International Symposium on Compound Semiconductors 2012
InGaAs/InP DHBTs with Emitter and Base Defined through
Electron-beam Lithography for Reduced Ccb
and Increased RF Cut-off Frequency
Evan Lobisser1,*, Johann C. Rode, Vibhor Jain2, Han-Wei Chiang, Ashish Baraskar3,
William J. Mitchell, Brian J. Thibeault, Mark J. W. Rodwell
Dept. of ECE, University of California, Santa Barbara, CA 93106, USA
(Now with 1Agilent Technologies, Inc., CA, 2IBM Corporation, VT, 3GlobalFoundries, NY)
Miguel Urteaga
Teledyne Scientific & Imaging, Thousand Oaks, CA 91360
Dmitri Loubychev, Andrew Snyder, Ying Wu, Joel M. Fastenau, Amy W. K. Liu
IQE Inc., Bethlehem, PA 18015
*evan.lobisser@agilent.com, +1 (707) 577-5629
Outline
•
•
•
•
•
Motivation
HBT Design & Scaling
Fabrication Process & Challenge
Electrical Measurements
Conclusion
2
Why THz Transistors?
Digital logic for
optical fiber circuits
0.1-1 Tb/s
optical fiber links
0.3- 3 THz imaging systems
THz amplifiers for
imaging, communications
High gain at microwave frequencies:
Precision analog design, high resolution ADCs, DACs
3
Type-I InP DHBTs at UCSB
Emitter:
n++ InGaAs cap
n InP
XX’:
z
Base:
p++ InGaAs
Doping grade
Drift collector:
n- InGaAs/InAlAs grade
n- InP
Sub-collector:
n++ InGaAs cap
n++ InP
Semi-insulating InP substrate
E
X’
CP
Collector
B
C
BP
Emitter
Base
X
z
4
HBT Scaling Laws
To double bandwidth of a mesa DHBT:
Keep constant all resistances and currents
Reduce 2:1 all capacitances and transport delays
Parameter
collector depletion layer thickness
base thickness
emitter junction width
collector junction width
emitter contact resistivity
base contact resistivity
current density
Change
decrease 2:1
decrease 1.41:1
decrease 4:1
decrease 4:1
decrease 4:1
decrease 4:1
increase 4:1
Epitaxial scaling
Lateral scaling
Surface prep
& doping
Keep lengths the same, reduce widths 4:1 for thermal considerations
5
Epitaxial Design
T(nm)
Material
Doping (cm-3)
Description
10
In0.53Ga0.47As
81019 : Si
Emitter cap
0.5
InP
51019 : Si
Emitter
15
InP
21018 : Si
Emitter
25
9.5
12
InGaAs
1-0.51020
:C
Base
In0.53Ga0.47 As
11017
: Si
Setback
InGaAs / InAlAs
11017
: Si
B-C Grade
1018
: Si
Pulse doping
3
InP
5
45.5
InP
11017 : Si
Collector
7.5
InP
11019 : Si
Sub Collector
41019
: Si
Sub Collector
5
In0.53Ga0.47 As
300
InP
11019 : Si
Sub Collector
3.5
In0.53Ga0.47 As
Undoped
Etch stop
Substrate
SI : InP
0
-0.5
Energy (eV)
15
-1
-1.5
-2
-2.5
-3
-3.5
0
50
100
150
Distance (nm)
Vbe = 1.0V, Vcb = 0.5V, Je = 0, 27 mA/m2
Thin (70 nm) collector for balanced fτ/fmax
High emitter/base doping for low Rex/Rbb
6
Sub-200 nm Emitter Anatomy
Hybrid sputtered metal stack for
low-stress, vertical profile
W/TiW interfacial discontinuity
enables base contact lift-off
High-stress emitters fall off
during subsequent lift-offs
Very thin emitter epitaxial layer
W
for TiW
minimal undercut
TiW
Semiconductor
wet etchfor low ρ
Interfacial Mo
blanket-evaporated
c
undercuts emitter contact
SiNx
100 nm
W
SiNx sidewalls protect emitter contact,
prevent emitter-base
shortsmetal has
Single sputtered
non-vertical etch profile
Mo
7
Lithographic Scaling and Alignment
Positive i-line lithography
Emitter
Negative e-beam lithography
Emitter
Base Mesa
E-beam lithography needed to define < 150 nm emitters and for
< 50 nm emitter-base contact misalignment
Base
Contact
Positive e-beam lithography
Negative i-line lithography
8
TiW
SiNx sidewall
W
Pt/Ti/Pd/Au
Wbc = 150 nm
Web = 155 nm
Tb + Tc = 95 nm
Wbc = 140 nm
Measurement
RF measurements conducted using Agilent E8361A PNA from 1-67 GHz
DC bias and measurements made with Agilent 4155 SPA
Off-wafer LRRM calibration, lumped-element pad stripping used to
de-embed device S-Parameters
Isolated pad structures used to provide clean RF measurements
25
De-embedded
Mason's Unilateral Gain (dB)
Mason's Unilateral Gain (dB)
25
20
Embedded
15
10
5
0 9
10
10
10
Frequency (Hz)
11
10
De-embedded
20
Embedded
15
10
5
0 9
10
10
10
11
10
Frequency (Hz)
10
DC Data
30
Peak f , f

max
14
Dotted: V = 0.2 V
2
cb
10-3
25/30/35 mW/m
12
n = 1.25
20
c
-4
Ic, Ib (A)
10
e
15
V =0V
10
cb
b,step
BV
0
0
0.5
1
8
-5
10
1.5
Vce (V)
ceo
2
= 2.44 V
2.5
6
n = 2.72
-6
A = 150 nm x 3 m
je
I
= 200 A
5
10

J (mA/m2)
25
10-2 Solid: V = 0.0 V
cb
b
10
I
I
b
c

4
10-7
10-8
2
0
0.2
0.4
0.6
Vbe (V)
0.8
1
0
β = 14 for 150 nm junction
VBceo = 2.44 V @ Je = 15 kA/cm2
Rex ≈ 2 Ω·µm2 (RF extraction)
Collector ρsheet = 14 Ω/□, ρc = 12 Ω·µm2
11
RF Data
U
20
H
15
I = 12.4 mA, V
c
10
ce
2
= 1.5 V
J = 27.6 mA/m , V
e
5
cb
= 0.54 V
600
f
f

3.5
3
200
V = 0.5 V
C
cb
cb
max
1010
1011
Frequency (Hz)
max
400
f = 530 GHz

f
= 750 GHz
0
109
4
Ccb (fF)
21
Cutoff frequency (GHz)
MAG/MSG
25
Gains (dB)
4.5
800
30
1012
0
0
5
10
15
20
2
Je (mA/m )
25
2.5
30
Peak RF performance at >40 mW/μm2
Kirk limit not reached
12
Equivalent Circuit Model
ρex = 2 Ω·μm2
Ccb = 3.0 fF
Ajc = 1.86 μm2 ~ 450 nm x 4 μm
(0.2 S)Vbeexp -jω(0.23 ps)
Ic = 12.4 mA
Vce = 1.5 V
Lowest ρex to date due to Mo contact, highly doped epi
Ccb lower than 100 nm collector epi designs due to E-beam litho
13
Performance Analysis
 nkBT

1
nkBT
 b c 
C je  
 Rex  Rc Ccb
2f
qI c
 qI c

230 fs
15 fs
45 fs
τec dominated by transit delays, high ideality factor reduces fτ ~ 10%
Expected base ρc = 4 Ω·μm2 and Rsh = 800 Ω/□
yields fmax > 1.0 THz for same fτ
E
B
Epitaxial design, process damage explain
high ηb, Rbb
Rsh increased by base contacts reacting with
2.5 nm of Pt diffuses ~ 8 nm 5 nm (20 %) of base
14
Conclusion
E-beam lithography used to define narrow emitter, narrowest base mesa
reported to date
Narrow mesa, low emitter ρc enable 33% increase in fmax from previous
UCSB results with 70 nm collector thickness
Epitaxial thinning increased fτ by 10% from 100 nm UCSB designs
1 THz bandwidth possible with improved base contact process
This work was supported by the DARPA CMO Contract No. HR0011-09-C-0060.
Portions of this work were done in the UCSB nanofabrication facility, part of the NSF-funded NNIN
network, and the MRL, supported by the MRSEC Program of the NSF under award No. MR05-20415.
15
Questions?
Extra Slides
Bipolar Scaling Laws
 b  Tb2 / 2 Dn
 c  Tc / 2vsat
Ccb  Ac / Tc
Wbc
Tb
Tc
Wgap
emitter
I c ,max  vsat Ae (Vcb  Vbi ) / Tc2

 Le 
1  ln  
 We 

Rex  c ,e / Ae
P
T 
Le
We   c ,b

Rbb 

  
Le  6
2
12  Ac ,b
 sh  Wbc
We
Wgap
length Le 
V. Jain
Fabrication: Emitter contact
High power
SF6/Ar ICP etch
Low power
SF6/Ar ICP etch
Cr
Cr
Cr
SiOx
SiOx
SiOx
Ti0.1W0.9
Ti0.1W0.9
Ti0.1W0.9
Cl2/O2 ICP etch
EBL
PR
W
W
W
n InGaAs, InP
n InGaAs, InP
n InGaAs, InP
p InGaAs
p InGaAs
p InGaAs
Mo
1
Fabrication: Emitter mesa
SiNx PECVD deposition
CF4/O2 ICP etch
InGaAs wet etch
2nd SiNx sidewall
InP wet etch
Ti0.1W0.9
Ti0.1W0.9
Cr
Ti0.1W0.9
W
W
W
n InGaAs, InP
p InGaAs
2
Base Post Cap
Ccb , po st 
 0 r  Apo st
Tc
Ccb,post does not scale with Le
 Adversely effects fmax as Le ↓
 Need to minimize the Ccb,post value
6
y = 1.09x - 0.02
4
C
cb
(fF)
Undercut below base post
No contribution of Base post to Ccb
2
0
0
1
2
3
L (m)
e
4
5
Transit time Modulation Causes Ccb Modulation
Qbase  constant Qbase
holes
Tc
electrons
Ccb  
Qbaseholes
Vcb
 0 qn( x ) A1  x / Tc dx  VbcA / Tc  f ( I c ,Vcb )
f 

Qbaseholes
C
 cb   f
I c
I c
Vcb
500
cb
400
0.0 V
0.2 V
cb
cb
-0.2 V
300
2
cb

L

Kirk Effect :
 f Vcb  0  Ccb I c  0
0.6 V
f (GHz)
2
f , -0.3 V

200
1
cb
0
eV
eV
ho les
Camnitz and Moll, Betser & Ritter, D. Root
Collector Velocity Modulation :
 f Vcb  0  Ccb I c  0
1
I b , ΔQba se
100
-1
0
2
4
6
8
2
J (mA/um )
10
12
0
-1
e
-2
300
400
Increase in τ c with Vcb  reduced Ccb
- strong effect in InGaAs SHBTs
- weak effect in InP DHBTs
-2
-0.2 V
0
7
100
200
nm
300
400
6
e
200
nm
5
cb
100
C /A (fF/m2)
0
8
4
0.0 V
Increase in Ccb is due to both
0.2 V
- base pushout into collector
3
V = 0.6 V
cb
2
0
2.5
5
7.5
J (mA/m2)
e
10
12.5
- and modulation of τ b by Vcb