Friction studies of silica slurries using a pin-on-disk rheometer with IC1000 pad surfaces

Catheryn L. Jackson¹ and David Mosley²

¹Core R&D and ²Electronic Materials, The Dow Chemical Company, Collegeville, PA

INTRODUCTION:  Understanding friction processes during the chemical mechanical planarization (CMP) of Si wafers for microelectronics applications is invaluable in slurry and pad design,¹ and has been used to improve performance.  Often large scale polishers are used for measurements^2,3 but this takes a large amount of material and is not ideal for research testing of new slurry chemistries or novel pad designs.  The polishing of silicon wafers in CMP is a controlled form of wear, and in order for wear to occur it is believed that boundary lubrication⁴ is needed so that the pad asperities can be in direct contact or very close proximity to the wafer. The friction coefficient in the CMP process, which varies between 0.2 and 0.8 depending on the slurry, wafer, and pad, creates most of the heat in the process leading to changes in slurry viscosity, pad modulus, and surface tribocorrosion rates. Stribeck curves during polishing often show features which can be interpreted as changes in the ‘lubricating’ layer as a function of sliding velocity. Additionally, wearing or “glazing” of pad asperities can cause instability in the CMP process and depends in part on particle morphology.⁵ The glazing effect is expected to alter the hydrodynamic state and degree of close contact at the interface.

We have used an Anton-Paar MCR-502 rheometer with the pin-on-disk attachment for tribology to simulate a CMP process.  The pin-on-disk has a contact area that allows for down pressures during pad conditioning and “polishing” runs to be comparable to commercial polishers, namely in the 1-7 psi range.  The instrument is capable of a ramp in sliding speed from 10^-7 to 1.4 m/s, and our experiments are in the 0.01 – 1 m/s range relevant to polishing processes. The instrument measures the friction factor as well as the ‘gap’ between the surfaces as a function of sliding velocity. This paper describes development of a method to measure the coefficient of friction (COF) of a commercial silica slurry with a conditioned commercial polyurethane CMP pad.

EXPERIMENTAL:  The Anton-Paar MCR-502 with the T-PID/44 pin-on-disk attachment has three ~1/4” stainless steel pins on the top fixture, which is mounted on a spring with adjustable stiffness (here set at an intermediate value of six) in the z-direction.  The pad pieces were attached to the pins with a pressure sensitive adhesive backing on the IC1000 top-pad, which had no subpad; a hole punch was used to cut pieces to fit the pins.  The bottom fixture has a threaded clamp with a sleeve insert to hold in place a 2.75” Teflon or other disk insert, here we used either a 2.5” diamond grit Saesol disk LPX-AR3B66 for conditioning or a 2” Si wafer (thermal oxide layer, 4.8 μm thick) for the polishing steps.  The wafer was attached to the Teflon bottom securely with Kapton double stick tape.  As is typical the pads are conditioned, in-situ, before polishing, to create asperities that will polish the surface of the wafer.  The temperature was controlled at 25 °C with the Peltier bottom plate.  The conditioning used distilled (DI) water, equilibrated at temperature for 1 minute at 1.5N, followed by a 60 second

velocity of 1.0 m/s at constant 1.5N force.  The pads were rinsed well with DI water and the cell cleaned out of pad debris.  The Saesol disk was then exchanged for the mounted Si wafer.  A Klebosol slurry at pH= 10.7 was studied at two concentrations, 12% and 24% by weight.  The slurry was added to the bottom cell and the polishing run was initiated.  This again allowed for equilibration at temperature for 1 min at 1.5N normal force, followed by a sliding velocity ramp from 0.001 – 1 m/s at 1.5 N normal force, with a logarithmic data sampling time protocol, to obtain the Stribeck curve.  Each run used fresh pads and a new Si wafer, as the wear of either could alter the results for duplicate runs. Confocal microscopy was used to study the pad surface morphology before and after the Stribeck curves.

RESULTS:  The Stribeck curves for water, 12% and 24% Klebosol slurries are shown in Figure 1. The water curve appears to show the transition between boundary and (elasto) hydrodynamic lubrication for a CMP pad on thermal oxide system. In some cases the tribometer can measure an increase in the gap between the pad and wafer when using water only. The slurries show behavior indicative of a more complex phenomenon. Comparing bulk viscosities of the slurries and water, a normal Stribeck analysis would predict an earlier onset of hydroplaning, but the friction curves are much flatter than in the water case. The tribometer gap stays the same or decreases in the presence of the slurries, which may indicate some pad glazing is being measured. Confocal microscopy images of pad pieces after the Stribeck curves were used to ascertain whether significant glazing occurred.

Figure 1.  Stribeck curve for 12% (circles) and 24% (triangles) Klebosol  slurries, and control of DI water at pH10.7 (squares).

REFERENCES:  1. Rosales-Yeomans, J. Electrochem. Soc. (2005).  2. Lee, Thin Solid Films (2014).  3.  Carter, J. Electrochem. Soc. (2007).  4.  Tichy, Int. J. Solid Structures (2000).  5.  Reddy, Int. Conf. Planarization/CMP Techn. (2010). 6.  Zhang, Friction (2015) and Yoshizawa, J. Phys. Chem. (1993).