Non-lensing defects and defect reduction for 193i

The water used in immersion lithography can cause defects including water marks, particles, and microbridges,^1–3 as well as the bubbles and anti-bubbles discussed in the last article. These five defects are important because they are found on almost all wafers processed by 193nm immersion (193i) lithography. Both simulations and experimental cross-sections have helped to explain how these defects form. Based on our understanding of these formation mechanisms, we propose a series of measures to reduce such defects.

Although water marks are circular like bubble defects, they are generated in a completely different way. As the exposure head moves across the wafer from die to die, water droplets can be left behind. Droplets change the resist sensitivity by leaching or water uptake. The photo-acids generated in the exposed area (as well as other resist components) can leach into the water droplet, as sketched in Figure 1(a). Water can also permeate into the resist film. If the water droplet is carried over to the post-exposure bake (PEB), the elevated temperature further enhances the local physical and chemical reactions.

Figure 1.(a) Water droplet left behind on the resist surface after immersion exposure changes the local resist sensitivity. (b) Scanning electron microscope (SEM) image of a water mark defect: the extent of the defect is shown by the dotted-line circle. The pitch of the lines is 220nm.

The assumption that a water droplet reduces the local resist sensitivity was confirmed by experimental results reported by Kenji Shitabatake's group.⁴ A drop of deionized water (DI) water was manually dispensed on the resist stack after exposure. After the water droplet dried, the wafer was sent to PEB and development. They measured a resist dissolution rate in the water droplet region that was lower than the rate where no droplet existed.

Figure 1(b) shows a scanning electron microscope (SEM) image of a typical water mark defect. The outline is circular. In the center, the resist lines are fat and almost connected together. Away from the defect center, the resist lines narrow until finally they are the same thickness as in areas not affected by the defect. Although the resist lines are fat in the defect area, the pitch remains the same, as outlined by the straight dotted lines in Figure 1(b). In contrast to bubble defects, no pattern magnification or fine fringes were observed with water marks. This was expected, since the latter are caused by chemical interaction of water droplets and the resist film. No lens effect is involved here.

We used a focused ion beam (FIB) technique to make and inspect a cross-section of one of these defects. We used a dual FIB/SEM instrument equipped with a gas injector system. First, a platinum film about 100nm thick was coated on the patterned resist film in the defect area. The coating method was ion-beam-induced deposition from a local gas supply of an organo-metallic vapor. Then, a finely-focused ion beam milled away a precise amount of material from the defect area. The sidewall exposed by this process is essentially a cross-section of the defect. After the ion-beam milling, we switched the tool to SEM mode and took high-resolution images.

Figure 2. (a) Top-down image of a water-mark defect. The pitch of the lines is 220nm. The dotted line shows where a cross-section was cut. (b) Cross-section of the water mark. Resist lines are dark, a platinum layer coated on the area appears light.

Figure 2 shows the cross-section of a water mark. For comparison, a top-down image of the defect is also included. The platinum film coated on the pattern gives the bright features in Figure 2(b). It covers the resist lines and fills the space between the lines as well. Resist lines show up dark in the image. `T-topping’ (the formation of insoluable heads) on the resist lines is observed in the defect area (as framed by the dotted box). In the center of the defect, the T-topping is so heavy that resist lines contact each other and form cavities, which prevents the platinum from coating the spaces between lines. This cross-section confirms that the water mark defects are actually formed by T-topping, due to the resist sensitivity loss at the surface.

The use of a resist or topcoat that causes a high receding contact angle in the water below the exposure head reduces the probability of leaving water droplets on the wafer. The critical value of the receding contact angle — in other words, the angle that ensures no droplet is left behind — depends on the scanning speed and exposure-head design. The defect density versus the receding contact angle has been measured for various resists and topcoats.^5,6 Based on experimental data, it has been widely accepted that the receding contact angle near 70° is needed to avoid leaving behind water droplets at full scanning speed.

The relationship between water mark defects and the surface receding contact angle is extremely non-linear. Near the critical receding angle, the number of defects drop sharply with angle increase.³ For example, increasing the receding angle from 55° to 70° reduced defect counts from 250 to 50.⁵

Rinsing the wafer with DI water immediately after exposure (post-rinse) reduces water mark defects. The post-rinse gets rid of water droplets as soon as possible. If the droplet is left on the surface too long, however, the post rinse becomes useless. The effectiveness of the rinse processes strongly depends on the resist stack. For some resists, the rinse can reduce the number of defects by up to 60%, while for others it provides a less than 5% reduction. This material dependency has been attributed to water uptake and is still under investigation.

Particles and clusters are another common type of defect, although they are not unique to immersion lithography: they are also observed on 193nm dry-processed wafers. However, immersion does introduce additional particle sources.

The immersion water inevitably contains particles and impurities. Particles in the immersion water can deposit on the wafer surface. Unwanted chemical components in the water can also aggregate on the wafer surface, forming particles. One milliliter of water that has two parts per trillion of iron—the detection limit for Fe—contains 2.2×10¹⁰ Fe atoms. As the water dries, these atoms can form a 640×640×640nm particle or, more likely, dozens of particles in the 100–200nm range.⁷

Another particle source is the wafer stage. A wafer is loaded on the stage, and the exposure head moves across it during alignment, taking the water meniscus with it. Particles on the wafer stage can be picked up by the water and brought onto the wafer surface, as sketched in Figure 3(a). These particles come from the previously processed wafers.

The force induced by the movement of the water meniscus can pick up any loose flakes or particles not only from the stage but also from the wafer's edge. The thickness of the resist stack changes from several hundred nanometers to zero at the edge. Particles and loose flakes are easily generated in this region. For example, some resists or topcoats do not adhere well to bare Si. If the resist film extends beyond the edge of the bottom antireflective coating (BARC) film, then it can easily form loose flakes: see Figure 3(b). When the exposure head moves across the wafer's edge, the loose flakes can be picked up by the water meniscus and transported to the wafer's center. Not only the resist and topcoat but also poorly-adhering process films such as oxide, nitride, low-k, and similar films could peel off at the wafer's edge. Peeled flakes or particles can also be transported to the wafer stage and cross-contaminate the next wafer.

Figure 3. (a) Particles on the wafer stage are picked up by the water meniscus. (b) A resist flake forms due to bad adhesion between the resist and bare Si.

Table 1. Formation mechanism of immersion-related defects and reduction strategies. Click to enlarge.

To reduce the number of particles on the wafer, ultra-pure DI water is needed for the immersion process. Standard clean room DI water — with resistivity >18.22 MΩ-cm — needs to be further treated before it can be injected into the scanner. A new specification for ultra-pure water has been suggested as a result.⁷

To further reduce particle/cluster defects, adhesion of the resist or topcoat on the substrate has to be improved. Usually, adhesion between BARC and bare Si is less problematic. For integrated wafers, the adhesion of BARC to the film stack on the wafer can be weak. Special treatment may be needed to enhance the adhesion. An edge bead removal (EBR) process removes loose flakes and keeps the edge region clean and robust to the water meniscus. Optimization of the EBR process is tedious but very important for the reduction of defects. The best process can only be obtained by numerous tests, and depends on the materials used. Wafer edge exposure (WEE) can also be employed, sometimes, to shape the resist film edge and, again, mitigate the problem.

After long use in production volumes, a wafer stage inevitably gathers particles and dry stains. Routinely cleaning the wafer stage reduces the possibility of these particles being picked up by a water meniscus and cross-contaminating the wafers.

Microbridges, as the name suggests, are adjacent resist lines that are connected by bridges that are (usually) smaller than about 500nm. Figure 4 shows two typical examples. This type of defect was first observed in 193nm dry-lithography-processed wafers, and was attributed to the BARC process⁸ and small opaque particles that block the exposure light. Immersion processed wafers have more microbridge defects than their dry-processed counterparts. This indicates that the 193nnm immersion process has additional sources of microbridge defects. For example, simulation shows that a bubble of ∼80nm diameter, located in the space area of a 100nm 1:1 dense pattern, can completely block the space between lines.⁹ Particles introduced by the immersion water can locally block the exposure light, and cause small under-exposed spots. An intermixing layer, which exists at the interface of the resist and topcoat, may be blamed for the microbridge defects.

Figure 4. SEM image of microbridge defects. The pitch of the line and space pattern is 220nm.

Bubble and anti-bubble defects, water marks, particles, and microbridges have been identified as 193nm immersion-related defects. Bubble defects and water marks are unique to the immersion. Anti-bubble defects, particles, and microbridges are also observed on dry processed wafers, but the immersion process provides additional sources of problems. The formation mechanism of immersion defects has been shown to involve scanner, material (BARC/resist/topcoat), and process setup: bubble defects and particles are introduced by the scanner; water marks, particles, and microbridges are introduced by materials; and particles from the wafer edge and anti-bubble defects can both be introduced by the process. To effectively reduce these immersion-related defects, different strategies have been proposed, which are summarized in Table 1. These strategies evolved from understanding defect formation mechanisms and have proven to be effective. A defect level of only about 50 defects per 300mm wafer has been demonstrated by an optimized 193nm immersion process, which was developed as a joint effort between the scanner supplier, the resist supplier, and integrated circuit manufacturers.

The author thanks Dr. Stefan Brandl of Qimonda AG for his collaboration. The FIB cross-section image in Figure 2 is his contribution.