Shining Some Light on Epitaxial Thin-film Silicon Solar Cells
Epitaxial silicon solar cells are based on a thin high-quality epitaxial layer on top of a highly-doped, inactive silicon substrate. Solar cell processing, although with its own specificities, can be similar to what is used for standard double-side contacted bulk silicon solar cells; therefore epitaxial silicon solar cells are also referred to as wafer-equivalent (EpiWE) . A cross-section of a typical epitaxial silicon solar cell can be found in Fig. 1.
|Figure 1. Cross-section of a typical epitaxial silicon solar cell with an epitaxial stack consisting of a BSF, a p-type active base and an n-type emitter.|
Depending on the quality of the epitaxial layer, a thickness of 20µm is usually considered as optimal for reducing the cost per Wp. Such a very thin active layer limits light absorption, resulting in a reduced cell performance. Therefore, implementation of a very specific light confinement structure is needed. For example, a light confinement structure consisting of a textured front surface and a porous silicon (PSi) intermediate reflector can be introduced .
Porous silicon reflector
The role of the PSi reflector is to boost the absorption in the active epitaxial layer of low energy photons, thus increasing the charge carrier generation and resulting in higher short-circuit currents of the cell. The reflector is located at the substrate-epi interface, as can be seen in Fig. 1, and is etched electrochemically into the highly-doped substrate. A very good control of the porosity and the etching depth is needed to form a reflective stack of PSi multi-layers with alternating high and low porosity. The refractive index of the layer is changed, but the optical thickness is kept constant. By applying the quarter wavelength principle, it’s possible to create constructive interference of the reflected light for a certain selected wavelength range (Bragg effect). Different types of these “Bragg reflectors” are evaluated. Reflection of the light over 90% for the targeted wavelength range can be obtained in this way (Fig. 2) .
|Figure 2. Total reflection of an epitaxial silicon solar cell with different types of porous silicon Bragg reflectors at the epi-surface interface. For the wavelength interval between 850nm to 1000nm, a reflection over 90% is obtained.|
Epitaxial active layers
After etching a reflector, the wafers are ready for deposition of the active layers. Layers are grown epitaxially on top of the PSi reflector. Therefore it is crucial that the top-surface is closed before growth starts. This is realized by applying a high-temperature bake, done in situ, just before the epitaxial growth. During this bake, the pore structure changes from columnar to large closed voids with a very thin dense layer at the surface (Fig. 3). Optical properties remain almost identical, but the latter structure makes it possible to grow a very high quality epitaxial stack.
Chemical-vapor-deposition (CVD) epitaxial growth is done at an atmospheric pressure at high temperature (1130°C) with trichlorosilane as silicon precursor: this makes it possible to obtain very high growth rates around 4µm/min. Boron doping is used for p-type regions. For n-type regions, arsine (instead of phosphorus which is more common in solar applications), is used for n-type doping to limit the out-diffusion at this high temperature. The CVD process as it is used nowadays is an expensive process, but it allows for the demonstration of the potential of the concept maximum freedom in the doping profile of the epitaxial stack that is applied . Also, n-type and p-type regions can be designed as desired and therefore, a complete stack consisting of a back surface field (BSF), base, emitter, and front surface field can be grown in situ. POCl3 emitter formation and the extensive cleans before and after emitter formation can be avoided. Several cell concepts with varying epitaxial stacks can be assessed. In this article, we will concentrate on two cell concepts: a p-type-base solar cell and an n-type-base solar cell.
|Figure 3. Reorganization of the porous silicon stack (left: as grown; right: after reorganization) occurs during the high temperature bake in situ just before epitaxial growth of the active layers. Optical properties of the PSi stack remain unaffected.|
Textured front surface
After epitaxial growth, which results in a very smooth top surface, roughening or ‘texturing’ of the front surface is done. To obtain this texturing, processes that are identical to those used for bulk silicon solar cells can be considered, but silicon removal of the expensive epitaxial stack must be limited to a minimum. Two different texturing processes were applied on the epitaxial layers: plasma texturing  and formation of random pyramids .
|Figure 4. SEM pictures of epitaxial layers after application of the two types of front-surface texturing methods that can be applied for epitaxial silicon solar cells: plasma texturing (left) and random pyramids (right).|
Plasma texturing has the advantage of being a dry process that is applied on one side. It can be tuned to obtain a reflectance below 16% at a wavelength of 600nm by removing only 1µm of silicon. Moreover, the process is independent of the crystal orientation, which makes it possible to process all kinds of substrates including multi- or even polycrystalline silicon. The resulting textured surface is fractal, which makes it difficult to passivate it efficiently. Figure 4 (left) shows a SEM picture of the epitaxial surface after plasma texturing.
Random pyramids are formed by wet etching in an alkaline solution (NaOH). Lower reflectivity can be obtained compared to plasma texturing, but the process can only be used on monocrystalline substrates (and therefore only for the demonstration of the potential of the concept), and a layer of around 4µm of silicon is removed on average during the formation of the pyramids. The main advantage of this process is the smoothness of the structures that are formed: at microscopic level, the surface is smooth (Fig. 4, right) and deposition of additional layers and/or passivation of this surface are easier than for plasma-textured surfaces.
Other steps of the solar cell process are identical to standard processes used for bulk silicon solar cells. In this article, we will describe two solar cell concepts: a p-type base solar cell in combination with plasma textured front surface, and an n-type base solar cell in combination with a random pyramid textured front surface.
p-type base solar cells with plasma textured front
In Fig. 5 (left), a cross-section of a p-type base solar cell is depicted. A porous silicon Bragg reflector is etched in highly-doped substrates. Two types of substrates are used: high-quality (but still highly-boron-doped) monocrystalline substrates to demonstrate the potential of the concept, and low-quality multi-crystalline substrates (from upgraded-metallurgic-grade Si), considered as potentially low-cost. After high-temperature anneal, an epitaxial stack is grown. A highly-doped (1E19 at/cm3) boron layer of 2µm acts as a BSF, and is followed by the active base (boron-doped 1E16 at/cm3) of 20µm. The switch to n-type doping is made in situ and an emitter is formed by only adding a few seconds to the whole process. This is an arsine-doped region of 2µm with a doping concentration of 1E18 at/cm3.
|Figure 5. Cross-section of a p-type base solar cell with plasma textured front surface (left) and an n-type base solar cell with a random pyramid textured front surface (right).|
Plasma texturing is performed on this substrate, removing 1µm average of the original 2µm-thick epitaxial emitter. A shallow front surface field is obtained by POCl3 diffusion. As with bulk silicon solar cells, front-surface passivation and anti-reflection coating are obtained by deposition of a silicon nitride layer.
For wafers on high-quality material, metallization is done with copper plating. For cells on low-quality multi-crystalline substrates, the metallization is realized with screen printing. The most important solar cell parameters for these two kinds of cells are summarized in the table.
On mono-crystalline substrates, a record efficiency of 16.2% is obtained . For low-cost substrates, with cells realized with processes that are fully compatible with standard industrial bulk solar cell processes, a 14.7% efficiency is obtained . The efficiencies obtained on epitaxially-grown “wafer equivalent” substrates are getting one step closer to the standard bulk silicon technology.
n-type cells with random pyramids
In Fig. 5 (right), a cross-section of an n-type base solar cell is shown. The porous silicon reflector formation and reorganization is identical to the p-type base solar cell concept.
For this concept, the p-type region, on top of the highly-doped BSF, will now act as a rear emitter with a boron doping concentration of 1E18 at/cm3, and is therefore limited to 2µm and has a doping concentration of 1E18 at/cm3. The n-type region now acts as the base and is originally 37µm-thick. Base doping is around 1E16 at/cm3.
Because the emitter is located at the rear, more process freedom is possible on the front. Random pyramids can be applied − in this case, removing 4µm from the n-type base. A front surface field is applied with a POCl3 diffusion without drive-in step, leading to a very shallow highly-doped n+ region of 50nm.
After passivation of the front surface, contacting is realized with evaporation of Ti/Pd/Ag on the front side and Al on the rear side. An efficiency of 16.9% is achieved with this solar cell concept .
Both examples show the flexibility of epitaxial “wafer equivalent” and the ease with which they can be processed with commonly known processing techniques. However, several challenges still have to be tackled before this concept can be successfully introduced to the industry. The availability of large-area cost-effective substrates is one of them. Currently, lower-purity materials such as UMG (upgraded metallurgical silicon material) and off-spec (out-of-spec recast multi-crystalline silicon material) are available only in small amounts. However, real price reductions can be obtained by fabricating cheap silicon sheets, avoiding the expensive sequence of the Siemens process, casting of the material and slicing of the wafers. Pilot projects exist, such as RGS (ribbon growth on substrate) and pressed poly-Si material, but do not produce wafers on an industrial scale . High-volume/large-area manufacturing of this type of highly-doped substrates still has to ramp up.
The second main challenge is the availability of high-throughput processing tools for both the porous silicon etching and the Si epitaxial CVD process. CVD reactors currently on the market are not designed to grow thick (20µm) epitaxial stacks in a fast way. Most of them are even single wafer reactors that cannot be used in an industrial solar cell line. Several institutes and companies are investigating this issue. At Fraunhofer ISE, the most encouraging prototype, an in-line atmospheric-pressure CVD reactor is being built .
1. S. Reber, et al., EU project SWEET on epitaxial wafer equivalents: results and future topics of interest., Proc. of the 21st EUPVSEC (2006), Germany, p 570.
2. F. Duerinckx, et al., Simulation and Implementation of a porous silicon reflector for epitaxial silicon solar cells, Prog. Photovolt: Res. Appl. 2008; 16: pp 399-407.
3. I. Kuzma-Filipek, et al., Chirped porous silicon reflectors for thin-film epitaxial silicon solar cells, J. Appl. Phys., vol. 104, no. 7, pp 073529-1 (2008).
4. K. Van Nieuwenhuysen, et al., Epitaxially grown emitter for thin film crystalline silicon solar cells, Thin Solid Films, 517 (2008) p383-384.
5. F. Duerinckx, et al., Light trapping for epitaxial thin film crystalline silicon solar cells, Proc. 20th EUPVSEC, (2005), Spain, p 1190.
6. Kuzma-Filipek, et al., >16% Large area epitaxial silicon wafer equivalent solar cells with rear epitaxial CVD emitters and Cu-based metallization, submitted to 26th EUPVSEC Hamburg, Germany 2011.
7. M. Recamán Payo, et al., Screenprinting process on 20um thick epitaxial multi-Si solar cells (efficiencies up to 14.5%), Proc. 25th EUPVSEC, (2010), Spain, p3646.
8. K. Van Nieuwenhuysen, et al., Epitaxial thin film silicon solar cells with efficiencies up to 16.9% by combining advanced light trapping methods and CVD emitters, Proc. 24th EUPVSEC (2009), Germany, p 2357.
9. G. Hahn et al., New crystalline silicon ribbon materials for Photovoltaics, J. Phys.: Condens. Matter 16, R1615-R1648 (2004).
10. S. Reber, et al., ConCVD and ProConCVD: development of high-througput CVD tools on the way to low-cost silicon epitaxy, 24th EUPVSEC (2009), Germany, p2560.
By Els Parton, Kris Van Nieuwenhuysen, Izabela Kuzma-Filipek, Jan Van Hoeymissen, Maria Recaman, Sivaramakrishnan Radhakrishnan Hariharsudan, Frédéric Dross, Jef Poortmans, Imec, Leuven, Belgium
May 16, 2011
Els Parton received her engineering degree and PhD in biological sciences at the Katholieke U. Leuven (K.U.Leuven), Belgium and is a scientific editor at imec, kapeldreef 75, B-3001 Heverlee, Belgium, +32 16 281467, firstname.lastname@example.org
Kris Van Nieuwenhuysen obtained her engineering degree in electronics at the De Nayer Institute for Science and Arts, Belgium, and is a Researcher Specialist at imec.
Izabela Kuzma-Filipek received her masters degree in material science at the U. of Science and Technology, Cracow, and her PhD from the Katholieke U. Leuven, Belgium, and is a Researcher Specialist at imec.
Jan Van Hoeymissen obtained a PhD in physical chemistry from the U. of Leuven, Belgium and is a Researcher Specialist at imec.
Maria Recaman received a PhD in chemical engineering from the Complutense U. of Madrid, Spain and is a Researcher Specialist at imec.
Hariharsudan Sivaramakrishnan Radhakrishnan received BEng (Hons.) in electrical and electronics engineering at Nanyang Technological U. and is a PhD student at imec.
Frédéric Dross received his PhD from the Ecole Nationale Supérieure des Télécommunications (ENST), Paris, France and is a principal researcher in the photovoltaics department at imec.
Jef Poortmans received his degree in electronic engineering and his PhD from the Katholieke U. of Leuven, Belgium. He is Program Director of the Strategic Program SOLAR+ at imec.