The changing art of measuring vacuum pressure
06/01/2004
Semiconductor manufacturing applications require measurement of a wide range of pressures to ensure proper vacuum conditions. Unfortunately, major process steps — such as CVD, PVD, and epitaxial applications — all have different working pressures and temperatures, and therefore, no single measurement principle can be applied to vacuum gauging in tools. This article summarizes trends in vacuum-pressure instrumentation, including common pitfalls in the application of pressure sensors and the emergence of active combination gauges.
Manufacturing semiconductors requires repeated cycles of coating and etching to build up the many layers that make up modern ICs. Semiconductors are fabricated under vacuum to reduce the risk of contamination from unwanted particles and residual gases. Typical vacuum pressure varies in wafer fab equipment depending on the process. For example, physical vapor deposition (PVD) usually occurs in the range of 5×10-3 to 0.01mbar, while chemical vapor deposition (CVD) is typically between 0.1 to 5mbar (Fig. 1). The loadlock operation will cycle between atmosphere pressure and ~100mbar, however, while the process chamber's background pressure in state-of-the-art processes is on the order of 10-10mbar.
 Figure 1. Typical ranges of vacuum-pressure measurement techniques and semiconductor-related processes. |
In general, many problems in the application of vacuum-pressure gauges for chipmaking stem from process contamination of the sensors. Initially conductive surfaces in sensors may become insulators, thereby causing a sensitivity reduction. This problem may eventually render gauges inoperable. Some chemicals are also very aggressive and may etch thin filaments and wires.
A variety of gauge designs are aimed at a range of vacuum applications with various levels of performance and vulnerability to certain operational conditions (see "Gauging the gauges for vacuums," p. 80). For example, the capacitance diaphragm gauge (CDG) measures gas type independent from total pressure. In clean vacuum conditions, CDGs are extremely stable, and serve as pressure transfer standards. On semiconductor tools, CDGs often show a drift, which can be the result of process by-products building up on the diaphragm. This results in uneven tensioning and weighting of the diaphragm, causing changes that are typically seen as a zero drift. Eventually the drift is larger than can be adjusted with the zero adjustment, and the gauge must be replaced.
The drift is both process-specific and specific to the particular design of a CDG. Certain types of CDGs are better for a particular process. Most CDGs use Inconel alloy as diaphragm material, which is useful for some processes, but for some others, alumina diaphragms prove to be more corrosion-resistant and are less affected by drift. Manufacturers try to minimize contamination (from process by-products) by heating the surfaces in contact with the process gases to prevent process gas condensation on the diaphragm.
The vacuum gauge industry is also trying to cope with the contamination problem by introducing special variants of gauges designed for a particular process. In these specialty gauges, certain materials are exchanged with less corrosion-prone materials for a particular process. For example, in Pirani gauges, either nickel or platinum filaments are substituted for the tungsten filament to provide longer sensor life in corrosive environments.
Gauge mounting on the process chamber also can trigger problems. Long, narrow tubing often is used to mount the gauges to the process chamber, resulting in reduced conductance of the gas through the tubing. The effect is that either it takes longer to reach an equilibrium pressure at the gauge, or the measured pressure does not correspond to the actual pressure in the chamber. Large errors in pressure readings can be introduced in this way.
As a general rule, the connection to the chamber should be as short as possible, and should have as large a cross-section as is practical. If the gauge is mounted near the pump, the pressure in the gauge may be considerably lower than in the rest of the system. Conversely, if the pressure gauge is placed near a gas inlet, the pressure indicated by the gauge may be much higher than in the rest of the system. In addition, certain gauges, such as the convection-enhanced Pirani model, must be mounted in a horizontal position to avoid erroneous pressure readings above 1.3mbar.
No single solution
There is no single pressure-sensing technique with a dynamic range that can measure from atmosphere down to ultrahigh-vacuum conditions. Therefore, measuring a wide pressure range requires several gauges, typically one flange and one controller per gauge.
Over the years, the trend had been to use one gauge controller to serve several gauges. Then digital gauges appeared on the market, integrating signal processing on the gauge itself, which eliminated the need for a separate controller and allowed easier interfacing with a computer. Now there are a handful of digital interfaces available, including RS232, RS485, DeviceNet, and Profibus.
More recently, individual measurement principles have been combined in one housing, thereby reducing the number of flanges needed at the vacuum chamber. This lessens the need for welding and leak testing on the process chamber, saves space, reduces the number of power and computer interface cables, and minimizes interface boards at the computer to reduce code complexity in the process control software and save time in acceptance testing of process chambers.
New "combo" gauges
One of the latest combination gauges is the TripleGauge (Fig. 2), which was launched last year by Inficon. In a single housing, this design combines a Bayard-Alpert gauge, a Pirani gauge, and a CDG, as well as an atmospheric-pressure sensor. This combination vacuum gauge has a built-in atmospheric switch for precise venting to atmosphere, which is useful in loadlock applications. The gauge allows measurement from 5×10-10 to 1500mbar, and Inficon believes it is the closest example yet of a single system that can measure all vacuum-pressure levels in a process application.
Recently, vacuum-pressure gauges based on silicon microelectromechanical system (MEMS) technology have appeared. It is not quite clear that these gauges are of great benefit to the semiconductor industry. An uncoated silicon pressure sensor on a vacuum chamber that is used to etch silicon will not last long, since the etchants will not only etch the wafer but will also etch the Si-MEMS sensor. Additionally, small MEMS structures are likely to be more prone to coating of the sensor in CVD or PVD applications than larger macroscopic structures. The Si-MEMS sensor may find application only in a relatively clean environment that is free from outgassing corrosive process gases.
As semiconductor manufacturers shrink feature sizes and introduce new processes, the vacuum control industry is also attempting to adapt concepts to new requirements. But one aspect in vacuum-pressure gauge development remains unchanged: to constantly improve repeatability and accuracy while lowering the cost of ownership.
Acknowledgments
TripleGauge is a trademark of Inficon. DeviceNet is a trademark of the ODVA organization. Inconel is a trademark of the Special Metals Corp. group of companies.
Martin Wüest is an R&D project manager for vacuum control at Inficon Ltd., Alte Landstrasse 6, FL-9496 Balzers, Liechtenstein; ph 423/388-3270, fax 423/388-3728, e-mail [email protected].
Gauging the gauges for vacuums
There are a variety of techniques to measure a vacuum, with pressure being defined as force/area. Unfortunately, no single method can measure from base pressure to atmosphere in semiconductor processes. Here is a comparison of a half-dozen types of commercially available measurement techniques, which are often used in vacuum gauges on semiconductor tools.
Capacitance diaphragm gauges
The capacitance diaphragm gauge (CDG) measures the pressure directly, according to the definition of pressure. The CDG can measure gas type independent from total pressure. The CDG consists of a very thin diaphragm mounted over a reference vacuum cavity. Under pressure load, the diaphragm is elastically bent. Therefore, at atmospheric pressure, the diaphragm is more deflected than in vacuum. The reference vacuum side of the diaphragm and the vacuum side of the base of the vacuum cavity are metallized to form a capacitance. The changing width between the diaphragm and base is a measure of the applied pressure, which is read out as a change in capacitance.
With the CDG pressure-sensing principle, total pressure can be measured from above the atmospheric pressure to approximately 10-5mbar. An individual sensor has a dynamic range of only 4 decades, however, which means that at least two sensors are needed to cover the pressure range from 10-5 to 103mbar.
Piezoelectric strain gauges
Piezoelectric strain gauges are quite similar to CDGs; however, the deformation of the diaphragm is measured by strain sensors. Sensitivities and dynamic range tend to be less than those of CDGs, but piezoelectric strain gauges are easier to miniaturize and tend to be lower in cost. Typically, these sensors are used in the pressure range from 1mbar to 50bar or higher.
Bayard-Alpert hot ionization cathode gauges
Hot ionization cathode gauges have a hot filament that thermally emits electrons, which ionize the residual gas in the sensor volume. The resulting ions are then collected at an anode. This ion current and the emission current are measured, with the resulting ratio proportional to the residual gas density. Using an ideal gas equation, the density is converted to a pressure value.
Since ionization is gas-species dependent, the hot ionization cathode gauge displays a gas-type dependent pressure. The sensors are usually calibrated with nitrogen gas and use correction factors to convert the pressure to the true pressure for another gas. Bayard-Alpert gauges are used in the pressure range between 10-11–10-2mbar. At higher pressures the filament burns out, and at lower pressures the electron-stimulated desorption presents the measuring limit.
Cold cathode ionization gauges
Similar to hot ionization cathode gauges, the cold cathode ionization gauge ionizes the residual gas; however, the method of generating the electrons is different. In a cold cathode ionization gauge, the electrons are generated in an electric discharge. A magnetic field forces the electrons on a longer, helical trajectory, thereby increasing the ionization probability.
Cold cathode ionization gauges are rugged and simple. This type of gauge may have difficulties starting a discharge at very low pressure. These gauges are used in the vacuum range from 10-11–10-2mbar.
Pirani gauges
The Pirani gauge uses thermal conductivity via hot wire to take measurements. As the pressure changes, the thermal conductivity changes, which causes a temperature change in the hot filament. The electronics measure the voltage required to keep the wire at the same temperature. Modern gauges have expanded the measuring range of the original Pirani gauge by enhancing the convection, different operations of the gauge, and temperature compensation.
Since the heat conductivity is gas specific, these gauges are also measuring a gas-type dependent pressure. Convection-type Pirani gauges should only be mounted in a horizontal position to obtain their specified accuracy. Properly designed Pirani gauges measure from 10-4mbar to atmosphere.
Quartz-crystal oscillator
In this technique, pressure changes cause a change in damping of the quartz crystal due to friction between the quartz surface and the gas molecules. Since the viscosity is also gas-specific, the pressure measured is gas-type dependent.
Modern quartz-crystal oscillator gauges use tuning fork designs and have a pressure measurement range from 10-4mbar to atmosphere.