Cylinder, purifier technologies for controlling contamination in CO
07/01/2002
By Carrie Wyse, Joseph Vininski, Tadaharu Watanabe, Matheson Tri-Gas Inc., Advanced Technology Center, Longmont, Colorado*
*Additional authors are listed in the Acknowledgments.
Overview
Methodical research has shown that a newly designed aluminum cylinder with a brass valve significantly reduces nickel and iron contamination in carbon monoxide. This was confirmed by dry etching wafers with additive CO from various types of cylinder packages. In addition, a new point-of-use inorganic purifier material can effectively remove residual metal carbonyls and moisture generated in a CO distribution system.
Carbon monoxide (CO) is a key component of the gas mixture used for selective etching of oxide films deposited over silicon nitride. It forms an oxygen-rich polymer layer at the oxide-nitride interface. This reduces the etch rate of silicon nitride and improves the selectivity of the process. CO is also used for silicon dioxide contact and via etch.
CO purity directly impacts overall process yield and quality. For plasma etch processes, selectivity and etch rate must be maintained at particular levels by precisely controlling process parameters and gas mixture concentrations. Metal contamination in CO is particularly detrimental because metal ions can be deposited onto an etched surface and can then diffuse during subsequent steps. This directly impacts the device's electrical characteristics, such as resistivity. Moisture contamination is also detrimental as it is an uncontrolled source of oxygen that affects process parameters.
It is well established that iron pentacarbonyl (Fe(CO)5) and nickel tetracarbonyl (Ni(CO)4) can form via gas-solid reactions under high-pressure CO [1], such as in a compressed gas cylinder. Fe(CO)5 and Ni(CO)4 are both relatively volatile complexes, and can therefore be transmitted in the gas phase throughout process-gas lines, ultimately contaminating devices being fabricated. CO is also known to contain low-level moisture.
To enhance contamination control in processes using CO, we have looked at an improved cylinder package and a point-of-use (POU) purifier, both designed to remove metal carbonyls and moisture.
Hydrolysis studies
Our Ultraline cylinder is iron- and nickel-free, specifically designed to minimize iron and nickel carbonyl contamination by eliminating steel- and nickel-wetted parts [2]. The new package design consists of an aluminum cylinder, a brass valve, and a pressure relief device containing a copper rupture disk in place of the standard nickel disk. Using a hydrolysis-sampling method, the new cylinder design was tested for metal contamination alongside 1) a standard carbon steel cylinder (alloy 4130) outfitted with a brass packed valve (commonly used in the gas industry); and 2) an aluminum cylinder (alloy 6061) outfitted with a 316L stainless steel cylinder valve having a diameter index safety system outlet.
Hydrolysis samples were collected from each cylinder by drawing 3.5kg of CO through 18MΩ water at 1 liter/min. These samples were analyzed using graphite furnace atomic absorption spectrometry (GFAA) for iron and inductively-coupled plasma mass spectrometry (ICP-MS) for other elements. Although the hydrolysis-sampling method has not been shown to be quantitative for metal carbonyl sampling, it is assumed that the amount of metals captured by hydrolysis is proportional to their concentration. Therefore, the method is useful for comparison. The average metals measured from two cylinders of each package were tabulated (Table 1).
 |
CO from the aluminum cylinder with a brass valve contained 1400x less iron than the carbon steel cylinder and 78x less iron than the aluminum cylinder with stainless steel valve. This trend of iron contamination measured in the CO directly corresponds to the relative iron levels found in each package and the relative rates of formation.
Nickel contamination in CO from the aluminum cylinder with a stainless steel valve was 6x greater than that found in the other two cylinders with brass valves. This is easily explained because stainless steel is 12% nickel.
The zinc levels were approximately 2x greater in the aluminum cylinder with a brass valve compared to the aluminum cylinder with a stainless steel valve, and 7x that in the carbon steel cylinder. Although zinc is a component of brass, there is no correlation between the brass content of the package and the zinc levels found in the CO.
Copper levels from all three cylinders were sub-ppb, and there is no correlation of copper contamination and cylinder materials.
Aluminum contamination was also at sub-ppb levels, but was slightly higher in the CO from the aluminum cylinder with a brass valve. However, no direct reactions between CO and solid aluminum, zinc, or copper were found in the literature, and based on their thermodynamic data, the formation of these compounds is not expected. Therefore, the presence of aluminum, zinc, and copper in the hydrolysis samples is not likely to be a result of reaction between the CO and the components, but rather a result of particle shedding.
Overall, the aluminum cylinder with a brass valve package, developed specifically for containing CO, reduces iron contamination by three orders of magnitude, and nickel contamination by a factor of five, according to wafer studies.
Wafer studies
Further investigation showed that the dramatic improvement in CO purity from the aluminum cylinder with a brass valve produces higher-quality silicon [3]. This was proven in a study in which CO was combined with a fluorocarbon etchant to plasma-etch clean silicon. Unetched control wafers were compared with wafers etched with CO from a steel cylinder with a stainless steel valve and nickel rupture disk, and CO from an aluminum cylinder with a brass valve. The wafers were analyzed for surface metals contamination via total reflectance x-ray fluorescence (TXRF). The data showed a dramatic reduction in metal contamination on the wafers prepared from CO packaged with the materials not containing iron- or nickel-wetted parts compared to the package containing an abundance of iron and nickel (Table 2).
 |
POU purification
Even when the highest grade of gas available is used in a particular process, impurities can still contaminate a system downstream of the cylinder. For example, moisture can enter a distribution system during cylinder change-out or inadequate purge procedures, and metal carbonyls can be generated in situ by the reaction between the CO matrix gas and iron- and nickel-containing components (i.e., stainless steel tubing, particle filters, Ni gaskets, mass flow controllers, etc.). The predominant material of choice for gas distribution tubing is normally 316L stainless steel, which contains 65% Fe and 12% Ni.
A gas may exit a cylinder with sub-ppb levels of volatile metal complexes and moisture, but the concentration of these contaminants will certainly increase as the gas progresses toward the tool. So POU purification is essential to maintain the integrity and consistency of an ultrapure process by removing the metal and moisture contaminants within the process gas stream.
The recently developed MetalX purifier removes both moisture and Fe(CO)5 from CO [4]. This technology should also remove Ni(CO)4, since its chemical and physical characteristics are similar to that of Fe(CO)5. (MetalX is an inorganic purifier material that selectively removes metal complexes, in addition to moisture, from various process gases including corrosive ones. It is chemically robust with respect to corrosive environments, and is supplied in an enhanced package designed to minimize exposure of the matrix gas to iron- and nickel-wetted parts.)
Fourier transform infrared (FT-IR) spectroscopy experiments were conducted to test this new purifier technology for the removal of Fe(CO)5 in CO. Due to the instability of Fe(CO)5, calibration standards are not trustworthy, so it was not feasible to generate quantitative calibration curves for Fe(CO)5 in matrix gases. However, using spectroscopic data [5] and assuming both Beer's Law and the absence of a matrix effect, it was possible to obtain semiquantitative estimates of the level of Fe(CO)5.
A moderate challenge of Fe(CO)5 was generated in CO gas by flowing 350sccm of CO over frozen Fe(CO)5. When Fe(CO)5 levels were monitored using the peak height of the most intense absorption band at 2013cm-1, and using a semiquantitative absorption coefficient, the concentration of Fe(CO)5 was calculated to be 4ppm. (There is overlap between the Fe(CO)5 absorption band at 2013cm-1 and CO absorption. Thus, the background spectrum of the CO gas was subtracted from each sample spectra, revealing Fe(CO)5 absorbance.) The challenge gas stream was then directed into our purifier or a bypass, followed by the FT-IR spectrometer. The spectrometer was equipped with a 10m-pathlength cell and a liquid-nitrogen-cooled MCT-A detector. With a 4ppm Fe(CO)5 challenge, no Fe(CO)5 was detected at the outlet of the purifier (Fig. 1). Thermal analysis of the exposed material showed no emission of iron carbonyl, indicating a strong interaction with the purifier medium.
Capacity studies in a nitrogen matrix were also conducted using a high challenge of Fe(CO)5. This was generated by combining 30sccm of N2 that was bubbled through Fe(CO)5 liquid at 0°C, with 450sccm N2. Although the IR signal was saturated at the bypass due to the high concentration of Fe(CO)5, quantitative analysis of the scrubber solution indicated there was ~700ppm Fe(CO)5 in the challenge. When our purifier was challenged with this high level of Fe(CO)5, none of it was detected at the purifier outlet until the purifier reached the saturation point for removing Fe(CO)5. At this point, the purifier reached the maximum capacity for impurity removal and could no longer remove Fe(CO)5 to low-level concentrations. Based on typical metal carbonyl levels found in CO, a 300mL MetalX purifier would last several years.
Moisture removal
The MetalX purifier was also quantitatively tested for moisture removal in N2 using atmospheric pressure ionization mass spectrometry (APIMS) and a calibrated moisture generator. When the purifier was challenged with 2ppm H2O at 2slpm for more than 20 hrs, the efficiency was <1ppb (Fig. 2). Even at a challenge of 7ppm H2O and a flow rate of 1slpm, the efficiency was still <1ppb. Capacity tests using FT-IR spectroscopy show that a 300mL purifier would last several years based on the typical moisture levels found in CO.
 Figure 2. Moisture removal purifier efficiency at different challenge levels and flow rates, measured by APIMS. |
The moisture performance of the purifier in N2 is expected to be similar to that in CO because both matrix gases have little chemical interaction with the purifier medium. This is supported by the fact that when the Fe(CO)5 experiments were conducted in CO, the moisture from the CO cylinder was removed by the purifier to levels below 50ppb, the detection limit of FT-IR spectroscopy.
Conclusion
Potential impurities in CO include moisture, Fe(CO)5, and Ni(CO)4. Iron and nickel carbonyls are generated in CO cylinders and distribution systems by reaction of CO with any iron- or nickel-containing components in the system. We have shown that a newly designed aluminum cylinder with brass valve package, containing no iron- and nickel-wetted parts, reduces nickel and iron contamination in CO by one and three orders of magnitude, respectively, compared to the conventional carbon steel package. These results were confirmed by dry-etching wafers with additive CO from various types of cylinder packages.
Further, a new POU purifier technology effectively removes any additional metal carbonyls and moisture generated in the CO distribution system. FT-IR spectroscopy studies demonstrate that this technology has significant capacity for Fe(CO)5 and moisture, with a moisture efficiency of <1ppb.
Acknowledgments
Additional authors include Rob Torres, Mark Raynor, and Virginia Houlding. We thank Ehrich Diede of Diede Precision Welding for the welding and assembly of manifolds. MetalX is a trademark of Matheson Tri-Gas Inc.
References
1. D. Shriver et al., Inorganic Chemistry, W.H. Freeman and Co., 1990.
2. P.C. Andersen et al., Semiconductor International, April 1998, p. 127.
3. G. Cooper et al., Semiconductor International, July 1997, p. 301.
4. C. Wyse et al., CleanRooms, June 2001, p. 7.
5. R.K. Tepe et al., Spectrochimica Acta Part B, 55, pp. 165-175, 2000.
Carrie Wyse received her BS in chemistry from Northern Arizona University and MA in chemistry from Princeton University. She is a research chemist at Matheson Tri-Gas Inc., Advanced Technology Center, 1861 Lefthand Cr., Longmont, CO 80501; ph 303/678-0700, fax 303/442-0711, e-mail [email protected].
Joseph Vininski received his BA in chemistry from St. Olaf College. He is a research chemist at the Matheson Tri-Gas Advanced Technology Center.
Tadaharu "Ted" Watanabe received his BS and MS in applied chemistry and PhD in materials science from Yamanashi University. He is a visiting senior research scientist at the Matheson Tri-Gas Advanced Technology Center, on leave from Nippon Sanso Corp., Tokyo, Japan.