STOCKHOLM SWEDEN 2019,
Advanced rear contact
design for CIGS solar cells
JORGE ALEXANDRE DE ABREU MAFALDA
Dept. of Electrical Engineering and Computer Science
Degree project in Engineering Physics (IF246X), Second Cycle
Advanced rear contact design for CIGS solar cells
Author: Jorge Alexandre de Abreu Mafalda Supervisors: Prof. Dr Mattias Hammar, KTH, Sweden
Carl Reuterskiöld Hedlund, KTH, Sweden
Prof. Dr Bart Vermang, IMEC & Hasselt University, Belgium Gizem Birant, IMEC & Hasselt University, Belgium
Examiner: Prof. Dr Mattias Hammar, KTH, Sweden
The current trend concerning the thinning of solar cell devices is mainly motivated by economic aspects, such as the cost of the used rare-earth elements, and by the re- quirements of emergent technologies. The introduction of ultra-thin absorber layers results in a reduction of used materials and thus contributes to a more cost-effective and time-efficient production process.
However, the use of absorber layers with thicknesses below 500nm gives rise to multiple apprehensions, including concerns regarding light management and the absorber’s quality.
Therefore, this experimental work presents a novel solar cell architecture that aims to tackle the issues of optical and electrical losses associated with ultra-thin absorber layers. To that end, a Hafnium Oxide (H f O2) rear side passivation layer was introduced in-between the copper indium gallium (di)selenide Cu(In, Ga)Se2, CIGS-based absorber layer and the Molybdenum (Mo) back contact. Then, the proposed Potassium Fluoride (KF) alkali treatment successfully established point contacts on the ALD-deposited oxide layer, resulting in a passivation effect with minimum current blockage.
The established cell architecture showed significant improvements regarding both open circuit voltage (Open-Circuit Voltage (Voc)) and efficiency when compared to unpassi- vated reference devices. The used solar cell simulator (SCAPS) attributes the observed improvements to a reduced minority carrier recombination velocity at the rear side of the device. Moreover, the provided photoluminescence (PL) results report a higher peak intensity and lifetime for passivated devices.
Furthermore, the overlay of the given external quantum efficiency (EQE) spectra with the performed simulations show that the HfO2passivation layer improves the optical reflection from the rear contact over a wavelength interval ranging from 500 to 1100 nm, resulting in a short circuit current (Jsc) improvement. An increased quantum efficiency observed throughout almost the entire measurement range, confirms that the enhance in Jscis also due to electronic effects.
Here, a produced solar cell device including a 3nm-thick HfO2rear passivation layer and a 500nm-thick 3-stage CIGS absorber, achieved a conversion efficiency of 9.8%.
Further, the approach of combining an innovative rear surface passivation layer with a fluoride-based alkali treatment resulted in the development and successful characterisa- tion of a 1-stage, 8.6% efficient solar cell. Such result, mainly due to a short circuit current (Jsc) enhancement, supports the introduction of more straightforward production steps, which allows a more cost-effective and time-efficient production process.
The produced device consisted of a 500nm-thick CIGS absorber, rear passivated with an ultra-thin (2nm) HfO2layer combined with a 0.6M KF treatment.
Keywords: CIGS, H f O2, surface passivation, alkali treatment, point contact openings, SCAPS
Den nuvarande trenden när det gäller solcellsanordningar huvudsakligen motiveras av ekonomiska aspekter, såsom kostnaden för att använda sällsynta jordartsmetaller, och av kraven i ny teknik. Införandet av ultratunna absorptionsskikt resulterar i en minskning av använda material och bidrar därmed till en mer kostnadseffektiv och tidseffektiv produktionsprocess.
Användningen av absorptionsskikt med tjocklekar under 500 nm ger emellertid upphov till flera bekymmer, beträffande ljushantering och absorptorkvalitet.
Därför presenterar detta experimentella arbete en ny solcellarkitektur som syftar till att ta itu med frågorna om optiska och elektriska förluster förknippade med ultra- tunna absorberlager. För detta ändamål infördes ett Hafnium Oxide (H f O2) bakre sidopassiveringsskikt mellan kopparindiumgallium (di) selenid Cu(In, Ga)Se2, CIGS- baserat absorberande skikt och Molybdenum (Mo) kontakt. Sedan upprättade den föreslagna kaliumfluorid (KF) alkali-behandlingen framgångsrikt punktkontakter på det ALD-avsatta oxidskiktet, vilket resulterade i en passiveringseffekt med minimal strömblockering.
Den etablerade cellarkitektur visade signifikanta förbättringar avseende både öppna kretsspänningen (Voc) och effektivitet i jämförelse med opassiverad referensanordningar.
Den använda solcellsimulatorn (SCAPS) tillskriver de observerade förbättringarna till en minskad minoritetsbärares rekombinationshastighet på enhetens baksida. Dessutom de tillhandahålls fotoluminescens (PL) resultat rapporterar en högre toppintensitet och livslängd för passive enheter.
Dessutom visar överläggningen av det givna externa kvantitetseffektivitetsspektrumet (EQE) med de utförda simuleringarna att passiveringsskiktet HfO2förbättrar den op- tiska reflektionen från den bakre kontakten över ett våglängdsintervall från 500 till 1100 nm, vilket resulterar i i en kortslutningsström (Jsc) förbättring. En ökad kvantverknings- grad observerats i nästan hela mätområdet, bekräftar att öka i Jscär också på grund av elektroniska effekter.
Här, en producerad solcellsanordning innefattande en 3 nm-tjock HfO2bakre passiver- ingsskikt och ett 500 nm-tjock 3-stegs CIGS absorber, uppnått en omvandlingseffektivitet på 9.8%.
Vidare resulterade tillvägagångssättet att kombinera ett innovativt bakre ytpassiver- ingsskikt med en fluoridbaserad alkalibehandling i utvecklingen och framgångsrik karaktärisering av en 1-stegs, 8.6% effektivitet solcell. Ett sådant resultat, främst på grund av en kortslutningsström (Jsc) förbättring, stöder införandet av mer enkla produk- tionssteg, vilket möjliggör en mer kostnadseffektiv och tidseffektiv produktionsprocess.
Den framställda anordningen bestod av ett 500 nm-tjock CIGS absorber, bakre pas- siverad med en ultra-tunn (2 nm) HfO2-skikt kombineras med en 0.6M KF behan- dling.
Nyckelord: CIGS, H f O2, surface passivation, alkali treatment, point contact openings, SCAPS
The financial support of IMEC, Energyville Consortium, Royal Institute of Technology (KTH) and Erasmus+ Internship grant is gratefully acknowledged. I thank the IMEC and Solliance teams for their assistance and kind collaboration.
I thank my supervisors Dr Bart Vermang and Gizem Birant for the wholehearted, in- sightful feedbacks and guidance. I also thank Dr Marc Meuris, for providing me with the support and invaluable learning experience at IMEC.
I also thank Dr Jessica de Wild for the 3-stage CIGS depositions and all the guidance, as well as valuable feedback throughout the entire process.
In addition, I am grateful to Dr Guy Brammertz for all the support and for developing the used MatLab scripts. Furthermore, a special thank you to Thierry Kohl for the assistance on the MatLab coding and for the 1-stage CIGS depositions.
I am also thankful to Eng. Kim Baumans for his guidance and time spent on equipment training.
I am grateful to Prof. Dr Mattias Hammar and Prof. Dr Carl Hedlund, members of my thesis committee, for their support and constructive feedback.
Furthermore, a special thank you to all my fellow student mates, David, Tomás, Moham- mad, Samaun, Sarallah, Sownder and Srivatsa.
Finally, I thank my family members and friends for their support.
Stockholm, July 29, 2019
Jorge Alexandre de Abreu Mafalda
1. General Introduction 1
1.1. Background . . . 1
1.2. Problem . . . 2
1.3. Motivation . . . 3
1.4. Goal . . . 4
1.4.1. Novel Passivation Layer . . . 4
1.5. Results highlight . . . 5
1.6. Outline . . . 6
2. Theoretical Background 8 2.1. Photovoltaic effect and Principle of solar operation . . . 8
2.2. CIGS solar cells . . . 11
2.2.1. Introduction . . . 11
2.2.1.1. Crystal Structure . . . 11
2.2.1.2. Defect physics . . . 13
2.2.2. Cell architecture and description . . . 16
2.2.2.1. CIGS vs c-Si . . . 17
2.2.3. Cell’s physics . . . 20
2.2.3.1. Loss mechanisms and efficiency limits . . . 24
2.2.4. Built-In Electric Field . . . 30
2.2.5. Tunnelling Conduction through ultra-thin oxide layers . . . 32
2.2.6. Motivation towards ultra-thin CIGS solar cells . . . 32
2.2.6.1. Scale-Up challenges for CIGS ultra-thin film modules . 33 2.2.7. State of the art of ultra-thin CIGS solar cells . . . 34
3. Methods and Techniques 40 3.1. Fabrication Process Overview . . . 40
3.1.1. Substrate cleaning . . . 41
3.1.2. Barrier layer and Back contact deposition . . . 42
3.1.3. Passivation layer deposition . . . 42
3.1.4. Alkali deposition . . . 44
3.1.4.1. Sodium Flouride . . . 45
3.1.4.2. Potassium Flouride . . . 46
3.1.5. Absorber layer co-evaporation . . . 46
3.1.6. Buffer layer (CdS) deposition procedure . . . 48
3.1.7. (i-)ZnO(:Al) window layer . . . 49
3.2. Devices characterisation . . . 50
3.2.1. Electrical characterisation . . . 50
3.2.1.1. Temperature dependency . . . 53
3.2.2. Optical characterisation . . . 54
3.2.2.1. External Quantum Efficiency (EQE) measurements . . . 54
3.2.2.2. Static Photoluminescense (STPL) measurements . . . . 54
3.2.2.3. Time-Resolved Photoluminescense (TRPL) measurements 55 3.2.3. Physical analysis and characterisation . . . 56
3.2.3.1. XRF mapping . . . 56
4. Passivation Layer Development and Characterisation 57 4.1. Scanning Electron Microscopy (SEM) . . . 57
4.1.1. The role of Selenium in the creation of contact openings . . . 58
4.1.2. The influence of pre-Selenisation outgassing . . . 64
4.2. Gwyddion analysis . . . 67
4.3. ImageJ analysis . . . 70
4.4. Chapter’s Conclusion . . . 74
5. Solar cell characterisation 76 5.1. 1-stage CIGS co-evaporation . . . 76
5.1.1. Batch 1 and 2 . . . 76
5.1.2. Batch 3 . . . 84
5.1.3. Batch 4 . . . 87
5.1.4. Batch 5 . . . 90
5.2. 3-stage co-evaporation . . . 93
5.2.1. Temperature dependent I-V measurements . . . 96
5.2.2. C-V analysis . . . 97
5.2.3. Photoluminescence analysis . . . 99
5.2.4. Two-diode model analysis and simulation . . . 102
5.2.5. External Quantum Efficiency measurements . . . 105
5.3. Chapter’s conclusion . . . 107
6. SCAPS simulations 109 6.1. CIGS baseline . . . 110
6.2. Introduction of the Hafnium Oxide rear passivation layer . . . 111
6.3. Chapter’s conclusion . . . 116
7. Concluding remarks 117 8. Outlook and Future Work 120 Bibliography . . . 120
A. Detailed derivation of the P-N junction’s saturation current and Photogen- erated current 132 A.1. P-N junction’s saturation current . . . 132
A.2. Photogenerated current . . . 134 B. Support information for Temperature-Dependent C-V measurements 136
2.1. Photon’s absorption mechanism in a semiconductor material with bandgap Eg. The photon with energy Eph = hν excites an electron from Ei to Ef. Moreover, at Ei a hole is created. (b) If Eph > Eg, a part of the energy is thermalised. Extracted from [1] . . . 9 2.2. (A) Direct band gap material. (B) Indirect band gap material. The schmatic
represents the electron thermal excitation mechanism. The electron from the occupied state at the VB is excited to the CB, where it occupies the previous unoccupied position. Extracted from [2] . . . 10 2.3. Electron-Hole pair generation and recombination in a solar cell. Extracted
from [3]. . . 10 2.4. Comparison between the primary stage (ZnS structure) and the evolu-
tion to the Chalcopyrite Tetrahedral crystal (Cu(In,Ga)Se2(CIGS)). The different colours represent the different atoms, as indicated in the figure.
Extracted from [4] . . . 12 2.5. Doped chalcopyrite crystal structure (CIGS). The Cu(In, Ga)Se2quater-
nary compund is formed when Gallium atoms partially substitute Indium atoms in the ternary CuInSe2system. . . 13 2.6. Electronic levels of intrinsic defects in the CuInSe2ternary compound.
Black columns represent acceptor levels, while white ones regard the donor-type levels.Extracted from [12]. . . 14 2.7. Best solar cell efficiencies achieved by different PV technologies. Released
on February 2019. Extracted from [5] . . . 15 2.8. Solar cell design. . . 16 2.9. CIGS stucture a), and band diagram b). Extracted from [1] . . . 21 2.10. Concentration of charge carriers in a p-n junction under forward bias.
Extracted from [6]. . . 23 2.11. Representation of the heat flows in the proposed heat engine. Extracted
from [1]. . . 24 2.12. Loss mechanisms taken into account by the Shockley-Queisser efficiency
limit. Extracted from [1]. . . 26 2.13. Solar cell equivelent circuit including a series resistance. Extracted from [7]. 29 2.14. Solar cell equivelent circuit including shunt resistance. Extracted from [7]. 30 2.15. n-P heterojuction. Extracted from [1]. . . 31 2.16. n-P heterojuction, considering a vacuum energy constant across the junc-
tion. Extracted from [1]. . . 31 3.1. Complete CIGS solar cell device fabrication flow. The first two illustrated
steps regard the ALD-deposition of the rear side HfO2passivation layer, and the introduction of the alkali species, respectively. . . 41
3.2. Preparation of the cleaning solutions and cleaning procedure. The se- quential bath into three different aquariums is as following: 1-NH3(30ml) + H2O (270ml). Rest time of six minutes; 2-EDI water. Rest time of three minutes; 3-IPA. Rest time of three minutes; 4-Drying of substrates using a
Nitrogen flow. . . 42
3.3. Schematic of an atomic layer deposition reaction. An ALD cycle corre- sponds to four steps identified in the Figure as (a): reactants introduction, (b): excess reactants purging using an inert gas, (c): introduction of the second precursor (counter-reactant) and (d): second purging to remove any excess reactants and reaction by-products. The illustrated cycle is repeated until the desired film thickness is achieved. Extracted from [8] . 43 3.4. Evolution of the spin-coating process over time. A) Thick film undergo- ing a thinning step, which is mostly dominated by the spun off of the remaining fluid. B) The intermediary point where thining by flow equals thining by evaporation. C) Immobilised thin-film. The evaporation of the solvent dominates the thinning. . . 45
3.5. CdS deposition procedure . . . 49
3.6. CdS deposition procedure: Rising and drying steps. . . 49
3.7. 5x5 cm2CIGS solar cell device after mechanical scribing. . . 50
3.8. P-N junction J-V characteristics under illumination and in the dark Ex- tracted from [1] . . . 52
3.9. CdS/CIGS interface band diagram schematic, illustrating the four main recombination processes. The major difference between the presented processes refer to their location within the band, and therefore, their different activation energies. Extracted from [9]. . . 53
3.10. Typical TRPL setup. Extracted from [10] . . . 56
4.1. SEM imaging analysis. . . 58
4.2. SEM imaging after alkali deposition on the 5x5 cm2sample’s surface. . . 59
4.3. SEM imaging after N2annealing (left) and Se annealing (right). . . 60
4.4. SEM imaging after N2annealing (left) and Se annealing (centre and right). 60 4.5. Example of Scanning Electron Microscope (SEM) characterisation after Se annealing . . . 61
4.6. SEM characterisation after Se annealing . . . 62
4.7. SEM characterisation of a test sample with a 3nm HfO2passivation layer and 0.5M KF alkali treatment spin-coated layer. Figures (a)-(d) evolve in magnification range. . . 63
4.8. SEM surface analysis of test sample passivated with a 6nm thick HfO2 layer, followed by the deposition of 0.1M NaF alkali treatment. . . 64
4.9. Test sample passivated with a 6nm thick HfO2layer and a 0.4M KF alkali after a 540oC N2annealing, followed by a 540oC Se annealing. . . 65
4.10. Test sample passivated with a 6nm thick H f O2layer and a 0.4M KF alkali after a 540oC Se annealing without a previous N2heat treatment. . . 66
4.11. Blister creation and subsequent burst as a result of the degassing of the trapped gases. . . 67 4.12. Set of test samples with a 3nm thick HfO2passivation layer. Additionally,
Figure a) has an average opening percentage of 23.4%, while the one on the right has 25.6% of its surface area opened. . . 68 4.13. Relevant set of test samples with a 6nm thick HfO2 passivation layer.
Additionally, the NaF treated test sample, from which Figure a) and b) are representative spots, has an average opening percentage of 19.8%, while the one treated with 0.4M KF, on the bottom has 12.9% average of surface openings. Additionally, measurement spots 2 and 4 are not shown due to their similarity to Spots 1 and 3, respectively. . . 69 4.14. Relevant set of test samples with a 9nm thick HfO2 passivation layer.
Additionally, the KF treated test sample, from which Figures a) and b) are representative spots, has an average opening percentage of 17.3%. . . . 70 4.15. Figures (a), (b), (e), (f) regard a test sample passivated with a 3nm H f O2
layer, followed by the spin-coating of 0.4M KF and subsequently a 20 minutes Se annealing at 540oC. Figures (c), (d), (g) and (h) followed the same processing steps but treated with 0.4M NaF instead. The KF sample presents a 21.5 opening percentage, while the NaF-treated sample has openings in 27.1% of the surface area. . . 71 4.16. Figures (a), (b), (e), (f) regard a test sample passivated with a 6nm HfO2
layer, on which a 0.4M KF aqueous solution is spin-coated. Subsequently, it undergoes a 20-minute Se annealing at 540oC. Figures (c), (d), (g) and (h) followed the same processing steps but treated with 0.4M NaF instead.
The KF sample presents a 17.3 opening percentage, while the NaF-treated sample has no significative openings. . . 72 4.17. Figures (a), (b), (e) and (f) correspond to the four measurement spots in
the KF treated sample, while Figures (c), (d), (g) and (h) regard the ones subject to the NaF alkali treatment. For visual comparison purposes, the four most left samples are presented with an 800x magnification, while on the most-right set (c) and (h) have a 31x magnification. . . 73 5.1. XRF maps concerning the solar cells CGI ratio. The four pictures on the
left correspond to the first batch, while the ones on the right regard Batch 2. . . 77 5.2. XRF maps concerning the solar cells GGI ratio. The four pictures on the
left correspond to the first batch, while the ones on the right regard Batch 2. . . 78 5.3. 2D XRF maps concerning the absorber’s thickness. X and Y axis are pre-
sented in cm and correspond to the lateral sizes of the sample. Moreover, the standing bar indicates the thickness variation, presented in µm. The four pictures on the left correspond to the first batch, while the ones on the right regard Batch 2. . . 79
5.4. Statistical presentation of the key electrical parameters of a solar cell device. The green boxes represent the two reference samples, one cor- responding to each batch, whereas the orange boxes concern the tested devices. Moreover, the best performing cell is highlighted in blue. Ad- ditionally, in the x-axis, it is presented the sample ID that follows HfO2 thickness/Alkali (KF) molarity. . . 81 5.5. Representative illuminated and Dark I-V curves for each of the produced
device on the third batch. . . 85 5.6. EQE and PL results for one reference, unpassivated cell and three passi-
vated cells with an oxide thickness ranging from 2 to 6nm and constant alkali molarity of 0.4M. The black curve labelled as 0nm/0.4M corre- sponds to the reference unpassivated device (CIGS peak marked by the blue arrow), whereas the three remaining curves correspond to the passi- vated ones, with an increasing HfO2thickness, as indicated in the label.
PL spectra are presented in logarithmic scale, allowing for a better com- parison and defect visualisation. . . 86 5.7. Illuminated and dark I-V curves correspondent to a measurement bias
range of -1 to 1V. . . 88 5.8. EQE spectra acquired over a 300-1300nm range. On another note, the PL
analysis used a varying excitement light pulse within the 950-1350nm range. The black curve labelled as 0nm/0.4M corresponds to the reference unpassivated device (peak marked by blue arrow), whereas the three remaining curves correspond to the passivated ones, with an increasing HfO2 thickness, as indicated in the label. PL spectra are presented in logarithmic scale, allowing for a better comparison and peak visualisation. 89 5.9. Batch 5 representative Illuminated and dark I-V curves, which correspond
to a measurement bias range of -1 to 1V. . . 91 5.10. Spectral response for both EQE and PL analysis as a function of the alkali
molarity change. PL spectra are presented in logarithmic scale, allowing for a better comparison. . . 92 5.11. Box plot statistical presentation of the most significant solar cell’s electri-
cal parameters. The presented results concern two solar cell devices with a 3-stage co-evaporated CIGS absorber. The x-axis present the samples’
identification codes which are named according to HfO2 thickness/e- vaporated NaF thickness. Furthermore, to allow a more straightforward comparison, the reference solar cell is presented in green, whereas the passivated device is shown in orange. . . 94 5.12. Voctemperature dependence for two distinct solar cells. . . 96 5.13. 2D -fdC/df mapping profiles. . . 98 5.14. Photoluminescence analysis and spectral response. The intensity scale
is presented in logarithmic scale to allow the identification of secondary peaks. . . 100 5.15. Lifetime extraction from Time-Resolved Photoluminescence analysis. . . 102
5.16. Illuminated J-V curve for the highest efficiency cell of the passivated 3- stage absorber device, incorporating the calculations of the two-diode model. . . 104 5.17. Open-circuit voltage (Voc) as a function of minority carrier lifetime. The
trend is theoretically calculated from Equation 5.10, by replacing the lifetime values from 0.1 until 100ns, in a 31-step process. . . 105 5.18. Overall average results comparison between the reference cell (Black) and
the 3nm H f O2passivated device. . . 106 5.19. EQE results corresponding to the five-cell average for each analysed
wavelength. . . 106 6.1. Variation of the CIGS layer thickness from 300 to 500nm, with a 50nm
increment for an unpassivated CIGS solar cell device. . . 110 6.2. Variation of the CIGS layer thickness from 300 to 500nm, within five steps
and using a constant HfO2thickness of 2nm. . . 114 6.3. a) IV and b) EQE measurements show overall improvement as a result of
the introduction of the rear-side passivation layer. . . 115 B.1. 2D fdC/df maps corresponding to temperature dependent C-V measure-
ments, ranging from−100oC to 40oC, with a 20oC step. Room temper- ature maps ∼ 20oC are given in Figure 5.13. Unfortunately, the map corresponding to the measurement of the unpassivated device at−20oC cannot be displayed due to data incongruences. . . 138 B.2. Frequency response maps obtained from SCAPS simulations. The first
row corresponds to a constant recombination velocity while varying the back contact’s surface height from 0.25eV to 0.35eV. Moreover, the second row illustrates the effects of the variation of the recombination velocity.
The reduction of the frequency response, i.e. the downward shift of the observable response is caused by the increasing of the back contact’s barrier height, but also by the reduction of the recombination velocity. . 139
2.1. Twelve possible point defects and respective oxidation number. Extracted from [11]. . . 14 3.1. Sodium Flouride experimental molarities. Room temperature NaF molar-
ity values extracted from [13]. . . 46 3.2. Potassium Flouride experimental molarities. Room temperature NaF
molarity values extracted from [14]. . . 46 4.1. Annealing process parameters. . . 59 4.2. Process parameters for study the outgassing role on the overall contact-
opening process. . . 64 5.1. 1-stage CIGS absorber samples. . . 76 5.2. Device’s best performant solar cell values. . . 81 5.3. Percentual differences between the produced test devices and correspon-
dent references. . . 83 5.4. Variation of the passivation layer thickness, while keeping the alkali
treatment and molarity constants. No Sodium was added to the solar cell device. As before, the sample ID naming follows the HfO2 thickness/
alkali molarity nomenclature. . . 84 5.5. Batch 3 electrical characterisation results and correspondent variation. . 84 5.6. Variation of the passivation layer thickness, while keeping the alkali
treatment and molarity constants. In this case, Sodium was added to the solar cell device to compare the obtained efficiencies with the ones of previous experiments. . . 87 5.7. Batch 4 electrical characterisation results and correspondent variation. . 88 5.8. Best-efficiency devices regarding each of the produced solar cells from
Batch 4. . . 89 5.9. Variation of the alkali molarity, while keeping the oxide thickness and
alkali treatment unchanged. In this experiment, the molarity of the used alkali treatment was altered by changing the salt concentration of the spin-casted solution. . . 90 5.10. Batch 5 average electrical characterisation results. . . 91 5.11. Highest efficiency solar cell values. . . 92 5.12. Electrical characterisation results for the best performant 3-stage solar
cell devices. Cell ID 0nm/5nm corresponds to the reference cell without an oxide layer, while ID 0nm/5nm refers to the passivated cell with a 3nm-thick passivation layer. Furthermore, both samples have a 5nm-thick NaF layer deposited via thermal evaporation. . . 95 5.13. Two-diode model simulations parameters . . . 103
Mo work function: 4.30eV [16]. Hafnium oxide data: [17] . . . 113 6.3. SCAPS simulation results regarding the variation of the CIGS thickness,
while keeping the passivation layer thickness constant (2nm). . . 115
1 General Introduction
1.1. Background
Solar energy is the most extensive, cost-effective and direct available source of power, and it can be used in the most varied ways. It is estimated that 120,000 TW hit the earth, far transcending the most demanding energy consumption scenarios and showing the viability for a further increase in solar cell energy installation. [18]. The most prominent exploitation of direct sunlight is aimed at heating and lighting buildings. Addition- ally, through the use of solar cells, it is possible to convert solar power into electricity or heating directly. Therefore, solar energy offers two significant possibilities, a solar thermal route and a Photovoltaic (PV) route. PV electricity generation is experiencing a crescendo, and consequently being widely introduced on grids worldwide.
According to the last International Energy Agency (IEA) publication (2015),∼70% of the global power consumption is derived from fossil fuels in the form of oil, coal and natural gas, [19]. These sources are non-reusable, non-sustainable and highly pollutant, leading to the increase of greenhouse gases emission and consequently, the world temperature increase that will trigger many other environmental issues. Thus, a transition towards renewable and sustainable energy sources is highly recommended.
Despite all the possible uses and abundance of collectable power, direct solar technolo- gies only account for 2.2% of the 2016 world energy supply [19]. From this 2.2% market share, 92% regards Silicon (Si) solar cells.
The development of the Si technology marked the beginning of the solar cell era. Tech- nological advances, combined with mass-production ability, led to dramatic price reduc- tions. Furthermore, both efficiency and reliability have experienced a crescendo since the start of its commercialisation. Mono-crystalline Si solar cells currently hold record efficiencies of 26.77% and 24.4% for lab-scale and terrestrial modules, respectively [20].
However, Si-wafer based solar cells have inherent limitations, mainly regarding material consumption, production throughput, the requirement for high-temperature processing and the capability to meet the requests of new market’s demands such as flexibility and form factor [21].
Nowadays the market attention is turned towards the Thin-Film Solar Cells (TFSC), and in how to reduce the efficiency gap between these and the well-established Si devices.
The increase in the market share and consequently in research is mostly due to the emergent technologies, such as wearable and auto that require a small area, flexible
and high-efficiency solar cells. TFSC devices offer a solution for the aforementioned Si limitations by integrating high absorption coefficient materials as absorbers, which results in an inherent reduction of the absorber layer thickness. Such a decrease is likely to enhance fabrication yield and throughput while contributing to a more cost-effective process by reducing the amount of utilised materials. Regarding its market share, TFSC account for the remaining 8% [22].
Commercially available Thin-Film (TF) devices mainly employ three different materi- als as absorbers: 1) amorphous Si, 2) Cadmium Telluride (CdTe) and 3) Cu(In, Ga)Se2
(CIGS). All the presented materials are polycrystalline, and therefore can be successfully produced by low-temperature processes, when compared to Si devices.
TFSC offer diverse and promising options to substantially reduce the cost of PV systems and devices [23]. Furthermore, TFSC offers a wide variety of choices in terms of the device design and used fabrication processes.
A variety of substrates (flexible or rigid, metal or insulator) can be used for deposition of different layers (contact, buffer, absorber, reflector) using different techniques (Physical Vapour Deposition (PVD), Chemical Vapour Deposition (CVD), Electro-Chemical Depo- sition (ECD), plasma-based, hybrid). Such versatility allows tailoring and engineering of the layers in order to improve device performance [24], as well as to confer a weight reduction. Despite, all the possibilities and advantages, TF efficiencies are still behind the ones obtained for conventional Si devices. For a more detailed comparison regarding up-to-date efficiency records, kindly refer to [20].
Within TF devices, CIGS and CdTe are, so far, the most promising technologies. The potential of a specific device can be measured according to the maximum theoretical efficiency, based on the Schottky-Queiser limit, as well as the record efficiency in addi- tion to the expected payback time. Therefore, the highest the theoretical maxima the higher is the maturity stage for a specific technology. Furthermore, one looks for higher efficiencies and lower payback times, which are mostly driven by cost-reductions in device manufacturing. This work focus on the development of the CIGS technology, whereas advancements concerning CdTe devices can be found elsewhere [25].
According to [20], CIGS Solar Cell (SC) hold a 22.9% efficiency record achieved by Solar Frontier.
1.2. Problem
As a continuous effort to decrease the cost of the devices, CIGS solar cells include an ultra-thin absorber layer (500 nm or lower), allowing to reduce both deposition time, material consumption and consequently process costs.
Several groups [26,27] have shown a reduction in efficiency, mostly caused by a decrease in Jsc, which is associated to an insufficient charge collection due to the reduction of the absorber’s thickness and back-surface recombination. To date, most studies have been focused on trying to translate the technologies and techniques used in traditional Si cells
to TF/flexible cells. However, despite the developments, multiple difficulties have been encountered.
Different types of problems such as incompatibility of the deposition methods with the used substrate, materials mismatches or even the loss of functionalities like flexibility due to the deposited layers may arise.
The problem addressed in this work concerns the complexity and cost of the current fabrication processes. Therefore, this work focused on trying to establish a simple, cost-effective and reproducible procedure, by comparing a less complex 1-stage CIGS co-evaporation step with the state-of-the-art 3-stage one. Hence, if demonstrating that it is possible to achieve similar efficiencies using a more straightforward and less time- consuming process, it would be possible to purpose a more cost-effective baseline procedure.
As stated before, ultra-thin absorber layers are an effective way to reduce process costs but, unfortunately, lead to efficiency losses mostly due to insufficient light absorption by the CIGS layer. Moreover, this low efficiency occurs because of the high charge carrier recombination on both front and rear surfaces, as well as due to reduced optical confinement.
Therefore, different techniques will be studied to improve TF solar cells efficiencies by reducing charge carrier recombination, while simultaneously aiming to increase the optical confinement.
1.3. Motivation
Whenever considering the improvement of a SC device, it is necessary to ensure an appropriate solar cell architecture that enhances carrier collection and provides a stable band structure. The presented work is part of IMEC’s Advanced Thin-Film Photovoltaics (ATFPV) project, from which this thesis focuses on the development of novel SC archi- tectures to improve the current collection by employing rear surface passivation layers.
Therefore, the main purpose of the presented work is to find a suitable rear contact pas- sivation layer/mechanism for ultrathin/thin (500 nm) CIGS solar cells. The passivation mechanisms combines an innovative ultra-thin (2-6 nm) Hafnium Oxide (HfO2) layer, with two distinct alkali treatments (Sodium fluoride (NaF) and Potassium fluoride (KF)).
The alkali treatments play a critical role in achieving a functional SC. Since HfO2is a dielectric, and it is sandwiched between the Molybdenum (Mo) back contact and the CIGS absorber layer (as illustrated in Figure2.8), it is fundamental to create contact openings allowing the photogenerated current to flow through the device. Without an effective contact between the absorber layer and the back contact, the produced current cannot be collected, and consequently, no energy is generated.
As a first step multiple oxide thicknesses and alkali molarities were tested to fabricate and reproduce contact openings in the micro- nanometre range. Moreover, a comparison
between the two alkali treatments was drawn, allowing to define the best element to use in the creation of the point contacts.
1.4. Goal
The study of CIGS solar cells and the development of new architectures aim to produce high-efficiency, flexible and semi-transparent solar cells, while reducing its production cost, and consequently increasing the profit.
The main goal of this project is to develop industrially viable ways to passivate the rear surface of TFSC. To be industrially viable, the production costs need to be reduced.
Hence, it is necessary to modify the conventional CIGS solar cell structure consisting of the following multilayer stack: 1) Soda-lime Glass (SLG), 2) Molybdenum (Mo), 3) CIGS, Cadmium Sulfide (CdS), 4) Intrinsic Zinc Oxide (i-ZnO), and finally 5) Aluminium- doped Zinc Oxide (n-ZnO:Al).
The three biggest money drivers are identified as the materials costs and availability, the large scale uniformity and finally the production throughput.
1.4.1. Novel Passivation Layer
The introduction of a passivation layer in the device’s rear side modifies the typical cell architecture consisting of a SLG/Mo/CIGS/CdS/i-ZnO/n-ZnO:Al multilayer stack.
The final cell structure is then illustrated in Figure2.8.
The main difference refers to the introduction of an ultra-thin oxide layer that is respon- sible for the rear-side passivation. Charge carrier recombination at interfaces is one of the most significant limiting processes in the development of high-efficiency CIGS SC.
Surface passivation techniques aim to reduce recombination of diode characteristics, as well as photocurrent at the absorber’s interfaces [28]. Surface passivation results are especially noticeable for very thin or non-graded absorbers, where simply a back surface field induced by bandgap grading as proposed by Lundeberg et al., [29], is not sufficient to prevent minority carrier recombination at the rear interface.
Over the last years, research focused on improving the absorber’s back interface since the buffer Cadmium Sulfide (CdS) layer partly acts as a passivation layer for the front side. Furthermore, multiple studies [30,31] show that rear surface passivation tech- niques improve carrier collection by reducing back contact recombination mechanisms.
Therefore, it is an efficient way to enhance the device’s conversion efficiency.
Previously, IMEC ATFPV team, studied and successfully developed active Aluminium Oxide (Al2O3) rear passivation layers, in which contact openings were created via Sodium Flouride (NaF) alkali treatments. Addition of alkali elements, mainly Na, has been proved to have beneficial effects on the bulk electronics properties, independently of the process stage in which it was introduced [32].
This work also has the objective to determine the effects of Hafnium Oxide (HfO2) treated with NaF as rear passivation, while simultaneously comparing the new oxide approach and the previous group work (Al2O3+NaF). Thus, it is fundamental to de- sign and build a simple and stable test structure, representative for the rear contact design, before developing and testing a complete solar cell. A rear test structure allows understanding which method is the most suited to deposit the alkali layer and form the contact openings. Additionally, other alkali treatment, Potassium Flouride (KF), is also studied. This new oxide-alkali treatment aims to increase the number of point contact openings while reducing their diameter. On one side, it is critical to increase the number of openings on the dielectric layer to avoid a current blockage, which will result in lower Jsc. On the other hand, high contact-opening densities, mostly constituted by micrometre-range openings, would possibly lead to the loss of the passivation effect.
Simultaneously a stability comparison between NaF and KF are also presented.
Previous work, [33], successfully achieved rear side passivation with the creation of nanosized point contacts, by combining an Atomic Layer Deposition (ALD) Al2O3back surface passivation layer and a Cadmium Sulfide (CdS) Chemical Bath Deposition (CBD) process. The CBD led to the formation of spherical particles on top of the Al2O3layer, that when removed generated the nanosized point contacts.
Previous studies [33,34] also show the influence of the absorber’s layer thickness on the final cell conversion efficiency. It has been shown that the thicker the absorber layer, the higher the efficiency. Further, this work has the objective to study the in- fluence of both HfO2and NaF layers’ thicknesses on the efficiency of the passivation layers and nano-sized contact opening process, but also on the overall cell conversion efficiency.
1.5. Results highlight
This work successfully achieved the proposed passivation effect by introducing a rear- side H f O2passivation layer in combination with two different alkali treatments, as well as utilising two distinct co-evaporation methodologies (1-stage and 3-stage).
Firstly, the research project succeeded in creating micrometre-size openings on varied oxide layer thicknesses (2-9nm) by employing both a spin-coated Potassium Fluoride (KF) pre-deposition alkali treatment combined with a Sodium Flouride (NaF) post- deposition treatment (PDT).
Secondly, regarding the 1-stage process, an 8.5% conversion efficiency was obtained for a passivated device comprising a 2nm-thick H f O2 rear passivation layer, a 0.6M KF pre-deposition treatment and finally a 7nm-thick NaF PDT. Such a result indicates a 4%
efficiency enhancement when compared to the average reference conversion efficiency value (4.5%).
Thirdly, the 3-stage co-evaporation mechanism led to a further enhancement of all the
key characterisation parameters and consequently to an improved conversion efficiency result. In this case, by combining a 3nm-thick H f O2 rear passivation layer with a 5nm-thick, thermally evaporated, pre-deposition alkali treatment a highest conversion efficiency of 9.8% was achieved. Besides, the higher obtained Voc, Jsc and efficiency values, the presented illuminated IV curves suggested a complete mitigation of both Shunt and Series resistance effects for the passivated curve.
Finally, it is possible to highlight that the great majority of the tested scenarios led to up-conversion efficiency values when regarding the passivated devices. Additionally, the presented External Quantum Efficiency (EQE) and Photoluminescence (PL) spectra strongly suggest an improved back contact quality and a reduced concentration of radiative recombination sites at the Molybdenum/CIGS interface due to the introduction of the rear-side H f O2passivation layer.
1.6. Outline
The presented Master Thesis report consists of 8 chapters, divided into three major parts.
The first (Chapter 2 and 3) regards all the theoretical background, used methods and materials as well as characterisation techniques.
Chapter 2introduces the reader to the specifics of the photovoltaic effect and principle of solar operation. Such a theoretical explanation introduces some of the mechanisms behind the conversion of solar light into electricity. Fundamental aspects such as the generation and recombination mechanisms are presented in detail, followed by a com- parison between direct and indirect bandgap materials, and why it is advantageous to use direct materials. In the second section, Chapter 2 gives an overview of CIGS solar cells, including its crystal structure, a detailed explanation of the cell architecture and commonly used materials, and finally an immersive explanation about the cell’s physics where topics like recombination and generation mechanisms, efficiency limits and loss mechanisms are included.
Further in the chapter, motivations to shift towards thinner cells and their scale-up challenges are introduced. As a final section of the chapter, the current state-of-the-art achievements and production methods are given.
Chapter 3introduces the methods and materials employed in this work and explains the various steps involved in cell fabrication. The passivation layer step is explained in detail.
Additionally, the used characterisation techniques such as Current-Voltage (I-V) and Capacitance-Voltage measurements (C-V), as well as Static Photoluminescence (STPL) and Time-Resolved Photoluminescense (TRPL) are introduced and briefly explained.
In the second part, the reader is introduced to the obtained results, which are split into three chapters: Passivation Layer Development (Chapter 4), Solar cell device analysis (Chapter 5) and a theoretical (SCAPS) study.
Chapter 4focuses on developing an efficient and stable rear passivation layer, aiming to
reduce the rear interface recombination on thin-film CIGS solar cells. Multiple attempts were made, using different passivation thicknesses and distinct alkali treatments to cre- ate the necessary contact openings on the oxide layer. Chapter 5 presents and compares the solar cell characterisation results.
Chapter 6summarises all the simulation work performed using a Solar Cell Capacitance Simulator (SCAPS) software. It was useful both as a pre-fabrication behavioural study, and post-characterisation comparison between the simulated (theoretical) and the exper- imentally obtained results.
Finally, Chapter 7 summarises and discusses the experimental findings, whereas Chap- ter 8provides future work possibilities and recommendations.
2 Theoretical Background
The given theoretical background has the objective to familiarise all readers to the pre- sented concepts and theories. This chapter starts with a generalised explanation of the photovoltaic effect, which consists of explaining the cell’s mechanism by illustrating the generation and recombination of the electron-hole pairs.
Next, the specific type of used solar cells is introduced and further developed into an explanation of its crystal structure, according to the solid-state physics. Subsequently, the proposed architecture is presented, and a comparison with Silicon technology drawn.
Finally, the chapter enumerates the dominant loss mechanisms and presents the state-of- the-art research that developed novel architectures to minimise such losses.
2.1. Photovoltaic effect and Principle of solar operation
The photovoltaic effect is the mechanism behind solar cells working principle. It consists of the generation of a potential difference, due to charge separation, at the junction of two different materials in response to electromagnetic radiation [1].
Up to the present time, solid-state junction devices, usually made of silicon, has domi- nated this field, profiting from the research and material abundance resulting from the semiconductor industry [35].
The photovoltaic effect can be divided into three fundamental processes:
1. Generation of charge carriers due to the absorption of photons by the junction materials.
When the cell’s material absorbs a photon, its energy is used to excite an electron from an initial energy level Ei to a higher energy level Ef, as illustrated in the figure2.1
Figure 2.1.:Photon’s absorption mechanism in a semiconductor material with bandgap Eg. The photon with energy Eph = hν excites an electron from Eito Ef. Moreover, at Eia hole is created. (b) If Eph> Eg, a part of the energy is thermalised. Extracted from [1]
Due to the quantised energy states, photons are only absorbed if the difference between the intrinsic and Fermi levels of energy matches the photon energy (hν
= Ef - Ei). In the so-called ideal semiconductors, the population of electronically- available states by electrons and holes are allowed below the Valence Band (VB) edge and above the Conduction Band (CB) edge. Consequently, between these two edges no energy states are allowed, and therefore no electron population exists.
Hence, this energy difference is called the bandgap, Eg=EC- EV.
However, in real semiconductors, the VB and CB edges are not flat, since they vary with the electron’s momentum described by a k-vector, as illustrated in figure2.2 (A).
Thus, the energy of an electron is dependent on its momentum because of the periodic structure of the semiconductor crystal. Furthermore, the material can be characterised according to the position of the VB maximum, relatively to the CB minimum. If the maximum and minimum have the same k-value, the semiconduc- tor has a direct bandgap. Otherwise, it is an indirect band gap material. Such a distinction can be observed in figure2.2. The later occurs when the electron cannot be excited from the VB into the CB without changing its momentum, i.e. without an additional phonon excitation or release. Thus, in the indirect band case, the electron can only be excited by changing its momentum, through exchanges with the crystal due to vibrations of the crystal lattice [36].
Figure 2.2.:(A) Direct band gap material. (B) Indirect band gap material. The schmatic represents the electron thermal excitation mechanism. The electron from the occupied state at the VB is excited to the CB, where it occupies the previous unoccupied position. Extracted from [2]
Whenever an electron is excited from the VB into the CB a hole is created in the VB. Then, the absorption of a photon leads to the creation of an electron-hole pair.
The described mechanism is illustrated in the figure2.3, (Step 1).
Furtherance, the photon’s radiative energy is converted into the electron-hole pair’s chemical energy, which efficiency is limited by thermodynamics [36].
Figure 2.3.:Electron-Hole pair generation and recombination in a solar cell. Ex- tracted from [3].
2. Subsequent separation of the photo-generated charge carriers in the junction After its creation, the electron-hole pair tend to recombine, i.e. the electron will fall back to its initial position as illustrated on the second step in Figure2.3, and
the energy released in two different ways. It can be either released as a photon (radiative recombination) or transferred to surrounding electrons or holes, as well as creating lattice vibrations (non-radiative recombination).
In most solar cells, it is necessary to use semipermeable membranes. These mem- branes are formed by n- and p-type materials, forming the so-called PN junction.
3. Photo-generated charge carriers collection at the junction terminals. The charge carriers are collected creating a current along the device (Illustrated in Figure 2.3, step 4). Then, the electron-hole pairs’ chemical energy is converted into electrical energy. Furthermore, after passing through the circuit, electrons and holes recombine at the metal-absorber interface (Figure2.3, step 5).
2.2. CIGS solar cells
2.2.1. Introduction
Copper Indium Selenide (CIS) is a ternary p-type absorber material which belongs to the I-III-VI family [37]. The first ever CIS cell was produced in 1953, followed by the achievement of a 12% efficiency device [38]. Following the trend, in 1976, the first CIS thin film solar cell with a CdS buffer layer was developed, achieving a 5% efficiency [39].
The first significant interest wave happened in 1981 when Mickelsen and Chan, [40], achieved an efficiency of 9.4% by using a novel co-evaporation procedure.
The next step was the introduction of Gallium (Ga) in the previous CIS alloy, becoming CIGS. The alloying of Ga into the CIS compound gave rise to meaningful improvements regarding the device efficiency since it allowed to engineer the CIGS layer bandgap from 1eV (pure CIS - Ga free) to 1.7eV (pure CGS-In free) [6]. Chalcogenide solar cells are the third class of TF solar cells. This work focus is on copper indium gallium selenide (CIGS) chalcogenide class. The chalcogenides term arises from the used chemical compounds consisting of at least one negatively charged chalcogen ion from group 16, which have one or more electropositive elements [41].
There are different groups of chalcogenide solar cells. The one relevant to this work are the chalcopyrite solar cells, which name is based on the class of materials used - chal- copyrite (Copper Iron Disulphide (CuFeS)). Additionally, regarding its crystallographic structure, it forms tetragonal crystals, as illustrated in figure2.4(b).
2.2.1.1. Crystal Structure
The chalcopyrite crystal structure is similar to the one of a zincblend, as illustrated in Figure2.4. The last has a cubic close packed, i.e. face-centered array, which is charac- terised by single-bonds between the enclosed atoms and a constant 1:1 ration between
the cations and anions. Moreover, the CIS Bravais lattice is body-centered tetragonal, while the Zinc blend structure has a face-centered Bravais lattice.
Hence, it is possible to afirm that the chalcopyrite structure is a superlatice of the zincblend structure with a c/a ratio close to 2, meaning that the unit cell is doubled along the crystal c-axis. The c/a ratio represents the tetragonal deformation, which varia- tions are intimately related with different bonding strenghts within the crystal since the strenghts of the Cu−Se, In−Se and Ga−Se bonds are different [42]. Thus, the smaller the ratio, the more closely packed the tetragonal unit cell is.
Furthermore, from an organic chemistry perspective, the presented tetrahedral coordi- nation indicates the predominancy of sp3hybridised covalent bonds, combined with a lower percentage ionic bonding due to the presence of distinct ionic species, which carry a different ionic character. Therefore, due to the superlattice behaviour, the low-valency cations in a regular zincblend structure are replaced by higher-valency ones (In/Ga) in the calcopyrite tetrahedral structure. Additionally, an extra original cation is replaced by a lower-valency specie, which occupies the cation sublattice and leading to the formation of the ordered structured in Figure2.4.
(a) Zinc Blend crystal structured illustrated by the most well-known example the ZnS structure.
(b) Tetrahedral chalcopyrite structure (CIS) before Ga doping.
Figure 2.4.:Comparison between the primary stage (ZnS structure) and the evo- lution to the Chalcopyrite Tetrahedral crystal (CIGS). The different colours represent the different atoms, as indicated in the figure. Ex- tracted from [4]
The altered and reduced symmetry, mostly due to the previously discribed cationic interchange, results in the creation of an eight-atoms primitive-cell, when compared
with a two-atom primitive cell in the zinc blend structure.
By dopping the previous CIS structure with Ga, via interstitial replacement of defects, CIGS is achieved, as illustrated in Figure2.5.
Figure 2.5.:Doped chalcopyrite crystal structure (CIGS). The Cu(In, Ga)Se2qua- ternary compund is formed when Gallium atoms partially substitute Indium atoms in the ternary CuInSe2system.
The oxidation numbers for the presented ionic species are +1 (Cu), +3 (In and Ga) and -2 (Se).
2.2.1.2. Defect physics
The ternary compound, CuInSe2is doped by its native defects, and therefore without needing an external doping source to specify the type of conduction. Consequently, the intrinsic defects on the deposited film shape the dominant donor type. The p-type CIGS semiconductor nature is caused by a Copper (Cu) defficiency, i.e. the existance of Cu vacancies (VCu). This intrinsic deffect, when formed, induces an extra energy level close to the valence band, (about 30 meV above the valence band), similar to the acceptor level in extrinsic semiconductors [12].
Cu(In, Ga)Se2have a high tolerance to off-stoichiometric compounds, mainly in intersti- tial positions without jeopardising its electronic properties. This highly advantageous characteristic is based on the two Copper (Cu) vacancies (VCu) with an Indium (In) on the Cu antisite defect (InCu), i.e. the 2VCu+InCu intrinsic defects [43,44], which can have very low or even negative formation energies, and therefore can be spontaneously formed after an equilibrium position is reached. Besides the three already mentioned, nine additional native point defects are present in undoped CuInSe2. Three donors - Selenium-vacancy (VSe); Copper interstitial (Cui); and Indium interstitial (Ini); two acceptors - Indium vacancy (VIn) and Selenium interstitial (Sei). Further the antisite disorder in the cation sublattice, similarly to the donor InCu, also criates the acceptor Copper atom in the Indium site CuIn. Additionally, the donors Selenium in the Copper site (SeCu), Indium in Selenium sites (InSe), the acceptors Selenium in the Indium site (SeIn) and Copper in Selenium sites (CuSe) are also present for higher energies.
The various point defects, corresponding electrical activity and position in the band-gap are illustrated in Table2.1and Figure2.6.
Table 2.1.:Twelve possible point defects and respective oxidation number. Ex- tracted from [11].
Figure 2.6.:Electronic levels of intrinsic defects in the CuInSe2 ternary com- pound. Black columns represent acceptor levels, while white ones regard the donor- type levels.Extracted from [12].
Most of the chalcopyrites are semiconductors since formed by elements from the first, third and sixth groups. Therefore, theoretically, all the following combinations are possible to be formed and used, i.e. the conceived material will have semiconductor properties.
Most commonly a mixture called copper indium gallium diselenide [Cu(InxGa1−x)Se2, CIGS], where x can vary between 0-1 is used on solar cells fabrication. Another possibility is to add sulphur to the previous mixture. Now, copper indium gallium diselenide/disul- phide [Cu(InxGa1−x)(SeyS1−y)2, CIGSS] is used. X varies between 0 and 1.
One of the most important features regarding CIGS solar cells is its tunable bandgap.
By tunning the x and y values, i.e. By tuning the In:Ga ratio x and the Se:S ratio y, the bandgap of CIGS can be tuned from 1.0 eV to 1.7 eV, with the optimum value of 1.5eV, [45]. This bandgap tuning is possible due to the combination of spin orbit coupling, the crystal field of the tetragonal structure, and the influence of Cu-3d electrons on the valence band [46]. Thus, the CIGS quaternary compound semiconductor is an excellent option to be used as an absorber material in solar cells. However, the CIGS conduction mechanism is entirely different when compared to elemental semiconductors as Si and Ge. The conductive behaviour occurs due to the non-stoichiometry between acceptors and donors [47].
Furthermore, CIGS(S) is a direct bandgap semiconductor material, which has a broad absorption coefficient, and therefore absorber’s thicknesses much lower than the ones required for silicon wafers, in crystalline silicon solar cells, can be used. The broad ab- sorption coefficient is a crucial feature towards the optimisation of solar devices, because it not only allows to reduce the amount of used material and so to turn the process more cost-effective, but also by depositing a thinner absorber layer it is possible to explore novel characteristics, such as flexibility and transparency of the produced solar devices.
The typical diffusion length values for electrons in a CIGS absorber layer are on the order of few micrometres, and so corroborates the previous assumption that a very thin absorber layer, within a 1−5µm thickness range, is enough to ensure a sufficient absorption of the incoming light fraction, with energies above the bandgap value. Stud- ies, [48], corroborate the effectiveness of CIGS absorber layers. Furthermore, it holds the highest value of solar radiation absorption, when compared with Amorphous Silicon (a-Si) and CdTe solar cells. CIGS cells have an absorption efficiency close to 100% at the first 3−4µm, being 95% of the incoming radiation absorbed in its 0.4−0.5µm. Figure 2.7shows the best researched solar cells, using different technologies, as reported so far.
Figure 2.7.:Best solar cell efficiencies achieved by different PV technologies. Re- leased on February 2019. Extracted from [5]
From the presented figure, it is easily understandable that among TFSC, the CIGS cell has the highest efficiency, corresponding to 22.6% for normal CIGS and 23.3% for CIGS cells using concentration technology.
2.2.2. Cell architecture and description
CIGS solar cells are constituted by a multiplicity of thin film layers with different operation purposes, in which Cu(In1−xGax)Se2nanocrystalline bulk semiconductor is used as the absorber material.
Figure2.8illustrates the developed CIGS structure.
Figure 2.8.:Solar cell design.
The CIGS structure fits in the so-called substrate configuration where the light enters from the top layer, a Transparent Conducting Oxide (TCO), passing through the CdS buffer layer until it reaches the absorber (CIGS) layer where it is absorbed. The light portion which is not absorbed reaches the back contact, commonly a Molybdenum Mo layer deposited on Soda Lime Glass (SLG).
CIGS-based solar cells present multiple advantages over the first generation of CIS solar cells. First, the bandgap can be tuned by adjusting the Ga/In ratio so that it can match the solar spectra energy range. If Ga replaces all In, the CIGS bandgap increases from 1.04eV to 1.68eV [49]. Further, the incorporation of Ga improves the Voc, since Voc ≈Eg/2 [37].
2.2.2.1. CIGS vs c-Si
Comparing CIGS cells with the Crystalline Silicon (c-Si) technology, the later needs a thick and rigid wafer (≈180µm) to have enough absorption to be commercialised [37], due to its low absorption coefficient (104cm−1) which results from having an indirect band gap [50]. In contrast, CIGS has an absorption coefficient beyond 105cm−1, and therefore requires a thickness hundred times lower than the one required for c-Si devices.
Regarding energy consumption, CIGS production uses lower thermal budgets (≈550oC) than c-Si technology (≈1100oC). Such reduction allows saving energy, turning the CIGS process into a more sustainable and economically viable operation [37,51], as well as it allows to reduce by an year the payback time of CIGS devices (1 year) when compared with c-Si cells (2years).
By reducing the absorber thickness, CIGS solar cells may become a highly competitive alternative for Si-based devices, as less crtitical raw materials, time and cost are involved in the development and production steps [52]. Further advantages and motivations to reduce the absorber’s layer thickness are discussed in2.2.6
Furthermore, CIGS technology uses more cost-effective materials, and so it has the potential to overcome the costs of the c-Si product [53], mostly due to the need of growing large and pure Si crystals which is both expensive and time consuming.
Additionally, the multiplicity of different layers in CIGS solar cells have a positive impact on Jsclosses. Further advantages arise from the possibility to monolithically integrate CIGS cells, whereas in c-Si devices it is not possible.
Another CIGS advantage regards the use of a lower contact thickness. The contact thickness is a fundamental device aspect mostly due to its influence on improving ohmic contacts, and consequently reducing the series resistance. Therefore, it is necessary to achieve a minimum grid thickness. In conventional c-Si devices the grids and rear Alluminium contacts are approximaatelly 20 and 40 µm, respectivelly, which is higher the ones used in the development of CIGS technology.
However, Si stability and efficiencies are still unmatched, mostly due to the research made to improve collection losses as well as in the field of encapsulation techniques.
Hence, c-Si solar cells are stable for as long as 20 years. Oppositely, CIGS encapsulation is still an ongoing research topic, and therefore not enough data is available to perform a valid comparison.
The typical cell is formed by low price substrates such as glass, polyimide foils and cost-effective metals such as stainless steel.
1. Substrate
The substrate is fundamental to the device’s development since the deposition of the Mo layer on the substrate of choice defines the selenisation conditions, such as temperature.
SLG is the most used material due to its smooth surface, stability, electrical insulat- ing features and more importantly its affordable price [54].
Additionally, SLG is preferred over regular glass because it is a source of Na to the absorber during both co-evaporation and selenisation processes. Sodium con- tamination has proven to be related to the increase in the cell’s overall efficiency.
A 0.1 at% Na supply benefits the CIGS devices regarding an increase of Vocand Fill-Factor (FF), leading to an efficiency increase [55]. The reason for the solar characteristics enhancement is mostly due to the passivating effect of sodium on the defects at the CdS buffer layer and CIGS junction [56,57].
Concerning physical parameters, SLG has a matching thermal expansion coefficient to the one of the CIGS material (5·10−6and 12·10−6K−1), which is fundamental to avoid adhesion and mismatch problems during the CIGS deposition process at high-temperatures [58]. Moreover, regarding physical aspects, SLG is the option of choice since it has good adhesion, low weight, and can undergo the processing temperatures without degradation.
2. Back contact
The metal back contact is the first layer to be deposited on the substrate. It is responsible for the optical reflectance, as well as to deliver the free carriers to the load (Figure2.3).
Few materials have been proposed as back contacts such as Mo, Tungsten (W), Tantalum (Ta), Manganese (Mn) and Titanium (Ti). However, whenever SLG is used as the substrate, Mo is by far the preferred one, due to its low contact resistance to the absorber layer and stability along the entire process, i.e. Mo does not react strongly with CIGS, and it does not degrade during the high-temperature CIGS deposition process. Such characteristics proved to limit the atomic diffusion, and therefore minimising adverse reactions during layers growth [59].
Multiple techniques are available to deposit the Mo layer on a rigid substrate, such as SLG. Direct Current (DC) and Radio Frequency (RF) sputtering are the most common choices due to characteristics of the obtained films, but other techniques such as Ion-beam sputtering and High-target-utilization sputtering (HiTUS) are also used [60]. The used deposition method and parameters strongly influence the quality of the deposit Mo thin film.
Furthermore, an ideal Mo back contact to be used in CIGS solar cells must be conductive, stress-free, well adherent and uniform. Single Mo layers can cause excessive tensile or compressive stress. Thus, a Bilayer structure is used to reduce residual stress [61].
Additionally, when depositing in large surface areas, crystalline Mo thin films are preferably grown with a (110) crystallographic orientation [62]. Therefore, by adjusting the deposition conditions regarding discharging power, working pressure, substrate temperature, and target-to-substrate distance the previous ideal characteristics can be achieved [63].
