The authors would like to thank Kevin Hart for film depositions and Steven Hegedus for helpful discussion. This work was partly supported by the National.
REAR SURFACE PASSIVATION OF INTERDIGITATED BACK CONTACT SILICON HETEROJUNCTION SOLAR CELL AND 2D SIMULATION STUDY 1,2
Meijun Lu , Ujjwal Das , Stuart Bowden , and Robert Birkmire Institute of Energy Conversion, University of Delaware, Newark, DE 19716 2 Department of Physics and Astronomy, University of Delaware, Newark, DE 19716 1
ABSTRACT Rear surface passivation by deposited intrinsic amorphous silicon (a-Si) buffer layer in interdigitated back contact silicon heterojunction (IBC-SHJ) solar cells significantly improves open circuit voltage (VOC) and short circuit current (JSC) but can lead to very low fill factor (FF) with an “S” shape J-V curve. In this paper, methods to optimize IBC-SHJ solar cell with improved FF are discussed and guided by two-dimensional numerical modelling. Two approaches to improve FF by modifying the buffer layer are evaluated: (1) increased conductivity, or (2) reduced band gap. Experimental results show that replacing the intrinsic a-Si layer in emitter with lightly doped p-type a-Si layer greatly improves fill factor, which is consistent with modelling prediction. However, the VOC and JSC are limited by the high recombination velocity of the unpassivated gap between the emitter and contact strips. The importance of gap passivation to achieve high efficiency in the IBC-SHJ structure was verified by 2D device simulation. The band gap of the intrinsic buffer layers have been reduced by changing the deposition conditions without substantially decreasing the passivation quality.
The IBC-SHJ cell reported earlier had a moderate 2 FF (74%), but relatively low JSC (28 mA/cm ) and VOC (605 mV) (see Fig.1), which is due to absence of intrinsic buffer layer passivation at the rear surface . Inserting an intrinsic a-Si buffer layer over the entire rear surface improves surface passivation, and results in a dramatic 2 improvement of VOC (691 mV) and JSC (>35 mA/cm on polished wafer), albeit with a low FF of 36% due to an “S” shape J-V curve  (also shown in Fig.1). However, the front junction SHJ cells fabricated with a-Si layers deposited with identical plasma conditions exhibited similar VOC but with much improved FF (74%) , implying necessity of further optimization of IBC-SHJ solar cells. In this work, optimization of IBC-SHJ solar cell performance with improved FF will be discussed, including two-dimensional numerical modeling and the guided preliminary experimental results.
INTRODUCTION A novel solar cell design, an interdigitated back contact silicon heterojunction (IBC-SHJ) solar cell was proposed recently . The IBC-SHJ combines the advantages of the interdigitated back contact (IBC)  and silicon heterojunction  solar cells. Having all the contacts at the back of the cell eliminates contact shading, leading to a higher short-circuit current (JSC). The rear junction also avoids trade-off between series resistance and reflectance, reducing the series resistance losses. It even simplifies cell stringing during module fabrication and improves the packing factor. Being a heterojunction device, IBC-SHJ also has the potential of higher open circuit voltage (VOC) due to the better surface passivation of the deposited amorphous silicon (a-Si) layer. The low temperature deposition, instead of the high temperature diffusion, decreases the thermal stress and reduces the bowing in thin wafers, which is the trend of future silicon solar cells. Also, the patterning is easier in heterojunctions than the diffused junctions in traditional rear junction cells since it is much easier to mask and etch depositions than diffusions, and further isolation between p/n a-Si layers is not always necessary.
Fig.1. Current density-voltage (J-V) curves for earlier reported IBC-SHJ solar cells with and without intrinsic a-Si buffer layer on the back surface [1,4]. TWO-DIMENSIONAL MODELING The interdigitated back-contact heterojunction device combines two technologies with challenging modeling requirements. Firstly, an all back contact device is an inherently two-dimensional structure that is difficult to model analytically or using a standard one-dimensional program such as PC1D. Secondly, heterojunction devices incorporating a-Si have also proven difficult to model, often requiring the development of new modelling programs.
The “Sentaurus Device” simulator (incorporating DESSIS) is extensively used for modeling opto-electronic devices and is well suited for modeling solar cells . The latest version includes complex defect models allowing the simulation of devices including a-Si. In the modelling, impurity scattering and carrier-carrier scattering were considered. The Auger and Shockley-Read-Hall recombination were modelled as a function of doping concentration. For a-Si, the critical parameters included in our study are energy distribution of the exponential band tails, and the Gaussian distribution of the mid-gap trap states. These are essential for the accurate modelling of any amorphous silicon device. In simulation, they are chosen based on reference  and were tuned to match the optoelectronic properties (e.g. dark and light conductivity) of our deposited a-Si layers. For c-Si/a-Si interfaces at the back surface, a thermionic emission model was used, in which the distribution function of the interface defect is modelled by two capture cross-sections, one for the holes and one for the electrons. A further benefit of the Sentaurus Device simulator is that it includes realistic optical effects such as absorption in the amorphous layer. An AM1.5G solar spectrum is used for the optical generation to simulate the J-V curve under standard one-sun illumination conditions 2 at intensity of 100 mW/cm . EXPERIMENTS The IBC-SHJ solar cells (Fig.2) were fabricated on n-type float zone Si (100) wafers with resistivity of ~ 1 Ω•cm. The rear interdigitated strips of n and p a-Si have the widths of 0.5 mm and 1.2 mm respectively and are used with and without undoped buffer layer. E-beam evaporated aluminium was used to contact the doped amorphous silicon strips. The front surface of the substrate was passivated with the intrinsic a-Si layer followed by the anti-reflection coating. All amorphous silicon depositions are performed in a direct current plasma enhanced chemical vapor deposition (DC PECVD) system at 200°C. We have also used RF plasmas (13.56 MHz) but found little practical difference . Presently the interdigitation is achieved through the use of photolithography, but the large millimeter-sized dimensions allow for other options to be investigated, e.g. laser patterning, ink-jet printing etc. The details of the fabrication process are reported elsewhere [1, 4].
Fig.2. Structure of IBC-SHJ cell with a-Si buffer layer on pstrip. In experiments, the a-Si buffer layer could also be on n-strip only or cover the whole back surface. IMPROVEMENT OF FF IN IBC-SHJ STRUCTURE SUGGESTED BY TWO-DIMENSIONAL SIMULATION In order to optimize IBC-SHJ solar cell performance with improved FF, simulations are first performed to guide the experiments. Both the experimental results and numerical simulation show that low FF with an “S” shape J-V appears only in the presence of intrinsic a-Si buffer layer on the p-type emitter strip. This suggests existence of hole transport barrier across the intrinsic a-Si buffer layer. Hence the variations of buffer layer are modeled: Conductivity of back surface buffer layer. In order to improve the hole transport across the buffer layer, the conductivity of the a-Si buffer layer in the emitter strip was varied by adding acceptor dopant. The device structure is similar to Fig.2., except that the front surface AR coatings and the intrinsic a-Si are replaced by a single passivation layer of SiNX (70nm). This ideal front surface simplifies the cell structure and it can still simulate the device performance very well with only a little difference in current due to the different anti-reflection effect. In the simulation, the gap between p- and n- strips is 100 µm, and an ideal surface passivation in the gap, with surface recombination velocity S equals zero, was assumed in the modeling. Poor passivation in the gap does not affect FF but deteriorates VOC and JSC and will be discussed later. The surface passivation effect of the buffer layers with different doping was assumed to be 11 same with a constant density of interface defects (1×10 -2 cm ). The simulated J-V curves under illumination are shown in Fig. 3, clearly indicate that the FF improves monotonically as the boron (p-type) dopant concentration (in other words, the conductivity) in a-Si buffer increases. But the VOC is similar since the same surface passivation quality was assumed. Increasing the doping concentration 18 -3 in the buffer layer to 2×10 cm , which is about one tenth 19 -3 of the doping level of the p-type emitter layer (2×10 cm ) used in simulation, FF attains a value of ~78% and the Sshape is removed.
This improvement in FF is due to modification in band alignment and offsets, which can be seen in Fig. 5 where the band diagrams with 1.72 eV (black color) and 1.65 eV (red color) intrinsic a-Si buffer layers are shown. It can be seen that the barrier for the minority carrier (hole) at valence band (EV) is less for the band gap of 1.65 eV, thereby resulting in higher FF.
Fig.3. Simulated J-V curves under illumination with varied boron dopant concentration in the a-Si buffer layer. By 18 increasing the doping concentration to 2×10 -3 cm , the FF attains a value of ~78%. Band gap of back surface buffer layer. The low FF and “S” shape J-V can also arise from the hole transport barrier due to enhanced valence band offset generated by the intrinsic buffer layer. Hence the effect of band gap of the intrinsic a-Si buffer layer on IBC-SHJ cell performance is studied. Again, the surface passivation effects of the different buffer layers are assumed to be same. Also, the electron affinity of the buffer layers are kept constant. Fig. 4 shows the illuminated J-V characteristics obtained from numerical 2D simulation with different band gap of intrinsic a-Si buffer. The figure shows that the band gap (Eg) of intrinsic buffer layer has a large effect on FF, which increases from 55% to >78% when Eg of the buffer layer reduces by 0.07 eV. However, the VOC was almost unaffected due to the assumption of the unchanged surface passivation.
Fig. 4: Simulated J-V curves for IBC-SHJ cells [structure shown in Fig. 2] under illumination with varied band gap (Eg) of the intrinsic a-Si buffer layer.
Fig. 5: Band diagram of the junction for two intrinsic buffer layer band gaps, 1.72 eV and 1.65 eV. The 1.65 eV case has smaller barrier at valence band which is easier for the carrier to transport. PRELIMINARY EXPERIMENTS TRYING TO IMPROVE FF IN IBC-SHJ SOLAR CELL Replacement of intrinsic a-Si buffer with lightly doped a-Si in IBC-SHJ structure. Guided by the simulation results, solar cells were fabricated to improve the FF. The conductivity of the back surface buffer layer was first varied. The IBC-SHJ solar cell structures were fabricated as shown in Fig. 2 with lightly boron doped a-Si buffer layers. The lightly doped aSi buffer was deposited with one tenth of the B2H6 flow rate than that of the p-type emitter layer, as guided by the simulation result depicted in Fig. 3. Fig. 6 shows the measured J-V curves under illumination, accompanied with cell structure. It can be seen that the FF improves dramatically by replacing intrinsic buffer with the lightly doped p-type buffer, which is consistent with numerical prediction. However, the VOC and JSC of the device is much lower compared to the device where the whole back surface including the gap between emitter and contact strip was passivated by intrinsic buffer (the red curve in Fig.6). The low VOC and JSC could be due to the poor passivation of the new buffer layer and/or an absence of gap passivation. The passivation effect of the lightly doped p-layer is studied by minority carrier lifetime test . The result is shown in Fig. 7, where the structure of the samples for lifetime test is also shown. It can be seen that the lifetime decreases with the addition of B2H6 in the
buffer layer, however, the implied VOC of lightly doped player is still as high as 660mV, indicating that very low VOC and JSC is not solely due to doped buffer layer. The lack of gap passivation is likely to lead to a reduced VOC and JSC. Poor gap passivation would result in recombination of carriers at the gap before they are collected at the junction. The importance of good gap passivation for high VOC and JSC was also verified by 2D device simulation, where the surface recombination velocity, S, of 100 µm gap between emitter and contact is varied. The result is shown in Fig. 8. When the S value increases from 150 cm/s to a moderate value of 5000 cm/s, VOC and JSC decrease ~30 mV and ~3 mA respectively. The simulation also demonstrates that the FF is unaffected by the gap passivation.
Fig. 8: Simulated J-V curves under illumination with varying surface recombination velocity in the gap between p- and n-strips. Larger gap recombination velocity would result in lower VOC and JSC. Cells incorporating improved gap passivation and lightly doped p-layer in the emitter strip are under further investigation. Decreasing band gap of intrinsic a-Si buffer.
Fig. 6: Measured J-V curves under illumination for solar cells with lightly p-type doped a-Si buffer layer on p-strips and intrinsic buffer layer on whole back surface. The device structures are shown in the figure.
Fig. 7: Minority carrier lifetime (injection level of 10 cm ) vs. B2H6 flow rate. Lightly doped p-layer still gives implied VOC of 660 mV. The structure for lifetime test is also shown in the figure.
According to the simulation result (Fig.4), narrowing the band gap of intrinsic a-Si buffer layer should also improve the FF. However, In order to keep the high VOC and JSC at the same time, passivation quality of the narrower band gap buffer layer should be comparable with our existing buffer layer. Intrinsic a-Si buffer layer with narrower Eg has been developed by varying plasma process parameters without alloying with other semiconductors like Ge. Fig. 9 shows the Tauc’s plot for two different a-Si layers, where the black curve represents the control buffer layer, and the red curve is the newly developed i-layer. It can be seen that the Eg of new i-layer has a band gap of about 0.03~0.04 eV narrower. The passivation quality of the new layer was determined from the effective minority carrier lifetime (τeff) measurement after depositing 8 nm a-Si layers on both sides of the wafer. The lifetime values shown besides the curves indicate that they have similar passivation quality. The solar cells with the narrower band gap buffer layer are presently under fabrication.
REFERENCES  M. Lu, S. Bowden, U. Das, and R. Birkmire, “Interdigitated back contact silicon heterojunction solar cell and the effect of front surface passivation”, Appl. Phys. Lett. 91, 2007 p. 063507.  M. D. Lammert and R. J. Schwartz, “The interdigitated back contact solar cell: a silicon solar cell for use in concentrated sunlight”, IEEE Trans. Electron Devices 24, 1977 p.337.  M. Taguchi, K. Kawamoto, S. Tsuge, T. Baba, H. Sakata, M. Morizane, K. Uchihashi, N. Nakamura, S. TM Kiyama, and O. Oota, “HIT Cells—High Efficiency Crystalline Si Cells with Novel Structure”, Prog.Photovolt: Res. Appl. 8, 2000 p. 503. Fig. 9: Tauc’s plot for control (black curve) and newly developed (red curve) intrinsic a-Si layers. They have similar passivation effects which can be seen from the effective lifetime shown beside the curves. CONCLUSION Two methods are proposed for the modification of rear surface buffer layer in IBC-SHJ structure by two dimensional numerical modelling; (1) increased conductivity (doping concentration), or (2) reduced band gap, to improve FF. Experimental results show that replacing the intrinsic buffer layer with a lightly doped ptype buffer layer greatly improves FF, consistent with simulation. The VOC and JSC are, however, limited by the slightly decreased passivation effect of lightly doped player and high recombination velocity in the unpassivated gap between p/n strips. The importance of gap passivation to improve VOC and JSC without affecting FF is also confirmed by 2D simulation. Band gap of the intrinsic buffer layers can be varied experimentally, and intrinsic buffer layers of narrower band gap with similar passivation quality have been achieved after process optimization. Cells with improved gap passivation and incorporating lightly doped p-layer and/or low band gap buffer layer in the emitter strip are under further investigation ACKNOWLEDGEMENT The authors would like to thank Kevin Hart for film depositions and Steven Hegedus for helpful discussion. This work was partly supported by the National Renewable Energy Laboratory under subcontract #ADJ-130630-12.
 M. Lu, S. Bowden, U. Das, and R. Birkmire, “a-Si/c-Si heterojunction for interdigitated back contact solar cell”, nd Proc. Of 22 European Photovoltaic Solar Energy Conference, 2007 p. 924.  M.Lu, S.Bowden, and R.Birkmire, “Two dimensional modelling of interdigitated back contact silicon th heterojunction solar cells”, Proc. of 7 International Conference on Numerical Simulation of Optoelectronic Devices, 2007 p. 55.  R. E. I. Schropp, and M. Zeeman, Amorphous and Micro-crystalline Silicon Solar Cells, 1998 p.183.  U.K. Das, M. Burrows, M. Lu, S. Bowden and R.W. Birkmire, “Surface passivation and heterojunction cells on Si (100) and (111) wafers using dc and rf plasma deposited Si:H thin films”, Appl. Phys. Lett. 92, 2008 p. 063504.  S. Bowden, U. K. Das, S. S. Hegedus and R. W. Birkmire, “Carrier Lifetime as a Developmental and Diagnostic Tool in Silicon Heterojunction Solar Cells”, WCPEC, 2006, p. 1295.