Plasmonic solar cells: status and prospects

The global energy crisis has driven a rapid increase in research in the field of renewable energy sources in recent years because of pressure to develop low-cost, clean alternatives to fossil fuels. Generating energy from the sun (photovoltaics) is such an example, although at present this solution is still too expensive to be broadly practical. A substantial fraction of the cost associated with photovoltaics is the hyperpure silicon used in making traditional solar cells. A more affordable approach would be to develop high-efficiency, thin solar cells, which would greatly reduce the consumption of hyperpure silicon.

Electric current is produced in a solar cell by creating electron-hole pairs (charge carriers) through absorption of photons in a semiconductor. Only photons with energy greater than the bandgap energy of the semiconductor are absorbed. Silicon is a poor absorber of photons with energy close to its bandgap. Consequently, a significant portion of solar radiation is transmitted through the solar cell without being absorbed. Wafer-based silicon solar cells employ random surface texturing to trap long-wavelength radiation inside the active volume of the cell, increasing the probability of absorption. The surface texture has feature sizes on the order of ~10μm.

Thin-film silicon solar cells with thicknesses typically of a few microns have greater transmission losses because of their reduced absorption volume, which leads to lower efficiencies. Conventional surface texturing is not suitable for light trapping in this case owing to large feature sizes. Novel, wavelength-scale light-trapping techniques are needed to enhance the efficiency of thin solar cells. One such technique, which we are exploring, involves trapping light inside the cell by exciting localized surface plasmons (waves of electrons) in isolated metal nanoparticles.

Figure 1. Effect of particle size on Q_scat (scattering cross-section of the nanoparticles) and f_sub (fraction of scattered radiation) and the particle's dielectric environment on Q_scat. SiO₂: Silicon dioxide. SiN: Silicon nitride.

The particles then behave as oscillating dipoles (or multipoles) and reradiate energy into the surroundings. A fraction, f_air, of the scattered radiation is coupled into photonic modes available in air, and another fraction, f_sub, into the modes available in silicon. Because of the larger refractive index of silicon, which leads to a high density of available photonic modes, most of the scattered light is coupled into silicon, i.e., f_sub≫f_air. Most of the light scattered into the substrate lies outside the escape cone (whose half-angle determines the critical angle for total internal reflection at the silicon-air interface) of silicon and is thus trapped inside the cell until it is completely absorbed.

Figure 2. Effect of nanoparticle shape and distance from the substrate on f_sub. All nanoparticles studied have a diameter of 200nm.

For efficient light trapping, we need to minimize parasitic absorption in the nanoparticles, maximize the scattering cross-section, Q_scat, of the nanoparticles, and maximize f_sub in the desired wavelength range (close to the bandgap of silicon). In general, silver has lower absorption than gold and is thus most suitable for light-trapping applications. The magnitude and peak position of Q_scat versus wavelength and f_subdepend critically on the particle size and shape, the effective dielectric constant of the surrounding medium, the position on the substrate (on illuminated or rear surface), and the distance from the substrate. The behavior of nanoparticles can be predicted using numerical simulations, enabling fabrication of optimal nanoparticle arrays for light-trapping applications.

We have shown that the position of the plasmonic resonance (peak of Q_scat versus wavelength) in the nanoparticles can be tuned to the wavelength region of interest by changing the particle's dimensions or its dielectric environment.¹ As the effective dielectric constant of the particle environment or the particle size increases, the Q_scat peak redshifts (see Figure 1). The Q_scat of the metal nanoparticles also increases with increasing particle volume and depends additionally on the local driving field (E_d) around the particle. For nanoparticles on the rear surface of the cell, E_d is determined by the field transmitted through the cell, while for particles on the illuminated surface, E_d is determined by interference between the incident and reflected fields.² Increasing particle volume to enhance Q_scat might not be advantageous because larger nanoparticles support multipole charge oscillations, which couple less scattered radiation into the substrate (in other words, reduce f_sub): see Figure 1. We have also shown^3,4 that the shape of the nanoparticles and their distance from the substrate are critical in determining f_sub (see Figure 2). As the separation between the particle and the substrate increases, f_sub decreases (see Figure 2).

As described above, Q_scat and f_sub cannot be maximized independently. However, acceptable Q_scat and f_sub can be obtained by tuning the nanoparticle properties. The light-trapping scheme employing plasmonic oscillations in metal nanoparticles can be implemented using simple techniques, without adding significant costs to the fabrication of solar cells. Since the metal nanoparticles are separated from the active silicon volume by a dielectric (surface-passivation) layer, the surface recombination in silicon is not affected either. This simple yet efficient light-trapping scheme is a step towards high-efficiency, thin silicon solar cells that could potentially replace fossil fuels as low-cost renewable energy sources. We aim to develop high-efficiency, thin plasmonic silicon solar cells by taking into account these design considerations.