News, views and contacts from the global Semiconductor industry

Immersion Imaging Under The Microscope

1 September 2004

The potential for immersion lithography to offer superior imaging is beyond doubt. The question is whether the benefits of extending its capabilities outweigh the problems. Burn J Lin, senior director of Taiwan Semiconductor Manufacturing Company Limited, outlines some of the issues that demand attention.

Immersion lithography extends the capabilities of scanner technology, similar to wavelength reduction. Using water, which has a refractive index of 1.44, image-coupling wavelength is reduced from 193nm to 134nm. Unlike other wavelength reduction schemes, such as 157nm and EUV lithography, light frequency remains unchanged. Drastic material changes, such as mask blank, pellicle, lens or photoresist, are unnecessary, and the light source does not need to be changed for each new wavelength.

In dry lithography, the Numerical Aperture (NA) of the space between the lens and the resist cannot exceed one. This is because the aperture angle u is restricted to <π/2 so that light can pass from the lens to the resist in a forward direction. When there is an immersion fluid in that space, the same angle can sustain a numerical aperture of n, the refractive index value, with the same restriction to u. In short, the immersion fluid extends the upper limit of 'coupling' (the flow of light from lens to resist) for higher spatial frequencies.

Since resolution W = k1l/NA, increasing the NA improves resolution. However, Depth Of Focus DOF = k3l/sin2(u/2). If NA is kept constant, a higher index-coupling medium reduces u to increase DOF. Hence, resolution and DOF can be traded off each other through the selection of the upper limit of the NA.


"Immersion fluid extends the upper limit of 'coupling' (the flow of light from lens to resist) for higher spatial frequencies."

The principles of immersion lithography are similar to those of immersion microscopy. Yet demonstrating this type of imaging using an immersion microscope will hardly convince the lithography community that it is a viable technology for the manufacturing environment. This is particularly obvious when one considers that the technique has to be adapted to a scanner that requires the wafer to be moved quickly in relation to the lens and the image CD controlled carefully.

There are many manufacturing-related issues that require attention. First, the high-index fluid must have sufficient transmission capability. Its viscosity must allow high-speed scanning. It has to be compatible with the photoresist – that is, it cannot affect the photoresist's imaging properties. Also, the fluid must not contaminate the photo-resist, the lens or anything else it comes into contact with.

The refractive index range of the fluid must be kept within narrow limits so it does not affect the image position, resolution or CD control. This can mean limiting the thickness of the coupling medium to 1~3mm and controlling the fluid temperature to 0.01˚C. During exposure, bubbles must not be allowed to reach a size and density that can affect the imaging quality. The flow of the fluid must not induce particulates. Fluid-handling time must be an insignificant part of the wafer throughput budget.


"Extending the bounds of immersion lithography has become something of a preoccupation."

Major suppliers have announced roadmaps for 193nm immersion scanners covering NA = 0.85 to >1.2. Some scanners have already been delivered. Meanwhile, resist and track suppliers have developed immersion-compatible systems, and full-chip device fabrication is ongoing.

Simulations that realistically incorporate polarisation-dependent stray light have proven the feasibility of 32nm logic node using 45nm half pitch. Some companies are developing immersion processes for the 65nm logic node, but most plan to use it for the 45nm logic node.


Extending the bounds of immersion lithography has become something of a preoccupation. Among the most popular research paths is development of a fluid with a refractive index in the 1.5–1.6 range. The principal driver for this is the prospect of a threshold of spatial frequency coupling that will allow better resolution and/or DOF.

Though this task seems easier than that of developing 157nm or EUV lithography, producing a fluid that has a higher index and characteristics similar to water is no simple matter. But the potential benefits of success and the relatively low research cost are motivating many large and small organisations to pursue this line of development.


dependent loss of contrast is not unique to immersion lithography."

Light in the TM mode loses contrast when forming high-spatial frequency images in the resist. A larger u leads to even worse levels of contrast. Two methods can be used to improve the situation. First, light can be polarised for the TE mode. Second, the refractive index of the photoresist can be increased to reduce u in the resist.

Passively filtering unwanted polarisation is easy, but too inefficient with regards to exposure energy expended. Developing an illuminator and an imaging lens that manipulate the polarisation is complicated and expensive, but it is feasible and will improve the contrast of the resist image. The improvement in DOF is less noticeable.

A further complication is correlating polarisation with pattern orientation at different field locations for optical proximity correction.


Polarisation-dependent loss of contrast is not unique to immersion lithography. Nevertheless, it will become a more pressing issue as the technology continues to increase the resolution in the resist beyond that which a dry system can support.

The success of immersion lithography will also push feature sizes deeper into the sub-wavelength range. Eventually, even the feature size at a 4X mask will be markedly sub-wavelength. A 32nm feature, for example, is 66% of the 193nm wavelength. An 8nm sub-resolution assist feature is 17% of 193nm. Such differences can render imaging even more non-linear and make polarisation in the image even more difficult to control. One solution is to fill and planarise the absorber side of the mask with a high-index material. If the index is 1.7, a 32nm feature will be 1.13l, and the 8nm assist feature will be 0.28l.


"The success of immersion lithography will push feature sizes deeper into the sub-wavelength range."

Will it be feasible to increase light frequency in immersion lithography further – can the vacuum wavelength be reduced enough to allow the use of immersion lithography? For example, in 157nm immersion lithography, several immersion fluids have been identified, but not one is suitable for scanner imaging. They all tend to absorb too much light, thus requiring unreasonably thin coupling thicknesses. They are also viscous, messy to handle and difficult to clean.

Considering the insurmountable difficulties experienced with dry 157nm systems, the lithography community generally agrees that developing a high-index fluid for 193nm (rather than for 157nm) is a much better option.

Interestingly, 248nm immersion is a possibility. The principal motivation is cost reduction. The cost of tools, materials and laser pulses at this level is much lower than for 193nm. There is, however, a conflict of interest between users and suppliers. The latter do not see any significant profit arising from the development of 248nm immersion lithography, while the former are urgently seeking to adopt less expensive tools and processes. Which side will eventually win out has yet to become clear.