Recent Progress in Understanding the Electrical Reliability of GaN - - PowerPoint PPT Presentation

recent progress in understanding the electrical
SMART_READER_LITE
LIVE PREVIEW

Recent Progress in Understanding the Electrical Reliability of GaN - - PowerPoint PPT Presentation

Recent Progress in Understanding the Electrical Reliability of GaN High-Electron Mobility Transistors J. A. del Alamo Microsystems Technology Laboratories Massachusetts Institute of Technology RSAMD 2014 Golden, CO, Sept. 7-9, 2014


slide-1
SLIDE 1

Recent Progress in Understanding the Electrical Reliability of GaN High-Electron Mobility Transistors

  • J. A. del Alamo

Microsystems Technology Laboratories Massachusetts Institute of Technology

RSAMD 2014

Golden, CO, Sept. 7-9, 2014

Acknowledgements:

  • F. Gao, J. Jimenez, D. Jin, J. Joh, T. Palacios, C. V. Thompson, Y. Wu

ARL (DARPA-WBGS program), NRO, ONR (DRIFT-MURI program),

slide-2
SLIDE 2

Outline

  • 1. Motivation
  • 2. Electrical and structural degradation of GaN HEMTs
  • 3. Hypotheses for GaN HEMT degradation mechanisms
  • 4. Paths for mitigation of GaN HEMT degradation

2

slide-3
SLIDE 3

Micovic, Cornell Conf 2010 94-95 GHz MMIC PAs: Micovic, MTT-S 2010

Pout>40 W/mm,

  • ver 10X GaAs!

Wu, DRC 2006

Breakthrough RF-µw-mmw power in GaN HEMTs

3

slide-4
SLIDE 4

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

4

slide-5
SLIDE 5

GaN HEMT: Electrical reliability concerns

High-voltage OFF and semi-ON: – Degradation of IDmax, RD, IGoff – VT shift – Electron trapping – Trap creation High-power: – Not accessible to DC stress experiments – Device blows up instantly ON: – Mostly benign

5

slide-6
SLIDE 6

ID, RD, and IG start to degrade beyond critical voltage (Vcrit) + increased trapping behavior – current collapse

IDmax: VDS=5 V, VGS=2 V IGoff: VDS=0.1 V, VGS=-5 V

Critical voltage for degradation in DC step-stress experiments

Joh, EDL 2008

G S D AlGaN GaN 2DEG

VGS VDS

1.E-06 1.E-05 1.E-04 1.E-03 1.E-02 1.E-01 1.E+00 1.E+01 0.85 0.9 0.95 1 1.05 1.1 1.15 1.2 10 20 30 40 50 |IGoff| (A/mm) IDmax/IDmax(0), R/R(0) VDGstress (V)

IDmax RS RD IGoff Vcrit

OFF-state, VGS=-10 V

=-10 V

6

slide-7
SLIDE 7

Critical voltage: a universal phenomenon

Meneghini, IEDM 2011 Ivo, MR 2011

GaN HEMT on SiC GaN HEMT on SiC

Demirtas, ROCS 2009

GaN HEMT on Si

Liu, JVSTB 2011

GaN HEMT on SiC

Ma, Chin Phys B 2011

GaN HEMT on sapphire GaN HEMT on Si

Marcon, IEDM 2010

7

slide-8
SLIDE 8

1.E-04 1.E-03 1.E-02 1.E-01 1.E+00 1.E+01 1.E+02 0.4 0.6 0.8 1 1.2 10 20 30 |IGoff| (mA/mm) IDmax/IDmax(0), R/R(0) VDGstress (V)

10 20 30 40 Device #1 #2 #3 #4 #5 IDmax IGoff

Joh, MR 2010

#6: unstressed

AlGaN GaN Gate

  • Small dimple in

early stages of IG degradation;

  • ID degradation

delayed

VDS=0 stress

Structural degradation: cross section

8

slide-9
SLIDE 9

Correlation between pit geometry and IDmax degradation

10 20 30 40 50 2 4 6 8 Permanent IDmax Degradation (%) Pit depth (nm) 5 10 15 20 2 4 6 8 Current collapse (%) Pit depth (nm)

Pit depth and IDmax degradation correlate:  both permanent degradation and current collapse (CC)

Pit depth Joh, MR 2010

9

slide-10
SLIDE 10

Structural degradation: planar view

200 nm 200 nm 200 nm 200 nm 200 nm

Unstressed VDG=15 V VDG=19 V (Vcrit) VDG=42 V VDG=57 V

  • Continuous groove appears for Vstress<Vcrit
  • Deep pits formed along groove for Vstress>Vcrit

OFF-state step-stress, VGS=-7 V, Tbase=150 °C Makaram, APL 2010

10

slide-11
SLIDE 11

2 4 6 8 10 12 50 100 150

Post-Stress Current Collapse (%) Average Defect Area (nm2)

2 4 6 8 10 12 50 100 150

Permanent IDmax Degradation (%) Average Defect Area (nm2)

IDmax degradation and pit cross-sectional area correlate

Makaram, APL 2010

Correlation between pit geometry and IDmax degradation

Cross-sectional area averaged over 1 µm

11

slide-12
SLIDE 12

Planar degradation: the role of time

10s 100s 1ks 10ks G A T E 0.5µm

VDS=0, VGS=-40 V, Tbase=150 °C

  • Very fast groove formation (within 10 s)
  • Delayed pit formation
  • Pit density/size increase with time
  • Good correlation between IDmax degradation and pit area

Joh, IWN 2010

12

slide-13
SLIDE 13

Time evolution of degradation for constant Vstress > Vcrit

IGoff and VT degradation:

  • fast (<10 ms)
  • saturate after 104 s

CC degradation:

  • slower
  • hint of saturation for long time

Permanent IDmax degradation:

  • much slower
  • does not saturate with time

10

  • 4

10

  • 2

10 10

2

10

4

10

6

0.05 0.1 0.15 0.2 0.25 Stress time (s) |∆VT| (V) 10

  • 8

10

  • 7

10

  • 6

10

  • 5

10

  • 4

|IGoff| (A)

IGoff |ΔVT| Initial Stress: VGS=-7 V and VDS=40 V 125 °C

10

  • 4

10

  • 2

10 10

2

10

4

10

6

0.05 0.1 0.15 0.2 0.25 Stress time (s) |∆VT| (V) 10

  • 8

10

  • 7

10

  • 6

10

  • 5

10

  • 4

|IGoff| (A)

IGoff |ΔVT| Initial Stress: VGS=-7 V and VDS=40 V 125 °C

Joh, IRPS 2011

13

slide-14
SLIDE 14

Joh, IRPS 2011

The role of temperature in time evolution

  • IG: weak T dependence
  • CC, IDmax: T activated

Incubation time

14

slide-15
SLIDE 15

Temperature acceleration

  • f incubation time

28 30 32 34 36

  • 5

5 10 15 1/kT (eV-1) ln(τinc) (s)

Permanent IDmax degradation Ea=1.12 eV Current collapse Ea=0.59 eV IGoff, Ea=0.17 eV

  • Different Ea for IGoff, CC, IDmax reveal different degradation physics
  • Ea for permanent IDmax degradation similar to life test data*

* Saunier, DRC 2007; Meneghesso, IJMWT 2010

15

slide-16
SLIDE 16

DC semi-ON stress experiments

Prominent pits and trenches under gate edge on drain side

Stress conditions: ID=100 mA/mm, VDS=40 or 50 V Step-T experiments: 50<T

a<230oC

SEM AFM

Wu, submitted to TED 16

slide-17
SLIDE 17

Trench/pit depth and width correlate with IDmax degradation

Wu, submitted to TED

ΔID=25.8%

Structural vs. electrical degradation

5 10 15 20 25 30 10 20 30 40 50 60 70 80 90

Trench/pit width, depth (nm) Permanent IDmax degradation (%) Trench/pit width Trench/pit depth

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0

  • 20
  • 15
  • 10
  • 5

5 10

Depth (nm) x (µm)

Gate Souce Drain Trench width Trench depth

17

slide-18
SLIDE 18
  • Pit/trench depth increase towards center of gate finger

 self heating + thermally activated process

  • Permanent IDmax degradation is thermally activated with Ea~1.0 eV

Wu, ROCS 2014

ΔID=21.6%

Thermally activated degradation

Drain Source Source Distance from center of gate finger Gate fingers

20 40 60 80 100 120 140 160 180 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0

Depth of damage (nm) Distance from center of gate finger (µm)

20.0 20.5 21.0 21.5 22.0 22.5 23.0 23.5 4.0 4.5 5.0 5.5 6.0 6.5 7.0 7.5

Ea=1.04 eV ln(1/|slope|) 1/kTchannel (eV

  • 1)

18

slide-19
SLIDE 19

1E-4 1E-3 0.01 0.1 1 10 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1

IDmax/IDmax(0) |IGoff| (mA/mm)

evolution of stress experiment

“Universal degradation” pattern:

  • IG degradation takes places first without ID degradation
  • ID degradation takes place next without further IG degradation

Wu, ROCS 2014

Sequential IG and ID degradation

Stress conditions: ID=100 mA/mm, VDS=40 or 50 V Step-Temperature: 50<T

a<230oC

START

19

slide-20
SLIDE 20

RF power degradation

Joh, IEDM 2010 Joh, ROCS 2011 Joh, MR 2012

  • RF power degradation

pattern matches that of OFF-state DC stress

  • But not always…

AFM SEM

20

slide-21
SLIDE 21

1. IG degradation

  • Fast
  • Electric-field driven
  • Little temperature sensitivity (Ea~0.2 eV)
  • Tends to saturate

Correlates with appearance of shallow groove and small pits

  • On S and D side (bigger on D side)
  • Groove/small pits appear for Vstress< Vcrit

Summary of electrical and structural degradation

21

slide-22
SLIDE 22

2. Current-collapse degradation (trapping)

  • Slower
  • Enhanced by temperature, electric field
  • Tends to saturate for very long times

Correlates with pit growth:

  • Pits randomly located on drain side
  • Pits grow with Vstress, time and temperature
  • Pits eventually merge

Summary of electrical and structural degradation

Dominant trap created by stress already present in virgin sample, Ea=0.56 eV Joh, IRPS 2011

22

slide-23
SLIDE 23

3. IDmax, RD degradation

  • Much slower
  • Temperature activated (Ea~1 eV)
  • Electric-field driven
  • Does not saturate

Correlates with geometry of pits and trench

  • Pits grow larger and merge into trench
  • Trench grows deeper

Summary of electrical and structural degradation

23

slide-24
SLIDE 24

defect state

ΔΦbi EC EF G S D AlGaN GaN 2DEG

AlGaN GaN

defect state

ΔΦbi EC EF G S D AlGaN GaN 2DEG

AlGaN GaN

Initial hypothesis: Inverse Piezoelectric Effect Mechanism

Strong piezoelectricity in AlGaN  |VDG|↑  tensile stress ↑  crystallographic defects beyond critical elastic energy Defects: Trap electrons  ns↓ → RD↑, ID ↓ Strain relaxation  ID ↓ Provide paths for IG  IG↑

Joh, IEDM 2006 Joh, IEDM 2007 Joh, MR 2010b

G S D AlGaN GaN 2DEG

24

slide-25
SLIDE 25

Model for critical voltage

z x x y

Vertical E-field E3 (MV/cm)

gate AlGaN GaN drain source

z x x y

Elastic energy density (MJ/m3)

gate AlGaN GaN drain source

VGS=-5 V, VDS=33 V 16nm 28% AlGaN

<Vertical Electric Field> <Elastic Energy Density>

3 31 33 33 13 10 33 2 13 12 11 1

) ( ) 2 ( E e C e C S C C C C T − + − + =

Mismatch stress Inverse piezoelectric stress

2 3 10 2 1 33 12 2 13 33 11 33

) ( 2 aE T T C C C C C C W + ∝ + − =

Joh, MR 2010

25

slide-26
SLIDE 26

Predictions of Inverse Piezoelectric Effect model borne out by experiments

To enhance GaN HEMT reliability:

  • Reduce AlN composition of AlGaN barrier (Jimenez, ESREF 2011)
  • Thin down AlGaN barrier (Lee, EL 2005)
  • Use thicker GaN cap (Ivo, IRPS 2009; Jimenez, ESREF 2011)
  • Use InAlN barrier (Jimenez, ESREF 2011)
  • Use AlGaN buffer (Joh, IEDM 2006; Ivo, MR 2011)
  • Electric field management at drain end of gate (many)

Can’t explain:

  • Groove formation/IG degradation below critical voltage
  • Presence of oxygen in groove/pit
  • Role of atmosphere during stress
  • Role of surface chemistry

26

slide-27
SLIDE 27

IG degradation for Vstress < Vcrit

Meneghini, IEDM 2011 Marcon, IEDM 2010

Vcrit=75 V

  • Sudden irreversible increase in IG,

enhanced by Vstress

  • No reported ID degradation
  • Preceded by onset of IG noise
  • Weakly temperature enhanced

(Ea=0.12 eV)

27

slide-28
SLIDE 28

IG degradation correlates with electroluminescence hot spots

Zanoni, EDL 2009 Meneghini, IEDM 2011

  • Gate current electrons produce EL in GaN substrate
  • EL spots tend to merge into a continuous line

VDS=0

28

slide-29
SLIDE 29

EL hot spots correlate with pits, pits are conducting

Montes Bajo, APL 2012

Shallow pits and groove responsible for IG degradation

Normal AFM Conducting AFM EL picture AFM topography

29

slide-30
SLIDE 30

Pits/Groove increase mechanical stress

Pit/groove increases mechanical stress due to inverse piezoelectric effect at drain end of gate Ancona, JAP 2012

  • 2 nm x 3 nm groove increases

mechanical stress in AlGaN from 4.6 GPa to 13 GPa

  • Groove has little effect in

current underneath

  • Pit formation brings major

loss of current

30

slide-31
SLIDE 31

Oxygen inside pit

  • O, Si, C found inside pit
  • Anodization mechanism for pit

formation? (Smith, ECST 2009)

  • Electrical stress experiments under

N2 inconclusive Park, MR 2009 Conway, Mantech 2007

EDX LEES

31

slide-32
SLIDE 32

Role of atmosphere on structural degradation

Surface pitting significantly reduced in vacuum

Stressed in ambient air Stressed in vacuum of 10-7 Torr

Off-state stress: Vds = 43 V, Vgs = -7 V for 3000 s in dark at RT

SEM Top View AFM Depth Profile TEM Cross Section

Gao, TED 2014

ΔID=0.5% ΔID=5.0%

32

slide-33
SLIDE 33

Impact of Moisture on Surface Pitting

Stressed in water- saturated gas (Ar)

SEM Top View TEM Cross Section

Stressed in dry gas (Ar)

  • Moisture enhances surface pitting
  • Results reproduced with dry/wet O2, N2, CO2 and air

Gao, TED 2014

ΔID=0.3% ΔID=28.8% Off-state stress: Vds = 43 V, Vgs = -7 V for 3000 s in dark at RT

33

slide-34
SLIDE 34

New hypothesis: AlGaN corrosion at edge of gate

  • Reduction of water:
  • Anodic oxidation of AlGaN:

2H2O + 2e- ↔ 2OH- + H2 2Alx Ga1-xN + 6h+ ↔ 2xAl3+ + 2(1-x)Ga3+ +N2 2xAl3+ + 2(1-x)Ga3+ + 6OH- ↔ xAl2O3 + (1-x)Ga2O3 + 3H2O

  • Complete redox electrochemical reaction:

2Alx Ga1-xN + 3H2O ↔ xAl2O3 + (1-x)Ga2O3 + N2 + H2 Electrochemical cell formed at drain edge of gate

e-

h+

H2O Gao, TED 2014

34

slide-35
SLIDE 35

Source of holes: trap-assisted tunneling

High electric field under gate edge Trap-assisted BTBT electron tunneling hole generation at AlGaN surface

Gao, TED 2014

35

slide-36
SLIDE 36

Source of water: diffusion through SiN

Gao, TED 2014

  • Water-vapor transmission rate (WVTR) through 100 nm
  • f PECVD SiN:

0.01~0.1 g/m2/day

  • Gao’s estimate of necessary WVTR to cause pits:

0.05~0.1 g/m2/day H2O H2O

36

slide-37
SLIDE 37

Tentative new model for GaN HEMT electrical degradation

Step 1: formation of shallow pits/continuous groove in cap

  • Pits/groove conducting: IG↑

Step 2: growth of pits through anodic oxidation of AlGaN

  • IDmax↓ as electron concentration under gate edge reduced
  • CC↑ due to new traps

Exponential dependence of tunneling current on electric field  “critical voltage” behavior

37

slide-38
SLIDE 38

1. Reduce hole production

– Mitigate electric field at gate edge:

  • gate edge design
  • field plate design

– Mitigate traps in AlGaN:

  • ptimize growth conditions
  • reduce AlN composition
  • thin down AlGaN
  • mitigate mechanical stress

2. Reduce water around gate edge

1. Reduce SiN permeability 2. Mitigate trapped moisture during process 3. Hermetic package

Paths for mitigation

38

slide-39
SLIDE 39

Many questions…

  • IG degradation:

– Detailed physics of onset of pits/groove? Also of electrochemical nature? – Why weak temperature activation? – Why does IG degradation saturate? – Detailed mechanism for electrical conduction of pits?

  • Trap formation:

– Why traps introduced during degradation have similar dynamic signature as virgin traps?

  • Mechanical stress:

– Does mechanical stress and inverse piezoelectric effect still play role in degradation?

  • Large variability in reliability:

– Why? Also need effective screening process for virgin devices

  • High-power RF stress

– Is there a pulsed stress mode that faithfully emulates high-power RF stress?

39