This paper investigates crystalline orientation in monolike silicon wafers and its effect on solar cell performance. Monolike silicon wafers from two different bricks cut from interior and corner region of an ingot were compared. The mono grain in the interior brick is nearly perfect, but there are some large oblong shaped sub-grains in the corner brick. The large sub-grains at corner brick wafers are oriented at (311), instead of (100) orientation. The (311) grains contain high density of dislocation and cannot be effectively textured by alkaline solution, therefore lowering the cell efficiency significantly. There is about 0.86% (abs) cell efficiency reduction on the monolike cells that contain large sub-grains.
Cast monolike silicon solar cells are a new type of photovoltaic (PV) cell that has drawn a lot of attention in PV industry recently [1-3]. Instead of pulling single crystalline silicon by Czochralski method, single crystalline silicon is grown from the melt by casting in monolike silicon wafers. The cast monolike silicon has an advantage of high cell efficiency close to cells made on Cz wafers, but at a cost similar to that of multicrystalline silicon wafers. There has been significant progress in monolike cell technology in the past few years. Cell efficiency of ~20% has been reported for the monolike cells fabricated with standard cell processing techniques [
Monolike silicon wafers are grown by casting from a seeded layer at the bottom of the crucible by heat exchange method [1,6]. Ideally, silicon grows vertically from bottom up in the same (100) orientation as the seeded layer, therefore creating single crystalline structure. However, due to the fast cooling at crucible walls, random nucleation may happen near the crucible walls, forming multicrystalline silicon around the perimeter of the ingots. Additionally, there may be silicon grains nucleated from the seed interfaces. The multicrystalline grains in the monolike silicon wafers degrade the wafer quality and bring up challenges in cell processing. For examples, the multicrystalline grain regions cannot be effectively textured with conventional alkaline solution. Variations in multicrystalline grain areas will also lead to a wide range of cell efficiency in a production line. It is therefore critical to control the non-mono grain regions and to optimize monolike cell process. In this paper, we study the crystalline structure of monolike silicon and its effect on cell performance.
Monolike silicon wafers of 156 × 156 mm2 were withdrawn from different locations, an interior and a corner, of a large industrial ingot (840 × 840 mm2). Wafers were cut by diamond wire sawing to a nominal thickness of 200 μm.
Solar cells were processed with a standard full aluminum back surface field (Al BSF) technology. First, wafers were treated in heated KOH solution to remove up to 5 μm sawing damage layer followed by KOH/IPA solution for texturing. After cleaning with 10% HF and 20% HCl solution, wafers were processed with phosphorous diffusion in a POCl3 tube furnace for a final sheet resistance of 65 Ω/sq followed by edge-isolated in acid solution and cleaning in 10% HF to remove phosphosilicate glass (PSG). Cells were coated with ~80 nm SiNx antireflection layer in a low frequency PECVD reactor. Electrical contacts were screen printed with DuPont PV17F Ag paste for front gridline and Monocrystal PASE-1203 Al paste for rear contact. The cell contacts were finally fired in TPSolar IR belt furnace.
Cell efficiency was measured by an automated I-V tester. Cell internal quantum efficiency (IQE) was determined by combination of wavelength-dependent spectral response and external surface reflectance measurements. Cells were further evaluated by a scanning Kelvin probe system assisted with local illumination [
Crystalline silicon structures from different grains in monolike silicon wafers were measured by X-ray diffactometer. Dislocation densities were further analyzed by optical microscopy after polishing and etched by Yang etching [
A series of monolike wafers were processed with standard Al BSF technology.
As the VOC represents an average value of full-sized cells which cannot show a spatial variation, therefore full-sized cells were further measured by a scanning Kelvin probe system.
tween the grain structure and surface potential. Grain structures vary by wafer location. The sub-grains on the interior brick cells (
The reasons for surface potential variation were analyzed by X-ray and dislocation density measurements. It is found that the mono grain has exactly the same crystalline orientation as silicon seeds at (100) orientation, but the sub-grains on the corner brick wafers are mostly (311) orientated. It is also found that several large (311) grains are closely packed but separated by the grains with seed orientation. It can be seen that the (311) orientation is outgrown from (100) surface. The special (311) grain orientation from the Si seed is interesting. Generally, it is known that there are several favorable growth orientations, such as (100), (110), (111), (311) and (533), in the cast multicrystalline silicon wafers [10,11]. However, there are only fewer growth orientations in the monolike wafers. The limited growth orientation is important for monolike wafers, but it may also create some large subgrains on the wafers, as shown in
The dislocation densities at different grains are measured by optical microscopy after polishing and etching.
orientation tends to encompass more dislocations than others.
The effect of dislocation and texture on the cell per-
formance is simulated by PC1D. Two representative cells from the interior and corner bricks were chosen for modeling. There are four distinct cell regions that may be classified which have individual effects on cell performance. Monolike silicon wafers may have 1) (100) oriented, textured single crystalline regions; 2) (100) oriented, textured single crystalline regions displaying dislocation patterns; 3) (311) oriented, planar single crystalline regions; and 4) planar, multicrystalline regions possessing many crystal orientations. Inspection of the representative interior brick cell reveals the following region ratio: 70% region 1, 15% region 2, and 15% region 3. Similar inspection of the representative corner brick cell reveals the following ratio: 36% region 1, 25%, region 2, 22% region 3, and 17% region 4. IQE and external surface reflectance curves were collected for each region of the representative cells and are shown in
Short and long wavelength IQE fitting using PC1D is necessary to derive the front and back surface recombination velocities (FSRV and BSRV, respectively) for accurate modeling of solar cell current and voltage parameters. FSRV is influenced by emitter quality and front surface passivation while BSRV is influenced by Al-BSF quality and τbulk. Using PC1D, BSRV is only accurate when τbulk is known. The variable defect densities observed for monolike wafers indicates that τbulk will also vary widely across the wafer. In order to model this variation in τbulk for the monolike Si wafers, BSRV is derived for a control cell made on Czochralski-grown Si
wafer with a known final τbulk and this value for BSRV is applied to the monolike cells. The modeled Czochralski cell was processed identically to the monolike cells, receiving the same front surface texture, phosphorus diffusion, and SiNx antireflection coating. The Czochralski cell was printed with the same Ag and Al metallization pastes and fired at the same peak wafer temperature. Final τbulk for this device was 250 μs, and IQE fitting using PC1D revealed a BSRV of 330 cm/s. Using the SRA profile for phosphorus-diffused emitter, FSRV was determined at 1.1 E5 cm/s. The combination of these parameters leads to an accurate determination of final VOCJSC, and efficiency for this device, which matched the experimental I-V data. Since the Czochralski cell and the monolike cells were processed identically, the value of 330 cm/s BSRV was applied to determine local variation in τbulk for the representative monolike cells modeled here.
aBased on fill factor of 78.3% for interior cell and 77.5% for corner cell.
aBased on fill factor of 78.3% for interior cell and 77.5% for corner cell.
Monolike silicon wafers are a new type of wafers which shows a high promise for solar cell manufacturing. Specially oriented silicon wafers can be created from a seeded growth technology. However, some unfavorable multicrystalline may be also formed on monolike silicon wafers. Our results clearly indicate the detrimental effect of non-mono grains on the monolike silicon cells. There is no standard classification method for monolike silicon wafers yet, but it’s highly desirable to have over 90% monolike region on the monolike wafers [
The cell efficiency of monolike silicon solar cells is affected by the multicrystalline grain regions, especially the (311) oriented grains. There is relatively low density of dislocations in the mono (100) grain, but a significant higher density of dislocation in (311) grains. The (311) grains result in long leaf-like or shingle surface after alkaline texturing. PC1D modeling indicates a large variation in τbulk across monolike silicon wafer surface most likely due to variation in dislocation density. Both dislocation and large planar surface affect the cell efficiency by 0.86% (abs) decrease compared to the cells fabricated on more uniform mono wafers.
The authors would like to thank the Silicon Solar Consortium (SiSoC) members for providing experimental samples.