Interdigitated back‐contact double‐heterojunction GaInP/GaAs solar cells

Interdigitated back‐contact (IBC) silicon solar cells are coming of age, but the potential of IBC configurations for compound semiconductor solar cells is yet to be explored. We outline an approach to generalize the diffusion‐driven charge transport (DDCT) method, previously studied for IBC light‐emitting diodes, to develop DDCT solar cells, enabling an IBC double‐heterojunction structure. In particular, we simulate and compare the electrical performance of a GaInP/GaAs DDCT solar cell with an ideal one‐dimensional reference cell to establish how the lateral dimensions of the DDCT structures affect their operation. Also, the suitability of the DDCT solar cells for concentration photovoltaics is explored. The results show that the DDCT solar cells with a finger pitch of approximately 10μm can match and even outperform the ideal reference structure under the AM1.5G solar spectrum, due to reduced Shockley‐Read‐Hall recombination. At high solar concentrations, the performance of the smallest pitch DDCT structure is essentially identical with the reference structure up to 100 suns. This suggests that combining the benefits offered by the IBC design with compound semiconductors could allow the development of an entire family of more efficient solar cells.


| INTRODUCTION
The past few decades have seen a rapid development of solar cells with the most advanced state-of-the-art devices being based on interdigitated back-contact (IBC) silicon solar cells, 1,2 efficient singlejunction 3,4 and multi-junction 5,6 III-As compound semiconductor cells, and their combinations. 7 As compared to silicon, the compound semiconductor cells offer several benefits due to their direct and tunable band gap and composition. Consequently, they presently hold the efficiency record for single-junction solar cells, with the the demonstration of 29.1% efficient GaAs-based thin-film devices, 3,4 but even they still fall relatively far behind their theoretical efficiency limit of approximately 33.5%. 8 This has also lead to the recently increased interest in ultrathin III-As solar cells 9-11 that can enable higher fill factors and open-circuit voltages. Unlike the further developed IBC silicon cells, however, the efficiency of III-As solar cells is still affected by contact-shading, which can notably reduce the illumination of the cell due to the presence of the metallic front contact-grid as discussed in previous studies. [12][13][14] Despite significant efforts to further increase the efficiency of the III-As cells, 13,15,16 conventional IBC designs cannot be directly applied to the presently prevailing compound semiconductor solar cells due to the very different carrier diffusion length scales, fabrication methods and surface passivation needs of these materials. Considering the potential benefits, developing IBC designs for III-As solar cells can, nevertheless, bridge the gap between the experimentally demonstrated efficiency for GaAs-based solar cells and its theoretical limit.
The first steps towards IBC technology for compound semiconductor devices were only taken very recently through modelling efforts of emitter-less back-surface alternating contact (EBAC) III-As solar cells 17,18 and diffusion-driven charge transport (DDCT) light-emitting diodes (LEDs). 19,20 On the surface, these devices are similar in the sense that both structures have contacts only on the back and that charge carriers migrate to the active region (AR) through the same surface. In more fundamental terms, however, they have one crucial difference. In the EBAC structure, the current is 'forced' to pass through the AR that is still sandwiched between the p-n junction, whereas in the DDCT structure, the AR is parallel to the p-n junction and the current is governed by carrier diffusion even more strongly than in conventional p-n junctions. Therefore, a clear functional difference between these designs is that even if the AR is removed from the DDCT structure, the p-n junction still forms a working device. As such, in terms of current transport, the EBAC solar cell design is very similar to the conventional two-side contacted structures, and it also leaves a significant part of the AR exposed to surface recombination which can substantially reduce the device efficiency, as recently discussed in Myllynen et al 19 within the LED context. The DDCT approach does not suffer from such limitation, since both surface and interface recombination can be readily minimized. 19 Nevertheless, harnessing the DDCT devices for solving the practical problems in developing IBC technologies for compound semiconductor devices has not been studied.
In this paper, we investigate if the recently proposed DDCT-LED structure 19 can enable the development of ultra-thin IBC III-As solar cells. In the studied design, the AR is electrically located parallel to the p-n junctions forming an array of lateral heterojunctions (LHJs) that function as selective contacts for electrons and holes, collecting the photogenerated charge carriers. The DDCT solar cell structure can be fabricated with a straightforward lithography process combined with a p-doping process. [21][22][23][24] Therefore, realizing DDCT cells should be simpler and easier as compared to the previously studied EBAC designs 17,18 relying on a selective-area growth (SAG) process, which is generally demanding for III-arsenide materials and reported only in a few papers, for example, in other studies. [25][26][27] In addition, originating from the high-power large-area LED context, the DDCT approach provides interesting possibilities for concentration photovoltaics. Also, using DDCT structures as the bottom and/or top cell of a multijunction solar cell can prevent current matching problems, as recently suggested. 7,28,29 Furthermore, DDCT solar cells can enable applications in more advanced devices such as thermophotonic heat pumps. 30,31 Indeed, our results suggest that the DDCT structures are very well suited for solar cell operation and can even match the performance of an ideal one-dimensional reference device, under both the AM1.5G solar spectrum (1, 000W/m 2 ) and concentrated sunlight.  Table 1 shows the used ABC recombination coefficients and the surface and interface recombination velocities based on well established literature values of state-of-the-art devices. In addition, the electron and hole mobilities for the materials forming the devices are estimated with the empirical mobility model discussed in Sotoodeh et al. 45 The optical calculations assume, for simplicity, that all incident photons from the AM1.5G solar spectrum having energies greater than the band-gap energy of GaAs (1.42 eV) are uniformly absorbed in the i-GaAs AR. This is justified, since the devices include an ARC, highly reflective back-contacts and ultimately also textured surfaces for redistributing the light through scattering, trapping most of the generated photons within the device. 9-11 Furthermore, since this assumption is made for both DDCT solar cells and the reference structure, the relative performance of the devices and therefore the conclusions of this work would remain the same, even if a more detailed optical model would be used. Even more importantly, however, the aim of the present work is to focus on charge transport and to compare different device geometries to one another, providing a detailed view of the electrical behaviour.

| CELL DESIGN AND MODELLING
For DDCT solar cells, the simulations are performed for one period of the device, extending from the middle of n-contact to the middle of p-contact. Therefore, the simulations include effectively the whole device area assuming a periodically large structure, as justified by device symmetries. Furthermore, the p-doped regions of the LHJs are approximated using square elements in the calculations for simplicity.  Figure 2A whose shape are identical to that of the η coll curves in Figure 3.

| RESULTS AND DISCUSSION
In Table 2, we present the figures of merit for the studied solar cells. In addition to the V oc and J sc already discussed after Figures 2   and 3, Table 2 shows the efficiency η mpp , current density J mpp , voltage