Developing supercritical carbon dioxide processing in microelectronics applications

Christopher Case, BOC Edwards; and James McClain, Micell Integrated Systems

The fabrication of integrated circuits includes many cleaning steps. For example, 125 such steps can be required to manufacture a typical 180-nm circuit. Cleaning challenges increase as feature sizes decrease. Hence water-based cleans do not always reach small, high-aspect-ratio features, and their residues are difficult to remove completely. In addition, drying wafers after water-based cleans can damage mechanically weak structures such as MEMS devices and porous low-k dielectrics.

Figure 1: SEM images of 150-nm isolated lines with a 6:1 aspect ratio that (a) have collapsed during traditional aqueous cleaning, and (b) have been maintained during processing using an inverse microemulsion formulation in supercritical CO2. (Samples provided by IBM Thomas J. Watson Research Center.)

To meet the challenges posed by shrinking feature sizes, the semiconductor industry has turned to the use of supercritical carbon dioxide (scCO₂) in microelectronics cleaning applications. In contrast to aqueous cleans, scCO₂ has low viscosity and negligible surface tension, so that the smallest structures can be protected from the damage associated with surface tension and capillary forces. At the same time, scCO₂'s gaslike properties facilitate its removal, obviating residue concerns. The effect of using a standard aqueous-based chemistry versus an scCO₂ formulation to remove residues from fragile structures is highlighted in Figure 1.

This article discusses the practical implementation of scCO₂ technology and reviews the results from copper and postash residue-removal tests.

The Use of scCO₂ in IC Cleaning Processes

Above certain pressures and temperatures, some materials do not experience a distinct phase change; they are in a supercritical state. Their molecules are too close together to form a gas, and at sufficiently high temperatures, the energy of the molecules is so great that the materials cannot condense into a liquid. Carbon dioxide, water, and propane are examples of materials that have widely studied supercritical phases.

CO₂ becomes supercritical at a relatively low temperature of 31°C and pressure of 75 bar. At those levels, it has liquidlike density, with very low viscosity and negligible surface tension. In addition, because scCO₂ is the second-most abundant and second-least costly solvent known, it is expected to play a crucial role in difficult and delicate clean and dry operations in the future. Consequently, a discussion of supercritical processing was incorporated into the 2001 edition of The International Technology Roadmap for Semiconductors (ITRS) as a possible solution for cleaning challenges in interconnect technology and MEMS devices.¹ Less recognized, but equally important, is the need to develop nonaqueous processes for problem materials such as germanium alloys and compound semiconductors.

Process steps in which scCO₂ is thought to be particularly advantageous include:

• Lithographic image development.

• Postdevelop photoresist drying.

• Postetch and postash cleaning of high-aspect-ratio structures.

• Postetch and postash cleaning of porous low-k structures.

• Cleaning, postclean drying, postclean lubrication, and etching of MEMS devices.

• Barrier layer and seed layer metal deposition.

• Porous ultra-low-k film formation.

• Compound semiconductor device metal lift-off.

Unlike water, CO₂ has no dipole moment and is not ionically active. Thus, very few substances spontaneously dissolve into CO₂. To remove postash residue from wafer vias, for example, it is necessary to add appropriate chemistries to scCO₂. While the addition of polar cosolvents yields a homogenous solution of increased polarity, cosolvent concentrations of <20% do not support ionic dissociation and solvation. Moreover, at higher concentrations, the process advantages of scCO₂ are lost. Therefore, specifically formulated surfactants can be added in small amounts to CO₂ to perform postetch or postash residue removal and metal cleaning. As illustrated in Figure 2, such surfactants create a nanometer-scale heterogeneous system of discrete polar microdomains dispersed in a continuous phase of nonpolar CO₂.² These discrete domains of microemulsions are often based on aqueous solutions, or on water itself.

Figure 2: Design of surfactants and the formation of micelles in scCO2: (a) CO2-philic structure, (b) CO2-phobic structure, and (c) micelles.

Water-in-CO₂ microemulsions were first explored in the early 1990s.^3,4 The surfactants added to CO₂ have a polar head group that points inward and a CO₂-philic tail that forms what are known as reverse micelles, which remove process residues. The polar domains in the microemulsion serve two important functions. First, they provide the medium for solvation and removal of water-soluble polar contaminants. (Experimental results show that >80% of postash residues can be removed with pure water-in-CO₂ microemulsions.) Second, they deliver the active chemistry that removes the rest of the residues.

In contrast to the chemistry system deployed in CO₂, an everyday surfactant such as laundry detergent contains oleophilic segments that enhance water's ability to solvate oily matter and hydrophilic portions that isolate the oily contaminant microscopically. In order to perform CO₂-based wafer cleans, active chemicals, analogous to those used in commercial cleaners, are modified to be compatible with CO₂. Typical aqueous-based commercial cleaners contain one or more active chemicals, such as acids, bases, chelants, etchants, and oxidants. Research has concluded that water-in-CO₂ microemulsion systems that contain one or more of these active chemistries are highly effective at removing residues rapidly and completely.

Processing Tests and Results

In scCO₂ tests, chemistry formulations for specific processes were first optimized using wafer fragments and a 5–25-ml laboratory apparatus called a view cell, which is equipped with sapphire windows allowing investigators to view the interior. It has been shown that formulations created in the view cell work similarly in 200-mm single-wafer processors as long as processing pressures, temperatures, and times are the same. Regardless of chamber size, scCO₂ processing equipment is generally configured as shown in Figure 3. Chemistries are injected into pressurized CO₂ because the solubility of the chemical adjuncts is typically a function of pressure.

Figure 3: Schematic drawing of scCO2 processing equipment.

Copper Cleaning. In damascene processing, copper ions are present on the surfaces of dielectric sidewalls and on top of Cu(0) lines after the bottom barrier has been etched away. Known as post-breakthrough, this ion residue is removed using aqueous cleaning processes that employ a suite of chemicals to enhance their solubility. It is critical to remove these residues in interconnect processing because they directly affect yields.

Figure 4: Photograph of 2-mg/ml copper ions effectively dissolved in scCO2 after treatment with a water-in-CO2 reverse microemulsion.

The ability of microemulsions to solubilize copper contamination in scCO₂ can be demonstrated visually. In one experiment, an oxidized copper complex—Cu(II)—was placed in a small, open pan in a view cell. The view cell was then closed and pressurized to 2500 psi at 40°C. The clear, colorless scCO₂ solution, visible through the window, remained unchanged under these conditions. The experiment was then repeated with a CO₂-philic surfactant and a water microemulsion, which were placed in a separate pan adjacent to the copper salt. The CO₂ turned green within one minute because of the presence of Cu(II) species in solution. The solution continued to change over the course of 3 minutes until it was deep green in color, at which time all of the copper complex had been dissolved. The results of the experiment are shown in Figure 4. The microemulsion formulation containing 2% surfactant and 17 mol of water per mole of surfactant rapidly solubilized 2 mg of Cu(II) complex per ml of scCO₂, representing a solubility level approximately six orders of magnitude larger than that of unmodified scCO₂.

After it had been demonstrated that microemulsions can solubilize oxidized copper, processes were developed to remove etch residues. First, three blanket copper wafers supplied by International Sematech were exposed to etch stop breakthrough gas (C₄F₈/O₂) to generate oxidized copper residues. Then the wafers were processed using either a commercial wet cleaning agent or two different CO₂-based microemulsions.

X-ray photoelectron spectroscopy (XPS) analysis results, presented in Table I, reveal that after breakthrough etch, only fluorine, oxygen, and carbon were present on the surface. No copper residue was present. After the wafers were cleaned, copper appeared on the surface while the fluorine peak decreased, signifying that the wafers were being cleaned. Significantly, the wafers cleaned with the scCO₂-based solvents had higher copper levels than the wafer cleaned with the commercial agent. The depth profiles in Figure 5, which compare atomic concentrations on the as-received wafer and those on one of the CO₂-cleaned samples, clearly demonstrate that copper residues can be removed using scCO₂-based microemulsions.

Figure 5: XPS depth-profile analysis showing concentration versus sputter time for (a) the as-received sample, and (b) a CO2-cleaned sample.

Postbarrier Breakthrough Cleaning and Residue Removal. It is somewhat more complicated to optimize postash residue-removal processes than copper cleaning processes. The chemical nature of postash residue depends on the interaction between the dielectric and the photoresist and etch gases. Baseline processes have been developed for the most common spin-on and plasma-enhanced chemical vapor deposition dielectrics.

To illustrate the effectiveness of scCO₂-based solvents in postash cleaning processes, post-breakthrough cleans were performed on patterned wafers containing two different low-k spin-on materials: JSR 5109 from JSR (Tokyo) and dense and porous SiLK from Dow Chemical (Midland, MI). Initially, no microemulsions were formed, since active chemistries, including acids, bases, etchants, and oxidizers, were mixed with scCO₂ without the addition of surfactants. The chemistries had been screened for their ability to clean wafers containing the low-k materials. Microscopic inspection showed that none of the formulations effectively removed visible residues from the post-breakthrough samples. However, when water-in-CO₂ microemulsions composed solely of water and a surfactant were tested, better cleaning results were achieved. Judging by the volume of residue material removed, the microemulsions appeared to be roughly 80% effective. Achieving optimal cleanliness, however, requires the addition of active chemicals.

To determine the effectiveness of scCO₂ processing, device structures on a wafer containing JSR 5109 dielectric and post-breakthrough structures were compared before cleaning, after cleaning with a process of record (POR) aqueous solvent, and after cleaning with a water-in-CO₂ microemulsion containing active chemistry.⁵ The results of this test are presented in Figure 6. The scanning electron microscope (SEM) image in Figure 6a of the as-received wafer sample shows granular residues on the floor and lower wall of the via and a white residue band halfway up the via wall. The z-contrast transmission electron microscope (TEM) image in Figure 6b of the as-received wafer clearly shows copper-containing residues on the lower half of the via wall. The SEM image in Figure 6c of the aqueous-cleaned sample shows that all residues were removed from the via floor but that some remained on the wall. The TEM image of the same sample in Figure 6d shows that all residues were removed from the via floor but that contamination remained on the via sidewalls. The SEM image in Figure 6e of the sample cleaned using a water-in-CO₂ microemulsion also shows that virtually all residues were removed without copper overetch and CD loss in the low-k dielectric. This test and comparable tests involving other dielectrics indicate that properly formulated CO₂ microemulsions are at least as effective as aqueous cleans at cleaning postash residues.

System Needs of CO₂-Based Processing

Since CO₂ becomes supercritical at a pressure level of 75 bar, the implementation of supercritical technology requires wafer processing in pressure vessels. In addition, the implementation of supercritical CO₂-based processes requires the development of new materials delivery and processing systems. In recent years, specialized scCO₂ pressure equipment and processes have been developed for microelectronics applications. The first production-worthy units are moving from the lab to the fab.

A schematic diagram of a production-scale CO₂ system is shown in Figure 7. Liquid carbon dioxide can be supplied either from a cooled bulk tank or from a room-temperature container such as a cylinder. The pressure of the liquid is increased to the supercritical regime, typically 3000 psi, with a pump. Because the solubility of chemicals in CO₂ is a function of pressure, chemicals are usually introduced into pressurized liquid CO₂. The pressurized mixtures are stored in accumulators, vessels capable of delivering requisite quantities of CO₂ into a processing chamber at high flow rates.

Figure 7: Schematic diagram of a production scale CO2 system.

After the process has been completed, the chamber cannot be depressurized directly, because pressure-dependent solubility can cause additives to deposit on the wafer. Instead, the chamber is purged at the processing pressure with pure CO₂, a lengthy procedure for which alternative technologies have been developed to decrease the cycle time.

Liquid CO₂ is discharged from the processing vessel into a pressure vessel known as a separator, which separates the chemistries from the CO₂. Cost-of-ownership calculations indicate that in a typical fab situation, a cost-effective implementation of the technology requires that CO₂ be recycled, as illustrated in Figure 7. Recycled CO₂ vapor is purified by distillation and supplemented with new liquid in the purifier.

CO₂ purity requirements for supercritical processing are unknown; no published studies causally link processing results with CO₂ purity. The starting grade of 99.9% (3N), known as beverage grade, is readily available commercially. As the technology matures, critical impurity levels will become better understood. If required, 3N purity can be readily transformed into 5N.

The processing chamber itself is a pressure vessel that must be engineered and constructed according to applicable engineering codes. Since most chamber designs accommodate a single wafer, cycle times must be kept as short as possible. The volume of the single-wafer chamber must be small, and the chamber closure and locking mechanism must be fast-acting. To integrate postash cleaning with dry etch/ash processes, high-pressure tools should be compatible with vacuum and controlled-atmosphere cluster tools.

Conclusion

Microelectronics processing with supercritical fluids is moving from the lab to the fab. Initially, the benefits of chemistry delivery by CO₂ rather than by water were thought to be applicable strictly to low-k materials, including specifically porous ones. That view is rapidly changing, however; now the technology is considered valuable because it helps to avoid costs generally associated with the removal of water in critical applications. It is too early to predict the extent to which CO₂-based processing will penetrate microelectronics processing. Nevertheless, the fabrication of some MEMS devices relies heavily on supercritical processes, while silicon processing is clearly moving in that direction. Since device dimensions will continue to shrink on a predictable schedule, either improved drying processes will be required to combat water retention in fine structures, or water will have to be avoided altogether.

References

1. The International Technology Roadmap for Semiconductors (San Jose: SIA, 2001).

2. JB McClain et al., "Design of Nonionic Surfactants for Supercritical Carbon Dioxide," Science 274 (1996): 2049–2052.

3. JL Fulton et al., "Aggregation of Amphiphilic Molecules in Supercritical Carbon Dioxide: A Small Angle X-Ray Scattering Study," Langmuir 11 (1995): 4241–4249.

4. KP Johnston et al., "Water-in-Carbon Dioxide Microemulsions: A New Environment for Hydrophiles Including Proteins," Science 271 (1996): 624–626.

5. MI Wagner et al., "Surfactant Enabled CO₂ Cleaning Processes for BEOL Applications: Post Barrier Breakthrough," in Proceedings of the Electrochemical Society Meeting (Pennington, NJ: Electrochemical Society, 2003), in press.

Christopher Case, PhD, is chief technology officer at BOC Edwards (Crawley, W. Sussex, UK). He chairs the ITRS interconnect technology working group, collaborating closely with industry experts to forecast IC trends. Before joining BOC Edwards, Case was an assistant professor of engineering at Brown University (Providence, RI). He has held positions at Solid State Solutions, Bell Labs, and International Sematech. He received a master's degree from Université Bordeaux in France and an ScB and PhD in engineering from Brown University in Providence, RI. (Case can be reached at +44 1293 528844 or chris.case@bocedwards.com.)

James McClain, PhD, is president and chief technical officer of Micell Technologies and its subsidiary, Micell Integrated Systems (Raleigh, NC). He is responsible for the direction of the company's work in the area of CO₂-based applications and is involved in applying its technologies to several aspects of microelectronic processing. He received a BS in chemical engineering from Lehigh University in Bethlehem, PA, and a PhD in chemistry from the University of North Carolina in Chapel Hill. (McClain can be reached at 919/313-2111 or jmcclain@micell.com.)