Tracking down causes of DUV sub-pellicle defects*
06/01/2000
Tracking down causes of DUV sub-pellicle defects*
Brian J. Grenon, Grenon Consulting, Colchester, VT; Charles R. Peters, Dominion Semiconductor, Manassas, VA; Kaustuve Bhattacharyya, KLA-Tencor, San Jose, CA**
overview
Crucial to the continued success of DUV lithography, particularly for low k1 processes, there is a new class of defects that can significantly impact mask performance and semiconductor chip manufacturing yields. Researchers have put together the techniques and defect-detection systems needed to identify the presence of these sub-pellicle or pellicle-related defects. They also explain the mechanism of defect formation and micro-analytical results identifying both the composition and possible sources of defects.
 |
The photomask pellicle has perhaps been one of the most significant inventions of the semiconductor industry [1]. Its use has allowed photolithographers the opportunity to continuously use the same mask without the interruption of cleaning and reinspection. The use of the pellicle resulted in the requirement for fewer masks and increased yields in the wafer fab. While these benefits are realized by all fabs, a poorly constructed pellicle can result in unwanted defect problems. With the advent of low k1 and shorter wavelength lithography, the materials and space between the pellicle film and mask surface can create a highly reactive environment. This environment can provide the opportunity for particle growth, which results in the formation of "killer" (printable) defects on the mask.
Fundamentals of pellicle construction
Historically, the materials used to manufacture a pellicle film have changed with the wavelength used for photolithography. The earliest films were manufactured with Mylar, nitrocellulose, cellulose acetate, poly(vinyl butyral) and most recently, polyfluorocarbons. The move from one pellicle film to another was often driven by problems discovered by photolithography engineers in the fab. These problems were either loss of pellicle transmission or some type of photo-degradation induced by repeated exposures.
 Figure 1. An overview of the defects found on one reticle. The top and bottom images are transmitted and reflected light images. |
The basic components of a pellicle are the transparent pellicle film, pellicle frame, frame adhesive and pellicle film adhesive [2]. The advent of 248nm and 193nm lithographies have made the selection of materials used for pellicle construction critical. While the construction of the pellicle is somewhat universal, the specific materials used may vary significantly. Table 1 provides a list of materials that are used in pellicle construction. Each manufacturer of pellicles may use different materials and processes to produce the final pellicle. As a result, different photochemical reactions can occur during reticle exposure, dependent upon the materials selected. A pellicle manufacturer that understands the possible effects of poor material selection can properly engineer a stable pellicle structure.
The materials in Table 1 have the potential to react under prolonged use and exposure to DUV. Perhaps the best-known example of a pellicle-induced contaminant on photomasks was the identification of Santovar [2,5-Di(tert-amyl)hydroquinone] crystallization on chromium images. This material is an antidegradant used in commercially available rubber compounds. Rubber compounds have been used as pellicle gasket (adhesives).
The mechanism by which Santovar caused defect formation was a result of ultraviolet light exposure causing the compound to form a quinone derivative and sublime and re-crystallize on the edges of the chromium images that acted as excellent nucleation sites for crystal growth. Outgassing of pellicle frame adhesives or pellicle film adhesives appears to be a major source of sub-pellicle defects.
Pellicle defect formation, detection case history
While other conditions may exist that can create defects similar to those experienced in this study, we focus on the specific manufacturing conditions under which these defects were formed and were detected. Dominion Semiconductor is a volume manufacturer of DRAM chips. As such, the reticles can see very high usage; over one million field exposures/month
eticle are possible. The DUV exposure systems use 248nm-wavelength light, emitted by a KrF excimer laser that produces approximately 100 mW/cm2 at the reticle plane.
 Figure 2. Transmission loss for five defects on two reticles. |
About six months after the introduction of a new product, the engineering staff started seeing unusual defects appear during routine monthly reticle inspections on KLA-Tencor STARlight SL3 systems. This finding dictated an increased frequency for these inspections. These relatively small defects were observed on the chromium side of the reticle and only under the pellicle. The defects were first noticed on and around large clear areas on bright field masks. Several reticles exhibited this problem at about the same time. Over time these small defects increased in size and in number on the reticle. Eventually, they formed large branched defects that were 50-100mm in diameter. The initially detected defects were <1.0mm. This finding led to the belief that the defect formation was a possible photochemical reaction occurring under the pellicle. It should be noted that these defects were not detected on the back of the reticle or on areas outside of the pellicle frame.
 Figure 3. Sub-pellicle defects on glass: a) on the glass surface, and b) growing on the edges of the chromium. |
The defects were first detected using a routine inspection periodically performed on each reticle to ensure they are free of contamination. The inspection system uses simultaneous transmitted and reflected illumination and a special contamination detection algorithm to detect flat, semitransparent defects, stains, and transmission errors, both in clear and opaque areas on the reticle. A single laser scans the reticle on a perpendicular axis, and two photosensors (reflected and transmitted) collect the emitted light. The digitized outputs are processed using advanced algorithms and any unexpected condition in the intensities of illumination indicates a defect.
The detection of the defects reported herein needed no special set-up on the inspection system and a standard program using 0.37mm-pixel sensitivity could easily detect them. Even a lower sensitivity 0.5mm-pixel inspection proved sufficient.
Sub-pellicle defect characterization
We used several analytical approaches to ascertain the source and composition of the defects. First, STARlight inspection provided a clear understanding of the size and density of the defects found on several masks that were analyzed (Fig. 1 on p. 159). Figure 1 shows an inspector report where nearly 3500 defects were detected on the reticle in the clear areas in just a small portion of the reticle. This was the case on the other reticles inspected under the same exposure and inspection conditions.
We used the inspection system to determine the transmission loss characteristics of several of the defects on two different reticles to determine the potential effects of the defects on photolithography (Fig. 2). These data clearly show evidence that the sub-pellicle defects reduce transmission through the mask such that they would either print or cause a significant loss in exposure on the wafer. At low k1 lithography, all of the defects measured on both reticles would be "killer" defects. For this reason it is important that reticles be re-qualified prior to use.
We evaluated the defects using polarized light optical microscopy to verify the results obtained by the mask inspection (Fig. 3). This showed us the defects on the glass surface (Fig. 3a), growing on the edges of the chromium images (Fig. 3b). In Figure 3b it can be seen that the chromium edges act as nucleation sites for defect formation.
We found that the defects were too thin to image with low voltage SEM (500-100eV) on an uncoated sample of the mask. However, higher voltage SEM with the samples coated with gold-palladium provides a set of useful micrographs (Fig. 4a-f).
It is important to note that the morphology of the defects changed once the pellicle was removed. The original morphology was that shown in Figs. 1 and 3. Special care was taken to assure that SEM sample preparation did not alter the defects. However, exposure to air caused the defects to change from thin crystalline-type defects to round droplets as seen in Fig. 4a-f. Figure 4d clearly shows the impact of heat on the defects. The crack in the small droplet was a result of exposure to the electron beam of the SEM. As a result of these observations, we concluded that the defects reacted with the environment and formed a different compound or were hygroscopic and absorbed water from the environment. Additionally, sensitivity to the electron beam either implies that the defects had a low melting point or had undergone a thermally induced transformation.
 |
Fourier transform infrared spectroscopy of the pellicle from a defective mask and an unused pellicle showed that the pellicle film did not undergo any decomposition or transformation. In addition, we did ultraviolet spectroscopy on the pellicle from a defective mask to determine whether the film had any change in transmission as a result of the sub-pellicle defect formation. This data indicated no transmission loss occurred on a pellicle removed from a defective mask; transmission characteristics at 365nm and 248nm were unchanged. This result indicates that no degradation of the pellicle film occurred and confirms the FTIR results. From these results we concluded that decomposition of the pellicle film did not contribute to the formation of sub-pellicle defects.
Energy dispersive x-ray analysis (EDX) on defective and nondefective masks determined the differences in elements on surfaces of the masks and pellicles (Table 2). Table 2 gives results of the elemental analysis of points on a defective and nondefective mask.
The data in Table 2 show several differences in elemental analyses. The outer surface of the anodized pellicle frame on the defective mask contains nickel; we believe the source is nickel acetate used to quench the anodize process. The EDX analysis indicated that the defects found on the reticle tested contained sulfur, as opposed to the quartz background. The silicon and oxygen found at the site of the defects is a result of the electron beam penetrating the defects and going into the underlying quartz.
The elements found on the inner frame wall are from the anodized aluminum and the inner frame wall adhesive. The inner frame wall adhesive supposedly acts as a gettering material for particles inside the inner frame pellicle-reticle area. The EDX of the cross-sectioned frame was done to confirm whether the presence of sulfur on the frame surface was from the anodization process. The presence of carbon and oxygen on the bottom surface of the pellicle frame is from the pellicle frame adhesive.
 Figure 5. Scenario No. 1. The formation of an ammonium carbonate sub-pellicle defect. |
We conducted ToF-SIMS analysis on two reticles with sub-pellicle defects. While other contaminants were found on the mask surfaces, a high concentration of ammonium ions was present in the large defect. Additionally, the ToF-SIMS results indicate the presence of ammonium ions on the chromium surface. However, the concentration was lower than on the quartz surface. It is important to note that the ToF-SIMS analysis did not indicate the presence of any negative ions. Only positive ions were detected on the surfaces of the quartz and chromium, perhaps indicating that the ammonium ions are directly bounded to the respective surfaces. The "starburst-type" defects seen in Figs. 1 and 3 were particularly high in ammonium ion concentration.
Defects scenarios
 Figure 6. Scenario No. 2. The formation of an ammonium sulfate sub-pellicle defect. |
The analytical procedures used to evaluate the defects indicate that there are two types of defects present on the reticles. The SEM micrographs provide evidence of two different morphologies. Likewise, the results from the EDX and ToF-SIMS analysis indicate the presence of sulfur and ammonium ions, respectively:
- The "starburst-type" is a thin semitransparent defect that undergoes transformation when exposed to the atmosphere. The defect does not transform while the pellicle is on the mask surface. The transformation changes the defects from thin crystals to round droplets. These defects are thermally sensitive, and shrink and crack when exposed to an electron beam. EDX analysis indicated the presence of sulfur in the defect. ToF-SIMS also indicated the presence of ammonium ions.
- The second type of defect is a small, randomly shaped defect that contains ammonium ions, but no sulfur as indicated by the ToF-SIMS analysis. This defect appears to be uniformly dispersed on the chromium surfaces.
We formulated several possible scenarios for the formation of these defects based on the currently available analytical data. Other sub-pellicle defects could be formed by other reaction mechanisms. The possible sources of the contaminants are the pellicle film adhesive, the sulfur-based black anodized pellicle frame surface, pellicle frame adhesive, and the mask cleaning processes. These scenarios may vary from one pellicle and mask manufacturer to another: - Scenario No. 1, Fig. 5 (p. 165). The contaminant is formed from ammonia outgassing from pellicle frame adhesive or film adhesive reacting with sulfate contaminant, carbon dioxide, and water to form ammonium carbonate. Crystals formed by sublimation of ammonium carbonate and calcium carbonate or carbon dioxide. Ammonium carbonate is a mixture of ammonium carbamate and ammonium bicarbonate. It is a colorless, hard translucent crystalline mass that decomposes in air to give off ammonia and carbon dioxide and form ammonium bicarbonate. It volatizes at 60°C [3]. This may explain why the material changed morphology when the pellicle was removed and exposed to air. Moisture in the air could cause the chemical transformation to occur. Additionally, exposure to the electron beam of the SEM could generate enough heat to cause the crystals to undergo a change in structure. This possibly is consistent with the behavior of the crystals when the SEM analysis was done on the contaminants (this effect can be seen in Fig. 4d).
While there are other possible scenarios that can be developed to explain the formation of sub-pellicle defects, one or more of these four scenarios are the most plausible based on the information from the defect and reticle analysis.
Conclusion
 Figure 7. Scenario No. 3. The formation of an ammonium-sulfate ammonium-bisulfate sub-pellicle defect. |
As shorter wavelengths of light are used for photolithography, the selection of materials and processes used to manufacture masks becomes more critical.
 Figure 8. Scenario No. 4. The formation of an ammonium sulfate sub-pellicle defect. |
The increased use of DUV lithography will most certainly create similar situations to the one reported here. Future introduction of 193nm lithography will only exacerbate the potential for more sub-pellicle defect formation on production reticles. This evaluation provides some insight into the possible catastrophic failures that can occur when potentially undetected reticle defects are formed during wafer production. Mask manufacturers need to develop materials and processes that do not create potentially high-risk situations in the wafer fab. Until such materials and processes are developed and understood, routine re-qualification of production reticles is suggested. Early detection of these sub-pellicle defects can prevent potentially costly product losses. It is also important to note that even though reticles are shipped from mask shops defect-free, there is no guarantee that they will remain that way once they are used in production.
Acknowledgments
William Volk, director of marketing at KLA-Tencor, is also an author of this article. We thank Larry McKinley, Glenn Storm, and Janice Paduano of DuPont Photomasks for providing their expertise and analysis of the pellicles; Bill Riley of SEM Consultants and George Bruno of KLA-Tencor AMRAY Division for the SEM and EDX analysis; George Slusser, Harold Linde, and Jay Burnham of IBM Microelectronics for FTIR and ToF-SIMS support; and Antonio Gallo and Riaz Mahjoor of Dominion Semiconductor.
Mylar is a registered trademark of DuPont. STARlight is a registered trademark of KLA-Tencor.
References
- V. Shea, W.J. Wojcik, "Pellicle cover for projection printing system," US Patent 4, 131, 363, December 26, 1978.
- J.S. Gordon, "Pellicles designed for high performance lithographic processes," SPIE, Vol. 2512, pp. 99-111, 1995.
- M. Windholz, Editor, The Merck Index, Ninth Ed., Merck and Co., Rahway, NJ, pp. 69-70, 1976.
*Based on a paper presented at BACUS '99 and published by permission of SPIE.
**Additional authors are listed in the Acknowledgments.
Brian J. Grenon received his degree in chemistry from the University of Vermont. He was responsible for mask technology development at IBM's Essex Junction, VT, facility for 19 years. Grenon is an independent consultant on mask technology. Grenon Consulting Inc., 92 Dunlop Way, Colchester, VT 05446; ph 802/862-4551, fax 802/658-8952, e-mail [email protected].
Charles R. Peters received his BS from the University of Illinois and his MS in physical chemistry from Auburn University. Peters is a senior process engineer at Dominion Semiconductor LLC.
Kaustuve Bhattacharyya received his MS in chemical engineering from New Mexico State University. He is an applications engineer at KLA-Tencor Corp.