ww.sciencedirect.com

i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n en e r g y 4 0 ( 2 0 1 5 ) 1 1 6 5 9e1 1 6 6 7

Available online at w

ScienceDirect

journal homepage: www.elsevier .com/locate/he

The effects of channel depth on the performanceof miniature proton exchange membrane fuelcells with serpentine-type flow fields

Dyi-Huey Chang a,*, Shin-Yi Wu b

a Department of Environmental Engineering and Management, Chaoyang University of Technology, Taiwanb Department of Mechanical Engineering, National Chin-Yi University of Technology, Taiwan

a r t i c l e i n f o

Article history:

Received 12 February 2015

Received in revised form

23 April 2015

Accepted 27 April 2015

Available online 4 June 2015

Keywords:

Channel depth

Fuel cell

Metallic bipolar plate

Microelectrical discharge machining

(micro-EDM)

* Corresponding author. Tel.: þ886 42 233230E-mail address: changdh@cyut.edu (D.-H

http://dx.doi.org/10.1016/j.ijhydene.2015.04.10360-3199/Copyright © 2015, Hydrogen Ener

a b s t r a c t

Flow channels are one of the key components of a fuel cell, because they perform various

essential functions that enable the system to operate correctly. In this study, three-path

serpentine and parallel-serpentine flow fields with various depths were analyzed experi-

mentally. Die-sinking microelectrical discharge machining was applied to fabricate mini-

ature SUS316L bipolar plates, of which both the rib and channel widths were 500 mm and

the channel depths varied among 200, 300, 400, and 600 mm in an active area of

20 mm � 20 mm. The clamping test was performed to examine the magnitude of mem-

brane electrode assembly deformation and the contact resistance. The pressure drops for

each cell were analyzed to determine the effects of channel depth. The results revealed

that a deep channel is required to leave sufficient space for reactant transportation and

water removal; however, too low flow velocity reduces the convective mass transport and

cell performance when the channel is too deep.

Copyright © 2015, Hydrogen Energy Publications, LLC. Published by Elsevier Ltd. All rights

reserved.

Introduction

Fuel cells are a promising power technology because of their

compactness, silent operation, and high power density, and

because they emit nearly no pollution. Among various types of

fuel cells, proton exchange membrane fuel cells (PEMFCs) are

the leading technology for portable electronic devices and

light-duty vehicles, because they operate at relatively low

temperatures (normally below 100 �C) and produce electrical

output to meet dynamic power requirements. A substantial

00x4530; fax: þ886 42 237. Chang).53gy Publications, LLC. Publ

knowledge base that focuses on the development of high-

efficiency PEMFCs exists [1e13]. It is generally agreed that

the geometric parameters and flow patterns of flow channels

strongly influence the performance of PEMFCs.

Bipolar plates with various flow fields can be fabricated

using various methods, namely machining, etching, and

embossing. For economic reasons, embossing is the preferred

process for mass production. However, embossing exhibits

limitations in forming channels with high depth-to-width

ratios, especially for narrow channels. In addition, plate

42365.

ished by Elsevier Ltd. All rights reserved.

i n t e rn a t i o n a l j o u r n a l o f h y d r o g e n en e r g y 4 0 ( 2 0 1 5 ) 1 1 6 5 9e1 1 6 6 711660

fractures or variances in channel dimensions might occur.

Research has been conducted on enhancing the embossing

process to form deeper narrow channels [14e16]. Therefore,

investigating the impact of channel depth on the performance

of miniature PEMFCs is worthwhile.

Various simulations have been performed to explore this

topic based on computational fluid dynamics. The results of

simulating a 200-cm2 cell indicated that that fabricating a

channel that is too deep tends to reduce the local gas pressure,

leading to variances in fuel distribution [10,17]. Modeling a

130-cm2 cell indicated that the optimal channel-to-rib width

ratio is approximately one, and a channel that is too deep

provides redundant fuel [18]. Simulating a 24.6-cm2 cell

revealed that the performance of a PEMFC with a serpentine

flow channel is insensitive to channel depth in certain oper-

ation conditions [3]. In the simulation of a 2.56-cm2 miniature

cell, the results indicated that cells with a high aspect ratio

yield higher current densities [19].

These simulation results reveal a trend that deeper chan-

nels are favored for miniature cells, whereas shallower

channels are preferred for larger cells. Several studies have

experimentally examined the impacts of channel dimension

on PEMFCs of various sizes [20e32]. Few of these studies have

focused on the impacts of channel depth on miniature

PEMFCs. A miniature PEMFC normally has narrow channel

and rib widths to provide sufficient channel length in a small

active area. Fabricating a PEMFC with fine, deep channels that

produces an efficient power density is challenging. For a

PEMFC with low power densities, the impact of channel di-

mensionsmight be less substantial because efficiency inmass

transportation is lower. This study examined the effects of

channel depth on the performance of highly efficient minia-

ture PEMFCs. Die-sinkingmicroelectrical dischargemachining

(micro-EDM) was applied to prepare micro flow channels in

metallic bipolar plates. Two serpentine-type flow fields were

created; both the rib and channel widths were 500 mm and the

channel depths varied among 200, 300, 400, and 600 mm in an

active area of 20 mm � 20 mm. The clamping test was per-

formed and pressure drops were measured to examine the

effects of channel depth.

Fig. 1 e The representation of (a) SFF and (b) PSFF and the

finished unipolar plate for (c) SFF and (d) PSFF.

Experimental and procedures

Fabricating the bipolar plates

In this study, SUS316L stainless steel was used as the plate

material because it features excellent heat and corrosion

resistance. Micro-EDM is applied to fabricate unipolar plates.

Micro-EDM can be classified into various categories based on

the tool electrodes. The process applied in this study involves

using cubic electrodes and is called the die-sinking micro-

EDM. During the process, copper tool electrodes were manu-

factured using micro-high-speed milling. A 1-mm-thick

SUS316L plate was cut into a 50 mm � 50 mm area by using a

wire machine and then served as the workpiece electrode. All

channels were processed on an area basis where electricity

was discharged in a single direction.

The serpentine-type flow field was applied as the flow

design. Previous studies [1,33] have reported that single-pass

serpentine flow fields tend to cause polarization problems

attributable to excessive channel length, and executing too

many passes tends to produce an uneven fuel distribution.

Therefore, the three-pass design was adopted. Fig. 1 shows

two types of commonly used serpentine flow field; Fig. 1(a)

illustrates the three-path serpentine flow field (SFF), and

Fig. 1(b) illustrates the three-path parallel-serpentine flow

field (PSFF). Fig. 1(c) and (d) show the finished unipolar plates

for SFF and PSFF, respectively. The plates were finished using

5 A of peak discharging current at a material removal rate of

7 mm3 min�1. The most obvious difference between SFF and

PSFF is that for PSFF there are single paths collecting and

distributing both gas andwater. The PSFF has been considered

a serpentine-type flow field in various studies [9,34]. For the

test samples, the channel and rib width are constant as

500 mm in a reaction area of 4 cm2. The rib-to-channel width

ratio was set as 100% because a previous study [18] has re-

ported that such a ratio enables high power densities to be

achieved. Channel depths are varied as 200, 300, 400 and

600 mm to test their impacts.

The clamping test

When assembling PEMFCs, a strong clamping force is usually

required to prevent the leakage of reactants and to reduce the

contact resistance. However, the clamping force causes the

membrane electrode assembly (MEA) to protrude into chan-

nels, leaving insufficient space for reactant flow and water

removal. This study explored the deformation of the MEA by

using the MEA compression test. During the test, unipolar

plates with parallel channels 200 mm in depth were cut in half

by using a wire machine and assembled to become a “half-

cell,” as shown in Fig. 2.The channel and rib widths are

500 mm, consistent to the bipolar plates in Fig. 1. Clamping

Fig. 2 e (a) The half bipolar plates, (b) half gasket, (c) testing

MEA, and (d) the clamping test.

Table 1 e General values used during simulations.

Description Value Units

Active area 400 mm2

Membrane thickness 0.1 mm

GDL thickness 0.4 mm

Catalyst layer thickness 0.01 mm

Cell temperature 333 K

Inlet flow rate 60 cm3/min

Active specific area 16000 1/m

Fig. 3 e The model geometry for PSFF with the depth of

600 mm with the cut plane for further observation.

i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n en e r g y 4 0 ( 2 0 1 5 ) 1 1 6 5 9e1 1 6 6 7 11661

torques ranging from 10 to 25 kgf cm were applied to examine

the magnitude of MEA deformation. The length of the MEA

protruding into channels (lp) is approximated as follows:

lp ¼ lch � lrib2

(1)

where lch and lrib are the thicknesses of the MEA in the chan-

nels and ribs, respectively.

In addition, to determine the optimal clamping force, the

contact resistances (rc) between unipolar plates and gas

diffusion layers (GDLs) are calculated as follows:

rc ¼ VJ

(2)

where J is the applied current density and V is the measured

electric potential in the test.

The single-cell test

In this study, the cell performance of metallic bipolar plates

with various channel depths was tested using the TEI-P300-

1AB2CS surveying instrument. In addition, the pressure

drops for the testing cells were measured to examine the ef-

fects of channel depth. The pressure drop is defined as the

difference between input and output gas pressure and was

obtained using a data acquisition (DAQ) system. During the

tests, a single cell 50 � 50 � 23 mm in dimensions was

assembled. The cell was fabricated with end plates, gaskets,

bipolar plates, and an MEA. The MEA comprised a proton ex-

change membrane and two GDLs with 0.25 mg cm�2 Pt for

both the anode and cathode. Pure hydrogen and oxygen were

supplied to the anode and cathode electrodes. The anode flow

rate was 60 cc min�1, and thermal gas humidification of 30 �Cwas applied; the cathode flow rate was controlled using stoi-

chiometric numbers of 1.2. The cell was operated under 1 atm

of pressure at room temperature.

The mathematical model

In this work, a three-dimensional model for a high tempera-

ture PEMFC has been approached and implemented for its

solution in COMSOL Multiphysics 4.4. The model is similar to

those already existing in scientific literature [9,25], in which it

has been applied to study flow channel geometry influence

amongst other variables. The values of most simulation pa-

rameters are identical to [9]. Specific values consistent with

our experiment are listed in Table 1. Fig. 3 illustrates the

model geometry for PSFF with the channel depth of 600 mm.

Considering the MEA deformation, the channel depth in the

geometry is reduced by 40 mm for all simulation cases. Several

simplifications assumptions are introduced in this model: (1)

all processes are under steady-state conditions; (2) the GDLs,

the catalyst layers and the membrane are isotropic and ho-

mogeneous porous medium; (3) the gas-phase follows ideal

gas law; (4) the flows in the channels and porous layers are

assumed to be laminar and incompressible; (5) the trans-

portations of electron and proton are respectively controlled

by solid-phase electronic potential and membrane ionic

potential.

Governing equations of the model are derived as follows

(Table 2 for nomenclature):

(1) The convection and diffusion for the gaseous species (in

channels, GDLs, catalyst layers):

V$��DiVCi þ ujCi

� ¼ Si (1.1)

The diffusion coefficient of gaseous species in the diffusion

layer can be expressed as:

D ¼ Deff$ðεÞ1:5 (1.2)

Table 2 e Nomenclature used in equations of the PEMFCmodel.

C Molar concentration ðmol m�3ÞD Diffusivity ðm2 S�1ÞF Faraday's constant ðC mol�1Þi Current density ðA m�2Þkr Relative permeability

K Permeability

N Molar fluxðmol m�2 s�1ÞP Pressure (Pa)

R Gas constantðmol m�3 s�1ÞH Mass source term

u Velocity ðm s�1ÞVcell Cell voltage (V)

S Specific surface area ðm�1ÞEcell Thermodynamic equilibrium

potential (V)

Greek alphabet

h Over potential

a Charge transfer coefficient

d Conductivity ðS m�1Þ∅ Cell potential (V)

r Density ðkg m�3Þm Dynamic viscosity ðN s m�2Þs Electronic conductivity ðSm�1ÞSubscripts

a Anode

c Cathode

i Gaseous species

j Gas or liquid

ref Reference value

eff Effective value

s Electronic value

l Free electrolyte value

i n t e rn a t i o n a l j o u r n a l o f h y d r o g e n en e r g y 4 0 ( 2 0 1 5 ) 1 1 6 5 9e1 1 6 6 711662

(2) Momentum equation for the gaseous species (in chan-

nels, GDLs, catalyst layers):

�uj$V

�uj ¼ F� 1

rj

Vpj þmj

rj

V2uj (2.1)

Fig. 4 e The thickness of MEA (mm).

V$u ¼ 0 (2.2)

The average velocity of phase j is calculated by continuous

function based on a Darcy's law:

V$

rj

� Kkrj

mj

Vpj

!!¼ Hj (2.3)

uj ¼ �k

hVpj (2.4)

(3) The transportations of electron (in bipolar plates, GDLs,

catalyst layers) and proton (in PEM):

The source term of electronic current is set to be zero

because of no electrochemical reaction in GDLs.

�Vð�ssV∅sÞ ¼ 0 (3.1)

Similarly, the source term for the ionic current in PEM is

zero.

�Vð�slV∅lÞ ¼ 0 (3.2)

The catalyst layer is the region where reaction happens

and the ionic balance and electron balance are described as

follows:

�V�� sl;effV∅l

� ¼ �S$iloc (3.3)

�V�� ss;effV∅s

� ¼ S$iloc (3.4)

The local current density (iloc) generated by the porous

electrode reaction in the catalyst layer can be described by

Bulter-Volmer equation:

iloc ¼ io�exp

�aaF

h

RT

�� exp

��acF

h

RT

��(3.5)

iloc ¼ io�exp

�aaF

h

RT

�� exp

��acF

h

RT

��(3.6)

h ¼ ∅s �∅l � Ecell (3.7)

Results and discussion

The clamping test

Fig. 4 shows the deformation of the MEA at various clamping

forces. The values of lp and contact resistances are listed in

Table 3. The clamping force of 20 kgf,cm was adopted in the

single-cell test because the contact resistance was low. The

contact resistance did not decrease substantially when

further clamping force was applied. However, when a

clamping force of 20 kgf,cm was applied, approximately

92 mm of the MEA protruded into the channel. Obviously, a

considerable portion of channel space is required for shallow

channels.

The single-cell test

Fig. 5 shows the curves of electric potential versus power

density (VeP curves) for both SFF and PSFF. In the following

sections, the cell performances are evaluated based on the

peak power densities, as shown in Table 4 for various channel

depths and flow fields. In general, the peak power densities

Table 3 e The length of MEA protruded to channels (lp)and contact resistance (rc) in various clamping forces.

Clamping forces (kgf cm) 10 15 20 25

lp (mm) 80 86 92 96

rc (mU cm2) 72 68 64 63

Table 4 e The peak power densities (mW,cm¡2) forvarious channel depths and flow fields.

Channel depths TPS TSS

200 mm 145 144

300 mm 500 395

400 mm 562 532

600 mm 531 215

i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n en e r g y 4 0 ( 2 0 1 5 ) 1 1 6 5 9e1 1 6 6 7 11663

occur around the average potential of 0.4e0.5 V. For lower

average potential, the cell performance decreases because of

the appearance of large amount of liquidwater in the cathode.

The cell performance of SFF was generally more favorable

than that of PSFF, except when the depth was 200 mm, at

which both flow fields exhibited poor performance. However,

the difference decreased as the channels deepened, because

the single path in PSFF has sufficient space. For both flow

fields, the cell performance initially increased as the channel

depth increased, and then decreased when the channel was

too deep (600 mm). The SFF with a channel depth of 400 mm

produced the highest power density of 562 mW cm�2. Fig. 5(a)

and (b) show the pressure drops for the anode and cathode

channels, respectively. The pressure drop was higher for the

PSFF than for the SFF because of the single paths. Similar to

the VeP curves, the difference decreased as the channel depth

increased.

Here, the impact of channel depth based on a cross in-

spection of Fig. 5 and Fig. 6 is discussed. At the channel depth

of 200 mm, both SFF and PSFF exhibited poor performance. The

pressure drop was high because of MEA deformation. The

cathode channel was completely clogged at a current density

of approximately 460 mA cm�2..At a channel depth of 300 mm, the performance of SFF was

muchmore favorable than that of PSFF. The cathode pressure

drop for PSFF was high and increased as the current density

increased, revealing that inefficient water removal occurs in

the single paths of PSFF flow fields because the channel space

is insufficient. It is also shown in Fig. 6 that the pressure drop

difference between SFF and PSFF at anode and cathode is

highest at 300 mm depth. This is because such depth provides

Fig. 5 e The VeP curves for the SFF and PSFF.

smaller channel cross-sectional area, comparing to deeper

channel. The single path in PSFF thus requires larger pressure

gradient to receive and distribute fluids. This is evidential for

cathode during lower cell potential, where more liquid water

is produced.

At a channel depth of 400 mm, both SFF and PSFF produced

their highest power densities compared with those at other

depths. The peak power density of SFF was the highest of all

flow fields; the pressure gradients were approximated to be

0.13 psi mm�1 and 0.83 psi mm�1 for anode and cathode

channels, respectively.

At a channel depth of 600 mm, the performance of both SFF

and PSFF was similar, indicating that deeper channels provide

sufficient space for the single path in PSFF to receive and

distributewater in the cathode. The pressure gradient for both

Fig. 6 e The pressure drops for (a) anode and (b) cathode

channels.

i n t e rn a t i o n a l j o u r n a l o f h y d r o g e n en e r g y 4 0 ( 2 0 1 5 ) 1 1 6 5 9e1 1 6 6 711664

flowfieldswas approximated to 0.10 psimm�1. The anode and

cathode pressure gradients were nearly identical, indicating

that the cathode water flows in the bottom of the channel,

leaving free space for gas flow. Two-phase flow considerably

complicates oxygen distribution. Therefore, the channel

depth cannot be increased excessively to improve the cell

performance.

A significant amount of researches have investigated the

effects of different channel geometrics on cell performance

based on experiments or simulations. The results of Scholta

et al. [18] suggest that the optimal channel-to-rib width ratio is

approximately one, as in Fig. 1. They also show that the

relationship between rib width and other channel geometrics

is less significant. This is because the rib width, although has

strong influences on the path length of electron transport, has

less or none effects on the gaseous flow. Therefore, the effects

of channel depth for various rib widths are theoretically

similar to the results described above. On the other hand,

Manso et al. [19] numerically investigated the effects channel

depth-to-width ratio using ten different channel geometrics.

Their results reveal that the cell performance increases for

higher channel depth-to-width ratio if the ratio value is less

than or close to one. Therefore, for different channel widths,

the overall cell performance might change while the effects of

channel depth remain the same.

Experiments of PSFF with various flow rates are evaluated

because the impacts of channel depth are much sensitive in

such a flow field. Fig. 7 shows the comparison of cell perfor-

mances at anode flow rates of 60 and 80 c.c. min�1. The

cathode flow rate was controlled using stoichiometric

numbers of 1.2. It can be seen at a channel depth of 300 mm

that the cell performance improves for higher flow rate. It is

because patterns of two-phase flow in the cathode channel

are more likely to change from plug flow to more efficient

annular flow due to high flow velocity. At a channel depth of

400 mm, the cell performance slightly decreases at higher flow

rates. It reveals that the cell performance at the flow rate of

60 c.c.min�1 is at its optimum. Too high flow velocity shortens

the gaseous retention time and reduces the cell performance.

Fig. 7 e The cell performances of PSFF with different

channel depths at anode flow rates of 60 and 80 c.c. min¡1.

At a channel depth of 600 mm, the increase of flow rates im-

proves cell performance a little. In such a flow rate, the IeP

curves is very similar for the cell with the depth of 400 and

600 mm. Deeper channel has larger cross-sectional area,

leading to lower flow velocity. It requires larger flow rate to

increase flow velocity and convective mass transport. As a

result, it is economically inefficient to fabricate bipolar plates

with too deep channel and operate at a high flow rate.

The numerical analysis

Fig. 8 compares experimental and simulated polarization

curves. In general, calculated results agree with experimental

data. However, the finite elemental method diverges and fails

to find the solution for the average potential smaller than

0.45 V. In addition, simulation results slightly over estimates

the current densities for smaller average potential. It appears

that the proposed model has limitation in dealing with the

complicity of two-phase flow. As described in existent litera-

ture [15,16] that the process of random droplet emergence and

liquid water clogging is not clearly understood because water

clogging is highly localized to certain regions in specific

channels. Nevertheless, the observation of individual channel

velocity provides some insight about the pattern of two-phase

Fig. 8 e Model validation for (a) SFF and (b) PSFF; Solid line:

experiment, Dotted line (Simulation).

Fig. 9 e Predicted flow velocity distribution (m/s) at 0.45 V. (a) SFF, (b) PSFF; (1) 300 mm, (2) 400 mm, (3) 600 mm.

i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n en e r g y 4 0 ( 2 0 1 5 ) 1 1 6 5 9e1 1 6 6 7 11665

flow. Fig. 9 shows the predicted flow velocity distribution at

0.45 V. As mentioned before, the cell with the depth of 600 mm

has lower average velocity and thus reduces the convective

mass transport and cell performance. However, the cell with

shallower depth, although provides higher average flow ve-

locity, shows highly heterogeneous in the distribution of ve-

locity. It can be observed that very low flow velocity exists in

most inner channels. In such a case, the plug flow may

dominate in the cathode channel and have more chance of

water clogging. On the other hand, deeper channel provides

sufficient cross-sectional area and may lead to annular flow.

Fig. 10 compares the vorticities in the cut plane at the six

channels near the cathode output (Fig. 3) for PSFF 300 mm and

600 mm. It can be seen in the channel depth of 300 mm that the

vorticities is significant small in the 4th to 6th channels. In

Fig. 10 e The vorticities (1/s) in the cut plane (Fig. 3

contrast, the vorticities is significant and uniform in all

channels with the depth of 600 mm. The vortices create cen-

trifugal forces to lead the liquid flow against the channel wall,

leaving channel central space for gaseous flow. Such an

annular flow removes water more efficiently and yields better

cell performance.

Conclusion

This study involved applying die-sinking micro-EDM to pre-

pare micro flow channels in metallic bipolar plates. The ef-

fects of channel depth were examined by measuring pressure

drops, observing MEA deformation, and numerical analysis.

The conclusions are listed as follows:

) for PSFF (a) 300 mm and (b) 600 mm at 0.45 V.

i n t e rn a t i o n a l j o u r n a l o f h y d r o g e n en e r g y 4 0 ( 2 0 1 5 ) 1 1 6 5 9e1 1 6 6 711666

1. An SFF with a channel depth of 400 mm produced the

highest power density of 562 mW cm�2.

2. The cell performance of SFF is generally more favorable

than that of PSFF. However, the cell performance becomes

similar as the channel depth increases, because the single

paths in PSFF cathode flow fields have sufficient space to

distribute water.

3. Because MEA deformation occupies portions of channel

space, a sufficient channel depth is required to enable

reactant transportation and water removal. However,

when the channel is too deep, too low flow velocity further

reduces the convective mass transport and cell perfor-

mance. The increment of flow rate slightly improves the

cell performance. However, it is economically inefficient to

fabricate bipolar plates with too deep channel and operate

at high flow rate.

Acknowledgments

This work was supported by the National Science Council of

Taiwan under grant numbers NSC 102-2221-E-324 -021 and

NSC 102-2622-E-324 -003 -CC3.

r e f e r e n c e s

[1] Aric�o AS, Cretı P, Baglio V, Modica E, Antonucci V. Influenceof flow field design on the performance of a direct methanolfuel cell. J Power Sources 2000;91:202e9.

[2] Choi K-S, Kim H-M, Moon S-M. Numerical studies on thegeometrical characterization of serpentine flow-field forefficient PEMFC. Int J Hydrog Energy 2011;36:1613e27.

[3] Ferng YM, Su A. A three-dimensional full-cell CFD modelused to investigate the effects of different flow channeldesigns on PEMFC performance. Int J Hydrog Energy2007;32:4466e76.

[4] Higier A, Liu H. Optimization of PEM fuel cell flow field vialocal current density measurement. Int J Hydrog Energy2010;35:2144e50.

[5] Hsieh S-S, Chu K-M. Channel and rib geometric scale effectsof flowfield plates on the performance and transient thermalbehavior of a micro-PEM fuel cell. J Power Sources2007;173:222e32.

[6] Inoue G, Matsukuma Y, Minemoto M. Effect of gas channeldepth on current density distribution of polymerelectrolyte fuel cell by numerical analysis including gasflow through gas diffusion layer. J Power Sources 2006;157:136e52.

[7] Jeon DH, Greenway S, Shimpalee S, Van Zee JW. The effect ofserpentine flow-field designs on PEM fuel cell performance.Int J Hydrogen Energy 2008;33:1052e66.

[8] Kumar A, Reddy RG. Effect of channel dimensions and shapein the flow-field distributor on the performance of polymerelectrolyte membrane fuel cells. J Power Sources2003;113:11e8.

[9] Lobato J, Ca~nizares P, Rodrigo MA, Pinar FJ, Mena E, Ubeda D.Three-dimensional model of a 50 cm2 high temperature PEMfuel cell. Study of the flow channel geometry influence. Int Jof Hydrogen Energy 2010;35:5510e20.

[10] Shimpalee S, Van Zee JW. Numerical studies on rib &channel dimension of flow-field on PEMFC performance. Int JHydrogen Energy 2007;32:842e56.

[11] Wang X-D, YanW-M, Duan Y-Y, Weng F-B, Jung G-B, Lee C-Y.Numerical study on channel size effect for proton exchangemembrane fuel cell with serpentine flow field. EnergyConvers Manag 2010;51:959e68.

[12] Yoon Y-G, Lee W-Y, Park G-G, Yang T-H, Kim C-S. Effects ofchannel configurations of flow field plates on theperformance of a PEMFC. Electrochimica Acta2004;50:709e12.

[13] Yoon Y-G, Lee W-Y, Park G-G, Yang T-H, Kim C-S. Effects ofchannel and rib widths of flow field plates on theperformance of a PEMFC. Int J Hydrog Energy 2005;30:1363e6.

[14] Hung J-C, Lin C-C. Fabrication of micro-flow channels formetallic bipolar plates by a high-pressure hydroformingapparatus. J Power Sources 2012;206:179e84.

[15] Mortazavi M, Tajiri K. Two-phase flow pressure drop in flowchannels of proton exchange membrane fuel cells: review ofexperimental approaches. Renew Sustain Energy Rev2015;45:296e317.

[16] Lu Z, Kandlikar SG, Rath C, GrimmM, Domigan W, White AD,et al. Water management studies in PEM fuel cells, part II: exsitu investigation of flow maldistribution, pressure drop andtwo-phase flow pattern in gas channels. Int J Hydrog Energy2009;34:3445e56.

[17] Shimpalee S, Greenway S, Van Zee JW. The impact ofchannel path length on PEMFC flow-field design. J PowerSources 2006;160:398e406.

[18] Scholta J, Escher G, Zhang W, Kuppers L, J€orissen L,Lehnert W. Investigation on the influence of channelgeometries on PEMFC performance. J Power Sources2006;155:66e71.

[19] Manso AP, Marzo FF, Mujika MG, Barranco J, Lorenzo A.Numerical analysis of the influence of the channel cross-section aspect ratio on the performance of a PEM fuel cellwith serpentine flow field design. Int J Hydrog Energy2011;36:6795e808.

[20] Hahn R, Wagner S, Schmitz A, Reichl H. Development of aplanar micro fuel cell with thin film and micro patterningtechnologies. J Power Sources 2004;131:73e8.

[21] Kuriyama N, Kubota T, Okamura D, Suzuki T, Sasahara J.Design and fabrication of MEMS-based monolithic fuel cells. JPower Sources 2008;145e146:354e62.

[22] Meyers JP, Maynard HL. Design considerations forminiaturized PEM fuel cells. J Power Sources 2002;109:76e88.

[23] Yu J, Cheng P, Ma Z, Yi B. Fabrication of miniature siliconwafer fuel cells with improved performance. J Power Sources2003;124:40e6.

[24] Zhang Y, Lu J, Shimano S, Zhou H, Maeda R. Development ofMEMS-based direct methanol fuel cell with high powerdensity using nanoimprint technology. ElectrochemCommun 2007;9:1365e8.

[25] Yang W-J, Wang H-Y, Lee D-H, Kim Y-B. Channel geometryoptimization of a polymer electrolyte membrane fuel cellusing genetic algorithm. Appl Energy 2015;146:1e10.

[26] Park BY, Madou MJ. Design, fabrication, and initial testing ofa miniature PEM fuel cell with micro-scale pyrolyzed carbonfluidic plates. J Power Sources 2006;162:369e79.

[27] Hung J-C, Chang D-H, Chuang Y. The fabrication of high-aspect-ratio micro-flow channels on metallic bipolar platesusing die-sinking micro-electrical discharge machining. JPower Sources 2012;198:158e63.

[28] Hung J-C, Lin C-C. Fabrication of micro-flow channels formetallic bipolar plates by a high-pressure hydroformingapparatus. Journal of Power Sources.

[29] Hung J-C, Yang T-C, Li K-C. Studies on the fabrication ofmetallic bipolar plateseusing micro electrical dischargemachining milling. J Power Sources 2011;196:2070e4.

[30] Lee S-J, Lai J-J, Huang C-H. Stainless steel bipolar plates. JPower Sources 2005;145:362e8.

i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n en e r g y 4 0 ( 2 0 1 5 ) 1 1 6 5 9e1 1 6 6 7 11667

[31] Lee S-J, Lee C-Y, Yang K-T, Kuan F-H, Lai P-H. Simulation andfabrication of micro-scaled flow channels for metallic bipolarplates by the electrochemical micro-machining process. JPower Sources 2008;185:1115e21.

[32] Lee S-J, Lee Y-M, Lee C-Y, Lai J-J, Kuan F-H, Chuang C-W. Theperformance of miniature metallic PEM fuel cells. J PowerSources 2007;171:148e54.

[33] Tuber K, Oedegaard A, HermannM, Hebling C. Investigation offractal flow-fields in portable proton exchangemembrane anddirect methanol fuel cells. J Power Sources 2004;131:175e81.

[34] Lobato J, Ca~nizares P, Rodrigo MA, Pinar FJ, Ubeda D. Study offlow channel geometry using current distributionmeasurement in a high temperature polymer electrolytemembrane fuel cell. J Power Sources 2011;196:4209e17.