Journal of Power Sources 288 (2015) 26e35

Contents lists available at ScienceDirect

Journal of Power Sources

j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / lo c a t e / j p o w s o u r

Sea water magnesium fuel cell power supply

Robert Hahn a, *, Jan Mainert a, Fabian Glaw a, K.-D. Lang b
a Fraunhofer IZM, Gustav-Meyer-Allee 25, D-13355 Berlin, Germany
b
Technische Universitat€ Berlin, TIB4/2-1, Gustav-Meyer-Allee 25, 13355 Berlin, Germany

highlights

Hydrogen is produced reproducibly in galvanic sea water cell by anodic dissolution of magnesium. Use of hydrogen
evolving Raney nickel cathodes with gas separation membrane.
Complete hydrogen generation fuel cell power supply demonstrated. Advantages of
magnesium AZ31 alloys demonstrated.
1200 Wh/kg, 3000 Wh/l electricity achieved from magnesium alloy sheets as the fuel.

article info abstract

Article history: An environmentally friendly magnesium fuel cell system using seawater electrolyte and atmospheric oxygen was tested under
Received 28 January 2015 practical considerations for use as maritime power supply. The hydrogen rate and therefore the power density of the system
Received in revised form were increased by a factor of two by using hydrogen evolution cathodes with a gas separation membrane instead of submerged
16 April 2015
cathodes without gas sepa-ration. Commercial magnesium AZ31 rolled sheet anodes can be dissolved in seawater for
Accepted 18 April 2015
Available online
hydrogen production, down to a thickness below 100 mm thickness, resulting in hydrogen generation efficiency of the anode
of over 80%. A practical specific energy/energy density of the alloy of more than 1200 Wh/kg/ 3000 Wh/l was achieved when
coupled to a fuel cell with atmospheric air breathing cathode. The per-formance of several AZ31 alloy anodes was tested as
Keywords:
Hydrogen generation well as the influence of temperature, electrolyte concentration and anode e cathode separation. The excess hydrogen produced
Magnesium alloy by the magnesium hydrogen evolving cell, due to the negative difference effect, is proportional to the cell current in case of
Fuel cell the AZ31 alloys, which simplifies system control considerably. Stable long-term operation of the system was demonstrated at
Anodic dissolution low pressures which can be maintained in an open-seawater-submerged hydrogen generator.
Gas diffusion electrode
Gas separation membrane
Energy density
Power density
© 2015 Elsevier B.V. All rights reserved.

1. Introduction
In response to the need for improved electrochemical energy storage in
mobile, stationary and portable applications, magnesium has recently
attracted new interest [1] as an anode material due to its high discharge
activity, large capacity and small environmental impact. Magnesium has a
negative standard electrode potential of 2.37 V (vs. SHE) which is more
negative than for zinc and aluminum. At 2.205 Ah/g, the Faradaic capacity of
magnesium is significantly higher than that of zinc. Its specific weight is lower
* Corresponding author.
E-mail addresses: robert.hahn@izm.fraunhofer.de (R. Hahn), fabian.glaw@izm.
fraunhofer.de (F. Glaw), klaus-dieter.lang@izm.fraunhofer.de (K.-D. Lang).
than that of aluminum and about 25% of that of zinc. Thus mag-nesium has
the potential for very high energy density and light-weight batteries.
Mg anodes are active in neutral aqueous electrolytes that contain chlorine
or other aggressive ions; thus the most important applications of magnesium
anodes so fare are seawater activated batteries and dissolved-oxygen seawater
batteries that were developed several decades ago. In these applications some
of the mayor drawbacks of magnesium battery anodes like the high self-
discharge rate and electrode clogging with reaction products (deposition of
Mg(OH)2 at the Mg surface) can be handled. On the other hand, technologies
for closed battery systems and secondary batteries based on Mg anodes, such
as magnesium-air [2] or magnesium-ion [3] batteries are still in their infancy,
but promise a huge market potential.

http://dx.doi.org/10.1016/j.jpowsour.2015.04.119 0378-
7753/© 2015 Elsevier B.V. All rights reserved.
 R. Hahn et al. / Journal of Power Sources 288 (2015) 26e35
Dissolved-oxygen batteries with Mg anodes have been devel-
oped for deep-water applications [4,5] and proven for long time
operation. Due to the low oxygen concentration in the sea water,
corresponding cathodes have a much larger surface than the
magnesium anode. In practical constructions, the distance be-
tween anode and cathode is relatively large which results in high
cell impedance and low current capacity. Aluminum anodes have
already been considered for operation of cathodes in the hydrogen
generation mode [6]. In case of hydrogen evolution cathodes the
current density is increased by one order of magnitude if the
separation between anode and cathode is kept at a constant low
distance during the discharge process, but the operating cell
voltage is very low (0.1 … 0.4 V). Therefore in the hydrogen evo-
lution mode, the power density can be increased but the energy
density is low due to the low cell potential. Most of the energy is
represented by the produced hydrogen which is released into the
sea.
On the other hand previous work indicates that a practical en-
ergy density near that of metal air batteries can be achieved at a
moderate power density if both the electrical power and the
hydrogen of a hydrogen evolving cell are used in combination with
a fuel cell [7e9]. Portable power supply demonstrators in the
1e4 W range [7] and power supplies with milliwatt output for
sensor applications [9] were tested. A similar approach was
investigated for aluminum hydrogen generation cells [10,11].
Hydrogen and electricity production in these closed cells shows
promising results, and additional studies are required to further
develop this concept. For instance, neither the influence of
continuously increasing electrolyte concentration during the
discharge due to water consumption, nor the performance at
higher currents and high hydrogen pressure is not fully understood
[12].
In this study, we combine the well proven magnesium sea
water discharge performance in the hydrogen evolution state,
with hydrogen utilization in a fuel cell [13], into a magnesium
fuel cell system. In contrast to dissolved-oxygen batteries, here
the high oxygen concentration of the air can be combined with
the high hydrogen generation rate of magnesium in sea water.
Due to the infinitely large electrolyte reservoir, reaction products
are continuously flushed from the cell and the electrolyte con-
centration remains constant, leading to long term stable
operation.
Power supplies of this type can on the one hand only be used
near sea level, but on the other hand they provide advantages over
other types of batteries in high energy density and low environ-
mental impact. Possible applications are sea based remote sensors
and signal buoys as well as auxiliary electrical power supplies for
sailing boats and leisure activities. Further developments may lead
to larger applications like power generation for submarine systems
in combination with additional oxygen sources or large storage
applications in combination with magnesium recycling.
The main reactions of the hydrogen evolution cell are:
Mg þ 2OH /MgðOHÞ2 þ 2e ðanodicÞ (1)
2H2O þ 2e /H2 þ 2OH ðcathodicÞ (2)
Mg þ 2H2O/MgðOHÞ2 þ H2ðoverallð1Þ þ ð2ÞÞ (3)
In addition to the main reactions according to equations (1)e(3),
a self-discharge reaction or hydrogen side reaction results in
additional hydrogen evolution directly at the anode or inside the
electrolyte. Impurities in the magnesium alloy anode may lead to
local corrosive hydrogen evolution directly at the metal alloy ac-
cording to (3) at a rate rH2-corr. In case of Mg anodes additional
27
hydrogen evolution reaction occurs as a function of the applied
potential which is called the negative difference effect (NDE) as
discussed in section 3.2. It is assumed that the anodic reaction does
2
not only lead to bivalent Mg þ ions according to (1) but also to
univalent magnesium ions Mgþ [17]. Thus hydrogen can be pro-
duced from a chemical reaction according to equation (4) as well.
þ 2þ
2Mg þ 2H2O/2Mg þ 2OH þ H2ðchemicalÞ (4)
Therefore, the resulting hydrogen rate rH2 is the sum of the
cathodic rH2-cat (1) and the chemical rH2-chem (4) and corrosive re-
action as indicated in equation (5).
r ¼r þr þr
H2 H2 cat H2 chem H2 corr (5)
The overall reaction of the hydrogen production inside the
hydrogen evolution cell and the fuel cell reaction with ambient
oxygen is shown in equation (6). As the water and oxygen are
available in unlimited amounts from the sea and ambient air
respectively, energy density calculation can be made which con-
siders only the Mg alloy as energy supply.
2Mg þ 2H2O þ O2/2MgðOHÞ2 (6)
The basic concept of the magnesium hydrogen fuel cell system is
shown in Fig. 1.
The galvanic hydrogen generation cell is submerged in the
seawater electrolyte. The planar magnesium anode is arranged
between two gas diffusion electrodes with hydrogen evolution
cathode catalyst and gas separation membranes. Additional
hydrogen which is generated directly at the anode or inside the
chamber can be collected at the middle hydrogen connecting port.
The reaction cell has openings at the bottom for sea water ex-
change and removal of reaction products. The collected hydrogen
is fed into the anode of a fuel cell above sea level which uses the
ambient air for cathode supply. The hydrogen evolution cells are
connected in parallel as they are all in contact with the same
electrolyte. The current of the hydrogen cell is controlled by the
DC/DC converter to produce adequate hydrogen flow which is
required for the fuel cell operation. The maximum hydrogen
pressure is low and depends on the submersion depth of the
hydrogen evolution cell. Therefore the fuel cell must be optimized
for low pressure operation.
2. Experimental
The experiments are divided into two stages. First, the general
influence of material parameters, temperature and electrolyte
concentration was investigated in a small flow cell were the
catalyst cathode was directly incorporated into the electrolyte
flow (Fig. 2a). Here, the Mg anode represents one sidewall of the
flow cell. Therefore it is not possible to fully dissolve the anode.
The discharge is stopped at the moment of the first occurrence of a
hole which results in electrolyte leakage. Second, a larger cell was
designed with two cathodes on both sides of the Mg anode to
study the influence of different Mg alloys and test the hydrogen
evolution gas diffusion electrodes (GDEs) with hydrogen separa-
tion membrane which were previously developed for the zinc
hydrogen evolution cells [7] (Fig. 2b). The separation between
anode and cathodes was varied by using anodes of different
thickness.
The hydrogen of the small cell was extracted at the outlet of the
electrolyte storage container. Most of the hydrogen of the larger
flow cell was extracted in the hydrogen collection chambers behind
the gas separation membranes of the cathode GDE on both sides.
The hydrogen which was produced at the anode and inside the
28 R. Hahn et al. / Journal of Power Sources 288 (2015) 26e35

Fig. 1. Concept of Magnesium hydrogen fuel cell power supply. 1) bottom opening for sea water exchange and removal of reaction products, 2) hydrogen connection to fuel cell, 3) gas liquid
separator, 4) hydrogen gas filter, 5) air filter, 6) secondary battery as electrical buffer.

electrolyte was taken from the electrolyte storage chamber similar to the
smaller cell set-up.
2.1. Test bed and electrochemical characterization
To simulate seawater operation, electrolyte from a 10 l brine
container was pumped at a flow of 5 l/min (pump: Eheim 1048) through the
flow-cells. Sea salt was taken from aquarium supply and dissolved in
deionized water. A salt concentration of 4 wt % was used for all experiments,
while in a special test saturated salt concentration was used. Anode and
cathode were connected to a resistive load or to a Basytec ® battery tester for
constant voltage,

Fig. 2. Experimental set-up and hydrogen evolving cells a) small cells with 20 cm2 submerged cathode b) larger flow cell with two 90 cm2 cathode GDEs and gas separation membrane on both sides
of the magnesium anode.
 R. Hahn et al. / Journal of Power Sources 288 (2015) 26e35 29

step-current and V/I scan measurements. The hydrogen rate was measured
with MKS flow meters. The measurement data of the electrical parameters,
flow and temperature were collected with help of a PC measurement unit. The
following test procedures were used:
1) 30 s measurement of open circuit potential,
2) Potentiostatic voltage reduction at 10 mV/s down to 300 mV,
3) 30 min at 300 mV to consume the oxygen dissolved in the system. At the
end of this step the voltage is reduced from typically 1.4 V to 0.6 … 0.7 V.
4) Constant current steps (each 10 or 20 min) with increasing current (20 mA
for small and 100 mA for large cells) until the cell voltage drops below 50
mV. This method is used to measure the hydrogen rate as function of the
outer current.
5) Polarization curves were measured with help of voltage scans at 10 mV/s
down to 50 or 100 mV.
6) Long term discharge at 50 or 100 mV. In case of the larger cells, step 3
and 4 were inserted into the constant voltage discharge every 4e5 Ah
discharge.
2.2. Electrolyte flow cell
The small flow cells are assembled using a PVC frame. Mg-anode and
stainless steel cathode holders are attached on both sides by elastomer
gaskets. The distance between anode and cathode is 11 mm at the beginning
3
of discharge. The electrolyte volume of the small cell is 60 cm . For
comparative purposes closed cells with no fluid connections were also
assembled. Here the separation be-tween anode and cathode was 7 mm which
3
results in a total elec-trolyte volume of 30 cm .
2 The AZ-alloys are characterized by a low self-corrosion rate and moderate
The large flow cells with 180 cm cathode area comprise a complete PVC
housing. The cathode GDEs are attached face to face at both sides of the
chamber at a distance of 17 mm. The gas sep-aration membranes are sealed
by a pressure sealing and adhesive to allow for an electrolyte-tight separation
of the hydrogen generated in the GDEs. The volume of the flow cell is 150
3
cm which is completely filled with the center anode and electrolyte. Silver
wire is used to connect the anodes. The wires and the interconnections are
sealed by a polymer film. The distance between anode and cathode at the
beginning of discharge is 2 mm (A1, A2), 7.75 mm (A3), 8 mm (A4) and 8.5
mm (A5) respectively. Polymer spacers are used to keep the anode in the
center of the cell during the discharge.
2.3. Cathodes
For the set-up according to Fig. 2a cathode GDEs were clamped between
two stainless steel grids.
For the setup according to Fig. 2b the hydrogen evolution cathode consists
of a three layer laminate as already tested with Zn cells [7]. The porous GDE-
includes Raney-nickel catalyst, poly-tetrafluoroethylene (PTFE) binder and a
nickel mesh current col-lector; it is the core of the system. A Teflon® based
gas separation
membrane is laminated onto the GDE at the outer side, in order to prevent gas
accumulation in form of gas bubbles at the catalyst electrolyte interface which
reduces the operating catalyst surface and therefore reduce the cathode current
density.
The separation layer is non-permeable for the electrolyte but allows
hydrogen to exit the cell. These electrodes which are already used in
hydrogen evolution coin cells for lubricant dispensers, were supplied by
VARTA micro battery [14]. For ease of handling, the GDE multi-layer foil is
mounted in a stainless steel frame (Fig. 2b). The PTFE gas separation
membrane is sealed against the metal frame with help of a sealing paste and
the metal frame is sealed in the PVC housing with help of gaskets. The
2
cathode area is 20 and 90 cm (on both sides) for the small and the large cells
respectively.
2.4. Anodes
The goal of this study is not only to come up with a high energy density
power supply but also to achieve an environment-friendly low-cost system.
Thus, only industrial proven and widely available Mg alloys were tested. Hg
and Pb containing materials are excluded from this investigation although it is
known that Mg-Ga-Hg [15] and MgePb [16] alloys show superior
performance for sea water activated batteries, such as higher discharge
voltage and lower self-corrosion potential. The dimensions of the anodes are
summarized in Table 1. Only AZ-alloys with the alloying elements Al and Zn
were tested. Table 2 contains information about the elemental concen-trations
and trace elements. Neat Mg foil 3N (Goodfellow®) was included in the tests
as a reference.
The alloying elements minimize self-discharge and improve discharge
performance by acceleration of the self-peeling of discharge products from
the metal surface [1].
discharge performance. Therefore they are mainly used for long term low
power dissolved oxygen batteries but not for sea water activated batteries.
The included aluminum plays an impor-tant role to inhibit the self-discharge
whereas zinc favors uniform dissolution of the magnesium [1]. The alloy
consists of the following metallurgical components: the a-phase which is a
MgeAleZn solid solution, the b-phase Mg17Al12 which acts pri-marily as
corrosion barrier, and Al and Zn grains [17]. In addition AleMn particles
serve as strong local cathodes to promote the discharge process [18].
For anode performance, not only the concentration of the alloying
elements, but also the micro structure and the processing of the materials were
proven critical. Thus, the use of hot rolling and subsequent annealing alters
the performance of AP65 alloys [19]. The observed stronger discharge
activity and utilization effi-ciency is attributed to fine grains, low density of
crystalline defects and good compositional homogeneity of the magnesium
matrix. Hot rolling and annealing increases the corrosion current density and
accelerates the Mg dissolution which is important for low activation times of
sea water activated batteries but may be detri-mental for a long-life power
supply. To study the influence of metallurgy we compared the behavior of
cast alloys (A1, A2) with rolled alloy sheets (A3, A4).

Table 1
Dimensions of tested magnesium anodes.

No. Alloy Supplier L w t [mm3] Surface [cm2] Mass [g] Nominal capacity [Ah]
A1 AZ31 e 80 60 10 e 87 191
A2 AZ31 Grillo 80 60 13 132.4 113.3 248
A3 AZ31 Grillo 98 70 1.5 142.2 19.5 43
A4 AZ31B Thyssen Krupp 109 69 1 149.0 13.6 30
A5 Mg 3N Goodfellow 100 75 0.5 146.0 6.3 14
30 R. Hahn et al. / Journal of Power Sources 288 (2015) 26e35

Table 2 occurred after 10 and 14 days for the saturated and brine electrolyte
Concentration of alloying elements of the tested magnesium metal and alloys ac-cording to respectively.
ASTM B90-93.
The experiment of Fig. 3a was conducted at room temperature and
No. Al Zn Mn Si Fe Cu Ni Ca Other changes of current density due to daily temperature variations in the
A2 2.6e3.5 0.7e1.4 laboratory are clearly visible in the measurement curves. The large
A4 2.5e3.5 0.6e1.4 0.2e1 0.1 0.005 0.05 0.005 0.04 0.3 temperature influence on the polarization curves was measured in a separate
A5 0.007 <0.002 0.017 0.005 0.028 0.002 <0.001 test with a heated flow cell and heated electrolyte (Fig. 4a).

In contrast to sea water application with unlimited amount of electrolyte,

3. Results and discussion
2 the Mg(OH)2 solubility limit was reached soon. To evaluate the influence of
3.1. Small cells with 20 cm submerged cathode
To demonstrate the importance of electrolyte flow, ano-deecathode
separation and electrolyte concentration, the discharge behavior of flow cells
were compared with closed cells at 4 wt % and saturated electrolyte
concentration (Fig. 3a). The salt concentration of sea water is between 3.5 and
3.8 wt%. At constant voltage 50 mV discharge, the current density at the start
is much higher for the closed cells with only 7 mm electrodes spacing (vs. 11
mm of the flow cells), and higher for higher electrolyte con-centrations. But
the current of the closed cells decreases rapidly within hours and becomes
much lower than the current density of the flow cells after 48 h. In saturated-
electrolyte case there is also a significant current reduction during the first 10
h but the current stabilizes during the remainder of the discharge. The lowest
current reduction was observed in case of the flow cell operated with 4 wt %
electrolyte concentration.
The reason for the current reduction are the dissolved Mgþþ cations which
precipitate on the Mg anode surface in the form of a Mg(OH)2 film when the
2 (Fig. 4a), used cathodes and anodes were replaced by new ones during the
space near the electrode surface is satu-rated with these Mg þ ions. This
happens much quicker in the closed cells. In the case of the flow cell the
electrolyte flow is suf-ficient to prevent clogging of the cell at 4wt %
electrolyte concen-tration. In saturated-electrolyte case, reaction products
accumulate at the anode even for the flow cells. Reaction products were
verified after disassembly of the cells, they result in a more pronounced
current decrease during discharge. Nevertheless it was possible to discharge
both flow cells until the metal was corroded throughout and the experiment
stopped due to electrolyte leakage. This
we used an electrolyte container of only 10 L. As a result the pH value of the
electrolyte increased to a value of ca. pH10 during the first discharge hour and
pH value and Mg(OH)2 concentration in a special experiment, the spent
electrolyte was replaced after 6 Ah discharge. There was no significant
difference between saturated Mg(OH)2 and fresh electrolyte as can be seen in
Fig. 4b. This finding is in agreement with the literature. It was shown that the
corrosion resistance of Mg only changes if the pH value increases above 10.5,
which corresponds to the pH of satu-rated Mg(OH)2 [20] and which was not
the case during our ex-periments. The reason is, that during Mg dissolution
the pH in the electrolyte layer near the Mg electrode can have a value near
pH10 even if the bulk solution has a pH as low as pH 4 [21]. In other words:
the Mg(OH)2 concentration near the Mg anode is high, even at the start of
discharge, independent of the concentration of the surrounding electrolyte.
The high open circuit potential of the V/I curve at start of discharge (t ¼
0) in Fig. 4b may be due to residual oxygen inside the cell.
To find the reason of the performance decline at the start of discharge
discharge of the next experiment. As can be seen in Fig. 4c a new cathode
does not change the performance (short circuit current of ca. 80 mA) of a used
anode (1.4 Ah, step 2 in Fig. 4c) while the start current was achieved again
with a new anode and the used cathode (step 3 in Fig. 4c). This is a clear
indication that performance reduction is due to the anode. The build-up of a
passivation layer on the magnesium anode with reduced ion transport
properties may be the reason for the rapid reduction of the maximum current
at the start of discharge.

Fig. 3. Discharge curves of anode A1 in small cell with 20 cm2 cathode. Comparison of the discharge current at 50 mV of closed and electrolyte flow cells using 4wt % sea water and saturated
electrolyte (a) and the corresponding temperature, current, hydrogen flow and the ratio of practical and theoretical hydrogen flow of the flow cell in sea water electrolyte (b).
 R. Hahn et al. / Journal of Power Sources 288 (2015) 26e35 31

Fig. 4. V/I curves of small cells with anode A1 (aec) and large Mg hydrogen evolving cells with anode A3 (d). Influence of temperature (a), influence of depth of discharge with fresh and used
electrolyte (b), replacing used anode and cathode by new ones during the course of discharge (c) and performance improvement over time which was obtained for anode A3 (d).

Concerning the cathode it was proven in practice, that the same catalyst was significantly higher in case of 4 wt % electrolyte concentration (Table 4).
GDEs can be used for the discharge of many Mg alloy sheets and several
thousand operational hours. This excellent long term behavior may be due to
the large amount of cost-efficient Ni catalyst.
3.1.1. Hydrogen generation
In both seawater-activated and seawater dissolved oxygen bat-teries,
In an attempt to find a possible countermeasure against the anode surface
generation of hydrogen leads to unwanted energy loss. Here such hydrogen is
layer, in one experiment the electrolyte flow was largely increased to purge
the main source of energy and can be transferred into electricity in the fuel
the flow cell. In that case a small improvement of the maximum current was
cell. Nevertheless corrosive hydrogen production which is not controlled by
achieved after 18 Ah discharge and electrolyte purge. This indicates that even
the cathode current may lead to unpredictable system control and self-
in the flow cells some reaction products accumulate between anode and
discharge. Therefore three cases of hydrogen generation were studied:
cathode which may be removed by forced electrolyte flow.

The polarization curves of sea water and saturated electrolyte solutions

1. Hydrogen generation of the anodes in electrolyte solution (Table 3),
have also measured as a function of depth of discharge. In both cases the
current is lower for higher ampere hours but nearly twice the current can be
2. Hydrogen generation in the Mg flow cell including GDE catalyst but zero
achieved in case of saturated electrolyte. On the other hand the anode
current flow (Table 3),
efficiency of the complete discharge
3. Hydrogen generation as function of current density (Fig. 5).

Table 3
Passive, corrosive hydrogen generation (*T ¼ 20 C, reaction time 72 h).

Anode Measurement condition sccm ml/(g*day) ml/(cm2*day)

2
A1 20 cm anode in electrolyte 0.2 e 14300
A1 in cell according to Fig. 2a with cathode 0.4 e 28700
A3 1 cm2 chip* after 20 Ah discharge in cell according to Fig. 2b e 3565 350
A3 1 cm2 chip* after 5 Ah discharge in cell according to Fig. 2b e 2902 290
32 R. Hahn et al. / Journal of Power Sources 288 (2015) 26e35

Table 4
Summary of Ah balance of the Mg alloy anode discharge experiments.

Cell type Small flow cells Large flow cells

Figures Fig. 3a,b Fig. 3a e Fig. 4d e e

Electrolyte concentration 4 wt.% Saturated 4 wt.%
Anode A1 A1 A2 A3 A4 A5
Current integral [Ah] 34.8 30.1 65.3 25.4 14.5 6.1
Hydrogen integral [Ah] 44.4 38.6 79.7b 31b 17.7b 7.4b
Percentage of hydrogen not generated at the cathode 21 %a 22 %a 22 %b 22 %b 22 %b 22 %b
Anode mass consumed [Ah] 54.4 72.1 e 43.02 30.02 0.49
Anode thickness consumed [mm] 6 7.6 4.17 1.47 0.98 0.49
anode efficiency for hydrogen production 81.6% 56.7% e 81.5% 64.6% 53.5%
Specific energy density of complete anodec [Wh/kg] 1045d 908d 976d 1271 1041 945
Energy density of complete anode [Wh/l] 2220 1930 2075 3009 2359 1985
a Mean value of total discharge.
b
Assuming the same relation measured with the small flow cells.
c
Assuming a sum voltage of fuel cell and hydrogen cell of 0.8 V.
d
For anode thickness of 8 mm.

In a first experiment the hydrogen evolution of Mg-alloy anode was
measured in the small cell with cathode and without cathode. It can be seen
from Table 4, that in case of A1 anodes the hydrogen evolution is doubled if
the catalyst cathode is present, even with no outer current applied. This
clearly indicates a multistep reaction of anode dissolution that incorporates
chemical reactions of hydrogen release inside the electrolyte or at the catalyst.
The hydrogen generation of anode chips of Mg alloy A3 was tested in an
industrial measurement set up for Zn pastes applied in primary batteries. Here
the hydrogen generation rate is much lower in comparison to A1 but still 100
times higher compared to the typical Zn anode material. It was also shown
that the gas evolution of the A3 anodes is higher if they are discharged for 20
Ah in the large flow cells, than if anodes are operated for only 5 Ah. This may
be due to the larger active surface area of the discharged material.
between generated hydrogen and the outer current is stable at ca. 140% but
shows a strong dependence on temperature. The large increase of the
corrosive hydrogen production with temperature is well known for
magnesium sea water batteries [1].
The hydrogen and current integral of the flow cell experiments of Fig. 3a
are summarized in Table 3 together with the results of the large flow cells.
With these data the hydrogen production was only ca. 21e22 % higher than
the amount of hydrogen obtained from the integral of the measured current
and Faraday's Law. We currently have no clear interpretation for this
discrepancy with the galva-nostatic step curve measurements.
2
3.2. Large flow cells with 180 cm cathode and hydrogen separation
membrane

3.2.1. Hydrogen generation

Current step curves were used to measure the hydrogen generated during
cell operation as function of the outer current.
The hydrogen flow as function of the electrical current was measured for
all cell configurations. The percentage of hydrogen flow in relation to the
hydrogen flow that would be expected ac-cording to the cathode current and
Faradays law showed some variation between the cells but at higher currents
the real hydrogen flow is ca. 40 %e50 % higher compared to the outer current
for all cells. Fig. 3b shows the hydrogen flow during a long term test with
large ambient temperature variations. Here again, the relation
Similar to the hydrogen generation measurements of the small cells
according to Fig. 5, the hydrogen generation characteristic of the large flow
cells is shown in Fig. 6 for the anodes A2-A5.
In all cases, the hydrogen rate is higher than the value which would be
expected according to Faraday's Law calculation. But here again, this higher
value is not a constant offset which might be expected from Table 3; instead,
it increases with current. This is a well-known electrochemical anomaly of
magnesium and its alloys, and it is called the negative difference effect
(NDE). The NDE was first explained for neat magnesium with help of a
model which assumes a partially protective film on the electrode surface [21]
and covers two aspects:

1) The difference between the spontaneous rate of hydrogen evo-lution

reaction on the Mg surface and the one which corre-sponds to the
measured hydrogen rate for the applied galvanostatic current is negative.

2) The dissolution current of the anodic magnesium increases faster than

expected.

It is assumed that the Mg oxidation results in both the univalent (Mgþ)

and the bivalent (Mgþþ) ions.
The surface-film NDE model was further extended to magne-sium alloys
as well [20]. A first attempt was made to describe the NDE behavior
quantitatively by supposing that the amounts of Mgþ and Mgþþ ions change
with the variations of the over potential [17]. As the relationship between the
rates of Mgþ and Mgþþ formation is not known, it can't be used to calculate
the real hydrogen rate as functions the relevant parameters like alloy
composition and temperature.
Fig. 5. Hydrogen flow as function of electrical current of the small closed cell for anode A1 at 4
wt % electrolyte concentration. Our main findings are:
 R. Hahn et al. / Journal of Power Sources 288 (2015) 26e35
Fig. 6. Hydrogen flow as function of current for anodes A2-A5 in large flow cells (a) and plotted in relation to the hydrogen generation which would be expected according to Faraday's Law
calculation with the outer current (b).
- For all investigated AZ alloys, the hydrogen rate is proportional to the cell
current but 20e50 % larger in relation to the hydrogen generation which
would be expected according to Faraday's Law calculation with the cell
current.
- Neat magnesium shows a different behavior. Here the hydrogen
generation at low current is larger by a factor of 3 compared to Faraday's
Law, but decreases at high currents below that of the magnesium AZ31
alloy. Current and hydrogen rate are not proportional.
- The hydrogen generation is strongly dependent on temperature but varies
little with change of electrolyte concentration and pH value.
- The relationship between hydrogen flow and current does not change with
depth of discharge.
- The hydrogen generation of the die cast block anode is slightly lower
compared to the rolled metal sheet anodes.
- The hydrogen evolution behavior is the same in flow cells with submerged
cathodes and cathodes with hydrogen separation membrane.
This stable relation between current and hydrogen rate under practical
operation conditions depends only on the type of anode material and the
temperature and facilitates the system control.
3.2.2. Discharge performance
The long term discharges of the large flow cells for the anodes A3 - A5
resulted in more reproducible results in comparison to the small cells because
they are not serve as part of the cell housing and can be discharged nearly
completely. Every 4 to 5 Ah, a voltage scan yielded V/I curves at progressing
depth of discharge. In addition, a current step curve of the hydrogen flow as
function of current and depth of discharge was obtained. The cast block anode
A2 shows a different behavior compared to the metal sheet anodes: the
current falls continuously but recovers during the break periods of the voltage
scan. This may be due to accumulated reaction products on the anode surface
which limit the ion transport and the ionic concentration equalization during
the breaks. On the other hand, the maximum current (1.4 A) is higher in
comparison with the metal sheet anodes A3eA5. The higher current of the
cast block anode at 100 mV discharge can be explained by the lower distance
(2 mm) between anode and cathode compared to the metal sheets (8 mm), i.e.
by the electrolyte-layer resistance.
Energy capacity obtained with the cast block anodes inside the large flow
cell varied greatly. Values between 20 and 65 Ah were
33
achieved. In all cases the discharges stopped due to complete clogging of the
cell space between anode and cathode by reaction products.
The behavior of the metal sheet anodes is quite different. Here a full
discharge is obtained until the thinness of the anode provokes a break-up into
larger pieces, resulting in a sudden reduction of connected anode area and in
clogging of the flow cell outlet.
The discharge current at 100 mV is quite stable with the exception of A4
which shows a prolonged start-up time of 3 Ah. A peculiarity is observed
during discharge of the A3 sheet anode. After stable operation at ca. 1 A the
current increases after 14 Ah and reaches 1.4 A maximum current. This may
be due to the for-mation of a rough surface with increased surface area. The
discharge current of the reference neat magnesium was between A3 and A4
alloy anode, but decreased during the cause of discharge.
The V/I scans of alloy anode A3 are shown in Fig. 4d. The first three V/I
curves are nearly identical followed, by an increase of current at 11.9 Ah and
stable operation at higher current between 16 and 20 Ah. In contrast, the V/I
curves of anode A2 are nearly identical until 16 Ah, which is followed by a
decrease in performance.
3.2.3. Energy density and efficiency
The total transferred energy was estimated with help of the current
integral, and H2-flow measurement. Mg utilization was determined by
measuring the anode weight difference. The hydrogen integral of the large
flow cells was not used because of some gas leakage in the system. Thus 22%
excess hydrogen gener-ation in relation to the outer current as measured for
the small cells was used. The real hydrogen flow might have been higher (Fig.
6b), thus the real energy density and hydrogen efficiency may be even higher
than estimated here.
The Faraday efficiency of the anode A1 is much lower in case of the
saturated electrolyte compared to the 4 wt % electrolyte. Optical inspection of
the cell after discharge revealed that the reaction products look quite different
to the sea water electrolyte. Therefore a different reaction mechanism can be
assumed for high electrolyte concentration. The hydrogen efficiency
calculation of neat mag-nesium results in a small value as well. This may be
due to the assumption of 22% hydrogen over-production while in reality the
hydrogen rate was much larger (Fig. 6a). At a current of 0.8 A the excess rate
was measured to be ca. 55% which would result in a hydrogen efficiency of
68%.
Homogeneous dissolution of the anode as shown for A3 is very
34 R. Hahn et al. / Journal of Power Sources 288 (2015) 26e35
important. Otherwise a higher amount of metallic particles may detach from
the anode and will be lost for energy generation. These detached anode parts
will be flushed out of the system, or may lead to clogging and breakdown. In
the closed system which was investigated here, detached anode particles
probably contribute to hydrogen generation in smaller amounts due to
corrosive hydrogen generation. Therefore the difference between the more
inhomo-geneous discharged anodes (A1, A2, A4) and the best one (A3)
would be even higher in an open system.
In fact, the utilization of A3 was so high, that a theoretical thickness of
only 30 mm was not used for hydrogen production which is an excellent
result. Remains of the used anodes are shown in Fig. 7.
For calculation of the specific energy and energy density, the real mass
and volume of A3eA5 was used while an anode thickness of 8 mm was
assumed for A1, A2 since these anodes were only partly discharged and the
same values will be obtained with 8 mm thick parts. Therefore there is no
direct correlation between anode efficiency for hydrogen production (using
the measured mass dif-ference) and specific energy density (using the weight
of the complete anode part). Specific energy and energy density was
calculated for a voltage of 0.8 V which results from the sum voltage of the
hydrogen generation cell of 0.1 V and the fuel cell voltage of 0.7 V.
The resulting electrical energy density of the Mg alloy anodes is quite
high. 1271 Wh/kg and 3009 Wh/l were achieved for the best anode A3 which
is ca. 20% better compared to the second-best A4. But the other anodes show
impressive energy density values as well, which is a clear indication of the
benefit of the investigated system as a shallow-depth sea water based power
source. Of course, these energy density values represent the energy density of
the metal “fuel” only and do not consider the weight and volume of the fuel
cell system. They can only be realized for the total system in the case of
prolonged usage. Weight and volume of the stored alloy anodes have to be
much larger than the fuel cell system itself. Magnesium alloy blocks and
sheets can be stored much easier and safer compared to hydrogen, chemical
hydrides or liquid fuels.
3.2.4. Power density
The potential difference between the Mg anode and the hydrogen
evolution cathode is much lower compared with the voltage of the metal air
battery and the discharge shows large ohmic over-potentials which results in a
relatively low power density. Nevertheless, the power density is increased by
nearly a factor of two with help of the gas separation membranes of the larger
flow cells. The mean discharge current at 100 mV of the small cells is 80 mA
while a stable discharge of the large cells was
Fig. 7. Appearance of used anodes after long term discharge (A3 and A4 in comparison with fresh anode).
2
obtained at 1.4 A. This translates to a current density of 4 mA/cm and 7.8
2
mA/cm for the submerged cathode GDEs and the GDEs with gas separation
membranes respectively. The reason for this improvement may be due to the
prevention of gas bubbles (which reduce the electrochemically active area) at
the catalyst electrolyte interface.
The other important reasons for over-potentials are the anode surface layer
and the electrolyte resistance. The cathode over-potential is not so important
in the current density range investi-gated here [22].
The large reduction of short circuit current at the start of discharge (Fig.
4b,c) may be the result of the formation of an anode solid electrolyte interface
layer (SEI) which may consist basically of Mg(OH)2. The anode discharge
behavior can be improved with help of additional alloying elements and
process technology as stated above [18, 19] but was not considered here
because this open sys-tem must be environmentally benign and the anodes
should be easily available at low cost.
Further increase of power density may be achieved by optimi-zation of the
anodeecathode distance. The large influence of the electrolyte specific
resistivity can be seen by comparing the V/I curves of thick cast anodes with
2 mm electrode separation and thin metal sheet anodes with 8 mm electrode
separation. At 50 mV a current of 1.7 A can be achieved with 2 mm distance
as opposed to 1.2 A at 8 mm distance. Therefore the electrode separation
should be kept at a minimum. A distance of 2 mm results in cell clogging and
incomplete anode discharge as was observed during long term discharges.
Anodes used with 7e9 mm distance to the cathodes could be fully discharged.
Therefore we conclude that current density can be improved by reducing the
electrode separation below 7 mm. But the distance will strongly depend on
electrode size and electrolyte flow conditions.
Assuming a total cell thickness of 12 mm (5.5 mm on both sides of the
anode and 1 mm anode thickness) and a sum voltage of 0.8 V as discussed
above, a power density of 10 W/l can be achieved. The energy density of the
fuel cell can be neglected here since it is much higher. This power density
seems to be quite low but it is higher compared to commercial direct
methanol fuel cells for energy autarkic and leisure applications [23] which are
successfully sold in the 25 W (1.1 W/l) to 110 W (4.8 W/l) range.
3.3. System test
Ambient air breathing, passively cooled micro fuel cells which were
developed at Fraunhofer IZM have been employed for the tests. The fuel cells
were optimized for lightweight portable application and passive humidity
management [24]. An automated
 R. Hahn et al. / Journal of Power Sources 288 (2015) 26e35
test bed was used which combined the hydrogen generation sys-tem shown in
Fig. 2b with the fuel cell according to Fig. 1. Instead of DC/DC converters
shown in Fig. 1 two electronic loads were used which could be programmed
independent for the fuel cell and the hydrogen cell.
The generated hydrogen is not fed directly into the fuel cell but into a
hydrogen buffer compartment which is connected to the fuel cell anode. The
hydrogen buffer is required for purges the fuel cell anode. Purge without a
buffer would not be possible since the available hydrogen rate from the
hydrogen generation cell is limited and has a sluggish response. On the other
hand, the buffer volume should be as small as possible as it increases the
start-up time of the system.
A typical system test is shown in Fig. 8. First the fuel cell is purged with
hydrogen, then the fuel cell load of 1.2 A is connected and the hydrogen cell
current fixed at 1.1 A. After pressure rises to a value of 100 mbar the
hydrogen current is reduced stepwise to determine the minimum required
value. Periodic pressure drops are due to fuel cell anode purges. Fuel cell
voltage and pressure keeps constant with a current of the hydrogen evolution
cell of ca. 1.03 A.
4. Conclusions and outlook
An environmentally friendly seawater magnesium hydrogen evolution
fuel cell system was tested under practical considerations for use as maritime
power supply. The power density was increased by a factor of two by using
hydrogen evolution cathodes with gas separation membrane, compared to
cathodes which are fully sub-merged in the electrolyte. Commercial
magnesium AZ31 rolled sheet anodes can be converted to Mg(OH)2 very
homogenously for hydrogen production, down to a thickness below 100 mm
resulting in hydrogen efficiency above 80%. The obtained energy density of
the optimized anodes is about 20% better than in the other tested samples. A
practical specific energy density of the alloy anodes above 1200 Wh/kg/3000
Wh/l is achieved, when feeding the pro-duced hydrogen into an ambient air
fuel cell. The excess hydrogen produced by the magnesium hydrogen
evolution cell because of the negative difference effect is proportional to the
cell current in case of the AZ31 alloys but not for neat magnesium. Therefore
the hydrogen rate of the hydrogen evolution cell can be controlled easily
based on the fuel cell current and temperature measurement. Stable long term
operation of the fuel cell system was shown at operating pressures below 100
mbar which can be achieved at an immersion depth of the hydrogen generator
of ca. 1 m.
Future work will include the optimization of the anode cathode
Fig. 8. System test of fuel cell and hydrogen generation cell, voltage (UHGC) and current
(IHGC) are shown as well as the fuel cell voltage (UFC) and the hydrogen pressure (p). The Fuel
cell current was fixed at 1.2 A.
35
distance of the magnesium alloy hydrogen evolution cell as func-tion of the
seawater movement and the possible influence of seawater aerosols on the
fuel cell long term stability. The catalyst cathode GDEs have been tested so
far for some hundred hours. Thus a long term test in brine water has to be
done as well and real life tests in maritime environment of the complete
system have to be performed.
Acknowledgments
This research was funded by the German Federal Ministry of Economics
and Technology under contract 03ET2028B. Special thanks go to Petra
Gehrke of Grillo AG for the supply of samples of Mg alloy anodes and the
measurement of H2 gasing rates of MgAZ31 samples. Thanks go to Thi van
Huyen Vu of ThyssenKrupp MgF for the supply of MgAZ31B sheet metal
samples. We also appreciate the fruitful cooperation and supply of hydrogen
evolu-tion GDEs by Martin Krebs of Varta micro battery. Big thanks go to
the micro fuel cells laboratory group at Fraunhofer IZM.
References
[1] N. Wang, R. Wang, C. Peng, C. Hu, Y. Feng, B. Peng, Research progress of magnesium
anodes and their applications in chemical power sources, Trans. Nonferous Met. Soc.
China 24 (2014) 2427e2439.
[2] T. Zhang, Z. Tao, J. Chen, Magnesium-air batteries: from principle to applica-tion, Mater.
Horiz. 1 (2014) 196.
[3] Y. Orikasa, T. Masese, Y. Koyama, T. Mori, M. Hattori, K. Yamamoto, High energy
density rechargeable magnesium battery using earth-abundant and non-toxic elements,
Nat. Sci. Reports 4 (July 2014) 5622.
[4] W. Wilcock, P. Kaufman, Development of seawater battery for deep-Water applications,
J. Power Sources 66 (1997) 71e75.
[5] Shinohara M., Araki E., Mochizuki M., Kanazawa T., Suyehiro K., Practical application
of a sea-water battery in deep-sea basin and its performance.
[6] P.K. Shen, A.C. Tseung, Design of a dome-shaped aluminium water battery,
J. Allied Electrochem. 24 (1994) 145e148.
[7] Robert Hahn, Andreas Gabler, Axel Thoma, Fabian Glaw, K.-D. Lang, Small fuel cell
system with cartridges for controlled hydrogen generation, Int. J. Hydrog. Energy 40 (15)
(April 2015) 5340e5345.
[8] S.M. Davis, K. Brandt, Electrochem. Cells 30 (2007) 1. US7169497.
[9] R. Hahn, S. Wagner, S. Krumbholz, H. Reichl, Optimization of Efficiency and Energy
Density of Passive Micro Fuel Cells and Galvanic Hydrogen Generators, DTIP of MEMS
&MOEMS, April 2008, pp. 9e11. ©EDA Publishing/DTIP 2008, ISBN: 978-2-35500-
006-5.
[10] Li Z.Y., Wang W., Aluminum-water hydrogen storage cell, Chinese Patent 02148850.9,
2003.
[11] L.S. Zhang, S.Y. Wang, W. Wang, Z.Y. Li, Structural design of aluminum water
hydrogen storage cell, Chin. J. Power Sources 31 (5) (2007) 393e395.
[12] Wang H.Z., Leung D.Y., Leung M.K., Ni M., A review on hydrogen production using
aluminum and aluminum alloys.
[13] Hahn R., Benz S., Electrochemical reactor useful for power supply of a system, comprises
a galvanic cell for hydrogen production with aqueous solution as an electrolyte, and a fuel
cell, DE 102009049084 B3.
[14] Krebs M., Hahn R., Wagner S., Hydrogen evolution device and fuel cell system, Patent
EP 2636768A1, US 20130236814 A1.
[15] K. Yu, X. Tin, Y. Hu, F. Chen, S. Li, Microstructure effects on the electrochemical
corrosion properties of Mg-4.1%Ga-2.2%Hg alloy as the anode for seawater-activated
batteries, Corros. Sci. 53 (2011) 2035e2040.
[16] R.K. Venkatesara, Performance evaluation of Mg-AgCl batteries for under water
propulsion, Def. Sci. J. 5 (2001) 161e170.
[17] L.J. Liu, M. Schlesinger, Corrosion of magnesium and its alloys, Corros. Sci. 51 (2009)
1733e1737.
[18] N.G. Wang, R.C. Wang, C.Q. Peng, Y. Feng, Effect of manganese on discharge and
corrosion performance of mangesium alloy AP65 as anode for seawater activated battery,
Corrosion 68 (2012) 388e397.
[19] Wang N.G., Wang R. C., Peng C. Q., Feng Y., Chen B., Effect of hot rolling and
subsequent annealing on electrochemical discharge behaviour of AP65 mag-nesium alloy
as anode for seawater activated battery.
[20] G. Song, A. Atrens, Understanding magnesium corrosion e a framework for improved
alloy performance, Adv. Eng. Mater. 5 (No. 12) (2003) 837e858.
[21] G. Song, A. Atrens, D. Stjohn, J. Nairn, Y. Li, The electrochemical corrosion of pure
magnesium in 1N NaCl, Corros. Sci. 39 (5m) (1997) 855e875.
[22] J. Blackledge, H.J. Kohnke, et al., Engineering of a single alkaline fuel cell,
J. Electr. Eng. 2 (3) (Jan.-March 2009) pp.8e17.
[23] http://www.udomi.de/produkte.php?lang¼de: Off Grid Power.
[24] S. Krumbholz, J. Kaiser, M. Weiland, R. Hahn, H. Reichl, Influences of current collector
foils with different opening ratios in passive polymer electrolyte membrane fuel cells, J.
Power Sources 196 (2011) 5277e5281.