by Robert Donovan
|
Is there a way to monitor contamination control's newest, most deceiving foe?
Airborne molecular contaminants (AMCs) are non-particulate contaminants that easily penetrate conventional HEPA and ULPA filters. Like particles, AMCs consist of molecules, or clusters of molecules. The major physical distinction between AMCs and aerosol particles is that AMCs are vapor-phase contaminants composed of fewer molecules and less total mass than aerosol particles, making AMCs too small to be detected by conventional particle counters. A functional definition of an AMC is any airborne contaminant not detected by a particle counter.
SEMI standard F21-95[1] subdivides AMCs into four categories: acids, bases, condensables and dopants. The table reproduces the classification designations presented in the F21-95 standard. Like particle standards, the class designation for each type of molecular contaminant specifies the maximum allowable concentration of that contaminant for the cleanroom to meet the requirements for that specific designation. For example, MA-10 means that the maximum allowable concentration of acidic molecular contaminants is 10 parts per trillion (ppt) calculated on a molar basis.
Similarly, an MC-100 designation means not more than 100 ppt of condensable molecular contaminants and so on. While each designation can be used alone, a complete description of cleanroom AMC quality consists of four independent designations, one for each category of AMCs.
Why are AMCs important?
On a mass basis, the concentration of AMCs in a typical cleanroom exceeds that of aerosol particles by orders of magnitude; similarly, AMC flux to a surface in this typical cleanroom is orders of magnitude greater than that of aerosol particle flux.[2] So why has the semiconductor industry primarily emphasized the control of aerosol particle concentrations in wafer environments while only lately considering AMCs, which have always existed in much greater numbers and mass in the cleanroom environment?
A partial explanation is that it has taken longer to recognize the dangers of AMCs compared to those of particle contamination. Consider that virtually all wafer processes include wet chemical cleaning steps and heat cycles. These steps and cycles remove many deposited AMCs and often reduce their concentrations below those levels that adversely affect product performance or manufacturability.
Particle penetration through fibrous filters. [5] |
However, this favorable arrangement does not always prevail. Perhaps the earliest demonstration of AMC impact and importance was that seen in the interactions of airborne amines with photoresists.[3] Exposure of resist- coated wafers to an ambient atmosphere containing such amines degrades photoresist and compromises dimensional control of the resulting printed patterns. This specific interaction is, however, just the beginning of a lengthy list of now documented, objectionable AMC interactions on semiconductor surfaces such as suppression of oxide growth, degradation of oxide quality, corrosion of metal layers, chemical doping and alteration of interfacial electrical properties.[2, 3]
AMC-induced product degradation is now a well-established phenomenon and is likely to become more serious as chip-critical dimensions continue to decrease. The need to employ reflective optics with the higher-energy light sources now required by state-of-the-art photolithographic processes (for example, 170-, 130-nm technologies) doubles the path length over which surface contaminants can interact with a light beam (both the incident and the reflected beam pass through the surface containment).[4]
In addition, high-energy photons can more easily dissociate AMCs that have been deposited on a surface creating reactive species that can form even larger contaminating compounds.[4] AMC-induced corrosion can also affect tools and structures that are part of the cleanroom.
HEPA/ULPA Filters
The size difference between AMCs and aerosol particles does not adequately explain AMC penetration through a particle filter. the figure plots particle penetration through a fibrous filter as a function of particle size.[5] This type of plot assumes that particle capture is dominated by diffusion transport for particles less than 0.1 µm in size and by interception and impaction interactions for particles larger than 1 µm. These names refer to the mechanisms that cause an aerosol particle to contact the surface of a fibrous filter. Transport of particles by diffusion becomes more significant as particle size decreasesparticle diffusion coefficients increase with decreasing particle size.
The size dependence of the interception and impaction mechanisms is just the oppositethese mechanisms of particle capture become more efficient as particle size increases. The transition between the regions in which each of these mechanisms dominate takes place over the 0.1 to 1.0 µm size decade and is marked by the maximum shown in the curves plotted in the figure.
Researchers have verified the general shape and behavior that is illustrated in the figure over the past 20 years. The demonstrated behavior is that particle penetration decreases as particle size becomes both larger and smaller than a size of maximum particle penetration.
At particle sizes smaller than the peaks of the curves in the figure, particle capture approaches 100 percent, even as it does for particle sizes larger than the peaks. Simple extrapolation of this plot to molecular clusters of the sizes typical of AMCs suggests that these fibrous filters should capture AMCs with high efficiency. This expectation proves untrueair molecules and AMCs readily penetrate fibrous filters.
Diffusion, interception and impaction are physical mechanisms leading to particle contact with the filter fibers. The filtration models on the plots of the figure are based on the principle that particle contact with a fiber represents particle capture. This assumption holds true for particles but not for AMCs.
What prevents this same model from accurately predicting AMC capture efficiency is the absence of the impact of the sticking coefficient (the number of adhering AMCs remaining on the fiber surface divided by the number of AMCs contacting the fiber).
Particle sticking coefficients are assumed to be 1. Sticking coefficients are less than 1 for airborne molecular clusters (Recent versions of the SIA Roadmap,[7] for example, list the sticking coefficients of various AMCs as 10-3 to 10-5.).
The same diffusion transport mechanism that causes small particles to collide with the fibers making up a fibrous filter also operates on AMCs. Lack of AMC contact with the fibers is not the problem. The difference between particle capture and AMC capture is that AMCs do not efficiently stick to the fibers while particles do.
Thermodynamic considerations clarify and justify such a conclusion. The thermal energy of a molecular cluster, regardless of size, is given by 3/2 kT, where k is the Boltzmann constant and T is absolute temperature (°Kelvin). Physical adhering forces between a molecular cluster and a surface are often dominated by Van der Waals forces, which depend directly on cluster size as well as material propertiesthe adhering forces decrease as cluster size decreases.
|
Thus, the ratio of thermal energy to adhesion force is larger for small-mass AMCs than it is for is for larger-mass particles. Most particles stick to the fiber; most AMCs do not. Thermal emission of AMCs, even at room temperature, keeps the equilibrium number of surface-adhering AMCs low. In addition, the surface heating that is part of many wafer processes raises thermal energy, which reduces the equilibrium concentration of adhering AMCs below room-temperature values. If filters based on physical transport and adhesion can't capture AMCs, how do you protect cleanrooms and work spaces from the deleterious effects of AMCs?
While many molecular clusters deposited on a surface can be removed simply by heating or by wet cleaning procedures commonly used in wafer processing, these solutions are post-deposition corrections. The preventive approach to be discussed here reduces AMC deposition by reducing the concentration of AMCs in the cleanroom environment to which process wafers are exposed.
Magnitude of the AMC problem
The concentrations of AMCs in cleanrooms vary from facility to facility and with time. Japanese researchers[8] report the following ranges of AMCs in cleanrooms not protected by chemical filters:
Acidic AMCs (MA): 10 – 1000 ppt
Basic AMCs (MB) 1000 – 100,000 ppt
Condensable AMCs (MC) (chiefly high boiling organic species[9]):
100-10,000 ppt
Dopant AMCs (MD): 0.1 – 100 ppt
For 0.25 µm technology, SEMATECH recommends the following maximum concentrations in each of these categories of AMCs:[10]
MA: 5 ppt
MB: 1000 ppt
MC: 1000 ppt
MD: 0.1 ppt
The Japanese researchers agree with these recommended limits except for the MC limit, which they regard as far too high.[8] Reference 8 recommends an MC limit of less than 1 ppt. So how does a cleanroom operator achieve these desired AMC concentrations?
Chemical Filters
The term chemical filters, as used here, refers to filters designed to remove AMCs from the air passing through them. These chemical filters are not filters designed to remove particles from liquid chemicals, although the same nomenclature is often used to describe such particle filters.
The discussion that follows will generally assume the filtration media of chemical filters to be that of a fibrous filter, although chemical filters can be made with other media forms, such as woven fabrics or non-woven felts.
Chemical filters remove AMCs by adsorption and/or chemical reactions that increase molecular adhesive forces above those typical of particle filters. Adsorption implies retaining AMCs on a surface by either physical forces between the AMC and the capturing surface (physisorption) or chemical bonding to the molecules of the capturing surface (chemisorption).
A more recent development is the chemical filter composed of ion exchange fibers that react with a specific AMC or class of AMCs. Such filters have selected, reactive chemicals impregnated into a base filter media. For example, activated charcoal impregnated with potassium permanganate effectively removes sulfur oxides.[11]
Physical adsorption is a low-energy interaction and is reversible. The relatively low bonding forces means that physically adsorbed species can readily desorb. Chemisorption is a higher-energy interaction characterized by higher bonding forces and is species dependent. A chemisorped AMC can be difficult to desorb and may desorb as a chemical species differing from what was initially adsorbed.
The most common adsorption-based AMC filters are the large variety of commercially available filters that use activated carbon. Such carbon-based filters are available as a carbon bed in a tray, as carbon granules incorporated among or between the fibers of a filter, as carbon impregnated into glass or polymer fibers and other configurations. Carbon is a workhorse adsorption media widely used to remove AMCs from air.
Acidic and basic AMCs (MA and MB), typically originating with ongoing processes within a cleanroom, have been controlled by chemisorption onto impregnated activated charcoal or ion exchange fibers tailored for removing either acidic or basic AMCs. A cationic species incorporated into the filtration media captures acidic AMCs; an anionic species, basic AMCs. A recent patent application discloses the use of a material called fraipontite which adsorbs both acidic and basic AMCs.[11]
Fraipontite is an aluminosilicate “having a double structure of which one side exhibits solid basic properties while the other side has solid acidic properties.”[11] Thus a chemical filter impregnated with fraipontite adsorbs both acidic and basic AMCs.
Condensable AMCs (MC) are often high boiling organic species that are efficiently physisorbed by activated charcoal. Dopant AMCs (MD) are species that modify the electrical properties of a semiconductor, such as boric acid, organo phosphates or arsenates. These AMCs are best controlled by eliminating their sources such as particle filters[6] or other material sources. Any detectable concentration of this category AMC is probably too much.[10]
Chemical filters, especially those containing activated carbon, sometimes shed objectionable particles that become airborne. Thus chemical filters are many times located upstream of a standard HEPA filter. Even so, materials used in the construction of chemical filters, like those used in the construction of particle filters, must be chosen with awareness of their propensity to introduce contaminants as well as remove them.
In general, AMCs can now be adequately measured and filtered. Should new generations of products demand AMC control beyond the present state-of-the-art, AMC characterization and control within minienvironments would become important.
References
- SEMI F21-95, “Classification of Airborne Molecular Contaminant Levels in Clean Environments,” Semiconductor Equipment and Materials International, 3081 Zanker Road, San Jose, CA 95134.
- Muller, A. J., L. A. Psolta-Kelty, H. W. Krautter and J. D. Sinclair, “Volatile cleanroom contaminants: sources and detection,” Solid State Technology 37 (9), September, 1994, pp 61-62, 64, 66, 68, 70, 72.
- Hartzell, A. L., “Deposition of Molecular Contaminants in Gaseous Environments,” Chap 8, pp. 289-295 in Contamination-Free Manufacturing for Semiconductors and Other Precision Products, Robert P. Donovan, editor, Marcel Dekker, Inc., 2001
- Kinkead, D. A., “Preventing Next-Generation Photochemical Contamination of Lithographic Optics,” MICRO 19(9), October 2001, pp. 40-43.
- Liu, B. Y. H., K. L. Rubow and B. Y. H. Pui, “Performance of HEPA and ULPA Filters,” pp 25-28 in the 1985 Proceedings of the Institute of Environmental Sciences (IEST, 940 East Northwest Highway, Mount Prospect, IL 60056).
- Lebens, J. A., W. C. McColgin, J. B. Russell, E. J. Mori and L. W. Shive, “Unintentional Doping of Wafers Due to Organophosphates in the Clean Room Ambient,” J. Electrochem. Soc., 143 (9) September, 1996, pp. 2906-2909.
- International Technology Roadmap for Semiconductors, 1999 Edition, p. 282 (Semiconductor Industry Association, 181 Metro Drive, Suite 450 San Jose, CA 95110 Phone: (408) 436-6600 )
- Fujimoto, T., K. Takeda, T. Nonaka, T. Taira and M. Sado, “Evaluation of Contaminants in the Cleanroom Atmosphere and on Silicon Wafer Surface (II): Organic Compounds Contaminants,” Vol. II Proceedings of the 1997 SPWCC, pp 157-166, Figures and Tables Addendum (Balazs Analytical Laboratory, 1219 Nightingale Court, Los Altos, CA 94024).
- Camenzind, M. “Identification of Organic Contamination in Cleanroom Air, On Wafers and Outgassing from Gloves and Wafer Shippers,” Proceedings of the 1996 SPWCC, pp 352-372 (Balazs Analytical Laboratory, 1219 Nightingale Court, Los Altos, CA 94024).
- Kinkead, D., M. Joffe, J. Higley and O. Kishkovich, “Forecast of Airborne Molecular Contamination Limits for the 0.25 Micron High Performance Logic Process,” SEMATECH Technology Transfer # 95052812A-TR, May 31, 1995 (Available at http://www. sematech.org).
- Sakata, S., H. Takahashi and K. Sata (to Takasago Thermal Engineering Co., Tokyo, Japan), “Air-Cleaning Filter, Method of Producing the Same, and High-Level Cleaning Device,” U. S. Patent 6,146,451, Nov. 14, 2000.
Robert P. Donovan is a process engineer assigned to the Sandia National Laboratories as a contract employee by L & M Technologies Inc., Albuquerque, NM.