Vaals, The Netherlands; June 8-11, 2015 IEEE International Image Sensor Society (IISS)
IISSIISS
IISS
IISS
P.1
Outlines
Benefits of 3D stacking
Observation of hot spots in stacked CIS
Photon emission test structures
Spatial distribution derivation and verification
Physical mechanism
Dependence on device types, sizes, and voltages
Correlation to impact-ionization substrate current
Practical metal shield design guidelines
Summary
P.2
Benefits of 3D Stacking
From FSI to BSI
Maximize the pixel level fill factor
3D stacked BSI
Maximize the chip level fill factor
Smaller footprint, lower Z-height, thinner camera modules
Decouple the process requirements for CIS & ASIC
Optical and electrical performance can be optimized independently
Well adopted by all leading-edge smart phones today
Potential future trend
Column- and row-based 3D connection pixel-based 3D connection
Oxide bond with TSV/TOV oxide and metal bond simultaneously
area Pixel
areaPD Unshieldedfactor fill level Pixel
area Chip
area array Pixelfactor fill level Chip
P.3
D1
D2 D4
D5
D6
D3 D7
D8
D9
A1 A2 A3
B1 C1
Pipeline ADC CDS PGA1 PGA2
Ref DAC
D1
D2 D4
D5
D6
D3 D7
D8
D9
A1 A2 A3
B1 C1
Pipeline ADC CDS PGA1 PGA2
Ref DAC Analog signal chain 2
Analog signal chain 1
Hot Spots Found in Test Chip Dark Image
1-to-1 correspondence with devices in various functional blocks
Stacked CIS CIS: N45BSI, 1P4M ASIC: N65LP, 1P4M
3MP test chip 1.1u pitch 2x2 shared 1.5T per pixel Dual signal chains
Dark image Analog gain=8 Exposure time=4s
circ. 2013 Q3 ASIC
Pixel Array
CIS
P.4
Enhanced Hot Spot Intensities
D1 D2
D3
D4
D5
D6
D7
D8
D9
D10
A1 A2 A3
D11 B1
C1
E1
D1 D2
D3
D4
D5
D6
D7
D8
D9
D10
A1 A2 A3
B1 C1 E1
D11
Dark image cropped to 1600x1200
Y-axis (DN) arbitrarily
scaled
D1D11 are scaled NMOS with almost identical bias conditions. Hot spot intensity is proportional to device size.
P.5
Test Element Group Design
Objectives: systematic study of the photon emission by device types, sizes, bias voltages, operation conditions, without or with various metal shields
TEG Description DUT
01 1.2V NMOS, unshielded 55
02 1.2V NMOS, 50u/0.12u, shielded 44
03 1.2V NMOS, 50u/0.06u, shielded 44
04 1.2V PMOS, unshielded 55
05 1.2V PMOS, 50u/0.12u, shielded 44
06 1.2V PMOS, 50u/0.06u, shielded 44
07 2.5/3.3V NMOS, unshielded 55
08 2.5/3.3V NMOS, 50u/0.6u, shielded 44
09 2.5/3.3V NMOS, 50u/0.28u, shielded 44
10 2.5/3.3V PMOS, unshielded 55
11 2.5/3.3V PMOS, 50u/0.6u, shielded 44
12 2.5/3.3V PMOS, 50u/0.28u, shielded 44
13 NPN, PNP BJT 6
14 P+/NW, N+/PW diode 6
15 Miscellaneous 6
L=3.0u
L=2.0u
L=1.5u
L=1.0u
L=0.9u
L=0.8u
L=0.7u
L=0.6u
L=0.5u
L=0.4u
L=0.28u
2.5/3.3V NMOS W=1, 10, 20, 40, 50u
A group of 5x11 NMOS that can be turned on 1-by-1 or altogether
P.6
Hot Spot Intensities Strongly Depend on Vds
Vds = 2.1V Vds = 2.2V Vds = 2.3V Vds = 2.4V
Exposure time = 4s; 1DN 0.0816e/s; Vgs=1.2V X,Y unit: pixel; Z unit: DN (12b ADC)
DN
1e/s
Dark current < 1e/s at 25°C
P.7
Focus on the Spatial Distribution
Vds=2.1V
2.2V
2.3V
2.4V
2.5V
2.6V
2.7V
2.8V
2.9V
3.0V
X,Y unit: pixel; Z unit: DN; note the different Z scales on left- & right-hand side.
NMOS L=0.28u W=1, 10, 20, 40, 50u Vgs =1.2V
P.8
Spatial Distribution of the Hot Spots
Study the hot spots spatial distribution
5 DUTs: W = 1u, 10u, 20u, 40u, 50u; L=0.28u
Fix Vgs with varying Vds or fix Vds with varying Vgs
Vds=2.7V
X,Y unit: pixel; Z unit: DN X unit: pixel; Y unit: DN (12b ADC)
0
1000
2000
3000
4000
0 50 100 150 200 250
Ph
oto
sig
nal
(DN
)
X position (pixel)
Vds=2.1
Vds=2.2
Vds=2.3
Vds=2.4
Vds=2.5
Vds=2.6
Vds=2.7
Vds=2.8
Vds=2.9
Vds=3.0
50/0.28
40/0.28
20/0.28
10/0.28
1/0.28
P.9
Point Spread Function (PSF)
Assuming an isotropic point source cosine-third-power law
Combination of inverse-square law & direction cosine law
3 cos' AA
r
hA’
A
Detector plane
Source plane
222
23222
3
2
2
22
2
2
1
1
44
1
4
yxrJdydxyxE
hrh
J
h
JyxE
h
hrAAA
h
JAE
;,
cos,
cos';'
P.10
Fitting of PSF to a Single Source
0
400
800
1200
1600
2000
40 60 80 100 120 140 160
Ph
oto
sig
nal
(DN
)
X position (pixel)
R100
Fit
1E-1
1E+0
1E+1
1E+2
1E+3
40 60 80 100 120 140 160
Ph
oto
sig
nal
(DN
)
X position (pixel)
R100
Fit
3.3V NMOS W/L=50/0.28 Vds =2.5V Vgs =0.8V
P.11
Fitting Multiple Emission Sources
Pn, xn, h are fitting parameters (n=15)
Pn is the strength of nth peak; xn is the location of nth peak
The empirical formula fits the data reasonably well
0
1000
2000
3000
4000
0 50 100 150 200 250
Ph
oto
sig
nal
(DN
)
X position (pixel)
Fitsum
Vds=3V
upixelhhxx
PxE
n n
n 056551
5
12322
..;
P.12
Comparing to BEOL Structure
The best-fit parameter h matches reasonably well with estimated distance
Extracted from emission data: h 6.05u
Estimated from process: h 6.25u
Test chip uses 1P4M ASIC & 1P4M BSI
In real silicon the BEOL dielectric stack is a complicated multi-layer structure with various refractive indices
Optical simulation is needed to account for the reflection, refraction, and diffraction at various interfaces
ASIC die
BSI die
M1A
M2A
M3A
M4A
M1B
M2B
M3B
M4B
6.25u
PD
DUT
P.13
Another Validation of the Empirical PSF
The ratio of the integrated photo signal to the peak signal is relatively constant, independent of device size and bias conditions
Measured data showed the same slope as the calculation predicted
0.E+00
1.E+05
2.E+05
3.E+05
4.E+05
5.E+05
6.E+05
0 1000 2000 3000 4000
Inte
gra
ted
sig
nal (D
N)
Peak signal (DN)
3.3V NMOSVgs=1.2VVds=2.1~3.3VW=10u
L=3.0u
L=2.0u
L=1.5u
L=1.0u
L=0.9u
L=0.8u
L=0.7u
L=0.6u
L=0.5u
L=0.4u
L=0.28u
Calc
5145
1
120
20
20
2023222
.
x y hyx
S
X,Y unit: pixel
P.14
Physical Mechanism
Hot carriers caused by high lateral electric field
(1) Brake radiation (bremsstrahlung) of energetic hot carriers
(2) Radiative recombination of e-h pairs generated by impact ionization
Space Charge Region
Inversion Layer
Region
Electron Energy
or Temp
Substrate current due to un-recombined holes
VSS VSS
(1) Hot carrier population (2) Hot carrier population Isub Hot carrier population (1) & (2) Isub
Part of the IR emission is not detected by Si PD (Eg1.1eV, cutoff 1.17um)
G
S D
P.15
Photon Emission vs. Vds
Photon emission is clearly not proportional to Ids
The strong dependence on Vds is similar to the impact ionization Isub
Electric field across the space-charge region (Vds – Vdsat)
PMOS shows much weaker photon emission than NMOS
Hole has larger effective mass, smaller mobility, smaller mean-free-path
1E+2
1E+3
1E+4
1E+5
2.1 2.3 2.5 2.7 2.9 3.1 3.3
Ep
hP
ho
to S
ign
al
(e/s
)
Vds (V)
3.3V NMOS, Vgs=1.2V, W=50uL=3.0u
L=2.0u
L=1.5u
L=1.0u
L=0.9u
L=0.8u
L=0.7u
L=0.6u
L=0.5u
L=0.4u
L=0.28u
1E+2
1E+3
1E+4
1E+5
2.8 3.0 3.2 3.4 3.6 3.8
Ep
hP
ho
to S
ign
al
(e/s
)
Vds (V)
3.3V PMOS, Vgs=1.3V, W=50uL=3.0u
L=2.0u
L=1.5u
L=1.0u
L=0.9u
L=0.8u
L=0.7u
L=0.6u
L=0.5u
L=0.4u
L=0.28u
P.16
5.E+2
5.E+3
5.E+4
5.E+5
0.5 1.0 1.5 2.0 2.5 3.0
Ep
hP
ho
to s
ign
al (e
/s)
Vgs (V)
3.3V NMOSVds=3.0VW=50u, t=1s
L=3.0u
L=2.0u
L=1.5u
L=1.0u
L=0.9u
L=0.8u
L=0.7u
L=0.6u
L=0.5u
L=0.4u
L=0.28u
0.0E+0
2.0E+5
4.0E+5
6.0E+5
8.0E+5
1.0E+6
1.2E+6
0.5 1.0 1.5 2.0 2.5 3.0
Ep
hP
ho
to s
ign
al (e
/s)
Vgs (V)
Vds=3.0VL=3.0u
L=2.0u
L=1.5u
L=1.0u
L=0.9u
L=0.8u
L=0.7u
L=0.6u
L=0.5u
L=0.4u
L=0.28u
Photon Emission vs. Vgs
Low-Vgs region: PE dominated by Ids
High-Vgs region: PE dominated by impact ionization
Vdsat Vgs – Vth ; Vds – Vdsat voltage across the pinch-off region
Vgs , Vdsat , (Vds-Vdsat) Photon emission
Vgs , Ids Photon emission
Emission dominated by impact ionization
Photon emission dominated by Ids
nkT
VIE gs
dsph exp
thgsds
phVVV
PE 1exp
P.17
Empirical Formulae for Photon Emission
Use the same equation as in BSIM4 Isub model
For PMOS, the data can be well described by Eq. (1)
For NMOS, we added a 2nd order term to the exponent in Eq. (2)
2
210
10
2
1
dsatdsdsatdsdsatdsdsph
dsatdsdsatdsdsph
VVPVVPVVIPE
VVPVVIPE
exp)(
exp)( P0, P1, P2 are parameters to be extracted from data
Under investigation
P.18
1E+2
1E+3
1E+4
1E+5
1E+6
1E+7
1E+8
0.2 0.3 0.4 0.5 0.6 0.7
Ep
h/I
ds/(
Vd
s –
Vd
sat)
(es
-1A
-1V
-1)
1/(Vds – Vdsat) (V-1)
3.3V NMOS & PMOS
N 50/2.0, Vgs=1.2
N 50/1.0, Vgs=1.2
N 50/0.8, Vgs=1.2
N 50/0.6, Vgs=1.2
N 50/0.4, Vgs=1.2
N 50/0.28, Vgs=1.2
P 50/2.0, Vgs=1.3
P 50/1.0, Vgs=1.3
P 50/0.8, Vgs=1.3
P 50/0.6, Vgs=1.3
P 50/0.4, Vgs=1.3
P 50/0.28, Vgs=1.3
1E-9
1E-8
1E-7
1E-6
1E-5
1E-4
1E-3
1E-2
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
I su
b/I
ds/(
Vd
s –
Vd
sat) (
V-1
)
1/(Vds – Vdsat) (V-1)
3.3V NMOS & PMOS
N 10/10, Vgs=0.8
N 10/10, Vgs=1.2
N 10/0.5, Vgs=0.8
N 10/0.5, Vgs=1.2
N 10/0.28, Vgs=0.8
N 10/0.28, Vgs=1.2
P 10/10, Vgs=0.8
P 10/10, Vgs=1.2
P 10/0.5, Vgs=0.8
P 10/0.5, Vgs=1.2
P 10/0.28, Vgs=0.8
P 10/0.28, Vgs=1.2
Photon Emission Correlated to Isub Except the detailed features, the photon emission and substrate current
show similar voltage (Vds, Vgs) dependence and NMOS vs. PMOS ratios
Only part of the photon emission is detectable by Si photodiodes
The percentage may depend on voltage, not a constant
Photon emission Substrate current
NMOS NMOS
PMOS PMOS
P.19
Comparing Photon Emission to Isub
Detected photo-carriers vs. Isub ratio is approximately 1 : 1010
Only a small portion of hot carriers generate photons
Hot carriers may lose energy via impact ionization or heat dissipation
Only a small portion of the photons are detected by pixel array
At least a hemisphere of photons are lost
Absorbed by poly-gate
Blocked by silicide & metal wires
Photons with E < 1.1eV are not detected
PD QE < 70%
N, Isub
P, Isub
N, Eph
P, Eph
1010
P.20
Hot Carrier Temperature
Photon energy distribution tail follows Maxwell-Boltzmann distribution
Inferred hot carrier temperature can be 1000°C or higher
Hot carrier energy can exceed eVds! (but average energy 3kT/2 << eVds)
T. Matsuda et al., “A test structure for spectrum analysis of hot-carrier-induced photoemission from scaled MOSFETs under DC & AC operation,” pp. 7174, ICMTS (2009)
A. Glowacki et al., “Electron temperature – the parameter to be extracted from backside spectral photon emission,” IRPS, pp. 5B.6.15B.6.7 (2013)
Slope =1/(kT ln10) kT0.15eV T1500°C
T1200°C
W/L=10u/0.12u Vds=1.2V, Vgs=0.8V
CCD
Vgs=1.0V
P.21
Metal Shield Design Guidelines
Step 1: Identify the emission sources (aggressors)
