Carbon Dioxide Snow Cleaning
by
Robert Sherman
Applied Surface Technologies
New Providence, New Jersey
Introduction
Carbon dioxide snow cleaning is a straightforward cleaning process than can remove both particulates and organic based contamination. The carbon dioxide snow cleaning process is simple; a dry ice stream is created by passing liquid or gaseous CO2 through an orifice and cleaning occurs by interactions between the dry ice stream and the contaminated surfaces.
In practice, generally, carbon dioxide cleaning is fast, gentle to the component being cleaned, and does not damage the environment. It is nontoxic, nonflammable, non-ozone depleting and human exposure poses minimal risk except for oxygen replacement, CO2 build up or frostbite (if applied directly to the skin). The cleaning process is residue-free and can carry away the contaminants for venting (or capture if hazardous).
As part of this forum, the different forms of surface cleaning with carbon dioxide -- macroscopic pellets, supercritical CO2, liquid phase, and snow cleaning are being discussed. Pellet cleaning has large dry beads accelerated at a surface and cleaning is done via a thermo-mechanical shock. Supercritical CO2 cleaning is a batch process that relies upon the solvent properties of the supercritical fluid. Liquid carbon dioxide cleaning is based upon solvency action. Snow cleaning, the subject of this paper, relies upon momentum transfer for particle removal and a solvent process for organic removal.
Carbon dioxide snow cleaning has been discussed in the literature previously at Precision Cleaning {1,2} and in other publications {3,4, 7-16}. Within this paper, we aim to review the basics and discuss several applications and case studies.
The Snow Cleaning Process
A- Snow Formation The thermodynamics of snow formation has been discussed extensively by Sherman previously {1,2}. The major point concerning snow formation is that the expansion of carbon dioxide within an orifice is a constant enthalpy*. With this condition for either gaseous or liquid CO2 feed , the pressure drop from the initial cylinder pressure (800 psi) to a lower pressure can lead to a mixture of liquid and gaseous phases within the nozzle if the pressure remains above 80 psi. Finally, below 80 psi, all the remaining liquid converts to solid - dry ice. For a gaseous CO2 source, this about 8% dry ice yield, for a liquid CO2 source, the dry ice percentage is about 45%. Of course the larger yield (increased cleaning speed) of dry ice with a liquid CO2 source goes along with increased CO2 consumption (cost). The actual percentages of dry ice formed depend on many factors including choice of liquid or gas CO2 feed, source temperature, pressure, nozzle and orifice design. Within improper attention to thermodynamic variables - a non constant enthalpy expansion, or improper operating factors, operating problems or inefficiencies can occur.
B - Cleaning Mechanisms -- Whitlock {4} in 1989 discussed the two primary mechanisms for CO2 snow cleaning; particles removal by momentum transfer; organic removal by a solvent process. Particle removal is accomplished by the combined actions of momentum transfer between the impinging dry ice and the surface contaminant and an aerodynamic drag force. A flowing gas, exerting an aerodynamic drag force alone cannot generate sufficient forces to remove micron and submicron sized particles. CO2 snow cleaning introduces mass, as dry ice particles, into the gas stream and the collisions between the impinging dry ice particles and surface particulates give rise to momentum transfer and particle removal. Once liberated from the surface, the particulate is easily carried away with the high velocity gas.
The organic removal mechanism involves the presence of liquid CO2 -- an excellent solvent for hydrocarbons and other nonpolar substances. During the short impact time, high stresses exist at the snow - surface interface and the pressure can easily exceed the dry ice yield stress and triple point pressure. The dry ice particle liquefies and acts as a solvent while in contact with the surface. When the particle starts to rebound off the surface, the interfacial pressures decrease and the dry ice particle re-solidifies, removing the contamination. M. Hills {5} presented data supporting the above liquid phase mechanism and she established a rule of thumb to determine what compounds can be removed. Generally, organics, which are easily absorbed in liquid CO2 are removed; organics with complex or other functional non-hydrocarbon based groups not readily soluble in liquid CO2 are not quickly removed. Instead, an abrasive and freeze removal process was proposed for these compounds. It appears that in practice, both organic mechanisms come into play depending on the contaminant and the its thickness.
The above mechanisms were for high velocity - small dry ice cleaning systems. E. Hill{6} studied particle removal using a low velocity - large snow flake system and proposed a non-contact thermophoretic process (forc es caused by thermal gradients) for particle removal. The low impact velocity in this mode does not lead to a liquid CO2 phase and hence there is no organic removal.
C - Examples - Historically, Hoenig{7} was the first to show excellent particle removing abilities with CO2 snow cleaning in the open literature. Next, Whitlock{4} quantified the removal efficiencies for micron and submicron particles and demonstrated removal efficiencies over 99.99% for a contaminated wafer in a cleanroom environment.
A simple example of particle removal was given by Sherman{8}by comparing the exact same areas at 1000x magnification before and after snow cleaning. A silicon wafer was scribed with a carbide tip generating many micron and submicron particles near the scratch (Fig. 1a). After CO2 snow cleaning, the micrograph also at 1000x magnification (Fig. 1b), shows no particles. This example demonstrates the cleaning of silicon dust and would be typical of particle removal from many different materials including metals, glass, ceramics, etc.
[FIGURE 1a and b]
Sherman and Whitlock{9} quantitatively demonstrated the effectiveness of carbon dioxide snow cleaning in removing organic residues from surfaces. Here, surface analysis was used to investigate the surface chemistry of the exact same areas on new silicon wafers, contaminated wafers, and then CO2 snow cleaned wafers. A fingerprint or facial grease was applied to the wafer and the extent of the contamination was measured; next, each wafer was CO2 snow cleaned and then the surface chemistries were measured. The results not only demonstrated that CO2 snow cleaning can remove contamination but also left the surface with less hydrocarbon contamination than on the new wafers. Further tests on cleaning new, uncontaminated wafers indicated the potential for reductions in hydrocarbons on the order of 50% or more and demonstrated cleaning down to the atomic scale.
Sherman{8} provided further microscopic evidence of organic removal by comparing the exact same areas before and after cleaning a facial grease residue. A pair of micrographs is shown at 1000x magnification of the same area of a scribed silicon wafer before and after cleaning. The initial wafer condition is shown in Fig. 2a with extensive contamination and after CO2 snow cleaning, no contamination is observed as shown in Fig 2b. This visual evidence of removing organic contamination is typical for many surfaces and materials.
[FIGURE 2a &b]
Equipment
CO2 snow cleaning systems are straightforward, they consist of a CO2 source, a nozzle with an internal orifice and the means to transport the CO2 from the source to the nozzle. A typical system, such as shown in Fig. 3, consists of a CGA320 cylinder fitting, flexible tubing, an on/off gun or valve and a nozzle. The available on/off controls include solenoid, pneumatics, or manual valves and hand guns.
The nozzle design is by far the most important factor in performing CO2 snow cleaning. The simplest nozzles are the single expansion nozzles and can be a variation of a Venturi orifice. The exit side is more typically a nosecone and this chamber is for dry ice nucleation. Its angle and length play an important role in determining the velocity of the stream and the snow size. Other nozzle designs are available including small diameter orifice tubes, metering leak valves and other concepts. With these nozzle choices, sudden expansions can occur that violate the constant enthalpy condition and these nozzles may not work as effectively with a gaseous CO2 source and not be as efficient. Different cleaning abilities can result from different designs.
Hoening{7} described the first instrument for cleaning with CO2 snow. The expansion of the liquid CO2 led to formation of soft, low velocity snow flakes, and thus, the term "snow cleaning". Whitlock, Weltmer and Clark{10} patented a double expansion nozzle with a coalescing chamber between the two orifices. Sherman has developed single expansion nozzles in similar fashion as Whitlock's and has also shown that the single and double expansion nozzles can be easily modified to produce the low velocity, large snow flake mode similar to that introduced by Hoening. Other nozzle designs have been introduced by other manufacturers, all are simple tube designs with either a metering valve or an on/off valve for control.
[FIGURE 3 ]
The nozzles described above are point sources and clean about a 1/4 inch diameter. Layden{11} introduced large area double expansion nozzles and made units ranging from 1/2 inch to over 28 inches in length. Other concepts for large area cleaning have also been introduced.
Applications
CO2 snow cleaning is versatile; its cleaning applications can span from simple laboratory cleaning functions to production contamination control problems. These applications include cleaning many different materials, optics, vacuum components, hard disk components, process tools and systems, etc. Some applications are discussed in greater detail on the Carbon Dioxide Snow Cleaning World Wide Web page introduced by Applied Surface Technologies {12}. Some applications will be summarized here - in general, customers do not reveal details on cleaning and details are not included.
A - Metal Cleaning for Vacuum Technologies Many applications in vacuum technologies were developed including the cleaning of stainless steel vacuum components, cleaning of electron, ion and x-ray optics, samples for surface analysis such as AES, XPS, AFM, {13 - 15}, cleaning of residual gas analyzers (RGA) and others. Cleaning of electropolished stainless steel parts appears to be as effective as high purity reagent grade solvents such as acetone and methanol. One interesting application is the cleaning of the stainless steel surfaces during vacuum system fabrication. Snow cleaning can remove the machining oils, particle residues, and other debris from the stainless steel surfaces and even weld rods and tips before welding. This process has led to less staining near the weld zones and less weld defects.
As an example, stainless steel parts for high and ultra high vacuum are usually solvent cleaned in order to remove contamination and electropolished to smooth and passivate the surface. However, there are certain parts and assemblies, i.e. flexible metal bellows, that can not be electropolished and proper cleaning can be a challenge. With flexible bellows assemblies, the most common cleaning method is trichloroethylene or other strong solvents. This cleaning procedure led to success usually but stains in the heat-affected zones could be observed on some sections after welding. These stains could not be removed except by electropolishing (almost impossible to remove all solution afterwards) or mechanical abrasion and were considered unacceptable for many applications and end users. Analysis of the stains suggested a burnt hydrocarbon residue, the source being insufficient cleaning of machining oils and lubricants. Carbon dioxide snow cleaning has been testing by several vacuum part manufacturers including Ranor, Inc., of Westminster, Mass. There, they clean the bellow section with CO2 snow and find an increased yield and acceptable surface finishes after orbital welding. Further, they also clean the welding tips and other parts of the orbital welding setup. Results include less to no staining in the heat affected zones and no need for subsequent cleaning.
Another example is the cleaning of a RGA by Layden and Wadlow{15} who investigated and compared CO2 snow cleaning to solvent cleaning. The RGA was designed to operate below 1 micron pressure, and initial pumpdown times proved to slow even after the first solvent cleaning. Solvent cleaning was done by total disas sembly of the unit and ultrasonically cleaning in isopropanol. Repeated solvent cleaning led to an improved pumpdown time but still not as fast as needed. For CO2 cleaning, just the filament and beam aperture were removed. All parts were cleaned simultaneously. Pumpdown times now were under 30 minutes or less. Analysis of outgassing elements after solvent cleaning showed hydrocarbons and alkaline based contamination; and after CO2 snow cleaning, these peaks were reduced or eliminated. This simple test by Layden demonstrates the ability to clean complex vacuum equipment with CO2 snow cleaning.
B - Medical Parts Another example of cleaning involved a rather large and complex metal assembly used in medical imaging technology. This assembly has to be kept free of moisture during manufacturing and particulates residues on one surface can lead to premature product degradation and rejection. Initial manufacturing methods and cleaning choices led to many failures, rejections, a low yield, and worse of all, field replacements. Quality control tests by an alcohol spray for particle collection and analysis showed many metallic particles from other parts on the assembly were present on the critical surface. Carbon dioxide snow cleaning was tested on the entire part and subsequent quality control tests showed acceptable low to no particle results. Further testing justified the installation of two custom cleaning stations. Return of investment was within 18 months, with major yield improvements. Rejected parts and field replacements are down by over 70%. Similar cleaning results have been found on other parts used in the medical field including polymeric parts, ceramics and others.
C - Surface Analysis - The removal of inadvertent or intentional contamination without altering the base material is sometimes required by surface scientists who use techniques such as XPS, AES SIMS or AFM{8, 13,14,16}. Here, the scientists explore the surface elements and chemistries of the top dozen or so atomic layers. Residual contamination from handling, field tests, or just atmospheric exposure can interfere the analysis. Generally solvent cleaning in reagent or better grades of acetone or ethanol can remove most inadvertent contamination, but there are cases where solvent cleaning is inadequate. In these cases, ultrasonic solvent cleaning may take too long, or may not be able to remove or adequately reduce the residues. This has been the case for several users within the automotive and chemicals industry who have been interested in exploring the surfaces of test parts or field failures. Users of carbon dioxide snow cleaning can quickly clean these parts in under a minute and removing enough residue to allow analysis. The cleaning is probably a combination of solvent action and freeze fracture. Other researchers have commented that a similar situation exists for sample with just the absorbed hydrocarbons - the contamination can be reduced quickly without altering the substrates. Other have commented on the speed of cleaning making the process economic compared to solvent cleaning. In another study involving Auger spectroscopy on ceramics, J. Geller{14} found not only surface cleaning but also no surface charge build up during the analysis of insulators.
D - Optics - The cleaning of glass substrates before coating has become a major application area for CO2 snow cleaning. This cleaning can serve as either an initial or final cleaning of uncoated and coated glass, ceramic or semiconductor substrates, or to clean individual optical components during production or assembly. Similar reductions in surface hydrocarbon contamination have been noted for plate glass and other glass substrates in cluding soda lime glass, quartz, and for coated glass substrates (InSnO, ZnO, and SnO). In these cases, the cleaning process removed particles and organics and did not alter, remove, or abrade the metal oxide coating. Carbon dioxide snow cleaning has also been used to clean optical components including mirrors, gratings, filters, and many other items. The extent of cleaning is better than most industrial solvents, and as good as reagent grade solvents and some acid treatments.
Another cleaning example involved a gyroscope subassembly. This part is made from a low thermal expansion glass and its geometry includes a total of over 24 faces and over 40 holes. Cleaning is done via an automated setup located in a particle free and moisture free hood. The automated setup selects the nozzle and the point; tube and large area nozzles are all used. Cleaning this part for particles and residual organic contamination is critical for the end application and carbon dioxide snow cleaning has been shown to fulfill the requirement.
E - Hard Disks Assemblies and Components - Previously, Sherman and Adams {1,2} wrote on this application and its challenges and methods. Sliders become contaminated with a variety of materials during their manufacturing and processing. Individual sliders, head gimbal assemblies and Voice coil motors have been cleaned on a repetitive basis without damage. A short burst of CO2 removed the contamination without any resi due. Although head stack assemblies are very delicate and difficult to clean, they have been successfully cleaned with a gentle stream of CO2 snow. Additionally, CO2 has cleaned disk media without damage to the surface. Further, cleaning of the head stack assemblies and components mentioned above are well suited for automation.
F - Cleanrooms, Process Equipment, and Tooling - One of the unique cleaning features of CO2 snow cleaning is the ability to clean cleanrooms. This may be opposite expectations because the CO2 stream would make the particles airborne, but that is the required step. The key is to clean counters and exterior of process equipment during maintenance, when critical surfaces and equipment are closed or covered. Particle that becomes airborne during cleaning can be captured by the HEPA filters during air exchange. Tests have been done and this concept has been verified.
The same concepts also apply to cleaning process equipment. The chamber is opened during routine maintenance and snow cleaning is used in addition to the normal cleaning process. Again, for cleanroom envi ronments, the HEPA filters remove the liberated particles. A series of tests was performed on a silicide depo sition system in a cleanroom during maintenance when critical surfaces are covered. The chamber was cleaned with the goal of reducing particle contamination that would be found on test wafers. After the cleaning process was instituted and followed for several months, it was noticed that the particle populations on test wafers were re duced by a factor of five even after longer times between measurements. This particle population reduction led to increased yields. It is imperative that this cleaning be done when the cleanroom is not in production to avoid inadvertent contamination of nearby critical surfaces. Similar results have been obtained in non-cleanroom environments.
Parameters for Effective Cleaning
Successful snow cleaning, as for any cleaning process, requires attention to process parameters including economics, moisture condensation, CO2 purity, static charge, and others. By addressing these issues in a test program scaling a bench top cleaning system shown earlier in Fig. 3 to systems involving automation and process control environments becomes straightforward and affordable.
A- Economics - Economics of cleaning were discussed in greater detail in reference ; and to summarize, cost per part cleaned can range from less than a $0.01 per part to over $1.00 per part. Factors that can influence cost are cleaning time, source purity, and many other factors. All cost estimates are based upon cylinder sources and significant cost reductions can be achieved through the use of bulk purchased CO2. Equipment costs range from about $1,500.00 (shown in Fig 3) to $5,000 for basic units while automated units costs are much higher, starting at $50,000.
B- Recontamination Issues - Recontamination risks for any cleaning process must be investigated before implementation. The major sources of recontamination are from the carbon dioxide and from the cleaning process. The cleaning process and the environment that it is done in can lead to re-contamination - be it from particles that were liberated and can fall back on to the sample or from the cleaning chamber or facility. For critical cleaning applications, the cleaning environment chosen is vital and must have proper dry air or nitrogen purge flows to carry the contamination away from the sample after cleaning. A good choice for critical cleaning applications is a laminar flow hood with HEPA filters and recirculating nitrogen. Hoods such as these are available from manufactures of carbon dioxide equipment and also from cleanroom suppliers and other sources. Costs can range about and above $15,000.
Whitlock and Sherman separately performed recontamination studies as related to CO2 sources. Whitlock{7} performed a series of tests in a cleanroom by counting particles on new wafers before and after exposure to a carbon dioxide snow stream. The initial particle population was raised from an average total of slightly less than 16 particles before cleaning to over 80 particles afterwards - corresponding to an increase of about 1 submicorn particle per square centimeter. The carbon dioxide source was the best available at the time of work, with roughly less than 100 ppm heavy hydrocarbons. EDS failed to identify any element heavier than sodium suggesting that these added particles are hydrocarbon-based. High temperature baking in air and also standard wafer cleaning methods eliminated these particles. Similar tests on commercial grades (industrial) led to extensive recontamination.
Sherman{8} studied the effects of different CO2 sources including welding grade, food grade, instrument (Coleman) grade, and supercritical fluid chromatography grade (SFC). The residual hydrocarbon contents on Si wafers were measured before and after CO2 snow cleaning. These results imply that the SFC grades are acceptable for cleaning "clean" samples while the other grades may not be acceptable for critical cleaning applications. Nevertheless, the four grades that led to an increase in the surface hydrocarbon concentration on clean wafers would still remove visible contamination and particles from surfaces as seen in Figures 1 and 2 (welding gas was used in these examples) and satisfy many cleaning applications outside of wafers, disk drives, precision optics or other critical cleaning situations.
C - Filtration - Critical cleaning applications require filtering the CO2. The cylinders, hoses and all the equipment have particles and the flowing gas or liquid can transport these particles out of the nozzle and on to the sample. Generally, the cleaning process will remove these particles, but there are risks that some will stay or land nearby. For critical cleaning applications, it is imperative to have in-line filters at the point-of-use; this means placing the filter right before the nozzle.
D - Process Environment/Moisture Condensation The cold CO2 snow stream lowers the surface temperature and moisture can condense. Generally, moisture condensation does not interfere with cleaning in many applications, unless the moisture "freezes" or stays on the surface too long. Cleaning setups usually have a method to minimize moisture condensation - the easiest method is use a hot plate as part of the sample support. Usual set point for the hot plate set about 30-35C, but higher temperature may be necessary in some cases. The samples can be held on the hot plate by vacuum chucks or other special rigs. Generally, with a thermal input, for samples with good thermal conductivity, moisture condensation does not occur. For samples with poor thermal conductivity, such as thick glass, an overhead hot air gun or lamp can be used. Other choices for moisture control are dry boxes, enclosed hoods or environmental chambers that are purged or heated. In critical cleaning situations, i.e. submicron particle removal from wafers or optics, moisture condensation must not occur or removal of these particles are hindered. Cleaning must be done in dry environmental chambers with HEPA filters and the other necessary factors to insure not only a dry environment but also a particle free environment to insure no recontamination.
E - Static Charge There is a potential for build up of static charge on surfaces during cleaning. This is caused by the ionization of a flowing CO2 gas. Obviously, this static charge buildup is not a problem for metal samples. From our experience, if the sample is grounded, static charge is not a problem. Charging is usually worse for glass samples or for electrically isolated parts on complex structures. For these cases, commercially available positive ionization sources can be obtained for charge compensation.
Summary and Conclusions
Taking advantage of the thermophysical properties of CO2, an effective cleaning stream containing solid CO2 snow can be created. Data has been presented showing that the process can remove particles of all sizes and organic residues. Many examples were provided covering the cleaning of different materials, wa fers, and process equipment including
