Using reverse-tone bilayer etch in ultraviolet nanoimprint lithography

S. V. Sreenivasan, Ian McMackin, Frank Xu, David Wang, and Nick Stacey, Molecular Imprints; and Doug Resnick, Motorola Research Laboratories

While nanoscale feature replication using imprinting or micromolding techniques has existed for several years, it was first suggested as a potential patterning approach nearly 50 years ago. Richard Feynmann, in his famous 1959 lecture "There's Plenty of Room at the Bottom: An Invitation to Enter a New Field of Physics," discussed the creation and mold-based replication of nanoscale features.¹ He correctly predicted that the original mold (template) can be written using a version of electron-beam lithography and discussed the possibility of taking a metal mold (template) and making multiple copies: "We would just need to press the same metal plate again into plastic and we would have another copy."

The first known large-scale attempt to use this technique involved a template with recessed structures that was impressed onto a thermoplastic material. With the combination of heat and pressure, the pattern in the template was transferred to the thermoplastic material. Compact discs were an early application of this technology. In the mid-1990s, research in the area of nanoscale replication created features as small as 10 nm.²

However, to be extended to a broad set of applications, nanoscale replication must overcome several practical challenges:

• It must be able to print fields with nonuniform pattern density at adequate throughput.

• It must etch nanostructures with appropriate critical dimension control.

• It must demonstrate precise alignment and overlay capability.

• It must minimize process-induced defects.

To meet these challenges, step-and-flash imprint lithography (S-FIL) was introduced.³ S-FIL is a step-and-repeat nano-replication technique based on low-viscosity ultraviolet (UV)- curable liquids. The use of low-viscosity monomers (which have viscosities of <5 cps) enables a low-imprint-pressure (<0.25 psi) process, resulting in low process defect levels. Furthermore, in situ nanoscale alignment corrections can be made in low-viscosity liquids prior to UV curing, resulting in <10 nm (3σ) alignment capability over 25 X 25-mm fields.⁴

A bilayer approach, the S-FIL process imprints features into a silicon-containing material that has been deposited onto an underlying organic layer. This approach enables the patterning of features with relatively low aspect ratios. The aspect ratios in the patterned material can then be amplified using an O₂ reactive-ion etch (RIE) step to etch the underlying organic material. S-FIL's ability to pattern low aspect ratios is key to minimizing defects, particularly when the template is separated from the UV-cured material.

Challenges to S-FIL and Other Imprint Techniques

In sub-50-nm imprint lithography, the etch process requires an increasingly thin residual film and increasingly high aspect ratios in the smallest features.⁵ The reason for this requirement is illustrated in Figure 1a. A thick residual layer with low-aspect-ratio features causes significant faceting in small isolated structures during the breakthrough etch step, potentially resulting in undesirable etch bias or even feature loss.

Figure 1: Schematic diagram showing (a) low-aspect-ratio features with thick residual layer, and (b) higher-aspect-ratio features. Minimizing the etch bias in small, isolated features in the S-FIL process requires the patterning of thin residual layers, the patterning of tall imprint structures, or both.

One way to avoid this problem is to print thin residual layers and/or tall structures. For example, if 30-nm isolated lines must be etched into the substrate and aspect ratios no larger than 3:1 are required to minimize defect propagation, the 30-nm features will be <90 nm tall. To minimize the undesirable effects of faceting in isolated features, the thickness of the residual layer must be less than one-third the feature height. However, this approach requires an etch transfer process that results in inferior in situ (in-liquid) alignment, defect levels, and throughput.

Based on the literature and on experiments performed by Molecular Imprints (Austin, TX) and Motorola Research Laboratories (Tempe, AZ), the four areas that affect the replication process can be summarized as follows.

Etching with Good Pattern Fidelity. All nanoimprint lithography techniques require a breakthrough etch step to eliminate the residual layer during pattern transfer. Good etching requires a thin residual layer and features with high aspect ratios. As illustrated in Figure 1b, the presence of a thin residual layer and high aspect ratios leads to minimal faceting during breakthrough etch.

Aligning Multiple Levels. In-liquid alignment has been shown to be promising for nanoresolution alignment. The low-viscosity monomer material that makes low-pressure printing possible also enables the template, while it is in close proximity to the wafer, to make nanoscale gliding movements, allowing precise in situ alignment corrections. The S-FIL process uses field-by-field, in-liquid, through-the-template alignment, which results in nanometer-scale alignment correction capability and in situ correction prior to UV exposure, where the alignment is locked in. In-liquid alignment at the sub-10-nm level before and after exposure is shown in Figure 2. However, it has been found that the in-liquid alignment approach can be disrupted if the template makes direct contact with the substrate or if a particle is trapped between the template and substrate, a situation that is aggravated by thin residual films. Hence, a residual layer of 100 nm or more is desirable to achieve robust in-liquid alignment.

Figure 2: In-liquid alignment before and after exposure. The average preexposure alignment is 7.8 nm (3σ), and the average postexposure alignment is 9.1 nm (3σ).

Low Defectivity Levels. Low defectivity requires thick residual film and features with low aspect ratios. Two classes of defects have the greatest impact on imprint lithography. The first involves the cohesive failure of film material when the template is separated from the cured material. Data indicate that this problem is aggravated in the regions where the highest-aspect-ratio structures exist. The schematic diagrams in Figures 3a and 3b compare features with low and high aspect ratios. In Figure 3b, the high-aspect-ratio, high-resolution feature on the left is most likely to fail during the separation of the template from the cured material.

Figure 3: Comparison between features with (a) low and (b) high height. Defects are caused by material failure in small features with high feature height. Hence, the circled feature is most likely to fail when the template is separated from the cured material.

The second class of defects that affects imprint lithography is caused by template damage. In the S-FIL process, a liquid film separates the template from the wafer, preventing direct contact and template defect creation during the printing process. However, when undesirable particles on the wafer (hard particles that are larger than the residual layer thickness) go undetected before imprinting, the template can make direct contact with them and suffer damage. The use of thick residual films (e.g., >0.25 µm) and lower-aspect-ratio structures makes it easier to detect undesirable particles and lower defect levels.

High Throughput. High throughput requires thick residual film and features with low aspect ratio. An important component of the throughput budget in imprint lithography is the time required for the UV-curable liquid to fill all the features fully. The use of thick residual layers leads to a wider channel for liquid to complete the filling process. It has been reported in the literature that filling speed is a cubic function of the width of the channel.⁶ Therefore, thicker residual layers allow liquid to redistribute significantly faster than do thinner layers. In addition, low-aspect-ratio features require less liquid redistribution than high-aspect-ratio ones, since the regions that contain the features on the template (recessed structures) require less liquid.

The matrix in Table I shows the performance of both thin and thick residual layers with either low-aspect-ratio or high-aspect-ratio features. From the standpoint of alignment, defectivity, and throughput, thick residual layers and low aspect ratios are optimal. However, the opposite is true for pattern transfer in the S-FIL process, since thin residual layers and high-aspect-ratio structures are needed to maintain the fidelity of isolated small structures. Consequently, a process that is significantly better in every aspect can be achieved by establishing an effective etch process for thick residual layers and low-aspect-ratio structures. That process is known as reverse-tone S-FIL (S-FIL/R).

Type of Layer Low-Aspect-Ratio Imprints High-Aspect-Ratio Imprints Thin residual layer E, T, D, A E, T, D, A Thick residual layer E, T, D, A E, T, D, A

Table I: Imprint lithography performance matrix for etch (E), throughput (T), defectivity (D), and alignment (A). (Green = good, blue = fair, red = poor.)

S-FIL/R and the Reverse-Tone Etch Process

A variant of the S-FIL process, S-FIL/R uses a reverse-tone bilayer etch technique. As shown in Figure 4a, step 1 of the process employs a planarizing topcoat (a spin-on or imprint-planarized silicon-containing organic material) on an organic imprint material. An optional organic resist material underneath the imprint layer may be used to provide better etch selectivity with the underlying substrate and/or better adhesion with the imprinted material. In step 2 (Figure 4b), the silicon topcoat is removed using either a dry or wet etch-back process or CMP to expose the organic imprint material below. In step 3 (Figure 4c), a selective O₂ RIE etch step is typically used to create a hard mask in the silicon-containing material. The hard mask enables the etching of thick organic films to create tall resist structures for pattern transfer.

Figure 4: The S-FIL/R process improves the ability of imprint lithography to align multiple layers, reduce defects, maintain etch CD control, and maximize throughput.

This approach is different from other bilayer approaches since the etch process reverses the tone of the lithographic structure created in the original patterned material. Moreover, blanket material removal in step 2 does not have a negative effect on the shape of the imprinted feature and, therefore, critical dimension (CD) control, while blanket breakthrough etch in the S-FIL process causes faceting of the imprinted features and CD control issues.

The S-FIL/R approach not only enables the printing of significantly thicker imprints without adversely affecting etch transfer, but it also enables the printing of low-aspect-ratio structures. The topcoat material can be readily designed to incorporate ultrahigh (>20%) silicon by weight, which is difficult to incorporate into imprint materials, given all the other process constraints. Therefore, the feature height required to achieve an acceptable etch mask in the bilayer resist can be smaller than half that of a traditional imprint process. These two key features—thicker residual layers and lower-aspect-ratio imprinted structures—result in better in-liquid alignment capability, lower susceptibility to defects, better CD control through etch, and the potential for higher throughput. A scanning electron microscope (SEM) image showing 60-nm dense lines after bilayer S-FIL/R resist etch is presented in Figure 5.

Figure 5: SEM image of 60-nm 1:1 features obtained using the S-FIL/R process.

Reverse toning has general applicability. It can be utilized for any imprint lithography technique, not just S-FIL. It is also applicable to lithographic structures created by E-beam lithography or photolithography. For example, if a negative- tone E-beam resist is desirable from a process and resolution standpoint while a dark-field photomask must be written with the process, the bright-field mask (opposite tone) can be patterned using an E-beam and the reverse-toning etch process can be used to obtain the desired dark-field photomask.

Printing and Etching Over Preexisting Topography

A major benefit of the S-FIL/R process is that it is more tolerant of preexisting topography than the S-FIL process.

Figure 6: Pattern transfer over preexisting topography using the S-FIL process: (a) plane of optical flat (dashed line) during imprint planarization; (b) breakthrough etch, after which region A requires additional etching to open up the underlying organic material and region B does not; and (c) postbreakthrough etch stage, where the residual layer in region A has been removed after additional etching and the residual layer in region B has been eliminated.

Pattern Transfer over Preexisting Topography Using S-FIL. Figure 6 shows a process flow for pattern transfer over preexisting topography. A similar etch transfer process has been described in the literature.³ The preexisting topography is first planarized using an imprint planarization step, in which imprinting is performed using an organic UV-curable liquid and an optical flat. This step is followed by imprinting in which patterned structures are created in the silicon-containing material. During breakthrough etch, minimal residual-layer variation is mandatory. More specifically, if the maximum and minimum residual-layer thickness of the imprinted material is t_max and t_min, the height of the imprinted feature is h, and the parameter s is a safety factor >1, the following inequality constraint must be satisfied:

That equation accounts for the faster deterioration of small isolated structures because of faceting. The variation in the residual layer (Δt) is affected by various factors, including shrinkage of the imprint planarization organic material during UV curing, flatness errors in the optical flat, and the mismatch between the optical-flat surface and the average substrate surface.

The faster deterioration of small isolated structures can be seen in a pattern-transfer case involving the creation of 50-nm structures over topography with 200-nm variations. Since the imprinting of features with aspect ratios greater than 2:1 causes defect problems, the feature height h in that case is limited to 100 nm. There are three types of variations:

• Δt₁: The total residual-layer variation resulting from the mismatch between template and substrate is ~25 nm TIR for double-sided polished silicon wafers. (That variation can be substantially greater for single-sided polished wafers and other substrates, such as smaller silicon, gallium arsenide [GaAs], and indium phosphide [InP] wafers.)

• Δt₂: Shrinkage of ~10% during UV curing causes planarization film nonuniformity of 20 nm.

• Δt₃: The best optical flats over a field can be of λ/20 quality, leading to an additional error of ~30 nm.

If the safety factor for maintaining the CD of isolated features is assumed to be 2, the total Δt must be <50 nm. However, Δt = (Δt₁ + Δt₂ + Δt₃) = 75 nm, an unacceptable error level that leads to the situation depicted in Figure 6c, in which the feature in region B has essentially disappeared by the time the breakthrough etch in region A has been completed and is ready for pattern transfer. That situation can be worse in the case of GaAs and InP substrates.

Pattern Transfer over Preexisting Topography Using S-FIL/R. Figure 7 shows a process flow for pattern transfer over preexisting topography using the S-FIL/R process. In this process, the silicon-containing planarizing topcoat is best applied using a planarizing spin-on film. This planarization process has two major advantages in conjunction with S-FIL/R: First, the faceting of features during blanket breakthrough etch is not a problem, as illustrated in Figure 7b; and second, the degree of planarization represented by Δt is not affected by substrate flatness variations that are of low spatial frequency (more than the order of a millimeter or greater).

Figure 7: Pattern transfer over preexisting topography using the S-FIL/R process (a) before and (b) after breakthrough etch. No faceting of features during etch has occurred.

Although the S-FIL/R process must satisfy the inequality constraint equation, the safety factor can be smaller than in the S-FIL process. In the case of S-FIL/R, Δt is largely affected by the height, size, and distribution of the features, factors that are well understood in the literature.

If S-FIL/R is used for pattern transfer of 50-nm structures over topography with 200-nm variations, the presence of silicon-containing materials makes it possible to imprint features in the organic material with aspect ratios of 1.5:1. The feature height is then about 75 nm.

In many nanoimprint lithography applications, including photonics, magnetic storage, and contact-hole creation in CMOS devices, printed features tend to be gratings, pillars, and contact-holes, which at any given lithographic level are substantially smaller than 1 µm. In addition, in CMOS applications, the data for a given lithographic level can be manipulated to include dummy features, which further assist the spin-on planarization process. Hence, in the S-FIL/R process, the planarization process is largely unaffected by preexisting topography, as depicted in Figure 7a, which shows pattern transfer before and after breakthrough etch. For 75-nm-tall features, a degree of planarization (Δt) that is approximately 10% of the height (~8 nm) has been obtained. In addition, it has been demonstrated that the S-FIL/R process is significantly more robust than the S/FIL process in the case of GaAs and InP substrates.

Patterning and Etching Contacts

Because patterning and etching high-resolution contacts is considered a major challenge for photolithography, these processes often necessitate the use of new lithographic technology. In S-FIL/R, these processes are performed by patterning pillars in the organic material and then conducting spin-on planarization using a silicon-containing material. This approach has two advantages over the standard S-FIL process:

1. The template-fabrication process in S-FIL/R is quite similar to the patterning of binary contact masks, since the total area exposed to an E-beam in a positive-tone resist is about 2% or less. In contrast, patterning an S-FIL contact template using a positive-tone resist is very challenging.

2. In the S-FIL process, holes are created directly in a silicon-containing material, and the height of the holes must be greater than that of the pillars because of faceting issues and the lower silicon content in the imprint material. Consequently, a high-resolution contact template and long features (sub-100-nm pins with aspect ratios of 2:1 or greater) protruding out from the template are required. As a result, the features are very fragile and likely to fail, causing repeating defects. In contrast, the S-FIL/R process uses robust templates with low-aspect-ratio holes.

Figure 8: Patterning using the S-FIL/R process: (a) imprinted organic pillars, and (b) contacts after reverse toning.

Figure 8 presents SEMs of patterned pillars that were reversed into contacts. Figure 9 shows 60-nm contact holes that have been patterned using the S-FIL/R process.

Figure 9: Patterning of 60-nm contacts using the S-FIL/R process.

Conclusion

Lithography tools are sometimes referred to as the milling machines of the 21st century. They have revolutionized the electronics industry and are continuing to enable many applications at the micro- and nanoscale.

It has been shown previously that the S-FIL process can fabricate sub-50-nm structures, complicated patterns, and 3-D structures, while providing precise overlay capability and low process defectivity.^3–5 The S-FIL/R process represents an improvement over the S-FIL process by enabling the printing of lower-aspect-ratio structures and thicker residual films during imprinting. It also has significant advantages over its predecessor in its ability to perform transfer patterning over preexisting topography.

S-FIL and S-FIL/R technology will likely be used to manufacture optical devices, microdisplays, and nanoscale electronics. Their impact on mainstream silicon fabrication will probably be a direct function of how well they can overcome a key manufacturing challenge: minimizing long-term defects to maximize yields. The challenge facing silicon fabrication is to develop and demonstrate an S-FIL process that can approach the long-term yield and productivity of photolithography.

Acknowledgments

The work described in this article was partially supported by grants from the DARPA Advanced Lithography program (Grants No. N66001-01-1-8964 and N66001-02-C-8011) and the NIST Advanced Technology Program (Grant No. 70NANB4H3012).

References

1. RP Feynmann, "There's Plenty of Room at the Bottom: An Invitation to Enter a New Field of Physics," presented at The American Physical Society Meeting, California Institute of Technology, December 29, 1959, published in Caltech's Engineering and Science, February 1960.

2. SY Chou, PR Krauss, and PJ Renstrom, "Nanoimprint Lithography," Journal of Vacuum Science and Technology B 14, no. 6 (1996): 4129–4133.

3. M Colburn et al., "Development and Advantages of Step-and-Flash Imprint Lithography," Solid State Technology 46, no. 7 (2001): 67–78.

4. BJ Choi et al., "Distortion and Overlay Performance of UV Step and Repeat Imprint Lithography" (paper presented at the Micro- and Nano-Engineering International Conference, Rotterdam, The Netherlands, September 19–22, 2004).

5. SV Sreenivasan, "Status of the Step and Flash Imprint Lithography Technology" (paper presented at the Micro- and Nano-Engineering International Conference, Rotterdam, The Netherlands, September 19–22, 2004).

6. M Colburn, "Step and Flash Imprint Lithography," PhD diss., University of Texas at Austin, 2001.

---

S. V. Sreenivasan, PhD, is cofounder and CTO of Molecular Imprints (Austin, TX). He is on a leave of absence from the University of Texas at Austin, where he is an associate professor of mechanical engineering and Thornton Centennial Fellow in engineering. Sreenivasan specializes in the design of ultraprecision machine systems for semiconductor processes. His specific research interests include the study of nanostructure fabrication using imprint lithography, flexure-based micro- and nanoprecision machines, precision optical systems, and real-time sensing architectures. He received a PhD in mechanical engineering from Ohio State University in Columbus. (Sreenivasan can be reached at 512/334-1210 or svs@molecularimprints.com.)

Ian McMackin, PhD, is manager of the process and applications group in the advanced lithography division of Molecular Imprints, which he joined shortly after the company's inception. The group is responsible for developing the lithography process from imprint through pattern-transfer etch, concentrating on CMOS-compatible applications. The group's work encompasses the formulation of imprint resists, development of etch recipes, and engineering of imprint tool subsystems that increase lithography performance. He received a BS in physics from Southern Illinois University in Edwardsville and a PhD from the Institute of Optics, University of Rochester in New York. (McMackin can be reached at 512/339-7760, ext. 215, or imcmackin@molecularimprints.com.)

Frank Xu, PhD, works on material development at Molecular Imprints, which he joined in 2003. Before joining the company, he spent nine years at 3M and Chorum Technologies working on optoelectronic materials. He received a PhD in polymer science and engineering from the University of Massachusetts at Amherst. (Xu can be reached at 512/339-7760, ext. 252, or frank@molecularimprints.com.)

David Wang is an etch process engineer at Molecular Imprints, which he joined in 2003. He received a BS in electrical engineering from the University of Texas at Austin. (Wang can be reached at 512/339-7760, ext. 287, or dwang@molecularimprints.com.)

Nick Stacey is a materials manager at Molecular Imprints, where he develops new imprint materials for use with customer processes. He has been with the company since 2003. Stacey has more than 15 years of experience in the electronics and imprint fields. Previously, he worked at 3M and the University of Texas at Austin. (Stacey can be reached at 512/339-7760, ext. 261, or nstacey@molecularimprints.com.)

Doug Resnick, PhD, is a fellow of the technical staff at Motorola Labs and a member of Motorola's scientific advisory board. He is responsible for developing Motorola's step and flash imprint lithography research program. He joined the company in 1990 and led the development effort for x-ray mask-pattern transfer. Before joining Motorola Labs, Resnick worked at AT&T Bell Laboratories. His development projects included x-ray lithography, GaAs direct-write, and plasma etching of photomasks. He has authored or coauthored more than 100 technical publications and holds 16 U.S. patents. He received a PhD in solid-state physics in 1981 from Ohio State University in Columbus. (Resnick can be reached at 480/413-7743 or doug.resnick@motorola.com.)