You are on page 1of 6

Radiation Effect Models in Solar Cells Comparison of Simulations

with Experimental Data


Alex Fedoseyev1, Tim Bald1, Marek Turowski1, Ashok Raman1, Cory Cress2, Robert Walters2, Jeff
Warner2, and Seth Hubbard3
1
CFDRC, Huntsville, Alabama 35805, USA; 2NRL, Washington, DC 20375, USA; 3RIT, Rochester, New
York 14623, USA

Abstract We present the radiation effect models for solar interface and bulk trap models enables simulations of charge
cells used in NanoTCAD device simulator and provide trapping and de-trapping in both steady-state and transient
comparison with experimentally measured on solar cell analyses. This provides the means for accurate simulations of
performance. The device modeled is a p+ n GaAs solar cell. Dark,
Light IV curves and corresponding performance parameters are solar cell performance and their degradation due to effects of
simulated and compared with experimental results for 2 MeV space radiation. To demonstrate the performance of the
protons at varying fluence. Majority carrier defect introduction models, we will present the 3D simulation results and
rates for n-type GaAs used in the NanoTCAD simulations were comparison with experimental data for the radiation effects on
taken from deep level transient spectroscopy (DLTS) data PV cells.
provided by the Naval Research Lab (NRL). Results show a good
match between simulations and experimental results.
Index Terms Solar cells, Radiation Effects Models, II. TRAPS MODELING IN CFDRCS NANOTCAD 3D SIMULATOR
Simulation, Experimental data, TCAD
In collaboration with NRL, we analyzed the basic physical
phenomena related to the radiation-induced degradation of
I. INTRODUCTION solar cells and associated data (e.g. obtained by DLTS). We
In the high-radiation open space environment the solar cells have implemented an advanced model of bulk defects/traps in
suffer from degradation, which depends on the materials and CFDRCs NanoTCAD three-dimensional (3D) device-physics
particular design of the photovoltaic cells. In the natural, high- simulator to represent the radiation-induced damage in the
radiation environment of space all solar cells suffer from appropriate solar cell layer(s). This model allows the user to
degradation. Although studies have been conducted and test specify various trap configurations, depending on the
data collected on the performance of solar cells in a radiation degradation to be modeled (radiation related, processing
environment, accurate models and software able to predict the related, etc.), and is based on descriptions provided in
radiation-induced degradation is not available in the literature [3], [4]. The traps occupancy equations are solved
commercial TCAD software. self-consistently with the basic semiconductor equations
The damage caused by radiation exposure consists of the (Poisson, carrier continuity) to yield modified device results.
generation of so-called Frenkel pairs. This generation is the The detailed description of models is provided in the Section
process in which atoms are dislodged from a lattice site to an below.
interstitial position, leaving behind a vacancy. The interstitial A. Trap Configurations and Semiconductor Equations
can migrate away until it finds a stable site, such as a surface.
This is different from the radiation effects observed in CFDRCs NanoTCAD three-dimensional (3D) device-
semiconductor submicron devices [1], [2]. physics simulator allows the user to specify various trap
The consequence of radiation damage is the formation of configurations, depending on the degradation to be modeled
centers where efficient non-radiative (phonon) recombination (radiation related, processing related, etc.). Four different
can occur. As a result, minority carriers have a shortened kinds of traps can be specified: (a) Donor traps (Nt-D):
lifetime when the material has been damaged by radiation. Positively charged when empty, Neutral when occupied; (b)
This in effect shortens the diffusion length and, therefore, Acceptor traps (Nt-A): Negatively charged when empty,
reduces the volume of material, which responds to the optical Neutral when occupied; (c) Neutral Electron traps (Nt-N-E):
excitation. Neutral when empty, Negatively charged when occupied; and
We are developing a comprehensive three-dimensional (3D) (d) Neutral Hole traps (Nt-N-H): Neutral when empty,
physical model for radiation-induced lattice defects (traps), Positively charged when occupied.
featuring: (i) multiple energy levels in semiconductor bandgap, Trapped electron and hole densities at a given energy level
(ii) various, user-defined trap densities for each energy-level are related to occupation probabilities for electrons (f0) and
(within bandgap), (iii) user-defined capture cross-sections for holes (1-f0).
each trap level. This model has been implemented into In the case of multiple traps distributed in energy, the
CFDRC 3D device simulator NanoTCAD. The addition of bandgap of the semiconductor is divided accordingly and the

978-1-4673-0066-7/12/$26.00 2011 IEEE 002801


occupied trap concentration at each energy level, Et, is nt N t nt
calculated as an integral between these sub-levels. Gn = , Gp =
In NanoTCAD, we commonly use the drift-diffusion (DD) en ep
approach to calculate the transport of charged carriers (5a), (5b)
(electrons and holes) and electric potential distribution in the
Here, Rn/p is the electron/hole capture rate, Gn/p is the
3D model, in the presence of various external fields and source
electron/hole emission rate, nt is the occupied trap density,
terms. The DD approach involves the solution of the Poisson
cn/cp is the electron/hole capture time constant, and en/ep is the
equation for electrical potential, and the carrier continuity
electron/hole emission time constant. The net generation-
equations for electron and hole densities.
recombination term used in the carrier continuity equations 2-3
In the presence of traps, the Poisson equation is modified as
is calculated as: GTn/p = Gn/p Rn/p.
follows:
The various time constants are calculated as shown below:
p n + N D+ N A + cn1 = vth, n n n (6a)

( N t D nt D )
+

( ) = q (1)
cp1 = vth , p p p
( N t A p t A ) +
(6b)
( p +
N H ) (n N E )
EC Et
en1 = vth, n n N C exp (7a)
where is the dielectric constant of the semiconductor kT
material, is the electric potential, q is the electronic charge,
p is the density of holes, n is the density of electrons, ND+ is
the density of donor ions (due to doping), NA- is the density of Et EV
ep1 = vth , p p N v exp (7b)
acceptor ions (due to doping), Nt-D is the total donor trap kT
concentration, nt-D is the trapped electron concentration of the
donor trap level, Nt-A is the total acceptor trap concentration, In these equations, vth,n/th,p is the electron/hole average
pt-A is the trapped hole concentration of the acceptor trap level, thermal velocity, n/p is the electron/hole average capture
Nt-N-E is the total neutral electron trap concentration, nt-N-E is cross-section, n/p is the electron/hole concentration, Nc/v is the
the trapped electron concentration of the neutral electron trap effective density of states of electrons/holes, Ec/v is the
level, Nt-N-H is the total neutral hole trap concentration, and pt- conduction/valence band energy, k is the Boltzmann constant
N-H is the trapped hole concentration of the neutral hole trap
(1.3810-23 J/K), and T is the absolute temperature.
level, and the other terms are as defined in the previous This model assumes that the electron and hole cross-sections
paragraphs. Additionally, implies summation over all the for capture and emission are constant and equal to each other.
energy levels included in the trap definition. The trap dynamics equation (to determine occupied trap
The carrier continuity equations are modified as shown density) is given by:
below: dnt n f N
= Rn Gn R p + G p = t + 0 t (8)
n
dt
q J n = q (G R + GT ) (2)
t
1 1 1 1 1
= + + +
p
cn en cp ep
q + J p = q (G R + GT ) (3) (9a)
t
1 1
where Jn and Jp are the electron and hole current densities, f0 = + (9b)

respectively, G and R represent the total generation and cn ep
recombination terms due to all other sources (e.g., Shockley-
Read-Hall (SRH), Auger, impact ionization, etc.), and GT After substituting with the expressions for the various time
represents the generation-recombination due to the traps. constants shown in equation 10-11, the rate of carrier flow for
The traps model itself is based on descriptions provided in each trap level can be calculated.
literature [3], [4]. The trap dynamics for single-level interface B. Steady State Solution of the Trap Dynamics Equation
traps may be represented as shown below:
The analyses of electrical and radiation response of solar
N t nt nt cells are typically performed under the steady state
Rn = Rp = assumption. This implies that dnt/dt = 0. Therefore, we can
cn cp
, (4a), (4b)

978-1-4673-0066-7/12/$26.00 2011 IEEE 002802


solve the electron capture probability function, f0, by simple substrate layer. Each contact is modeled to be perfectly ohmic.
resolution of equations 6-9. This model was designed such that the layers under the top
The charge contribution is calculated for each trap level and contact would be opaque while the active region would absorb
then integrated over all trap energy levels to determine the the incident light similar to an actual photovoltaic device.
total contribution to the Poisson equation.
Please note that, .The recombination term for each trap level
can be reduced to the standard SRH form:

np ni2
Rt = (10)
p (n + nl ) + n ( p + p1 )
if we define electron and hole lifetimes as
n p
p = 1 v n Nt
th and n = 1 v p Nt
th respectively
[4]. Also, as described earlier, the total recombination term in
the continuity equations due to traps is calculated by
summation across all traps levels.

III. DEVICE MODEL Fig. 1: Schematic of NRL p+ n GaAs solar cell.

A large amount of data for a p+n GaAs solar cell sample was
received by CFDRC from NRL which provided detailed
results of the effect that radiation damage has on solar cell IV. SIMULATION RESULTS AND DISCUSSION
performance. The data provided by NRL included dark and We were able to utilize the newly formulated trap model in
light IV curves and EQE for various levels of fluence at 2 the NanoTCAD solver to simulate the effects of radiation
MeV Proton irradiation as well as DLTS (Deep Level damage on the p+n GaAs solar cell. The carrier trap data used
Transient Spectroscopy) results. Some of these results are
for the post radiation simulations were derived from several
published in [5], [6].
sources. The DLTS data supplied by NRL provided the
The device for this data set was grown on an n-type GaAs
introduction rates of 2 MeV protons for majority carrier n-type
substrate by MBE. A buffer layer doped with Si to a
traps. The activation energies and trap cross sections of
concentration of 1x1018 cm-3 was first grown on the substrate
majority carrier n-type traps were derived from the DLTS
at a thickness of 0.5 m. This was followed by a 3 m n-type
results of [5], [7]. An example of this data is shown in Table 1
base region with a doping concentration of 2x1016 cm-3. A 0.5
below. Each activation energy is given with respect to the
m p-type emitter was then grown and doped with Be to a
valance band. In addition, the activation energy and capture
level of 4x1017 cm-3. A 30 nm p-type Al0.85Ga0.15As window
doped with Be to a concentration of 2x1018 cm-3 was then cross section for each trap is accompanied with its respective
grown, followed by a 0.5 m GaAs cap layer doped to a standard deviation.
concentration of 1x1019 cm-3 as shown in Fig. 1.
A scale model of the device described above was made to Table 1: Majority carrier n-type traps
simulate the performance of the device for pre and post-
radiation. The device model area was set to 6 m x 1 m with Trap E(eV) S.D. (eV) (cm^2) S.D. (cm^2)
the cap layer area set to 1 m x 1 m. This was included to PR1 0.710 0.083 2.00E-12 1.00E-12
capture the transport effects caused by the GaAs-AlGaAs- PR2 0.854 0.042 2.10E-13 8.00E-14
GaAs hetero interface for photogenerated carriers. The
PR4' 1.105 0.017 2.20E-14 1.10E-14
remaining area of 5 m x 1 m acted as the active
photoabsorption region of the model cell. All layers, PR4'' 1.145 0.009 7.80E-15 1.80E-15
thicknesses and doping layers were consistent with the device PR5 1.359 0.004 4.00E-15 1.90E-15
described above. The substrate in the device model was given
a thickness of 5 m. This was to ensure adequate For the post radiation simulations, traps were applied to the
photoabsorption and limit the size of the model as actual appropriate regions. Electron traps, labeled as PR in Table
substrates are hundreds of microns thick. Profile and 1, were applied as majority carrier traps in the n-type base. For
perspective views of this model can be seen in Fig. 1. The top each trap, two separate definitions were made to account for
contact of the device is located on top of the cap layer and the the empty and filled states of the trap. For an electron trap,
bottom contact is located beneath the entire area of the the empty state corresponds to when the trap is positively

978-1-4673-0066-7/12/$26.00 2011 IEEE 002803


charged and has a capture cross section, n. When the trap is shows the light penetration and absorption for 560 nm and 800
filled with an electron, it becomes a neutral hole trap with a nm.
cross section that is 100 times smaller than when it was empty. For a more detailed look into the behavior of the simulated
This is done for simulation purposes due to the lack of device under light conditions, the external quantum efficiency
minority carrier trap data. The same approach is used when (EQE) was calculated for pre and post-radiation. From the
applying hole traps to the p-type emitter. results shown in Fig. 4, the simulated results showed a good
To ensure that this model would behave as expected, dark match for nearly all wavelengths for both pre and post
IV simulations were done for pre and post-radiation radiation conditions as well as an excellent match to the
conditions. The post-radiation simulation was done for 2 MeV degradation trend for increasing fluence. However,
proton irradiation for fluences ranging from 6x1010 cm-2 to discrepancies are apparent in the medium length wavelengths,
5x1012 cm-2. Fig. 2 shows the results of the dark IV particularly between 500 nm and 700 nm for the post-rad
simulations as compared to measured data. An excellent match simulations. This could be, in part, due to the choice of capture
was obtained for the dark IV simulations. cross sections used in the simulations. The cross sections used
were within the reported standard deviations, but using slightly
5.00E-03
larger cross sections would provide a better agreement for
4.50E-03 Measured Pre-Rad
(a) individual wavelengths.
TCAD Pre-Rad
4.00E-03 Measured 6E10
TCAD 6E10
3.50E-03 Measured 2E11
TCAD 2E11
(a) (b)
3.00E-03 Measured 5E11
Current (A)

TCAD 5E11
2.50E-03 Measured 1E12
TCAD 1E12
2.00E-03 Measured 5E12
TCAD 5E12
1.50E-03

1.00E-03

5.00E-04

0.00E+00
0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 1.3
Voltage (V)

1.00E+00

1.00E-01

1.00E-02
(b)
Fig. 3: Photoabsorption at 560nm (a) and 800nm (b).
1.00E-03

1.00E-04

Measured Pre-Rad
From the results of the EQE simulation, the short circuit
1.00E-05
Current (A)

TCAD Pre-Rad current (Isc) was calculated by multiplying the EQE by the
1.00E-06 Measured 6E10
TCAD 6E10 AM0 spectrum and integrating over all wavelengths. This was
1.00E-07
Measured 2E11
TCAD 2E11 done for both pre and post radiation conditions.
1.00E-08
Measured 5E11
TCAD 5E11 1.00
1.00E-09
Measured 1E12
0.90
1.00E-10 TCAD 1E12
Measured 5E12 0.80
1.00E-11
TCAD 5E12
1.00E-12 0.70
0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 1.3
Voltage (V) 0.60
EQE

0.50
Measured Pre-Rad
Fig. 2: Pre and Post radiation dark IV simulations for TCAD model 0.40 TCAD Pre-Rad
Measured 6E10
of NRL cell. (a) Linear scale, (b) Log scale 0.30 TCAD 6E10
Measured 2E11
TCAD 2E11
0.20 Measured 5E11
The model was also used to simulate the behavior of TCAD 5E11
Measured 1E12
0.10
performance under light conditions. The device was simulated TCAD 1E12
0.00
under an AM0 spectrum for fluences between 6x1010 and 300 400 500 600 700 800 900 1000
1x1012 cm-2. In conjunction with the same trap data presented Wavelength (nm)

above, the model was simulated from 300 nm to 960 nm in 20


nm increments. Refractive index data supplied by SOPRA data Fig. 4: Measured and simulated EQE results for the NRL cell.
tables and NRL were used by the solver to calculate the Simulated results show a good match for each wavelength as well as
absorption coefficients for each wavelength and material. An the degradation trend for increasing fluence.
example of photoabsoption is shown in Fig. 3. The figure

978-1-4673-0066-7/12/$26.00 2011 IEEE 002804


Degradation of Performance Parameters
1.05
The Isc was then added to the dark IV results displayed in 1 Measured

Norm Jsc
Fig. 2 to produce the simulated light IV curves. These curves, 0.95 TCAD
0.9
shown in Fig. 5, demonstrate an excellent match of the overall 0.85
degradation for increasing fluence. However, it can be seen 0.8
that there are discrepancies in the absolute values of the Isc and 0.75
open-circuit voltage (Voc) between the measured and simulated 1.E+10 1.E+11 1.E+12 1.E+13
data. These values differ by an average of ~5% for each
fluence. 1

Norm Voc
0.95

Measured Data 0.9


0 0.85
0 0.2 0.4 0.6 0.8 1
-0.1 0.8
Measured Pre-Rad
1.E+10 1.E+11 1.E+12 1.E+13
-0.2
Measured 6E10
-0.3 1
Measured 2E11
Current (mA)

Norm FF
-0.4 Measured 5E11 0.95

-0.5 Measured 1E12 0.9

-0.6 0.85

-0.7 0.8
1.E+10 1.E+11 1.E+12 1.E+13
-0.8
1
-0.9
Voltage (V) Norm Eff
0.9

Simulated Data
0.8
0.0
0.7
0.0 0.2 0.4 0.6 0.8 1.0 1.2
-0.1 0.6
0.5
-0.2
TCAD Pre-Rad
1.E+10 1.E+11 1.E+12 1.E+13
-0.3 TCAD 6E10 -2
Proton Fluence (cm )
Current (mA)

TCAD 2E11
-0.4
TCAD 5E11
TCAD 1E12
Fig. 6: Normalized degradation of performance parameters for
-0.5 fluences ranging from 6x1010 to 1x1012 cm-2.
-0.6

-0.7

-0.8
V. CONCLUSION
-0.9
Voltage (V) Radiation effect models for solar cells used in NanoTCAD
device simulator have been presented. Comparisons with the
Fig. 5: Measured and simulated light IV curves for pre and post experimentally measured data on solar cell performance from
radiation conditions. An excellent match is observed for the
degradation trend of the curves. Absolute values for Isc and Voc
NRL are provided. The device model, a p+ n GaAs solar cell,
differ by an average of ~5% between measured and simulated data. has been used. Dark and light IV curves and corresponding
performance parameters are simulated and compared with
experimental results for 2 MeV proton irradiations at varying
Using the results from the light IV simulations, the fluence. Majority carrier defect introduction rates for n-type
performance parameters (JSC, VOC, Jmax, Vmax, Pmax, FF, and GaAs used in the NanoTCAD simulations were taken from
Efficiency) were calculated. Each parameter was normalized deep level transient spectroscopy (DLTS) data provided by the
with respect to the pre-radiation data to ascertain the extent of Naval Research Lab (NRL). Results show a good match
degradation and compared to that of the NRL results. The
between simulations and experimental results.
comparisons shown in Fig. 6 indicate a good match for each
performance parameter, particularly for the overall efficiency.
ACKNOWLEDGEMENT
Support from DTRA SBIR program is greatly appreciated.

978-1-4673-0066-7/12/$26.00 2011 IEEE 002805


REFERENCES

[1] M. Turowski, A. Raman, and R. Schrimpf, Non-Uniform


Total-Dose-Induced Charge Distribution in Shallow-Trench
Isolation Oxides, IEEE Trans. on Nuclear Science, 51, 2004,
3166.
[2] A. I. Fedoseyev, M. Turowski, A. Raman, E. W. Taylor b, S.
Hubbard, S. Polly, Q. Shao and A. A. Balandin Space radiation
effects modeling and analysis of quantum dot based
photovoltaic cells, Proc. SPIE 7467, 746705 (2009);
doi:10.1117/12.826691.
[3] P. Habas, Analysis of Physical Effects in Small Silicon MOS
Devices, Ph.D. Thesis, Tech. University of Vienna, Institute for
Microelectronics, 1993; Online at:
http://www.iue.tuwien.ac.at/phd/habas/.
[4] Robert F. Pierret and Gerold W. Neudeck (editors), Advanced
Semiconductor Fundamentals, Modular Series on Solid State
Devices Volume VI, Addison-Wesley Publishing Company,
July 1989.
[5] J. H. Warner, G. P. Summers, R. J. Walters, S. R. Messenger,
Energy dependence of majority carrier defect introduction rates
in p+ n GaAs photodiodes irradiated with protons, Journal of
Applied Physics., vol. 96, no. 12, pp. 72257228, 2004.
[6] X. J. Chen, H. J. Barnaby, J. H. Warner, S. R. Messenger, R. J.
Walters, S. A. Ringel, J. Park, Non-Linear Behaviors of Dark
Current Slope in P p+n GaAs Solar Cells Following Proton
Irradiations, in 34th IEEE Photovoltaic Specialist Conference,
2009, p. 001565.
[7] J. H. Warner, R. J. Walters, S. R. Messenger, G. P. Summers, S.
M. Khanna, D. Estan, L. S. Erhardt, A. Houdayer, High-
Energy Proton Irradiation Effects in GaAs Devices, IEEE
Transactions on Nuclear Science., vol. 51, no. 5, pp. 2887
2895, 2004.

978-1-4673-0066-7/12/$26.00 2011 IEEE 002806

You might also like