Introduction

Why extend DUV lithography from 248 to 193 nm?

Many state-of-the-art photoresists now function at a Rayleigh k value of 0.50 under conventional illumination with chrome-on-glass masks. These resists are said to be "linear" to a minimum feature size of /2(NA), with exposure latitude of ±10%, or 250 nm (0.25 µm) in the case of 248-nm lithography at 0.50 NA. While it is possible to fabricate development-level devices under these conditions, the total process window afforded under these circumstances is rarely adequate to manufacture semiconductor product in high yield. Hence, "optical enhancements" such as off-axis illumination and/or phase masks are required to make this lithographic system (comprising tool and resist) function with a process window commensurate with manufacturing requirements. It is thus likely that the transition from 248- to 193-nm lithography will depend largely on the trade-offs among the cost/performance of 248-nm, 0.6-NA optically enhanced lithography, conventional 193-nm, 0.6-NA lithography, and "optically enhanced" 193-nm, 0.6-NA lithography.

If the semiconductor industry ultimately has available 0.6-NA exposure tools, optical enhancements, and both single-layer resists (SLR) and bilayer resists or top-surface imaging (BLR/TSI) resists at 248 and 193 nm, numerous resist/lithography technology choices will be available for the 0.25-0.15-µm device generations. These choices are listed in Table 1. While it is difficult at present to determine which of these options will be chosen for each of the next three generations of semiconductor manufacturing, it is possible that the current need for higher levels of integration and performance in microprocessors and DRAM will lead to economic feasibility for 0.25-0.15-µm manufacturing.

If history is any guide, and if the economics of manufacturing permit, the answer to the question "Why switch to 193-nm lithography?" may be the following:

1. 193-nm lithography offers the possibility to extend DUV lithography with optical enhancements to 0.15-µm dimensions, and possibly to smaller dimensions.
2. "Mild" optical enhancements of 193-nm lithography are likely to afford manufacturable lithographic processes at 0.18-µm dimensions.
3. Backward introduction of 193-nm lithography to 0.25-0.18-µm device manufacturing would afford low-cost, conventional DUV processes without optical enhancements.

Single-layer resist (SLR) approaches

Figure 1

It would indeed be fortunate if the transition from 248- to 193-nm lithography could occur in a relatively smooth manner, like the transition from 436 to 365 nm in the late 1980s. Many 436-nm (G-line) photoresists functioned effectively at 365 nm (I-line) because both the optical properties and the photochemistry of the diazonaphthoquinone-ovolak photoresists were similar at these wavelengths. However, this was not the case in the very difficult subsequent transition from 365-nm lithography to 248-nm imaging. Both the materials and the imaging mechanism (including photochemistry) were necessarily changed completely with the introduction of chemical amplification (CA) (in the form of an acid-catalyzed deprotection chemical reaction)[3].

The degree of difficulty in the transition to 193-nm lithography falls somewhere between these past two lithography transitions. The optical properties of the current 248-nm DUV resists are very different at the 193-nm exposure wavelength. The optical absorbance of a typical p-hydroxystyrene-based DUV resist is shown as a function of wavelength in Figure 2. At a wavelength of 193 nm, the 0.3 absorbance per micron of a 248-nm resist increases to more than 10 per micron. At this high absorbance, 193-nm radiation cannot penetrate through to the bottom of a 248-nm resist layer, rendering commercially available DUV resists useless for exposure at 193 nm. The acid-catalyzed chemical amplification chemistry used in commercial 248-nm DUV resists is completely applicable to 193-nm imaging, however. One complication in this imaging chemistry is the difference in chemical reaction pathways for the generation of photoacid at these different wavelengths, which is discussed below. The introduction of 193-nm lithography thus requires the development of a family (or families) of photoresist materials and chemistries specifically tailored to function at this wavelength.

Figure 2

A high-performance single-layer photoresist is needed for the successful introduction of 193-nm lithography for a number of reasons: First, a simple lithographic process must be available to develop 193-nm DUV lithography exposure tools. Second, industry acceptance of a new lithography wavelength is likely to be most effective when SLR processing accompanies the new lithography. Finally, 193-nm lithography will probably replace 248-nm lithography when and if the replacement is economical and provides significant improvement in chip performance (or manufacturing yield). Consequently, new 193-nm single-layer resists with high-performance imaging characteristics (resolution, depth of focus, process latitude, adhesion, sensitivity), plasma etch resistance equivalent or even superior to conventional DUV resists, and compatibility with industry-standard processing chemicals (aqueous-base developers) must be developed.

New polymer materials with high optical transparency are required for 193-nm (single-layer) resists. This optical transparency at the exposure wavelength must be combined with a set of properties possessed by hydroxystyrene polymers, which are the standard building block primarily of conventional DUV (248-nm) resists:

1. Hydrophilicity (for good positive-tone development characteristics).
2. High T (130-170°C), for good thermal properties and the latitude to perform higher post-exposure bakes.
3. Aromatic rings in high concentration (for good etch resistance).
4. An easily blocked hydroxyl group [for incorporation of acid-cleavable functionality, i.e., a functional group which is easily attached to an acidic moiety such as a carboxylic acid that is reactive and removable (deblockable) in the presence of a strong (photogenerated) acid, thereby providing a significant solubility change].

Acrylic polymers are the new basis for 193-nm resist design because of their excellent optical transparency (see Figure 2) and easily tailored structure. The characteristics, advantages, and limitations of acrylic chemically amplified resist systems are listed below:

• Transparency at 193 nm The aliphatic nature of the polymer and the low extinction coefficient of the vibronic transition ( at 210-220 nm) are characteristics of a very transparent polymer. Figure 2 shows a comparison of the absorbance properties of poly(4-hydroxystyrene) (PHOST) with poly(methyl methacrylate) (PMMA).
• Property diversity Substitution of the ester "R" group permits great flexibility in modifying polymer properties. Additionally, copolymerization can dramatically expand the property spectrum. A virtually infinite variety of polymer characteristics can emanate from this family. Fortunately, the optical transparency at 193 nm is virtually 100%, providing an excellent platform on which to build a resist.
• Ease of synthesis Acrylic polymers prepared for 193-nm single-layer resists are products of a direct synthesis, unlike traditional DUV polymers, which are made via a two-, three-, or even four-step process, adding significantly to complexity and cost of preparation. In another departure from DUV materials, acrylic polymers are synthesized without the need for acids or bases. In short, resist component manufacturability, an enormous impediment to the move to DUV (248-nm) technology, may not be nearly as problematical in 193-nm SLRs.
• Plasma etch resistance Simple acrylics have etch rates in aggressive plasma environments found in many semiconductor processes that are approximately twice those of phenolic resists. Therefore, the polymer must be substantially modified to provide suitable etch resistance.

• Development of acid-catalyzed deprotection chemistry (chemical amplification) in the early 1980s, including acid-catalyzed reactions of esters of methacrylic acid by Ito and coworkers at IBM [4].
• Development of all-acrylic, aqueous-developing positive resists employing the methacrylate terpolymer concept, by IBM in the late 1980s [5].
• Recognition by workers at Fujitsu that plasma etch resistance and 193-nm transparency can reside simultaneously in an acrylic resin through incorporation of alicyclic components [6].

Tool evaluation resist
The first high-speed, high-resolution positive single-layer resist was developed in 1993 by the IBM/MIT Lincoln Laboratory team [7]. The primary purposes of this (IBM Version 1) resist were to evaluate the imaging performance of the SVGL (Silicon Valley Group, Lithography) prototype 193-nm step-and-scan system, and to demonstrate the feasibility of a 193-nm SLR. To properly evaluate a new tool, it was of the utmost importance to design the Version 1 resist with dual-wavelength (193-, 248-nm) capability. With this design, the resist can be exposed with a 248-nm stepper with known characteristics, and questions about stepper (optics) quality and resist performance are separable.

The Version 1 resist (Figure 3) comprises two components, an iodonium triflate onium salt and a methacrylate terpolymer originally developed for printed-circuit-board lithography [5]. Each monomer serves a separate function in the terpolymer. T-butyl methacrylate (TBMA) provides an acid-cleavable side group which is responsible for creating a radiation-induced solubility change. Methyl methacrylate (MMA) promotes hydrophilicity for photoinitiator solubility and positive-tone development characteristics, while also improving adhesion and mechanical properties and minimizing shrinkage after exposure/bake processing steps. Methacrylic acid (MAA) controls aqueous development kinetics. This polymer is prepared in a single step from readily available, inexpensive components. By selecting the terpolymer composition and molecular weight, imaging properties (such as dissolution properties, photospeed, contrast) can be altered to a significant extent.

Figure 3

Version 1A (initial formulation) resist was used extensively on the SVGL Micrascan® 193 prototype. In fact, the resist was integral to the optical characterization of the tool and accelerated optimization of the prototype's optical characteristics. The highest resolution obtained at 193 nm (NA = 0.5) for Version 1A was 0.22 µm in 0.75 µm of resist. Figure 4 shows 193-nm imaging of this resist using the SVGL prototype step-and-scan system.

Figure 4

An advanced version of this resist (Version 1B) was developed in 1994. Version 1B has been used principally on the GCA (XLS) 248-nm stepper at MIT Lincoln Laboratory. The resist is linear to 0.30 µm at a resist thickness of 0.75-0.85 µm with good process latitude, and is able to resolve features down to 0.25 µm with this DUV stepper when overexposed. Figure 5 shows the resolution capability in Version 1B exposed with this 248-nm stepper at dimensions of 0.30 µm, pictured through a dose range of 5.2-6.4 mJ/cm². Recent imaging results with 193-nm exposure show a further gain in resolution to 0.14 µm.

Figure 5

It should be noted that most of the acrylic-based resists develop with base (hydroxide) concentrations (0.01-0.05N) significantly lower than are typically used with the current production concentrations (0.21-0.26N) of phenolic (248- or 365-nm) resists. This difference could affect the introduction of this class of 193-nm resists into a manufacturing environment.

Etch-resistant resists
The Version 1 resists described above have etch properties surprisingly similar to those of conventional DUV resists (e.g., IBM APEX-E) in CF-based (oxide) etch recipes. More aggressive etch chemistries (e.g., aluminum and polysilicon etching) demand substantial increases in etch resistance. For example, the etch rate of the Version 1 resist can be three times as high as that of a novolak-based resist under halogen etch conditions used widely in the semiconductor industry. The approach currently being used to impart etch resistance without degrading the optical transparency of 193-nm single-layer resists involves incorporation of alicyclic (aliphatic/cyclic) compounds into the polymer structure.

The etch resistance of acrylic resists can be substantially improved by increasing the carbon content of the polymer via the incorporation of cyclic aliphatic functionality. For example, the etch rate of poly(isobornyl methacrylate) is less than half that of poly(methyl methacrylate) in a hydrogen bromide plasma [6]. This approach is based on earlier studies which correlated the etch rate with the carbon/hydrogen ratio of a series of polymers [8]; it is currently the method most commonly used to develop etch-resistant, transparent 193-nm photoresists.

It is apparent from published resist approaches that high-resolution lithography using a 193-nm SLR is possible. It is equally likely that plasma etch resistance approaching DUV aromatic resists is feasible. Much less certain is the development of a resist formulation which possesses a combination of both imaging quality and plasma etch resistance. The lack of flexibility in a traditional resist design approach, coupled with the often competing considerations of hydrophilic-hydrophobic balance, lithographic contrast, high T, thermal stability, and etch resistance, limits the options of the 193-nm resist designer.

To provide greater flexibility in resist design, we have developed three-component resists consisting of a methacrylate polymer slightly modified from Version 1, an alicyclic dissolution inhibitor compound, and finally, a photoacid generator. After evaluation of large numbers of dissolution inhibitor compounds, it became apparent that a class of compounds (5-B steroids) was available (from natural sources) with the following very desirable properties:

• High solubility in resist and PGMEA.
• Strong dissolution inhibition.
• High exposed dissolution rate.
• 193-nm transparency.
• Moderating influence on T.
• Plasma etch resistance.
• Good thermal stability (>200°C).

Figure 6 Figure 7

193-nm wavelength effects on photoacid generator efficiencies and substrate reflectivity
In addition to the polymer transparency issues, the primary impact of the switch from 248 to 193 nm on the design of CA resists involves resist chemistry, specifically related to the photoacid generator (PAG). The effect of the change in exposure wavelength is confined to photochemical processes, since the thermal acid-catalyzed reactions occurring in CA resist are independent of exposure wavelength. Although the significantly higher photon energy at 193 nm compared to 248 nm [6.42 eV (148 kcal/mol) vs. 4.89 eV (115 kcal/mol), respectively] opens the possibility of new photochemical pathways for the polymer matrix (i.e., scission of the polymer backbone, as seen in PMMA resists), it is likely to be relatively insignificant because of the low doses employed in CA resist systems. Of greater importance will be the availability of PAGs which are efficient photoacid generators at 193 nm. Ideally, one would like to use the same PAG materials which were developed for 248-nm applications. It appears (with the exception of the PAGs whose photochemical mechanism of acid generation relies on photoinduced electron transfer sensitization with a phenolic polymer matrix) that most of the PAGs developed for 248-nm lithography will function at the 193-nm wavelength. Transparency problems can arise, however, because of the aromatic nature of most PAGs. Recently, more transparent PAGs have been specifically developed for 193-nm resists [9].

In summary, it seems that while only a subset of DUV 248-nm PAGs can be used at 193 nm, these compounds, coupled with new PAGs which are being developed, should provide a material set diverse enough for the formulation of 193-nm chemically amplified resists.

The evolution from 365 to 248 nm has presented lithography engineers with new challenges caused by increased thin-film interference effects in resists during exposure. These challenges were the result of the inherently higher reflectivity which occurs at shorter wavelengths and the development of new nonbleachable photoresists, and have fueled the development of both top- and bottom-antireflective-layer (ARL) technology for use at 248 nm. The extension of production photolithography to 193 nm will also require reflection suppression for certain lithography levels. Extensive experimental and modeling work has already been done at 248 nm to better understand requirements, limitations, and performance of ARL technology [10]. This framework will have a direct bearing on developments at 193 nm, since much of this modeling can be generically applied to any exposure wavelength. The limitation at 193 nm, however, lies in knowledge of the optical properties of the materials in use, since few commercially available instruments can measure the complex refractive index at wavelengths of less than 200 nm.

Most bottom ARLs designed for use at 248 nm are composed of strongly absorbing phenyl-containing polymeric systems; as was noted above, these materials are less than ideally suited for 193-nm exposure. These very high absorptivities lead to reflection coefficients at 193 nm that, although small (2-5%), are several times higher than reflection coefficients at 248 nm. However, prototype 193-nm ARLs have already been developed which incorporate a more transparent methacrylate functionality [11] and have refractive indices of 1.5-1.6. These ARLs better match the real part of the refractive index of the 193-nm resist systems, which are also methacrylate-based. From these early experiments, it seems that 193-nm ARL development should be considerably easier than development of the corresponding single-layer resists, and such development should be limited only by the availability of precision measurement tools at 193 nm.

Resists for 193-nm lithography: Summary and challenges

New polymers Workers at IBM, in cooperation with the University of Texas and B. F. Goodrich, have demonstrated that 193-nm resists can be developed using (nonacrylic) polymers based on the polymerization of cyclic olefins (e.g., norbornene derivatives) [12]. The imaging properties do not yet approach those of photoresists based on acrylic polymers, but plasma etch rates lower than those of novolak have been demonstrated. Workers at AT&T (Lucent Technologies) have published information on a new resist system that is effectively a cyclic olefin/acrylic hybrid with attractive resist performance [13].

New high-performance resists Workers at Fujitsu [14], NEC [15], and Toshiba [16] have recently published information on new (second-generation) acrylic resists which combine good etch resistance and clearly improved imaging characteristics. Lithographic performance data on 193-nm steppers with these new resists remain to be demonstrated.

Significant challenges face the industry in this era of early research and development of 193-nm resists, especially in the following areas:

• The need for an etch-resistant single-layer resist, compatible with 193- and 248-nm exposure tools, with resolution capability (k-factor) equivalent to the new generation of DUV resist technology.
• Single-layer resists compatible with industry-standard developers (e.g., 0.26N TMAH).
• Alternate polymer families (beyond acrylics) for 193-nm SLRs.
• Negative-tone 193-nm SLRs.
• High-performance, aqueous-developing silicon-containing resists for 193-nm bilayer systems.
• Top-surface-imaging (TSI) resists with faster photospeeds than the current resists.
• Timely development of optical exposure tool infrastructure.

Acknowledgment

Micrascan is a registered trademark of SVG Lithography Systems, Inc.

References

Received February 9, 1996; accepted for publication November 25, 1996