the setup was isolated from the pump. The initial fluctuations were most likely caused by vibrations picked up by the SRG after isolating the valve and by influx which is caused by gaseous molecules absorbing into the stainless steel, resulting in lower initial pressures.

46 38 Figure 3.8 Experiment 2: Repeatability Study Results Average Error (ΔP vs. Time). Figure 3.9 Experiment 2: Leakage and Outgassing vs. Time.

47 Leakage and Outgassing Predictability Study and Steady Outgassing Rate Study Based on the results of the bakeout temperature determination and the leakage and outgassing repeatability studies, it was concluded that LO rates are repeatable to ±8 %RDG after baking at 150 C for 1 hour. Although baking produced low LO rates that were fairly repeatable, two additional studies were conducted to determine if the variability could be reduced further. The first study, the Outgassing Predictability Study, demonstrates how to accurately predict outgassing rates from previously measured LO rates. The second study, the Steady Outgassing Rate Study, was used to distinguish the point at which the outgassing rate becomes constant and only leakages through seals can be detected by the measurement instrumentation. The methodologies and the results of each study are discussed in the next two sections Leakage and Outgassing Predictability Study To further reduce the variability in the data obtained from the outgassing repeatability study, the LO rates were analyzed to determine if shortened collection periods could be used to forecast future changes in the data. In order to perform the evaluation, subsets of the data acquired from the leakage and outgassing repeatability study were used to forecast results over preselected intervals in time. The LO pressure rise data was used to perform the analysis by fitting 10, 15, 20, 35 minute intervals of data by use of a 1 st, 2 nd, and 3 rd order polynomial equation. The coefficients acquired from the equations were then used to forecast new values over the next 5, 10, 15, 35 minutes. The forecasted pressures were then evaluated against the empirical results to determine the effectiveness of fit for each polynomial equation, data collection interval, and the point in time after the test setup was isolated from the pump. The effectiveness of the fit was determined by comparing the variability of the estimation errors with the variability of the original values, commonly known as R 2 [28]. R 2 values were calculated by

48 40 (3.2) the sum of squared errors, SSE is ( ) (3.3) the regression sum of squares, SSR is ( ) (3.4) and the total sum of squares, SST is (3.5) where is the actual pressure, is the predicted pressure, and is the mean of the actual pressures. Due to the amount data being analyzed, a database was developed to process over 16,000 records into 1,050 unique R 2 values. The SQL code is provided in Appendix G. Next, the R 2 values were used to determine which polynomial equation would produce the most favorable results. Three cumulative distribution function (CDF) plots (Figure 3.10) were created for the 1 st, 2 nd, and 3 rd order polynomial equations to compare the R 2 distributions. As shown in Figure 3.10, the CDF plots show that the second order polynomial has the highest percentage of R 2 values above where 1 is the optimal condition. Figure st, 2 nd, and 3 rd Order Polynomial CDF Plots.

49 41 After determining that the 2 nd Order Polynomial coefficients could be used to most accurately forecast the LO rates the question then was, at what predicted interval (5 min, 10 min, 15 min, 30 min) would give the best results? The statistical software package, JMP, was used to model the predicted intervals against the average R 2 values as shown in the one-way analysis plot in Figure Figure 3.11 Predicted Intervals vs. Average R 2 One-way Analysis. The green triangles represent the Analysis of Variance (ANOVA) where the centerline is the mean of the response and the height of the triangle is the 95% confidence interval. The red boxes represent the quantile distributions of data from amongst each predicted interval. Because the quantile distribution boxes were small for intervals 10 through 30, there was little variance amongst the R 2 values during those predicted intervals. The Tukey-Kramer and Hsu s MCB mean comparison circles show that the best R 2 values were obtained from prediction intervals 25 and 30 and the worst were those at a 5 minute intervals. The same analysis was used to determine how long after isolating from the pump should the data collection period begin. The analysis considered only 2 nd order polynomial fits and the 25 to 30 minute prediction intervals. JMP was used to model the start of data collection intervals against the average R 2 values as shown in the one-way analysis plot in Figure 3.12.

50 42 Figure 3.12 Data Collection Start Time vs. Average R 2 One-way Analysis. As shown in the plot, the quantile distribution boxes show less variance for collections periods starting at 15 minutes through 50 minutes and the student-t mean comparison circles show that the worst R 2 values are those at 10 minutes. Therefore, the best collection time to start collecting data is from 15 to 50 minutes after the vacuum pump has been isolated. To verify that the assumptions made in the statistical analysis were correct; the data collected from the repeatability test was filtered to show only 2 nd Order Polynomial coefficients having 25 and 30 minute prediction intervals with collection periods starting after 15 minutes. Percent errors were then obtained from the actual pressure readings and forecasted pressure readings. The Pareto plot, shown in Figure 3.12, illustrates the interactions amongst the different treatments and ordered relative to the size of the effects.

51 43 Figure 3.13 Treatment Interactions Pareto Plot. The Pareto plot shows that the least significant effects (green bars) are those having a prediction interval of 25 minutes and a data collection time starting from 15 to 45 minutes after the pump has been isolated. In conclusion, it was found that when the data collected begins 15 minutes after the pump was isolated, the data could data could be forecasted over the next 25 minutes with errors of 0.5% or less Steady Outgassing Rate Study The outgassing rate study was conducted to find the point in time where leakages can be distinguished from the outgassing of molecules that are desorbing from the interior surfaces of the vacuum chamber. Leakages are caused by molecules entering the vacuum chamber from atmosphere through small orifices, whereas outgassing is caused mostly by water molecules, H 2, CO, and CO 2 that have absorbed and are being desorbed from the interior surfaces. The vacuum components used throughout this research were made mostly of stainless steel, which created an additional source of outgassing caused by the reaction of iron oxides in the stainless steel, producing H 2 O [27]. Measurements taken during the repeatability study consisted of both leakage and outgassing. Leakages are known to be linear with time, while outgassing is expected to stabilize to a constant value after a period of time. When measured simultaneously, these two contributions will add together resulting in a linear relationship between the change over time where the intercept will represent the stabilized outgassing value. Correspondingly, the slope will

52 44 represent the stabilized leak rate. As shown in Figure 3.14, the change in slope plotted over a period of time never reached a constant state, indicating that outgassing did not stabilize over the 600 minute collection period. Figure 3.14 Change in Slope over 10 Minute Intervals. 3.6 Outgassing Quantification Study and Leak Calibration Testing From the Leakage and Outgassing Predictability Study, it was found that leakages and outgassing rates can be accurately predicted over 25 minute intervals with errors of 0.5% or less. To test these assumptions, the experimental setup was modified to incorporate two sets of valves and volumes as shown in Figure 3.15.

53 45 Figure 3.15 Experiment Design 2: Design 1 Modified to Include Two Volumes. The additional volumes and valves enabled direct comparisons between the LO rates of the manifold and the volumes to be measured separately, therefore generating a means to differentiate LO rates between the three. Testing was performed by measuring the changes in pressure before, during, and after the volumes were isolated for 30 minute periods. The data that was collected prior to and during isolation was used to predict the LO rate for each volume. The prediction data was then evaluated against the data collected during the entire 1-1/2 hour collection period. Lastly, the experimental study was modified once again so that a 10-7 atm cc/sec helium leak could be incorporated into the design for testing. The methodologies used and the results of the outgassing quantification study and leak calibration test is discussed in the next section Outgassing Quantification Study Results Two tests were carried out by baking at 150 C for one hour before cooling for a total of three hours. Like the previous tests, the spinning rotor gauge was placed onto the ball/flange assembly and measurements were taken for 30 minutes prior to starting the test to verify pressure stability. At t=0, the data collection period was started by actuating the diaphragm valve, V Pump, isolating the SRG from the pump. Figure 3.16 displays a schematic of the experimental setup where MV 1 and MV 2 is represented by the sections

54 46 shaded in red and blue respectively. The middle section, shown in a red/blue pattern, represents the overlapping sections. Figure 3.16 Experiment Design 2: Schematic Layout. At t=30 min, the data collection began and only the rise in pressures of LO was measured in the manifold M Act. At t=60 min, the diaphragm valve, V 1, was closed so that only LO from MV 2 were measured by the SRG. The pressure was allowed to accumulate in MV 2 until t=90min, V 1 was reopened allowing the pressure accumulated in Volume 1 to expand back into the manifold M Act. Plots of the first expansion with V 1 closed then reopened are shown for Test 1 and Test 2 in Figures 3.17 and 3.18, respectively.

55 47 Figure 3.17 Experiment Design 2: Test 1, Valve 1 Expansion (ΔP vs. Time). Figure 3.18 Experiment Design 2: Test 2, Valve 1 Expansion (ΔP vs. Time).

56 48 The double lines shown for M ACT (light blue) and M V2_ACT (light red) represent the rise in pressure caused by the LO that was directly recorded by the SRG. To simplify the plotted data, two sets of lines, light blue and dark blue, were used to show the predicted manifold pressures while the lower set of lines show the resulting LO of the volume. Predictions for the missing M ACT LO data between 60 and 90 minutes were based on the data collected between 30 and 60 minutes, and were found using the forecasting methods mentioned in the previous section. The predictions are illustrated by the upper M Pre_30-60 lines in light blue. Next, the missing M ACT LO data for the periods between 60 and 90 was predicted again by fitting a 2 nd order polynomial line, M Pre_30-60:90-120, using the data collected between the 30 to 60 minute and 90 to 120 minute periods. Because the M Pre_30-60: prediction periods provide a better approximation of the missing M ACT LO data, the M Pre_30-60: prediction periods were used to evaluate the effectiveness of the M Pre_30-60 prediction. Lastly, the LO rates of the volume were determined by taking the differences between the M Pre_30-60 and M Pre_30-60: predicted periods and the actual LO rate of the manifold M V2_Act. The predicted LO rates of the volumes are shown by the lower V 1_Pre_30-60 and V 1_Pre_30-60: lines in light blue and dark blue, respectively. The results of the M Pre_30-60 predictions in Test 1 and Test 2 showed that the 30 to 60 minute collection period in Test 2 produced a better prediction of the LO rates over 60 to 90 minutes than that of Test 1. The percent error between the M Pre_30-60 and M Pre_30-60: the instant before opening V 1 was calculated at 3.9% for Test 1 and 0.9% for Test 2. The high percent errors of Test 1 are suspected to be caused by the data collected between 30 and 60 minutes. The data collected during this period appeared to be linear compared to the LO rate of M ACT, which caused a higher predicted LO rate. After 120 minutes the test was performed again except V 2 was isolated instead of V 1 and the pressure was measured from M V1_Act rather than MV2_Act during the isolation period. Plots of the second expansion with V 2 closed then reopened are shown for Test 1 and Test 2 in Figures 3.19 and 3.20, respectively.

57 49 Figure 3.19 Experiment Design 2: Test 1, Valve 2 Expansion (ΔP vs. Time). Figure 3.20 Experiment Design 2: Test 2, Valve 2 Expansion (ΔP vs. Time).

58 50 The results of the M Pre_ predictions in Test 1 and Test 2 showed that the 90 to 120 minute collection periods produced highly accurate predictions of the LO rates of the manifold and volume 2. The percent error between the M Pre_30-60 and M Pre_30-60: the instant before opening V 2 was calculated to 0.1% for Test 1 and 0.2% for Test 2. In conclusion, the accuracy of LO rate predictions appeared to increase at higher pressures and longer periods after isolating from the pump Leak Calibration Testing The leak calibration test was performed using the outgassing quantification method to characterize and to correct for LO. The experimental setup was modified to incorporate a 10-7 atm cc/sec helium standard leak into the design, as shown in Figure Figure 3.21 Initial Leak Calibration Testing Design. Due to the size of the oven chamber and for safety purposes, the standard leak was not connected to the system until after baking. Even though initial isolation and pressure

59 51 stability tests did not show a significant amount of outgassing from the exposed piping and fittings, additional modifications were made to minimize the exposed surface area. The modified design reduced the exposed surface area from 4.6 cm 2 to 0.12 cm 2. The reduced piping also allowed room for a second pumping manifold to be incorporated into the design. The volumes, VOL 1 and VOL 2, were replaced with larger 25cc nominal volumes. Figure 3.22 displays a schematic of the leak calibration setup. Figure 3.22 Experiment Design 3: Schematic Layout. The leak calibration test was carried out by baking the experimental setup at 150 C for one hour before cooling for a total of three hours. The standard leak and spinning rotor gauge were then placed onto the stand and allowed to pump for 24 hours to reduce the moisture content on the surfaces exposed to atmosphere. After the prolonged pumpdown, three repeatability tests were performed to verify that LO was reduced and would not lead to additional errors during the calibration. The tests were conducted by isolating valves V P1, V P2, V LE for 30 minutes before opening valve V P1 and pumping for 30 minutes. As seen in Figure 3.23, the results of the LO repeatability test were mostly linear, indicating that outgassing was minimal. Because the LO pressure increased incrementally after each test, the base pressure could not easily be attained unless the setup was pumped for another 24 hours.

60 52 Figure 3.23 Experiment Design 3: Pretesting LO Repeatability Test. All valves except for valves V P2 and V LE were placed in the open position prior to testing. At t=0 min, the test was started by actuating the diaphragm valve, V P1, isolating the SRG from the high vacuum pump, At t=30 min, the data collection began and the rise in pressure was measured for M 1, M 2, VOL 1, and VOL 2. At t=60 min, the diaphragm valve, V 2, was closed so that only M 1, M 2, and VOL 1 were measured by the SRG. The pressure was allowed to accumulate in VOL 2 until t=90 min, V 2 was reopened allowing the pressure accumulated in Volume 1 to expand back into M 1, M 2, and VOL 1. At t=120 min, the M 2 was isolated from M 1, M 2, VOL 1, and VOL 2 by actuating the leak manifold valve, V LM. The remaining valves, V LE and V P2, were immediately opened after V LM causing the helium and LO to evacuate into the hold manifold pump, P HM. From 120 to 210 minutes the LO quantification test was repeated again except by isolating VOL 1 at 150 minutes before reopening at 180 minutes. A plot of the first two expansions with V 2 closed then reopened followed by V 1 closed then reopened is shown in Figure 3.24.

61 53 Figure 3.24 Experiment Design 3: VOL 2 and VOL 1 Predictions. The data sets with double lines represent the rise in pressure caused by the LO that was directly recorded by the SRG. The missing M 1 +M 2 +VOL 1 +VOL 2 LO data for the periods between 55 and 90 was determined by fitting a 2 nd order polynomial line, [M 1 +M 2 +VOL 1 +VOL 2 _PRE_30-55:90-120], using the data collected between the 30 to 55 minute and 90 to 120 minute periods. Next, the missing M 1 + +VOL 1 +VOL 2 LO data for the periods between 150 and 180 was predicted by fitting a 2 nd order polynomial line, [M 1 + +VOL 1 +VOL 2 _PRE_ : ], using the data collected between the 120 to 150 minute and 180 to 210 minute periods. The LO rates of the volumes were determined by taking the differences between the measured periods and the predicted periods. After the missing volume data was quantified, the missing M 2 LO data from 120 to 150 minutes was found by projecting out [M 1 +M 2 +VOL 1 +VOL 2 _PRE_30-55:90-120] another 30 minutes and taking the difference from M 1 + +VOL 1 +VOL 2. Reversely, the

62 54 missing M 2 LO data from 90 to 120 minutes was determined by projecting back [M 1 + +VOL 1 +VOL 2 _PRE_ : ] 30 minutes and taking the difference from M 1 +M 2 +VOL 1 +VOL 2. The predicted manifold LO rates, [M 2 _PRE_ : ] and [M 2 _PRE_30-55:90-120], are shown in Figure Figure 3.25 Experiment Design 3: M 2 Predictions. Because both M 2 predictions from 90 to 120 and 120 to 150 are correlated, the predicted LO rates from 90 to 120, [M 2 _PRE_ : ], and from 120 to 150 minutes, [M 2 _PRE_30-55:90-120], were combined and fit with a 2 nd order polynomial. The correlated M 2 predictions, [M 2_Correlated ], were projected out from 150 to 210 minutes and the predicted LO rates of M 1 +M 2 +VOL 1 +VOL 2 were determined by taking the difference between [M 1 + +VOL 1 +VOL 2 _PRE_ : ] and [M 2_Correlated ]. The correlated M 2 predictions, [M 2_Correlated ], and the projected values of M 1 +M 2 +VOL 1 +VOL 2, [M 1 +M 2 +VOL 1 +VOL 2 _PRE_30-210_Projection], are shown in Figure 3.26

63 55 Figure 3.26 Experiment Design 3: Correlated M 1 +M 2 +VOL 1 +VOL 2 Projection. At t 209 min, valve V P2 was closed and valves V LE and V LM were opened allowing the accumulated LO pressure to expand into M 2, as shown in Figure 3.27.

64 56 Figure 3.27 Experiment Design 3: Standard Leak Corrected for LO. At t 210 min, the first reading from the SRG showed that the helium depleting from the standard leak allowed the manifold to recover to the same pressure in less than a minute. At t 220 min, valve V 2 was closed temporarily allowing the standard leak and LO rates to increase until reopening at t 249 min. The initial pressure drop from 209 to 210 minutes was used to correct for the LO for the remainder of the test, as shown by Standard Leak (HE) LO and V2_Closed LO in dark blue and dark red, respectively. Although the leak rate of the standard leak was not calculated for this mock calibration, the leak rate could be determined from the opening and closing of V 1 or V 2. The results of the leak calibration test proved that the outgassing quantification method can be used to accurately measure and predict LO of the manifold and volumes during pretesting. Additionally, a 2 nd order polynomial can be used to accurately predict the LO during a leak calibration. The percent error of the LO prediction during the final

65 57 opening and closing of V2 from 220 to 249 minutes is suspected to be the same as the final pretesting LO from 150 to 180 minutes, 0.2%. Until further tests are conducted, an overall conservative approximation of the uncertainty of the outgassing quantification method is 2% RDG. 3.7 Development of the Prototype A fully functional prototype for calibrating standard leaks using the outgassing quantification method has been developed based on the results of the LO studies. Unlike the experimental designs used for testing, the prototype will have butt-welded joints to reduce atmospheric leakages through seals. To further reduce initial base pressure and outgassing, the prototype will utilize the following independent evacuation manifolds (shown in Figure 3.28): A) The Hold Manifold will be used for initial pumping after atmospheric exposure and to hold a leak before and after calibration, B) The UHV Manifold will be used for pumping the manifold encompassing the calibrated volumes, CDG, and SRG when the pressure is less than 10-3 Torr, C) The Mass Spectrometer Manifold will be used for directing and measuring the elemental composition of gases by use of a highly accurate quadruple mass spectrometer. Figure 3.28 Prototype Design: Evacuation Manifold Connections (Top View).

66 58 Each pumping manifold will use UHV magnetically levitated turbo molecular pumps backed by smaller UHV turbo molecular pumps to reduce hydrocarbons back streaming into the system. This reduction will ultimately achieve lower base pressures. A complete list of the pumps, pumping speeds, and their respective manifolds is located at the bottom of Appendix A. The prototype will take advantage of the lower uncertainty aspects from the techniques mentioned in Section 1.4 as well as the implementation of the Rate-of-Rise (ROR) methodology. It will be equipped with two calibration ports for performing batch calibrations or a high precision NIST calibrated standard can be added for implementing the accumulate dump and the direct comparison techniques. The accumulate dump and the direct comparison techniques will utilize the mass spectrometer and NIST calibrated standard leak by directly comparing the different leakage rates between the known standard and DUT. These techniques will be used mostly for helium leak calibrations while the outgassing quantification method will be used for calibrating leaks backfilled with mixed gases. The outgassing quantification method will also incorporate the ROR method to measure and correct for nonlinear pressure readings caused by initial fluctuations in LO after isolation, measurement errors from the instrumentation, and minute changes in LO rates during leak calibrations. Development of the prototype began in July The design was revised approximately 200 times to reflect increased knowledge of the process and the variables contributing to high calibration uncertainties. In January 2012, a final design for the prototype was selected and mechanical drawings for a custom vacuum manifold and mounting bracket were created. A multiview of the assembled prototype is shown in Figure 3.29 and detailed mechanical drawings of the mounting bracket and vacuum manifold are shown in Appendices F and G.

67 59 Figure 3.29 Prototype Design: Multiview. The fabrication of the prototype is made of two major parts: the mounting bracket and the vacuum manifold. The mounting bracket consists of four smaller brackets that can be completely assembled and disassembled at any time. The vacuum manifold will consist of valves, fittings, and tubing that will be welded and fastened to the mounting bracket. Valves are positioned closely together in order to minimize the overall line volume and surface area. The lower line volume enables higher pressures to be achieved, and the smaller surface area reduces outgassing from the interior surfaces. This creates an ideal environment for calibrating slower leaks at pressures with an accuracy of 1% using the CDG opposed to 1.5% to 2% using the SRG. But, the reduced volume could restrict access to the valves and tubing during welding. However, use of the mounting bracket as

68 60 a fixturing jig will help to assure that the valves and fittings have been aligned properly and to allow the fabricator to weld in such a sequence that all welds will be accessible. Because welding vacuum chambers is a highly specialized work, Vacuum Technology Incorporated (VTI) was contracted for the fabrication of the mounting bracket and vacuum manifold because of their unique capabilities and expertise in the field of vacuum science. To ensure that the prototype is fabricated in a manner that would result in the least amount of leakage and outgassing, a detailed statement of work (SOW), outlining the work activities and requirements of the final product was made between the purchasing agency and VTI prior to work start. The key requirements in the SOW are listed below: All tubing provided by the vendor should be 316L Stainless Steel, ¼"OD and 0.035" Wall Thickness. All welds shall be "butt welds." Socket welds or any welding technique that could create virtual leaks or could contribute to a higher outgassing rate must be avoided. The vacuum manifold shall be baked at 150 C for 12 hours. Test 1: The vacuum manifold shall be successfully leak tested to 10-9 atm cc/sec ±30%RDG with helium and provide report. Test 2: The vacuum manifold shall be able to successfully reach pressures of 10-8 Torr or lower within 12 hours of continuous pumping. The manifold may be baked at 150 C prior to testing. Provide report with bake out time, temperature, and all data collected during the test(s).

69 61 Chapter 4 Conclusion and Recommendations 4.1 Conclusion The main contributors of high calibration uncertainty were internal leaks and outgassing from the internal surfaces, fittings, and valves within the vacuum chamber. Leakage and outgassing is undesirable when calibrating standard leaks because it results in a condition of initial pressure rise that yields a calculated value of leakage rate that is greater than the true value [1]. The outgassing quantification method is a primary calibration technique that is an extension of the method, which can be used to distinguish leakage and outgassing of the vacuum manifold from the leakage of the standard leak. This method can be used to determine the actual rise in pressures of the standard leak by subtracting the leakage and outgassing rate from the total pressure in the vacuum manifold. The results obtained from the experiments proved that the outgassing quantification method can be used to accurately predict the accumulation of outgassing in a known volume to within 0.2% of the actual measured value. Based on the results of the leak calibration test, the method was also proven to accurately measure and predict leakage and outgassing rates of the manifold and volumes during mock calibrations of a standard leak. This research concluded that the outgassing quantification method can reduce-if not completely eliminate-the once unknown pressure measurement uncertainty of leakage and outgassing when calibrating leaks rates from 1x10-2 atm cc/sec to 9.9x10-9 atm cc/sec. Fabrication of the prototype is expected to be complete in May The prototype will undergo extensive testing upon receipt. A LabVIEW application is being developed to automate testing of the outgassing quantification method. A verity of standard leaks having different leak rates will be used to evaluate the repeatability and reproducibility of the new system against leak rates calibrated at a higher level lab or NIST. If needed, additional improvements or modification will be made until the prototype is deemed adequate for production use. The modular design and extruded

70 62 aluminum mounts will allow for additional prototypes to easily be adapted to a leak calibration stand similar to that shown in Figure 4.1. Figure 4.1 Prototype Design: Conceptual Rendering of the Leak Calibration Stand. 4.2 Recommendations Although the leakage and outgassing rates could be accurately determined during the leak calibration test, the re-evacuation of the leak manifold (M 2 ) could introduce additional uncertainties. The evacuation tests that were performed prior to the leak calibration testing in section suggested that leakages and outgassing rates will increase incrementally after the isolated manifold is evacuated and re-isolated. Reevacuation of the leak manifold is unavoidable because the accumulation of gas from the standard leak must be evacuated prior to calibration. Measures must be taken to stabilize pressures in the leak manifold to avoid additional uncertainties caused by the increased leakage and outgassing rates after re-evacuation. Plans for future testing and modifications; 1. Due to the vibrations caused by pneumatic actuation of valves and because the SRG is highly sensitivity to vibrations, significant pressure measurements errors were observed during testing. To minimize the measurement errors caused by

71 63 these vibrations, slight modification have been made to the prototype to isolate the SRG. A collar clamp will be used to secure the SRG conflate flange to a granite surface in hopes of eliminating additional calibration uncertainties. 2. Due to the increased leakage and outgassing rates that were observed after reevacuation of the leak manifold, a pressure control valve will be added to the leak manifold to evacuate the leak at a controlled rate. Interfacing the pressure control valve and a CDG gauge with LabVIEW will allow for the manifold to be evacuated at approximately the same rate as the leakage from the standard leak, and therefore, eliminate changes in outgassing rate from the leak manifold.

72 64 List of References 1. Ehrlich, C., & Basford, J. (1992). Recommended practices for the calibration and use of leaks. Journal of Vacuum Science & Technology A: Vacuum, Surfaces, and Films, 10, McGregor, K. (2008, September 9). Tracer Gas Calibrated Leak Standard Model Selection Guide. Cincinnati Test Systems. Retrieved February 2, 2012, from 3. CalMaster Calibrated Leaks Manual (PDF). (2010, December 2). Calibrated Leaks from LACO Technologies. Retrieved February 5, 2012, from 4. Calibrated Gas Leaks. (2010, May 11). Calibrated Gas Leaks. Retrieved February 5, 2012, from 5. Buxton, K. (2004, September 3). Calibrator Selection Guide for Helium Leak Detectors and Systems. Vacuum Technology Inc. Retrieved February 5, 2012, from df 6. Abbott, P., & Jabbour, Z. (2011, November 16). An Introduction to Mass Metrology in Vacuum. Cal Lab, 18, Number 4, Hanlon, J. F. (2003). A user's guide to vacuum technology (3 rd ed.). Hoboken, NJ: Wiley-Interscience. 8. Lipt k, B. G. (2012). Instrument engineers' handbook (4 th ed.). Boca Raton: CRC Press. 9. MKS Instruments - SRG-3 Spinning Rotor Gauge System. (2 nd ed.). MKS Instruments the world's broadest source of instruments, components, and subsystems, for vacuum and gas-based processes. Retrieved February 11, 2012, from Šetina, J., Zavašnik, R., & Nemanič, V. (1987). Vacuum tightness down to the mbar l s-1 range, measured with a spinning rotor viscosity gauge. Journal of Vacuum Science & Technology A: Vacuum, Surfaces, and Films, 5(4), Redhead, P. (2002). Recommended practices for measuring and reporting outgassing data. Journal of Vacuum Science & Technology A: Vacuum, Surfaces, and Films, 20(5), Vacuum History & Technology. (2011, October 20). McAllister Technical Services Home Page. Retrieved February 20, 2012, from Danielson, P. (2004, May 25). Choosing the Right Vacuum Materials. A Journal of Practical and Useful Vacuum Technology. Retrieved February 20, 2012, from Danielson, P. (2006, November 3). Nitpicking Your Vacuum System. A Journal of Practical and Useful Vacuum Technology. Retrieved February 24, 2012, from

73 Bernhardt, K. (2010). Vacuum Technology Compendium. Asslar, Germany: Pfeiffer Vacuum. 16. Danielson, P. (2003, October 4). How To Choose a Roughing/Backing Pump for the Turbo and Drag Family. A Journal of Practical and Useful Vacuum Technology. Retrieved February 26, 2012, from Biltoft, P., Benapfl, M., & Swain, T. (2003, February 24). Chapter 4: Vacuum System Components. Las Positas College Vacuum Technology 60A & 60B. Retrieved March 1, 2012, from lpc1.clpccd.cc.ca.us/lpc/tswain/chapt4.pdf 18. Carison, M., Wheeler, W. (1965). U.S. Patent No. 3,208,758. Washington, DC: U.S. Patent and Trademark Office. 19. Vacuum Components KF/ISO Series. (2007, March 7). HIS. Retrieved March 1, 2012, from KF (QF) Flanges Technical Notes. Kurt J. Lesker Company Vacuum Scien our Business. Retrieved March 1, 2012, from VCR Metal Gasket Face Seal Fittings. (2010, February 17). Metal Gasket Seal VCR Fitting - Swagelok. Retrieved March 7, 2012, from Hoffman, D., Singh, B., & Thomas, J. (1998). Handbook of Vacuum Science and Technology. Oxford, UK: Academic Press. 23. Callahan, F., Falls, C., Wennerstrom, E. (1970). U.S. Patent No. 3,521,910. Washington, DC: U.S. Patent and Trademark Office. 24. Liptak, B. (2003). Pressure Measurement. Instrument Engineers' Handbook (4th ed., pp ). London: CRC Press. 25. Ahrens, C. (2008). Essentials of meteorology: An Invitation to the Atmosphere (5th ed.). New York, NY: Thomson Brooks Cole. 26. Sedlmayr, E., Eichler, H., Kronfeldt, H., Jousten, K., & Seidel, J. (2005, November 30). Investigation of TDLAS for its Application as Primary Standard for Partial Pressure Measurements. KOBV - Opus & Archivierung. Retrieved March 13, 2012, from opus.kobv.de/tuberlin/volltexte/2005/1161/pdf/padilla_gerardo.pdf 27. Jousten, K. (1999, June 3). Thermal Outgassing. Conseil Européen pour la Recherche Nucléaire. Retrieved March 10, 2012, from R-Squared Definition and Other Information. Hedge Fund Consistency Index, Hedge Funds Research. Retrieved March 20, 2012, from Section 2 - Valves. (2009, October 23). Kurt J. Lesker Company Home Page Vacuum Science is our Business. Retrieved March 10, 2012, from

74 E Pressure/Flow Controller and 245 Control Valve. (2009, August 10). MKS Instruments - Pressure Measurement & Control. Retrieved March 10, 2012, from Atomic Layer Deposition, Atomic Layer Deposition (ALD) Diaphragm Valves, (MS , R6). (2008, June 3). Diaphragm-Sealed Valves, Radial Diaphragm, Ultra High Purity, High Flow - Swagelok. Retrieved March 10, 2012, from

75 67 Appendix A Prototype Schematic Diagram

76 68 Appendix B Prototype Model

77 69 Appendix C Vacuum Valve Specifications Angle Valves Seal Flanges Actuation Copper Bonnet, Fluorocarbon Bonnet, Viton, Chemraz CF, KF, ASA, VCR, ISO Manual, Pneumatic, Electromagnetic Temp. Rating Pressure Rating Lest Test Sensitivity <=200 C 10-9 Torr to 15 psig 2x10-10 atm cc/sec Gate Valves Seal Flanges Actuation Copper Bonnet, Fluorocarbon Bonnet, Viton, Chemraz CF, KF, ASA, VCR, ISO Manual, Pneumatic, Electromagnetic Temp. Rating Pressure Rating Lest Test Sensitivity <=200 C 10-9 Torr to 15 psig 2x10-10 atm cc/sec Pressure Control Valve Seal Flanges Actuation Metal-to-Metal Seal CF, VCR, NPT Stepper Motor Temp. Rating Pressure Rating Lest Test Sensitivity <=50 C Torr to 15 psig 2x10-11 atm cc/sec Diaphragm Valve Seal Flanges Actuation Elgiloy Diaphragm and PFA Seat VCR, Tube Butt Weld Manual, Pneumatic, Electromagnetic, Temp. Rating Pressure Rating Lest Test Sensitivity <=200 C < 35 psig 1x10-9 atm cc/sec Figures and Specifications for the valves were obtained from the following sources; Angle Valve - Kurt J. Lesker Catalog [29], Gate Valve - Kurt J. Lesker Catalog [29], Pressure Control Valve - MKS Instruments [30], Diaphragm Valve Swagelok Company [31].

78 70 Appendix D Outgassing Predictability Study Results Index Appendix Title Page D.1 Outgassing Predictability Study 1 st Order 71 Polynomial R 2 Results D.2 Outgassing Predictability Study 2 nd Order 72 Polynomial R 2 Results D.3 Outgassing Predictability Study 3 rd Order Polynomial R 2 Results 73

79 71 Appendix D.1 Outgassing Predictability Study 1 st Order Polynomial R 2 Results Measure Period (Minutes) Predicted Period (Minutes) R 2 Results Start End Duration Start End Duration Avg. R 2 Rank

80 72 Appendix D.2 Outgassing Predictability Study 2 nd Order Polynomial R 2 Results Measure Period (Minutes) Predicted Period (Minutes) R 2 Results Start End Duration Start End Duration Avg. R 2 Rank

81 73 Appendix D.3 Outgassing Predictability Study 3 rd Order Polynomial R 2 Results Measure Period (Minutes) Predicted Period (Minutes) R 2 Results Start End Duration Start End Duration Avg. R 2 Rank

82 74 Appendix E Average R 2 Database SQL Code Table Description: Table Name: Field Names\Field Type: Table Description: Table Name: Field Names\Field Type: Query Description: Query Name: SQL: Query Description: Query Name: SQL: Query Description: Query Name: SQL: Query Description: Query Name: SQL: Actual Pressure and Predicted Pressure Input Table tbl_data String_ID\Text, Time (Minutes)\Number, Rep_1_Actual\Number, Rep_2_Actual\Number, Rep_3_Actual\Number, Rep_4_Actual\Number, Rep_5_Actual\Number, Rep_6_Actual\Number, Rep_7_Actual\Number, Rep_1_Predicted\Number, Rep_2_Predicted\Number, Rep_3_Predicted\Number, Rep_4_Predicted\Number, Rep_5_Predicted\Number, Rep_6_Predicted\Number, Rep_7_Predicted\Number Test Names and Unique Fields tbl_test_names ID\Number, String_ID\Text, Known_Start\Number, Known_End\Number, Unknown_Start\Number, Unknown_End\Number, Order\Text Calculate average of each subset qry_mean SELECT DISTINCTROW tbl_data.string_id, Avg(tbl_Data.Rep_1_Actual) AS AvgOfRep_1_Actual, Avg(tbl_Data.Rep_2_Actual) AS AvgOfRep_2_Actual, Avg(tbl_Data.Rep_3_Actual) AS AvgOfRep_3_Actual, Avg(tbl_Data.Rep_4_Actual) AS AvgOfRep_4_Actual, Avg(tbl_Data.Rep_5_Actual) AS AvgOfRep_5_Actual, Avg(tbl_Data.Rep_6_Actual) AS AvgOfRep_6_Actual, Avg(tbl_Data.Rep_7_Actual) AS AvgOfRep_7_Actual, Avg(tbl_Data.Rep_1_Predicted) AS AvgOfRep_1_Predicted, Avg(tbl_Data.Rep_2_Predicted) AS AvgOfRep_2_Predicted, Avg(tbl_Data.Rep_3_Predicted) AS AvgOfRep_3_Predicted, Avg(tbl_Data.Rep_4_Predicted) AS AvgOfRep_4_Predicted, Avg(tbl_Data.Rep_5_Predicted) AS AvgOfRep_5_Predicted, Avg(tbl_Data.Rep_6_Predicted) AS AvgOfRep_6_Predicted, Avg(tbl_Data.Rep_7_Predicted) AS AvgOfRep_7_Predicted FROM tbl_data GROUP BY tbl_data.string_id; Calculate sse for each record ([Rep_Actual]-[Rep_Predicted])^2 qry_sse1 SELECT tbl_data.string_id, tbl_data.[time (Minutes)], ([Rep_1_Actual]-[Rep_1_Predicted])^2 AS sse1_1, ([Rep_2_Actual]-[Rep_2_Predicted])^2 AS sse1_2, ([Rep_3_Actual]-[Rep_3_Predicted])^2 AS sse1_3, ([Rep_4_Actual]-[Rep_4_Predicted])^2 AS sse1_4, ([Rep_5_Actual]-[Rep_5_Predicted])^2 AS sse1_5, ([Rep_6_Actual]-[Rep_6_Predicted])^2 AS sse1_6, ([Rep_7_Actual]-[Rep_7_Predicted])^2 AS sse1_7 FROM tbl_data; Calculate ssr for each record ([Rep_Predicted]-[AvgOfRep_Actual]-[)^2 qry_ssr1 SELECT tbl_data.string_id, tbl_data.[time (Minutes)], ([Rep_1_Predicted]-[AvgOfRep_1_Actual])^2 AS ssr1_1, ([Rep_2_Predicted]-[AvgOfRep_2_Actual])^2 AS ssr1_2, ([Rep_3_Predicted]- [AvgOfRep_3_Actual])^2 AS ssr1_3, ([Rep_4_Predicted]-[AvgOfRep_4_Actual])^2 AS ssr1_4, ([Rep_5_Predicted]-[AvgOfRep_5_Actual])^2 AS ssr1_5, ([Rep_6_Predicted]-[AvgOfRep_6_Actual])^2 AS ssr1_6, ([Rep_7_Predicted]-[AvgOfRep_7_Actual])^2 AS ssr1_7 FROM qry_mean INNER JOIN tbl_data ON qry_mean.string_id = tbl_data.string_id GROUP BY tbl_data.string_id, tbl_data.[time (Minutes)], ([Rep_1_Predicted]- [AvgOfRep_1_Actual])^2, ([Rep_2_Predicted]-[AvgOfRep_2_Actual])^2, ([Rep_3_Predicted]- [AvgOfRep_3_Actual])^2, ([Rep_4_Predicted]-[AvgOfRep_4_Actual])^2, ([Rep_5_Predicted]- [AvgOfRep_5_Actual])^2, ([Rep_6_Predicted]-[AvgOfRep_6_Actual])^2, ([Rep_7_Predicted]- [AvgOfRep_7_Actual])^2; Calculate SSE by taking sum of qry_sse1 grouping on String_ID qry_sse SELECT qry_sse1.string_id, Sum(qry_sse1.sse1_1) AS SSE_1, Sum(qry_sse1.sse1_2) AS SSE_2, Sum(qry_sse1.sse1_3) AS SSE_3, Sum(qry_sse1.sse1_4) AS SSE_4, Sum(qry_sse1.sse1_5) AS SSE_5, Sum(qry_sse1.sse1_6) AS SSE_6, Sum(qry_sse1.sse1_7) AS SSE_7 FROM qry_sse1 GROUP BY qry_sse1.string_id;avgofrep_3_actual])^2, ([Rep_4_Predicted]- [AvgOfRep_4_Actual])^2, ([Rep_5_Predicted]-[AvgOfRep_5_Actual])^2, ([Rep_6_Predicted]- [AvgOfRep_6_Actual])^2, ([Rep_7_Predicted]-[AvgOfRep_7_Actual])^2;

83 75 Query Description: Query Name: SQL: Query Description: Query Name: SQL: Query Description: Query Name: SQL: Calculate SSR by taking sum of qry_ssr1 grouping on String_ID qry_ssr SELECT qry_ssr1.string_id, Sum(qry_ssr1.ssr1_1) AS SSR_1, Sum(qry_ssr1.ssr1_2) AS SSR_2, Sum(qry_ssr1.ssr1_3) AS SSR_3, Sum(qry_ssr1.ssr1_4) AS SSR_4, Sum(qry_ssr1.ssr1_5) AS SSR_5, Sum(qry_ssr1.ssr1_6) AS SSR_6, Sum(qry_ssr1.ssr1_7) AS SSR_7 FROM qry_ssr1 GROUP BY qry_ssr1.string_id; Calculate SST, SST=[SSE]+[SSR] qry_sst SELECT qry_ssr.string_id, [SSE_1]+[SSR_1] AS SST_1, [SSE_2]+[SSR_2] AS SST_2, [SSE_3]+[SSR_3] AS SST_3, [SSE_4]+[SSR_4] AS SST_4, [SSE_5]+[SSR_5] AS SST_5, [SSE_6]+[SSR_6] AS SST_6, [SSE_7]+[SSR_7] AS SST_7 FROM qry_sse INNER JOIN qry_ssr ON qry_sse.string_id = qry_ssr.string_id; Calculate R 2, (1-[SSE]/[SST]) qry_r_squared SELECT qry_sst.string_id, tbl_test_names.known_start, tbl_test_names.known_end, tbl_test_names.unknown_start, tbl_test_names.unknown_end, tbl_test_names.order, (1- [SSE_1]/[SST_1]) AS R2_1, (1-[SSE_2]/[SST_2]) AS R2_2, (1-[SSE_3]/[SST_3]) AS R2_3, (1- [SSE_4]/[SST_4]) AS R2_4, (1-[SSE_5]/[SST_5]) AS R2_5, (1-[SSE_6]/[SST_6]) AS R2_6, (1- [SSE_7]/[SST_7]) AS R2_7 FROM tbl_test_names INNER JOIN (qry_sse INNER JOIN qry_sst ON qry_sse.string_id = qry_sst.string_id) ON tbl_test_names.[string_id] = qry_sse.string_id;

84 76 Appendix F Mounting Bracket CAD Drawing Index Appendix Drawing # Drawing Name Page F.1 MB1EXP Assembly Exploded View 77 F.2 MB1N MB1 1/4 Laser Cut Nested Layout 78 F.3 MB1BN MB1B 1/2 Laser Cut Nested Layout 79 F.4 MB1DN MB1DN 1/4 Laser Cut Nested Layout 80 F.5 MB1A01 MB1 Plate A01 Machining 81 F.6 MB1B01 MB1 Plate B01 Machining 82 F.7 MB1B02 MB1 Plate B02 Machining 83 F.8 MB1B03 MB1 Plate B03 Machining 84 F.9 MB1B04 MB1 Plate B04 Machining 85 F.10 MB1BAWL MB1B Assembly Weld Locations 86 F.11 MB1C01 MB1 Plate C01 Machining 87 F.12 MB1C02 MB1 Plate C02 Machining 88 F.13 MB1CAWL MB1C Assembly Weld Locations 89 F.14 MB1D01 MB1 Plate D01 Machining 90 F.15 MB1DAWL MB1D Assembly Weld Locations 91

85 Appendix F.1 Mounting Bracket: Assembly Exploded View 77

86 Appendix F.2 Mounting Bracket: MB1 1/4 Laser Cut Nested Layout 78

87 Appendix F.3 Mounting Bracket: MB1B 1/2 Laser Cut Nested Layout 79

88 Appendix F.4 Mounting Bracket: MB1DN 1/4 Laser Cut Nested Layout 80

89 Appendix F.5 Mounting Bracket: MB1 Plate A01 Machining 81

90 Appendix F.6 Mounting Bracket: MB1 Plate B01 Machining 82

91 Appendix F.7 Mounting Bracket: MB1 Plate B02 Machining 83

92 Appendix F.8 Mounting Bracket: MB1 Plate B03 Machining 84

93 Appendix F.9 Mounting Bracket: MB1 Plate B04 Machining 85

94 Appendix F.10 Mounting Bracket: MB1B Assembly Weld Locations 86

95 Appendix F.11 Mounting Bracket: MB1 Plate C01 Machining 87

96 Appendix F.12 Mounting Bracket: MB1 Plate C02 Machining 88

97 Appendix F.13 Mounting Bracket: MB1C Assembly Weld Locations 89

98 Appendix F.14 Mounting Bracket: MB1 Plate D01 Machining 90

99 Appendix F.15 Mounting Bracket: MB1D Assembly Weld Locations 91

100 92 Appendix G Valve Layout CAD Drawing Index Appendix Drawing # Drawing Name Page G.1 VLOP Valve Layout Parts 93 G.2 VLOP3D Valve Layout ORMC Provided Parts 94 G.3 VLOPARTS Valve Layout Parts Model 95 G.4 VLM_TOP Valve Layout Multiview Top 96 G.5 VLM_LHS Valve Layout Multiview Left Hand Side 97 G.6 VLM_FRONT Valve Layout Multiview Front 98

101 Appendix G.1 Valve Layout: Parts 93

102 Appendix G.2 Valve Layout: ORMC Provided Parts 94

103 Appendix G.3 Valve Layout: Parts Model 95