No abstract

Spectacular progress in developing advanced Si circuits with reduced size, along the track of Moore's law, has been relying on necessary developments in wet cleaning of nanopatterned Si wafers to provide contaminant free surfaces. The most efficient cleaning is achieved when complete wetting can be realized. In this work, ordered arrays of silicon nanopillars on a hitherto unexplored small scale have been used to study the wetting behavior on nanomodulated surfaces in a substantial range of surface treatments and geometrical parameters. With the use of optical reflectance measurements, the nanoscale water imbibition depths have been measured and the transition to the superhydrophobic Cassie-Baxter state has been accurately determined. For pillars of high aspect ratio (about 15), the transition occurs even when the surface is grafted with a hydrophilic functional group. We have found a striking consistent deviation between the contact angle measurements and the straightforward application of the classical wetting models. Molecular dynamics simulations show that these deviations can be attributed to the long overlooked atomic-scale surface perturbations that are introduced during the nanofabrication process. When the transition condition is approached, transient states of partial imbibition that characterize intermediate states between the Wenzel and Cassie-Baxter states are revealed in our experiments.

Laser cleaning is one of the new promising dry cleaning techniques considered by semiconductor companies to replace wet cleans in the near future. A dry laser cleaning tool was tested that uses an inert gas jet to remove particles lifted off by the action of a DUV excimer laser. A model was developed to simulate the cleaning process and analyze the influence of experimental parameters on laser cleaning efficiency. The best cleaning efficiencies obtained with 1.0 μm SiO2, ∼0.3 μm Si3N4, and 0.3 μm SiO2 particles deposited on Si wafers were 84±8%, 33±4%, and 12±7%, respectively. This is in qualitative agreement with theoretical calculations showing the existence of a size threshold for the removal of nonabsorbing particles by dry laser cleaning. Among the process parameters tested to optimize the process efficiency, fluence showed the highest influence on removal efficiency, before the number of laser pulses and the laser repetition rate. The use of high fluences was limited by the damaging of the wafer surface, which was not homogeneous on a macroscopic scale. The optimum number of laser pulses per unit area depended on the type of particle. The laser repetition rate had no significant influence on cleaning efficiency and can be used to reduce process time. The influence of capillary condensation on the process was demonstrated by the higher removal efficiency of 0.3 μm SiO2 and Si3N4 particles, 88±6% and 78%, respectively, upon exposure of wafers to air saturated with moisture prior to laser processing. This was attributed to the explosive evaporation of capillary condensed water, similar to the mechanism proposed for liquid assisted laser cleaning.

This work proves that blistering is the partial de-lamination of a thick enough Al2O3 layer caused by gaseous desorption in the Al2O3 layer upon thermal treatments above a critical temperature: the Al2O3 layer acts as a gas barrier and bubble formation occurs. First, using an atmospheric pressure rapid thermal processor with an atmospheric pressure ionization mass spectrometry, desorbing species upon heating of Si / Al2O3 samples are identified: evident desorption peaks are observed around 400 °C for all spectra. The spectrum for m/e = 18, an indication of H2O, illustrates that gaseous desorption from Al2O3 and from the Si substrate itself continues up to 600 °C and 700 °C, respectively. Also, it is shown that in the case of a 30 nm Al2O3 layer, blistering starts at same annealing temperatures as gaseous desorption begins. In the case of a thin enough (≤ 10 nm) Al2O3 film, blistering does not show. To complete the proof, elastic recoil detection measurements clearly show that after annealing a thick Al2O3 film above 400 °C the H content is higher near the c-Si interface as compared to the near surface.Fortunately, effective lifetime and capacitance voltage measurements show that 5 to 10 nm Al2O3 layers can still be adequate passivation layers after being annealed in N2 environment at temperatures up to 500-700 °C: (i) interface trap densities (Dit) can remain below 1x10 11 cm -2 and (ii) fixed charge densities (Qf) stay negative and in the order of -3x10 12 cm -2 .Random local Al back surface field (BSF) solar cells, fabricated using a blistered film as rear surface passivation and no additional contact opening step, clearly show that random local BSFs are created upon firing of a blistered rear passivation layer covered by metal. Therefore, it is clear that blistering should be avoided, since it will reduce the overall rear surface passivation.

Brush scrubbers are among the most commonly used instruments for wafer-cleaning applications nowadays. However, the removal mechanisms of nanosized particles are far from clear, especially because no direct experimental data are available to backup theoretical models in the literature. This study combines a theoretical approach based on a force analysis with an experimental study of the removal of nanosized slurry particles. In the theoretical part, all forces affecting the adhesion and the removal of particles are evaluated to determine which are dominant in two extreme removal mechanisms: lifting and rolling. In the experimental part, the removal efficiency of 34nm SiO2 particles is investigated by using the haze approach. Based on a study of the aging of contaminated wafers, conditions are selected where no chemical bonds are formed between a particle and a substrate. Force analysis and experimental observations both show that nanosized particles cannot be lifted directly by a brush. Instead, rolling should be the main particle-removal mechanism. The average fluid film thickness between brush and wafer surface is determined based on power measurements of the brush motor and a friction analysis, indicating that the system is in a hydrodynamic lubrication regime across a wafer in average. In this frame, results also show that the hydrodynamic drag force is the dominant removal force for nanosized particles.