TOPCon shingle solar cells: Thermal laser separation and passivated edge technology

This work shows the first demonstration of thermal laser separation (TLS) and post‐metallization passivated edge technology (PET) applied to tunnel‐oxide passivated contact (TOPCon) shingle solar cells. The shingle solar cells with 26.46 mm × 158.75 mm size are separated from industrial full‐square TOPCon host solar cells. The singulation is performed either by TLS from the front side (emitter side) or by conventional laser scribe and mechanical cleaving (LSMC) from the rear side (emitter‐free side). The TLS optimized in this work yields up to 0.2%abs more efficient shingle cells after separation in comparison with LSMC‐separated shingle cells. The most promising PET sequence identified for the singulated TOPCon cells consists of depositing an 8‐nm‐thin aluminum oxide layer by thermal atomic layer deposition at a temperature below 200°C in conjunction with subsequent hotplate annealing at 250°C. Application of the PET yields a boost of up to 0.5%abs in energy conversion efficiency for edge‐passivated TOPCon shingle cells in comparison with their performance directly after separation. This efficiency‐increasing impact of the PET sequence is found not to be strongly dependent on the separation process applied. The most efficient TOPCon shingle cell after PET achieves an efficiency of 22.0% and has been singulated by TLS.


| INTRODUCTION
Cutting large-area solar cells in at least two sub-cells is nowadays very common in the solar cell industry. 1,2 Separated cells result in lower current per cell and a quicker increase of module voltage in series interconnection. In addition to the approach of cutting a large cell into, for example, two or three sub-cells, there is also the approach of singulating even more sub-cells, also known as shingle solar cells. 3,4 Research activities on shingle solar cells were and are being carried out for various solar cell types as for, for example, passivated emitter and rear cells (PERC), 4-8 silicon heterojunction (SHJ) cells, [9][10][11][12][13] or tunnel-oxide passivated contact (TOPCon) cells. 13 SHJ and TOPCon solar cells implement passivating contacts, resulting in increased voltages and higher efficiencies compared with PERC, which is still the standard in industry today. 14 The amorphous silicon-based SHJ solar cells are usually processed at temperatures of up to around 200 C to maintain their advantages, 15,16 whereas the polycrystalline siliconbased contacts of TOPCon solar cells can withstand similar hightemperature processing as for PERC solar cells. This makes them compatible with existing mass fabrication and of high interest in research and industry.
Surface recombination in solar cells reduces the number of excited charge carriers on all surfaces (leading to power loss) but is more pronounced at newly created, that is, unpassivated, edge surfaces of cut solar cells. As the perimeter-to-area ratio increases with decreasing sub-cell size, edge recombination becomes more and more important the smaller the cells get. 17,18 The impact of edge recombination becomes even more significant with cells of higher efficiency potential. Although edge recombination is also present in today's halfor third-cells, it is of extraordinary importance for shingle cells. Thus, for fabrication of highly efficient shingle cells/modules, the challenge consists in minimizing edge recombination: i. The cutting of the cells should be performed with low-damage technologies. The reference separation technology is a conventional laser scribe and mechanical cleaving (LSMC) process. To create the breaking point for LSMC, a laser is used to form a continuous scribe (i.e., laser ablation) over the entire length of the silicon cell at the desired positions. This scribe then enables the mechanical cleave into sub-cells. In contrast to LSMC, thermal laser separation (TLS) 19 needs only a very short initial laser scribe to create a starting crack that can then be propagated through the wafer in any direction by a cleave laser combined with a water-air aerosol jet. This results in cuts with very smooth edge surfaces. Several publications have shown application of TLS on PERC 7,20-22 or SHJ 10,21,23 devices. Although it is also applicable to TOPCon solar cells, the authors are not aware of any publication on this.
ii. The charge carrier recombination at the newly created edge surfaces should be minimized. This can be achieved by, for example, edge passivation. 17,24,25 After successful experimental proof of an approach for edge passivation and modeling of the respective potential, 18 Fraunhofer ISE filed a patent application in 2018. In 2019, Fraunhofer ISE introduced the post-metallization passivated edge technology (PET) and demonstrated edge passivation on PERC shingle cells. 7,26 The PET consists of the deposition of a passivation layer after cell separation, for example, an aluminum oxide (Al 2 O 3 ) layer-which is known for excellent passivation quality 27 -with subsequent activation of the same by means of elevated temperature exposure, also referred to as annealing.
The temperatures for both processes can be kept at or below 250 C and thus make the PET concept also very attractive for SHJ and TOPCon cells. For SHJ, the PET has been also demonstrated on half-cut cells, 23 and the concept of the PET approach was taken up by INES in 2020 10 and demonstrated on SHJ shingle cells. [10][11][12]28 To the authors' knowledge, the application of the PET has so far not been examined for TOPCon shingle cells in already published work.
Hence, this paper demonstrates its applicability in conjunction with TLS for TOPCon cell cutting.

| APPROACH AND EXPERIMENT
This study aims to develop improved processes for cell separation and edge passivation that can be applied for fabricating high-quality TOPCon shingle solar cells. Conventional LSMC serves as a reference process for cutting silicon cells. We compare this reference process with TLS and optimize the latter for the TOPCon cell architecture.
TLS is performed with a "microDICE" tool made by 3D-Micromac. 29 We then proceed to optimize edge passivation for TOPCon shingle solar cells. Finally, the optimized PET sequence is examined on shingle solar cells that have been separated either by LSMC or by TLS.

| TOPCon host solar cells
The experiment is performed using industrial TOPCon host solar cells.
The devices, shown in Figure

| Experiment plan
The experiment plan is schematically illustrated in Figure 2. It contains three sub-experiments denoted as experiments A to C: • Experiment A targets the optimization of the TLS process for low-damage cutting of the TOPCon host cells into shingle cells.
• Experiment B targets the optimization of the thermal atomic layer deposition (ALD) of Al 2 O 3 for edge passivation.

| Experiment A: TLS optimization
The starting point for optimizing the TLS process is the process previously used at Fraunhofer ISE for completely internally fabricated TOPCon shingle solar cells. Based on previous experience, it is a good idea to optimize the TLS process with regard to the host cells actually used. This optimization is performed in three steps: (A-I) Optimization of the cleave step to minimize/prevent surface damage close to the dividing line.Therefore, different cleave processes are applied either on the front side (emitter side) or on the rear side (emitter-free side) of host cells without initial scribes (i.e., the samples are not cut). This allows an assessment of the cleave step using photoluminescence (PL) imaging as described in Baliozian et al. 21

| Experiment B: ALD optimization Al 2 O 3 layer
For the optimization of the thermal ALD of Al 2 O 3 on shingle cells, the optimized TLS process from Section 2.2.1 is applied from the front side. The ALD process is performed in a "FlexAl" system from Oxford Instruments using trimethylaluminum and water vapor as precursors.

| Experiment A: TLS optimization
The TOPCon host cells are only exposed to the cleave step within the TLS sequence either on the front or the rear side without initial laser scribing. With the host cells still in one piece, potential damage of their surface passivation due to the cleave process can be quantified and minimized.
Exemplary PL images before and after laser cleaving with two different laser powers are shown in Figure 3. The PL images before the cleave step show a bright area between the shingle cells a and b as there are no busbar contacts present on the front side. After performing the cleave process with higher laser power, this area between a and b is no longer bright but dark. This indicates a degradation of the passivation quality in the area where the spatially extended laser beam of the cleave step interacts with the surface passivation. In contrast, for the lower laser power, the PL images before and after laser cleave are identical. Thus, the PL intensity is not lowered by the cleave step meaning that the surface passivation quality is not  Figure 4A,B, respectively, grouped by the ALD Al 2 O 3 passivation treatment.

Cutting the host cells by TLS into shingle cells leads to losses in
open-circuit voltage V OC of ΔV OC = À4 mV and in pseudo fill factor pFF of ΔpFF = À1.2% abs (on average for all five groups). A clear improvement is then seen in Figure 4B for the first four groups after deposition of the ALD Al 2 O 3 layer (blue vs. orange, on average): F I G U R E 3 Cutouts from photoluminescence (PL) images of a lower part of the host cells (see Figure 1) taken before and after the cleave process on the front side with higher and lower laser power. The cleave process has been applied for all five dividing lines. The results of this test are easiest to see between the two shingle cells a and b (region marked with dashed lines) because of the lack of front side busbar contacts.
ΔV OC = +2 mV and ΔpFF = +1.1% abs . The subsequent annealing step further improves the cell parameters for three of the four groups (not for the first group "T 1 ,8") and yields ΔV OC = +3 mV and ΔpFF = +1.3% abs compared with the values measured after TLS (green vs. orange). The ALD process "T 2 ,8" shows the largest mean gains after annealing (Δη = +0.5% abs , ΔV OC = +3 mV, and ΔpFF = +1.5% abs ) and thus has been chosen to be applied for edge passiv-  Table 1 in relation to the data for its host cell. This shingle cell has been processed with the "T 2 ,8" ALD Al 2 O 3 passivation recipe, and it was located within the host cell at position c. This means that this shingle cell features two cut edge surfaces.
It seems that cutting the host cells into shingle cells and applying the PET enables to even achieve larger pFF values for the edge-passivated shingle cells than for the host cells. One potential explanation for this observation can be that, in addition to the newly applied surface passivation on the cut edges, the already existing surface passivation in the cell area is also improved by the PET process steps. As is seen for the fifth group "Anneal" in  T A B L E 1 I-V data for the most efficient TOPCon shingle cell from position c (see Figure 1) in different processing states, expressed in each case as an absolute offset from its host cell I-V data.

Cell type
State show a mean pFF loss of À0.3% abs after 11 months, process "T 2 ,8"which has been chosen for experiment C-results in a mean pFF decrease of less than À0.1% abs during the same period. For the processes "T 1 ,14" and "T 2 ,14," the mean pFF decrease accounts to À0.1% abs and À0.2% abs , respectively. Thus, the ALD process for forming the Al 2 O 3 passivation layer influences not only the absolute performance of the shingle cells but also their long-term stability. Process "T 2 ,8" yields extremely promising PET long-term stability data.

| Experiment C: Comparison of separation techniques
As in the previous section, the I-V data for the TOPCon host and shingle cells are again shown relative to the specified reference groups in Figure 6A,B, respectively. The shingle cells in groups C 1 and C 2 have been separated by two different LSMC processes from the rear side. The shingle cells in group C 3  On average over the three groups C 1 to C 3 , the larger performance drop for the inner shingle cells after separation is Δη = À0.2% abs , ΔV OC = À2 mV, and ΔpFF = À0.9% abs .
The LSMC process in group C 1 leads to comparable efficient shingle cells as the LSMC process in group C 2 . On the other hand, the TLS process in group C 3 is superior to both and yields more efficient shin- After PET, a clear improvement is seen for all groups in Figure 6. The shingle cells in group C 1 , separated by LSMC, benefit the most from the PET (Δη = +0.5% abs for "C 1 ,2") followed by group C 3 , singulated by TLS (Δη = +0.4% abs for "C 3 ,2"). But also, for the LSMC separation in group C 2 , there is a clear improvement by the PET despite the partly very rough edge surface. The positive impact of the PET approach is therefore not limited to a certain separation process, as it shows a clearly efficiency-increasing effect in all cases.
It is remarkable that the PET leads to the fact that the I-V data difference in Figure 6B between the outer and inner shingle cells decreases in comparison with their larger difference directly after separation. The results suggest that the smoother the edge surface is, the lower is the difference between the outer and inner shingle cells after edge passivation by the PET approach: For C 3 and TLS (smoothest edge surface), the smallest I-V data difference is seen, whereas for C 2 and LSMC (roughest edge surface), the largest I-V data difference is seen.