Multifunctional Titanium Oxide Layers in Silicon Heterojunction Solar Cells Formed via Selective Anodization

Herein, a novel strategy is introduced to reduce the consumption of scarce materials in silicon heterojunction solar cells by combining approaches for Ag replacement in the metallization and a reduction of the indium tin oxide layer thickness: a Ti layer deposited by physical vapor deposition serves both as the contact layer of a copper‐based metallization and after electrochemical oxidation as capping layer enabling the use of a thinner transparent conductive oxide. Further, the TiOx layer can build an encapsulation layer. While oxygen evolution and metal dissolution are found to be critical side reactions, a nonaqueous electrolyte is found in which these reactions can be avoided. The application on silicon heterojunction solar cells shows promising first results, exhibiting a short circuit current density of 35 mA cm−2 and a cell efficiency of close to 21% despite nonoptimized layer thicknesses.


Introduction
Silicon heterojunction (SHJ)-solar cells are gaining increasing relevance, because of their high efficiency and attractive levelized cost of energy (LCOE) potential.However, critical factors in terms of production costs and security of supply are the silver used for the screen-printing metallization and the rare element indium, which is the main part of the indium tin oxide (ITO) electrode layer.Strategies must be found to reduce or replace these elements so that the SHJ technology can increase its share in the photovoltaic market, which will soon grow into the terawatt range of annual production. [1]egarding silver reduction, several copper electroplating approaches were evaluated in the last decade, [2,3] for example, using an organic resist masking as published by Maxwell/Sundrive [4] or Centre Suisse d'Electronique et de Microtechnique. [5]Another option is the NOBLE (native oxide barrier layer for selective electroplating) process sequence, [6] developed at Fraunhofer ISE, which allows to replace silver screen printing by copper electroplating without using costly organic masking, as shown in Figure 1a.Metal layers are deposited via physical vapor deposition (PVD) sequentially without vacuum break and work as adhesion (Ti), plating seed (Cu), and masking (Al) layers.Currently, applying the standard NOBLE process sequence, the complete metal layer stack is removed by wet chemical etching after the selective plating process in the nongrid areas.
The approaches to reduce the indium content of SHJ-solar cells have been proposed, for example, by Meyer Burger, [7] Kaneka [8] or Fraunhofer ISE, [9] applying SiN x , SiO x , or TiO x layers on top of a thinned transparent conductive oxide (TCO).These layers are deposited in an additional step after metallization.The partial replacement of ITO by aluminum-doped zinc oxide (AZO) has also been investigated. [10]n this contribution, we introduce a new concept to replace the indium partly or fully.This concept can elegantly be integrated into the NOBLE-process sequence, but potentially also be combined with other SHJ-metallization routes or applied to perovskite solar cells.We propose to use titanium oxide (TiO x ) as a capping layer, which is made by electrochemical oxidation (anodization) of a thin PVD Ti layer.This anodization step takes place in a liquid electrolyte via setting a positive potential to the Ti layer.The anodization current can be distributed by the solar cell and/or by the TCO layer below.
][13] 2) The TiO x works as an encapsulation layer.This can increase the module stability and lifetime and can even enable the application DOI: 10.1002/solr.202300418Herein, a novel strategy is introduced to reduce the consumption of scarce materials in silicon heterojunction solar cells by combining approaches for Ag replacement in the metallization and a reduction of the indium tin oxide layer thickness: a Ti layer deposited by physical vapor deposition serves both as the contact layer of a copper-based metallization and after electrochemical oxidation as capping layer enabling the use of a thinner transparent conductive oxide.Further, the TiO x layer can build an encapsulation layer.While oxygen evolution and metal dissolution are found to be critical side reactions, a nonaqueous electrolyte is found in which these reactions can be avoided.The application on silicon heterojunction solar cells shows promising first results, exhibiting a short circuit current density of 35 mA cm À2 and a cell efficiency of close to 21% despite nonoptimized layer thicknesses. of chemically less stable, but cheaper and indium-free TCOs (e.g., aluminum-doped zinc oxide, AZO) instead of ITO. [14]he layers below the TiO x , are protected from the chemical liquids used for plating, etching, and anodizing as the process is based on oxygen and titanium ion movement through the solid material; and 3) The challenging selective wet etching step of the Ti layer, when Ti is applied in the standard NOBLE process as an adhesion layer, is replaced by the anodization step.Therefore, the application of hazardous chemicals for selective chemical etching of the Ti [15] is avoided and the process route remains lean, as no additional process step is required.
The adapted process scheme is depicted in Figure 1b, for example, of NOBLE metallization.For alternative metallization approaches, the titanium layer can also be deposited on top of the TCO (without vacuum break) before any other metallization step, as it is shown in Figure 1c for the example of screen-printing metallization.The thereby inserted Ti layer between the metal contact and TCO could additionally reduce the contact resistivity.
In the literature, many publications and patents can be found concerning the anodization of Ti, introducing manifold electrolytes.The type of electrolyte strongly influences the anodization process, as it determines possible side reactions such as oxygen evolution or metal dissolution and also affects the morphology and crystallinity of the TiO x layer. [16]A typical growth characteristic of %2 nm TiO x per applied anodization voltage is found in the literature. [16]In general, TiO x is known as chemically stable material exhibiting a low absorption in the visible range and a tunable refractive index depending on crystallinity and crystal phase. [9,17,18]For our application, the main challenge is to achieve a full, homogeneous, and selective anodization of the thin Ti layer without damaging the metallic grid or solar cell.These challenges are not investigated in the literature, as the available publications are focused on massive titanium metal pieces.A compact and hole-free TiO x layer is required to provide encapsulating properties and the optical properties (e.g., refractive index) should be preferably similar to the underlying TCO to match antireflection conditions.
In a comparative study, we have tested three different electrolytes: one aqueous electrolyte and two nonaqueous electrolytes.We analyzed their performance regarding the morphology of the resulting TiO x layer, damage to the metallic grid on solar cell, and optical performance in solar cells.

Visual Investigation
Screen-printed SHJ solar cells with ITO and a thin Ti layer on top were anodized in the three electrolytes.Figure 2 shows photographs of four anodized samples (b-e) and a reference sample (a, without Ti).The top part of the anodized samples displays the grayish metallic color of the Ti layer on the ITO on textured SHJ solar cells (only the bottom part was immersed in the electrolyte).The bottom, anodized part of the samples appears blue due to the transition from metallic Ti to transparent TiO x .Compared to the reference cell the anodized samples exhibit a brighter blue, due to a change in interference color caused by the additional TiO x layer.The three solar cell samples (b-d) show slightly different shades of blue.The blue of the sample anodized in the aqueous electrolyte (sample b) is less bright compared to the two other samples.The sample using AZO instead of ITO (sample e) exhibits a homogeneous blue color in the anodized part, proving a homogeneous anodization and an intact AZO layer underneath the TiO x .This result demonstrates the capability of the Ti/TiO x layer to work as a protection and encapsulation layer for the underlying TCO.For availability reasons, we used only ITO samples for the further investigations.

Electron Microscopy Measurements
For a deeper insight into the layer properties, samples were investigated with high-angle annular dark-field (HAADF) scanning transmission electron microscopy (STEM), to image the structures, together with electron energy loss spectroscopy (EELS) for elemental/phase mapping using multilinear least square fitting (MLLS).The results obtained from the HAADF-STEM-EELS analyses are shown in Figure 3.The Ti reference reveals a compact Ti layer (Figure 3a).The layer anodized with the aqueous electrolyte (Figure 3b) appears also compact and hole-free.However, there are some metallic Ti (green) residues visible, especially at the interface to the ITO.As Ti is a strong light absorber, this is of course not acceptable for the final application.This incomplete anodization might be caused by the strong side reaction of oxygen evolution at the metal grid.This hypothesis is supported by preliminary results with gridfree samples, where no oxygen evolution was observed.These samples show a higher degree of anodization.The layer obtained from anodization in nonaqueous electrolyte 1 seems to contain no or just a very small amount of Ti residues (Figure 3c), no accumulation is visible at the interface to the ITO, indicating an almost complete anodization.This layer is rather porous, containing holes of 5-10 nm size, which could be problematic in view of the encapsulating properties of the layer.The anodization in nonaqueous electrolyte 2 (Figure 3d) also leads to almost full conversion to TiO x , no Ti islands are visible at the interface to the ITO.The layer exhibits some small voids but seems to be dense and compact in general.For all electrolytes, the layer thickness is increased by anodization to 25-30 nm.This is, additionally to the partly porous nature of the layers, caused by the higher molar volume of TiO x compared to Ti. [19] To analyze the selectivity of the anodization to Ti versus the silver grid, scanning electron microscopy (SEM) pictures of some metal fingers have been made.They are shown in Figure 4.The exemplary metal finger in Figure 4a is unaffected by the anodization process, the metal has not been dissolved.The samples anodized in the aqueous and the nonaqueous electrolyte 2 behaved like this.Contrary to that, as shown in Figure 4b, the anodization in nonaqueous electrolyte 1 leads to the dissolution of the silver, leaving behind only the organic matrix of the screenprinted finger.Testing of a sample with a copper grid revealed the same behavior: the copper was also completely dissolved during anodization in nonaqueous electrolyte 1.This is of course not acceptable, making nonaqueous electrolyte 1 unsuitable for our final application.

Light Beam-Induced Current Measurements
The contour plots in Figure 5 display the results from light beaminduced current (LBIC) [20] measurements.This technique scans  Figure 5a shows the reference sample before Ti sputtering.An absorption of almost 100% is visible.In contrast, for the nonanodized sample in Figure 5b nearly all of the incident light is absorbed by the only 15 nm thin metallic Ti layer, leading to very low current feedback.For the sample anodized in the aqueous electrolyte in Figure 5c, the external quantum efficiency (EQE) is increased compared to the Ti reference.Nevertheless, the overall level of 65% absorption is still low.This observation is well in line with the STEM-EELS analysis, revealing Ti residues at the interface and therefore an incomplete anodization.Contrarily, the samples anodized in nonaqueous electrolytes 1 and 2 (Figure 5d,e, respectively) exhibit a high EQE of roughly 92% and 93%, respectively.These results fit also well with the STEM-EELS analysis which indicated a higher degree of conversion from Ti to TiO x in the nonaqueous electrolytes.

IV-Measurements
IV-measurements on SHJ solar cells anodized in nonaqueous electrolyte 1 were impossible due to the completely dissolved grid.The cells anodized in the aqueous electrolyte had a very bad overall performance, as already expected from the LBIC and STEM-EELS results.Therefore, Table 1 lists only the IV results for the two reference cells (without any Ti layer and after Ti sputtering) and the sample anodized in nonaqueous electrolyte 2. The short circuit current density ( J SC ) is strongly increased from 12.5 to 35.1 mA cm À2 by the anodization step, leading to a solar cell efficiency of almost 21%.Nevertheless, a gap of 3 mA cm À2 in J SC and roughly 2% abs in efficiency is observed compared to the reference cell without Ti.This loss can be partly explained by the nonoptimized layer thicknesses in the stack (see Section 2.5), as the 25-30 nm of TiO x are added on top of the 75 nm ITO, which leads to unwanted reflection losses.Additionally, the process sequence is not optimized regarding setup and processing voltage and time, and solely a limited number of samples was processed to prove the principle.

Optical Simulations
To analyze the observed loss in J SC , the wafer ray tracer tool of PV lighthouse has been used to simulate the J SC of different layer stacks.The results are given in Table 2. Three different datasets for the TiO x layer were used to demonstrate the dependence of the J SC loss on the refractive index of the TiO x .A loss of 0.7-2.4mA cm À2 in J SC can be explained by the nonoptimized thicknesses of the layer stack, mainly due to reflection losses at the front side.Unfortunately, the refractive index of our anodized TiO x is not known to date, as the processing of suitable samples to determine the refractive index is challenging (anodizing process depends on the substrate).
A residual Ti layer of 1 nm thickness would lead to an additional loss of 3.5 mA cm À2 , the total loss sums up to 4.2, 5.1, and  Table 1.IV-results of the references with screen-printed Ag grid and the solar cell anodized in nonaqueous electrolyte 2. 5.8 mA cm À2 , respectively.With this result, it can be concluded that the anodized cell, presented in the previous section, contains a maximum or less than 1 nm residual Ti (maybe distributed inhomogeneously over the cell surface), as the loss in J SC was only 3 mA cm À2 .The STEM-EELS characterization at one position shows no residual Ti for the best process, demonstrating that it is possible to convert the Ti layer completely to TiO x .However, this analysis only gives a limited insight regarding Ti residuals at other positions on the solar cell.Our hypothesis on the cause of the 3 mA cm À2 current loss is that either there are few isolated sub-nm Ti residuals scattered over the cell surface (which would be hard to detect), or the refractive index of the TiO x layer matches so that we experience this loss.In both cases, further adaptations to the anodization process would need to be made, so that the process sequence becomes competitive.
As the refractive index of our anodized TiO x is most probably too high, voids in the layer (as observed for the first nonaqueous electrolyte) might help to reduce the effective refractive index.The mixed value of air and TiO x should be lower than the pure TiO x value.But, in contrast, voids might reduce the encapsulation properties of the TiO x layer.It is expected that by a variation of electrolyte and processing conditions, a lowering of the refractive index is possible.The simulation based on a reduced ITO layer thickness (optimized to the TiO x thickness) and a suitable refractive index of the TiO x demonstrates that our approach enables at least a similar J SC in principle.

Conclusion and Outlook
In this contribution, we introduced a completely new approach to reduce indium consumption in SHJ-solar cell manufacturing.This approach can be well combined with the NOBLE approach, enabling a simultaneous reduction of silver and indium consumption and other metallization approaches.It is applicable to large-scale production, as it can be easily put into practice as an inline process, but the overall fabrication process remains lean.We propose to use an electrochemically oxidized TiO x layer as a capping and encapsulation layer and investigate three different types of electrolytes for the selective anodization of thin Ti layers to TiO x .Aqueous electrolytes seem to be unsuitable due to the strong side reaction of oxygen evolution at the metal grid, preventing full anodization.The nonaqueous electrolyte 1 allows (almost) full anodization but leads to porous layers and is nonselective, as it dissolves the metal grid.Therefore, it does not fulfill one of the major requirements and is unsuitable for our application.The third type of electrolyte (nonaqueous electrolyte 2) gives the most promising results, showing an (almost) complete and selective anodization of the Ti layer, resulting in a J sc of 35 mA cm À2 .Nevertheless, further investigation is required, to better understand the anodization process and the resulting TiO x layer (e.g., determination of the refractive index).Possible future steps include the evaluation of other nonaqueous electrolytes, manufacturing of solar cells using the NOBLE process and/or AZO as TCO with an optimized layer thickness, and the integration of solar cells into modules and subsequent stability tests (e.g., UV behavior of the TiO x layer).

Experimental Section
Small-sized (2.1 Â 2.1 cm 2 ) silver screen-printed textured SHJ-solar cells, with rear pþ emitter, fabricated according to the study of Luderer et al. [21] were used.Additionally, large M2 SHJ solar cells were used as test samples.A thin Ti layer (%15 nm) was sputtered on top of the metalized solar cells.This is not the targeted order, but for simplicity, it is applied for the pretests for evaluating the electrolytes.The M2 solar cells were cut into small pieces (30 Â 50 mm 2 ) and anodized in the three different electrolytes.The anodization was performed at 20-30 V for 30-300 s.
Thin sections were removed from the anodized samples using focused (Ga) ion-beam (FIB)-SEM.The thin sections were then analyzed using a probe-corrected JEOL ARM 200F transmission electron microscope operated at 200 kV.HAADF STEM imaging sensitive to atomic number variations within the sample, and EELS spectrum imaging (SI) were performed using the DualEELS mode on a Gatan GIF Quantum ERS spectrometer.Elemental phase maps consisting of Ti, TiO x , and In 2 O 3 were extracted by performing multilinear least squares fitting (MLLS) on the 3D data cube after splicing the low-loss and high-loss SIs together.Reference spectra for the background, Ti, TiO x , and In 2 O 3 phases were extracted from the experimental data set and used in the MLLS analysis.

Figure 1 .
Figure 1.a) Standard NOBLE process sequence, b) adapted NOBLE process sequence, the chemical etching step of Ti is replaced by the anodization step, and the total number of process steps remains the same.c) Process sequence for implementing the Ti anodization in standard screen-printing metallization route.

Figure 2 .
Figure 2. a) Photographs of a reference textured SHJ solar cell with screen-printed Ag grid prior to Ti sputtering.b-d) Three exemplary samples anodized with the aqueous and the two nonaqueous electrolytes and e) one anodized test sample with AZO instead of ITO layer.The bluish bottom part is anodized and the grayish top part displays the nonanodized titanium layer.

Figure 4 .
Figure 4. SEM pictures of screen-printed silver fingers after anodization in nonaqueous electrolyte a) 2 and b) 1.

Figure 5 .
Figure 5. LBIC measurements of a 20 Â 20 mm 2 area at 658 nm on textured SHJ solar cell samples.a) Reference sample, without Ti layer.b) Reference sample with Ti layer, without anodization.c-e) Samples anodized in aqueous electrolyte (c), nonaqueous electrolyte 1 (d), and nonaqueous electrolyte 2 (e).The metal grid (busbar and fingers) appears red/brighter due to reflection of the incident light at these positions.