A global transition away from
energy dense and cheap fossil fuels will require the commercial implementation
of numerous new energy technologies, each of which must be scaled
to large enough sizes to impact existing markets and substantially
lower global CO2 emissions. The electrochemical conversion
of CO2 to chemicals, fuels, and feedstocks has shown enough
promise in the past decade to justify both academic and industrial
sectors to increase efforts to better assess the potential economic
and technical feasibility of the technology. Intensifying the process
toward an industrial scale can be achieved by higher production rates,
either by simply increasing the total area of catalyst in a reactor
or by increasing the reaction rate (current density) for a given area
of catalyst. Both approaches will be needed to economically produce
an industrial quantity of product in a single plant (e.g., >100
tons
product/day). In electrochemical CO2 reduction, however,
increasing current densities to those needed for commercial operation
(e.g., >200 mA/cm2) requires researchers to use cell
designs
that can supply enough CO2 to the catalyst layer to fuel
the reaction, as opposed to traditional H-cell reactors. For these
reasons, more and more researchers have begun using catalysts deposited
onto gas-diffusion layers (GDLs), where high concentrations of CO2 can be maintained
in close proximity to the catalyst layer
even at high reaction rates. GDLs can also reduce overall cell potentials
by directly improving catalytic activity, while a more system-focused
testing platform can help reduce major system losses such as ohmic
heating. Researchers operating these experimental devices at higher
current densities, however, have discovered a number of operational
intricacies that can make the direct switch away from lower current
density experiments in an H-cell challenging. This Viewpoint is meant
to describe some of these unique operational considerations that can
impact catalytic activity and our ability to accurately collect data,
while acting as a starting guide for researchers to transition to
using gas-diffusion layers as a platform for benchmarking novel catalysts
at commercially viable current densities.
Specific
discussion points here may apply broadly across all platforms
for performing high current density CO2 reduction, while
others pertain specifically to our presented cell configuration consisting
of three chambers and a catalyst layer deposited onto a GDL (Figure 1
). Membrane electrode
assembles, which show promise in the reduction of overall cell potentials
in future systems, are a similar cell configuration with additional
operational intricacies brought on by a stagnant or extremely small
catholyte layer.
1−5
For researchers new to high current density testing, however, the
three-chamber configuration can provide an easy platform to rapidly
test new catalysts and GDLs.
6−9
This Viewpoint briefly discusses the intricacies
of operating GDLs for CO2 electroreduction through three
main subtopics: assembly, operation, and postanalysis of data.
Figure 1
(a) A standard
electrochemical for CO2 electrolysis
for testing catalysts in a three-electrode configuration. (b) Side
profile showing chambers for the CO2 (left), catholyte
(center), and anolyte (right). (c) Isometric view of the exploded
three-chamber cell showing individual components.
In electrochemical systems, a catalyst is typically deposited
onto
a substrate, with carbon paper, glassy carbon, and metal foils being
common supports. In high current density experiments, a GDL
10−12
commonly acts as a support for the catalyst. A catalyst is typically
deposited directly onto the microporous layer side of the GDL, which
is hydrophobic and helps to form a gas–liquid interface between
the microporous layer and the catalyst. The creation of new GDLs specifically
with mechanical and chemical properties suited for CO2 reduction
applications will be an important avenue for new research going forward,
13,14
but most current systems report using commercially available GDLs
such as Sigracet 39BC or variations from Freudenberg or Covestro.
The hydrophobic nature of bare carbon-based GDLs and an as-prepared
GDL with a 100 nm layer of sputtered silver is demonstrated in Supporting Video 1.
Here, we further show that
after the surface has been briefly placed under reducing potentials,
a separately prepared 150 nm thick Cu catalyst layer becomes fully
wetted when a droplet of water is pipetted on top. The bare GDL without
a catalyst layer, however, maintains its hydrophobic surface even
after being placed in reductive conditions. During operation, commonly
used metallic and porous catalyst layers (e.g., Ag, Au, Cu) are then
assumed to be fully covered by the electrolyte, while CO2 diffuses across a gas–liquid
interface provided by the hydrophobic
surface of the microporous layer inside of the GDL. As we discuss
in more detail in our recent perspective, we then assume that CO2 travels a short
distance and reacts in dissolved form rather
than a gaseous form with a three-phase interface.
15
The diffusion of CO2 from a nearby gas
phase not only
allows for much higher current densities than an H-cell but has consequences
for the morphology, activity, and stability of the catalyst layer.
In an H-cell, for example, even a nanostructured catalyst functions
as a planar electrode, relying on CO2 to diffuse 40–120
μm to reach the catalyst layer. In this configuration, almost
all CO2 reduction measurements are then performed under
some degree of mass transport limitations, resulting in concentration
polarizations and making intrinsic activity difficult to accurately
determine and potentially overestimated.
16
For a rough nanostructured electrode with a surface porosity of
several microns, only the outermost surfaces may have access to sufficient
reagent, while the base can be depleted of CO2.
17
In a GDL, however, the catalyst layer must allow
for access to CO2 on one side and electrolyte on the other
side, resulting in a porous electrode structure compared to the planar
system in an H-cell. In this configuration, a large electrochemically
active surface area is then possible, with all surfaces of the catalyst
having comparatively greater access to CO2 due to the short
diffusion pathway compared to that in an H-cell. Thus, the greater
surface area for CO2 reduction can then result in greater
geometric activity at lower overpotentials than H-cell experiments.
Further, a GDL can allow for gaseous products to diffuse into the
gas phase prior to nucleating at the surface and blocking active sites.
These structural differences between H-cells and GDLs would similarly
result in different pH and concentration gradients along catalyst
nanostructures, meaning that structural benefits of catalysts derived
in an H-cell may not have performance that translates directly into
a GDE configuration.
A challenge that results from this orientation,
however, is the
potential for hydroxide and carbonates to build up inside of the structure
before they can be transported to the bulk electrolyte. Crystallization
of salts can then be a common occurrence within the porous catalyst
layer, especially at higher current densities where substantial hydroxide
is generated.
18
These considerations should
then be factored into the design of catalyst structures for CO2 reduction, in addition
to the design and composition of the
material itself. Finally, in an H-cell configuration, impurities are
easily deposited onto the outermost catalyst sites, which can cause
fast deactivation of CO2 reduction in favor of hydrogen
evolution, as these sites possess the greatest access to CO2 coming from the bulk
electrolyte. Due to the opposite direction
of CO2 transport within a GDL system, however, impurities
are likely to deposit onto the catalyst surfaces furthest from the
source of CO2.
15
As illustrated
in Supporting Video 2, the electrical connection
and sealing of a GDL within a cell is
also important. Here, conductive tape is applied on the gas side (noncatalyst
side) of the GDL composed of carbon fibers to provide an electrical
connection to the potentiostat but also to physically fix it to the
sealing gasket during cell assembly. More importantly, conductive
tape is applied around the entire electrode instead of just at the
top in order to minimize the in-plane distance that electrons travel
through the GDL to reach the CO2 reduction catalyst. Sufficient
current collection is extremely important for higher current density
operation due to the relatively high resistivity of commonly used
GDLs (∼106× more resistive than pure Cu), which
can cause potential/voltage variations across the GDL and catalyst
layer that scale with the applied current density (via Ohm’s
Law; V
drop_GDL = IR
GDL). These potential variations can then result
in heterogeneous local current distributions across the catalyst layer.
In terms of CO2 reduction and catalyst characterization,
the nonuniform potential and current density throughout the catalyst
layer could then result in location-dependent product formation and
differences in the local reaction environment (e.g., pH), irrespective
of the catalyst used.
This intricacy of high current density
CO2 electrolysis
is particularly important for CO2 reduction on copper electrodes
due to the sensitivity of pH and current density on H2,
CO, and C2 product formation.
13,19,20
These implications can be illustrated in Figure 2
a by assuming an ideal catalyst with an overall
activity of 120 mV/dec (after iR and local pH correction).
In this figure, we can see the differences in the sensitivity of the
overpotential and the local pH as a function of the geometric current
density. An applied potential difference of only 36 mV induced by
resistive losses on this electrode could then result in current density
variations from 100 to 200 mA/cm2, which could then result
in large spatial differences in the local reaction environment across
the electrode surface. The lower current density region (e.g., 0.01–10
mA/cm2), however, is less sensitive to these effects due
to the exponential relationship between the overpotential and current
density and the dependency of the GDL’s resistive losses on
the current density. These considerations are then unique to only
high current density catalyst testing and applications. It is thus
important to emphasize that catalytic activity at higher current densities
can be strongly influenced by noncatalytic electrode variations, such
as the resistivity of the GDL, how current is collected, and the overall
electrode size.
Figure 2
Examples of important considerations when benchmarking
catalytic
CO2 reduction performance under high current density operation.
(a) Comparative effect of current density on the ideal required overpotential
and the local reaction environment, showing how the system behaves
differently as a function of current density. (b) Effect of high current
operation on the measured peak area of H2 as recorded by
a gas chromatograph (GC), showing deviation from the limit of linearity
(dashed) at high concentrations (2.25 cm2 geometric area).
(c) Measured solution resistance between the working and reference
electrodes as a function of applied current density using both current
interrupt (CI) and electrochemical impedance spectroscopy (EIS) (1
M KHCO3 is used as an electrolyte). This illustrates changes
in the electrolyte conductivity as a function of operation and current
density.
Assembly of the entire cell for
electrochemical testing can be
seen step-by-step in the Supporting Information, while a video showing cell operation
is also available (Supporting Video 3). Here and in Figure 1
, we can see the overall simplicity
of the cell design but also the typical ancillary equipment needed
to run the device such as a pump to flow liquid through the electrolyte
channels and a GC to measure the formed gaseous products. In high
current density experiments, it is also important to note that the
concentration of product gases entering the GC increases significantly
as compared to H-cell testing, which can impact product measurement.
Multipoint calibration of a GC across the concentration range of all
of the expected gas products (H2, CO, CH4, C2H4) is then necessary as the concentrations
produced
may no longer be within a linear range, known as the Limit of Linearity
(illustrated using H2 in Figure 2
b). This can lead to overall Faradaic efficiencies
(FEs) greater than 100% if the lower-concentration calibrations typically
used in H-cell testing are extrapolated. This can be particularly
misleading if liquid products are not measured or reported in the
total FE measurement as the observed total FE of gas products may
still be below 100%. While it is possible to increase the gas flow
rate to reduce the GC’s measured concentrations back to a linear
calibration range, this may cause pressure imbalances between the
gas/liquid phases within the GDL while reducing the single-pass conversion
efficiency of CO2. Similarly, because the GC peak areas
are related to the total applied current, the overall geometric catalyst
area can also be reduced. Either way, it is essential that the GC
is properly calibrated within the range of concentrations that are
produced, instead of extrapolating from lower-concentration calibration
data.
During operation of a GDL or membrane–electrode
assembly,
other operational factors may cause instability or complicate analysis
of electrochemical behavior if not taken into consideration. At higher
current densities, much higher ohmic drops between the working and
reference electrodes can be expected due to the larger charge passed
through the electrolyte. Not only can this be a major contributing
factor to the overall cell potential in a two-electrode setup but
it must be taken into account when correcting for the potential of
the working CO2 reduction cathode in a three-electrode
system. While electrochemical impendence spectroscopy (EIS) is typically
performed even in H-cells to determine a system’s ohmic drop
between the working and reference electrodes, the measured electrolyte
resistance value between the reference and working electrodes can
also be shown to vary as a function of the current density (Figure 2
c). Here, using two
different methods and several replicates, the solution resistance
is found to decrease with the applied current density. This means
that iR determination and correction may require
resistance measurements at various conditions to capture the real
working potential of a cathode. In the case of the 1 M KHCO3 electrolyte used to acquire
this data, the change in electrolyte
resistance is likely due to an overall increase in electrolyte conductivity
due to the generation of hydroxide at the cathode surface. This will
change the electrical properties of the electrolyte within the diffusion
region but can also change the overall pH and conductivity of the
bulk catholyte over time, particularly if the buffer breaks down.
Alternatively, if 1 M KOH is used as a catholyte, long exposure to
gaseous CO2 via the GDL may cause the pH to steadily decline
due to the spontaneous formation of bicarbonate, along with a corresponding
reduction in the conductivity of the electrolyte. Unless accounted
for, these factors can cause the determined working potential of the
cathode to differ from an iR correction performed
at 0 mA/cm2.
Additionally, the large ohmic drop throughout
the system driven
by high current density operation can result in large temperature
changes to the electrolyte, further affecting solution conductivity,
electrode activity (via Arrhenius’ Law), and the solubility
of CO2 diffusing across the gas–liquid interface.
Without a sufficient electrolyte volume and passive cooling of the
electrolyte chambers to mitigate the heating in the system, these
temperature changes will affect the observed electrochemical results.
Even worse, over a multihour stability test, the temperature can increase
gradually, providing transient operating conditions. Additional strategies
to avoid this include minimizing the electrolyte distance between
the anode and cathode and using a high-conductivity electrolyte to
facilitate charge transport.
A final complexity of operating
GDLs pertains to the delicate gas–liquid
interface that provides gaseous CO2 in close proximity
to the catalyst. Even slight overpressures on either the gas or liquid
side of the GDL can cause gas to bubble into the liquid phase or result
in flooding of the GDL. If possible, the pressure of both phases across
the GDL should then be regulated to ensure that catalytic activity
can be determined without additional uncertainties introduced by stability
of the gas–liquid interface. One notable source of pressure
imbalance can even come from in-line GCs connected to the cell. Here,
a constant backpressure from the GC, as well as pressure increases
during injections, can cause gas to enter the liquid phase if the
pressure spike is too high. For GCs that use syringe injections, however,
the CO2 gas pressure is easier to maintain near atmospheric
pressure.
In summary, this Viewpoint discusses several operational
intricacies
of using GDLs for electrochemical CO2 reduction, which
is becoming increasingly important as the number of reports at higher
current densities grows. The differences in testing/optimizing catalyst
performance between traditional aqueous H-cells and gas diffusion
electrodes is not trivial, and many new protocols must be used to
ensure proper sample preparation, recording of data, and product identification.
However, we believe that, with proper care and attention, the large
field of catalyst researchers working on CO2 electroreduction
can leverage existing infrastructures to expedite the scientific development
of this technology. We hope this article acts as a source of information
for catalyst-focused researchers looking to move to high current density
catalyst testing and reinforces the need for fundamental research
performed under practical conditions.