Low-cost fuel-cell based sensor of hydrogen

production in lab scale microbial electrolysis cells

Nu ria Montpart a, Mireia Baeza b, Juan Antonio Baeza

a,*,

Albert Guisasola

Article history:
Received 17 October 2015
Received in revised form
29 July 2016
Accepted 24 September 2016

abstract

A fuel cell sensor was demonstrated as a low cost on-line monitoring tool
for hydrogen producing microbial electrolysis cells (MECs) at lab scale.
Hydrogen produced in the MEC was oxidized at the anode of the fuel cell
generating electricity that could be easily monitored. The total electrical
current obtained was proved to correlate with the hydrogen supplied (r2
0.99) and was equally efficient as other reference analytical
methodologies based on gas chromatography. Signal was repetitive (2.3%
variation coefficient, n = 12) and did not show interferences by the
presence of other compounds like methane and carbon dioxide. The fuel
cell was coupled to single and double chamber MECs validating its
applicability as on-line monitoring portable device for quantifying
hydrogen production. The use of the fuel cell as an on-line hydrogen
sensor offers an alternative to more complex methodologies and offers
the possibility to implement control systems and optimization strategies.
Introduction
In the current oil dependency situation, hydrogen has raised interest
because of being a renewable clean energy vector whose combustion has
no impact on the greenhouse effect. Biohydrogen production technologies
are an interesting alternative to the common steam reforming of fossil
fuels to chemically produce hydrogen, especially when using waste
streams as feedstock. Microbial electrolysis cells (MECs) simultaneously
allow the treatment and valorization of wastewater by producing
hydrogen.
MEC operation relies on the presence of a group of microorganisms
that have the ability to use an external insoluble electrode as final
electron acceptor (known as exoelectrogens or anode respiring bacteria,
ARB). ARB consume organic matter available in wastewater under
anaerobic conditions, donating the last electron involved in their
metabolic pathway to the electrode. The supply of some electrical energy
to overcome thermodynamic limits allows the flow of electrons generated
on the anode to the cathode, where the abiotic
hydrogen production takes place. The interest of this process lies on the
facts that (i) energy requirements to drive this process are much lower
than those required for water electrolysis (ii) wastewater depuration and

valorization are accomplished and (iii) higher hydrogen yields than with
other biohydrogen producing technologies can be reached.
In MEC lab scale studies, analytical methods to monitor the system
performance in terms of hydrogen production may be an issue because of
practical reasons, such as the cost of the equipment, the system location
or the number of replicate reactors to monitor. Operational reasons, such
as too low hydrogen production, can also impede its detection. The use of
a gas chromatograph (GC) is a common practice in lab scale studies when
aiming at hydrogen production monitoring, either by off-line
measurements or by coupling the reactor
with an on-line GC. Both methodologies also require the connection of a
flow-meter to fully quantify the production. Gas production can also be
measured on a GC sparing a flow-meter with a double analysis strategy as
presented by Ambler and Logan. Other analytical techniques described in
the literature for hydrogen detection comprise the use of membrane inlet
mass spectrometry together with a flowmeter system or specific hydrogen
sensors. These techniques are notably expensive and require regular
maintenance. The on-line estimation of biohydrogen production based on
conductivity measurements has also been proposed by Gueguim Kana et
al.. The main advantage of measuring hydrogen production on-line is the
possibility to develop process control techniques that could improve the
system
performance.
As a sensor, a device should present certain characteristics, such as
high sensitivity and selectivity, robustness, small response time, ability to
transduce the sensor signal to electrical signal, operation at room
temperature and low power consumption. Current materials for gas
sensors are semiconductor metal oxides, where the absorption and
desorption of the gas on its surface produces a change on the material
conductivity that can be correlated with the gas concentration. A small
amount of catalytic metal is often added to
improve the selectivity and the sensitivity of the gas sensor. Other
currently used methods to monitor hydrogen in biogas plants are
electrochemical sensors.
A never reported low-cost strategy for on-line hydrogen monitoring
in MEC is presented in this work. A fuel cell is examined to work for
hydrogen production quantification and, therefore, its potential to give a
fast and stable response, a long lifetime and easy calibration and
operation are tested. The possibility of monitoring hydrogen production
with the system proposed in this work is validated coupling it to an MEC in
single chamber and double chamber configuration.

Materials and methods

Fuel cell sensor

A single reversible fuel cell provided with a cation exchange
membrane was used in this work (54 mm 54 mm 17 mm, SKU 632000,
Fuel Cell Store, USA). A needle was tightly connected to the anodic
chamber inlet port, which allowed the connection to gas collection bags or
to the system headspace (Fig. 1). Epoxy glue (Araldit, Ceys) was used in
joints to minimize leakages. The cathodic inlet port was left uncapped to
permit free air diffusion into the cathode.

Fig. 1 - Fuel cell sensor with connection modification in the anodic

chamber inlet port.

Gas bags with low hydrogen permeability (Ritter, Cali-5- Bond, 100
mL and 500 mL) were used to supply the gas sample to the fuel cell for its
calibration. A Teflon rubber septum permitted the connection with the
needle from the fuel cell. The bags were rinsed three times with nitrogen
gas and emptied with a vacuum pump prior to gas sample introduction.
The gas sample was introduced to the gas bag via a gas tight syringe
(Hamilton Samplelock Syringe, 1 mL and 10 mL). High purity gases
(hydrogen, nitrogen, carbon dioxide and methane, Air Products, Inc.) were
used to prepare the gas samples.

Microbial electrolysis cells

Different MECs were coupled to the fuel cell sensor for validation on
a real biohydrogen production system. The single chamber membraneless MEC was a 28 mL cylindrical methacrylate vessel provided with a 16
mL glass cylinder at the top tightly sealed with a PTFE rubber cap that
enabled gas collection and connection to the fuel cell. The anode was a
graphite fiber brush (20 mm diameter 30 mm length; 0.21 m2 specific
surface area; made with fibers of diameter 7.2 mm, type PANEX33 160K,
ZOLTEK) with a titanium wire core. The
cathode consisted of graphite fiber cloth (3.8 cm diameter, 7 cm2 total
exposed area) coated with platinum (5 mg Pt/cm2, ElectroChem Inc.). Both
electrodes, spaced 2.5 cm apart, were connected to a power source (HQ
Power, PS-23023) applying a potential of 0.8 V. Current intensity was
measured quantifying the voltage drop across a 12 U external resistance
serially connected to the circuit. The cell was easily converted to a double
chamber MEC by coupling an identical module and placing an anion
exchange membrane (AEM) in between (AMI7001S, Membranes International INC). The membrane was soaked

overnight in a 10% sodium chloride solution. Under this configuration the

distance between electrodes increased to 7 cm. Hold-up volume per
module was 35 mL. MEC run in fed-batch mode at room temperature with
acetate as sole carbon source (0.5e1 g L-1 initial concentration). The
medium used contained per liter: 0.2 g NH4Cl, 4 mg FeCl2, 6 mg Na2S, 5
mL of mineral media solution and 172 mL phosphate buffer solution (PBS).
Mineral media solution contained per liter: 1 g EDTA, 0.164 g CoCl2$6H2O,
0.228 g CaCl2$2H2O, 0.02 g H3BO3, 0.04 g Na2MoO4$2H2O, 0.002 g
Na2SeO3, 0.02 g Na2WO4$2H2O, 0.04 g NiCl2$6H2O, 2.32 g MgCl2, 1.18
g MnCl2$4H2O, 0.1 g ZnCl2, 0.02 g CuSO4$5H2O and 0.02 g AlK(SO4)2.
The PBS stock solution contained per liter 70 g Na2HPO4 and 12 g
KH2PO4. 2-Bromoethanesulfonate was used as methanogenic activity
inhibitor at a concentration of 50 mM. In the double chamber MEC
configuration, the catholyte was a 100 mM PBS solution. The cell content
was completely replaced with fresh medium when the batch cycle
finished. MEC were sparged with nitrogen for 10 min (50 mL N min1) after
feeding to guarantee anaerobic conditions. The fuel cell was also rinsed
with nitrogen before its connection to MEC.

Performance indexes
Voltage response in the fuel cell and MEC was monitored by means
of a 16-bit data acquisition card (Advantech PCI-1716) connected to a
personal computer with a software developed in LabWindows CVI 2013 for
data acquisition. Current intensity (I) in the fuel cell and MEC was
calculated measuring the voltage drop (V) across a resistor (Rext)
according to Ohm's law (Equation (1)).
I = V/Rext
(1)
The electrical charge circulated in the system was calculated as presented
in Equation (2):

Charge(coulombs) =

tf

I (t)dt

(2)

ti

Theoretical coulombs supplied to the fuel cell as hydrogen could be

estimated from the volume of hydrogen fed to the
fuel cell as presented in Equation (3):

Charge (coulombs) H2 = nH2 . 2 . F

(3)

where nH2 is the moles of hydrogen, calculated with the ideal gas law at
the current temperature, 2 are the moles of electrons available per mole
of hydrogen and F is Faraday's constant (96,485 C/mol-e).
Coulombic recovery for standards was calculated as the percentage
of coulombs measured by the fuel cell (Equation (2)) with respect to the
theoretical coulombs calculated from hydrogen supplied (Equation (3)).
Gas chromatography (Agilent Technologies, 7820-A) was used as
reference method for hydrogen analysis using a thermal conductivity
detector and argon as carrier gas (oven temperature 40 C, HP-mole sieve
column, detector temperature 220 C). Gas quantification was performed
according to the procedure presented by Ambler and Logan, following a
double run methodology. In this methodology, the sample is analyzed in a
first run, then a known volume of nitrogen is added as a reference
compound and the resulting mixture is again analyzed. Mass balances
calculations including the change in sample composition and the volume
added allow the calculation of the initial gas volume as presented by
Equation (4):

(4)

where Vtotal i is the initial total gas volume in the bag, Vadded;N2 is the
known volume of nitrogen added, Vrun 1 is the volume injected in the GC in the
first analysis, and x is the molar fraction of nitrogen in the first analysis or the
second as indicated on the subscript. The output of gas chromatography
analyses is the volumetric fraction (or molar fraction) per compound, thus,
knowing the total initial volume, the initial volume per compound can be
calculated. Acetate concentration was analyzed from 0.22 mm filtered samples
with gas chromatography using a flame ionization detector and helium as carrier
gas (oven temperature 85e130 C, ramp of 3 C min1, 130-220 oC, ramp of 35 C
min-1, nitroterephthalic acid modified PEG capillary column, detector temperature
275 C).

Results and discussion

Description of the system

The ability of the fuel cell to work as a sensor was initially tested.
Two fuel cells were used in parallel in the tests performed in this work to
account for the device variability, observing no significant differences. For
each test, the fuel cell was connected to a gas bag containing a known
volume of hydrogen gas. Hydrogen supplied diffused to the anodic
chamber of the fuel cell where it was oxidized. Protons migrated through
the cation exchange membrane whereas the electrons flowed along the
electrical circuit to the cathode producing a voltage signal. In the cathodic
chamber, oxygen was reduced to water. Voltage reached a maximum in
the first minutes of the test and, then, decreased progressively (Fig. 2A).
Current intensity was calculated with Equation (1) using the measured
voltage and then Equation (2) was used to calculate the total electrical
charge. Fig. 2B shows the current intensity obtained with different external
resistances for an identical hydrogen volume supplied. As can be
observed, the system reached higher current intensity for an external
resistance of 12 U, showing less limitation than the external resistances of
100 and 350 U.

From a chemical engineering point of view, the fuel cell operation is

a reaction driven by a non-porous heterogeneous catalyst and hence there
is external mass-transfer limitation of hydrogen from the bulk gas to the
anodic catalyzer surface, and chemical reaction with its three steps of
reactant (hydrogen) adsorption, surface reaction and product (protons)
desorption. In addition to these classical heterogenic catalysis steps, the
external electrical resistance of the fuel cell limits the maximum rate of
the surface reaction, reducing the electron flow through the electric
circuit. Regarding the experiments performed with different external
electrical resistances, the current intensity profiles obtained with 100 and
350 U were lower than the profile obtained with 12 U. As the only
operational change is the increased external resistance, it means that the
surface reaction is the slower step in the overall process, at least in the
cases of 100 and 350 U. The exact value of external resistance where
surface reaction becomes limiting depends on the flow convection around
the cathode as well as on the cathode physical characteristics. In our
system, no external mass transfer limitation seems to exist for the 100
and 350 U resistances except after 10 h of operation, when the change in
the intensity profile indicates some external mass transfer limitation effect
due to very low hydrogen concentration in the bulk. The observed
constant decrease of current intensity is related to the decrease of
hydrogen concentration in the gas bag (i.e. current intensity seems to be
proportional to the bulk hydrogen concentration). On the other hand, the
results for the external resistance of 12 U show that external mass
transfer seems to be limiting the observed reaction rate. The shape of the
current profile depends on the diffusion of hydrogen from the bulk gas of
the bag through the needle and the inlet port of the fuel cell to the surface

of the catalyzer, where several zones are exposed to different hydrogen

concentration and these zones are changing with the decrease of bulk
hydrogen concentration during the operation of the cell. Other factors
such as water accumulation in the anode that decreases the catalytic
surface may be also modifying the fuel cell response.

Selection of external resistance and system calibration When the fuel

cell is conceived to work in series with another device, for example if it is to be

used as a sensor like in this work, it is crucial that the fuel cell is not the limiting
step of the process. The rate at which the electrochemical reactions take place in
a fuel cell is dependent on the external resistance that is used to close the
electrical circuit. Hence, a set of external resistances (i.e. 12, 50, 100 and 350 U)
were tested to evaluate which load would offer the maximum quantification of
the volume of hydrogen supplied to the fuel cell (Fig. 3). Such quantification was
determined in terms of percentage of coulombs recovered, i.e. coulombs
measured referred to hydrogen supplied in units of coulombs.
As shown in Fig. 3, higher external loads resulted in lower coulombic
recoveries, which were also more dispersed. These observations were most
probably caused by the electrical current limitations mentioned before, since
high external resistances led to slower surface reactions and, thus, to higher
hydrogen retention time in the system which would have led to an increase in
hydrogen leakage to the exterior. Hydrogen leakage in these systems can
become a significant issue if it is not well addressed. The highest coulombic
recovery was achieved with the 12 U external resistance, allowing an average
recovery of 71%. This value can be considered high because the usual efficiency
reported for fuel cells is around 40-60% of useful output energy versus total
input energy. Considering the results of Fig. 3, the 12 U resistance was used in all
the following tests.
Fig. 3 e Hydrogen recovered (in units of coulombs) as a function of the
external resistance in the fuel cell. Error bars indicate standard deviation, which
was calculated based on triplicate tests.
Even though current intensity was dependent on the external resistance,
reaching lower values for higher loads, a signal response was immediately
detected in all the cases. This device should have a response time at least one
order of magnitude lower than the process that is being monitored. Thus, a fast
response is needed in a measuring device aiming at fast hydrogen detection, so
that a corrective control action can be quickly applied. However, a slower
response of the measuring instrument can be accepted for a slow biological
dynamics as in an MEC.
Once the adequate external resistance was set, the fuel cell was calibrated
to get a correlation between the coulombs measured in the fuel cell and the
actual moles of hydrogen supplied. The system was considered to be at 1 atm
pressure during these calibration tests. As shown in Fig. 4, a good linear
correlation with high repeatability was obtained for the quantification. The
highest variability was observed for the highest quantity of hydrogen tested.
Higher volumes have longer analysis time, which favors hydrogen diffusion and

increases the probability of human error when sampling the gas, increasing the
standard deviation of the results.
Fig. 4 e Fuel cell sensor calibration for coulombs measured vs. moles of
hydrogen supplied. Error bars indicate standard deviation, which was calculated
based on triplicate tests.
In terms of hydrogen volume, also a lower detection limit for the fuel cell
was of interest. Indeed, 4 mL of hydrogen was found to be the lowest volume
that permitted its quantification at the highest attainable coulombic recovery, i.e.
71%.
A total of 12 consecutive experiments (containing 10 mL of hydrogen
each) were used to assess the repeatability of the fuel cell as an electrochemical
sensor. The fuel cell quantified an average of 71.09 C with a variation coefficient
of 2.3%. Values for the variation coefficient lower than 5% are generally
describing high repeatability.

Validation of the fuel cell sensor

As already mentioned, gas chromatography is the most common analytical
method to measure hydrogen concentration. Hence, it has been considered as a
reference method to compare with the one presented in this work.
Standard mixtures of hydrogen and nitrogen were quantified in triplicate
by gas chromatography and with the fuel cell. A strong positive correlation (slope
very close to the unity) was obtained when comparing both methods, indicating
that these methodologies were analogous regarding hydrogen quantification (Fig.
5). The least-square linear regression test fitting the five averaged points was
used to make the statistical evaluation of the fuel cell sensor and the errors
associated to slope and intercept. No significant difference was observed at 95%
confidence level (dashed line in Fig. 5). The slope and intercept were 0.98 0.05
and 0.00 0.03, respectively (r2 = 0.996; n = 5; 95% confidence level).
Fig. 5 e Validation of a fuel cell as hydrogen monitoring device with a
reference method (gas chromatography, GC). The linear regression equation
obtained was y = 0.98(0.05) x + 0.00(0.03), r2 = 0.996. Average values from
triplicate tests for GC and fuel cell (symbols) and their corresponding standard
deviation, linear regression (solid line) and 95% confidence intervals (dashed
line).

Evaluation of possible interferences from other biogas

compounds
Methane and carbon dioxide were tested independently in the device to
check that they were not interfering and therefore no electrical signal was being
detected. Neither methane nor carbon dioxide showed current generation in the
fuel cell, indicating that they did not react and no interference existed. Another
verification test with hydrogen was performed afterward to confirm that no
catalytic damages were caused by these compounds. In both cases, the signal
was totally matching the signal before the tests with carbon dioxide or methane,

and therefore no catalytic loss was experienced in the system. Methane and
carbon dioxide are therefore regarded to be inert in the fuel cell.
Carbon dioxide was proven to be inert when fed independently in the fuel
cell. Moreover, mixtures of carbon dioxide with hydrogen were also tested to
investigate about the possible poisoning effects of carbon monoxide, which can
be produced by the reverse water gas shift reaction. Carbon monoxide is
reported to adsorb on the catalyst, decreasing the efficiency of the fuel cell.
Nevertheless, no catalytic loss was observed in the system studied along all the
operational period of this study.
Regarding the lifetime of the device, the system was tested for a period of
four months without seeing any loss in its measuring capacity. A periodical
control analysis and, if necessary, a calibration of the device would be
recommended. An option to avoid catalytic loss by carbon monoxide formation
includes the possibility of trapping carbon dioxide in a sodium hydroxide solution
before the gas mixture arrives to the fuel cell.

Implementation in a hydrogen-producing microbial electrolysis cell A

last test using the fuel cell as an on-line monitoring tool was performed
connecting it in series with an MEC, both in single chamber (Fig. 6A) and double
chamber configuration (Fig. 6B). Fig. 6C and D show, respectively for each
configuration, the current intensity monitored for the MEC and the fuel cell. Both
signals followed a similar trend although the response for the fuel cell was lower
than for the MEC. Lower current intensities in the double chamber configuration
were result of a lower applied voltage and higher spacing between electrodes.
Fig. 6 e Implementation of the fuel cell as an on-line hydrogen production
monitoring tool in a single chamber MEC (left) and in a double camber MEC
(right). A, B: Experimental setup. C, D: MEC current intensity signal (solid) and
fuel cell response (bold). E, F: Hydrogen production measured by the fuel cell
(bold), estimated according to the MEC current intensity (solid), and the
maximum attainable according to substrate fed (dashed).
The cumulative moles of hydrogen produced (Fig. 6E and F) were
quantified (i) according to the fuel cell response by applying Equation (2) with
the intensity measured in the fuel cell and using the calibration shown in Fig. 4,
and (ii) according to the MEC response by applying also Equation (2) but with the
intensity measured in the MEC and using Equation (3), assuming that the charge
calculated with Equation (2) was devoted to hydrogen production. As can be
observed in Fig. 6E, the theoretical production of hydrogen according to the MEC
response for the single chamber MEC configuration was higher than the
maximum attainable moles according to the substrate fed. This observation and
the coulombic efficiency higher than 100% were indicators of a case of hydrogen
recycling in single-chamber MEC. In such situation, hydrogen electrochemically
produced in the MEC is used as electron donor by homoacetogenic bacteria
producing acetate, which contributes to additional current generation in the
system. Another consequence of hydrogen recycling is that hydrogen measured
is lower than expected based on the MEC signal. Actually, hydrogen measured by
the fuel cell only accounted for 44% of the theoretical production based on MEC
current intensity and 55% of the maximum attainable hydrogen production
based on the acetate provided. However, these values lower than 100% are not
indicating an incorrect quantification of the hydrogen produced by the fuel cell.

Conversely, this quantification is more accurate than both methodologies. The

quantification based on MEC current intensity has the problem created by the
recycling phenomena, while the theoretical value based on the amount of
acetate added does not consider any other biological process that could occur as
biomass growth, methanogenesis or hydrogen recycling.
On the other hand, the monitored signal matched perfectly the hydrogen
production in a double chamber MEC configuration (i.e. in absence of any
hydrogen recycling phenomena). Fig. 6D and F show a perfect description of
hydrogen production with a 100% quantification of the theoretical hydrogen
production. The overlapping of hydrogen production estimation by the fuel cell
and by the MEC indicates a perfect operation of the double chamber MEC, where
all the acetate consumed generates electrical current that is used by the cathode
to generate hydrogen. Double chamber MECs are better for producing hydrogen,
avoiding the problems of microbial competence for substrate as in single
chamber cells (especially homoacetogenesis and methanogenesis). However,
double chamber MECs have also other problems as increased pH gradient
between cathode and anode, which decreases greatly the hydrogen production
performance.

Practical implications
The use of a fuel cell at lab scale to monitor hydrogen production is proposed
and evaluated as an alternative to more complex analytical methods, such as
gas chromatography. There are advantages and drawbacks for both systems, but
in the end the possibility of using one or another is highly dependent on the
system to be monitored (configuration, use, location, budget, etc.).
Regarding gas chromatography, the advantages of the system include the
possibility to detect more than one gas compound, not only hydrogen, and the
availability of commercial on-line GC. Nevertheless, the device is notably
expensive (thousands of euros) and it requires the consumption of utility gases.
Also, when it comes to the analysis per se, a double analysis is required in order
to be able to quantify the volume produced (as in Ambler and Logan).
Alternatively the system can be connected to a flow meter to measure total gas
production, increasing the costs.
On the other hand, the fuel cell as a device to monitor the hydrogen
produced represents a completely affordable instrument (less than 50 V in this
case), which can easily be set on-line with the system to monitor and that only
needs a conventional data acquisition system to record voltage. Being selective
for hydrogen, it can be a perfect option in systems where pure hydrogen is
produced, such as double chamber MECs. In a single chamber MEC configuration,
or in other anaerobic systems, only hydrogen produced or its relative
composition can be measured.
The time required to perform the analysis can be quite long when using it
off-line, in the range of few hours, since hydrogen arrives to the reacting
catalytic surface only by diffusion, which can be slow due to the reduced driving
force when the hydrogen concentration decreases. This possible problem could
be minimized by using a fuel cell with higher capacity.

The fact that the fuel cell will consume the product of interest is not an
important drawback at lab scale. More importantly, having a device able to
monitor on-line with high reliability the hydrogen produced also allows the
implementation of control strategies, which can end up improving the whole
performance of the system.

Conclusions
A low cost fuel cell is presented in this work as an alternative,
transportable and robust methodology to quantify hydrogen production in MEC at
lab scale. Tests revealed the high repeatability of the system, showing a high
correlation of the signal with the amount of hydrogen supplied. A strong
correlation was also obtained when comparing the methodology presented here
with gas chromatography as a reference method and no significant differences
were observed at 95% confidence level. In addition, neither interference due to
other biogas compounds nor loss of catalytic activity during all the operational
period were observed. The use of the fuel cell as on-line monitoring device
measuring biohydrogen production in MEC at lab scale was also demonstrated in
single and double chamber configurations.