PROCESS EQUIPMENT—LITHOGRAPHY

Using CD SEM to evaluate material compatibility with DUV photoresists

Zhou Lin and Anne F. VanNatter, Silicon Valley Group

The adverse effect of airborne molecular contaminants, such as ammonia or various amine compounds, on the control of critical dimensions (CDs) during lithography of sub-0.5-µm linewidths is well understood in the semiconductor industry. As device dimensions continue to shrink, this phenomenon and its effects on defect density and product yield have become an increasingly important concern. In response, both manufacturers and users of photolithography equipment must implement procedures that will help ensure that the process chamber remains free of molecular contamination. Because of the need to review the potential of all materials in the processing area to contribute airborne contaminants, the amount of required testing can be significant.

Among the contaminants that particularly affect lithography, the most noteworthy examples are organic amines and ammonia in parts-per-billion concentrations, which poison chemically amplified deep-ultraviolet (DUV) photoresists.^1,2 Other researchers have reported that when a chemically amplified photoresist was exposed to a 17% ppb concentration of ammonia for 10 minutes, there was an ~20% change in linewidth.³ Unlike i-line photoresist, chemically amplified DUV resist forms an acid during the exposure step. Acid loss at the resist/air interface can then cause capping on the resist profile. Such a loss may be caused by a reaction with a base adsorbed from the air into the resist film or by acid evaporation during postexposure bake (PEB).⁴ If a material outgasses organic bases, such as amine, ammonia, or 1-methyl-2-pyrrolidone (NMP), the acid in the resist film can be neutralized. In extreme cases, the result is a complete insolubility of the film in developer, leaving only the latent image on the wafer.

There has been a consistent and largely successful effort on the part of photoresist suppliers to improve the resistance of the film to these contaminants. In addition, there has been a progression of changes in the filtration of the air supply to photo tools, and a chemical air filter integrated in the cluster tool lessens the possible airborne contamination level.⁵ In parallel, monitors have been developed to measure ambient contamination levels. Nevertheless, the materials used in the fabrication of lithography equipment and other materials in the process area must be tested for their potential to emit molecular contamination within the process chamber. Such materials include building materials, solvents, chemicals, epoxies, adhesives, tapes, cleaners, foams, rubbers, lubricants, oils, paints, coatings, sealant, and composite materials.

The chemical and physical properties of these materials vary considerably, and there is no easy way to know if a particular material will fail a compatibility test unless there are very obvious contaminants listed on its material data sheet. Complicating the issue further is the fact that almost all the materials present in IC fabrication areas are made up of mixtures. Even a tiny amount of other components in a known noncontaminating material such as PVC will affect its compatibility with the DUV resist since these photoresists are sensitive to contaminants in the parts-per-billion range.

Because of the reluctance of manufacturers to reveal all the chemical components of a product for proprietary reasons, the usual method of ensuring that a material is DUV photoresist compatible is to compare differences in linewidth on processed control wafers and test wafers that have been exposed to the material. The original procedure adopted for such testing was cross-section scanning electron microscopy (SEM), which requires breaking wafers. A desire to reduce test wafer consumption while providing repeatable, quantified results led to the development of CD SEM, a nondestructive linewidth measurement technique. Approval of a material using the CD SEM method is determined by evaluating the percentage change between linewidths on the control and test wafers. Although some profile information is not available with this method, CD SEM provides a more accurate way to determine whether a material meets pass/fail criteria than cross-section SEM, because interpreting such profiles is very subjective at low contamination levels.

Experiments

To compare the results that can be obtained from cross-section and CD SEM, a number of material samples were tested. The comparative research was conducted using the Micrascan step-and-scan lithography tool (Silicon Valley Group, San Jose) and a positive DUV photoresist at a thickness of 8000 Å. The Micrascan was connected with an SVG 90SE process track to provide a cluster environment, which contained a citric acid—impregnated, charcoal-filtered chamber. The experiments were not intended to result in optimum process control or resist profiles but to set up a simple, fairly quick way to quantify the effect of a test material on CD control when a process delay is encountered.

After an initial wafer was exposed and processed without induced delays to determine best exposure, two wafers were used for each subsequent sample: a control wafer and a test wafer. The control wafer was not exposed to the test material but was delayed 10 minutes during PEB in a clean desiccator within the citric acid—impregnated, charcoal-filtered enclosure. The test wafer was processed identically with the exception that it was placed in proximity with a sample of the material being tested during the postexpose delay time. The same dessicator was used for both types of wafers. A Model S8820 SEM (Hitachi Scientific Instruments; Mountain View, CA) was used to measure linewidth and an XL40 (Philips Analytical; Mahwah, NJ) for cross-section SEM. In both cases, 350-nm horizontal group lines were measured.

During the photolithography process, both the control wafer and the test wafer were exposed to the same dose of DUV photoresist in each of six fields (A, B, C, D, E, and F), which were set up to receive differing doses. Field C received the nominal dose (100%) for 350-nm group lines, while the doses for fields A, B, D, E, and F were 90, 95, 105, 110, and 115%, respectively. Each field was exposed in 15 0.2-µm focus steps. The resulting horizontal group lines in fields B, C, and D were measured using CD SEM. The measurements were taken at seven sites in each of these fields, and each site was measured in five focus positions. Then the average of each field was calculated. After the control and test wafers had been measured using CD SEM, they were broken and prepared samples were measured using cross-section SEM.

Wafer Type (nm)	Field B	Field C	Field D
Control wafer	389	376	365
Test wafer	455	430	420
% change	17	14	15

Table I: CD-SEM results for the control and test wafers shown in Figure 1.

Wafer Type (nm)	Field B	Field C	Field D
Control wafer	380.6	358.8	347.7
Test wafer	380.4	363.0	348.9
% change	0	1	0

Table II: CD-SEM results for the control and test wafers shown in Figure 2.

Results and Discussion

Two examples of cross-section and CD SEM results are provided here to demonstrate the differences and commonalities between the two approaches. Figures 1 and 2 show cross-section SEM results for the control and test wafers in test samples 1 and 2, respectively. Although the differences are not visually dramatic, Figure 1 shows some t-topping on the test wafer, which suggests that the material being tested is not compatible with DUV photoresist. The corresponding CD SEM results, which are listed in Table I, showed an average increase in linewidth of 15% on the test wafer compared to the control wafer, thus confirming the failure of this test material. Similarly, Figure 2 shows no significant visual difference in linewidth between the control and test wafers, and the corresponding CD SEM results, shown in Table II, provide quantitative evidence that the percentage change was insignificant. Based on these quantified results, it can be clearly seen that sample 2 passed the compatibility test.

Figure 1: Cross-section SEM profiles for a control wafer and a test wafer that failed. Thus, the test material would not be approved for use. The micrographs show 350-nm horizontal features in nominal focus.

A primary benefit of the CD SEM technique is that comparisons of the control and test wafer can be based on quantitative measurement data, whereas visual comparisons of cross-section SEM micrographs are sometimes subjective. With the availability of quantitative data, an easily understood pass/fail criterion can be established in terms of percentage change, and the confidence level in the test results will be high. CD SEM also is less susceptible to operator error than cross-section SEM, which requires very strict operational procedures that are difficult to control. For example, if the break in the test or control wafer is not perpendicular to the cross section of the line, the linewidth will appear larger than it actually is, thus creating the potential for false positives and false negatives.

CD SEM also offers other advantages over cross-section SEM for evaluating a material's potential for emitting airborne molecular contamination in lithography process areas. First, this method is more cost-efficient. Cross-section SEM requires the destruction of one send-ahead wafer for establishing exposure dose and two test wafers for each material being evaluated. In contrast, when CD SEM is used to measure linewidths, the test wafers can be recycled and reused many times. Thus, the cost savings associated with CD SEM can be substantial.

Figure 2: Cross-section SEM profiles for a control wafer and a test wafer that passed. Thus, the test material would be approved for use. The micrographs show 350-nm horizontal features in nominal focus.

CD SEM is more efficient than cross-section SEM as well. Obtaining linewidth measurements using CD SEM requires approximately 15 minutes per wafer; cross-section SEM requires about 1 hour because it involves breaking the wafer and preparing the sample in addition to taking the SEM picture. Also, using CD SEM enables operators to obtain more information. Because of the time required for cross-section SEM, only a limited number of scans can be taken for each wafer. Thus, using CD SEM improves productivity.

Conclusion

While there have been continuing, significant improvements in both photoresists and environmental air filtration, the presence of molecular contamination in lithographic process areas still poses a substantial risk to CD control. In order to manage this risk, lithographic equipment manufacturers and the fabs that utilize these systems need to establish testing programs to determine the compatibility of materials with DUV resist processing. Using CD SEM rather than cross-section SEM for such testing has the advantages of increasing the accuracy of test results, reducing testing costs substantially, and improving test operators' productivity.

References

1. SA MacDonald et al., "Airborne Chemical Contamination of a Chemically Amplified Resist," in Proceedings of Advances in Resist Technology and Processing VIII (Bellingham, WA: International Society for Optical Engineering [SPIE], 1991), 1—12.

2. JC Vigil, MW Barrick, and TH Grafe, "Contamination Control for Processing DUV Chemically Amplified Photoresists," in Proceedings of SPIE's International Symposium on Microlithography (Bellingham, WA: SPIE, 1995), 210.

3. Y Kawai et al., "The Effect of an Organic Base in a Chemically Amplified Resist on Patterning Characteristics Using KrF Lithography," in Digest of Papers from MicroProcess '94, the Seventh MicroProcess Conference, (1994), 202—203.

4. SA MacDonald et al., "Airborne Chemical Contamination of a Chemically Amplified Resist. 1. Identification of a Problem," Chemical Materials 5 (1993): 348—356.

5. OP Kishkovich and MA Joffe, "Proposing an Industrial-Standard Methodology for Testing Chemical Air Filters," MICRO 14, no. 6 (1996): 83—98.

Zhou Lin is a system engineer at Silicon Valley Group's lithography division in Ridgefield, CT. Since joining SVG in 1995 he has been involved in continuous improvement and development of photolithography systems, overlay and metrology improvement, CD and linewidth control, process development, contamination control, and photoresist evaluation and development. Lin holds an MS in chemical engineering with an emphasis on semiconductor materials processing from Arizona State University, Tempe. He also has MS and BS degrees in chemical engineering from Tsinghua University, Beijing. (Lin can be reached at 203/894-2171 or lin@svg.com.)

Anne F. VanNatter is a senior staff engineer at the Silicon Valley Group lithography division, where she manages the engineering systems lab, which includes the molecular contamination program. During her 22 years of experience in the semiconductor industry, including time spent at National Semiconductor, she has held various management and engineering positions with responsibilities in resist process development and contamination control.

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