The continued quest for miniaturization of feature sizes in microelectronic, optoelectronic, and microelectromechanical systems places strong demands on the lithographic and patterning processes used to fabricate such devices. In particular, semiconductor device manufacturing is continuing to push the limits of the high volume optical lithography processes and materials used for fabricating the current sub-100nm generations of device features. Developing so-called “single layer” resist (SLR) materials and processes that can enable such sub-100nm patterning methods in the face of the extremely limited focus latitude available in modern optical projection lithography tools and yet still provide sufficient etch barrier toughness continues to be a significant challenge and potential future roadblock. In SLR approaches, these two issues are coupled because they both depend directly on the resist film thickness. One way to overcome these issues is to decouple the depth of focus limitation from the etch barrier toughness by transitioning to bilayer resist materials and processes. The basic idea in these bilayer methods is to perform the initial pattern in an extremely thin film which is coated onto a thicker etch barrier layer and subsequently transfer that pattern through the etch barrier layer to form the final resist pattern. Such bilayer methods have been developed but have suffered from a number of problems including material compatibility limitations which inhibit design flexibility, and process complexity of the multiple coating steps required. A related but different approach is to employ so-called “top surface imaging” approaches in which a single layer resist is subsequently chemically modified in its near surface region after pattern-wise exposure to render selected regions of the resist film surface more resistant to subsequent etching processes. Again, the final patterned resist is formed by transferring the pattern formed by exposure and chemical modification throughout the entire resist film thickness by an etching technique.

Such top surface imaging (TSI) techniques using vapor or liquid phase silylation have been investigated extensively as alternatives to conventional resist processing. Indeed, a variety of different process schemes including Si-CARL, DESIRE, PRIME, SUPER, and Digital-Top-Surface-Imaging have been considered over the past 20 years. Ideally, TSI via silylation methods involves the selective silylation of specific regions of a resist while the other regions are not silylated, i.e. the incorporation of silicon into either only the exposed or unexposed regions of the resist film. Patterns are obtained by selectively etching away the unsilylated regions with an oxygen plasma, whereas the silicon present in silylated regions forms a glassy SiO2 layer on the resist surface which serves as an oxygen plasma etch barrier and prevents etching of the underlying organic resist layer.

TSI imaging schemes have several potential benefits. Since only the top portion of the resist film is imaged, the resist may be relatively opaque at the wavelength of the imaging tool, thus allowing the use of a wider variety of organic resist materials with a particular exposure wavelength. In addition, since the imaging reaction in principle only needs to occur in the surface layer of the resist film, the required depth of focus for the imaging tool and process can be minimized, thereby widening the process window of the imaging process. A further advantage is that extremely high aspect ratio resist features can be produced since the development of the pattern in the organic film is achieved by pattern transfer from the surface image into the bulk of the organic resist film using a plasma etch which can be tailored to be highly anisotropic. However, a variety of difficulties have limited the successful application of such TSI approaches. One of the main challenges with TSI processes has been finding ways to reduce the line edge roughness (LER) of the TSI-generated patterns to within acceptable values. Earlier imaging schemes such as DESIRE relied on crosslinking of the polymer resist film in the exposed areas to slow the diffusion of the silylating agent into the exposed areas, thus producing different concentrations of Si in the exposed and unexposed regions. Since some Si remained even in exposed regions where it was not desired, a plasma descum process was required to remove the thin layer of silicon incorporated into crosslinked regions before performing the O2 plasma pattern transfer etch.1, 4 Furthermore, differences in the crosslinking density and subsequent amount of silicon incorporation across the width of an optically exposed feature led to non-uniform silylation profiles, commonly referred to as “bird's beak” profiles, which ultimately led to difficulty with critical dimension (CD) control of the feature and increased the LER of the overall process. So-called “digital silylation” techniques were developed to overcome these problems with the cross-linking based silylation processes. Digital silylation offered the potential to eliminate the need for a descum step because the active sites on the polymer which participate in the silylation are generated during exposure and post exposure baking (PEB) of the resist. In this manner, the unexposed areas are nominally unreactive towards the silylating agents and thus a descum etch is in principle not required. However, due to the non-uniform energy deposition profile across a feature which results from the non-ideal aerial image produced using optical projection tools, the concentration of active sites in the exposed polymer varies across the feature width. The amount of Si incorporated in such features using the digital silylation processes thus decreases from the feature center to the feature edges. Therefore, the thickness of the silicon etch barrier formed at the feature edges is still thinner than the bulk of the feature and so it does not serve as an effective etch mask in the oxygen plasma which still results in CD control problems and unacceptable LER.

Recently, significant efforts have been undertaken to develop area selective atomic layer deposition techniques (ASALDT). During atomic layer deposition (ALD), thin film growth is controlled by self limiting surface reactions. Thus, if a surface can be modified to prevent the surface reactions involved in the ALD process, nucleation and film growth during ALD can also be prevented on those regions of the surface. ASALDT utilizes this approach to block nucleation in certain areas while allowing film deposition only in desired areas. Previous studies reported the use of octadecyltrichlorosilane SAMs to inhibit nucleation of ALD films. More recently, we have shown that patterned polymer films can provide a much more robust and simpler method for achieving selective ALD. These recent investigations demonstrated that nucleation of titania films produced by ALD can be successfully blocked on polymer materials that do not contain OH groups in their backbone. In addition, we observed that under surface conditions where ALD films nucleate on the polymer layers, a conformal and defect free film can be formed directly on the polymer.

In this presentation, a novel method for utilizing area selective ALD to perform TSI is presented. The approach involves the selective deposition of a inorganic etch barrier , i.e. a metal oxide in this work, on predefined areas on the surface of a polymer film which have been delineated by radiation exposure, followed by selective etching of the regions of the polymer film not masked by the etch barrier layer to obtain the desired pattern. This TSI method offers the potential to overcome the problem of thin etch barrier thickness at the feature edge encountered in other TSI methods (e.g. digital silylation) which result from low silicon incorporation at the feature edge. In the ALD-based TSI approach presented here, the amount of etch barrier material incorporated into the film structure does not strictly depend on the total concentration of active sites generated in the exposed polymer volume as is the case with previous TSI approaches. The nucleation of the ALD film growth depends only on the presence of active sites on the polymer film surface, and thus the amount of ALD etch barrier deposited does not depend on the total hydroxyl concentration in the volume of the polymer film but instead depends on the number of ALD growth cycles performed. Furthermore, since the thickness of the etch barrier formed at the feature does not depend directly on the energy deposition profile and resulting volume polymer hydroxyl content, it should be possible to use ALD to grow thicker more robust etch barrier layers that can withstand the plasma pattern transfer etching processes and thus help to reduce LER in such features. Finally, because deposition of the etch barrier material can be extremely selective, a plasma descum etch is not required which offers the potential to simplify the overall imaging process. While there are many questions that must be answered to fully validate the promise of this new technique, this presentation will give a status report of results of experiments that both demonstrate the feasibility of achieving TSI using an area selective ALD process on polymer films and on experiments directed at understanding the fundamental behavior and limitations of the method.