非晶-微晶异质结电池两维电脑模拟

TWO-DIMENSIONAL COMPUTER ANALYSIS OF a-Si:H/ c-Si HETEROJUNCTION SOLAR CELLS Muhammad Nawaz 1,2 and Arve Holt 2 1 University graduate centre (UNIK), P. O. Box 70, N-2027 Kjeller, Norway 2Institute for Energy Technology (IFE), P. O. Box, 40, N-2027 Kjeller, Norway Email: 1Nawaz@unik.no , 2Muhammad.Nawaz@ife.noABSTRACT: A theoretical design analysis using two dimensional computer aided design tool (i.e., TCAD) is presented for a-Si:H/ c-Si heterojunction solar cells. A set of device physical and geometrical parameters such as thicknesses, doping, bandgap discontinuities, contact resistivities, contact width, minority carrier life time, interface traps have been studied using AM1.5 solar spectrum at room temperature. A thermionic field emission transport model at the heterointerface is found to be critical for accurately predicting the solar cell efficiency comparable to that of real devices. With this, an efficiency over 20 % has been obtained over a range of band discontinuity values (ΔE C = x. ΔE g, where x is a variable in our simulation and ΔE g is the difference of the bandgap of two materials). Keeping almost the same efficiency performance, a bandgap discontinuity limitation is fairly relaxed by introducing a thin step graded a-Si layer which facilitates the transport of the carriers at the heterointerfaces. An n-doping in the c-Si absorber of 0.1 – 1 x10 17 cm-3 is found to be optimal for achieving maximum solar efficiency. A solar cell efficiency significantly degrades for interface traps density exceeding 5 x 10 11 cm -3 values. This efficiency decay is faster for larger trap cross-section (> 1.0 x 10 14 cm2). Keywords: Modelling, Solar Cell, Design, Simulation, Heterojunctions 1 INTRODUCTION The challenge of developing photovoltaic (PV) technology to a cost-competitive alternative for established energy sources can be achieved using simple, high-throughput mass-production compatible processes. Issues to be addressed for large scale PV deployment in large power plants are enhancing the performance of solar energy systems by increasing solar cell efficiency, using low amounts of materials which are durable, stable, and abundant on earth, and reducing manufacturing and installation cost. One very promising solar cell design to answer these needs is the silicon heterojunction (HJ) solar cell [1 - 6], of which the emitter and back surface field are basically produced by a low temperature growth of ultra-thin layers of amorphous silicon. In this design, amorphous silicon ( a-Si:H) constitutes both “ emitter ” and “ base-contact/back surface field ” on both sides of a thin crystalline silicon wafer-base ( c-Si) where the photogenerated electrons and holes are generated; at the same time, a-Si:H passivates the c-Si surface. Low thermal budget, excellent passivation property of a-Si material, high efficiency potential, simplicity of the manufacturing process (i.e., same process on front and back side of the wafer for a-Si deposition, TCO etc), excellent stability (i.e., avoids degradation: Staebler Wronski effect) are the main differentiating but motivating parameters for heterojunction solar design compared to the standard c-Si solar cell. Recently, cell efficiencies of 22.8 and 23 % (by Sanyo HIT design: [1,3]) have been demonstrated for double heterojunction (DHJ or HIT) solar cell structures of c-Si absorber of thickness 100 μ m and > 200 μm respectively. Depending on the growth condition, a range of band gap values have been reported for a-Si material. It is this layer which is sandwiched between the TCO (transparent conducting oxide) and c-Si absorber layer on front and back side of the device (i.e., thus forms a double heterojunction device: DHJ). While a range of uncertainties in the bandgap and electron affinity of TCO also exist, the solar cell efficiency is affected and often results in the non-uniform I-V characteristics of the device which is linked with transport at the heterojunction at the front and bottom side of the device. It is the purpose of this work to explore the range of bandgap discontinuity values where the device performs admirably well. The DHJ structures are further studies by varying doping and thickness of the c-Si absorber layer with different carrier life time and by inserting traps at the interfaces with different capture cross-sections. A commercial two dimensional computer aided design tool (i.e., Atlas from Silvaco [7]) has been used in this work. 2 DEVICE SIMULATION SETUP Solar cell structures used in the device simulation are shown in figure. 1. A 100 nm thick TCO layer is used at the top surface of the device. Detail of the modeling parameters is shown in table 1. A transfer matrix method is used as a beam propagation model which found out to be more suitable for multi-layer structures which takes into account coherence and interference effects. Disordered materials such as hydrogenated amorphous silicon ( a-Si:H) contain a large number of defect states within the band gap of the material. To accurately model the devices made of amorphous materials, a continuous density of defect states is used for a-Si. Here, the density of defect states (DOS) is described as a combination of exponentially decaying band tail states (one for donor like and another for acceptor like) and Gaussian distributions of the mid-gap states. The characteristic decay energy of conduction and valence band tails are 22 and 55 meV, respectively. Similarly, the characteristic decay energy of donor and acceptor type midgap Gaussian states is 15 meV each. The peak energy distribution of acceptor and donor type Gaussian states are 0.62 and 0.78 eV, respectively. Electron (hole) capture cross-section for acceptor tail and Gaussian states are set to 1 x 10 -16 (1 x 10 -14) cm 2. Similarly, hole (electron) capture cross-section for donor tail and Gaussian states are set to 1 x 10-16 (1 x 10-14) cm2. A set of physical models implemented for device simulation include bandgap narrowing effect, doping and temperature dependent mobility models and using Fermi 25th European Photovoltaic Solar Energy Conference and Exhibition /5th World Conference on Photovoltaic Energy Conversion, 6-10 September 2010, Valencia, Spain412Dirac statistics. Defect recombination at the a-Si/c-Si interface and at the surface (i.e., interface of TCO and a-Si:H) is modeled by the recombination velocity parameter. Defect recombination in the bulk and at the surface (i.e., semiconductor – semiconductor interface and semiconductor-conductor interface) is modeled with the Schokly-Read-Hall (SRH) recombination. Furthermore, Auger recombination is also taken into account using default model values. Figure 1: A simulated layer structure along with photogenerated profile (top) and complete DHJ layer design (bottom) Table 1: Parameter detail of different materials used in the device simulation Thermionic field emission coupled with the tunneling model for transport at the heterointerface is specially implemented. Simulation data presented here deals with AM1.5 solar spectrum and at room temperature. 3 RESULTS AND DISCUSSION Influence of the contact resistivitiy and contact width on solar cell performance is illustrated in figure 2. The efficiency remains approximately unaffected with varying contact resistivity however, the forward current density decreases with increasing resistivity values. A decay in the efficiency both for single heterojunction and double heterojunction is noticed with increasing contact width as a result of decrease in the photogenerated current density due to more shadowing of the metal contact. Figure 2: The current density versus voltage for different contact resistivity values (top) and efficiency versus contact width (bottom) for a-Si/c-Si heterojunction solar cells Efficiency as a function of bandgap parameter is illustrated in figure 3. With fixed bandgap discontinuity (ΔE C = x. ΔE g and ΔE V = (1- x). ΔE g where x is variable in our simulation and ΔE g is the difference of the bandgap of two materials) and using standard transport model, we noticed that the HIT solar cell performance is undervalued (i.e., current density Jsc of 32 mA/cm 2, open circuit voltage Voc of 0.73 V, and efficiency η of 18.5 %). This is the result of the poor photo-generated carrier collection efficiency due to the presence of heterojunction at the respective n and p contacts of the device. While smooth I-V characteristics have been obtained, varying bandgap discontinuities does not influence the efficiency values (see fig. 3). Using thermionic field emission tunneling model at the heterojunction interfaces, we observed an improved performance (i.e., Jsc of 40 mA/cm 2, Voc of 0.73 V, and η of 23.8 %), comparable to that of Sanyo HIT module. Using tunneling model at the heterojunction, we noticed that the device electrical behavior is smooth over limited range of band gap discontinuities (i.,e varying x from 0.4 to 0.6 or 40 % to 60% of the bandgap difference of the two materials). While Voc remains approximately the 25th European Photovoltaic Solar Energy Conference and Exhibition /5th World Conference on Photovoltaic Energy Conversion, 6-10 September 2010, Valencia, Spain413Figure 3: Efficiency as a function of band discontinuity parameter with and without tunneling model and with/without grading Figure 4: Efficiency (a), open circuit voltage (b) and current density (c) as a function of wafer thickness for different carrier life time. Also shown is the efficiency versus cSi thickness (d) for different n-doping. same with varying bandgap discontinuities, efficiency drops due to the large values of the bandgap discontinuity. This drop is either due to the large ΔE C(electron potential barrier) or large ΔE V (holes potential barrier) values. Introducing further thin grading layers with increasing bandgap from the c-Si to amorphous Silayer at two sides of the junction ( p-i interface at a-Si/c-Siside and i-n interface at c-Si/a-Si side), I-V characteristics were further improved over large range of the bandgap discontinuities. The grading of the bandgap of a-Si layer with increasing bandgap value from top side of c-Si/a-Si interface is possible with fine tuning the growth conditions of PECVD. Thin graded layers indeed facilitates the transport of the carrier at both sides of the device and possible high V OC may be obtained for top layer of large bandgap value. In order to reduce the power generating cost, possibility of using thin a-Si/ c-Si heterojunction solar cells need to be explored. Since the HJ cell structure is symmetrical, the effect of thermal stress buildup is fairly relaxed and hence bowing of the thin wafer is also diminished. However, the use of thin wafer for improved efficiency potential of HJ cell is only realizable with excellent surface passivation. Efficiency versus wafer thickness for different minority carrier life is shown in Figure 4. Minority carrier lifetime is very important in solar cells, as longer lifetimes lead to higher cell efficiency caused by enhanced charge collection at the cell terminals. For a given minority carrier life time, the efficiency remains approximately constant in the thickness range from 100 to 300 μ m, while a significant drop in efficiency is noticed for very thin wafers (< 50 um). For a given thickness, the efficiency increases with the increase in carrier life time as expected. The low efficiencies for thin wafers are resulting from the decrease in JSC for decreasing wafer thicknesses (fig. 4c). This is simply a consequence of the reduced optical absorption in very thin wafers. At the same time, an increase in V OC can be seen for decreasing wafer thicknesses. The increased V OC dampens the efficiency drop for thin wafers, as it opposes the reduced J SC. This effect was also reported by Taguchi et al., when demonstrating high efficiencies on < 100 μ m HIT solar cells[1]. Our simulation predicts a fill factor in the range 80-85 % for most samples, except for the very high minority carrier life time (10 ms) of c-Si material where the fill factor was around 85 - 87 %. Experimentally FF up to 80 % has been achieved for high efficiency DHJ cells. The very high FF for the 10 ms carrier life time wafers is a result of (too) high V OC and (too) low JSC. This gives the current-voltage curve an unrealistic high squareness. Efficiency as a function of wafer thickness for different p-doping is shown in figure 4d. For a given wafer thickness, efficiency first increases with the increase of doping, reaches a maximum value and then drop for very high doping in the c-Si material. The first improvement in efficiency is resulted because of slight increase in the V OC (i.e., improved built in voltage due to doping) and fill factor with the doping. For very large doping, the majority of photogenerated carriers is uncollected and produces significant recombination at the interface and in the bulk region of the device and hence efficiency drops. Due to different bandgap and diss-similarities of material properties (of materials like for example TCOs, a-Si, c-Si) at front and bottom side of the device, clean 25th European Photovoltaic Solar Energy Conference and Exhibition /5th World Conference on Photovoltaic Energy Conversion, 6-10 September 2010, Valencia, Spain414and good quality heterointerface with minimum surface state densities is prerequisite step for improved efficiency performance of heterojunction solar cells. The interface properties were studied by inserting interface traps at Figure 5: Efficiency as a function of interface trap density (a) and trap capture cross-section (b). Also shown is the photogeneration rate at low trap density (c: 1 x 10 10cm-2) and high trap density (d: 1 x 10 12 cm-2) fixed locations such as semiconductor-semiconductor and semiconductor-conductor interfaces of the device. It is well known that the minority carrier lifetime depends on the number and energy levels of recombination centers or traps. Single acceptor trap level has been introduced at 0.7 eV below the conduction band. Note that, the trap centers, whose associated energy lies in a forbidden gap, exchange charge with the conduction and valence bands through the emission and capture of electrons. Depending on the location, the trap centers influence the density of space charge in semiconductor bulk, at the interface and the recombination statistics. Influence of trap densities and trap cross-section is shown in figure 5. Both parameters are linked with the carrier life time through Schokly Read Hall recombination process in the used model [7]. With fixed trap cross-section and recombination velocity, efficiency remains almost constant upto 1.0 x 10 11 cm-2. A drop in efficiency is noticed for larger trap densities. For constant trap densities, efficiency decreases with the increase of the trap cross-section. A decay in the efficiency is faster for DHJ than SHJ device where the drop in efficien