A Distributed SPICE-model of A Solar Cell: IEEE Transactions On Electron Devices June 1996
A Distributed SPICE-model of A Solar Cell: IEEE Transactions On Electron Devices June 1996
A Distributed SPICE-model of A Solar Cell: IEEE Transactions On Electron Devices June 1996
net/publication/3062547
CITATIONS READS
87 36,670
2 authors, including:
SEE PROFILE
All content following this page was uploaded by Abdelhalim abdelnaby Zekry on 11 January 2021.
Abstract-In this paper, a distributed SPICE-model for a solar the solar cell area. For instance, in the dark case, the junction
cell is worked out. Special attention is paid to the problems voltage decreases as one moves away from the edge of a
of nonhomogeneous current distribution and the effective series metallization finger. It reaches its minimum value midway
resistance. Elaborate experimental techniques have been used to
determine the model parameters. They are based on the I-V between two adjacent fingers. This voltage drop is caused
characteristics, the small signal impedance, and the open-circuit by the forced horizontal flow of cell current in the relatively
voltage decay measurements. The SPICE simulation results are high resistance top layer. Accordingly, the current density
compared with the measured static and dynamic characteristics decreases as the distance from the metallization edge increases
of a solar cell. We found satisfactory agreement concerning
reaching a minimum value midway between the fingers. But
the static characteristics* whereas there is discrepancy in the
dynamic characteristics. This is because the SPICE-model of a the changes in the current density J is much greater than the
p - n junction diode contains only one time constant. It is found changes in the junction voltage V, since the current density
that two time constants are necessary for accurate description J depends exponentially on the junction voltage according to
of the dynamic performance. The techniques and methodologies the law of the junction. Under illumination if the solar cell
developed here are applicable to other junction diodes operating is short circuited the junction voltage increases away from
at high frequencies or under transient conditions.
metallization edge because of the series resistance. It reaches
maximum value midway between the fingers.
I. INTRODUCTION Consequently, a part of the photo current will be lost as a
dark current in the junction. Since the photo-generation process
I N the solar cell literature [l], [ 2 ] , a lumped circuit model
is normally used to describe the terminal electrical perfor-
mance of solar cells. Because of the partial metallization of
is homogeneous, the net collected current decreases away from
the metallization edge. Here again, any small change in the
junction V, leads to a larger one in the dark current. The
the front surface in solar cells to receive the incident solar
radiation, the cell current flows laterally in thin top layer of general case is the superposition of the previous two cases
where the solar cell is illuminated and has certain terminal
the cell. Since the current is distributed throughout the cell,
the lateral IR drop increases with increasing distance from output voltage V ( o ) according to the load value. Since the
the metallization edge. Consequently, the junction voltage also photocurrent makes V, to increase and V ( o ) makes it to
decreases leading to edge crowding of the cell current. This decrease, it is expected that the two effects, at least partially,
compensate each other resulting in a weaker dependence of
effect can badly be modeled by a one lumped series resistance.
Now, such simplification of the reality is not necessary because V, on position for the solar mode of operation.
Consequently, a distributed model rather than a lumped
of the presence of powerful circuit simulators such as SPICE.
A more accurate model is useful for both understanding the one must be basically used to more accurately describe the
internal electronic processes in the solar cell itself [ 3 ] and for electrical behavior of a solar cell and to see how massive is the
current crowding process under different operating conditions.
photovoltaic array applications [4]. Consequently, our purpose
In principle, the solar cell is divided into identical segments
is to develop a distributed SPICE-model for a solar cell.
in which the current distribution is the same. A segment is
Experimental techniques will be presented to determine the
model parameters. Comparison between SPICE simulation enclosed within dashed lines in Fig. l(a). It comprises the cell
area from the center line of a metallization finger to the line
results and the measurement results will be presented.
half way to the neighboring finger. The number of segments
n, amounts to 2 n f , where n f is the number of fingers. All
11. DISTRIBUTED MODELOF A SOLAR CELL these segments are connected in parallel. The problem now is
A conventional solar cell geometry is shown schematically reduced to the modeling of the segment itself. Therefore, each
in Fig. l(a). The front electrode consists of a metallic grid segment is divided into sufficiently enough small sections.
with a main collecting busbar. The active device itself is From the principle point of view, as the number of sections
an nf-p junction supplied by an ohmic metallic contact at or elements ne increases the accuracy will be improved to a
the back side and an antireflection coating at the front side. certain limit determined by the approximations made in the
Because of the partial metallization of the front surface the mathematical processing. In each section the current density
solar cell current is nonhomogeneously distributed throughout can be considered constant with its average value. Conse-
quently, it can be modeled with the equivalent circuit shown in
Manuscript received December 22, 1994; revised July 24, 1995. The review Fig. l(b) where the photocurrent &he represents the generated
of this paper was arranged by Editor P. K. Bhattacharya. photocurrent in the eth section, De the junction diode, R,, the
The authors are with the Department of Electrical Engineering, College of
Engineering, King Saud University, Riyadh 11421, Saudi Arabia. shunting resistance, and R,, the series resistance, basically of
Publisher Item Identifier S 0018-9383(96)03366-7. the top layer. It must be pointed out that the first section under
0018-9383/96$05.00 0 1996 IEEE
692 IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 43, NO. 5 , MAY 1996
*P
\
Fig 1 (a) A conventional solar cell geometry, and (b) a distributed solar cell model for one segment in a solar cell
metallization has no current source and a negligible series conductivity modulation is negligibly small. The theoretical
resistance. values of Rste calculated according to Appendix I are also
This model can be analyzed using SPICE circuit simulator plotted in Fig. 3 for sake of comparison with the measured
if the parameters of the circuit components are known. In the results. It is interesting to notice that Rste at low currents is
next section we will show how these model parameters can equal to half of the real value of the total series resistance
be determined. Rst,, where
OF THE MODELPARAMETERS
111. DETERMINATION
In this. section, methods to determine the static and the e=l
small signal parameters of the model will be presented. SPICE
and R,,, = R,, at low currents.
will be used to calculate the static I-V characteristics and the
For equal sections in the segment all series resistances
small signal impedance of the solar cell. Comparison of the
R,,, are equal and Rsto = (ne- 1)RS,,/2nf. From Fig. 4,
simulation results with the measured results will be carried out
Rsto = 0.3 ohm. The number of segments in the specimen
in order to see how accurate is the distributed model.
is 12 and the number of elements ne is assumed to be 5.
Substituting the above values in (1) one gets R,,, = 0.9 R.
A. Static Characteristics
The intersection of the extrapolated straight-line portion of
I ) Dark Characteristics: From the measured dark I-V char- the log I-V characteristics with the current axis at V = 0
acteristics of the solar cell shown in Fig. 2 we determined the results in the total saturation current Io of the solar cell
ideality factor n of the diodes in the equivalent circuit, the specimen. From Fig. 2 we get
reverse saturation current I,, of any diode D , and values for
the series resistance R,, and for the parallel resistance Rpe. Io = 1.2 x l o p 7 A.
An equivalent total series resistance Rsteof the solar cell is
Since all elements at low currents are approximately equally
determined by dividing the voltage difference AV between
biased, 10will be distributed among the elements in proportion
the measured curve and the extrapolated line shown in Fig. 2
to their respective areas. Therefore
by the solar cell current at this voltage difference. The total
series resistance Rsteas a function of the solar cell current I is
shown in Fig. 3. It is clear from this figure that Rste decreases
with the current I. In previous work [ 6 ] ,the series resistance where A is the cell area and SA is the element area. With
is modeled by a constant resistance independent of the current. SA = 0.066 cm2 and A = 4 cm2, I,, = 2 x lo-' A. The
This can be caused either by conductivity modulation of the slope of the straight line portion of the log1 versus V curve
lightly doped base region or to the current crowding near the results in an ideality factor n = 1.60.
edge of the metallization finger. After estimating the highest The shunt resis-tance Rp is the inverse slope of the I-V
possible injection level in the base it has been found that the characteristics at the origin. Rp amounts to 21 kR for this solar
~
50
40
30
20
10
-
U
E -10
L
c -20
!
2 -30
-'t -40
U
-901
. ""
C e l l voltage ( V I
.1 6 I I
v measured
For instance, from curve IV in Fig. 4, I p h = 78 mA. Substi-
- calculated
tuting this value together with 2nf = 12 and ne = 5 in ( 2 )
E 14
-E
I one gets I P h = 1.625 mA.
By inspecting the illuminated I-V characteristics in Fig. 4,
:.12 we notice that there is no regular downward shift of the I-
U
V curves as the light intensity increases. The curves rather
4
3 10
intersect each other. The second remark is the relatively small
open circuit voltage for the highest illumination levels. This
2
H
.Y
is mainly because of the increased cell temperature with
.08 - illumination level. A secondary cause for the intersection of
the I-V curves is the current crowding effect.
E
.06[ I I, I I I . I 1 B. Small Signal Impedance
0.10 0.20 0.40 0.60 0.80 1.0
For the sake of completeness, a small signal description
I (A)
of the solar cell is required. This is useful for understanding
Fig. 3. The series resistance Rste as a function of the dark current I. the internal electronic processes underlying the operation of
the solar cell. In principle, the small signal behavior at low
cell. For the elementary solar cells the shunting resistance R,, frequency reflects the static performance of the cell whereas
amounts of 1260 kR. the high frequency small signal behavior reflects the dynamic
2 ) The Illuminated I-V Characteristics: To determine the performance. That is, the small signal parameters sums the
remaining parameters in the circuit model of a solar cell, static and dynamic response. On the other side, that static
694 IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 43, NO. 5, MAY 1996
20kHZ /
0.0 0.2 0-1 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4 2.6 2.8
R (SI)
Fig. 7. Impedance diagrams for I = 15 mA.
(a)
Fig. 8. (a) The lumped model of a solar cell. (b) The magnitude of the
TABLE I admittance Y as a function of frequency at a dc cell current I = 5 mA. (c)
THEEFFECTIVE
TRANSIT
TIMEAND SERIES RESISTANCE AT The phase of the admittance versus frequency for dc cell current I = 5 mA.
DC CURRENTS
DIFFERENT DETERMINED BY DIFFERENT METHODS (d) The magnitude of the admittance Y versus frequency for dc cell current
I = 30 mA. (e) The phase of the admittance versus frequency for dc cell
current I = 30 mA.
VT
For more accurate description of the measured terminal Gn(0) = - (10)
Irn
small signal impedance and consequently more accurate values
of the SPICE parameters rT and Rs, we developed a lumped where I,,, and I,, are the dc recombination currents in the
model for the solar cell which takes into consideration the space charge region and neutral regions respectively.
dependance of Cm and gm on frequency. The space charge recombination conductance G,, is inde-
1 ) A More Accurate Model for the Solar Cell: The new pendent of frequency and it has the expression of (9). The
small signal model is shown in Fig. 8(a) where an inductance neutral regions conductance G, depends on frequency and
L must be added in series with a resistance RB and can be written as
the parallel combination of the diode capacitance C, and G, = IY,lcos0. (11)
the junction resistance rp. The addition of a parasitic
inductance is necessary because of the high capacitance of The resistance rp and the capacitance e?, are then given by
the large area solar cell can build with it a resonance circuit. Tp = l/(Gn + G,,(O)) (12)
+
RB = R,t0/2 Reo. The capacitance C, is composed of and
three internal capacitances; the junction capacitance C, which
is independent of frequency, the capacitance of the stored c,= c,+ c,,+ c,. (13)
excess charges in the space charge region C,, which is also From the equivalent circuit of Fig. 8(a) the total series resis-
independent of frequency, and the capacitance of the neutral tance Tt and the total reactance xt of the solar cell can be
regions C, which depends on frequency because of the long expressed by
diffusion path in the neutral regions around the transition
region.
The resistance rp is composed of two parallel resistances; and
the resistance due to recombination in the space charge region wc,rg
T,, and the resistance due to the recombination in the neutral Xt=wL-
region r,. The capacitances can be expressed by the following +
1 (wcprp)2'
relationships: The total admittance IKl and its phase Bt have the form
A
0.8
E
.e
..c
c
Y
0.6
.c
0
0
c
& 0.4
Eo
2
I-
o .2
0.0
10’ 1o2 lo3 10‘ io5 IO6 ,lo7
The frequency ( H t )
(b)
100
+measured
+simulated
50
0
ca
0
U
v
-50
.c
c
IO0
-1 50
10’ 10 * lo4 10’
The frequency ( H t )
(C)
Fig. 8. (Continued.)
2) Determination of G,,(O) and G,(O): In order to deter- first term in (18) is equal to I,,, and the other is equal to I,,.
mine thme conductances the recombination currents I,,, and It i s found that (18) is satisfied with I,,, = 1.09 x lo-’ A
I,, must be known. This is achieved by using the equation and I,, = 3.193 x A. Then one can calculate I,,, and
Irnat any given voltage V . By substituting these currents in
I = IosceV/2Vt + IoneV’Vt (I8) (9) and (10) one obtains the conductances GSn(0)and G,(O).
where V- is the voltage on the solar cell. I,,, and I,, are the The measured admittance and its phase is fitted by (16)
reverse saturation currents of I,,, and Irn, respectively. The and (17) at different biasing currents and representative fitting
ZEKRY AND AL-MAZROO. DISTRIBUTED SPICE-MODEL OF A SOLAR CELL 691
2
- measured
6 simulated A
1.5
c.
E
I
0
z
c
Y
z
c
0 1
0,
3
T0i
E
r .5
c
40
1.30 mA
20
0
n
t
8g -20
Y
0
x
5 -40
*
I
I-
-60
0.7 0.7
T=40 k Tz40.C
0.6 0.6
0.5 0.5
5: 0.4 - 0.4
v
c5 Vi = 0.35 V
1
X
CI
w
’0.3 0.3
0.2 L V1’0.2 v
0.2
01
t I I I 0
0 0.25 0.50 0.75 1.00 0 0.25 0-50 0.75 00
X (mm) X (mm)
(a) (b)
Fig. 9. (a) Junction potential distribution in darkness. (b) Potential distribution I*-(z)for an illuminated solar cell with I,, = 78 mA.
values determined from the small signal measurement is static I-V characteristics, the junction potential distribution,
because of the high contact resistance value R, added to the and the magnitude and phase of the solar cell admittance as
internal series resistance of (RSt,/2).The different values of a function of frequency at different dc operating points were
the parasitic inductance for the magnitude of the admittance obtained as outputs from SPICE.
IYI and its phase attributed to the four point measurement Figs. 2 and 4 show the simulated I-V characteristics together
techniques used here where there is a parasetic inductance with the measured ones for the solar cell specimen at different
with the cell itself and with the current sensing resistance. illumination intensities and temperatures. We also plotted the
While as T~~ is constant at about 10 ,us, the space charge I-V characteristics from the lumped model with R, = 0.15 R
lifetime T~~ increases with the biasing current from 1.3 11s to in Fig. 2 for sake of comparison. There is a good agreement
about 7 ps. In order to confirm the lifetime measurement we between the simulated and measured I-V characteristics at
measured the lifetime by using the well-recommended open T = 27°C for the lower photocurrents and at T = 40°C
circuit voltage decay method. The results are listed in Table I. for the higher ones which means that the solar cell is getting
We see from this table that 7,c-d is in good agreement with hotter during the measurement because of the power dissipated
the transit time from the impedance measurement. Both ?-,C,d in it. It is clear from Fig. 2 that the lumped model is less
and rz are greater than r,, because they take into consideration accurate than the distributed model at high current densities.
the effect of C, and C,. From the foregoing analysis, it has The junction voltage as a function of the distance II: measured
become clear that two time constants T, and T,, are required from metallization edge is shown in Fig. 9(a). The decrease
to accurately fit the admittance versus frequency curve. It is of the junction voltage with increased x due to R, is only
known that the built-in SPICE-model of the diode contains apparent at higher junction voltage; namely at VI = 0.6 V.
only one single time constant, the transit time T T . in the next The potential distribution according to (A. 1) in Appendix 1
section we will see how accurate the SPICE simulation is if is also shown in the same figure. It is clear from the figure
TT is taken to equal T Z , the effective transit time resulting that there is good agreement between the SPICE simulation
from the impedance diagram. results and the exact solution which indicates that the number
of elements taken here is sufficient for accurate simulation. In
addition, the drop in the junction voltage due to R, is smaller
Iv. SPICE SIMULATION OF THE SOLAR CELLS than one thermal voltage at the highest operating junction
Now, all the required SPICE parameters of the solar cell voltage of about 0.57 V for conventional solar cells.
equivalent circuit shown in Fig. 2 are determined. Then the Fig. 9(b) shows the junction voltage as a function of the
nodes of the circuit are numbered and the circuit is input to position 2 measured from the metallization edge at I p h = 78
the computer according to the known SPICE instructions. The mA. It is clear from this figure that when the solar cell is short
ZEKRY AND AL-MAZROO: DISTRIBUTED SPICE-MODEL OF A SOLAR CELL 699
1.0 I I I I I 1 I
+ Simulated by Pspice ylP I=5mA
+ Lumped model admittance YL r
'ii 0.8 - WE
-
5
c
CI
0.6 -
i
0
U
Y
*E O 4 -
L.
E
0
f 0.2 -
Fig. 10. Comparison of the lumped admittance with the Spice simulation
and measured admittance for a dc cell current I = 5 mA.
well as the small signal impedance. There is some discrepancy
between the measured and SPICE simulation results because
circuited, VI = 0, the lateral voltage drop will be maximum the built-in SPICE-model of the p-n junction diode neglects
with an increasing junction voltage as x increases which is the the frequency dependance of the neutral admittance. Our
opposite of the potential distribution in the dark case. As VI investigations showed that two time constants are required
increases, the junction voltage distribution becomes more flat to accurately account for the performance of the solar cell.
until it levels out at VI M 0.6 V. It must be mentioned that the characterization techniques
The magnitude of the SPICE-simulated admittance Y ( f ) and methodologies developed here are directly applicable to
of the solar cell as a function of frequency at a bias current photodetectors and other diodes.
of 5 mA is shown in Fig. 10 together with the lumped
model admittance and measured admittance. Because of the APPENDIX
small bias current of the cell the effect of distributed R, is
negligibly small. It is seen from these curves that the SPICE- The Equivalent Series Resistance RSt
simulated admittance deviates more than the lumped model
admittance from the measured admittance because the SPICE- According to Fig. 11 and based on Hauser theory of emitter
model built in the program does not take into consideration the current crowding [5] one gets the junction potential distribu-
frequency dependence of Y, while the lumped model takes tion
into consideration this frequency variation. Accordingly, the
built-in SPICE-model for the p-n diode must be modified to ac-
count for the frequency dependence of the neutral admittance.
For our specimens, however, the discrepancies between the
three admittances lie within acceptable accuracy limits. This
confirms the proposed model of a solar cell and the extraction
techniques of its parameters. where R,,, is the lateral series resistance of one section in a
segment. R,,, = ((~EWS)/(WE.LF)). IL(O)is the total cell
V. CONCLUSION current at x = 0.
We can now express the equivalent total series resistance
In this work a distributed SPICE-model for a solar cell is of one segment R,,, as
developed. Elaborate experimental methods have been deviced
to accurately determine the model parameters. They are based
on the static I-V characteristics and the small signal impedance (A.3)
measurements. The model parameters for an element in a
segment are the series resistance R,,, the shunting resistance Substituting x = W, in (A.1) and combining it with (A.2)
R,,, the subtrate resistance R d , , the reverse saturation current and (A.3) it follows that
I,,, the ideality factor N , the zero biase capacitance C,,,, the
built-in voltage Q, the grading coefficient m, the transit time
T T , the temperature, and the semiconductor material.
R,,, =
ln(I/ cos z )
z t a n z )Rsso. (
After determining the SPICE parameters of a solar cell For small values of IL(O) = I such that tan z = z and
we run the SPICE program and obtained simulation results. cosz = (1- ( z 2 / 2 ) ) ,(A.4) can be approximated to
It has been found that these results are in good agreement
with the measured illuminated and dark I-V characteristics as R,,,.= (Rsso/2). (A.5)
700 IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 43, NO. 5, MAY 1996
According to (A.5) the effective resistance of one segment at Abdelhalim Zekry was born in Menofia, Egypt,
relatively small values of the currents amounts only to half its on August 8, 1946. He received the B.Sc. and
M.Sc. degrees from Cairo University in 1969 and
real series resistance value Rss,. The total equivalent resistance 1973, respectively, and the Dr. Ing. degree from the
of the cell Rste can be expressed by Technical University of Berlin, in 1981.
He joined the Institute of Electrical Materials,
Technical University of Berlin, in 1975 as a sci-
entific coworker. In 1982, he became an Assistant
Professor, Department of Electronics and Comput-
where 2nf is the number of segments in the cell. ers, Faculty of Engineering, Ain Shams University,
Cairo In 1988, he moved to King Saud University,
Riyadh, Saudi Arabia, as an Associate Professor Since August 1992, he has
REFERENCES been a Professor of Electronics at the same university His main research
activities are devoted in semiconductors, and to chemcal vapor deposition
. _S. M. Sze. Phvsics
111 , ofSemiconductorDevices.
“ New York Wilev, 1981, Dr Zekry was awarded the best research prize of Ain Shams University
pp. 790-838. in 1987
[2] E. S. Yang, Microelectronic Devices. New York Wiley, 1988, pp.
357-386.
[3] F. D. No and T. D. Morgan, “SPICE-modeling of cascade solar cells,”
in IEEE Proc. Southeast Con$, IEEE Service Center, Piscataway, NJ,
1991, pp. 776-780. Abdulhameed Yousef Al-Mazroo (SM’86-M’88)
[4] N. A. Kennedy, A. F. Imece, and C. A. Gross, “Analysis of utility received the B Sc. degree from King Saud Univer-
interconnected photovoltaic systems,” in IEEE Proc. North American sity (formerly Riyadh University), Riyadh, Saudi
Power Symp., IEEE Service Center, Piscataway, NJ, 1987, pp. 131-141. Arabia, in 1980, the M Sc degree from University
[5] I. R. Hauser, “The effect of distributed base potential on emitter current of Michigan, Ann Arbor, in 1984, and the P h D
~ injection density and effective base resistance for a strip transistor degree from Virginia Polytechnic Institute and State
geometries,” IEEE Trans. Electron Devices, vol. ED-1 1, pp. 238-242, University, Blacksburg, in 1988, all in electrical
1964. engineering.
[6] D. L. Crook and J. R. Yeargan, “Optimization of silicon solar cell design From July 1980 to September 1981, he worked as
for use under concentrated sun light,” IEEE Trans. Electron Devices, a Teaching Assistant in the Department of Electrical
vol. ED-24, pp. 330-336, 1977. Engineering, King Saud University He joined the
[7] M. A. AlSweidan, “Development of vector-impedance meter,” Master same department as an Assistant Professor in January 1989 His current
Thesis, College of Engineering, King Saud University, Riyadh, Saudi research interests include thick film circuits, modeling of semiconductor
Arabia, 1994. devices, and characterization of magnetic materials and components