Opt Quant Electron (2014) 46:1379–1384

DOI 10.1007/s11082-014-9898-y

Complete optoelectronic simulation of patterned silicon

solar cells

Arghavan Arjmand · Dylan Mcguire

Received: 10 September 2013 / Accepted: 12 February 2014 / Published online: 25 February 2014
© Springer Science+Business Media New York 2014

Abstract In this paper, patterned silicon solar cells are fully modelled optically and electri-
cally using Lumerical Solutions’ optical simulation software, FDTD Solutions and electrical
simulation solver, DEVICE. The optical simulation calculates the spatial distribution of pho-
ton absorption in the silicon when the cell is illuminated by unpolarized sunlight with the
AM1.5 solar spectral intensity. The photon absorption data is converted into a spatial genera-
tion rate of electron-hole pairs. The electrical simulation uses this generation rate to calculate
the collection efficiency of electron and hole carriers, accurately accounting for surface and
bulk recombination in silicon. Two main designs are simulated, each with periodic structures:
a grating structure and a square pyramid structure. The short-circuit current as well as the
overall conversion efficiency for the two devices are calculated.

1 Introduction

There has been growing interest in increasing the efficiency of solar cells by adding wave-
length scale patterning or plasmonic structures to enhance the absorption of light and reduce
unwanted reflections. In order to compare different designs, it is essential to calculate the
maximum power conversion efficiency from sunlight to electricity. We simulate the 3D pat-
terns optically using the finite difference time domain (FDTD) method (Taflove and Hagness
2005) in order to calculate the full spatial distribution of absorbed photons, and therefore
the spatial distribution of generated electron-hole pairs, under solar illumination. This spatial
distribution is then included in an electrical simulation which solves the Poisson and drift-
diffusion equations. The combined optical and electrical simulations allow us to account for
both optical effects, such as unwanted reflections, as well as electrical effects, such as bulk
and surface recombination, to calculate the full power conversion efficiency for the design.
The absorption in these thin crystalline silicon solar cells is lower than the absorption of the
conventional much thicker amorphous silicon solar cells, hence yielding lower short circuit

A. Arjmand (B) · D. Mcguire

Lumerical Solutions, Vancouver, BC, Canada
e-mail: aarjmand@lumerical.com

123
1380 A. Arjmand, D. Mcguire

Fig. 1 Schematic of
a rectangular groove and b
square pyramid solar cell
structure

currents as well. This study aims to report and compare the performance of the two thin silicon
solar cell designs from the work of Ref. Chong et al. (2012), optically and electrically.

2 Simulation models

2.1 Optical simulation

The FDTD Solutions software (Lumerical Solutions Inc) is used for optical modelling of
the 2D and 3D silicon solar cell designs by calculating the absorption spectrum and hence
carrier generation in silicon. A planewave source is used to model the sunlight incident on
the device. The incoherent average of two orthogonally polarized sources is necessary to
correctly represent the incoherent unpolarized light. The two designs from Chong’s paper
studied here are shown in Fig. 1. For each of the designs, one unit cell of the grating is
simulated and periodic boundaries are used. Table 1 summarizes the design parameters for
the two structures. The effective thickness of silicon is kept at 3um for both cases. In the
rectangular groove design, silicon is etched to achieve the grating. In the square pyramid
design, the grating is made of TiO2 . In both devices, Palik’s optical material data (Palik
1998) are used for silicon and silica. CRC optical material data (Shiles et al. 1980) is used for
the aluminium contact and a constant refractive index of 2.5 is assumed for TiO2 . Because the
structures are assumed to be infinitely periodic, the contacts are not simulated in the optical
simulation.

Table 1 Design parameters

Name of design Grating parameters

Duty cycle (%) h (nm) d (nm)

Rectangular groove 50 1,000 600

Square pyramid 100 900 750

123
Complete optoelectronic simulation 1381

Fig. 2 Doping profile (log scale) of the rectangular groove (left) and square pyramid (right) solar cell

The absorption per unit volume is calculated by

1
L (r, λ) = − ω |E (r, λ)|2 I m [ε (r, λ)] (1)
2
In this equation, L is absorbed power per unit volume at wavelength λ, E is the electric field,
ω is the angular frequency and is the permittivity of the material. Absorbed power, electric
field and the permittivity are functions of position and wavelength λ. Absorption data is post
processed to reflect the AM1.5G illumination over the wavelength range 0.4–1.1µm.
The generation rate is calculated assuming every photon absorbed yields one electron-hole
pair.
 
λ
G(r )α ∫ L (r, λ) dλ (2)
hc
Here, G is the generation rate per unit volume, h is Planck’s constant and c is the speed of
light in vacuum. The generation rate is then used in the electrical simulation to study the
distribution of charge and current collection at the contacts.

2.2 Electrical simulation

The solar cells are electrically characterized using DEVICE software (Lumerical Solutions
Inc). The Poisson and drift-diffusion equations are solved on a finite element mesh. The
dopant concentration in silicon is introduced in the Poisson equation and the optical stimulus
imported from the FDTD simulation is accounted for by the continuity equation. Aluminium
and silver are used for the base and emitter contacts respectively. In both designs, the substrate
is assumed to have a background p-type doping concentration of 21016 cm−3 . The front and
back contact regions are n- and p-type diffusion doped respectively. Figure 2 shows a cross
section of the doping profile in both structures. In both devices, the work function values for
the contacts are 4.28 eV for the base aluminium and 4.26 eV for the emitter silver. The DC
permittivity values for TiO2 and silica are 80 and 3.9 respectively.
For silicon, the following is used: relative dielectric permittivity of 11.7, intrinsic work
function of 4.2 eV, electron and hole normalized effective mass of 1.18 and 0.8098 and
bandgap of 1.11452 eV (Selberherr 1984). The mobility due to lattice scattering is treated as

123
1382 A. Arjmand, D. Mcguire

a basic input into the DEVICE semiconductor model, and may be entered as a constant value
or with a temperature dependence described by the temperature model,
 η
T
μn,
L
p (T ) = μn,
L
p (300)
300

where μn, L (300) is the value of the mobility at T = 300 K, and η is a temperature exponent
p
and subscripts n and p refer to electrons and holes, respectively. Here, the system is simulated
at T = 300 K and electron and hole lattice scattering coefficients of 1,471 and 470.5 cm2 /v-s
were used (Masetti et al. 1983). The Masetti model is used in these DEVICE simulations to
account for the influence of impurities on the carrier mobility in silicon and the coefficients
in Masetti et al. (1983) are used yielding a minority carrier diffusion length of 274 µm.
In the electrical simulation, bulk and surface recombination loss mechanisms are modelled:
For bulk silicon, trap-assisted recombination, Auger recombination and radiative recombi-
nation processes are taken into account. For trap assisted recombination, the electron and
hole lifetimes are 1e−6 and 4e−7 s. For Auger recombination, the electron and hole capture
coefficients are 2.8e−31 and 9.9e−32 cm6 /s, respectively. For radiative recombination, the
capture rate is 1.6e−14 cm3 /s (Selberherr 1984).
Surface recombination at the interface of silicon with SiO2 is modelled assuming a carrier
recombination velocity of 100 cm/s (Eades and Swanson 1985). For the square pyramid
design, the carrier recombination velocity at the interface of silicon with TiO2 is 1,000 cm/s,
chosen much larger than that of the native oxide. Surface recombination at the interface
of silicon with the two contacts is modelled assuming a carrier recombination velocity of
1,000 cm/s (Kelzenberg et al. 2008), reflecting the minority carriers away from the contacts
and towards the junction.
The simulation for the rectangular groove design is carried out in 2D. The simulation for
the square pyramid design can be done in both 2D and 3D. For the 2D simulation, the 3D
generation rate data calculated in the optical simulation is averaged in the y direction for use
in the 2D electrical stimulation.

3 Results and discussion

The spatial distribution of the generation rate under incoherent illumination for the two
designs is shown in Fig. 3. The generation rate pattern closely follows the absorption pattern
in silicon.
The ideal short circuit current from the optical simulation for each design is listed in
Table 2, assuming every absorbed photon turns into an electron hole pair and is collected at
the contact. Even though refractive index of TiO2 is much smaller than that of silicon, the
pyramid patterning provides better light coupling into the bulk silicon layer yielding greater
absorption.
Figure 4 shows the I − V and power curves from the three electrical simulations including
the loss mechanisms such as bulk and surface recombination. The bias across the contacts
is varied from 0 to 0.7 V. The short circuit current and the overall conversion efficiency can
be extracted from these plots. Table 2 lists the short circuit current values for each design.
In both designs, the short circuit current is reduced due to bulk recombination. 2D and 3D
short circuit current values for the pyramid design are in close agreement implying that for
this particular symmetry, computation time and resources can be saved by running only the
optical simulation in 3D and the electrical simulation in 2D.

123
Complete optoelectronic simulation 1383

Fig. 3 Generation rate of electron-hole pairs for the rectangular groove (left) and square pyramid (right) solar
cells

Table 2 Solar cell short circuit

Name of design Jsc ideal (mA/cm2 ) Jsc (mA/cm2 ) η (%)
current and efficiency
Rectangular groove 23.1 18.9 9.2
Square pyramid 2D 25 19.9 9.4
Square pyramid 3D 25 20.1 9.5

Fig. 4 Current desity and power versus voltage for solar cell designs

Despite having a larger short circuit current, the pyramid design has a smaller open circuit
voltage. This is because the reverse bias saturation current is larger for the pyramid design
with TiO2, than the rectangular design using the native oxide SiO2. This can be accounted
for based on the difference in the surface recombination velocities at the interface of silicon
with TiO2 (1,000 cm/s) versus SiO2 (100 cm/s). This effect is particularly exaggerated since
the interface is not doped.
Assuming an input power of 100 mW/cm2 , the overall conversion efficiency of the solar
cells is calculated and shown in Table 2. The gratings in both designs minimize reflections
from the surface; however, for the same effective thickness of silicon, the rectangular groove
grating has increased silicon-oxide surface area which leads to more surface recombination

123
1384 A. Arjmand, D. Mcguire

at this interface, though the silicon/TiO2 interface has higher carrier recombination velocity.
The conversion efficiency of the square pyramid design is higher than that of the rectangular
grating design by about 3 %. For complex designs such as the ones studied here, this proves
the importance of carrying out both optical and electrical simulations to characterize and
further optimize the structure in order to determine the domaninant contributing mechanisms
in the behaviour of the device.

4 Conclusion

Patterned silicon solar cells have been fully simulated optically and electrically using the
FDTD method and a drift-diffusion solver. Optical simulation account for the incoherent
nature of sunlight and the 1.5AM solar spectrum. Electrical simulations account for the optical
stimulus, doping concentrations, bulk and surface recombination processes. Such complete
characterization of these devices allows for the calculation of full optical-to-electrical figures
of merit which is essential for optimizing the power conversion efficiency of new solar cell
designs.

References

Chong, T.K., Wilson, J., Mokkapati, S., Catchpole, K.R.: Optimal wavelength scale diffraction gratings for
light trapping in solar cells. J. Opt. 14, 024012 (2012)
Eades, W.D., Swanson, R.M.: Calculation of surface generation and recombination velocities at the Si-SiO2
interface. J. Appl. Phys. 58(11), 4267–4276 (1985)
Kelzenberg, M.D., Filler, M.A., Kayes, B.M., Putnam, M.C., Turner-Evans, D.B., Lewis, N.S., Atwater, H.A.:
Single-nanowire Si solar cells. In: Proceedings of the 33rd IEEE PVSC, p. 1 (2008)
Lumerical Solutions Inc., Canada. http://www.Lumerical.com
Masetti, G., Severi, M., Solmi, S.: Modeling of carrier mobility against carrier concentration in arsenic-,
phosphorus-, and boron-doped silicon. IEEE Trans. Electron Devices ED–30, 764–769 (1983)
Palik, E.D.: Handbook of Optical Constants of Solids. Academic, San Diego (1998)
Selberherr, S.: Analysis and Simulation of Semiconductor Devices. Springer, Vienna (1984)
Shiles, E., Sasaki, T., Inokuti, M., Smith D.Y.: Self-consistency and sum-rule tests in the Kramers-Kronig
analysis of optical data. Application to aluminium. Phys. Rev. B. 22, 1612–1628 (1980)
Taflove, A., Hagness, S.C.: Computational Electrodynamics: The Finite-Difference Time- Domain Method,
3rd edn. Artech House, Boston (2005)

123