S-1
Supporting Information
Macroscale and nanoscale photoelectrochemical behavior of p-type Si(111) covered by a single
layer of graphene or hexagonal boron nitride
Annelise C. Thompson, Burton H. Simpson, and Nathan S. Lewis*
Division of Chemistry and Chemical Engineering, California Institute of Technology, Pasadena,
CA, 91125, United States
*Corresponding author:
nslewis@caltech.edu
S-2
Table of Contents
X-ray photoelectron spectra of p-SiO
x
, p-Si/Gr, and p-Si/h-BN electrodes
S2
Ultraviolet photoelectron spectra of p-Si-H, p-Si/Gr and p-Si/h-BN electrodes
S4
Raman maps for typical Gr and h-BN samples
S5
UV/Vis spectra for typical Gr and h-BN samples
S7
Cyclic voltammetry of macroscale p-SiO
x
electrodes
S8
Additional fits for macroscale p-Si/Gr
V
oc
data
S9
SECCM line scans on Gr and h-BN
S10
Effective redox potential
S11
Redox couple compatibility with nanoscale junctions
S12
Supporting Data
X-ray photoelectron spectra of p-SiO
x
, p-Si/Gr, and p-Si/h-BN electrodes
Figure S1. X-ray photoelectron spectra of p-Si surfaces. (A) The Si 2p region of a p-SiO
x
electrode after fabrication. (B) The Si 2p region of a p-Si/Gr electrode after fabrication. (C) The
Si 2p region of a p-Si/h-BN electrode after fabrication. The peaks at 99.2 and 99.8 eV are
assigned to the Si 2p
1/2
and 2p
3/2
peaks, respectively, whereas the small peak at 102.5 eV is
attributable to the oxide formed during the fabrication process.
The oxide introduced to the surface through the fabrication process was quantified using a
simple substrate-overlayer model:
푑
=
휆
표푣
sin
휃
[
ln
(
1
+
퐼
표
푆푖
퐼
표
표푣
∗
퐼
표푣
퐼
푆푖
)
]
where
λ
ov
is the attenuation factor through the oxide overlayer (2.6 nm),
1
θ
is the angle from the
surface to the detector,
is the instrument normalization factor expected from
퐼
표
푆푖
퐼
표
표푣
uncontaminated Si and SiO
2
samples, which was taken as 1.3 for this instrument,
is the
퐼
표푣
measured intensity of the silicon oxide peak found around 103 eV, and is the measured
퐼
푆푖
intensity of the silicon. Using this equation, an oxide of ~ 0.22 nm in thickness was detected on
the p-Si/Gr electrodes and an oxide of ~ 0.22 nm in thickness was detected on the p-Si/h-BN
electrodes. A monolayer of oxide is about 0.35 nm thick, so the average electrode with a 2D
overlayer tested in this work had less than a monolayer (ML) of oxide growth after the
S-3
fabrication process. Substantially more oxide of ~0.50 nm in thickness was detected on the p-
SiO
x
electrodes, equivalent to 1.5 ML.
Figure S2. X-ray photoelectron spectra of A) the N 1s and B) the B 1s region of a typical h-BN
sample and C) the C 1s region of a typical Gr sample.
Figure S2 shows the X-ray photoelectron spectra for typical p-Si/Gr and p-Si/h-BN samples after
an anneal under forming gas. Figure 2a-b displays the N 1s and B 1s regions of a h-BN sample
with single peaks at 398.3 eV and 190.6 eV respectively, yielding a 1:1 ratio of B:N from the
peak area ratio, after adjusting for elemental relative sensitivity factors. The C 1s region of a p-
Si/Gr sample in Figure 2c had three peaks at 284.4 eV, 285.0 eV, and 286.1 eV, ascribable to
the sp
2
carbon bonds of graphene and the residual C-C and C-O bonds of the PMMA transfer
support, respectively.
S-4
Ultraviolet photoelectron spectra of p-Si-H, p-Si/Gr and p-Si/h-BN electrodes
Figure S3. Ultraviolet photoelectron spectra of annealed p-Si-H, p-Si/Gr and p-Si/h-BN samples,
corrected with reference to a sputter-cleaned Au sample.
The work function of each sample was calculated by extrapolating the slope of the secondary-
electron cutoff to its intercept with the x-axis, and subtracting that value from the excitation
energy of He I (21.21 eV). The data before and after annealing are shown in Table S1 and are
the results of measurements on at least four samples. The magnitude of the work function for p-
Si-H was different from the value calculated from the dopant density (5.02 eV) but is consistent
with the band-bending and dipole expected on this surface
2
. The work function and surface
dipole of the p-SiO
x
surface was the same as the p-Si/Gr and p-Si/h-BN surfaces, within error.
Table S1. Secondary electron cutoff, dipole, and work function of samples*
Before anneal
After anneal
SEC
E
VBM
S
δ
W
F
S
SEC
E
VBM
S
δ
W
F
S
Sample
eV
eV
eV
eV
eV
eV
eV
eV
p-Si–H
16.94
0.53(1)
-0.37(9)
4.27(10)
16.80
0.51(2)
-0.25(5)
4.40(7)
p-SiO
x
16.60
0.29(20)
-0.27(16)
4.60(6)
16.28
0.25(6)
+0.01(13)
4.92(17)
p-Si/Gr
16.51
0.40(3)
-0.07(2)
4.70(2)
16.47
0.58(6)
+0.14(5)
4.73(7)
p-Si/
h-BN
16.75
0.49(2)
-0.22(10)
4.45(9)
16.44
0.40(2)
0.00(8)
4.77(6)
*Standard deviations for E
VBM
S
,
δ,
and W
F
S
are shown in parentheses and are in units of
hundredths of an eV (ceV).
S-5
Raman maps for typical Gr and h-BN samples
Figure S4. Raman spectra of a Gr sample on 300 nm thick SiO
2
. (A) A typical Raman spectrum
for a Gr sample. (B) Contour plot of I
G
/I
2D
for a typical monolayer Gr sample from a Raman map.
The plot displays data for a 75x75
μm
region on the sample.
Figure S4 shows both the typical Raman spectrum for a spot on a graphene sample and the
contour plot of the intensity of the G peak to the intensity of the 2D peak (I
G
/I
2D
) ratio. The 95%
of values observed for this ratio were between 0.4 and 0.5, indicating the presence of intact
monolayer Gr.
3
Some regions of lower intensity were observed, but the majority of the area
was covered with pristine monolayer graphene.
S-6
Figure S5. Raman spectra of an h-BN sample on 300 nm SiO
2
. (A) A typical Raman spectrum
for a h-BN sample. (B) Contour plot of the peak intensity for a typical monolayer h-BN sample
from a Raman map. Although the intensity varied across the sample, the 1370 cm
-1
peak was
present in each spectra. The plot displays data for a 75x75
μm
region on the sample.
In Figure S5, the Raman spectrum of h-BN had a single peak at 1367 cm
-1
and thus the contour
plot showed only the relative intensity of different spots on a sheet of h-BN.
6
The intensity varied
across the h-BN sheet, but no voids were visible at this scale. Each contour plot exceeded the
size expected for a single grain of Gr or h-BN, given an average diameter of 50
μm
and 5
μm,
respectively. The scale for each plot was set so that regions with no intensity should be bright
blue, to differentiate between regions of variable and no intensity.
S-7
UV/Vis spectra for typical Gr and h-BN samples
Figure S6. UV/Vis spectra for (A) Gr and h-BN samples on quartz slides and (B) Tauc plots for
h-BN and Gr. The optical band gap of h-BN was determined to be 6.07 eV from extrapolation of
the linear region of the Tauc plot to an intercept with the x-axis. The fit is shown in the inset. The
same plot for the graphene showed no sharp increase that would indicate of the presence of a
band gap.
Figure S6 displays the UV/vis spectra for Gr and h-BN on quartz. h-BN shows almost no
absorption across the visible light region (400-700 nm) while Gr absorbs 2.7% of light in the
same region. The optical band gap of h-BN was determined to be 6.07 eV from extrapolation of
the linear region of the Tauc plot to an intercept with the x-axis, which is within the typical
bounds for the polycrystalline material.
8
The same plot for the graphene showed no such sharp
increase that would indicate of the presence of a band gap.
S-8
Figure S7. Photocurrent density vs potential (
J
-
E
) behavior of p-Si–H and p-SiO
x
electrodes in
contact with Cp
2
Fe
+/0
(A) and Cp
2
Co
+/0
(B), in CH
3
CN–0.50 M LiClO
4
under 100 mW cm
-2
of
ELH-type simulated solar illumination. The dashed lines show scans of the same electrodes
without illumination.
Figure S7 presents the representative macroscale current density versus potential (J-E)
behavior for p-Si–H and p-SiO
x
electrodes in the dark and under illumination. The p-SiO
x
electrodes exhibit similar behavior to the p-Si–H electrodes, although the magnitude of
V
oc
and
the fill factor as lower for p-SiO
x
.
S-9
Figure S8. Possible linear fits for
V
oc
measurements on macroscale p-Si/Gr electrodes
extrapolated to the plateau for the macroscale p-SiO
x
electrodes, with the effective solution
potential measured vs a saturated calomel electrode (SCE). Fit #1 is the linear fitting of the
V
oc
values in contact with CoCp*
2
+/0
, CoCp
2
+/0
, Me
8
Cp
2
Ni
+/0
, and MV
2+/+•
with an R
2
of 0.970 and
excludes NiCp
2
. Fit #2 excludes Cp*
2
Co
+/0
as well as NiCp
2
and has a lower R
2
of 0.773,
whereas Fit #3 shows where a plateau in
V
oc
could be, given the maximum
V
oc
value for p-Si/Gr
electrodes as indicated by the measurements in contact with CoCp
2
+/0
.
Three possible fits of the data are shown in Figure S8. Each fit is a linear fit and extrapolation of
the data including direct weighting of the standard deviation. Fit #1 is the fit shown in the main
text, with a slope of -0.24. Fit #1 and #2 are shown with a theoretical plateau at 400 mV, which
is the plateau value of
V
oc
for p-SiO
x
and p-Si/h-BN measured in this work, while Fit #3 takes
into consideration that the plateau may be lower for p-Si/Gr. Fit #3 has a plateau at 300 mV, the
mean value for p-Si/Gr in contact with Cp*
2
Co
+/0
. Given data for previous work modifying a Si
surface with a monolayer of carbon atoms, each set of data here is expected to have a parallel
slope, making Fit #2 (slope of -0.42) a viable fit when compared with the slopes from the other
macroscale electrodes (slopes of -0.41-0.45, Table S2). Further macroscale measurements at
more negative potentials would help to clarify the behavior of p-Si/Gr under these conditions. If
the additional measurements gave credence to Fit#2, the differences between Gr and h-BN
highlighted in the main text would still be giving rise to a much more negative onset in the slope
of V
oc
, a conclusion which is still corroborated by the nanoscale measurements performed in this
work
.
S-10
Table S2. Linear fit data for macroscale measurements from main text*
Sample
Slope
X-axis intercept (V)
R
2
p-Si–H
-0.414(33)
0.423(174)
0.969
p-SiO
x
-0.431(104)
0.073(12)
0.800
p-Si/Gr
-0.244(25)
-0.262(33)
0.970
p-Si/h-BN
-0.454(67)
-0.060(6)
0.919
*The error for slope is given in milli-units as slope is unitless. The error for the x-intercept is
given in mV.
S-11
Figure S9. SECCM line scan of
V
oc
in 10
μm
steps on Gr (upper) and h-BN (lower) in contact
with Cp
2
Co
+/0
, Cp
2
Ni
+/0
, and Me
2
Cp
2
Fe
+/0
, respectively. Each value is the average of six
measurements at the same spot. Standard deviations are represented by the shaded regions
surrounding each scan.
The measurements in Figure S9 were taken in consecutive 10
μm
steps, although each scan
with a different redox couple was collected on a separate region of the sample, to avoid cross
contamination of the redox species. For p-Si/Gr, the measurements for Cp
2
Ni
+/0
and
Me
2
Cp
2
Fe
+/0
were very similar and showed little spatial variation across the measured region.
The
V
oc
values for pSi/Gr in contact with Cp
2
Co
+/0
were distinctly higher and showed a variation
between steps of a magnitude consistent with expectations for line defects or grain boundaries
(10-30 mV).
4
The
V
oc
of the p-Si/h-BN sample in contact with the same redox couples shifted
over a much wider range of potentials but exhibited a relatively low variation between steps. The
measurements of
V
oc
in contact with Cp
2
Ni
+/0
showed the largest standard deviations across the
S-12
surface on both the p-Si/Gr and p-Si/h-BN, consistent with expectations given the low solubility
of this redox couple on the test electrolyte solution.
S-13
S1. Effective solution potential
As the solubility of the redox couples in this work were not identical, the concentrations
used in each solution varied enough to affect the resulting measurements of
V
oc
. Grimm et. al.
have previously explained a method by which the solution potential (
E
(A/A
-
)) can be converted
to the effective solution potential (
E
eff
(A/A
-
)) to account for the differences in concentration of the
minority acceptor, using a derivative of the Nernst equation. The full derivation and explanation
for these shifts can be found in Ref. 31. The effective cell potential (
E
eff
) was calculated using
the equation
퐸
푒푓푓
(
A
A
―
)
=
퐸
(
A
A
―
)
―
k
B
T
q
ln
[
퐴
푒푓푓
]
[
퐴
]
where
E
(A/A
-
) was the cell potential of the redox couple vs. SCE, [A] was the minority acceptor
concentration, and [A
eff
] = 10 mM. Table 1 shows the concentrations actually used to make each
solution, but the values for
V
oc
are normalized to [A
eff
] = 10 mM.
S2. Redox couple compatibility with nanoscale junctions
Carrier transport accounts for much of the incongruency between macroscale and
SECCM measurements, but SECCM presents unique experimental challenges when working
with some specific redox couples. Measurements with
Cp*
2
Co
+/0
and Cp
2
Ni
+/0
were respectively
impacted by issues associated with the reactivity and solubility of these compounds. Because of
the extremely negative potential of the Cp*
2
Co
+/0
couple, Cp*
2
Co
0
can be oxidized to Cp*
2
Co
+
by
reaction with oxygen. The need for a solvent saturated atmosphere to ensure stability of the
liquid junction requires that the SECCM be operated in a flush box rather than a recirculating
glove box, because the evaporated solvent would poison the catalyst used to remove water in a
recirculating system. However, the flush box had a relatively higher oxygen content, resulting in
faster conversion of Cp*
2
Co
0
, and consequently producing a positive shift in
E
eff
for this couple
during experiments. This shift was observed to occur so rapidly that only a few measurements
could be taken before the solution potential was no longer well-defined by Cp*
2
Co
+/0
. During
bulk electrolysis of Cp
2
Ni
0
, the Cp
2
Ni
+
was observed to have very limited solubility (~ 1 mM) in
the LiClO
4
solution. While enough Cp
2
Ni
+
could be generated by bulk electrolysis to produce a
stable
E
eff
, local Cp
2
Ni
+
concentrations at the interface went above the solubility limit during
cathodic operation. This insolubility would result in unstable interfaces and
V
oc
measurements
during extended operation at a single point. These limitations can in principle be overcome by
careful selection of redox couples, but the need for several facile redox couples across a wide
range of potentials necessitated the use of these non-optimized redox species.
References
1.
Hochella, M. F.; Carim, A. H., A reassessment of electron escape depths in silicon and thermally
grown silicon dioxide thin films.
Surface Science
1988,
197
(3), L260-L268.
2.
Gleason-Rohrer, D. C.; Brunschwig, B. S.; Lewis, N. S., Measurement of the Band Bending and
Surface Dipole at Chemically Functionalized Si(111)/Vacuum Interfaces.
The Journal of Physical
Chemistry C
2013,
117
(35), 18031-18042.
3.
Ferrari, A. C., Raman spectroscopy of graphene and graphite: Disorder, electron–phonon
coupling, doping and nonadiabatic effects.
Solid State Communications
2007,
143
(1–2), 47-57.
S-14
4.
Long, F.; Yasaei, P.; Sanoj, R.; Yao, W.; Král, P.; Salehi-Khojin, A.; Shahbazian-Yassar, R.,
Characteristic Work Function Variations of Graphene Line Defects.
ACS Applied Materials & Interfaces
2016
.