Improved interconnect materials for next-generation Solid Oxide Fuel Cells

IN

DEGREE PROJECT MATERIALS DESIGN AND ENGINEERING, SECOND CYCLE, 30 CREDITS

STOCKHOLM SWEDEN 2019,

Improved interconnect materials for next-generation Solid Oxide Fuel Cells

PABLO COLLANTES

KTH ROYAL INSTITUTE OF TECHNOLOGY

SCHOOL OF INDUSTRIAL ENGINEERING AND MANAGEMENT

A BSTRACT

Solid Oxide Fuel Cells (SOFC) are attractive candidates in the search for cleaner and energy efficient production due to their numerous advantages such as fuel flexibility, modularity and exceptional efficiencies when combined heat and power is harnessed. A key element in its design are the interconnects which are mainly manufactured from custom ferritic stainless steels to carry the electricity between two adjacent cells. However, the high operation temperatures increase the chromia scale thickness in those steels, which reduces their conductivity. At the same time, chromium (VI) volatilization due to the wet atmosphere poisons the electrodes and reduces the cell life. Therefore, the narrow of the selection of suitable materials and high production costs have hindered their commercialization. Recent advances in lower temperature SOFC operation have opened a window for new interconnect materials and innovative processes. A Ce/Co nanocoating can be applied in the readily available AISI 441 ferritic steel to form a protective spinel oxide layer that reduces the effect of both degradations in the interconnects. The coating is applied in a continuous roll-to-roll process and then the interconnect shape is pressed in the material, manufactured as Sanergy HT 441 by Sandvik.

However, mechanical stresses cause microcracks that expose the substrate material, which can impact the oxidation behaviour negatively. Fortunately, pre-treatments can achieve the spinel to diffuse over short distances and combine with elements in the substrate, homogenizing the protective effect. This phenomenon known as self-healing has not been studied with sufficient depth for the Sanergy HT 441.

Thus, different series of temperature and short pre-treatments times were tested, and self-healing properties were observed by means of SEM surface characterization and chemical analysis. The results indicate that self-healing can be obtained within short times using isothermal pre-treatments at temperatures over 750 °C.

Keywords: SOFC, Interconnect, Self-healing properties, Corrosion, Cr Vaporization, Pre-treatment, Reactive Element

S AMMANFATTNING

Solid Oxide Fuel Cells (SOFC) är attraktiva kandidater i jakten på renare och energieffektiv produktion på grund av deras många fördelar som bränsleflexibilitet, modularitet och exceptionella effektivitet när kombinerad värme och kraft utnyttjas. Ett viktigt element i dess utformning är sammankopplingar som huvudsakligen tillverkas av anpassade ferritiska rostfria stål för att transportera elektricitet mellan två angränsande celler. De höga driftstemperaturerna ökar emellertid tjockleken på kromskalan i dessa stål, vilket minskar deras konduktivitet. Samtidigt förorenar krom (VI) på grund av den våta atmosfären elektroderna och reducerar celllivslängden. Därför har det smala urvalet av lämpliga material och höga produktionskostnader hindrat deras kommersialisering. De senaste framstegen inom SOFC-drift med lägre temperatur har öppnat ett fönster för nya sammankopplingsmaterial och innovativa processer. En Ce / Co-nanocoating kan appliceras i det lättillgängliga AISI 441 ferritstålet för att bilda ett skyddande spinelloxidskikt som minskar effekten av båda nedbrytningarna i sammankopplingarna. Beläggningen appliceras i en kontinuerlig rulle-till-rullningsprocess och sedan pressas sammankopplingsformen in i materialet, tillverkat som Sanergy HT 441 av Sandvik. Mekaniska spänningar orsakar emellertid mikrokrackar som exponerar underlagsmaterialet, vilket kan påverka oxidationsbeteendet negativt.

Lyckligtvis kan förbehandlingar uppnå att spinellen diffunderar över korta avstånd och kombineras med element i underlaget och homogeniserar den skyddande effekten. Detta fenomen som kallas självhelande har inte studerats med tillräckligt djup för Sanergy HT 441. Således testades olika serier av temperatur och korta förbehandlingstider, och självhelande egenskaper observerades med hjälp av SEM-ytkarakterisering och kemisk analys. Resultaten indikerar att självläkning kan uppnås inom korta tider med användning av isotermisk förbehandling vid temperaturer över 750 ° C.

P REFACE

I would like to thank my main supervisors Carlos Bernuy-López and Jörgen Westlinder for giving me the opportunity to write this thesis at Sandvik Materials Technology. From the beginning of the project, they have been a constant support in a professional and personal level. Considering this last element, a special mention goes to Laura Rioja as well: your happiness is contagious and your suggestions invaluable. Carlos and you have made me feel like at home, and for that I owe you my deepest gratitude.

I also appreciate the opportunity to be integrated with the team in the numerous meetings, conversations and even conferences. Sebastian, thanks for your patience to show me your work and its intricacies. Your sound technical knowledge and critical thinking added with your sharp humour was always a motivation for me. I truly hope we can keep in touch in the future. Anna, Claire, Robert, Mikael and Ulf: thank you for all the conversations we had together, I have always felt that I could confidently come to you with any problems and shortcomings and there would be time to discuss it. Very special thanks to Carlos again for your patience and understanding, and likewise to Jan Andersson, who dealt with all my requests and provided me with all the tools that made my experimental work possible. I enjoyed working on your side, and I learnt many things in the way. Last but not least, I would like to thank Mats Lundberg, my KTH and also industrial supervisor. Our discussions and the presentations with Carlos really helped me through the project.

Finally, I would like to express my most sincere gratitude to my parents, who always cheered me up through endless Skype calls and messages (and nice food care packages). Wiebke, you have a very special place in the list too, bearing with me in my darkest hours and providing me with endless support.

I came this far because of you, and it fills me with joy to have you by my side every day. Thank you so much.

L IST OF ACRONYMS

𝑉_𝑎𝑐𝑐 Acceleration Voltage AFC Alkaline Fuel Cell BSE Backscattered Electrons CVD Chemical Vapour Deposition

EDS Energy Dispersive X-Ray Spectroscopy IEA International Energy Agency

IUPAC International Union of Pure and Applied Chemistry MCFC Molten Carbonates Fuel Cell

PAFC Phosphoric Acid Fuel Cell PVD Physical Vapour Deposition PEMFC Polymeric Electrolyte Fuel Cell SEM Secondary Electron Microscopy SE Secondary Electrons

S/N Signal to noise SOFC Solid Oxide Fuel Cell

TEC Thermal Expansion Coefficient WD Working Distance

C ONTENTS

1 Introduction ... 1

1.1 Aim of the Thesis ... 3

1.2 Social impact ... 3

2 Theory of high temperature corrosion ... 5

2.1 Fuel Cells ... 5

2.2 Solid Oxide Fuel Cells ... 6

2.2.1 Electrolyte and electrode materials and manufacturing ...7

2.2.2 Interconnect materials ...8

2.3 Oxidation of metals ... 8

2.3.1 Thermodynamics ...9

2.3.2 Kinetics ... 10

2.4 Corrosion in metallic interconnects ... 12

2.4.1 Chromium evaporation ... 13

2.4.2 Dedicated SOFC interconnect steels design ... 13

2.4.3 Coatings to mitigate Cr evaporation ... 14

3 Materials and Methods: ... 15

3.1 Material and sample preparation ... 15

3.2 Pre-oxidation Exposures ... 17

3.3 Analytical Methods ... 19

3.3.1 SEM Analysis ... 19

3.3.2 Simulation techniques: CASINO V2.5 ... 22

4 Results ... 24

4.1 Cracks development analysis through SEM imaging (Secondary Electrons) ... 24

4.1.1 Isothermal treatments ... 24

4.1.2 Isochronous treatments ... 27

4.2 Oxide growth through SEM analysis (EDS) ... 30

4.2.1 Point EDS... 30

4.2.2 Chemical mapping EDS ... 34

5 Discussion ... 40

5.1 Influence of pre-oxidation treatment design on the self-healing properties ... 40

5.2 Chemical development of the self-healing properties ... 42

6 Summary and Outlook ... 45

7 References ... 47

8 Appendix ... 51

8.1 Ampliation of Figure 19: Surface Development in Isothermal treatments ... 51

8.2 Ampliation of Figure 26: Surface Development in Isochronous treatments ... 52

1 I NTRODUCTION

Cutting down greenhouse emissions is part of our most current pressing challenges. At the same time, the soaring demand for energy for a growing population and applications demands the development of new technologies for a cleaner and more effective energy production.

According to the International Energy Agency (IEA), our means to produce and use energy today are patently unsustainable [1]. Not only energy is produced with an increasing amount of emissions but also as the IEA estimated in 2016, around 1.1 billion people still need access to electricity, most of them living in rural areas [2]. It is from this understood that there is a need for decentralized and accessible production of clean energy. To that end, new sustainable technologies and systems in the roadmap have already been envisioned [1].

Fuel cells are part of this joint development since, among other things, they obtain electrical efficiencies up to 60% [3], which exceeds the limitations of conventional heat engines as they do not follow the Carnot cycle. If also heat is harnessed in the operation, this figure can be increased to 80% [4].

Moreover, they present other practicalities: the output energy is scalable and directly proportional to the number of cells in the stack. The operation is silent, clean and free from noxious combustion by- products (NOx, SOx).

During operation, fuel cells transform chemical energy stored in the fuel into electricity through a chemical reaction. The concept of associating chemical reactions with electricity was first demonstrated by Sir Humphry Davy in electrolysis experiments with aqueous solutions in 1800 [5]. Later in 1834, Faraday introduced fundamental concepts in electrochemistry and conduction theory, leading to the discovery of fuel cells by William Grove in 1839 [6]. First applications of an alkaline fuel cell system were found by Francis Bacon in 1959, and years later implemented in NASA Apollo missions in the late ‘60s.

[6] The development of different types of fuel cells and designs was paced with the introduction of new techniques such as vapour deposition or plasma spraying and materials in the cell.

Nowadays, Polymer Electrolyte Membrane Fuel Cells (PEMFCs) have the highest attention in scientific publications and it is the most mature technology. However, they are restricted to operate in low temperature (≈80 °C) with high purity hydrogen only, and deal with extra costs of the platinum catalyst [3] On the other hand, Solid Oxide Fuel Cells can use a variety of fuels such as biogas, natural gas or methanol and because of the high operation temperatures (>600 °C) they do not require a catalyst [7].

The components of any fuel cell are three: an anode, a cathode, and an electrolyte. The voltage that a single cell output is small for any conventional application (<1V). To increase the voltage, the cells are stacked in series. Each cell is separated and connected electrically to the next by an interconnect. [3]

The requirements for the interconnect material are high: it needs to withstand long operation cycles submitted to high-temperature corrosion on a dual atmosphere (both facing anode and cathode) while being impermeable to gases. Moreover, it must remain electrically conductive and have a similar Thermal Expansion Coefficient (TEC) as the adjacent electrodes [5].

In the early development of interconnect materials, Siemens-Westinghouse attempted to produce single cells depositing conductive ceramics based on Mg-doped Lanthanum Chromite by plasma spraying on tubular cell designs. Two main problems were observed. First, the sintering process was problematic as high densifications were difficult to obtain in oxidizing atmospheres. As a result, there were formed liquid phases that leaked in the electrode porous material. Additionally, the material increased its expansion in a hydrogen atmosphere, producing cracks in every element of the cell [5]

Metallic interconnects based on ferritic alloys and high chromium alloys have been developed to avoid those problems. The 𝑇𝐸𝐶 matches closely the one of the electrodes and electrolytes and with the benefit of higher toughness and strength. This has made possible to develop thinner planar (Sulzer design) interconnects enhancing the power density and consider new stack designs such as the microtubular design, evolved from the Siemens-Westinghouse. [5]

During the operation, chromium diffuses at high temperature and creates a thick passivating chromia (Cr2O3) scale layer of several microns. [3] However, in this process Cr (VI)species volatilize from the material, poisoning the cathode. At the same time, the contact resistance of the chromia layer decreases the power output. [5, 3, 7, 8]. Research has been conducted to minimize those effects optimizing the composition of the steel: avoiding Si and Al that producing non-conductive oxides in favour of adding stabilizing elements such as Nb or Ti and spinel forming oxides such as Mn have shown positive effects on the performance of the interconnects [7, 8, 5, 9].

A complementary action to mitigate the chromium vaporization and avoid high corrosion resistance is the implementation of protective coatings [10]. Surface layers of Reactive Element Oxides (REOs), rare earth perovskites and composite spinel oxides have also shown good properties to reduce oxide growth kinetics, increase oxide scale conductivity and improve oxide scale-to-metal adhesion [11]. The combination of composite layers of REOs (Ce, La, Y) and composite spinels of the Mn-Co system effectively reduce the high-temperature oxidation rate [12] while still retaining high conductivity and reducing chromium evaporation [13].

Spinel coatings are difficult to deposit on substrates with complex geometries due to their low toughness and thickness [14]. They would have to be applied after the flow patterns have been imprinted, meaning that each interconnect would be individually coated in a batch process. This increases the production times and the cost as extra machinery to produce the different flowing patterns is required. Instead, coatings such as Co-Ce can be deposited in metallic state with techniques such as Physical Vapor Deposition (PVD) [15] or electrodeposition [16] as a conversion coating in a continuous roll-to-roll process [14]. The metallic pre-coating withstands the forming operations without spalling [14]. After forming, the pre-coating is oxidized at high temperatures and reacts with the Mn diffusing from the steel, generating the desired final (Co,Mn)3O4 spinel coating [17, 18] over a seed layer of Ce.

It has been observed that the forming operations can crack the surface of the pre-coating exposing the substrate materials, eliminating the protection against Cr-evaporation. Nevertheless, previous studies have shown that the micrometre size cracks can be self-healed at high temperatures due to the instant volume expansion of the coating as it oxidizes [19]. In order to make SOFC cost-effective and commercially attractive, longer service lives of the materials are needed to decrease the cost per kWh.

An important cost lowering factor is process optimization and devise new operating conditions (temperatures and times) that can help to obtain durable materials in scalable production ways.

In the production of material for metallic interconnects, Sandvik Materials Technology (SMT) reconciles the need for a high production output at competitive prices by the design of in-line coating processes.

By this means, the number of steps to achieve the final product is greatly reduced, and due to the minimal manufacturing, the material can later be used in a variety of ways depending on the customer needs. However, increasing the flexibility of the possible applications comes at a cost. The target group of companies that can find a field of application for the material is potentially increased, and so are accordingly the customer demands for specific solutions. Not only it would be counterproductive from an industrial strategy standpoint to dedicate its actives to give tailored solutions for each and every case, but also it would defeat the main purpose to obtain a simpler process route. Moreover, due to confidentiality agreements, it is common for customers not to disclose to sensitive information concerning manufacturing processes and vice versa.

To bypass the limitations of development due to industrial secrecy, Sandvik Materials Technology performs research on the behaviour of the material under different conditions to add value and provide pre-treatment recommendations for the customer. Some research on the present material undergoing long term exposures at high temperatures has been performed in cooperation with Chalmers. However, not much information about the self-healing properties had been obtained for pre-oxidation treatments in relatively short times at lower temperatures. This was the starting point to start filling the existing gaps of knowledge and to settle a ground to conduct further investigations.

1.1

A

T

The purpose of this project is to obtain knowledge about the self-healing properties of the interconnect.

Prior to being introduced to the fuel cell stack, forming flow patterns on the interconnects can create cracks on the coated material in the in-line production. These cracks leave unexposed metal to react with the oxidizing atmosphere in the operation of the cell, producing chromia vaporization from the steel. The chromia poisons the active points of catalysis of the electrodes and shortens the cell life. At the same time, the steel oxidizes more rapidly, weakening the electrically conductive and mechanical properties.

Conversion coatings like Ce/Co produce a protective layer that effectively stops chromia evaporation by different mechanisms. With increased temperatures, some elements in the steel can be diffused upwards and form different oxides when combined with the elements in the coating. These oxides experience a volume expansion that is able to cover the exposed areas in the material where there is no coating present, protecting them in what is called the ‘self-healing’ effect. Although previous studies have shown that this mechanism is effective in pre-treatments at high temperature (>800°C), investigations about the effect of self-healing properties at medium to low temperatures are limited.

Starting from this point, the study aims to answer two inferential research questions proposed below.

From a manufacturing perspective, the design of more efficient and cost-effective pre-treatments could be achieved by obtaining knowledge about how the independent variables of temperature and time affect the progression of the self-healing properties. Hence, the first research question is: How do the combination of temperature or the time variables in the pre-oxidation treatments influence the achievement of self-healing? This was conducted by SEM image analysis of the selected samples before and after the treatments.

Although it is relevant for the self- healing process to observe that oxides are produced on the top layer, a more precise manner to quantify its degree is to characterize them by chemical analysis. Therefore, a second research question was proposed: Are the observed self-healing properties reflected also in microstructural and chemical changes? For this case, compositional EDS data by point and map techniques were used.

1.2

S

As it was stated in the introduction, it is a must to find alternative and sustainable ways to obtain and use energy. Nowadays it is already possible to produce great quantity of energy by renewable means, for example using solar and wind power which are considerably mature technologies. However, the effective use of that energy is hindered by a number of reasons. The intermittent nature of its production and the lack of distributing and mostly storing infrastructure limits its potential utilisation.

The introduction of the Hydrogen Economy and the possibility of using and storing hydrogen as fuel makes possible to bridge the gap between production and consumption with only small adaptations in the way we use energy today. Fuel cells play a key role in this development as they are efficient chemical conversion systems that transform the hydrogen in energy to fit the demand when its needed and vice

versa. This technological challenge is eventually a material challenge, as the design and optimization of more efficient, cost-effective and long-lasting fuel cells and its components requires intensive research.

2 T HEORY OF HIGH TEMPERATURE CORROSION

2.1

F

C

Fuel cells are energy conversion devices that can transform chemical energy directly into electricity and heat through coupled redox reactions. Because they are not limited by the Carnot efficiency, they offer a higher efficiency than combustion engines, especially if the heat generated is also harnessed [4].

Besides, they present additional benefits such as the possibility to produce decentralised energy in scalable systems, avoid greenhouse emissions during silent operation and the possibility to function on a constant fuel supply. There are only three essential components that are needed to put a cell together:

an ion-conducting electrolyte, two electrodes and an electrical connection circuit. When fuel is fed into the system, the reaction is produced in the electrodes. In the anode electrode, oxidation of the fuel is produced while the reduction of oxygen in air takes place in the cathode. The conduction of ionic species is made through the electrolyte and the current is transferred through an external circuit. Although this working principle is common for all fuel cells, several types of electrolytes have been developed to work with various kinds of fuel and operation temperature ranges to allow the mobility of the different ions as it can be seen in Figure 1.

Figure 1: Types of Fuel Cells and their mobile ions at working temperatures (drawn by the author): SOFC (Solid Oxide Fuel Cell), MCFC (Molten Carbonates Fuel Cell), PAFC (Phosphoric Acid Fuel Cell), PEMFC (Polymer

Electrolyte Fuel Cell), AFC (Alkaline Fuel Cell).

Fuel cells operating at lower temperatures (<100 °C) such as Proton-Exchange Membrane Fuel Cells (PEMFC) demand high diffusivity of ions which is only achievable with pure hydrogen and the use of expensive platinum catalysts. On the other hand, they can provide a faster start-up suitable for mobile purposes which has made them lead the development of this technology in the recent years. However, membranes that work at high temperatures (>600 °C) such as Solid Oxide Fuel Cells (SOFC) are not so strictly limited by the ionic conduction, allowing to use of different sources of fuel (from H2 to hydrocarbons such as diesel) without the need of catalysts. Different applications of this technology include stationary units in CHP plants, where long uninterrupted cycles and few start-up times are expected [5].

2.2

S

O

F

C

The name of SOFC stems from the use of an ion-conducting solid ceramic electrolyte. The high operation temperatures enable the oxygen ion conduction through the membrane from the cathode and oxidize several types of fuels in the anode (H2 , NH3 , ethanol, biogas, biodiesel). Figure 2 illustrates the SOFC working principle when H2 is used as fuel.

Figure 2: Working principle of a SOFC (drawn by the author)

The oxidation of H2 in the anode creates water vapour (Equation 1) and liberates electrons. Those are transferred to the corresponding O2 reduction semi reaction in the cathode to produce the oxygen ions (Equation 2) that travel through the membrane due to the potential gradient and can start the cycle again. The overall reaction in the cell is presented in Equation 3.

𝐴𝑛𝑜𝑑𝑒 𝑝𝑎𝑟𝑡𝑖𝑎𝑙 𝑟𝑒𝑎𝑐𝑡𝑖𝑜𝑛 𝐻₂ + 𝑂²⁻→ 𝐻₂𝑂 + 2𝑒⁻ (1)

𝐶𝑎𝑡ℎ𝑜𝑑𝑒 𝑝𝑎𝑟𝑡𝑖𝑎𝑙 𝑟𝑒𝑎𝑐𝑡𝑖𝑜𝑛 1

2𝑂₂+ 2𝑒⁻ → 𝑂^{2− (2)}

𝐺𝑙𝑜𝑏𝑎𝑙 𝑟𝑒𝑎𝑐𝑡𝑖𝑜𝑛 1

2𝑂₂+ 𝐻₂ → 𝐻₂𝑂 ⁽³⁾

During the operation of the SOFC, both ohmic resistance from heat production and irreversible mix of gases (when fuel meets the oxidant) reduce the ideal performance of the cell. Some of the heat produced in the SOFC is directly used for the conversion of different fuel streams into hydrogen and CO2, but the excess heat can be also retrieved to power heat engines or gas turbines [5]. Thus, near- reversible process efficiencies can be obtained as the SOFC when it is integrated into larger systems.

2.2.1 Electrolyte and electrode materials and manufacturing

The ceramic membranes and the electrodes in the SOFC are exposed to high temperatures (600 to 1000

°C) and must be designed to last operative for years with minimal degradation. These challenging operating conditions came with several demands. A high chemical stability is required in both oxidizing and reducing atmospheres in the electrodes. Additionally, good oxygen ionic conduction has to be ensured between both while keeping gas tightness to avoid contamination and insulating properties to electron currents. On top of that, mechanical and thermal stability are necessary to avoid stresses in all components. These requirements initiated a quest to find both better materials and production techniques. In the early years of ceramic processing, producing thin components by dry powder compaction did not yield good results and it was neither a cost-effective alternative The introduction new manufacturing techniques such as chemical vapour deposition (CVD), tape casting and extrusion processes allowed for better control of the dimensions, the density, thermal shock resistance and reduce substantially the production costs [5, 4, 10].

Among the favoured materials used for SOFC electrolytes are the fluorite structured Yttria-Stabilised Zirconia (YSZ) and Scandia Stabilized Zirconia (ScSZ). Other fluorite structure oxide ion conductors such as Gadolinium Doped Ceria (GDC) have been also developed for lower temperatures (600 to 800 °C) [20, 21]. In the recent years, perovskites such as Lanthanum Strontium Gallium Magnesium Oxide (La0.9Sr0.1Ga0.8Mg0.2O3) and brownmillerites like Barium Lanthanide Oxide (Ba2Ln2O5) materials have raised attention for their high conductivity, but they still lack the sufficient mechanical strength [22, 20, 5].

Contrarily to the electrolyte, the electrodes have to be porous to offer the most possible catalytic surface area. Perovskite oxides such as Strontium Doped Lanthanum Magnetite (LSM) and Lanthanum Strontium Cobalt Ferrite (LSCF) are also used in the cathode, which offer a good conductivity and high chemical resistance in oxidising environments. Good electrical conductivity paired with active catalytic properties are also key requirements for the anode. Since 1960’s nickel has been used for this purpose, as it combines good performance and economics but has a significant Thermal Expansion Coefficient (TEC) mismatch with the electrolyte. Ni-cermets of YSZ/Ni-oxide offer much better adhesion properties [5].

2.2.2 Interconnect materials

Individual cells need to be stacked in series for practical due to their low voltage output and the interconnect material functions as electrical connection between them (Figure 3).

Figure 3: a) SOFC planar stack design consisting of three cells (drawn by the author). b) A commercial 3 kW stack (E3000), manufactured by Elcogen OY, Finland, consisting of 119 cells separated and electrically connected with

interconnects [7].

Their main function is to separate the semi reactions in the cathode and the anode while allowing electrical connection between them. The required properties are a combination of the anode and cathode, limiting the selection to a few materials. Initially, ceramic materials such as doped rare earth chromites (La,Sr,Ca)(Cr,Mg)O3 were the only oxides that could satisfy such demands and operate above 800 °C to ensure good ionic conduction. However, their high brittleness and manufacturing cost made them a difficult alternative for mass production. Nevertheless, thanks to the development of electrolytes that can work in lower temperatures, the introduction of metallic materials such as stainless steels has been made possible. They feature a BCC structure due to high Cr additions (18-28%) and provides them a 𝑇𝐸𝐶 of 10 𝑥 10⁻⁶𝐾⁻¹ , which is very similar to the adjacent electrodes. Although their lower costs for mass-production with increased electrical, thermal and mechanical properties makes them attractive compared to the ceramic materials, their lifetime under SOFC conditions remains to be demonstrated [5, 23].

2.3

O

The International Union of Pure and Applied Chemistry (IUPAC) defines corrosion as “an irreversible interfacial reaction of a material (metal, ceramic, polymer) with its environment which results in consumption of the material or in dissolution into the material of a component of the environment.“ [24].

When applied to metals, reactions with the environment is a natural phenomenon to reach the stable equilibrium. It can also be regarded as the inverse action of extractive metallurgy, where more pure forms of metals are derived from their compounds [25]. As it can be understood, fighting the natural state of equilibrium takes effort, in form of economic losses from direct (maintenance and substitution

of the corroded elements) and indirect consequences of corrosion (stop the industrial operation, loss of equipment efficiency, contamination).

Many classifications exist for the corrosion processes but attending to the localization of the electrochemical reactions, generally it can be classified in electrochemical corrosion and high temperature corrosion (also called direct corrosion). Electrochemical corrosion requires the existence of an electrolyte, normally in solution and two differentiated reaction zones are formed. [24] The material is dissolved into the medium in the anodic zones, while the cathodic ones remain protected.

On the other hand, high temperature corrosion involves the relation between metals and their surrounding gaseous phase at high temperatures. No electrolyte is formed, and the reaction products are formed homogeneously on the surface of the material while both the electrons and ions are exchanged through the oxide layer. The corrosion resistance and the protective quality of such layer depends on the properties of the corrosion products. [26]

Both corrosion mechanisms achieve eventually the same outcome but in different ways. The most relevant case in this investigation is high temperature corrosion, as the oxide layer acts as the electrolyte for the ion and electron conduction and the temperatures reach 600-850 °C

2.3.1 Thermodynamics

The nature of corrosion as a chemical reaction implies its dependence with temperature and the activities of the species involved. At high temperatures, metals can react with the neighbouring oxygen to form oxides, carbides nitrides or other products as described by Equation 4

𝑥𝑀(𝑠) + ^𝑦₂𝑂

2(𝑔) ↔ 𝑀_𝑥𝑂_𝑦(𝑠). (4)

The spontaneity of the reaction at a particular temperature and pressure can be predicted by the Gibbs free energy of the system according to Equation 5

∆𝐺 = ∆𝐻 − 𝑇∆𝑆 (5)

where 𝐺 is the Gibbs energy, 𝐻 is the enthalpy, 𝑆 is the entropy and 𝑇 the temperature in Kelvin (K). If the delta of energy is negative (ΔG<0), the oxidation is thermodynamically spontaneous and vice-versa.

The definition of the Gibbs energy for the independent components of a multiphase reaction is a complex function of the chemical potentials of the possible species that can form in the process [25].

Leaving aside the intricacies of the thermodynamic description, the exchanged Gibbs energy of the reaction can be described with Equation 6.

∆𝐺 = ∆𝐺⁰+ 𝑅𝑇𝑙𝑛𝐾_𝑒𝑞 = ∆𝐺⁰+ 𝑅𝑇𝑙𝑛 𝑎_{𝑀𝑥𝑂𝑦}(𝑠) 𝑎_𝑀^𝑧(𝑠)𝑎_𝑂

2 𝑦 2 (𝑔)

(6)

Where 𝛥𝐺⁰ is the standard free energy change and 𝑅 is the gas constant. The ratio of the activities of products and reactants is incorporated in 𝐾𝑒𝑞 ,the equilibrium constant for the reaction. Considering that in this close system pure solids are used as the reactants and obtained as products, their activity is unity in the equilibrium constant. Therefore, the whole equation can be described as a variable of the oxygen pressure and temperature only (Equation 7). This is far more convenient when it comes to elaborate predictions of oxide stability, for instance using the Ellingham Diagram (Figure 4).

∆𝐺 = ∆𝐺⁰+ 𝑅𝑇𝑙𝑛 1 𝑝_𝑂

2 𝑦 2 (𝑔)

(7)

Figure 4: Ellingham-diagram for the formation of oxides based on their standard free energy of formation over temperature [27]

2.3.2 Kinetics

However, a material can be thermodynamically active and still not be corroded if the ambient conditions are not favourable. The different stages of growth and the types of corrosion products can affect the rate at which the material is oxidized. The growth mechanisms of the oxide scales can be fundamentally divided in three steps as described in Figure 5 below:

Figure 5: Schematic illustration of the Step 1: adsorption, Step 2: oxide nucleation and Step 3: continuous growth of an oxide scale. Drawn by the author.

Initially, the oxygen ions reach the surface of the bare metal by Van der Waal interaction, in what is known as the chemisorption process. In this process, the oxygen molecules are reduced and dissociate in their ions (O^2-). They bond with the charged cations (Mn²⁺) of the metal to form small oxide nuclei that diffuse across the surface, growing until it forms a stable scale. Consecutively, this oxide layer extends over the whole surface by growth coalescence of nuclei. At this point, the dominating mechanism for oxide growth is solid state diffusion of ions through the oxide scale [10]. The diffusion of metal cation ions or the oxygen ions have to happen in parallel, but the type of scale growth depends on which is the predominant diffusion process. A predominant metal diffusion defines an outward growth, while inward scale development is produced when the oxygen ions penetrate into the metal at a higher rate. The slower mechanism controls the process [28] The reaction continues until there is not a potential difference (driving force for the growth) to continue the process.

The rate and protective behaviour of the oxides is related with the ratio between oxide and metal volume. This can be quantified using the Pilling- Bedworth ratio (𝑃. 𝐵) in Equation 8:

𝑃. 𝐵. = 𝐴₀× 𝜌_𝑂 𝑎 × 𝐴₀× 𝜌_𝑀

(8)

Where 𝐴₀ is the molecular weight of the oxide, 𝑎 Is the valence of the metal and 𝜌_𝑂 and 𝜌_𝑀 are respectively the densities of the oxide and the metal. When 1 < 𝑃. 𝐵 < 2 , the behaviour of the oxide layer is protective. On the other hand, the oxides outside of that range are not protective, either being thin and porous (𝑃. 𝐵 < 1 ) or because they grow excessively and lose cohesivity with the metal (𝑃. 𝐵 > 2 ). Other important factors that can greatly affect the adhesiveness between metal and oxide while it grows are the 𝑇𝐸𝐶 matching, and the oxide plasticity at high temperatures. [29]

One of the most direct ways to measure the corrosion kinetics is thus the mass increase by unit of area as a function of time. This results in a series of known relationships from where the oxidation law can be derived fitting the experimental data into a theoretical curve.

When the oxidation is controlled by ion diffusion (Step 3) and the corrosion products are protective, a parabolic behaviour can be described according to the following Equation 9:

𝑊²= 𝐾1× 𝑡 + 𝐾2 (9)

Where 𝑊 is the increase of weight by unit of area and 𝑡 is the time. 𝐾1 and 𝐾2 are temperature- dependent coefficients. On the other hand, when the oxides spall off and they for porous structures that are not protective, this can be described by a linear behaviour as described by Equation 10:

𝑊 = 𝐾₃× 𝑡 (10)

𝐾₃ is also a constant in the equation. In this case, oxygen reacts directly with the unprotected metal surface. Additional mechanisms such as the logarithmic behaviour lie outside the reach of this study, as it takes place at relatively low temperatures ( < 400 °C). [30, 29]

Figure 6: Equations for different oxidation rates. Reproduced from [30]

The mechanism of high temperature corrosion in ferritic interconnects is a mixture of both parabolic and linear contribution. As it will be described in the following section, under specific circumstances the otherwise protective scales can spall off and the rapidly increase the oxidation rate. This is known as breakaway corrosion, and it can be detected as a first step of parabolic mass behaviour followed by a rapid linear phase. [3] This is the case when the interconnect cannot supply more protective oxide- forming atoms to the metal-oxide interface.

2.4

C

The recent research of the stability of the electrodes have turn the focus to the interconnects to guarantee operation lives of >10000 h [31]. The decreasing working temperatures have also made possible to simplify the interconnect manufacturing processes by using metallic materials [19, 5].

Nevertheless, a cost-effective production might be the only direct advantage: not only there are high mechanical requirements (creep resistance, 𝑇𝐸𝐶 matching), but also most challenges to sustain a long- term operation with these materials involve complex problems with corrosion behaviour. In first place, the interconnects need to remain gas-tight although they are continuously exposed to extremely oxidant and reducing conditions from both electrodes, which accelerates the corrosion rate.

Furthermore, the corrosion products have to remain conductive to ensure excellent electrical performance (< 10 mΩ · cm²) [31, 32]. However, there are not many choices for conductive oxides that can generate a continuous protective layer and match 𝑇𝐸𝐶 of the adjacent elements at high temperatures. High Cr stainless steels are used for their ability to grow a protective chromia layer. Cr is extensively used as an alloy element to form chromia (Cr2O3), which is several orders of magnitude

more conductive than the counterparts Alumina (Al2O3) or Silica (SiO2) [33, 32]. Still, the major drawback is that the volatilization of the corrosion products (Cr species) produce membrane and electrode poisoning reducing the interconnect mechanical properties and therefore, the cell life. This process is known as chromium evaporation.

2.4.1 Chromium evaporation

It is well established since the early 1990s that early decrease in power output was connected with Cr poisoning of the cathode while using chromia forming alloys at high temperature. Thermodynamically, Cr2O3 scales formed in dry atmosphere remain stable even at high temperatures (1000 °C) [7, 3].

However, due to the presence of humidity in air coming to the cathode, the breakaway corrosion is triggered and chromia forms hydroxides at much lower temperatures following Equation 11:

1

2𝐶𝑟₂𝑂₃(𝑠) + 𝐻₂𝑂 (𝑔) +3

4𝑂₂(𝑔) → 𝐶𝑟𝑂₂(𝑂𝐻)₂(𝑔) (11) The hydroxides contain volatile Cr (VI) species which are deposited in the cathode in form of Cr (III) or interact with the membrane and the rest of the cell components [21], eventually blocking the catalysis active points. The formation of the chromia scale in the interconnect acts as a coupled process, supplying Cr to the surface. Due to the high inlet flows of air, the kinetics are improved, and the reaction equilibrium reached either when the totality of Cr is evaporated from the scale, or the mass flow reaches the kinetic limits of the hydroxide forming [3, 26].

2.4.2 Dedicated SOFC interconnect steels design

The minimum Cr needed to obtain homogeneous passivation is around 18-22% in steels. However, to prevent breakaway corrosion higher Cr content is desirable (22- 25%) [31, 8]. Special chromium-based alloys have been engineered to offer the maximum performance at very high temperatures. The Oxide- Dispersion Strengthened (ODS) CFY alloy manufactured by Plansee is an extreme example: with ~95%

Cr, 5% Fe and Y in the composition, this allow matches the 10⁻⁶ 𝑇𝐸𝐶 of the membrane and holds excellent properties at 850 °C [34]. However, it has to be manufactured by powder metallurgy, and the extra processing steps can be as costly as the ceramic interconnects that it intends to replace [5]. More moderate ferritic steels such as Crofer 22 H and Sanergy HT (produced by SANDVIK AB) are commercially available. They both feature high Cr compositions, around 22%. [3] To prevent the formation of the insulating SiO2 phase, they are vacuum melted and added W, Nb or Mo that encapsulates the Si in Laves- phase precipitates. Additionally, to enhance the protective behaviour of the chromia scales, Mn is added. This element forms a spinel (Cr,Mn)3O4 that reduces the Cr evaporation [26, 5, 35]. The addition of reactive elements (Ce, La, or Y) in small proportion is known to improve the oxidation properties blocking the outward diffusion of the metallic cations in the grain boundaries [3, 36] which reduces the thickness of the chromia layer and thus the electrical resistance. Simultaneously, it enhances the metal- oxide adhesion, allegedly by its desulphurization effect. Lastly, the mechanical properties are affected by the changes of the microstructure due to grain refinement. [10]. Nonetheless, the chromium evaporation is not reduced by this procedure. With the reduction of the operation temperatures, there are less factors that justify the investment in custom-designed ferritic steels and point more in the direction of optimizing the properties of economical and readily available steels such as AISI 441 or AISI 430 for interconnect applications.

2.4.3 Coatings to mitigate Cr evaporation

As shown by Fontana et al. [18], the enhancement of the corrosion properties due to the reactive elements is also found when they are applied in a coating. Besides, the design flexibility of coatings broads the choice of surface chemistry and eases the obtention of an appropriate microstructure. The most common deposited materials are perovskites and spinels. Perovskite coatings, with a general ABO3

formula, can be also based on reactive elements such as Lanthanum Strontium Cobaltites (LSC) or Lanthanum Strontium Ferrites (LSF). They can be deposited by magnetron sputtering and provide good ionic conduction, aside from the oxidation resistance already described for reactive elements. However, alike them, they do not limit Cr evaporation successfully and they are extremely brittle [11, 5]. The state of the art of the coating chemistry are the spinel-based coatings, with a general formula of AB2O4 . Their doping capabilities allows them to reach a suitable 𝑇𝐸𝐶 with the ferritic substrate, while being electrically conductive and preventing the volatilization of Cr (VI) species. From this group, (Mn,Co)3O4

stands out as the more promising candidate. Unlike the perovskites, the spinel formers can be deposited as conversion coatings: the material is laid down in metallic state by electrodeposition or PVD and then is het treated to obtain it into the final microstructure. When Co is deposited, it rapidly oxidizes into a Co3O4 layer, and if the temperature is sufficiently high, upwards Mn diffusion from the steel forms the (Mn,Co)3O4 layer as studied by Falk-Windisch [3] and Grolig [7]. Furthermore, Falk-Windisch also proved that regardless of the amount of Mn in the spinel, a 640 nm Co-coating can still block the Cr volatilization by a factor of 10 at 850°C [3, 37]. Compared to the direct deposit of the spinel, the conversion coatings provide better control of the thickness, homogeneity, porosity and they allow for plastic deformations such as forming before the heat treatments is performed [14, 13]. When reactive element coatings are combined with spinel coatings the material equals or outperforms the custom-designed steels regarding chromium evaporation [3]. As an added advantage, the electrical properties and the oxidation resistance are greatly improved.

3 M ATERIALS AND M ETHODS :

Based on the structure of research questions proposed in the Aim, the study follows an experimental study design according to the methodological research design proposed by Creswell [38] .

To address the first descriptive research question 1) How do the combination of temperature or the time variables in the pre-oxidation treatments influence the achievement of self-healing? and the second inferential research question 2) Are the observed self-healing properties reflected also in microstructural and chemical changes? a pre-experimental pre-test-post-test approach was adopted. First, several samples of the material were distributed in randomized series without a control group. Secondly, they were tested in sequential treatments and the results were compared before and after them. Because different variables such as temperature and time were expected to change the outcome of the experiment, a factorial comparison was used to examine their individual and simultaneous effect within the series and with the rest. Details about the formal structure of the study are described below.

3.1

M

The material for analysis was a ferritic stainless steel Sandvik Sanergy® HT 441, which incorporates a conversion coating of Co/Ce. This material has been developed from the AISI 441 stainless steel to perform in high-temperature applications and it has been specially designed to be used as an interconnect material in SOFC. The version used for the study is a single coated applied on a single side.

The composition is listed in the following Table 1 [7]:

Table 1: Substrate material composition

Fe Cr C Si Mn Nb Ti P Ni

AISI 441 wt% bal. 17.83 0.012 0.55 0.26 0.48 0.14 0.024 0.13

The coating tested on the material is described in Table 2 [7]. It was deposited in metallic state using a Physical Vapor Deposition roll-to-roll process (PVD) in Sandvik Material Technology AB.

Table 2: Investigated coating system on the substrate material

Sample designation Inner Coating Outer Coating

Ce/Co 10 nm cerium 600 nm cobalt

To enhance the gas distribution of the interconnects, flow patterns are imprinted on the surface of their coated surface. The large deformations that the material withstands in the creation of this channels causes severe mechanical stresses. Figure 7 shows the creation of the three distinct areas on the surface of a planar interconnect after its forming process:

Figure 7: Cross section of a planar coated steel interconnect after formed. The top (T), slope (S) and valley (V) topological features are represented.

The stress condition where most cracks are formed is in tension, as it can be seen in the top (T) part of the interconnect flow patterns. However, small variations in the selected area during the analysis can produce misrepresentative information due to the tilting towards the slope (S). A forming technique that could show a representative amount and size of cracks, but homogeneously distributed over a flat area would be more convenient for analysis. Figure 8a shows a conventional deep drawing Erichsen test [39]. The standard application of the test would produce a curved dome and render the same problems as with the formed samples. However, a modified Erichsen test with a cylindrical die could yield a much more ideal flat surface at the dome.

Prior to this study, Sandvik Sanergy® 441 HT samples were obtained in the form of 20 cm² sheets by 0.3 mm in thickness and they were submitted to the modified Erichsen test. An initial circumference of 5 cm in diameter was marked on top of the surface, and the sheet was punched until a 12% biaxial deformation was obtained following Equation 12.

𝜀 = ln (𝐷2

𝐷1) ⁽¹²⁾

Figure 8b and c present the initial state of the material in the study. A visual inspection of the material did not reveal any macroscopic cracking or coating spallation due to the deformation process.

Figure 8: Schematic of the a) original Erichsen cupping test, b) the resulting sample from the coating side and c) the same sample from the steel side.

As it is indicated in Figure 9, first the flat circular area in the dome was cut out to obtain a square of 4 cm². From that piece of material, smaller samples of 1 cm² were obtained. Previous research done on the same material showed that to able to make any qualitative analysis of the self-healing properties, it was necessary to track the same area [19]. Therefore, the samples were marked for identification

drawing an arrow at the top right of the sample with the tip of a diamond scriber. The 0.5 µm wide scratch mark allowed to define an analysis area consistently through the analysis.

Figure 9: Sample obtention and marking

The samples 1 cm² samples were stored randomly in different plastic trays and were labelled subsequently for the different series of treatments.

3.2

P

-

E

There has not been conducted extensive research about the response of Sandvik Sanergy® 441 HT to pre-oxidation treatments. Thus, information was needed for a vast array of the possible manufacturing conditions depending on time and temperature presented in Table 3. In order to meet the end of the experimental part within the time limitations of the project, a preselection step of the treatments was performed. The selected pre-oxidation treatments are indicated in the outlined boxes in the table, in total dwell times.

The treatments were grouped in two experimental sets to simulate different manufacturing conditions, isothermal and isochronous. Three series of isothermal pre-oxidation treatments and another three isochronous were designed. A sample was assigned to each of the series. To be able to study the properties in the same area, the experiments had to be performed alternating analysis and treatment.

In latter stages of the experimental work, a refinement of the initial matrix was performed introducing two new isothermal series. The objective was to validate the corrosion state observed in the discontinuous series and to cover a broader area in the lower temperature region. A new sample was introduced in each of the series, one at 650 °C and other at 750 °C. The sample in the lower temperature series was tested in two exposures, 5 and 24 hours respectively. The other one at higher temperature was submitted to a single exposure of 24 hours. Both series were conducted in identical experimental conditions as the rest of isothermal series described below. This new set of experiments is outlined in blue in the matrix.

Table 3: Pre-oxidation treatment matrix

T/t AS FORMED 1h 2h 5h 12h 24h 48h

600 A0 A1 A2 A3 A4 A5 A6

650 B0 B1 B2 B3 B4 B5 B6

700 C0 C1 C2 C3 C4 C5 C6

750 D0 D1 D2 D3 D4 D5 D6

800 E0 E1 E2 E3 E4 E5 E6

850 F0 F1 F2 F3 F4 F5 F6

A Nabertherm programmable box furnace unit with a lever door was used. The furnace had no circulation and ambient air was used. The samples were held on a ceramic boat on the bottom of the furnace, with the coated face facing upside. The initial and final temperature for the exposures was always below 250 °C. For every treatment, the heating and cooling ramps were set to 5 °C/min, though the actual time was influenced due to the natural cooling of the furnace. A Type K thermocouple connected to a Data Logger TC-08 was installed next to the sample to performed on-line temperature measurements.

By reading the matrix horizontally, three different series of isothermal treatments at 600, 750 and 850

°C were designed. The total dwell time at the same temperature was considered in every stage of the series, so the time for each individual exposure was calculated from the sum of the previous. An illustrative diagram of the different exposures is presented in Figure 10 below.

Figure 10: Schematic of the 3 series of isothermal treatments

Similarly, three isochronous treatments were designed at 1, 5 and 24 hours reading the matrix vertically.

Figure 11 illustrates the design of the exposure series. In this case, the exposures were performed in a separate and smaller box furnace unit with a Eurotherm 2416 multi-programmer. The ambient in the furnace, heat cycle and the positioning of the sample were identical as in the isothermal set.The variable factor was the increasing temperatures. Same as in the isothermal exposures, each treatment in the series was given the same total dwell length at the corresponding temperature. That results for individual dwell times to be the same length as the total dwell. The reason to apply this criterion was to be able to compare the final stages of the treatments in the matrix for both sets. To be able to distinguish

them from the isothermal samples in the matrix nomenclature, a T was added before the number as it can be seen in Figure 11. For example, the states that correspond to 5 hours of treatment at 750 °C are the sample D3 in isothermal series and the D3T in isochronous.

Figure 11: Schematic of the 3 series of isochronous treatments

A Type K thermocouple connected to a Data Logger TC-08 was used in both furnace units to obtain online measurements of the experimental temperature profile. The thermocouple tip was situated directly above the sample in both furnaces, trying to obtain the most similar position as it was possible.

Some deviations from the designed treatment were observed in both units. The unit used in isothermal experiments was found to consistently overshoot the temperature around 20 °C regardless of the designed treatment temperature. On the other hand, the smaller furnace unit used in the isochronous set presented a sustained negative trend to hold the temperature 10 °C lower on average through the treatments. Additionally, it presented a more linear cooling pattern closer to the introduced parameters.

3.3

A

M

The analysis process and the exposures were conducted in alternate stages synchronously. Scanning Electron Microscopy (SEM) was used for both imaging and chemical analysis techniques. Secondary Electrons (SE) were used for top-down surface characterization, while Point and Map Energy Dispersive Spectroscopy (EDS) was employed in the chemical analysis part. Prior to experimental analysis, the simulation software Casino V2.5 [40] was used to model and obtain suitable parameters for the study.

3.3.1 SEM Analysis

Scanning electron microscopy (SEM) is a useful imaging tool for microstructural surface characterization of metals and their corrosion behaviour. With a much shorter wavelength than conventional light microscopy, the SEM emits an accelerated focused beam of electrons in vacuum which interacts with the electrons in the sample. Different signal detectors produce images by continuous scanning of both the elastic and inelastic interactions of the electrons coming from the sample. The types of interactions detected from electrons with the sample can be seen in Figure 12:

Figure 12: Different types of interaction of the incoming electrons with the sample atoms [8].

• Secondary electrons (SE) are produced from the emission of the valence electron of the atoms in the specimen as they are inelastically scattered by the incident beam. The energy of these electrons is very small, so only those generated at the surface (<2 nm below the surface of the sample) [3] of the specimen can be emitted outside. Because of the high surface sensitivity, they provide the most information about the topological information with high resolution. [26, 41]

• Backscattered electrons (BSE) are elastically reflected by the sample atoms depending on the 𝑉_𝑎𝑐𝑐 and the specimen type. Their higher energy increases the penetration depth and the interaction volume into the sample, lowering the resolution. Given that they interact with the nuclei of the atoms in the material, they can provide compositional information in what is known as Z-contrast. The heavier elements deflect more electrons back to the signal detector and they appear to be brighter. This can be used to give information about the distribution of chemical elements in it and the texture of the phases. [26, 41]

• Photons with energies in the range of Characteristic X-rays are detectable through Energy Dispersive X-Ray Spectroscopy (EDS). This radiation is when an electron from an outer shell fills the hole that is produced in an inner shell of the atom as a result of a collision of an incoming very energetic electron. The energy of the photon is characteristic of the material and is released with a specific intensity depending on the difference in the energy states between the hole and the filling electron. The precision of this technique increases the greater Z is, not being able to quantify elements with Z <11 (C, N, O) even though they can be detected. However, as the generation of characteristic X-Rays takes place deepest in the interaction volume (Figure 13), the resolution is poor and can only be increased with long analysis times improving the statistics. [26, 41]

• Photons with energies in the visible range of the spectrum are resulting from the recombination of electron-hole pairs that have previously been created by the incidence of the electron beam on the sample. They allow for the analysis of the local composition, identification of band structures and details about the sample phases growth. [26, 41]

Figure 13: Interaction volume of the incoming electron beam with the sample [7]

3.3.1.1 High-resolution imaging in SEM

A Sigma EVO50 VP scanning electron microscope has been used in this research. It features a LaB6

Thermal Emission (TE) filament type. It is equipped with detectors for secondary electrons, backscattered electrons and transmitted electrons as well as microanalysis system by energy dispersion Oxford Inca, infrared camera and 5 motorized axes. It features a variable pressure high vacuum system as well with an antipollution system of the sample and a theoretical resolution of 2nm (30 keV) [42]

The heated cathode emits electrons that have acquired energy over the work function through vacuum and reach the sample surface which acts as the anode. The emission is dependent on the 𝑉_𝑎𝑐𝑐 between the anode and cathode. The intensity of the probe current can be modified to achieve higher brightness.

Although it is a basic physical mechanism, it has some disadvantages: the source emitting area (10-30 µm) and the energy spread can be unstable because of the thermal energy supplied by heating, reaching a few electron volts. [43]

For a higher brightness, resolution and control over the energy spread, field emission guns (FEG) should be used. A Schottky field emission gun or a cold emission gun offer three orders of magnitude higher brightness and a spread of just a few tenths of electron volts from the nanometric tip. On the other hand, FEGs have associated costs because of the high vacuum pumps, which must keep around 10⁻⁸ to 10⁻⁹ Pa to keep the tip clean and avoid electric discharge. [43]

Independently of the emission type, the electron beam must be focused into the sample with a system of magnetic lenses as it was mentioned before. A magnetic field generated from two coils deflects the pathway of the incident beam and travels along the yoke, which is the gap in their axis. Later designs of SEMs incorporate ‘in-lens’ detectors in the yoke that allow for higher resolution as more electrons are collected and the aberrations are minimized decreasing the focal length. However, the physical design of this technique restricts the specimen size due to the extremely close working distance (WD). [44].

The optical imperfections of this otherwise simple technique arise from the fact that it is impossible to focus parallel electrons to a point in the focal plane to produce the maximum resolution without experiencing aberrations. The most common are spherical and chromatic aberrations, astigmatism and diffraction errors. Spherical and chromatic aberrations depend on the focal length used and the convergence angle of specific wavelengths. The diffraction is visible when small apertures are used, limiting the size of the lens. Last, astigmatism affects the symmetry of the focal length along the lens axis. Corrections for one type of aberration are generally a compromise to the others [43]. Selecting the

optimal working conditions for high-resolution imaging is as important as incorporating correction mechanisms like monochromators or multipoles. [44]

The minimum probe size that yields the maximum resolution can be expressed as a function of the aberrations with this mathematical expression [43]:

𝑑_𝑝= √(𝑑_𝑖)²+ (0.5𝐶_𝑠𝛼³)²+ [0.43∆𝐸 𝐸₀𝐶_𝑠𝛼]

2

+ [1.22𝜆 𝛼 ]

2 (13)

The first term (𝑑_𝑖) in Equation 13 is the size of the electron source demagnified by the convergent and objective lenses. The second and third are depending on the spherical aberration 𝐶_𝑠, which are related to the WD linearly. The fourth corresponds to the diffraction error and the change in aperture.

Furthermore, the ratio ^∆𝐸_𝐸

0 is the energy spread of the electron gun over the energy of incident electrons, 𝛼 is the convergence angle and 𝜆 is the electron wavelength. [43]. As shown in the equation, it is not trivial to achieve a high resolution with the TE SEM as the size of the electron source is dependent on the tip size. Reducing the first term 𝑑𝑖 makes the current very small and the signal to noise (S/N) ratio decreases. Likewise, if the energy spread ∆𝐸 in the third term is higher even at low 𝐸₀, making imaging at low voltages where chromatic aberration becomes problematic. [43]

There are multiple ways to overcome these problems, but they are not exempt from limitations. If a small aperture is selected, the convergence angle decreases, and the diffraction error and spherical aberration are minimized. On the other hand, this has a similar effect as 𝑑𝑖 on the S/N ratio. Another variable that can be optimized for a better resolution is the WD, as reducing the focal length decreases spherical aberration. Finally, the energy of incident electrons 𝐸0 also plays an important role beyond obtaining a good resolution. From Equation 13, it is seen that a high 𝑉_𝑎𝑐𝑐 (large 𝐸₀ and small 𝜆 ) contributes to a smaller probe size decreasing the chromatic aberration and the diffraction error. The higher 𝑉_𝑎𝑐𝑐limits the ∆𝐸, increasing the stability of the beam [43].

Nevertheless, a better resolution is not the only parameter that has to be considered when finding optimal imaging conditions. As can be seen in Figure 13, the trajectories of the electrons and the interaction with the sample define the interaction volume. The extension of this region is primarily dependent on the energy of the incoming electrons as well as the composition of the material. The greater the energy input and lighter the atomic weight of the material, the bigger the interaction volume and vice versa [43].

3.3.2 Simulation techniques: CASINO V2.5

The commercially available modelling software Casino (version 2.5), based on Monte Carlo simulation, allows calculating the electron trajectories and their interactions in the material using different physical models [40]. It also provides two-dimensional displays of the interaction volume for electrons of different energies, and the intensity of characteristic X-Rays emitted by the different elements as a function of depth.

The material was modelled in Casino software (version 2.5) using a three-layer composite: a substrate layer of ferritic steel (AISI 441) with a 10 nm intermediate Ce coating and a 600 nm Co3O4 top coating to recreate the analysis conditions after the oxidation treatments. The simulation was run for 20000

electrons in energies between 5 to 20 keV in increments of 5keV (Figure 14). The physical models and the probe size were taken as default.

The results from the simulation predicted that interaction volume would be sufficient to analyse the coating and the substrate at 15 to 20 keV by BSE (blue lines). However, the X-rays (red lines in Figure 14) start carrying chemical information from the substrate at 20 keV. As it was explained in 3.1, the interaction of the SE for surface analysis is very superficial, but X-rays penetrate deep into the material to extract chemical information. Selecting 20 keV as 𝑉_𝑎𝑐𝑐 the surface can still be inspected with high resolution. At the same time, chemical information from the coating and the top part of the substrate (about a hundred nm) is visible. This way, it is possible to tell if the self-healed parts of the coating can cover the exposed cracked parts by EDS mapping.

a) b)

c) d)

Figure 14: Simulated interaction volumes with different acceleration voltages in Casino Software V2.0: a) 5 keV b) 10 keV c) 15 keV d) 20 keV. Red lines represent BSE and blue lines X-rays.