Chemical Sensing and Advanced Process Control forAlGaN/GaN HEMT Manufacturing

Soon Cho, Michael E. Aumer, Darren B. Thomson, Dan Janiak, Gary W. Rubloff, Deborah P. Partlow

[Lab orResearch Group

Logo]

High Performance Applications

GaN-based electronic devices for high frequency, high power applications, such as radar electronics

Challenges

Growth of GaN epi films and GaN-based alloys with reproducibility and quality constraints that are much higher relative to optoelectronic applications

deposition pattern generation etching metrologymetrology

run-to-runfeedback control

run-to-runfeedforward

control fault classification& response

real-timecontrol

sensor

real-time faultdetection

sensor

Approach

Advanced Process Control (APC) already pervasive in the Si ULSI industry is applied to the GaN-based processes to bring reproducibility and process optimization

SensorSensor--based based REALREAL--TIMETIME Advanced Process ControlAdvanced Process ControlGaN MOCVD Process

∫ (Signal) dt

Film

Thi

ckne

ss

TARGET

Metrology Model

Post-Process Characterization

Real-time In-situ Mass Spectrometry

Process Variability is a Concern...

AlGaN Cap Layer Thickness MetrologySensorSensor--based metrology can predict/control AlGaN thickness to based metrology can predict/control AlGaN thickness to within 3A or 1%within 3A or 1%

GaN-based Heterostructure Design

Multiple material parameters across the multi-layer structure must be co-optimized Direct consequences in terms of device performance

0.01.6 1.8 2.0 2.2 2.4 2.60

100

150

200

250

300

350

G379

G380

G382

G383

G386

Better Quality

Thick film anomaly

Gan

XR

D F

WH

M @

(002

) (ar

csec

)

Methane/Ethane ratio during GaN epitaxy

XRD FWHMXRD FWHM

0.01.6 1.8 2.0 2.2 2.4 2.605

6

7

8

9

10

G379

G380

G382

G383

G386

Better Quality

GaN

PL

band

edg

e in

tens

ity (a

.u.)

Methane/Ethane ratio during GaN epitaxy

PL Band EdgePL Band Edge

0.0 5.0x10-4 1.0x10-3 1.5x10-3 2.0x10-305

6

7

8

9

10

G379

G380

G382

G383

G386

Better Quality

GaN

PL

band

edg

e in

tens

ity (a

.u.)

Normalized H2O level at the end of pre-deposition anneal

Band Edge

0.01.0x10-5 1.2x10-5 1.4x10-5 1.6x10-5 1.8x10-5 2.0x10-50.00

0.01

0.02

0.03

0.04

0.05

G379G380

G382G383

G386

Better Quality

GaN

PL

deep

leve

l int

ensi

ty (a

.u.)

Normalized O2 level at the end of pre-deposition anneal

Deep Level

… GaN Epilayer Photoluminescence Quality PredictionBand edge and deep level must be optimized concurrently

Pre-process Contamination Control is a key

Low impurity condition must be established prior to GaN growth sLow impurity condition must be established prior to GaN growth steptepCan predict, but little room for course correction during GaN sCan predict, but little room for course correction during GaN step itselftep itself

•• Reduce impurity levels in reactor through preReduce impurity levels in reactor through pre--process purge, process purge, anneal, and R2R bakeanneal, and R2R bake--outout

•• Watch for impurity levels to drop below critical limit to guaranWatch for impurity levels to drop below critical limit to guarantee tee acceptable PL qualityacceptable PL quality

CONCLUSIONS

• In-situ metrology is key to achieving real-time APC• GaN for advanced electronic application faced with serious

manufacturing hurdles

• Chemical sensing in GaN MOCVD for Advanced Process Control• Accelerated learning at R&D stage• Manufacturing reproducibility• Insight into intrinsic chemistry & process

• Extensive fault detection & management in use already• Precision metrology for critical AlGaN layer thickness (20-25nm)

to ~1% in real-time demonstrated• Accurate prediction of product quality (crystal quality, PL) in

real-time demonstrated

• APC benefits exploited by Northrop Grumman development• AlGaN thickness control• Pre-process contamination control

GaN Epilayer Crystal Quality Prediction• Based on intrinsic process chemistry – two major

byproducts Methane & Ethane representing different reaction pathways

•• Monitoring the Me/Et ratio predicts product crystal quality to Monitoring the Me/Et ratio predicts product crystal quality to ~ 5% precision~ 5% precision

GaN Epilayer Photoluminescence Quality Prediction …Same Me/Et ratio predicts PL band edge intensity to ~ 5% precisiSame Me/Et ratio predicts PL band edge intensity to ~ 5% precisionon

Lower Me/Et ratio produces better quality (XRD, PL band edge)Lower Me/Et ratio produces better quality (XRD, PL band edge)Surface reaction (Ethane) pathway desiredSurface reaction (Ethane) pathway desired

Band Edge

0.01.5x10-3 2.0x10-3 2.5x10-3 3.0x10-3 3.5x10-05

6

7

8

9

10

G379

G380

G382

G383

G386

Better Quality

GaN

PL

band

edg

e in

tens

ity (a

.u.)

Normalized H2O level during GaN epitaxy

Deep Level

0.0 1.6x10-5 1.8x10-5 2.0x10-5 2.2x10-5 2.4x10-50.00

0.01

0.02

0.03

0.04

0.05

G379G380

G382G383

G386

Better Quality

GaN

PL

deep

leve

l int

ensi

ty (a

.u.)

Normalized O2 level during GaN epitaxy