Diamonex - Improvements in Oxide CMP Using Diabond Pad Conditioners

Technical Articles

Diamond Bonding | Diamonex CMP PAD Conditioner Product List
Technical Articles | Diamonex CMP Pad Conditioner Application Form (PDF)

Improvements in Oxide CMP Using DIABOND^® Pad Conditioners

J. W. Zimmer - sp³ Corporation, Mountain View, CA
A. B. Stubbmann - DIAMONEX, Allentown, PA

Introduction

Consumables in oxide CMP such as pads, slurry, conditioners, and test wafers represent about 25-30% of the cost of ownership and oxide removal rates have a direct percentage influence. Therefore, any improvements in conditioning of the polishing pads can have a positive impact on the overall cost of ownership if these improvements result in either increased throughput, longer pad lifetimes or reduced maintenance and qualification costs. This paper shows how a simple change in the type of pad conditioner can dramatically impact cost of ownership in a production environment.

Discussion

This experiment consisted of replacing the standard Ni plated diamond pad conditioner on a production IPEC 472 polisher with a DIABOND^® diamond bonded diamond pad conditioner which was engineered specifically for CMP applications. This conditioning disk consists of diamond grit epitaxially bonded to a CVD diamond film to yield a continuous diamond film with abrasive characteristics tailored to work well with a Rodel IC 1000 type pad. The DIABOND^® conditioner was installed as a direct replacement for the existing Ni plated diamond grid type conditioner (end effector) and no process changes were made other than overall polishing time to adjust for any rate increase. The processes tested were a BPSG film and an undoped TEOS film on a Rodel IC 1400 pad with Cabot SC 112 slurry. Results shown in graphs are from unpatterned test wafers with equivalent process film thickness. These are interspersed with production wafers to establish process control points.

The results were both immediate and long lasting. Graph 1 shows polishing rate curves for both a typical DIABOND^® conditioned pad as well as a pad conditioned with a traditional Ni plated diamond grid. The most obvious result is a substantial increase in oxide polishing rate averaging about 30%. The next most obvious feature is the increase in pad lifetime. The normal pad lifetime with the Ni plated conditioner was typically 250 wafers with a substantial rate fall off after 150 wafers. The DIABOND^® conditioned pad allows 300 or more wafers to be processed within the same tolerance window as a 150 wafer run using the original conditioning disk. This represents about double the pad lifetime for the same process window. Pad erosion measurements taken on used pads confirm a 50% reduction in pad wear rate through the life of the pad. This is shown in Graph 4. Graph 2 shows the equivalent comparison of wafer uniformity which shows that there is also an improvement in consistency compared to the traditional conditioning methods. Graph 3 shows that similar results can be obtained with undoped TEOS films without modifying the equipment setup.

These are typical results and are supported by ongoing production use which has reached several thousand wafers across 25 pads with no degradation in the results. Rate stability is within 10% during more than a month of continuous production use. This indicates that the relative lifetime of the DIABOND^® conditioner itself is several times longer than traditional conditioners. Subsequent examination of the used conditioner disk showed that the wearout mechanism is simple abrasive wear on the diamond grit rather than significant grit loss as is the case with traditional nickel plated conditioners.

Graph 1

Graph 2

Graph 3

Graph 4

Conclusions

The simple replacement of a traditional Ni plated diamond pad conditioner with a DIABOND^® diamond bonded diamond conditioner resulted in a substantial improvement in the cost of ownership for this step based on the improvement in pad lifetime, polishing rate, and conditioner lifetime. Doubling the pad lifetime reduces pad costs by 50% and also cuts requalification costs by 50% in both labor, machine time, and test wafers. Labor and requalification time for conditioner disk replacements are reduced by 75% or more based on a longer conditioner lifetime. Polishing rate improvements have an impact directly proportional to the percentage of total cycle time represented by the polishing time. These improvements will also reduce troubleshooting costs related to rate and uniformity problems since the conditioner is now a stable constant across thousands rather than hundreds of wafers.

« Back to Technical Articles

A Division of Morgan Advanced Ceramics
Morgan Technical Ceramics is a Global Business Unit of the Morgan Crucible Company plc.
It comprises Morgan Advanced Ceramics and Morgan Electro Ceramics.