Quantitative In-Situ Measurement of Asperity Bending

Under the Wafer During Polishing

Caprice Graya, Daniel Aponea, Chris Rogersa, Vincent P. Mannoa, Chris Barnsb, Mansour Moinpourb, Sriram Anjurc, Ara Philipossiand

aTuftsUniversity, Department of Mechanical Engineering, Medford, MA, USA

bIntel Corporation, Santa Clara, CA, USA

cCabot Microelectronics Corporation, Aurora, IL, USA

dUniversity of Arizona, Department of Chemical Engineering, Tucson, AZ, USA

ABSTRACT

The interaction of the wafer, slurry and pad determines the material removal rate during Chemical Mechanical Planarization (CMP). Dual emission laser induced fluorescence (DELIF) provides a means to measure the slurry layer thickness between the wafer and a Fruedenbergy FX9 pad during CMPwith high spatial (4.3 m/pixel) and temporal resolution (2 Hz). In this paper we present some preliminary measurements of pad compression using DELIF to measure the standard deviation of asperity height. Static slurry layer images were captured at high (70 kPa) and low (0 kPa) down-force applied to the wafer. In-situ, dynamic images at 10 kPa down-force applied to etched wafers were imaged.Two wafers were etched such that they contain square wells, one wafer with 27 m and the other will 14.5 m deep wells. In the static case, asperity compression is directly related the amount of fluid displaced. In the dynamic case, asperity compression is 35% greater under the 27 m wells than the 14.5 m wells.

INTRODUCTION

Slurry layer behavior interactions with the polishing pad and wafer during Chemical Mechanical Planarization (CMP) have been extensively studied both by computer modeling [1-3] and experimentation [4,5]. Knowledge of the pad/slurry/wafer interface provides insight into the physical processes that govern material removal rate.Until recently, it has been difficult to obtain measurements of slurry layer thickness directly. Dual Emission Laser Induced Fluorescence (DELIF) is a technique that was developed and reported by Copetta and Rogers[6] in 1998, to view properties of the slurry layer measured in-situ during the polishing process. Since then, DELIF has been used to attain spatially and temporally averaged values for variables that include pH [6], temperature [6,7] and slurry film thickness [8]. Recently the technique has been refined such that we can now capture high spatial and temporal resolution images during CMP [9]. This paper will explore the feasibility of studying pad characteristics such as mean asperity compression and pad surface roughness based upon slurry film thickness measurements. Optical limitations of the technique will also be discussed.

EXPERIMENTAL OVERVIEW

All experiments are performed a laboratory scale Struers RotoPol-31 table top polisher unit which is a 1:2 model of a SpeedFam-IPEC 472. Figure 1 illustrates the polishing setup. The polishing head has been replaced by an aluminum shaft that is driven by a Dayton ½ HP DC

Figure 1. Polishing Setup. A Struers RotoPol polisher carries the platen. The polishing head is replaced by a shaft. A force table is under the polisher.

motor. The 30 cm (12”) platen is controlled via a LabVIEW interface and is fitted with a Mitsubishi Freqrol frequency modulator for speed control. The platen and the shaft are rotated in the same direction and speed, and dynamic experiments were performed at rotation rates of 30 RPM. Slurry is injected at a rate of 50 cc/min to the center of the pad during polishing by Masterflex peristaltic pump. The polisher sits atop an AMTI Force table capable of measuring forces and torques in all 3 dimensions. A weighted lever arm at the top of the shaft (not pictured) allows for the application of down-force to the wafer. The entire apparatus is isolated from external vibrations by a steel table.

A complete description of DELIF can be found elsewhere [6,10]. However, a brief explanation will be provided here discussing the details specific to this experiment. In order to view the slurry layer, it is necessary to replace the silicon wafer with an optically transparent BK7 wafer. The wafer is scaled down to 76.2 mm (3”) in diameter to match the scaling of the polisher and is 12.7 mm (0.5”) thick to prevent slurry from splashing onto the top of the wafer. Figure 2 illustrates the optical characterization of DELIF used here. First a Nd/YAG laser emits a pulse at 355 nm. That energy is absorbed by the polishing pad material and re-emitted with an emission peak at 392 nm. The slurry is mixed with Calcein dye at a concentration of 1 g/L. There is very little spectral overlap between the laser emission and the Calcein absorption, therefore the almost all of the Calcein emission is due to absorption of light emitted by the pad. The Calcein reemits the light with a spectral peak at 530 nm. The emitted light passes through a zoom lens to a dichroic beam splitter, which reflects wavelengths shorter than 496nm and transmits longer wavelengths. The reflected and transmitted light passes through some additional filters and is collected by 2 cameras, A and B. The filter regions for the cameras are indicated in figure 2. The images are affected equally by non-uniformities in the excitation source and therefore taking the ratio of B/A corrects the overall intensity data from the 2 cameras. All data processing discussed is done on these B/A images. Image intensities correlate linearly with slurry film thickness. A method for attaining a calibration factor to convert intensity values to slurry layer thickness is discussed elsewhere [11] and will not be discussed here. The calibration factor used for data reported is 70 m per ratio unit. Figure 3a is an

Figure 2. Emission and Absorption Spectra of DELIF components. Optical filter regions for Cameras A and B are highlighted.

example of a ratio intensity map of the fluid layer between a flat Fruedenberg FX9 polishing pad and a BK7 glass wafer. Figure 3b is an image of a patterned wafer over a flat Fruedenberg FX9 pad.

Two experiments will be discussed here. The first is a static case in which we compare asperity compression and mean slurry film thickness due to applied down force. The image of the 0 kPa was taken with the wafer sitting atop a conditioned Fruedenberg FX9 pad and no shaft or lever arm pressure. The 0 kPa case was compared to a 0.5 kPa (10 psi) case in which the wafer was pressed against the pad in the same location as the 0 kPa case. The second experiment was a dynamic case in which patterned wafers were rotated against a conditioned Fruedenberg FX9 polishing pad. Wafers were etched with HF such that they contained arrays of square wells

Figure 3. DELIF images at 10 kPa on a flat Fruedenberg FX9 polishing pad. (a) A flat wafer over the pad. (b) A wafer with etched 27 um deep wells.

ranging from 0.25 mm2 to 4 mm2. The two wafers tested had well depths of 14.5 and 27 um. Material removal was minimized by diluting the slurry 10X with water. These etched wells were also used to correlate image intensity values tofluid layer thickness.

RESULTS

Before discussing the image analysis, pad topography should be considered. Figure 4 shows a profilometer scan across the 76.2 mm region of the pad; this sample length was chosen because it is the same as the wafer diameter. Note that in addition to pad asperities (micron-scale topography) there is also a fair amount of global pad topography. The imaged regions of interest (ROI) are approximately 6 mm2 regions, which will only show the micron-scale topography. The wafer will be resting only on the highest portion of the pad, therefore, the average slurry film thickness in each image varies depending on the pad’s global topography. In addition, the image-to-image slurry film thickness during a run and applied global down-force to the wafer correlate poorly due to the relatively small size of the ROI.

For the static experiment, 14 different points on the pad were imaged at 0 and 70 kPa down-force. Figure 5 shows the histograms for one of these points on the pad at the two down-forces. The 0 kPa case refers to images of the slurry layer under only the weight of the wafer. Asperities are assumed to be minimally compressed in this case because the histograms closely resemble Gaussian curves with minimal skew. Histograms of images for the 70 kPa case exhibit skew towards darker intensities, or a thinner slurry layer. Since the slurry completely fills the spacesbetween the wafer and the polishing pad surface, the shape of the slurry layer thickness distribution over an image should be the same as the asperity size distribution. Average asperitycompression over an image, , due to the applied down force can then be estimated using the standarddeviations of the slurry film thickness at 0 kPa and 70 kPa as follows:

,(1)

Figure 4. A profilometer scan of the 76.2 mm region atop which the wafer sits. The zoomed-in image is the 2.5 mm viewing window that DELIF captures.

Figure 5. Histograms of static images at 0 kPa and 70 kPa down-forces at the same location on the polishing pad. The means, 70 and 0, and the standard deviations,  and 0, were used to calculate fluid displaced and asperity compression, respectively.

where and are the standard deviations of the 70 kPa and 0 kPa cases, respectively. By comparing the mean intensity values for the 70 KPa (70) and 0 kPa (0) images taken over the same portion of the polishing pad, the change in slurry layer thickness (T = 70 – 0) can be calculated at that pad location. Figure 6 shows that over the 17 static images, the asperity compression varies linearly with T.

During the dynamic experiments, the down force applied to the etched wafers was approximately 10 kPa. As mentioned earlier, figure 3b is an image of the slurry layer between a patterned wafer with 27 m deep wells and the pad. Means () and standard deviations ()were determined for sub-images denoted by regions 1 and 2 labeled in figure 3b. The standard deviations for both regions were compare to evaluate asperity compression from inside the patterned well to outside the well. Figure 7 shows the distributions of the  values in both regions for well depths of 27 m and 14.5 m. In region 1, the  for both well depths is roughly the same. However in region 2,  for the 27 m deep wells is greater than the 14.5 m deep wells. For both wafers, the asperities seem to be expanding under the well, but the expansion is greater for deeper wells.

Figure 6. Local asperity compression linearly increases with the amount of slurry displaced as70 kPa global down-force is applied to the wafer.

Figure 7. Asperity compression under wafer features during polishing.

CONCLUSION

In this paper we have shown that DELIF is a technique capable of providing measurements of both the slurry layer thickness and pad topography on an asperity size scale. Global pad topography results in widely varying fluid displacement due to applied down-forces to the wafer over small areas (6 mm2). Analysis of asperity compression in static images after applying 70 kPa down-force shows that average asperity compression increases with the reduction of the average slurry layer thickness. Dynamic measurements show that asperity expansion increases with well depth.

REFERENCES

1.A. T. Kim, J. Seok, J. A. Tichy, T. S. Cale, J. Electroche.l Soc., 150 (9), G570 (2003).

2.Y. R. Jeng, H. J. Tsai. J. Electrochem Soc., 150 (6), G348 (2003).

3.J. L. Yuan, B. Lin, Z. W. Shen, J. J. Zheng, J. Ruan, L. B. Zhang, Advances in Abrasive Processes Key Engineering Materials, 202, 85, (2001).

4.D. G. Thakurta, C. L. Borst, D. W. Schwendeman, R. J. Gutmann, W. N. Gill, Thin Solid Films, 366 (1-2), 181, (2000).

5.H. Liang, F. Kaufman, R. Sevilla, S. Anjur, Wear, 211 (2), 271, (1997).

6.J. Copetta, C. Rogers, Experiments in Fluids, 25, 1, (1998).

7.J. Cornely. Master's Thesis, TuftsUniversity, 2003.

8.J. Lu, C. Rogers, V. P. Manno, A. Philipossian, S. Anjur, M. Moinpour, J. Electrochem. Soc., 151 (4), G241, (2004).

9.C. Gray, D. Apone, C. Rogers, V. Manno, C. Barns, S. Anjur, M. Moinpour, A. Philipossian. Electrochem. Solid-State Lett., in press (2005).

10.C. H. Hidrovo, D. P. Hart, Meas. Sci. Tech., 12, 467, (2001).

11.C. Gray, D. Apone, C. Rogers, V. P. Manno, C. Barns, M. Moinpour, S. Anjur, A. Philipossian. CMP-MIC Proceedings, Freemont, CA, Feb. 22-25, 2005.