Effects of etch barrier densification on step and flash imprint lithographyS. Johnson, R. Burns, E. K. Kim, M. Dickey, G. Schmid, J. Meiring,S. Burns, and C. G. Willsona�
University of Texas at Austin, Austin, Texas 78712
D. Convey, Y. Wei, P. Fejes, K. Gehoski, D. Mancini, K. Nordquist,W. J. Dauksher, and D. J. Resnickb�
Motorola Labs, Tempe, Arizona 85284
�Received 21 January 2005; accepted 6 September 2005; published 29 November 2005�
Step and flash imprint lithography �SFIL� is an alternativelow cost lithographic technique that has demonstrated sub-40 nm pattern replication. The SFIL process facilitates highresolution pattern transfer with minimal feature placementand overlay errors via the use of a transparent template topattern a low viscosity prepolymer at low pressure and roomtemperature.1 In the SFIL process, a photocurable mixture ofmonomers referred to as the etch barrier is dispensed onto asubstrate. A relief patterned template is then pressed onto theetch barrier allowing the liquid to completely wet the inter-face between the template and substrate. Once the liquid etchbarrier has assumed the topography of the template, it isphotocured via ultraviolet �UV� exposure and the template isremoved. At this point, an inverse replica of the templatepattern has been captured in the cured etch barrier on thesubstrate. Subsequent dry etch steps are then used to transferthe pattern in the polymerized etch barrier to the underlyingsubstrate.
Previous work has characterized the SFIL process in de-tail and investigated aspects such as template generation,2
curing kinetics,3 etch transfer processes,4 and functional de-vice fabrication.4 Further work is focusing on the refinementof process steps, such as UV cure and template separation,that impact critical dimensions. As the etch barrier polymer-izes and interactions between molecules shift from van derWaals interactions to covalent bonds, densification on a mac-roscopic scale has been observed.5 Although previous workhas shown no pattern placement errors due to densification,the impact of polymerization induced shrinkage on the im-printed feature profile has remained an area of interest.Changes in critical dimension are of particular importance.This article presents two models of the polymerization pro-
a�Author to whom correspondence should be addressed; electronic mail:[email protected]
2553 J. Vac. Sci. Technol. B 23„6…, Nov/Dec 2005 0734-211X/2005
cess and documents the impact of polymerization-inducedshrinkage on imprinted feature dimensions and profile.6
II. MESOSCOPIC MODEL
Etch barrier polymerization and densification have beensimulated with a first-principles mesoscopic model, as previ-ously reported.7 Unlike finite element approaches, which usecontinuum approximations, mesoscopic models consider ma-terials on a molecule-by-molecule basis. This approach hasthe advantage of being able to capture the stochastic,molecular-scale effects that can become pronounced at smallsize scales. For example, mesoscopic modeling has beenused to explore the effects of individual acid molecules online edge roughness in chemically amplified photoresists.7
Mesoscopic simulation of polymerization and densifica-tion is conducted by a two-step process, to be describedbriefly. First, etch barrier component molecules including themonomer, crosslinker, and photoinitiator are placed ran-domly on a three-dimensional lattice �Fig. 1�. Polymerizationis performed by allowing the molecules to randomly diffuseand react inside the lattice via a Monte Carlo simulation.Parameters such as component concentrations, exposuredose, and reaction rate constants can be varied to effect po-lymerization results. Second, densification is simulated bycalculating the equilibrium positions of all molecules. Thisstep is performed by assigning intermolecular harmonic�spring� potentials to all molecules in the lattice. An iterativeprocedure is then used to gradually adjust the positions of themolecules until the forces between them reach a minimum.The strengths of the potentials are determined from the bond-ing network generated in the first step; covalent bonds aregiven a much stronger potential than van der Waals bonds.
As previously reported, good agreement has been shownbetween simulated and experimental extents of conversion.7
Results for this model also indicate that densification willcause the feature to shrink largely in the vertical direction.This behavior is thought to be caused by adhesion of the
cured etch barrier and spin-cast transfer layer to the rigid
2554 Johnson et al.: Effects of etch barrier densification on step 2554
substrate. Figure 2 depicts calculated and experimental re-sults for vertical shrinkage as a function of feature aspectratio, reproduced from Burns et al.7 for purposes of compari-son. Agreement with experimental results is quite good. Asshown, increasing feature aspect ratios correspond to a de-creasing percentage vertical contraction as measured at thecenter of the top surface. That is to say, short, wide featureswill exhibit a larger percent of vertical contraction than nar-row, tall features. It is interesting to note that the finite ele-ment model discussed in the following section reproducesthis same trend.
III. FINITE ELEMENT MODEL
A finite element model �FEM� of the etch barrier was alsoused to examine the effects of polymerization inducedshrinkage. This model utilizes continuum mechanics andmeasured bulk material properties to simulate final etch bar-
FIG. 1. SFIL mesoscopic model.
FIG. 2. Percent vertical contraction as a function of aspect ratio.
rier feature profiles after template separation. For this model,the liquid etch barrier is assumed to completely wet the tem-plate. It is also assumed that after UV cure, the polymerizedetch barrier adheres to the template and does not shrink untilthe template is removed. Distributed stress and strain energythroughout the etch barrier account for the lack of shrinkage.Once the template is removed, the etch barrier deforms tominimize strain energy based on initial template geometry,percent shrinkage, and the elastic modulus and Poisson ratioof the cured etch barrier.
A finite element model consisting of 200, 100, 70, 40, and30 nm lines has been constructed. Imprinted features are100 nm in height on top of an 80 nm residual layer. Only theetch barrier material is included in the model; the underlyingrigid substrate is represented by a fixed boundary conditionpreventing movement of the bottom interface of the etchbarrier. For bilayer sample configurations utilizing a spin-cast transfer layer on the wafer prior to imprinting, the un-derlying transfer layer is assumed to have mechanical prop-erties similar to those of the cured etch barrier. This allowsboth polymer films to be modeled as one thicker film. Edgesof the residual layer are also assigned a symmetric boundarycondition to simulate the effect of a residual layer that coversan entire imprinted die. This model uses etch barrier materialproperties of 10% volumetric shrinkage, 100 MPa elasticmodulus, and 0.4 Poisson ratio.
Figure 3 presents FEM model results of imprinted featurenode displacement due to shrinkage. The shape of the etchbarrier indicates initial template geometry, and shading indi-cates displacement due to shrinkage. Figure 3�a� shows thelateral component of feature displacement. At the base offeatures �a point commonly used to determine critical dimen-sion �CD�� show very little displacement or change in dis-placement. However, the tops of features are far enoughaway from the constraining residual layer to exhibit some
FIG. 3. Finite element model of imprinted lines. Node displacement data for200 nm �left� through 30 nm �right� lines are shown: �a� horizontal compo-nent, �b� vertical component, and �c� total displacement magnitude.
motion. Larger features show more lateral contraction at the
2555 Johnson et al.: Effects of etch barrier densification on step 2555
top of lines: the 200 nm line shows 30 nm of lateral contrac-tion, and the 70 nm line shows 10 nm of lateral contraction.These numbers, when taken with any change in featureheight, can also be used to determine sidewall angles.
Figure 3�b� shows the vertical component of feature nodedisplacement. The top of each of the 100 nm tall featuresexhibits a total displacement of 15 nm towards the top of theresidual layer. One can also begin to see the aspect ratiodependence on vertical shrinkage predicted by the mesos-copic model in this data. Note that the larger lines show morecurvature and slightly more displacement than the smallerfeatures.
Figure 3�c� shows the total displacement of features �bothvertical and lateral components summed together�. A com-parison of the total displacement results with the vertical andlateral displacements shows a strong correlation between thetotal and vertical displacements. That is to say, the verticalcomponent accounts for the majority of the displacement offeatures, and most of the polymerization induced shrinkagewill manifest itself as a decrease in feature height.
IV. EXPERIMENT
Experimental work was performed with the goal of gen-erating cross sections of both imprinted samples and the tem-plate used for printing. Given both template and imprintedcross sections, one can then determine the change in etch
FIG. 4. Cross-section SEM image of features printed with template shown inFig. 5.
FIG. 5. Cross-section TEM image of template features: 200, 100, 70, 40, and
30 nm features shown.
JVST B - Microelectronics and Nanometer Structures
barrier geometry due to shrinkage. Three etch barrier formu-lations predicted to exhibit different amounts of bulk volu-metric shrinkage were prepared for imprinting. Monomermolecules with bulky pendant groups are expected to exhibitless shrinkage upon polymerization than those with smallpendant groups.5 Formulations of 4% of the photoinitiatorDarocur 1173 �Ciba�, 30% ethylene glycol diacrylate, and66% of either lauryl acrylate, hexyl acrylate or ethylene gly-col diacrylate were prepared. Based on bulk measurementsof volumetric shrinkage upon polymerization describedpreviously,5 these formulations are expected to vary between10% and 20% densification upon photocure. These formula-
FIG. 6. Tabulated TEM and SEM feature dimension data.
FIG. 7. Linewidths at base: comparison of FEM simulation, template dimen-
sion, and imprinted line dimension.
2556 Johnson et al.: Effects of etch barrier densification on step 2556
tions were imprinted on a Molecular Imprints Imprio-100 atMotorola Labs in Tempe, AZ. Samples were exposed at365 nm with 74.2 mJ/cm2 for 60 s. Figure 4 shows cross-section scanning electron microscope �SEM� images of 200,100, 70, 40, and 30 nm lines.
After imprinting, the template used to image the im-printed samples was prepared for detailed metrology. Thetemplate was coated with a thin film of chromium for imag-ing contrast. It was then coated with a film of oxide to main-tain sample integrity during subsequent processing. A thincross section appropriate for transmission electron micros-copy �TEM� use was prepared by focused ion beam milling.Figure 5 shows cross section TEM images of the templateused to print the features shown in Fig. 4. Widths at the baseand top of features and feature height were recorded formany lines of varying width and pitch. Figure 6 presents asummary of this data.
V. RESULTS AND CONCLUSIONS
Figures 7–9 summarize dimensions of etch barrier, tem-plate, and finite element model features. The horizontal axisdenotes nominal feature dimension and the vertical axis de-notes actual dimensions. Figure 7 compares linewidths at thebase of the features. Template, imprinted, and FEM featuresall show the same dimensions. That is to say, model andexperimental data show no change in linewidth at the base offeatures. Figure 8 compares linewidths at the top of features;note that this data can also be interpreted as an effectivemeasure of sidewall angle when combined with featureheight information. Template, imprint, and model dimen-sions again agree closely. Two hundred nm features begin toshow some shrinkage as indicated by the finite elementanalysis �FEA� model, but these changes in dimension areapproaching the limit of SEM resolution. Note that the di-mensional variations of the 40 and 30 nm lines are a metrol-ogy artifact related to the shape of these small features in thetemplate. Figure 9 presents feature height, defined as the dis-
FIG. 8. Linewidths at top: comparison of FEM simulation, template dimen-sion, and imprinted line dimension.
tance from the topmost point of a feature to the top of theresidual layer. Measured template dimensions are roughly15 nm larger than their corresponding imprinted and FEMcounterparts. These results indicate that polymerization in-duced shrinkage does indeed manifest itself largely as achange in the height of imprinted features with a slightchange of sidewall angle. Adhesion to the rigid substrateprevents lateral motion during shrinkage. Thus, photo-loymerization induced shrinkage manifests itself primarily asa change in feature height.
In summary, model and experimental studies of photopo-lymerization induced bulk shrinkage have shown minimalimpact on the CD and line profile of imprinted features. Bothmodel and empirical data show no change in CD at the baseof features. One hundred nm tall features on an 80 nm re-sidual layer exhibited a total change of 15 nm in height, andlarger features show small decreases in sidewall angle. Nodedisplacement data from the finite element model is able topredict the final profile of the imprinted etch barrier.
ACKNOWLEDGMENTS
The authors gratefully acknowledge the support ofDARPA �Grant No. N66001-01-1-8964�. The authors wouldalso like to thank Adolfo Rios, Eric Ainley, Jennifer Clift,and Kathy Palmer at Motorola Labs for help with templategeneration and metrology.
1B. J. Choi, M. Meissl, M. Colburn, T. Bailey, P. Ruchhoeft, S. V.Sreenivasan, F. Prins, S. Banerjee, J. G. Ekerdt, and C. G. Willson, Proc.SPIE 4343, 436 �2001�.
2T. C. Bailey, S. C. Johnson, S. V. Sreenivasan, J. G. Ekerdt, C. G. Will-son, and D. J. Resnick, J. Photopolym. Sci. Technol. 15, 481 �2002�.
3E. K. Kim, N. A. Stacey, B. J. Smith, M. D. Dickey, S. C. Johnson, B. C.Trinque, and C. G. Willson, J. Vac. Sci. Technol. B 22, 131 �2004�.
4S. C. Johnson et al., Proc. SPIE 5037, 197 �2003�.5M. Colburn, I. Suez, B. J. Choi, M. Meissl, T. Bailey, S. V. Sreenivasan,J. Ekerdt, and C. G. Willson, J. Vac. Sci. Technol. B 19, 2685 �2001�.
6G. M. Schmid, M. D. Stewart, V. K. Singh, and C. G. Willson, J. Vac. Sci.Technol. B 20, 185 �2002�.
7R. L. Burns et al., Proc. SPIE 5374, 348 �2004�.
FIG. 9. Feature height: comparison of FEM simulation, template dimension,and imprinted line dimension.
