Using an immersion-type BEOL cleaner with hydroxylamine and fluorine chemistries

Jae-Inh Song and Richard Novak, Akrion

As design rules have shrunk down to the sub-quarter-micron level, the numbers of metal lines in semiconductor devices have increased dramatically, and back-end-of-line (BEOL) cleaning has become increasingly critical to the IC industry. Two distinct developments have characterized BEOL interconnect technology since the advent of sub-quarter-micron design rules. The first is the extreme narrowing of aluminum-copper metal-line widths and spaces, prompted by the delay in implementing copper/low-k interconnect technology, particularly in memory devices.¹ (Recent semiconductor trends have favored the development of logic and communications-oriented device applications.) The second development is the implementation of mid- and low-k dielectrics using copper for the interconnect material.

These developments have made it incumbent on the industry to reduce defect density levels. At the same time, the high cost of BEOL cleaning chemicals can account for as much as 36% of the total cost of ownership of BEOL cleaning applications.² Such high costs have pressured companies to prolong chemical-bath lifetime in order to cut costs.

Since the early 1990s, highly organic, solvent-based chemistries such as hydroxylamine have dominated the polymer-removal process after metal-line and via etch/ash processes.³ However, numerous alternative chemistries have been introduced, including fluorine-based aqueous and semiaqueous formulas.⁴ The advantages of these latter technologies are shorter process times, which result in higher process equipment throughput; residue removal efficiency at lower process temperatures; and a reduction in environmental concerns and disposal costs. However, a process of record has not been established, particularly in copper/low-k applications, because the compatibility of the alternative chemistries with most low-k dielectrics has not been definitively confirmed. In addition, these chemistries have a relatively low process window (i.e., etch rate and polymer removal ability versus process temperature).

This article reports on the features and capabilities of an advanced immersion-type wet cleaning tool using both amine- and fluorine-based chemistries for postetch cleaning applications. The effect of various rinse conditions on defect density is discussed, and postclean SEM profiles are included.

Figure 1: Top-view schematic of the 300-mm BEOL immersion-type wet cleaning tool. Solvent tanks 1 and 2 include ultrasonics and operate at 40°–80°C. They are followed by two rinse tanks, one of which includes a CO2 injection and infuser system; an end-effector wash-dry (EWD) tank; and a low-consumption IPA dryer.

All experiments were performed at the Akrion Applications Laboratory in Allentown, PA, using the GAMA 300-mm BEOL wet cleaning tool, shown schematically in Figure 1. The tool's chemical process tanks, for both cleaning and IPA rinsing, are made from 316 stainless steel and have a closed recirculation loop. Some wafers were prepared by patterning and etching 0.4-µm metal-line features composed of TiN, Al-Cu, TiN-Ti, and FSG or PE-TEOS dielectric films. After patterning and etching, these wafers underwent a dry-plasma ashing process. Other wafers were were prepared by patterning and etching 0.3–0.7-µm via structures composed of TiN, Al-Cu, TiN-Ti, and FSG or PE-TEOS dielectric films, or by undergoing a standard short-loop process flow.

Hydroxylamine-Based Chemistry

Most commercial BEOL cleaning products based on hydroxylamine contain 18 to 22% DI water by volume. Keeping the DI-water content constant is very important to achieving stable polymer removal. However, because a hydroxylamine-based chemistry performs best at temperatures in the 60°–70°C range, the total DI-water content of an aqueous solution can decrease by as much as 50% over a 20-hour period. Figure 2 shows the relationship between DI-water content and bath life for a hydroxylamine-based organic polymer remover at 70°C. The DI-water percentages were measured by gas chromatography–mass spectrometry.

Figure 2: DI-water content versus chemical bath life for a hydroxylamine-based organic polymer remover at 70°C (y = –0.012x + 20.923; R2 = 0.985).

The DI-water content in a hydroxylamine-based solution also influences the rate of cleaning-chemistry-induced surface corrosion on metallic substrates. Figure 3, for example, shows that significant corrosion can occur on Al-Cu film surfaces when DI-water content is between 3 and 50% of the chemical solution. To maintain solution equilibrium in the BEOL tool, the chemical tanks include a DI-water spiking feature.

Figure 3: DI-water content versus corrosion rate on an Al-Cu surface.

IPA Rinsing. To date, the best way to minimize chemical corrosion is to rinse wafers with a nonpolar solution such as isopropyl alcohol (IPA) immediately after an organic polymer-removal treatment, particularly when the clean follows a plasma process such as line or via etch. However, the concentration of polymer remover in an IPA rinse tank gradually increases, creating a potential source of corrosive chemical residues on the wafer surface during the subsequent DI-water rinse. The ability to control the amount of organic polymer remover in an IPA bath could minimize that possibility. To that end, the wet cleaning tool used in this study incorporates an indirect, pH-based method to monitor and adjust the concentration of organic polymer remover in its IPA rinse tank. Figure 4 shows the relationship between hydroxylamine content and pH variation in an IPA solution. Pure IPA has a pH of 7.5; as seen in the figure, this level increases sharply as soon as the IPA is exposed to the cleaning chemistry. Saturation is reached when the concentration of organic polymer remover reaches about 3% by volume.

Figure 4: pH variation versus hydroxylamine content in an IPA rinse tank.

DI-Water Rinsing. Most chemically induced surface corrosion on Al-Cu film occurs during the DI-water rinse step.⁵ Because wafers have a chemical-carryover layer after immersion in an organic polymer remover, even after an initial IPA rinse, a subsequent DI-water rinse, which attacks this carryover layer, can produce highly reactive species such as OH. Such species, in turn, can react on metallic substrates and corrode the surface.

Two techniques have been introduced in the wet cleaning tool to prevent such corrosion: one is CO₂ injection during the DI-water rinse; the other is the use of megasonic energy. CO₂ is very soluble in DI water and generates carbonic acid (H-COOH), which acts as a corrosion inhibitor on metallic substrates.⁶ However, the pH of DI water with CO₂ injection turns acidic at around 4.2–4.5, at which point the wafer surface's affinity for particulates increases (based on the zeta potential theory).

Figure 5: Comparative particle behavior on oxide (PE-TEOS) for DI-water rinses using various CO2 concentrations and megasonic power levels.

The implementation of megasonics prevents particles from readhering during the rinsing process. As shown in Figure 5, the use of megasonics with the DI-water/CO₂ rinse can result in fewer surface particles than rinsing without the megasonic enhancement. While megasonics are not recommended for post-metal-line-etch cleaning processes because of the potential for pattern damage, such as line lifting, their use is extremely beneficial for post-via-etch processes, which do not show any pattern degradation, as can be seen in Figure 6.

Figure 6: Comparative SEM profiles of a via (a) before and (b) after exposure to sonic energy during both an organic polymer-removal treatment and DI-water rinse process.

On Al-Cu surfaces, the DI-water temperature during the rinse process correlates to the surface corrosion metric known as pit corrosion density (PCD). As shown in Figure 7, PCD is lower at 17°C than at rinse temperatures above 25°C, whether or not CO₂ injection is incorporated into the rinse process. At 33°C, however, the difference between the results with and without CO₂ is significant.

Figure 7: Pit corrosion density (PCD) versus DI-water rinse temperature.

Polymer-Removal Ability. The keys to achieving dry-etch residue removal without surface corrosion are maintaining an efficient liquid flow in the process tanks, applying ultrasonic energy during the chemical immersion steps, and maintaining tight control of such process parameters as temperature and DI-water content. The BEOL cleaning tool under study satisfies all three requirements.

Figure 8: SEM profiles of a via (a) before, and (b) and (c) after cleaning with a hydroxylamine chemistry. The cleaning tool safely removes sidewall etch residues (as shown in b) and does not create a T-top profile (as shown in c).

The first two requirements are especially important in high-aspect-ratio via contact applications because heavy polymer-residue generation cannot be avoided during the etch process (resulting from the need for a positive slope). As shown in Figure 8, when used with a hydroxylamine chemistry, the system removes etch residues even in a misaligned via contact without causing damage or leaving a T-top profile.

Based on the experimental results, the recommended process recipe for the tool with a hydroxylamine-based chemistry consists of five steps:

• Etch-residue remover immersion for 10 minutes in tank 1, with efficient fluid recirculation, ultrasonics for via applications, and DI-water spiking to maintain equilibrium.

• A repeat of step 1 in tank 2.

• IPA rinse for 10 minutes in a tank equipped with an in-line chemical concentration-control system.

• High-flow DI-water cascade rinsing for 7–10 minutes in a tank equipped with a CO₂ injection and verification system, and megasonics.

• Standard IPA dry.

Aqueous and Semiaqueous Fluorine Chemistries

The dry-etch-residue removal mechanism of fluorine-based chemistries is a combination of chemical diffusion into the residue and liftoff of polymer films by the chemical's moiety, followed by their dissolution in water during a DI-water rinse step. Unlike conventional amine-based chemical cleaning processes, most fluorine-based solutions are effective at relatively low (ambient) process temperatures.

Figure 9: SEM profiles of a via after cleaning with a fluorine-based chemistry.

Polymer Removal Ability. Figure 9 illustrates the BEOL cleaning tool's effectiveness when used with fluorine following a via etch process. No etch residues remain, and there is no sign of surface corrosion on the metal surface at the bottom of the via contact or of significant structural or topological damage.

Based on study results, polymer removal ability in most fluorine-based chemistries is not dependent on the chemical immersion time, which can therefore be optimized in the range of 3–5 minutes per step. The recommended five-step process recipe with a fluorine chemistry is as follows:

• Etch-residue remover immersion for 3–5 minutes, with efficient fluid recirculation, megasonics, and DI-water spiking.

• High-flow DI-water cascade rinsing for 3–10 minutes with megasonics.

• Etch-residue remover immersion for 3–5 minutes, with efficient fluid recirculation, megasonics, and DI-water spiking.

• High-flow DI-water cascade rinsing for 3–10 minutes with megasonics.

• Standard IPA dry.

Figure 10: Particle performance of the BEOL cleaning tool using a fluorine-based chemistry.

Particle Performance. The particle performance of the BEOL cleaning tool with a fluorine-based chemistry is shown in Figure 10. The process recipe used during the data collection period consisted of two 5-minute immersions in the cleaning chemistry, each of which was followed by a 5-minute DI-water rinse and a final standard IPA dry.

Particle levels on a 1000-Å PE-TEOS surface with 3-mm edge exclusion were measured using an SP1 tool from KLA-Tencor (San Jose). As the figure indicates, particle performance was controlled at around 12 ≥0.12-µm particle adders per 200-mm wafer. Based on these results, it is apparent that fluorine can etch away the contaminated top layer of a wafer while additives in the chemistry suppress particle readhesion.

CD Variation. Generally, fluorine-based chemistries exhibit extremely low etch rates on most nonpatterned, blanket-wafer dielectrics. However, on patterned wafers, the etch rate on contact sidewalls that have been structurally damaged by high-density plasma can be significant. In the worst case, a CD loss of 300 Å per side has been reported.

Figure 11: Process temperature versus CD variation for fluorine-based chemistries. The respective CD specifications were 0.3, 0.35, 0.5, and 0.7 µm.

Preventing such variation requires maintaining tight control of the dry-etch/ash and cleaning processes. The relationship between CD variation and cleaning-chemical process temperature is shown in Figure 11, indicating that a CD may vary by 15%, particularly in small-diameter contacts, depending on process temperatures. In most applications using a fluorine-based chemistry, lowering the process temperature can minimize the CD skew. The design of the BEOL tool addresses this issue by including a heater/heat exchanger in the recirculation loop to control chemical temperature.

Conclusion

A BEOL cleaning tool has been found to provide efficient sidewall polymer removal in postetch applications without structural damage using either hydroxylamine- or fluorine-based chemistries. Bath life can be prolonged with a DI-water spiking method during chemical recirculation. For hydroxylamine recipes, use of both an in-line concentration control system in the IPA rinse tank and a bubble-free DI-water rinse with injected CO₂ minimizes surface corrosion on metallic substrates.

Particle defect densities are kept low by using megasonics in both the chemical and DI-water-rinse tanks to prevent readhesion. The CD loss issue in fluorine-based chemical applications has been minimized by tightly controlling process temperature with a high-capacity heater/heat exchanger in the process tank recirculation loop.

References

1. M Igarashi et al., "The Best Combination of Aluminum and Copper Interconnects for High Performance 0.18 µm CMOS Logic Devices," in Proceedings of the International Electron Devices Meetings (Gaithersburg, MD: IEDM, 1998), 829–832.

2. J Song, Akrion marketing data (Allentown, PA: Akrion, 2000).

3. C Helms, "Contamination Control and Defect Reduction in Semiconductor Manufacturing III," EKC Technology technical data (Danville, CA: EKC Technology, 1994), 22.

4. D Follweiler and D Peters, "Recycling of Solvents in Used Photoresist Strippers," in Proceedings of the Fourth International Symposium on Environmental Issues with Materials and Processes for the Electronics and Semiconductor Industries (Pennington, NJ: Electrochemical Society, 2001).

5. P Marcus and J Oudar, "Corrosion Mechanism in Theory and Practice" (New York: Marcel Dekker, 1995).

6. E Flick, Corrosion Inhibitors, Industrial Guide (Westwood, NJ: Noyes Publications, 1993).

Jae-Inh Song, PhD, is vice president of process applications and marketing at Akrion (Allentown, PA). His interests include manufacturing-oriented surface preparation and characterization technology, and cleaning technology to enhance yields. Song has authored or coauthored more than 30 journal publications and 40 conference presentations. He is also the holder or coholder of approximately 100 semiconductor process and equipment-related patents. He received a PhD in solution chemistry in 1989 from the University of Glasgow, UK. (Song can be reached at 610/530-3607 or jsong@akrion.com.)

 

Richard Novak, PhD, is vice president of advanced technology and CTO at Akrion. He was a member of the technical staff at RCA Laboratories and has more than 25 years of experience in the IC industry. He cochaired the first four International Symposia on Wafer Cleaning. Novak received a PhD in ceramic engineering from the University of Illinois in Urbana-Champaign. (Novak can be reached at 610/530-3449 or rnovak@akrion.com.)