Postmetallization “Passivated Edge Technology” for Separated Silicon Solar Cells

—This article introduces a postmetallization “passi-vated edge technology” (PET) treatment for separated silicon solar cells consisting of aluminum oxide deposition with subsequent annealing. We present our work on bifacial shingle solar cells that are based on the passivated emitter and rear cell concept. To separate the shingle devices after metallization and ﬁring, we use either a conventional laser scribing mechanical cleaving (LSMC) process or a thermal laser separation (TLS) process. Both separation processes show similar pseudo ﬁll factor (pFF) drops of − 1.2% abs from the host wafer to the separated state. The pFF of the TLS-separated cells increases by up to + 0.7% abs from the as-separated state after PET treatment due to edge passivation, while the pFF of LSMC-separated cells increases by up to + 0.3% abs . On cell level, the combination of TLS and PET allows for a designated area output power density of p out = 23.5 mW/cm², taking into account an additional 10% rear side irradiance.


I. INTRODUCTION
T HE REVIVAL of the shingling interconnection approach of solar cells [1] is an option to obtain higher photovoltaic module output power densities p out . Shingling of solar cells is done by interconnecting the rear busbar of a cell to the front busbar of a neighboring one. The overlap and the interconnection of the busbars lead to: (i) the removal of cell spacing and therefore to an increase of active cell area within the module, (ii) the decrease of shading losses due to the absence of visible busbars and interconnectors, and (iii) the decrease of electrical resistance losses on the interconnection level. Shingling solar cells have first been used for niche applications such as satellite devices [2], electronic devices [3], and electrical vehicle prototypes [4]. The potential shown by the shingling approach leads to the recent increase in interest not only shown in publications [5]- [8] Manuscript received September 2, 2019; revised November 9, 2019; accepted December 6, 2019. This work was supported by the German Federal Ministry for Economic Affairs and Energy within the research project "PV-BAT400" under Contract 0324145. (Corresponding author: Puzant Baliozian.) The and patents [9]- [11], but also in already existing commercially available shingle modules [12], [13]. Similar to half-size cells [14], shingle cells are usually separated from metallized and fired host wafers. The separation process leads to a decrease in solar cell efficiency due to edge recombination induced by the separation [15]. Edge recombination becomes even more significant for small-sized cells with higher perimeter-to-area ratios, leading to losses mainly in fill factor FF (initially caused by losses in pseudo fill factor pFF) in addition to losses in open-circuit voltage V OC .
Edge recombination can be reduced by: (i) the depletion of charge carriers of a kind from the edge and/or (ii) the reduction of density of traps at the edge. As review of potential methods for edge passivation, we find that some previously investigated methods to reduce edge recombination of solar cells consist of introducing an emitter window commonly created by the use of diffusion barriers [16], [17]. The emitter window cuts off the emitter conduction towards the edge, thus reducing the support of minority charge carriers tremendously. Having the same aim, the removal of the emitter by creating passivated isolation trenches is proposed as another approach [18]. In this method, the trenches are formed by ablating the emitter, followed by wet-chemical etching, and finally, passivating the trenches by thermally grown silicon dioxide (SiO 2 ) or polysilicon. In Ref. [19], a strong doping at the separation path is suggested to create a repulsion of carriers from the edge by a surface field effect. It has also been reported that treating the edge wetchemically allows the growth of SiO 2 on the edge, which shows a passivation effect [20]. In Ref. [21], the surface passivation of two sawed slits tens of micrometers far from the cell active area is achieved by the deposition of aluminum oxide, capped by silicon nitride, and silicon oxide layers. The process is done after partially damage etching the slit surfaces.
However, fabricating such cells requires several additional premetallization process steps [22] or postmetallization chemical etching processes, which make the implementation on industrial fabrication scale challenging. Simulation results confirm the positive effect of passivated edges on shingle cells by reducing the effective surface recombination velocity at the edge, S eff,edge [23], [24]. Nevertheless, no potentially industrial feasible process is found in literature.
This article introduces the concept of postmetallization "passivated edge technology" (PET). The PET is demonstrated on bifacial p-type silicon shingled passivated edge, emitter, and rear (pSPEER) solar cells, which have been initially reported without PET in Ref. [25]. The bifacial pSPEER cell concept is based on the passivated emitter and rear cell (PERC) [26] architecture, where an additional laser-assisted separation process is applied to obtain the cell stripes. The separation processes investigated in this article are conventional laser scribing and mechanical cleaving (LSMC) and thermal laser separation (TLS) [27], [28]. The effect of the separation processes and the edge passivation are discussed together with the first solar cell results.

A. pSPEER PET Solar Cell Concept
The separation process leads to an increase of edge recombination, especially when the emitter extends to the separated edges; see Fig. 1(a) for a schematic cross section of the pSPEER cell. The cells only have a natively grown SiO 2 layer at the edge. For that reason, an additional postmetallization/separation PET aims to decrease S eff,edge by coating the edges with an intended dielectric passivation layer. The "pSPEER PET " solar cell is obtained after edge passivation by the PET; see Fig. 1 In this article, the PET consists of two steps: a postmetallization/separation deposition of an aluminum oxide (AlO X ) layer and a postdeposition annealing (PDA) for the activation of the deposited AlO X passivation layer. Thermal atomic layer deposition (ALD) is used for the AlO X deposition. This has several advantages: (i) ALD is known for a very conformal deposition, e.g., the wafer edge is well coated after the process; (ii) AlO X can provide an excellent surface passivation because it combines enhanced chemical passivation with a strong field effect passivation induced by a high amount of fixed negative charges [29]; (iii) it is transparent for very thin films of thicknesses d ranging between 5 nm ≤ d ≤ 15 nm [30]. OPTOS simulations done on module level confirm that such thin AlO X layers deposited on the front and rear surfaces of the cells do not induce optical losses; (iv) the passivation effect of thermal ALD-deposited AlO X is known to be activated at low annealing temperatures T ann < 225°C [31], [32]. The latter hinders thermally induced metal contact degradation which has been reported [33].

B. Solar Cell Fabrication
Industrial 6-inch gallium-doped Czochralski-grown silicon (Cz-Si:Ga) PERC precursors are used for the fabrication of the pSPEER and pSPEER PET cells investigated in this article. Fig. 2 shows the corresponding fabrication process. The precursors' base resistivity ranges between 0.3 Ω cm ≤ ρ B ≤ 0.9 Ω cm. The front phosphorous-doped emitter is passivated by a silicon nitride (SiN X ) layer; the rear silicon base is passivated by a layer stack consisting of AlO X capped by SiN X . The precursors are optimized for monofacial use, which explains the yellowish color of the rear side; see Fig. 3(a). The fabrication process performed in this article starts from the laser contact openings. The rear side silver busbars (external contacts) are screen printed first. Next, the rear-side aluminum contact grid and the front-side silver contact grid are printed. Contact firing is then performed in an industrial fast firing oven. The printed metallization layouts are designed to obtain six shingle devices from one host   Fig. 3(c).
The performed separation processes in this article are (i) conventional LSMC and (ii) TLS, both implemented by using the microDICE tool from 3D-Micromac [34]. The following items describe the processes: 1) LSMC is done by a nanosecond infrared pulsed laser that scribes along the entire separation path and ablates up to one-third of the substrate's thickness. Consequently, the samples are manually and mechanically cleaved. 2) TLS is performed by creating an initial scribe by means of the same pulsed laser used in the LSMC process to start a crack of length between 0.5 mm ≤ l ≤ 1 mm. For the cleave process, an infrared continuous wave laser in combination with a simultaneous water/air cooling jet is used to thermally induce a compressive stress and consequent tensile stress along the separation path. This leads to the cleavage of the substrate. Planar view scanning electron microscope (SEM) images of exemplary cell edges for LSMC and TLS processed samples are shown in Fig. 4(a) and (b), respectively. The edge obtained by LSMC has a rough laser-scribed region which is about one-third of its thickness. The mechanically cleaved part of the edge is rather smooth. In contrast, the cell edge obtained by the TLS process shows an entirely smooth surface.

C. SunsV OC and Current-Voltage Characterization
In this article, SunsV OC measurements [36], [37] are performed to characterize the effect of the separation and passivation processes. By this method, the effect of edge recombination on the solar cell performance is independent of possible changes   identical way. A reference cell not undergoing any process steps is measured continuously throughout the experiments to track the reproducibility of the SunsV OC measurement.
The illuminated current-voltage (IV) measurement is done at standard test conditions on each side separately using an industrial cell tester calibrated by a reference pSPEER cell measured at Fraunhofer ISE CalLab PV Cells. During the IVtesting procedure, the cells are placed on a black, nonconducting background, which ensures negligible reflection well below 5% over the relevant spectrum. The contacting is performed at the front-and rear-side busbars by pin arrays.

A. Experimental Plan for PET on Cells
The aim of this experiment is to observe the effect of the separation processes from host wafer to separated state by using SunsV OC measurements; see the experimental process flow in Fig. 6. In addition, the effect of the AlO X deposition and subsequent annealing on cells is examined. LSMC is done 6 h before AlO X deposition. The TLS-diced host wafers are separated also 6 h prior to the AlO X deposition, while another group is TLS-diced 19 h prior to the AlO X deposition. The different separation dates are to investigate the influence of the waiting time between separation and deposition. In air at room temperature, native SiO 2 growth occurs on exposed silicon [38]. Due to the logarithmic growth of SiO 2 as a function of exposure time to air, similar layer thicknesses are expected for both waiting times in this article.
The AlO X deposition process is performed in a FlexAL reactor of Oxford Instruments [39] using trimethylaluminum and water vapor as precursors. Two deposited AlO X layer thicknesses are investigated: d 1 = 7 nm and d 2 = 14 nm, as measured with spectroscopic ellipsometry on a planar silicon process control sample. Three to four separated cells have been processed in each deposition run (see Fig. 7). Height spacers are used to ensure a deposition on the rear side as well.
SunsV OC measurements before and after deposition are completed by contacting at five measurement positions per busbar. Postdeposition hotplate annealing (PDA) is performed at temperature T ann < 225°C. Control cells (without deposition) from each separation process are kept and are not coated with AlO X To understand the effect of the deposition and annealing processes and estimate the effective surface recombination velocity S eff , a symmetrical lifetime sample (n-type float-zone silicon, Quasi-steady-state photoconductance (QSSPC) measurements [40] are performed to obtain the minority carrier lifetime τ eff dependency on the excess carrier density Δn at the following states: (i) as-deposited and (ii) annealed. S eff is then approximated by using [41] 1 / τ eff = (1 / τ bulk ) + (2S eff / W ) . (1) Noting that τ bulk is the intrinsic bulk lifetime and τ eff is extracted at Δn = 10 15 cm − 3 .

A. Effect on Charge Carrier Lifetime
The QSSPC measurement of the lifetime sample in Fig. 8 shows the injection-dependent lifetime obtained by means of QSSPC for the as-deposited and annealed states. After annealing, τ eff = 432 μs at Δn = 10 15 cm − 3 is measured.
Using (1), S eff = 22 cm/s is obtained, showing the possibility of attaining low S eff values by the current deposition/annealing method. Note that the planar S eff is not necessarily equal to that of the edge due to different surface and geometrical conditions.

1) Effect of PET:
From the SunsV OC measurements, we extract V OC values at an illumination intensity of 1000 W/m² and the pFF. The host wafers attain mean V OC = 669 mV. The changes in V OC throughout the process steps are minor, with a maximum variation of ΔV OC = ± 2 mV. This small impact on V OC is in agreement with simulation results for similar sample types and V OC levels [23], [42].  Hence, we focus on the pFF data which are shown in Fig. 9. The values are given for the four different sample states: (i) host wafer, (ii) separated, (iii) as-deposited, and (iv) annealed. They are divided into the different separation groups and AlO X layer thicknesses. The results are discussed considering the state-tostate changes.
To start with, LSMC and TLS separation processes lead to a decrease in pFF (ΔpFF = − 1.2% abs ) from the host wafer to the separated state due to similar edge recombination effects.
Noticeably, the introduction of the AlO X layer with thickness d 1 or d 2 leads to an improved edge passivation and thus to a gain in pFF. The deposition on TLS-separated cells leads to the highest increase from separated to as-deposited state reaching ΔpFF = +0.4% abs .
After annealing, the TLS-separated and coated cells show the highest increase, ΔpFF = +0.4% abs from as-deposited to annealed states. Such an increase after PDA is due to the activation of the passivation layers that coat the smooth edge surface. Annealing of control cells without deposition leads to an enhanced pFF. This is probably due to the grown SiO 2 on the edges, which is allowed by the absence of the AlO X coating. As an assessment of PET, the TLS-separated cells show higher pFF gains than the LSMC-separated ones when both PET process steps (deposition and PDA) are completed. The deposition on a smooth surface with reduced defects obtained by TLS and the activation of the deposited passivation layer lead to the highest pFF increase of +0.7% abs and the highest measured pFF values. TLS-separated cells regain up to 50% rel of their initial pFF loss due to the separation process (observed for both deposition thicknesses and separation groups), and control cells that are not coated and are only annealed do not show such a regain. LSMC-separated cells regain around 16% rel after PET. This shows that PET leads to edge passivation for both separation processes; however, a larger effect is recorded for TLS-separated cells.
2) Edge Passivation Stability: Based on the pFF results, the passivation stability investigation (see Fig. 10) shows that the cells coated with d 2 = 14 nm remain almost constant with minimal (ΔpFF ≈ 0.1% abs ) to no measured decrease. These results show a stable edge passivation over time for the cells coated with d 2 = 14 nm AlO X , while cells with d 1 = 7 nm or without degrade with time. The explanation for the time-based passivation stability on the deposited layer thickness is under further investigation.

C. IV Measurement Results
Based on the results of the previous section, the favored processes for the fabrication of edge-passivated cells are used to fabricate cells for IV measurements. Host wafers are separated into shingle cells by means of TLS. The cells are then coated using an identical thermal ALD deposition process to the one used before. This time this leads to around 13-nm-thick postmetallization AlO X layer. The cells are then annealed on a hotplate.
IV measurements of TLS-separated edge-passivated pSPEER PET solar cells are done in comparison to LSMC-separated pSPEER solar cells that do not undergo any further processes after separation. The front-and rear-side IV measurement data from the most efficient cells are summarized  TABLE I  IV MEASUREMENT DATA FOR THE CELLS WITH HIGHEST OUTPUT POWER  DENSITIES FOR TLS+PET-PROCESSED AND LSMC-SEPARATED CELLS IN  ADDITION TO AVERAGE VALUES OF THE FRONT-SIDE MEASUREMENTS Characteristic data for the front-and rear-side measurements are shown. Front-and rear-side measurements are separately performed at G = 1000 W/m 2 . Since the busbar is intended to be covered in the module, a designated area j SC,des excluding the busbar area is considered. p out considers an additional 10% irradiance from the rear side.
in Table I. For the front-side measurements, average values (Av.) are also shown. The TLS-separated cell after PET treatment yields a designated front-side energy conversion efficiency η f = 22.1%, featuring V OC = 669 mV, FF = 81.4%, and short-circuit current density j SC,des = 40.5 mA/cm 2 . Due to the PET treatment, the cell features a high pFF = 83.2%. Even though PET is included as postfiring thermal process, a low r S = 0.38 Ω cm 2 is achieved, hinting that no significant degradation at the metal contacts occurs. The LSMC-separated cell without PET attains η f = 21.7%. In comparison to the pSPEER cell, the pSPEER PET cell features a higher front-side efficiency Δη f = +0.4% abs , which can be attributed to the higher short-circuit current density Δj SC = +0.2 mA/cm 2 and ΔFF = +0.5% abs . This difference in FF can be explained by the higher pFF of +0.8% abs , while featuring a slightly higher r S . Mainly, the improved pFF is a result of an enhanced edge quality obtained by the TLS process and the additional PET processes.
The rear-side measurement results also show similar trends, where the pSPEER PET cell attains a designated rear-side energy conversion efficiency η r = 14.7% that is +0.3% abs higher than the LSMC-separated pSPEER cell that attains η r = 14.4%. The difference in ΔpFF = +0.8% abs is entirely projected on the ΔFF = +0.8% abs due to the identical rear side r S = 0.35 Ω cm 2 between the two compared cells. The TLS-separated and edgepassivated solar cell attains a bifaciality β = η r /η f = 0.67, whereas the LSMC-separated cell without PET attains β = 0.66.
Considering an additional 10% rear-side irradiance G r = 100 W/m 2 , the edge-passivated cell attains an output power density p out = 23.5 mW/cm 2 . This value is greater by Δp out = +0.4 mW/cm 2 in comparison to the nonpassivated LSMC-separated cell.
The result of the TLS + PET processed cell shows its clear advantage in terms of efficiency. Therefore, we have demonstrated the functionality of the proposed PET postmetallization edge passivation approach. No changes in the premetallization processes in an industrial production line are required. The integration of PET into the fabrication after the separation of silicon solar cells (i.e. shingle cells, half-cut cells, or small-area cells) can, thus, lead to a significant boost in cell efficiency with a relatively reasonable integration within the already existing production infrastructure. This makes the proposed PET attractive for industrial application and incorporation in production lines.

V. CONCLUSION
This article introduces a postmetallization/separation PET for separated solar cells. The effect of PET on bifacial p-type silicon shingled passivated edge, emitter, and rear solar cells (pSPEER) is tested. After edge passivation, we call the PET-treated shingle cells "pSPEER PET " solar cells.
The TLS process leaves the separated cells with visibly smoother edges in comparison to separation by LSMC. By just considering the separation processes without further PET, the separation processes lead to similar drops in pseudo fill factor pFF of ΔpFF ≈ − 1.2% abs .
For the PET deposition process, low-temperature thermal ALD of AlO X for two layer thicknesses d 1 = 7 nm and d 2 = 14 nm is performed. TLS-separated PET-treated pSPEER PET cells have regained half the loss that is induced due to separation. The pFF increase from as-separated TLS-cut cells to edge-passivated cells is found to be up to +0.7% abs , while LSMC-separated cells gain up to +0.3% abs after PET. Thereby, the cells with d 2 = 14-nm-thick AlO X layer show constant pFF values within the experimental time frame of 167 h after passivation, thus indicating the stability of the passivation.
After TLS and PET, the best pSPEER PET solar cell attains a designated front-side energy conversion efficiency η f = 22.1%. Assuming an additional rear-side irradiance G r = 100 W/m 2 , this cell features a total output power density p out = 23.5 mW/cm 2 . TLS followed by PET leads to a higher output power density of Δp out = +0.4 mW/cm² compared to a PET-free LSMC-separated cell. The PET treatment introduced in this article shows a clear benefit and the thus processed and separated solar cells achieve higher p out values.
The proposed PET is a promising postmetallization edge passivation treatment that might be realized in the industry without requiring adjustments in premetallization stages.