Plasma-assisted cleaning of extreme UV optics
Alexandre Dolgov is a PhD Student in the Industrial Focus Group XUV Optics. His supervisor is Professor Fred Bijkerk from the Faculty of Science and Technology.
The goal of this work was to characterize the plasma chemistry induced by EUV radiation in a physisorbed layer of molecular species. Model systems were developed, both experimentally and theoretically, allowing contamination scaling laws to be studied. Applied to the case of photolithography, this work focuses on revealing the relevant physics and photochemical processes for cleaning. To highlight important physical processes, contamination and cleaning mechanisms were studied as a function of the surface and plasma environment. The plasma impact on chemistry and on the cap-layer surfaces are discussed in this thesis.
Using an experimental setup that directly reproduces Extreme UV-lithography relevant conditions, a numerical model for obtaining the spatio-temporal behavior of EUV-induced plasma evolution was validated. To achieve this, a setup, equipped with Langmuir probes was used to obtain time resolved experimental data. The experimentally obtained current-voltage curves were used to validate the model. The applicability and accuracy of in situ plasma diagnostics is still a subject of discussion. When using Langmuir probes, it must be taken into account that all the quantities extracted from the current-voltage curves are spatially averaged values. The probes average over the sheath that the probe itself creates during the plasma decay. Thus, the probe current-voltage curves may be dominated by changes to the plasma that the presence of the probe induces. As a result, the plasma may be characterized inaccurately.
The actual plasma parameters corresponding to the experimental results are replicated using Particle In Cell-Monte Carlo numerical modeling. For the case of cylindrical symmetry, the model is an accurate instrument for studying plasma formation during the EUV pulse and to discover the most important processes during plasma formation and decay. The model was then used in conjunction with experiments to characterize the plasma in the volume above a multilayer mirror.
The knowledge of EUV-induced plasma formation and evolution allows plasma-induced effects and photo-induced effects to be distinguished. The best illustration of the influence of high-energy photons on cleaning mechanisms has been obtained experimentally for amorphous carbon (a-C). By comparing a-C removal in a surface wave discharge (SWD) plasma and an EUV-induced plasma, the physics of cleaning for photoactive hydrogen and noble helium environments were determined. For helium gas, the results for both plasmas were similar, while for hydrogen, an EUV plasma was found to create a larger ratio of atoms to ions at the sample surface. In the EUV-induced hydrogen plasma, the C-atom yield per ion is higher than in the SWD plasma. The presence of high-energy photons leads to more efficient plasma cleaning due to surface activation processes. We refer to this combination of processes as reactive ion sputtering (RIS).
However, in the presence of high-energy photons and ions, contamination is a function of the surface environment. It has been shown that, in the case of a grazing incidence collector for extreme ultraviolet (EUV) lithography, the molecular contamination layer may have physical properties similar those of a diamond-like carbon (DLC) film. Such a DLC contamination layer may require ten times more H3+ ions to remove one C atom during hydrogen plasma treatment. This highlights the importance of understanding the growth of contamination in order to optimize cleaning processes.
Low-temperature hydrogen plasmas can also be effective for reducing oxidized ruthenium top surfaces of a MLM. And that, for water-induced oxidation, reduction and oxidation can be balanced by controlling the ratio of the water and hydrogen partial pressures. Our experiment shows that this balance follows a linear relationship, indicating that the chemistry is relatively simple. Follow up experiments on more deeply oxidized Ru samples showed that reduction of sub-surface oxide is very slow, and may not be reversible using EUV-induced plasma reduction.
Our proof-of-concept experiments demonstrate that an in situ, EUV generated plasma cleaning technology is feasible. The EUV pulse generates a low-temperature plasma, along with photo-induced surface activation. Together these combine to yield a highly reactive environment that quickly and efficiently removes amorphous carbon and reduces ruthenium oxide. We show that by controlling the background gas partial pressure, and by biasing the mirror, the plasma environment and cleaning rates can be controlled with a relatively high level of precision.
Our studies have largely fulfilled the industrially relevant goal of assessing EUV-induced cleaning methods, as listed in Chapter 1:
- EUV-induced plasma cleaning can be performed without interrupting wafer exposure.
- EUV-induced plasma cleaning can be performed at high speed: 3nm/h for carbon and 0.1 nm/h for oxygen at ion energies of ~100 eV.
- EUV-induced plasma cleaning is highly selective.
where = 0.45 is the carbon removal rate, and is the Ru removal rate, which is less than 10-4 at energies over 100 eV.
In conclusion, the studies in this thesis provide better understanding of the mechanism of cleaning (reduction) methods that use EUV radiation generated plasma. The measurement results showed that carbon can be etched, and ruthenium reduced, with high efficiency and high selectivity using hydrogen ions. In performing these experiments, we also showed that numerical modeling, combined with Langmuir probe measurements, can provide accurate spatio-temporal dependent estimates of the plasma characteristics. These can then be used to further refine cleaning methods and to guide the choice of plasma-facing mirror surfaces.
