Enhanced light absorption in
ultrathin-Si film using nanowire
arrays as light couplers
Master Thesis, 60 EC
Dominique Lucille van Poorten
Daily supervisor:
Nasim Tavakoli
Supervisor:
dr. Esther Alarc´
on-Llad´
o
Second reviewer:
dr. Elizabeth von Hauff
Solar energy is a sustainable energy source that is widely used on earth. With the increasing demand for energy, the search for high efficient solar cells continues. One promising tandem cell design is proposed by Tavakoli and Alarc´on-Llad´o, where a gallium arsenide (GaAs) nanowire (NW) array is not only used as a top cell but also as a way to couple light in a thin film silicon (Si) bottom cell. In this thesis have made the first steps in proving this theory. We have found a method to successfully transfer a NW array from one substrate to our sample. In addition, we showed a method to do photo-conductive measurements on thin-film Si. Lastly, we explored the possibility of using perovskites, namely; MAPbBr3 and MAPbI3, as a NW array top cell. We
calculated the short circuit current density (Jsc) in the perovskite and the Si bottom cell. We
found that MAPbI3 is preforming better than MAPbBr3. However, both perovskites NW arrays
1 Introduction 1
2 Background 3
2.1 The Maxwell equations in a single nanowire . . . 4
3 Simulations 6 3.1 Perovskite . . . 6
3.2 Lumerical . . . 6
3.3 Single nanowire . . . 7
3.3.1 Set-up . . . 7
3.3.2 Forward scattering and absorption cross section . . . 9
3.3.3 Results . . . 9 3.4 Nanowire array . . . 11 3.4.1 Set-up . . . 11 3.4.2 Method . . . 12 3.4.3 Results . . . 12 3.4.4 Coated NW . . . 15 3.5 Conclusion . . . 16
4 Towards a proof of concept 17 4.1 Photo-conductive measurements . . . 17
4.2 Fabrication process, theory . . . 20
4.2.1 Contact fabrication . . . 20
4.2.2 Wire bonding . . . 23
4.2.3 Nanowire transfer and placement . . . 23
4.3 Fabrication process, practice . . . 25
4.4 Measurements . . . 29
5 Conclusion 31
Bibliography 33
Appendix A Jsc data for the different layers 37
Appendix B Jsc as a function of filling fraction 39
Appendix C Trial Si-wafer sample 41 Appendix D PDMS recipe 44 Appendix E NW breaking 45
Introduction
The warming of the earth due to human actions and the emission of greenhouse gasses becomes an increasingly bigger problem. In order to keep up with the energy demand, fossil fuels are being burned whilst emitting greenhouse gas CO2. A solution to keep up with the energy demand with a
small carbon footprint, is using sustainable energy sources. One such a sustainable energy source is solar energy. Sunlight consists of different energy photons, whose total intensity is 982 W/m2 [1]. The spectral irradiance of the sun as a function of wavelength can be seen in figure 1.1 as
the black line. In semiconductor solar cells, each absorbed photon excites an electron from the valence to the conduction band, if these electrons are collected by an external circuit electricity is generated. There are some limitations to the amount of energy that can be generated, the two most common limitations are; first of all, only photons that have an energy higher than the band gap energy can excite an electron. Secondly, above band gap energy photons fall back to the band gap energy, the excess energy is lost and is called a thermalization loss.
This causes an efficiency limit called the Shockley-Queisser efficiency limit (S-Q limit)[2]. In the S-Q limit it is assumed that every photon above the band gap excites an electron, it can thus be called an ideal solar cell. The absorption spectrum in terms of spectral irradiance for two different semiconductors, namely gallium arsenide (GaAs) in red and silicon (Si) in blue is shown in figure 1.1. Here, we can see that the absorption spectrum of Si is below that of GaAs for the same wavelength. This is due to the earlier mentioned thermalization losses, which are higher if the difference between band gap energy and photon energy is bigger. Due to these losses the S-Q efficiency is decreasing with lower band gap energy.
Figure 1.1: The spectral radiance (black line) plotted as a function of wavelength. In addition, the spectral radiance that can be gathered from a perfect GaAs (red) and Si (blue) solar cell is plotted.
Currently, over 90% of the solar cell production is based on Si[3]. Si is an elemental
semicon-ductor with a band gap energy of 1.14 eV (λ = 1088 nm) [4] and is one of the best understood materials in electronics and thus photovoltaics (PV). Besides being one of the best understood materials in electronics Si is cheap and easy to make [3]. However, Si does not have the best
efficiency. An overview of the efficiency record holding materials as a function of their band gap is shown in figure 1.2. It can be seen that GaAs and Si are holding the two top efficiencies[5,6].
GaAs is an interesting but non-abundant PV-material and has band gap energy of 1.43 eV (λ = 867 nm)[7]. GaAs is a semiconductor compound, which is a combination of gallium from column III and arsenic from column V in the Periodic Table of Elements[8].
Figure 1.2: An overview of the efficiency record holding for different material solar cells plotted against their corresponding band gap. The black line represents the theoretical Shockley-Queisser efficiency limit. In addition, the grey lines show 75% and 50% of the limit. Figure taken from Ref. [9].
A promising way to go beyond the S-Q limit is stacking multiple PV materials in one solar cell, this is called a tandem cell. The idea behind the tandem cell is that the higher energy photons are absorbed in a semiconductor with a high band gap energy, the lower energy photons will continue to the next semiconductor with a lower band gap energy and so on. With this technique the thermalization losses in a solar cell can be minimized. However, there are some problems in making such a tandem cell. For example, not all materials are compatible with each other. As they have different lattice constants the problem of lattice mismatch is occurring in combinations between Si and any semiconductor of the III-V group [10]. In addition, some semiconducting materials are scarce and expensive, making them unsuitable for large scale solar cell production.
An interesting way to solve these problems is the use of nanowires (NWs). In this thesis we will therefore try to make an efficient and potentially cost-effective thin film solar cell made from an ultra-thin Si film with a perovskite or GaAs NW array on top.
Background
In order to create a solar call that is efficient, cheap and thin there are some problems to overcome. One promising design to go around these problems was proposed by Tavakoli and Alarc´on-Llad´o in 2019. The proposed design is a tandem cell consisting of a GaAs NW array on top of a Si ultrathin film of 2 µm[11], with silver (Ag) back reflector. It was found that, when a GaAs NW cell was
used on top of a thin Si, the NW array had a dual function; it functioned as a top cell, as well as a way to couple light below the band gap of GaAs to the waveguide of the Si bottom cell. This way, near infrared light that is normally not absorbed well in Si thin film, is now better absorbed[11].
The results are shown at the end of this section, but first we will shortly elaborate on the benefits of using NWs instead of a thin film of the same material.
A major benefit of using a GaAs NW array instead of a GaAs thin film is that, due to the small footprint of a NW, lattice mismatch is no longer a problem. This means that many material combinations are possible, even Si/III-V and thus Si/GaAs[12,13]. In addition, in earlier research
it has been found that a NW can absorb more light than their geometrical cross-section. In other words, the same amount of light can be absorbed while using less material [14,15,16]. Thus,
problems with using high-cost and non-abundant materials are solved, as less material is needed. Lastly, as mentioned earlier the NW arrays can be used to couple light into the waveguide of the Si bottom cell, this is due to diffraction pattern which is created by the periodic NW array[17].
In the work of Tavakoli and Alarc´on-Llad´o the NWs had a length of 6 µm and as a variable they changed the distance between the NWs, known as the pitch, and the diameter (d) of the NWs (figure 2.1 a). As mentioned, the idea behind the NW top cell is that the light can be guided from the NW array to the Si bottom cell, where the light is coupled to the waveguide of Si this is schematically shown in figure 2.1 b).
(a) (b)
Figure 2.1: a) Schematic figure of the GaAs/Si tandem solar cell with Ag back reflector. b) Schem-atic representation of the electric field in the waveguide of the NW array, being coupled into the waveguide of the Si bottom cell. Figure taken from Ref. [11].
They assumed that an ideal solar cell, e.g. there are only radiative losses. In addition, the design was chosen to be a four terminal tandem cell. This way, there are no restrictions in matching voltages or currents. However, four terminal tandem cells have a lower total efficiency to be lower, which is mostly due to some optical losses[18,19]. By calculating the current density in mA per
cm2 (J
sc) Tavakoli and Alarc´on-Llad´o showed that there is an absorption enhancement in the Si
bottom cell for the near infrared light, the calculation is shown in section 3.4.2. The results are shown in figure 2.2, here the Jsc) is shown as a function of NW diameter. The coloured lines
represent the different pitch distances and the dashed line shows the Jscof a 2µm Si-film without
NW on top, which from now on will be called the bare film. In conclusion, depending on the pitch and the diameter the absorption in Si could be enhanced by a factor 3.8 compared the bare film.
Figure 2.2: The Jsc in mA/cm2 if the Si bottom cell as a function of NW diameter in nm. The
different colours represent the pitch of the NW array, the bare film is shown as the dashed line. It can be seen that for pitch 800 nm the increase in absorption is the most and is a factor 3.8 higher compared to the bare film. Figure taken from Ref. [11].
2.1
The Maxwell equations in a single nanowire
In order to understand the physics behind the interaction of light with a single NW, we will briefly go into the Maxwell equations. Normally, when light propagates through a medium it can be seen as the light being coupled to an optical mode supported by that medium. However, if we decrease the size of the medium to the dimensions of the wavelength of light, there are fewer optical modes supported due to the boundaries of the medium. In the case of a NW there are only a few of modes where the light can couple to, these modes are called the eigenmodes of a cylinder and can only occur if the refractive index of the NW (ncyl) is higher than that of the surrounding medium
(nmed), thus ncyl> nmed [20]. At the interface of the medium and the NW the light can scatter,
the relation between the scattered fields and the amplitudes of the incoming field can be calculated using the Maxwell’s equations at the interface of the two media (the surrounding medium and the NW). The media are characterized by their dielectric permittivities med and N W and their
magnetic permeabilities µmed and µN W. To solve the Maxwell’s equations there are two boundary
conditions[21,22]:
1. At the interface the tangential electric (Ek) and the magnetic (Hk) field components are continuous, thus Ekmed= EkN W and Hkmed = HkN W
2. The components normal to the plane tangent to the interface of the electric field (D⊥= E⊥) displacement and the magnetic induction field (B⊥ = µH⊥) are conserved. Thus D⊥med = D⊥N W and B⊥med= B⊥N W
The eigenmodes can be calculated using the Maxwell’s equation in cylindrical coordinates and the Hankel and Bessel functions[23], the exact way to do this is beyond the scope of this thesis.
However, in order to be able to identify the type of modes in a single NW the electric field lines of lower order modes are shown in figure 2.3. Here the HE (magnetoelectric) mode is a mode where the magnetic field is dominant over the electric field. In the EH (electromagnetic) mode its the other way around[21]. In this thesis we focus on the EH and HE modes, however more modes can
be found; transverse magnetic (TM) and transverse electric (TE). The reason we do not focus on pure TM and TE modes is that we do not encounter them in the simulations done in chapter 3.
Figure 2.3: The electric field lines of the HE11, HE12, EH11 and EH12 mode in the cross section
of an infinitely long NW. Figure adapted from Ref. [21], therein reproduced from Ref. [24, 25, 26] From the electric field lines in figure 2.3 the type of mode can be found in a single NW, this will be used in chapter 3. Even though the electrical fields within a single NW will be different compared to that of a NW array, the single NW can give us an idea of interesting diameters for a NW array. These principles are shown in chapter 3, where we simulated a single perovskite NW and a perovskite NW array.
Simulations
In chapter 2 a new design for a thin film tandem cell was described, where the GaAs NW array top cell not only functioned as top cell but in addition was used for light trapping. In this chapter we will explore another material as top cell, namely perovskites. We will explore this with help of simulations. Based on the results of these simulation we can decide to explore these materials in experiments. First, we will briefly explain what a perovskite is. Secondly, the different types of simulations will be discussed with the corresponding results and conclusions.
3.1
Perovskite
As an alternative to a GaAs NW array, methylammonium (MA) lead halide perovskite NW array was explored. A methylammonium lead halide perovskite is a semiconductor with a specific crystal structure namely, the ABX3 formula. Where A is a halide, B an anode such as lead and lastly the
X3is MA (CH3NH3 the chemical formula for MA) as shown in figure 3.1[27].
Figure 3.1: Schematic representation of the crystal structure of a perovskite. In this figure A is a large cation, B a anion and X a small cation. Figure taken from Ref. [27].
We chose MAPbBr3 and MAPbI3 as they low-cost to make and can be made at AMOLF [28,29,30]. The refractive index of both perovskites is shown in figure 3.2.
3.2
Lumerical
Simulations in Lumerical were done in order to compare light-matter interaction in perovskite NW arrays with GaAs NW arrays. Lumerical solves the Maxwell’s equation using the Finite-Difference Time-Domain (FDTD) method[31,32,33]. Using this method, first a 6 µm long single NW was
simulated, from which the absorbed light in the NW and the scattered light from the NW was calculated. Second, a NW array with 6 µm long NWs on a 2 µm Si film with a 200 nm silver back reflector was simulated, where the absorption in the different materials was calculated.
The first simulation was done only on MAPbBr3surrounded in air (n = 1) and also in a polymer
NWs have to be transferred to the membrane while remaining in an array. Thus, embedding the NWs in a polymer is necessary, n = 1.5 was chosen as it resembles the refractive index of the polymer polydimethylsiloxane (PDMS)[34]. The second simulation is done on both perovskites, and are shown only for a surrounding n = 1.5, as this is a more realistic scenario for experiments. All simulations were done for a range of different sizes for the perovskite NW, diameters between 200 nm and 900 nm. This was done in order to understand the influence of the NW diameter on the scattered and absorbed light. In the sections below, the two types of simulations are explained and the result of each simulation is shown.
(a) (b)
Figure 3.2: The refractive index (n,k) of (a) MAPbBr3 and (b) MAPbI3, These graphs have been
produced based on Ref. [35, 36].
3.3
Single nanowire
Simulations in a single NW were done to calculate the absorption and the forward scattering cross section. These simulations were done using a Total-Field Scatter-Field (TFSF) light source. In the TFSF light source the incoming field and the scatter field are separated, enabling the absorption and forward scattering to be calculated. In this section the simulation is explained, followed by the results of the simulation.
3.3.1
Set-up
Figure 3.3 schematically shows the simulation setup. From outside to inside the following boxes can be found: 1) the FDTD simulation area, 2a and 2b) a box made out of DFTMonitors that measure the transmission power as a function of frequency in Hz, 3) the TFSF source and 4) the absorption box. The different boxes are explained below.
Figure 3.3: Schematic overview of the setup, where the different color boxes represent the simulation areas. 1) The total FDTD simulation area. 2) DFTMonitor that calculates the transmission per frequency in Hz, this consist of two part the upper part that calculates back scattering and the lower part that calculates forward scattering. 3) TFSF source and 4) a group of DFTMonitors that calculates the power that is absorbed. In the middle of the figure the NW can be found.
The first box (Box 1) represents the FDTD area, which is the complete simulation area. Within this box the Maxwell’s equations are solved at the nodes of small 3D blocks, which are called the mesh. For an as realistic possible simulation, these blocks should be as small as possible. However, the calculations are time-consuming because of which there is a trade-off: an as small as possible mesh for an accurate simulation or a simulation with lower run-time. One way to go around this trade-off is by choosing a mesh that is smaller in and around the NW, and increases in mesh size further away from the NW. Box 1 has an additional function which is the boundary conditions. In this simulations the boundary condition was chosen to be a Perfectly Matched Layer (PML), which means that all the light that leaves the simulation area is absorbed at the boundaries. Due to this complete light absorption, there will be no light reflected back into the simulation area.
As a light source, a TFSF-source is used (figure 3.3, Box 3). The TFSF-source is similar to a normal plane wave, except that at the edges of this box the incoming plane wave is ”subtracted”. This means that if there is no object that can scatter the plane wave within the box, there will be no plane-wave measured outside of the TFSF box. However, if there is scattering the scattered light will pass the edge of the TFSF box and can thus be measured using a box of FDTMonitors. These monitors then measure the total scattering from the NW. In our case we are interested in separating the scattered light from the bottom of the NW which is the forward scattered light and the backward scattering. This can be measured using two boxes, where one is open at the top and one at the bottom. The forward scattering can be measured with a box of FDTMonitors where the top is open (Box 2b). In Box 2a the backward scattered light is measured using a box of FDTMonitors where the bottom is open, thus measuring all the light scattered from the sides and the top of the NW. This way, we can separate the light that is scattered from the NW to the top and bottom hemisphere around the wire.
The last box is Box 4, this is a box made out of FDTMonitors and measures the power output and input of the box. The difference between the two is the absorbed power and thus the absorbed light.
3.3.2
Forward scattering and absorption cross section
The forward scattering cross section can be calculated with the transmission measured in box 2 and source power and intensity from box 3. Lumerical has the function ”transmission” that provides the transmission (T) divided by the source power (Psource), both as a function of frequency (f) in
Hz. However, the ideal plane wave has infinite power while the transmissions power is finite. In order to overcome this problem the cross section can be calculated by dividing the transmission by the source intensity (Isource):
σ = T (f ) ∗ Psource(f ) Isource(f )
(3.1) This calculation is done for box 2a and 2b, where 2a gives the backward scattering cross section σbsand 2b the σfs. lower and upper part of box 2, where the lower part gives the forward scattering
and the upper part the back scattering. The σ values for the different wavelengths per NW diameter were retrieved from Lumerical and saved as a .txt file.
3.3.3
Results
From the data obtained in the previous section, contour plots were made using Wolfram Mathem-atica (version 11.3). In figure 3.4 the results are shown, on the left for a single NW in medium n = 1 and on the right for n = 1.5, the top shows the σfs and at the bottom the absorption is shown.
The color scale represents the area in µm2 that is being absorbed/scattered by the single NW.
Figure 3.4: On the left (right) contour plots for surrounding medium n = 1 (n = 1.5). At the top (bottom) the effective forwards scattering (absorption) in µm2is shown, as a function of wavelength
and NW diameter.
In figure 3.4 the bottom graphs show the effective absorption. It can be seen that light is strongly absorbed for above energy band gap photons. For diameters between 300 nm and 800 nm with surrounding n = 1 and n = 1.5 the absorption cross-section is higher than the geometrical
cross-section of the single NW. This phenomenon has been found in earlier research as well[15,16]
and is also visible in the two top graphs where light is scattered in a bigger area than the geometrical cross-section of the single NW. The pronounced lines in two top graphs of figure 3.4 correspond to the modes in the single NW, meaning that light from these wavelengths can easily couple to the NW[21]. It can be seen that for a bigger difference between the refractive index of the medium
and that of the NW, the modes are slightly blue shifted[37]. In order to see which modes can be
seen in figure 3.4 the electric field intensity (left) and the electrical field lines (right) are shown for surrounding n = 1.5 in figure 3.5. The electric field lines correspond to three different modes as shown in chapter 2 (figure 2.3), namely: HE11, HE12 and EH11.
(a) d = 390 nm, λ = 1080 nm. The electric field lines (right) show a HE11
mode.
(b) d = 590 nm, λ = 561 nm. The electric field lines (right) show a HE12
mode.
(c) d = 590 nm, λ = 566.5 nm. The electric field lines (right) show a EH11
mode.
Figure 3.5: On the left the | ~E2| and on the right the electric field lines of a single NW for different
3.4
Nanowire array
In this section, a NW array with pitch 1000 nm, height 6 µm and diameter ranging from 200 nm to 900 nm is simulated on top of a 2 µm Si film, with a 200 nm silver back panel underneath and a surrounding medium of n = 1.5. The pitch was chosen to be a 1000 nm due to the results in a GaAs NW array where for a bigger pitch distance the absorption enhancement in Si increased the most, due to introducing a higher number of diffracted orders into Si[11] This was done for MAPbBr
3
and MAPbI3, from the simulations the absorption was measured and the Jscwas calculated in the
different materials.
3.4.1
Set-up
For this simulation again the FDTD method of Lumerical was used. Within the FDTD simulation area the different material layers were made, this is schematically shown in figure 3.6. There are four DFTMonitors that measure the output power and a plane wave light source that emits light of wavelength 400 nm ≤ λ ≤ 1100 nm. These wavelengths are chosen because the refractive index in this range is known for both perovskites (see figure 3.2). Before explaining the different monitors that calculate power at specific points, it has to be noted that this type of simulation, namely one that simulates a NW array, are rather difficult. In order to reduce the time needed per simulation, the symmetry of the NW array can be used by choosing the boundary conditions to be symmetric and anti-symmetric. The DFTMonitors are placed such that the absorption can be measured in-between each material, above the light source a monitor that measures the reflection is placed. The power of the incoming light is measured by the ”Top” monitor and is placed below the light source. Below the NW the ”Middle” monitor is placed, this monitor measures the power of the light after the NW. Lastly, the power of the light just before exiting the Si film is measured by the ”Bottom” monitor.
Figure 3.6: Schematic overview of the setup, where we simulated a 6 µm long NW on top of 2 µm thick Si with a 200 nm Ag back reflector. The purple arrow indicates the direction of the plane-wave light source. The power of the light is measured at four different places, this is shown in the figure as Reflection, Top, Middle and Bottom.
3.4.2
Method
From the monitors the absorption in the total absorption (AbsT), the absorption in Si (AbsSi) and
the NW (AbsN W) per wavelength can be calculated by Lumerical using the following formulas:
AbsN W = T op − M iddle
AbsSi= M iddle − Bottom
AbsT = T op − Bottom
This absorption data is given as output from Lumerical and is saved in a .txt file. Using the absorption per wavelength [abs(λ)], the AM1.5G solar spectrum [Nph(λ)] and the elementary
charge (e), the photo-generated current (Jph) can be calculated with the following equation[11]:
Jph= e
Z 1100nm
400nm
Nph(λ)abs(λ)dλ (3.2)
Here, we integrate from 400 nm up to 1100 nm to calculate the Jsc of the complete cell as we
have the refractive index values for the two perovskites in this range. It has to be noted sunlight has light with a higher wavelength, the absorption can thus be higher than we calculate in this range. If we neglect the series or shunt resistance losses, the Jphand the Jscare the same. In this
thesis we will be using the Jsc as a figure of merit.
3.4.3
Results
Using equation 3.2, the Jsc in the top and bottom cell was calculated for every NW diameter.
In order to see if there is a difference between the absorption in Si with and without a NW, the Jsc of a bare Si film was also calculated and compared to the Jsc in Si with a NW array on top.
In addition, we compared the absorption in the perovskite and Si for both perovskite NW arrays as top cell and a 6 µm perovskite film as top cell. This was done to see what the influence of a NW array is in comparison to their thin film counterparts. The perovskites have a different band gap, this means that low band gap wavelength material MAPbBr3 will transfer more light
to the Si bottom cell than MAPbI3, making a fair comparison between the performance of the two
perovskite difficult. Therefore, it was chosen to calculate the Jscfor both perovskites in the bottom
cell (Si) for a wavelength from λ = 800 nm, which is in the near infrared and where it is predicted that a NW array top cell can enhance the absorption in Si as was seen in earlier simulations with GaAs NW arrays[11]. The J
scin the bottom cell for both perovskites can then be compared with
each other in the same range.
The results are shown in figure 3.7 for MAPbBr3 (green) and MAPbI3(red). Both plots show
the Jsc for bare Si (black dotted line) for 800 nm < λ < 1100 nm, which is equal to 1.8 mA/cm2.
The coloured dotted line represents the Jscin Si with a MAPbBr3(green) and MAPbI3(red) film
as top cell. It can be seen in figure 3.7 that there is no difference between the Jscin bare Si, and Si
with a MAPbBr3film while in the absorption in Si is higher if a film of MAPbI3is placed on top of
the Si, as less light is reflected. However, the biggest difference in Jsccan be seen for the MAPbI3
NW array, where at d = 750 nm the Jsc is almost double. In addition, we can see multiple peaks
in Jsc for both perovskites, these peaks indicate that for the corresponding diameter the light is
effectively coupled to the bottom cell. Thus, we see that the diameter is of great importance for how well the light of the top cell is coupled to the waveguide of the bottom cell.
The performance difference between the two perovskites mostly depends on the refractive index of the perovskites after their band gap, after which the light is transferred to the Si. What happens in this region is that the guided mode is more confined in the case of MAPbI3as the refractive index
difference is bigger. Due to this confinement the light within the wire is transferred to the Si where a diffraction pattern is being created due to the NW pattern which acts as a grating. However, if the light is less confined, which is the case for MAPbBr3due to the lower refractive index after the
band gap, part of the electric field is outside of the NW. This field exponentially decays as the field propagates along the cylinder, there is thus a loss of energy. The modes for which this happens
are called ”leaky modes” as the energy leaks out of the NW[20,38]. The energy that leaks out is
not confined in the NW, because of which it is not coupled into the waveguide of the Si, e.g. the grating becomes less effective. If the difference between the NW and its surrounding is bigger, the light is more confined thus there are less leaky modes. Hence, the MAPbI3 preforms better than
the MAPbBr3 NW array.
Figure 3.7: Jsc in Si for in mA/cm2 for 800 nm < λ < 1100 nm. The dotted black line represents
the Jsc in bare Si and is calculated to be 1.8 mA/cm2, in green the Jsc of MAPbBr3 is shown and
in red the Jsc of MAPbI3, in this graph the dashed line is for a film of the perovskite while the solid
line is the NW array with correspond diameters on the x-ax.
However, to conclude if one of these perovskites is interesting as a top cell the Jsc of the whole
four-terminal cell is calculated and shown in figure 3.8 for wavelength 400 nm < λ < 1100 nm. It can be seen that for a MAPbBr3 NW array, a MAPbBr3 film or bare Si there is barely any
difference in Jsc of the whole four-terminal cell. In the case of MAPbI3there is a almost a factor
1.5 increase in absorption if a MAPbI3 film or MAPbI3 NW array is placed on Si compared the
bare Si. However, the absorption difference between a MAPbI3 NW array or a MAPbI3 film is
barely visible.
Interestingly, it can be seen that the strong peaks in Jsc seen in figure 3.7 are not visible in
figure 3.8. This indicates that light above the band gap energy of the perovskites is coupled to and absorbed by the Si bottom cell, which is thus not visible in figure 3.7. In addition, for MAPbBr3
we can see that for smaller diameter NW the film is actually preforming better than the NW array. This indicates that there is more reflection in the lower wavelength regime for smaller diameter. The data that supports this claim is shown in Appendix A.
Figure 3.8: Jsc in the top and bottom cell for in mA/cm2 for 400 nm < λ < 1100 nm. The dotted
black line represents the Jsc in bare Si and is calculated to be 17.9 mA/cm2, the dotted grey line
is a film of a) MAPbBr3 and b) MAPbI3, and the purple line represents Jsc for different the NW
diameter shown on the x-axis.
Another way to present these results is by calculating the filling fraction of the NW. We calculated this to check if there was a correlation between NW filling fraction and Jsc, however we
did not find such a correlation. The results are shown in Appendix B.
The previously shown figures show the Jsc in the bottom or complete cell as a function of NW
diameter. However, to gain insight in why there is more light is absorbed in the bottom cell, it is interesting to look at the absorption spectrum of the bottom cell. Therefore, we calculated the absorption spectrum of a MAPbI3 NW array with pitch 1000 nm and diameter 500 nm which is
shown in figure 3.9. Here, the absorption in the Si bottom cell is shown as a function of wavelength. In yellow the absorption spectrum of bare Si is shown, where we can clearly see the yellow line oscillating, this is the Fabry-P´erot resonance. Fabry-P´erot resonances arise as followed: when the light enters the thin film Si, part of the light is reflected, while the other part travels through the Si. However, at the bottom of the Si, again part of the light is transmitted while another part is reflected. Due to this reflection and transmission, a part of the light is remaining in the Si, which causes a constructive or destructive interference with itself, which is visible in the absorption plot as oscillations[39,40]. The absorption spectrum in Si with a MAPbI
3top cell (d = 500 nm) is shown
in blue. It can be seen that for above MAPbI3 band gap energy photons there is some absorption
in the Si. However, the absorption increases for below MAPbI3 band gap energy photons, which
is expected as these photons do not have enough energy to excite an electron from the valence to the co-valence band in the MAPbI3. In addition, it can be seen that around the Fabry-P´erot
resonance there are sharp absorption peaks in the blue line. This increase in absorption around the Fabry-P´erot is due to the light trapping caused by the NW on top of the thin film Si.
Figure 3.9: Absorption spectrum in the Si bottom cell. In yellow for bare Si and in blue in the Si with a MAPBI3 NW array with pitch 1000 nm and diameter 500 nm.
3.4.4
Coated NW
In addition to the NW array as described in section 3.4.1, we simulated a MAPbI3 NW array for
d = 600 nm and with a 10 nm (n = 4) coated layer on the NW. The idea behind this is that the light in the coated NW is more confined to the NW, as explained in section 3.4.3. We choose the MAPbI3 NW array as the factor 1.5 increase in light absorption seemed more promising than the
MAPbBr3 NW array. It has to be noted, that there is no suitable material (e.g. with negligible
absorption) that has n = 4. Suitable materials are commonly used as an anti-reflective coating and have a low refractive index, an example is Silicon Nitride (SiNx) with a refractive index n ≈
2.1[41,17]. Due to their low refractive index, these materials are not suitable for our purposes. However, to get an idea of the influence of a high refractive index material as a coating for a NW array we decided to do the simulation regardless of the real life possibilities.
Using the same method as described in the previous sections, we calculated the Jscfor Si with
λ > 800 nm and the whole tandem cell in the range 400 nm < λ < 1100 nm. These results are shown in table 3.1 and compared with the Jsc of the bare MAPbI3 NW array and bare Si.
JSi, λ>800nm
sc (mA/cm2) Total Jsc (mA/cm2)
Bare Si 1.7
Bare NW array 2.4 27.7 Coated NW array 2.8 27.8
Table 3.1: Calculated Jsc for bare Si, a MAPbI3 NW array at d = 600 nm and for the same NW
array but then coated with a 10 nm layer of a material of n = 4.
The results show us that the bare Si has the does not absorb light well compared to Si with a NW array on top, as was already shown in section 3.4.3. However, there is an increase of 0.4 mA/cm2in the Si with a coated NW array on top compared to the bare NW array. Interestingly, the absorption of the whole tandem cell does not change a lot, as the coated NW array on top of Si only has an increase 0.1 mA/cm2. We think that this is due to a better light coupling of the light to the outer side of the coated NW, from which the light is better transferred to the Si but
not absorbed in the MAPbI3. In order to check this hypothesis we plotted the intensity of the
electric field in the middle of on of the NWs in the NW array as shown in figure 3.10.
Figure 3.10: In this plot the intensity of the electric field (| ~E2|) in the unit cell is shown for λ = 906 nm, on the left the bare and on the right the coated MAPbI3 NW array. The white circle
represents a NW with d = 600 nm.
The | ~E2| for the coated NW as plotted in figure 3.10 (right) is not only strongly present within the NW but also on the right and left side of the NW. This is in agreement with our hypothesis that the light confined to outer sides of the NW. However, this only happens at the left and right side of the NW and not around the whole NW. This is due to the fact that the intensity of the electric field is strongly dependent on the polarization of light[42,43]. It has to be noted that by
coating the NW we influence the mode, thus it can explain why the intensity of the electric field is different for the same wavelength. Based on this simulation we conclude that coating the NW array with a 10 nm layer of a material with n = 4 does not increase the absorption of the whole cell and therefore we should not pursue making this type of perovskite NW array tandem cell.
3.5
Conclusion
We found that for the single MAPbBr3 NW the refractive index difference between the NW and
its surrounding plays a role in the absorption: if the difference is low the absorption peaks are blue shifted. This gives an idea about the materials that are suitable for a NW tandem cell, namely NW with a high refractive index compared to their surrounding. This is also visible in the simulated NW array, for MAPbI3with n = 2.2 after the band gap the absorption is higher than for MAPbBr3
which has n = 2 after the band gap. Even though a tandem cell with a MAPbI3 NW top cell
absorbs a factor 1.5 more light than bare Si, it does not absorb much more than using a MAPbI3
film as top cell.
Using a coating around the MAPbI3 NW array for d = 600 nm as a way to create the bigger
difference in refractive index did show an increase of 0.4 mA/cm2 in the Si. However, as the
absorption in the whole cell only changed by 0.1 mA/cm2 coating the NW array was not found
to be as useful. We think that the coating caused better light transfer to the Si, where this light is above the band gap of MAPbI3. Thus, more light that could be absorbed by the MAPbI3 was
transferred and absorbed in the Si.
Comparing the most promising perovskite simulated in this, the MAPbI3 NW array, that has
at most an absorption increase of factor 1.5 to the GaAs NW simulated by Tavakoli and Alarc´ on-Llad´o (2019), that has a absorption increase of almost factor 4, we can conclude that the GaAs NW arrays are much more suitable as a top cell for the suggested tandem cell design[11].
Towards a proof of concept
As explained in Chapter 2, the GaAs NW array top cell acts a a waveguide to couple light into the Si thin-film bottom cell to increase the photo-current in ultrathin Si-based solar cells. However, this is currently shown in simulations and not yet in experiments. This chapter explains the method we use in order to experimentally prove the enhanced light absorption. First, the type of measurement will be explained. Secondly, the fabrication of the sample is step by step explained. Lastly, the measurement itself is shown together with the result.
4.1
Photo-conductive measurements
In order to prove that there is more light absorption in a Si thin-film if there is a GaAs NW array on top, photo-conductive (IV) measurements will be done on a Si membrane with and without a GaAs NW array that is optically coupled to the Si. Although the two layers are not electrically connected we can still measure the change in photo-resistivity in Si. The idea behind this measurement is that if there is more light absorption in the Si, there are more excited electrons thus more free carriers and thus less photo-resistance. This can be measured using the van der Pauw method, which is a four-terminal measuring method and can be used on conducting sample of arbitrary shape, as long as: (1) the contacts are at the circumference of the sample, (2) the contacts are sufficiently small, (3) the sample is uniformly thick, (4) the sample does not contain any isolated holes, and (5) the contacts are Ohmic[44,45]. The Si thin-film that we have is a p-type <100>
Si membrane from Norcada, and satisfies these constraints, the contacts we can make ourselves such that it satisfies the constraints. Given figure 4.1a, in the van der Pauw method a current is applied in point 1, from which the current flows through 3 and 4 to the grounded point 2 (I12).
The voltage drop is measured in parallel between point 3 and 4 (V34= V3− V4)[44,45,46]. This
is schematically shown in figure 4.1a, from the measured I12 and V34 the resistance R12,34 can be
calculated with the following formula:
R12,34=
V34
I12
Different configurations for R, such as R23,14, are defined similarly. Using Ohms law it follows
that the resistance is lower if there are more free carriers in the system, which is the case for a measurement in light compared to a measurement in the dark (figure 4.1b). Furthermore, a similar change in resistance is expected for Si thin-film with GaAs NW array on top, compared to bare Si.
(a) (b)
Figure 4.1: (a) Electrical circuit for a van der Pauw measurement, where a current is applied in 1, 2 is grounded and the voltage drop is measured between 3 and 4. (b) Expected IV-curve for the measured resistance in light and dark. During a measurement in light, the system has more free carriers, resulting in a higher current for the same voltage. Figure adapted from Ref. [47].
The correlation between the resistance and the resistivity averaged over the sample thickness, more commonly known as the sheet resistance (Rsh), is given by[44,45]:
Rsh=
π ln(2)
R12,34+ R23,14
2 F (4.1)
Here, F is the correction factor and depends on the ratio Rr= R12,34 R23,14 as shown in equation 4.2. Rr− 1 Rr+ 1 = F ln(2)arcosh( exp[ln(2)/F ] 2 ) (4.2) Using equation 4.1 and 4.2, the sheet resistance can be calculated. The sheet resistance does not immediately give the absorption of the Si bottom cell. Fortunately, we can derive the absorption from the sheet resistance using the following relation:
Rsh=
ρ t = 1
σ ∗ t
Where ρ is the resisitivity, t the thickness of the membrane and σ the conductivity. As the conductivity is proportional to the photocurrent and the photocurrent is proportional to the ab-sorption, we can thus derive the absorption from Rsh [48,49,50]:
Rsh∼
1
absorption (4.3) In order to do the IV-measurements a suitable contact design is needed. In the design for the contact pattern we have to take into account the size of the NW array, which is 300 µm × 300 µm. Therefore, the area where we measure the IV should not be bigger than the NW array. In addition, we work on a membrane that is rather fragile. The membrane used in this thesis is a 7.5 mm × 7.5 mm frame with a membrane of 3.5 mm × 3.5 mm in the middle. Rather than placing probes on the membrane we had to place them on the frame. Taking these limitations into account we came up with the contact pattern in figure 4.2. Fabricating this pattern has two different steps, the first step is to insulate the membrane with SiO2except for the measuring areas, and the second step is
making the contacts. The first step is shown in figure 4.2 1A. Because the contacts are brought to the frame of the membrane, there is a large area where the contacts and the silicon are in contact with each other. As it is unwanted to we measure current outside of the measurement area, we
decided to sputter an insulating oxide on the Si. We choose a 40 nm layer of SiO2as insulator, as
it shows insulating properties at small thickness[51,52]. As a result, the SiO2 has to be etched
down from the measurement area, creating a new fabrication step (see section 4.2). Figure 4.2 1B shows the contacts. In total there are 3 measurement areas and have a size of 50 µm × 200 µm, 100 µm × 200 µm and 1500 µm × 200 µm. At the corners of these areas a 10 µm × 10 µm contact point is made (figure 4.3a), the points are brought to the frame of the membrane by a 1800 µm × 300 µm arm, that ends with a 400 µm × 400 µm contact pad where the probes can be attached to (figure 4.3b).
Figure 4.2: Schematic overview of the contact design for the van der Pauw measurement, on the left the etch pattern is shown and on the right the contacts are shown.
(a) (b)
Figure 4.3: On the left: close up of one of the holes in the SiO2 with the contact points in the
corner. On the right: Close up of one of the contact arms that brings the contact point to the frame.
This design was sent to DeltaMask, which is a company that creates masks for photo-lithography. They made the design in chrome on a 5 inch quartz wafer. With the mask the Ohmic contacts could be made, it was chosen to use a 150 nm gold (Au) layer as Ohmic contact with a 5 nm Chromium (Cr) layer underneath as an adhesion layer[53].
4.2
Fabrication process, theory
In this section the fabrication process on the membrane will be explained. As the membrane did not have any contacts on it, these needed to be made at AMOLF. The process to make the contacts is shown in section 4.2.1. However, the contacts are rather delicate and can be easily scratched during probe measurement. Using wire bonding insured us that the contacts would not be damaged during measurements, this method is explained in subsection 4.2.2. In order to do measurements with the NW on the membrane, the NW needed to be transferred from the GaAs wafer where they were grown on, to the membrane. This transfer method is explained in subsection 4.2.3.
4.2.1
Contact fabrication
With photo-lithography a thin layer of photoresist is applied to the sample, which reacts to a specific wavelength. In this fabrication process a 1 µm thick negative photoresist layer is applied which, after baking, becomes harder. When the photoresist is exposed to light with λ = 365 nm, it reacts such that the exposed parts of the resist remain on the sample after washing with the chemical Ma-D 533s. This creates the desired pattern of indents. A schematic overview can be found in figure 4.4. The places where the photoresist is present, are protected from etching or from the metals that are evaporated on the sample in order to create the contact (see figure 4.5). The fabrication recipe is a standard recipe, depending on the thickness of the resist the power of the source (dose) should be changed. This method has been used many times before at AMOLF, as the dose and the thickness of the resist match up. The exact steps are shown below.
(a) (b)
(c) (d)
Figure 4.4: A schematic overview of the photo-lithography process. (a) The sample is cleaned, (b) after cleaning a negative photoresist layer is spincoated on the sample. Followed by UV exposure, the black part corresponds to the chromium part of the mask. The negative photoresist reacts with the UV light (λ = 365 nm) and becomes hard (c). The soft parts can then be washed away during the development (d). The desired pattern is now made in the negative photoresist.
SiO
2and sample cleaning
The membranes from Norcada are already clean, therefore we could immediately evaporate SiO2
on the sample using the E-flex. The E-flex process is described in subsection ”E-flex”. After the evaporation of 40 nm SiO2 the sample was cleaned for 12 seconds using oxygen plasma in the
Spincoating
The sample was cooled with an N2gun and transferred to the spincoater. In order to have a better
bonding between the photo-resist and the sample, a few drops HDMS was applied until the sample was covered in HDMS and following recipe was run.
Step Rotation [rpm] Acceleration [rpm/s] Time [s] Lid open (0), lid closed (1) Function 1 0 20 5 1 Close lid 2 40000 1000 45 1 Spin HDMS 3 0 1500 0 0 Stop spin and open
lid
After applying the HDMS, the sample was placed on the hotplate for 1 minute on 150 ◦C. In order to spincoat a 1 µm layer of MaN-1420 the following spincoat-recipe was used.
Step Rotation [rpm] Acceleration [rpm/s] Time [s] Lid open (0), lid closed (1) Function 1 0 20 5 1 Close lid 2 2500 500 45 1 Spin MaN-1420 3 0 1500 0 0 Stop spin and open
lid
After spincoating the photoresist, the sample was baked for 90 seconds on 150 ◦C. The sample now has a 1 µm thick layer of negative photoresist, as schematically shown in figure 4.4b.
This spincoat recipe should create a photoresist thickness of 1 µm. However, if the container of the photoresist is opened frequently the solvents in the photoresist can evaporate. Which can make the photoresist sturdier, causing a thicker layer. Therefore, the spincoating has to be done on a dummy sample and the thickness should be measured before before the membrane is spincoated. With a thicker layer of photoresist the dose for the UV exposure changes.
UV exposure
For the UV exposure (figure 4.4c) the MA/BA6 high-precision mask and bond aligner from SUSS MicroTech was used. We used the soft contact program with the below described parameters.
Parameter Value nominal dose [mW] 25.0 dose for MaN 1µm [mJ] 450.0 exposure time [s] 18 Alignment gap 100 µm WEC type [s] 18 WEC-offset [s] 18
Development
After exposing, the sample has to be developed (figure 4.4d). During development the photo-resist that has become soft due to exposure is washed away, which creates a pattern that can be etched or sputtered with chromium (Cr) and gold (Au) to create the contact.
For the development the sample will be soaked in Ma-D 533s at 21◦C. Followed by H2O for
Chemical Time [s] Ma-D 533s at 21◦C 120 H2O 15
H2O 15
After development the sample is ready to be etched or provided of a metal contact. After etching the steps described from ”Spincoating” to ”Development” have to be done again.
Etching
The etching was done with the Oxford PlasmaLab 80+ using a mixture of fluoroform (CHF3) and
argon (Ar) gas (figure 4.5a). To etch 30 to 40 nm of SiO2, the recipe needs to be run for 1 minute.
The photoresist protects the areas where etching is unwanted, thus creating the desired pattern. After etching the sample has to be thoroughly cleaned, as there is some left over SiO2 particles
that are not wanted in the process.
After the etching the sample was cleaned in base Piranha for 15 minutes. This was done two times, to make sure the photo resist residues are completely gone.
(a) (b)
(c) (d)
Figure 4.5: Schematic overview of the etching (a, b) or evaporation process (c, d) that can be done after creating the desired pattern in the photoresist layer. The photoresist protects the area where no etching/evaporation should occur. In (a) the SiO2 is etched from the unprotected area. After
etching the sample the left over negative photoresist is cleaned from the sample (b). In (c) a gold (Au) layer is evaporated on the sample and the photoresist. During the lift-off process in (d) the photoresist dissolves with the Au on top of it, only the Au evaporated on the sample remains.
E-Flex
Chromium (5 nm) and gold (150 nm) where deposited using Electron Beam Physical Vapor De-position (EBPVD). This is a technique where in a vacuum chamber an electron beam is shot on a material, in this case Cr and Au. Due to the electron beam the metal becomes hot and changes from the solid phase to the gas phase. The gas floats to the top of the chamber where the sample is stationed, this way the sample is coated with the desired metal. In the NanoCenter at AMOLF the Polyteknik Flextura M508 E was used for EBPVD. A schematic overview of the metal layer on top of the resist is shown in figure (figure 4.5c).
After evaporation, the remaining photoresist with Cr and Au was dissolved in acetone (figure 4.5d). This process, called lift-off, took around 60 minutes. As acetone leaves traces on the sample, the sample was cleaned with IPA and blow dried with an N2 gun.
4.2.2
Wire bonding
In order to do the measurements without the possibility of breaking the membrane, we used wire bonding from the contact patches on the membrane to a sample holder. A schematic representation of the sample holder with sample is shown in figure 4.6. On the sample holder we can easily attach cables to preform our measurements with.
Figure 4.6: Schematic representation of the sample holder with the sample. The dimensions are in mm. From the contact patches of the sample we can wire bond to the contact points of the sample holder. From the contact points of the sample holder, wires go to the sides from where we can easily connect a cable. The four biog holes on the corners are to mount the sample holder to a set-up. Figure courtesy of mr. Dion Ursem, AMOLF.
4.2.3
Nanowire transfer and placement
For this project our group, the 3D photovoltaics group at AMOLF has a collaboration project with ´Ecole Polytechnique F´ed´erale de Lausanne (EPFL) in Switzerland. At EPFL the NW arrays are grown on a GaAs wafer. Using e-beam lithography holes are made into the wafer with a SiO2
mask. After making the holes the wires are grown using the liquid-solid-vapor technique[54].
The NWs have to be transferred from the GaAs wafer to the membrane, this is done by first spincoating the NWs (figure 4.7a). After spincoating the NWs can be peeled off (figure 4.7b)[55]
(a) (b)
(c)
Figure 4.7: (a) A thin layer of PDMS is spincoated on the GaAs wafer with NWs. (b) When the PDMS has dried, the PDMS embedded NWs can be lifted off the wafer. (c) After lift-off, the PDMS embedded NWs can be transferred to the Si membrane.
Spincoating PDMS
In order to transfer the NWs from the wafer to the membrane, a layer of PDMS (recipe shown in Appendix D) was spincoated on the GaAs wafer. The spincoating was done with an open bowl using the following recipe.
Step Rotation [rpm] Acceleration [rpm/s] Time [s] Lid open (0), lid closed (1) Function 1 0 20 5 1 Close lid 2 3000 1000 45 1 Spin PDMS 3 0 1500 0 0 Stop spin and open
lid
After spincoating the GaAs wafer was placed in the oven for 24 hours at 50 ◦C. After retriev-ing the GaAs wafer from the oven the PDMS was kept for at least another 24 hours outside of the oven, after this time the PDMS was hard enough to be peeled-off and transferred to the membrane.
Peel-off and transfer
In order to have the NWs in the PDMS, without the PDMS breaking, a square around the NW was cut. Then the PDMS was pulled using tweezers while at the same time cutting the NW at the bottom using a surgical knife, this is schematically shown in figure 4.7b.
The PDMS embedded NWs were ready to be placed on the membrane. For the correct place-ment the zoom function on a camera was used, by zooming in the location of the wires and the contacts was better visible. As the PDMS sticks to the membrane, it was difficult to make little
adjustments in the location of the PDMS. Therefore, acetone was used as a way to temporarily move the PDMS. The acetone evaporates shortly after applying to the membrane, the sample is now ready for measurements with the NW arrays.
4.3
Fabrication process, practice
The above described fabrication process was found too difficult to execute on a membrane. There-fore, we had to make some changes. In this section I will describe the challenges we found but also the successes.
Challenges
First of all, working with membranes is a challenging and difficult process due to their fragility. As we could not make the membranes ourselves, there was a limited supply of membranes.
During the UV-lithography we encountered the first problem. In order to explain this problem, I will first shortly describe how the MA/BA6 high-precision mask and bond aligner works between placing the sample and alignment.
1. The sample is placed on a stage and kept in place with a vacuum.
2. The stage moves up to the mask, where the sample is in contact with the mask. During this contact the stage will move such that the sample is completely parallel to the mask, this is important as the pattern on the sample will now completely the same as the pattern on the mask after exposure.
3. After the sample is completely parallel to the mask, the stages moves down by 100 µm. 4. With the sample being 100 µm below the mask the sample can be aligned with the structures
on the mask.
The alignment is very important between the first and the second exposure step, as we want to make sure that the contact point are on the Si and not on the SiO2. However, due to the vacuum
the membrane is curved making an exact alignment between the first and second step of exposure difficult. We therefore decided to get rid of the vacuum during second exposure step, this was done by placing a piece of tape on the vacuum hole. This was found to be destructive for the sample, as the somewhat sticky photoresist stuck to the mask. The membrane remained stuck on the mask while the stage went down, causing the membrane to break. This is schematically shown in figure 4.8.
In principle the photoresist should not be sticky, as we bake it at 150◦C for 90 seconds. We could try to bake it longer, but there is no guarantee that this will solve the problem. Therefore, we decided a different approach: instead of creating holes in the SiO2 and then align the contacts
we will first make the contacts on the SiO2and then etch holes using 400 nm of CSAR-62 undiluted
photoresist resist and the Voyager electron-beam lithography. On these holes we will then again evaporate the Cr and Au with the E-Flex, such that we have a metal contact to the Si membrane. However, this means that the NW array that will be placed on the membrane is not directly in contact with the membrane as there is 40 nm layer of SiO2in between. The additional layer can
influence the measurements, as there is another layer in which the light will refract. However, even if we were to etch the holes a native oxide layer would appear after time. The thickness of the native oxide layer depends on factors such as temperature and humidity, this is explained in more detail in Ref. [56]. Earlier research showed that the thickness of the native oxide on a p-type Si wafer was around 2 nm [57], which is an order of magnitude lower than the 40 nm of SiO
2 we
evaporated on the membrane. However, Tavakoli and Alarc´on-Llad´o showed that even with an 100 nm SiO2 layer underneath the PDMS embedded GaAs NW array there is a higher Jsc in Si[11].
Figure 4.8: Schematic representation of the different steps the MA/BA6 high-precision mask and bond aligner takes between sample placement and exposure in the case of vacuum (left column) and without vacuum (right column). Step 1. shows the sample (in grey) placed on the stage. 2. The stage moves up such that the sample and the mask are in contact. 3. The stage moves down such that the sample can be aligned in step 4. It can be seen that if there is a vacuum the membrane is hollow due to the vacuum. If there is no vacuum the membrane with photoresist sticks to the mask, such that in the last two steps the membrane breaks.
The above described method should, in principle, work. However, during the first lift-off our membrane broke. We therefore concluded that we need to stabilize the membrane. One way to stabilize the membrane is by adding support at the back of the membrane, such that the problem shown in figure 4.8 does not happen. We decided to do this by applying a droplet of S1813 photoresist, after baking the photoresist on the top of the membrane. S1813 is a positive resist that should be baked for 3 minutes at 115◦C. However, this baking process is made for thin layers of photoresist, while in our case we have a thick droplet. Therefore, we decided to bake S1813 for 24 hours in the over at 50◦C. We found that the resist was hardened, in addition this method was also beneficial for the Ma-N 1420 as this also hardened during the 24 hours. As we suspected, it was now possible to do UV exposure without any problem. The S1813 could be removed from the membrane with acetone. However, we found that there is some residual S1813. This residue can interfere with future measurements. However, if there is a metal applied before the S1813 this is no problem. This metal can be silver which functions as a back reflector or for example gold which can act like a back contact (see figure 4.9). At the end of the fabrication process we can then apply another droplet of S1813 and use this as stabilizer during the measurements.
Figure 4.9: Schematic representation of the proposed stabilization method. On the left the bare membrane, on the right the membrane with underneath it a metal (gold for contact or silver as a back reflector) followed by S1813 photoresist for stabilization.
Successes
Besides challenges we fortunately also encountered successes. First of all, the method of making the contacts as described in section 4.2 was found to be working on a piece of Si wafer. We made contacts on Si wafer with and without SiO2, as the diffusion length of Si is bigger than the distance
between the contacts[58,59], we found that we indeed need the SiO
2. This is shown in Appendix
C. Another success is the spincoating and transfer of the GaAs NW array in PDMS. With the method described in section 4.2.3, we successfully transferred from one substrate to the other. The successful transfer is shown in figure 4.10, here we places the GaAs NW array (shown inside the red circle) on top of Si wafer piece. It has to be noted that during the course of this thesis we did not receive the final GaAs NW array. Fortunately, we did receive a trial sample on which different diameters and pitch distance GaAs NW arrays were grown, however the height of these GaAs NW arrays was not yet perfected.
Figure 4.10: On the left: in PDMS embedded GaAs NW arrays with different pitch and diameters on top of a piece of Si wafer with gold contacts. The red circle shows the location of the GaAs NW arrays on the sample. Picture taken with a mobile phone. On the right: 20x zoom on the GaAs NW arrays, it can be seen that depending on the diameter of the NWs and the pitch the colour of the arrays changes. Picture taken with an optical microscope.
In order to see how well we can transfer the NW from one substrate to another we made Scanning Electron Microscope (SEM) pictures from the original substrate before (figure 4.11) and after transfer (figure 4.12). In figure 4.11, the SEM images part of the GaAs NW array can be seen before we spincoated with PDMS. The images are made under an angle of 10◦, such that the height of the NWs can also be seen. From these images we can see some defects in the NW array and it can be seen that the pitch is different; in a) the pitch distance is 300 nm while in b) the pitch distance is 600 nm.
(a) (b)
Figure 4.11: SEM images of a GaAs NW array under an angle of 10◦with a) p = 300 nm, d = 70 nm and b) p = 600 nm, d = 70 nm. It can be seen that some of the NWs have a different height, however there is already a periodic array.
Figure 4.12 a) shows the substrate on which the NW array with pitch 300 nm and diameter 70 nm was grown after lift-off under an angle of 45◦. It can be seen that wires with defects, e.g. they are grown onto each other, remain on the substrate. In figure 4.12 a close up of one of the wires (pitch 600 nm and diameter 100 nm) is shown under an angle of 45◦. It can be seen that the wire is not cut of straight, however the biggest remainder is only 51.89 nm.
(a) (b)
(c)
Figure 4.12: SEM images of the remaining GaAs NWs after nanowire transfer under an angle of 45◦with a) p = 300 nm, d = 70 nm, b) p = 300 nm, d = 100 nm and c) p = 600 nm, d = 100 nm. In a) it can be seen that the defects remain on the substrate while the single NWs have been remove. b) Shows a 120.000 x magnification of the removed NW array. In c) a 500.000 x magnification of the remains of a single NW. It can be seen that at most 51.89 nm remains. In addition, we can see that the NW is cut under an angle, as the right part of the NW is lower in height than the left part.
In addition to cutting the NWs from the substrate we tried to break them by first rubbing over the PDMS embedded NWs and then remove the PDMS with tweezers. However, we found that this method was not successful as most of the NWs remained on the substrate and were only partly broken. This is shown in Appendix E.
4.4
Measurements
The photo conductive measurements will be done using the Oriel QuantX-300 Quantum Efficiency System made by Newport, the QUANTX-300 has the possibility to measure the current and voltage while changing the wavelength in steps of 1 nm. This way, the change of resistance and thus Rsh
can be measured per wavelength. As Rsh is proportional to 1/absorption (see section 4.1) it is
possible to find possible Fabry-P´erot resonances and thus prove the enhanced light absorption in thin film Si using GaAs NW arrays, as presented by Tavakoli and Alarc´on-Llad´o [11].
A schematic figure of the QUANTX-300 is shown in figure 4.13. In the QUANTX-300 the light is produced by a Xenon lamp (1), after which the light passes a chopper wheel (2) where the frequency of the light can be set, followed by a filter (3) and a monochromator (4). After the monochromator the light passes a beam-splitter (5), where halve the light is detected by the
Monitor (6), whose signal directly correlates to the optical power at the sample at each wavelength step. The other halve is directed to the sample (7). Lastly, the reflection is measured (8). Using the software we can apply a current to the sample while measuring the voltage, the electrical scheme is shown in figure 4.1b.
Figure 4.13: Schematic representation of the QUANTX-300. Where the numbers correspond to (1) a Xenon lamp, (2) chopper wheel, (3) filters, (4) monochromator, (5) beamsplitter, (6) monitor, (7) sample and (8) reflection monitor. Figure adapted from Ref. [60]
At the moment of writing, we were not able to do the measurements as our sample was not ready. However, we hope to be able to do these measurements in the near future.
Conclusion
In this chapter we will try to answer the two main research questions of this thesis, namely; Is it beneficial to use a perovskite top cell to couple the light to the Si thin film bottom cell? (Chapter 3). And can we experimentally prove the enhanced light absorption in Si thin film if a GaAs NW array with different pitch and diameter is used as a light coupler? (Chapter 4)
As described in chapter 3 we simulated a single 6 µm long MAPbBr3 NW, with changing
diameter from 200 nm till 900 nm and for a surrounding medium n = 1 and n = 1.5, for which we calculated the absorption and forward scattering efficiency. We found that by changing the refractive index the modes are are slightly blue shifted for higher refractive index. In addition, we saw that the light absorption and forward scattering is bigger than the geometrical cross section of the single NW. Meaning that less material is needed while absorbing the same amount of light. However, the single NW does not show the influence of a periodic NW array on the light trapping withing a bottom cell, which is the proposed tandem cell design. Therefore, we simulated MAPbBr3 and a MAPbI3 NW array with height 6 µm on top of 2 µm thick Si, where we kept
the pitch distance at 1.000 nm while changing the diameter from 200 nm till 900 nm. In these simulations we kept the surrounding medium at n = 1.5. To compare the different perovskites we calculated the Jsc the complete cell and the bottom cell, we did with a NW array top cell,
perovskite film top cell and without a top cell. The Jsc in thin film Si was found to be 17.9
mA/cm2. We found that using a MAPbBr
3 NW array as top cell does increase the absorption,
and thus the Jsc within the Si bottom cell. However the Jsc of the complete cell (maximum Jsc =
21.8 mA/cm2, for d = 850), which is an increase of almost 4 mA/cm2compared to bare Si, however
it was not improved much compared to the Jsc of the MAPbBr3 film (Jsc = 20.5 mA/cm2). As
it is rather difficulty to make a NW array top cell, while the difference in Jsc between the thin
film and the NW array top cell we conclude that MAPbBr3 is not feasible for our tandem cell
design. In addition, we found that Jsc was increased to a maximum of 29.2 mA/cm2 for a NW
array with d = 650 nm. However, using a MAPbI3 film we calculated the Jsc to be 27.0 mA/cm2.
This indicates that there is an increase in absorption if the refractive index between the NW and the surrounding medium is bigger. Therefore we coated a d = 600 nm MAPbI3 NW with a high
refractive index material of n = 4. Although we found an increase in Jλ>800nm
sc for the bottom cell,
we found that the overall absorption was not changed much; from 27.7 mA/cm2 for an bare wire to 27.8 mA/cm2for the coated wire.
Based on the above described results we concluded that the use of MAPbBr3and MAPbI3NW
array top cell is not beneficial. However, current perovskite solar cells do not have a perovskite layer of the 6 µm that we simulated, instead they have a thickness comparable to the wavelength of light
[61,62]. It might therefore be interesting to do simulations where the length of the perovskite NW
is changed for a certain diameter. However, as the effect of refractive index will remain, I would propose to run these simulations only for MAPbI3 which is at the moment the most promising
material.
In chapter 4 the experiment is described. We showed successful method to etch holes in a layer of SiO2that is on a piece of Si wafer, we can then align these holes such that we can create contacts
found that making these contacts on the membrane was challenging, as the photoresist stuck to the mask needed for photo-lithography which broke one of the membranes.
We therefore propose a new method; first the gold contacts are created (without etching the SiO2) on the membrane using photo-lithography. After making these contacts, photoresist is
spincoated in order to do e-beam lithography. Using the e-beam lithography holes are made in the photoresist around the contact points, here part of the gold and the SiO2underneath is etched away
following evaporation of chromium and gold. This way, some gold is touching the bare membrane such that it is possible to do photo-conductive measurements.
Another solution could be to stabilize the membrane. One could think of using PDMS, as it remains flexible while supporting the fragile membrane. However, PDMS is a polymer and can melt when it is too hot, which causes a problem during the evaporation of gold in the E-Flex. Therefore, a more suitable materials could be sol-gel, which starts as a gel, but becomes SiO2
when dried[63]. During the drying the sol-gel shrinks, which might cause the membrane to brake.
The final solution that we would like to propose for future fabrication is stabilizing the membrane with photoresist, as described in the ”challenges” in section 4.3. So far, we found that this method is working for a photo-lithography fabrication process.
Lastly, we have shown a method to successfully transfer a NW array from one substrate to another. Due to the polymer PDMS the NW array remains in its periodic array, while it can stick to a new substrate. This way it is possible to transfer NW array to different samples, without breaking them. However, this method is not useful for big NW arrays, as the thin PDMS layer is fragile.
With the above described conclusions, the second research question remains unanswered for now. However, as an experimental outlook we propose the photo-conductive measurements as described in section 4.4. We expect that photons with an energy below GaAs band gap energy will be coupled to the Si, where Fabry-P´erot resonances will be seen in the absorption plot as predicted by Tavakoli and Alarc´on-Llad´o [11].
