UPTEC K 18018
Examensarbete 30 hp Juni 2018
Compositional gradients in sputtered thin CIGS photovoltaic films
Daniel Boman
Teknisk- naturvetenskaplig fakultet UTH-enheten
Besöksadress:
Ångströmlaboratoriet Lägerhyddsvägen 1 Hus 4, Plan 0
Postadress:
Box 536 751 21 Uppsala
Telefon:
018 – 471 30 03
Telefax:
018 – 471 30 00
Hemsida:
http://www.teknat.uu.se/student
Abstract
Compositional gradients in sputtered thin CIGS photovoltaic films
Daniel Boman
Cu(In,Ga)Se2 (CIGS) is a semiconductor material and the basis of the promising thin-film photovoltaic technology with the same name. The CIGS film has a typical thickness of 1-2 mm, and solar cells based on CIGS technology has recently reached efficiencies of 23.3%. Ultra-thin CIGS solar cells use sub-micrometer thick films that require significantly less material and can be manufactured in a shorter amount of time than films with typical thicknesses. With decreasing thickness, both electrical and optical losses get more significant and lower the overall performance. Electrical losses can be decreased by increasing the overall film quality and by utilising a graded band gap throughout the CIGS layer. The band gap can be changed by varying the [Ga]/([Ga]+[In]) (GGI) ratio. Higher overall film quality and a higher band-gap towards the back of the absorber are expected to increase the performance.
In this work, sputtered CIGS solar cells were made with different CIGS layer thicknesses, that ranged between 550-950 nm. Increased heat during deposition was examined and shown to increase the film quality and performance for all thicknesses.
Two different ways of doping CIGS with Na was examined and it was found that higher Na content lead to an increasing predominance of the (112) plane. The band gap was graded by varying the GGI composition throughout the CIGS layer and depth profiles were made with Glow-Discharge Optical Emission Spectroscopy (GDOES). It was found that a sputtered CuGaSe2 (CGS)layer below the CIGS-layer lead to a steep increase of the GGI near the back contact. When CGS made up 10% of the total CIGS layer thickness, a significant increase in performance was observed for all thicknesses. CIGS-absorbers with a less graded region with low GGI, making up 30%
or 60% of the total CIGS layer thickness were made. A decrease in GGI in that region, was shown to increase the current but lower the voltage. No substantial increase in total performance compared to a fully graded CIGS layer was seen regardless of layer thickness. For further work the optical losses needs to be addressed and work on increasing the optical path in the CIGS layer needs to be done.
ISSN: 1650-8297, UPTEC K 18018 Examinator: Peter Broqvist Ämnesgranskare: Tobias Törndahl
Handledare: Esko Niemi, Charlotte Platzer Björkman
Populärvetenskaplig sammanfattning
Att det idag finns ett allt större klimathot råder det inget tvivel om. En omställning mot en mer hållbar energiförsörjning är nödvändig för vårt klimat och vår civilisations skull [1]. Solenergi är en av de alter- nativa energikällorna som växer snabbast och som också har störst potential. På alla ytor där solen lyser finns potentialen att placera en solpanel som i upp till 25-30 år kan leverera energi med minimalt underhåll.
Majoriteten av alla solpaneler idag är gjorda av kisel, s.k kiselmoduler. Det är de billigaste modulerna och som också produceras i störst kvantiteter. Den stora nackdelen med kiselmoduler är dock att de är tunga och sköra. Det innebär att de är bra att placera på stora, öppna och relativt plana ytor i form av solparker. Ofta är dessa parker placerade där människor inte bor och konkurrerar med annan verksamhet så som jordbruk.
Framför allt tak, men även fasader däremot är stora, öppna och annars oanvända ytor som finns där män- niskor faktiskt bor. För att kunna utrusta befintliga byggnader med kiselmoduler, så kräver det ofta stora ingrepp och i vissa fall förstärkning av tak och fasader. Estetiken är inte tilltalande för många, och det kan vara svårt att få dem att passa in i en befintlig stadsbild.
En av de alternativa teknologierna som idag finns bygger på materialet CuInxGa(1−x)Se2(CIGS). Det kallas även CIGS-teknologi, och är en så kallad tunnfilmsteknologi där det aktiva materialet bara utgör c:a 1-2.5 μm. Jämfört med kisel, som är 200 - 500 μm, är det endast en bråkdel av tjockleken. CIGS kan göras till flexibla och lätta moduler som även dem bara är en bråkdel så tjocka som kiselmoduler. Det gör att de enkelt kan integreras i befintliga tak och på fasader på ett sådant sätt att de smälter in i byggnaden.
CIGS-moduler idag ändå dyrare än motsvarande för kisel. Råmaterialens priser fluktuerar mer än av kisel, effektiviteten på modulerna är lägre och tillverkningskostnaderna är högre [2]. En del av lösningen kan vara att göra skikten ännu tunnare, s.k ultra-tunna, då utgör det aktiva materialet <1 μm. De resulterande solcellerna kräver då mindre material och fler moduler kan produceras på kortare tid. Problemet är dock att när skikten görs ännu tunnare uppstår ökade förluster under användning i form av mindre infångat ljus, men också elektriska förluster till följd av förändringar i materialet. Dessa elektriska förluster uppstår främst till följd av att en större del av ljusabsorptionen sker närmre den bakre kontakten. Det leder i sin tur till att mer av den infångade energin blir till värme istället för elektrisk energi.
CIGS kan tillverkas med flera olika metoder, i det här arbetet har sputtring använts. Där utgår man från ett pulver som under högt tryck och temperatur pressats samman till en kompakt skiva. Beroende på vilken tunnfilm man vill belägga, så pressas olika pulver (Mo, CIGS, ZnO etc). Skivan placeras mitt emot en bit stålplåt som agerar som strukturell komponent då solcellen i sig är för tunn för att klara sig på egen hand.
Vakuum pumpas i utrymmet och fylls sedan med argon. En pulserande spänning läggs mellan skivan och plåten vilket joniserar argonatomerna och får dem att slå ut atom för atom från skivan, som sen fastnar på stålplåten. Efter c:a 10-30 sekunder av denna process har en heltäckande film, som är 20-100 nm tjock, bildats på plåten. Detta utförs i en sekvens från ett antal skivor med olika sammansättning för att få den önskade lagerstrukturen som sen utgör själva solcellen.
Arbetet har gjorts tillsammans med Midsummer AB i Järfälla, Stockholm. De tillverkar kompletta turn-key system för att tillverka flexibla CIGS-moduler. Deras huvudsakliga produkt är ett sputtersystem som heter DUO. Det är ett högvakuumsystem som innefattar 25 st sputterstationer som belägger framför allt tunna plåtskivor av rostfritt stål. I det här projektet har alla celler producerats utav ett av dessa verktyg.
I det här arbetet har det på ett par olika sätt undersökts hur de elektriska förlusterna kan minskas för olika
ska det leda till ett bättre material med mindre förluster, vilket även visade sig stämma. Effekten av att ha små mängder natrium i CIGS-skiktet har även undersökts och där det observerades att filmkvalitén ökade med mängden natrium, något som kan leda till mindre förluster.
I CIGS kan man variera mängden Ga i förhållande till In på olika djup i skiktet, vilket påverkar både de elek- triska och optiska egenskaperna. Man skapar en så kallad gradient. I det här arbetet har det undersökts om en viss gradient kan ge upphov till minskade förluster. Det observerades att ett väldigt Ga-rikt område nära den bakre delen av solcellen minskade förlusterna och ökade prestandan. Det observerades också att genom att ha en mer plan gradient av Ga genom skiktet och en i genomsnitt låg nivå, kunde man öka strömmen utan att förlora för mycket av spänningen. Den motsatta effekten erhölls dock om den plana delen utgjorde för stor del av skiktet.
Effekterna som observerades för samtliga av de olika variationerna av gradienter visade sig vara likadana för alla tjocklekar och inte bara för ultra-tunna. Effektiviteten för de ultra-tunna kunde höjas med 1-2%, men trots det så är skillnaden i prestanda jämfört med normal tjocklek fortfarande stor. För fortsatt arbete rekommenderas det att rikta uppmärksamheten mot den optiska aspekten av förlusterna. Det innebär att försöka förlänga vägen som ljuset färdas genom solcellen genom att ha ett reflekterande skikt i bakre delen av solcellen, samt att försöka sprida det inkommande ljuset.
Acronyms and Abbreviations
Absorber The CIGS layer in the solar cells
CB Conduction band
CGI [Cu]/([Ga]/[In]) ratio
CGS CuGaSe2
CIGS CuInxGa(1−x)Se2
CIS CuInSe2
GGI [Ga]/([Ga]/[In]) ratio Isc Short circuit current SCR Space charge region
TCO Transparent conductive oxide
VB Valence band
Voc Open circuit voltage
Contents
1 Introduction 1
2 Aim 2
3 Theory 3
3.1 Solar cell principle . . . 3
3.2 CIGS . . . 4
3.3 CIGS solar cell structure . . . 5
3.4 Ga-grading . . . 6
3.5 Ultra-thin CIGS . . . 7
3.6 Sputtering . . . 8
4 Experimental 9 4.1 Deposition of thin film layers . . . 9
4.1.1 Heater power . . . 10
4.1.2 Na doping . . . 11
4.1.3 CGS layer . . . 11
4.1.4 Grading plateau . . . 12
4.2 Analytical equipment . . . 12
4.2.1 Glow discharge optical emission spectroscopy . . . 12
4.2.2 IV-measurement . . . 13
4.2.3 Quantum efficiency (QE) . . . 14
4.2.4 Scanning electron microscopy (SEM) . . . 14
4.2.5 X-ray diffraction (XRD) . . . 14
5 Results and discussion 15 5.1 Influence of varying heater power . . . 15
5.2 Influence on morphology by Na . . . 18
5.3 Effect on Ga-grading in thin CIGS . . . 20
5.3.1 Varying back surface field . . . 20
5.3.2 Varying grading plateau . . . 24
6 Conclusion 27
7 Acknowledgements 28
1. Introduction
With an increasing climate threat, the need for clean energy has never been so great. Solar cells offers a viable option to produce clean and sustainable energy on almost any surface exposed to the sun. Progress made in the last decades have made solar cells more efficient as well as cheaper and have begun to reach a level where it can compete with fossil alternatives. Still there is much work to be done, currently more than 60% of the worlds electricity comes from fossil fuels and an energy revolution is needed [3].
Today the global market of photovoltaics (PV) is dominated by mono- and polycrystalline silicon solar cells, with over 90% of the PV-market [3]. The rapidly decreasing production costs for this kind of solar cells and modules has led to a point now where on some places, solar power is cheaper than coal power [4]. Due to their low price and high efficiency, they are ideal to be placed in solar parks. Unfortunately, silicon modules does not offer a universal solution and have some limitations. These mostly revolve around the fact that silicon solar modules are both rigid, fragile and heavy. Solar parks also require large open spaces and to avoid having to transport the electricity over large distances, they need to be placed close to were people live. Unfortunately this then means that they have to compete with other businesses such as farming. It is often difficult to equip buildings with these panels and the aesthetics are not always appreciated.
One of the areas that are expected to grow the most in the coming years is building integrated photovoltaics (BIPV) where modules are placed on roofs and facades of buildings. This allows for the energy to be created close to where it is used, and especially roofs are otherwise large and unused spaces that face the sun. In those applications, in order to install them on existing buildings without the need for any modification or additional structural requirements, properties such as weight and flexibility are highly valued.
CIGS is an alternative photovoltaic thin-film technology based on the material CuInxGa(1−x)Se2. Thereby, the abbreviation CIGS. In recent times it has reached an efficiency of 23.3% [5] and offers modules that are flexible, thinner and significantly lighter. These modules are only a few millimetres thick and can be placed on any building, regardless of shape, strength or location. Unfortunately the price is still higher than that of silicon modules, but in the coming years it is expected for it to be able to compete in some applications [3][6].
Progress is still needed in the CIGS field though, both in terms of efficiency as well as in price. One could reduce the thickness of the CIGS layer. Unfortunately when making the CIGS layer thinner, increased op- tical and electrical losses occur. This is due to decreased absorption of photons, decreased film quality and increased recombination near the back contact. This results in a decreased performance of the final solar cell. Work has been done that show that some of the electrical losses may be recovered by increasing the film quality and/or by having a compositional gradient throughout the cell. This can ultimately lead to more efficient ultra-thin solar cells, that require less material than today, but still can be made in to competitive solar modules.
Midsummer AB is a company located in Stockholm that produce complete equipment for the manufacturing of CIGS solar cells. The technology revolves around sputtering multiple layers inside a high-vacuum cham- ber. A decrease in cell thickness would mean that the cycle time for each solar cell would be decreased, thereby inceasing the production capacity, and also less raw material being used. That would allow their customers to produce more cost efficient solar modules that can more easily compete on the market.
2. Aim
The overall aim of this study was to explore how to reduce the losses in ultra-thin CIGS layers. The losses can be divided into electrical losses and optical losses, where the electrical ones was the focus of this study.
The aims of the study were:
• Explore how the compositional gradients varies as a function of thickness and supplied heat for a given deposition process. Possible smearing of a gradient at high temperature may occur, and more discrete steps might be seen at low temperatures. The grading might also be different for different thicknesses due to reduced deposition time.
• Increase the overall film quality for ultra-thin CIGS layers by increasing the temperature for a given process.
• Explore the influence of Na and examine see whether different ways of supplying Na could affect the crystallinity of the CIGS.
• Create and evaluate different Ga-gradients from two different aspects. This was to see how different gradients at different thicknesses affect the performance of the solar cells.
1. Creating a back surface field by having a steep increase in [Ga]/([Ga]+[In]) (GGI) near the back contact. This may be more beneficial for ultra-thin absorbers due to more recombination occurring near the back contact.
2. Creating a flat non-graded region (plateau) within the absorber with a low band gap by sputtering a larger part of the absorber from a single target. It was suggested that this might increase the photocurrent generation without losing too much of the open circuit voltage (Voc).
3. Theory
3.1 Solar cell principle
The working principle of a solar cell builds upon the photovoltaic effect, which is the creation of additional charge carriers in a material due to the absorption of photons. An incident photon of appropriate energy is absorbed by the material. This leads to the excitation of an electron and puts it into a higher energy state.
The electron can then be lead through an external circuit, and be utilized to power almost any electrical device. More specifically, the type of material required is known as a semiconductor. The defining property is that there exists an energy gap, commonly known as the band gap between the valence band (VB) and the conduction band (CB) in the material. The CB lies higher in energy than the VB and between these bands no energy states exist.
If an incident photon with energy equal to or larger than the size of the band gap is absorbed by the material, an electron-hole-pair is created and the electron is excited from the VB to the CB. This raises the energy of the electron by an amount equal to that of the photons. It leaves behind a so called hole in the VB that can be regarded as a positively charged pseudo-particle.
In silicon solar cells, two different Si-layers are joined together. One of them is said to be n-doped, which has an excess of electrons, and the other is p-doped with an excess of holes. At the interface between these two layers a pn-junction is formed (see fig 3.1). The principle in CIGS solar cells is slightly different, where a CIGS layer is combined with a layer made from one of several possible materials to form this junction.
Regardless of the solar cell technology, a region called the space charge region (SCR) is formed due to the difference in concentration of electrons/holes in the two layers. As a consequence of this, an electrical field is formed which asserts a force on electrons towards the n-doped side, and a force on the holes towards the p-doped side. If a photon is absorbed by the material near this region and an electron-hole-pair is created, the electrical field will act to separate the two charge carriers. This physically forces them towards different ends of the solar cell, where they can ultimately be collected by metallic contacts. This accumulation of differently charged carriers at two distinct physical regions of the solar cell creates the positive and negative pole of the solar cell. Without the pn-junction and its resulting effects, it would be nearly impossible to extract any energy from the absorbed photons. The electrons and holes would have no driving force to take different routes and would simply cancel each other out (known as recombination) within the material, effectively releasing the absorbed energy [7].
Figure 3.1: PN-junction in a solar cell with a physical junction depicted on top. The negative charges are p- doped material with an excess of electrons. Positive charges are n-doped material with a deficit of electrons.
Lower part shows the band structure of the junction and the charge carrier generation.
3.2 CIGS
CuInxGa(1−x)Se2 (CIGS) is a I − III − V I2 type of semiconductor with a tetragonal chalcopyrite crystal structure. It has a band gap (Eg) that can be continuously varied by substituting In with Ga while still retaining the same crystal structure. At the edge of the compositional variations there are CuInSe2 (CIS) where Eg = 1.02 eV and CuGaSe2(CGS) with Eg = 1.69 eV [8]. This allows for the band gap to be tuned to match the solar spectrum, as well as having a gradually changing band gap throughout the structure.
Non-graded CIGS has been found to have optimal performance at around 20-30% Ga with a band gap of 1.2eV [9]. One of the central properties of CIGS is that the optoelectronic performance is fairly insensitive to variations in the atomic composition. This is very beneficial for large scale production where very precise process control can be difficult or expensive [10]. Incorporation of Na (and other alkali-metals) has also been shown to increase the phase field of the chalcopyrite phase and increase the compositional tolerance [10]. CIGS is utilised in thin-film solar cells as the absorber material and acts as the p-type layer in the pn-junction. CIGS is intrinsically p-doped resulting from prevalent Cu-vacancies (VCu) within the material and the average carrier concentration is 1015− 1016cm3[11]. CIGS has a direct band gap and an absorption coefficient of >3x104cm−1for >1.3eV photons, much higher than that of those materials with indirect band gap (i.e silicon) [10]. This allows for the absorber being much thinner than otherwise, while still absorbing the same amount of photons. For example, 95% of the incident photons from solar illumination are absorbed with an absorber of only 1 μm [10]. For silicon the corresponding thickness is in the range of a few hundred μm. This contributes to making it a promising material for usage in large scale thin-film solar cells where thin cells result in cheaper and lighter panels.
3.3 CIGS solar cell structure
To make a complete CIGS solar cell, several layers besides the CIGS absorber is needed. In figure 3.3, starting from the bottom, below the first actual layer is the main structural component, the substrate. Most commonly glass is used for this, but to be able to make flexible modules, steel or polymers is used. Mid- summer uses a 0.15mm thick ASTM 430 stainless steel substrate with the shape of a 6" square with rounded corners. Steel is primarily used for its structural and flexible properties but is is also a decent conductor and acts as the cathode in the final cell. Due to the properties of steel at high temperatures, a diffusion barrier is needed to prevent diffusion of Fe and Cr. These will otherwise diffuse into the absorber and lower the performance drastically.
A molybdenum layer is the bottommost layer that the CIGS absorber comes in contact with. It is considered to be the rear contact of the CIGS solar cell, even though the physical contact of the cell is soldered to the steel below. The molybdenum layer reacts with the Se (in CIGS or present in the reaction chamber) and MoSe2 is formed. This compound has been shown to be beneficial in creating a good ohmic contact with the CIGS [12]. This is due to MoSe2 being a p-type semiconductor with a band gap of 1.3eV. In a standard, non-graded cell, it acts as an electronic mirror for the electrons, while at the same time providing a low-resistance contact for the holes [9]. MoSe2is a layered compound though, so careful control over the process as well as a high temperature (>550◦C), is needed to make sure the layers are perpendicular to the substrate, avoid adhesion problems and increase the electronic properties [10].
The actual CIGS layer makes up the majority of all the deposited layers with regard to thickness, most of the charge carriers are generated here. The average thickness of this layer has traditionally been in the range of 1 - 2.5 μm [13].
On top of the absorber is a In2S3 based buffer layer that improves the electronic interface between the ab- sorber and the layers above.
Above the buffer is an intrinsic ZnO based layer deposited to act as a high-resistive window layer, mainly to avoid shunting and pin-holes [10]. On the top is the contact layer, consisting of a transparent conductive oxide (TCO) which conducts the electrons out of the cell. The electrons are transported to a silver grid, which is printed on top of the cell in a separate step. For an in depth explanation of the structure see [10].
Figure 3.2: Cell structure and different layers
3.4 Ga-grading
By altering the amount of Ga in relation to In, the [Ga]/([Ga]+[In]) ratio (GGI) is changed. Ideally this is done without changing the [Cu]/([Ga]+[In]) ratio (CGI). The change in GGI results in the band gap also being changed. That occurs mainly through a change in the conduction band minima, and the band gap can be varied between 1.02 eV (pure CIS) and 1.68 eV (pure CGS). The band gap changes according to Eg[eV ] = 1.04 + 0.67 · x + b · x(x − 1), where x is the GGI, and b is an optical bowing coefficient that varies between 0.11 and 0.24 [8]. If the GGI is continuously changed throughout the CIGS layer, so will also the band gap, and a band gap grading will be formed (see figure 3.4). Most commonly a higher band gap is present towards the back, and a lower towards the front, called normal grading. This is will be handled in this study. However, it should be mentioned that double grading can also be made. This is when a notch near the front is made so that the minimum band gap is somewhere near the front but not at the interface.
Normal grading has mainly two consequences on the electrical properties of the CIGS material. The first
Figure 3.3: Band edge diagram of a CIGS absorber with normal grading. Upper dotted line show the conduction band minimum for a normal grading. Valence band remains unchanged. ξAshow the electrical field obtained.
one is a higher band gap towards the back contact which results in less probability of recombination. The second one is that an additional electrical field is created that is proportional to the slope of the GGI. In the case of a normal grading the electrical field is directed towards the front. These two altered electrical properties then together in theory lead to both increased Vocand short circuit current (Isc).
The increased Vocis due to decreased recombination near the back contact where the recombination velocity is relatively high [14].
The second effect of normal grading is an improved Iscdue to an additional carrier collection. The average time before an excited electron in the CB recombines with a hole in the VB is called the minority carrier lifetime. On average this is the time an electron has to reach the SCR before recombining. The additional electrical field throughout the absorber creates a force that acts on the excited electrons, causing them to drift towards the front. In non graded absorbers the act of getting photoelectrons to the SCR relies completely on diffusion. The force of the electrical field increases the probability of the electron to reach the SCR during one minority carrier lifetime. On average, the number of collected electrons increase, which leads to an increased current. The additional length that an electron can move due to this additional force, within the material during one minority carrier lifetime, can then easily increase to where it is longer than the actual absorber. This effect is likely to be more important in sputtered films than in evaporated where the overall film quality tends to be lower and the minority carrier lifetime is expected to be shorter.
Raising the band gap towards the back may also results in a decrease in the collection of low energy photons.
This since the absorption probability is inversely proportional to the band gap [15]. This in turn should then lead to a decrease in Isc. However because the photo generation of carriers near the back in CIGS devices with regular thickness is quite small, the net effect on most devices is still positive. This could be a problem for thinner CIGS devices where the absorption of light near the back contact contributes more to the photo generated current. Depending on the height and slope and of the gradient, the net improvement might therefore be quenched [16].
In sputtered CIGS absorbers at Midsummer, a gradient is achieved by sputtering a cell from several different
quaternary CIGS targets with different GGI. This however creates the possibility that the resulting gradient has discrete steps rather than a desired continuous. The idea is that these discrete steps are smoothened out due to high diffusion. However it is uncertain if this effect only is valid at certain temperatures and/or thicknesses. And especially for ultra-thin absorbers it is unclear how the resulting gradient will be.
3.5 Ultra-thin CIGS
Even though CIGS is considered to be a thin-film technology compared to silicon, there is still a strive to- wards making it even thinner. Standard devices of today are usually between 1 - 2.5 μm thick CIGS layer, however standard devices at Midsummer lies in the range of 0.9 - 1μm. That already puts them close to ultra-thin. Because of the potential increase in cycle time and reduction in use of raw materials, an even further decrease is desired. However with decreasing thickness there are more challenges.
The main challenge is the absorption of light. At 1 μm thick, approximately 95% of the incident solar illumi- nation is absorbed [10]. Below this value, inevitable optical losses will occur regardless of the film quality.
Light will be absorbed by the Mo back contact without creating any photocurrent. This loss mainly affects the low energy photons of the solar spectrum, since they are most likely to be absorbed deep in the absorber due to their lower absorption probability [15]. This is also something that previously has been confirmed by QE-measurements on ultra-thin films [17].
To combat this, two strategies, one optical and one electrical are pursued to improve device performance.
The electrical, which is the focus of this study is coupled to the fact that recombination at the back-contact will be more significant. Therefore the importance of current collection throughout the device increases.
That is were band gap grading, increased crystallinity and improved passivation of grain boundaries might be beneficial. Increased temperature during deposition as well as different types of Na incorporation might lead to higher crystallinity and improved film quality [18].
Other methods of preventing recombination at the back contact revolves around a thin nano-structured passi- vated rear surface layer of Al2O3or equivalent [19]. This however is still very experimental and still difficult to implement in a production environment.
The optical aspect revolves around creating an optically reflective back contact. This allows for the unab- sorbed light to instead of being parasitically absorbed by the back contact, be reflected and travel through the absorber an extra time. This increases the effective path in which the light travels through the absorber, and increases the likelihood of it generating photocurrent. The previously mentioned Al2O3is one candidate for this kind of back contact, but using ZrN or a TCO as back contact has also been suggested [20] [21].
However, yet of today these back contacts have not been successfully implemented in the current process due to the highly reactive environment.
Another option is to add a layer on top of the cells, of so called a superstrate, that through a randomly tex- tured surface, scatters the incoming light and increases the optical path through the solar cell. This can be done either in the form of laminate on module scale, surface nano particles, or as a textured TCO [22] [23].
It is also possible to combine the two latter techniques and achieve light trapping. However implementing any of this in the current process presents new challenges.
3.6 Sputtering
Sputtering is a thin-film deposition process commonly used in various industries. It is the process used for all deposited layers in this study. A typical sputtering chamber is depicted in figure 3.4.
Fundamentally, sputtering is the ejection of atoms from a target that is bombarded by energetic particles.
The target is composed of the same material as that of the desired film. The energetic particles are most commonly ions made from Argon, but noble gases are also possible to use. These ions can be produced by either using an ion gun or a plasma. In this case only plasma will be considered. Plasma can be described a gas that has a high amount of ions and free electrons. This causes it to be electrically conductive and respond to electromagnetic forces.
The sputtering process requires a process chamber to be pumped down to a pressure of typically 10−4Pa [24]. Argon (or any other noble gas) is then introduced into the chamber, raising the pressure to 1-10 Pa depending on the process [24]. Applying a high enough potential between the substrate and the target causes gas ions and electrons present in the gas to accelerate and collide with neutral gas atoms. If the particles have high enough energy, the neutral atoms will get ionised too and start to accelerate in the electrical field.
This causes a chain reaction and the plasma is "ignited" where it also produces a visible glow.
Some of the accelerated particles will hit the target. If they have enough energy to overcome the surface energy, atoms from the surface will be ejected. These mostly neutral target atoms will then travel across the chamber and be deposited on the substrate. If the target has a low conductivity it is possible to apply a pulsed-DC or RF-potential instead of the standard DC-potential.
To decrease both the required gas pressure in the chamber and increase the efficiency of the sputtering process, magnets can be placed behind the target. This is called magnetron sputtering and with a proper magnetic setup, a magnetic field is created so that free electrons are trapped near the target surface.
If the desired film composition for different reasons can not be acquired simply by sputtering from a solid target, a reactive gas can also be supplied. By supplying a reactive gas in the chamber together with the noble gas, sputtered atoms can react with this gas and create new compounds. This is called reactive sputtering [24]. A more detailed description of sputtering can be found in [24].
Figure 3.4: Schematic representation of a magnetron sputtering chamber
4. Experimental
4.1 Deposition of thin film layers
All of the cells in this work have been produced in a DUO tool, designed and manufactured by Midsummer.
The system has a layout as shown in figure 4.1 and consists of 25 individual sputtering stations divided into two main chambers. Each main chamber has a hub with 14 arms, where each arm can hold one substrate.
An intermediate chamber is used to transfer the substrates between the main chambers. The entire system is a closed high vacuum system and all the layers are deposited without breaking the vacuum.
The process begins with a steel substrate being loaded into the system at the LL-position as seen in figure 4.1. The substrate is a piece of 0.15mm thick stainless steel, that has the shape of a 6” square with rounded corners. The substrate is grabbed by one of the 14 arms and is in sequence sputtered at station 1 - 25. The central hub on which the arms are attached to will rotate and move all the arms at the same time. This causes the substate to remain on the same arm until it switches chamber.
Figure 4.1: System overview of the DUO tool [25]
The process of making a cell starts with a diffusion barrier being deposited. On top of this is then the alkali-doped molybdenum layer deposited. The substrate then passes in to the main chamber B through the intermediate chamber. In there at a higher temperature (500-650◦C), the absorber is deposited by reactive sputtering from compound targets at several different stations.
Additional heat is supplied by resistive heating near the substrate, for which the power can be individually adjusted for each arm. By sputtering from targets with different compositions with regards to GGI ratio, a compositional gradient through out the absorber can be achieved. The absorber deposition starts with a layer of CuGaSe2 (CGS) and subsequently CIGS targets with varying GGI-ratio are sputtered to achieve the desired gradient. After the absorber is deposited, the cell is rapidly cooled to below 250◦C and then transferred back into the main chamber A. There a In2S3buffer is deposited followed by window layers and an ITO-based top contact. After the top contact, the cell exits the machine and is finalised in a screen-printer
solid, the cells can then be soldered in strings and put together into flexible modules.
The machine has two different operation modes, an one-by-one mode (OBO) and a production mode. The OBO mode allows for only one cell to be in the machine at a given point in time, and the same arm is used for every cell. Due to some individual variation in the heaters on each arm, this increases the repeatability in experiments but reduces the production rate. The production mode allows for all the arms to be occupied, which means that 30 cells can be in the machine at a given time which gives a higher throughput. This has been shown to be less suitable for research purposes due to less repeatability in small sample series. In all the experiments done in this work, OBO mode is used. To reduce differences between OBO and production mode, a pre-heating step of the substrate with the arm heaters is performed after the Mo-deposition before the CIGS is deposited. This allows the substrate to reach a temperature more similar to one that it would reach during production mode.
Generally when regarding thin-film deposition in an academic environment, the amount deposited is mea- sured in either mass or thickness. However due to the practical difficulties in measuring thickness in a production line and on a run-to-run basis, the thicknesses are here defined as the amount of energy supplied during a sputtering process. This is specified in kJ, and for a given target and process parameters, it scales linearly to the thickness. It should however be noted that it does occur differences between different tar- gets. This means that identical sputter parameters such as pressure and power may result in different sputter yields and thicknesses from different targets. For most purposes though, it does scale linearly enough to be a sufficient measurement of thicknesses.
4.1.1 Heater power
The substrate temperature was altered by varying the power supplied to the heater coils in the arm that holds the substrate. Five different heater powers were supplied during the deposition of four different absorber thicknesses (225, 300, 350 and 400 kJ) according to table 4.1. 8 cells were made for every heater power and thickness (20 different variations). The specified heater power was supplied to the cell during the pre- heating step, and during the entire CIGS deposition. The values are based on the current standard heater power (390 W/584 W) and adjusted to ±10-20%.
A fixed amount of barrier layer was deposited with the same parameters for all cells, and with an identical Mo based back contact on top of it. Na was supplied by sputtering from a doped Mo target (Mo:Na), allow- ing for it to be diffused to the absorber during deposition. CIGS was sputtered with an equal amount from 6 different targets of varying GGI according to figure 4.2. Pressure was at 0.005 mbar and sputtering power was 4 W/cm2. The cells were completed with buffer and top contacts which were deposited with identical parameters for all cells. Finally a silver grid was printed on the cells.
Table 4.1: Heater powers
Heater setting H1 H2 H3 H4 H5
Inner power [W] 312 351 390 429 468 Outer power [W] 467 526 584 642 701
Figure 4.2: Sputtering sequence for CIGS targets with respective target CGI and GGI 4.1.2 Na doping
Six different types of cells where made with different methods of incorporating Na, and with the CIGS deposited at two different temperatures (H1 and H5). The first kind had Na incorporated by sputtering a Na-doped Mo-target (Mo:Na). The second one had Na incorporated by sputtering a NaF-doped CIGS-target (CIGS:NaF). Finally the third one had a combination of the two. All of them had the same total amount of diffusion barrier and Mo deposited as back contact. The absorber consisted of 300 kJ sputtered from a single CIGS-target at 0.005 mbar and 4 W/cm2sputtering power. The cells were finalised with buffer and top contacts which were deposited with the same parameters for all cells.
Heater setting H1 H5
MoNa x - x x - x
CIGS:NaF - x x - x x
CIGS non-doped x - - x - -
Table 4.2: Na doped cells. Each column represents one variation
4.1.3 CGS layer
Four different relative thicknesses of CGS layers (0, 10, 20 and 30% of the total absorber) were made with subsequent sputtering of CIGS on top. Three different total thicknesses of 225, 300 and 400 kJ were made for each of the different CGS layers. These 12 variations were made in one series with 8 cells of each variation. Diffusion barrier, Mo back contact, buffer and top contacts were identically sputtered for all, and the cells were finalised with silver grids. The CIGS was sputtered at 0.005 mbar with a sputtering power of 4 W/cm2and with a heater setting at H3. The sputtering sequence of CGS and CIGS is shown in figure 4.3.
Figure 4.3: Sputtering sequence for CIGS targets with respective target CGI and GGI with varying CGS thickness. X-axis shows the unitless relative depth of the absorber. Y-axis shows the unitless compositional ratio (GGI and CGI).
4.1.4 Grading plateau
Two different plateau lengths (30 or 60% of total absorber thickness) were made from each of three different targets with varying GGI (0.185, 0.222 or 0.259). This was performed for three different thicknesses (225, 300 and 400 kJ), resulting in 18 different variations. Sputtering parameters for the CIGS were 0.005 mbar and 4 W/cm2and the heater setting was H3. 8 cells were made for each variation and the sputtering sequence can be seen in figure 4.4 and 4.5. The cells had the same diffusion barrier and Mo back contact. They were also completed with a buffer and top contact and finalised with a silver grid.
Figure 4.4: Sputtering sequence for CIGS targets with respective target CGI and GGI with 60% plateau of varying GGI. X-axis shows the unitless relative depth of the absorber. Y-axis shows the unitless composi- tional ratio (GGI and CGI).
Figure 4.5: Sputtering sequence for CIGS targets with respective target CGI and GGI with 30% plateau of varying GGI. X-axis shows the unitless relative depth of the absorber. Y-axis shows the unitless composi- tional ratio (GGI and CGI).
4.2 Analytical equipment
4.2.1 Glow discharge optical emission spectroscopy
Glow discharge optical emission spectroscopy (GDOES) is a spectroscopic method used for qualitative and quantitative analysis of metallic and non-metallic solid materials. Traditionally it has been used for both bulk analysis and depth profiling of several different types of samples. In this work GDOES has been used for depth profiling where it measures the elemental composition of a sample as a function of the depth.
The principle behind GDOES is that a sample with a flat surface is placed towards a small circular opening surrounded by an o-ring. This opening is part of a cylindrical copper tube called the anode, see figure 4.6.
The o-ring and the sample creates an air tight seal and is then pumped down to a pressure of about 0.5 - 10 hPa. The enclosed volume is then flushed with Ar and pumped a few times to remove any surface con- tamination. It is then kept at a certain application specific Ar pressure and a potential is applied between the cylindrical anode and the sample, which then acts as the cathode. For some applications with highly
conductive materials, a high direct voltage (DC) is applied. For materials with low conductivity, an alternat- ing voltage at radio frequencies (RF) is required. A glow discharge plasma is then formed and Ar cations are accelerated towards the sample surface. They hit the surface with sufficient energy to knock atoms out of the sample in a process known as sputtering, same as during the deposition but in a reversed manner.
These knocked out atoms then diffuse into the plasma where they collide with high-energy electrons, neu- trals and ions causing them to get excited. When they return to their ground state a photon is released with a wavelength that is characteristic for that atomic element. This light is then lead further into the machine through an entrance slit, to a grating and then to a detector. The detector can be of two kinds, the first one is a photo multiplier tube (PMT) that is physically placed at a position corresponding to a certain specific wavelength. The second type is a Charge Couple Device (CCD) which generally is less sensitive than the PMT but can handle multiple wavelengths in parallel. The intensity of the light is directly proportional to the concentration of the element in the plasma, assuming that there is no overlap at that wavelength between multiple elements. This allows for a quantification of that element to be made if there is a sufficient amount of reference samples with that element previously measured.
By measuring the variation of intensity over time, and comparing it to reference samples with known con- centrations and thicknesses, the intensity over time can be converted into atomic concentration over depth.
To achieve the best possible quantification, an adequate number of reference samples that have concentra- tions close to that of the unknown sample and ideally, a sputtering rate that is similar, are needed.[26]
In this work, all depth profiles were made on a Spectruma GDA750 in RF-mode with a 2.5mm anode located at Uppsala University. The calibration were made using reference samples supplied by the solar cell group at the Division for Solid State Electronics measured with XRF and profilometry.
Figure 4.6: Schematic of a Spectruma GDOES instrument [27].
4.2.2 IV-measurement
One of the most important electrical characterisation methods done on solar cells are IV-measurements (also known as JV) which stands for current-voltage. The cell is exposed to a light source as similar as possible to that of the AM1.5 spectrum at 1000 W /m2. A sweeping voltage is applied to the cell and the current produced by the cell is measured at specific intervals of the sweep. The resulting curve gives the parameters Voc, Isc, fill factor (FF) and efficiency which are probably the most important parameters within photovoltaics. Further evaluation of this curve and fitting the diode equation can also yield parameters such as series resistance and shunt resistance. For IV-measurements a Newport 91193-1000 solar simulator was used in conjunction with an Advantest R6244. In this setup the fill factor has been shown to have a large spread due to noise from the measurement setup. This can be seen when measuring the same cell multiple times and induces a uncertainty in both efficiency and fill factor.
4.2.3 Quantum efficiency (QE)
Quantum efficiency is an optoelectronic characterisation method that can reveal both electrical and optical parameters of the solar cell. The cell is exposed to a monochromatic light beam and the resulting current is measured. The wavelength of the light is then swept over a certain range (typically 300-1200nm) and the current for each point is recorded. This results in a spectral response known as the external quantum efficiency (EQE) and if the reflectance is subtracted, the internal quantum efficiency (IQE) is obtained. The EQE is given as a percentage, and is defined as the ratio between collected electrons and incident photons at a certain wavelength. The IQE is also a percentage, but the ratio between collected electrons and absorbed photons at a certain wavelength. It should be noted that absorbed in this case also includes the photons absorbed by the back contact that does not contribute to the photo-current. To remove the effect of the back contact, transmission measurement is needed which is far more difficult to perform.
The IQE gives more detailed information about the actual absorber, while the EQE provides information about the entire cell. Both give valuable information about the cells, but different aspects of it and where certain losses occur. For QE-measurements a Bentham PVE300 was used.
4.2.4 Scanning electron microscopy (SEM)
Scanning electron microscope (SEM)was used to provide cross-section images of the solar cells. Details on the underlying principles is found elsewhere. The instrument used was a Hitachi S4800.
4.2.5 X-ray diffraction (XRD)
X-ray diffraction (XRD) was used to determine crystal structure and crystallinity of the solar cells. Details on the underlying principles is found elsewhere. The instrument used was a Siemens D500 diffractometer.
5. Results and discussion
5.1 Influence of varying heater power
Increasing the heater power and thereby increasing the substrate temperature, lead to an increased Voc and Isc for all thicknesses. The fill factor remained unchanged, but an increase in efficiency was seen. Due to the measurement setup, a large spread in fill factor was observed which causes the trend in total efficiency to be less defined. As shown by figure 5.1, the increase in current was ~0.25 A for the thinnest absorbers, and ~0.1 A for the thickest. Figure 5.1 shows that the Vocon average gain ~40 mV for the thinnest, and ~10 mV for the thickest. In total, the increased temperature lead to an ~1.5% absolute increase in efficiency for the thinnest, and ~1% for the thickest. The GDOES estimated the thickness of the 225 kJ absorbers to be 550±30 nm, 300 kJ 720±30 nm, 350 kJ 840±30 nm and 400 kJ 950±30 nm.
Figure 5.1: Cell Iscand Vocfor various thicknesses and heater settings. Each box in the box plots represents the measurement data for all samples in one variation. Each color is a specific heater setting.
Figure 5.2: Cell FF and Eff for various thicknesses and heater settings. Each box in the box plots represents the measurement data for all samples in one variation. Each color is a specific heater setting.
thickness. This was most easily observed for the thicker absorbers, but the trend can still be observed in the thinner ones.
Figure 5.3: SEM images of cross-sections of a) 550 nm H1, b) 550 nm H5, c) 950 nm H1, d) 950 nm H5.
Layers are from bottom up: CGS, CIGS, buffer, window layer and top contact.
GDOES depth-profiles show that the GGI was mostly unaffected by the variation in temperatures. That applies for all thicknesses and in the majority of the absorber depth, an example can be seen in figure 5.4.
The gradient near the rear contact becomes slightly less steep with increasing temperature, while a slightly more defined knee was seen towards the back contact for H1 and H2. A more linear gradient was therefore seen at higher temperatures, most likely due to higher diffusion of Ga. The effect was most visible for the 720 nm and thicker absorbers. The back surface field should in theory then get slightly shifted towards the rear contact due to its correlation with the slope of the GGI. In none of the cases did the gradient show any tendencies towards discrete steps. This confirms that even at very thin absorbers with low heat, the diffusion was still high enough to create a smooth gradient.
The average CGI between samples varied between 0.74 - 0.86 but it had a median of 0.84 which was close to the target composition. This variation lied in [Ga] + [In] rather than [Cu] which was very stable between samples. However, there was no general correlation between this variation and the temperature. The integral of the relative intensities from the GDOES measurements, can be considered a good number for comparison of the total amount of an element present in the film. Based on this, the amount of In and Ga seemed to be very similar between samples. However, in the bulk of the absorber, the intensities were lower. This indicates that In and Ga were smeared out, rather than differing in total amount.
Looking at the Na signal of the depth profiles, it can be seen in figure 5.5 that the overall amount of Na
decreased with increasing temperature, but especially at the interface between CIGS/Mo as well as in the buffer. The decrease in Na at higher temperatures could lead to lower crystallinity as described in section 5.2. That however, depends on when the Na evaporated during the deposition, and if it was available during grain growth. Both higher crystallinity and larger grains is however, often coupled to higher temperatures due to increased diffusion, which may counteract the loss of Na. The reason why less Na was seen at higher temperatures is most likely due to the high vapour pressure of Na which caused it to evaporate. Since a pre-heating step was done before the CGS deposition, it is unclear if the same amount of Na was present regardless of heater power. This since the higher heater power was also supplied during the pre-heating step, which could lead to more evaporation. That would mean that there were less Na available during the CIGS deposition. Further testing and analysis between process steps, should reveal if there is any substantial loss of Na due to this. A better understanding of Na could lead to an even higher film quality. Regardless of this, higher temperatures seem to give larger grains.
Figure 5.4: CGI and GGI for 950 nm absorber and various heater powers.
The biggest relative and absolute increase in Voc, Iscand efficiency with heating power, was seen in the thin- ner absorbers (figure 5.2 and 5.1). This indicates that thinner absorbers were more affected by an increased temperature than thicker absorbers were.
Since diffusion is a process that depends on both temperature and time, thinner absorbers may therefore be compensated for the shorter deposition time with a higher temperature. It could also be beneficial with an annealing step for the thinner absorbers. However, a process step like that is very difficult to implement in the current process.
The decrease in Vocseen in low temperatures (figure 5.1) could possibly be explained by lower crystallinity and smaller grains as a consequence of less diffusion. That is because the recombination tends to be greater when the grain size is smaller and at decreased crystallinity. The decreased Voccould also be explained by an increased number of defects or a decrease in carrier concentration. Further analysis is needed to confirm this. Once again the thinner absorbers were most affected since they were exposed to high temperatures for a shorter amount of time. Not to mention that the sputtering process itself provides a large amount of heat.
Higher temperatures could therefore compensate for the lack of time and speed up diffusion to some extent for thin absorbers, as described in section 3.5. This was supported by the fact that with the highest heating power, the Voc for the thinnest absorbers was increased to levels close to that of the thickest absorbers. It is difficult to say if even higher temperatures or longer time would increase Voc even further. Judging by the trend, it is most likely the case for 550 nm absorbers. That is because unlike 720 nm and thicker, the
Figure 5.5: Na depth profiles for 950 nm absorber and various heater powers.
Isc was mainly affected by thickness and not as much by heater setting. This was somewhat expected, but some correlation with heater setting was still seen. It is most likely linked to the same mechanisms as for the Voc(less recombination).
As a note to this, the correlation here was done between heater power and cell performance. Due to difficul- ties in measuring the temperature in this in-line process, it is not necessarily so that the temperature scales linearly with heater power. To further explore this, more accurate measurement of temperature should be done. The process equipment has pyrometers that measure the temperature. The problem with pyrometers lies in the fact that they are sensitive both to temperature and thickness of the deposited film. Therefore no correlation to the actual temperature can be drawn since small variations in thickness can lead to a large difference in estimated temperature.
5.2 Influence on morphology by Na
Figure 5.6 shows that the amount of Na in the Mo/CIGS interface increases with an increasing supply of Na.
At the same time, the amount in the bulk of the absorber remains constant. Figure 5.7 shows that the (112) peak was significantly larger for a given heater setting, in the cases where Na was available during the CGS deposition. The heater setting was an equally important parameter and for the height of the (112) peaks. It was even larger when Na was also supplied during the actual deposition of the CIGS in the form of NaF.
The other peaks does not show any clear trend and only have small changes. SEM images (figure 5.8) show no clear distinction in grain size. However the largest grains and also the roughest top surface tends to be seen when both MoNa and NaF were present.
Figure 5.6: Na content near the back contact with different methods of incorporating Na for H5.
Figure 5.7: a) XRD peaks for peak for H1 and H5 with different methods of incorporating Na, b) zoomed view of (112) peak
Figure 5.8: Cross sections of absorbers with different Na content, a) H5 MoNa CIGS:NaF, b) H5 Mo CIGS:NaF, c)H5 MoNa CIGS. Layers are from bottom up: CGS and CIGS.
By varying the way that Na was incorporated, both the amount of Na available during the growth of the CGS, and subsequently during the growth of the CIGS, was changed. When Na was incorporated through doping of the Mo-layer (Mo:Na) there were already some amount available when the CGS was deposited, which was not the case otherwise. This seems to increase the predominance of the (112) plane as its peak height was significantly lower without Mo:Na. It seems equally important that there was Na already present when the CGS was deposited, as well as when the CIGS was deposited. Leaving either one of those out, decreases the height. Since the other peaks remain approximately the same and does not show the same trend, it is difficult to say if the overall cristallinity was changed and how much was just variations in tex- ture. Further analysis with GIXRD or TEM would be needed to examine this. It still remains unclear exactly how Na influences the properties of CIGS layers and it has been suggested that it leads to both larger and smaller grains [28]. It also needs to be further explored how this influences the cell performance.
The actual amount of Na present at the interface between CIGS/Mo seems to be additive. The more Na present during deposition, the more was accumulated in that interface. This despite the increase in crys- tallinity and the unchanged average concentration in the "bulk" of the absorber. This means that despite there being larger grains, less surface area and possibly higher crystallinity, more Na was still locked up in that interface. This could be the result of higher diffusion in the grain boundaries between large grains compared to small grains. Further investigation of this is needed to confirm this.
Despite the large difference in (112)-height, SEM images does not show that any major difference in grain size in the different samples. It does appear though that the grain size decreases in the order of the peak height in the XRD-spectrum, but no reliable FWHM-comparison of peaks have been made. This still sup- ports previous findings that more Na increases the grain size and crystallinity.
5.3 Effect on Ga-grading in thin CIGS
5.3.1 Varying back surface field
Figure 5.9 shows that varying the amount of CGS in the absorber (0, 10, 20 or 30% of total absorber thick- ness) most prominently results in a larger Ga-rich region near the back contact. It also slightly increases the overall Ga-content, stretching into the bulk of the absorber. The steep increase of GGI near the back-contact should in theory result in a back-surface field in which the strength of the field is determined by the slope of the GGI [16]. This however was not explicitly measured in this work.
Figure 5.9: GDOES depth profile for 720 nm absorber with 0, 10, 20 or 30%.
The resulting IV-curves are shown in figure 5.10, in which it can be seen that the Iscincreases from 0 to 10%
CGS, but then decreases when made thicker than that. The Voc decreases with increased amount of CGS in all absorbers (seen in figure 5.10), even though at 550 nm between 0 and 10% CGS this difference was smaller than for the rest. The fill factor show no obvious trend due to the large spread, but somewhat of an decrease can be seen towards more CGS. The resulting efficiencies show that a maxima exists at 10% CGS for all thicknesses, even though the difference between 10 and 20% at 720 nm was negligible.
Figure 5.10: IV-data for 550, 720 and 950 nm thick absorbers with 10, 20 or 30% of the absorber consisting of CGS.
Figure 5.11: SEM images of 950 nm absorbers with a) 0% CGS, b) 10% CGS, c) 20% CGS and d) 30%
CGS. Layers are from bottom up: CGS, CIGS, buffer, window layer and top contact.
SEM-images (figure 5.11) show that near the back contact there was a region of small grains that increase with the amount of CGS and it appears to be more pores in the structure at 0% CGS.
The Na content has a maxima at 10% CGS, both in total amount and at the back contact interface for all thicknesses. The amount of Na in the buffer and absorber remains the same, regardless of CGS content.
The reason for this could lie in a certain GGI near the back contact reaching a level in which the Na affinity was at its peak. It has been shown that CGS shows a higher affinity for Na than CIGS [29]. The affinity in the high GGI region between CGS and CIGS has however not been studied, and therefore the affin- ity in high GGI CIGS remains unknown. The change in Na does not show any strong correlation with the IV-curves (figure 5.10), but it is still possible that this to some extent contributed to the measured parameters.
In theory, by adding CGS at the back of the absorber a highly beneficial back-surface field is created that increases the current collection and decreases recombination. The increase in Iscwhen going from 0% CGS to 10% CGS (figure 5.10) shows the effect on current collection. However increasing the CGS-content above 10% has the opposite effect, and more CGS than that leads to less Isc for all thicknesses. This can be understood by looking at the GDOES depth profiles (fig 5.9), where it can be seen that the slope of the GGI towards the back does not get steeper with increasing CGS above 10%. Instead the GGI maxima gets higher and the shape gets more curved rather than linear. Since the slope of the GGI is coupled to the elec- trical field, only a small additional gain in current collection was therefore expected. At the same time, the position where the steep increase towards the back begins, was shifted towards the front of the absorber.
No clear changes in the QE can be seen between the different samples indicating that the surface band gap does not change. This shift towards the front along with a slight overall increase in GGI towards the front, lead to the absorber getting less efficient at absorbing low energy photons. This was because low energy photons have a lower probability of absorption and therefore more of them were absorbed near the back of the absorber [15]. This requires however that the band-gap is lower than the photon energy, and the steep increase in GGI causes a steep increase in band-gap which inhibits this. This decreased absorption counter- acts the additional small electrical field that >10% CGS gave and therefore the total Iscdecreased.
This change in low energy photon absorption should be seen in QE-measurements, however the QE-curves look almost identical. If calculating the Isc from the curves, the result was a value comparable to that from the IV-measurement in figure 5.10. No visible change in the long wavelength cutoff was seen. This could be because according to the GDOES, the thickness of the absorber increases with increasing amount of CGS.
The sum of CGS+CIGS should however always be the same. This was most likely due to a higher sputtering rate of the CGS target compared to the CIGS target. The result was a thicker absorber that gave a higher current, and compensates for the expected change in long wavelength cutoff that otherwise would have been seen. If adjusted for this effect, the 20 and 30% CGS samples would most likely show lower Isc than for 10%. The difference in estimated thickness could also be the result of changed film properties which causes the GDOES to incorrectly calculate the thicknesses in one or several samples. Further depth measurements with profilometer or other tools would be needed for this.
The Voc, which according to theory is expected to increase with a back surface field [14], show the opposite behaviour. Vocwas reduced with an increasing amount of CGS in a linear fashion for all thicknesses. This could be because the overall increase in Ga-content lead to deteriorated electronic properties [30]. Sufficient increase in overall Ga-content would outweigh the decrease in recombination due to a higher band gap.
Adding to this, additional small grains near the back contact were seen (figure 5.11) with increasing amount of CGS. The additional grain boundaries could also be a contributing factor to the decreasing Voc. This is because grain boundaries are known contributors to recombination. However, grain boundaries are generally considered benign, but depending on the nature of them, they can still cause a great amount of recombination [10]. Even in the most benign case, they will likely always have a higher recombination rate than the bulk, and more grain boundaries should result in more recombination [31].
Between thicknesses, the biggest difference was once again the Isc. It was mostly dependent on the thickness of the absorber, and the back surface field can only compensate for a small part of the difference. The total efficiency follows the same pattern as that of the I , and the best performance was given by 10% CGS.
This shows that some back surface field was as expected, highly beneficial, but the effects were very similar regardless of the thickness. Therefore the thinner absorbers did not gain more from this than the thicker ones did.
5.3.2 Varying grading plateau
The resulting GDOES depth profiles show that the desired flat non-graded region (plateau), was not reached for any of the samples regardless of thickness. An example of this can be seen in figure 5.12. There were always at least a slight grading, but differences in both CGI and GGI can still be seen between samples. This was especially seen in the samples with a 60% "plateau". The 30% samples differ less, and the GGI only varies close to the front of the absorber. The IV-curves (figure 5.13) show that the Iscincreases and the Voc
decreases with decreasing overall GGI. The Voc, Isc and the efficiency were all more affected by changes in GGI when a wider plateau was made. Fill factors show no visible trend and remain relatively constant regardless of thickness and GGI. The efficiency remains almost constant at the different thicknesses for the 30% samples. For all thicknesses except 950 nm, the efficiency decreases with increasing Ga-content for 60% samples.
Figure 5.12: GDOES depth profile for 550 nm thick absorbers with either 30 or 60% of the absorber con- sisting of a non-graded 0.19,0.22 or 0.26 GGI flat region.
The results above are in line with what the theory suggests. Since higher GGI results in higher band gap, it produces less recombination and promotes higher Voc. It also leads to less low energy photons being absorbed, and therefore lower Isc. These effects were visible for all thicknesses with equal magnitude, but the 30% samples were substantially less affected. Most likely because the different gradients were very similar due to only 30% of the absorber being altered by a small magnitude. For the thinnest absorbers the Vocremained the same despite changes in the GGI. This effect was not seen in any of the other thicknesses.
The higher current seen for the samples with lowest Ga content, also show the highest efficiency. This was because the increase in current that comes with lower GGI was proportionally larger than the drop in voltage.
Figure 5.13: IV-data for 550, 720 and 950 nm thick absorbers with either 30 or 60% of the absorber sputtered from one 0.19,0.22 or 0.26 GGI target.
It should also be noted that the CGI for some samples was up to ~0.1 lower in average compared to what was seen in previous sections for similar recipes. The cause of this seems to lie in a higher raw signal of In and a lower of Cu. The Se signal was also lower which indirectly affects the CGI calculation as well as the calculation of the other elements. This could be due to different concentrations as a result of varying conditions during the deposition. An example of this might be a thinner CGS layer. It would cause the In sputtered from the CIGS targets to smear out less, give a lower Cu content and lower the overall CGI. It does not explain the Se though.
More likely it was due to the quantification model being less accurate in a certain range of raw signals.
The reference samples lie in a quite narrow range of raw signals and the behaviour outside this region was only being estimated by extrapolation. Besides a change in concentration, a change in raw signal can be the result of a slightly different sample morphology or crystallinity, which may alter the plasma concentration of the elements, but also the sputtering rate. If the quantification model can not compensate for this, then the quantification will be incorrect. Besides the CGI, the GGI should also be affected by this but due to that being one of the parameters that were expected to vary, it was difficult to identify this. It can not be excluded that the GDOES instrument itself might also be the cause. Because of this, the absolute values of the GGI and CGI may not be trusted, but the relevant comparision between samples in this series should still be valid.
It has previously been shown that a CGI between 0.78-0.84 does not affect performance substantially [32].
Even though this may not be true for all CIGS absorbers, it can still be used as a guideline. In this series, many samples were below that, and that would likely have caused the absorber to be too Cu-poor, resulting
concentrations were within the expected range and that the GDOES measurements have some error in them.
To confirm this, reference samples in a wider compositional range would have to be made and the samples measured again. The morphology and crystallinity should also be examined in detail with SEM and XRD.
In total, the consequences of low GGI plateaus were more or less the the same regardless of thickness, and there is little indicating that thinner absorbers would benefit more than thicker.
It is likely that the current will increase even more with lower GGI. The Voc will unfortunately likely also decrease further but it is unknown how the total efficiency will be affected. The electronic properties will increase with decreasing GGI, but only down to 0.25. Beyond that and the electronic properties will get slightly worse [30]. The practical implications of this are hard to predict. Therefore, in future work, a plateau with even lower GGI should be made to examine this.
