Reliability of a Semiconductor Power Switch in a Power Electronics Converter Krishna Shenai, PhD Senior Fellow Computation Institute, The University of Chicago, Chicago, IL (USA) Adjunct Professor EECS Department, Northwestern University, Evanston, IL (USA) July 25, 2018 IEEE PELS DL @ SFBAC
Today s Topics for Discussion Current methodology used to design power converters Current approach used to assess semiconductor power switch reliability in a power converter Case study: Lessons learned from extensive fieldreliability investigation of high-density power supplies Silicon vs. Wide Bandgap (WBG) power devices Future of WBG power devices Moving forward: How to design power converters with built-in field-reliability
What is Field-Reliability? Current reliability assessment methods are only good to evaluate infant mortality
Field-reliability of a power converter is among the least understood topics today! We are not able to design power converters with built-in field-reliability!
What is Needed? Information vs. Power Processor Information Age Energy Age MTBF > 100,000 hrs. MTBF > 1,000,000 hrs. Cost Increased switching speed at reduced power consumption Increased energy efficiency with smaller profile Cost End-of-Life = 17 years End-of-Life = 125 years
Trends in Power Conversion V SUP L O A D i L (t) Increased current density, higher switching frequency, and higher T jmax Power Converter Higher system integration Increased power density Increased cooling density Smaller converter, lower cost and higher efficiency
Today s Power Converter Design Approach K. Shenai et al, Proceedings of the IEEE, vol. 102, no. 1, pp. 35-52, Jan. 2014
Power Chip in an OEM T jmax to T a AMBIENT (T a ) OEM Power Converter Power Module Chip T jmax Miniaturization demands high-frequency power conversion and system integration
Today s Semiconductor Power Switch Reliability Assessment Approach High-Temperature Reverse Bias (HTRB) Test High-Temperature Gate Bias (HTGB) Test Temperature Humidity Bias (THB) Test Thermal Cycling (-40 C to 125 C) Power Supply Operating Life Test
The HTRB Test Test performed to accelerate failure mechanisms Typical stress conditions are: - T a = 125 C to 150 C - V dc V BR Test duration ~ 1000 hours Failure rate (λ) is estimated by considering the dependence on temperature (T), relative humidity (RH) and electric field (E) λ = A e φ kt e B RH e CE
EPC egan FET Reliability - Example R. Strittmatter et al, EPC egan FETs Reliability Testing Phase Six, www.epc-co.com, 2014
Reliability of High-End Computer Server Power Supplies a Case Study
Cost ( /W) Efficiency (%) MTBF (a.u.) Cooling Cost (a.u.) Failures of Computer Supplies IBM eserver 900 30% volume taken by PS 10% volume by cooling ROAD BLOCK! Power loss in components and packaging! 10 kw Power Supply 50 khz to 75 khz = 50% reduction in size 12 90 100 4 8 85 10 2 4 Si 80 1 1 50 100 150 Frequency (khz) 1 2 3 Power Density (a.u.) P. Singh et.al., IBM J. Res. & Dev., Nov. 2002
At UI-Chicago (1995-2004) Founded and directed world s first industryuniversity-government consortium to improve power supply reliability (1998-2004) K. Shenai, IEEE Spectrum, vol. 37, No. 7, pp. 50-55, July 2000 (invited paper)
Power Supply Arcing P. Singh et al, IEEE APEC Digest, pp. 225-229, 2001
Power Supply Arcing Partial Vacuum Test identifies location of failure P. Singh et al, IEEE APEC Digest, pp. 225-229, 2001
Power Supply Arcing Zinc Whisker Spray Test identifies minimum spacing between features P. Singh et al, IEEE APEC Digest, pp. 225-229, 2001
Safe Operating Area (SOA) Degradation Zero voltage transition (ZVT) boost converter High turn-off dv/dt and avalanche stress on Q 2 Degradation of transfer curve of Q 2 with time Degradation of SOA of Q 2 with time N. Keskar, M. Trivedi and K. Shenai, IEEE IAS Digest, pp. 1098-1102, 1999
Silicon MOSFET Failures due to Dynamic Avalanching N. Keskar, M. Trivedi and K. Shenai, IEEE IAS Digest, pp. 1639-1645, 1999
Accelerated HTRB Stress Test Good MOSFET K. Shenai, 12 th Annual Automotive Reliability Workshop, Nashville, TN, May 2007 Bad MOSFET
Power MOSFET Failure: Effect of Die Size SEB Failure Rates of 1000V Silicon Power MOSFETs Actual measured data down to 83% of rated voltage No data for lower stresses due to very low failure rates Low-level leakage results in significant device de-rating. K. Shenai, 12 th Annual Automotive Reliability Workshop, Nashville, TN, May 2007
Field-Failures of Power MOSFETs in Power Supplies Residual material defects in silicon caused field-failures of power MOSFETs in high-end server power supplies K. Shenai, IEEE NAECON 2010, Dayton, OH, July 2010
Silicon IGBT Failures During Short-Circuit and Inductive Switching Conditions Simultaneous High-Voltage and High-Current Situation
Short Circuit Failure in Silicon IGBTs 600V/50A punch-through IGBT Failure after 18 μs M. Trivedi and K. Shenai, IEEE Trans. Power Electronics, vol. 14, no. 1, pp. 108-116, Jan. 1999
Silicon IGBT Failure Under Clamped Inductive Stress M. Trivedi and K. Shenai, IEEE Trans. Power Electronics, vol. 14, no. 1, pp. 108-116, Jan. 1999
Physics of Hot Spot Formation Poynting vector S = E x H Dipole radiation pattern Electric field strength (color) Poynting vector (arrows) Energy conservation law δu = grad S J. E δt u = electromagnetic energy density u = 1 2 (E.D + B.H) K. Shenai et al, IEEE Proceedings, vol. 102, no. 1, Jan. 2014, pp. 35-52
Si vs. WBG Power Devices Future of WBG Power Devices
System-Level Benefits of WBG Power Devices 50,000 cm 3 18 kg WBG Power Switch Silicon Power Switch Shenai s Figure of Merit - Q F2 A E M 2400x improvement 4,500 cm 3 0.2 kg for Increased energy savings Reduced system cost Robust & reliable system WBG Power Switch Shenai et al, IEEE TED, pp. 1811-1823, 1989 At the system level, the objective should be to increase power and cooling densities
Smaller Chip : Lower Cost
Si vs. WBG Material Properties
W H Y W B G? Increased Energy Efficiency Replace Si with WBG Smaller Converter Profile Increased switching frequency and system integration Higher Junction Temperature Improved package and thermal management Wide bandgap (WBG) semiconductors, such as SiC and GaN devices, offer superior electrical and thermal performances compared to silicon K. Shenai et al, "Optimum Semiconductors for High-Power Electronics," IEEE Trans. Electron Devices, vol. 36, no. 9, pp. 1811-1823, September 1989.
Status of Commercial WBG Power Devices Vertical SiC JBS Power Diodes ( 300 V < V BR < 1700 V) Lateral Low-Voltage ( V BR < 650 V) GaN Power Transistor Vertical High-Voltage ( 900 V < V BR < 1700 V) SiC Power MOSFET Cost of WBG device is 2-3X higher than that of Si device Circuit design complexities Gate driver issues System-level benefits of WBG devices are minimal WBG device field-reliability is unknown
SiC Wafers Current & Future Parameter Silicon SiC Growth Temperature < 1000 C > 2000 C Method Czochralski PVT Defect Density < 1/cm 2 Very High Cost Low Very High
Defect Engineering of 4H-SiC Wafers BPD Density (cm -2 ) TSD Density (cm -2 ) 45000 40000 4500 4000 35000 3500 30000 3000 25000 2500 20000 2000 15000 1500 10000 1000 5000 500 0 0 2000 2002 2004 2006 2008 2010 2012 2014 2000 2002 2004 2006 2008 2010 2012 2014 TED Density (cm -2 ) 5000 4500 4000 3500 3000 2500 2000 1500 1000 500 0 2000 2002 2004 2006 2008 2010 2012 2014 45000 40000 35000 30000 25000 20000 15000 Dislocation Density (cm -2 ) BPD TSD TED 10000 5000 0 2000 2002 2004 2006 2008 2010 2012 2014
Defects in State-of-the-Art Commercial 4H-SiC Wafers Micropipe s (a) (b) (c) High resolution synchrotron monochromatic X-ray topographs recorded at Argonne s Advanced Photon Source (APS) facility. (a) Back-reflection X-ray topograph (g = 0004) images of close-core threading screw dislocations (TSDs) and basal plane dislocations (BPDs) in a (0001) 4H SiC wafer; (b) Grazing incidence X-ray topograph (g = 11-28) of 4H-SiC substrate showing TSDs (right and left handed) and TEDs; (c) Transmission X-ray topograph showing the images of BPDs.
c-axis Formidable Material Challenge Growth in the c-axis direction, enabled by screw-dislocations providing steps! SiC seed Vertical (c-axis) 4H-SiC boule growth proceeds from top surface of large-area seed via hundreds to thousands of threading screw dislocations (TSDs). After Ohtani et al. J. Cryst. Growth 210 p. 613. Threading screw dislocation growth spirals (THE sources of steps for c-axis growth) found at top of grown 4H-SiC boule. Contention: Elimination of screw dislocations from power devices not possible while maintaining commercially viable crystal quality and growth rate and via this approach. Crystal grown at T > 2200 C --> High thermal gradient & stress --> More dislocations
Role of Crystal Defects on the Electrical Characteristics of PiN Diode
Avalanche Testing of Power Devices First proposed by Shenai C. S. Korman et al, in Dig. Int. High-Frequency Power Conversion, pp. 128-139, 1988
Measured Avalanche Energy (E AVL ) of Power Diodes 600V/8A 4H-SiC JBS diode 600V/6A silicon MPS diode Why E AVL of SiC Power Diodes < E AVL of Si Power Diodes? Most WBG data sheets do not list E AVL First demonstrated by Shenai K. Shenai et al, Proc. IEEE, Feb. 2014
Peak Diode Current (A) Measured dv/dt of SiC Power Diodes 6 SW2 PD2 4.5 Failure Instant 3 1.5 T c = 25 C 0 0 15 30 45 60 dv/dt (V/nsec.) SW2: 600V/6A 4H-SiC JBS diode PD2: 600V/8A silicon MPS diode SD1: 300V/10A 4H-SiC JBS diode SW1: 200V/12A silicon MPS diode Why dv/dt Capability of SiC Power Diodes < dv/dt Capability of Si Power Diodes? Most WBG data sheets do not provide dv/dt ratings First demonstrated by Shenai s group K. Acharya and K. Shenai, Power Electronics Technology, pp. 672-677, Oct. 2002
Safe Operating Area (SOA) of Power MOSFETs IXFB30N120P Why SiC SOA is smaller than silicon? C2M0080120D K. Shenai et al, Proc. IEEE, Feb. 2014
4H-SiC Material Defects and MOS Gate Oxide Reliability K. Yamamoto et al, Influence of threading dislocations on lifetime of gate thermal oxide, Mat. Sci. Forum, vols. 717-720 (2012), pp. 477-480.
Commercial SiC JBS Diodes Measured @ 25ºC Measured @ 25ºC Punch-through Design Silicon Avalanche breakdown Minimum on-resistance SiC Punch-through (leaky) Not optimized Too much fat left in SiC diodes K. Shenai and A. Chattopadhyay, IEEE Trans. Electron Devices, vol. 62, no. 2, pp. 359-365, Feb. 2015.
Field-Induced Lattice Deformation in 600V 4H-SiC JBS Diode Defect delineation study performed using hard X-rays at Argonne s Advanced Photon Source (APS). At 900V reverse bias, TSDs in the vicinity of the metalsemiconductor junction were excited and acted as charge generation centers that led to diode breakdown. Collaborators: Stony Brook University Brookhaven National Labs K. Shenai unpublished work, 2014
Lessons from the Past: Higher chip cost and thermal limitations rendered GaAs chip technology always a technology of the future
Why Power Electronics Converters Fail in the Field?
Why Power Supplies Failed in the Field? Output capacitor leakage; reactive charge dumping from transformer leakage inductance; power supply arcing caused by zinc whiskers K. Shenai, IEEE Spectrum, vol. 37, No. 7, pp. 50-55, July 2000 (invited paper)
Failures in Electronic Systems What is the Junction Temperature T j?
Industry Response Question: Please indicate which components you consider most important to be addressed by future research to improve the reliability of power electronics converter systems? J. Falck al, IEEE Industrial Electronics Magazine, vol. 12, no. 2, June 2018, pp. 24-35
Industry Response Question: Please rank the following options for achieving high reliability for power electronics systems? J. Falck al, IEEE Industrial Electronics Magazine, vol. 12, no. 2, June 2018, pp. 24-35
Physics of Failures What is the role of material defects on cost, performance and reliability of a semiconductor power switch?
Power Semiconductor Switch v SUP L O A D i L (t), di L (t)/dt Load is inductive off C A on B v AB (t), dv AB (t)/dt Low on-state resistance (R DS(ON) ) - to reduce conduction power loss (I 2 R DS(ON) ) Low capacitances - to reduce switching power losses (CV 2 f) Good reliability Low chip cost
Losses in a Power MOSFET Power converters are designed by considering mainly P ON
Drift-Region Design Gate (G) Source (S) C GS R G R W n + source p-body R D C GD n - epi C DS n buffer n + substrate Drain (D)
On-State Power Dissipation ΔT = T j T c = I2 ON R ON R jc T jmax = 150 C is industry standard > 200 C is desired
Thermal Management of WBG Power Devices GaN (< 1 micron) SiC (few microns) S G D Heat Source 2DEG S G S Silicon Substrate (~ 200 microns) AlGaN Buffer (a few microns) SiC Substrate (~ 350 microns) Lateral GaN Power Transistor D Vertical SiC Power Transistor
Moving Forward: How to design power converters with built-in field-reliability? Industry-Driven Consortium Component Suppliers Inverter & Converter Manufacturer Major OEMs Research Labs OEMs Academia EV Grid Aerospace Computer/Telecom
Need to Account for Material Device System Interactions Top-down systems-driven reliability engineering approach
Systems-Driven Reliability Engineering Cost (T a, MTBF) OEMs Motor Control, Utility Grid, EVs, Power Supplies, etc. Power Density (Cooling Density) Converter Suppliers Inverters & DC-DC Converters Current Density (T jmax & V BD ) Switch & Module Manufacturers Power Semiconductor Chips & Modules Defect Density (Growth Rate & Wafer Size) Material Suppliers SiC & GaN Wafers
QUESTIONS? Thank You
OPTIONAL SLIDES
What is Reliability? Reliability = measure of continuous service accomplishment (or time to failure) Metrics Mean Time To Failure (MTTF) measures reliability Failures In Time (FIT) = 1/MTTF, the rate of failures Traditionally reported as failures per 10 9 hours of operation Ex. MTTF = 1,000,000 hours FIT = 10 9 /10 6 = 1000 Mean Time To Repair (MTTR) measures Service Interruption Mean Time Between Failures (MTBF) = MTTF+MTTR
The Famous Bathtub Curve
Accelerated HTRB Stress Test K. Shenai, 12 th Annual Automotive Reliability Workshop, Nashville, TN, May 2007
Weibull Probability Model
Acceleration Parameters
MOSFET Field-Reliability Model Predicted Failure Rate = Acceleration Factor (AF) x Acceleration Test Failure Rate
SiC MOSFET Gate Oxide Failures
Role of Bulk Material Defects on SiC MOSFET Gate Oxide Reliability