CHAPTER 2 LITERATURE REVIEW
2.2.3 Chemical Annealing Model
The chemical annealing model has been constructed in order to explain the crystallization of amorphous silicon when exposed to a pure H2 plasma [19]. Indeed, µc-Si:H can be produced by using a layer-by-layer technique by alternatively depositing a thin a-Si:H layer and expose it to a pure H2 discharge which crystallizes the thin layer. Neither the etching model, nor the surface diffusion model, can explain this phenomenon. This third model is based on the chemical reaction of atomic hydrogen coming from the plasma with hydrogen bonded to silicon at the film surface or sub-surface as depicted in Fig. 2-7. The reaction creates a silicon dangling bond and a hydrogen molecule. This reaction is exothermic, and the resulting structure thermal vibration promotes surface and bulk structural rearrangement leading to the energetically more favorable µc-Si:H [15]. The Si dangling bond created is then transformed into a more stable and rigid Si-Si bond or, if placed at thefilm surface, re-hydrogenated by atomic hydrogen from the plasma. In the selective etching model, the H atoms attach to silicon and re-hydrogenate the bulk silicon until silane desorption. But in the chemical annealing model, the H atoms recombine with surface or sub-surface hydrogen.
Fig. 2-7 Schematic representation of the chemical annealing model. The hydrogen atoms from the plasma recombine with hydrogen bonded to surface or sub-surface silicon atoms, delivering vibrational energy which favors silicon crystallization [15].
20
Consequently, the atomic hydrogen flow rate towards the surface has to be large compared to the silicon radical flux to promote the growth of µc-Si:H. This is to increase the surface diffusivity by a fully H-covered surface, to remove undesirable a-Si:H by selective etching by H atoms, or to generate vibrational energy by hydrogen surface or sub-surface recombination, according to the surface diffusion, the selective etching or the chemical annealing models, respectively.
21
Chapter 3
Experimental Technique
3.1 Plasma-Enhanced Chemical Vapor Deposition
The deposition method used to produce the μc-Si:H film is rf (the excitation frequency is 27.12MHz )PECVD. The plasma provides some of the activation energy required for the chemical reaction, in effect reducing the processing temperature required during the film deposition. It is done by collisions with electrons, which originate as secondary electrons in the plasma and build up their energy by acceleration in an electric field. The μc-Si:H film is deposited by attaching reactive particles of dissociated silane molecules, called radicals, to the surface of growing film. Some of the energy transferred to silane molecules in the collisions with electrons is radiated as visible light, for which reason the deposition method is also referred to as glow discharge.
3.2 Raman Spectrum
Fig. 3-1 Micro-Raman spectrometer composed of an He-Ne laser source, a microscope, a notch filter, a grating and a CCD camera [20].
22
Raman spectroscopy consists in the observation of inelastic scattering of an incident light beam by a media which could be a gas, a liquid or a solid. The Raman effect, first reported by Raman and Krishna in 1928 [20], has been used in this work to evaluate the degree of crystallinity of the deposited silicon layers. The Raman apparatus was a Renishaw RM series with an He-Ne laser source (633 nm).
Silicon films with mixed amorphous-nanocrystalline phase composition, the spectrum consists of a broad low-frequency component peaking around 480 cm-1 and related to the amorphous phase and a substantially narrower peak at 520cm-1 whose increase in intensity indicates nucleation of the nanocrystalline phase. The total scattering intensity I( )ω in the
frequency range under study can be written as the formula.
( ) ( )
( ) c a
I ω =I ω +I ω Eq. 3-1
where Ic
( )
ω is the intensity of the line related to the nanocrystalline phase (520cm-1), and(
Ia ω
)
is that of the line associated with the amorphous phase (480 cm-1).Now, considering the relation between the integrated raman intensity and the volume fractions of the amorphous and nanocrystalline phases. This relation is usually written in the Eq. 3-2, where y is the scattering factor.
c a And this is a good agreement with the estimate y=0.88. Thus the formula of crystallinity is described as Eq. 3.5.
Chapter 4
Results and Discussions
4.1 Hydrogenated Amorphous Si Solar Cell
Hydrogenated amorphous Si (a-Si:H) material have received a great attention because of their use in low-cost solar cells. In order to analyze and optimize the performance of thin-film a-Si:H solar cells, it is important to set up an accurate numerical model to simulate the transport mechanisms in the device. Device simulation comprises methods of basic carrier transport and continuity equations for any type of structures defined by two dimensional doping profiles. Silvaco (Atlas), which is based on finite element analysis, is widely used by research institutes and industries. Through Atlas, we input several parameters including:
device structure, doping concentration, and bias condition; then, using numerical methods such as: electron and hole current equation, electron and hole continuity equations, and Poisson equation; finally getting the outputs like mobile carriers, electric fields, potentials, and currents in the form of two-dimensional contours and vectors as well as quasi three-dimensional contours. Besides simulating the external current-voltage characteristics, it allows a detailed simulation of physical behaviors of devices in both steady-state and in transient regime. Device simulation has been widely used to study a-Si:H p-b-i-n solar cell performance, which is sensitive to the material properties.
4.1.1 Simulation Model
The physical models that we have used are Klaassen’s concentration dependent SRH model (KLASRH), Klaassen’s Auger model (KLAAUG), Klaassen’s concentration dependent lifetime model, and Klaassen’s low field mobility model (KLA). The photo-generation model which includes a ray tracing algorithm is used to calculate the transmission and absorption of light in the bulk as well as reflection and refraction at the interfaces. The p-i-n solar cell is
24
operated under the global standard solar spectrum (AM 1.5G) illumination with total incident power density which is 100 mW/cm2 and the light intensity for each wavelength is calculated in the wavelength from 0.3 to 1.2 μm. A structure consists of flat glass/ITO/a-Si:H p-i-n/Ag, is considered as the solar cell model for simulation, in which light penetrate through the p-layer.
The figure is shown as Fig. 4-1.
Fig. 4-1 Schematics of an a-Si solar cell structure, which consists of glass/ITO/a-Si:H p-i-n/Ag
We set the solar cell in length and width both as 1 μm . The thicknesses of i-layer is 450 nm. We assumed that interfaces are all flat for this simulation. The doping concentration of p- and n-type set as 3x1018 cm-3 and 9x1018 cm-3at first. Each dangling bond density in p-layer and n-layer is set as 5x1018 cm-3 and 9x1018 which is higher than that in other layers. It is because the dangling bond density would increase with higher doping concentration in a-Si:H material. Also, we have analyzed the effect of light soaking would increase dangling bond density (Ndb) in a-Si:H layer. The electrical, optical, and structural parameters in the
25
Table 4-1 which were adapted from the literature [22, 23] are used for simulating the device performance. The distribution of density of states (DOS) in the forbidden energy gap of undoped a-Si:H i-layer is used in the simulation.
Table 4-1 Device Parameters of a-Si:H solar cell used in Simulations.
Parameter p-a-SiC:H b-a-SiC:H i-a-Si:H n-a-Si:H
Layer thickness (nm) 5~20 1~10 250~500 10~30
Mobility gap (eV) 2 1.96 1.86 1.8
Donor doping density (cm-3) 3x1018
acceptor doping density (cm-3) 9x1018
Electron mobility(cm2/V s) 20 20 20 20
Hole mobility(cm2/V s) 4 4 4 4
Electron life time (μs) 0.01 0.01 0.01 0.01
Hole life time (μs) 0.1 0.1 0.1 0.1
Effective DOS in the valance and conduction bands (cm-3)
2x1020 2x1020 2x1020 2x1020
Exponential tail Prefactors NTD, NTA (cm-3eV-1)
4x1021 4x1021 4x1021 4x1021
Characteristic energy WTD (VB tail) (eV) 0.12 0.11 0.08 0.05
Characteristic energy WTA (CB tail) (eV) 0.07 0.055 0.05 0.03
Gaussian distribution density NGD, NGA (cm-3eV-1)
5x1018 5x1017 5x1015 9x1018
Characteristic energy for Gaussian distribution WTD (donor like state) (eV)
0.2 0.2 0.2 0.2
Characteristic energy for Gaussian distribution WTA (acceptor like state) (eV)
0.2 0.2 0.2 0.2
Peak of donor like Gaussian distribution EGD (eV)
0.8 0.82 0.83 0.78
Peak of acceptor like Gaussian distribution EGA (eV)
0.9 0.67 0.67 0.52
Correlation energy U (eV) 0.3 0.47 0.36 0.5
Transmittance of glass / ITO 0.9
Reflectivity of n layer / metal contact 0.9
Surface recombination velocity 1x107 1x107
a measured from the conduction band edge ; b measured from the valance band edge
26
Fig. 4-2 is the general standard model for DOS of a-Si:H i-layer and is described by : 1. A parabolic conduction band and exponentially decaying conduction band tail.
( ) exp C 2. A parabolic conduction band and exponentially decaying conduction band tail.
( ) exp V
3. Two equal Gaussian distributions of states around the middle gap separated from each other by a correlation energy (U) for representing the defect state related to dangling bonds (DB+/0 and DB0/-).
Fig. 4-2 Distribution of acceptor/donor-like trap states across the forbidden energy gap used in i-layer in the simulation.
27
The device performance of the a-Si:H solar cell is studied by using Atlas. The electric characteristics of a-Si:H solar cell are as following:
Fig. 4-3 The illuminated J-V characteristic curve of a-Si:H p-i-n solar cell
The computer simulation of a-Si:H p-i-n solar cell was carried out by atlas. The short circuit current (Jsc), open circuit voltage (Voc), fill factor (FF), and the efficiency of a-Si:H solar cell are 13.3 mA/cm-2 , 0.83V, 0.74, and 8.37 %, respectively in Fig. 4-3. We considered the lower efficiency resulted from two reasons. First, the flat interfaces would reduce the light refraction and reflection. Second, the lack of TCO between the n-layer/Ag interface also made the efficiency low. The performance of the cell within light soaking effect is analyzed by increasing Ndb from 5x1015 cm-3 to 5x1018 cm-3 as shown in Fig. 4-4 and Fig. 4-5. The open-circuit voltage Voc, fill factor FF, and short-circuit density Jsc were all decreased due to the Ndb increased. a-Si:H would degrade upon exposure to sunlight. This phenomenon, called the Staebler-Wronski effect or light soaking effect, causes large increases in defect density and is reversible when annealed at temperatures above 150 oC. The metastable defects are
28
believed as dangling bonds formed by breaking weak bonds in the random network. In order to eliminate this effect, reducing i-layer thickness and hydrogen content in the film are two better ways so far. Thus, the a-Si:H i-layer of tandem solar cell is also designed to be thinner to achieve this purpose.
Fig. 4-4 Photovoltaic performance of a-Si:H solar cell with different Ndb of i-layer.
29
Fig. 4-5 The J-V characteristics curve with different Ndb in i layer of a-Si:H cells.
4.2 Microcrystalline Silicon solar cell
The structure of microcrystalline p-i-n solar cell is as following: flat ITO/p-μc-Si:H / i-μc-Si:H/n-a-Si:H/Ag. The open-circuit voltage, short-circuit current density, and the fill factor determine the output properties of a solar cell; however, unlike the other two parameters, the behavior of the open-circuit voltage (Voc) is not easy to be understood. It is well known the Voc in hydrogenated microcrystalline silicon (μc-Si:H) thin-film solar cell declines sharply when the crystalline fraction increases from 60% to 90% or even higher. The tendency of different crystallinities of μc-Si:H cells has been studied until now, but it still lacks detailed quantification of the impact of individual material properties on solar cells.
Here, we used a powerful electrical-optical computer modeling program (Atlas) to simulate the optimal μc-Si:H thin-film solar cells to understand the difference between the a-Si:H and μc-Si:H solar cell.
30
4.2.1 Simulation Model
The μc-Si:H solar cell structure is shown as Fig. 4-6. The length and width of solar cell are both set as 1 μm . The thicknesses of i-layer is 2200 nm. We assumed that interfaces are all flat in this simulation. Each doping concentration of p- and n-type initially set as 1x1019 cm-3and 3x1018 cm-3to optimize cell performance.
Fig. 4-6 μc-Si:H solar cell structure.
To model all aspects of the solar cell performance accurately, we had to assume that the more crystallized material has a lower band gap, higher carrier mobilities, and both higher mid-gap defect density and narrower band tails. A lower band gap for more crystallized μc-Si:H has been previously measured [24] by in situ Kelvin probe analysis and the “Flat Band Hetero-junction” technique; it also measured by photoluminescence (PL) in the study of Merdzhanova et al. [25] . In a-Si:H, the bonding length and bonding angle altered slightly but still in rule in short range, whereas atoms arrange randomly in long range. It makes the tail distribution is broader than the crystalline material. The physical simulation model are
31
Klaassen’s concentration dependent SRH model (KLASRH), Klaassen’s Auger model (KLAAUG), Klaassen’s concentration dependent lifetime model, and Klaassen’s low field mobility model (KLA).
Table 4-2 Device Parameters of μc-Si:H solar cell used in Simulations.
Parameter p-μc-Si:H b-μc-Si:H i-μc-Si:H n-a-Si:H
Layer thickness (nm) 10~30 5~10 1500~2500 10~20
Mobility gap (eV) 1.2 1.4 1.4 1.8
Donor doping density (cm-3) 3x1018
acceptor doping density (cm-3) 1x1019
Electron mobility(cm2/Vs) 100 32 32 20
Hole mobility(cm2/Vs) 25 8 8 4
Electron life time (μs) 0.01 0.01 0.01 0.01
Hole life time (μs) 0.1 0.1 0.1 0.1
Effective DOS in the valance and
conduction bands (cm-3) 2x1020 2x1020 2x1020 2x1020
Exponential tail Prefactors NTD, NTA (cm-3eV-1)
4x1021 4x1021 4x1021 4x1021
Characteristic energy WTD (VB tail) (eV) 0.04 0.045 0.05 0.05
distribution WGA (acceptor like state) (eV) 0.2 0.2 0.2 0.2
Peak of donor like Gaussian distribution
EGD (meas. From Valance edge) (eV) 0.4 0.7 0.7 1
Peak of acceptor like Gaussian distribution
EGA(meas. From Conduc. edge) (eV) 0.6 0.5 0.5 0.6
Correlation energy U (eV) 0.2 0.2 0.2 0.2
Transmittance of glass / ITO 0.9
Reflectivity of n layer / metal contact 0.9
Surface recombination velocity 1x107 1x107
a measured from the conduction band edge ; b measured from the valance band edge
32
The photogeneration model, which includes a ray tracing algorithm, is used to calculate the transmission and absorption of light in the bulk as well as reflection and refraction at the interfaces. The μc-Si:H solar cell is operated under the global standard solar spectrum (AM 1.5G) illumination with 100 mW/cm2 total incident power density and the light intensity for each wavelength is calculated in the wavelength ranger from 0.3 to 1.2 μm. Table 4-2 indicates electrical, optical and structural parameters used for simulating the device performance [26].
Fig. 4-7 Distribution of acceptor/donor-like trap states across the forbidden energy gap used in i-layer in the simulation study.
Fig. 4-7 is the general standard model for DOS of μc-Si:H i-layer. We assumed the Gaussian defect density in μc-Si:H i-layer is 4x1016 cm-3 which is higher than in a-Si:H i-layer (5x1015 cm-3), because the defects fill with the grain boundaries. Due to the μc-Si:H material, it makes lower band gap, sharper band tails and higher Gaussian defect density. Simulation
33
results indicate a higher Jsc but lower Voc in higher crystallinity material as shown in Fig. 4-8.
The short current ,open voltage, fill factor, and the efficiency of a-Si:H solar cell are 21.81 mA/cm2, 0.47V, 0.73, and 7.49 %, respectively. As we have stated, the electrical characteristics are differ from the a-Si:H solar cell in Table 4-3. The crucial point to reach solar grade quality µc-Si:H is to have good quality grain boundaries containing a low amount of unstable defects. Indeed, low quality amorphous silicon grain boundaries or cracks along grain boundaries which can be observed in highly-crystalline µc-Si:H.
Fig. 4-8 The illuminated J-V characteristic curve of μc-Si:H p-i-n solar cell
Table 4-3 The electrical characteristics for a-Si:H and μc-Si:H solar cells.
Description Jsc (mA/cm2) Voc (V) FF Efficiency (%)
a-Si:H solar cell 13.3 0.83 0.74 8.37
μc-Si:H solar cell 21.81 0.47 0.73 7.49
34
4.3 Tandem Solar Cell with and without x Layer
The Tandem solar cell is constituted by the optical and electrical series connection of an amorphous silicon (a-Si:H) top cell and a microcrystalline silicon (µc-Si:H) bottom cell. As a consequence of the electrical series connection, the short-circuit current density Jsc of the whole tandem is limited by the absorber (top or bottom cell) with the lower current generation capabilities. The thickness of the a-Si:H intrinsic layer must be made as thin as possible to minimize the Staebler-Wronski effect. It is generally thin (about 250 nm) in order to collect a maximum of the photo-induced electrons. The µc-Si:H intrinsic layer has to be thicker (about 2 µm) because of its indirect band-gap and the necessity to match the photo-current generated by the two stacked cells.
Fig. 4-9 Tandem solar cell without x layer.
Tandem solar cell with the following structure as shown in Fig. 4-9 was used in the simulations: flat glass / ITO / p-a-SiC:H / i-a-Si:H(di,top) / n-a-Si:H / p-µc-Si:H /
35
i-µc-Si:H(di,bot) / n-a-Si:H / Ag. The Ag forms the BR. The absorber layers were relatively thin (di,top 200 nm and di,bot 2.2µm), and no interlayer was applied. The calibrated optical and electrical parameters of undoped, doped a-Si:H and µc-Si:H layers were used in the simulations.
4.3.1 Simulation Model
Since the tandem cell was fabricated under the same conditions as the single-junction cells we used this set of parameters for modeling of the tandem cells. Also, The physical models we used here are identical with a-Si:H and µc-Si:H solar cell as mentioned before except thickness.
Fig. 4-10 Tandem solar cell with x layer.
The physics controlling the electric transport in n-a-Si:H and p-µc-Si:H interface is generally called tunneling recombination junction (TRJ) as shown in Fig. 4-9 which is
36
explored with atlas. When modeling the tandem cell as a complete stacked structure (pinpin device), and not using a x layer for the TRJ between the two-component cells, we could not obtain an realistic J-V curve for illuminated tandem cell unless adding a strong-recombination layer (which we call an x-layer) sandwiched between the n- and p-layers of the two inter-cell contact regions as shown in Fig. 4-10. We also have found that the parameters of the TRJ, such as doping concentration and the defect density in the doped layers of the TRJ, must be optimized in order to get the realistic illuminated characteristics of the tandem cell. The parameters for models, such as the mobility gap and the defect density of the x-layer, were also sensitive for obtaining the realistic tandem cell characteristics. After determining the parameters of the TRJ models, we obtained an excellent J-V curve of tandem cell with added-x-layer models. A typical J-V curves for both cases (with and without x layer) are shown in Fig. 4-11.
Fig. 4-11 The J-V curves of tandem solar cells with and without x layer.
37
The parameters which were used to describe TRJ are reported in Table 4-4 and Fig. 4-12 shows the energy band diagram of the tandem cell in thermal equilibrium for insertion of special layers (x-layer) in the tunnel/recombination junction on tandem cells. Fig. 4-13 shows (a) the distribution of hole current density, (b) electron current density and (c) the recombination rate in the cell. In the cell carrier's current density increases as the carriers move toward the "tunnel junctions". But at "tunnel junction" the hole current is seen to drop very low in the n region, while correspondingly the electron current is seen to drop very low in the p region. This indicates a very strong recombination process is happening in the "tunnel junction" which manifests itself in the Fig. 3c, the plot of recombination rate in the contacts.
This good recombination in contacts is needed for continuity of currents. Careful examination of Fig. 4-13(b) shows that there actually is a small component of electron current in the second sub-cell moving toward the first "tunnel junction", which is due to the unbalance of the net photo-carrier generation. Hence, the modeling shows the delicate balancing going on at these contacts [28]. The crucial role of recombination in the "tunnel junction" contacts of multi-junction solar cells is very different than the physics in the true tunnel junctions of tunneling diodes, in which electrons tunnel through the band gap from valence band to conduction band. Here electrons must fall from the conduction band to the valence band through a recombination process and fill in holes. Because the electrical field in the
"tunneling junction" is so strong due to our wanting to dope heavily to increase the field across the absorbers and because of this field's orientation (see Fig. 4-12), it acts against the holes and the electrons moving into the x-layers in contact region. Hence, our modeling shows that supplying carriers to this recombination can be a problem. It is in this supply role that tunneling is needed through the n-layer for electrons to supply the x-layer and through p-layer for holes to supply the x- layer. The key process in the functioning of the contact region is recombination. Any material layer that enhances this recombination will reduce the dipole modification and will enhance cell performance if it does not strongly absorb light.
38
Hence, the TRJ region cannot be represented with a resistor or diode model, and such a layer
Hence, the TRJ region cannot be represented with a resistor or diode model, and such a layer