A1358 Journal of The Electrochemical Society, 153 共7兲 A1358-A1364 共2006兲

0013-4651/2006/153共7兲/A1358/7/$20.00 © The Electrochemical Society

Analysis of Mass Transport of Methanol at the Anode of a

Direct Methanol Fuel Cell
C. Xu,a Y. L. He,b T. S. Zhao,a,*,z R. Chen,a and Q. Yea
a
Department of Mechanical Engineering, The Hong Kong University of Science and Technology, Clear
Water Bay, Kowloon, Hong Kong SAR, China
b
School of Energy and Power Engineering, Xi’an Jiaotong University, Xi’an 710049, China

We present an analysis of the mass transport of methanol at the anode of a direct methanol fuel cell 共DMFC兲 and show that the
overall mass-transfer coefficient can be determined by measuring the cell limiting current density. We measured the cell limiting
current density of an in-house-fabricated DMFC with different flow fields for various methanol concentrations and flow rates of
methanol solution. The experimental data showed that the overall mass-transfer coefficient was nearly independent of current
density, although the rate of CO2 gas bubble liberation changed with current density. We found that the overall methanol-transfer
coefficient in the serpentine flow field could be significantly increased with increased methanol flow rate due to the enhanced
under-rib convection. We developed the correlation equations predicting the overall methanol transfer coefficient in terms of the
methanol flow rate for a given DMFC hardware. Finally, we showed that the polarization curves predicted based on the correlation
equation of the overall mass-transfer coefficient in the DMFC with the serpentine flow field for different flow rates were in fairly
good agreement with the experimental data.
© 2006 The Electrochemical Society. 关DOI: 10.1149/1.2201467兴 All rights reserved.

Manuscript submitted November 25, 2005; revised manuscript received January 24, 2006. Available electronically May 16, 2006.

The liquid feed direct methanol fuel cell 共DMFC兲, based on a small CO2 bubbles grew and coalesced with the slugs, resulting in a
solid polymer electrolyte, has received much attention as a leading pulsed growth behavior. It was concluded that CO2 bubble removal
candidate power source for portable electronic devices, electric ve- from the backing layer by flow-induced removal and large CO2
hicles, and other mobile applications because of its unique advan- bubble coalescence. Yang et al.8 studied the CO2 bubble behavior in
tages such as high energy-conversion efficiency, easy delivery and the anode flow field of an in-house-fabricated transparent DMFC
storage of liquid fuel, ambient temperature operation, and simple with a 4.0 ⫻ 4.0 cm active area. The effects of cell orientations,
construction.1-4 However, the commercialization of the DMFC is methanol solution flow rates, and operation temperatures were in-
still hindered by several technological problems, including a sub- vestigated. Furthermore, the channel-clogging phenomenon caused
stantial methanol crossover through polymer membranes, low elec- by CO2 slugs was never encountered in the single-serpentine flow
troactivity of methanol oxidation on the anode, and severe cathode field. More recently, Wong et al.9 investigated the performance and
flooding.4 Over the last decade, different fundamental aspects of the CO2 bubble evolution behavior in an in-house-fabricated micro-
DMFC system have been studied extensively.2-14 DMFC with various-sized microchannels down to 400.0 ␮m. They
It has been understood that mass transport of reactants in a fuel found a periodical gas-slug blockage behavior and analyzed the ef-
cell is one of the crucial issues in improving cell performance. To fect of this behavior on the mass transfer of methanol.
alleviate the problem of methanol crossover, a DMFC typically has The study of mass transfer in DMFCs and in proton exchange
to be operated with diluted methanol solution. However, rather di- membrane fuel cells 共PEMFCs兲 has mainly been focused on flow
luted methanol solution results in large mass-transport polarization, field geometry11,13,18-22 and characteristics of the diffusion layer
particularly at high current densities, leading to lower cell perfor- 共DL兲.23,24 Scott et al.12 measured the limiting current density data to
mance. Therefore, it is essential to optimize the mass-transport pro- study the effect of temperature, air cathode pressure, methanol fuel
cess of methanol at the DMFC anode such that both the rate of flow rate, and methanol concentration on the mass-transfer resis-
methanol crossover and the mass-transport polarization can be mini- tance and cell performance based on a parallel anode flow field.
mized. To this end, it is essential to gain a better understanding of They concluded that the mass-transport limitation arose from the
mass-transport phenomena in fuel cells. Compared with other types diffusion of methanol in the carbon cloth covering the membrane
of fuel cells, relatively few papers have been reported on the study electrode assembly 共MEA兲 and from the hydrodynamic influence of
of the methanol transport process at the anode of a DMFC.10-14 This gas bubbles in the flow channel. Yang et al.14 found that the in-
is mainly because the mass transport of methanol occurs in a liquid– creased liquid velocity due to the increase in void fraction of the gas
gas two-phase flow, consisting of methanol solution and reaction- bubbles in flow channels enhances the mass transfer of methanol
produced gas CO2, under typical DMFC operating conditions, from the flow channel to the gas diffusion layer and hence improves
which results in a rather complicated mass-transport process. cell performance. Wong et al.13 found that smaller channel depth
Clearly, the mass transport of methanol is affected by the gas CO2 affected the cell performance adversely due to the longer gas slugs
flow behavior and the gas CO2 liberation rate, as well as the geo- in flow channels.
metric design of DMFC flow fields. Recently, the effect of under-rib convection in the DL on the
The study of two-phase flow patterns in the anode flow field has mass transfer both in interdigitated flow fields and in serpentine flow
been reported recently.8-10,15-17 Argyropoulos et al.15 studied the fields has been studied mainly in PEMFCs.24-31 Pharoah et al.26
CO2 bubble flow characteristics in a DMFC based on two different performed a computational fluid dynamics 共CFD兲 simulation to
parallel flow channels. It was observed that there were three two- study the importance of convective transport as a function of the DL
phase flow patterns, including bubbly, slug, and annular flow ac- permeability. They found that for typical flow-field dimensions and
cording to different levels of gas fraction, and gas-slug clogging fuel cell operating conditions, convective transport occurs in the DL
often occurred in some of the flow channels of the parallel flow when the permeability exceeds 10−13 m2. Wang et al.27 conducted a
fields. Mench et al.16 observed the order of 0.1–0.5 mm CO2 bubble 3D CFD simulation and found that there exists a large pressure drop
growth and ejection from the different locations within the channel between two adjacent channels, which leads to a reactant-flow short
diffusion layer 共DL兲 interface region with video microscopy. Those circuit. Their results showed that 0.1 m/s velocity in the porous DL
can be induced by the channel-to-channel pressure difference, and
convection is dominant in the lateral species transport through the
* Electrochemical Society Active Member. DL. Oosthuizen et al.28 studied the species under-rib crossover due
z
E-mail: metzhao@ust.hk to the channel-to-channel pressure difference and the effect of this
 Journal of The Electrochemical Society, 153 共7兲 A1358-A1364 共2006兲 A1359

cess of methanol occurs in a liquid-gas two-phase flow. Although

different gas CO2 bubble behaviors in DMFCs have been studied
with the help of visualization techniques,8-10,15-17 the effect of gas
bubbles on the methanol transport still remains unclear.
The mass flux of methanol, J, is affected by two mass-transfer
resistances: one is the hydrodynamics mass-transfer resistance from
the flow channel to the channel/DL interface, causing a concentra-
tion difference 共Cb − Cw兲, and the other is the mass-transfer resis-
tance in the porous DL, causing a concentration difference 共Cw
− Ccl兲. In the flow channel, methanol transport depends not only on
the hydrodynamics in the flow channel but also on the gas-bubble
liberation behavior at the surface.12 Gas bubbles and slugs in the
flow channel affect the hydrodynamic mass transfer through the
channel/DL interface to the DL in two opposite directions. On one
hand, the presence of gas slugs accelerates the velocity of the metha-
nol solution, increasing the convection mass-transfer coefficient at
the channel/DL interface.13,14 On the other hand, gas slugs may
Figure 1. Schematic diagrams illustrating the transport mechanisms of
methanol at the DMFC anode: 共a兲 diffusion alone in the DL and 共b兲 diffusion
cover some areas of the permeable wall, reducing the effective area
with under-rib convection in the DL. for methanol transfer from the channel to the DL.4,13 Therefore, the
methanol flux through the channel/DL interface can be expressed as

J = 共A␦ − Ab兲h共Cb − Cw兲 = A␦共1 − ␤兲h共Cb − Cw兲 关3兴

crossover on the pressure distribution in the channel and the tem-
perature distribution in the flow plate. In their situations, channel- where A is the area of electrode, ␦ is the open ratio of the flow field,
to-channel flow crossover is only significant when the porosity of h is the mass-transfer coefficient at the channel/DL interface, and ␤
the diffusion layer exceeds approximately 0.65. Williams et al.31 is the ratio of the gas coverage fraction on the channel/DL interface.
studied the influence of under-rib convection through the DL on the Although it is difficult to quantify both ␤ and h at this stage, Eq. 3
cathode limiting current in PEMFCs using three DLs with distinc- indicates that both ␤ and h affect the mass-transport of methanol. As
tively different gas permeabilities. They found that under-rib con- an approximation, the mass-transfer coefficient h can simply be ob-
vection contributed to the limiting current density of PEMFCs even tained from the heat and mass-transfer analogy for a fully developed
when using a non-interdigitated flow field pattern. Our literature single-phase laminar duct flow.4 For example, for a fully developed
review indicates that the effect of under-rib convection on cell per- laminar flow through a square channel with a constant mass flux
formance in DMFCs has not been studied yet. applied at one surface and no flux at the others, the Sherwood num-
In this work, we present an analysis of the mass transport of ber is given by32
methanol at the anode of a DMFC. We show that the overall mass-
transfer coefficient is related to the cell limiting current density. We H
Sh = h = 2.71 关4兴
measured the overall mass-transfer coefficient of an in-house- D
fabricated DMFC with different flow fields for various methanol
concentrations and flow rates of methanol solution. We paid particu- where H is the equivalent diameter of channel and D is the diffu-
lar attention to the effect of both the gas liberation rate and pressure- sivity of methanol in the water.
driven under-rib convection in the DL on the DMFC performance. In the DL, the mass transport of methanol is rather complicated
for the following reasons. First, the gas CO2 emerged in the DL may
Analytical occupy some pores, blocking the diffusion of methanol molecules.
Consider the mass transport of methanol solution on the anode of Second, the interfaces between the flow field ribs and the DL 共see
a DMFC, as shown in Fig. 1a, which consists of a flow field 共flow Fig. 1兲 result in a nonuniform distribution of methanol underneath
channels and collector ribs兲, a DL, and a catalyst layer 共CL兲. Metha- the ribs and a lengthened diffusion path of methanol in the DL.
nol solution is forced to flow along the length of the flow channel Third, as illustrated in Fig. 1b, methanol solution flows in serpentine
while penetrating the DL to the CL, where part of methanol is oxi- or interdigitated flow fields may cause a pressure difference between
dized to form protons, electrons, and carbon dioxide as follows two adjacent channels, leading to channel-to-channel convection un-
der ribs 共referred to as under-rib convection hereafter兲, which may
CH3OH + H2O → CO2 + 6H+ + 6e− 关1兴 also contribute to the methanol transfer in the DL. To incorporate all
while the remaining methanol crosses over through the membrane to these effects, an effective mass-transfer coefficient keff is introduced
the cathode. According to Faraday’s law, the flux of methanol across and accordingly, the methanol flux through the DL can be expressed
the channel/DL interface toward the catalyst layer can be expressed as
as
J = Akeff共Cw − Ccl兲 关5兴
I
J= M + Jm 关2兴 Note that the effects of both the rib shading and the under-rib con-
6F
vection are taken into account by keff, which depends on the thick-
where I is the total current of the cell, M is the molecular weight of ness, permeability, and porosity of the DL, methanol diffusivity,
methanol, and Jm is the crossover flux of methanol. In the meantime, width of the collector ribs and channels, and the channel-to-channel
the produced CO2 in the CL transports backward through the DL to pressure difference. If neglecting the effects of rib coverage and
the flow channel and is swept by the stream of liquid solution to- under-rib convection for approximation, the mass-transfer coeffi-
ward the exit of the flow channel. Gas CO2 appears in the DL and cient keff can be simply obtained by the following equation
the flow channel when the methanol vapor pressure exceeds its satu-
rated value corresponding to the operating cell temperature. For ex- Deff D␧1.5
ample, for a typical DMFC operating with a methanol flow flux of keff = = 关6兴
l l
0.1 mL/min cm2 and at 75°C, CO2 gas bubbles appear even at a
rather low current density of 10.0 mA/cm2 due to its low solubility. where ␧ and l represent the porosity and the thickness of the DL.
Therefore, under typical operating conditions, the mass-transfer pro- Combining Eq. 3 and 5, we obtain
A1360 Journal of The Electrochemical Society, 153 共7兲 A1358-A1364 共2006兲

J Cb − Ccl
j= = = ktot共Cb − Ccl兲 关7兴
A 1 1
+
␦共1 − ␤兲h keff
where the overall mass-transfer coefficient 1/ktot = 共1/␦共1 − ␤兲h兲
+ 共1/keff兲 represents the combined mass-transfer coefficient both at
the channel/DL interface and in the DL. Combining Eq. 2 and 7, we
obtain
i
+ jm = ktot共Cb − Ccl兲 关8兴
6F
For a given bulk methanol concentration Cb, increasing the current Figure 2. Designs of the anode flow fields 共drawn to scale兲.
density leads to a decrease in the methanol concentration in the
catalyst layer Ccl. As the limiting current density is reached 共i
= ilim兲, the methanol concentration in the catalyst layer decreases to i
zero, i.e., Ccl = 0, and the methanol crossover goes to null 共 jm 6F
= 0兲. Therefore, Eq. 8 reduces to ktot = 关16兴
iA
共Cin兲lim −
ilim = 6FktotCb 关9兴 12FQ
We now define the bulk methanol concentration Cb as the mean With Eq. 16, the influence of current density, i.e., the CO2 gas lib-
concentration between the channel inlet and outlet, i.e. eration rate, on the mass transport of methanol can be studied.

Cin + Cout Experimental

Cb = 关10兴
2 The DMFC.— The in-house-fabricated DMFC consisted of a
MEA, with an active area of 3.0 ⫻ 3.0 cm, sandwiched between
Based on the mass balance of methanol over the entire flow channel two bipolar plates, which were fixed by two fixture plates. The MEA
ilimA consisted of a Nafion 115 membrane and two E-TEK single-side
Cout = Cin − 关11兴 ELAT electrodes. The carbon cloth 共E-TEK, type A兲 with 30% poly-
6FQ tetrafluoroethylene 共PTFE兲 wetproofing treatment was used as the
where Q is the volumetric flow rate of methanol solution. Substitut- backing support layer both in the anode and in the cathode elec-
ing Eq. 11 into 10 yields trodes. 4.0 mg/cm2 with unsupported 关Pt/Ru兴 Ox 共1:1 atom %兲 and
2.0 mg/cm2 using 40% Pt on Vulcan XC-72 were used as the cata-
ilimA lyst loading on the anode and cathode side. Furthermore,
Cb = Cin − 关12兴 0.8 mg/cm2 Nafion was applied onto the catalyst layer surface of
12FQ
each electrode. The membrane was pretreated by a well-known stan-
Substituting Eq. 12 into 9, we obtain dard procedure.8 Finally, the MEA was formed by hot pressing un-
der the condition of 135°C and 4 MPa for 3 min.
6Fktot
ilim = Cin 关13兴 The flow-field design.— To avoid corrosion, the flow-field plates
Aktot
1+ were made of 316 L stainless steel plates with a thickness of
2Q 1.0 mm, into which the flow fields were fabricated by wire-cut tech-
nique. To reduce the electrical resistance, the flow fields were sput-
Equation 13 indicates that for a given methanol flow rate, there
tered with a thin layer of Pt 共0.2 ␮m兲. A single-serpentine cathode
exists a limiting current density corresponding to each methanol
concentration at the inlet of the flow channel. Alternatively, Eq. 13 flow field was used in all the experiments. Two anode flow fields, a
can also be rewritten as single-serpentine flow field 共SSFF兲 and a parallel flow field 共PFF兲,
as shown in Fig. 2, were tested. The open ratios of all the flow fields
Aktot are 0.5, and both the channel width and the rib width are 1.0 mm.
1+
2Q The test rig.— The experiments were carried out in the test rig
共Cin兲lim = i 关14兴
6Fktot detailed elsewhere.8 On the anode, aqueous methanol solution was
fed by a digital high-performance liquid chromatography 共HPLC兲
where 共Cin兲lim represents the critical methanol concentration at micropump 共series III兲. Before entering the cell, the methanol solu-
which i = ilim. Equation 14 implies that there exists a critical metha- tion was preheated to the desired temperature by a heater connected
nol concentration 共Cin兲lim at which the methanol concentration in the to a temperature controller. On the cathode, 99.999% high-purity
catalyst layer becomes zero, corresponding to each current density i. oxygen was supplied without humidification. A mass-flow meter
Solving for the overall mass-transfer coefficient ktot from Eq. 13 共Omega FMA-7105E兲, along with a multiple-channel indicator
gives 共Omega FMA-5876A兲, was used to control and measure the flow
rate of oxygen.
ilim
Measurement instrumentation and test conditions.— In this
6F work, the Arbin BT2000 共Arbin Instrument兲 electroload interfaced
ktot = 关15兴
ilimA with a computer was employed to control the cell operation and
Cin − measure voltage-current 共polarization兲 curves. All the experiments
12FQ
were performed at 75°C and used the same cathode flow field, hav-
Equation 15 indicates that for a given methanol concentration Cin ing the same configuration and geometry as SSFF and fed with a
and a flow rate Q, the overall mass-transfer coefficient ktot can be constant oxygen gas flow rate of 200.0 sccm at atmospheric pres-
determined by measuring the limiting current density. Similarly, Eq. sure. Such an extremely high oxygen feed rate to the cathode en-
14 can also be rewritten as sured that the measured limiting current density of the DMFC in this
 Journal of The Electrochemical Society, 153 共7兲 A1358-A1364 共2006兲 A1361

Figure 3. Polarization curves for different methanol concentrations in the

cell with the PFF 共flow rate 1.0 mL/min; temperature 75°C兲.

work was caused by the mass-transport limitation of methanol,

rather than by the mass-transport limitation of oxygen. Methanol
concentration was varied from 0.125 to 1.0 M, while the flow rate
was changed from 0.2 to 25.0 mL/min.
Results and Discussion
Measured limiting current density.— Figure 3 shows the polar-
ization curves of the DMFC fed with 0.5 and 2.0 M methanol con-
centrations at the same flow rate of 1.0 mL/min. It can be seen that
for the lower methanol concentration of 0.5 M, increasing the cur-
rent density resulted in a sharp drop in cell voltage, meaning that the Figure 4. Polarization curves for different flow rates 共methanol concentra-
methanol in the catalyst layer was depleted and the limiting current tion 0.125 M; temperature: 75°C兲: 共a兲 SSFF and 共b兲 PFF.
density was reached. However, for the higher methanol concentra-
tion of 2.0 M, with increasing current density the cell voltage
dropped almost linearly toward zero but there was no sharp drop in
1
the cell voltage. This polarization behavior means that at higher
methanol concentrations, the voltage loss is predominately caused 6F
ktot = 关17兴
by the sluggish reaction and the mixed potential as a result of the Cin A
increased rate of methanol crossover but not by the transport limi- −
ilim 12FQ
tation of methanol. Therefore, there exists no limiting current den-
sity at sufficiently high methanol concentrations, unless the metha- Because Cin /ilim ⬇ constant, Eq. 17 implies that the overall mass-
nol flow rate is extremely low. The objective of this work was to transfer coefficient depends on the flow rate of methanol solution Q
determine the overall mass-transfer coefficient with Eq. 15 by mea- only. This finding will be addressed in more detail in discussing Fig.
suring the limiting current densities. All the experimental data re- 6.
ported hereafter were collected at low methanol concentrations and Comparing Fig. 5a and b indicates that the influence of the
low flow rates, under which the limiting current density occurred. methanol flow rate on the limiting current density for the serpentine
Figure 4 shows the polarization curves of the DMFC with either flow field 共Fig. 5a兲 is much more significant than for the parallel
the SSFF 共Fig. 4a兲 or the PFF 共Fig. 4b兲. The experiments were flow field 共Fig. 5b兲. The more significant effect of the flow rate on
conducted by varying the flow rate while fixing the concentration of the limiting current density for the serpentine flow field can be at-
methanol solution at 0.125 M. It is seen that for both flow fields the tributed to two aspects. First, at the same flow rate the liquid veloc-
limiting current density increased with increasing flow rate as a ity in the serpentine flow channel is higher than in the parallel flow
result of the improved mass transport of methanol. For the serpen- channels, leading to a higher mass-transfer coefficient h at the
tine flow field, the limiting current density increased from channel/DL interface; second, the larger pressure drop from the inlet
51.0 to 153.0 mA/cm2, nearly three times, when the flow rate was to outlet in the serpentine flow field may cause the under-rib con-
increased from 1.0 to 25.0 mL/min. However, for the parallel flow vection between two adjacent channels, increasing keff.
field, the same increment in the flow rate resulted in a relatively Mass-transfer coefficient.— The experimental data shown in Fig.
smaller increase in the limiting current density 共from 5 can also be presented by plotting the overall mass-transfer coeffi-
33.0 to 58.0 mA/cm2兲. Figure 4 also indicates that at the same flow cient against current density based on Eq. 16 for each flow rate and
rate the serpentine flow field yielded better cell performance. each methanol concentration corresponding to each limiting current
The variations in the limiting current density with methanol con- density. The results are shown in Fig. 6a for SSFF and in Fig. 6b for
centration for different flow rates are shown in Fig. 5a for SSFF and PFF. It is interesting to notice from Fig. 6a that for the serpentine
Fig. 5b for PFF. It is clear that for each given flow rate the limiting flow field, when the flow rate was below 5.0 mL/min, the overall
current density increased nearly linearly with methanol concentra- mass-transfer coefficient was nearly independent of current density,
tion for both flow fields, i.e., Cin /ilim ⬇ constant. To address the as mentioned in discussing Fig. 5. The CO2-gas liberation rate in-
implication of this behavior, we rewrite Eq. 15 as creases with current density. An increase in current density will lead
A1362 Journal of The Electrochemical Society, 153 共7兲 A1358-A1364 共2006兲

Figure 6. Overall mass-transfer coefficient vs cell current density for differ-

ent flow rates 共temperature: 75°C兲: 共a兲 SSFF and 共b兲 PFF.
Figure 5. Limiting current density vs methanol concentration for different
flow rates 共temperature: 75°C兲: 共a兲 SSFF and 共b兲 PFF.
and low methanol flow rate, gas-bubbles in the channel are continu-
ously swept by methanol flow and no gas-slug clogging occurs.8
to an increase in the gas-void fraction in both the DL and the flow While in parallel flow fields, immobile gas-slug clogging often oc-
channel. At the same flow rate of methanol solution, increasing the curs in some channels.15 The immobile gas clogging in channels is
gas void fraction in the flow channel increases the liquid-phase ve- considered to cover the DL badly, which greatly reduces the effec-
locity, resulting in a higher hydrodynamic mass-transfer coefficient tive area for methanol transport through the channel/DL interface
共h兲 at the channel/DL interface. However, increasing the gas void and causes severe mass-transfer resistance. However, when the gas
fraction in the DL increases the mass-transfer resistance of metha- clogging is removed by liquid flow, the moving gas slugs in the
nol. The current density independence of the overall mass-transfer channel do not cause severe mass-transfer resistance anymore. As
coefficient at low flow rates of methanol solution presented in Fig. discovered in our previous study,9 when the length of gas clogging
6a indicates that the increased hydrodynamic mass-transfer coeffi- exceeds a critical value, the gas clogging is removed by feed flow.
cient 共h兲 at the channel/DL interface balanced the increased mass- For higher current density, the gas-clogging exceeds the critical
transfer resistance in the DL as a result of an increase in current length more quickly, and thus with the increase of current density
density. It is also seen from Fig. 6a that at relatively high methanol the gas-clogging flow changes to slug-swarming flow without gas
flow rates, ktot decreased slightly with an increase in current density. clogging. Therefore, in the parallel flow field the effect of gas clog-
For instance, at 25.0 mL/min, ktot was reduced by about 15% when ging is lowered with the increase of current density, and the metha-
current density was increased from about 100.0 to 400.0 mA/cm2. nol transfer is improved due to the fading gas clogging, as shown in
At higher flow rates of methanol solution, the bubble removal rate Fig. 6b. In general, ktot increases somewhat and does not change
from the flow channel becomes rather high. As a result, the flow in greatly for the parallel flow field when current density is increased.
the channel approaches single-phase flow and both h and ␤ become Therefore, it is safe to conclude that the CO2 gas liberation rate has
independent of current density. Therefore, the mass-transfer resis- little effect on the overall methanol mass-transfer coefficient for
tance at the channel/DL interface remains more or less the same both types of flow fields.
with increasing current density. However, the increased gas-void It is also seen from Fig. 6 that the effect of methanol flow rate on
fraction in the DL as a result of the increased current density results ktot is significant, especially for the serpentine flow field. For the
in a lower keff in the DL. Therefore, the overall mass-transfer coef- serpentine flow field, when the flow rate was increased from
ficient ktot decreased slightly with increasing current density. 0.2 to 25.0 mL/min, the value of ktot increased from about
For the parallel flow field 共Fig. 6b兲, it can be seen that ktot in- 0.55 to 2.0 ⫻ 10−5 m/s, nearly quadruply. For the parallel flow
creased slightly with current density. The slight increase of ktot with field, the increase in ktot with increasing flow rate is relatively
current density in the parallel flow field is different with the slight smaller than for the serpentine flow field. For instance, when the
decrease in the serpentine flow field. This difference may be ex- flow rate was increased from 1.0 to 25.0 mL/min, ktot increased
plained by the different gas bubble behaviors in the different flow from about 0.55 to 0.85 ⫻ 10−5 m/s, about 1.5 times.
fields. In serpentine flow fields, except at very low current density Figure 7 presents the variation in ktot with the methanol flow rate
 Journal of The Electrochemical Society, 153 共7兲 A1358-A1364 共2006兲 A1363

Figure 7. Overall mass-transfer coefficient vs methanol flow rate for differ-

ent current densities 共temperature: 75°C兲. Figure 8. Comparison between the experimental data and Eq. 19 共solid line兲
for different flow rates in the cell with the SSFF 共methanol concentration
0.25 M; temperature 75°C; E0 = 0.75 V, bcell = 0.6 V dec−1 and C1
= 0.07 V兲.
based on the same experimental data shown in Fig. 6. It is seen that
although scattered slightly due to the influence of current density,
the measured ktot in general increased in a linear fashion with the
flow rate. Moreover, as evidenced from the slope of each curve, the tion 18 is also compared with the experimental data in Fig. 7. The
increment of the measured ktot as a result of the increased methanol maximum relative error between Eq. 18 and the experimental data is
flow rate for the serpentine flow field is much larger than for the about 13% for the serpentine flow field and 15% for the parallel flow
parallel flow field. In general, the hydrodynamic mass-transfer coef- field.
ficient h at the channel/DL interface for both flow fields increases
Mass-transport polarization.— With the overall mass-transfer
with increasing flow rate. However, as mentioned earlier, in the
coefficient given by Eq. 18, we can now predict the mass-transfer
serpentine flow field the methanol transport in the DL occurs by
polarization. To this end, we express the cell voltage as a result of
both diffusion and under-rib convection, whereas in the parallel flow
various losses as33
field the methanol is transported in the DL predominately by diffu-
sion. Therefore, the change in the overall mass-transfer coefficient is Ecell = E0 − bcell log共i兲 − Rei + C1 ln共1 − i/ilim兲 关19兴
more sensitive to a flow-rate change in the serpentine flow field than
where the equilibrium potential E0 is determined by reaction kinet-
in the parallel flow field.
ics as well as activity of species and methanol crossover, both Tafel
In order to further elaborate that the mass-transfer resistance in
constant bcell and C1 are determined by reaction kinetics, and Re is
the DL is larger than at the channel/DL interface, we now conduct a
the internal resistance. Note that the limiting current density ilim can
one-dimensional mass-transfer analysis based on Eq. 4 and 6. Based
be obtained by substituting Eq. 18 into 13, i.e.
on Eq. 4, the hydrodynamic mass-transfer coefficient at the
channel/DL interface is h = 1.40 ⫻ 10−5 m/s, while based on Eq. 6 6F共a + bQ兲 ⫻ 10−6
the mass-transfer coefficient in the DL is keff = 0.86 ⫻ 10−5 m/s ilim = Cin 关20兴
A共a + bQ兲 ⫻ 10−6
共assuming porosity is 0.7 and thickness of DL is 0.35 mm兲. The 1+
relative smaller value of the mass-transfer coefficient in the DL 2Q
means that the main methanol transfer resistance occurs in the DL. It Taking the cell with the serpentine flow field fed with 0.25 M
should be recognized that in practice the flow in the channel is not a methanol solution as an example, different flow rates of methanol
single-phase fully-developed laminar flow, and the liberation of gas solution were tested, leaving all other operating parameters constant.
bubbles disturbs the boundary layer greatly and enhances the metha- One set of parameters were assumed representing the reaction kinet-
nol transfer, leading to a value much higher than 1.40 ⫻ 10−5 m/s. ics of all tests for simplicity. Substituting the parameters and Eq. 20
In the DL, as seen from Fig. 1, due to the coverage of the flow into 19, we can obtain the polarization curves for different flow rates
channel ribs, the path for methanol diffusion is lengthened, and the of methanol solution. The predicted polarization curves and the ex-
mass-transfer coefficient in the DL should be less than 0.86 perimental data are compared in Fig. 8. The predicted mass-transfer
⫻ 10−5 m/s. Therefore, it is reasonable to conclude that the polarization for each flow rate is in fairly good agreement with the
methanol-transfer resistance is relatively smaller at the channel/DL experimental data.
interface than in the DL.
Based on the above analysis and the comparison in the measured Conclusion
ktot between the serpentine flow field and the parallel flow field Mass transport of methanol at the anode of a DMFC has been
shown in Fig. 7, one can conclude that the more significant increase analyzed and it was found that the overall mass-transfer coefficient
in the overall methanol-transfer coefficient ktot as the result of the can be determined by measuring the cell limiting current density.
increased flow rate of methanol solution in the serpentine flow field Experiments were then carried out in an in-house-fabricated DMFC
mainly results from the enhanced under-rib convection. with an active area of 3.0 ⫻ 3.0 cm and equipped with either a
Because current density has little effect on ktot in both flow fields, serpentine flow field or a parallel flow field at the anode for various
as an approximation, the measured ktot can be correlated in terms of methanol concentrations and methanol flow rates. Salient findings
the methanol flow rate Q. The following correlation was obtained and conclusions are summarized as follows:
based on a least-squares fit of the experimental data shown Fig. 7 for
the DMFC hardware used in this work 1兲 The overall mass-transfer resistance of methanol transport
from the flow channel to the catalyst layer is composed of the hy-
ktot = 共a + bQ兲 ⫻ 10 −6
共m/s兲 关18兴
drodynamics mass-transfer resistance at the channel/DL interface
where a = 8.373 and b = 0.478 for the serpentine flow field, and the mass-transfer resistance in the DL. The experimental results
whereas a = 6.1861 and b = 0.0921 for the parallel flow field. Equa- showed that the latter was much larger than the former.
A1364 Journal of The Electrochemical Society, 153 共7兲 A1358-A1364 共2006兲

2兲 The experimental results showed that the overall mass- 4. Z. H. Wang and C. Y. Wang, J. Electrochem. Soc., 150, 508 共2003兲.
5. Z. Qi and A. Kaufman, J. Power Sources, 110, 177 共2002兲.
transfer coefficient in the serpentine flow field decreased slightly 6. A. Z. Weber and J. Newman, Chem. Rev. (Washington, D.C.), 104, 4679 共2004兲.
with the increase of current density but it increased slightly with 7. Q. Ye, T. S. Zhao, H. Yang and J. Prabhuram, Electrochem. Solid-State Lett., 8,
current density in the parallel flow field. Generally, it can be con- A52 共2005兲.
cluded that the overall mass-transfer coefficient is nearly indepen- 8. H. Yang, T. S. Zhao and Q. Ye, J. Power Sources, 139, 79 共2005兲.
dent of current density, particularly at low flow rates of methanol 9. C. W. Wong, T. S. Zhao, Q. Ye, and J. G. Liu, J. Electrochem. Soc., 152, A1600
共2005兲.
solution. 10. G. Q. Lu and C. Y. Wang, J. Power Sources, 134, 33 共2004兲.
3兲 It has also been found that the significantly increased overall 11. J. C. Amphlett, B. A. Peppley, E. Halliop, and A. Sadiq, J. Power Sources, 96, 204
methanol-transfer coefficient in the serpentine flow field as the result 共2001兲.
of increased methanol flow rate can mainly be attributed to the en- 12. K. Scott, W. M. Taama, S. Kramer, P. Argyropoulos, and K. Sundmacher, Electro-
chim. Acta, 45, 945 共1999兲.
hanced under-rib convection. Therefore, it is critical to obtain an 13. C. W. Wong, T. S. Zhao, Q. Ye, and J. G. Liu, J. Power Sources 155, 291 共2006兲.
optimal design of the flow channel ribs and the DL to enhance the 14. H. Yang, T. S. Zhao, and Q. Ye, Electrochem. Commun., 6, 1098 共2004兲.
under-rib convection. 15. P. Argyropoulos, K. Scott, and W. M. Taama, Electrochim. Acta, 44, 3575 共1999兲.
4兲 Based on the above findings, the overall mass-transfer coef- 16. M. M. Mench, S. Boslet, S. Thynell, J. Scott, and C. Y. Wang, in Direct Methanol
Fuel Cells, S. Narayanan, T. Zawodzinski, and S. Gottesfeld, Editors, PV 2001-4,
ficient can be correlated in terms of the flow rate of methanol solu- p. 241, The Electrochemical Society Proceedings Series, Pennington, NJ 共2001兲.
tion. Therefore, two respective correlation equations for predicting 17. T. Bewer, T. Beckmann, H. Dohle, J. Mergel, and D. Stolten, J. Power Sources,
the overall mass-transfer coefficient for the serpentine and parallel 125, 1 共2004兲.
flow fields have been developed. 18. K. W. Lum and J. J. McGuirk, J. Power Sources, 143, 103 共2005兲.
19. A. Kumar and R. G. Reddy, J. Power Sources, 113, 11 共2003兲.
5兲 It has been shown that the predicted mass-transport polariza- 20. Y. G. Yoon, W. Y. Lee, G. G. Park, T. H. Yang, and C. S. Kim, Electrochim. Acta,
tions based on the correlation equation of the overall mass-transfer 50, 709 共2004兲.
coefficient were in fairly good agreement with the experimental 21. J. Scholta, G. Escher, W. Zhang, L. Küppers, L. Jörissen, and W. Lehnert, J. Power
data. Sources 155, 66 共2006兲.
22. M. Grujicic and K. M. Chittajallu, Chem. Eng. Sci., 59, 5883 共2004兲.
23. M. Prasanna, H. Y. Ha, E. A. Cho, S.-A. Hong, and I.-H. Oh, J. Power Sources,
Acknowledgments 131, 147 共2004兲.
24. M. V. Williams, E. Begg, L. Bonville, H. R. Kunz, and J. M. Fenton, J. Electro-
The work described in this paper was fully supported by a grant chem. Soc., 151, 1173 共2004兲.
from the Research Grants Council of the Hong Kong Special Ad- 25. G. Inoue, Y. Matsukuma, and M. Minemoto, J. Power Sources, In press.
26. J. G. Pharoah, J. Power Sources, 144, 77 共2005兲.
ministrative Region, China 共project no. HKUST6101/04E兲. 27. Y. Wang and C. Y. Wang, J. Power Sources, 147, 148 共2005兲.
Hong Kong University of Science and Technology assisted in meeting the 28. P. H. Oosthuizen, L. Sun, and K. B. McAuley, Appl. Therm. Eng., 25, 1083 共2005兲.
29. C. Y. Soong, W. M. Yan, C. Y. Tseng, H. C. Liu, F. L. Chen, and H. S. Chu, J.
publication costs of this article. Power Sources, 143, 36 共2005兲.
30. A. Kazim, P. Forges, and H. T. Liu, Int. J. Energy Res., 27, 401 共2003兲.
References 31. M. V. Williams, H. R. Kunz, and J. M. Fenton, J. Electrochem. Soc., 151, 1617
1. J. Larminie and A. Dicks, in Fuel Cell Systems Explained, p. 1, John Wiley & 共2004兲.
Sons, Ltd, England 共2003兲. 32. G. J. Hwang, Y. C. Cheng, and Y. W. Chiang, Int. J. Heat Mass Transfer, 40, 481
2. C. Y. Wang, Chem. Rev. (Washington, D.C.), 104, 4727 共2004兲. 共1997兲.
3. X. Ren, P. Zelenay, S. Thomas, J. Davey, and S. Gottesfeld, J. Power Sources, 86, 33. P. Argyropoulos, K. Scott, A. K. Shukla and C. Jackson, J. Power Sources, 123,
111 共2000兲. 190 共2003兲.