1. Introduction

Harnessing an underutilized electro-magnetic (EM) spectral region can become an essential technology for high efficiency solar cells set to provide a sustainable energy supply. In the milli-/micro-meter wave spectral region, the rectenna is used to capture and rectify propagating EM radiation [1, 2]. In ultraviolet (UV), visible (VIS) and near-infrared (NIR) wavelength regions, photovoltaic solar cells are promising EM energy harvesters and represent a mature contemporary technology [3]. However, in the intermediate mid-infrared region there are still no effective solutions for energy harvesting. The tail of solar spectrum extending down to 4000 nm wavelengths is not utilised. Furthermore, harvesting close to 10¹⁷ W of mid-IR thermal radiation from Earth’s emissions is a significant concern for the renewable energy community [4].

Effective use of light absorption for solar energy harvesting is of primary research interest for ensuring a sustainable energy supply. Recently, black silicon (b-Si), which has a surface covered with pyramidal nano-spikes, was used for efficient light of sub-bandgap energy photons [5], surface enhanced Raman scattering (SERS) sensors [6], and as a bactericidal surface [7]. Photo-voltaic solar cell technology is limited in its energy conversion efficiency by the light absorption window of the active medium and the ray optical limit for light capture [8, 9]. In the case of crystalline Si, conversion efficiencies around 25% have been reached [10], however, a significant part of the solar spectrum at IR wavelengths falls through the transparency window. Altogether, including heat losses approximately 80% of total solar light energy remains unutilized, as shown in Fig. 1(a). To expand the spectral range accessible for energy harvesting into the near-IR up to 1200 nm wavelengths, a textured surface of Si can be used [11]. A typical semiconductor solar cell has a limit for energy conversion down to 1300 nm wavelength [12]. Conversely, black body radiation at around 5772 K from the Sun’s surface is cut off by atmospheric absorption at wavelengths above 2500 nm. Hence, utilization of the residual thermal part of solar energy at IR wavelengths can provide further advantages for solar energy harvesting [3].

 

Fig. 1 (a) Air mass 1.5 solar light energy spectrum [23] with portions utilized by a crystalline Si solar cell, losses due to transparency and heating highlighted separately. (b) Schematic illustrating the concept of a hybrid cell for solar-to-electrical energy conversion by harnessing the full solar spectrum; the n-type Si region is formed on b-Si by phosphorus doping the the p-type b-Si substrate.

Download Full Size | PDF

Black-Si solar cells with efficiencies above 18% have been demonstrated [13] without an anti-reflection coating, with an even higher 22.1% conversion efficiency reached when the surface was passivated using atomic layer deposition [14]. The tapered needles create a gradual refractive index change, which reduces reflectivity [15]. Antireflective natural nanostructured surfaces have a similarly tapered spike morphology [16,17]. The porosity, surface chemistry and passivation of Si can be controlled by electrochemical etching [18]. Solar cells with light absorbing efficiencies of over 90% within thickness of single micrometers have been proposed using slanted conical holes [19]. Recently, quantum efficiency close to 100% was demonstrated in photodiodes using b-Si with alumina top-layer made by atomic layer deposition [20].

In previous work [5] it was demonstrated that solar energy absorption by b-Si can be used for thermal-to-electrical energy conversion. Herein, a Ge-Sn layer was added for IR light absorption to further improve the light absorption efficiency. Ge and Sn, just like Si, are IV group elements, however, exhibit narrower band gaps: 0.67 eV (Ge) and ∼ 0 eV (Sn) at room temperature and form a solid solution at all atomic mixing ratios. By adding Sn to Ge, it is possible to reduce the band gap to ∼ 0.2 –0.4 eV at 15% of Sn. Therefore the absorption edge can be tuned towards longer wavelengths as the concentration of Sn is increased [21]. With an indirect band gap of 1.11 eV Si can absorb the shorter wavelength segment of the solar spectrum up to 1100 nm. However, the spectrum of solar light reaching the surface of the Earth extends up to 2500 nm for Air mass 1.5 condition, as shown in Fig. 1(a). Those wavelengths could potentially be harnessed and utilized by using narrower band gap materials required for absorption of long wavelength light. For the same purpose, graphene or graphite could be used. However, considering the ease of deposition process and durability, a Ge-Sn system is better for practical applications.

Figure 1(a) highlights the segment of the solar spectrum that can be converted into electricity by means of Si photovoltaics (assuming 25% efficiency). The neglected IR part of the spectrum can be harvested using a hybrid cell conceptually outlined in Fig. 1(b). The upper segment of the device has a p-n junction and operates as a standard high efficiency Si solar cell. Textured b-Si top surface serves to reduce reflectivity and thereby to increase the achievable absorption efficiency of solar radiation. The lower section of the Si solar cell has a Ge-Sn absorber layer deposited in order to harvest the IR portion (λ ≥1100 nm) that is beyond the Si absorption edge.

Here, a concept of a hybrid solar cell is proposed to maximize the utilization of solar energy and to increase the total efficiency of system. It is based on b-Si for photovoltaic harvesting over visible wavelengths of solar spectrum and complimented with a film of Ge-Sn to collect the IR portion of the solar energy as heat and convert it to voltage using a thermoelectric Seebeck element. A schematic illustration of the concept is shown in Fig. 1(b).

2. Samples and methods

2.1. Ge-Sn alloy as IR light absorber

Formation of the near-IR absorbing layer was carried out by thermal evaporation of Ge and Sn onto Si substrates. Ge and Sn were simultaneously evaporated from the same crucible. In the resulting film the atomic ratio of Sn to Ge was 20%. Structural and compositional characterization of Ge-Sn films was carried using scanning electron microscopy (SEM) and energy-dispersive X-ray spectroscopy (EDS). As shown in Fig. 2(a) and its inset, Ge-Sn were thoroughly intermixed in the evaporated film.

 

Fig. 2 (a) Ge-Sn film on b-Si: scanning electron microscopy (SEM) and energy-dispersive X-ray spectroscopy (EDS) images. Single crystal p-type Si was used as the substrate in the final solar cell design shown in Fig. 1. (b) Reflection spectra of b-Si/Ge-Sn surface from visible-to-near-IR wavelengths in comparison with b-Si (Data from ref. [24]) and a mirror-polished Si substrate (IR detector was not optimized for the short wavelength visible part of the solar spectrum). Interference due to a thin-film etalon effect is apparent in the b-Si/Ge-Sn reflectivity. Background shaded region shows the characteristic intensity profile of the solar spectrum. (c) FDTD simulated spectral dependence of absorbance in the structure depicted in panel (a), illustrating the separate contributions of b-Si and Ge-Sn layers. (d) FDTD simulated spatial distributions of absorbance at wavelengths λ = 500, 800, and 2000 nm.

Download Full Size | PDF

Solar cell reflectance and light source spectra were measured using conventional microscopic/macroscopic techniques. In VIS-NIR wavelength region spectral acquisition was performed using a mini-spectrometer (C10083CAH, Hamamatsu Photonics, 350–1050 nm) and an optical spectrum analyzer (Q8344, ADVANTEST, 400–1600 nm). For mid-IR region, a FT-IR spectrometer, customized with an external light source and a microscopy unit (FT-IR 4200, JASCO 1000–4000 nm), was used with mercury cadmium telluride (MCT) and triglycine sulfate (TGS) detectors. The light source spectrum was used without calibration to the sensitivity of the detector.

2.2. Numerical simulations

Reflectance, transmittance and absorbance of b-Si/Ge-Sn composite materials was modeled using a commercially available finite-difference time-domain (FDTD) photonic simulation engine (FDTD Solutions, Lumerical Solutions Inc.). Surface topology of b-Si anti-reflectance layer was digitally reconstructed based on SEM images. Spectral dependencies of complex refractive index values were used as reported in the literature for Ge-Sn [21] and Si [22]. Near-IR absorber alloy composition was estimated to be Ge_0.85Sn_0.15 based on EDS data.

Simulations were performed for two separate geometry types: for the full device setup as outlined in Fig. 1(b), and for the thin-film configuration of the test structure in Fig 2(a). In all cases the 3D calculation region spanned a 2.5 × 2.5 μm surface segment. Periodic boundary conditions were used on the planes bordering the simulated absorber stack. Conversely, perfectly matched layer boundaries terminated the simulation region along the light propagation direction. Optical excitation was simulated using a linearly polarized plane wave source, with an extended spectrum spanning the λ = 300−2500 nm range relevant for solar energy harvesting. Simulations of the full device structure (Fig. 1(b)) were performed separately for the top b-Si anti-reflectance layer and for the bottom Ge-Sn absorber. The mutual interactions of those two sections through the intermediary 500 μm Si wafer were subsequently modeled using the Beer-Lambert law.

2.3. Photovoltaic/photothermal hybrid cell

The hybrid device was fabricated by first preparing a photovoltaic solar cell with b-Si antireflective nanotexturing, and subsequently applying a photothermal NIR absorber module on its back-surface.

The photovoltaic cell was made by first dry-etching the top surface Si layer by inductive coupled plasma reactive ion etching (ICP-RIE) method using a Samco RIE-101iPH tool to form b-Si. Subsequently, a p-n junction was formed by means of a sol-gel process. The surface of intrinsically p-doped Si was spin coated with a phosphorous based dopant agent (EPLUS SC-909, Tokyo Ohka Kogyo Co., LTD,) at 3500 rpm for 20 s. Then, the substrate was pre-annealed at 200°C for 60 s followed by a successive annealing at 950° for 10 min. This gave rise to a n-doped surface region, and a resultant p-n junction. The oxidized surface layers were removed by dipping into a 1% by volume aqueous HF solution for 15 min. Patterned Ag electrodes were formed on both sides of the Si substrate by using a metal mask and thermal evaporation.

The photothermal NIR absorber module was created by first electrically insulating the bottom surface of the photovoltaic cell. For this purpose a ∼100 nm thickness SiO₂ was deposited. Then, the ∼ 4 μm thickness Ge-Sn layer was evaporated to provide near-IR absorption at wavelengths transmitted through the photovoltaic cell, as described previously. This hybrid device unit was attached to a commercially available Seebeck element (TEP1-1264-1.5, Nihon Techno Ltd.) using a heat conductive grease (X-23-7877, Shinetsu silicone) to ensure good thermal contact. Conversion efficiency of the Seebeck device depends on the ambient temperature. At room temperature the maximum efficiency is around 4%. The opposite side of the Seebeck element was put on an air cooled heat sink in order to keep the device at room temperature. While the device can operate without a cooling system, thermal control is useful in ensuring the stability of photo-thermal measurements. Furthermore, the heat-to-electrical power conversion of the device was benchmarked against either an additional air cooled heat sink and/or Peltier element (UT-7040WJ-HS100, with temperature controller), as reported previously [5].

3. Results and discussion

3.1. Characterization and performance of Ge-Sn layer as IR light absorber

Figure 2(a) shows the SEM and structural composition EDS images of a sample’s cross section, revealing fractions of Si, Ge and Sn, respectively. The EDS image reveals a gradual shift in composition of the Ge-Sn absorber layer as its thickness increases. Due to the more rapid evaporation of Sn, which has a lower melting point and vapor pressure than Ge, the Ge/Sn atomic ratio shifts from an initial Ge_0.65Sn_0.35 to Ge_0.93Sn_0.07 throughout the course of the deposition.

The antireflective top b-Si surface exhibits sub-micrometer (0.1 − 0.9 μm) scale roughness. A leaf-like fractal surface with different scale roughness is effective in the reduction reflectivity over spectrally broad spectral range [24]. Fabrication of cascading nano-micro feature-size structures with extremely low-reflectivity down to 0.1%, high-transparency, and low scattering of light, to maximise the total absorbed energy in Si, has been demonstrated by combining several Si processing techniques [24]. The reduced reflectivity due to a gradient nature of refractive index change on b-Si surface can be beneficial for enhanced absorption in the hybrid b-Si/Ge-Sn surface as shown in Fig. 2(b). The fabricated surface shows less than 5% reflectivity up to an IR wavelength of 3000 nm and no light transmission through Ge-Sn layer could be observed, hence, most of incident light which falls within the transparency window of Si has been absorbed in Ge-Sn layer.

Experimental observations of the composite absorber performance are corroborated by FDTD simulation results depicted in Fig. 2(c). Absorbance exhibited by the b-Si/Ge-Sn stack ranges from above 90% at the visible wavelengths, down to around 70% at near-IR. Spectral dependence of absorbance shows how the two different materials, namely Si and Ge-Sn, provide complimentary contributions to the overall optical energy uptake. The thin 1 μm thickness b-Si top layer substantially decreases the total reflectance of the material stack, however, due to its low optical path, absorbs only the shortest wavelength portion of light. Remainder of the energy at wavelengths λ > 700 nm is preferentially absorbed in the 4 μm thickness Ge-Sn layer. The interference fringes come as a result of the multiple reflections inside the top b-Si film and are also observed in the experimental reflectance shown in Fig. 2(b). FDTD simulations likewise show that a majority of optical losses are due to reflectance.

Spatial distribution of absorbance at select wavelengths is shown in Fig. 2(d). Evidently, at visible λ ∼ 500 nm wavelengths, most of the absorbance occurs in the top b-Si layer. Silicon absorbance of near-IR λ ∼ 800 nm light is fairly weak and the short 1 μm optical path proves insufficient, hence most of energy uptake occurs in Ge-Sn. This radiation exhibits photon energies in excess of the indirect band gap of Si, therefore if absorption length was longer it could partake in photovoltaic conversion more productively than by photothermal absorption in Ge-Sn. Lastly, long wavelength radiation λ ∼ 2000 nm cannot be efficiently absorbed in Si by any mechanism save for weak free carrier absorption, therefore, is harvested only in Ge-Sn.

Figure 3(b) shows the photothermal response from b-Si and Ge-Sn/b-Si materials used as absorbers in a thermal-to-electrical conversion setup. Performance of the Ge-Sn/b-Si absorber was benchmarked against textured b-Si, which is a hierarchical combination of an alkaline etched micro-pyramidal surface texture of Si, overlayed by subwavelength b-Si spikes formed through plasma etching. The surface has a fractal-like self-affinity and exhibits a lower reflectivity, especially towards the near-IR wavelengths [24]. Illumination by a metal halide lamp, whose spectrum spans the visible range and has no appreciable output at wavelengths λ ≥ 1000 nm, results in similar thermoelectric voltage response irrespective of the choice of absorber type shown in Fig. 3(b). This is mainly due to source spectrum matching the absorbance of silicon.

 

Fig. 3 (a) Spectra of light sources used to simulate visible and visible-to-near-IR illumination. The combined halogen and metal halide light source had a total intensity I_l = 72.8 mW/cm². Inset shows a simplified schematic of experiment: light source is illuminating nanotextured surface with tailored absorption at near-IR spectral range, all placed on a Peltier element. The right-panel spectra were acquired at the near-IR to mid-IR spectral range, as measured using FT-IR. (b) Photo-thermal response of b-Si, Ge-Sn on b-Si and textured b-Si.

Download Full Size | PDF

However, when a halogen lamp with an extended spectrum up to λ ≥ 3000 nm is used to co-illuminate the light harvesting surfaces, as shown in Fig. 3(a), performance differences become apparent. The Ge-Sn with b-Si shows stronger output voltage response as compared to the other absorbers. This is a direct indication of stronger heating of the illuminated surface. Despite having the lowest reflectance, textured b-Si does not posess a mechanism for absorbing sub-band gap wavelengths, hence, this radiation cannot be used for photo-thermal generation to the detriment of total efficiency. From this result it is clear that Ge-Sn layer operates as the light absorber at longer than 1100 nm wavelengths. The demonstrated IR light-to-electricity conversion can contribute to a spectral widening of the overall light-to-electricity conversion.

3.2. Photovoltaic/thermoelectric hybrid solar cell

In order to transfer the broadband absorber concept outlined in Fig. 2 into a hybrid solar cell the portion of optical energy available for photovoltaic conversion needs to be maximized. Therefore, when no additional light-trapping mechanisms are used, silicon substrate should be at least 100 μm thick to maintain an adequate photovoltaic performance. The device prototype sketched in Fig. 1(b) uses a 500 μm thickness Si layer, with a Ge-Sn absorber applied on its back side.

Normalized absorbance FDTD simulation results for such a structure are displayed in Fig. 4(a). Similarly to the spectral distribution of absorbance shown in Fig. 2(c), here short wavelength radiation is absorbed in front-facing Si, and the residual near-IR contribution in the Ge-Sn layer. However, in this case Si absorbance extends up to λ ∼ 1000 nm wavelengths. This difference in relative energy uptake in Si and Ge-Sn is especially apparent in the spatial distribution of absorbance shown in Fig. 4(b). Near-IR interference spectroscopic features in this case appear as a result of multiple reflectance in the 100 nm thickness SiO₂ spacer layer between Si and Ge-Sn. Another noteworthy difference is that absorbance at near-IR averages around 60%, hence, slightly lower than exhibited by the thin-film absorber in Fig. 2. This can be attributed to substantial reflectance at either Si/SiO₂ or SiO₂/Ge-Sn interface. Therefore, reduction of such reflection by, for example, using high refractive index dielectrics like as Al₂O₃, TiO₂, HfO₂, would improve overall absorption and efficiency.

 

Fig. 4 (a) FDTD simulated spectral dependence of absorbance in the hybrid photo-voltaic/thermovoltaic solar cell depicted in Fig. 1(b), illustrating the separate contributions of b-Si and Ge-Sn layers. (b) FDTD simulated spatial distributions of absorbance at wavelengths λ = 500, 800, and 2000 nm.

Download Full Size | PDF

Figure 5 shows the photovoltaic response of the hybrid solar cell, schematically outlined in Fig. 1(b), with and without a Ge-Sn bottom absorber layer. Regardless of presence of the Ge-Sn layer, the photovoltaic response was approximately the same. This result indicates that the bottom thermal harvesting section does not affect the performance of photovoltaic solar-to-electrical conversion. Therefore it can be concluded that the 100 nm thickness SiO₂ insulator layer was sufficient in preventing degradation of the photovoltaic performance. When the Ge-Sn absorber was deposited without the SiO₂ spacer the observed photocurrent and voltage were both lower, presumably due to electrical conduction losses.

 

Fig. 5 Photovoltaic performance of the hybrid cell with combined solar-to-electrical and thermal-to-electrical action. The I − V characteristic of a solar cell with and without Ge-Sn. The inset shows the photothermal response of the cell with a Ge-Sn layer. Key parameters are: short circuit current I_sc = 11.06 mA, open circuit voltage V_oc = 0.46 V, maximum power P_max = 1.72 mW, fill factor FF = P_max/(I_scV_oc) = 33.8% (for solar harvesting surface area S = 1 cm²). The top-side was textured by wet etching with μm-sized pyramidal structures.

Download Full Size | PDF

The inset in Fig. 5 shows the additive photothermal response of the hybrid cell with Ge-Sn absorber layer. Illumination of the combined halogen and metal halide lamps had the cumulative intensity of I_l = 72.8 mW/cm². Photovoltaic conversion efficiency was ${\eta }_{\mathit{PV}}\equiv \frac{P_{\mathit{max}}}{P_{\mathit{in}}}=\mathit{FF}\times I_{\mathit{sc}}{V}_{\mathit{oc}}/I_l\simeq 2.4\%$ calculated for a short circuit current I_sc = 11.06 mA/cm², open circuit voltage V_oc = 0.46 V, maximum power P_max = 1.72 mWcm², and fill factor of FF = P_max/(I_scV_oc) = 33.8%. Conversely, thermoelectric generation added an additional V_th ≃ 10.6 mV voltage, I_th ≃ 12.2 mAcm² current density, and P_max = 0.129 mW/cm² power to the total output of the hybrid device. Hence, thermal energy can add 7.5% of conversion efficiency to the photovoltaic system in this experiment.

The major advantage of this hybrid solar cell system is that, both photovoltaic and thermo-electric output can be used not only in conjunction but also independently. For example, if thermoelectric energy conversion efficiency starts to approach that of the photovoltaic solar cell, it is possible to combine the output of both cell components. Otherwise, thermal energy can be used to drive the peripheral equipment, such as temperature monitoring and irradience sensors or for angular orientation adjustments of the cell panels to maximize power generation. In addition, this hybrid system can provide cooling for the photovoltaic device through the Seebeck element. Thereby additional water cooling that is used in commercial solar panel system to maintain cell performance could be relinquished.

The performance of both photovoltaic and thermoelectric cells could be improved in the future of our work. For instance, adding surface passivation by silicon oxide or silicon nitride layer would mitigate surface recombination and further improve anti-reflectance with minimal effect on near-IR transmittance. Further improvements would involve appropriately balancing the electron-hole density, optimization of the p-n junction profile, as well as the thickness of substrate. The energy conversion efficiency of thermal cells is still fairly low, however new concepts of thermoelectrical converter such as a spin-Seebeck device [25–27] are expected to increase the conversion efficiency towards approaching that of solar cells. These Seebeck devices are also producible by various deposition methods, hence would be compatible with our devices.

Presented analysis proves that a hybrid cell equiped with a Ge-Sn absorber can, indeed, harvest the IR part of solar spectrum and convert it to electrical energy without compromising the photovoltaic efficiency over the visible solar spectrum at a maximum of Seebeck device conversion efficiency. The sun facing surface of b-Si also serves a function of reducing reflectivity over IR wavelengths [28] which contributes to better absorption of sub-band gap reaching the buried Ge-Sn absorber layer. Heat removal from p-n junction by photo-thermal conversion also contributes to a higher efficiency of photo-voltaic conversion, especially in hot-climate operation conditions.

4. Conclusion

A hybrid photovoltaic/thermoelectric cell based on b-Si with a Ge-Sn IR radiation absorber is demonstrated. The Ge-Sn layer works as an absorbing layer for the near-IR wavelengths, longer than 1100 nm, and contributes up to 7% to the output voltage by thermal-to-electrical conversion (without consideration of an improved photovoltaic conversion due to cooler p-n junction). It is demonstrated that by using such a composite material it is possible to harvest the full spectrum of sunlight by absorbing the residual radiation that falls within the Si band gap and converting it to thermal energy, which is then in turn converted into electricity. The hybrid solar cell with dual photovoltaic and photothermoelectric action was made by a simple process of n-p junction formation on p-type b-Si by thermal annealing and by addition of a Ge-Sn layer for IR absorption deposited on the bottom-side of solar cell. Thin films absorbing at IR wavelengths can also find applications in the field of fast and simple bolometers [29]. The presented idea to add a thermal energy harvesting is applicable not only to solar cells, but to rectennas or any other EM radiation harvesting system.