Chemically amplified photoresists are out of steam where EUV is concerned.
As EUV exposure tools, sources, and photomasks have become more capable, the lithography sector’s attention has turned to EUV photoresist. After all, once the exposure system can produce a high quality image at the wafer, the resist still has to capture it for pattern transfer. Unfortunately, the increasing emphasis on photoresist has made the limitations of current formulations even more obvious.
Ever since the introduction of 248nm exposure wavelengths, lithographers have depended on chemically amplified (CA) resists. In CA resists, incoming light does not directly cause the backbone polymer to become soluble. Rather, the light is absorbed by photoacid generators (PAGs), which release multiple photoacid molecules per photon. The photoacid reacts with protective groups on the polymer backbone, “de-protecting” it and causing it to become soluble in developer. In EUV resists, the incoming light can also excite photoelectrons, which in turn can drive the photoacid generation reaction.
Thus, at the very heart of chemically amplified resist chemistry we find a tradeoff between resolution and sensitivity. To maximize resolution, resist designers would like to have small backbone molecules, and for the photoacid de-protection reaction to take place close to the location of the PAG. Both the photoacid diffusion distance and the amount of resist de-protected by each photoacid molecule should be small.
To maximize sensitivity, on the other hand, designers look for nearly opposite characteristics. The photon capture cross-section should be large, to make use of as many photons as are available. Each photon should generate many photoacid molecules, which should rapidly diffuse into the resist layer to de-protect as many polymer chains as possible. Unfortunately, the large diffusion distance this behavior implies contributes to blur, the uncertainty in the dimensions of an exposed feature.
The implications of this tradeoff for EUV resists in particular are quite troubling. Proposed manufacturing insertion points for EUV are at the 7nm node or below, with critical dimensions of 40nm or less for some layers. Yet, as previously discussed, EUV light sources are struggling to achieve reasonable power levels. With very few photons available, the EUV process needs highly sensitive photoresists to achieve acceptable throughput.
While designers of 193nm photoresists also face a resolution/sensitivity tradeoff, the short wavelength and high energy of EUV photons pose additional challenges. Organic molecules have poor EUV absorbance; a thick photoresist is desirable to capture as much of the available light as possible. Thick resist layers also can tolerate longer, more aggressive etch processes. Unfortunately, high aspect ratio features are prone to pattern collapse.
High-energy photons also exacerbate the problem of shot noise. Each part of the image is exposed to a very small number of photons, which will be distributed randomly along the intensity profile. The resulting noise is a major contributor to line edge roughness. As line edge roughness becomes large relative to feature width, it limits the effective resolution that can be achieved. (A future article will look at EUV and line edge roughness in more detail.)
In fact, overall resist performance is increasingly defined by randomness: the random distribution of polymer chain lengths, the random distribution of PAG molecules in the resist, and the random distribution of incoming photons combine to increase the uncertainty in the resist image.
Inpria is one of several companies looking for an alternative to chemically amplified resists. Their approach is based on tin-oxide nanoclusters, which are smaller than polymer chains but have a high EUV cross-section. According to Inpria CEO Andrew Grenville, these clusters absorb four or five times as much EUV radiation as CA resists at a constant dose. The high absorbance should make it possible to increase sensitivity without undermining resolution. Moreover, the resists do not depend on a secondary de-protection reaction. Rather, the EUV light and any secondary electrons it generates can directly break bonds, inducing differential solubility within the film. The materials deliver both superior etch resistance, Grenville said, and up to 40:1 etch contrast with carbon-doped oxide dielectrics. Such a high contrast greatly simplifies the process stack, reducing the need for intermediate hard mask and pattern transfer layers.
At this point, Grenville said, the company can print 13nm line and space patterns, with 35 mJ/cm2 sensitivity. These results are comparable to the best CA resists, while the Inpria approach has achieved superior line edge roughness. In results printed at this year’s SPIE Advanced Lithography conference, the company demonstrated successful integration with standard wafer processing equipment, with no metal transfer between wafers. Commercialization efforts are proceeding on multiple fronts, as the company is both seeking manufacturing partners and working to improve their in-house manufacturing processes.
CA resists have served the industry well for more than 20 years, but may be reaching their limits in the EUV era. If the alternatives to CA resists are as successful, the industry will be in good hands.