Gate current degradation in W-band InAlN/AlN/GaN …...Gate current degradation in W-band...

Gate current degradation in W-band InAlN/AlN/GaN HEMTs

under Gate Stress Yufei Wu and Jesús A. del Alamo

Microsystems Technology Laboratories (MTL)Massachusetts Institute of Technology (MIT)

Sponsor: National Reconnaissance OfficeApril 04, 2017

1

Purpose

To understand degradation of gate leakage current

in ultra-scaled InAlN/AlN/GaN HEMTs

under positive gate stress

2

1. Motivation 2. Devices and experimental approach3. Gate stress experiments

• Harsh step-stress (-recovery) experiments• Mild constant gate stress experiment

4. Thermal stress5. Gate current: dominant charge transport

mechanisms6. Conclusions

Outline

3

1. Motivation 2. Devices and experimental approach3. Gate stress experiments

• Harsh step-stress (-recovery) experiments• Mild constant gate stress experiment

4. Thermal stress5. Gate current: dominant charge transport

mechanisms6. Conclusions

Outline

4

Benefits of GaN for RF: • Wide bandwidth• High power density• Excellent energy efficiency• Small volume

Promising applications:

Benefits of GaN for RF

5

• High spontaneous polarization in InAlN high 2DEG density • InAlN thickness scaling gate length scaling

W- and V-band applications

Al0.2Ga0.8N/GaN ln0.17Al0.83N/GaN

Δ P0 (cm-2) 6.5 x 1012 2.7 x 1013

Ppiezo (cm-2) 5.3 x 1012 0

Ptotal (cm-2) 1.2 x 1013 2.7 x 1013

[J. Kuzmik, EDL 2001]

InAlN as barrier material

6

In0.17Al0.83N

In0.17Al0.83N lattice matched to GaN Potentially better reliability!

[M. A. Laurent, JAP 2014]

InAlN as barrier material

7

1. Motivation 2. Devices and experimental approach3. Gate stress experiments

• Harsh step-stress (-recovery) experiments• Mild constant gate stress experiment

4. Thermal stress5. Gate current: dominant charge transport

mechanisms6. Conclusions

Outline

8

9

Devices (E-mode)

Thermal models available

BFTEM of virgin device

Gate metal

passivation

InAlN

AlNGaN

InAlN/AlN/GaN HEMTs:• E-mode• Lg = 40 nm• Lgs=Lgd=1 µm• W-band

[Saunier, CSICS 2014]

5 nm1 nm

FOMs:• IDmax : ID at VGS = 2 V and VDS = 4 V• VTsat : VGS extrapolated from ID at peak gm point at VDS = 4 V• IGoff : IG at VGS = -2 V, VDS = 0.1 V• IDoff : ID at VGS = -2 V, VDS = 0.1 V• RD : at IG = 20 mA/mm• RS : at IG = 20 mA/mm

Detrapping & initialization:• 100 °C bake for 1 hour

10

FOMs for benign characterization, detrapping methodology

-2 -1 0 1 210-3

10-2

10-1

100

101

102

103

VDS = 4 Vvirgin device

I D (m

A/m

m)

VGS (V)

11

ΔIDmax/IDmax(0)[%]

ΔVTlin[mV] RD/RD(0) RS/RS(0)

After initialization

0 0 1 1

After 200 characterizations

0.70 -23.4 0.95 0.89

After detrapping 0 2.3 1.01 1

Impact of 200 successive characterizations

Nearly complete recovery after thermal detrapping step characterization suite is benign and detrapping step is effective

Impact of characterization and detrapping

1. Motivation 2. Devices and experimental approach3. Gate stress experiments

• Harsh step-stress (-recovery) experiments• Mild constant gate stress experiment

4. Thermal stress5. Gate current: dominant charge transport

mechanisms6. Conclusions

Outline

12

3600 4800 6000 7200101

102

103

I Gst

ress (m

A/m

m)

time (s)

VGS,stress (V)1.3 1.5 1.7 1.9 2.1 2.3

2.3 V

13

Stress conditions: • VGS,stress = 0.1 – 2.5 V in 0.1 V steps, VDS = 0 V, RT • Recovery: VDS = VGS = 0 V• Stress time = recovery time = 150 s• Characterization every 15 s

• IGstress ↑ at constant VGS,stress for VGS,stress ≥ 2.3 V

0 1200 2400 3600 4800 6000 7200

10-810-710-610-510-410-310-210-1100101102103

I Gst

ress (m

A/m

m)

time (s)

VGS,stress (V)0.1 0.5 0.9 1.3 1.7 2.1

RT Positive-VG step-stress-recovery experiment

0 1200 2400 3600 4800 6000 720010-4

10-3

10-2

10-1

100

101

102 2.5

finaldetrapped

stressrecovery

|I Gof

f| (m

A/m

m)

time (s)

2.10.1 0.5 0.9 1.3 1.7VGS,stress (V)

1.7 V

2.3 V

0 1200 2400 3600 4800 6000 720002468

101214161820

2.5

stressrecovery

R D (oh

m.m

m)

time (s)

finaldetrapped

2.10.1 0.5 0.9 1.3 1.7VGS,stress (V)

2.3 V

Time evolution of IGoff and RD

Two mechanisms:• From VGS,stress = 1.7 V: IGoff ↑, trapping ↑ mechanism 1: new defects

generated in AlN• From VGS,stress = 2.3 V:

o IGoff ↑ ↑o RD and RS ↑ ↑

14

Mechanisms 2 ?

Before and after stress: permanent degradation

• IDoff ↑↑• IDmax ↓• ΔVTsat > 0

15

Mechanisms 2: consistent with gate sinking

IDoff

VTsat

IDmax

Stress conditions: • VGS,stress = 0.1 – 2.5 V in 0.1 V steps, VDS = 0 V, Tstress = 150 °C • Stress time = 60 s• Characterization every 15 s

High T Positive-VG step-stress experiment

16

0 240 480 720 960 1200 144010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

101

102

103 2.4VGS,stress (V)

I Gst

ress

(mA/

mm

)

time (s)

0.1 0.4 0.8 1.2 1.6 2.0

960 1080 1200 1320 1440 1560102

103

VGS,stress (V)

I Gst

ress (m

A/m

m)

time (s)

1.6 1.8 2.0 2.2 2.4

2.0 V

• IGstress ↑ at constant VGS,stress for VGS,stress ≥ 2.0 V

High T Positive-VG step-stress experiment

0 240 480 720 960 1200 1440

2

4

6

8

10

12

14

finaldetrapped

R D (oh

m.m

m)

time (s)

VGS,stress (V)0.1 0.4 0.8 1.2 1.6 2.0 2.4

2.0 V

0 240 480 720 960 1200 144010-4

10-3

10-2

10-1

100

101

VGS,stress (V)

finaldetrapped

|I Gof

f| (m

A/m

m)

time (s)

0.1 0.4 0.8 1.2 1.6 2.0 2.4

1.4 V

2.0 V

• From VGS,stress = 1.4 V: IGoff ↑• From VGS,stress = 2.0 V: IGoff, RD, and RS ↑ ↑• Lower threshold for degradation than at RT Both mechanisms

thermally enhanced17

1. Motivation 2. Devices and experimental approach3. Gate stress experiments

• Harsh step-stress (-recovery) experiments• Mild constant gate stress experiment

4. Thermal stress5. Gate current: dominant charge transport

mechanisms6. Conclusions

Outline

18

Stress conditions: • VGS,stress = 2 V, VDS = 0 V, RT • Characterization every 15 s

RT Constant-VG stress experiment

• IGoff becomes noisy at tstress ~ 500 s; increases afterwards consistent with trap generation in AlN layer (mechanism 1)

• RD changes little throughout experiment• IGstress keeps decreasing no Schottky barrier degradation

19

Before and after stress: permanent degradation

• IDoff ↑↑• No significant IDmax degradation• No significant subthreshold characteristics degradation

20

IDoff

Two degradation mechanisms identified:1. Under mild gate stress:

oObservation: IGoff ↑, trapping ↑, thermally enhancedoProposed mechanism 1: high electric field induced

defect generation in AlN interlayer

2. Under harsh gate stress:oObservation: IGoff ↑, RD and RS ↑, ΔVT > 0, IDmax ↓,

thermally enhanceoProposed mechanism 2: self-heating induced Schottky

gate degradation, or gate-sinking

21

Summary so far

1. Motivation 2. Devices and experimental approach3. Gate stress experiments

• Harsh step-stress (-recovery) experiments• Mild constant gate stress experiment

4. Thermal stress5. Gate current: dominant charge transport

mechanisms6. Conclusions

Outline

22

-2 -1 0 1 210-3

10-2

10-1

100

101

102

103

IDmax

VTsat

I D (m

A/m

m)

VGS (V)

before stress after 400 °C after 500 °C

VDS = 4 V

IDoff

Thermal stress experiment

Impact of thermal stress: 1 min. RTA in N2

Permanent degradation: • Prominent IDmax ↓ with T• Positive VTsat shift• IDoff ↑

Same signature as that of degradation mechanism 2 consistent with gate sinking

23

1. Motivation 2. Devices and experimental approach3. Gate stress experiments

• Harsh step-stress (-recovery) experiments• Mild constant gate stress experiment

4. Thermal stress5. Gate current: dominant charge transport

mechanisms6. Conclusions

Outline

24

Virgin device: Thermionic Field Emission fitting

• T dependence well explained by Thermionic Field Emission (TFE) theory• Extracted effective Schottky barrier height (φb): 0.95 eV

25

0.1 0.2 0.3 0.4 0.5 0.610-10

10-8

10-6

10-4

10-2

100

experimentTFE fitting

I G (m

A/m

m)

VGS (V)

T: -50 °C to 200 °C

20 24 28 32 36-40

-36

-32

-28

-24

ln[I sk

Tcos

h(E 00

/kT)

]1/E0 (eV-1)

φb = 0.95 eV

After harsh gate stress: Poole-FrenkelEmission fitting

• T dependence well explained by Poole-Frenkel Emission (P-F) theory• Extracted effective trap energy level (φt): 0.11 eV

26

Stress conditions: VGS,stress = 0 – 2.5 V in 0.1 V steps, VDS = 0 V

0.4 0.5 0.6 0.7 0.8-12

-11

-10

-9

-8

-7T: -50 °C to 200 °C

experimentP-F fitting

ln(|I

G|/V

G)

VG0.5 (V)0.5

20 25 30 35 40 45 50 55

-16

-15

-14

-13

φt = 0.11 eV

y-in

t. [ln

(I G/V

G) v

s. V

G0.

5 ]1/kT (eV)-1

After mild gate stress: Poole-FrenkelEmission fitting

• T dependence well explained by Poole-Frenkel Emission (P-F) theory• Extracted effective trap energy level (φt): 0.36 eV• Close to donor level of N vacancy in AlN of 0.5 eV [T. L. Tansley, PRB

1992] 27

Stress conditions: VGS,stress = 2 V, VDS = 0 V

0.45 0.50 0.55 0.60 0.65-18

-16

-14

-12

-10

experimentP-F fitting

ln(|I

G|/V

G)

VG0.5 (V)0.5

T: -50 °C to 200 °C

20 25 30 35 40 45 50 55-26

-24

-22

-20

-18

-16

φt = 0.36 eV

y-in

t. [ln

(I G/V

G) v

s. V

G0.

5 ]1/kT (eV)-1

1. Motivation 2. Devices and experimental approach3. Gate stress experiments

• Harsh step-stress (-recovery) experiments• Mild constant gate stress experiment

4. Thermal stress5. Gate current: dominant charge transport

mechanisms6. Conclusions

Outline

28

Identified two degradation mechanisms:1. Under mild gate stress:

o Observation: IGoff ↑, trapping ↑, thermally enhancedo Proposed mechanism 1: high electric field induced defect

generation in AlN interlayer

2. Under harsh gate stress:o Observation: IGoff ↑, RD and RS ↑, ΔVT > 0, IDmax ↓, thermally

enhancedo Proposed mechanism 2: self-heating induced Schottky gate

degradation, or gate-sinking

Transport model for IG in low forward regime:• Virgin device: TFE with φb = 0.95 eV• After degradation mechanism 1: P-F with φt = 0.36 eV• After degradation mechanism 2: P-F with φt = 0.11 eV

Conclusions

29