 |
 |
 |
 |
|
Comments? Suggestions? Send us your feedback.
MicroMagazine.com |
Ultrapure Photochemical Processing
Investigating defect mechanisms for a low-volume resist coating recipe with a solvent prewet stepChris Lansford, Clint Miller, and Dan Sutton, Advanced Micro Devices; and Jason Zanotti, formerly with AMD Monitoring of defect excursions and research into their generation mechanisms can facilitate the integration of a cost-saving prewet step into a photoresist process. As i-line and deep ultraviolet (DUV) photolithography processes have grown more complex, the cost of the photoresists used in these procedures has also increased. More masking layers are also required, adding further to a fab's per-wafer expenditures. To help keep costs down, IC manufacturers and tool vendors have developed special coating recipes and pump parameters that reduce dispense volumes somewhat, but this approach has limitations since there must be a sufficient resist volume to prevent edge voiding. One technique that could potentially reduce resist volumes even further is the use of a solvent prewet step to overcome the surface energy of the primed wafer. Because the addition of a chemical step to the resist coating recipe can produce defects that can adversely affect die yield, however, an understanding of the mechanisms for defect generation in the solvent prewet process is critical. To further that understanding, this article shares some of the information Advanced Micro Devices (AMD) has learned during its implementation of the prewet step at Fab 25 in Austin, TX. Examples of potential defects are discussed, and the results of experiments in which process parameters were varied in an attempt to control a defect mechanism are presented. Such factors as volume, dispense time, and spin speeds as well as the solvent type and the vapor primer were examined and found to be significant. Background Implementation of the solvent prewet technique for reducing photoresist dispense volumes was accomplished at the AMD fab using existing facilities and equipment, although some modification of the track systems—Mark 8 and ACT8-series tools from Tokyo Electron (TEL; Austin, TX)—was required. The solvent chosen was already being used for backside wash and edge-bead removal to improve photoresist solubility on the wafer edge. The two major components of this commercially available solvent are ethyl lactate and 4-methyl-2-pentanone. Other components may also be present as by-products of the manufacturing process. The solvent has been used at the site for some time and is available through the bulk chemical delivery system. The low-volume resist process implemented was basically the vendor-recommended recipe, with slight modifications made to compensate for site-specific differences such as temperature and choice of photoresists, which can affect coating uniformity. The process consists of dispensing a small volume of solvent on the wafer (as shown schematically in Figure 1a), then spinning it off at a high speed. As the solvent is evaporating on the surface, the photoresist is applied (Figure 1b), also at a high spin speed. The solvent vapor allows the photoresist to wet the wafer surface more readily, overcoming the surface energy introduced earlier by vapor priming. This dispense process has proved very stable for producing consistent wafer-to-wafer resist coatings at greatly reduced resist volumes. Defect Mechanisms During the first phase of the implementation project, defect density studies were performed to qualify the new low-volume resist process, and split-lot results indicated it was comparable to the previous recipe. When implementation proceeded over additional tracks, however, defect excursions that were possibly related to the solvent prewet step began to occur. Such events were expected, because of the large number of parameters that are involved in the process recipe, the critical timing between the solvent prewet and resist dispense steps, and the advanced age of many of the fab's photoresist track tools. Several types of defects were identified and their generation mechanisms were investigated.
Embedded Particles. The defect densities on wafers processed by the various tracks were charted using phototrack monitor wafers. These wafers are processed and then scanned at a very high magnification to detect embedded and surface defects associated with the resist coat, patterning, and development.1,2 When the low-volume resist process was released to some older tracks, the baseline defect density immediately shifted slightly upwards and took some time to return to normal. A typical control chart for such tools is shown in Figure 2.
Microscopic inspection revealed that the defects were very small, pointlike particles in or under the photoresist. To determine if the particles were added during the solvent prewet step, some unpatterned monitor wafers were processed through the coating modules with solvent dispensing only, and others with photoresist dispensing only.3 In the latter runs, fewer than five particles were added for a typical wafer, indicating that the defect source was the solvent step. A physical examination of some of the older tracks was then performed, and the solvent dispense valves and lines that had not been used for some time were replaced, which solved the particle generation problem. As mentioned above, the normal sequence for these particle tests involved sending the test wafers through the coating modules only. However, on one series of tests, the vapor prime, or adhesion, modules were also included. As the sample wafer map in Figure 3 shows, there was a large increase in defects on these test wafers. In addition, tracks using different vapor primers were shown to have different defect levels. It is believed that these results are related to the multiple components of the prewet solvent mixture. Since the various components have slightly different evaporation rates, the solvent may break up into droplets, forming droplet residues. There might also be a small amount of chemical residue in the solvent (similar to the dissolved carbon in ultrapure water), so the defects may be similar to water droplet defects. The generation mechanism for chemical-residue defects is discussed in more detail later in this article.
Solvent Aerosol Particles. Because an additional chemical is being dispensed onto and spun off the wafer during the solvent prewet step, aerosol formation is also a concern. The same mechanisms that generate photoresist aerosol particles can lead to solvent aerosol particles.4 If the photoresist coating track is properly maintained, solvent aerosol particles are generally rather benign. However, if the cup exhaust is not controlled properly, or the splash rings and cup are not maintained, any aerosolized solvent can be redeposited onto the wafer during the resist application. Fortunately, there was very little evidence of this effect following implementation of the low-volume process recipe. One subtle problem related to aerosolized solvent did arise: the cup design had to be reconfigured to handle the additional solvent so that wafer backsides would not be contaminated by aerosol particles from either the solvent or the photoresist. Figure 4 shows a wafer map of a backside contamination monitor for a DUV process, along with a micrograph of one of the many aerosol particles detected on the wafer. This backside contamination problem has been described in detail elsewhere.5
Color Variations. A more subtle type of defect seen on a macro level was called the parentheses defect, since it appeared on wafer maps as one or two arc shapes, as seen in Figure 5. Microscopic review revealed that the shapes were created by slight color variations. Comparative resist-thickness measurements were then performed on wafers processed using the new and old resist coating recipes, and the results indicated that thickness variations occurred only on the solvent prewet wafers. It appeared that too much solvent was being dispensed during the prewet step, so the subsequent spin-off period was insufficiently long to remove all of the liquid. When the resist was dispensed, the remaining solvent would interact with it, causing localized viscosity variations, which were seen as color variations. This phenomenon is of great concern for device layers where linewidths must be strictly controlled, since linewidth is modulated by the resist thickness. Reducing the volume of solvent dispensed and making daily dispense-volume calibrations helped to prevent the thickness variation problem from recurring.
Chemical-Residue Defects. A fourth type of defect was not readily detectable on the patterned phototrack monitor wafers, but could be seen on product scans after regions of oxide and silicon had been defined. Specifically, very thin scumming defects were detected on a particular product at a particular masking step. An example is shown in Figure 6. These residues appeared to be design-layout sensitive, but were not killer defects.
The initial attempts to prevent such defects involved increasing dose, development, and DI-water rinse times, but none of these parameters appeared to have the desired effect. To determine whether the defects might be related to the solvent prewet, a test similar to the one used for embedded particles was developed and run, but instead of monitor wafers, product wafers were processed to investigate the effect of topography and layout. Wafers were vapor primed using one of two chemical formulations—100% hexamethyldisilazane (HMDS) and HMDS plus N,N-diethylaminotrimethylsilane (DEATS)—and then subjected to only the solvent prewet portion of the recipe. Examples of the respective results are shown in Figure 7. These test results suggested that the defect mechanism was related to the solvent prewet and depended strongly on the surface energy of the vapor primer.
To investigate this defect mechanism further, a series of experiments was performed in which the parameters in the solvent prewet portion of the recipe that affect solvent evaporation were varied. The principal parameters considered were the solvent spin-off time prior to the application of the photoresist, and the spin speed during this step. Both affect the amount of solvent or solvent vapors present on the wafer surface as the photoresist is dispensed. Duplicate test runs were performed using the two types of vapor primer. Results for the test runs that used 100% HMDS are shown in Table
I. Many of these test wafers showed no chemical-residue defects, and the
experiment was considered inconclusive because the defect mechanism was
not prevalent when this chemical was used for priming.
It was known from previous defect studies involving DI-water rinse residues that droplet formation is likely to occur when DI water is dispensed on a wafer primed with DEATS, and therefore it was suspected that such wafers would also be susceptible to the solvent prewet defect mechanism. The tests run with the HMDS-plus-DEATS primer confirmed this hypothesis. The results, shown in Table II, revealed that for a HMDS-plus-DEATSprimed surface, which has a very high contact angle, the longer and faster the prewet solvent spin step was, the more defects were produced.
The proposed explanation for this variation is that during spinning, the pressure at the wafer surface decreases enough to induce solvent evaporation, which enables a low-vapor-pressure, low-concentration component of the solvent to become concentrated enough to precipitate out of solution onto the wafer. Because the tendency of the highly thinned prewet solvent to form droplets just before evaporating to dryness is much greater on a high-contact-angle surface, the precipitate is more localized on the surfaces of HMDS-plus-DEATSprimed wafers than on 100% HMDSprimed surfaces. It is possible that this effect is exacerbated by the use of a two-component prewet solvent. For example, if one component evaporates before the other, the surface energy could still be adequately depressed for the solvent prewet step to be effective. If the low-concentration solvent component is less soluble in the remaining primary component than in the original mixture, it could be more likely to precipitate out of solution rather than be spun off the wafer. Regardless of the precise mechanism, it seems clear that a residue remains on HMDS-plus-DEATSprimed wafers in concentrations high enough to interfere with photoresist development. Future work will evaluate whether similar defects occur with a single-component prewet solvent. Conclusion As photoresists continue to increase in complexity and cost, IC manufacturers and their pump and track suppliers will continue to explore ways to reduce chemical consumption. Different techniques for applying low volumes of photoresist to the wafer may exhibit different defect mechanisms, however, and defect excursions may be related to subtle differences in track parameters when new recipes are used across multiple tools. When a photoresist coating recipe with a solvent prewet step was implemented at AMD, the defects that were discovered included particles embedded in the solvent, solvent aerosol particles, color and resist thickness variations, and chemical-residue defects that were dependent on the vapor primer. Wafers primed with a vapor containing DEATS were more likely to have residues than those primed with a 100% HMDS vapor. It was also found that the parameters used in the prewet step had an effect on the defect mechanisms. On the other hand, because the fab had an overall defect control strategy in place that combined monitor wafers and product wafer scans, the defect excursions related to implementing the low-volume resist process could be studied, understood, and controlled. Such capabilities are critical to the successful integration of any new methods for reducing photoresist consumption. Acknowledgments This article was originally presented at Interface 2000, Arch Chemicals Microlithography Symposium, San Diego, November 2000. Used with permission. The authors would like to thank Greg Goodwin, John Lepper, and Bruce Pickelsimer for their encouragement of this project; the Fab 25 technicians who assisted in the processing and analysis of the photoresist track monitoring data and also helped with the related experiments; and Carol Midboe, Tim Jackson, Christine Fischer, and the fab's maintenance technicians, who were instrumental in the implementation of the solvent prewet process. References
Chris Lansford, PhD, has worked in the diffusion, photolithography, and etch areas at Advanced Micro Devices in Austin, TX, since January 2000. He has a BS in chemistry and a BA in Plan II from the University of Texas (Austin) and also holds a PhD in analytical chemistry from the University of Illinois in Champagne-Urbana. (Lansford can be reached at 512/602-6452.) Clint Miller is a senior process engineer at AMD in Austin and a member of the Fab 25 photolithography department, in charge of TEL ACT8 and Mark 8 coat/develop tracks. He has worked for AMD since 1995, with previous responsibilities in the areas of shift sustaining and shift supervision. He received a BS in engineering physics from Southwestern Oklahoma State University in Weatherford. (Miller can be reached at 512/602-1296 or clint.miller@amd.com.) Dan Sutton is a senior member of the technical staff at AMD in Austin. He is a member of the Fab 25 photolithography department, where he is in charge of tool-based monitors and defect reduction on all photo equipment. He has worked for AMD since 1985, with responsibilities for coat/develop tracks, spin-on glass tracks, wafer scrubbers, 1´ scanners, mix-and-match programs between 1´ scanners and 5´ reduction steppers, and defect reduction. He has a BS in physics from Purdue University (Lafayette, IN) and an MSEE from the University of Texas (Austin). (Sutton can be reached at 512/602-4858 or dan.sutton@amd.com) Jason Zanotti completed several cooperative stints with AMD and joined the company full-time in its engineer rotational program following graduation from the University of Idaho, where he obtained a BS in chemical engineering. He recently joined Intel as a process engineer, where he works with TEL ACT12 track systems. (Zanotti can be reached at jasonzanotti@yahoo.com.)
 |
Buyers Guide | Media Kit
All rights reserved.