Heuristic modelling of multijunction solar cells using a parallel genetic algorithm

Čičić, S

Abstract

In order to fabricate solar cells with the highest possible values of efficiencies, material type, layer thickness and doping have to be properly selected. It can be achieved if light absorption and carrier generation are maximized and losses minimized. The same parameters that increase carrier generation, can increase certain types of losses. Parameters which reduce one type of losses, tend to increase the other types. Structural complexity of these devices combined with already mentioned conflicting requirements create vast parameter space which is highly uneven and contains a huge number of local minima and maxima. It makes it difficult for the most of search methods to locate the global maximum. Therefore, heuristic optimization is crucial for solving problems as complex as this one.

To find the optimal combination of these parameters, genetic algorithm is used with drift-diffusion model and all material parameters calculated as a function of energy gap. This way we have very realistic material parameter set which, together with detailed losses modeling, provide reliable results. To test the model, findings were compared with the record setting devices. Results were in agreement, which makes the model trustworthy. Two types of devices were optimized: multi-junction solar cells (MJSC) and photon energy converters (PEC). In case of MJSCs, parameters which were optimized are thicknesses, impurity concentrations, energy gaps and optimal current. And in case of PECs, thicknesses, impurity concentrations and optimal current were optimized. The optimization was repeated with different types of losses accounted in order to see how each one of them affects the overall efficiency. From the results of optimization it was possible to see what are the main drawbacks in the device efficiency and how to overcome them.


Calculations were carried out with ASTM G173−03 Global tilted solar spectrum in case of MJSCs and laser with intensity of 5W/cm2 and wavelength of 855nm in case of PECs. The absorption was calculated from kppw code.

The maximum efficiencies achieved for the unconstrained device are 30.158%,41.479%, 45.669%, 50.775% and 53.653% for MJSC devices with one, two, three, four and five subcells, respectively, when all types of losses are taken into account. In case of the series constrained device, the results are 31.080%, 42.467%, 48.276%, 50.777%, 53.653%, 54.917% and 55.317% for devices with up to seven subcells, respectively, when all types of losses are taken into account, as well. The values for the PECs are 69.431%, 68.838%, 66.676% and 65.698% for one, five, ten and fifteen subcells, respectively. This time all losses are accounted as well. If the model is applied to the record setting devices, the results are 32.34% for 2JSC and 38.1%, while actual, measured, values are 31.6±1.5% and 37.9±1.2%, respectively, which is an outstanding match.

Detailed device parameters obtained through the optimization process are presented. Examination of those results leads to possible recipe how to fabricate the highest possible efficiency devices. It was concluded that the radiative recombination is the most dominant type of losses in III-V semiconductors and can be suppressed by increasing the material’s energy gap. Diffusion dark current can be suppressed by increasing the energy gap as well, while the doping levels shoud be increased. On the other hand, Auger recombination can be reduced by decreasing the doping, while the increase of energy gap reduces Auger much more than the other two. This leads to significant drop in efficiency when the algorythm tries to suppress the Auger, in comparison when only the two other types of losses are accounted. Nevertheless, the suppression of all losses leads to more efficient devices. This analysis can be a guide the future experiments and indicate how much more efficiency can be achieved with these devices, which materials to target and how to correctly balance between various contradicting requirements imposed by the nature of semiconductor materials.

Citation

Čičić, S. (in press). Heuristic modelling of multijunction solar cells using a parallel genetic algorithm. (Thesis). University of Salford

• Thesis Type: Thesis
• Acceptance Date: Mar 29, 2019
• Deposit Date: Apr 5, 2019
• Publicly Available Date: Apr 5, 2019