Abstract

Composite silicon dioxide-titanium dioxide (SiO2-TiO2) films are deposited on a large area of 15.6 × 15.6 cm2 textured multicrystalline silicon solar cells to increase the incident light trapped within the device. For further improvement of the antireflective coatings (ARCs) quality, dimethylformamide (DMF) solution is added to the original SiO2-TiO2 solutions. DMF solution solves the cracking problem, thus effectively decreasing reflectance as well as surface recombination. The ARCs prepared by sol-gel process and plasma-enhanced chemical vapor deposition (PECVD) on multicrystalline silicon substrate are compared. The average efficiency of the devices with improved sol-gel ARCs is 16.3%, only 0.5% lower than that of devices with PECVD ARCs (16.8%). However, from equipment depreciation point of view (the expiration date of equipment is generally considered as 5 years), the running cost (USD/watt) of sol-gel technique is 80% lower than that of PECVD method for the first five years and 66% lower than that of PECVD method from the start of the sixth year. This result proves that sol-gel-deposited ARCs process has potential applications in manufacturing low-cost, large-area solar cells.

1. Introduction

Antireflective coatings (ARCs) are now widely used in mass produced silicon photovoltaics [13]. High quality ARCs must be designed accurately on mono- and multicrystalline silicon solar cell surfaces, not only to passivate the surface dangling bonds acting as recombination centers [48], but also to minimize the loss of optical absorption. The conventional plasma-enhanced chemical vapor deposition (PECVD) technique is employed generally to deposit antireflection layers such as Si3N4 and SiO2 [913]. However, this technique is always associated with vacuum processes and dangerous process gases. Therefore, replacing PECVD by other cost-effective methods can considerably reduce production costs of manufacturers and raise the security of the clean-room environment. Sol-gel spin-coating method applied extensively in preparing lots of different kinds of coated films has become a popular technique for the past thirty years due to its high process speed, low-cost, continuous production, and suitability for the large-area process [1417]. Nevertheless, ARCs deposited on textured silicon wafers by sol-gel method present worse uniformity than those deposited by PECVD technique. The severest nonuniformity areas resulting from centrifugal force appear on four corners of the substrate. For overcoming it, we control the spin rate of the spin coater, spin duration, and adjust the concentration of sol-gel solutions. Further, adding dimethylformamide (DMF) solution to the composite films is a critical step to solve the problem of film cracking. Mostly, cracks are caused owing to different coefficients of thermal expansion between the air and ARCs films [1820]. Although the basic properties of most kinds of sol-gel-deposited films have been discussed, the application of composite films in solar cell research is not well investigated.

This study proposes a simple low-cost sol-gel method of preparing high-quality ARCs on large area multicrystalline wafers. DMF solution can help ARCs to be effectively coated on irregular and rough wafer surfaces treated by acid texturing. The average efficiency of solar cells with sol-gel-deposited antireflective films approaches that of ready-made crystalline solar cells. Thus, using sol-gel method instead of PECVD technique to deposit ARCs has potential applications in mass production of solar cells.

2. Experiment

2.1. Preparation of Sol-Gel Solution and Process of Spin Coating

The experimental SiO2-TiO2 colloid solutions in this study contained varying volume ratios of SiO2 and TiO2 solutions. The SiO2 solution was prepared by reacting metal alkoxide with a mixture of a critical amount of water and hydrochloric acid (HCl) catalyst in a medium diluted with ethyl alcohol solvent. Tetraethyl orthosilicate (Si(OC2H5)4) was then added into the resultant solutions and stirred with a magnetic stirrer. The TiO2 solution was prepared similarly except that titanium isopropoxide (Ti(OC3H9)4) was added instead of (Si(OC2H5)4). The SiO2 and TiO2 were then mixed separately in different volume ratios. Tables 1, 2, and 3 show the detailed information for all solutions. Finally, a small amount of DMF was added into the SiO2-TiO2 synthetic colloid solution.

After adding several drops of synthetic solution to the 15.6 × 15.6 cm2 multicrystalline wafers, the wafers were spin-coated for two rounds. The first coating was applied at a speed of 150 rpm to ensure uniform composite films, and the second round was performed at 330 rpm to obtain the desired film thickness. Each film was then prebaked at 80°C for 20 min and postbaked at 200°C for 40 min in atmosphere. Then, samples were further annealed in a quartz furnace from 100°C to 850°C for around 30 seconds firstly. Then furnace temperature cooled down from 850°C to around 100°C and lasted for 30 seconds.

Optical characteristics such as refractive index, thickness, and reflectance of films were subsequently measured by ellipsometer and fiber coupled CCD array spectrometer, respectively.

2.2. Design of Antireflection Coatings

More incident light is essential for mono- and multicrystalline solar cells to enhance their short-circuit current (). In some applications, zero reflectance is needed throughout a narrow spectrum band or at a single wavelength. For a substrate with a single ARC, the light from the ARC-substrate interface is reflected back to the ambient (air)-ARC interface with a phase change of 180° and interfere with the light reflected from the interface between ambient (air) and ARC in the opposite direction. In this state, the reflectance can be described as where , , and are the refractive indices of the ARC material, air, and substrate, respectively. To obtain zero reflectance, the term of must be zero. The resulting boundary condition can then be described as Moreover, the optimum thickness and refractive index with a minimum reflectance for a single-layer ARC can be deduced from where is the midrange wavelength of 630 nm, and where and are the refractive index and layer thickness, respectively [21].

2.3. Solar Cell Performance

Solar cells were fabricated from boron-doped multicrystalline Si wafers with resistivity of approximately 15 Ω-cm. Wafers were textured using hydrofluoric acid. After forming the emitter region by thermal diffusion of phosphorous atoms in a quartz tube furnace at 850°C, the back surface field was formed by annealing the aluminum contact deposited using screen printing method. The composite SiO2-TiO2 films and Si3N4 films were then deposited on the front of solar cells by spin-coating and PECVD techniques, respectively. Quasi-Steady-State Photoconductance lifetime measurement technique (QSSPC) was carried out to evaluate the surface passivation effect of the ARCs on silicon substrate. Front silver contact prepared through screen printing method was then annealed at 700°C for 30 min. Finally, effective illuminated light (AM 1.5 G) from the solar simulator was slightly decreased by the grid-type collecting electrode, which occupied about 8% of the front surface area.

3. Results and Discussion

When designing an ARC, the goal is to determine the refractive index and thickness of SiO2-TiO2 films that provide optimal performance along the desired spectrum. Figure 1(a) shows refractive index spectra for pure TiO2 and SiO2-TiO2 composite films deposited on monocrystalline substrates. The average refractive index is defined as the average value of all points on each curve. It can be found that as the pure TiO2 solution adds more volumes of SiO2 solution, the average refractive index decreases obviously from 2.1 to 1.45. Figure 1(b) further shows the different refractive indices corresponding to each ratio of SiO2 to TiO2 (SiO2/TiO2) at a wavelength of 630 nm. At this wavelength, spectral photon intensity is the highest, and the strongest destructive interference happens, causing light to be trapped in devices [21]. Based on the boundary condition equation (2) mentioned previously, the appropriate refractive index 1.85 is obtained while the SiO2/TiO2 ratio is 0.1.

For analyzing the properties of composite films, X-ray photoelectron spectroscopy (XPS) spectra for pure SiO2 film, pure TiO2 film, and their mixture films in different SiO2/TiO2 ratios is carried out within the binding energy of 0–1200 eV as shown in Figure 2(a). The Si 2p, Si 2s, Ti 2p, and O 1s binding energy locate at 103.1, 155, 456.6, and 532.4 eV, respectively. The C 1s peak is 284.7 eV, which is attributable to carbon contamination at the outer surface of the composite films. The Ti 2p peaks gradually disappear with increasing of SiO2/TiO2 ratio, whereas the Si 2s and Si 2p peaks strengthen, which validates that the components of all composite films changed momentarily. Figure 2(b) shows the results of further analysis of the behaviors of O 1s peaks in composite films with various SiO2/TiO2 ratios. The peak shifts from low binding energy (Ti–O bonds) to high binding energy (Si–O bond). Characteristics of composite film get close to that of pure SiO2 film. From the materials analysis perspective, these changes explain why the refractive index of SiO2-TiO2 composite films changes as different SiO2/TiO2 ratios.

Since the ratio of SiO2/TiO2 has been optimized to 0.1, we start to investigate the properties of composite films based on this ratio in later experiment. Figure 3 is a scanning electron microscope (SEM) cross-sectional image of the SiO2-TiO2 composite films before and after annealing process. It can be seen in Figure 3(a) that, by adjusting spinning speed, the SiO2-TiO2 composite film can be coated to the surface of textured multicrystalline wafer smoothly, but is still loose before annealing. After annealing at 850°C for 30 seconds, thickness of ARC decreases from 93.3 nm to 75.4 nm as shown in Figure 3(b). The film becomes denser, probably attributed to firing of the organic solvent and removal of resin [22]. However, in this case, a desirable thickness approximating 85 nm is calculated using (3) shown previously. A 10 nm deviation exists between theoretical value and experimental result, leading to a very slight influence on cell’s performance. It’s really difficult to control the film thickness accurately of using sol-gel technique; therefore, how to improve this disadvantage is a critical issue in the future.

The SEM images of surface morphology of composite SiO2-TiO2 films with DMF ranging from 0% to 15% are schematically presented in Figure 4. Figure 4(a) exhibits lots of cracks in the composite SiO2-TiO2 film without DMF. From the zoomed insert, visible cracks spread on the bottom of notches formed by acid texturing. The cracks, which probably formed during the annealing process, decrease the absorption of incident photons. Worse light absorption may lead to lower light-generated carriers, thus decreasing the of solar devices. Figures 4(b) to 4(d) show that adding 5% to 15% DMF to SiO2-TiO2 solutions gradually improves the crack problems. When the DMF ratio reaches 15%, the cracks are almost entirely eliminated. This phenomenon may be probably explained that DMF solution can not only diminish the reaction rate of sol hydrolyzing-polycondensation and prevent the little sol particle from growing up, but also promotes the lateral connection of particles by combining the hydrogen bonds with intermediates of sol during gelling process. Moreover, the pore tension of mixture solution could be diminished by DMF solution in thermal process, leading to lower vapor pressure. Hence, the composite films are coated uniformly. According to QSSPC measurement (not shown here), the lifetime of excess minor carrier of each composite film with 0% to 15% DMF solution are 2, 8, 17, and 23 μs, respectively, revealing that as the cracks in the films reduce, the lifetimes becomes higher. Compared to the lifetime of monocrystalline silicon substrate, these relative low lifetime values can be attributed to lots of grain boundaries in multicrystalline silicon substrate, which act as recombination centers.

Figure 5 displays the optical reflectance spectra of pure TiO2 films, SiO2-TiO2 composite films with different volume of DMF, and Si3N4 films fabricated by PECVD. The results show that the PECVD ARC has a higher reflectance compared to other films prepared by sol-gel technique at wavelengths lower than 500 nm. This relatively high value is attributed to its relatively higher refractive index of Si3N4 at short wavelength [16]. The PECVD ARC has an average reflectance of approximately 9.5%, and its minimum reflectance is approximately 1.5% at wavelengths approximating 770 nm. Further comparison of the curves of the other composite films with and without DMF shows that lowest average reflectance is obtained while the mixed SiO2-TiO2 solution plus 15% DMF, presenting an average reflectance of 9% and minimum reflectance of approximately 4% at the wavelength around 670 nm. Obviously, composite films without cracks provide suitable path for incident light, reducing the reflection loss considerably. In summary, the entire measured tendency corresponds well with the SEM images shown in Figure 4, revealing the feasibility of adding DMF solution.

Figure 6 compares performances of solar cells between PECVD-ARC cells (ready-made multicrystalline cells for mass production) and improved sol-gel-ARC cells. All performance parameters for the fifty devices with optimal sol-gel ARCs are highly consistent, which confirms the reliability of the process. The optimum average efficiency is approximately 16.3%, about 0.5% lower than the average efficiency (around 16.8%) of PECVD-ARC cells. This 0.5% discrepancy is caused mainly by short-circuit current and open-circuit voltage (). The chief reason of the lower of those sol-gel-ARCs devices can be discussed in two factors: reflectance and lifetime. As noted in Figure 5, the average reflectance of sol-gel-ARCs devices plus 15% DMF solution is 9%, which is lower than 9.5% of the PECVD-ARCs devices. It is expected that sol-gel-ARC device should absorb more incident light resulting in higher . However, average of the sol-gel-ARCs devices is 34.9 lower than 35.2 mA/cm2 of the PECVD-ARC devices, which implies that the carrier lifetime substantially affects . Noted that the mean lifetime of PECVD-ARCs devices is around 27 μs and of the sol-gel-ARCs devices is around 23 μs. Moreover, lower mean lifetime of the sol-gel-ARCs devices also affects their performance. Michl et al. indicated that the excess minor carrier lifetime substantially affects the in multicrystalline devices [23]. This result is consistent with the experimental trend in this study. Hence, further studies are needed to improve the interface quality between a sol-gel-ARC and a silicon substrate.

Figure 7 compares the appearance of a PECVD-ARC device (Figure 7(a)) with a sol-gel-ARC device (Figure 7(b)). Before fabrication of the ARCs, both devices initially have the same multicrystalline silicon substrates. It can be seen that the colors of both devices are nearly the same after fabricating ARCs. Very little color variation is caused by the different reflectance of both devices at visible wavelength spectrum. The deep blue sol-gel-ARC is quite uniform without any uncoated areas formed on the four corners of the substrate, indicating that the proposed sol-gel process can potentially replace the PECVD process for the mass production of single- and multicrystalline silicon solar cells.

Table 4 provided by Gallant Precision Machining Co., Ltd., Taiwan compares equipment cost, materials cost, and running cost between various ARCs fabrication methods. The expiration date of equipment is generally considered as five years. From equipment depreciation point of view, the running cost (USD/watt) of sol-gel technique is around 80% lower than that of PECVD method for first five years and around 66% lower than that of PECVD method from the start of the sixth year. Although the efficiency of sol-gel-ARC device is lower than that of PECVD-ARC device, a huge difference in manufacturing cost still makes sol-gel-ARC device have enough competitiveness.

For further increasing efficiency of the proposed sol-gel-ARCs cells, it is necessary to fabricate double or even triple layers ARCs. A single layer ARC is able to be nonreflective only at one wavelength, often at the middle of the visible region, but multiple layers are more effective over the whole visible spectrum. Moreover, high conversion efficiency of monocrystalline silicon solar cells can probably be realized with the same sol-gel process.

4. Conclusion

A study aiming to improve the quality of sol-gel ARCs that could help increase the efficiency of multicrystalline solar cells is presented in this work. The effective refractive index and thickness of the films are adjusted by controlling the spin rate and annealing temperature. Average reflectance is eventually reduced to 9% by adding DMF solution to reduce cracks in the film while the average efficiency of devices is improved to 16.3%. Although the efficiency 16.3% still trails that of the ready-made multicrystalline solar cells by approximately 0.5%, the reliability and uniformity as well as lower running cost are highly attractive to manufacturers. In the future, use of the sol-gel technique, which combines spin-coating with furnace annealing to prepare ARCs, could enable a more efficient and a continuous mass production of Si solar cells compared to the use of PECVD method, which requires an expensive vacuum process and dangerous process gases.

Acknowledgments

This work is sponsored by Gallant Precision Machining Co., Ltd., Taiwan and the National Science Council of Taiwan under the Grants nos. NSC 100-2628-E-451-002-MY2(2/2) and NSC 101-3113-E-451-001-CC2.