R Inspection E
T
I
C
L
E
The Case of the “Growing” Reticle Defect at 193-nm Lithography Kaustuve Bhattacharyya, and William Volk, KLA-Tencor Corporation Brian Grenon, Grenon Consulting Inc. Darius Brown, and Javier Ayala, IBM Microelectronics Corporation
Defect formation on advanced photomasks used for DUV lithography has introduced new challenges at low k1 processes industry wide. Especially at 193-nm scanner exposure, the mask pattern surface, pellicle film and the enclosed space between the pellicle and pattern surface can create a highly reactive environment. This environment can become susceptible to defect growth during repetitive exposure of a mask on DUV lithography systems due to the flow of high energy through the mask. Due to increased number of fields on the wafer, a reticle used at a 300-mm wafer fab receives roughly double the number of exposures without any cool down period, as compared to the reticles in a 200-mm wafer fab. Therefore, 193-nm lithography processes at a 300-mm wafer fab put lithographers and defect engineers into an area of untested mask behavior.
Introduction
The photomask pellicle has perhaps been one of the most significant inventions of the semiconductor industry.1 A pellicle is made of transparent polymer film that protects the pattern surface of the mask from outside contaminants. This allows lithographers to use a mask for extended period of time 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, with the advent of low k1 and shorter wavelength lithography, the materials and environment trapped between the pellicle film and the mask surface can create a highly reactive environment. This environment can trigger photochemical reaction providing the opportunity for particle growth, which results in the formation of “killer” (printable) defects on the mask 2 (Figure 1).
The defect growth mechanism could be a function of the following factors: individual component of a mask (chrome, quartz, attenuator, pellicle, pellicle frame, adhesives etc.), reticle container,4 storage, fab environment, stepper environment, mask cleaning process residuals (reactive ions left on the mask surface after cleaning processes), cumulative energy through the mask (due to repetitive exposures of the mask on scanners), wavelength of the exposure and possibly time. It is definitely not a trivial task to explain this complex mechanism that may not only involve many of these above factors, but also their interaction. Authors of this article have seen and reported reticle defect growth cases at I-line (365-nm) and UV (248-nm) wavelength scanner exposures. For 248-nm exposure a
F i g u re 1. Components of a photomask .
78
Summer 2003
Yield Management Solutions
R
severe defect growth problem was found and was traced back to the ammonia outgassing from the pellicle frame adhesive.2 For I-line scanner exposure, we continue to see similar defect growth problems: masks arrive from mask houses perfectly clean (determined by fab incoming inspection) and over the course of only few hundred wafer exposures in the fab, some masks show catastrophic defect growth. But the scenario at DUV (193-nm) is more serious; especially in a 300-mm wafer fab, where the reticles are exposed almost twice as much to print a wafer (as 300 mm wafers have roughly about double the F i g u re 2. TeraStar number of chips compared to 200-mm wafers). Due to the significance of this problem at 193-nm scanner wavelength exposure, the authors decided to run the following experiment. The experiment and the results are explained below. Experiment
An attenuated PSM mask (gate level) was repetitively exposed on a 193-nm scanner using 200-mm wafers. A reticle defect inspection was performed using KLA-Tencor’s TeraStar (STARlight) tool after every hundredth wafer exposure. Within the experimental conditions we have over-exaggerated the normal use conditions by using a higher dose on the scanners to speed up the process.
E
T
I
C
L
E
(STAR light) arc h i t e c t u re .
Results The total number of defects (patterned surface defects) was recorded after each inspection and was plotted against the number of wafer exposed. Defects started to grow on both clear and attenuated surfaces, but the major intensity of growth remained on the clear and especially clear/ attenuator interfaces. A slow growth in defect count was observed until the 600 th wafer exposure, followed by an explosion in defect count at the 700th exposure (Figure 3).
Inspection Method An important criteria of this experiment was detection of any defect increase on the clear area of the mask, as well as detection of defect increase on the attenuator and chrome surface of the mask. The TeraStar (STARlight) inspection system used in this study utilizes simultaneous transmitted and reflected UV illumination and a special contamination detection algorithm to detect particles, stains, thin-chrome, ESD, flat, semi-transparent defects, and transmission errors, both in the clear and opaque areas (Figure 2). A single UV laser source scans the photomask on a perpendicular axis, and two photo-sensors (reflected and transmitted) collect the emitted light. The two digitized outputs are processed using advanced algorithms, and any unexpected condition in the intensities of illumination indicates a defect. Thus, it was possible to detect any defect increase not only on the clear area of the mask but also on the attenuator and chrome surface of the mask. A 0.25-micron pixel inspection was used to achieve 180-nm defect sensitivity on the mask.
F i g u re 3. Number of 200- mm wafe rs exposed vers us d efect gro w t h .
The amount of energy at the reticle plane (corresponding to the number of exposures from above figure) can be calculated from the following equation: Pellicle_dose =
Wafer_dose • Fields/wafer • number_of_wafers Mag2 • Lens_transmission
(Wafer Dose = 20 mJ; Fields/Wafer = 69; Mag = 4; Lens_transmission = 0.5)
Summer 2003
Yield Management Solutions
79
R
E
T
I
C
L
E
F i g u re 4. Energy at the reticle pl ane versus the defect growth.
Energy at the reticle plane versus the defect growth (Figure 4) shows a correlation until just before the defect explosion. The following pictures show the defects captured by the STARlight tool. The reticle inspection after the 700th exposed wafer showed a drastic increase of defects on the reticle (Figure 5).
F i g u re 5. STARlight inspection resul t showing greater than 10,00 0
Most of the defects are located in the clear area or at clear/attenuator interface. The majority of all the defects consist of two distinct kinds: one that looks like a large contamination (Figure 6) and the other resembles a comet with a long narrow tail and a prominent nucleus (Figure 7).
Raman spectroscopy is a light scattering technique often referred to as the sister of complementary technique to infrared spectroscopy. Raman spectroscopy provides vibration information about the analyzed compounds and depends on a change in the induced dipole moment or polarization to produce Raman scattering. When a beam of photons strikes a molecule, the photons are scattered
F i g u re 6. A defect (0.5 µm x 1 µm) showing 13 percent transmission loss.
F i g u re 7. A very narrow comet defect with 5 percent transmission loss.
80
Summer 2003
Yield Management Solutions
defects after 70 0 wafers were exposed.
Defect composition analysis – Raman spectroscopy
R
E
T
I
C
L
E
elastically (Rayleigh scattering) and inelastically (Raman scattering) generating Stokes and anti-Stokes lines. The Raman spectra were collected using a Renishaw Model 2000 Raman spectrometer equipped with a 633-nm laser. All measurements were made without removing the pellicle. The beam size is approximately one micron in diameter. The results show that there are three different kinds of defects involved with this growth – ammonium sulfate, cyanuric acid, and an unknown compound (Figure 8). Background Raman spectra on mask surface (a nondefective spot) were collected from both inside and outside of the pellicle covered area to cancel out the possibility of contamination due to transport and handling of the analysis (Figures 9 and 10). Background spectra for this plate can be compared with the spectra on a defect of interest (this defect was one of those large contamination type growths that was found by STARlight inspection). This clearly shows two extra Raman shift peaks at 975 cm-1 and 705 cm-1 (Figures 11 and 12). These peaks correspond to ammonium sulfate and cyanuric acid respectively.
F i g u re 9: Background Raman spectra on mask.
F i g u re 11. A def ect used to collect R aman sp ectra.
F i g u re 8. Ra man Scattering.
Figures 13 and 14 show the second defect of interest (this was the comet type defect) and its Raman spectra indicating intense peak at 975 cm-1 and 703 cm-1. This also indicates the presence of ammonium sulfate and cyanuric acid. There was also another defect, which was analyzed with Raman spectroscopy: Figures 15 and 16 showing different composition with two peaks at 1064 cm-1 and 790 cm-1. The compounds for these peaks are not identified yet. The authors believe that these could possibly be organic compounds.
F i g u r e 10. Background sp ectra o n mask outside th e pellicle.
F i g u re 12. Raman spectra indicating pea ks at 975 cm - 1 and 705 cm - 1 .
Summer 2003
Yield Management Solutions
81
R
E
T
I
C
L
E
F i g u re 13. A defect shaped as a comet.
F i g u re 14. Raman spectra on the nucleus of this comet defect indicating p e a k s at 975 cm - 1 an d 70 3 cm - 1 .
F i g u r e 15. Another defect for a nal ysis.
F i g u re 16. Spect ra of unknown def ect at 1064 cm - 1 an d 790 cm - 1 .
So, the above results show that there are three different kinds of defects involved with this growth – ammonium sulfate (Raman peak at 975 cm-1), cyanuric acid (Raman peak at 703 cm -1 or 705 cm-1) and some unknown compound(s) (with two peaks at 1064 cm-1 and 790 cm-1). In the following section the mechanisms involved in these kinds of defect growth are described. Defect formation mechanism — search for a root cause
As found from the defect composition analysis, the majority of the defects are cyanuric acid and ammonium sulfate. This points towards the participation of ammonia (possibly introduced through the rinse water or final rinse), carbon dioxide (ambient) and sulfuric acid (introduced during the sulfuric acid/hydrogen peroxide strip process).
or ammonium ions and carbon dioxide (CO2) occur. A possible source for ammonia could be the residual ammonium ions remaining on the mask surface after the rinse. However, due to the fact that a very dilute ammonium hydroxide is used in the rinse water, (in the equilibrium) water has a significantly higher presence than the NH 3. It is important to note that ammonium ions are relatively labile ions and can easily migrate in a highly reactive environment. Any conceivable mechanism should involve a dehydration reaction, and that is the case when a reticle endures high energy passing through it during the exposure on a scanner. Figure 17 illustrates a direct mechanism for cyanuric acid formation:5
Cyanuric acid formation — ammonium ions, carbon dioxide, water Cyanuric acid (C 3O3N3H3) can be formed on mask surface if a chemical reaction between ammonia (NH3) 82
Summer 2003
Yield Management Solutions
F i g u re 17. Cyanuri c acid (C 3 O 3 N 3 H3 ) .
R
E
T
I
C
L
E
Other possible mechanisms Other possible sources such as reticle container,4 storage, fab environment, stepper environment can also contribute to the root cause and should also be considered.
F i g u re 18. Cyanuric acid formation (one of the majority of the defects) .
Cyanuric acid formation — photo-degradation of cyano acrylate pellicle film adhesive While this scenario may not be as likely as previously described one, cyanoacrylate adhesive contamination can occur from the out-gassing of these materials (Figure 18). Although it may not be clearly obvious that the degradation of these compounds could result in the formation of cyanuric acid, it should be noted that their degradation could result in the elimination of a labile nitrile group. Cyano-acrylate adhesives are monomeric cyanoacrylates that polymerize spontaneously in the presence of moist air (Figure 19).
F i g u re 19. A nother pos sible mechanism f or cyanuric acid form a t i o n — photo-degra dat ion of pellicle ad hesi ve.
Ammonium sulfate formation — ammonia, ammonium ions, sulfuric acid or sulfate ions, water Residual ammonium ions left on the mask surface from rinse reacts with the residual sulfuric acid or sulfate ions from strip process in a high energy environment is a possible explanation for ammonium sulfate [(NH4)2SO4] formation illustrated in Figure 20.
F i g u re 20. Ammonium sulfate for ma tion (the second highest type of defects).
Defect formation is a dynamic process
It takes multiple exposures on 193-nm scanners to trigger the above-explained chemical reactions (by exceeding the necessary activation energy for each reaction). The goal should be to detect the defect growth well ahead of time, when the defects are just starting to form and the transmission loss through these nascent defects is not enough to make them printable defects yet. Following are pictures from a similar experiment being conducted using a COG mask (a different mask than the one used for experiment above). This is a chrome-on-glass mask with 193-nm pellicle and is being exposed on a 193-nm scanner. Incoming fab STARlight inspection showed that this was a clean mask with no defects. After forty 300 mm wafers exposed at 100mJ/field (equivalent to 160 wafers at 25mJ/field), another STARlight inspection was performed and some defects were detected (near the chrome/clear edge) that look like a yellowish stain, as shown in Figure 21. But the transmission loss was not enough to make these defects a critical problem at that point. After another nineteen 300-mm wafers were exposed at 100mJ/field (total of 236 300-mm-wafers at 25mJ), the STARlight inspection showed that more prominent defects (about 0.5 micron in size and some with even 75 percent transmission loss) started to form (Figure 22). Conclusion
With the 193-nm stepper process, lithographers are entering an untested territory of mask behavior. Especially at a 300-mm wafer fab, due to the increased number of fields on the wafer, a reticle receives roughly double the number of stepper exposures compared to a reticle at a 200-mm wafer fab. Therefore, the 193-nm stepper process at a 300-mm wafer fab not only introduces a new wavelength of light for mask exposure, but reticles in such a fab also endure more prolonged exposure at this wavelength, without as much cool-down time between wafer-to-wafer exposures as at a 200-mm wafer fab. The above experiments showed that at 193-nm scanner illumination, the enclosed space between the pellicle film and the mask pattern surface can create a highly reactive environment that, at a certain energy level (activation energy), can trigger photochemical reaction, forming critical defects on the mask. Within Summer 2003
Yield Management Solutions
83
R
E
T
I
C
L
E
More prominent defects are forming with the increase in reticle exposures
F i g u re 21. Reticle defect image aft er 160 modeled wa fers wer e
F i g u re 22: Defect image after 236 modeled wafers were exp osed
exposed (300 mm wafers).
(3 00 mm wafer s).
the experimental conditions, we have grossly overexaggerated the normal use conditions by using a higher dose on the scanners to speed up the process. For example, due to the extreme nature of the exposure conditions used in this experiment, the exposure versus defect growth relationship (Figure 3) might not be applicable for another mask at normal operating conditions. However, the above investigation provides some insights to the seriousness of this defect-growth problem on masks at 193-nm exposure wavelength, and reveals the composition of some of these defects. In addition, it also tries to highlight the possible formation mechanisms of this defect growth; however, more work needs to be done on this subject to understand all the possible scenarios involved. This defect growth could be a function of the many following factors: individual components of a mask (chrome, quartz, attenuator, pellicle, pellicle frame, adhesives etc), reticle container,4 storage, fab environment, stepper environment, mask cleaning process residuals (reactive ions left on the mask surface after cleaning process), cumulative energy through the mask (due to repetitive exposures of the mask on scanners), wavelength of the exposure, mask level pattern density (clear field/dark field) and, possibly, time. It is a very difficult task to explain all the aspects of the interactions among the above-mentioned factors. It is almost impossible to keep all the possible contributing factors (that are responsible for defect growth) constant between two experiments; hence, another experiment may show a significant defect growth at a different exposure or energy level. Understanding the mechanism of this defect growth is very important, and this may help the mask makers to develop materials and processes that can possibly prevent such defect growth; but, this is not a trivial task. Until such materials and
processes are developed and tested, there is no guarantee that masks shipped defect-free will remain so once they are used in production. The ideal reticle quality control goal should be to detect the defect growth well ahead of time when the defects are just starting to form, and the transmission loss through these nascent defects is not high enough to make them yield limiting yet. To achieve this, it is recommended that a carefully developed reticle inspection strategy be implemented with the goal of minimizing meantime to detect (MTTD) any defect growth resulting from prolonged reticle use.
84
Summer 2003
Yield Management Solutions
Acknowledgements
The authors would like to thank John Muscatel and Jose Estabil of KLA-Tencor for their valuable contributions. 2002 BACUS "Best Paper Award." BACUS Symposium on Photomask Technology, October 2002, Monterey, California, USA. References 1 . V. Shea, W. J. Wojcik, “Pellicle Cover for Projection Printing System”, US Patent 4, 131, 363, December 26, 1978. 2 . B. J. Grenon, C. R. Peters, and K. Bhattacharyya, “Tracking Down Causes of DUV Sub-pellicle Defects”, Solid State Te c h n o l o g y, June 2000. 3 . J. S. Gordon, “Pellicles Designed for High Perf o rmance Lithographic Pro c e s s e s ” , SPIE, Vol. 2512, pp. 99-111, 1995. 4 . T. Kozeki, S. Shigematsu, H. Nakagawa, “Longevity of 193-nm/ArF Excimer Pellicle (Influences of Organic Va p o r s ) ” , Photomask Japan, April, 2000. 5 . U l l m a n n ’s Encyclopedia of Industrial chemistry, Sixth Edition, June 2001.