Strategies for realizing high-efficiency silicon heterojunction solar cells

,


Introduction
The photovoltaic industry is a technologically diverse market despite that different types of solar cells share the same basic working principle, i.e., the photovoltaic (PV) effect [1].Nowadays, the commercial PV market is mainly shared by wafer-based crystalline silicon (c-Si) technologies and thin-film technologies.Thanks to the abundance of Si, processing maturity, high efficiency and long service time (over 25 years), c-Si technologies own around 95% of the global annual PV production in 2021 [2].It is well-accepted that the continuously improved efficiency of solar cells (modules) and reduced manufacturing costs are strongly backing the success of the PV industry.Among different factors that promote the advancement of the PV industry, uninterrupted technological innovations are of great importance.

Evolution of c-Si PV technologies towards carrier-selective passivating contacts
Starting with the aluminium-doped back surface field (Al-BSF) solar cells, they feature simple architecture (see Fig. 1(a)) and were the main working horse of the PV industry in the past decades till 2013.However, mainly due to the full-area direct contact of metal and Si at the rear side, Al-BSF solar cells exhibit high recombination losses of photogenerated electrons and holes, thus limiting the open-circuit voltage (V OC ) to below 650 mV.In addition to the high parasitic absorption and non-optimum internal reflection at the rear side, the best Al-BSF solar cell has an efficiency of 20.29% [3].Commercially, the Al-BSF cells exhibit around 20% efficiencies [4].The recombination losses, especially at the rear side of Al-BSF cells, can be reduced by introducing a dielectric passivating layer between the Si and rear metal contact.This type of cell is referred to as the passivated emitter and rear cells (PERC) as first reported in 1989 [5] and in 2022 it possesses the largest portion of the market [6].As seen from a typical commercial PERC solar cell sketched in Fig. 1(b), the rear side of typical PERC cell features point contacts together with heavily doped p + regions, while the non-contacted surface is covered by the layer stack such as AlO x /SiN y for suppressing the surface recombination and enhancing the internal reflectance [7,8].PERC-based architectures (PER'X') such as passivated emitter rear locally diffused (PERL) (Fig. 1(c)) [9] and passivated emitter rear totally diffused (PERT) (Fig. 1(d)) [10] solar cells, featuring either locally-or fully-diffused boron-doped (B-doped) regions, respectively, are also industrial appealing [6].Overall for the PER'X' family, a long-held record efficiency of 25% was achieved with a high V OC of 706 mV back in 1999 [10][11][12].In mass production till 2022, the PERC cells have efficiencies of about 24% [13].
For both Al-BSF and PER'X' solar cells, although diffused doping of wafer surfaces can partially reduce the recombinations by minimizing the concentration of minority carriers at the metal/Si interface, it also introduces significant Auger recombination, free carrier absorption (FCA), and bandgap narrowing that limit their efficiencies [14][15][16][17].To alleviate, basically, the direct contact between the metal and silicon wafer, solar cells with carrier-selective passivating contacts (CSPCs) are developed.As the name suggests, stacks of thin-film layers are deposited on both sides of the wafer to passivate the surface defects of the wafer and selectively collect the photogenerated charge carriers.The prominent examples are low-thermal budget silicon heterojunction (SHJ) solar cells [18] and high-thermal budget tunnel-oxide passivating contacts (TOPCon) [19] or doped polysilicon (poly-Si) on oxide junction (POLO) [20] solar cells (see Fig. 1(e)-(g)).For SHJ and TOPCon solar cells, optimized thin (<10 nm) hydrogenated intrinsic amorphous silicon ((i) a-Si:H) layers and ultra-thin (<2 nm) silicon oxide (SiO x ) layers are utilized, respectively, to provide excellent chemical passivation of c-Si surfaces.Then, typically, doped a-Si:H or poly-Si layers, or their alloys with oxygen or carbon [21][22][23][24][25][26][27][28][29][30][31][32][33], are implemented to achieve carrier selectivity, respectively.SHJ solar cells have reached a record efficiency of 26.81% [34] with a high V OC of 751.4 mV in a front/back-contacted (FBC) configuration, and 26.7% in an interdigitated back-contacted (IBC) architecture [35].Till the end of 2022, the best TOPCon solar cell efficiency has reached 26.4% [36] and POLO-IBC demonstrated an efficiency of 26.1% [37].Besides such record devices, in production lines till 2022, solar cells with CSPCs deliver efficiencies of around 25% [38,39].
These concepts of CSPCs entail the use of either doped a-Si:H or poly-Si layers which exhibit high parasitic absorption when placed on the illumination side of the solar cells.Alternatively, dopant-free transition metal oxides (TMOs), which are more transparent than the abovementioned silicon-based layers, can mitigate this optical loss while providing sufficient carrier selectivity.Materials with high work function (WF), such as molybdenum oxide (MoO x ) [40][41][42][43], tungsten oxide (WO x ) [44,45] and vanadium oxide (VO x ) [46], can be used as hole transport layers (HTLs).Materials with low WF, such as lithium fluoride (LiF x ) [47], magnesium oxide (MgO x ) [48], magnesium fluoride (MgF 2 ) [49], titanium oxide (TiO x ) [50,51], or strontium fluoride (SrF 2 ) [52] can be used as electron transport layers (ETLs) [47,53].In particular, in the Photovoltaic Materials and Devices (PVMD) group of Delft University of Technology, we recently demonstrated the promising use of MoO x as the front HTL in SHJ solar cells with a certified record efficiency of 23.83% [54].As sketched in Fig. 1(h) [55], researchers presented a 21.4%-efficient solar cell with dopant-free contacts on both sides, where MoO x /ITO/Ag and ZnO/LiF x /Al contact stacks are used for hole and electron collections, respectively.Besides, both unencapsulated SHJ solar cells and encapsulated modules endowed with MoO x were reported to retain over 95% of their initial efficiencies after standard damp heat stability tests [56][57][58], demonstrating the promising viability of this technology.Nevertheless, additional research is required for further progressing TMO-based SHJ solar cells.
Overall, SHJ solar cell that holds the highest single-junction c-Si efficiency is one of the most promising c-Si technologies nowadays and it is predicted to own a 20% market share by 2032 [59].Its application, together with high bandgap perovskite solar cells to reduce losses induced by spectral mismatch, is critical for reducing the levelized cost of electricity (LCOE) of PV power plants.Till May of 2023, two-terminal (2T), three-terminal (3T) and four-terminal (4T) perovskite/SHJ tandem solar cells have achieved remarkable efficiencies of 33.7% (area: 1.0035 cm 2 ), 29.56% (area: 24.5 cm 2 ), and 30.79% (area: 64 cm 2 ), respectively [60][61][62].Challenges remain in scaling up and mitigating the instability of the perovskite subcells.Altogether, thanks to the substantial research progress made with perovskite/c-Si tandem solar cells [63][64][65][66], the momentum of continuous solar cell efficiency increment is secured.

A brief recap of the efficiency progression of SHJ solar cells
The SHJ with (i)a-Si:H layers, also initially known as 'Heterojunction with Intrinsic Thin-layer' (HIT) solar cell was first introduced by Panasonic (Sanyo) with an efficiency of 18.1%, significantly marking better V OC and FF values at cell level than those achieved in similar c-Si solar cell architectures without the (i)a-Si:H passivating layer [18].Apart from the continuous developments from Panasonic (Sanyo) that resulted in an efficiency of 23.7% with a 98-μm-thin wafer (area: 100.7 cm 2 ) in 2011 [71][72][73][74][75][76], the high-efficiency SHJ technology also attracted numerous research interests from universities, research institutes and companies worldwide [53,[77][78][79], especially around the expiration period of 'core' SHJ patents [80].Two years later, in 2013, Panasonic (Sanyo) improved the efficiency to 24.7% by reducing the interface recombination of (i)a-Si:H/c-Si together with minimizing the resistive losses in the solar cells [81].In 2014, they realized another jump in Fig. 1.Schematic representations of the evolution of c-Si PV technologies towards carrier-selective passivation contacts (CSPCs): a) Al-doped Back Surface Field (Al-BSF) solar cells [53]; b) Passivated Emitter and Rear Cells (PERC) [67]; c) Passivated Emitter and Rear Locally diffused cells (PERL) [68]; d) Passivated Emitter and Rear Totally diffused cells (PERT) [69]; e) Solar cell with a tunnel oxide passivating contact (TOPCon) on the rear side [19]; f) Solar cell with TOPCon on both sides [70]; g) Silicon heterojunction (SHJ) solar cells [69]; h) Dopant-free SHJ solar cells [55].The presented structures are typical device designs that are adapted from their corresponding references.
Y. Zhao et al. efficiency to 25.6% by implementing an IBC configuration [82].Nevertheless, soon after, FBC-SHJ solar cell fabricated by Kaneka reached an efficiency of 25.1%, which also focused on recombination mitigation at the (i)a-Si:H/c-Si interface [83].In 2017, Kaneka achieved an efficiency of 26.7% for an IBC-SHJ solar cell [35], which was a long-held record efficiency for single-junction c-Si solar cells until November 2022.At that time, LONGi announced a record-breaking efficiency of 26.81% for an FBC-SHJ solar cell (area: 274.4 cm 2 ), which features a J SC of 41.5 mA/cm 2 , a V OC of 751.4 mV, a FF of 86.07%[84].The certified record efficiency progression of low-thermal budget SHJ solar cells since the first 'HIT' solar cell is illustrated in Fig. 2.
Such optimized opto-electrical performance of the SHJ solar cell must require a thorough theoretical understanding of device designs, experimental optimizations and careful processing.Tremendous efforts are being put internationally into the technology development and multi-gigawatt industrialization of SHJ solar cells and modules [53,59,77,78,[85][86][87][88][89][90].However, the increasing demand of raw materials such as silver and indium, which are critical to this PV technology, has triggered comprehensive studies on the growth perspectives of this PV technology, from both the industrialization and sustainability standpoints [59,90,91].
In this contribution, we present experimental efficiency improvements for FBC-SHJ solar cells from around 18% to above 24% at the laboratory scale by following the guidelines provided by our rigorous device, material and optical simulations.We here provide our experience as (i) a roadmap showing the progressively enhanced device FF guided by simulations, (ii) the optimized high-efficiency monofacial and bifacial solar cells featuring nc-Si:H-based and/or MoO x -based contact stack, (iii) SHJ structures with less indium-based transparent conductive oxide (TCO) and silver consumption.Finally, based on numerical simulations, we present different innovative SHJ-based solar cell designs addressing advantages and drawbacks compared with state-of-art SHJ technology.
For cells with Molybdenum oxide (MoO x ), various plasma treatments were additionally applied after the HPTs, and MoO x layers were thermally evaporated (Provac PRO500S) using a stoichiometric MoO powder source.Afterwards, RF magnetron sputtering (Polyteknic AS) was used for depositing TCO layers.For wafers that feature TCO layers deposited at room temperature, they were annealed in air at 180 • C for min.Eventually, solar cells were finished with different metallization approaches, namely, Ag screen-printing, Ag evaporation and/or Cuelectroplating.Solar cells with screen-printed Ag contacts went through a curing step in the air at 170 • C for 40 min.Alternatively, monofacial cells were electroplated at the front side with Cu grids and thermally evaporated at the rear side with 500-nm-thick Ag, whereas bifacial solar cells were electroplated with Cu grids on both sides.To further improve the anti-reflection effect of solar cells, around 100-nmthick SiO x or MgF 2 layer was optionally deposited via e-beam evaporation on the front side of the devices for monofacial devices and both sides for bifacial devices.A geometrical factor of 1.7 was used for calculating the deposition durations for the textured surface based on the depositions on flat substrates.Solar cells reported in this work have a designated area of 4 cm 2 .

Thin-film and device characterizations
Spectroscopic ellipsometry (SE) (M-2000DI system, J.A. Woollam Co., Inc.) was utilized for extracting information such as the thickness, optical constants and optical bandgap of thin film layers deposited on either flat glass (Corning Eagle XG) or wafer substrates.Temperaturedependent dark conductivity (σ d ) measurement was conducted to extract the E a and σ d of doped Si thin-film layers.For TCO films, the Hall effect measurement system HMS-5000 from ECOPIA CORP was used to extract the conductivity type, resistivity, carrier mobility (μ e ) and carrier density (N TCO ) of TCO films.Spectrophotometer setups (PerkinElmer Lambda 950 and 1050 systems) were used to obtain the wavelengthdependent transmittance (T) and reflectance (R) of the testing layer, thus obtaining the absorptance (A) of the testing layer.
The photoconductance lifetime tester (Sinton WCT-120) was used for measuring the passivation quality during the development of contact stacks for SHJ solar cells and monitoring the passivation quality along the fabrication of solar cells.The effective minority carrier lifetime (τ eff ) values reported in this work were extracted at the minority carrier Fig. 2. The certified record efficiency progression of low-thermal budget SHJ solar cells since the first 'HIT' solar cell was reported by Panasonic (Sanyo) [18,34,35,60,[71][72][73]76,[81][82][83][84][92][93][94][95].Note, the 'HIT' solar cell reported in 1992 was not certified.The star symbol represents the 26.81%FBC-SHJ solar cell.Both LONGi and Huasun achieved a certified efficiency of 25.26% in June and July of 2021, respectively, while LONGi's cell having an area of 244.53 cm 2 and Huasun's cell having an area of 274.5 cm 2 , respectively.
Y. Zhao et al. density equals to 10 15 cm − 3 .The dark current-voltage setup (Cascade Summit probe station CAS33 Microtech) was utilized for characterizing the contact resistivity (ρ c ) of the contact stacks implemented in solar cells.The solar simulator (AAA class Wacom WXS-90S-L2) was used to characterize the illuminated current-voltage performance of fabricated solar cells.Sinton Suns-V OC -150 Illumination-Voltage Tester was employed to assist in the extraction of the series resistance of the device (R s ).An in-house-built external quantum efficiency (EQE) setup with a Fraunhofer ISE-calibrated photodiode was used for evaluating the spectral response of fabricated solar cells.The independently certified I-V parameters were obtained from the CalTeC of the Institute for Solar Energy Research Hamelin (ISFH) or Fraunhofer ISE CalLab PV Cells.

Numerical simulations
Technology computer-aided design (TCAD) Sentaurus from Synopsys Inc. was utilized for device simulations [98].The optical simulation model Genpro4 was used for optimizing the device design from the light management perspective [99].First-principle density-functional theory (DFT) calculations were performed for investigating MoO x and TCO films using Vienna Ab initio Simulation Package (VASP) with projector augmented wave (PAW) method [100,101].

Selective transport of charge carriers: the transport mechanism
In SHJ solar cells, the collection of charge carriers entails their transport across heterointerfaces or energy barriers.The effectiveness of this process poses strong indications to the eventual performance of solar cells.Therefore, a comprehensive understanding of the underlying physical mechanisms behind the selective transport of charge carriers is crucial for guiding the experimental development of high-efficiency SHJ solar cells.Our TCAD simulations [102,103] suggest the indispensable role of tuning the electrical properties of doped layers and TCO layers for achieving efficient selective transport of charge carriers from c-Si bulk to the TCO.Fig. 3 illustrates a schematic energy band diagram of a typical SHJ solar cell and the corresponding transport mechanisms of charge carriers.
As depicted in Fig. 3, the band bending formed inside c-Si near the c-Si/(i)a-Si:H interface is determined by the properties of the (n)-type contact stack (c-Si/i/n/TCO/metal) or (p)-type contact stack (c-Si/i/p/ TCO/metal), which is affected by the relative Fermi level position in each compositional material.Besides, due to different bandgaps of c-Si (~1.12 eV) and the (i)a-Si:H layer (~1.7 eV), band offsets (ΔE C for conduction band, ΔE V for valence band) are formed at the c-Si/(i)a-Si:H interface.Typically, ΔE C (~0.22-0.31eV) is smaller than the ΔE V (~0.39-0.61eV), and the latter is more sensitive to increase with a higher hydrogen content of the (i)a-Si:H film [77,104].Thanks to the band bending and the band offsets, there are highly asymmetric concentrations of electrons and holes within c-Si near each c-Si/(i)a-Si:H interface.The large difference in carrier population at c-Si interface suggests a high-quality selective contact [102,105,106].Subsequently, the accumulated electrons/holes near the c-Si/(i)a-Si:H interface can cross the potential barrier formed by the (i)a-Si:H/(n)-layer or (i)a-Si: H/(p)-layer via thermionic emission (TE) or with the support of tunneling processes such as direct tunneling (DT) and trap-assisted-tunneling (TAT).The significance of the specific mechanisms varies for the transport of electrons and holes [103,107].
Further, the negative charge collection is based on the transport of electrons towards the conduction band energy.The band bending and the band offset between adjacent materials are mainly defined by the work functions (WF) and electronic properties of the (n)-layer and TCO.Electrons cross the potential barriers formed at hetero-interfaces, via TE, DE and TAT processes.As suggested by our simulation study, a (n)-layer with a low activation energy (E a ) can enhance the band bending near the c-Si/(i)a-Si:H interface, and enable a reduced transport barrier from the (n)-layer to the TCO [102].Besides, a low WF TCO is preferred due to the favourable reduced WF mismatch between the TCO and the (n)-layer [102,103,108,109].
The collection of positive charge is based on holes transport up to the (p)-layer and electrons in the adjacent TCO layer.Note that at (p)-layer/ TCO interface charge exchange occurs via band-to-band (B2BT) or TAT tunneling processes [103].The energy alignment between the valence band of the (p)-layer and the conduction band of TCO is essential to discriminate if B2BT or TAT mechanism dominates the charge transfer.In the presence of the aforementioned energy alignment, the transport is based on B2BT in direct energy transition between electrons from the TCO conduction band jumping into (p)-layer valence band energy state, and vice versa.In the absence of such energy alignment, the charge exchange is based on TAT mechanisms as indirect energy transitions or recombination of electrons from TCO with holes from c-Si at defect Fig. 3.The schematic energy band diagram of a typical SHJ solar cell with an n-type c-Si wafer under dark and thermal equilibrium (the solid black line is the bottom of the conduction band; the solid pink line is the top of the valence band).The ΔE C and ΔE V represent the conduction and valence band offsets, respectively.The colored arrows in green and red are for the transport of holes and electrons, respectively.The corresponding transport mechanisms of charger carriers are thermionic emission (TE), direct tunnneling (DT), trap-assisted-tunneling (TAT) and bandto-band tunnelling (B2BT).The dashed lines indicate energy levels of different defect states within the bandgap.Those dashed lines also illustrate that the energy band bending can affect the energy levels of defect states.Figure adapted from Ref. [103].
Y. Zhao et al. states inside the (p)-layer occur near the (p)-layer/TCO interface.In general, B2BT is preferred because it implies only direct energy transitions [103,110].The proper energy alignment can be achieved by adjusting the E a of (p)-layer and/or the WF of TCO.In particular, (p)-layer with a low E a enables more flexibility in TCO electronic properties.Despite a low E a required, the band bending near the c-Si/(i) a-Si:H interface can be also enhanced with a (p)-layer that exhibits a high E g [102].
The transport of charge across heterointerfaces is typically evaluated by the contact resistivity (ρ c ) [103].It has been proven experimentally that the minimization of ρ c of SHJ contact stacks is essential for achieving high-efficiency SHJ solar cells [47,111,112].To understand the physical material parameters affecting ρ c , we constructed a detailed TCAD simulation framework replicating the transfer length method (TLM) to assess the ρ c of contact stacks in interest [113,114].We thus correlate ρ c to the V OC and FF of solar cells [103].As discussed before, the transport of charge carriers is ultimately driven by material properties that define the Fermi-level energy of layers forming the contact stacks.Therefore, we present in Fig. 4 (for (p)-type contact stack) the guidelines for improving the performance of SHJ solar cells with two practically accessible material properties, namely, E a of doped layers and carrier concentration (N TCO ) of TCO.As seen in Fig. 4, different combinations of these two parameters lead to various ρ c values, which provide insight into distinct dominating transport mechanisms of charge carriers.More details on correlating various transport mechanisms (e.g., B2BT, TAT) and various combinations of E a and N TCO for both (n)-and (p)-type contact stacks can be found in our previous work [103].It is worth noting our simulations were also validated with experimentally fabricated contact stacks that feature various combinations of E a and N TCO [115].
Overall, as known from our simulations, the V OC of solar cells is solely determined by the quality of the band bending inside the c-Si near its interface with (i)a-Si:H, whereas the FF is additionally influenced by the quality of the charge carrier transport at the doped-layer/TCO interface.Moreover, reducing the E a while increasing the N TCO leads to a reduction of ρ c of contact stacks.Last but not least, the optimization of (p)-type contact stack requires greater effort as it is more sensitive to layer properties as compared to (n)-type contact stack.

High-quality (i)a-Si:H passivating layer
To realize the high V OC and FF of SHJ solar cells, surface recombination must be suppressed by depositing a few nanometers thick optimized hydrogenated intrinsic amorphous silicon ((i)a-Si:H) passivating layers on both sides of the c-Si wafer.Typically, the development of (i)a-Si:H layer is mainly based on assessing the τ eff and implied V OC of test samples.To enhance the passivation quality of c-Si interfaces by using at thin (i)a-Si:H layer, various processing techniques such as tuning the PECVD deposition parameters (e.g., pressure, power, hydrogen dilution, substrate temperature) of (i)a-Si:H monolayer [116][117][118][119][120], stacking two or multiple intrinsic layers [93,[121][122][123][124][125][126][127], applying hydrogen plasma treatment (HPT) [128][129][130][131][132][133][134] and performing post-annealing were explored [135][136][137][138].The purpose of this research is twofold: i) to prevent possible detrimental epitaxial growth on the c-Si interface and ii) to ensure passivation support provided by hydrogen atoms inside the layer [136].We present in Fig. 5(a) the τ eff of symmetrical samples incorporating the bilayer and HPTs passivation approaches.Note that in Fig.
(a), 9-nm-thick (i)a-Si:H monolayer (or i-2) was deposited with a highly hydrogen-diluted plasma.Our bilayer approach combines firstly a 1-nm-thick (i)a-Si:H deposited without additional hydrogen dilution (or i-1) and then stacking on top 7 or 8-nm-thick i-2 layer for subsequential deposition of (n)-layer or (p)-layer stack, respectively [118].The HPTs that consist of a hydrogen plasma treatment and a VHF (i)nc-Si:H treatment were applied after the deposition of the (i)a-Si:H bilayer [131].
As depicted in Fig. 5(a), the (i)a-Si:H monolayer delivers a τ eff of 4.3 ms, which can be slightly improved to 4.8 ms when the bilayer is applied.While a notable enhancement of τ eff to 11.2 ms was observed when the bilayer is treated with HPTs.As suggested by our Fouriertransform infrared (FTIR) spectroscopy measurements that characterize the evolution of Si-H bonding configurations and microstructural properties of (i)a-Si:H layers, the additional HPTs led to a less-dense and more-hydrogenated (i)a-Si:H layer due to the increased formation of large volume deficiencies in the film [118].In other words, the resulting film containing higher concentrations of polyhydrides can be advantageous for contributing to the passivation of c-Si surface dangling bonds by serving as hydrogen reservoirs during processes with elevated temperatures [139].Subsequential deposition of (n)-layer stack on the HPTs treated bilayer showed further boosted τ eff to 22.3 ms thanks to improved field-effect passivation and enhanced chemical passivation.Particularly, the latter can be attributed to the beneficial hydrogen incorporation during the deposition of (n)nc-Si:H, which was deposited with highly hydrogen-diluted plasma conditions [140][141][142].However, only 4.6 ms was obtained after deposition of the (p)-layer stack.This could be attributed to defect formation in (i)a-Si:H layer as a result of Fermi level shift due to the overlaying (p)-layer stack [143][144][145][146], excessive annealing of the initially deposited i/p stack during the deposition of the second i/p stack [143,144], and/or indicates the necessity of further optimization of the deposition conditions of the (p)-layer stack.It is worth noting that a more careful choice of i-1 and i-2 based on their microstructure to form the bilayer, and even with an additional (i)a-SiO x :H or wet-chemical oxidic buffer layer before the deposition of the bilayer, are expected to further benefit the passivation quality [34,93,127].
Although superior passivation quality is essential for realizing highefficiency SHJ solar cells, the bulk quality of (i)a-Si:H layer is also crucial in influencing the charge carrier collection from the c-Si bulk to the TCO, thus affecting the FF of solar cells [127,147,148].We present in Fig. (b)-(d) the effects of (i)a-Si:H deposition temperature underneath both doped layers on the passivation quality of solar cell precursors before TCO sputtering, and the illuminated J-V performance, especially, the V OC and FF of completed solar cells [118].It should be noted that the (i) a-Si:H in Fig. 5(b)-(d) refers to 'bilayer + HPTs' as illustrated in Fig. 5 (a).As seen in Fig. 5(b), cell precursors (before TCO sputtering) that feature (i)a-Si:H layers deposited at the temperature of 140 • C deliver the highest average τ eff , and consequently, the highest average V OC in completed solar cells (see Fig. 5(c)).However, those cells exhibit deteriorated FFs as compared to higher temperature counterparts (see Fig. 5  (d)).This observation is comprehended by the FTIR-characterized underdense and hydrogen-rich film obtained with a lower deposition temperature [118].On the one hand, the (i)a-Si:H deposited at a low temperature (140 • C) can relatively enhance hydrogen accumulation [139], promote easier H-diffusion [149,150], prevent possible epitaxial growth caused by i-2 [121] and mitigate potential defect formation induced by HPTs [121,151], thus benefitting the V OC of the solar cells.On the other hand, the underdense or porous (i)a-Si:H bulk can increase series resistance and contact resistivity in the device [121,148], therefore lowering the FFs of solar cells.
In summary, we have optimized the (i)a-Si:H passivating layers based on a multi-layer approach together with the additional HPTs.Our research identifies two crucial requirements for optimizing (i)a-Si:H layers in high-efficiency SHJ solar cells: (i) achieving excellent surface passivation to minimize losses induced by surface recombinations, and in the meanwhile (ii) ensuring high bulk quality for uninterrupted charge carrier collection.

Hydrogenated nanocrystalline silicon (nc-Si:H)-based doped layer (stack) for SHJ solar cells
The requirement of a low E a for doped layers points to implementing materials such as hydrogenated nanocrystalline silicon-based (e.g., nc-Si:H and nc-SiO x :H) layers (E a lower than 50 meV) other than typically used a-Si:H (E a of around 200-300 meV) [34,131,140,142,[152][153][154][155][156][157][158][159][160][161][162][163][164][165][166].The nc-Si:H-based materials contain tiny crystalline grains embedded in the amorphous matrix, and might possess a certain volume fraction of voids [167][168][169].Thanks to the more efficient doping in their embedded nanocrystals [170], nc-Si:H-based layers can achieve lower E a values as compared to that of a-Si:H.Notwithstanding the advantageous electrical properties of nc-Si:H-based layers, their implementation into SHJ solar cells is challenging.This is due to the growth of nc-Si:H, especially, the evolution of their compositions (nanocrystals and amorphous phase) depends on the nature of the substrate and the thickness of the layer [167,171,172].Those substrate-and thickness-dependent growth characteristics are critically relevant for SHJ solar cells, which feature (i)a-Si:H passivating layers and might require limited thicknesses of nc-Si:H-based layers for the optimum device performance.Therefore, efforts have been devoted to facilitating fast nucleation of nanocrystals (minimizing the amorphous incubation phase), which enables efficient band bending near the c-Si surface [34,111,131,140,141,163,[173][174][175][176].Accordingly, we developed doped nc-Si:H-based layers with low E a values (below 50 meV) and (p)nc-SiO x : H layers with high E g values (>2.3 eV in E 04 ) by tuning the PECVD deposition parameters, and implemented them into SHJ solar cells [131].
For electron collection using a single (n)nc-SiO x :H layer, we experimentally assessed a minimum thickness of approximately 8 nm is essential to maintain the passivation quality and be effective for electron transport, while the minimum thickness required for (n)nc-Si:H and (n) a-Si:H can be as thin as 3-4 nm [142,156,177].This difference is mostly related to their dissimilar capability to incorporate active doping in such thin thicknesses (<10 nm), where (n)a-Si:H results to be the most conductive, then followed by (n)nc-Si:H and lastly (n)nc-SiO x :H.The less conductive (n)nc-Si:H-based layers can be linked to their low doping gas (phosphine) flow and highly hydrogen-diluted deposition conditions.The former can lead to lower phosphorous content in the film, while the latter can result in hydrogen-induced dopant inactivation, and alter the amorphous incubation phase of nc-Si:H-based layers to have a microstructure that differs from standard (n)a-Si:H [178][179][180].Nevertheless, for a 3-nm-thick (n)nc-Si:H-based layer that features an E a value of at least around 600 meV, and yet, it delivers FF above 78.6% in solar cells when combined with our aforementioned optimized HPTs (see Fig. 6 (a)).The optimized HPTs lead to a H-rich and underdense (i)a-Si:H, which provide hydrogen atoms to significantly suppress the surface recombination and may enhance the nucleation of those nc-Si:H-based layers, thus contributing to more efficient selective transport of electrons as can be seen in Fig. 6(a) [103,[181][182][183].Moreover, the nucleation could be also promoted thanks to a larger H coverage on the growing surface that increases the surface mobility of SiH x radicals [184].It is worth noting that despite the limited active doping in such a thin (n)nc-Si:H layer, electrons are capable of tunneling directly from the c-Si to the TCO, making the electron transport to be less dependent on the condition of (n)-layer/TCO interface after processing.As a result, as long as the passivation quality is well-maintained, efficient electron collection can still be achieved.A thicker (n)nc-Si:H-based layer with well-developed nanocrystals can also result in similar electron transport [103]; however, the device performance can be then limited by its higher parasitic absorption.The above results and discussions suggest the use of a combination of thin (n)-layers that ensure a low E a with a limited thickness.Therefore, we propose the use of 3 nm (n)nc-Si:H and 2 nm (n)a-Si:H as (n)-layer stack to enhance the FF of solar cells (see Fig. 6(a)).The additional thin (n)a-Si:H can (i) lower the E a of the (n)-contact, (ii) preserve the passivation quality after TCO sputtering, and (iii) prevent surface oxidation of (n)nc-Si:H induced by the TCO sputtering [142].
As for hole collection, we combined the optimized (p)nc-SiO x :H and (p)nc-Si:H as (p)-layer stack for SHJ solar cells.This bilayer stack improves the hole accumulation at the c-Si/(i)a-Si:H interface thanks to the high E g of (p)nc-SiO x :H and the low E a of (p)nc-Si:H, and the latter also promotes the charge carrier exchange at the (p)-layer stack/TCO interface.For instance, for cells with only 20-nm-thick (p)nc-SiO x :H as the (p)-layer, a high average device series resistance (R S ) of 2310 mΩ⋅cm 2 and limited FFs of around 71.5% were obtained.The addition of a (p)nc-Si:H significantly lowered the R S to approximately 950 mΩ⋅cm 2 and boosted the FF to 79.5% when combining a 4 nm (p)nc-SiO x :H with a 12 nm (p)nc-Si:H (see Fig. 6(b)).In the meanwhile, the V OC was also improved from 692 mV to 714 mV by using the optimized (p)-layer stack as shown in Fig. 6(b).Those results highlight the critical roles of using the layer with a low E a for contacting the TCO and additionally enhancing the band bending at the c-Si/(i)a-Si:H interface.Besides, by increasing the thickness of the (p)nc-Si:H layer, we generally observed increased V OC and FF [185].This is attributed to the decreased E a of the (p)-layer stack with a thicker (p)nc-Si:H.Moreover, the optimized HPTs can significantly decrease the R S of the (p)-contact stack from 2920 mΩ⋅cm 2 to 1440 mΩ⋅cm 2 , thus contributing an increment in FF from 70.9% to 77% for solar cells with 4 nm (p)nc-SiO x :H and 16 nm (p)nc-Si: H (see Fig. 6(b)).

Dopant-free molybdenum oxide (MoO x ) as efficient hole collectors
Alternative to Si-based (p)-layers that can induce significant parasitic absorption when placed on the illumination side [186], highly transparent Molybdenum oxide (MoO x ) (E g = 3 eV) was proven to be a promising replacement for efficient hole collection [42,43,47,187].MoO x is an (n)-type material with a high WF that mismatches with that of the c-Si.As a consequence, band bending near the c-Si interface can be formed and it is in favor of accumulating holes while repelling electrons [188].Both DFT and TCAD simulations from our group have suggested the importance of a high WF of MoO x for achieving efficient solar cells [54,189].While, as shown in Fig. 7(a), our DFT simulation indicates the formation of a strong dipole at the (i)a-Si:H/MoO x interface.This dipole induces a WF attenuation of MoO x and the degree of attenuation increases with a thicker MoO x , thus weakening the band bending.Besides, as illustrated in Fig. 7(b), the WF of MoO x can vary with its stoichiometry, or, the oxidation state of Mo [54].Especially in practice, the WF of MoO x is often degraded due to oxygen migration from MoO x that reacts with adjacent materials.This aspect was also emulated with WF profiles shown in Fig. 7(c) for different (i)a-Si:H/MoO x interfaces, where a thinner MoO x is needed for saturating the WF when deposited on a less-reactive interface.Therefore, we have developed a plasma treatment technique that is capable to preserve the high WF of MoO x as verified by XPS measurements [54,189].Furthermore, through both the TCAD simulations (see Fig. 7(d)) and experimental solar cell results, a MoO x layer as thin as ~1.7 nm can deliver the optimal trade-off between the WF and WF attenuation, which depends on the thickness of the MoO x .Following the theoretical findings, we demonstrated a record efficiency of 23.83% for a monofacial SHJ solar cell with this ultra-thin MoO x as an efficient hole collector [54].This cell also delivers an active area J SC,EQE of 41.63 mA/cm 2 , showcasing the optical advantageous employment of MoO x for FBC-SHJ solar cells.

High-mobility TCOs
Despite knowing the significance of the TCO properties in relation to the transport of charge carriers (introduced in Section 3.1), typically, an ideal TCO layer for high-efficiency SHJ solar cells should feature (i) a low sheet resistance (R sh ) for allowing efficient lateral transport of charge carriers to the metal electrodes while maintaining broadband transparency in the wavelength range of interest (300-1200 nm), (ii) low contact resistance with the surrounding layers, (iii) suitable refractive index for maximizing the light in-couping into solar cells and (iv) a minimum N TCO of around 10 20 cm − 3 for enabling efficient charge carrier transport (see Fig. 4).Note that an increased N TCO can lead to enhanced parasitic free carrier absorption (FCA) for near-infrared light (wavelength above 700 nm) as a result of intra-band transitions within the conduction band [190][191][192].To achieve a low R sh , both the μ e and the N TCO of the TCO can be increased.Therefore, we performed comprehensive studies on developing and implementing various TCOs, especially high-mobility tungsten-and fluorine-doped indium oxides (IWO and IFO), for high-efficiency SHJ solar cells.The material properties of different TCOs and their influence on solar cell performance were explored both experimentally and theoretically (via DFT modellings) [193][194][195].With DFT simulations, we were able to calculate the equilibrium geometric, electronic structure, effective electron mass and WF of ITO, IFO, and IWO [195,196].Especially, the last (simulated WF) provides additional insights for interpreting the effect of different TCO layers on the contact resistivity of doped contact stacks [195].
With optimized deposition conditions, a remarkably high μ e of 87 cm 2 V − 1 s − 1 was achieved together with an N TCO of 1.2 × 10 20 cm − 3 of an optimized IFO:H film.The resulting solar cell exhibited an enhanced J SC, EQE of 1.53 mA/cm 2 (integrated from external quantum efficiency    measurement), compared to our lab-standard ITO counterparts, while maintaining the FF [194].Furthermore, optimized room-temperature sputtered IWO films featuring μ e of around 45 cm 2 V − 1 s − 1 and N TCO of approximately 2 × 10 20 cm − 3 were implemented into SHJ solar cells, which also delivered higher J SC,EQE and slightly improved FF as compared to cells with our lab-standard ITO [193].We present in Fig. 8 the opto-electrical properties of various TCO layers deposited on different substrates.As shown in Fig. 8(a)-(c), it is worth noting that we observed substrate-dependent opto-electrical properties of TCO layers.Besides, those properties could be also affected by performing post-annealing on TCO layers.Those changes in TCO properties can be related to effused hydrogen originating from underlying thin-film Si layers, defect passivation by hydrogen, oxygen incorporation and increased crystallization during post-annealing treatment [193,195,197,198].Therefore, it is necessary to consider the substrate-and process-dependent opto-electrical properties of TCO layers when evaluating the performance of solar cells and performing simulations with properly extracted material properties [195,197].As depicted in Fig. 8 (d), the reduced absorptance of the IFO and IWO films in the near-infrared wavelength range can be ascribed to their lower N TCO values as indicated in Fig. 8(a).This observation is in accordance with the classical Drude Theory [191].The absorptance difference in the short-wavelength range reflects the various optical bandgaps of those TCO layers, which are 3.78 eV, 3.80 eV and 3.86 eV for ITO, IFO and IWO films, respectively, as extracted from Tauc plots [199].This again demonstrates the optical benefits of IFO and IWO films in comparison to the ITO layer [195].
In summary, with optimized (n)-and (p)-type contact stacks featuring nc-Si:H-based layers, we can minimize ρ c down to 33 mΩ⋅cm 2 and 144 mΩ⋅cm 2 , respectively [185].A low ρ c of 177 mΩ⋅cm 2 was also obtained for the MoO x -based contact stack [54].Especially, in the view of reducing scarce material usage, such as indium, we experimentally demonstrated that IWO and ITO as thin as 25 nm deliver promisingly low ρ c of around 80 mΩ⋅cm 2 and 114 mΩ⋅cm 2 when combined with nand p-layer, respectively [195].For cells with such thin TCO layers, both low ρ c values are essential for enabling c-Si wafer-assisted lateral transport [200].

Cu-electroplating for monofacial and bifacial solar cells
Ag screen printing (SP) is mostly used for manufacturing SHJ solar cells nowadays.The high Ag consumption imposes challenges for the terawatt (TW) implementation of SHJ technology [91].Besides, the minimum achievable finger width allowed by screen printing is around 20 μm-30 μm [201], which also caps the attainable efficiency due to relatively high shading losses (metal coverages are around 3.25% in our lab-standard Ag SP process).To decrease shading losses while maintaining a low finger resistivity, metal electrodes with a high aspect ratio are required.Such metal electrodes can also facilitate less front metal coverage and allow metal grids with smaller pitch sizes.To fulfil all aforementioned requirements, we developed and optimized a Cu-electroplating metallization approach which can be applied for monofacial plating and simultaneous bifacial plating [202,203].The latter is extremely desirable for fabricating bifacial SHJ solar cells that can deliver more power with less metal usage than the monofacial counterparts [204,205].
The schematic sketch of the simultaneous bifacial Cu-electroplating setup is illustrated in Fig. 9 (a), where two groups of standard threeelectrode cells were immersed in a CuSO 4 -based electrolyte.The three-electrode consists of the reference electrodes (RE, Ag/AgCl in this case), the working electrode (WE, wafer sample in this case) and the counter electrodes (CE, Cu sheets in this case).We utilized two independent Metrohm Autolab potentiostat tools for independently controlling the Cu-electroplating on both sides.Specifically, a full area of 100-nm-thick thermally evaporated Ag seed layer was first deposited on top of the TCO layer.Then, photolithography steps including applying photoresist, exposure and development were conducted to cover the area that was not intended to form Cu grids with photoresist.Then, the patterned area with the Ag seed layer exposed was electroplated with Cu.Afterwards, the remaining photoresist and the Ag seed layer underneath were removed.Detailed descriptions of the Cu-plating process are available in our publications [202,203].
With this approach, the usage of Ag is significantly minimized.
Typically, the line resistivity is 10.0 ± 5.0 μΩ⋅cm for our lab-standard Ag SP, while it is only 1.7 ± 0.1 μΩ⋅cm in the case of Cu-electroplated contacts.In addition, unlike our lab-standard screen-printed Ag grids, the electroplated Cu grids appear to be void-free in the bulk and conformal on the surface [203].Overall, Cu-plating allows more optimum design of metal grids with smaller pitch sizes (915 μm) than Ag screen printing (around 1300-1900 μm) while only featuring a small metal coverage of 1.55%, which significantly enhanced both the electrical and optical performance of our SHJ solar cells [203].

Optimum light management
Apart from minimizing the metal shading loss as discussed in section 3.6, parasitic absorption and reflection losses should also be minimized [186].The parasitic absorption is reduced by applying at the front side our developed (n)-type nc-Si-based or MoO x -based layer stacks, as either of them is more transparent than (n)a-Si:H or Si-based (p)-layers, respectively, mainly thanks to their higher bandgaps [54,142].Moreover, the use of high-mobility IFO:H and IWO TCO films can also lower Y. Zhao et al. the parasitic absorption as compared to standardly used ITO layers [193,194].To suppress the reflection loss, the optical simulation model Genpro4, which integrates ray-tracing and wave-optics, was employed for optimizing the optical design of devices [99].With spectroscopic ellipsometry, the optical constants and thicknesses of all layers applied in solar cells were experimentally extracted and used as inputs for optical simulations.For instance, a SiO x or MgF 2 layer with an optimum thickness was added to form a double-layer anti-reflection coating (DLARC) with the adjacent TCO layer [193,206].The application of optimized DLARC, which allows maximum light coupling into the solar cells, can facilitate the TCO reduction in both monofacial and bifacial SHJ solar cells [195].In Fig. 10, we present the schematic sketch of the fabricated 24.18%-efficient FBC-SHJ solar cell (see Table 1), its measured EQE and the simulated absorptance spectra with Genpro4.As illustrated in Fig. 10, the overall good correspondence between the measured EQE spectrum and simulated absorptance of the c-Si absorber provides additional insights into the optical constraints of the device, which can guide further improvements in device light management.

The evolution of FF and the best monofacial and bifacial FBC-SHJ solar cells
As a result of the abovementioned design strategies for achieving efficient selective transport and collection of charge carriers for highefficiency SHJ solar cells, we present in Fig. 11 the followed roadmap of FF development for FBC-SHJ solar cells that employ nc-Si:H-based contact stacks in the PVMD group.Initially, due to the limited insights on charge carrier transport and the insufficient control over crucial processing steps and materials, we observed a rather large spread of FF for cells with unoptimized nc-Si-based contact stacks.After performing rigorous device simulations, we gained a better comprehension of the transport mechanisms of charge carriers.The simulations also revealed key accessible material parameters that can be utilized to efficiently develop and optimize the contact stacks.Specifically, with different optimization objectives as depicted in Fig. 11, we progressively improved in a few years the maximum achievable FF and constrained the distribution of the obtained FFs.In this contribution, we present the best attained FF value of 83.3%, which can be further enhanced with the continuous process and material optimizations on the passivation, the effectiveness of the contact stack and the metallizations.
Lastly, to evaluate the developed contact stacks with optimized DLARC and Cu-electroplating metallization, monofacial and bifacial FBC-SHJ solar cells were fabricated with the structures shown in Fig. 12. Rear junction solar cells with doped nc-Si:H-based contact stacks are sketched in Fig. 12(a) and (b).Front junction solar cells with MoO xbased contact stack are illustrated in Fig. 12(c) and (d).External parameters of fabricated best solar cells are presented in Table 1.Monofacial FBC-SHJ solar cells featuring doped nc-Si:H-based contact stacks reached an efficiency of 24.18% with a FF of 83.30%, while bifacial solar cells demonstrated efficiency approaching 23% (n-side illumination) with room for further optimization.A bifaciality factor of 0.95 is calculated for this bifacial solar cell.Alternatively, front junction SHJ solar cells with MoO x window layer obtained efficiencies of 23.83% and 23.25% (p-side illumination) for monofacial and bifacial configurations, respectively.A high bifaciality factor of 0.98 was obtained for the bifacial cell endowed with MoO x .All devices feature less TCO usage (11%-56% less) as compared to our lab-standard TCO thicknesses.Especially, bifacial solar cells provide a more effective way to reduce TCO and metal consumption as compared to their monofacial counterparts.

'Beyond' standard FBC-SHJ architectures
The aforementioned results show our lab-scale R&D experience in processing and optimizing low-thermal budget FBC-SHJ solar cells.With a similar approach, the industry reported recently a record conversion efficiency of 26.81% for single-junction c-Si devices using SHJ technology, which achieved almost its maximal potential [34].This excellent result indicates that SHJ technology is ready to explore alternative device architectures to simplify the process while being highly performant.In this context, we propose here novel solar cell structures and point out their corresponding advantages and potential drawbacks or limitations.To do so, we utilized TCAD simulations tools [98], which were consistently coupled with Genpro4 [99], to simulate different solar cell architectures under one sun illumination accounting for the same c-Si parametrization (i.e. 2 Ω⋅cm, 100-μm-thick with 2% front metallization) as the optimized reference baseline SHJ solar cell.Fig. 13 shows the schematic sketches of the proposed device configurations with reference stacks of layers.Note that for the sake of simplicity, we report the structure with better light management between front or rear junction (FJ or RJ) schemes considering realistic choices of thin-film layers.As a result, we evaluated architectures featuring dopant-free, TCO-selective and buried junctions with FJ structures, whereas the reference baseline SHJ solar cells and devices with localized contacts were assessed with RJ configurations.Table 2 summarizes the simulated external parameters of the various SHJ-based device structures.
As seen in Table 2, for the reference baseline SHJ device (see Fig. 13  (a)), we simulated an efficiency of 26.82% with a V OC of 0.753 V, a FF of 0.869 and a J SC of 40.99 mA/cm 2 .Note that we considered (n)-layer on the front as such a functional layer is typically less absorptive than a functional (p)-layer [131].Further, SHJ solar cells with dopant-free passivating contacts are a viable alternative to substitute rather parasitically absorptive doped Si-based thin-film layers when placed on the illumination side.With this device configuration as shown in Fig. 13(b), we performed simulations based on our optimized ultra-thin MoO x layer on the front side [54], which led to an increment in J SC to 41.13 Fig. 10.The schematic sketch of the fabricated 24.18%-efficient SHJ solar cell (left), and its EQE and simulated absorptance spectra (right).The mismatch between the EQE spectrum and the simulated absorptance profiles can be attributed to (i) the nonideal collection efficiency of charge carriers in EQE measurement while the simulation considers only optical effects, (ii) underestimated absorption of TCO in the near-infrared range due to limited sensitivity of the reflective-type SE used for extracting the extinction coefficient [207], (iii) possible non-negligible plasmonic absorption of rear Ag is not considered in the simulation [208]., SHJ solar cells with buried junctions [210] and SHJ solar cells with localized contacts [211].
Y. Zhao et al. mA/cm 2 .As a result, the calculated performance of the dopant-free structure is 0.14% abs higher than the simulated baseline SHJ solar cells.Moreover, aiming at having more transparent contacts and simpler processes, we propose a solar cell structure in which front and rear TCOs act as HTL or ETL layers, without doped or dopant-free layers (see Fig. 13(c)) [209].As expected, the external parameters depend on the opto-electrical properties of the TCO, in particular the one acting as the HTL.In this regard, promising TCO candidates are those that can induce an electrical field within the bulk of the c-Si absorber due to their relatively high work function.For instance, IWO is one such candidate [193].Similarly, Al-doped zinc oxide (AZO) exhibits the properties of being a promising candidate as the ETL [212].As a result, for TCO-selective devices, we calculated 41.19 mA/cm 2 , 0.753 V, 0.863 and 26.76% as J SC , V OC , FF and conversion efficiency, respectively.Such parameters are similar to those calculated for the baseline SHJ structure.Nevertheless, the main advantage of such solar cells stands in the fact that fewer layers are used, besides the eventual practical challenges, such as realizing a thin (i)a-Si:H layer (stack) being resilient against subsequent TCO processing.Moreover, the use of less damaging TCO deposition techniques, such as pulsed-laser deposition [213] and reactive plasma deposition [214,215], is expected to also contribute to the experimental optimizations of TCO-selective SHJ solar cells.
Taking the TCO-selective structure as a reference but to allow more flexibility in choosing TCO materials, we propose to combine this approach with a c-Si wafer that has surfaces been previously Boronand/or Phosphorus-diffused (see Fig. 13(d)).The purpose of this doping profile is to enable a built-in internal electrical field inside the bulk fulfilling the junction condition similar to the in-diffusion of cells with poly-Si-based carrier-selective passivating contacts [216][217][218].With this structure, we were able to select rather transparent TCOs on both sides that lead to a simulated J SC of 41.33 mA/cm 2 , which is 0.2 mA/cm 2 higher than the dopant-free option.Despite the optical gain, we observed a V OC drop of 0.12 V and a FF reduction of 1.7% abs as compared to the baseline SHJ solar cells, due to the increased intrinsic recombination in the highly doped regions.Accordingly, the calculated conversion efficiency is 26.12% for SHJ solar cells with buried junctions.
Last but not least, we evaluated a solar cell structure with the same contact stack as our baseline SHJ devices, but the front contact layers, namely, (i)a-Si:H, (n)-layer and TCO are localized beneath the front metal contact as depicted in Fig. 13(e).With this structure, we obtained a 1.35 mA/cm 2 increase in J SC which results in 0.78% abs efficiency gain with respect to the baseline SHJ solar cells, leading to a conversion efficiency of 27.60%.The experimental development of this solar cell architecture might require the fabrication of proof-of-concept intermediate structures such as cells with localized TCO, or devices with localized TCO and doped layers.It is worth noting, besides all these proposed architectures as shown in Fig. 13 can be adapted to be bifacial, they can be also modified to be used as bottom cells for tandem applications, for example, with perovskite top cells.

Conclusions
This work presents an overview of strategies for realizing highefficiency FBC-SHJ solar cells with the assistance of advanced optoelectrical simulations.Different optimization aspects are briefly discussed, namely, the development of high-quality (i)a-Si:H passivating layer, efficient selective transport of charge carriers from c-Si to TCO, optimization of metallizations and light management of the solar cells.Developed nc-Si-based contact stacks are implemented into Cuelectroplated monofacial and bifacial FBC-SHJ solar cells, achieving efficiencies of 24.18% and 22.84%, respectively.Alternative MoO xbased contact stacks deliver efficiencies of 23.83% and 23.25% for monofacial and bifacial configurations, respectively.As compared to monofacial solar cells, bifacial solar cells provide a more effective way to significantly reduce the usage of In and Ag.Lastly, we propose multiple innovative solar cell designs based on SHJ technology.Their potential performance was evaluated and compared.The FBC-SHJ solar cells that feature localized contacts were simulated to achieve a practical maximal efficiency of 27.60%, which surpasses that of the baseline SHJ solar cells mainly due to the significantly reduced parasitic absorptions.

Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Fig. 4 .
Fig. 4. Contour plot of (a) ρ c and (b) FF of IBC-SHJ solar cell as function of E a and N TCO at (p)-type contact stack.Figure adapted from Ref. [103].

Fig. 5 .
Fig. 5.The effective minority carrier lifetime (τ eff ) of (a) symmetrical samples as a function of minority carrier density (MCD), (b) the τ eff of solar cells before metallization and the (c) V OC and (d) FF of completed solar cells as a function of deposition temperature of (i)a-Si:H underneath both doped layers.In Figure (a), HPTs stand for hydrogen plasma treatments, the n represents 3 nm (n)nc-Si:H + 2 nm (n)a-Si:H, whereas the p denotes 5 nm (p)nc-SiO x :H + 30 nm (p)nc-Si:H.Note in (a), only (i)a-Si:H layers were deposited at a substrate temperature of 160 • C while all other layers and HPTs were conducted at a substrate temperature of 180 • C. In (b), the same (n)-layer stack is used as that in (a), the (p)-layer stack is 5 nm (p)nc-SiO x :H + 16 nm (p)nc-Si:H.Figure adapted from Ref. [118].

Fig. 7 .
Fig. 7. (a) The differential charge at the Si/MoO x interface for MoO x featuring different stoichiometry.The inset illustrates the cross-sectional view of the Si/ MoO x interface, with the dipole forming within the 2.5 Å-thick interface indicated by the arrow.The light blue atoms represent Si, while the grey and the red atoms denote the Mo and O in the MoO x , respectively.(b) The simulated work functions of MoO x as a function of the x for different oxygen deficiencies: O3 and O2 removal.The unit cell of MoO 3 as modelled in DFT simulations is depicted in the inset.(c) Different work function profiles utilized for simulating a reactive (blue squares) and a less-reactive (i)a-Si:H/MoO x interface (black circles).(d) The simulated conversion efficiency of SHJ solar cells with MoO x as the hole collector using work function profiles taken from (c).Figure adapted from Ref. [54].

Y
.Zhao et al.

Fig. 9 .
Fig. 9.The (a) schematic sketch of the simultaneous bifacial Cu-electroplating setup, the (b), (c) scanning electron microscope (SEM) and (d), (e) optical microscope images of electroplated Cu fingers, and screen-printed Ag contacts.In (a), WE represents the working electrode (wafer sample), RE stands for the reference electrode (Ag/AgCl) and CE refers to the counter electrode (Cu sheets).

Y
.Zhao et al.

Fig. 11 .
Fig. 11.The followed roadmap of FF development for FBC-SHJ solar cells with nc-Si:H-based contact stacks in the PVMD group.The roadmap includes detailed optimization objectives and approximate periods of different development steps.The data points in the plot represent the FF of individual solar cells.

Table 1
Best fabricated monofacial and bifacial FBC-SHJ solar cells with Cu-electroplating and thinner TCO.
a Compared to our lab-standard 75 nm front and 150 nm rear TCO thicknesses.b Independently certified at ISFH CalTeC or Fraunhofer ISE CalLab PV Cells.

Table 2
Simulated external parameters of various SHJ-based device architectures.