Effect of aggressive gases on capacitance manometers
10/01/2002
overviewCapacitance manometers are often used as pressure sensors on plasma etching and chemical vapor deposition tools where they are exposed to corrosive reactants and byproducts. Conventional wisdom has it that ceramic-diaphragm capacitance manometers are immune to such corrosive atmospheres. Recent analysis has shown, however, that ceramic diaphragms are less resistant than Inconel-alloy diaphragms.
On an Inconel-diaphragm manometer, all joints are welded without using brazing alloy or lead glass solders (Fig. 1) [1]. Inconel 600 is a high-nickel alloy (72% Ni, 15.5% Cr, 8% Fe) specifically developed to be corrosion-resistant. This alloy is resistant to a wide range of corrosive media and can be up to 10 times more corrosion-resistant than 316 stainless steel [2]. The chromium content gives better resistance than Inconel Alloys 200 and 201 under oxidizing conditions, and the high content gives good resistance to reducing conditions.
Inconel is virtually immune to fluorine and to chlorine ion stress corrosion cracking and is resistant to organic acids. Little or no attack occurs at room and elevated temperatures in dry chlorine or hydrogen chloride gases, and the alloy is also resistant to ammonia bearing atmospheres, nitrogen, and carburizing gases. At temperatures up to 550°C in these media, Inconel has been shown to be one of the most resistant of the common alloys [2, 3].
 Figure 1. An all-welded Inconel-diaphragm capacitance manometer has only one material in the wetted path. |
null
 Figure 2. A ceramic-diaphragm capacitance manometer, with five different wetted path materials. |
null
Ceramic-diaphragm gauges
A ceramic diaphragm is sealed between two ceramic parts that form the vacuum reference side (top) and process measurement side of the sensor (Fig. 2). Construction materials include aluminum oxide ceramic (typically 99.5% alumina) and a lead-containing solder glass used to bond the ceramic parts. The metallic portions of the device are Vacon 70 alloy, while a silver-copper brazing alloy is used in the ceramic-to-metal seal at the gauge inlet connection. Vacon 70 is a high-iron, low-thermal-expansion alloy (49% Fe, 28% Ni, 23% Co), similar to Kovar, which is designed for sealing to aluminosilicate glass and ceramics. It was not specifically developed for corrosion resistance.
Although pure aluminum oxide is corrosion-resistant, certain process chemicals, such as hydrofluoric acid or hydrogen fluoride gas plus moisture, will etch the lead-silicate glass used to mount and seal the diaphragm, as well as the ceramic itself.
Instability and drift
Capacitance manometer zero drift can be described as a variation in output signal compared to the original electrical-mechanical zero setting when measured at high vacuum. Many factors can cause zero drift in a capacitance manometer: corrosion, chemical- or ion-etching of the diaphragm, material deposition on the diaphragm, temperature variation, and repeated cycling to atmospheric pressure.
Innovations in the latest Inconel capacitance manometers address these factors. Corrosion-resistance was discussed above. Although ceramic has a lower thermal expansion coefficient than Inconel, specifications given for a ceramic-diaphragm manometer indicate a thermal stability similar to Inconel-diaphragm gauges. By adding thermal control, particle baffles (Fig. 1), and application-specific sensor configurations, material deposition on the diaphragm can be greatly minimized. The effects of overpressure on stability are described later.
 Figure 3. Zero drift vs. cumulative etch time for deep silicon trench etch cycles. |
null
To assess the reliability of capacitance manometers in semiconductor process tools, the zero stability of Inconel-diaphragm and ceramic-diaphragm sensors was compared after exposure to a silicon etch process that included nitrogen trifluoride (NF3) plasma clean cycles, after cycling to atmospheric pressure, and after exposure to vibration (as might be caused by turbomolecular or dry vacuum pumps).
Exposure to silicon etch
During silicon trench etching for shallow trench isolation, a capacitance manometer is exposed to various combinations of bromine- and fluorine-containing gases. We ran tests where the etch gas contained hydrogen bromide and the plasma clean step used oxygen and NF3.We plotted the change in the zero voltage level vs. cumulative etch time in RF-hours, which is the time that the manometer is exposed to an active plasma (Fig. 3). Data from two 1torr full-scale Inconel-diaphragm units are compared with data from one 1torr full-scale ceramic-diaphragm manometer, all with a 10V full-scale output signal. Over 1500 hrs, the zero level of the ceramic unit decreased by more than 50mV, equivalent to a 5mtorr drift in pressure reading. Simultaneously, on adjacent chambers, the two Inconel manometers drifted less than ±6mV, equivalent to ±0.6mtorr.
Exposure to plasma clean
Many CVD systems use an in situ RF or microwave plasma clean to remove deposited films on the fixtures and chamber walls. Single-wafer systems run a clean process after each deposition cycle and batch systems require a plasma clean after a fixed number of wafers are processed. For these cleans, it is common to use fluorine-bearing gases (NF3 or C3F8) that produce very reactive atomic fluorine species in a microwave discharge.
 Figure 4. Effect of cycling to atmospheric pressure on zero drift. |
null
We looked at the effect of a simulated NF3 plasma chamber clean cycle on the long-term zero stability of Inconel- and ceramic-diaphragm capacitance manometers. During 124 hours of exposure to NF3 plasma, the ceramic-diaphragm manometer required frequent re-zeroing (approximately hourly) to maintain process requirements. The unit ultimately failed as the output shifted by 11% of full scale, greater than the available coarse and fine zero adjustment ranges. The Inconel-diaphragm manometer performance was unaffected by exposure to chamber plasma clean cycle and did not require continued re-zeroing.
Cycling to atmospheric pressure
It is claimed that since ceramic is more rigid than metal, it is much less likely to deform or fatigue as a result of repeated pressure cycling or overpressure. In addition, compared to a metal diaphragm, a ceramic diaphragm should be better able to return to the same position for a given pressure, giving more repeatable measurements over time and eliminating the need for an isolation valve. Published specifications (see table) do not support these claims, however. For example, an Inconel-diaphragm capacitance manometer can withstand an overpressure of 45psia compared to an overpressure limit of 20psia for ceramic-diaphragm manometers.
In actuality, acute deformation and stretching of the Inconel diaphragm does not occur when cycling to atmospheric pressure. In an overpressure situation, the reference-side disk mechanically supports the diaphragm. The diaphragm never reaches its yield stress limit. Inconel-diaphragm manometers routinely withstand thousands of cycles from vacuum to atmospheric pressure, which is supported with data showing the effect of 1000 cycles between vacuum and atmospheric pressure on zero drift of a 1torr full scale ceramic-diaphragm manometer and three 100mtorr full scale Inconel-diaphragm manometers (Fig. 4). The ceramic-diaphragm gauge shows a significantly larger and variable zero drift.
Since ceramic diaphragms are reported to reach their final (equilibrium) position faster than metal ones, one would expect this to reduce the response time. Actually, aspects of the design other than diaphragm material have a greater effect on the response time or return-to-zero. The greater effects of gas dynamics and electronic measuring and filtering circuit delay-times mask the mechanical response time of the diaphragm. Our tests have shown the zero-recovery time for ceramic-diaphragm manometers to be no better than Inconel-diaphragm manometers.
Position and vibration
The ceramic manometer diaphragm has twice the mass/unit area of the Inconel diaphragm. This difference can be demonstrated by taking devices mounted in a vertical position and rotating both devices 90° to a horizontal position to remove the effects of gravity, then measuring the output voltage. The ceramic manometer typically produces an offset two times greater than the Inconel-based manometer. The increased mass effect increases the sensitivity to vibration.
In tests, ceramic-diaphragm and Inconel-diaphragm 100mtorr full-scale manometers were subjected to rms vibration levels in the range 0.2–0.5g and the effect of vibration on the signal output voltage and standard deviation of each manometer was noted. We found that even at low g-levels there is a response difference between the ceramic and Inconel units. The voltage output of the Inconel-diaphragm unit was up to four times less sensitive to vibration than the ceramic-diaphragm manometer.
Best for the application
We conclude that Inconel alloy diaphragm gauges are better suited for applications where they are exposed to such aggressive gases. In addition, since the ceramic diaphragm is stiffer and thicker, with more mass than an Inconel diaphragm, it is less sensitive to pressure changes and more sensitive to orientation, vibration, and shock. Inconel-diaphragm manometers showed greater immunity to zero shift caused by cycling between vacuum and atmospheric pressure.
Acknowledgments
Inconel is a registered trademark of Special Metals Corp.; Kovar is a registered trademark of CRS Holdings Inc.
References
- D. Jacobs, Vacuum & ThinFilms, pp. 30–35, Feb. 1999.
- Special Metals Corp., INCONEL alloy 600 data sheet, www.specialmetals.com.
- www.hpalloy.com/DataSheets/600.htm; "Selection of Compatible Materials for Use with Fluorine," NASA Guideline GD-ED-2206.
Bob Hyman received his BSEET from Northeastern University, Boston. He is product marketing manager at MKS Instruments Inc., 6 Shattuck Rd., Andover, MA 01810; ph 978/975-2350, fax 978/975-7663, [email protected].
 |
null