Paper Trail

cmd = pdftohtml -s -i apache/papers/16/paper16.pdf

ret = Page-1
Page-2
Page-3
Page-4
Page-5
Page-6
Page-7
Page-8
Page-9
Page-10
Page-11
Page-12
Page-13

numPages = 27

pageNum = 1

14010
Phys. Chem. Chem. Phys., 2012,
14 , 14010–14022
This journal is c the Owner Societies 2012
Cite this:
Phys. Chem. Chem. Phys., 2012, 14 , 14010–14022
Identifying active surface phases for metal oxide electrocatalysts: a study of manganese oxide bi-functional catalysts for oxygen reduction and water oxidation catalysisw
Hai-Yan Su,
a Yelena Gorlin,b Isabela C. Man,a Federico Calle-Vallejo,a
Jens K. Nørskov,
bc Thomas F. Jaramillo*b and Jan Rossmeisl*a
Received 16th March 2012, Accepted 23rd August 2012 DOI: 10.1039/c2cp40841d
Progress in the ﬁeld of electrocatalysis is often hampered by the diﬃculty in identifying the active site on an electrode surface. Herein we combine theoretical analysis and electrochemical methods to identify the active surfaces in a manganese oxide bi-functional catalyst for the oxygen reduction reaction (ORR) and the oxygen evolution reaction (OER). First, we electrochemically characterize the nanostructured a-Mn2O3 and ﬁnd that it undergoes oxidation in two potential regions: initially, between 0.5 V and 0.8 V, a potential region relevant to the ORR and, subsequently, between 0.8 V and 1.0 V, a potential region between the ORR and the OER relevant conditions. Next, we perform density function theory (DFT) calculations to understand the changes in the MnOx surface as a function of potential and to elucidate reaction mechanisms that lead to high activities observed in the experiments. Using DFT, we construct surface Pourbaix and free energy diagrams of three diﬀerent MnOx surfaces and identify 1/2 ML HO* covered Mn2O3 and O* covered MnO2, as the active surfaces for the ORR and the OER, respectively. Additionally, we ﬁnd that the ORR occurs through an associative mechanism and that its overpotential is highly dependent on the stabilization of intermediates through hydrogen bonds with water molecules. We also determine that OER occurs through direct recombination mechanism and that its major source of overpotential is the scaling relationship between HOO* and HO* surface intermediates. Using a previously developed Sabatier model we show that the theoretical predictions of catalytic activities match the experimentally determined onset potentials for the ORR and the OER, both qualitatively and quantitatively. Consequently, the combination of ﬁrst-principles theoretical analysis and experimental methods oﬀers an understanding of manganese oxide oxygen electrocatalysis at the atomic level, achieving fundamental insight that can potentially be used to design and develop improved electrocatalysts for the ORR and the OER and other important reactions of technological interest.
1.
Introduction
Fundamental understanding of electrochemical reactions on surfaces has improved signiﬁcantly in recent years, yet many
microscopic processes occurring during electrochemical reac- tions are still poorly understood due to diﬃculties in simulating electrochemical reactions computationally and in pinpointing active sites experimentally. The ultimate challenge in electro- catalyst development is to identify the active sites on a given catalytic surface and determine the reaction mechanisms on those sites. If one can achieve such level of fundamental under- standing, one could accelerate the design and development of improved electrocatalysts.
1–5
The electrochemical oxygen reduction reaction (ORR) and
oxygen evolution reaction (OER) are of great interest as they are processes involved in energy conversion between fuel and electricity and vice versa. The development of a bi-functional catalyst for both reactions is an important challenge in electro- chemistry; such a catalyst could be particularly useful for energy storage applications. For example, the catalyst could be employed in a unitized regenerative fuel cell (URFC), which is
a Center for Atomic-Scale Materials Design (CAMD), Department of
Physics, DTU, Technical University of Denmark, DK-2800 Kgs, Lyngby, Denmark. E-mail: jross@fysik.dtu.dk
b Department of Chemical Engineering, 381 North-South Mall,
Stanford University, Stanford, CA 94305–5025, USA. E-mail: jaramillo@stanford.edu
c SLAC National Accelerator Laboratory, Stanford, CA 94025-7015,
USA
w Electronic supplementary information (ESI) available: (1) The phase-diagram of MnOx surfaces calculated as a function of the potential at pH = 0; (2) free-energy diagram for oxygen reduction and oxygen evolution on all the non-self-consistent MnOx surfaces; (3) the method to calculate the number of O (NO) coordinated with Mn on various MnOx surfaces; (4) the method to construct Pourbaix diagram for both bulk and surface. See DOI: 10.1039/c2cp40841d
PCCP
Dynamic Article Links
www.rsc.org/pccp
PAPER
Published on 18 September 2012. Downloaded by Swansea University on 17/10/2016 09:49:59.
View Article Online / Journal Homepage / Table of Contents for this issue

pageNum = 2

This journal is c the Owner Societies 2012
Phys. Chem. Chem. Phys., 2012,
14 , 14010–14022 14011
an energy storage device that can utilize intermittent renewable energy such as wind or solar. During operation, a URFC splits water into H2 and O2, in the electrolysis mode and consumes H2 to produce electricity, in the fuel cell mode.
6,7 Both the fuel cell
reaction and the water electrolysis reaction require large over- potentials at the oxygen electrode – no current catalyst material operates near the equilibrium potential for either the ORR or the OER. Consequently, improved oxygen electrode catalysts will increase the eﬃciency of the full conversion cycle from electricity to hydrogen and back to electricity in an URFC.
8,9
Furthermore, it is desirable to develop highly active catalysts from cheap and abundant materials, i.e. alternatives to IrO2/Pt or IrO2–RuO2/Pt catalysts, which are the best catalysts known for these reactions.
10–12
In this study, we investigate ORR/OER catalytic activity of
manganese oxides (MnOx) surfaces. Several characteristics of MnOx motivate our study: (1) Mn changes oxidation states from +2 to +3 to +4 near the equilibrium potential for the ORR and the OER,
13
suggesting that Mn can exchange
oxygen atoms with the electrolyte at relevant potentials – a property that could potentially facilitate ORR and OER chemistry, (2) manganese is an inexpensive, earth-abundant element, and thus is scalable for large-scale energy applica- tions, and (3) there is precedent for Mn oxides eﬀectively catalyzing the OER: the Oxygen–Evolving Complex (OEC) in Photosystem II is a Mn-oxo cluster that catalyzes the OER during photosynthesis.
14–17 Historically, a number of manganese
oxides have shown promising electrocatalytic activity for either the ORR or the OER, but not for both.
18–30 Recently, it was
shown that a nanostructured a-Mn2O3 exhibited excellent bi-functional ORR and OER activity similar to that of the best known precious metal nanoparticle catalysts: Pt, Ru, and Ir.
31
However, the bi-functional ORR/OER activities of the nano- structured a-Mn2O3 and precious metal nanoparticles are still short of an ideal reversible oxygen electrode.
In principle, it should be possible to develop an ideal
reversible oxygen electrode – a material that eﬀectively catalyzes both the ORR and the OER. With such a catalyst, one would be able to obtain a high reduction current at potentials just cathodic of the equilibrium potential and a high anodic current at potentials just anodic of the equilibrium potential. Such a catalyst would likely undergo minimal changes in surface structure, swinging from one reaction to the other, as it would always operate near the equilibrium potential. For imperfect catalysts, there are large overpotentials associated with both the ORR and the OER, which means that the two reactions operate at signiﬁcantly diﬀerent potentials away from equilibrium in opposite directions. The diﬀerent operating conditions will likely result in diﬀerent surface conditions within each potential window of activity and likely diﬀerent oxidation states of the catalyst surface at the relevant potentials for the ORR and the OER.
To understand surface conditions of imperfect catalysts
under ORR/OER relevant potentials, a variety of in situ and ex situ
spectroscopic techniques have been employed.
25,32–35
We believe that density functional theory (DFT) calculations can also be used to help elucidate active catalyst surfaces. While DFT methods have problems describing transition metal oxides accurately, they have been shown recently to
describe trends in reactivity of metals and metal oxides for the OER and the ORR quite well.
36–40 Due to the complexity of
the systems of interest in describing these processes, DFT calculations are the only methods available to us at the moment. The calculations can be used to construct surface Pourbaix diagrams, which describe surface oxidation and dissolution processes at a given pH and potential,
41 making
it possible to identify thermodynamically stable surface phases during reaction conditions (as a function of pH and potential), the catalytic activity of those surfaces, and the associated mechanistic pathways for the reactions of interest. The knowl- edge of active surfaces and reaction mechanisms gained from DFT studies will shed light onto the surface chemistry of catalyst materials in ways that are extremely diﬃcult to obtain with modern experimental tools. DFT can thus play a unique role in contributing to the design and development of improved materials.
In the work described herein, we present DFT calculations
in combination with electrochemical characterization to eluci- date the active surfaces and reaction mechanisms for the ORR and the OER on a bi-functional Mn oxide catalyst. First, the electrochemical characterization of a recently developed nano- structured manganese oxide catalyst demonstrates excellent ORR and OER activity, but suggests that the catalyst under- goes a change in the oxidation state in the onset region of ORR activity as well as in the potential region between ORR and OER activity. Attempts to characterize the surface oxida- tion state under operating conditions using ex situ X-ray photoelectron spectroscopy have yielded some information on the active surfaces involved.
35 To gain greater insight into
the surface chemistry of this catalyst during reaction condi- tions, we turn to DFT calculations to identify the precise surface structures involved as well as associated reaction mechanisms for both oxygen reduction and oxygen evolution.
Our study involves the following elements. The ﬁrst step is to
determine which surface structures of manganese oxide are present as a function of pH and applied potential versus the reversible hydrogen electrode (RHE). We obtain this information by employing DFT calculations to generate surface Pourbaix diagrams for diﬀerent adsorbate (e.g. O* and HO*) covered surfaces. To link the calculated surface structures of manganese oxide to ORR or OER conditions, we then use DFT to calculate binding energies for all reaction intermediates involved in the ORR and the OER and predict ORR and OER overpotentials for each surface structure. These overpotentials are then inserted into the previously developed Sabatier model
42 to produce a
computationally derived linear sweep voltammogram (LSV). The computationally derived LSV reveals activity as a function of applied potential for manganese oxide surfaces in a self- consistent manner, meaning that reaction turnover can only occur on surface phases that are identiﬁed to be present at a given potential. Our results indicate that the active surface for the ORR is 1/2 ML HO* covered Mn2O3 and for the OER, O* covered MnO2. Having identiﬁed the active surfaces involved in these reactions, our DFT calculations can oﬀer further insight into mechanistic pathways: the associative mechanism is the expected pathway for the ORR pathway on 1/2 ML HO* covered Mn2O3 and the direct recombination mechanism is the most likely OER pathway on O* covered MnO2.
Published on 18 September 2012. Downloaded by Swansea University on 17/10/2016 09:49:59.
View Article Online

pageNum = 3

14012 
 Phys. Chem. Chem. Phys., 2012, 
 14 , 14010–14022 
 This journal is c the Owner Societies 2012 
 When the computationally derived LSV is compared to 
 the experimental LSV measured on a recently developed nano-structured manganese oxide catalyst, the theoretical predictions closely match experimental onsets for ORR and OER catalytic activities. The close match between theory and experiment validates the application of a ﬁrst-principles theoretical analysis to the electrochemical oxygen reduction and oxygen evolution reactions on surfaces at the atomic level. By focusing our analysis on reaction energetics, namely the binding energies of reactive intermediates, we expect our approach to be robust and not very dependent on the computa- tional setup and the exchange and correlation functional applied in the DFT simulations. 
 2. 
 Methods 
 2.1 
 Computational methods 
 The spin-polarized DFT calculations are performed at the generalized gradient approximation (GGA) RPBE level 
 43 
 using the plane wave implementation in Dacapo and the Atomic Simulation Environment Ultra-soft pseudo-potentials are used to deal with the ion cores. 
 44 Therefore the electronic 
 wave-functions can be represented well by a plane wave basis set with a cutoﬀ energy of 350 eV. The electron density is treated on a grid corresponding to a plane wave cutoﬀ at 500 eV. A Fermi smearing of 0.1 eV and Pulay mixing are used to ensure the fast convergence of the self-consistent electron density. Atomic positions are relaxed until the sum of the absolute forces is less than 0.05 eV A˚ 
 1. For reference, the 
 calculated equilibrium lattice constants of MnOx are 4.5 A˚/ MnO, 5.78 A˚ (a), 9.59 A˚ (c)/Mn3O4, 9.51 A˚/a-Mn2O3 and 4.43 A˚ (a), 2.86 A˚ (c)/b-MnO2, in good agreement with the experimental measurements and previous DFT studies. 
 45–53 
 The starting point for this analysis is calculations on four 
 well-deﬁned manganese oxide surfaces (Fig. 1). For the OER and the ORR it is likely that the facets control surface activity rather than surface defects since defects are expected to be covered by oxygen at the very oxidizing conditions relevant for OER and ORR. In this work, we speciﬁcally consider four close packed MnOx surfaces 
 46 
 and examine their trends 
 in behavior: MnO(001), b-MnO2(110), Mn3O4(100) and a-Mn2O3(110). The surface structures with the most stable terminations are shown in Fig. 1. For Mn3O4 (in Fig. 1a) all the surface Mn atoms are equivalent and each Mn atom coordinates with four oxygen atoms in the same plane and one oxygen in the second layer (see Fig. 1a). The a-Mn2O3(110) surface has four diﬀerent types of Mn atoms (Fig. 1b): two Mn atoms coordinate with ﬁve oxygen atoms: four oxygen atoms in the same plane and one in the second layer (site 1), and three oxygen atoms in the same plane and two in the second layer (site 4). The other two atoms coordinate with four oxygen atoms: three oxygen atoms in the same plane and one oxygen in the second layer (site 2), and two oxygen atoms in the same plane and two in the second layer (site 3). b-MnO2 has a rutile phase 
 54,55 and two types of Mn atoms on the surface: ﬁve- 
 coordinated Mn (coordinated unsaturated site, site 1 in Fig. 1c), with four oxygens in the same plane and one in the second layer, and six-coordinated Mn (bridge site, site 3 in Fig. 1c) that is 
 considered to be the inactive sites. Our calculations show that the MnO(001) surface (Fig. 1d) reconstructs immediately in the presence of oxygen, and thus this oxide phase is not considered any further. 
 A periodically repeating 4–8 layer slab is employed in the 
 model to determine the most stable MnOx surfaces in our calculations (see Fig. 1). A vacuum of at least 20 A˚ is used to separate the slab from its periodic images. Supercells with periodicity (2 
  1) have been employed to simulate adsorption 
 and electrochemical reaction, with Monkhorst–Pack type of k 
 -point sampling of 4 
  4  1 for MnO(100) and b-MnO2(110), 
 and 2 
  4  1 for Mn3O4(001). For the complex crystal structure 
 of a-Mn2O3(110), only (1  1) unit cell and 2  3  1 Monkhorst–Pack type of k-point sampling are used. The 2–4 top layers as well as possible adsorbates are fully relaxed. 
 We apply a previously developed method, the computa- 
 tional standard hydrogen electrode (CSHE) for modeling the thermochemistry of electrochemical reactions. 
 37,38 
 In 
 this method the only way the potential aﬀects the relative free energy is through the chemical potential of the electrons in the electrode. This ‘‘ﬁrst order’’ inclusion of the potential has been used to predict the activity trends for the ORR on metal and metal alloys and in the design of electrocatalysts. 
 36,37 
 Furthermore, we have shown that thermochemical features such as phase diagrams in water are also well described by 
 Fig. 1 
 The schematic structures (top view) of diﬀerent manganese oxide 
 phases, Mn atoms in blue, O atoms in red. (a) Mn3O4(001) – white rectangle indicates the (2 
  1) unit cell with the equivalent ﬁve-fold coordinated active 
 sites 1,2,3,4; (b) Mn2O3(110) white rectangle indicates the (1  1) unit cell with four types of sites: 1 – ﬁve-fold coordinated (with four oxygen atoms in the same plane), 4 – ﬁve-fold coordinated (three oxygen atoms in the same plane and two in the second layer), 2 – four-fold coordinated (three oxygen atoms in the same plane and one in the second layer) and 3 – four-fold coordinated (two oxygen atoms in the same plane and two in the second layer), and (c) MnO2(110) surfaces – a rutile type stoichiometric surface. The dashed line indicates a (1 
  2) unit cell. Positions 1 and 2 are equivalent and 
 represent the active sites (cus). Sites 3 and 4 are equivalent six-fold coordinated and are called the bridge sites; (d) MnO(100) with (1 
  1) unit 
 cell. 1 and 2 are equivalent ﬁve-fold coordinated active sites. 
 Published on 18 September 2012. Downloaded by Swansea University on 17/10/2016 09:49:59. 
 View Article Online

pageNum = 4

This journal is c the Owner Societies 2012
Phys. Chem. Chem. Phys., 2012,
14 , 14010–14022 14013
this method.
56 The only eﬀect of the pH is the change of
chemical potential of the solvated protons. At standard conditions (pH = 0), H+(aq)+e
is in equilibrium with 1/2
H2(g) at zero potential vs. the SHE. At ﬁnite pH and potential vs.
SHE (USHE) the chemical potential of a proton and an
electron is:
(H
+(aq))+(e) = 1/2 H
2(g) – eUSHE+kBT ln10 pH
(2.1)
2.2
Pourbaix diagrams
To construct the surface Pourbaix diagram for the MnOx system, we ﬁrst generate a calculated bulk Pourbaix diagram by considering the equilibrium between diﬀerent bulk oxide phases and bulk manganese metal. In the same diagram we also include the equilibrium between diﬀerent soluble products and solid substances. As we start by exploring phase beha- viour in the bulk, all these values can be adapted from the Pourbaix atlas, the reference system is the bulk metallic phase.
13 More details on the construction of the diagrams
can be found in the ESI.w After constructing bulk Pourbaix diagrams which are equivalent to the diagrams found in the Pourbaix atlas, we go one step further and identify the adsorbates (e.g. O
* or HO*) that are present and their coverage
(e.g. 1/4 ML, 1/2 ML, etc.). We thus identify the most stable surface structures for each oxide phase at respective pH and potential, key information that is not found in the Pourbaix atlas. The calculations employed to determine the stable surfaces as a function of pH and potential employ a previously developed model.
41 In short, the surface is in equilibrium with protons and
liquid water at 298 K so that oxygen and hydroxyl may be exchanged between the surface and a reference electrolyte.
Consider a clean surface with a quantity of X* available sites
onto which oxygen or hydroxyl can potentially adsorb. At a given pH and potential, the surface will interact with the inter- facial water layer such that some of the H2O molecules at the interface will dissociate onto available sites, producing adsorbed O or HO and releasing protons and electrons in the process. We introduce the variables NO and NHO to reﬂect the number of adsorbed O and HO species, respectively, and the variable N* which represents the number of remaining free sites at the given pH and potential after the adsorption processes have occurred. Thus, the total number of available sites to begin with, X
¼
NO þ NHO þ N; i.e. after adsorption sites either contain O, HO, or remain free sites. The stoichiometric equation reﬂecting this process is as follows:
X
þ ðNO þ NHO ÞH2Oð1Þ ! ðNO þ NHO þ N
Þ
ads
þ ð2NO þ NHO ÞH
þ þ ð2N
O þ NHO Þe

ð2:2Þ
We can thus calculate the free energy change of the surface
covered with adsorbates relative to the clean surface, on the DFT scale as follows:
Gsurf ¼ E
DFT
NO þNHO þN
ð
Þads E
DFT
X
NO þ NHO
ð
ÞE
DFT
H2OðgÞ
þ
2NO þ NHO
ð
Þ
2
E
DFT
H2ðgÞ þ DZPE T DS
2NO þ NHO
ð
Þ eU kBT ln aHþ
ð
Þ
ð2:3Þ
where EDFT
ðNO ;NHO;NÞ
; EDFT
X
ð
Þ ; E
DFT
H2OðgÞ; E
DFT
H2ðgÞ are the calculated
ground state energies of the surface with the adsorbates, of the clean surface and of the references molecules in the gas phase. Zero point energies (ZPE) corrections are calculated using DFT calculations of the vibrational frequencies and standard tables of molecules. The changes in entropy (TS
0, T = 298 K)
are calculated from the standard tables for gas phase mole- cules .
57 Detailed description about how to perform all correc-
tions can be found in ref. 37 and 38.
2.3
Experimental methods
The electrochemical characterization was performed on a-Mn2O3 nanostructured thin ﬁlms electrodeposited onto polished glassy carbon disks (GC, 0.196 cm
2, SigradurG
HTW Hochtemperatur-Werkstoﬀe GmbH) as described previously.
31 The ﬁlms were characterized using cyclic voltam-
metry (CV) in a three electrode electrochemical cell in a rotating disk electrode (RDE, Pine Instruments) conﬁgu- ration. All CVs were iR-compensated and measured using a Bio-Logic potentiostat (VMP3) in 0.1 M KOH electrolyte, in nitrogen or oxygen saturated environments, with a scan rate of 5 mV s
1 and a rotation rate of 1600 rpm. Platinum wire was
used as a counter electrode and Hg/HgO electrode was used as a reference electrode. The potential scale was calibrated to a reversible hydrogen electrode (RHE) and all potentials are reported vs. RHE. CVs in nitrogen were used to identify oxida- tion state changes in an inert environment, while CVs in oxygen identiﬁed potentials relevant for the ORR and the OER. Base CVs in nitrogen and ORR CVs in oxygen were performed from 0.05 V to 1.1 V vs. RHE, while OER linear sweep voltammo- grams (LSV) were performed from 0.05 V to 1.9 V vs. RHE.
To compare ORR and OER activities of the nanostructured
a-Mn2O3 to active precious metals and metal oxides, electro- chemical characterization was also performed on commercial carbon-supported platinum (20 wt% Pt/C, Etek) and ruthe- nium (20 wt% Ru/C, Premetek) nanoparticles which were previously shown to have a comparable surface area to the nanostructured a-Mn2O3.
31 Catalyst dispersions of precious
metal nanoparticles were prepared by adopting a known literature procedure.
51 Brieﬂy, 14 mg of conditioned catalyst
powder were ultrasonically dispersed in 2 ml of isopropanol, 3 ml of Millipore water, and 20 ml of 5 wt% Naﬁon solution (Sigma-Aldrich). For characterization, 10 ml of the dispersed catalyst was drop-casted onto a polished glassy carbon electrode and allowed to dry in room air. To capture both ORR and OER activities in one linear sweep, characterization was performed between 0.05 V and 1.7 V for Ru/C, 1.9 V for a-Mn2O3, and 2.2 V for Pt/C. Diﬀerent anodic potentials were used in diﬀerent catalytic systems in order to reach an OER current of 10–20 mA cm
2 in each case; the highest value of
2.2 V used in Pt/C system was not applied to all other catalysts to mitigate carbon oxidation at high anodic potentials. Although the nanoparticles are prepared as metals, at high anodic potentials relevant to OER, the surface of the nano- particles is converted to a metal oxide. Consequently, while the ORR is observed on metal or on partially oxidized metal surfaces, the OER is observed on the electrochemically formed metal oxide surfaces.
Published on 18 September 2012. Downloaded by Swansea University on 17/10/2016 09:49:59.
View Article Online

pageNum = 5

14014
Phys. Chem. Chem. Phys., 2012,
14 , 14010–14022
This journal is c the Owner Societies 2012
3.
The oxygen reduction and oxygen evolution
reactions
In an acid environment the ORR and the OER can be written as:
O2 + 4H
+ + 4e
2 2H2O
(3.1)
We consider two possible ORR/OER reaction mechanisms:
an associative mechanism that involves a HOO* species, where * represents the active site on the metal surface, and a direct O2 dissociation/recombination mechanism.
In acid, the associative mechanism goes through the following
elementary steps (the ORR proceeds from top to bottom, eqn (3.2)
- eqn (3.5), where the OER proceeds from bottom
to top, eqn (3.5)
- eqn (3.2):
O2 + H
+ + e
2 HOO*
(3.2)
HOO* + H
+ + e
2 O* + H2O
(3.3)
O* + H
+ + e
2 HO*
(3.4)
HO* + H
+ + e
2 H2O
(3.5)
In an alkaline electrolyte, H2O rather than H3O
+ may act as
the proton donor, resulting in the overall ORR and OER equation:
O2 + 2H2O + 4e

2 4OH
(3.6)
The analogous associative mechanism in base is as follows:
O2 + H2O + e

2 HOO* + OH
(3.7)
HOO* + e

2 O* + OH
(3.8)
O* + H2O + e

2 HO* + OH
(3.9)
HO* + e

2 OH
(3.10)
Notice that the surface intermediates (HOO*, O*, HO*) are
the same in both environments and that they all contain at least one oxygen atom. It is through this oxygen that the intermediates bind to a Mn ion at the surface.
The mechanism via direct O2 dissociation/recombination
mechanism consists of the following elementary steps (for simplicity, only the steps in an acid environment are shown):
1/2O2
2 O*
(3.11)
O* + H
+
+ e

2 HO*
(3.12)
HO* + H
+ + e
2 H2O
(3.13)
The ORR and OER mechanisms considered in this study
neglect the eﬀect of the electric ﬁeld in the double layer and do not treat barriers which may depend on whether the proton donor is H2O or H3O
+. Thus, at a ﬁxed potential on the RHE
scale, there is no diﬀerence in the free energy of the ORR/OER intermediates calculated in acid versus in base for the following reasons: (1) all reactions involve the same intermediates and the same number of protons and electrons and (2) aH+ and aOH– are directly related by means of a pH/pOH scale since water is in equilibrium with H
+ and OH–.38 As such, we will
use the equations derived for the acid solution and apply them to a basic environment to be commensurate with the
experimental data in the base presented herein. Although this method cannot accurately model absolute kinetic rates, the consistent set of assumptions will allow for direct comparison of relative trends in activity. Despite the points made above, we note that for a number of catalyst systems the ORR activity has been found experimentally to be a function of pH. This could arise for a number of reasons, for instance the possibility of an O2
reaction pathway in which the step producing O
2

does not involve binding to the catalyst surface.
58 Never-
theless, for the most active catalysts like Pt and Ru, the ORR/OER overpotentials are not particularly sensitive to pH and do not proceed through an O2
reaction pathway.
4.
Results
4.1.
Electrochemical characterization
Fig. 2 shows experimental results from our electrochemical characterization of a nanostructured a-Mn2O3 electrode performed in nitrogen and oxygen saturated 0.1 M KOH. Three diﬀerent data sets are presented in the ﬁgure: (1) a base CV in a nitrogen-saturated environment, (2) a CV in an oxygen-saturated environment of the same potential window, and (3) a linear sweep voltammogram (LSV) in a wide potential window in an oxygen-saturated solution. The base CV performed in the nitrogen-saturated environment was used to identify oxidation/reduction features on the nanostructured a-Mn2O3 surface. As seen in the inset of the ﬁgure, two oxidation features are observed in the anodic sweep – one between 0.5 and 0.8 V and another between 0.8 and 1.0 V. These features likely correspond to the oxidation of Mn3O4 to Mn2O3 and Mn2O3 to MnO2, consistent with the thermo- dynamic standard potentials for these processes, which are 0.69 V and 1.01 V, respectively.
13
The reductive feature
occurring between 0.90 and 0.65 V in the cathodic sweep of the N2-saturated CV pertains to the discharge reaction of MnO2 to Mn2O3, as assigned in the literature.
59
Fig. 2
Electrochemical characterization of an a-Mn2O3 nano-
structured thin ﬁlm. Direct comparison of a base CV in nitrogen (also shown in the inset), a LSV in oxygen. Later in this paper the DFT-produced surface Pourbaix diagram of Fig. 6 shows that the relevant surface for the ORR is Mn2O3 and the relevant surface for the OER is MnO2.
Published on 18 September 2012. Downloaded by Swansea University on 17/10/2016 09:49:59.
View Article Online

pageNum = 6

This journal is c the Owner Societies 2012
Phys. Chem. Chem. Phys., 2012,
14 , 14010–14022 14015
The CV in the oxygen-saturated environment was used to
identify the onset potential for the ORR. The catalyst ﬁrst draws ORR current at approximately 0.83 V in the cathodic direction (0.80 V in the anodic direction) and reaches a diﬀusion limited current at 0.60 V. The onset potential of 0.83 V corresponds to an overpotential of 0.4 V, which indicates a highly active non-precious metal ORR catalyst. The experimental LSV reveals that the onset potential for the OER occurs at 1.5 V, which corresponds to an overpotential of 0.27 V and provides evidence of high catalytic activity for the OER. The location of the second oxidation feature between 0.8 V and 1.0 V suggests that during the anodic sweep, the catalyst changes its oxidation state in the potential region between ORR and OER activity. Additionally, since the ORR region is located at the end of the reduction feature seen in the cathodic sweep of the base CV (0.90 V to 0.65 V) and overlaps with an oxidation feature in the anodic sweep of the base CV (0.50 V to 0.80 V), the active surface may undergo changes at diﬀerent potentials of ORR activity. Spectroscopic methods have been used to study oxidation state changes in MnOx, however speciﬁc identiﬁcation of the surface phases has remained elusive.
35 To identify these phases, we employ theory.
4.2.
DFT studies of stable surface structures of MnOx as a
function of pH, applied potential, and starting bulk material
We aim to understand how MnOx bulk and surface structures change across the pH-potential window, and how these changes impact ORR and OER activity. To do so, we ﬁrst investigate the relative stability of diﬀerent adsorbate surface structures for each of three diﬀerent bulk oxide phases: (a) Mn3O4(001), (b) Mn2O3(110) and (c) MnO2(110). The MnOx phase, crystal structure, and the surface adsorbates present during reaction conditions (pH and applied potential) will likely depend on how the material was synthesized in the ﬁrst place, e.g. starting MnOx crystal structure, nanoparticle size,
60
etc.
Here we present a thermodynamic analysis for all possible
bulk and surface structures. Though only one combination of a bulk and surface structure can be the most thermodynami- cally stable at a given pH, temperature, and applied potential, it is possible that other structures might be present due to kinetic control, and thus those structures could also contribute to OER and ORR activities.
13
The calculated free energies for all possible surface adsor-
bate structures on each of the three bulk structures (Mn3O4, Mn2O3, and MnO2) are plotted versus potential at pH = 0, shown in Fig. S1 in the ESI.w The structure with the lowest free energy at a given potential determines the most likely surface structure as it is the most thermodynamically stable. Fig. 3 then incorporates the eﬀect of pH to produce three surface Pourbaix diagrams in which the most stable surface for each bulk oxide is constructed as a function of pH and the electrode potential vs. SHE (USHE). Later in Section 5, we ultimately combine this information along with thermo- dynamic data for the bulk oxide phases in order to construct a single General Surface Pourbaix diagram that allows for phase changes both at the surface as well as deeper within the bulk of the near-surface region. We ﬁrst discuss details of the surface changes for each of the bulk MnOx phases, as shown in Fig. 3.
Fig. 3 shows that at low potentials in acidic solutions,
dissolution to Mn
2+ is spontaneous for all MnO
x bulk phases.
In alkaline solutions this process is suppressed, and instability is not as problematic as it is in acidic solutions. In both types of electrolytes, corrosion is most severe at potentials higher than 1.46 V (RHE) where the MnOx can be oxidized and dissolved into MnO4
. We thus focus our discussion on
the alkaline environment and within that region identify the most stable surface structures as a function of potential on (a) Mn3O4, (b) Mn2O3, and (c) MnO2. Note that in Fig. 3
Fig. 3
Surface Pourbaix diagram on (a) Mn3O4(001), (b) Mn2O3(110),
and (c) MnO2(110). Lines a and b represent the reversible hydrogen electrode (RHE) line and the O2/H2O equilibrium line. The notation ‘‘b’’ within the Fig. 3(c) legend represents the adsorbates at the bridge sites and coordinated unsaturated sites.
Published on 18 September 2012. Downloaded by Swansea University on 17/10/2016 09:49:59.
View Article Online

pageNum = 7

14016
Phys. Chem. Chem. Phys., 2012,
14 , 14010–14022
This journal is c the Owner Societies 2012
the potential versus RHE at any pH can be read oﬀ as the vertical distance from line ‘a’ to the region of interest.
(a) Mn3O4(001), Fig. 3(a): for the case of Mn3O4(001), at
low potentials (0.46 V
o URHE o 0.95 V) the clean surface
(i.e. no adsorbates) is the most stable surface structure. As the potential increases (0.95
o URHE o 1.29 V), water dissocia-
tion begins, leading to the presence of HO* with increasing coverage, e.g. 1/4 ML HO/Mn3O4 and 1 ML HO/Mn3O4. At potentials URHE > 1.29 V the adsorbed hydroxyls are oxidized further to O* to form 1ML O/Mn3O4.
(b) Mn2O3(110), Fig. 3(b): for Mn2O3(110), water will
dissociate to HO* at potentials URHE > 0.53 V, with HO* coverage increasing from 1/4-1ML HO/Mn2O3 all the way to 1 ML HO/Mn2O3 by URHE = 1.23 V. At this point, the hydroxyls are oxidized further to produce 1 ML O/Mn2O3.
(c) MnO2(110), Fig. 3(c): at low potentials (0.78 V
o URHE o
1.1 V), the surface Pourbaix diagram of MnO2(110) shows that the bridge sites of MnO2(110) are occupied by HO* (2OHb/MnO2). The bridge HO* then gradually dissociates into O* within the potential region of 1.1 V
o URHE o 1.38 V
(Ob+OHb/MnO2 and 2Ob/MnO2). At higher potentials O* adsorbs at the coordinated-unsaturated sites to form 3O/MnO2 and 4O/MnO2.
4.3.
Activity of stable MnOx surface structures for the
OER/ORR
Having identiﬁed the most thermodynamically stable surface structures as a function of pH and potential for each of the bulk phases of MnOx, we now look to identify which of those surfaces are likely to be present during OER/ORR operating conditions. To accomplish this goal, we ﬁrst use ORR/OER free energy diagrams generated by DFT to calculate the theoretical overpotentials for OER/ORR on all the relevant MnOx surfaces (for more details see Fig. S2 and S3 in ESIw). The ‘‘theoretical overpotential’’ to which we are referring is the overpotential beyond which all reaction steps become thermodynamically downhill. The ‘‘theoretical overpotential’’ is related to, but not identical to, the ‘‘onset’’ potential that is often used as a ﬁgure of merit in experimental LSVs. Previously described kinetic models of electrocatalytic reactions show that the experimental ‘‘onset’’ potential is expected to occur approximately 0.15 V prior to the ‘‘theoretical over- potential.’’
42 Our calculations of the reaction energetics for
the OER/ORR are not shown for every possible surface conﬁguration in Fig. 3, but rather only for the ‘self-consistent’ catalytic surfaces; that is, the surfaces that are thermodynami- cally stable, according to the Pourbaix diagrams of Fig. 3, at the overpotential
at which catalyst is operating.
At the high potentials required to drive the OER, the self-
consistent surfaces for each of the three MnOx bulk phases are quite similar – they are all completely covered by oxygen. This is in agreement with our previous work investigating the OER on rutile oxide surfaces. Since the oxide surfaces are covered with oxygen at OER relevant potentials, no active sites are available for water adsorption,
38 and thus the eﬀect of water
and its interactions with adsorbed reaction intermediates can be neglected. This simpliﬁes the analysis of OER reaction energetics substantially.
At ORR potentials, however, the eﬀect of water cannot be
neglected as there are available sites for water to adsorb and potentially dissociate into HO* and O*. This leads to considerably diﬀerent MnOx adsorbate surface structures for each of the three bulk structures, namely clean Mn3O4(001), 1/2 ML HO* covered Mn2O3(110) and MnO2(110) with HO* at bridge sites as spectators. Notice that for all of these surfaces, there are empty sites where water can adsorb and impact the adsorption energies of ORR intermediates, parti- cularly with HO* and HOO* as these adsorbates can form H bonds with adjacent water molecules. Therefore, the eﬀect of water is included in the free energy diagram for intermediates involved in the ORR.
Detailed studies on metals
61
have shown that water
stabilizes surrounding HO* and HOO* species by
0.3 eV.
In our preliminary studies of this eﬀect for metal oxide surfaces, we investigated a single neighboring water molecule interacting with HO* and HOO* adsorption on an MnO2(110) surface. We obtained similar stabilization eﬀects of
0.5 eV
and
0.35 eV, respectively. For the purposes of this work, we
choose to use
0.3 eV for the stabilization eﬀect of water on
both HO* and HOO* intermediates and note that more detailed studies of the eﬀects of water at metal oxide interfaces will be considered in future studies.
We note that in this work we identify surface structures
based solely on static equilibrium considerations. Under reac- tion conditions the local coverage of reaction intermediates is in a very dynamic state, and these dynamics could very well play a role on the reaction chemistry. For low rates of reaction, however (i.e. near the experimental ‘onset’ potential, which occurs before the theoretical overpotential), the surface Pourbaix diagram is a good model for determining the self- consistent surface.
Free energy diagrams constructed for the self-consistent
surfaces, shown in Fig. 4, provide insight into the mechanistic pathways involved in oxygen reduction and oxygen evolution. They also point out the source of reaction overpotentials for each surface, exactly the kind of information needed to facilitate the development of improved catalysts.
The free energy calculations for the OER indicate that for
O* covered Mn3O4(001) and O* covered Mn2O3(110), the associative pathway is energetically favorable compared to the direct mechanism. For the O* covered MnO2(110) surface, however, the direct pathway is slightly favored. Previous studies have shown that the OER activity of metal oxides follow a ’volcano’ relationship based on scaling relations that correlate binding energies for the diﬀerent reaction inter- mediates.
62 The O* covered MnO
2(110) surface is close to the
top of the volcano where the intermediates have a better compromise in interaction strength, which results in a more ﬂexible reaction mechanism. In this case, the direct mechanism by recombination of oxygen atoms has a slightly lower free energy than associative mechanism by only 0.08 eV, as described in the Fig. S3(b) (ESIw).
To gain more information about the sources of over-
potential on the self-consistent MnOx surfaces, we compared those free energy diagrams with that of an ideal oxygen evolution/reduction catalyst, shown in Fig. 4(a). The ideal catalyst is deﬁned by a free energy reaction diagram in which
Published on 18 September 2012. Downloaded by Swansea University on 17/10/2016 09:49:59.
View Article Online

pageNum = 8

This journal is c the Owner Societies 2012 
 Phys. Chem. Chem. Phys., 2012, 
 14 , 14010–14022 14017 
 the four charge transfer steps have identical reaction free energies of 1.23 eV = 4.92 eV/4 for an electrode held at USHE = 0. 
 If one is able to tune the binding energy of each intermediate on a surface to achieve this optimal situation, that catalyst surface would approach the activity of an ideal oxygen electrode. How- ever, as illustrated in recent work, 
 62,63 there is a universal scaling 
 relationship on a wide range of metals and oxides that governs the binding energy of the HOO* intermediate with respect to HO*, resulting in an approximately constant diﬀerence between the two energy levels (DGHOO  DGHO  3:2 eV). This is far oﬀ of an optimal catalyst which would exhibit an energy diﬀerence of 2.46 eV (2e 
  1.23V) between those two particular 
 energy levels. Thus, the ‘universal’ 3.2 eV energy diﬀerence between HOO* and HO* levels can be used to deﬁne the lowest possible ‘‘theoretical overpotential’’ for the OER and the ORR [(3.2 eV 
 2.46 eV)/2e 
 E 0.37 V] on a wide variety of 
 materials. The scaling relationship between HOO* and HO* holds for MnOx just as well, as shown in Fig. 4(b), (c) and (d), with values of 3.18 eV, 3.1 eV and 3.12 eV. The slight deviation of DGHOO  DGHO from 3.2 eV can be attributed to adsor- bate coverage eﬀects. 
 Indeed, the scaling relationship between the HOO* and 
 HO* binding energies explains one major source of reaction overpotential, however additional sources of overpotential can also arise from sub-optimal O* binding. It has been previously shown that the potential-determining step for the OER is either the second water dissociation step eqn (3.3) or the HO* oxidation step eqn (3.4). 
 38 Both steps involve O* and 
 either HOO* or HO*; as the latter two species scale linearly with one another, the expression (DGO  DGHO ) contains information regarding the binding energies for all three species and is introduced as the universal descriptor of oxygen evolu- tion activities. 
 We can see from Fig. 4(b) and (c) that for the OER, the O* 
 covered 
 Mn3O4(001) and Mn2O3(110) have the same 
 potential-determining step, the second water dissociation step eqn (3.3) in which the third (of four) H 
 + and e are trans- 
 ferred. The O* covered Mn3O4(001) surface exhibits a lower ‘‘theoretical overpotential’’ than the O* covered Mn2O3(110) surface, 0.6 V vs. 0.79 V. This originates from the placement of the O* energy level with respect to the energy levels of the intermediates below (HO*) and above (HOO*) it; closer to half-way is better in order to prevent large uphill reaction steps. For the O* covered MnO2 surface, the second water dissociation eqn (3.3) is also the potential-determining step, according to the associative mechanism (see Fig. 4(d)), leading to a ‘‘theoretical overpotential’’ of 0.6 V. 
 For the ORR, the same scaling relationship holds between 
 HOO* and HO*. Thus, much like the OER, one part of the ORR overpotential originates from this correlation while the other part arises from sub-optimal O* binding. The free energy diagrams of the intermediates for the ORR on the self- consistent MnOx surfaces are shown in Fig. 5. Our previous studies have shown that the potential-determining ORR step is either the formation of HOO* eqn (3.2) or the reduction of HO* eqn (3.5). 
 38 As HO* and HOO* scale linearly with one 
 another, DGHO can be introduced as a universal descriptor of oxygen reduction activities. We can see that all three self- consistent MnOx surfaces – clean Mn3O4(001), 1/2 ML HO* covered Mn2O3(110) and 1/2 ML HO* (bridge) MnO2(110) – are active for the ORR. The potential-determining step is HO* 
 Fig. 4 
 Free-energy diagram for the oxygen evolution reaction on (a) 
 the perfect catalyst, and O covered (b) Mn3O4(001), (c) Mn2O3(110) and (d) MnO2(110) at U = 0, pH = 0 and T = 298 K. DGHOO  DGHO (vertical solid lines) values of the three manganese oxides in (b), (c), and (d), are close to 3.2 eV, the average value found on a wide range of metals and oxides as shown in our recent paper in ref 56. The optimum value on the perfect catalyst is 2.46 eV. 
 Published on 18 September 2012. Downloaded by Swansea University on 17/10/2016 09:49:59. 
 View Article Online

pageNum = 9

14018
Phys. Chem. Chem. Phys., 2012,
14 , 14010–14022
This journal is c the Owner Societies 2012
reduction for the ﬁrst two oxides while for MnO2 it is HOO* formation. Calculated ‘‘theoretical overpotentials’’ are approximately 0.55–0.57 V for all cases.
As mentioned above, water can stabilize ORR intermediates
during reaction and, as a result, changes in water coverage and the number of hydrogen bonds could inﬂuence the calculated overpotentials. A more detailed study of water adsorption and its eﬀects on stabilizing reaction intermediates would provide a more accurate estimate of ‘‘theoretical overpotentials.’’
5.
Discussion
The Pourbaix diagrams generated in Fig. 3 describe changes in adsorbate surface structure as a function of pH and potential for three diﬀerent MnOx bulk structures. One question that arises is what if the bulk structure of the catalyst could change with respect to pH and potential? In Fig. 6, we present a general MnOx Pourbaix diagram that accounts for phase transitions in both the bulk (e.g. in the near-surface region) and at the very surface of the catalyst. As seen in Fig. 6, from 0.46 V
o URHE o 0.69 V, the most stable MnOx bulk/surface is
a clean (adsorbate-free) Mn3O4(001) surface. From 0.69 V
o
URHE
o 0.98 V, the material is oxidized into 1/2 ML HO*
covered Mn2O3(110), assuming no kinetic diﬃculties. From 0.98 V
o URHE o 1.01 V the Mn2O3(110) surface remains,
however HO* coverage increases to 3
4 ML. Between 1.01 V o
URHE
o 1.21 V, the bulk is oxidized from Mn2O3 to
MnO2(110) and at the surface, bridge sites are covered with HO*. As the potential increases above 1.21 V, the surface is further oxidized until the surface is completely covered by O*. At even higher potentials the MnO4
dissolution becomes
thermodynamically favourable at any pH.
In combination with the ‘‘theoretical overpotentials’’ for the
ORR and the OER on relevant surface structures of Mn3O4(001), Mn2O3(110), MnO2(110) as described in Fig. 4 and Fig. 5, we can use the DFT-calculated general MnOx Pourbaix diagram to identify the active surfaces during operating conditions. We ﬁnd that for the ORR, the active surface in the onset region is a 1/2 ML HO* covered Mn2O3(110), while for the OER, the active surface is O* covered MnO2(110). The predicted change in the oxidation state from Mn(III) in Mn2O3 to Mn(IV) in MnO2 in the potential region between the ORR and the OER is supported by the oxidation feature between 0.8 V and 1.0 V seen in the anodic sweep of the base CV (Fig. 2).
‘‘Theoretical onset potentials’’ for the ORR and OER were
calculated for these two surfaces by subtracting 0.15 V from the ‘‘theoretical overpotentials’’ based on the kinetic models described earlier, resulting in calculated values of 0.40 V and 0.45 V, respectively. These values are in good agreement with the experimentally observed onset potentials of 0.4 for the ORR and 0.27 for the OER measured on the nanostructured a-Mn2O3 electrocatalyst.
To visually relate theoretical predictions of ORR/OER
activities to experimental results, we used the Sabatier model
42
to create theoretical LSVs for the ORR and the OER (Fig. 7). Constructing the theoretical LSVs could only be possible by having ﬁrst identiﬁed the most thermodynamically stable bulk and surface structures present during reaction conditions.
In producing these theoretical LSVs, diﬀusion limitations for the ORR are included by invoking the Koutecky–Levich equation for a rotating disk at 1600 RPM.
64 Fig. 7(a) shows theoretical LSVs
for the self-consistent surface structures pertaining to bulk Mn3O4(001), Mn2O3(110) and MnO2(110), constructed as if no changes in bulk MnOx stoichiometry were induced by the electro- chemical potential. In other words, the bulk structure was ﬁxed throughout the entire potential window – only the surface was allowed to change as shown in Fig. 3. Fig. 7(a) thus reveals the intrinsic catalytic activities of bulk Mn3O4(001), Mn2O3(110) and MnO2(110) structures. All three bulk MnOx structures are
Fig. 5
Free-energy diagram for oxygen reduction on (a) Mn3O4(001),
(b) 1/2 ML HO* covered Mn2O3(110) and (c) MnO2(110) with HO* at bridge sites as spectators at U = 0, pH = 0 and T = 298 K.
Published on 18 September 2012. Downloaded by Swansea University on 17/10/2016 09:49:59.
View Article Online

pageNum = 10

This journal is c the Owner Societies 2012
Phys. Chem. Chem. Phys., 2012,
14 , 14010–14022 14019
shown to be highly active ORR catalysts. However, for the OER only O* covered MnO2(110) and O* covered Mn3O4(001) are highly active.
Fig. 7(b) shows a theoretical LSV in which both the bulk
structure and the surface structure are allowed to change with applied potential. Fig. 7(b) is most relevant for MnOx catalysts with negligible kinetic barriers to phase changes in the near- surface region, thus allowing both the bulk and the surface to reach their thermodynamically stable structures. For such a catalyst, the ORR has two relevant active surfaces. At the ORR onset potential of 0.83 V, the active surface is a 1/2 ML HO* covered Mn2O3(110). However, as the potential decreases below 0.69 V and the current approaches diﬀusion- limited values, the 1/2 ML HO* covered Mn2O3(110) is predicted to be reduced to clean Mn3O4(001) surface. This DFT-predicted change in the oxidation state of MnOx is supported by the oxidation feature between 0.5 V and 0.8 V seen in the anodic sweep of the base CV, Fig. 2.
In the OER region of Fig. 7(b), the theoretical OER activity
of the self-consistent MnOx surface is also shown. At the high potentials of the OER, O* covered MnO2 is the expected bulk- surface combination. Fig. 7(b) also compares the theoretical LSVs of MnOx in both the ORR and OER regions to those of self-consistent Ru and Pt, in which phase transitions to RuO2 and PtO2 at oxidative potentials were taken into account.
13
According to theoretical LSVs shown in Fig. 7(b), the pre- dicted activity order for the OER is RuO2 > MnO2 > PtO2, and for the ORR is Pt > Mn2O3 > Ru. We note that that this same model has previously been successful in predicting the trends in ORR activity for metal-alloy catalysts.
36,39 We now
turn our attention to comparing theoretical predictions with experimental measurements for MnOx, Pt, and Ru catalysts.
Fig. 7(c) shows experimental LSVs for the nanostructured
a-Mn2O3, Ru/C and Pt/C. Pt/C demonstrates the best ORR activity, while the oxidized Ru/C demonstrates the best OER activity. The nanostructured a-Mn2O3 shows high activity for both reactions. Under reductive potentials relevant to the ORR, the Mn2O3 surface outperforms Ru/C and approaches activity of Pt/C, while under oxidative potentials relevant to the OER, the MnO2 surface outperforms the oxidized Pt/C
and approaches the activity of the oxidized Ru/C. For both the ORR and the OER, the experimental activity trends are identical to those predicted by the DFT models. There is also excellent quantitative agreement between theory and experiment.
Fig. 6
General surface Pourbaix diagram for MnOx catalysts. The
oxidation state of the surface and the ORR and OER potential are constant versus the reversible hydrogen electrode (RHE). Lines a and b represent the RHE line and the O2/H2O equilibrium line.
Fig. 7
Calculated current density for (a) Mn3O4, Mn2O3 and MnO2;
(b) self-consistent curves from DFT calculation for MnOx, Ru and Pt; (c) experimental curves for MnOx, Ru and Pt.
Published on 18 September 2012. Downloaded by Swansea University on 17/10/2016 09:49:59.
View Article Online

pageNum = 11

14020
Phys. Chem. Chem. Phys., 2012,
14 , 14010–14022
This journal is c the Owner Societies 2012
Only minor discrepancies are observed between the theore-
tically-predicted and the experimentally-observed onset poten- tials for MnOx. One possible source of the discrepancies could be due the assumptions made about the number of oxygen atoms, NO, coordinated to each Mn atom at the MnOx surface in the DFT calculations. This theoretical study focuses solely on b-MnO2 amongst the MnO2 family as this is the most stable of its phases. However, the presence of a-MnO2 and g-MnO2 phases can be expected in experimental MnO2 electrodes,
19,21,22,26,28 and
the coordination environment of these phases will exhibit diﬀerent values of NO. Furthermore, the theoretical calculations of the catalyst surface structure as a function of electrochemical potential (Fig. 3) examined the changes in the MnOx structure only in the top-most layer. During the experiments, however, it is quite possible that complete or incomplete stoichiometric changes could penetrate deeper into the material and impact NO, which can lead to signiﬁcant variation in electrochemical activity. Some, but not all, of the possibilities are accounted for in Fig. 6, the general Pourbaix diagram. Fig. 8 shows the origin of this particular eﬀect, exhibiting the relationships among (1) the free energy of HO* (DGHO , descriptor for ORR activity), (2) the free energy diﬀerence between O* and HO* (DGO DGHO , descriptor for OER activity), and (3) NO, the number of oxygen atoms coordinated to surface Mn (see ESI for a sample calculation of NO). For the case of Mn2O3 surfaces, as the oxygen coordination number increases, DGHO changes only marginally, while DGO DGHO changes by 0.7 eV, which means that the catalytic activities of manganese oxides can sometimes, but not always, be sensitive to oxygen coordination at the surface.
The close match between theoretical predictions and experi-
mental results suggests that we have successfully modelled the surface structure of MnOx catalysts, and in particular how metal oxide surfaces change with pH and applied electro- chemical potential. We have also successfully simulated the catalytic activity of those surfaces for the ORR and the OER, having identiﬁed the active surface structure as well as the reaction pathways involved. This insight gained from DFT calculations can now be used to improve the design principles
for OER/ORR catalysts. For example, our theoretical calcula- tions have identiﬁed that the stabilization of intermediates through hydrogen bonds with water is an important contri- butor to ORR overpotential on MnOx catalysts. Therefore, a rational design of more hydrophobic catalyst structures, resulting in a reduced number of water molecules adsorbed on the surface, can lead to a signiﬁcant improvement in ORR activity of MnOx catalysts. Our calculations have also demon- strated that an O* covered MnO2 surface is close to the top of OER volcano. Therefore, to improve the activity of O* covered MnO2 for OER, it will be necessary to modify the catalyst surface in such a way as to break the scaling relation- ship between the energies of HOO* and HO* intermediates. Finally, our ﬁnding that the number of oxygen atoms coordi- nated to each Mn atom at the MnOx surface has a signiﬁcant impact on the binding energy of reaction intermediates suggests that manipulation of the surface coordination environment, through approaches such as nanostructuring, doping, and alloying, can also lead to improved manganese oxide electrocatalysts for the ORR and the OER.
6.
Conclusions
The surface electrochemistry of metal oxide catalysts is complex. Phase changes are prevalent both at the surface and in the near-surface region that depend greatly on pH and applied potential. And the structure of the material, both at the surface and within the bulk, has a signiﬁcant inﬂuence on catalyst activity. In this work, we combine theory and experiment to understand this chemistry for the speciﬁc case of MnOx materials that catalyze the ORR and the OER. The theoretical models developed in this work, however, are more broadly applicable to other metal oxides as well as to other electrocatalytic reactions.
Experimentally, we have shown that a nanostructured
a-Mn2O3 is an excellent bi-functional catalyst for the ORR and the OER, and that the catalyst likely undergoes phase changes at the surface as a function of applied potential, in particular at ORR potentials and in the potential window between the ORR and the OER. In an eﬀort to understand surface changes under reaction conditions, as well as how they impact catalytic activity and reaction pathways for both reactions, we developed theoretical models using density func- tional theory (DFT). DFT calculations were employed to construct surface Pourbaix diagrams for MnOx and then to identify ‘‘theoretical overpotentials’’ for the surfaces present during reaction conditions across the pH and potential window. Our calculations reveal that the active surfaces for the ORR and the OER are 1/2 ML HO* covered Mn2O3 and O* covered MnO2, respectively. As shown in Fig. 7(b), this phase transition between the two operating conditions is beneﬁcial in that MnO2 is a better catalytic surface for the OER than Mn2O3. Thus an active catalyst phase is formed under each of the two reaction conditions.
The calculations also suggest mechanistic pathways for the
ORR and the OER on the relevant surface structures: the ORR proceeds by the associative pathway, while for the OER, the direct pathway is favored slightly. With these calculations we were able to construct theoretical LSVs for MnOx which
Fig. 8
The free energy of HO* (DGHO , solid circle) and the free
energy diﬀerence between O* and HO* (DGO DGHO , open circle) plot against the number of O (NO) coordinated with Mn on Mn2O3(110) and MnO2(110). I, II and III represent three diﬀerent types of Mn atoms on the Mn2O3(110) surface respectively.
Published on 18 September 2012. Downloaded by Swansea University on 17/10/2016 09:49:59.
View Article Online

pageNum = 12

This journal is c the Owner Societies 2012 
 Phys. Chem. Chem. Phys., 2012, 
 14 , 14010–14022 14021 
 make predictions on catalytic activity for the ORR and the OER. The theoretical LSVs agree well with the experimental LSVs measured on the active nanostructured MnOx bi-functional catalyst; the close match between theory and experiment suggests that the theoretical model is accurate and robust. 
 By combining ﬁrst-principles theoretical analysis and 
 experimental methods, atomic-level insight into the catalyst chemistry can be achieved. This allows one to determine principles for improving catalyst design. For the ORR, our DFT model predicts that decreasing the surface’s aﬃnity for water adsorption should signiﬁcantly increase catalytic activity, as it is desirable to destabilize the reaction inter- mediates HOO* and HO*. For the OER, our calculations show that to improve the activity of MnOx, it is necessary to design a surface structure that can break the scaling relation- ship between the energies of HOO* and HO* intermediates. If future ORR and OER catalysts are developed with these design principles in mind, superior activity for both reactions can be achieved. 
 Acknowledgements 
 We gratefully acknowledge funding from the Danish Strategic Research Council’s HyCycle program, the Danish Council for Technology and Innovation’s FTP program. This research was supported in part by the European Commission (Marie Curie Research Training Network MRTN-CT-2006-032474), by the Danish Council for Strategic Research via SERC project through grant no. 2104-06-011 and by the Catalysis for Sustainable Energy (CASE) initiative. This work was partially supported by the IMI Program of the National Science Foundation under Award No. DMR 0843934. YG, TFJ, and JKN were supported by the Center on Nanostruc- turing for Eﬃcient Energy Conversion (CNEEC) at Stanford University, an Energy Frontier Research Center funded by the U.S. Department of Energy, Oﬃce of Science, Oﬃce of Basic Energy Sciences under Award Number DE-SC0001060. 
 Notes and References 
 1 R. Forgie, G. Bugosh, K. C. Neyerlin, Z. C. Liu and P. Strasser, 
 Electrochem. Solid-State Lett. 
 , 2010, 13, D36–D39. 
 2 H. Dau, C. Limberg, T. Reier, M. Risch, S. Roggan and 
 P. Strasser, ChemCatChem, 2010, 2, 724–761. 
 3 M. W. Kanan and D. G. Nocera, Science, 2008, 321, 1072–1075. 4 K. Macounova, M. Makarova and P. Krtil, Electrochem. Commun., 
 2009, 11, 1865–1868. 
 5 M. V. Makarova, J. Jirkovsky, M. Klementova, I. Jirka, K. Macounova 
 and P. Krtil, Electrochim. Acta, 2008, 53, 2656–2664. 
 6 G. Y. Chen, S. R. Bare and T. E. Mallouk, J. Electrochem. Soc., 
 2002, 149, A1092–A1099. 
 7 N. S. Lewis and D. G. Nocera, Proc. Natl. Acad. Sci. U. S. A., 
 2006, 103, 15729–15735. 
 8 S. Chretien and H. Metiu, J. Chem. Phys., 2008, 129, 074705, DOI: 
 10.1063/1.2956506. 
 9 P. Mani, R. Srivastava and P. Strasser, J. Phys. Chem. C, 2008, 
 112, 2770–2778. 
 10 T. Ioroi, N. Kitazawa, K. Yasuda, Y. Yamamoto and H. Takenaka, 
 J. Appl. Electrochem. 
 , 2001, 31, 1179–1183. 
 11 H. Liu, B. L. Yi, M. Hou, J. F. Wu, Z. J. Hou and H. M. Zhang, 
 Electrochem. Solid-State Lett. 
 , 2004, 7, A56–A59. 
 12 L. L. Swette, A. B. Laconti and S. A. McCatty, J. Power Sources, 
 1994, 47, 343–351. 
 13 M. Pourbaix, Atlas of Electrochemical Equilibria in Aqueous Solutions, 
 Pergamon Press, New York, 1966. 
 14 K. N. Ferreira, T. M. Iverson, K. Maghlaoui, J. Barber and 
 S. Iwata, Science, 2004, 303, 1831–1838. 
 15 B. Loll, J. Kern, W. Saenger, A. Zouni and J. Biesiadka, Nature, 
 2005, 438, 1040–1044. 
 16 J. Rossmeisl, K. Dimitrievski, P. Siegbahn and J. K. Norskov, 
 J. Phys. Chem. C 
 , 2007, 111, 18821–18823. 
 17 P. E. M. Siegbahn and M. Lundberg, J. Inorg. Biochem., 2006, 100, 
 1035–1040. 
 18 M. L. Calegaro, F. H. B. Lima and E. A. Ticianelli, J. Power 
 Sources 
 , 2006, 158, 735–739. 
 19 Y. L. Cao, H. X. Yang, X. P. Ai and L. F. Xiao, J. Electroanal. 
 Chem. 
 , 2003, 557, 127–134. 
 20 F. Y. Cheng, Y. Su, J. Liang, Z. L. Tao and J. Chen, Chem. Mater., 
 2010, 22, 898–905. 
 21 M. I. A. M. S. El-Deab, A. M. Mohammad and T. Ohsaka, 
 Electrochem. Commun. 
 , 2007, 9, 2082–2087. 
 22 M. S. El-Deab and T. Ohsaka, J. Electrochem. Soc., 2008, 155, 
 D14–D21. 
 23 F. Jiao and H. Frei, Chem. Commun., 2010, 46, 2920–2922. 24 F. H. B. Lima, M. L. Calegaro and E. A. Ticianelli, J. Electroanal. 
 Chem. 
 , 2006, 590, 152–160. 
 25 F. H. B. Lima, M. L. Calegaro and E. A. Ticianelli, Electrochim. 
 Acta 
 , 2007, 52, 3732–3738. 
 26 A. M. Mohammad, M. I. Awad, M. S. El-Deab, T. Okajima and 
 T. Ohsaka, Electrochim. Acta, 2008, 53, 4351–4358. 
 27 I. Roche, E. Chainet, M. Chatenet and J. Vondrak, J. Phys. Chem. 
 C 
 , 2007, 111, 1434–1443. 
 28 L. Q. Mao, T. Sotomura, K. Nakatsu, N. Koshiba, D. Zhang and 
 T. Ohsaka, J. Electrochem. Soc., 2002, 149, A504–A507. 
 29 L. Q. Mao, D. Zhang, T. Sotomura, K. Nakatsu, N. Koshiba and 
 T. Ohsaka, Electrochim. Acta, 2003, 48, 1015–1021. 
 30 J. S. Yang and J. J. Xu, Electrochem. Commun., 2003, 5, 306–311. 31 Y. Gorlin and T. F. Jaramillo, J. Am. Chem. Soc., 2010, 132, 
 13612–13614. 
 32 N. Ramaswamy, R. J. Allen and S. Mukerjee, J. Phys. Chem. C, 
 2011, 115, 12650–12664. 
 33 N. Sivasankar, W. W. Weare and H. Frei, J. Am. Chem. Soc., 2011, 
 133, 12976–12979. 
 34 B. S. Yeo and A. T. Bell, J. Am. Chem. Soc., 2011, 133, 5587–5593. 35 Y. Gorlin and T. F. Jaramillo, ECS Trans., 2011, 41, 1701–1707. 36 J. Greeley, I. E. L. Stephens, A. S. Bondarenko, T. P. Johansson, 
 H. A. Hansen, T. F. Jaramillo, J. Rossmeisl, I. Chorkendorﬀ and J. K. Norskov, Nat. Chem., 2009, 1, 552–556. 
 37 J. K. Norskov, J. Rossmeisl, A. Logadottir, L. Lindqvist, 
 J. R. Kitchin, T. Bligaard and H. Jonsson, J. Phys. Chem. B, 2004, 108, 17886–17892. 
 38 J. Rossmeisl, Z. W. Qu, H. Zhu, G. J. Kroes and J. K. Norskov, 
 J. Electroanal. Chem. 
 , 2007, 607, 83–89. 
 39 V. Stamenkovic, B. S. Mun, K. J. J. Mayrhofer, P. N. Ross, 
 N. M. Markovic, J. Rossmeisl, J. Greeley and J. K. Norskov, Angew. Chem., Int. Ed. 
 , 2006, 45, 2897–2901. 
 40 H. A. Hansen, I. C. Man, F. Studt, F. Abild-Pedersen, T. Bligaard 
 and J. Rossmeisl, Phys. Chem. Chem. Phys., 2010, 12, 283–290. 
 41 H. A. Hansen, J. Rossmeisl and J. K. Norskov, Phys. Chem. Chem. 
 Phys. 
 , 2008, 10, 3722–3730. 
 42 J. Rossmeisl, G. S. Karlberg, T. Jaramillo and J. K. Norskov, 
 Faraday Discuss. 
 , 2008, 140, 337–346. 
 43 B. Hammer, L. B. Hansen and J. K. Norskov, Phys. Rev. B: 
 Condens. Matter Mater. Phys. 
 , 1999, 59, 7413–7421. 
 44 D. Vanderbilt, Phys. Rev. B: Condens. Matter Mater. Phys., 1990, 
 41, 7892–7895. 
 45 V. Bayer, C. Franchini and R. Podloucky, Phys. Rev. B: 
 Condens. Matter Mater. Phys. 
 , 2007, 75, 035404, DOI: 10.1103/ 
 PhysRevB.75.035404. 
 46 V. Bayer, R. Podloucky and C. Franchini, Phys. Rev. B: 
 Condens. Matter Mater. Phys. 
 , 2007, 76, 165428, DOI: 10.1103/ 
 PhysRevB.76.165428. 
 47 A. K. Cheetham and D. A. O. Hope, Phys. Rev. B: Condens. 
 Matter Mater. Phys. 
 , 1983, 27, 6964–6967. 
 48 C. Franchini, V. Bayer, R. Podloucky, J. Paier and G. Kresse, 
 Phys. Rev. B: Condens. Matter Mater. Phys. 
 , 2005, 72, 045132, 
 DOI: 10.1103/PhysRevB.72.045132. 
 Published on 18 September 2012. Downloaded by Swansea University on 17/10/2016 09:49:59. 
 View Article Online

pageNum = 13

14022 
 Phys. Chem. Chem. Phys., 2012, 
 14 , 14010–14022 
 This journal is c the Owner Societies 2012 
 49 C. Franchini, R. Podloucky, J. Paier, M. Marsman and G. Kresse, 
 Phys. Rev. B: Condens. Matter Mater. Phys. 
 , 2007, 75, 224120, 
 DOI: 10.1103/PhysRevB.75.224120. 
 50 B. E. F. Fender, A. J. Jacobson and F. A. Wedgwood, J. Chem. 
 Phys. 
 , 1968, 48, 990–995. 
 51 S. Mukherjee, A. K. Pal, S. Bhattacharya and J. Raittila, Phys. 
 Rev. B: Condens. Matter Mater. Phys. 
 , 2006, 74, 104413, DOI: 
 10.1103/PhysRevB.74.104413. 
 52 D. B. Rogers, R. D. Shannon, A. W. Sleight and J. L. Gillson, 
 Inorg. Chem. 
 , 1969, 8, 841–849. 
 53 W. G. Wykoﬀ, Crystal Structures, Wiley, New York, 1963. 54 S. A. Chambers and Y. Liang, Surf. Sci., 1999, 420, 123–133. 55 P. Jones and J. A. Hockey, Trans. Faraday Soc., 1971, 67, 2679–2685. 56 J. Rossmeisl, J. K. Norskov, C. D. Taylor, M. J. Janik and 
 M. Neurock, J. Phys. Chem. B, 2006, 110, 21833–21839. 
 57 P. W. Atkins, Phys. Chem., Oxford University Press, Oxford, 1998. 58 N. Ramaswamy and S. Mukerjee, J. Phys. Chem. C, 2011, 115, 
 18015–18026. 
 59 A. Kozawa and R. A. Powers, J. Electrochem. Soc., 1966, 113, 
 870–&. 
 60 A. Navrotsky, C. C. Ma, K. Lilova and N. Birkner, Science, 2010, 
 330, 199–201. 
 61 J. Rossmeisl, A. Logadottir and J. K. Norskov, Chem. Phys., 2005, 
 319, 178–184. 
 62 I. C. S. Man, H. Y. Calle-Vallejo, F. Hansen, H. A. Martinez, 
 J. I. Inoglu, N. G. Kitchin, J. Jaramillo, T. F. Norskov and J. K. Rossmeisl, ChemCatChem, 2011, 3, 1159–1165. 
 63 M. T. M. Koper, J. Electroanal. Chem., 2010, 660, 254–260. 64 A. J. L. R. F. Bard, Electrochemical Methods: Fundamentals and 
 Applications 
 , Wiley, 2000. 
 Published on 18 September 2012. Downloaded by Swansea University on 17/10/2016 09:49:59. 
 View Article Online

pageNum = 14

pageNum = 15

pageNum = 16

14012 
 Phys. Chem. Chem. Phys., 2012, 
 14 , 14010–14022 
 This journal is c the Owner Societies 2012 
 When the computationally derived LSV is compared to 
 the experimental LSV measured on a recently developed nano-structured manganese oxide catalyst, the theoretical predictions closely match experimental onsets for ORR and OER catalytic activities. The close match between theory and experiment validates the application of a ﬁrst-principles theoretical analysis to the electrochemical oxygen reduction and oxygen evolution reactions on surfaces at the atomic level. By focusing our analysis on reaction energetics, namely the binding energies of reactive intermediates, we expect our approach to be robust and not very dependent on the computa- tional setup and the exchange and correlation functional applied in the DFT simulations. 
 2. 
 Methods 
 2.1 
 Computational methods 
 The spin-polarized DFT calculations are performed at the generalized gradient approximation (GGA) RPBE level 
 43 
 using the plane wave implementation in Dacapo and the Atomic Simulation Environment Ultra-soft pseudo-potentials are used to deal with the ion cores. 
 44 Therefore the electronic 
 wave-functions can be represented well by a plane wave basis set with a cutoﬀ energy of 350 eV. The electron density is treated on a grid corresponding to a plane wave cutoﬀ at 500 eV. A Fermi smearing of 0.1 eV and Pulay mixing are used to ensure the fast convergence of the self-consistent electron density. Atomic positions are relaxed until the sum of the absolute forces is less than 0.05 eV A˚ 
 1. For reference, the 
 calculated equilibrium lattice constants of MnOx are 4.5 A˚/ MnO, 5.78 A˚ (a), 9.59 A˚ (c)/Mn3O4, 9.51 A˚/a-Mn2O3 and 4.43 A˚ (a), 2.86 A˚ (c)/b-MnO2, in good agreement with the experimental measurements and previous DFT studies. 
 45–53 
 The starting point for this analysis is calculations on four 
 well-deﬁned manganese oxide surfaces (Fig. 1). For the OER and the ORR it is likely that the facets control surface activity rather than surface defects since defects are expected to be covered by oxygen at the very oxidizing conditions relevant for OER and ORR. In this work, we speciﬁcally consider four close packed MnOx surfaces 
 46 
 and examine their trends 
 in behavior: MnO(001), b-MnO2(110), Mn3O4(100) and a-Mn2O3(110). The surface structures with the most stable terminations are shown in Fig. 1. For Mn3O4 (in Fig. 1a) all the surface Mn atoms are equivalent and each Mn atom coordinates with four oxygen atoms in the same plane and one oxygen in the second layer (see Fig. 1a). The a-Mn2O3(110) surface has four diﬀerent types of Mn atoms (Fig. 1b): two Mn atoms coordinate with ﬁve oxygen atoms: four oxygen atoms in the same plane and one in the second layer (site 1), and three oxygen atoms in the same plane and two in the second layer (site 4). The other two atoms coordinate with four oxygen atoms: three oxygen atoms in the same plane and one oxygen in the second layer (site 2), and two oxygen atoms in the same plane and two in the second layer (site 3). b-MnO2 has a rutile phase 
 54,55 and two types of Mn atoms on the surface: ﬁve- 
 coordinated Mn (coordinated unsaturated site, site 1 in Fig. 1c), with four oxygens in the same plane and one in the second layer, and six-coordinated Mn (bridge site, site 3 in Fig. 1c) that is 
 considered to be the inactive sites. Our calculations show that the MnO(001) surface (Fig. 1d) reconstructs immediately in the presence of oxygen, and thus this oxide phase is not considered any further. 
 A periodically repeating 4–8 layer slab is employed in the 
 model to determine the most stable MnOx surfaces in our calculations (see Fig. 1). A vacuum of at least 20 A˚ is used to separate the slab from its periodic images. Supercells with periodicity (2 
 1) have been employed to simulate adsorption 
 and electrochemical reaction, with Monkhorst–Pack type of k 
 -point sampling of 4 
 4 1 for MnO(100) and b-MnO2(110), 
 and 2 
 4 1 for Mn3O4(001). For the complex crystal structure 
 of a-Mn2O3(110), only (1 1) unit cell and 2 3 1 Monkhorst–Pack type of k-point sampling are used. The 2–4 top layers as well as possible adsorbates are fully relaxed. 
 We apply a previously developed method, the computa- 
 tional standard hydrogen electrode (CSHE) for modeling the thermochemistry of electrochemical reactions. 
 37,38 
 In 
 this method the only way the potential aﬀects the relative free energy is through the chemical potential of the electrons in the electrode. This ‘‘ﬁrst order’’ inclusion of the potential has been used to predict the activity trends for the ORR on metal and metal alloys and in the design of electrocatalysts. 
 36,37 
 Furthermore, we have shown that thermochemical features such as phase diagrams in water are also well described by 
 Fig. 1 
 The schematic structures (top view) of diﬀerent manganese oxide 
 phases, Mn atoms in blue, O atoms in red. (a) Mn3O4(001) – white rectangle indicates the (2 
 1) unit cell with the equivalent ﬁve-fold coordinated active 
 sites 1,2,3,4; (b) Mn2O3(110) white rectangle indicates the (1 1) unit cell with four types of sites: 1 – ﬁve-fold coordinated (with four oxygen atoms in the same plane), 4 – ﬁve-fold coordinated (three oxygen atoms in the same plane and two in the second layer), 2 – four-fold coordinated (three oxygen atoms in the same plane and one in the second layer) and 3 – four-fold coordinated (two oxygen atoms in the same plane and two in the second layer), and (c) MnO2(110) surfaces – a rutile type stoichiometric surface. The dashed line indicates a (1 
 2) unit cell. Positions 1 and 2 are equivalent and 
 represent the active sites (cus). Sites 3 and 4 are equivalent six-fold coordinated and are called the bridge sites; (d) MnO(100) with (1 
 1) unit 
 cell. 1 and 2 are equivalent ﬁve-fold coordinated active sites. 
 Published on 18 September 2012. Downloaded by Swansea University on 17/10/2016 09:49:59. 
 View Article Online

pageNum = 17

pageNum = 18

2 4OH
(3.6)
The analogous associative mechanism in base is as follows:
O2 + H2O + e

2 HOO* + OH
(3.7)
HOO* + e

2 O* + OH
(3.8)
O* + H2O + e

2 HO* + OH
(3.9)
HO* + e

pageNum = 19

pageNum = 20

pageNum = 21

This journal is c the Owner Societies 2012 
 Phys. Chem. Chem. Phys., 2012, 
 14 , 14010–14022 14017 
 the four charge transfer steps have identical reaction free energies of 1.23 eV = 4.92 eV/4 for an electrode held at USHE = 0. 
 If one is able to tune the binding energy of each intermediate on a surface to achieve this optimal situation, that catalyst surface would approach the activity of an ideal oxygen electrode. How- ever, as illustrated in recent work, 
 62,63 there is a universal scaling 
 relationship on a wide range of metals and oxides that governs the binding energy of the HOO* intermediate with respect to HO*, resulting in an approximately constant diﬀerence between the two energy levels (DGHOO DGHO 3:2 eV). This is far oﬀ of an optimal catalyst which would exhibit an energy diﬀerence of 2.46 eV (2e 
 1.23V) between those two particular 
 energy levels. Thus, the ‘universal’ 3.2 eV energy diﬀerence between HOO* and HO* levels can be used to deﬁne the lowest possible ‘‘theoretical overpotential’’ for the OER and the ORR [(3.2 eV 
 2.46 eV)/2e 
 E 0.37 V] on a wide variety of 
 materials. The scaling relationship between HOO* and HO* holds for MnOx just as well, as shown in Fig. 4(b), (c) and (d), with values of 3.18 eV, 3.1 eV and 3.12 eV. The slight deviation of DGHOO DGHO from 3.2 eV can be attributed to adsor- bate coverage eﬀects. 
 Indeed, the scaling relationship between the HOO* and 
 HO* binding energies explains one major source of reaction overpotential, however additional sources of overpotential can also arise from sub-optimal O* binding. It has been previously shown that the potential-determining step for the OER is either the second water dissociation step eqn (3.3) or the HO* oxidation step eqn (3.4). 
 38 Both steps involve O* and 
 either HOO* or HO*; as the latter two species scale linearly with one another, the expression (DGO DGHO ) contains information regarding the binding energies for all three species and is introduced as the universal descriptor of oxygen evolu- tion activities. 
 We can see from Fig. 4(b) and (c) that for the OER, the O* 
 covered 
 Mn3O4(001) and Mn2O3(110) have the same 
 potential-determining step, the second water dissociation step eqn (3.3) in which the third (of four) H 
 + and e are trans- 
 ferred. The O* covered Mn3O4(001) surface exhibits a lower ‘‘theoretical overpotential’’ than the O* covered Mn2O3(110) surface, 0.6 V vs. 0.79 V. This originates from the placement of the O* energy level with respect to the energy levels of the intermediates below (HO*) and above (HOO*) it; closer to half-way is better in order to prevent large uphill reaction steps. For the O* covered MnO2 surface, the second water dissociation eqn (3.3) is also the potential-determining step, according to the associative mechanism (see Fig. 4(d)), leading to a ‘‘theoretical overpotential’’ of 0.6 V. 
 For the ORR, the same scaling relationship holds between 
 HOO* and HO*. Thus, much like the OER, one part of the ORR overpotential originates from this correlation while the other part arises from sub-optimal O* binding. The free energy diagrams of the intermediates for the ORR on the self- consistent MnOx surfaces are shown in Fig. 5. Our previous studies have shown that the potential-determining ORR step is either the formation of HOO* eqn (3.2) or the reduction of HO* eqn (3.5). 
 38 As HO* and HOO* scale linearly with one 
 another, DGHO can be introduced as a universal descriptor of oxygen reduction activities. We can see that all three self- consistent MnOx surfaces – clean Mn3O4(001), 1/2 ML HO* covered Mn2O3(110) and 1/2 ML HO* (bridge) MnO2(110) – are active for the ORR. The potential-determining step is HO* 
 Fig. 4 
 Free-energy diagram for the oxygen evolution reaction on (a) 
 the perfect catalyst, and O covered (b) Mn3O4(001), (c) Mn2O3(110) and (d) MnO2(110) at U = 0, pH = 0 and T = 298 K. DGHOO DGHO (vertical solid lines) values of the three manganese oxides in (b), (c), and (d), are close to 3.2 eV, the average value found on a wide range of metals and oxides as shown in our recent paper in ref 56. The optimum value on the perfect catalyst is 2.46 eV. 
 Published on 18 September 2012. Downloaded by Swansea University on 17/10/2016 09:49:59. 
 View Article Online

pageNum = 22

pageNum = 23

pageNum = 24

pageNum = 25

This journal is c the Owner Societies 2012 
 Phys. Chem. Chem. Phys., 2012, 
 14 , 14010–14022 14021 
 make predictions on catalytic activity for the ORR and the OER. The theoretical LSVs agree well with the experimental LSVs measured on the active nanostructured MnOx bi-functional catalyst; the close match between theory and experiment suggests that the theoretical model is accurate and robust. 
 By combining ﬁrst-principles theoretical analysis and 
 experimental methods, atomic-level insight into the catalyst chemistry can be achieved. This allows one to determine principles for improving catalyst design. For the ORR, our DFT model predicts that decreasing the surface’s aﬃnity for water adsorption should signiﬁcantly increase catalytic activity, as it is desirable to destabilize the reaction inter- mediates HOO* and HO*. For the OER, our calculations show that to improve the activity of MnOx, it is necessary to design a surface structure that can break the scaling relation- ship between the energies of HOO* and HO* intermediates. If future ORR and OER catalysts are developed with these design principles in mind, superior activity for both reactions can be achieved. 
 Acknowledgements 
 We gratefully acknowledge funding from the Danish Strategic Research Council’s HyCycle program, the Danish Council for Technology and Innovation’s FTP program. This research was supported in part by the European Commission (Marie Curie Research Training Network MRTN-CT-2006-032474), by the Danish Council for Strategic Research via SERC project through grant no. 2104-06-011 and by the Catalysis for Sustainable Energy (CASE) initiative. This work was partially supported by the IMI Program of the National Science Foundation under Award No. DMR 0843934. YG, TFJ, and JKN were supported by the Center on Nanostruc- turing for Eﬃcient Energy Conversion (CNEEC) at Stanford University, an Energy Frontier Research Center funded by the U.S. Department of Energy, Oﬃce of Science, Oﬃce of Basic Energy Sciences under Award Number DE-SC0001060. 
 Notes and References 
 1 R. Forgie, G. Bugosh, K. C. Neyerlin, Z. C. Liu and P. Strasser, 
 Electrochem. Solid-State Lett. 
 , 2010, 13, D36–D39. 
 2 H. Dau, C. Limberg, T. Reier, M. Risch, S. Roggan and 
 P. Strasser, ChemCatChem, 2010, 2, 724–761. 
 3 M. W. Kanan and D. G. Nocera, Science, 2008, 321, 1072–1075. 4 K. Macounova, M. Makarova and P. Krtil, Electrochem. Commun., 
 2009, 11, 1865–1868. 
 5 M. V. Makarova, J. Jirkovsky, M. Klementova, I. Jirka, K. Macounova 
 and P. Krtil, Electrochim. Acta, 2008, 53, 2656–2664. 
 6 G. Y. Chen, S. R. Bare and T. E. Mallouk, J. Electrochem. Soc., 
 2002, 149, A1092–A1099. 
 7 N. S. Lewis and D. G. Nocera, Proc. Natl. Acad. Sci. U. S. A., 
 2006, 103, 15729–15735. 
 8 S. Chretien and H. Metiu, J. Chem. Phys., 2008, 129, 074705, DOI: 
 10.1063/1.2956506. 
 9 P. Mani, R. Srivastava and P. Strasser, J. Phys. Chem. C, 2008, 
 112, 2770–2778. 
 10 T. Ioroi, N. Kitazawa, K. Yasuda, Y. Yamamoto and H. Takenaka, 
 J. Appl. Electrochem. 
 , 2001, 31, 1179–1183. 
 11 H. Liu, B. L. Yi, M. Hou, J. F. Wu, Z. J. Hou and H. M. Zhang, 
 Electrochem. Solid-State Lett. 
 , 2004, 7, A56–A59. 
 12 L. L. Swette, A. B. Laconti and S. A. McCatty, J. Power Sources, 
 1994, 47, 343–351. 
 13 M. Pourbaix, Atlas of Electrochemical Equilibria in Aqueous Solutions, 
 Pergamon Press, New York, 1966. 
 14 K. N. Ferreira, T. M. Iverson, K. Maghlaoui, J. Barber and 
 S. Iwata, Science, 2004, 303, 1831–1838. 
 15 B. Loll, J. Kern, W. Saenger, A. Zouni and J. Biesiadka, Nature, 
 2005, 438, 1040–1044. 
 16 J. Rossmeisl, K. Dimitrievski, P. Siegbahn and J. K. Norskov, 
 J. Phys. Chem. C 
 , 2007, 111, 18821–18823. 
 17 P. E. M. Siegbahn and M. Lundberg, J. Inorg. Biochem., 2006, 100, 
 1035–1040. 
 18 M. L. Calegaro, F. H. B. Lima and E. A. Ticianelli, J. Power 
 Sources 
 , 2006, 158, 735–739. 
 19 Y. L. Cao, H. X. Yang, X. P. Ai and L. F. Xiao, J. Electroanal. 
 Chem. 
 , 2003, 557, 127–134. 
 20 F. Y. Cheng, Y. Su, J. Liang, Z. L. Tao and J. Chen, Chem. Mater., 
 2010, 22, 898–905. 
 21 M. I. A. M. S. El-Deab, A. M. Mohammad and T. Ohsaka, 
 Electrochem. Commun. 
 , 2007, 9, 2082–2087. 
 22 M. S. El-Deab and T. Ohsaka, J. Electrochem. Soc., 2008, 155, 
 D14–D21. 
 23 F. Jiao and H. Frei, Chem. Commun., 2010, 46, 2920–2922. 24 F. H. B. Lima, M. L. Calegaro and E. A. Ticianelli, J. Electroanal. 
 Chem. 
 , 2006, 590, 152–160. 
 25 F. H. B. Lima, M. L. Calegaro and E. A. Ticianelli, Electrochim. 
 Acta 
 , 2007, 52, 3732–3738. 
 26 A. M. Mohammad, M. I. Awad, M. S. El-Deab, T. Okajima and 
 T. Ohsaka, Electrochim. Acta, 2008, 53, 4351–4358. 
 27 I. Roche, E. Chainet, M. Chatenet and J. Vondrak, J. Phys. Chem. 
 C 
 , 2007, 111, 1434–1443. 
 28 L. Q. Mao, T. Sotomura, K. Nakatsu, N. Koshiba, D. Zhang and 
 T. Ohsaka, J. Electrochem. Soc., 2002, 149, A504–A507. 
 29 L. Q. Mao, D. Zhang, T. Sotomura, K. Nakatsu, N. Koshiba and 
 T. Ohsaka, Electrochim. Acta, 2003, 48, 1015–1021. 
 30 J. S. Yang and J. J. Xu, Electrochem. Commun., 2003, 5, 306–311. 31 Y. Gorlin and T. F. Jaramillo, J. Am. Chem. Soc., 2010, 132, 
 13612–13614. 
 32 N. Ramaswamy, R. J. Allen and S. Mukerjee, J. Phys. Chem. C, 
 2011, 115, 12650–12664. 
 33 N. Sivasankar, W. W. Weare and H. Frei, J. Am. Chem. Soc., 2011, 
 133, 12976–12979. 
 34 B. S. Yeo and A. T. Bell, J. Am. Chem. Soc., 2011, 133, 5587–5593. 35 Y. Gorlin and T. F. Jaramillo, ECS Trans., 2011, 41, 1701–1707. 36 J. Greeley, I. E. L. Stephens, A. S. Bondarenko, T. P. Johansson, 
 H. A. Hansen, T. F. Jaramillo, J. Rossmeisl, I. Chorkendorﬀ and J. K. Norskov, Nat. Chem., 2009, 1, 552–556. 
 37 J. K. Norskov, J. Rossmeisl, A. Logadottir, L. Lindqvist, 
 J. R. Kitchin, T. Bligaard and H. Jonsson, J. Phys. Chem. B, 2004, 108, 17886–17892. 
 38 J. Rossmeisl, Z. W. Qu, H. Zhu, G. J. Kroes and J. K. Norskov, 
 J. Electroanal. Chem. 
 , 2007, 607, 83–89. 
 39 V. Stamenkovic, B. S. Mun, K. J. J. Mayrhofer, P. N. Ross, 
 N. M. Markovic, J. Rossmeisl, J. Greeley and J. K. Norskov, Angew. Chem., Int. Ed. 
 , 2006, 45, 2897–2901. 
 40 H. A. Hansen, I. C. Man, F. Studt, F. Abild-Pedersen, T. Bligaard 
 and J. Rossmeisl, Phys. Chem. Chem. Phys., 2010, 12, 283–290. 
 41 H. A. Hansen, J. Rossmeisl and J. K. Norskov, Phys. Chem. Chem. 
 Phys. 
 , 2008, 10, 3722–3730. 
 42 J. Rossmeisl, G. S. Karlberg, T. Jaramillo and J. K. Norskov, 
 Faraday Discuss. 
 , 2008, 140, 337–346. 
 43 B. Hammer, L. B. Hansen and J. K. Norskov, Phys. Rev. B: 
 Condens. Matter Mater. Phys. 
 , 1999, 59, 7413–7421. 
 44 D. Vanderbilt, Phys. Rev. B: Condens. Matter Mater. Phys., 1990, 
 41, 7892–7895. 
 45 V. Bayer, C. Franchini and R. Podloucky, Phys. Rev. B: 
 Condens. Matter Mater. Phys. 
 , 2007, 75, 035404, DOI: 10.1103/ 
 PhysRevB.75.035404. 
 46 V. Bayer, R. Podloucky and C. Franchini, Phys. Rev. B: 
 Condens. Matter Mater. Phys. 
 , 2007, 76, 165428, DOI: 10.1103/ 
 PhysRevB.76.165428. 
 47 A. K. Cheetham and D. A. O. Hope, Phys. Rev. B: Condens. 
 Matter Mater. Phys. 
 , 1983, 27, 6964–6967. 
 48 C. Franchini, V. Bayer, R. Podloucky, J. Paier and G. Kresse, 
 Phys. Rev. B: Condens. Matter Mater. Phys. 
 , 2005, 72, 045132, 
 DOI: 10.1103/PhysRevB.72.045132. 
 Published on 18 September 2012. Downloaded by Swansea University on 17/10/2016 09:49:59. 
 View Article Online

pageNum = 26

14022 
 Phys. Chem. Chem. Phys., 2012, 
 14 , 14010–14022 
 This journal is c the Owner Societies 2012 
 49 C. Franchini, R. Podloucky, J. Paier, M. Marsman and G. Kresse, 
 Phys. Rev. B: Condens. Matter Mater. Phys. 
 , 2007, 75, 224120, 
 DOI: 10.1103/PhysRevB.75.224120. 
 50 B. E. F. Fender, A. J. Jacobson and F. A. Wedgwood, J. Chem. 
 Phys. 
 , 1968, 48, 990–995. 
 51 S. Mukherjee, A. K. Pal, S. Bhattacharya and J. Raittila, Phys. 
 Rev. B: Condens. Matter Mater. Phys. 
 , 2006, 74, 104413, DOI: 
 10.1103/PhysRevB.74.104413. 
 52 D. B. Rogers, R. D. Shannon, A. W. Sleight and J. L. Gillson, 
 Inorg. Chem. 
 , 1969, 8, 841–849. 
 53 W. G. Wykoﬀ, Crystal Structures, Wiley, New York, 1963. 54 S. A. Chambers and Y. Liang, Surf. Sci., 1999, 420, 123–133. 55 P. Jones and J. A. Hockey, Trans. Faraday Soc., 1971, 67, 2679–2685. 56 J. Rossmeisl, J. K. Norskov, C. D. Taylor, M. J. Janik and 
 M. Neurock, J. Phys. Chem. B, 2006, 110, 21833–21839. 
 57 P. W. Atkins, Phys. Chem., Oxford University Press, Oxford, 1998. 58 N. Ramaswamy and S. Mukerjee, J. Phys. Chem. C, 2011, 115, 
 18015–18026. 
 59 A. Kozawa and R. A. Powers, J. Electrochem. Soc., 1966, 113, 
 870–&. 
 60 A. Navrotsky, C. C. Ma, K. Lilova and N. Birkner, Science, 2010, 
 330, 199–201. 
 61 J. Rossmeisl, A. Logadottir and J. K. Norskov, Chem. Phys., 2005, 
 319, 178–184. 
 62 I. C. S. Man, H. Y. Calle-Vallejo, F. Hansen, H. A. Martinez, 
 J. I. Inoglu, N. G. Kitchin, J. Jaramillo, T. F. Norskov and J. K. Rossmeisl, ChemCatChem, 2011, 3, 1159–1165. 
 63 M. T. M. Koper, J. Electroanal. Chem., 2010, 660, 254–260. 64 A. J. L. R. F. Bard, Electrochemical Methods: Fundamentals and 
 Applications 
 , Wiley, 2000. 
 Published on 18 September 2012. Downloaded by Swansea University on 17/10/2016 09:49:59. 
 View Article Online

Home