Removal of nanoparticles from surfaces using ultrasound at megahertz frequencies

Within the last decade, the semiconductor industry has overstepped the frontier between micro- and nanotechnology. With the ever shrinking dimensions of the functional elements now being located in the nanometer range, a multitude of new technological problems will have to be tackled. As surface and interface processing steps form an important part of device fabrication, they will have a crucial influence on subsequent processing steps as well. One major issue is the removal of nano-particulate contamination residing on the wafer surface in the form of residues from abrasive process steps such as chemical mechanical polishing, resist removal or plasma etching. Since substrate losses will have to be limited to sub-nanometer levels, corrosive chemistries might only be used when strongly diluted. However, the loss in chemical aggressiveness severely affects the efficiency of pure-chemical cleaning. Hence, the chemical particle removal will have to be physically assisted through the introduction of nano-mechanical forces to the particles in question. Several techniques have been proposed as solutions to that dilemma, ranging from laser ablation to spray- and megasonic cleaning. The latter utilizesultrasonic agitation of the cleaning liquid at frequencies in the megahertz region, which induces a multitude of possible cleaning mechanisms. Most of these mechanisms involve oscillating microbubbles in the ultrasonically agitated cleaning fluid. As the strength of the bubble oscillation and therefore the magnitude of those effects are strongly depending on the bubble radius, precise control over the size and number of the active bubbles is crucial. To achieve that, the process conditions such asthe gas content of the liquid, the extent of acoustic reflections in the cleaning tank and the sonication conditions have to be chosen carefully in order to yield a signicant amount of acoustic cavitation (i.e. cavitation activity) and particle removal. To obtain a complete fundamental study of acoustic caviation, information about a multitude of complementary experimental parameters needs to be combined to gain insight into the underlying physical processes. At frequencies of 1 MHz, the oscillating bubbles are so small that it is becoming difficult to observe them in a direct way, e.g. by high-speed imaging. Hence, a metrology based on secondary effects of acoustic cavitation is more suitable to study cavitation activity. In a first part of this work, sonoluminescenceand cavitation noise measurements were combined with synchronized High-speed stroboscopic Schlieren imaging in order to obtain a complete picture of the initialization and temporal evolution of acoustic cavitation in a single wafer cleaning tank. Results obtained during continuous sonication of argon-saturated water at various nominal power densities indicate that acoustic cavitation occurs in a cyclic manner, during which periods of stable (and possibly cleaning) and increased inertial (and possibly damaging) cavitation partially alternate. It is shown that a correlation exists between sonoluminescence, Schlieren contrast, number of visible bubbles and the ultraharmonic and broadband signals extracted from the cavitation noise spectra. This signal correlation helps to identify different caviation regimes. The occurrence of higher and ultraharmonics in the acoustic emission spectra is characteristic for the stable cavitation state. The inertial cavitation state is characterized by a strong attenuation of the sound field, an explosive "nucleation" of bubbles, sonoluminescence and the occurrence of broadband components in the acoustic spectra. Both states can only be sustained at sufficiently high intensities of the sound field. At lower intensities, their occurrences are limited to short, random bursts and particle removal eciency is reduceduntil it vanishes when a continuous state of cavitation activity ceasesto exist. These bursts are accompanied by the occurrence of fast-movingbubbles. The origin of these "bubble streamers" has been investigated and their role in the initialization and propagation of cavitation activity throughout the sonicated liquid is discussed.In a second part of this work, it is shown that cavitation activity as measured by means of ultraharmonic cavitation noise can be significantly enhanced when pulsed sonication is applied to a gas supersaturated liquid under traveling wave conditions. It is demonstrated that this enhancement coincides with a dramatic increase in particle removal and is therefore of great interest for megasonic cleaning applications. Furthermore, it is shown that the optimal pulse parameters are determined by the dissolution time of theactive bubbles, whereas the amount of cavitation activity depends on the ratio between pulse-off and pulse-on time as well as the applied acoustic power. The optimal pulse-on time is independent of the corresponding pulse-on time but increases significantly with increasing gas concentration. It is shown that supersaturation is needed to enable acoustic cavitation under aforementioned conditions, but has to be kept below values, for which active bubbles cannot dissolve anymore and are therefore lost during subsequent pulses. For the applicable range of gas contents between 100% and 130% saturation, the optimal pulse-on time reaches values between 150 and 340 ms, respectively. Full particle removal of 78 nm-diameter silica particles at a power density of 0.67W/cm² was obtained for the optimal pulse-on times. The optimal pulse-on time values were derived from the dissolution time of bubbles with a radius of 3.3 μm and verified experimentally. The bubble radius used in the calculations corresponds to the linear resonance size in the applied sound field, which demonstrates that the recycling of active bubbles is the mainenhancement mechanism. The optimal choice of the pulsing conditions, however, is constrained by the trade-off between the effective sonication time and the desire to have a sufficient amount of active bubbles at lower powers, which might be necessary if very delicate structures have to be cleaned. The different interdependent effects that accompany acoustic cavitation strongly depend on the bubble size. Therefore, bubble size control is required for the establishment of a damage-free, yet efficient low-power cleaning process. In the last part of this thesis, a methodology for the evaluation of the bubble size distribution by means of a combination of cavitation noise measurements and ultrasonic pulsing is presented. The key component of this methodology is the definition of an upper threshold size below which bubbles have to dissolve inbetween subsequent pulses in order to be sustainably recycled. The experimentally determined bubble size distributions for different power densities are interpreted in the frame of numerical calculations of the oscillatory responses of the bubbles to the intermittent driving sound field. The distributions are found to be shaped by the size dependent interplay between bubble pulsations, rectified diffusion, coalescence and the development of parametrically amplied shape instabilities. Furthermore, a relation between cleaning and acoustic cavitation activity is established while considering the changed acoustic conditions bubbles experience in close vicinity to a solid surface. The influence of the inertialtype of cavitation activity on the structural damage that is occurring during the cleaning process was evaluated. It is found that the damage to the fragile structures that ought to be cleaned is signicantly reduced, when the applied power is kept below values where inertial cavitationis observed.Based on these results, a reactivation-deactivationmodel is proposed to explain the enhancement of cavitation activity. Inthis model, the pulse-on time determines the magnitude of the reactivation of partially dissolved bubbles and the deactivation of active bubbles by coalescence. It is shown that the subsequent recycling of active bubbles leads to an amplication of cavitation activity, which saturates after a certain number of pulses. The model was fitted to the experimental data for the cavitation activity measured by means of ultraharmonic cavitation noise as a function of the pulse duration. Measurements of thedevelopment of the cavitation noise and sonochemiluminescence over a sequence of pulses for different pulse-on and -off times confirm the overall validity of the proposed model to describe the accumulation ofcavitation activity, but also demonstrate its limitations with regard totemporal fluctuations of the cavitation activity. High-speed images of the cavitation field relate the deactivation of active bubbles by coalescence to the increase in volume concentrations of larger bubbles.nrpages: 216status: publishe