Slide 1: A Study of Temperature Induced Polonium Diffusion on SRAM SER Performance
A Study of Temperature Induced Polonium Diffusion on SRAM SER Performance Ramya Ramarapu, Richard Wong– Cisco Systems Brett Clark– Honeywell Tian-Jing Shen – GSI Technology
Slide 2: Outline
Outline • Introduction to alpha induced SER • Recent reported research review • Experimental design and data collection • Results &Conclusions
Slide 3: Introduction
Introduction • Recent data show Sn can have alpha emitting Po at higher concentrations below the Sn surface. • Alpha particle induced SER experiment was conducted on 65nm 36Mb SRAM units to measure the soft error rate of the cells exposed to two Sn sources with varying emissivities to evaluate polon
Slide 4: Alpha Induced SER
Alpha Induced SER • The Po decay produces the 5.3 MeV alpha particle. 5.3 MeV α particle • Alpha particle creates electron hole pairs in the silicon • If the current is large enough, storage nodes such as SRAMs can switch states – Causing an error in the state R. C. Baumann, IEEE Trans. Device Mater
Slide 5: Introduction: Pb free solders
Introduction: Pb free solders • Alpha emission increasing in some Sn solders over time • Alpha emission increasing with depth in Sn Alpha Emission vs Depth: 100 μm 80 70 y = 0.86x + 12 R² = 0.98 Counts·khr-1·cm-2 A 60 B 50 C 40 y = 0.22x + 8.4 R² = 0.99 30 20 10 y = 0.019x + 0.78 R² = 0.3574 0 0 20
Slide 6: Polonium Diffusion Reported
Polonium Diffusion Reported • Alpha emission increasing over time from 210Po diffusion • Mechanism confirmed – Doped low alpha Sn with 210Po – Sample heated & compared against room temp control • Diffusion rate strong function of temperature 40 Emissivity vs Time: Sn doped with 210Po Alpha Spectrum
Slide 7: Polonium Transport Conclusions
Polonium Transport Conclusions • Processes utilizing solid/liquid phase changes are susceptible to microsegregation effects. • Polonium diffusion rate increases 230x from 298K to 498K • Polonium present at levels below analytical detection limits presents significant SER risk • Diffusion can increas
Slide 8: SRAM Application Experiment
SRAM Application Experiment • Measure SRAM SER over time – Decapped 36Mb SRAMs • Exposure conditions – Ambient conditions – Ultra low alpha Sn <2 α·khr-1·cm-2 – 99.99% grade non low alpha Sn 2000 α·khr-1·cm-2 • Initial run room temperature • Heated to 100°C for 1 hour • Final run
Slide 9: Experimental Procedure
Experimental Procedure Purpose Runs Alpha Source Verify Rn and neutron SER Dry Run No Sn source SER due to ULA Sn ULA run ULA Sn at 2 α/khr/cm2 SER due to 99.99% Sn 99.99% grade Sn Run LA Sn at 2000 α/khr/cm2 SER due to 99.99% Sn post Bake 99.99% grade Sn LA Sn post bake bake run 2600 α/khr/cm2
Slide 10: Experimental Setup with Sn sources
Experimental Setup with Sn sources 3 boards in the chassis Sockets with Sn sources on top
Slide 11: Baseline Run
Baseline Run • Ambient Rn levels 1.2 Bq/m3 – 40 FIT/Mb • Neutron SER measured at TSL – 240 FIT/Mb • Run duration 744 hours • No SER events observed • Results match with the estimation within statistical error Errors Hrs Devices Alpha Expected Measured emission FIT/Mb(Rn+n) FIT/Mb 0 744 186 none 280
Slide 12: Sn Source Placement
Sn Source Placement • Sn alpha source used for SER measurements. • Sources were made rest on the top surface of the package to minimize the air gap to 0.53 mm and to fit within the socket • Devices at room temperature during experiment Packaged Exposed silicon die Tin alpha Socket SRAM source Silico
Slide 13: ULA Sn Run
ULA Sn Run • Sn sources rest on top of the package • Sources were handled using latex gloves to avoid contamination • Run duration 960 hours • No SER events detected • Results within statistical error Errors Hrs Devices Alpha Expected Measured emission FIT/Mb FIT/Mb a/cm2/khr (Rn+n+a) 0 960 186 <2 <
Slide 14: 99.99% Sn test run
99.99% Sn test run • Source 0.12 mm thick 99.99% Sn sheet • Measured with XIA UltraLo 1800 Alpha detector – 5.3 MeV Po alpha 4N Tin Alpha Emission Spectrum particle present in 70 spectrum 60 Initial • Sn source attached to 50 5.3 MeV 210Po α Counts/hr 40 the base of the ULA 30 source with double 20
Slide 15: Low Alpha Results
Low Alpha Results • 53 events observed in 410 hrs on 62 devices • Calculated Emission: – Low Alpha Sn FIT rate/ (200 FITs/Mb at 1 a/cm2/khr) Errors Hrs Devices Measured Calculated FIT/Mb alpha emission a/cm2/khr 53 410 62 6 *104 300
Slide 16: 99.99% Sn Post Bake
99.99% Sn Post Bake 4N Tin Alpha Emission Spectrum 70 • Low alpha Sn sources 60 Initial were baked 50 Counts/hr 100 °C, 1 hour 40 – 100 C for 1 hour 30 • 256 SER events 20 10 observed on 62 0 1 2 3 4 5 6 7 devices over 315 hrs Energy (MeV) Errors Hrs Devices Measured Calculated FIT/Mb alpha emission
Slide 17: SER Rate Temperature Dependence
SER Rate Temperature Dependence • Alpha spectrum changed significant above 5 MeV after heating – 2x increase in 5-7 MeV range • SER rate increased 7x . • Potential increased SER sensitivity >5MeV 4N Sn source emissivity α/khr/cm2 1-5 MeV 5-5.5 MeV 5-7 MeV 1-7 MeV Initial Alpha 994 348 465 1654 100 C
Slide 18: Conclusions
Conclusions • ULA Sn SER contribution negligible • SRAM sensitivity to alpha induced SER confirmed • Diffusion increased both alpha flux and effective alpha energy impinging on the device • Device SER response alpha energy dependent – 2x increase in high energy alphas, 7x increase in SER rate – Need
Slide 19: Further Study
Further Study • Resolve the 2x increase in flux to the 7x increase SER. • Verify the error rate counts • Cross section the SRAM to investigate the metal layers to the Si. • Simulate various energy alpha particles going through the Si.
Slide 20: Acknowledgements
Acknowledgements • Thank you and acknowledge – Peng Su, David Chen, Tae-kyu Lee, Gnyaneshwar Ramakrishna, Shamil Toma.(Cisco Systems) – John Senechal, Anthony Leung(GSI) – Stuart Coleman(XIA) – Wayne Meyer (Honeywell)