Quantum Transport across Amorphous-Crystalline Interfaces in Tunnel Oxide Passivated Contact Solar Cells Direct versus Defect-Assisted Tunneling

Chinese Physics Letters PAPER Quantum Transport across Amorphous-Crystalline Interfaces in Tunnel Oxide Passivated Contact Solar Cells: Direct versus Defect-Assisted Tunneling To cite this article: Feng Li et al 2021 Chinese Phys. Lett. 38 036301 View the article online for updates and enhancements. CHIN.PHYS.LETT. Vol.38, No.3(2021)036301 Quantum Transport across Amorphous-Crystalline Interfaces in Tunnel Oxide Passivated Contact Solar Cells: Direct versus Defect-Assisted Tunneling Feng Li(a78a240)1;2 , Weiyuan Duan(a239a31a67)2, Manuel Pomaska2, Malte Köhler2, Kaining Ding(a1a239a129)2, Yong Pu(a110a199)1 , Urs Aeberhard2, and Uwe Rau2 1College of Science, Nanjing University of Posts and Telecommunications, Nanjing 210023, China 2IEK-5 Photovoltaik, Forschungszentrum Jülich, 52425 Jülich, Germany (Received 24 October 2020; accepted 9 December 2020; published online 2 March 2021) Tunnel oxide passivated contact solar cells have evolved into one of the most promising silicon solar cell concepts of the past decade, achieving a record efficiency of 25%. We study the transport mechanisms of realistic tunnel oxide structures, as encountered in tunnel oxide passivating contact (TOPCon) solar cells. Tunneling transport is affected by various factors, including oxide layer thickness, hydrogen passivation, and oxygen vacancies. When the thickness of the tunnel oxide layer increases, a faster decline of conductivity is obtained computationally than that observed experimentally. Direct tunneling seems not to explain the transport characteristics of tunnel oxide contacts. Indeed, it can be shown that recombination of multiple oxygen defects in a-SiOx can generate atomic silicon nanowires in the tunnel layer. Accordingly, new and energetically favorable transmission channels are generated, which dramatically increase the total current, and could provide an explanation for our experimental results. Ourworkprovesthathydrogenatedsiliconoxide(SiOx:H)facilitateshigh-qualitypassivation,andfeatures good electrical conductivity, making it a promising hydrogenation material for TOPCon solar cells. By carefully selecting the experimental conditions for tuning the SiOx:H layer, we anticipate the simultaneous achievement of high open-circuit voltage and low contact resistance. DOI: 10.1088/0256-307X/38/3/036301 Passivating and carrier-selective contacts count among the key elements to further increase the effi- ciency of silicon-based solar cells. One of the main representatives of device architectures implementing such contacts is the tunnel-oxide passivated contact (TOPCon) solar cell, consisting of an ultrathin amor- phous SiOx (a-SiOx) layer on an Si wafer, in combina- tion with a heavily doped silicon film.[1] TOPCon so- lar cells have undergone rapid development in the past decade. By applying TOPCons as a full-area rear con- tacts, a record efficiency of more than 25% could be achievedforanareaof 2 2cm2,[2] andmorethan21% for areas larger than 100cm2.[3;4] Careful engineering is key to the achievement of high efficiency: (1) high quality interface passivation of the tunnel oxide layer (long carrier lifetime), (2) efficient doped layers in c-Si (high Voc) and (3) efficient majority carrier transport (high fill factors). While several new materials for passivated con- tacts besides a-SiOx have been reported, e.g., a-SiC,[5] a-SiOx is the most widely used, as it exhibits extraor- dinary characteristics as an effective buffer layer, com- bining several beneficial properties: it is anticipated not only to effectively reduce recombination and lower production costs (withstanding industrial contact fir- ing) but also to reduce optical losses.[6] Due to the wide bandgap of a-SiOx, band-like electronic trans- port is suppressed, and tunneling becomes the main transport mechanism when the thickness of the ox- ide layer is at the nanometer level.[7] In this respect, the transport processes at a-SiOx passivated contacts differfromthoseattheclassicala-Si:Hpassivatedcon- tact in silicon heterojunction devices, where a signif- icant fraction of the current is assumed to be ther- mally activated.[8] In spite of the critical role of the oxide barrier, a microscopic understanding of the re- lation between local electronic structure (in terms of nanoscale thickness fluctuations, defect states, etc.) and the transmission characteristics of the oxide bar- rier is currently lacking. In order to narrow this gap in understanding, we explore the electronic properties of a-SiO2 and a- SiOx passivating contacts, as used in the TOPCon solar cell architecture, from a microscopic perspective. We first compute the atomic configuration and elec- tronic structure of c-Si, a-SiO2 and a-SiOx, respec- tively, based on first principles calculations. Next, by coupling the Hamiltonians of the different materials, a sandwichstructureof c-Si/a-SiOx=c–Siisconstructed, using the lead-conductor-lead (LCR) model[9;10] in an open boundary condition. For this contacted sand- wich structure, the tunneling transport properties are analyzed based on an integrated approach, combin- ing the density functional theory (DFT) with a non- equilibrium Green’s function picture of ballistic quan- tum transport. One significant finding is that the Supported by the National Natural Science Foundation of China (Grant Nos. 61704083 and 61874060), the Natural Science Foun- dation of Jiangsu Province (Grant No. BK20181388), and NUPTSF (Grant No. NY219030). Corresponding authors. Email: lifeng@njupt.edu.cn; puyong@njupt.edu.cn ©2021 Chinese Physical Society and IOP Publishing Ltd 036301-1 CHIN.PHYS.LETT. Vol.38, No.3(2021)036301 transport across the tunnel oxide layer depends not only on the passivation and the thickness of the bar- rier layer, but also on the presence and configuration of oxygen defects in a-SiOx. Inthecaseofmultiplede- fects, these were found to recombine, forming a type of atomic silicon nanowires in the a-SiOx layer, due to structural reorganization. The presence of such “monofilament pinholes” significantly enhances con- ductance as new transmission channels are opened. Methods. We computed the electronic trans- port property of c-Si/a-SiOx:H/c-Si sandwich struc- tures by using an integrated approach, combining the DFT with non-equilibrium Green’s function for- malism (NEGF). Similar approaches have been used to assess leakage currents in gate oxides.[8] In the first step, an ab initio molecular dynamic simula- tion (AIMD) was performed to build the structures of a-SiO2 and c-Si/a-SiOx:H/c-Si, using tools from the VASP package.[11;12] Next, the Hamiltonian ma- trix was obtained, based on static electronic computa- tion, usingthePWSCFcodeavailableintheQuantum ESPRESSO suite.[13] Here, 10 10 10, 4 4 4 and 2 2 1 Monkhorst-Pack k-point grids were consid- ered for the Brillouin zones of c-Si, a-SiO2, and c-Si/a- SiOx:H/c-Si, respectively. A cutoff energy of 450eV for the wavefunctions provided a converged electronic density of states. The convergence of energy and force were set at 10 6 eV and 0.001eV/Å, respectively. We then used the maximally localized Wannier functions (MLWFs), together with the Landauer formula[14] as implemented in the WanT code,[15] to compute the total conductance in the lead-conductor- lead (LCR) model. The computational details are de- scribed in the Supplementary Materials. Results and Discussion—c-Si/a-SiO2:H/c-Si. Constructing the geometric structure at the amorphous-crystalline interface represents a starting point from which to examine the impact of interfacial properties on device characteristics. The microscopic picture of the complex interface region is captured by means of an ab initio description, allowing for a com- prehensive assessment of the device-relevant states. In order to construct the relevant structural model, a-SiO2 isinsertedasatunneloxidelayerbetweencrys- talline silicon layers, resulting in a c-Si/a-SiO2:H/c-Si sandwich structure. We first consider the structural, electronic and transport properties of ideal bulk c-Si, in order to verify the correctness of the integrated simulation ap- proach. To achieve a linear scaling of required com- puting time, the electron Bloch wavefunctions ob- tained from the ab initio calculation are translated into MLWF forms by means of a unitary transforma- tion. The computed electronic band structure and transmission gap of c-Si are shown in Figs.S1 and S2 in the Supplementary Materials. We observe a clear indirect bandgap of 0.60eV, which is very close to the DFT result of 0.61eV. The band edges around the Fermi level of c-Si are also provided, via the anal- ysis of the layer-resolved density of states (LRDOS, Fig.S2). This result proves that the MLWFs method conservestheaccuracyofthefirst-principleselectronic structure calculations, confirming the reliability of the integrated method used in this work. We therefore used atomic orbitals as the initial Wannier functions for MLWFs in all of our computations. The efficiency of charge carrier extraction across tunnel oxide barriers depends on a range of configu- rational parameters, including (a) tunnel oxide layer thickness, (b) interfacial passivation, and (c) oxygen defects in a-SiOx:H (i.e., the stoichiometry of the ox- ide). In view of their potential impact on solar cell efficiency, we will discuss these factors below, both in- dividually and from a microscopic perspective. (b) (a) (c) 0.6 nm 0.9 nm 1.2 nm (d) L-lead R-leadConductorBuffer Buffer L-lead R-leadConductor H01-L H-L H-C H-R H01-RH-LC H-CR Fig. 1. (a) Schematic representation of the lead- conductor-lead (LCR) model. Fully optimized geometric structures of three cases, in which the thicknesses of a- SiO2 layers are (b) 0.6, (c) 0.9 and (d) 1.2nm. Here c-Si layers of 1.6nm thickness are inserted between the cen- tral oxide layer and the leads in order to release the local stress at the interfaces. Role of Tunnel Oxide Layer Thickness. In the TOPCon solar cell fabrication process, precise con- trol of the thickness of the a-SiO2 tunnel oxide layer is challenging. In this situation, it is to be expected that the fluctuation of a-SiO2 layer thickness will affect electronic transport across the amorphous-crystalline interfaces (Fig.S3). The effect of tunnel oxide layer thickness becomes even more critical when the thick- ness of the a-SiO2 layer is as small as only a few nanometers, since the geometric and electronic prop- erties of thinner buffer layers are more readily affected by the substrate. As shown in Fig.1, three different thicknesses of a-SiO2 layers are investigated in this work: 0.6, 0.9, and 1.2nm, respectively, which are 036301-2 CHIN.PHYS.LETT. Vol.38, No.3(2021)036301 close to the experimentally estimated ideal thickness of 1nm.[16] To fully release the local stress at the in- terfaces, two additional c-Si buffer layers with a thick- ness of 1.6nm are inserted between the central a-SiO2 layer and the c-Si leads. Since oxygen has a much higher electronegativity (3.4) than silicon (1.9), the original surface of Si (100) will be seriously modified after full optimization, leaving some interfacial silicon atoms with dangling bonds. To passivate these de- fects, H atoms are used to saturate all the dangling bonds. 4 0 5 10 15 20 25 30 35 40 Buffer c-Si Buffer c-Sia-SiO2 0.6 nm 1.2 nm 0.9 nm En er gy (e V) En er gy (e V) En er gy (e V) Distance (Å) 4.5 5 5.5 6 6.5 7 7.5 4 4.5 5 5.5 6 6.5 7 7.5 4 4.5 5 5.5 6 6.5 7 7.5 100 80 60 40 20 0 (a) (b) (c) Fig. 2. Layer-resolved density of states (LRDOS) of the c-Si/a-SiO2:H/c-Si sandwich structure, with a-SiO2 tun- nel oxide layer thicknesses of (a) 0.6nm, (b) 0.9nm and (c) 1.2nm. The continuous transmission channels across the tunnel oxide layer rapidly disappear when the width of the a-SiO2 layer increases from 0.6nm to 1.2nm. As transport requires the availability of current- carrying states at the relevant energies close to the band edges of the contact materials, we computed and analyzed the LRDOS for the c-Si/a-SiO2:H/c-Si structure, as shown in Fig.2. The slight asymmetry of band structure is due to the dissymmetry of inter- facial geometric structures. In order to compare the energy levels between the different cases, the valence and conduction band positions are evaluated based on the model-solid theory,[17] in which the average of the electrostatic potential is defined as the reference level. The computational details of the model-solid theory are given in Fig.S1. The positions determined in this way amount to 5.32eV (0.6nm), 5.00eV (0.9nm) and 4.82eV(1.2nm)fortheVBM,andto5.95eV(0.6nm), 5.68eV (0.9nm) and 5.46eV (1.2nm) for the CBM. In Fig.2, thesepositionsarehighlightedbyyellowdashed lines. The downshift in VBM and CBM as a function of the oxide layer thickness is attributed to the in- crease in the proportion of oxygen. Based on the LRDOS image, we found that some local density of states is formed at the interfaces of c-Si/a-SiO2. According to their characteristics, the associated states can be divided into two groups: one group consists of strongly localized states, as enclosed by the green circles. These are located at specific en- ergy levels, e.g., at 5.90, 5.35, and 5.13eV, respec- tively, and originate mainly from interfacial defects; the other group contributes a relatively weak density of states, as indicated by the green arrows. The cor- responding DOS is distributed over a wide range of energies, and originates in the exponentially decaying wave functions of the states on both sides of the barri- ers. As a common feature, both types of state extend into a larger fraction of the a-SiO2 tunnel oxide layer. Furthermore, a clear and deep band gap of 6.5eV is found, located in the middle area of the c-Si/a- SiO2:H/c-Sistructure, whichisindicativeofana-SiO2 layer. This result is close to the 7.4–8.1eV experimen- tal band gap value for bulk a-SiO2.[18] In addition, the band offset for the valence bands at the c-Si/a- SiO2:H interface is estimated to be 2.2eV, which is much larger than that of 0.25–0.45eV at the c-Si/a- Si:H interface; meanwhile, the band offset for the con- duction bands at the c-Si/a-SiO2:H interface is 3.7eV, which again is much larger than that of 0.15eV at the c-Si/a-Si:H interface. Although the energy bar- riers at the interfaces are very high, the oxide layer does not prevent the majority of charge carrier trans- port across the barrier, as can be inferred from the fact that TOPCon solar cells are capable of exceeding an 82% of fill factor.[19] As such, the observed con- ductivity is attributed to tunneling transport through the ultra-thin oxide layer. To begin the process of shedding light on the dependence of transport charac- teristics on nanostructure configuration parameters, we first investigated the relationship between quan- tumconductanceandthethicknessofthetunneloxide layer. Figure 3 shows the quantum conductance results, computed for 0.6, 0.9, and 1.2nm thick tunnel ox- ide layers, displa