Development of an Semi-Rigid Abrasive Pad for
Chemical Mechanical Polishing
HO-YOUN KIM, HYOUNG-JAE KIM, HAE-DO JEONG|
School of Mechanical Engineering, Pusan National University
Zip Code 609-735, San 30, Jang-Jun Dong, Geum-Jung Gu, Pusan, Korea
Chemical mechanical polishing (CMP) has been widely accepted for the planarization of multi-layer structures in semiconductor manufacturing. The results of CMP can be optimized by several process parameters such as equipment, pad, carrier film and slurry. However, there have been serious problems in repeatability of results.
CMP plays a key role in semiconductor manufacturing process and is known as most effective method to achieve good global planarization. This CMP process consist of slurry, pad, carrier film, equipment and other process conditions. Among them, slurry and pad are is the most important process parameters. Slurry play an important role in terms of material removal and surface quality of polished wafer. Pad also affects the polishing result significantly such as within wafer non-uniformity(WIWNU) and wafer to wafer non-uniformity (WTWNU) .
However, there are many defects in CMP process, such as dishing, erosion, recess etc., as shown in Figure 1. Many problems such as low breakdown strength, high leakage currents and poor CV behavior in manufactured chip are caused by these defects and should be solved.
Several researchers studied about an abrasive embedded pad to solve these problems . This paper describes the developed semi-rigid abrasive pad and a dishing generating mechanism. MRR(Material Removal Rate) and WIWNU values were used to estimate the performance of the developed pad and results were compared with that of conventional CMP pad.
|2. Dishing and Erosion|
Dishing phenomenon is one of the most serious problems in the copper and tungsten CMP. It is thought that a dishing is caused by material selectivity, pressure distribution and cavitation. Among these reasons, the uneven pressure distribution is considered as important reason of dishing. It causes a deviation of material removal rate, thus the dishing phenomenon becomes grow. Therefore, dishing effects can be reduced by appropriate selection of materials with similar modulus of elasticity and by using semi-rigid semi-rigid abrasive pad.
|2. 1. Modeling of dishing and erosion|
Appropriate analysis model is suggested to investigate the pressure distribution between different materials. The analysis model is composed of dielectric layer and metal layer. Each material's modulus of elasticity is shown in Table 1.
This model is shown in Figure 2. The analysis was performed in the static state. It was assumed that there is no sliding between materials. And, all materials behave within elastic deformation region.
As shown in Figure 2, the model was composed of metal and ILD. To investigate the pressure distribution between different materials, each material should be transferred to same material. In this study, the metal layer was transferred to ILD layer with keeping the height of pattern.
If there is no sliding on the interface, eq. 1 and eq. 2 are inferred when pressure is applied. In this place, it was found that the stress ratio between two different materials is same as that of modulus of elasticity.
The force, Fm, applied to metal is Πmbm. If Fm can be changed to stress of ILD, ΠI, the metal width is turned to the equivalent width, bme(Figure 2). Therefore, ΠI can be gained by eq. 3 and Πm can be determined by eq. 4.
ΠmEbm = ΠIEbme
|2. 2. Pressure distribution on different materials|
It was found in eq. 2 that the uneven pressure between different materials was dominated by the modulus of elasticity of each material. It is independent of the pad properties. In this way, pressure ratio which is applied to TEOS and metal(copper, tungsten) was evaluated. The result is shown in Table 2.
It was found that the pressure of tungsten was 5.3 times and that of copper was 1.73 times greater than that of TEOS. Therefore, the removal rate of metal becomes relatively greater than that of TEOS. Thus, it can be explained that the uneven pressure distribution is one parameter which generates a dishing phenomenon. After dishing generation, this phenomenon is thought to be expanded by cavitation.
|3. Semi-rigid abrasive pad|
It is expected that the defects can be reduced by controlling abrasives, if the uneven pressure and cavitation dominate a dishing phenomenon. Thus, a semi-rigid abrasive pad was developed.
|3. 1. Manufacturing method|
The semi-rigid abrasive pad consists of a polymer binder and CeO2(55vol%) abrasives. The abrasive layer was formed on IC1000/SUBAW(double layer) of the 1o thickness. If the bonding strength of the binder is too strong, scratches can be generated on the wafer. On the other hand, if it is too loose, the wear of the abrasive layer becomes to increase. Therefore, the bonding strength should be optimized. In other words, the bonding strength should be near the friction force generated between the wafer and the pad to prevent both extreme wear of the pad and scratches on the wafer surface, as shown in Figure 3.
The surface of the semi-rigid abrasive pad is shown in Figure 4. The surface structure of pad was observed by scanning electron microscope(SEM). As shown in Figure 4, CeO2 abrasives were coated by binder and abrasives have a diameter with a range of 1GPA 5GPA. The manufactured pad have a 400o diameter and a uniform structure across the whole area.
The properties of the manufactured pad surface were evaluated by a hardness test. In general, the increase of the surface hardness is in its favour in order to increase the selectivity of pattern structure in the device wafer process. The manufactured pad was compared with IC1400 in terms of surface hardness and compression ratio. The hardness was measured at 9 points in the direction of a radius and the value was represented as 89-91(Asker-C). The compression ratio had also a similar tendency with that of IC1400, as shown in Figure 5.
|3. 2. Experimental conditions|
To investigate characteristics of the semi-rigid abrasive pad, new CMP was compared with conventional CMP in the same conditions using TEOS film wafer (10000π), as shown in Table 3. In the conventional CMP, ILD 1300 slurry and IC1400 pad were used. In new CMP using the semi-rigid abrasive pad, DI water and KOH chemical were supplied. The CMP equipment used in this experiment is a machine with one head and one platen.
|3. 3. Result and discussion|
Experimental results of conventional CMP and new CMP using the semi-rigid abrasive pad are shown in Table 4.
As a result, the material removal rate of semi-rigid abrasive pad with DI water was 0.8 times and the rate with KOH was 1.44 times higher than that of conventional CMP. This result is due to a high efficiency machining of the semi-rigid abrasive compared to a free abrasive slurry.
On the wafer surface polished by semi-rigid abrasive pad, some scratches were generated in case of using DI water. But there was no scratch in case of using KOH. It is shown that the semi-rigid abrasive pad can achieve no scratch surface and a high efficiency machining within U region at Figure 3.
To evaluate the non-uniformity for a global planarization, a residual film thickness was measured by elipsometer. The measurement method is 52-points cartesian coordinate with edge exclusion 3o and 1Π. The non-uniformity was obtained using eq. 5.
The global planarization showed a good characteristic in all conditions. It was found that a global planarization can be achieved by the semi-rigid abrasive pad with only DI water and pure chemicals(Figure 6).
However, a path length in rotary CMP have a difference according to each point of pad. This variation of path lengths causes a difference of wear amount and a waveness structure on pad is generated. To solve these problems, a new polishing type and conditioning method must be developed.
This paper introduced the generation mechanism of dishing by uneven pressure distribution between different materials. And, the semi-rigid abrasive pad was developed for scratch-free, high efficiency and echo-process.
Figure 1. Metal and ILD CMP defect type 
Figure 2. Analysis modeling
Figure 3. Relation between bonding strength and friction force
|Figure 4. Surface of semi-rigid abrasive pad|
Figure 5. Surface hardness versus time
(SRAP : Semi-Rigid Abrasive Pad)
Figure 6. Non-uniformity by experimental condition (EE 3o, 1Π)