Rippey Corporation

The performance of post-CMP PVA brush scrubbing operations is not only influenced by the cleaning chemistry and scrubber parameters, but more importantly by the physical and mechanical properties of the polymer material. While hydrodynamic forces and surface charge characteristic effects assist in removing particles from a surface, an equally important mechanism in PVA brush scrubbing is physical contact. Contact cleaning, as the name implies, involves the PVA sponge brush coming into direct contact with the particles and sweeping them off the surface. The material properties of the brush therefore become significant.

This paper will discuss the influence of PVA material physical properties on cleaning performance and how these properties are effected by scrubber parameters. Also discussed will be the effects of cell structure, porosity, brush compression and rotational velocity on fluid flow through the PVA brush roller.

PVA brush scrubbing technology is the most common form of cleaning in post-CMP applications¹. Although there have been numerous papers published on the science of brush scrubbing, few have addressed the brush itself. During semiconductor manufacturing, the PVA brush is one of the few solid objects, which comes into direct contact with the face of the wafer. Cleaning performance during brush scrubbing is not only dependent on the chemistries used and the tool design, but also on the physical properties of the brush material².

The current technology for the introduction of cleaning solutions, in post-CMP brush scrubbing operations is the use of flow-through brush mandrel (brush core) techniques. In this approach, a hollow core with perforations forms the brush support. DI water or cleaning chemistries are delivered onto the PVA brush through the core. The solution then flows from the inside, through the brush, and is evenly distributed to the surface being cleaned. This flow-through design provides a better distribution of the cleaning solution, improves cleaning performance and increases brush life. Liquid flow through the brush helps to reduce particle accumulation (brush loading) on the brush, thereby extending the useful lifetime of the brush and reducing the cost of ownership for the scrubbing process.

This paper presents the advantages of the flow-through technology, compared to spray or drip techniques, and the influence of PVA brush material properties on cleaning performance.

The manufacturing techniques used for pore formation in the PVA sponge material can have a significant effect on the flow characteristics and other material properties. Common methods for sponge manufacturing are based on gas formation or gas injection (foaming technique). The Rippey/Kanebo PVA sponge material uses a proprietary technique of reactive phase separation for the formation of pores. This method consists of adding a pore-forming agent to the liquid polymer mixture. As the polymer is converted to its non-soluble state, the pore-forming agent assists in the stabilization of the developing insoluble polymer network and contributes to exclusion volume of the pores. The resulting differences in pore structure/distribution are shown below (Figure 1).

As can be seen from the above photographs of the internal pore structure, the reactive phase separation techniques, using starch as the pore forming agent, produce a highly consistent and open cell structure. The advantages of this method are tighter control of porosity, higher consistency and a more open cell structure. The uniform and open architecture of the brush roller results in a reduced pressure drop and a uniform flux over the surface. Air or gas formation (foaming) produces a higher concentration of closed cell pores and a less consistent pore distribution due to the difficulty in controlling the distribution of air bubbles.

Unlike nylon bristle brushes, which are set to come close but not touch the surface of the substrate, soft PVA brushes clean by coming into direct contact with the wafer surface. As a result, cleaning performance is dependent on compression of the brush against the wafer surface. The further the brush is compressed against the wafer surface, cleaning performance continues to increase until a maximum is reached at a compression of approximately 2 mm. After this optimization is reached, cleaning performance starts to degrade with further compression³.

The brush is in the form of an elastic open cell polymer sponge. As the brush, or the brush nodule, is compressed against the wafer surface, several physical property changes will take place. The brush will have a localized increase in density and corresponding drop in porosity as a direct result of the compression and the collapse of the pore structure (see Figure 2 and 3). The compression of the brush will also induce a localized pressure increase around the nodule resulting from the pumping action of expelling the liquid from the pores. This pressure will vary in proportion to the porosity of the sponge and the viscosity of the liquid (cleaning solution). The extent of these localized effects will depend on the rotational velocity of the brush and the mechanical properties of the brush material. The resulting local density and pressure increase coupled with the pore size decrease from compression will hinder flow through the brush until the structure rebounds to the original volume.

The unique pore structure created by reactive phase separation is most beneficial when using flow-through techniques. Flow-through brush cores can offer distinct advantages over solid core designs. With liquid or cleaning solutions flowing through from the inside, the brush is continually rinsed or flushed from the inside out. This flushing substantially reduces the effect of brush loading on cleaning performance as demonstrated in Figure 4. This experiment was conducted on an Oliver Design dual rail disk scrubber equipped with a new flow-through core. The brushes were dipped into diamond slurry prior to scrubbing polished disk media. The experiment was conducted both with and without flow-through. The DI water flow to the cores was set at 0.95 gpm for a set of four brushes. This experiment measured the number of runs before the slurry was removed and particle adders returned to baseline.

Some preliminary experiments have been conducted on the flow characteristics of the PVA brush when used in conjunction with the flow-through core⁴. The total flow through the core was measured at varying rotational velocities (0 – 1000 rpm) for 15 seconds, both with and without brushes installed. It should be noted that some of the rpms are much higher than commonly used in post-CMP cleaning. The results are detailed in Figure 5. The initial line pressure feeding the cores was 25 psi with a flow of 0.95 gpm With no rotation, the total flow measured was approximately the same for cores with and without brushes. This indicates that at relatively low flow (a flux of 0.578 ml/cm² • sec) the brush offers little resistance. What was not expected from this experiment was a drop in flow with brushes installed as the rotational velocity increased. As with the core itself, we had expected an increase in flow from the added centrifugal force. This is only preliminary data, but it appears to indicate that the brush has higher backpressure as the rotational velocity increases. This could be caused by several possible factors, which would need further investigation:

The tangential force resulting from the rotational acceleration could distort the flexible polymer network decreasing cell volume and thereby inhibiting flow.
Increased resistance due to a longer path the fluid must take through the brush caused by rotation.
Turbulent flow within the brush from increased fluid velocity.
A pressure front within the brush, resulting from the brushes being compressed together, would increase as a function of the rotational velocity resulting in a decreased flux.

Work is in progress to gain a better understanding of the factors involved and their influence.

In order to obtain clean wafers after CMP, you need to overcome the particle adhesion forces (mainly van der Waals) holding the particles to the wafer surface. Brush scrubbing utilizes direct contact between the brush asperities (surface roughness) and the particles as the main removal force⁵. According to Zhang et al., this removal force is a combination of asperity contact forces and asperity–particle adhesion (resulting from the particle colliding with a soft material of high surface energy) forces. Hydrodynamic and zeta potential forces will aid in both particle removal and preventing particle reattachment to both wafer and brush. All of these forces are aided by the nodule brush configuration, which by its design has a higher surface pressure (for a given compression distance) and reduced hydroplaning potential when compared to flat or ridged brush designs. This enables the nodule brush to come into closer contact with the particles and remove them with greater force.

The authors would like to thank Yassin Mehmandoust, of Oliver Design and the Oliver Design Applications Laboratory for their assistance in generating the data referenced in this report.

W.C. Krusell, I.J. Malik, F. Mohr, D.J. Hymes, Double-Side Wafer Scrubbing Beyond Post-CMP Cleaning, Electrochemical Society, Fourth International Symposium on Cleaning Technology Semiconductor Device Manufacturing 1995.
A.A. Busnaina, N. Moumen, J. Piboontum, Contact Post-CMP Cleaning of Thermal Oxide Wafers, CMP-MIC Conference, 1999.
W.C. Krusell, I.J. Malik, M.A. Ravkin, J.J. Farber, J. Zhang, J.M. de Larios, The Resurgence of Mechanical Brush Scrubbing III, SEMICON Korea, January 1995.
K.G. Bahten, D. McMullen, Y. Mehmandoust, Liquid Flow Dynamics for Flow-Through Core Post-CMP Brush Scrubber Designs and the Influence of PVA Brush Pore Structure, VMIC, VLSI Multilevel Interconnection Conference, Santa Clara, CA, 1999.
F. Zhang, A.A. Busnaina, G. Ahmadi, Particle Adhesion and Removal in Chemical Mechanical Polishing and post-CMP Cleaning, Journal of The Electrochemical Society, 146 (7) 1999.