Download - Low Noise Avalanche Photodiodes
Low Noise Avalanche Photodiodes
Low Noise Avalanche Low Noise Avalanche PhotodiodesPhotodiodes
John P. R. DavidJohn P. R. DavidElectronic & Electrical EngineeringElectronic & Electrical Engineering
University of Sheffield, U.K.University of Sheffield, U.K.
Talk OutlineF APD backgroundF Low noise mechanisms in
thin p+-i-n+sF Temperature dependenceF APD speedF Conclusions
Talk OutlineF APD backgroundF Low noise mechanisms in
thin p+-i-n+sF Temperature dependenceF APD speedF Conclusions
Low Noise Avalanche Photodiodes
National Centre for III-V Technologies,University of Sheffield, UK
H Established in Department of Electronic & Electrical Engineering, University of Sheffield in 1978
H Mission: To provide III-V wafers & devices to the UK academic community
H Current Capability: 2 MBE, 3 MOVPE, Device Fabrication, Characterisation
H Staff: 10 scientists, 6 techniciansH Growth output: 750 wafers/yearH Optical wafers & devices: Lasers, LEDs,
VCSELs, RC-LEDs, waveguides, modulators, AFPMs, pins, APDs, Q-Dot lasers, Q-Cascade lasers
H Electrical wafers & devices: HBTs, HEMTs, diodes
H Established in Department of Electronic & Electrical Engineering, University of Sheffield in 1978
HH MissionMission: To provide III-V wafers & devices to the UK academic community
HH Current CapabilityCurrent Capability: 2 MBE, 3 MOVPE, Device Fabrication, Characterisation
HH StaffStaff: 10 scientists, 6 techniciansHH Growth outputGrowth output: 750 wafers/yearHH Optical wafers & devicesOptical wafers & devices: Lasers, LEDs,
VCSELs, RC-LEDs, waveguides, modulators, AFPMs, pins, APDs, Q-Dot lasers, Q-Cascade lasers
HH Electrical wafers & devicesElectrical wafers & devices: HBTs, HEMTs, diodes
180 miles180 miles
Low Noise Avalanche Photodiodes
APD Research at Sheffield
•• Impact ionization coefficient investigation • New materials and novel structures• ‘Dead-space’ characterization• Excess noise m easur ements • Temperature dependence• Analytical and numerical m odelling• Low-noise, high -speed avalanche photodiodes• Single photon avalanche photodiodes ( SPADs)
Devices
Physics
Low Noise Avalanche Photodiodes
Generic communication system
Signal carryingmedium
Transmitter Receiver
inputdata
bits/s
output
Repeaterspacing X km
Communication capacity - Bit rate length product= repeater spacing*data rate {X [km]*Bandwidth [b/s]}
Using an APD can increase repeater spacing
Low Noise Avalanche Photodiodes
p-i-n vs. APD
Reverse Bias Voltage
Ph
oto
curr
ent
p-i-n operating range
APD operating range
p-i-n] low voltage] bias insensitive] temperature
insensitive] simple bias circuit] fast] simple bias circuit] cheap
APD] high sensitivity] single photon detection
(possibility)
Low Noise Avalanche Photodiodes
Technology Comparison
PhotomultipliersPhotomultipliers
+ High gain (~106)+ Low dark current+ Low noise? Reliability
+ Inexpensive+ Compact+ Rugged+ High detectivity+ High reliability+ Simpler, cheaper filters+ Reasonable gain+ High efficiency+ Low voltage (<100V)
Avalanche Avalanche photodiodesphotodiodes
- Expensive - Poor efficiency- Complex filters- Bulky- Fragile- High voltage (~1kV) - High noise
- Temperature sensitive
Low Noise Avalanche Photodiodes
Comparison of three detector types
PIN-FET Ge APD InGaAs APD
Sensitivity X dBm (X+4) dBm (X+8) dBm
Cost Moderate Moderate High
Wavelength 1.3µm & 1.5µm 1.3µm 1.3µm & 1.5µm
Reliability 1011 hrs 106hrs 106hrs
u An InGaAs APD requires 8dB less optical power to produce the same signal as a PIN-FET.
u PIN-FET is cheaper, faster and more reliable.
Low Noise Avalanche Photodiodes
p-i-n vs. APDSe
nsiti
vity
(Se
nsiti
vity
( dB
mdB
m))
Bit Rate (GB/s)Bit Rate (GB/s)
Comparison of sensitivity
ð APDs can have an extra -8dBm sensitivity c.f. p-i-n
ð APD advantage reduces as speed increases
ð APDs can have an extra -8dBm sensitivity c.f. p-i-n
ð APD advantage reduces as speed increases
Low Noise Avalanche Photodiodes
InGaAs/InP SAM-APD
Low Noise Avalanche Photodiodes
Electric field distribution in a SAM-APD
EInP = 400kV/cm
EInP = 350kV/cm
EInP = 150kV/cm
EInGaAs ~230kV/cm
EInGaAs <170kV/cm
Ele
ctri
c F
ield
wm≈ 500nmwc=150nm
wa ≈ 1500nm
EInGaAs = 0kV/cm
InGaAs absorption region
Q1.5-1.1 grade
InP
multiplication
region
Slow, poor quantum efficiency, low gain
Optimum speed, quantum efficiency & gain
Slow, high gain, high dark current
Low Noise Avalanche Photodiodes
Punch-through ~ 16V Breakdown ~ 39V
SAM-APD Current-Voltage Characteristics
Reverse bias (V)10 20 30 40
Pho
to/D
ark
Cur
rent
(A)
10-11
10-10
10-9
10-8
10-7
10-6
10-5
10-4
10-3
10-2
0
20
40
60
80
100
Photocurrent
Dark current
Gain
Vph ~ 16V
Photocurrent and dark current
Low dark current even at high multiplication gains (> 50) (< 70nA @ 0.9 Vbd)
Low dark current even at high multiplication gains (> 50) (< 70nA @ 0.9 Vbd)
Low Noise Avalanche Photodiodes
APD Vendors
Vendors Sensitivity(dBm)
VBD (V) ∆ VBD (V/°C) 3DB bandwidth(GHz)
Misc:
JDSU ERM577 -32 40-70 1.8 10nA@ VBD -1.58.5uA/uW@ VBD -1.5
MitsubishiFU319SPA
-33 35-75 0.12 1.9
AgereP173A
-34 45-70 0.07-0.14 2
FujitsuFRM5W232BS
-34 50 0.12 2.5
Nova CrystalsNVR251
- 29-32 0.03 Fused InGaAs/Si
NEC 8501 40-80 0.1 3 [email protected] VBD
Alcatel 1914 30-75 3.5 [email protected] VBDSensors UnlimitedSU-10ATR
-24 25-42 6 5nA@ VBD -1
MitsubishiFU321SPA
-25 20-40 0.05 7
JDSU ERM578 -22 20-40 7NEC 4270 -25 16-32 0.02 8 Superlattice APD
(k=0.4)
Low Noise Avalanche Photodiodes
Ionization - Threshold energy
final electron valence electron
hole
hot electron
mlh
mhh
ms-o
me
Eth
Eg
EEgg = Band gap= Band gapEEthth = Threshold energy= Threshold energy‘minimum energy for ‘minimum energy for ionization’ionization’
For parabolic bands For parabolic bands and equal masses,and equal masses,EEthth = 1.5 = 1.5 EEgg
For For ionizationionization::1)1) energy and momentum conservationenergy and momentum conservation2)2) minimisation of energyminimisation of energy
Impact ionization schematic diagram
Low Noise Avalanche Photodiodes
High field region 1injectedelectron
7 collectedelectrons
Me = 7
electronhole
tim
e
Multiplication buildup time required to achieve M
Multiplication process
J Multiplication= current out/current in
J Avalanche takes time to build-up
J Multiplication= current out/current in
J Avalanche takes time to build-up
Low Noise Avalanche Photodiodes
Excess avalanche noise
Reverse bias in Volts0 10 20 30 40 50 60
Mea
n M
ultip
licat
ion,
<M
>
0
5
10
15
20
25
30
35
Mean Multiplication <M>Noise on Multiplication
Multiplication buildup time required to achieve M Ø An APD can give us
gain.
Ø Unfortunately the avalanche noise can degrade the S/N ratio.
Ø An optimum value for < M> exists.
Ø An APD can give us gain.
Ø Unfortunately the avalanche noise can degrade the S/N ratio.
Ø An optimum value for < M> exists.
Low Noise Avalanche Photodiodes
An ideal avalanche photodiode
Mean SquareNoise current=
2 2eI BMphoto
APDbiasedto Gain = M
Iphoto MIphoto
noise current=
2eI Bphoto
/Non ideal APDS = 2eIphotoBM2 F
where F = excess noise factor
/For an ideal APD F = 1
/Non ideal APDS = 2eIphotoBM2 F
where F = excess noise factor
/For an ideal APD F = 1
Low Noise Avalanche Photodiodes
ò APD’s rely on internal gain to improve S/N ratio
ò Impact ionization process ⇒ stochastic ⇒ avalanche noise
ò Excess avalanche noise limits APD’s maximum useful gain, M
ò In bulk structure, large β/α (or α/β) ratio required for low excess noise
ò APD’s rely on internal internal gaingain to improve S/N ratio
ò Impact ionization process ⇒ stochasticstochastic ⇒ avalanche noise
ò Excess avalanche noise limits APD’s maximum useful gain, M
ò In bulk structure, large β/α (or α/β) ratio required for low excess noise
Background
Multiplication factor
Log
(Pow
er) AP
D noise
APD signal
M=1 Mopt
OptimumSNR
Preamplifiernoise floor
Reverse bias voltage
Mea
n m
ultip
licat
ion Ideal
Real
Low Noise Avalanche Photodiodes
Photodetectors - S/N
p-i-n photodiode and amplifier
equiv. amp. noisesource
__ i2
AMP
A[W] Vo( )S
N
I
iph
amp
=
2
2( )
Iph
__ i2AMP__
i2AVAL(M)
M VoA[W]
( )22
2
AVALamp
ph
ii
MINS
+=
Iph
Avalanche photodiode and amplifier
Low Noise Avalanche Photodiodes
McIntyre’s avalanche noise theory (1966)
F M Mk
kM
M( ) (
( ))= +
− −
11 1 2
k = βα
ê α = electron probability of ionization per unit length [m-1]
ê β = hole probability of ionization per unit length [m-1]
where
Assumes:Ê Multiplication process does not depend
on carrier history.Ë k = β /α is a constant
Low Noise Avalanche Photodiodes
McIntyre’s model for electron injection F v Me
Multiplication, M
2 3 4 5 6 7 8 91 10
Exc
ess
Noi
se F
acto
r, F
1.5
2.5
3.5
4.5
1.0
2.0
3.0
4.0
5.0
k = 0
k = 1
0.4
0.2
0.6
The excess noise depends only on the ionization coefficient ratio (k) and the multiplication value.Larger ionizing carrier type should initiate avalanche.
k = β / ak = β / a
Low Noise Avalanche Photodiodes
GaAs Ionization Coefficients
1/E (cm/V)
0e+0 2e-6 4e-6 6e-6
Ioni
zati
on c
oeff
icie
nts
(cm
-1)
102
103
104
105
106
β
α
α≈β
0.5µm
1µm
0.1µm
k = β/α = 1 0.9 0.8
Field dependence of GaAsionization coefficients
v Most III-V semiconductors have 0.4 ≤ k ≤ 2.5
v High excess noise expected, especially at higher electric fields when k→1
Low Noise Avalanche Photodiodes
Ionization Coefficients
Reverse Voltage (V)0 5 10 15 20 25 30 35
Pho
tom
ultip
licat
ion
(arb
. uni
ts)
1
2
3
4
5
6
M e(Si)
Mh(Si) M e(GaAs)
Mh(GaAs)
α = −−
1 1W
MM M
MM
e
e h
e
h
ln
β =−
−
1 1W
MM M
MM
h
h e
h
e
ln
Me and Mh for 1µm Si and GaAs pinsSimple relationship between α, β andmultiplication characteristics in pins.
GaAs & Si have similar Vbd butvery different α/β ratios.
Low Noise Avalanche Photodiodes
Excess noise in Si and GaAs, Me
Multiplication, M
2 3 4 5 6 7 8 91 10
Exc
ess
Noi
se F
acto
r, F
1.5
2.5
3.5
4.5
1.0
2.0
3.0
4.0
5.0
McIntyreGaAs Si
k=0
0.2
0.4
0.60.8k=1.0
Ü In thick structures, the excess noise F is determined by k, the β/α ratio.
Ü Silicon has a small kcompared to GaAs, hence low noise.
Low Noise Avalanche Photodiodes
Enhancement of Ionization by MQW’s
Inverse electric field (x10-5 cm/V)
0.2 0.3 0.4 0.5 0.6
Ele
ctro
n io
nisa
tion
coef
ficie
nts
(/cm
)
101
102
103
104
105
Bulk GaAs
MQWCapasso's data
AlGaAs
GaAs∆Ec
∆Ev
Ec
Ev
p+
n+
electrons
Chin et al. (1980) postulated that a large êEc would enhance α
Capasso et al. (1982) reported a large enhancement in α ?in
AlGaAs/GaAs MQW’s
Low Noise Avalanche Photodiodes
Schematic of a SAM-APD
JLight is absorbed in thick InGaAs layer
JPhotogenerated holes impact ionize in InP
JConventional designs involve thick multiplication layers, so that α/β ratio is small, to achieve low excess noise
JLight is absorbed in thick InGaAs layer
JPhotogenerated holes impact ionize in InP
JConventional designs involve thick multiplication layers, so that α/β ratio is small, to achieve low excess noise
Low Noise Avalanche Photodiodes
p-i-n diode schematic
Pure Me & Mh obtained by illuminating thick p+ & n+
layers with short wavelength illumination.
n+-i-p+ s also grown to obtain Mh more easily.
VR
1µm p+intrinsic
1µm n+
hv
hv
Me
Mh
Intrinsic thickness, w varies from 1µm to 0.05µm.
Low Noise Avalanche Photodiodes
Multiplication from GaAs p+-i-n+s
Reverse bias (V)0 5 10 15 20 25 30 35
Mul
tipl
icat
ion,
M
0
1
2
3
4
black: Me, red: Mh
1µm
0.1µm
0.5µm
0.025µm
0.05µm
Multiplication factors Ø Me and Mh were measured in different thickness p+-i-n+s.
Ø Lock-in techniques allow Meand Mh to be determined in the presence of large dark currents.
Ø Me ≈ Mh as ‘w’ decreases, suggesting that α ≈ β
Low Noise Avalanche Photodiodes
Excess noise in GaAs p+-i-n+s
Multiplication, Me
2 3 4 5 6 7 8 91 10
Exc
ess
Noi
se F
acto
r, F
1
2
3
4
5w = 1µmw = 0.5µmw = 0.3µmw = 0.2µmw = 0.1µmw = 0.049µm
Expected convergencewith decreasing w
k = 0
k = 1 0.4
0.2
0.6
The excess noise decreases as w decreases, instead of increasing as k→1
Electron initiated noise measurements showed unexpected and significant noise reduction as w became smaller
Low Noise Avalanche Photodiodes
Excess noise in GaAs n+-i-p+s
The excess noise decreasesas w decreases, instead of increasing according to k.
Hole initiated noise measurements also showed unexpected and significant noise reduction as w became smaller
Behavior cannot be explained by McIntyre theory
Multiplication, Mh
2 3 4 5 6 7 8 91 10
Exc
ess
Noi
se F
acto
r, F
1
2
3
4
5w = 0.2µmw = 0.1µmw = 0.05µm
Expected convergencewith decreasing w
k = 0
k = 1 0.4
0.2
0.6
Low Noise Avalanche Photodiodes
Multiplication characteristics in InP
Reverse bias (V)0 20 40 60 80 100
Mul
tiplic
atio
n, M
2
4
6
8
0
10
w = 0.24µmw = 0.33µmw = 0.48µmw = 0.90µmw = 2.40µm
Fujitsu SAM-APD
Measured Me (symbols)
Calculated Me (solid lines)using bulk ionization coefficients
Measured Me (symbols)
Calculated Me (solid lines)using bulk ionization coefficients
Multiplication factors
Low Noise Avalanche Photodiodes
Excess noise factor in InP
Multiplication, M
1 10
Exc
ess
nois
e fa
ctor
, F
5
10
15k=2.4
k=1
k=0
decreasing w
Fujitsu SAM APD
k=0.4
F(M)ð Same symbols as beforeð Noise measured using
wrong (electron) carrier type
ð Fujitsu SAM-APD gives β/α = 1.4 ∴ keff = 0.7
ð Structure with w = 0.24 gives keff = 0.4 - much better than SAM-APD with hole multiplication
ð Low noise possible even with electron injection with thin w
ð Same symbols as beforeð Noise measured using
wrong (electron) carrier type
ð Fujitsu SAM-APD gives β/α = 1.4 ∴ keff = 0.7
ð Structure with w = 0.24 gives keff = 0.4 - much better than SAM-APD with hole multiplication
ð Low noise possible even with electron injection with thin w
Low Noise Avalanche Photodiodes
Multiplication characteristics in Silicon
Reverse bias (V)
5 10 15 20 25 30
Mul
tiplic
atio
n, M
0
2
4
6
8
10
w = 0.18µm
w = 0.12µm
w = 0.32µm
w = 0.84µm
(n+-i-p+)
Measurement of Mh and Mmix on 0.84µm n+-i-p+
Measurement of Me and Mmix on 0.32, 0.18, 0.12µm p+-i-n+s
Multiplication factors
Low Noise Avalanche Photodiodes
Local noise model prediction vs. experiment in submicron Si p+-i-n+s
Multiplication, M2 4 6 8 10
Exc
ess
Noi
se F
acto
r, F
2
4
6
8
10
k = 0.2
k = 0
blue: Me
black: McIntyre model
Fe(Me) 4 Local field noise model gives increasing excess noise from k = 0.4-0.7 as w decreases from 0.32-0.12µm.
4 Experiment shows that F(Me) however is virtually constant at k ≈ 0.2.
Low Noise Avalanche Photodiodes
Modeling of thin APD behavior
Modeling of thin APD behavior
Low Noise Avalanche Photodiodes
McIntyre Noise Model
x
Pro
babi
lity
of io
niza
tion
Electronsα1
x
Pro
babi
lity
of i
oniz
atio
n
Holes
β1
H McIntyre’s noise model assumes that a carrier’s ionization probability is independent of distance probability density function (PDF) is exponential
H This assumption leads to the McIntyre expression for excess noise factor
H Avalanche noise depends on the β /α ratio
Probability density function of ionization
Low Noise Avalanche Photodiodes
Dead Space Models
x
Pro
babi
lity
of io
niza
tion
deadspace
x
Pro
babi
lity
of io
niza
tion
deadspace
α1
β1
�More realistic picture of ionization probability shows significant dead space at high electric field
�Presence of dead space reduces CoV makes multiplication more deterministic less noisy
�A significant dead space reduces the importance of the β/α ratio & the carrier type initiating multiplication
Probability density function of ionization
Low Noise Avalanche Photodiodes
Monte Carlo Estimation of F
w
Excess Noise Factor,
l Multiplication viaimpact ionization
1NeNh
NMMMM
M NN )...( 2121 −− ++++>=<
NMMMM
M NN )...( 22
21
22
212 −− ++++
>=<
Mtrial = 1 + Ne + Nh
2
2
><><
=MM
F
Low Noise Avalanche Photodiodes
Probability distribution of electron ionization path lengths (<M> = 5)
0.0 0.5 1.0 1.5 2.00
1
2
3
Electron Ionization Path Length, le 0.00 0.02 0.04 0.06 0.08
Nor
mal
ised
Pro
babi
lity,
µm
-1
0
20
40
60
d
d
High Field: E = 960 kV/cm
Low Field: E = 380 kV/cm
(µm)
(µm)
<le> = 0.39µmCoV = 0.86
<le> = 0.032µmCoV = 0.31
Probability density function of ionization
CoV = stand. dev. in le / <le>CoV = stand. dev. in le / <le>
ÜAt low fieldsð relatively small dead space & low ionization probability
ÜAt high fieldsð relatively large dead space & higher ionization probability ðnarrow ionizationprobability distribution.
Low Noise Avalanche Photodiodes
Distribution of Multiplication for <M> = 5
0.0001
0.001
0.01
0.1
1
Multiplication, M e
0 10 20 30 40 50 60 70
Prob
abili
ty fu
nctio
n of
Mul
tiplic
atio
n, P
e(M
)
0.0001
0.001
0.01
0.1 w = 0.05µm E = 960 kV/cm F = 2.008
w = 0.5µm E = 380 kV/cm F = 2.933 2 There are more high
order multiplication events at lower electric fields, giving rise to more noise
Probability function of multipliplication
Low Noise Avalanche Photodiodes
Electr ic F ie ld kV/cm4 0 0 600 8 0 0 1 0 0 0
Pat
h le
ngth
, µ
m
0.01
0.1
1
Sca
tter
s/io
niz
atio
n e
ven
t1 0 1
1 0 2
1 0 3
1 0 4
Typical path lengths as a function of electric field
Monte Carlo model results
Scatters per ionization eventBallistic dead space
d = 2.1eV/qEMean ionization path length <le>
2 Scattering becomes less important as the electric field increases
2 Ionization tends towards ballistic ideal, i.e. like PMT
Low Noise Avalanche Photodiodes
Excess noise in p+-n diodes
P or N doping varies from2 x 1016 cm-3 to 3 x 1017 cm-3.
VR
1µm p+p / n type
1µm n+
hv
MeøReducing w in p+-i-n+s
reduces excess noise.
øHow does increasing doping i.e. electric field gradient affect noise?
Low Noise Avalanche Photodiodes
Electron Multiplication, Me
0 5 10 15 20
Exc
ess
Noi
se F
acto
r, F
0
5
10
15
Nd=2x1016cm-3
Nd=2x1017cm-3
Nd=1x1017cm-3
Nd=5x1016cm-3
P-I-Ns: w=1µm & 0.1µm
Simulated excess noise in P+N junctions(... primary electrons are injected into the HIGH field)
The excess noise is reducedby increasing the doping
For the same total depletionthickness, F(P+N) < F(P+-I-N+)
Simulated excess noise results
Low Noise Avalanche Photodiodes
Simulated excess noise in PN+ junctions(... primary electrons are injected into the LOW field)
The excess noise is againreduced by increased doping
For the same dopingmagnitude, F(PN+) isslightly greater than F(P+N)
Electron Multiplication, Me
0 5 10 15 20
Exc
ess
Noi
se F
acto
r, F
0
5
10
15
NA=2x1016cm-3
NA=2x1017cm-3
NA=1x1017cm-3
NA=5x1016cm-3
Low Noise Avalanche Photodiodes
Experimental results with electron injection
Electron multiplication, Me
0 5 10
Exc
ess
Noi
se F
acto
r, F
1
2
3
4 P+N:ND=6x1016
cm-3
PN+:NA=1x1017
cm-3
PN+:NA=3x1017
cm-3
Increasing the doping in thePN+ devices reduces noise
Experiment corroborates theory
Low Noise Avalanche Photodiodes
Effect of temperature variation on APD
performance
Effect of temperature variation on APD
performance
Low Noise Avalanche Photodiodes
l APD multiplication is very temperature sensitiveNot a problem when input signal is large - BER increases when at the limit of sensitivity
l Breakdown variation is ~ 0.06-0.2V/°C
Temperature (0C)
-10 0 10 20 30 40 50 60 70
VA
PD
(V)
35
40
45
50
55
60
65
M = 12M = 4
M=4
M=12
Active circuit required to vary bias to ensure constant multiplication
Bias required for M = 4 & M = 12 at different temperatures
Temperature dependence of avalanche multiplication
Low Noise Avalanche Photodiodes
Reverse bias (V)
0 5 10 15 20 25 30 35 40 45
Mea
sure
d cu
rren
t (A
)
10-11
10-10
10-9
10-8
10-7
10-6
10-5
10-4
10-3Increasing
temperature
Dark current
Photocurrent
1µm GaAs p-i-n l Photocurrent, dark current and breakdown measured on different thickness GaAsp-i-n diodes, from 20K-500K
l Sharp Vbd observed at all temperatures.
l Dark currents increase with temperature
l Avalanche multiplication reduces with increasing temperature
Temperature dependent I-V for 1µm GaAs
Low Noise Avalanche Photodiodes
Reverse bias (V)0 5 10 15 20
Mea
sure
d cu
rren
t (A
)
10-11
10-10
10-9
10-8
10-7
10-6
10-5
10-4
10-3
Increasingtemperature
Dark currentPhotocurrent
Reverse bias (V)0 1 2 3 4 5 6 7 8 9
Mea
sure
d cu
rren
t (A
)
10-11
10-10
10-9
10-8
10-7
10-6
10-5
10-4
10-3Increasing
temperature
Dark currentPhotocurrent
• Similar behavior seen in thinner avalanche width structures• Thinner devices are less affected by changes in temperature
0.5µm GaAs p-i-n 0.1µm GaAs p-i-n
Temperature dependent I-V for 0.5µm & 0.1µm GaAs
Low Noise Avalanche Photodiodes
Temperature (K)0 100 200 300 400 500
Bre
akdo
wn
Vol
tage
(V)
5
10
15
20
25
30
35
40
45
The breakdown change is more significant in thicker structures
w=1µm
w=0.5µm
w=0.1µm
Temperature (K)0 100 200 300 400 500
Vbd
( T) /
Vbd
(300
K)
(%)
70
80
90
100
110
120
Decreasing AvalancheWidth
w = 1.0 µm, 0.5µm & 0.1 µm Percentage change in Vbd
Temperature coefficientdecreases from 0.032V/oC to 0.004V/oC
Change in Vbd with Temperature
Low Noise Avalanche Photodiodes
1/E (x10-6 cm/V)1 2 3 4 5
Ioni
zatio
n C
oeff
icie
nts
(cm
-1)
103
104
105
Increasing Temperature
αβ
500K 300K
100K
GaAs
GaAs ionization coefficients l Ionization coefficients derived from multiplication data
l Ionization coefficients decrease with increasing temperature
l The change is much larger at lower electric fields
l Thinner avalanche widthsoperate at higher electric-fields
l Phonon scattering relatively less important at higher electric fields
Temperature dependent ionization coefficients
Low Noise Avalanche Photodiodes
Reverse bias (V)0 2 4 6 8 10 12 14 16 18
Mea
sure
d cu
rren
t (A
)
10-12
10-11
10-10
10-9
10-8
10-7
10-6
10-5
10-4
10-3
10-2
Increasing temperature
photocurrent
Idar
k
InP I-V characteristicsl Similar measurements on
w = 0.3µm InP p-i-n from 20K-300K
l Low dark current and l Sharp breakdown
observed
l Vbd decreases as temperature decreases
l Similar results observed in structure with w = 0.5µm.
Temperature dependent I-V in thin InP
Low Noise Avalanche Photodiodes
Temperature (K)0 50 100 150 200 250 300
Vbd
(T)/V
bd(3
00K
) (%
)
40
50
60
70
80
90
100
w = 0.30µmw = 0.50µmZappa et al. Fujitsu SAM APD
é Lower temperature coefficient of breakdown voltage, ηo as w decreases.
é Zappa et al (IPRM’96): ηo ~ 0.225V/oC
é Fujitsu SAM-APD: ηo ~ 0.09 - 0.15V/oC
é w = 0.3µm: ηo ~ 0.012V/oC é w = 0.5µm: ηo ~ 0.02V/oC
InP percentage change in Vbd
InP temperature coefficient of Vbd
Low Noise Avalanche Photodiodes
Effect of thin avalanching widths
on APD speed
Effect of thin avalanching widths
on APD speed
Low Noise Avalanche Photodiodes
Gain-bandwidth Characteristics
Multiplication Gain (M)1 10
Ban
dwid
th (G
Hz)
1
10
Gain -Bandwidth ~120GHz
Carrier trapping
v Bandwidth decreases at low multiplication -carrier trapping
v Bandwidth decreases at high multiplication -multiple transits
Low Noise Avalanche Photodiodes
w1position
time
electroninjected
0
M = 3
w1
time
0
M = 5
v APD is slow c.f. p-i-n diodes due to multiple transits required for high gains
v Difficult to achieve 10 Gb/s operation with thick avalanching structures
APD speed limitations-multiplication build-up time
Low Noise Avalanche Photodiodes
position
time
electroninjected
0 w1
M = 3
w2<w1
electroninjected
time
M = 3
T1
T2 < T1
Decreasing w results in shorter transit times -higher speed
Thin avalanche region multiplication build-up time
Low Noise Avalanche Photodiodes
Frequency (GHz)0.1 1 10 100
Nor
mal
ised
Gai
n (d
B)
-10
-5
0M = 1B = 40GHz
M = 5B = 8GHz
M = 10B = 4GHz
u APD frequency response approximates a 1st order system
u Figure of merit - Gain bandwidth product (GBP)
u Motivation of thin avalanche regions < 1µm to increase GBP
Frequency response of APDs for fixed reverse bias
APD limitations - frequency response
Low Noise Avalanche Photodiodes
o
electron multiplication factor, M10 100
norm
aliz
ed b
andw
idth
(2π
f 3dBτ)
0.1
1
¯ Factors affecting APD speed: Carrier transit time, RC time constant, carrier diffusion andmultiplication buildup time (f3dB)
J Conventional (Emmons’) model [Emmons, 1967]
J Negligible dead space ð d = 0
J Constant carrier speed ð v = vsat(saturated drift velocity)
J Constant gain-bandwidth product,GBP
J GBP scales with τ , carrier transit time
α/ β = 1
increasing α/ β
Constant GBP [Emmons]
Multiplication-limited bandwidth
Low Noise Avalanche Photodiodes
time (ps)0 10 20 30 40 50
Impu
lse
curr
ent (
A)
10-9
10-8
10-7
10-6
without dead spacewith dead space
¬ Comparison of d = 0 cf. non-local model with d ≠ 0
¯ v = vsat
¯ Avalanching region width of w = 0.1µm
¯M = 12.5
¯ d ñ, avalanche current impulse response decays more slowly ð lower f3dB [Hayat and Saleh, 1992]
Effect of dead space on speed
Low Noise Avalanche Photodiodes
l fm of APDs with avalanche width of w, α = β and <M> = 20
l Compare d/w = 0, 0.2 and 0.3
w (µm)0.0 0.5 1.0 1.5 2.0
f m (
GH
z)
0
5
10
Emmons' analysisd/w = 0d/w = 0.2d/w = 0.3
v Obtain fm by Fourier transforming the current impulse response
v fm obtained agrees with Emmons’ prediction for d/w = 0
v As d/w increases, fm falls below the predicted values
v d/w is larger in thin APDsð absolute decrease in fm is larger in thin APDs
Effects of dead space
Low Noise Avalanche Photodiodes
time (ps)0 10 20 30 40 50
Impu
lse
curr
ent (
A)
10-9
10-8
10-7
10-6
Monte Carlo modelconstant v = v sat
v Monte Carlo model cf. constant v = vsat modelv Same dead space, d
¯ w = 0.1µm, M = 12.5¯ Enhanced speed in
MC model leads to faster decay of current impulse response ð higher f3dB
¯ Dead space and enhanced speed effects compete!
[Hambleton et al, 2002]
Carrier speed assumptions
Low Noise Avalanche Photodiodes
Ü Constant GBP
Ü f3dB (Monte Carlo) > f3dB (Emmons) for all w and all MÜ Enhanced carrier speed dominates dead space
M1 10
f 3dB (G
Hz)
10
100
M1 10
f 3dB
(GH
z)10
100
M1 10
f 3dB
(GH
z)
10w = 1.00µm w = 0.20µm w = 0.05µm
Monte CarloEmmons
Simulation result comparisons
Low Noise Avalanche Photodiodes
w (µm)0.1 1
gain
-ban
dwid
th p
rodu
ct (G
Hz)
100
1000v Monte Carlo á more
rapidly as wâ thanEmmons and (vsat + dead space)
v GBP enhancement áas w â
v Worst case is using (vsat+ dead space)
GBP (Monte Carlo) > GBP (Emmons) > GBP (vsat + dead space)
Monte Carlo modelEmmons’ modelvsat + dead space
GBP comparison
Low Noise Avalanche Photodiodes
Ü Emmons’ model predicts GBP200nm = 2 × GBP400nm
Ü But larger d/w in w = 200nm device slows frequency response
Ü Suggests v200nm > v400nm
¤ Lenox et al. (PTL 1999) measured f3dB of InAlAs RCE APDs¤w = 400 nm GBW = 130 GHz¤w = 200 nm GBW = 290 GHz
¤ GBP200nm > 2 × GBP400nm
Published experimental f3dB
Low Noise Avalanche Photodiodes
Conclusions
H The excess noise decreases as the avalanche width decreases below 1µm, in disagreement with the theory of McIntyre
H The low noise results from a more deterministic impactionization process at high fields as dead space becomes more important
H The carrier type initiating the multiplication becomes unimportant at high fields
H Thin avalanching regions should be less temperature sensitive
H Thin avalanching regions should be capable of high speedoperation
H 40 Gb/s APDs highly probable
Low Noise Avalanche Photodiodes
Graham Rees, Peter Robson, Richard Tozer,
Bob Grey, Mark Hopkinson, Geoff Hill, J.S. Roberts
J.S. Ng, B.K. Ng, C.H. Tan, K.S. Lau, C. Groves, D.J. Massey,
B. Jacob, P.J. Hambleton, C.N. Harrison, M. Yee,
D.S. Ong, C.K. Chia, R. Ghin, K.F. Li, S.A. Plimmer, G.M. Dunn
IEEE LEOS, EPSRC, DERA, EU
Graham Rees, Peter Robson, Richard Graham Rees, Peter Robson, Richard TozerTozer, ,
Bob Grey, Mark Bob Grey, Mark HopkinsonHopkinson, Geoff Hill, J.S. Roberts, Geoff Hill, J.S. Roberts
J.S. Ng, B.K. Ng, C.H. Tan, K.S. Lau, C. Groves, D.J. Massey, J.S. Ng, B.K. Ng, C.H. Tan, K.S. Lau, C. Groves, D.J. Massey,
B. Jacob, P.J. B. Jacob, P.J. HambletonHambleton, C.N. Harrison, M. Yee, , C.N. Harrison, M. Yee,
D.S. D.S. OngOng, C.K. , C.K. ChiaChia, R. , R. GhinGhin, K.F. Li, S.A. , K.F. Li, S.A. PlimmerPlimmer, G.M. Dunn, G.M. Dunn
IEEE LEOS, EPSRC, DERA, EUIEEE LEOS, EPSRC, DERA, EU
Acknowledgements
Low Noise Avalanche Photodiodes
l Applications requiring UV detection
2 Atmospheric UV remote sensing
2 UV astronomy2 Combustion control2 Detection of fire,
corona discharge on HV lines
2 Aircraft & missile detection
ll Applications requiring Applications requiring UV detectionUV detection
2 Atmospheric UV remote sensing
2 UV astronomy2 Combustion control2 Detection of fire,
corona discharge on HV lines
2 Aircraft & missile detection
UV DetectionUVUV--enhanced enhanced Si Si photodiodesphotodiodes
Low Noise Avalanche Photodiodes
Why SiC for UV APDs?
l Wide bandgap (3.25eV for 4H-SiC)⇒ excellent for UV detection⇒ very low dark current⇒ high temperature operation
l Large β/α ratio in 4H-SiC⇒ desirable for thick APD structures⇒ performance in thin structures unknown
Why Why SiCSiC for UV for UV APDsAPDs??
ll Wide Wide bandgapbandgap (3.25eV for 4H(3.25eV for 4H--SiCSiC))⇒⇒ excellent for UV detectionexcellent for UV detection⇒⇒ very low dark currentvery low dark current⇒⇒ high temperature operationhigh temperature operation
ll Large Large β/αβ/α ratio in 4Hratio in 4H--SiCSiC⇒⇒ desirable for thick APD structuresdesirable for thick APD structures⇒⇒ performance in thin structures unknownperformance in thin structures unknown
Motivation
Low Noise Avalanche Photodiodes
4H-SiC Device Structures
l 2° +ve bevel edge & multistep junction extension terminationl Square mesas with areas : 50 × 50 ~ 210 × 210 µm2
l Passivated with SiO2 & SiNxl Al/ Ti top contact with optical access
l 2° +ve bevel edge & multistep junction extension terminationl Square mesas with areas : 50 × 50 ~ 210 × 210 µm2
l Passivated with SiO2 & SiNxl Al/ Ti top contact with optical access
Low Noise Avalanche Photodiodes
Wavelength (nm)230 250 270 290 310 330 350 370
Res
pons
ivit
y (m
A/W
)
0
20
40
60
80
100
120
140
160
250 300 350 400
100
101
102
70%
60%
50%
40%
30%
20%
10%
Responsivity at Unity Gain, Beveled APDs
l Similar to typical 6H-SiC photodiodes
l Responsivity cutoff at ~380 nm ⇒ visible-blind
l Peak responsivity of 144 mA/W at 265 nm ⇒quantum efficiency of ~ 67%
l Similar to typical 6H-SiC photodiodes
l Responsivity cutoff at ~380 nm ⇒⇒ visible-blind
l Peak responsivity of 144 mA/W at 265 nm ⇒⇒quantum efficiency of ~ 67%
Low Noise Avalanche Photodiodes
Reverse IV Characteristics
l Avalanche breakdown is sharp & well-defined at Vbd = 58.5V & 124.0Vl Carriers injected with 230 ~ 365 nm light to initiate multiplicationl Iph is 1 ~ 3 orders of magnitude > Idark
l AC measurements corroborate DC results
l Avalanche breakdown is sharp & well-defined at Vbd = 58.5V & 124.0Vl Carriers injected with 230 ~ 365 nm light to initiate multiplicationl Iph is 1 ~ 3 orders of magnitude > Idark
l AC measurements corroborate DC results
Reverse bias voltage (V)0 20 40 60 80 100 120
Cur
rent
(A)
10-11
10-10
10-9
10-8
10-7
10-6
10-5
10-4
Reverse bias voltage (V)0 10 20 30 40 50 60
Cur
rent
(A)
10-12
10-11
10-10
10-9
10-8
10-7
10-6
10-5
10-4
DarkDark
297 nm297 nm365 nm365 nm
230 nm230 nm
DarkDark
297 nm297 nm365 nm365 nm
230 nm230 nm
Beveled APDs, 160×160µm2 Reach-Through APDs, 150×150mm2
Low Noise Avalanche Photodiodes
Multiplication Characteristics
l M of > 200 measuredl M at various λ more disparate for thicker APD structurel Smaller M from shorter λ
⇒ Mh > Me ⇒ β > α
l M of > 200 measuredl M at various λ more disparate for thicker APD structurel Smaller M from shorter λ
⇒⇒ MMhh > > MMee ⇒⇒ ββ > > αα
Reverse bias voltage (V)50 60 70 80 90 100 110 120
Mul
tipl
icat
ion
fact
or, M
2468
101214161820
365nm297nm265nm250nm240nm230nm
Reverse bias voltage (V)25 30 35 40 45 50 55 60
Mul
tipl
icat
ion
fact
or, M
2468
101214161820
365nm297nm265nm250nm240nm230nm
Beveled APDs Reach-Through APDs
Low Noise Avalanche Photodiodes
Multiplication factor, M5 10 15 20 25 30 35 40 45
Exc
ess
nois
e fa
ctor
, F
10
20
30
40
50
60
k = 0.15
k = 2.8
Multiplication factor, M10 20 30 40 50
Exc
ess
nois
e fa
ctor
, F
5
10
15
20
25
30
k = 0.8
k = 0.1
¢¢ 230 nm230 nm�� 365 nm365 nm
¢¢ 230 nm230 nm�� 365 nm365 nm
Excess Avalanche Noise Characteristics
l Excess noise measured for M > 40⇒ good quality of APDs, very stable avalanche multiplication
l Very low excess noise of k = 0.1 & 0.15 measured with 365 nm lightl Excess noise from electron injection (230 nm) gave k = 0.8 & 2.8
l Excess noise measured for M > 40⇒⇒ good quality of APDs, very stable avalanche multiplicationgood quality of APDs, very stable avalanche multiplication
l Very low excess noise of k = 0.1 & 0.15 measured with 365 nm lightl Excess noise from electron injection (230 nm) gave k = 0.8 & 2.8
Beveled APDs Reach-Through APDs
Low Noise Avalanche Photodiodes
Comparison with Si & Al0.8Ga0.2As
Reverse bias voltage (V)20 25 30 35 40 45 50 55 60
Mul
tipl
icat
ion
fact
or, M
4
8
12
16
20
4H-SiC
Reverse bias voltage (V)0 2 4 6 8 10 12 14
Mul
tipl
icat
ion
fact
or, M
4
8
12
16
20
Al0.8Ga0.2AsSil Vbd of 4H-SiC is 10× & 5× of Si &
Al0.8Ga0.2As respectivelyl Me & Mh closer for Si, Al0.8Ga0.2As
l 4H-SiC ⇒ lowest excess noise in a w = 0.1 µm structure
l Vbd of 4H-SiC is 10× & 5× of Si & Al0.8Ga0.2As respectively
l Me & Mh closer for Si, Al0.8Ga0.2As
l 4H-SiC ⇒⇒ lowest excess noise in a w = 0.1 µm structure
Multiplication factor, M10 20 30 40 50 60 70
Exc
ess
nois
e fa
ctor
, F
2
4
6
8
10
12
k=0.1
k=0
k=0.2k=0.3k=0.4
Si and Si and AlAl0.80.8GaGa0.20.2AsAs
365 nm365 nm
230 nm230 nm
MMeeMMhh
Low Noise Avalanche Photodiodes
+ 4H-SiC APD’s exhibit good visible-blind performance
+ Photomultiplication characteristics⇒ Large M in excess of 200 measured⇒ show unambiguously that β > α⇒ β/α ratio remains large in short devices
+ Very low excess noise of k = 0.1 achieved with mainly holes-initiated multiplication
+ 4H-SiC is a suitable material for high-gain, low noise UV avalanche photodiodes
+ 4H-SiC APD’s exhibit good visible-blind performance
+ Photomultiplication characteristics⇒⇒ Large M in excess of 200 measured⇒⇒ show unambiguously that β > α⇒⇒ β/α ratio remains large in short devices
+ Very low excess noise of k = 0.1 achieved with mainly holes-initiated multiplication
+ 4H-SiC is a suitable material for high-gain, low noise UV avalanche photodiodes
Conclusions
Low Noise Avalanche Photodiodes
l AlxGa1-xAs material system is widely used in HBTs and IMPATTs
l Use in telecom wavelength APDs limited by lack of lattice-matched material that absorbs at long wavelength
l GaInAsN has recently been demonstrated 11 absorbs long wavelength11 lattice-matched to AlxGa1-xAs
l GaAs-based APDs is possible and may require AlxGa1-xAs multiplication region for optimum performance
MotivationMotivation
Al0.8Ga0.2As : A Very Low Excess Noise Multiplication Medium for GaAs-based APDs
Low Noise Avalanche Photodiodes
Device structuresDevice structuresl Homojunction p-i-n/n-i-p
grown by conventional MBE with w = 1 µm
l 1 heterojunction p-i-n with w=0.8 µm to obtain Me & Mhfrom same diode
l Optical access window fabricated by wet etching
l Pure carrier injection obtained with 442nm & 633nm light
l 542nm light used to produce mixed carrier injection
l Homojunction p-i-n/n-i-p grown by conventional MBE with w = 1 µm
l 1 heterojunction p-i-n with w=0.8 µm to obtain Me & Mhfrom same diode
l Optical access window fabricated by wet etching
l Pure carrier injection obtained with 442nm & 633nm light
l 542nm light used to produce mixed carrier injection
Al0.8Ga0.2As : A Very Low Excess Noise Multiplication Medium for GaAs-based APDs
Low Noise Avalanche Photodiodes
Avalanche excess noise of homojunction pAvalanche excess noise of homojunction p--ii--n diodesn diodes
• k~ 0.19 for electron multiplication• Larger F for mixed carrier injection
• Lower M for mixed carrier injection ⇒ α > β
Al0.8Ga0.2As : A Very Low Excess Noise Multiplication Medium for GaAs-based APDs
Reverse bias voltage (V)0 10 20 30 40 50
Mu
ltip
licat
ion
fact
or,
M
1
2
3
4
5
6
7
8
9
10
11
12
Multiplication factor, M2 4 6 8 10 12 14 16 18 20
Exc
ess
no
ise
fact
or,
F1
2
3
4
5
6
7
8
9
10
k=0
k=1MMee
MMmixedmixed
Low Noise Avalanche Photodiodes
Avalanche excess noise of thin diodesAvalanche excess noise of thin diodes
• Comparable excess noise for bulk and thin diodes
• Vbd ↓ with decreasing w
Al0.8Ga0.2As : A Very Low Excess Noise Multiplication Medium for GaAs-based APDs
Reverse bias voltage (V)2 3 4 5 6 78 10 20 30 40 60
Mu
ltip
licat
ion
fact
or,
M
1
2
3
4
5
6
7
8
9
10
11
12
1µm0.1µm
Multiplication factor, M2 4 6 8 10 12 14 16 18 20
Exc
ess
no
ise
fact
or,
F
1
2
3
4
5
6
7
8
9
10
k=0
k=1
Low Noise Avalanche Photodiodes
Comparison with InPComparison with InP--based APDsbased APDs
Al0.8Ga0.2As : A Very Low Excess Noise Multiplication Medium for GaAs-based APDs
Multiplication factor, M1 2 3 4 5 6 7 8 9 10 11 12
Exc
ess
nois
e fa
ctor
, F
1
2
3
4
5
6
7
8
Al0.8Ga0.2As
CommericalInP-based APDFujitsu VN206
ki=0
ki=1
l Commercial InP-based APD give excess noise ofki=~0.7 with hole initiated multiplication
l Much lower excess noise can be obtained with Al0.8Ga0.2As as avalanche medium
l Commercial InP-based APD give excess noise ofki=~0.7 with hole initiated multiplication
l Much lower excess noise can be obtained with Al0.8Ga0.2As as avalanche medium
Low Noise Avalanche Photodiodes
Comparison with lower aluminium AlComparison with lower aluminium AlxxGaGa11--xxAsAs
l AlxGa1-xAs (x ≤ 0.6) has large avalanche excess noise
l Excess noise of Al0.8Ga0.2As is much lower
l Al0.8Ga0.2As also has lower excess noise than a commercial InP-based APD
l At M=10, excess noise of Al0.8Ga0.2As is at least 2 times lower
l AlxGa1-xAs (x ≤ 0.6) has large avalanche excess noise
l Excess noise of Al0.8Ga0.2As is much lower
l Al0.8Ga0.2As also has lower excess noise than a commercial InP-based APD
l At M=10, excess noise of Al0.8Ga0.2As is at least 2 times lower
Al0.8Ga0.2As : A Very Low Excess Noise Multiplication Medium for GaAs-based APDs
Aluminium composition, x0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
k i
0.15
0.20
0.30
0.40
0.50
0.60
0.700.80
1.00
CommercialInP-based APDs
Low Noise Avalanche Photodiodes
Ionization coefficientsIonization coefficients
Al0.8Ga0.2As : A Very Low Excess Noise Multiplication Medium for GaAs-based APDs
Inverse electric field (cm/V)1x10-6 2x10-6 3x10-6 4x10-6 5x10-6
Impa
ct io
niza
tion
coe
ffic
ient
s (c
m-1
)
101
102
103
104
105
Al0.8Ga0.2As
AlxGa1-xAs (x=0, 0.15, 0.3, 0.6)l Large α/β ratio as
compared to AlxGa1-xAsof lower x⇒ Lower excess noise
l β/α ratio of InP is small ⇒ Higher excess noise
l Large α/β ratio as compared to AlxGa1-xAsof lower x⇒ Lower excess noiseLower excess noise
l β/α ratio of InP is small ⇒ Higher excess noiseHigher excess noise
Low Noise Avalanche Photodiodes
ConclusionsConclusions
+Bulk Al0.8Ga0.2As diodes give lower excess noise than AlxGa1-xAs (x ≤ 0.6) or InP
+Consequence of the larger α/β ratio in Al0.8Ga0.2As
+Low noise APDs may be achievable on GaAssubstrates using Al0.8Ga0.2As as the gain medium
+Bulk Al0.8Ga0.2As diodes give lower excess noise than AlxGa1-xAs (x ≤ 0.6) or InP
+Consequence of the larger α/β ratio in Al0.8Ga0.2As
+Low noise APDs may be achievable on GaAssubstrates using Al0.8Ga0.2As as the gain medium
Al0.8Ga0.2As : A Very Low Excess Noise Multiplication Medium for GaAs-based APDs
Low Noise Avalanche Photodiodes
APD noise measurement system
Excess noise+ 10 MHz center frequency+ ENBW of 4 MHz+ AC technique with
modulated light source+ F from M & noise power+ Shot noise of Si p-i-n as
reference
Excess noiseExcess noise+ 10 MHz center frequency+ ENBW of 4 MHz+ AC technique with
modulated light source+ F from M & noise power+ Shot noise of Si p-i-n as
reference