Bypassing the single junction limit with advanced photovoltaic architectures

Abstract

In single-junction photovoltaic (PV) devices, the maximum achievable power conversion efficiency (PCE) is mainly limited by thermalization and transmission losses, because polychromatic solar irradiation cannot be matched to a single bandgap. Several concepts are being investigated to reduce these losses, such as the classical vertical multijunction cells, 'lateral' tandem cells, and multi-exciton generation in the form of photon up- and down-conversion. While in theory, efficiencies exceeding 90% are possible (Landsberg or thermodynamic limit), there are severe practical limitations in terms of processability, cost, and spectral sensitivity. Here, we present a simulation environment based on Bayesian Optimization that is able to predict and optimize the electrical performance of multi-junction architectures, both vertical and lateral, in combination with multi-exciton materials. With respect to vertical stacks, we show that by optimizing bandgap energies of multi-exciton generation (MEG) layers, double junctions can reach efficiencies beyond those of five-junction tandem devices (57%). Moreover, such combinations of MEG and double junction devices would be highly resilient against spectral changes of the incoming sunlight. We point out three main challenges for PV material science to realize such devices. With respect to lateral architectures, we show that MEG layers might allow reducing nonradiative voltage losses following the Energy Gap Law. Finally, we show that the simulation environment is able to use machine learned quantitative structure-property relationships obtained from high-throughput experiments to virtually optimise the active layer (such as, the film thickness and the donor-acceptor ratio) for a given architecture. The simulation environment thus represents an important building block towards a digital twin of PV materials.