Radiation hardness of thin film solar cells

Thesis report

Paraskevi Danaki

Supervisor: Charlotte Platzer-Bj¨orkman

Department of Physics and Astronomy Uppsala University

Sweden June 18, 2019

In this master project, the proton radiation effects in thin film solar cells were studied.

Electrical characterization was performed before and after radiation to determine the effects on Cu₂ZnSnS₄ (CZTS) and CuIn_xGa_(1−x)Se₂ (CIGS) cells. Proton radiation caused degradation on both types of cells. Room temperature annealing gave improved performance, but the recovery was slow. Thermal annealing at 100^◦ C had notable im- pact on the basic current-voltage parameters. In CZTS cells, after 17 hours annealing, the open circuit voltage recovered completely, short-circuit current recovered almost com- pletely and the fill factor had minor recovery. Quantum efficiency measurements showed degradation for both CIGS and CZTS devices in all wavelengths after radiation, and partial recovery after annealing. Doping density measurements were also implemented with two different methods, capacitance-voltage and drive level capacitance profiling. The radiation did not cause any significant change in the doping density. Additionally, admit- tance measurements that were performed in both types of cells, could not give conclusive results. Lastly, a model was also created before and after the radiation for a CZTS cell, in order to quantify the caused defects. To fit the experimental data after the radiation, an increase of defect density in the CZTS layer of 87% was needed.

First of all, I would like to thank my supervisor, Charlotte Platzer-Bj¨orkman. She trusted me from the very beginning and supported me throughout the whole project. She gave me guidance and valuable feedback whenever I needed, while allowing this project to be my own work. Thank you Lotten.

Secondly, I would like to offer my special thanks to my mentor on my master studies, Matthias Weiszflog. His valuable suggestions and discussions during my studies helped me choose the path I wanted to take.

I am also particularly grateful for the assistance given by the solar cell group at Up- psala University. In particular Volodymyr Kosyak shared with me his expertise on the experimental devises and Sethu Saveda Suvanam on radiation effects and the space en- vironment.

My special thanks are extended to my parents for supporting me throughout my life.

Finally, my master thesis would have been impossible without the aid and enthusiastic encouragement of Anastasios Gorantis. He helped me stay focused, organized and evolve as a physicist.

1. Sammanfattning p˚a svenska 5

2. Introduction 6

3. Theory 11

3.1. Solar cells . . . . 11

3.1.1. Semiconductors . . . . 11

3.1.2. pn junction . . . . 12

3.1.3. One diode model . . . . 15

3.1.4. Quantum efficiency . . . . 16

3.1.5. Capacitance measurements . . . . 17

3.1.6. Main losses . . . . 17

3.2. Radiation . . . . 18

3.2.1. Theory . . . . 18

3.2.2. Effects of radiation on solar cells . . . . 20

4. Experimental Method 25 4.1. Experimental procedure . . . . 25

4.2. Equipment . . . . 27

4.2.1. IV . . . . 27

4.2.2. QE . . . . 27

4.2.3. CV . . . . 28

4.2.4. DLCP . . . . 28

4.2.5. TAS . . . . 29

4.3. Modeling . . . . 29

5. Results 34 5.1. Electrical Characterization . . . . 34

5.1.1. CZTS . . . . 34

5.1.2. CIGS . . . . 41

5.2. Device Modeling . . . . 45

5.2.1. Before radiation . . . . 45

5.2.2. After radiation . . . . 47

6. Discussion and Conclusions 53 6.1. Discussion . . . . 53

6.2. Conclusions . . . . 54

6.3. Future work . . . . 54

Appendices 59

A. Admittance 59

Solceller ¨ar en v¨axande del av energitillf¨orseln inte bara i Sverige utan i hela v¨arlden.

En av f¨ordelarna med solceller ¨ar att de kan anv¨andas i frist˚aende system. En typ av frist˚aende system ¨ar rymdtill¨ampningar. Rymduppdrag som utf¨ors n¨armare ¨an planeten Mars, anv¨ander solceller f¨or energitillf¨orseln. Ett problem ¨ar dock att i rymden finns det gammastr˚alning och andra partiklar som kan skada solceller och f¨or att skydda solceller och f¨orl¨anga deras livstid, m˚aste man k¨anna till str˚alningseffekter.

P˚a Jorden anv¨ands kisel solceller, men de ¨ar tunga och degraderas av str˚alning. D¨arf¨or anv¨ands andra material i rymden som har h¨og effektivitet och ¨ar tunnare och l¨attare ¨an ki- sel. I det h¨ar examensarbetet, unders¨oktes om tv˚a material kan anv¨andas, CuInxGa(1−x)Se2

(CIGS) och Cu₂ZnSnS₄ (CZTS). Solcellerna utsattes f¨or str˚alning med protoner och de- ras elektriska egenskaper m¨attes f¨ore och efter str˚alningen. Dessutom, utvecklades en mo- dell f¨or CZTS beteende f¨ore och efter str˚alningen f¨or att f¨orst˚a och kvantifiera str˚alningens effekter p˚a solcellerna.

B˚ada materialen degraderades efter str˚alningen eftersom CIGS effektivitet sj¨onk med 58% och CZTS med 40%. Proven v¨armebehandlades i 100 ^◦C och genom detta ¨okade deras effektivitet delvis. Den utveklade modellen visar att defektdensiteten i CZTS ¨okade med 87% efter str˚alningen.

Sammanfattningvis har CZTS och CIGS solceller h¨og potential att anv¨andas i rymd- till¨ampningar eftersom de visar en h¨og tolerans f¨or str˚alning, om deras effektivitet skulle

¨okas.

In the last 60 years the world has changed significantly, with electrical devices being an integral part of our daily life. Electrical energy is required for all of them to work, so most of us cannot imagine living without a reliable energy supply. The increasing demands request increased supply, but one source is not enough as nothing comes without any disadvantages. The sun is the main source of energy on planet Earth. It heats air masses creating wind, which powers wind turbines. It sets the water cycle in motion, powering hydroelectric power plants. It plays a crucial role in photosynthesis feeding organisms and vegetation that are now used as fossil fuel, and of course, it provides energy directly with the help of solar cells.

The history of solar cells started in 1839 when Alexandre Edmond Becquerel discovered the photovoltaic effect, the main principle behind solar cells. The first solar cell, with an efficiency of around 1%, was produced in 1885 by C. Fritts [1]. In 1905, Albert Einstein explained the photoelectric effect, an effect similar to the photovoltaic, which laid the theoretical foundation for the study and development of solar cells. Some years later, in 1954, a working solar cell with 6% efficiency was developed at Bell laboratories [2].

The next revolution came in 1976, when Carlson and Wronski fabricated the first thin film solar cell from amorphous silicon [3]. In the next decade, the two sibling materials that will be our focus, namely CuInxGa_(1−x)Se2 (CIGS) and Cu2ZnSnS4 (CZTS), were fabricated for thin film solar cell uses.

The number of applications is increasing dramatically as the costs drop. From calcu- lators to solar parks, photovoltaics are an essential and growing part of energy supply.

The ability to build off-grid systems is one of the reasons the field of solar cells is grow- ing. Space applications are a type of off-grid systems, although probably not the most common ones. Most missions that are closer than the orbit of Mars, use solar cells as electricity supply. Further than Mars, sunlight is not enough to power missions, so other methods are used.

The first satellite with solar cells was launched in 1958 named Vanguard 1 [4]. In the beginning, crystalline silicon solar cells were used for space missions, since Si was the only available material. The disadvantages however were many. Silicon cells are heavy, do not have high efficiency compared to their mass, and are affected by solar flare particles.

When solar cells based on other materials with better properties were manufactured, they dominated. Now, triple junction solar cell are mainly used, although they also have both advantages and disadvantages. The research for radiation hard cells with high efficiency is still ongoing.

The aim of this project was to investigate the effects of proton radiation on thin film solar cells. Understanding the effects of radiation is essential in order to prevent damages and improve efficiency and lifetime in space applications. To that end, one needs to mea- sure the cells’ electrical characteristics before and after the radiation. There are many characterization methods that can be used, such as electroluminescence, capacitance mea- surements, optical techniques etc. In this work we focused on electrical characterization and capacitance measurements. The effects of radiation on capacitance based measure-

ments have not been included in radiation hardness research of CZTS cells yet, so part of this project was to attempt to determine them. Modeling the device before and after the radiation was also performed, since it can offer insights in the damages that radiation causes.

The rest of this thesis is organized as follows. First, in chapter 3, we review the working principles of solar cells and the characterization methods we used. Radiation and why it affects solar cells will also be revised. In chapter 4 we explain the experimental procedure we followed, and review the equipment that we used. Then, in chapter 5 we present the experimental results. Finally, in chapter 6 we discuss these results, draw some conclusions and provide an outlook.

3.1. Solar cells

The main principle behind solar cells is the photovoltaic effect, which was first demon- strated by Becquerel in 1839. When materials are exposed to light, electric current and voltage are created. This happens because light transfers its energy to the electrons of the material, which get excited. The photovoltaic effect is similar to the photoelectric effect, but in the photoelectric phenomenon the emitted electrons escape the material while in the photovoltaic, they stay inside it.

3.1.1. Semiconductors

The energy of electrons bound to an atom is quantized, so it can only take certain values called energy levels, which are given by the Schr¨odinger equation. The lowest energy level that an electron can occupy is the ground state, and if the electron absorbs enough energy, it can go to a higher, excited state.

In crystalline materials, the energy levels are so close to each other that they create continua called bands. The band that contains the ground states with higher energy is called the valence band and the band that contains the excited states with the lowest energies is the conduction band. If a range of energies has no allowed states, it is called a band gap. Depending on the size of the band gap, materials are divided into the categories metal, insulator and semiconductor. In metals, the conduction and valence bands are very close or overlap, and there is a small or no band gap. In insulators there is a large band gap, which prevents electrons from going from the valence to the conduction band. Semiconductors are in the middle, with small band gaps that, in certain cases, allow the excitation of electrons from valence to conduction band.

There are two types of band gaps: direct and indirect (Figure 3.1). The type of band gap is an intrinsic property of the material. An example of a direct band gap material is GaAs and an example of an indirect one is Si. In a direct band gap material, the maximum energy of the valence band and the minimum energy of the conduction band have the same crystal momentum. Therefore, only a change in energy is required for an electron to go from the valence to the conduction band. In an indirect band gap material, on the other hand, in addition to a photon, an electron needs to interact with a phonon, making the excitation to the conduction band less likely. However, the requirement of interaction among more particles, also makes it more difficult for an electron and hole to recombine, as the presence of a phonon with the appropriate momentum is required.

Momentum

Energy Eg

Momentum

Energy Eg

Direct band gap Indirect band gap

Photon absorption

Phonon absorption

Valence band Conduction band

Valence band

Conduction band

Photon absorption

Figure 3.1.: Direct and indirect band gap

The occupation probability of the energy states follows the Fermi-Dirac distribution described with:

F (E) =



exp E − E_F kT

 + 1

−1

(3.1) where F is the occupation probability, E is the energy of the state, E_F is the energy at which the occupation probability is 0.5 (Fermi level), k is the Boltzmann constant, and T is the temperature.

When the material is doped with impurities, the Fermi level moves towards the valence or the conduction band, depending on the type of the doping, and the electrical properties of the semiconductor are modified. Pure silicon has four valence electrons (electrons in the outer shell of the atom), which bind to the neighboring atoms creating the lattice.

If silicon is doped with a group III element, such as boron (B), which has three valence electrons, one electron of the neighboring silicon atoms is unpaired creating a vacant state of electron hole. If, on the other hand, silicon is doped with a group V element, like phosphorus (P), which has five available electrons, there is one extra free electron in the lattice. When the material contains more holes it is called “p-type” and when it has more free electrons it is called “n-type”. The type of semiconductor indicates which is the material’s majority carrier. It should be clarified here that there is no additional charge, so the crystal is still neutral.

The addition of impurities can help to increase the number of free carriers. When the material is at temperature 0K, carriers do not have enough energy to move in the lattice.

With increasing temperature, some electrons become thermally excited and are free to move in the lattice, leaving behind empty holes, which can also move. Doping atoms add energy levels inside the band gap, giving the opportunity to electrons to move to or from these states. At room temperature, there are enough electrons and holes available to move freely, so the material conducts current.

3.1.2. pn junction

Things become even more interesting when a p-type material comes into contact with an n-type material, creating a pn junction. When this happens, the excess electrons from the n-type diffuse into the p-type and the holes from the p-type diffuse into the n-type.

When they meet, they recombine, leaving ions of opposite charge in each side. The ions create an electric field and the area in which they are located, is called the depletion region or space charge region (Figure 3.2). The electric field created by the depletion region drives a current named drift current, that counteracts the diffusion of electrons to the p-type and holes to the n-type. The process continues until an equilibrium between the two forces is achieved.

+

+ + + + + + + + + + + + + + +

- -

- -

- -

- -

- -

- -

- -

- - n-type

p-type Space

charge region

Diffusion length

Figure 3.2.: The pn junction

There are two options to connect a pn-junction to a circuit. Either the positive terminal can go to the p-type and the negative terminal to the n-type or the opposite. The first is called forward bias and the last reverse bias.

When the junction is connected with a forward bias, the voltage drives the holes in the p-type and the electrons in the n-type to the depletion region. The charge carriers recombine with each other reducing the width of the space charge region and the resistivity of the junction. With the shortened depletion region, it is easier for majority carriers to cross the junction and recombine with the minority carriers in the other side. The result is a constant carrier flow, i.e. a current. Electrons and holes are oppositely charged, but they also flow in opposite directions so their current is added and the junction behaves as a conductor.

In a reverse bias connection of a pn junction, the majority carriers are pulled away from the depletion region increasing its width. The majority carriers cannot cross the junction, the resistivity increases and the junction behaves as an insulator.

The current of an ideal pn junction is described with:

J (V ) = J₀



exp qV kT



− 1



(3.2) where J is the current density, J₀ is the saturation current density, q is the electron charge and V is the applied voltage. A solar cell in the dark is actually a pn junction, which means a diode, and it follows this equation. The ideal IV curve is shown in

Figure 3.3. The saturation current density is given by:

J₀ = A qD_en²_i

L_eN_A + qDhn²_i L_hN_D



(3.3) where A is the cross-sectional area, D_e and D_h are the diffusion coefficients for electrons and holes, n_i is the intrinsic carrier concentration, L_e and L_h are the diffusion lengths for electrons and holes, and N_A and N_D are the acceptor and donor densities in the n and p type respectively.

I

V Forward

Bias Reverse

Bias

Break-down voltage

Knee voltage

Figure 3.3.: Ideal IV curve

When a solar cell is illuminated, there are three possibilities. Firstly, the photons can be reflected in the surface and not interact with the cell at all. Secondly, the photons can pass through the solar cell, also without interacting. In the last case, energy from the light can be deposited into the electrons in the valence band, exciting them to the conduction band. The photon should have energy greater than the band gap to excite an electron. An electron-hole pair is generated and the carriers are added to the majority and minority carriers. As the concentration of minority carriers increases, the drift current increases. The minority carriers cross the junction, become majority carriers and exit the cell recombining in the external circuit and depositing their energy.

The generation of electron-hole pairs and their flow to the external circuit, creates an extra current named light current. The new equation that describes the diode under illumination is:

J (V ) = J₀



exp qV kT



− 1



− J_L (3.4)

where J_L is the light current density. The light IV curve is shown in Figure 3.4.

I

V

Voc

Jsc MPP

V_m

J_m

FF = V_mJ_m J_scV_oc

Dark

Light

Figure 3.4.: Ideal IV curve

3.1.3. One diode model

The most fundamental characterization method is to measure the current of the cell while altering the applied voltage. The IV curve under illumination can give very useful infor- mation about the solar cell, since important parameters such as J_sc (short circuit current density), Voc(open circuit voltage) and Pmp(maximum power point) are determined from it. The open circuit voltage is given by:

V_oc = kT

q ln J_L J₀ + 1



(3.5) Another helpful parameter extracted by the IV curve is FF (fill factor, Equation 3.6) which has no physical meaning but describes the “squareness” of the IV curve, i.e. how close to the maximum power point the cell works.

F F = F F0(1 − rs) (3.6)

where

r_s= R_s

R_ch, with R_ch= V_oc

J_sc (3.7)

where Rs is the series resistance, and

F F₀ = v_oc− ln(v_oc+ 0.72)

v_oc+ 1 (3.8)

where

voc= V_oc

nkT /q (3.9)

If one has to choose only one parameter that will judge the performance of a solar cell, then this parameter must be the efficiency (η). It describes how well the cell converts the energy of the incoming light into electric energy. All the parameters mentioned previously are included in η which is described with:

η = P_out

P_in = P_mp

P_in = J_scV_ocF F

P_in (3.10)

However, when a deeper analysis is needed, η is not enough because it does not give information about losses.

Another parameter that characterizes solar cells is the ideality factor (n). It describes how close is the solar cell to an ideal one. For an ideal solar cell, where there is no recombination in the space charge region, n = 1, but for experimental devices, n can take other values.

The equivalent circuit of a solar cell according to the one diode model is shown in Figure 3.5. The diode is the pn junction in the dark, the current source is the light generated current (J_L) and there are two resistances that represent losses. The shunt resistance (R_sh) gives J_Lan alternative path so it needs to be as high as possible, because we want the current to go out of the circuit, not return back. The series resistance symbolizes the resistances in the materials and the surrounding circuit and needs to be minimized because there is voltage drop in it. Both resistances affect the fill factor and thus the efficiency of the solar cell. The current in this case is given by:

J (V ) = J₀exph q

nkT(V − R_sJ )i

+ G_shV − J_L (3.11)

where G_sh is the shunt conductance.

I_L

ID

Rsh

R_s

Ish

I

V +

- Ideal cell

Figure 3.5.: Solar cell equivalent circuit

3.1.4. Quantum efficiency

Quantum efficiency is another characterization method, used among other things to mea- sure the band gap of the material and locate losses. The tested solar cell is exposed to photons of different energies, and the number of carriers collected by the cell are measured. Photons with energy below the band gap cannot create electron-hole pairs therefore quantum efficiency (QE) is zero in those wavelengths.

There are two types of quantum efficiency, external (EQE) and internal (IQE). External QE takes into account all the incident photons. Internal QE on the other hand, takes into account only the photons that are absorbed by the solar cell, so it only depends on photogeneration and collection. If the reflection and transmission profile of the cell is available, IQE can be obtained from EQE.

The QE curve of an ideal solar cell has a square shape. However, real cells have many losses as we will discuss later. If losses happen in the front surface of the cell, QE drops in high energy photons (blue light) because they are absorbed only in the front surface (Figure 3.2). The green portion of the QE curve is affected by losses in the bulk, and the red part, which can penetrate more, can be absorbed in the rear surface. Other factors that reduce QE are shading losses, which are independent of the wavelength, and reflection.

3.1.5. Capacitance measurements

When we discussed previously about the formation of the pn junction, we mentioned that in the depletion region, there is an electric field created by ions. The concentration of negative ions on the one side and of positive ions in the other side, can also be seen as a capacitor. Measuring the conductance and capacitance of the junction can give valuable information about doping density, free carriers, deep states and other.

In a standard capacitor C = Q/V , where C is the capacitance, Q is the charge and V is the voltage. But in a solar cell, Q does not vary linearly with voltage so a more general equation in needed, and this is C = δQ/δV . A small ac voltage of the form V = V_ac[cos ωt + j sin ωt] is applied to the cell, and the capacitance is measured. We will discuss the information that can be obtained from this data in chapter 4.

3.1.6. Main losses

The world record efficiency of a single junction solar cell is 29.1% [5] which means that 70.9% of the incident energy is lost. There are many mechanisms through which the energy is lost during the process and we will discuss some of them.

• Optical losses When we talk about optical losses we mainly talk about reflection and transmission. Reflection can occur at the front contact, on the front surface of the cell or on the rear surface. In the first two cases photons are not absorbed, and therefore get lost. Shading by the grid, dead areas and parasitic absorption are also important causes of optical losses. They mostly effect the current generated by the solar cell. There are a number of ways to reduce these kind of losses. Anti- reflection coating, surface texturing and appropriate top contact design are the most important ones.

• Generation losses Generation losses are the ones that are caused by material properties. When incoming photons have less energy than the band gap of the material, they cannot excite electrons and create electron-hole pairs. They might be absorbed as heat or simply not interact.

• Thermalization losses

When an electron absorbs energy from an incoming photon and goes from the va- lence to the conduction band, it initially goes to a higher state inside the conduction

band. However, it rapidly relaxes to an energy equal to the band gap. The rest of the energy of the photon is lost in the material through thermalisation.

• Recombination losses When an electron in the conduction band loses its energy and goes to the valence band, it can meet a hole. In this case, they recombine and their energy is lost. Recombination losses are of the more complex and difficult to prevent and they affect both the current and the voltage. There are three types of recombination: radiative, Shockley-Read-Hall (SRH) and Auger recombination.

Radiative In this type of recombination, the energy of the electron hole pair is emitted as a photon with energy equal to the band gap. The photon is not likely to be absorbed because its energy is exactly in the threshold, so the energy is most probably lost. Radiative recombination is more probable in direct band gap materials, because in indirect ones there must be a phonon interaction as well.

SRH When defects with energy states in the forbidden region are present in the crystal, it is possible for two carriers to meet there and recombine. The defects could be added on purpose (dopants) or through contamination during the manufacturing process. Radiation affects solar cells because it increases the defect density. Grain boundaries and interfaces can also work as recombination centers. The released energy is in the form of a photon or heat. It is more probable than radiative recombination because the energy difference is smaller.

Auger Auger is the recombination that happens when an electron and a hole recom- bine and transfer their energy to another electron in the conduction band or a hole in the valence band, which gets excited to a higher energy state. It looses the extra energy by collision with other atoms. This type of recombination is more probable when the material has higher minority carrier concentration, such as materials with higher doping.

• Resistive losses

The existence of parasitic resistances (series and shunt), results resistive losses due to voltage drop in the series resistance and current loss in the shunt resistance.

The series resistance has its root in bulk resistances of the semiconductor, metallic contacts and interconnections while shunt resistance is mainly due to leakage at the edges of the cell across the pn junction. Crystal defects and impurities in the junction can also cause decrease of shunt resistance. The parameter that is mostly effected by resistances is the fill factor.

3.2. Radiation

3.2.1. Theory

Radiation in space can be divided into electromagnetic and particle radiation. The first is wave radiation and will not be discussed in detail. The other type, particle radiation, will be the focus of this project. The effects of each type depends on properties of the incoming particles such as energy, mass, and charge and on properties of the target material such as atomic mass and density. However, the effects are difficult to separate, because many of the incoming particles induce secondary effects.

Radiation

Electromagnetic ^Particles

Ionizing Non Ionizing Charged Uncharged

Protons ...

Electrons ...

Neutrons ...

Figure 3.6.: Types of radiation in space

Electromagnetic radiation can be divided into ionizing and non-ionizing particles of zero rest mass, with the border between the two types roughly defined. Non-ionizing particles do not have sufficient energy to excite atoms in the target material, but induce photochemical reactions and have thermal effects. Ionizing radiation has more energy and interacts with matter through the photoelectric effect, Compton effect, and pair production depending on the energy.

Particles with non-zero rest mass can either be charged or uncharged. The interaction of uncharged particles, such as neutrons, with matter is similar to the electromagnetic one.

Charged particles on the other hand, transfer their energy through Coulomb interaction, therefore behave differently when interacting with matter. When they enter a material, they gradually lose their energy by ionizing material atoms. How deep into the material they will penetrate, depends on the initial energy, the type of particle and the material it interacts with. The energy loss per unit path length is called stopping power and usually shows a peak before the particle completely stops (Bragg peak, Figure 3.7). The existence of this peak means that a large part of the energy is deposited at a specific depth in the material. Taking into account the material properties, the energy of the radiation can be adjusted so that most of it is deposited to the desired depth.

Path Length

Stoping Power

Figure 3.7.: Bragg peak

3.2.2. Effects of radiation on solar cells

Charged particles can affect solar cells in two ways. Depending on their energy they can ionize lattice atoms or they can cause displacement damages. The incoming particles can remove lattice atoms out of their positions creating point defects (Figure 3.8). The atom might leave its position (vacancy) or it can exchange positions with another lattice atom (antisite). Other types of point defects are interstitials, where atoms take positions that are not normally occupied, substitutions, where lattice atoms are substituted with other elements, and Frenkel pairs, which are vacancy-interstitial pairs. Radiation can produce Frenkel pairs or move already existing defects such as substitutes, interstitials, or antisites. The creation of Frenkel pairs is more important because in the other case there is no creation of new defects but rather conversion of one type of defect into another [6].

As an example, if an already existing vacancy or substitute moves to a different position due to radiation, there are no new defects created, so only minor changes are expected.

Vacancy

Antisite

Substitute

Frenkel pair

Interstitial

Figure 3.8.: Defect types

Of the two materials that we investigated, namely CZTS and CIGS, the first one has not been thoroughly studied yet regarding radiation effects. The other one is more mature, therefore defects and stability issues are better known. The three possible defects that radiation could cause are Cu_i− V_Cu, In_i− V_In and Se_i− V_Se [6]. Cu is highly mobile at room temperature so Cu defects can easily be self-annealed. Although In is less mobile, the In vacancy can also possibly be annealed, with the help of Cu-exchanging reactions.

Lastly, Se defects are much less likely because they have high formation energy, and they can possibly also be annealed with the help of Cu-exchanging reactions.

Regarding CZTS, the probable defects and their effects have been studied, but not radiation caused defects. However, even these studies can give us important information.

According to Baranowski et al [7], the Cu_Zn defect is the most dominant one, and its energy level is 0.12eV above the valence band maximum, so it is an acceptor level. The V_Cu, V_Zn, Zn_Sn and Cu_Sn defects have also relatively low formation energies so they are also probable. The V_Cu, V_Zn and Zn_Sn have shallow acceptor levels, but experimental results have found deep acceptor defects that were attributed to Cu_Sn [8]. This means that Cu_Sn could function as a recombination center.

An overview of radiation studies of thin film solar cells in literature is given in Table 3.1.

In most cases, cells show degradation under radiation. However, depending on the energy

and flux of the radiation, even some performance improvement is possible. This is the case for lower fluxes. The effect is considered to be equivalent to light-soaking [9], [10], [11] and to hydrogen-passivation of recombination centers [12].

Degradation starts when flux increases and the higher the flux, the more the cells degrade. This can be attributed to many causes, and since it is a complex matter, it is difficult to be certain which of them is the real cause of the degradation. Ionization effects [9], [13], recombination centers generation [13], [12], [14] and optical transmittance losses [12] are some of the proposed reasons.

Annealing was also used in some of the experiments in order to accelerate any cell recovery. Lamb et al [15] used 100^◦C annealing and a dramatic change was observed in the performance. Jasenek et al [13] used room temperature annealing for six months, and the cells recovered partially.

Table 3.1.: List of studied papers. “e” represents electron radiation and “p” proton radiation.

Energy Dose (cm⁻²) Cell Characterization method Reference

e

60-250 keV 10¹⁴− 10¹⁷ CIGS IV [9]

100 keV 10¹⁵ CIGS EL [10]

60-600 keV 10¹⁴− 10¹⁷ CIGS IV, EL, CV [11]

0.5-3 MeV 10¹¹− 5 · 10¹⁸ CIGS IV, CV, CF [13]

2 MeV 10¹⁴− 2 · 10¹⁷ CZTS IV, PL [16]

p

380 keV 10¹²− 3 · 10¹⁶ CZTS IV, PL [16]

0.5 MeV 10¹²− 10¹⁴ CdTe IV, CV, QE, SCAPS [15]

3 MeV 10¹⁰− 10¹³ CZTS IV, QE [14]

4 MeV 10¹¹− 10¹⁴ CIGS IV, CV, CF [13]

15 MeV 10¹²− 10¹⁵ CdTe IV, Optical Transmittance [12]

4.1. Experimental procedure

The radiation was performed at The Tandem Laboratory in ˚Angstr¨om. The solar cells were irradiated with proton radiation of 250 keV and a dose of 10¹² protons/cm². The reason we chose this energy is because the Bragg peak (Figure 3.7) in this case is within the device, therefore most of the energy would be deposited there. The stopping range can be calculated with SRIM simulation [17], but it was considered beyond the scope of this project, so the energy was chosen based on previous research [14]. As discussed in section 3.2, protons induce both ionization and displacement damages, but at low energies the displacement damages dominate. The type of particle was chosen based on what causes problems in space applications and what is feasible as an experimental procedure in ˚Angstr¨om. Most satellites orbit Earth in a low orbit (∼ 2.000 km above Earth), where there is strong influence by the inner Van Allen radiation belt. This area contains large numbers of protons that are trapped by Earth’s magnetic field. The dose was chosen to be 10¹² protons/cm², which corresponds to ∼ 10 years in low Earth orbit.

We attempted to investigate two different thin film materials: CZTS and CIGS, with the second considered more as reference. An overview of the samples and the measure- ments we performed in each of them is shown in Figure 4.1. For each material, we had two samples in which we performed various experiments. The reason we chose to have two samples of each material, is because we wanted to perform frequent measurements in at least one of them. However, in order to measure, one needs to connect the sample to the measurements devices and this is done with small metal probes. Multiple connections can damage the cell’s surface, therefore, we decided to take the risk for only one, and have a second as reference sample. Furthermore, it was safer to have two, in case one of them got damaged at any point during the process.

CZTS1 CZTS2 CZTS

CIGS

IV, QE ^{IV, TAS,} CV, DLCP

CIGS1 CIGS2

Figure 4.1.: The samples and the measurements we performed on them. The fabrication names of the samples are CZTS1: 23SSe8, CZTS2: 23SSe7, CIGS1: T62a2 and CIGS2: T62a12

The CZTS samples were manufactured in ˚Angstr¨om by Ross et al [18]. The two samples had slightly different properties before getting exposed to radiation, because they had different S/(S + Se) ratio. The CIGS samples were also manufactured in ˚Angstr¨om by Frisk et al [19]. It was originally one sample that was cut in two pieces for the reasons we mentioned before.

For CZTS1 we performed IV before the radiation, and after the radiation we measured IV every day for 2 weeks. After that, thermal annealing was performed at 100^◦C and an IV measurement followed each annealing step. 100^◦C was chosen based on previous research [15] and because this temperature can be reached in space. The annealing steps were: 10 minutes, 30 minutes, 1 hour, 2 hours, 5 hours and 8 hours. The annealing process lasted about 2 weeks and, when it was completed, we decreased the frequency of the IV measurements finishing the measurements after 2 weeks. QE was measured before the radiation, after it, and after 2 hours annealing.

For CZTS2 we had a different approach. The measurements (IV, TAS, CV, DLCP) were not frequent but after certain points. The points we chose were: before radiation, after radiation, 2 weeks after radiation, after a 10-minute thermal annealing, after a 1.5- hour annealing and after a 17-hour total annealing. We continued measuring IV two more times but with the Newport IV setup, which was easier to use and had a lamp that simulates better the sun spectrum. A timeline of the electrical characterization we performed to the CZTS samples, is shown in Figure 4.2.

CZTS1

Before radiation Week 1 Week 2 Week 3

IV, QE IV every day IV every day Radiation

QE QE

IV after each annealing step

IV after each annealing step

Week 4 Week 5 Week 6

IV x2 IV

CZTS2

Before radiation Week 1 Week 2 Week 3

IV, TAS, CV, DLCP

Radiation

anneal 10 min

Week 4 Week 5 Week 6

Newport IV

* * * * *

*: IV, TAS CV, DLCP

anneal 17 h anneal

1.5 h Newport IV

Figure 4.2.: Timeline for CZTS samples.

For CIGS1 we performed IV before the radiation, and after the radiation we measured IV every day for 2 weeks. QE was measured before the radiation, after it, and 2 weeks after it. The annealing process started one month after that, and was done in the same way as in CZTS1.

For CIGS2 we started by measuring IV, TAS, CV and DLCP before radiation, after radiation, and 2 weeks after radiation. The investigation was stopped after this point because we didn’t see any improvement. Furthermore, the admittance measurements did not give clear results even before the radiation. Besides that, during the measurement after the radiation, we faced some problem with the LCR meter and our data were not reliable. Due to the problem with the LCR meter we could not trust either of the methods to get the doping after the radiation. Therefore, our analysis for CIGS was based only on the IV curves of CIGS2 and the rest of the measurements will not be discussed. A timeline of the electrical characterization we performed on them, is shown in Figure 4.3.

CIGS1

Before radiation Week 1 Week 2 Week 14

IV, QE IV every day IV every day Radiation

QE QE

IV after each annealing step

IV after each annealing step

Week 15 Week 16

IV x2

CIGS2

Before radiation Week 1 Week 2 IV, TAS,

CV, DLCP Radiation

* *

*: IV, TAS, CV, DLCP

Figure 4.3.: Timeline for CIGS samples.

4.2. Equipment

In order to electrically characterize our solar cells as thoroughly as possible, we used different characterization methods described below.

4.2.1. IV

In order to have a fair comparison of cells, the cell parameters should be measured in the same conditions. The standard test conditions (STC) is a standard according to which all solar cells and panels should be measured at temperature T = 25^◦C, solar irradiance P_in = 1000W/m² and an air mass 1.5 (AM1.5) spectrum. The last condition is the spectrum that corresponds to the spectrum of the sun, when light has passed through 1.5 atmospheres thickness.

The basic IV parameters (J_sc, V_oc, FF and η) are easy to extract based on Figure 3.4.

The rest of the parameters, however, need more effort and are extracted gradually by the one diode equation (Equation 3.11) [20]. First, a plot of the derivative g(V ) = dJ/dV against V in the neighborhood of J_sc, gives the shunt conductance and therefore the shunt resistance. Secondly, the intercept of the plot of r(J ) = dV /dJ against (J + J_sc)⁻¹ gives the series resistance and the slope gives the ideality factor. The last step is a semilogarithmic plot of (J + J_sc) against (V − RJ ), the intercept of which gives the saturation current density. The slope of the same plot is the ideality factor, so a comparison of the two values is possible. Writing the codes needed to extract the parameters, was also part of the project.

Two different setups were used to measure IV. The first will be denoted as Newport IV and the equipment it included were: a Keithley 2401 multimeter and a Newport Sol2A Class ABA Solar Simulator. The second will be denoted as ICVT and the equipment it included were: an Agilent 4284A LCR meter and a Keithley 2401 multimeter.

4.2.2. QE

Quantum efficiency measurements are used to extract an estimation of the band gap of the material [21]. Quantum efficiency is given by:

QE = 1 − exp (αW )

αL + 1 (4.1)

where α is the absorption coefficient, L is the diffusion length and W is the depletion region width. Close to band gap, where the absorption is weak, the denominator can be neglected. By Taylor expansion of the equation QE = 1 − exp (αW ), we get that QE(λ) ∝ α(λ), where λ is the wavelength. The absorption coefficient for direct band gap materials is α(λ) ∝ pE − E_g therefore QE(λ) ∝ pE − E_g. Plotting QE² against energy around the band gap and extrapolating to the x-axis, yields an estimation of the band gap.

The equipment we used to measure QE were: an Oriel Apex Monochromator Illumi- nator, a Cornerstone^TM 130 1/8m Holographic Grating Monochromator, a Stanford Re- search system SR570 low-noise current preamplifier, a Stanford Research system SR810 Lock-in amplifier and a Stanford Research system SR830 Lock-in amplifier.

4.2.3. CV

The Capacitance-Voltage measurements were used to determine the doping of our sam- ples [22], [23]. In the interest of analyzing our data, we had to accept the depletion approximation, which is not accurate for thin film cells. According to this approxima- tion, the depletion region is defined precisely, ends abruptly, and is fully depleted of free carriers. In this case, the capacitance that comes from the depletion edges is given by:

C = εε₀A

W (4.2)

where ε is the semiconductor relative permittivity, ε0 is the vacuum permittivity, A is the cell’s area, and W is the width of the depletion region. For an one-sided (Schottky) diode the width is:

W = s

2εε₀(V_bi− V_dc)

qN_B (4.3)

where Vbi is the built-in voltage, Vdc is the applied voltage, q is the electron charge, and N_B is the doping concentration. Combining these equations, we get that the doping is:

N_CV = − 2 qεε₀A²

 dC⁻² dV

⁻¹

(4.4) In the depletion approximation, the capacitance only comes from the depletion region, so the previous equation is correct even when the doping is not constant across the semiconductor. The position in this case is:

x_CV = εε₀A

C (4.5)

The equipment we used to measure CV were: an Agilent 4284A LCR meter and a Keithley 2401 multimeter.

4.2.4. DLCP

Drive level capacitance profiling is another method to determine the doping [24]. The applied ac voltage has varied amplitude and the capacitance is:

C = C₀+ C₁dV + C₂dV²+ . . . (4.6)

The doping density is given by:

N_DLCP = − C₀³

2qεε₀A²C₁ (4.7)

and the position is x = εε₀A C₀ .

If the sample has deep traps, DLCP gives more accurate results at high frequencies, since the depletion region approximation assumed for CV is no longer valid. At low frequencies, where the traps have time to respond to the applied ac voltage, the two measurements should give similar results.

The equipment we used to measure DLCP were: an Agilent 4284A LCR meter and a Keithley 2401 multimeter.

4.2.5. TAS

The temperature dependent admittance spectroscopy measurement as a function of fre- quency is another helpful measurement, since it can yield the thickness of the film, the position of the Fermi level and the density of defect states. However, it can only detect defects between the edge and mid-gap.

The frequency range of the applied voltage is chosen so that it crosses the transition frequency of trapped states. When the temperature is low or the frequency high, carriers do not have time to respond to the frequent voltage change, so they do not shift in and out of the depletion edge. As the temperature increases or frequency decreases, trap states respond to the applied voltage, creating a cut-off step, characteristic for every certain (T, f ) point. The derivative of capacitance will have a peak in the capacitance step.

Plotting the frequency of the peak and the temperature in an Arrhenius plot returns a straight line, the slope of which is the activation energy and the intercept is the attempt- to-escape frequency. The code required to extract these information was written as part of the project. The temperature range was from 85K to 305K, and the frequency range was from 10² to 10⁶ Hz.

The equipment we used to measure TAS were: an Agilent 4284A LCR meter, a Keithley 2401 multimeter and a LakeShore 325 Temperature controller. Liquid nitrogen was used to cool the sample.

4.3. Modeling

In order to understand the changes that radiation caused to our solar cells, we created a model of our devices before radiation and compared it with our experimental results for CZTS1. After that, we tried to modify our model to fit the experimental data that we obtained after radiation. In this way we could get an indication of which parameters were affected by the radiation and resulted in the degradation of the cells. The model was built with the Solar Cell Capacitance Simulator (SCAPS-1D) [25] version 3.2.01 and was based on the model developed by Frisk et al [26].

SCAPS 1D is a free numerical tool that can simulate thin film solar cells with up to seven layers plus their interfaces and contacts. For the derivation of the model, it uses the Poisson equation (Equation 4.8a, relationship between charge and electric field) and con- tinuity equations (Equation 4.8b and Equation 4.8c, tracking of carriers in terms of move- ment, generation, and recombination), taking into account the current density equations

(Equation 4.8d and Equation 4.8e). The equations are solved in their one-dimensional form. It can simulate current-voltage, capacitance-voltage, capacitance-frequency and quantum efficiency.

The input parameters required for each layer are the ones in Equation 4.8. Furthermore, the user can define absorption, reflection and recombination. Optical filters at the front and back contact can also be added. Lastly, there is the possibility to specify illumination, temperature, resistances, voltage working point and many other parameters.

dξ dx = q

ε(p − n + N_D − N_A) 1

q dJ_e

dx = U − G 1

q dJ_h

dx = −(U − G) J_e= qµ_enξ + qD_edn

dx J_h = qµ_hpξ + qD_hdp dx

(4.8a) (4.8b) (4.8c) (4.8d) (4.8e) The equations are combined with boundary conditions at the interfaces between the materials and at the contacts, producing a system of coupled differential equations. The structure is discretized creating a mesh with more points close to interfaces and contacts, and fewer points in areas where properties are fairly constant. For the given structure, the basic equations are solved by a combination of a Gummel iteration scheme and Newton- Raphson, and steady states are calculated. Small signal analysis is used for capacitance measurements.

5.1. Electrical Characterization

5.1.1. CZTS

IV

The two different CZTS samples (CZTS1 and CZTS2) had similar IV parameters before the radiation. For CZTS1 we used only the Newport IV setup and the curves are shown in Figure 5.1. We can see the IV curves of CZTS2 with ICVT in Figure 5.2 and with Newport IV in Figure 5.3. Their IV parameters before the radiation are shown in Table 5.1.

−0.4 −0.2 0.0 0.2 0.4 0.6 0.8

Voltage (V) (40

(20 0 20 40 60

Current (

mA cm²⁾

IV curve with Newp rt IV in sample CZTS1

After radiati n

After 10 minutes anneal After 17 h urs anneal 2 weeks after radiati n Bef re radiati n

Figure 5.1.: IV curve

−0.4 −0.2 0.0 0.2 0.4 0.6 0.8 Voltage (V)

−20 0 20 40 60

Current (

mA cm²⁾

IV curve with ICVT in sam le CZTS2

After radiation

After 10 minutes anneal After 17 hours anneal 2 weeks after radiation Before radiation

Figure 5.2.: IV curve with ICVT

−0.4 −0.2 0.0 0.2 0.4 0.6 0.8

Voltage (V) (20

0 20 40 60

Current (

mA cm²⁾

IV curve with Newp rt IV in sample CZTS2

After 10 minutes anneal After 17 h urs anneal 2 weeks after radiati n Bef re radiati n

Figure 5.3.: IV curve with Newport IV

A world record CZTS cell with 12.6% efficiency was reported in 2014 [27], but our samples were far from that. Before the radiation, we had a sample with 7.55% efficiency (CZTS1) and one with 7.28% (CZTS2). We need to notice at this point the difference between the measurements in the two devices (ICVT and Newport IV) for CZTS2. ICVT gives less reliable results than Newport IV, since the position of the sample relative to