ACTAUNIVERSITATIS

UPSALIENSISUPPSALA

2020

Digital Comprehensive Summaries of Uppsala Dissertationsfrom the Faculty of Science and Technology 1944

Defect Engineering in KesteriteMaterials for Thin Film Solar Cells

KATHARINA RUDISCH

ISSN 1651-6214ISBN 978-91-513-0963-7urn:nbn:se:uu:diva-407820

Dissertation presented at Uppsala University to be publicly examined in Polhemsalen,Ångströmlaboratoriert, Lägerhyddsvägen 1, Uppsala, Wednesday, 20 May 2020 at 13:15 forthe degree of Doctor of Philosophy. The examination will be conducted in English. Facultyexaminer: Prof. Dr. Alain Lafond (Université de Nante).

AbstractRudisch, K. 2020. Defect Engineering in Kesterite Materials for Thin Film Solar Cells.Digital Comprehensive Summaries of Uppsala Dissertations from the Faculty of Science andTechnology 1944. 82 pp. Uppsala: Acta Universitatis Upsaliensis. ISBN 978-91-513-0963-7.

Cu2ZnSnS4 has great potential to be applied as an earth abundant and non-toxic absorbermaterial in thin film solar cells, based on its suitable optical properties. However, severalchallenges have prevented the achievable efficiencies from exceeding 12.6 %, which is wellbelow marketable efficiencies compared to competing solar cell technologies. One of thestruggles in the development of Cu2ZnSnS4 solar cells is the high number of harmful defectsleading to severe potential fluctuations. This thesis investigates different strategies of defectengineering in Cu2ZnSnS4, in particular to reduce Cu-Zn disorder.

Cu2ZnSnS4 thin films are produced by a two-step process, where Cu-Zn-Sn-S precursors aredeposited by co-sputtering and then annealed at high temperature to yield crystalline films.The material properties are investigated with Raman spectroscopy, photoluminescence andspectrophotometry.

In the scope of this thesis, the following approaches to defect engineering are investigated:thermal treatments, varying partial pressures during the annealing step, and cation exchange toform the compound Cu2MnSnS4. Thermal treatments substantially enhance the degree of orderin Cu2ZnSnS4. However, for the first time the severe limitations of such treatments are shown,indicating their insufficiency to reduce cation disorder to a level where potential fluctuations nolonger affect Cu2ZnSnS4 solar cells. Furthermore, the stannite Cu2MnSnS4 suffers from cationdisorder just like kesterite Cu2ZnSnS4 demonstrating that cation disorder is not restricted to thekesterite crystal structure and posing new challenges for finding new solar cell materials.

On the other hand, the presented results demonstrate a strong composition dependence of theordering kinetics. Compositions with high densities of vacancies or interstitials significantlyenhance the ordering rate by reducing the activation energy while the critical temperatureis constant for the investigated compositions. Furthermore, the important effect of S2 andSnS partial pressures during the annealing step of the fabrication is predicted from chemicalmodels and experimentally verified by investigation of composition-spread Cu2ZnSnS4 thinfilms. Increasing both partial pressures leads to higher solubility of vacancies in Sn-richCu2ZnSnS4 further amplifying the positive effect of composition on the order-disorder transition.Investigation of composition-spread thin films further revealed the interplay between materialproperties and composition as well as secondary phases. In particular, the photoluminescenceyield was drastically enhanced in the presence of SnSx secondary phases. This thesis discussesthese results in the context of the current understanding of Cu2ZnSnS4.

Keywords: CZTS, thin film solar cells, defect engineering, composition-spread films, orderdisorder transition, intrinsic defects, secondary phases

Katharina Rudisch, Department of Materials Science and Engineering, Solar CellTechnology, Box 534, Uppsala University, SE-751 21 Uppsala, Sweden.

© Katharina Rudisch 2020

ISSN 1651-6214ISBN 978-91-513-0963-7urn:nbn:se:uu:diva-407820 (http://urn.kb.se/resolve?urn=urn:nbn:se:uu:diva-407820)

Errata: The printed version of this thesis was not assigned isbn number and series number.

“Don’t let anyone rob you of your imagination,your creativity, or your curiosity.

It’s your place in the world; it’s your life.Go on and do all you can with it, and make it the life you want to live.”

Mae JemisonFirst female African American astronaut in space

List of papers

This thesis is based on the following papers, which are referred to in the textby their Roman numerals.

I Order-disorder transition in B-type Cu2ZnSnS4 and limitations of

ordering through thermal treatments.

Katharina Rudisch, Yi Ren, Charlotte Platzer-Björkman, and JonathanScragg, Applied Physics Letters 108 231902 (2016)

II The effect of stoichiometry on Cu-Zn ordering kinetics in

Cu2ZnSnS4 thin films.

Katharina Rudisch, Alexandra Davydova, Charlotte Platzer-Björkman,and Jonathan Scragg, Journal of Applied Physics 123, 161558 (2018)

III The Single-Phase Region in Cu2ZnSnS4 Thin Films from Theory

and Combinatorial Experiments.

Alexandra Davydova, Katharina Rudisch, and Jonathan J. S. Scragg,Chemistry of Materials 30, 4624-4638 (2018)

IV Structural and Electronic properties of Cu2MnSnS4 from

Experiment and First-Principle Calculations.

Katharina Rudisch, William F. Espinosa-García, Jorge M.Osorio-Guillén, Carlos M. Araujo, Charlotte Platzer-Björkman, andJonathan J. S. Scragg, Physica Status Solidi B 1800743 (2019)

V Prospects for defect engineering in Cu2ZnSnS4 solar absorber

films.

Katharina Rudisch, Alexandra Davydova, Lars Riekher, JoakimAdolfsson, Luciano Quaglia Casal, Charlotte Platzer-Björkman,Jonathan Scragg, submitted manuscript (2020)

Reprints were made with permission from the publishers.

Author’s contributions

I Substantial contributions to the design of the experiment, samplefabrication, sample characterization and data collection, data analysisand interpretation, paper drafting and writing with input from co-authors.

II Substantial contributions to the design of the experiment, substantialcontribution to sample characterization and data collection, data analysisand interpretation, paper drafting and writing with input from co-authors.

III Contributed to sample preparation (with thermal ordering treatments),supported analysis of Raman mapping and contributed to the discussionof the order parameter, input to manuscript writing.

IV Substantial contributions to experimental design, sample fabrication,sample characterization, experimental data analysis and interpretation,paper drafting and major part of writing with input from co-authors.

V Contributions to experimental design, supervision of a master thesisstudent, part of sample fabrication, part of sample characterization anddata collection, data analysis and interpretation, paper drafting andwriting with input from co-authors.

Related work

The following contributions were made during my studies, but are not includedin this thesis:

1. S.Y. Li, C. Hägglund, Y. Ren, J. Scragg, J.K. Larsen, C. Frisk, K. Rud-isch, S. Englund, C. Platzer-Björkman, “Optical properties of reactivelysputtered Cu2ZnSnS4 solar absorbers determined by spectroscopic el-lipsometry and spectrophotometry.” Solar Energy Materials and SolarCells 149 pp. 170-178 (2016)

2. Y. Ren, N. Ross, J.K. Larsen, K. Rudisch, J. Scragg, C. Platzer-Björkman“Evolution of Cu2ZnSnS4 during Non-Equilibrium Annealing with Quasi-in Situ Monitoring of Sulfur Partial Pressure.” Chemistry of Materials29 (8), pp. 3713-3722 (2017)

3. N. Ross, S. Grini, K. Rudisch, L. Vines, C. Platzer-Björkman “Sele-nium Inclusion in Cu2ZnSn(S,Se)4 Solar Cell Absorber Precusors forOptimized Grain Growth.” IEEE Journal of Photovoltaics 8 (4), pp.1132-1141 (2018)

4. A. Davydova, J. Eriksson, R. Chen, K. Rudisch, C. Persson, J. Scragg“Thio-olivine Mn2SiS4 thin films by reactive magnetron sputtering: Struc-tural and optical properties with insights from first principles calcula-tions.” Materials & design 152, pp. 110-118 (2018)

Contents

Sammanfattning på svenska . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

Zusammenfassung auf Deutsch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191.1 Kesterite for thin film photovoltaics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201.2 Scope and aim of this thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

2 Structural properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 262.1 The CZTS crystal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 262.2 Fabrication of thin film CZTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 262.3 Models for phase stability and decomposition into secondary

phases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 302.4 Experimental investigation of the single phase region . . . . . . . . . . . . . . . . 33

3 Electronic properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 393.1 The band gap of CZTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 393.2 Defects in CZTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 413.3 Defect complexes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44

4 Cation disorder in CZTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 484.1 Disorder in CZTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 484.2 The concept of disorder in solids – putting a number on

disorder . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 504.3 Cu-Zn disorder in kesterite CZTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54

5 Defect Engineering in CZTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 585.1 Reducing Cu-Zn disorder in CZTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 585.2 Alloying CZTS with other elements for defect engineering . . . . . . . 615.3 Exploring cation exchange for materials with preferable

properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63

6 Concluding remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 666.1 Summary of the results of this thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 666.2 Outlook . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68

Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72

Sammanfattning på svenska

Styrning av defekter i kesteritmaterial för tunnfilmssolceller

Greta Thunberg rör för närvarande vid en öm punkt världen över med sinadelvis aggressiva uppmaningar att äntligen uppfatta klimatkrisen som en sådanoch agera för att bromsa den dramatiska globala uppvärmningen. En viktigbyggsten för en mer hållbar livsstil är övergången från fossila till förnybaraenergikällor. Redan nu erbjuder solceller en billig och praktisk lösning för enomfattande energiförsörjning baserad på solljus.

Idag dominerar solceller av kristallint kisel solcellsmarknaden. Produktio-nen av kiselsolceller kräver relativt stora mängder material och energi, dettapå grund av att kisel inte absorberar ljus särskilt bra. Därför behöver kiselsol-celler vara relativt tjocka vilket begränsar tillämpningarna och möjligheternatill kostnadsminskning i produktionen. Detta står i kontrast till tunnfilmssol-celler som tack vare mycket god förmåga att absorbera ljus klarar sig med entjocklek på mindre än en tiondels hårstrå.

Det finns olika typer av tunnfilmssolceller. Det som skiljer dem åt är fram-för allt materialet som utgör det ljusabsorberande skiktet. Vissa typer av tunn-filmsolceller har verkningsgrader som liknar dem för solceller av kristallintkisel, framför allt de som består av Cu(In,Ga)Se2, ofta kallat CIGS, och CdTe.Både CIGS och CdTe-solceller tillverkas också kommersiellt. Dessa materialär emellertid delvis gjorda av sällsynta eller giftiga ämnen, vilket gör pro-duktionen dyrare, mer komplicerad och på sikt riskerar att begränsa mängdensolceller som kan produceras.

Materialet Cu2ZnSnS4 (CZTS) består enbart av grundämnen som finns istora mängder i jordskorpan och är inte giftigt. CZTS har ett direkt bandgappå ca 1.5 eV, vilket ger hög absorbtionskoefficient, dvs att det absorberar ljusväldigt bra. Därför har den stor potential att kunna användas som ett ljusab-sorberande skikt i tunnfilmssolceller.

Den bästa verkningsgraden som uppnåtts för CZTS tunnfilmssolceller hit-tills är 12,6 %. Detta är emellertid fortfarande för lågt för att konkurrera medandra tunnfilmssolceller. Ett stort hinder för att nå bättre verkningsgrad är demånga kristalldefekter som orsakar rekombination av laddningsbärarna innande kan bidra till kretsen. Denna doktorsavhandling undersöker därför olika sättatt manipulera och kontrollera kristalldefekter i CZTS.

CZTS-proverna som undersöktes för denna avhandling framställdes genomså kallad samsputtring, där en substrat belades med tunna filmer bestående av

11

de fyra komponenterna koppar (Cu), zink (Zn), tenn (Sn) och svavel (S). I ettandra steg värmebehandlades de tunna filmerna i över 550 ◦C. Under dennaprocess arrangerar atomerna sig i en periodisk struktur, ett så kallat kristallgit-ter. I fallet med CZTS har detta gitter kesteritstruktur, som kan härledas fråndiamantstrukturen.

Andra föreningar kan också bildas under värmebehandlingen, så kallade se-kundärfaser, till exempel ZnS, CuS, SnS eller Cu2SnS3. Sekundärfaserna kanha negativa effekter på solcellens prestanda. Den här avhandlingen visar attCu3SnS4 har en särskilt negativ effekt på CZTS materialegenskaper, medanSnSx har en positiv effekt och minskar rekombinationsprocesserna i CZTS. Isynnerhet måste de förhållanden som råder under värmebehandlingen kontrol-leras mycket noggrant för att förhindra att CZTS sönderdelas till andra faser. Iavhandlingen demonstreras att det är fördelaktigt att berika atmosfären undervärmebehandlingen med S och SnS i gasform för att minimera avdunstning avdessa komponenter från CZTS-proven.

Den elektroniska bandstrukturen av ett material beskriver hur laddningsbä-rare (till exempel elektroner) kan röra sig i kristallen och bestämmer förmå-gan till ljusabsorption hos materialet. I en halvledare har bandstrukturen ettbandgap, dvs ett gap mellan de elektroniska banden och elektronerna kan in-te befinna sig i det gapet. Det gör att ett material bara absorberar ljus medenergi större än bandgapsenergin. CZTS absorberar emellertid en relativt stormängd ljus i energiområdet under bandgapsenergin, vilket det inte borde. Det-ta indikerar att det finns många kristallfel materialet, som bildar energinivåer ibandgapet. Dessa energinivåer kan orsaka att rekombinationshastigheten ökardramatiskt. När laddningsbärare rekombinerar betyder det att ledningsbärareförlorats innan de kan bidra till kretsen.

En kristall kan ha många olika gitterdefekter: vissa bildas på grund av attden kemiska sammansättningen avviker från det ideala förhållandet. Om detfinns för lite koppar, till exempel, kan kopparvakanser bildas i kristallen. Eller,om det finns för mycket koppar kan kopparatomer sätta sig som interstitia-ler, alltså mellan gitterplatserna. Om det finns för mycket koppar men för litezink kan kopparatomerna också ta upp gitterplatser där zink egentligen skasitta (även kallat substitution). Vissa gitterfel påverkar materialets elektriskaledningsförmåga mer negativt än andra. Tenn på zinkplatser eller tenn på kop-parplatser bildar energinivåer som väsentligen kan öka rekombinationsgraden.Kopparvakanser å andra sidan inducerar ytterligare laddningsbärare i kristal-len utan att generera energinivåer som främjar rekombinationsgraden. Där-för förbättrar kopparvakanser ledningsförmågan hos CZTS. Genom att justerasammansättningen av koppar, zink och tenn kan vissa gitterdefekter undvikaseller stimuleras. Den här avhandlingen visar att CZTS har ganska hög gradav rekombination oberoende av atomernas sammansättning. Rekombintaions-graden är bara minskad i områden där det finns SnSx-sekundärfasen. Det ärganska oroande eftersom det betyder att defekten som orsaker rekombinatio-nen inte kan påverkas genom att ändra atomsammansättningen.

12

Koppar- och zinkatomerna har samma storlek i CZTS-kristallen. Detta gördet enkelt för dem att byta plats. Dessa substitutioner förekommer så oftai CZTS att i typiska CZTS-prover distribueras koppar- och zinkatomer heltslumpmässigt på gitterplatserna. Därför kallas det också Cu-Zn-oordning. Så-ledes är parets substitution Cu på Zn-plats (CuZn) och Zn på Cu-plats (ZnCu)det överlägset vanligaste gitterfel i CZTS. Eftersom båda substitutionerna upp-träder i lika stora andelar vid Cu-Zn-oordningen är Cu-Zn-oordningen obero-ende av förhållandet av koppar, zink och tenn.

Oordning i kristaller kan påverkas av temperaturen. Vid höga temperatu-rer har atomerna mer kinetisk energi och det är lättare för dem att lämna sinagitterplatser och bilda parsubstitutioner. Vid lägre temperaturer tenderar ato-merna att stanna kvar på sina faktiska gitterplatser. Men vid låga temperaturerhar de mindre energi att utföra en substitution. Eftersom CZTS kristalliserasvid ganska höga temperaturer på över 500 ◦C fördelas koppar- och zinkatomer-na i fullständig oordning efter värmebehandling. Under den snabba kylning-en “fryser” de flesta av de oordnande atomerna fast på gitterplatsen och vidrumstemperatur sker sedan nästan ingen substitution av atomerna. Däremotgör en långsam kylning att fler atomer hinner distribueras i det ordnade till-ståndet. Således kan termiska behandlingar avsevärt minska antalet oordnadekoppar- och zinkatomer. Baserat på teorier som beskriver oordning i kristal-ler kan förutsägelser göras om beteendet av Cu-Zn-oordningen. För CZTS hardet påvisats i den här avhandlingen att extremt långa värmebehandlingar påflera år skulle vara nödvändiga för att minimera effekterna av Cu-Zn-oordningpå solceller. Därför behövs nya lösningar för att främja effekten av termiskabehandlingar och ytterligare minska Cu-Zn-oordningen.

Cu-Zn-oordning kan också påverkas av andra kristalldefekter, eftersom des-sa kan påverka rörligheten hos Cu- och Zn-atomerna och därmed ändra hursnabbt atomerna kan ordnas om under en termisk behandling. Vakanser el-ler interstitialer underlättar substitutionen av atomerna och underlättar såledesomordningen av Cu- och Zn-atomerna. Således kan effekten av termiska be-handlingar förbättras och en högre grad av Cu-Zn-ordning kan uppnås. Va-kanser och interstitialer finns huvudsakligen i Cu-rika och Cu-fattiga CZTS iform av Cu-interstitialatomer, eller Cu-vakanser eller som Zn-vakanser ellerZn-interstitialer.

Dessutom kan förhållandena under värmebehandlingen justeras med högapartialtryck för SnS och S. Under dessa förhållanden ökar lösligheten för kop-parvakanser i CZTS och fler vakanser kan bildas. Detta leder till en ytterligareförbättring av ordningshastigheten under termiska behandlingar.

Ett annat sätt att minska Cu-Zn-oordningen är att ersätta ett av elementen. Idetta arbete ersattes Zn av mangan (Mn), som inte är giftigt och förekommer ihög koncentration i jordskorpan. Föreningen Cu2MnSnS4 (CMTS) kristallise-ras i stannitstrukturen, som är mycket lik kesteritstrukturen. Även om Cu ochMn skiljer sig åt i storleken av kristallen och CMTS kristalliseras i en annanstruktur än CZTS har det visats i den här avhandlingen att oordning av ato-

13

merna fortfarande finns i kvar i kristallen. Det förväntas därför att CMTS intekommer att ge en förbättring för användning i solceller jämfört med CZTS.Baserat på kunskapen från avhandlingen kan rekommendationer härledas förvidare sökning efter framtida solcellsmaterial.

14

Zusammenfassung auf Deutsch

Defekt-Engineering in Kesterit-Materialien fürDünnschichtsolarzellen

Greta Thunberg trifft derzeit weltweit einen Nerv mit ihren teils aggressivenAufforderungen doch endlich die Klimakrise als solche wahrzunehmen und zuhandeln, um die bisherige dramatische Erderwärmung zu bremsen. Ein wich-tiger Baustein für eine nachhaltigere Lebensweise ist der Umstieg von fossilenauf erneuerbare Energiequellen. Solarzellen bieten bereits jetzt eine günstigeund praktikable Lösung für eine weitreichende Energieversorgung auf der Ba-sis von Sonnenenergie.

Bisher wird ein Großteil der Solarzellen aus kristallinem Silizium herge-stellt. Für die Herstellung von Silizium-Solarzellen ist ein großer Material-und Energieaufwand nötig, was die Möglichkeiten für Anwendungen und zurKostenreduzierung in der Herstellung begrenzt. Dem gegenüber stehen Dünn-filmsolarzellen, welche mit einer kompletten Dicke von weniger als ein Zehn-tel vom Durchmesser eines Haares auskommen.

Technologien für Dünnfilmsolarzellen werden anhand des Materials unter-schieden, welches die Absorbierschicht ausmacht. Einige dieser Technologi-en, zum Beispiel Cu(In,Ga)Se2 (auch CIGS) oder CdTe basierte Solarzellen,haben Wirkungsgrade ähnlich dem von Silizium und werden bereits kommer-ziell hergestellt. Allerdings sind diese Materialien aus seltenen Erden oder gif-tigen Stoffen hergestellt, was die Herstellung komplizierter und teurer macht.

Das Material Cu2ZnSnS4 (CZTS) ist frei von seltenen Erden und ungiftigund hat zudem vorteilhafte Eigenschaften um als lichtabsorbierende Schicht(auch Absorber) in Dünnfilmsolarzellen angewendet zu werden. Es hat einedirekte Bandlücke mit einer Energie von etwa 1.5 eV, was zu einem hohen Ab-sorptionskoeffizientenführt, das heißt, dass CZTS Licht sehr gut absorbierenkann. Auf der Basis von CZTS als Absorbermaterial wurden bisher Solarzel-len mit bis zu 12.6 % Wirkungsgrad hergestellt. Dies ist allerdings noch zuwenig um mit anderen Dünnschichtsolarzellen konkurrieren zu können. EinHindernis für bessere Wirkungsgrade sind die vielen Gitterfehler, welche einehohe Rekombinationsrate zur Folge haben. Rekombination der Ladungsträ-ger bedeutet, dass diese verloren gehen, bevor sie zum Stromkreislauf beitra-gen können. Daher untersucht diese Doktorarbeit verschiedenen Möglichkei-ten zur Manipulation und Kontrolle von Gitterfehlern in CZTS.

Die CZTS Proben, welche für diese Arbeit untersucht wurden, wurdendurch Co-Sputtering hergestellt. Daraus resultieren Dünnfilme mit den vier

15

Bestandteilen Kupfer (Cu), Zink (Zn), Zinn (Sn) und Schwefel (S). In einemzweiten Schritt werden die Dünnfilme bei über 550 ◦C ausgeheizt. Dabei ar-rangieren sich die Atome in eine periodische Struktur, dem Kristallgitter. ImFall von CZTS hat dieses Gitter die kesterite Kristallstruktur, welche sich ausder Diamantstruktur ableiten lässt.

Beim Ausheizen können sich auch andere Verbindungen bilden, so genann-te Sekundärphasen, zum Beispiel ZnS, CuS, SnS, oder Cu2SnS3. Diese Sekun-därphasen können negative Auswirkungen auf den Wirkungsgrad der Solarzel-le haben. In dieser Arbeit wurde gezeigt, dass Cu3SnS4 sich besonders negativauf die Materialeigenschaften von CZTS auswirkt, während SnSx eine positi-ve Wirkung hat und die Rekombinationsrate in CZTS reduziert. Im Speziellenmüssen die Bedingungen, die während des Ausheizens herrschen, genau kon-trolliert werden, um den Zerfall von CZTS in Sekundärphasen zu unterbinden.Es ist zum Beispiel vorteilhaft die Atmosphäre während des Ausheizens mitgasförmigen S und SnS anzureichern, um Verdampfen dieser Komponentenvon der Probe zu minimieren.

Die elektronische Bandstruktur eines Materials beschreibt, wie sich die La-dungsträger (zum Beispiel Elektronen) im Kristall bewegen können und be-stimmt auch die Absorption eines Materials. In einem Halbleiter weist dieBandstruktur eine Bandlücke auf, also eine Lücke zwischen den energetischenNiveaus der elektronischen Bänder, in welcher sich die Elektronen nicht bewe-gen können. Das führt dazu, dass ein Material nur Licht mit höherer Energieals die Bandlückenenergie absorbieren kann. CZTS absorbiert jedoch relativviel Licht im Energiebereich unterhalb der Bandlückenenrgie, was darauf hin-weist, dass viele Gitterdefekte (oder auch Gitterfehler) im Material vorhan-den sind, welche zusätzliche Energieniveaus innerhalb der Bandlücke bilden.Diese Energieniveaus können die Rekombinationsrate gravierend anwachsenlassen und die Leitungsfähigkeit des Materials beeinträchtigen.

Ein Kristall kann viele mögliche Gitterdefekte aufweisen. Manche Gitter-defekte bilden sich aufgrund der abweichenden Komposition von dem idea-len Verhältnis der Atome. Gibt es zum Beispiel zu wenig Kupfer, könnensich Kupferleerstellen im Kristall bilden. Bei zu viel Kupfer im Material kön-nen sich Kupferatome als Zwischengitteratome ansiedeln. Wenn zu viel Kup-fer, aber zu wenig Zink vorhanden ist, können die Kupferatome auch Zink-Gitterplätze einnehmen (CuZn, Substitution). Manche Gitterfehler haben schwer-wiegende Auswirkungen auf die Leitfähigkeit eines Materials. Zinn auf Zink,oder Zinn auf Kupfer Substitutionen bilden Energieniveaus, welche die Re-kombinationsrate besonders anwachsen lassen können. Hingegen induzierenKupferleerstellen zusätzliche Ladungsträger im Kristall ohne Energieniveauszu generieren, welche die Rekombinationsrate fördern. Damit verbessern siedie Leitfähigkeit von CZTS. Indem das Verhältnis von Kupfer, Zink und Zinnangepasst wird, können so bestimmte Gitterdefekte vermieden oder stimuliertwerden.

16

In dieser Doktorarbeit wurde gezeigt, dass die Rekombinationsrate in CZTSunabhängig von der Komposition sehr hoch ist. Der Rekombinationsgrad konn-te nur durch die Präsenz der Sekundärphase SnSx verringert werden. DieseBeobachtung ist bersorgniserregend, da dies bedeutet, dass die Gitterfehler,welche zur Rekombination führen, nicht durch die Komposition von CZTSbeeinflusst werden können.

Die Kupfer- und Zinkatome im CZTS-Kristall haben die selbe Größe alsGitterionen und es fällt ihnen dadurch leicht, ihre Kristallplätze zu vertauschenund Substitutionen zu bilden. Diese Gitterfehler kommen in CZTS so häufigvor, dass in typischen CZTS Proben Kupfer- und Zinkatome komplett unge-ordnet auf den Gitterplätzen verteilt sind. Dies wird auch Cu-Zn-Unordnunggenannt. Damit ist die Paarsubstitution Cu auf Zn (CuZn) und Zn auf Cu (ZnCu)der bei weitem häufigste Gitterdefekt in CZTS. Da bei Cu-Zn-Unordnung bei-de Defekte zu gleichen Anteilen auftreten, ist Cu-Zn-Unordnung unabhängigvon der Komposition.

Unordnung in Kristallen lässt sich durch Temperatur beeinflussen. Bei ho-hen Temperaturen haben die Atome mehr Bewegungsenergie und es fällt ihnenleichter ihre Gitterplätze zu verlassen und Paarsubstitutionen zu bilden. Beiniedrigeren Temperaturen tendieren die Atome eher dazu auf ihren eigentli-chen Gitterplätzen anzuordnen. Allerdings haben sie bei niedrigen Tempera-turen weniger Energie um eine Substitution durchzuführen. Da CZTS bei rechthohen Temperaturen von über 500 ◦C kristallisiert, sind die Cu und Zn Ato-me nach dem Ausheizen in kompletter Unrodnung verteilt. Beim schnellenAbkühlen wird ein Großteil der ungeordneten Atome in ihren Gitterplätzen“eingefroren” und bei Zimmertemperatur finden so gut wie keine Substitu-tionen der Atome statt. Hingegen ermöglicht ein langsames Abkühlen, dassmehr Atome in den geordneten Zustand übergehen können. So kann durchthermische Behandungen die Anzahl der ungeordneten Cu und Zn Atomedeutlich verringert werden. Auf der Grundlage von Theorien, welche Unord-nung in Kristallen beschreiben, können Vorhersagen zum Verhalten der Cu-Zn-Unordnung errechnet werden. Für CZTS wurde in dieser Arbeit gezeigt,dass extrem lange Temperaturbehandlungen von mehreren Jahren nötig wärenum die Auswirkungen von Cu-Zn-Unordnung auf Solarzellen zu minimieren.Daher sind weitere Lösungsansätze notwendig um Cu-Zn-Unordnung zu re-duzieren.

Cu-Zn-Unordnung kann weiterhin durch andere Gitterdefekte beeinflusstwerden, da diese Auswirkungen auf die Beweglichkeit der Cu und Zn Atomehaben können und somit die Ordnungsrate verändern. Leerstellen oder Zwi-schengitteratome erleichtern die Substitution der Atome und fördern somitdas “ordnen” der Cu und Zn Atome während einer Temperaturbehandlung.So kann der Effekt von Temperaturbehandlungen verstärkt werden und einhöherer Grad an Cu-Zn-Ordnung erreicht werden. Leestellen und Zwischen-gitteratome finden sich vor allem in Cu-reichen und Cu-armen CZTS in Form

17

von Cu-Zwischenatomen, bzw. Cu-Leerstellen oder als Zn-Leerstellen, bzw.Zn-Zwischenatomen.

Des Weiteren können die Bedingungen während des Ausheizens durch ho-he SnS und S Gasdrücke angepasst werden. Unter diesen Bedingungen ist dieLöslichkeit von Cu-Leerstellen in CZTS erhöht und es können mehr Leerstel-len gebildet werden. Dies führt zu einer weiteren Verbesserung der Ordnungs-rate während Temperaturbehandlungen.

Eine andere Möglichkeit zum Reduzieren der Cu-Zn-Unordnung ist ein be-troffenes Element auszutauschen. In dieser Arbeit wurde Zink durch Mangan(Mn) ersetzt, welches ebenfalls ungiftig ist und nicht zu seltenen Erden ge-hört. Die Verbindung Cu2MnSnS4 (CMTS) kristallisiert in der Stannitsturk-tur, welche der Kesteritstruktur sehr ähnlich ist. Obwohl Cu und Mn sich inihrer Größe im Kristall unterscheiden und CMTS in einer anderen Strukturals CZTS kristallisiert, hat sich gezeigt, dass dennoch Unordnung im Kristallvorhanden ist. Es ist daher zu erwarten, dass CMTS im Vergleich mit CZTSkeine Verbesserung in der Anwendung für Solarzellen darstellt. Des Weiterenkönnen weitreichende Empfehlungen für die weitere Suche nach zukünftigenAbsorbermaterialien abgeleitet werden.

18

1. Introduction

Humanity currently faces its greatest challenge yet - tackling the unavoidableconsequences of a human-caused, drastic change in the earth’s climate, andtrying to shift from a lifestyle which is based on growth and consumerism,to a sustainable one taking the limits of our planet and consequences of ourlifestyle on other living beings into consideration. A new generation led bythe teenager Greta Thunberg is electrifying the whole world by the clear andinescapable message demanding immediate action to save their future. Mean-while the older generation still in charge is struggling to find and implementsuitable solutions, a latest example is the Green Deal initiated by the EuropeanCommission president Ursula von der Leyen with the vision to make Europethe first climate-neutral continent1. One of the key issues is to develop andimplement sustainable, climate friendly ways to generate electricity for theworldwide growing demand. Solar cells play a central role in future scenariosfor sustainable electricity generation.

The sun is a powerful energy source. The earth continuously receives aglobal mean surface radiation of 185 W m−2 from the sun2, or 94.4×106 GWover the whole surface. A direct way to utilize this energy is through photo-voltaic devices, i. e. solar cells, which can transform light energy directly intoelectricity.

Photovoltaics represents a clean and renewable source for electricity gener-ation, and has been an active field of research for more than half a century. Wehave come a long way from the first silicon solar cell demonstrated by Belllaboratories in 1954 with 6 % conversion efficiency3 to today, where multi-junction solar cells reach conversion efficiencies of up to 47.1 %4. By theend of 2018, the worldwide cumulative installed photovoltaic capacity was512 GW, providing 2.9 % of the global electricity demand and the solar cellmarket has continued to grow rapidly5.

Beside being a green technology for electricity generation, solar cells yieldseveral other unique advantages: Solar cells can be installed independent ofthe electric grid, and can provide electricity anywhere, where there is sun-light. Their electricity output is scalable, which presents the opportunity toadapt size of the installation to the demand and circumstances on-site. Fur-thermore, the cost of solar cell installations has decreased drastically over thelast decades making them one of the cheapest electricity sources available (forinstallation in certain parts of the world) and reducing the payback time of so-lar cell installations below any other technology for electricity generation6,7.

The main challenge to integrate solar cell installations into the electric gridis that the output is fluctuating heavily from day to night time and due to

19

weather conditions. This presents the necessity for large scale energy storageto reduce the electricity output to the grid during times of peak production andto bridge periods of less solar irradiation. Unfortunately, there are still no largescale solutions to this challenge.

Up to now, the majority of the solar cells produced are based on crystallinesilicon. Silicon is an abundant and cheap material and the industry for solarcell production has had a long time to evolve. However, due to the indirectband gap and poor absorption, thick layers of around 200 μm of silicon withhigh crystalline quality are necessary8. This increases the production costsand constrains the applications of silicon solar cells. Thin film solar cells onthe other hand rely on a thin stack of materials of only a few μm9 which allowsthe production of extremely light weight and even flexible solar modules. Thereduction of the material allows decreasing production costs and faster payback times. Materials used for thin film solar cells yielding competitive effi-ciencies are CdTe and Cu(In,Ga)Se2 (CIGS)4. Unfortunately, these materialsinclude rare and expensive or toxic elements which hinders their widespreadapplication and up-scaling of the production to meet the growing energy de-mand. However, Kesterite Cu2ZnSnS4 (CZTS), which is studied in this thesis,could be a cheap and non-toxic alternative absorber material for thin film solarcell.

1.1 Kesterite for thin film photovoltaicsKesterite Cu2ZnSnS4 (CZTS) has received considerable attention over the last15 years as an alternative thin film absorber material, because it consists ofabundant and non-toxic elements10. CZTS has high absorption and a directband gap at 1.5 eV, which can be tuned between 1.5 and 1.0 eV by partially re-placing S with Se11,12. Therefore, only a thin layer of 1−2 μm is sufficient forlight absorption. It has been shown that CZTS thin films can be fabricated byseveral different deposition techniques including vacuum based processes13

as well as non-vacuum techniques14 yielding well performing solar cells.The conversion of the energy of photons (i. e. of light) to an electric current

is achieved through two main processes in the solar cell: (1) The photon isabsorbed by the absorber layer of the solar cell which generates an electron inthe conduction band and a hole in the valence band. (2) The electron and thehole are separated and collected by the contacts before they can recombine.

A typical stack of a CZTS solar cell is sketched in Figure 1.1 (a). Thethickest layer of the solar cell stack is the CZTS absorber layer. As the namesuggests, the absorber layer is the layer of the solar cell where (ideally) themajority of the photons are absorbed leading to the excitation of free electronsand holes, i. e. carrier generation. In order to collect the generated carriersfrom the absorber layer, an n-type buffer layer is introduced on top of theCZTS layer (in Figure 1.1 the buffer layer is made of CdS). The p-type CZTS

20

Figure 1.1. Structure of a solar cell with kesterite Cu2ZnSnS2 as absorber layer.Sketch of band diagram and photovoltaic effect in the CZTS solar cell at short-circuitconditions (V = 0).

absorber layer and the n-type buffer layer form a pn-junction which leads tothe formation of the space charge region at the interface and an internal elec-tric field. The electrons and holes which are generated within the space chargeregion or which diffuse into the space charge region will be accelerated inopposite directions causing charge separation. The electrons are acceleratedtowards the buffer layer, while the holes are accelerated to the opposite direc-tion, away from the buffer layer. The schematic band diagram of a CZTS solarcell and charge generation and separation is illustrated in Figure 1.1 (b). InCZTS solar cells the buffer layer is often formed by a CdS layer. However,based on the toxicity of CdS and the unfavorable cliff-like band alignment be-tween CZTS and CdS (see Figure 1.1) alternative buffer layers could be anadvantage. Therefore, CZTS solar cells with the alternative Cd-free bufferlayers Zn(O,S)15 and Zn1−xSnxOy

16 have been demonstrated.Finally, the carriers are collected by the front and back contacts. The metal-

lic back contact collects the free holes and is usually made of molybdenumMo. The electrons are collected by a transparent conducting oxide (TCO)layer, which is deposited on top of the buffer layer. Both the buffer and TCOlayer should be as transparent as possible in order to reduce parasitic absorp-tion. But at the same time they should also be good electron conductors. ZnOand Al doped ZnO are often used as TCO layers in CZTS solar cells. For so-lar cells produced in the lab, a narrow metal grid is often deposited on top toreduce the resistive losses in the TCO.

As other electrical devices, the performance of solar cells is most com-monly analyzed by their current-voltage characteristics. Figure 1.2 shows thecurrent-voltage curve (I-V curve) of a CZTS solar cell with a stack as de-scribed in Figure 1.1 (a). Without being illuminated, the I-V curve of a solar

21

-0.2 0.0 0.2 0.4 0.6 0.8-20

-10

0

10without illuminationilluminated

Current(mA)

Voltage (V)

ISC

VOC

Pmax = Imp x Vmp

Figure 1.2. I-V curve of a CZTS solar cell.

cell resembles the exponential behavior of a pn-junction. Under illumination,the I-V curve is shifted along the current axis by the amount of the light currentIL. Therefore, the I-V characteristics of an ideal solar cell under illuminationcan be described by the adapted ideal diode equation17

I(V ) = I0

[exp

(qV

nkBT

)−1

]− IL . (1.1)

I0 is the reverse saturation current which is driven by recombination, q is theelemental charge, n is the ideality factor, kB is the Boltzmann constant and T isthe temperature. It should be noted that the current I is often expressed as cur-rent density with the formula sign J in A/m2 to remove the dependence on thecell area. The ideal diode equation and Equation 1.1 do not accurately describethe I-V characteristics of real solar cells, as the one shown in Figure 1.2. Thatis because equation 1.1 does not consider resistive effects, and the so calledcross-over where the IV-curve under illumination crosses the IV-curve withoutillumination cannot be explained with Equation 1.1.

The I-V curve of a solar cell is characterized by two distinct points: theopen circuit voltage VOC, the voltage that builds up when no current flowsbetween the contacts, and the short circuit current ISC, the current at zero volt-age. Furthermore, the maximum power point along the curve is found wherethe product V · I is largest. The fill factor can then be defined by the ratio

FF =Vmp · Imp

VOC · ISC. (1.2)

22

Finally the conversion efficiency η can be defined by the parameters extractedfrom the I-V curve

η =VOC · ISC ·FF

Pin, (1.3)

where Pin is the power input, in case of a solar cell this is the power of theilluminating light17.

From equation 1.3, the three parameters VOC, ISC and FF should be as highas possible to maximize the conversion efficiency of the solar cell. Accordingto equation 1.1, the short circuit current ISC(V = 0) should be equal to the lightcurrent IL for an ideal solar cell. It becomes clear, that any effect reducing thelight absorbed in the absorber layer, e. g. reflection or parasitic absorption inthe buffer layer, reduces IL and has a negative effect on ISC.

An important property defining the absorption in the absorber layer is theband gap of the absorber material. A small band gap increases the range ofphotons that can be absorbed, because photons with photon energy smallerthan the band gap are insufficient to excite electrons to the conduction band.Therefore, a smaller band gap EG is preferable with respect to the short circuitcurrent ISC. On the other hand, after charge generation all energy transferredfrom the photon to the electron greater than EG will be lost due to thermal-ization. Therefore, EG is directly related to the achievable output voltage VOCand higher EG will improve VOC. This conflicting dependence of VOC and ISCleads to an optimal range of the band gap between 1.1 eV and 1.6 eV for theabsorber material in single junction solar cells, where the maximum achiev-able efficiency is around 30 %18 (also known as Shockley-Queisser limit). Theband gap of CZTS with EG = 1.5−1.6 eV falls within this range and is there-fore a suitable semiconductor for photovoltaic applications.

An expression for VOC can be derived from equation 1.1 by applying I = 0,which gives

VOC =kBT

qln(

IL

I0+1

). (1.4)

A higher light current IL has a positive effect on VOC, while the reverse satura-tion current I0 should be as small as possible to maximize VOC. Because I0 isproportional to the recombination in the solar cell, VOC is heavily affected forexample by defects acting as recombination centers. The shortcoming of VOCis often expressed as a function of EG as the open circuit voltage deficit ΔVOC

ΔVOC =EG

q−VOC . (1.5)

In CZTS solar cells the large VOC deficit is most troublesome. Record effi-ciencies of solar cells based on Cu2ZnSn(S,Se)4 have stagnated at 12.6 %14,19

and fall behind other thin film solar cells like CdTe, CIGS and perovskitetechnologies. Several bottle necks have been identified that are likely to limitthe conversion efficiency in CZTS solar cells: (1) Sulfurization of Mo at the

23

metallic back contact leads to the formation of of MoS2 and to decompositionof CZTS into secondary phases20 causing a higher series resistance and in-creased recombination at the back interface21. (2) the unfavorable band align-ment between CdS and CZTS forming a so-called cliff band alignment at theinterface22 (see also Figure 1.1 (b)). This causes interface recombination andVOC losses. The alternative buffer layers Zn(O,S)15 and Zn1−xSnxOy

16 seemto solve this problem in CZTS solar cells. However, the absence of a majorimprovement of the solar cell performance leads to the suspicion that othershortcomings still remain.

Which leads to bottleneck number (3): the high intrinsic defect densities inthe bulk CZTS absorber layer present a serious challenge on the path towardsimproving CZTS solar cell conversion efficiencies. The problems in the CZTSbulk have been identified as the main impediment to increasing the device ef-ficiency of CZTS solar cells. The intrinsic defects and defect complexes causedetrimental band tailing and potential fluctuations which cause CZTS to sufferfrom a high rate of recombination limiting the VOC

23,24. This issue has beenbroached in numerous studies25,26,27. However, so far no solution towardshigher efficiencies has been found. Cu-Zn disorder and deep defects, such asSnZn, have been identified that could potentially lead to such a detrimentaleffect on the material properties.

1.2 Scope and aim of this thesisThe research presented in this thesis attempts to address the final challengepresented in Section 1.1: the intrinsic defects and crystalline quality of CZTS.When the work for this thesis started in January 2015, it was widely acceptedin the research community that CZTS crystallizes in the kesterite crystal struc-ture with abundant Cu-Zn disorder28. Detrimental potential fluctuations hadbeen identified as a possible limitation of VOC and characterization of the in-trinsic defects as well as Cu-Zn disorder was a hot topic29,30,31.

Achieving control over the defects and their effect on a material is referredto as “defect engineering”. In this work, different possibilities for defect en-gineering in CZTS are explored in order to control the nature of the defectswith the aim to improve the properties of the CZTS material. A special focusis placed on the reduction of Cu-Zn disorder in CZTS. The strategies investi-gated in particular involve

1. Temperature treatments to reduce disorder (Paper I)2. Composition tuning for fast ordering kinetics (Paper II)3. Tuning the annealing step in the fabrication process to find ways to in-

fluence the composition and the nature of intrinsic defects (Paper III andV)

4. Cation exchange to form related materials to explore the possibility ofavoiding cation disorder (Paper IV).

24

The following chapters present the results of these works in the wider con-text of the current understanding of CZTS as a solar cell absorber material.The thesis is arranged in four chapters covering the structure of CZTS in Chap-ter 2, and defects and electronic properties of CZTS in Chapter 3. The cationdisorder as the most abundant defect in CZTS is discussed in detail in Chap-ter 4. Finally possibilities for defect engineering in CZTS are elaborated inChapter 5 with a focus on the results from Papers I to V. The thesis is com-pleted by Chapter 6 which summarizes the main conclusions and presents anoutlook from this work.

25

2. Structural properties

The crystalline structure of a solid is defined based on the atoms that it is buildfrom, e. g. their size and the occupation of the electron shells. For each com-bination of atoms, there is one preferred structure in which the atoms align.This crystal structure in turn defines essential properties of the solid, such asthe electronic band structure. This chapter describes the structural propertiesof Cu2ZnSnS4 (CZTS) and the fabrication methods exploited in this thesisfor CZTS thin film fabrication. The fabrication of this quaternary compoundimplicates some new challenges with regard to decomposition based on thevolatility of Sn and SnS at high temperatures. Therefore, the chapter also cov-ers different models describing the decomposition of CZTS and segregationof secondary phases and their effect on devices and on the CZTS materialproperties are discussed.

2.1 The CZTS crystalAs depicted in Figure 2.1, the Cu2ZnSnS4 lattice can be derived from thezincblende ZnS lattice by aligning a Zn, a Sn and two Cu atoms on the cationsites around one S atom forming S-Cu2ZnSn tetrahedral motifs that satisfythe octet rule. Several polymorphs can be constructed from these tetrahedraincluding the stannite (space group I4̄2m), kesterite (I4̄), the PCMA structure(primitive-mixed CuAu structure with space group P4̄2m) and another tetrag-onal structure with space group P4̄2c. First principles Density functional the-ory (DFT) calculations predicted the kesterite structure to be the most stablecrystal structure33,34. The kesterite ground state structure of CZTS could beconfirmed with neutron diffraction experiments35. The kesterite phase is sta-ble up to 876 ◦C where a phase transition to cubic sphalerite structure (spacegroup F 4̄3m) occurs28. Experiments by neutron diffraction further gave ev-idence for disorder in the Cu-Zn planes of the kesterite structure (for moreinformation about Cu-Zn disorder see Chapter 4). The other two tetragonalpolymorphs of CZTS with space group P4̄2m and P4̄2c can be formed un-der certain fabrication conditions and can be differentiated from the kesteritestructure by Raman spectroscopy36 (see Box 2.1).

2.2 Fabrication of thin film CZTSFabrication of CZTS thin films with good crystallinity faces several chal-lenges. On one hand, crystallization of CZTS requires temperatures of 500 ◦C

26

Figure 2.1. Comparison of CZTS with kesterite and stannite crystal structure, whichcan both be derived from the zincblende structure of ZnS. Both kesterite and stan-nite CZTS are build from S-Cu2ZnSn tetrahedral motifs. Produced with the SoftwareCrystalMaker 32.

or higher37,38. On the other hand, SnS phases become highly volatile at suchhigh temperatures which drives the decomposition of CZTS39, as is discussedin detail in Section 2.3. This makes it difficult to implement a single-step pro-cess for fabrication. There have been few approaches to produce CZTS thinfilms by co-evaporation40,41 and sputtering42 in a single step. However, thegeneral route to produce CZTS thin films involves a two step process. In thefirst step, a precursor containing Cu, Zn, Sn and S is deposited at rather lowtemperatures and during the second step the precursor is annealed at a highertemperature in a controlled environment to reduce evaporation of SnS. Vari-ous techniques have been employed for the precursor deposition ranging fromvacuum techniques, such as sputtering43,19,44 and evaporation45, to solution-based processing14,46.

The precursors for the samples investigated in Papers I to V were producedby co-sputtering from alloy and metal targets (either CuS, Zn and Sn, in activemode with H2S, or CuS, ZnS and SnS). The substrates were soda-lime glasscovered by DC-sputtered molybdenum to resemble the processing of solar cellfabrication where a molybdenum layer is used as the back contact. In the caseswhere a transparent substrate was needed for characterization (Paper IV), theprecursors were directly deposited on the soda-lime glass. Soda-lime glass isnot only a cheap and heat durable choice for the substrate, but it also containssodium (Na) which out-diffuses into the Cu-Zn-Sn-S layer during the anneal-ing step and substantially improves crystallization47. During deposition of theprecursors the substrate holder was heated to 250 ◦C. The substrate holder isusually rotated to achieve uniform composition across the sample area. In Pa-pers II, III and V the Cu-Zn-Sn-S precursors were deposited without substrate

27

Box 2.1: Raman spectroscopy

Raman spectroscopy measures the spectrum of lattice vibrations of a material,also called phonons. Lattice vibrations depend on the mass of the vibratingatoms and the characteristics of the bonds to their neighboring atoms, e. g.bond lengths and strength. This leads to a characteristic set of phonon modesfor every material and enables phase identification based on Raman spectra.

The principle of Raman spectroscopy is illus-trated in the Figure: 1. An incoming photonexcites an electron. 2. The electron scatterswith the lattice, which creates (annihilates) aphonon. 3. The electron recombines emittinga photon with an energy lower (higher) thanthe incoming photon ΔE equal to the created(annihilated) phonon energy. The phonon spec-trum is acquired by measuring the energy dif-ference ΔE. Therefore it is important to use amonochromatic light source, typically a laser,as excitation source.

1

2

3Ener

gy

Stokes Rayleigh Anti-Stokes

PhononScattering processes

Inte

nsity

ELaser ELaser+EPhononELaser–EPhonon

–

Phonon

In order to stimulate photon-electron interaction, the incoming photons musthave an energy equal or higher than the band gap of the material. ResonantRaman conditions apply if the energy of the incoming photon is close to theband gap energy or another optical transition energy. Then the photon-electroninteraction probability is greatly enhanced which leads to higher intensity ofthe Raman spectrum.

LASER

CCD

0 200 400 600 800 1000

0

2000

4000

6000

8000

10000exc= 785nm

Mn2SnS3

Inte

nsity

(arb

. unit

)

Wavenumber (cm-1)

Sample

Grating

Edge Filter

28

Box 2.2: Composition measurements

Energy-dispersive X-ray spectroscopy (EDS or EDX) and X-ray fluores-

cence spectroscopy (XRF) utilize the characteristic X-ray emission to quan-tify the elemental composition of a sample. EDX uses an electron beam asexcitation source. Because of the high interaction probability of electrons withmatter, the probing volume is rather small which leads to a low probing depthand uncertainties in the resulting atomic composition of about ±5 %50. XRFon the other hand relies on high energy X-rays as excitation source, whichprovides a deep probing depth and more accurate compositions. However, thesetup needs to be calibrated by a sample with known composition.For the composition studies of this thesis, the calibration sample for XRF wasanalyzed by Rutherford back scattering (RBS) which is an ion beam tech-nique. This technique delivers rather exact composition results but requires anelaborate experimental setup.

rotation which led to a continuous gradient in the cation composition acrossthe sample area. These samples are referred to as composition-spread CZTSthin films. The precursors deposited by co-sputtering from alloy targets werefully sulfurized and featured a poorly crystalline structure of the zincblendetype with small grains48.

A crucial aspect during the precursor deposition is the control of the cationcomposition, because the composition of the precursor impacts the secondaryphase formation and the nature of the defects in the final CZTS film. In gen-eral, the integral composition of the Cu-Zn-Sn-S precursor films was mea-sured by X-ray fluorescence using a calibration sample that had been char-acterized by Rutherford back scattering49. The composition of composition-spread CZTS films was mapped with energy dispersive X-ray spectroscopy(see Box 2.2).

In the second processing step, the precursors were annealed in a tube fur-nace to enhance the crystallinity and grain growth. The samples were placed ina graphite box together with elemental sulfur (or other sulfur and SnS sourcesin Papers III and V). Argon served as an inert process gas in the tube furnace.The graphite box was inserted from a cold loading zone into a pre-heated hotzone and removed after the annealing time which ranged from 2 min to 13 min.The annealing temperature was around 550-580 ◦C. The small volume in thegraphite box together with the short annealing times allowed to sustain a sul-fur rich atmosphere for the duration of the high temperature anneal and by thatreduce sulfur and SnS loss from the sample by evaporation. Nevertheless, acomposition shift is typically observed when comparing the annealed CZTSfilm with the precursor film due to Sn-loss. The mechanisms of Sn-loss duringthe high temperature anneal will be discussed in Section 2.3.

29

The annealing step needs to be carefully optimized in order to yield CZTSthin films of high quality. Because the annealing process in the tube furnaceis a non-equilibrium process, there is a trade-off between long annealing time,yielding larger grains, and maintaining the sulfur and SnS partial pressures,which degrade for longer annealing times51,52. Other parameters that need tobe optimized are the annealing temperature and the tendency of the CZTS filmto peel off based on residual stresses from sputtering and on stress and strainexerted on the film and the substrate under fast temperature changes.

Prior to further deposition, the surface of the CZTS film is conditioned byKCN etching to remove oxides, sulfates and some of the secondary phasesfrom the CZTS surface53. For the complete stack of a solar cell, the trans-parent ‘window’ layers are deposited on top of the (etched) CZTS absorberlayer. A common buffer layer to form the pn-junction with Cu2ZnSn(S,Se)4(CZTSSe) is cadmium sulfide CdS which is deposited by chemical bath de-position13,19. However, CdS does not form an optimal band alignment withCZTS and is often criticized for its toxicity. Therefore, alternative buffer lay-ers have been investigated, yielding promising candidates such as Zn(O,S)15

and Zn1−xSnxOy16. The front contact is then completed with an intrinsic ZnO

layer and an Al-doped ZnO layer and in some cases a metal grid for reducedresistivity losses. The complete stack of a CZTS solar cell is illustrated inFigure 1.1 (a).

2.3 Models for phase stability and decomposition intosecondary phases

A reoccurring challenge in the fabrication of CZTS thin films is to avoid sec-ondary phase formation, because they can have a detrimental impact on thedevice performance of solar cells. Common secondary phases in CZTS thinfilms are shown in the ternary phase diagram in Figure 2.2 and their effect onsolar cell characteristics according to 55 are listed in Table 2.1. The best de-vices are fabricated with Zn-rich CZTS in order to avoid formation of the detri-mental phases CuxS and Cu2SnS3

14,19. It should be noted that Figure 2.2 rep-resents the equilibrium state of the quaternary Cu-Zn-Sn-S system at 400 ◦C.The phase diagram is expected to differ for conditions that apply during thinfilm fabrication. This explains why other secondary phases are commonly ob-served in CZTS thin films as well, such as SnS, or Cu3SnS4 which is observedin composition-spread CZTS thin films.

Secondary phases can form directly from the precursor during the growth,which is considered segregation of secondary phases, or they can form whenCZTS decomposes into ternary or binary phases. Several reasons have beenidentified that facilitate secondary phase formation during the growth ofCZTS. Even though kesterite Cu2ZnSnS4 is the stable phase for the quaternarysystem Cu-Zn-Sn-S, its phase stability region is rather narrow according to

30

Figure 2.2. Sketch of the ternary phase diagram at 400 ◦C, according to Ref. 54. Onlydiscussed secondary phases are included. Some secondary phases, such as Cu3SnS4,CuS and SnS are not shown, as they do not lie in the same plane due to their S content.

Table 2.1. Typical secondary phases observed in Cu2ZnSnS4 thin films and their effecton solar cell characteristics according to Ref. 55.

Phase Property of phase Effect on deviceZnS Eg = 3.6 eV reduces effective area,

can increase series resistance

CuxS 1.2eV < Eg < 2.5 eV, severe shunting,highly conductive/metallic recombination reduces VOC

SnSx 1.1-2.2 Eg barriers for carrier collection

Cu2SnS3 0.98-1.35 Eg recombination and lower VOC,reduced carrier collection

31

theoretical investigations56,57, i. e. it cannot tolerate large deviations from thestoichiometric 2:1:1:4 atom ratios without precipitation of secondary phases.To prevent segregation of CuS, SnS or SnS2, ZnS and Cu2SnS3 the process-ing conditions, including the composition, need to be carefully adjusted. Evensmall deviations can lead to secondary phase formation and it is not unusualthat several secondary phases are observed together in the same sample alongwith the kesterite CZTS phase58.

Decomposition of CZTS into the binaries is driven by the instability ofSn(IV) in CZTS at high temperatures and low S2 partial pressure. Reaction 2.1depicts the decomposition of CZTS into its binary secondary phases leading tothe reduction of Sn(IV) in CZTS to Sn(II) in SnS.51 The high vapor pressureof SnS then drives Reaction 2.2 where SnS evaporates by forming gaseousSnS.

Cu2ZnSn(IV)S4(s) ⇀↽ Cu2S(s)+ZnS(s)+Sn(II)S(s)+12

S2(g) (2.1)

SnS(s) ⇀↽ SnS(g) (2.2)

These reactions represent equilibria, i. e. they are reversible and expected tooccur in both directions simultaneously. The rate at which the reaction andits reverse reaction occur and the ratio of the products of the system in equi-librium are determined by the boundary conditions of the system. Therefore,the decomposition described in Reactions 2.1 and 2.2 can be prevented by en-suring sufficient S2 and SnS partial pressures. This will limit SnS loss anddecomposition of CZTS which can have a detrimental effect on the solar cellperformance59.

One reaction initiating the decomposition of CZTS into its secondaryphases arises from the sulfurization of the molybdenum back contact whichoccurs during the high temperature anneal. Typically, a MoS2 layer of a fewhundred nm is formed at the back contact. The formation of MoS2 is not onlyproblematic based on the negative impact on the series resistance of CZTS so-lar cells60,21, but also because it leads to sulfur deficiency in the CZTS layercausing decomposition of CZTS into its binary secondary phases61,20

2Cu2ZnSn(IV)S4 +Mo −→ 2Cu2S+2ZnS+2Sn(II)S+MoS2 .(2.3)

Another approach to understand the phase stability of CZTS is given in Pa-per III, where a chemical model for the single phase region is established basedon equilibria of the defect complexes that are formed in the off-stoichiometricCZTS phase (see also Section 3.3). The single phase region encompasses thecomposition range where CZTS is formed without secondary phase segrega-tion. This region should contain the point of stoichiometric composition. Forany other point in the single phase region, the off-stoichiometric composi-tion implies the formation of defect complexes and the density of these defectcomplexes increases further away from the stoichiometric point (defect com-plexes in CZTS are discussed further in Section 3.3). At a certain composition

32

the formation of defect complexes will be less favorable than segregation ofsecondary phases. Therefore the edge of the single phase region marks thesolubility range of the defect complexes in that composition range. The sol-ubility of the defect complexes is not constant for the material, but changeswith temperature and processing conditions, such as partial pressure of S2 andSnS. Hence, in this model the equilibrium is formed between CZTS contain-ing a defect complex and CZTS with secondary phases. For example, for theB-type defect complex (2ZnCu +ZnSn) the equilibrium reaction is

2Cu2ZnSnS4 with (2ZnCu +ZnSn) ⇀↽ Cu2ZnSnS4 +4ZnS . (2.4)

The equilibria involving other defect complexes and secondary phases can bederived similarly.

Equilibria involving SnS2 as a segregated secondary phase will again beaffected by the SnS and S2 partial pressures based on the reaction

SnS2(s) ⇀↽ SnS(g/s)+12

S2(g) , (2.5)

and the solubility of these defect complexes will depend heavily on the pro-cessing conditions. The conclusion regarding solubility limits of defect com-plexes in CZTS is that not only the composition of the precursors determinesthe nature of the intrinsic defects, but also the fabrication conditions with re-gards to S2 and SnS partial pressures.

These results explain, why phase segregation occurs even if the integralcomposition is inside the (equilibrium) phase region illustrated in Figure 2.2.During processing of CZTS thin films for solar cell applications, we shouldinstead consider the non-equilibrium single phase region including the impli-cations on secondary phase segregation and solubility of intrinsic defect.

2.4 Experimental investigation of the single phaseregion

Previous experimental investigations of the phase stability of CZTS in terms ofcomposition and the extent of the single phase region are based on single crys-tal and powder samples fabricated by solid state reaction29,62,63. Based on thedifferent processing conditions compared to thin film fabrication techniques,these findings are not expected to match the situation of thin film processingvery well. In Papers III and V a different approach is presented to study thephase stability and the single phase region of CZTS under thin film process-ing conditions using composition-spread CZTS thin films (see Box 2.3). InPaper III, this method was applied to validate the model of the phase stabilitybased on the solubility of defect complexes.

In order to identify the single phase region on the composition-spread sam-ples, the secondary phase boundaries (boundary marking the region where

33

Box 2.3: Investigation of composition-spread CZTS samples

In Papers III and V an experimental approach has been established to identifythe single phase region of CZTS composition-spread samples combining X-raydiffraction and Raman spectroscopy for secondary phase analysis. With theknown composition boundaries for secondary phase formation, it is possibleto distinguish effects of secondary phases from effects of composition changesin CZTS. This knowledge is vital for understanding the potential for defectengineering because the formation energy of defects depends on the chemicalpotential which is again tied to the composition.The approach to use composition-spread samples presents the opportunity tostudy the phase stability region of CZTS under thin film processing conditions.This would be impossible by means of individually produced CZTS films with-out composition-spread, but each with a slightly different composition to coverthe full single phase region in compositions (or at most very tedious to real-ize as an experiment and to interpret the results). By the approach of usingcomposition-spread samples, it is secured that the full composition range isexposed to the exact same fabrication process in terms of sputtering, anneal-ing and temperature history, but also rest times in between processing steps.However, this approach also comes with some challenges, such as handlingand interpretation of the huge amount of data from mapping the sample areaby different analysis techniques. Furthermore, it is undeniable that the pro-cessing conditions of composition-spread CZTS do lead to some differencesin phase formation in the ternary Cu-Sn-S secondary phases: typical Zn-poorCZTS thin films exhibit the ternary phase Cu2SnS3. For composition-spreadCZTS films on the other hand, segregation of Cu3SnS4 is observed (see64 andPapers III and V). This difference could be caused by the interactions of thedifferent composition regions via the gas phase.

34

Box 2.4: X-ray diffraction

X-ray diffraction (XRD) is a common technique to study the crystal structureof a material. A monochromatic X-ray beam (typically the Kα radiation of Cuwith λ = 1.5406 Å) is elastically scattered by the electrons of the atoms in thecrystal. At certain angles θ to the incoming beam, the scattered x-rays interfereconstructively according to Bragg’s law

nλ = 2d sinθ

and the diffraction pattern will feature a peak at that angle. From the anglesθ of constructive interference the distance d between the crystal planes can becalculated.For the analysis of thin films, the incoming beam can be fixed at a very lowangle, typically between 0.5◦ − 2◦, to increase the probed volume. This tech-nique is called Grazing incidence X-ray diffraction or GI-XRD. It is importantthat the films are polycrystalline and non-oriented to ensure that all diffractionpeaks of the pattern are recorded.

a secondary phase could be resolved) were analyzed across the full samplearea. In Papers III and V X-ray diffraction (XRD, see Box 2.4) was usedto identify Sn-S related secondary phases and to find their phase boundaries.However, the diffraction pattern of Cu2ZnSnS4 overlaps with other binaryand ternary sulfides based on the similarities in their structure. Therefore,XRD is unsuited to identify the secondary phases ZnS and Cu3SnS4. Addi-tionally, the identification of CuS proved to be difficult by XRD. Instead, thesecondary phases ZnS, CuS and Cu3SnS4 were analyzed by multiwavelengthRaman spectroscopy which was shown to be a suited complementary tech-nique to XRD for secondary phase identification65,66. For multiwavelengthRaman spectroscopy, the Raman spectrum is recorded under different excita-tion wavelengths that match optical transitions of potential (secondary) phasesof the sample in order to measure their response in resonant condition. Thisway the Raman signal of even small amounts of a phase can be detected.

XRD can scan the full sample stack including the Mo back contact. Ramanspectroscopy on the other hand is a surface-sensitive technique and secondaryphases that occur deeper within the film are more difficult to detect, while theRaman response from secondary phases close to the surface may be enhanced.In Paper V the secondary phase assignment was therefore cross-checked withcombined scanning transmission electron microscopy and energy dispersiveX-ray spectroscopy (STEM-EDX).

Figure 2.3 shows the outline of a composition-spread CZTS thin film in acomposition diagram. The secondary phase boundaries from XRD and Raman

35

SnSx

ZnSCuS

Cu3SnS4

Figure 2.3. Composition region covered by a composition-spread CZTS sample withsecondary phase boundaries indicated by white dotted and dashed lines. The singlephase region is marked with a white star.

mapping are indicated by white dotted and dashed lines. The region enclosedby these boundaries represents the CZTS single phase region.

With the outline of the single phase region established, the properties of theCZTS phase within this region are investigated in Papers III and V by study-ing the Raman characteristics. These yield information about the crystallinequality of the materialand about defects present in CZTS, in particular the Qparameter which can be extracted from resonant Raman spectra of CZTS67

(see Chapter 4 for more information about Q).The trends observed for the Q parameter across the composition-spread

sample are discussed in more detail in Section 5.1 and Figure 5.2 shows theQ parameter for two composition-spread CZTS films annealed under differentconditions. In Paper III these trends are used to confirm the model describingthe stability of the single phase region based on the solubility of defect com-plexes which was described in the previous section. The Q parameter reflectsthe expected trends for defects in off-stoichiometric CZTS: within the singlephase region the density of defects increases away from the stoichiometriccomposition and stagnates at the composition where secondary phases appear.In other words, the boundary of the single phase region marks the solubilitylimit of certain defects in the CZTS crystal. In regions that are outside thesingle phase region, the density of defects remains almost constant. Instead ofimplementing more defect complexes in the CZTS phase, secondary phasesare formed. Therefore, segregation of secondary phases can be regarded as aconsequence of limited solubility of defects in the CZTS phase. The effect of

36

Figure 2.4. Cross section images of a composition-spread CZTS thin film at dif-ferent compositions. The images were acquired by scanning transmission electronmicroscopy. Adapted from Paper V.

defects on the properties of the CZTS phase will be further discussed in thefollowing chapters of this thesis.

The areas of the composition-spread sample outside the single phase regionyield information about the interplay of the CZTS phase and the secondaryphases. Paper V discusses the effect of secondary phases on CZTS based onobservations from Raman spectra and photoluminescence (see Box 3.2). Fig-ure 3.4 shows the photoluminescence intensity across two composition spreadsamples annealed under different conditions. Comparing the trends across thesamples with the outline of the secondary phase boundaries reveals that ZnSand CuxS do not influence the photoluminescence intensity. However, theternary phase Cu3SnS4 severely reduces the photoluminescence intensity. Asimilar, detrimental effect of this phase can be observed on features of the Ra-man spectrum (see Figure 5.2). In Paper V it is proposed that Cu2ZnSnS4 andCu3SnS4 form a solid solution. The presence of Sn-S phases on the other handenhances the photoluminescence intensity by several orders of magnitude.

Figure 2.4 shows scanning transmission electron microscopy cross sectionimages extracted from different positions across the composition-spread sam-ples. The cross section images reveal that the morphology of the CZTS filmis enhanced for Sn-rich compositions of the composition-spread CZTS filmscompared to stoichiometric or Zn-rich regions. Sn-rich CZTS exhibits largergrains and a reduced number of visible extended defects (i. e. stacking faultsor dislocations). However, it is unclear whether the enhanced morphology iscaused by the presence of SnSx phases or by the Sn-rich composition of theCZTS phase.

The exact origin and mechanisms behind the interplay of the SnSx sec-ondary phases and the CZTS phase leading to higher photoluminescence in-tensity and possibly enhanced morphology and grain growth are not under-stood. Some possibilities are discussed in Paper V, e. g. passivation of inter-

37

faces by SnSx phases, the effect of secondary phases on the growth mecha-nisms and inhibited decomposition of CZTS based on the presence of SnSxphases.

This chapter summarizes the challenges in the fabrication of Cu2ZnSnS4 thinfilms connected to decomposition and segregation of secondary phases. It isexplained how the growth conditions (in particular during the annealing step)influence the secondary phase formation and that sufficient S2 and SnS shouldbe supplied to counteract secondary phase segregation and decomposition ofCZTS. These insights are particularly important for the optimization of theconditions during the anneal to grow CZTS thin films. Furthermore, the ef-fect of secondary phases on CZTS is discussed. The ternary phase Cu3SnS4has a detrimental effect on the CZTS material properties. SnSx on the otherhand improves the photoluminescence yield in the CZTS phase. However, itis difficult to conclude whether the presence of SnSx can possibly lead to animprovement in CZTS device performance.

38

3. Electronic properties

The electron orbitals that contribute to the bonds in a crystal lattice form con-tinuous bands. Based on the periodicity of the lattice, the properties of thiselectronic band structure are continuous throughout the whole material. Sothe crystal structure shapes the electronic band structure of a solid, definingfor example its band gap and the density of states within the bands. Both arevery important for the absorption of photons and therefore for materials for so-lar cell applications. Defects in the structure create deviations from the idealband structure. They cause doping and free carriers in the material, whichare important for conductivity. On the other hand, defects also cause poten-tial fluctuations and act as recombination centers, which can have detrimentaleffects on the solar cell efficiency. This chapter sheds light on the electronicstructure of Cu2ZnSnS4 and the nature of common defects in the material aswell as their effect on solar cell devices.

3.1 The band gap of CZTSCu2ZnSnS4 is a direct band gap material with strong light absorption. Theo-retical investigations predict the band gap at 1.5 eV33,68. The most commontechnique to probe the band gap of a material is from absorption spectra (seeBox 3.1). Other parameters of a material relevant for photovoltaic devicesthat can be extracted from the absorption spectra are the absorption coefficientand sub-band gap absorption due to band tails. The absorption coefficient ofCZTS is typically above 104 cm−1 which is reasonable for an absorber mate-rial for solar cells. The band gap of CZTS has been assigned with energiesranging from 1.45-1.65 eV based on multiple techniques, including absorp-tion69,70, spectroscopic ellipsometry71,72, quantum efficiency73,13 or photolu-minescence excitation12. For single crystal CZTS quantum efficiency analysisgave a slightly larger band gap energy of 1.64-1.68 eV and electroreflectancespectroscopy gave even larger band gaps of 1.71 eV74. Another study reportsthe band gap for CZTS single crystals at 1.46 eV based on electroreflectancespectroscopy75.

The wide spread of reported values for the band gap is unexpected, becausethe energy gap between valence and conduction band should be characteris-tic for the material. However, in CZTS several factors were identified, thatcan modify the observed band gap. One of the more prominent factors isthe degree of cation order (see Chapter 4), which has shown to cause band

39

Box 3.1: Band gap estimation from absorption spectra

When light shines on an object, the light is either reflected, absorbed or trans-mitted through the object. (The light can also be scattered, however this partis small and can be neglected if the surface is rather smooth.) If the incidentlight intensity is known, the transmission and reflection of a sample can bemeasured to determine the absorption of the sample. The absorption of a ma-terial describes the ability of the material to transform the energy of incominglight to (mainly) electronic excitations in the material and is dependent on thethickness of the material by the Lambert-Beer law

I(d) = I0 · exp(−α(E)d) ,

where I0 is the intensity of the light entering the material, I is the light intensityafter transmitting through distance d of the material and α is the absorptioncoefficient. The absorption coefficient of a material strongly depends on theenergy of the incident light E. For direct band gap materials, the absorptioncoefficient increases above the band gap energy due to the drastic increase ofthe joint density of states JDOS according to the relation

α(E) ∝ JDOS ∝1E

√E −Eg ,

Sub-band gap absorption is attributed to band tail absorption and can be differ-entiated from absorption above the band gap energy by its logarithmic behaviorwith energy76.

40

gap changes of more than 100 meV77,78,13 (see also Figure 4.3 (a)). Anotherfactor that can change the band gap energy is the cation composition79. Ad-ditionally, secondary phases which are easily formed during CZTS synthesis(see also Section 2.3) can affect the absorption spectra and hamper the bandgap extraction.

The large discrepancy of reported values for the band gap in CZTS mightalso be caused by the relatively large contribution of band tails to sub-bandgap absorption which complicates the extraction of the correct band gap80,24.Such band tails are detrimental for the device performance because they en-hance recombination and reduce the open circuit voltage VOC

81,82,83,84. Com-pared with other competing materials for thin film solar cells, the absorptionspectrum of CZTS features by far the strongest contribution of band tails85.

The origin of band tails in CZTS are potential fluctuations, which are eitherdue to band gap fluctuations or electrostatic potential fluctuations. The dif-ference between both kinds of potential fluctuations is depicted in Figure 3.1.Electrostatic potential fluctuations cause both bands to shift in parallel (to-gether with the vacuum energy) throughout the material and the band gap,i. e. the energy difference between valence and conduction band, remains con-stant. Electrostatic potential fluctuations are caused by charged point defects,by structural defects or by impurities86. In general, potential fluctuations areexpected to occur at regions with a high defect density, such as dislocations,grain boundaries and interfaces. If on the other hand band gap fluctuationsare present conduction band and valence band fluctuate independently withreference to the vacuum energy throughout the material84. As a consequencethe energy difference between valence and conduction band is not constantanymore but the band gap varies throughout the material. Band gap fluctua-tions can be induced by composition or crystalline inhomogeneities, secondaryphases or stress in the material.

In any real material a combination of both kinds of fluctuations is expectedto occur. However, several studies have identified band gap fluctuations as thedominating cause for band tails in CZTS81,80. Considering Cu-Zn disorder,as discussed in Chapter 4, both types of potential fluctuations are possibleconsequences. On one hand Cu-Zn disorder causes band gap changes whichcould lead to band gap fluctuations. On the other hand, clustering of Zn-rich and Cu-rich motifs87 could create charged domains which could generateelectrostatic potential fluctuations.

3.2 Defects in CZTSDefects are features in the crystal where the crystal symmetry is broken, i. e.where the atomic arrangement deviates from the ideal crystal structure. Suchdefects can be extended defects, such as dislocations, stacking faults, grainboundaries, interfaces or the surface. Defects that only involve one lattice

41

Figure 3.1. Effect of potential fluctuations vs. band gap fluctuations. Both result inthe formation of tail states in the band diagram.

Figure 3.2. Illustration of intrinsic point defects in a crystal.

site are called point defects. Different kinds of point defects are depicted inFigure 3.2. Point defects that cause deviation from the octet rule are easilyionized by binding an electron (acceptor) or releasing an electron (donor). InCZTS the Cu on Zn antisite (CuZn) and the copper vacancy (VCu) have thelowest formation energies and are therefore expected to be the most commondefects88,56. Both of these defects are acceptors and therefore generate holesas carriers in the crystal, yielding p-type conductivity for CZTS. Indeed, p-type conductivity has been confirmed in several experimental reports89,90.

Free electrons or holes can interact with ionized defects and by neutralizingthem, the free carriers are bound or trapped by the defect. If the ionizationenergy is small, in the order of the thermal energy kBT , the carrier is easilyreleased again and the effect of the defect is less severe. Defects with smallionization energy are called shallow defects. Analogously, defects with anionization energy much larger than kBT are called deep defects. Instead ofreleasing the bound carrier after ionization, it is more likely that the carrier re-

42

Box 3.2: Photoluminescence

Photoluminescence is the emission that is released upon radiative recombina-tion of an electron and a hole, which were prior excited by an incoming photon.After excitation, the carriers thermalize to the band edges and possibly defectstates within the band gap before recombining. A photoluminescence spec-trum therefore contains information about the recombination mechanisms of amaterial. The energy of the photoluminescence indicates the energy differenceof the bands or the energy position of defects within the bands. The photolu-minescence yield gives information about the rate of radiative recombination.

Different recombination processes can be identified inphotoluminescence spectra: (a) band-to-band recombi-nation, (b) free-to-bound recombination and (c) bound-to-bound recombination, also called donor-acceptor-pair (DAP) recombination (see Figure)). They can bedifferentiated by the energy of the emission, their de-pendence on the intensity of the incoming light, theirtemperature behavior or by their transients.

combines with another hole or electron. Therefore, deep defects act as recom-bination centers and are detrimental to devices. The VCu in CZTS is regardedas a shallow defect, while CuZn has a higher ionization energy and is morelikely to act as a recombination center30. Recombination has a detrimentaleffect on the open circuit voltage VOC of solar cells (see also Equation 1.4) andthe investigation of recombination processes is therefore of wide scientific in-terest. A common technique to investigate radiative recombination processesin a material is by photoluminescence spectroscopy (see Box 3.2).

So far only intrinsic defects have been considered. Extrinsic defects areformed if another element is introduced to the crystal, which is not part of theideal lattice. A typical dopant for CZTS is sodium Na. Na in the CZTS thinfilm improves the device performance by enhancing the growth and improv-ing the electrical properties. However the exact reason for the improvementis still under debate. One assumption, which has been proposed for the simi-lar material Cu(In,Ga)Se2, is that Na forms antisites on Cu sites NaCu duringthe growth but is released later leaving behind the shallow acceptor VCu

91.Another assumption is that Na locates at the grain boundaries and passivatesthem92,93. A positive effect has been also observed for other light alkali el-ements94,91 and hydrogen95. Another dopant with positive effect on deviceperformance is germanium Ge. A reduction of band tailing and efficienciesof up to 12.3% were achieved with a low open voltage deficit of 548 meV bysupplying relatively high doping levels of up to 1 % of Ge96. Again, the ex-act mechanisms for the improvement are unclear. Grain boundary passivation,reduction of Sn-related defects and a positive influence on the growth mecha-

43

nism reducing Sn loss have been hypothesized. The research covered by thisthesis mainly concerns intrinsic point defects and defect complexes. Dopingwith other elements and extrinsic defects will not be elaborated further.

3.3 Defect complexesDue to the coulomb interaction between two ionized defects, it is favorable toform defect complexes compared to independent point defects. This is illus-trated by the following reaction describing the equilibrium between two singledefects and a defect complexes

V−Cu +Zn+Cu

⇀↽ (V−Cu +Zn+Cu) . (3.1)

In CZTS, defect complexes have much lower formation energies and appearin much larger numbers than independent point defects. Defect complexesdo not contribute to the doping level of the material, because the charge ofthe involved ionized defects is compensated and the defect complexes have anoverall neutral charge.

The defect complex (CuZn +ZnCu) has the lowest formation energy97,88,which explains its abundance and Cu-Zn disorder in CZTS. The Cu-Zn anti-site defect complex affects the band gap77 and may be one of the reasons forband gap fluctuations. This defect complex and Cu-Zn disorder is discussedthoroughly in Chapter 4.

The Cu-Zn defect complex and the other defect complexes of opposite an-tisites, (CuSn + SnCu) and (SnZn +ZnSn), do not cause composition fluctua-tions. Off-stoichiometric single phase CZTS is therefore connected with theformation of other defect complexes. As a result of the width of the phasestability region, a large amount of defects are expected for off-stoichiometricCZTS and it is reasonable to assume that the majority of these defects arrangein defect complexes. A set of likely defect complexes was established to ex-plain off-stoichiometric CZTS. The proposed defect complexes were based onthe observed compositions of single phase Cu2ZnSn(S,Se)4 and the point de-fects identified by neutron diffraction on Cu2ZnSnSe4 powders29,63,98,62. Thedefect complexes are summarized in Table 3.1 and the resulting compositionrange based on each defect complex is visualized in the composition diagramin Figure 3.3. CZTS can contain a mixture of two neighboring types in orderto yield compositions across the whole single phase composition region.

So far, the experimentally verified defect complexes include A-, B-, D-, F-,and G-type99,100,62,58,101. Defect complexes forming in Sn-rich compositionscould not be studied by neutron diffraction due to the limitations in powderfabrication to produce Sn-rich CZTS as discussed in Section 2.3. Based on theframework established in Paper III, which was also explained in Section 2.3,particularly Reaction 2.5, defect complexes A, E, H and J will be affected bythe SnS and S2 partial pressures during fabrication and indeed these are the

44

Figure 3.3. CZTS types against composition. The typical composition region of pre-cursors used for device fabrication is marked in green.

Table 3.1. CZTS types and their intrinsic defect complexes, according to Refs. 29,99,100,62,58.

CZTS-type defect complex CZTS-type defect complexA-type [VCu +ZnCu] D-type [CuZn +Cui]

B-type [2ZnCu +ZnSn] C-type [2CuZn +SnZn]

E-type [2VCu +SnZn] or F-type [2Cui +ZnSn] or[2SnCu +VZn +VCu] [CuSn +Zni +Cui]

G-type [ZnSn +Zni] H-type [SnZn +VZn]

I-type [CuSn +3Cui] J-type [SnCu +3VCu]

45

CZTS types that have proven to be challenging to achieve in powder CZTSsamples. It should be noted, that the affected defect complexes are also theones that contain vacancies.

The charge compensation in defect complexes implies that they are not asharmful recombination centers as independent, charged defects. However,since defect complexes disrupt the crystal periodicity they can still act asscattering centers and reduce the life time of free carriers. Furthermore, firstprinciple DFT calculations predict that defect complexes cause changes in theband gap which lead to band gap fluctuations97,30. Especially defect com-plexes involving the defects SnZn and SnCu are expected to cause band gapfluctuations of up to 0.4 eV30. In contrast, A- and B-type defect complexesare more benign to the material properties and should not have a major effecton the device performance.

Due to the dependence of the occurrence of defects on the CZTS composi-tion, the composition has a major impact on the solar cell performance. Basedon theoretical predictions, Cu-rich CZTS is expected to yield a very high den-sity of CuZn antisites, causing high doping levels and detrimental recombina-tion centers88. The deep defect SnZn in Sn-rich CZTS will have severe effectson the non-radiative recombination rate, considering that not all defects willbe paired as (more benign) defect complexes. On the other hand, the shallowacceptor VCu should have a positive effect on the solar cell performance inCu-poor CZTS88. The A-type defect complex is expected to have the mostbenign effects on the solar cell performance97. The theoretical predictions ex-plain why CZTS solar cells with highest efficiencies are fabricated with CZTSwith Cu-poor and Zn-rich composition13,19,14.

The experimental investigation regarding the nature of defects in CZTSis challenging based on the high defect densities. The investigation ofcomposition-spread samples can offer a broader picture of the impact ofcomposition on the effects of intrinsic defects. Devices fabricated fromcomposition-spread CZTS samples yielded best device performance in the re-gion of A-type and B-type CZTS, however with different maxima with regardto composition for the current JSC and the voltage VOC

102,103,104.In Paper V the effect of CZTS composition on the photoluminescence in-

tensity is studied using composition-spread CZTS thin films. Figure 3.4 showsthe PL yield against composition for two composition-spread CZTS thin filmsannealed under different conditions. The photoluminescence intensity doesnot vary significantly throughout the single phase region of both sampleswhich implies that the main path for non-radiative recombination is not causedby composition dependent defect complexes. Instead, the overall low photo-luminescence yield throughout the single phase region indicates the presenceof a detrimental defect or defect complex unaffected by the composition. Dif-ferent explanations are possible, such as a charged defect with low formationenergy throughout the single phase region, or S-related defects such as VS,which are predicted to act as a detrimental recombination sites from theoreti-

46

Figure 3.4. Photoluminescence yield of composition-spread CZTS thin films annealed(a) under typical anneal conditions and (b) in higher S and SnS partial pressures.Secondary phase boundaries are marked with white dotted and dashed lines. Adaptedfrom Paper V.

cal calculations105. Another possible explanation could be recombination dueto extended defects, such as dislocations, stacking faults or grain boundaries.Such defects appear in much higher number in CZTS with stoichiometric com-position compared to Sn-rich CZTS, as discussed in the previous chapter (seeFigure 2.4).

A low photoluminescence yield implies that non-radiative recombinationis dominating in a material. Because recombination has severe effects on theopen circuit voltage VOC (as explained by Equation 1.4), the photolumines-cence yield is also an indicator for the potential of an absorber material toreach a satisfactory open circuit voltage106. Hence, the low photolumines-cence yield throughout the whole CZTS single phase region is worrisome,because it indicates that the VOC will be limited for CZTS solar cells indepen-dent of the the cation composition.

This chapter summarizes the electronic properties of CZTS including the bandstructure and the effect of different types of defects. Band gap fluctuations andelectrostatic potential fluctuations are exploited as origin for the severe levelof band tailing observed for CZTS and the effects of extended defects, pointdefects and defect complexes are discussed. Based on findings from litera-ture, the defects SnZn and SnCu act as detrimental recombination centers andtherefore Sn-rich compositions should be avoided for CZTS devices. Our ownresults from photoluminescence investigations on composition-spread CZTSfilms do not reflect the expected dependence on the composition. Instead, itappears that a composition-independent defect is limiting the photolumines-cence yield in CZTS throughout the whole single phase region.

47

4. Cation disorder in CZTS

The concept of order and disorder in solids is a wide field of research on itsown. It stretches from short-range order in amorphous materials to long-rangeordered crystals which are categorized by their lattice symmetries and is ap-plied to understand the solid-liquid phase transition. This chapter summa-rizes the current understanding of cation disorder in Cu2ZnSnS4 where thelong-range order is broken. Different theories are introduced describing long-range order in solids and their concepts are used to interpret cation disorder inCu2ZnSnS4 and its implications.

4.1 Disorder in CZTSThe CZTS crystal is made up of copper (Cu), zinc (Zn), tin (Sn) and sulfur (S)atoms (see also Fig. 2.1). Possible candidates for antisite pair formation, whichcould lead to disorder in the crystal, are the cations Cu, Zn and Sn. In theliterature the term “disorder” can also be found applied to the occupation of theanion site in Cu2ZnSn(S,Se)4 (CZTSSe) containing sulfur and selenium107.However, this thesis is concerned with the pure sulfur compound and disorderinvolving the anion site is not considered.

The crystal radii and charge states of the cations of CZTS are listed in Ta-ble 4.1. Zn and Cu ions have the same size in the CZTS crystal which shouldfacilitate antisite formation between both atom kinds, whereas the Sn cationis somewhat smaller and exchange of Sn with either Zn or Cu is less likely.Another indicator is the charge of the cations. The more similar the cationsare in their charge, the more likely it is for them to form an antisite defect pair.This is reflected by the calculated formation energies88 which are listed for allcombinations of the three cations in Table 4.2. As predicted from the cationradii and charge, the formation energy of the antisite pair involving Cu and Znis the lowest and therefore the most likely antisite pair formed in CZTS.

The formation energies for all three antisite pairs are positive and the defectpair is formed at the cost of energy. The crystal without defect pairs should bethe energetically favorable state for the system. However the free energy F ,which indicates if a system is energetically favorable, is defined as

F =U −TS . (4.1)

The formation energy is contained in the internal energy U . The second termin Equation 4.1 is a product of the temperature T and the entropy S which is

48

Table 4.1. Effective cation radii for crystals of coordination number 4 (tetrahedralstructures) and the respective charge states of the cation, according to Ref. 108.

Ion Charge Cation radius (in Å)Cu +1 0.74Zn +2 0.74Mn +2 0.8Sn +4 0.69

Table 4.2. Formation energies (in eV) for antisite pairs from Ref. 88.

Defect cluster Cu−Zn + Zn+Cu Cu3−Sn + Sn3+

Cu Zn2−Sn + Sn2+

Zn

Formation energy (eV) 0.21 1.99 0.86

defined asS = kB lnΩ . (4.2)

The entropy S depends on the Boltzman constant kB and on the number ofpossible states or configurations of a system Ω. If Ω increases, the entropy ofa system increases. Based on the negative sign in Equation 4.1 higher entropyis favorable for a system (note that this applies at finite temperature, i. e. T >0). In the case of CZTS, the perfect crystal without any defects has onlyone possible configuration and the entropy is zero. If a Cu-Zn antisite pairis introduced, four different configurations are possible considering only theCu-Zn planes of one unit cell. Therefore, the entropy of the system is increasedby introducing a Cu-Zn disorder pair, which results in a lower free Energy Fand an energetically more favorable state of the system.

Cu-Zn disorder was experimentally verified by neutron diffraction on CZTSpowders35. It was mainly observed within the Cu-Zn planes (Rietvelt posi-tions 2c and 2d), while the 2a Rietvelt position in the Cu-Sn planes appearedunaffected and only occupied by Cu. Later studies show contradicting resultsas to whether the Cu atoms on 2a positions are involved in the Cu-Zn disordertransition109,100. Theoretical studies also claim that Cu-Zn disorder shouldnot be restricted to the 2c and 2d positions of Cu-Zn planes, but also involve2a Cu sites in the Cu-Sn planes. Although, there may be a preference for Znto form antisites within the Cu-Zn planes97,110. To this date it is not certainwhether the 2a sites are involved in the Cu-Zn disorder and if so under whichconditions. It is plausible that composition and morphology of CZTS play arole.

The direct measurement and quantification of Cu-Zn disorder in CZTS isdifficult, especially for thin film samples. Common methods to quantify Cu-Zn disorder are neutron diffraction63,28, resonant X-ray diffraction111, syn-chrotron radiation X-ray diffraction112 and anomalous diffraction113. Thesemethods allow the analysis of the occupation of certain crystallographic sitesincluding the fraction of disordered and ordered Cu and Zn atoms. How-ever, they require larger samples sizes, typically in the form of powder sam-

49

ples and are unsuited to study thin film samples. One method to observecation disorder in thin film CZTS is atomic resolution transmission electronmicroscopy114,115. But this method is impracticable to study disorder on amacroscopic scale and for a large sample number. Instead, Cu-Zn disorder inCZTS thin films is investigated indirectly by studying effects of disorder onoptical and structural properties of the material. The effect of Cu-Zn disorderon the material properties of CZTS and possible methods to study disorder inCZTS thin films will be further explained in Section 4.3.

4.2 The concept of disorder in solids – putting a numberon disorder

Bragg and Williams published a series of papers about disorder in solids dur-ing the 1930’s to explain experimental observations of atomic re-arrangementin crystalline Au-Cu and Fe-Al metal alloys for temperature treatments wellbelow their melting temperatures116. The order-disorder transition (ODT) in asolid differs from other phase transitions in being a continuous change of theatom arrangement rather than an abrupt reorientation as is observed for a tran-sition between two crystal structures. This is why this transition is classifiedas a second order phase transition.

To demonstrate the concept of disorder in solids one can define a binarycompound with a ground state structure where the lattice sites α are occupiedby the atom type A and lattice sites β are occupied by atom type B. If thecompound experiences disorder, some lattice sites α may be occupied by atomtype B and some lattice sites β by atom type A. Bragg and Williams introducethe degree of order S for a material116 based on Pα

A , the current probability ofthe site α to be occupied by aatom type A, and r, the probability of the site αto be occupied by A at complete disorder,

S =Pα

A − r1− r

. (4.3)

Note that the degree of order S is not identical to the entropy S of the systemwhich is defined in Equation 4.2. In Equation 4.3 the degree of order is definedin such a way that S = 0 if A and B are randomly distributed among α and βsites and S = 1 if all α sites are occupied by A (and all β sites by B).

For a theoretical interpretation, Cu-Zn disorder in CZTS is usually reducedto disorder in the Cu-Zn planes for simplification. This way, the kesteritestructure can be reduced to a two-dimensional crystal of Cu and Zn atoms andthe degree of order S from Equation 4.3 can be rewritten to

S = 2PCuCu −1 , (4.4)

50

Figure 4.1. a) Sketch of the equilibrium state of the degree of order Se at temperature Twith Tc the critical temperature. b) Equilibrium curve for CZTS based on experimentalobservations. Taken from Ref. 117.

with PCuCu the probability to find a Cu atom on a Cu site (on the Cu-Zn planes).

If the crystal is fully ordered, PCuCu = 1 and S = 1. If the crystal is completely

disordered, PCuCu reduces to 0.5 and S becomes zero.

The formation of disorder in a crystal is driven by the entropy, as has beenexplained in the previous section. According to Equation 4.1 the effect ofentropy on the free energy depends on the temperature. Therefore, also theequilibrium state of the degree of order Se depends on the temperature T ofthe system. Complete order with Se = 1 is achieved only at T = 0 K becausethen the term with the entropy S in Equation 4.1 becomes zero and the mostfavorable state of the system is the ordered crystal with the lowest formationenergy.

High temperature implies that the entropy has a stronger effect on the freeenergy and disorder becomes more favorable. At the same time thermal agi-tation increases, which facilitates the formation of antisite pairs when atomsexchange sites with their neighboring atoms. As the temperature increases,the number of antisite pairs, i. e. the degree of disorder, increases and the sys-tem will approach an equilibrium state with a degree of order 1 > Se > 0. Atsome temperature the lattice sites α and β are randomly occupied by A andB, i. e. Se = 0. This temperature is called the critical temperature Tc. For anytemperature above Tc the degree of order will be zero. The dependence of thedegree of order Se on the temperature for a system in equilibrium is depictedin Figure 4.1.

Scragg et al.117 investigated the degree of order of CZTS after thermal treat-ments at several temperatures and determined the equilibrium curve for CZTS(see Fig. 4.1). They found the critical temperature of the ODT in CZTS at265 ◦C. In the selenide compound Cu2ZnSnSe4 (CZTSe) the critical tempera-ture was found somewhat lower at 200 ◦C77.

Both critical temperatures are well below typical annealing temperaturesthat CZTS and CZTSe samples are exposed to during fabrication (see Sec-

51

tion 2.2). The kesterites should be completely disordered in the Cu and Znlattice sites during the high temperature anneal. Once the annealing time isup, the samples are often cooled down quickly and the Cu and Zn atoms donot have enough time to undergo ordering transitions. Instead, the disorderedstate of the system is frozen and at low temperatures the ordering rate is neg-ligible. Therefore, it can be assumed that the samples are highly disorderedwithout further ordering treatments.

The rate at which a system approaches its equilibrium state Se(T ) dependson certain system parameters and on the temperature. For higher temperaturesthe rate will be higher, for lower temperatures it can require a relatively longtime to reach the equilibrium state. This way it is possible to “freeze” a crystalin a state of low order by fast cooling to lower temperatures where the orderingkinetics are very slow. This implies, that the system is not in equilibriumanymore (i. e. S �= Se).

The theory of Vineyard118 is more suited to study the degree of order for asystem that is not in equilibrium. Vineyard derives a relation for the orderingkinetics dS

dt by introducing the rate of ordering KO and the rate of disorderingKD for a system. If the system is at equilibrium KO = KD, and KO > KD (KO <KD) if the degree of order is lower (higher) than Se(T ). For a stoichiometriccompound (Nα = NA and Nβ = NB) the relation for the ordering kinetics isgiven as

dSdt

=1fB

[KO fB fA(1−S)2 −KD( fA + fBS)( fB + fAS)

], (4.5)

where fA and fB are the fractions of atom kind A and B.The theory of Vineyard was applied to the disorder in the Cu-Zn planes of

Cu2ZnSnSe4 by Rey et al.77 and Equation 4.5 can be rewritten to

dSdt

=12[KO(1−S)2 −KD(1+S)2] . (4.6)

Assuming a direct exchange mechanism, the rate constants KO and KD fortwo-dimensional Cu-Zn planes can be expressed as

KOKD

= 4 f exp(−U

kBT

)exp

(±3vSkBT

), (4.7)

with v the nearest neighbor interaction energy, f the frequency of the latticevibration of the atomic interchange and U the activation energy associatedwith the atomic interchange.

The nearest neighbor interaction energy v is directly proportional to thecritical temperature with the relation

vkB

=23

Tc , (4.8)

52

Table 4.3. Kinetic parameters of the order-disorder transition in B-type CZTS fromPaper I.

f UkB

vkB

9×1012 Hz 18 500 K 358 K

1 100 10000 1000000 1E80.0

0.2

0.4

0.6

0.8

1.0S=0.85~ 27 years

Calc.orderparameterS

Time (s)

Degree of order S

S=0.4~ 1.5 h

S=0.6~ 10 h

S=0.7~ 4.5 days

S=0.8~ 1 year

0

50

100

150

200

250

300Optimized cooling

Temperature(°C)

Figure 4.2. Optimized cooling profile with maximum achievable degree of orderthrough thermal treatments. Taken from Paper I.

which was derived in Paper II. In Paper I, the remaining two kinetic parametersf and U were determined for the ODT in B-type CZTS as listed in Table 4.3.Once the parameters f , v and U are known, the degree of order S can beestimated for an arbitrary temperature profile T (t) by numerical integration ofEquation 4.6 with reasonably small time intervals dt.

Furthermore, in Paper I an optimized cooling profile is derived from Equa-tion 4.6 which gives the highest possible degree of order by thermal treatmentswithin a certain time. The optimized cooling profile together with the evolu-tion of the degree of order S are illustrated in Figure 4.2. On one hand, thisfigure can be regarded as a guide to optimize thermal treatments to achieve ahigh degree of cation order in CZTS. On the other hand, it reveals that ther-mal treatments alone cannot effectively reduce Cu-Zn disorder based on theslow ordering kinetics at lower temperatures. Though, it should be noted thata higher degree of cation order S > 0.8 upon thermal treatments has been ob-served in single crystal CZTS98,74, which indicates that the ordering kineticsdiffer for thin film samples and single crystals.

Applying theories of disorder in solids to Cu-Zn disorder in CZTS natu-rally comes with limitations and weaknesses. The theories are limited to com-pounds with stoichiometric composition, and an initial state of order S which

53

is not too far away from the equilibrium state of order. Both conditions typi-cally do not apply to CZTS, which is commonly fabricated with Cu-poor andZn-rich composition and a degree of order much lower than the equilibriumstate. Furthermore, the assumption that the Cu-Zn disorder is limited to Cu-Zn planes has been questioned by several studies. Other complications maybe considered, such as clustering of disordered motifs within the CZTS crystalwhich was proposed by Zawadzki et al.87. Despite the mentioned weaknesses,these theoretical models provide a better general understanding of the natureof disorder in CZTS and the ordering process and yield the possibility to es-timate the effect of certain treatments on the degree of order and reasonablelimitations of the material.

4.3 Cu-Zn disorder in kesterite CZTSThe effect of Cu-Zn disorder on the CZTS material properties can be stud-ied by exposing the material to different thermal treatments below the criticaltemperature which should change the degree of order of the lattice. It is impor-tant to distinguish between permanent structural changes that could arise uponthermal treatments and changes related to the order-disorder transition, whichshould be completely reversible upon further thermal treatments. This wayit is possible to correlate the observed changes upon thermal treatments withchanges in the degree of Cu-Zn order and the findings of such studies are sum-marized in this section. For the experimental studies of cation disorder of thisthesis, the low temperature treatments to promote cation order were performedin the same tube furnace which was also used for crystallization anneals of theprecursors as described in Section 2.2. This setup allowed the treatments totake place in a constant argon atmosphere and it allowed exact temperaturecontrol and monitoring of the sample temperature during the whole treatment.

As was discussed in Section 4.1, methods for direct quantification of thedegree of disorder in CZTS are not suited for the investigation of thin filmssamples. Therefore, other methods were developed to estimate the degreeof disorder in CZTS thin films by studying the effects of Cu-Zn disorder ondifferent CZTS properties.

The degree of Cu-Zn order has a drastic effect on the band gap of CZTS,which is narrowed substantially when decreasing the degree of Cu-Zn order.Valentini et al.119 observe a band gap change of up to 150 meV for one sampleupon thermal treatments. Theoretical predictions based on DFT calculationspredict changes of the band gap of 130 meV120 and 200 meV117 based onCu-Zn disorder which are in the range of experimental observations. The nar-rowing of the band gap is explained by a larger density of the defect pair(CuZn+ZnCu) which creates defect levels within the band gap and by thatcontributes to an overall reduction of the band gap. The effect of band gapchanges upon ordering can be observed by different techniques, such as spec-

54

Figure 4.3. a) Band gap widening after thermal ordering treatments observed in quan-tum efficiency spectra. b) Changes in the Raman spectra of CZTS thin films that canbe correlated with different degrees of order. Taken from Paper I.

trophotometry77, photoluminescence spectra121 or quantum efficiency (seeFigure 4.3 (a)). Consequently, Rey et al. utilized the effect of Cu-Zn dis-order on the band gap shift of CZTSe thin films to estimate the degree of orderS77 and proposed a method to analyze the degree of order for thin films.

Analogously, the effects of Cu-Zn disorder on the Raman spectra of CZTShave been correlated with the degree of order S31 to yield a method to ana-lyze Cu-Zn disorder in CZTS thin films. The Raman spectrum of CZTS (inbackscattering configuration) consists of the intense A mode at 338 cm−1 andanother A mode at 289 cm−1. Under resonant excitation, B and E modes be-tween 250 cm−1 and 380 cm−1 become clearly visible as well. Two majoreffects can be detected with reducing cation order: widening of the full widthhalf maximum (FWHM) of the main peak at 338 cm−1 and a change in thepeak height ratios involving the peaks at 289 cm−1, 305 cm−1, 367 cm−1, and377 cm−1 (see Figure 4.3 (b)).

The widening of the FWHM for more disordered CZTS seems plausiblebased on the assumption that disorder in the Cu-Zn planes compromises thecrystallinity of the CZTS material as a whole. With reduced crystal quality,the phonons in the crystal will be confined to a smaller volume. Accordingto the phonon confinement model, this causes the Raman modes to widen andshift slightly to smaller wavenumbers122, as is observed in the Raman spectrain Figure 4.3 (b) for CZTS with less cation order.

Paris et al.67 focus on the changes of the relative peak heights in the Ramanspectrum after thermal ordering treatments and define the two parameters Qand Q′ based on the peak heights of the Raman modes

Q =H289

H305and Q′ =

H338

H367 +H377. (4.9)

In Paper I it is proposed that the relative peak heights defining the parameterQ change because of a change of the crystal symmetry from ordered kesterite

55

I4̄ to disordered kesterite I4̄2m, which is the same symmetry group as thestannite crystal. Crystals of the kesterite symmetry group I4̄ feature three A-modes in their Raman spectra. In the symmetry group I4̄2m, two of theseA-modes are degenerate and instead only two A-modes are distinguishablein the Raman spectrum. Therefore, it is expected that one of the A-modesreduces in intensity as CZTS becomes more disorder and resembles a crystalwith I4̄2m symmetry. This could explain the reduction in intensity of the modeat 289 cm−1 which has been identified as one of the A-modes for CZTS. Thiswould imply, that a completely disordered CZTS crystal with I4̄2m symmetryshould not feature any intensity for the Raman mode at 289 cm−1, resulting inQ = 0.

In Paper I the Q parameter was used to quantify the degree of order for aseries of B-type CZTS thin films that had been exposed to different thermalordering treatments. The temperature profiles of the ordering treatments wereadditionally evaluated with the theory by Vineyard (Equations 4.6 and 4.7)and a linear relation S(Q) was derived

S = (0.18±0.02)Q . (4.10)

It should be noted, that this linear relation passes through zero, which is ex-pected if the changes in Q originate from a symmetry change of the CZTScrystal due to disorder as explained in the previous paragraph.

In Paper II the correlation between the parameters Q and Q′ and disorder inCZTS was verified for the full composition region of CZTS by investigationof composition-spread thin films. It was demonstrated that both parameterscorrelate with the FWHM of the main CZTS Raman peak at 338 cm−1 whichled to the conclusion that both parameters reflect the state of order of the CZTSmaterial. However, the Q parameter attained very different values for somecomposition regions (this is further discussed in Section 5.1) indicating thatthe relation S(Q) given in Equation 4.10 only applies to B-type CZTS.

Despite the direct effects of Cu-Zn disorder observed on the band gap andthe vibration modes of CZTS, the effect on solar cell devices is still debated.Based on the experimentally observed and calculated effects on the band gapone would predict that an increased degree of order implies less band gapfluctuations which should reduce the open circuit deficit in CZTS based solarcells. Instead, it has not been possible to reduce the open circuit voltage deficitby thermal ordering treatments23,123,124,25. The open circuit voltage was im-proved, but the band gap increases by the same amount and the deficit of theopen circuit voltage remained at the same value. Even though the open circuitvoltage deficit could not be reduced, thermal ordering treatments enhancedseveral material properties in CZTS, such as reduction of free carriers from1018 cm−1 to 1016 cm−1 and higher mobilities125, and higher luminescence in-tensity implying a reduction in non-radiative recombination124. Furthermore,a positive effect on the fill factor has been observed which led to an overallimprovement of device efficiencies25,123. A possible explanation why order-

56

ing treatments fail to improve the open circuit deficit is discussed in Paper I.It is argued that thermal treatments do not achieve the necessary degree of or-der in thin film CZTS to reduce the band tailing to a degree where it is notdetrimental to the open circuit voltage in solar cells.

57

5. Defect Engineering in CZTS

Defect engineering aims to control the defects in a material in order to tunethe materials fundamental properties, like the band gap and absorption, or(electrical) conduction and recombination processes126,127. In Cu2ZnSnS4 inparticular, defect engineering aims to reduce or to passivate Cu-Zn disorderand detrimental point defects like SnZn and to enhance the hole doping byVCu instead of CuZn. This has led to Cu-poor and Zn-rich compositions asthe preferred compositions for Cu2ZnSnS4 devices. In Papers I – V differentstrategies for defect engineering are exploited with the aim to control and re-duce the Cu-Zn disorder and subsequent band gap fluctuations. This chapterdiscusses these strategies and puts them into the context of strategies discussedin literature.

5.1 Reducing Cu-Zn disorder in CZTSThe most common way to control Cu-Zn disorder, i. e. the density of(CuZn+ZnCu) defect pairs, is by temperature treatments as described in Chap-ter 4. However as discussed in Paper I and illustrated in Figure 4.2, tem-perature treatments alone do not succeed in reducing Cu-Zn disorder suffi-ciently in conventional CZTS thin film absorbers, because they would requirean extremely long time to produce highly ordered CZTS. Other strategies areneeded to reduce the favorability of disorder (i. e. raise Tc) or to enhance theordering kinetics and by that boost the effect of temperature treatments.

As derived in Equations 4.6 and 4.7 the kinetics of the ordering process dSdt

are governed by the frequency of the lattice vibration of the atomic interchangef , the nearest neighbor interaction energy v, and the activation energy associ-ated with the atomic interchange U . In order to change the ordering kinetics,one has to find a way to manipulate one or more of these parameters. Thefrequency of the lattice vibration f depends on the mass of the involved atomsin the lattice and their bonds. This frequency is not very sensitive to minorchanges to the lattice such as doping and effects on this frequency should bevisible in the Raman spectrum. The nearest neighbor interaction v describesthe energy difference of the system before and after the cation exchange and isdirectly proportional to the critical temperature of the order-disorder transitionas was shown in Equation 4.8. U on the other hand is the activation energyof the cation exchange. Figure 5.1 visualizes the two parameters v and U . Toenhance the ordering kinetics, v should be increased, to boost the driving force

58

Figure 5.1. Energy diagram sketchingthe activation energy for the atomic inter-change U and the nearest neighbor interac-tion energy v. Taken from Paper II.

towards the ordered state of the system, while U should be lowered to enhancethe cation exchange process.

In Paper II the effect of the cation composition on the ordering kinetics isinvestigated on a composition-spread CZTS thin film with a continuous cationgradient. The Q parameter, which is proportional to the degree of Cu-Zn order(see Equations 4.9 and 4.10), was evaluated from resonant Raman spectra andis plotted against the composition in Figure 5.2. The ternary phase Cu3SnS4leads to a drastic reduction in the Q parameter in Zn-poor compositions out-side the single phase region. Otherwise, the composition of Zn within thesingle phase region does not seem to affect the ordering kinetics. However theQ parameter changed drastically with the Cu/Sn ratio, with lowest values forQ where the Cu/Sn ratio is stoichiometric (Cu/Sn= 2) and higher Q values inCu-rich and Cu-poor compositions. Further investigations indicated that thecritical temperature remains constant for different compositions. Therefore,the change in ordering kinetics can not be due to a change in v, but must becaused by a lowered activation energy U . The composition regions with higherQ parameter correlate with the composition regions where defect complexeswith vacancies or interstitials are expected. Such defects are known to enhancediffusion processes in other materials128. Therefore, the observed compositiondependence is interpreted as an enhancement of the ordering transition by de-fect assisted diffusion processes. These results show, that defect engineeringusing the cation composition to reduce Cu-Zn disorder in CZTS is possibleand that higher degrees of order can be achieved in compositions with a highdensity of interstitials or vacancies.

In Paper III (see also Section 2.3), it was shown that the solubility of defectcomplexes containing vacancies strongly depends on the processing condi-tions, in particular on the SnS and S2 partial pressures during the high temper-ature anneal. By supplying high partial pressures of S2 and SnS, the solubilityof defect complexes containing vacancies can be enhanced leading to highdensities of vacancies in Sn-rich/Cu-poor CZTS. This should further enhance

59

Figure 5.2. Q parameter from resonant Raman spectra across composition-spreadsamples annealed (a) under typical anneal conditions and (b) in higher S2 and SnSpartial pressures. The secondary phase boundaries are indicated with white dottedlines. Adapted from Paper V.

the impact of defect assisted diffusion on the ordering kinetics during ther-mal treatments. Therefore, the partial pressures during the fabrication processshould be considered when aiming for a strong effect of vacancy assisted Cu-Zn ordering in Sn-rich compositions.

This is demonstrated in Papers III and V with composition-spread CZTSthin films annealed in different S2 and SnS partial pressures, see Figure 5.2 (a)and (b). For the sample annealed in higher S2 and SnS partial pressures (Fig-ure 5.2 (b)), the secondary phase boundary of SnSx is shifted to a lower Cu/Sncomposition ratio, indicating that the CZTS phase is less likely to decomposefor Sn-rich compositions and higher defect densities in CZTS should be at-tained. Indeed, the Q parameter is strongly enhanced in the Sn-rich composi-tion region indicating a higher density of vacancies. The Q parameter was evenfurther enhanced in Zn-poor composition regions close to the secondary phaseboundary of Cu3SnS4. This effect has not been explained yet. It could be con-nected to the presence of CuZn antisites, either as independent point defectsor in defect complexes. The full width half maximum of the main vibrationmode in the Raman spectra showed the same trends across the composition-spread samples as the Q parameter, confirming that Q predicts changes in thecrystalline quality.

The results presented in Papers III and V are promising that there is stillroom for improvement in the ordering kinetics through defect engineeringby tuning the density of the internal CZTS defect complexes. The resultsdemonstrate that it is important to consider not only the cation composition ofthe precursor, but also the annealing conditions to reach CZTS with desirableproperties.

CZTS based solar cells suffering from band tailing due to the cation disor-der could benefit from a higher density of vacancies in Sn-rich compositions,especially with a high density of A-type defect complexes (VCu + ZnCu). First

60

Figure 5.3. Elements reported as doping or alloying candidates for CZTS. Adaptedfrom Ref. 129.

principle calculations predict that the reduction of the band gap by the Cu-Zn antisite complex can be counteracted, or passivated, by the A-type defectcomplex due to its opposite effect on the CZTS band gap97. Other studieshave reported beneficial effects of Sn-rich composition on CZTS properties,including blue shift of the band gap and reduced sub-band gap absorption79

and a difference of the ordering kinetics depending on the composition to-wards A-type CZTS121. However, a direct comparison of the reported CZTScompositions and the observed effects on cation disorder and potential fluc-tuations is difficult. The defect nature and defect densities are likely to differbased on the different employed fabrication conditions and the difficulty toderive the cation composition of the CZTS phase from the integral film com-position. Unfortunately, not many studies are available about the effect ofCu-rich compositions. For such compositions the effects should be easier tocompare, since the partial pressures during the anneal should not have such astrong impact on the defect nature of the CZTS phase.

5.2 Alloying CZTS with other elements for defectengineering

Another method for defect engineering is by doping or alloying with an im-purity element. Both methods have been exploited with several different ele-ments (see the periodic table depicted in Figure 5.3) with the hope to enhanceCZTS material properties and to improve the device performance. Dopingis typically referred to when adding only small amounts of an impurity ele-

61

ment with the goal to tune electronic properties, but without changes to thecrystal structure of the host material. While the term alloying is referred towhen larger amounts ( > 1 mol.-%) of another material are introduced to thehost material, forming a solid solution. Initially alloying was mainly used todescribe metal alloys. In CZTS alloying is often regarded as a partial atomreplacement27.

The most common element for alloying with CZTS is selenium (Se). It re-places sulfur in the host lattice and is used to tune the band gap between 1.6 eVwith no selenium down to 1.0 eV when replacing all sulfur by selenium11,12.The highest efficiencies for kesterite solar cells were achieved with relativelyhigh selenium contents (Se/(Se+S)>0.75)19,14. This atomic exchange alsoeffects the order-disorder transition by reducing the critical temperature toTC,CZTSe = 200 ◦C77 and by lowering the activation energy U for the cationexchange compared to CZTS. Despite the lower activation energy, CZTSe suf-fers from severe Cu-Zn disorder. The highest degree of order S reported forCZTSe thin films is at around 0.777, which is similar to the amount of disorderexpected in CZTS after low temperature ordering treatments (see Paper I).

Other impurity elements were reported to have a positive effect on the Cu-Zn disorder. One of them is silver (Ag), which replaces Cu in the host CZTSlattice. Theoretical calculations predict a higher formation energy of the Zn-Ag antisites130,131 and a neutron diffraction study shows that in fact a partialsubstitution of Cu by Ag suppresses Cu-Zn disorder132. Characterizing theoptical properties of (Cu,Ag)2ZnSnS4 revealed reduced band tailing when theAg content is increased133,134. The downside of alloying with Ag is that thealloy converts to n-type for higher silver contents135 which results in poordevice performance using the conventional devices stack for kesterite solarcells.

Cadmium (Cd) has also been alloyed with CZTS where it replaces Zn. Cal-culations predict that already small amounts of Cd can have a positive effecton the Cu-Zn disorder131. Relatively high efficiencies were achieved withdevices that were produced with Cu2(Zn,Cd)SnS4

136,137 and one of the ob-served improvements through Cd alloying was a reduction in band tailing136.However, alloying can have several effects on the material properties, e. g. thealloyed element can act as a fluxing agent during the growth (i. e. facilitate thegrowth and crystallization of CZTS) and suppress secondary phase segrega-tion131. Therefore, the observed reduction of band tailing may be caused byother effects of alloying rather than a reduction of Cu-Zn disorder.

The candidates for alloying with CZTS presented here have been chosen asexamples for the most promising results in terms of device performance andtheir effect on Cu-Zn disorder. However, when choosing suitable elementsfor device fabrication other considerations should not be neglected, such astoxicity (Cd) and scarcity (Se and Ag)10. Alloying CZTS with toxic and scarcematerials contradicts the initial attraction that caused the immense interest inCZTS in the first place.

62

Figure 5.4. Kesterite and stannite crystal structure for Cu2MnSnS4. Stannite is theground structure for CMTS based on first principles calculations. Taken from Pa-per IV.

5.3 Exploring cation exchange for materials withpreferable properties

An extreme version of alloying is to completely exchange one element in thecompound and to form a new compound with the goal to reduce unfavorabledefects while sustaining beneficial properties. In the case of CZTS, a cationexchange should ideally reduce cation disorder, but maintain an electronicstructure with a direct band gap around 1.5 eV and a high absorption constant.In Paper IV the compound Cu2MnSnS4 (CMTS) is investigated as a potentialalternative to CZTS. The Zn cation is replaced by manganese (Mn), which isa non-toxic and abundant transition metal just as Zn. Compared to other abun-dant transition metals, Mn features the largest difference in cation radius withCu, which should reduce the probability for cation disorder. Calculations pre-dict the compound to be stable towards decomposition in the stannite groundstate structure138, which has been confirmed by neutron diffraction139. Thecation arrangement in stannite CMTS is depicted in Figure 5.4. The com-pound CMTS has a direct band gap at 1.61 eV with high absorption140.

In Paper IV CMTS thin films of different cation compositions were fab-ricated by adapting the two stage process described in Section 2.2. Thethin films were carefully characterized by X-ray diffraction and Raman spec-troscopy, which gave evidence for several secondary phases besides the mainCMTS phase. Absorption spectra revealed band gap energies in the range 1.42– 1.59 eV depending on the cation composition and a high absorption coeffi-cient in the order of 104 cm−1. However, low temperature treatments between160 ◦C and 300 ◦C cause reversible changes in the band gap, the Urbach en-ergy and the absorption constant which suggest that CMTS suffers from anorder-disorder transition with a critical temperature of 230±10 ◦C (see Fig-

63

160 180 200 220 240 260 280 3001.60

1.61

1.62

1.63

1.64

Band gap

Bandgap(eV)

Annealing temperature (°C)

70

80

90

100

110

Urbach tail

Urbachtail(meV)

1.4

1.6

Absorptionconstant

AbsorptionconstantA(105cm

-1eV

-1/2)

Rangefor TC

Figure 5.5. Reversible changes in band gap energy, Urbach tail energy, and absorp-tion constant after thermal treatments at different temperatures. The inferred criticaltemperature of the order-disorder transition is marked by the grey area. Taken fromPaper IV.

ure 5.5). Considering the calculated low formation energy of the Cu-Mn an-tisite pair in CMTS120 it is proposed that the order-disorder transition occursamong the Cu and Mn atoms and causes the observed effect of low tempera-ture treatments.

This finding is curious as it demonstrates that cation disorder is probablynot restricted to the kesterite lattice, but also exists in the stannite structure.The stannite structure was thought to be immune to cation disorder, becauseobservations based on neutron diffraction found no sign for cation disorderin the stannite Cu2FeSnS4

35. Furthermore, cation disorder in CZTS is oftenargued to be motivated by the almost equal cation size and the arrangementof the cations in neighboring positions within the same lattice planes. Bothfeatures do not apply to stannite CMTS.

The interpretation that stannite CMTS suffers from Cu-Mn disorderin the same way as kesterite CZTS is supported by the calculations ofantisite-formation energies in several compounds with the chemical structureCu2XSnS4

120. Replacing Zn by beryllium (Be), magnesium (Mg), Mn, iron(Fe), or nickel (Ni) does not notably affect the formation energy of the defect

64

pair of the (CuX+XCu) antisite and cation disorder is a likely consequence inall of these compounds, unaffected by their ground state structure.

Nevertheless, different interpretations of the presented results of Paper IVare possible. Previous studies did not consider the effect of disorder duringtheir assignment of the crystal structure of CMTS and it is possible that thestannite phase was assigned by mistake, similar to the discussion of the cor-rect crystal structure of CZTS28. Therefore, two alternative interpretationsare given in Paper IV, namely that either kesterite is the actual ground statestructure of CMTS, or even though stannite is the ground state structure, theformation energy of disordered kesterite could be more favorable than the stan-nite configuration. In conclusion, a disorder-free material should (i) not favordisorder in its ground state structure and (ii) lack structural polymorphs thatexhibit cation disorder and have similar formation energies as the ground statestructure.

The conclusions from Paper IV demonstrate that cation disorder and associ-ated band tailing is not easily avoided by cation replacement with manganeseor even other cations. Cation disorder seems to be an inherent characteristicof the kesterite and the stannite structures. More drastic structural changes arenecessary to diminish the effect of cation disorder. For example, the compoundCu2BaSnS4 crystallizes in a trigonal structure which impedes the formation ofCu-Ba antisites. In fact, a less severe effect of band tailing is observed in thephotoluminescence and absorption characteristics of the compound141. Nev-ertheless, solar cells fabricated with a Cu2BaSnS4 absorber layer only reachedefficiencies of 5.2 %142.

Even though plenty of studies explore the possibilities of alloying andcation exchange in CZTS26,27, none succeeded to improve the efficiency ofkesterite solar cells. This issue could be related to the difficulty in optimizingcritical material properties (such as composition and intrinsic defects), and tofinding suitable materials to compliment the material of the absorber layer inthe full solar cell stack.

65

6. Concluding remarks

The previous chapters discuss the findings of the research conducted through-out this PhD in the broader context of the general understanding of CZTS andits material properties. This chapter, however, solely summarizes the resultsand conclusions that were attained in the scope of this PhD in terms of defectengineering in kesterite Cu2ZnSnS4. Section 6.2 is a personal reflection ofthese results and provides an outlook for future research efforts.

6.1 Summary of the results of this thesisDefect engineering aims to control the densities of important or detrimentaldefects in a material. The understanding of the kinetic and thermodynamiceffects that govern the formation of these defects can be the key to achieve thisgoal, or imply limitations of defect engineering in a material. In a defect richmaterial like Cu2ZnSnS4 (CZTS), developing methods for defect engineeringpresents additional challenges because it becomes difficult to distinguish theeffect of individual defects or of specific treatments for defect engineering onthe material.

Nevertheless, the research conducted within the scope of this thesis investi-gated various techniques for defect engineering in CZTS thin films. The am-bition of this thesis was to bring a knowledge-led approach to the challengeof defect engineering in CZTS, by connecting observable material propertieswith the current understanding and perception regarding defect formation anddefect kinetics. Raman spectroscopy was the main experimental techniqueto analyze the change in crystal quality and degree of order upon differentmethods for defect engineering. Spectrophotomery and photoluminescencegave further insights about the absorption and recombination characteristicsof the material, respectively. The investigation of composition-spread sam-ples yielded valuable insights about the interplay of fabrication conditions andphase stability as well as the influence of the cation composition on materialproperties. Based on the results, the potential of different methods for de-fect engineering in CZTS to enhance the solar cell conversion efficiency wasdiscussed.

Paper I determined the kinetic parameters for the Cu-Zn order-disordertransition in CZTS. The knowledge of the kinetic parameters allowed to pre-dict an optimized temperature profile for thermal ordering treatments and forthe first time the practical limitations for such treatments could be estimated.

66

The results reveal the slow ordering kinetics and weak driving force for or-dering in CZTS with severe implications: thermal treatments alone will notsucceed in reducing Cu-Zn disorder in CZTS to an extent where the conse-quential potential fluctuations are harmless to the open circuit voltage of solarcells, because the treatment would require years. Instead other strategies needto be explored.

Paper II represents the first report to demonstrate the influence of the cationcomposition of CZTS in a wider range on the order-disorder transition. A clearenhancement of the degree of order after thermal treatments was observed forcompositions away from the stoichiometric composition ratio of Cu/Sn= 2.Further investigations resolved that the composition does not affect the criti-cal temperature of the order-disorder transition, which is proportional to thenearest neighbor interaction energy, i. e. the driving force towards the orderedstate. It is concluded that instead the activation energy of the transition isinfluenced by the Cu/Sn ratio. The enhanced ordering kinetics suggest de-fect assisted cation exchange mechanisms in Cu-poor and Cu-rich CZTS inthe presence of vacancies or interstitials. Therefore, composition tuning isidentified as a potential tool for defect engineering in CZTS to reduce cationdisorder.

As in Paper II, the experimental approach in Papers III and V utilizedcomposition-spread CZTS thin films, i. e. samples with a continuous varia-tion of cation composition across their area. Through both papers, a methodhas been developed to identify the secondary phase boundaries and the CZTSsingle phase region by combining composition analysis from energy disper-sive X-ray spectroscopy with phase analysis based on multiwavelength Ramanspectroscopy and X-ray diffraction. This method yields the unique opportu-nity to investigate how defect formation is affected by the chemical potentialsin CZTS synthesis.

Paper III demonstrates how the solubility of vacancies in Cu-poor CZTScan be increased by raising the partial pressures of S2 and SnS during theannealing step of the fabrication. The higher density of vacancies results in afurther enhancement of the ordering kinetics in Cu-poor CZTS compared tothe observations from Paper II. A chemical model describing the solubility ofcomposition dependent defect complexes in CZTS was developed and verifiedthrough the observed results. The implications of this model indicate furtheropportunities for defect engineering in CZTS.

The effect of composition within the single phase region and the influenceof secondary phases outside of it were studied in Paper V. The positive effectof higher partial pressures of S2 and SnS on the density of copper vacancies inCu-poor CZTS leading to an enhancement in the ordering kinetics, which wasalready observed in Paper III, was further strengthened. Surprisingly, the pho-toluminescence intensity followed a different trend with CZTS compositioncompared to the ordering kinetics. The highest photoluminescence intensityand enhanced morphology were not observed within the single phase region,

67

but correlated with composition regions where SnSx secondary phases werepresent. This indicated that SnSx secondary phases are necessary for passivat-ing or eliminating detrimental defects. On the other hand, the generally lowphotoluminescence yield throughout the single phase region was worrisome,as it indicated detrimental defects in CZTS which appear to be unaffected bychanges in the chemical potential. Therefore, the results of Paper V imply thepotential of defect engineering to enhance the ordering kinetics in CZTS, butalso limitations of defect engineering through composition tuning to reducedetrimental defects leading to non-radiative recombination.

Paper IV followed a different approach to defect engineering and inves-tigated the potential of cation exchange yielding related compounds to avoidcation disorder. Exchanging one element from the compound should effect theparameters governing the order-disorder transition, i. e. the critical tempera-ture and the activation energy for ordering should be affected. Here, the struc-tural and electronic properties of the compound Cu2MnSnS4 were studied.Even though Cu2MnSnS4 crystallizes in the stannite structure, the compoundnevertheless suffers from an order-disorder transition implying that cation dis-order is not restricted to kesterite Cu2ZnSnS4. The existence of cation disor-der in Cu2MnSnS4 points towards a general problem extending to a variety ofcompounds containing atom species with similar radii.

6.2 OutlookTo my understanding after investigating the CZTS material for a little morethan five years, CZTS solar cell conversion efficiencies are currently limitedby not one but several deficiencies in the bulk CZTS material: (1) deep defectsleading to recombination, e. g. SnCu or SnZn, (2) Cu-Zn disorder and (3) mor-phological problems related to decomposition of CZTS into secondary phases.CZTS (as well as CZTSe) stands out among other solar cell absorber materi-als as very defect rich and with serious band tailing. Several candidates arediscussed to cause these detrimental band tails. Deep defects, such as SnZn, orCu-Zn disorder are the main suspects.

The effect of Cu-Zn disorder on CZTS devices remains unknown. Severalstudies reported that ordering treatments failed to reduce the open circuit volt-age deficit25,123, however there is no possible way to be sure as long as we failto produce highly ordered CZTS thin films. Personally, I am still convincedthat Cu-Zn disorder is detrimental to the solar cell performance and (at least)one of the reasons for the observed band tailing. The dramatic change of theband gap with ordering treatments makes it difficult to believe that Cu-Zn dis-order could be benign to the electrical properties of CZTS. This being said,the results from Paper I about the limitations of thermal treatments to increasethe degree of order should be an eye opener that we either have to learn to live

68

with Cu-Zn disorder in CZTS and find ways to passiviate its effects, or comeup with new creative solutions to reduce Cu-Zn disorder.

Some ideas to tackle Cu-Zn disorder are presented and discussed in thisthesis: The results from investigations on composition-spread samples sug-gest that further composition tuning together with novel annealing conditionsimplementing high S2 and SnS partial pressures can lead to an immense in-crease in the degree of order after thermal treatments. The characterizationtechniques were limited to a qualitative comparison of the degree of order. Itwould be interesting to investigate the actual level of Cu-Zn order achieved bytuned fabrication conditions.

As intriguing it may be to eliminate Cu-Zn disorder in CZTS, Paper V sug-gests that other defects limit the photoluminescence yield and drive recombi-nation especially in the single phase region. Instead the limitation seems liftedonly in the presence of SnSx secondary phases.

The photoluminescence yield of composition-spread samples in Paper Vsuggests a wide composition region that should be preferable in terms of re-combination rates, i. e. compositions where SnSx secondary phases form, inthe range of A- to E-type CZTS. However, CZTS absorbers of high efficiencysolar cells typically exhibit a narrow composition region in the (A+B)-typecomposition range103,104,102,13. This discrepancy raises uncomfortable ques-tions: Is the high PL-yield for even J- and E-type CZTS only an artifact ofcomposition-spread CZTS? Or is the device stack not optimized for composi-tions other than B- and A-type CZTS and therefore yields worse efficiency out-side this composition region? The positive effects of SnSx secondary phaseson the material properties poses further questions: Can we achieve highly ef-ficient CZTS solar cells (past 12.6 %), if secondary phases are a necessity forenhanced crystallinity and reduced non-radiative recombination? And whatrole do SnSx secondary phases play in the absorber exactly? Why is the pho-toluminescence enhanced in the region where the deep Sn-related defects areexpected? Last but not least, there is no evidence so far if the high photolu-minescence yield and high degree of Cu-Zn order in Sn-rich regions of thecomposition-spread samples correlate with an improvement in the band tail-ing.

The research on CZTS is at a turning point as the trend in the record effi-ciencies stagnates. Turning to new materials is tempting, but our results fromPaper IV show that the effect of disorder is most likely not eliminated as longas the cations chosen are similar in their sizes and the compound crystallizesin structures similar to kesterite or stannite.

In the end, my research could not deliver the one solution for an earth-abundant, non-toxic solar cell absorber material. However, I hope that myconclusions can inspire and present new pathways to improve the crystallinequality of thin film solar cell absorbers.

69

Acknowledgments

First, my thanks goes to my supervisor team for giving me the opportunityto work on this exciting project with you. To my main supervisor Jonathan:Thank you for your guidance and mentoring through these years. I will alwayscherish the engaging and vibrant scientific discussions we had. Your support-ing, kind and patient nature has made this PhD a very enjoyable journey forme. I admire your ambition at work and your strive towards high quality whileat the same time keeping a healthy work-life-balance. Thank you Lotten, foralways providing a more applied device perspective when Jonathan and I gotlost in specific material science questions, for always having an open doorwhen I had questions and for all your input on organizational issues. I couldalways count on your knowledge and experience to provide a new angle andto not loose track of the bigger picture of our research.

Thank you Marika for the nice group outings to which you invited us, andyour warm, cheerful, and energetic way to lead the solar cell group. I considermyself very lucky to have had the chance to work in a group with not one, buttwo ambitious and engaging female leaders. Thank you Lotten and Marika forbeing such great role models! The world needs more of you.

I big thank you goes to my co-authors, who have made it possible to realizethe research that led to this thesis: Alexandra, thank you for starting the workon composition-spread CZTS. You made it much easier for me to continue thework. Thank you William and Moyses for providing a theoretical perspec-tive to our experimental work on CMTS and giving us more clues about howCMTS works. Thank you Lars R. for the times in the dark TEM lab, dis-cussing CZTS compositions and secondary phases as well as life. Thank youJoakim for trusting me to supervise your master project and fighting throughall that Raman data from composition-spread samples. Thank you Luciano forsetting up the matlab code to combine the different measurement techniquesand visualize our data in an accessible way.

Thank you to Carl for taking the time to read my thesis and your valuablefeedback and thank you to Gabriella and Sven for making the Swedish sum-mary for my thesis readable and understandable.

I would also like to acknowledge the help that I got in the lab throughoutthese years: Uwe, thank you for saving Dorothea and me from a burning BAKin the clean room. Also, it was a great experience teaching with you in the labsfor the measurement techniques course. Thank you to Jes, Ray, and Sven forhelping me to fabricate devices and to Christopher for helping with electricalmeasurements of devices. Thank you to Shuyi for introducing me to the spec-trophotometer and for helping me with the measurements and the analysis.

70

Tove, thank you for helping me to set up the calibration for XRF measure-ments of CMTS. Thank you Kostya for the discussion of defect solubility andsolid solutions and possible effects of defects in CZTS.

Thank you Jörgen, för your support in questions regarding the PhD stud-ies. I am thankful to the administration staff who made everything work sosmoothly during the past years: Ramy, Ingrid, Maria Nordengren and MariaSkoglund. And to the MSL staff of the clean room, who made everything runsmoothly in the lab.

Finally, I would like to acknowledge the people around me, who havebeen an incredible source of energy and motivation throughout the past years:Thanks to all of FTE and especially the Thin Film Solar Cell group for allthe smiles and ’hej’s in our corridor, for the interesting discussions over lunchand during Friday seminars, for cheerful Christmas dinners and for fun groupoutings to Vik Slot and Oslo and for the Friday beer club of course.

Thank you for the positive spirit Adam and Jan. For kind chats Xi, Mandy,Natalia, Nina and Ngan. Thank you Nils for entrusting us with your cat fora weekend, in the end it led to us adopting Paul. Thank you Johan, Piotr,Patrice, Olivier, and Dorothea for defending before me and showing that it canbe done. Thank you Sven for defending before me via zoom in these strangetimes. And to Fredrik, Faraz, Nishant, and Corrado: all the best, fun and luckfor your remaining journey.

A special thanks goes to the coffee crew Lukas, Asta, Malkolm, Patrik andUmut. I enjoyed our regular coffee breaks. Thanks for listening to my com-plaints and for distracting me with your own troubles. Thank you Shabnam,for introducing me to the massage chairs in the basement and for the leaves onthe wall above your desk.

Thank you to the inspiring people that I met during my time as a represen-tative on various boards throughout the university: Gwenna, Anna, Susanne,Åsa, Camilla, Neil and Megha.

I am also thankful for all the new friendships that I found in Uppsala dur-ing my PhD: The V-dala Syjunta, you were my Swedish lessons every week.Tobi, Sara, Agne, Tatjana, Henry, and Shirin for the spontaneous pub evenings.The Villa-PhD group Thomas, Anna, Kristofer, Henning, Ola, Francesco andCharlotte for always making me feel as if I was one of you.

I am incredibly thankful to you Johann for your input on my thesis and foryour support, mentally and scientifically, during the past years and during thepast weeks in particular. Thank you to Paul Dirac and Simon for making thelast weeks of thesis writing in home office during Corona-virus times morechaotic but so much more fun (aka “Fluffig und Knuffig”). My thank alsogoes to family and friends who have supported me during my PhD in Uppsala,particularly my parents, Klaus and Astrid, my parents in law, Holger and Bir-git, my sister and her family Franziska, Christian and Gustav, and Christianand Luise. All your visits and interest in my work have made my PhD muchmore enjoyable and were highly appreciated.

71

References

[1] European Commission, “The European Green Deal,” 11 December 2019.[2] M. Wild, D. Folini, C. Schär, N. Loeb, E. G. Dutton, and G. König-Langlo,

“The global energy balance from a surface perspective,” Climate Dynamics,vol. 40, no. 11, pp. 3107 – 3134, 2013.

[3] D. M. Chapin, C. S. Fuller, and G. L. Pearson, “A New Silicon p-n JunctionPhotocell for Converting Solar Radiation into Electrical Power,” Journal ofApplied Physics, vol. 25, no. 5, pp. 676–677, 1954.

[4] M. A. Green, E. D. Dunlop, J. Hohl-Ebinger, M. Yoshita, N. Kopidakis, andA. W. Ho-Baillie, “Solar cell efficiency tables (Version 55),” Progress inPhotovoltaics: Research and Applications, vol. 28, no. 1, pp. 3–15, 2020.

[5] G. Masson and I. Kaizuka, “Trends in Photovoltaic Applications 2019.”International Energy Agency, Photovoltaic Power Systems Programme, 2019.

[6] K. P. Bhandari, J. M. Collier, R. J. Ellingson, and D. S. Apul, “Energy paybacktime (EPBT) and energy return on energy invested (EROI) of solarphotovoltaic systems: A systematic review and meta-analysis,” Renewable andSustainable Energy Reviews, vol. 47, pp. 133 – 141, 2015.

[7] S. A. Mann, M. J. de Wild-Scholten, V. M. Fthenakis, W. G. van Sark, andW. C. Sinke, “The energy payback time of advanced crystalline silicon PVmodules in 2020: a prospective study,” Progress in Photovoltaics: Researchand Applications, vol. 22, no. 11, pp. 1180–1194, 2014.

[8] M. A. Green, “Photovoltaic technology and visions for the future,” Progress inEnergy, vol. 1, p. 013001, jul 2019.

[9] M. Powalla, S. Paetel, D. Hariskos, R. Wuerz, F. Kessler, P. Lechner,W. Wischmann, and T. M. Friedlmeier, “Advances in Cost-Efficient Thin-FilmPhotovoltaics Based on Cu(In,Ga)Se2,” Engineering, vol. 3, no. 4, pp. 445 –451, 2017.

[10] T. Unold and H. Schock, “Nonconventional (Non-Silicon-Based) PhotovoltaicMaterials,” Annual Review of Materials Research, vol. 41, no. 1, pp. 297–321,2011.

[11] R. Haight, A. Barkhouse, O. Gunawan, B. Shin, M. Copel, M. Hopstaken, andD. B. Mitzi, “Band alignment at the Cu2ZnSn(SxSe1−x)4/CdS interface,”Applied Physics Letters, vol. 98, no. 25, p. 253502, 2011.

[12] L. Q. Phuong, M. Okano, G. Yamashita, M. Nagai, M. Ashida, A. Nagaoka,K. Yoshino, and Y. Kanemitsu, “Free-carrier dynamics and band tails inCu2ZnSn(SxSe1−x)4: Evaluation of factors determining solar cell efficiency,”Phys. Rev. B, vol. 92, p. 115204, Sep 2015.

[13] J. K. Larsen, F. Larsson, T. Törndahl, N. Saini, L. Riekehr, Y. Ren, A. Biswal,D. Hauschild, L. Weinhardt, C. Heske, and C. Platzer-Björkman, “CadmiumFree Cu2ZnSnS4 Solar Cells with 9.7% Efficiency,” Advanced EnergyMaterials, vol. 9, no. 21, p. 1900439, 2019.

72

[14] W. Wang, M. T. Winkler, O. Gunawan, T. Gokmen, T. K. Todorov, Y. Zhu, andD. B. Mitzi, “Device Characteristics of CZTSSe Thin-Film Solar Cells with12.6% Efficiency,” Advanced Energy Materials, vol. 4, no. 7, p. 1301465,2014.

[15] T. Ericson, J. J. Scragg, A. Hultqvist, J. T. Wätjen, P. Szaniawski, T. Törndahl,and C. Platzer-Björkman, “Zn(O, S) Buffer Layers and Thickness Variations ofCdS Buffer for Cu2ZnSnS4 Solar Cells,” IEEE Journal of Photovoltaics,vol. 4, pp. 465–469, Jan 2014.

[16] T. Ericson, F. Larsson, T. Törndahl, C. Frisk, J. Larsen, V. Kosyak,C. Hägglund, S. Li, and C. Platzer-Björkman, “Zinc-Tin-Oxide Buffer Layerand Low Temperature Post Annealing Resulting in a 9.0% Efficient Cd-FreeCu2ZnSnS4 Solar Cell,” Solar RRL, vol. 1, no. 5, p. 1700001, 2017.

[17] S. M. Sze, Semiconductor Devices: Physics and Technology. John Wiley &Sons, 2nd ed., 2002.

[18] W. Shockley and H. J. Queisser, “Detailed Balance Limit of Efficiency of p-nJunction Solar Cells,” Journal of Applied Physics, vol. 32, no. 3, pp. 510–519,1961.

[19] D.-H. Son, S.-H. Kim, S.-Y. Kim, Y.-I. Kim, J.-H. Sim, S.-N. Park, D.-H. Jeon,D.-K. Hwang, S.-J. Sung, J.-K. Kang, K.-J. Yang, and D.-H. Kim, “Effect ofsolid-H2S gas reactions on CZTSSe thin film growth and photovoltaicproperties of a 12.62% efficiency device,” J. Mater. Chem. A, vol. 7,pp. 25279–25289, 2019.

[20] J. J. Scragg, T. Kubart, J. T. Wätjen, T. Ericson, M. K. Linnarsson, andC. Platzer-Björkman, “Effects of Back Contact Instability on Cu2ZnSnS4Devices and Processes,” Chem. Mater., vol. 25, no. 15, pp. 3162–3171, 2013.

[21] F. Zhou, F. Zeng, X. Liu, F. Liu, N. Song, C. Yan, A. Pu, J. Park, K. Sun, andX. Hao, “Improvement of Jsc in a Cu2ZnSnS4 Solar Cell by Using a ThinCarbon Intermediate Layer at the Cu2ZnSnS4/Mo Interface,” ACS AppliedMaterials & Interfaces, vol. 7, no. 41, pp. 22868–22873, 2015.

[22] M. Bär, B.-A. Schubert, B. Marsen, R. G. Wilks, S. Pookpanratana, M. Blum,S. Krause, T. Unold, W. Yang, L. Weinhardt, C. Heske, and H.-W. Schock,“Cliff-like conduction band offset and KCN-induced recombination barrierenhancement at the CdS/Cu2ZnSnS4 thin-film solar cell heterojunction,”Applied Physics Letters, vol. 99, no. 22, p. 222105, 2011.

[23] A. Redinger, J. Sendler, R. Djemour, T. P. Weiss, G. Rey, P. J. Dale, andS. Siebentritt, “Different Bandgaps in Cu2ZnSnSe4: A High TemperatureCoevaporation Study,” IEEE J. Photovolt., vol. 5, pp. 641–648, March 2015.

[24] S. Siebentritt, G. Rey, A. Finger, D. Regesch, J. Sendler, T. P. Weiss, andT. Bertram, “What is the bandgap of kesterite?,” Solar Energy Materials andSolar Cells, vol. 158, pp. 126 – 129, 2016. Nanotechnology for NextGeneration High Efficiency Photovoltaics: NEXTGEN NANOPV SpringInternational School & Workshop.

[25] G. Rey, T. Weiss, J. Sendler, A. Finger, C. Spindler, F. Werner, M. Melchiorre,M. Hála, M. Guennou, and S. Siebentritt, “Ordering kesterite improves solarcells: A low temperature post-deposition annealing study,” Solar EnergyMaterials and Solar Cells, vol. 151, pp. 131 – 138, 2016.

73

[26] D. Shin, B. Saparov, and D. B. Mitzi, “Defect Engineering in MultinaryEarth-Abundant Chalcogenide Photovoltaic Materials,” Advanced EnergyMaterials, p. 1602366, 2017.

[27] Y. E. Romanyuk, S. G. Haass, S. Giraldo, M. Placidi, D. Tiwari, D. J. Fermin,X. Hao, H. Xin, T. Schnabel, M. Kauk-Kuusik, P. Pistor, S. Lie, and L. H.Wong, “Doping and alloying of kesterites,” Journal of Physics: Energy, vol. 1,p. 044004, aug 2019.

[28] S. Schorr, “The crystal structure of kesterite type compounds: A neutron andX-ray diffraction study,” Sol. Energy Mater. Sol. Cells, vol. 95, no. 6, pp. 1482– 1488, 2011. Special Issue : Thin film and nanostructured solar cells.

[29] A. Lafond, L. Choubrac, C. Guillot-Deudon, P. Deniard, and S. Jobic, “CrystalStructures of Photovoltaic Chalcogenides, an Intricate Puzzle to Solve: theCases of CIGSe and CZTS Materials,” Zeitschrift für anorganische undallgemeine Chemie, vol. 638, no. 15, pp. 2571–2577, 2012.

[30] S. Chen, A. Walsh, X.-G. Gong, and S.-H. Wei, “Classification of LatticeDefects in the Kesterite Cu2ZnSnS4 and Cu2ZnSnSe4 Earth-Abundant SolarCell Absorbers,” Advanced Materials, vol. 25, pp. 1522–1539, 3 2013.

[31] J. J. S. Scragg, L. Choubrac, A. Lafond, T. Ericson, and C. Platzer-Björkman,“A low-temperature order-disorder transition in Cu2ZnSnS4 thin films,” Appl.Phys. Lett., vol. 104, no. 4, p. 041911, 2014.

[32] D. C. Palmer, “CrystalMaker,” 2014. CrystalMaker Software Ltd, Begbroke,Oxfordshire, England.

[33] S. Chen, X. G. Gong, A. Walsh, and S.-H. Wei, “Crystal and electronic bandstructure of Cu2ZnSnX4 (X=S and Se) photovoltaic absorbers: First-principlesinsights,” Appl. Phys. Lett., vol. 94, no. 4, 2009.

[34] S. P. Ramkumar, Y. Gillet, A. Miglio, M. J. van Setten, X. Gonze, and G.-M.Rignanese, “First-principles investigation of the structural, dynamical, anddielectric properties of kesterite, stannite, and PMCA phases of cu2znsns4,”Phys. Rev. B, vol. 94, p. 224302, Dec 2016.

[35] S. Schorr, H.-J. Hoebler, and M. Tovar, “A neutron diffraction study of thestannite-kesterite solid solution series,” European Journal of Mineralogy,vol. 19, pp. 65–73, 02 2007.

[36] M. Dimitrievska, F. Boero, A. P. Litvinchuk, S. Delsante, G. Borzone,A. Pérez-Rodríguez, and V. Izquierdo-Roca, “Structural Polymorphism in“Kesterite” Cu2ZnSnS4: Raman Spectroscopy and First-PrinciplesCalculations Analysis,” Inorganic Chemistry, vol. 56, no. 6, pp. 3467–3474,2017.

[37] E. A. Pogue, M. Goetter, and A. Rockett, “Reaction kinetics of Cu2−xS, ZnS,and SnS2 to form Cu2ZnSnS4 and Cu2SnS3 studied using differential scanningcalorimetry,” MRS Advances, vol. 2, no. 53, pp. 3181–3186, 2017.

[38] T. Özdal and H. Kavak, “Determination of crystallization thresholdtemperature for sol-gel spin coated Cu2ZnSnS4 thin films,” CeramicsInternational, vol. 44, no. 15, pp. 18928 – 18934, 2018.

[39] A. Weber, R. Mainz, and H. W. Schock, “On the Sn loss from thin films of thematerial system Cu-Zn-Sn-S in high vacuum,” Journal of Applied Physics,vol. 107, no. 1, p. 013516, 2010.

74

[40] I. Repins, C. Beall, N. Vora, C. DeHart, D. Kuciauskas, P. Dippo, B. To,J. Mann, W.-C. Hsu, A. Goodrich, and R. Noufi, “Co-evaporated Cu2ZnSnSe4films and devices,” Solar Energy Materials and Solar Cells, vol. 101, pp. 154 –159, 2012.

[41] B.-A. Schubert, B. Marsen, S. Cinque, T. Unold, R. Klenk, S. Schorr, andH.-W. Schock, “Cu2ZnSnS4 thin film solar cells by fast coevaporation,”Progress in Photovoltaics: Research and Applications, vol. 19, no. 1,pp. 93–96, 2011.

[42] B.-T. Jheng, P.-T. Liu, and M.-C. Wu, “A promising sputtering route for denseCu2ZnSnS4 absorber films and their photovoltaic performance,” Solar EnergyMaterials and Solar Cells, vol. 128, pp. 275 – 282, 2014.

[43] S. Englund, N. Saini, and C. Platzer-Björkman, “Cu2ZnSn(S,Se)4 fromannealing of compound co-sputtered precursors – Recent results and openquestions,” Solar Energy, vol. 175, pp. 84 – 93, 2018. Recent Progress inPhotovoltaics, Part 1.

[44] K.-J. Yang, D.-H. Son, S.-J. Sung, J.-H. Sim, Y.-I. Kim, S.-N. Park, D.-H.Jeon, J. Kim, D.-K. Hwang, C.-W. Jeon, D. Nam, H. Cheong, J.-K. Kang, andD.-H. Kim, “A band-gap-graded CZTSSe solar cell with 12.3% efficiency,” J.Mater. Chem. A, vol. 4, pp. 10151–10158, 2016.

[45] H. Sugimoto, H. Hiroi, N. Sakai, S. Muraoka, and T. Katou, “Over 8%efficiency Cu2ZnSnS4 submodules with ultra-thin absorber,” in 2012 38thIEEE Photovoltaic Specialists Conference, pp. 002997–003000, June 2012.

[46] J. Kim, H. Hiroi, T. K. Todorov, O. Gunawan, M. Kuwahara, T. Gokmen,D. Nair, M. Hopstaken, B. Shin, Y. S. Lee, W. Wang, H. Sugimoto, and D. B.Mitzi, “High Efficiency Cu2ZnSn(S,Se)4 Solar Cells by Applying a DoubleIn2S3/CdS Emitter,” Advanced Materials, vol. 26, no. 44, pp. 7427–7431,2014.

[47] W. M. Hlaing Oo, W.M., J. Johnson, A. Bhatia, E. Lund, M. Nowell, andM. Scarpulla, “Grain Size and Texture of Cu2ZnSnS4 Thin Films Synthesizedby Cosputtering Binary Sulfides and Annealing: Effects of ProcessingConditions and Sodium,” Journal of Electric Materials, vol. 40, no. 11,p. 2214, 2011.

[48] J. J. Scragg, T. Ericson, X. Fontané, V. Izquierdo-Roca, A. Pérez-Rodríguez,T. Kubart, M. Edoff, and C. Platzer-Björkman, “Rapid annealing of reactivelysputtered precursors for Cu2ZnSnS4 solar cells,” Progress in Photovoltaics:Research and Applications, vol. 22, no. 1, pp. 10–17, 2014.

[49] T. Ericson, “Reactive sputtering and composition measurements of precursorsfor Cu2ZnSnS4 thin film solar cells.” Uppsala University, 2013. Licenciatethesis.

[50] D. E. Newbury and N. W. M. Ritchie, “Is Scanning ElectronMicroscopy/Energy Dispersive X-ray Spectrometry (SEM/EDS)Quantitative?,” Scanning, vol. 35, no. 3, pp. 141–168, 2013.

[51] J. J. Scragg, T. Ericson, T. Kubart, M. Edoff, and C. Platzer-Björkman,“Chemical Insights into the Instability of Cu2ZnSnS4 Films duringAnnealing,” Chemistry of Materials, vol. 23, no. 20, pp. 4625–4633, 2011.

[52] Y. Ren, N. Ross, J. K. Larsen, K. Rudisch, J. J. S. Scragg, andC. Platzer-Björkman, “Evolution of Cu2ZnSnS4 during Non-Equilibrium

75

Annealing with Quasi-in Situ Monitoring of Sulfur Partial Pressure,”Chemistry of Materials, vol. 29, no. 8, pp. 3713–3722, 2017.

[53] M. Buffière, G. Brammertz, S. Sahayaraj, M. Batuk, S. Khelifi, D. Mangin,A.-A. El Mel, L. Arzel, J. Hadermann, M. Meuris, and J. Poortmans, “KCNChemical Etch for Interface Engineering in Cu2ZnSnSe4 Solar Cells,” ACSApplied Materials & Interfaces, vol. 7, no. 27, pp. 14690–14698, 2015.

[54] I. Olekseyuk, I. Dudchak, and L. Piskach, “Phase equilibria in theCu2S-ZnS-SnS2 system,” Journal of Alloys and Compounds, vol. 368, no. 1,pp. 135 – 143, 2004.

[55] M. Kumar, A. Dubey, N. Adhikari, S. Venkatesan, and Q. Qiao, “Strategicreview of secondary phases, defects and defect-complexes in kesteriteCZTS-Se solar cells,” Energy Environ. Sci., vol. 8, pp. 3134–3159, 2015.

[56] A. Nagoya, R. Asahi, R. Wahl, and G. Kresse, “Defect formation and phasestability of Cu2ZnSnS4 photovoltaic material,” Phys. Rev. B, vol. 81,p. 113202, Mar 2010.

[57] D. Han, Y. Y. Sun, J. Bang, Y. Y. Zhang, H.-B. Sun, X.-B. Li, and S. B. Zhang,“Deep electron traps and origin of p-type conductivity in the earth-abundantsolar-cell material Cu2ZnSnS4,” Phys. Rev. B, vol. 87, p. 155206, Apr 2013.

[58] P. Schöppe, G. Gurieva, S. Giraldo, G. Martínez-Criado, C. Ronning,E. Saucedo, S. Schorr, and C. S. Schnohr, “Discrepancy between integral andlocal composition in off-stoichiometric Cu2ZnSnSe4 kesterites: A pitfall forclassification,” Applied Physics Letters, vol. 110, no. 4, p. 043901, 2017.

[59] A. Redinger, D. M. Berg, P. J. Dale, and S. Siebentritt, “The Consequences ofKesterite Equilibria for Efficient Solar Cells,” Journal of the AmericanChemical Society, vol. 133, no. 10, pp. 3320–3323, 2011.

[60] C.-W. Jeon, T. Cheon, H. Kim, M.-S. Kwon, and S.-H. Kim, “Controlledformation of MoSe2 by MoNx thin film as a diffusion barrier against Se duringselenization annealing for CIGS solar cell,” Journal of Alloys and Compounds,vol. 644, pp. 317 – 323, 2015.

[61] J. J. Scragg, J. T. Wätjen, M. Edoff, T. Ericson, T. Kubart, andC. Platzer-Björkman, “A Detrimental Reaction at the Molybdenum BackContact in Cu2ZnSn(S,Se)4 Thin-Film Solar Cells,” Journal of the AmericanChemical Society, vol. 134, no. 47, pp. 19330–19333, 2012.

[62] L. E. V. Rios, K. Neldner, G. Gurieva, and S. Schorr, “Existence ofoff-stoichiometric single phase kesterite,” Journal of Alloys and Compounds,vol. 657, pp. 408 – 413, 2016.

[63] G. Gurieva, L. E. Valle Rios, A. Franz, P. Whitfield, and S. Schorr, “Intrinsicpoint defects in off-stoichiometric Cu2ZnSnSe4: A neutron diffraction study,”Journal of Applied Physics, vol. 123, no. 16, p. 161519, 2018.

[64] E. A. Lund, H. Du, W. M. Hlaing OO, G. Teeter, and M. A. Scarpulla,“Investigation of combinatorial coevaporated thin film Cu2ZnSnS4 (II):Beneficial cation arrangement in Cu-rich growth,” Journal of Applied Physics,vol. 115, no. 17, p. 173503, 2014.

[65] P. Fernandes, P. Salomé, and A. da Cunha, “Study of polycrystallineCu2ZnSnS4 films by Raman scattering,” Journal of Alloys and Compounds,vol. 509, no. 28, pp. 7600 – 7606, 2011.

76

[66] X. Fontané, L. Calvo-Barrio, V. Izquierdo-Roca, E. Saucedo,A. Pérez-Rodríguez, J. R. Morante, D. M. Berg, P. J. Dale, and S. Siebentritt,“In-depth resolved Raman scattering analysis for the identification ofsecondary phases: Characterization of Cu2ZnSnS4 layers for solar cellapplications,” Applied Physics Letters, vol. 98, no. 18, p. 181905, 2011.

[67] M. Paris, L. Choubrac, A. Lafond, C. Guillot-Deudon, and S. Jobic,“Solid-State NMR and Raman Spectroscopy To Address the Local Structure ofDefects and the Tricky Issue of the Cu/Zn Disorder in Cu-Poor, Zn-Rich CZTSMaterials,” Inorg. Chem., vol. 53, no. 16, pp. 8646–8653, 2014.

[68] C. Persson, “Electronic and optical properties of Cu2ZnSnS4 andCu2ZnSnSe4,” Journal of Applied Physics, vol. 107, no. 5, p. 053710, 2010.

[69] J. J. Scragg, P. J. Dale, and L. M. Peter, “Towards sustainable materials forsolar energy conversion: Preparation and photoelectrochemicalcharacterization of Cu2ZnSnS4,” Electrochemistry Communications, vol. 10,no. 4, pp. 639 – 642, 2008.

[70] H. Katagiri, K. Jimbo, M. Tahara, H. Araki, and K. Oishi, “The Influence ofthe Composition Ratio on CZTS-based Thin Film Solar Cells,” vol. 1165 ofMRS Proceedings, 2009.

[71] A. Crovetto, A. Cazzaniga, R. B. Ettlinger, J. Schou, and O. Hansen, “Opticalproperties and surface characterization of pulsed laser-deposited Cu2ZnSnS4by spectroscopic ellipsometry,” Thin Solid Films, vol. 582, pp. 203 – 207,2015. E-MRS 2014 Spring Meeting, Symposium A, Thin-Film ChalcogenidePhotovoltaic Materials.

[72] S.-Y. Li, C. Hägglund, Y. Ren, J. J. Scragg, J. K. Larsen, C. Frisk, K. Rudisch,S. Englund, and C. Platzer-Björkman, “Optical properties of reactivelysputtered Cu2ZnSnS4 solar absorbers determined by spectroscopicellipsometry and spectrophotometry,” Solar Energy Materials and Solar Cells,vol. 149, pp. 170 – 178, 2016.

[73] T. P. Dhakal, C.-Y. Peng, R. R. Tobias, R. Dasharathy, and C. R. Westgate,“Characterization of a CZTS thin film solar cell grown by sputtering method,”Solar Energy, vol. 100, pp. 23 – 30, 2014.

[74] T. M. Ng, M. T. Weller, G. P. Kissling, L. M. Peter, P. Dale, F. Babbe,J. de Wild, B. Wenger, H. J. Snaith, and D. Lane, “Optoelectronic andspectroscopic characterization of vapour-transport grown Cu2ZnSnS4 singlecrystals,” J. Mater. Chem. A, vol. 5, pp. 1192–1200, 2017.

[75] S. Levcenco, D. Dumcenco, Y. Wang, Y. Huang, C. Ho, E. Arushanov,V. Tezlevan, and K. Tiong, “Influence of anionic substitution on the electrolyteelectroreflectance study of band edge transitions in single crystalCu2ZnSn(SxSe1−x)4 solid solutions,” Optical Materials, vol. 34, no. 8,pp. 1362 – 1365, 2012.

[76] F. Urbach, “The Long-Wavelength Edge of Photographic Sensitivity and of theElectronic Absorption of Solids,” Phys. Rev., vol. 92, pp. 1324–1324, Dec1953.

[77] G. Rey, A. Redinger, J. Sendler, T. P. Weiss, M. Thevenin, M. Guennou,B. El Adib, and S. Siebentritt, “The band gap of Cu2ZnSnSe4: Effect oforder-disorder,” Appl. Phys. Lett., vol. 105, no. 11, p. 112106, 2014.

77

[78] C. Krämmer, C. Huber, C. Zimmermann, M. Lang, T. Schnabel, T. Abzieher,E. Ahlswede, H. Kalt, and M. Hetterich, “Reversible order-disorder relatedband gap changes in Cu2ZnSn(S,Se)4 via post-annealing of solar cellsmeasured by electroreflectance,” Applied Physics Letters, vol. 105, no. 26,p. 262104, 2014.

[79] C. Malerba, F. Biccari, C. L. A. Ricardo, M. Valentini, R. Chierchia,M. Müller, A. Santoni, E. Esposito, P. Mangiapane, P. Scardi, and A. Mittiga,“CZTS stoichiometry effects on the band gap energy,” Journal of Alloys andCompounds, vol. 582, pp. 528 – 534, 2014.

[80] G. Rey, G. Larramona, S. Bourdais, C. Choné, B. Delatouche, A. Jacob,G. Dennler, and S. Siebentritt, “On the origin of band-tails in kesterite,” SolarEnergy Materials and Solar Cells, vol. 179, pp. 142 – 151, 2018.

[81] T. Gokmen, O. Gunawan, T. K. Todorov, and D. B. Mitzi, “Band tailing andefficiency limitation in kesterite solar cells,” Appl. Phys. Lett., vol. 103, no. 10,2013.

[82] J. E. Moore, C. J. Hages, R. Agrawal, M. S. Lundstrom, and J. L. Gray, “Theimportance of band tail recombination on current collection and open-circuitvoltage in CZTSSe solar cells,” Applied Physics Letters, vol. 109, no. 2,p. 021102, 2016.

[83] J. Mattheis, U. Rau, and J. H. Werner, “Light absorption and emission insemiconductors with band gap fluctuations – A study on Cu(In,Ga)Se2 thinfilms,” Journal of Applied Physics, vol. 101, no. 11, p. 113519, 2007.

[84] J. H. Werner, J. Mattheis, and U. Rau, “Efficiency limitations ofpolycrystalline thin film solar cells: case of Cu(In,Ga)Se2,” Thin Solid Films,vol. 480 - 481, pp. 399 – 409, 2005.

[85] M. Nishiwaki, K. Nagaya, M. Kato, S. Fujimoto, H. Tampo, T. Miyadera,M. Chikamatsu, H. Shibata, and H. Fujiwara, “Tail state formation in solar cellmaterials: First principles analyses of zincblende, chalcopyrite, kesterite, andhybrid perovskite crystals,” Phys. Rev. Materials, vol. 2, p. 085404, Aug 2018.

[86] B. G. Mendis, M. D. Shannon, M. C. J. Goodman, J. D. Major, A. A. Taylor,D. P. Halliday, and K. Durose, “The nature of electrostatic potentialfluctuations in Cu2ZnSnS4 and their role on photovoltaic device performance,”Journal of Physics: Conference Series, vol. 471, p. 012014, nov 2013.

[87] P. Zawadzki, A. Zakutayev, and S. Lany, “Entropy-Driven Clustering inTetrahedrally Bonded Multinary Materials,” Phys. Rev. Applied, vol. 3,p. 034007, Mar 2015.

[88] S. Chen, J.-H. Yang, X. G. Gong, A. Walsh, and S.-H. Wei, “Intrinsic pointdefects and complexes in the quaternary kesterite semiconductor Cu2ZnSnS4,”Phys. Rev. B, vol. 81, p. 245204, Jun 2010.

[89] H. Katagiri, K. Saitoh, T. Washio, H. Shinohara, T. Kurumadani, andS. Miyajima, “Development of thin film solar cell based on Cu2ZnSnS4 thinfilms,” Solar Energy Materials and Solar Cells, vol. 65, no. 1, pp. 141 – 148,2001. PVSEC 11 Part I.

[90] J. J. Scragg, P. J. Dale, L. M. Peter, G. Zoppi, and I. Forbes, “New routes tosustainable photovoltaics: evaluation of Cu2ZnSnS4 as an alternative absorbermaterial,” physica status solidi (b), vol. 245, no. 9, pp. 1772–1778, 2008.

78

[91] T. Maeda, A. Kawabata, and T. Wada, “First-principles study on alkali-metaleffect of Li, Na, and K in Cu2ZnSnS4 and Cu2ZnSnSe4,” physica status solidi(c), vol. 12, pp. 631–637, 6 2015.

[92] T. Gershon, B. Shin, N. Bojarczuk, M. Hopstaken, D. B. Mitzi, and S. Guha,“The Role of Sodium as a Surfactant and Suppressor of Non-RadiativeRecombination at Internal Surfaces in Cu2ZnSnS4,” Advanced EnergyMaterials, vol. 5, no. 2, p. 1400849, 2015.

[93] C.-Y. Liu, Z.-M. Li, H.-Y. Gu, S.-Y. Chen, H. Xiang, and X.-G. Gong,“Sodium Passivation of the Grain Boundaries in CuInSe2 and Cu2ZnSnS4 forHigh-Efficiency Solar Cells,” Advanced Energy Materials, vol. 7, no. 8,p. 1601457, 2017.

[94] S. G. Haass, C. Andres, R. Figi, C. Schreiner, M. Bürki, Y. E. Romanyuk, andA. N. Tiwari, “Complex Interplay between Absorber Composition and AlkaliDoping in High-Efficiency Kesterite Solar Cells,” Advanced Energy Materials,vol. 8, no. 4, p. 1701760, 2018.

[95] J. Park, J. Huang, J. Yun, F. Liu, Z. Ouyang, H. Sun, C. Yan, K. Sun, K. Kim,J. Seidel, S. Chen, M. A. Green, and X. Hao, “The Role of Hydrogen fromALD-Al2O3 in Kesterite Cu2ZnSnS4 Solar Cells: Grain Surface Passivation,”Advanced Energy Materials, vol. 8, no. 23, p. 1701940, 2018.

[96] A. D. Collord and H. W. Hillhouse, “Germanium Alloyed Kesterite Solar Cellswith Low Voltage Deficits,” Chemistry of Materials, vol. 28, no. 7,pp. 2067–2073, 2016.

[97] D. Huang and C. Persson, “Band gap change induced by defect complexes inCu2ZnSnS4,” Thin Solid Films, vol. 535, pp. 265 – 269, 2013.

[98] A. Ritscher, M. Hoelzel, and M. Lerch, “The order-disorder transition inCu2ZnSnS4 – A neutron scattering investigation,” Journal of Solid StateChemistry, vol. 238, pp. 68 – 73, 2016.

[99] A. Ritscher, A. Franz, S. Schorr, and M. Lerch, “Off-stoichiometric CZTS:Neutron scattering investigations on mechanochemically synthesizedpowders,” Journal of Alloys and Compounds, vol. 689, no. Supplement C,pp. 271 – 277, 2016.

[100] G. Gurieva, M. Dimitrievska, S. Zander, A. Pérez-Rodríguez,V. Izquierdo-Roca, and S. Schorr, “Structural characterisation ofCu2.04Zn0.91Sn1.05S2.08Se1.92,” phys. status solidi (c), vol. 12, no. 6,pp. 588–591, 2015.

[101] R. Gunder, J. A. Márquez-Prieto, G. Gurieva, T. Unold, and S. Schorr,“Structural characterization of off-stoichiometric kesterite-type Cu2ZnGeSe4compound semiconductors: from cation distribution to intrinsic point defectdensity,” CrystEngComm, vol. 20, pp. 1491–1498, 2018.

[102] A. Fairbrother, M. Dimitrievska, Y. Sanchez, V. Izquierdo-Roca,A. Pérez-Rodríguez, and E. Saucedo, “Compositional paradigms in multinarycompound systems for photovoltaic applications: a case study of kesterites,” J.Mater. Chem. A, vol. 3, pp. 9451–9455, 2015.

[103] M. Valentini, C. Malerba, E. Salza, M. T. D. Luca, M. Capizzi, and A. Mittiga,“Combinatorial study of co-sputtered Cu2ZnSnS4 thin-film stoichiometry forphotovoltaic devices,” 2014 IEEE 40th Photovoltaic Specialist Conference(PVSC), pp. 0439–0442, 2014.

79

[104] M. Dimitrievska, A. Fairbrother, E. Saucedo, A. Pérez-Rodríguez, andV. Izquierdo-Roca, “Secondary phase and Cu substitutional defect dynamics inkesterite solar cells: Impact on optoelectronic properties,” Solar EnergyMaterials and Solar Cells, vol. 149, pp. 304 – 309, 2016.

[105] S. Kim, J.-S. Park, and A. Walsh, “Identification of Killer Defects in KesteriteThin-Film Solar Cells,” ACS Energy Letters, vol. 3, no. 2, pp. 496–500, 2018.

[106] O. D. Miller, E. Yablonovitch, and S. R. Kurtz, “Strong Internal and ExternalLuminescence as Solar Cells Approach the Shockley-Queisser Limit,” IEEEJournal of Photovoltaics, vol. 2, pp. 303–311, July 2012.

[107] P. Bais, M. T. Caldes, M. Paris, C. Guillot-Deudon, P. Fertey, B. Domengès,and A. Lafond, “Cationic and Anionic Disorder in CZTSSe KesteriteCompounds: A Chemical Crystallography Study,” Inorganic Chemistry,vol. 56, no. 19, pp. 11779–11786, 2017.

[108] R. D. Shannon, “Revised effective ionic radii and systematic studies ofinteratomic distances in halides and chalcogenides,” Acta CrystallographicaSection A, vol. 32, pp. 751–767, Sep 1976.

[109] C. J. Bosson, M. T. Birch, D. P. Halliday, K. S. Knight, A. S. Gibbs, and P. D.Hatton, “Cation disorder and phase transitions in the structurally complex solarcell material Cu2ZnSnS4,” J. Mater. Chem. A, vol. 5, pp. 16672–16680, 2017.

[110] Y.-F. Zheng, J.-H. Yang, and X.-G. Gong, “Cu-Zn disorder in stoichiometricand non-stoichiometric Cu2ZnSnS4/Cu2ZnSnSe4,” AIP Advances, vol. 9,no. 3, p. 035248, 2019.

[111] A. Lafond, L. Choubrac, C. Guillot-Deudon, P. Fertey, M. Evain, and S. Jobic,“X-ray resonant single-crystal diffraction technique, a powerful tool toinvestigate the kesterite structure of the photovoltaic Cu2ZnSnS4 compound,”Acta Crystallogr. B, vol. 70, pp. 390–394, Apr 2014.

[112] T. Washio, H. Nozaki, T. Fukano, T. Motohiro, K. Jimbo, and H. Katagiri,“Analysis of lattice site occupancy in kesterite structure of Cu2ZnSnS4 filmsusing synchrotron radiation x-ray diffraction,” Journal of Applied Physics,vol. 110, no. 7, p. 074511, 2011.

[113] D. M. Többens, R. Gunder, G. Gurieva, J. Marquardt, K. Neldner, L. E.Valle-Rios, S. Zander, and S. Schorr, “Quantitative anomalous powderdiffraction analysis of cation disorder in kesterite semiconductors,” PowderDiffraction, vol. 31, no. 3, pp. 168 – 175, 2016.

[114] N. A. Kattan, I. J. Griffiths, D. Cherns, and D. J. Fermin, “Observation ofantisite domain boundaries in Cu2ZnSnS4 by atomic-resolution transmissionelectron microscopy,” Nanoscale, vol. 8, pp. 14369–14373, 2016.

[115] B. G. Mendis, M. D. Shannon, M. C. Goodman, J. D. Major, R. Claridge, D. P.Halliday, and K. Durose, “Direct observation of Cu, Zn cation disorder inCu2ZnSnS4 solar cell absorber material using aberration corrected scanningtransmission electron microscopy,” Progress in Photovoltaics: Research andApplications, vol. 22, no. 1, pp. 24–34, 2014.

[116] W. L. Bragg and E. J. Williams, “The Effect of Thermal Agitation on AtomicArrangement in Alloys,” Proc. Roy. Soc. A, vol. 145, no. 855, pp. 699–730,1934.

[117] J. J. S. Scragg, J. K. Larsen, M. Kumar, C. Persson, J. Sendler, S. Siebentritt,and C. Platzer Björkman, “Cu-Zn disorder and band gap fluctuations in

80

Cu2ZnSn(S,Se)4: Theoretical and experimental investigations,” phys. statussolidi (b), vol. 253, no. 2, pp. 247–254, 2016.

[118] G. H. Vineyard, “Theory of Order-Disorder Kinetics,” Phys. Rev., vol. 102,pp. 981–992, May 1956.

[119] M. Valentini, C. Malerba, F. Menchini, D. Tedeschi, A. Polimeni, M. Capizzi,and A. Mittiga, “Effect of the order-disorder transition on the optical propertiesof Cu2ZnSnS4,” Applied Physics Letters, vol. 108, no. 21, p. 211909, 2016.

[120] R. Chen and C. Persson, “Electronic and optical properties of Cu2XSnS4 (X =Be, Mg, Ca, Mn, Fe, and Ni) and the impact of native defect pairs,” Journal ofApplied Physics, vol. 121, no. 20, p. 203104, 2017.

[121] M. Lang, T. Renz, N. Mathes, M. Neuwirth, T. Schnabel, H. Kalt, andM. Hetterich, “Influence of the Cu Content in Cu2ZnSn(S,Se)4 solar cellabsorbers on order-disorder related band gap changes,” Applied PhysicsLetters, vol. 109, no. 14, p. 142103, 2016.

[122] M. Dimitrievska, A. Fairbrother, A. P. rez Rodrí guez, E. Saucedo, andV. Izquierdo-Roca, “Raman scattering crystalline assessment of polycrystallineCu2ZnSnS4 thin films for sustainable photovoltaic technologies: Phononconfinement model,” Acta Materialia, vol. 70, pp. 272 – 280, 2014.

[123] C. Krämmer, C. Huber, T. Schnabel, C. Zimmermann, M. Lang, E. Ahlswede,H. Kalt, and M. Hetterich, “Order-disorder related band gap changes inCu2ZnSn(S,Se)4: Impact on solar cell performance,” in PVSC 2015 IEEE42nd, pp. 1–4, June 2015.

[124] S. Bourdais, C. Choné, B. Delatouche, A. Jacob, G. Larramona, C. Moisan,A. Lafond, F. Donatini, G. Rey, S. Siebentritt, A. Walsh, and G. Dennler, “Isthe Cu/Zn Disorder the Main Culprit for the Voltage Deficit in Kesterite SolarCells?,” Advanced Energy Materials, vol. 6, no. 12, p. 1502276, 2016.

[125] D. M. Bishop, B. McCandless, T. Gershon, M. A. Lloyd, R. Haight, andR. Birkmire, “Modification of defects and potential fluctuations in slow-cooledand quenched Cu2ZnSnSe4 single crystals,” Journal of Applied Physics,vol. 121, no. 6, p. 065704, 2017.

[126] W. Bergholz, Defects and Impurities in Silicon Materials. Lecture Notes inPhysics, vol. 916, ch. Defect Engineering in Silicon Materials. Springer,Tokyo, 2015.

[127] A. J. Atanacio, T. Bak, K. A. Rahman, and J. Nowotny, “Chapter One - DefectEngineering of Photosensitive Oxide Materials. Example of TiO2 SolidSolutions,” in Materials for Sustainable Energy (R. van Eldik and W. Macyk,eds.), vol. 72 of Advances in Inorganic Chemistry, pp. 1 – 47, Academic Press,2018.

[128] C. Freysoldt, B. Grabowski, T. Hickel, J. Neugebauer, G. Kresse, A. Janotti,and C. G. Van de Walle, “First-principles calculations for point defects insolids,” Rev. Mod. Phys., vol. 86, pp. 253–305, Mar 2014.

[129] Dmarcus100., Nov. 2016.

Accessed: 2020-03-31.[130] E. Chagarov, K. Sardashti, A. C. Kummel, Y. S. Lee, R. Haight, and T. S.

Gershon, “Ag2ZnSn(S,Se)4: A highly promising absorber for thin filmphotovoltaics,” The Journal of Chemical Physics, vol. 144, no. 10, p. 104704,

81

2016.[131] G. Sai Gautam, T. P. Senftle, and E. A. Carter, “Understanding the Effects of

Cd and Ag Doping in Cu2ZnSnS4 Solar Cells,” Chemistry of Materials,vol. 30, no. 14, pp. 4543–4555, 2018.

[132] G. Gurieva, A. Franz, J. M. Prieto, T. Unold, and S. Schorr, “Structural andOptoelectronic Characterization of (AgxCu1−x)ZnSnSe4 solid solution,” in2018 IEEE 7th World Conference on Photovoltaic Energy Conversion(WCPEC) (A Joint Conference of 45th IEEE PVSC, 28th PVSEC 34th EUPVSEC), pp. 0808–0811, June 2018.

[133] C. J. Hages, M. J. Koeper, and R. Agrawal, “Optoelectronic and materialproperties of nanocrystal-based CZTSe absorbers with Ag-alloying,” SolarEnergy Materials and Solar Cells, vol. 145, pp. 342 – 348, 2016.

[134] T. Gershon, K. Sardashti, O. Gunawan, R. Mankad, S. Singh, Y. S. Lee, J. A.Ott, A. Kummel, and R. Haight, “Photovoltaic Device with over 5% EfficiencyBased on an n-Type Ag2ZnSnSe4 Absorber,” Advanced Energy Materials,vol. 6, no. 22, p. 1601182, 2016.

[135] T. Gershon, Y. S. Lee, P. Antunez, R. Mankad, S. Singh, D. Bishop,O. Gunawan, M. Hopstaken, and R. Haight, “Photovoltaic Materials andDevices Based on the Alloyed Kesterite Absorber (AgxCu1−x)2ZnSnSe4,”Advanced Energy Materials, vol. 6, no. 10, p. 1502468, 2016.

[136] C. Yan, K. Sun, J. Huang, S. Johnston, F. Liu, B. P. Veettil, K. Sun, A. Pu,F. Zhou, J. A. Stride, M. A. Green, and X. Hao, “Beyond 11% Efficient SulfideKesterite Cu2ZnxCd1−xSnS4 Solar Cell: Effects of Cadmium Alloying,” ACSEnergy Letters, vol. 2, no. 4, pp. 930–936, 2017.

[137] R. Sun, D. Zhuang, M. Zhao, Q. Gong, M. Scarpulla, Y. Wei, G. Ren, andY. Wu, “Beyond 11% efficient Cu2ZnSn(S,Se)4 thin film solar cells bycadmium alloying,” Solar Energy Materials and Solar Cells, vol. 174, pp. 494– 498, 2018.

[138] T. Fukushima, K. Yamauchi, and S. Picozzi, “Magnetically inducedferroelectricity in cu2mnsns4 and cu2mnsnse4,” Phys. Rev. B, vol. 82,p. 014102, Jul 2010.

[139] T. Fries, Y. Shapira, F. Palacio, M. C. Morón, G. J. McIntyre, R. Kershaw,A. Wold, and E. J. McNiff, “Magnetic ordering of the antiferromagnetCu2MnSnS4 from magnetization and neutron-scattering measurements,”Physical Review B, vol. 56, pp. 5424–5431, Sept. 1997.

[140] R. R. Prabhakar, S. Zhenghua, Z. Xin, T. Baikie, L. S. Woei, S. Shukla, S. K.Batabyal, O. Gunawan, and L. H. Wong, “Photovoltaic effect in earth abundantsolution processed Cu2MnSnS4 and Cu2MnSn(S,Se)4 thin films,” SolarEnergy Materials and Solar Cells, vol. 157, pp. 867 – 873, 2016.

[141] D. Shin, B. Saparov, T. Zhu, W. P. Huhn, V. Blum, and D. B. Mitzi,“BaCu2Sn(S,Se)4: Earth-Abundant Chalcogenides for Thin-FilmPhotovoltaics,” Chemistry of Materials, vol. 28, no. 13, pp. 4771–4780, 2016.

[142] D. Shin, T. Zhu, X. Huang, O. Gunawan, V. Blum, and D. B. Mitzi,“Earth-Abundant Chalcogenide Photovoltaic Devices with over 5% EfficiencyBased on a Cu2BaSn(S,Se)4 Absorber,” Advanced Materials, vol. 29, no. 24,p. 1606945, 2017.

82

Acta Universitatis UpsaliensisDigital Comprehensive Summaries of Uppsala Dissertationsfrom the Faculty of Science and Technology 1944

Editor: The Dean of the Faculty of Science and Technology

A doctoral dissertation from the Faculty of Science andTechnology, Uppsala University, is usually a summary of anumber of papers. A few copies of the complete dissertationare kept at major Swedish research libraries, while thesummary alone is distributed internationally throughthe series Digital Comprehensive Summaries of UppsalaDissertations from the Faculty of Science and Technology.(Prior to January, 2005, the series was published under thetitle “Comprehensive Summaries of Uppsala Dissertationsfrom the Faculty of Science and Technology”.)

Distribution: publications.uu.seurn:nbn:se:uu:diva-407820

ACTAUNIVERSITATIS

UPSALIENSISUPPSALA

2020