The etching of Si(100) is studied quantitatively in a molecular beam setup. After exposing the silicon surface to XeF2 doses between 102 and 104 monolayers (MLs) of XeF2, thermal desorption spectroscopy is used to study the SiFx content of the reaction layer. Large values of the fluorine content (up to 30 ML) are observed. These are explained by surface roughening due to the etching process, which increases the effective surface by up to a factor of 6 after 105 ML of XeF2, corresponding to 104 ML of etched Si. This implies a porous or canyon-like structure. The dose dependence of the fluorine content is not single exponential; it increases rapidly for the first 102 ML and much more slowly thereafter, saturating after approximately 104 ML. This behaviour is described by the rapid formation of a monolayer of SiFx species in the initial regime, followed by the slow formation of a steady state multilayer of SiF–SiF2–SiF3 and SiF2–SiF3 chains. The etch rate shows a similar dose dependence, increasing from 12% of the incident flux in the monolayer regime to 20% in the multilayer regime. Finally, it is shown that fluorine diffusion into the silicon bulk is not an important process in the reaction layer formation.

The temperature dependence of the Si͑100͒/XeF 2 etch reaction is studied quantitatively in a molecular beam setup. At a sample temperature of 150 K the reaction probability reaches unity initially, after which the XeF 2 condenses on the surface and blocks the etching process. For increasing temperatures the XeF 2 reaction probability initially decreases from 100% at 150 K down to 20% around 400 K, but for temperatures above 600 K it increases again up to 45% at 900 K. In a simple reaction scheme the high etch rate at low temperatures is explained by a XeF 2 -precursor, with an activation energy for desorption of 32Ϯ4 meV. Furthermore the increased etch rate at high temperatures is explained by the desorption of SiF 2 with an activation energy of 260Ϯ30 meV. The steady-state fluorine content of the SiF x reaction layer, measured using thermal desorption spectroscopy, reaches a maximum of 5.5 monolayers at 300 K. For increasing temperatures it decreases to a submonolayer coverage above 700 K. The temperature dependence of the formation of the reaction layer is described well by including the XeF 2 -precursor in a previously developed adsorption model.

No abstract

No abstract

Articles you may be interested inCharacterization of out-of-band radiation and plasma parameters in laser-produced Sn plasmas for extreme ultraviolet lithography light sources Extreme ultraviolet lithography (EUVL) is the leading candidate for lithography beyond the 22 nm half-pitch device manufacturing node. These geometries impose tighter requirements for standard critical dimension metrology and call for new strategies able to quantify and monitor extreme ultraviolet (EUV) specific parameters. In this paper, the approaches to measure two key EUV imaging parameters, namely flare and out-of-band (OoB) radiation, are discussed. EUV sources are known to emit a broad spectrum of wavelengths ranging from EUV to deep ultraviolet (DUV) and beyond. As the DUV can contribute to the photoresist exposure and degrade imaging performance, it is critical to accurately determine the amount of DUV OoB in EUVL exposure tools at the wafer level. In this paper, a methodology using an aluminum-coated reticle to measure the DUV=EUV ratio in resist is discussed. Such a mask is able to provide quantitative in situ information on the scanner DUV content thanks to its ability to transmit DUV and absorb EUV. The experimental OoB results for two EUVL tools are reported and compared with modeling predictions. Flare in EUVL is caused by light scattered by the surface roughness of the optical elements and has a larger impact as compared to optical lithography. As a consequence, a precise and accurate flare metrology is essential to guarantee a proper qualification of the effect, as well as to implement an effective compensation strategy. However, the flare level estimate has been historically based on operator and tool-dependent procedures that are unable to meet the requirements for accuracy and precision dictated by EUVL. A robust in-line approach to flare metrology is developed and qualified. As in the case of OoB, experimental flare results for two EUVL tools are reported. The experimental data are compared to full-chip simulations using the point spread function of the tool's optical system.