Development and advantages of step-and-flash lithography

07/01/2001

M. Colburn, T. Bailey, B.J. Choi, J.G. Ekerdt, S.V. Sreenivasan, C.G. Willson, Texas Materials Institute, University of Texas, Austin, Texas

overview
A yet unheralded alternative for future lithography, step-and-flash imprint lithography, appears to be an inexpensive method for pattern generation capable of sub-100nm resolution on silicon wafers. Researchers at the University of Texas have been analyzing various process options on a tool in place there and have shown that this process has several advantages over comparable compression imprinting techniques for applications that require precision layer-to-layer alignment error measurement. It is capable of high-resolution patterning at room temperature with 2-3 pounds applied pressure. The process uses chemicals that are commercially available for sub-100nm patterns. Development work has been so promising that commercialization is under way.

Does optical lithography have a limit? A combination of improvements in optics, further reduction in wavelength, and introduction of more complex masks and processes will surely enable printing features <100nm. Unfortunately, the cost of optical exposure tools is increasing exponentially [1]. In addition, despite the advantages associated with each next generation lithography (NGL) technique, each of these is also expensive.

Looking for yet another alternative, we have been studying imprint lithography (IL) as an inexpensive method for pattern generation capable at sub-100nm resolution on silicon wafers. IL has several important advantages over conventional optical lithography and NGLs. The parameters in the classic photolithography resolution formula (R = kλ/NA) are not relevant to IL because the technology does not use reduction lenses. The resolution of imprint techniques in the sub-100nm regime is well documented [2, 3] and appears to be limited only by the resolution of structures that can be generated in a template or mold.

Click here to enlarge image

Figure 1. SSIL begins with a) spinning an organic thermoset onto a substrate, followed by a spun-on silylated thermoplastic coating. A topography patterned template is b) brought into contact with the top layer at >40psi and >T_g silylated thermoplastic, resulting in its displacement. The template is c) removed and d) the imprinted pattern is transferred to the underlying layer by a short halogen reactive ion etch (RIE) followed by oxygen RIE.

Imprint templates are typically fabricated using electron beam writers that provide high resolution, but lack the throughput required for mass production. IL thus takes advantage of resolution offered by e-beam technology without compromising throughput.

Imprint lithography
There are many IL techniques, all variations on a common theme. The basic premise is that a template or mold with a prefabricated topography is pressed into a displaceable material. That material takes on the shape of the master pattern defined in the template, and through some curing process, the shaped material is hardened into a solid. The process is by nature a contact patterning process that transfers patterns without scaling. So there are common challenges to all of these imprint techniques, the foremost being the dependence of this technology on 1x imprint master resolution, and the potential for defect production and propagation.

Researchers systematically studied imprint lithography techniques in the 1990s [2-7]. This research is divided into two camps, one preferring imprinting into a thermoplastic or thermoset polymer, the other into an ultraviolet (UV) light-curable material. Several researchers [2, 6-8] have followed the same basic technology: A polymer heated above its glass transition temperature (Tg) is imprinted with a mold. The system is cooled below the Tg of the polymer while the mold is in contact, thus curing the shape of the imprint. This process has demonstrated remarkable resolution with features as small as 10nm [2].

Early in our research program, we investigated the prospect of imprinting a silicon thermoplastic at elevated temperatures and pressures [3]. Our goal was to generate a bilayer structure analogous to that produced by bilayer or trilayer lithographic processes [11]. We dubbed the process "step and squish imprint lithography" (SSIL, Fig. 1). Results from this compression molding study [3] illustrated serious problems. Imprinting with varying pattern density results in incomplete displacement of the thermoplastic (Fig. 2) even at elevated temperature and high pressure for

long periods. Partial pattern transfer, failure to displace material completely, release difficulties, and harsh process conditions seemed to limit the potential of this approach.

Indeed, other researchers have reported these problems in compression molding of PMMA derivatives [7]. More important, we decided that the use of high temperatures and high pressures would severely limit our ability to achieve the layer-to-layer alignment required for microelectronic device fabrication.

Figure 2. With SSIL, a) isolated protruding features and b) periodic patterns are replicated relatively easily, but c) isolated recessed patterns are difficult to imprint successfully.Click here to enlarge image

The second route to imprint lithography relies on curing a low-viscosity, photosensitive material with UV light. This method has been used in the production of optical disks [12]. Philips Research has demonstrated a photopolymer process of this sort that produces high-resolution polymer features [4]. In this process, a liquid acrylate formulation was photopolymerized in a glass template to generate the required topographical features. While the Philips process shows promise for creating high-resolution images, it did not produce high-aspect-ratio images, and the patterned acrylate polymers lack the etch resistance required for semiconductor manufacturing.

Step-and-flash imprint lithography
Evaluating our experience and that of others, we choose to refocus our efforts on a different technique that we named "step and flash imprint lithography" (SFIL, see "Step and flash methods"). SFIL is a high-throughput, low-cost approach to generating relief patterns with sub-100nm linewidth. It does not use projection optics or lenses, and operates at room temperature. The SFIL process relies largely on chemical and low-pressure mechanical processes to transfer patterns.

The two key differences between SFIL and other imprint lithography techniques are its use of a low-viscosity photo-curable organosilicon liquid and a transparent rigid template. The low viscosity of the photo-curable liquid eliminates the need for high temperatures and pressures. The rigid imprint template is transparent to allow flood exposure of the photopolymer to achieve cure, and to enable classical optical techniques for layer-to-layer alignment.

With SFIL, an organic transfer layer is spin-coated onto a silicon substrate. Then a low-viscosity photopolymerizable organosilicon solution is dispensed on the wafer to form an etch barrier in the area to be imprinted. A surface-treated (i.e., a release layer) transparent template bearing patterned relief structures is aligned over the coated silicon substrate. The template is lowered into contact with the substrate, thereby displacing the etch barrier filling the imprint field and trapping the photopolymerizable liquid in the template relief. Irradiation with UV light through the backside of the template cures the photopolymer. The template is then separated from the substrate, leaving an organosilicon relief image on the surface of the coated substrate that is a replica of the template pattern. A short halogen etch is used to clear any undisplaced etch barrier material so that the underlying transfer layer is completely exposed in the shallows of the pattern. Finally, an oxygen reactive ion etch into the transfer layer amplifies the aspect ratio of the imprinted image.

The imprinting process is conducted at room temperature, and since the template is transparent, all alignment schemes that have been used successfully with mask aligners, steppers, and scanners can be used with SFIL. While the process is simple in concept, every step in the process presents interesting challenges in engineering and materials science.

SFIL advantages
Linewidth variations resulting from statistical fluctuations in processing conditions in conventional optical lithography cause a distribution of gate lengths and a corresponding distribution in device performance that requires "product sorting." These so-called DL variations are the manifestations of accumulated variance in parameters such as focus, exposure dose, development time, humidity, temperature, and amine concentration, all of which affect critical dimensions (CDs). While there is a level of variability in the production of SFIL templates, once a template is on a tool, wafer-to-wafer variance in CDs is very small because imprint pattern definition is essentially process independent. So, we expect to see a very small variance in CDs using imprint lithography and a very large improvement in wafer-to-wafer and lot-to-lot CD variability.

Within the realm of low-cost imprint lithography, SFIL has several advantages over other imprint techniques. The low-pressure, room-temperature replication conditions used in SFIL provide the potential for precise layer-to-layer alignment by minimizing the force applied to the substrate-template interface, hence minimizing material distortion of both the substrate and template. In addition, patterning at room temperature eliminates thermal expansion mismatch problems.

The UV and visible transparency of SFIL templates allows for the application of existing overlay alignment techniques.

An additional benefit is that the production of SFIL templates takes advantage of existing reticle fabrication techniques and infrastructure (discussed below). The high-resolution, high-aspect ratio features produced by SFIL have yet to be demonstrated by other imprint techniques.

Imprint template
Imprint lithography is a 1x-pattern transfer process. The design and production of a high-quality template is therefore key to its success. Currently, templates are prepared following standard phase-shift mask manufacturing techniques: A resist-on-chromium-coated quartz mask blank is patterned with an electron beam, and the exposed resist is developed away (i.e., a positive tone process). Then, the exposed chromium is removed with a dry etch process and the quartz is etched using a standard phase-shift etch process, creating topography in mask quartz. The remaining chromium is then stripped, and 1 in.2 templates are cut from the standard 6 in. x 6 in. x 0.25 in. quartz plate to fit the holder on our imprint apparatus.

We chose this particular template size because it represents a larger area than the maximum die size called for in the International Technology Roadmap for Semiconductors (ITRS). Our most recent templates were produced at the DuPont Reticle Technology Center, but IBM-Burlington, Agilent, and the Naval Research Laboratory have also provided templates.

The ultimate resolution of imprint technologies is limited by the resolution of the imprint template or mold. It is therefore desirable to extend the ability to pattern these templates to coincide with the ITRS, which calls for 65nm minimum resist features for microprocessor gate length and 130nm minimum mask feature size for optical proximity correction features by 2005. For 1x pattern transfer with imprint lithography, there would be need to accelerate mask feature size targets in the ITRS to coincide with the resist feature targets. Patterns near this resolution are being printed today, but only as optical proximity correction features. This resolution goal is far easier to achieve on rigid planar SFIL templates than on wafers, particularly since the throughput of the maskmaking process is not constrained.

Release layer
It is imperative, following exposure, that the etch barrier remain adhered to the underlying transfer layer, releasing easily and completely from the template. It is therefore necessary to modify the surface energy of the template to promote selective release at the template-etch barrier interface.

Figure 3. Water contact angle on an untreated SFIL template (left), a newly treated template (center), and a treated template used and cleaned vigorously for two months (right).Click here to enlarge image

The surface treatment procedure currently used in the SFIL process is based on a self-assembling fluorinated trichlorosilane treatment [13]. This surface treatment reaction has yielded surface energies ~12 dynes/cm. (In comparison, polytetrafluoroethylene or PTFE Teflon has a surface energy of 18 dynes/cm [14].) The surface treatment must maintain its release characteristics through hundreds or thousands of imprints in a manufacturing process. Preliminary results indicate that our current technique could provide films with the required durability. Our tests comparing untreated and treated templates show that the latter maintain release functionality for up to two months (Fig. 3); we have seen no evidence of catastrophic loss of release function. Work is under way to quantify film durability for a variety of treatment conditions.

Etch barrier
Several critical issues must be considered in designing the SFIL etch barrier chemistry, including adhesion, photopolymerization kinetics, shrinkage, and etch selectivity. Tailoring surface properties is crucial. The etch barrier fluid must wet the template well to facilitate filling the topography, yet it must release from the template readily after exposure. These requirements are conflicting, and trade-offs must be analyzed and understood.

We formulated our first etch barrier solutions from a free radical generator dissolved in a solution of organic monomer, silylated monomer, and a dimethyl siloxane (DMS) oligomer. Each component serves a specific role in meeting the end use requirements. The free radical generator initiates polymerization upon exposure. The organic monomer ensures adequate solubility of the free radical generator and adhesion to the organic transfer layer. The silylated monomers and the DMS provide the silicon required to give high-oxygen etch resistance. Both monomer types help maintain the low viscosity required for filling. The silylated monomer and DMS derivative also serve to lower the surface energy, allowing for template release.

Our current formulation is made from commercially available monomers and DMS derivatives [3]. The dose of 365nm radiation required to achieve cure is approximately 20-50 mJ/cm² [14]. The surface energy is ~28 dynes/cm [3]. The etch barrier contains greater than 15 (w/w) silicon providing excellent oxygen reactive ion etch selectivity with respect to the underlying organic transfer layer [15].

Transfer layer
The transfer layer is an organic film that can be tailored in thickness and properties for specific applications, but must follow certain guidelines. It cannot be soluble in the liquid etch barrier, and must adhere well to the cured etch barrier. The transfer layer must remain intact after exposure. There should be significant etch selectivity between the silicon-containing etch barrier and the transfer layer to obtain aspect ratio magnification. These and other issues are being considered as materials development continues. We have used a variety of commercially available antireflection layers, PMMA, and Olin HR100 as transfer layer materials.

Figure 4. UT's step-and-flash imprint stepper.Click here to enlarge image

Equipment
Imprint lithography relies on the parallel orientation of imprint template and substrate. Inaccurate orientation may yield a layer of cured etch barrier that is nonuniform across the imprint field. It is thus necessary to develop a mechanical system whereby template and substrate are brought into co-parallelism during etch barrier exposure. This is achieved in SFIL by way of a two-step orientation scheme. In step one, template stage and wafer chuck are brought into course parallelism via micrometer actuation. The second step is a passive flexure-based mechanism that takes over during actual imprint [16, 17].

For our work, IBM donated a 248nm Ultratech stepper, which we converted to an imprint step-and-repeat tool. As modified, the major machine components include:

• a micro-resolution z-stage that controls the average distance between the template and the substrate and the imprinting force;
• an automated x-y stage for step-and-repeat positioning;
• a pre-calibration stage that enables parallel alignment between the template and substrate by compensating for orientation errors introduced during template installation;
• a fine-orientation flexure stage that provides a highly accurate, automatic parallel alignment of the template and wafer to the order of tens of nanometers across an inch;
• a flexure-based wafer calibration stage that orients the top of the wafer surface parallel with respect to the plane of the x-y stage;
• an exposure source that is used to cure the etch barrier;
• an automated fluid delivery system that accurately dispenses known amounts of the liquid etch barrier; and
• load cells that provide imprinting and separation force data.

Our "multi-imprint apparatus" (Fig. 4) is currently configured to handle 1 in.2 square templates. It is used to produce more than 20 imprints on 200mm wafers for defect studies. Templates and wafers are loaded and unloaded manually. Printing operations, including x-y positioning of the wafer, dispensing etch barrier liquid, z-translation of the template to close the gap between the template and wafer, UV curing of etch barrier, and controlled separation are all automated [16], controlled with a LabVIEW interface.

What about defects?
Imprint lithography skeptics insist that this type of contact patterning will necessarily create defects in the resulting pattern.

Figure 5. Even very large defects a) completely disappeared b) after eight imprints.Click here to enlarge image

A study is under way to investigate creation and propagation of defects in SFIL. The process defects one may encounter in contact patterning may be divided into three groups (neglecting template pattern errors): particles or contaminants that originate on the imprint template; bubbles formed during the etch barrier displacement; and pattern defects caused by features adhering to the template and pulling away from the substrate.

The first type is discussed below. It can be shown, although it is not done so here, that the second type defects are not seen with low-viscosity etch barrier solutions, even in very high-resolution features. And we generally do not see third defect types as long as the surface treatment and etch barrier are prepared according to specification.

Figure 6. a) 100nm Ti lines patterned via metal lift-off and b) a micropolarizer array illuminated with polarized light.Click here to enlarge image

Wafers with multiple imprints were carefully analyzed for defects. One region of the imprint field was tracked by optical microscopy through multiple consecutive imprints. The size of the defects was estimated and tracked. Figure 5a shows a field of severe defects, which was followed through consecutive imprints. After eight imprints, the region was visually free of defects (Fig. 5b). We believe that the defects are removed by entrainment in the etch barrier, thus cleaning the template for subsequent imprints. It appears that the process is self-cleaning for contaminants on the template; the best way to clean a template is to use it.

Results
The UT-Agilent collaboration [15] has resulted in some significant process development; e.g., we have used a high-resolution template with an array of orthogonal 100nm lines and spaces to create a micropolarizer array. Following imprinting and etch transfer, we deposited titanium (Ti) and immersed the substrate in an acetone bath to lift off resist features, leaving only the Ti that was adhered to the substrate. The result was an alternating array of orthogonal metal gratings that made up the micropolarizers (Fig. 6).

Figure 7. 250nm features transferred through an organic planarization organic layer on a substrate with a 700nm tall pre-existing grating structure.Click here to enlarge image

The ultimate goal of the Agilent collaboration was to test a process for patterning a non-flat substrate using imprint lithography. To that end, PMMA was spin coated to provide a thin organic layer over a prepatterned substrate, typically a grating structure. Then, a planarization layer of an acrylate photopolymer was cured over the hard-baked PMMA using a nonpatterned optical flat for the imprint. Finally, the etch barrier was patterned over the planarized organic layer. In the end, we generated high-aspect ratio resist features using the same etch transfer process as that used for flat substrates. Features as small as 250nm were etched over 700nm topography (Fig. 7).

Conclusion
SFIL appears to have several advantages over comparable compression imprinting techniques for applications that require precision layer-to-layer alignment error measurement. SFIL is capable of high-resolution patterning at room temperature with 2-3 PSI applied pressure. The process uses chemicals that are commercially available for sub-100nm patterns.

We have constructed a tool to run automated SFIL imprinting experiments. Imprint templates, made by standard photomask processes, were treated with a low-surface energy self-assembled monolayer to aid selective release at the template-etch barrier interface. This surface treatment was shown to be quite durable, surviving repeated imprints and multiple aggressive physical cleanings without loss of function. Imprints are made in a low-viscosity, photopolymerizable formulation containing organosilicon precursors. The etch barrier cures with approximately 20-50 mJ/cm² of broadband radiation.

The SFIL process appears to be self-cleaning. The number and size of imprinted defects resulting from template contamination decreased with each successive imprint; the imprint field was contamination-free after eight imprints. Contamination on the template was observed to be entrained in the polymerized etch barrier. The SFIL multilayer scheme has been successfully applied to the patterning of 60nm lines with 6:1 aspect ratio, and 80nm features with 14:1 aspect ratio. Using metal lift-off, we have successfully patterned 100nm metal lines and spaces and generated a functional micropolarizer array. 250nm features with high-aspect ratio were transferred over 700nm topography. The SFIL technology is now being commercialized by Molecular Imprints in Austin, TX [18].

Acknowledgments
An additional author of this article is Annette Grot, who is now with Agilent Technologies, Palo Alto, CA.

The authors thank DPI-RTC, IBM, Ultratech Stepper, ETEC, 3M, NRL, International Sematech, and Compugraphics for generous gifts and technical consultation. Special thanks to R. "Jaga" Jagannathan, Franklin Kalk, and David Markle for their assistance. We gratefully acknowledge the financial support of DARPA (MDA972-97-1-0010) and SRC (96-LC-460).

LabVIEW is a registered trademark of National Instruments.

References

1. 3rd International Sematech Next Generation Lithography Workshop, Colorado, 1999.
2. S.Y. Chou, P.R. Krauss, P.J. Renstrom, "Nanoimprint lithography," J. Vac. Sci. Tech. B, 14(6): p. 4129, 1996.
3. M. Colburn, "Step and Flash Imprint Lithography: An Alternative Approach to High-Resolution Patterning," Proc. of SPIE Symposium on Microlithography: Emerging Lithographic Technologies III, 1999.
4. J. Haisma, et al., "Mold-assisted Nanolithography: A Process for Reliable Pattern Replication," J. Vac. Sci. Tech. B, 14(6): p. 4124, 1996.
5. Y. Xia, G.M. Whitesides, Soft Lithography, Angew, Chem. Int. Ed. Engl., 37, p. 550, 1998.
6. R.W. Jaszewski et al., "Hot Embossing in Polymers as a Direct Way to Pattern Resist," Microelectronic Engineering, 41/42, p. 575, 1998.
7. H-C. Scheer et al., "Problems of the Nanoimprinting Technique for nm-scale Pattern Definition," Electron, Ion and Photon Beam Technol. and Nanofabrication, Chicago, 1998.
8. H. Schultz et al., "New Polymer Materials for Nanoimprinting," J. Vac. Sci. Tech. B, 18(4): p. 1861, 2000.
9. M. Li, et al., "3D Microfabrication Using Deep Nanoimprint Lithography," Electron, Ion and Photon Beam Technol. and Nanofabrication, 2000.
10. W. Zhang, S.Y. Chou, "Multilevel Nanoimprint Lithography With Submicron Alignment on 4" Wafers," Electron, Ion and Photon Beam Technol. and Nanofabrication.
11. J. Shaw et al., "Polysiloxanes for Optical Lithography," Solid State Technology, 30(6): p. 83, 1987.
12. G. Bouwhuis et al., "Principles of Optical Disc Systems," Bristol: Adam Hilger, 1985.
13. T. Bailey et al., "Step and Flash Imprint Lithography: Template Surface Treatment and Defect Analysis," J. Vac. Sci. Tech. B, 18(6): p. 3572, 2000.
14. A. Ulman, An Introduction to Ultrathin Organic Films from Langmuir-Blodgett to Self-Assembly, Academic Press, Boston, 442, 1991.
15. M. Colburn et al., "Step and Flash Imprint Lithography for Sub-100nm Patterning," SPIE's 25th Intl. Symp. Microlithography: Emerging Lithographic Technologies IV, Santa Clara, CA, 2000.
16. B.J. Choi et al., "Partially Constrained Compliant Stages for High-Resolution Imprint Litho.," ASME DETC2000/MECH-14145, Baltimore, MD, 2000.
17. B. J. Choi et al., "Design of Orientation Stages for Step and Flash Imprint Lithography," ASPE 1999 Annual Meeting.
18. For more information, contact Norman E. Schumaker, CEO; ph 512/899-8539, e-mail [email protected].

Mathew Colburn received his BS at Purdue University and his MS at the University of Texas. Coburn is the Richard W. Moncrief Endowed Graduate Fellow and is pursuing his PhD in chemical engineering at U. of Texas.

Todd Bailey received his AS at Hudson Valley Community College and his BS at Massachusetts Institute of Technology. He is a Thrust Fellow at the University of Texas' Austin College of Engineering and is pursuing his PhD in chemical engineering.

Byung Jin Choi received his BS at Hanyang University, Seoul, and his PhD in mechanical engineering at the University of Texas, where he is a post-doctorate fellow.

S.V. Sreenivasan received his PhD in mechanical engineering at Ohio State University. Sreenivasan is an associate professor of mechanical engineering at the U. of Texas.

John G. Ekerdt received his BS at University of Wisconsin and PhD at the University of California. He is the Z.D. Bonner Professor in Chemical Engineering and Department Chair at U. of Texas.

C. Grant Willson received his BS and PhD from the University of California at Berkeley and his MS from San Diego State University. Willson is a professor of chemistry and chemical engineering and holder of the Rashid Engineering Regents Chair at the U. of Texas, Willson Research Group, WELCH 5.240, Austin, TX 78712; ph 512/471-3975, fax 512/471-7222, e-mail [email protected].

__________________________________________

Click here to enlarge image

Step and flash methods
The University of Texas (UT) and UT-Agilent SFIL processes (see illustration) both use a low-viscosity monomeric photopolymerizable etch barrier applied as a liquid over a transfer layer and a transparent imprint template that allow backside UV illumination. The UT process uses a template-to-substrate alignment scheme that traps the etch barrier, closing the alignment gap so the force creates a thin layer. The UT-Agilent process uses a laminator roller (depicted as a black oval) and a compliant, transparent template to minimize the etch barrier base layer. In both processes, the imprint is illuminated through the backside of the template to cure the etch barrier. The template is withdrawn, leaving low-aspect ratio, high-resolution features in the etch barrier. The residual etch barrier at the bottom of patterns is etched away with a short halogen plasma etch. Then the pattern is transferred to the underlying layer with an anisotropic oxygen reactive ion etch, creating high-aspect ratio, high-resolution features in the organic transfer layer that can be utilized as-is or as an etch mask for transferring the features into the substrate.