Figure 2a–e shows the top-view SEM images of the different process steps for the silicon surface texturing. Figure 2a shows the 50–80 nm porous Si on the surface of Si wafer etched by MACE method in the mixed solutions of AgNO3/HF/H2O2. Subsequently, the porous Si is modified by the isotropic etching in the mixed aqueous solutions containing HF/HNO3 and turns to be nanohole structures with a diameter of 800 nm as shown in Fig. 2b. Finally, the micron inverted pyramids (IPs) with different sizes (Fig. 2c–e) are obtained by sodium hydroxide in aqueous solution at 60 °C for 30, 60, and 90 s, respectively. From Fig. 2c–e, we can see that after alkali treatment, the IPs structure sizes for three etching time of 30, 60, and 90 s are ~ 1, 1.3, and 1.8 μm, respectively, meaning an increasing size of IP with the increase of alkali treatment time. Also, we notice that the IPs tend to collapse and transit to be the upright pyramids with the increase of the etching time. As known, the inverted pyramids have the advantage of light trapping over upright ones because light will undergo extra one or two bounces in inverted pyramids than that in upright pyramids. Therefore, the structures with shorter etching time are suitable for the light-trapping textures of PV devices because of the advantage in the short-wavelength antireflection. Figure 2f is the compared photos for different surface structures corresponding to Fig. 2a–e.
Now we turn to the optical properties of Si IP-strus. From the reflectance over the whole wavelength range of 300–1100 nm (Fig. 3a), we observe that the porous Si has a low reflection because of the excellent light-trapping performance of nanostructures [22,23,24]. For nanohole structures, the reflectance in the whole wavelength range has an obvious increase, which is attributed to the decrease of density and increase of feature size of nanoholes. After NaOH treatment for 30 s, benefiting from 3–4 bounces between the (111) planes of the IP, the IPs structures display lower reflection over the 300–1100 nm wavelength range, especially in the short-wavelength range of 300–500 nm. With the alkali etching time increasing, the IPs become larger and tend to be the upright pyramids, resulting in an increasing reflectance. When all samples were covered with the same stack SiO2/SiNx coating, the reflectance drop sharply by more than 10%, which is attributed to the combined reflectance from the optical interference of the stack SiO2/SiNx thin films and the surface structures. In this case, the reflection spectra of samples from different processes are mainly different in the wavelength range of 300–600 nm, which is caused by the difference of feature size of IPs. In particular, Si IP-strus covered by the stack SiO2/SiNx layers displays better short-wavelength antireflection ability than the others, indicating the excellent external quantum efficiencies (EQEs) in the short-wavelength range.
Furthermore, we calculate the average solar reflectivity Rave (see Fig. 3b) over the wavelength range of 300–1100 nm and compare the reflectivity of Si IP-strus with other structures corresponding to different intermediate processes shown in Fig. 2a–c. Rave can be calculated by the expression of
$$ R\mathrm{ave}=\frac{\int_{300\ \mathrm{nm}}^{1100\ \mathrm{nm}}\mathrm{R}\left(\uplambda \right)\ast \mathrm{S}\left(\uplambda \right)\ast \mathrm{d}\uplambda}{\int_{300\ \mathrm{nm}}^{1100\ \mathrm{nm}}\mathrm{S}\left(\uplambda \right)\ast \mathrm{d}\uplambda} $$
(1)
where R(λ) and S(λ) denote the measured reflectance and AM1.5 solar photon spectral distribution, respectively. As shown in Fig. 3b, the Raves of porous Si, nanoholes, IPs, and IPs with SiO2/SiNx coating are 8.22, 17.96, 15.18 (group 1—30 s)/17.35% (group 2—60 s)/20.3% (group 3—90 s), and 3.91% (group 1—30 s)/4.48% (group 2—60 s)/5.60% (group 3—90 s), respectively. The Raves show that the IP-strus have a better antireflection ability than nanoholes and show a decreasing trend with the increase of feature size. When IP-Strus are coated by the stack SiO2/SiNx layers, the lowest Rave is 3.91%, revealing an ideal light-trapping structure for the PV device.
The stack SiO2 (~ 2 nm)/SiNx (~ 75 nm) passivation for the Si IP-based n+ emitter is an effective way for achieving well electrical performance of IP-based PERC and their passivation effect [1] and mechanism have been systematically studied in our previous work [14]. To show the electrical superiority of the stack Al2O3/SiNx passivation layers at the rear of our device, we investigate the influence of the different annealing and light-soaking conditions on the effective minority carrier lifetime (τeff) with respect to the injection level (Δn), as shown in Fig. 4a. Notice that the polished Si wafers have the bulk minority carrier lifetime of ~ 350 μs, and the stack Al2O3/SiNx layers are symmetrically deposited on both sides of polished Si wafers. The thickness of inner Al2O3 and the outer SiNx layer is estimated as ~ 3 and ~ 125 nm, respectively. Two annealing conditions are performed in the air atmosphere: 300 °C and 800 °C for 15 min. Then the wafers are illuminated at 25 °C under the full-wave ranged halogen lamp with a power intensity of 50 mW cm−2 for 100 s. As can be seen from Fig. 4a, the 48 μs τeff (300 °C) and 126 μs τeff (800 °C) after annealing are much higher than the 22 μs τeff of the as-deposited Al2O3/SiNx passivated samples at the injection level of 1.2 × 1015 cm−3.
Importantly, the effective minority lifetime of annealed samples after 100 s of illumination are 230 μs and 150 μs, respectively, much higher than 126 μs and 48 μs before illumination, demonstrating a very clear light-enhanced c-Si surface passivation of Al2O3/SiNx layers. The charge trapping effect during light soaking [25,26,27,28] could be one of the main mechanisms for the light-enhanced c-Si surface passivation of Al2O3/SiNx films. As Al2O3 films are reported to have a negative fixed charge density [29,30,31,32], some of the excess electrons generated by light were likely to be injected or tunneled into trap states in the inner Al2O3 film, resulting in an increased level of field-effect passivation. Interestingly, the light-enhanced passivation effect at 300 °C annealing is better than that at 800 °C, meaning that light-soaking at a lower temperature annealing is a more effective way to the application of PV device.
To study the effect of the annealing process on the surface modification, we compare the Fourier transform infrared spectroscopic (FTIR) absorption spectra of the annealed samples with that of the as-deposition sample. Figure 4b manifests that the Si–N, Si–O, Si–H, and N–H bonds correspond to the stretching absorption peaks at the wavenumbers of ~ 840, 1070, 2200, and 3340 cm−1, respectively. We see that the densities of both the Si–N and Si–O bonds show an obvious increase after annealing; meanwhile, the density of the Si–H bonds increases slightly. The increases of the Si–O and Si–H bond density implies the decrease of the dangling bonds at the interface of Si/SiO2, resulting in a better passivation effect [33]. Also, the annealing process promotes the density of Si–N bonds, indicating a more dense structure which can effectively prevent the out diffusion of H from entering into the environment instead of into Si bulk. However, for excessively high annealing temperature, the H in Si–H and N–H groups can escape from the bulk Si and the dielectric layers to the environment, which causes the decline of the passivation effect. The result of FTIR is consistent with that of the effective minority lifetime.
To further understand the difference of passivation mechanism between thermal annealing and light-soaking treatment, we analysis the density of fixed charges (Nf) and the density of interface traps (Nit) at the interface of Si and Al2O3 (ALD)/SiNx (PECVD) stack layers by using capacitance–voltage (C-V) measurements from a rigorous metal–oxide– semiconductor (MOS) model.
Nf can be obtained from the following equation:
$$ {\mathrm{N}}_{\mathrm{f}}=\frac{{\mathrm{Q}}_{\mathrm{f}}}{\mathrm{S}\times \mathrm{e}}=\frac{{\mathrm{C}}_{\mathrm{OX}}\times \left({\mathrm{V}}_{\mathrm{MS}}-{\mathrm{V}}_{\mathrm{FB}}\right)}{\mathrm{S}\times \mathrm{e}} $$
(2)
where the following expression can calculate VFB
$$ {V}_{\mathrm{FB}}={V}_{\mathrm{MS}}-\frac{Q_f}{C_{\mathrm{OX}}} $$
(3)
Note that S is the area of metal electrode, e is electronic charge, COX is the capacitance of dielectric film layer, VMS is the difference of the work function between the metal electrode and p-type Si, and VFB is flat band voltage.
Using the Lehovec method [34], we can obtain Nit from the C-V curve:
$$ {\mathrm{N}}_{\mathrm{it}}=\frac{\left({\mathrm{C}}_{\mathrm{OX}}-{\mathrm{C}}_{\mathrm{FB}}\right){\mathrm{C}}_{\mathrm{FB}}}{3{\left(\updelta \mathrm{C}/\updelta \mathrm{V}\right)}_{\mathrm{FB}}\mathrm{ekTS}}-\frac{{\mathrm{C}}_{\mathrm{OX}}^2}{\left({\mathrm{C}}_{\mathrm{OX}}-{\mathrm{C}}_{\mathrm{FB}}\right)\mathrm{S}{\mathrm{e}}^2} $$
(4)
where (δC/δV)FB is the slope near-flat band and is taken as the absolute value. CFB, e, and k are capacitance of MOS structure in a flat band, electronic charge, and Boltzmann constant, respectively.
It can be seen from Fig. 4c that the measured C-V curve of the Al2O3/SiNx stack layers shows obvious accumulation region, depletion region, and inversion region. According to the C-V curves and Eq. (2–4), we obtain the interface properties of the prepared MOS structures, as shown in Table 1.
Table 1 Nf and Nit at the interface between Al2O3/SiNx stack layers and Si The fixed negative charge densities show a significant increase by an order of magnitude after thermal annealing meanwhile the interfacial states densities significantly decrease, indicating that annealing enhanced the chemical passivation and field-effect passivation of dielectric films. By further light-soaking treatment, the densities of interfacial states keep the same level, while the densities of fixed negative charges increase further. As mentioned above, some of the excess electrons generated by light were likely to be injected or tunneled into trap states in the inner Al2O3 film, which means that light soaking can enhance the field-effect passivation of the dielectric film. Although the value of Nit is high, the sample by 300 °C annealing and 100 s light-soaking has the highest τeff of 230 μs due to the highest Nf of − 2.87 × 1012 cm−2, meaning that field-effect passivation has an advantage over chemical passivation in this case.
Figure 4d shows the photoluminescence and electroluminescence photos of 1, 1.3, and 1.8 μm IP solar cells with the same passivation process. The brightness of the three groups of photos for both photoluminescence and electroluminescence keeps basically the same level, meaning that the three groups of solar cell devices perform equally well in the passivation of defects. That is to say, the passivation process determines the electrical performance of the solar cell instead of the feature size of IPs, which will be confirmed by the following output parameters of the fabricated solar cells.
Based on the excellent optical and electrical performance of the simultaneous SiO2/SiNx stack layers passivated front Si IP-based n+ emitter and Al2O3/SiNx stack layers passivated rear reflector, we fabricated the Si IPs-based PERC.
Figure 5a shows the internal quantum efficiencies (IQEs) and front surface reflections of the fabricated Si IP-based PERCs. We can observe that 30-s alkali-etching IP-based device (group 1—30 s) shows the lowest reflectance in the short wavelength of 300–600 nm due to its smaller feature size of IPs. Importantly, group 1—30 s has the highest IQEs in this wavelength range, and thus yields the highest external quantum efficiencies (EQEs) as shown in Fig. 5b. Also, the fabricated devices display almost the same EQEs in the long-wavelength range because of the same level of reflectance and IQEs in this range. Therefore, group 1—30 s with smaller feature size possesses better output performance than the other two groups, which is further confirmed by the I-V and P-V curves of devices (see Fig. 5c). Figure 5d shows the η of our champion device reached 21.41%, as well as the Voc of 0.677 V, Isc of 9.63 A, and FF of 80.30%. By our knowledge, it is the highest η among MACE-IP-based solar cells. The inset of Fig. 5d is a photograph of the front and rear surface of the champion device.
Furthermore, Table 2 shows the detailed parameters of the fabricated devices. Obviously, the average Isc (9.63 A) of the group 30 s device is higher than that of the other two groups, which lies in its best anti-reflection ability of front surface as mentioned above. The difference of Iscs mainly determines the output performances of the devices. Besides, the higher FF and the lower series resistance Rs guarantees the higher η of group 30 s. It is worth to note that all the average Vocs of the Si IP-based PERCs are in the range of 674–676 mV, demonstrating that the same excellent passivation for the front and rear surface of all groups. Finally, benefiting from the gain of optical and electrical performance, we have successfully achieved the highest η of 21.4% of Si IP-based PERC solar cell.
Table 2 Detailed output parameters of Si IPs-based PERC