Precision Cleaning
September 2005

As the semiconductor industry moves to 300 mm, it's being challenged to create processes that are more uniform and repeatable. Non-contact acoustic cleaning could be a key to the puzzle.

Megasonic Cleaning Charts a Course to the Big Time

By Mark Beck

Non-contact acoustic cleaning – megasonics – is becoming a more readily recognized technology now that semiconductor manufacturers are converting to 300 mm processing and are ready to meet the more stringent requirements of single-wafer cleaning.

This high-frequency acoustic cleaning is widely known as the method for assisting many current wet cleaning and processing steps because the process parameters can be more precisely controlled and cavitation-induced damage to patterned layers can be eliminated.

Parameters that affect megasonic processing include frequency, sonic intensity, chemistry, process time and temperature. The most important physical parameters, however, are turning out to be the cavitation energy, acoustic boundary layer and acoustic micro-streaming induced forces.

How it works

In megasonic cleaning, a piezoelectric crystal array transducer converts alternating electrical energy into mechanical energy using the piezoelectric effect whereby certain materials change dimension when an electrical charge is applied. A ceramic piezoelectric crystal is excited by high-frequency AC voltage, between 500 and 2,000 kHz, causing the ceramic material to rapidly change dimension or vibrate.

Figure 1: Megasonic cleaning uses the piezoelectric effect to produce acoustic waves that move through the cleaning liquid.

These vibrations are transmitted by the resonant masses of the transducer and directed into the liquid, producing acoustic waves in the cleaning fluid. Acoustic cavitation produced by pressure variations in the sound waves moving through the liquid and the effects of acoustic streaming cause particles to be removed from the material being cleaned (see Figure 1).

Acoustic cavitation is the generation and action of cavities, or bubbles, in a liquid. Acoustic waves move through a liquid and produce variations in the liquid's pressure. When the liquid pressure momentarily drops below the vapor pressure during the low-pressure portion of the acoustic wave, small evacuated areas, or cavities, are formed. These cavities quickly become filled with gas (a foreign contaminant such as dissolved oxygen or air) and/or vapor (a gaseous form of the surrounding liquid). These tiny bubbles are set in motion by the acoustic wave.

Figure 2: Plot of bubble radius versus frequency.

The tiny bubbles can expand and contract in the liquid. Bubble expansion is caused by reducing the ambient pressure in the liquid. The bubbles can become large enough to be seen by the unaided eye. The bubbles may contain gas or vapor or a mixture of both. If the bubbles contain gas, their expansion can be caused by rectified diffusion.

Rectified diffusion is the diffusion of dissolved gas from the liquid into the bubble, and vice versa, with the pressure oscillations resulting in a net diffusion into the bubble. This net inward diffusion occurs because the bubble surface area increases during inward diffusion and decreases during outward diffusion; a higher surface area leads to more diffusion.

Figure 3: Table of physical parameters affecting the cavitation threshold.

If the ambient liquid is not saturated with gas, then rectified diffusion must compete with ordinary diffusion from the bubble to the liquid. In that case, the sound pressure amplitude must exceed a certain value for the bubbles to significantly increase in size.

The pressure oscillations that created the bubbles can also cause them to expand and contract. If the pressure variation is great enough to reduce the local liquid pressure down to or below the vapor pressure in the negative parts of the acoustic cycle moving through the liquid, any minute cavities or bubbles that are present will grow larger. If the range of the pressure variation is increased to produce zero and then negative pressures locally in the liquid, then bubble growth is increased. Gas from the liquid diffuses into a bubble during expansion, and leaves the bubble during contraction.

Figure 4: Comparison of boundary layer in ultrasonic and megasonic cleaning.

When the bubble reaches a size that can no longer be sustained by its surface tension, the bubble will expand and then collapse, or implode, which is an important action of the cavitation phenomenon. The bubble action of cavitation has sufficient energy to overcome particle adhesion forces and to dislodge particles attached to substrates in the stream of bubbles. Imploding cavitation bubbles generate shock waves that dislodge particles from substrate surfaces. Cavitation breaks down the molecular force by which a particle is held to surface either by direct impact from bubble implosion or by the fatiguing action caused by repeated bombardment.

Cavitation implosion force varies with the size and contents of the bubble. Larger bubbles are unstable and implode with larger force; smaller bubbles are stable and collapse with less force. Vapor collapses more quickly, resulting in larger implosion force, whereas gas cushions and slows the collapse, resulting in a smaller implosion force.

Figure 5: Table of physical parameters affecting the acoustic boundary layer.

The intensity and effect of cavitation on materials being cleaned is related to the type of acoustic cavitation produced. Two types of acoustic cavitation have been identified: transient cavitation and stable cavitation. Transient acoustic cavitation is produced by ultrasonic cleaning frequencies, between 20 and 350 kHz, which transform low-energy-density sound waves into high-energy-density collapsing bubbles. In transient cavitation, the mostly vapor-filled bubbles exist for only a few acoustic cycles, followed by a rapid and violent collapse. This type of cavitation is likely to produce violent events in the acoustic field, such as radiation of light (sonoluminescence) and shock waves. The level of violence produced is believed to be dependent on the maximum size of transient bubbles that is related to the acoustic frequency. Because transient cavitation concentrates energy into very small volumes and tends to produce very high local temperatures and pressure, it can cause surface erosion and damage to sensitive substrates.

Bubble size decreases as acoustic frequency increases, and the smaller the maximum bubble size, the less violent the cavitation produced. The high frequencies used in megasonic cleaning, from 500 to 2,000 kHz, produce controlled acoustic cavitation that is characterized by mostly small, gas-filled cavities. Unlike the violent implosion associated with vapor-filled cavities in transient cavitation, controlled cavitation bubbles exhibit less violent collapse, producing lower temperatures and pressure. As a result, megasonic cleaning substantially minimizes surface erosion and damage to substrates being cleaned. Stable cavitation produces light in the visible range (violet), while the light produced by transient cavitation is primarily in the UV range (with a peak at 270-290 nm) (see Figure 2).

Cavitation threshold

Cavitation does not occur until a specific threshold is reached. The cavitation threshold is defined as the minimum pressure amplitude required to induce cavitation.

The cavitation threshold has been found to increase with increasing hydrostatic pressure (under most conditions) and to decrease with increasing surface tension, with increasing temperature and with an increasing number of solid contaminants. A reduction in the number of hydrophobic ions (such as C1- and F-) will also decrease cavitation threshold because these ions collect at bubble surfaces and prevent cavitation bubbles from dissolving.

A lower cavitation threshold enables cavitation to occur more readily. This suggests that cavitation could be mitigated under the following conditions: low surface tension, high hydrostatic pressure, low temperature and the presence of as few solid surfaces and contaminants as possible (see Figure 3).

Boundary layer

During megasonic cleaning, the cleaning solution flows swiftly past the substrate being cleaned, forcing chemistry into contact with contaminant particles, removing them from the surface and carrying them away. On a microscopic scale, during acoustic cleaning, fluid friction at the surface of the substrate being cleaned causes a thin layer of solution to move more slowly than the bulk solution.

Figure 6: Table of physical parameters affecting the acoustic streaming velocity.

This layer of slow-moving fluid at the surface is called the boundary layer. The boundary layer effectively shields the substrate surface from fresh chemistry and shields contaminant particles from the removal forces of the bulk fluid (see Figure 4).

Megasonic cleaning has proven especially effective at removing submicron particles in part because it reduces the boundary layer. The higher frequencies of megasonic cleaning reduce the boundary layer to less than 0.5 micron, compared to the boundary layer of 2.5 microns produced by ultrasonic cleaning frequencies. The primary affect of acoustic streaming is the bulk fluid motion of the cleaning solution. The thickness of the boundary layer decreases as the velocity of bulk fluid motion increases.

Figure 7: Ultrasonic vs. megasonic cleaning frequencies and particle sizes.

Reduction of the boundary layer yields several benefits. It allows fresh chemistry to come closer to the substrate and come into contact with smaller particles. This higher chemistry refresh rate results in faster cleaning. Boundary layer reduction increases the effectiveness of the acoustic streaming removal forces by allowing the cleaning solution to rush past the substrate closer to the substrate's surface, forcing chemistry onto particles, removing them from the surface and carrying them away.

The small, controlled cavitation bubbles generated by megasonics are able to remove contaminants within the thinner boundary layer. This effect is especially important in removing small particles and accessing small surface features. Reducing the boundary layer results in increased removal of submicron particles, particles that were previously protected by the boundary layer, as well as increased particle removal overall (see Figure 5).

Acoustic streaming is considered another primary particle removal mechanism of megasonic cleaning. Acoustic streaming is a time-independent fluid motion generated by a sound field. This motion is caused by the loss of acoustic momentum by attenuation or absorption of a sound beam. Acoustic streaming enhances particle dissolution and the transport of detached particles away from surfaces, thereby decreasing particle re-deposition.

Acoustic streaming velocity is a function of energy intensity, geometry, energy absorption, liquid density and viscosity and sound speed in the liquid. Streaming velocity has been found to increase linearly with acoustic intensity (power). Velocity also increases linearly with frequency. Streaming velocity also decreases with distance from the source, due to attenuation.

Figure 8: Single wafer megasonic concept: vertical orientation.

Acoustic streaming comprises several important effects: 1) bulk motion of the liquid, 2) microstreaming and 3) streaming inside the boundary layer.

The primary effect of acoustic streaming is bulk motion of the liquid, the strong localized flow of cleaning solution. The shear force of the bulk liquid motion is the primary particle removal agent. In a closed tank, forces due to sound pressure variation create this bulk fluid motion that carries particles away from the substrate once the molecular attraction of the particle to the surface is broken and the particle is dislodged. Bulk fluid motion increases linearly with acoustic intensity. The bulk fluid motion shear force combines with the other effects of acoustic streaming to increase particle removal.

Figure 9: Single-wafer megasonic concept: horizontal orientation - knife.

A second effect of acoustic streaming is microstreaming. Microstreaming, also known as Eckart streaming, occurs near oscillating bubbles or any compressible substance in the liquid. Microstreaming occurs at the substrate surface, outside the boundary layer, due to the action of bubbles as acoustic lenses that focus sound power in the immediate vicinity of the bubble.

This is a powerful type of streaming in which the bubbles scatter sound waves and generate remarkably swift currents in localized regions. The currents are most pronounced near bubbles that are undergoing volume resonance and are located along solid boundaries. Microstreaming aids in dislodging particles and contributes to megasonic cleaning.

Figure 10: Single-wafer megasonic concept: Busnaina design.

Most of the flow induced by acoustic streaming occurs in the bulk liquid outside the boundary layer. However, there is a third effect of acoustic streaming-Schlichting streaming-which is associated with cavitation collapse and is believed to assist in the removal of small particles and their transport away from surfaces. Schlichting streaming occurs outside the boundary layer and is characterized by very high local velocity and vortex (rotational) motion.

The vortices are of a scale much smaller than the wavelength. Schlichting streaming results from interactions with a solid boundary. Steady viscous stresses are exerted on the boundaries where this type of rotational motion occurs, and these stresses may contribute significantly to removal of surface layers.

The combined effects of acoustic streaming produced in megasonic cleaning may slide, roll or lift a particle from its initial position on a substrate, depending on the size and shape of the particle, as well as the nature of the hydrodynamic force being applied. Acoustic streaming, both inside and outside the boundary layer, enhances cleaning and other chemical reactions. The strong currents and small boundary layer thickness that result from acoustic streaming aid particle transport significantly (see Figure 6).

Ultrasonic cleaning frequencies, between 20 and 350 kHz, produce transient acoustic cavitation. Megasonic cleaning operates at much higher frequencies, from 500 to 2,000 kHz, which produce controlled acoustic cavitation. Controlled cavitation is characterized by stable bubbles that are relatively permanent, can exist for many acoustic cycles, and do not cause damage to substrate surfaces because the cavitation radii is much smaller at higher frequencies and has less energy upon collapse. Thus, megasonic controlled acoustic cavitation is best suited for sensitive substrate surfaces that cannot withstand the heat and pressure of transient cavitation.

In addition, ultrasonics simultaneously cleans all surfaces of a submerged object. This means that ultrasonic cleaning subjects areas of the substrate that do not need to be cleaned to the previously described effects of transient acoustic cavitation. Megasonics accomplishes line-of-sight cleaning; it affects only those surfaces of the object that are in the path of the acoustic wave.

The mechanical effects of both ultrasonic and megasonic cleaning can be helpful in speeding particle dissolution and in displacing particles. Ultrasonics and megasonics have also been demonstrated to speed or enhance the effect of many chemical reactions. In addition, residual cleaning chemicals can be removed quickly and completely by either ultrasonic or megasonic rinsing. However, there are applications for which megasonic cleaning clearly would be favored.

The effects of ultrasonics and megasonics on substrate surfaces and particle removal results provide the basis for identifying the best applications for each process. Ultrasonic cleaning is most appropriate for strong, heat-tolerant substrate materials requiring multi-surface cleaning. Ultrasonics is also well suited for the removal and/or dissolution of large particles from chemically tolerant substrates.

Megasonics is most appropriate for heat- or chemical-sensitive substrates that cannot withstand the heat and pressure of transient cavitation and for applications requiring line-of-sight-dependent cleaning. Parts that cannot be cleaned with ultrasonics because they are sensitive to the frequency or transient cavitation effects can often be cleaned with megasonics. Megasonic cleaning is also applicable in the removal or dissolution of small particles (less than 0.3 micron). For example, this cleaning technique has been proven effective for removing 0.10-micron particles from silicon wafers and other cavitation-sensitive products, without causing substrate damage (see Figure 7).

Megasonic cleaning may be used with a variety of chemistries, including water, neutral aqueous solutions, alkaline aqueous solutions, acidic aqueous solutions, ethyl lactate, alcohol, acetone, N-methyl pyrollidone, dibasic esters and glycol ethers.

Although megasonic cleaning is used primarily for particle removal, it can also be used to increase the efficiency of chemical cleaning with surfactants or detergents. Removal of other contaminants depends on the solution in the tank.

Cleaning chemistries play a significant role in megasonic cleaning, because the chemical composition of the cleaning solution may affect the zeta potential, or the atomic attraction forces between particles. The zeta potential is neutral or negative in high pH chemistries. The zeta potential does not change with up to two orders of magnitude change in the concentration of chemicals.

The addition of megasonic energy to the chemistry substantially enhances particle removal. Chemists have succeeded in getting very dilute solutions to clean effectively with the addition of megasonics to the cleaning process.

Almost 80 percent of the current manufacturing is already single wafer. The industry is moving to 300 mm and this will challenge processes to be more uniform and repeatable. Cleaning is being required to be at the point of use, not in central service locations. The process parameters are narrowing with feature sizes pushing sub-100 nm. While current 200 mm cleaning is predominantly batch, scaling these batch processes to 300 mm is challenging.

Megasonic has been the standard for featured layer cleaning for some time; damage has recently become an issue with concentrating megasonic products like nozzles. The need to prevent damage has resulted in the application of the acoustic energy as close to the point of use as possible. Current single-wafer products and concepts that meet these needs have been proposed and tested.

Full wafer transducers that closely couple the megasonic energy to the wafer surface as proposed by Ahmed Busnaina patent application, or the current Verteq Goldfinger single wafer megasonic rod, are examples of point-of-use applications of acoustic energy. Energy density control allows for reduced power and no damage. Intensities below 5 W/cm2 have proven to be very effective at removing 0.1 µm particles.

One of the historical drawbacks of megasonic cleaning has been the throughput of the batch systems. The standard clean has been a two- to three-step process. With the batch systems, process time per tank can be 10-20 minutes while current single-wafer technology with the acoustic energy delivered to the precise point of use has resulted in process times in the sub-one-minute range.

Additionally, multiple chemistries can be mixed and the processing time further reduced. With production facilities like foundries operating a wide mix of products with many split or small lots, single wafer lends itself to the flexible, quick turn, custom process that manufacturing currently requires (see Figures 8, 9 and 10).

The most important new process parameter driving the development of new megasonic technologies is the need to provide a more uniform acoustic field in which the substrate is processed. Not all of the future processes are clean; there are many applications for acoustically assisted chemical processes, which leads us into the field of Sono-chemistry.

Acoustic energy has proven to assist cleaning. It has also been proven to remove particles in deep trenches and vias. The oscillations of the bubbles effectively pump and stir the chemistry on the microscopic level inside the boundary layer near the surface. The danger is that damage can still occur if the uniformity and intensity of the energy is not precisely controlled. The process is also more repeatable wafer to wafer if the energy is constant over the whole wafer surface.

The future is Sono-chemistry

The chemical effects of ultrasound are diverse and include dramatic improvements in stoichiometric and catalytic reactions. In some cases acoustic sound irradiation can increase the reactivity by nearly a million fold.

Ultrasonics has long been used to assist the chemical reactions in baths. Applications include electroplating where sonic streaming will induce flows that can counteract the depletion of the diffusion layer between electrode and electrolyte ions, which ordinarily limit the current yield in plating. Ultrasonics has not been applied to wafer processing to assist chemical processes due to concerns over damage. Megasonics is currently being applied to many of these processes.

About the author

Mark J. Beck is Chief Executive Officer of ProSys Inc. (Campbell, CA), a company he founded in 1996. He previously worked for National Semiconductor, Semiconductor Technologies, and Cypress Semiconductor. He received a BS in mechanical engineering and manufacturing management from the University of California, Berkeley and is the co-inventor of tow patents for megasonic cleaning equipment.