USING OPTICAL EMISSION SPECTROSCOPY TO IMPROVE EQUIPMENT UPTIME FOR AN AL2O3 ALD PROCESS * JOHN LOO Samsung Austin Semiconductor 12100 Samsung Blvd. Austin, Texas 78754 * Presented at the AVS 5 th International Conference on Atomic Layer Deposition August 8-10, 2005, San Jose California
Installed OES unit with heater blankets
Typical Lightwind Emission Spectrum - ALD Process 5000 Wavelength05 4000 Signal Intensity 3000 2000 Wavelength03 Wavelength04 Wavelength01 Wavelength02 Wavelength06 1000 0 200 300 400 500 600 700 800 900 1000 Wavenength (nm) Typical emission spectra of an Al2O3 ALD process
Wafer01 Wafer02 Wafer03 Wafer04 Wavelength Intensity Wafer Processing Setpoint Window Typical monitoring of chamber affluent wavelength
SYSTEM INFORMATION Communication Link To Process Tool Process Chamber Optional Data System Optical Fiber Multi-channel Optical Emission Spectrometer Downstream ICP Plasma Emission Source (Sensor) Interface To Pump Lightwind Chemical Monitoring System RF Generator Interface Data Acquisition & Control Computer The Lightwind system monitors process chemistry via plasma-induced Optical Emission Spectroscopy (OES).
Wavelength Intensity Wavelength time trend of a long chamber coat. There is a distinct shape of the trend which indicates that the chamber is ready for post chamber maintenance release.
1st 5K Coat 1st 5K Coat 1st 5K Coat 1st 5K Coat 1st 5K Coat 1st 5K Coat 1st 5K Coat 2nd 5K Coat 2nd 5K Coat 2nd 5K Coat 2nd 5K Coat 2nd 5K Coat 2nd 5K Coat 2nd 5K Coat wavelength01 OH=309nm 600 550 500 450 wavelength01 OH=309nm 400 350 Wavelength time trend of two successive short coats ( ½ the time of the long coat ). This shows a repeatable seasoning curve in the same process chamber. Keeping this signature in mind a higher gas flow coat was attempted.
Improper trend indicating seasoning did not properly occur 1-5 2 5K Coats wavelength01 OH=309nm Proper trend indicating seasoning is complete 390 370 350 330 310 290 270 250 wavelength01 OH=309nm 1st 5K 1st 5K 1st 5K 1st 5K 1st 5K 1st 5K 1st 5K 2nd 5K 2nd 5K 2nd 5K 2nd 5K 2nd 5K 2nd 5K 2nd 5K 49pt Contour Map performed after coating 49pt Contour Map performed after coating
DALD02-5 100cc TMA Ar Bubbler Flow Coat 6min 4500 4000 6 minutes 3500 3000 2500 2000 1500 wavelength01 OH=309nm Wavelength 1 1000 500 0 1 40 79 118 157 196 235 274 313 352 391 430 469 508 547 DALD02-5 5K Coat After the 6min 100c Coat These charts indicate that the same seasoning curve can be attained with a simple change to the coating recipe. 1500 3 hours 1300 1100 900 700 wavelength01 OH=309nm Wavelength 1 500 300 1 1199 2397 3595 4793 5991 7189 8387 9585 10783
DALD02-5 5K Coat After the 6min 100c Coat 1500 1300 1100 900 Wavelength 1 wavelength01 OH=309nm 700 500 6 hours of Chamber Coating DALD02-5 100cc TMA Ar Bubbler Flow Coat 6min 300 1 1199 2397 3595 4793 5991 7189 8387 9585 10783 3 hours of Chamber Coating 4500 4000 3500 3000 2500 2000 wavelength01 OH=309nm 6 min of Chamber Coating 1500 1000 500 0 1 40 79 118 157 196 235 274 313 352 391 430 469 508 547 By understanding the response from the chamber affluent during chamber coating, qualification time ( time required to prepare a processing chamber for production material processing ) can be reduced from 6hours to 6min.
Chemical Supplier A Chemical Supplier B Wavelength Intensity Wavelength01 Wavelength02 Wavelength03 Wavelength04 In this example, precursor chemical suppliers were compared by overlapping full spectra. Critical wavelengths were previously indentified as shown above. Using OES provided additional information regarding the precursor chemical composition from both suppliers.
These charts show how OES monitoring of chamber affluent can detect changes to the reaction liquid fill level. As the bottle level drops, the wavelength intensity decreases. Bottle Fill = 100g Reference Line Bottle Fill Level = 63g Bottle Fill Level = 43g Bottle Fill Level 20g
Equipment A DALD03-5 700 600 500 400 Pre Wet Clean Wet Clean Release Trap install Trap Removed ALO_0039.bch PM5_Wavelength1.obs.avg ALO_0039.bch PM5_Wavelength8.obs.avg ALO_0040.bch PM5_Wavelength1.obs.avg 300 ALO_0040.bch PM5_Wavelength8.obs.avg 200 ALO_0042.bch PM5_Wavelength1.obs.avg 100 ALO_0042.bch PM5_Wavelength8.obs.avg 0 2 nd window install 2004-12-14-00:00:00 2004-12-19-00:00:00 2004-12-24-00:00:00 2004-12-29-00:00:00 2005-01-03-00:00:00 2005-01-08-00:00:00 Here OES monitoring was used to evaluate a new pump trap design. Shortly after installation, the intensity of the monitored wavelength decreased leading to an early detection of a trap failure.
The use of OES provides critical information regarding processing system integrity before and after a schedule maintenance. Here is an example whereby OES data was utilized after a precursor bottle fill. This coupled with an in-situ precursor fill system drastically reduced the total time required. Bottle Removal PrePurge ( 4 hours ) Bottle Removal ( 1 hour ) New Bottle Post Purge ( 1 hour ) Qualification Thk/PCs (.5 hour ) Pre Fill Tool Setup ( 2 min ) Auto Fill Sequence Includes 2 purge sequences ( 20 min ) Qualification on 1 st two warmup wafers after PM release ( 8 min) By switching to an AutoRefill System and then confirming chamber integrity post the maintenance with OES data, total time reduced by 6 hours.