Analyzing and characterizing 193nm resist shrinkage
05/01/2001
COVER FEATURE
Bo Su, Guy Eytan, Applied Materials Inc., Santa Clara, California
Munirathna Padmanaban, Andrew Romano, Clariant Corp., Somerville, New Jersey
overview
In addition to stability and collapse issues facing 193nm resists, there is concern regarding a decrease in linewidth when exposed to an electron beam during critical dimension measurements. This problem can be observed using any scanning electron microscope. Such an interaction between the measurement system and sample materials poses a great challenge in process development for 193nm lithography.
Critical issues, such as stability and collapse encountered during 193nm-resist develop processing, may delay the deployment of 193nm (ArF) lithography, even though 193nm scanners are commercially available. There is also the issue of 193nm-resist line shrinkage when exposed to an electron beam (e-beam) during critical dimension (CD) measurements [1-3]. When an ArF resist line is subject to changing CDs because of shrinkage, repeated CD measurements using a CD-SEM - commonly done during process-development tests - are highly undesirable.
Resist shrinkage when exposed to UV light has been observed even in 248nm resists in its early development phase [4]. Modified resist formulations helped to overcome these problems. Whole wafer e-beam flooding has also been investigated to improve 248nm-resist stability [5]. This work showed that both physical and optical properties of the studied resists were altered through certain degrees of e-beam exposure. Subsequent, continuous improvements of 248nm resist formulations, however, proved that such a stabilization technique was unnecessary in IC production.
Since 193nm-resist technology is still in its development phase, there is so far no clear advantage to one resist formulation over others. Thus, we chose to investigate three different 193nm-resist formulations from Clariant (simply labeled Nos. 1, 2, and 3). Using Applied Materials' VeraSEM 3D system, we observed vertical and lateral 193nm-resist shrinkage under e-beam exposure. Overall, we observed different CD changing behaviors for lines and spaces. For example, repeated CD measurements on a space magnify the CD changing effect because there is three to five times more resist volume exposed to the e-beam than that with a line. Hence, the influence of other competing effects (i.e., the "background noise") from line edge roughness, carbonization, etc. are reduced.
We did find that by measuring a space or an edge width at a tilted view, which is part of the capability of the VeraSEM 3D system, the severity of e-beam-induced resist shrinkage of different resist types can be compared directly with a high level of confidence. This system can electronically tilt the e-beam, enabling exposure of only one sidewall of a resist feature.
Resist tests
Clariant resist No. 1 is an acrylate based on development work at Fujitsu, and No. 2 is based on work at Hyundai. Resist No. 3 is a new Clariant formulation. We coated bare silicon wafers with a bottom anti-reflection coating (BARC) and a ~450nm-thick layer of resist. Exposure was done with a stepper (0.6 NA, 0.7s) at the specified exposure dose for each resist.
We observed that lines with resist No. 1 started to collapse at 140nm at small (280nm) pitch, but we did not see any resist collapse with resists Nos. 2 and 3, even down to 120nm. Resists Nos. 2 and 3 did, however, have a slower photospeed (i.e., they required 33% more exposure dose than No. 1). We also noted that resist No. 1 exhibited a better profile at 150nm than Nos. 2 and 3.
Both 193nm resist lines and spaces were measured repeatedly using 500 eV e-beam landing energy and 10 pA beam current, unless otherwise specified. Multiple parameters, for example the edge width [6] (i.e., the single secondary electron peak width), as well as the top and bottom readings of lines/spaces, can be collected using a unique measurement algorithm. VeraSEM 3D is capable of collecting up to four parameters during a single measurement without adding extra time.
Characterizing resist shrinkage
We can make direct observations of resist shrinkage (Fig. 1) when it occurs after severe e-beam exposure (i.e., continuous e-beam scanning for a couple of minutes resulting in a dose of ~0.1 Coulomb/cm2).
In addition to the lateral shrinkage detailed in Fig. 1, we also observed vertical shrinkage. That is, the tilted beam (i.e., only the resist top and one sidewall exposed to the e-beam) caused resist lines to twist in the direction of the exposed sidewall. The effect was more pronounced with resist No. 3. Obviously, this happened because one sidewall shrinks and the other does not.
Analyzing repeated CD-SEM measurements, we observed that linewidths shrink quickly for the first five measurements and then level off and stay more or less the same after 10 measurements (Fig. 2). Repeated measurements of a space width, however, show a quick initial change followed by a near linear widening thereafter (see Fig. 2). Here, our reasoning is that when measuring a space, five times more resist is exposed to the e-beam compared to lines (for the measurement window width and line and space sizes used in our study). Thus, it takes much more e-beam exposure to saturate resist shrinkage.
 |
We also observed that the top of an isolated line shrinks faster that the bottom (Fig. 3). It is reasonable to expect such behavior, since the top portion of a line or space receives more e-beam exposure than the bottom due to limited e-beam penetration depth compared to resist thickness (i.e., a 500eV e-beam penetration depth is <100nm for photoresist). Repeated SEM measurements of a space showed similar behavior - the space top opens faster than the bottom. Our data here revealed a good linear fit with resist No. 1 top CD opening at 0.34nm/measurement and the bottom CD reading 0.26nm/measurement. We found that all three resists were subject to different top and bottom resist shrinkage rates, but the rate was significantly faster with resist No. 3 (i.e., 0.64nm/measurement vs. 0.34nm and 0.29nm).
Since bottom space width trending is subject to the sidewall slope angle, direct comparison of bottom space trending is not as meaningful if sidewall angles are unknown. Thus, to take sidewall angle influence into account, we examined space edge widths that can be easily measured using a tilted view. We saw that the edge width of a space increases as resist shrinks and that the increase is more or less linear with increasing numbers of measurements.
Edge widths of a space at tilted views are influenced by two opposite trends: Resist height shrinkage tends to decrease edge width, while faster top to bottom shrinkage tends to increase edge width. For a small tilted beam angle (~5° in this study), resist height shrinkage is independent of beam tilt angle in the first order since there is no difference in e-beam penetration depths compared to top down SEM analysis. However, shrinkage rates of top and bottom space widths depend strongly on space sidewall angles; when using a tilted view, a sloped sidewall slows down any trends for edge widths to increase.
Starting with an edge width of ~54nm for resist No. 1 and using 450nm resist height and linear slope approximation, we estimated the sidewall angle at ~88°. Similarly, we estimated space sidewall angles for resists Nos. 2 and 3 as 84° and 86°. Comparing edge width trending slope, it was clear that resist No. 3 shrank much faster than the other two resist types (i.e., 0.32 nm/measurement vs. 0.12nm and 0.09nm). These findings were consistent with top CD-SEM measurements of the spaces. So, since edge width measurements at a tilted view take into account vertical as well as lateral shrinkage, it can be used as a good indicator for resist shrinkage rate.
The edge width of a line using tilted view CD-SEM is much more complicated than for a space. One must also consider the twisting that occurs and how it causes the e-beam to "see" less edge width as measurements progress. Thus, the trending behavior of an edge width of a line at a tilted view is not as meaningful if line twist effect under e-beam is unknown.
 Figure 2. Linewidth slimming and space widening due to resist shrinkage under e-beam exposure of resist No. 1. |
We also observed that nested lines exhibit a different CD decreasing trend as a resist shrinks under an e-beam (Fig. 4), as compared to an isolated line. Nested lines show slower rate of CD change as CD-SEM measurements progress, especially at the beginning of the measurements. While the data in Fig. 4 were taken from resist No. 2, they are representative of all three resists that we tested. We believe that the reason for the better performance with nested lines is the different sidewall angles in nested and isolated lines. In general, for given lithography conditions, isolated lines have sloped sidewalls compared to those of nested lines, and sloped sidewalls increase e-beam exposed resist volume.
Resist shrinkage model
Based on our work, we have constructed a simple model that describes the shrinkage of 193nm resists when irradiated by an e-beam. While we do not fully understand chemical reactions that cause the resist shrinkage, we have decided to attribute it to polymer cross-linking for the sake of discussion.
Our reasoning is that at low e-beam dose, only a small percentage of available polymers undergo cross-linking. As the e-beam dose increases with repeated SEM measurements, more polymers undergo cross-linking until all available polymers have done so.
 Figure 3. Normalized top and bottom CD readings of an isolated No. 1 resist line. |
Further, shrinkage begins at the depth of e-beam penetration. Two mechanisms keep shrinkage progressing. First, as repeated measurements continue, the increased e-beam dose induces more polymer cross-links within the e-beam penetration area until all available polymers are depleted. Second, as the resist shrinks, the e-beam reaches a little deeper into the resist, inducing more resist shrinkage until the cross polymerized denser resist prevents further penetration (i.e., an e-beam has less penetration into denser material). Shrinkage eventually slows then stops at some amount of e-beam exposure.
Our model, then, for resist volume lost due to e-beam-induced resist shrinkage (ΔV) is:
ΔV = a x Ve(d, D, Θ)
where a is a resist-dependent number and Ve is the resist volume exposed to the e-beam. Ve depends on electron penetration depth (d), the electron exposure dose (D), and the angle (q) that the e-beam strikes a sidewall. A tilted beam increases q, thus increasing the shrinking effect through increased sidewall penetration depth.
 Figure 4. Line shrinkage comparison between isolated (space-to-line ratio >>5:1), semi-nested (3:1), and nested (1:1) lines for resist No. 2. |
Lower beam energy does lessen resist linewidth shrinkage (or space width widening) during CD-SEM measurements; lower beam energy means less penetration depth, hence less e-beam exposed resist volume (Ve). For example, the data in Fig. 5 show that 20 consecutive CD-SEM measurements induce ~5% linewidth reduction at 0.3 keV, while the reduction increased to ~11% at 1keV.
 Figure 5. The normalized linewidth slimming, for resist No. 1, as a function of e-beam exposure at different e-beam landing energies. |
Our model only considers primary electrons, not secondary and back-scattered electrons. Other research has clearly shown, however, that substrate materials (i.e., organic and inorganic BARCs) affect 193nm resist shrinkage behavior under e-beam [2]. Such observation suggests that secondary electrons and back-scattering electrons generated from the substrates also contributed to bottom resist CD changes during CD-SEM measurements.
Based on our work and model, we have derived the following recommendations to reduce e-beam-induced resist shrinkage for a given 193nm resist:
- Using lower energy electrons reduces the interaction volume by reducing e-beam penetration depth.
- Using lower beam currents and less exposure time reduces e-beam dose.
- Using off-site automatic SEM focusing (i.e., focusing on a nearby feature) avoids charging and lessens e-beam dose at a measurement site.
- Creating near-vertical resist-line profiles lessens resist shrinkage.
In the long term, improved 193nm resist formulations will also help minimize e-beam induced shrinkage and may prove to be more cost-effective.
Conclusion
We have observed vertical and lateral 193nm resist shrinkage under e-beam exposure using sidewall CD-SEM imaging technology. This technique has helped us to characterize different CD changing behaviors for lines, spaces and edge widths. Repeated CD measurements on a space enhance the e-beam-induced resist shrinkage effect because more resist volume is exposed to the e-beam during CD-SEM measurements. Hence, the resist shrinkage rates of different resists can be directly compared by measuring a space top width. The edge widths of spaces seen with tilted SEM views indicate both the severity of resist shrinkage of different resist types and sidewall slope influence.
We have also identified critical parameters affecting resist shrinkage in CD-SEM measurements. Experimental data show that ArF resist shrinkage is dependent on resist formulation materials, electron beam energy and penetration, electron dose, and feature sidewall angle. With the three resists used in this work, clearly there was no winner when we compared photo speed, collapse resistance, and shrinkage under e-beam (see table). Resist manufacturers have to consider one more property in resist formulation improvements.
References
- M. Neisser, et al., "Mechanism studies of SEM measurement effects on 193nm photoresists and the development of improved linewidth measurement methods," Proceedings of Interface 2000, pp. 43-52.
- L. Pain, et al., "Study of 193nm resist behavior under SEM inspection: How to reduce linewidth shrinkage effect," Proc. of Interface 2000, pp. 233-248.
- B. Su, A. Romana, "Study on 193nm photoresist shrinkage after electron beam exposure," Proceedings of Interface 2000, pp. 249-264.
- N.R. Bantu, et al., "Low shrinkage DUV resist," Proc. of Interface 1998, p. 229.
- M.F. Ross, et al., "Characterization of electron beam stabilization of deep-UV resist," Proceedings of Interface 1997, p. 119.
- B. Su, et al., "Sidewall angle measurements using CD-SEM," Proc. of Advanced Semiconductor Manufacturing Conference (ASMC), pp., 259-261, 1998.
Bo Su is a senior member of the technical staff in the global product management group of CD metrology in the process diagnostics and control group (PDC) of Applied Materials Inc., 3050 Bowers Ave., Santa Clara, CA 95054; ph 408/727-5555, fax 408/748-9943, [email protected].
Guy Eytan is an application development engineer in the PDC group at Applied Materials.
Munirathna Padmanaban is group leader in the development of 193nm resist materials at Clariant Corp., Somerville, NJ.
Andrew Romano is global product manager for advanced products at Clariant Corp.