As introduced in the first part of this series, 193nm immersion lithography has three resist-process approaches: resist with a solvent-soluble topcoat, resist with a developer-soluble topcoat, and resist without a topcoat. The mainstream for current processes appears to be the second of these. The topcoat is coated on the resist film and serves as a barrier layer to prevent leaching and water uptake. To be successful, it must fulfill numerous requirements and different experiments have been designed to evaluate the topcoat performance. First, the topcoat and resist stack is tested for leaching to ensure that the topcoat can effectively prevent resist components from leaching into water (see the second part of the series). Next, the topcoat has to be physically and chemically compatible with the resist film underneath. Different combinations of topcoats and resists have been exposed and the process window, line-edge roughness (LER), and pattern profile are measured to identify the most compelling combination. Finally, the process parameters of the best combination of topcoat and resist need to be optimized. Specifically, the bake temperatures of the resist and topcoat must be aligned to achieve the best performance.Physical and chemical compatibility of topcoat and resist
During exposure, the topcoat and resist stack is immersed in water. Light can reflect from the interface of water and topcoat as well as from the interface of topcoat and resist. The refractive index of the topcoat (ntopcoat) must match that of the water (nwater) and resist (nresist) to interfere destructively, thus reducing overall reflection:
Typical 193nm resist has a refractive index of 1.7, and water has a refractive index of 1.44 at a wavelength of 193nm. According to Equation (1), the ideal refractive index of topcoat for water immersion is ∼1.55.
Figure 1. Scanning-electron micrograph of a blob defect. It consists of many small particles that surround a big particle in a circular shape, like satellites.
The thickness of the topcoat is another important parameter that affects its lithography performance. This has to be optimized to ensure that reflections from interfaces interfere destructively. For the light with an incident angle of zero, i.e., corresponding to the small numerical aperture (NA) situation, the topcoat thickness should be around:
For light with a high incident angle or high NA situation, the optimal thickness corresponding to the first reflective minimum needs to be calculated using a simulation tool. In addition to optical performance, a thicker topcoat better prevents leaching and water penetration. Depending on the acceptable leaching level, a target thickness is defined. The recommended thickness for most developer-soluble topcoats falls in the range of 30–90nm.
The solvent for the topcoat should not react with the resist. Otherwise, the solvent could partially dissolve the resist surface in the topcoat coating process, and form an intermixing layer. Most 193nm resists use propylene glycol methyl ether acetate (PGMEA) and propylene glycol monomethyl ether (PGME) as solvents, and topcoats have an alcohol-based solvent system. Solvent incompatibility may occur when the resist and topcoat are supplied by different vendors. To plumb a new topcoat, special attention has to be paid to the solvent cleaning procedures for the line.
The developer-soluble topcoat is removed by aqueous tetra-methyl ammonium hydroxide (TMAH) in the develop module. A high dissolution rate in the developer is essential for the effective removal of the topcoats. A high dissolution rate also reduces the ‘blob defect’ counts. Blob defects, also known as ‘satellite spot defects’, are mostly found in low pattern density areas. They tend to have diameters of 10–50μm (see Figure 1). Blob defects are typically composed of topcoat materials re-deposited on the surface during the development and rinse steps. The mechanism of how the blob defects are formed is complex and still under investigation. It has been attributed to non-uniformity of developer as well as the abrupt pH change by the de-ionized water rinse. The dissolution rate of most topcoats is in the range from 100–1000nm/s.
As explained in a detail in the second section, the water meniscus of the 193i exposure head moves across the topcoat surface during exposure. The smaller the receding contact angle, the more likely it is that water droplets will be left behind. If the receding contact angle is zero, a water film will be left behind—so called ‘film pulling.’1 The water left behind will then form circular water marks. To reduce or eliminate these a large receding contact angle is needed. This can be realized by increasing the hydrophobicity of the topcoat surface. However, this requirement seems contradictory to the requirement of reducing blob defects. It has been observed that the hydrophobic topcoat tends to have a low dissolution rate in the developer. A careful optimization has to be implemented here. The formation mechanism and detailed reduction strategies will be discussed in a later article in this series.
The solvent-soluble topcoat approach has the advantage that, after exposure, it is removed by a solvent and not aqueous TMAH developer. Without considering the dissolution rate in the aqueous TMAH developer, the solvent-soluble topcoat can be designed to be very hydrophobic. The water receding contact angle can be above 100°. It has been demonstrated that with the solvent-soluble topcoat, the immersion defect counts can be reduced2 to 0.1/cm2.Lithographic assessment of topcoat and resist combinations
So far, there is no universal topcoat that can work well with all 193nm resists, although it is a very attractive goal for topcoat suppliers. Different developer-soluble topcoats are on the market and numerous combinations of topcoats and resists have been tested and evaluated.3 By comparing the overall lithographic performance, a well-performing resist/topcoat stack can be identified.
Figure 2(a) shows top-down images of dense features with linewidths of 90nm and 90nm space widths between them (hereafter described as 90nm/90nm), made in Resist A with different topcoats (TC1, TC2, and TC3). The images were taken at the best dose and focus. For easy comparison, all pictures have the same size: 1μm by 1μm. A strongly incompatible situation was observed with the combination of Resist A and TC2, in which resist lines are broken and distorted. An LER (3σ) of the resist lines with TC1 is ∼8.7nm, and with TC3 is ∼8.0nm. These results indicate that topcoats do have some influence on the LER. The process windows are plotted in Figure 2(b) with a critical dimension (CD) target of 90nm and tolerance of ±10%. No process window is obtained with TC2, and the biggest process window is obtained with TC1. Cross-sections give much better insight into the profile of the resist lines as well as severity of the resist film loss. The cross-sectional images of the resist and topcoat combinations at best doses and focus (corresponding to the top-down images) are also presented in Figure 2(a). The incompatible combination, Resist A/TC2, shows significant resist loss, with the height of the resist pattern less than 40% of the nominal resist thickness. Top-rounding was observed both with TC1 and TC3, which suggests that topcoats may cause the resist loss.
Figure 2. (a) Top-down images of 90nm/90nm dense features taken from different resist/topcoat stacks. The field of view (FOV) is 1μm by 1μm. The corresponding x-sections are also included. (b) Process window of resist and topcoat combinations.
To further investigate resist film loss, contrast curves were measured from stacks of Resist A/ TC1, TC3, and TC4 (TC2 was skipped because of the poor compatibility with resist A and TC4 was added). The resist film's thickness before exposure is 200nm. Figure 3(a) shows the measured contrast curves. With the dose increasing to E0, the resist with TC4 loses the most thickness. (E0 is the minimum dose value needed to completely expose the resist film.) TC1 causes the least resist loss. Figure 3(a) also reveals that the resist dark loss with all three topcoats—i.e., the change in the resist thickness after the removal of the topcoat without exposure—is negligible. This suggests that the topcoat-induced resist loss may be due to a photo-reaction. The stacks of Resist A and TC1, TC3, and TC4 were exposed at the same dose and focus for 90nm/90nm dense patterns. Cross-sectional images were taken from the resist patterns with different topcoats—see Figure 3(b). About 50nm of resist loss is measured from the Resist A/TC4 stack, while with TC1 the resist loss is only about 30nm. By comparing the overall lithographic performance described previously (process windows, LER, profiles, resist loss etc), the Resist A/ TC1 stack is identified as the optimal combination among the tested variations.
Figure 3. (a) Contrast curves measured from stacks of Resist A with three different topcoats. (b) Cross-sectional images from 90nm/90nm resist stacks with different topcoats. The height of the resist patterns is labeled in the pictures.
The resist/topcoat stack has to go through three baking steps: the post-apply bake (PAB) of resist; the PAB of topcoat; and the post-exposure bake (PEB). The three bake steps have specific effects. The PAB of the resist and the topcoat remove most of the solvents. The evaporation of the resist solvent also effects the photo-acid generator (PAG) distribution in the resist. The PEB affects de-protection, diffusion of photo-acid, and the dissolution rate of the topcoat in the developer. The bake temperatures need to be optimized in order to get the best performance.
An experimental design (called ‘design of experiments’ or DOE) was carried out with the Resist A/TC1 stack. All other parameters were kept the same, only the bake temperatures changed. In additional to the temperatures recommended by vendors, the DOE matrix includes 5°C above and below the resist PAB and PEB temperatures as well as 10°C above and below the topcoat PAB temperature. Contrast curves were measured at different bake temperatures. Figure 4(a) shows the results. Strong influence of temperatures on the contrast curve is observed, especially in the shoulder regime. In the shoulder regime, the temperature combination of 105°C/80°C/110°C (PAB temperature of resist/PAB temperature of topcoat/PEB temperature) gives the smallest resist loss. Figure 4(b) shows the cross-section images of the resist pattern baked at temperatures of 105°C/80°C/110°C and vendor recommended temperatures (the process of record or POR was 110°C/90°C/115°C). The exposure conditions (illumination setting, NA, mask, dose, and focus) are the same for the cross-sectioned wafers. Apparently, the resist pattern processed with the POR bake temperatures has additional resist loss of ∼10nm compared to that of 105°C/80°C/110°C.
Figure 4. (a) Contrast curves measured at different bake temperatures. (b) Cross-section images of resist patterns processed at bake temperatures of 105°C/80°C/110°C and process-of-record (POR) temperatures (110°C/90°C/115°C).
The optimized PAB temperatures of both the resist and topcoat are lower than that of POR, and the PEB temperature doesn't need to be changed. Decreasing the PAB temperatures has to do with the PAG distribution in the resist. In the resist PAB step, thermal energy drives the resist solvent to the surface where it evaporates. Part of the PAG in the bulk of the resist is also dragged to the resist surface by the solvent flow. In the topcoat PAB step, the resist solvent evaporation and PAG movement to the resist surface happen again because the topcoat PAB also bakes the resist. The POR resist PAB temperature was recommended by vendor. It was optimized for a 193nm ‘dry’ lithographic process, which uses a single layer resist on a bottom anti-reflection coating. However, in the immersion process, the resist PAB temperature has to be decreased in order to accommodate the topcoat bake. With POR temperatures the resist is actually over-baked by the topcoat PAB, and more PAG is moved to the resist and topcoat interface. The high PAG concentration causes extra resist loss after exposure and development, as we observed. In principle, the resist PAB temperature can be the same as the ‘dry’ litho process if the PAB of the topcoat could be omitted. However, this is not preferred because barrier performance of the topcoat degrades with decrease of the bake temperature.193i resist process without topcoat
Resist processes without top protection coatings are the favored solution for introducing 193 immersion lithography into mass production. These simplified processes go without separate coating and baking steps for the topcoat material and thereby promise reduced cost of ownership and fewer sources of defects. However, the challenge is to simultaneously provide both minimal leaching in water and superior overall lithographic performance. Innovative resists are the key to ensure the success of this approach. Resist suppliers are actively pursuing this goal.
Various 193i-dedicated resists have been evaluated and compared with the topcoat process.4 The process-window of the 90nm dense features obtained with the binary mask is plotted in Figure 5. With the line-CD target of 90nm and a CD-tolerance of ±10%, the maximal DOF is 1.0μm and the maximal exposure latitude is 8%. The process window of our baseline process is also plotted in Figure 5 for comparison. This baseline process was selected out of numerous combinations of dry resists and developer-soluble topcoats and still shows superior performance. Compared to the baseline process, the process window of current 193i resists without a topcoat is about 15-20% smaller, indicating the dilemma for resist vendors to select resist components that simultaneously meet requirements of high lithographic performance and low leaching behavior when in direct contact with water.
Figure 5. Process-window plot of 90nm-dense features, made using a binary image mask, with a target critical dimension (CD) of 90nm and a CD tolerance of ±10%. DOF: depth of focus. EL: exposure latitude.
Without any topcoat, amine contamination is a concern for chemical amplified resists. PEB delay was evaluated to test the resist line CD change vs. the PEB delay time. The wafers were exposed with fixed dose and focus. After exposure the wafers were stored in the track for 1min, 3min, and 10min delay times. After this delay, the wafers were sent to PEB and development. The CDs of 90nm/90nm resist patterns were measured across the wafer: both the top and bottom of the resist lines to better monitor the resist profile change. The measured CDs were averaged all across the wafer, and plotted in Figure 6 as a function of PEB delay time.
Figure 6. Resist line CD versus the post-exposure bake (PEB) delay time
Neither top and bottom CDs in Figure 6 changed with PEB delay times as long as 10min. This means that the resist profile stays the same for PEB delays up to 10min inside the track. We also made cross-sections of the wafers with the different PEB delay time. The profile of the resist patterns does not change with a PEB delay of up to 10min, which indicates that the resist process is very robust to PEB delay. A possible explanation is that the track and scanner are linked, and the amine level inside the Lithius track is ∼0.5ppb.
Although resist processes without topcoats have potential and are preferred solutions for 193i lithography, the performance of current samples is still behind the topcoat process. Developer-soluble topcoats are currently mainstream, but one has to pay attention to compatibility issues with resists.
Yayi Wei, David Back
Qimonda North America Corp.
Yayi Wei was a staff engineer of Qimonda North America Corp. when this article was written, but is now at AZ Electronic Materials and can be contacted at Yayi.Wei@az-em.com. He is a member of SPIE working on advanced lithographic process development and photoresist evaluation.
1. C. Van Pesk, A. Grenville, R. Engelstad, G. Nellis, T. Shedd, S. Schuetter, K. Doxtator, S.-H. Lin, D. C. Owe-Yang, Film pulling and meniscus instability as a cause of residual fluid droplet, Int'l Symp. Immersion Lithography, Brugge, Belgium, 2005.
3. Yayi Wei, K. Petrillo, S. Brandl, F. Goodwin, P. Benson, R. Housley, U. Okoroanyanwu, Selection and evaluation of developer-Soluble topcoat for 193nm immersion lithography, Proc. SPIE 6153, pp. 615306, 2006.
Read the other articles in the series: