1. Introduction

Photovoltaic solar cells are best known as an important technology for the future energy production [1–8]. To make power from photovoltaics competitive with fossil-fuel technologies, it is desirable to reduce the cost and increase the photocurrent conversion efficiency of solar cells. Usually a dielectric layer is deposited on the solar cell surface as a passivation layer to reduce the surface recombination, as well as an antireflection film to reduce the surface reflection. Silicon dioxide (SiO₂), silicon nitride (Si₃N₄), and titanium dioxide (TiO₂) are commonly used in photovoltaic cell fabrication with refractive indices of 1.5, 2.0, and 2.5, respectively. Another important way to enhance the solar cell absorption is the surface plasmon effect. In recent years, metal nanoparticles via surface plasmon have been widely utilized to enhance light absorption in solar cells because of the unique characters such as large normalized scattering cross section and adjustable resonance characteristics by changing nanoparticle geometries, etc [9–12]. Considering the above two approaches, the antireflection effect is more pronounced for a thick dielectric layer, while the increased overlap of the surface plasmon with the low-index dielectric layer causes a blue shift in resonance, which is not beneficial for the absorbance of near-bandgap light. So there is a dilemma where the antireflection effect is negative correlation with the light trapping induced by surface plasmons.

In this article, we investigated the effect of the antireflection film and the surface plasmon on the photocurrent conversion efficiency of silicon solar cells. Because of the dynamic depolarization and radiation damping for large particles and more absorption for small particles, we made use of self-assembled nanoparticles with an appropriate size range (10-140 nm) [13,14]. The effect of spacer thickness in the range of 0-100 nm was studied systematically, in particular for the thin ones where the characteristics of surface plasmon vary significantly. We found that there are optimum thicknesses (9-35 nm)of the spacer for the maximum conversion efficiency, increased by 38.3% compared to the nanoparticle-decorated cells without the spacer.

2. Experiments

As a demonstration, crystalline solar cells were employed here for their easy availability. Figure 1 shows a schematic diagram of the silicon solar cell geometry. The test p⁺-n-n⁺ cells are realized by applying the ion implantation method on a 4-in n-type silicon wafer with a thickness of 380 μm and the thickness of the p+ emitter layer is about 400 nm. Then the aluminum electrodes are fabricated on both sides of the device.

 

Fig. 1 Schematic diagram of the experimental design.

Download Full Size | PDF

After these processes, the wafer was cut into many single solar cells and eight of them were selected as the experimental samples. The SiO₂ layers with varying thicknesses (3, 9, 15, 35, 50, 75, 100 nm) were sputtered on seven of the samples using an Oxford Optofab3000 ion beam sputter. Deposition was performed at a pressure around 10⁻⁴ torr. The refractive index of the sputtered oxide was characterized by a Cary 5E spectrophotometer around 1.50. Meanwhile, a sample without SiO₂ layer is remained for comparison. Then the 12-nm silver layer was deposited on all of the samples using a Denton Explorer 5E electron beam evaporator, followed by a rapid thermal annealing in N₂ ambient. The size of the resulted nanoparticles was controlled by the annealing conditions such as the temperature and the time. The current-voltage (I-V) and external quantum efficiency (EQE) measurements were performed on the solar cell samples before and after the formation of nanoparticle arrays on oxide with different thicknesses.

Figure 2(a) shows SEM images of Ag particles on the silicon wafers, separated from the substrate by SiO₂ layer with different thicknesses (the pictures for other thicknesses show the same situation). These SEM images are analyzed using the software Image J [15], as shown in Fig. 2(b). The diameter of the particle is taken to be equivalent to the longest distance between any two points along the particle boundary. Figure 2(a) shows that the quasi-circular Ag islands have the surface coverage of approximately 30%, irrespective of the oxide thickness. Particle size distribution is quite broad, with particle diameter in the range from 10 to 140 nm. Moreover, more than 78% of the particles have diameters between 20 and 100 nm.

 

Fig. 2 (a) Scanning electron microscopy (SEM) of Ag nanoparticles on Silicon substrates coated with the SiO₂ layers with different thicknesses. (b) The corresponding distribution of particle sizes. The numbers at the top right corner indicate the thicknesses of the SiO₂ layers.

Download Full Size | PDF

Figure 3 shows the I-V characteristics of solar cells obtained under the illumination from a 100 Wcm⁻², AM1.5 spectrum by an Oriel solar simulator (Model 94041A). All solar cells are coated by Ag nanoparticles and SiO₂ films with different thicknesses. Details of the I-V parameters are summarized in Table 1 . The efficiency increases firstly with increasing spacer thickness, from 4.3% for the solar cell with no oxide to 6.1% for 35 nm SiO₂ layer. Afterwards, the efficiency decreases and achieves the minimum 4.9% for the 100 nm thick spacer. Figure 3 and Table 1 also show that the open circuit voltage V_oc of all samples is almost equal, which has a margin of error of plus or minus 3.3 percent. The fill factor (FF) also changes little, with a margin of error of plus or minus 3.9 percent. While the short circuit current density I_sc varies significantly with increasing spacer thickness. The largest photocurrent, enhanced by 41.7% compared to the case of no spacer, was achieved for the sample with a 35-nm-thick SiO₂. The maximum I_sc lead to the best performance in the power conversion efficiency η = FF I_sc V_oc/P_in, where I_sc, and P_in are the short circuit current and input power, respectively. According to the description above, I_sc is a determining factor in improving the photocurrent conversion efficiency and depends obviously on the SiO₂ layer thickness.

 

Fig. 3 Photovoltaic I-V curves for the samples after Ag particles deposition under one-sun illumination (AM1.5, 100 mWcm⁻²) using a solar simulator. The numbers at the down left corner indicate the thicknesses of SiO₂ layers.

Download Full Size | PDF

Table 1. Photovoltaic performance of the solar cells coated by different thickness of oxide after Ag deposition

View Table

3. EQE analysis

Figure 4 shows the EQE of the solar cells measured for further investigation on the light trapping provided by nanoparticles placed atop SiO₂ layers with different thicknesses. The extra phase lag between the light reflected from the front and rear surfaces of the spacer is more pronounced at short wavelengths where the path length in the spacer layer is significant relative to the light wavelength, consistent with the EQE peaks at the short wavelength for 75 nm and 100 nm SiO₂ layer. As can be seen in Fig. 4, the curves are broader for the thin SiO₂ layer, which can be ascribed to that the surface plasmon under the influence the SiO₂ layer and the Si substrate has a wider resonance. A local maximum in EQE peak, of up to 83.9%, is observed for the 35 nm SiO₂. An interesting observation is that the EQE peak is linearly red shifted roughly with the oxide thickness, under the influence of the antireflection effect.

 

Fig. 4 The external quantum efficiency (EQE) of the Si solar cells with Ag nanoparticles, separated from the Si substrate with different thickness of SiO₂ layer. The numbers at the bottom shows the oxide thicknesses.

Download Full Size | PDF

Figure 5 illustrates that the I_sc obtained from the EQE spectra as a function of the SiO₂ layer thickness. The sun spectrum I_AM 1.5 is taken from the link [16]. The same trends are observed for the two curves. The I_sc increases firstly and reduces afterward with the oxide thickness and the maximums are observed in the thickness range of 9-35 nm, indicating that a suitable thickness of SiO₂ layer is desirable to achieve the best photocurrent conversion efficiency.

 

Fig. 5 The I_sc obtained from the EQE spectra by integrating over the sun spectrum in comparison to the values of I_sc measured with the sun simulator.

Download Full Size | PDF

To investigate the effect of surface plasmons for different thickness of the SiO₂ layer, the EQE spread of solar cells coated by different thickness of SiO₂ layer after and before Ag particles deposition is shown in Fig. 6 . It must be mentioned that the EQE of the bare solar cell is higher than the same cell after Ag deposition in the entire wavelength range, which can be attributed to two mechanisms, the first is that the recombination of electron-hole pairs at the metal-semiconductor interface is rapid, especially for the cell without SiO₂ passivation layer; the second is that the destructive interference between incident and scattered light in the Si substrate occurs at wavelengths below resonance which localizes in the long wavelength range. Figure 6 shows that the EQE changes obviously with the SiO₂ layer thickness due to the presence of the surface plasmon, except at the longer wavelengths where the EQE spread is almost 0, indicating that the effect of Ag nanoparticles is negligibly small. In other word, the surface plasmon determines EQE in the range where the resonance is strong, demonstrating the resonance wavelength is an important factor in determining the light coupling. Figure 6 also shows that there is a crossover point at which the EQE spread is equal to 0. In Mie scattering, there is a phase mismatch between the scattered and the incident light, which can varies with wavelength and lead to the constructive or destructive interference. Especially, this phase mismatch is amplified near the plasmon resonance, corresponding to the experimental peaks and dips. Figure 6 also shows that there tends to be a reduction in crossover wavelength for the thin oxide, e.g. from 580 nm for 3 nm SiO₂ layer to 440 nm for 9 nm SiO₂ layer, whereas a linear increase for the thick ones, demonstrating the tunability of the EQE of nanoparticle-decorated cells by the spacer thickness that are typical for solar cell manufacturing. The dip wavelength correlated with the plasmon resonance is blue shifted as the SiO₂ layer is very thin, and show little change as the oxide thickness is beyond 9 nm, which can be ascribed to that the surface plasmon depends on the SiO₂ layer and the Si substrate for the thin spacer, while the effect of the SiO₂ layer dominates over that of the Si substrate as the spacer is thick enough. Moreover, the dip value increases with the thickness for the thin spacer, and decreases subsequently as the oxide thickness is beyond 9 nm. The largest reduction in EQE after the Ag nanoparticle deposition, of up to 46.1%,is observed at a wavelength of 490 nm for the sample with 75-nm-thick oxide. It may be noted that the peak is also clearly red shifted, as the thickness is beyond 9 nm. In conclusion, the light trapping provided by the random sized particles on the thin SiO₂ layer can overcome the suppression of photocurrent as a result of high gain, little suppression and blue shifted crossover wavelength.

 

Fig. 6 Measured EQE as a function of wavelength for cells after Ag deposition, shown as difference relative to reference EQE of the same cells prior to the nanoparticles deposition.

Download Full Size | PDF

4. Optical simulations

In order to further investigate the experimental EQE trends with different SiO₂ thicknesses, numerical simulations were also performed using the FDTD solutions package from Lumerical. A simplified structure was employed, with a SiO₂ layer between the particles and the Si substrate. As the experimental nanoparticles were roughly semispherical shapes, nanoscale Ag semisphere was used in the simulations. Nanoscale Ag semisphere arrays on a Si substrate are modeled using periodic boundary conditions so that the interaction between Ag particles are considered, and the period in x and y direction is 120 nm to keep the same surface coverage with the experiment. The Z dimension is truncated by the perfectly matched layers. It is advisable to set particle diameter to 75 nm, because the particle diameters in Fig. 2 are symmetrically 75 nm. A normally incident Total-field scattered-field source [TFSF] is used to illuminate the substrate. The semisphere is completely localized in the total field. The dielectric functions were modeled using a Drude model for Ag, and a Drude-Lorentz model for Si, while the optical data for the oxide was taken from Palik.

Figure 7 shows the normalized scattering cross section (Q_scat), the fraction of the scattered light into the Si substrate (F_sub), and the fraction coupled into the substrate for Ag nanoparticle with 75-nm diameter on Si substrate coated by SiO₂ spacer with different thicknesses. Clear peaks in the Q_scat curves are observed, corresponding to the surface plasmon resonance. There is only one peak for the thin SiO₂ layers, indicating the effect of the Si substrate on surface plasmon dominates over that of the SiO₂ layer. Two peaks are seen as the SiO₂ layer thickness is 15 nm, which can be ascribed to the fact that the effect of the SiO₂ layer is the same order as that of the Si substrate. It must be noted that the height of the Q_scat peak is as high as 4.3 for 9 nm thick SiO₂. In such case, a substrate covered with a 23.3% areal density of particles could fully scatter the incident light. Clearly, the resonance is blue shifted with increasing overlap of the surface plasmon with the low-index dielectric layer, e.g. the resonance is significantly blue shifted to 575 nm for the 9nm thick SiO₂, compared to the value 900 nm for 0 nm of the spacer. Finally, the peak wavelength converges to 455 nm, which supports the experimental result in Fig. 6 that the EQE spread dip related with the resonance show little change as the oxide spacer is thick enough. Moreover, the Q_scat of the surface plasmon on the SiO₂ layer is far less than 1 at long wavelengths, indicating the influence of surface plasmon on light scattering is negligible, which agrees qualitatively with the experimental result that the EQE difference between before and after Ag particles deposition is negligibly small beyond a wavelength of 1000 nm. Figure 7(b) shows that F_sub is greater than 0.5 at majority of the wavelengths for the thin spacer. In particular for the case of no spacer, about 80% of scattered light is directed into the Si substrate. Moreover, F_sub reduces with increasing oxide thickness almost for all wavelengths, due to the reduced coupling into modes inside the Si. Clear peaks and dips, induced by surface plasmons, can be seen in Fig. 7(c). For wavelengths below resonance, the scattered light is out of phase with the incident light, leading to destructive interference, corresponding to the dips in Fig. 6. However, this suppression can be avoided by locating nanoparticle arrays on the rear of solar cells [17]. At the short wavelengths, the absorption curves agree qualitatively with the EQE ones, and the difference between them can be attributed to the wide range of particle sizes in the experiment. However, the absorption is much greater than the EQE in the long wavelength range, indicating that the diffusion length of minority has a significant influence. According to the Eq. (1),

 

Fig. 7 (a) Calculated normalized scattering cross section (Q_scat). (b) Fraction of the scattered light into the substrate [F_sub] for Ag nanoparticles. (c) Fraction of light coupled into the substrate. The particles are separated from the Si substrate by SiO₂ films with different thicknesses. The numbers at the corner indicate the thickness of SiO₂ layer.

Download Full Size | PDF

5. Conclusions

In summary, we investigated the influence of the antirefelction film and the surface plasmon on the silicon solar cell performance, and discovered that the effect of the surface plasmon on light absorption dominates over that of the antireflection film at resonance, so the resonance wavelength is an important factor in determining the light trapping. For self-assembled Ag particles on the surface of solar cells in our experiments, appropriate spacer thickness (9-35 nm) were desirable to broaden the plasmon resonance, narrow photocurrent suppression range, weaken the suppression amplitude and strengthen the gain at the resonance wavelength, while still providing antireflection effect. Our results show that there is a trade-off between the benefit of the antireflection effect and the influence of the surface plasmon.