The harvesting of hot carriers produced by plasmon decay to generate electricity or drive a chemical reaction enables the reduction of the thermalization losses associated with supra-band gap photons in semiconductor photoelectrochemical (PEC) cells. Through the broadband harvesting of light, hot-carrier PEC devices also produce a sensitizing effect in heterojunctions with wide-band gap metal oxide semiconductors possessing good photostability and catalytic activity but poor absorption of visible wavelength photons. There are several reports of hot electrons in Au injected over the Schottky barrier into crystalline TiO2 and subsequently utilized to drive a chemical reaction but very few reports of hot hole harvesting. In this work, we demonstrate the efficient harvesting of hot holes in Au nanoparticles (Au NPs) covered with a thin layer of amorphous TiO2 (a-TiO2). Under AM1.5G 1 sun illumination, photoanodes consisting of a single layer of ∼50 nm diameter Au NPs coated with a 10 nm shell of a-TiO2 (Au@a-TiO2) generated 2.5 mA cm–2 of photocurrent in 1 M KOH under 0.6 V external bias, rising to 3.7 mA cm–2 in the presence of a hole scavenger (methanol). The quantum yield for hot-carrier-mediated photocurrent generation was estimated to be close to unity for high-energy photons (λ < 420 nm). Au@a-TiO2 photoelectrodes produced a small positive photocurrent of 0.1 mA cm–2 even at a bias of −0.6 V indicating extraction of hot holes even at a strong negative bias. These results together with density functional theory modeling and scanning Kelvin probe force microscope data indicate fast injection of hot holes from Au NPs into a-TiO2 and light harvesting performed near-exclusively by Au NPs. For comparison, Au NPs coated with a 10 nm shell of Al2O3 (Au@Al2O3) generated 0.02 mA cm–2 of photocurrent in 1 M KOH under 0.6 V external bias. These results underscore the critical role played by a-TiO2 in the extraction of holes in Au@a-TiO2 photoanodes, which is not replicated by an ordinary dielectric shell. It is also demonstrated here that an ultrathin photoanode (<100 nm in maximum thickness) can efficiently drive sunlight-driven water splitting.

1. S1
SUPPORTING INFORMATION
Harvesting Hot Holes in Plasmon-coupled Ultrathin Photoanodes
for High-Performance Photoelectrochemical Water Splitting
Ehsan Vahidzadeh,1
* Sheng Zeng,1
Kazi M Alam,1,2
Pawan Kumar,1
Saralyn Riddell,1
Narendra
Chaulagain,1
Sergey Gusarov,2
Alexander E. Kobryn2
and Karthik Shankar1
1
Department of Electrical and Computer Engineering, University of Alberta, 9211 - 116 St,
Edmonton, AB T6G 1H9, Canada.
2
Nanotechnology Research Centre, National Research Council Canada, 11421 Saskatchewan
Drive, Edmonton, Alberta, T6G 2M9, Canada
SECTION S1. MATERIALS AND METHODS
S1.1. Chemicals: KOH with a purity of 85% and Na2SO4 with a purity of 99% were purchased
from Sigma-Aldrich. Methanol with a purity of 99% was purchased from Fisher Chemicals. All
chemicals were used as received.
S1.2. Synthesis of TiO2 and Au@a-TiO2 photoelectrodes: Figure S1 illustrates the fabrication
process of Au@a-TiO2 photoelectrodes. All the photoelectrodes were fabricated using fluorine-
doped tin oxide (FTO) coated glass as substrates. First, the FTO was washed using soap, acetone,
and methanol under strong sonication, respectively. Following that, FTO substrates were loaded
into a DC magnetron sputtering system, and a thin layer of Au with a thickness of 10 nm was
deposited on top of the FTO substrates. Next, the Au-coated FTO substrates were annealed for 1
hr under an air atmosphere in a tube furnace at 550 ℃. After this step, hemispherical gold
nanoislands were formed on top of FTO through spontaneous dewetting of the Au film.
Subsequently, the Au nanoisland samples were loaded into an Oxford Instruments atomic layer
deposition system (ALD) using TiCl4 as the Ti precursor and a N2/H2 plasma as the N precursor.
The substrate temperature was 200 °C and a deposition rate of 0.024 nm/cycle was employed. A
thin (10 nm) layer of amorphous TiO2 layer was deposited on top of the photoelectrodes to
fabricate the Au@a-TiO2 core-shell films. a-TiO2 photoelectrode was fabricated by direct ALD
deposition of a layer of amorphous TiO2 on top of a bare FTO electrode.

2. S2
Figure S1. Schematic illustration of the Au@a-TiO2 fabrication process
S1.3. Characterization: The surface morphology of the photoelectrodes was investigated using a
field emission scanning electron microscope (Zeiss Sigma FESEM) and a Helium Ion Microscope
(Zeiss Orion). The optical properties of the samples were investigated using a UV-Vis
spectrophotometer (Perkin-Elmer, Lambda 1050), and the FDTD simulation was conducted using
Lumerical electromagnetic simulation package (Lumerical Solutions Inc). The crystalline
structures of the samples were recorded using X-ray Powder Diffraction (XRD) technique (Bruker,
D8 X), the elemental composition and binding energies of the samples were acquired by x-ray
photoelectron spectroscopy (Kratos Analytical, Axis-Ultra) instrument equipped with an Al-Kα
source (15 kV, 50 W) under ultra vacuum. The XPS depth profile was also measured using a PHI
VersaProbe III Scanning XPS Microprobe equipped with an Ar+
ion sputtering gun. The surface
potential measurements were conducted using a Kelvin probe force microscopy (KPFM) equipped
with an Atomic Force Microscope (Dimension FastScan, Bruker) and with 520 nm and 635 nm
diode lasers.
S1.4. Deposition of Au and TiO2 thin films: Deposition of Au layer was implemented in a DC
magnetron sputtering system (Kurt J. Lesker Co.) operating under high vacuum (850 mTorr).
Before the deposition, the Au target was pretreated for 60 seconds. The deposition duration was
30 seconds. Deposition of amorphous TiO2 layer was done on an ALD apparatus (Oxford
Industries ALD, FlexAl), Titanium isopropoxide was used as the Titanium precursor, and oxygen
plasma was the source of oxygen, the substrate temperature was suet to 120 0
C, the chamber was
firstly cleaned for 50 cycles then the actual deposition was done during 227 cycles with a
deposition rate of 0.44 A0
/cycle (approximately 10 nm in thickness).
S1.5. Photoelectrochemical H2 evolution and Faradaic efficiency measurement: The
photoelectrochemical measurements were conducted in a standard three-electrode system. The
fabricated photoelectrodes were used as working electrodes. Pt was chosen as the counter
electrode, and a standard Ag/AgCl electrode was used as a reference electrode. The electrolytes
used for the experiments were an aqueous solution of 1 M Potassium Hydroxide (KOH) or Sodium
sulfate (Na2SO4). The linear sweep voltammetry (LSV) measurements were conducted by
sweeping the applied potential from -0.8 to +1 V, and the dark photocurrent was collected for the
comparison purpose under the same condition. The photocurrent versus time (I-t measurements)
was conducted either under the constant applied potential of +0.6 (photoanode) in a 1M aqueous

3. S3
KOH electrolyte or applied potential of -0.6 (photocathode) in a 1 M aqueous Na2SO4 electrolyte.
The scavenger test was conducted using the same procedure except that the solution was a 1M
KOH in 10-volume percent methanol in water for the hole scavenger test. The solar simulator used
in these experiments was a standard Class A solar simulator (Newport Instruments) with an
intensity of 100 mW cm -2
. A 420 nm long pass optical filter with an optical density of 5.0 was
used in front of the solar simulator to create UV-filtered solar illumination. A set of LEDs were
also used as illumination light sources. The LEDs' output power intensity was calibrated to be 10
mW cm -2
.
The evolved H2 was measured in an H-Cell made of quartz under standard 1 Sun illumination. The
electrolyte was a solution of 1 M KOH in water, the counter electrode was Pt, and the reference
electrode was a standard Ag/AgCl electrode. The applied potential was fixed at +0.6 V. The
experiment was conducted for 30 minutes and the amount of H2 produced at the counter electrode
was measured using a gas chromatograph (Shimadzu GC-2014) equipped with Porapak Q and
molecular sieve columns, and a pulsed discharge detector (PDD). The carrier gas of the GC was
ultra-high purity (UHP) helium. The retention time of the H2 was 2.75 min (Figure S9). The
Faradaic efficiency, which is the ratio between the observed hydrogen in experimental condition
to theoretical hydrogen evolution based on photocurrent value, was measured using the following
equation:
= 100 ∗
( )
( ). ( ). "(# )
2. ( ). %&( ')
J is the photocurrent density, A is the photoelectrode area, T is the irradiation time, e is the
standard charge of an electron (1.602 × 10-19
C). NA is the Avogadro number (6.02 x 1023
mol-1
).
S1.6. FDTD Simulations: Lumerical software (ANSYS Inc.) was used to implement the FDTD
simulations. Using the graphical user interface (GUI) of Lumerical, one can define materials with
different shapes and structures even in the nanoscale size regime. The first step of simulations is
defining the structures. 5 hemispherical Au nanoparticles with diameters ranging from 30 nm to
70 nm were defined inside the simulation environment to represent the polydisperse nature of the
sample. For constructing the Au@a-TiO2 sample, a hemispherical shell made of TiO2 with a
thickness of 10 nm was added on top of the each of the Au hemispheres. The boundary condition
in all directions was set to perfectly matched layer (PML). A total field scatter field (TFSF) light
source propagating in Z direction with a 300-800 nm bandwidth and 1000 frequency points was
chosen as the light source. The absorption cross-section was recorded using the proper monitor.
To simulate the near-field electric field intensity, the effect of polydispersity of the Au@a-TiO2
was ignored, and the electric field intensity was recorded using a frequency domain field and power
monitor on a single nanoparticle at plasmon resonance (650 nm) under unpolarized light excitation.
To mimic an unpolarized light, these simulations were performed using the same TFSF source two
times with 90-degree difference in polarization of the TFSF source. The electric field obtained

4. S4
from these two simulations was averaged and reported. A uniform mesh size of 1 nm was
introduced in the simulations to ensure the accuracy of the results.
S1.7. Building the structures for DFT: We have built the TiO2-Au composite structures for DFT
calculations based on collected Raman and XRD characterizations data (Figure S2 a,b). Due to the
semi-crystalline structure of a-TiO2, we constructed the semi-crystalline a-TiO2 by taking the
dominant TiO2 anatase (101) and perturbing it, using a molecular mechanics force field called
universal force field (UFF) assisted optimization protocol.1
We have also performed computations
for the ordered TiO2 anatase (101) and this UFF-optimized system for comparison purposes. The
energy cut-off value and the threshold for convergence criterion for the self-consistent loop were
220 eV and less than 10-4,
respectively. Projected density of states (PDOS) plots were constructed
using the Gaussian broadening method, with a broadening parameter of 0.08 eV. The highest
occupied molecular orbital (HOMO) and lowest unoccupied molecular orbital (LUMO) were
constructed using VMD (visual molecular dynamics) software.
S1.8 DFT modeling: Density functional theory (DFT) based quantum chemical calculations were
performed in two steps, geometry optimization, and electronic properties calculations using
OpenMx 3.9.2 (Open source package for Material eXplorer) package,2
where norm-conserving
pseudopotentials,3
and pseudo-atomic localized basis functions are used. Generalized gradient
approximation (GGA)4
with Perdew–Burke–Ernzerhof (PBE) exchange-correlation functional has
been employed in all the computations. Periodic boundary conditions were used for all the systems.
We considered Hubbard U-corrections in our DFT model, where the U-value for Ti d-orbitals was
supposed to be 3.3 eV.5
This Hubbard U-correction provides additional on-site Coulomb
interaction, which is needed for reproducing experimental bandgaps. Such additional functions
force the electrons of a particular orbital to be more localized.6
SECTION S2. ESTIMATION OF THE QUANTUM YIELD
In the grey-colored photoelectrochemical i-t curve of Figure 3a, the value of the current density
(jfull) at 165 seconds is taken to be the steady-state photocurrent density under full spectrum
AM1.5G one sun illumination. This value equals 2.43 mA cm-2
. The corresponding current density
under UV-filtered AM1.5G one sun illumination (using 420 nm long pass filter) is 0.48 mA cm-2
.,
which we term juvf. Therefore, the photocurrent generation attributable purely to high energy (he)
photons with wavelengths less than 420 nm is given by jhe = jfull - juvf = 1.95 mAcm-2
. Now, the
number of photons present in the AM1.5G irradiance spectrum (per data published by NREL and
NASA published data) with wavelengths less than 420 nm is 1.334 ×1016
cm-2
. If every single one
of these photons contributed to photocurrent, the expected photocurrent density would be 2.14 mA
cm-2
. The measured photocurrent density of 1.95 mAcm-2
for high energy photons is 91.5% of the
expected value, indicating a near-unity external quantum yield (also known as the incident photon-
to-electron conversion efficiency or IPCE). There are important caveats to consider here. The first
is that the density of gold nanoparticles varies from sample-to-sample. Some samples have a
higher optical density than others in the 300-420 nm spectral range. Secondly, the presence of a
current doubling process for high energy photons of chemical origin intrinsic to the Au/a-

5. S5
TiO2/KOH system (similar to that observed upon the addition of alcohols) cannot be ruled out.
Therefore, we stress that the quantum yield calculation is a rough estimate at this point and more
detailed studies are needed to clarify the precise value and origin of the high external quantum
yield in Au@a-TiO2 photoanodes. The more important result is the proof-of-concept
demonstration of hot hole harvesting, which is unexpected in a heterojunction consisting of n-type
TiO2 and Au NPs.
Figure S2. FESEM images of (a) Au nanoislands and (b) Au@a-TiO2.

6. S6
Figure S3. Core‐level high-resolution XPS analysis of Au@a-TiO2, (a) Survey scan (b) O1s (c) Ti 2p
and (d) Au 4f regions; (e) The XPS depth profile of Au@a-TiO2 by Argon etching.

7. S7
Figure S4. (a) The linear sweep voltammetric (LSV) curves of Au@a-TiO2 and a-TiO2 using
illumination from a solar simulator and a solar simulator with UV cut-off filter under applied potential
vs. NHE reference electrode in 1 M KOH electrolyte. (b) The LSV curves of Au@a-TiO2 under different
LEDs illumination under applied potential vs. NHE reference electrode in 1 M KOH electrolyte.
Figure S5. (a) Measured i-t curve under cathodic bias (-0.6 V vs Ag/AgCl) of Au@a-TiO2 under UV
filtered solar illumination in 1 M KOH electrolyte. The inset is a schematic illustration of the Au@a-
TiO2 photoelectrode under negative bias (b) Measured i-t curve under cathodic bias (-0.6 V vs
Ag/AgCl) of a-TiO2-Au photoelectrode under UV-filtered solar illumination. The inset is a schematic
illustration of the a-TiO2- Au photoelectrode under negative bias.

8. S8
Figure S6. Linear sweep voltammograms showing the comparison of the photoelectrochemical
performance in 1 M KOH under AM1.5G solar illumination of FTO/Au@a-TiO2 and FTO/a-TiO2-Au@a-
TiO2 photoanodes.

9. S9
Figure S7. (a) UPS work function of Au, a-TiO2, and Au@a-TiO2 and (b) Valence band spectra of
Au@a-TiO2 and a-TiO2 photoelectrodes. c) Mott-Schottky plots of Au@a-TiO2 and a-TiO2 (d) Long-
wavelength absorption spectra of the Au@a-TiO2 before and after visible light irradiation (the NIR
absorption corresponds to absorption related to the free charge carriers). The inset shows the
difference between LSPR peak normalized absorption spectra of Au@a-TiO2, before and after visible
light irradiation.

10. S10
Figure S8. (a) Photoelectrochemical measured i-t curves of Au@ Al2O3 under AM1.5G 1 sun
illumination and UV-filtered solar illumination and (b) Champion photoelectrochemical measured i-t
curves of Au@c-TiO2 using the AM1.5G one sun illumination (black curve) and UV filtered solar
illumination (red curve) under +0.6 V applied potential vs. Ag/AgCl reference electrode in a 1 M KOH
electrolyte. An Au@a-TiO2 sample exhibiting excellent photoelectrochemical performance was
annealed at 500℃ to crystallize the amorphous TiO2. The resulting photoelectrode consisting of Au
nanoislands on FTO coated with crystalline TiO2 is termed Au@c-TiO2. The photocurrents for solar
and UV-filtered solar illumination are still anodic for the annealed photoanode (Au@c-TiO2), but way
less than the values for the unannealed sample (Au@a-TiO2). Annealing the TiO2 eliminates the mid-
gap states and reduces the probability of extraction of hot holes from Au.
Figure S9. (a) Projected density of states (PDOS) of selected atoms for Au -TiO2 heterostructures.
TiO2 (101) without any perturbation and oxygen vacancies and Au (111) were chosen for DFT
modeling. (b) Orbital-resolved projected density of states (PDOS) for Au (111) in Au@a-TiO2
heterostructure.

11. S11
Figure S10. GC chromatogram of evolved H2 in PEC water splitting on Pt counter electrode, using
Au@a-TiO2 and a-TiO2 photoelectrodes.
REFERENCES
1. Rappé, A. K.; Casewit, C. J.; Colwell, K.; Goddard III, W. A.; Skiff, W. M., UFF, a full periodic table
force field for molecular mechanics and molecular dynamics simulations. Journal of the American chemical
society 1992, 114 (25), 10024-10035.
2. Ozaki, T., Variationally optimized atomic orbitals for large-scale electronic structures. Physical
Review B 2003, 67 (15), 155108.
3. Bachelet, G.; Hamann, D.; Schlüter, M., Pseudopotentials that work: From H to Pu. Physical Review
B 1982, 26 (8), 4199.
4. Perdew, J. P.; Burke, K.; Ernzerhof, M., Generalized gradient approximation made simple. Physical
review letters 1996, 77 (18), 3865.
5. Dhifallah, M.; Iachella, M.; Dhouib, A. n.; Di Renzo, F.; Loffreda, D.; Guesmi, H., Support Effects
Examined by a Comparative Theoretical Study of Au, Cu, and CuAu Nanoclusters on Rutile and Anatase
Surfaces. The Journal of Physical Chemistry C 2019, 123 (8), 4892-4902.
6. Capdevila-Cortada, M. a.; Łodziana, Z.; López, N. r., Performance of DFT+ U approaches in the
study of catalytic materials. ACS Publications: 2016.