异质结电池表面钝化的研究

SURFACE PASSIVATION OF MULTI-SI FOR HETEROJUNCTION SOLAR CELLS

A.S. Asha*, M. Canino*, C. Summonte*, S. Binetti, M. Acciarri, J. Libal, D.Cavalcoli∇, A. Cavallini∇

*CNR-IMM – via Gobetti 101 – 40129 Bologna – Italy

Università Milano Bicocca – Milano, Italy

∇

Physics. Dept, University of Bologna, viale C Berti Pichat 6/II I-40127 Bologna, Italy

ABSTRACT: We studied the passivation effect of plasma deposited intrinsic and doped hydrogenated amorphous silicon layers on industrial grade, n-type, 1 Ω⋅cm multicrystalline silicon, aimed at the fabrication of Heterojunction (HJ) solar cells based on such material. The investigated variables include the process temperature, hydrogen dilution of silane during the intrinsic layer deposition, and passivating layer thickness. The structure of the passivating layers was analysed by optical measurements. Surface photovoltage and quasi-steady state photoconductance were used to characterize untreated and treated wafers. Thin a-Si:H layers exhibited the best passivating performance. Keywords: Heterojunction, Multicrystalline Silicon, Passivation.

1 INTRODUCTION

The application of the heterojunction (HJ) scheme to multicrystalline silicon combines several advantages over other technologies and/or substrates, that span from moderate cost of the substrate, low temperature processing, compatibility with thin substrates, reduced fabrication time, efficiency in the medium-high range for crystalline silicon, relative matureness of the technology that makes it an appealing candidate for short term transfer to production. In particular, the medium-high efficiency range is strictly linked to the possibility of obtaining efficient surface and grain boundary passivation, which has indeed recently raised multicrystalline silicon to compete with the more costly monocrystalline silicon substate [1].

We analysed the passivating properties of a thin intrinsic (amorphous or epitaxial) layer placed between the bulk base and the emitter, with respect to its thickness and deposition temperature. The transport mechanisms of an HJ device fabricated by employing such a technology were studied by dark J-V measurements.

2 EXPERIMENTAL

The material used for this study is industrial grade, n-type, 1 Ω⋅cm, 300 µm thick multicrystalline silicon. The substrates were etched to remove the saw damage in a solution of HF(50%):HNO3(65%):CH3COOH(100%)= 3:43:7 for 20 minutes. The passivating layers were deposited on 5x5 cm2 samples by Plasma Enhanced Chemical Vapour Deposition (PECVD) at 13.56 MHz frequency and 4W rf power.

In a p/n HJ solar cell, the junction is basically formed by depositing an i+p double layer on the substrate front surface. To reproduce this situation, for the analysis of the passivating properties of the intrinsic layer we fabricated symmetrical structures by depositing an i+p double layer on both sides of each sample. The deposition parameters analysed were i) the hydrogen dilution in the silane mixture: either no hydrogen, which results in an amorphous passivating layer, or a mixture of SiH4:H2 = 0.8:108, which, in the case of (100) CZ silicon, is shown to result in an epitaxial layer even at temperatures as low as low as 120 °C [2,3]; ii) the deposition time, 1’30’’ or 3’’; iii) the deposition

temperature, from 100°C to 200 °C. The p-type layer was deposited under 0.95 hPa partial pressure, same substrate temperature as the passivating layer, 3’26” deposition time, SiH4:B2H6:H2 = 2.5:0.04:21 sccm gas flow rates. In order to study the transport properties, a complete p/n device was fabricated with the following scheme: deposition of a thin a-Si:H layer at 160°C, followed by deposition of the p-type layer on the front side; deposition of an n-type layer at 16W rf power and 13.56 MHz frequency, 170°C substrate temperature and 0.6 mbar partial pressure in a mixture of PH3:H2=2:100 for 10 minutes on the reverse side. As antireflecting transparent conducting oxide, ITO was deposited by sputtering through a 1x1 cm2 shadow mask [4]. An Al front grid, and Ag on the back side, were then evaporated to form the device contacts. The i+p double layers were characterized by reflectivity measurements, Surface Photovoltage (SPV) and Quasi-steady state photoconductance (QSSPC) measurements. Details on the SPV apparatus are reported in [5]. The complete p/n device was characterized by dark J-V measurements.

3 RESULTS

3.1 Optical measurements

The hemispherical reflectance was measured as a function of the incident light wavelength using an integrating-sphere system to minimize any signal loss due to the sample surface roughness. The curves were simulated by Optical [6], using the film/substrate model, in order to obtain information about the deposited film structure. The film was modeled using the Effective Medium Approximation including a-Si:H and void fraction, whereas the c-Si optical parameters were used for the substrate. Total deposited thickness and void fraction were the free parameters in the simulation. No distinction between passivating layer and p-layer was attempted.

Fig.1 shows the measured and the simulated hemispherical reflectance of samples passivated at increasing deposition temperature. The void fraction and the (i+p) film thickness obtained by Optical simulations are also reported in the figure. The 362 nm peak typical of c-Si is visible in the as-etched sample. This peak is masked by the a-Si:H film, highly absorbing at this wavelength, in all other cases. The weak R interference

maximum, located at 550 nm for the 33 nm thick film, shifts to 700 nm due to thickness increase up to 60 nm.

The simulation shows an increase of deposition rate with temperature, and a decrease of void fraction,

representative of lower hydrogen content.

shown in Fig. 2.

Note the very good agreement of the thickness determined by optical simulation with that observed by SEM (47 and 50 nm in this case).

3.2 QSSPC measurements

Quasi Steady State Photoconductance (QSSPC) measurements [7] were made before and after the i+p double layer deposition. In the first case, the samples were passivated by means of a methanol-iodine solution., in order to minimize the surface contribution and determine the bulk diffusion length. The surface recombination velocity Seff after the PECVD treatment can be obtained by comparison between the two measurements using equation

(1):

(1)

where τ b is the bulk lifetime, τ eff is the lifetime value measured on the samples passivated as described above, and W is the wafer thickness [8].

Fig. 3 shows the Seff as a function of excess carrier density, obtained by QSSPC, measured on two samples that differ by the thickness of the passivating layer. The deposition temperature was 120 °C. For the sake of comparison, the two samples were cut off the same multicrystalline silicon wafer, that, before the PECVD treatment, showed 57±1 µs recombination time at low injection, as measured by QSSPC using the iodine-methanole solution. As the illuminating flash is wider than the sample size, the result is representative of an average over the entire sample surface. Note that both τeff and τb are taken as dependent on injection in Eq. (1), yet, an Seff independent on ∆p is observed in Fig.3. The surface recombination velocity observed in the case of 9 nm is a factor of 3 lower than in the 22 nm thick passivating layer, indicating that once the passivation of the surface states has been achieved, further treatment is detrimental. In the design of heterojunction solar cells, this allows keeping the passivating layer thickness to a minimum, thus avoiding series resistance and fill factor degradation.

Figure 1: Hemispherical reflectivity curves of samples obtained at different deposition temperatures (black: experimental data; red: result of the simulation). The as-etched sample is reported for comparison. The thickness and void fraction used in the simulation are also indicated.

Figure 2: Cross section SEM micrograph of the sample obtained at 120 °C

The simulated reflectance overestimates the actual reflectance for the as-etched sample, evidencing the antireflecting properties of the surface roughness of the untreated wafers. The effect is somewhat reduced for increasing thickness of the i+p double layer. Indeed cross section Scanning Electron Microscopy (SEM) images of the samples show that the deposited layers act in the sense of reducing the cuspidic surface shape, as it is

Figure 3: Surface recombination velocity as a function of minority carrier injection. The curves are labeled with the passivating layer thickness.

Fig. 4 reports the surface recombination velocity for two samples fabricated using different substrate temperature, computed as described above. Again, the 5x5 cm2 samples were cut off the same mc-Si wafer, that

showed an initial bulk recombination time τ b = 25±1 µs at low injection, as measured by QSSPC using the iodine-methanole solution. The result obtained on the sample that includes the epitaxial passivating layer is

also reported.

Figure 4: Surface recombination velocity as a function of minority carrier injection for mc-Si wafers passivated using a-Si:H layers obtained at different temperatures. The values obtained using an epitaxial layer are reported for comparison.

The lifetime data obtained by QSSPC as a function of deposition temperature of the passivating layer are reported in Table I. It is obtained that a best temperature exists, that is attributed to a trade-off between the decrease of Si-H2 complex related defects, and insufficient hydrogen defect passivation. Similarity of the passivation properties of the amorphous layer, and the epitaxial layer when fabricated under appropriate conditions, has already been shown in ref [2].

Table I: Effective lifetime as a function of deposition temperature of the passivating layer.

T τ eff °C µs 100 9.7 ± 0.5

120 10.0 160 12.5 200 9.8

3.3 Surface Photovoltage measurements With reference to the samples reported in Fig. 3, Fig. 5 shows the SPV signal [9] as a function of incident radiation energy measured on the sample with 9 nm passivating layer. The actual absorption profile of the sample was computed using Optical. It is obtained that the inverse absorption coincides with what expected for bare silicon. In fact, in the wavelength range used for SPV measurements, the refractive index contrast with the substrate is limited, and the a-Si:H is transparent. This results in limited difference in reflectance, whereas the radiation is absorbed in the substrate. The SPV spectrum reported in Fig.5 clearly shows the Si absorption edge. At photon energies above the band-gap energy the SPV signal decreases, possibly due to e-h recombination at Si / a-Si:H interface, while at 1.3 eV it decreases due to recombination at the surface.

3.4 Transport properties

The transport properties of final devices fabricated using the most convenient passivation among the ones analysed above were studied through the dark J-V characteristics. The passivating layer was deposited at 13.56 MHz frequency, 4W rf power, 0.6 hPa, 160°C substrate temperature, 1’13” deposition time, 26 sccm SiH4 flow rate. Similar deposition conditions, when used to fabricate solar cells on 1 Ω cm (100) CZ silicon wafers in a fully low temperature process (no implantation, no Al diffusion, thus no back surface field), are shown to result in a device efficiency of 16.2% [10].

Figure 5: SPV signal as a function of incident photon energy for a sample passivated with 9 nm a-Si:H.

Figure 6: Dark J-V characteristics of a plasma passivated p/n device measured at different temperatures. Symbols: experimental data. Lines: analytical fit

The dark J-V curves of the device measured at different temperatures are reported in Fig. 6. The values of J0 extracted as a function of temperature by fitting the data through the diode equation are reported in Fig. 7. The reverse saturation current J0 shows little, if any, dependence on temperature. This behaviour indicates that

intraband tunnelling processes govern the recombination in these devices, in contrast to what normally observed on CZ-Si based devices, that show multitunneling through gap states followed by capture-emission processes [11].

Figure 7: Dark saturation current J0 on a plasma passivated p/n device as a function of temperature.

4 CONCLUSIONS

The passivating properties of layers obtained by PECVD on mc-Si have been analysed with respect to the structure of the layer (epitaxial or amorphous), layer thickness, and deposition temperature. QSSPC measurements show that thinner layers provide better passivating effect. Optical analyses evidence that the deposited layer thickness increases with increasing deposition temperature for fixed time, whereas the void fraction is observed to decrease due to lower H incorporation. Comparison with SEM observation confirms the validity of optical analysis. The relationship between the surface recombination velocity and the deposition temperature confirms the necessity to

optimize the deposition temperature, minimizing the Si-H2 complexes content while maintaining a good passivation by H.

ACKNOWLEDGEMENT

This work is supported by the Fondazione Cassa di Risparmio in Bologna.

REFERENCES

[1] M.A. Green, Prog. Photovolt: Res. Appl. 13 (2005)

447

[2] J. Plà, E. Centurioni, C. Summonte, R. Rizzoli, A.

Migliori, A. Desalvo, F. Zignani, Thin Solid Films 405 (2002) 248

[3] R. Rizzoli, E. Centurioni, J. Plà, C. Summonte, A.

Migliori, A. Desalvo, F. Zignani, J. of Non- Crystalline Solids 299-302 (2002) 1203

[4] M.G. Zebaze Kana, E. Centurioni, D. Iencinella, C.

Summonte, Thin Solid Films 500 (2006) 203

[5] A. Castaldini, D. Cavalcoli, A. Cavallini, and M.

Rossi, Sol. Energy Mater. Sol. Cells 72 (2002) 559 [6] E. Centurioni, Appl. Opt. 44 (2005) 7532

[7] R. A. Sinton, A. Cuevas, Appl. Phys. Lett. 69 (1996)

2510

[8] L. Fresquet, S. Olibet, E. Vallat-Sauvain, A. Shah, C.

Ballif, 22th EPVSEC, 3-7 Sept 2007 Milan

[9] L. Kronik and Y. Shapira, Surf. Sci. Rep. 37 (1999) 1 [10] D. Iencinella, E. Centurioni, M.G. Zebaze Kana, R.

Rizzoli, C. Summonte, A. Desalvo, F. Zignani, 20th EPVSEC, Barcellona, E, 6-10 June 2005, p.992

[11] Zignani, R. Galloni, R. Rizzoli, M. Ruth, C.

Summonte, R. Pinghini, Q. Zini, P. Rava, A. Madan, Y.S. Tsuo, Mat. Res. Soc. Symp. Proc. Vol. 420 (1996) 45