www str soft com

STR Groupwww.str-soft.com

Modeling of III-Nitride MOVPEModeling of III Nitride MOVPE

May 2012

Prehistory of STR:1984: Start of the MOCVD modeling activities at Ioffe Institute, St. Petersburg, Russia

1993-1996: Group for modeling of crystal growth and epitaxy at University of Erlangen-Nuernberg, Germany

History of software development2000: Launch of development of the first specialized software2000: Launch of development of the first specialized software2003: First release of commercial software package 2004: First release of the software for device engineering

STR Today:

2More than 40 scientists and software engineers

STR, GmbH2, Erlangen, Germany INFOTECH Co.3,

Kyunggi-Do, South KoreaSTR G Ltd 1

STR US, Inc.2, Richmond, VA

SimSciD Corporation3, Yokohama Japan

STR Group, Ltd.1, St.Petersburg, Russia

Pitotech, Ltd.3,Chang Hua City, Taiwan

Yokohama, Japan

1 – Headquarter and R&D (consulting service and software development)2 – Local offices3 – Local distributors

Bulk crystal growth modeling (Si Ge SiGe GaAs InP SiC AlN Al O Optical Crystals)

3

Bulk crystal growth modeling (Si, Ge, SiGe, GaAs, InP, SiC, AlN, Al2O3, Optical Crystals)Epitaxy and deposition modeling (Si, SiGe, SiC, AlGaAs, AlGaInP, AlGaInN, high-k oxides)Modeling of device operation (LEDs, Laser Diodes, FETs/HEMTs Shottky diodes)

Software & consulting services :

- Modeling of crystal growth from the melts and solutions: CGSim- Modeling of polysilicon deposition by Siemens process: PolySim- Modeling of bulk crystal growth of SiC, AlN, GaN: ViR- Modeling of epitaxy of compound semiconductors: CVDSim

M d li f t l t i d l t i d i Si LED- Modeling of optoelectronic and electronic devices: SimuLED

Customer base:

• More than 160 companies and universities worldwide• Top LED, LD and solar cell manufacturers • Top sapphire, GaAs, GaP, GaN, AlN and SiC wafer manufacturers• Top MOCVD reactor manufacturers• Top MOCVD reactor manufacturers

OutlineOutline

• Motivation for MOVPE modelingMotivation for MOVPE modeling

• Specific features of nitride growth. CVDSim™ as the dedicated software for simulation of nitride growth by MOCVD

• Modeling of nitride growth in industrial reactors with focus on • Modeling of nitride growth in industrial reactors with focus on AlN/AlGaN growth in CCS reactor

• Unsteady effects in growth of GaN-based quantum-well heterostructures

Basic overview of CFD modelingBasic overview of CFD modeling

• (C)omputation (F)luid (D)ynamics – approach to simulate a MOVPE reactor and process using simulate a MOVPE reactor and process using computer modeling

• CFD is used to model the flow heat transfer and • CFD is used to model the flow, heat transfer and mass transport/chemistry in the reactor

• Day by day in house CFD modeling has started in • Day-by-day in-house CFD modeling has started in MOCVD companies about 10-15 years ago when commercial CFD packages became availablecommercial CFD packages became available

CFD modeling: what for?CFD modeling: what for?

N CFD d li i d b ll h j MOCVD • Now CFD modeling is used by all the major MOCVD equipment manufacturers (Aixtron, Veeco, TNSC, Applied Materials ) to design the new reactorsApplied Materials…) to design the new reactors.

• CFD modeling is used also by MOCVD end-users as Hitachi Cable Osram Opto Philips Lumileds Samsung Hitachi Cable, Osram Opto, Philips Lumileds, Samsung LED and others.

M d li f fl d h t t f i it ti • Modeling of flow and heat transfer is quite routine procedure now, focus is shifted to the combination of CFD with advanced chemistry modelsCFD with advanced chemistry models…

Modeling of III nitride MOVPE: main issuesModeling of III-nitride MOVPE: main issues• Complicated gas-phase chemistry and gas-

Creighton et al.,

(2004)

phase nucleation at high pressures

• Complicated surface chemistry – effect of p ytemperature and carrier gas

• Parasitic deposition in wide range of Zavarin et al., (2005)• Parasitic deposition in wide range of temperature variation

O t t f th d li

Parikh et al., (2006)• Optimization of the growth recipe

Output from the modeling

p g p• Understanding of the underlying mechanisms• Optimization of the reactor design• Optimization of the reactor design• Increase of the process yield

CVDSim™Chemical Vapor Deposition Simulator

Simulation Tool for Modeling of CVD Processes in Industrial Reactors

CVDSim Nitride Edition:CVDSim Nitride Edition:• Prediction of the growth rate and composition• Parasitic reactions and particle formationp• Parasitic deposition on injectors and walls (low temperature kinetics,

condensation of the adducts and non-volatile products)Eff t f th l tti i t h ll iti• Effect of the lattice mismatch on alloy composition

Joint Aixtron/STR presentation at ICMOVPE XV, 2010

Modeling is actively used for the ti i ti d d i process optimization and design

of commercial production-scale multi-wafer reactors

GaN growth rate vs pressure in 6x2” CCS reactor

Combination of simulation and experimental analysis has been applied to improve the

performance of the existing reactors and to develop a novel reactor concept

Contour lines of indium compositionp

The use of the inverted precursor supply allows 40 runs of 1 5 mm thick supply allows 40 runs of 1.5 mm thick

GaN without removing parasitic deposits instead of 10 growth runs for

the conventional growth processthe conventional growth process

Deposition rate over the ceiling(modeling results)

µm/hThickness mapping: from convex to concave profile

µm/h

AlN and AlGaN growth in Close Coupled Showerhead reactorCoupled Showerhead reactor

Kinetic model for AlGaN depositionKinetic model for AlGaN deposition• Adsorption and incorporation of all Ga- and some Al-containing

ispecies• Etching of GaN part of AlGaN alloy, etching rate depends on

iti f AlG Ncomposition of AlGaN

H2 NH3Ga

Dandling bondsg

S f t lSurface crystal lattice layer

Gas-phase reaction mechanism in AlN MOVPEGas-phase reaction mechanism in AlN MOVPE

1.

2.

3.

4.

5.

• Adduct formation• Elimination of methane

1.2.

• Adduct reaction with ammonia• Formation of oligomers• AlN gaseous molecules formation from

3.4.5. g

decomposition of dimers and trimers containing both Al and N

Gas-phase reaction and cluster nucleation Gas-phase reaction and cluster nucleation mechanism in AlN MOVPE

Ways of aluminum losses:formation of oligomers (n3) and AlN particles

S• Species initiating cluster nucleation: AlN vapor• Nucleation and subsequent growth of clusters is due to reactions between AlN nuclei and Al-containing species such as AlN, DMAl:NH2, (DMAl:NH2)2g p , 2, ( 2)2

Gas phase nucleation and cluster Gas phase nucleation and cluster growth/evaporation

Main assumptions used to simulate gas phase nucleation:

Due to their small size, the clusters can be regarded as apseudo gas of heavy molecules

Transport of the clusters and their growth/evaporation aretreated using the first three moments of the size distributionfunction

Particle growth due to interactions with other group-IIIspecies

Effect of the thermophoretic forceEffect of the thermophoretic force

Thermophoretic force prevents particles from reaching the substrate, moving p p p g , gthem in direction opposite to the temperature gradient

surface at ~1000Csurface at 1000 C

Particle density, 3

GaN particle size ranges from 10100 nm, density

Laser scattering experiments,standard MOVPE conditions,

kg/m3from 16108 cm-3sta da d O co d t o s,conventional precursors(TMGa, TMAl, TMIn, ammonia);carrier gas - hydrogen/nitrogen The AlN and AlGaN MOVPE models

were previously successfully verified

J.R. Creighton et al, (2004)were previously successfully verified by the data obtained in the vertical rotating disk and planetary reactors

AlN model verification: CCS 3x2'' reactorAlN model verification: CCS 3x2 reactor

Reaction pathways at different V/III ratiosReaction pathways at different V/III ratios

Low NH3 flow (0.36 slm): TMAl -TMAlNH3 equilibrium shifts TMAlNH3 equilibrium shifts back toward TMAl; growth efficiency is highTMAl concentration

High NH3 flow (6 slm): formation of heavy-molar-mass / low-diffusivity species (i.e. [DMAlNH2]2) followed

AlN t tiby production of AlN vapor and AlN particles

AlN concentration

AlN model verification: CCS 3x2'' reactorAlN model verification: CCS 3x2 reactorEffect of the TMAl flow rate

2,0

2,5 (QNH3=0.36slm)

experiment computation transport limitro

n/hr

Increase of V/III ration results in

1 0

1,5

transport limit

rate

, mic

r

p = 40 Torr

increase of material loss and growth rate reduction

0,5

1,0

AlN

gro

wth

(QNH3=6slm)experiment computationtransport limit

1,82,0

T=1100oC

n/hr exp QTMAl=30mol/min

comp.

Effect of the V/III ratio

0 5 10 15 20 25 30 35 40 45 50 55 600,0

A

QTMAl, mol/min

transport limit

1,01,21,41,6

rate

, mic

ron p

exp QTMAl=15mol/min comp.

p = 40 Torr

TMAl

0,40,60,81,0

AlN

gro

wth

AlN Growth rate depends on ammonia

0 2000 4000 6000 80000,00,2

V/III ratio

flow rate

High growth rates of AlGaN in CCS 3x2'' reactor High growth rates of AlGaN in CCS 3x2 reactor (cooperation with TU Berlin)

Additional losses of gallium:i t ti f G t i i

TMGa/TMAl+NH3+H2

hinteraction of Ga-containing species with AlN particles (proceeds in kinetically-

3x2” Close Coupled Showerhead reactor

limited conditions)

Process parameters:Measurements:i it E iR TT DA UV t Process parameters:

Reactor height (h): 6-21 mmTotal flow: 8 slm Ammonia flow: 1.5 slm

in-situ: EpiR-TT-DA-UV systemex-situ: XRD

Pressure: 50-500 mbarTemperature: 1017 C - 1052 CTemplates GaN/Al2O3Layer thickness: 100-700 nm

Target: growth of AlGaN layers with high aluminum content and growth rate

Effect of the pressure on the AlGaN growth rate and Effect of the pressure on the AlGaN growth rate and composition (h=const)

growth rate compositiongrowth rate composition2

25

30

ate,m

/h

1

1.5

tent

,%

15

20

grow

thra

0.5

1

experiment

Alco

nt

10

15

i

pressure mbar100 150 200 250 300 350 4000

experimentcomputation

pressure, mbar100 150 200 250 300 350 4000

5 experimentcomputation

Both growth rate and composition decrease with pressure due to enhanced intensity of particle generation in the gas phase

pressure, mbar pressure, mbar

Data: J. Stellmach et al., Journal of Crystal Growth 315 (2011) 229J. Stellmach et al., presented at ICMOVPE XV, 2010

Effect of the gap height on the AlGaN growth rate and

th t composition

Effect of the gap height on the AlGaN growth rate and composition (p=const)

growth rate composition

504

ent,

%

30

40

e,m

/h 3

Alco

nte

20

30

experiment (SR)experiment (XRD)

grow

thra

t

1

2

5 10 15 200

10p ( )

experiment (PL)computation

5 10 15 200

1 experimentcomputation

Increase of the reactor height results in lowering the growth rate due to particle formation effects, however, the influence on the AlGaN composition is not so

gap, mmgap, mm

, , pstraightforward

Data: J. Stellmach et al., Journal of Crystal Growth 315 (2011) 229J. Stellmach et al., presented at ICMOVPE XV, 2010

Further adjustment of the growth conditionsFurther adjustment of the growth conditions

5 05,5

insitu SR90

100 XRD

growth rate composition

3 03,54,04,55,0 XRD

h = 6mm

(탆/h

)

60

70

80

90

h = 6mm

insitu SR PL

(%)

h=h1

h=h1

1 01,52,02,53,0

owth

rate

(

20

30

40

50

Al c

onte

nt 1

0,0 0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8 0,9 1,00,00,51,0 Reducing

TMGah = 15mmgr

o

TMAl / (TMAl +TMGa) molar fraction0,0 0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8 0,9 1,00

10

20Reducing

TMGa

ATMAl / (TMAl + TMGa) molar fraction

h = 15mmh=h2>h1 h=h2>h1

TMAl / (TMAl +TMGa) molar fraction TMAl / (TMAl + TMGa) molar fraction

Reproducible growth of AlGaN in the entire compositional range has been achieved with the growth rate above 3 µm/hachieved with the growth rate above 3 µm/h

Data: J. Stellmach et al., Journal of Crystal Growth 315 (2011) 229J. Stellmach et al., presented at ICMOVPE XV, 2010

High growth rates of AlN in planetary 6x2'' reactorHigh growth rates of AlN in planetary 6x2 reactor

Prediction and explanation of the onset of iti h i t f i V/III ti

Growth rate with TMAl flow

parasitic chemistry for various V/III ratios, involving quite low V/III=1.5

The conditions allowing AlN epitaxy with the 6

7

8

9

H2 = 34 slmm/h

r Exp.

G

The conditions allowing AlN epitaxy with the growth rates exceeding 8μm/h have been

found3

4

52

V/III - 1.5

wth

rate

,

1,6

2,0

/hr

Growth rate variation with ammonia

0 100 200 300 400 500 600 7000

1

2

Gro

0,8

1,2

1,6

Exp.Comp.h

rate

, m

/

P=100mbar, T=1045-1070oC

TMAl flow rate, mol/min

0 0

0,4

0,8

H2 = 34 slmTMAl = 165 mol/min

Comp.

Gro

wth

Data: Lundin et al., ACCGE 15, 2011, to be published in Journal of Crystal Growth

10 100 10000,0

NH3 flow rate, sccm

Effect of growth conditions on characteristics of

GaN-based quantum-well heterostructuresGaN based quantum well heterostructures

2012

STR G L dSTR-Group, Ltd.

Epitaxy-to-Device Engineering Modeling

H t t tP i Heterostructurecharacterisics

Process recipe

Pressure in the reactor

Growth temperature of cess

I V

Internal Quantum Efficiency

pthe each layer

Variation of precursor flow rates pr

oc

I-V curve

LED spectrum

flow rates

Introduction of growth interruption stages

VP

E

Carrier gas flow rate and composition

MO

V

Dopant supply M

Variation of the growth temperature and precursor supply specified by epi-engineer for growth of LED heterostructure

T Growth interruption and

Quantum well

BarrierEBL

p-AlGaNQuantum

well

temperature ramping Barrier

TMIn

Typical layer structures of

TEGa

Nominal structure, which can be j t d f th t d t t lib ti f th fl t

Typical layer structures of state-of-the-art InGaN-based visible LEDs

projected from the steady-state calibration of the precursor flow rates

Ga, In, Al

Composition profiles in QW, barrier, and EBL in blue SQW heterostructure grown under various

conditionsDuring QW growth, InN fraction does not reach its steady-state value

Intensive Indium desorption during growth interruption results in sharp

conditions

Recipes simulated in growth of Single QuantumWell structure

1 - QW and barrier are grown at constantt t T 750oC QW(750)/G N(750)

0.25

results in sharp decrease of In content in growing layer

temperature, T=750oC: QW(750)/GaN(750)

2 - QW growth followed by interruption (GI) andnext by barrier growth. QW is grown at 750oC.Temperature was ramped from 750oC up to 800oCnt 0 15

0.2

1 Temperature was ramped from 750 C up to 800 Cduring 30s. Barrier was grown at 800oC:QW(750)/GI/GaN(800)

InN

cont

en

0.1

0.15 1

0.05In fraction exhibits the presence of

QuantumWell

GaN barrier

2

Blue SQW LEDthickness, nm

0 2 4 6 80

the presence of indium (In “tail”) in the barrier

Composition profiles in QW, barrier, and EBL in blue MQW heterostructures grown under

various conditionsvarious conditions

Blue MQW LEDA- Quantum Well, B – GaN Barrier, C – AlGaN EBL

0.25constant T = 750 C + ramping (AlGaN)T(barrier)=800 C + ramping

A B C

onte

nt0.15

0.2 3

InN

co

0.1Recipes simulated in growth of Single

0.054

p g gQuantum Well structure

3 – [QW(750)/GaN(750)]x5/AlGaN

thickness, nm0 10 20 30 40 5004 – [QW(750)/GI/GaN]x5/AlGaN

Emission spectra and peak wavelength of SQW heterostructures grown under various

conditions

conditions

Bl e SQW LED0.8

1.0 1 3

(a.u

.) 12

600

Blue SQW LED

0.4

0.6

n Sp

ectr

um

j 100A/ 2

∆λ

550

600

1 3

h (n

m)

12

0 0

0.2

Emis

sion j~100A/cm2

500

k W

avel

engt

h

400 420 440 460 480 500 520 540 560 580 6000.0

Wavelength (nm)

10-3 10-2 10-1 100 101 102 103 104

450

PeakIn comparison with isothermal growth of QW

and barrier (1), the recipe with growthinterruption followed by high-temperaturebarrier growth (2) results in some depletion

10 10 10 10 10 10 10 10

Current Density (A/cm2)of GaN barrier by Indium and as a result in ashift ∆λ of dominant wavelength. Blue shiftwith current is suppressed

I-V curve and Internal Quantum Efficiency of SQW heterostructures grown under various

conditionsconditions1.0

Blue SQW LED

104 0.6

0.8

QE

102

103

(A/c

m2 )

0.2

0.4 1 3

IQ 12

10-1

100

101

1

ent D

ensi

ty (

1

10-3 10-2 10-1 100 101 102 103 1040.0

Current Density (A/cm2)

2 6 2 8 3 0 3 2 3 410-3

10-2

10 1 3

Cur

re 12

Current Density (A/cm )

In comparison with isothermal growth of QWand barrier (1), the recipe with growthinterruption followed by high-temperature2.6 2.8 3.0 3.2 3.4

Bias (V)

interruption followed by high temperaturebarrier growth (2) results in some depletion ofGaN barrier by Indium and decrease of forwardvoltage

Electrical-Optical characteristics of MQW heterostructures grown under various conditions

1 0

104

0.6

0.8

1.0 9 10

ctru

m (a

.u) 3

4

101

102

103

10

9 10

y (A

/cm

2 ) 34

0.2

0.4

Emis

sion

Spe

c

j~100A/cm2

2

10-1

100

101

urre

nt D

ensi

ty

Along with positionof dominant wave-

400 420 440 460 480 500 520 540 560 580 6000.0

E

Wavelength (nm)2.6 2.8 3.0 3.2 3.4

10-3

10-2C

Bias (V)600

1 0

of dominant wavelength and blueshift of theemission spectrumwith current the

550

600

8 10

gth

(nm

) 34

0.6

0.8

1.0with current, therecipe specified forgrowth of MQWLED has an effecton LED nonideality

500

Peak

Wav

elen

g

0.2

0.4

0.6

910

IQE

34

on LED nonidealityfactor

10-3 10-2 10-1 100 101 102 103 104

450

P

Current Density (A/cm2)

10-3 10-2 10-1 100 101 102 103 1040.0

Current Density (A/cm2)

4

Blue MQW LED

Electrical-Optical characteristics of heterostructures with the samenominal design depend on the recipe specified in MOCVD growth. Inparticular, Indium segregation results in:• Shift of the emission spectra to longer wavelengthShift of the emission spectra to longer wavelength• Increased blue shift of the dominant wavelength with current

Simulations performed demonstrate that the detailed modeling ofdevice structure growth accounting fordevice structure growth, accounting for• unsteady effects like surface segregation,coupled with the

t ti f i t t d• computations of carrier transport, and• light emission spectrais a powerful tool for deliberate control of growth conditions andoptimization of LED characteristics

Th k f ki d tt ti !Thank you for kind attention!