Hydrodynamics
Change
Tafel Slopes
in
Electrochemical
CO
2
Reduction
on Copper
Nicholas
B. Watkins,
†
Zachary
J. Schiffer,
†
Yungchieh
Lai, Charles
B. Musgrave
III, Harry
A. Atwater,
William
A. Goddard
III, Theodor
Agapie,
*
Jonas C. Peters,
*
and John M. Gregoire
*
Cite This:
ACS Energy
Lett.
2023,
8,
2185−2192
Read Online
ACCESS
Metrics
& More
Article
Recommendations
*
sı
Supporting
Information
ABSTRACT:
The
hydrodynamics
of electrochemical
CO
2
reduction
(CO
2
R) systems
is an insufficiently
investigated
area of research
that has broad
implications
on catalyst
activity
and selectivity.
While
most
previous
reports
are limited
to
laminar
and CO
2
-sparged
systems,
herein
we address
a wide
range
of hydrodynamics
via electrolyte
recirculation
systems.
We find that increased
hydrodynamics
at the electrode
surface
results
directly
in changes
to the ethylene
and methane
Tafel
slopes,
demonstrating
that mass
transport
is on equal
footing
with
catalyst
active
sites
in determining
reaction
mechanisms
and
the ensuing
product
distribution.
Mass
transport
is
traditionally
considered
to be in the purview
of systems-level
engineering,
yet the present
work
shows
that CO
2
R mechanistic
work
must
be considered
in the context
of the mass
transport
conditions.
We extend
our analysis
to organic
coatings,
demonstrating
that the films
shield
the active
sites from
variability
in hydrodynamics
and increase
the residence
time of CO so
that it may be further
reduced
to desirable
products.
H
eightened
anthropogenic
atmospheric
CO
2
levels
motivate
the
development
of
processes
to
mitigate
CO
2
emissions
associated
with
climate
change.
1,2
Electrochemical
CO
2
reduction
(CO
2
R)
is being
pursued
as
one
solution
toward
a carbon-neutral
economy.
3,4
However,
many
challenges
remain,
including
those
associated
with
selectivity
for
multi-carbon
fuel
candidates.
Current
trends
focus
on
copper’s
crucial
ability
to form
appreciable
quantities
of carbon-coupled
products.
As
yet
it is not
possible
to reliably
tune
the
downstream
C
2+
products
produced,
such
as ethylene
and
ethanol,
and
the
mechanisms
defining
such
selectivities
are
not
firmly
established
(Scheme
1).
5
Even
on
a planar
catalyst,
perhaps
the
simplest
practical
system
for
mechanistic
studies,
there
are
many
effects
to
consider,
including,
but
not
limited
to,
concentration
polarization
effects
from
dissolved
CO
2
,
intermediates,
electrolytes,
fluctuations
in
local
pH,
and
the
faceting
and
restructuring
of the
electrode
surface.
6
Tafel
analysis,
derived
from
Butler
−
Volmer
kinetics,
is one
powerful
tool
that
has
long
been
used
to
aid
in
the
understanding
of
electrochemical
mechanisms.
By
analyzing
the
logarithm
of current
versus
applied
potential,
information
can
be
obtained
regarding
the
number
of proton
and
electron
transfers
occurring
before
the
rate-determining
step
(RDS)
(eq
1).
7
Through
better
mechanistic
understanding,
catalysts
and
systems
can
be
optimized
to
reduce
energy
barriers
and
alter
selectivity.
Recently,
it has
come
to
light
that
this
approach
provides
only
limited
reliable
insight,
in
part
due
to
over-
reliance
on
Tafel
slopes
and
biases
toward
reporting
cardinal
values.
8
Influences
from
mass
transport
can
obscure
observa-
tion
of
the
desired
Tafel
behavior.
CO,
an
intermediate
on-
path
to carbon-coupled
CO
2
R products,
has
been
used
as the
reactant
to
try
to
simplify
mechanistic
studies.
9
−
11
However,
such
strategies
cannot
replicate
the
catalyst
microenvironment
Received:
February
27,
2023
Accepted:
April
11,
2023
Published:
April
13,
2023
Scheme
1. Electrochemical
Reduction
of CO
2
to CO
Requires
Two
Proton
and Electron
Transfers
a
a
Reduction
of
CO
is proposed
to
be
the
rate-determining
step
for
producing
carbon-coupled
products,
but
the
exact
species
it is reduced
to upon
dimerization
are
still
under
investigation.
5
.
Letter
http://pubs.acs.org/journal/aelccp
©
2023
The
Authors.
Published
by
American
Chemical
Society
2185
https://doi.org/10.1021/acsenergylett.3c00442
ACS Energy
Lett.
2023,
8,
2185
−
2192
Downloaded via CALIFORNIA INST OF TECHNOLOGY on May 22, 2023 at 21:59:46 (UTC).
See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.
resulting
from
CO
2
/HCO
3
equilibrium,
pH
gradients,
and
equilibration
of
the
electrode
with
CO
as
a product,
highlighting
the
importance
of
directly
interrogating
CO
2
reduction
systems.
8
Equation
1: A simplified
equation
for the Tafel
slope,
α
−
1
,
can be derived
from
the Butler
−
Volmer
equation,
where
n
and
q
are the number
of electron
transfers
before
and
during
the RDS,
respectively.
n
q
Tafel slo
pe
60 mV dec
/2
1
=
+
(1)
Gas
diffusion
electrodes
and
gas
sparging
have
been
investigated
to
address
the
issues
of
low
CO
2
solubility
(34
mM)
and
low
mass
transport
in aqueous
CO
2
reduction.
12
Gas
diffusion
assemblies
enhance
CO
2
transport
to the
catalyst
by
decreasing
the
diffusion
path
length
of CO
2
to
the
electrode
surface,
but
they
inherently
create
a distribution
of
mass
transport
conditions
and
local
electrochemical
environments
that
impede
their
utilization
for
studying
catalyst
mecha-
nisms.
13,14
H-cells,
in
which
CO
2
is
typically
bubbled
across
the
electrode
surface,
are
widely
adopted
for
mechanistic
studies.
Their
limitations
in
elucidating
intrinsic
catalyst
behavior
include
the
documented
sensitivity
of product
distribution
to
the
bubble
rate
and
size.
15
−
17
Follow-up
work
on
copper
demonstrated
that
at
−
1.1
V vs RHE
a difference
in flow
rate
between
5 and
10
sccm
would
afford
a 20%
swing
in Faradaic
efficiency
in
favor
of
ethylene
over
hydrogen.
18
Even
with
vigorous
bubbling,
the
electrolyte
flow
closest
to the
electrode
is
essentially
stagnant,
leading
to
well-defined
Nernstian
diffusion
boundary
layers.
19
As
the
flow
rate
increases,
the
boundary
layer
becomes
thinner
and
reduces
the
diffusion
path
length
for
reactants
to
travel
from
the
bulk
to
the
surface
(Figure
1a).
In CO
2
R, the
increased
exchange
between
species
in
the
boundary
layer
and
the
bulk
solution
mitigates
deviations
in
pH
from
the
bulk
electrolyte
value.
20
While
bubble
formation
from
gaseous
products
can
disrupt
the
flow
field,
model
laminar
systems
have
been
successfully
deployed
to
modulate
CO
2
availability
via
flow
rate
and
to
facilitate
cascade
catalysis
using
multiple
CO
2
R
catalysts
(Ag
and
Cu).
21,22
Rotating
disk
electrodes
(RDEs)
also
serve
as laminar
systems
of
interest
for
inspecting
electrochemical
CO
2
reduction,
but
they
are
challenged
by
bubble
formation
on
the
electrode
surface.
23,24
Rotating
cylinder
electrodes
(RCEs)
mitigate
this
issue
by
operating
in a turbulent
regime
at nearly
all
rotation
rates
(Figure
1b).
While
the
uniformity
of
the
boundary
layer
is disrupted
in this
system,
the
mass
transfer
to
the
electrode
surface
can
still
be
described
well
by
mass
transfer
coefficient
values.
25
Such
studies
have
been
performed
on
copper
and
have
revealed
that
increased
transport
afforded
by
turbulent
flow
results
in a proportional
increase
in partial
current
densities
for
CO
(
j
CO
) and
decrease
in
rates
for
carbon-coupled
products.
26,27
This
result
supports
the
hypothesis
that
CO
engages
in
a dynamic
desorption
−
readsorption
to the
electrode
surface.
In
this
study,
we
explore
recirculation
flow
systems
in
a
range
of cell
geometries,
creating
a combination
of control
and
flexibility
that
complements
traditional
H-cell
and
RCE
reactors.
Studying
the
influence
of a range
of reactor
designs
and
mass
transport
conditions
elucidates
the
sensitivity
of
Tafel
analysis
to
external
factors
in
CO
2
R
catalysis.
Importantly,
the
rate-determining
step
in the
CO
2
reduction
reaction
is pushed
later
in
the
mechanism
with
increasing
convection.
Additionally,
our
observation
of
systematic
variation
in product
selectivity
with
hydrodynamics
conditions
provides
an
opportunity
to better
understand
our
team’s
recent
high-throughput
study
of copper
alloys.
This
study
revealed
a
scaling
relationship
between
j
C
d
2
H
d
4
and
j
CH
d
4
that
spanned
more
than
3 orders
of
magnitude
of
partial
current
densities
and
showed
that
organic-additive-derived
film
coatings
can
break
this
scaling
relationship.
28
The
use
of
hydrodynamics
to
observe
or alter
scaling
relationships
highlights
the
importance
of
evaluating
the
sensitivity
to
mass
transport
in
both
mechanistic
studies
and
reactor
design.
To
probe
the
influence
of hydrodynamics
on
catalytic
behavior,
we
focus
on
two
main
cell
designs:
H-cells
similar
to those
reported
by
the
Jaramillo
and
Ager
laboratories
and
a recirculation-based
flow
cell
designed
by
the
Gregoire
group
as part
of the
high-throughput
analytical
electrochemistry
(ANEC)
system
(see
the
Support-
ing
Information
(SI),
Figure
S1a,b,
for
cell
geometries).
15,16,29
ANEC
was
designed
to deliver
CO
2
-saturated
electrolyte
to
the
electrode
via
an
inlet
oriented
20
°
from
horizontal
toward
the
catalyst
surface,
which
enhances
electrolyte
convection
and
mitigates
bubble
formation
on
the
electrode.
29
Recirculating
flow
conditions
were
also
incorporated
in a more
traditional
H-
cell
with
2 different
flow
paths:
a horizontal
inlet,
hereafter
referred
to as the
parallel-flow
H-cell
(Figure
S1c),
and
a 20
°
inlet
oriented
toward
the
working
electrode,
hereafter
referred
to as the
angled-flow
H-cell
(Figure
S1d).
COMSOL
calculations
were
performed
on
the
three
variations
of recirculation
cells
to quantify
the
hydrodynamics
(Figure
2a
−
d)
and
evaluate
its
influence
on
extraction
of
desorbed
reaction
intermediates
(Figure
2e
−
h).
The
modeling
of the
cell
with
CO
2
-sparged
electrolyte
is challenged
by
the
ill-
defined
hydrodynamic
conditions,
although
we
expect
that
the
parallel
flow
of gas
will
produce
hydrodynamics
that
are
most
similar
to
those
of
the
parallel-flow
H-cell.
While
the
fluid
dynamics
calculations
qualitatively
describe
the
convection
of
CO
2
to
the
electrode
surface,
they
do
not
directly
provide
information
about
the
residence
time
or
wicking
of
reaction
products
from
the
electrode
surface.
To
resolve
this,
the
flux
of
a chemistry-agnostic
species
was
monitored
using
a Dirichlet
Figure
1. Graphical
representation
of the change
in concentration
of surface
species
due to electrolyte
flow near the electrode
surface
with a) laminar
flow and b) turbulent
flow.
Relative
velocities
and
depth
are indicated
by arrow
size and color,
respectively
(not
to
scale).
ACS Energy
Letters
http://pubs.acs.org/journal/aelccp
Letter
https://doi.org/10.1021/acsenergylett.3c00442
ACS Energy
Lett.
2023,
8,
2185
−
2192
2186
boundary
condition
to
mimic
that
of
surface-generated
CO
(Figure
2e
−
h).
We
find
that
the
loss
of
the
species
is
commensurate
with
the
simulated
hydrodynamics;
namely,
the
flux
away
from
the
electrode
surface
is largest
in the
ANEC
cell,
followed
by
the
angled-flow
H-cell,
and
then
smallest
in
the
parallel-flow
H-cell.
This
is consistent
with
experimental
findings
using
potassium
ferricyanide
reduction,
with
a 3-fold
smaller
average
diffusion
layer
in the
ANEC
cell
and
angled-
flow
H-cell
than
in the
parallel
H-cell
(see
Figure
S2,
Table
S1).
The
relative
convection
in each
cell
can
be
further
compared
using
the
relationship
established
by
RCEs
between
transport
and
j
CO
. We
observe
that
the
differences
in
convection
are
commensurate
with
the
observations
from
Figure
2,
where
ANEC
is the
highest
transport
and
the
H-cell
is the
lowest
(Figure
3).
Since
the
total
CO
2
R current
density
is comparable
among
the
systems
in
question,
we
attribute
the
changes
in
selectivity
to the
changes
in intermediate
transport
and
not
to
differences
in CO
2
availability
(Figure
S3a).
The
differences
in
partial
current
density
are
more
apparent
at
higher
over-
potentials,
likely
due
to increased
activity,
which
is consistent
with
prior
studies.
30
The
changes
in
CO
partial
current
densities
motivate
the
comparison
of their
potential-dependent
performance
toward
products
more
reduced
than
CO,
commencing
with
j
C
d
2
H
d
4
in
Figure
4a
and
continuing
with
additional
products
in Figure
5a
−
c
(for
all
selectivity
data,
see
Figure
S4,
Table
S2).
To
quantify
the
differences
in product
distribution
in the
context
of Tafel
analysis,
the
recently
reported
Bayesian
framework
was
adopted
to generate
the
distribution
of Tafel
slopes
for
each
of
the
three
H-cell
flow
conditions
as well
as for
ANEC.
31
The
maximum
likelihood
Tafel
slope
values
for
the
ANEC
cell
are
39
and
41
mV
dec
−
1
, respectively.
8
These
values
are
markedly
different
than
those
observed
using
data
established
by
Kuhl
et
al.
for
the
sparged
H-cell,
99
and
80
mV
dec
−
1
, respectively
(Figure
4a,
Figure
S5a).
15
Applying
the
same
analysis
to the
parallel-flow
and
angled-
flow
H-cells,
we
observe
that
increasing
convection
results
in
increasing
j
CO
and
decreasing
Tafel
slopes
for
ethylene
and
methane
(Figure
4b,
Figure
S5b,
full
probability
distributions
in Figures
S6
−
S13).
This
observed
relationship
between
CO
production
and
Tafel
slopes
may
be
rooted
in the
sensitivity
of
methane
and
ethylene
production
to the
surface
concentration
of
CO.
30
We
turn
to
Tafel
analysis
to
characterize
how
the
apparent
RDS
of copper-catalyzed
CO
2
reduction
to ethylene
changes
with
convection.
The
results
suggest
a shift
from
an
initial
electron
transfer
with
the
sparged
H-cell
(expected
120
Figure
2. COMSOL
simulations
of electrolyte
velocity
in the a) parallel-flow
H-cell,
b) angled-flow
H-cell,
and c) ANEC
cell are shown
as
cross-sections
through
the middle
of each
cell, as shown
for ANEC
in d), where
the electrolyte
flow is from
left to right,
the electrode
is on
the bottom,
and the vertical
axis is the height
above
the electrode
surface.
The distribution
in flow velocity
within
1 mm of the electrode
surface
is markedly
different
among
the 3 flow
geometries.
By setting
the concentration
of a test species
at the electrode
to 1 mM,
the
simulated
test species
flux from
the electrode
surface
can be estimated
for each
flow geometry
in e), f), and g), respectively.
This
is shown
schematically
for ANEC
in h). The increasing
transport
from
a) to b) to c) therefore
has a marked
impact
on CO extraction
from
the
electrode
surface.
Figure
3. Measured
partial
current
densities
for CO in each
cell
geometry
with respect
to potential.
Each
data point
corresponds
to
an individual
experiment,
with the exception
of the sparged
H-cell
data,
reproduced
from
ref 15, which
is an average
of three
experiments.
15,16,29
The
shaded
regions
correspond
to a 95%
confidence
interval
around
a trend
line fitting
function
calculated
via locally
weighted
scatterplot
smoothing
(LOWESS;
see SI for
calculation
details).
ACS Energy
Letters
http://pubs.acs.org/journal/aelccp
Letter
https://doi.org/10.1021/acsenergylett.3c00442
ACS Energy
Lett.
2023,
8,
2185
−
2192
2187
mV
dec
−
1
for
n
= 0 and
q
= 1, see
Scheme
1 and
eq
1) toward
a
later
step
in the
CO
2
R mechanism
with
ANEC
(
n
≥
1),
such
as a subsequent
chemical
step
(
q
= 0, Tafel
slope
≤
60
mV
dec
−
1
) or
electron
transfer
step
(
q
= 1, Tafel
slope
≤
40
mV
dec
−
1
). Furthermore,
upon
changing
the
electrolyte
in ANEC
from
0.1
to
0.25
M
KHCO
3
, we
find
no
apparent
change
in
methane
and
ethylene
Tafel
slopes,
indicating
that
the
effect
is
likely
driven
by
convection
and
not
changes
in local
pH,
which
are
suppressed
by
the
increased
buffer
strength
(Figure
S14).
The
lowered
Tafel
slopes
provide
an
opportunity
for
substantially
increasing
partial
current
densities
with
a modest
increase
in overpotential.
Figure
4a
indicates
that
the
present
cell
designs
cannot
fully
realize
that
opportunity
due
to a mass
transport
limit
when
the
partial
current
density
reaches
ca.
10
mA
cm
−
2
; we
expect
future
work
on
scalable
reactor
design
can
leverage
the
impact
of
convection
on
the
Tafel
slope
to
increase
partial
currents
toward
highly
reduced
products.
Given
the
systematic
variation
in observed
Tafel
slope
with
level
of convection,
we
identify
two
predominant
explanations
for
the
underlying
cause:
increased
CO
2
availability
or
decreased
surface
coverage
of
CO
with
increased
hydro-
dynamics.
The
former
explanation
implies
that
insufficient
Figure
4. a) Tafel
plots
for ethylene
comparing
data from
ANEC
and the sparged
H-cell.
A small
plateau
exists
on the bottom
of each
ANEC
plot due to the noise
floor
of the instrument.
Each
data point
corresponds
to an individual
experiment,
with the exception
of the sparged
H-
cell data,
reproduced
from
ref 15, which
is an average
of three
experiments.
15
b) Probability
density
curves
for ethylene
Tafel
slope
values
using
Bayesian
statistics
for each
cell geometry
investigated.
An identical
figure
for methane
exists
in Figure
S5.
Figure
5. Distribution
of gaseous
products
shown
as stack
plots
for a) bare
15
and b) 1-Br
2
-modified
34
polycrystalline
Cu electrodes
with
respect
to potential
in a CO
2
-sparged
H-cell.
The product
distributions
from
c) bare and d) 1-Br
2
-modified
polycrystalline
Cu electrodes
in
ANEC
are shown
with
smooth
trend
lines
calculated
using
LOWESS
(see SI for calculation
details).
Each
data
point
corresponds
to an
individual
experiment,
with
the exception
of the sparged
H-cell
data,
reproduced
from
ref 15, which
is an average
of three
experiments.
15
The presence
of 1-Br
2
mitigates
the influence
of the different
flow conditions
in the H-cell
and ANEC
cell.
ACS Energy
Letters
http://pubs.acs.org/journal/aelccp
Letter
https://doi.org/10.1021/acsenergylett.3c00442
ACS Energy
Lett.
2023,
8,
2185
−
2192
2188