Doping of Wide Bandgap Semiconductors GaN
SiC
ZnO
Diamond
n-type dopant
Si on Ga site (~ 15 meV)
N on C site (~ 85 meV)
B on Zn (~30-60 meV)
N (~ 1.7 eV)
p-type dopant
Mg on Ga site
Al on Si site (~ 200 meV)
N on O site (~ 170-200 meV)
B (~ 370 meV)
(160 meV)
GaAs Si 6meV
Si P 45meV
C (Be) B ~28 meV 40meV
Zn ~ 340 meV n-conductivity*
~ 0.002 Ωcm
~ 0.01 Ωcm
~ 0.02 Ωcm
> 1000 Ωcm
p-conductivity*
0.2-2 Ωcm
0.5-2 Ωcm
0.5-40 Ωcm
10-100 Ωcm
* Experimental values
~ 0.001 Ωcm
Why difficult? • Dopant binding (or activation) energy ∝ m* • Maybe the ubiquitous H plays a big role?
Van Der Walle et al. Nature
Commonly Detected Deep Levels in High Quality uid n-GaN Grown by MOCVD and MBE (Steve Ringel, OSU) Defects: dislocation, stacking fault, VN, GaI, antisites, Impurities: C, O, H …, defect complexes: H-VGa, O-VN, H-MgGa … Deep Level Optical Spectroscopy (DLOS) and Deep Level Transient Spectroscopy (DLTS)
EC – ET
Ec •All concentrations reflect typical distributions seen for high quality UCSB material1,2
0.25 eV 0.60 eV 0.90 eV
•Individual concentrations can be significantly varied by growth conditions, impurities, etc.3
1.35 eV
•SI GaN has considerably different trap spectrum4
2.40-2.80 eV 3.04 eV 3.22 eV 3.28 eV
Ev
1013 1014 1015 Approx trap density (cm-3)
1. A. Hierro, et al., APL 76, 3064 (2000) 2. A. Hierro, et al., Phys. Stat.Sol. 228, 937 (2001) 3. A. Hierro, et al., APL 80, 805 (2002) 4. S.A. Ringel, et al., Proc. 2003 ISCS, in print.
Hydrogenic model for impurities Ionization of acceptor or donor impurities can be considered in the same fashion as the electron energy levels and radii are calculated for an isolated hydrogen atom. 4
2
∗
q Z m The impurity levels Edn are given by: Ec − Edn = 2 2 2n (4πε ) where n is a positive integer and Z is the number of unit charge of the ionized donor atom, i.e. Z = 2, for a doubly ionized donor.
Z In terms of hydrogen ∆Edn = Ec − Edn = 13.6 atom ionization energies : nε r 2 n The orbital radii of the r = 0.53 ε r m Å dn Z m∗ electrons are given by:
2
m∗ m
Doping of semiconductors I For a semiconductor in thermal equilibrium, For flatband (charge neutral) condition,
np =
ni2
− Eg = N c N v exp kT
ρ (r ) ∆.E = = N d+ − N a− + p − n = 0 εε 0
For flatband (charge neutral) condition,
N a− 1 = N a 1 + g a exp Ea − E f
For flatband (charge neutral) condition,
N d+ 1 = N d 1 + g d exp E f − Ed kT
[(
[(
) kT ] ) ]
Here ga and gd are the acceptor and donor degeneracy factors. For GaAs: gd = 2, ga = 4 (heavy and light hole bands); For Si (six minima): gd = 12, ga = 4; For GaN: gd = 2, ga = 4
Doping of semiconductors II From the charge neutrality condition:
ρ (r ) ∆.E = = N d+ − N a− + p − n = 0 εε 0
We have,
Nd Na − 1 + g d exp E f − E d kT 1 + g a exp E a − E f
[(
E f − Ec = N c exp kT
) ]
[(
Ev − E f − N v exp kT
) kT ]
The above equation is the generic equation governing the position of the Fermi level in a semiconductor at equilibrium. If the donor and acceptor concentrations and their activation energies are known, then the position of the Fermi level can be calculated. Note that if (Ef – Ed) and (Ef – Ea ) are >> 0, then impurities are fully ionized
p and n type doping of GaN • N-type doping for GaN is simple and similar to other common semiconductors – Si has Ea ~ 20 meV (mobility ~800 – 1000 cm2V-1s-1), Ge not so good for doping due to low incorporation
• P-type doping is complicated and quite different from other semiconductors – Mg has activation energy of ~0.2 eV (mobility few tens of cm2V-1s-1) – Mg forms complexes with hydrogen which has to be broken first, for Mg to act as acceptors – Mg-H complexes can be broken down by • Annealing in nitrogen atmosphere • Low energy electron beam irradiation (LEEBI); ~10KeV, 60 µA
Variation of resistivity with annealing temperature
Technological impact of doping problems • Fabrication of HBTs very difficult (poor base resistance) • Contact resistance high for LEDs and lasers (high series resistance, bad diode characteristics) • Other device structures involving p-type doping are not easy to fabricate • N-type doping of AlGaN with high Al composition is also difficult (activation energy increases with bandgap) • P-type doping for higher Al composition is even more difficult • InN samples are usually degenerately n-type doped. It is difficult to even make a mild p-type device
S.I. GaN… Growth GaN deposited on sapphire substrate typically shows n-type conductivity. SIMS of GaN film near sapphire interface: Oxygen thermally etched or diffused from the sapphire substrate
Residual impurities in precursors and gasses
• minimize O incorporation • compensation via C or dislocations
S.I. GaN… Fe doping Growth of GaN films grown on sapphire, first 0.3 µm Fe doped. Fe conc. (cm-3)
Sheet Resistance (W/sq.)
XRD FWHM (arcsec)
0
3×103
251/491
1.7×1018
2×105
-
1.3×1019
7×109
*
!
253/481
* XRD FWHM for (002) and (102) reflections
Highly insulating films of high structural quality. Sten Heikman et al., UCSB 2002
Crystal growth:n-type doping Ga(In)N: Si – shallow, [Si] < 1019 cm-3 because of SixNy formation and step
poisoning
AlxGa1-xN: Si forms DX center in at x > 0.5 + self-compensation through formation of VGa and VAl = triple acceptors
wide band gap!
C. Stampfl and C.G. Van de Walle, Appl. Phys. Lett. 72, 459 (1998)
1e+19
Example: on Al0.63Ga0.37N-on-sapphire base layers
n [cm-3]
0.3 µm thick AlxGa1-xN:Si films
1e+18
xAl= 0.2 xAl= 0.5
xAl= 0
xAl= 0.58 1e+17 1e+16
xAl= 0.6
0 1e-5
1e-4
Si/(Al+Ga)
1e-3
Crystal growth: p-type doping GaN:Mg ~ 160 - 230 meV deep, H-passivated
(p~ 1 x 1018 cm-3)
shallower in InxGa1-xN (~ 80 – 160 meV)!
GaN:Mg, Xing et al
InGaN:Mg K. Kamakura et al., J.Cryst. Growth 221 (2000) 267
problem: p-type doping of AlGaN with high Al mole fractions
Modulation Doping
Doping statistics
Kozodoy et al. 1999
Ion Implantation in GaN Manufacturability, reproducibility, flexibility and reliability But… •
Accurate dose control
Intrinsically low temperature process Impurity Concentration profile and the structural changes can be tailored Insensitivity to the lattice structure and defects, and the presence of impurities Not constrained by the thermodynamics Simplify the device processing
High resistivity to annealing. Inability to epitaxially recrystallize the implantation induced amorphous layer by annealing Complete recovery requires 1500oC annealing(2/3 of melting point), thus high N2 over 15Kbar N2 overpressure, and proper capping layers
Requirements for implantation doping by activation annealing Remove the implantation-produced compensation defects and the lattice disorder Electrically/optically activate implanted species by moving the interstitial dopants to substitutional sites via short-range diffusion. No considerable redistribution of dopants after post-implantation annealing. Haijiang Yu, UCSB 2004
Thermodynamic analysis and the thermal processing Thermal Thermal Processing Processing Chamber Chamber
Processing Temperature: ~1500oC
Pressure: 1500 psi (or ~100 bar) N2
2 inch wafer accommodation
Temperature 44% ~1500oC
RF Generator
Output power %
Graphite or Moly susceptor Thermocouple in alumina sheath
100 bar N2 1minute Sample surface temperature Thermocouple temperature
Phase Diagram
Expecting point 10 kbar needed
Our pressure
GaN stability at equilibrium
1500oC annealing with 100 bar N2:
- 100 bar N2 ~100oC
- Rapid processing
- 1500oC processing >10 kbar N2
-Thermally stable capping layers (Davydov, et.al., phys.stat.sol.(a) 2001)
Rapid high pressure annealing of planar GaN •
GaN protected: – ~1500oC – ~Optimized – sputtered AlN – ~100 bar N2
~1300oC annealing no AlN cap ~100nm Sputtered AlN ~2um UID GaN Sapphire
Typical UCSB MOCVD GaN template (0002)~300 arcsec
(10 12) :~950 arcsec
Before annealing
~1500oC annealing 100 nm AlN cap
Ion implantation results 2
Rs (Ohm/square)
-3
Si concentration (cm )
10
15
100 keV, 5 X 10
-2
cm
1019 18
10
2
80 60
2
90 cm / V.s
70
40
60 2
80 cm / V.s
20
50
2
17
10
0
o
After 1500 C activation anneal As-implanted
16
10
90
95 cm / V.s
80
0
0.1
0.2
0.3
0.4
0.5
0
2
4
6
8
55 cm / V.s 10 12 14 16 15
Si activation (%)
Si implant in SI GaN at
20
100
105 cm / V.s
100
40
-2
Implanted Si dose (10 cm ) 0.6
Depth (µm) 7 After contact annealing
• Minimal dopant redistribution • Very low sheet resistance: ~ 20 ohm/square, normal doping: ~ 500 ohm/square • Contact resistance: 0.07 ohm mm for non-alloyed contacts 0.02 ohm mm for alloyed contacts normal doping: ~ 0.2-0.5 ohm mm Haijiang Yu et al. APL, 2004
Resistance ( Ω)
6
Before contact annealing
5 4 3 2 1 0
0
10
20
30
40
Distance (µ µm)
50
60
Polarization-induced doping
Debdeep Jena, 2001
Polarization doping Ga-face Al composition x increases
AlxGa1-xN
ρ = −∇ ⋅ P 3D
To maintain space charge neutrality, ns =ρ3D D. Jena et al. Phys Rev. B 67 153306 (2003)
• Interesting from theoretical standpoint to study such 3D charges which was found to have mobility limited by only alloy scattering at very low temperatures
Polarization-induced doping
• Wherever high conductivity bulk material is needed – - in MESFETs, JFETs, Regrown contacts, etc. Debdeep Jena, 2001
Alternative structure: Polarization-doped FET Graded AlXGa1-XN
Advantages over conventional MESFETs GaN
• Higher mobility: no impurity scattering • No carrier freeze-out • High Schottky barrier height • High breakdown field in AlGaN
Advantages over HEMTs • Ability to tailor gm curve for better linearity Charge = 1.6e18 cm-3 Mobility = 800 cm2/V.s
• Different grading schemes lead to different charge profiles
PolFET: Device Characteristics 100 Gain(dB), POUT(dBm)
Gain POUT PAE
24
80 60
20 16
40
12
20 0
4
8
12
16
20
Power-Added Efficiency (%)
28
PIN(dBm)
150µm x 0.7µm devices IDS = 850 mA/mm gm = 93 mS/mm fT = 19 GHz fMAX = 51 GHz
• Output Power = 4.3 W/mm @ 4GHz • PAE = 63% • Gain = 10 dB (VDS = 30V, IDS = 40 mA/mm)
• On sapphire Siddharth Rajan et al. 2004
Nitride MBE (Jim Speck, UCSB) rf plasma MBE growth diagrams • Basic considerations • Morphology control
Bulk transport and growth • n-GaN • p-GaN Polarization reversal
Alloys • AlGaN to AlN 2DEGs (AlGaN/GaN, AlN/GaN) • InGaN Wetting layer and purity issues Challenges: High In composition InGaN InN GaN, InN quantum dots Gen 930 at ND (Albert)
RF MBE GaN Growth Diagram Constant N* = 15.2 nm/min
Accumulation of Ga
Ga-flux (nm/min)
20
18
Ga-droplet Ga-stable
16
C
B
Intermediate Ga-stable
Steady-state Ga coverage
A 14
N-stable 12 550
600
650
Substrate Temperature (°C)
700
FGa < N*
Morphology: AFM Ga-droplet
Intermediate
5µm
Ga-flux (nm/min)
20
18
C
16
B
N-stable
A 14
12 550
600
650
Substrate Temperature (°C)
700
Structure: Cross-Section TEM
Ga-droplet
g(2110)
Intermediate
g(2110) {1013}
g(0110)
500nm
N-stable
Pits initiate at mixed and edge dislocations Decrease in pits Surface kinetics Ga-adlayer coverage
n-GaN: Transport Structure
Growth structure: Hall mobility measured on 30µm x 30µm vdP patterns
MBE-GaN film n-p-n+ isolation layers MOCVD-GaN template
1-2 µm thick NTD = 108 -1010cm-2
Sapphire substrate
Experiments: 1) Determine the effect of Dislocation Scattering by growing on templates with different dislocation densities 2) Determine the effect of growth regime by changing III/V ratio
Doped GaN: Transport Properties nH (cm-3)
nCV (cm-3)
784
2.0x1016
3.7x1016
858
2.9x1016
Doped with Si Ga-droplets Intermediate
15 0 550
1098
600 650 700 750 Substrate Temperature (°C)
800
Both decent n-type and p-type GaN achieved (GaN:Mg was grown under Ga crossover regime and Hall measurement results on the right) B. Heying et al. and I.P. Smorchkova et al.
2.9x1016
851 1.2x1016 semi-insulating
N-stable
2.3x1016
25
2
30
Hole mobility µ p (cm /Vs)
Ga-flux (nm/min)
45
Constant N-flux 3.3 nm/min
µH (cm2/Vs)
20 15 10 5
0
5 10 Hole concentration p (10
17
cm-3 )
15
Why MBE for Mg in GaN? Accurate p-n junction placement is possible
• Very sharp doping profiles
MOCVD
1020
1021
1018 10
Mg Conc [cm-3]
20
10
Concentration (atoms/cc)
1019
~ 9 nm/dec
~ 2.5 nm/dec
3
Mg concentration (atoms/cm )
MBE
1019 18
10
17
10
17
Si Conc [cm-3] 16
10
10
230 nm 400 nm
16
0
100
200 300 400 Depth (nm)
500
600
15
10
0
0.5
1 Depth (µm)
1.5
Growth Transition Series (constant Mg flux): Morphology
5 µm
Increasing Ga flux
Mg Growth Transition Increasing Ga flux
5 µm
‘Crossover’ p = 4.8x1017 cm-3 µ = 13 cm2/Vs
‘Droplet’ p = 5.7x1017 cm-3 µ = 11 cm2/Vs
‘Droplet’ p = 5.4x1017 cm-3 µ = 11 cm2/Vs
Intermediate or N-rich growth: insulating Optimal properties at the crossover!
Polarity Inversion: Sample Structures Single Mg-doped layers
Multiple Mg-doped layers
Ga-stable vs. N-stable
Ga-stable vs. N-stable
0.4 µm Mg-GaN Mg BEP = 1x10-8 torr (for Ga-rich: p = 8-9 x 1017)
120 nm Mg-GaN, Mg BEP = 1.8x10-8 torr 120 nm UID MBE GaN 120 nm Mg-GaN, Mg BEP = 1x10-8 torr 120 nm UID MBE GaN
30 nm UID MBE GaN
120 nm Mg-GaN, Mg BEP = 7.8x10-9 torr 120 nm UID MBE GaN
1 – 2 µm MOCVD GaN
120 nm Mg-GaN, Mg BEP = 3.5x10-9 torr 30 nm UID MBE GaN
Tgrowth = 650ºC
1 – 2 µm MOCVD GaN
N-Rich Single Mg-Doped Layers: Complete Polarity Inversion CBED
X-Section TEM
MBE
MBE MOCVD
MOCVD
0.2 µm
Ga-Rich Single Mg-Doped Layers: ‘Spike-Shaped’ Inversion Domains ID Formation TDs Isolated
Plan-View
Density ~109 cm-2 100 nm
MBE MOCVD
Cross-Section
400 nm
Multilayer Structures: N-rich 120 nm Mg-GaN, Mg BEP = 1.8x10-8 torr 120 nm UID MBE GaN 120 nm Mg-GaN, Mg BEP = 1x10-8 torr 120 nm UID MBE GaN 120 nm Mg-GaN, Mg BEP = 7.8x10-9 torr 120 nm UID MBE GaN 120 nm Mg-GaN, Mg BEP = 3.5x10-9 torr 30 nm UID MBE GaN
1 – 2 µm MOCVD GaN
MBE layers are fully inverted
Multilayer Structures: Ga-rich 120 nm Mg-GaN, Mg BEP = 1.8x10-8 torr 120 nm UID MBE GaN 120 nm Mg-GaN, Mg BEP = 1x10-8 torr 120 nm UID MBE GaN 120 nm Mg-GaN, Mg BEP = 7.8x10-9 torr 120 nm UID MBE GaN 120 nm Mg-GaN, Mg BEP = 3.5x10-9 torr 30 nm UID MBE GaN
1 – 2 µm MOCVD GaN
Inversion begins at first Mg layer
Model for Polarity Inversion: Growth on Dry Surfaces Ga wetting layer development 300 Å
GaN WL
ID nucleation on dry regions
630 Å c+ GaN
Growth c- Mg-GaN
1250 Å
2500 Å
Ga wetting layer develops gradually (Solution: pre-wet surface!)
c+ Mg-GaN
c+ GaN (no Mg)
Alloy Growth: AlGaN Idea: N-flux Ga-flux Al-flux
Excess Ga
• Full range of Al composition available
Al Incorporated Ga 0
0.1
0.2
0.3
0.4
• Composition is controlled with the group III component with the highest sticking coefficient (Al)
0.5
Beam Equiv. Press.(µtorr)
• III/V > 1 • Al comp. = 10 %
0.6
AlGaN/GaN III/V Study All Samples: Al0.12Ga0.88/GaN Ga Rich No Ga Droplets Spirals 77Κ µ= µ=11,200 cm2/Vs Ga Rich Ga Droplets Spirals 77Κ µ= µ=11,200 cm2/Vs
Ga-flux (nm/min)
45 No droplets Ga stable
30 15 0 550
Ga Rich No Ga Droplets AlGaN Spheres 4nm x 150nm 77Κ µ=8000 cm2/Vs
Ga droplets Ga stable
N stable 600 650 700 750 800 Substrate Temperature (°C)
N-flux 3.3 nm/min
Ga Rich No Ga Droplets AlGaN Spheres 5nm x 120nm 77Κ µ=3000 cm2/Vs
Transport Properties of Al.09Ga.91N/GaN Heterostructure
50000
30000 20000 Previous Record
10000
12
nsh 2.2 X 1012cm-2
0 10
1.25 1.20
Rxx/Rxx(B=0)
13
40000
10
4.2 K
60000
Magnetoresistance
10
Shubnikov-de Haas Oscillations
Al.09Ga.91N/GaN Tg = 750°C
Mobility (cm2/Vs)
Sheet Carriers (cm-2)
Temperature Dependent Hall Properties
1.15 1.10 1.05 1.00 0.95
100
Temperature (K)
0.90 0
MBE-AlGaN MBE-GaN Spacer
1
2
3
4
B(T)
MOCVD-Template
nSdH = 2e/(hD(1/B))
ref
5
6
AlN/GaN Growth of AlN under Ga wetting layer
AlN GaN
37 Å AlN
49 Å AlN
77K
300K 100 Å AlN
I.P. Smorchkova et al., JAP 90, 5196 (2001)
Alloy Growth: InGaN Growth Idea:
Example: InGaN Growth
N-flux Ga-flux In-flux
Excess In
• Composition is controlled with the group III component with the highest sticking coefficient (Al>Ga>In) • Full range of In composition available
Ga Incorporated In Beam Equiv. Pressure
C. Poblenz et al. 2003
• III/V > 1 for optimal morphology and properites
In Flux Series: Morphology and Transport
12.9% In, droplets µ = 110 cm2/Vs n = 1.5x 1018 cm-3
Excess In (nm/min)
600ºC 1.1x10-7 Ga BEP Variable In flux
60 40
In droplets In stable
20 0 400
N stable
• • • •
No droplets In stable
N-flux 2.3 nm/min
5.3% In, no droplets µ = 134 cm2/Vs n = 7.0x 1017 cm-3
500 600 Substrate Temperature (°C)
Structure: 0.5 µm InGaN on npn isolation structure 11% In, no droplets µ = 87 cm2/Vs n = 1.56x 1018 cm-3
Hall: 30 µm x 30 µm Greek Cross
0% In, no droplets
All AFM: 5 µm x 5 µm
Towards 2DEGs: AlGaN/InGaN 1.00E+14
Nsh
Al0.12Ga0.88N/In0.06Ga0.94N µ (77K) = 500 cm2/Vs InGaN/AlGaN AlGaN/GaN
1.00E+13
Al0.09Ga0.91N/GaN µ (77K) = 18,470 cm2/Vs 1.00E+12 0.001
0.01
0.1
1/T(K)
Low temperature freeze-out not observed in InGaN
Impurity Incorporation in MBE-In0.17Ga0.83N: SIMS Ga B.E.P.
N
N
In
In
5
5
10
150 nm
4
10
Oxygen 3
10
Indium 200
GaN 750°C
4
Oxygen 3
10
Indium
Boron
2
10
0
150 nm
10
Boron 2
InGaN 600°C
GaN 600°C
MOCVD GaN
Indium Desorption
Counts per second
Counts per second
10
GaN 750°C
InGaN 600°C
GaN 600°C
No Indium Desorption
MOCVD GaN
B.E.P.
Ga
400
600
Depth (a.u.)
800
1000
10
0
200
400
600
Depth (a.u.)
800
1000
Sensitivity of Ga Wetting Layer to Excess In Desorption of excess metal
Ga-rich GaN growth at 600ºC No desorption
Counts per second
1 .4 x 1 0
4
1 .2 x 1 0
4
1 .0 x 1 0
4
8 .0 x 1 0
3
6 .0 x 1 0
3
4 .0 x 1 0
3
2 .0 x 1 0
3
MOCVD GaN
GaN
GaN (In)
GaN
GaN (In)
GaN
150 nm
AlGaN markers
(In) indicates excess In present during growth
Oxygen Aluminum Boron
0 .0 0
500
1000
D e p th ( a .u .)
1500
2000
Structure of Ga and In Wetting Layers Ga-rich growth
In-rich growth
Laterally contracted Ga bilayer
In wetting layer
Why such high UID in InGaN? Difference in wetting layer (Ga bilayer vs In ‘monolayer’)! J.E. Northrup, J. Neugebauer, R.M. Feenstra, and A.R. Smith, Phys. Rev. B 61, 9932 (2000) H. Chen, R.M. Feenstra, J.E. Northrup, T. Zywietz, J. Neugebauer, and D.W. Greve, J. Vac. Sci. Tech. B 18, 2284 (2000)
Effect of Ga Flux on InGaN Composition and Growth Rate 0.18 0.16
%In by XRAY
0.14 0.12
0.14
0.1
0.08 0.06
0.07
0.04
Growth Rate (µ µm/hr)
In (BEP) = 4.05 x 10-7 Torr In Droplet Regime, 600°C
Maximum of 17% In incorporation at 600°C
0.02 0 0.00E+00
2.00E-08
4.00E-08
6.00E-08
8.00E-08
1.00E-07
1.20E-07
Ga Flux (B.E.P.)
FGa = 0
FGa > 0
Complete In wetting layer: ‘monolayer’ structure
Ga + In wetting layer: Complete Ga wetting layer: structure unclear fluid-like bilayer
Origin of Impurities Source
Impurity Possibility Reason ?
Outgassing of Ga, In, or N source
O, B
NO
SIMS
Background pressure in the system
O
NO
System base pressure is in mid 10-11 Torr range
O, B
NO
TEM
N source gas
O
YES
Low cracking efficiency of plasma source
N source
B
YES
Plasma could sputter crucible during growth
Reduced growth rate in the InGaN layer
~1% N source efficiency: 1018 UID = 1 ppm oxygen from total N source