A crystallographic optimization study in Ta TaN bi-layer sputtering to reduce semiconductor manufacturing queue time constraint
The Physical Vapor Deposition (PVD) magnetron sputtering method is the most extensively used technique for depositing metallic thin film in the semiconductor wafer fabrication industry. The PVD equipment manufacturer has specified stringent control, in this case, a queue time restriction between the...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English English |
| Published: |
2022
|
| Subjects: | |
| Online Access: | http://eprints.utem.edu.my/id/eprint/26909/1/A%20crystallographic%20optimization%20study%20in%20Ta%20TaN%20bi-layer%20sputtering%20to%20reduce%20semiconductor%20manufacturing%20queue%20time%20constraint.pdf http://eprints.utem.edu.my/id/eprint/26909/2/A%20crystallographic%20optimization%20study%20in%20Ta%20TaN%20bi-layer%20sputtering%20to%20reduce%20semiconductor%20manufacturing%20queue%20time%20constraint.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | The Physical Vapor Deposition (PVD) magnetron sputtering method is the most extensively used technique for depositing metallic thin film in the semiconductor wafer fabrication industry. The PVD equipment manufacturer has specified stringent control, in this case, a queue time restriction between the Copper (Cu) barrier seed layer and the following Cu electroplating layer deposition processes. This restriction is formed without providing detailed data or reasoning. This is applied as a condition to safeguard the interconnect quality and performance from functionality and reliability performance failures. Consequently, the queue time has imposed plenty of manufacturing challenges. In cases when the PVD and electroplating processes have to be stopped, the queue time cannot be exceeded and this creates constraints that lead to evasive steps that can create a capacity bottleneck that risks failure in the on-time delivery. The objective of this research is to investigate the film characteristics change in Cu electroplating film against queue time impact in order to verify the need to uphold queue time requirement recommended by the tool manufacturer. In pursuit of this, baseline film ageing characteristics that include sheet resistance and microstructure studies over an ageing period are planned out and compared to the peer studies. The comparisons prompted improvement in Ta bi-layer α-Ta crystal texture content which is shown to result in higher Cu (111) crystal texture and promoted sheet resistance (Rs) stability in the interconnect stack. In order to improve the metal film stability, a screening analysis of variance (ANOVA) experiment was conducted. A fractional factorial experiment was planned out to study the impact of TaN bias power, nitrogen flow rate, Cu bias power, and re-sputter layer on both the Cu barrier seed layer and the Cu electroplating stacked on Cu barrier seed layer. The impact study is primarily on film Rs ageing properties. A statistically significant parameter that influences the Rs stability for both layers is re-sputter treatment layer. Microstructural x-ray diffraction (XRD) analysis is done to understand the ageing mechanism. Next, selected screening experiment parameter settings, along with control, have been used in a wide range of queue time intervals between Cu barrier seed and Cu electroplating processes to understand the impact of the queue time on Cu Electroplating film Rs stability. The results show a stable Rs to an interval of 40 hours. It is also learnt that the cell with re-sputter done on TaN layer has stability up to almost four times compares to the standard manufacturing process cell. The application of TaN layer re-sputter has also shown a higher level of α-Ta ratio content with respect to β-Ta which in turn produces more, of the preferred Cu (111) grains. This research went a step further to have a prediction model to minimize Rs mean and nonuniformity changes during ageing by using the TaN re-sputter process in an RSM experiment. TaN re-sputter impact has again shown its dominance over other modeled parameters causing an insignificant model fit for mean Rs. Additionally, a prediction model is successfully generated for Rs non-uniformity at 48 hours ageing period. |
|---|