Analysis of Thin Bifacial Silicon Solar Cells with Locally Diffused and Selective Back Surface Field

In the last decade, significant cost reductions in the kWh obtained from photovoltaic modules were related to the “economy-of-scale” benefits, the solar cell efficiency increase and the lower consumption of high purity silicon per Wp. For instance, wafer thickness was reduced from 350 mm to 180 mm. However, the silicon wafer still represents 58% of the cell cost and 42% of the PV module production costs1. The use of thinner wafers can lead to cost reduction, but high quality surface passivation is needed to achieve high efficiency devices2. Under the point of view of industrial processing, equipments need to be adapted for handling thin wafers3 and the standard n+pp+ silicon solar cells, with the rear surface fully covered by the Al thick layer, have to be modified in order to avoid the wafer bowing4,5. The Al layer forms a back surface field that reflects the minority charge carriers away from the rear contact, reducing the recombination. The bow is produced during the thermal step performed to diffuse Al into a silicon wafer and it is a result of the different thermal coefficients of expansion of aluminum layer and silicon. The wafer bowing can lead to cracking during soldering and lamination processes used to assemble the PV modules. In order to avoid the wafer bowing, solar cells with a rear metal grid and surface passivation based on dielectric films have been studied. For instance, Lee et al.6 produced large area PERC (passivated emitter rear cell) solar cells and achieved the efficiency of 17.7% by using 148.6 cm2 FZ (float-zone) Si wafers and the efficiency of 17.2% in 156.8 cm2 multicrystalline Si wafers. The wafers used had a thickness of 130 mm and the rear side was passivated with the SiOx/SiNx/SiOxNy stack deposited by PECVD (plasma-enhanced chemical vapor deposition) and also by implementing the local back contacts using a laser. Front emitter was doped with phosphorus diffused in a quartz tube furnace by using POCl3 as source. With the same goal of reducing bowing, Gu et al.5 implemented a BSF (back surface field) region by using a thin layer composed by boron and aluminum pastes deposited by screen-printing. The authors reported that the metallization paste thickness was reduced while the efficiency was not diminished. Therefore, thinner layer of doping paste fired at temperatures up to 800 °C can help the development of low bowing silicon solar cells. Bifacial cells have been studied since the 1960s years7. This kind of cell is active in both faces and can convert the radiation reflected by the surrounding areas8. The bifacial cells also can be applied in static concentrators with different designs9,10,11. Several bifacial structures were developed by using silicon as starting material7,12. Simplified bifacial cells are based on the standard monofacial structure with one pn junction, one BSF region and the metal grid on both faces. This way, the use of a metal grid on the rear face avoids the bow and thinner wafers can be used. To obtain high electric current when the solar cell is illuminated by BSF region, the minority charge carrier diffusion length needs to be larger than the wafer thickness. Thus, the carriers photogenerated far from pn junction could be collected. Taking into account similar front and rear passivation, minority carrier diffusion length has to be over twice the wafer thickness to allow the production of symmetric bifacial cells, that is, cells with similar short-circuit electric current for front and rear side illumination13,14. High efficiency bifacial cells based on n+pp+ structure and using standard 200 mm thick Cz-grown (Czochralski) Analysis of Thin Bifacial Silicon Solar Cells with Locally Diffused and Selective Back Surface Field


Introduction
In the last decade, significant cost reductions in the kWh obtained from photovoltaic modules were related to the "economy-of-scale" benefits, the solar cell efficiency increase and the lower consumption of high purity silicon per Wp. For instance, wafer thickness was reduced from 350 mm to 180 mm. However, the silicon wafer still represents 58% of the cell cost and 42% of the PV module production costs 1 . The use of thinner wafers can lead to cost reduction, but high quality surface passivation is needed to achieve high efficiency devices 2 . Under the point of view of industrial processing, equipments need to be adapted for handling thin wafers 3 and the standard n + pp + silicon solar cells, with the rear surface fully covered by the Al thick layer, have to be modified in order to avoid the wafer bowing 4,5 . The Al layer forms a back surface field that reflects the minority charge carriers away from the rear contact, reducing the recombination. The bow is produced during the thermal step performed to diffuse Al into a silicon wafer and it is a result of the different thermal coefficients of expansion of aluminum layer and silicon. The wafer bowing can lead to cracking during soldering and lamination processes used to assemble the PV modules.
In order to avoid the wafer bowing, solar cells with a rear metal grid and surface passivation based on dielectric films have been studied. For instance, Lee et al. 6 produced large area PERC (passivated emitter rear cell) solar cells and achieved the efficiency of 17.7% by using 148.6 cm 2 FZ (float-zone) Si wafers and the efficiency of 17.2% in 156.8 cm 2 multicrystalline Si wafers. The wafers used had a thickness of 130 mm and the rear side was passivated with the SiO x /SiN x /SiO x N y stack deposited by PECVD (plasma-enhanced chemical vapor deposition) and also by implementing the local back contacts using a laser. Front emitter was doped with phosphorus diffused in a quartz tube furnace by using POCl 3 as source. With the same goal of reducing bowing, Gu et al. 5 implemented a BSF (back surface field) region by using a thin layer composed by boron and aluminum pastes deposited by screen-printing. The authors reported that the metallization paste thickness was reduced while the efficiency was not diminished. Therefore, thinner layer of doping paste fired at temperatures up to 800 °C can help the development of low bowing silicon solar cells.
Bifacial cells have been studied since the 1960s years 7 . This kind of cell is active in both faces and can convert the radiation reflected by the surrounding areas 8 . The bifacial cells also can be applied in static concentrators with different designs 9,10,11 . Several bifacial structures were developed by using silicon as starting material 7,12 . Simplified bifacial cells are based on the standard monofacial structure with one pn junction, one BSF region and the metal grid on both faces. This way, the use of a metal grid on the rear face avoids the bow and thinner wafers can be used. To obtain high electric current when the solar cell is illuminated by BSF region, the minority charge carrier diffusion length needs to be larger than the wafer thickness. Thus, the carriers photogenerated far from pn junction could be collected. Taking into account similar front and rear passivation, minority carrier diffusion length has to be over twice the wafer thickness to allow the production of symmetric bifacial cells, that is, cells with similar short-circuit electric current for front and rear side illumination 13,14 .
High efficiency bifacial cells based on n + pp + structure and using standard 200 mm thick Cz-grown (Czochralski) silicon wafers were reported by Yang et al. 15 . Homogeneous n + and p + regions were obtained by P and B diffusion performed at 870 °C and 1020 °C, respectively. POCl 3 and BBr 3 liquid sources were used in the diffusion processes. SiN x antireflection/passivation coating was deposited on both faces as well as the screen printed metal grid. The large area devices (149 cm 2 ) achieved the front and rear efficiencies of 16.6% and 12.8%, respectively. Janben et al. 16 also used p-type Cz-Si and screen-printed contacts, but instead to implement a uniform p + region on the rear face, an Al metal grid was deposited to produce a local BSF. Front and rear surfaces were covered with an a-SiN x :H layer deposited by PECVD. The front and rear efficiency of 17% and 10.3%, respectively, were reported 15 .
Standard p + nn + structure with homogenous doped regions was applied to produce small (4 cm 2 ) FZ and Czsilicon solar cells, achieving efficiencies of 19.1% /18.1% (front / rear illumination) and 17.7% / 15.2%, respectively. The wafers employed were 250 mm thick and the metal grid was defined by photolithography and deposited by e-beam evaporation at high vacuum 17,18 . Buck et al. 19 obtained efficiencies of 15.9% / 13.4% (front/rear illumination) in large area (144 cm 2 ) p + nn + devices fabricated by using FZ silicon wafer (200 mm thick) and with metal grid deposited by screen-printing. They also used BBr 3 and POCl 3 as diffusion sources to form p + and n + homogeneously doped regions.
The thin silicon bifacial cells have been developed by implementing n + pp + , p + nn + or n + p structures. In small area devices (4 cm 2 ) with evaporated contacts, front and rear efficiencies of 17% / 14.9% and 16% / 13% were reported for p + nn + and n + pp + cells, respectively 20 . Homogenous n + and p + regions were obtained by P and B diffusion in 140 mm thick Cz wafers. Using screen printed emitter and BSF as well as evaporated contacts, Pérez et al. 21 reported small area bifacial devices that achieved the efficiencies of 14.2% / 13.6%. When screen-printed metal grids were implemented, efficiencies were lower: 11.6% / 10.8% for n + pp + cells and 8.3% / 8.0% for p + nn + ones. Bifacial n + np + cells with a thickness of 130 mm were developed by using a fully industrial in-line process 22 . The n + and p + regions were implemented by screen-printing deposition of phosphorus and boron pastes and the dopants were co-diffused into silicon wafers in the belt furnace. Metal grids were also deposited by screen-printing. Small area (4 cm 2 ) devices achieved the front/rear efficiency of 13.6% / 10.4% 22 . Recart 23 developed larger bifacial devices (24.7 cm 2 ) in 120 mm thick Cz wafers by using screen-printed boron pastes to obtain p + BSF, but n + emitter was formed by phosphorus diffusion utilizing POCl 3 as source. The n + pp + solar cells reached the efficiencies of 14.3% and 10.8% for front and rear side illumination, respectively. Thin bifacial cells with n + front emitter and the rear surface passivated with a SiN x layer was presented by Steckemetz et al. 24 . The SiN x layer produced an effective passivation on the rear face and 14.6% / 13.6% (front/rear illumination) efficient devices were reported. Wafer thickness was 140 mm and devices had the area of 4 cm 2 .
The aim of this paper is to present the development and comparison of thin n + pp + bifacial cells produced by using the locally aluminum doped back surface field and by using the selective region doped with aluminum and boron. The process to obtain the selective p + region was based on boron deposition by spin-on and diffusion in the standard quartz tube followed by Al grid screen-printing and diffusion in a belt furnace. As far as we know, the combination of both processes to accomplish the selective BSF was used for the first time to produce thin bifacial solar cells. The devices were fabricated on solar grade Cz silicon wafers with the thickness of around 150 mm by using industrial techniques. Surface passivation of a SiO 2 thin layer was also analyzed.

Materials and Methods
The starting material was Cz-silicon, p-type, boron doped, 1-20 Ω.cm, <100> orientation. The thickness of the wafers was of around 150 mm. The saw damage of the wafers was removed by immersing them in a KOH solution and the etching time defined the thickness of the wafers. Texture etch was carried out in an alkaline solution also based on KOH, but with lower concentration.
To develop the n + pp + solar cells with locally diffused Al BSF, the process sequence shown in Figure 1a was used. Wafers were cleaned in RCA standard solutions and a 100 nm thick SiO 2 layer was grown at 1000 °C in a quartz tube furnace. Photoresist was spin coated on one face and the oxide layer was etched away in a buffered HF solution. After cleaning the wafers, the phosphorus diffusion was accomplished in a quartz tube furnace at 875 °C with POCl 3 as source. Phosphorus profile was measured by ECV (electrochemical capacitance-voltage) technique. The diffusion processes formed the n + emitter with a sheet resistance of 32 Ω/□, with a junction depth of around 0.6 mm and surface concentration (C S ) of approximately 3.3×10 19 cm -3 . Phosphorus silicate glass (PSG) and SiO 2 layer were removed in an HF bath and wafers were cleaned in RCA solutions. A 10 nm thick SiO 2 layer was grown at 800 °C in order to passivate the surfaces. Cells without oxide were also processed to evaluate the passivation effectiveness of the SiO 2 layer. The titanium dioxide antireflection coating (ARC) was deposited by e-beam evaporation at high vacuum on the front face and the metal grid was deposited by screenprinting on both faces. In the first batches, ARC was not deposited on the rear face in order to avoid the misleading effects related to the TiO 2 etching process performed during the firing of Al paste. The silver paste was deposited on the front face and the Al paste was screen-printed on the rear face. The metal pastes were dry and co-fired in a belt furnace. The diffusion/firing temperature (T Firing ) used for cells without oxide layer was 840 °C and for those with a 10 nm thick SiO 2 layer was 850 °C 25 . The metal grid was a standard H-pattern with two busbars, covering 9.5% of the surfaces (front and rear). Pseudo-square cells (80 mm × 80 mm) were obtained after performing the edge isolation step.
The n + pp + cells with selective BSF region doped with boron and Al were processed as summarized in Figure 1b. After the texture etch and the RCA cleaning, p + region was obtained by boron spin-on deposition and diffusion in a quartz tube furnace at 1000 °C. Boron diffusion was performed before phosphorus in order to avoid the production of a very deep pn junction and thicker dead layer, which occurs when phosphorus is diffused in supersaturation conditions. The oxide layer needed to mask the p + region from the phosphorus diffusion was grown in the same thermal step of the boron diffusion. Photoresist was spin coated on the boron-doped face and the oxide layer was etched away in a buffered HF solution. Phosphorus was diffused at 875 °C (POCl 3 as source). Phosphorus and boron silicate glasses (PSG and BSG) were etched away by means of HF solution and RCA cleaning was performed. Silicon dioxide layer was grown and TiO 2 ARC was deposited. Ag and Al metal grids were screen-printed on the front and rear face, respectively, and were co-fired in the belt furnace. The diffusion/firing temperatures were the same used for cells with local Al-BSF. Boron diffusion resulted in the doped region with a thickness of around 1 mm, surface concentration of 3.6×10 19 cm -3 and sheet resistance of 30 Ω/□. The n + emitter presented the sheet resistance of approximately 32 Ω/□.
All the devices were characterized under standard conditions (100 mW/cm 2 , AM1.5G and 25 °C) in a solar simulator calibrated with silicon solar cells previously measured at CalLab -FhG-ISE (Fraunhofer-Institut für Solare Energiesysteme), Germany. Two-dimensional distribution of minority charge carrier lifetime (τ) was obtained by m-PCD (microwave photoconductivity decay) with the WT1000-PV device of Semilab. Minority charge carrier diffusion length (L D ) was calculated from LBIC (light beam induced current) measurements performed with the WT1000-PV equipment. Table 1 presents the open circuit voltage (V OC ), the shortcircuit current density (J SC ), the fill factor (FF), the efficiency (η) and the current symmetry factor (CSF) of the bifacial solar cells with local Al-BSF. The average values were calculated from the results of at least three cells. Quality of bifacial cells can be evaluated by the current symmetry factor CSF (also called "bifaciality") defined as the ratio of the short-circuit current when the cell is illuminated on the rear face to the one when the cell is illuminated on the front face.

Results and Discussion
The solar cells with SiO 2 passivated surfaces presented an average V OC of 9-10 mV higher than that from unpassivated ones. The short-circuit current was increased by implementing SiO 2 film when the cell was illuminated  on the front face. The CSF was very low for uncovered and SiO 2 covered solar cells. Thus, the rear passivation was not effective and/or the minority carrier lifetime was very low, resulting in the minority carrier diffusion length smaller than the wafer thickness. The Figure 2 shows the two-dimensional distribution of the minority charge carrier diffusion length in passivated and unpassivated devices. The bifacial solar cells without SiO 2 layer presented an average diffusion length of 70 mm. In oxide-passivated devices, the average L D was of around 90 mm, also lower than the wafer thickness. Table 2 summarizes the electrical characteristics of the best solar cells produced with local BSF. In order to assess the values of minority carrier lifetimes, front surface recombination velocity (S f ), rear surface recombination velocity (S r ) and specific series resistance (r s ), the bifacial cells were simulated by using the PC1D computer program 26 and these parameters were adjusted to fit in well the simulated V OC , J SC , FF and η to the experimental ones. Bearing in mind that Al-BSF covers 9.5% of the rear surface, an effective surface recombination velocity (S r-ef ) was considered by using the method presented by Lago-Aurrekoetxea 27 . The surface recombination velocity was set in 1×10 7 cm/s for the surfaces not covered by Al or SiO 2 . The following parameters were set to obtain the results presented in Table 2 for the cells without oxide: S f = 1×10 7 cm/s, S r-ef = 9.05×10 6 cm/s, τ = 10 ms and r s = 2.8 Ω.cm 2 . When the cells were passivated with the thermally-grown SiO 2 layer, the parameters used in the simulations that led to the best I-V (electric current-voltage) curve fitting were: S f = 1×10 5 cm/s, S r-ef = 2.5×10 4 cm/s, τ = 10 ms and r s = 2.8 Ω.cm 2 . Minority carrier lifetime was very low and high phosphorus doped emitter normally presents a high recombination velocity of around 10 5 cm/s, even if surfaces are SiO 2 coated. The combination of low minority carrier lifetime and high surface recombination led to the low current symmetry factor of these bifacial cells.
In order to evaluate the contamination and gettering mechanisms during the process, minority carrier lifetime (or bulk lifetime) was measured after each processing step and the results are shown in the Figure 3 and   CP4 one (HNO 3 :HF:CH 3 COOH). During the measurements, samples were immersed in an iodine+ethanol solution to passivate the surfaces. The metallization comprises three thermal steps in belt furnace, two of them for drying of the metal pastes (from 150 °C to 300 °C) and one step for firing these pastes (840-850 °C). To analyze the effect of this fabrication step on bulk lifetime, wafers were thermally processed without the metal grid. The results showed that bulk lifetime was enhanced by P gettering and the metallization step degraded the minority carrier lifetime. In the Table 3 are summarized the results obtained from n + pp + cells with selective BSF formed by boron diffusion over the whole rear surface and by Al locally diffused. Comparing the cells with local Al-BSF (Table 1) to those with Al/B-BSF, we can observe that the current symmetry factor was increased from 0.14 to 0.73 for passivated devices and from 0.16 to 0.64 in unpassivated cells. Oxide layer did not improve the cell output parameters.
The electrical parameters of the best solar cells with selective Al/B-BSF are presented in Table 4. The most efficient device presented the efficiency of 13.7% under front illumination (pn junction) and 8.9% under rear illumination (BSF region). Results from similar devices with smaller area (24.7 cm 2 ) were reported by Recart 23 , achieving front efficiency slightly higher, but with the busbars outside the active area. The rear efficiency was 1.9% (absolute) higher than the obtained in this work because ARC was used.
Deposition of the ARC on the rear face can enhance the J sc of around 2 mA/cm 2 and the efficiency can be improved to 10%.
The PC1D computer program was used for modeling these cells and the internal parameters such as τ, S f , S r and r s were adjusted to fit simulation results to the experimental ones. For passivated cells, input parameters set to simulate the cell performance were: S f = 7×10 5 cm/s, S r = 7×10 5 cm/s, τ = 10 ms and r s = 3.4 Ω.cm 2 . To model the devices with nonpassivated surfaces, parameters used were: S f = 1×10 7 cm/s,   S r = 1×10 7 cm/s, τ = 20 ms and r s = 2 Ω.cm 2 . The thin SiO 2 layer did not produce an effective surface passivation for cells with Al/B-BSF. Likewise the solar cells with local-Al BSF, the bulk lifetime, between 10 ms to 20 ms, used to simulate the solar cell with selective BSF, was also very low. The average L D obtained from LBIC measurements reached values as high as 460 mm (devices without SiO 2 ) and 500 mm (oxide-passivated cells), as shown in Figure 5. The high L D was not due to the high bulk lifetime, but due to the boron BSF that reduced the recombination rate on the rear face and it was considered in the estimation of L D from LBIC measurements.

Conclusions
Local Al-doped BSF and selectively doped Al/B-BSF were implemented in thin bifacial silicon solar cells.
Concerning the solar cells with locally doped BSF, the analysis of the minority charge carrier lifetime showed that phosphorus diffusion produced gettering, but due to the contamination in belt furnace, bulk lifetime decreased and the final value was similar to the initial. Besides, the minority carrier diffusion length calculated from LBIC measurements was smaller than the wafer thickness. In addition, the silicon dioxide growth at 800 °C did not result in an effective surface passivation. The combined effect of low bulk lifetime and unpassivated rear surface led to the small CSF, lower than 0.16. Therefore, the structure was not suitable for bifacial devices.
The solar cells with a selective Al-B BSF presented higher current symmetry factors of around 0.70. Although the silicon dioxide layer established a poor rear surface passivation, the BSF in whole face allowed the achievement of average minority carrier diffusion lengths three times higher than the wafer thickness. The most efficient bifacial cell was fabricated without SiO 2 layer and achieved an efficiency of 13.7% and 8.9% for front (n + face) and rear (p + face) illumination, respectively, the latter without ARC. The simulation of this cell with ARC on the rear face showed an improvement of around 2 mA/cm 2 in the J SC and, consequently, the efficiency could rise to approximately 10%. The results obtained in this work were similar to that reported to bifacial devices with small area, but larger solar cells with spin-on boron deposition were developed and busbars were considered inside active area.
To sum up, thin bifacial solar cells need a selective BSF region when high quality surface passivation cannot be implemented in the industrial process.