Release Layers for Contact and Imprint Lithography

Contact print technology has been widely used since the 1960s to pattern integrated circuits. In addition to having very high throughput, the equipment is relatively inexpensive and far less complex than modern projection printers and scanners. Because of its many advantages, contact printing remained a significant lithography technology in the IC industry well into the 1980s.

Despite these capabilities, contact printing has become obsolete for most critical applications due in part to the very high level of defects that result from the printing process.¹ Modern projection printers have equaled the throughput, and far surpassed the resolution and overlay accuracy of the most advanced contact print systems. However, it is the high defect level associated with vacuum contact printing that has kept this technology from being used for applications where its resolution and overlay capabilities are acceptable.

Using phase-shift masks (PSMs), it has already been demonstrated that 193 nm photolithography can produce sub-100 nm features.² A combination of improved optics, a reduced wavelength to 157 nm, and the introduction of more complex processing will surely enable further reductions in feature size.

Along this path, such improvements come with an ever-increasing cost for photolithographic tools. As conventional projection lithography reaches its limits, next-generation lithography (NGL) tools may provide a means to further pattern shrinks, but are expected to come at a price tag that is prohibitive for many companies.

A significant part of the cost of a projection printing tool is tied up in the optics and the light source. Looking ahead, this trend is not expected to change as tool manufacturers first address the challenges of building lenses from calcium fluoride (CaF₂) for 157 nm tools, and later deal with the issues of building multilayer reflective optics for extreme ultraviolet (EUV, 13 nm) systems.

1. These SEMs show the resolution durability of 750 nm dense lines printed with vacuum contact lithography. The lines printed with an uncoated mask on the first exposure (left) did not hold up by exposure #8 (middle). With a coated mask, on the other hand, the durability was maintained by exposure #200 (right).

For either contact lithography or imprint lithography, defects mainly result from damage imparted to the resist layer caused by direct contact between a mask and a resist-coated substrate. Conventional chromium-coated quartz photomasks have high surface energies and thus have a high potential to adhere to photoresist. When intimately contacted, resist tends to be pulled from the wafer and remain on the mask.

Thus, not only is the resist coating of the immediate wafer damaged, but subsequent layers are likely to be affected, especially if the defect falls into a clearfield area of the mask. The defect will also create a gap between the wafer and mask, potentially degrading resolution in nearby areas. As a result, defects adhering to a mask have a cumulative effect, propagating to subsequent layers until the mask is cleaned.

A potential solution to these problems is to apply a low surface energy release layer directly onto the mask. If a robust release layer can be developed, the impact to lithography could be significant. At the very least, a good release layer will minimize the number of mask cleans necessary in the contact print process. Optimistically, it could help to enable an imprinting process extendable to the 10 nm regime.

Attempts have been made to reduce defect levels by employing a measured gap between the substrate and mask. This very commonly used technique, known as proximity printing, is effective at reducing defect levels. However, resolution is rapidly lost as the gap increases, creating a compromise between defect density and maximum resolution.

Surface treatments for masks have also been used to improve their scratch resistance.³ However, such techniques do not lower surface energy and thus do not decrease the tendency for particulates to adhere. Coatings have also been applied to masks to keep their surfaces clean.⁴ Such strippable lacquer coatings are applied in relatively thick films, then peeled from the mask surface just prior to use. These coatings were never intended for in situ use on masks during the contact printing process for several reasons: They are too thick to permit the level of contact needed for high-resolution printing; they are not anti-sticking, and may create or attract more defects when contacted with resist; and, depending on their chemistry and thickness, they may absorb significant amounts of the exposing light intended for the photoresist.

Conversion coatings using fluorinated silane-based monomers have also been applied to masks to alter their surface energy.⁵ This process, known as SURCAS (surface conversion for anti-sticking), successfully lowers the surface free energy of a coated mask, but lacks adequate durability in consecutive runs required by a production contact print operation.

Attempts have also been made to apply low surface energy release coatings directly to resist-coated wafers.⁶ A formulation of poly vinyl alcohol mixed with a surfactant and a lubricative monomer was applied to a wafer and baked following a normal resist application process. Significant improvements in device yield and mask durability resulted when used in hard contact compared with trials done with uncoated wafers. However, this process suffers from the obvious drawback that each wafer must be taken through an additional coating and baking step.

Amorphous fluoropolymers look especially promising for release-layer technology.⁷ Teflon AF is a family of amorphous fluoropolymers made by DuPont Fluoroproducts (Wilmington, Del.). These materials are made by the copolymerization of 2,2-bistrifluoromethyl-4,5-difluoro-1,3-dioxole (PDD) with other fluorine-containing monomers.

2. These 750 nm dense lines were vacuum contact printed using a Teflon AF coated and contaminated mask. Although the lines failed to be resolved in the first exposure (left), resolution is improved by the second exposure (middle), and completely restored by exposure #25 (right).

As with other fluoropolymer resins, Teflon AF has good thermal stability and chemical resistance along with a very low surface energy. However, unlike other fluoropolymers such as polytetrafluoroethylene (PTFE), Teflon AF is amorphous. This characteristic offers many additional properties that make it particularly useful as a mask coating. Because it has none of the crystallites found in semi-crystalline materials such as PTFE, coatings (<200 µm) of Teflon AF are virtually transparent to light with wavelengths greater than 200 nm.^8,9

In addition, Teflon AF is soluble at room temperature in several fluorocarbon-based solvents, allowing it to be easily spin-coated onto a photomask. It can also be easily removed using the same solvents. The Teflon AF structure is depicted below. For this particular formulation, x was 0.35.

Two series of tests were performed to determine coating durability in normal vacuum contact mode.¹⁰ Two masks of identical resolution were used for this test, one of which was coated with 75 nm of Teflon AF and the other left bare. Each mask was used to sequentially expose several resist-coated wafers without benefit of any interim mask cleans.

It was noted that, during the test, wafers exposed using the uncoated mask would often adhere to the mask following exposure, requiring operator assistance. Nevertheless, the uncoated mask was able to successfully resolve dense lines of 750 nm across the wafer for the first few exposures. However, by the eighth wafer, resolution of these lines had degraded completely.

In contrast, the coated mask exhibited superior durability. Resolution of 750 nm dense lines was maintained for 200 consecutive wafers, where the test was terminated. SEM micrographs from both data sets are shown in Figure 1. Following the 200th exposure, it was noted that the mask coating was slightly marred where the outer periphery of wafers had made contact during exposure. However, the center region of the mask remained not only unscratched, but particle-free when examined under a bright light. The uncoated mask was heavily contaminated with particles after only eight exposures.

Because a surface coated with Teflon AF has a significantly lower surface energy than a resist-coated surface, it is reasonable to assume that masks that become contaminated after Teflon coating may demonstrate a self-cleaning capability. That is, particles may be more likely to adhere to higher-energy wafer/resist surfaces than to coated masks.

To investigate this possibility, a coated resolution mask was left in a dirty environment for 48 hours, becoming contaminated with a wide variety of particles. It was then used to expose 25 wafers consecutively in vacuum contact without mask cleaning. Figure 2 shows the first wafer exposed where, as expected, 0.75 µm dense lines failed to resolve. The other two SEMs in Figure 2 show the same lines as produced on wafers #2 and #25. Resolution has clearly been restored to a level of quality only obtainable with a clean mask.

Cytop is another family of amorphous fluoropolymer materials manufactured by Asahi Glass (Tokyo). Cytop fluoropolymers have a cyclized ring structure that prevents crystallinity in a manner similar to Teflon AF. The physical and mechanical properties of these materials coupled with their amorphous nature also make them suitable as a release coating film for contact printing.

3. Although an uncoated mask shows defects after 10 prints (left), there are no apparent defects on a mask coated with Cytop after 250 imprints (right).

A key performance metric of a release coating used for contact printing is the ability to maintain mask cleanliness for many consecutive exposures. To demonstrate this capability, a resolution mask was coated with 100 nm of Cytop CTL-107M and used to expose 250 consecutive wafers in vacuum contact. An identical uncoated mask was also used for a series of 10 consecutive vacuum exposures. Figure 3 shows photographs of both masks after the test, clearly demonstrating the effectiveness of using a release coating.

Imprint lithography techniques are essentially micromolding processes in which the topography of a template defines the patterns created on the substrate. As in contact lithography, a release layer is necessary to avoid the transfer of resist from the substrate to the template. Because this technology may be suitable for dimensions as small as 10 nm, it is not possible to spin-coat materials such as Teflon AF and Cytop onto the template surface. Good conformality cannot be obtained and the release layer must be extremely thin so that the template feature size is not impacted.

Whitesides et al have formed a template by applying a liquid precursor to polydimethylsiloxane over a master mask produced using either electron-beam or optical lithography.¹¹ The liquid is cured, and the PDMS solid is peeled away from the original mask. The PDMS template can then be coated with a thiol solution, which is subsequently transferred to a substrate and coated with a thin layer of gold.

The process of curing the PDMS against the master can result in unwanted adhesion of the PDMS to the exposed regions of the silicon wafer. The PDMS is sealed irreversibly to the silicon master surface by a cross-linking reaction. To prevent this adhesion, the master surface is passivated by the gas phase deposition of a long-chain, fluorinated alkylchlorosilane (CF₃(CF₂)₆(CH₂)₂SiCl₃).

4. This schematic drawing illustrates the Step and Flash Imprint Lithography process.

Because the PDMS is easily deformable, the technology is not well suited for devices requiring precise pattern placement. Nanoimprint lithography, developed by Chou et al, uses a solid mold, such as silicon or nickel.¹²

The imprint process is accomplished by heating a resist above its glass transition temperature and imparting a relatively large force to transfer the image into the heated resist. To minimize possible adhesion between the resist and the mold, a fluorinated material is typically added to the resist. Features as small as 10 nm have been imaged using this approach.

Devices that require several lithography steps and precise overlay require an imprinting process capable of addressing registration issues. Step and Flash Imprint Lithography (S-FIL), a technique developed by Willson et al, solves the problem of overlay by using a transparent quartz template.¹³

The steps required for patterning are schematically depicted in Figure 4. The process employs a template/substrate alignment scheme to set the template parallel to the substrate. A low-viscosity liquid etch barrier material is then injected between the template and substrate. The gap is closed and ultraviolet light is illuminated through the template, thereby curing the etch barrier. The template is withdrawn, leaving a precisely replicated inverse of the pattern on the template. The viscosity of the etch barrier is sufficiently low, so that minimal pressure (~2-4 psi) and no additional heating is necessary to move the liquid into the stencil. Finally, because the template is transparent, conventional overlay schemes can be used to align patterns.

5. This drawing depicts the formation of a low surface energy release layer.

6. A dirty template was used to imprint several die on a silicon wafer. Defects that start on the template embed themselves in the etch barrier and, by the seventh imprint, there are no detectable particles.

When this functional group is synthetically attached to a long fluorinated aliphatic chain, a bifunctional molecule suitable as a template release film is created. The silane-terminated end bonds itself to a template's surface, providing the durability necessary for repeated imprints. The fluorinated chain, with its tendency to orient itself away from the surface, forms a tightly packed comb-like structure and provides a low-energy release surface. Annealing further enhances the condensation, creating a highly networked, durable, low surface energy coating. The release layer formation process is depicted in Figure 5.

7. Lines measuring 40 and 30 nm were resolved in a UV-curable etch barrier.

As new nanolithography technologies continue to be developed, their potential as cost-effective alternatives to optical lithography methods continues to grow. Surface acoustic wave (SAW) devices, photonic crystals, waveguides and high-density memories are just a few of the products that can be produced provided defect levels can be reliably controlled.

Low surface energy release layers comprised of amorphous fluoropolymers have already been proven to be effective for improving defect levels in contact print lithography. This patented technology is used routinely within Motorola's Physical Sciences Research Labs, and has been shown to provide dramatic improvements in both mask life and mask cleaning frequency.

In addition, release layers have been successfully applied to the surfaces of imprint templates with similar results. It is unclear at this early stage of development whether nano imprint lithographies will supplant more established optical methodologies. However, it is clear that a successful release layer technology is required to control defects and enable the potential of imprint lithography to be achieved.