Volume 62, Issue 2, April 2018 Published by Johnson Matthey © Copyright 2018 Johnson Matthey

ISSN 2056-5135

Johnson Matthey’s international journal of research exploring science and technology in industrial applications

Johnson Matthey Technology Review is published by Johnson Matthey Plc.

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License. You may share, copy and redistribute the material in any medium or format for any lawful purpose. You must give appropriate credit to the author and publisher. You may not use the material for commercial purposes without prior permission. You may not distribute modiﬁ ed material without prior permission.

The rights of users under exceptions and limitations, such as fair use and fair dealing, are not affected by the CC licenses. www.technology.matthey.com www.technology.matthey.com

Johnson Matthey’s international journal of research exploring science and technology in industrial applications

Contents Volume 62, Issue 2, April 2018

132 Editorial: Electrochemistry at Johnson Matthey By Jonathan Sharman

134 Rechargeable Multi-Valent Metal-Air Batteries By Laurence J. Hardwick and Carlos Ponce de León

150 Challenges and Opportunities in Fast Pyrolysis of Biomass: Part II By Tony Bridgwater

161 Lithium Recovery from Aqueous Resources and Batteries: A Brief Review By Ling Li, Vishwanath G. Deshmane, M. Parans Paranthaman, Ramesh Bhave, Bruce A. Moyer and Stephen Harrison

177 All-Solid-State Batteries and their Remaining Challenges By Jitti Kasemchainan and Peter G. Bruce

181 “Electrochemistry: Volume 14” Reviewed by John Blake, Angus Dickinson and Massimo Peruffo

185 The International Flow Battery Forum 2017 Reviewed by Marion van Dalen and Julia O’Farrelly

189 Effect of Temperature and Catholyte Concentration on the Performance of a Chemically Regenerative Fuel Cell By David B. Ward and Trevor J. Davies

204 21st International Conference on Solid State Ionics Reviewed by Thomas Bartlett and James Cookson

208 Johnson Matthey Highlights

211 Inter-Diffusion of Iridium, Platinum, Palladium and Rhodium with Germanium By Adrian Habanyama and Craig M. Comrie

231 Toward Platinum Group Metal-Free Catalysts for Hydrogen/Air Proton-Exchange Membrane Fuel Cells By Frédéric Jaouen, Deborah Jones, Nathan Coutard, Vincent Artero, Peter Strasser and Anthony Kucernak https://doi.org/10.1595/205651318X696855 Johnson Matthey Technol. Rev., 2018, 62, (2), 132–133

www.technology.matthey.com

Guest Editorial Electrochemistry at Johnson Matthey

Historically, Johnson Matthey has had a long volumetric or mass-specific energy density. Metal- association with electrochemistry, and perhaps air cells are attractive here, because there is no this was inevitable because of the importance need for a material to store charge at the cathode, of the platinum group metals (pgm) to Johnson which contributes mass and volume to the cell; Matthey’s early development. Platinum, in rather the charge is supplied by the reduction of particular, has been incredibly useful in the field, oxygen from the air on an electrocatalyst, as in a because of its exceptional electrocatalytic activity fuel cell (known as an air, or oxygen, de-polarised and impressive inertness in most environments. electrode). One of the key challenges for metal- Famously, Johnson Matthey supplied the platinum air cells is coping with direct contact between for Grove’s 1839 experiment that demonstrated the electrolyte and the air, as this can lead to the fuel cell effect (1, 2). After a rather long evaporation, oxidative degradation or, in the case incubation period, this is now paying dividends: of alkaline electrolytes, loss of conductivity by Johnson Matthey has a business unit designing carbonation. The latter effect has tended to limit and manufacturing membrane electrode the application of zinc-air cells to single-discharge assemblies for proton exchange membrane (PEM) applications. New solid electrolytes will be needed fuel cells, including direct methanol fuel cells, for the development of compact, robust metal-air and electrodes for phosphoric acid fuel cells. systems. Recently, applications for electrochemistry have been undergoing something of a renaissance Fuel Cells Without Platinum and Johnson Matthey is active in a diverse range of areas where electrochemistry plays a part, Following the pioneering work of researchers such a number of which are covered in this issue of as Jean-Pol Dodelet, oxygen reduction catalysts Johnson Matthey Technology Review. for fuel cells have been developed that contain no pgms. Whilst many problems remain to be solved Batteries for these (typically) transition metal-doped carbon materials, application in real products is starting The scope of electrochemical activities within to take place. Johnson Matthey is active in this Johnson Matthey is now much broader than field, working with a strong team in a European fuel cells, with lithium-ion batteries the focus of project called CRESCENDO and members of this intense research, development and investment, team have reviewed the field in this issue (3). motivated by the prospect of a very large market Lower cost is an obvious attraction for non-pgm for electric vehicles. The drive for increased energy catalysts, but care is needed here; in 2015 a US density, safety, charging rate and resilience to Department of Energy (DOE)-sponsored cost capacity loss during charge and discharge has led analysis for an automotive system showed that to multiple research strands for each component expensive precursors, processing costs and the of these batteries. The size of the market means need for two stacks made the non-pgm system that workers are already looking at extraction of 28% more expensive than the platinum-based lithium from new sources and at routes to recycle system (4). Higher active site density, to give the lithium, and many of the other elements within thinner catalyst layers with reduced mass the cells, especially those such as cobalt that show transport losses, and improved stability are price-sensitivity as volumes ramp up. Alternative two of the main ambitions for these types of chemistries are also in the frame for transport catalyst, but the insensitivity to certain gas-phase applications, where they hold promise of higher contaminants is also of interest.

Power to Chemicals and Storage largest, best-established industrial electrochemical processes, which makes chlorine and caustic soda Fuel cells and batteries represent the conversion from salt water and electricity (7). The use of GDEs of chemical energy to electrical energy, but the as oxygen-depolarised electrodes for chlor-alkali prospect of increasing quantities of electrons saves hundreds of millivolts, leading to 30% lower from renewable electricity sources such as wind energy costs, and avoids the need to deal with and solar power is fuelling interest in the reverse hydrogen gas that is the normal byproduct. When process, to produce useful chemicals. Alternatively, it comes to consumer products however, efficiency electrochemical systems can be used to store this is a less certain guide and convenience and other energy until it is needed. factors matter; the current domination of the For storage, lithium-ion batteries can be used transportation market by the internal combustion en masse to provide highly dynamic capacity for engine has not come about because these engines buffering electricity grids over short time periods. are efficient. Advocates of redox flow batteries point to the It is an exciting time to be working in industrial capacity fade of lithium-ion and suggest that over, electrochemistry, not least because of the say, 20 years, a flow battery is a better investment. linkages with the growth of renewable energy. A summary of a recent conference is in this issue Electrochemistry is inherently multi-disciplinary, (5) as is an article describing a fuel cell that uses bringing in aspects of chemistry, physics, materials the same type of redox technology, as a route to a science, electrical engineering and chemical pgm-free cathode (6). engineering, and Johnson Matthey, with its diverse For conversion of power to chemicals, in the technical workforce and network of business units, simplest case, water can be electrolysed to form is well placed to act on many of the developments oxygen and hydrogen, with the hydrogen used and opportunities covered in this issue. Watch out to power fuel cells, feed into gas grids or for for progress updates in future issues of Johnson established applications. PEM electrolysers are Matthey Technology Review. growing in abundance because of their robustness to intermittent operation and ability to ramp up quickly to high current densities, compared to JONATHAN SHARMAN traditional liquid, alkaline systems. Better oxygen Johnson Matthey, Blounts Court, Sonning evolution reaction (OER) electrocatalysts than Common, Reading, RG4 9NH, UK iridium oxide are desirable and Johnson Matthey Email: [email protected] recently completed a European funded project on PEM electrolysers that encompassed this challenge. References More active catalysts were developed, but they had lower stability in the very harsh, high potential, 1. C. F. Schœnbein, Phil. Mag., 1839, 14, (85), 43 oxidising environment of the electrolyser anode. 2. W. R. Grove, Phil. Mag., 1839, 14, (86–87), 127 Good progress was made however, by significantly 3. F. Jaouen, D. Jones, N. Coutard, V. Artero, P. reducing hydrogen crossover in thin, reinforced Strasser and A. Kucernak, Johnson Matthey membranes. Technol. Rev., 2018, 62, (2), 231 Synthesis of other chemicals via electrochemical 4. B. D. James, ‘2015 DOE Hydrogen and Fuel methods is also being investigated. For example, Cells Program Review: Fuel Cell Vehicle and Bus Johnson Matthey has just started an ambitious Cost Analysis’, Strategic Analysis Inc, Arlington, European project on the electrochemical Virginia, USA, 10th June, 2015 conversion of gaseous carbon dioxide to methanol. 5. M. van Dalen and J. O’Farrelly, Johnson Matthey Success requires selective, efficient conversion Technol. Rev., 2018, 62, (2), 185 of course, and the design and synthesis of new 6. D. B. Ward and T. J. Davies, Johnson Matthey electrocatalysts is the key. A few millivolts matter Technol. Rev., 2018, 62, (2), 189 greatly in electrochemical conversions. For power to 7. J. Kintrup, M. Millaruelo, V. Trieu, A. Bulan and E. chemicals, each extra millivolt required adds to the S. Mojica, Electrochem. Soc. Interface, 2017, 26, cost of generation, as do parasitic side reactions. (2), 73 Efficiency is vital in industrial electrochemistry, as it has a direct bearing on cost. A recent example of this is the deployment of gas diffusion electrodes (GDE) for the chlor-alkali process, one of the

www.technology.matthey.com

Rechargeable Multi-Valent Metal-Air Batteries A review of research and current challenges in secondary multivalent metal-oxygen batteries

By Laurence J. Hardwick* hereafter referred to as metal-O2 batteries, could Stephenson Institute for Renewable Energy, play an increasingly important role as an energy Department of Chemistry, The University of storage technology in consumer electronics, Liverpool, Peach Street, Liverpool, L69 7ZD, UK electric vehicles, stationary storage and defence, because of their high specific energy (energy Carlos Ponce de León# per unit weight) and energy density (energy per Faculty of Engineering and the Environment, unit volume). A schematic representation of the

University of Southampton, Highfield, rechargeable metal-O2 cell is shown in Figure 1.

Southampton, SO17 1BJ, UK In general, for non-aqueous metal-O2, on discharge, metal ions formed at the metallic anode Email: *[email protected]; are transported across the electrolyte and into the # [email protected] pores of the air cathode. O2 from the atmosphere enters the cathode, and dissolves into the electrolyte within the pores. It is then reduced at Rechargeable metal-oxygen cells could exceed the porous carbon electrode surface by electrons the stored energy of today’s most advanced from the external circuit and combines with the lithium-ion cells. However challenges exist that metal ion (Mn+) from the electrolyte, leading to must be overcome to bring this technology the formation of a solid metal oxide (MxOy) as the into practical application. These challenges final discharge product. Remarkably the reaction include, among others, the recharge and is, to varying degrees, reversible. The metal cyclability efficiency, materials development and oxide can be electrochemically oxidised, releasing improvements in fundamental understanding oxygen gas, thus making this an energy storage of the electrochemistry and chemistry inside device. the cell. The common challenges for the anode, In aqueous metal-O2, because the solubility of including corrosion, passivation and dendrite oxygen in water is low at atmospheric pressure formation and those for the air cathode and the (0.2 mM, at 25°C, 1 atm.) it is necessary to use electrolyte are summarised in this review for oxygen in the gas phase not the liquid. Oxygen cells based on magnesium, calcium, aluminium, from the atmosphere diffuses into the porous silicon, zinc and iron. carbon electrode, known as the gas diffusion electrode, by difference in pressure of oxygen 1. Introduction between the outside and inside of the cell. Catalysts facilitate the reduction of oxygen to Unlike a conventional battery where the reagents hydroxyl ions (in an alkaline electrolyte), with are contained within the cell, metal-air batteries electrons generated from the oxidation of the utilise dioxygen (O2) from the atmosphere, and metal anode. This may lead to the precipitation of consequently can be thought of as a battery-fuel a solid metal hydroxide (Mx(OH)y) within the cell. cell hybrid. Rechargeable metal-air batteries, The process is known as a three-phase reaction:

(a) (b)

Electrocatalyst supported on C black e– e– e– e– Electrocatalyst Hydrophobic layer Air cathode and current collector

Metal Metal anode anode + – M O2 OH O2

Electrolyte Aqueous electrolyte

1 n+ – – – M M + ne M + yOH 1 M(OH)y + ye MxOy n+ – y Gas diffusion electrode, ca. 0.5 mm xM + ne + /2O2 1 MxOy 1 – – /2O2 + 2e + H2O 1 2OH

Fig. 1. Typical metal-O2 batteries: (a) non-aqueous; and (b) aqueous

catalyst (solid), electrolyte (liquid) and oxygen as it will either be precipitated out by hydroxide (gas). A comparison of the phase boundaries of ions, or form a negatively charged complex, in aqueous and non-aqueous metal-O2 batteries is the vicinity of the anode (for example the zincate 2– shown in Figure 2. anion: Zn(OH)4 ), which will never reach the The reactions in non-aqueous and aqueous cathode. metal-O2 cells are very different, and as Despite the recent major research activity such present differing challenges to enable into rechargeable Li-O2 cells in particular (1–3), their technological realisation, which will be metal-O2 batteries have in fact been under discussed further within this review. Notably in consideration from as early as 1868, with the first non-aqueous electrolytes the oxygen reduction demonstration of a Zn-O2 battery by Leclanché at the cathode results in the formation of a metal (4). Since this time some metal-O2 technologies oxide precipitate within the porous cathode itself. have moved from a purely academic interest to However, in alkaline aqueous electrolyte the a commercial product, most notably the primary multivalent metal ion does not reach the cathode, Zn-air battery.

(a) (b) Organic electrolyte Aqueous + Air O2 electrolyte M O2 + O2 M O2 ion MxOy

– e– e Solid Solid

Fig. 2. Phase boundaries for (a) aqueous; and (b) non-aqueous metal-O2 batteries

A substantial amount of work has been carried Table I, where they are compared with those out on metal-O2 batteries, but among the present for present day Li-ion cells. In the case of the –1 challenges is the fundamental understanding and non-aqueous Li-O2 cell, a value of 11,586 Wh kg subsequent control of the electrochemistry and is often quoted; however this is based on the mass chemistry that takes place within metal-O2 cells of Li metal alone. All metal-O2 cells gain mass (O2) and the discovery of new materials that deliver as they discharge, so the mass of O2 should be the desired superior lifetime and performance (at included for all metal-O2 systems. a competitive cost). There are already a number The claim is often made that the increase in specific of highly detailed reviews and articles on Li-O2 energy, in particular over conventional battery

(1, 5–8) and Na-O2 (9, 10), as well as primary systems (Li-ion), is that the reactant dioxygen does Zn-air systems (11). This review focuses on the not have to be carried within the cell. However, this recent work in rechargeable multivalent (+2, is a misconception. The leap forward in theoretical

+3, +4) metal-O2 cells, their operation, energy specific energy on migrating from Li-ion to either storage capability and the challenges they face in aqueous or non-aqueous metal-O2 (using Li-O2 as realisation of becoming a practical technology. the example) arises because Li2O2 in the cathode

stores more Li, and hence charge, than say LiCoO2 2. Theoretical Energy Storage per unit mass and Li metal stores more charge per unit mass than a graphite (C6Li) anode. Theoretical –1 The theoretical specific energies (Wh kg ) energy densities for aqueous metal-O2 cells are (gravimetric energy densities) and energy also attractive. The theoretical energy density densities (Wh dm–3) (volumetric energy densities) can be calculated by taking into account both the for a range of metal-O2 batteries are given in volume of the metallic anode in the charged state

Table I Data for Conventional Batteries Electrochemical Reactions for Common Metal-O2 Couples that Form the Basis of Energy Storage Devices

Cell Theoretical Theoretical energy Battery voltages, specific energy, density, Wh dm–3 (active V Wh kg–1 components in brackets)

Li-ion Today ½C6Li + Li0.5CoO2 = 3C + LiCoO2 3.8 387 1015

Li metal/LiCoO2 ½Li + Li0.5CoO2 = LiCoO2 3.9 534 2755

Li-O (aq) 2234 2 3.2 3582 2Li + ½O2 + H2O = 2LiOH (Li + H2O + LiOH)

Li-O2 (non-aq) 2Li + O2 = Li2O2 2.959 3457 3459 (Li + Li2O2) 4Li+ O2 = 2Li2O 2.913 5226 3823 (Li + Li2O)

Mg-O2 (aq) 3.11 2859 1671 (Mg + H2O + Mg(OH)2) a Mg+ ½O2 + H2O = Mg(OH)2 1.2 1103 645 (Mg + H2O + Mg(OH)2)

Ca-O2 (non-aq) Ca + O2 = CaO2 3.38 2514 2403 (Ca + CaO2)

Al-O2 (aq) 2.71 2794 1160 (Al + H2O + Al(OH)3) a 2Al + ³/²O2 + 3H2O = 2Al(OH)3 1.2 1237 514 (Al + H2O + Al(OH)3)

Si-O2 (ionic liquid) 2.39 4264 2492 (Si + SiO2) a Si + O2 = SiO2 1.1 1962 1146 (Si + SiO2)

Fe-O2 (aq) 1.28 764 715 (Fe + H2O + Fe(OH)2) Fe + ½O2 + H2O = Fe(OH)2

Zn-O (aq) 6091 (ZnO) 2 1.65 1086 Zn + ½O2 = ZnO 2316 (Zn + ZnO) a Experimentally recorded potentials on discharge of specific metal-O2 cells – the theoretical energy storage of these systems are subsequently diminished (in italic). Theoretical energy density includes volume of all active components within the cell. Data taken from (1) or calculated directly from open source thermodynamic data.

and the volume of the discharge product (MxOy results from the oxidation of the metal in an or Mx(OH)y) formed from 100% utilisation of the aqueous environment, either to a soluble metal anode. When H2O is involved in the reaction, its ion or complex or to an insoluble oxide, hydroxide volume requirement should also be considered, or other salt. In metal-O2 batteries, the oxidising which lowers maximum energy densities of aqueous agent is commonly dioxygen, proton or water; systems close to present day Li-ion. hence the overall chemical change can occur via The theoretical energy density is also greater for one or more of the reactions in Equations (i)–(iv):

Li-O2 than Li-ion but the gain is not as great as n n n+ – M + /4O2 + /2H2O → M + nOH (i) for specific energy. Certainly if the lithium metal anode would operate successfully in a Li-O cell, 2 M + n/ O + nH+ → Mn+ + n/ H O (ii) then the positive energy storage benefit for the Li 4 2 2 2 metal/LiCoO cell would also be realised (Table I), 2 M + nH+ → Mn+ + n/ H (iii) as the energy density will be comparable to most 2 2 metal-O2 cells. n+ n – M + nH2O → M + /2H2 + nOH (iv)

3. Common Challenges of Oxidation by a proton can be avoided by the use of alkaline aqueous electrolytes (Equation (iv)). Rechargeable Metal-Air Cells An example of corrosion is given for the Fe-O2 3.1 Challenges for the Metallic Anode cell in Equations (v)–(ix). During spontaneous dissolution of iron, the electrode gradually

Metallic anodes have been studied for many becomes coated with a Fe(OH)2 film as a result of decades for numerous alternative battery systems the following corrosion reaction: other than metal-O2, and the main challenges are Fe → Fe2+ + 2e– well understood. The common issues with metallic E = –0.44 V vs. SHE (v) anodes are corrosion, passivation and dendrite c formation (Figure 3). – – 2H2O + 2e → 2OH + H2 Ea = –0.83 V vs. SHE (vi) 3.1.1 Corrosion Overall: Corrosion of the metal anode is the major side reaction that limits the performance and shelf Fe + 2H2O → Fe(OH)2 + H2 life of metal-O2 cells. Most commonly corrosion Ecell= +0.39 V vs. SHE (vii)

Dense M O Electrolyte x y Electrolyte surface layer n+ – – n+ – M O2 + 4e + 2H2O " 4OH Ions M " M + ne

Cathodic e – Anodic zone zone Metal, M Metal, M

Corrosion Passivation

Risk of short circuit Uneven replating of metal Growth of dendritic structures Mn+ + ne – " M

Metal, M Metal, M Metal, M

Dendrite formation

Fig. 3. Schematic showing corrosion, passivation and dendrite formation processes at metal anode

Remembering that for a corrosion reaction to be enough conductivity to deliver electrons to the spontaneous: reaction site efficiently with low overall impedance. High surface areas imply small pores that could Ecell = Ec – Ea > 0 (viii) become blocked by solid metal oxides, such as

–1 CaO2, so a compromise is necessary between high With DG = –nFEcell = –75.5 kJ mol (ix) active surface area and adequate pore size. For

Therefore as DG < 0, the reaction is practical metal-O2 cells, an exterior O2 permeable thermodynamically favourable. membrane may be required to prevent the ingress of water and carbon dioxide, whilst still allowing 3.1.2 Passivation the free passage of oxygen. In general the porous structure of the air cathode is not optimised for Soluble species formed at the air cathode can gas evolution. The mechanical pressures, which migrate across to the anode where they can be the carbon structure is subjected to as a result reduced either chemically or electrochemically to of the evolution of dioxygen produced trying to form a non-conductive layer on the metal surface escape the electrode, will cause mechanical increasing the internal electrical resistance of the breakdown of the electrode, in addition to the cell and preventing metal dissolution. high positive potential of oxygen evolution that leads to carbon corrosion. 3.1.3 Dendritic Formation and In aqueous metal-O2 air cathodes, a homogenous distribution of a nano-sized catalyst is required to Deformation maximise the performance by increasing the round- During the cycling of the metallic anode, the trip efficiency by lowering the voltage gap between metal ion will not necessarily be plated where it charge and discharge processes (12). The structure has been stripped. As a consequence the metal of the air cathode is critical in order to maximise electrode will start to change shape; its surface will power capability. Nonetheless, the risk of mechanical roughen, producing an electrode with an uneven degradation of the air cathode via O2 evolution will thickness. The restructuring of the metal will lead increase at higher current densities. In aqueous to the production of zones in the metal anode, systems gas diffusion electrodes are employed in which become less active, resulting in loss of which the gas is carried in hydrophobic channels performance. After repeated stripping and plating then dissolves in the electrolyte within hydrophilic cycles, shedding of active material occurs leading channels that are in very close proximity. In this to an irreversible loss of capacity. In worse cases way the O2 is transported mainly in the gas phase, the roughened surface results in dendritic growth rather than the much slower diffusion in the liquid. of the metal which can eventually penetrate the Assuming a suitable gas diffusion electrode that cells’ glass fibre separator and then form a short ensures facile O2 transport can be constructed, then a circuit with the air cathode. In aqueous systems high electrode surface area is necessary. The proper this will result in the end of life of the cell and optimisation of the air cathode should lessen the it will need to be replaced. In non-aqueous cells risk of electrode flooding from the electrolyte which the result can be more catastrophic with the would lead to a reduction in oxygen accessibility. short-circuit leading to thermal runaway. A further challenge is to prevent the ingress of

CO2 which will lead to carbonate formation within 3.2 Challenges for the Air Cathode the air cathode through reaction with the alkaline electrolyte. The challenges for the air cathodes for both non-aqueous and aqueous metal-O2 cells are 3.3 Electrolytes for Metal-O2 Batteries shown in Figure 4. For non-aqueous systems a typical air cathode is comprised of a high surface Understanding, controlling and hence eliminating area carbon mixed with a polymeric binder. The side reactions in metal-O2 cells is a significant porous carbon air cathode is required to ensure undertaking. The electrolyte must be stable to O2 a large electrolyte/electrode surface area and and its reduced species, as well as compounds that accommodate the insoluble discharge product form on discharge; it must exhibit sufficient M+

(MxOy), as well as to facilitate oxygen diffusion conductivity, O2 solubility and diffusion to ensure to the reaction site through the cathode film. In satisfactory rate capability, as well as being able to addition, the porous carbon network must provide wet the electrode.

(a) Substrate effects Pore design Electrolyte • Forming the pure desired • Heirarchical pore structure • Wide electrochemical discharge product • Efficient diffusion of 2O into stability window • Minimisation of side pores • Solvent and/or salt stable reactions • Avoid pore blockages against intermediate species • Surface vs. solution mechanism Reaction pathway • Conductivity Voltage gap between • Volatility, evaporation from discharge and charge cell processes H2O, CO2 • O2 solubility, diffusivity Charge O2

O2 permeable external membranes • Allow the passage of O but prevents Potential Potential 2 Discharge the ingress of major contaminants • Stable against electrolyte Capacity • Stable against reduced oxygen species • Prevent evaporation of electrolyte

(b) Electrolyte Catalysts • Drying out through • Sluggish oxygen reactions water evaporation to • Limitations of energy/power open air H2O and efficiency • Air electrode • Optimisation of type, flooding through distribution, loading electrolyte ingress – into air cathode OH O2 pores leading to Side reactions from air decrease in oxygen • Atmospheric CO2 reacts accessibility with alkaline electrolyte CO • Optimisation of gas 2 leading to carbonate diffusion layer and precipitates hydrophobic layer

Fig. 4. Challenges facing: (a) the non-aqueous; and (b) the aqueous air cathode

The characteristics which a suitable metal-O2 cell Developing stable electrolytes that fulfil most, if electrolyte must fulfil are listed below: not all, of the above criteria remains one of the • resistive to nucleophilic attack from superoxide major research challenges in this field. intermediates • able to dissolve the electrolyte salt to sufficient 4. Alkaline-Earth Metal-O2 concentration. In other words it should have a high dielectric constant (e) 4.1 Magnesium-O2 • high ionic conductivity (>1 mS cm–1) in order

to minimise internal resistance Present Mg-O2 research is limited and recent • for liquid electrolytes, it should have low studies have focused on non-aqueous systems, viscosity h, to facilitate mass transport magnesium alloys, suitable stable electrolytes and

• high O2 solubility and diffusivity bifunctional electrocatalysts for oxygen reduction • stable in a wide electrochemical potential (13–18). A recent review article summarises the window research to date (19). • good thermal and chemical stability For the aqueous based system the predicted cell • compatibility with other cell components voltage, according to Equations (x)–(xii), is 3.11 V, • low cost, low toxicity and low flammability. corresponding to a specific energy of 2859 Wh kg–1:

Anode: coworkers (21). Recently Ponrouch et al. (22) demonstrated reversible Ca plating in an organic Mg → Mg2+ + 2e– carbonate solvent with Ca(BF4)2 salt at 100°C Ea = – 2.7 V vs. SHE (x) and this has reignited research in calcium-based Cathode: battery systems. In particular, electrochemical

– – studies on the cathode side have followed (23, 24) O2 +4e +2H2O → 4OH that show some reversibility. However, finding an Ec = 0.401 V vs. SHE (xi) electrolyte that is good from reversible Ca plating/ Total: stripping and oxygen reduction/evolution reactions remains a major task. The challenges facing 2Mg + O2 + 2H2O → 2Mg(OH)2 non-aqueous alkaline-earth metal-O2 cells in Ecell = 3.11 (xii) particular are summarised in Figure 5. As a primary battery, in aqueous solutions, In aqueous-based systems, calcium metal

Mg-O2 suffers from high polarisation resulting corrodes rapidly in water and Ca-O2 batteries in a practical cell voltage of 1.2 V. Despite the use an electrolyte containing 2:1 methanol:water advantages of magnesium being abundant and a ratio to decrease its reactivity. This electrolyte very active lightweight metal with relatively low with 10% (NH4)2SO4 plus 5% LiCl was used with toxicity, the cell has low coulombic efficiency due to a platinum air breathing electrode to produce a the slow kinetics of the oxygen reduction and the cell able to deliver 70 mA cm–2 at 1 V. Due to the high rates of magnesium corrosion. Magnesium can limitations of the air cathode, a high temperature be mechanically replaced, but whilst operational Ca-O2 using a Ca-Si anode alloy in a molten salt suffers from irreversible polarisation and high with 29% mol of CaO-71 mol CaCl2 stabilised with self-discharge rates. In alkaline electrolytes, 5 wt% Zr-O at 850°C has been proposed (25). magnesium anodes passivate, but still show high The charge-discharge processes are associated corrosion rates, therefore aqueous Mg-O2 systems to the reversible formation of CaSi and CaSi2, use a sodium chloride solution. Even in this Equation (xiii): electrolyte, pure magnesium passivates over time 1 2CaSi + /2O2 CaO + CaSi2 (xiii) and therefore some alloys such as Mg/Al/Zn have been proposed to mitigate against this (20). With an open circuit voltage of 2.28 V the battery operated for 52 h with a Faradaic efficiency of 95%.

Other room temperature Ca-O2 batteries use Ca or 4.2 Calcium-O2 Ca alloy in acetonitrile or DMF and gas diffusion Until fairly recently there were no reports of cathode (Co, Ni, Fe, and/or Mn or transition metal reversible Ca metal stripping/plating in an organic porphyrin or phthalocyanines). Cooper et al. solvent. The oxide passivation layer on Ca metal (26) reported the passivation of Ca electrodes in was found to be electronically and ionically 3 mol dm–3 NaOH but the addition of Cl– produced insulating within a seminal study by Aurbach and stable electrodes. The open circuit voltage of the

Discharge Anode e– Cathode Mg and Ca – • Discharge • Formation of e – + products not fully passivating oxide characterised layer • Use of alloys 2+ M “M ” O2 Cell • Optimal system not Electrolyte yet established • Search for stable Non-aqueous Porous Mx(Oy)z electrolytes electrolyte carbon

Fig. 5. Challenges for non-aqueous alkaline-earth metal-air batteries

140 © 2018 Johnson Matthey https://doi.org/10.1595/205651318X696729 Johnson Matthey Technol. Rev., 2018, 62, (2) cell was 3.1 V for <3 mol dm–3 OH–, <0.1 mol of the anode. The competing electrochemical dm–3 Cl– and coulombic efficiency proportional to processes on the Al surface are: the concentration of Na approaching 100%. Based • formation and/or dissolution of an initial Al2O3 on these results the Ca-O2 cell might operate at a and subsequent Al(OH)3 layer thermodynamic efficiency of 35–40% for a cathode • oxidation of Al to Al3+ –2 – polarisation of 0.4 V and 100 mA cm current • formation of corrosion species – Al(OH)4 and density. Al(OH)3

• parasitic corrosion resulting in H2 formation via water reduction. 5. Aluminium-O2 Self-corrosion at open circuit prevents the

The ideal reactions for the Al-O2 battery during storage of wet Al-O2 cells and reduces their discharge are Equations (xiv)–(xvi): discharge efficiency. The corrosion and oxidation Anode: of Al in alkaline electrolytes depends on electrolyte

– – – properties, temperature and purity. Indeed one of Al + 4OH → Al(OH)4 + 3e the methods for suppressing parasitic corrosion Ea = –2.31 V vs. SHE (xiv) whilst maintaining the electrode activity can be Cathode: achieved by employing high purity aluminium

– – (>99.999% purity) or by alloying the aluminium O2 + 2H2O + 4e → 4OH with minor amounts of indium, gallium, zinc, Ec = 0.401 V vs. SHE (xv) magnesium, tin or thallium. This of course will Total: dramatically increase the cost of the anode (29). It is essential to the battery performance that 4Al + 3O2 + 6H2O → 4Al(OH)3 both the Al anode and air cathode can operate at Ecell = 2.71 V (xvi) a current density greater than 100 mA cm−2. With The theoretical specific energy is 2793 –1Wh kg a neutral brine electrolyte, the Al(III) is largely considering the product Al(OH)3 and the cell voltage formed as a solid oxide and/or hydroxide and the is 2.71 V. But in practice the available specific energy performance of the battery depends critically on is lower as the operating voltage is usually around the form of this precipitate; it must not form a 1.1 V. The battery is a primary energy storage passivating film on the aluminium surface nor inhibit device since the electrodeposition of aluminium the air cathode. In addition, the Al material used as from alkaline solutions is not thermodynamically the negative electrode must be stable to corrosion feasible due to its negative standard potential, during battery storage, i.e. the following chemical leading to hydrogen evolution before any aluminium reactions (Equations (xix)–(xx)) should not occur can be deposited. Mechanically recharging has been at open circuit potential or during discharge: proposed by replacing the anode. Alternatively 3+ – 4Al + 3O2 + 6H2O → 4Al + 12OH (xix) hydrogen evolution could be potentially removed by using ionic liquid electrolytes, where some 3+ – 2Al + 6H2O → 2Al + 3H2 + 6OH (xx) studies show good plating reversibility (27).

The performance of an Al-O2 battery is The competing demands of stability to corrosion determined by the electrochemical and corrosion and rapid anodic dissolution makes difficult the properties of the aluminium alloy electrode and identification of the appropriate aluminium alloy. the associated hydrogen gas evolution (28), Several alloys to suppress the hydrogen evolution Equations (xvii)–(xviii): and lower the self-corrosion of aluminium have

– – been identified. Nevertheless the high cost of 2H2O + 2e → 2OH + H2 these high-purity aluminium alloys still restricts E = –0.83 V vs. SHE (xvii) the commercialisation of Al-O2 batteries and its use has been limited. Figure 6 shows typical 2Al + 6H2O → 2Al(OH)3 + 3H2 (xviii) challenges and performance of an Al-O2 battery, Aluminium passivates in air or in aqueous solutions using gel electrolyte, bipolar electrodes and a by formation of an oxide layer of several nanometre flowing electrolyte (30). thicknesses which is detrimental to battery Despite the operational barriers of the aluminium- performance and restricts the ability to achieve air battery mentioned above, a number of studies the reversible potential causing delayed activation dealing with hydrogen evolution and rechargeable

Discharge Anode e– Cathode Problems of aluminium – e – + • Al(OH)3 precipitate • Formation of nm thick blocking pores 3OH– OH– 4OH– oxide layer (Al2O3) • Parasitic corrosion of Al resulting in H evolution O2 2 Al Rechargeability • Al(OH)3 precipitate on – • Mechanical regeneration surface Al(OH)3 Al(OH)4 2H2O of Al Alkaline or Porous neutral carbon + electrolyte catalyst Electrolyte Cell 2.0 • Challenge to reversibly Gel electrolyte • Open circuit potential 1.8 Bipolar cell plate Al in most known Flowing reduced from 2.71 V to ca.

, V electrolyte electrolytes, therefore 1.6 1.6 V due to competing only primary cell cell electrochemical reactions E 1.4 • Practical operating voltages of ca. 1.1 V 1.2 • Self-corrosion at open 1.0 circuit prevents the storage of wet Al-air cells Cell voltage, Cell voltage, 0.8

0.6 1 10 100 Current density, j, mA cm–2

Fig. 6. Challenges for Al-air and typical performance curves for different primary Al-air configurations. Performance curve data provided by C. Ponce de Leon taken from (30) © Elsevier 2013

issues have been reported. Mori (31–33) for system has a specific energy of 4264 Wh –1 kg . It example, used ceramic barriers such as aluminium consisted of a doped silicon anode and carbon- oxide or aluminium tungsten oxide between the based air cathode with the addition of MnO2 as aqueous electrolyte, the aluminium anode and the an oxygen reduction reaction (ORR) catalyst. The air cathode in order to suppress the accumulation of electrolyte employed within the cell was a room byproducts. However, the current densities remained temperature hydrophilic ionic liquid, 1-ethyl-3 low. The same author also suggested aluminium methylimidazolium oligofluorohydrogenate (EMI terephthalate metal-organic framework (MOF) for (HF2.3F)), which is synthesised from EMI Cl and HF. the air cathode and 1-ethyl-3-methylimidazolium The discharge of the cell results in the oxidation chloride as an ionic liquid electrolyte in order to of silicon at the silicon wafer anode and in the obtain more stability over repeated electrochemical reduction of atmospheric oxygen at the air cathode. reactions. Deyab used 1-allyl-3-methylimidazolium The room temperature ionic liquid (RTIL) electrolyte bis(trifluoromethylsulfonyl)imide ionic liquid to anions participate in both electrode reactions and overcome the hydrogen evolution reaction (HER) are therefore key in permitting the operation of the (34). They suggested that 1.5 × 10–3 mol dm–3 battery. added to a 4.0 mol dm–3 NaOH electrolyte resulted The reaction at the anode is Equation (xxi): in an enhanced capacity from 1720 to 2554 mA –1 – – – h g . Other attempts to use aluminium include a Si + 12(HF)2F → 8(HF)3F + SiF4 + 4e (xxi) mechanical rechargeable Al-O2 cell consisting of And the reactions at the cathode are Equations wedge anodes able to reach up to 1500 W h kg–1 (xxii)–(xxiii): energy density (35). – – – O2 + 12(HF)3F + 4e → 2H2O + 16(HF)2F (xxii)

6. Silicon-O2 – – SiF4 + 2H2O + 4(HF)2F → SiO2 + 4(HF)3F (xxiii) A Si-O2 battery was first reported by Ein-Eli and co-workers in 2009 (36) and theoretically the Leading to the overall reaction, Equation (xxiv):

–1 –2 Si + O2 → SiO2 (xxiv) capacity of 1206 mAh g at 0.1 mA cm with an operating voltage of 1 V being reported, Equations Cells tested show an open circuit potential of (xxv)–(xxvii). ca. 1.4 V rather than the theoretically predicted Anode: 2.89 V, the operating potential for this cell is in the – – Si + 4 OH → Si(OH)4 + 4e 0.8–1.1 V region. The role of water in the Si-O2 cell Ea = –1.69 V vs. SHE (xxv) is central in the production of SiO2 at the cathode, see Equation (xxiii). SiO2 formation results in the Cathode: – reaction of H2O with a fluororacidic anion [(HF)2F ] O + 2H O + 4e– → 4OH– and with SiF . Indeed it has been demonstrated 2 2 4 E = 0.401 V vs. SHE (xxvi) that this ionic liquid electrolyte with an addition c of 15 vol% of water can increase cell capacity to Leading to the overall reaction: 72.5 mAh cm–2 from 50 mAh cm–2 for water free → electrolyte, at a current density of 0.3 mA cm–2. Si + O2 + 2H2O Si(OH)4 Ecell = 2.091 V (xxvii) The main challenges of Si-O2 cells are illustrated within Figure 7. In alkaline electrolyte there is a side reaction that Another variation of a Si-O2 cell has also been increases with respect to basicity, Equation (xxviii): considered in which the cell alternatively operates – 2– in an alkaline electrolyte (37–40) with specific Si + 2OH + 2H2O → SiO2(OH)2 + 2H2 (xxviii)

Discharge

Anode e – Cathode Problems of silicon Membrane required – • Passivation of Si from e – + • To maintain optimal water surface accumulation of content within electrolyte Si(OH)4 or SiO2 O2 and to prevent the cell • Self-corrosion of silicon drying out Si “Si4+” O2

Alkaline or Porous O2 ionic liquid carbon + Si(OH) electrolyte catalyst 4 Modified or SiO2 Si surface

• Optimise gas diffusion electrode for Si-O2 cell Electrolyte • Deactivation of catalyst by electrolyte

Limited Si(OH)4 solubility in H2O 1.6 • Si(OH) precipitate 4 1.4 blocking electrode pores 1.2

Cell 1.0 • Practical operating voltage 0.8 of 0.8–1.1 V

Voltage, V Voltage, 0.6 –2 –2 –2 –2 Rechargeability 0.4 300 mA cm 100 mA cm 50 mA cm 10 mA cm 26.7 mAh 15.25 mAh 12.5 mAh 3 mAh • Mechanical regeneration 0.2 of Si 0 0 100 200 300 400 500 600 700 Time, h

Fig. 7. Specific challenges facing the Si-O2 battery and cell performance data using EMI·(HF)2.3F RTIL electrolyte at different constant current densities. Cycling data from figure reproduced from (41) © Elsevier 2010

7. Zinc-O2 offers a tantalising glimpse of its great promise, if the primary technology performance could be With a theoretical specific energy of 1086 Wh kg–1 replicated in a secondary system. The principal the Zn-O2 battery is one of the most studied disadvantage at present is the low cycle life of metal-air systems offering low cost and relatively the cell when electrochemically recharged and a abundant materials (zinc, potassium hydroxide, voltage gap (0.8 V under a load of 20 mA cm–2) carbon, manganese dioxide) (3, 42). The battery between discharge and charge processes. has an open circuit potential of 1.65 V in 6 M Rechargeable Zn-O2 suffers from two key

KOH and results from the elimination of the H2 drawbacks that do not affect the primary cell: evolution reaction. This is due to the presence at the poor stability of the air cathode when used to the surface of the Zn metal electrode, of a thin charge the battery, and as in the earlier example ionically conducting, but electronically insulating of lithium metal, the formation of dendrites layer of ZnO. Water is therefore excluded from on the zinc anode leading to short circuits and the electrochemical interface, thus the effective shedding of zinc. Dendrite formation occurs as stability range of the electrolyte is extended the Zn is replated from an electrolyte containing 2– (Equations (xxix)–(xxxi)): non-uniform zincate, Zn(OH)4 composition Anode: gradients and uneven ZnO at the surface.

– 2– – Gravitational demixing causes the concentration of Zn + 4OH Zn(OH)4 + 2e zincate ions to increase at lower positions, resulting in Ea = –1.199 V vs. SHE (xxix) slight differences in the conductivity of the electrolyte. In solution: Consequently there is a gradual redistribution of

2– – the zinc, so that the lower portions of the electrode Zn(OH)4 ZnO + 2OH + H2O (xxx) become thicker as it is stripped and plated.

Air cathode: Rechargeable Zn-O2 cells require zinc precipitation

– – from the water-based electrolyte to be closely O2 + 2H2O + 4e 4HO controlled. Challenges include dendrite formation, Ec = 0.401 V vs. SHE (xxxi) non-uniform zinc dissolution and limited solubility Leading to the overall reaction, Equation (xxxii): in electrolytes. Electrically reversing the reaction at

1 a bifunctional air cathode, to liberate oxygen from Zn + /2O2 → ZnO Ecell = 1.6 V (xxxii) discharge reaction products, is difficult; membranes Zn is easily recyclable, thereby offering both a tested to date have low overall efficiency. Charging sustainable and low toxic energy storage device. voltage is much higher than discharge voltage, The recycling process is as follows: producing cycle energy efficiency as low as 50%. 1. disassembly of the used Zn source (typically Providing charge and discharge functions by 80% Zn utilisation) separate unifunctional cathodes increases cell size, 2. reaction of the Zn anode oxidation product weight and complexity. (ZnO) with the electrolyte (KOH), Equation A satisfactory electrically recharged system (xxxiii): potentially offers low material cost and high specific energy. Figure 8 summarises the main ZnO + 2KOH + H2O → K2Zn(OH)4 (xxxiii) characteristics and challenges of a Zn-O2 cell and 3. electrowinning of the zincate solution, Equation shows the charge-discharge cycles of a cell with (xxxiv): a hierarchical meso-macroporous lanthanum –2 manganite (LaMnO3) cathode at 25 mA cm K2Zn(OH)4 → Zn + 2KOH + H2O + ½O2 (xxxiv) current density (43). 4. reassembly of the Zn source containing the There has been a recent increase in interest electrowon Zn. in rechargeable Zn-air with a number of studies

Millions of primary Zn-O2 (Zn-air) batteries are reported. In particular, Wu et al. (44) used a sold annually for hearing aids and other medical simple scalable procedure to manufacture porous devices. Commercially available cells can have bifunctional electrodes for the Zn-O2 battery. very high specific energies and energy densities, The polyhedral carbon electrodes reported use a up to 442 Wh kg–1 and 1672 Wh dm–3 respectively Zn-doped Co-based zeolitic imidazolate frameworks for the Duracell® Zn-air button cell. The primary (ZnCo-ZIFs) precursor. The authors state that Zn-air cell is a mature technology, which is not Zn doping produces small Co nanoparticles with currently the case for rechargeable Zn-O2 and high nitrogen content that enhances both the

Discharge

e – Anode Cathode Problems of zinc e– – + • Low cycle energy efficiency as low • Dendrite formation as 50% • Deformation of 4OH– 2OH– • Large voltage gap of 0.8 V at –2 electrode from current loads of 20 mA cm gravitational demixing 2– Membrane required Zn Zn(OH)4 O2 • Blocking CO2 whilst allowing O2 to ZnO + 2OH– pass H2O • To maintain optimal water content + H2O within the electrolyte and prevent Alkaline Porous the cell drying out Electrolyte electrolyte carbon + catalyst • Formation of Zn(CO3)2 from CO2 ingress • Cell drying out from water 3.0 O2 evaporation CO 2.5 2 , V

cell 2.0 Cell E • Voltage gap >1.5 V 1.5 • Large overpotential required for OER

Cell voltage, Cell voltage, 1.0

–2 Zn/O2, ±25 mA cm charge/discharge, 600 s 0.5 0 500 1000 1500 2000 2500 3000 Time, s

Fig. 8. Challenges and performance for rechargeable Zn-O2. Charge, discharge cycle figure reproduced from performance curve data from (43) © Elsevier 2014

–2 evolution and reduction of oxygen. The ZnCoNCx hysteresis between 0.6–1 V at 10–20 mA cm . materials are highly graphitised with a porous These include transition metal hydroxysulfides polyhedral structure and large surface area. The (46), pyrolysised MOFs containing Zn and Co authors reported a cell able to cycle up to 100 (47), Fe3Mo3C-supported IrMn Clusters (48), –2 times at 7 mA cm with only 10 mV difference Ni3FeN-supported Fe3Pt intermetallic nanoalloys between charge and discharge cycles. In another (49), CoO0.87S0.13 (50), carbon nitrides (51) and example, a composite material consisting of Co3O4 Fe0.1Ni0.9Co2O4 spinels (52). nanoparticles decorated with carbon nanofibers (CNFs) was used as a bifunctional electrode for 8. Iron-O2 the Zn-O2 battery. Metal ion containing polymers were electrospun and carbonised to obtain the Alkaline Fe-O2 batteries were under intense composite. The authors claimed that the electrode industrial investigation during the 1970s and was able to yield four electrons during the reduction early 1980s, most notably by The Westinghouse of oxygen, competing with commercial platinum Electric Corporation, USA, and the Swedish electrodes, and that the prototype cell had an National Development Corporation, Sweden, for energy efficiency of 64% when charged/discharged the application of an electric vehicle battery. The –2 at 1 mA cm (45). alkaline Fe-O2 battery has a predicted open circuit Recent reports have shown greatly improved cell voltage of about 1.28 V and a theoretical cyclability under ambient conditions, with 100+ specific energy of 764 Wh kg–1 with the formation cycles being routinely presented. A number of of Fe(OH)2 (53). catalysts have shown bifunctionality in order Iron electrodes have been used for over 70 years in to overcome the sluggish kinetics of the oxygen Ni/Fe batteries, and the notable advantages of this evolution reaction (OER) and ORR with voltage system are the availability, abundance and low cost

145 © 2018 Johnson Matthey https://doi.org/10.1595/205651318X696729 Johnson Matthey Technol. Rev., 2018, 62, (2) of iron. Iron electrodes are fabricated into sintered anodic dissolution reaction, poor low-temperature structures. The iron electrode is formed from high performance, and low charge acceptance at high purity iron oxide powder in order to increase the ambient temperatures due to a parasitic HER. The overpotential for H2 evolution. However, this adds challenges and performance of a Fe-O2 cell are significant cost to electrode fabrication. The Fe-O2 shown in Figure 9. battery substitutes the nickel electrode for an air More recently a solid electrolyte based Fe-O2 cell breathing electrode which in theory could increase has been reported (56, 57). An oxide ion conductor the specific energy by 100%. is employed as the electrolyte and the cell operates The reactions at the anode are complex and at temperatures greater than 600°C. The solid still a matter of discussion and involve several oxide iron-air battery utilises the FeO-Fe phase intermediates. The reaction is also complicated by equilibrium as a means of storing electrical-chemical hydrogen evolution if the metallic iron corrodes to energy via a series of H2/H2O-mediated reversible

Fe(II), forming Fe(OH)2, and the electrode gradually electrochemical-chemical looping reactions. The becomes coated with a film of Fe(OH)2 as a result of capacity and efficiency characteristics of this the following reactions, Equations (xxxv)–(xxxviii): battery were found to be strongly dependent on the degree of iron utilisation: higher charge Fe → Fe2+ + 2e– E° = –0.447 V vs. SHE (xxxv) and energy storage capacity, but lower round-

– – trip efficiency, can be produced at higher iron Fe + 2OH Fe(OH)2 + 2e E° = –0.88 V vs. SHE (xxxvi) utilisation, and lower charge and energy storage capacity, but higher round trip efficiency, can be – – produced at lower iron utilisation. 2H2O + 2e → 2OH + H2 E° = –0.8277 V vs. SHE (xxxvii) A bifunctional carbon based air electrode containing 0.5 mg cm–2 loading of 30% palladium on carbon, Total reaction: reported by McKerracher et al. (58), was stable after 1000 charge/discharge cycles at 10 mA cm–2 in an Fe + 2H2O → Fe(OH)2 + H2 (xxxviii) Fe-O2 alkaline battery. At higher current densities If the HER, Equation (xxxviii), can be avoided (20–80 mA cm–2), the electrode showed better the efficiency of the Fe-O2 battery is greater. The performance than some commercially produced formation of Fe(OH)2 is followed by oxidation to platinum on carbon gas diffusion electrodes. The

Fe(III) forming magnetite, Fe3O4, Equation (xxxix): authors suggested that the better performance is

– – due to the well-dispersed, nanoscale Pd particles. 3Fe(OH)2 + 2OH Fe3O4 + 4H2O + 2e Figueredo-Rodríguez et al. (59) reported an Fe-O2 Ea = –0.76 V vs. SHE (xxxix) battery with a specific energy of 764 W h kg−1 Fe The reaction at the cathode is the reduction of and 453 W h kg−1 Fe energy density with a charge oxygen, Equation (xl): capacity of 814 mA h g−1 Fe when it was cycled at a −2 – – current density of 10 mA cm . The cell employed ½O2 + H2O + 2e → 2OH nanocomposite iron electrodes manufactured by Ec = 0.401 V vs. SHE (xl) iron oxide synthetised via a molten salt fusion The overall reaction where magnetite (as method followed by mixing with a high surface area opposed to Fe(OH)2) is the discharge product is carbon by ball milling. The gas diffusion electrodes Equation (xli): comprised three main parts bound together by hot pressing at 180°C and 250 kPa: 3Fe + 2O2 → Fe3O4 Ecell = 1.15 V (xli) 1. a gas diffusion layer composed of 80 wt% high The manufacturing of the iron electrode involves surface area (ca. 64 m2 g−1) carbon mixed with additives such as bismuth and sulfur to minimise 20 wt% polytetrafluoroethylene the corrosion of the iron electrode and the evolution 2. a catalyst layer formed by 30 wt% Pd/C catalyst of hydrogen in an attempt to favour Equations in a 5 wt% Nafion solution, and

(xxxix) and (xl). The formation of magnetite Fe3O4 3. an expanded nickel mesh current collector. Equation (xli) leads to a lower cell voltage. Problems that adversely affect the performance 9. Conclusion of the iron anode are spontaneous corrosion in the charged state (leading to a high rate of Modern society requires energy storage devices self-discharge), a low faradaic efficiency for the with much higher levels of energy storage than

Discharge

Anode e – Cathode Problems of iron • Low cycle energy efficiencies, – • Parasitic corrosion of Fe e – + as low as 40% resulting in H evolution • Large voltage gap of 0.8 V at 2 –2 6OH– 8OH– current loads of 25 mA cm • Fe(OH)2 precipitate on surface

• Low faradaic efficiency for 3Fe(OH)2 O anode dissolution reaction Fe 2 +2OH–

4H O Fe3O4+ 4H2O 2 Alkaline Porous electrolyte carbon + catalyst Electrolyte 1.8 • Cell drying out from water evaporation 1.6

Cell , V 1.4 (a) (b)

• To maintain optimal water cell content within electrolyte and E 1.2 to prevent the cell drying out 1.0

0.8 Cell voltage, Cell voltage, Charge 25 mA cm–2 0.6 Discharge 30 mA cm–2 Charge/discharge at + 40 mA cm–2 0.4 0 2 4 6 8 10 12 14 16 18 Time, h

Fig. 9. Challenges and performance for Fe-O2 and comparison of two charge-discharge cell potential curves –2 –2 for an Fe-O2 battery: (a) 25 mA cm charge discharged at 30 mA cm of a prototype stack containing 190 air and 95 iron electrodes of 10 × 20 cm divided into 19 cells connected in series (54); (b) charge and discharge at 40 mA cm–2, the cell contained air electrodes of 100 cm2 area separated by 3–4 mm from the Fe electrode (55). Performance curve data provided by C. Ponce de Leon

ever before. Rechargeable metal-O2 cells are a cell is not proof of true reversibility. Major gaps amongst the few contenders that can exceed in fundamental knowledge of all types of metal-O2 the stored energy of the current state-of-the-art cells remain, for example the realisation of Li-ion cells. Multivalent metal-air (O2) batteries electrolyte and materials composition that provide described herein possess significant promise for reversible stripping and plating of metal anodes high energy storage applications, but in general and reversible oxide/hydroxide production and most examples (with the exception of primary decomposition. As a result, much ground-breaking zinc-air that has been commercialised for several and exciting science remains to be done. years) are a long way from being a technological product. Fundamental challenges in all aspects of the cell, anode, electrolyte and cathode, need to References be addressed. The major issue for aqueous based systems is the ability to recharge and cycle at high 1. P. G. Bruce, S. A. Freunberger, L. J. Hardwick and efficiency, to be resolved by the development of J.-M. Tarascon, Nature Mater., 2012, 11, 19 ever more effective bifunctional OER and ORR 2. G. Girishkumar, B. McCloskey, A. C. Luntz, S. catalysts. Swanson and W. Wilcke, J. Phys. Chem. Lett., Rechargeable batteries need to be based on 2010, 1, (14), 2193 truly reversible reactions and the ability to cycle 3. J.-S. Lee, S. T. Kim, R. Cao, N.-S. Choi, M. Liu, K.

T. Lee and J. Cho, Adv. Energy Mater., 2011, 1, 25. N. U. Pujare, K. W. Semkow and A. F. Sammells, J. (1), 2 Electrochem. Soc., 1988, 135, (1), 260 4. G. L. Leclanché, ‘Improvement in Combining 26. J. F. Cooper and P. K. Hosmer, ‘The Behavior of Generating and Secondary or Accumulating the Calcium Electrode in Aqueous Electrolytes’, Galvanic Battery’, US Patent, 64,113; 1867 Abstract 54, 152nd Meeting, The Electrochemical 5. D. Aurbach, B. D. McCloskey, L. F. Nazar and P. G. Society, Atlanta, USA, 9th–14th October, 1977, Bruce, Nature Energy, 2016, 1, (9), 16128 p. 25

6. A. C. Luntz and B. D. McCloskey, Chem. Rev., 27. Y. Zhao and T. J. VanderNoot, Electrochim. Acta, 2014, 114, (23), 11721 1997, 42, (11), 1639

7. X. Zhang, X.-G. Wang, Z. Xie and Z. Zhou, Green 28. M. Nestoridi, D. Pletcher, J. A. Wharton and R. J. K.

Energy Environ., 2016, 1, (1), 4 Wood, J. Power Sources, 2009, 193, (2), 895

8. F. Li and J. Chen, Adv. Energy Mater., 2017, 7, 29. Q. Li and N. J. Bjerrum, J. Power Sources, 2002,

(24),1602934 110, (1), 1 30. D. R. Egan, C. Ponce de León, R. J. K. Wood, R. 9. C. L. Bender, D. Schröder, R. Pinedo, P. Adelhelm L. Jones, K. R. Stokes and F. C. Walsh, J. Power and J. Janek, Angew. Chem. Int. Ed., 2016, 55, Sources, 2013, 236, 293 (15), 4640 31. R. Mori, RSC Adv., 2014, 4, (57), 30346 10. P. Adelhelm, P. Hartmann, C. L. Bender, M. Busche, C. Eufinger and J. Janek,Beilstein J. Nanotechnol., 32. R. Mori, J. Electrochem. Soc., 2015, 162, (3), 2015, 6, 1016 A288 11. Y. Li, M. Gong, Y. Liang, J. Feng, J.-E. Kim, H. Wang, 33. R. Mori, RSC Adv., 2017, 7, (11), 6389 G. Hong, B. Zhang and H. Dai, Nature Commun., 34. M. A. Deyab, Electrochim. Acta, 2017, 244, 178

2013, 4, 1805 35. A. R. Despić, J. Appl. Electrochem., 1985, 15, (2), 12. Y. Liu, Q. Sun, W. Li, K. R. Adair, J. Li and X. Sun, 191 Green Energy Environ., 2017, 2, (3), 246 36. G. Cohn, D. Starosvetsky, R. Hagiwara, D. D. 13. T. Zhang, Z. Tao and J. Chen, Mater. Horiz., 2014, Macdonald and Y. Ein-Eli, Electrochem. Commun., 1, (2), 196 2009, 11, (10), 1916 14. T. Khoo, A. Somers, A. A. J. Torriero, D. R. 37. X. Zhong, H. Zhang, Y. Liu, J. Bai, L. Liao, Y. Huang MacFarlane, P. C. Howlett and M. Forsyth, and X. Duan, ChemSusChem, 2012, 5, (1), 177 Electrochim. Acta, 2013, 87, 701 38. Y. E. Durmus, Ö. Aslanbas, S. Kayser, H. Tempel, 15. N. Wang, R. Wang, Y. Feng, W. Xiong, J. Zhang F. Hausen, L. G. J. de Haart, J. Granwehr, Y. Ein- and M. Deng, Corrosion Sci., 2016, 112, 13 Eli, R.-A. Eichel and H. Kungl, Electrochim. Acta, 16. S. Yuan, H. Lu, Z. Sun, L. Fan, X. Zhu and W. 2017, 225, 215 Zhang, J. Electrochem. Soc., 2016, 163, (7), 39. A. Garamoun, M. B. Schubert and J. H. Werner, A1181 ChemSusChem, 2014, 7, (12), 3272 17. H. Xiong, K. Yu, X. Yin, Y. Dai, Y. Yan and H. Zhu, 40. D.-W. Park, S. Kim, J. D. Ocon, G. H. A. Abrenica, J. Alloys Comp., 2017, 708, 652 J. K. Lee and J. Lee, ACS Appl. Mater. Interfaces,

18. G. Vardar, J. G. Smith, T. Thompson, K. Inagaki, 2015, 7, (5), 3126 J. Naruse, H. Hiramatsu, A. E. S. Sleightholme, J. 41. G. Cohn and Y. Ein-Eli, J. Power Sources, 2010, Sakamoto, D. J. Siegel and C. W. Monroe, Chem. 195, (15), 4963 Mater., 2016, 28, (21), 7629 42. J. Fu, Z. P. Cano, M. G. Park, A. Yu, M. Fowler and 19. C.-S. Li, Y. Sun, F. Gebert and S.-L. Chou, Adv. Z. Chen, Adv. Mater., 2017, 29, (7), 1604685 Energy Mater., 2017, 7, (24), 1700869 43. Y.-C. Lee, P.-Y. Peng, W.-S. Chang and C.-M. Huang, 20. R. P. Hamlen, E. C. Jerabek, J. C. Ruzzo and E. J. Taiwan Inst. Chem. Eng., 2014, 45, (5), 2334 G. Siwek, J. Electrochem. Soc., 1969, 116, (11), 44. X. Wu, G. Meng, W. Liu, T. Li, Q. Yang, X. Sun and 1588 J. Liu, Nano Res., 2018, 11, (1), 163 21. D. Aurbach, R. Skaletsky and Y. Gofer, J. 45. B. Li, X. Ge, F. W. T. Goh, T. S. A. Hor, D. Geng, G. Electrochem. Soc., 1991, 138, (12), 3536 Du, Z. Liu, J. Zhang, X. Liu and Y. Zong, Nanoscale, 22. A. Ponrouch, C. Frontera, F. Bardé and M. R. 2015, 7, (5), 1830 Palacín, Nature Mater., 2016, 15, 169 46. H.-F. Wang, C. Tang, B. Wang, B.-Q. Li and Q. 23. P. Reinsberg, C. J. Bondue and H. Baltruschat, J. Zhang, Adv. Mater., 2017, 29, (35), 1702327 Phys. Chem. C, 2016, 120, (39), 22179 47. B. Chen, X. He, F. Yin, H. Wang, D.-J. Liu, R. Shi, 24. T. Shiga, Y. Kato and Y. Hase, J. Mater. Chem. A, J. Chen and H. Yin, Adv. Funct. Mater., 2017, 27, 2017, 5, (25), 13212 (37), 1700795

48. Z. Cui, Y. Li, G. Fu, X. Li and J. B. Goodenough, 1978, 2, (3), 287 Adv. Mater., 2017, 29, (40), 1702385 55. H. Cnobloch, Proceedings of the 88th Convention 49. Z. Cui, G. Fu, Y. Li and J. B. Goodenough, Angew. of the Battery Council International Future Clean Chem. Int. Ed., 2017, 56, (33), 9901 Silent Power, Mexico City, Mexico, 25th–29th April, 1976, pp. 39–48 50. J. Fu, F. M. Hassan, C. Zhong, J. Lu, H. Liu, A. Yu and Z. Chen, Adv. Mater., 2017, 29, (35), 56. A. Inoishi, S. Ida, S. Uratani, T. Okano and T. 1702526 Ishihara, RSC Adv., 2013, 3, (9), 3024

51. W. Niu, Z. Li, K. Marcus, L. Zhou, Y. Li, R. Ye, K. 57. X. Zhao, N. Xu, X. Li, Y. Gong and K. Huang, RSC Liang and Y. Yang, Adv. Energy Mater., 2018, 8, Adv., 2012, 2, (27), 10163 (1), 1701642 58. R. D. McKerracher, C. Alegre, V. Baglio, A. S. Aricò, C. Ponce de León, F. Mornaghini, M. Rodlert and F. 52. Y.-T. Lu, Y.-J. Chien, C.-F. Liu, T.-H. You and C.-C. C. Walsh, Electrochim. Acta, 2015, 174, 508 Hu, J. Mater. Chem. A, 2017, 5, (39), 21016 59. H. A. Figueredo-Rodríguez, R. D. McKerracher, M. 53. K. Vijayamohanan, T. S. Balasubramanian and A. Insausti, A. G. Luis, C. Ponce de León, C. Alegre, V. K. Shukla, J. Power Sources, 1991, 34, (3), 269 Baglio, A. S. Aricò and F. C. Walsh, J. Electrochem. 54. L. Öjefors and L. Carlsson, J. Power Sources, Soc., 2017, 164, (6), A1148

The Authors

Laurence Hardwick is Professor of Electrochemistry and Director of the Stephenson Institute for Renewable Energy within the Department of Chemistry at the University of Liverpool, UK. His research interests focus on understanding reaction mechanisms of metal-air batteries and the development of surface sensitive in situ electrochemical Raman and infrared spectroscopies to probe battery electrode-electrolyte interfaces.

Carlos Ponce de León Albarrán is Senior Lecturer, Energy Technology Group within Engineering and the Environment at the University of Southampton, UK. He is working from an electrochemical engineering perspective in aqueous metal-air batteries, water treatment, metal ion removal, characterisation of novel electrode materials, electrochemical strategies for pollution control and redox flow cells for energy conversion.

www.technology.matthey.com

Challenges and Opportunities in Fast Pyrolysis of Biomass: Part II Upgrading options and promising applications in energy, biofuels and chemicals

By Tony Bridgwater • An appreciation of the potential for fast Bioenergy Research Group, European pyrolysis as a pretreatment method, i.e. for Bioenergy Research Institute, Aston University, bio-oil to be an effective energy carrier Birmingham B4 7ET, UK • Greater interest in bio-oil as a precursor for second generation biofuels for transport Email: [email protected] • Greater awareness of the potential for fast pyrolysis and bio-oil to offer more versatile process routes to a wider range of products Fast pyrolysis for liquids has been developed and contribute to biorefinery concept in recent decades as a fast and flexible method development to provide high yields of liquid products. An • Considerably greater interest in upgrading overview of this promising field is given, with a bio-oil sufficiently for it to be used for heat, comprehensive introduction as well as a practical power and other applications with greater guide to those thinking of applying bio-oils or fast confidence by users. pyrolysis liquids in various applications. It updates Figure 1 summarises the possibilities for the literature with recent developments that have applications for bio-oil and the main developments occurred since the reviews cited herein. Part I are expanded below. gave an introduction to the background, science, feedstocks, technology and products available for fast pyrolysis (1). Part II details some of the promising applications as well as pre-treatment and bio-oil upgrading options. The applications include use of bio-oil as an energy carrier, precursor Conversion Process Product to second generation biofuels, as a biorefinery Process Separation Chemicals concept and upgrading to fuels and chemicals. heat Gas Conversion Biofuels Fast Liquid 1. Applications of Bio-Oil Engine,turbine Electricity pyrolysis bio-oil Bio-oil can substitute for fuel oil or diesel in many Combustion Process Heat applications including boilers, furnaces, engines Char Boiler and turbines for electricity generation which was heat thoroughly reviewed in 2004 (2). Although many Char aspects have not changed very much, the most Fig. 1. Applications for fast pyrolysis products significant changes since then include:

1.1 Pretreatment Method for Energy 1.2 Co-Firing Carrier Co-processing of biomass with conventional fuels Biomass is a widely dispersed resource that has is potentially a very attractive option that enables to be harvested, collected and transported to full economies of scale to be realised as well as the conversion facility. The low bulk density of reducing the problems of product quality and clean biomass, which can be as low as 50 kg m–3, means up. Most current co-firing applications are those that transport costs are high and the number of where the biomass fuels are added to the coal vehicle movements for transportation to a large feed and this is widely practised at up to 5% on scale processing facility are also very high, with the energy demand of the power station. A few consequent substantial environmental impacts. applications involve conversion to a fuel gas via Conversion of biomass to a liquid by fast pyrolysis gasification followed by close coupled firing to the at or near the biomass source will reduce transport power station boiler. There are also some successful costs and reduce environmental concerns as examples of co-firing fast pyrolysis liquids in coal the liquid has a density of 1200 kg m–3 – more fired and natural gas fired power stations (4, 5). than ten times higher than low density crops and residues. This not only reduces the number 1.3 Fast Pyrolysis Based Biorefinery of vehicle movements and costs by up to 87%, it also reduces costs of handling and transportation While biorefineries are not new, the recognition of by virtue of it being a liquid that can be pumped. their strategic and economic potential is recent. This leads to the concept of small decentralised A biorefinery can be defined as the optimised fast pyrolysis plants of 50,000 to 250,000 tonnes performance of the use of biomass for materials, per year for production of liquids to be transported chemicals, fuels and energy applications, where to a central processing plant. It is also possible to performance relates to costs, economics, markets, consider mixing the byproduct char with the bio-oil yield, environment, impact, carbon balance and to make a slurry to improve the energy content social aspects. In other words, there needs to be of the product, but the pyrolysis process will then optimised use of resources, maximised profitability, require that its process energy needs are met from maximised benefits and minimised wastes. another source. The large majority of chemicals are manufactured Adoption of decentralised fast pyrolysis with from petroleum feedstocks. Only a small proportion transportation of the resultant liquid to a central of the total oil production, around 5%, is used gasification and fuel synthesis plant has both tech in chemical manufacture but the value of these nical and economic advantages and disadvantages chemicals is high and contributes a comparable as summarised in Table I. The impact of inclu revenue to fuel and energy products. There is a sion of fast pyrolysis as a pretreatment step on clear economic advantage in building a similar biomass-to-liquids (BTL) cost and performance has flexibility into the biofuels market by devoting part been analysed (3). of the biomass production to the manufacture of

Table I Comparison of Bio-Oil Gasification to Solid Biomass Gasification to Generate Syngas (3)

Impact on capital cost Impact on overall Bio-oil vs. solid biomass and product cost process performance

Transport costs for bio-oil Lower Higher

Handling and storage costs for bio-oil Lower None

Very low alkali metals in bio-oil Lower Higher

Liquid bio-oil feeding to a gasifier, Lower Higher particularly pressurised

Changed gas cleaning requirements when Lower Higher using bio-oil

Need for additional fast pyrolysis process Higher Lower

Other processing Fig. 2. Fast pyrolysis based biorefinery Hydro- treating

Hydrogen separation

Vapour Cracking Chemicals Fast pyrolysis Liquid Gasification

Transport Slurry Synthesis Refining Char fuels

Electricity Gas Heat chemicals. In fact, this concept makes even more have taken place in recent years concerning sense in the context of biomass because it is approaches to upgrading, especially for biofuels. chemically more heterogeneous than crude oil and There are a number of objectives for upgrading of conversion to fuels, particularly hydrocarbons, is which the main ones are: not so cost effective. Figure 2 shows fast pyrolysis (a) Improvement of bio-oil quality to overcome or at the heart of a biorefinery. reduce one or more of the quality deficiencies A key feature of the biorefinery concept is the (summarised in Part I, Table III (1)) co-production of fuels, chemicals and energy. As (b) Production of chemicals explained earlier, there is also the possibility of (c) Removal of oxygen to provide hydrocarbon gasifying biomass to make syngas, a mixture of biofuels. hydrogen and carbon monoxide for subsequent synthesis of hydrocarbons, alcohols and other 2.1 Bio-Oil Quality Improvement chemicals. However, this route is energy intensive so much of the energy content of the biomass is lost The most important properties that inhibit in the processing. Therefore, electricity generation widespread use of bio-oil are: may be the most efficient use of biomass (6). • Phase separation from use of wet feedstock Since the empirical chemical composition and/or secondary cracking of vapours leading of biomass, approximately (CH2O)n, is quite to high water content in the liquid product. different from that of oil (CH2)n, the range of Phase separation cannot be reversed except primary chemicals that can be easily derived through relatively high additions of co-solvents from biomass and oil are quite different. Hence, such as ethanol any biomass chemical industry will have to • Incompatibility and immiscibility with be based on a different selection of simple conventional fuels from the high oxygen content ‘platform’ chemicals than those currently used in of the bio-oil the petrochemical industry. Since the available • High solids content that affect catalysts and biomass will inevitably show major regional utilisation in engines and burners differences, it is quite possible that the choice • High viscosity that hinders pumping and of platform chemicals derived from biomass will combustion and which cannot readily be show much more geographical variation than in controlled by raising temperature as for heavy petrochemical production. fossil fuels due to temperature sensitivity • High water content that lowers heating value 2. Bio-Oil Upgrading but also lowers viscosity • Chemical or thermal instability which limits Bio-oil can be upgraded in a number of ways: the use of higher temperatures for controlling physically, chemically and catalytically. While this properties has been extensively reviewed (2, 7–10), some • High acidity leading to corrosion in storage and interesting and potentially important developments utilisation.

2.1.1 Filtration (15). The University of Florence, Italy, has worked on emulsions of 5 to 95% bio-oil in diesel (16–18) Hot-vapour filtration can reduce the ash content of to make either a transport fuel or a fuel for power the oil to less than 0.01% and the alkali content generation in engines that does not require engine to less than 10 ppm, much lower than reported modification to dual fuel operation. There is limited for biomass oils produced in systems using only experience of using such fuels in engines or burners, cyclones. This gives a higher quality product but substantially higher levels of corrosion/erosion with lower char (11), however accumulated char were observed in engine applications compared to on the filter medium is catalytically active and bio-oil or diesel alone, sometimes to the extent potentially cracks the vapours, reduces yield by up of limiting operation to less than 1 hour. A further to 20%, reduces viscosity and lowers the average drawback of this approach is the cost of surfactants molecular weight of the liquid product. There is that provide longer term stability and the high limited information available on the performance energy required for emulsification. or operation of hot vapour filters, but they can be specified and perform similarly to hot gas filters in 2.1.4 Blends gasification processes. Diesel engine tests performed on crude and on More recently, some success has been achieved hot-filtered oil showed a substantial increase in through production of blends of bio-oil with a variety burning rate and a lower ignition delay for the latter, of co-solvents and other sustainable or green fuels due to the lower average molecular weight for the as well as conventional transport fuels. Bio-oil filtered oil (12). Hot gas filtration has not yet been by itself is considered too demanding for simple demonstrated over a long-term process operation. direct use due to acidity, ageing, particulates and A little work has been done in this area by the US incompatibility with fossil fuels. Therefore some Department of Energy National Renewable Energy exploratory work was initiated in 2012 to produce Laboratory (NREL), USA and VTT Energy, Finland homogenous blends of bio-oil with bio-diesel and (11), and by Aston University (13), but very little an alcohol co-solvent – both ethanol and butanol has been published. (19). A key result was that single phase and stable Liquid filtration to very low particle sizes of blends of bio-oil, biodiesel and either ethanol or below around 5 µm is very difficult due to the butanol could be prepared which utilised the physico-chemical nature of the liquid and usually whole bio-oil including the water content. Areas of requires very high pressure drops and self-cleaning miscibility and non-miscibility were identified and filters, although improvement is claimed with filter the work was published (19). The shorter term pores of around 10 µm. objective is to address ferry needs rather than intercontinental shipping and also to satisfy the 2.1.2 Solvent Addition new requirements for low sulfur fuels. A key requirement is to maximise the use of Polar solvents have been used for many years to bio-oil, maximise the sustainability of the resultant homogenise and reduce the viscosity of biomass blend by use of renewable solvents, and satisfy oils. The addition of solvents, especially methanol, marine oil specifications, of which flash point showed a significant effect on the oil stability. above 60°C is key. Subsequently, the early work Diebold and Czernik (14) found that the rate of was extended to consider diesel and marine gasoil viscosity increase (‘ageing’) for the oil with 10 wt% as hydrocarbons in a four-component blend. The of methanol was almost twenty times less than second phase of this work is nearing completion for the oil without additives. Use of co-solvents to after testing a wide range of co-solvents. compatibilise bio-oil with other sustainable liquid fuels as blends is covered below. 2.2 Chemicals

2.1.3 Emulsions Although bio-oil contains in excess of 1000 individual chemicals, few are present in sufficient Pyrolysis oils are not miscible with hydrocarbon concentrations to justify recovery. This has been fuels but they can be emulsified with diesel oil reviewed by Radlein (20). The largest single with the aid of surfactants. A process for producing component in bio-oil is in fact water. Other stable micro-emulsions with 5–30% of bio-oil in chemicals of value include food flavouring often diesel has been developed at CANMET, Canada known as ‘liquid smoke’ (21) and until recently

153 © 2018 Johnson Matthey https://doi.org/10.1595/205651318X696738 Johnson Matthey Technol. Rev., 2018, 62, (2) the only commercial application of fast pyrolysis • Gasification to syngas followed by synthesis to was production of liquid smoke by Red Arrow hydrocarbons or alcohols. of Wisconsin, USA. Hydroxacetaldehyde or glycolaldehyde can be isolated from bio-oil (22) 2.3.1 Hydrodeoxygenation and is considered as the most reactive browning compound participating in the Maillard reactions Hydro-processing of liquid bio-oil rejects oxygen (23). Levoglucosan, an anhydrosugar, can be as water by catalytic reaction with hydrogen. This readily recovered in high purity and high yield is a separate and distinct process to fast pyrolysis but until recently was perceived as having limited that can therefore be carried out remotely. market value. This has been reviewed (24). More The process is typically high pressure (up to attention is currently being paid to the potential for 20 MPa) and moderate temperature (up to 400°C) hydrolysis to sugars (25). High purity acetic acid is and requires a hydrogen supply or source (27). recovered from slow pyrolysis liquids by Profagus Most attention is now focused on multiple step in Germany (26), together with other chemicals processes with increasingly severe conditions when market conditions are right. starting with a stabilisation step to improve temperature stability followed by more orthodox 2.3 Hydrocarbon Biofuels hydrotreating. More or less full hydrotreating gives a naphtha-like product that requires orthodox Direct production of high yields of liquids by fast refining to derive conventional transport fuels. This pyrolysis inevitably caused attention to focus on would be expected to take place in a conventional their use as biofuels (sustainable transport fuels) refinery to take advantage of know-how, existing to supplement and replace fossil fuel derived processes and economies of scale. A projected transport fuels. However, the high oxygen content typical yield of naphtha equivalent from biomass of bio-oil and non-miscibility or incompatibility with is about 20% by weight or 55% in energy terms hydrocarbon fuels has prevented simple adoption excluding provision of hydrogen (9). Inclusion of of bio-oil as a transport fuel. hydrogen production adds a significant inefficiency The main methods for upgrading bio-oil to due to use of biomass to generate hydrogen for transport fuels are summarised in Figure 3: example by gasification and shifting the CO. This • Hydrodeoxygenation of bio-oil to a substantially reduces the yields to around 15 wt% or 40% in de-oxygenated product energy terms. The process can be depicted by the • Catalytic vapour cracking of fast pyrolysis following conceptual reaction (Equation (i)): vapours (i.e. close coupled) to aromatics that CH1.32O0.43 + 0.77 H2 → CH2 + 0.43 H2O (i) can be followed by hydrodeoxygenation and/or introduction into a refinery The catalysts originally tested in the 1980s and • Partial upgrading by hydrodeoxygenation 1990s were based on sulfided cobalt molybdenum followed by introduction into a refinery or nickel molybdenum supported on alumina • Direct introduction of crude bio-oil into a or aluminosilicate and the process conditions refinery are similar to those used in the desulfurisation

Indirect routes Direct routes Fig. 3. Overview of fast pyrolysis Biomass upgrading methods

Fast pyrolysis

Liquid bio-oil Gasification Cracking Syngas Modify, Blends e.g. esters Conversion,e.g. Hydro- Fischer-Tropsch process

Refining

Alcohols Hydrocarbons, Chemicals Fuels SNG, diesel, gasoline, etc.

154 © 2018 Johnson Matthey https://doi.org/10.1595/205651318X696738 Johnson Matthey Technol. Rev., 2018, 62, (2) of petroleum fractions. However a number of recovered and recycled as only a fraction of the fundamental problems arose including that hydrogen would be utilised due to the need for the catalyst supports of typically alumina or high hydrogen partial pressures. Recovery and aluminosilicates were found to be unstable in the recycling of unused hydrogen is both technically high water content environment of bio-oil and the and economically challenging. sulfur was stripped from the catalysts requiring There is increasing interest in supercritical constant re-sulfurisation. The main activities were processing of bio-oil to either improve the based at Pacific Northwest National Laboratory properties of bio-oil or to de-oxygenate it to a (PNNL), USA by Elliott (28–30) and at Université hydrocarbon fuel. The supercritical fluids studied catholique de Louvain (UCL) in Louvain la Neuve include water, CO2, methanol, ethanol, butanol in Belgium by Maggi et al. (31, 32). This area has and cyclohexane using traditional CoMo type been thoroughly reviewed (7). A recent design catalysts, precious metals such as platinum, study of this technology for a biomass input of palladium and ruthenium on inert supports such 2000 dry tonnes per day for production of gasoline as carbon or cracking catalysts including HZSM-5. and diesel has been carried out by PNNL (33). The results are mixed with no clear conclusions on A comprehensive review of unsupported metal the efficacy of this route. High pressures are still sulfide hydrotreating catalysts was published in required as well as recovery of the fluids involved. 2007 (34). Many researchers report an improvement in More recently, attention turned to precious bio-oil properties such as lower acid levels and higher metal catalysts on less susceptible supports, and esters, but there has been a disappointing absence considerable academic and industrial research has of significant moves towards real hydrocarbon been carried out. Of note is the work by UOP in bio-fuels. There continues to be an interest in use Chicago, USA, (now Honeywell UOP) with PNNL of model compounds even though it is impossible in the USA to address the scientific and technical to adequately represent the complexity of bio-oil challenges and develop a cost effective process with single compounds or even groups of so called (35). Model compounds were used initially to representative compounds. understand the basic processes (36) and both whole oil and fractions have been evaluated. 2.3.2 Catalytic Vapour Cracking Tests have been carried out on both batch and (Close Coupled) continuous flow processes focussing on an initial low temperature stabilisation step followed by Cracking, usually over zeolites, rejects oxygen as more extensive catalytic de-oxygenation using CO2, as well as water, summarised in the conceptual different metal catalysts and processing conditions overall reaction below (Equation (ii)): to give a range of products including petroleum C H O + 0.26 O → 0.65 CH + refinery feedstock. Remaining challenges include 0.99 1.32 0.43 2 1.2 0.34 CO2 + 0.27 H2O (ii) complete deoxygenation especially of phenols without saturation with hydrogen. The process takes place in two stages: firstly A key aspect is production of hydrogen. Since the cracking which deposits carbon or coke on the hydrogen requirement is significant, it should be catalyst surface, which is then burned off in renewable and sustainable. Few refineries have a second reaction. In this case the oxygen is a hydrogen surplus, so this has to be provided. ultimately mostly rejected as CO2, with some water, There are many ways of providing hydrogen such from burning off the carbon on the coked catalyst. as gasification of biomass followed by shifting to H2 This lowers the carbon efficiency of the process then scrubbing CO2. Product bio-oil or the aqueous compared to hydrodeoxygenation, but avoids the phase from a phase separated product can be need for hydrogen and pressure. steam reformed to hydrogen; or hydrogen can be Cracking takes place at atmospheric pressure generated locally by electrolysis of water preferably either with in situ catalyst or in a close coupled using renewably produced electricity. Supply process. There is no requirement for hydrogen or of hydrogen from external sources is unlikely to pressure. The projected yield is around 18 wt% be feasible due to very high cost of storage and aromatics and the process is understood to be transport. The necessary purity of hydrogen is the basis for the recently abandoned Kior process unknown, but some CO shifting may take place in (37, 38). This process is believed to have been the hydroprocessing reactor removing the need for based on a first stage of zeolite cracking, possibly dedicated shift reactors. The high cost of hydrogen modified with metals, followed by hydrotreating means that unused hydrogen would have to be to deliver hydrocarbon transport fuels. Although a

155 © 2018 Johnson Matthey https://doi.org/10.1595/205651318X696738 Johnson Matthey Technol. Rev., 2018, 62, (2) large demonstration plant was built in Mississippi, miscibility of bio-oil with conventional refinery deliveries of products were consistently below streams and the potential for catastrophic blockage claims and expectations, and the projected of a non-mixed bio-oil component in upgrading yields were not met and appeared optimistically processes such as hydrocracking. Therefore high. A similar process is under development by partial hydrodeoxygenation to an upgraded and Anellotech (39). hydrocarbon miscible product was seen as one Early work by NREL added a close coupled of the more attractive solutions. However, due secondary reactor to the fast pyrolysis process to immiscibility, simple addition of bio-oil to any in which vapours passed through a close coupled refinery stream would lead to two phase flow into fixed bed of ZSM catalyst (40). This has the whatever upgrading process was selected. Since all advantage of providing independent control of the upgrading processes operate at moderate to high temperature and residence of pyrolysis vapours temperatures, this will result in phase separation over the catalyst. of the bio-oil above around 100°C and subsequent Among cracking catalysts, ZSM-5 has attracted polymerisation at higher temperatures leading to most attention due to its shape selectivity to blockage of the preheaters or the upgrading unit aromatics, with promoters such as gallium or which would be very costly to remedy. Since bio-oil nickel (41). A key disadvantage is that the catalyst hydrodeoxygenation is conventionally carried out rapidly cokes which requires frequent regeneration in stages, partial hydrodeoxygenation will require as in a fluid catalytic cracking (FCC) unit in a less hydrogen and result in a lower cost process. conventional refinery. Oxygen is thus substantially One problem is defining how much oxygen has removed as CO and CO2 (as well as H2O) compared to be removed for miscibility and secondly how to solely H2O in hydrodeoxygenation. Production of the miscibility of an upgraded bio-oil which can aromatics is also likely to be of significant interest be black, can be measured when mixed with a to the chemicals sector, where aromatics are the conventional black refinery stream. This approach second largest global petrochemicals sector. has not progressed very far for both these reasons A complementary approach is to incorporate as well as substantial doubts about the extent of cracking catalysts in the pyrolysis reactor which development required in the refinery. However, offers a more compact reaction system, but here is still good reason to believe that this compromises have to be made between optimum approach offers significant potential. pyrolysis conditions and optimum catalysis conditions. This area has attracted much increased 2.3.4 Direct Addition of Bio-Oil to a interest in recent years. Although some advantages Fluid Catalytic Cracking Unit result in improvements to yield and quality of liquids, the catalyst has to operate at the same The realisation that conventional refineries temperature as pyrolysis (or vice versa) and the provide an enormous asset in fuel processing and necessary contact times for fast pyrolysis are not production with their technical know-how and optimal for catalytic cracking. However this could economies of scale, has led to wider consideration operate as the first step in a multi-stage process of partial upgrading to a refinery compatible followed by secondary vapour processing utilising material intermediate for subsequent refinery process conditions more suitable for vapour phase processing. cracking. This approach offers technical and A report by Hydrocarbon Processing for the future economic advantages especially when combined of FCC and hydroprocessing in modern refineries with catalyst development and is the approach states that: “Biomass-derived oils are generally best adopted by Inaeris Technologies, USA (42); this upgraded by HZSM-5 or ZSM-5, as these zeolitic also allows for recycling and processing of used catalysts promote high yields of liquid products and catalyst. propylene. Unfortunately, these feeds tend to coke easily, and high TANs and undesirable byproducts 2.3.3 Partial Upgrading by such as water and CO2 are additional challenges” (43). It was recognised that some upgrading may Hydrodeoxygenation Followed by be necessary prior to introduction of bio-oil (44). Introduction into a Refinery Integration into refineries by upgrading through Direct incorporation of bio-oil into a refinery was cracking or hydrotreating has been reviewed by long thought to be unacceptable due to the poor Huber and Corma (45).

Most attention focused on hydrotreating bio-oil in Figure 2. Although there is a small energy as a means of reducing the oxygen content to a penalty from the lower pyrolysis energy efficiency, level that is compatible with refinery operation transportation energy and additional bio-oil or is substantially or totally miscible with refinery gasification stage, this is more than compensated streams or has sufficiently low oxygen content to by the economies of scale achievable on a be stable and temperature insensitive. Another commercial sized gasification and transport fuel approach has been deoxygenation of bio-oil synthesis plant (3). over zeolites followed by hydrodeoxygenation as Although the concept of very large gasification practiced by KiOR Inc, USA, but this activity has plants of 5 GW or more has been promoted (50) now been wound up. based on importation of biomass on a massive One approach is to introduce raw bio-oil into the scale to an integrated plant, there are significant riser of a FCC unit. As bio-oil cannot be preheated obstacles to be overcome. Decentralised fast without decomposing and blocking the feeder, it pyrolysis plants of up to 100,000 tonnes per has to be fed cold which imposes an additional year or 12 tonnes per hour are currently feasible thermal load. In addition, when the bio-oil and close to being commercially realised. Bio-oil contacts hot regenerated catalyst it will evaporate gasification in an entrained flow oxygen blown and crack into vapours and char or coke. There is pressurised gasifier is also feasible such asa therefore a significant loss of efficiency in carbon Texaco or Shell system, with the added advantage utilisation of potentially 35% from Equation (ii) and that feeding a liquid at pressure is easier than solid an increased coke burden in the FCC regenerator. biomass, offers lower costs and the gas quality Hydrodeoxygenation tends to retain the majority under such conditions is likely to be higher than of the carbon while losing the oxygen as water, but from solid biomass. Future Energy (now Siemens, there is a significant inefficiency if the hydrogen is Germany) has successfully conducted pressurised generated from biomass, such as by gasification oxygen blown gasification tests on both bio-oil and and shifting. However the concept appears bio-oil/char slurries (51, 52) and this approach, promising as the requirement for independent known as the bioliq process, is under development upgrading is obviated. Some preliminary results at Karlsruhe Institute of Technology (KIT) in have been published (46, 47). Historically, one Germany where bio-oil/char slurries from a twin approach to upgrading bio-oil over zeolites was screw pyrolysis reactor are gasified and converted re-evaporation of bio-oil and passing the vapours to biofuels (53). through zeolites. This work was pioneered by There is increased interest in smaller scale Bakhshi et al. and this provides some insight into economic synfuels technology such as the Velocys® , the potential of this approach as well as the effect USA, microchannel reactor (54). This is claimed of different catalysts (48, 49). to overcome the scale problems of conventional The ultimate ambition of most work in this area is Fischer Tropsch technology in that economic a ‘one-pot’ approach where full deoxygenation to operation is feasible at small scale making it more an acceptable product would be accomplished. To suitable to biomass based processes. date this has not been as successful with low yields and extensive byproducts requiring utilisation or 2.3.6 Other Methods and Routes disposal. There appears to be a realisation that fast pyrolysis is a crude primary conversion step A wide variety of methods and catalysts have been giving an unusually complex product that is likely investigated in recent years with some examples to be best processed in an optimised sequence of listed below. Many are attempts to conflate carefully considered conversion steps. different processes, reactions and catalysts to move towards the ‘one-pot’ approach mentioned 2.3.5 Gasification of Bio-Oil for earlier, but most have misjudged the chemical complexity of bio-oil. It is important to emphasise Synfuels the importance of maximising yield and minimising A recent concept that has attracted much interest unwanted reactions especially minimising residues is the decentralised production of bio-oil or bio-oil/ since these will have to be disposed of at a potential char slurries for transportation to a central process cost as well as lowering efficiency: plant for gasification and synthesis of hydrocarbon • Acid cracking in supercritical ethanol transport fuels, for example by Fischer Tropsch • Aqueous-phase reforming + dehydration + synthesis or alcohols. This is depicted above hydrogenation

• Dicationic ionic liquid such as C6(mim)2−HSO4 References • Esterification of pyrolysis vapours 1. A. V. Bridgwater, Johnson Matthey Technol. Rev., • Esterification of liquid bio-oil 2018, 62, (1), 118 • Hydrogenation-esterification over bifunctional Pt catalysts 2. S. Czernik and A. V. Bridgwater, Energy Fuels, • Reactive distillation 2004, 18, (2), 590 2– • Solid acid catalysts such as 40SiO2/TiO2−SO4 3. A. V. Bridgwater, ‘Technical and Economic

• Solid base catalysts such as 30K2CO3/Al2O3- Assessment of Thermal Processes for Biofuels’, NaOH Report for Life Cycle and Techno-Economic • Steam reforming over orthodox and more Assessment of the Northeast Biomass to Liquids unusual catalysts Project, NNFCC Project 08/018, Consultants on Process Engineering Ltd, Solihull, UK, June, 2009 • ZnO, MgO and Zn-Al and Mg-Al mixed oxides. 4. D. Chiaramonti, A. Oasmaa and Y. Solantausta, Renew. Sust. Energy Rev., 2007, 11, (6), 1056 3. Conclusions 5. Q. Dang, M. M. Wright and R. C. Brown, Environ. The liquid bio-oil produced by fast pyrolysis has Sci. Technol., 2015, 49, (24), 14688 the considerable advantage of being storable and 6. B. Kamm, P. R. Gruber and M. Kamm, ‘Biorefineries transportable, as well as having the potential to – Industrial Processes and Products’, in “Ullmann’s supply a number of valuable chemicals. In these Encyclopedia of Industrial Chemistry”, Wiley-VCH respects it offers a unique advantage. Fast pyrolysis Verlag GmbH & Co KGaA, Weinheim, Germany, has some basic and essential requirements if good 2007 yields of good quality bio-oil are to be derived. The 7. R. E. Maggi and D. C. Elliott, ‘Upgrading Overview’, potential of bio-oil is increasingly being recognised, in “Developments in Thermochemical Biomass with a rapid growth in research into improving bio Conversion”, eds. A. V. Bridgwater and D. G. oil properties particularly for dedicated applications B. Boocock, Springer Science+Business Media and for biofuel production. Much of the research is Dordrecht, The Netherlands, 1997, pp. 575–588 still at a fundamental scale even to the use of model 8. A. V. Bridgwater, Catal. Today, 1996, 29, (1–4), compounds and mixtures of model compounds that 285 purport to represent whole bio-oil. It is doubtful 9. A. V. Bridgwater, Appl. Catal. A: Gen., 1994, 116, if a limited component mixture can adequately (1–2), 5 represent the complexity of bio-oil. Some of the 10. Q. Zhang, J. Chang, T. Wang and Y. Xu, Energy most interesting and potentially valuable research Convers. Manage., 2007, 48, (1), 87 is on more complex and more sophisticated 11. J. P. Diebold, S. Czernik, J. W. Scahill, S. D. Philips catalytic systems and these will require larger and C. J. Feik, ‘Hot-Gas Filtration to Remove Char scale development to prove feasibility and viability. from Pyrolysis Vapours Produced in the Vortex Biorefineries offer considerable scope for Reactor at NREL’, NREL/CP-430-7215, Biomass optimisation of fast pyrolysis based processes Pyrolysis Oil Properties and Combustion Meeting, and products, and these will require development Estes Park, USA, 26th–28th September, 1994, pp. of component processes in order to optimise an 90–108 integrated system. They will necessarily include 12. A. L. Shihadeh, ‘Rural Electrification from Local provision of heat and power for at least energy self Resources: Biomass Pyrolysis Oil Combustion sufficiency. in a Direct Injection Diesel Engine’, PhD There is an exciting future for both fast pyrolysis Thesis, Department of Mechanical Engineering, and bio-oil upgrading as long as these are focused Massachusetts Institute of Technology, on delivering useful and valuable products. Massachusetts, USA, September, 1998 13. J. Sitzmann and A. V. Bridgwater, ‘Upgrading Acknowledgement Fast Pyrolysis Oils by Hot Vapour Filtration’, 15th European Energy from Biomass Conference & This is a revised and updated version of an Exhibition, Berlin, Germany, 7th–11th May, 2007, original text published by Taylor and Francis (55). pp. 1378–1383 Reproduced with permission from Taylor and 14. J. P. Diebold and S. Czernik, Energy Fuels, 1997, Francis Group LLC Books. 11, (5), 1081

15. M. Ikura, S. Mirmiran, M. Stanciulescu and H. “Energy from Biomass and Wastes X”, ed. D. L. Sawatzky, Natural Resources Canada, “Pyrolysis Klass, IGT, Chicago, USA, 1987, pp. 765–784 Liquid-in-Diesel Oil Microemulsions”, US Patent, 28. E. G. Baker and D. C. Elliott, ‘Catalytic Upgrading 5,820,640; 1998 of Biomass Pyrolysis Oils’, in “Research in 16. P. Baglioni, D. Chiaramonti, M. Bonini, I. Thermochemical Biomass Conversion”, eds. A. Soldaini and G. Tondi, ‘Bio-Crude-Oil/Diesel V. Bridgwater and J. L. Kuester, Elsevier Science Oil Emulsification: Main Achievements of the Publishers Ltd, London, UK, 1988, pp. 883–895 Emulsification Process and Preliminary Results 29. E. G. Baker and D. C. Elliott, ‘Catalytic Hydrotreating of Tests on Diesel Engine’, in “Progress in of Biomass-Derived Oils’, in “Pyrolysis Oils from Thermochemical Biomass Conversion”, ed. A. V. Biomass: Producing, Analyzing and Upgrading”, Bridgwater, Blackwell Science Ltd, Oxford, UK, eds. J. Soltes and T. A. Milne, ACS Symposium 2001, pp. 1525–1539 Series, Vol. 376, American Chemical Society, 17. D. Chiaramonti, M. Bonini, E. Fratini, G. Tondi, K. Washington, DC, USA, 1988, pp. 228–240 Gartner, A. V. Bridgwater, H. P. Grimm, I. Soldaini, 30. D. C. Elliott and G. N. Neuenschwander, ‘Liquid A. Webster and P. Baglioni, Biomass Bioenerg., Fuels by Low-Severity Hydrotreating of Biocrude’, 2003, 25, (1), 85 in “Developments in Thermochemical Biomass 18. D. Chiaramonti, M. Bonini, E. Fratini, G. Tondi, K. Conversion”, eds. A. V. Bridgwater and D. G. B. Gartner, A. V. Bridgwater, H. P. Grimm, I. Soldaini, Boocock, Vol. 1, Springer Science+Business A. Webster and P. Baglioni, Biomass Bioenerg., Media, Dordrecht, The Netherlands, 1997, pp. 2003, 25, (1), 101 611–621

19. A. Alcala and A. V. Bridgwater, Fuel, 2013, 31. R. Maggi and B. Delmon, ‘Characterization of 109, 417 Bio-Oils Produced by Pyrolysis’, in “Advances in 20. D. Radlein, ‘The Production of Chemicals from Fast Thermochemical Biomass Conversion”, ed. A. V. Pyrolysis Bio-Oils’, in “Fast Pyrolysis of Biomass: Bridgwater, Vol. 1, Springer Science+Business A Handbook”, CPL Press, Newbury, UK, 1999, pp. Media, Dordrecht, The Netherlands, 1993, pp. 164–188 1086–1094 21. ‘Food Flavouring (Smoke)’, Task 34, IEA Bioenergy, 32. P. Grange, E. Laurent, R. Maggi, A. Centeno and B. International Energy Agency, Paris, France: Delmon, Catal. Today, 1996, 29, (1–4), 297 http://task34.ieabioenergy.com/food-flavoring/ 33. S. B. Jones, J. E. Holladay, C. Valkenburg, D. J. (Accessed on 22nd February 2018) Stevens, C. W. Walton, C. Kinchin, D. C. Elliott 22. ‘Hydroxyacetaldehyde’, Task 34, IEA Bioenergy, and S. Czernik, “Production of Gasoline and Diesel International Energy Agency, Paris, France: http:// from Biomass via Fast Pyrolysis, Hydrotreating task34.ieabioenergy.com/hydroxyacetaldehyde/ and Hydrocracking: A Design Case”, PNNL-18284, (Accessed on 22nd February 2018) Pacific Northwest National Laboratory, Richland, 23. J. E. Hodge, J. Agric. Food Chem., 1953, 1, (15), Washington, USA, February, 2009, 76 pp 928 34. S. Eijsbouts, S. W. Mayo and K. Fujita, Appl. Catal. 24. ‘Levoglucosan: Historical Developments and A: Gen., 2007, 322, 58 Current Status of Levoglucosan Production and 35. ‘UOP and Ensyn to Form Joint Venture to Uses’, Task 34, IEA Bioenergy, International Offer Second Generation Biomass Technology’, Energy Agency, Paris, France: http://task34. Honeywell UOP, Illinois, USA, 10th September, ieabioenergy.com/levoglucosan/ (Accessed on 2008 22nd February 2018) 36. D. C. Elliott and T. R. Hart, Energy Fuels, 2009, 25. P. Blanco-Sanchez, J. Lad and A. V. Bridgwater, 23, (2), 631 ‘Glucose Production from Anhydrosugars by Acid 37. J. Lane, ‘KiOR: The Inside True Story of a Company Hydrolysis’, 6th International Symposium on Energy from Biomass and Waste, International Gone Wrong’, Biofuels Digest, Miami, USA, 17th Waste Working Group, Venice, Italy, 14th–17th May, 2016 November, 2016 38. K. Fehrenbacher, ‘A Biofuel Dream Gone Bad’, 26. ‘Acetic Acids: Made from Beechwood’, proFagus Fortune, Time Inc, New York City, USA, 4th GmbH, Bodenfelde, Germany: http://profagus. December, 2015 de/en/acetic-acids/ (Accessed on 9th November, 39. ‘Technology’, Anellotech: http://anellotech.com/ 2017) technology (Accessed on 28th January 2018) 27. D. C. Elliott and E. G. Baker, ‘Hydrotreating 40. J. P. Diebold and J. W. Scahill, Energy Prog., 1988, Biomass Liquids to Produce Hydrocarbon Fuels’, in 8, (1), 59

41. Y. T. Cheng and G. W. Huber, Green Chem., 2012, Technol., 1995, 45, (3), 185 14, (11), 3114 50. H. P. A. Calis, J. P. Haan, H. Boerrigter, A. van 42. ‘What We Do’, Inaeris Technologies: http://www. der Drift, G. Peppink, R. van den Broek A. P. C. inaeristech.com/technology.html (Accessed 9th Faaij and R. H. Venderbosch, ‘Preliminary Techno- November 2017) Economic Analysis of Large-Scale Synthesis 43. “Future Roles of FCC and Hydroprocessing Units Gas Manufacturing from Imported Biomass’, in in Modern Refineries Fluid Catalytic Cracking”, “Pyrolysis and Gasification of Biomass and Waste”, Hydrocarbon Publishing Co, Paoli, USA, March, ed. A. V. Bridgwater, Proceedings from Strasbourg, 2009 30th September–1st October, 2002, CPL Press Newbury, UK, 2003, pp. 403–418 44. C. J. Freeman, S. B. Jones, A. B. Padmaperuma, M. Santosa, C. Valkenburg and J. Shinn, “Initial 51. H. Morehead, ‘Siemens Gasification and IGCC Assessment of U.S. Refineries for Purposes of Update’, Gasification Technologies 2008, Potential Bio-Based Oil Insertions”, PNNL-22432, Washington, DC, 7th October, 2008 Pacific Northwest National Laboratory, Richland, 52. D. Volkmann, ‘Update on Technology and Projects’, Washington, USA, April, 2013, 35 pp Gasification Technologies Conference, Washington 45. G. W. Huber and A. Corma, Angew. Chem. Int. DC, USA, October, 2004 Ed., 2007, 46, (38), 7184 53. N. Dahmen, J. Abeln, M. Eberhard, T. Kolb, H. 46. A. de Rezende Pinho, M. B. B. de Almeida, F. L. Leibold, J. Sauer, D, Stapf and B. Zimmerlin, Mendes, M. S. Talmadge, C. M. Kinchin and H. L. WIREs Energy Environ., 2017, 6, (3), e236 Chum, Fuel, 2017, 188, 462 54. H. J. Robota, L. A. Richard, S. Deshmukh, S. 47. N. Thegarid, G. Fogassy, Y. Schuurman, C. LeViness, D. Leonarduzzi and D. Roberts, Catal. Mirodatos, S. Stefanidis, E. F. Iliopoulou, K. Surv. Asia, 2014, 18, (4), 177 Kalogiannis and A. A. Lappas, Appl. Catal. B: 55. A. V. Bridgwater, ‘Pyrolysis of Biomass’, in Environ., 2014, 145, 161 “Transformations to Effective Use: Biomass Power 48. R. K. Sharma, and N. N. Bakhshi, Energy Fuels, for the World”, eds. W. van Swaaij, S. Kersten, 1993, 7, (2), 306 and W. Palz, Pan Stanford Publishing Pte Ltd, 49. J. D. Adjaye and N. N. Bakhshi, Fuel Process. Singapore, 2015, pp. 473–514

The Author

Tony Bridgwater is a professor of chemical engineering at Aston University, Birmingham, UK. He has worked at Aston University for most of his professional career and is currently director of the European Bioenergy Research Institute. He has a world-wide research portfolio focussing on fast pyrolysis as a key technology in thermal biomass conversion for power, heat, biofuels, and biorefineries. He is a fellow of the Institution of Chemical Engineers and a fellow of the Institute of Energy.

www.technology.matthey.com

Lithium Recovery from Aqueous Resources and Batteries: A Brief Review A review of the methods to produce lithium and approaches to recycling from end-of-life lithium-ion batteries

Ling Li portable electronics. To alleviate the potential Chemical Sciences Division, Oak Ridge risk of undersupply, lithium can be extracted from National Laboratory, Oak Ridge, Tennessee raw sources consisting of minerals and brines or 37831, USA from recycled batteries and glasses. Aqueous lithium mining from naturally occurring brines Vishwanath G. Deshmane and salt deposits is advantageous compared to Materials Science and Technology Division, extraction from minerals, since it may be more Oak Ridge National Laboratory, Oak Ridge, environmentally friendly and cost-effective. In Tennessee 37831, USA this article, we briefly discuss the adsorptive behaviour, synthetic methodology and prospects M. Parans Paranthaman* or challenges of major sorbents including spinel Chemical Sciences Division, Oak Ridge lithium manganese oxide (Li-Mn-O or LMO), National Laboratory, Oak Ridge, Tennessee spinel lithium titanium oxide (Li-Ti-O or LTO) 37831, USA and lithium aluminium layered double hydroxide

*Email: [email protected] chloride (LiCl·2Al(OH)3). Membrane approaches and lithium recovery from end-of-life LIB will also Ramesh Bhave be briefly discussed. Materials Science and Technology Division, Oak Ridge National Laboratory, Oak Ridge, 1. Introduction Tennessee 37831, USA Due to the accelerated expansion of the LIB Bruce A. Moyer industry, the global demand for lithium is Chemical Sciences Division, Oak Ridge expected to increase significantly with an National Laboratory, Oak Ridge, Tennessee annual growth rate of 8.9% through 2019 to 37831, USA 49,350 metric tonnes (1–3). The estimated global lithium end-use applications are summarised in Stephen Harrison Table I. Batteries used in portable electronics, Alger Alternative Energy LLC, Brawley, hybrid cars and electric vehicles consume ~35% California 92227, USA of the total lithium market share. The ceramics and glass sector with a ~32% market share is the second highest consumer of lithium. Primary lithium resources are from pegmatites, The demand for lithium is expected to increase continental brines and geothermal brines, and drastically in the near future due to the increased the secondary resources are from clays and usage of rechargeable lithium-ion batteries (LIB) seawater (4). Lithium carbonate (Li2CO3), which in electric vehicles, smartphones and other is the major commercial lithium product, is

Table I List of Estimated Global Lithium End-Use Applicationsa

Market Applications Products share, %

Li CO ; LiOH; Li metal; lithium hexafluorophosphate Batteries – portable electronics; 2 3 (LiPF ) electrolyte salts; lithium chloride (LiCl); Li alloys; hybrid cars; electric vehicles; 35 6 lithium cobalt oxide (LiCoO ); and other Li electrode grid storage applications 2 compositions

Ceramics and glass 32 Spodumene – LiAl(SiO3)2; Li2CO3 Lubricants and greases 9 LiOH

Air treatment; continuous casting mould flux powders; Li organometallics; Li metal; LiCl; lithium aluminium 5; 5; 4; 1 polymer production; primary Al hydride (LiAlH4); butyl lithium; lithium citrate production

Other uses such as in medicine as antidepressants, bipolar 9 Li compounds disorder a Summarised from (5, 9, 10) mostly prepared through mining, extracting and short review on the recovery of lithium and other treating spodumene ores and salt lake brines (4). valuable metals from end-of-life LIBs are discussed Lithium carbonate is losing market share to lithium in Section 5. hydroxide (LiOH), which is increasingly favoured for LIB cathode applications (2). Currently, two brine 2. Recovery of Lithium from Brines operations in Chile and a spodumene operation in by Adsorption and Ion Exchange Australia account for the majority of global lithium production (5). Spinel Li-Mn-O, spinel Li-Ti-O and LiCl·2Al(OH)3 Extraction from brines would be advantageous have been identified as potential sorbents for relative to extraction from ores, since it is more lithium extraction from aqueous resources. In the environmentally friendly and cost-effective (6). It is section below, we discuss these sorbents including estimated that the lithium production cost from salt their synthetic methods, structures, adsorption lake brines is US$2–3 kg–1, whereas that from the mechanisms, morphologies and adsorption or ores or spodumene is US$6–8 kg–1 (4). The major ion exchange capacities from different aqueous lithium-containing brine resources around the world resources. are listed in Table II. The lithium concentration –1 of brines ranges from 100–1000 mg l , whereas 2.1 Lithium Manganese Oxides average lithium concentration present in seawater (Li-Mn-O) is merely 0.17 mg l–1 (4). Therefore, salt lake and geothermal brines are the most promising aqueous Spinel-type Li-Mn-O are attractive candidates for resources for industrial scale lithium extraction (7). commercial lithium extraction owing to their high In fact, a majority of lithium is currently produced capacity and superior selectivity towards lithium. Li through solar evaporation, followed by the removal Mn-O are synthesised as precursor materials, from of impurities through precipitation. However, which the ion sieves are obtained by replacing the this method is time consuming (usually 18–24 Li+ with H+. Li-Mn-O can be synthesised via various months) and requires large land areas. Besides, methods including solid state reaction, sol-gel, the presence of excessive cations such as sodium, hydrothermal or reflux, yielding different particle potassium, calcium and magnesium plus chloride sizes and morphologies, which lead to different ion ions in the brines makes it challenging to obtain a exchange capacities of the lithium de-intercalated high purity product. sorbents. In general, the Li-Mn-O precursors can

Finally, the rechargeable LIB industry has be expressed by the formula (Li)[LixMn2-x]O4, expanded significantly with the maturation of where A-site (mostly Li) and B-site (Li and/or Mn) clean and sustainable energy technologies. A represent 8a tetrahedral and 16d octahedral sites

Table II World Brine Compositionsa,b

Source Li, wt% Na, wt% Mg, wt% K, wt% Ca, wt%

Clayton Valley, USA 0.0163 4.69 0.019 0.4 0.045

Salton Sea, USA 0.01–0.04 5.00–7.00 0.07–0.57 1.30–2.40 2.26–3.9

Salar de Atacama, Chile 0.157 9.1 0.965 2.36 0.045

HombreMuerto, Argentina 0.068–0.121 9.9–10.3 0.018–0.14 0.24–0.97 0.019–0.09

Salar de Uyuni, Bolivia 0.0321 7.06 0.65 1.17 0.0306

Searles Lake, USA 0.0054 11.8 – 2.53 0.0016

Great Salt Lake, USA 0.0018 3.70–8.70 0.5–0.97 0.26–0.72 0.026–0.036

Dead Sea, Israel 0.0012 3.01 3.09 0.56 1.29

Sua Pan, India 0.002 6 – 0.2 –

Bonneville, USA 0.0057 8.3 0.4 0.5 0.0057

Zabuye, China 0.0489 7.29 0.0026 1.66 0.0106

Taijinaier, China 0.031 5.63 2.02 0.44 0.02 a Adapted from (8, 9) b Please note only cations with high concentrations are provided in addition to Li

(10), and the acid treated ion sieves have a general room temperature. Nevertheless, the nano-sized formula of MnO2· xH2O. The primary Li uptake Li1.33Mn1.67O4 prepared by a gel process exhibited mechanism for the spinel-type sorbents is the a slightly lower lithium uptake of 28.2 mg g–1 + + + Li /H exchange, in which the Li can be intercalated/ from artificial seawater (31). In fact, Li1.33Mn1.67O4 de-intercalated into the octahedral interstices, with prepared from different precursors exhibited an intact spinel structure (11). Furthermore, the Li+ different lithium uptake even though the synthetic can be cycled in and out freely within a relatively method and temperature are exactly the same wide range of Li:Mn molar ratios (12, 13), resulting (27). A comparative study showed that ion sieves in several common manganese oxide precursors derived from Li4Mn5O12 (Li1.33Mn1.67O4) exhibited a including LiMn2O4 (10, 12, 14–18), Li1.6Mn1.6O4 (11, higher capacity compared to those derived from –1 –1 19–26) and Li1.33Mn1.67O4 (19, 27–32). Desorption/ LiMn2O4 (46.6 mg g vs. 23.9 mg g ) (10). regeneration of the spinel-type sorbents requires LiMn2O4 related ion sieve has a relatively lower contacting the sorbents with acid. ion exchange capacity and weak stability due to

Table III lists the ion exchange properties of the Jahn-Teller distortion with cycling. The MnO2 the lithium ion sieves derived from Li-Mn-O with preparation was first reported in 1981via treating various Li:Mn molar ratios. The lithium extraction LiMn2O4 with acid (34). It was further confirmed capacity depends on various parameters in 1984 that lithium can be cycled in and out of including the synthetic condition of the precursor the [Mn2]O4 framework over a wide range of x to materials (20, 33), actual Li:Mn molar ratio (33), form Li1–xMn2O4 (12). The acid treated ion sieve temperature and pH of the contact solution (22). MnO2 obtained from LiMn2O4 nanowire exhibited Therefore, the reported ion exchange behaviour an ion exchange capacity of ~16.8 mg g–1 from of a given sorbent can vary between different LiCl solutions (15). In later years, the same research groups. To date, the maximum ion research group synthesised LiMn2O4 nanorods exchange capacity of the manganese oxide is (15–20 nm in diameter and several micrometers 54.65 mg g–1 which was realised recently in in length) via a one-step soft chemistry method,

Li1.33Mn1.67O4 synthesised from Li2CO3 and MnCO3 and the related ion sieve showed a slightly higher –1 (30). The as-prepared Li1.33Mn1.67O4 powders extraction capacity of 20.5 mg g from LiCl were mixed with a chitosan binder and extruded solutions (14). into cylinder-shaped material (chitosan–LMO, Li1.6Mn1.6O4 related ion sieve MnO2· 0.5H2O diameter of 0.7 mm). The extraction was carried has an overall relatively high capacity, which is out in a column system with seawater flowing at attributed to the availability of strong acidic sites

Table III List of Some Common Li-Mn-O Precursors Synthesised Under Different Conditions

Capacity, Precursors Synthesis Morphology Solution Ref. mg g–1

H Mn O (19) 1.33 1.67 4 – – pH = 6.6 27–30 H1.6Mn1.6O4 400°C using lithium nitrate (LiNO ) as a 47.1 3 Various – (27) flux from different (optimum) precursors

Mixed solution Low-temperature solid- with Li+, Na+, K+ , Nanorod 46.6 (10) phase reaction (673 K) Mg2+ and Ca2+ of 10.0 mmol l–1

Spherical with 2–3.5 mm A combination of in diameter, hydrothermal reaction polyvinyl LiCl, pH = 10.1 23.5 (35) and solid-phase chloride (PVC) calcinations manganese(IV) oxide (MnO2) Li1.33Mn1.67O4 Tartaric acid gel process with lithium acetate Nano Artificial seawater 28.2 (31) (CH3COOLi) and Mn(CH3COO)2· 4H2O Seawater through a column Cylinder-shaped, setup packed chitosan-LMO Solid state with chitosan- 54.7 (30) granules diameter LMO, room of 0.7 mm temperature, pH = 6.6

Modelling a column – – – (28) system

Controlled low- Nanowire with 5 temperature nm diameter and LiCl, pH = 9.19 16.8 (15) hydrothermal synthesis 400 nm in length

Manganese(II) nitrate tetrahydrate MnO nanorods (Mn(NO ) ), LiOH and 2 LiCl (10.0 mmol 3 2 with 15–20 nm in 20.5 (14) LiMn O hydrogen peroxide l–1 Li+), pH = 10.1 2 4 diameter (H2O2) mixed solution at 383 K for 8 h

Mixed solution High-temperature with Li+, Na+, K+ , Nanorod 23.9 (10) calcinations (1003 K) Mg2+ and Ca2+ of 10.0 mmol l–1

Lithium-enriched salt lake brine (pH = 6; main Meso- or metallic ions: Li+ Citrate method macroporous 1.5 (36) Molar Mn/Li = 237 mg l–1, Na+ foam 1.125 3591 mg l–1, K+ –1 Li1.6Mn1.6O4 3118 mg l and Mg2+ 109 g l–1)

Sol-gel with Mn(NO ) One-dimensional Saltern bittern, (21) 3 2 10.5 and LiOH (1D) nanowire pH = 10

Continued

Capacity, Precursors Synthesis Morphology Solution Ref. mg g–1

Materials prepared by Calcination of lithium the reflux manganese dioxide Molar Mn/Li = method was less Seawater (0.17 (LiMnO ) which was 40 (20) 1.125 2 crystalline as mg l–1 Li+) made by hydrothermal compared to the and reflux methods hydrothermal method

Calcination of LiMnO2 which was made by a Simulated brine hydrothermal method Particle size (270 mg l–1 Li+), 27.2 (22) using manganese(III) 100–300 nm 50°C, pH = 5.35 oxide (Mn2O3) and LiOH

Calcination of LiMnO2 which was made by a LiCl (69.4 mg hydrothermal method l–1 Li+, with the using potassium Particle size presence of Na+ , 42.1 (24) permanganate ≤200 nm K+, Ca2+ and (KMnO ), 4 Mg2+), pH = 10.1 manganese(II) chloride Li1.6Mn1.6O4 (MnCl2) and LiOH Li1.16Sb0.29Mn1.54O4 Calcination of LiMnO 2 Qarhan salt lake which was made by brine (179 mmol a controlled redox l–1 Li+, 15,190 precipitation using Particle size mmol l–1 Na+ , manganese(II) 26.9 (25) ≤200 nm 13,729 mmol l–1 hydroxide (Mn(OH) ), 2 K+, 429 mmol LiOH and ammonium l–1 Ca2+, 80,125 persulfate mmol l–1 Mg2+) ((NH4)2S2O8) LiCl enriched Wet chemistry and seawater (5 mg 40 (37) hydrothermal at 120°C l–1 Li+)

+ inside the solid (20). Li1.6Mn1.6O4 is relatively trend was found in Li uptake with increasing Li:Mn difficult to synthesise, usually by calcination molar ratio (33). Furthermore, the extraction of LiMnO2 in O2 at an appropriate temperature capacity of Li1.6Mn1.6O4 in simulated brines –1 + (8LiMnO2 + 2O2 → 5Li1.6Mn1.6O4). To date, the (270 mg l Li ) increases with increasing highest reported ion exchange capacity is 42.1 temperature (30–50°C) and increasing pH values mg g–1 (6.06 mmol g–1) from LiCl solution at a (1–12) (22). The high selectivity for lithium ions pH of 10.1 (24). However, the lithium uptake of was confirmed, with high separation coefficients of the same sorbent from salt lake brine dropped αLi/Mg = 109.5, αLi/Na = 220.7, αLi/K = 125.5 (22). to 28.3 mg g–1 (4.08 mmol g–1) and was further In addition, there have been studies on ion sieves reduced to 25.1 mg g–1 after six cycles (24). In derived from antimony (37), Mg (39, 40) and Fe addition, the ion exchange capacity increases with (41) doped Li-Mn-O. The ion exchange capacity increasing stacking fault concentrations in the (from Li+ enriched seawater) of ion sieves derived –1 precursor LiMnO2 (24, 38). Li1.6Mn1.6O4 prepared from Li1.16Sb0.29Mn1.54O4 reached 40 mg g (37). by the hydrothermal method showed a slightly Mg-doped spinel Li-Mn-O ion sieve exhibited an higher lithium uptake and cycling stability than optimum ion exchange capacity of 37.4 mg g–1 that prepared by the reflux method (20). Lithium from LiCl solution (200 mg l–1 Li+, pH = 12) (39). extractive materials prepared with LiOH·H2O and Nevertheless, MgMn2O4 exhibited a small ion –1 manganese(II) carbonate (MnCO3) usually have exchange capacity (from seawater) of 8.5 mg g higher Li+ ion exchange capacity than materials and the equilibrium time is 96 hours, indicating a prepared with Li2CO3 and MnCO3, and an ascending slow ion exchange (42).

In summary, Li-Mn-O ion sieves exhibited a ions in the sodium bicarbonate (NaHCO3)-added high ion exchange capacity and high selectivity salt brine and the ion exchange capacity reached for lithium ions from various aqueous resources. 32.6 mg g–1 at a pH of 6.5 (43). However, the The acid generated during lithium uptake can be ion exchange rate is slow, taking 24 hours to get recycled for regenerating the sorbents. This could to equilibrium. This work has since stimulated potentially reduce the cost of the acid consumption great efforts investigating the ion exchange itself. However, the dissolution of Mn2+ during the behaviour of this emerging ion sieve (44–49, 52, regeneration process with acid degrades the ion 53). The isotherm of H2TiO3 exhibited a Langmuir exchange capacity and results in a poor cycling type behaviour, following the pseudo-second stability. This key issue seriously limits Li-Mn-O’s order rate model (45, 46). The ion exchange + potential for upscaling. Further studies are needed capacity of H2TiO3 increases with increasing Li to improve the stability during cycling to realise concentration and decreasing pH values of the a stable ion exchange capacity. Simplicity of the aqueous resources (46, 49). Specifically, the ion regeneration process is also desirable. exchange capacity of H2TiO3 increased from 11.26 to 31.27 mg g–1 when initial concentration of Li+ 2.2 Lithium Titanium Oxides (Li-Ti-O) was increased from 500 to 2500 ppm (pH = 13.46) (49). To further elucidate the effects of other

Titanium-based spinel oxides share most of factors on the ion exchange capacity of H2TiO3, a the advantages with the manganese-based comprehensive orthogonal test with five factors spinel oxides, with an addition of being more (pre-calcination temperature, Li:Ti molar ratio, environmentally friendly, as the titanium is an earth reaction temperature, ion exchange temperature, abundant element, is stable and does not dissolve Li+ concentration) was performed (52). The –1 in acid. In particular, metatitanic acid (H2TiO3) has highest ion exchange capacity of 57.8 mg g been considered as an emerging environmentally is achieved under the optimum conditions: Li+ friendly sorbent for lithium extraction from concentration = 4.0 g l–1 (highest among the tested), aqueous resources. The precursor lithium titanate ion exchange temperature = 60°C (highest among

(Li2TiO3) was first synthesised in 1988 and various the tested), molar ratio of Li:Ti = 2.2, reaction synthesis methods are now available in the temperature = 650°C, pre-calcination temperature literature, including solid-state reaction (43–47), = 25°C. To make H2TiO3 more economically hydrothermal (48) and sol-gel (49, 50). efficient, low‑grade titanium slag was used as the Debate persists about the crystal structures of starting material and the optimal capacity reached –1 Li2TiO3 and H2TiO3, in which Chitrakar et al. (43) 27.8 mg g (47). indexed both compounds as monoclinic with a Li4Ti5O12 is one of the common anode materials used space group C2/c, but later Yu et al. (51) pointed in LIB (54) and the related H4Ti5O12 is a common ion out that H2TiO3 should be more reasonably indexed sieve for lithium extraction from aqueous solutions. with the 3R1 space group with an LDH structure. H4Ti5O12 derived from Li4Ti5O12 nanotubes (~70 nm

Typically, layered H2TiO3, derived from a layered in diameter) exhibited an ion exchange capacity of –1 –1 + Li2TiO3 precursor upon treatment with HCl solution, 39.43 mg g from LiCl solution (120 mg l Li , will go through ion exchange with lithium ions from pH = 9.17). In summary, H2TiO3 is an attractive the geothermal brines at a pH >7 to form Li2TiO3 sorbent for selective lithium extraction with superior

(H2TiO3 + 2LiOH → Li2TiO3 + 2H2O). Lithium can advantages including high ion exchange capacity, be recovered from Li2TiO3 by treating with HCl high selectivity, high stability, environmental solution (Li2TiO3 + 2HCl → H2TiO3 + 2LiCl). The friendliness and economic efficiency. However, it theoretical ion exchange capacity of H2TiO3 is is still at the laboratory scale, partly due to the up to 142.9 mg g–1 (48), whereas the highest acid requirement during the regeneration process, experimental ion exchange capacity so far is which produces secondary wastes. 94.5 mg g–1 (46). This is actually the maximum + achievable capacity, as only 75% of the H occupied 2.3 Lithium Aluminium Layered ion exchange sites in H TiO are exchangeable 2 3 Double Hydroxide Chloride with Li+ (44). Table IV summarises the adsorptive behaviours of H2TiO3 synthesised under different While the Li-Mn-O and Li-Ti-O sorbents have conditions from various research groups. attracted significant attention from academia,

It was first demonstrated in 2014 that2 H TiO3 LiCl·2Al(OH)3.xH2O (referred to as Li/Al LDH) is an exhibits an extremely high selectivity toward lithium attractive candidate for application in large scale

Table IV List of the Adsorptive Properties of H2TiO3 Synthesised Under Different Conditions

Ion sieve Capacity, Synthesis Method Solution Ref. morphology mg g-1

Solid state Uniform particle, 1–2 µm LiOH (694.1 mg l–1 Li+) 39.8 (45)

Li enriched salt lake brine (1630 mg l–1 Li+ , Plate like particles with collected from Salar Solid state average diameter of de Uyuni, Bolivia) 32.6 (43) 100–200 nm added with sodium bicarbonate (NaHCO3), pH = 6.5

Solid state using titanium LiOH (2.0 g l–1 Li+), – 39.2 (44) dioxide (TiO2) and LiOH·H2O 25°C Particles size ranges LiOH (4.0 g l–1 Li+), Optimum Sol-gel (52) from 20–70 nm 60°C 57.8

Low grade Ti, solid state 100–300 nm LiOH (2.0 g l–1 Li+) 27.8 (47)

Plate-like particle, Solid state 700°C LiOH+LiCl 94.5 (46) 100–300 nm

H2TiO3 mixed with poly(vinyl alcohol) Solid state Seawater pH = 7.64 30.3 (53) (PVA) matrix, porous composite foam

Solid state from LiOH·H O 2 Plate-like particle LiOH (2.0 g l–1 Li+) 76.7 (48) and TiO2 Sol-gel using CH COOLi and Optimum, 3 60–80 nm LiOH (4.0 g l–1 Li+) (49) Ti(OC4H9)4 27.4

industrial plants due to its various advantages, (atomic ratio Li:Al ~0.38) was used in a large scale including low cost, environmental friendliness column system packed with 25 tonnes of sorbent and easy regeneration. Li/Al LDH materials have for selective lithium extraction from magnesium- + - a general formula [LiAl2(OH)6] B ·nH2O, where containing brines for more than 200 cycles, B = Cl, Br. They are crystallised in hexagonal demonstrating the good stability of this sorbent symmetry with the Li+ located in the vacant (64). Li/Al LDH has a good selectivity for LiCl (the octahedral sites within the aluminium hydroxide form of Li salts in brine and seawater) compared + (Al(OH)3) layer (55). The [LiAl2(OH)6] layers are to other cations, because the distance between separated by water molecules and hydroxide ions Al(OH)3 layers is at the nanoscale such that only (55). Li/Al LDHs can be synthesised by intercalating ions with small radii can be intercalated (64). Even the Li+ (in the form of LiCl, LiOH, lithium sulfate though the ionic radii of Mg2+ (0.074 nm) and Li+

(Li2SO4)) into aluminium hydroxides, which are (0.068 nm) are close, the large polarisability of in the form of naturally occurring minerals such the Li–Cl bond as compared to the Mg–Cl bond still + as gibbsite (α-Al(OH)3) or bayerite (β-Al(OH)3) makes this sorbent Li selective (64). (55–59). Recently, alternative synthetic routes Simbol Inc, USA (65) developed a column system such as a solvent-free mechanochemical method packed with Li/Al LDHs for the extraction of lithium have been demonstrated (60, 61). salts from geothermal brines sourced from the To the best of our knowledge, there exist limited Salton Sea, California, USA. The geothermal brine articles in the literature discussing the adsorptive has a bulk composition of about 260 ppm Li+ , properties of Li/Al LDH. It was first discovered to 63,000 ppm Na+, 20,100 ppm K+, 33,000 ppm be a selective sorbent for lithium extraction by Ca2+ and other ions (65). The Li/Al LDHs Simbol Dow Chemical Inc in 1980 (62). The synthesis Inc prepared have a high Li:Al atomic ratio of up to method was later modified, leading to an increase 0.5, which maximises the number of lithium sites in the molar fraction of LiX in LiX/Al(OH)3 from available in the layered structure for the intercalation 0.2 to 0.33 (63). Commercial granular Li/Al LDH and de-intercalation of lithium from a brine solution

(65). Note that this invention of extraction process Mg2+ rejection compared to only 15% for Li+, which is applicable to geothermal brine as well as other was attributed to its higher hydraulic permeability brine sources. Recently, Li et al. demonstrated to pure water and 0.1 M sodium chloride (NaCl) safe LIB using Li4Ti5O12 (LTO) electrode materials solution, and its lower critical pressure. Recently, prepared from Li2CO3 extracted from geothermal novel positively charged polyamide composite brine solutions using Li/Al LDH sorbents with good nanofiltration membranes were fabricated by cyclability (65). These demonstrations provide the interfacial polymerisation of DAPP and TMC a promising way for making low cost, large and supported on PAN ultrafiltration hollow fibre scale LTO electrode materials for energy storage membrane (21). The advantage of using hollow applications. In summary, LiCl·2Al(OH)3· xH2O is fibre compared to the mostly reported flat‑sheet an attractive candidate to be applied in large scale configuration is that the hollow fibres have high plant for extraction of lithium salts from various packing density, lower energy and maintenance cost brines. A detailed study on this sorbent regarding and easy fabrication of the modules. The rejection the isotherms is still needed. order of this composite hollow fibre membrane was

magnesium chloride (MgCl2) > magnesium sulfate 3. Recovery of Lithium from Brines by (MgSO4) > NaCl ≥ LiCl (21). Functionalisation of the positively charged Membranes membrane (fabricated by interfacial polymerisation Membrane processes offer several advantages of TMC and BPEI supported on polyetherimide compared to conventional processes, such as lower sheets) with EDTA showed good separation energy requirements and capital investments, performance with a Li+/Mg2+ separation factor simple and easy to operate systems, smaller of ~9.2. This was attributed to the tendency footprints, ease of scalability and many other of EDTA to form complexes with the divalent specific application related advantages. For cations. It was suggested that the combination of example, in sorbent based separations in packed Donnan exclusion, dielectric exclusion and steric and fluidised bed systems, there is a significant hindrance governed the mass transport inside the pressure drop and loss of sorbent particles. nanofiltration membranes. Furthermore, it was However, both these limitations can be eliminated also indicated that when membrane pore size is by the fabrication of mixed matrix membranes close to the ionic radius, steric hindrance plays a including Li+ selective sorbent. Although there is an significant role in the separation (21, 66, 67). increasing interest in membrane based Li+ recovery An electrolysis method employing the typical processes, there are only limited published reports anion exchange membranes (MA-7500, SYBRON discussing techniques such as nanofiltration (2, 5, and American IONAC®) and lithium iron phosphate

21, 66–69), electrolysis (70–72), electrodialysis (LiFePO4)/iron(III) phosphate (FePO4) electrodes (73–76), dialysis (74), membrane solvent extraction was investigated for the extraction of Li+ from (77–79) and membrane type adsorbents or mixed salt lake brines (70–72). The effect of different matrix membranes (80–84). The summary of these parameters on the Li+ extraction performance studies is provided in Table V. was studied. At optimised operating conditions, The first study on the application of nanofiltration electrodes exhibited a noteworthy Li+ exchange for the recovery of lithium from brines used a capacity of 38.9 mg g–1 (72). spiral-wound Desal-5 DL 2540C membrane (GE Recovery of lithium from seawater was also Osmonics), which showed a 61–67% retention of demonstrated by an electrodialysis based the Mg2+, while Li+ passed through the membrane, technique, which uses organic membranes giving a Li+/Mg2+ separation factor of 3.5 (66). A impregnated with an ionic liquid (73, 75). The Desal-DK membrane (GE Osmonics) showed a Li+/ separation of lithium was mainly achieved based Mg2+ separation factor ranging between 2 to 3.2 on its relatively lower or higher permeation rates depending upon the feed Li+ and Mg2+ concentration compared to other cations. However, it was and their ratio (5, 68). The higher operating pressure, suggested that the poor durability of the ionic lower pH and higher feed Li+:Mg2+ ratio improved membrane is a major issue preventing long-term the separation (68). The relative Li+ separation lithium recovery (74). The applied voltage, feed performance of nanofiltration‑NF90 (Dow) and low velocity, feed Li+:Mg2+ ratio and pH significantly pressure reverse osmosis-XLE (Dow) membranes influenced the Li+/Mg2+ separation factor (76). was evaluated with salt lake brine (2). NF90 Supported liquid membranes (SLMs) have also membrane appeared more efficient, showing 100% attracted interest, borrowing selectivity from

Table V Summary of Reported Studies of Lithium Extraction Using Membrane Processes

Mechanism Lithium Separation or separation Membrane system Ref. source factor process

Spiral-wound Desal-5 DL 2540C, spiral- wound Desal DK (GE Osmonics, USA); NF90 and XLE (Dow, USA); Spiral-wound DK-1812 (Suntar Membrane Tech, China), (2, 5, DL-2540; 1,4-bis(3-aminopropyl)piperazine Li+/Mg2+: Nanofiltration Salt lake brine 21, (DAPP) and trimesoyl chloride (TMC) 2–42 66–69) polymerised on the polyacrylonitrile (PAN) hollow fibre; ethylenediaminetetraacetic acid (EDTA) functionalised TMC and branched polyethyleneimine (BPEI)

Anion exchange MA-7500 (SYBRON); Highly SelemionTM CMV with ionic liquid TMPA-TFSI; Electrolysis, selective Salt lake brine, Gore-Tex® impregnated with ionic liquid (72– electrodialysis, recovery seawater (PP13-TFSI); Li ion conductive glass-ceramics 76) dialysis of Li+, Li+/ (Ohara Inc, Japan), ACS (Anion exchange) Mg2+: 12–77 and CIMS (Cation exchange) (ASTOM, Japan)

α-acetyl-m-dodecylacetophenone (LIX54) and tri-n-octylphosphine oxide (TOPO) in kerosene embedded in Celgard® 2500 >90% Membrane Geothermal membrane; tributylphosphate (TBP) + FeCl extraction 3 (77– solvent water, salt lake in kerosene with polyethersulfone (PES) + of Li in 79) extraction brine and sulfonated poly(phthalazinone ether 2 h, high Li+ sulfone ketone) (SPPESK) blend; TBP + selectivity iron(III) chloride (FeCl3) in kerosene with poly(ethylene-co-vinyl alcohol) (EVAL)

+ + Binary mixtures Li /Na : 35, PSS threaded HKUST-1 metal-organic + 2+ Grotthuss of Na+, K+ and Li /Mg : (80) framework (MOF) 1815 Mg2+ with Li+*

Li Mn O /PVC, Li Mn O 1.33 1.67 4 1.33 1.67 4 >90% Li+ Seawater, encapsulated in polysulfone (PSf)/Kimtex, recovery, (81– Sorption geothermal Li Mn O /PSf/PAN mixed matrix 1.33 1.67 4 complete Li+ 85) brine (nanofibre), LDH‑polyvinylidene fluoride selectivity (PVDF)/PVDF hollow fibres

the incorporated solvent extraction reagents. To improve the stability of the SLM for Li+ Ma et al. (77) reported the first study on the extraction, a nanoporous ion exchange membrane extraction of lithium from geothermal water with was fabricated by blending PES with sulfonated the SLM technique. A mixture of extractants poly(phthalazinone ether sulfone ketone) consisting of LIX54 (the main component is (SPPESK) as a extractant stabiliser (79). With

α-acetyl-m-dodecylacetophenone) and TOPO were PES/SPPESK blend membrane and TBP and FeCl3 immobilised in the Celgard® 2500 membrane mixed in kerosene as an extractant, Li+ extraction having 37–48% porosity. The SLM showed 95% was performed both in a single-stage extraction extraction of Li+ in just 2 hours; however, it and a sandwiched membrane extraction contactor exhibited stable performance for only up to 72 system. The best Li+ extraction performance was hours before the flux dropped drastically. The obtained at a PES:SPPESK ratio of 6:4 and a decreased stability was attributed to the pressure polymer concentration of 30 wt%. However, these difference over the membrane sheet, the solubility membranes had limited stability in benzene and of the liquid membrane in the adjacent solutions toluene despite being stable in kerosene (78, 79). and emulsion formation of the liquid membrane in To further improve the stability of the membrane aqueous solutions (77). with different solvents, EVAL membranes were

169 © 2018 United States Government https://doi.org/10.1595/205651317X696676 Johnson Matthey Technol. Rev., 2018, 62, (2) fabricated. These membranes showed exceptional the strip is carried out simultaneously, eliminating + + stability in Li extraction with TBP/FeCl3/kerosene the need to employ a separate step for Li recovery. for about 1037 hours. This higher stability was In summary, although there are many published attributed to the unique structure of EVAL, reports on membrane-based separation processes consisting of both hydrophobic ethylene and for lithium extraction, the technology is currently hydrophilic vinyl alcohol units (78). at the laboratory scale with significant potential for In a recent study, novel polystyrene sulfonate further development and process scale-up in the (PSS) incorporated HKUST-1 MOF membranes future. were fabricated for Li+ recovery from brines through an in situ lithium confinement process 4. Recovery of Lithium from Brines by (80). These MOF membranes showed exceptionally Other Methods good performance in Li+ recovery with separation selectivities (molar) of 35, 67 and 1815 over Na+ , There have been reports of lithium extraction using K+ and Mg2+, respectively. It was established other methods such as precipitation and solvent that the perm-selectivity followed a trend of extraction. The precipitation method was used to Li+ > Na+ > K+, which was determined by the extract lithium from the Dead Sea in 1981 (86). Later corresponding binding affinities of these cations a two-stage precipitation process was developed to + to the sulfonate groups. The transportation of Li extract Li2CO3 from brines collected from Salar de through the membrane is proposed to be governed Uyuni, Bolivia (700–900 mg l–1 Li+) (8). Solvent by the Grotthuss mechanism, wherein the charge extraction has been widely used to extract metals is transported by the coordinated hopping of Li+ from the aqueous phase due to the simplicity of the between sulfonate groups of PSS threaded through equipment and operation. In fact, it was applied the cavities of HKUST-1 (80). Another successful to extract lithium from aqueous solutions of alkali membrane-type adsorbent of spinel manganese metal salts as early as 1968 (87). The extraction + oxide (H1.33Mn1.67O4) was prepared by a solvent of Li ions into the organic phase is associated exchange method using PVC as a binder (81). This with the cation exchange mechanism. Various membrane-type adsorbent has an uptake capacity solvents including tri-n-butyl phosphate (88), of 10.6 mg g–1 Li+ from seawater (0.17 mg l–1 Li+). ionic liquid added 1-ethyl-3-methylimidazolium

A membrane reservoir system with encapsulated bis(trifluoromethylsulfonyl)imide ([C2mim][NTf2])

Li1.33Mn1.67O4 in PSf and Kimtex (Korea Non-woven mixed with tri-n-butyl (89) and so on, have been Tech Ltd, South Korea) was tested for Li+ recovery reported for lithium extraction. Organophosphorus from seawater. The Kimtex based systems showed ligands in the presence of ammonia were tested for best results with ~84% Li+ recovery in one day lithium extraction, in which the highest extraction due to the easy wetting and water penetration percentages in the presence of H-PHO, H-PHI and in the reservoir (82). The Li1.6Mn1.6O4-PSf/ H-BIS ligands were 43.2%, 45.7% and 90.0%, PAN‑based composite mixed matrix nanofibres respectively (90). as a flow through membrane Li+ absorber was highly permeable to water under minimal 5. Recovery of Lithium from Recycled trans-membrane pressure (83, 84). The balance Lithium-Ion Batteries between kinetic and dynamic Li+ adsorption capacity could be obtained at optimal seawater A rechargeable LIB mainly comprises a and membrane contact time (84). lithium-containing oxide cathode, an anode, an Bhave et al. (85) have fabricated novel LDH organic electrolyte and a separator. Table VI lists (LIS)/Kynar®-PVDF mixed matrix membranes the chemical composition of a typical LIB. The supported on PVDF hollow fibres (Arkema Inc, cathode is usually made of LiCoO2, lithium nickel + France) for Li recovery from geothermal brines. dioxide (LiNiO2) and lithium manganese(III,IV)

Due to the high temperature of the geothermal oxide (LiMn2O4) and the anode is typically graphite. brines, robust membranes are required to operate Aluminium and copper are used as current at temperatures up to 95°C. Preliminary results collectors. The recoverable materials from an showed the potential of these membranes to end-of-life battery include aluminium, copper, LiOH obtain a high lithium separation factor with nearly or Li2CO3, cobalt oxide, nickel oxide and manganese complete rejection of other monovalent and oxide. There have been a number of articles in the divalent cations in the brine solution. The selective literature focused on recovery of metals such as sorption/diffusion of Li+ and back-extraction into cobalt, lithium and nickel from spent LIBs (91–95).

For example, a three-step process (100) was Table VI Chemical Composition of a Typical LIBa developed to recover cobalt and lithium from the cathode materials: Component wt% (a) leaching of the cathode materials with HCl

LiCoO2 27.5 (b) separation of cobalt from lithium with solvent extraction Steel/Ni 24.5 (c) precipitation of lithium as carbonate. Cu/Al 14.5 Employing the same technique, with organic Carbon 16 citric acid as the leachant, 90% cobalt and 100% lithium were recovered from end-of-life LIBs (101). Electrolyte 3.5 Alkaline solution was used to leach the battery’s Polymer 14 internal substances followed by dissolving the a Adapted from (95) residue in sulfuric acid (H2SO4) solution, yielding Li2CO3 (102). The effect of different parameters

such as the concentration of the leachant H2SO4,

Processes to recycle LIBs were first developed for temperature, pulp density and reductant H2O2 the sake of environmental considerations, since the concentration on the leaching of the waste was waste is usually flammable and toxic. It can also investigated (103). An optimum condition of pulp −1 achieve some economic benefits as driven by the density, 100 g l , 2M H2SO4, 5 vol% of H2O2, with prices of cobalt and possibly lithium, though they a leaching time of 30 min and a temperature of fluctuate drastically depending on their availability. 75°C, was identified (103). Very recently, several Figure 1 presents a flow sheet of a typical methods to recover lithium and other high value hydrometallurgical process, which is the most metals such as cobalt from spent LIB have been common process to recover lithium from spent reported (104–110). From both the viewpoints of LIBs. The whole procedure involves physical and environmental friendliness and economic benefits, chemical processes to complete the following recovery of lithium from spent LIB is desirable. steps: Nevertheless, most of the recycling processes are (a) pretreatment of the spent LIBs – dismantling the still at laboratory scale and much effort needs cells, thermal treatment and mechanochemical to be directed into this area. In addition, safety process precautions should be emphasised when LIB are (b) dissolution and leaching of metals from the dismantled. cathode material with hydrochloric acid (HCl), bioleaching Summary and Outlook (c) separation of lithium and other metals via solvent extraction, chemical precipitation and Aqueous lithium mining of continental brines appears electrochemical process (96–99). to be a promising approach to realise economically

Anode Electrolyte Dismantling and separation Current collectors (Cu/Al) End-of-life LIBs Stainless steel case Residue Other materials

Cathode Leaching, Extraction and materials precipitation Mixture of separation of metals lithium and other metals

Fig. 1. Flow sheet of a typical recycling process for spent LIBs

171 © 2018 United States Government https://doi.org/10.1595/205651317X696676 Johnson Matthey Technol. Rev., 2018, 62, (2) and environmentally attractive lithium production. DOE Public Access Plan. All the authors have no Extraction from seawater would be relatively costly competing financial interests. due to the extremely low lithium concentration of 0.17 ppm, though it would be of interest in coastal countries that have neither mineral nor continental References brine resources. Alternatively, brines such as salt 1. U. Wietelmann and R. J. Bauer, ‘Lithium and lake brines or geothermal brines serve as a rich Lithium Compounds’, in “Ullmann’s Encyclopedia resource. However, evaporation is a slow process of Industrial Chemistry”, Wiley-VCH Verlag GmbH that takes up to 24 months and the final products & Co KGaA, Weinheim, Germany, 2000 usually have low purity, whereby sorbents and 2. A. Somrani, A. H. Hamzaoui and M. Pontie, membranes are effective alternatives. The spinel- Desalination, 2013, 317, 184 type sorbents exhibit excellent ion exchange 3. S. Ziemann, M. Weil and L. Schebek, Resour. capacity and high selectivity, although the Conserv. Recycl., 2012, 63, 26 regeneration process could be expensive. On the 4. C. Grosjean, P. H. Miranda, M. Perrin and P. other hand, LiCl·2Al(OH)3 offers moderate capacity, Poggi, Renew. Sustain. Energy Rev., 2012, 16, but this material has other advantages such as low (3), 1735 cost and easy regeneration, which are essential for 5. G. Yang, H. Shi, W. Liu, W. Xing and N. Xu, Chin. industrial applications. Further research needs to be J. Chem. Eng., 2011, 19, (4), 586 carried out to better control the defects of the spinel precursor materials. Alternative methods such as 6. P. K. Choubey, M. Kim, R. R. Srivastava, J. Lee and J.-Y. Lee, Min. Eng., 2016, 89, 119 solvent extraction could be used to extract lithium from salt lake brines or geothermal brines. The need 7. P. Meshram, B. D. Pandey and T. R. Mankhand, for large quantities of lithium domestic supply in the Hydrometallurgy, 2014, 150, 192 USA remains a key priority, for example. Scale-up 8. J. W. An, D. J. Kang, K. T. Tran, M. J. Kim, T. Lim trials are essential to realise industrial operations to and T. Tran, Hydrometallurgy, 2012, 117–118, 64 meet the US domestic demand. This requirement justifies continued investment in the extraction of 9. D. A. Boryta, T. F. Kullberg and A. M. Thurston, critical lithium from salt lake and geothermal brines. Cemetall Foote Corp, ‘Production of Lithium Compounds Directly from Lithium Containing In addition, recovery of lithium from recycled LIB Brines’, US Patent Appl., 2011/0,123,427 needs a major investment in the near future. 10. S.-Y. Sun, X. Song, Q.-H. Zhang, J. Wang and J. G. Yu, Adsorption, 2011, 17, (5), 881 Acknowledgement 11. M. J. Ariza, D. J. Jones, J. Rozière, R. Chitrakar This work was supported by the Critical Materials and K. Ooi, Chem. Mater., 2006, 18, (7), 1885 Institute, an Energy Innovation Hub funded by 12. M. M. Thackeray, P. J. Johnson, L. A. de Picciotto, the US Department of Energy, Office of Energy P. G. Bruce and J. B. Goodenough, Mater. Res. Efficiency and Renewable Energy and Advanced Bull., 1984, 19, (2), 179 Manufacturing Office. 13. Q. Feng, Y. Miyai, H. Kanoh and K. Ooi, Langmuir, 1992, 8, (7), 1861 14. Q.-H. Zhang, S.-P. Li, S.-Y. Sun, X.-S. Yin and J. Additional Information G. Yu, Chem. Eng. Sci., 2010, 65, (1), 169 This manuscript has been authored by UT-Battelle, 15. Q.-H. Zhang, S. Sun, S. Li, H. Jiang and J.-G. Yu, LLC under Contract No. DE-AC05-00OR22725 Chem. Eng. Sci., 2007, 62, (18–20), 4869 with the US Department of Energy (DOE). The 16. Q. Feng, Y. Higashimoto, K. Kajiyoshi and K. United States Government retains and the Yanagisawa, J. Mater. Sci. Lett., 2001, 20, (3), publisher, by accepting the article for publication, 269 acknowledges that the US Government retains a 17. C. Özgür, Solid State Ionics, 2010, 181, (31–32), non-exclusive, paid-up, irrevocable, world-wide 1425 license to publish or reproduce the published 18. L. Li, W. Qu, F. Liu, T. Zhao, X. Zhang, R. Chen form of this manuscript or allow others to do and F. Wu, Appl. Surf. Sci., 2014, 315, 59 so, for US Government purposes. The DOE will 19. R. Chitrakar, Y. Makita, K. Ooi and A. Sonoda, provide public access to these results of federally Chem. Lett., 2012, 41, (12), 1647 sponsored research in accordance with the 20. R. Chitrakar, H. Kanoh, Y. Miyai and K. Ooi, Ind.

Eng. Chem. Res., 2001, 40, (9), 2054 Ind. Eng. Chem. Res., 2014, 53, (9), 3682 21. L. Liu, H. Zhang, Y. Zhang, D. Cao and X. Zhao, 42. Y. Miyai, K. Ooi and S. Katoh, Sep. Sci. Technol., Colloids Surf. A: Physiochem. Eng. Aspects, 1988, 23, (1–3), 179 2015, 468, 280 43. R. Chitrakar, Y. Makita, K. Ooi and A. Sonoda, 22. X. Shi, D. Zhou, Z. Zhang, L. Yu, H. Xu, B. Chen Dalton Trans., 2014, 43, (23), 8933 and X. Yang, Hydrometallurgy, 2011, 110, (1– 44. L. Zhang, D Zhou, G. He, F. Wang and J. Zhou, 4), 99 Mater. Lett., 2014, 135, 206 23. R. Chitrakar, H. Kanoh, Y. Miyai and K. Ooi, Chem.

Mater., 2000, 12, (10), 3151 45. X. Shi, Z. Zhang, D. Zhou, L. Zhang, B. Chen and L. Yu, Trans. Nonferrous Met. Soc. China, 2013, 24. J.-L. Xiao, S.-Y. Sun, J. Wang, P. Li and J.-G. Yu, 23, (1), 253 Ind. Eng. Chem. Res., 2013, 52, (34), 11967 46. C. P. Lawagon, G. M. Nisola, J. Mun, A. Tron, R. E. 25. S.-Y. Sun, J.-L. Xiao, J. Wang, X. Song and J.-G. C. Torrejos, J. G. Seo, H. Kim and W.-J. Chung, J. Yu, Ind. Eng. Chem. Res., 2014, 53, (40), 15517 Ind. Eng. Chem., 2016, 35, 347 26. R. Chitrakar, K. Sakane, A. Umeno, S. Kasaishi, 47. D. Tang, D. Zhou, J. Zhou, P. Zhang, L. Zhang and N. Takagi and K. Ooi, J. Solid State Chem., 2002,

169, (1), 66 Y. Xia, Hydrometallurgy, 2015, 157, 90

27. X. Yang, H. Kanoh, W. Tang and K. Ooi, J. Mater. 48. S. Wang, P. Li, W. Cui, H. Zhang, H. Wang, S. Zheng

Chem., 2000, 10, (8), 1903 and Y. Zhang, RSC Adv., 2016, 6, (104), 102608

28. K. Ooi, Y. Makita, A. Sonoda, R. Chitrakar, Y. 49. L. Zhang, D. Zhou, Q. Yao and J. Zhou, Appl.

Tasaki-Handa and T. Nakazato, Chem. Eng. J., Surf. Sci., 2016, 368, 82 2016, 288, 137 50. A. Deptuła, M. Brykała, W. Łada, T. Olczak, B.

29. H.-J. Hong, I.-S. Park, T. Ryu, J. Ryu, B.-G. Kim Sartowska, A. G. Chmielewski, D. Wawszczak

and K.-S. Chung, Chem. Eng. J., 2013, 234, 16 and C. Alvani, Fusion Eng. Des., 2009, 84, (2–6), 681 30. T. Ryu, Y. Haldorai, A. Rengaraj, J. Shin, H.-J.

Hong, G.-W. Lee, Y.-K. Han, Y. S. Huh and K.-S. 51. C.-L. Yu, F. Wang, S.-Y. Cao, D.-P. Gao, H.-B. Hui, Chung, Ind. Eng. Chem. Res., 2016, 55, (26), Y.-Y. Guo and D.-Y. Wang, Dalton Trans., 2015, 7218 44, (35), 15721

31. K. S. Chung, J. C. Lee, E. J. Kim, K. C. Lee, Y. S. 52. G. He, L. Zhang, D. Zhou, Y. Zou and F. Wang, Kim and K. Ooi, Mater. Sci. Forum, 2004, 449 Ionics, 2015, 21, (8), 2219 452, 277 53. L. A. Limjuco, G. M. Nisola, C. P. Lawagon, S.-P. Lee, J. G. Seo, H. Kim and W.-J. Chung, Colloids 32. Y. Miyai, K. Ooi, T. Nishimura and J. Kumamoto, Surf. A: Physiochem. Eng. Aspects, 2016, 504, Bull. Soc. Sea Water Sci., Jpn., 1994, 48, (6), 267 411 54. G.-N. Zhu, Y.-G. Wang and Y.-Y. Xia, Energy 33. L. Wang, W. Ma, R. Liu, H. Y. Li and C. G. Environ. Sci., 2012, 5, (5), 6652 Meng, Solid State Ionics, 2006, 177, (17–18), 1421 55. J. P. Thiel, C. K. Chiang and K. R. Poeppelmeier, Chem. Mater., 1993, 5, (3), 297 34. J. C. Hunter, J. Solid State Chem., 1981, 39, (2), 142 56. A. M. Fogg, A. J. Freij and G. M. Parkinson, Chem. Mater., 2002, 14, (1), 232 35. G. Xiao, K. Tong, L. Zhou, J. Xiao, S. Sun,

P. Li and J. Yu, Ind. Eng. Chem. Res., 2012, 51, 57. A. V. Besserguenev, A. M. Fogg, R. J. Francis, S. J.

(33), 10921 Price, D. O’Hare, V. P. Isupov and B. P. Tolochko, Chem. Mater., 1997, 9, (1), 241 36. L.-W. Ma, B.-Z. Chen, Y. Chen and X.-C. Shi, 58. A. M. Fogg and D. O’Hare, Chem. Mater., 1999, Micropor. Mesopor. Mater., 2011, 142, (1), 147 11, (7), 1771 37. R. Chitrakar, H. Kanoh, Y. Makita, Y. Miyai and K. 59. G. R. Williams and D. O’Hare, J. Phys. Chem. B, Ooi, J. Mater. Chem., 2000, 10, (10), 2325 2006, 110, (22), 10619 38. L. Croguennec, P. Deniard, R. Brec and A. Lecerf, 60. S.-L. Wang, C.-H. Lin, Y.-Y. Yan and M. K. Wang, J. Mater. Chem., 1997, 7, (3), 511 Appl. Clay Sci., 2013, 72, 191

39. L. Tian, W. Ma and M. Han, Chem. Eng. J., 2010, 61. J. Qu, X. He, B. Wang, L. Zhong, L. Wan, X. Li, S.

156, (1), 134 Song and Q. Zhang, Appl. Clay Sci., 2016, 120, 40. Q. Feng, Y. Miyai, H. Kanoh and K. Ooi, Chem. 24

Mater., 1993, 5, (3), 311 62. J. M. Lee and W. C. Bauman, The Dow Chemical 41. R. Chitrakar, Y. Makita, K. Ooi and A. Sonoda, Company, ‘Recovery of Lithium from Brines’, US

Patent Appl. 1979/4,159,311 84. M. J. Park, G. M. Nisola, E. L. Vivas, L. A. Limjuco,

63. W. C. Bauman and J. L. Burba III, FMC Corp, C. P. Lawagon, J. G. Seo, H. Kim, H. K. Shon and W.-J. Chung, J. Membrane Sci., 2016, 510, 141 ‘Composition for the Recovery of Lithium Values from Brine and Process of Making/Using Said 85. R. Bhave, V. Deshmane and D. Kim, ‘Selective Composition’, US Patent, 6,280,693; 2001 Recovery of Lithium from Geothermal Brine Using Novel Mixed Matrix Membranes Supported 64. N. P. Kotsupalo, A. D. Ryabtsev, I. A. Poroshina, on Hollow Fiber and Inorganic Supports’, 2018, in A. A. Kurakov, E. V. Mamylova, L. T. Menzheres preparation and M. A. Korchagin, Russ. J. Appl. Chem., 2013, 86. J. A. Epstein, E. M. Feist, J. Zmora and Y. Marcus, 86, (4), 482 Hydrometallurgy, 1981, 6, (3–4), 269 65. J. L. Burba III, R. F. Stewart, B. E. Viani, S. 87. D. A. Lee, W. L. Taylor, W. J. McDowell and J. Harrison, C. E. Vogdes and J. G. S. Lahlouh, S. Drury, J. Inorg. Nucl. Chem., 1968, 30, (10), Simbol Inc, ‘Improved Sorbent for Lithium 2807 Extraction’, World Patent Appl., 2015/171,109 88. C. Shi, Y. Jing and Y. Jia, J. Mol. Liq., 2016, 215, 66. X. Wen, P. Ma, C. Zhu, Q. He and X. Deng, Sep. 640 Purif. Technol., 2006, 49, (3), 230 89. C. Shi, Y. Jing and Y. Jia, Russ. J. Phys. Chem. A, 67. Q. Bi, Z. Zhang, C. Zhao and Z. Tao, Water Sci. 2017, 91, (4), 692

Technol., 2014, 70, (10), 1690 90. B. El-Eswed, M. Sunjuk, Y. S. Al-Degs and A. 68. S.-Y. Sun, L.-J. Cai, X.-Y. Nie, X. Song and J.-G. Shtaiwi, Separ. Sci. Technol., 2014, 49, (9), Yu, J. Water Process Eng., 2015, 7, 210 1342 69. W. Li, C. Shi, A. Zhou, X. He, Y. Sun and J. Zhang, 91. E. G. Pinna, M. C. Ruiz, M. W. Ojeda and M. H. Sep. Purif. Technol., 2017, 186, 233 Rodriguez, Hydrometallurgy, 2017, 167, 66

70. Z. Zhao, X. Si, X. Liu, L. He and X. Liang, 92. G. P. Nayaka, J. Manjanna, K. V. Pai, R. Vadavi,

Hydrometallurgy, 2013, 133, 75 S. J. Keny and V. S. Tripathi, Hydrometallurgy, 2015, 151, 73 71. X. Liu, X. Chen, Z. Zhao and X. Liang, 93. M. Joulié, E. Billy, R. Laucournet and D. Meyer, Hydrometallurgy, 2014, 146, 24 Hydrometallurgy, 2017, 169, 426 72. X. Liu, X. Chen, L. He and Z. Zhao, Desalination, 94. S.-H. Joo, D. ju Shin, C. Oh, J.-P. Wang, G. 2015, 376, 35 Senanayake and S. M. Shin, Hydrometallurgy, 73. T. Hoshino, Fusion Eng. Des., 2013, 88, (11), 2016, 159, 65

2956 95. C. K. Lee and K.-I. Rhee, J. Power Sources, 2002, 74. T. Hoshino, Desalination, 2015, 359, 59 109, (1), 17 75. T. Hoshino, Desalination, 2013, 317, 11 96. J. Xu, H. R. Thomas, R. W. Francis, K. R. Lum, J. Wang and B. Liang, J. Power Sources, 2008, 76. Z. Ji, Q. Chen, J. Yuan, J. Liu, Y. Zhao and 177, (2), 512 W. Feng, Sep. Purif. Technol., 2017, 172, 168 97. M. Contestabile, S. Panero and B. Scrosati, J. 77. P. Ma, X. D. Chen and M. M. Hossain, Sep. Sci. Power Sources, 2001, 92, (1–2), 65 Technol., 2000, 35, (15), 2513 98. S. Castillo, F. Ansart, C. Laberty-Robert and J.

78. L. Xing, J. Song, Z. Li, J. Liu, T. Huang, P. Dou, Y. Portal, J. Power Sources, 2002, 112, (1), 247 Chen, X.-M. Li and T. He, J. Membrane Sci., 2016, 99. D. C. R. Espinosa, A. M. Bernardes and J. A. S. 520, 596 Tenório, J. Power Sources, 2004, 135, (1–2), 79. J. Song, X.-M. Li, Y. Zhang, Y. Yin, B. Zhao, C. Li, 311 D. Kong and T. He, J. Membrane Sci., 2014, 471, 100. P. Zhang, T. Yokoyama, O. Itabashi, T. M. Suzuki 372 and K. Inoue, Hydrometallurgy, 1998, 47, (2–3), 80. Y. Guo, Y. Ying, Y. Mao, X. Peng and B. Chen, 259 Angew. Chem., 2016, 128, (48), 15344 101. L. Li, J. Ge, F. Wu, R. Chen, S. Chen and B. Wu,

81. A. Umeno, Y. Miyai, N. Takagi, R. Chitrakar, K. J. Hazard. Mater., 2010, 176, (1–3), 288 Sakane and K. Ooi, Ind. Eng. Chem. Res., 2002, 102. J. Nan, D. Han and X. Zuo, J. Power Sources, 41, (17), 4281 2005, 152, 278 82. K.-S. Chung, J.-C. Lee, W.-K. Kim, S. B. Kim and

K. Y. Cho, J. Membrane Sci., 2008, 325, (2), 503 103. B. Swain, J. Jeong, J. Lee, G.-H. Lee and J.-S. Sohn, J. Power Sources, 2007, 167, (2), 536 83. W.-J. Chung, R. E. C. Torrejos, M. J. Park, E. L. Vivas, L. A. Limjuco, C. P. Lawagon, K. J. 104. J. Ordoñez, E. J. Gago and A. Girard, Renew. Parohinog, S.-P. Lee, H. K. Shon, H. Kim and Sustain. Energy Rev., 2016, 60, 195 G. M. Nisola, Chem. Eng. J., 2017, 309, 49 105. Y. Guo, F. Li, H. Zhu, G. Li, J. Huang and W. He,

Waste Manage., 2016, 51, 227 Manage., 2017, 64, 171 106. S. P. Barik, G. Prabaharan and B. Kumar, Waste 109. L. Li, E. Fan, Y. Guan, X. Zhang, Q. Xue, L. Wei, Manage., 2016, 51, 222 F. Wu and R. Chen, ACS Sustainable Chem. Eng., 107. X. Chen, H. Ma, C. Luo and T. Zhou, J. Hazard. 2017, 5, (6), 5224 Mater., 2017, 326, 77 110. J. Xiao, J. Li and Z. Xu, Environ. Sci. Technol., 108. L.-P. He, S.-Y. Sun, X.-F. Song and J.-G. Yu, Waste 2017, 51, (20), 11960

The Authors

Dr Ling Li obtained her PhD in Materials Science and Engineering from the University of Tennessee, Knoxville, USA, in 2015. Then she worked as a Postdoctoral Fellow at Oak Ridge National Laboratory (ORNL), USA, from 2015 to 2017. She is currently a Materials Scientist at Magic Leap Inc. in Austin, Texas, USA.

Dr Vishwanath G. Deshmane is a Postdoctoral Research Associate in the Materials Science & Technology Division of ORNL. He earned his PhD degree with Chemical Engineering major from North Carolina A&T State University, Greensboro, USA, in 2012 and Master’s degree in Chemical Engineering from the Institute of Chemical Technology, Mumbai, India, in 2007. He joined ORNL in 2016 after completing his postdoctoral research at North Carolina A&T State University. He has more than seven years of research and development experience in catalysis and membrane separations and one year of industrial process engineering experience. He has more than 16 peer-reviewed publications with 450+ citations.

Dr Mariappan Parans Paranthaman is a Corporate Fellow and a Group Leader at ORNL. He is a fellow of the National Academy of Inventors, the American Association for the Advancement of Science (AAAS), American Ceramic Society, ASM International and the Institute of Physics, London, UK. He earned his PhD degree in Chemistry from the Indian Institute of Technology, Madras, India, in 1988. After completing his postdoctoral research at the University of Texas at Austin and the University of Colorado, Boulder, USA, he joined ORNL in May 1993. He has authored or co-authored more than 400 journal publications with an “h-index” of 59 (Google scholar citation) and issued 37 US Patents related to superconductivity, energy storage and solar cells.

Dr Ramesh Bhave is a Distinguished Staff and a Team Leader at ORNL. He earned his PhD degree in Chemical Engineering from the University of Bombay, India. After completing his postdoctoral research at the Stevens Institute of Technology, Hoboken, New Jersey, USA, he joined industry and worked for Alcoa, US Filter and Pall Corporation in the USA for a span of over 21 years before joining ORNL in early 2008. He has authored or co authored more than 50 journal publications and has 12 issued US Patents related to membrane separations covering a wide range of applications.

Dr Bruce A. Moyer is a Corporate Fellow at ORNL, leading the Chemical Separations Group. He received his BS degree in chemistry from Duke University, USA, in 1974 and his PhD in inorganic chemistry from the University of North Carolina at Chapel Hill, USA, in 1979. In addition to leading programmes in fundamentals of extraction, nuclear-fuel cycle separations and critical materials, he led the chemical development of the caustic side solvent extraction process in use for caesium removal from millions of gallons of nuclear waste. Dr Moyer is Co-editor of the journal Solvent Extraction and Ion Exchange and the book series Ion Exchange and Solvent Extraction.

Dr Stephen Harrison is the Chief Technology Officer (CTO) of Alger Alternative Energy, LLC, USA. Previously, he served as the CTO of Simbol, Inc, USA. Dr Harrison led Simbol’s development of Li, Mn and Zn extraction for existing geothermal power plants and invented the process used by Limtech, Inc, USA, for the production of high-purity lithium carbonate. He has more than 40 patents and 12 publications. Dr Harrison has a Chemistry degree from Loughborough University, UK, and a PhD in Chemical Engineering from the University of Newcastle upon Tyne, UK. He has spent many years developing sustainable chemical processes firstly with Hydro‑Quebec, Canada, and more recently in California with AIC Labs Inc, USA.

www.technology.matthey.com

All-Solid-State Batteries and their Remaining Challenges A potential route towards safer, higher performing batteries

Jitti Kasemchainan*, Peter G. State-of-the-Art in All-Solid-State Bruce** Batteries Department of Materials, University of Oxford, Parks Road, Oxford OX1 3PH, UK The main component of all-solid-state batteries is a solid electrolyte, which can be ceramic, Email: *[email protected], glass, polymer or a mixture. The differences in **[email protected] the electrical, electrochemical and mechanical properties of a solid electrolyte compared to the more familiar liquid electrolytes are key to the All-solid-state batteries, which utilise a solid challenges in all-solid-state batteries. At room electrolyte in place of liquid electrolytes, have temperature, the Li-ion conductivity of a solid the potential for higher energy densities and electrolyte is usually at least two or three orders of greater safety than current lithium-ion batteries. magnitude lower than that of a liquid electrolyte, However they still face many challenges before especially in the case of solid polymers (4). This the technology is ready to be commercialised. can result from the solid electrolyte’s intrinsic This short report summarises the current state properties or from existing grain boundaries. of knowledge in all-solid-state batteries including However, the conductivities of some sulfide-based the electrical, electrochemical and mechanical electrolytes like Li10GeP2S12 (LGPS), Li7P3S11 and properties of the electrolytes, and the challenges Li6PS5X (X = Cl, Br, I) are comparable to or even that remain to be overcome in their development higher than those of liquid electrolytes (5–7). and processing. It seems widely accepted and reported that the electrochemical oxidation potentials of Introduction solid electrolytes are superior to those of liquid electrolytes. Solid electrolytes may be stable There is increasing worldwide motivation to above 5.0 V vs. Li0/Li+ . In the case of liquid research and develop all-solid-state batteries electrolytes, it has been shown that decomposition in order to achieve better safety, higher power occurs above 4.0 V vs. Li0/Li+ (8). In fact, recent and energy density, as well as wider operating results from density functional theory (DFT) temperature energy storage (1–3) as compared computations (9–11) on the thermodynamic to conventional lithium-ion batteries. Liquid stability of various solid electrolytes hint that electrolytes used in Li-ion batteries are based on above 4.0 V vs. Li0/Li+ most solid electrolytes organic solvents, which are intrinsically volatile and can be oxidised and decomposed into different highly flammable. In contrast, solid electrolytes phases. Even though there is little experimental are usually able to withstand high temperatures evidence supporting such computational data, (>80°C). Several challenges related to solidifying work by Auvergnoit et al. (12) gives an idea of batteries still remain to be addressed from the possible decomposition of Li6PS5Cl towards fundamental understanding before the technology ordinary positive electrode materials of either will be ready for widespread commercialisation. LiCoO2 (LCO), LiNi1/3Co1/3Mn1/3O2 (NCM)

or LiMn2O4 (LMO). With X-ray photoelectron However, there is no fully developed mechanism spectroscopy (XPS), Auger electron spectroscopy to describe how Li dendrites start to grow at the (AEM) and scanning electron microscopy (SEM), interface of a solid electrolyte with a Li metal it was possible to detect elemental sulfur, lithium electrode and the properties of solid electrolytes polysulfides, P2Sx (x ≥5), phosphates and LiCl at the responsible for Li dendrite growth are not yet interface of Li6PS5Cl with the positive electrodes. well established. Newman and Monroe related Similarly, Koerver et al. (13) used XPS and SEM to dendrite growth to the shear modulus of polymer discover interphase formation between a β-Li3PS4 electrolytes (25, 26). Porz et al. (27) proposed that solid electrolyte and a NCM electrode material. a surface crack or Griffith-like fracture in some If the aim is for all-solid-state batteries that can glassy, polycrystalline and single crystalline solid achieve higher energy density than conventional electrolytes can invoke progressive crack opening Li-ion batteries containing liquid electrolytes, for Li metal to plate and penetrate through. implementing Li metal as the negative electrode is Unlike liquid electrolytes, solid electrolytes are in crucial (2). Reactivity of solid electrolytes against general not readily deformable. Thus, preparation Li metal has to be considered. The electrochemical of ceramic or glassy electrolytes including assembly reduction potential of solid electrolytes needs with electrodes and conductive carbonaceous to be close to the redox reaction of Li metal, materials is quite specific. For example, the which is set as a reference at 0 V vs. Li0/Li+. A process requires significant external pressure

Garnet-type (Li7La3Zr2O12 (LLZO)) is the only stable higher than 100 MPa for cell assembly (1, 12, solid electrolyte towards Li metal; up to now there 28–31). Note that the active negative electrodes is no real proof of its reactivity after exposure to Li used in the referenced literature were indium, metal. Other solid electrolytes will undergo reduction graphite or Li4Ti5O12 (LTO), possibly to avoid the reactions by consuming Li-ions and electrons from problem of Li metal dendrites. During cycling of Li metal to form an interphase layer, which is the all-solid-state batteries, currently available so-called solid electrolyte interphase (SEI). The electrode materials such as sulfur (32), NCM (13) electrical properties of this SEI layer play a role in or even LCO (28) experience a volumetric change how the reaction between solid electrolytes and Li when being lithiated and delithiated. This diminishes metal continues (14–16). If this layer is a mixed the physical contact between the electrode and conductor of Li-ions and electrons, it will continue to solid electrolyte phases, subsequently impeding grow as long as contact is maintained between the the batteries’ cyclability or capacity retention. solid electrolyte and Li metal. Many solid electrolytes are in this category: for example, Li10GeP2S12 (15, Remaining Challenges 16), Li10SiP2S12 (15, 16), Li10Si0.3Sn0.7P2S12 (15,

16) and Li1+xAlxGe2–x(PO4)3 (17). If the layer is a Thick composite positive electrode layers (high Li-ion conductor and minimally an electron active mass loading) and thin solid electrolyte conductor, the SEI acts as a protective layer layers need to be considered for all-solid batteries to prevent further reactions between the solid so as to achieve favourable energy and power. electrolyte and Li metal. This behaviour is observed Most all-solid-state batteries in the literature for lithium phosphorous oxynitride (LiPON) (18), (1, 12, 13, 28) exhibit areal capacities less than –2 Li7P3S11 (19) and Li6PS5X (X = Cl, Br) (19). 1.0 mAh cm and operate at a C-rate around Another important issue in all-solid-state batteries 0.1 C, particularly for charging. Commercial when using a Li metal electrode is Li dendrite Li-ion battery electrodes (33) for cell assembly growth, potentially causing short circuit. This is with liquid electrolytes can attain capacities over the case for solid electrolytes that do not react 2.0 mAh cm–2 and C-rates of 0.2 C. This implies with Li metal or react and form a protective layer. that the performance of all-solid-state batteries is Many studies that can be found elsewhere (20–23) limited conceivably by kinetics or mass transfer. point out the growth of Li dendrites in Garnet-type Moreover, there is a lack of feasible processes for electrolytes, even when an electrolyte pellet was assembly and scale-up for all-solid-state batteries densified to obtain a relative density higher than (34, 35). It is also questionable whether available 99.5%. Another interesting report about the sulfide processing techniques from manufacturing electrolyte 80Li2S∙20P2S5 illustrates locations in standard Li-on batteries can be directly exploited SEM micrographs that could be Li dendrites (24). in all-solid-state battery systems.

Conclusion 17. M. Zhang, K. Takahashi, N. Imanishi, Y. Takeda, O. Yamamoto, B. Chi, J. Pu and J. Li, J. Electrochem. Though all-solid-state batteries have extensively Soc., 2012, 159, (7), A1114 attracted attention in academia and industry, a 18. A. Schwöbel, R. Hausbrand and W. Jaegermann, number of known and unknown challenges make the Solid State Ionics, 2015, 273, 51 technology immature. To accomplish comprehension 19. S. Wenzel, S. J. Sedlmaier, C. Dietrich, W. G. Zeier and eventual commercialisation of the technology, and J. Janek, Solid State Ionics, 2017, In Press, long-term interdisciplinary research and development Corrected Proof in science and engineering will be required. 20. E. J. Cheng, A. Sharafi and J. Sakamoto, Electrochim. Acta, 2017, 223, 85 References 21. F. Aguesse, W. Manalastas, L. Buannic, J. M. Lopez del Amo, G. Singh, A. Llordés and J. Kilner, ACS 1. Y. Kato, S. Hori, T. Saito, K. Suzuki, M. Hirayama, Appl. Mater. Interfaces, 2017, 9, (4), 3808 A. Mitsui, M. Yonemura, H. Iba and R. Kanno, 22. R. H. Basappa, T. Ito, T. Morimura, R. Bekarevich, Nature Energy, 2016, 1, (4), 16030 K. Mitsuishi and H. Yamada, J. Power Sources, 2. J. Janek and W. G. Zeier, Nature Energy, 2016, 1, 2017, 363, 145 (9), 16141 23. A. Sharafi, H. M. Meyer, J. Nanda, J. Wolfenstine 3. A. L. Robinson and J. Janek, MRS Bull., 2014, 39, and J. Sakamoto, J. Power Sources, 2016, (12), 1046 302, 135 4. J. G. Kim, B. Son, S. Mukherjee, N. Schuppert, 24. M. Nagao, A. Hayashi, M. Tatsumisago, A. Bates, O. Kwon, M. J. Choi, H. Y. Chung and S. T. Kanetsuku, T. Tsuda and S. Kuwabata, Phys. Park, J. Power Sources, 2015, 282, 299 Chem. Chem. Phys., 2013, 15, (42), 18600 5. N. Kamaya, K. Homma, Y. Yamakawa, M. Hirayama, 25. C. Monroe and J. Newman, J. Electrochem. Soc., R. Kanno, M. Yonemura, T. Kamiyama, Y. Kato, S. 2004, 151, (6), A880 Hama, K. Kawamoto and A. Mitsui, Nature Mater., 2011, 10, (9), 682 26. C. Monroe and J. Newman, J. Electrochem. Soc., 2005, 152, (2), A396 6. Y. Seino, T. Ota, K. Takada, A. Hayashi and M. Tatsumisago, Energy Environ. Sci., 2014, 7, 27. L. Porz, T. Swamy, B. W. Sheldon, D. Rettenwander, (2), 627 T. Frömling, H. L. Thaman, S. Berendts, R. Uecker, W. C. Carter and Y.-M. Chiang, Adv. Energy Mater., 7. M. A. Kraft, S. P. Culver, M. Calderon, F. Böcher, T. 2017, 7, (20), 1701003 Krauskopf, A. Senyshyn, C. Dietrich, A. Zevalkink, J. Janek and W. G. Zeier, J. Am. Chem. Soc., 2017, 28. W. Zhang, D. A. Weber, H. Weigand, T. Arlt, I. 139, (31), 10909 Manke, D. Schröder, R. Koerver, T. Leichtweiss, P. 8. Q. Li, J. Chen, L. Fan, X. Kong and Y. Lu, Green Hartmann, W. G. Zeier and J. Janek, ACS Appl. Energy Environ., 2016, 1, (1), 18 Mater. Interfaces, 2017, 9, (21), 17835 9. Y. Zhu, X. He and Y. Mo, J. Mater. Chem. A, 2016, 29. Y. Ito, Y. Sakurai, S. Yubuchi, A. Sakuda, A. 4, (9), 3253 Hayashi and M. Tatsumisago, J. Electrochem. Soc., 2015, 162, (8), A1610 10. Y. Zhu, X. He and Y. Mo, ACS Appl. Mater. Interfaces, 2015, 7, (42), 23685 30. T. Ohtomo, A. Hayashi, M. Tatsumisago, Y. Tsuchida, S. Hama and K. Kawamoto, J. Power 11. W. D. Richards, L. J. Miara, Y. Wang, J. C. Kim and Sources, 2013, 233, 231 G. Ceder, Chem. Mater., 2016, 28, (1), 266 12. J. Auvergniot, A. Cassel, J.-B. Ledeuil, V. Viallet, 31. D. H. Kim, D. Y. Oh, K. H. Park, Y. E. Choi, Y. J. V. Seznec and R. Dedryvère, Chem. Mater., 2017, Nam, H. A. Lee, S.-M. Lee and Y. S. Jung, Nano 29, (9), 3883 Lett., 2017, 17, (5), 3013 13. R. Koerver, I. Aygün, T. Leichtweiß, C. Dietrich, W. 32. M. Nagao, A. Hayashi, M. Tatsumisago, T. Ichinose, Zhang, J. O. Binder, P. Hartmann, W. G. Zeier and T. Ozaki, Y. Togawa and S. Mori, J. Power Sources, J. Janek, Chem. Mater., 2017, 29, (13), 5574 2015, 274, 471 14. S. Wenzel, T. Leichtweiss, D. Krüger, J. Sann and 33. ‘Li-ion Battery Electrode/Li Chips’, MTI Corp: http:// J. Janek, Solid State Ionics, 2015, 278, 98 www.mtixtl.com/li-ionbatteryelectrodelichips. aspx (Accessed on 30th January 2017) 15. P. Bron, B. Roling and S. Dehnen, J. Power Sources, 2017, 352, 127 34. K. Kerman, A. Luntz, V. Viswanathan, Y.-M. Chiang 16. S. Wenzel, S. Randau, T. Leichtweiß, D. A. Weber, and Z. Chen, J. Electrochem. Soc., 2017, 164, J. Sann, W. G. Zeier and J. Janek, Chem. Mater., (7), A1731 2016, 28, (7), 2400 35. Y.-S. Hu, Nature Energy, 2016, 1, (4), 16042

The Authors

Jitti Kasemchainan is a postdoctoral researcher with Peter G. Bruce’s group at the Department of Materials, University of Oxford, UK. He completed an Erasmus Mundus Master Course of Materials for Energy Storage and Conversion (MESC) in 2011. He carried out a doctoral thesis at Robert Bosch GmbH in Gerlingen, Germany, and received his PhD in Mechanical Engineering from Karlsruhe Institute of Technology (KIT), Germany, in 2015. His current research is focused on solid electrolytes and the electrode-electrolyte interfaces for rechargeable batteries.

Peter Bruce’s research interests embrace materials chemistry and electrochemistry, especially lithium and sodium batteries. Recent efforts have focused on the synthesis and understanding of new materials for lithium-ion batteries, on understanding anomalous oxygen redox processes in high capacity Li-ion cathodes, the challenges of the lithium- air battery and the influence of order on the ionic conductivity of polymer electrolytes. His research has been recognised by a number of awards and fellowships, including from the Royal Society, the Royal Society of Chemistry, the German Chemical Society and The Electrochemical Society. He was elected to the Royal Society in 2007 and the Royal Society of Edinburgh in 1994. As well as directing the UK Energy Storage Hub, SUstainable PowER GENeration and supply (Supergen), Peter is Chief Scientist in the Faraday Institution, the UK centre for research on electrochemical energy storage.

www.technology.matthey.com

“Electrochemistry: Volume 14” Edited by Craig Banks (Manchester Metropolitan University, UK) and Steven McIntosh (Lehigh University, Bethlehem PA, USA), Specialist Periodical Reports in Electrochemistry, Royal Society of Chemistry, Cambridge, UK, 2017, ISBN: 978-1-78262-114-0, £314.95

Reviewed by John Blake*, Angus This work considers both the direct molecular Dickinson, Massimo Peruffo adsorption and the dissociative adsorption of Johnson Matthey, Lydiard Fields, Great borohydride with a view to understanding the Western Way, Swindon, SN5 8AT, UK selectivity of this reaction over the competing hydrogen evolution reaction. The paper looks at *Email: [email protected] surfaces of increasing levels of complexity starting with single facet metal surfaces through to metal alloy surfaces. In all cases robust calculations produce adsorption energies, which are presented Introduction along with diagrams showing the catalytic surface and adsorption sites clearly. The correlation of “Electrochemistry: Volume 14” is a collated book these results to experiment is discussed although of five papers edited by Craig Banks (Manchester not comprehensively compared. Metropolitan University, UK) and Steven McIntosh The effects of solvation are next investigated (Lehigh University, Bethlehem PA, USA), both along with some calculations on metal of whom are well established in the field with nanoparticles, bringing the discussion onto a research interests covering the topics in the book. more complex but realistic thread. The solvation The book is one of a series which aims to collate study highlights effects such as changes in the and summarise the key topics receiving attention orientation of the adsorbing molecule with the within the electrochemical literature. concomitant change in adsorption energy. The nanoparticle study draws attention not only to Electrochemistry and Materials the presence of multiple facets which all need to be considered but also to the difference in Development accessibility to the active sites on nanoparticles Chapter 1 by Mary Clare Sison Escaño is and the consequence of this on the selectivity of titled ‘Borohydride Electro-Oxidation on Metal a given catalyst. Electrodes: Structure, Composition and Solvent Chapter 2 by Zhongyang Wang et al. is ‘Recent Effects from DFT’. Within this work the electro- Progress in the Development of Anion Exchange oxidation of borohydride on well-defined catalytic Membranes for Electrochemical Devices’. surfaces is investigated by using density functional This article is a general review of the latest theory (DFT) modelling to calculate the adsorption developments in the field of anion exchange energy of the reactive intermediates onto specific membranes (AEM) used in electrochemical sites on the catalyst surface, Equation (i): devices. The first two chapters review a number – – BH4 + * (metal site) → BHy* + (4–y)H* + e (i) of polymers reporting the advantages and

181 © 2018 Johnson Matthey https://doi.org/10.1595/205651318X696756 Johnson Matthey Technol. Rev., 2018, 62, (2) weaknesses of the backbone and functional group Pt alloys and the other palladium alloys. Both chemistries. The authors highlight the superior sections give a detailed review of the mechanisms durability (1) for a long-alkyl chain quaternary of oxidation on the catalyst and how this is altered ammonium onto polyphenylene (2) and reports based on the type, supporting material and a number of strategies to increase the chemical morphology of the catalyst used. The material durability of the functional groups, among others presented is comprehensive and nicely combines the removal of the alpha hydrogen to suppress the multiple analytical techniques to give a well-rounded Hoffman elimination (Figure 1). review of the subject area. Particular attention In Chapter 3, ‘Anodic Materials for Electrooxidation has been paid to the effect of the morphology of of Alcohols in Alkaline Media’, Sadia Kabir and Pd-based catalysts on the electrochemical activity Alexey Serov have produced a well-rounded of the materials. This chapter does well in explaining overview on the topic that would be useful for the reasons for the benefits in electrochemical people just starting in the field. A summary of the performance seen by combining Pt and Pd with workings of an AEM fuel cell is given that contrasts different materials rather than just quoting the it nicely with the better known proton-based higher performance. fuel cells. The different oxidation mechanisms in In Chapter 4, ‘Newer Polymer Electrolytes and alkaline media for a range of fuels are presented Electrocatalysts for Direct Alcohol Fuel Cells’, P. including methanol, ethanol and ethylene glycol. Sridhar, S. D. Bhat and A. K. Sahu have produced The advantages of the different fuels are discussed, a good summary of recent advances in membranes as well as some of the potential downfalls, such and electrocatalysts for direct alcohol fuel cells as incomplete oxidation for larger molecules. This (DAFC). The chapter is well structured with the section is slightly let down by typographical errors material being broken down into proton conducting in some of the mechanistic equations presented, membranes, anion conducting membranes and though the overall presentation is clear and easy catalysts for DAFC. to follow. The authors make good use of tables to The section on proton conducting membranes highlight important reference data. has different membranes and inorganic materials The bulk of the chapter gives an in-depth review organised into different sections. The authors of both the different catalytic materials capable give clear definitions for the difference between of electrochemically oxidising alcohols in alkaline composite and hybrid membranes, highlighting the media and the advantages of these materials over effects that the different methods for incorporating the traditional platinum catalyst. The review is the inorganic into the membrane have on its broken down into two sections, one dealing with properties. These differences in properties are

Fig. 1. Degradation N mechanism for + quaternary ammonium- H2C + N based anion exchange H2C membranes OH– OH

+ CH3OH CH CH3 H2C + 3 H2C N N CH3 CH3 CH3 OH–

H + OH– R + + N N H2O R

182 © 2018 Johnson Matthey https://doi.org/10.1595/205651318X696756 Johnson Matthey Technol. Rev., 2018, 62, (2) then related to in-cell performance data. However, within electrocatalysis. These include using MOFs the chapter’s reproduction quality lets the subject as catalytically active materials and as supports material down as it can be hard to discern what and precursors to electrocatalysts for a number data belongs to which sample. This is a shame as of electrochemical transformations. Similar to the structural images of the inorganic materials previous chapters this article gives a succinct shown are very crisp and clear. summary of the state of MOF research in the A whole section is dedicated to sulfonated electrochemical literature. poly(ether ether ketone) (sPEEK)-based The book highlights other applications for AEM membranes rather than just concentrating on particularly as a separation membrane in vanadium Nafion based materials. The section on sPEEK redox flow batteries (VRFB). The increase in reviews a range of different materials made by demand for VRFB is linked to the increasing demand blending additives into the polymer material and for storage systems to support renewable energy. the effect these additives have on the structure and properties of the membrane. This section is then Conclusions nicely rounded off with a direct comparison of the different membrane types. The membrane section This book gives a good feel for the type of areas ends with a brief overview on alkaline membrane in which electrochemistry can be used to develop materials. the understanding of electrocatalysis and materials The authors then move onto presenting a nice development. While this book does not give a summary of recent advances in electrocatalysis detailed review of the literature for any specific for DAFC. This covers catalysts for both methanol field of study it does provide an interesting sample and ethanol oxidation, as well as the corresponding of papers across the field. This book is not suitable tolerant cathode catalysts. Overall the chapter is a for a researcher in the field looking for a detailed good review on the subject matter and would be literature review, however it is successful in useful reference material for people interested in providing a glimpse into the key topics and themes this topic. currently receiving the most attention. In the third and fourth chapters the main "Electrochemistry: degradation modes are described along with their Volume 14" impact on the AEM performances related to the loss of ionic exchange capacity (3). The authors also examine key characterisation techniques to detect and quantify mechanical and chemical degradation mechanisms. Chapter 5 by Y. Luo and N. Alonso-Vante reports the performances and durabilities of a number of AEM under a well-established test protocol (4, 5). It is titled ‘Application of Metal Organic Framework (MOF) in the Electrocatalytic Process’. This chapter starts by providing a simple summation of why MOF materials are becoming of greater interest in the electrocatalytic arena, namely as an alternative References to the high cost of the commonly used platinum group metals (pgms). It then goes on to summarise 1. J. Parrondo, M. J. Jung, Z. Wang, C. G. Arges and the wide range of MOF materials available V. Ramani, J. Electrochem. Soc., 2015, 162, (10), containing many complex three dimensional (3D) F1236 structures which are of interest in their own right. 2. M. R. Hibbs, J. Polym. Sci., Part B: Polym. Phys., These materials however can often possess some 2013, 51, (24), 1736 interesting catalytic properties and with the wide 3. C. G. Arges, L. Wang, J. Parrondo and V. Ramani, range of structures it could be possible to ‘fine J. Electrochem. Soc., 2013, 160, (11), F1258 tune’ these catalytic properties. 4. C. G. Arges, J. Parrondo, G. Johnson, A. Nadhan Following on from the previously described and V. Ramani, J. Mater. Chem., 2012, 22, (9), scene-setting discussion is a number of literature 3733 reviews on key areas where MOF materials are 5. C. G. Arges, L. Wang, M. Jung and V. Ramani, J. being investigated in the academic arena for use Electrochem. Soc., 2015, 162, (7), F686

The Reviewers

John Blake joined Johnson Matthey in 2012. He is currently a Senior Scientist within the Technology group in Fuel Cells. His work focuses on technical customer support and the development of next generation anode components.

Massimo Peruffo joined Johnson Matthey in 2015. He is currently a Lead Scientist, Quality Control and Characterisation Laboratory Manager in Fuel Cells. The main focus of his research is to define and develop new characterisation tools to support the technology department.

Angus Dickinson joined Johnson Matthey in 2006. He is currently a Lead Scientist in the Technology group in Fuel Cells. The main focus of his work is to develop new MEA

www.technology.matthey.com

The International Flow Battery Forum 2017 A wide range of topics were covered from academic studies to challenges faced in industry

Reviewed by Marion van Dalen*, The conference opened with a keynote address Julia O’Farrelly by the developer of the VRFB, Professor Emeritus Johnson Matthey, Blounts Court, Sonning Maria Skyllas-Kazacos (University of New South Common, Reading RG4 9NH, UK Wales (UNSW), Australia). She gave an overview of the various flow battery chemistries, the status *Email: [email protected] of their commercial installation and outlined the Battery Storage and Renewable Energy Fund in South Australia. Her talk was accompanied by Introduction plenty of lively discussion and questioning. Jonathan Radcliffe (Birmingham Energy Institute, The Eighth International Flow Battery Forum, UK) spoke about the need for decarbonisation organised by Swanbarton Ltd, UK, focused on of energy, particularly heating and cooling. One industrial applications of redox flow batteries of the proposed solutions is electrification; he (RFB). The conference was held from 27th to spoke about the ‘value of flexibility’ in electricity 29th June 2017 at the Mercure Piccadilly Hotel, storage. The biggest cost benefits will be for Manchester, UK. It was attended by 212 delegates longer duration storage and RFBs are a candidate from all over the world, including flow battery for this. developers, material and component suppliers and academics. Vanadium Redox Flow Battery The conference covered both the vanadium redox flow battery (VRFB) – the most mature The first morning focused on recent progress in technology – and alternative chemistries including the flow battery industry and its prospects for organic redox flow batteries (ORFB). It celebrated further commercialisation. Presentations were the increasing number of large-scale flow battery given by several companies commercialising installations to date. As renewables such as wind VRFBs. Scott McGregor (RedT, UK) is targeting power and photovoltaics (PV) play an increasing the renewables market and regards a flow battery part in energy generation worldwide, the need as a machine for energy storage, offering multiple for storage of electricity becomes crucial. Flow stacked services and heavy use, rather than a batteries are well suited to this application, battery (provision of power, infrequent use for a particularly when storage for multiple hours is single application). Sumitomo Electric Industries, required. Throughout the conference, there was Ltd has been developing VRFBs for over 30 years; a focus on further commercialisation of RFB Toshikazu Shibata (Sumitomo Electric Industries, technology and what is needed to achieve this. Ltd, Japan) presented performance data for its All participants saw the need to further promote 15 MW/60 MWh demonstration plant in Hokkaido, understanding of the benefits of flow batteries Japan. This plant is integrated with a large PV amongst energy storage system developers and installation, it was started up in 2015 and is able investors. to achieve both frequency regulation and longer

185 © 2018 Johnson Matthey http://dx.doi.org/10.1595/205651318X696819 Johnson Matthey Technol. Rev., 2018, 62, (2) duration renewables integration. Peter Fischer of vanadium salts. A 2 M vanadium electrolyte (Fraunhofer ICT, Germany) gave an update on the solution is stable to precipitation between 15°C project RedoxWind, which will link a 2 MW/20 MWh and 40°C; decreasing the vanadium concentration VRFB with a 200 kWh lithium-ion battery and a to 1.6 M increases the stability window but wind turbine. decreases the energy density. Inorganic additives Xiangkun Ma (Dalian Rongke Power Co, Ltd) developed at UNSW have allowed the use of 3 M described the construction of its 200 MW/800 MWh vanadium solutions with increased energy density demonstration plant on the Dalian peninsula, China. and without significant precipitation. The first phase (400 MWh) was due for completion –1 end-2017. The factory has capacity for 300 MW year Vanadium Supply Chain stack manufacturing capacity; all electricity used during the production process is from PV. Professor Terry Perles (TTP Squared, Inc, USA) presented Dr Huamin Zhang (Dalian Institute of Chemical supply and price data for vanadium and noted that Physics, Chinese Academy of Sciences, China) three quarters of the global vanadium supply is as a described further developments by Rongke Power coproduct of steel slag. Alberto Arias (Arias Resource and the Dalian Institute of Chemical Physics which Capital (ARC) Management LP, USA) highlighted included a non-fluoride ion conducting membrane that vanadium is the largest cost component of a with very good capacity retention, improved bipolar VRFB and that price volatility is a major challenge plate technology and high power density stacks. in their widespread deployment. Arias proposed Liyu Li (UniEnergy Technologies (UET), LLC, USA) that a vanadium leasing model, where a financial presented a competitive value proposition for VRFB intermediary maintained title to the vanadium, compared to Li-ion. He demonstrated the increased could help reduce the upfront capital expenditure cycle life of VRFB compared to other battery (CAPEX) and smooth out vanadium volatility risks. technologies and showed how the economics of VRFB improve at longer charge/discharge Safety durations, making them particularly suitable for renewable energy integration. UET is using mixed Jonathan Buston (Health and Safety Laboratory, UK) hydrochloric/sulfuric acid electrolytes developed at talked about safety issues in the battery industry. Pacific Northwest National Laboratory, USA. RFB resemble chemical plants and face standard Mianyan Huang (Pu Neng, China) described a chemical engineering challenges of process and proprietary engineered plastic membrane with low heat control. Additional complexity comes from permeation rates, high efficiency and low cost. the integration of the RFB control system with the The membrane will be used in a prototype VRFB at National Grid control system. 500 kW scale in 2017. Both Li and Huang agreed that growth in the renewable energy storage Technical Issues market is led by China. In 2016, 15% of total wind energy and 20% of PV generation were curtailed; Dr Fikile Brushett (Massachusetts Institute of an opportunity for storage. Technology (MIT), USA) probed pore-scale mass Lars Moellenhoff (Glex Energy Storage (formerly transport in RFBs and its effect on the performance Gildemeister), Germany) highlighted that the key of the system. He looked at different reactor scales hurdle for developing RFB projects is to convince and developed a way to use a smaller reactor which investors to think beyond the 5–10 years typical replicates the flow of a larger system. of the energy storage industry and to consider the total cost of ownership (TCO) benefits. GLex’s Component Development for Flow positioning as an integrated, independent power Batteries producer should help to overcome this. Thorsten Seipp (volterion GmbH, Germany) Energy Density presented the development of fully welded stack technology. Continuous graphite plates are welded RFBs for energy storage are generally housed in onto thermoplastics, giving a compact new stack containers of limited volume; hence it remains design with very thin (<2.5 mm) half cells. The important to maximise the energy density of the smaller, thinner stack allows material savings and vanadium electrolyte solution. Professor Emeritus automated production. Skyllas-Kazacos described work in her group Jan Girschik (Fraunhofer UMSICHT, Germany) to develop additives to prevent precipitation is developing a continuous process for making

186 © 2018 Johnson Matthey http://dx.doi.org/10.1595/205651318X696819 Johnson Matthey Technol. Rev., 2018, 62, (2) extra-large bipolar plates. The requirement for added to increase the safety and the energy density a bipolar plate is that it is liquid- and gas-tight, of the electrolyte but they reduce the performance corrosion and chemical resistant, with high of the membranes. Elestor has investigated the use conductivity and mechanical stability. Low cost is of competing ions, different operating conditions also an important feature. and regeneration of the membrane. Pekka Peljo (École polytechnique fédérale Organic Redox Flow Batteries de Lausanne (EPFL), Switzerland) described a copper-acetonitrile (Cu-ACN) RFB which can be There were several presentations on ORFBs. charged by electricity or low-grade heat (~100°C), This technology has attracted interest because see Figure 1. Replacing water with propylene the organic electrolyte offers the potential of carbonate electrolyte solvent helps maintain the significantly lower capital cost compared to cell voltage at 1.3 V. vanadium systems. In a keynote lecture, Professor Ian Whyte (WhEST, UK) has developed generic Michael Aziz (Harvard University, USA) described flow battery stack technology for use with a quinones and other electrolytes for aqueous organic range of battery chemistries, for example an iron RFBs. In principle, low cost organic electrolytes with ferricyanide system. Michael Tucker (Lawrence a shorter lifetime could be replaced periodically Berkeley National Laboratory, USA) described a and still offer a low cost solution. low-cost all-Fe flow battery for use in, for example, Thibault Godet-Bar (Kemwatt, France) is making mobile phone charging and emergency lighting in prototype ORFBs and plans to build an industrial the developing world. The electrolyte is a non-toxic scale, 200 kWh ORFB in 2018. It has identified disposable iron solution. a low-cost, readily scalable organic electrolyte and achieved 7000 cycles. Its cost target is Conclusions US$300 kWh–1 in three years’ time. Steven Reece (Lockheed Martin Energy, USA) This very interesting conference covered the whole noted that design for low cost must be approached spectrum of topics, from academic studies to from the system level. The Lockheed Martin challenges faced in the further commercialisation system offers a tuneable metal-ligand electrolyte, of flow batteries on an industrial scale. Much of an ion-selective membrane with low crossover the conference focused on longer duration (multi and optimised stack design. The chemistry is hour) energy storage for integration of PV and based on earth-abundant transition metals and wind energy. In order to compete with technologies easy-to-synthesise ligands. such as Li-ion, capital cost reduction remains a key focus.

Other Chemistries for Flow Batteries Reference

Natalia Mazur (Elestor, The Netherlands) described 1. P. Peljo, D. Lloyd, N. Doan, M. Majaneva and next generation flow batteries based on the K. Kontturi, Phys. Chem. Chem. Phys., 2014, 16, (7), 2831 HBr/Br2 couple. Bromine complexing agents are

Redox flow battery Thermal Cu regeneration Cu g Cu+ + e– Heat (100ºC) Electricity Cu2+ + e– g Cu+ 2Cu+ g Cu + Cu2+ Cu2+

Cu+

Fig. 1. Schematic of the Cu-ACN system (Reproduced from (1) with permission of the PCCP Owner Societies)

The Reviewers

Marion van Dalen holds a MSc degree from Delft University of Technology, The Netherlands, in Chemical Engineering, specialising in Process Engineering. After graduating, she joined Johnson Matthey as a Research Scientist in the field of membranes.

Julia O’Farrelly holds a BSc (Hons) degree in Chemistry and works as Principal Information Analyst in the Technology Forecasting and Information group at Johnson Matthey. She is interested in the development and commercialisation of new technologies.

www.technology.matthey.com

Effect of Temperature and Catholyte Concentration on the Performance of a Chemically Regenerative Fuel Cell POM-based catholytes for platinum-free polymer electrolyte fuel cells

By David B. Ward, fuel combustion for both transport and stationary Trevor J. Davies*# applications (1, 2). Despite the recent dominance Department of Natural Sciences, Faculty of of lithium-ion batteries in the ‘electrochemical Science & Engineering, University of Chester, power’ sector, hydrogen PEFCs are beginning to Thornton Science Park, Pool Lane, Ince, make a market impact, for example, with the Chester CH2 4NU, UK launch of the Toyota Mirai and the increasing uptake of PEFCs in the materials handling sector *Email: [email protected] (3, 4). In addition, governments are beginning the transition to a hydrogen economy, with #Present address: Electrochemical Germany having the most ambitious target of Technology Technical Centre, INOVYN, South 400 hydrogen refuelling stations by 2023 (5). Parade, Runcorn, Cheshire WA7 4JE, UK However, issues around cost and durability have plagued PEFC development and continue to inhibit their widespread commercialisation (6). Chemically regenerative redox cathode (CRRC) In a conventional PEFC, hydrogen is oxidised to polymer electrolyte fuel cells (PEFCs) are attracting protons and electrons at the anode (Equation (i)) more interest as a platinum-free PEFC technology. and oxygen (in air) is reduced to water at the These fuel cells utilise a liquid catalyst or catholyte, cathode (Equation (ii)): to perform the indirect reduction of oxygen, + – H2 → 2H + 2e (i) eliminating the major degradation mechanisms that plague PEFC durability. A key component of – + O2 + 4e + 4H → 2H2O (ii) a CRRC PEFC system is the catholyte. This article reports a thorough study of the effect of catholyte Overall: concentration and temperature on CRRC PEFC 2H2 + O2 → 2H2O (iii) system performance for H7PV4Mo8O40 and

Na4H3PV4Mo8O40, two promising polyoxometalate The main cause of the cost and durability issues (POM)-based catholytes. The results suggest 80°C with conventional PEFCs is the direct four-electron and a catholyte concentration of 0.3 M provide reduction of oxygen at the cathode (Equation the optimum performance for both H7PV4Mo8O40 (ii)). Due to relatively slow kinetics (1, 7), high and Na4H3PV4Mo8O40 (for ambient pressure Pt loadings are required to catalyse the oxygen operation). reduction reaction (ORR), increasing the cost of the membrane electrode assembly (MEA) (8). Up 1. Introduction to 80% of the Pt in a conventional fuel cell can be on the cathode. Furthermore, the large overpotential For decades, PEFCs have been promoted as the required to induce the oxygen reduction reaction future replacement for power generation via fossil (even with high Pt loading) typically results in a

200 mV drop in operating voltage at relatively low to 1.23 V vs. the standard hydrogen electrode as current densities of ~0.2 A cm–2 (9). The presence possible; and (b) display fast regeneration kinetics of air at the cathode is also the key component (i.e. the chemical reaction between oxygen and the in the major mechanisms of fuel cell degradation, reduced catholyte) to ensure the catholyte is highly including chemical degradation of membranes via oxidised when leaving the regenerator. Tolmachev peroxide species (10, 11) and voltage transients and Vorotyntsev recently published a thorough at cell start-up and shut-down that oxidise the review of CRRC PEFCs (14), with a focus on all catalyst’s carbon support (12, 13). catholyte chemistries researched since the CRRC An alternative solution to the problem of cost concept was first proposed by Posner in 1955 (15). and durability is the CRRC PEFC, illustrated in At present, three aqueous catholyte chemistries Figure 1. These systems utilise the indirect dominate CRRC research and development. These reduction of oxygen and resemble a hybrid of are P-V-Mo POMs that utilise V4+/5+ redox chemistry a fuel cell and flow battery (14). The anode is (16–18); Fe2+/3+ systems with organic ligands essentially identical to that of a conventional selected to enhance the rate of the re-oxidation PEFC, with hydrogen gas oxidised to protons and reaction (19); and an adaptation of the Otswald – electrons at a catalyst (Pt) coated membrane. At process exploiting NO/NO3 redox chemistry (20). the cathode, a catholyte (Cath) is reduced at a Table I lists the best CRRC PEFC performance two-phase liquid | electrode boundary. With facile obtained to date with the three different catholyte kinetics, this electrochemical reaction utilises a chemistries. Transient peak power refers to the low cost porous graphite electrode, replacing the performance of the cell whereas steady state requirement for Pt. The reduced catholyte (H-Cath) peak power is associated with the whole CRRC then travels to an air bubbling device called the system (i.e. regenerator + cell) (16). Given the US ‘regenerator’, where oxygen is reduced to water Department of Energy (DOE) 2020 target for peak and the catholyte is returned to its original oxidised power in transportation PEFCs is 1000 mW cm–2 state. Consequently, gaseous air never enters the (21), the data in Table I suggests CRRC PEFCs cathode, eliminating the major PEFC degradation could soon compete with conventional fuel cells. mechanisms. In addition, the catholyte ensures In terms of power density, the most promising the membrane remains well hydrated, allowing the catholyte formulations are the POMs: H7PV4Mo8O40 use of dry hydrogen. for transient operation and Na4H3PV4Mo8O40 for A key component of a CRRC PEFC system is the steady state operation (18). Furthermore, these catholyte, a redox active electrolyte that reacts with catholytes are associated with excellent durability. dioxygen in its reduced form. To maximise the cell Using 0.3 M Na4H3PV4Mo8O40 in a single cell operating voltage the catholyte should: (a) possess CRRC PEFC system, Ward et al. recently reported a high redox potential in its oxidised state, as close 200 hours operation (at constant current) with

Polymer electrolyte membrane

Hydrogen Catholyte Exhaust (O2 depleted air + H2O)

Anode Cathode Regenerator

+ – + – H2 " 2H + 2e Cath + H + e " H-Cath 4H-Cath + O2 " 4Cath + 2H2O

Air bubbles

Depleted hydrogen Air Pump

Fig. 1. Schematic diagram of a CRRC PEFC system with a hydrogen anode

a Table I Details of Best Performing CRRC PEFC Systems Reported to Date (with H2 as the Fuel) Catholyte Cathode Transient Steady state Reference reaction maximum maximum (simplified) power, power, mW cm–2 mW cm–2

0.5 M FeSO / 1 M H SO 4 2 4 Fe3+ + e– Fe2+ 249 – (19) / iron phthalocyanine →

+ 2 HNO3 + 6 H + – 5 M HNO3 6 e → 2 NO(g) 730 – (20) + 4 H2O

0.45 M POMb – 510 – (17)

5+ – 4+ 0.3 M H6PV3Mo9O40 V + e → V 1000 380 (16) 5+ – 4+ 0.3 M H7PV4Mo8O40 V + e → V 1078 566 (18) 5+ – 4+ 0.3 M Na4H3PV4Mo8O40 V + e → V 864 578 (18) aAll catholytes are aqueous bUndisclosed polyoxometalate negligible loss in performance (18). Indeed, a using the best CRRC catholytes in the literature, similar system was almost commercialised by ACAL H7PV4Mo8O40 and Na4H3PV4Mo8O40. Catholyte Energy Ltd, UK, who achieved over 10,000 hours concentration ranges from 0.2 M to 0.45 M and operation on an automotive test cycle without any temperature from 40°C to 90°C (at ambient significant signs of degradation (22). In their fully pressure). High temperature operation (i.e. above oxidised forms, the vanadium and molybdenum 80°C) is desirable as it has been shown to improve POM constituents have oxidation states +5 and +6, peak power density in both conventional fuel cells respectively. Upon electrochemical reduction in and flow batteries (25, 26). In addition, operating the cell, vanadium(V) is reduced to vanadium(IV) PEFCs above 80°C provides heat management whilst the molybdenum remains unchanged (23). and CO tolerance benefits (27). As such, a reduction level of 100% corresponds to a catholyte where the vanadium and molybdenum 2. Experimental Details have oxidation states +4 and +6, respectively (the reduction of molybdenum occurs at lower A brief description of the experimental equipment potentials). Under most operating conditions, the and procedures is provided below, with more catholytes are partially reduced, with a mixture details in the Supplementary Information. of vanadium(V) and vanadium(IV). Upon contact H7PV4Mo8O40 catholyte (HV4) synthesis followed with oxygen, the vanadium(IV) is chemically the ‘Metallomax’ procedure (28) using the reagents: oxidised to vanadium(V) and O2 is reduced to deionised water (with a resistivity of 18.2 MΩ cm); water, regenerating the catholyte for its next visit vanadium(V) oxide (V2O5) powder (99.2%, Alfa to the cell. Aesar, UK); Mo powder (99.9%, Alfa Aesar, UK);

Little is known regarding the effect of phosphoric acid (H3PO4) (85.0%, Sigma Aldrich, concentration and temperature on CRRC PEFC UK); and molybdenum trioxide (MoO3) (99.5%, systems using H7PV4Mo8O40 and Na4H3PV4Mo8O40 Alfa Aesar, UK). Na4H3PV4Mo8O40 catholyte (NaV4) catholytes, with 0.3 M and 80°C the only reported was produced from HV4 by the addition of NaOH parameters (18). Matsui et al. studied the effect of (98%, Alfa Aesar, UK). Catholyte concentration concentration and temperature on the performance was determined gravimetrically using 25 ml of a similar catholyte, H6PV3Mo9O40 (24). They glass density jars and pre-determined density vs. found a concentration of 0.3 M and a system concentration calibration curves. Adjustment was temperature of 80°C gave the best performance. achieved by either deionised water addition or However, the two values were the limits of their heating and evaporation as required. The matrix of test matrix and the cell performance achieved was concentrations and temperatures studied (for both relatively poor compared to a study with the same HV4 and NaV4) is given in Table II. catholyte (16). In this article, we report the study The cell components and build procedure of a larger temperature-concentration matrix (single cell, 5 × 5 cm active area) were identical

(VWR International, UK) and 827 pH Lab pH meter Table II Temperature and Catholyte (Metrohm, UK). Catholyte redox potentials were Concentration Values Investigated (for measured using a mercurous sulfate reference both HV4 and NaV4) electrode and a JP945 graphite rod (Merson UK) Temperature, Catholyte concentration, as a working electrode. Fully oxidised samples °C M of all catholytes were submitted for 31P nuclear magnetic resonance (NMR) analysis (more details 0.2 0.3 0.4 0.45 in the Supplementary Information). 40 ü ü ü 50 ü 3. Results and Discussion 60 ü 3.1 Catholyte Composition 70 ü

80 ü ü ü ü Aqueous solutions of H3+xPVxMo12–xO40 where x > 1 contain a mixture of species in dynamic ü ü ü 90 equilibrium (29–32). Typically, the vanadium is + present in keggin anions or in its free form of VO2 to that reported previously (16), apart from the (vanadium(V)). Likewise, if the POM is reduced, membrane electrode assemblies. The latter were vanadium(IV) can exist as keggin-bound or free NR212 membranes with a ‘standard’ anode catalyst VO2+. To provide an indication of the speciation in coating (0.3 mg Pt cm–2) and a naked cathode side the catholytes, fully oxidised samples underwent (supplied by Ion Power GmbH, Munich, Germany). 31P NMR analysis. Example spectra are given in Although this is a thicker membrane than the the Supplementary Information. Peak analysis previously used Gore Primea (resulting in poorer followed that of Pettersson and co-workers fuel cell performance), it was found to be robust (30, 31), from which it was possible to estimate + to large temperature variations and gave more concentrations of different species: VO2 (Free V); repeatable results. PV1Mo11O40 keggins (V1-keggin); PV2Mo10O40

The CRRC test rig and experimental procedures keggins (V2-keggin); PV3Mo9O40 keggins were identical to those reported previously (16). (V3-keggin); PV4Mo8O40 keggins (V4-keggin); and All cell tests were conducted using a catholyte flow free phosphate (phosphate). Figure 2 illustrates –1 -1 31 rate of 140 ml min and air flow rate of 1 l min the P NMR speciation results for Na4H3PV4Mo8O40

(into the regenerator at ambient pressure). The and H7PV4Mo8O40 across a range of concentrations anode hydrogen pressure was 1 bar and the at 298 K. cell was operated ‘dead-ended’ with occasional In general, increasing the catholyte concentration hydrogen purge events to remove excess water. increases the concentration of all the species involved The cell temperature was controlled using heating in the dynamic equilibrium. However, pH also plays rods (RS, UK) inserted in the steel end plates whilst a role, which is evident in the difference between the catholyte temperature was controlled with a NaV4 and HV4. For a given POM concentration, CAST-X 300 inline heater (Cast Aluminum Solutions, increasing the pH of the catholyte by adding NaOH USA). An HCP-803 potentiostat (Bio-Logic, France) to convert HV4 to NaV4 results in less free vanadium was used for all cell current-voltage (I–V) curves, and more V4- and V3-keggins. This agrees with the monitoring open circuit voltage and electrochemical equilibrium proposed by Souchay and co-workers impedance spectroscopy. All cell tests were (33) (Equation (iv)): repeated for a range of catholyte reduction levels 4– + (13 – n)[Hn–1PVnMo12–nO40] + 16H apart from impedance measurements, which were 4– (12 – n)[Hn–2PVn–1Mo13–nO40] + only conducted with catholytes at reduction levels + 12VO2 + H3PO4 + 12H2O (iv) corresponding to fuel cell open circuit voltage values of 800 ± 25 mV (see the Supplementary Information POM speciation plays a key role in the regeneration for more details). All cell and rig components reaction. Reduced keggins containing three or displayed good chemical compatibility with the POM more vanadium centres are the only species that catholytes and no degradation of components was react with oxygen at an appreciable rate (34–36). observed throughout testing. Thus, higher concentrations of V3 and V4-keggins Catholyte conductivity and pH were measured are favourable and Figure 2 suggests NaV4 should using a pHenomenal CO1300L conductivity meter undergo faster regeneration than HV4.

(a) (b) 0.50 0.50

0.45 0.45

0.40 Phosphate 0.40

0.35 V4-keggin 0.35 V3-keggin 0.30 0.30 V2-keggin

0.25 V1-keggin 0.25 Free V 0.20 0.20

0.15 0.15 Species concentration, M Species concentration, Species concentration, M Species concentration, 0.10 0.10

0.05 0.05

0 0 0.2 0.3 0.4 0.5 0.2 0.3 0.4 0.5 POM concentration, M POM concentration, M

Fig. 2. Estimated concentrations of species present in fully oxidised samples at 298 K (determined from 31P NMR) for: (a) HV4; (b) NaV4 in a range of catholyte (POM) concentrations

3.2 Ex Situ Thermodynamic Properties Thus, all the redox couples in the catholytes are pH sensitive and their redox potentials should Figures 3(a) and 3(b) illustrate the effect of increase as the pH decreases (i.e. a Nernstian temperature and catholyte concentration on the relationship). This agrees with the trends observed redox potential of HV4 (H7PV4Mo8O40) and NaV4 in Figures 3(b) and 3(d) – both catholytes are

(Na4H3PV4Mo8O40) catholytes at a reduction level acids, so concentration has a strong effect on pH, of 5% (i.e. 5% vanadium(IV), 95% vanadium(V)). leading to a positive correlation between proton Note, measured catholyte redox potentials correlate concentration and redox potential. However, the well with the corresponding CRRC cell open circuit subtle changes in pH with temperature do not voltages (OCV) (16). Similar results were achieved follow the general decline in redox potential as for a range of reduction levels. As seen previously temperature increases, suggesting speciation (18), the protonic catholyte (HV4) has considerably effects may play a role. The conductivity results in higher redox potentials than its sodium analogue Figures 3(e) and 3(f) are expected. Increasing (a consequence of pH). For both catholytes, temperature results in faster moving ions whilst increasing temperature generally decreases the increasing catholyte concentration increases the redox potential and consequently, the cell OCV. number of ions in the solution. Thus, increasing Concentration has the opposite effect. Further concentration should have a positive effect on insights are provided by the pH measurements in cell performance, increasing the catholyte redox Figures 3(c) and 3(d). Both catholytes contain free potential (and consequently cell OCV) and reducing and keggin-bound vanadium. Free vanadium(V) the cell ohmic resistance, whereas the case for undergoes a one-electron two-proton reduction, temperature is more complicated. whereas keggin bound vanadium(V) reduction is thought to proceed via a one-electron one-proton 3.3 Cell Performance mechanism (37) (Equations (v) and (vi)):

+ – + 2+ Figures 4(a) and 4(b) illustrate the transient cell VO2 + e + 2H  VO + H2O (v) performance of 5% reduced HV4 for a range of IV V 4– – [Hn+m–1PV mV n–mMo12–nO40] + e + catholyte concentrations at 80°C (the same trends + IV V 4– H  [Hn+mPV m+1V n–m–1Mo12–nO40] (vi) were observed for a range of reduction levels).

(a) (b) 0.35 0.35 4 4 SO SO 2 2 0.30 0.30

HV4 HV4 vs . Hg/Hg vs . Hg/Hg NaV4 0.25 0.25 NaV4

0.20 0.20 Redox potential, V Redox Redox potential, V Redox

0.15 0.15 40 50 60 70 80 90 0.2 0.3 0.4 0.5 Temperature, ºC Concentration, M (c) (d) 1.2 1.2 HV4 1.0 1.0 NaV4

0.8 0.8

0.6 0.6 HV4 pH pH 0.4 NaV4 0.4 0.2

0.2 0

0 –0.2 40 50 60 70 80 90 0.2 0.3 0.4 0.5 Temperature, ºC Concentration, M (e) (f) 450 500

400 450

350 400 HV4 –1 –1 350 300 NaV4 300 250 250 200 200 150 150 Conductivity, mS cm Conductivity, HV4 mS cm Conductivity, 100 100 NaV4 50 50

0 0 40 50 60 70 80 90 0.2 0.3 0.4 0.5 Temperature, ºC Concentration, M Fig. 3. Thermodynamic properties of 5% reduced HV4 and NaV4 catholytes at: (a), (c) and (e) a range of temperatures for 0.3 M catholyte concentration; and (b), (d) and (f) a range of catholyte concentrations at 80°C

(a) (b) 1.1 1.0 0.2 M HV4 1.0 0.9 0.3 M HV4 0.9 0.8

0.4 M HV4 –2 0.8 0.7 0.45 M HV4 0.7 0.6 0.6 0.5 0.5 0.4

Cell voltage, V Cell voltage, 0.2 M HV4 0.4 0.3 0.3 M HV4 0.3 W cm density, Power 0.2 0.4 M HV4 0.2 0.1 0.45 M HV4 0.1 0 0 0.5 1.0 1.5 2.0 2.5 3.0 0 0.5 1.0 1.5 2.0 2.5 3.0 –2 –2 (c) Current density, A cm (d) Current density, A cm 0.21 0.06 0.2 M HV4

0.20 2 0.3 M HV4 0.05 2 8.75 A 0.45 M HV4

0.19 W cm 15 A W cm 0.04 0.18 25 A

0.17 0.03

0.16 0.02 0.15 Area specific HFR, 0.01

0.14 Area specific reactance,

0.13 0 0.2 0.3 0.4 0.5 0.15 0.20 0.25 0.30 0.35 2 Catholyte concentration, M Area specific resistance,W cm

Fig. 4. CRRC fuel cell performance at 80°C: (a) i–V; (b) corresponding power density curves generated with HV4 catholyte at four different concentrations (labelled); (c) high frequency resistance obtained at three current loads for 0.2–0.45 M HV4; (d) Nyquist plots obtained at 0.6 A cm–2 for 0.2 M, 0.3 M and 0.45 M HV4

Note, the peak power densities are smaller than was not achieved at 0.2 V for the higher catholyte the previously reported value of 1.078 W cm–2 concentrations and, therefore, the performance of for 0.3 M HV4 at 80°C because of the thicker the cell can be considerably improved. membrane (18). The current density-voltage The impedance results in Figures 4(c) and 4(d) (i–V) curves in Figure 4(a) show a considerable provide insights into the cause of the performance improvement in performance on increasing HV4 benefits gained by increasing concentration. The concentration from 0.2 M to 0.3 M, which translates area specific high frequency resistance (HFR) to a 25% increase in peak power. In addition, there values obtained at 0.35 A cm–2, 0.6 A cm–2 and is a noticeable increase in the maximum current 1 A cm–2 (8.75 A, 15 A and 25 A) in Figure 4(c) density. However, the gains from further increasing all show an increase in cell ohmic resistance as the catholyte concentration are much smaller, with catholyte concentration increases – the opposite a 6% increase in peak power achieved for the of that expected from the conductivity results in 0.3→0.45 M transition. Given the latter corresponds Figure 3(f). The reason for this is unclear, but + 2+ to a 50% increase in catholyte cost, the extra power may be linked to VO2 and VO species occupying gained may not be economical. In addition, there sites within the membrane. The Nyquist plots (at is little change in the maximum current density 0.6 A cm–2) in Figure 4(d) show the total cell on increasing the catholyte concentration from resistance (the low frequency intercept) decreasing 0.3 M to 0.45 M. This suggests a limiting current with concentration. Although not shown, the larger

195 © 2018 Johnson Matthey http://dx.doi.org/10.1595/205651318X696800 Johnson Matthey Technol. Rev., 2018, 62, (2) arcs on the right were found to vary with catholyte gained by increasing concentration. As such, flow rate (the smaller arcs on the left were increasing the concentration beyond 0.3 M results unaffected), suggesting those arcs are associated in relatively minor performance improvements. with impedance due to catholyte mass transport. The corresponding situation for 5% reduced NaV4 Therefore, the left arcs must be related to anode is illustrated in Figure 5 (the same trends were impedance and/or catholyte kinetics. As catholyte observed for a range of reduction levels). Note concentration increases, there is a sharp decrease the transient cell performance is poor compared in the size of the right arc, consistent with the to HV4 due to the large difference in open circuit expected decrease in concentration voltage loss (i.e. voltages (18). Figures 5(a) and 5(b) suggest the cathode mass transport). This is accompanied by a best cell performance is obtained with 0.3 M NaV4. subtle decrease in the left arc, most likely caused As before, there is a significant increase in limiting by an increase in the cathode exchange current. current and peak power density on increasing the However, there is also an increase in cell ohmic NaV4 concentration from 0.2 M to 0.3 M. However, a resistance which diminishes the improvements further increase in concentration has a detrimental

(a) (b) 1.0 0.8 0.2 M NaV4 0.9 0.7 0.3 M NaV4 0.8 0.4 M NaV4

–2 0.6 0.7 0.45 M NaV4 0.5 0.6 0.4 0.5

Cell voltage, V Cell voltage, 0.3 0.4 0.2 M NaV4

0.3 W cm density, Power 0.2 0.3 M NaV4 0.4 M NaV4 0.2 0.1 0.45 M NaV4 0.1 0 0 0.5 1.0 1.5 2.0 2.5 3.0 0 0.5 1.0 1.5 2.0 2.5 3.0 –2 Current density, A cm Current density, A cm–2 (c) 0.18 (d) 0.05 0.2 M NaV4

2 0.3 M NaV4 0.17 0.04 0.45 M NaV4 2 W cm 0.16 W cm 0.03

0.15 0.02 0.14 8.75 A 0.01 Area specific HFR,

0.13 15 A Area specific reactance,

25 A 0 0.12 0.2 0.3 0.4 0.5 0.12 0.16 0.20 0.24 0.28 2 Catholyte concentration, M Area specific resistance,W cm

Fig. 5. CRRC fuel cell performance at 80°C: (a) i–V; (b) corresponding power density curves generated with NaV4 catholyte at four different concentrations (labelled); (c) high frequency resistance obtained at three current loads for 0.2–0.45 M NaV4; (d) Nyquist plots obtained at 0.6 A cm–2 for 0.2 M, 0.3 M and 0.45 M NaV4

196 © 2018 Johnson Matthey http://dx.doi.org/10.1595/205651318X696800 Johnson Matthey Technol. Rev., 2018, 62, (2) effect on performance. The Nyquist plots in Figure level (38). Conversely, more vanadium(IV) will be 5(d) show the right-hand arc, associated with keggin bound (i.e. anionic) in reduced NaV4 than in catholyte mass transport, decreasing in size as reduced HV4 (38). catholyte concentration increases, as seen with Figures 6(a) and 6(b) illustrate the effect of HV4. However, unlike HV4, the cell ohmic resistance temperature on the performance of a CRRC fuel (Figure 5(c)) is lowest for the 0.3 M catholyte cell using 0.3 M HV4 catholyte (5% reduced). concentration, leading to the best observed At low current density, the i–V curves show the performance. Comparing Figures 4(c) and 5(c), positive effect of temperature on reaction kinetics, the increase in HFR from 0.2 M to 0.45 M catholyte with smaller activation losses as the temperature concentration is much larger for HV4 than for NaV4. increases. However, at higher current densities, One reason for this could be differing amounts of operation above 80°C has a detrimental effect free vanadium(IV), which could occupy membrane on performance. For the temperatures studied sites, displacing protons. Due to differences in pH, (40–90°C), the cell peak power density ranges from the concentration of VO2+ in reduced HV4 will be 692 mW cm–2 to 860 mW cm–2, corresponding to much larger than NaV4, for the same reduction a 24% power increase (from 40°C to 90°C). In a

(a) (b) 1.1 0.9 HV4 313 K 1.0 0.8 HV4 323 K 0.9 0.7 HV4 333 K 0.8 –2 HV4 343 K 0.6 0.7 HV4 353 K 0.5 0.6 HV4 363 K HV4 313 K 0.4 0.5 HV4 323 K Cell voltage, V Cell voltage, 0.3 0.4 HV4 333 K

Power density, W cm density, Power 0.2 HV4 343 K 0.3 HV4 353 K 0.2 0.1 HV4 363 K 0.1 0 0 0.5 1.0 1.5 2.0 2.5 3.0 0 0.5 1.0 1.5 2.0 2.5 3.0 Current density, A cm–2 Current density, A cm–2 (c) (d) 0.26 0.06 HV4 313 K HV4 343 K

0.24 2 0.05 HV4 353 K 2 0.22 W cm HV4 363 K 0.04 W cm 0.20 0.03 0.18

8.75 A 0.02 0.16 15 A Area specific HFR, 0.01 0.14 Area specific reactance, 25 A

0.12 0 40 50 60 70 80 90 0.15 0.20 0.25 0.30 0.35 0.40 Temperature, ºC Area specific resistance,W cm2

Fig. 6. (a) i–V; (b) corresponding power density curves generated with 0.3 M HV4 catholyte at a range of temperatures (labelled); (c) high frequency resistance obtained at three current loads over a range of temperatures for 0.3 M HV4; (d) Nyquist plots obtained at 0.6 A cm–2 over a range of temperatures for 0.3 M HV4

197 © 2018 Johnson Matthey http://dx.doi.org/10.1595/205651318X696800 Johnson Matthey Technol. Rev., 2018, 62, (2) study with a conventional PEFC and fully humidified 80°C. Although not shown, the left-hand arc

H2 and air, Song et al. found raising cell temperature has been found to increase when the anode is from 40°C to 100°C increased peak power from poisoned, suggesting anode kinetics can affect its ~500 mW cm–2 to ~800 mW cm–2, a 60% increase shape. In this case, Figure 6(d) would agree with (25). This suggests the CRRC PEFC in Figure 6 is other studies of high temperature conventional quite robust to temperature and cold starts would PEFCs, where increasing temperature was found not be as problematic as in conventional PEFCs. to decrease electrochemical Pt surface areas and The HFR results in Figure 6(c) show cell ohmic exchange current densities (25, 40, 41). Thus, resistance continuously decreases with increasing high temperature CRRC PEFC performance may be temperature, even at 90°C. Thus, although the fuel limited by anode kinetics. is dry hydrogen, the membrane drying observed The effect of temperature on NaV4 catholyte in other temperature studies with PEFCs is not performance was similar to that for HV4. The present here due to the aqueous cathode (39). power density curves for 0.3 M NaV4 (25% Rather, temperature has the expected positive reduced) over a temperature range of 40–90°C effect on ohmic resistance via increasing the are shown in Figure 7(a), along with HFR results ionic conductivity of the electrolytes. The Nyquist in Figure 7(b). As with HV4, 80°C gave the best plots in Figure 6(d) provide insights into the performance and cell ohmic resistance decreased performance trends in Figures 6(a) and 6(b). over the whole temperature range. Thus, no As observed, increasing temperature from 40°C membrane drying was observed and the reduction to 70°C reduces all three sources of voltage loss: in performance on moving to 90°C operation can the right-hand arc (catholyte mass transport) be tentatively attributed to a decrease in the total and left-hand arc (associated with kinetics) are area of the Pt | electrolyte | hydrogen 3-phase both dramatically reduced in size and cell ohmic boundary on the anode. resistance also decreases. On moving from 70°C to 80°C, the two arcs remain similar in shape and 3.4 Regenerator Performance size, the main difference being the shift to a lower HFR, hence lower overall cell resistance. However, Following the method of Gunn et al. (16), on increasing to 90°C, there is a significant growth regeneration sweeps were performed with in the left-hand arc, which more than offsets the reduced catholytes over the range of conditions decrease in HFR. Thus, the overall cell resistance in Table II (see the Supplementary Information increases and the cell performs worse than at for more details). Using data generated from

(a) (b) 0.24 0.6 0.22

0.5 2

–2 0.20 W cm 0.4 0.18

0.3 NaV4 313 K 0.16 NaV4 323 K 0.2 NaV4 333 K 0.14 8.75 A Power density, W cm density, Power Area specific HFR, NaV4 343 K 15 A 0.1 0.12 NaV4 353 K 25 A NaV4 363 K 0 0.10 0 0.5 1.0 1.5 2.0 2.5 40 50 60 70 80 90 Current density, A cm–2 Temperature, ºC

Fig. 7. (a) Current density-power density curves generated with 0.3 M NaV4 catholyte at a range of temperatures (labelled); (b) high frequency resistance obtained at three current loads over a range of temperatures for 0.3 M NaV4

198 © 2018 Johnson Matthey http://dx.doi.org/10.1595/205651318X696800 Johnson Matthey Technol. Rev., 2018, 62, (2) these experiments, the rate of the regeneration temperature appears to benefit catholyte reaction (i.e. the rate of vanadium(IV) oxidation) regeneration apart from the fastest reactions (i.e. can be expressed as a ‘regeneration current’ currents greater than ~35 A) where increasing to (16). Figure 8 illustrates regeneration currents 90°C is detrimental. evaluated at different reduction levels for HV4 and Zhizhina et al. studied the regeneration of 0.2 M

NaV4 over the concentration/temperature matrix H7PV4Mo8O40 at various levels of reduction over investigated. As observed previously, regeneration a range of temperatures (40–90°C) in 1 atm O2 currents for NaV4 noticeably exceed those for HV4 (37). The researchers used a basic shaking method (at the same conditions) due to the difference in to mix gas and liquid. The maximum volumetric pH of the two catholytes (18). For both catholytes, regeneration current they measured was concentration has a relatively minor effect on the approximately 400 A l–1, recorded at 90°C for 68% rate of the regeneration reaction, with varying reduced catholyte. This corresponds to ~80 A l–1 direction. In general, for high levels of reduction, for the same reaction in air. In the CRRC system, increasing catholyte concentration has a positive 65% reduced 0.2 M HV4 produced a regeneration effect on the regeneration current, whereas the current of ~30 A. Given the volume of catholyte in opposite occurs for low reduction levels. Increasing the regenerator is ~250 ml, this corresponds to a

(a) (b) 45 60

40 55 35 IV 70% V 50 30 65% VIV 25 45 60% VIV 20 40 55% VIV 15 35 50% VIV

Regeneration current, A Regeneration 10 current, A Regeneration IV 30 5 45% V

0 25 0.2 0.3 0.4 0.5 0.2 0.3 0.4 0.5 Catholyte concentration, M Catholyte concentration, M

40 50 35 IV 70% V 45 30 65% VIV 25 40 60% VIV 20 35 55% VIV 15 30 50% VIV

Regeneration current, A Regeneration 10 current, A Regeneration IV 25 5 45% V

0 20 40 50 60 70 80 90 40 50 60 70 80 90 Temperature, ºC Temperature, ºC

Fig. 8. Regeneration currents for a range of catholyte reduction levels plotted against concentration for: (a) HV4; (b) NaV4 at 80°C. Regeneration currents for a range of catholyte reduction levels plotted against temperature for: (c) 0.3 M HV4; (d) 0.3 M NaV4

199 © 2018 Johnson Matthey http://dx.doi.org/10.1595/205651318X696800 Johnson Matthey Technol. Rev., 2018, 62, (2) volumetric current of ~120 A l–1, which is similar PEFC systems is often not reported. It corresponds in magnitude to that recorded by Zhizhina et al., to the cell operating voltage when the regeneration the increase likely caused by the better liquid-gas current equals the cell current. In this case, the mixing (37). For reduction levels greater than 50%, redox potential of the catholyte entering the fuel Zhizhina et al. found the rate of regeneration, w, cell does not change, resulting in a constant cell could be described by Equation (vii), where k is an operating voltage (providing the cell is durable). apparent rate constant, [VIV] is the concentration Figure 9 illustrates the measured steady state cell of vanadium(IV) in the reduced catholyte, [H+] voltage at 1 A cm–2 for both HV4 and NaV4 catholytes is the concentration of protons, pO2 is the partial at all the conditions in Table II. The highest pressure of oxygen, R is the universal gas constant steady state cell voltage of 0.47 V is achieved by and T is temperature (37). 0.45 M NaV4 at 90°C. Reducing the concentration 10 and temperature to 0.3 M and 80°C, respectively, – IV 2.8 + –2.5 RT w = k[V ] [H ] pO2e (vii) only results in a 10 mV drop in cell voltage, which is tiny considering the difference in cost between Equation (vii) suggests increasing the 0.3 M and 0.45 M catholyte. Thus, 0.3 M and 80°C temperature and concentrations of oxygen and appear to be the optimum parameters for NaV4. reduced catholyte all have a positive effect on the The same is true for HV4. However, in this case rate of regeneration, as expected. The negative temperature has a greater effect on performance, dependence on proton concentration is less with much larger variations in cell voltage observed obvious and arises from POM-speciation effects. across the matrix of conditions. The reason for this Only keggin-bound vanadium(IV) can reduce is the relatively low regeneration currents achieved molecular oxygen. Decreasing the pH favours the with HV4, making the system more sensitive to formation of VO2+ (38) and thus, results in slower temperature. Thus, 0.3 M NaV4 produces better regeneration kinetics. Equation (vii) also explains steady state performance than 0.3 M HV4 and is the trends in Figure 8. Increasing the catholyte more robust to changes in temperature. concentration for a given reduction level increases Using the method of Gunn et al. (16), the steady both vanadium(IV) and proton concentration (see state peak power densities were estimated from Figure 3(e)). In addition, increasing the reduction the transient cell performance and regeneration level increases catholyte pH. So, the positive results for each catholyte in Table II. The effect of increasing catholyte concentration is only results are summarised in Figure 10 and show observed at the highest reduction levels. Regarding NaV4 outperforms HV4 at all conditions. For both temperature, the trends in Figures 8(c) and 8(d) catholytes at 0.3 M concentration, increasing agree with the exponential term in Equation (vii), temperature improves steady state peak power apart from some of the results at 90°C. The latter up to 80°C, after which performance diminishes is caused by the effect of temperature on oxygen (Figure 10(a)). Likewise, Figure 10(b) suggests solubility. When the kinetics are fast, the reaction 0.3 M is the optimum temperature for both NaV4 can be limited by the mass transport of dissolved and HV4 catholytes. oxygen. In this case, increasing temperature may have a negative effect on the regeneration rate 4. Conclusions by decreasing oxygen solubility. This appears to be the case for regeneration currents over 35 A, Although there are subtle differences in how corresponding to 140 A l–1, at 90°C. Note, this temperature and concentration affect the negative effect of temperature was not observed regenerator and cell performance of HV4 and NaV4 by Zhizhina et al. because their (comparative) catholytes, in terms of the overall CRRC system, volumetric currents were significantly lower than the results are identical: 0.3 M and 80°C result in 140 A l–1 (37). the optimum steady state performance for both HV4 and NaV4 catholytes (at ambient pressure). 3.5 Steady State Performance No meaningful benefit in system performance can be obtained from further increasing catholyte For many applications, fuel cells are required to concentration. Likewise, increasing the temperature perform at a given power for a prolonged time. In beyond 80°C appears to be detrimental for both this case, transient cell performance results, like HV4 and NaV4. Surprisingly, the CRRC PEFCs could those in Figures 4–7, are misleading for CRRC operate with dry hydrogen at 90°C and ambient PEFCs. The true steady state performance of CRRC pressure with no evidence of membrane drying.

(a) 0.2 M HV4 0.3 M HV4 0.4 M HV4 0.45 M HV4

0.43 0.41 0.42 0.50 0.42 0.40 0.39 0.45 0.33 0.40 0.33 0.33 0.35 0.26 0.30 0.15

0.25 0.16 0.20 Cell voltage, V Cell voltage, 0.15 0.45 0.10 0.4 0.03 0.05 0.3 0.2 0

90 80 70 60 50 40 Concentration, M Temperature, ºC

(b) 0.2 M NaV4 0.3 M NaV4 0.4 M NaV4 0.45 M NaV4

0.47 0.46 0.46 0.46 0.44 0.45 0.50 0.42 0.33 0.43 0.42 0.45 0.38

0.40 0.33

0.35

0.30

0.25

0.20 0.17 Cell voltage, V Cell voltage, 0.15 0.45 0.10 0.4 0.05 0.3 0.2 0

90 80 70 60 50 40 Concentration, M Temperature, ºC

Fig. 9. Measured steady state cell operating voltage at 1 A cm–2 for a CRRC fuel cell using: (a) HV4; (b) NaV4 catholytes at a range of concentrations and temperatures

(a) (b) 0.55 0.55

0.50 0.50 –2 –2 0.45

0.45 0.40

0.35 0.40 NaV4 0.30 0.3 M NaV4 0.35 HV4

Peak power density, W cm power density, Peak 0.25 W cm power density, Peak 0.3 M HV4

0.20 0.30 40 50 60 70 80 90 0.2 0.3 0.4 0.5 Temperature, ºC Catholyte concentration, M

Fig. 10. Estimated steady state peak power densities for a CRRC fuel cell using: (a) 0.3 M HV4 and 0.3 M NaV4 over a range of temperatures; (b) various concentrations of HV4 and NaV4 catholytes at 80°C

However, the step from 80°C to 90°C operation 4. “Fuel Cells: Data, Facts and Figures”, eds. D. was detrimental for both cell and regenerator Stolten, R. C. Samsun and N. Garland, Wiley-VCH performance. The latter was attributed to oxygen Verlag GmbH & Co KGaA, Weinheim, Germany, solubility issues whilst the former was linked to 2016, 408 pp anode kinetics. Thus, high temperature operation 5. ‘Germany: H2 MOBILITY Targets 400 Hydrogen may yield benefits at pressures above 1 atm. This Fueling Stations by 2023’, Hydrogen Mobility will be explored in future work. Europe, Fuel Cells and Hydrogen Joint Undertaking, Brussels, Belgium, 5th May, 2016 6. “Fuel Cell Technical Team Roadmap”, U.S. DRIVE, Acknowledgements Office of Energy Efficiency and Renewable Energy, Washington, DC, USA, June, 2013 This study was part funded by the Higher Education Funding Council for England (HEFCE) Innovation 7. F. T. Wagner, B. Lakshmanan and M. F. Mathias, J. Fund. The authors also thank Dr Corinne Wills Phys. Chem. Lett., 2010, 1, (14), 2204 (Newcastle University, UK) for the NMR analysis 8. O. T. Holton and J. W. Stevenson, Platinum Metals and Dr Natasha Gunn (University of Chester), Dr Rev., 2013, 57, (4), 259 Matthew Herbert (University of Chester), Joshua 9. H. A. Gasteiger, S. S. Kocha, B. Sompalli and F. T. Denne (Advanced Propulsion Centre, UK) and Wagner, Appl. Catal. B: Environ., 2005, 56, (1–2), 9 Nadine Uwigena (University of Chester) for their 10. F. D. Coms, ECS Trans., 2008, 16, (2), 235 help and advice. 11. E. Endoh, S. Terazono, H. Widjaja and Y. Takimoto, Electrochem. Solid-State Lett., 2004, 7, (7), A209 12. C. A. Reiser, L. Bregoli, T. W. Patterson, J. S. Yi, J. References D. Yang, M. L. Perry and T. D. Jarvi, Electrochem. Solid-State Lett., 2005, 8, (6), A273 1. H. A. Gasteiger and N. M. Marković, Science, 2009, 324, (5923), 48 13. E. Brightman and G. Hinds, J. Power Sources, 2014, 267, 160 2. M. F. Mathias, R. Makharia, H. A. Gasteiger, J. J. Conley, T. J. Fuller, C. J. Gittleman, S. S. Kocha, 14. Yu. V. Tolmachev and M. A. Vorotyntsev, Russ. J. D. P. Miller, C. K. Mittelsteadt, T. Xie, S. G. Yan Electrochem., 2014, 50, (5), 403 and P. T. Yu, Electrochem. Soc. Interface, 2005, 15. A. M. Posner, Fuel, 1955, 34, 330 14, (3), 24 16. N. L. O. Gunn, D. B. Ward, C. Menelaou, M. A. 3. T. Yoshida and K. Kojima, Electrochem. Soc. Herbert and T. J. Davies, J. Power Sources, 2017, Interface, 2015, 24, (2), 45 348, 107

17. R. Singh, A. A. Shah, A. Potter, B. Clarkson, A. 28. N. Martin and M. Herbert, ACAL Energy Ltd, Creeth, C. Downs and F. C. Walsh, J. Power ‘Synthesis of Polyoxometalates’, World Patent Sources, 2012, 201, 159 Appl. 2015/097,459 18. D. B. Ward, N. L. O. Gunn, N. Uwigena and T. J. 29. L. Pettersson, Mol. Eng., 1993, 3, (1–3), 29 Davies, J. Power Sources, 2018, 375, 68 30. L. Pettersson, I. Andersson, J. H. Grate and A. 19. S.-B. Han, D.-H. Kwak, H. S. Park, I.-A. Choi, J.-Y. Selling, Inorg. Chem., 1994, 33, (5), 982 Park, K.-B. Ma, J.-E. Won, D.-H. Kim, S.-J. Kim, 31. A. Selling, I. Andersson, J. H. Grate and L. M.-C. Kim and K.-W. Park, ACS Catal., 2016, 6, Pettersson, Eur. J. Inorg. Chem., 2000, (7), 1509 (8), 5302 32. I. V. Kozhevnikov, Chem. Rev., 1998, 98, (1), 171 20. S.-B. Han, D.-H. Kwak, H. S. Park, I.-A. Choi, 33. P. Souchay, F. Chauveau and P. Courtin, Bull. Soc. J.-Y. Park, S.-J. Kim, M.-C. Kim, S. Hong and K. Chim. France, 1968, (6), 2384 W. Park, Angew. Chem. Int. Ed., 2017, 56, (11), 2893 34. I. V. Kozhevnikov, Izv. Akad. Nauk SSSR: Ser. Khim., 1983, 4, 721; translated into English in 21. ‘3.4: Fuel Cells, 2016’, in “Fuel Cell Technologies Russ. Chem. Bull., 1983, 32, (4), 655 Office Multi-Year Research, Development, and Demonstration Plan”, Office of Energy Efficiency 35. V. M. Berdnikov, L. I. Kuznetsova, K. I. Matveev, and Renewable Energy, Washington, DC, USA, N. P. Kirik and E. N. Yurchenko, Koord. Khim., May, 2017 1979, 5, (1), 78 22. ‘ACAL Energy Fuel Cell Achieves 10,000 Hour 36. I. V. Kozhevnikov, Yu. V. Burov and K. I. Matveev, Endurance’, Fuel Cell Today, Royston, Hertfordshire, Izv. Akad. Nauk SSSR: Ser. Khim., 1981, 11, UK, 27th June, 2013 2428; translated into English in Russ. Chem. Bull., 23. V. F. Odyakov, E. G. Zhizhina and K. I. Matveev, J. 1981, 30, (11), 2001 Mol. Catal. A: Chem., 2000, 158, (1), 453 37. E. G. Zhizhina, V. F. Odyakov, M. V. Simonova and 24. T. Matsui, E. Morikawa, S. Nakada, T. Okanishi, H. K. I. Matveev, Kinet. Catal., 2005, 46, (3), 354 Muroyama, Y. Hirao, T. Takahashi and K. Eguchi, 38. A. Selling, I. Andersson, J. H. Grate and L. ACS Appl. Mater. Interfaces, 2016, 8, (28), 18119 Pettersson, Eur. J. Inorg. Chem., 2002, (3), 743 25. C. Song, Y. Tang, J. L. Zhang, J. Zhang, H. 39. L. Wang, A. Husar, T. Zhou and H. Liu, Int. J. Wang, J. Shen, S. McDermid, J. Li and P. Kozak, Hydrogen Energy, 2003, 28, (11), 1263 Electrochim. Acta, 2007, 52, (7), 2552 40. J. Zhang, Y. Tang, C. Song, J. Zhang and H. Wang, 26. C. Zhang, T. S. Zhao, Q. Xu, L. An and G. Zhao, J. Power Sources, 2006, 163, (1), 532 Appl. Energy, 2015, 155, 349 41. Y. Song, J. M. Fenton, H. R. Kunz, L. J. Bonville and 27. F. A. de Bruijn, R. C. Makkus, R. K. A. M. Mallant M. V. Williams, J. Electrochem. Soc., 2005, 152, and G. J. M. Janssen, Adv. Fuel Cells, 2007, 1, 235 (3), A539

The Authors

David B. Ward is a Leading Research Fellow at the University of Chester, UK. His interests lie in the study and development of environmentally considerate engineering technologies and practices. David graduated in 2001 with a PhD from the Department of Chemical and Process Engineering, University of Sheffield, UK. He spent several years working as a Senior Engineer for a research and development (R&D) start-up, ACAL Energy Ltd, developing a high efficiency gas liquid oxidation reactor for a novel hydrogen proton exchange membrane (PEM) fuel cell engine. David continues to advance this work with respect to both fuel cells and water treatment applications.

Trevor J. Davies was a Senior Lecturer at the University of Chester and led the Electrochemistry Research Group from 2014 to 2018. His research interests include electrolysers, fuel cells, flow batteries and sensors. Trevor completed his PhD in Chemistry in 2005 at the University of Oxford, UK. Before joining Chester in 2014, he was a fuels scientist at Shell, UK, and subsequently led the cathode development programme at ACAL Energy Ltd, UK. Trevor recently left the University of Chester to join INOVYN’s Electrochemical Technology Business.

www.technology.matthey.com

21st International Conference on Solid State Ionics Highlights of the latest developments in solid-state batteries for energy storage

By Thomas Bartlett and James Solid-State Electrolytes Cookson* Johnson Matthey, Blounts Court, Sonning The garnet-type group of solid-state electrolytes Common, Reading, RG4 9NH, UK (SSEs) were a focus of numerous talks with

two sessions dedicated to the Li7La3Zr2O12 *Email: [email protected] (LLZO) ceramic. S. Uhlenbruck (Jülich Research Centre, Germany) opened one of the sessions with an overview of the recent developments in the field. He mentioned that the discrepancy in Introduction many reported conductivity results could in some part be explained by preventable sample aging. The 21st International Conference on Solid State LLZO is known to react with water and carbon Ionics (SSI-21) was held in Padova, Italy, from dioxide to form lithium carbonate and Uhlenbruck 18th to 23rd June, 2017. The conference saw showed that samples stored in an argon box were ~1300 people attend over the six days, covering still affected by the ppm levels of water present, four macro areas: leading to carbonate formation and depletion of • energy and environment lithium at the electrolyte surface. Uhlenbruck also • communication and robotics showed that the cycle life of LLZO-based SSB • biological systems and life sciences could be enhanced when a gold nano-layer was • fundamental theory. deposited at the electrolyte/Li-metal interface. The energy and environment macro area saw The development of LLZO electrolytes was 30 topics including: beyond lithium-ion batteries, discussed further by A. Aguadero (Imperial advanced lithium and sodium batteries, solid-state College London, UK) with her work on aluminium/ batteries (SSB), redox-flow batteries, polymeric gallium doped LLZO; these elements are known to batteries, solid oxide fuel cells and many others. stabilise the favourable, highly Li-ion conducting The attendance was truly international with a cubic phase. Focused ion beam secondary ion mass significant proportion of attendees travelling spectrometry (FIB-SIMS) was used to locate the from academic institutions outside Europe. A few position of the dopant within the LLZO electrolyte, companies were also in attendance, predominantly with Al found concentrated at the particle grain from the USA and Japan. boundaries. Aguadero suggested the possibility This review will focus on the recent advancements of different dendrite formation mechanisms for Al presented in the field of SSB. This topic was the and Ga doped samples when used with Li-metal largest of those on offer within the energy and anodes. She linked this difference in mechanism environment macro area and was well attended to the inhomogeneity of cation distribution leading over the six days. The talks covered herein have to uneven polarisation in the electrolyte. The been grouped by topic and include: mechanism of lithium dendrite growth through • solid-state electrolytes solid electrolytes, in particular LLZO, was further • analytical techniques for SSB discussed by B. Sheldon (Brown University, • all-solid-state Li-ion batteries (ASSLiB). USA). He argued that dendritic growth does not

204 © 2018 Johnson Matthey https://doi.org/10.1595/205651318X696774 Johnson Matthey Technol. Rev., 2018, 62, (2) necessarily require grain boundary pathways to The lithium sits in pseudo-octahedral sites causing propagate. Instead, he proposed that dendritic polymerisation of the C60 as shown in Figure 1. Due growth is in fact caused by surface imperfections. to the high electrical conductivity of the polymer, Large surface flaws allow Li penetration beyond the its use as an electrolyte for all-solid-state Li-ion crack into the bulk crystal, driven by strain energy batteries was limited, however Pontiroli et al. release at the dendrite tip. He is continuing his showed that the ionic and electronic conductivities work towards a general model for dendrite growth may originate from different causes. This suggests in SSEs with more results soon to be published. that it could be possible to shut off the electrical In addition to garnet-based electrolytes, sulfide conductivity whilst maintaining the high Li-ion electrolytes featured heavily throughout the week. conductivity of the polymer. A notable speaker was R. Kanno (Tokyo Institute A nano-composite electrolyte comprised of an of Technology, Japan), one of the initiators of insulating matrix of silica particles and lithium research into highly conductive sulfide based salts was presented by P. M. Vereecken (University

Li10GeP2S12 (LGPS) electrolytes. Kanno presented of Leuven, Belgium). These systems work by the his recent work on overcoming the electrochemical selective adsorption of the salts at the insulating stability limitations of these SSEs through element particle surface, leading to an accumulation of substitution. Although substitution of silicon with vacancies at the interface and higher conductivities. oxygen was found to decrease the conductivity The ionic liquid and lithium salt (ILE) system performance, higher electrochemical stability was discussed was shown to reach conductivities also indicated. Other sulfide systems were tested approaching 1 mS cm–1, rivalling conventional by Kanno with the metal removed; the idea being liquid organic electrolytes. that the removal of the metallic elements would increase their stability towards Li-metal anodes. Analytical Techniques for Solid-State Both showed good coulombic efficiency so were Batteries believed to be more electrochemically stable than their germanium containing counterparts. A number of new techniques aimed at probing the Alternatives to garnet and sulfide based electrolytes structure/performance relationship in both SSEs were also discussed in the SSE sessions. One such and SSB were presented across the week. S. Taibl electrolyte was the Li salt/C60 polymeric electrolyte (Vienna University of Technology, Austria) presented presented by D. Pontiroli (Università di Parma, Italy). her work on the spatially resolved electrochemical

(a) (b)

Fig. 1. (a) Structure of Li4C60 as obtained from geometry optimisation, in the space group I2/m. Grey sticks are carbon bonds, magenta and brown spheres are carbon atoms involved in the covalent intermolecular bonds: [2+2] bridges and single bonds respectively. Red and yellow spheres represent LiT and LiO respectively; (b) view along the c axis of one polymeric plane (the Li ions have been omitted for clarity) (Reprinted with permission from (1) Copyright 2015 by the American Physical Society)

205 © 2018 Johnson Matthey https://doi.org/10.1595/205651318X696774 Johnson Matthey Technol. Rev., 2018, 62, (2) characterisation of SSEs. Conventional impedance material, with the Li distribution during charge/ spectroscopy uses macro-electrodes to probe discharge. Advances in probing the structure diffusion within the bulk of the SSE. Here, Taibl et of more conventional electrode systems were al. used individually accessible micro-electrodes, discussed by J. Joos (Institute for Applied Materials deposited by lithography, to electrochemically (IAM-WET), Karlsruhe Institute of Technology map the conductivity of the sample surface. This (KIT), Germany). His talk gave an update on their electrochemical map was combined with elemental model development for correlating bulk X-ray mapping obtained through laser ablation inductive tomography with TEM three-dimensional (3D) laser coupled plasma mass spectrometry (LA-ICP-MS) ablation; the concept behind this model is to use to correlate the Al and lanthanum surface the high resolution data from small areas of sample concentration with Li-ion conductivity. analysed by TEM 3D laser ablation and apply this to A look at combinatory techniques to probe SSB the comparatively fast and non-destructive X-ray during operating was delivered by K. Yamamoto tomography of the bulk sample. Joos et al. showed (Japan Fine Ceramics Center, Japan). The talk how such a method could allow the voids within the highlighted his recent work on the study of SSB electrode ink to be calculated, and hence inform on by transmission electron microscopy (TEM). The the particle distribution required to minimise void systems presented were interesting in that the space and maximise performance. anode is formed in situ from the SSE. The SSE is reduced to form a Li2O-Al2O3-TiO2-P2O5-based All-Solid-State Lithium-Ion Batteries electrode (2) and was followed by ex situ spatially resolved electron emission loss spectroscopy E. Wachsman (University of Maryland, USA) opened (SR-EELS) to resolve the electronic structural the first ASSLiB session, presenting an overview of changes occurring. This was correlated with in situ different techniques for improving the performance electron holography (EH) experiments to correlate of garnet-based ASSLiBs. An Al-nanolayer coating the electronic structural changes occurring to the at the Li-metal anode/SSE interface (Figure 2) Ti/O in the lithium titanium aluminium phosphate was shown to permit wetting of the SSE by the

(a) (b) (c)

(d) (e) (f)

Fig. 2. Wetting behaviour and interfacial morphology characterisation of Li/garnet SSE and Li/Al-coated garnet SSE: (a) wetting behaviour of molten Li with garnet SSE and Al-coated garnet SSE. The inset is a schematic showing the contact angles of a molten Li droplet wetting the surface of both uncoated and Al- coated garnet SSEs. Improved Li wettability is demonstrated after Al-coating the garnet surface; (b) and (c) SEM images of Li/garnet SSE, showing the poor Li wettability of uncoated garnet; (d) to (f) SEM images of Li/Al-garnet SSE-Al exhibiting superior Li wettability with Al-coated garnet (Reprinted from (3) © The Authors, some rights reserved; exclusive licensee American Association for the Advancement of Science. Distributed under a Creative Commons Attribution Non Commercial License 4.0 (CC BY-NC))

Li-metal and reduce the contact resistance Conclusions significantly. Wachsman also presented a method for fabricating flexible solid-state conductive The conference was well attended by the membranes; this consisted of a polymer matrix international community covering all aspects and 3D ceramic network formed by electrospinning of solid-state ionics. Within the battery macro- a garnet salt and polyvinylpyrrolidone (PVP) area, particular emphasis was put on research polymer. This polymer-ceramic composite into SSB. All aspects of academic research in possessed a conductivity of 2.5 × 10–4 S cm–1 the field were represented including electrolyte and supressed dendrite formation over 300 h at development, interfacial studies and routes for 0.5 mA cm–2 when fabricated with a Li-anode. ASSLiB fabrication. The presentations were of high These types of flexible polymer-ceramics were quality and audience interaction with the speakers shown to be compatible with scalable tape-casting was both frequent and impactful. processes and were reported by Wachsman to achieve current densities of up to 3 mA cm–2 . The final ASSLiB session was opened by E. References Ivers-Tiffée (KIT, Germany). She presented a 1. S. Rols, D. Pontiroli, C. Cavallari, M. Gaboardi, M. modelling approach to predict the structure Aramini, D. Richard, M. R. Johnson, J. M. Zanotti, required for LGPS-based batteries to compete with E. Suard, M. Maccarini and M. Riccó, Phys. Rev. B: conventional Li-ion batteries. The model indicated Condens. Matter Mater. Phys., 2015, 92, (1), 14305 that it was possible to match Li-ion in terms of power 2. Y. Iriyama, C. Yada, T. Abe, Z. Ogumi and K. and energy density for bulk-type ASSLiB systems, Kikuchi, Electrochem. Commun., 2006, 8, (8), with an electrolyte thickness of 1 μm. Ivers-Tiffée 1287 discussed how this model could be applied to other 3. K. Fu, Y. Gong, B. Liu, Y. Zhu, S. Xu, Y. Yao, W. systems, such as garnet-based ASSLiBs, to help Luo, C. Wang, S. D. Lacey, J. Dai, Y. Chen, Y. Mo, direct battery design and achieve systems that can E. Wachsman and L. Hu, Sci. Adv., 2017, 3, (4), compete with conventional Li-ion technologies. e1601659

The Reviewer

Thomas Bartlett was a Research Scientist at Johnson Matthey’s Technology Centre, Sonning Common, UK. He obtained an undergraduate degree with Master’s in Chemistry from the University of Bath, UK, and a doctorate in Nano- Electrochemistry from Oxford University, UK. He worked in the Battery Materials Team on a range of research projects.

James Cookson is a Research Manager at Johnson Matthey’s Technology Centre, Sonning Common, UK. He obtained a DPhil in Inorganic Chemistry at the University of Oxford, UK. James leads Johnson Matthey’s research activities in battery technologies; exploring both Li-ion as well as next generation technologies.

www.technology.matthey.com

Johnson Matthey Highlights A selection of recent publications by Johnson Matthey R&D staff and collaborators

Hierarchical Three-Dimensional Fe3O4@Porous product and the naturally occurring azaspiracid-3 Carbon Matrix/Graphene Anodes for High but there were chromatographic and spectroscopic Performance Lithium Ion Batteries differences between the two structures. S. Hao, B. Zhang, Y. Wang, C. Li, J. Feng, S. Ball, High Performance Mixed Matrix Membranes M. Srinivasan, J. Wu and Y. Huang, Electrochim. (MMMs) Composed of ZIF-94 Filler and 6FDA-DAM Acta, 2018, 260, 965 Polymer A facile annealing of Mil-53(Fe) templates was M. Etxeberria-Benavides, O. David, T. Johnson, used to synthesise a hierarchical 3D Fe3O4@ M. M. Łozińska, A. Orsi, P. A. Wright, S. Mastel, porous carbon matrix (PCM) and Fe3O4@PCM/ R. Hillenbrand, F. Kapteijn and J. Gascon, J. graphene and their electrochemical performances Membrane Sci., 2018, 550, 198 were tested as anode materials for Li-ion High performance mixed matrix membranes batteries. The Fe3O4@PCM/graphene electrode has a much better cycling performance with a (MMM) comprising ZIF-94 filler and 6FDA-DAM high reversible capacity: up to 1000 mAh g–1 after polymer matrix were developed. Mixed gas tests (15CO2:85N2) at 25ºC and 1–4 bar transmembrane 500 cycles at 1 C, compared with Fe3O4@PCM. This enhanced performance can be attributed to pressure difference were carried out to assess the the high conductivity and structure stability given CO2/N2 separation performance. Adding ZIF-94 by the special graphene based hierarchical 3D increased the CO2 membrane permeability and nanostructure. maintained a constant CO2/N2 selectivity of ~22. At 40 wt% ZIF-94 loading the biggest increase

Total Synthesis of (6 R,10 R,13 R,14 R,16 R,17 in CO2 permeability was observed, reaching the R,19 S,20 R,21 R,24 S,25 S,28 S,30 S,32 R,33 highest permeability at an equivalent selectivity R,34 R,36 S,37 S,39 R)-Azaspiracid-3 Reveals to 6FDA-DAM MMM in the literature. For the first Non-Identity with the Natural Product time non-hazardous solvents (THF and methanol) N. T. Kenton, D. Adu-Ampratwum, A. A. Okumu, instead of DMF were used to synthesise ZIF-94 Z. Zhang, Y. Chen, S. Nguyen, J. Xu, Y. Ding, P. MOF crystals with particle size smaller than 500 nm McCarron, J. Kilcoyne, F. Rise, A. L. Wilkins, C. O. in a scalable process. SEM, AFM and nanoscale IR Miles and C. J. Forsyth, Angew. Chem. Int. Ed., imaging by s-SNOM were used to characterise the 2018, 57, (3), 805 membranes.

A late stage Nozaki-Hiyama-Kishi coupling Origin of Phase Inhomogeneity in Lithium Iron was used to form the C21–C22 bond with the Phosphate during Carbon Coating C20 configuration unambiguously created Y. Liu, J. Wang, J. Liu, M. N. Banis, B. Xiao, A. from L–(+)-tartaric acid in a convergent and Lushington, W. Xiao, R. Li, T.-K. Sham, G. Liang stereoselective total synthesis of an assigned and X. Sun, Nano Energy, 2018, 45, 52 structure of azaspiracid-3. Postcoupling steps included oxidation to an ynone, an amended A non-conductive Fe2P2O7 phase was found to Stryker reduction of the alkyne, global deprotection be formed on LiFePO4 during the carbon coating and oxidation of the ensuing C1 primary alcohol process. This phase formation depends on particle to the carboxylic acid. Mass spectrometry showed size, temperature and annealing atmosphere. The that there was a match between the synthetic changes were directly linked to the change of the

reducing potential. Improved understanding of the Triphos-Ru(CO)H2, 1, a stable and accessible Ru parameters required during carbon coating will help dihydride complex, has been found to catalyse to control the phase purity of carbon coated LiFePO4 aldehyde hydrogenation under neutral conditions. It and achieve better electrochemical performance. shows high activity and can be used without solvent. It has good activity at low catalyst loadings for the The Role of Catalyst Support, Diluent and Co- Catalyst in Chromium-Mediated Heterogeneous reductive amination of aldehydes under mildly acidic Ethylene Trimerisation conditions. Challenging chemoselectivity examples are presented where C-halogen groups, alkene or M. J. Lamb, D. C. Apperley, M. J. Watson and P. W. ketone functionality are not reduced. Improved Dyer, Top. Catal., 2018, in press chemoselectivity and activity are obtained by Initiator systems for the oligomerisation or using the pre-formed complex, 1, compared to in polymerisation of ethylene were made by situ formed catalysts from Triphos and Ru(acac)3 consecutive treatment of an already calcined solid especially at low reaction temperatures. oxide support (SiO2, γ-Al2O3 or mixed SiO2-Al2O3) with solutions of Cr{N(SiMe3)2}3 (0.71 wt% Cr) and Microstructural Analysis and Transport Resistances a Lewis acidic alkyl Al-based co-catalyst (15 molar of Low-Platinum-Loaded PEFC Electrodes equivalents). The impact of the oxide support, F. C. Cetinbas, X. Wang, R. K. Ahluwalia, N. N. calcination temperature, co-catalyst and reaction Kariuki, R. P. Winarski, Z. Yang, J. Sharman and diluent on both the productivity and selectivity of D. J. Myers, J. Electrochem. Soc., 2017, 164, (14), the immobilised Cr initiator systems were studied. F1596 The top performing combination (SiO2–600, modified methyl aluminoxane-12 {MMAO-12}, heptane) Microstructural characterisation and polarisation created a mixture of hexenes (61 wt%, 79% data analysis were carried out for PEFC cathodes 1-hexene) and polyethylene (16 wt%) with an with low pgm loadings. 3D pore morphology and –1 –1 ionomer distribution were resolved using nano- activity of 2403 gCr h . Two competing processes are proposed to explain the results: trimerisation CT. A model accounts for energy, charge and via a supported metallacycle-based mechanism and mass transport and the effect of liquid water polymerisation through a classical Cossee-Arlman flooding. Flooding in the electrode was shown chain-growth pathway. to contribute significantly to transport losses especially at high operating pressures while the Using a Freeman FT4 Rheometer and Electrical pressure-independent resistance at the catalyst Capacitance Tomography to Assess Powder Blending surface due to transport through the ionomer film G. Forte, P. J. Clark, Z. Yan, E. H. Stitt and M. is significant at low temperatures and low catalyst Marigo, Powder Technol., 2018, in press, corrected loading. The importance of electrode roughness proof (electrochemically-active surface area/geometric A Freeman FT4 powder rheometer and electrical electrode area) in determining the mass transport capacitance tomography (ECT) were used to losses is also highlighted. measure the influence of segregation or mixing on flow properties as well as powder mixing and CO Oxidation and Site Speciation for Alloyed Palladium–Platinum Model Catalysts Studied by in mixedness. Two powders with different properties Situ FTIR Spectroscopy such as particle size, density, basic flowability and electrical permittivity were used in two different N. M. Martin, M. Skoglundh, G. Smedler, A. Raj, initial arrangements: (a) a heavier, smaller powder D. Thompsett, P. Velin, F. J. Martinez-Casado, Z. at the top which would be expected to mix readily; Matej, O. Balmes and P.-A. Carlsson, J. Phys. Chem. (b) the inverse which would be expected to resist C, 2017, 121, (47), 26321 the axial blending mechanism in the FT4. During 30 Transient CO oxidation over a series of bimetallic cycles of the FT4 impeller passing into and back out Pd-Pt catalysts with different Pd:Pt molar ratios was of the powder layer, the torque and thus flow energy studied using in situ FTIR. The catalysts contained were tracked. Simultaneous ECT measurements both alloyed PdPt nanoparticles with particle sizes using a two plane sensing system were taken. 25–35 nm and monometallic Pd nanoparticles below Mixing was clearly shown for (a) and the absence 10 nm. In the absence of O , adsorbed carbonyl of blending for (b) using reconstructed tomograms 2 species formed on both Pd and Pt. CO adsorbed and the basic average permittivity data. linearly on top of Pt and in bridged configurations on

Hydrogenation and Reductive Amination of Pd. It was shown that adding Pd to Pt/Al2O3 shifted Aldehydes using Triphos Ruthenium Catalysts the CO-poisoned state to lower temperatures, F. Christie, A. Zanotti-Gerosa and D. Grainger, therefore increasing the temperature range for low- ChemCatChem, 2018, 10, (5), 1012 temperature CO oxidation (Figure 1).

Fig. 1. Reprinted with Pd Pt CO permission from N. M. Martin, M. Skoglundh, G. Smedler, A. Raj, D. Thompsett, P. Velin, F. J. Martinez-Casado, Z. Matej, O. Balmes and P.-A. Carlsson, J. Phys. Chem. C, 2017, 121, (47), 26321. Copyright 2017 American Chemical Society

Renewable Acrylonitrile Production detailed model. It approximates the behaviour E. M. Karp, T. R. Eaton, V. Sànchez i Nogué, V. of the full model; accuracy is sacrificed for speed Vorotnikov, M. J. Biddy, E. C. D. Tan, D. G. Brandner, of evaluation. A few runs of the detailed model R. M. Cywar, R. Liu, L. P. Manker, W. E. Michener, M. are used to train the initial surrogate, then the Gilhespy, Z. Skoufa, M. J. Watson, O. S. Fruchey, D. surrogate model is used in place of the detailed R. Vardon, R. T. Gill, A. D. Bratis and G. T. Beckham, model with a simplex optimisation method. The Science, 2017, 358, (6368), 1307 kinetic parameters are evaluated with the full model and the surrogate model can be updated with the Acrylonitrile (ACN), presently derived from new information. propylene and used as a commodity chemical for the production of plastics and fibres, is a candidate for A Study of the Soot Combustion Efficiency of an displacing petroleum feedstocks with biomass. The SCRF® Catalyst vs a CSF During Active Regeneration starting material in this work is 3-hydroxypropionic L. Cumaranatunge, A. Chiffey, J. Stetina, K. acid (3-HP) produced via microbes from sugars. McGonigle, G. Repley, A. Lee and S. Chatterjee, A TiO2 catalyst is used to dehydrate and nitrilate Emission Control Sci. Technol., 2017, 3, (1), 93 using NH3. Yields are over 90% and could be scaled up to 98%. Hazards are reduced as the process is A Pt-based catalysed soot filter (CSF) was found endothermic and the production of HCN is avoided. to have a significantly higher soot combustion efficiency compared to a Cu-SCR catalyst-coated Application of Surrogate Modelling to the soot filter (SCRF® catalyst) under typical active Optimisation of Kinetic Parameters in an Emissions regeneration conditions at 550–600ºC. There was Control Catalyst Model Using Vehicle Drive Cycle sufficient NO production capacity in the CSF and Data 2 the impact of NO2 to the overall soot combustion J. E. Etheridge, G. John and T. C. Watling, Emission efficiency under active regeneration conditions was Control Sci. Technol., 2017, 3, (4), 310 significant. It is thought that NO2 produced in situ Surrogate modelling was used to optimise kinetic in the CSF is quickly reacted with soot to drive the parameters in a vehicle emissions control catalyst thermodynamic equilibrium of the NO oxidation. model using engine or vehicle test data. This This leads to significantly higher soot combustion approach reduces the time needed for optimisation efficiency in the Pt-coated CSF compared to the by reducing the number of evaluations of the SCRF® catalyst or an uncoated filter.

www.technology.matthey.com

Inter-Diffusion of Iridium, Platinum, Palladium and Rhodium with Germanium Improved materials for the next generation of electronic devices

Adrian Habanyama* platinum, palladium and rhodium. Our approach Department of Physics, Copperbelt University, was essentially twofold. Firstly, conventional thin PO Box 21692, Jambo Drive, Riverside, Kitwe film couples were used to study the sequence 10101, Zambia; Department of Physics, of phase formation in the germanide systems. University of Cape Town, Rondebosch 7700, Conventional thin film couples were also used to South Africa identify and monitor the dominant diffusing species during the formation of some of the germanides *Email: [email protected] as these can influence the thermal stability of a device. Secondly, we observed and analysed Craig M. Comrie several aspects of the lateral diffusion reactions in iThemba LABS, National Research Foundation, these four systems, including activation energies PO Box 722, Somerset West 7129 South and diffusion mechanisms. Lateral diffusion Africa; Department of Physics, University of couples were prepared by the deposition of thick Cape Town, Rondebosch 7700, South Africa rectangular islands of one material on to thin films of another material. Rutherford backscattering spectrometry (RBS) and microprobe-Rutherford The down-scaling of nanoelectronic devices to backscattering spectrometry (µRBS) were ever smaller dimensions and greater performance used to analyse several aspects of the thin film has pushed silicon-based devices to their physical and lateral diffusion interactions respectively. limits. Much effort is currently being invested in X-ray diffraction (XRD) and scanning electron research to examine the feasibility of replacing microscopy (SEM) were also employed. Si by a higher mobility semiconductor, such as germanium, in niche high-performance metal 1. Introduction oxide semiconductor (MOS) devices. Before Ge can be adopted in industry, a suitable contact Ge has several attractive properties such as high material for the active areas of a transistor must mobility of charge carriers and very low carrier be identified. It is proposed that platinum group freeze-out temperatures (1–4). There is currently metal (pgm) germanides be used for this purpose, much research on high mobility semiconductors, in a similar manner as metal silicides are used such as Ge, with the view of using them to replace in Si technology. Implementation of Ge-based Si in niche high-performance MOS devices (5–7). technology requires a thorough understanding of Before Ge can be adopted by industry a suitable the solid-state interactions in metal/Ge systems contact material to the active areas (source, drain in order to foresee and avoid problems that may and gate) of a transistor must be identified. It is be encountered during integration. We present a proposed that pgm germanides be used for this systematic study of the solid-state interactions in purpose, similar to the manner in which metal germanide systems of four of the pgms: iridium, silicides are used in present Si technology (6, 8).

Our work investigates the solid-state interactions (23) reported a scanning tunnelling microscopy in germanide systems of four of the pgms: Ir, Pt, (STM) and spectroscopy study of the formation of Pd and Rh, in thin film and lateral diffusion couples. Pt/germanide phases on Ge (111). This study gave In the design of transistors the contact material a demonstration of the structural dependence of should be stable over a wide temperature range. electronic properties in the Pt-Ge system. Schottky Conventional thin film couples are well suited for barrier diodes have been used in many applications investigating the phase formation sequence and such as gates for metal semiconductor field-effect temperature stability of the phases of a system. We transistors (MESFET), solar cells and detectors have also used thin film couples to identify some (24–27). A reduction of the PtGe/Ge electron of the dominant diffusing species during the phase Schottky barrier height by rapid thermal diffusion formation. For device integrity it is important to of phosphorus was reported by Henkel et al. (28). identify the dominant diffusing species during the The results showed that rapid thermal diffusion formation of the respective germanides as this can from a solid diffusion doping source was effective in influence their thermal stability. reducing Schottky barrier heights of Pt germanide The samples used for studying lateral diffusion Schottky barrier diodes on Ge. Chawanda et al. reactions were composed of a thick island of one (29) investigated the change in the current-voltage material on top of a much thinner film of another (I–V) electrical properties of Pt Schottky contacts material. Upon annealing the island material on Ge (100) at different annealing temperatures. would react with the underlying film through Their results showed that the as-deposited barrier vertical diffusion, going through a sequence of heights had values that were near the band gap phases until the most island-material rich phase of Ge for Pt/Ge (100) Schottky diodes resulting in is formed. Since no further vertical reaction with good Schottky source/drain contact materials in the underlying film is possible, the most island- p-channel Ge-MESFETs for the hole injection from material rich phase may then grow laterally until source into inverted p-channel (30). Chawanda it attains a critical width, after which other phases et al. (31) also studied the electrical properties appear and grow simultaneously. This is a case of of Pd Schottky contacts with Ge (100). I–V and multiple phase formation as would be found in bulk capacitance-voltage (C–V) measurements were diffusion couples. Lateral diffusion couples thus performed under various annealing conditions. provide the transition between thin film and bulk Only one Pd germanide phase, PdGe, was formed. behaviour. A hole trap at 0.33 eV above the valence band was Since the island material is abundant for diffusion observed after annealing at 300°C. In another in the lateral diffusion couples, phase formation and study Chawanda et al. (32) investigated the change reaction kinetics can be studied to a greater extent in the electrical properties of Ir Schottky barrier than in thin film planar structures. Lateral diffusion diodes on Ge (100). Electrical characterisation of structures can be used to simulate bulk diffusion these contacts using I–V and C–V measurements couples because phase formation could extend to was performed under various annealing conditions. lengths of around 100 µm (9). In kinetic studies Thermal stability of the Ir/Ge (100) sample was of thin film planar structures the diffusion lengths observed up to an annealing temperature of are typically less than 0.5 µm. One can therefore 500°C. The results also showed that the onset study the transition from thin film to bulk diffusion temperature for agglomeration (binding of primary couple behaviour. The study of lateral diffusion particles leading to phase formation) in 20 nm Ir/ couples is particularly well suited for dealing with Ge (100) samples occurs between 600–700°C. the challenges of achieving the required lateral Gaudet et al. (7) carried out a systematic study abruptness of semiconductor junctions. Excessive of thermally induced reactions of 20 transition diffusion of the substrate element, in this case metals with Ge substrates. They monitored Ge, during germanide formation could result in metal/Ge reactions in situ during ramp anneals overgrowth and bridging in devices (10). at 3°C s–1 using time-resolved XRD and diffusion Various early techniques were developed to light scattering. They also carried out resistance study lateral diffusion couples (11–20). In later measurements. Their results showed that the pgms studies, µRBS was used (9, 21, 22). The major Pd and Pt were among the six most promising advantage of this technique is its ability to give candidates for microelectronic applications, the depth information. other candidates being nickel, cobalt, copper and Some previous work has been carried out in the iron. Ni is the most used metal for reducing contact research field of pgm/Ge junctions. Saedi et al. resistance. An example of previous research in the

212 © 2018 Johnson Matthey https://doi.org/10.1595/205651318X696639 Johnson Matthey Technol. Rev., 2018, 62, (2) area of Ni/Ge junctions is the work reported by plane in monitoring the direction of flow of atoms. A Peng et al. (33) on the I–V characteristics of Ni/ thin layer of Ti acted as an inert marker interposed Ge (100) Schottky diodes and the Ni germanide between coupling layers of metal and Ge; the Ge induced strain after subjecting the Schottky layer being at the surface of the sample. Upon contacts to rapid thermal annealing in the annealing of this structure both Ge and metal atoms temperature range of 300–600°C. Their results could have diffused across the marker at different showed that the orthorhombic structure of NiGe rates causing it to shift in position towards the induces epitaxial tensile strain on Ge substrates dominant diffusing species. The marker Ti signal due to the difference in lattice constants. They also was monitored by RBS for different annealing suggested that the increase in barrier height with times. The dominant diffusing species (DDS) during increasing annealing temperature may have been phase formation was determined by observing the due to the conduction band edge shift by the strain relative shift of the marker. after the germanidation process. The lateral diffusion couples were also prepared Hallstedt et al. (34) studied the phase by electron beam evaporation at a base pressure in transformation and sheet resistance of Ni on the low 10–5 Pa range. A thin film of one material was single crystalline SiGe(C) layers after annealing deposited first. An ordinary Si wafer with an array treatments at 360–900°C. The role of strain of 390 × 780 mm2 rectangular windows (referred relaxation or compensation in the reaction of Ni to as a Si mask), made by photolithographic on Si1–x–yGexCy layers due to Ge or carbon out- techniques and selective etching, was then brought diffusion to the underlying layer during the phase into contact with the substrate without breaking transformation was investigated. Formation of vacuum. Island material was deposited through crystalline Ni(SiGe) was complete at 400–450°C the mask to form structures consisting of metal but the thermal stability decreased rapidly with islands on Ge films and Ge islands on metal films. increased Ge amount due to agglomeration. This Figure 1 shows a schematic illustration of the thermal behaviour was shifted to higher annealing sample preparation setup. The source metal or Ge temperatures when C was incorporated. Ni(SiGeC) sample pieces were each placed in one of the three layers formed at 500–550°C after which there crucibles on a water-cooled copper hearth. The was Ge segregation to the underlying layer and positions of the crucibles could be adjusted from the C accumulation at the interface. Thanailakis outside of the electron-beam evaporation chamber et al. (35) established a relationship between as- so that each material to be evaporated could be deposited Pd/Ge (111) and Ni/Ge (111) Schottky placed in the path of the electron beam in turn. Each barrier height values, the metal work functions and crucible was shielded from the adjacent one by a the density of surface states of the Ge substrate. 2 cm high partition to prevent cross-contamination during deposition. Above the crucibles was a shutter 2. Experimental that could be opened and closed from outside the chamber using a bar magnet. The sample changer, A study of the phase formation sequence, using which accommodated a maximum of three sample conventional thin film couples, was carried out prior holders, hovered above the shutter. The sample to the lateral diffusion study. This study was carried changer could be rotated using an external handle out using RBS and XRD for phase identification. in such a way as to place one sample holder at a Thermally oxidised single crystal Si wafers with a time in the line of sight of the target vapour. In (100) crystal orientation were used as substrates this way, it was possible to prepare up to three in all studies. In the studies of the Ir-Ge system, a sets of different samples (each on one sample thin layer of titanium (2 nm) had to be deposited holder) in one experimental run. Between the onto the SiO2 prior to the deposition of the coupling shutter and the sample changer hung the Si mask materials; this Ti layer reacted with SiO2 forming a holder which had a provision where a Si wafer ‘glue’ without which the structure could not adhere. with several rectangular 390 × 780 mm2 openings Electron beam vacuum deposition of coupling could be placed. The mask holder could be swung layers of metal and Ge was carried out at pressures from side to side without breaking the vacuum. in the low 10–5 Pa range. Likewise, its height could be adjusted externally. In a further preliminary investigation a marker The main features of the bottom compartment of technique was used to monitor atomic mobility the chamber were a system of vacuum pumps and during phase formation. The term marker refers to a cryopanel. A baffle valve is closed between the a material deposited in the sample as a reference upper and lower compartments to ensure that the

Mask holder rotation handle

Sample holder rotation handle

Sample holder Sample changer Thickness monitor Samples Shutter Position of mask

Aperture

Crucibles Leak valve

Cryopanel Turbo pump

Ti-sublimation VacIon pump pump

Fig. 1. Schematic illustration of the high-vacuum system used for electron-beam evaporation of thin films and lateral diffusion couples. The upper section contains the sample holder, thickness monitor and electron gun. A baffle valve is closed between the upper and lower sections to ensure that the lower section of the system is maintained under vacuum when the system is not in use or during sample changing

lower section of the system is maintained under were recorded along with position information. vacuum when the system is not in use or during Typically, 128 × 128 spectra were generated in sample changing. each run and once it had been established that After removing the samples with islands from no variation in composition was observed parallel the evaporation chamber they were cleaved into to the interface, spectra were summed along this twelve identical samples, each with two or three direction. This reduced the analysis to a one- islands. Furnace-annealing was used to activate dimensional traverse of 128 spectra perpendicular solid-state interaction after which the samples to the interface, thereby optimally exploiting beam were analysed. The samples were examined using position while improving on statistics. The RBS SEM to distinguish the various reaction regions and data was analysed using RUMP (36) simulations. measure their diffusion lengths. Representative samples were selected for further analysis by 3. Results mRBS. The distribution of elements as a function of lateral position was obtained using mRBS. This 3.1 Iridium-Germanium System technique also provided information regarding 3.1.1 Thin Film Couples the elemental distribution as a function of depth and the thickness of the films. A 2 MeV a-beam Ir-Ge conventional thin film couple samples for focused down to a pixel size of about 1 × 1 µm was the study of the phase formation sequence had scanned across a well-defined area of the samples. the structure: SiO2/Ge (550 nm)/Ir (90 nm). The This area, typically of 400 × 400 mm2, was chosen samples were annealed in vacuum for various to include all reaction regions observed in the SEM periods of time at temperatures ranging from 350°C micrographs. Sample orientation was adjusted in to 800°C. The results showed the appearance such a way that the interfaces of the regions of of the compounds IrGe and Ir4Ge5 at annealing interest lay horizontally in line with the original temperatures of around 350°C. Observation of the island edge so that the microprobe beam scanned separate formation of either compound as a first parallel to the original island interface. RBS spectra phase could not be achieved. An XRD spectrum

214 © 2018 Johnson Matthey https://doi.org/10.1595/205651318X696639 Johnson Matthey Technol. Rev., 2018, 62, (2) obtained following annealing at 400°C for 80 minutes is shown in Figure 2. Energy, MeV 1.4 1.6 1.8 250 The figure shows the presence of IrGe and Ir4Ge5 together with unreacted Ir indicating co-existence.

After these two phases, Ir3Ge7 formed; from our 200 Ti data it was not possible to tell whether all the Ir –2 –2 –2 2 –2 was consumed before Ir3Ge7 appeared. The phase 5

150 cm Ir SiO IrGe is the most Ge rich in the Ir-Ge system, it was Ge

4 4 Si <> IrGe Ge 180 cm the final phase observed and only formed above 150 cm Ir cm 35 275 800°C. There was no evidence of the presence of 100 Ti

Ir3Ge4 during the whole reaction. Normalised yield A 1.2 nm layer of Ti acted as an inert marker 50 interposed between coupling layers of Ir and Ge to monitor atomic mobility during phase formation. 0 This structure was annealed at 400°C for different 300 350 400 450 500 times. Figure 3 is a diagram showing the results Channel obtained from a RUMP simulation (36) on an RBS Fig. 3. A diagram showing the results obtained spectrum of a sample annealed at 400°C for 20 from a RUMP simulation on an RBS spectrum minutes. The results show the Ti marker as lying of a sample annealed at 400°C for 20 minutes. between the phase Ir4Ge5 and the unreacted Ge at The inset shows the result of the simulation with 15 –2 the surface. thickness in units of 10 cm Figure 4 shows the Ti ‘glue’ and Ti marker signals before and after annealing. The Ti marker signal shifted by 0.02 MeV to higher energies, i.e. towards the surface. From these results a conclusion can be Energy, MeV drawn if certain assumptions are made. Firstly, if 1.25 1.30 1.35 1.40 1.45 1.50 1.5 IrGe was the first phase to form, then Ge would be As deposited the sole moving species for both IrGe and Ir4Ge5 400ºC, 20 min formation. This is so because had Ir been a moving A species during either IrGe or Ir4Ge5 formation, it 1.0 B would have been observed to diffuse across the marker, which was not the case. Secondly, if Ir4Ge5 ‘Glue’ Ti were the first phase to form, then all we canbe sure of is that Ge was the sole moving species 0.5 Marker Ti Normalised yield

300

Si 0 250 Ge 300 320 340 360 380 Ir Channel IrGe 200 Ir4Ge5 Fig. 4. Ti ‘glue’ and Ti marker signals before and after annealing. The marker signal shifted by 0.02 150 MeV to higher energies (A to B), while the Ti ‘glue’ signal did not shift 100 Yield, arbitrary units Yield, arbitrary

50 during Ir4Ge5 formation. During IrGe formation 0 there would be no discernible indicator as to which 0 20 40 60 80 species was moving, the marker being at the 2q, º interface between Ir4Ge5 and Ge. The only firm Fig. 2. X-ray diffraction spectrum of a sample of conclusion we can draw from the marker results is composition SiO2/Ge (550 nm)/Ir (90 nm) after that Ge was the sole diffusing species during Ir4Ge5 annealing at 400°C for 80 minutes formation.

3.1.2 Lateral Diffusion Couples Energy, MeV 1.2 1.4 1.6 1.8 2.0 The lateral diffusion couple results with Ir islands on Ge films showed little germanide growth Ir upon annealing with a reaction region that was 2.0 Region A (Ge island) too narrow to properly resolve and monitor. The Region B Ir4Ge5 reverse configuration with Ge islands on Ir films 1.5 Region C Region D showed substantial lateral diffusion. Figure 5 Unreacted Ir film Ir3Ge7 shows an SEM micrograph of part of a lateral 1.0 IrGe4 diffusion couple, with a Ge island (250 nm) on an Ge

Normalised yield Ir Ir film (25 nm), which had been annealed at 800°C for 30 minutes. 0.5 Four different reaction regions were observed and are labelled A to D in Figure 5. The original edge 0 of the island is clearly visible as the right edge of 300 350 400 450 500 the white vertical strip in region C. This sample Channel then underwent mRBS in order to determine the Fig. 6. Superposition of selected RBS spectra composition and thickness of the various regions. from each of the reaction regions and from the An area to be scanned was chosen to include all unreacted Ir film. Ir peak heights of the various regions observed. phases and surface positions of Ge and Ir are A representative RBS spectrum picked from each indicated of the four reaction regions labelled A to D is shown in Figure 6. A spectrum picked from the unreacted

8 Original interface Ir film is also included in Figure 6. Peak heights of the spectra taken from regions D, C and B show the phases Ir

Ge , Ir Ge and IrGe respectively. x 4 5 3 7 4 6 The germanides in these three regions are seen Ge + at the surface position. On the other hand, the Ir IrGe4 IrGe4 peak from region A lies below the surface position, 4 indicating that there was no Ir at the surface. Region A therefore consisted of a layer of unreacted Ir3Ge7 Ge on top of the IrGe4 phase. Composition, IrGe 2 Decomposition Ir Ge The RBS data from the scanned area were analysed 4 5 to get stoichiometric information from integrated counts of the Ir and Ge peaks. Figure 7 shows 0 0 100 200 300 400 Lateral position, mm

Fig. 7. Stoichiometric information of a Ge island (250 nm) on an Ir film (25 nm) annealed at 800°C for 30 minutes, derived from integrated counts of the Ir and Ge peaks as a function of position

A B C D the results as a function of lateral position. The

figure shows a region of the phase Ir3Ge7 inside the

original island interface. The width of this Ir3Ge7 region was observed to increase with annealing time. This suggests a gradual decomposition of the

phase IrGe4 into Ir3Ge7 by the reaction 3IrGe4 → 30 mm Ir3Ge7 + 5Ge. As in the thin film work, the major difference Fig. 5. SEM micrograph of part of a Ge island (250 between the lateral diffusion couples prepared by nm) on an Ir film (25 nm) annealed at 800°C annealing at temperatures below 800°C and those for 30 minutes showing all the different reaction at and above 800°C was the absence of the IrGe regions observed 4 phase below 800°C; hence no region of IrGe4

Si 3.2 Platinum-Germanium System 250 Ge Pt3Ge2 3.2.1 Thin Film Couples 200

Pt-Ge conventional thin film couple samples for 150 the study of the phase formation sequence had the structure: SiO2/Ge (500 nm)/Pt (120 nm) 100 15 –2 or equivalently, SiO2/Ge (2270 × 10 cm )/ Yield, arbitrary units Yield, arbitrary 15 –2 Pt (780 × 10 cm ). These were annealed in 50 vacuum for various periods of time at temperatures ranging from 150°C to 300°C. The results showed 0 0 20 40 60 80 the appearance of Pt Ge as the first phase formed 2 2q, º at an annealing temperature of around 190°C. Figure 8 shows an RBS spectrum obtained Fig. 9. X-ray diffraction spectrum of a sample of following annealing at 190°C for 2 hours, together composition SiO2/Ge (500 nm)/Pt (120 nm) after annealing at 220°C for 80 minutes. The non- with its RUMP simulation. The inset in Figure 8 congruent phase Pt3Ge2 is observed to form second shows that at this stage of the reaction the sample consisted of Si covered with SiO2 as the substrate, followed by Ge, then Pt2Ge and some unreacted our data it was not possible to tell whether all the

Pt at the surface. The results shown in Figure 8 Pt was consumed before Pt3Ge2 appeared. The were consistent with results obtained from the XRD next detected phase was PtGe at 250°C. The last analysis performed on this same sample. phase observed was PtGe2 which formed from the

Pt3Ge2 was observed as the second phase to interaction of PtGe with unreacted Ge. form at 220°C. Figure 9 shows an XRD spectrum A 1.2 nm layer of Ti acted as an inert marker of a sample annealed at 220°C for 80 minutes interposed between coupling layers of Pt (300 × 1015 –2 15 –2 indicating the presence of Pt3Ge2. The result was cm ) and Ge (430 × 10 cm ) to monitor atomic consistent with the result of a RUMP simulation on mobility during phase formation. Figure 10 shows RBS data obtained from the same sample. From

Energy, MeV Energy, MeV 1.0 1.2 1.4 1.6 1.8 1.2 1.4 1.6 1.8 300 250 Ti –2

–2 –2 –2

250 2 –2 –2 –2 2 –2 200 cm Pt Ge SiO 2 cm Ge Ge Ge SiO Si <> cm 70 Ge 2 Pt Pt 110 cm 2 Si <> 255 200 cm 55 Pt PtGe 235 cm 2175 cm 2175 600 Pt 275 cm 150 150 100 100 Normalised yield Normalised yield 50 50

0 0 250 300 350 400 450 500 200 250 300 350 400 450 500 Channel Channel Fig. 8. RBS spectrum of a sample of composition Fig. 10. RBS spectrum and RUMP simulation of 15 –2 15 –2 a marker structure of composition SiO / Pt (300 SiO2/Ge (2270 × 10 cm )/Pt (780 × 10 cm ) 2 15 –2 15 –2 after annealing at 190°C for 2 hours, together with × 10 cm )/Ti (1.2 nm)/Ge (430 × 10 cm ) its RUMP simulation. The inset shows the result of after annealing at 250°C for 20 minutes. The inset the simulation with thickness in units of 1015 cm–2. shows the result of the RUMP simulation with thickness in units of 1015 cm–2 The phase Pt2Ge is observed to form first

217 © 2018 Johnson Matthey https://doi.org/10.1595/205651318X696639 Johnson Matthey Technol. Rev., 2018, 62, (2) the results obtained from a RUMP simulation on an It is therefore unlikely for the first process to RBS spectrum of a sample annealed at 250°C for have taken place above the marker. The second 20 minutes. The inset shows the Ti marker as lying reaction is the one most likely to have taken place between two layers of Pt2Ge. We observe some above the marker. The PtGe was therefore formed unreacted Ge at the surface and some unreacted from Pt2Ge and Ge above the marker where there Pt below the marker. The results show a significant was no discernible indicator as to which species 15 –2 amount (235 × 10 cm ) of PtGe above the was moving during its growth. The phase Pt3Ge2 marker, between Pt2Ge and unreacted Ge. There is was not observed to form in the presence of the no PtGe below the marker. Ti marker; PtGe was formed after Pt2Ge skipping

Figure 11 shows the Ti marker signals before Pt3Ge2, which was observed in the absence of the and after annealing at 250°C for 20 minutes. The Ti marker during the study of the phase formation

Ti marker signal shifted from A to B by 0.025 MeV sequence using the sample structure: SiO2/Ge to lower energies, i.e. towards the substrate. It is (500 nm)/Pt (120 nm). clear that both Pt and Ge atoms migrated across the marker to interact, forming Pt2Ge on either 3.2.2 Lateral Diffusion Couples side of the marker. The amount of Pt that crossed the marker and was observed above it, in units of The lateral diffusion couple results for Pt islands 1015 cm–2, is shown in Equation (i): on Ge films showed little lateral diffusion upon annealing with a reaction region that was too # Pt = (2/3) × (55) (from Pt2Ge) + (1/2) × narrow to resolve and monitor properly. The (235) (from PtGe) = 154 (i) reverse configuration with Ge islands on Pt films The amount of Ge that crossed the marker and showed substantial lateral germanide growth. was observed below it is shown in Equation (ii): Figure 12 shows an SEM micrograph of part of a lateral diffusion couple of a Ge island (145 nm) on # Ge = (1/3) × (110) (from Pt2Ge) = 37 (ii) a Pt film (35 nm) annealed at 450°C for 24 hours. The atomic diffusion ratio of Pt to Ge is therefore The figure shows a part of the island that is close about 4 to 1. to one of its corners. Five different regions, labelled There are two possible mechanisms by which the A to E, are observed where A starts in the middle PtGe above the marker could have been formed. of the island and E is the outermost region. The

These are: Pt2Ge → PtGe + Pt and Pt2Ge + Ge → original edge of the island is clearly visible between 2PtGe. No PtGe was seen below the marker; the regions C and D. first of these two processes did not take place in The sample in Figure 12 underwent mRBS analysis that region as it would have left some PtGe there. to obtain stoichiometric information by RUMP simulation of the Pt and Ge peaks. The scanned

Energy, MeV 1.25 1.30 1.35 1.40 1.45 1.50 1.5

B A 1.0

0.5 C E Normalised yield A B D

30 mm

0 300 320 340 360 380 Channel Fig. 12. SEM micrograph of a Ge island (145 nm) Fig. 11. Upon annealing at 250°C for 20 minutes, on a Pt film (35 nm) annealed at 450°C for 24 the marker Ti signal shifts by about 0.025 MeV hours, showing representative regions in a part of from channel 342 to channel 335 (A to B) the island which is close to one of the corners

218 © 2018 Johnson Matthey https://doi.org/10.1595/205651318X696639 Johnson Matthey Technol. Rev., 2018, 62, (2) area was chosen to include all regions observed in the SEM micrograph. Figure 13 shows the results Energy, MeV 1.0 1.2 1.4 1.6 1.8 as a function of lateral position. 2.0 The regions B, C, D and E are found to consist of Pt PtGe , Pt Ge , PtGe and unreacted Pt respectively. 2 2 3 Region A Region A appeared to be composed of a mixture of 1.5 Region B PtGe and unreacted Ge. To show this more clearly, 2 Region C PtGe a representative spectrum picked from each of the Region D Pt2Ge3 five regions labelled A to E is shown in Figure 14. 1.0 Region E PtGe2 Downward pointing arrows are used to indicate the surface positions of Ge and Pt. The spectrum from Normalised yield Shoulder region E shows a peak of unreacted Pt and no Ge. 0.5 Ge Pt Peak heights of the spectra taken from regions D,

C and B show the phases PtGe, Pt2Ge3 and PtGe2 respectively. The Ge in these three regions is seen 0 at the surface position. It is seen from the solid 250 300 350 400 450 Channel line in the figure that the region A consisted of Fig. 14. Superposition of selected RBS spectra from unreacted Ge and the phase PtGe . The Pt peak 2 each of the five regions. Pt peak heights of the of the solid line lies at the surface position. This various phases and the surface positions of Ge and shows that there was some Pt at the surface. The Pt are indicated ‘shoulder’ marked in the figure however indicates that there was less Pt at the surface than deeper down. Region A therefore consisted of PtGe2 at the signals. This is by virtue of their having atomic bottom while at the top there was unreacted Ge numbers that lie relatively close to each other in the together with the phase PtGe2. periodic table. At the same time, samples needed to comprise of elemental layers thick enough to 3.3 Palladium-Germanium System induce an appreciable X-ray yield. The structure used was SiO2/Ge (500 nm)/Pd (70 nm). 3.3.1 Thin Film Couples In this system, reaction was induced at relatively low temperatures. Figures 15 and 16 show RBS Of major concern while determining the best spectral data for samples annealed at temperatures thickness of our Pd-Ge sample structure was the of 100°C and 150°C for 2 hours and 80 minutes likelihood of overlap between Pd and Ge RBS respectively. Data from XRD analysis were also obtained, these are displayed alongside the corresponding RBS data. Layer thicknesses obtained 3.0

Original interface by RUMP simulations (solid lines) are shown. Rather Ge + straightforward behaviour is observed in this 2.5 PtGe2 system with the two congruent phases Pd2Ge and PtGe x 2 PdGe being the only ones observed. Pd Ge was the 2.0 2 first to form at around 100°C. Pt2Ge3 1.5

PtGe 3.3.2 Lateral Diffusion Couples 0.1 Samples for lateral diffusion study were prepared Composition, PtGe by deposition of thick Ge islands on thin Pd films. 0.5 This configuration was chosen on the basis of the Pt results observed in the Ir-Ge and Pt-Ge systems. 0 0 50 100 150 200 250 Several lateral diffusion samples were annealed Lateral position, mm at various temperatures for different lengths of time. The Pd-Ge system exhibited relatively low Fig. 13. Stoichiometric information of a Ge island temperature reaction. It was therefore necessary on a Pt film annealed at 450°C for 24 hours derived from RUMP simulation of the Pt and Ge to carry out the investigation for this system at peaks, as a function of lateral position much lower temperatures than those used for the other systems.

(a) (b) Energy, MeV 300 0.8 1.0 1.2 1.4 1.6 Si 250 Ge 100 Pd Pd2Ge Pd 200 80 240 cm–2

150 60 Pd2Ge 220 cm–2 100 40 Ge –2 Normalised yield Yield, arbitrary units Yield, arbitrary 2035 cm 50 20 SiO2 Si <> 0 0 20 40 60 80 0 2q, º 150 200 250 300 350 400 450 Channel

Fig. 15. X-ray diffraction and corresponding RBS spectrum, of a sample of composition SiO2/Ge (500 nm)/ Pd (70 nm) after annealing at 100°C for 2 hours. The inset shows the result of the RUMP simulation with 15 –2 thickness in units of 10 cm . The phase Pd2Ge is observed to form first

(a) (b) Energy, MeV 0.8 1.0 1.2 1.4 1.6 300

Si 100 150ºC, 80 min 250 Ge Simulation of Pd-Ge/Pd-Ge/Ge/Si-O/Si Pd2Ge PdGe 80 Pd2Ge 200 360 cm–2 60 150 PdGe 300 cm–2 100 40 Ge

Normalised yield –2 Yield, arbitrary units Yield, arbitrary 1860 cm

50 20 SiO2 Si <> 0 0 20 40 60 80 0 2q, º 150 200 250 300 350 400 450 Channel

Fig. 16. X-ray diffraction and corresponding RBS spectrum of the sample of composition SiO2/Ge (500 nm)/ Pd (70 nm) after annealing at 150°C for 80 minutes. The inset shows the result of the RUMP simulation 15 –2 with thickness in units of 10 cm . The phase PdGe is observed to form after Pd2Ge

Figure 17 shows an SEM micrograph of one The spectrum from region C shows a peak of representative sample with a 100 nm thick Ge unreacted Pd and no Ge. The peak height of the island on a 20 nm thick Pd film, annealed at 325°C spectrum taken from the region B shows the phase for 2 hours, showing three distinct regions labelled Pd2Ge. This phase is seen at the surface position. A to C. Areas which were chosen to include all From the solid line in the figure, it can be seen that the reaction regions observed were scanned on region A consisted of unreacted Ge and PdGe. The the nuclear microprobe for analysis by mRBS. A Pd peak of the solid line lies at the surface position, spectrum picked from each of the three regions showing that there was some Pd at the surface. of the Pd-Ge lateral diffusion sample is shown in The ‘shoulder’ marked in the figure indicates that Figure 18. there was less Pd at the surface than deeper

Energy, MeV 0.8 1.0 1.2 1.4 1.6 0.8 C Pd

Pd2Ge B 0.6 Region A (Ge island) Region B A Region C (Pd film) 0.4 PdGe Ge

Normalised yield 0.2 Shoulder Pd 30 mm

0 200 250 300 350 400 Fig. 17. SEM micrograph of a Ge island (100 nm) Channel on a Pd film (20 nm) annealed at 325°C for 2 hours showing the different reaction regions Fig. 18. Superposition of selected RBS spectra from each of the regions of the Pd-Ge lateral diffusion sample. Pd peak heights of the various phases and down. Region A therefore consisted of PdGe at the surface positions of Ge and Pd are indicated bottom while at the top there was unreacted Ge intermingled with PdGe. reported on in this study; therefore great care was 3.4 Rhodium-Germanium System taken to abate this. The sample structure used in this study was SiO2/Ge (500 nm)/Rh (60 nm). 3.4.1 Thin Film Couples Figures 19, 20 and 21 show XRD results alongside RBS spectra with RUMP simulations for

Only four equilibrium phases exist in the Rh-Ge samples with the structure SiO2/Ge (500 nm)/Rh system: Rh2Ge, Rh5Ge3, RhGe and Rh17Ge22. The (60 nm) annealed between 320°C and 400°C. Our chance of getting excessive overlap of RBS peaks RBS data strongly suggest the formation of the was greater in this system than in any other yet non-congruent phase Rh2Ge as the first phase but

(a) (b) Energy, MeV 300 0.8 1.0 1.2 1.4 1.6 Si 250 Ge 100 320ºC, 2 h Simulation of Rh-Ge/Rh-Ge/Ge/Si-O/Si RhGe

200 80 Rh2Ge 265 cm–2 150 60 RhGe 345 cm–2 100 40

Normalised yield Ge Yield, arbitrary units Yield, arbitrary –2 50 1890 cm 20 SiO2 0 Si <> 0 20 40 60 80 0 2q, º 150 200 250 300 350 400 450 Channel

Fig. 19. X-ray diffraction and corresponding RBS spectrum of a sample of composition SiO2/Ge (500 nm)/ Rh (60 nm) after annealing at 320°C for 2 hours. The inset shows the result of the RUMP simulation with thickness in units of 1015 cm–2

(a) (b) Energy, MeV 300 0.8 1.0 1.2 1.4 1.6 Si 250 Ge 100 330ºC, 2 h Simulation of Rh-Ge/Ge/Si-O/Si RhGe 200 80 –2

150 RhGe

60 740 cm

–2 100 40 Ge Normalised yield Yield, arbitrary units Yield, arbitrary 50 cm 1800 20 SiO2 Si <> 0 0 20 40 60 80 0 2q, º 150 200 250 300 350 400 450 Channel

Fig. 20. X-ray diffraction and corresponding RBS spectra of a sample of composition SiO2/Ge (500 nm)/Rh (60 nm) after annealing 330°C for 2 hours. The inset shows the result of the RUMP simulation with thickness 15 –2 in units of 10 cm . From RBS data the non-congruent phase Rh2Ge is observed to convert to RhGe, Rh2Ge X-ray peaks could not be observed

(a) (b) Energy, MeV 300 0.8 1.0 1.2 1.4 1.6 Si 250 Ge 100 400ºC, 20 min Simulation of Rh-Ge/Ge/Si-O/Si Rh17Ge22 22 200 80 –2 Ge 17 810 cm 150 60 Rh

–2

100 40 Ge Normalised yield Yield, arbitrary units Yield, arbitrary

1720 cm 1720 50 20 SiO2 Si <> 0 0 20 40 60 80 0 2q, º 150 200 250 300 350 400 450 Channel

Fig. 21. X-ray diffraction and corresponding RBS spectra of a sample of composition, SiO2/Ge (500 nm)/Rh (60 nm) after annealing at 400°C for 20 minutes. The inset shows the result of the RUMP simulation with 15 –2 thickness in units of 10 cm . The phases Rh17Ge22 is observed

there is no firm evidence of this from the X-ray 3.4.2 Lateral Diffusion Couples data. RBS data showed that the Rh2Ge started to form around 280°C and later gave way to RhGe. Samples for lateral diffusion study were prepared

Figure 21 shows that Rh17Ge22 was formed after by deposition of thick Ge islands on thin Rh films. the RhGe phase. Only RhGe and Rh17Ge22 peaks This configuration, as for the Pd-Ge lateral diffusion were observed in the X-ray results, probably samples, was chosen on the basis of the results because the Rh2Ge layer was too thin to induce a observed in the Ir-Ge and Pt-Ge systems. Several good X-ray yield. samples were annealed at various temperatures

222 © 2018 Johnson Matthey https://doi.org/10.1595/205651318X696639 Johnson Matthey Technol. Rev., 2018, 62, (2) while time was monitored in the usual way. two regions are seen at the surface position. It can Shown in Figure 22 is an SEM micrograph of one be seen from the solid line in the figure that the representative sample, with a 100 nm Ge island region A consisted of unreacted Ge and the phase on a 20 nm Rh film, annealed at 600°C for 15 Rh17Ge22. The Rh peak of the solid line lies below minutes. The figure shows four reaction regions the surface position. This shows that there was no labelled A to D. Areas which were chosen to include Rh at the surface. Region A consisted of unreacted all the reaction regions observed were scanned on Ge overlaying the phase Rh17Ge22. For very long the nuclear microprobe for analysis by mRBS. annealing times at relatively high temperatures RBS spectra picked from each of the four regions, (around 600°C and above) the phase RhGe was A, B, C and D are shown in Figure 23. The observed to slowly stretch across the original spectrum from region D shows a peak of unreacted interface into the island region. Rh and no Ge. Peak heights of the spectra taken from regions C and B show the phases RhGe and 4. Discussion Rh17Ge22 respectively. The germanides in these In this work we used conventional thin film as well as lateral diffusion couples to study the germanide systems of four pgms: Ir, Pt, Pd and Rh.

D 4.1 Iridium-Germanium System C 4.1.1 Thin Film Couples

B Using conventional thin film couples, IrGe and

Ir4Ge5 where observed to be the first phases A to form in the Ir-Ge system and co-existed at annealing temperatures of around 350°C. Ir3Ge7 formed after these two phases while the IrGe 100 mm 4 phase only appeared above 800°C. By interposing a thin layer of Ti (1.2 nm) to act Fig. 22. SEM micrograph of a Ge island (100 nm) as an inert marker between coupling layers in the on an Rh film (20 nm) annealed at 600°C for 15 Ir-Ge system, the direction of atomic mobility was minutes successfully monitored during the initial stages of the reaction. In the marker samples, IrGe and

Ir4Ge5 were again found to coexist from the first Energy, MeV stages of reaction. The movement of the marker 1.2 1.3 1.4 1.5 1.6 1.7 1.8 0.4 indicated that Ge was the sole moving species Rh during Ir4Ge5 formation. It was not certain whether Region A (Ge island) RhGe Ge was also the sole moving species during IrGe 0.3 Region B formation. Region C Rh17Ge22 Region D (Rh film) 0.2 4.1.2 Lateral Diffusion Couples Ge Rh The phases observed to form in lateral diffusion couples of the Ir-Ge system were the same as those Normalised yield 0.1 observed in the thin film study on this system with

the exception of IrGe, i.e. IrGe4, Ir3Ge7 and Ir4Ge5. 0 The phase Ir3Ge7 was seen to stretch across the 300 350 400 450 original island interface at all temperatures. As in

Channel the results of the thin film couples, the phase IrGe4 Fig. 23. Superposition of selected RBS spectra was only observed to nucleate at temperatures from each of the four regions of the Rh-Ge lateral above 800°C. diffusion sample. Rh peak heights of the various The graph in Figure 24 is a plot of the growth in phases and surface positions of Ge and Rh are width with annealing time for a sample annealed indicated at 800°C. In this figure the growth in width of

40 –27.0

Ge + Ir4Ge5 Ir IrGe4 Ir Ge –27.5 IrGe4 3 7 30 –28.0

Ir4Ge5 X b b –28.5 Ea = 1.6 ± 0.1 eV 20 lnK Xb + Xg –29.0

Anneal time, min 10 –29.5

–30.0 0 10.5 11.0 11.5 12.0 12.5 –100 –50 0 50 100 150 200 1/k T eV–1 Reaction length, mm

Fig. 24. Plot of growth width with time of anneal Fig. 25. Arrhenius plot, lnKb versus 1/kbT, showing temperature dependence of Ge diffusion rate for the phases Ir3Ge7 and Ir4Ge5, for a sample with Ge island (250 nm) on an Ir film (25 nm) annealed through Ir4Ge5, yielding an average activation energy of 1.6 ± 0.1 eV at 800°C. Xb refers to the growth width of the Ir3Ge7 region while Xg refers to that for Ir4Ge5

This was because the mechanisms at play during the Ir3Ge7 phase region is labelled as Xb while Ir3Ge7 growth were not the same at 800°C as at the width of the Ir4Ge5 region is labelled as Xg. It temperatures below. At 800°C the Ir3Ge7 observed should be pointed out that the origin in Figure 24 inside the island region was from the decomposition does not correspond to the original island interface of IrGe4 while that outside was formed by the but refers to the interface between the IrGe4 and interaction between Ir4Ge5 and the outward diffusing

Ir3Ge7 regions. The growth curve for the region Ge atoms. Below 800°C the Ir3Ge7 outside was also labelled as Ir3Ge7 was a result of Ir3Ge7 grown formed by the interaction of Ge and Ir4Ge5 but that from Ir4Ge5 and that from the decomposition of inside was observed as a result of exposure as Ge

IrGe4. The growth characteristics observed are was consumed from the source region. On the other parabolic with time. hand, Ir4Ge5 was formed by the interaction of Ge

The growth of the phases Ir3Ge7 and Ir4Ge5 were with the unreacted Ir film, both below and above monitored at temperatures of 700°C, 750°C and 800°C and thus at all three temperatures. 800°C. Different annealing times were chosen to obtain a reasonable range of growth widths at 4.2 Platinum-Germanium System each of the three temperatures. The squares of the growth widths were plotted against the times of 4.2.1 Thin Film Couples annealing at each temperature and the diffusional growth constants, Kb, were obtained from the In our work on the Pt-Ge system, Pt2Ge was the slopes. The logarithms of the diffusional growth first phase formed. The second phase observed constants were then plotted against the reciprocals was Pt3Ge2. The next phase detected was PtGe of the product, kbT, of the Boltzmann constant and the last was PtGe2. The non-congruent phase and the absolute temperatures. The resulting Pt2Ge3 was skipped between the last two phases.

Arrhenius plot for the phase Ir4Ge5 is shown in The marker technique was also applied to the Pt-Ge

Figure 25. The average activation energy, Ea, was system. Pt was found to be the dominant diffusing obtained from the slope of the straight line fit of species during Pt2Ge formation. Some Ge diffusion the Arrhenius plot. The value determined for the was also observed to take place. The atomic diffusion diffusion process in the Ir4Ge5 phase was Ea = 1.6 ratio of Pt to Ge was measured as being about 4 to 1. ± 0.1 eV. Unfortunately the results from the samples 4.2.2 Lateral Diffusion Couples annealed at the temperatures 700°C, 750°C and 800°C could not be used to obtain a value Three phases were observed to form in the lateral of the activation energy for the Ir3Ge7 phase. diffusion couples of the Pt-Ge system: PtGe2 and

Pt2Ge3 inside the original island region and PtGe the phases PtGe2, Pt2Ge3 and PtGe are shown in outside. The Pt2Ge and Pt3Ge2 phases which were Figure 27. observed in the thin film study of the system The activation energies, Ea, were obtained from were absent in the lateral diffusion couples. The the slopes of the straight line fits. The average graph in Figure 26 is a plot of the various growth activation energies determined from the lateral widths against the time of annealing for a sample growth rates of Pt2Ge3 and PtGe were 0.9 ± 0.1 annealed at 550°C. The origin in this figure does eV and 1.3 ± 0.4 eV, respectively. The activation not correspond to the original island interface but energy corresponding to the apparent lateral refers to the interface between the regions labelled growth rate of the PtGe2 region was 1.5 ± 0.2 eV. as A and B in Figure 12. The growth widths of the phase regions, PtGe2, Pt2Ge3 and PtGe are labelled 4.3 Palladium-Germanium System as Xa, Xb and Xg respectively. It must be pointed out that whereas the lateral 4.3.1 Thin Film Couples growth of the Pt2Ge3 and PtGe in the regions labelled as C and D in Figure 12 respectively are The only phases observed to form in the thin film due to reaction mechanisms, the growth of region study of the Pd-Ge system were the two congruent

B is due to exposure of PtGe2 by the consumption of phases PdGe and Pd2Ge. overlaying Ge. The growth characteristics observed for all regions are parabolic with time. 4.3.2 Lateral Diffusion Couples The lateral widths, Xa, Xb and Xg, were monitored at the temperatures 450°C, 500°C and 550°C. The two phases, PdGe and Pd2Ge, which were Different annealing times were chosen to obtain a observed to form in the thin film study of the Pd- reasonable range of growth widths at each of these Ge system were also the only two observed in the three temperatures. The squares of the growth lateral diffusion couples. The growth region outside widths were plotted against the times of annealing the original island interface, labelled as region at each temperature and the diffusional growth B in Figure 17, consisted of the phase Pd2Ge. constants, Kβ, were obtained from the slopes. The The original island region, labelled as region A in logarithms of the diffusional growth constants were Figure 17, consisted of PdGe at the bottom while then plotted against the reciprocals of the product, at the top there was unreacted Ge intermingled kbT, of the Boltzmann constant and the absolute with PdGe. There was no region of completely temperatures. The resulting Arrhenius plots for exposed PdGe without intermingled unreacted Ge, it was therefore not possible to obtain data for the growth or exposure rates of the phase PdGe. 6

Ge + 5 PtGe Pt PtGe2 2 Pt2Ge3 PtGe –27 4

–28 X 3 a PtGe X +X a b Ea =1.3±0.4eV –29 Xa +X b +X Anneal time, h g 2 b lnK –30 1 PtGe2 Pt2Ge3 Ea =1.5±0.2eV Ea =0.9±0.1eV 0 –31 –50 0 50 100 150 200 250 300 Reaction length, mm –32 13 14 15 16 17 Fig. 26. Plot of annealing time against reaction –1 1/kb T eV widths for the phases PtGe2, Pt2Ge3 and PtGe in a sample with Ge (145 nm) on Pt (35 nm) at a constant annealing temperature of 550°C. The Fig. 27. Arrhenius plot, lnKb versus 1/kbT, showing temperature dependence of the Ge diffusion rate growth widths of the phase regions, PtGe2, Pt2Ge3 through Pt2Ge3, PtGe and PtGe2 and PtGe are labelled as Xa, Xb and Xg respectively

The growth of the Pd2Ge region was monitored at determined from the plot in Figure 29 was the temperatures 275°C, 300°C and 325°C. The Ea = 1.0 ± 0.1 eV. periods of annealing were chosen so as to obtain a reasonable range of growth widths at each of 4.4 Rhodium-Germanium System the three temperatures; results are presented in Figure 28. Parabolic growth characteristics were 4.4.1 Thin Film Couples observed. Arrhenius plots were obtained from the data presented in Figure 28, in the same way as Our RBS data strongly suggested the formation of explained for the Ir-Ge and Pt-Ge lateral diffusion Rh2Ge as the first phase in the Rh-Ge system but couples. Figure 29 is an Arrhenius plot showing there was no firm evidence of this from the X-ray the temperature dependence of the Ge diffusion data because this phase was too thin to give a good rate through Pd2Ge. The average activation energy X-ray yield. This was a consequence of the fact that the samples had to be kept thin enough to avoid

excessive RBS peak overlap. The Rh2Ge phase 20 appeared to give way to RhGe while Rh17Ge22 was Ge + Pd formed as the last phase. PdGe Pd2Ge

15 4.4.2 Lateral Diffusion Couples

275ºC The phase Rh17Ge22 was observed inside the original island region in lateral diffusion couples of 10 the Rh-Ge system while RhGe grew outside in most 300ºC cases. Under certain conditions, the latter phase Anneal time, h was observed to slowly stretch across the original 5 325ºC interface into the island region. This suggested a

slow decomposition of Rh17Ge22 into RhGe. There were four distinct regions observed in the lateral 0 –10 0 10 20 30 40 50 diffusion couples of the Rh-Ge system; these Reaction length, mm are represented schematically in Figure 30. The relative position of the original island interface Fig. 28. A plot of reaction length against the time and different phases are shown. Knowledge of the of annealing for the phase Pd Ge in Ge (100 nm) 2 position at which the original interface lay was vital on Pd (20 nm) at temperatures 275°C, 300°C and 325°C in the analysis of the reactions taking place between different phase regions. This was particularly so in this system with wide reaction regions and a slight shift of the reaction interface between the –29.0 Rh17Ge22 and RhGe regions with respect to the original interface for different times of annealing. –29.5 Our RBS data from the thin film study of this system strongly suggested the formation of the –30.0 non-congruent phase Rh Ge but this phase was Pd2Ge 2 Ea =1.1±0.1eV not observed in the lateral diffusion couples. The b –30.5 growths of Rh17Ge22 and RhGe in the lateral diffusion lnK

–31.0

Original interface –31.5

Ge + –32.0 Rh17Ge22 Rh Rh17Ge22 RhGe 19.0 19.5 20.0 20.5 21.0 21.5 –1 1/kb T eV

Fig. 29. Arrhenius plot, lnKb versus 1/kbT, showing temperature dependence of Ge diffusion rate Fig. 30. Diagram showing different phase regions observed in lateral diffusion couples of the Rh-Ge through Pd2Ge, yielding an average activation energy of 1.0 ± 0.1 eV system

226 © 2018 Johnson Matthey https://doi.org/10.1595/205651318X696639 Johnson Matthey Technol. Rev., 2018, 62, (2) couples were monitored at the temperatures 450°C, 5. Summary and Conclusion 500°C and 600°C. The temperature range and annealing times were chosen in such a way that The results of our thin film study are summarised in the decomposition of Rh17Ge22 into RhGe was not Table I. Temperatures at which the first reactions significant. Results for carefully chosen annealing were observed to begin are indicated. times at 500°C are presented in Figure 31. The The lateral diffusion samples used in this study growth characteristics were observed to be parabolic. were prepared with the configuration of Ge islands Arrhenius plots obtained from the data shown in evaporated onto metal films. In all systems, the Figure 31 are presented in Figure 32, showing germanide phases were seen to spread out from the temperature dependence of the Ge diffusion the source region in decreasing order of Ge rate through RhGe and Rh17Ge22. An average content. The growth characteristics observed in all activation energy of 1.2 ± 0.3 eV was obtained for phases of the four systems studied were parabolic both Rh17Ge22 and RhGe. with time. This is indicative of a diffusion controlled

80 –25 Ge + Rh Rh17Ge22 Rh17Ge22 RhGe –26 60 RhGe

–27 Ea =1.2±0.3eV b Rh Ge 40 17 22 lnK –28 Ea =1.2±0.3eV

Anneal time, min 20 –29

–30 13 14 15 16 17 0 –1 –50 0 50 100 150 200 1/kb T eV Reaction length, mm Fig. 32. Arrhenius plots, lnKb versus 1/kbT, showing Fig. 31. Plots of annealing time against reaction temperature dependence of Ge diffusion rate length for the phases Rh Ge and RhGe in Ge through Rh17Ge22 and RhGe, yielding an average 17 22 (100 nm) on Rh (20 nm) at 500°C activation energy of 1.2 ± 0.3 eV for both phases

Table I Summary of the Phase Formation Sequence Results for the Four Systems Studied, with the Temperatures at which the First Reactions were Observed to Start

Temperature at which System Phase formation sequence first reaction begins

1st IrGe and Ir4Ge5 (co-existing) Ir-Ge 2nd Ir3Ge7 350°C 3rd IrGe4

1st Pt2Ge 2nd Pt Ge Pt-Ge 3 2 190°C 3rd PtGe 4th PtGe2 1st Pd Ge Pd-Ge 2 100°C 2nd PdGe

1st Rh2Ge Rh-Ge 2nd RhGe 280°C 3rd Rh17Ge22

227 © 2018 Johnson Matthey https://doi.org/10.1595/205651318X696639 Johnson Matthey Technol. Rev., 2018, 62, (2) process which, as modelled by Kidson (37), results insight and to make comprehensive suggestions on in parabolic growth even in multiphase systems. the ideally suited choice of pgm/Ge combinations Table II gives a summary of the various phases for particular applications in semiconductor observed in the lateral diffusion couples of each technology applications such as gates for metal system and the corresponding activation energies semiconductor field-effect transistors, solar cells obtained. and detectors.

The magnitudes of the activation energies, Ea, calculated for all phases, as shown in Table II, Acknowledgements suggest that the lateral diffusion reactions in all four systems were not driven by surface diffusion The authors wish to thank the University of but rather by diffusion through the interior of the Cape Town, the South African National Research lateral diffusion couples; typical values for surface Foundation and the University Science, Humanities diffusion being around 0.6 eV (38). and Engineering Partnerships in Africa (USHEPIA) In the current design and processing of transistors for financial assistance. They also wish to thank the the contact material should exhibit low sheet and Materials Research Group at the iThemba Labs at contact resistances, form at a low temperature Faure, South Africa, for the use of their facilities, and be stable over a wide temperature range. A Miranda Waldron in the Electron Microscope unit at previous systematic study of the thermally induced the University of Cape Town, Ms Terry Davies of the reaction of a large number of transition metals with X-ray unit in the Geological Science Department, Ge substrates revealed that NiGe and PdGe are the University of Cape Town. most promising candidates when taking the above requirements into account (7, 8). From Table I we References see that Pd germanides form at lower temperatures than any of the other pgm germanides. Pt 1. C. O. Chui, H. Kim, D. Chi, B. B. Triplett, P. C. germanides were also found to be promising McIntyre and K. C. Saraswat, ‘A Sub-400/spl deg/C candidates for microelectronic applications (7). Germanium MOSFET Technology with High-/spl There has been very little work carried out on Kappa/Dielectric and Metal Gate’, International the electrical properties of Ir/Ge junctions (32). Electron Devices Meeting, San Francisco, USA, 8th–11th December, 2002, Institute of Electrical Our survey of previous work in this research field and Electronics Engineers Inc, Piscataway, USA, showed no evidence of any systematic study of the pp. 437–440 electrical properties of Rh/Ge junctions. Our work 2. A. Ritenour, S. Yu, M. L. Lee, N. Lu, W. Bai, A. Pitera, has looked at several aspects of the interfacial E. A. Fitzgerald, D. L. Kwong and D. A. Antoniadis, phase growth and inter-diffusion kinetics at ‘Epitaxial Strained Germanium p-MOSFETs pgm/Ge junctions. However, more work needs to with HfO/sub 2/Gate Dielectric and TaN Gate be carried out regarding the electrical properties of Electrode’, International Electron Devices Meeting, these junctions. Our current results could then be Washington, DC, USA, 8th–10th December, 2003, used in conjunction with the results based on the Institute of Electrical and Electronics Engineers study of electrical properties in order to draw more Inc, Piscataway, USA, pp. 18.2.1–18.2.4

Table II Summary of the Germanide Phases Seen to Spread Out from the Source Region in Their Decreasing Order of Germanium Content During Lateral Diffusion, with Corresponding Activation Energies Obtained

System Phases observed during lateral diffusion and activation energies obtained

IrGe Ir Ge Ir Ge Ir-Ge 4 3 7 4 5 Ea = 1.6 ± 0.1 eV

PtGe2 Pt Ge PtGe Pt-Ge 2 3 Ea = 1.5 ± 0.2 eV Ea = 0.9 ± 0.1 eV Ea = 1.3 ± 0.4 eV PdGe Pd Ge Pd-Ge 2 Ea = 1.0 ± 0.1 eV Rh Ge RhGe Rh-Ge 17 22 Ea = 1.2 ± 0.3 eV Ea = 1.2 ± 0.3 eV

3. H. Shang, H. Okorn-Schmidt, K. K. Chan, M. Copel, 18. J. C. Liu, J. W. Mayer and J. C. Barbour, J. Appl. J. A. Ott, P. M. Kozlowski, S. E. Steen, S. A. Cordes, Phys., 1988, 64, (2), 651 H.-S. P. Wong, E. C. Jones and W. E. Haensch, ‘High 19. J. C. Liu, J. W. Mayer and J. C. Barbour, J. Appl. Mobility p-Channel Germanium MOSFETs with a Phys., 1988, 64, (2), 656 Thin Ge Oxynitride Gate Dielectric’, International 20. J. C. Liu and J. W. Mayer, J. Mater. Res., 1990, 5, Electron Devices Meeting, San Francisco, USA, (2), 334 8th–11th December, 2002, Institute of Electrical and Electronics Engineers Inc, Piscataway, USA, 21. P. J. Ding, R. Talevi, W. A. Lanford, S. Hymes and pp. 441–444 S. P. Murarka, Nucl. Instr. Meth. Phys. Res. Sect. B: Beam Int. Mater. Atoms, 1994, (1–4), 85, 167 4. C. O. Chui, S. Ramanathan, B. B. Triplett, P. C. 22. H. S. de Waal, “The Effect of Diffusion Barriers, McIntyre and K. C. Saraswat, IEEE Electron Dev. Stress and Lateral Diffusion on Thin-Film Phase Lett., 2002, 23, (8), 473 Formation”, PhD thesis, University of Stellenbosch, 5. C. Claeys and E. Simoen, “Germanium-Based South Africa, 1999 Technologies: From Materials to Devices”, Elsevier 23. A. Saedi, B. Poelsema and H. J. W. Zandvliet, Surf. BV, Oxford, UK, 2007, 449 pp Sci., 2011, 605, (5–6), 507 6. D. P. Brunco, B. De Jaeger, G. Eneman, J. Mitard, 24. E. Hökelek and G. Y. Robinson, Solid State G. Hellings, A. Satta, V. Terzieva, L. Souriau, F. E. Electron., 1981, 24, (2), 99 Leys, G. Pourtois, M. Houssa, G. Winderickx, E. Vrancken, S. Sioncke, K. Opsomer, G. Nicholas, 25. E. H. Rhoderick and R. H. Williams, “Metal- M. Caymax, A. Stesmans, J. Van Steenbergen, Semiconductor Contacts”, 2nd Edn., Monographs P. W. Mertens, M. Meuris and M. M. Heyns, J. in Electrical and Electronic Engineering, Vol. 19, Electrochem. Soc., 2008, 155, (7), H552 Clarendon Press, Oxford, UK, 1988, 252 pp 26. G. A. Baraff and M. Schlüter, Phys. Rev. B, 1986, 7. S. Gaudet, C. Detavernier, A. J. Kellock, P. 33, (10–15), 7346 Desjardins and C. Lavoie, J. Vac. Sci. Technol. A, 2006, 24, (3), 474 27. S. Asubay, Ö. Güllü and A. Türüt, Appl. Surf. Sci., 2008, 254, (11), 3558 8. J. A. Kittl, K. Opsomer, C. Torregiani, C. Demeurisse, S. Mertens, D. P. Brunco, M. J. H. Van 28. C. Henkel, S. Abermann, O. Bethge, G. Pozzovivo, Dal and A. Lauwers, Mater. Sci. Eng.: B, 2008, S. Puchner, H. Hutter and E. Bertagnolli, J. 154–155, 144 Electrochem. Soc., 2010, 157, (8), H815 29. A. Chawanda, C. Nyamhere, F. D. Auret, W. 9. R. S. Nemutudi, C. M. Comrie and C. L. Churms, Mtangi, M. Diale and J. M. Nel, J. Alloys Compd., Thin Solid Films, 2000, 358, (1–2), 270 2010, 492, (1–2), 649 10. S.-L. Zhang and M. Östling, Crit. Rev. Solid State 30. H. B. Yao, C. C. Tan, S. L. Liew, C. T. Chua, C. K. Mater. Sci., 2003, 28, (1), 1 Chua, R. Li, R. T. P. Lee, S. J. Lee and D. Z. Chi, 11. L. R. Zheng, L. S. Hung, J. W. Mayer, G. Majni and ‘Material and Electrical Characterization of Ni- and G. Ottaviani, Appl. Phys. Lett., 1982, 41, (7), 646 Pt-Germanides for p-channel Germanium Schottky 12. L. R. Zheng, L. S. Hung and J. W. Mayer, J. Vac. Source/Drain Transistors’, Sixth International Sci. Technol. A, 1983, 1, (2), 758 Workshop on Junction Technology, Shanghai, China, 15th–16th May, 2006, Institute of Electrical 13. L. R. Zheng, L. S. Hung and J. W. Mayer, Thin Solid and Electronics Engineers Inc, Piscataway, USA, Films, 1983, 104, (1–2), 207 pp. 164–169 14. S. H. Chen, L. R. Zheng, J. C. Barbour, E. C. Zingu, 31. A. Chawanda, C. Nyamhere, F. D. Auret, W. Mtangi, L. S. Hung, C. B. Carter and J. W. Mayer, Mater. T. T. Hlatshwayo, M. Diale and J. M. Nel, Physica B, Lett., 1984, 2, (6), 469 2009, 404, (22), 4482 15. B. Blanpain, J. W. Mayer, J. C. Liu and K. N. Tu, J. 32. A. Chawanda, S. M. M. Coelho, F. D. Auret, W. Appl. Phys., 1990, 68, (7), 3259 Mtangi, C. Nyamhere, J. M. Nel and M. Diale, J. 16. B. Blanpain, J. W. Mayer, J. C. Liu and K. N. Tu, Alloys Compd., 2012, 513, 44 Phys. Rev. Lett., 1990, 64, (22–28), 2671 33. C.-Y. Peng, Y.-H. Yang, C.-M. Lin, Y.-J. Yang, C. 17. B. Blanpain, ‘Lateral Diffusion Couples and F. Huang and C. W. Liu, ‘Process Strain Induced Their Contribution to Understanding Thin Film by Nickel Germanide on (100) Ge Substrate’, Reactions’, in “Crucial Issues in Semiconductor 9th International Conference on Solid-State and Materials and Processing Technologies”, eds. Integrated-Circuit Technology, Beijing, China, S. Coffa, F. Priolo, E. Rimini and J. M. Poate, 20th–23rd October, 2008, Institute of Electrical Springer Science+Business Media, Dordrecht, The and Electronics Engineers Inc, Piscataway, USA, Netherlands, 1992, pp 421–425 pp. 681–683

34. J. Hållstedt, M. Blomqvist, P. O. Å. Persson, L. 36. L. R. Doolittle, Nucl. Instr. Meth. Phys. Res. Sect. Hultman and H. H. Radamson, J. Appl. Phys., B: Beam Int. Mater. Atoms, 1986, 15, (1–6), 227 2004, 95, (5), 2397 37. G. V. Kidson, J. Nucl. Mater., 1961, 3, (1), 21 35. A. Thanailakis and D. C. Northrop, Solid State 38. L. Eckertová, “Physics of Thin Films”, 2nd Edn., Electron., 1973, 16, (12), 1383 Plenum Press, New York, USA, 1986

The Authors

Adrian Habanyama is a Senior Professor Craig Comrie is an Lecturer at the Copperbelt Emeritus Associate Professor University, Zambia. His research at the University of Cape Town, areas are solid-state physics South Africa. His research areas and nanotechnology. are solid-state physics and nanotechnology.

www.technology.matthey.com

Toward Platinum Group Metal-Free Catalysts for Hydrogen/Air Proton-Exchange Membrane Fuel Cells Catalyst activity in platinum-free substitute cathode and anode materials

Frédéric Jaouen*, Deborah Jones ultralow Pt content electrodes have demonstrated Institut Charles Gerhardt Montpellier, CNRS - good performance, but alternative non-pgm Université Montpellier - ENSCM, Place Eugene anode catalysts are desirable to increase fuel cell

Bataillon, 34095 Montpellier cedex 5, France robustness, decrease the H2 purity requirements

and ease the transition from H2 derived from # Nathan Coutard, Vincent Artero natural gas to H2 produced from water and Laboratoire de Chimie et Biologie des Métaux, renewable energy sources. Université Grenoble Alpes, CNRS, CEA, 17 rue des Martyrs, 38054 Grenoble cedex 9, France 1. Introduction Peter Strasser 1.1 Opportunities for PEMFCs Technische Universität Berlin, Institut für Chemie, Strasse des 17. Juni 124, 10623 Electrochemical devices, and PEMFCs in particular, Berlin, Germany are under intense development for a cleaner and more efficient use of energy, including the use Anthony Kucernak of renewable electricity for transportation (1). Department of Chemistry, Imperial College While rechargeable batteries directly store and

London, South Kensington Campus, London, discharge electric power, H2/air PEMFCs convert

SW7 2AZ, UK the chemical energy of H2 into electricity and

heat. Today, the lion’s share of H2 production

Email: *[email protected], comes from natural gas. In the future, H2 could #[email protected] however be produced at competitive price and with lower environmental impact from water and renewable energy. The power-to-gas and gas The status, concepts and challenges toward to-power consecutive conversions needed to catalysts free of platinum group metal (pgm) use H2 as an energy carrier might seem a priori elements for proton-exchange membrane fuel less attractive than the reversible storage of cells (PEMFC) are reviewed. Due to the limited electricity in batteries due to its higher complexity reserves of noble metals in the Earth’s crust, a and lower roundtrip energy efficiency. However, major challenge for the worldwide development technical requirements and customer acceptance of PEMFC technology is to replace Pt with pgm can favour fuel cells over batteries for certain free catalysts with sufficient activity and stability. markets. In the automotive sector for example, The priority target is the substitution of cathode lithium-ion battery electric vehicles (BEV) offer catalysts (oxygen reduction) that account for shorter driving range than internal combustion more than 80% of pgms in current PEMFCs. engine (ICE) vehicles. In contrast, H2-powered

Regarding hydrogen oxidation at the anode, PEMFC vehicles have already demonstrated a

231 © 2018 Johnson Matthey https://doi.org/10.1595/205651318X696828 Johnson Matthey Technol. Rev., 2018, 62, (2) driving range of 500 km and refuelling time of being reached in an unconfined space. BEVs are less than 4 min (2, 3). Whether battery or fuel distinctly associated with low driving range and cell, the electrification of the automobile could the difficulty in knowing the instantaneous state- significantly cut carbon dioxide emissions (25% of of-charge of a battery (5). Considering systems

CO2 emissions originated from the transport sector able to deliver the same electric power (kW), the in 2010 (4)) and reduce our reliance on fossil fuels. weight of a H2-tank/PEMFC stack system becomes

Otherwise, CO2 emissions from road transportation systematically lower than that of a rechargeable Li will continue increasing over the next decades, due battery above a certain threshold amount of energy to an increased global fleet of vehicles (Figure 1). (kWh). While the exact threshold value depends

H2 is often perceived as more dangerous than on the technology status and also on the mass of gasoline-fuelled ICE. However the low density of the car, threshold values of 20–30 kWh have been

H2 naturally prevents the detonation limit from estimated (1, 2). As a comparison, 10 kWh is the

(a) 2200 2000 1800 1600 1400 1200 1000 800 Cars, millions 600 400 200 0

2005 2010 2015 2020 2025 2030 2035 2040 2045 2050 Gasoline type ICEV Diesel type ICEV Gasoline type hybrid Diesel type hybrid Liquid fuel plug-in Gas fuel ICEV Gas fuel hybrid Hydrogen hybrid Hydrogen fuel cell Electric vehicle (b) 4 –1 3

2 emissions, Gt year 2 1 CO

2005 2010 2015 2020 2025 2030 2035 2040 2045 2050

Africa Asia Brazil Canada China EEUR WEUR FSU India LAM MEA Mexico Pacific Russia USA

Fig. 1. (a) Assumed technology mix for cars until 2050 (million vehicles); (b) predicted CO2 emissions –1 from cars (gigatonnes CO2 year ). Reproduced from (4). The model assumes that global car fleet remains dominated by gasoline and diesel ICE (78%), a significant share of hybrid vehicles (18%) and a small fraction of EVs (4%). Key: ICEV = internal combustion engine vehicle; EEUR = Eastern Europe; WEUR = Western Europe; FSU = Former Soviet Union; LAM = Latin America; MEA = Middle East and Africa. Used by permission of the World Energy Council, London, www.worldenergy.org

232 © 2018 Johnson Matthey https://doi.org/10.1595/205651318X696828 Johnson Matthey Technol. Rev., 2018, 62, (2) electrical energy needed to move a mid-size car more energy efficient than fuel cells (2). Pure BEVs over 100 km (1). and fuel cell electric vehicles (FCEV) may therefore While a BEV with longer driving range needs a target different segments of the automotive market proportionally higher mass of battery to store more (1). Other important applications of PEMFCs are as energy, a H2-tank/PEMFC system only needs a larger backup power systems, and for combined heat and tank to store more energy, with low associated power (CHP) (6). For backup power, the chemical incremental mass. This simple fact favours energy contained in H2 can be stored for years

H2-PEMFC systems for transport applications without ‘discharge’. For CHP, the cell voltage of the requiring long driving range (Figure 2). However, PEMFC can be advantageously controlled during when a short driving range is acceptable, BEVs are operation to tune the electric and thermal power

(a) NiMH battery Li-ion battery PbA battery 800

700

600 Fuel cell + 35 MPa hydrogen tanks 500

400

300

200 Fuel cell + 70 MPa 100 hydrogen tanks

Energy storage system volume, l system volume, Energy storage 0 0 100 200 300 400 500 600 Range, km 300 miles (480 km)

(b) PbA battery EV NiMH battery EV

Li-ion battery EV 4000

3500

3000

2500

2000

1500

Vehicle test mass, kg Vehicle 1000

500 Fuel cell electric vehicle 0 0 100 200 300 400 500 600 Range, km 300 miles (480 km)

Fig. 2. (a) Calculated energy storage system volume for an electric vehicle (EV) equipped either with a compressed-H2 storage/PEMFC system or with various battery technologies as a function of the vehicle range; (b) same comparison but for the estimated mass of the energy storage and conversion system. The power trains are adjusted to provide a zero to 97 km h–1 acceleration time of 10 s. Reproduced from (2) with permission from Elsevier

233 © 2018 Johnson Matthey https://doi.org/10.1595/205651318X696828 Johnson Matthey Technol. Rev., 2018, 62, (2) outputs. About 140,000 PEMFC units for CHP have above a certain threshold of volume production. been installed in Japan between 2009 and 2015 This may impede reaching (or staying at, in the and this application is also taking off in South case of millions of units produced per year) the –1 Korea and Europe (6). cost target of US$20 kWelectric for an automotive The remaining technical challenges for PEMFCs PEMFC stack (8). Reaching but also staying at this are lowered cost and improved durability, with cost is necessary for PEMFCs to be cost competitive the electrode catalysts and membrane ionomer with ICE and affordable to the wide public. materials being at the heart of the vital functions The particular importance of the cathode catalyst and lifetime of a PEMFC. While at low production (oxygen reduction reaction (ORR) catalysis) is volumes, the global demand of Pt for PEMFCs is not introduced in Section 2, while the anode catalyst high and the share of the Pt cost in the PEMFC stack (hydrogen oxidation reaction (HOR) catalysis) cost is not excessive, both those scalars would opportunities are discussed in Section 3. This review increase dramatically in the case of a massive gives a focused account of recent achievements deployment of PEMFCs. The share of Pt catalyst to and discusses the needs and possibilities toward fuel cell stack cost would increase due to decreased the rational design of improved non-pgm cathode cost for all other components through economies of layers and opportunities in non-pgm anode scale (Figure 3) (7). Assuming a constant Pt price catalysts. Comprehensive reviews and book over time and unchanged Pt mass per rated power chapters on pyrolysed metal-nitrogen-carbon of PEMFC (kWelectric), the stack cost would reach (Me-N-C, where ‘Me’ is a transition metal) cathode a lower-value plateau, nearly incompressible upon catalysts and inorganic non-pgm anode catalysts further increased production volumes. In parallel, for PEMFCs can be found elsewhere (9–12). the ratio of Pt-to-stack cost would increase to ca. 50% (Figure 3). It must be noted that this 2. Non-pgm Cathode Catalysts percentage is a conservative estimation, reached assuming a constant Pt price. It is likely, however, 2.1 The Need for Non-pgm Cathode that the scarcity of Pt combined with increased Catalysts demand would lead to increased price. This could possibly lead to an increased stack cost (instead of a Due to the much slower kinetics of the ORR than levelled-off cost above ca. 100,000 units per year), those of the HOR on Pt surfaces, 80–90% of Pt in

160 Share of catalyst to PEMFC stack cost, % 40 –1 140 electric 120 30 100

80 20

40 10

PEMFC stack cost, US$ kW 20 Stack cost Share of catalyst ink to total cost 0 0 1000 10,000 100,000 1,000,000 Annual volume production, units per year

Fig. 3. Predicted automotive PEMFC stack cost as a function of annual volume production. The predicted –2 cost is based on the 2016 automotive technology, including in particular 0.21 gPt kWel (0.116 mgPt cm –2 at cathode, 0.018 mgPt cm at anode). Graph drawn from data in (7), data used with permission from Strategic Analysis Inc, USA

H2-fuelled PEMFCs is currently positioned at the conductive carbon matrix distinguishes them from cathode (13). In this context, the development of Pt these well-defined organic compounds, and is based catalysts with higher ORR activity normalised critical to reach high current densities in PEMFC. per mass Pt is under intense investigation (14–16), with the practical objective of reducing the Pt content 2.2 Concepts for the Design of Me-N-C in automotive PEMFC stacks down to the current Catalysts and Catalyst Layers pgm content in catalytic converters of ICE-powered automobiles (1–4 g depending on vehicle size, engine Research and development (R&D) efforts and types and local regulations on air quality). The very interest in non-pgm catalysts for the ORR in acid high ORR activity of some Pt nanostructures (such medium have never been so intense, as witnessed as Pt nanoframes or jagged nanowires) recently by a rising number of research groups working on observed in rotating-disk-electrode measurements the topic but also an increasing number of R&D remain to be transposed to the PEMFC environment, funding calls dedicated to this class of catalysts, and their long-term durability proven. In addition, a for example in Europe from the Fuel Cells and very high turnover frequency for the ORR on a small Hydrogen 2 Joint Undertaking (FCH 2 JU) (26), number of active sites in the cathode, while being a and in the USA from the US Department of Energy dream for electrocatalysis scientists, may turn out (DOE) (27). In the USA, R&D efforts in non to be an issue for membrane electrode assembly pgm catalysts for PEMFC are now organised by

(MEA) developers due to enhanced local O2 diffusion the Electrocatalysis Consortium (ElectroCat). A barriers in a PEMFC cathode with low volumetric broader incentive to reduce the reliance on Critical density of active sites (17, 18). The replacement of Raw Materials and increase their recycling was also Pt-based catalysts with pgm-free cathode catalysts initiated by the European Union (EU) for various is considered a holy grail. Ideally, highly active and new energy technologies (28). Two companies are durable pgm-free cathode catalysts could replace currently engaged in the development of Me-N-C Pt-based cathodes in PEMFCs designed for all and other non-pgm catalysts for PEMFCs (Pajarito types of markets. Alternatively, pgm-free cathodes Powder, USA and Nisshinbo Holdings, Japan). In not meeting the stringent durability and power September 2017, Ballard, Canada, and Nisshinbo performance targets of the automotive industry may Holdings announced the first portable PEMFC (30 W) be competitive for other applications (for example, commercialised with a non-pgm cathode catalyst backup power, mobile applications or CHP) (16). (29). This interest in Me-N-C catalysts and closer The pgm-free materials that have hitherto shift towards application is the result of important displayed the highest ORR activity when tested progress in the field since 2009, with breakthroughs in aqueous acid medium or in single-cell PEMFC achieved in the ORR activity reached at high cell are pyrolysed Me-N-C catalysts, with the metal voltage in single-cell PEMFC (30), power density at being iron or cobalt (19, 20). Me-N-C catalysts cell voltage experienced during practical operation have been prepared from numerous precursors of (0.5–0.7 V range) (22, 31), and understanding metal, nitrogen and carbon via the optimisation of of the nature of the active sites (21, 22, 23, 25) the precursor ratio, metal content and pyrolysis and how they catalyse the ORR at atomistic level conditions that must be adapted to each system (24, 32, 33). Table I gives examples of synthesis of precursors (10). The nature of the metal- strategies and corresponding ORR current density based active sites in such Me-N-C catalysts is measured in PEMFC at 0.9 V (ORR ‘activity’) for fundamentally different from that in Pt-based some of the most active Fe-N-C catalysts to date. catalysts (compare Figures 4(b), 4(c) and 4(d) The cathode catalyst loading used in each work is –2 to Figure 4(a)). The most active sites for ORR also indicated in the second column (mgFe-N-C cm ). –2 in pyrolysed Me-N-C catalysts are, from the most Within a certain range (typically 0.5 mgFe-N-C cm –2 recently established knowledge, single metal-ions to 5 mgFe-N-C cm ), the current density at 0.9 V strongly coordinated with nitrogen ligands (MeNx increases proportionally with cathode catalyst moieties, x being on average 4), and these MeNx loading. Based on this, the current density at –2 moieties are covalently integrated in graphene, 0.9 V expected for a loading of 5 mgFe-N-C cm is or disordered graphene, sheets (Figure 4) (21, indicated in the third column in Table I. It must be

22, 23–25). The local coordination of such MeNx noted that proportionally increased current density moieties resembles the metal-ion coordination in with increased cathode loading is restricted to low phthalocyanine and porphyrin compounds (23), current. At high current density (>200 mA cm–2), but their covalent integration in the electron- the cell performance is also impacted by mass- and

(a) (b)

(111)

5 nm 1 nm

1 nm 1 nm

Fig. 4. The crystallographic ordering of Pt atoms in Pt-based catalysts and the atomically-dispersed nature of Fe and Co atoms in pyrolysed Fe(Co)-N-C catalysts, as revealed by high-resolution scanning transmission electron microscopy (HR-STEM) images: (a) Pt3Ni nanoframe, from (14). Reprinted with permission from AAAS; (b) Co-N-C catalyst obtained via pyrolysis in ammonia of a cobalt salt and graphene, showing atomically dispersed cobalt in the N-doped carbon matrix, reproduced from (21); (c) Fe-N-C catalyst prepared from a ferrous salt, aniline and cyanamide and pyrolysed in N2, from (22). Reprinted with permission from AAAS; (d) Fe-N-C catalyst prepared from a ferrous salt, phenanthroline and ZIF-8, and pyrolysed in argon (CNRS catalyst, transmission electron microscopy (TEM) image provided by Goran Drazic, National Institute of Chemistry, Slovenia)

Table I State-of-the-Art Oxygen Reduction Reaction Activity of Iron-Nitrogen-Carbon Cathode Catalysts in Single Cell PEMFC Measured under Pure Oxygen and Hydrogena

Current Loading, Expected Back Cathode catalyst Ref. density at 0.9 mgFe-N-C current density pressure, description V, mA cm–2 cm–2 at 0.9 V at 5 bar –2 mgFe-N-C cm , mA cm–2

5 4 6 1 Polyimide nanoparticles (NPs) (35) (60 nm), multipyrolysis and leaching steps, final pyrolysis in NH3 7 1 35 1 Fe(II)salt + phen + ZIF-8, (36) pyrolysis in argon at 1050°C then in NH3 at 950°C 3 1.5 10 1 Fe-porphyrin + Co-porphyrin + (37) silica template, pyrolysed in N2 at 1000°C, HF leaching

Continued

Current Loading, Expected Back Cathode catalyst Ref. density at 0.9 mgFe-N-C current density pressure, description V, mA cm–2 cm–2 at 0.9 V at 5 bar –2 mgFe-N-C cm , mA cm–2

10 2 25 1 ZIF-8 + Fe electrospun with (34) polyacrylonitrile (PAN) and poly(methyl methacrylate (PMMA), pyrolysed at 1000°C in Ar then 900°C in NH3, acid leached

4 4 5 0.5 Fe(II) salt + nicarbazin + silica (38) template, pyrolysis in N2, HF leach, pyrolysis in NH3

5 1 25 1 ZIF [Zn(eIm)2 rho] + Fe(II) + (39) phen, single pyrolysis in NH3 at 950°C aPEMFC conditions: 80°C, 100% RH feed gases, Pt/C anode

charge-transport across the cathode active layer While the target in the early stage of non-pgm (Figure 5). Oxygen transport limitation in a thick catalyst development exclusively focused on the cathode is particularly exacerbated when it is fed activity at high potential (volumetric activity or with air (31, 34), which is the case for almost all current density at 0.8 V or 0.9 V), the activity target PEMFC applications. is now accompanied by a power performance target

(a) (b) Increased (c) volumetric activity H+

– O2, e

10 mm 100 mm

(e) (d)

A B C

Improved transport properties Energy D E

Log(J)

Fig. 5. Interplay between Fe-N-C cathode volumetric activity, thickness and transport properties in determining the cathode performance at low and high current densities: (a) thin Fe-N-C cathode –2 (representative for 0.4 mgFe-N-C cm ) with reference volumetric activity; (b) thick Fe-N-C cathode –2 (representative for 4.0 mgFe-N-C cm ) with reference volumetric activity; (c) thick Fe-N-C cathode with enhanced volumetric activity; (d) thick Fe-N-C cathode with reference volumetric activity but enhanced mass-transport properties; (e) next generation Fe-N-C cathode with improved volumetric activity and mass transport. The graph shows schemed Tafel plot presentations (potential vs. logarithm of the current density) of the cathode polarisation curves. The linear part corresponds to the current density region where the cathode is only controlled by ORR electrokinetics (no limitation by transport of O2, protons and electrons)

(PEMFC operating point of typically 0.6–0.7 V) the electrode thickness is ca. 10 µm). The layer (see Table II). The volumetric-activity concept thickness for Pt/C and Fe-N-C catalysts alike is was specifically defined for non-pgm catalysts by governed by the carbon loading, with apparent 3 the General Motors Fuel Cell group, USA, in 2003 density of 0.37–0.40 gcarbon from catalyst per cm (40). The volumetric activity is defined as the areal of electrode usually observed (30, 31, 40)). We current density of a non-pgm cathode normalised can extract from this a rule-of-thumb of 25 µm by the cathode thickness (A cm–3, reported at electrode-thickness increment per 1 mg cm–2 of 0.8 V or 0.9 V). As is valid for the Fe-N-C loading carbon material from the catalyst. Future efforts effect within certain conditions, the proportionality should thus focus on improving both the volumetric between current density of the cathode (when it is activity of Me-N-C catalysts and the mass-transport controlled by electrokinetics) and cathode thickness properties of Me-N-C layers (Figures 5(c) can be assumed to be valid within a certain range and 5(d), respectively), to ultimately combine of thickness, Equation (i): advances in activity and transport properties to compete with Pt-based cathode layers on the whole J (at 0.8 or 0.9 V) = jV t (i) range of current density (Figure 5(e), curve E). –2 with J the current density (A cm ), jv the volumetric For Pt-based catalysts in contrast, there is no activity (in A cm–3 at 0.8 V or 0.9 V) and t the incentive to increase performance by increasing the cathode thickness (cm). As one can see, a possible Pt loading at the cathode, because Pt is expensive. approach to increase the current density at high The trend is opposite, with attempts to reach potential may consist of increasing the thickness of break-even performance (same current density at the cathode layer (Figures 5(a) and 5(b)). From a given cell voltage) but with lower Pt loading than a cost perspective, this is feasible, but it faces today. The key parameter for pgm-based catalysts practical limitations due to increasing average path is the mass activity, iM. Equation (ii): lengths for O2, protons and electrons in order to J (at 0.9 V) = iM L (ii) reach the active sites. The opposite directions of the –2 O2 and protons flow can lead to particularly severe with J the current density (A cm ), iM the mass –1 mass-transport limitations if they co-act to form activity (in mA mgPt at 0.9 V) and L the Pt loading –2 gradients of O2 concentration and electrochemical at the cathode (mgPt cm ). Reporting the activity potential, respectively (schemed as fading arrows of pgm-based catalysts as a mass activity (A –1 in Figure 5). With current MEA technology, 100 µm gpgm ) and of non-pgm catalysts as a volumetric is considered the upper realistic thickness limit for activity (A cm–3 cathode) arises therefore from the a Fe-N-C layer. This is already 10 times thicker different nature of the main limitations (cost for than usual Pt-based cathode layers (with a 50% pgm-based cathodes and performance for non –2 Pt/C catalyst and a cathode loading of 0.4 mgPt cm , pgm cathodes) (16). The targets set for non-pgm

Table II Non-pgm Cathode Current Density Targets at 0.9 V (‘Activity’) and 0.7 V (‘Power Performance’) in Single-Cell PEMFC (8)a

Cathode feed FCH 2 JU targets US DOE targets

Current density Current density Experimental Experimental at 0.9 V cell at 0.9 V cell conditions conditions voltage voltage

O2 H /O , 0.5 bar H /O , 1 bar gauge/1 2 2 2 2 gauge/0.5 bar bar gauge, 80°C, 100% 75 mA cm–2 44 mA cm–2 (8) gauge, 80°C, 100% RH RH

Current density Experimental at 0.7 V cell – – conditions voltage Air H2/air, 2.5 bar gauge/2.3 bar gauge, 600 mA cm–2 – – 80°C, 50% RH/30% RH a –1 –2 US DOE target is equivalent to a Pt-cathode with mass activity of 44 A g Pt at 0.9 V and a loading of 0.1 mgPt cm

catalysts for automotive application in the recent with the cathode fed with fully humidified 2 O FCH 2 JU call of 2017 (26) and that of the US (Figures 6(a)–6(c)) or fully humidified air DOE Office of Energy Efficiency & Renewable (Figure 6(d)). At 0.6 V, the current density –2 Energy (EERE) (27) are reported in Table II. By reaches 1.0–1.2 A cm in pure O2 and the peak comparing Table I and Table II, one can see that power density is nearly 1 W cm–2 at around today’s most active Fe-N-C catalysts could reach, 0.4–0.5 V (Figures 5(a), 5(b) and 5(c)) (22, 31, –2 at a cathode loading of 5 mgFe-N-C cm , ca. one- 34). The use of a zinc-based zeolitic imidazolate third to half of the current density targets at 0.9 V metal-organic-framework (ZIF-8) as sacrificial set for the next generation of pgm-free catalysts. precursor of carbon and nitrogen resulted in 2011 While promising, such cathode layers should in a more open structure and higher accessibility also show mass-transport properties appropriate to O2 of the FeNx sites formed during pyrolysis to operation at high current density. Figure 6 (Figure 5(a)) (31). Since then, ZIF-8 has been shows representative examples of the best extensively studied for the preparation of highly power performance single-cell PEMFC obtained microporous Fe-N-C and also N-C materials (41). with Fe-N-C cathodes and Pt-based anodes, Other metal organic frameworks (MOFs) have

(a) (b) 1.0 1.0

0.8 0.8 Pt/C 0.6 0.6 20% Pt/C Fe-N-C (Black Pearls)Fe-N-C (ZIF-8)

0.4 V Voltage, 0.4 Fe/N/CF Cell voltage, V Cell voltage, 0.2 0.2 Fe/N/KB

0 0 0 0.5 1.0 1.5 2.0 2.5 0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 –2 Current density, A cm Current, A cm–2 (c) (d) 1.0 1.0 1.0 1.0 35% Nafion 50% Nafion Power density, W cm Power density, W cm 0.8 0.8 60% Nafion 0.8 0.8 –2 Pt/C(0.1mg Pt cm )

0.6 0.6 0.6 0.6 Voltage, V Voltage, Voltage, V Voltage, 0.4 0.4 0.4 0.4 pair 1.0 bar

p 0.3 bar –2 O2 –2 0.2 0.2 0.2 p 1.0 bar 0.2 O2 p 2.0 bar O2 0 0 0 0 0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 0 0.2 0.4 0.6 0.8 1.0 1.2 Current density, A cm–2 Current density, A cm–2

Fig. 6. Examples of state-of-the-art power performance obtained with Fe-N-C cathodes in single-cell PEMFC: (a) H2/O2 polarisation curve with a cathode prepared from ferrous acetate, phenanthroline and –2 ZIF-8, pyrolysis in Ar then in NH3 (blue curve), 3.9 mgFe-N-C cm cathode, Pt-based anode, 80°C, 2 bar –2 gas pressure on each side (0.5 bar is water vapour), the green curve is for a 0.4 mgPt cm cathode (31); (b) H2/O2 polarisation curve with a cathode prepared by co-electrospinning Fe(phen)3 complex, ZIF-8 and –2 –2 polyacrylonitrile (labelled Fe/N/CF), pyrolysis in Ar then NH3 (34), 3 mgFe-N-C cm cathode, 0.3 mgPt cm anode, 2 bar gas pressure on each side, 80°C, 100% relative humidity (RH); (c) H2/O2 polarisation curve with a cathode prepared from ferric salt, aniline and cyanamide, first pyrolysis in 2N -acid-leaching-second –2 –2 pyrolysis in N2, 4 mgFe-N-C cm cathode, 2 mgPt cm anode, 80°C, dry O2 partial pressure 0.3 bar, 1 bar or 2 bar (22); (d) H2/air polarisation curve, dry air partial pressure 1 bar, 100% RH, otherwise same conditions as for (c). From (22). Reprinted with permission from AAAS

239 © 2018 Johnson Matthey https://doi.org/10.1595/205651318X696828 Johnson Matthey Technol. Rev., 2018, 62, (2) also been investigated, but Zn-based MOFs are are typically about 80–100 μm thick while Pt/C so far the best candidates, and in particular the cathode thickness is ≤20 μm. subcategory of ZIFs (39, 42, 43). The advantage of Another practical advantage of pgm-free Me-N-C such ZIFs is the low boiling point of zinc (907°C). cathode catalysts is their strong resistance to During the pyrolysis at T > 950°C, most Zn poisoning, while Pt-based catalysts suffer from (undesired in final Fe-N-C catalysts) is evacuated severe poisoning from various species, including as volatile products while Fe stays. Acid leaching some gases that can be present at trace amounts in of excess Zn is thus avoidable. While high specific fossil-derived H2 (Figure 7) (50) but also anions, area and high microporous area in particular (pores including chloride anions that are commonly with width ≤2 nm) are important for reaching high encountered in field applications. electrocatalytic activity with Fe-N-C catalysts (22, In summary, further improvement of the power 30, 44), the connection between micropore-hosted performance of Me-N-C cathode layers in PEMFCs

FeNx moieties and the macroporous structure of the can be reached by either increasing the catalyst electrode is an important key for proper accessibility activity (the cathode can then be made thinner, by O2. Mesoporosity can be introduced during the while preserving the ‘apparent’ cathode activity) or catalyst synthesis preparation (for example, with by increasing the reactant transport properties of a silica template approach or pore-forming agents the cathode layer, including long-distance transport (38, 45, 46)), but macroporosity often depends (through the porous cathode) and short-distance on the electrode preparation method as a whole, transport (which can be modulated by catalyst not only on intrinsic catalyst morphology (47). An morphology or agglomerate size). The major original approach to combine the microporosity of experimental efforts have hitherto focused on ZIF-8 derived Fe-N-C catalyst with macroporosity increasing the ORR activity of Me-N-C materials. in the electrode resorted to the electrospinning of Such efforts are still critical to further increase the Fe-doped ZIF-8 with a carrier polymer (34). The activity, selectivity and durability of such catalysts, carrier polymer forms fibrous structures, imparting but work on cathode layer design and catalyst inter-fibre macroporosity in the final electrode morphology is also critical to improve non-pgm structure. Another broad approach for the synthesis cathode behaviour at high current density when of highly active Fe-N-C catalysts has involved the fed with air. use of sacrificial monomers or polymers as C and N sources (48, 49). Figure 5(c) shows the H2/O2 2.3 Deconvoluting Activity of polarisation curve with Fe-N-C cathode prepared Me-N-C Catalysts into Site Density from the pyrolysis of an iron salt and two different and Turnover Frequency monomers, aniline and cyanamide. The two different monomers helped in forming a bimodal If 100–125 µm remains the upper limit of practical porosity, with the addition of cyanamide increasing Me-N-C layer thickness in the future, then further greatly the microporous surface area in the final increasing the current density at 0.9 V will require Fe-N-C catalyst (22). increasing the volumetric activity. This can mainly

While the beginning-of-life H2/O2 polarisation be achieved via increasing either the number of curves of single-cell PEMFC comprising Fe-N-C active sites per unit volume (site density, SD, i.e. cathodes now approach those of Pt/C cathodes, the number of sites that can be electrochemically the performance in H2/air conditions (needed for addressed) or the specific activity for ORR (turnover use in technologically-relevant conditions) is facing frequency, TOF) of single sites, Equation (iii): severe mass-transport issues. Figures 6(c) and jV (at 0.8 or 0.9 V) = SD TOF(at 0.8 or 0.9 V) e (iii) 6(d) show the effect, for a same Fe-N-C cathode, of –3 switching from pure O2 to air. The power density is where SD has units of sites cm , TOF has units reduced by a factor >2, and the polarisation curves of electrons site–1 s–1 and e is the electric charge on air are characterised by a strong bending above of one electron (C electron–1). On one hand, ca. 0.4 A cm–2. This bending occurs at lower current increasing the wt% Pt on a support (typically, densities than when using Pt/C cathodes, likely due carbon powder) nearly proportionally increases the to a much greater thickness of Fe-N-C cathodes at a apparent activity of the Pt-supported catalyst. Pt –2 loading of 4 mgFe-N-C cm , relative to Pt/C cathodes nanoparticles of 2–3 nm size can now be grown –2 –2 of typically 0.4 mgPt cm (0.6 gcarbon cm , assuming on carbon supports up to ca. 50 wt% Pt on a typical 40% Pt/C catalyst). Such Fe-N-C cathodes carbon (51) before the average Pt particle size

(a) (b) 1.0 0.6 83 ppm H2S On Off 0.8 0.5

0.4 0.6 Preventive shut 0.3 0.4 down of Pt PEFC Cell potential, V Pt Cell potential, V 0.2 Pt 0.2 ODAN Fe-ODAN-1% ODAN 0.1

0 0.5 1.0 1.5 0 500 1000 1500 2000 Current density, A cm–2 Time, s

0.6 77 ppm toluene 0.6 163 ppm benzene

On Off On Off 0.5 0.5

0.4 0.4

0.3 0.3 Preventive shut Preventive shut down of Pt PEFC down of Pt PEFC Cell potential, V

Cell potential, V 0.2 0.2 Pt Pt 0.1 ODAN 0.1 ODAN

0 500 1000 1500 2000 2500 3000 0 500 1000 1500 2000 2500 3000 Time, s Time, s

–2 Fig. 7. (a) H2-O2 polarisation curve for an MEA comprising either a Fe-N-C cathode (4 mg cm loading) or a –2 Pt/C cathode (0.4 mgPt cm ); cell voltage before and after addition of: (b) 83 ppm H2S; (c) 77 ppm toluene; and (d) 163 ppm benzene in the cathode gas stream. Reproduced with permission from (50). Published by the Royal Society of Chemistry

significantly increases. This allows high Pt mass ORR activity in acidic medium, many observations and Pt surface area per volume of electrode to be show that their intrinsic activity is much less than reached. As a consequence, Pt-based cathodes are that of the atomically dispersed Me-Nx moieties highly active but at the same time relatively thin (52). As a consequence, Me-N-C catalysts display

(5–15 μm), which secures high accessibility by O2, a much lower weight content (<5 wt%) of active protons and electrons. For pyrolysed Fe-N-C or metal relative to Pt/C catalysts (50 wt%). From a Co-N-C catalysts, the situation is different. While given supposed wt% of potentially active metal in the metal atoms are atomically dispersed as Me-Nx Me-N-C material (MeNx moieties in the bulk or on moieties at low metal content (up to ca. 3 wt% the surface), the atomic mass and the utilisation metal (Fe or Co) on N-C, as is the case for the factor (ratio of electrochemically addressable MeNx catalysts of Figure 4), at higher metal contents moieties to total number of moieties in a given the metal atoms aggregate during pyrolysis to form catalyst), it is possible to calculate expected SD reduced metallic particles or metal carbides. Such values. Table III shows that only a slightly lower crystalline structures are often partially or totally SD-value is calculated for Fe-N-C (3 wt% Fe) vs. surrounded by graphitic shells during pyrolysis, 50 wt% Pt/C catalyst if one assumes that the which protects them somewhat from the acidic site utilisation of Fe-Nx moieties is 100%. Also of environment during electrochemistry. While such practical interest, it in turn implies that the loading 2 core-shell Metal@N-C structures may have some of FeNx sites per cm of MEA is only four times

Table III Examples of Site Density and Site Loading Numbers for Iron-Nitrogen-Carbon and R R′ n+ Platinum on Carbon Catalysts for Oxygen Reduction Reaction as well as Ni(P2 N2 ) a, b, c

Catalyst description SD, number of SL, number of sites Corresponding sites per cm3 of per cm2 electrode catalyst loading, mg electrode cm–2

20 18 d –2 d Fe-N-C, 3 wt% Fe, 1.29 × 10 1.29 × 10 4 mgFe-N-C cm 100% site utilisation

19 17 d –2 d Fe-N-C, 3 wt% Fe, 25% 3.23 × 10 3.23 × 10 4 mgFe-N-C cm site utilisation

20 17 e –2 e Pt/C, 50 wt% Pt, 25% 3.09 × 10 3.09 × 10 0.4 mgPt cm site utilisation c

R R′ n+ 18 16 f –2 f [Ni(P2 N2 )2] , 100% 7.5 × 10 1.5 × 10 0.04 mgNiP2N2 cm site utilisation, (53)

19 16 –2 Pt/C 50 wt% Pt, 25% 3.9 × 10 3.9 × 10 0.05 mgPt cm site utilisation a R R′ n+ The amount of electrochemically addressable active sites in Ni(P2 N2 ) was determined from cyclic voltammetry b 3 Ultra-low loading of Pt/C for HOR. The value of 0.4 gcarbon per cm of electrode volume was assumed for all catalysts. For Fe-N-C, it was assumed that either 100% or 25% of the Fe-Nx moieties are surface-exposed (participate in the ORR), while for Pt/C, it was assumed that ¼ of the Pt atoms are surface exposed (ratio corresponding to Pt particles of ca. 2–3 nm) cThis Pt site utilisation expresses only the ratio of surface Pt atoms to all Pt atoms. We highlight however that not all Pt surface sites are equivalent in terms of TOF, with terrace sites and concave coordinated Pt sites being more active for ORR (54, 55) dFor a 100 µm-thick electrode eFor a 10 µm-thick electrode fFor a 25 µm-thick electrode higher for a 100 µm thick Fe-N-C cathode than rather than statistical distribution on the surface for a 10 µm thick Pt/C cathode (Table III). Thus, and in the bulk, could increase the utilisation factor as a result of combined constraints in cathode of FeNx sites (in electrocatalysis, only the sites at layer thickness and in active metal content in the solid-electrolyte interface are electrochemically pyrolysed Fe-N-C catalysts, these straightforward active) (step C in Figure 8). The utilisation factor calculations imply that, in order to reach a same of statistically-dispersed FeNx moieties could also current density at 0.8 V or 0.9 V, such a Fe-N-C be increased via increased carbon surface area cathode must comprise FeNx active sites with (decreased average number of stacked graphene a TOF that is in fact very comparable to that of sheets in the material, step D in Figure 8). The surface-exposed Pt atoms. Table III also reports TOF of single metal-atoms in MeNx moieties might the numbers calculated assuming that only ¼ of also be increased, either through the preferential the Fe-Nx moieties are utilised (on the surface). formation of certain MeNx moieties (for example The assumed utilisation factor of 0.25 is in line edge vs. in-plane, if edge defects are more active, or with very recent quantifications of site utilisation vice versa if in-plane defects are shown to be more for such materials (see later). active) or the formation of more complex sites. One

Some possible pathways toward increased activity such possibility is the formation of binuclear Fe2Nx of Me-N-C catalysts are schemed in Figure 8, sites, where the Fe–Fe distance is commensurate reached via increasing either the SD or TOF (Equation with the O=O bond distance, allowing the two Fe

(iii)). While increasing the metal content without centres to work faster than two individual FeNx formation of metallic particles during pyrolysis may moieties (32). Other additional parameters have remain limited, there might be some gain possible been proposed or investigated to tune the TOF, relative to the present status (see Figure 4). such as bi-metallic catalysts (for example Fe-Mn Increased density of defects in graphene sheets or Fe-Co) (56, 57), and co-doped carbon by (in-plane sites) or increased edge length per mass nitrogen and another light element (such as sulfur, of carbon (edge sites) could allow increase of the phosphorus or boron) (58). The introduction of density of FeNx active sites (step B in Figure 8). At chemical elements other than Fe(Co), N and C could a fixed bulk metal content, preferential formation indeed offer broader perspectives on the electronic of FeNx sites on the surface of the carbon material, properties of the graphene sheets, and thereby of

In-plane binuclear Edge FeNx In-plane FeNx Fe2Nx moiety moiety O2 moiety

Edge Site binuclear density Fe2Nx moiety E O2 A B TOF

Site utilisation

D C

Fig. 8. Scheme of possible ways to increase the volumetric activity of Me-N-C catalysts: A typical Fe-N-C catalyst with four stacked graphene sheets including in-plane and edge FeNx sites. Only the FeNx sites on the top and bottom sheets are O2 accessible; B Fe-N-C catalyst with higher Fe content, including O2-accessible and O2 inaccessible sites; C Fe-N-C catalyst with preferential location of FeNx sites on the top and bottom graphene sheets (O2-accessible surfaces); D Fe-N-C catalyst with higher specific surface area (lower number of stacked graphene sheets); E Fe-N-C catalyst featuring binuclear Fe2Nx sites with possible cooperative O=O bond dissociation on the two Fe centres

the electron density at the Fe centres (59). The thereby breaking the relationship between stacking

O2 binding energy and therefore the TOF might be number and ORR activity. Such a relationship is increased further with one of those approaches. otherwise expected, with the hypothesis that most

Whether the MeNx sites are mostly located in- FeNx sites are located in-plane. plane or on the edge of graphene or disordered In guiding experimental efforts, disentangling graphene sheets is important (22, 23, 30, 60, 61). the overall volumetric activity into SD and TOF Answering that question would indicate whether (Equation (iii)) is not only of scientific importance the in-plane size of graphene sheets must be but also has technological implications on the decreased or the average number of stacked layers development of promising catalyst preparation decreased. For a set of Fe-N-C catalysts prepared routes and on the design of high-performance via the silica template method and using nicarbazin non-pgm cathode layers. For example, a low and iron nitrate as N, C and Fe precursors, it was SD-value implies stringent requirements on local shown that decreased stacking led to increased mass-transport properties near active sites, for a activity, implying that most of the active sites are given current density of the cathode layer. Once in-plane moieties for this synthesis (Figure 9) (60). methods are developed for quantifying SD in Interestingly, the carbon matrix was highly graphitic Me-N-C catalysts, TOF values can then be deduced for this set of catalysts (60), in contrast with what from the combined knowledge of the experimental is usually observed on most Fe-N-C catalysts that values of activity and SD. As an example, TOF are highly ORR-active (31, 62–64). The average values much higher than that of surface-located number of stacked graphene layers is typically only Pt atoms in Pt/C catalysts may lead to additional 4–6, as estimated from Raman spectroscopy, for mass-transport issues at short range (higher local high-surface-area amorphous carbon structures in diffusion flux of2 O needed towards individual Fe-N-C catalysts. Formation of large in-plane voids active sites). In addition, Me-N-C catalysts with (several carbon atoms removed, see for example high TOF but low SD may not only lead to local

Figure 4(b)) could also allow O2 access to in- O2 starvation when operating at high current plane FeNx moieties that are not situated in the density in PEMFC, but may also lead to the build uppermost graphene layer of a graphitic crystallite, up of hydrogen peroxide concentration gradients

(a) (b)

O2 access In-plane defects gr pyr pyrrolic FeN4+1 FeN4 FeN3 N N N 1.0 1.0 1.0 Graphene stack # FeN2+1 FeN2+2 –OH 0.9 0.8 0.8

FeN2+2 Power density, W cm Npyridinic =O 0.8 Npyrrolic 0.6 0 0.1 1.0 10 100 0.6 –COOH Inaccessible Edge defects sites 0.4 Voltage, V Voltage, 0.4

40 0.2 0.2

O2 voltage Air voltage

20 O2 power Air power 0 0 0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 –2 Current density, mA cm Current density, 0 Current density, A cm 15 20 25 Number of graphitic layers

Fig. 9. Effect of average number of stacked graphene sheets on ORR activity of pyrolysed Fe-N-C catalyst: (a) Scheme of in-plane and edge FeNx sites and O2 (in)accessibility; (b) PEMFC polarisation curves with H2/ O2 or H2/air feed; (c) negative correlation between current density at 0.8 V read on the H2/air polarisation curve and the average number of stacked graphene layers for a set of 12 Fe-N-C catalysts. Test conditions: 80°C, 100% RH, 1.5 bar (H2/O2) and 2.5 bar (H2/air) total gas pressures. Reprinted with permission from (60). Copyright (2017) American Chemical Society

(high concentration locally around active sites), dispersed Fe-ions) can now be synthesised, with expected dramatic influence on the long term the fact that Mössbauer spectroscopy is a bulk durability of Me-N-C cathodes (65). For all these technique implies that it cannot directly distinguish reasons, disentangling SD and TOF values from bulk sites from surface sites. The same observation the overall ORR activity will be important for the applies to X-ray absorption spectroscopy, the other further development of Me-N-C catalysts. broadly applied technique to characterise the local Whereas established methods exist for Pt-based environment around Fe and Co centres in pyrolysed catalysts (carbon monoxide chemisorption, Me-N-C catalysts. Recent years have witnessed electrochemical hydrogen sorption) they do not the development of a few ex situ (gas-solid) and work at room temperature for Me-N-C catalysts. in situ (liquid-solid) sorption techniques to assess Reliable quantification of the catalytic sites on the the SD of such catalysts. These rely on the strong surface of pyrolysed Me-N-C catalysts is an ongoing interaction of a small probe molecule with surface challenge. Attempts have been made in the past to adsorption sites resulting in poisoning of the site. estimate the values of SD and TOF of some Fe-N-C Early non-pgm catalyst poisoning studies realised catalysts (66, 67), however those estimations have that COgas, an intuitive choice of poison for Fe always included one or more hypotheses such centres, is unable to block Fe-N-C sites quantitatively as: (a) full utilisation of all Fe atoms in Fe-N-C under ambient temperature and pressure conditions –3 (NumberFe cm = SD); or (b) full utilisation of the (68, 69). In contrast, the cyanide ion was identified

FeNx moieties (a less crude hypothesis than (a), as a suitable in situ poisoning ligand for FeN4 but still probably far from the real Fe utilisation). centres (70, 71). Owing to its irreversible adsorption 57Fe Mössbauer spectroscopy is powerful in however, CN– adsorption could not be utilised for distinguishing FeNx moieties from crystalline Fe quantitative SD evaluation. Recently, an ex situ low- structures. Even though clean Fe-N-C catalysts with temperature (–100°C) CO gas pulse chemisorption an 57Fe Mössbauer spectroscopic signature showing based technique for the quantification of the SD only quadrupole doublets (assigned to atomically- of Fe-Mn-N-C and Mn-N-C catalysts was reported

244 © 2018 Johnson Matthey https://doi.org/10.1595/205651318X696828 Johnson Matthey Technol. Rev., 2018, 62, (2) by Strasser’s group (Figures 10(a) and 10(b)). outside the electrolyte, and thus relies on the Quantitative values of SD were directly obtained assumption that all sites probed by gas molecules from the total amount of adsorbed CO derived from remain active and accessible in an electrochemical consecutive CO pulses (72). Subsequent thermal environment. More recently still, an electrochemical desorption of the adsorbed CO during heating ramps in situ technique to probe, evaluate and quantify from –100°C to about +400°C provided additional the SD at the surface of a powder catalyst electrode insight on the desorption kinetics and, indirectly, into was reported by Kucernak’s group (73). The authors the relative CO chemisorption energies of different demonstrated a protocol that allows the quantification

FeNx or dissimilar Me-Nx sites. This SD estimation of SD in Me-N-C catalysts operating under acidic technique is straightforward, robust and may be conditions by means of nitrite adsorption, followed applicable to various non-pgm metal centres. Its by reductive stripping (Figures 10(c) and 10(d)). drawback consists of the fact that it is performed The method showed direct correlation to the catalytic

(a) (b)

–1 150 Fe-N-C Mn-N-C-3HT-2AL Cat 5 (Fe,Mn)-N-C (Fe,Mn)-N-C-3HT-2AL 100 4 Mn-N-C , nmol mg CO

3 Fe-N-C-3HT-2AL 50

2 0.1 M HClO4 Fe-N-C-2HT-1AL n CO uptake, 0 Metal-free N-C 1 0 10 20 30 40 50 –1 Normalised intensity, arbitrary units arbitrary Normalised intensity, Metal-free N-C Mass activity, Im (0.8 V), mA mgCat 0 0 20 40 60 80 100 Time, min

Potential, V vs. RHE (c) (d) –0.3 –0.2 –0.1 0 0.1 0.2 0.3 0.4

0.1 mA cm–2 + H , O2 H2O + + – N/C NH4 H ,e NO Fe-N/C N N Stripping N N Fe Fe Poisoning Nitrosyl N N N N Unpoisoned reduction – Poisoned charge ORR active NO2 ORR inactive Recovered

–0.8 –0.6 –0.4 –0.2 Potential, V vs. SCE

Fig. 10. Scheme of methods recently developed to quantify SD in pgm-free cathode catalysts: (a) quantification of moles CO adsorbed on monometallic Fe-N-C and Mn-N-C materials, or bimetallic Fe-Mn-N-C materials (labelled as (Fe,Mn)-N-C, although not necessarily implying binuclear active sites) by pulsed CO chemisorption at low temperature; (b) correlation between CO uptake (mole per mgcatalyst) and ORR activity; (c) scheme showing the poisoning of Fe site by nitrite leading to stable Fe-NO adducts and its removal by reductive stripping, leading to the regeneration of the ORR-active Fe site; (d) the number of Fe sites is determined from the electric charge associated with Fe-NO adducts during electrochemical reduction. Reproduced from (72) and (73)

245 © 2018 Johnson Matthey https://doi.org/10.1595/205651318X696828 Johnson Matthey Technol. Rev., 2018, 62, (2) activity and was demonstrated for a number of on whether the main instability (decreased ORR non-pgm catalyst materials. Lastly, a recent study activity over time) of pgm-free Me-N-C catalysts in from Los Alamos National Laboratory showed that PEMFC mostly originates from a decreasing SD or a specific doublet in the Mössbauer spectrum of an from a decreasing average TOF over time. Fe-N-C catalyst was modified in the presence of NO (74). After an electrochemical reduction treatment III 3. Non-pgm Anode Catalysts applied to convert potential Fe N4 moieties into II Fe N4 moieties, the introduction of NO-gas strongly 3.1 Limitations of Platinum Catalysts modified only one doublet. That doublet accounted for 24% of the relative absorption area while the sum of The anode Pt loading in PEMFC is currently around –2 all doublets (all types of FeNx moieties) accounted for 0.05 mg cm and cannot be further decreased 63% of the absorption area. This defines a utilisation without unacceptably increasing the anode factor of 0.38 for that specific catalyst, in line with sensitivity to H2-fuel contaminants. Indeed in the expected utilisation factor if FeNx moieties are floating electrode configuration, Pt nanoparticles statistically dispersed in graphene sheets, and with supported on carbon black (20–50 wt% Pt/C) –2 an average stacking (as determined experimentally) at ultra-low loadings <5 µgPt cm have shown –2 of five graphene sheets. HOR exchange current density of 100 mA cmPt –1 In the near future, one or several of these (80 A mgPt ‘mass-normalised’ exchange current methods and possibly new ones will certainly be density) at room temperature, with the performance regularly applied by research groups in the field. doubling when the temperature is increased to 60°C This will give more detailed information on both (75). However, ultra-low loaded Pt catalyst layers –2 the SD and TOF values in such catalysts and will (1–5µgPt cm ) are extremely sensitive to a range guide the design of such catalysts and catalyst of contaminants (CO, hydrogen sulfide) present in layers (Figure 11). Such methods that allow the fuel or leached from stack components that deconvolution of SD and TOF values will also reversibly or irreversibly deteriorate their activity. inform on how these values change after various Hence, the design of catalysts with ultra-low Pt electrochemical or chemical aging of the catalysts. content that are more tolerant to contaminants, This will lead to novel understanding in particular or of non-Pt HOR catalysts that are immune to

1023 0.9 V 1022

1021 –2

1020

1019 Pt/C

1240× site density 1018 30× improvement

Site density, sites cm Site density, 10 40× NPMC, 1 mg cm–2 1016 1240× activity 1015 improvement 0.0001 0.001 0.01 0.1 1 10 100 Site activity, s–1

Fig. 11. Master plot showing the linear relation between site loading (number of active sites per cm2 geometric area of cathode) and site TOF (number of electrons reduced per site and per second, at 0.9 V). The orange line corresponds to an iso-activity curve of 44 mA cm–2 at 0.9 V for the cathode (US DOE target for O2-fed cathode, see Table II), which can be reached with different combinations of SL and TOF values. The SL value is itself a combination of the SD and cathode thickness. The operating point for Pt nanoparticles –2 on carbon is indicated with the orange circle. Two Fe-N-C cathodes with loading of 1 mgFe-N-C cm are represented (open purple circles and filled red circle) as well as three possible paths to reach the cathode activity target, for a fixed cathode thickness

246 © 2018 Johnson Matthey https://doi.org/10.1595/205651318X696828 Johnson Matthey Technol. Rev., 2018, 62, (2) contaminants would not only further reduce the enzymes that reversibly catalyse HOR close to the total Pt content in PEMFCs but would also facilitate equilibrium potential and with turnover frequencies –1 the use of lower cost H2 reformed from natural gas exceeding 1000 s for both HOR and HER at 0.2 V or produced from biomass. This would ease the overpotential (82). Hydrogenases only contain Ni transition between ‘fossil’ H2 and renewable H2. and metal atoms in a sulfur-rich and organometallic environment at their active sites (Figure 13) (83, 3.2 Tungsten and Molybdenum 84). Inspired by the structure of these active sites, [Ni(P RN R′) ]n+ catalysts based on a nickel(II) Carbides 2 2 2 centre coordinated to two diphosphine ligands, Except for pgms, nickel as well as several Ni-alloys and bearing two pendant amine groups in a can drive HOR catalysis in highly alkaline conditions distorted square-planar geometry were designed (76). For a long time, tungsten and molybdenum by DuBois (Figure 13) (85, 86). Such amine carbides possibly doped with Co or Ni (WC, M/WC, groups mimicking the pendant base found at M = Ni, Co and Co/MoC) were the only pgm-free the [FeFe]-hydrogenase active site act as proton HOR catalysts that are stable under acidic conditions relays in close proximity to the metal-bound (77–79). Anodes based on these materials mixed hydride to promote H–H bond formation during with carbon black have exhibited current densities up the hydrogen evolution reaction (HER), and as a to 20–40 mA cm–2 at 0.1 V vs. reversible hydrogen polariser of the H–H bond, to promote its cleavage R R′ n+ electrode (RHE) (77–79). Such materials have during HOR (87, 88). Bioinspired [Ni(P2 N2 )2] proven quite resistant to CO (80) and have been complexes display bidirectional activity in HER and successfully implemented as MEA anodes together HOR with a few derivatives being reversible HER/ with Pt-based cathodes and displayed a maximum HOR catalysts, albeit with a kinetic bias towards power density of ~20 mW cm–2. The replacement of one or other direction (89–92). Almost reversible Pt by such catalysts in single-cell PEMFC has so far HER/HOR catalysis is observed in fully aqueous Cy Arg 7+ resulted in a factor-10 lower power density (Figure electrolyte with [Ni(P2 N2 )2] , although at 12) (81). However, they suffer from limited activity elevated temperatures (91). Maximal HOR TOFs and also limited stability due to carbide oxidation have been reported in the range of 102 s–1 at pH

(and release of CO2) coupled to the formation of the values between 0 and 1, with dramatic decrease as corresponding metal oxides. soon as the pH exceeds 2. R R′ n+ Immobilisation of [Ni(P2 N2 )2] complexes 3.3 Bioinspired Nickel-Diphosphine on carbon nanotubes (CNTs) deposited onto gas diffusion layers (GDL) yields very efficient Catalysts reversible catalytic materials for HER/HOR (93– More recently, a bioinspired approach for pgm-free 95). Three distinct immobilisation modes (covalent, HOR catalysts was developed. Hydrogenases are p-stacking and electrostatic) have been developed to attach the bioinspired catalytic site onto such nanostructured electrodes (96). To that aim, various R R′ n+ 1.0 200 [Ni(P2 N2 )2] structures incorporating distinct

Power density, mW cm anchoring groups were used. Figure 13 shows the Cy Arg 7+ Cy Ester 2+ 0.8 structure of [Ni(P2 N2 )2] , [Ni(P2 N2 )2] 150 Cy Py 2+ and [Ni(P2 N2 )2] containing arginine, activated 0.6 ester and pyrene anchoring groups, respectively. 100 Pt/C With different anchoring groups on the molecular 0.4 catalyst come different receiving groups on the CNTs Voltage, V Voltage, Pt/C 50 (Figure 13(b)). Standard grafting strategies were 0.2 CoMo/C

–2 employed: polycationic/polyanionic electrostatic interaction (95), covalent amide linkage (94) and 0 0 0 200 400 600 800 π-stacking of a pyrene moiety directly onto CNTs Current density, mA cm–2 (93). An alternative procedure was developed to R R′ n+ construct molecular [Ni(P2 N2 )] catalytic sites Fig. 12. Fuel cell polarisation curves with a typical in a stepwise manner on the CNT-based electrode Pt/C catalyst at the cathode and, at the anode, (53), in which the diphosphine ligand was firstly Co-Mo carbide or Pt/C. Reproduced from (81) with immobilised via amide coupling and the nickel permission from Elsevier

(a) (b) O O– + HO O H2N R1 NH NH+ A B NH N P O H2N (CH ) Cy B 2 3 P CN O– Cy Ni Cy S S S P N R1 S Ni Fe CN O Cy P S NC Fe Fe S S CO C CN O– +HN OC H O CO + H2N N O [NiFe]-hydrogenase active site [FeFe]-hydrogenase active site (CH2)3 NH2 OH

Rn Rn L, n+ N N – Cy Arg 7+ MWNT-COO / [Ni(P2 N2 )2] Cy Cy GDL P P Ni P P N N R2 Cy Cy Rn Rn O R R′ n+ N N N [Ni(P2 N2 )2] H O P P Cy Cy + Ni NH2NH3 Cy Cy P P N R HO O 2 R = H N 1 + N NH2 [Ni(P CyN Arg) ]7+ 2 2 2 N O H NH 2 O L = H+ , n = 7 MWNT-NH + / [Ni(P CyN Ester) ]2+ GDL 3 2 2 2 O O

R3 O N [Ni(P CyN Ester) ]2+ R2 = 2 2 2

O N R L = none, n = 2 Cy 3 N P P Ni CyN P Cy P Cy N R3 = Cy Pyrene 2+ [Ni(P2 N2 )2]

R3 L = none, n = 2

MWNT / [Ni(P CyN Pyrene) ]2+ GDL 2 2 2

Fig. 13. (a) Three molecular complexes for H2 oxidation and evolution with anchoring groups, inspired from the active sites of [FeFe]-hydrogenases. Here, the active site of C. reinhardtii HydA (B) is represented in its native state and a H2 molecule can coordinate where the red arrow points (83). The NiFe active site of E. coli Hyd1 is represented in a similar fashion (A) (84); (b) three bio-inspired molecularly-engineered – Cy Arg 7+ + Cy Ester 2+ nanomaterials for H2 oxidation: MWCNT-COO /[Ni(P2 N2 )2] , MWCNT-NH3 /[Ni(P2 N2 )2] and Cy Pyrene 2+ MWCNT/[Ni(P2 N2 )2] . Adapted from (96) with permission of the Royal Society of Chemistry

centre was then introduced in a second step in the of sites per geometric area of electrode) of form of a commercially available nickel salt (97). 1–3 × 10–9 mol cm–2 when densely packed CNT Electrochemical characterisation of the final electrodes are used. This loading is increased by electrodes in acetonitrile allowed quantification one order of magnitude when carbon microfibres R R′ n+ of the amount of [Ni(P2 N2 )2] species that are are used as templates to provide the CNT electrodes electrochemically addressable and display a two- with three-dimensional structuring. This value electron wave in cyclic voltammetry. In brief, all (2 × 10–8 mol cm–2) was used to determine SD and techniques provide typical site densities (number site loading (SL) data in Table III (53).

–2 –1 The electrocatalytic activity of all these up at ca. 15 mA cm (7.5 A mgNi ) at room materials was first assessed in 0.5 M sulfuric acid temperature and 0.3 V vs. RHE, and ca. 40 mA cm–2 –1 aqueous solution in a nitrogen or H2 atmosphere (>20 A mgNi ) at 85°C, a technologically relevant provided from the back of the porous substrate operating temperature, and 0.3 V vs. RHE (53). in half-cell configuration (floating electrode This catalytic performance approaches that technique, Figure 14). In some cases, the of a Pt nanoparticle-based electrode (Tanaka, R R′ n+ –2 [Ni(P2 N2 )2] -coated electrodes were coated 0.05 mgPt cm ) benchmarked under identical with a Nafion membrane to form a stable, pgm- conditions. Proton reduction catalysis at room free, air-resistant MEA. Under such conditions, temperature reaches at 100 mV overpotential + –2 reversible electrocatalytic activity for H /H2 a current density of 7 mA cm at 25°C, and interconversion was observed (Figure 14) (94). 38 mA cm–2 at 85°C (Figure 14) (53). Some of these Hydrogen is evolved at potentials just slightly electrode materials furthermore proved quite stable negative compared to the thermodynamic with unchanged catalytic response over 10 hours equilibrium (no overpotential required) and of continuous operation. The H2 oxidation current R pyrene anodic current density corresponding to hydrogen measured on GDL/MWCNT/[Ni(P2 N2 )2] oxidation is measured for potentials positive to the electrodes was found stable in the presence of thermodynamic equilibrium under H2 supply. 50 ppm CO, a feature likely shared by covalently R Performance was assessed for GDL/MWCNT/ immobilised NiP 2 species (93). Resistance to CO Cy Ester 2+ [Ni(P2 N2 )] (MWCNT = multi-walled carbon poisoning is thus another advantage of this series nanotube) in a half-cell floating electrode -set of catalysts over Pt nanoparticles, the surface of

(a) (b)

50 50 2.0 2.0 40 85ºC 40 1.5 1.5 1.0 30 30 1.0 Log(j), mA cm j, mA cm 20 20 –2 0.5 0.5 –2 10 25ºC 10 0 0 0 0 –0.5 –0.5 –2 –2 j, mA cm –10 –10 –1.0 –1.0

–20 –20 Log(j), mA cm –1.5 –1.5 –30 –30 –2.0 –2.0 –40 –40 –2.5 –2.5 –0.1 0 0.1 0.2 0.3 –0.15 –0.10 –0.05 0 0.05 0.10 0.15 E, V vs. SHE E, V vs. SHE (c) 300 Fig. 14. Linear sweep voltammograms recorded Cy Ester 250 in 0.5 M H2SO4 for a GDL/MWCNT-[Ni(P2 N2 )]

–1 MEA in a hydrogen atmosphere at different 200 temperatures: (a) adapted from (53) with

metal Cy Pyrene [Ni(P2 N2 )2](BF4)2 / MWNT/GDL permission of the Royal Society of Chemistry; (b) –2 mg 150 Commercial Pt electrode, 0.5 mtPt cm Tafel plot of the same catalyst (black) compared

–2 –2 to a commercial Pt electrode (0.05 mgPt cm ) 100 (red) at 25°C (dashed trace) and 85°C (solid

mA cm trace) in 0.5 M H2SO4 (adapted from (53) with 50 permission of the Royal Society of Chemistry); Cy Pyrene Normalised current density, Normalised current density, (c) chronoamperometry of Ni(P2 N2 )2] (BF4)2/MWNT/GDL (red) and a commercial Pt 0 –2 0 20 40 60 80 electrode (0.5 mgPt cm ) (black) in a 50 ppm CO Time, min atmosphere (adapted from (93) with permission of Wiley)

249 © 2018 Johnson Matthey https://doi.org/10.1595/205651318X696828 Johnson Matthey Technol. Rev., 2018, 62, (2) which is irreversibly poisoned within a few minutes output of 14 mW cm–2 at 0.47 V and 60°C, only under such conditions (Figure 14(c)). seven times lower than a full-Pt PEMFC similarly In order to gain structural insights regarding the built and operated under the same conditions (95). active species, X-ray absorption spectra (XAS) R pyrene at the Ni edge on GDL/MWCNT-[Ni(P2 N2 )2] electrodes were measured (93). The XAS of 4. Conclusions R pyrene 2+ immobilised [Ni(P2 N2 )2] species are quite similar, although not identical, to that of standalone Major breakthroughs have been achieved over the R R′ 2+ [Ni(P2 N2 )2] complexes. As-prepared GDL/ last decade in the design of catalysts based on R pyrene MWCNT-[Ni(P2 N2 )2] electrodes also contain Earth-abundant metals for catalysing the ORR or Ni(II) ions coordinated to light atoms attributed HOR, that are compatible with PEMFC technology to solvent or water molecules, but these species and operate with overpotential requirements are washed off during electrochemical equilibration similar to those of conventional Pt catalysts. These in aqueous electrolytes (93). Of note, the XAS catalysts are also more selective and therefore less recorded at the Ni edge are found unchanged sensitive to poisoning, a major asset for worldwide after 1 h of H2 evolution or H2 oxidation catalysis introduction of PEMFC technology if such innovative in aqueous H2SO4 0.5 M solution, attesting for the catalysts can be implemented in full devices while stability of the grafted species (93). retaining other key specifications, i.e. power These materials were implemented and shown performance and durability. Still, progress has to be to be operational in compact PEMFC prototypes made in two directions. First, the electrochemical (95, 93). An early fully operational Pt-free PEMFC activities of such catalysts are still lower than those Ph Pyrene 2+ was developed with MWCNT-[Ni(P2 N2 )2] at of optimised Pt-based catalysts. Closing the gap the anode, and a Co-N-C ORR catalyst (98, 99) at can be pursued by increasing the site density or the the cathode (93). An output power of 23 µW cm–2 turnover frequency of the active sites, both for ORR was obtained (Figure 15). More recently, the and HOR noble metal-free catalysts. Additionally, maximum power of a fuel cell integrating a specific optimisation of the catalyst layer structures Cy Arg 7+ SWNT-[Ni(P2 N2 )2] (SWNT = single-walled for such catalysts could help promoting the power nanotube) anode catalyst was measured just density reached for pgm-free H2/air PEMFCs below 2 mW cm2 (95). In that case, the limiting through a better control of protons and substrate/ component is however a biocathode based on product diffusion together with avoidance of bilirubin oxidase immobilised on CNTs. Replacing flooding. The other direction, in which it is urgent this biocathode by a Pt-based cathode yielded a to invest, is the stability of the catalyst materials Cy Arg 7+ SWNT-[Ni(P2 N2 )2] -Pt/C PEMFC with a power during representative drive cycles.

0.8 30 P,

0.6 m W cm 20 0.4 E, V

10 –2 0.2

0 0 0 20 40 60 80 100 j, mA cm–2 Co-N-C

H2 PEM O2 Ni-CNT

Fig. 15. Schematic representation of the PEMFC assembly from (93); the inset shows polarisation and –1 power density curves recorded at 60°C with supply of partially humidified 2H (20 ml min ) at the anode and passive air convection at the cathode. Adapted from (93) with permission of Wiley

Acknowledgements 13. H. A. Gasteiger, J. E. Panels and S. G. Yan, J. Power Sources, 2004, 127, (1–2), 162

This work was supported by the FCH Joint 14. C. Chen, Y. Kang, Z. Huo, Z. Zhu, W. Huang, H. L. Undertaking (CRESCENDO Project, Grant Xin, J. D. Snyder, D. Li, J. A. Herron, M. Mavrikakis, M. Chi, K. L. More, Y. Li, N. M. Markovic, G. A. Agreement n°779366) and the French National Somorjai, P. Yang and V. R. Stamenkovic, Science, Research Agency (Labex programme, ARCANE, 2014, 343, (6177), 1339 ANR-11-LABX-0003-01). Goran Drazic from the 15. M. Li, Z. Zhao, T. Cheng, A. Fortunelli, C.-Y. Chen, National Institute of Chemistry, Ljubljana, Slovenia, R. Yu, Q. Zhang, L. Gu, B. V. Merinov, Z. Lin, E. is acknowledged for his work on the HR-STEM Zhu, T. Yu, Q. Jia, J. Guo, L. Zhang, W. A. Goddard microscopy on the Fe-N-C catalyst from Centre III, Y. Huang and X. Duan, Science, 2016, 354, National de la Recherche Scientifique (CNRS). (6318), 1414 16. D. Banham and S. Ye, ACS Energy Lett., 2017, 2, References (3), 629 17. A. Kongkanand and M. F. Mathias, J. Phys. Chem. 1. F. T. Wagner, B. Lakshmanan and M. F. Mathias, J. Lett., 2016, 7, (7), 1127 Phys. Chem. Lett., 2010, 1, (14), 2204 18. N. Nonoyama, S. Okazaki, A. Z. Weber, Y. Ikogi 2. C. E. Thomas, Int. J. Hydrogen Energy, 2009, 34, and T. Yoshida, J. Electrochem. Soc., 2011, 158, (15), 6005 (4), B416 3. J. Garback, T. Ramsden, K. Wipke, S. Sprik and J. Kurtz, ‘Controlled Hydrogen Fleet and 19. Z. Chen, D. Higgins, A. Yu, L. Zhang and J. Zhang,

Infrastructure Demonstration and Validation Energy Environ. Sci., 2011, 4, (9), 3167 Project’, Workshop: Compressed Natural Gas 20. F. Jaouen, E. Proietti, M. Lefèvre, R. Chenitz, J. and Hydrogen Fuels: Lessons Learned for the P. Dodelet, G. Wu, H. T. Chung, C. M. Johnston Safe Deployment of Vehicles, National Renewable and P. Zelenay, Energy Environ. Sci., 2011, 4, Energy Laboratory of the US Department of (1), 114 Transportation and US Department of Energy, 21. H. Fei, J. Dong, M. J. Arellano-Jiménez, G. Ye, N. Washington, DC, USA, 10th–11th December, 2009 D. Kim, E. L. G. Samuel, Z. Peng, Z. Zhu, F. Qin, J. 4. “Global Transport Scenarios 2050”, World Energy Bao, M. J. Yacaman, P. M. Ajayan, D. Chen and J. Council, London, UK, 2011, 76 pp M. Tour, Nature Commun., 2015, 6, 8668 5. O. Gröger, H. A. Gasteiger and J.-P. Suchsland, J. 22. H. T. Chung, D. A. Cullen, D. Higgins, B. T. Sneed, Electrochem. Soc., 2015, 162, (14), A2605 E. F. Holby, K. L. More and P. Zelenay, Science, 6. H. R. Ellamla, I. Staffell, P. Bujlo, B. G. Pollet and 2017, 357, (6350), 479 S. Pasupathi, J. Power Sources, 2015, 293, 312 23. A. Zitolo, V. Goellner, V. Armel, M.-T. Sougrati, 7. B. D. James, J. M. Huya-Kouadio, C. Houchins T. Mineva, L. Stievano, E. Fonda and F. Jaouen, and D. A. DeSantis, “Mass Production Cost Nature Mater., 2015, 14, (9), 937

Estimation of Direct H2 PEM Fuel Cell Systems for 24. N. Ramaswamy, U. Tylus, Q. Jia and S. Mukerjee, Transportation Applications: 2016 Update”, Rev. J. Am. Chem. Soc., 2013, 135, (41), 15443

3, Strategic Analysis Inc, Arlington, USA, January, 25. A. Zitolo, N. Ranjbar-Sahraie, T. Mineva, J. Li, Q.

2017, 244 pp Jia, S. Stamatin, G. F. Harrington, S. M. Lyth, P. 8. “Fuel Cell Technologies Office: Multi-Year Krtil, S. Mukerjee, E. Fonda and F. Jaouen, Nature Research, Development, and Demonstration Commun., 2017, 8, 957 Plan: Planned Program Activities for 2011-2020”, 26. “2017 Annual Work Plan and Budget”, Fuel Cells Office of Energy Efficiency and Renewable Energy, and Hydrogen 2 Joint Undertaking (FCH2JU), Washington, DC, USA, 2013 Brussels, Belgium, 2017 9. G. Wu and P. Zelenay, Acc. Chem. Res., 2013, 46, 27. ‘Fuel Cell Technologies Office Annual FOA’, Funding (8), 1878 Number DE-FOA-0001647, Fuel Cells Technology 10. F. Jaouen, ‘Heat-Treated Transition Metal-NxCy Office, Office of Energy Efficiency and Renewable Electrocatalysts for the O2 Reduction in Acid Energy (EERE), Washington, DC, USA, 18th PEM Fuel Cells’, in “Non-Noble Metal Fuel Cell November, 2016 Catalysts”, eds. Z. Chen, J.-P. Dodelet and J. 28. ‘Communication from the Commission to the Zhang, Wiley-VCH Verlag GmbH & Co KGaA, European Parliament, the Council, the European Weinheim, Germany, 2014, pp. 29–118 Economic and Social Committee and the 11. A. Serov and C. Kwak, Appl. Catal. B: Environ., Committee of the Regions on the Review of the 2009, 90, (3–4), 313 List of Critical Raw Materials for the EU and the 12. A. Brouzgou, S. Q. Song and P. Tsiakaras, Appl. Implementation of the Raw Materials Initiative’, Catal. B: Environ., 2012, 127, 371 COM/2014/0297 final, 26th May, 2014

29. Ballard Power Systems Inc, ‘Ballard to Offer 49. G. Wu, K. Artyushkova, M. Ferrandon, A. J. Kropf, World’s First PEM Fuel Cell Product Using Non D. Myers and P. Zelenay, ECS Trans., 2009, 25, Precious Metal Catalyst’, PR Newswire, New York, (1), 1299 USA, 12th September, 2017 50. D. Malko, T. Lopes, E. Symianakis and A. R. 30. M. Lefèvre, E. Proietti, F. Jaouen and J.-P. Dodelet, Kucernak, J. Mater. Chem. A, 2016, 4, (1), 142 Science, 2009, 324, (5923), 71 51. T. R. Ralph and M. P. Hogarth, Platinum Metals 31. E. Proietti, F. Jaouen, M. Lefèvre, N. Larouche, Rev., 2002, 46, (3), 117 J. Tian, J. Herranz and J.-P. Dodelet, Nature 52. J. A. Varnell, E. C. M. Tse, C. E. Schulz, T. T. Fister, Commun., 2011, 2, 416 R. T. Haasch, J. Timoshenko, A. I. Frenkel and A. 32. E. F. Holby, G. Wu, P. Zelenay and C. D. Taylor, J. A. Gewirth, Nature Commun., 2016, 7, 12582 Phys. Chem. C, 2014, 118, (26), 14388 53. T. N. Huan, R. T. Jane, A. Benayad, L. Guetaz, P. D. 33. W. Liang, J. Chen, Y. Liu and S. Chen, ACS Catal., Tran and V. Artero, Energy Environ. Sci., 2016, 9, 2014, 4, (11), 4170 (3), 940 34. J. Shui, C. Chen, L. Grabstanowicz, D. Zhao and 54. F. Calle-Vallejo, J. Tymoczko, V. Colic, Q. H. Vu, M. D.-J. Liu, Proc. Natl. Acad. Sci. USA, 2015, 112, D. Pohl, K. Morgenstern, D. Loffreda, P. Sautet, W. (34), 10629 Schuhmann and A. S. Bandarenka, Science, 2015, 350, (6257), 185 35. Y. Nabae, S. Nagata, T. Hayakawa, H. Niwa, Y. Harada, M. Oshima, A. Isoda, A. Matsunaga, K. 55. F. Calle-Vallejo, M. D. Pohl, D. Reinisch, D. Loffreda, Tanaka and T. Aoki, Sci. Rep., 2016, 6, 23276 P. Sautet and A. S. Bandarenka, Chem. Sci., 2017, 8, (3), 2283 36. F. Jaouen, V. Goellner, M. Lefèvre, J. Herranz, E. Proietti and J. P. Dodelet, Electrochim. Acta, 2013, 56. S. Brüller, H.-W. Liang, U. I. Kramm, J. W. 87, 619 Krumpfer, X. Feng and K. Müllen, J. Mater. Chem. A, 2015, 3, (47), 23799 37. J. Y. Cheon, T. Kim, Y. Choi, H. Y. Jeong, M. G. Kim, Y. J. Sa, J. Kim, Z. Lee, T.-H. Yang, K. Kwon, 57. A. Serov, M. H. Robson, M. Smolnik and P. O. Terasaki, G.-G. Park, R. R. Adzic and S. H. Joo, Atanassov, Electrochim. Acta, 2012, 80, 213 Sci. Rep., 2013, 3, 2715 58. J. Y. Cheon, J. H. Kim, J. H. Kim, K. C. Goddeti, J.

38. A. Serov, K. Artyushkova, E. Niangar, C. Wang, N. Y. Park and S. H. Joo, J. Am. Chem. Soc., 2014,

Dale, F. Jaouen, M.-T. Sougrati, Q. Jia, S. Mukerjee 136, (25), 8875 and P. Atanassov, Nano Energy, 2015, 16, 293 59. U. I. Kramm, I. Abs-Wurmbach, I. Herrmann- 39. V. Armel, S. Hindocha, F. Salles, S. Bennett, D. Geppert, J. Radnik, S. Fiechter and P. Bogdanoff,

Jones and F. Jaouen, J. Am. Chem. Soc., 2017, J. Electrochem. Soc., 2011, 158, (1), B69 139, (1), 453 60. M. J. Workman, A. Serov, L. Tsui, P. Atanassov and K. Artyushkova, ACS Energy Lett., 2017, 2, (7), 40. H. A. Gasteiger, S. S. Kocha, B. Sompalli and F. T. 1489 Wagner, Appl. Catal. B: Environ., 2005, 56, (1–2), 9 61. F. Charreteur, F. Jaouen, S. Ruggeri and J.-P. 41. A. Morozan and F. Jaouen, Energy Environ. Sci., Dodelet, Electrochim. Acta, 2008, 53, (6), 2925 2012, 5, (11), 9269 62. F. Jaouen, J. Herranz, M. Lefèvre, J.-P. Dodelet, U. I. 42. D. Zhao, J.-L. Shui, L. R. Grabstanowicz, C. Chen, Kramm, I. Herrmann, P. Bogdanoff, J. Maruyama, S. M. Commet, T. Xu, J. Lu and D.-J. Liu, Adv. T. Nagaoka, A. Garsuch, J. R. Dahn, T. Olson, S. Mater., 2014, 26, (7), 1093 Pylypenko, P. Atanassov and E. A. Ustinov, ACS 43. P. Su, H. Xiao, J. Zhao, Y. Yao, Z. Shao, C. Li and Appl. Mater. Interfaces., 2009, 1, (8), 1623 Q. Yang, Chem. Sci., 2013, 4, (7), 2941 63. C. Gumeci, N. Leonard, Y. Liu, S. McKinney, B. 44. F. Jaouen, M. Lefèvre, J.-P. Dodelet and M. Cai, J. Halevi and S. C. Barton, J. Mater. Chem. A, 2015, Phys. Chem. B, 2006, 110, (11), 5553 3, (43), 21494 45. J. Anibal, H. G. Romero, N. D. Leonard, C. Gumeci, 64. U. I. Kramm, I. Herrmann, S. Fiechter, G. Zehl, I. B. Halevi and S. C. Barton, Appl. Catal. B: Environ., Zizak, I. Abs-Wurmbach, J. Radnik, I. Dorbandt 2016, 198, 32 and P. Bogdanoff, ECS Trans., 2009, 25, (1), 659 46. I. Herrmann, U. I. Kramm, S. Fiechter and P. 65. V. Goellner, V. Armel, A. Zitolo, E. Fonda and F. Bogdanoff, Electrochim. Acta, 2009, 54, (18), Jaouen, J. Electrochem. Soc., 2015, 162, (6), H403 4275 66. F. Jaouen and J.-P. Dodelet, Electrochim. Acta, 47. S. Stariha, K. Artyushkova, A. Serov and P. 2007, 52, (19), 5975 Atanassov, Int. J. Hydrogen Energy, 2015, 40, 67. U. I. Kramm, J. Herranz, N. Larouche, T. M. Arruda, (42), 14676 M. Lefèvre, F. Jaouen, P. Bogdanoff, S. Fiechter, 48. G. Wu, K. L. More, C. M. Johnston and P. Zelenay, I. Abs-Wurmbach, S. Mukerjee and J.-P. Dodelet, Science, 2011, 332, (6028), 443 Phys. Chem. Chem. Phys., 2012, 14, (33), 11673

68. M. S. Thorum, J. M. Hankett and A. A. Gewirth, J. 84. R. Hidalgo, P. A. Ash, A. J. Healy and K. A. Vincent, Phys. Chem. Lett., 2011, 2, (4), 295 Angew. Chem. Int. Ed., 2015, 54, (24), 7110 69. Q. Zhang, K. Mamtani, D. Jain, U. Ozkan and A. 85. A. D. Wilson, R. H. Newell, M. J. McNevin, J. T. Asthagiri, J. Phys. Chem. C, 2016, 120, (28), Muckerman, M. R. DuBois and D. L. DuBois, J. Am. 15173 Chem. Soc., 2006, 128, (1), 358 70. S. Gupta, C. Fierro and E. Yeager, J. Electroanal. 86. C. J. Curtis, A. Miedaner, R. Ciancanelli, W. W. Chem. Interfacial Electrochem., 1991, 306, (1– Ellis, B. C. Noll, M. R. DuBois and D. L. DuBois, 2), 239 Inorg. Chem., 2003, 42, (1), 216 71. W. Li, A. Yu, D. C. Higgins, B. G. Llanos and Z. 87. D. L. DuBois, Inorg. Chem., 2014, 53, (8), 3935 Chen, J. Am. Chem. Soc., 2010, 132, (48), 17056 88. B. Ginovska-Pangovska, A. Dutta, M. L. Reback, 72. N. R. Sahraie, U. I. Kramm, J. Steinberg, Y. Zhang, J. C. Linehan and W. J. Shaw, Acc. Chem. Res., A. Thomas, T. Reier, J.-P. Paraknowitsch and P. 2014, 47, (8), 2621 Strasser, Nature Commun., 2015, 6, 8618 89. S. E. Smith, J. Y. Yang, D. L. DuBois and R. M. 73. D. Malko, A. Kucernak and T. Lopes, Nature Bullock, Angew. Chem. Int. Ed., 2012, 51, (13), Commun., 2016, 7, 13285 3152 74. J. L. Kneebone, S. L. Daifuku, J. A. Kehl, G. Wu, 90. A. Dutta, S. Lense, J. Hou, M. H. Engelhard, J. A.

H. T. Chung, M. Y. Hu, E. E. Alp, K. L. More, P. S. Roberts and W. J. Shaw, J. Am. Chem. Soc.,

Zelenay, E. F. Holby and M. L. Neidig, J. Phys. 2013, 135, (49), 18490 Chem. C, 2017, 121, (30), 16283 91. A. Dutta, D. L. DuBois, J. A. S. Roberts and W. J. Shaw, Proc. Natl. Acad. Sci. USA, 2014, 111, 75. C. M. Zalitis, J. Sharman, E. Wright and A. R. (46), 16286 Kucernak, Electrochim. Acta, 2015, 176, 763 92. A. Dutta, J. A. S. Roberts and W. J. Shaw, Angew. 76. W. Sheng, A. P. Bivens, M. Myint, Z. Zhuang, R. Chem. Int. Ed., 2014, 53, (25), 6487 V. Forest, Q. Fang, J. G. Chen and Y. Yan, Energy Environ. Sci., 2014, 7, (5), 1719 93. P. D. Tran, A. Le Goff, J. Heidkamp, B. Jousselme, N. Guillet, S. Palacin, H. Dau, M. Fontecave and 77. V. Sh. Palanker, R. A. Gajyev and D. V. Sokolsky, V. Artero, Angew. Chem. Int. Ed., 2011, 50, (6), Electrochim. Acta, 1977, 22, (2), 133 1371 78. M. Nagai, M. Yoshida and H. Tominaga, Electrochim. 94. A. Le Goff, V. Artero, B. Jousselme, P. D. Tran, Acta, 2007, 52, (17), 5430 N. Guillet, R. Métayé, A. Fihri, S. Palacin and M. 79. S. Izhar, M. Yoshida and M. Nagai, Electrochim. Fontecave, Science, 2009, 326, (5859), 1384 Acta, 2009, 54, (4), 1255 95. S. Gentil, N. Lalaoui, A. Dutta, Y. Nedellec, S. 80. D. R. McIntyre, G. T. Burstein and A. Vossen, J. Cosnier, W. J. Shaw, V. Artero and A. Le Goff, Power Sources, 2002, 107, (1), 67 Angew. Chem. Int. Ed., 2017, 56, (7), 1845 81. S. Izhar and M. Nagai, J. Power Sources, 2008, 96. N. Coutard, N. Kaeffer and V. Artero, Chem. 182, (1), 52 Commun., 2016, 52, (95), 13728 82. K. Pandey, S. T. A. Islam, T. Happe and F. A. 97. R. T. Jane, P. D. Tran, E. S. Andreiadis, J. Pécaut Armstrong, Proc. Natl. Acad. Sci. USA, 2017, 114, and V. Artero, Comptes Rendus Chimie, 2015, 18, (15), 3843 (7), 752 83. G. Berggren, A. Adamska, C. Lambertz, T. R. 98. A. Morozan, P. Jégou, B. Jousselme and S. Palacin, Simmons, J. Esselborn, M. Atta, S. Gambarelli, Phys. Chem. Chem. Phys., 2011, 13, (48), 21600 J.-M. Mouesca, E. Reijerse, W. Lubitz, T. Happe, 99. A. Morozan, P. Jégou, M. Pinault, S. Campidelli, B. V. Artero and M. Fontecave, Nature, 2013, 499, Jousselme and S. Palacin, ChemSusChem, 2012, (7456), 66 5, (4), 647

The Authors

Frédéric Jaouen obtained his PhD at the Royal Institute of Technology (KTH), Stockholm, Sweden, in 2003 under the supervision of Professor Göran Lindbergh. From 2004 to 2011, he was a research associate in Professor Jean-Pol Dodelet’s group at Institute National de la Recherche Scientifique (INRS), Canada, where he focused on non-precious metal catalysts for oxygen electro-reduction. In 2011, Frédéric Jaouen was awarded an excellence chair from the ANR and moved to Université de Montpellier, France, to pursue his research on novel pgm-free catalysts for electrochemical energy conversion devices as a CNRS research fellow. He currently coordinates the H2020 project CREATE on anion-exchange membrane fuel cells and electrolysers and is strongly involved in several other European

and national projects focusing on catalysts free of critical raw materials for O2 and CO2 reduction. He was awarded the academic research prize from the Energy division of the French Chemical Society in 2017.

Deborah J. Jones received her PhD in 1982 from the University of London, King’s College, UK, under the supervision of John Emsley. After a period at Southampton University, UK, she moved to France first with a Royal Society Fellowship and then a European Commission Sectoral Grant in Non-Nuclear Energy at the University of Montpellier, France. She is currently Director of Research at CNRS, Associate Director of the Institute Charles Gerhardt for Molecular Chemistry and Materials in Montpellier and serves as vice-president of the International Society of Electrochemistry. Her interests encompass the development of membrane, catalyst and electrode materials for proton exchange membrane fuel cells and electrolysers. She is a Fellow of the Electrochemical Society (2015) and recipient (2016) of the Sir William Grove award of the International Association for Hydrogen Energy.

Nathan Coutard graduated in 2015 from Université Pierre et Marie Curie in Paris, France. His master’s internship in Professor Kylie Vincent’s lab at University of Oxford, UK, aimed to study the regulatory hydrogenase from Ralstonia eutropha using in operando spectro electrochemistry. His PhD project in Dr Vincent Artero’s group aims to optimise and integrate bio-inspired catalysts in a functional proton-exchange membrane fuel cell. His work includes electrode nanostructuration, electrochemistry, catalyst design, synthesis and tinkering.

Vincent Artero studied at the Ecole Normale Supérieure, Ulm, Germany, and graduated with his PhD in Inorganic Chemistry at the Université Pierre et Marie Curie in Paris in 2000. After a post-doctoral stay in RWTH Aachen, Germany, in the group of Professor Ulrich Kölle, he moved to the CEA centre in Grenoble (Fundamental Research Division) in 2001 to develop bioinspired chemistry related to hydrogen production and artificial photosynthesis. He received the ‘Grand Prix Mergier-Bourdeix de l’Académie des Sciences’ (2011) and was granted a Starting Grant from the European Research Council (ERC 2012– 17). He currently acts as Chair of the Scientific Advisory Board of the ARCANE Excellence Laboratory Network (LABEX) for bio-driven chemistry in Grenoble and as co-chair of the French Research Network (GDR) on solar fuels.

Peter Strasser is the chaired professor of Electrochemistry and Electrocatalysis in the Chemical Engineering Division of the Department of Chemistry at the Technical University Berlin, Germany. Prior to his appointment, he was Assistant Professor at the Department of Chemical and Biomolecular Engineering at the University of Houston, USA. Before moving to Houston, Professor Strasser served as Senior Member of staff at Symyx Technologies, Inc, USA. In 1999, he received his doctoral degree in Physical Chemistry and Electrochemistry from the Fritz-Haber-Institute of the Max-Planck-Society, Berlin, Germany, under the direction of Gerhard Ertl. Peter Strasser was awarded the Otto-Hahn Research Medal by the Max-Planck Society, the Otto Roelen medal for catalysis awarded by the German Catalysis Society and the Ertl Prize awarded by the Ertl Center for Catalysis.

Anthony Kucernak received his PhD from Southampton University, UK, in 1991 and carried out postdoctoral research at Cambridge University, UK. He subsequently moved to Imperial College London, UK, where he has held the post of Professor of Physical Chemistry since 2009. He has published more than seventy papers on novel methods for screening catalysts under realistic operating conditions; the study of individual catalyst particles and the modification of surface composition to probe the fundamentals of electrocatalysis. He has been an invited plenary speaker at the 2005 International Fuel Cell Workshop, Japan, and co-chaired the Gordon Conference on fuel cells in 2004. In 2006 he was awarded the Helmut Fisher medal for his work on fuel cell electrode structure and fundamental results in electrocatalysis.

255 © 2018 Johnson Matthey This page is intentionally blank www.technology.matthey.com Johnson Matthey Technology Review is Johnson Matthey’s international journal of research exploring science and technology in industrial applications www.technology.matthey.com Editorial team

Manager Dan Carter Editor Sara Coles Editorial Assistant Ming Chung Senior Information OfficerElisabeth Riley

Johnson Matthey Technology Review Johnson Matthey Plc Orchard Road Royston SG8 5HE UK Tel +44 (0)1763 253 000 Email [email protected] www.technology.matthey.com