Journal of Electroanalytical Chemistry 535 (2002) 49
/55

www.elsevier.com/locate/jelechem

Array membrane electrode assemblies for high throughput screening

of direct methanol fuel cell anode catalysts
Renxuan Liu a, Eugene S. Smotkin a,b,*
a
NuVant Systems, 10 West 33rd Street, Suite 127, Chicago, IL 60616, USA
b
Department of Chemistry, University of Puerto Rico at Rio Piedras, San Juan, PR 00931, USA

Received 6 February 2002; received in revised form 15 July 2002; accepted 13 August 2002

Abstract

A fuel cell using an array membrane electrode assembly has been developed for the high throughput screening of fuel cell
electrocatalysts. Standard membrane and electrode assembly methods are used. The use of modified fuel cell hardware permits the
use of realistic catalyst exposure histories and steady state reaction conditions. The array fuel cell requires no supplemental
electrolytes. The performance of the array fuel cell is demonstrated by the testing of one prepared in-house and three commercially
available fuel cell catalysts. Within the potential range of a DMFC anode (i.e. 0.3
/0.4 V), the catalyst rankings were PtRu (Johnson
Matthey) /PtRu oxide (E-Tek) /PtRu (reduced by NaBH4) /Pt.
# 2002 Elsevier Science B.V. All rights reserved.

Keywords: Array fuel cell; Membrane electrode assemblies

1. Introduction lead to artifacts in the discovery of new catalysts. Array

Direct methanol fuel cells (DMFCs) convert the
chemical energy of methanol to electrical power by
direct oxidation of methanol at a fuel cell anode. The
poor anode and cathode kinetics of methanol/air
electrochemistry precludes commercialization of
DMFCs. Over 30 years ago PtRu was found to be the
best, yet inadequate, catalyst for DMFCs [1
/5]. The
anodic reaction of adsorbed CO with H2O is rate
limiting at DMFC anodes. Thus catalysts that activate
H2O at more negative potentials are needed. Other
issues include catalyst ripening and degradation of
catalyst j electrolyte and catalyst j support (typically
carbon) interfaces. The search for better catalysts has
motivated the development of high throughput screen-
ing methods [6
/10].
Reports of combinatorial methods applied to electro-
catalysis typically include catalyst array preparative
methods tailored and linked to a novel high throughput
screening method. The tight connection between the
array preparative method and the screening method can
* Corresponding author
E-mail address: esmotkin@goliath.cnnet.clu.edu (E.S. Smotkin).
0022-0728/02/$ - see front matter # 2002 Elsevier Science B.V. All rights reserved.
PII: S 0 0 2 2 - 0 7 2 8 ( 0 2 ) 0 1 1 4 4 - 0
preparative methods include electrodeposition [7,8],
sputter deposition [11,12], and deposition of metal ion
precursors followed by borohydride reduction [9]. These
preparative methods are not methods typically used for
fuel cell catalyst preparation, although bulk catalysts
have been augmented by sputter deposition of Pt on the
catalytic surface [13]. The need for high metal dispersion
in mixed metal DMFC catalysts limits preparation of
bulk mixed metal catalysts to low temperatures. Thus
phase segregation is prevalent and the number of
phases, how the component metals distribute amongst
those phases and the role of each phase in catalysis is
directly linked to the preparative method. Typically,
when catalyst arrays are prepared using metal salt
precursors, a single reduction method is used across
the array [9]. The use of a single reduction method
across a library of elements with extreme nominal
composition variations, combined with un-optimized
well-less chemistry can result in varying degrees of
incorporation of the metal ion components across the
array. Thus no a priori knowledge of the relationship
between nominal and actual compositions is possible.
Examples exist where the ‘best catalyst’ identified in an
array was an artifact of the preparative method. A
50 R. Liu, E.S. Smotkin / Journal of Electroanalytical Chemistry 535 (2002) 49
/55
Fig. 1. Fuel cell schematic (not to scale).
negative result in compositional phase space can be
considered negative only for the preparative method
used. For example, PtRu has been ranked as inferior to
PtRuOs [14] and PtRuOsIr [9] when both are prepared
by borohydride reduction. However, when PtRu is
prepared by a sulfite complex route [4], it proved
superior to borohydride prepared PtRuOs and PtRuO-
sIr. However, the sulfite complex route has not been
successfully applied to a ternary or quaternary catalyst.
Catalyst candidates must be compared (based on
geometric electrode area) to state-of-art catalysts in
real DMFCs because performance is highly dependent
on the membrane and electrode fabrication method and
the electrochemical conditioning process used to attain
steady state DMFC performance. Screening device
designs require maximal consideration and adaptation
of the actual DMFC reactor features. Our strategy is to
disconnect the array preparative method from the
screening method and focus on realistic and reliable
array testing. This enables comparison of bulk catalysts,
prepared by any method, to state-of-art commercially
available catalysts. The following single cell DMFC
description provides the design basis for our screening
device.
The DMFC electrode/electrolyte system, an iono-
meric membrane (NafionTM 117) sandwiched between
the anode and cathode catalyst layers, is a three-layer
membrane and electrode assembly (MEA). A fuel cell
assembly housing the MEA (Fig. 1) includes a pair of
grooved graphite blocks (flow fields) to distribute
electrode reactants and products. The flow fields contact
highly porous electronically conductive carbon fabric
gas diffusion layers (GDLs). The opposite faces of the
GDLs contact the catalyst layers and collect current
from those layers while minimally impeding mass
transport of reactants and products between the elec-
trode surfaces and the flow fields. MEAs are fabricated
by application of catalyst inks to: (1) the membrane; or
(2) the GDLs (five-layer GDL integrated MEA).
We now demonstrate a high-throughput screening
device for fuel cell anode catalysts that incorporates
state-of-art MEA fabrication methods for catalyst
arrays and requires no moving parts (reference electro-
des (REs), etc).1 Steady state polarization curves for all
array spots are obtained simultaneously. Cathode per-
formance variances are avoided by use of a common
cathode of area fivefold greater than the sum of the
array anodes. The use of modified fuel cell hardware
permits genuine fuel cell conditioning of all array spots,
a prerequisite to steady state performance. Conditioning
involves fuel cell operation (1
/3 days) at a selected cell
voltage or current. Recent in-situ XANES studies
confirm that conditioning does not involve chemical
changes of the bulk catalyst [15] and is more likely
related to rewetting of the catalyst surfaces with the
ionomeric electrolyte [16]. Delamination of the nano-
sized catalyst surfaces from the ionomer probably
occurs during the heat treatment2 required for MEA
fabrication. Initial catalyst screening requires at least 3
days of steady state data acquisition: the single cell
testing of one electrode takes 4
/6 days of conditioning/
1
Patent pending.
2
Dehydration of Nafion† during heating can result in up to 15%
volumetric reduction, which may cause delamination from the catalyst
surface.
 R. Liu, E.S. Smotkin / Journal of Electroanalytical Chemistry 535 (2002) 49
/55 51

data acquisition. Establishment of confidence intervals

requires the repetitive testing of four to six samples of
the catalyst: the reliable testing of a single composition
requires at least 25 days of continuous test stand
operation. The comparison of five catalysts reliably
would take 125 days of testing not including the time
required to prepare the electrodes. Finally, the success of
a comparative study using a series of single cell test
stands relies on extremely reproducible behavior of the
cathode counter electrodes (CEs) (25 in the above
example), a daunting demand.
The screening device MEA has an array of catalytic
spots at the anode side and a conventional cathode Fig. 2. Sensor electrode.
electrode. The MEA cathode catalytic area contacts a
GDL of matching geometric area while the opposite side
of the polymer membrane faces the 25 separate cata-
lyzed GDL disks. The GDL disks fill the voids of and
are coplanar with a hole-punched Teflon gasket. The
uncatalyzed faces of the disk GDLs contact the sensor
electrode flow fields. Humidified H2 is delivered to the
cathode electrode, which serves as dynamic hydrogen
RE. Fuel (methanol, reformate, etc.) is delivered to the
anode array. Polarization curves are obtained by scan-
ning the potential applied between the CE/RE and the
entire array electrode simultaneously. At any instance,
the array electrodes are equipotential versus the com-
mon CE/RE, although electronically insulated from
each other. The steady state currents from each and
every array electrode are fed to unique current followers,
sensed by the data acquisition card and recorded
digitally. Thus 25 catalyst polarization curves are
obtained simultaneously, yet independently of each
other. The control experiment data show that the Fig. 3. Ceramic sensor electrode array flow fields.
stoichiometric ratio of the fuel in the flow fields is
sufficiently high that no significant depletion of fuel and have a thermal expansion coefficient similar to that
occurs along the serial path over the array electrodes. of graphite. The ceramic material used for the array
Arbitrary location of five spots for each catalyst block meets both criteria. The grooves on the array flow
candidate within the array further reduces the possibility field block connect the holes that accommodate press
of systematic errors. We demonstrate the use of the fitting of 25 sensors in series.3 The graphite sensors are
device to rank four different catalysts. These catalysts glued into the ceramic array block with the sensor flow
are: (1) PtRu (50:50) by Johnson Matthey, stock#39723, fields oriented with the array block flow field.
lot#E31G29; (2) PtRu made by the NaBH4 method [17]; Fig. 4 shows the graphite common CE/RE flow field
(3) PtRu oxide powder by E-TEK, lot#2883PR04; (4) Pt block. The flow field grooves guide the reactant gas/
by Johnson Matthey, stock#12755, lot#J27J16. liquid over the uncatalyzed GDL side. The opposite side
of the GDL is catalyzed and contacts the Nafion† . Fig.
5 shows the exploded schematic of the high throughput
2. Experimental array fuel cell. The five layer MEA anode disks are
incorporated into the array punched Teflon gasket. The
2.1. Fuel cell assembly
array block is placed against the anode array side of the
MEA with each graphite sensor electrode contacting the
Fig. 2 is a single graphite sensor electrode. A flow
GDL of a catalyst candidate spot. The fuel cell assembly
field for contact to the GDL is at one end, and a
threaded hole for the screw lead of a pin jack (to op-amp is bolted between two end plates.
current follower) is at the other. The sensor flow field
meshes with the array block flow field (Fig. 3). The
3
array block material must be electronically insulating Flow fields accommodating 256 sensors are in preparation.
52 R. Liu, E.S. Smotkin / Journal of Electroanalytical Chemistry 535 (2002) 49
/55
Fig. 4. Graphite common electrode flow field.
2.2. Instrumentation
A voltage follower (Fig. 6) controls the potential
difference between CE/RE and the array working
electrodes (WEs) [18]. The buffer enables currents up
to 10 A. A standard operational amplifier circuit for
current-to-voltage conversion (current follower) inter-
faces the sensor electrodes to the data acquisition card
(Fig. 7) [19]. The buffer enables up to 400 mA per
sensor. In this study there are 25 current follower
modules. A computer controlled National Instruments
card (PCI 6031E I/O card) is used to program the
potential input to the voltage follower (Fig. 6) and
acquire the signal from all current followers (Fig. 7). A
LabVIEW program controls the potential, collects and
digitally stores current and potential data.
Fig. 5. Schematic of exploded high throughput device.
2.3. MEA preparation
Anode and cathode catalyst inks are prepared by
dispersing the catalysts in a solubilized Nafion† solu-
tion (Aldrich) as described by Wilson and Gottesfeld
[20]. The solubilized Nafion† encases the catalyst
particles and facilitates contact of the catalyst particles
with the 7-mil (1 mil /25.4 mm) thick membrane
electrolyte. Teflon dispersion is added to the cathode
ink to facilitate water rejection during fuel cell opera-
tion. After extensive stirring the ink is applied to one
side of the GDL and baked. The GDL is then pressed
against the polymer membrane during fuel cell assem-
bly. Alternatively, the catalysts can be decal transferred
to the Nafion† layer [20].
The GDL carbon fiber paper (Toray paper, E-TEK,
TGPH-060) used in our single cell DMFCs is also used
for the array GDL system. Toray paper is cut to size (10
cm2) for the CE/RE. Pt black (Johnson Matthey,
stock#12755, lot#J27J16) ink is applied to the GDL.
In this study, the array side of the MEA is prepared with
six duplicate samples for each of four catalyst samples,
and one blank sample. Six catalyzed GDLs, for each
catalyst candidate, are prepared by evenly coating a
square of GDL material and then stamping out 0.95 cm
diameter disks.
The large CE/RE was hot pressed onto Nafion† 117
(160 8C at 1800 lb-force (1 lb-force /4.44822 N) for 5
min). The 25 WEs were hot pressed onto the other side
of the Nafion† sheet (160 8C at 1000 lb-force for 5
min). This is a five layer GDL integrated MEA. Prior to
the testing of catalyst candidates, a control experiment
with 25 PtRu electrodes was conducted. Then four
catalysts, each catalyst occupying six positions distrib-
uted randomly on the array and one dead blank spot,
were tested.
 R. Liu, E.S. Smotkin / Journal of Electroanalytical Chemistry 535 (2002) 49
/55 53

Fig. 6. Voltage follower schematic for potential control of the array electrode versus the CE.

2.4. Testing procedure

The fuel cell was conditioned by passing wet hydrogen

through both sides of the assembled fuel cell at 60 8C
over night. A linear scan between 0 and 0.2 V was
applied for several times. The anode stream was then
switched to methanol (0.5 M, 10 ml min1) while the
humidified hydrogen flow to the CE/RE side was
maintained. Linear scans between 0 and 0.7 V were
applied repetitively until steady state performance was
attained, indicating completion of the conditioning
process. The initial array electrode potentials were
stepped from 0 to 0.8 V in 10 mV increments at 5 s
intervals and then reset to 0 V at the completion of the
scan. The fuel cell was thermostated at 60 8C. Fig. 8. Raw I
/V curves for control experiment PtRu array WE.

lot#C13K06; 1.62 mg cm 2). At potentials negative of

3. Results and discussion 0.4 V, all 25 curves overlap each other. At potentials
above 0.4 V the curves are obviously mass transfer
Fig. 8 shows the I
/V curves of the control experiment affected and limiting currents are achieved at /0.8 V.
array electrode (PtRu Johnson Matthey, stock#41171, The limiting currents are used to obtain the kinetic

Fig. 7. Current follower circuit. Input from the back of a sensor electrode.
54 R. Liu, E.S. Smotkin / Journal of Electroanalytical Chemistry 535 (2002) 49
/55

currents by mass transfer correcting the raw data

according to the following equation [21].

1 1 1
 
I Ikinetic Ilimiting

Fig. 9 shows mass transfer corrected the I
/V plots.

The curves overlap as expected since all electrodes are of
the same material and the same loading. This control
experiment confirms that depletion of the anode flow
field reactant stream is negligible.
Fig. 10(a) shows the I
/V curves for the array
electrode containing PtRu (Johnson Matthey,
stock#39723, lot#E31G29), PtRu (reduced by
NaBH4), PtRu oxide (E-TEK, lot#2883PR04) and Pt
(Johnson Matthey, stock#12755, lot#J27J16). All cata-
lyst spots are randomly positioned and have a repetition
of six. All spot data of the same composition are
averaged and standard deviation error bars are included
as shown in Fig. 10(b). Potentials above 0.45 V versus
DHE are outside the useful range of DMFCs. We do
not mass transport correct the data of Fig. 10 because
the control experiments (Figs. 8 and 9) indicate that
mass transport corrections are not needed within the
fuel cell regime (i.e. less than 0.5 V). The spots were
randomly distributed throughout the array and the
polarization curves were averaged.
Within the potential range of a DMFC anode (i.e.
0.3
/0.4 V), the catalyst ranking is PtRu (Johnson
Matthey) /PtRu oxide (E-Tek) /PtRu (reduced by Fig. 10. (a) I
/V curves for array electrode containing randomly
NaBH4)/Pt. The blank (essentially a carbon electrode) positioned PtRu (Johnson Mathey), PtRu (reduced by NaBH4), PtRu
oxide (E-TEK) and Pt (Johnson Matthey). (b) The ranking of four
was flat with no current. The ranking of the four
catalysts in DMFC anode operating region.
catalysts was easily accomplished within 1 week. In
comparison, we have completed similar comparisons
extensive studies are being conducted. More recently,
with a conventional single cell fuel cell testing device
we have ranked in order Johnson Matthey PtRu /In
within a period of 6 months. The rankings of the
house prepared PtRu Watanabe catalyst /Borohydride
Johnson Matthey PtRu, PtRu (reduced by borohydride)
prepared PtRu /Alfa Aesar Pt. These rankings were
and Pt have been extensively corroborated with full fuel
consistent with full fuel cell performance curves.
cell performance curves [22,23]. The E-Tek catalyst was
New DMFC membrane electrolytes are needed with
purchased for the purpose of this study and more
lower methanol permeability. New membrane candi-
dates will require new MEA fabrication methods. This
will typically involve making a dispersion of the catalyst
in a liquid blend consisting of ordinary organic solvents
(e.g. alcohol, glycerol, etc.) and the solubilized un-cross
linked polymer [24]. This dispersion is known as catalyst
‘ink’. Optimizing the preparative method for catalyst
inks is a complex, laborious process [25]. The para-
meters requiring optimization include:

1) Selection of ink components.

2) Quantity and order of addition of components.
3) Type and length of stirring.
4) Catalyst deactivation method (if needed).

Fig. 9. Mass transport corrected I
/V curves for control experiment The device of Fig. 5 can simultaneously screen 25
array. different ink preparative methods. An MEA with a
 R. Liu, E.S. Smotkin / Journal of Electroanalytical Chemistry 535 (2002) 49
/55
common cathode or a common anode can be used to
screen anode or cathode catalyst ink formulations,
respectively. In this comparative study of commercially
available catalysts against borohydride reduction pre-
pared PtRu, the above parameters were optimized for
PtRu in a conventional fuel cell test stand. The
optimized parameters were used and the bulk catalyst
samples were randomly distributed on the array elec-
trode.
4. Summary and conclusion
A high throughput screening device for fuel cell
electrocatalysis has been developed that incorporates
actual fuel cell design features. No artifacts due to the
use of supplemental electrolytes with mobile anions are
possible since only Nafion† is used as the electrolyte.
State-of-art membrane electrode assembly fabrication
methods are used, permitting catalyst testing with
realistic reactant exposure histories (conditioning) and
steady state reaction conditions. The device uses no
moving parts and state-of-art flow field delivery of
reactant streams to the porous electrode surfaces. The
device of this study accommodates an array of 25 spots
permitting repetitive testing of catalysts. Although
control experiments verify even performance across the
array, multiple samples are randomly distributed across
the array to minimize possible systematic errors.
Acknowledgements
This work is funded by NuVant Systems and the
Army Research Laboratory, Collaborative Technology
Alliance in Power and Energy.
References
[1] D.F.A. Koch, Australian Patent, 46, 1964, p. 123.
[2] B.D. McNicol, D.A.J. Rand, K.R. Williams, J. Power Sources 83
(1999) 15.
[3] J.O’M. Bockris, H. Wroblowa, J. Electroanal. Chem. 7 (1964)
428.
55
[4] M. Watanabe, M. Vehida, S. Motoo, J. Electroanal. Chem. 229
(1987) 395.
[5] M.P. Hogarth, G.A. Hards, Platinum Metals Rev. 40 (1996) 150.
[6] M.G. Sullivan, H. Utomo, P.J. Fagan, M.D. Ward, Anal. Chem.
71 (1999) 4369.
[7] (a) A. Gorer, World Patent Application WO 00/54346, Septem-
ber, 2000;
(b) P. Strasser, S. Gorer, M. Devenney, in: O. Yamamoto (Ed.),
International Symposium on Fuel Cells for Vehicles, Electro-
chemical Society of Japan, Nagoya, November 2000, p. 153.
[8] C.J. Warren, R.C. Haushalter, L. Matsiev, US Patent 6,187,164.
[9] E. Reddington, A. Sapienza, B. Gurau, R. Viswanathan, S.
Sarangapani, E.S. Smotkin, T.E. Mallouk, Science 280 (1998)
1735.
[10] J. Pope, D. Buttry, World Patent Application WO 2000037718,
June 2000.
[11] P. Cong, R.D. Doolen, Q. Fan, D.M. Giaquinta, S. Guan, E.W.
McFarland, D.M. Poojary, K. Self, H.W. Turner, W.H. Wein-
berg, Angew. Chem. Int. Ed. Engl. 38 (1999) 484.
[12] P. Strasser, S. Gorer, M. Devenney, in: S. Narayanan, S.
Gottesfeld (Eds.), Proc.-Electrochem. Soc. (Direct Methanol
Fuel Cells, PV2001-4), p. 191.
[13] C.K. Witham, W. Chun, T.I. Valdez, S.R. Narayanan, Electro-
chem. Solid-State Lett. 3 (2000) 497.
[14] K.L. Ley, R. Liu, C. Pu, Q. Fan, N. Leyarovska, C. Segre, E.S.
Smotkin, J. Electrochem. Soc. 144 (1997) 1543.
[15] R. Viswanathan, R. Liu, G. Hou, E.S. Smotkin, S.R. Bare, F.
Modica, G. Mickelson, C.U. Segre, N. Leyarovska, J. Phys
Chem. Sect. B 106 (2002) 3458.
[16] Personal communication with David Wilkinson of Ballard Power
Systems.
[17] K.L. Ley, R. Liu, C. Pu, Q. Fan, N. Leyarovska, C. Segre, E.
Smotkin, J. Electrochem. Soc. 144 (1997) 1543.
[18] A.J. Bard, L.R. Faulkner, Electrochemical Methods: Fundamen-
tals and Applications, 2nd ed., John Wiley & Sons, Inc, New
York, 2001, p. 643.
[19] P. Horowitz, W. Hill, The Art of Electronics, 2nd ed., Cambridge
University Press, Cambridge, 1989, p. 184.
[20] M.S. Wilson, S. Gottesfeld, J. Appl. Electrochem. 22 (1992) 1.
[21] E. Gileadi, Electrode Kinetics for Chemists, Chemical Engineers,
and Materials Scientists, VCH Publishers, Inc, Weinheim, 1993,
p. 5.
[22] B. Gurau, Direct Methanol Fuel Cells: Catalysis and Engineering
Aspects, Ph.D. Thesis, Illinois Institute of Technology, Chicago,
Illinois, May 2002.
[23] H. Lei, S. Suh, B. Gurau, B. Workie, R. Liu, E.S. Smotkin,
Electrochim. Acta 47 (2002) 2913.
[24] E. Antolini, L. Giorgi, A. Pozio, E. Passalacqua, J. Power Sources
77 (1999) 136.
[25] M. Hogarth, P. Christensen, A. Hamnett, A. Shukla, J. Power
Sources 69 (1997) 113.