Testing Conditioning Discs on CMP Tester

 

1. Test Goal

Perform pad wear tests to differentiate between conditioning disks used for CMP.

2. Test Equipment

The CETR CMP Tester was used to characterize the conditioning disks. This stand-alone bench-top CMP machine provides a fully instrumented CMP process with up to 4.25 conditioning disks and up to 9 pads. The tester is capable of precision translational and rotational motions, with speeds ranging from 0.1 to 1,000 rpm. A normal load is applied by a closed-loop servomechanism, and can be kept constant or linearly increasing, ranging from 0.5 mN to 1 kN. Therefore, actual pressures and velocities are achieved for the CMP process, which allows for development, incoming inspection and ongoing functionality testing of pads, slurries, conditioning discs and retaining rings, as well as for CMP process development. Friction force and torque, normal load, pad wear depth and contact acoustic emission can all be measured in-situ (real-time) and recorded at a total sampling rate of 20 kHz. 

A schematic for using the CMP Tester for Conditioner-Pad Tests is shown in Figure 1. 


Figure 1. Schematic of CMP Tester For Conditioner-Pad Tests


3. Test Specimens

The specimens were four 2.5 conditioning disks (made by Abrasive Technology), identified as 7, 8, 9 and 10. The opposing surfaces were 6 pad coupons, one per each test, cut from the same IC pad (made by Rodel) with an easy-to-use handheld mechanical cutter. 

4. Test Procedure and Parameters

The pad coupon was attached to the platen with its adhesive. The conditioning disc was fastened to a suspension holder that in turn was attached to the upper stage. The suspension holder has been designed so that the conditioning disc is always parallel to the pad surface. The upper stage had both rotational and translational motions.

One test per each specimen was performed with simultaneous measurements of the dynamic friction coefficient and pad wear. Test duration varied somewhat between the tests because the goal was to run the tests until approximately 0.030 of pad wear, but no longer than 10 hours; so, some tests were terminated earlier, if the pad wear exceeded 0.030.

A total of four tests were performed. A normal load was maintained constant at 30 N, rotational speeds were 90 rpm for the conditioning disc and 100 rpm for the pad, disc translational speed was 5 mm/s on the stroke of 15 mm. DI water was continuously fed onto the pad with a pump.


5. Experimental Results

The results from all the four tests are given in Figures 2 to 5, where friction coefficient and pad wear are shown during the test period.

For disc 7 (Fig. 2) friction coefficient slowly increased by about 8% for about 5,000 - 6,000 seconds, and very slowly decreased afterwards by about 1%. At first glance, pad wear appeared generally linear over the test, and reached about 0.78 mm after as little as 2.5 hours, which corresponds to a wear rate of 0.32 mm/hr. Closer observation of the data, however, reveals a subtle decrease in slope of the wear data after about 5000 seconds, the same approximate time where friction coefficient leveled off and then began to decrease.

For disc 8, except for the first 1,000 seconds, the friction coefficient also slowly increased by about 17% for about 11,000 seconds, and then remained constant, within 0.1% for the rest of the test (Fig. 3). Within the first 1000 seconds, the COF exhibited a small peak before starting to monotonically increase. The pad wear appeared linear after about 1,500 seconds and reached 0.67 mm after 5.5 hours, which corresponds to a wear rate 0.12 mm/hr. Before 1,500 seconds, the slope of the pad wear appeared less steep than the slope afterwards. 

Figure 2. Friction Coefficient and Pad Wear for Disc 7 Figure 3. Friction Coefficient and Pad Wear for Disc 8

This change in pad wear slope and friction coefficient at early times is even more pronounced for disk 9 (Fig. 4). For this disc, two significant thresholds may be observed, at about 1,800 and 12,000 seconds. At both times, changes in friction coefficient and pad wear are noticeable. During the first 1,800 seconds, friction coefficient was constant and pad wear was negligible. After 1,800 seconds, the friction coefficient increased, and then remained relatively constant until about 12,000 seconds. During the period between 1,800 and 12,000 seconds the pad wear monotonically increased, but at a gradually reduced rate. After about 12,000 seconds, the friction coefficient slowly decreased to the end of the test, and the rate of pad wear first noticeably decreased and then increased by the end of the test. The pad wear reached 0.54 mm after 4.8 hours, which corresponds to a wear rate of 0.11 mm/hr. These parameter changes during a particular test could be due to a variety of factors, such as wearing of pad grooves or disc particles, or pad non-uniformities in either density or elasticity.

Disc 10 (Fig. 5) behaved similar to disc 7 in that friction coefficient slowly increased until about 6,000 seconds, and then remained relatively constant. Pad wear rate was constant up to the same 6,000 second, and then slowly decreased for the rest of the test. Total pad wear was 0.83 mm during 3.0 hours, which corresponds to a wear rate of 0.28 mm/hr.

Figure 4. Friction Coefficient and Pad Wear for Disc 9 Figure 5. Friction Coefficient and Pad Wear for Disc 10

Figures 6 and 7 show comparisons of pad wear and friction coefficient, respectively, for the four tests. The data suggest two groups of the discs, 7 and 10 that produced an average pad wear rate of 0.30 mm/hr, and discs 8 and 9 with the average wear of 0.11 mm/hr. The test results are also consistent in that discs with a higher friction coefficient caused higher pad wear. 

Figure 6. Pad Wear Comparison Identifies Two Disc Groups Figure 7. Friction Coefficient Correlates With Pad Wear

6. Conclusions

1. The bench-top CMP Tester is fully capable of characterizing conditioning discs for CMP, including for detailed QA/QC tests, such as to determine pad wear rate variations either within a manufacturing lot or lot-to-lot.

2. Of the four conditioning disks tested, the data suggested two groupings. Discs 7 and 10 in one group produced an average pad wear rate 2.6 times larger than the average produced by discs 8 and 9 in the other group. 

3. The friction coefficients of the discs 7 and 10 in one group were larger than those of the discs 8 and 9 in the other group. This correlates with the pad wear rates in that larger friction coefficient results in higher wear rate.