Monitoring, Detection and Identification of Metallic Contaminationn in a Production Enviro onment Philippe Maillot R&D Metrology STMicroelectronics Rousset philippe.maillot@st.com
2 Outline Context and Motivation Metal behavior Present industrial mon nitoring Limitation Work axis
3 Context Business mandates aggressive performances / reliability specifications But how much time /money can we afford to minimize yield loss and TTM impact?? Area of interest
4 ncreased sensitivity of devices to race metal contamination Ex: 0% yield loss at 300ppt average metal surface concentration Normalized die fail ratio versus metal X average concentration in at/cm² Die fail ratio 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0. 0.00E+08 6.00E+07 3.00E+08.40E+09 9.00E+08 6.00E+08 8
5 Basic Metal behavior i Diffusion Ex: Fe Segregation Ex: Ti Precipitation Ex: Cu SiO2 Si
6 Basic Metal behavior Implant/Recoil or Diffusion Bulk monitoring best fitted Wet Clean, Plasma Surface monitoring best fitted i SiO2 Si
7 ilution effect! Metal y Slow diffuser Identical contamination level and diffusion cycle 5. 0 9 /cm² 5. 0 9 /cm² Metal x Fast diffuser.00 3 /cm 3 Over 5µm 6.60 0 /cm 3 Over 750µm
8 Bottom line: Potential impact to device depends on Metal type Process Design Technology : Bipolar, CMOS, Imager..
Where to focus on? FEOL! Maximum risk if process step includes interaction of liquid or plasma chemistry with Si or very thin ox Implant/recoil process Second order effect, i.e contamination passed on from previous step must always be considered as well 9 9
Which equipement to focus on? All FEOL equipment Wet benches Resist stripper Oxide/Poly etch Poly/Nitride furnace RTP/RTO Implanters Scrubbers 0
Standard industry monitoring Combination of sweeping TXRF and SPV ( or other lifetime technique..) Some comments to follow..
PV can be blind to slow diffuser Reference Wet bench RCA Anneal (T,,t ) Wet bench RCA Anneal (T 2,,t 2 ) OK Adapt pre SPV anneal nok 2
few alternatives to SPV Good zone ELYMAT Bad zone SPV Ideal case since dataa interpretation between lifetime type techniques can be misleading and requires strong expertise 3 H
TXRF becomes limited in terms of detection limits Move to automated VPD ICP-MS Unit: 0 8 /cm² TXRF ICP-MS witn Manual VPD Ti 2 3 Cr 2 3 Mn 2 0. Fe 2 3 Co 2 2 Ni 2 2 Cu 2 2 Zn 2 2 Mo 0 W 20 ICP-MS with Automated VPD 0.5 0. 2 2 0.05 0.0
Standard monitoring is blind to energetic contaminants Thus implant needs specific bulk VPD monitoring *Maillot/ Roux (STMicroelectronics). 8th international conference on ion implantation technology (IIT200), Kyoto, Japan, June 200 Specie Surface Bulk Calcium Sodium Magnésium Aluminum Potassium Titanium.3 2.9 0.57 27.24 0.65 0.32 3.34.29 0.85 34.67 0.35 2.59 Chromium 0.22 0. Iron.43 0.92 Cobalt ND ND Nickel 0.08 0.4 Copper ND 3.8 Zinc 0.3 0. Molybdenum 0.29 278.9 Tungsten 0.3.45 VPD ICP-MS data in units of 0 0 at/cm²
Check handling equipement! Handling arm induced diffused from backside contamination Frontside SPV Backside TXRF 6
cces to DLTS : a plus for metal ID Reference Low yield area Area of interest Identification of yield impacting metal using on scribe line dedicated DLTS structures Codegoni.D et al (STMicroelectronics), Solid State Phenomena Vols. 45-46 (2009) pp 23-26, 2009 7
Best compromise monitoring Surface Large screening through Sweeping TXRF Additional screening of «key» tools with automated VPD ICPMS Bulk : Post anneal SPV VPD on thick oxide for implant Acces to DLTS for bulk metal identification if killer is not Fe. + Regular backside contamination baseline 8
Industrialization equirements Parameters to be considered Metrology capacity limitation Out-of-control rate versus proven risk and thus downtime Optimized control plan will always be a trade off between risk minimization, cost and productivity. 9
Then Establish whole tool set baseline through VPD ICP-MS, TXRF and SPV Put in place tight regular monitoring of key tools, like class wet benches for instance. Put in place a looser control on rest of the tool set Capability review and optimization of all SPC control limits per tool type based on history, litterature data and ITRS roadmap *Maillot.P et al.,stmicroelectronics, ASMC Conference, 2009 20 2
Quick root cause finding is key! First through metal ID through metrology or analytical Then through in depth work with equipement enginnering and tool supplier Partitioning Check all part dataa sheet ( alloy composition) Least expected source can be where the problem is, check second order effect! 2 2
But. In most cases We have very few data linking trace level contamination and potential yield loss caused by electrical parameter degradation We have limited knowledge of metal behavior through simulation We are very limited in our capability to probe locally 22 2
Thus there are few gaps to fill Correlation study between contamination level and degradation of electrical parameters/yield Modelisation of metal physical behaviour in ICs structures vs process steps Local trace metal identification Analytical techniquee development Indirect : innovative test structures University/Industryy COMET project 23 2
Acknowledgment to the many colleagues at ST and. Thank you for your attention 24 2