Recent Progress in Understanding the Electrical Reliability of

GaN High-Electron Mobility Transistors

J. A. del AlamoMicrosystems Technology LaboratoriesMassachusetts Institute of Technology

2015 MRS Spring MeetingSymposium AA: Materials for Beyond the Roadmap Devices in Logic, Power and MemorySan Francisco, CA, April 6-10, 2015

Acknowledgements: C. Y. Chen, F. Gao, J. Jimenez, D. Jin, J. Joh, T. Palacios, C. V. Thompson, Y. WuARL (DARPA-WBGS program), NRO, ONR (DRIFT-MURI program),

Outline

1. A few “universal” observations 2. Hypotheses for degradation mechanisms3. Many questions…

2

Counter-IED Systems (CREW)

200 W GaN HEMT for cellular base station Kawano, APMC 2005

100 mm GaN-on-SiC volume manufacturingPalmour, MTT-S 2010

GaN HEMT: breakthrough RF power technology

Sumitomo Remote Radio Head for

Japanese Base Station3

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

4

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

GS D

AlGaN

GaN

2DEG

VGSVDS

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

|I Gof

f| (A

/mm

)

I Dm

ax/I D

max

(0),

R/R

(0)

VDGstress (V)

IDmax

RS

RD

IGoff

Vcrit

OFF-state, VGS=-10 V

=-10 V

5

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 20106

Structural degradation; correlation with electrical degradation

0

10

20

30

40

50

0 2 4 6 8

Perm

anen

t IDm

axDe

grad

atio

n (%

)

Pit depth (nm)

• Pit at edge of gate• Pit depth and IDmax degradation correlate

Pit depth

Joh, MR 2010

Chowdhury, EDL 20087

Structural damage at gate edge: a universal phenomenon

Barnes, CS-MANTECH 2012 Dammann, IIRW 2011 Marcon, MR 2010

Chang, TDMR 2011

Liu, JVSTB 2011

Christiansen, IRPS 2011

Cullen, TDMR 2013

8

0

2

4

6

8

10

12

0 50 100 150

Perm

anen

t ID

max

Degr

adat

ion

(%)

Average Defect Area (nm2)

Structural degradation: planar view

200 nm

200 nm

Unstressed

• Vstress>Vcrit: pits along gate edge • Pit cross-sectional area correlates

with ID degradation

OFF-state stress:VDG=57 V, Tbase=150 °C

Makaram, APL 2010

averaged over 1 µm

9

Holzworth, ECST 2014

Structural damage at gate edge: a universal phenomenon

Barnes, CS-MANTECH 2012

Whiting, MR 2012

Brunel, MR 2013

Monte Bajo, APL 2014

10

Time evolution of degradation for constant Vstress > Vcrit

IGoff and VT degradation:• fast (<10 ms)• saturate after 104 s

Permanent IDmax degradation:• much slower• does not saturate with time

10-4 10-2 100 102 104 1060

0.05

0.1

0.15

0.2

0.25

Stress time (s)

|∆V T| (

V)

10-8

10-7

10-6

10-5

10-4

|I Gof

f| (A)

IGoff

|ΔVT|

Initial

Stress: VGS=-7 V and VDS=40 V125 °C

10-4 10-2 100 102 104 1060

0.05

0.1

0.15

0.2

0.25

Stress time (s)

|∆V T| (

V)

10-8

10-7

10-6

10-5

10-4

|I Gof

f| (A)

IGoff

|ΔVT|

Initial

Stress: VGS=-7 V and VDS=40 V125 °C

Joh, IRPS 201111

The role of temperature in time evolution

Different degradation physics:• IG: weak T dependence• IDmax: T activated, Ea similar

to life-test data*

Incubation time

28 30 32 34 36-5

0

5

10

15

1/kT (eV-1)

ln(τ in

c) (s)

Permanent IDmax degradationEa=1.12 eV

Current collapseEa=0.59 eV

IGoff, Ea=0.17 eV

* Saunier, DRC 2007; Meneghesso, IJMWT 2010

Joh, IRPS 2011

Incubation time

12

DC semi-ON stress experiments

• Pits and trenches under gate edge on drain side

• Trench/pit depth and width correlate with IDmax degradation

Stress: ID=100 mA/mm, VDS=40 or 50 V Step-T experiments: 50<Ta<230oC

(Tj~110-330oC)

SEM

AFM

Drain

Wu, JAP 2015

0 5 10 15 20 25 300

10

20

30

40

50

60

70

80

90

Tren

ch/p

it w

idth

, dep

th (n

m)

Permanent IDmax degradation (%)

Trench/pit width

Trench/pit depth

Average of 5 1 µm x1 µm scans at finger center

13

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

/|slo

pe|)

1/kTchannel (eV-1)

• Pit/trench depth increase towards center of gate finger self heating + thermally activated process

• Permanent IDmax degradation thermally activated with Ea~1.0 eV

Wu, MR 2014ΔID=21.6%

Thermally activated degradation

DrainSource

Source

Distance from center of gate finger

Gate fingers

0 20 40 60 80 100 120 140 160 1800.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

4.0

Dep

th o

f dam

age

(nm

)

Distance from center of gate finger (µm)

ID degradation rate

14

1. IG degradation• Fast• Electric-field driven• Weak temperature

sensitivity (Ea~0.2 eV)• Tends to saturate

Correlates with appearance of shallow groove and small pits

Summary of electrical and structural degradation under OFF and Semi-ON bias

2. IDmax degradation• Much slower• Electric-field driven• Temperature activated (Ea~1 eV)• Starts after IG saturated• Does not saturate

Correlates with growth of pits andmerging into trenches

15

defectstate

ΔΦbi

EC

EF

GS D

AlGaN

GaN2DEG

AlGaN GaN

defectstate

ΔΦbi

EC

EF

GS D

AlGaN

GaN2DEG

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 2006Joh, IEDM 2007Joh, MR 2010b

16

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 • Sequential nature of IG and ID degradation• Presence of oxygen in pit• Role of atmosphere during stress

17

IG degradation not critical; TDDB*-like

Meneghini, IEDM 2011Marcon, IEDM 2010

Vcrit=75 V

• IG starts increasing for Vstress<Vcrit• Onset enhanced by Vstress

• Weibull distribution• Preceded by onset of IG noise

* TDDB = Time-Dependent Dielectric Breakdown18

IG correlates with EL; EL hot spots correlate with pits, pits are conducting

Montes Bajo, APL 2012

Shallow pits responsible for IG degradation

Normal AFM

Conducting AFM

EL picture

AFM topography

Zanoni, EDL 2009

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

I Dm

ax/I D

max

(0)

|IGoff| (mA/mm)

evolution ofstress experiment

“Universal” degradation pattern:• IG degradation first without ID degradation• ID degradation next without further IG degradation• “Corner” of IG and ID same for all samples

Wu, MR 2014

Sequential IG and ID degradation

Semi-ON stress: ID=100 mA/mm, VDS=40 or 50 V Step-Temperature: 50<Ta<230oC

START

Wu, ROCS 2014

20

Oxygen inside pit

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

formation? (Smith, ECST 2009)

Park, MR 2009

Conway, Mantech 2007

EDXLEES

21

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, air and vacuum

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

Role of atmosphere on structural degradation

22

New phenomenon: AlGaN corrosion

2AlxGa1-xN + 3H2O ↔ xAl2O3 + (1-x)Ga2O3 + N2 + H2

Electrochemical cell formed at drain edge of gate

Electrochemical reaction (requires holes):

Source of holes: trap-assisted BTBT

Gao, TED 2014

Source of water: diffusion through SiN

23

Tentative complete model? Step 1: formation of shallow pits/continuous groove in cap• TDDB-like formation of small conducting

paths: IG↑

Step 2: growth of pits through anodic oxidation of AlGaN• IDmax↓ as electron concentration under gate

edge reduced

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

200 nmVDG=19 V (Vcrit)

Makaram, APL 2010

200 nmVDG=57 V

24

Many questions…• Why weak temperature activation of IG degradation?• Why does IG degradation tend to saturate?• Why does ID degradation start as IG degradation saturates?• Does mechanical stress and inverse piezoelectric effect play role?

• Why large variability in reliability? • Is this all relevant under RF power conditions?

Small pit (2 nm x 3 nm) increases mechanical stress in AlGaN by 3X

Ancona, JAP 2012

25