 |
 |
 |
 |
|
Comments? Suggestions? Send us your feedback.
MicroMagazine.com |
Ultrapure Gas DeliveryDevising specifications for optimizing point-of-use gas pressure regulatorsDan Morgan, Parker Hannifin Veriflo A point-of-use pressure regulator must perform properly so that it can provide steady gas pressure during delivery and maintain a consistent set-point pressure when the gas stick is opened and closed. Strides in computer performance and the decreasing cost of manufacturing semiconductor wafers relative to size are the result of many small, incremental improvements in wafer design and manufacturing processes. These improvements rely, in turn, on the quality of all of the components that are used in the fab, from the environmental system to the wafer processing tools themselves. Therefore, advancements in the quality of such components, no matter how small, are important to the entire semiconductor industry because they either support or spawn other advancements. This article examines point-of-use pressure regulators and the need to develop specifications that can optimize their use and bolster the performance of the processing tools to which they are attached. Process Tool Gas Delivery System One of the requirements of advanced wafer processing is that tool chambers must receive a consistent, accurate, and repeatable supply of process gases without perturbations or flow spikes that can affect the uniformity of the wafer chemistry. Gas delivery subsystems, often referred to as "gas sticks," fulfill this requirement. In a typical semiconductor fabrication process line, gas sticks are used to initiate, deliver, and shut off the flow of gases into process tools, as dictated by the process recipe. These subsystems generally consist of several automatic shutoff valves, a pressure regulator, and a mass-flow controller (MFC). A gas stick receives its supply from a "gas room," which contains the various high-pressure process gases used throughout the facility. These gases are plumbed to a medium-pressure distribution manifold that moves the gases from the lower-floor gas room up to the wafer-processing tools located in the upper-floor cleanrooms. A given gas can be delivered to several different tools and several chambers within each tool. The gas stick incorporates valves that permit the inflow and outflow of process gas and enable the stick to be purged of air prior to start-up. The MFC is used to control the flow of gas to the tool chamber for the duration of the process reaction. Because process steps can be as short as several seconds in duration, it is important that the MFC ramp up from zero to the desired flow as quickly as possible. The gas stick's point-of-use regulator maintains a relatively constant pressure to the MFC inlet to help stabilize flow. To ensure proper gas flow in the fab, it is useful to understand how pressure regulators function and how they relate to the MFC in a point-of-use gas delivery system. Regulator Basics A schematic diagram of a typical point-of-use regulator is presented in Figure 1. This device controls pressure in a quite simple way. A handwheel compresses a spring when turned. This spring, called a range spring, pushes down on a metal diaphragm, which separates the process gas from the external environment with a metal-to-metal seal. The diaphragm, in turn, drives open the regulator poppet valve, which allows the high-pressure gas to flow into the diaphragm cavity. When the gas pressure acting on the diaphragm area generates a force equal to that of the range spring, the diaphragm returns to a neutral position, allowing the poppet to close. The regulator diaphragm maintains the balance between the range spring and the process gas pressure. Several other forces within the regulator contribute to this force balance, including the spring rate of the diaphragm itself multiplied by the diaphragm stroke, the internal bias spring load, and the inlet pressure acting on the poppet area. This force balance is mathematically represented in the following equation:
(Pout x Adiaphragm) + Fdiaphragm + Fbias spring + (Pin x Apoppet) = Frange spring
As the regulator opens to provide increasing gas flow, the poppet, diaphragm, and springs move to create the required opening at the poppet and seat interface for the set flow. While this "downward" repositioning of the poppet results in a lower range-spring force because of reduced compression on the range spring as the diaphragm moves downward, it also causes the diaphragm and the bias spring to exert increased force. Consequently, outlet pressure drops as flow increases. This relationship is shown in Figure 2, a flow curve for a point-of-use regulator that depicts the change in outlet pressure as a function of flow. The flow curve illustrated in Figure 2 is actually composed of two curves, each of which represents regulator outlet pressure at any given flow rate, as flow is increased and then decreased. The measured pressure difference between these two curves at any point is the hysteresis of the regulator. Hysteresis is caused by friction between the regulator components. Many regulator designs incorporate a friction device that helps to dampen the runaway oscillation which occurs in an underdampened pressure regulator. The steep slope at the beginning and termination of flow is described as the "creep" or "lockup" region of the flow curve.
Under steady-state flow conditions, the regulator maintains a force balance between the regulator outlet pressure applied to the diaphragm and the regulator range spring, while negligible force is exerted on the regulator poppet and seat interface. When flow is terminated, outlet pressure continues to rise until sufficient force develops between the poppet and seat to effect a seal. When flow is initiated through the regulator, the creep pressure is drained off before the regulator poppet lifts off the seat. If there is any appreciable friction between the poppet and the seat, the outlet pressure continues to drop slightly below the outlet pressure set point before jumping back up when the static friction has been overcome. Referred to as transient overshoot, this sudden reversal in the slope of the regulator outlet pressure relative to time can cause the MFC to spike. The timing of this event can be critical to the process and is dependent on system volume, flow rate, and the amount of creep pressure buildup. The response of a regulator relative to time as a change in the flow rate occurs, known as transient response, is an important aspect of regulator performance. While any change in flow rate affects regulator outlet pressure, the change from zero flow to a low-flow setting has the greatest impact on semiconductor processes. Low-flow transient response is illustrated in Figure 3, which presents a graph of regulator outlet pressure as a function of time as flow is initiated in a regulator exhibiting excessive friction. This graph shows that the outlet pressure of this regulator overshot the desired pressure set point by 1.0 psi.
Testing Regulator Performance A series of tests was performed to investigate the relationship between regulator performance and the stability of gas delivery from a point-of-use gas delivery system under the flow-rate conditions that are typical in semiconductor fabrication processes. A schematic of the gas delivery system used to perform the tests is illustrated in Figure 4. The system consisted of a system inlet pressure transducer, a pressure regulator, a regulator outlet pressure transducer, an air-operated isolation valve, a mass-flow controller inlet pressure transducer, a mass-flow controller, and an air-operated outlet valve. The system's inlet pressure was set at 65 psia while the regulator outlet pressure was set at 30 psia at zero flow. The flow settings of 20 and 100 std cm3/min were 20 and 100% of full MFC scale. Control signals to the air-operated valves and the MFC were given simultaneously. To maintain consistency relative to the MFC in-line leak rate, the system was opened and then closed for 10 seconds before each test was begun.
The effect of regulator friction on pressure stability was studied in a step function transient response test. As described in Semaspec Standard 90120392B-STD ("Test Method for Determination of Regulator Performance for Gas Distribution System"), this test demonstrates the effect of friction on regulator outlet pressure as gas flow is quickly changed from 25 to 50% of the maximum rated flow for the regulator. The maximum rated flow is defined as the maximum flow within the regulator's control range, or the maximum flow recommended by the manufacturer. A test using the maximum flow within the control range of a state-of-the art point-of-use regulator provided the following values:
Because most semiconductor fabrication facilities use flow rates that are far lower than the test flows provided by this method and, as this study demonstrated, the step function transient response test does not provide an adequate measure of pressure stability in typical low-flow applications, a low-flow transient response test was explored to evaluate regulator response to a sudden initial demand for flow. This low-flow transient response test differs from the SEMI transient response test in Semaspec 90120392B-STD because regulator response is evaluated as the flow rate is varied quickly from 0 to 350 std cm3/min. Data were collected at a rate of 1000 samples per second using Labview data acquisition software (National Instruments, Austin, TX). Three regulators displaying widely varying low-flow transient response behavior were tested. The Semaspec method was used to test these components for step function transient response. The test data were then analyzed to determine whether there was a quantifiable difference between the regulators. Finally, the regulators were installed in the type of gas-flow system that is typically used in semiconductor fabs and tested to determine the effects of the varying degrees of low-flow transient response. In test 1, regulator outlet pressure and gas flow were measured as a function of time as flow was initiated. The test flow rate was 20 std cm3/min and the MFC was set at 20% of full flow capacity. In test 2, the flow rate was 100 std cm3/min and the MFC was set at 100% of full flow capacity while the other parameters remained the same. Tests 3 and 4 were the same as tests 1 and 2, respectively, except the pressure across the MFC was allowed to equalize to demonstrate the effects of a relatively long delay before the initiation of gas flow. A range of parameters related to the operation of the components shown in Figure 4 were monitored as a function of time: system inlet pressure (a), regulator outlet pressure (c), the pressure-control signal to the air-operated valves (d and g), the control signal to the MFC (f), MFC inlet pressure (e), and the MFC flow rate. The test data were analyzed to determine regulator transient overshoot, transient response, regulator stabilization time, MFC stabilization time, and flow spike as a percentage of set flow (only spikes that occurred after the MFC had stabilized are reported). Values for these parameters are presented in Tables I and II.
In all the tests, flow spikes were initially registered as the isolation valve (d) was opened. These spikes, a product of the MFC design, were caused not by flow but by voltage. They were most severe when the pressure across the MFC was equalized before flow was initiated. As depicted in Figure 5, the rush of pressure into the MFC caused it to close, delaying the flow of gas.
The data for regulator 3, test 2, illustrated in Figure 6, revealed that a regulator outlet pressure transient overshoot of 1.0 psi is sufficient to cause an MFC to react, giving a flow signal in excess of 5% of the flow set point. Consequently, a maximum transient overshoot specification of 0.5 psi for regulator transient response is required to provide adequate pressure stability and ensure stable gas delivery.
Conclusion The low-flow transient response test described in this article can quantify the performance of a regulator at the start of gas delivery. This test enables process engineers to qualify a regulator for point-of-use semiconductor applications. In order to ensure adequate pressure stability and guarantee that the MFC used in these tests provided stable gas delivery, the regulator transient response overshoot specification could not exceed 0.5 psi. However, different MFCs‹even from the same manufacturer‹display varying levels of sensitivity to transient overshoot. Advances in regulator response must keep pace with faster-opening MFCs so that a fast-acting, stable gas delivery system can be established. By understanding the interaction of the various components used in a gas stick, the requirements for each component can be specified, ensuring that semiconductor manufacturers have access to the best possible gas delivery system. Acknowledgments The author would like to thank Warner Thelan, a test engineer at Parker Hannifin's Veriflo division in Richmond, CA, whose expertise in testing pressure regulators contributed to the substance of this article, and to Bill Stubbs, an applications engineer at Veriflo, who contributed to the article's development aspect. Dan Morgan is the research and development manager at Parker Hannifin's Veriflo division in Richmond, CA. He joined the company (formerly called Veriflo Corp.) in 1984 as a manufacturing engineer before being promoted to the position of design engineer. Morgan is an expert in high-purity regulator design for the semiconductor industry, an area in which he holds two patents. He received a BS in mechanical engineering from Washington State University in Pullman, WA. (Morgan can be reached at 510/412-1253 or danmorgan@veriflo.com.) MicroHome |
Search | Current Issue | MicroArchives Questions/comments about MICRO Magazine? E-mail us at cheynman@gmail.com. © 2007 Tom Cheyney |
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
