Synthesis and characterisation of delafossite CuFeO2 for solar energy applications

UPTEC Q 16006

Examensarbete 30 hp Juni 2016

Synthesis and characterisation of delafossite CuFeO for

solar energy applications

2

Axel Forslund

Teknisk- naturvetenskaplig fakultet UTH-enheten

Besöksadress:

Ångströmlaboratoriet Lägerhyddsvägen 1 Hus 4, Plan 0 Postadress:

Box 536 751 21 Uppsala Telefon:

018 – 471 30 03 Telefax:

018 – 471 30 00 Hemsida:

http://www.teknat.uu.se/student

Abstract

Synthesis and characterisation of delafossite CuFeO

for solar energy applications

Axel Forslund

ISSN: 1401-5773, UPTEC Q 16006 Examinator: Åsa Kassman Rudolphi Ämnesgranskare: Tomas Edvinsson Handledare: Gerrit Boschloo

Delafossite CuFeO₂ is an intrinsic p-type semiconductor with a band gap around 1.5 eV. Further, it is composed of relatively abundant, nontoxic elements, and therefor have potential to be an attractive material for solar energy harvesting.This work examines three routes to synthesise this material.

The first includes a sol-gel deposition and then relies on solid state reaction above 650 degrees Celsius in inert gas atmosphere. In this work, no delafossite is obtained with this method.The second method is a hydrothermal route to make particles under hydrostatic pressure in an autoclave. Delafossite is obtained mixed with other phases.The third route includes aqueous precipitation similar to the second route, but a temperature of 70 degrees Celsius and ambient pressure is sufficient to produce a pure delafossite particle phase. It provides a robust and simple way to make delafossite CuFeO₂

particles.The resulting particles are deposited and compressed on glass into thin films.The films have a band gap slightly below 1.5 eV and show some photoactivity in electrochemical measurements.

i

Syntes och karakterisering av CuFeO

-delafossit för solenergitillämpningar

Axel Forslund

Som en del i att minska den mänskliga påverkan på klimatet måste användningen av fossila bränslen reduceras. Samtidigt som det globala energibehovet ökar, är också flera nya förnybara energislag på väg upp. Produktionen av solceller har de senaste åren ökat och priserna på moduler har fallit. Effektiv energiproduktion från solinstrålningen med tillgängliga och billiga material kan vara viktigt för att uppnå de klimatmål vi har satt upp.

Kommersiella solceller domineras fortfarande av kiselsolceller. Tunnfilmsceller finns dock på marknaden och även färgämnessensiterade solceller, med den engelska förkortningen DSSC, används idag. Många alternativ använder dock ovanliga ämnen som inte alltid bryts för sitt eget värde och därmed blir dyra eller prisberoende av marknadsfluktuationer. Andra ämnen är mycket giftiga och medför risker vid utvinning och tillverkning av solceller.

Nya solcellsmaterial finns på forskningsstadiet, så som CZTS-solceller (koppar- zink- tenn- sulfid/selenid) och hybridorganiska perovskitsolceller. En del av dessa använder vanliga och lättillgängliga ämnen och skulle möjligen kunna tillverkas med lösningsbaserade tekniker, vilket kan sänka tillverkningskostnader.

Utöver att generera elektricitet direkt från solstrålningen kan man bland annat också tänka sig produktion av vätgas med hjälp av solens strålning.

CuFeO2-delafossit är ett material av koppar- och järnoxid som kan absorbera solljus och skulle kunna användas i solenergitillämpningar. Det utgörs dessutom av relativt tillgängliga och ofarliga material. Potential finns för att tillverka filmer med enkla och energisnåla tekniker.

I det här arbetet undersöks tre metoder för att tillverka CuFeO2-delafossit.

1. I den första metoden beläggs ett substrat med en gel av koppar- och järnsalter. Gelen bränns sedan bort tillsammans med saltresterna och kvar blir en amorf film, utan kristallstruktur. Tanken är sedan att anlöpning vid över 650 ˚C i ett argonflöde ska skapa en syrefattig atmosfär där den deponerade filmen av termodynamiska skäl omvandlas till delafossit. I det här arbetet erhålls dock ingen delafossit, utan enbart andra faser av koppar- och järnoxid bildas.

Det är alltså en svår metod, som dessutom kräver ett mycket värmetåligt substrat av specialglas. Vidare medför den stor energiåtgång med flera anlöpningar vid höga temperaturer och ett högt flöde av inert gas.

I de två andra metoderna tillverkas partiklar av CuFeO2-delafossit som sedan kan beläggas på att substrat.

ii

2. Med en “hydrotermisk” metod tillverkas partiklar vid förhöjt tryck och temperatur i vatten i en tryckkokare, eller autoklav. I det här arbetet används temperaturer mellan 100 ˚C till 200 ˚C. För att bilda delafossit krävs en basisk lösning och NaOH används för att höja pH- värdet innan reaktion i autoklaven sker. Så fort NaOH tillsätts fälls ett förstadium till partiklar ut, som sedan åldras i autoklaven. Som utgångsmaterial används en lösning av koppar- och järnsalter, i det här fallet Cu(NO3)2 och FeCl2. I det här arbetet erhålls CuFeO2-delafossit- partiklar tillsammans med en blandning av andra partiklar av koppar- och järnfaser.

3. Med en betydligt enklare metod tillverkas liknande CuFeO2-delafossitpartiklar vid enbart 70 ˚C i vatten i en laboratorieglasflaska i en ugn. NaOH används även här för att fälla ut ett förstadium till partiklar, men dessa behöver inte lika hög temperatur för att bilda enbart delafossit. En del av orsaken till den lägre temperaturen är att sulfater, Cu(SO4) och Fe(SO4), används här – istället för Cu(NO3)2 och FeCl2. Dessa partiklar blir flakformade och ca 200 nm breda, men betydligt tunnare.

I det här arbetet har förhållandet i det initiala skedet av reaktionen, vid tillsättanden av NaOH och utfällningen, visat sig vara mycket viktigt i styrandet av bildade faser. Det är kon- centrationen av NaOH, snarare än mängden NaOH under åldringen, som påverkar vilka faser som bildas. I det här arbetet har pellets av ren NaOH tillsatts, vilket gett en mycket robust syntes av delafossit. Delafossit har bildats som enda kristallina fas synlig i röntgen- kristallografi, även då den tillsatta mängden NaOH varit för liten för att alla koppar- och järnjoner ska kunna reagera.

Det är viktigt att ha en tillräckligt basisk miljö för att kunna reducera Cu²⁺ till Cu¹⁺. Vid tillräckligt basiska förhållanden kan då delafossit bildas, med oxidationstillstånden Cu¹⁺Fe³⁺O2. Men när inte enbart delafossit bildas på grund av ändrade initiala förhållanden, bildas ofta Cu¹⁺2O som bifas. Även där är oxidationstalet Cu¹⁺ och kopparjonen är reducerad från Cu²⁺. För att utreda detta vidare krävs analys av de initialt bildade faserna vid olika förhållanden.

Det pulver som tillverkats har deponerats på glas med en ledande beläggning på, flourdopad tennoxid – även kallat FTO-glas. ”Spin coating” och ”doctor blading” har använts som deponeringsmetoder. De deponerade filmerna har mycket dåliga mekaniska egenskaper, och pulvret lossnar lätt. För att förbättra filmerna har de tillverkade proverna pressats under 15 kN/cm² till 20 kN/cm². De pressade filmerna har bättre mekanisk stabilitet och stannar på substratet även vid nedsänkning i vatten. De visar även fotoaktivitet i elektrokemiska mätningar. Vidare förbättring av filmerna behövs dock innan en fungerande fotovoltaisk eller fotoelektrokemisk anordning är aktuell.

Metod 1 kan jämföras med metod 3 utan hänsyn till om filmer erhållits eller ej, och utan att se till eventuella filmers prestanda. Metod 1 kräver mer energi och arbete med att belägga utgångsmaterial, men ger en film direkt på ett substrat. Metod 3 utgör en enkel och robust metod att tillverka delafossit, men kräver efterarbete för att tillverka en tunnfilm. Å andra sidan kan partiklarna beläggas på en mängd olika substrat, vilker ger större flexibilitet och kanske även en prismässig fördel.

Examensarbete 30 hp på civilingenjörsprogrammet Teknisk fysik med materialvetenskap

Uppsala universitet, juni 2016

Acknowledgements

This work have been conducted with professor Gerrit Boschloo as supervisor and I wish to thank him for this opportunity to investigate an interesting material. As co-supervisors, Malin Johansson and Xiaoliang Zhang have also provided valuable input and helped me with mea- surements and synthesis. Thank you, Malin for all the constructive suggestions and comments.

During this work, I have also had help from many people in the group at Physical Chemistry, which I am very thankful for. Leif Häggman helped me in the lab, Wenxing helped me with electrochemical measurements and you all helped explaining areas which were new to me. I have also had a great time with you in the group.

I will address a special thank to Pedro Berastegui at Inorganic Chemistry, who helped me with the tube furnaces. I also wish to thank Tomas Edvinsson for his useful comments and valuable input.

Acronyms

CIGS copper indium gallium selenide CZTS copper zinc tin sulfide/selenide DSSC dye sensitised solar cell

FTO fluorine doped tin oxide

GIXRD gracing incidence X-ray diffraction LSV linear sweep voltammerty

PV photovoltaic

RHE reversible hydrogen electrode SE secondary electron

SEM scanning electron microscopy SHE standard hydrogen electrode UV-Vis ultraviolet-visible spectroscopy XRD X-ray diffraction

Contents

Populärvetenskaplig sammanfattning i

Acknowledgements iii

Acronyms iv

List of Figures vii

List of Tables viii

1 Introduction 1

2 Background 3

2.1 Semiconductors . . . 3

2.1.1 Direct and indirect band gap . . . 3

2.1.2 Doping . . . 4

2.1.3 The pn-junction . . . 4

2.1.4 Photovoltaic effect . . . 4

2.2 Solar energy . . . 5

2.2.1 Solar water-splitting . . . 5

2.3 Methods for characterisation and analysis . . . 7

2.3.1 XRD . . . 7

2.3.2 UV-Vis . . . 7

2.3.3 SEM . . . 8

2.3.4 Electrochemical measurements . . . 9

2.4 Delafossite CuFeO2 . . . 9

2.4.1 Properties . . . 9

2.4.2 Applications . . . 10

2.4.3 Synthesis . . . 10

3 Experimental 13 3.1 Synthesis . . . 13

3.1.1 Sol-gel route with subsequent annealing . . . 13

3.1.2 Hydrothermal direct growth on substrate . . . 14

3.1.3 Powder synthesis . . . 15

3.1.4 Compression film making . . . 17

3.2 Characterisation methods/equipment . . . 17

3.3 Chemicals . . . 18

4 Results and discussion 19

4.1 Sol-gel route . . . 19

4.2 Powder synthesis . . . 22

4.2.1 Hydrothermal process . . . 22

4.2.2 Low temperature aqueous precipitation process . . . 22

4.3 Compression of film . . . 26

4.4 Choosing an ‘easy route’ . . . 32

5 Summary and conclusion 33

6 References 34

Appendix A SEM 39

Appendix B UV-Vis 42

Appendix C XRD 43

Appendix D LSV 47

Appendix E Synthesis conditions 48

Appendix F Peak broadening of powder XRD with Scherrer equation 51

List of Figures

2.1 Two variations of the delafossite structure . . . 11

4.1 Photos of annealed films on glass . . . 19

4.2 SEM picture of amorphous film . . . 20

4.3 SEM cross section of amorphous film . . . 20

4.4 Diffractograms from samples annealed in 575^◦C argon flow . . . 20

4.5 Diffractograms from samples annealed in 700^◦C argon flow . . . 21

4.6 Comparison of diffractograms from annealed sample at 700^◦C and 800^◦C . . . 22

4.7 SEM pictures of LT12d. . . 23

4.8 Diffractogram from LT12 at different stages in synthesis . . . 24

4.9 SEM pictures of compressed film . . . 28

4.10 SEM cross section of compressed film . . . 29

4.11 UV-Vis from compressed film . . . 30

4.12 Absorption coefficient for compressed film . . . 30

4.13 Tauc band gap estimation of compressed film . . . 31

4.14 Urbach tail band gap estimation of compressed film . . . 31

List of Tables

2.1 Classification system for solar water photolysis . . . 6

3.1 Chemicals used in sol-gel route . . . 13

3.2 Annealing conditions for sol-gel route samples on sola lime glass . . . 14

3.3 Annealing conditions for sol-gel route samples on aluminoborosilicate glass . . . 15

3.4 Precursor solution composition for hydrothermal syntheses . . . 16

3.5 Conditions for different hydrothermal syntheses . . . 16

3.6 LT12 precursor solution composition . . . 16

3.7 Conditions for low temperature synthesis LT12 . . . 17

4.1 Parameters and results for low temperature synthesis LT12 . . . 25

Benefits and drawbacks of sol-gel preparation . . . 32

Benefits and drawbacks of low temperature process . . . 32

1 Introduction

Non-fossil energy is becoming more and more important in our energy system and in our strive to reduce the emissions of carbon dioxide and reduce our effect on the global climate. The global energy demand is still increasing and fossil fuels provide the majority of the energy supply today [1]. Renewable energy is emerging and in recent years we have seen decreasing prices in solar energy and increasing production of photovoltaic devices [2]. Solar energy harvesting with high efficiency and without using materials with high environmental impact might play a key role in reducing our environmental footprint.

Silicon solar cells remains a commercially well-established technique, but the manufacturing is energy intensive and requires extremely pure silicon. The silicon layer also has to be relatively thick due to the indirect band gap, requiring more material. There are some alternatives commercially available such as thin film GaAs, copper indium gallium selenide (CIGS) and CdTe. However, some are made up of exotic materials with low abundance in the earths crust, making them expensive with possibly unstable and market dependent prices and some are quite toxic and require special care, both during manufacturing and when in use. Further, many alternatives have simply not been around long enough for long term stability guaranties.

Electrochemical solar sells such as dye sensitised solar cells (DSSCs) use an approach dif- ferent from the ordinary pn-junction and exist on the market today, but they traditionally use metal-organic complexes with precious metals such as ruthenium as dye and suffer from stability issues. To achieve the best efficiencies they also need to use a liquid electrolyte with the risk of leakage [3], making them less practical if not used mainly for their tunable colour.

New solar cell materials such as copper zinc tin sulfide/selenide (CZTS) and hybrid organic perovskites could provide a solution with relatively abundant materials and possibly also made with solution based techniques cutting the manufacturing costs. Other materials could also be considered for photovoltaic applications.

Another route utilising the solar energy is solar water-splitting. Hydrogen is an attractive energy carrier as it combusts only into water and is functional in a fuel cell. Production of this fuel directly, from only water and sunlight seems like an ideal system. It also reduces the intermittency problem associated with renewable power generation from sun and wind, as the fuel produced can be stored for an arbitrary time (and transported), until combusted or used in a fuel cell.

In theory, solar water-splitting can reach high efficiencies, close to the Shockley-Queisser limit above 30 % for 1 Sun illumination, more practical limits have been estimated to between 10 % and 20 % depending on the configuration [4, 5]. The later estimation requires a tandem system.

Delafossite CuFeO2 is made up of comparatively abundant materials and could possibly be used as an absorber material in a solar cell or to reduce water in a solar water-splitting device.

It has been reported to be an intrinsic p-type semiconductor and has a band gap of about 1.5 eV. However it has also been reported to be challenging to synthesise [6, 7].

In this study three earlier reported routes synthesising delafossite CuFeO2 are examined and thin films are made either directly in the synthesis or from the product resulting from the

delafossite synthesis. In only one of the synthesis routes, delafossite CuFeO2 is produced as the only phase after slight altering of the reported method. In the other routes, no delafossite or a mix of phases are produced.

Out of these three routes, two are compared from a synthesis point of view: A sol-gel route requiring annealing and a low temperature particle route requiring deposition and compression after the synthesis.

2 Background

In the following section an introduction about semiconductors, solar energy and the measuring techniques used in this work is given. Next, a short background follows about the material system examined in the work. Some earlier reports for the synthesis of and measurements on the material is included.

2.1 Semiconductors

A semiconductor is a solid with a conductivity that is intermediate that of a metal and an insulator [8]. Alternatively, it can be more appropriate to define a metal as having it’s Fermi level where the density of states is not vanishing [9, 10]. A semiconductor or insulator on the other hand, have the Fermi level in a region of vanishing density of states.

Bands of electron energy states and band gaps arise when introducing quantum mechanical models together with periodic potentials in solid crystals. The crystal structure of a crystalline chemical compound determines the band structure with energy states and the electron proper- ties of the compound determines the Fermi level and in which way the electrons occupy these states.

Metals can have bands that are not completely filled, and can easily — with a small amount of energy — excite electrons into free electronic states within the band to be able to act nearly as free electrons. Metals are therefor often good conductors. But a material can also be a metal if bands overlap with each other. This gives a similar effect, as there are no band gaps hindering the electrons to get excited, but a continuum of states still allow them to easily get excited.

Semiconductors and insulators on the other hand, have band gaps with the valence band fully occupied. The electrons therefor need larger amounts of energy to reach a free state, which reduces their ability to conduct current. The difference between semiconductors and insulators is not always clearly defined, but generally the band gap of a semiconductor is small enough to allow electrons from the valence band to get thermally excited into the conduction band at reasonable temperatures.

2.1.1 Direct and indirect band gap

There are two types of band gaps: Direct and indirect. A direct band gap have its valence band maximum at the same position in the Brillouin zone as the conduction band minimum. An electron will thus need extra energy to cross the gap. An indirect semiconductor on the other hand, does not have its valence band maximum in the same place in reciprocal space as the conduction band minimum. To cross the gap, an electron will therefor need not only photon energy, but also energy from a crystal momentum.

As photons carry little momentum, the excitation by a photon needs to be combined with some momentum from the crystal lattice, which drastically decreases the probability for light

absorption for indirect band gap semiconductors. Two examples are GaAs as a direct band gap semiconductor and Si as an indirect band gap semiconductor. In practical terms this forces a Si solar cell to be much thicker than a GaAs solar cell.

2.1.2 Doping

To increase the carrier density and thus the conductivity of a semiconductor, small amounts of impurities can be implanted into the crystal lattice, so called doping. There are two types of doping, n-type and p-type. For n-type doping electron donors are inserted, elements with one valence electron more than the semiconductor. They bind by forming hybrid orbitals with the surrounding semiconductor atoms, but the extra electron is then left over, not included in any binding hybridisation. The electron is then just loosely bound to the positively charged dopant atom, otherwise acting like a free electron.

By using the similarity to the hydrogen atom — substituting the electron mass to an ef- fective electron mass and the dielectric constant of vacuum to the dielectric constant of the semiconductor — the binding energy can be calculated relative to the conduction band mini- mum. The energy is rather small and the donor electrons are easier to excite than the intrinsic electrons. In some intermediate temperature region when the donors are thermally ionised but the intrinsic are not the electron density is practically the same as the donor concentration in the semiconductor.

The same reasoning is used for p-type doping where an element with one electron less than the semiconductor is inserted. This creates an extra state to where an electron can easily be excited and thereby create a hole in the valence band.

2.1.3 The pn-junction

A pn-junction is in principle made up of an n-type semiconductor and a p-type semiconductor joined together. In the boundary between the two types of semiconductors the electrons and holes diffuse to the other side. As they recombine a depletion layer will appear where only the charged, fixed dopant atoms are left. As only the ionized atoms remains the layer is polarised and hinders the charges to diffuse further over the boundary, and equilibrium can be reached.

If a minority electron from the p-side enters the depletion region it will drift to the n-side driven by the depletion region electric field. This giver rise to a drift current that is compensated by the diffusion current of electrons still diffusing from the n-side to the p-side at equilibrium.

If external bias is applied, the chemical potential on either side can be changed. This in turn changes the diffusion current to increase for forward bias, or decrease for reverse bias (eventually approaching zero, leaving almost only the drift current at a practically constant value).

2.1.4 Photovoltaic effect

There are several effects that can give rise to the buildup of a voltage in a semiconductor. One is the direct excitation of electrons by photons.

If a semiconductor is exposed to light, electrons in the valence band can be excited to the conduction band by absorbing the energy of a photon. The energy of the photon must then be at least as high as the energy gap between the valence band and the conduction band of the semiconductor. For smaller energy, there is no density of states for the electron to go to. For absorption of higher energy photons, the electron gets excited up above the conduction band edge, but will soon loose excess energy and relax to close to the conduction band edge.

If the electron in the conduction band and the resulting hole in the valence band can be separated, then a voltage is built up and work can be extracted from the material. If an

electron-hole pair is generated in the depletion region in a pn-junction, they are effectively separated by the built in voltage. This is the way for conventional solar cells to utilise light to build up a voltage that can be used to perform work in a circuit [9, 10], and these solar cells are called photovoltaic (PV) devices, or cells, in this work.

The maximum efficiency of a PV device is predicted by Shockley & Queisser [11] and reaches just above 30 % for a single junction device.

2.2 Solar energy

The basic functioning of a conventional PV device is described in section 2.1.4. The most common material for PV devices is Si, and Si-cells have in recent years decreased in price, and the worldwide installed capacity increases [12].

However, there are also other ways of generating a voltage from sunlight, for example in a photoelectrochemical cell, where a semiconductor is immersed in an electrolyte.

DSSCs is a subtype of these and use dyes attached to a large band gap semiconductor.

The dye acts as an absorber, absorbing photons to excite electrons into higher energy states.

These excited electrons ideally injects into the high band gap semiconductor. Simultaneously the empty state in the dye attracts an electron from — or injects a hole into — the electrolyte, oxidising a species in the electrolyte. This species is transported to the counter electrode where it can get into a reduced state again [13].

The electrolyte can be made of a solution with I^- /I3- redox species, or of some hybrid metal organic complex with e.g. Co [14].

When the excited electrons injects into the large band gap semiconductor the risk of recom- bination decreases and the electrons can be transferred to a contact more efficiently.

2.2.1 Solar water-splitting

Previously, methods of generating electricity from sunlight have been mentioned. Another way of extracting energy from the solar radiation is to use it directly to split water into hydrogen and oxygen. This can be done in a way similar to the electrochemical solar cell, but the reversible redox reaction for the electrolyte is substituted to oxidation of water into oxygen at the anode and reduction of the thereby released protons into hydrogen at the cathode. Either of, or both, the anode and the cathode can be a photodevice, absorbing photons and driving the cell half-reactions [5].

The formal electrochemical potential required to drive the water-splittingis around 1.23 V.

and can be divide into the following half reactions. The potential is given relative to the reversible hydrogen electrode (RHE):

Reduction reaction:

2H⁺+ 2e⁻→ H² (E0^red = 0.0 V v.s. RHE) Oxidation reaction:

H2O + 2h⁺ → ¹₂O2+ 2H⁺ (E0^ox = −1.23 V v.s. RHE) Complete reaction:

H2O → H2+¹₂O2 (∆E = −1.23 V v.s. RHE)

The overall reaction has a negative standard potential and energy must be added for the reaction to occur. Even though 1.23 V theoretically is enough for some reaction, higher voltages

Table 2.1 – Classification system for solar water photolysis schemes. S: singel photosystem; D: dual photosystem operating at different threshold wavelength λ1

and λ2. After Bolton et al. [15].

Scheme

classification No. of

photosystems Minimum no.

of absorbed photons per H2

Reaction

S1 1 1 H2O −−→ H^1hv 2+¹₂O2

S2 1 2 H2O −−→ H^2hv 2+¹₂O2

S4 1 4 H2O −−→ H^4hv 2+¹₂O2

D2 2 2 H2O −−−−−→ H^hv1+hv2 2+¹₂O2

D4 2 4 H2O −−−−−−→ H^2hv1+2hv2 2+¹₂O2

are required in practice. The excess voltage required is called overvoltage and is caused by the difference in chemical potential required to drive the reaction and by losses in the system [15].

The extra chemical potential required is inevitable and can be calculated [16], and some losses will also aways be in the system but have to be estimated to a reasonable value. The maximum efficiencies for a photolytic system are also different depending on which principal features are utilised.

Bolton et al. [15] sort some different photoelectrochemical (PEC) systems into five different classes, denoted S1, S2, S4, D2 and D4. These notations indicates the number of photosystems used (S for singel photosystem and D for dual photosystem) and the number of photons (1,2 or 4) used to produce one H2 molecule. These classes are summarised in table 2.1.

In the same paper, Bolton et al. suggests the ideal efficiency limit for the S2 system to be 30.7 %, at λg = 775 nm, and for the D4 system to be 41.0 %, with λ1 = 655 nmand λ2 = 930 nm. However, Bolton et al. also estimates some more realistic efficiencies of around 17 % and 27 % for the same systems respectively, still assuming 100 % quantum yeld, absorption of all incoming sunlight and neglecting losses from reflections at interfaces. Further, they predict practical efficiencies of 10 % and 16 % for the S2 and D4 systems respectively, by counting in likely losses [15].

The optimal band gap energy for the maximum efficiency varies with the assumed losses, but in the analysis of Bolton et al. [15] and the similar analysis of Prévot & Sivula [5] the optimal band gap energy is around 1.3 eV and 1.9 eV for each photoelectrode respectively. Hu et al. [17] get optimum band gap levels of 1.0 eV to 1.2 eV and 1.6 eV to 1.8 eV respectively, for different resistive losses, losses due to a fill factor <1, etcetera.

So in theory, solar water-splitting can reach high efficiencies, close to the Shockley-Queisser limit, but it has been suggested that reasonable practical limits could lie at 10 % and 16 % for single photosystems and dual photosystems respectively. Prévot & Sivula [5] use a similar analysis but with a tandem device with two different semiconductors with different band gaps and predicts a solar to hydrogen efficiency of around 22 %.

In their discussion Prévot & Sivula compare this with a tested solar to hydrogen efficiency of 9.3 % in a system with a commercial PV silicon cell and a high pressure electrolyser. This comparison could be questioned as their calculated efficiency is for a tandem device while the tested system is for a single junction, commercial silicon cell with about 15 % solar to current efficiency (not a state-of-the-art cell). Further, the produced hydrogen in the tested system comes out pressurised — an inevitable step for practical hydrogen storage which is not counted into the 22 % efficiency claimed by Prévot & Sivula.

Further, as stated by Hu et al. [17], PV cells coupled with electrolysers can be tuned

and wired to match their I-V characteristics so that the total efficiency of the system will simply be the product of the efficiency of the solar cell and the electrolyser. In the case of a state-of-the-art system with a multijunction PV cell under concentrated solar illumination, 43.5 %× 73 % = 31.8 % efficiency, or recently even higher, should be possible to reach now [17].

By contrast, Jacobsson et al. [18] argue that PV/electrolyser systems and PEC cells do not differ considering the physical processes involved. This would mean that they essentially are affected by the same losses and therefor have the same theoretical efficiencies. Factors that would then matter are more coupled to properties of materials required for e.g. stability in PECs, inevitable light absorption losses for solution immersed systems or grid losses in grid connected PV/electrolyser systems.

From this point of view a PV/electrolyser system is the most developed and commercially mature, due to the fact that each technique can — and have been — developed independently.

An obstacle for commercialisation have been the low price for natural gas derived hydrogen, rather than technology development [18].

2.3 Methods for characterisation and analysis

2.3.1 XRD

To characterise the crystal structure of the samples made in solid state reaction, hydrothermal and low temperature synthesis, X-ray diffraction (XRD) was used. An X-ray beam with a specific wavelength is sent towards the sample. The wavelength is small — around the size of the lattice spacing of the sample, e.g. Cu-Kα was used in this work, which is around 1.54 Å — and thus the x-ray interacts with the sample through constructive and destructive interference.

Constructive interference is achieved for certain angles determined by the Bragg-condition:

nλ = 2d sin(θ)

where n is an integer number, λ is the wavelength, d is the distance between two crystal planes and θ is the incident angle of the beam.

For thin films a constant grazing incidence angle can be used to enhance the relative intensity of the thin film at the surface compared to the substrate. The method is called gracing incidence X-ray diffraction (GIXRD).

2.3.2 UV-Vis

In a typical ultraviolet-visible spectroscopy (UV-Vis) measurement the absorbance and/or the reflectance of a sample is recorded over a span of wavelengths. Light from a light source is lead through a monochromator (for example a prism and a slit) to obtain a small span of wavelenghts. The light is then allowed to pass through a sample or reflect against a sample before collected in a detector. This gives an absorbance or reflectance measurement respectively.

After interaction with the sample the light can be collected with an integrating sphere as to include scattered light for detection. The integrating sphere ideally consists of a sphere whose walls scatters all light.

Estimation of the band gap energy can be done following the Tauc method [19, 20], using the following equation for measurement data in the absorption edge in the absorption spectrum:

αhv = B(hv− E^g)^r

Here α is the absorption coefficient, h is Boltzmann’s constant, v is the photon frequency, B is a constant, Eg is the band gap energy and r is an index associated to indirect and direct

band gap, among other things [21]. r = 2 corresponds to an indirect band gap and r = 1/2 to a direct band gap. If (αhv)^1/r is plotted against the photon energy (hv), then Eg can easily be determined by extrapolating a line in the absorption edge — for some suitable r — to αhv = 0, where then the corresponding hv = Eg.

The absorption coefficient in case of uniform absorbance through the film is defined as [19]:

α =−1 dln(T )

where d is the thickness and T the transmittance of the layer of interest. During measure- ments, some of the incident light is reflected before entering the material and the reflectance contribution must be extracted from the measured transmittance:

T ≈ 1− R^measured Tmeasured .

Here R is the measured reflectance. This is an approximation and does not take multiple reflexes from the several interfaces into account. The resulting formula is

α =−1 dln

!1− Rmeasured

Tmeasured

"

There could be several layers other than that of interest in the sample and these have to be compensated for as well. For example fluorine doped tin oxide (FTO) coated glass is used in this work, and even though the glass is assumed to be completely transparent the FTO needs to be taken into account. A first order correction αd = αtotdtot− αF T OdF T O can be used as an approximation to remove the FTO absorbance.

2.3.3 SEM

A scanning electron microscopy (SEM) accelerates electrons from an electron emitting source (e.g. a LaB6 crystal or a hot W-filament can be used.) through a system of lenses and detects scattered electrons resulting from interactions with the accelerated electrons and the sample.

The beam of accelerated electrons is scanned stepwise across the sample, and a detector counts electrons resulting from interactions for each point at which the beam is focused. Then, a computer creates an image of the gathered information from these scanned points.

The SEM thus gives a black and white picture of the sample with contrast due to geometry and compositional differences at the sample ‘surface’. The information given in the picture results from interactions in the interaction volume of the electron beam and the sample at a specific point. The interaction volume is determined by the spot size and the penetration depth of the electrons. The electron penetration depth depends on the acceleration voltage and the material, but is typically around 0.5 µm to 5 µm, though the depth from which different mechanisms give information always differs.

The spot size is the cross section of the cone formed by the electron beam, at the sample surface. It is changed by focusing the beam, but limited by not only the lens system but also the sample roughness and height differences in the sample as in that case not a single object plane can be obtained — some depth of field is needed to see the whole picture clearly. A tilting sample is also to be avoided, as the beam then has to refocus during the scanning to get a clear picture over the whole scanning area.

There are several types of detectors used in SEM. One is the secondary electron (SE) detector which is situated not right above the incident electrons, but at an angle from the normal of the surface. This detector detects the SEs that have escaped from the material after energy transfer from the incident electrons. As they have to escape the material, they loose energy and

therefor only the electrons close to the surface can escape outside the sample. This detector is therefor more surface sensitive and suitable for topological measurements.

Two other types of detectors used in SEM are back scattered electron (BDE) detectors and energy dispersive X-ray spectrometry (EDS or EDX) detector. However, they are not used in this work.

2.3.4 Electrochemical measurements

To measure the behaviour of a species at different potentials, electrochemical methods can be used. The measurements depend not only on the electrical field in the cell, but also on chemical events during the measurement. Voltages are often given relative to some standardised reference.

The standard hydrogen electrode (SHE) is defined to be the reference electrode for which, under standard conditions and at all temperatures, E⁰(H⁺/H2) = 0, i.e. the reduction potential of a hydrogen ion to hydrogen gas is zero. In an aqueous solution at 25^◦Cthis is calculated to be between 4.4 V to 4.5 V v.s. the potential of an electron at rest in vacuum.

In electrochemical standards the voltage is given with the positive direction reversed com- pared to the physical convention of decreasing potentials into the negative region from the electron at rest in vacuum.

In practical measurements a reference electrode is used. One such is the Ag/AgCl electrode, where Ag and AgCl is in equilibrium with a KCl-solution. The potential of Ag/AgCl depends on the concentration of KCl in the solution, but is usually around 0.2 V v.s. SHE. If the solution in the reference electrode is saturated the potential is 0.197 V v.s. SHE.

If the pH in the electrolyte is not 0, as in the SHE, the reduction potential of hydrogen will be different. Therefor RHE is often defined to have the potential of hydrogen reduction at any pH. It therefor coincide with SHE at pH 0, but can generally be estimated from the relationship:

E = 0.0000 V− 0.0591 V × pH (v.s. SHE)

E.g. if measured voltages are given with respect to a saturated Ag/AgCl reference electrode in 1 m NaOH solution, then Ag/AgCl is

0.197 V (v.s. SHE), and SHE is

0.0591 V× 14 = 0.8274 V (v.s. RHE),

which gives a total difference of about 0.197 + 0.827 = 1.024 ≈ 1.0 V for Ag/AgCl v.s. RHE.

So the reduction potential for hydrogen at pH 14 is then around −1.0 V v.s. Ag/AgCl.

In electrochemistry, a cathodic current goes into the electrode, i.e. electrons goes into the solution and reduce the electrolyte. Correspondingly, an anodic current goes from the electrode into the solution, i.e. the electrons move from the solution to the electrode, oxidising the electrolyte.

A measurement that scans the voltage in a constant rate from one voltage to another, and measures the current during the scanning is called linear sweep voltammerty (LSV).

2.4 Delafossite CuFeO

2.4.1 Properties

Delafossite crystals typically have the chemical formula A⁺¹B⁺³O2 [22]. It is constructed by alternating layers of close packed standing O–C⁺–O dumb-bells and Fe³⁺O6 octahedrons. There

are two common crystallised states of delafossite: One with rhombohedral 3R-(R3m) symmetry (3R delafossite) and one with hexagonal 2H- (P6c/mmc) symmetry (2H delafossite), with the stacking sequences ‘AaBbCcAa. . . ’ and ‘AaBbAa. . . ’ respectively. An illustration of the structures can be seen in figure 2.1. (The illustrations are made with ‘VESTA’, a crystal structure illustration program [23].)

CuFeO2 can be an intrinsic p-type conductor and conducts well compared to some other intrinsic p-type materials [24]. The good conductivity have suggestively been attributed to Cu vacancies and interstitial O in the delafossite crystal structure [24], but also to the strong covalent nature of the Cu-O bond [7]. The conductivity of single crystal delafossite CuFeO2

is higher perpendicular to the c-axis compared to parallel to the c-axis, and in the order of 1 Ω⁻¹cm⁻¹ [24–27].

2.4.2 Applications

Delafossite CuFeO2 has reportedly an optical band gap¹ at around 1.1 eV to 1.6 eV [6, 7, 24, 27–30] with the conduction band positioned at around −0.4 eV relative to RHE at pH 13.6, according to Prévot et al. [6]. This is about 3.2 V below vacuum level and is in the same region as measured earlier [27, 31]. This is a suitable position for the reduction of water into hydrogen as a photocathode in a tandem water-splitting cell.

One of the greatest advantages of delafossite CuFeO2is its long term chemical stability under neutral and alkaline conditions [32]. It has been tested to show small declining in photoactivity in electrochemical measurements after hours and even months of operation [6, 32].

It could also possibly be used as a hole conductor in PV devices. However, the conductivity features some anisotropy [25] with good conductivity perpendicular to the c-axis, but worse conductivity parallel to the c-axis.

2.4.3 Synthesis

Solid state reaction

The conventional way of preparing delafossite CuFeO2 uses a high temperature reaction in oxygen deficient atmosphere, where delafossite CuFeO2 seems to be thermodynamically stable [20]. Delafossite CuFeO2 may form in temperatures above 650^◦C, but in many earlier works temperatures of up to more than 1000^◦C have been used [6, 7, 25, 27, 29, 31, 33].

Sol-gel preparation

A sol-gel process should include the transition of a sol into a gel [34]. A sol is often defined as a suspension of small particles in a liquid that remains dispersed during the process, and a gel as a three dimensional network extending uniformly over the liquid phase. There are more and less rigorous definitions of the expression, but according to a wider definition the gel can be formed from e.g. metal-organics and produces a uniform solid or highly viscous liquid matter with no precipitation of the compounds.

The ‘Pechini route’ [35] includes using an alpha-hydroxycarboxylic acid like citric acid in combination with metal oxides or salts. After adding a polyhydroxy alcohol such as ethy- lene glycol, a polyesterification creates a polymer gel from metal chelate complexes with the carboxylic acid and the alcohol [34, 36].

Prévot et al. [6] have earlier prepared films with a sol-gel² route, or modified Pechini route, to get a uniform thin-film with the desired Cu-Fe stoichiometry. They spin coat an equimolar

1Whether the band gap is direct or indirect differs in the reports.

2According to a broader definition [34].

(a) (b)

(c) (d)

Figure 2.1 – There are two structures of delafossite CuFeO2: (a) and (c) with rhombohedral 3R-(R3m) symmetry (3R delafossite) viewed from the a and c axis, respectively; (b) and (d) with hexagonal 2H-(P6c/mmc) symmetry (2H delafossite), also viewed from the a and c axis, respectively. The stacking sequences of the Cu–Fe layers are ‘AaBbCcAa. . . ’ for 3R delafossite and ‘AaBbAa. . . ’ for 2H delafossite.

solution with Cu and Fe nitrates in ethanol with citric acid and ethylene glycol. In the solution, the alpha-hydroxycarboxylic acid (the citric acid) forms a chelate with the metal ions and during spin coating this reacts with the polyhydroxy alcohol (the ethylene glycol) into a polymer gel when ethanol is evaporating, evenly distributing the different metal ions through the film [34, 36].The organic polymer is then burned away at elevated temperature and the first non-organic film obtained is amorphous and must undergo a solid state high temperature reaction in oxygen deficient atmosphere to form delafossite CuFeO2, as described under ‘Solid state reaction’.

Hydrothermal synthesis

Instead of heating the reactants up to > 650^◦C, high pressures can be used in combination with relatively low temperatures in a liquid to prepare delafossite structured powders [22, 37–43].

In autoclave, delafossite powders have been obtained at temperatures as low as 100^◦C[37]. In these syntheses, large amounts of NaOH is used [37, 38] and the pH during the reaction can be quite high. However, some of the NaOH seems to be consumed in the reaction and it is not clear whether the high pH catalysts the reaction or just acts as a ‘consumable’.

John et al. [44] used something similar to a hydrothermal synthesis using Cu(SO4)•5H2O and Fe(SO4)•7H2O in water at 70^◦Cadding differently concentrated NaOH-solutions. However, the vessel was not sealed to build up a pressure and was thereby not strictly a hydrothermal synthesis. In this report, it will be named ‘low temperature (synthesis/route/method)’. This low temperature method is reported to result in precipitated delafossite CuFeO2 powder in a NaOH solution after annealing in an oven at 70^◦C.

Compression

There have been earlier reports of making metal oxide films on glass with a compression method [45]. Lindström et al. use a suspension of 20 %wt to 25 %wt TiO2 powder in ethanol, doctor bladed on a glass substrate. When the ethanol has evaporated the sample is compressed using a hydraulic press with a pressure in the order of 10 kN/cm².

By this method µm-thick films are made in a simple way. The films are conducting and the porosity can be controlled by adjusting the pressure. Powder of delafossite CuFeO2 could possibly be prepared in a similar way.

3 Experimental

In this part of the report the experimental details are given. All the analytical equipment used are listed in section 3.2 and the chemicals used are specified in section 3.3.

3.1 Synthesis

3.1.1 Sol-gel route with subsequent annealing

The ‘sol-gel route’¹ used in this thesis is based upon the synthesis method used by Prévot et al.

[6]. The method includes the following initial steps:

1. Mixing of equimolar Cu(NO3)2•3H2O and Fe(NO3)3•9H2O in ethanol until dissolved 2. Adding citric acid

3. Stirring for 2 h

4. Adding ethylene glycol 5. Stirring overnight

6. Spin coating on FTO at 3000 rpm for 1 min 7. Drying on hot plate at 100^◦C for 10 min

8. Annealing in a furnace at 450^◦C for 30 min to burn away organic material. (2 h is used by Prévot et al.)

The amounts of precursor chemicals used are listed in table 3.1.

1The route could be called ‘sol-gel route’ by a broader definition. Modified ‘Pechini route’ could possibly be considered more correct, see section 2.4.3 under ‘Sol-gel preparation’.

Table 3.1– Chemicals used in sol-gel route with subsequent solid state reaction

Chemical Amount (mmol) Amount

Cu(NO3)2•3H2O 2.00 482 mg

Fe(NO3)3•9H2O 2.00 808 mg

Ethanol — 10 ml

Citric acid 4.00 768 mg

Ethylene glycol 4.48 0.25 ml

Table 3.2– Annealing conditions for sol-gel prepared samples on FTO-coated soda lime glass. The ramping rate while heating was 10^◦C/min, while the samples where cooled naturally in the furnace. The gas flow could not be measured with the furnace used in this case.

Sample Temperature (^◦C) Type of gas

A2 500 Air

C3 575 Air

C5 650^a Air

C2 500 Ar

C4 575 Ar

aAt this temperature the substrate softened, so no annealing in argon was done at 650^◦C

This results in a brownish, transparent film. The described process can be repeated several times to increase the thickness of the thin film. After each cycle the colour changes into more dark brown and the film appears less transparent upon ocular investigation. It should be noted that repetition of the process was tried without the last annealing step. However, the resulting films did not change colour, which seems reasonably to be due to dissolution of the previous layer – so that it washes away during spin coating of the new layer.

Further, according to earlier investigations with termogravimetric analysis [6, 20, 46, 47]

crystalline phases of CuFe2O4 and CuO start forming at 450^◦C and above. Therefore, the temperature of the first annealing should not exceed this temperature to obtain a uniform, non-crystalline film.

After this first deposition of Cu and Fe nitrates, the film is annealed for a second time in higher temperature:

• Annealing in tube furnace at >500^◦Cin controlled atmosphere

Prévot et al. use a temperature of 700^◦C in argon. In this work both lower and higher temperatures are also investigated for the second annealing. The conditions used for samples on FTO-coated soda lime glass are presented in table 3.2. Sample C5 was annealed in air at 650^◦C but did not withstand the temperature, so no sample was annealed in argon for this temperature.

When annealing at 700^◦C a special glass must be used as the glass transition temperature and softening point of usual soda lime glass lies slightly above 500^◦C [48] and cannot stand higher temperatures, as noticed for sample C5. Therefore FTO-coated aluminoborosilicate glass from Solaronix SA was used above these temperatures. The annealing conditions for these samples are presented in table 3.3.

Before heating the samples in table 3.3, argon gas was set to flow through the tube furnace for 3 h to 6 h at room temperature to decrees the amount of oxygen present. Prévot et al. [6]

use an argon flow of 300 ml/ min, and this flow rate was used also in this work. Significantly lower argon flow rates were investigated but resulted in e.g. metallic tin phases from the FTO.

3.1.2 Hydrothermal direct growth on substrate

In section 3.1.3 under ‘Hydrothermal process’ a route to make particles is described. In an attempt to grow a film by precipitating particles directly on FTO-coated soda lime glass in hy- drothermal conditions, substrates were immersed into an autoclave together with the precursor solution. The full description of the hydrothermal process can be seen in the named section.

Table 3.3– Annealing conditions for sol-gel prepared samples on FTO-coated alu- minoborosilicate glass.

Sample Temperature (^◦C) Ar-gas flow (ml/ min) Comment

F1 700 300 old film

E1 700 300 Spin coated on glass

E2 700 300 Spin coated on FTO

However the glass did not stand the highly alkaline environment used for making the particles during heating in the autoclave and the glass substrates were destroyed.

Further, precipitation from the precursor solution seems to occur directly when adding NaOH. Thus, direct growth from this system seems hard for this reason, not only for the hydrothermal process, but also the low temperature process described under ‘Low temperature aqueous precipitation process’.

3.1.3 Powder synthesis

To synthesise powder a hydrothermal method and a low temperature method has been tested.

The hydrothermal method [22, 37, 39–43] involves heating an aqueous solution in an autoclave container maintaining a constant volume and thus raising the pressure within the container as the temperature increases. In the synthesis a high pH have been used before, as noted in section 2.4.3 under ‘Hydrothermal synthesis’, and NaOH seems to be consumed in the process.

The low temperature method [38, 44] also uses an aqueous solution but in a temperature below 100^◦C. John et al. [38] use 50^◦C to 90^◦Cto obtain delafossite CuFeO2.

After taking the samples named in the following sections out of the oven, they have been washed with deionised water twice and in ethanol once. To extract the particles from the washing liquid the suspension was centrifuged for 3 min at 4000 rpm.

Hydrothermal process

The hydrothermal synthesis method was taken mainly from Xiong et al. [37] although many other similar routs have been presented before [22, 39–43]. The route included:

1. Dissolving Cu(NO3)2•3H2O and FeCl2•4H2O in deionised water by stirring and some heating

2. Adding NaOH 3. Stirring for 10 min

4. Pouring the solution into autoclave and sealing

5. Putting the autoclave into an oven at a temperature in the range of 100^◦Cto 160^◦C for 17 h to 40 h

6. Natural cooling to room temperature

The composition of the precursor solutions and the amount of added NaOH is presented in appendix E, but a typical example is shown in table 3.4.

The hydrothermal synthesis was altered in search for a method giving the most delafossite CuFeO2 phase. It was altered as presented in table 3.5.

Table 3.4– Example of precursor solution composition for hydrothermal syntheses.

Chemical Amount (mmol) Amount

Water 1100 20 ml

Cu(NO3)2•3H2O 4.2857 1035.4 mg

FeCl2•4H2O 4.2857 852.0 mg

NaOH 31.4 1260 mg

Table 3.5– Conditions for different hydrothermal syntheses.

Sample Temperature (^◦C) Added NaOH (g) Time in oven (h)

HT1 100 1.257 21

HT4 120 1.25 17

HT5 140 2.5 66

HT6 160 1.25 22

HT7 160 2.5 22

HT8 160 2.5 40

Low temperature aqueous precipitation process

For the low temperature powder synthesis with aqueous precipitation the route followed was mainly that of John et al., with some differences worth of notice for sample LT12. LT10 and LT11 were prepared as close to the route used in John et al. [38] as possible. However, the adding of NaOH solution showed to be a critical moment, and monitoring the pH as the NaOH was added proved to be very hard.

Therefor the NaOH was weighted and added in solid form for sample LT12, and the pH was measured with indicator paper for the fresh samples just as the precipitate had settled at the bottom of the vessel, and after 24 h ageing in the same vessel.

Hereafter, the term ageing will be used for the period during which the samples are stored at elevated temperature (70^◦C), after the NaOH have been added and the initial reaction have occurred.

Table 3.6 shows the composition of the precursor solution for the low temperature particle synthesis of LT12 samples. A summary of the added amount of NaOH into samples LT12a, LT12b, LT12c, LT12d and LT12e is shown in table 3.7. The temperature is held constant at 70^◦C for all samples. The molar amount of Cu and Fe approximately corresponds to a concentration of 10 g/l.

Table 3.6– LT12 precursor solution composition.

Chemical Amount (mmol) Amount

Water — 10 ml

Cu(SO4)•5H2O 1.574^a 393.0 mg

Fe(SO4)•7H2O 1.574^a 437.6 mg

aCorresponds to an approximate Cu²⁺/ Fe²⁺-concentration of 10 g/l.

Table 3.7 – Conditions for low temperature synthesis LT12.

Sample Temperature (^◦C) Added NaOH (g)

LT12a 70 0.20

LT12b 70 0.25

LT12c 70 0.30

LT12d 70 0.40

LT12e 70 0.50

In table 3.7 the conditions for synthesis LT12 are listed. The reaction temperature is 70^◦C for all samples and NaOH is added to the sample solutions with the amount increasing for the samples in alphabetical order from LT12a to LT12e.

After the synthesis, the samples were cooled naturally from the reaction temperature and washed with deionised water twice and in ethanol once. The samples were stored in ethanol until characterisation could be undertaken.

3.1.4 Compression film making

In this work the principle when making CuFeO2is the same as for the method used by Lindström et al. [45]. CuFeO2powder is mixed with ethanol to a suspension. The suspension is then doctor bladed or spin coated onto a FTO glass substrate, or in some cases to a SnO2 coated FTO glass.

The resulting film is then compressed in a hydraulic press under a pressure of around 15 kN/cm² to 20 kN/cm². For the compression, the sample is placed in between two polished stainless steel cylinder plates. Aluminum foil is separating the sample film and one of the steel plates to avoid contamination of the plate.

3.2 Characterisation methods/equipment

• SEM: Zeiss LEO 1550 FEG

• XRD:

– Powder XRD

∗ Siemens D-5000 Th-Th

∗ Bragg-Brentano setup

∗ Motorised slits

∗ Instrumental broadening around 0.1^◦ – GIXRD:

∗ Siemens D-5000 Th-2Th

∗ Parallel beam setup

∗ X-ray mirror

∗ Instrumental broadening around 0.3^◦ – Both use a 1.54 Å Cu-Kα X-ray-source

• UV-Vis: Perkin Elmer Lamba 900 spectrometer – Double-monochromator

– 50-mm in diameter BaSO4 coated integrating sphere

3.3 Chemicals

• Cu(NO3)2•3H2O, 99.5 %, analysis grade, Merck KGaA

• Fe(NO3)3•9H2O, 98 %–101 %, Fluka chemie AG

• Citric acid, 99.5 %, reagent grade, Sigma-Aldrich Chemie GmbH

• CuSO4•5H2O, 99 %, analysis grade, Merck KGaA

• FeSO4•7H2O, 99 %, reagent grade, Sigma-Aldrich Chemie GmbH

• FeCl2•4H2O, 99 % analysis grade, Riedel-de Haën AG

• SnO2, colloidal, 15 %wt SnO2, K⁺ pH 10, NYASOL SN15, The PQ Corporation

• Polyethylene glycol (PEG), 20 000(16 000–24 000), Fluka chemie AG

• NaOH, 98 %, Sigma-Aldrich Chemie GmbH

• Ethylene glycol, Sigma-Aldrich Chemie GmbH

• Terpineol, Sigma-Aldrich Chemie GmbH

• Salicylic acid, 99 %, Sigma-Aldrich Chemie GmbH

• Ethanol, VWR

4 Results and discussion

Three methods for synthesising delafossite CuFeO2 have been investigated and the results are described and discussed below.

4.1 Sol-gel route

(a) (b) (c)

Figure 4.1 – Photos of annealed films on glass. (a) Film of 1 layer spin coating sol-gel prepared film on FTO coated aluminoborosilicate glas annealed at 450^◦C. (b) ‘6 layer’-film on FTO coated aluminoborosilicate glas annealed at 450^◦C. (c) ‘6 layer’-film on FTO coated aluminoborosilicate glas annealed at 700^◦C.

In section 3.1.1 the method used to make oxide films on glass have been described. Some films in different stages of this process can be seen in figure 4.1. Figure 4.1a and 4.1b is a picture of a single layer precursor film and a ‘6-layer’ precursor film, respectively, on high temperature resistant aluminoborosilicate glass and annealed at 450^◦C. XRD measurements on such films showed no crystalline features. Only FTO can be seen in a diffractogram with an incident angle of 0.2^◦, which should have enhanced any peaks for crystalline phases of the topmost film relative to the FTO (the XRD diffractograms are found in appendix C, figure C.1). Such films also look very uniform in SEM, as can be seen (on ordinary soda lime glass with FTO) in figure 4.2. The thickness of a 6-layer film is around 400 nm as determined from SEM pictures showed in figure 4.3.

Figure 4.1c show a ‘6-layer’ film prepared as the films just described, and then annealed a second time at higher temperature in an atmosphere with low oxygen content. The tem- peratures tried here range from 500^◦C to 800^◦C. None of the samples showed any delafossite phase peaks in XRD measurements. Only CuO and CuFe2O4 could be identified in XRD measurements seen in figure 4.4.

The samples annealed at 700^◦C in 300 ml/ min argon flow — which is exactly the condi- tions used by Prévot et al. [6] — did not result in any delafossite phase in this study. XRD measurement results from an as prepared film on glass are shown in figure 4.5. Only Cu2O and maghemite Fe2O3 peaks are present in the diffractogram. The diffractogram peaks from

Figure 4.2 – SEM pictures of an amorphous film on FTO coated soda lime glass after 6 cycles of sequential spin coating and annealing at 450^◦C.

Figure 4.3– SEM cross section of an amorphous film deposited on FTO-glass by sequential spin coating and annealing at 450^◦C.

15 20 25 30 35 40 45 50 55 60 65 70 75 80

0 500 1,000 1,500 2,000

2θ(^◦)

A.u.

C4_575deg_Ar FTO_ref

CuFe2O4_34-0425 CuO_45-0937 SnO2tinoxide_41-1445

Figure 4.4 – Diffractogram from annealed sample in 575^◦C in an argon flow and a reference diffractogram from only FTO-glass. The furnace used could not measure the argon flow. Reference peaks are shown at the bottom with colours corresponding to each reference in the legend.

15 20 25 30 35 40 45 50 55 60 65 70 75 80 0

500 1,000 1,500 2,000

2θ(^◦)

A.u.

E1_on_glass_700deg_Ar Cu2O_05-0667

Fe2O3maghemite_39-1346

Figure 4.5 – Diffractograms from annealed samples in 700^◦C in an argon flow of 300 ml/min.

Reference peaks are shown at the bottom with colours corresponding to each reference in the legend.

maghemite Fe2O3 closely resembles them from magnetite Fe3O4 separated only by a small shift.

However, almost every peak in question are shifted in favour for maghemite, and the picture of the film in figure 4.1c is red — whereas magnetite is black — which should indicate that the phase is maghemite Fe2O3.

Annealing at even higher temperature, 800^◦C, resulted in maghemite Fe2O3 as the major phase and Cu2O peaks were no longer present in the diffractogram, which can be seen in figure 4.6.The ovens used here and by Prévot et al. [6] are both tube furnaces with adjustable gas flow and the same diameter. The amorphous precursor films seen herein and in [6] also seem similar. Possibly, the oven used in this work could differ from the one used by Prévot et al., not holding exactly the same temperature as measured by the oven thermometer or leaking gas. However, the oven is used frequently by others and errors with the oven does not seem likely. Further, Chen & Wu [29] vary the annealing parameters in a similar process with thin films prepared by spin coating and obtained varying relative amount of delafossite phase for temperatures under 700^◦C.

The non delafossite phases found for lower temperatures, CuFe2O4 and CuO, are the same as found earlier for similar temperatures [29]. Cu2O and Fe2O3, found for 700^◦C, have also been reported before [33, 49], though as a product of decomposition of CuFeO2 at 1180^◦C. As seen when comparing figure 4.5 and 4.6 for the sample annealed at 700^◦Con glass and on FTO coated glass respectively, the appearance of Cu2O correlating peaks compared to those of SnO2

and Fe2O3 seem to be prohibited by the FTO coating. This could indicate that the formation of crystalline Cu2O is less favourable at FTO glass than on bare glass, at least in the presence of Fe2O3.

Two samples similar to sample E2 were annealed at 700^◦C in nitrogen flow and at signifi- cantly lower flow rate, around 20 ml/min. These films were gray with white tints and showed traces of metallic tin in XRD measurements, seen in appendix C, figure C.3.

Whether this was caused by too high oxygen partial pressure or by the nitrogen is not inves- tigated in this work. Though Chen & Wu [29] use a nitrogen atmosphere to obtain delafossite.

However, this was on quartz substrates, and not on FTO coated glass. Thus the combination of FTO, Cu and Fe precursors and nitrogen cannot be excluded as a reason for the metallic tin formation.