High Voltage DC and RF Power Reliability of GaN HEMTs J. A. del Alamo and J. Joh* Microsystems Technology Laboratories, MIT, Cambridge, MA (USA) *presently with Texas Instruments, Dallas, TX (USA) ICNS 2011 Glasgow, July 10-15, 2011 Acknowledgements: ARL (DARPA-WBGS program), ONR (DRIFT-MURI program) Accel-RF Corporation
Breakthrough RF mmw power in GaN HEMTs Micovic, MTT S 2010 94 95 GHz MMIC PAs: Micovic, Cornell Conf 2010 P out >40 W/mm, over 10X GaAs! Wu, DRC 2006 2
GaN HEMTs in the field Counter IED Systems (CREW) 200 W GaN HEMT for cellular base station Kawano, APMC 2005 100 mm GaN on SiC volume manufacturing Palmour, MTT S 2010 3
Recent great strides in RF power reliability 28 V RF oper. life > 2 years (Xband, 3 db comp., ~150 o C) Kolias, MTT S 2010 MTTF=1x10 7 h at 47 V (Cband, 5 db comp., ~150 o C) Yamasaki, MTT S 2010 MTTF=7x10 7 h at 28 V (40 GHz, 1.5 db comp., ~150 o C) Heying, MTT S 2010 4
Dominant degradation mechanisms In general: under RF stress? RF stress P out, Gain, I Dmax, I G, V T shift, dispersion RF introduces more degradation than DC RF stress accelerated by V DQ, P in, T j Conway, IRPS 2007; Joh, ROCS 2008, IEDM 2010, ROCS 2011; Chini, IEDM 2009 Chini, EUMW 2009 Indications of two competing mechanisms: Trap creation and trapping? Field driven structural degradation? Dammann, IRPS 2010 Rozman, ROCS 2009; Chini, IEDM 2009 5
Outline 1. RF power reliability concerns 2. Methodology for RF reliability experiments 3. Electrical and structural results 4. Discussion: the role of gate placement 5. Conclusions 6
RF power reliability concerns ON DC stress: Mostly benign High power DC stress: Not accessible to DC stress experiments Device blows up instantly OFF and semi ON high voltage DC stress : Degradation of I Dmax, R D, I Goff V T shift Electron trapping Trap creation Formation of grooves and pits under drain end of gate 7
RF experiment flowchart: conventional approach START RF Stress P out, PAE, Gain, I DQ, I GQ END T stress Limitations: Bias point shifts during stress Limited RF characterization No DC characterization No trap characterization If examining different RF conditions, RF characterization confusing 8
RF experiment flowchart: improved approach (I) T base START Short Characterization (DC, RF) T stress RF Stress End? NO YES Short Characterization (DC, RF) END New features: RF and DC characterization under standardized conditions At beginning, end and periodically through experiment Limitations: Limited characterization Characterization temperature cannot be too different from stress temperature Cannot separate trapping from permanent degradation 9
RF experiment flowchart: improved approach (II) RT Full Characterization (DC, RF, CC) T base START Short Characterization (DC, RF) T stress YES Detrapping RF Stress Key Event? NO New features: Comprehensive DC, RF and pulsed characterization under standardized conditions (RT) At beginning, end, and during experiment Detrapping step to enable trap characterization END: detrapping + Full characterization 10
Setup for RF reliability studies Accel RF AARTS RF10000 4/S system: two 2 4 GHz channels two 7 12 GHz channels Max P in =30 dbm T base =50 200 C DC/Pulsed Characterization KeithleySources Agilent B1500A Accel RF System Hardware Switching Matrix RF/DC Units Windows based PC DUT Heater T base MIT RF/DC Characterization Suite DC FOMs Current collapse Accel RF Software RF measurement Temperature control Stressing Augmented with: external instrumentation for DC/pulsed characterization software to control external instrumentation and extract DC and RF FOMs 11
RF stress experiments START Full Characterization (DC, RF, CC) RT Short Characterization (DC, RF) T base =50 C T stress YES Detrapping RF (DC) Stress Key Event? END: detrapping + Full characterization NO T base =100 C for 30 mins Full DC I V sweeps RF power sweep @ V DS =28 V, I DQ =100 ma/mm Current collapse (after 1 V DS =0, V GS = 10 V pulse) Room temperature DC FOMs: I Dmax, R S, R D, V T, I Goff, RF FOMs @ V DS =28 V, I DQ =100 ma/mm Saturated conditions (P in =23 dbm): P out,sat, G sat, PAE Linear conditions (P in =10 dbm): G lin Every few minutes at T base =50 C 12
RF stress experiments: P in step stress Motivation: higher P in larger V waveform at output MMIC: single stage internally matched 4x100 μm GaN HEMT (OFF-state V crit >60 V at RT) Gate centered in S D gap Step P in stress: V DS = 40 V, I DQ = 100 ma/mm P in = 0 (DC), 1, 20 27 dbm 300 min stress at each step Gain (db) 14 12 10 8 6 4 2 0 Gain PAE T stress =50 C (T j =110 230 C) V DS =40 V, I DQ =100 ma/mm Joh, ROCS 2011 10 15 20 25 30 P in (dbm) 25 20 15 10 5 0 PAE (%) 13
Evolution of RF stress P in (dbm) I DQ (ma/mm) 30 25 20 15 10 5 DC P in 0 0 10 20 30 40 50 60 Time (hr) 400 300 200 100 0 I DQ P out (dbm) Gain PAE (%) 33 32 31 30 DC P out P out 29 0 10 20 30 40 50 60 Time (hr) 30 25 20 15 10 5 PAE START Full Characterization (DC, RF, CC) RT Short Characterization (DC, RF) T base =50 C T stress YES Detrapping RF (DC) Stress Key Event? END: detrapping + Full characterization NO -100 0 20 40 60 Time (hr) 0 0 10 20 30 40 50 60 Time (hr) P in changing RF FOMs changing Degradation apparent but not easily quantifiable 14
RF FOM during short characterization START Detrapping P out at P in =23 dbm, G lin at P in =10 dbm V DS =28 V, I DQ =100 ma/mm, T base =50 C Full Characterization (DC, RF, CC) RT Short Characterization (DC, RF) T base =50 C T stress YES RF (DC) Stress Key Event? NO END: detrapping + Full characterization Mild degradation under DC and low P in Adding RF increases degradation: P in P out 15
DC FOM during short characterization I Dmax /I Dmax (0), R/R(0) 1.2 1.1 1 0.9 0.8 DC P in =1 20 21 22 23 24 25 26 27 dbm I Goff T base =50 C R D R S I Dmax 0 1000 2000 3000 Time (min) 1.E+01 1.E+00 1.E-01 1.E-02 1.E-03 1.E-04 1.E-05 1.E-06 I Goff (ma/mm) START Full Characterization (DC, RF, CC) RT Short Characterization (DC, RF) T base =50 C T stress YES Detrapping RF (DC) Stress Key Event? END: detrapping + Full characterization NO Mild degradation under DC and low P in At P in =20 dbm, step degradation in I Goff Beyond P in =20 dbm, increasing degradation of I Dmax and R D 16
DC/RF/CC full characterization Permanent I Dmax Degradation (%) Current Collapse (%) 9 8 7 6 5 4 3 2 1 0 T base =RT Initial DC RF P out Current Collapse Δ I Dmax 10 0 10 20 30 Stress Input Power P in (dbm) Beyond P in =20 dbm: Sharp P out degradation Permanent degradation of I Dmax Increased CC evidence of new trap creation 33 32 31 30 29 Saturated P out (dbm) START 100 C Detrapping Full Characterization (DC, RF, CC) RT Short Characterization (DC, RF) T base =50 C T stress YES RF (DC) Stress Key Event? END: detrapping + Full characterization NO 17
Structural degradation (planar view) SEM AFM Pit formation along drain end of gate edge Similar to DC high voltage OFF state stress DC OFF state stress, V DG =50 V, 1000 min, ~150 o C Makaram, APL 2010 18
HV OFF state DC vs. RF power degradation Similar pattern of degradation: HV OFF state DC RF power I Dmax beyond V crit beyond P in crit R D beyond V crit beyond P in crit R S small increase small increase I Goff beyond V crit beyond P in crit Current Collapse beyond V crit beyond P in crit Permanent I Dmax beyond V crit beyond P in crit Pits under drain end of gate Yes Yes Pits under source end of gate No No High V end of load line responsible for degradation 19
Step P in stress: Offset Gate Offset gate devices (L GS <L GD ): OFF-state V crit > 80 V at T=150 C Saturated P out (dbm) 32.5 32 31.5 31 30.5 30 RF FOMs DC RF P in =20 23 26 dbm T j ~170 C by adjusting T base 15 14.5 Gain 14 P out 13.5 Inner loop (50 C) 13 0 300 600 900 1200 Time (min) Small Signal Gain G lin (db) I Dmax /I Dmax (0), R/R(0) 3 2.5 2 1.5 1 0.5 0 DC FOMs DC RF P 1.E+01 in =20 23 26 dbm Inner loop (50 C) R S R D I Dmax 1.E+00 1.E-01 1.E-02 I Goff 1.E-03 0 300 600 900 1200 Time (min) I Goff (ma/mm) Joh, IEDM 2010 Increased degradation under high P in No I Goff degradation Degradation of I Dmax and R S, not R D 20
HV OFF state DC vs. RF power degradation Different pattern of degradation: HV OFF state DC RF power I Dmax beyond V crit beyond P in crit R D beyond V crit beyond P in crit R S small increase beyond P in crit I Goff beyond V crit No Current Collapse beyond V crit beyond P in crit Permanent I Dmax beyond V crit beyond P in crit Pits under drain end of gate Yes No Pits under source end of gate No No High V end of load line NOT responsible for degradation 21
High power pulsed stress High power stress not accessible in DC pulsed stress Offset gate and centered gate devices on same wafer: Normalized R S, R D 1.4 1.3 1.2 1.1 1 100 pulses, 500 μs, 0.05% duty I Dpulse =950 ma/mm Offset gate R S R D Centered gate R D R S 0 20 40 60 80 Stress V DS (V) High power region of load line responsible for degradation Pulsed stress reproduces R S degradation in offset gate device No R S degradation in centered gate 22
Summary New RF reliability testing methodology developed Under RF stress, degradation worse than at DC bias point Different patterns of RF degradation observed: In some device designs, it reproduces HV OFF state DC degradation (field driven) In other device designs, degradation pattern correlates with high power pulsed stress (power driven?) DC reliability not good predictor for RF reliability Need for fundamental studies of RF reliability 23