Lect01 PDF

Lecture Notes 1
Silicon Photodetectors
• Light Intensity and Photon Flux

• Photogeneration in Silicon
• Photodiode
◦ Basic operation
◦ Photocurrent derivation
◦ Quantum efficiency
◦ Dark current
• Direct Integration
• Photogate
• Appendices
◦ Appendix I: Derivation of Continuity Equation
◦ Appendix II: Depletion Width for PN Junction
◦ Appendix III: MOS Capacitor
◦ Appendix IV: Useful Data
EE 392B: Silicon Photodetectors 1-1

Sfrag replacements
Preliminaries
• Photodetector is the front end of the image sensor. It converts light

incident on it into photocurrent that is (hopefully) proportional to its
intensity
• Conversion is done in two steps:
◦ Incident photons generate e-h pairs in the detector (e.g., silicon)
◦ Some of the generated carriers are converted into photocurrent
• Photocurrents are typically very small (10s to 100s of fA)
◦ Direct measurement is difficult
ADC
◦ Usually integrated into charge on a capacitor and then converted to
Gain
voltage before readout
DN
Photonflux Current density Charge Voltage
Quantum Efficiency Integration Conversion
space/time Gain
ph/cm2·sec A/cm2 Col V

Visible Light
• We are mainly concerned with visible light image sensors

• Recall that the energy of a photon is given by Eph = hc/λ, where
h = 4.135 × 10−15eV.sec is Planck’s constant, c = 3 × 108m/s is the speed
of light, and λ is the wavelength
• Visible light wavelengths (λ) range from 400 nm to 700 nm
Violet: 400 nm (Eph = 3.1 eV)
Blue: 450 nm (Eph = 2.76 eV)
Cyan: 500 nm (Eph = 2.48 eV)
Green: 550 nm (Eph = 2.27 eV)
Yellow: 600 nm (Eph = 2.08 eV)
Red: 700 nm (Eph = 1.77 eV)
Infrared: > 800 nm (Eph < 1.55 eV)

• The amount of light incident on an image sensor surface depends on
◦ The light source
◦ The surface reflectance of the object being imaged
◦ The imaging optics used
• Different visible light sources, e.g., daylight (D65), incandescent, halogen,
fluorescent have different power spectra

Radiometry and Photometry
• Two ways to measure the intensity of light incident on a surface:

◦ Radiometry measures it as irradiance E W/m2
◦ Photometry measures it as illuminance Eν in lux or lumens/m2, which
1
is defined as 683 W/m2 at λ = 555nm
• Illuminance takes into account the sensitivity of the human eye to
different wavelengths; λ = 555nm is the wavelength for which the human
eye is most sensitive and the value for which the photopic vision curve is
normalized
1
0.9
0.8
relative sensitivity
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
350 400 450 500 550 600 650 700 750
wavelength [nm]

• Translating from irradiance to illuminance: Denote the vision photopic
curve as Y (λ) and the irradinace density E(λ) W/m2.nm, then the
illuminance is given by
Z 700
Eν = 683 Y (λ)E(λ) dλ lux
400

Photon Flux
• Photon flux F0 is the number of photons per cm2.sec incident on a surface

• Using the photon energy Eph(λ), we can readily translate irradiance
density E(λ) into photon flux
Z 700 −4
10 E(λ)
F0 = dλ photons/cm2.sec
400 Eph(λ)
• Translating from illuminance to photon flux:
◦ At λ = 555nm, Eph = 35.8 × 10−20Joule; thus 1 lux corresponds to
F0 = 1016/683 × 35.8 = 4.09 × 1011photons/cm2·sec, or, 133 photons
strike a 1µm × 1µm surface per 1/30 sec
◦ A typical light source (e.g., D65) has a wide range of wavelengths
and 1 lux roughly corresponds to F0 ≈ 1012photons/cm2.sec, or, 333
photons strike a 1µm× 1µm surface per 1/30 sec

• Photon flux values encountered vary over a very wide range:
clear sky ≈ 104 Lux, or F0 = 1017
room light ≈ 10 Lux, or F0 = 1013
full moon ≈ 0.1 Lux, or F0 = 1011
moonless night ≈ 10−4Lux, or F0 = 108

Photocharge Generation in Semiconductors
• Incident photon energy must be > band gap energy (Eg ) to generate an
electron-hole pair
◦ Electrons go to the conduction band (EC )
◦ Holes go to the valence band (EV )
• Energy band diagram of silicon:
Ec
PSfrag replacements Eg = 1.124eV
Ev
• Coincidentally (and luckily) photons in the visible range have enough

energy to generate e-h pairs
◦ No photon can generate more than one e-h pair
• Energy gap of other semiconductors: Ge (0.66 eV), GaAs (1.42 eV)
Photocharge Generation Rate in Silicon
• Assume a monochromatic photon flux F0 photons/cm2.sec at wavelength

λ incident at the surface (i.e., x = 0) of silicon
photon flux
0
PSfrag replacements silicon
e-h pair
• The photon absorption in a material is governed by its absorption

coefficient α(λ) cm−1
• Let F (x) be the photon flux at depth x, then the number of photons
absorbed per second between x and x + ∆x is given by
F (x) − F (x + ∆x) ≈ αF (x)∆x,
We can write this equation in the limit as
dF (x)
= −αF (x)
dx
Solving we obtain
F (x) = F0e−αx photons/cm2.sec
Thus the rate of e-h pairs generated at x is

d
G(x) = (F0 − F (x)) = αF0e−αx e-h pair/cm3.sec
dx

Absorption Coefficient of Silicon
7
10
6
10
Absorption Cofficient [cm−1]
5
10
4
10
3
10
2
10
rag replacements
1
10
200 300 400 500 600 700 800 900 1000
Wavelength [nm]
E. Palik, ”Handbook of Optical Constant of Solids,” Academic, New York, 1985

Absorption Length of Visible Light in Silicon

Light Absorption in a Silicon Slab

Comments
• F (x) and G(x) are average values assuming a large ensemble of photons
(approaching continuum values)
◦ The photon absorption process is actually discrete and random
• Note that:
◦ 99% of blue light is absorbed within 0.6 µm
◦ 99% of red light is absorbed within 16.6 µm
• These depths (surprisingly) are quite consistent with the junction and well
depths of a CMOS process
• But, this is not the whole story . . .
◦ Photocharge needs to be collected and converted into electrical signal

Photodetectors in Silicon
• A photodetector is used to convert the absorbed photon flux into

photocurrent
• There are three types of photodetectors used, photodiode, which is a
reverse biased pn junction, photogate, and pinned diode
• In a standard CMOS process there are three types of photodiodes available
◦ nwell/psub
◦ n+/psub
◦ p+/nwell
and two types of photogates
◦ nMOS transistor gate to drain
◦ pMOS transistor gate to drain

• In this lecture notes we discuss the photodiode and photogate operation.
The pinned diode will be discussed in the following lecture notes

Photodiode Operation
• Assume the depletion approximation of a reverse biased pn junction

PSfrag replacements photon flux
quasi-neutral n-type
n-region
vD > 0
depletion
region iph
quasi-neutral
p-region p-type
• The photocurrent, iph, is the sum of three components:

◦ Current due to electrons generated in the depletion (space charge)
region, isc
ph
◦ Current due to holes generated in the quasi-neutral n-region, ipph
◦ Current due to electrons generated in the quasi-neutral p-region, inph

• Most electrons generated in the depletion region are converted into
current by strong electric field
• Carriers generated in the quasi-neutral regions need to diffuse to the
depletion region to be collected
◦ Some charge is lost through recombination
◦ The diffusion length determines the fraction of charge that is not
recombined

Photocurrent Derivation
• Assumptions
◦ Abrupt pn junction
◦ Depletion approximation
◦ Low level injection, i.e., flux induced carrier densities << majority
carrier densities
◦ Short base region approximation, i.e., junction depths << diffusion
lengths. This is is quite reasonable for advanced CMOS processes
• Our results are inaccurate but will help us understand the dependence of
iph on various device parameters
References:
• F. Van de Wiele, “Photodiode Quantum Efficiency,” in P. G. Jespers, F. van de Wiele, M. H. White eds. “Solid State Imaging,” p.
47, Noordhoff (1976).
• J.C. Tandon, D.J. Roulston, S.G. Chamberlain, Solid State Electronics, vol. 15, pp. 669 – 685, (1972).
• R.W. Brown, S.G. Chamberlain, Physica Status Solidi (a), vol. 20, pp. 675 – 685 (1973)

PSfrag replacements
• Consider the depletion approximation for a reverse biased pn junction

photon flux
v 0
quasi-neutral
i n-type
n-region
x1
depletion
region
x2
quasi-neutral
p-region p-type
x3
x
• Assume a monochromatic photon flux F0 photon/cm2·sec incident at the

surface (x = 0), the e-h generation rate at depth x is given by
G(x) = αF0e−αx ph/cm3.sec
• Assuming all generated electrons in the space charge region are collected,
the current density due to generation in the space charge region is
sc
jph = qF0(e−αx1 − e−αx2 ) A/cm2,
where q = 1.6 × 10−19Col is the electron charge
• The current density due to generation in n-type quasi-neutral region,
which is diffusion current (since there is no field in this region), is given by
0
p ∂p (x)
jph = −qDp n

∂x x=x1
0
where pn is the photogenerated minority carrier (hole) density, and Dp is
the diffusion constant of holes (in cm2/sec)
• To find the current density due to generation in the n-type quasi-neutral
0
region, we first need to find pn(x). This can be done by solving the
continuity equation (see derivation in Appendix I) with current density
p
substituted for by the jph expression above
0 2 0
∂pn ∂ p
= Dp 2n + G(x) − R(x),
∂t ∂x
where G(x) is the hole photogeneration rate and R(x) is their
recombination rate
By the short base assumption, the recombination rate is negligible and we
set R(x) = 0

Now, assuming steady state, the continuity equation simplifies to
2 0
dp
0= Dp 2n + G(x),
dx
which has solution of the form
0 F0 −αx
pn(x) = a + bx − e
αDp
To find a and b, we assume that:
0
◦ at x = 0, we have an ohmic contact, which gives pn(0) = 0
0
◦ at x = x1, i.e., at the edge of the depletion region, pn(x1) = 0
Substituting, we obtain
0 F0 x
pn(x) = (1 − (1 − e−αx1 ) − e−αx)
αDp x1

0
pn (x) 0
quasi-neutral
n-region
PSfrag replacements
x1
depletion
region
We can now find the diffusion current density

0
p ∂p (x)
jph = −qDp n

∂x x=x1
qF0
= (1 − (αx1 + 1)e−αx1 )
αx1

• The current density due to generation in the p-type can be similarly
found, and we obtain
n qF0
jph = ((α(x3 − x2) − 1)e−αx2 + e−αx3 )
α(x3 − x2)
Here we assumed that an ohmic contact at x = x3, which is quite
arbitrary (you will derive it with more reasonable assumptions in HW1)
• The total photogenerated current density is thus given by
−αx1 −αx2 −αx3

qF0 (1 − e ) (e −e )
jph = − A/cm2
α x1 (x3 − x2)
• To find x1 and x2, we can use the simplifying assumptions to derive the
depletion region width (see Appendix II), and we obtain
s
2s 1 1
x2 − x 1 = (vD + φn + φp) + ,
q Na Nd
and use the fact that xn/xp = Na/Nd, where
s = 10.45 × 10−13F/cm is the permittivity of Si
Nd and Na are the donor and acceptor densities in cm−3
φn and φp are the potentials in the n and p regions

Example
• Consider the nwell/psub diode in the generic 0.5µm CMOS process

described in Handout 4 with vD = 2V and F0 = 4.09 × 1012
photons/cm2·sec at λ = 555nm (room light), find the photocurrent
density components
• Using the depletion equation, we find that x1 = 2 − 0.176 = 1.824µm and
x2 = 3.76µm
The photocurrent density components are
sc
jph = 120 nA/cm2
p
jph = 192 nA/cm2
n
jph = 28 nA/cm2
Thus the total photocurrent density jph = 340 nA/cm2
So, for a photodiode of area 30µ2, iph = 102fA

Factors Affecting Photocurrent
• iph is linear in F0, i.e., proportional to illumination

• iph is nonlinear in α and λ
• iph increases as x1 decreases and x2 increases, i.e., as the depletion width
(x2 − x1) increases, which can be achieved by a combination of:
◦ shallow pn junction,
◦ low doping, and/or
◦ by increasing reverse bias voltage
• Depletion region width, however, increases slowly with reverse bias voltage
(high reverse bias voltage also increases dark current as we shall soon see)

Quantum Efficiency
• Quantum efficiency QE(λ) is the fraction of photon flux that contributes

to photocurrent as a function of the wavelength λ
• Using our derived photocurrent equation, we obtain
−αx1 −αx2 −αx3

1 (1 − e ) (e −e )
QE(λ) = − electrons/photons
α x1 (x3 − x2)
• This, in addition to being inaccurate due to the approximations we made,
ignores:
◦ reflection at the surface of the chip
◦ reflections and absorptions in layers above the photodetector
◦ variation of jph over the photodetector area (edge effects)

Example
Consider the nwell/psub diode with vD = 2V

0.7
0.6
0.5
Quantum Efficiency
0.4
0.3
0.2
Sfrag replacements 0.1
0
400 500 600 700 800 900 1000
Wavelength [nm]

Dark Current
• There are sources other than photon flux that lead to current in the
photodetector – the sum of these currents is called ”dark current”
• It is called ”dark current” because it is the the photodetector current with
no illumination present (in the dark)
• Dark current is bad. It limits the image sensor performance:
◦ Introduces unavoidable shot noise
◦ Can vary substantially over the image sensor array causing Dark
Signal Non-uniformity (DSNU)
◦ Reduces signal swing

Sources of Dark Current

Dark Current Contributions
• Dark current due to thermal generation (dominated by traps with energy

in the middle of the bandgap) – can be calculated
◦ Generation current is exponentially dependent on temperature
• Dark current due to interface defects, material (crystal) defect, and metal
contamination, such as:
◦ Edge of the STI or LOCOS
◦ Si/SiO2 interface
◦ Edge of junctions (end-of-range implant damage)
These sources are difficult to model and can only be experimentally
measured
◦ Highly fabrication process dependent
◦ Mitigated by careful pixel layout and dopant profiling (more on this
later)

Calculation of Generation Dark Current
• Thermally generated dark current density due to bulk defects consists of

three components:
p
◦ Current due to carrier diffusion from the quasi-neutral regions, jdc
n
and jdc (vD > 0)
sc
◦ Current due to generation in the space charge region, jdc
• We first analyze the first two components in the same way we analyzed
p n
jph and jph , assuming abrupt pn junction and short base approximation let
pn(x) be the thermally generated minority carriers in the n-type
quasi-neutral region
Ignoring recombination and assuming steady state, the continuity
equation reduces to
d2pn(x)
0 = Dp ,
dx2
with the general solution
pn(x) = ax + b
PSfrag replacements
Assuming ohmic photon
contact at x = 0 we get pn(0) = pn0, the minority carrier
concentration at thermal equilibrium, and assuming no free carriers at the
edge of the depletion
v region, we have pn (x1) = 0
i
Thus
x
pn(x) = pn0 1 −
x1
0 pn0
pn
n-type
quasi-neutral
n-region
x1
x xn
depletion
region
xp
x2
quasi-neutral
p-region p-type
x3 np
np0

The (diffusion) current density is given by
p dpn(x)
jdc = −qDp
dx x=x1

pn0 n2i
= qDp = qDp
x1 Nd x 1
Similarly
n n2i
jdc = qDn
Na(x3 − x2)
• Derivation of the current due to generation in the space charge region,
sc
jdc , is more complicated (see below). It yields

sc qni xn xp
jdc ≈ n + p ,
2 τ0 τ0
where τ0n and τ0p are the excess carrier lifetimes in the n and p type
material, respectively

In practice the depletion region is much wider on one side. In this case we
can express the current density as
sc qnixd
jdc ≈ ,
2τo
where xd = xn + xp is the depletion width and τo = τn = τp is the excess
carrier lifetime of the wider side
• Example: again consider the nwell/psub diode with vD = 2V, at room
temperature
sc
jdc = 3.977 nA/cm2
p n
jdc + jdc = 1.9611 nA/cm2
jdc = 5.938 nA/cm2
So, for a photodiode of area 30µm2, idc ≈ 1.78fA

Factors Affecting idc
• idc increases dramatically with temperature T , since it increases with the

Eg
1.5 − 2kT
intrinsic carrier concentration ni, which is proportional to T e
• isc
dc is the dominant component of idc
1
◦ it increases with doping concentration, since τ is proportional to N
◦ it decreases with decrease in depletion width, thus reducing reverse
bias voltage reduces idc (but also reduces iph!)
◦ the calculated isc
dc is only valid for low electric field, at higher electric
fields (which occurs in the shallower and more highly doped junctions
of advanced processes), iscdc increases much faster with reverse bias
voltage

Generation-Recombination in Depletion Region
• Here, we derive the generation-recombination current in the depletion

region of a reverse biased pn-junction
• The analysis is referred to as the Shockley, Read, Hall (SRH) model
• Salient features of the SRH model:
◦ Generation and recombination of carriers occur through localized
states (recombination centers) with energy within the bandgap
◦ Overall population of the recombination center is fairly constant
◦ Recombination centers quickly capture the majority carriers, but
have to wait for the arrival of a minority carrier

• The generation-recombination rate is given by
Ntvthσnσp(pn − n2i )
U = h i h i
Ei −Et Et −Ei
σp p + ni exp kT + σn n + ni exp kT
(pn − n2i )
= h i h i ,
Ei −Et Et −Ei
τn p + ni exp kT + τp n + ni exp kT
where Nt is the generation-recombination center density, σn,p is the
capture cross section, and the minority carrier lifetimes are given as
τn,p = 1/Ntvthσn,p
We assume that τn = τp ≡ τo (see Handout 4)
Recombination: U > 0
Generation: U < 0
• In the depletion region

Ef n − Ei(x)
n = ni exp ni
kT
Ei(x) − Ef p
p = ni exp ni
kT
• Then
2
−n
U ∼
= i
τpni exp EtkT
−Ei
+ τnni exp EikT−Et
−ni
= h i
τo exp EtkT−Ei
+ exp EikT−Et
• Define
Ei − Et
UT = exp
kT
Then, the maximum generation rate is obtained when
∂U
= 0 ⇒ U T = 1 ⇒ E t = Ei ,
∂UT
and is given by
ni
Gmax = ,
2τo
• The generation current is
Zx2
qnixd
jdcsc = q Gmax dx ≈
2τo
x1

Generation Current at Si/SiO2 Surface
• The semiconductor surface has plenty of localized states having energies

within the bandgap
• The kinetics of generation-recombination at the surface is similar to trap
states in the bulk except that the trap density Nst is an areal density (# /
cm2)
• Again, using the SRH model, the surface generation-recombination rate is
given by
Nstvthσnσp(psns − n2i )
Us = h i h i
σp ps + ni exp EikT
−Est
+ σn ns + ni exp EstkT−Ei
2
∼ (p s n s − n i)
= Nstvthσ h i ,
Ei −Est
ps + ns + 2ni cosh kT
where again we assumed that σn = σp ≡ σ

Recombination: Us > 0
Generation: Us < 0
• Surface generation depends on carrier density
◦ When the surface has plenty of carriers, either due to inversion or
accumulation,
2
(p s n s − n i)
Us ∼
= Nstvthσ h i
Ei −Est
ps + ns + 2ni cosh kT
is small
◦ When the surface is depleted, ps and ns are small, and
Nstvth sni
Us ∼
=− σni = − ,
2 2
Here we assume that Est = Ei and s = Nstvthσ has the unit of
velocity (cm·sec−1)
Surface generation current density is thus
s qsni
jdc =
2

Activation Energy of Dark Current Components
• Notice that the generation currents due to bulk traps and surface traps
are proportional to ni
• And the diffusion current in the quasi-neutral region is proportional to n2i
• Therefore, in a plot of dark current vs log(1/T )
◦ the activation energy of bulk trap and surface trap dark current is Eg
◦ the dark current that is due to diffusion in the quasi-neutral region
has activation energy of Eg /2

Surface Recombination Velocity
• It is difficult to ”derive” the surface recombination velocity

• It is obtained experimentally

• Experimental procedure:
◦ Reverse bias the gated diode
◦ Sweep gate bias from inversion to accumulation
◦ Measure DC current from substrate in inversion (I3), depletion (I2),
and accumulation (I1)

• The currents are given by:
s
qniW AB qnisASB qn2i Dp
I1 = + + AB
2τo(diode) 2 NB τp
qniWGAG qnisASG
I2 = I1 + +
2τo(gate) 2
qnisASG
I3 = I2 −
2
qnisASG
I2 − I3 =
2
qniWGAG
I3 − I1 =
2τo(gate)

Direct Integration
• As discussed earlier, photocurrent is typically too small to measure directly

• The most commonly used mode of photodiode operation in an image
sensor is direct integration, where the photocurrent (and dark current) are
directly integrated over the diode capacitance
vD Q
PSfrag replacements PSfrag replacements
reset high light
reset vD
CD Qmax
iphvo+ idc
vo low light
Q
iph + idc
Qmax CD
high light
low light
tint
t tint t
◦ The photodetector is reset to the reverse bias voltage vD

◦ The diode current discharges CD for tint seconds, which is called
integration time or exposure time
◦ At the end of the integration time the accumulated charge Q(tint) (in
electrons) or voltage vo(tint) is read out
• Assuming that the photo and dark currents do not change with reverse
bias voltage, we obtain
1
Q(tint) = (iph + idc)tint electrons
q
Assuming that CD does not vary with reverse bias voltage, we get
(iph + idc)tint
vo(tint) = vD − V
CD
• The maximum nonsaturating photocurrent is thus given by
qQmax
imax
ph = − idc
tint
Qmax is called the well capacity
• To avoid blooming, i.e., overflowing of charge to neighboring
photodetectors in the image sensor, we ensure that the diode is reverse
biased , i.e., vo(tint ) > 0V
Thus qQmax ≤ vD × CD (very often the voltage swing is lower than vD
resulting in well capacity lower than vD × CD )

Example
• Consider the nwell/psub diode with vD = 2.2V, and area AD = 30µm2

the photodiode capacitance
s
CD = AD = 1.55f F
xn + xp
Note: this is unrealistically small since it does not include edge
capacitance and the capacitances of interconnect and other devices
connected to the photodetetcor
• Thus (ignoring these other capacitances) the well capacity
Qmax = 3.41 × 10−15/1.6 × 10−19 = 21312 electrons
• Assuming tint = 20ms and dark current idc = 2fA we get that the
maximum nonsaturating photocurrent
imax
ph = 167.55fA,
which corresponds to 16.5 lux at λ = 555nm
• To put this in perspective, a typical DRAM cell holds ∼ 160,000 to
200,000 electrons
Finding Q(tint) and vo(tint) Numerically
• Since the depletion region width changes with the reverse bias voltage,
CD , jph, and jdc are not constant during integration
• The output charge and voltage can be found numerically
◦ Set λ and F0 to desired values
◦ Set vo(0) = vD and Q(0) = 0 and calculate vo(k∆t) and Q(k∆t)
iteratively beginning with k = 1 and ending with k = t∆t
int
◦ To calculate vo((k + 1)∆t) and Q((k + 1)∆t):

1. Calculate the depletion region width and CDk (using vo(k∆t))
2. Calculate the current densities jph(k∆t) and jdc(k∆t) and the
charge accumulated ∆Qk = (jph(k∆t) + jdc(k∆t))∆t
∆Qk
3. Set vo((k + 1)∆t) = vo(k∆t) − k
CD
and
Q((k + 1)∆t) = Q(k∆t) + ∆Qk
• The following graphs provide computed Q(tint) and vo(tint) as a function
of F0 for vD = 2.2V

−9 nwell/psub Diode, Wavelength=600nm
x 10
7
Direct Integration
Photo Charge (C/cm2)
0
0 0.5 1 1.5 2 2.5 3 3.5
2 12
Photon Flux (photons/cm s) x 10

−9 nwell/psub Diode, Wavelength=600nm
x 10
6
Direct Integration
4
Total Charge (C/cm2)
0
0 0.5 1 1.5 2 2.5 3 3.5
2 12

nwell/psub Diode, Wavelength=600nm
2.4
2.2
Final voltage due to photocharge (V)
Direct Integration
1.8
1.6
1.4
1.2
1
0 0.5 1 1.5 2 2.5 3 3.5
2 12

nwell/psub Diode, Wavelength=600nm
2
1.9
1.8
Direct Integration
1.7
Final total voltage on diode (V)
1.6
1.5
1.4
1.3
1.2
1.1
1
0 0.5 1 1.5 2 2.5 3 3.5
2 12

Photogate
• Photogate is used in CMOS sensors (CMOS APS), frame transfer CCD

(FT-CCD) and time-delay-and-integration CDD (TDI-CCD)

Photogate Operation
• Gate voltage vG is set high enough to bias the MOS capacitor into the
deep depletion regime (this requires vG >> vT ) (see Appendix)
• Electrons generated in the depletion region are collected in the potential
well
• Electrons generated in the quasi-neutral region will
◦ Recombine with holes
◦ Diffuse to depletion region and get collected in the potential well if it
is within the diffusion length of the minority carriers
• Holes will be collected in the substrate
• How many of the photo-generated carriers are collected depends on:
◦ Diffusion length of minority carriers
◦ Location and length of the depletion region

Quantum Efficiency of Photogate
• Photocurrent has two components

◦ Current due to generation in the depletion region, isc
ph , again almost
all carriers contribute to the current
◦ Diffusion current due to generation in the quasi-neutral p-region, inph
• To calculate the current we make the depletion approximation and use the
basic MOS capacitor equations to find the depletion region width (see
Appendix III) (you will derive it in HW2)
• A disadvantage of the photogate is lower quantum efficiency, especially
for shorter wavelengths (blue), due to absorption in the polysilicon gate
(which has the same α as crystalline silicon)
• Photogate is also used in direct integration mode; charge accumulated on
gate is transferred to another capacitor (as we shall see later)

Quantum Efficiency of Photogate
• It can be shown that QE for photogate, not including absorption in the

polysilicon gate, is given by
2 −αL −αxd

αL e −e cosh ((L − xd)/Ln)
QE(λ) = 1−e−αxd + 2 2 n αe−αxd +
α Ln − 1 Ln sinh ((L − xd)/Ln)
• Limiting Cases:
◦ Very long diffusion length (Ln → ∞):
−α(L−xd)

1−e
QE = 1 − e−αxd + e−αxd 1 −
α(L − xd)
◦ Very thick substrate (L → ∞):
−αxd
αL n e
QE = 1 − e−αxd +
(αLn + 1)

Appendix I – Derivation of 1-D Continuity Equation
• Consider minority carrier (electron) current flow in p-type silicon

• In a slab x to x + ∆x
Gn (x)
PSfrag replacements
Ec
jn (x) jn (x + ∆x)
Ev
x Rn (x) x + ∆x
jn(x): electron current density at x

Gn(x): generation rate (electrons/cm3·s)
Rn(x): recombination rate (electrons/cm3·s)
n(x): electron density at x (electrons/cm3)

• The rate of electron density increase in the slab
∂n(x) 1
∆x ≈ − (jn(x) − jn(x + ∆x)) + (Gn(x) − Rn(x))∆x,
∂t q
which in the limit, gives
∂n(x) 1 ∂jn(x)
= + (Gn(x) − Rn(x))
∂t q ∂x
assuming no electric field, the current is only due to diffusion and is given
by
∂n(x)
jn(x) = qDn ,
∂x
where Dn is the diffusion constant for electrons in cm2/s
substituting, we get the continuity equation
∂n(x) ∂ 2n(x)
= Dn + (Gn(x) − Rn(x))
dt ∂x2
• Similarly for holes,
∂p(x)
jp(x) = −qDp
∂x
and the continuity equation is
∂p(x) ∂ 2p(x)
= Dp 2
+ (Gp(x) − Rp(x))
∂t ∂x
• Assuming low level injection, i.e., that excess carrier concentration <<
majority carrier concentration, we get that
np − np0
Rn =
τn
where np0 is the intrinsic minority carrier concentration, and τn is the

carrier lifetime

Appendix II – Depletion Width for PN Junction
PSfrag replacements
• Energy band diagrams at thermal equilibrium
Ec Ec
E fn
qφn
Ei Ei
qφp
E fp
Ev Ev
p-type n-type
Here φn = kT
q ln Nnid and φp = kT
q ln Nnia , where
k = 8.62 × 10−5eV K−1 is the Boltzman constant
T is the temperature in Kelvin
q = 1.6 × 10−19Col is the electron charge
ni is the intrinsic carrier concentration ≈ 1.45 × 1010cm−3 at room
temperature
Nd and Na are the donor and acceptor densities in cm−3

PN Junction Energy Band Diagram
• The energy band diagram for reverse biased pn junction

p-type n-type
PSfrag replacements
qφp
E fp Ec
qvD
E fn
qφn
Ei
Ev
ρ col/cm3
−xp qNd
xn x
−qNa
E V/cm
x
Emax
φV

E and φ are found by solving the Poisson equation
d2 φ dE ρ(x)
2
= − = − ,
dx dx s
where s = 10.45 × 10−13F/cm is the permittivity of Si
So in the p-type region, we obtain
qNa
E(x) = − (x + xp)
s
and
qNa
Emax = − xp
s
Similarly, in the n-type region we have
qNd
E(x) = (x − xn)
s
and
qNd
Emax = − xn
s
Thus
x n Na
=
x p Nd

Now Z xn
φ(xn) = − Edx
−xp
qNdx2n qNax2p
= +
2s 2s
= vD + φ n + φ p
Combining the last two equations, we obtain that the depletion width
r
2s 1 1
xd = xn + xp = (vD + φn + φp)( + )
q Na Nd
• Example (nwell/psub diode): assuming vD = 2V, φn = 0.3486V, and
φp = 0.289V, we get xn = 0.176µm and xp = 1.76µm
• The (small signal) diode capacitance per unit area is defined as
dQ
C= ,
dvD
where the charge Q = qNdxn = qNaxp. Thus,
s
C= F/cm2
xn + xp
For the previous example C = 5.4 × 10−9F/cm2

Appendix III – MOS Capacitor
• First consider the energy band diagrams under thermal equilibrium for
PSfrag replacements
polysilicon, oxide, and silicon
E0
0.95eV Ec
4.05eV
Ef ≈ Ec Ec
Ei
qφp
Ef
polysilicon Ev
polsilicon p-type
oxide
E0 is the free electron energy

E0 − Ec = 4.05eV is the semiconductor electron affinity
E0 − Ec oxide = 0.95eV is the oxide electron affinity

• The PSfrag
energyreplacements
band diagram for the MOS system under thermal equilibrium
assuming vG = 0
qv0
3.1eV
3.1eV Ec
qφp Ei
Ef
Ev
qψs0
tox
poly oxide p-type
ρ
0 xd x
−Q
−qNa

We can find v0, ψ0, and xd by writing the flat-band voltage in two ways
and solving the Poisson equation
Eg
vF B = + φp = v0 + ψs0,
2q
qNaxd
v0 = , and
Cox
qNax2d
ψs0 =
2s
Cox = tox F/cm2, and ox = 34.5 × 10−14F/cm
ox
PSfrag replacements
• Energy band diagram in the deep depletion regime
qv0
Ec
Ei
Ef
Ev
qvG
qψs
Ef tox
poly oxide p-type

The MOS system is in deep depletion when ψs > 2φp (this is the same
condition as for strong inversion except that here we are interested in the
transient response before the onset of strong inversion)
This gives that

1 p
vG > 2φp − vF B + 4qNasφp = vT ,
Cox
where vT is the threshold voltage (assuming no threshold adjust implant
is used)
• To find the depletion region depth xd, note that
v0 + ψ s = v G + v F B ,
where
qNax2d
ψs = , and
2s
qNaxd
v0 =
Cox
solving for ψs we get
q
ψs = v1 + v2 − v22 + 2v1v2,

where
v1 = vG + vF B , and
qNas
v2 = 2
Cox
the depletion width can then be determined
Note: for the MOS capacitor to stay in the deep depletion we set vD = ψs

Appendix IV
Bulk Mobility

Minority Carrier Diffusion Length

Minority Carrier Lifetime

Lect01 PDF

Uploaded by

Copyright:

Available Formats

You might also like

Lect01 PDF

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Lect01 PDF

Uploaded by

Copyright:

Available Formats

Lecture Notes 1

• Light Intensity and Photon Flux

EE 392B: Silicon Photodetectors 1-1

• Photodetector is the front end of the image sensor. It converts light

ph/cm2·sec A/cm2 Col V

• We are mainly concerned with visible light image sensors

EE 392B: Silicon Photodetectors 1-3

EE 392B: Silicon Photodetectors 1-4

• Two ways to measure the intensity of light incident on a surface:

EE 392B: Silicon Photodetectors 1-5

EE 392B: Silicon Photodetectors 1-6

• Photon flux F0 is the number of photons per cm2.sec incident on a surface

EE 392B: Silicon Photodetectors 1-7

EE 392B: Silicon Photodetectors 1-8

PSfrag replacements Eg = 1.124eV

• Coincidentally (and luckily) photons in the visible range have enough

• Assume a monochromatic photon flux F0 photons/cm2.sec at wavelength

PSfrag replacements silicon

• The photon absorption in a material is governed by its absorption

Thus the rate of e-h pairs generated at x is

EE 392B: Silicon Photodetectors 1-11

EE 392B: Silicon Photodetectors 1-12

EE 392B: Silicon Photodetectors 1-13

EE 392B: Silicon Photodetectors 1-14

EE 392B: Silicon Photodetectors 1-15

• A photodetector is used to convert the absorbed photon flux into

EE 392B: Silicon Photodetectors 1-16

EE 392B: Silicon Photodetectors 1-17

• Assume the depletion approximation of a reverse biased pn junction

• The photocurrent, iph, is the sum of three components:

EE 392B: Silicon Photodetectors 1-18

EE 392B: Silicon Photodetectors 1-19

EE 392B: Silicon Photodetectors 1-20

• Consider the depletion approximation for a reverse biased pn junction

• Assume a monochromatic photon flux F0 photon/cm2·sec incident at the

EE 392B: Silicon Photodetectors 1-22

EE 392B: Silicon Photodetectors 1-23

We can now find the diffusion current density

EE 392B: Silicon Photodetectors 1-24

EE 392B: Silicon Photodetectors 1-25

• Consider the nwell/psub diode in the generic 0.5µm CMOS process

EE 392B: Silicon Photodetectors 1-26

• iph is linear in F0, i.e., proportional to illumination

EE 392B: Silicon Photodetectors 1-27

• Quantum efficiency QE(λ) is the fraction of photon flux that contributes

EE 392B: Silicon Photodetectors 1-28

Consider the nwell/psub diode with vD = 2V

Sfrag replacements 0.1

EE 392B: Silicon Photodetectors 1-29

EE 392B: Silicon Photodetectors 1-30

EE 392B: Silicon Photodetectors 1-31

• Dark current due to thermal generation (dominated by traps with energy

EE 392B: Silicon Photodetectors 1-32

• Thermally generated dark current density due to bulk defects consists of

EE 392B: Silicon Photodetectors 1-34

EE 392B: Silicon Photodetectors 1-35

EE 392B: Silicon Photodetectors 1-36