Chemistry and physics of the PEB process in a CA resist
08/01/2000
William Hinsberg, John Hoffnagle, Frances Houle,
IBM Almaden Research Center, San Jose, California
Additional authors are listed in the Acknowledgments.
This article first appeared in the Spring 00 edition of Microlithography World.
overview
Measurements and modeling of the post-exposure bake process of a chemically amplified resist reveal that diffusion plays little role and that the observed spreading of exposed regions results primarily from chemical kinetics in regions where a photo-acid gradient exists. Analysis of volatile products during PEB reveals new side-reaction pathways.
The chemically amplified (CA) photoresist materials developed for DUV (248nm) lithography are in many ways superior to alternate materials [1], displaying, for example, very high contrast and radiation sensitivity. It is anticipated that CA resists will be used well into the future, potentially down to 35nm ground rules. The current understanding of their imaging process, however, is not sufficient to inspire confidence at such demanding dimensions.
 Figure 1. Comparison of the chemistry of DNQ-novolac (i-line) and chemically amplified (CA) resists. |
CA resists are based on a radically different principle than the i-line resists they superceded [2]. Figure 1 compares the imaging chemistries of the two. While the i-line resist is rendered soluble in aqueous base by the photochemistry, a CA resist requires a post-exposure heating step (typically 90-140°C for 1-2 min) to activate the deprotection reaction that modifies the solubility of the film. This is because the chemical species formed by exposure of the film is an acid catalyst, a substance that induces the reaction of a large number of functional groups within the resist during post-exposure bake (PEB) without being consumed itself.
The chemistry during the PEB step plays a critical role in determining the overall imaging properties of a CA resist system. While simple in concept, the PEB step is complex in its details. It encompasses a multistep, catalyzed, chemical reaction in a solid film (often accompanied by a parallel uncatalyzed reaction that produces similar changes in polymer structure); volatilization of some of the reaction products (accompanied by film shrinkage); and transport of species within the film.
PEB and image spreading
It has been recognized since the earliest stages of CA resist R&D at IBM that movement or transport of the acidic catalyst within the film could limit the ultimate achievable resolution. Thermodynamics requires the acid to move from regions of high concentration (the exposed pattern) to regions of low concentration (the surrounding unexposed area), but other factors determine the rate. Since the acid catalyst is not consumed, and if the acid were sufficiently mobile, then in principle, a single, small, exposed region in a film could cause the entire resist film to undergo the catalyzed reaction and develop. Clearly this does not happen with today's process conditions and length scales, but image spreading might yet limit the resolution of CA resists.
Detecting acid diffusion by analyzing how an exposed resist feature changes in size (Fig. 2a) has proved difficult. Diffusion rates are related to the concentration gradient of the diffusant, and the rate of the deprotection reaction is related to the acid concentration [3], so chemical kinetics and acid diffusion are tightly intertwined and will vary in a complex way as a function of temperature, time, and location in the film.
We have developed a two-step experimental approach to characterize image spreading during PEB using infrared spectroscopy that allows these two effects to be analyzed separately [4]. First, the detailed deprotection chemistry is analyzed under conditions where there is no gradient of acid in the polymer film, leading to a quantitative description of the chemical kinetics of the system that is accurate over a wide range of temperatures, acid concentrations, and polymer compositions [3]. In a second analysis, the experiment is repeated using 193nm light to make a 200nm-thick layer of exposed resist containing acid at the surface of a much thicker (~1000nm) unexposed layer of strongly absorbing photoresist.
A schematic of the experiment is shown in Fig. 2b. In the limit where the photogenerated acid does not diffuse to any significant degree, then only the initial acidified surface layer will undergo the acid-catalyzed chemistry. In the case where the acid does diffuse into the initially acid-free underlayer, then both the surface layer and some portion of underlayer will react. The overall extent of deprotection in the film is readily quantified by monitoring the intensity of a characteristic absorption in the infrared spectral region, and the chemical deprotection kinetics are precisely and quantitatively known. Thus, the details of acid transport can be discerned by comparing the experimental behavior of the system to that predicted by reaction-diffusion modeling.
The CA resist we have focused on is based on the poly((t-butoxycarbonyloxy)styrene) (PTBOCST) polymer, one of the simplest and best-characterized CA resist polymers for DUV. In the work described here, PTBOCST is combined with di-(t-butylphenyl)iodonium perfluorobutanesulfonate (TBI-PFBS), a photo-acid generator used in contemporary CA resist systems, which generates perfluorobutanesulfonic acid (PFBS-A) upon irradiation.
Modeling
To analyze the results, the chemistry and physics of the resist film were modeled using a stack of 120, 10nm-thick resist subfilms, each initialized at time = 0 with the fully protected polymer and an amount of acid catalyst formed by the 193nm exposure. The initial acid distribution accounted for the dose attenuation by the strongly absorbing film.
Within each layer, the full range of deprotection chemistry can take place, and transport of the acid can occur between adjacent layers. The model contains only the minimal elements judged necessary to capture the main chemistry and physics: the chemical deprotection chemistry (known from independent experiments); plus a means for acid transport in the direction of its concentration gradient, using Fick's Law of diffusion to describe the concentration dependence of the rate [5].
Chemistry suggests that the mobility of the acid (an ionic species) will differ between the nonpolar environment of the initial protected polymer, and the polar environment that results from deprotection. The model thus contains two diffusion coefficients to characterize diffusion in each of those two extremes. The model is formulated so that the overall diffusion rate of acid at any location in the film at any instant depends on both its concentration gradient and on the composition of the local polymer environment.
Results
We have found that this simple model can provide an accurate description of our experimental kinetics measurements. In the description that emerges, within the temperature range from 65-105°C, the diffusion of PFBS-A in the initial protected PTBOCST polymer increases from a value of DTBOCST = 0.05 to 0.15nm2/sec. The temperature dependence of DTBOCST can be described by an Arrhenius plot with a 36.5kcal/mole energy of activation. As the polymer undergoes deprotection, the mobility of PFBS-A decreases sharply. In the fully deprotected polymer (largely poly(hydroxystyrene), or PHOST) the diffusion constant is DHOST = 0.001 to 0.05nm2/sec in the same temperature range. The activation energy decreases to 22.1kcal/mole. These diffusion coefficients are much less than might have been expected, and lead to effective acid diffusion lengths on the order of 5nm during typical PEB (100°C for 120 sec).
 |
Figure 3 plots a set of composition profiles for our PEB experiment, derived from the reaction diffusion analysis. The top traces in 3a show how the initial acid distribution in the film changes during a 600-sec PEB at 100°C. The bottom plot in Fig. 3a shows the concurrent evolution of the deprotection reaction, as measured by the amount of the TBOCST protecting groups remaining on the polymer. The acid profile changes only slightly during PEB, while the "front" of deprotection moves substantially with time; in this particular circumstance, the front has moved ~100nm in 2 min. The resist solubility profile follows the deprotection.
 |
Figure 3. Composition profiles of surface-exposed resist films. The top graphs show the time evolution of the acid profiles a) with, and b) without diffusion. The lower plots show the insoluble polymer concentrations, again a) with the experimentally derived acid diffusion coefficients, and b) without diffusion. The dashed lines indicate the half-maximum position of the acid and deprotection after a 2-min PEB (with diffusion).
Figure 3b presents the behavior predicted without diffusion. Even under these limiting conditions, the deprotection front advances with PEB time. In fact, the results are virtually indistinguishable from the case in Fig. 3a for typical PEB times. Therefore, the model indicates that there is considerable spreading of the profile of deprotected polymer with increasing PEB time, far in excess of that expected on the basis of acid diffusion alone (Fig. 3a). Since the deprotected polymer is removed during development, this spreading is manifested as an apparent increase in the widths of bright features with increasing PEB time.
The deprotection spreading during PEB results primarily from the kinetics of the acid-catalyzed deprotection reaction, not diffusion. The rate of deprotection is an increasing function of acid concentration. In the highly exposed region near the top of the film, the acid concentration is maximum and full deprotection occurs quickly. By a depth of about 150nm, the acid concentration has tapered off due to decreasing exposure, as it would at the edge of a bright image feature. Lowered acid concentration dramatically slows the local rate of the deprotection reaction. As the PEB time lengthens, more and more polymer is converted to the deprotected state at greater and greater distances from the maximum acid concentration. The "front" of soluble deprotected polymer advances into the insoluble region, even if the acid concentration profile remains fixed.
The apparent image spreading is primarily a consequence of the fact that the rate of the deprotection reaction varies with acid concentration, which, in turn, reflects the spatial profile of the exposure dose. While Fig. 3 demonstrates that the diffusion of acid can increase the rate with which the spreading occurs, in this instance it constitutes a minor contribution to overall spreading of the deprotected polymer region on the time scale (<2 min) typical of PEB processing.
Implications for patterning
To better understand the practical implications of our observations, we have carried out reaction-diffusion simulations for line-space arrays exposed with sinusoidal aerial images (85nm half-pitch in Fig. 4), using exposure dose and PEB process conditions typical for PTBOCST resist. Figure 4a depicts the evolution of the acid profile, and that of the profile of the deprotected polymer (which comprises the developable image) for the case where the acid diffusion kinetics are those derived from our experimental measurements. The aerial image contrast was 100%.
 |
Figure 4b displays the same profiles for the case without diffusion. The polymer is considered to be soluble in developer when the deprotection reaction has reached at least 80% conversion. Dashed lines indicate this threshold on the graphs. The regions where a curve falls below that line are soluble and will be removed upon development. Thus the sets of curves can be used to describe how linewidth will decrease and the space width increase with increasing PEB time. Comparing Figs. 4a and b shows that acid diffusion has negligible impact for 85nm features and typical resist-processing conditions.
But what happens as we move to smaller linewidths? Figure 5 shows the developed linewidth of a dark feature as a function of nominal image size along with the linewidth shifts due to chemistry alone and due to diffusion alone. The dimensions and shifts are plotted as percentages of the nominal linewidth. The scatter in the data points reflects statistical variations in simulations similar to Fig. 4. Figure 5 shows that this resist system produces lines that are a constant ~93% of the nominal width down to ~60nm. The constant 7% shift is due to the deprotection chemistry dynamics, the range of which scales with the dimensions of the sinusoidal aerial image, even below 30nm. As the linewidth decreases below 70nm, however, the previously negligible effects of acid diffusion become significant and ultimately become dominant, dramatically narrowing the linewidth of the developed feature.
Even though the effective diffusion length of our PAG is only 5nm, our results imply that it can be further reduced by increasing the size of the spectator anion that maintains charge neutrality. Thus, it may be possible to approach the ideal "no-diffusion" condition in real resists. Image spreading will still occur, however, driven by the catalytic nature of the image-recording process, which allows even very small amounts of acid to produce substantial deprotection. There is also the possibility that a trade-off between sensitivity and resolution will be required as we approach the dimensional limits of CA resist, even in the absence of diffusion. The model described here provides a framework to examine that trade-off.
Side-reaction processes
Though acid-catalyzed deprotection chemistry is common to all positive-tone CA resist systems, the reaction has been rather casually studied. The question of how cleanly such deprotection reactions take place in thin solid films has been largely ignored, even when NMR studies of model systems in solution have shown other reaction pathways leading to products that could impact lithographic properties. Detailed spectrometric studies using inverse-gated proton-decoupled 13C NMR [6, 7] and atmospheric pressure mass spectrometry (MS) [8] show that the compositions and chemistry of the processed PTBOCST resist system is far more intricate than first thought. Figure 6 shows the desired and expected reaction in a), leading to PHOST and volatile products.
 Figure 5. The developed linewidth as a percentage of nominal bark feature width and the contributions to narrowing from diffusion and chemical dynamics. |
NMR analysis of normally processed resist reveals that a significant fraction of the polymer has undergone reactions that yield modified PHOST. Two distinct structures are formed when the scission occurs at the expected bond (A in Fig. 6b). In the major side product, the t-butyl cation has become bound directly to the six-carbon ring, but there is a minor product where it attaches to the pendant oxygen atom. These side reactions will modify the dissolution properties of the resist film, possibly lowering contrast and sensitivity. After a 110° PEB, ~8% of the polymer repeat units have these modified structures. If the PEB temperature is increased (130°C for 60 sec), the amount of repeat units undergoing side reaction doubles, but only the ring-substituted side product appears in the film. This illustrates that even small variations in process conditions can induce significant changes in the composition of a processed resist film.
 Figure 6. a) Deprotection chemistry, and b) side reactions during PEB in PTBOCST resist. |
Mass spectral analysis of the volatile materials released during PEB using a trace atmospheric gas analyzer (TAGA) reveals a second set of products that are not derived from reactions of t-butyl cation [8], but rather from a t-butoxy radical. This unstable intermediate must be produced by scission at B in Fig. 6b, and then undergo further reaction to yield stable products. The acetone and t-butyl alcohol detected by the TAGA are characteristic products of t-butoxy radical reactions. The importance of this newly identified reaction path depends on the process conditions; for example, an increase in exposure dose (which will increase the concentration of acid available during PEB) causes an increase in the ratio of products resulting from this new pathway relative to the dominant reaction. This reinforces the notion that seemingly minor changes in process variables (e.g., exposure dose) can have unexpected results.
Summary
PTBOCST resist contains the key functionality common to all positive-tone CA resists in the simplest possible chemical composition, making it an ideal system for fundamental study. The initial work summarized here has already yielded new insights into the inner workings of the PEB process, and the connection between resist chemistry and imaging. The presence of previously unrecognized side products and reaction pathways adds complexity to the tasks of resist design and processing, and their impact on resist functional properties must now be assessed.
We are now applying the techniques we have developed to resist systems of greater complexity, with the goal of systematically characterizing the influence of structural and compositional changes in the resist on the chemistry and physics of the PEB process.
Anowledgments
Additional authors of this article include Hiroshi Ito, Martha Sanchez, Mark Sherwood, and Greg Wallraff. We thank Michael Morrison, Dean Pearson, and Michelle Poliskie (Grinnell College) for their technical assistance. This work has been supported in part by National Science Foundation Grant DMR-9400254 (CPIMA).
References
- H. Ito, IBM J. Res. Devel., 41, 69, 1997.
- R. Dammel, Diazonaphthoquinone-based Resists, Bellingham, WA, SPIE Optical Engineering Press, 1993.
- G. Wallraff et al., SPIE Advances in Resist Technology and Processing XVI, 3678, 138, 1999.
- G. Wallraff, J. Hutchinson, et al., Vac. Sci. Tech. (B), 12, 3857, 1994.
- The chemical kinetics model, using stochastic kinetics simulations methods, is described in detail in reference 4. The application of stochastic methods in simulation of transport kinetics is described in F. Houle, W. Hinsberg, Applied Physics A, 66, 143, 1998. The combination of these two for the present study is detailed in a manuscript submitted to J. Vac. Sci. Tech. B.
- H. Ito, M. Sherwood, SPIE Advances in Resist Technology and Processing XVI, 3678, 104, 1999.
- H. Ito, M. Sherwood, J. Photopolym. Sci. Tech., 12, 625, 1999.
- The MS ionization source is described in B. Prime, B. Shushan, Anal. Chem., 61, 1195, 1989. Full details of the instrumentation and experimental method are contained in F. Houle, G. Poliskie, W. Hinsberg, D. Pearson, M. Sanchez, SPIE Advances in Resist Technology and Processing XVI, 3999, in press, 2000.
For more information, contact William Hinsberg at IBM Almaden Research Center, 650 Harry Rd., San Jose, CA 95120; ph 408/927-1642, fax 408/927-3310.