Scholarly article on topic 'Optimized Laser Doped Back Surface Field for IBC Solar Cells'

Optimized Laser Doped Back Surface Field for IBC Solar Cells Academic research paper on "Materials engineering"

CC BY-NC-ND
0
0
Share paper
Academic journal
Energy Procedia
OECD Field of science
Keywords
{"back surface field" / "laser doping" / IBC / "back contact" / "high efficency" / n-type}

Abstract of research paper on Materials engineering, author of scientific article — Morris Dahlinger, Kai Carstens

Abstract We present the optimization of the laser doped back surface field (BSF) for interdigitated back contact solar cells (IBC). The POCl3 flow limits the phosphorus concentration in the phosphorus silicate glass (PSG) during furnace diffusion, hence limits the sheet resistance when used as dopant source for laser doping. The saturation current densities of quasi steady state photo conductance (QSSPC) samples correlate with the sheet resistance dependent Auger contribution simulated with EDNA 2. Utilizing the measured saturation current density and contact resistance for various sheet resistances, we optimize the BSF doping for the recently presented 23.24% efficient laser processed IBC solar cell by numerical 3D solar cell simulation.

Academic research paper on topic "Optimized Laser Doped Back Surface Field for IBC Solar Cells"

CrossMark

Available online at www.sciencedirect.com

ScienceDirect

Energy Procedia 92 (2016) 450 - 456

6th International Conference on Silicon Photovoltaics, SiliconPV 2016

Optimized laser doped back surface field for IBC solar cells

Morris Dahlinger, and Kai Carstens

Institute for Photovoltaics, Pfaffenwaldring 47, 70569 Stuttgart, Germany

Abstract

We present the optimization of the laser doped back surface field (BSF) for interdigitated back contact solar cells (IBC). The POCl3 flow limits the phosphorus concentration in the phosphorus silicate glass (PSG) during furnace diffusion, hence limits the sheet resistance when used as dopant source for laser doping. The saturation current densities of quasi steady state photo conductance (QSSPC) samples correlate with the sheet resistance dependent Auger contribution simulated with EDNA 2. Utilizing the measured saturation current density and contact resistance for various sheet resistances, we optimize the BSF doping for the recently presented 23.24% efficient laser processed IBC solar cell by numerical 3D solar cell simulation.

© 2016 The Authors.Publishedby ElsevierLtd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.Org/licenses/by-nc-nd/4.0/).

Peer review by the scientific conference committee of SiliconPV 2016 under responsibility of PSE AG. Keywords: back surface field; laser doping; IBC; back contact; high efficency; n-type

1. Introduction

One of the challenges in back contact back junction solar cells (IBC) is to locally dope the emitter as well as the back surface field BSF, and, to define and allocate the contacts and metallization to both regions. We use laser processes for local doping and ablation. Recently, we presented a laser processed IBC solar cell with an efficiency -q = 23.24% [1]. The manufacturing process includes two laser doping steps: one dopes the boron emitter, the other one the phosphorus back surface field. Furthermore, a third laser step locally ablates the rear dielectric to define the contact areas to the emitter and the BSF; a fourth step uses laser ablation to structure the rear metallization. Apart from the geometry and the dimensions of the differently doped and contacted areas of the solar cell, the precise doping levels and profiles play an important role in the efficiency optimization.

One option to further optimize the cell efficiency, is to adjust the doping level of the emitter and the BSF. The cell consists of line shaped emitter and BSF regions, which are contacted by the metallization grid via point shaped areas through the rear dielectric stack. Arguing, that the dimensions and geometries of the differently doped regions, e.g. pitch, emitter and BSF width, are optimized within the technological limitations, the doping layout is kept

1876-6102 © 2016 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

Peer review by the scientific conference committee of SiliconPV 2016 under responsibility of PSE AG. doi: 10.1016/j .egypro. 2016.07.126

unchanged. The doping concentration affects the sheet resistance RshBSF, the saturation current density J0BSFpass in passivated and contacted J0BSFcont regions. Consequently, the total recombination changes, and the cell efficiency does. Furthermore, the surface doping of the BSF affects the specific contact resistance p between the metal and the semiconductor, which contributes to the series resistance, thus decreasing the fill factor and the cell efficiency.

This manuscript shows that the doping level of the BSF, characterized by the sheet resistance RshBSF, is controlled by the limited dopant source. The saturation current density of laser doped test samples follows the trend of the simulated Auger-contribution in the diffused region, suggesting that no traceable laser induced damage is present. We optimize the cell efficiency via 3D cell simulations by varying the doping concentration and the contact distance in the BSF region, utilizing fit functions to measured data.

2. Experimental

We measure the saturation current density J0BSF in the passivated and unpassivated regions of the BSF, and the specific contact resistance on test samples with differently laser doped surfaces, hence varied sheet resistances. In the experiment, we emulate the solar cell manufacturing process as close as possible. We use fit functions to the measured data, which are then used as input for the 3D simulation. Except the varied parameters, the simulated unit cell, as well as all the unchanged parameters, are identical to our manufactured IBC cell. Details of the manufacturing process and modelling of the best manufactured IBC cell will be published elsewhere.

2.1. Sheet resistance

Six inch pseudo square 3 Qcm p-type silicon wafers with the same surface treatment as the cell rear side serve as substrates. After removal of the natively grown silicon oxide in diluted hydrofluoric acid (HF), a high temperature diffusion furnace diffuses a shallow phosphorus doping in the wafer surface. During this process, a phosphorus silicate glass PSG grows. Four groups of wafers are processed with the POCl3 flow as the only varied parameter in the diffusion recipe. Subsequently, a nanosecond pulsed line focused laser beam irradiates areas of 2x2 cm2 with varied pulse energy density H on the wafers. After stripping off the PSG in 5% HF solution, a four-point-probe measurement measures the sheet resistance in the laser irradiated and non-irradiated areas.

Fig. 1. Sheet resistance Rsh of laser doped BSF depends on POCl3 flow and laser pulse energy density H before (a) and after (b) etch back and thermal oxidation. The Rsh does not change significantly by the etch back and oxidation. The dopant source is limited by the POCl3 flow, hence the sheet resistance Rsh. Once the threshold for doping is reached, the Rsh only slightly decreases with increasing pulse energy density H. The sheet resistance of the FSF (unirradiated) increases during etch back and thermal oxidation.

Figure 1(a) shows the sheet resistance Rsh, of laser doped BSF for varied POCl3 flow and laser pulse energy density H. Without laser irradiation, the furnace diffusion is very shallow and yields resistances of some hundred Q/sq, used as front surface field in the IBC cells. Exceeding the threshold laser pulse energy density H ^ 1.2 J/cm2 ,

additional doping is achieved. The sheet resistance drops until H ^ 1.5 J/cm2 and only slightly decreases for H< 3.5 J/cm2 . This saturation suggests that an increased pulse energy density H does not lead to a higher phosphorus incorporation in the laser molten silicon surface. The level at which Rsh saturates is limited by the POCl3 flow. A reduced POCl3 flow depletes the phosphorus contained in the PSG and in the shallow diffusion in the wafer, hence limiting the dopant source. The sheet resistance of the front surface field FSF increases with reduced POCl3 flow, since the phosphorus concentration in the shallow diffusion is reduced. Figure 1(b) shows the sheet resistances after further processing the samples similarly to the solar cells. The samples are etched back in an acidic silicon etching solution, and then thermally oxidized. The BSF sheet resistance does not change significantly, but the FSF resistance increases for the investigated POCl3 flows. The BSF already exhibits a rather deep diffusion after laser irradiation, hence the doped layer removed during etch back is negligible. Varying the etching duration, enables to adjust the FSF resistance and leaves the BSF basically unaffected (not shown here). Utilizing a POCl3 diffusion as dopant source, the laser doping provides an elegant way to tailor the FSF and BSF sheet resistance in IBC solar cells.

2.2. Specific contact resistance

A GP 4-Test pro measures the specific contact resistance p of evaporated aluminum contacts, utilizing the transfer length method (TLM). The same manufacturing process as for the cells are used, including chemical polishing, cleaning, furnace diffusion, laser doping, thermal oxidation, deposition of an insulation dielectric layer stack, and laser ablation. In contrast to the cell process, the aluminum is evaporated through a shadow mask, instead of laser structuring the aluminum layer. Prior measurement, the samples are annealed for 5 min at 350°C on a hot

2.3. Saturation current density

To evaluate the saturation current density of the BSF, we prepare symmetrically doped samples on the same 2 Qcm n-type CZ wafers as used for the solar cells. Four groups of wafers are diffused with varied POCl3 flow; subsequently, eight areas of 5 x5cm2 are laser irradiated full area on both sides. The laser pulse energy density H is varied in the range 1.83 J/cnP < H< 3.45 J/cm2. The same manufacturing process as for the cells are used, including chemical polishing, cleaning, furnace diffusion, laser doping, thermal oxidation. We omit the dielectric layer stack to ensure that the passivation layer can be removed chemically without altering the diffused surfaces. We measure the quasi-steady state photo conductance (QSSPC) with a WCT-120 Tester in the passivated state and after stripping the silicon dioxide (SiO2) passivation in HF solution. We extract the saturation current density J0BSF as proposed by Kimmerle et al. using equation (4) of Ref. [2].

sheet resistance [n/sq] sheet resistance R [o/sq]

Fig. 3. (a) Saturation current density J0BSF per side in passivated (SiO2) and unpassivated state. The passivated J0BSF follow the trend of the Auger contribution simulated for varied depths factors Z using EDNA2 [3]. A furnace diffused sample shows J0BSF in the same range as the laser doped. (b) The calculated Voclim and Vocjmpi match well for the QSSPC samples.

Figure 3(a) shows the saturation current densities J0BSF per side, either passivated with 20 nm SiO2 or unpassivated. Furthermore, the Auger contribution in the BSF is simulated using EDNA2 [3] for Gaussian doping profiles with depth factors Z = 0.3 to 0.8 and disabled Shockley-Read-Hall (SRH)-recombination at the surface or in the volume of the diffused region. The applied models in the EDNA simulation are listed in the appendix. Unfortunately, measured doping profiles of our samples are not available at present. Considering former measured doping profiles, the wide range in simulated depths likely cover the present profiles. The measured J0BSF follows the trend of the Auger recombination of passivated samples. Note, that the EDNA2 simulation was done for a temperature T = 300 K, whilst the measured J0BSF are evaluated for T = 298.15 K. With the effective intrinsic charge carrier density nieff as the main driver of the change in J0BSF with temperature [4], one can approximate the simulated Jobsf with J0BSF,300K/n2ieff,300K = JoBSF,298.i5K/n2ieff,298.i5K. The corresponding simulated Jo values for T = 298.15 K should be about 0.73 times lower than the presented ones. A furnace diffused (not laser irradiated) diffusion, highlighted as a star, passivated with the same thermal oxide, shows an offset in the same range as the laser doped samples. For more clarity, the data points are not labeled with their laser pulse energy densities, since no significant correlation between H and J0BSF is observed. In unpassivated state, the J0BSF increases with Rsh since the band bending induced by the doping decreases, thus minority charge carriers reach the highly recombination active surface more easily. Simulation results, setting the surface recombination velocity to vtherm = 1.7x107 cm/s (as the thermal velocity for hole at T = 300 K), well represent the trend of the measured J0BSF.

We calculate the open circuit voltage limit

Voc,lim = Vth I"

f J ^ 2 J

V 0 BSF y

here the thermal voltage at 298.15K is Vth = 25.69 mV, the short circuit current density is, in accordance with the QSSPC measurement, Jsc = optical factor x 38mA/cm2 = 26.6 mA/cm2, and the saturation current density J0BSF per side.

The implied open circuit voltage Voc,impl derives from

Voc,impl Vth

An (An + Nd )

li,eff

Here, An is the excess charge carrier concentration, ND the wafer base doping concentration, and nieff the effective intrinsic charge carrier concentration. Figure 3(b) shows, that the calculated Voclim and Vocimpl agree well, which validates the assumptions of negligible SRH-recombination in the wafer bulk for the J0 evaluation.

3. Optimization by 3D solar cell simulation

We model the current/voltage curve in the dark and with illumination, as well as the quantum efficiency of the solar cell with an efficiency -q = 23.24% presented in Ref. [1], using the 3D simulation tool QUOKKA 2 [5], following the guidelines given in [6]. The models used are listed in the appendix. The nieff is consistently used in the evaluation of the J0BSF and in 3D simulation. We fit the specific contact resistance (Figure 2) and the J0BSF data, passivated and unpassivated, representing the contacted area (Figure 3). The empiric fit functions are then swept in a simulation to optimize the BSF doping for IBC solar cells. The simulated unit cell neglects busbars and perimeter effects (charge carrier diffusion and recombination at the sample edges), and features line shaped doping with point contacts centered in the middle of the BSF and emitter regions. The circular point contacts are approximated by squares of the same area in the simulation. Arguing, that the unit cell is optimized within the technological limitation in precision, yield, and throughput, we vary the contact distance dcBSF along the laser doped line and the sheet resistance Rsh of the BSF. The specific contact resistance p and J0BSF follow the sheet resistance Rsh according to their fit functions. All other parameters are kept constant.

Table 1 shows the measured and simulated efficiency of the best manufactured laser processed IBC solar cells. The input parameters of the differently doped surfaces, emitter, FSF, BSF, and gap in the simulation are established the same way as described in this manuscript. The simulation results including perimeter effects, conducted with a multi-unit cell approach as described in [7], agrees well with the measured results. As simplification, the optimization of the BSF doping is done only excluding perimeter effects. Thus, the simulated efficiency of the fabricated cell excluding perimeter effects serves as reference for comparison.

Table 1. Characteristic current density/ voltage-values of the best measured laser processed IBC cell and their modelling results with and without perimeter effects.

Efficiency Open circuit voltage Short circuit current density Fill factor

^ [%] Voc [mV] Jsc [mA/cm2] FF [%]

Measured 23.24 ± 0.47* 681.6 ± 2.3 41.34 ± 0.79 82.47 ± 0.54

Simulation including perimeter 23.38 681.5 41.44 82.79

Simulation excluding perimeter 23.52 683.5 41.44 83.05

*certified measured by Fraunhofer ISE CalLab

Figure 4 shows the simulated efficiency depending on the contact distance dcBSF and sheet resistance Rsh. The modelled efficiency of the fabricated cell, neglecting perimeter effects, reached -q = 23.52%. The optimized efficiency r| = 23.56% with a gain A-q = 0.04% shows, that further efficiency gain by BSF doping optimization is unlikely. Instead, we find a wide parameter window for 200 ^m > dcBSF > 400 ^m and 50 Q/sq < RshBSF < 80 Q/sq with efficiencies -q > 23.5%. For RshBSF > 100 Q/sq, the efficiency drops due to the increasing contact resistance, hence a reduction of the fill factor. Consequently, further efficiency optimizations in the non pn-junction area need the change from line shaped BSF doping to point shape doping just underneath the contact areas.

Fig. 4. Simulated efficiency r) depending on contact distance dcBSF and sheet resistance ^shBSF of the BSF. There is a broad maximum in efficiency r) for 200 ^m > dcBSF > 400 ^m and 50 fi/sq < RshBSF < 80 fi/sq with an efficiency r) ^ 23.5%.

4. Conclusions

We have optimized the efficiency of our laser processed IBC solar cell using a 3D simulation, by changing the back surface field doping and point contact distance. The POCl3 flow during the growth of the PSG elegantly adjusts the sheet resistance of the laser doped BSF. The saturation current densities of laser doped passivated QSSPC test samples correlate with the Auger-contribution in the diffused surface. Empiric fit functions to measured saturation current densities in passivated and unpassivated state, and a fit to the sheet resistance dependent specific contact resistance serve as input for the 3D solar cell simulation. Compared to the simulation results of a manufactured solar cell, neglecting perimeter effects show minor efficiency gain. Instead, a wide parameter window of the BSF sheet resistance and contact distance is predicted.

Acknowledgements

The authors thank E. Hoffmann, H. Moldenhauer, L. Beisel, M. Saueressig, B. Lutz, L. Bauer, and B. Winter for technological support, as well as E. Hoffmann, R. Zapf-Gottwick, and J. H. Werner for carefully reading the manuscript and steady support. This work is funded by German Federal Ministry of Economics and Technology (BMWi) project No. 0325714A.

Appendix A. Models used in EDNA 2 and QUOKKA2

The effective intrinsic equilibrium charge carrier density nieff = 8.62*109cm"3 at a temperature T = 25°C is consistently used in the evaluation of the saturation current density J0BSF and Quokka 2 simulation.

The following lists the Models used in EDNA2 with the same terminology as in the software.

Radiative: Auger:

Mobility Model: Intrinsic band gap: Density of states: Dopant ionisation: Carrier statistics: Band gap narrowing: Temperature:

Trupke2003 fit Richter2012 Klaassen1992 Passler2002

Sentaurus2008 DOS Form. 2

Altermatt2006

Fermi-Dirac

Schenk1998

300.0K

Model and parameters used in QUOKKA2 with the same terminology as in the software.

Bulk.Auger:

Bulk.mobility:

Bulk.Brad:

Temperature:

Bulk.nieff:

Richter2012 Klaassen 4.73 x10"15 cm3/s 298.15 K

References

8.62x109 cm-3

[1] M. Dahlinger, K. Carstens, E. Hoffmann, S. Wansleben, J.R. Köhler, R. Zapf-Gottwick, J.H. Werner, 23.2% Efficiency with Laser Processed IBC Solar Cells, in Proc. 31th Europ. Photovolt. Solar Energy Conf.,; 2015; Hamburg Germany; p. 462-465.

[2] A. Kimmerle, P. Rothhardt, A. Wolf, R. A. Sinton, Increased Reliability for J0-analysis by QSSPC, Energy Procedia 2014; 55: 101-106.

[3] K.R. Mcintosh and P.P. Altermatt, A freeware 1D emitter model for silicon solar cells, in Proc. 35th IEEE Photovolt. Spec. Conf.; 2010; Honolulu, Hawaii, USA; p. 2188-2193.

[4] D. E. Kane and R. M. Swanson, Measurement of the Emitter Saturation Current by a Contactless Photoconductivity Decay Method, in Proc. 18th IEEE Photovolt. Spec. Conf.; 1985; Las Vegas, Nevada, USA; p. 578-583.

[5] A. Fell, A free and fast three-dimensional/two-dimensional solar cell simulator featuring conductive coundary and quasi-neutrality approximations, IEEE Transactions on Electron Devices 2013; 60; 733-738.

[6] A. Fell, K. R. McIntosh, P. P. Altermatt, G. J. M. Janssen, R. Stangl, A. Ho-Baillie, H. Steinkemper, J. Greulich, M. Muller, B. Min, K.C. Fong, M. Hermle, I.G. Romijn, M.D. Abbott, Input Parameters for the Simulation of Silicon Solar Cells in 2014, IEEE J. Photovoltaics 2015; 5; 1250-1263.

[7] A. Fell, K. Fong, K. R. McIntosh, E. Franklin, A.W. Blakers, 3-D Simulation of Interdigitated-Back-Contact Silicon Solar Cells With Quokka Including Perimeter Losses, IEEE J. Photovoltaics 2014; 4; 1040-1045.