The impact of airborne molecular bases on DUV photoresists
08/01/2001
David Ruede, Extraction Systems Inc., Franklin, Massachusetts
Monique Ercken, Tom Borgers, IMEC, Leuven, Belgium
overview
Exhaustive examination of resist characteristics for well-known parameters such as isolated and dense feature resolution, adhesion, and etch selectivity is ongoing. Sensitivity of resists to airborne molecular bases adds yet another parameter to be characterized. In an effort to bring the latest-generation processes into production more quickly, many device manufacturers have implemented several levels of safeguards to protect resists from bases such as ammonia, NMP, and TMA. This article compares the sensitivities of four commercially available DUV resists to a number of airborne molecular bases.
 |
Over the past five years, the semiconductor industry has undergone numerous changes with the advent of 0.25µm geometries. Exposure tools have moved from i-line to deep ultraviolet (DUV), and a new vocabulary containing words like chemically amplified, photo-acid generator, T-topping, and molecular bases has become part of the process engineer's jargon. New issues have emerged also. Critical dimension (CD) shifts witnessed in the early stages of DUV process development have become the source of intense study and numerous publications [1-3]. Chemically amplified DUV photoresist sensitivities to molecular bases such as ammonia, 1-methyl pyrrolidione (NMP), trimethylamine (TMA), and related compounds have led a search for production-worthy solutions, resulting in decreased sensitivity in some resists to molecular base exposure [4].
The selection of resist formulations often includes sensitivity to T-topping as one criterion, although this characteristic will often have a lesser priority than the ability to image the required features or resistance to etch processes. In order to gain some insight into actual production processes, this article characterizes the relative sensitivity of some commonly used DUV resists in an effort to provide a basis for comparison and a discussion of the needs for airborne molecular contamination monitoring and control.
Test procedure
In this study, photoresists available at IMEC's deep submicron silicon pilot line were selected: three 248nm formulations and one 193nm formulation. Each resist was coated onto 200mm wafers on a BARC layer using spin-on standard processes. The resist layers were exposed to molecular bases, generally consisting of ammonia as the major constituent and amines in lesser concentrations, in a cleanroom environment for 15, 30, and 45 min. Several die on each wafer were exposed at best focus and at various energies to determine optimal dose. The resists were characterized in terms of linewidth and CD changes for isolated and dense lines. Control wafers were placed inside a chemically filtered exposure system enclosure to separate acid diffusion from airborne contamination. During the test period, total molecular base concentrations were continuously monitored both in the cleanroom and in the exposure system environmental enclosures.
 Figure 1. Total molecular base concentration at the ASML 5500/300 wafer stage remains stable compared to a cleanroom environment. |
Wafers were exposed on two DUV tools, an ASML 5500/300B KrF stepper and an ASML 5500/900 ArF scanner, upgraded with a lens used in the ASML 5500/950 scanner. Resist processing was performed on two Tokyo Electron Ltd. track systems, a Mark 8 for the 248nm processes and an Act 8 for the 193nm processes. Molecular base concentrations were detected and reported with Extraction Systems' TMB-RTM real-time, total molecular base monitor. CDs and the appearance of T-topping were measured using Hitachi Model S-8820 and S-9200 CD-SEMs; cross sections were measured on a Philips XL-30 SEM.
Resists tested and general coating conditions are described in Table 1. Appropriate antireflective layers were selected for each resist type, an inorganic BARC used with the 193nm resist. Resist parameters were chosen based on IMEC's standard coating recipes.
Figure 1 displays the total molecular base levels during the test period, indicating an average of 6ppb(v) total molecular base in the cleanroom ambient. The significance of using volume (v) is that it effectively removes the weight of the molecule from consideration. At the same time, the exposure tool enclosure maintained average molecular base levels of <1ppb(v). These readings were used as a baseline for the reported sensitivity values.
Data analysis
CD data for dense and isolated lines were collected for each resist and used to create energy vs. linewidth curves from which ideal curves were derived. The curves can be extended beyond the limits of the data to provide values not available from raw CD data. For Resist L dense lines at best focus, the best fit of CD data is shown in Table 2.
Data from fitted value tables were used to generate CD shift values. To compensate for acid diffusion, CD shift values were calculated for each delay time using the following formula:
CD Shifttime = n = [Delay in Cleanroomtime = n - (No Delay + (Delay in Steppertime = n - No Delay))]/Delay Timetime = n
In this calculation, the CD shift at delay time = n is determined by dividing the difference between the CDs measured in test wafers and control wafers by the total delay time.
Discussion
Although the study is limited in terms of the number of resists tested, several observations can be derived from the results. Analogous to parameters such as etch resistance, resist sensitivity to molecular contamination remains a factor that must be considered when selecting a process and setting up procedures for volume manufacturing. For example, within the stepper environment, little change occurred in molecular base levels from no delay to a 45-min delay, while the coated wafer sat in the exposure tool chamber. However, molecular base levels after a 45-min delay in the cleanroom could increase by as much as 55%. These results can be used to determine the appropriate resists at particular design rules.
 Figure 2. The impact of molecular base contamination on CD is particularly visible with the ArF Resist O, which displays a 35% variation at the 20ppb level. |
In an examination of CD shifts, sensitivities of 248nm photoresists were shown to vary by as much as an order of magnitude, from 0.037nm/min/ppb for resist M to 0.368nm/min/ppb for resist L (Fig. 2). The degree of T-topping can also indicate sensitivities to molecular bases. In particular, the T-topping effect of molecular bases on Resist L (KrF high-activation-type resist) was particularly prominent when exposed to the cleanroom ambient for 45 min (Fig. 3), whereas its impact on Resist N (KrF Acetal-type resist) was minimal (Fig. 4). These results led to the selection of Resist N in IMEC's 0.25µm process with no requirement for additional filtration within the exposure tool.
 |
Resist swelling or T-topping occurs in all exposed wafers. Unmanaged, it can significantly affect CD control and lead to wafer reworking or, in the worst case, yield loss. Generally, however, T-topping can be kept to an acceptable level through the use of chemical filters and careful management of timing delays in processing wafers. Once the resist is developed, the risk of T-topping is eliminated. T-topping manifests itself as an increase in linewidth or CD at the top of the imaged feature. Molecular bases such as ammonia and amines in the ambient react with acidic species produced during exposure of the resist, interfering with the chemical reactions that form the desired features. Reactions of ammonia and the photo-generated acids in the resist occur most strongly at the surface of the resist where the concentration is highest and the reaction with nearby resist acid species more likely. The linewidth increase can be minimal, as seen in Fig. 3b, or so severe that the space between adjacent features disappears, as seen in areas of Fig. 5c. If the CD shift is great enough, changes in the resulting linewidth can cause the electrical characteristics and performance of the devices to change, possibly to the point where the device will not perform to specification and its speed or function is diminished.
 |
Resist L is a high-activation resist, allowing for thinner resist thicknesses and imaging 0.18µm features. Though T-topping was visible in Resist L after long exposure in the cleanroom (Fig. 3c), it displayed little T-topping after delays in the exposure tool (Fig. 3b), making it suitable for IMEC's 0.18µm process. The ability of filters and monitoring to measure and control molecular contamination inside the exposure and track systems is well established, and resist sensitivity could be somewhat sacrificed for other, more critical properties such as resolution.
Figure 5 shows the molecular contamination sensitivity of Resist O, an ArF COMA-type resist, for 0.15µm features. This resist exhibited the greatest sensitivity of those tested — extreme T-topping after a 45-min delay in a cleanroom ambient (Fig. 5c). This result is not surprising given the developmental nature of 193nm processes and early resist formulations. However, Resist O is used on IMEC's 193nm exposure system because, as in the cases of Resists L and N, there is negligible T-topping within the exposure tool (Fig. 5b).
Interestingly, Resists M and N had the lowest amount of T-topping, but are older resists chosen a few years ago for the larger linewidths 0.25µm and 0.30µm. They were chosen for their insensitivity when ammonia filters were not well understood and their performance questionable. In contrast, Resists L and O are newer resists for the more aggressive CDs 0.18µm and 0.15µm. They were chosen more for their printing performance and to a much lesser extent for their ammonia sensitivity, after tool filters were able to keep ammonia levels low and TMB available to monitor problems.
Table 3 presents the data for Resist L dense features. The maximum CD shift value at best exposure dose is shown in blue. This process was repeated for each resist and the CD shift values at each delay time are summarized in Fig. 6.
Figure 2 shows the relative maximum sensitivity of all four resists and the impact of molecular base contamination on CDs. Data were based on the "worst case" CD shift for each resist for delay times up to one hour. Plots were produced using the maximum CD shift values for each resist and assume a linear dependency to concentrations of molecular contamination. The data highlight the potential CD change for even the least sensitive resists at elevated concentrations, useful information during the process development. For example, if Resist M is exposed to a concentration of 20ppb, a level that can be seen occasionally in even the most controlled environment, a CD change of 1nm/min is possible. A 10-min delay can mean a 10nm total CD shift. This may not be a problem if the wafer is in a well-filtered photo-cluster environment, and may be even less of a problem for less critical processes at 0.25µm.
Even within the tool, levels as high as 20ppb could occur if procedures are not followed closely, since unfiltered, ambient air can enter during routine maintenance. Insufficient purge time before a process is restarted can also leave elevated residual levels, particularly for less-volatile molecular bases that can adhere to surfaces. Further, imbalance in air flow between photo-cluster components can lead to ambient air infiltration in some cases or to flow from areas of high potential contamination (e.g., hexamethyldisilazane [HMDS] processes) to areas where sensitive wafers sit for extended time. Mapping the molecular contamination levels both inside and outside the photocell can help process engineers visualize potential problem areas and develop procedures that take appropriate precautions to avoid long delay times. Resist data in use at their sites can be used to characterize both the resist sensitivity and the ambient, track, and exposure tool molecular base background levels to assist in making judgments about acceptable delay times for their particular processes.
The difference in sensitivity between dense and isolated lines is shown in Fig. 2, and is particularly visible with the ArF Resist O, which displays a 35% variation at the 20ppb level. Somewhat less visible but still significant is the 25% variation at the 20ppb level in Resist N. It is possible that the 35% difference is due to less mature ArF resists, but this does not sufficiently explain variation in Resist N. More study would be needed to determine the role of resist maturity, the differences between KrF and ArF processes, and the role of feature-specific resist formulations to completely understand this occurrence.
Sensitivity of some resists (Fig. 6) showed nonlinearity during the molecular base exposure time of 0-45 min. Resists N and O exhibited an increase in the rate of CD shift between 15- and 30-min delay times, but from 30-45 min the rate of change was reduced. A more detailed investigation on the role of the bulk resist layer in the production and diffusion of photo-acid species would shed light on the significance of resist thickness and feature size, and whether or not the nonlinearity was repeatable.
Monitoring both the cleanroom ambient and the equipment enclosures for molecular bases can provide useful data regarding the effect of resist contamination on CD shifts. The knowledge of this parameter alone can help process engineers set limits for molecular base exposure, help them define the need to inspect or rework wafers, and alert them to potential catastrophic product loss.
Filtration systems
Depending on the design rules, chemical filtration and monitoring is required for some, though not all, DUV resists. Other solutions to minimizing or eliminating the effects of molecular base contamination include resist improvements that decrease sensitivity to molecular contamination. But the most straightforward approach is controlling airborne contamination with robust filters in the exposure tool. Chemical filters, in some cases, have moved from traditional, treated carbon-based media to new formulations, including highly porous polymeric materials that can provide improved removal efficiency and capacity. In the latest ArF exposure tools, filters have the ability to remove molecular bases and protect resists with formulations that remove condensable materials like large organic molecules. These condensable species are of particular concern because they can form films on optical surfaces, leading to lens performance degradation. Similarly, condensable materials can degrade wafer surface quality and electrical performance of the devices. Work continues to combine materials and produce filters to address multiple concerns.
Additionally, device makers are adopting chemical contamination monitoring schemes for several purposes. Traditional chemical sample capture and analysis using impinger (chemical trap or bubbler) methods and gas chromatography/mass spectrometry (GC/MS) analysis in a laboratory can allow process engineers to look at average concentrations on a given day. Regular sampling, monthly or quarterly, for example, can provide trend data. In either case, sampling is a "snapshot" of the chemical environment at a particular point in time and does not provide discrete event analysis. In situ NH3 sensors with low-ppb-level sensitivities have been more recently adapted to monitor filter performance and ambient concentrations. These methods provide a look at short-term events, particularly for catastrophic accidents such as releases of HMDS or intrusion of NH3 from adjacent work areas.
 Figure 6. When exposed to delays in the cleanroom, CDs can vary widely between isolated and dense lines. |
The usefulness of these devices is limited by the need for regular calibration and the lack of discrimination among molecular bases, particularly those that occur only occasionally and without warning. Triethylamine and related by-products of tetramethyl ammonium hydroxide (TMAH), NMP, and non-process-related bases such as corrosion inhibitors used in recirculating coolant systems are not generally detected by such devices. Accidental releases of these compounds can occupy hours if not days of downtime before detection and correction. In contrast, total molecular base monitors provide a complete picture of the molecular contamination environment. For this reason, they are not only being adopted to check filter performance, but are also being implemented in process control decisions, particularly at the <0.18µm node.
Conclusion
Four commercially available, commonly used resists were tested for sensitivity to molecular base degradation and shown to vary in sensitivity by an order of magnitude. The potential effects of molecular base contamination can, however, be controlled or eliminated if resists with reduced sensitivity are used with robust chemical filters designed to maintain particle counts in process equipment enclosures below 1ppb. Work has just begun to characterize the costs of resist degradation due to airborne molecular base exposure [5]. The findings clarify the need for chemical filtration and monitoring in particular 248nm and 193nm photolithography processes.
Acknowledgments
The authors would like to acknowledge Myriam Moelants and Nadia Vandenbroeck for their invaluable contribution in performing the experiments, gathering data, preparing wafers, and metrology.
References
- S.A. MacDonald et al., "Airborne Chemical Contamination of a Chemically Amplified Resist," Advances in Resist Technology and Processing VIII, SPIE, Vol. 1446, 2-12, 1991.
- O. Kishkovich et al., "Sampling Cleanroom Air for Organic Amines," presentation at Sematech DUV Workshop, Austin, Texas, October, 1997.
- J.C. Vigil et al., "Contamination Control for Processing Chemically Amplified Resists," SPIE, Vol. 2438, 630-632, 6/1995.
- H. Ito, "Chemically Amplified Resists: Past, Present, and Future," Advances in Resist Technology and Processing XVI, SPIE, Vol. 3678, 2-12, 1999.
- D. Kinkead, W. Goodwin, K. Turnquist, "Modeling and Controlling the Effects of Base Contamination in DUV Lithography Resists," MICRO, 18(9), pp. 71-84, 2000.
David Ruede earned his BS and MS in chemistry from Central Connecticut State University, New Britain, CT. An industry veteran with more than 16 years of experience, he joined Extraction Systems in 2000. Extraction Systems Inc., 10 Forge Park, Franklin, MA 02038; ph 508/553-3900, fax 508/553-3901, e-mail [email protected].
Monique Ercken received her MS and PhD in chemistry (polymer chemistry) from the Universities of Leuven and Diepenbeek, Belgium, respectively. Since joining IMEC, she has been involved in resist-related issues, including contamination, and is currently responsible for 193nm frontend resist screening. IMEC vzw., Kapeldreef 75, B-3001, Leuven, Belgium; ph 32 0 16 281 338, fax 32 0 16 281 214, e-mail [email protected].
Tom Borgers recently joined IMEC as a technical assistant in the photolithography department.