ORIGINAL PAPER

Instability of Supported Platinum Nanoparticlesin Low-Temperature Fuel Cells

Y. Shao-Horn Æ W. C. Sheng Æ S. Chen ÆP. J. Ferreira Æ E. F. Holby Æ D. Morgan

� Springer Science+Business Media, LLC 2007

Abstract This paper discusses the mechanisms of surface

area loss of supported platinum (Pt) electrocatalysts in low-

temperature fuel cells. It is argued that submonolayer dis-

solution of Pt nanoparticles governs the surface area loss at

high voltages by increasing the loss of Pt from carbon and

coarsening of Pt nanoparticles on carbon.

Keywords Fuel cell � Catalyst � Platinum �Nanoparticles � Durability � Dissolution � Coarsening �Crystal migration � Transmission electron microscopy �TEM � Ostwald ripening � Solubility � Coalescence

1 Introduction

A fuel cell directly converts chemical energy to electricity

via electrochemical reactions. Although conversion of

hydrocarbon fuels is highly desirable in low-temperature

fuel cell technologies in terms of system energy density,

known electrocatalysts cannot provide conversion effi-

ciencies and rates acceptable for practical applications.

Therefore, hydrogen is typically used as the fuel and

oxygen in air is employed as the oxidant. We here use

proton exchange membrane (PEM) fuel cells to illustrate

the components and the operation of low-temperature fuel

cells. The schematic of a PEM fuel cell is shown in Fig. 1.

Protons produced from oxidation of hydrogen in the anode

flow to the cathode through the membrane, and combine

with oxygen to form water in the cathode. Compensating

charge in the form of electrons flows through the external

circuit and provides the useful work done by the fuel cell.

The three-layer structure of anode, cathode and proton-

conducting medium is referred commonly to as membrane

electrode assembly (MEA). PEM fuel cell electrodes typ-

ically have a thickness of *10 lm, and consist of

electrically connected catalysts in three dimensions (3D)

and an impregnated proton-conducting phase. This elec-

trode structure provides active sites for electrochemical

reactions as well as electron and ion conduction pathways

in 3D. Electrochemical oxidation of hydrogen and reduc-

tion of oxygen can only occur at interfacial boundary

regions, where gas molecules reach catalyst surfaces that

are connected electrically to the external circuit through a

3D conducting network (usually carbon) and are accessible

to the 3D proton-conducting medium. Conventional PEM

fuel cell catalysts typically consist of Pt nanoparticles on

the order of 2–3 nm in diameter, which are dispersed on

the surface of primary carbon particles of 20–50 nm to

maximize the surface area of Pt per unit Pt mass, as shown

in the schematic in Fig. 2a. These primary carbon particles

can form carbon particle aggregates of 100–300 nm in

diameter, which are connected electrically in 3D. Pt

Electronic supplementary material The online version of thisarticle (doi:10.1007/s11244-007-9000-0) contains supplementarymaterial, which is available to authorized users.

Y. Shao-Horn (&) � S. Chen

Electrochemical Energy Laboratory, Department of Mechanical

Engineering, Massachusetts Institute of Technology, Cambridge,

MA 02139, USA

e-mail: shaohorn@mit.edu

W. C. Sheng

Electrochemical Energy Laboratory, Department of Chemistry,

Massachusetts Institute of Technology, Cambridge, MA 02139,

USA

P. J. Ferreira

Materials Science and Engineering Program, University of Texas

at Austin, Austin, TX 78712, USA

E. F. Holby � D. Morgan

Department of Materials Science and Engineering, University of

Wisconsin-Madison, Madison, WI 53706, USA

123

Top Catal

DOI 10.1007/s11244-007-9000-0

loading (measured in mgPt/cm2electrode) and Pt dispersion in

the cathode significantly influence the activation overpo-

tential for oxygen reduction, the efficiency, and the cost of

PEM fuel cells [1, 2]. The state-of-the-art Pt loading is

0:4 mgPt=cm2electrode in the cathode [2] while that required

in the anode can be much lower due to fast electrochemical

oxidation of hydrogen on Pt. Not only is it important to

lower Pt loading in order to reduce fuel cell cost but also it

is essential to maintain the activity of Pt catalysts during

the lifetime of fuel cells.

Coarsening of Pt nanoparticles, which decreases the

ratio of Pt surface atoms to bulk atoms and the availability

of triple phase boundary (TPB) regions in electrodes, and

loss of Pt from the 3D carbon network, increases the

activation overpotential and results in decreased cell volt-

age, efficiency, and lifetime of fuel cells. A dispersion of Pt

nanocrystals is thermodynamically unstable to the growth

and formation of larger crystals or particles driven by

reduction of surface energies. As approximately 5,500 h of

fuel cell operation are required in a vehicle [3], long-term

stability of Pt nanoparticles on carbon is critical to meet the

required lifetime and efficiency of PEM fuel cells. Multiple

processes can degrade the electrochemically active surface

area, particularly at the cathode, and long-term stability of

cathode activity has now emerged as one of the major

challenges to developing commercially viable PEM fuel

cells for automotive applications. Essential to developing

better catalyst stability is an understanding of the funda-

mental mechanisms leading to catalyst activity loss or

surface area loss. Unfortunately, the mechanism of surface

area loss of supported Pt nanoparticles on carbon in contact

with an ion-conducting phase at temperatures lower than

200 �C are not well understood. The mechanism of surface

area loss in the cathode in low-temperature fuel cells has

been a subject of debate in the last three decades with a

resurgence of work in PEM fuel cells in the last 5 years.

In this paper, the microstructure of conventional carbon

supported Pt nanoparticle electrocatalysts is first discussed,

which includes the size, morphology and distribution of Pt

nanoparticles. We then analyze reported evidence for the

two main mechanisms of Pt surface area loss in low-tem-

perature fuel cells, which are Pt migration and coalescence,

and Pt dissolution and redeposition, and summarize dif-

ferent processes that can contribute to Pt surface area loss

in fuel cell electrodes. It is proposed from these analyses

that Pt dissolution governs the surface area loss of fuel cell

electrodes that are exposed to voltages greater than 0.8 V

vs. RHE. The surface area loss results from Pt coarsening

on carbon, and Pt loss from fuel cell electrodes and pre-

cipitation in the ion-conducting phase. We further examine

assumptions used in the debate over the coarsening

mechanism of Pt nanoparticles on carbon, via particle

migration and coalescence and Ostwald ripening. Although

there is no unique evidence to support crystal migration of

Pt nanoparticles on carbon at low temperatures, we argue

that it cannot be ruled out on theoretical grounds.

As controlling Pt dissolution is essential to maintaining

activity of fuel cell electrodes, we focus the latter part of

the paper on the dissolution thermodynamics and kinetics

of Pt nanoparticles. Reported Pt solubility data in acid

solutions and fuel cells are compiled, which show that Pt

solubility is potential and temperature dependent as pre-

dicted by Pourbaix. However, Pt dissolution seen in many

experiments exhibits much weaker potential dependence

than that expected from the Nernstian principle. We pro-

pose that this phenomenon can be attributed to dissolution

Cathode

H2 O2/Air

H2O

Electrical Load

e- e-

H+ OH+

H2O H2O

Proton ExchangeMembrane

Membrane Electrode Assembly

Carbon SupportedPt Particles

Anode

H2 => 2H+ + 2e- 2H+ + 2e- + 1/2O2 => H2O

Fig. 1 Schematic of a typical PEM fuel cell. Hydrogen oxidation and

oxygen/Air reduction take place at the anode and the cathode,

respectively. The electrons generated upon hydrogen oxidation in the

anode pass through the external circuit, combine with oxygen and

protons to form water in the cathode and provide work

(a)

Three-Dimensional Carbon Network

(b)

(111)

(010)

(111) (111)

(001)

(100)

-

-

Equilibrium shape of Pt crystals

Fig. 2 (a) Schematic of Pt

nanoparticles dispersed on

primary carbon particles in

supported Pt electrocatalyst

samples. Pt nanoparticles are

about 2–3 nm in diameter and

the size of the primary carbon

particles is 20–50 nm; (b):

truncated octahedron consisting

of {111}Pt and {100}Pt facets,

which is the equilibrium shape

of Pt crystals at 0 K

Top Catal

123

within the top monolayer, which may be governed largely

by surface defects and defect energetics. Moreover, it is

postulated that the Gibbs–Thomson effect on Pt nanopar-

ticles can lead to dramatically increased driving force for

dissolution and faster dissolution kinetics at room tem-

perature from a theoretical standpoint. This hypothesis is

supported by recent experiment results, which have shown

that increasing average Pt nanoparticle sizes from *2 to

*4 nm, and reducing the number of Pt nanoparticles

smaller than 2 nm, can greatly improve the stability and

reduce Pt surface loss during voltage cycling. Finally, we

conclude with some general remarks about the status of our

understanding in crystal migration, Pt dissolution and

redeposition of Pt nanoparticles at the atomic scale, and

potential opportunities to enhance the stability of supported

Pt electrocatalysts in low-temperature fuel cells.

2 Microstructure of Pristine Supported Pt

Nanoparticles

Supported Pt electrocatalysts can be prepared by either (1)

impregnation of carbon support with Pt precursor solution

and precipitation of Pt nanoparticles by reduction [4]; (2)

adsorption of colloidal Pt particles produced from reactions

of H2PtCl6 and sodium salts on carbon support [5].

Detailed synthetic processes and the effects of carbon

supports on the size and distribution of Pt nanoparticles can

be found in review articles of Auer et al. [6] and Antolini

[4]. We here discuss the size, morphology and distribution

of Pt nanoparticles supported on carbon in conventional

electrocatalysts for PEM fuel cells.

We here show the microstructure of two commercial Pt

catalyst powder samples [46 wt% Pt/C, Tanaka Kikinzoku

Kogyo (TKK)] supported on Vulcan, and a proprietary

high surface area carbon, which are commonly used in

PEM fuel cells. Transmission electron microscopy (TEM)

images of these two samples at low and high magnifica-

tions are shown in Figs. 3a–d and 4a–d, respectively. The

high-resolution TEM images in Figs. 3d and 4d show that

individual Pt nanoparticles are single crystals. It is inter-

esting to note that Pt nanoparticles on carbon typically

appear quasi-circular, on the order of 2–3 nm, and exhibit a

lack of facets. Although it is difficult to tell the 3D shapes

of Pt nanoparticles in the TEM images that project 3D

objects into two-dimensional (2D) images, a few Pt nano-

particles present on the edge of primary carbon particles in

Figs. 3b–d and 4b–c reveal that they are disk-like. From a

thermodynamic point of view, the equilibrium shape of Pt

nanoparticles will be determined by minimization of sur-

face energy per unit area of exposed Pt surfaces provided

that edge and curvature effects are negligible. For ideal

face-centered-cubic (fcc) metals such as Pt, the surface

energy (c) of atomic planes with high symmetry follow the

order of c {111}Pt \ c {100}Pt \ c {110}Pt, as expected

from the surface atomic density [7, 8]. The equilibrium

shape of a Pt crystal at zero temperature has been shown to

be a truncated octahedron (cubo-octahedron) consisting of

Fig. 3 Typical TEM images of

46 wt% of Pt supported on

Vulcan (TKK) from low to high

magnifications (a, c, b to d). The

Pt nanoparticles appear to be

quasi-circular and on the order

of 2–3 nm. High-resolution

TEM images show that

individual Pt nanoparticles are

single crystals (d). The Pt

dispersion is poor on Vulcan.

Some Pt particles agglomerate

near the junctions among

primary carbon particles and

some areas on the surface of

Vulcan carbon have no Pt

nanoparticle (b). The Pt

nanoparticles aspect ratio is of

1.5–4.5

Top Catal

123

{111}Pt and {100}Pt facets [9], as shown in Fig. 2b. Pt

nanoparticles supported on carbon in these TKK fuel cell

electrocatalysts have flattened shapes that deviate from the

equilibrium cubo-octahedron geometry. Some Pt nanopar-

ticles supported on high surface area carbon appear to

exhibit pronounced {111}Pt surfaces as marked in Fig. 4d,

which is consistent with the observations by Ross et al.

[10] for Pt nanoparticles less than 3.5 nm. The shapes of Pt

nanoparticles are dependent on synthesis conditions, which

may vary the nucleation and growth rates of Pt clusters on

carbon surface, the energy of Pt surfaces, and the interface

between Pt and carbon support [11]. The disk-like geom-

etries can provide more contact area with carbon surface,

which may be important to anchoring of Pt nanoparticles

on carbon. In addition, they can yield Pt surface facets

other than the {111}Pt and {100}Pt, and change the size-

dependent surface distribution of Pt atoms as reported by

Kinoshita [12].

Measurements of individual Pt nanoparticles (particle

agglomerates are excluded) in the TEM images show that

Pt nanoparticles supported on Vulcan and high surface area

carbon have very similar sizes and size distributions, which

are shown in Fig. 5a–b. Both sizes and distributions of

these two 46 wt% TKK catalysts are comparable to those

of 20 wt% E-TEK catalysts reported previously [13], and

those found by Markovic et al. [14]. The histograms of Pt

nanoparticles can be used to compute the volume/area

averaged surface area, which is comparable to surface area

measured by electrochemical methods. Typically electro-

chemically active surface area is equal to *70% of the

Fig. 4 Typical TEM images of

46 wt% of Pt supported on high

surface area carbon (TKK) from

low to high magnifications (a, b,

c to d). The Pt nanoparticles

appear to be quasi-circular and

on the order of 2–3 nm. High-

resolution TEM images show

that individual Pt nanoparticles

are single crystals (d). Pt

nanoparticles are well dispersed

on high surface area carbon

support. The Pt nanoparticles

aspect ratio is of 1–3. Some Pt

nanoparticles exhibit

pronounced {111}Pt surfaces as

shown in (d)

2 4 6 8 10 12 14 16 182 4 6 8 10 12 14 16 180

10

20

30

40

50

60

70

80

(a) (b)

0

10

20

30

40

50

60

70

80

Mean=2.5SD = 0.53Min = 1.17Max = 3.8

Sample = 200 particles

Mean=2.8SD = 0.51Min = 1.3Max = 4.0

Sample = 200 particles

Data: Data2_BModel: Gauss

Chi^2/DoF = 3.4848R^2 = 0.99771

y0 -4.88163 ±4.44841xc 2.75149 ±0.01592w 1.30434 ±0.09672A 116.79278 ±14.42983

Data: Data1_BModel: Gauss

Chi^2/DoF = 74.06649R^2 = 0.95321

y0 3.79311 ±5.94694xc 2.98076 ±0.06151w 0.93339 ±0.15965A 84.23166 ±17.44387

Num

ber

of P

artic

les

Num

ber

of P

artic

les

Spherical Particle Size (nm)Spherical Particle Size (nm)

Fig. 5 Histograms of the Pt

nanoparticle size distribution of

the two catalysts in Figs. 3 and

4: (a) 46 wt% of Pt supported

on Vulcan (TKK); and (b)

46 wt% of Pt supported on high

surface area carbon (TKK).

Both of the two catalysts have

similar Pt nanoparticle sizes and

size distributions

Top Catal

123

volume/area averaged surface area from TEM measure-

ments [15, 16]. This difference might be attributed to the

contact area between the Pt nanoparticles and carbon

support, which is considered in the volume/area averaged

surface area calculation but would not be accessed in the

electrochemical measurements.

While the sizes and distributions of Pt nanoparticles are

similar, Pt nanoparticles supported on high surface area

carbon (Fig. 4) show better dispersion than those found on

Vulcan (Fig. 3) as evidenced by the higher number of Pt

nanoparticle agglomerates on Vulcan. Primary Vulcan

carbon particles have a diameter on the order of *40 nm

while that of the high surface area carbon is smaller

(*20 nm). Some Pt particles agglomerate near the junc-

tions among primary carbon particles, and some areas on

the surface of Vulcan carbon have no Pt nanoparticles, as

shown in Fig. 3b. Poor dispersion of Pt nanoparticles on

Vulcan found in the TEM images is consistent with the fact

that the specific area of Pt nanoparticles on Vulcan deter-

mined from measuring the charge of hydrogen adsorption

or desorption in cyclic voltammetry (210 mC=cm2Pt is

assumed) is only 63 m2/gPt [17], which is considerably

lower than Pt supported on high surface area carbon (80–

96 m2/gPt) [1, 17].

3 Instability of Supported Pt Nanoparticles in Low-

Temperature Fuel Cells

Activity loss or electrochemically active surface area loss

of electrodes that consist of Pt nanoparticles supported on

carbon or Pt/C electrocatalysts have been studied exten-

sively in acid solution, phosphoric acid (PA) fuel cells [18–

21] and in PEM fuel cells [3, 15, 22–28]. The loss of

electrochemically active surface area of Pt with time is a

major source of reduced fuel cell voltage. This effect can

be seen in Fig. 6, which shows the degradation during

constant potential holds for a PEM fuel cell short stack

(operated with H2 air (stoichiometric flows of 2/2) under

80 �C, 150 kPaabs, and 100% relative humidity (RH) [15].

The measured surface area loss in Fig. 6a can be related to

the cathode kinetics to show that surface area loss makes a

major contribution to the drop in cell potential shown in

Fig. 6b, particularly at the open circuit voltage condition

[15]. Two major mechanisms have been proposed to

describe Pt surface area loss in fuel cell electrodes: (i)

crystal migration and coalescence, and (ii) Pt dissolution

and redeposition. In the following, we analyze experi-

mental evidence in support of each mechanism, and

summarize potential mechanisms responsible for surface

area loss of supported Pt electrocatalysts in low-tempera-

ture fuel cells.

3.1 Reported Evidence in Support of Pt Coarsening via

Crystal Migration

The two primary arguments in support of Pt coarsening via

crystal migration are based on the absence of voltage

dependence and the shape and growth of the particle size

distributions. Many earlier studies [21, 29, 30] of Pt

nanoparticles supported on carbon fail to observe a sig-

nificant potential dependence (0 V to *0.8 V) of the Pt

area loss over time in phosphoric acid (H3PO4), from

which it was concluded that simple Pt dissolution and

redeposition cannot be the dominating mechanism. Gruver

et al. [30] have shown that the change in the electro-

chemical surface area of Pt nanoparticles supported on

carbon varies to a small extent with the applied potential

[up to 0.75 V vs. the reversible hydrogen electrode (RHE)],

and no change in Pt loading is noted in the cathode, where

the electrodes have been preflooded with *100% H3PO4

and purged with nitrogen at 191 �C.

The particle migration mechanism has implications for

the growth rate and asymptotic shape of the particle size

distribution. The mechanism can be described generally by

surface migration of adatoms on the Pt crystal surface, 2D

motion of Pt crystals on the carbon support, and coales-

cence when crystals meet [31, 32]. In the crystal migration

model, Granqvist and Buhrman [32] have used a simple

Fig. 6 Cell voltage (a) and Pt

surface area (b) as a function of

short-stack runtime. Dashed line

in (b) corresponds to the

calculated Pt surface area loss

[15]. Short stacks were operated

with hydrogen/air

(stoichiometric flows of 2/2)

under 80 �C, 150 kPaabs, and

100% RH

Top Catal

123

statistical method, to show that the asymptotic particle size

distribution is log-normal with respect to volume (with a

distinctive tail at large particle sizes) and Ruckenstein and

Pulvermacher [31] have shown that the rate of particle

growth is dependent on particle loading on support. These

properties have been used to identify when crystal migra-

tion is occurring. For example, Blurton et al. have shown

an asymptotic Pt particle size distribution of Pt nanoparti-

cles with a tail at large particle sizes in an electrode

exposed to air and held at *0.6 V and 163 �C for 645 h

[21]. Based on the shape of this asymptotic distribution,

these authors have related the surface area loss of Pt

nanoparticles in the PA fuel cell environment to the

mechanism of crystal migration and coalescence. The

correlation between steady-state particle size distribution

and the coarsening process has been used later by Wilson

et al. [24] and Tada et al. [25] in understanding the deg-

radation of Pt nanoparticles supported on Vulcan aged

under a constant voltage (0.5 V) or constant current density

(1.5 A/cm2) over a few thousands of hours of operation in a

hydrogen/oxygen PEM fuel cell. As asymptotic Pt particle

size distributions with a tail toward large particle sizes

were found in their cathode, a growth mechanism via

crystal migration was proposed. Although the surface area

loss of Pt nanoparticles is found to be potential dependent,

the difference noted between the cathode and anode has

been attributed to higher levels of hydration [24] or carbon

corrosion [25] at the cathode.

3.2 Reported Evidence in Support of Pt Coarsening and

Loss by Pt Dissolution and Redeposition

During steady-state or transient operation with electrodes

exposed to voltages greater than 0.8 V vs. RHE, the

following observations have been made in support of Pt

surface area loss induced by Pt dissolution and redepo-

sition: (i) the extent of Pt surface area loss is potential

dependent; (ii) loss of Pt from fuel cell electrodes and

detection of Pt nanocrystals in the ion-conducting phase

of low-temperature fuel cells; (iii) the rate of Pt surface

area loss can be accelerated by voltage cycling and

increases with the upper voltage limit; (iv) changes in

the size and distributions of Pt nanoparticles. We will

discuss these four aspects in detail in the following

section.

3.2.1 Potential Dependent Pt Surface Area Loss

Several studies have found that the degree of coarsening of

Pt nanoparticles supported on carbon clearly exhibits

strong voltage dependence at voltages higher than 0.8 V

vs. RHE. This observation supports coarsening of Pt

nanoparticles by the mechanism of Pt dissolution and

redeposition first proposed from studies of Pt black in

H3PO4 by Tseung [33]. Honji et al. [20] have found that Pt

nanoparticles coarsen to larger sizes and the loading of Pt

in the cathode decreases with increasing applied potentials

after a 100-h constant voltage hold (0–0.9 V vs. RHE) at

205 �C in an electrochemical cell setup flooded with

H3PO4 and purged with nitrogen gas. These authors have

thus proposed that the growth of Pt nanoparticles occurs by

redeposition of dissolved Pt in the acid, and reduction of Pt

loading in the cathode is caused by Pt dissolution at volt-

ages greater than 0.8 V vs. RHE. This process of Pt

coarsening is analogous to classical Ostwald ripening,

which describes the growth of large particles at the expense

of small particles through 3D transport of atomic and

molecular species [34–36]. In this context, Ostwald rip-

ening of Pt nanoparticles involves dissolution of Pt from

small nanoparticles, transport of dissolved Pt species and

redeposition (reduction) of Pt species from solution onto

large nanoparticles, as first conceptualized by Ross [37].

It should be mentioned that a different Ostwald ripening

process has been suggested by Bett et al. [29], where Pt

transport would occur via transport of molecular Pt species

on the carbon support, referred to here as ‘‘2D Ostwald

ripening’’. This hypothesis has been proposed based on

experimental findings that neither the effect of Pt loading

(5 wt% Pt/C vs. 20 wt% Pt/C) nor the potential depen-

dence (0.1 to 1 V vs. RHE) on the Pt area loss is significant

[29]. However, to the authors’ knowledge, no definitive

experimental evidence of Pt atom transport on carbon has

been reported to date.

3.2.2 Pt Deposition in the Ion-Conducting Phase

Aragane et al. [19, 38] have first reported the presence of

Pt particles in the matrix of PA fuel cells and Pt loss from

the cathode, which has provided direct evidence for Pt

dissolution from supported Pt nanoparticles in the cathode

and reduction of soluble Pt species with permeated

hydrogen gas molecules in the membrane. Similarly,

Yasuda et al. [28, 39] have detected the appearance of

numerous Pt single crystals on the order of 10–100 nm in

the membrane after potential holds at high voltages such as

0.8 V and 1.0 V in air and nitrogen in PEM fuel cells. In

addition, the amount of Pt found in the matrix and thus the

amount of Pt dissolution has been shown to increase with

increasing cathode potentials [19, 28, 38]. The presence

and diffusion of soluble Pt species in fuel cell electrodes is

further confirmed by the fact that soluble Pt species have

been found in the water collected from the reactant gases

exiting the cell [27].

Top Catal

123

3.2.3 Enhanced Pt Loss upon Voltage Cycling

Pt surface area loss is accelerated upon voltage cycling

compared to extended holds at constant potentials. Ki-

noshita et al. first studied potential cycling effects on Pt

black and Pt nanoparticles on carbon [40] in 1 M sulfuric

acid (H2SO4) purged with nitrogen at 23 �C. They have

shown that the surface area of Pt nanoparticles significantly

decreases upon voltage cycling from 0.05 to 1.25 V vs.

RHE, a result of Pt dissolution into the acid and growth of

large Pt nanoparticles at the expense of small ones. This

phenomenon has been reported for the first time in PEM

fuel cells by Patterson [41], where accelerated Pt area loss

of a hydrogen-air PEM cathode is observed when subjected

to square-wave voltage cycles between 0.87 and 1.20 V

(65 �C and 60 s per cycle). Post-mortem analysis of the

MEA after *100 h reveals the presence of Pt in the

membrane, which implies that Pt diffusion occurs on the

micrometer scale (i.e., *10 lm from the cathode into the

membrane in comparison to *10 nm from a Pt nanopar-

ticle to a neighboring nanoparticle) [41]. Darling and

Meyers [42, 43] have recently developed a numerical

model for the Pt area loss induced by voltage cycling,

which includes a potential-dependent dissolution of Pt

(Pt $ Pt2+ + 2e-, i.e., a 2.303RT/2F potential depen-

dence), chemical dissolution of Pt oxide

(PtO + 2H+ $ Pt2+ + H2O), and surface tension driven

growth of Pt nanoparticles. This analysis has been used to

reasonably model and reproduce Patterson’s voltage

cycling data [41]. Although the model considers diffusion

of Pt2+ from the cathode to the anode and electrochemical

reduction into Pt (Pt2+ + 2e- ? Pt), it does not explicitly

account for the fact that Pt particles have been detected

inside the membrane, as reported by Patterson [41].

The rate and extent of electrochemical surface area loss of

Pt becomes larger with an increase in the upper voltage limit

from 0.8 to 1.2 V (with flowing nitrogen to the cathode)

during voltage cycling, which is accompanied by an

increasing number of Pt crystals observed in the membrane

[44]. It is noted that the distribution of Pt deposition in the

membrane is unaffected by the lower voltage limit [44]. The

presence of hydrogen in the membrane has been shown to be

essential to the appearance of these Pt crystals in the mem-

brane, which supports the claim that these Pt crystals form

by chemical reduction of Pt soluble species with hydrogen

gas [44]. The mechanism of hydrogen induced reduction and

precipitation is in good agreement with the observation that

the location of Pt deposition in the membrane is dependent

on the membrane hydrogen partial pressure and moves

toward the anode with flowing oxygen relative to flowing

nitrogen in the cathode [44]. Although electrochemical

surface area loss of Pt has been qualitatively related to the

amount of Pt deposited in the membrane, the coarsening of

Pt nanoparticles on carbon in the cathode during cycling is

not considered in this study [44].

3.2.4 Changes in the Size and Morphology of Pt

Nanoparticles

Aragane et al. [19, 38] have first analyzed Pt particle size

distributions on carbon in the cathode using TEM. It has

been shown that Pt nanoparticles coarsen to the same

degree from *4 to *7 nm throughout the thickness of the

cathode, and that the loading of Pt on carbon decreases in

the cathode approaching the cathode–electrolyte (matrix)

interface after PA fuel cell operation at 0.7 V for 1,500 h at

190 �C (presumably in 99% H3PO4 and air input for the

cathode). Similar observations have been made in PEM

fuel cell MEA cathode samples (Pt nanoparticles supported

on Vulcan and supported on high surface area carbon) after

voltage cycling between 0.6 V and 1.0 V (with fully

humidified nitrogen at the cathode) at 80 �C [15, 16]. We

show and discuss these findings in some detail. Four TEM

micrographs taken from four different locations, 1, 4, 7 and

10 lm from the diffusion-medium (DM) and cathode

interface, are shown in Fig. 7, respectively. Supported Pt

nanoparticles on the order of 2–3 nm in the pristine MEA

have grown to *6 nm on carbon in the cycled MEA

cathode, which is comparable across the entire cathode.

This observation supports that the coarsening of spherical

Pt particles on carbon across the cycled MEA cathode

thickness proceeds in one single process—Ostwald ripen-

ing on the nanometer-scale. As Ostwald ripening is driven

by reduction of surface energy of Pt particles, very small Pt

particles should disappear, and the small particle size end

of the distribution should shift to larger values in the cycled

MEA cathode relative to the pristine sample. This is con-

sistent with the fact that the number of Pt particles no

larger than 5 nm is reduced significantly in the cycled

MEA cathode in Fig. 7. Moving toward the membrane, the

weight of Pt on carbon decreases while the amount of Pt

deposited in the ionomer phase increases in the cycled

MEA cathode. The increasing weight percentage of Pt in

the ionomer phase towards the cathode/membrane interface

can be explained by the fact that the flux of cross-over

hydrogen required for Pt precipitation in the ionomer phase

is the highest at the cathode/membrane interface in the

cathode, and it decreases considerably with increasing

distance from the cathode/membrane interface upon fast

electro-oxidation and depletion of hydrogen in the cathode.

Relative contributions of Pt coarsening on carbon and loss

of Pt from carbon in the cathode have been quantified from

analysis of cross-sectional TEM images, as shown by

previous studies [15, 16]. The decrease in the weight of Pt

remaining on carbon has been quantified from the

Top Catal

123

histograms of Pt nanoparticles across the cathode thickness

[15], from which the amount of Pt crystals precipitated in

the ionomer phase is obtained. Although the Pt precipita-

tion mechanism dominates near the cathode–membrane

interface, the contributions of Pt loss and Pt coarsening to

the overall Pt surface area loss averaged over the entire

cathode have been found similar. It is assumed that each

mechanism is responsible for 50% of total Pt area loss

(reduction of roughly 1/3 of the original Pt surface area in

the fresh MEA cathode), as shown in Fig. 8 [15, 16]. It

should be noted that cycled cathode samples that consist of

Pt nanoparticles supported on Vulcan and high surface area

carbon exhibit similar changes across the cathode thickness

[15, 16], which may indicate that the impact of support in

Pt surface area loss in this cycling experiment is small. No

significant change is noted in the anode in the cycled MEA,

as shown in Fig. 9, which further confirms that Pt disso-

lution activated at high voltages is key to Pt loss from fuel

cell electrodes and coarsening of Pt nanoparticles on car-

bon during voltage cycling.

3.3 Proposed Mechanisms for Instability of Pt

Nanoparticles in Low-Temperature Fuel Cells

In general, there are four processes that have been con-

sidered relevant to the loss of electrochemically active

surface area of Pt in the fuel cell electrodes:

1. Ostwald ripening based coarsening of individual Pt

nanoparticles on carbon, which may involve dissolu-

tion of Pt from small particles, diffusion of soluble Pt

species from small to large particles in the ionomer

phase and redeposition/reduction of soluble Pt species

onto large particles on the nanometer-scale, as shown

in Fig. 10a. This process is analogous to the Ostwald

ripening process [34–36] that typically involves trans-

port of atoms or molecules from small particles to

large particles and growth of large particles at the

expense of small ones, driven by reduction in the

surface energy. This mechanism is likely to be in part

responsible for surface area loss in the fuel cell

Fig. 7 Top: Schematic of the cycled cross-sectional MEA cathode

and TEM images were obtained sequentially in the direction

perpendicular to the membrane surface from the diffusion medium–

cathode interface to the cathode/membrane interface. Middle: (1)–(4)

are corresponding TEM micrographs obtained from locations 1, 2, 3

and 4 in the cross-sectional MEA cathode shown in the schematic

(top). Bottom: (1)–(4) Pt particle size histograms measured from

TEM micrographs (middle) of the cross-sectional cycled MEA

cathode. All four histograms were obtained from a measurement of

100 particles [15].

Top Catal

123

cathode when the cathode is exposed to voltages

higher than 0.8 V vs. RHE (steady-state or transient),

where the solubility of Pt nanoparticles is significant

[15, 20, 45–48].

2. Pt crystal migration and coalescence, which involves

motion of Pt particles and coalescence where they

meet on the carbon support, as shown in Fig. 10b.

Crystallite migration and coalescence in the absence of

electrolyte is observed clearly in gas-phase sintering

studies of Pt/C catalysts but is insignificant at temper-

atures below 500 �C in the gas-phase [49]. Although

unique experimental evidence of crystal migration of

Pt nanoparticles in low-temperature fuel cells has not

been reported, it has been proposed [24, 25] that this

process is responsible for surface area loss in the fuel

cell cathode when the cell voltages are low, since the

solubility of Pt nanoparticles is believed to be negli-

gible at voltages lower than 0.7 V vs. RHE from bulk

Pt measurements [46].

3. Detachment of Pt nanoparticles from the carbon

support and agglomeration of Pt nanoparticles, gener-

ally induced by carbon corrosion, as shown in

Fig. 10c. The extent by which this process contributes

to the loss of electrochemical activity of fuel cell

electrodes is dependent on the cell voltage, the nature

of interactions between Pt nanoparticles and the carbon

support, the degree of graphitization of carbon support,

and potentially other factors, such as the RH value.

Although corrosion of conventional carbon supports

such as Vulcan is considered negligible at cell voltages

lower than 0.8 V in low-temperature fuel cells [50],

carbon corrosion and weight loss has been shown to be

significant at voltages higher than 1.1 V vs. RHE

[50, 51].

4. Dissolution and reprecipitation of Pt single crystals in

the ionomer and membrane by chemical reduction of

soluble Pt species with hydrogen molecules, which

result in loss of Pt from the carbon support and loss of

electrode activity, as shown in Fig. 10d. A consider-

able number of hydrogen molecules can permeate

through the proton-conducting membranes in the

MEA, and the location of Pt deposition in the

membrane is dependent on the partial pressure of

oxygen in the cathode [39, 44] (the higher partial

pressure of oxygen, the closer Pt deposition relative to

the anode in the MEA). Pt coarsening via this process

Fig. 8 Relative Pt surface area loss (solid triangles) on the nanome-

ter-scale (Ostwald ripening on the carbon support) and the

micrometer-scale (diffusion of oxidized Pt species and precipitation

of Pt particles in the ionomer phase) mechanisms across the cathode

after 10,000 voltage cycles. The lower line depicts the percentages

(solid circles) of the remaining electrochemically active Pt surface

area (on the carbon support) across the electrode (lower gray region).

The upper gray region corresponds to the Pt surface area lost by Pt

precipitation in the ionomer phase. The white region depicts the

relative Pt surface area lost associated with Ostwald ripening of Pt on

the carbon support. The inserted upper TEM image: nonspherical

particles off carbon support are increasingly more abundant toward

the cathode/membrane interface. The inserted lower TEM image:

spherically shaped Pt nanoparticles on carbon predominate near the

DM/cathode interface [15].

Fig. 9 TEM micrographs of Pt

supported on high surface area

carbon at the cathode (a) and

the anode (b) after voltage

cycling between 0.6 and 1.0 V

with fully humidified nitrogen.

Detailed experimental

conditions can be found in [15].

Pt nanoparticles in the cathode

are clearly much larger than

those at the anode

Top Catal

123

occurs on the micrometer-scale, which leads to

substantial loss in Pt available for electrochemical

reactions at the cathode.

A large amount of uncertainty still surrounds the active

coarsening mechanism (Ostwald ripening or crystal

migration) under different conditions, which we discuss in

detail below.

3.3.1 Ostwald Ripening vs. Crystal Migration

Whether coarsening of Pt nanoparticles on carbon occurs by

crystal migration and coalescence or Ostwald ripening via Pt

dissolution and redeposition has been a long-standing debate

in the literature. We here discuss underlying assumptions

used to distinguish these two mechanisms and discuss their

validity. Ample experimental evidence for the presence of

soluble Pt species and growth of large Pt nanoparticles at the

expense of small particles has been reported, which is

indicative of the Ostwald ripening mechanism. However,

detailed atomic processes associated with Ostwald ripening

via Pt dissolution and redeposition are not understood. The

idealized crystal migration model implies an asymptotic

particle size distribution that is log-normal in volume (with a

distinctive tail at large particle sizes) [32] and a rate of par-

ticle growth that depends on particle loading [31]. In

contrast, Ostwald ripening results in an asymptotic particle

size distribution with a tail on the small particle end, a

maximal particle size cutoff, and a Pt-loading-independent

growth rate [36, 52, 53]. These characteristics have been

used frequently in the identification of coarsening mecha-

nisms underlying experimental data in the previous studies.

Unfortunately, these comparisons have failed to yield defi-

nite results for Pt coarsening mechanisms in PEM and PA

fuel cells. We believe this is largely due to the fact that

experimental and computational limitations have forced the

use of approximate models for complex real situations. There

are several major challenges in the deduction of Pt

(a) Growth via Modified Ostwald Ripening

Carbon support

Pt => Pt x+ + xe - Ptx+ + xe - => Pt

Dissolution Redeposition

Transportof Pt x+ complex

(c) Detachment from carbon support

Carbon support

Detachment

(b) Coalescence via Crystal Migration

Carbon support

(d) Dissolution and Precipitation in the Ion Conductor

Pt => Pt x+ + xe -

Ptx+

complex

Membrane/ionomer

H2 Pt single crystals

H2 + Pt x+ => Pt + 2H +

Fig. 10 Proposed mechanisms

for instability of Pt

nanoparticles in low-

temperature fuel cells: (a):

dissolved Pt species from

smaller particles diffuse through

ionomer phase and redeposit

onto the surfaces of larger

particles (Ostwald ripening);

(b): Pt nanoparticles migrate on

the surface of carbon support

and coalesce; (c): Pt

nanoparticles detach from the

carbon support and/or

agglomerate due to carbon

corrosion; (d): Soluble Pt

species from the cathode are

reduced and precipitated out in

the ionomer and membrane by

chemical reduction by

permeated hydrogen molecules

from the anode

Top Catal

123

coarsening mechanisms based on comparing to theoretical

load dependence, surface area loss rate laws, and asymptotic

particle size distributions. First, uncertainties reside in

comparison between experimental and computational par-

ticle size distributions as experimentally determined particle

size distributions of Pt might not reach the asymptotic limit.

Second, multiple coarsening processes can simultaneously

give rise to the experimental particle size distributions. This

can be particularly problematic in the crystal size distribution

of Pt particles in the aged cathode determined from X-ray

powder diffraction, which averages over many areas and

may consist of Pt crystals that have been coarsened by dif-

ferent processes. A particle size distribution that involves

multiple processes cannot be used to deduce the mechanism

of Pt coarsening by comparison to single process models. For

example, Ferreira et al. have shown [15, 16] Pt nanoparticles

averaged across the thickness of the cycled MEA cathode

show a particle size distribution with a tail at large Pt crystal

sizes, where most particles greater than 10 nm correspond to

Pt crystals off the carbon support. Although the particle size

distributions of image 2–4 in Fig. 7 have a tail towards large

Pt crystals, it is apparent from these TEM images and the

location of these large Pt crystals (off the carbon support) that

they cannot be formed by crystal migration and coalescence.

A third problem associated with particle size distributions

comes from tying mechanisms to the overall rate law for

surface area loss. The reaction rate laws of the different

coarsening mechanisms overlap enough that it is very diffi-

cult to differentiate these processes using kinetic parameter

fitting alone. Finally, observation of necked Pt nanoparticles

on carbon has been used to support crystal migration, which

is an important aspect of cycled cathodes not captured by the

particle size distribution based analyses discussed above. An

example of necking of Pt crystals is shown in Fig. 11, where

several single crystals of Pt on the order of*5 nm are joined

together on the carbon support. While necking is expected

during the coalescence of migrating particles, the observa-

tion of these necked Pt nanoparticles does not provide

unambiguous evidence for crystal migration as these necked

regions can be developed by effective capture of soluble Pt

species and preferred Pt deposition in the region between

adjacent Pt nanoparticles. Therefore, to the authors’

knowledge, there is no unique experimental evidence in the

literature to support the mechanism of crystal migration and

coalescence in PEM fuel cells.

Corrosion of carbon support can contribute to the

movement of Pt nanoparticles, which can lead to coales-

cence of Pt nanoparticles and the formation of necked

regions, which further complicates the identification of

active crystal migration. Although it is believed that carbon

corrosion is not significant upon voltage cycling between

0.6 and 1.0 V with fully humidified nitrogen, a few primary

carbon particles in the MEA cathode have been found to

exhibit rougher particle edges and non-uniform transmitted

intensities within the particle in comparison to pristine

Vulcan particles, as shown in Fig. 12a–b. It is surprising to

note, however, that carbon corrosion in the cathode has

been shown to be severe in a fuel cell held at a low voltage

of *0.67 V vs. RHE with flowing air in a recent work of

Guilminot et al. [54]. Localized thinning (corrosion) of

carbon particles may weaken the attachment of Pt nano-

particles on the carbon support and may result in migration,

coalescence and agglomeration of Pt nanoparticles.

Although the electrochemical area loss of Pt in the MEA

cathode upon voltage cycling with the upper voltage limit

higher than 0.8 V has been attributed primarily to coars-

ening of individual Pt nanoparticles on carbon and Pt loss

from the carbon support associated with Pt dissolution and

redeposition, movement and coalescence of Pt nanoparti-

cles on carbon can also contribute the MEA cathode

surface loss. The extent may depend on the interactions

between Pt nanoparticles and the carbon support, the

degree of graphitization of carbon support, etc. Although

there is no definitive experimental evidence for crystal

migration of Pt nanoparticles under PEM fuel cell condi-

tions, the possibility is still under active consideration,

particularly for Pt particles supported on graphitized sur-

faces. Therefore, we here discuss what might influence the

mobility of Pt particles from theoretical grounds.

3.3.2 Mobility of Pt Nanoparticles on Carbon

A more quantitative understanding of Pt particle mobility

on carbon is desired to establish if crystal migration is

likely to play a significant role. This would allow one to

assess the likely contribution of crystal migration under

PEM fuel cell temperatures. Some insight on the possible

physics underlying crystal migration can be obtained by

stepping back from the full PEM fuel cell liquid environ-

ment and considering the simpler case of diffusion in the

gas environment. If we assume a typical interparticle dis-

tance of l, we can then assume that significant coarsening

occurs in a time s over which particles move approximately

this distance. An effective diffusion constant D can be

estimated by D = l2/s. By considering l = 10 nm (corre-

sponding to a particle density of 1012/cm2) and a

characteristic time of about s = 105 s (approximately the

time to lose a quarter of the surface area at 600 �C in the

data from Bett et al. [49]) we get D(600 �C) = 10-21 m2/s.

While this is very approximate, it can be usefully compared

to recent molecular dynamic simulations by Chen and

Chan [55], who have found D values around 10-12 m2/s at

room temperature for Pt nanoparticles up to 500 atoms

(about 2.5 nm diameter) on an idealized graphite surface.

Although the simulations make use of quite simplified

Top Catal

123

interatomic potentials, the vast difference of an increase of

9 orders of magnitude, which would only be further

increased by performing the molecular dynamics at

600 �C, strongly suggests that a model of Pt particles dif-

fusing on idealized graphite is inadequate to explain the gas

phase data of Bett et al. [49]. The discrepancy suggests that

other Pt–carbon interactions, presumably trapping at car-

bon defects, must dominate the physics of Pt particle

migration. This hypothesis was suggested by Bett et al.

[49] who used a trapping model from Phillips et al. [56] to

explain anomalously fast coarsening for high Pt loading. In

order to explain their surface area changes in nitrogen for

5 wt% Pt on carbon at 600 �C, Bett et al. [49] proposed a

trapping energy of 1.8 eV. This large value would likely

rule out any significant transport at 80 �C. However, it is

not clear that this large trapping energy would persist in a

liquid environment. For example, a likely candidate for

trapping sites are step edges, whose dangling bonds have

been shown to strongly bond Au atoms in ab initio studies

[57]. Competitive binding from ions in the electrolyte for

these step edges might greatly reduce the effective trapping

barrier, enhancing mobility of Pt particles by many orders

of magnitude. If such a model were valid, a strong

dependence of Pt coarsening on the carbon defect con-

centration is expected. Further work is necessary to

determine realistic trapping models to quantitatively

describe the Pt particle motion in an electrolyte environ-

ment and assess the conditions under which particle

migration and coalescence might be active. Such models

would provide new insights into synthesis of stable metal

nanoparticles supported on graphitized surfaces.

4 Solubility of Bulk Pt and Pt Nanoparticles in Acid

Solutions

Understanding and reducing Pt dissolution is key to

increase Pt catalyst stability in low-temperature fuel cells.

In particular, dissolution underlies the 3D Ostwald rip-

ening mechanism, which can enable the growth of large

Pt nanoparticles at the expense of small ones (Fig. 10a)

and loss of Pt from the carbon support via precipitation of

Pt in the ionomer and membrane (Fig. 10d). Therefore, it

is of great importance to understand processes and

parameters that influence Pt dissolution so that they can

be controlled. The starting point for understanding Pt

dissolution in PEM fuel cells is the equilibrium concen-

tration of dissolved Pt at a given potential and

temperature in acidic environments (consistent with those

found in PEM fuel cells).

Fig. 12 TEM micrographs

show the carbon structures in

the MEA cathode before (left)

and after (right) voltage cycling

between 0.6 and 1.0 V with

fully humidified nitrogen.

Detailed experimental

conditions can be found in [15].

The right image exhibits

rougher particle edges and non-

uniform transmitted intensities

within the particle in

comparison to carbon particles

in the pristine Pt/Vulcan sample

Fig. 11 TEM images from low to high magnifications (left to right)

show Pt nanoparticle necking in the cycled MEA cathode after

voltage cycling between 0.6 and 1.0 V with fully humidified nitrogen.

Detailed experimental conditions can be found in [15]. Several single

crystals of Pt on the order of *5 nm joint together on the carbon

support

Top Catal

123

4.1 Potential and Temperature Dependent Pt Solubility

Both potential and temperature can strongly influence

equilibrium Pt solubility. In order to understand Pt disso-

lution thermodynamics relevant to stability of Pt

nanoparticles in low-temperature fuel cells, we here com-

pare equilibrium Pt solubility behavior as predicted by

Pourbaix and reported steady-state Pt solubility data in the

literature.

4.1.1 Nernstian Prediction of Pt Solubility

Pt is electrochemically stable in acids at low potentials

relative to RHE potential but Pt dissolves in acidic solu-

tions at high potentials. At 25 �C, Pourbaix [45] has

suggested that Pt dissolution can occur at voltages greater

than 0.85 V vs. RHE and pH values in the range of 0 and -

2. Following Pourbaix [45], Pt may be oxidized by the

following two reactions:

PtðsÞ $ Pt2þðaqÞ þ 2e� ð1aÞ

PtðsÞ þ H2O$ PtOðsÞ þ 2Hþ þ 2e� ð1bÞ

where PtO(s) may be dissolved chemically in acid to form

Pt2+. The equilibrium concentration of soluble Pt2+ in

reaction (1a) can be related to the applied potential E using

the Nernst equation:

cPt2þ ¼ exp2FðE � E0ðTÞÞ

RT

� �ð2Þ

where the activity coefficient of soluble Pt2+ ions is

assumed to be 1 and cPt2þ is the Nernstian concentration of

soluble Pt2+ in molar (M) under the applied voltage E. E

and E0(T) are the applied and reversible voltages versus

RHE at a given temperature T, respectively. It should be

noted that as the Nernstian concentration of soluble Pt2+ in

Eq. 2 is independent of pH. RHE and standard hydrogen

electrode (SHE) may be used interchangeably in the

following discussion. At 25 �C, E0 = 1.188 V vs. SHE or

RHE [45]. The Nernstian concentration of Pt2+ as shown by

Pourbaix is very low but increases over three orders of

magnitude from 0.8 to 0.95 V vs. RHE, as shown in

Fig. 13. The Nernstian principle in Eq. (2) predicts that, on

a log plot vs. potential, the overall magnitude of the Pt2+

concentration for zero potential is set by -E0(T) 9 2F/

(2.303RT). The slope is such that a 10-fold increase of the

Pt2+ concentration is correlated to every 2.303RT/2F

increase in potential, which translates to 29.5 mV at

25 �C and 46.5 mV at 196 �C. Bindra and Yeager [46]

have measured dissolution of Pt foil in the voltage range of

0.8 to 0.95 V in 96% H3PO4 at 196 �C, and have shown

Nernstian potential dependence that is in good agreement

with the reaction in Eq. 1 as proposed by Pourbaix, as

shown in Fig. 13. It should be noted that the concentration

of soluble Pt species increases considerably with

temperature. This is primarily due to the RT term in the

denominator of the exponential, but also in part due to the

reduction in the equilibrium potential E0(T) at elevated

temperatures. The equilibrium potential E0(T) is related to

the change in Gibbs free energy of the chemical reaction

associated with oxidation of Pt and reduction of hydrogen,

which can be written as:

DG0ðTÞ ¼ �ð2FÞE0ðTÞ ð3ÞAlthough the temperature-dependent data of DG0(T) are

not available in the literature to the authors’ knowledge,

one can estimate DG0(T) and E0(T) at other temperatures

from the values at 25 �C [45] and 196 �C [46] from the

following equation [59]:

DG0ðT2Þ � DG0ðT1Þ ¼ DC0PðT2 � T1Þ

� DS0ðT298 KÞðT2 � T1Þ � DC0PT2 ln

T2

T1

ð4Þ

We can further simplify Eq. 4 and obtain a linear

temperature dependence by assuming that DC0P for the

chemical reaction is very small:

DG0ðT469 KÞ � DG0ðT298 KÞ � CðT469 K � T298 KÞ ð5Þ

Here C is a constant that can be fit to experimental

dissolution data. One may estimate DG0(T) and E0(T) at a

Fig. 13 The equilibrium dissolved Pt concentrations taken from

previous studies of Ferreira [15], Wang-pc [47], Wang-np [58], Honji

[20], Bindra [46], and predicted with a Nernstian relationship for

Pt ? Pt2+ + 2e-. Pourbaix data are calculated from Eq. 2 for

temperatures 25, 80, and 196 �C [45]. Experimental details for each

of the other data sets are given in Table 1

Top Catal

123

given temperature between 25 and 196 �C from the

following:

DG0ðTÞ � DG0ðT298 KÞ þ CðT � T298 KÞ ð6Þ

�2FE0ðTÞ ¼�DG0ðT298 KÞ þ

�DG0ðT469 KÞ � DG0ðT298 KÞ

T469 K � T298 K

�

� ðT � T298 KÞ�

ð7Þ

DG0(T298 K) can be calculated from the Pourbaix

E0(298 K) using Eq. 3 while the value of DG0(T469 K)

can be estimated from Bindra and Yeager’s Pt

concentration data [46] using the Nernst principle and

Eq. 3. The solubility of Pt as a function of potential at

80 �C can then be interpolated from Eq. 7 and is shown as

the dashed line in Fig. 13.

4.1.2 Reported Pt Solubility Data

Solubility data of Pt in acid reported in the literature are

compared in Fig. 13, and key experimental conditions of

these studies such as Pt morphology (e.g., due to Pt foil vs.

nanoparticles), temperature, and atmosphere are summa-

rized in Table 1 (a more complete set of solubility data can

be found in Supporting Information—Fig. S1). We here

compare experimental findings of Pt solubility with the

proposed surface oxidation mechanism of Pt and the

Nernstian prediction in detail. Although the reported con-

centrations of soluble Pt species are found to generally

increase with increasing temperature and voltage at volt-

ages lower than *1.0 V vs. RHE, the potential dependence

and the magnitude of solubility considerably deviate from

the Nernstian behavior. This deviation can be thought of as

occurring in two ways. First, the solubility of polycrystal-

line Pt (Wang-pc in Fig. 13) and Pt nanoparticles (Wang-

np in Fig. 13) has been shown to first increase in the

voltage range from 0.85 to 1.0 V vs. RHE, and decrease or

reach a plateau at voltages greater than about 1.1 and 1.0 V

vs. RHE, respectively [47, 58]. This observation is con-

sistent with the proposed surface oxidation mechanism of

Pt, where the growth and place-exchange PtO atomic layers

can passivate the Pt surface and reduce Pt dissolution [61,

62] (see further discussion of the oxidation process below).

In addition, Darling and Meyers have used a kinetic model

to show that complete coverage of PtO on Pt leads to sharp

reduction in the solubility of Pt nanoparticles at voltages

greater than 1.1 V vs. SHE at 50 �C and 176 �C [42].

Second, the solubility of Pt in acid has been found to

increase considerably with increasing temperature and

voltage, which is in general agreement with the Nernstian

principle. However, with the exception of the data reported

by Bindra and Yeager [46], which show the Nernstian

potential dependence (one decade increase in the Pt2+

concentration for every 46.5 mV increase in voltage at

196 �C in Fig. 13), all other reported data of Pt solubility

show a much weaker voltage dependence than that

expected from the Nernstian behavior. This discrepancy

can be seen clearly by comparing the experimental data of

Wang et al., [47, 58], Ferreira et al. [15], and Honji et al.

[20] to the Nernstian potential-dependent solubility values

(the Pourbaix lines at 25 �C, 80 �C and 196 �C in Fig. 13)

as predicted from Eq. 2 at voltages lower than 1.1 V vs.

RHE. The results of Pt nanoparticles reported by Ferreira

et al. and that of Pt polycrystalline wire reported by Wang

et al. (Wang-pc) show an approximately n = ½ Nernstian

behavior in regions between 0.85 and 1.1 V, where n

represents the number of electrons involved in the

Table 1 Summary of characteristics of studies of dissolved Pt

concentration in acid. Temperature column lists the temperature at

which the dissolution equilibrates, acid is the liquid into which the Pt

is dissolved, Pt morphology is the structure of the solid phase Pt

source, max. dissolved monolayers gives the equivalent number of Pt

monolayers from the surface needed to yield the maximum measured

dissolved Pt, and atmosphere gives the gas environment in which the

experiment is performed

Name Temperature (�C) Acid Pt morphology Max. dissolved monolayers Atmosphere

Ferreira 80 0.5 M sulfuric Nanoparticle 2 9 10-2 N2

Wang-pc 23 0.57 M perchloric Polycrystalline 2 9 10-1 Ar

Wang-np 23 0.57 M perchloric Nanoparticle 1 9 10-2 Ar

Ota-a 23 1 M sulfuric Pt black 5 9 10-2 O2/Air

Ota-b 51 1 M sulfuric Pt black 1 9 10-1 O2

Ota-c 76 1 M sulfuric Pt black 3 9 10-2 Air

Honji 206 19 M phosphoric Nanoparticle 5 9 10-1 N2

Bindra 196 18.6 M phosphoric Foil \6-1440 N2

Komanicky 25 0.6 M perchloric Single crystal surfaces 5 9 10-1 Ar

The names refer the first authors in the following references: Ferreira [15], Wang-pc [47], Wang-np [58], Ota-a,b,c [48], Honji [20], Bindra[46],

Komanicky [60]. Wang-pc and Wang-np refer to polycrystalline and nanoparticle samples, respectively. In all data sets potential is controlled by

an electrode except for Ota-a,b,c, where the potential is set by equilibration with an oxygen containing atmosphere

Top Catal

123

electrochemical dissolution reaction and affects the

Nernstian potential dependence by 2.303RT/(nF). This

n = 1/2 behavior cannot be explained by the reaction

model in Eq. 1 from the Nernstian point of view. In

addition, Wang et al. [58] have shown that Pt nanoparticles

exhibit even lower potential dependence (n = *0.2) and

similar concentrations of soluble Pt species in comparison

to the Pt polycrystalline sample in the voltage range from

0.85 to 1 V at 23 �C. Moreover, it should be mentioned

that Komanicky et al. [60] have recently studied the solu-

bility of Pt single crystal surfaces at 25 �C in 0.6 M

perchloric acid and have found that the solubility is not

monotonically dependent on potential in the voltage range

from 0.7 and 1.1 V vs. RHE (see Supporting Information—

Fig. S2). As their concentrations of soluble Pt species in

acid were measured after only 24 h, which, under room

temperatures conditions, was found by Wang et al. [47] to

be too short to reach steady-state Pt concentrations, the

results of Komanicky et al. are not included in Fig. 13 and

are not discussed in the context of the Nernstian principle.

Lastly, considerably higher concentrations of soluble Pt

species than those predicted by Pourbaix are noted at

voltages lower than 0.95 V vs. RHE for a given tempera-

ture, which is particularly pronounced for the solubility

data of Pt nanoparticles reported by Wang et al. [47, 58],

Ferreira et al.[15], and Honji et al. [20], as shown in

Fig. 13.

The discrepancy between reported Pt solubility data and

predicted values of the Pourbaix model is not presently

understood. Provided that impurity effects and analytical

measurement errors can be excluded from our experiments,

the deviation from the Nernstian behavior may be attrib-

uted to: (i) multiple chemical or electrochemical processes

governing the concentration of soluble Pt species; (ii)

surface morphologies of Pt samples that influence the

thermodynamics and the kinetics of Pt dissolution if less

than one monolayer of Pt is dissolved; (iii) particle size

effects on the solubility as the surface energy of nanopar-

ticles can influence the thermodynamics and kinetics of Pt

dissolution. (i) If Pt dissolution were controlled by the

reaction in Eq. 1a alone, the concentration of soluble Pt

species should be independent of pH, as shown in Eq. 2.

However, Ota et al. [48] have shown that the solubility of

Pt nanoparticles at 1.03 and 1.08 V vs. RHE decreases with

increasing pH, which suggests that chemical reactions such

as dissolution of PtO in acid and/or other electrochemical

reactions similar to the reaction in Eq. 1b play an important

role in Pt dissolution. Further experimental and theoretical

studies are required to elucidate the process of Pt dissolu-

tion in acid at the atomic scale; (ii) surface morphologies of

Pt samples that influence the thermodynamics and the

kinetics of Pt dissolution if less than one monolayer of Pt is

dissolved. We will examine and discuss the amount of Pt

dissolved in the previous studies listed in Table 1 [15, 20,

46–48, 58, 60] with respect to the observed potential

dependence; (iii) particle size affects on the solubility as

the surface energy of nanoparticles can influence the

thermodynamics and kinetics of Pt dissolution. For exam-

ple, the experimental data reported by Pourbaix [45] and by

Bindra et al. [46] were conducted with polycrystalline Pt

foils while those of Ferreira et al. [15] were conducted on

Pt nanoparticles, and differences between these might

therefore be expected. We will discuss the influence of the

Gibbs–Thomson term on the stability and dissolution rate

of Pt as a function of particle size below.

4.1.3 Surface Morphological Effects

We examine the amounts of Pt being dissolved in these

previous studies [15, 20, 46–48, 58, 60] in order to

understand the cause of the non-Nernstian behavior of

reported Pt solubility. Methods to detect Pt dissolution in

the literature include weight loss of Pt [46, 63], inductive

coupled plasma (ICP) [15, 47, 60], dithizone—benzene

method [48], and atomic absorption spectrophotometry

[64]. Based on the measured Pt concentrations and elec-

trolyte volumes it is straightforward to estimate the total

Pt dissolved from the samples. This can be combined with

measured sample areas and a density of Pt atoms on the

surface (estimated to be 1.50 9 1015 atoms/cm2 for the

{111}Pt surfaces based on a lattice parameter of 3.92 A

for bulk Pt) to assess how many monolayers (on average)

of Pt are dissolved during the experiment. These values

are given in Table 1 and are discussed in detail below. It

is interesting to note that the solubility data of Bindra and

Yeager are associated with dissolution of multiple mon-

olayers of Pt from a polycrystalline Pt sample while less

than one monolayer of Pt is dissolved in the experiments

of the other studies [15, 20, 47, 48, 58, 60]. Although the

true electrochemical surface area was not measured in the

study of Bindra and Yeager, we can estimate an upper

bound on the number of monolayers dissolved from the

nominal surface area of the electrode. Dissolution of 6,

14, 144, and 1,440 monolayers can be found at 0.8, 0.85,

0.9, and 0.95 V, respectively. In general, these are quite

large numbers, showing that a significant amount of the Pt

is accessed and dissolved. When a large number of Pt

monolayers are involved, it is reasonable to expect that

surface morphology, surface defects, surface impurities

and impurities in solution have negligible effects on

overall equilibrium properties, which is in good agree-

ment with the fact that the measured Pt solubility of

Bindra and Yeager [46] follows the Pourbaix model

(Fig. 13). On the other hand, dissolution of Pt in the top

one monolayer or less could be strongly influenced by

Top Catal

123

surface structure and chemistry of Pt samples, and is not

well described by the bulk equilibrium thermodynamic

analysis—the Nernstian principle formulated in Eq. 2.

The onset of non-Nernstian behavior for sub-monolayer

Pt dissolution is supported by the observation that the

concentration of soluble Pt species at 0.8 V reported by

Bindra and Yeager [46] slightly deviates from the

Nernstian behavior. At 0.8 V only the top *2 monolayers

were dissolved (this was determined by assuming a

roughness factor, the true surface area of Pt over the

nominal electrode surface area, of *3 [65] for the Pt

polycrystalline foil sample). Less stable surface Pt sites

near step edges, grain boundaries, dislocations, and

impurities may be dissolved preferentially to Pt sites

located on low-index facets such as {111}Pt and {100}Pt.

Analysis of the experiments of Ferreira et al. [15], Wang

et al. [47, 58], and Honji et al. [20] shows that only a

maximum of 0.02, 0.2 (polycrystalline) and 0.0004

(nanoparticles), and 0.5 monolayers are dissolved,

respectively. Therefore, we propose that the weak

potential dependence and the deviation from the Nernstian

behavior found in these studies (shown in Fig. 13) results

from the fact that these experiments are within the sub-

monolayer dissolution limit, where Pt dissolution is

strongly influenced by surface structure and chemistry of

Pt. Although the steady-state condition is not reached,

Komanicky et al. [60] have shown that Pt dissolution and

passivation are strongly influenced by surface structure

upon dissolution of less than 0.5 monolayer. The {111}pt

surface showed the lowest Pt solubility at 0.95 V versus

RHE relative to the {100}Pt, {110}Pt and the {111}Pt—

{100}Pt nanofacets, as shown in Fig. S2. The proposed

importance of non-equilibrium Pt dissolution is consistent

with the theory that accelerated Pt dissolution rates upon

voltage cycling compared to extended holds at constant

potentials [40, 47, 66] are due to the difference in the

surface structure of Pt developed between the transient

and steady-state operation. Voltage cycling may continu-

ously create high-energy Pt surface sites, which might be

associated with the relatively high dissolution rates. While

it is likely that the non-Nernstian behavior of previous

solubility studies [15, 20, 47, 48, 58, 60] is connected to

the non-equilibrium nature of the submonolayer dissolu-

tion process, it is still an open challenge to create a

realistic model of the surface thermodynamics that can

quantitatively explain the dissolved concentrations.

4.1.4 Proposed Surface Oxidation Processes of Pt

Any realistic surface model that explains Pt solubility will

have to take into account oxidation, which changes the

chemistry, stability, and kinetics of the surface. We here

give a brief overview of surface species formed upon

electrochemical oxidation of bulk Pt surfaces as reported

in the literature. Conway et al. [67] have proposed the

following process upon electrochemical oxidation of Pt

from a quantitative potentiodynamic cyclic voltammetry

(CV) study of polycrystalline Pt in 0.5 M H2SO4 at

25 �C:

4Pt + H2O! Pt4OH + Hþ þ e�(� 0.89 V vs. RHE)

Pt4OH + H2O! 2Pt2OH + Hþ + e�(� 0.95 V vs. RHE)

Pt2OH + H2O! 2PtOH + Hþ + e�(� 1.05 V vs. RHE)

Place - exchange: PtOH ! OHPt;

OHPt ! PtO + Hþ + e�(� 1.05 V vs. RHE)

Oxide formation : PtO! PtOnð[ 1:1 V vs:RHEÞ

These proposed intermediate species and processes have

been investigated extensively in subsequent studies. Allen

et al. [68] and Winograd et al. [69] have both confirmed

the presence of PtO species from their X-ray photoemission

spectroscopy measurements by electrochemical oxidizing

smooth Pt foil in acid at constant voltages above 2 V vs.

RHE at room temperature. In addition, You et al. [70] have

confirmed the place-exchange step by surface X-ray scat-

tering and proposed two possible mechanisms of the place-

exchange step: (i) PtOH ? HOPt, as proposed by Conway

et al. [67]; (ii) PtO ? OPt. Jerkiewicz et al. [61] have

recently performed combined studies of CV, electro-

chemical quartz-crystal nanobalance, and Auger electron

spectroscopy on sputtered Pt samples, which significantly

improve our understanding of Pt electro-oxidation and

refine the mechanism proposed by Conway et al. [67].

These authors have revealed that (i) the Pt surface is not

fully passivated (the monolayer coverage of chemisorbed

oxygen is less than unity) at *1.2 V vs. RHE; (ii) the

place-exchange step does not happen with PtOH, but

occurs with PtO (PtO ? OPt) on the unpassivated surface

of sputtered Pt samples at *1.2 V vs. RHE. We speculate

that these oxidized surface species formed between *0.9

and 1.2 V can dissolve chemically, and one or multiple

oxidation processes proposed above can govern the sub-

monolayer dissolution thermodynamics and kinetics of Pt

nanoparticles. In addition, surface defects of Pt nanoparti-

cles could shift the onset voltages of these processes

relative to polycrystalline Pt. Therefore, understanding the

nature of surface species at different voltages is key to

explain the reported submonolayer Pt solubility data sum-

marized in Fig. 13 and improve the stability of Pt

nanoparticles.

Top Catal

123

4.2 Particle Size Effects on Pt Dissolution

Thermodynamics and Kinetics

The dissolution thermodynamics and kinetics of Pt nano-

particles can be significantly different from that of bulk Pt

as Pt nanoparticles have more surface per unit volume,

which creates an effective pressure and drives up the

activity of the solid phase. The change in the chemical

potential, EGT, of a particle as a function of diameter, d, can

be described using the Gibbs–Thomson formula (see, e.g.

[71]):

EGT ¼ lðdÞ � lð1Þ ¼ 4cX=d ð8Þ

where c is the surface energy and X is the volume per atom.

EGT is plotted as a function of particle diameter in Fig. 14,

where cPt = 0.148 eV/A2 and XPt = 15.4 A3/atom [42] are

used. We here estimate and discuss the effect of EGT on the

dissolution thermodynamics and kinetics of Pt nanoparti-

cles as a function of particle diameter.

4.2.1 Effects on Dissolution Thermodynamics

From a theoretical perspective, the impact of EGT on the

dissolution thermodynamics should be quite significant in

the particle size range of supported Pt nanoparticles in low-

temperature fuel cell catalysts. The Gibbs–Thomson effect

can be included in Eq. 2 by modifying E0 to yield

cPt2þ ¼ exp2F E � EoðTÞ þ EGT=2ð Þ

RT

� �

¼ exp2F E � EoðTÞ þ 2cX=dð Þ

RT

� �ð9Þ

where EGT is measured in eV/atom and the two comes from

the number of electrons in the Pt dissolution reaction. It can

be seen from Eq. 9 that EGT/2 enters into the Nernst equation

for Pt concentration as an effective voltage—the energy

change per electron. Within this simple framework solubility

for a nanoparticle is formally equivalent to solubility of a

macro-scale particles at a higher potential. Based on Eq. 9

and Fig. 14, it is clear that the Gibbs–Thomson energy

should have a large impact on solubility at small diameters.

In particular, the dependence is such that EGT changes most

dramatically in the relevant energy scales below 5 nm.

Between d = ? and 5 nm EGT increases by only 0.18 eV

while between d = 5 nm and 1 nm it increases by 0.73 eV.

Therefore, the potential that a particle feels driving disso-

lution to Pt2+ (2 electrons involved) effectively increases by

0.37 V from 5 to 1 nm, which should create a strong

dependence of the equilibrium concentration of dissolved Pt

on the particle size. However, there is little experimental

evidence indicating that the Gibbs–Thomson effect is play-

ing a major role in the solubility data as summarized in

Fig. 13. The most direct test for the influence of Gibbs–

Thomson energetics is to compare the two data sets of Wang-

pc and Wang-np [47, 58]. Assuming the Pt wire used in the

Wang-pc study can be treated as having infinite diameter

particles and assuming an average particle size of 2.5 nm for

the Wang-np study, Eq. 9 implies that the Wang-np data

should behave as if effectively destabilized by 0.18 V

compared to Wang-pc. However, there does not seem to be

any clear enhancement of dissolution from the Gibbs–

Thomson effect in the Wang-np compared to Wang-pc data,

as shown in Fig. 13. It should be noted that a much smaller

fraction of a monolayer was dissolved in the solubility

measurement of Pt nanoparticles relative to that of Pt poly-

crystalline wires. The absence of discernable solubility

enhancement associated with small nanoparticles as a result

of Gibbs–Thomson effect is consistent with our hypothesis

that the solubility of Pt nanoparticles is largely dominated by

defect energetics on the particle surface when less than one

monolayer of Pt is being dissolved in the solubility

measurement.

4.2.2 Effects on Dissolution Kinetics

The Gibbs–Thomson energetics would not only increase Pt

solubility but also enhance the dissolution rates of Pt

nanoparticles. We explore this effect using simulation

techniques. The simulation follows the model of Darling

and Meyers [42] although some parameters have been refit

and the effect of Pt precipitation through hydrogen cross-

over from the anode has been included (details of this

model will be published in a forthcoming paper [72]). The

Fig. 14 The Gibbs–Thomson energy as a function of a spherical

nanoparticle diameter calculated from Eq. 8 with parameters for Pt.

The surface energy and the atomic volume of Pt are taken to be

cPt = 0.148 eV/A2 and XPt = 15.4 A3/atom [42], respectively

Top Catal

123

result of the simulation is shown in Fig. 15, where a con-

stant potential of 0.95 V, 0.4 mg/cm2 Pt loading, and

10 lm of electrolyte adjacent to the cathode are assumed.

Figure 15 shows the simulated particle dissolution rate as a

function of particle diameter, where a very strong depen-

dence of dissolution rates on size is noted with an

enormous enhancement for smaller particles (\5 nm). The

dissolution rates can be related to a lifetime of the particle

if we assume the dissolution rate continues at a fixed flux of

Pt per unit area until the particle dissolves completely. It is

clear that from 4 to 2 nm a critical crossover occurs

between lifetimes on the scale of many thousands of hours

to ones on the scale of tens to hundreds of hours. It should

be noted that these ‘‘lifetimes’’ are the time to complete

dissolution assuming the particle keeps a constant disso-

lution flux per unit area equal to the one calculated

instantaneously at a given diameter. This lifetime will

therefore overestimate the true lifetime of the particle,

since the flux will increase due to increased Gibbs–

Thomson energy during dissolution. The lifetimes should

be understood as a convenient way to think about the

magnitude of the flux per unit area coming off the particle

at each diameter. The calculated flux values are not very

accurate, since the exponential dependencies in the disso-

lution process are sensitive to small errors in parameter

values, but the qualitative picture is quite clear. Near 5 nm

particles have the stability to survive (at relatively high

constant potential holds) for the thousands of hours

required in PEM fuel cell automotive applications [3], but

for sizes near 1–2 nm the particles dissolve almost

instantly compared to PEM fuel cell lifetime goals. The

size dependence shown in these simulations strongly sug-

gests that even relatively small increases in particle sizes

from the approximately 2–3 nm usually used today could

greatly enhance stability.

Similar to dissolution thermodynamics, there is no clear

experimental verification of the Gibbs–Thomson effect on

dissolution rates in the literature. Under constant voltage

hold at 0.9 V vs. RHE, Wang et al. [47] have reported that

dissolution rates for Pt polycrystalline wire and Pt nano-

particles are nearly identical (1.7 9 10-14 and 1.4 9 10-

14 g/cm2 s, respectively). It is likely that dissolution rates

for submonolayer dissolution would be driven by non-

equilibrium properties quite distinct from those described

by Gibbs–Thomson and the theoretical model based on

Nernstian behavior.

4.2.3 Gibbs–Thomson and the Stability of Heat-Treated Pt

Nanoparticles upon Voltage Cycling

Given the stability implications of the Gibbs–Thomson

equation, one expects that increasing Pt particle sizes and

decreasing the number of particles smaller than 2 nm

would significantly enhance stability. Recently Makharia

et al. [50] have reported that the stability of Pt nanopar-

ticles upon potential cycling in PEM fuel cells can be

improved by heat-treatments at high-temperatures [50].

However, neither the mechanism by which the stability

was obtained nor the detailed heat-treatment condition

was reported. Heat-treatments at high temperatures can

increase particle sizes and decrease the number of small

Pt nanoparticles by crystal migration, as reported previ-

ously [21, 49, 53]. Here we describe a recent experiment

that demonstrates the connection between changes in Pt

particle size through heat treatments and reduced Pt sur-

face area loss and enhanced Pt stability during potential

cycling in acid.

We here examine three Pt/C samples: pristine TKK

28 wt%, and two heat-treated samples at 900 �C in argon

atmosphere for 1 min (HT–1) and 15 min (HT-2). The

histograms of these three samples are compared in Fig. 16,

where Pt particle sizes increase upon heat-treatments. It is

interesting to note that the fraction of Pt particles in the

range of 1–2 nm is reduced considerably upon heat-treat-

ments, which can be seen clearly in Fig. 16 and in the high-

resolution TEM images in Fig. 17a–b. Small Pt particles

presumably migrate on the carbon support, join with other

particles and coalesce into large particles during the heat-

treatment, as reported previously [21, 49, 53]. These three

samples were then cycled from 0.6 to 1.0 V vs. RHE in

0.5 M H2SO4 at a scanning rate of 20 mV/s at 23 �C. Heat-

treated samples show greatly reduced Pt surface area loss

and improved stability upon potential cycling, as shown in

Fig. 18a. It is hypothesized that heat-treated samples hav-

ing lower fractions of Pt nanoparticles smaller than 2 nm

relative to the pristine sample have reduced Pt solubility,

which reduces Pt loss and surface area loss upon voltage

Fig. 15 Predicted dissolution properties as a function of Pt particle

diameter. The simulation is based on an extension of the model of

Darling and Meyers [42] to include particle size distributions and

hydrogen crossover [72]. The simulation is based on a uniform

particle size with a constant potential of 0.95 V, 0.4 mg/cm2 Pt

loading, and 10 lm of electrolyte adjacent to the cathode

Top Catal

123

cycling. As no significant change in the Pt particle size

before and after voltage cycling was noted in the HT-2

sample, it is believed that the surface area loss shown in

Fig. 18a largely result from Pt dissolution. In addition, the

rate of surface loss in the first 400 cycles is markedly

reduced in the heat-treated samples. We propose that the

rapid surface area loss of Pt nanoparticles in the pristine

sample upon initial potential cycling largely results from

rapid dissolution of small Pt nanoparticles of 1–2 nm into

the acid solution. This is in good agreement with experi-

mental TEM observations that most small Pt particles

disappeared in the pristine Pt/C sample after potential

cycling, as shown in the histograms of Pt particles before

and after voltage cycling in Fig. 18b. These experimental

observations support the theoretical picture developed

above of the dramatically decreased stability of particles in

the 1–2 nm region associated with the Gibbs–Thomson

effect.

5 Concluding Remarks

Key findings from previous studies of electrochemical

surface area loss of supported Pt electrocatalysts in low-

temperature fuel cells, and the current status in the

understanding of fundamental molecular scale mechanisms

that drive Pt degradation can be summarized:

1. Electrochemically active surface area is degraded

through Pt loss from fuel cell electrodes and coarsen-

ing of Pt nanoparticles.

2. The underlying mechanism of Pt loss is likely through

dissolution of Pt at relatively high potentials (greater

than *0.8 V).

3. The underlying mechanism of coarsening of Pt nano-

particles likely occurs by Pt dissolution and

redeposition, which is analogous to Ostwald ripening.

4. The role of particle migration and coalescence in the Pt

surface area loss at low-temperature fuel cells is still

uncertain. Although there is no unique experiment

evidence in support of crystal migration and coalescence,

they cannot be ruled out from a theoretical standpoint.

5. The nature of Pt dissolution is still poorly understood.

It is proposed that equilibrium thermodynamic con-

siderations do not apply to the submonolayer

dissolution processes often relevant for fuel cells,

and that Nernstian behavior cannot be expected for

submonolayer dissolution of Pt nanoparticles. How-

ever, not even a qualitative theory that can explain

trends in submonolayer Pt solubility with temperature

and potential has been put forward.

It is clear that Pt dissolution, leading either to coarsening

through Ostwald ripening or loss of Pt through precipita-

tion in the ion-conducting phase, is a major source of

degradation of the Pt electrocatalysts in low-temperature

fuel cells. The non-Nernstian behavior of measured Pt

dissolution, coupled with the observation that dissolution

typically involves only a fraction of a monolayer, suggests

Fig. 16 (a): Histograms of Pt nanoparticle size distribution of

28 wt% Pt supported on high surface area carbon (TKK). Black:

pristine; grey shaded: a sample (HT-1) heat-treated at 900 �C under

argon for 1 min; light grey: a sample (HT-2) heat-treated at 900 �C

under argon for 15 min

Fig. 17 High resolution TEM

images of (a) pristine 28 wt% Pt

nanoparticles supported on

surface area carbon (TKK) and

(b) a sample (HT-1) heat-treated

at 900 �C in argon atmosphere

for 1 min. Note that the number

of Pt nanoparticles smaller than

2 nm is decreased considerably

after the heat-treatment

Top Catal

123

that non-equilibrium processes may dominate. The reali-

zation of the importance of submonolayer Pt dissolution

suggests that controlling Pt surface structure and chemistry

could be a very powerful tool to reduce degradation of Pt

electrocatalysts. For a 2 nm diameter particle about half the

atoms are in the first monolayer, which implies that very

significant amounts of Pt and Pt surface area can be lost

through dissolving only the first monolayer. There is

already some evidence that manipulating surface structure

and chemistry could increase durability. Komanicky et al.

[60] have shown that dissolution rates can vary by factors

of 5 between different types of surfaces at different

potentials, and that the sites of the dissolved Pt on the

surface can vary widely (e.g., from pits vs. step edges).

Perhaps most compelling, Zhang et al. [73] have recently

shown that Au clusters on Pt nanoparticles could greatly

enhance Pt stability, which has been attributed to the

alteration of the oxidation characteristics of the Pt surface.

It is believed that Au atoms are deposited electrochemi-

cally on high-energy surface sites of Pt particles, which

might reduce dissolution of the least stable sites on Pt

nanoparticles. The success of Zhang et al.’s approach

demonstrates how enhancing stability of Pt nanoparticles

from the perspective of modifying the surface structure and

chemistry might be particularly fruitful.

Acknowledgments Some Pt/C and aged MEA samples used in this

study were obtained from GM Fuel Cell Activities. The authors thank

P. Strasser for providing the TKK Pt/C 28 wt% sample used in this

study, and H.A. Gasteiger, R. Makharia, S. Kocha, F. Wagner, D.

Myers, J.P. Meyers, R. Darling, and D. Rolison for stimulating dis-

cussion. This work is supported by the DOE Hydrogen Initiative

program under award number DE-FG02-05ER15728, and made use

of the Shared Experimental Facilities supported by the MRSEC

Program of the National Science Foundation under award number

DMR 02-13282. Y.S.H. acknowledges financial support from GM

Fuel Cell Activities and an Air Products Faculty Excellence grant and

D.M. gratefully acknowledges a 3M Nontenured Faculty Award.

References

1. Gasteiger HA, Kocha SS, Sompalli B, Wagner FT (2005) Appl

Catal B 56:9

2. Gasteiger H, Gu W, Makharia R, Mathias MF, Sompalli B (2003)

In: Vielstich W, Lamm A, Gasteiger H (eds) Handbook of fuel

cells—fundamentals, technology and applications, vol. 3. John

Wiley & Sons, Chichester, UK, pp 593

3. Matthias M, Gasteiger H, Makharia R, Kocha S, Fuller T, Pisco

(2004) J Prepr Pap-Am Chem Soc Div Fuel Chem 49:471

4. Antolini E (2003) J Mater Sci 38:2995

5. Mukerjee S, Srinivasan S (2003) In: Vielstich W, Lamm A,

Gasteiger H (eds) Handbook of fuel cells—fundamentals, tech-

nology and applications, vol 3. John Wiley & Sons, Chichester,

UK pp 503

6. Auer E, Freund A, Pietsch J, Tacke T (1998) Appl Catal A

173:259

7. Skriver HL, Rosengaard NM (1992) Phys Rev B 46:7157

8. Foiles SM, Baskes MI, Daw MS (1986) Phys Rev B 33:7983

9. Frenken JWM, Stoltze P (1999) Phys Rev Lett 82:3500

10. Sattler ML, Ross PN (1986) Ultramicroscopy 20:21

11. Ferreira PJ, Shao-Horn Y (2007) Electrochem Solid-State Lett

10:B60

12. Kinoshita K (1990) J Electrochem Soc 137:845

13. Antolini E, Giorgi L, Cardellini F, Passalacqua E (2001) J Solid

State Electrochem 5:131

Fig. 18 (a) Degradation of electrochemical surface area of Pt

nanoparticles of the three samples in Fig. 16 along with the number

of cycles. The loading of Pt on glassy carbon electrode (GCE) for the

three samples is 10 lg/cm2, by depositing 20 lL of 0.215 mg/mL of

Pt/C water suspension, followed by 10 lL of 0.03 wt% of Nafion

solution after catalysts dried on GCE. After slow drying of the

electrodes, they were cycled in 0.5 M H2SO4, between 600 and

1,000 mV (vs. RHE) for 1,200 scans at 80 �C. CV measurements

were taken each 200 cycles at room temperature at 50 mV/s. The

surface area was calculated by integrating the hydrogen–Pt desorption

area. (b): the histograms of the pristine sample before and after

voltage cycling. TEM images of Pt nanoparticles were taken on JEOL

2010F TEM, operating at 200 kV. The average diameter and size

distribution were measured from 200 Pt particles

Top Catal

123

14. Markovic NM, Radmilovic V, Ross PN (2003) In: Wieckowski

A, Savinova ER, Vayenas CG (eds) Physical and electrochemical

characterization of bimetallic nanoparticles electrocatalysts,

catalysis and electrocatalysis at nanoparticles surfaces. Marcel

Dekker, New York and Basel, pp 311

15. Ferreira PJ, la O’ GJ, Shao-Horn Y, Morgan D, Makharia R,

Kocha S, Gasteiger H (2005) J Electrochem Soc 152:A2256

16. Shao-Horn Y, Ferreira P, la O GJ, Morgan D, Gasteiger HA,

Makharia R (2006) ECS Trans 1:185

17. Merzougui B, Swathirajan S (2006) J Electrochem Soc

153:A2220

18. Landsman DA, Luczak FJ (2003) In: Vielstich W, Lamm A,

Gasteiger H (eds) Handbook of fuel cells—fundamentals, tech-

nology and applications, vol 4. John Wiley & Sons, Chichester,

UK, pp 811

19. Aragane J, Murahashi T, Odaka T (1988) J Electrochem Soc

135:844

20. Honji A, Mori T, Tamura K, Hishinuma Y (1988) J Electrochem

Soc 135:355

21. Blurton KF, Kunz HR, Rutt DR (1978) Electrochim Acta 23:183

22. Fowler MW, Mann RF, Amphlett JC, Peppley BA, Roberge PR

(2002) J Power Sources 106:274

23. Knights SD, Colbow KM, St-Pierre J, Wilkinson DP (2004) J

Power Sources 127:127

24. Wilson MS, Garzon FH, Sickafus KE, Gottesfeld S (1993) J

Electrochem Soc 140:2872

25. Tada T, Kikinzoku T (2003) In: Vielstich W, Gasteiger H, Lamm

A (eds) Handbook of fuel cells—fundamentals, technology and

applications, vol 3. John Wiley and Sons, p 481

26. Xie J, Wood DL, Wayne DM, Zawodzinski TA, Atanassov P,

Borup RL (2005) J Electrochem Soc 152:A104

27. Xie J, Wood DL, More KL, Atanassov P, Borup RL (2005) J

Electrochem Soc 152:A1011

28. Akita T, Taniguchi A, Maekawa J, Sirorna Z, Tanaka K, Kohy-

ama M, Yasuda K (2006) J Power Sources 159:461

29. Bett JAS, Kinoshita K, Stonehart P (1976) J Catal 41:124

30. Gruver GA, Pascoe RF, Kunz HR (1980) J Electrochem Soc

127:1219

31. Ruckenstein E, Pulvermacher B (1973) J Catal 29:224

32. Granqvist CC, Buhrman RA (1976) J Catal 42:477

33. Tseung ACC, Vassie PR (1976) Electrochimica Acta 21:315

34. Lifshitz IM, Slyozov VV (1961) J Phys Chem Solids 19:35

35. Wagner C (1961) Z Elektrochem 65:581

36. Voorhees PW (1985) J Stat Phys 38:231

37. Ross PN Jr (1987) In: Petersen EE, Bell AT (eds) Catalyst

deactivation. Marcel Dekker, New York, p 165

38. Aragane J, Urushibata H, Murahashi T (1996) J Appl Electro-

chem 26:147

39. Yasuda K, Taniguchi A, Akita T, Ioroi T, Siroma Z (2006) J

Electrochem Soc 153:A1599

40. Kinoshita K, Lundquis JT, Stonehar P (1973) J Electroanal Chem

48:157

41. Patternson T (2002) In: Igwe GJ, Mah D (eds) Fuel cell tech-

nology topical conference proceedings, 2002 AIChE Spring

National Meeting (March 10–14), New York, p 313

42. Darling RM, Meyers JP (2003) J Electrochem Soc 150:A1523

43. Darling RM, Meyers JP (2005) J Electrochem Soc 152:A242

44. Yasuda K, Taniguchi A, Akita T, Ioroi T, Siroma Z (2006) Phys

Chem Chem Phys 8:746

45. Pourbaix M (1966) Atlas of electrochemical equilibria in aqueous

solutions. Pergamon Press, Oxford

46. Bindra P, Clouser SJ, Yeager E (1979) J Electrochem Soc

126:1631

47. Wang XP, Kumar R, Myers DJ (2006) Electrochem Solid-State

Lett 9:A225

48. Ota K-i, Koizumi Y, Mitsushima S, Kamiya N (2006) ECS Trans

3:619

49. Bett JA, Kinoshita K, Stonehart P (1974) J Catal 35:307

50. Makharia R, Kocha SS, Yu PT, Sweikart MA, Gu W, Wagner FT,

Gasteiger HA (2006) ECS Trans 1:3

51. Stevens DA, Hicks MT, Haugen GM, Dahn JR (2005) J Elect-

rochem Soc 152:A2309

52. Wynblatt P, Gjostein NA (1975) In: McCaldin JO, Somorjai G

(eds) Progress in solid state chemistry, vol 9. Pergamon, New

York pp 21

53. Ehrburger P, Lahaye J (1984) Abstr Paper Am Chem Soc 187:28

54. Guilminot E, Corcella A, Charlot F, Maillard F, Chatenet M

(2007) J Electrochem Soc 154:B96

55. Chen J, Chan KY (2005) Mol Simul 31:527

56. Phillips WB, Desloge EA, Skofronick JG (1968) J Appl Phys

39:3210

57. Jensen P, Blase X (2004) Phys Rev B 70:165402

58. Wang XP, Kumar R, Kariuki N, Myers DJ (2007) J Electrochem

Soc (unpublished)

59. Anderson GM, Crerar DA (1993) Thermodynamics in geo-

chemistry. Oxford University Press, New York

60. Komanicky V, Chang KC, Menzel A, Markovic NM, You H,

Wang X, Myers D (2006) J Electrochem Soc 153:B446

61. Jerkiewicz G, Vatankhah G, Lessard J, Soriaga MP, Park YS

(2004) Electrochimica Acta 49:1451

62. Alsabet M, Grden M, Jerkiewicz G (2006) J Electroanal Chem

589:120

63. Ota KI, Nishigori S, Kamiya N (1988) J Electroanal Chem

257:205

64. Rand DAJ, Woods R (1972) J Electroanal Chem 35:209

65. Rose TL, Robblee LS (1990) IEEE Trans Biomed Eng 37:1118

66. Woods R (1976) In: Bard AJ (eds) Electroanalytical chemistry,

vol 9. Marcel Dekker, New York pp 1

67. Angerste H, Conway BE, Sharp WBA (1973) J Electroanal Chem

43:9

68. Allen GC, Tucker PM, Capon A, Parsons R (1974) J Electroanal

Chem 50:335

69. Hammond JS, Winograd N (1977) J Electroanal Chem 78:55

70. Nagy Z, You H (2002) Electrochim Acta 47:3037

71. Lupis CHP (1983) Chemical thermodynamics of materials. North

Holland, New York

72. Holby E, Morgan D, Shao-Horn Y (2007) (unpublished)

73. Zhang J, Sasaki K, Sutter E, Adzic RR (2007) Science 315:220

Top Catal

123