ULTRAPURE MATERIALS—CHEMICALS

Using a buffered rinse solution to minimize metal contamination after wafer cleaning

Robert J. Small, Maria L. Peterson, and Aaron Robles, EKC Technology; Danielle Kempa, National Semiconductor; and Jill Knittel, Wacker Siltronic

Wet chemical processes are considered effective methods of removing particles, metal ion contaminants, and native oxides from semiconductor devices, with such wafer-cleaning steps accounting for up to 30% of the 300 to 500 process steps for a typical IC. As critical dimensions continue to shrink and junctions become shallower, however, individual devices are becoming sensitive to increasingly low levels of contaminants and the ever-present demand for more-advanced cleaning technologies is intensifying. The smaller geometries, including thinner gate oxides and metal films, will require different types of cleaning chemistries that are both environmentally friendly and have controllable process characteristics (etch rate, particle removal, and rinsability, for example). It is also critical that cleaning solution residuals be removed rapidly and thoroughly.

The detrimental effects on IC devices of metallic contaminants from cleaning chemistries and other sources have been studied extensively.^1—4 Heavy metals in silicon can form efficient charge generation or recombination centers, increasing the leakage current of junctions. Sodium and potassium ions have high diffusivities in oxide films, which can lead to threshold instabilities and a deterioration of gate oxides.⁵ Other metal ions such as iron, calcium, copper, zinc, and nickel also affect dielectric integrity and the generation and recombination of charge centers. Although front-end-of-line (FEOL) processes remain the industry’s focus of concern regarding metal contamination issues, devices at the back end of line (BEOL) are similarly susceptible because metals adsorbed into the various interlevel dielectrics may change the dielectric constants and possibly diffuse to contacts or metal junctions. The Semiconductor Industry Association’s predictions of maximum metal contamination levels for FEOL and BEOL from its 1994 National Technology Roadmap for Semiconductors are listed in Table I.⁶

Table I: SIA roadmap predicted acceptable residual metal contaminant levels for 1995 (64-Mb DRAMs) and 2010 (1-Gb DRAMs).⁶

As the table indicates, for 64-Mb DRAMs, the FEOL residual alkali ion level should be no higher than 1 x 10¹¹ atoms/cm² with transition metal ions at <5 x 10¹⁰ atoms/cm². By the time the industry is producing 1-Gb DRAMs, alkali ions will have to be reduced to <5 x 10⁹ atoms/cm² and transition metals to <2.5 x 10⁹ atoms/cm².⁷ Similarly, BEOL contaminant levels will have to drop from 5 x 10¹¹ to <5 x 10⁹ atoms/cm².

At present, even parts-per-billion-grade ultrapure chemicals can contribute iron concentrations on wafer surfaces in excess of 3 x 10¹² atoms/cm² for processes that have not been optimized.⁸ In addition, recent research with 8—20-nm silicon gate oxides contaminated with 10 x 10¹⁰ to 10 x 10¹⁴ atoms/cm³ has shown that reliability and breakdown voltage can be correlated with iron contamination levels: 10-nm oxides cannot tolerate iron contamination above 8 x 10¹⁰ atoms/cm³ and even at 20 nm the residual iron should be below 16 x 10¹² atoms/cm³.⁹ After discussing the adsorption of metal ions onto oxidized IC substrates and its importance to surface-cleaning processes, this article focuses on the postclean rinse, comparing a recently developed buffered solution with deionized (DI) water and isopropyl alcohol (IPA).

Mechanisms for Metal Ion Adsorption

In the presence of dissolved oxygen, or even oxygen in air, silicon and some metal substrates used in IC devices can oxidize, forming a monolayer (or thicker layer) of oxide on their surface. Thus, a cleaning solution often contacts an oxide layer rather than the bare metal of a device metal line. When this contact occurs, the oxide surfaces may acquire a charge that can enhance or repel ion adsorption; therefore, understanding the adsorption tendencies of mobile metal ions is critical to maximizing cleaning solution effectiveness. A summary of such characteristics is provided below.

The adsorption of hydrogen (H⁺) or hydroxide (OH^—) ions from water—or of other dissolved species such as metals, organics, or surfactants—can produce a surface charge on a wafer in solution. In acidic solutions, the high concentration of H⁺ ions tends to induce a positive surface charge, whereas the abundance of OH^— ions in alkaline solutions tends to induce a negative charge. As seen in Figure 1, each type of oxide that may form on a wafer surface has a point at which its surface is neutral, which is known as the zero point of charge (ZPC). When the charge is caused only by the substrate, water, and H⁺ and OH^— ions, without any other adsorbates considered, the ZPC is equivalent to another measure of neutral surface charge, the isoelectric point (IEP). The IEP of a metal oxide can be predicted by the solid composition, cationic charge, and radius of the metal ion.¹⁰

Figure 1: Effect of solution pH on the surface charge of various oxides.

As Figure 1 indicates, silica (SiO^₂) has a negative surface charge across most of the pH range; titanium dioxide (TiO^₂) and alumina (Al^₂O^₃) change from a positive surface charge to a negative charge at about pH 6 and 9, respectively; and magnesia (MgO) has a positive surface charge up to pH 12. Positively charged mobile metal ions, therefore, tend to adsorb to negatively charged silica surfaces over a wide pH range, even when only electrostatic attraction (preferential adsorption of positively charged ions on negatively charged surfaces) is considered (see Figure 2). Above pH 9, cations would also be expected to adsorb to alumina surfaces, but only under extremely basic (high) pH conditions would electrostatic cation adsorption to MgO surfaces be expected to occur. (Specific adsorption, or chemisorption, where positively charged ions bind to a positively charged surface or negatively charged ions bind to a negatively charged surface, also occurs but is more difficult to predict.) Other researchers have observed that the tendency for adsorption increases with ionic radius: K⁺ > Na⁺ > Li⁺ > Ca²⁺ > Mg²⁺, and the adsorption affinity of transition metals is affected by electronic configuration (Irving-Williams order): Cu²⁺ > Ni²⁺ > Co²⁺ > Fe²⁺ > Mn²⁺.¹¹

Figure 2: Typical isotherm for cationic adsorption on an oxide surface. The position and shape of the adsorption edge may shift depending on the affinity of the metal ion for the surface.

Based on these tendencies, to minimize metal wafer contamination cleaning solutions should have either a high pH and rely on ion exchange, or a low pH and rely on the tendency for cations to desorb. For example, the commonly used high-pH SC-1 cleaning solution (NH^₄OH + H^₂O^₂ + DI water) removes many cationic and organic species from surfaces by exchanging NH^₄+ for the adsorbed metals. Alternatively, a low-pH solution will cause desorption of metals simply by modifying the surface charge of the substrate. Including a chelating compound in the solution will encourage the desorption process and keep the dissolved metals from readsorbing on the wafer or precipitating into the solution.

Postclean Rinsing

The postclean rinse step, whether a part of an FEOL or BEOL process, is as crucial to maximizing device yields as the cleaning step itself. The rinse operation must preserve the surface cleanliness achieved by the clean and remove the residue from the chemical bath.¹² Each wafer has a certain carryover layer from the process tank, which varies in thickness between 30 and 100 µm. Ideally, this residual chemistry, whether acidic or basic, should be removed immediately to avoid any detrimental effects that may be caused by the shifting pH of the rinse solution. For example, Figure 3 shows that an oxide layer can protect aluminum from corrosion between pH 4 and 10.3. If the rinse solution pH shifts beyond these limits, the metal oxide layer will be etched, which can lead to scrapped wafer product. Other studies have shown that during the rinsing process a piranha solution changes from pH 1 to pH 6.8, but that this change takes 4 minutes in a quick-dump rinser (QDR) compared with 12 minutes in a cascade-type rinser. Similarly, pH neutralization of SC-1 rinse solutions requires 4 minutes in a QDR and 24 minutes in a cascade rinser.^13—16

Figure 3: Pourbaix diagram for an aluminum-water system at 25°C and solution aluminum concentration of 0.01 mole/kg. Dashed lines bracket the stability limit of water and solid lines define stability fields for the solid or aqueous aluminum species.

Another recent report indicated that DI water is not always successful at removing transition metal ion contamination on polished TEOS wafers.¹⁷ The effective removal (10—20% improvement) of sodium, aluminum, and calcium could only be achieved by adding NH_⁴OH to the DI water. The aluminum species may undergo a chemical change at the higher pH of the enhanced rinse solution, forming Al(OH)^₄— species that are stable in solution and thus releasing the metal from the oxide surface; the sodium and calcium are released by ion exchange as NH^₄+ ions compete with the alkali ions for surface binding sites. One possible explanation for the inability of plain DI water to desorb metal ions from surfaces is that the adsorbed metal ions require a driving force before they will desorb; that is, an ionic medium is needed to balance the ionic charges once the cation is in solution. Currently, 18-M‡ DI water has essentially no ionic impurities except for the residual H⁺ and OH^— ions, which, at pH 7, are close to 1 x 10^-7 mole.

It has also been found that transition metal ion contamination, including zinc, copper, and iron, can only be removed effectively with ammonium hydroxide and citric acid solutions.^18,19 The metal oxide—TEOS bonding must be altered and the metal ion in solution must be stabilized with chelators. This stabilization process is actually an alteration of the reduction oxidation potential of the metallic species and a reduction of the chemical activity of the ions that will inhibit their redeposition on the wafer.

A Buffered Postclean Chemistry

To maximize the effectiveness of the rinse step, a buffered postclean treatment (PCT) solution has been designed to neutralize amine-based cleaning chemistries instantly. This capability eliminates the diffusion time and slow pH change of DI rinsing via QDR or cascade rinsers, thereby eliminating potential metal corrosion problems caused by changing pH conditions in the rinse bath. Buffered PCT rinsing also facilitates the removal of metal ion contamination by modifying the surface charge of the substrate materials. The chemistry is nonflammable and can be disposed of safely into a fab’s acid drains.

In experiments comparing the effectiveness of the PCT chemistry with that of DI water and IPA, 3-in. wafers (n-type, phosphorous doped, <100>) with a TEOS blanket film were soaked in 100-ppb metal ion solutions—sodium, potassium (K), calcium (Ca), iron (Fe), or copper (Cu)—for 1 hour, then suspended into the rinse solutions for 5 minutes. All wafers were then rinsed with DI water and dried with nitrogen. Residual surface metal levels were measured using total reflection x-ray fluorescence.

The results, shown in Figure 4, indicate that the buffered pH 4 PCT solution reduced surface contamination from transition metal ions (iron and copper) and alkali metal ions (potassium and calcium) more effectively than IPA or DI water. The performance of the buffered solution for iron is notable because neither DI water or IPA significantly reduced surface concentrations of this metal.

Figure 4: Surface metal levels on test wafers following various rinses after contamination in a solution containing 100 ppb of selected metal ions.

Other experiments were conducted in a fab to qualify amine-based cleaning chemistries for postetch residue removal. Residual sodium values were determined by secondary ion mass spectrometry on ashed and unashed wafers after treatment in amine-based solutions followed by a rinse of DI water, IPA, or PCT solution. Figure 5 shows the postrinse sodium levels for wafers that were processed using one of two types of amine-based chemistries, EKC311 and EKC270 (EKC Technology, Hayward, CA). Results indicate that the ionic medium and chelating ability of the buffered PCT chemistry effectively reduced sodium contamination. In contrast, the IPA rinse was not successful at removing the sodium cation. IPA has a dielectric constant of 18, compared with the PCT solution’s dielectric constant of ~79. Solvents with high dielectric constants perform best at solvating charged species; if such stabilizing forces are not available, the cation species will not be released from the oxide surface. The results also show that the ashing step affects residual sodium contamination.

Figure 5: Surface sodium levels on ashed and unashed wafers following various cleaning chemistries and rinse solutions.

Conclusion

Simple DI-water rinses alone are not always sufficient for removing residual metal ions from IC surfaces. Buffered rinsing chemistries such as the PCT solution have an intrinsic advantage because they rely on the ionic nature of the solution to neutralize amine chemistries and remove metal contamination. Removal of adsorbed metals occurs by ion exchange, diffusion, or competitive adsorption of solution ions for surface ions followed by complexation of desorbed metals with chelating agents to stabilize the metals in solution. The PCT rinse chemistry instantly neutralizes the basic amine cleaning chemistry while removing residual adsorbed metals.

References

1. Bergholz W, Zoth G, Gelsdorf F, et al., “Metal Contamination in ULSI Technology,” in Proceedings of the Second Symposium on Defects in Silicon. Defects in Silicon II, Bullis WM, and Gosele U (eds), Washington, DC, ECS, pp 21—39, 1991.

2. Verhaverbeke S, Meuris M, Mertens PW, et al., “The Effect of Metallic Impurities on the Dielectric Breakdown of Oxides and Some New Ways of Avoiding Them,” in IEDM Technical Digest, Washington, DC, IEEE, pp 71—74, 1991.

3. Norga GJ, Platero M, Black KA, et al., “Mechanism of Copper Deposition on Silicon from Dilute Hydrofluoric Acid Solution,” Journal of the ECS, 144: 2801—2810, 1997.

4. Schmidt HF, Meuris M, Mertens PW, et al., “Evaluation of the Effects of Chemical Purity within the RCA Wafer-Cleaning Process,” Microcontamination, 11(9):27—32, 1993.

5. Wolf S, Silicon Processing for the VLSI Era, Sunset Beach, CA, Lattice Press, pp 1—120, 1995.

6. The National Technology Roadmap for Semiconductors, San Jose, Semiconductor Industry Association, 1994.

7. Deal D, “Coming Clean: What’s Ahead in Silicon Wafer Cleaning Technology,” Precision Cleaning, II(6):24—30, 1994.

8. Hall RM, Rosato JJ, and Lindquist PG, “Effect of SC-1 Process Parameters on Particle Removal Efficiency and Surface Metallic Contamination,” in Proceedings of the Materials Research Society, San Francisco, MRS, p 285, 1996.

9. Henley WB, Jastrzebski L, and Haddad NF, “Effect of Iron Contamination in Silicon on Thin Oxide Breakdown and Reliability Characteristics,” Journal of Non-Crystalline Solids, 187:134—139, 1995.

10. Parks GA, “The Isoelectric Points of Solid Oxides, Solid Hydroxides, and Aqueous Hydroxo Complex Systems,“ Chemical Reviews, 65:177—198, 1965.

11. Stumm W, Chemistry of the Solid-Water Interface, New York, John Wiley, p 344, 1992.

12. Hall RM, Rosato JJ, Lindquist PG, et al., “Improving Rinse Efficiency with Automated Cleaning Tools,” Semiconductor International, 19(12):151—160, 1996.

13. Olesen MB, Hill TV, and Phan T, “Sunburst Turbo Quick Dump Rinsing Study,” in Second International Symposium on Ultra Clean Processing of Silicon Surfaces, Bruges, Belgium, p 10, 1994.

14. Tonti A, “A Simple Model for Rinsing,” in Second International Symposium on Cleaning Technology in Semiconductor Device Manufacturing, Ruzyllo J, and Novak RE (eds), Phoenix, ECS, pp 41—47, 1992.

15. Helms CR, “Comparison of Diffusion and Convection in Aqueous Cleaning Tools,” in Contamination Control and Defect Reduction in Semiconductor Manufacturing III, Schmidt DN (ed), San Francisco, ECS, pp 222—227, 1994.

16. Rosato JJ, Walters RN, Hall RM, et al., “Studies of Rinse Efficiencies in Wet Cleaning Tools,” in Cleaning Technology in Semiconductor Device Manufacturing, Ruzyllo J, and Novak RE (eds), New Orleans, ECS, pp 140—152, 1994.

17. Huber A, Erdmann V, Zielonka G, et al., “Metrology and Analytics for the Optimization of CMP Processing,” in 14th International VLSI Multilevel Interconnection Conference, Santa Clara, CA, pp 343—348, 1997.

18. Burggraaf P, “Keeping the ’RCA’ in Wet Chemistry Cleaning,” Semiconductor International, 17:87—89, 1994.

19. Morinaga H, and Fujisue M, “Method and Compositions for Cleaning Surface of Substrate,” in PCT Int. Appl. WO 97 05, 228, Mitsubishi Chemical, 1997.

Robert J. Small, PhD, is research director of the R&D group at EKC Technology (Hayward, CA). In this position, he is involved in developing new chemistries for postetch residue removal and postclean treatment as well as with purification processes for raw materials. Small received a BS in chemistry from Norwich University (Northfield, VT), an MS in organic chemistry from Texas Tech, and a PhD in organic photochemistry from the University of Arizona. (Small can be reached at 510/784-5846.)

Maria L. Peterson, PhD, has worked for EKC as an R&D chemist since 1996. Her work involves developing and testing formulations for removing metal contamination and particles from IC substrates and magnetic disks, and formulating and testing CMP and post-CMP chemistries for metal and oxide substrates. She received a BA in geology from Carleton College (Northfield, MN), MS in geology from UC Berkeley, and PhD in surface geochemistry from Stanford University. (Peterson can be reached at 510/784-5845.)

Aaron Robles is a laboratory technician in R&D at EKC. His work involves testing chemical formulations for metal compatibility and residue removal, and analysis of corrosion and etch rates. Robles has a BS in biology from CSU Hayward.

Danielle Kempa conducted Class 100 cleanroom projects in the applications engineering group at EKC in 1996; she now works at National Semiconductor (South Portland, ME). Kempa attended Central Michigan University and received a BS in chemistry from CSU Hayward.

Jill Knittel worked at EKC Technology as an applications engineer developing process conditions for wet cleaning. She is now a final cleaning process engineer at Wacker Siltronic (Portland, OR). She has a BS in chemistry from CSU Hayward.

MicroHome | Search | Current Issue | MicroArchives
Buyers Guide | Media Kit