Supporting Information
Optical and electrochemical effects of H
2
and
O
2
bubbles
at
upward-facing Si
photoelectrodes
Paul
A. Kempler
1
, Zachary P.
Ifkovits
1
, Weilai Yu,
1
Azhar I. Carim,
1
Nathan S.
Lewis
1,2*
1
Division
of
Chemistry
and
Chemical
Engineering,
127-72,
210
Noyes
Laboratory,
California
Institute of
Technology, Pasadena, CA 91125
2
Beckman Institute, California Institute
of Technology, Pasadena, CA 91125
*Corresponding Author: nslewis@caltech.edu
1
Electronic
Supplementary
Material
(ESI)
for
Energy
&
Environmental
Science.
This
journal
is
©
The
Royal
Society
of
Chemistry
2020
Materials:
All
chemicals
were
commercially
available
and
used
as received.
Potassium
ferrocyanide(II)
trihydrate
(K
4
Fe(CN)
6
,
ACS
Reagent
99%),
was
obtained
from
Aldrich,
gallium-indium
eutectic
(GaIn,
99.99%,
metals
basis)
was
obtained
from
Alfa
Aesar,
and
concentrated
ammonium
hydroxide
(NH
4
OH,
28%-30%)
was
obtained
from
JT
Baker.
Hydrochloric
acid
(HCl,
ACS
grade
36.5-38%),
acetone,
and
isopropyl
alcohol
were
obtained
from
Millipore.
Buffered
oxide
etchant
(6:1
(v/v)
40%
NH
4
F
to
49%
HF)
was
obtained
from
Transene
Inc
and
hydrogen
peroxide
(H
2
O
2
,
ACS
grade
30%)
was
obtained
from
Macron
Chemicals.
Sulfuric
acid
(H
2
SO
4
,
TraceMetal
grade)
was
obtained
from
Fisher
Scientific
and
potassium
hydroxide
(KOH,
99.98%
metals
basis)
was
obtained
from
ACROS
Organics.
Solutions
were
diluted
to
the
specified
concentration
with
water
having
a resistivity
of
18.2
MΩ·cm,
obtained
from
a Millipore
deionized
(DI)
water
system.
N-type
Si
wafers
with
a
resistivity
of
0.4
Ω-cm
and
p-type
Si
wafers
with
a resistivity
of
10–20
Ω-cm,
having
diameters
of
100
mm,
thicknesses
of
525
μm,
and
<100>
orientation,
were
obtained
from
Addison
Engineering.
Fumasep
FAAM-15
and
Nafion
TM
117
were
obtained
from
Fuel
Cell
Store
and
nickel
wire
(Ni,
99.5%)
with
a diameter
of
0.5
mm,
was
obtained
from
BeanTown
Chemical.
Calomel
(CHI150)
and
mercurous
oxide
(CHI152)
reference
electrodes
were
obtained
from
CH
Instruments.
Preparation
of
μW
substrates:
Silicon
wafers
were
cleaned
with
acetone
and
isopropyl
alcohol,
exposed
to
a hexamethyldisilazane
primer,
and
spin-coated
with
S1813
photoresist
(Shipley)
at
4000
rpm.
The
resist
layer
was
photolithographically
patterned
via
UV-exposure
through
a
chrome
mask
which
had
a square
grid
of
circular
holes
6 μm
in
diameter
and
14
or
28
μm
in
spacing.
Al
2
O
3
etch
masks,
125
nm
in
thickness,
were
evaporated
into
the
exposed
hole
array
via
electron
beam
evaporation
at
1 Å
s
-1
.
Substrates
were
cleaved,
mounted
onto
a 6 in
diameter
Si
2
carrier
wafer
with
thermally
conductive
oil,
and
loaded
into
an
Oxford
Instruments
Dielectric
System
100
ICP-RIE.
Microwires
were
formed
via
deep
reactive-ion
etching,
DRIE,
of
Si
at
-
120-130°C
in
a SF
6
/O
2
plasma
at
a capacitively
coupled
power
of
3 W
and
an inductively
coupled
power
of
900
W.
Silicon
was
cleaned
prior
to metallization
via
a modified
Radio
Corporation
of
America
(RCA)
standard
clean
1 (5:1:1
(vol)
H
2
O:NH
4
OH:H
2
O
2
at 70-75
°C)
for
at
least
10
min
followed
by
an RCA
standard
clean
2 (6:1:1
(vol)
H
2
O:HCl:H
2
O
2
at 65-70
°C)
for
at
least
10
min.
The
samples
were
dipped
in
HF
between
the
cleaning
steps,
which
also
resulted
in the removal of the Al
2
O
3
etch
mask.
Fabrication
of
electrodes:
Metallization
was
performed
in an Orion
Series
Sputtering
system
(AJA)
at a base
pressure
of
10
-7
torr.
Ni
was
sputtered
in a 5 mtorr
Ar
plasma
at a deposition
rate
of
~1.5
nm
min
-1
.
An
ohmic
back
contact
was
made
to
the
rear
side
of
the
Si
with
an In-Ga
eutectic
(Alfa
Aesar),
and
the
electrode
was
affixed
to a Cu-Sn
wire
via
conductive
Ni
or
Ag
epoxy
(Ted
Pella).
The
electrode
was
sealed
into
a piece
of
6 mm
diameter
borosilicate
glass
tubing
using
a chemically
resistant
epoxy
(Hysol
9460)
that
was
cured
for
>12
h at
room
temperature. The electrode
areas were measured in ImageJ with a ruler serving as a
scale bar.
Scanning-electron
microscopy:
Scanning-electron
micrographs
(SEMs)
were
obtained
with
a
FEI
Nova
NanoSEM
450
at an
accelerating
voltage
of
10.00
kV
with
a working
distance
of
5
mm
and
an
in-lens
secondary
electron
detector.
Plan-view
SEMs
were
acquired
with
a resolution
of
14
pixels
μm
-1
.
Cross-sectional
SEMs
were
acquired
from
perspective
85°
away
from
the
substrate normal
with a resolution of 21 pixels μm
-1
.
Mass-transport velocity measurements:
The
current
derived from oxidation of Fe(CN)
6
4–
was
calculated based on the
change in absorptivity of the electrolyte in the cell before vs after a bulk
electrolysis. The electrolyte was not stirred during testing but
was
vigorously stirred by a
3
magnetically-powered Teflon stir bar prior to sampling
the electrolyte. A
calibration was
performed at
a large area Ni coil held potentiostatically
at 1.3 V vs
RHE. At this potential, the
current associated with oxygen evolution was
negligible and
the current derived from Fe(CN)
6
4–
oxidation was limited
by mass transport.
Figure S7
presents the change in absorptivity
at 420 nm
as a function
of charge passed for a series
of such experiments. The concentration
of Fe(CN)
6
3–
in the cell was calculated
as Q / nFV
where Q is the total charge passed at 1.3 V
vs. RHE, n is
the number
of
electrons
required for the oxidation
(1),
F is Faraday’s constant
(96485 C mol
-1
)
and V is
the volume
of electrolyte in the
cell
during oxidation (0.097 L or 0.047 L
for
cells
incorporating the upward-facing
and downward-facing electrodes, respectively). The extinction
coefficient of 1059 M
-1
cm
-1
was calculated from a
linear regression of the measured absorbances
and calculated concentrations.
Sample characterization
: Absorbance spectra were collected on a
Agilent 8453 UV-vis
Spectrometer using polystyrene cuvettes
with a
path length of 1 cm. Optical constants were
measured by use of a variable-angle spectroscopic
ellipsometer with
a rotating analyzer (J.A.
Woolam Co., Inc.). Measurements were recorded at
angles of incidence
of 60°,
65°, and 70° in
5
nm increments
in wavelength over
a range from 300 to 1000 nm.
Reflectance coefficients
were
calculated from the best-fit
optical constants and
thicknesses for the films.
Image processing:
Images of downward-facing
photoelectrodes were processed in
MATLAB.
The manually defined
electrode area
restricted the pixel area for data collection and
was used to
calibrate the pixels per mm
2
scale at the electrode surface. The location and
diameter of bubbles
were recorded for each image,
and
the fractional coverage was calculated
relative to
the
geometric electrode area. Manual quantification
of bubbles was supplemented by automatic
detection of
similarly sized bubbles
using
a Hough transform.
4
Convection due to thermal gradients:
A solar fuels
devices
exhibiting absorbing 90% of the
solar spectrum and converting
solar energy to fuel at an efficiency of 10% will require an
outgoing flux of 8.1
⨉
10
2
W m
-2
of
radiation
and heat at steady state. In the absence of bubble-
driven convection, this heat will be
removed
by conduction across temperature gradients and
convection due to resulting density gradients.
The maximum temperature
gradient expected to
occur for a large area, planar absorber, can be found
using Fourier’s Law, assuming
temperature
drops uniformly across the 0.01 m thick electrolyte
layer (Equation S1)
(S1)
푞
=
‒
푘
푑푇
푑푧
Where
q
is
the heat flux in W m
-2
,
T
is the temperature in
the
electrolyte
in
K,
k
is the thermal
conductivity of water, 0.60 W
m
-1
K
-1
, and
z
is the thickness coordinate extending positively
from zero at the surface of the top-facing light absorber. A temperature
gradient of 14 K across a
0.01 m electrolyte layer will sustain the removal of 8.1
⨉
10
2
W m
-2
of heat via conduction
alone. The relative contribution
of thermal convection to conduction
across a layer
of thickness
L
with temperature
difference
ΔT
is given by the
Rayleigh number, Ra (Equation S2),
(S2)
푅푎
=
푔
퐿
3
훽Δ푇
훼휈
where
g
is the acceleration due to gravity
(9.81 m s
–2
),
β
is the thermal expansion coefficient
(210
⨉
10
–6
K
–1
),
⍺
is the
thermal diffusivity (0.14
⨉
10
–6
m
2
s
–1
), and
ν
is the kinematic
viscosity (1.00
⨉
10
–3
Pa · s). A fluid layer between a solid surface and
a free surface
is predicted
to be stable for Ra < 1101.
1
Ra is 200 for
a 0.01 m liquid layer above an absorber
surface
that is
14 K greater than the ambient temperature, indicating that
the
fluid layer
will likely be stable.
Increased thicknesses of the liquid layer or increased irradiance could lead to
Ra > 1101 such
that thermal convection cannot be ignored.
5
Figure S1
: Schemes of cells used for (
A
) upward-facing and
(
B
) downward-facing
photoelectrochemical experiments in 0.50 M H
2
SO
4
(aq) and
1.0 M
KOH(aq).
Figure S2
: (
A
)
Calculated
spectral
reflectance
of Si, black line
, Si/TiO
2
(46 nm)/Ni(4.5 nm), red
line, and Si/TiO
2
(46 nm)/Ni(10.2 nm), red dashed-line, surfaces in
air. Film thickness and
spectral reflectance was calculated from measured (
B
)
ψ
(B) and
(
C
)
Δ.
6
Figure S3
: (
A
)
Cross-sectional scanning-electron micrograph of a
μW n-Si/TiO
2
/Ni sample.
(
B
)
Plan-view scanning-electron
micrograph of a μW
n-Si/TiO
2
/Ni
sample.
Figure S4
: (
A
)
J-E
behavior for upward-facing n-Si/TiO
2
/Ni
photoanodes in
stagnant 1.0 M
KOH(aq) recorded at
200 mV
s
-1
under
illumination from an
ELH lamp.
(
B
) Photographs of a
planar
n-Si/TiO
2
/Ni electrode at 60 mA cm
-2
(
C
) μW
6|14 n-Si/TiO
2
/Ni electrode at 65 mA cm
-2
Scale bars represent 1 cm.
7
Figure S5
: (
A
)
J-E
behavior for μW 6|14
(red)
and
μW 6|28 (black) upward-facing n-Si/TiO
2
/Ni
photoanodes in stagnant
1.0
M KOH(aq) recorded
at 200 mV s
-1
under illumination from
an ELH
lamp. Forward scans were
recorded
before (continuous) and reverse scans were recorded after
(dashed) 1 hour of continuous O
2
evolution at
1.8 V vs RHE. A
planar n-Si/TiO
2
/Ni photoanode
with 10 nm of Ni is shown in blue for comparison
of the optical
properties of the thicker
Ni films
(
B
)
J-t
behavior for μW 6|14 and
μW 6|28 photoanodes in (A) held at 1.6 V vs RHE.
8
Figure S6
: (
A
)
Cyclic
voltammogram of a
polished, 0.5 mm diameter Ni wire
embedded
in
epoxy in 1.0 M KOH(aq). The scan
rate
is specified in
mV s
-1
. (
B
) Cyclic voltammogram of 10
mM Fe(CN)
6
4–
in 1.0 M KOH(aq) at a polished, 0.5 mm
diameter
Ni wire embedded in
epoxy.
(
C
) Cyclic
voltammogram of 10 mM Fe(CN)
6
4–
in 1.0
M KOH(aq) at
a Au
wire
embedded
in
borosilicate glass
tubing.
(D)
Expanded view of voltammetry of 10 mM Fe(CN)
6
4–
in 1.0 M
KOH(aq) at a polished, 0.5 mm diameter
Ni wire embedded in
epoxy at a scan
rate of 200 mV s
-
1
.
(E)
Linear sweep voltammograms recorded at
an upward-facing μW 6|28 n-Si/TiO
2
/Ni
photoanode under illumination from a
630 nm light-emitting diode before (red)
and after (black)
a constant potential hold at 1.5 V vs RHE for 5 min.
9
Figure S7
: Electrolyte absorbance at 420 nm versus [Fe(CN)
6
3–
] as
measured via the anodic
charge passed
at a Ni wire in a solution of 10 mM Fe(CN)
6
4–
in
1.0 M KOH(aq).
Figure S8
: Calculated
j
Fe(III)
as a function
of
j
O2
as measured via quantification of Fe(CN)
6
3–
before and after bulk electrolysis
in 10 mM Fe(CN)
6
4–
, 1.0 M KOH(aq). Measurements
at planar,
blue squares,
μW 6|14, dark-red circles, and μW 6|28 light-red circles,
n-Si/TiO
2
/Ni.
Upward-
facing
measurements are represented as filled
markers, downward-facing measurements as open
markers. Error bars
represent one standard
deviation from three independent
experiments.
10
Figure S9
: Unaltered side-view photographs of illuminated gas streams
emanating from upward-
facing
photoanodes at
J
ph
= 20 mA cm
-2
.
Figure S10
: Time-dependent
J
ph
/
J
ph
bare
behavior
for
planar (A),
μW 6|28 (B), and μW 6|14 (C)
n-Si/TiO
2
/Ni electrodes in 10 mM Fe(CN)
6
4-
(aq),
1.0 M KOH(aq) as a function of illumination
intensity. Optical losses did not exceed
2% of the maximum photocurrent.
11
Figure S11
: (
A
)
Ray-tracing simulation for multiple surface-attached gas bubbles with
휃
b
=
160°.
(
B
)
Optical absorption distribution at
a surface
containing
bubbles with
R
b
= 100 μm and
휃
b
=
160°
for varied
center to center distances, recorded as a function
of position. The
absorption at individual
positions, as bins, was
normalized
to the total power incident on each
bin in the absence
of reflections or refractions.
12
Figure S12
: (
A
)
Graphical representation
of
Equation 2. (
B
) Plot of the
dimensionless
scattering
distance (
s/R
) versus the
ray position (
r
=
x/R
) for
bubbles with
휃
b
= 90°
and
160°.
The scattered distance,
s
(
r
),
can be determined
from the scattered angle off normal incidence and
height of
the ray intersection off the absorber surface. This height can be calculated as the sum of
H
and
R
b
sin(
휃’),
where
퐻
=
푅
푏
sin
(
휃
푏
‒
휋
2
)
such that
푠
=
푅
푏
(
sin
휃
'
‒
cos
휃
푏
)
tan
2
휃
'
To allow
easier integration on the projected
plane of incident the equation can
be
expressed in
terms of
r
, using the
fact that
r =
R
b
cos(
휃
b
)
and
sin
(
휃
b
)
=
(1
–
r
2
).
13
Figure S13
: (
A
)
J-t
behavior
of
an inverted p-Si/H electrode under chopped illumination. The
light source was turned off and
an image
was recorded
under
diffuse light at
times
marked with
arrows. (
B
)
J-t
behavior
of an inverted n-Si 6|14 electrode under chopped illumination.
Figure S14
: Extended
J
vs
t
data for
a n-Si 6|14/TiO
2
/Ni photoanode
under 1-Sun
(dotted
black)
and 2-Suns (dotted
red) illumination in 1.0 M KOH(aq). The potential
was
stepped from 1.4 V
(solid red), to 1.6 (solid blue, black),
to 1.8 V vs RHE
(solid purple). Sweeps to measure
J
ph
were
recorded before and after
individual constant potential holds.
14
1
.
2
1
.
4
1
.
6
1
.
8
E
/
V
v
s
R
H
E
-
5
0
5
1
0
1
5
2
0
2
5
J
/
m
A
c
m
-
2
Figure S15:
iR
-corrected
J
vs
E behavior
of an
n-Si/TiO
2
/Ni photoanode in 1.0 M
KOH(aq)
under illumination from an ELH lamp.
The forward and
reverse sweeps were recorded before
(black) and after (red)
a 1 hr constant potential
hold at 1.6 V
vs.
RHE. Original data
and data
corrected for the
iR
potential are shown
as
continuous
and dashed lines, respectively.
References
1.
W. M. Deen,
Analysis of Transport Phenomena,
Oxford University Press, Oxford,
1998
15