Development and Investigation of an Efficient Electrolysis Process for the Conversion of Carbon Dioxide to Formate Von der Fakultät Chemie der Universität Stuttgart zur Erlangung der Würde eines Doktors der Naturwissenschaften (Dr. rer. nat.) genehmigte Abhandlung Vorgelegt von Armin Löwe aus Kirchheim unter Teck Hauptberichter: Prof. Dr.-Ing. Elias Klemm Mitberichter: Prof. Dr. rer. nat. Andreas Friedrich Prüfungsvorsitzender: Prof. Dr. rer. nat. Bernhard Hauer Tag der mündlichen Prüfung: 01.03.2021 Institut für Technische Chemie der Universität Stuttgart 2021 Die Verantwortung für den Inhalt dieser Veröffentlichung liegt beim Autor. The author is responsible for the content of this publication. Erklärung über die Eigenständigkeit der Dissertation Ich versichere, dass ich die vorliegende Arbeit mit dem Titel „Development and Investigation of an Efficient Electrolysis Process for the Conversion of Carbon Dioxide to Formate“ selbständig verfasst und keine anderen als die angegebenen Quellen und Hilfsmittel benutzt habe; aus fremden Quellen entnommene Passagen und Gedanken sind als solche kenntlich gemacht. Declaration of Authorship I hereby certify that the dissertation entitled „Development and Investigation of an Efficient Electrolysis Process for the Conversion of Carbon Dioxide to Formate“ is entirely my own work except where otherwise indicated. Passages and ideas from other sources have been clearly marked. Stuttgart, September 11th, 2020 Place, date Signature - Armin Löwe Acknowledgment My greatest gratitude goes to Prof. Dr.-Ing. Elias Klemm for providing me the opportunity to work on this exceptional interesting topic that I hope will accompany my future career. Special thanks also for allowing me to pursue own ideas, providing a great working environment, imparting knowledge and for the detailed corrections on the present work. I want to extend my gratitude also to Dennis Kopljar who introduced me to this, at that time, com- pletely unknown topic. Also, to Fabian Bienen. Thanks to the both of you for the large number of fruitful discussions, helpful tips, for assisting with preparations and, Dennis, your great work I could start from. Thanks a lot to the whole institute, but most importantly to the electrochemical reaction engineer- ing group, for the great teamwork, wonderful discussions no matter day or night, awesome working environment and a lot of fun. Furthermore I want to thank all colleagues for the conducted measurements. Heike Fingerle for trying so hard to measure our unwilling samples. Barbara Gehring, Dennis Beierlein and Jan Florenski for measuring all those TGA samples and Efi Hadjixenophontos for taking the TEM images. To all colleagues, i’d like to thank you for the wonderful time at this institute! You live your life by a code, an ethos. Every man does. It’s your shoreline. It’s what guides you home. And trust me, you’re always trying to get home. - Act of Valor, 2012 Dedicated to my parents. Contents 1 Abstract 1 2 Zusammenfassung 5 3 Introduction 9 4 Theoretical Background 13 4.1 Electrochemical basics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 4.1.1 The electrical double layer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14 4.1.2 Cell voltage and electrode potentials . . . . . . . . . . . . . . . . . . . . . . . . 16 4.1.3 Kinetic aspects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18 4.1.4 Quantitative relations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24 4.1.5 Important aspects of electrochemical cells . . . . . . . . . . . . . . . . . . . . . 25 4.2 Carbon dioxide reduction reaction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27 4.2.1 Fundamentals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27 4.2.2 Equilibrium potentials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29 4.2.3 Electrocatalytic conversion of carbon dioxide . . . . . . . . . . . . . . . . . . . 31 4.2.4 Influence of operating conditions . . . . . . . . . . . . . . . . . . . . . . . . . . 37 4.3 Gas diffusion electrodes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43 4.3.1 Wetting behavior . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46 4.3.2 Mass transport in gas diffusion electrodes . . . . . . . . . . . . . . . . . . . . . 49 4.3.3 Equivalent circuit model of porous electrodes . . . . . . . . . . . . . . . . . . . 50 4.4 Process concepts for carbon dioxide electrolysis . . . . . . . . . . . . . . . . . . . . . . 53 4.4.1 Aspects to consider for alkaline reaction conditions . . . . . . . . . . . . . . . . 53 4.4.2 Coupling with downstream bipolar membrane electrodialysis . . . . . . . . . . . 54 Contents 4.4.3 Alternative anode reactions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56 5 Motivation and Objectives 59 6 Experimental Section 63 6.1 Preparation methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63 6.1.1 Homogeneous precipitation of supported metal oxides . . . . . . . . . . . . . . 63 6.1.2 Oxidative pretreatment of the Vulcan XC 72 support . . . . . . . . . . . . . . . 65 6.1.3 Synthesis of supported palladium nanoparticles . . . . . . . . . . . . . . . . . . 65 6.1.4 Preparation of gas diffusion electrodes . . . . . . . . . . . . . . . . . . . . . . . 65 6.2 Experimental setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67 6.2.1 Details for the semi-batch mode of operation . . . . . . . . . . . . . . . . . . . 68 6.2.2 Details for the continuous mode of operation . . . . . . . . . . . . . . . . . . . 68 6.2.3 Used and developed electrochemical cells . . . . . . . . . . . . . . . . . . . . . 70 6.3 Analytical methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75 6.3.1 Quantification of gaseous products . . . . . . . . . . . . . . . . . . . . . . . . . 75 6.3.2 Quantification of dissolved products . . . . . . . . . . . . . . . . . . . . . . . . 75 6.3.3 Catalyst and electrode characterization . . . . . . . . . . . . . . . . . . . . . . 76 6.4 Electroanalytical methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79 6.4.1 Determination of electrode activity and wetting . . . . . . . . . . . . . . . . . . 79 6.4.2 Determination of faradaic and energetic efficiencies . . . . . . . . . . . . . . . . 82 6.4.3 Estimations on the catalyst’s oxidation state . . . . . . . . . . . . . . . . . . . . 84 6.5 Data evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85 6.5.1 Analysis of semi-batch operation . . . . . . . . . . . . . . . . . . . . . . . . . . 85 6.5.2 Analysis of continuous operation . . . . . . . . . . . . . . . . . . . . . . . . . . 87 6.5.3 Double layer capacitance and activity evaluation . . . . . . . . . . . . . . . . . 88 6.6 Fundamental investigations on new process concepts . . . . . . . . . . . . . . . . . . . 89 6.6.1 Bipolar membrane electrodialysis for downstream processing . . . . . . . . . . 89 6.6.2 Selective alcohol oxidation as alternative anode reaction . . . . . . . . . . . . . 93 7 Results and Discussion 97 7.1 Reproducibility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99 7.1.1 Reproducibility of galvanostatic experiments . . . . . . . . . . . . . . . . . . . 99 7.1.2 Reproducibility of polarization curves and double layer capacitance . . . . . . . 102 II Contents 7.2 Investigations on the catalyst . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105 7.2.1 Type of catalyst . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105 7.2.2 Synthesis parameters for tin oxide based catalysts . . . . . . . . . . . . . . . . . 110 7.2.3 Tin oxide loading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111 7.3 Investigations on process conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117 7.3.1 The effect of temperature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117 7.3.2 Effects of the electrolyte . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125 7.3.3 Requirements for reactant and electrolyte purity . . . . . . . . . . . . . . . . . . 137 7.4 Further optimization of the electrode’s matrix . . . . . . . . . . . . . . . . . . . . . . . 143 7.4.1 Influence of preparation parameters . . . . . . . . . . . . . . . . . . . . . . . . 143 7.4.2 Fine tuning the electrode’s hydrophobicity . . . . . . . . . . . . . . . . . . . . . 143 7.5 Merging single parameter optimizations . . . . . . . . . . . . . . . . . . . . . . . . . . 147 7.6 Temporal behavior . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149 7.6.1 Long-term stability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149 7.6.2 Transient behavior . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154 7.7 Scale-up . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161 7.7.1 Scale-up of the catalyst synthesis . . . . . . . . . . . . . . . . . . . . . . . . . 161 7.7.2 Scale-up of the electrode preparation . . . . . . . . . . . . . . . . . . . . . . . 161 7.7.3 Scale-up of the electrochemical cell . . . . . . . . . . . . . . . . . . . . . . . . 163 7.8 Process concepts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171 7.8.1 Coupling carbon dioxide electrolysis with bipolar membrane electrodialysis . . . 171 7.8.2 Coupling carbon dioxide reduction with selective alcohol oxidation . . . . . . . 179 8 Conclusion and Outlook 183 8.1 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183 8.2 Outlook . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187 9 Appendix 211 9.1 Carbon dioxide reduction reaction - Standard potential calculation . . . . . . . . . . . . 213 9.2 Preparation methods - Additional derivations . . . . . . . . . . . . . . . . . . . . . . . 215 9.2.1 Preparation of the supported catalyst . . . . . . . . . . . . . . . . . . . . . . . . 215 9.2.2 Preparation of gas diffusion electrodes . . . . . . . . . . . . . . . . . . . . . . . 216 9.3 Electroanalytical methods - Supporting explanations . . . . . . . . . . . . . . . . . . . 217 9.4 Investigations on the catalyst . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220 III Contents 9.4.1 Type of catalyst . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220 9.4.2 Tin oxide loading - Estimations on the quantity of metallic tin . . . . . . . . . . 222 9.5 Investigations on process conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223 9.5.1 The effect of temperature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223 9.5.2 Concentration effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 224 9.5.3 Electrolyte purity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 225 9.6 Temporal behavior - Long-term stability . . . . . . . . . . . . . . . . . . . . . . . . . . 227 9.7 Process concepts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 228 9.7.1 Operating carbon dioxide electrolysis in the recycled product mixture . . . . . . 228 9.7.2 Coupling carbon dioxide reduction with selective alcohol oxidation . . . . . . . 229 IV Contents Listings Table 1: List of abbreviations. Abbreviation Description AB acetylene black ABS acrylonitrile butadiene styrene ACS American Chemical Society AEM anion exchange membrane AOR alcohol oxidation reaction AsB angle selective backscattered electron ATR-IR attenuated total reflection infrared spectroscopy ATR-SEIRAS attenuated total reflectance surface-enhanced infrared absorption spectroscopy BPED bipolar membrane electrodialysis BPM bipolar membrane CB carbon black CDL double layer capacitance CE counter electrode CEq chemical equation CEM cation exchange membrane CI current interrupt CL catalytic layer CO carbon monoxide CO2EL CO2 electrolysis CO2RR CO2 reduction reaction CV cyclic voltammetry DC direct current DD double-distilled DFAFC direct formic acid fuel cell DFFC direct formate fuel cell V Contents Abbreviation Description DFT density functional theory DI deionized DLR German Aerospace Center, Stuttgart, Germany (dt. Deutsches Zentrum für Luft- und Raumfahrt) EC electrical connection ECE energetic cathode efficiency ECSA electrochemical active surface area EDL electrical double layer EE energy efficiency EIS electrochemical impedance spectroscopy FA formic acid CA carbonic acid FE faradaic efficiency FID flame ionization detector GC gas chromatopraphy GDE gas diffusion electrode GDL gas diffusion layer GHG green house gas GLS gas-liquid separator GWP global warming potential HER hydrogen evolution reaction HPLC high performance liquid chromatography ICP-OES inductively coupled plasma optical emission spectroscopy IHP inner Helmholtz plane ITC Institute of Technical Chemistry, University of Stuttgart, Germany KESD polyethylene[polystyrenesulfonic acid-co-divinylbenzene] LSV linear sweep voltammetry MFC mass flow controller MFM mass flow meter VI Contents Abbreviation Description MOR methanol oxidation reaction MPI Max Planck institute, Stuttgart, Germany NBR nitrile butadiene rubber OC outer connection OCP open circuit potential OER oxygen evolution reaction OHP outer Helmholtz plane ORR oxygen reduction reaction p.a. proanalysis PEEK polyether ether ketone PEM polymer electrolyte membrane ph. Eur. European Pharmacopoeia (fr. Pharmacopoeia Europaea) PMMA poly(methyl methacrylate) PTFE poly(tetrafluoroethylene) RDE rotating disk electrode RE reference electrode RID refractive index detector SDS sodium dodecyl sulfate SEM scanning electron microscopy SHE standard hydrogen electrode TCD thermal conductivity detector TEM transmission electron microscopy TGA thermogravimetric analysis VBA Excel’s visual basic for applications VX Vulcan XC 72 WE working electrode XAS X-ray absorption spectroscopy XES X-ray emission spectroscopy XRD X-ray diffraction VII Contents Table 2: List of symbols. Symbol Description Unit α activity coefficient - αs symmetry factor - ∆fHθ standard formation enthalpy kJ mol−1 ∆RGθ standard Gibb’s free energy kJ mol−1 ∆RHθ standard reaction enthalpy kJ mol−1 ∆RSθ standard reaction entropy J mol−1 K Sθ standard entropy J mol−1 K η overpotential V κ electrical conductivity S m−1 λ wavelength m ρ density g m−3 ρr resistivity Ω m σ loading g m−2 Φ volume fraction - ϕ electric potential V ϕH Helmoltz layer potential V ϕp porosity - A area m2 c concentration M C capacitance F m−2 geo Cg conversion factor - D diffusion coefficient m2 s−1 d distance m d̄ mean diameter m E electrical energy W s ECE energetic cathode efficiency % EE energetic efficiency % VIII Contents Symbol Description Unit FE faradaic efficiency % f CO2 excess factor - fmixture mixing factor - I current A j current density A m−2 k reaction rate constant s−1 km mass transfer coefficient s−1 l length m m mass g M molar mass g mol−1 n molar amount of substance mol p pressure Pa P electric power W q electric charge C R resistance Ω r reaction rate mol s−1 S solubility mol L T temperature ◦C t time s V voltage V Vf volume L V̇F volumetric flow rate of converted CO2 according to Faraday’s law L s−1 V̇f volumetric flow rate L s−1 Vm molar volume L mol−1 v stoichiometric factor - w mass fraction - z number of transferred electrons - IX Contents Table 3: List of indices. Index Description Ω ohmic A anode A spec. area specific ad adsorbed AM anode to membrane AR anode to reference electrode aq solved in water C cathode CB carbon black CM cathode to membrane CR cathode to reference electrode CT charge transfer diff diffusion DL double layer eff effective el electrolyte F Faraday FA formic acid X Contents Index Description g gaseous geo geometrical GSE galvanostatic electrolysis intr intrusion IR IR compensated lim limiting M membrane max maximum Me metal min minimum mix mixed ox oxidation/oxidized/oxidized form p particle prec precursor red reduction/reduced form res residue ZC zero charge Table 4: List of constants. Constant Value Unit Description F 96485.3 A s mol−1 Faraday constant R 8.314 J K−1 mol Gas constant XI Contents XII 1 | Abstract To date, the very basis of the vast majority of chemical products, fuels and energy is fossil based. Espe- cially chemical industry is often bound to such resources due to a lack of alternatives or for economical reasons. In order to be able to move from fossil fuel consumption towards a more sustainable, carbon neutral supply new processes need to be established. This includes not only replacing carbon dioxide (CO2) emitting ones, but also developing such processes that use this green house gas as carbon resource. Besides thermocatalytically activated conversion processes, the electrochemical reduction of CO2 to value-added products recently gained a lot of interest. The basic idea is to convert CO2 in highly concentrated exhaust gas streams with renewable electric power. Depending on the applied catalyst different products including formate/formic acid, oxalic acid, methane, light alcohols, carbon monoxide or synthesis gas are reported. All these products represent platform chemicals for numerous upstream processes. The present work focuses on the development of an energy efficient electrolysis process that converts CO2 into formate salts or formic acid at high current density. In contrast to the huge amount of research that focuses on catalyst development, systematic variations of reaction conditions like temperature, gas feed purity or electrolyte composition are far less reported. They were investigated with the aim of enhancing the energetic efficiency of formate formation. The thereby used gas diffusion electrode (GDE) is also a unique type of electrode that is barely reported for CO2 reduction reaction (CO2RR) to formate, which was further optimized to reach, to date, unreported current densities of up to −1500 mA cm−2. The focus was thereby not only set to maximize said current densities but also to allow for industrial relevant lifetimes. The obtained insights were then used to scale up the cell from 1 cm2 to 10 cm2 and finally to 70 cm2 of geometrical electrode surface. Regarding the reaction conditions, an unexpected dependency of the limiting current density on tem- perature was found. In contrast to literature, which consistently suggests lower operating temperatures to enhance CO2 solubility, an optimum performance at 50 ◦C was observed. This value can be explained by the contrary development of CO2 solubility and diffusivity with temperature leading to a maximum Chapter 1. Abstract in mass transport at said temperature. Experimental data and calculations from a simple model support this explanation. On top of that, the used type of GDE shows a limitation in product mass transport for certain conditions that, to the best of the author’s knowledge, has not been reported yet. One of the view notable influences of the used electrolyte on CO2RR was attributed to the applied cation. Alkali metal ions with a large ionic radius increased the electrode’s activity significantly and allowed for a longer lifetime in the straightforward order of Na+ < K+ < Rb+ < Cs+. However, due to relatively high material costs for salts other than sodium based ones, cation recovery needs to be established in a final process. Highly concentrated mixtures of formate and bicarbonate salts were found to be unsuitable, mainly due to a strong product mass transport limitation and excessive CO2 gas formation inside the GDE. Both aspects cause a strong shift of selectivity in favor of hydrogen formation. On the other hand, the type of anion or other specifications like pH (7 to 14) and salt concentration (0.1 M to 2 M) showed no notable influence. Instead of using a typical multi-layered GDE design with a more distinct spatial separation of gas and liquid phase, a homogeneous single-layer design was used. A bimodal pore structure of specific hydrophobicity allows for an efficient CO2 transport through open macropores towards a large number of liquid filled pores. The latter are formed by carbon agglomerates bearing the electrocatalyst, which was investigated regarding the type of metal/metal oxide, loading and synthesis method. Tin oxide was confirmed to be the material of highest energetic efficiency, even though there are more active or selective alternatives. Due to the potentioselective character of tin oxide and a loading dependent deactivation, a tin loading of 0.5 mg cm−2 geo was found to be best suited for long-term operation, while a higher load- ing of 7.5 mg cm−2 geo allows for the highest activity in short-term operation. Deactivation effects were investigated even though the causes could not be identified or quantified with absolute certainty. During short-term operation the optimized cathode side can be operated with as little as −620 mV of IR com- pensated overpotential at a current density of −200 mA cm−2. This represents a top value compared to literature (up to August 2019). Partly exchanging the poly-(tetrafluoroethylene) (PTFE) binder with Nafion significantly increased the limiting current density. It is suspected that a local increase in hydrophilicity results in an extended amount of wetted agglomerates while the highly hydrophobic gas transport pores remain open. A GDE matrix composition of 15:20:65 Nafion:PTFE:Acetylene Black was found to be optimal. Combining this matrix with an increased electrode loading of 2 mg cm−2 geo and the best performing electrolyte ([CsCl] = 2 M, pH = 10) yielded a current density of −1500 mA cm−2 at 79 % of formate selectivity and −900 mV of IR compensated overpotential. 2 During long-term operation, long diffusion pathways for the formed product, reduction of the elec- trocatalytically active species and a mechanical degradation of the GDE were found to limit the time on stream. Target-oriented adjustments in the electrode’s composition prolonged the lifetime from 24 h to 250 h. Furthermore, an observed shift from CO2 reduction to hydrogen formation is caused by reduction of the electrocatalytically active tin oxide to metallic tin. Here, the selectivity was partly recovered by periodically applying a slight oxidative polarization. The gained insights were used to perform an energy efficient scale-up of the cell with completely revised geometry and components. A filter press type cell with minimized electrode distance, effective electric contacting and oxygen bubble removal was designed and fabricated. It reaches a total cell voltage of 2.7 V to 3.0 V at a current density of 200 mA cm−2 and a total current of 14 A (active electrode area = 70 cm2). This corresponds to an energy efficiency of 26 % to 23 % and again represents a top value compared to literature (up to August 2019). Finally, new process concepts are examined on a basic level. First, a bipolar membrane electrodialy- sis (BPED) cell is investigated for potential coupling with CO2 electrolysis (CO2EL). Such process can be used for downstream protonation of formed formate, product purification and concentration, as well as for metal hydroxide regeneration. Results will demonstrate its general functionality but, until now, dialysis is limited to low formic acid concentrations of 0.5 M to 1 M. Trying to assess higher concentra- tions than that results in high diffusion losses of uncharged formic acid through adjoining membranes. Second, selective alcohol oxidation is evaluated for its potential to replace the hitherto used oxygen evolution reaction (OER). Using a nickel mesh anode yielded an anodic formate formation with 86 % selectivity at −200 mA cm−2. 3 Chapter 1. Abstract 4 2 | Zusammenfassung Bis heute basiert der Großteil der Produktion von chemischen Produkten, Kraftstoffen und Wärme auf fossilen Rohstoffen. Speziell die chemische Industrie ist dabei oft, aufgrund mangelnder Alternativen oder aus ökonomischen Gründen, auf solch fossile Rohstoffe angewiesen. Um in der Lage zu sein auf eine nachhaltigere, kohlenstoffneutrale Produktion umschwenken zu können, müssen neue Prozesse eta- bliert werden. Dies beinhaltet nicht nur den Ersatz von Kohlendioxid (CO2) emittierenden Prozessen, sondern auch die Entwicklung von solchen, die CO2 als Kohlenstoffquelle nutzen. Neben thermokatalytisch aktivierten Umwandlungsprozessen hat die elektrochemische Reduktion von CO2 zu höherwertigen Produkten kürzlich deutlich an Interesse gewonnen. Die grundlegende Idee ist, CO2 aus hochkonzentrierten Abgasströmen mittels Strom aus erneuerbaren Energien umzusetzen. Abhängig vom verwendeten Katalysator sind dabei unterschiedliche Produkte zugänglich, wie zum Bei- spiel Formiat/Ameisensäure, Oxalsäure, Methan, kurzkettige Alkohole, Kohlenmonoxid oder Synthese- gas. All diese Produkte stellen Plattformchemikalien für zahlreiche weiterführende Prozesse dar. Die vorliegende Arbeit befasst sich mit der Entwicklung eines energieeffizienten Elektrolysepro- zesses, der CO2 zu Formiatsalzen, bzw. ferner Ameisensäure, bei hohen Stromdichten umwandelt. Im Gegensatz zu den zahlreichen Forschungsarbeiten, die sich mit der Entwicklung neuer Katalysatoren beschäftigen, sind systematische Untersuchungen von Reaktionsbedingungen, wie Temperatur, CO2- Reinheit oder Elektrolytzusammensetzung, deutlich weniger beschrieben. Solche Reaktionsbedingungen wurden mit dem Ziel untersucht, die Energieeffizienz der Formiat/Ameisensäureproduktion zu steigern. Die dabei verwendeten Gasdiffusionselektroden (GDEs) bilden einen einzigartigen Typ Elektrode, der bisher für die CO2 Reduktionsreaktion (CO2RR) zu Formiat kaum berichtet und im Laufe dieser Arbeit weiter optimiert wurde. Damit wurden bisher unveröffentlichte Stromdichten von bis zu −1500 mA cm−2 erreicht. Der Fokus der Entwicklung wurde dabei nicht nur auf die Maximierung der Stromdichte ge- legt, sondern auch darauf industriell relevante Standzeiten zu erreichen. Die gewonnenen Erkenntnisse wurde dann verwendet, um die Elektrolysezelle schrittweise von 1 cm2 auf 70 cm2 an geometrischer Kathodenoberfläche hochzuskalieren. Kapitel 2. Zusammenfassung Im Hinblick auf die untersuchten Reaktionsbedingungen wurde eine unerwartet starke Abhängigkeit der limitierenden Stromdichte von der Temperatur beobachtet. Im Gegensatz zur Literatur, welche kon- sistent möglichst geringe Temperaturen empfiehlt um die CO2 Löslichkeit zu erhöhen, wurde eine opti- male Temperatur von 50 ◦C bestimmt. Dieser Wert kann durch die entgegengesetzten Abhängigkeiten der CO2 Löslichkeit und des CO2 Diffusionskoeffizienten von der Temperatur erklärt werden, was zu einem Maximum an CO2 Massentransport bei der genannten Temperatur führt. Experimentelle Daten und Be- rechnungen aus einem einfachen Modell unterstützen diese Erklärung. Zudem zeigt der verwendete Typ GDE für bestimmte Reaktionsbedingungen eine Produkt-Massentransportlimitierung welche, nach dem besten Wissen des Autors, für diese Reaktion bisher noch nicht beschrieben wurde. Einer der wenigen merklichen Einflüsse des Elektrolyten auf die CO2RR bezieht sich auf die Wahl des Kations. Alkali- metallionen mit großem Ionenradius erhöhten die Elektrodenaktivität signifikant und erlaubten deutlich längere Standzeiten in der Reihenfolge Na+ < K+ < Rb+ < Cs+. Allerdings müsste durch die in analo- ger Reihenfolge ebenso ansteigenden Materialkosten eine Kationenrückgewinnung im finalen Prozess stattfinden. Hochkonzentrierte Mischungen von Formiat- und Bicarbonatsalzen wurden als ungeeignet bewertet, da sie zu einer starken Produktdiffussionslimitierung und übermäßiger CO2-Gasbildung inner- halb der GDE führen. Beide Aspekte bewirken eine Verschiebung der Selektivität hin zu einer verstärkten Wasserstoffbildung. Hingegen zeigten andere Arten von Anionen, sowie Unterschiede im pH Wert (7 - 14) oder der Salzkonzentration (0.1 M - 2 M) keinen nennenswerten Einfluss. Im Gegensatz zu den typischerweise verwendeten, mehrschichtigen GDEs, welche eine deutlichere räumliche Trennung von Gas- und Flüssigphase zeigen, wurden in der vorliegenden Arbeit homogene, einschichtige GDEs verwendet. Eine bimodale Porenstruktur mit spezifischer Hydrophobizität ermög- licht dabei eine starke Verzahnung von Gastransport- und flüssigkeitsgefüllten Poren. Letztere werden durch Kohlenstoffagglomerate geformt die den Elektrokatalysator tragen. Dieser wurde hinsichtlich des verwendeten Metalls/Metalloxids, der Beladung und seiner Synthesemethode untersucht. Zinnoxid wur- de als der Elektrokatalysator mit der höchsten energetischen Effizienz bestätigt, obwohl es entweder ak- tivere oder selektivere Alternativen gäbe. Aufgrund des potentioselektiven Charakters von Zinnoxid und einer beladungsabhängigen Deaktivierung ergab sich eine Zinnbeladung von 0.5 mg cm−2 geo als bestgeeig- net für den Langzeitbetrieb. Hingegen zeigte eine Zinnbeladung von 7.5 mg cm−2 geo die höchste Aktivität im Kurzzeitbetrieb von nur 30 min. Auftretende Deaktivierungseffekte wurden untersucht, wobei die zu- grundeliegenden Effekte nicht mit absoluter Sicherheit identifiziert oder gar quantifiziert werden konn- ten. Im Kurzzeitbetrieb kann die optimierte GDE mit nur −620 mV an IR-kompensiertem Überpotential bei einer Stromdichte von −200 mA cm−2 betrieben werden, was einem im Vergleich zur vorhandenen Literatur (Stand August 2019) Spitzenwert entspricht. 6 Ein teilweiser Austausch des verwendeten Polytetrafluorethylen (PTFE) Bindemittels durch Nafion ergab eine deutliche Steigerung der limitierenden Stromdichte. Es wird vermutet, dass eine Erhöhung der lokalen Hydrophilie die Anzahl benetzter Agglomerate steigert, ohne dabei die nötigen CO2 Transport- poren zu fluten. Eine Matrixzusammensetzung von 15:20:65 an Nafion:PTFE:Acetylene Black wurde dabei als optimal befunden. Die Kombination dieser optimierten Matrix mit einer erhöhten Beladung von 2 mg cm−2 geo und der Verwendung des besten Elektrolyts ([CsCl] = 2 M, pH = 10) ermöglichte eine Stromdichte von −1500 mA cm−2 bei 79 % Formiat Selektivität und nur −900 mV an IR-kompensiertem Überpotential. Im Langzeitbetrieb wurden die langen Diffusionswege für das Produkt Formiat, eine Reduktion des aktiven Metalloxides und eine mechanische Degradation der GDE als limitierende Faktoren für die ma- ximale Betriebsdauer festgestellt. Eine Anpassungen der Elektrodenzusammensetzung verlängerte die Betriebsdauer dabei von 24 h auf 250 h. Weiter wurde festgestellt, dass die Reduktion des Metalloxides für eine Verschiebung des Produktspektrums weg von CO2RR Produkten hin zur Wasserstoffbildung verantwortlich ist. Hier konnte die Selektivität durch Regeneration des Katalysators mittels periodischer Rückoxidation über geringe anodische Polarisierung teilweise wieder hergestellt werden. Die während dieser Arbeit gewonnen Erkenntnisse wurden schließlich dazu verwendet, die ver- wendete Elektrolysezelle unter Berücksichtigung der maximalen Energieeffizienz hochzuskalieren. Eine komplett überarbeitete Geometrie im Filter-Press-Stil mit optimierten Komponenten erlaubt eine deut- liche Minimierung des Elektrodenabstands, die effiziente elektrische Kontaktierung, sowie eine verbes- serte Sauerstoffabfuhr. Die Zelle erreicht dabei Zellspannungen zwischen 2.7 V und 3 V bei einer Strom- dichte von 200 mA cm−2 und einem Gesamtstrom von 14 A (geometrische Elektrodenfläche = 70 cm2). Dies bedeutet eine Energieeffizienz zwischen 26 % und 23 % und entspricht im Vergleich zur bisherigen Literatur (Stand August 2019) ebenso einem Spitzenwert. Letztlich wurden neue Prozesskonzepte als Proof-of-Concept Studien untersucht. Dazu zählt zum einen eine mögliche Kopplung von CO2-Elektrolyse und Bipolarmembran Elektrodialyse. Solch ein nachgeschalteter Dialyseprozess kann dazu verwendet werden in der Elektrolyse gebildetes Formiat zu protonieren, Produktströme aufzureinigen und anzureichern, sowie verbrauchtes Alkalimetallhydroxid zurückzugewinnen. Die generelle Funktion der Bipolarmembran Elektrodialyse wurde bestätigt, ist aber noch durch die zur Verfügung stehenden Membranen und die damit verbundene hohe Diffusionsrate der Ameisensäure auf geringe Ameisensäurekonzentrationen von 0.5 M bis 1 M begrenzt. Zum anderen wurde die selektive Alkoholoxidation als Anodenreaktion bzw. als Ersatz für die bisher verwendete Sau- erstoffbildungsreaktion untersucht. Unter Verwendung eines Nickelkatalysators wurde anodenseitig die Bildung von Formiat mit einer Selektivität von 86 % bei einer Stromstärke von 200 mA cm−2 beobachtet. 7 Kapitel 2. Zusammenfassung 8 3 | Introduction Within the last hundred years consumption of fossil-based resources rose in an exponential manner. To date, chemical industry, electricity production and the transportation sector are the biggest consumers.[1] The related emission of CO2 is thereby well known to be a major cause of climate change, induced by excessive green house gas (GHG) emission. Even though CO2 has the lowest global warming potential (GWP) of all GHGs defined in the Kyoto protocol, its sheer mass of emission makes it the major cause.[2] Compared to the year 1990, atmospheric CO2 concentration in 2018 has already risen by around 16 % to 407 ppm.[3] As a consequence, average global temperature increased by around 0.5 ◦C compared to 1990 and by around 1 ◦C compared to 1950 with effects already being noticeable.[4] For the last decades, awareness for the excessive emission of green house gases like CO2 strongly rose, not only among scientists but also in the public and among business stakeholders. Interest for and emphasis on products being produced sustainably increased along with that. However, in light of the current fossil-based industry, it becomes obvious that new sustainable and competitive pathways need to be developed. A carbon neutral economy should not only be addressed by reducing the amount of CO2 emissions but also by utilizing this GHG as a feedstock to produce value-added substances. At the same time, renewable energy sources have a continuously rising share in the electricity production.[5] However, their fluctuating nature leads to periods of excess and deficit, which limits their share for a stable power supply. Thus, renewable energies require large-scale storage solutions. A promising and intensively studied pathway addressing both, CO2 conversion and energy storage, is the electrochemical reduction of CO2 using excess electricity. The obtained products can be stored and used for time-delayed power recovery e.g. in fuel cells.[6] Different products including formic acid, alcohols and light aliphatic hydrocarbons are reported that can suit this purpose. Among multiple products that are accessible by electrochemical reduction of CO2, CO (or synthesis gas) and formic acid have been shown to be producible with industrial relevant selectivity, energy effi- ciency, current density and long-term stability.[7] Both products are produced by a two-electron transfer process, which explains their easy accessibility. While CO or synthesis gas are of specific interest for Chapter 3. Introduction synthetic fuel production via the Fischer-Tropsch process, applications for formic acid are of broader range. Besides its use in non-chemical industries such as dyeing, textile and leather processing,[8] it could be used as fuel[6] or feedstock for further upgrading[9]. Formic acid and formate are both described as fuels utilized in a direct formic acid or direct formate fuel cell (DFAFC or DFFC). However, some major issues within such devices prevail. In DFAFC CO is a hardly suppressed side product which deactivates the noble metal catalyst. Also, the high volumetric energy density of concentrated formic acid comes along with the downside of a strong corrosivity. In DFFC, corrosivity of the fuel and CO evolution are widely eliminated but the low energy density of an aqueous formate solution limits its application to stationary ones.[6] More promising would be a downstream upgrading step to higher hydrocarbons. Formatotrophic microbes have been shown to be able to convert formic acid into a variety of different fuels, chemicals or even long chain molecules. To name a few examples ethanol, isobutene, propane, acetone, fatty acids or proteins are reported.[9–11] For applications in chemical industry, formic acid can be used as save and transportable source of carbon monoxide and/or hydrogen e.g. for carbonylation,[12] hydrogenation[13] or hydrocarboxylation[14] reactions. Also, formic acid can be decomposed to either hydrogen or CO, depending on the catalyst and reaction conditions used.[15, 16] This fact makes it a safe and transportable precursor for decentralized hydrogen, synthesis gas or CO production. Electrochemical reduction of CO2 to various products has recently experienced a boost in attention. Most of the investigations are focusing on the catalyst itself with all kinds of concepts, including met- al/metal oxide or alloy based systems, organometallic or single-site and structured catalysts.[17, 18] Even though some catalyst systems have been shown to exhibit high activity and/or selectivity, their limited stability or material costs often exclude them from industrial applications.[17] The high overpotential necessary for CO2RR at sufficiently high current densities remains challenging with energy efficiencies barely exceeding 30 %. So far, simple nano-structuring of earth abundant metals like tin seems the best option for industrialization.[17] Another major subject of investigation is the transfer to electrode systems that allows access to in- dustrial relevant current densities. Gas diffusion electrodes (GDEs) must be mentioned as the by far most common system for this task. Their optimization is of highly complex nature and requires controlling the electrode’s composition, while considering multiple aspects of reaction conditions. With such opti- mized electrodes, results can then be transferred into continuous mode of operation and into scaled-up electrolyzers.[17–19] 10 11 Chapter 3. Introduction 12 4 | Theoretical Background Within this chapter important basics of electrochemistry, CO2 reduction reaction (CO2RR) and gas dif- fusion electrodes (GDEs) are discussed. The electrochemical background includes basic concepts of electrified solid/liquid interfaces and respective descriptions of thermodynamics, kinetics and quantita- tive relationships. In contrast to these long-known fundamentals, CO2RR with application of GDEs is a rather novel field of research. Literature results regarding important parameters like electrode composi- tion or reaction conditions are therefore summarized. Finally, a short outlook into a conceivable process concept and alternative anode reactions are provided. 4.1 Electrochemical basics Electrochemical reactions in general can be divided into two main classes. One are those in a galvanic element, being spontaneous redox reactions to provide electrical energy. Most known examples here are batteries and fuel cells. The second class are reactions of redox active compounds, which consume electric power and produce substances of higher energy content. Excluding batteries, the respective electrochemical cell is then called an electrolyzer. Characteristic for both classes is the local separation of reduction and oxidation reaction, as well as of electron and mass transport, enabling the storage or supply of electric power. The present work focuses on the second class to store electrical energy in form of formate/formic acid. Therefore, no detailed discussion of galvanic elements shall be given here. During electrolysis endergonic redox reactions are forced to proceed by an external direct current (DC). As for this fact, the standard Gibb’s free energy (∆RGθ ) is positive and the standard cell voltage is defined as being negative. For most applications such an electrochemical cell is composed of two parts, an anode and a cathode side, often divided by a separator (see figure 4.1). Electrons are conducted through an outer circuit from the anode, where the oxidation reaction takes place (positively polarized), to the cathode side, where the reduction reaction takes place (negatively polarized). A separator in between allows charge balancing while keeping anolyte and catholyte apart to avoid contamination or Chapter 4. Theoretical Background re-oxidizing and re-reduction of formed products. In general, three different types of separators are available. While anion exchange membranes (AEMs) and cation exchange membranes (CEMs) are selective to the respective ion, diaphragms allow an unspecific transport.[20] C at h o d e (- ) A n o d e (+ ) Power source Ion transfer e- [ox1] [red1] e-[ox2] [red2] e-e- Figure 4.1: Schematic illustration of an electrolysis cell. One of the seemingly easiest examples is electrolytic water splitting under acidic conditions. The overall reaction, represented by the chemical equation (CEq), (CEq 4.3) is separated into an anodic ox- idation (CEq 4.1) and a cathodic reduction (CEq 4.2). At the positively polarized anode water is split into molecular oxygen, protons and electrons. While the produced oxygen remains in or can be released from the anode compartment, electrons are transferred through the outer electrical circuit to the nega- tively polarized cathode. Produced protons can move through the CEM or diaphragm and are reduced to hydrogen. Reaction (CEq 4.3) has a ∆RGθ of −237.13 kJ mol−1, which corresponds to a standard cell voltage of −1.23 V.[20] Oxidation: 2H2O O2 +4H++4e− (CEq 4.1) Reduction: 4H++4e− 2H2 (CEq 4.2) Overall reaction: 2H2O O2 +2H2 (CEq 4.3) 4.1.1 The electrical double layer Whenever an electronic conductor comes into contact with an ionic conductor redox reactions take place, forming a potential difference at the phase boundary. This applies to the two electrodes of an electro- chemical cell, which are both in contact with the ion conducting electrolyte. These two solid-liquid interfaces are of special importance as they represent the location where the electrochemical conversion takes place. Considering a metal electrode in contact with an electrolyte the isotropic character of the bulk liquid phase is interrupted, inducing an orientation of solvent dipoles and charged species by inter- 14 4.1. Electrochemical basics actions with the solid phase. This orientation of dipoles and charges results in an electrification of the electrolyte close to the phase boundary. In addition, the mobile electrons in the solid metal phase react to this electrification by moving towards the boundary. The formed parallel layers of ions in the solid phase and counter ions in the liquid phase are the origin of the term electrical double layer (EDL).[21] This separation of charges results in a measurable potential difference ϕ . During electrolysis the potential difference is enforced by the outer circuit electronics, controlling the excess charge in the solid phase. An illustration of this electrified interface and the corresponding course of electric potential is given in figure 4.2. Since the solid electrode is usually highly conductive it cannot sustain an electric field inside itself. Any excess charge is therefore residing with maximum distance to other electrons at the surface. The first liquid phase mono-layer in contact with the charged electrode surface is mainly consisting of solvent dipole molecules, while the second mono-layer is occupied by solvated ions of opposite charge. The latter is called outer Helmholtz plane (OHP). In this simplest model, introduced by and named after H. v. Helmholtz, the amount of charge in the OHP is equal to the excess charge of the solid electrode and the absolute value of potential decreases linearly to zero (model a) in figure 4.2). This structure of two parallel planes of opposite charge can be seen as a classical plate capacitor. The Helmholtz model, however, was not able to explain the observed dependence of capacitance upon the electrode potential. L. Gouy and G. Chapman came up with a model that was based on the idea that charge carriers, in contrast to the fixed electrode surface charges, exist in a diffusive layer (model b) in figure 4.2). Ions in the vicinity of an electrode are not only influenced by electrostatic forces but also by thermal movements. Due to the decreasing influence of the electrode, counter charge concentration is the highest close to the phase boundary, dropping exponentially with increasing distance from the charged electrode surface. A simplification made in this theory was a point-charge assumption, allowing counter charges to approach the electrode surface arbitrary close. O. Stern dropped this approximation by defining a minimum distance from the electrode (model c) in figure 4.2). Basically, the Stern theory combines Helmholtz’s and Gouy-Chapmann’s ideas with some charges being fixed in a parallel mono-layer, similar to the OHP, and other being located in a diffusive layer. Thus, the absolute value of potential shows a linear disclaim turning into an exponential course.[21–23] Although solvent molecules occupy the electrode, ions of contrary and equal charge may adsorb on the surface. The layer formed by the centers of these ions is called inner Helmholtz plane (IHP) and influences the potential difference depending on the type of ion adsorbed (model d) in figure 4.2). Electrode surface, IHP and OHP are also known as triple layer. However, not all ions are able to adsorb on the electrode. In order to allow for this process Gibb’s free energy of adsorption has to be negative. First, solvent molecules have to be removed from the electrode’s surface to free up space. Second, solvent 15 Chapter 4. Theoretical Background OHP Potential c) OHP Potential d) OHP E le ct ro de qM = qOHP Potential a) Distance in solution Potential b) IHP 0 Distance in solution Distance in solution Distance in solution qM > qOHPqM > qOHP 000 E le ct ro de E le ct ro de E le ct ro de E le ct ro de E le ct ro de E le ct ro de E le ct ro de Figure 4.2: Illustration of different electrical double layer models and respective course of electric potential on a negatively polarized electrode. a) Helmholtz model: a rigid plane model with equal charge amounts within the electrode (qM) and outer Helmholtz plane (qOHP) resulting a linear change in potential. b) Gouy-Chapmann model: counter ions are represented by point-charges in a diffusive layer. c) Stern Model: an extension of the rigid Helmholtz model by a diffusive layer. d) Extension of the Stern model by adsorbed negative ions increasing the metal charge density. Figure adapted from Bockris and Reddy.[21] molecules in the ion hydration shell need to be stripped off before adsorption can take place. While changes in water-electrode interactions are positive, changes in ion-electrode interactions are negative. Ion-water interactions are therefore typically decisive whether the total change in Gibb’s free energy of adsorption is negative or positive. Especially for weakly hydrated ions like iodide or cesium negative values can be obtained, allowing for adsorption.[21, 22] 4.1.2 Cell voltage and electrode potentials In order to force an endergonic electrochemical reaction to occur, a certain minimum voltage is neces- sary. Both half-cell reactions contribute to this voltage with their respective equilibrium potential ϕ0, which depends on temperature, pressure and concentration of substances. When measured under stan- dard conditions, meaning a temperature of 25 ◦C, reactant and product activities equal to 1 mol L−1 as well as a partial pressure of 101.3 kPa, it is called the standard electrode potential ϕθ . The cells standard voltage V θ and equilibrium voltage V 0 is then given by: V θ = ϕ θ cathode−ϕ θ anode (4.1.1) V 0 = ϕ 0 cathode−ϕ 0 anode (4.1.2) 16 4.1. Electrochemical basics V θ results from plain thermodynamics and can be calculated from the reaction’s ∆RGθ as follows [20]: V θ =−∆RGθ zF (4.1.3) Whereat z is the number of electrons transferred and F the Faraday constant (F = 96 485.3 A s mol−1). Individual electrode potentials cannot be measured, thus, comparing half-cell reactions requires a refer- ence potential. Often the redox couple H+/H2, the so called standard hydrogen electrode (SHE), is used for this purpose with its defined ϕθ of 0 V. Hence, recording half-cell potentials is typically done in a three electrode configuration illustrated in figure 4.3. The current is passed from working electrode (WE) to counter electrode (CE), while the voltage is measured between WE and reference electrode (RE). A high ohmic resistance prevents the flow of current between WE and RE. V (I=0) WE RE CE Power source I Figure 4.3: Electrode arrangement and points of measurement in a three electrode configuration. Figure adapted from V. Schmidt.[20] The electrode potential is a state function only depending on temperature, concentration and pressure. In terms of differing temperatures the equilibrium potential is given by:[20] ϕ 0(T ) = ϕ θ + ( dϕ0 dT ) P,c · (T −T θ ) (4.1.4) with: ( dϕ0 dT ) P,c = ∆RS zF (4.1.5) Describing the dependency of reactant and product concentrations on the electrode potential is possible with the Nernst equation:[20] ϕ 0 = ϕ θ + RT zF · ln i ∏ k=1 α νox, k ox, k j ∏ m=1 α νred, m red, m (4.1.6) 17 Chapter 4. Theoretical Background With α being the activities of the components reacting according to the general redox reaction: νox,1Ox1 + ...+νox,iOxi +ze− νox,1Red1 + ...+νox,jRed j (CEq 4.4) The present work focuses on the CO2 electroreduction around ambient pressure, meaning a maxi- mum of 20 mbar relative pressure. For this reason the influence of pressure on the electrode potential is neglected here since a ∆p of 20 mbar (maximum over pressure obtained in all experiments) equals to a ∆ϕ0 of only 0.2 mV. 4.1.3 Kinetic aspects The so far discussed electrode potentials and cell voltages exclusively consider thermodynamics, mean- ing all participating reactions are in equilibrium. Whenever a current is applied to this system certain reactions are accelerated and equilibria are disturbed. At the electrode-electrolyte interface electronic conductance turns into ionic conductance to close the electrical circuit. This charge transfer across the phase boundary causes chemical transformations resulting the electrolysis products, either directly or by subsequent chemical steps. 4.1.3.1 Different types of overpotential Whenever a current is passed through the cell, measured electrode potentials, respectively cell voltage, will differ from the calculated equilibrium potential/voltage. This difference is called overpotential or overvoltage η . ηhal f−cell( j) = ϕmeasured( j)−ϕ 0 (4.1.7) ηcell( j) =Vmeasured( j)−V 0 (4.1.8) Overpotential arises from multiple causes, which can be divided into three categories: activation, concentration and ohmic overpotentials. Activation overpotentials relate to limitations in the fundamen- tal steps of a reaction like ad- or desorption and surface reaction rates. Concentration overpotentials refer to all mass transport related resistances. This includes depletion of reactants or accumulation of products at the active site, caused by high reaction rates and/or slow diffusion rates. They also arise from bubble formation which reduces the active electrode surface area or liquid cross section area for ion transport. 18 4.1. Electrochemical basics Last, ohmic overpotentials are originating from the finite conductivity of electrodes, electrolytes, mem- brane or outer circuit.[20, 24, 25] The measured electrode potential can therefore be expressed as: ϕ( j) = ϕ 0 +ηactivation( j)+ηconcentration( j)+ηohmic( j) (4.1.9) ηohmic |j| / mA cm-2 φ / V ηconcentrationηactivation φ0 φ Figure 4.4: Contributions of activation, concentration and ohmic overpotential to the electrode potential as func- tion of current density. Figure adapted from Endrodi et al.[26] Basically, the different contributions, activation, concentration and ohmic overpotential relate to, catalyst, electrode and cell design, respectively. As indicated in equation 4.1.9 all three types of overpo- tential are a function of the current density passed through the cell, giving individual contributions to the electrode potential as illustrated in figure 4.4. Ohmic overpotentials ideally correlate in a linear manner with current density according to Ohm’s law (V = IR). The outer circuit and metal electrodes themselves usually show very high electronic con- ductivity and do not contribute significantly to voltage losses. On the other hand, ionic charge transport through the electrolyte and membrane, as well as electrical conductivity through electrodes with large amounts of insulating binders is comparatively low. Ohmic losses therefore strongly depend on the cell design and electrolyte, namely the electrode-electrode distance and electrode/electrolyte conductivity. Nevertheless, to be still able to compare different electrode systems investigated in individual cells, the ohmic overpotential is usually measured and the electrode potential is compensated (IR-drop compensa- tion). For some applications electrochemical impedance spectroscopy (EIS) is a powerful technique to determine this ohmic resistance but its implementation and evaluation is rather complex. Another, much easier way, is the current interrupt (CI) method, depicted in figure 4.5. The equivalent circuit, known as Randle’s equivalent circuit, contains a charge transfer resistance parallel to a capacitor representing the chemical reaction and EDL, respectively. Additionally, two ohmic resistors in series 19 Chapter 4. Theoretical Background represent the electron and ion conducting phase. During electrolysis the applied current I0 is periodically turned off and the decay in voltage is analyzed. All ohmic losses instantaneously vanish and the voltage drops from V0 to V1. The non-ohmic charge transfer resistance RCT however, is still influenced by the parallel double layer capacitor CDL. Hence, it gives a discharge curve with finite voltage decay, indicated in the top diagram of figure 4.5 between t0 and t1. The total off-time t1− t0 is thereby in a range of 100 µs and the total ohmic resistance RΩ is given by:[24, 27] RΩ = V0−V1 I0 (4.1.10) t / µs V / V V0 V1 t0 t1 t / µs I / A 0 I0 RE WE RΩ RCT CDL CE RΩ t0 t1 a) b) Figure 4.5: a) Randle’s equivalent circuit with b) idealized course of voltage and current after current interrup- tion. CE: counter electrode; RE: reference electrode; WE: working electrode. Figure adapted from van der Merwe et al.[24] Concentration or diffusion overpotentials exist whenever an electrochemical reaction is occurring. By consuming or producing substances, concentrations close to the active site differ from the bulk phase, influencing the electrode potential. When the rate of the electrochemical reaction (given by current density) exceeds the rate of the mass transport towards or away from the electrode’s surface, concentra- tion overpotential rises exponentially. In other words, diffusion of reactants, products or charge carriers becomes the rate determining step and a further increase in the applied voltage does not yield higher currents.[25, 28] This, however, does only apply to simple reaction systems. In more complex systems like the reduction of CO2 in aqueous solutions, other reducible substances (e.g. water) can accept electrons leading to an indeed increasing current but yielding other products such as hydrogen. The exponential correlation between activation overpotential and current density is discussed in the next section. 20 4.1. Electrochemical basics 4.1.3.2 The Butler-Volmer equation As indicated in figure 4.4, activation overpotentials are dominating at low current densities. In analogy to the Arrhenius equation describing the temperature dependence of reaction rates, J. Butler and M. Volmer introduced a description for the influence of electrode potential on current density. Figure 4.6 represents the Gibb’s free energy as a function of the reaction coordinate. By applying an electric potential the energy state of electrons inside the solid phase (e-(ox)) is raised by zF∆ϕ , while the one of the reduced charge carrier species in the liquid phase (e-(red)) is not influenced. Thus, the energy state of the activated complex is raised by αszF∆ϕ . αs is determining the symmetry of the activation barrier and is therefore also called symmetry factor (0 < αs < 1). Note that the following discussion refers solely to the cathode (forward reaction in figure 4.6 is a reduction), at which both oxidation (ox) and reduction (red) can possibly occur due to reversibility.[21, 28] Reaction coordinate G ib b' s fr ee e ne rg y αszFΔφ zFΔφ ΔG≠red(φ1) ΔG≠red(φ2) ΔG≠ox(φ1) ΔG≠ox(φ2) [ox] + z e- ⇌ [red] e- (ox) e- (red) zF φ 1 zF φ 2 Electrode Electrolyte Figure 4.6: Gibb’s free energy for the transition of an electron from a negatively polarized electrode to a charge carrier in the electrolyte. Figure adapted from Hamann and Vielstich.[28] In general, the rate of a first-order chemical reaction is given by a rate constant k and the reactant con- centration c: r = k · c (4.1.11) The rate constant of a reaction is thereby influenced by the Gibb’s energy of activation (∆G6=) according to the Arrhenius equation: k = k0 · exp [ −∆G 6= RT ] (4.1.12) 21 Chapter 4. Theoretical Background The rate of reduction reaction is therefore given by: rred = kred · c(ox) = k0 red · c(ox) · exp [ −∆G6=red(ϕ1) RT ] (4.1.13) With c(ox) being the reactant concentration. When applying an electrode potential ∆ϕ the electron’s Gibb’s free energy is raised from zFϕ1 to zFϕ2. Together with jred =−zFrred, partial current density of the forward reaction or reduction jred can be described as: jred(ϕ2) =−zFk0 red · c(ox) · exp [ −∆G6=red(ϕ2) RT ] (4.1.14) Taking figure 4.6 into account, ∆G6=red(ϕ2) can be rewritten as: jred(ϕ2) =−zFk0 red · c(ox) · exp [ −∆G6=red(ϕ1)+(1−αs)zF∆ϕ RT ] (4.1.15) with ∆ϕ = ϕ2−ϕ1. Setting ϕ1 as the potential equal to the reference electrode, the term−∆G6=red(ϕ1)/RT becomes constant and can be included into a derived rate constant k0′ red. ϕ2 then corresponds to the measured electrode potential ϕ . jred(ϕ) =−zFk0′ red · c(ox) · exp [ −(1−αs)zF RT ·ϕ ] (4.1.16) In the same way, the partial current density of the backward reaction or oxidation can be expressed by: jox(ϕ) = zFk0′ ox · c(red) · exp [ αszF RT ·ϕ ] (4.1.17) Equations 4.1.16 and 4.1.17 describe the partial current densities of reduction and oxidation reaction, depending on the electrode potential ϕ . At an open circuit condition all components are in equilibrium and both reactions are equally fast. The respective current density is called exchange current density j0 = jox(ϕ 0)= | jred(ϕ 0)|. Applying Nernst’s equation (equation 4.1.6) and the definition of overpotential according to equation 4.1.7, both partial current densities can be expressed as: 22 4.1. Electrochemical basics jred(η) =− j0 · exp [ −(1−αs)zF RT ·η ] (4.1.18) jox(η) = j0 · exp [ αszF RT ·η ] (4.1.19) with: j0 = zFk0′ red · c(ox) · exp [ −∆G6=red(ϕ 0) RT ] (4.1.20) j0 = zFk0′ ox · c(red) · exp [ −∆G 6=ox(ϕ 0) RT ] (4.1.21) Finally, adding up both partial current densities results in the Butler-Volmer equation:[21, 28] j(η) = jox(η)+ jred(η) = j0 · { exp [ αszF RT ·η ] − exp [ −(1−αs)zF RT ·η ]} (4.1.22) Obviously, the j(η) function depends on the exchange current density and symmetry factor αs. Es- pecially j0 depends directly on the electrode’s activity, determining ∆G 6=red(ϕ1). Figure 4.7 illustrates the Butler-Volmer equation with some arbitrary values for j0 and αs. j0 = 10 ·10−4 A cm−2 corresponds to relatively active reaction system like the H+/H2 couple at metal electrodes. j0 = 10 ·10−6 A cm−2, on the other hand, corresponds to a moderately active system like the CO2/HCOO- couple on a tin oxide electrode.[28, 29] As mentioned, αs describes the symmetry of the curve. At αs = 1/2 equal absolute val- ues of negative or positive overpotential result in the same absolute value of current. With differing αs reduction or oxidation are accelerated differently.[21] 0 0 jox = j0 · e αszF RT η jred = −j0 · e− (1−αs)zF RT η j = jox − jred η / V j / m A cm = 2 −0.1 0 0.1 −0.5 0 0.5 αs = 0.2 αs = 0.5 αs = 0.8 j0 = 0.1mAcm=2 αs = 0.5 j0 = 1µAcm=2 η / V j / m A cm = 2 Figure 4.7: The dependence of current density on overpotential according to Butler-Volmer.[21, 28] a) Contribution of reductive and oxidative current density to the observed one (αs = 0.5). b) Effect of different values for αs and j0. 23 Chapter 4. Theoretical Background 4.1.4 Quantitative relations Quantitative relations between the introduced electric charge and converted substance are an absolute fundamental in electrolysis and were first described by Faraday in 1834.[30] Faraday described the de- posited molar amount of substance nF as being proportional to the passed electric charge q and reversed proportional to the substance valance. Mathematically speaking this can be expressed as: nF ∼ q z (4.1.23) The proportionality constant F is named after Faraday and the so-called Faraday’s first law reads as follows: q = z ·F ·nF (4.1.24) With q = I · t (4.1.25) the molar amount of substance can be expressed as: nF = I · t z ·F (4.1.26) nF therefore represents the maximum amount of product to be formed. However, electrolysis is rarely completely selective. In electrochemical processes it is common to express this selectivity as faradaic efficiency (FE) towards product i, defined as: FE i = ni nF ·100% = ni · z ·F I · t ·100% (4.1.27) This value includes side reactions and short circuit currents. The latter is especially important in industrial scaled electrolyzers with a high total current. Short circuit currents strongly depend on the reactor design, while the extent of side reactions is mostly based on catalyst, electrode design and reaction conditions. With the FE to the desired product i and the equilibrium voltage of the overall reaction (equation 4.1.2) an energy efficiency (EE) of the electrolysis can now be calculated. EEi = V 0 cell Vcell ·FEi (4.1.28) 24 4.1. Electrochemical basics For most electrochemical investigations cathode potential is measured against a reference electrode in- stead of the cell voltage. In order to still be able to compare the performance of reported systems, the energetic cathode efficiency (ECE) is introduced and defined as: ECEi = ϕ0 C, vs SHE ϕC, vs SHE, IR ·FEi (4.1.29) Note that the ECE does not represent a real energetic efficiency, since ohmic overpotentials and the counter reaction are not considered in this value. Also, the ECE depends on the reference electrode chosen for obtaining ϕ0 and ϕ . It does, however, allow a comparison of the cathode’s performance. A high energetic efficiency is only achieved by a combination of high selectivity and low overpotentials. Both strongly depend on the chosen catalyst, type of electrode and applied reaction conditions. The cell design in turn influences mainly the ohmic overpotentials. The total power consumption P of an electrolyzer is simply calculated by the product of cell voltage and current passed. P =Vcell · I (4.1.30) Multiplied by time gives the electrical energy E: E = P · t (4.1.31) In order to be able to compare different types of electrodes, the measured current is typically normalized to the geometrical surface area Ageo of the electrode and given as current density. j = I Ageo (4.1.32) 4.1.5 Important aspects of electrochemical cells The term electrochemical cell describes any type of reactor in which redox reactions are conducted using or generating electric power. Depending on the type of application different basic designs are common. Fundamental research on a nanometer scale like the investigation of catalyst systems usually requires to work with low amounts of material and thus small reactor volumes. The latter is especially important when chemical analysis of produced or consumed substances is targeted. Usually, non-industrial elec- trodes like glassy carbon supports with deposited electrocatalytically active components are employed, 25 Chapter 4. Theoretical Background which need to be operated at low current densities to avoid mass transport limitation. The resulting low amounts of product or little changes in reactant concentration are challenging to detect if the cell volume is too large. For these reasons, H-type cells or low-volume vessels incorporating special electrodes like rotating (ring) disk or dropping mercury electrodes are applied. For industrial applications an energy efficient production is targeted. Multiple designs are therefore known, which need to be selected depending on aspects e.g. regarding electrode structure, type of elec- trolyte, operating temperature or mass transport. A main issue to be considered is the reduction of power consumption while maintaining high currents. This implies the reduction of voltage losses caused by different, already discussed, effects. Reaction overpotentials due to moderately active electrodes, mass transport limitations and ohmic resistances are thereby to recall. Optimizing the electrochemical cell therefore needs to address all these issues, while also considering aspects of manufacturing, material costs and long-term stability. The ability to scale up a cell either by geometrical expansion or by numbering-up is also of great importance. Increasing the active electrode area by multiplying single cells is called stacking. While different ways to connect multiple cells are known, putting them in series is probably the most common way. Here, the anode of one cell is electrically connected to the cathode of the next cell either by a bipolar plate or by external wiring. Such a stack incorporating thin plates of electrodes and electrolyte flow fields is illustrated in figure 4.8. External electrical connection Bipolar plate connection Stack with bipolar plate connections Cathode Membrane Anode Figure 4.8: Schematic illustration of multiple cells being stacked in a filter press. Figure adapted from V. Schmidt.[20] 26 4.2. Carbon dioxide reduction reaction 4.2 Carbon dioxide reduction reaction The chemical conversion of CO2 is a long known and studied topic with multiple possible pathways and products. Basically, these pathways can be divided into two groups of reactions. First, reactions where the oxidation state of the carbon atom, +IV, is maintained, for example during production of urea, sali- cylic acid, polycarbonates or other inorganic carbonates.[7, 31] These reactions require comparatively low amounts of energy and sometimes not even a catalytic process. The second group of reactions contains pathways in which the final product has a lower oxidation state than +IV. Examples would be oxalic acid (+III), formic acid or CO (+II), methanol (-II) and methane (-IV). These reactions require much larger amounts of energy, but their general demand is also higher. Production of such substances can be done thermocatalytically like in the methanol synthesis process or in the Sabatier reaction.[7] While these reactions require quite harsh reaction conditions, electrochemical reduction allows a conversion at ambient pressure and temperature. CO2 electrolysis (CO2EL) has been studied intensively for the past five decades including a variety of different products.[32] The following sections focus on different as- pects of the electrochemical conversion of CO2 like thermodynamics, catalysts, mechanisms and reaction conditions. 4.2.1 Fundamentals Depending on the catalyst and electrolyte used, CO2 can be converted into several products. The most common ones are listed in equations (CEq 4.5) - (CEq 4.7). Besides the well investigated pathways to formate/formic acid, CO and methane, higher hydrocarbons with a brought range of products like alkenes[33, 34], alcohols[35, 36], aldehydes and ketones[37] are also known. When using aprotic solvents a direct coupling of two CO2 radical anions forming oxalate is also possible.[38] Except for oxalate formation a source of protons is of course necessary. This, however, leads to the problem that these protons themselves can be reduced to hydrogen in a reaction called hydrogen evolution reaction (HER) (CEq 4.8). This undesired product represents one of the main side reactions in CO2RR. Its formation has to be suppressed by applying the right catalyst, electrode and reaction conditions. 27 Chapter 4. Theoretical Background CO2 + H2O+2e− HCOO−+ OH− (CEq 4.5) CO2 + H2O+2e− CO +2OH− (CEq 4.6) CO2 +4H2O+8e− CH4 +8OH− (CEq 4.7) H2O+2e− H2 +2OH− (CEq 4.8) Figure 4.9 shows a schematic illustration of an alkaline electrolysis cell converting CO2 and a metal hydroxide to the respective metal formate, bicarbonate and oxygen. Reaction (CEq 4.9) proceeds on the cathode, forming formate and hydroxide ions. Another CO2 molecule can then react with the formed hy- droxide ion in a consecutive hydroxylation to bicarbonate. On the anode, hydroxide ions are oxidized to oxygen leaving behind water and the cation of the metal hydroxide (CEq 4.11). These cations are trans- ferred through the membrane to maintain charge neutrality, leading to the overall reaction (CEq 4.12). Reduction: CO2 +H2O+2e− HCOO−+OH− (CEq 4.9) Consecutive hydroxylation: OH−+CO2 HCO3 − (CEq 4.10) Oxidation: 2KOH 1/2O2 +H2O+2K++2e− (CEq 4.11) Overall reaction: 2CO2 +2KOH KHCOO+KHCO3 + 1/2O2 (CEq 4.12) C at h o d e A n o d e Power source 2K+ 2e- H2O + CO2 HCOO- + OH- CO2HCO− 3 2e- H2O + 1/2O2 2OH- e-e- Figure 4.9: Schematic illustration of an electrolysis cell converting CO2 and potassium hydroxide into the re- spective formate/bicarbonate salt mixture and oxygen. 28 4.2. Carbon dioxide reduction reaction 4.2.2 Equilibrium potentials Using equation 4.1.1 and 4.1.3, the standard electrode potential for the CO2RR to the most common C1- products can be calculated. For a detailed calculation and thermodynamic literature values of standard formation enthalpy and entropy see appendix section 9.1, p. 213. The following values are given vs SHE at pH = 0, T = 25 ◦C and p = 101.3kPa. CO2 +2H++2e− HCOOH ϕ θ = −0.166V (CEq 4.13) CO2 +2H++2e− CO+H2O ϕ θ = −0.104V (CEq 4.14) CO2 +8H++8e− CH4 +2H2O ϕ θ = 0.17V (CEq 4.15) 2H++2e− H2 ϕ θ = 0V (CEq 4.16) When formulating the reaction equation for the reduction of CO2 to formate/formic acid, there are multiple possibilities depending on the present species, which again depend on pH and respective pka values. Between a pH of 0 and 3.77 (pka, formic acid) formic acid is the main product while at higher pH formate is formed. On the other hand, the thermodynamically stable form of CO2 within a pH range of 6.35 (pka, carbonic acid) to 10.3 (pka, bicarbonate) is bicarbonate. Above that range carbonate is the main form. This leads to the following reaction equations. For an easier discussion later on, all of them are also represented in their proton utilizing form. CO2 +2H++2e− HCOOH pH = 0 to 3.77 (CEq 4.17) CO2 +H2O+2e− HCOO−+OH− pH = 3.77 to 6.35 (CEq 4.18) (CO2 +H++2e− HCOO−) HCO3 −+H2O+2e− HCOO−+2OH− pH = 6.35 to 10.3 (CEq 4.19) (HCO3 −+2H++2e− HCOO−+H2O) CO3 2−+H2O+2e− HCOO−+3OH− pH = 10.3 to 14 (CEq 4.20) (CO3 2−+3H++2e− HCOO−+H2O) Transforming equation 4.1.6 with the definition of the pH (pH = −log(αH+)) and with the convention that under standard conditions all other activities are equal to 1, leads to the following general ϕ0(pH) relation. ϕ 0 = ϕ θ −0.059V · νH+ z ·pH (4.2.1) 29 Chapter 4. Theoretical Background It is obvious that the slope of this function depends on the νH+/z ratio, which is different for reactions (CEq 4.17) to (CEq 4.20). Based on this ratio and the pka values for formic acid, carbonic acid and bicarbonate the ϕ0(pH) function can be plotted in a so called Pourbaix diagram (see figure 4.10). While the slope of the CO2, HCO3 -, CO3 2-/HCOOH, HCOO- equilibrium potential differs for different pH ranges, it is constant for the CO2 (HCO3 -, CO3 2-)/CO; CO2 (HCO3 -, CO3 2-)/CH4 and H+/H2 couples with a slope of −0.059 V. These courses of the equilibrium potential over pH are well known from literature.[39–41] However, bicarbonate and carbonate salts itself cannot be converted electrochemically to formate/formic acid.[42] Instead, physically solved CO2 is the active species.[42] This consideration leads to the fact that chemical equation (CEq 4.18) applies to a pH range of 3.77 to 14. The resulting equilibrium potential is given by the upper solid line in figure 4.10 (CO2/HCOOH, HCOO-). Calculated equilibrium potentials show that at high pH values CO2RR to formate is favored against production of CO and hydrogen. 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 −1 −0.8 −0.6 −0.4 −0.2 0 0.2 pH ϕ 0 / V (v s S H E ) CO2/H2CO3 HCO3 - CO3 2- HCOOH HCOO- CO2/HCOOH, HCOO- H+/H2 CO2/CH4 CO2/CO CO2, HCO3 -, CO3 2-/HCOOH, HCOO- Figure 4.10: Pourbaix diagram for hydrogen evolution and various CO2 reduction reaction products. The lower solid line represents the equilibrium potential for the CO2/bicarbonate/carbonate - formic acid/for- mate redox couple, when the thermodynamically stable species are considered. The upper solid line considers CO2 as reactant over the whole pH range. The present work focuses mainly on the reduction reaction of CO2EL. However, a counter reaction is always necessary for electron donation. In most studies oxygen evolution reaction (OER) serves as counter reaction. Mainly due to its simple handling. Figure 4.11 shows the equilibrium cell voltage for CO2RR to formate/formic acid and OER as function of pH. Usually, electrolyzers work at the same pH value for both half-cells due to the ability of proton/cation exchange through the membrane. Thermody- 30 4.2. Carbon dioxide reduction reaction namically speaking this makes a pH of 14 the best working point with a minimum potential difference of V 0 min = −1.095V. 0 2 4 6 8 10 12 14 1 0.5 0 −0.5 H2O/O2 CO2/HCOOH, HCOO- V 0 min = 1.095V V 0 max = 1.395V pH ϕ 0 / V (v s S H E ) Figure 4.11: Pourbaix diagram for CO2 reduction to formic acid/formate and oxygen evolution reaction. The difference between both represents the standard cell voltage. 4.2.3 Electrocatalytic conversion of carbon dioxide Carbon dioxide with its linear geometry is a very stable molecule (∆Gf = −394kJ mol−1)[43]. This fact requires a relatively high energy input, or energy rich co-reactants, to convert it. Schwarz and Dodson reported the first electron transfer step, forming the CO2 radical anion, to occur at −2.24 V (vs. SHE), which is −2.07 V to −1.55 V larger than the equilibrium potential.[44] The need for an effective catalyst is obvious. It is not only needed because of the high activation energy but also to achieve a high selectivity since the equilibrium potentials of various CO2RRs and HER are similar. A catalyst with low overpo- tential for the desired product and high overpotential for all undesired reactions is necessary. In light of this, it becomes intelligible that the vast majority of current studies on electrochemical CO2 conversion focuses on the development of highly active and selective catalyst systems. Metal, metal oxide and alloy based systems are the most intensively studied ones. [45–47] Metal chalcogenide,[47] homogeneous[48] and metal-free[49] catalyst systems, however, are also well known. In the present work solely metal and metal oxide catalysts were used and the following discussion exclusively addresses these systems. 4.2.3.1 Metal/metal oxide catalysts Metal/metal oxide based catalysts can be divided into four major groups, depending on the main product. Metals that form mainly formate/formic acid (Sn, Pb, Hg, In, Cd, Bi), metals that form mainly CO (Au, Ag, Zn) and metals that form high amounts of hydrocarbons (Cu).[32, 39] The last group of metals is not active for CO2RR and produces mainly hydrogen (Pt, Ni, Fe).[32] Within the first group, tin oxide is 31 Chapter 4. Theoretical Background the most investigated metal oxide, mainly due to its high performance, low price and non-toxicity.[50] However, FE and ECE of the respective catalyst can vary significantly, depending on the type of metal, its oxidation state, particle size, morphology or the presence of co-catalysts.[17] Figure 4.12 gives a slight overview over the enormous amount of literature, illustrating the achieved FE and ECE or EE over the applied current density. 0 100 200 300 400 500 0 20 40 60 80 100 a) |j| / mA cm-2 F E F o rm a te / % 0 20 40 60 80 100 E E / % 0 500 1,000 1,500 0 20 40 60 80 100 |j| / mA cm-2 E C E / % Literature (until August 2019) Results of this work Figure 4.12: Summary of a) faradaic efficiencies to formate, [19, 29, 41, 51–79] and b) energetic cathode efficien- cies [29, 41, 51–75] and energy efficiencies [19, 51, 76–79] reported in literature (up to August 2019) over the respective current density. Only formate selective metal, metal oxide and metal alloy based catalyst systems of the elements Sn, Pb, In, Bi and Cu are represented. The oxidation state of the catalyst is probably the most important factor determining selectivity and activity. For tin based systems the presence of surface oxide species is mandatory. Metallic tin shows no significant activity towards CO2RR. In fact it even is a moderate catalyst for HER.[58, 80] Surface tin ox- ide forms easily when exposed to air, but can be reduced during cathodic operation.[58, 81] Whenever the native oxide layer is removed, either by acidic or electrochemical etching, HER partial current density strongly increases.[58] Figure 4.13 shows the thermodynamically stable regions for tin and tin oxides. From a pure thermodynamic point of view tin based catalysts should not be able to convert CO2. The on- set potential for reduction of CO2 to formate is around −1 V vs SHE with a typical window of operation between −1.3 V to −1.6 V vs SHE.[59] At these potentials tin should exist in its metallic form. Fortu- nately, the reduction of tin(II)-(IV) oxide is kinetically hindered.[82] In-operando spectroscopic methods like Raman and X-ray absorption spectroscopy (XAS) revealed that a metastable oxidized state is main- tained even at much more negative potentials than suggested by thermodynamics.[81, 82] Depending on the production method of the catalyst, a tin(II) mixed oxide/hydroxide is formed on the surface with a core of metallic tin or tin(IV) oxide.[82, 83] 32 4.2. Carbon dioxide reduction reaction 5 6 7 8 9 10 11 12 13 14 −1 −0.5 0 Hydrous Sn+IV oxide Metallic Sn Hydrous Sn+II oxide S ta n n at e (+ IV ) io n s pH ϕ 0 / V (v s S H E ) Figure 4.13: Pourbaix diagram of tin. Figure adapted from Dutta et al.[81] These findings, however, cannot be simply transferred to other metals. While indium shows a similar behavior for anodized or etched surfaces as tin, lead and bismuth show a higher selectivity when fully reduced.[80] In general, it is also hard to state a ranking for different metal catalysts, as their performance depends on too many factors. While some groups report activity and selectivity in the order of Pb > Sn and Pb > Sn, respectively,[62, 66] others report it in the order of Sn > Pb and Sn > Pb,[84] Pb > Sn and Sn > Pb[85] or Sn > Pb and Pb > Sn.[39] CO2RR is known to be a surface sensitive reaction, meaning that activity and selectivity of a catalyst depend on its particle size and morphology. For example, Zhang et al. prepared tin(IV) oxide particles of different particle sizes ranging from 3 nm to 200 nm in diameter, supported on carbon black. As a key message the authors reported a maximum in formate FE at a particle size of d̄p = 5nm. Their suggested explanation, despite not being proven, is that the binding energy between key intermediates and the active site shows an optimum for 5 nm particles.[74] Another explanation for such an optimum might by derived from DFT studies for the conversion of CO2 to CO, whereat edge sites of Au nanoparticles favor CO2RR and corner sites favor HER.[86] Apart from adjusting the size of spherical nanoparticles, much effort is invested in creating de- fined morphologies. Examples are simple rod-,[87] coralline-[88] or urchin-like[89] structures, as well as dendritic[90] structures and sheets forming hierarchical mesoporous systems[91]. The motivation for structuring catalysts is versatile. One aspect is the higher amount of active sites per geometrical elec- trode surface or per catalyst mass. Compared to polycrystalline metal foils, structured catalysts obtain a much larger surface area and more unsaturated atoms.[92] Another aspect is mass transport. Besides the intrinsic activity of an active site its accessibility by the reactant is a crucial point, which can be addressed by a special electrode design. For example Li et al. demonstrated a superior performance of 33 Chapter 4. Theoretical Background thin sheets, consisting of highly porous tin oxide, compared to planar metal foils and even some GDEs. They attributed the higher activity to an increased active surface area and enhanced CO2 mass transfer.[91] Last, non-oriented primary particles create numerous grain boundaries within the catalyst particle, which hinder an efficient electron transport. In contrast to that, straight rod- or wire-like structures can provide undisturbed electron transport.[93] However, highly structured materials again come with their own dis- advantages. Synthesis often requires templates and hydrothermal conditions. The resulting particles are typically in a range of 100 nm to 2 µm.[88, 91, 94, 95] This can complicate an efficient introduction in elec- trode systems like GDEs or lead to low mass activity. Using GDEs the particle size reduction to a target size of 5 nm achieves a comparable or even better performance than the mentioned special structured materials.[19] 4.2.3.2 Catalyst deactivation Chemical or physical changes to the catalyst with negative impact on performance, in general referred to as catalyst deactivation, have been reported repeatedly by different groups. Various mechanisms for deactivation are known, e.g. induced either by unsuitable cathodic polarization or other interfering substances. Highly negative potentials may change the catalysts oxidation state, e.g. from tin(II) to tin(0), which is just moderately active for CO2RR as previously mentioned.[81] Anawati et al. reported a minimum in cathode deactivation at −1.6 V vs SHE. Less negative potentials increased the oxide layer thickness, while significantly more negative potentials lead to the formation of KSn. The colloidal in- termetallic compound formed between catalyst and electrolyte was found to be responsible for material loss and cathode deactivation.[96] This transition is reported to occur on a time-scale of several hours and might be an explanation for material loss reported by multiple groups, especially at higher current densities.[53, 63, 97] Other present substances like heavy metal impurities are also well known to contribute to a cathode deactivation. Impurities in the electrolyte can be deposited during cathodic polarization and interfere with the CO2RR catalyst either by poisoning the active site itself, or by providing HER active sites.[39] Hori et al. showed that pre-electrolyzing the electrolyte can prevent deactivation of a copper based catalyst effectively by removing heavy metals from the electrolyte.[98] Last, the reactant CO2 it- self can lead to deactivation. Reports state a black deposit, which was analyzed to be mostly graphitic, blocking active sites. Different recovering techniques like deep cathodic polarization or anodic polariza- tion turned out to be able to remove these deposits either by mechanic stripping with high amounts of hydrogen being produced or by anodic oxidation. 34 4.2. Carbon dioxide reduction reaction 4.2.3.3 Reaction mechanisms Carbon dioxide electroreduction can lead to a brought variety of products with different reaction path- ways competing. The extent at which each pathway is undergone depends mainly on the catalysts, leading to the four different groups of metals described in section 4.2.3.1. The following description of proposed reaction mechanisms will focus on formate production with some remarks about CO and hydrocarbon formation. Each overall reaction pathway consists of multiple elemental steps. The formate producing pathway can be summarized as follows: CO2(g) CO2(aq) (CEq 4.21) CO2(aq) CO2(ad) (CEq 4.22) CO2(ad)+e− CO2(ad) q− (CEq 4.23) CO2(ad) q−+H2O HCOO(ad) q +OH− (CEq 4.24) HCOO(ad) q+e− HCOO−(ad) (CEq 4.25) HCOO−(ad) HCOO−(aq) (CEq 4.26) Equation (CEq 4.21) and (CEq 4.22) refer to an initial equilibrium, which is connected to the CO2/carbonic acid/bicarbonate/carbonate equilibrium resulting the required adsorbed CO2.[42, 99] This adsorbed species is then activated by a single electron transfer yielding the CO2 radical anion. Analysis of different electrode-electrolyte systems showed similar characteristics in obtained polarization curves indicating that all reaction pathways share the same rate determining step. The only shared pathway for all CO2RR products is the single electron transfer of equation (CEq 4.23) (see figure 4.14). Subsequent protonation of the CO2 radical anion and a second single electron transfer leads to formate which desorbs from the active site.[99] One major difference which determines the main selectivity of a catalyst arises from equation (CEq 4.24). CO2 orientation towards the electrode’s surface during protonation is a crucial parameter. Feaster et al. calculated binding energies for two cases with either a *COOH or *OCOH adsorption orientation by density functional theory (DFT) and correlated their results with observed partial current densities of CO or formate. Vulcano plots for both cases were thereby obtained which fit very well to experimental data. CO selective metals like gold and silver are on top of the vulcano shaped function jCO(*COOH binding energy), while other metals bind this intermediate either too strong (platinum) or too weak (tin). When using the binding energy of the *OCOH intermediate as descriptor, tin is on top of the jHCOO-(*OCOH 35 Chapter 4. Theoretical Background CO2 O O C H H2O - OH- e- HCOO- H2O - OH- O OH C H2O + e- - OH- O C CO H2O + e- - OH- n (H2O + e-) -n (OH-) CH4, C2H4, ... e- e- O O C O O C O C H Figure 4.14: Mechanisms of different CO2 reduction reaction pathways leading to the three major groups of products, namely formate, CO and hydrocarbons. Figure adapted from Zhu et al.[47] with contents of Feaster et al.[100]. binding energy) plot. These results suggest that CO selective metals tend to bind the CO2 radical anion at the carbon atom, while formate selective metals bind it via the oxygen atoms.[100] Coordination of CO2 towards a tin surface via the oxygen atoms is in good accordance with other investigations stating that under electrolysis conditions a surface carbonate species is formed.[83] As mentioned in section 4.2.3.1 multiple groups reported the importance of an oxide surface layer which contains a mixture of surface oxides and hydroxides.[58, 80, 81, 101] Without these oxygen species CO2 cannot adsorb in the necessary orientation. Instead, protons can adsorb more easily to the metallic surface, hence shifting the selectivity towards hydrogen.[101] Another pathway that is based on *COO adsorption is the formation of hydrocarbons. This way is based on the fact that CO does not desorb easily from a copper surface as it is the case for e.g. silver based electrodes.[102] Adsorbed CO is thereby further reduced in multiple single electron transfer and protonation reactions.[99] The first hydrogen atom addition can proceed by two different ways. First, by the previously de- scribed addition of a proton from free or adsorbed water to the CO2 radical anion. This reaction is proposed to occur on most solid metal catalysts (equation (CEq 4.24)). Second, it can proceed via inser- tion into a metal hydride as follows: MH+CO2 MCOOH (CEq 4.27) The major difference for both reactions is the activation step. Within the first described mechanism CO2 itself needs to be activated by an initial single electron transfer, which requires a comparatively large amount of energy.[39] The second described mechanism is only known for palladium based nanoparticu- late catalysts operated at low overpotential (at least for heterogeneous catalyst systems). Instead of CO2 36 4.2. Carbon dioxide reduction reaction activation water is used to form a metal hydride in an initial step, which consecutively reacts with CO2. In addition, specially designed homogeneous catalyst were also reported to be able to promote this mech- anism. It allows a significant reduction of the overall activation barrier and related overpotential.[99, 103] Thus, the reaction mechanism changes to: CO2(g) CO2(aq) (CEq 4.28) CO2(aq) CO2(ad) (CEq 4.29) H2O H(ad) ++OH− (CEq 4.30) H(ad) ++e− H(ad) q (CEq 4.31) H(ad) q +CO2(ad) COOH q (ad) (CEq 4.32) COOH(ad) q +e− HCOO−(ad) (CEq 4.33) HCOO−(ad) HCOO−(aq) (CEq 4.34) 4.2.4 Influence of operating conditions In recent years several studies have been published focusing on different operating conditions like type and concentration of electrolyte, pH or temperature. All these parameters will be described in the follow- ing subsections. The main problem when trying to assess optimum reaction conditions is that presented studies are barely comparable due to the high complexity of such electrolysis systems. However, the large number of publications allows the identification of trends and starting points for the herein investigated system. 4.2.4.1 Effects of the electrolyte In every electrochemical cell the electrolyte, which can exist in liquid or solid state, is responsible for ion transport between both current bearing electrodes. In polymer electrolyte membrane (PEM) electrolyzers for example the electrolyte consist of a solid CEM sandwiched between the two electrodes. This concept is especially applicable for producing gaseous products like CO or CH4, which can be extracted by a humidified CO2 feed. For alkaline CO2RR a liquid electrolyte between cathode and membrane is usually applied to remove the produced formate salt. In general this liquid electrolyte consists of a solvent with a conductive salt and should fulfill the following requirements.[20] • Chemically and electrochemically stable within the used potential range. • High solubility for reactants and products. • High mobility of reactants and products to facilitate fast mass transport. 37 Chapter 4. Theoretical Background • Stabilization of reaction intermediates. • Low resistivity to avoid ohmic losses. • High heat conductivity to remove or supply heat locally. • Low vapor pressure to avoid dilution of the CO2 feed, bubble formation and uncontrolled reconden- sation. Even though the cathodic stability of water is rather moderate (HER competing with CO2RR) and the solubility of CO2 is very low (SCO2, 20 ◦C = 35 mmol L−1)[38], the lack of alternatives leads to water as the most widely used solvent. The type of cation in the electrolyte is known to influence the CO2RR significantly. While the ob- servations of different research groups are quite consistent with larger cations increasing the electrode’s activity and selectivity, the given explanations are more diverse. Even though the presented effects are reported for silver or gold based electrodes they are believed to also apply to tin based systems since the adsorption of CO2 is not affected and the rate determining step is equal.[104] Thorson et al. attributed the measured trend in activity of Li+ < Na+ < K+ < Rb+ < Cs+ to basically two effects, which are both based on cation adsorption on the electrode surface. First, small cations in general show a higher hydration power, meaning more water molecules are contained within the hydration shell. This leads to a lower cation adsorption on the negatively charged cathode. Adsorbed cations, however, can stabilize reaction intermediates like the CO2 radical anion. Accordingly, large cations show a better stabilization. Second, adsorbed cations have an influence on the potential of the outer Helmholtz layer, ϕOHP. Due to a higher propensity for larger cations to adsorb on the electrode, ϕOHP becomes less negative.[104] This difference in turn results in a decrease in adsorbed H+ according to equation 4.2.2.[105] Since H+ is involved in the rate determining step of the HER (RDS: H+ + e– H q)[106] a lower concentration of H+ means a more negative equilibrium potential. Thus, the partial current density of hydrogen at a given electrode potential is smaller for larger cations. [H+]Electrode = [H+]Bulk · exp [ −ϕOHP ·F RT ] (4.2.2) In contrast to these explanations, Mills et al. calculated the adsorption equilibrium potentials for alkali metal cations on transition metals via DFT. Their results showed that cation adsorption is unfa- vored at typical electrode potentials in CO2RR of −1.2 V to −1.6 V (vs SHE) and requires even higher potentials in the range of −2.1 V to −2.6 V (vs SHE). Furthermore, the steric effect of larger cations has been investigated using a modified Poisson-Boltzmann theory and was found to be negligible.[107] Singh et al. concluded from these findings that neither cation adsorption nor steric effects provide sufficient 38 4.2. Carbon dioxide reduction reaction justification and that another effect has to exist. The authors used a multiphysics model to investigate the effect of cation hydrolysis.[108] Metal ions in aqueous solution form a hydration shell which can undergo an acid-base reaction according to: M+(H2O)x MOH(H2O)x−1 +H+ (4.2.3) This reaction creates a potential buffer system but for bulk phase cations the pKa value is rather high, e.g. 13.6 for Li+ and 14.7 for Cs+. Thus, for medium alkaline electrolytes cation hydrolysis is negligible. This changes significantly when the cation approaches the strongly negative polarized electrode. Due to stronger electrostatic interactions, hydrolysis proceeds much faster and the pKa value drops especially for larger cations, e.g. to 11.6 for Li+ and to 4.3 for Cs+. Therefore, larger cations provide a greater buffer ca- pacity and prevent the formation of bicarbonates or carbonates close to the electrode. The larger amount of available CO2 reduces the reaction’s overpotential and increases the electrode’s activity.[108] These theoretical findings were recently confirmed by in-situ attenuated total reflectance surface-enhanced in- frared absorption spectroscopy (ATR-SEIRAS), although to a lesser extend than predicted.[109] In contrast to the straight forward size dependence of alkaline metal cations, the effect of anions is more complex due to their greater variety and more versatile chemical characters. Depending on the type of anion, effects on pH and pH buffer capacity, strength of adsorption or CO2 buffer capacity have to be considered. In general bicarbonate is the most widely used anion, mainly due to its ability to buffer pH and CO2 concentration. During CO2RR in neutral to alkaline electrolytes hydroxide ions are produced in stoichiometric quantities (see (CEq 4.9)), which leads to a higher local pH, depending on the current density. The carbonic acid/bicarbonate/carbonate system can compensate for this increase to a certain extend which allows higher concentrations of solved CO2 at the active site, compared to non-buffered systems.[38] Note that bicarbonate or carbonate ions themselves cannot be reduced and have to be pro- tonated and dehydrated to CO2 first in order to be available for conversion.[42, 110] The pH dependency of the FECO2RR products : FEH2 ratio is known to have a maximum in neutral to slightly alkaline media with acidic values favoring HER by excessive proton availability and highly alkaline conditions shifting the carbonic acid equilibrium away from free CO2.[51, 78, 111, 112] This, however, strongly depends on the type of electrode and current density as both have a large influence on the local pH.[111] Different research groups reported the electrolyte’s pH value to be one of the most crucial parame- ters in CO2RR, especially when working in an H-type cell with long reactant diffusion pathways. The pH dependency is thereby reported to arise from the related CO2/carbonic acid/bicarbonate/carbonate equilibrium instead of protons being involved in the rate determining step.[41, 113, 114] However, CO2RR 39 Chapter 4. Theoretical Background results in the production of one or more hydroxide ions, depending on the product formed. Especially at high current densities this gives a much higher local pH compared to the bulk. Burdyny and Smith predicted the local pH to quickly rise with increasing current density and to show similar values at −200 mA cm−2, regardless of the bulk electrolyte. Accordingly, the initial large difference in solved CO2 vanishes with increasing reaction rates.[111] Besides bicarbonate, hydroxide, sulfate, fluoride, chloride, bromide and iodide salts are repeatedly reported. Halides have been investigated especially on copper and silver electrodes and are reported to improve production rates of methane, higher hydrocarbons [115–117] and CO[118] in the order F- < Cl- < Br- < I-. The increasing adsorption strength of F- to I- is assumed to be the decisive factor. Large anions like Br- help to stabilize the adsorbed CO2 molecule during the geometry changing first single electron transfer step by donating electron density to the positively polarized carbon atom. Furthermore, it is suggested that the adsorption of halides suppresses the adsorption of protons and thus increases the HER overpotential. Hence, a stronger halide adsorption favors CO2RR over HER.[117, 118] The ranking for activity and selectivity between hydroxide, sulfate, bicarbonate and halides in contrast is not consistent in literature and should be investigated for each electrolysis system separately.[71, 112, 119, 120] In order to minimize ohmic overpotentials and raise energeti