Comparative Functional Tests of Polishing Pads on CMP Tester Mod. CP-4

Objective

Quantitative comparison of two pads, pad 1 (which showed better field performance) and pad 2 (which showed worse performance in the field).

Test Procedure and Parameters: 

The contact interface was a 6” polishing pad coupon as the lower specimen, rotating at 60 rpm, and a 2” rotating wafer coupon as the upper specimen, laterally oscillating back and forth by 1” with a speed of 5 mm/s (optional in-situ conditioning disc was not used per customer request). Polishing slurry was fed into the interface continuously.
Both the dynamic coefficient of friction COF and energy of contact high-frequency acoustic emission AE were continuously monitored during the tests, while the resistance of top wafer layers was measured periodically. Also, pad wear was monitored in-situ.
Two series of tests were conducted at the same contact pressure of 5 psi:

- with tungsten wafers (W on ILD on Si) and corresponding slurry for 10 min each,

- with copper wafers (Cu on metal on SiC) and corresponding slurry for 5 min each.

Observations:

1. Tests with tungsten wafers yielded the following results:
1a - each of the pads was tested twice with two different wafer coupons, with both COF and AE plots being very repeatable (Fig-s 1 and 2). 
1b - clearly visible several polishing zones: 
- zone of tungsten polish with constant COF and high AE, 
- first transition zone, in which both remaining micro-islands of tungsten and just-exposed micro-areas of a dielectric under-layer were polished, 
- small zone of dielectric polish (its short duration is likely due to the low mechanical polish-resistance of the ILD), 
- second transition zone (with both ILD and substrate micro-areas polished),
- zone of substrate polish, in which COF was constant and AE was negligible.
1c - both electrical surface resistance measurements and visual observations fully confirmed presence of the above zones. 
1d - both transition zones (from W to ILD and from ILD to Si) were shorter for pad 1 than for pad 2, which means that pad 1 produces less non-uniformity on the wafer with mixed micro-areas of different layers.
1e - both W and ILD layers were removed by pad 1 faster than by pad 2 (see Fig. 3).
1f - wear of pad 1(50 micron) was much smaller than wear of pad 2 (130 micron) (see Fig. 4). 

2. Tests with copper wafers confirmed the differences between the pads:
2a - two tests of each pad with two different wafer coupons showed great repeatability for both COF and AE, again (Fig-s 5 and 6), with clear polish zones for different wafer layers and a transition zone of mixed micro-areas.
2b - comparison plot (Fig. 7) shows a difference in rate of material removal between the pads, with pad 1 polishing faster than pad 2.
2c - pad wear (Fig. 8) was about 25 micron for pad 1 and 75 micron for pad 2. 

Conclusions:

1. CMP Tester mod. CP-4 provides for the effective control of functional properties of polishing pads. It can be used for both quality control and incoming inspection.

2. All the parameters measured in-situ, namely COF, AE and pad wear, produced very repeatable and significant differences between the pads, fully confirmed by the end-user in the fab.