† Corresponding author. E-mail:
Project supported by the Natural Science Foundation of Yunnan Province, China (Grant No. 2015FB123), the 18th Yunnan Province Young Academic and Technical Leaders Reserve Talent Project, China (Grant No. 2015HB015), and the National Natural Science Foundation of China (Grant No. U1037604).
To more in depth understand the doping effects of oxygen on SiGe alloys, both the micro-structure and properties of O-doped SiGe (including: bulk, (001) surface, and (110) surface) are calculated by DFT + U method in the present work. The calculated results are as follows. (i) The (110) surface is the main exposing surface of SiGe, in which O impurity prefers to occupy the surface vacancy sites. (ii) For O interstitial doping on SiGe (110) surface, the existences of energy states caused by O doping in the band gap not only enhance the infrared light absorption, but also improve the behaviors of photo-generated carriers. (iii) The finding about decreased surface work function of O-doped SiGe (110) surface can confirm previous experimental observations. (iv) In all cases, O doing mainly induces the electronic structures near the band gap to vary, but is not directly involved in these variations. Therefore, these findings in the present work not only can provide further explanation and analysis for the corresponding underlying mechanism for some of the experimental findings reported in the literature, but also conduce to the development of μc-SiGe-based solar cells in the future.
As the world’s population is continuously increasing exponentially, the need for energy is forecasted to double the today’s requirement for energy by the end of 2025. At present, world energy demand is mostly satisfied by the fossil fuels, which have limited available resources (for example: oil exploitation limit is 43 years; coal exploitation limit is 231 years; natural gas exploitation limit is 62 years; nuclear exploitation limit is 73 years).[1–3] Furthermore, another important consideration of increasing fossil fuel consumption is the influence on environment,[4] for example, burning coal will emit 322.8 g/kW·h of CO2, 1.8 g/kW·h of NOx, and 3.4 g/kW·h of SOx. Therefore, the urgent challenge is to develop and exploit renewable, environment-friendly energy sources. Out of different renewable energy sources known today, solar energy is one of the most abundant, clean, and effective energy sources. The photovoltaic (PV) technology is the most important solar energy application, which can directly harvest and convert solar energy into usable and storable electric energy. However, there are two most drawbacks of PV technology, which hinder its industrial application, i.e., the low conversion efficiency and high production cost.
In the case of device design for solar cells, light management is one of the most important topics. Because most Si-based materials belong to indirect band gap semiconductors, so thin film whose thickness is less than 2 μm could not guarantee to absorb enough solar light. Thus, improving the efficiency of Si-based solar cells must improve the solar light absorption. For this purpose, some light trapping structures (for example, p–i–n or n–i–p device structures, textured substrates, etc.) have been designed to enhance light absorption and improve energy conversion efficiencies.[5] On the other hand, Si–Ge alloying could also achieve this purpose. Matsui et al. fabricated microcrystalline (μc) Si1−xGex thin film by plasma-enhanced chemical vapor deposition, which exhibits infrared response even higher than that of double-thickness microcrystalline Si thin film, and found the optical absorption increases with the increase of Ge content. They have achieved a 6.3% efficiency of a single junction p–i–n solar cells by using a 1-μm-thick μc-Si0.8Ge0.2:H i-layer, which shows excellent performance stability under prolonged light soaking.[6,7] Due to the optical band gap of SiGe alloy decreases with the increase of Ge content and, therefore its optical band gap could be continuous adjusted in a wide range from ∼ 1.1 eV to ∼ 0.66 eV via changing the component proportion in the film. It has the ability to expand absorption of solar spectrum and improve the conversion efficiency of solar cells. This advantage of continuously adjustable band gap makes SiGe alloy thin film used as the intermediate or bottom absorber layer in the multi-junction tandem solar cells.[8] In addition, because its absorption coefficient is 1–2 orders of magnitude higher than that of silicon material, so it could be prepared into thinner film. The latter could shorten the process time and reduce the cost of raw material, which are also important for the industrial production. These advantages make it arouse a great deal of research interesting solar cells applications. Matsui et al. replaced the conventional μc-Si:H with the μc-Si1−xGex:H as infrared absorber in double junction tandem solar cells, in which the bottom cell thickness can reduce more than half while preserving the current matching with hydrogenated amorphous silicon (a-Si:H) top cell. Finally, they obtained an initial efficiency of 11.2% for a-Si:H/μc-Si0.9Ge0.1:H solar cell with bottom cell thickness less than 1 μm.[9]
However, if Ge is incorporated into Si material, the properties of structure and electrics are also consequently changed. Importantly, the bond energies of Ge–Ge bond and Ge–H bond are weaker than those of Si–Si bond and Si–H bond, leading to H atoms’ prior bonding with Si atoms. As a result, the passivation of Ge dangling bond is deficiency, and there are more Ge dangling bonds in SiGe thin film. And when the Ge content is larger than 0.3, the photovoltaic performance begins to degrade: the conduction band tail width is broadened, the electron transfer rate decreases, and the carrier transport is worsened.[10–12] Matsui et al. also found the solar cell parameters for larger Ge contents are lowered by increasing the charge carrier recombination in the μc-Si1−xGex:H layer, and the photocarrier transport in the solar cells for x > 0.2 is dominated by the carrier recombination due to the increased dangling bond defects and the illumination-induced field distortion in the i-layer.[6,7] Recently, Bidiville et al. found that the short-circuit current density of μc-SiGe p–i–n solar cells increases almost 4 mA/cm2 due to oxygen doping. And, they attributed this effect to the oxygen doping compensating for the space charges caused by the germanium dangling bounds rather than the direct defect passivation.[13] In their experiment, the oxygen doping of the i-layer was conducted by using 1% CO2 diluted in H2, because the required gas flow is so small that a diluted gas must be used. During crysctalization of μc-SiGe, oxygen impurities adsorb onto the surfaces or grain boundaries (i.e., surface doping, including the occupation of surface vacancy or replacement of surface atom), and then diffuse into bulk (i.e., bulk doping, including occupation of inner vacancy or replacement of host atom).
In fact, the conversion efficiency of ∼ 1.2-μm-thick μc-Si:H solar cells is deteriorated if the oxygen content in absorber layers exceeds the range from 1.2 × 1019 cm−3 to 2 × 1019 cm−3. Efficiency losses due to the fact that the oxygen impurities lead to the decreases of both fill factor and external quantum efficiency at wavelengths of >500 nm.[14,15] However, a few published articles are available to compare the doping effects of oxygen between μc-Si and μc-SiGe. Motivated by the experimental work in Ref. [13], we believe that it is necessary to systematically study the relationship between electronic structure and photovoltaic performance of O-doped μc-SiGe, in order to develop efficient μc-SiGe-based solar cells. Thus, in the present work, the crystal structure and electronic structure of O-doped SiGe are investigated by using density functional theory (DFT) calculations. Moreover, the relationship between electronic structure and photovoltaic properties will be discussed and compared. Based on the calculated results, possible explanations and underlying mechanism for previous experimental observations will be provided.
In the present work, all of the calculations are carried out by using the periodic density functional theory (DFT) package of Cambridge Serial Total Energy Package (CASTEP) codes.[16] The core electrons (O:[He], Si:[Ne], Ge:[Ar]) are treated as the ultrasoft pseudopotential. The exchange–correlation effects of valence electrons (O:2s22p4, Si:3s23p2, Ge:4s24p2) are described by the revised Perdew–Burke–Ernzerhof for solid (PBEsol) within generalized gradient approximation (GGA).[17] In order to obtain accurate electronic structure, the method of GGA + U is adopted to overcome the well-known shortcoming of GGA,[18] which underestimate the band gap value of semiconductor, via adjusting the parameters of electronic configuration In the present work, the U value of Si\Ge\O-p orbital is set to be Ueff = U − J = 7.90 eV, which have been proved to be efficient for Si-based materials in our previous work.[19,20] Using this GGA+U method, we obtain the accurate band gaps of c-Si (1.119 eV), c-Ge (0.743 eV), α-SiO2 (8.718 eV), which are much close to the experimental measurements. In the DFT calculations, the quality setting affects all relevant task parameters that control the precision of the simulation, which includes the basis set, k-point, and SCF convergence criteria, along with the convergence criteria for the geometry optimization. In order to ensure the accuracy of the results, as well as the convenience of direct comparison, we set it up to the highest standard (i.e., ultra-fine) in the present work. The Monkhorst–Pack scheme k-points grid sampling is set to be 4 × 4 × 2 for the irreducible Brillouin zone. A 60 × 60 × 80 mesh is used for fast Fourier transformation. An energy cutoff of 380 eV is used for expanding the Kohn–Sham wave functions. The Broyden–Fletcher–Goldfarb–Shanno (BFGS) scheme is chosen as the minimization algorithm.[21] Its convergence criteria are set as follows: the force on the atoms is less than 0.03 eV/Å, the stress on the atoms is less than 0.05 GPa, the atomic displacement is less than 1 × 103 Å, and the energy change per atom is less than 1 × 105 eV. Based on the optimized crystal structure, the electronic structure and the optical properties are then calculated. The band structures are calculated along the paths connecting the high-symmetry points.
For the Si1−xGex alloy, there are lots of possible structural configurations, most of which are disorder alloys. In order to clearly understand the doping effect of oxygen, we use cubic Si0.5Ge0.5 alloy configuration (in the context, it is abbreviated by SiGe) with the symmetry of
In this original model, the Ge content is equal to Si content, and Ge atom occupies the equivalent lattice site of Si atom. Choosing this composition we can directly compare the effects of doping oxygen on different positions. In the cases of bulk doping, a 2 × 2 × 2 supercells are used for constructing the O-doped SiGe models. In other words, the original supercells are 11.191 × 11.191 × 11.191 Å3, which is large enough to avoid the self-interaction between impurities in the periodic boundary conditions. To set the substitutional doping model, which is denoted as SiGe:O@Si or SiGe:O@Ge in the following context, one Si or Ge atom in the supercell is replaced by an O atom; while, to set the interstitial doping model, which is represented by SiGe:O@Vac in the following context, an O atom is inserted into the lattice vacancy. In these supercells, the total number of atoms reaches to 64 or 65. In the cases of surface doping, a 3 × 3 super-slab (11.7464 × 11.7464 Å2) for (001) surface or a 2 × 3 super-slab (11.0746 × 11.7464 Å2) for (110) surface (These two surfaces are separated by a 20-Å-thickness vacuum layer) is used for constructing the O-doped SiGe models. Each surface model contains 8 (for (001) surfaces) or 12 (for (110) surfaces) SiGe atomic layers, and thus the size of model along the surface normal direction is larger than 40 Å, which could produce reasonable surficial structure and properties for SiGe. Moreover, in order to mimic the influence of bulk phase, the below half-layers are fixed. To set the substitutional doping model, one Si or Ge atom on the topmost layer is replaced by an O atom; while, to set the interstitial doping model, an O atom is placed onto the surface vacancy. In these surpercells, the total number of atoms reaches to 144 or 145. For the case of interstitial doping model, we first consider all kinds of lattice vacancies, and then compare the total energies, and finally choose the model with the minimum total energy as the stable model in the present work.
Owing to the ionic bonding existing between O impurity and Si or Ge atom, the bond lengths of Ge–O (2.1531 Å) and Si–O bond (2.2310 Å) in the substitutional doping models (SiGe:O@Si, and SiGe:O@Ge) are shorter than the bond length of Si–Ge bond (2.4229 Å), and their crystal volumes are slightly contracted. At the same time, the symmetry of substitutional doping models is changed into
The calculated band structures of pure SiGe and different O-doped SiGe’s are shown in Fig.
Combined with above mentioned band structures, the total and partial density of states (DOS) of pure SiGe and O-doped SiGe are plotted in Fig.
For the SiGe (001) surface, there are two possible terminated planes: the utmost-top plane is for all Si atoms or Ge atoms. That is to say, all the dangling bonds on the terminated plane are either Si dangling bonds or Ge dangling bonds, and they are denoted by SiGe (001)–Si surface or SiGi (001)–Ge surface in the following context. In the present work, the calculated surface energies of these two surfaces are respectively 1.620 J/m2 for SiGe (001)–Si surface and 1.724 J/m2 for SiGe (001)–Ge surface, so the Si atoms could be mainly exposed on the SiGe (001) surface. The most stable surface doping configurations that are obtained by geometry optimization are illustrated in Figs.
For the pure surface, the surface relaxation is very significant: the outmost top atoms are upward shifting by ∼ 0.2 Å. In the case of SiGe (001)–Ge surface, the surface reconstruction is negligible; while in the case of SiGe (001)–Si surface, the surface reconstruction is very significant: Si atoms on the outmost top layer are shifted by ∼ 1 Å along the
In Figs.
The surface band structures and the corresponding densities of sates of different O-doped SiGe (001) are illustrated in Figs.
The above mentioned features can be further explored in the information provided by DOS. For pure SiGe (001) surface, the surface states are mainly caused by the dangling bonds: the Si-3p states dominantly form the top of VB and the bottom of CB for the SiGe (001)–Si surface; while the Ge-4p states dominantly form the top of VB and the bottom of CB for the SiGe (001)–Ge surface. Moreover, in the case of O-doped SiGe (001)–Si surface, the Si-3p states also dominantly form the impurity energy states above the VB. The O doping causes the DOS peaks on the top of VB and the bottom of CB to become stronger than the scenario in the pure SiGe (001) surface. These calculated results suggest that the surface impurity energy states caused by O doping not only enhance the infrared light absorption, but also improve the behaviors of photo-generated carriers (including: generation, transfer, and so on).
In the present work, the calculated surface energy of SiGe (110) surface is 1.290 J/m2, which is quite obviously lower than the surface energy of SiGe (001) surface. Thus, the (110) surface is easily exposed for SiGe material, especially in μc-SiGe. After surface relaxation, the Ge atoms on the topmost layer are outward shifted by ∼ 0.511 Å, while the Si atoms on the topmost layer are inward shifted by ∼ 0.314 Å. Furthermore, on the second atomic layer, the Ge/Si atoms are outward shifted by ∼ 0.149 Å/∼ 0.233 Å. On the other hand, on the unrelaxed (110) surface, Ge atoms and Si atoms are on the identical plane. This means that the surface relaxation is significant on SiGe (110) surface. Along the surface later directions, Ge atoms on the topmost layer are not obviously shifted, while Si atoms on the topmost layer are shifted by ∼ 0.412 Å along the
As shown in the above Fig.
The surface band structures and the corresponding densities of sates of different O-doped SiGe (110) surfaces are illustrated in Figs.
Firstly, because the surface energy of SiGe (110) surface is obviously lower than those of SiGe (001) surfaces, this surface may predominately be exposed in the μc-SiGe material. Thus, for p–i–n absorber layer, the surface property of SiGe (110) surface is a key factor to the final photovoltaic performance of μc-SiGe-based solar cells. In order to understand the electronic structure of O-doped SiGe, the bulk, (001) surface, and (110) surface are chosen as the research objects, and the relevant micro-structures and properties are calculated by DFT + U method in the present work. The lattice distortions caused by O doping for bulk SiGe are not obvious. In the case of surface doping, O impurity leads to the significant surface relaxation or surface reconstruction on the topmost layers. The impurity formation energy of O interstitial doping is lower than that of O substitutional doping for surface model, implying that O impurity prefers to occupy the lattice vacancy site on SiGe surface. By O doping, the band gaps are slightly narrowed, except in the case of O-doped SiGe (110) surface. Importantly, there are abundant impurity energy states in the band gap of O@Vac surface model. These doping effects not only enhance the infrared light absorption, but also improve the behaviors of photo-generated carriers. On the other hand, the surface charge and potential decrease in the cases of SiGe (001)–Ge surface: O@Vac and SiGe (110) surface O@Vac, in which (001)-Ge surface and (110) surface have the lower surface energies, and the corresponding impurity formation energy of O interstitial doping is lower. Thus, the reducing surface work function could confirm previous experimental observation: oxygen doping can compensate for the space charges caused by the germanium dangling bonds. In all cases, O doing mainly induces the electronic structures near the band gap to vary, but is not directly involved in these variations. The findings in the present work could well explain the experimental observations in the literature, and conduce to the development of μc-SiGe-based solar cells.
1 | |
2 | |
3 | |
4 | |
5 | |
6 | |
7 | |
8 | |
9 | |
10 | |
11 | |
12 | |
13 | |
14 | |
15 | |
16 | |
17 | |
18 | |
19 | |
20 | |
21 | |
22 |