Realizing a facile and environmental-friendly fabrication of high-performance multi-crystalline silicon solar cells by employing ZnO nanostructures and an Al2O3 passivation layer

Nowadays, the multi-crystalline silicon (mc-Si) solar cells dominate the photovoltaic industry. However, the current acid etching method on mc-Si surface used by firms can hardly suppress the average reflectance value below 25% in the visible light spectrum. Meanwhile, the nitric acid and the hydrofluoric contained in the etching solution is both environmental unfriendly and highly toxic to human. Here, a mc-Si solar cell based on ZnO nanostructures and an Al2O3 spacer layer is demonstrated. The eco-friendly fabrication is realized by low temperature atomic layer deposition of Al2O3 layer as well as ZnO seed layer. Moreover, the ZnO nanostructures are prepared by nontoxic and low cost hydro-thermal growth process. Results show that the best passivation quality of the n+ -type mc-Si surface can be achieved by balancing the Si dangling bond saturation level and the negative charge concentration in the Al2O3 film. Moreover, the average reflectance on cell surface can be suppressed to 8.2% in 400–900 nm range by controlling the thickness of ZnO seed layer. With these two combined refinements, a maximum solar cell efficiency of 15.8% is obtained eventually. This work offer a facile way to realize the environmental friendly fabrication of high performance mc-Si solar cells.


Results and Discussion
Optimization of the Al 2 O 3 Passivation Layer. To realize the best passivation quality of the Al 2 O 3 film, the minority carrier lifetime (τ ) on the 70 Ω per square n + doped mc-Si surface are investigated as a function of the Al 2 O 3 film thickness (determined by the ALD growth cycles) as well as the annealing duration of the Al 2 O 3 /mc-Si stacks. The results are shown in Fig. 2. As is evidently shown in Fig. 2, the unannealed mc-Si/Al 2 O 3 stacks and the annealed ones exhibit quite different variation tendency in τ . The τ of the unannealed stack exhibit a low τ of 40 μ s. Then the τ value increases monotonically to 352 μ s as the growth cycle of the Al 2 O 3 reaches 800. While for the annealed stacks, the value of the Al 2 O 3 growth cycle determines three different regions of the τ performance. For a given annealing duration, the τ increases (Region I) as the Al 2 O 3 growth cycle increases from 0 to 75. A peak value of 425 μ s is obtained by the 300 s annealed sample as the growth cycle reaches 75. Then the τ value decreases (Region II) as the growth cycle further increases from 75 to 200. Finally, a slight increase trend in τ is observed again at highest growth cycle (Region III). At each value of the growth cycle, the prolonged annealing time leads to a higher τ value in Region I and II. In contrast, the opposite trend is observed in Region III. For example, at the growth cycle of 75, the τ value of the samples increases from 193 μ s to 425 μ s as the annealing duration increases from 0 to 300 s. Oppositely, at the growth cycle of 800, the τ value decreases from 352 μ s to 238 μ s as the annealing duration increases. Furthermore, the passivation effect of the structure of Si/Al 2 O 3 /ZnO NS with 100 s Al 2 O 3 annealing time is further checked by measuring its minority carrier lifetime. As shown in the violet line of Fig. 2, the minority carrier lifetime changes little after the formation of the ZnO NS on the mc-Si/Al 2 O 3 sample. It indicated that the subsequent ZnO NS preparation has little effect on the passivation quality of the Al 2 O 3 on the Si surface. As a result, the subsequent ellipsometry and C-V tests are just conduct on the substrates with only Al 2 O 3 coating for obtaining more accurate SiO 2 thickness and flat band voltages. Given consideration to both passivation quality as well as the heat budget, the 100 s annealed stack with 75 cycle  Al 2 O 3 growth (~12 nm thick and with a relatively high τ of 407 μ s) is selected as the optimum process conditions for the following solar cell fabrication.
Further investigations are made to understand the passivation mechanism of the Al 2 O 3 film with different thickness on the n + doped mc-Si surface. During the annealing treatment on Si/Al 2 O 3 stack, the excess oxygen inside the Al 2 O 3 layer will diffuse to the Al 2 O 3 /Si interface and forms a SiO 2 interlayer 49 . The SiO 2 layer brings chemical passivation effect by saturating the dangling Si bonds on mc-Si surface 41,45,49 . As a result, the thickness of the SiO 2 interfacial layer is analyzed. Before the annealing treatment, owning to the low growth temperature (200 °C) of the ALD-Al 2 O 3 , the SiO 2 layer is unable to be recognized by the ellipsometer. Table 1 shows the thickness evolution of the SiO 2 interfacial layer of the annealed Al 2 O 3 samples with growth cycles increased from 50 to 200. After the annealing procedure, the thickness of the SiO 2 layer evidently increases. As the Al 2 O 3 growth cycles increases from 50 to 100, the thickness of the SiO 2 increased by 57% (1.84 to 2.81 nm). At this moment, the increment of SiO 2 (8.84 to 12.98 nm) coordinates with that of the Al 2 O 3 film. However, as the growth cycles increased from 150 to 200 cycles, the thickness increment of the SiO 2 greatly reduced to 7%, and 1%, respectively. It can be predicted that the thickness of the SiO 2 layer will stand-still at ~3.1 nm at higher growth cycles. This phenomenon can be explained by the annealing condition of the all the Al 2 O 3 samples are the same. So, the amount of the oxygen which can diffuse to the Si/Al 2 O 3 interface is limited. Thus, the chemical passivation effect due to the formation of SiO 2 layer is most effective at a relative low Al 2 O 3 growth cycle ranging from 0 to 100. After that, the negative charge concentration of the Al 2 O 3 films are investigated. Because the negative charges can lead to field passivation effect, which can impact the electron density at the mc-Si surface 40,46,50 . Figure 3a,b describes the C-V characteristics of the Mo/Al 2 O 3 /n + -type mc-Si capacitor, in which the Al 2 O 3 growth cycles ranges from 50 to 200. As shown in Fig. 3a, before annealing, the 50 cycles grown Al 2 O 3 film exhibit electric break-down in accumulation capacitance owning to the relatively low film thickness. The C-V curves of the samples with 100-200 growth cycles exhibit a wide flat band hysteresis window, indicating a high interface trap density. We speculate this phenomenon is mainly caused by the existence of the Si dangling bonds at the mc-Si/Al 2 O 3 interface. This observation corresponds with the poor τ performance of the unannealed Al 2 O 3 samples shown in Fig. 2. After the annealing treatment, the forged SiO 2 interfacial layer saturated most of the dangling Si bonds. In sequence, the flat band hysteresis windows of the samples are greatly narrowed in Fig. 3b. More importantly, as the Al 2 O 3 growth cycles increases, their corresponding flat band voltage shifts toward positive direction, indicating a gradually increased negative charge concentration. One part of the negative charges come from the formation of the tetrahedrally coordinated Al site in the Al 2 O 3 film after the annealing. Another part of the negative charges comes from the SiO 2 interfacial layer 45 . For the p + type silicon passivation, electrons are the minority carriers. The negative charges inside the Al 2 O 3 film plays a positive role as it can prevent the electrons from diffusing to the surface of the p + type Si surface. In our case, the negative charges plays a negative role, because it will accelerate the diffusion of the holes to the n + type Si surface. The τ evolution in the three regions of Fig. 2 can be explained by the combined affection of the two passivation effects mentioned above. In region I, the chemical passivation by the formation of SiO 2 layer dominates the passivation mechanism. Because at this moment, the thickness of the SiO 2 increases quickly. Thus, more and more Si dangling bonds are saturated. Meanwhile, the negative charge concentration inside the passivation layer is low, the corresponding negative influence is minimized. As a result,  the τ value increases quickly in region I. While in region II, the growth of the SiO 2 layer stops, but the negative charge inside the Al 2 O 3 layer continue to increase. In this situation, the field passivation effect plays the main role. Unfortunately, as is mentioned above, the negative charge concentration will accelerated the recombination rate on the n + Si surface, the τ value decreases. At even higher Al 2 O 3 growth cycles in region III, the increment of negative charge concentration is mainly contributed by the formation of Al 2 O 3 at outer side film, which can hardly influence the holes inside the n + type mc-Si. In consequence, the τ value gradually increases as a result of the increased thickness of the passivation layer. Results show that a thin Al 2 O 3 layer can saturate most of the Si dangling bonds while eliminating the negative influence of negative charges on n + doped Si. As a result, the best passivation quality is obtained.

Optical Optimization of the ZnO NS.
There have been many investments on anti-reflection character of the ZnO NS synthesized from spin-coated seed layer [25][26][27] . However, the film thickness cannot be controlled precisely by spin-coating. To overcome this difficulty, the precise film thickness control of the ZnO seed layer is achieved by ALD growth in this work, taking the advantage of the layer-by-layer growth mechanism of the ALD. The anti-reflection character of the ZnO NS synthesized from ALD grown seed layer with different thickness is systematically investigated in this work. At first, the crystallization behavior of the ALD grown ZnO seed layer is investigated because it will greatly affect the surface morphology of the later grown ZnO NS 25 . Figure 4 shows the XRD patterns of the ALD-ZnO seed layers with different growth cycles (ranging from 50 to 200) prepared on the surface of the optimized mc-Si/Al 2 O 3 stack. The seed layer based on 50 growth cycles exhibit a weak ZnO (002) peak at 2θ = 34.4°, which shows that the as-deposited seed layer is mainly amorphous. As the seed layer growth cycle reaches 100, the ZnO (002) peak become stronger, indicating some small ZnO crystal grains started to form inside the seed layer. The calculated ZnO crystal grains size inside the seed layer using the Scherrer formula 51 are summarized in Table S1 in the supporting information. The full-width half-maximum (FWHM) of the ZnO (100) peak decreases as the growth cycle increases, demonstrating a gradually increased grain size of the ZnO crystal form 14.2 to 31.7 nm in the seed layer. Figure 5a-d show the morphology evolution of the ZnO NS synthesized from the seed layers demonstrated in Fig. 4. As shown in Fig. 5a, the ZnO NS prepared on the 50 cycles grown seed layer exhibit a disordered morphology and distribution. One reason is the poor crystalline degree of the seed layer as indicated in Fig. 4. Another reason is that the seed layer is relatively thin, which is unable to efficiently avoid the inhibiting effect of the Al 3+ (contained beneath the Al 2 O 3 layer) on the growth of the ZnO NS 52 . The ZnO NS started to exhibit an ordered morphology after the seed layer growth cycle reaches 100. The TEM image of one ZnO NS shown in Fig. S1 in the supporting information exhibit that the as-grown ZnO NS are single crystalline. As the seed layer growth cycle increases from 100 to 200, the obtained ZnO NS exhibit an increase in diameter and length. On the contrary, the density of the ZnO NS is reduced. The increased diameter is owning to the increased grain size of the ZnO crystal in the seed layer. Because the single crystal epitaxial growth of the ZnO NS are most likely to start from the crystal grains inside the seed layer. We speculate that the increased ZnO crystal grain size is achieved by linking the nearby small grains together. Thus, the number of the grains reduces, leading to the reduced density of the ZnO NS. As fewer NS can be grown within a certain area, the growth of each wire-shaped NS is accelerated, leading to the increased length of the NS. Results show that the surface morphology of the ZnO NS can be controlled by altering the thickness of the ALD grown ZnO seed layer.
To evaluate the light trapping ability, the reflectance spectra of the bare mc-Si, the optimized mc-Si/Al 2 O 3 stack, the mc-Si/Al 2 O 3 /ZnO NS stack prepared in Fig. 5a-d are shown in Fig. 6. Significantly, more than 35% of the incident light is reflected away from the bare mc-Si surface. Deposition of the Al 2 O 3 layer only helps a little, the reflectance loss still excess 30%. It is encouraged that the reflectance is largely suppressed after the growth of the ZnO NS. It is not surprising that the ZnO NS synthesized from the 50 cycles grown seed layer show the highest reflection, because the distribution of the ZnO NS is inhomogeneous (indicated in Fig. 5a). As the seed layer growth cycle increases from 100 to 200, the lowest reflectance value of each mc-Si/Al 2 O 3 /ZnO NS stack gradually decreases (from 7.6% of 100 cycles seed layer growth to 4.3% of 200 cycles seed layer growth). Meanwhile, the lowest reflectance value point shifts from the short wavelength toward the long wavelength (from 7.6% of 100 cycles seed layer growth to 4.3% of 200 cycles seed layer growth). The reduced lowest reflectance value is owing to the increased length of the NS, which enhances the multiple reflections of the incident light. We speculate that the spectrum shift of the lowest reflectance value point is as a result of the increased thickness of the ZnO seed layer. For single anti-reflection layer, the relationship between the film thickness (d) and the lowest reflectance wavelength can be explained as: where N is the refractive index of the film, d is the film thickness, θ is the angle of the incident light, m is any natural number. From this equation, it is evident that the increased film thickness will lead to the "red shift" of λ. This trend corresponds with our observations. The calculated solar energy weighted (AM 1.5) reflectance (R w ) of the nanostructured surface with seed layer growth cycle ranges from 100 to 200 are 9.3%, 8.5% and 8.7% respectively. Results show that the seed layer growth cycle of 150 is the optimum value for the fabrication of the solar cells. Two dimensional (2D) finite difference time domain (FDTD) analysis is carried out to gain insight into the light harvesting mechanism of the optimized mc-Si/Al 2 O 3 /ZnO NS stack. The bare mc-Si surface as well as the optimized mc-Si/Al 2 O 3 stack are also added for comparison. In the simulation, the wavelength of the incident   Fig. 5c. The light intensity distribution (|E y |) for the three structures mentioned above are shown in Fig. 7a1-c1. The strong |E y | on both of the mc-Si (Region A 1 ) and the mc-Si/ Al 2 O 3 stack (Region B 1 ) indicates a high surface reflectance, confirming with the reflectance results shown in Fig. 6. It is evidently shown in Region C 1 that the the intensity of reflected light is greatly reduced after the formation of the ZnO NS. The strong |E y | distribution at the Region C 3 indicating the ZnO NS play a crucial role in suppressing light reflection. One reason is that ZnO has a relatively appropriate refractive index of ~2 in visible light spectrum 53 . Meanwhile, the NS morphology provides a density-graded interference between the air and the substrate 26,54 . Moreover, the effective path length of the incident light is prolonged by multiple reflection effect between the NS. These three effects further boost the light trapping ability of the ZnO NS. As shown in Region C 3 , the bright field inside each ZnO NS reveals that most of the incident light couples into the cylinder-shaped NS. The red field shown in region C 3 suggest that another part of the incident light as well as the escaped light from the NS undergo multiple bounces between the nearby NS. At each bounce, more light is coupled into the ZnO NS. Owing to the wide bandgap character of both ZnO and Al 2 O 3 , these two materials are highly transparent in visible light spectrum. Thus, the light couples into the ZnO NS are then transmitted to the ZnO seed layer and the beneath Al 2 O 3 layer, eventually absorbed by the mc-Si. Benefited from the excellent anti-reflection ability of the ZnO NS, the |E y | distribution in region C 2 is obviously stronger than that in Region A 2 and B 2 , demonstrating an enhanced light absorption of the mc-Si substrate.  Fig. 8a, the obtained cell parameters are summarized in Table 2. The bare mc-Si solar cell exhibit a poor V oc and J sc performance, which is related to the high surface recombination rate and consistent with its low τ of ~43 μ s shown in Fig. 2. The cell based on the optimized mc-Si/Al 2 O 3 stack exhibit a simultaneous increase in V oc and J sc , leading to an evidently improved CE. Our previous work indicate that the improvement is as a result of suppressed surface recombination rate on mc-Si by Al 2 O 3 passivation (echo with the τ increment in Region I of Fig. 2). A further increment in J sc of ~6.1% is observed by the cell based on the optimized mc-Si/Al 2 O 3 /ZnO NS stack. The improvement in J sc is mainly contributed by the significantly lowered surface reflectance on mc-Si/ Al 2 O 3 stack after the formation ZnO NS (indicated in Fig. 6), leading to an increase in photon generated carriers. A maximum conversion efficiency of 15.8% is achieved by the optimized mc-Si solar cell with two functional layers. Meanwhile, the V oc performance of the solar cell with mc-Si/Al 2 O 3 /ZnO NS and mc-Si/Al 2 O 3 /ZnO NS stack are identical. It indicated that the passivation quality of the Al 2 O 3 on Si surface has not degraded very much after the ZnO NS growth, which agrees with the minority carrier lifetimes shown in Fig. 2. Moreover, the cell based on The external quantum efficiency (EQE) analyzation is performed on the representative cells listed in Table 2. The obtained results are shown in Fig. 8b. Comparing with the solar cell based on bare mc-Si, the Al 2 O 3 passivated cell exhibit an enhanced photon conversion ability in the whole 300-1100 range owning to the suppressed surface recombination rate. After the ZnO NS are further added on the Al 2 O 3 film, the obtained cell exhibit a reduced photon conversion rate in the range of 300-400 nm. This can be explained by the band edge absorption of the ZnO for short wavelength photons. Therefore, less photons in that range can be absorbed by the mc-Si substrate. It is encouraged that the EQE of the cell based on the Al 2 O 3 /ZnO NS stack show an evident increase in 450-700 nm range. The increase is owing to the improved anti-reflection ability in the visible spectrum by the application of the ZnO NS. Fortunately, the solar irradiation mainly concentrates in that range 55 , leading to a net increase in J sc performance of 1.9 mA/cm 2 of the obtained cell. Comparing with the cell with the bare mc-Si surface, the cell based on the mc-Si/ZnO NS also exhibit an improved EQE performance. However, the improvement is limited by the high surface recombination rate owing to the poor passivation quality of the ZnO. Therefore, the cell based on the mc-Si/ZnO NS show lower EQE in high solar energy range (450-700 nm) than the Al 2 O 3 passivated cell, leading to the overall poor J sc performance shown in Table 2. Results show that if the Al 2 O 3 layer is removed away from the mc-Si/Al 2 O 3 /ZnO NS stack, the merit in photon capturing of the ZnO NS will be overwhelmed by the poor passivation quality of the ZnO seed layer on the mc-Si surface.
To further prove the effectiveness of the mc-Si solar cell based on the Al 2 O 3 /ZnO NS stack demonstrated in this work, comparisons are made among the a-Si, sc-Si and mc-Si solar cells based on the ZnO NS reported in recent years. The photovoltaic performance of these cells are listed in Table 3. Although the a-Si solar cells developed by Nowak et al. 28,29 exhibit relatively low CEs below 10%, the fabrication procedure was simple and cost-effective. For the mc-Si solar cells, the CE of the mc-Si solar cell with ZnO NS anti-reflection surface demonstrated by Chen et al. 31 was limited by its relatively low V oc of ~500 mV (typically > 610 mV for commercial sc-Si cells). The identical results were also obtained by Aurang et al. in sc-Si solar cell applications 33 . This was owing to the poor passivation quality of the ZnO on Si surface as mentioned above. In the current work, this problem was settled by adding an Al 2 O 3 interfacial layer between the mc-Si and the ZnO NS as described in Table 2, which is similar to those demonstrated in refs 34 and 35. The ALD-Al 2 O 3 passivation layer have shown several merits over the thermal-SiO 2 34 and the PECVD-SiN x 35 films. As compared to the thermal-SiO 2 , the deposition temperature of the passivation layer can be reduced greatly from 850 °C to 200 °C by utilizing the ALD-Al 2 O 3 . Moreover, the Al 2 O 3 /ZnO seed layer stacks can be prepared in an ALD chamber. On the other hand, the ALD-Al 2 O 3 layer was pin-hole free. Comparing with the PECVD-SiN x thin film. During the hydro-thermal growth procedure of the ZnO NS, the pin-hole free passivation layer can prevent the reaction solution from reaching the Si/front electrode area more efficiently. This is beneficial for obtaining high quality solar cells with good V oc and FF performance. However, The J sc of the cell demonstrated in this work was not as high as that of the others. Because there is no surface texturing process on mc-Si in this work. On the positive side, the usage of highly toxic or corrosive chemicals such as HNO 3 and HF was avoided, which is very important for realizing the eco-friendly production.  Fabrication of the solar cell. Commercial grade p-type mc-Si wafers with a resistivity of 1-3 Ω•cm were used as the starting substrate. At first, the wafers are immersed in 20 wt. %, 80 °C NaOH solution for 5 min for removing saw damages. Then, an n + emitter with a sheet resistance of 70 Ω per square was formed using liquid POCl 3 diffusion. The formed phosphor silicon glass was removed by immersing the wafers in diluted HF solution.
After that, the front/back electrode was formed by screen-printing of silver/aluminum paste. The area-fraction on the surface was ~9% (the power loss induced by grid shadow was not excluded in the J sc and CE calculation), the active area of each cell was 3.24 cm 2 . The front/rear electrode metallization were realized by annealing the samples in the RTA furnace in air with an peak temperature of ~720 °C. The samples were then shortly immersed into diluted HF to remove the oxide layer and contaminant formed on the front surface of the samples during the formation/annealing step of the electrodes. Then, the Al 2 O 3 /ZnO NS stack were prepared on the front surface of the cells.
Sample characterization. The minority carrier lifetime on the mc-Si surface was obtained using a life time tester (Semilab WT-1000). The C-V curves were measured by a semiconductor device analyzer (Agilent B1500A). The XRD measurement was carried out in Bruker D8 system. The surface morphologies were characterized by a SEM (Hitachi SU1510). The reflectance spectra were measured using a spectrometer (Shimadzu UV3600) equipped with an integrating sphere. The solar cell performance was obtained under a standard 1-sun illumination with a sun simulator (Oriel-940401A) and a sourcemeter (Keithley-2400). The data obtained was based on an average of about 6 wafers/cells.  Table 3. Literature comparison in photovoltaic performance of the Si solar cells based on ZnO NS published in recent years.