Integrated MEMS UHP pressure transducers
10/01/1998
Integrated MEMS UHP pressure transducers
Jim Hoffman, Dave Wagner, SenSym Inc., Milpitas, California
Andrew Dribnak, Veriflo Division, Parker Hannifin Corp., Richmond, California
Integrating a piezoresistive MEMS chip with a stainless-steel membrane produces a compact, ultra-high-purity pressure transducer. The combination of the proper materials and an optimum design improves both signal stability and reliability in microelectronic processing applications.
Pressure transducers are widely used in the semiconductor industry for ultra-high-purity gas (UHP) pressure measurement in process lines. Unfortunately, current UHP transducers rely on older technologies that have problems with zero and span drift, thermal shift, and case stress. Recalibrating to correct transducer errors requires ongoing maintenance that increases downtime and cost of ownership.
Solving these high-purity gas and liquid flow problems has been extremely difficult. The transducer must be designed to resist corrosive media and survive the effects of a harsh industrial environment. In addition, 5- and 10-micro in. (5- and 10-Ra) finishes are required to maintain the purity of the delivered fluid.
To overcome these problems, the Veriflo Division of Parker Hannifin Corp. and SenSym formed a joint venture to manufacture a new MEMS-style, micromachined, silicon-based transducer that minimizes zero drift, thermal shift, and case stress (see photo at right and Fig. 1).
|
A new, MEMS-style, micromachined, silicon-based UHP pressure transducer minimizes process downtime and instability. (Photo courtesy of Bill Gutoff)
|
Figure 1. MEMS-style, micromachined, silicon-based transducer.
Transducer requirements
A pressure transducer develops an electrical signal that is proportional to the pressure applied. An ideal pressure transducer design minimizes the effects of temperature, mechanical stress, and aging on the output signal so as not to compromise the semiconductor fabrication process. In actuality, transducers respond to environmental effects, and only some of the effects can be easily compensated through signal conditioning. Others are impossible to correct and add unpredictable variation to pressure measurement.
Correctable errors are typically proportional to temperature or pressure. Transducers can have linear and nonlinear (typically second-order) temperature effects for both zero-pressure output and span. The raw span and offset signals with respect to pressure may also require correction and can be nonlinear. These repeatable errors can be readily corrected by the transducer manufacturer using a variety of analog or digital compensation schemes.
In classic analog compensation, linear errors are routinely compensated with passive or active DC circuits. The second-order errors, while more difficult, can be handled by nonlinear analog feedback via temperature sensors (thermistors) or by digital correction. As long as errors are repeatable, they are correctable and can be eliminated with proper signal conditioning.
While correctable (repeatable) errors are somewhat problematic for the transducer manufacturer, it is the noncorrectable
(unrepeatable) errors that cause concern for the transducer user. Environmental effects, such as humidity, that can affect the accuracy of a transducer can be unrepeatable, and are much more difficult or impossible for the transducer manufacturer to predict and correct. Examples of noncorrectable effects include case stress susceptibility, hysteresis, creep, and long-term stability.
An example of case stress would be a shift in the offset of the device due to mounting-/installation-related stress. The degree of stress applied is different in each application and is hard to predict. If the raw sensor element and package are not designed to reject such stresses, it will be impossible for the manufacturer to eliminate the offset instability caused when the user mounts the sensor in the field. Creep, hysteresis, and long-term offset shift are three other effects that can be time, temperature, humidity and/or pressure related and impossible to eliminate. If a transducer exhibits any of these characteristics, it will require frequent calibration that always adds process variation, unpredictability, and replacement cost, and possibly even downtime.
High-purity transducer applications involve stringent material control and manufacturing techniques to ensure fluid purity of semicorrosive liquids and gases. Gas delivery lines are often welded into place, which makes component replacement difficult and costly, and also induces welding stress that can cause transducer errors.
High-purity chemical processing, whether for semiconductor fabrication or any other application, inherently involves highly corrosive environments and requires an absence of chemical and particulate contamination. Any surface exposed to process gases and liquids must be resistant to corrosion, and maintain smooth, non-particle-retaining surfaces. Any transducer used in the system must also adhere to these requirements.
Standard transducer designs
The high-purity transducer market has traditionally used capacitive- and Wheatstone-bridge-based technologies in the fabrication of pressure transducers. Each technology has advantages and disadvantages, and selection has always required certain compromises.
Capacitive technologies rely on one or more semiparallel plate capacitors, which have an air gap that changes with applied pressure. Typically, a stainless-steel diaphragm both isolates the media and functions as one plate of the capacitor; this design has the advantage of low, uncompensated, temperature errors. It has been used in low-pressure absolute reference sensors, as the technology is more stable with vacuum-sealed plates. Gauge or differential sensors of this design have offset instability due to humidity changes between the plates.
The main disadvantage of capacitive transducers is that they require large diaphragms with large displacements to produce reasonable signals. Large diaphragms and volumes tend to add unswept dead volumes that can trap particles and increase purge cycle times. Even with large plates, however, the actual change in capacitance is so small (on the scale of pF) that extensive signal conditioning is required to amplify and filter a signal to a usable level. Signal conditioning can be particularly difficult in electrically noisy applications, since one of the plates is electrically connected to the process tubing.
Metal-foil Wheatstone-bridge designs use strain-sensitive resistors strategically placed on a diaphragm or beam, such that the resistance increases proportionally to the strain change. Attached with an adhesive, the gauges (typically four) are arranged into a fully active bridge with two increasing and two decreasing in resistance. The Wheatstone-bridge design adds "some" intrinsic compensation, as each leg of the bridge reacts similarly to temperature. Metal-foil gauges are constructed of metallization on a carrier (typically polyimid).
Metal-foil gauge manufacturing is labor intensive and produces a large distribution of performance due to the fact that each gauge must be glued in place by hand. The adhesive that holds the gauges in place, like any polymer, will soften and "creep" with time and pressure. Any relaxation of the adhesive will cause a change in transducer output. In both technologies (capacitive or Wheatstone-bridge based), the stability of the sensor depends on the stability of the sensor attachment method and the stability of the measured structure (diaphragm). Both gauge technologies are easily compensated and have reasonably good signal-to-noise characteristics.
MEMS-based design
MEMS-based transducers generally rely on piezoresistive strain gauges to translate pressure into electrical signals (see "Silicon MEMS transducers" on p. 124). The piezoresistive effect is based on the change in the mobility of charge carriers in a resistor due to a change in mechanical stress (thus changing the resistance). Piezoresistive bridges typically produce over 60 times more signal than foil gauges for the same applied pressure.
Silicon MEMS transducers
Silicon micromachining uses both isotropic and anisotropic etchants to sculpt 3-D structures in silicon. Commonly used anisotropic etchants include potassium hydroxide (KOH), tetramethylammonium hydroxide (TMAH), and ethylene diamine-pyrocatechol (EDP). Etch rates vary depending on the exposed crystal planes, allowing for precise control of silicon diaphragm dimensions based on the tolerances of the lithography. Etch rates in the <100> plane are typically between 50 and 140 times the etch rate in the <111> plane. Thin silicon diaphragms can be created with well-defined dimensions bounded by <111> planes.
Isotropic etchants include hydrofluoric (HF)/Nitric (HNO3) mixtures and offshoots. Xenon difluoride (XeF2) is also a popular dry isotropic etchant. All of these silicon etchants must be compatible with the materials on the wafer such as silicon dioxide and silicon nitride. Hence, micromachining is often done as early in the process as is feasible.
Although there are ways to lap and polish an entire wafer to piezoresistive diaphragm thickness, handling exremely thin dice would be difficult without very specialized equipment. However, selectively thinning part of the wafer while leaving enough thick areas to maintain mechanical strength is well within the capability of MEMS technology. Sensor dice can be fabricated wholly within thinned areas. Diaphragms on chips can be as thin as 20 ?m, with control to ?0.4 ?m.
Electrical circuitry on the chip is handled through standard bipolar processing. Batch wafer fabrication leads to low-cost manufacturing with thousands of sensors/wafer. Resistors are implanted together at the wafer scale using submicron photolithography techniques, such that alignment of the resistors with respect to each other is nearly perfect.
A joint venture between Veriflo and SenSym has produced a MEMS-based pressure transducer family for the high-purity market (the EXACT Series). To date, MEMS-based transducers have used an oil-filled interface to transfer the pressure to the chip in corrosive applications. However, since the high-purity market requires a media-isolated transducer that will not introduce contaminants (no oil-filled transfer fluids), a completely new structure was required.
The new design incorporates a fully integrated Wheatstone-bridge silicon pressure sensor bonded to a Hastelloy C-22 stainless-steel diaphragm. Due to the mismatch in the thermal coefficients of expansion between the Hastelloy and silicon, the chip must be extremely small. The correct physical dimensions must be tightly controlled to minimize mechanical creep in the transducer.
As corrosive media pressurizes the chemically inert Hastelloy diaphragm, it deflects. Diaphragm deflection induces strain in the sensor chip, and the chip`s piezoresistors change resistance in response to the induced strain. A resultant electrical signal is thus proportional to pressure. Using case stress isolation techniques, stress signals due to installation and vibration are rejected. In addition, the case stress is effectively reduced by strategically placing the piezoresistive sensor die so that non-pressure-related signals are effectively canceled. Using a clean signal to start with greatly simplifies the signal conditioning, and easily allows for a choice of output signals.Hastelloy nickel-based alloy is extremely chemically inert. Material evaluation indicated a 50% improvement with the use of Hastelloy in standard corrosion tests over 316L Stainless Steel. (Ferric chloride is typically used to evaluate stainless steels.) In addition to its superior corrosion resistance, Hastelloy`s coefficient of expansion and modulus of elasticity are more closely matched to those of the sensor die and the die-attach medium. The sensor die is bonded to the diaphragm using a high-temperature Thermo-Fused process that results in greatly improved long-term stability and reliability of the sensing elements. In addition, the difficult-to-correct errors of creep, hysteresis, and long-term offset shift are practically eliminated (Figs. 2 and 3).
|
Figure 2. Creep test at elevated temperatures (70?C), EXACT Series vs. competitors
|
Figure 3. Room temperature drift test (two weeks), EXACT Series vs. competitors.
Conclusion
The use of MEMS components and Thermo-Fused die-attach technologies produces a more stable, accurate pressure transducer for ultra-high-purity applications. Recalibration issues and case stress sensitivity are virtually eliminated. The new pressure transducers are capable of withstanding the extreme environmental conditions inherent in semiconductor manufacturing processes, without requiring the continuous maintenance needed by other designs.
Acknowledgment
Hastelloy C-22 is a registered trademark of Haynes International Inc. Thermo-Fused Technology is a patent pending of SenSym Inc.
JIM HOFFMAN has designed sensor-based integrated circuits and MEMS processes for 18 years, working primarily with pressure transducers and accelerometers. A former employee of National Semiconductor, he is presently the director of wafer fab engineering at SenSym Inc., a division of BTR. E-mail Hoffman at [email protected]
DAVE WAGNER has worked in MEMS-based sensor design and manufacturing for 15 years, and was formerly employed at EGG IC Sensors, where he manufactured accelerometers and pressure sensors. He is now director of engineering at SenSym, a division of BTR. E-mail Wagner at [email protected]
ANDREW DRIBNAK is currently corporate director of product management at Veriflo. He has also served on SEMI and SEMATECH committees for gas delivery systems since 1994. Veriflo Div., Parker Hannifin Corp., 250 Canal Boulevard, Richmond, CA 94804; ph 510/412-1143, fax 510/232-7396, e-mail [email protected].