Calhoun: The NPS Institutional Archive

Faculty and Researcher Publications Faculty and Researcher Publications

2005

The design and optimization of

advanced multijunction solar cells using

the Silvaco ATLAS software package

Michael, Sherif

þÿ�S�o�l�a�r� �E�n�e�r�g�y� �M�a�t�e�r�i�a�l�s� �&� �S�o�l�a�r� �C�e�l�l�s�,� �V�o�l�.� �8�7�,� �(�2�0�0�5�)�,� �p�p�.� �7�8�5 ��7�9�4

http://hdl.handle.net/10945/46458

ARTICLE IN PRESS

Solar Energy Materials & Solar Cells 87 (2005) 785–794

0927-0248/$ -

doi:10.1016/j

�Correspo

E-mail ad

www.elsevier.com/locate/solmat

The design and optimization of advancedmultijunction solar cells using the Silvaco ATLAS

software package

Sherif Michael�, Andrew Bates

Space Systems Academic Group, Department of Electrical & Computer Engineering, Naval Postgraduate

School, Monterey, CA 93943, USA

Received 15 May 2004; received in revised form 16 July 2004; accepted 19 July 2004

Abstract

In this paper, the design and optimization of advanced multijunction photovoltaic devices,

utilizing a newly introduced modeling technique (J. Sol. Energy Mater. Sol. Cells, submitted

for publication), is demonstrated. In our opinion, the introduction of this modeling technique

to the photovoltaic community will prove to be of great importance in aiding the design and

optimization of advanced solar cells. A model of an InGaP/GaAs/InGaNAs/Ge four-junction

solar cell is prepared and is fully simulated. The major stages of the process are explained and

the simulation results are compared to published theoretical and experimental data. An

example of cell parameters optimization is also presented. The flexibility of the proposed

methodology is demonstrated. The methodology and design considerations for this process are

explained.

r 2004 Elsevier B.V. All rights reserved.

Keywords: Multi-junction solar cells modeling; Virtual growth of crystals; Simulation; Optimization

see front matter r 2004 Elsevier B.V. All rights reserved.

.solmat.2004.07.051

nding author. Tel.: +1831 656 2252; fax: +1 831 656 2760.

dress: michael@nps.edu (S. Michael).

ARTICLE IN PRESS

S. Michael, A. Bates / Solar Energy Materials & Solar Cells 87 (2005) 785–794786

1. Introduction

In our research of the existing literature, we have found a number of significantpublications which fully describe various aspects of semiconductor devicecharacteristics and modeling. However, all of them were focused on very specificissues and lacked the breadth of a complete simulation tool. The ATLAS devicesimulator from Silvaco International has the capability to model a wide variety ofphysical device characteristics. These include DC, AC small-signal, and full time-dependency, drift-diffusion transport models, energy balance, lattice heating, gradedand abrupt heterojunctions, optoelectronic interactions with general ray tracing,amorphous and polycrystalline materials, stimulated emission and radiation,Fermi–Dirac and Boltzmann statistics, doping effects, trap dynamics, SRH,radiative, Auger, and surface recombination. The combination of these powerfulmodeling techniques into one software package gives ATLAS the unique capabilityto accurately model a wide range of solar cell operating characteristics. This abilityshows remarkable promise towards streamlining the complex task of advancedmultijunction solar cell design.

2. Modeling process

ATLAS predicts the electrical characteristics of physical structures by simulatingthe transport of carriers through a 2D grid. These simulations can be done muchcheaper and faster than physical experimentation and can provide solar cellinformation that is difficult or impossible to measure. The accuracy of thisphysically-based simulation tool depends greatly on the accuracy of the materialparameters used in constructing the solar cell model. Important parameters neededfor solar cell modeling in ATLAS include band gap energy, electron and hole statedensities, electron and hole mobilities, permittivity, electron affinity, radiativerecombination rate and optical parameters. One of the most critical parameters foradvanced solar cell modeling is the correct definition of the refractive index, n, andthe extinction coefficient, k, for a material. Many of the advanced ternary andquaternary materials have limited published optical parameters. Good approxima-tions of the n and k values may be obtained through interpolation between simplercompounds. Fig. 1 shows an example of interpolation between GaAs and InAs toarrive at optical parameters for InGaNAs.

Once a solar cell is simulated in ATLAS, it may be illuminated with a constantwavelength of light or a complex spectrum such as AM0 (Fig. 2). A wide variety ofoutputs are available to the solar cell designer. These include I2V characteristics,photogeneration rate, spectral response, potential build-up (Fig. 3) and electrostaticfield (Fig. 4). A good measure of the accuracy of the ATLAS model is seen in Fig. 5which shows the I2V characteristics of both an InGaP and GaAs cell under AM0illumination. These simulated results compare favorably with numerous publishedexperimental results.

ARTICLE IN PRESS

Fig. 1. Interpolations of n and k for InGaNAs (from Ref. [1]).

Fig. 2. Air-mass zero (AM0) solar spectrum (NREL).

S. Michael, A. Bates / Solar Energy Materials & Solar Cells 87 (2005) 785–794 787

3. The InGaP/GaAs/InGaNAs/Ge quad-junction cell

The design of a multijunction solar cell is complicated by both the desire to havemaximally efficient junction layers and the need to match the current produced ineach junction layer under optimal load conditions. ATLAS greatly aids in thisprocess by providing I2V characteristics and spectral responses for each junctionlayer in a candidate design. This allows the solar cell designer to adjust the

ARTICLE IN PRESS

Fig. 3. Potential build-up for InGaP/GaAs cell (after Ref. [2]).

S. Michael, A. Bates / Solar Energy Materials & Solar Cells 87 (2005) 785–794788

parameters of the junction layers to produce an optimal design. To produce aworking InGaP/GaAs/InGaNAs/Ge four-junction solar cell, several variations ofjunction layer thickness were tested to match output current between layers. Thethickness of the InGaP layer was varied to allow more or less light through to theGaAs cell below it. The GaAs and the InGaNAs layers could likewise vary their cellthicknesses to balance their currents. The Ge cell consisted of a substrate of constantlarge thickness. The final cell as tested (Fig. 6) produced 41.6mW/cm2 for anefficiency of 30.75%. Fig. 7 shows the combined I2V characteristic of the cell andbreaks out the individual junction layer contributions. Fig. 8 shows the spectralresponse for each junction layer. Fig. 9 shows the photogeneration occurring atvarious depths in the multijunction cell over a range of incident wavelengths.

Further work on improving the InGaP/GaAs/InGaNAs/Ge cell design isunderway. A method for the incorporation of a genetic search algorithm [3] todirect the ATLAS simulations have been met with early success. Fig. 10 shows an

ARTICLE IN PRESS

Fig. 4. Depletion region electrostatic field in GaAs cell (from Ref. [4]).

Fig. 5. I2V characteristics of individual GaAs and InGaP solar cells (from Ref. [4]).

S. Michael, A. Bates / Solar Energy Materials & Solar Cells 87 (2005) 785–794 789

example of construction of a chromosome from binary genes. Fig. 11 illustrates theshort-circuit current convergence over successive iterations of current matchingroutine. The improvement of output power from a single GaAs cell over successivegenetic search generations is shown in Fig. 12. In addition, an iterative current

ARTICLE IN PRESS

Fig. 6. InGaP/GaAs/InGaNAs/Ge cell as simulated (from Ref. [1]).

S. Michael, A. Bates / Solar Energy Materials & Solar Cells 87 (2005) 785–794790

matching routine has been developed which has the ability to search out the junctionlayer thickness combination with the highest power output for a multijunction cell(Table 1). This is an improvement over the previously obtained results reportedin [5].

4. Conclusion

The ability of the ATLAS device simulator [6], to accurately model solar cellcharacteristics has been shown. The detailed outputs available to the solar celldesigner allow for efficient and effective simulation and optimization of even themost advanced solar cell designs. Using these tools, an InGaP/GaAs/InGaNAs/Gecell was designed and simulated showing spectral and I2V characteristics in linewith expectations. Further work on improving this design is continuing and showspromise for even higher cell efficiencies.

ARTICLE IN PRESS

Fig. 7. I2V characteristics of InGaP/GaAs/InGaNAs/Ge cell (from Ref. [1]).

Fig. 8. Spectral response of InGaP/GaAs/InGaNAs/Ge solar cell (from Ref. [1]).

S. Michael, A. Bates / Solar Energy Materials & Solar Cells 87 (2005) 785–794 791

ARTICLE IN PRESS

Fig. 9. Photogeneration in InGaP/GaAs/InGaNAs/Ge cell (after Ref. [1]).

Fig. 10. Construction of a chromosome from binary genes.

S. Michael, A. Bates / Solar Energy Materials & Solar Cells 87 (2005) 785–794792

ARTICLE IN PRESS

Fig. 11. Short-circuit current convergence over successive iterations of current matching routine.

Fig. 12. Power output from GaAs cell over successive genetic search algorithm generations.

S. Michael, A. Bates / Solar Energy Materials & Solar Cells 87 (2005) 785–794 793

ARTICLE IN PRESS

Table 1

Maximum power comparison between genetic algorithm and junction layers proposed in Ref. [3]

Layer material (layer thickness)

InGaP (0.50 mm) GaAs (1.00mm) InGaNAs (1.54mm) Ge (304 mm)

Pmax3 (mW/cm2) 22.79 22.61 17.41 13.33

Pmax optimized (mW/cm2) 23.41 28.49 18.75 20.83

% Improvement 2.72 26.01 7.70 56.23

S. Michael, A. Bates / Solar Energy Materials & Solar Cells 87 (2005) 785–794794

References

[1] S. Michael, M Green, Innovative approach for the design and optimization for multijunction

photovoltaic devices, NCPV and Solar Program Review Meeting, 2003, pp. 737–740.

[2] S. Michael, A novel approach for the modeling of advanced photovoltaic devices using virtual wafer

fabrication tools, J. Sol. Energy Mater. Sol. Cells, submitted for publication.

[3] K.F. Man, K.S. Tang, S. Kwong, Genetic algorithms: concepts and applications, IEEE Trans. Ind.

Electron. 43 (1996) 519–534.

[4] S. Michael, P. Michalopoulos, Application of the SILVACO/ATLAS software package in modeling

and optimization of state-of-the-art photovoltaic devices, Proceedings of the 45 Midwest Symposium

on Circuits and Systems, Tulsa, OK, August 5–7, 2002.

[5] M. Green, The verification of Silvaco as a solar cell simulation tool and the design and optimization of

a four-junction solar cell, MSEE Thesis, Department of Electrical and Computer Engineering, Naval

Postgraduate School, Monterey, CA, 2003.

[6] ATLAS User’s Manual, vols. 1–2, Silvaco International, 2000.