1. Introduction

Controlling light absorption in very thin semiconductor layers is of prime importance for thin film photovoltaics, photodetectors and infrared imaging. Such a control may be achieved using various kinds of light trapping techniques allowed by the maturity of nanophotonics. Light trapping efficiency depends on the increase of light absorption over controlled spectral range and large angular acceptance. In the specific case of photovoltaic (PV) solar cells, ideal light trapping structures integrated over very large areas, should enable absorption control over the whole solar spectrum, and for a wide range of incidence angles.

Absorption enhancement using surface plasmons has been proposed during the last years, either with self-assembled nanoparticles [1] or regular arrays of nanostructures fabricated using top-down technological processes [2]. In these approaches, the absorption enhancement around the plasmon resonances is balanced by light absorption in the metallic nanostructures. Photonic Crystals (PCs) have also been considered to realize either a back reflector [3] or selective filters for tandem solar cells [4]. It was recently proposed to pattern directly the absorbing medium of a PV solar cell as a planar photonic crystal (PPC), in order to control the absorption of incident light [5–10].

In this work we investigate both theoretically and experimentally the absorption of an hydrogenated amorphous silicon (aSi:H) PPC and compare it to a plain layer; optical properties of this material are detailed in ref [11]. In section 2, the basic configuration of the absorbing one-dimensional (1D) PPC structures is introduced, and we discuss the choice of appropriate geometrical parameters, optimised by using electromagnetic simulations. A simple way to pattern wide areas of aSi:H layers as PPCs is then described in section 3. In section 4, we discuss the absorption spectra measured on PPCs and plain layers, and we compare the results with optical simulations.

2. Absorbing photonic crystal structure design and simulation

For this demonstration, we chose to work with aSi:H with a thickness around 100 nm, deposited on glass, see Fig. 1 . Such a thickness is expected to ensure a maximum subsequent carrier collection, as the minority carrier diffusion length is evaluated to ~200nm in this absorbing material [12]. Our PPCs consist in 1D lattices of air slits drilled in a silicon layer, positioned between layers of moderate refractive indices like silicon oxide or transparent conductive oxide layers (TCOs). We use the Bloch modes of the PPC which stand above the light line [13,14]. Incident light may then be trapped in such resonances at a wavelength and during a photon lifetime closely related to the parameters of the structure: lattice parameter (L), semiconductor filling factor (ff: ratio between the width w of the silicon patterns and L), thickness of the layer and surrounding media refractive indices. The dispersion characteristics of these Bloch modes exhibit a low curvature, which results in an increased density of optical states, and a low dependence over light angle of incidence. The maximum absorption is then achieved if the photon lifetime at resonance is matched with the lifetime corresponding to absorption losses in the unpatterned absorbing medium [5]. Over a wide spectral range, as we will consider here, increasing the absorption involves multiple Bloch mode resonances, and the PPC structure parameters should be systematically scanned to achieve an optimal configuration.

 

Fig. 1 (a) Schematics of a 1D PPC patterned in an absorbing layer (violet) deposited on a glass substrate (clear blue). (b) The material surface filling factor (ff) and the lattice parameter (L) of the 1DPC are indicated. Incoming light is normal to the layer, the two orientations of the electric field are shown.

Download Full Size | PDF

We first performed simulations using Rigorous Coupled Wave Analysis (RCWA) [15] to obtain the absorption spectrum of such a stack, in the 380-750nm wavelength range, as a function of the 1D PPC parameters (L and ff). The investigation is limited to wavelengths below 750nm, i.e. in the spectral range where aSi:H exibits a non zero absorption coefficient [11], and we considered the AM1.5g solar spectral intensity distribution. The corresponding contour mapping of wavelength-integrated absorption efficiency is shown on Fig. 2 . Using the same method, we simulated that a simple 100nm unpatterned layer of aSi:H on glass absorbs about 30% of the same solar light spectral range. In the case of the 1D PPC, the absorption is increased to more than 42%. One can note on Fig. 2 the integrated absorption tolerance to a change in the PC parameters: a 10% relative change in the filling factor or in the PPC period affects integrated absorption only by 2%.

 

Fig. 2 Contour mapping of the integrated absorbency of a 100nm patterned aSi:H layer deposited on glass. The proportion of absorbed solar light intensity, integrated over 380-750nm, is shown as a function of the PC parameters (L, ff, see Fig. 1).

Download Full Size | PDF

3. Fabrication

95nm thick aSi:H films were deposited on a glass substrate in a standard capacitively coupled radio frequency glow discharge system [16]. In order to generate the PPC pattern over a large-area, we used laser holography. Two coherent planar waves originating from a single 325nm He-Cd laser were combined to interfere with each other and to produce a series of grating lines on a resist mask [17,18] which was transferred into a hard mask of silica. aSi:H was then selectively etched using Reactive Ion Etching, with a gas plasma based on SF₆ and Ar, at low pressure (15 mT), yielding an aSi:H PPC structure with a period of 335nm and ff = 58 +/− 2%. These parameters do not strictly correspond to the optimal structure, due to technological inaccuracies. The silica hard mask was left above the stripes as a protection, adding about 100nm of transparent medium, which has a negligible effect on the absorption properties, and may account for a transparent layer to be used as a TCO for carrier collection in a complete stack, or as an antireflection coating. This process enables to pattern aSi:H layers with thicknesses up to 400nm, as shown on Fig. 3 , where the shown sample (used only for topographical characterization) has a 1D PPC period of 400nm and a lateral stripe size of 225 +/− 5nm. As shown in Fig. 3(a), the etched side-walls are fairly vertical, although the top half of the silicon patterns exhibit a 5-10° slope. The line profile can be seen on Fig. 3(b) from above, the latter is not perfectly straight, but stays well below the 10% tolerance limit given by simulations.

 

Fig. 3 Close-up (a) and general scanning electron microscope (SEM) view (b) of a 400nm thick processed sample, covered by a thin chromium layer to increase the image quality.

Download Full Size | PDF

4. Optical characterisation and results

To characterize the optical properties of the structures, we performed spectrally resolved reflectivity (respectively transmittivity) experiments integrated over the whole half space around the front (resp. back) sample surface, using an integrating sphere with a 1mm spot diameter; the experimental uncertainty on each measurement is of 5%. The results obtained for unpolarized light are shown on Fig. 4(a) . In non absorbing PPCs, the addressed Bloch modes generally exhibit a very high reflectivity at normal incidence [13]. But in the case of our material, this maximum is damped and its amplitude is well below 100% since a significant amount of incident light is resonantly absorbed. Given the absence of any other light source in this structure, and the low light intensity we are using, the transmittance T and reflectance R spectra enable a direct measurement of the effective absorbency A as shown by the relation A + T + R = 1.

 

Fig. 4 (a) Reflection and transmission spectra of the 95nm thick 1D PPC (shown for unpolarized light). (b) Derived absorption spectra of the same structure for Ex and Ey polarizations. (c) Absorption spectra of the 1D PPC structure (blue squares), of an unpatterned layer with the same thickness (red crosses), and of an unpatterned layer of 400nm (black plain circles with line). Simulations are also shown for the 1D PPC (blue continuous line) and the corresponding unpatterned structure (red continuous line).

Download Full Size | PDF

Figure 4(b) displays the absorption spectra derived from R and T measurements for incident light with an electric field polarized along x (E_x) and y (E_y) axes. Due to the strong anisotropy of the 1D PPC, the photon lifetimes and wavelengths of resonant modes strongly depend on the incident light polarization. This results in a higher absorption increase at low (resp. high) wavelengths for E_x (resp. E_y) polarization. Such a polarization dependant behavior could be reduced in the case of a 2D PPC structure.

In the case of unpolarized incident light, the spectra are shown on Fig. 4(c), together with spectra calculated by RCWA, for both the 1D PPC structure and the unpatterned aSi:H layer including the remaining SiO₂ layer on top. These absorption measurements, which fairly agree with the prediction, show an absorption increase of the patterned layer, with respect to the unpatterned one, over the whole measured spectrum. The increase is of 50% between 380 and 750nm: ~44% of AM1.5g solar intensity is absorbed instead of ~29%. These spectra exhibit a decrease in absorption efficiency above a wavelength of λ = 600 nm. This is due to the fast decrease of aSi:H extinction coefficient k towards zero [11]. For comparison, we measured on the same set-up the absorption spectrum of a simple 400nm thick aSi:H layer, plotted on the same graph on Fig. 4(c). The latter absorbs ~40% of solar light on the same spectrum, still less than the 95nm thick 1D PPC, where the volume of absorbing material is more than seven times lower. The absorption spectrum corresponding to the 400nm thick unpatterned layer exhibits oscillations attributed to Fabry-Perot modes ; these are visible at long wavelengths, i.e. provided the absorption coefficient is low enough to enable the observation of such resonances.

The measured values of the absorption are higher than the numerically predicted ones, in particular at high wavelengths. This is first attributed to the measurement uncertainties, which are higher at high wavelengths because of the small measured values. Second, two topographical characteristics of the real PPC structure were not considered in the simulations: the patterns exhibit disorder, and the sidewalls are not strictly vertical, which is expected to increase the incident light coupling efficiency [19]. As shown on Fig. 5 , the ratio between the unpatterned and patterned layer absorption is around 1.35 between 400 and 500nm, where the absorption efficiency of the unpatterned layer is constant at around 50%. In this spectral range, the extinction coefficient of aSi:H is so large that the absorption length is lower than the layer thickness; the PPC patterns offer a way to couple the incident light with a reduced reflectance, which explains measured and predicted ratio. The effect can therefore be considered antireflective. Above 500nm, the absolute values of the absorption are reduced, due to the lower absorption coefficient of aSi:H [11] so photons lifetime corresponds to an optical path length much longer than the layer thickness, allowing light to couple with Bloch resonances of the structure . This leads to the apparition of various damped absorption peaks, observable in Fig. 4(c) as well as on Fig. 5. This is attributed to the impact of the PPC Bloch modes which resonantly increase the absorption. The absorption ratio is thus increasing along with the decrease of the extinction coefficient k. This absorption enhancement finally indicates an increased generation of photocarriers, which is more precisely proportional to the spectrally integrated solar light absorption indicated above, provided appropriate processes suited to a-Si:H solar cells are used to passivate the etched surfaces of the PPC, in order to keep surface recombination low and so as to fully benefit from this photonic engineering scheme.

 

Fig. 5 Experimental absorption ratio between the 95nm-thick patterned and unpatterned layers. No points are shown above 700nm as error bars are larger than the ratio, given the extremely small measured absolute values.

Download Full Size | PDF

5. Conclusion

In conclusion, we have shown both theoretically and experimentally that patterning an aSi:H layer as a PPC yields a 50% increase of the absorption, in a wavelength range as wide as 380-750nm. A patterned structure of 95nm then absorbs incident light with a higher efficiency than a 400nm thick unpatterned layer. Given the limited carrier diffusion of aSi:H, this is of high importance, since such a reduction of the thickness will improve the collection of the photogenerated carriers and thus the solar cell efficiency. Using this very promising concept, we are now developing the fabrication of complete solar cells using such an absorbing structure, together with TCO layers. Given the controllability of the absorption enhancement through photonic band engineering, we believe that this method could also be used for other absorbing materials such as crystalline silicon [20], CIGS, III-V or II-VI semiconductors, but also in order to improve the characteristics of detectors on given spectral bands [21].