A CMOS Photodiode Model

Wei-Jean Liu, Oscal T.-C. Chen

Signal and Media Laboratories,
Dept. of Electrical Engineering,
National Chung Cheng University,
Chia-Yi, 621Taiwan

Li-Kuo Dai, Ping-Kuo Weng, Kaung-Hsin Huang, Far-Wen Jih

Solid-State Devices Materials Section,
Materials & Electro-Optics Research Division,
Chung-Shan Institute of Science & Technology,
Tao-Yuan, Taiwan

Abstract
2. The Proposed CMOS Photodiode Model
This paper utilizes the concepts of equilibrium equations and
minority carrier continuity to build and derive at a CMOS Photo current happens because incident light incites
photodiode model. By using the TSMC 1P3M 0.5 µ m CMOS electron-hole pairs. When a potential applied on two sides of a
technology, photodiodes with two different types of photodiode separates the electron-hole pairs, the photo current is
Nwell-PEpitaxial-Psubstrate and Ndiffusion-Pwell-PEpitaxial-Psubstrate were generated. Two major sources of photo currents are: (1) from the
fabricated. Measurements were done to test the photoresponse diffusion current due to an imbalance of carrier concentration
from 400nm to 1000nm. To simulate the photoresponses of the outside of the depletion region; (2) from the drift current due to the
Nwell-PEpitaxial-Psubstrate and Ndiffusion-Pwell-PEpitaxial-Psubstrate separation of electron-hole pairs within the electric field of the
photodiodes using the proposed model, if an inadequate value of depletion region. In analyzing the physic characteristics of a
the surface recombination velocity which is affected by the surface semiconductor, there are two assumptions in general: (1) electric
defects in the manufacture process is utilized, the proposed model field outside of the depletion region is set to zero, meaning that
would yield a little larger error in estimating the photoresponse in a drift current only happens within the depletion region; (2) the state
shorter wavelength. To overcome this problem, the curve mapping of low-level injection, meaning that the variation of majority
scheme is applied to find the adequate boundary condition of the carrier concentration can be neglected.
surface recombination velocity with the least mean squared error A few parameters must be illustrated before going into details
thus allowing the proposed model to achieve a good performance. on deriving a CMOS photodiode model. First the “excess minority
The concept of the model proposed herein could be utilized in generation rate” using the function of G (x, λ) represents the
designing any CMOS photodiode for simulating its number of excess minority carrier produced within each unit,
photo-electronic characteristics in various industrial applications. 1/(cm3 × sec). The x represents the depth of the distance from the
incident light contact area of the semiconductor, whileλ stands
1. Introduction for the wavelength of the incident light. This mathematical
The CMOS process has increasingly becoming the mainstay in function explains that the excess minority generation rate is a
the IC market due to advantages of its low-power dissipation and function of the incident wavelength and the depth of distance from
high-density integration. Its applications will also be more the incident light contact area. When x is at the depth of the
prevalent as the photo-electronic conversion becomes one of the distance from the contact area of the semiconductor, exponential
enabling technologies. Figure 1(a) illustrates a conventional decay would happen to excess minority generation rate G (x, λ)
photodiode with a single layer [1]. However with advances in as shown below:
technologies, multi-layers as shown in Fig. 1(b) and 1(c) are now G ( x, λ ) = ϕ (x )× α × exp(−αx) (1)
available and prevailing. This work conducts a complete analysis,
The parameter of φ(x) is a photon flux, and x indicates the
simulation and measurement of the photo response characteristics
distance away from the incident surface. An exponential decay
of the CMOS photodiode. First, the physic characteristics of the
semiconductor will be analyzed. With equilibrium equations and following this distance will happen. If φ(0) means the photon
adequate boundary conditions, a CMOS photodiode model useful flux at the incident surface, φ(0) and φ(x) have the following
in any architecture design will be derived. Next the 1P3M CMOS relationship:
0.5 µ m process provided by the TSMC is used to design P
ϕ(0) = in (1− R) (2)
photodiodes with two different types of Nwell-PEpitaxial-Psubstrate and hν
Ndiffusion-Pwell-PEpitaxial-Psubstrate. Measurements are then done to test ϕ(x) = ϕ(0) × exp(−αx) (3)
their photo responses. Lastly, the results from measurement and Where R is a reflection rate, h is the Planck constant, Pin is the
simulation are compared and analyzed. By using the curve power of incident light, ν is the frequency of incident light, and
mapping scheme to find the adequate boundary condition with the α is the absorption coefficient of the photo-receiving material.
least mean squared error, the proposed model would be able to
The absorption coefficient is an inverse of the absorption length
achieve an optimized performance. Therefore, the CMOS
that indicates the length where the photon flux decays to 1/e of the
photodiode model proposed herein can be integrated with various
previous one. For the silicon material, the relationship between the
circuit simulators to provide system-on-chip simulations of
absorption length and light wavelengths is shown in Figure 2. If
industrial apapplications with CMOS sensors.
the wavelength isλ, then α could be represented by [2]:
 log10 α = 13.2131− 36.7985λ + 48.1893λ2 − 22.7562λ3 (4) considered here are electrons, with signs opposite of Jn.
In deriving the proposed CMOS photodiode, the basic J p = − qD n (∂ n p ∂ x )x = x (13)
d
structure as shown in Figure 1(a) is adopted first. As this research The above equations only solve the simplest photodiode
assumed from the beginning that the electric field could only exist model shown in Figure 1(a). Next, in comparing the CMOS
within the depletion region, the drift current then could only be photodiodes shown in Figure 1(b) and Figure 1(c) with a
generated within the depletion region. This is possible by conventional one, the most significant difference is the multi-layer
performing integration on the length of the depletion region based structure in the p-type semiconductor. Where the two P layers meet
on the carrier generation rate to get the number of all carriers. Each in the Nwell-PEpitaxial-Psubstrate photodiode as shown in Figure 1(b),
carrier is then multiplied by its electron charge to derive at the the new boundary condition is generated. This boundary condition
intensity of the drift current within the depletion region, as shown could be looked at from the physic characteristics of the current
below: density and minority carrier concentration. First look at the current
xd
(5) density. Due to the continuity of currents, the current densities at
J drift = − q ∫ G ( x , λ ) dx where the two layers meet would be equal thus creating the first
xn
boundary condition as follows:
Next the diffusion current, caused by the density imbalance
outside of the depletion region, is derived. As the photocurrent is J = −q× Dn1 × ( )
∂np1
∂x x=xp
= −q× Dn2 × ( )
∂np2
∂x x=xp
(14)
generated from the minority carrier, we can study its behavior at
Dn1 and Dn2 represent the diffusion coefficients in PEpitaxial and
the equilibrium state. In an N-type semiconductor where hole is
Psubstrate, respectively, whereas np1 and np2 are the densities of the
the minority carrier, the equilibrium equation is:
minority carriers in PEpitaxial and Psubstrate, respectively. The second
d 2 ( pn ( x) − pn 0 ) pn ( x) − pn 0 (6)
Dp − + G (λ , x) = 0 boundary condition is achieved from the continuity of
dx2 τp concentration of the minority carrier, as shown below:
In Eq. (6), Pn0 represents the density of the minority carrier [ ]
n p1 x p = n p 2 x p [ ] (15)
(hole) when the N-type semiconductor is at the equilibrium state; It is evident to see that from using the continuity relationship
Pn is all the minority carrier concentration after light incident; Dp of the current density and minority carrier concentration, the photo
as the hole’s diffusion coefficient and τp as lifetime of the hole. current equation of a multi-layer photodiode could be solved.
After deriving the general solution, two sets of boundary condition Hence, to derive the model of the Ndiffusion-Pwell-PEpitaxial-Psubstrate
must be included to derive the particular solution. The first set of photodiode in Figure 1(c), the continuity of current density and
boundary condition is at the surface area of x being 0: minority carrier concentration between Pwell-PEpitaxial and
[
Dp × ∂( pn (∂xx)−pn0 ) = − ( pn (x) − pn0 ) × S p (7) ]
x=0
PEpitaxial-Psubstrate can be utilized.
Lastly, one thing to note is the dark current. It is caused by
The Sp in Equation (7) is the surface recombination velocity
many reasons, such as temperature, density, biased voltage, and
of the hole at the semiconductor’s surface area. For the purpose of
defects in the manufacture process. The dark current concerned
simulation simplicity, this parameter is firstly set to infinity to
here is generated from bias voltage, and this term should be
derive at the boundary condition of Pn(0) being equal to Pn0. The
removed. To derive at the dark current generated by the biased
second set of boundary condition would be the boundary condition
voltage and carry densities, Pin of energy for incident light is set to
on the edges of the depletion region of x being xn. Assuming the
0 in the equation. The following equation derives the total current
low-level injection, the boundary condition is represented by the
density, where Jdark stands for the dark current:
following:
 qVbias 
Jtotal=Jn+Jdrift+Jp-Jdark (16)
− 
pn ( xn ) = p n 0 e  kT  (8) 3. Simulation and Analysis of the Photodiode Models
Vbias is the inverse-biased voltage added at the two sides of The simulation of the photo responses of
the photodiode, k as the Boltzmann constant, and T as the absolute Nwell-PEpitaxial-Psubstrate and Ndiffusion-Pwell-PEpitaxial-Psubstrate
temperature. The particular solution of Pn(X) could then be photodiodes is shown in Figure 3, and that at each region of the
derived using these two sets of boundary conditions. Lastly, photodiodes is shown in Figure 4 and Figure 5. The effect of that
incorporating the expression below would achieve the current the currents generated from these regions are quite different is
density generated in an N-type semiconductor: made by the ionic doping concentration and doping depth. Doping
J n = qD p (∂p n ∂ x ) (9)
x = xn concentration mainly affects the current density. The minority
Similarly, the current density in a p-type semiconductor could carriers, that need diffuse to the edge of the depletion region and
also be derived using the equilibrium equation as in Eq. (10) along then pass through this region, become photocurrents successfully.
with Eq. (11) representing the bottom layer of the semiconductor Hence, the more ionic doping concentration is the higher
and Eq. (12) as the boundary condition beneath the depletion probability of minority recombination is. In other words, the
region. induced current is decreased as the ionic doping concentration is
d 2 (n p ( x) − n p 0 ) n p ( x) − n p 0
increased[3]. We can observe that a greater ionic doping
(10)
Dn − + G (λ , x ) = 0 concentration causes a shorter minority carrier lifetime. In addition
dx 2 τn
to doping concentration, the doping depth affects the absorption
Dn ×
∂(n p ( x)−n p 0 )
∂x
[
= − (n p ( x) − n p 0 )× S n ]
x = xh
(11) wavelength. From Figure 2, a longer wavelength can penetrate into
 qVbias 
a deeper region. Therefore, a deeper junction depth has a better
− 
photo response for a longer wavelength and vice versa. Here, we
n p ( xd ) = n p 0 e  kT 
(12)
compare the photo responses of the above two photodiodes as
As in Eq. (13), photo currents generated by the p-type shown in Figure 4 and Figure 5. Due to the Nwell having a less
semiconductor could be accomplished. Note that the carriers doping concentration and a deeper depth than those of the Ndiffusion,
the photo response of Nwell has a greater value, and a response peak two photodiodes were conducted to find the optimized boundary
in a longer wavelength. Since the light wavelength nearby 400nm conditions. It is evident that the proposed photodiode model
is absorbed more by the Nwell region and less by the Ndiffusion region, would be useful in various circuit simulators to understand the
the current density in the P-type semiconductor of the responses of the CMOS photodiodes.
Ndiffusion-Pwell-Pepitaxial-Psubstrate photodiode has a better response in a

References
short wavelength. When comparing photo responses of the N-type
and P-type semiconductors, the P-type semiconductor has a less [1] J. Reginald and K. Arora, ”Using PSPICE to simulate the
ionic doping concentration and a deeper depth, and thereby it has a photoresponse of ideal CMOS integrated circuit photodiodes,”
large photo response to the incident light with a long wavelength. Proc. of IEEE Southeastcon Bringing Together Education,
Science and Technology, pp. 374 –380, 1996.
4. Comparison of Measurement and Simulation Results [2] H. Melchior, “Demodulation and photodetection techniques,”
Figure. 6 illustrates the measurement results of the photo in Laser Handbook Vol. 1, F. T. Arecchi and E. O. Schulz-Dubios,
responses of these two photodiodes. The Eds. Amsterdam:North Holland, pp. 725-835, 1972.
Ndiffusion-Pwell-Pepitaxial-Psubstrate photodiode shows a better response [3] D. Schroder., Semiconductor Material and Device
as we predicted in simulation. Next, simulation and measurement Characterization, New York: John Wiley and Sons, 1990.
results are compared. Figure 7 shows the measured and simulated [4] S. Sze, Physics of Semiconductor Devices, New York: John
photo responses under a biased voltage being 0V and the surface Wiley and Sons, 1980.
recombination velocity being infinity. When a wavelength is less N-well Xn N-diffusion
Depletion Region
Xn

N-well Xn Depletion Region Xd Xd

than 750nm, there is a larger difference between the simulation and Depletion Region Xd Epitaxy Layer X e Xd
X
P-well Xp

X
X Epitaxy
measurement results. As previously mentioned, the surface Xe

~
~
~
~
~
~
recombination velocity affects the number of minority carriers

~
~
~
P-substrate P-substrate
Xh Xh P-substrate Xh

generated by the minority carriers on surface contact areas. As

(a) (b) (c)
explainable from Eqs. (7) and (11), if the surface recombination
Fig.1 Cross sections of photodiodes.
velocity is set at infinity, the two sets of boundary conditions then (a)Nwell-Psubstrate.
could take on the following expressions: (b)Nwell-PEpitaxial-Psubstrate.
pn(0)=pn0 (17) (c)Ndiffusion-Pwell-PEpitaxial-Psubstrate.
np(xh)=np0 (18)
Equation (18) holds according to the law of physics. As
shown in Figure 2, under the influence of the absorption
coefficient, and the thickness of a few hundred micro meter for the
entire CMOS photodiode, the minority carrier generated by the
incident light is close to zero at the bottom of the semiconductor
[4]. The minority carrier is actually the number of minority carrier
under a balance state. However, it is impossible to make the
number of minority carrier equal to zero at the top surface of the
semiconductor. Hence, Eq. (17) may not be true. This means that
there are always more minority carriers in this section than the
number of minority carriers under a balance state. Setting the Fig.2 Relationship between absorption coefficients and
boundary condition at contact area of x being 0 is also another wavelengths.
reason behind the difference between measurement and simulation.
The effect of the surface recombination velocity is shown in Figure
8. The surface photocurrent yields the maximum when the surface
recombination velocity is zero and zero when that is infinite.
Figure 9 shows the photo responses from 400nm to 700nm under
several different surface recombination velocities.
To improve upon this error, the curve mapping scheme is
adopted to come up with a boundary condition set that allows the
least mean squared error between simulation and measurement.
After computation, the surface recombination velocity at contact
areas for Nwell-PEpitaxial-Psubstrate and Ndiffusion-Pwell-Pepitaxial-Psubstrate
photodiodes are 2x103(cm/sec) and 1x104(cm/sec), respectively.
Table 1 lists the comparison of the proposed and conventional Fig.3 Simulation results of Nwell-PEpitaxial-Psubstrate. and
CMOS photodiode models[1]. The model proposed herein can Ndiffusion-Pwell-PEpitaxial-Psubstrate photodiodes.
accommodate many structures with adaptive parameters to achieve AêW AêW AêW

a precise estimation of photoresponse. 0.4

0.35 0.35
0.4
0.35
0.4

0.3 0.3 0.3

5. Conclusion 0.25
0.2
0.25
0.2
0.25
0.2
0.15 0.15 0.15
A CMOS photodiode model useful in all kinds of CMOS 0.1 0.1 0.1
0.05 0.05 0.05
photodiode structures was accomplished. By using the TSMC λ λ λ

1P3M 0.5 µ m CMOS technology, photodiodes with two different

0.5 0.6 0.7 0.8 0.9 1 0.5 0.6 0.7 0.8 0.9 1 0.5 0.6 0.7 0.8 0.9 1

(1)N-type Semiconductor (2)Depletion Region (3)P-type Semiconductor

types of Nwell-PEpitaxial-Psubstrate and Ndiffusion-Pwell-PEpitaxial-Psubstrate (a) (b) (c)

were implemented. The simulation and measurements of these Fig.4 The simulated currents of the Nwell-PEpitaxial-Psubstrate
photodiode in each region.
 (a)N-type semiconductor. depthH umL 0
(b)Depletion region. 0.5
1
(c)P-type semiconductor.
1.5
AêW AêW AêW
0.5 0.5 0.5

0.4 0.4 0.4 0.05

0.3 0.3 0.3

Aê W
0
0.2 0.2 0.2

0.1 0.1 0.1 -0.05

λ λ λ
0.5 0.6 0.7 0.8 0.9 1 0.5 0.6 0.7 0.8 0.9 1 0.5 0.6 0.7 0.8 0.9 1

(1)N-type Semiconductor (2)Depletion Region (3)P-type Semiconductor 0.25

0.5
(a) (b) (c) 0.75
Fig.5 The simulated currents of the Ndiffusion-Pwell-PEpitaxial- wavelengthH nmL 1

Psubstrate photodiode in each region. (b)

(a)N-type semiconductor. Fig.8 The effect of the surface recombination velocity to the
(b)Depletion region. current density in the semisonductor surface.
(c)P-type semiconductor. (a)Surface recombination velocity being ∞.
(b)Surface recombination velocity being 0.

Fig.6 Measured responsivities of Nwell-PEpitaxial-Psubstrate and

Ndiffusion-Pwell-PEpitaxial-Psubstrate photodiodes.

(a)

(a)

(b)
Fig.9 The photo responses from 400nm to 750nm under several
different surface recombination velocities.
(a) Nwell-PEpitaxial-Psubstrate
(b) Ndiffusion-Pwell-PEpitaxial-Psubstrate
(b)
Table 1. The comparison of the proposed and the
Fig.7 Measured and simulated responsivities for zero bias
voltage and infinite surface recombination velocity .
conventional models.
Conventional The proposed
(a) Nwell-PEpitaxial-Psubstrate Models CMOS model
(b) Ndiffusion-Pwell-PEpitaxial-Psubstrate photodiode
depthH umL
Conditions model[1]
0
0.5 Surface Infinite Utilizing curve
1
recombination mapping to find
1.5
velocity the adequate
value
0.2
Parameter Bias voltage Zero only Any value
AêW
0.15
consideration Dark current Neglect Dependent on
0.1 from bias bias voltage
0.05 voltage
0
Single layer only Adaptive for all
0.25
0.5
Structures kinds of
0.75 photodiodes
wavelengthH nmL 1

(a)