Arcing prevention by dry clean optimization at Shallow Trench Isolation (STI) Etch in AMAT MxP by use of plasma parameters

Transcription

1 Page 1 Arcing prevention by dry clean optimization at Shallow Trench Isolation (STI) Etch in AMAT MxP by use of plasma parameters nd AEC/ Conference Europe, April 18 th -20 th 2001, michael.hartenberger@tu-cottbus.de lars.christoph@infineon.com andreas.steinbach@infineon.com

2 Acknowledgement The authors of this poster would like to thank Siegfried Bernhard Ines Uhlig Astrid Kretschmar Thomas Kuenzelmann, Germany The work for this paper was supported by the EFRE fund of the European Community and by funding of the State Saxony of the Federal Republic of Germany, project, project number The authors are responsible for the content of the paper. Page 2 2 nd AEC/ Conference Europe, April 18 th -20 th 2001, Acknowledgement

3 Outline Page 3 Introduction Shallow Trench Isolation (STI) etch process Arcing at STI etch in AMAT MxP chamber SEERS measurement technique Development of a new dry clean using plasma parameters for real time process monitoring Basic experiments (not shown in this publication) Optimization of clean step Optimization of conditioning step Optimization of clean-/conditioning time Long term evaluation Summary 2 nd AEC/ Conference Europe, April 18 th -20 th 2001, Outline

4 STI etch process - Overview a) Starting profile b) Resist ARC SiN Mask-open step Resist ARC SiN c) Trench etch step Resist ARC SiN Si Si Si deep trench d) Oxide CVD e) Planarization f) A-B: top-cd Oxide Oxid A SiN SiN Si Si Oxide bottom-cd Page 4 2 nd AEC/ Conference Europe, April 18 th -20 th 2001, B Introduction - STI etch process

5 Arcing what s that? Arcing in terms of technology: = higher particle density caused by aluminum particles Arcing in terms of physics: = temporary local instability of plasma Fig.: Arcing generated Aluminum particle on wafer surface Page 5 2 nd AEC/ Conference Europe, April 18 th -20 th 2001, Arcing at STI etch in MxP

6 Why Arcing at STI etch in our AMAT MxP chambers appeared Basic reasons for arcing: Hardware defects (i.e. scratches) Normal waste of anodization Potential differences between chamber parts Critical process conditions (i.e. high RF power, polymers) Main arcing reason at STI etch in our MxP chambers: Change of product mix: a new product uses a high polymerizing etch chemistry The usual dry clean was not efficient enough for the new process mix, thicker polymer layers at the chamber wall Arcing! Page 6 2 nd AEC/ Conference Europe, April 18 th -20 th 2001, Arcing at STI etch in MxP

7 How to get rid of arcing at STI etch in AMAT MxP chamber? Extending clean time or increasing gasflows for the standard clean recipe is not a viable solution. It caused two effects: Less arcing, yes. But also a very strong first wafer effect on critical dimensions of product wafers, which could not be tolerated! Development of a new dry clean recipe was necessary: With effective polymer reduction on chamber walls. And with small tolerable CD degradation on product wafers. Development of a new dry clean using a different gas mixture. Page 7 2 nd AEC/ Conference Europe, April 18 th -20 th 2001, Arcing at STI etch in MxP

8 Strategy to develop a new dry clean 1. Basic experiments: Test of other gas mixtures, e.g. CF 4 /O 2, using resist test wafers Target: High polymer etch rate 2. Evaluation of the new dry clean on product wafers: Page 8 Target: First wafer effect (critical dimensions!) negligible. 3. Long term evaluation of new dry clean on product wafers: Target: New dry clean must be efficient enough Optimization between long term clean effect and first wafer effect on critical dimensions For the efficient development of the new dry clean we used a quick and dirty real time in- situ process monitor - SEERS. 2 nd AEC/ Conference Europe, April 18 th -20 th 2001, Development of new dry clean

9 Principle of SEERS RF RF current RF RF voltage Page 9 FFT FFT Model SEERS Electron collision rate rate Electron density Bulk Bulk power DC DC bias bias voltage 2 nd AEC/ Conference Europe, April 18 th -20 th 2001, Self Excited Electron Plasma Resonance Spectroscopy Passive electrical method Integral physical parameters Development of new dry clean

10 Experiments to optimize the new dry clean The new CF 4 /O 2 dry clean recipe consists of two parts (like old one): 1. Clean step to remove the polymer 2. Conditioning step to minimize the first wafer effects on the following product wafers Experiments: Optimization of clean step time, without subsequent conditioning step, using resist test wafers to monitor first wafer effect Application of standard conditioning step, using product wafers Fine tuning of conditioning time / recipe parameters on product wafers Long term evaluation of new dry clean Targets of all these optimization experiments: Maximize the polymer reduction at chamber wall Minimize the first wafer effect on critical dimensions Page 10 2 nd AEC/ Conference Europe, April 18 th -20 th 2001, Development of new dry clean

11 Optimization of clean step time (without conditioning step) using test wafers median electron collision rate [s -1 ] electron collision rate vs. time [median] 3.00E E E E E E E+00 Clean Resist Dummy 14:52 15:21 15:50 16:19 16:48 17:16 17:45 time electron collision rate vs. time [median] E E E E E E+07 Clean 1 is longer than Clean 2 median electron collision rate [s -1 ] 3.00E E E E E E E+00 15:21 15:50 16:19 16:48 17:16 17:45 Clean step is performed without following conditioning step to monitor the impact of the cleaning process on chamber conditions. First wafer effect depends on cleaning process step time: time 4.80E E E E E E+07 Dry clean 1: first wafer effect on following resist dummies Dry clean 2: first wafer effect on 1 following resist dummy only Page 11 2 nd AEC/ Conference Europe, April 18 th -20 th 2001, Optimization of clean step

12 Conclusions from optimization of clean step time Dry clean 2: Has a significant lower impact on chamber conditions. Therefore we decided to use it for further optimization. Conditioning step after cleaning process: Is applied to remove the first wafer effect completely. But in case of to much conditioning the whole clean becomes inefficient, arcing could appear again. Finally long term tests of new cleaning process are necessary. Page 12 2 nd AEC/ Conference Europe, April 18 th -20 th 2001, Optimization of clean step

13 Application and optimization of conditioning step Page 13 median electron collision rate [s -1 ] electron collision rate vs. time [median] 6.00E E E E E E+07 Product Clean 3.00E+07 17:45 18:57 20:09 21:21 time 3.90E E E E E+07 2 nd AEC/ Conference Europe, April 18 th -20 th 2001, top-cd top-cd vs. time [mean] 17:45 18:14 18:43 19:12 19:40 20:09 20:38 21:07 Squares in diagrams represent first product wafer after each dry clean Electron collision rate indicates first wafer effect on product wafers Electron collision rate also indicates drift of chamber conditions (decline of average represented by falling line in diagram) First wafer effect is also indicated by critical dimensions, measured on product wafers time Optimization of conditioning step

14 Conclusions from application of conditioning step Page 14 Tested on product wafers First wafer effect, caused by dry clean, is indicated by electron collision rate. Additionally electron collision rate indicates drift of chamber conditions (declining average). First wafer effect is indicated by critical dimensions as, measured on product wafers. Chosen dry clean process with standard conditioning step has still significant impact on critical dimensions at STI etch. Conditioning time and / or conditioning recipe has to be changed to reduce influence on critical dimensions. 2 nd AEC/ Conference Europe, April 18 th -20 th 2001, Optimization of conditioning step

15 Modified conditioning step recipe electron collision rate vs. time [median] 3.80E+07 top-cd vs. time [mean] median electron collision rate [s -1 ] 3.60E E E E+07 top-cd [nm] Page E+07 14:52 15:21 15:50 16:19 16:48 17:16 17:45 18:14 18:43 19:12 time DRAM product Squares in diagrams represent first product wafer after each dry clean Now conditioning step has no CF 4 /O 2, but clean-step is still with CF 4 /O 2 No significant first wafer effect after each dry clean Now dry clean does not seem to have a big influence on critical dimensions & electron collision rate But: is dry clean still as effective? (Polymers are generated in the conditioning step and without CF 4 / O 2 even more possibly the clean effect is spoiled by the following conditioning step?) 2 nd AEC/ Conference Europe, April 18 th -20 th 2001, 14:52 15:21 15:50 16:19 16:48 17:16 17:45 18:14 18:43 19:12 time Optimization of conditioning step

16 Clean- and conditioning time optimization long vs. short conditioning electron density vs. time [median] electrondensity vs. time [median] long clean / long conditioning long clean / short conditioning 5.00E E E E+09 median electron density [cm -3 ] 4.60E E E E E+08 14:24 14:52 15:21 15:50 16:19 16:48 17:16 time 1.96E E E E E+09 median electron density [cm -3 ] 4.60E E E E E E+09 14:24 14:52 15:21 15:50 16:19 16:48 17:16 Squares in diagrams represent first product wafer after each dry clean Optimization from economic point of view: Influence of cleaning- and conditioning time on chamber conditions Longer conditioning time does not seem to cause changes in trend of electron density for following resist wafers here time 1.96E E E E+09 Page 16 2 nd AEC/ Conference Europe, April 18 th -20 th 2001, Time optimization

17 Clean- and conditioning time optimization long vs. short clean median electron density [cm -3 ] 5.00E E E E E+08 electrondensity vs. time [median] long clean / short conditioning 3.00E E+09 14:24 14:52 15:21 15:50 16:19 16:48 17:16 time 2.00E E E E E+09 median electron density [cm -3 ] 5.00E E E E E E+08 electrondensity vs. time [median] short clean / short conditioning 17:24 17:52 18:21 18:50 19:19 19:48 20:16 Squares in diagrams represent first product wafer after each dry clean Shorter clean time reduces first wafer effect on following resist test wafers But shorter clean time means less polymer reduction at chamber walls! time 2.00E E E E E E+09 Page 17 2 nd AEC/ Conference Europe, April 18 th -20 th 2001, Time optimization

18 Long term evaluation of dry clean 5.00E+07 Electron Collision Rate [Median] vs. Date (IT-Mask-Step) 2.00E+07 Page 18 Median Electron Collision Rate [s -1 ] 4.50E E E E E E E+07 Dry clean DRAM product 1.00E Date Long term evaluation of new dry clean in normal production shows no significant influence of dry clean on electron collision rate of DRAM- product wafers Only normal scattering (maybe because of pre- process tolerances) in electron collision rate 2 nd AEC/ Conference Europe, April 18 th -20 th 2001, E E E E E E E E E E+00 Long term evaluation

19 Dry clean impact on critical dimensions - Comparison of old and new dry clean CD width CD width vs. wafer Old dry clean Critical dimensions in spec wafer DRAM product lower Spec upper Spec New dry clean Critical dimensions in spec too New dry clean has no significant influence on critical dimensions CD width CD width vs. wafer wafer DRAM product lower Spec upper Spec Page 19 2 nd AEC/ Conference Europe, April 18 th -20 th 2001, Long term evaluation

20 Summary I: Benefit of plasma parameters Electron collision rate and electron density have been used to optimize a new clean recipe at Shallow Trench Isolation etch in AMAT MxP chamber. Both plasma parameters show in real time Dry clean impact on chamber conditions and first wafer effect on product wafers And superimposed long term drift effects of chamber conditions Because plasma parameters do indicate chamber condition drifts, they can be used to monitor cleaning efficiency with respect to wall polymers. Therefore plasma parameter measurements can significantly help to improve efficiency & reduce costs of dry clean process development. Page 20 2 nd AEC/ Conference Europe, April 18 th -20 th 2001, Summary

21 Summary II: Dry clean optimization Benefits of new dry clean: Has no significant impact on critical dimensions of following product wafers (as shown in short- and long- time observations) Is shorter than old dry clean higher throughput Better clean efficiency than old dry clean (as observed during chamber opening for maintenance purposes) Chamber up time has been extended by longer dry clean period All targets of dry clean optimization (maximize polymer reduction from walls and minimize first wafer effect on critical dimensions) have been reached in short period of time using in-situ plasma monitoring techniques. Risk of arcing reduced. Page 21 2 nd AEC/ Conference Europe, April 18 th -20 th 2001, Summary