Metastability in Hydrogenated Nanocrystalline

Mater. Res. Soc. Symp. Proc. Vol. 910 © 2006 Materials Research Society

0910-A02-01

Metastability in Hydrogenated Nanocrystalline Silicon Solar Cells Guozhen Yue, Baojie Yan, Gautam Ganguly, Jeffrey Yang, and Subhendu Guha United Solar Ovonic Corporation, 1100 West Maple Road, Troy, MI, 48084 ABSTRACT We have recently reported that hydrogenated nanocrystalline silicon (nc-Si:H) solar cells exhibit metastability behavior that is significantly different from hydrogenated amorphous silicon (a-Si:H) solar cells. First, we studied the light-induced change of nc-Si:H cells under different light spectra, and observed that no light-induced degradation occurs when the photon energy is lower than the bandgap of a-Si:H, while light-induced degradation indeed occurs when the photon energy is higher than the bandgap of a-Si:H. Based on this observation, we conclude that the light-induced defect generation occurs mainly in the amorphous and/or grain boundary regions. Secondly, we found that forward-bias current injection in the dark does not cause any degradation in the performance of nc-Si:H cells, in contrast to observations in a-Si:H cells. This phenomenon can be explained by assuming transport percolation through crystallites, where excess carrier recombination does not cause degradation. Third, we found that a reverse electrical bias does not reduce, but rather enhances the light-induced degradation in the nc-Si:H cell performance. This enhancement increases with the magnitude of the reverse bias. We carried out a systematic study including measurements of quantum efficiency losses and fill factors under different color filters. In view of the heterogeneity of the material structure, we proposed a “back-to-back” diode model which explains most of the experimental results. Finally, we present results of the improvement of stability of nc-Si:H cells made using hydrogen dilution profiling. We find that the nc-Si:H cells with an optimized hydrogen dilution profile show very little lightinduced degradation though these cells have a high amorphous volume fraction. This indicates that the amorphous volume fraction is not the only factor determining the degradation. Other factors, such as the structure and distribution of the amorphous phase, as well as the properties of the grain boundaries, apparently play important roles in the overall stability of nc-Si:H cells. INTRODUCTION Light-induced defect generation in hydrogenated amorphous silicon (a-Si:H) due to the Staebler-Wronski effect [1] has been a subject of intense study because it degrades the performance of a-Si:H-based solar cells. Although significant effort has been made toward understanding the mechanism and reducing the amount of the light-induced degradation, it remains an unresolved issue. It is believed that the defect creation is caused by recombination of the excess electron-hole pairs generated by photo-excitation or forward bias current injection. Applying a reverse bias to a cell during light-soaking reduces the recombination probability of the electron-hole pairs, and hence reduces the degradation in a-Si:H based solar cells. In recent years, hydrogenated nanocrystalline silicon (nc-Si:H) has attracted remarkable attention due to both its ability to absorb long wavelength photons and its better stability under prolonged light soaking. nc-Si:H is an inhomogeneous material comprised of an amorphous phase, crystalline grains, and grain boundaries with a wide range of crystalline volume fraction. It also exhibits preferential orientation, and an anisotropic transport may be expected. Its metastability behavior, therefore, shows a substantial complexity and controversy, even more

than its a-Si:H counterpart. Previously, it has been reported that no light-induced degradation is observed in nc-Si:H cells [2,3]. Recently, it has been found that certain nc-Si:H cells show lightinduced degradation in their performance. Klein et al. [4] reported that nearly no light-induced degradation was observed in the nc-Si:H cell with the highest crystalline volume fraction, but cells with high amorphous volume fractions exhibit up to 10% degradation in the cell efficiency. Meillaud et al. [5] also found that the defect density and the stability of nc-Si:H cells made by very high frequency glow discharge method are related to the crystallinity. Although the degradation rate decreases with increasing crystallinity, the defect density has a minimum value for the cells deposited in the nanocrystalline/amorphous transition region. For solar cells made at high deposition rate (~ 3.5 nm/s), Gordijn et al. [6] demonstrated an improved solar cell performance after prolonged light-soaking. They explained that the series resistance is reduced, and a new defect equilibration could be reached by light-soaking. In this paper, we summarize our recent results on metastability study of nc-Si:H solar cells under various illumination and bias conditions [7-9]. The results show that stability behavior of nc-Si:H cells is significantly different from that of a-Si:H solar cells. EXPERIMENTAL DETAILS Single-junction, 1-µm thick nc-Si:H and 0.2-µm thick a-Si:H nip solar cells were deposited on 4 cm × 4 cm Ag/ZnO back reflector (BR) coated stainless steel (SS) substrates and bare SS substrates, respectively. The intrinsic layers were deposited using a modified very high frequency glow discharge method using a silane-hydrogen mixture at high deposition rates of ∼38 Å/s. A 0.11-µm thick a-SiGe:H bottom cell used in the forward current injection experiment was made by a conventional radio frequency glow discharge method at a low rate of ~1 Å/s. The solar cells were completed with indium-tin-oxide dots having an active-area of 0.25 cm2 on top of the p layer. Current density versus voltage, J-V, characteristics were measured in the dark and under an AM1.5 solar simulator at 25 oC. J-V curves were also measured under light through narrow band-pass filters centered at 390 and 670 nm to probe the i/p interface as well as the bulk properties of the cells. Quantum efficiency measurements were conducted at room temperature under the short circuit condition and also under a reverse bias of 1 V in some cases. Light soaking experiments were carried out using 100 mW/cm2 of white light at 25 oC and 50 oC under the open-circuit condition or under an electrical bias. Light-soaking was also done using different light spectra. A red light was obtained by using white light source with a 665 nm cut-on filter, and blue light with a 650 nm cut-off filter. The light intensity for different light spectra was adjusted by a convex lens to have the same short-circuit current density (Jsc) for given solar cells. Forward bias carrier injection experiments were carried out in the dark at room temperature. The samples were annealed under vacuum at 150 oC for 2 hours before and after light/current soaking. RESULTS A. Light-soaking with different light spectra First, we present the light-soaking results for nc-Si:H cells after illumination with different spectra. For comparison, a-SiGe:H cells on Ag/ZnO that are typical for the bottom cell in a triple junction structure were also studied. Figure 1 shows the normalized maximum power,

Pmax (t)/Pmax(0)

1.00

Red, a-SiGe:H

0.95

Blue, a-SiGe:H 0.90

Red, nc-Si:H Blue, nc-Si:H

0.85 10

100

1000

Time (minutes) Fig. 1. Light-soaking kinetics of the normalized efficiency of nc-Si:H single-junction solar cells and a-SiGe:H singlejunction solar cells for different spectral regions.

Table I. Summary of light-soaking results for nc-Si:H single junction solar cells. IN denotes the initial state, LS the light soaked state, and DEG the percentage of degradation. Light State Jsc FF Eff Voc (mA/cm2) (V) (%) IN 22.20 0.474 0.621 6.53 white LS 22.00 0.467 0.598 6.14 DEG 0.9 % 1.5 % 3.7 % 6.0 % red IN 22.19 0.472 0.620 6.49 LS 22.11 0.473 0.619 6.47 DEG 0.4 % 0% 0% 0.3 % blue IN 21.98 0.476 0.621 6.50 LS 21.78 0.469 0.591 6.04 DEG 0.9 % 1.5 % 4.8 % 7.1 % Table II. Summary of light soaking results for a-SiGe:H bottom cells. FF Eff Light. State Jsc Voc (mA/cm2) (V) (%) IN 23.10 0.628 0.601 8.72 white LS 22.88 0.601 0.567 7.80 DEG 1.0 % 4.3 % 5.7 % 10.6 % red IN 23.91 0.614 0.587 8.62 LS 23.48 0.601 0.563 7.94 DEG 1.8 % 2.1 % 4.1 % 7.9 % blue IN 23.50 0.621 0.572 8.35 LS 22.91 0.591 0.539 7.30 DEG 2.5 % 4.8 % 5.8 % 12.6 %

Pmax(t)/Pmax(0), as a function of light-soaking time for nc-Si:H single-junction solar cells and aSiGe:H single-junction cells under red and blue light. For the nc-Si:H cell, the light intensities for the three spectra were adjusted to generate Jsc∼44 mA/cm2, which is approximately twice the Jsc under AM1.5 illumination. For the a-SiGe:H cells, the light intensities were adjusted to generate Jsc∼46 with the same mA/cm2 consideration. Tables I and II summarize the J-V characteristics of the two series of cells in their initial and light-soaked states. The data clearly show that there is no light-induced degradation for the nc-Si:H cell under red light, but some degradation under white light, and more degradation under blue light. Obviously, the absence of lightinduced degradation under red light illumination is due to the lack of absorption in the amorphous phase and no lightinduced defect generation in the crystalline phase. In contrast to red light illumination, white and blue light illumination show noticeable degradation in the cell performance. The difference in degradation between the whiteand blue-light soaked nc-Si:H cells could be explained in two ways. First, the blue light has a large portion of photons absorbed in the amorphous phase, which produces defects. Second, the blue light is absorbed near the i/p interface, a region to which the cell performance is very sensitive. For the a-SiGe:H cell,

we consider the material to be homogenous having uniform bandgap of about 1.4 eV. The differences among the light-soaking results in the three spectral regions are mainly due to the second mechanism discussed above. Overall, the light-induced degradation in nc-Si:H is much less than that in the a-SiGe:H cell even under white and blue light. These experiments confirm that the light-induced defect generation in amorphous phase and the grain boundary region is responsible for the degradation in the nc-Si:H solar cells. B. Forward bias carrier injection in the dark Next, we present the results of the electrical bias effect on the metastablity in 0.99 2 nc-Si:H solar cells. To a-Si:H (3 mA/cm ) check whether a forward bias injection causes 0.97 degradation, we applied a 1 V forward bias to the a0.95 Si:H and nc-Si:H cells in 2 the dark at room a-SiGe:H (72 mA/cm ) temperature. The initial 0.93 forward current density was 1 10 100 1000 10000 60 mA/cm2 for the nc-Si:H Time (minutes) cell and 3 mA/cm2 for the Fig. 2. Normalized FF as a function of time with a forward bias in a-Si:H cell. The kinetics of an a-Si:H, an a-SiGe:H and an nc-Si:H single-junction cells. the FF under the forward Table III. Behavior of Voc and FF for an a-Si:H (cell 1), an a- bias is plotted in Fig. 2. SiGe:H (cell 2) and three nc-Si:H (cells 3-5) cells under various One can see that the FF of the a-Si:H cell degrades forward bias conditions in the dark. continuously during the Sample State Voc FF ∆Voc ∆FF/FFin experiment time of 17 (V) (%) (mV) hours. This is simply due a-Si:H Initial 0.998 0.696 to new defects generated by Cell 1 20 h, 1V, 0.978 -20 0.664 -4.6 the recombination of excess 3 mA/cm2 carriers injected by the a-SiGe:H Initial 0.623 0.638 applied forward bias, which 0.597 -26 0.596 -6.5 Cell 2 20h, 1V, deteriorates the cell 72 mA/cm2 performance. The nc-Si:H nc-Si:H Initial 0.469 0.575 cell, on the other hand, 0 0.579 +0.7 Cell 3 15 h, 0.5V, 0.469 shows no degradation in the 7 mA/cm2 FF although the forward Cell 4 Initial 0.469 0.591 current density was much 0.470 1 0.588 -0.5 15 h, 1V, higher than the a-Si:H cell. 58 mA/cm2 One may suspect that Cell 5 Initial 0.472 0.587 thermal annealing caused 0.474 2 0.590 +0.5 64h, 1V, by the high current density 57 mA/cm2 might have occurred in the 1.01

2

FF(t)/FFin

nc-Si:H (60 mA/cm )

nc-Si:H cell. Therefore, we performed another experiment with an a-SiGe:H cell. The forward injected current density at 1 V for this cell is around 72 mA/cm2, which is higher than that of the nc-Si:H cell, but we found significant degradation in FF as shown in Fig. 2. In addition, the temperature of the samples was monitored using a thermocouple on the sample surface. No noticeable temperature increase was observed during the forward current injection. The experiment has been carried out for different nc-Si:H cells and under different bias conditions. The result is reproducible. Table III summarizes the data of open-circuit voltage (Voc) and FF for one a-Si:H, one a-SiGe:H, and three nc-Si:H cells. For the a-Si:H cell, Voc decreased by 20 mV and FF by 4.6% for a 1 V forward bias after 20 hours of current soaking, even though the forward current density was only 3 mA/cm2. On the other hand, for the nc-Si:H cells, no current induced degradation was observed in either Voc or FF for all three cells with one cell (Cell 5) having been subjected to a 57 mA/cm2 forward-current injection for more than 60 hours. For the a-SiGe:H cell, the injected current density is also high, ~ 72 mA/cm2, but it caused a degradation of 26 mV in Voc and 6.5% in FF after 20 hours of current soaking. Based on the experiments above, it can be concluded that the nc-Si:H cells do not have current-induced degradation even though they suffer from light-induced degradation. C. Enhancement of light-induced degradation in nc-Si:H cells under reverse bias Next, we present the results of our study showing the effect of electrical bias on light-induced degradation in both the a-Si:H and nc-Si:H solar cells. For each experiment, two cells having similar initial J-V characteristics on the same substrate were selected for simultaneous lightsoaking. One cell served as a reference and was light-soaked under the open-circuit condition. A reverse bias of 2 V was applied to the other cell during light soaking. It should be pointed out that prior to this experiment, we had first confirmed that the two cells on the same substrate showed the same amount of degradation after light soaking under the same conditions. Thus, any difference in the degradation magnitude for the two cells can be attributed to the effect of the applied bias. Again, the experiment was repeated many times, and the results are reproducible. Typical results are listed in Table IV. One notes that, for the a-Si:H cell under the open-circuit condition (Cell 1), 69 hours of light soaking cause a decrease of 30 mV in Voc and 8.6% in FF; while for the cell under a 2 V reverse bias (Cell 2), there are no light-induced changes within the measurement uncertainty. Table IV. Voc and FF values for the a-Si:H (cells 1, 2) and ncThis result is consistent Si:H (cells 3, 4) cells before and after AM1.5 light soaking at with previous reports 25oC with and without a –2 V bias. [10,11]. The Sample State Voc FF ∆Voc ∆FF/FFin reduced/eliminated light(V) (%) (mV) induced degradation by a a-Si:H Initial 0.984 0.695 reverse bias in the a-Si:H Cell 1 69h, no bias 0.954 -30 0.635 -8.6 cell is commonly explained Cell 2 Initial 0.985 0.704 by the reduced 69h, -2 V 0.983 -2 0.708 +0.6 recombination rate. This is because the enhancement in nc-Si:H Initial 0.476 0.591 the electric field effectively Cell 3 63h, no bias 0.467 -9 0.559 -5.4 separates the electron-hole Cell 4 Initial 0.476 0.594 pairs. In contrast, the 63h, -2 V 0.427 -49 0.523 -12.0 reverse bias effect on the

light-induced degradation in the nc-Si:H cells is quite different from that in the a-Si:H cell. As shown in Table IV, the Voc of the biased cell (Cell 4) decreased by 49 mV after 63 hours of light soaking, while that of the cell under the open-circuit condition (Cell 3) decreased by only 9 mV. Similarly, the FF shows more degradation in the cell with the reverse bias (12.0%) than under the open-circuit condition (5.4%). Apparently, a reverse bias in the nc-Si:H cell does not reduce, but rather enhances the light-induced degradation. The enhanced degradation in the cell performance was completely restored after subsequent annealing in vacuum at 150 oC for 2 hours. Since the electrical bias has a significant effect on light-induced degradation in nc-Si:H solar cells, it is very interesting to investigate the bias dependence of the degradation. To perform this experiment, eight cells on the same substrate were light-soaked simultaneously under 100 mW/cm2 white light at 50 oC with various bias conditions of –1 V, short circuit (0 V), open circuit (+0.48 V), and +1 V. To check the reproducibility of the degradation rate, we included two cells at each bias condition. Table V lists the J-V characteristics of the initial and degraded states reached after 307 hours of light soaking. The values of FF under blue and red light illumination, FFb and FFr, were measured using narrow band-pass filters centered at 390 and 670 nm, respectively. One can see that the two cells with the same bias condition have similar degradation rates, hence the results are reproducible in terms of different cells. The light-induced degradation of the FF under AM1.5 illumination as a function of applied bias is plotted in Fig. 3. The short circuit corresponds to 0 V bias condition, and the open circuit is considered to be a forward bias condition with a bias value of Voc, which is around 0.48 V. It is clearly seen that the light-induced degradation in FF was increased by the increase of the magnitude of the reverse bias. Under the 1 V reverse bias condition, the FF degrades on the Table V. J-V characteristics of the initial (In.) and degraded (Deg.) nc-Si:H solar cells. The light-soaking was done under 100 mW/cm2 white light with various bias conditions of –1 V, short circuit, open circuit, and +1 V at 50 oC. FFb and FFr are fill factors measured with a blue filter and a red filter, respectively, as given in the text. FFbin. and FFrin. are the initial values of FFb and FFr, respectively. Cell No.

Bias (V)

Status

Voc (V)

1

-1 V

In. Deg. In. Deg. In. Deg. In. Deg. In. Deg. In. Deg. In. Deg. In. Deg.

0.479 0.464 0.479 0.464 0.483 0.472 0.480 0.474 0.477 0.470 0.478 0.470 0.482 0.467 0.478 0.465

2 3

Short circuit

4 5

Open circuit

6 7 8

+1 V

∆Voc (mV)

-15 -15 -9 -6 -7 -8 -15 -13

FF

0.634 0.563 0.633 0.566 0.618 0.580 0.631 0.574 0.622 0.600 0.628 0.603 0.608 0.587 0.629 0.604

∆FF /FFin. (%)

-11.2 -10.6 -6.1 -9.0 -3.5 -4.0 -3.5 -4.0

FFb

0.663 0.624 0.659 0.624 0.664 0.636 0.663 0.638 0.650 0.617 0.655 0.618 0.655 0.607 0.656 0.612

∆FFb /FFbin. (%)

-5.9 -5.3 -4.2 -3.8 -5.1 -5.6 -7.3 -6.7

FFr

0.664 0.579 0.664 0.580 0.660 0.617 0.663 0.599 0.653 0.622 0.655 0.622 0.655 0.621 0.651 0.624

∆FFr /FFrin. (%)

-12.8 -12.7 -6.5 -9.7 -4.7 -5.0 -5.2 -4.1

average by 10.9%, while the FF degrades only by 7.6% under the short circuit (a) condition. For the open circuit and +1 V 10 bias conditions, however, the degradation in the FF is the same, around 3.8%. 8 FFb and FFr are very useful measurements in determining the 6 uniformity of material quality in solar cells. Normally, FFb is a measure of the properties near the i/p interface since the 4 blue light is mainly absorbed in this region. Moreover, since the holes have a 2 very short distance to traverse under blue -1.5 -1 -0.5 0 0.5 1 1.5 illumination, FFb is also determined by the Bias (V) electron transport properties. Similarly, Fig. 3. Light-induced degradation in FF versus FFr is a measure of the hole transport the electrical bias during light soaking, where property in the bulk of the i layer. In Table ∆FF denotes the variations in FF. The light V, one notes that, under the reverse bias soaking was carried out at 50oC under one sun and short circuit conditions, the FFr has more degradation than the FFb, whereas white light for 309 hours. the FFb has more degradation under the 1.10 forward bias. These indicate that the Light-soaked under reverse bias degradation under different bias conditions 1.08 is not uniform. Banerjee et al. [12] used a quantum efficiency loss technique to study a-Si:H 1.06 and a-SiGe:H alloy cells with different i/p interfaces and intrinsic layers, and 1.04 identified the carrier collection losses in Light-soaked under open circuit the front surface and the bulk of the i layers. We use the same technique to 1.02 Initial investigate light-induced degradation in nc-Si:H cells. The quantum efficiencies of 1.00 the solar cells under short circuit condition 350 450 550 650 750 850 950 (Q(0)) and 1 V reverse bias (Q(-1)) were Wavelength (nm) measured. The quantum efficiency loss, Fig. 4. Qloss spectra of nc-Si:H cells before and Qloss, is defined as the ratio of Q(-1) to after light-soaking with and without reverse bias. Q(0). Usually, the Qloss values at 390 and Qloss is defined as the ratio of quantum efficiency 670 nm have correlations with the values under 1 V reverse bias and short circuit condition. of FFb and FFr, respectively. High values of FFb and FFr are associated with low values of Qloss at 390 and 670 nm, and vice versa. Figure 4 shows the Qloss as a function of the wavelength for two cells. One cell was light-soaked under the reverse bias condition, and the other under open circuit condition. For their initial states, the Qloss spectra are the same, indicating that the two cells have similar carrier collections. After light-soaking, the Qloss curves become significantly different. The cell under the reverse bias condition has much higher Qloss Qloss

∆ FF/FFin (%)

12

than the cell under the open-circuit condition in the long wavelength region, and lower Qloss in the short wavelength region. The results are consistent with the FFb and FFr measurements. The result of the Qloss, combined with FFb and FFr measurements, suggests that the degradation in nc-Si:H cells is not uniform. Under a reverse bias, the light-induced degradation could be the result of extra defects generated in the bulk of the intrinsic layer since the intrinsic layer absorbs most of the long-wavelength photons. Similarly, under forward bias conditions, the degradation could take place mostly in the region close to the p layer. D. Improvement of nc-Si:H cell stability by hydrogen dilution profiling technique We have made significant progress in understanding the mechanism of light-induced degradation in nc-Si:H solar cells. We have also tried to improve the stability of the nc-Si:H cell by optimizing the deposition process and device design. As previously reported [13], hydrogen dilution profiling is an effective method of controlling the nanocrystalline evolution during the deposition of nc-Si:H, and improves the initial solar cell performance. Further study shows that an appropriate hydrogen dilution profiling also improves the nc-Si:H cell stability. Table VI lists the J-V characteristics of two sets of nc-Si:H single-junction solar cells at their initial and stable states. The stable states were reached under 100 mW/cm2 white light at 50oC for 1000 hours. The intrinsic nc-Si:H layers in the first set (the first two cells) were deposited using RF. MVHF was used for the nc-Si:H intrinsic layer deposition in the second set (the last two cells). The first RF cell using a constant hydrogen dilution during the intrinsic layer deposition show a large light-induced degradation of ~ 15% in efficiency, mainly due to reductions in open-circuit voltage (Voc) and fill factor (FF). The second cell, with an optimized hydrogen dilution profiling, shows only a 3.5% light-induced degradation. Similarly, for the MVHF cells, the cell with a step hydrogen dilution profiling shows an 8.5% light-induced degradation, which is somewhat lower than the RF cells with constant hydrogen dilution, but is still much larger than the dynamically profiled cell. The Voc and FF in the MVHF cell (13348) with an optimized hydrogen dilution profiling did not degrade after prolonged light soaking; in fact, the FF of this Table VI. Initial (A) and stable (B) performance of nc-Si:H cells. (C) refers to the percentage of light-induced change. The crystalline volume fraction (fc) was extracted from the ratio of the component integrated area to the total area of the Raman spectra excited by red laser. FF Run Dep. H fc State Eff Jsc Voc (V) # Method Dilution (%) (%) (mA/cm2) A 7.85 23.06 0.499 0.682 10514 RF Constant 24.3 B 6.73 22.44 0.470 0.638 C -14.3% -2.7% -5.8% -6.5% 10505 RF Dynamic 16.5 A 7.56 22.76 0.520 0.638 Profiling B 7.30 22.91 0.517 0.616 C -3.5% +0.7% -0.6% -3.4% 12085 MVHF Step 30.5 A 6.62 21.76 0.479 0.635 Profiling B 6.06 21.65 0.470 0.596 C -8.5% -0.5% -1.9% -6.1% 13348 MVHF Dynamic 20.8 A 7.82 22.72 0.524 0.657 Profiling B 7.72 21.85 0.527 0.670 C -1.3% -3.9% +0.6% +2.0%

Intensity (a.u.)

Intensity (a.u.)

cell improved slightly. Raman analysis was used to examine the material structure on the six samples. Figure 5 shows the Raman spectra of the sample 13348 excited with a green (532 nm) laser and a red (633 nm) laser. The green light probes the material structure in the top layer near the i/p interface, while the red light reveals the information from the bulk of the intrinsic layer. The Raman Cell 13348 spectra were deconvoluted into different 532 nm components of amorphous LA (∼310 cm-1), LO (∼380 cm-1), TO (∼480 cm-1), intermediate (∼500 cm-1), and crystalline (∼520 cm-1) modes. The crystalline volume fraction was extracted from the ratio of the component integrated area to the total area. Compared to Raman spectra for all the samples, three important phenomena are 200 300 400 500 600 observed. First, the crystalline volume fraction Raman shift (cm-1) is higher for the green laser than the red laser in the samples with constant hydrogen dilution Cell 13348 (Sample 10514), as normally observed in the 632.8 nm nanocrystalline evolution with thickness. The optimized hydrogen dilution profiling (Samples 10505 and 13348) reversed this trend and resulted in a lower crystalline volume fraction in the region near the i/p interface as previously reported. Second, the stable cells have lower crystalline volume fractions than those with 200 300 400 500 600 high light-induced degradation, especially at the -1 Raman shift (cm ) i/p interface region probed by the green light. Fig. 5. Raman spectra and their Third, all the cells do contain a large amount of deconvolutions of cell 13348 excited with amorphous phase. Therefore, light-induced (upper) a green laser and (lower) a red laser. degradation does not have a direct correlation with the amorphous volume fraction. DISCUSSION A. Back-to-back connected diode model We now discuss possible mechanisms for the observed phenomena on electrical bias effect. In general, nc-Si:H can be thought of as a mixed material comprised of nanometer size grains and amorphous tissues. Thus, it is not difficult to understand the results of forward-current injection in the dark. The conductivity of the amorphous regions is much lower than that of the crystalline regions. Theoretically, a percolation threshold occurs at a crystalline volume fraction of ~ 33 % with a random distribution of grains in a three dimensional system. However, the carrier transport in nc-Si:H is complicated by the columnar growth of grains and complex structures at grain boundaries. The percolation threshold could be much larger than the theoretical value for the transport parallel to the surface, but could also be much smaller for perpendicular transport as indicated by a sharp drop of Voc for the cells with a very small amount

(a)

of crystalline volume fraction. Therefore, it is reasonable to assume that the forward injected current basically flows through the percolation Amorph ou s Phase path of grains in the nc-Si:H cells studied. The light-soaking experiments under different Crystalline Phase Crystalline Phase light spectra have shown that only the recombination in the amorphous phase is responsible for the degradation in the nc-Si:H + + + cell. Since there are very few excess carriers + + injected into the amorphous phase, current induced degradation does not occur in nc-Si:H solar cells. The reverse bias effect on the light(b) Forward biased diode Reverse biased diode induced degradation in nc-Si:H cells is more (1) difficult to explain, and we offer some - (3) conjectures. When exposed to light, the amorphous phase is expected to participate in (2) the carrier transport because photocarriers are generated in both the amorphous regions and + + the grains. For the current path involving the (5) amorphous phase, the material is a randomly + distributed hetero-junction system formed by an amorphous phase and large grains, which + + + + have different bandgaps and activation (4) energies. Considering a given amorphous Fig. 6. The band diagram of a “-crystalline- region, if both sides are large grains with a amorphous-crystalline-” region in nc-Si:H low bandgap, the interface region between the material to illustrate the back-to-back diode amorphous and grains can form a microscopic model under (a) open circuit, and (b) reverse diode. Thus, a small unit of “-crystallinebias during light-soaking. In (b), (1) represents amorphous-crystalline-” is equivalent to two the electron injection into the amorphous phase microscopic back-to-back connected diodes. from the left crystalline phase in the forward If we further assume that the grains are more biased diode; (2) photo-generated electron-hole n-type than the amorphous phase, electrons pairs recombine in the forward biased region of will be accumulated in the amorphous side the amorphous phase; (3) photo-generated and depleted in the crystalline side, when they electrons are swept out of the the reverse form a diode. The charge redistribution will biased region in the amorphous phase by an create a local internal electric field at the electric field; (4) holes accumulate in the grain boundary between the amorphous and boundary in the right crystalline region; and (5) crystalline phases. In this case, the conduction electrons recombine with accumulated holes in and valence bands of the amorphous side are bent down, and the bands of the crystalline the grain boundary region. side are bent up. To illustrate this picture, a schematic band diagram of a “-crystalline-amorphous-crystalline-” region is drawn in Fig. 6(a), where two microscopic “back-to-back” connected diodes are also illustrated. One can see that the excess electrons in the amorphous phase generated by illumination have a tendency to move out of the amorphous phase and into the grain boundary region, where they will recombine with -

-

-

the holes accumulated in the crystalline side. Therefore, we expect that, in addition to the lightinduced defect generation in the amorphous region, light-induced defects could also be generated in the grain boundary region. The grain boundary region might have much higher defect density than the amorphous phase, and play a significant role in the carrier transport and solar cell performance. Based on this picture, it is not difficult to imagine that the light-induced degradation in nc-Si:H solar cells depends not only on the properties of the amorphous phase, such as the amorphous volume fraction, but also on the grain boundary properties, such as the hydrogen concentration and the structure of the material. This could explain that some nc-Si:H cells with similar crystalline volume fraction show different degradation behaviors. B. Explanation for the experimental results When a reverse bias is applied to a nc-Si:H cell, one of the diodes in the back-to-back diode configuration will be forward biased and the other reverse biased. The reverse bias will increase the electric field in the reverse biased region and cause a substantial drop in the field in the forward-biased region. The band diagram of the “-crystalline-amorphous-crystalline-” structure under the reverse bias is plotted in Fig. 6(b). In this case, there are two mechanisms that give rise to the enhanced degradation in nc-Si:H cell performance. First, for the forward biased diode (left side in Fig. 6(b)), photo-generated electrons in the crystalline regions will move toward the amorphous regions. These excess electrons, together with the electrons generated in the amorphous region, will recombine with the holes generated in the amorphous region as illustrated by the processes (1) and (2) in Fig. 6(b). This will increase the defect generation rate in the amorphous region. On the right side, the enhanced electric field would sweep out the excess electrons in amorphous phase more efficiently and reduce the recombination of the electron-hole pairs. This is similar to the case in a conventional a-Si:H cell that the light-induced degradation is significantly suppressed by a reverse bias. Net additional defects will be determined by the internal field distribution within the amorphous region, and in turn by the thickness of these regions and the nature of the boundaries between the amorphous and the nanocrystalline phase. For the regions with very thin amorphous phase, the electric field would be sufficiently high to prevent recombination either in the open-circuit or the reverse bias conditions. Light-induced defect generation would take place only in the forward biased region when a reverse bias is applied. In this case, there will be more light-induced defects generated by the application of an external reverse bias than in the open-circuit condition. In addition, we need to consider the second process as illustrated by (3), (4), and (5) in Fig. 6(b). Driven by the enhanced electric field, the electrons from the amorphous region will recombine with the holes from the crystalline region in the grain boundary as described in (5). It would be reasonable to assume that the grain boundary region has poor structure and a higher hydrogen concentration. Therefore, the recombination at the grain boundary will enhance the total defect generation. To summarize the proposed back-to-back diode model, we identified two processes that can cause the enhanced light-induced degradation by a reverse bias. First, the reduced electric field in the forward-biased diode causes additional recombination in the amorphous region. Second, the electrons driven out of the amorphous region by the electric field in the reverse diode recombine with the accumulated holes in the grain boundaries, which also produce additional defects. An immediate challenge for this model is to answer why a forward external bias does not increase the light-induced degradation. First, it is difficult to apply a high forward bias without damaging the solar cell. For example, a 0.48 V forward bias corresponds to the open circuit

condition. Second, a forward bias causes a significant current injection, and the injected space charges make the bias voltage drop mainly at the i/p interface region. The intrinsic layer experiences a low field similar to the open circuit condition. Therefore, no measurable difference was found in light-induced degradation between forward bias and open circuit condition. SUMMARY We have studied the behavior of light-induced degradation in nc-Si:H solar cells under different conditions. By using different light spectra during light-soaking, we first identified that amorphous phase/grain boundary is responsible for the light-induced degradation of nc-Si:H cells. Secondly, we found that there is no forward-current-induced degradation in the dark. Third, we found an enhanced light-induced degradation under reverse bias, which is contrary to the results obtained in a-Si:H cells. We propose a model to interpret the reverse bias enhanced light-induced degradation in nc-Si:H solar cells. A “-crystalline-amorphous-crystalline-” region in the nc-Si:H layer could be considered as a “back-to-back” connected diode structure. Finally, we provided the data to improve the stability of nc-Si:H cells by using hydrogen dilution profiling technique. ACKOWLEDGEMENTS The authors thank S.R. Ovshinsky and H. Fritzsche for their constant encouragement, T. Palmer and E. Chen for sample preparation and measurements, and C. Teplin at NREL for Raman measurements. This work was supported by NREL under the thin-film partnership program Subcontract Nos. ZDJ-2-30630-19 and ZXL-6-44205-14. REFERENCES: [1] D. L. Staebler and C. R. Wronski, Appl. Phys. Lett. 31, 292 (1977). [2] J. Meier, R. Flückiger, H. Keppner, and A. Shah, Appl. Phys. Lett. 65, 860 (1994). [3] K. Yamamoto, IEEE Transactions on Electron Devices 46, 2041 (1999). [4] S. Klein, F. Finger, R. Carius, T. Dylla, B. Rech, M. Grimm, L. Houben, and M. Stutzmann, Thin Solid Films 430, 202 (2003). [5] F. Meillaud, E. Vallat-Sauvain, X. Niquille, M. Dubey, J. Bailat, A. Shah, and C. Ballif, Proceeding of the 31st IEEE Photovoltaic Specialists Conference, Lake Buena Vista, FL, January 3-7, 2005, p. 1412. [6] A. Gordijn, J. Francke, L. Hodakova, J.K. Rath, and R.E.I. Schropp, Mater. Res. Soc. Symp. Proc. 862, 87 (2005). [7] B. Yan, G. Yue, J.M. Owens, J. Yang, and S. Guha, Appl. Phys. Lett. 85, 1925 (2004). [8] G. Yue, B. Yan, J. Yang, and S. Guha, Appl. Phys. Lett. 86, 092103 (2005). [9] G. Yue, B. Yan, J. Yang, and S. Guha, J. Appl. Phys. 98, 074902 (2005). [10] L. Yang, L. Chen, J. Y. Hou, and Y. M. Li, Mater. Res. Soc. Symp. Proc. 258, 365 (1992). [11] J. Yang, K. Lord, B. Yan, A. Banerjee, and S. Guha, Proc. of 29th IEEE Photovoltaic Specialists Conference, (IEEE, New Orleans, Louisiana, 2002), p.1094. [12] A. Banerjee, X. Xu, J. Yang, and S. Guha, Appl. Phys. Lett. 67, 2975 (1995). [13] B. Yan, G. Yue, J. Yang, S. Guha, D.L. Williamson, D. Han, and C. Jiang, Appl. Phys. Lett. 85, 1955 (2004).