By Carl Remenyik ORNL has developed a calibrator to improve the accuracy of measurements of the flow of gases essential to the production of integrated semiconductor circuits for computers and other electronic devices. This device is designed for use with toxic and corrosive gases. ORNL has developed an instrument primarily for the calibration of gas flow controllers used in the semiconductor industry. This instrument collects and weighs the gas that flowed through the controller to be calibrated. The sensitive balance or load cell weighing the gas is relieved of the weight of the heavy cylinder collecting the gas by submerging it in water and offsetting its weight with the cylinder's buoyancy. This instrument received an R&D 100 award in 1995 from R&D magazine. Processes for manufacturing semiconductor devices used in televisions, radios, computers, and many other electronic devices require accurately controlled flows of many different gases into a vacuum chamber. Some of the gases used in industry are among the most corrosive and toxic, such as tungsten hexafluoride, silane, hydrogen bromide, chlorine, dichlorosilane, and boron trichloride. A plate or "wafer" inside the vacuum chamber serves as a substrate on which the semiconductor devices are formed. By various means, the molecules in the gases are caused to deposit in layers at selected locations on the surface of this substrate; to diffuse into existing layers to change their electric, thermal, or chemical properties; or to etch off some material from the surface. The final product is an integrated circuit chip for use in the many different electronic devices. As the speed and data-storing capacity of computers have increased over the past few decades, the required dimensions of circuit components were made ever smaller and the components were packed more densely on the wafers. At the same time, the wafers have been made bigger. These developments have increased the cost of producing finished wafers so much that their dollar values are now typically six-digit numbers for each wafer. It is not surprising then that manufacturers are very concerned about the rates at which rejects turn up among their products. To help reduce the reject rate related to improper gas flow, ORNL has developed a calibrator to improve gas flow measurements.   Calibrating Gas Flow Controllers   The quality of semiconductor circuits depends greatly on the accuracy with which the flow of gases into the vacuum chamber is controlled. The most frequently used instruments that regulate gas flows are mass flow controllers (MFCs). They measure and control the mass of the gas that flows through them. MFCs must operate with an error not more than 1% to make the desired reproducibility and quality of the product possible. Before MFCs can be used in fabrication systems, they must be calibrated. The calibration standard must be even more accurate than the device it calibrates. Experience teaches us that the calibration standard should not have an error greater than about 0.1% if a device needs to be calibrated with an error of 1%. Manufacturers and laboratories use various calibration methods. Several methods determine the rate of mass flow indirectly. With some, the gas pressure, temperature, and volume must be measured, and in addition, the gas constant and compressibility factor must be known. From these, the mass is calculated with a thermodynamic relationship. These methods suffer from the fact that each of the quantities entering the calculations contain some measurement error, affecting the results cumulatively. Furthermore, some of these methods cannot be applied to toxic or corrosive gases. Other indirect methods of calibration themselves require calibration by one of the other methods and, therefore, cannot be more accurate. Some methods of calibration weigh gases leaving or entering the calibrated device during a measured time. Measuring weight may be regarded as equivalent to measuring mass and, therefore, such gravimetric methods of calibration may be considered direct methods. In a typical procedure, a pressure cylinder is evacuated, weighed, and then connected through pipes and valves to the calibration setup with the device to be calibrated. During a measurement, the elapsed time is measured from the instant the valve is opened, when gas begins to flow into the cylinder, until the valve is shut. Then, the cylinder is disconnected and weighed again. The increase in weight is the weight of the collected gas. From the measured weight and time, the average mass flow rate of the gas streaming through the device and into the cylinder can be calculated. The measurement can be performed with an initially pressurized cylinder from which gas is discharged during the measuring process. Generally, the error of a precision balance is a percentage of the full range of the balance. When the pressure cylinder is weighed before and after the operation, the weight of the gas—the difference between the two weighings—is determined with an error corresponding to the total weight on the balance, which includes the heavy, steel cylinder. If the amount of the collected gas weighs only a small fraction of the weight of the cylinder, this error is too large a percentage of the gas weight. Because of this problem, the measurement procedure must continue until a sufficient amount of gas has been collected. At low flow rates, the collection may require many hours or even days. During all that time, the entire system has to operate steadily with very small drifts and oscillations of temperature, pressure, and instrument settings because, at the end, only the average flow rate over the entire time can be calculated, and, therefore, the accuracy of the calibration depends on the steadiness of the operating conditions. One such measurement determines just one point on the calibration curve. This method is time-consuming and cumbersome because the cylinder must be disconnected from the pipes of the calibration system for every weighing. This method introduces occupational hazards when the gas is toxic or corrosive, necessitating purging of all or part of the system before the pipe connection may be opened. The length of the required time makes calibration very costly, even for one gas. But mass flow controllers respond differently to different gases, necessitating separate calibration for each gas (which is prohibitively expensive). Some existing methods allow calculation of an approximate calibration coefficient for one gas from the results obtained for another gas, but these methods are not satisfactory.   ORNL's Contribution   The semiconductor industry needs improved calibration technology to overcome the shortcomings of existing methods, to improve the accuracy and speed of calibration, to make measurements involving highly reactive gases, and to avoid worker exposure to toxic gases, thus increasing the safety and efficiency of the industry's manufacturing operations. This need prompted an effort in ORNL's Instrumentation and Controls Division to develop a new instrument that measures the weight of the process gas directly, eliminates the need for disconnecting the instrument from the flow system during the calibration process, reduces the time needed for a measurement by orders of magnitude, may be used with the corrosive and toxic process gases involved in the fabrication of semiconductors, and increases accuracy tenfold over the accuracies of commonly used instruments. The new ORNL-developed gravimetric calibrator meets these criteria. The first prototype was made of a stainless steel, that allows the instrument to resist attack by the specified reactive gases. Our instrument is being used in ORNL's Mass Flow Controller Testing Laboratory, which was originally set up to carry out work for SEMATECH, a research consortium of the semiconductor industry, and is now a user facility available for industrial firms, universities, and other organizations. The patented instrument has been in operation since 1994, and it is the only instrument of its type known to exist. The co-inventor is James Hylton of the Instrumentation and Controls Division who collaborated throughout the development of the instrument. The idea behind this invention is to make an empty steel vessel seem almost weightless by submerging it in water to balance most of its weight by buoyancy. In other words, the water pushes upward against the 23 kilogram (50-pound) container enough so that it feels no heavier than one-tenth of an ounce. The submerged container is suspended from a type of scale called a load cell, an electrical device that can measure weight by sensing the elastic deformation of one of its components caused by the weight. It contains a strain gage, whose electrical resistance is increased the more it is stretched (thus decreasing the measured current passing through it by an amount related to increases in applied force). The one used here has a load-bearing limit of 50 grams and would instantly collapse under the container's weight. Therefore, the container is submerged in water and its weight is balanced by buoyancy. Tare weights ensure that there is a small load, 3 to 4 grams, on the load cell even when the container is evacuated. Without these small weights, buoyancy might take the container off the load cell. Fig. 1. Carl Remenyik checks a mass flow controller that feeds gas to the gravimetric gas flow calibrator that he developed for the semiconductor industry. Below is the submerged cylindrical vessel in which a corrosive or toxic gas can be safely weighed. Next to his left shoulder is the load cell from which the submerged vessel is suspended. For this device Remenyik received an R&D 100 award in 1995. Photograph by Bill Norris. Fig. 2. The gravimetric flow calibrator (1 = container, 2 = connecting capillary tube, 3 = vertical pipe connection, 4 = support frame, 5 = load cell, 6 = support ring for container, 7 = capillary vent pipe, 8 = support plate for load cell or balance, 9 = tare weight.. The main part of the gravimetric calibrator (shown in Fig. 1 and Fig. 2) is a closed, double-walled cylindrical container that is nearly 2 meters (6 feet) long. The space enclosed by the inner wall communicates with the outside through a capillary tube. This tube connects with a system of gauges, flow controllers, and remotely controlled valves, which can ultimately connect the container either to the instrument to be tested or to a vacuum pump. Or the valve can be closed, isolating the system. In preparation for a series of measurements, the container is first evacuated. The measurement of the gas flow rate through the device to be calibrated begins when a valve switches the flow from a by-pass into the container. The load cell continuously measures the increasing weight of the accumulating gas and sends electrical signals to a computer, which records the increasing weight 2000 times a second. The measurement may last seconds or minutes, depending on the magnitude of the flow rate, and it is terminated by switching the valve back to its initial bypass direction. The next measurement may follow immediately. As the accumulating gas applies an increasing weight on the load cell, the cell deflects slightly, lowering the container. Even the maximum displacement of the container is only a fraction of the thickness of writing paper, but it still twists the capillary tube, resulting in a small change of the force acting on the load cell. That is an error, but the capillary tube was designed and arranged so that the error caused is always more than 10,000 times smaller than the weight of the gas; because it is much smaller than the specified error, it is ignored. However, this small effect can either be calculated or measured, and the error caused by it in the results may be corrected, should that ever become necessary. The double walls of the container also affect the accuracy. When the container is being evacuated, its dimensions shrink a little even though it is made of steel. The force of buoyancy acting on an object in water changes if the volume the object occupies changes. When gas enters the container during a measurement, the pressure rises inside the container and expands it, increasing its buoyancy and causing a false change in the gas weight. This source of error is eliminated by inserting the container inside a shell that is large enough to leave a gap around the container. When this assembly is submerged in water, the buoyancy is determined by the outside dimensions of the shell, which remain unaltered by changes in the volume of the container inside. A second prototype of this instrument is now under development. Experience with the first instrument and changes in the semiconductor industry have prompted improvements. The planned second prototype is more versatile and much smaller than the first instrument, possibly compact enough to become a portable, table-top device. If the improved instrument works as planned, it will likely be adopted for use by the semiconductor industry. BIOGRAPHICAL SKETCH CARL REMENYIK is a consultant with ORNL's Instrumentation and Controls Division. He also has been a professor of fluid mechanics at the University of Tennessee at Knoxville. He made important contributions to the development of the cytriage (blood cell separator; winner of an R&D 100 award) in the 1960s, and he worked with Jim Hylton to develop the gas flow calibrator for which he received an R&D 100 award in 1995. He has a Ph.D. degree in fluid mechanics from Johns Hopkins University. He first began working at ORNL in 1965 when he joined the Reactor Division. Before coming to Tennessee, he worked for the Glenn L. Martin Company, which eventually became Lockheed Martin Corporation.