‘Collaborate and thrive’ model enables fast path to high-volume production
08/01/2006
Anticipating high customer demand for its new finer linewidth chips, the Taiwan-based fabless Flash manufacturer, Eon Silicon Solution Inc., asked Dongbu Electronics to be ready to not only drive the shift of its Flash memory designs from the 0.22µm to 0.18µm linewidth, but also achieve a fast ramp to high-volume production.
While Dongbu Electronics had the framework for processing Flash chips at the 0.18µm node, the Korean foundry encountered a number of processing hurdles that it had to overcome to meet the expectations of fabless manufacturer Eon Silicon Solution. Moreover, the two companies faced potential language and cultural barriers that needed to be quickly surmounted. Despite these challenges, Dongbu and Eon demonstrated how a strong fabless-foundry relationship and a collaborative spirit can drive a win-win scenario.
Minimizing word-line stress
Experiencing lower than expected yields at the onset of Flash memory processing in early 2005, the two companies worked together to isolate the problem to an abnormally high ‘word-line stress’rate. This rate compares closely to reliability metrics applied to a typical Flash cell. Action was taken to modify the cell profile. Significant differences were seen in test results depending on equipment used. By using tunnel oxidation, Pre_CLN and GC CD retargeting, the word-line stress was reduced from 15.8% to 2.92% (Fig. 1).
 Figure 1. Use of tunnel oxidation with Pre_CLN and GC CD retargeting enabled a sharp reduction in word-line stress and a return to higher yields. |
null
Overcoming poly-1 bridging
In Flash memory chips, poly-1 serves as a floating-gate to store information. Bridging between floating-gates can cause a loss of capacitance needed to store data. When Eon experienced high failures in one lot and low failures in another, the varying results were isolated to poly-1 bridging. Since Dongbu processes two wafer lots simultaneously, unwanted particles from one lot affecting the other lot were suspected. Dongbu evaluated all factors that could cause the poly-1 bridging, and quickly determined through checkboard testing (Fig. 2) that the problem was caused by wafer-to-wafer contact on the cleaning module. More specifically, the unwanted particles that caused poly-1 bridging were generated when the front side of one wafer (from lot 1) faced the back side of another wafer (from lot 2) on the cleaning module. By laying the wafers on the cleaning module so that they did not face each other, unwanted particles were not generated, and thus the poly-1 bridging was eliminated.
 Figure 2. Checkboard testing before and after changing wafer placement on the cleaning module shows the wide ranging results caused by poly-1 bridging. |
null
Isolating elusive DC failures
About nine months after Dongbu initiated development for Eon, chip yields began to fluctuate wildly from lot-to-lot. The yield fluctuations were caused by erratic and elusive DC failures, which in turn were found to be caused by cross-contamination (Figure 3).
 Figure 3. Elusive DC failures from cross-contamination were causing erratic fluctuations in yield from lot-to-lot. |
Just as the unusual DC failures occurred, Eon was beginning to see the growing customer demand it had expected for its 0.18µm chips. Driven by the urgency to resolve this problem, the two companies further collaborated to isolate and eliminate the DC failures. Accordingly, Dongbu conducted “dummy processing” of three wafer lots over a short time period. This analysis identified the wafer lot that was the source of contamination. The contaminating lot was then transferred to other equipment for lot processing and contamination removal.
Conclusion
The graph (Fig. 4) shows how processing of FLASH chips at the 0.18µm node was stabilized over time to meet Eon’s criteria. In working closely to resolve processing issues, Dongbu and Eon demonstrated a collaborate-and-thrive business philosophy. Both companies have developed a strong mutual respect and an understanding of cross-cultural differences; close-working protocols have also been established. The two companies have recently announced that their ongoing collaboration has resulted in completing the pure-play foundry industry’s first development of a 130nm NOR Flash memory process.
 Figure 4. Yield stabilization, the ultimate goal of close fabless-foundry collaboration. |
null
Jae Song received his PhD in solution electrochemistry from the U. of Glasgow in 1989 and is EVP of strategic business development at Dongbu Electronics, 32 Dongbu Financial Center, 891-10, Daechi-Dong, Gangnam-Gu, Seoul, Korea; ph 82/2-3484-2470; e-mail [email protected].
UPCOMING EVENT
SEMICON Taiwan
September 11-13, 2006 Taipei World Trade Center Taipei, Taiwan
At SEMICON Taiwan, exhibitors highlight the latest in semiconductor manufacturing equipment, materials, and services. Programs in conjunction with the exposition offer technical and business education addressing manufacturing challenges, the latest technologies, and industry trends.
For more information on programs, contact Ms. Ming Wu at [email protected], ph 886/3-573-3399, ext. 222, or visit www.semi.org.