Distinguishing Direct and Indirect Photoelectrocatalytic OxidationMechanisms Using Quantitative Single-Molecule Reaction Imagingand Photocurrent MeasurementsJustin B. Sambur and Peng Chen*

Department of Chemistry and Chemical Biology, Cornell University, Ithaca, New York 14853, United States

*S Supporting Information

ABSTRACT: Light-driven semiconductor-catalyzed oxidationreactions are of fundamental importance in photocatalysis andphotoelectrocatalysis for removing organic contaminants inwastewater, solar energy conversion, and fine chemical synthesis.The underlying reaction mechanism is often unclear because it isdifficult to measure directly and specifically the semiconductor-catalyzed reaction rates. For example, an organic molecule couldbe oxidized “directly” by photogenerated holes that aretransported from the semiconductor interior to the semi-conductor−electrolyte interface or “indirectly” by photogen-erated intermediates (e.g., hydroxyl radical, superoxide anion, orhydrogen peroxide) that are produced at the semiconductorsurface in aqueous solution. New experimental approaches thatcan distinguish these pathways are thus desirable. Here we introduce quantitative single-molecule, single-particle fluorescenceimaging to measure the photoelectrocatalytic oxidation rate of a model organic substrate, amplex red, on the surface of individualrutile TiO2 nanorods. Our approach probes the oxidation product selectively before it becomes further degraded (whichcomplicates bulk reaction kinetics measurements) while also avoiding interparticle charge transfer kinetics. By examining thereaction rate scaling relations versus light intensity at fixed potential and versus potential at fixed light intensity, together with thecorresponding photocurrent scaling reactions, we demonstrate that amplex red oxidation on a TiO2-nanorod photoanodeproceeds via an indirect mechanism.

1. INTRODUCTIONSemiconductor photo(electro)catalysis is a multistep processthat involves photoexcitation of a semiconductor, followed bycharge-carrier separation within the semiconductor andsubsequent charge transfer reactions at the semiconductor−liquid interface.1,2 Solar-light-driven semiconductor-catalyzedchemical reactions are of broad interest for environmentalremediation (e.g., decomposition of organic contaminants inwastewater),3 solar energy conversion,4 and fine chemicalsynthesis.2,5

Decades of research have been devoted to addressing thefundamental questions regarding the underlying photo-(electro)catalytic oxidation mechanism. For example, a varietyof oxidizing intermediate species may be produced by aphotoexcited semiconductor in aqueous solution.6 Photo-generated electrons can react with dissolved O2 to generatesuperoxide anions (O2

•−), which can then oxidize a substrate.O2

•− can also further react with protons and additionalphotoexcited electrons to form hydrogen peroxide (H2O2) andhydroxyl radical species (OH•), which can also act as oxidants.7

On the other hand, photogenerated holes can directly oxidize asubstrate or react with surface-adsorbed water molecules togenerate OH•.8 Since both photogenerated electrons and holescan potentially produce oxidative intermediates, identifying the

true oxidant in a photo(electro)catalytic oxidation reaction canbe challenging. Even when anaerobic conditions are used toremove the O2-induced oxidation pathway, photogeneratedholes could react “directly” or “indirectly” (e.g., via intermediateOH•) with organic molecules, and therefore the dominantreaction pathway needs to be differentiated.8,9

Previous studies used ensemble-level photocatalysis andphotoelectrochemical characterization tools to distinguishbetween the direct or indirect hole-induced oxidationmechanisms. Most experiments use TiO2-based photo(electro)-catalysts because the materials are commercially available,inexpensive, nontoxic, and photochemically stable.6 Forphotoexcited semiconductor particles in aqueous solution(i.e., photocatalysis), reaction mechanisms can be probed byanalyzing product distributions and formation rates. By use ofthese analyses, the direct, indirect, and simultaneous direct−indirect oxidation mechanisms have been observed (see ref 10and references therein); the mechanism depends on the

Special Issue: Richard P. Van Duyne Festschrift

Received: February 23, 2016Revised: May 16, 2016Published: May 19, 2016

Article

pubs.acs.org/JPCC

© 2016 American Chemical Society 20668 DOI: 10.1021/acs.jpcc.6b01848J. Phys. Chem. C 2016, 120, 20668−20676

photocatalyst, solution composition, and organic substrate. Insome cases, such as formic acid oxidation by TiO2 photo-catalysts, some authors propose an indirect mechanism11,12

while others propose a direct mechanism.2,13−15 Thiscontroversy is partly due to that in photocatalysis bothphotogenerated electrons and holes react on the same particlesurface, making it difficult to differentiate which charge-carrierlimits the oxidation rate of the organic substrate.16

To address this controversy, Salvador and co-workerscombined electrochemical and photocatalytic methods (i.e.,photoelectrocatalysis) to separate the oxidation and reductionprocesses to a TiO2 photoanode and a metal cathode,respectively.10,17 This approach effectively removed theelectron-induced oxidation pathway by rapidly extractingphotogenerated electrons via the external circuit. By analyzingthe organic substrate-induced photocurrent enhancement as afunction of light intensity at a fixed applied potential, theyconcluded that formic acid and methanol are oxidized by directand indirect pathways, respectively. A critical requirement forthis approach is that the organic substrate is oxidized by the so-called “current doubling” mechanism.18 Several other studieshave been reported in using the scaling relation of photocurrentenhancement versus light intensity for other photoanode−substrate combinations.9,19,20 On the other hand, the scalingrelation of substrate-induced photocurrent enhancementsversus the applied potential has not been exploredquantitatively.Considering the importance of photo(electro)catalytic

oxidation reactions and the complications faced whendetermining the underlying mechanism, new or alternativeexperimental approaches are desirable to distinguish betweenpotential mechanistic pathways.10 Here, building upon ourrecent work,21 we use a novel single-molecule, single-particleapproach to distinguish between photoelectrocatalytic oxida-tion mechanisms. We chose solution-processable rutile TiO2nanorods with well-defined surface facet orientations ({100}sidewalls and {011} end tips) as model semiconductorphotoanodes.22 We chose to study amplex red as a modelorganic substrate because it can be irreversibly oxidized to ahighly fluorescent product resorufin (Scheme 1) and the

reaction rate can be monitored via absorption and fluorescencespectroscopy at the ensemble level and via fluorescence at thesingle-molecule level.23,24 Amplex red is becoming increasinglypopular as a probe reaction for (photo)catalytic oxidationreactions,25−29 but the reaction mechanism is often not defined.Recent ensemble-level photocatalysis studies showed thatphotoexcited TiO2 nanoparticles could further oxidize theproduct resorufin,21 which complicates the quantitation ofamplex red to resorufin reaction kinetics because thefluorescent product is continuously generated as well asconsumed during the bulk fluorescence assay experiment.

Here we use single-molecule fluorescence microscopy toquantitatively image individual product molecules while theyare generated on individual TiO2 nanorod photoanode surfaces(i.e., single reaction resolution). The single-molecule, single-particle approach to study photoelectrocatalysis has severalunique advantages compared with ensemble-level measure-ments. First, fluorescence microscopy selectively probesproduct formation prior to its consumption. Second, theproduct fluorescence signal is distinct from the photo-electrochemical current signal, which is not selective to thesubstrate oxidation process and contains many othercontributions such as water oxidation in an aqueous electrolyte.Third, as we only study isolated nanorods, complex charge-carrier transport processes between nanorods30 do notcontribute to the observed reaction rates. Fourth, studyingsingle particles also removes ensemble-averaging effects andallows for identification of champion or inactive catalysts.However, we note that the technique is limited to fluorogenicreactions.Using single-molecule fluorescence microscopy to measure

product formation rates on individual TiO2 nanorod photo-anodes, we demonstrate here that the scaling relations versuslight intensity at fixed potential and versus potential at fixedlight intensity support that AR oxidation follows an indirectoxidation mechanism.

2. MECHANISTIC MODELHere we describe a mechanistic model for quantifying thephotoelectrocatalytic oxidation kinetics of an organic substrate(R) on the surface of a semiconductor photoanode in aqueouselectrolyte. Assuming the rate-limiting step of R photo-electrooxidation is first order with respect to both R and theoxidizing species, a general expression of the photoelectroox-idation rate of R (vR) can be written as

=v k [R] [ox]R R s s (1)

Here kR is a rate constant. [R]s is the surface concentration of Ron the photoanode, and it depends on the bulk solutionconcentration of R. [ox]s is the surface concentration of theoxidizing species that reacts with R. In the direct mechanism, oxwould be the surface-accessible photogenerated holes (h+)s; inthe indirect mechanism, ox would be one of the surface speciesresulting from the photoelectrooxidation of water, for example,surface adsorbed hydroxyl radical (OH•)s, hydrogen peroxide(H2O2)s, or superoxide anion (O2

•−)s (this last pathway can beminimized to an insignificant level by removing O2 from thesolution.We use the Gartner model to describe the photogenerated

hole transport rate to the semiconductor photoanode surface(step i in Scheme 2).31 The major assumptions of this modelare the following. (1) Charge carrier generation follows aBeer−Lambert absorption profile into the semiconductorphotoanode. (2) The applied potential E is entirely manifestedas a potential drop within the depletion region of thesemiconductor whose thickness is W (i.e., the depletionapproximation; see Scheme 2 for schematic illustration). (3)All photogenerated holes within the depletion region aretransported to the surface (i.e., no recombination with electronswithin W). (4) Holes generated outside the depletion region(i.e., x > W in Scheme 2) are transported to the surface if theydiffuse into the depletion region. (5) Due to the ∼3.0 eV bandgap of TiO2, the thermally excited hole population at roomtemperature is orders of magnitude lower than the light-

Scheme 1. Oxidative Deacetylation of Amplex Red (AR) toResorufin (P) by a Photoexcited TiO2 Photoanode

The Journal of Physical Chemistry C Article

DOI: 10.1021/acs.jpcc.6b01848J. Phys. Chem. C 2016, 120, 20668−20676

20669

induced hole population, and therefore thermally excited holesare not considered. Under these conditions, the photo-generated hole transport rate vh+ to the photoanode surface isgiven by

αα

= = −−

+

++

⎛⎝⎜⎜

⎞⎠⎟⎟v

tI

WL

d[(h ) ]d

1exp( )

1hs

0p (2)

Here q, I0, α, and Lp are the elementary charge (C), intensity oflight impinging on the semiconductor photoanode (photon/s),absorption coefficient (cm−1) of the semiconductor at theexcitation light wavelength, and the hole diffusion length (cm).

εε= −W E E qN2 ( )/( )0 fb d (cm), where ε is the relativedielectric constant of the semiconductor, ε0 is the vacuumpermittivity, E is the applied potential relative to the Ag/AgClreference electrode, Efb is the flat band potential, and Nd is thedoping density of the semiconductor.In eq 2, αLp and αW account for the fraction of absorbed

photons within a diffusion length distance from W and thefraction of photons absorbed within W, respectively. For therutile TiO2 nanorods we studied here, α375nm = 104 cm−1 32 andLp = 20−30 nm,33,34 yielding αLp = 0.03, which is less than 5%and therefore can be neglected in the (1 + αLp) term in eq 2. Inother words, the contribution of interior photogeneratedcarriers that diffuse to W can be neglected (i.e., for x > W inScheme 2). To estimate the magnitude of the potentialdependent αW term, we evaluate W at E − Efb = 0.9 V(corresponding to an (E − Efb) value that is larger than anyapplied potential used in single-molecule imaging experiments,

e.g., +0.2 V versus Efb = −0.63; see section 4.2 for Efbdetermination) using an estimated Nd = 1018 cm−3,21 whereW = 100 nm and αW = 0.1. Thus, this small magnitude of αWallows for expanding the exponential term in eq 2 into a Taylorseries and neglecting the higher order terms to yield asimplified Gartner equation (eq 3), as done by Butler:35

α αεε

= = =−+

+vt

I W IE EqN

d[(h ) ]d

2 ( )h

s0 0

0 fb

d (3)

Assuming all holes transported to the surface will undergointerfacial charge transfer reactions on the photoanode (e.g.,with H2O or R), the corresponding photocurrent i is given by

α αεε

= =−

i qI W qIE EqN

2 ( )0 0

0 fb

d (4)

Equation 4 predicts that i is proportional to (E)1/2 and I0. Thephotocurrent increases with increasingly anodic potentialbecause it is proportional to the number of photons absorbedwithin W, whose thickness scales with (E)1/2. Equation 4 alsopredicts that the flat band potential (Efb) is also thephotocurrent onset potential (i.e., when E > Efb, i > 0).To describe the major interfacial charge transfer reactions by

the photogenerated holes, we considered a mechanistic modelthat includes a limited number of major irreversible reactionsteps for water oxidation toward O2 on a TiO2 photoanodesurface, following Salvador et al.10,36,37 The relevance of thesereactions is supported by experimental data obtained with bulkrutile TiO2 photoanodes.36−38 These major reactions areschematically shown in Scheme 2, and their rate equationsare listed in Table 1. In this model, water oxidation byphotogenerated holes is approximated to start with theoxidation of surface adsorbed H2O or OH− to generate surfaceadsorbed OH• (step ii in Scheme 2 and eq 5 in Table 1). Wenote a more recent work by Salvador and co-workers usingisotopically labeled Ti18O2 photocatalysts in anhydrousacetonitrile showed that terminal oxygen atoms in TiO2 caninitially capture photogenerated holes.39 The relevance of theaforementioned model in the context of this work is discussedin Supporting Information section 1, and the implications ofthis mechanism will be discussed following the derivation of themodel. As a further approximation, we do not differentiate theoxidation rates of surface adsorbed H2O and OH−. The surfaceadsorbed OH• can subsequently combine to form H2O2 (stepiii in Scheme 2 and eq 6 in Table 1), which can be furtheroxidized by holes to generate O2 (step iv in Scheme 2 and eq 7in Table 1). The oxidation of H2O2 to O2 consumes two (h+)s,but we assume that they are sequential reactions, as three-bodyreactions are microscopically unlikely, and assume that the firsthole oxidation step is rate limiting so that the formation rate of

Scheme 2. Illustration of Photoelectrooxidation of anOrganic Substrate (R) to a Product (P) and of H2O/OH

− toO2 on a TiO2 Photoanode

a

aSee the main text for details on steps i−vii. Step i: hole transport tothe TiO2 surface. Steps ii−iv represent the major water oxidationreaction steps. Step v: direct oxidation pathway of R by surface holes.Steps vi and vii: indirect oxidation pathway of R by OH• and H2O2,respectively.

Table 1. Major Kinetic Reactions Included in the Model for Water Oxidation on Photoexcited TiO2 under Anodic Conditionsa

step reaction rate law equation no.

ii + ⎯ →⎯⎯⎯⎯− + ••(H O or OH ) (h ) (OH )

k2 s s s

OH = =•

+ −• •vt

kd[(OH ) ]

d[(h ) ][(OH ) or (H O) ]OH

sOH s s 2 s 5

iii + ⎯ →⎯⎯⎯⎯⎯• •(OH ) (OH ) (H O )k

s s 2 2 sH2O2 = = •v

tk

d[(H O ) ]d

[(OH ) ]H O2 2 s

H O s2

2 2 2 26

iv + ⎯ →⎯⎯ ++ +(H O ) 2(h ) O 2Hk

2 2 s s 2O2 ν = = +

tk

d[O ]d

[(h ) ][(H O ) ]O2

O s 2 2 s2 27

aThese reactions are schematically shown in Scheme 2.

The Journal of Physical Chemistry C Article

DOI: 10.1021/acs.jpcc.6b01848J. Phys. Chem. C 2016, 120, 20668−20676

20670

O2 is first order with respect to holes. It is also assumed that theapplied potential is sufficiently positive so that all recombina-tion reactions between electrons and holes, as well as electronsand photogenerated intermediates (e.g., OH• and H2O2), aresufficiently slow to be negligible kinetically, similar to what wasassumed by Salvador10,36 and consistent with the Gartnermodel.Depending on the mechanism, R can react directly with (h+)s

(step v in Scheme 2) or with any of the intermediates of wateroxidation reactions (e.g., OH• and H2O2; steps vi and vii inScheme 2). Superoxide, another potential oxidant that can formby reducing O2 by the photogenerated electron and with whichR can react, can be minimized by degassing the electrolyte andoperating under dominantly anodic conditions. In either case,we further assume that the concentration of R is sufficientlysmall compared with that of H2O or OH− so that it contributesinsignificantly to the overall kinetics and thus the photocurrentfrom all photoelectrocatalytic oxidation reactions on TiO2surfaces (we will justify this approximation in analyzing theexperimental data in section 4.3). In other words, R can betreated as a probe molecule that samples the surfaceconcentration of oxidizing species but does not significantlyperturb the species’ concentration.Taking into account eq 2, as well as eqs 5−7 in Table 1, we

can apply the steady-state approximation to the intermediates(h+)s, (OH

•)s, and (H2O2)s:

= − − =+

+ •

tv v v

d[(h ) ]d

2 0sh OH O2 (8)

= − =•

•

tv v

d[(OH ) ]d

2 0sOH H O2 2 (9)

= − =t

v vd[(H O ) ]

d02 2 s

H O O2 2 2 (10)

To help solve for [(h+)s], [(OH•)s], and [(H2O2)s] from eqs

8−10, we approximate that the rate constant for hole captureby either surface adsorbed OH− or H2O is similar in magnitudeto that by surface-adsorbed H2O2 (i.e., kOH

• ≈ kO2).36 In

addition, due to the large excess of bulk OH− or H2O, weapproximate that the steady-state surface concentrations ofOH• and H2O2 are both significantly less than surface adsorbedOH− and H2O. This approximation is supported by that themaximum theoretical surface coverage of OH− on bulk rutileTiO2 is about (5 to 15) × 1014 cm−2 whereas the calculatedvalues of [(OH•)s] and [(H2O2)s] are 8 × 1012 and 8 × 1013

cm−2, respectively, from experimental photocurrent data.8,36

Using these approximations, we obtained

α εε=

−+−

•

Ik

E EqN

[(h ) ][(H O or OH ) ]

2 ( )s

0

OH 2 s

0 fb

d (11)

α εε=

−• Ik

E EqN

[(OH ) ]2

2 ( )s

0

H O

0 fb

d2 2

4

(12)

=−

•kk

[(H O ) ][(H O or OH ) ]

2 2 sOH 2 s

O2 (13)

Equations 11, 12, and 13 indicate that [(h+)s] and [(OH•)s]differ in their dependences on I0 and E, whereas [H2O2]s ispotential and light independent, and its magnitude depends onthe ratio of rate constants for OH• formation and O2 evolution.

By use of eq 1, for R to react with (h+)s, (OH•)s, or (H2O2)s

the corresponding rate equations are

αεε

= =

−

+

−•

v k

kk

IE EqN

[(R) ][(h ) ]

[(R) ][(H O or OH ) ]

2 ( )R R s s

R s

OH 2 s0

0 fb

ds (14)

α εε

= =

−

•v k

kI

kE EqN

[(R) ][(OH ) ]

[(R) ]2

2 ( )R R s s

R s0

H O

0 fb

d2 2

4

(15)

= =−

•

v k

kk

k

[(R) ][(H O ) ]

[(R) ][(H O or OH ) ]

2

R R s 2 2 s

R sOH 2 s

O2 (16)

The three equations predict distinct scaling relations for vR vsthe light intensity I0 or the applied potential E. For the directmechanism, vR scales with I0 and (E)1/2. For the indirectmechanism via (OH•)s, vR scales with (I0)

1/2 and (E)1/4. Andfor the indirect mechanism via (H2O2)s, vR would beindependent of I0 and E.We note here that the vR dependences on E and I0 in eq 15

for the indirect mechanism are unaffected whether photo-generated holes are assumed to react with either adsorbedH2O/OH

− to generate oxidizing OH• species or terminaloxygen atoms in TiO2 to generate oxidizing O•− species (seeSupporting Information section 1); additional characterizationtools would be required to distinguish between the twooxidants.39 In section 4 below we will apply eqs 4 and 14−16 toanalyze the results from the photoelectrooxidation of amplexred (AR), where we also validate many of the approximations inthe above mechanistic model.

3. EXPERIMENTAL SECTION3.1. Nanorod Synthesis. Rutile TiO2 nanorods were

synthesized via a molten-flux salt method, following Liu et al.22

The nanorods were synthesized by grinding a mixture of 250mg of P25 TiO2 nanoparticles (Acros Organics), 1 g of NaCl(Aldrich), and 250 mg of Na2HPO4 (Aldrich) to a fine powder.The powder was transferred to a crucible and annealed in a boxfurnace (Thermo Scientific) at 825 °C for 12 h. After cooling toroom temperature, the reaction product was washed 5 timeswith boiling 0.1 M hydrochloric acid to dissolve residual salts(presumably from NaCl and Na2HPO4 precursors40). Thenanorods were dispersed in a 4:1 (by volume) ethanol to H2O(all water used was np-H2O, unless specified otherwise).

3.2. Photoelectrochemical Measurements. The designof the electrochemical flow cell reactor and experimental setuphas been described in detail.21 The working electrode was aTiO2 nanorod-coated ITO electrode prepared by spin-casting300 μL of a ∼1 mg mL−1 nanorod solution onto an ITOsubstrate at 750 rpm. The nanorod−ITO electrode was rinsedwith H2O and annealed in air at 450 °C for 30 min prior to use.The cell was mounted on the stage of an Olympus IX71microscope and connected to a CH Instruments 1200Apotentiostat. The nanorod-coated ITO working electrode wasexcited with a continuous wave, circularly polarized 375 nmlaser (32.9−151.3 W/cm2, CrystaLaser DL-375-020) usingprism-type total internal reflection (TIR) illumination with anillumination area of ∼80 × 95 μm2. We measured photo-

The Journal of Physical Chemistry C Article

DOI: 10.1021/acs.jpcc.6b01848J. Phys. Chem. C 2016, 120, 20668−20676

20671

current−time (i−t) data in the absence of amplex red with 20ms time resolution under chopped illumination (10 mHz, 50%duty cycle) at a constant applied potential. The background(dark) current measured from the macroscopic ITO electrodewas subtracted from the photocurrent signal using a two-pointlinear background subtraction method to an average of fiveconsecutive data points before and after illumination.3.3. Single-Molecule Reaction Imaging. A continuous

wave (18 mW), circularly polarized 532 nm laser beam(CrystaLaser CL532-025-L) was focused onto a 50 × 20 μm2

area (630 W cm−2) of the ITO electrode to induce thefluorescence of the reaction product resorufin via TIRgeometry. The TiO2 nanorods were also excited via TIRillumination using a 375 nm laser (60 W cm−2). Thefluorescence was collected by a 60× NA1.2 water-immersionobjective (Olympus UPLSAPO60XW), filtered by a 580 ± 30nm band-pass filter (HQ580m60, Chroma), and detected by aback-illuminated ANDOR iXon EMCCD camera (DU897D-CS0-#BV) operated at a 15 ms frame rate. A 1.6× magnificationchanger was used to magnify the image 96× overall. The finalimage pixel size was 155.8 nm × 174.9 nm. The slightasymmetry of the pixel size was due to a Princetonspectrograph (Acton SpectraPro) in the detection path thatwas used to acquire fluorescence spectra from the sample.In a typical experiment, the electrode potential, laser powers,

and reactant solution flow rate were fixed (i.e., steady-stateconditions). A series of fluorescence images (i.e., catalyticmovie) were acquired after a steady-state photocurrentresponse was reached (approximately 1 min after 375 nmlaser excitation). The potential was held for 7.5 min at eachpotential and stepped anodically by 0.1 V from −0.6 V to +0.2V. Following imaging experiments, the flow cell wasdisassembled and scanning electron microscopy (SEM, LEO1550) imaging of the sample was performed at the CornellCenter for Materials Research (CCMR).3.4. Image Processing and Data Analysis. Image

processing and data analysis methods were described in detailin our earlier study.21 Fluorescence images were backgroundsubtracted according to ref 41, and all fluorescent bursts whosepixel intensity values were greater than the image mean pixelintensity plus 6 standard deviations were considered aspotential candidate molecules. Each candidate was fitted witha two-dimensional (2D) Gaussian function. The width andintegrated intensity of each fit were analyzed and used toquantitatively count single molecules. We developed analgorithm21 that considers (1) “hot” pixels or those with fitwidths too narrow to represent a single molecule, (2) singlemolecules that appear in consecutive frames (i.e., productadsorption), (3) multiple molecules that appear within adiffraction-limited volume, and (4) single product moleculesthat diffuse on the surface during a single frame. Only thosecandidates whose fit widths and intensities were consistent withthose of single product molecules were included in constructingthe super-resolution reaction images. The sample drift wasmonitored in a frame-by-frame fashion by fitting the emissionfrom individual 100 nm Au nanoparticle fluorescent markersadsorbed on the ITO electrode; the positions of products weresubsequently corrected for in a frame-by-frame fashion.

4. RESULTS AND DISCUSSION4.1. Observation of Photocurrent Transients from

Chopped Illumination and a Qualitative Explanation ofthe Potential-Dependent Transient Behavior. This

section describes the photoelectrochemical current responseof these single-crystalline rutile TiO2 nanorods initiated by asquare-wave 375 nm laser pulse. The nanorods have an averagelength and diameter of 263 ± 113 nm and 62 ± 16 nm,respectively.21 To our knowledge, the photoelectrochemicalproperties of these materials have not been reported in theliterature. The majority of nanorods are well-isolated from eachother on the ITO electrode surface (surface coverage of ∼2%),lying down with their long axes parallel to the electrode(Supporting Information Figure S2); charge transfer betweenadjacent or stacked nanorods does not contribute significantlyto the photocurrent response.Representative current−time data under 10 mHz chopped

375 nm illumination at different potentials are shown in Figure1A. The photocurrent data in Figure 1 was obtained in the

absence of AR; adding AR to the electrolyte does not affectmuch the photocurrent response presumably because of its lowconcentration. The 375 nm laser beam excites ∼1000 TiO2particles, containing mostly nanorods and pseudosphericalparticles (see also Supporting Information Figure S2). Thechopping frequency is slow here so that each illumination cyclelasts 50 s to mimic the steady-state 375 nm illuminationcondition used in single-molecule reaction imaging conditions.Three distinct features were observed in the current−time data(denoted on the −0.4 V data indicated by the black trace inFigure 1A): (1) an initial photocurrent spike (iinitial) when thelight is turned on, (2) a decay of the photocurrent on a

Figure 1. (A) Representative current−time responses at +0.2 V (greenline), −0.3 V (blue), −0.4 V (black), and −0.55 V (red) during asingle on−off cycle at 10 mHz (50% duty cycles). The initial (iinitial),steady state (iss), and cathodic (icath) photocurrent components of thecurrent−time response are denoted on the −0.4 V data. [AR] = 0 nM.(B) Reconstructed i−E data obtained from the chopped photocurrentmeasurements in (A). (C) iinitial

2 and iss2 are linearly proportional to E

for E ≥ −0.3 V. (D) i at fixed +0.2 V versus I0 (photon/s) indicatesthat the photocurrent is linearly proportional to incident laser powerimpinging on the nanorods (i.e., I0), where I0 = 0.02Iincident (where 0.02accounts for the fraction of the ITO electrode surface covered by theTiO2 nanorods and pseudospherical particles; see main text fordetails).

The Journal of Physical Chemistry C Article

DOI: 10.1021/acs.jpcc.6b01848J. Phys. Chem. C 2016, 120, 20668−20676

20672

millisecond to second time scale while the light is on toward asteady-state current (iss), and (3) an initial negative, cathodicphotocurrent spike (icath) when the illumination is turned off,which decays to the background current level also on amillisecond to second time scale. For E = −0.8 V to +0.2 V, wenever observed (1) initial cathodic current spikes upon lightillumination, (2) steady-state cathodic currents under con-tinuous illumination (anodic currents have positive sign inFigure 1), or (3) photocurrent under sub-bandgap 532 nmillumination only.The photocurrent transient behavior in Figure 1A can be

qualitatively interpreted based on previous photoelectrochem-ical studies.42−51 Prior to illumination of the TiO2 nanorods,the system is in a dark equilibrium condition, and for an N2-purged aqueous electrolyte, the dominant h+-accepting specieson the nanorod surface is H2O or OH− (and AR in ourexperiments, whose low nanomolar concentration makes itscontribution negligible). Electron−hole pairs are generated inthe TiO2 nanorods when the square-wave light pulse impingeson the electrode. An anodic photocurrent is measured when thephotogenerated electron−hole pairs are initially separated;electrons are transported to and collected by the ITO electrode,and holes are transported to the semiconductor−electrolyteinterface and oxidize H2O or OH−. Depending on the potential,the photocurrent decays with illumination time on a milli-second to second time scale until iss is reached. The magnitudeof iss depends on the applied potential and reflects the net rateof charge transfer across the interface at the steady state,52

where the surface concentration of h+ acceptors (e.g., H2O orOH−) is lower than the initial, dark equilibrium condition.Upon turning off the light illumination, photogeneratedelectrons can recombine with photogenerated surface inter-mediates, such as surface trapped holes or chemisorbedintermediates that can accept electrons (e.g., OH•, H2O2,),leading to a cathodic photocurrent spike that decays while thephotogenerated electrons and/or the surface intermediates aredepleted.4.2. Establishing a Potential Regime Where the

Ga rtner Model Adequately Describes the PhotocurrentDependences on Potential and Light Intensity. Asdescribed in section 2, the Gartner model assumes that allrecombination reactions between electrons and holes, as well aselectrons and photogenerated intermediates, are sufficientlyslow to be negligible kinetically. Here we establish a potentialregime where this approximation is valid.First, we establish a potential regime where the photo-

current−time behavior is adequately described by the Gartnermodel. Following Salvador36 and Sagara,43 we constructed i−Ecurves by plotting the components of the current−timeresponse versus potential (Figure 1B). For E ≥ −0.3 V, iss is>70% of iinitial, and icath is almost negligible. The observationthat >70% of the initial photocurrent is retained implies thatthe photocurrent−time behavior of these TiO2 nanorods isadequately described by the Gartner model because this modeldoes not consider photocurrent decay behavior. The cathodiccurrent spike is negligible over the same potential rangepresumably because the larger bending within the depletionregion (curved black lines within W in Scheme 2) effectivelyseparates and transports photogenerated electrons to the ITOelectrode, and therefore a negligible amount of photogeneratedelectrons are transported to the surface for recombination.Second, we establish a potential regime where the photo-

current dependences on potential and light intensity

qualitatively follow the Gartner model. For E ≥ −0.3 V, iinitialand iss are proportional to (E)1/2 (see linear i2 vs E plot inFigure 1C). At fixed +0.2 V, where iss is 97% of iinitial, i scaleslinearly with I0 (Figure 1D). We note here that for ourillumination geometry I0 = 0.02Iincident, where Iincident is the totalincident light power on the ITO electrode and the 0.02 factoraccounts for the fraction of ITO surface area covered by theTiO2 nanorods and pseudospherical particles. We thencalculated the apparent surface recombination probability(SRP) using SRP (%) = [(iinitial − iss)/iinitial] × 100 (SupportingInformation Figure S3), following Kong et al.53 For E ≥ −0.3V, SRP < 30%, and in the same potential region i ∝ I0 and i ∝√E (Figure 1C,D). Therefore, for the potential regime of E ≥−0.3 V, the photocurrent dependences on time, potential, andlight intensity are adequately described by the simplified formof the Gartner model (eq 4).Equation 4 also predicts that a linear fit to i2−E data yields

Efb at i = 0 A. Following refs 36 and 43, we extrapolated thelinear fit of the iinitial

2−E data in Figure 1C to i = 0 nA, yieldingan ensemble-averaged Efb = −0.63 ± 0.01 V, which is inagreement with literature values for bulk rutile TiO2 at pH8.54,55 This linear i2−E extrapolation method was alsopreviously shown36,43 to yield an Efb value equivalent to thatdetermined in the dark from Mott−Schottky analysis ofcapacitance−potential data. The steady-state photocurrentfalls to zero at about −0.6 V because the steady-statephotocurrent is proportional to the depletion region thicknessW, which is nearly 0 cm at −0.6 V. According to eq 4, i = 0 nAwhen E = Efb. Therefore, −0.6 V is very close to the determinedflat band potential (Efb = −0.63 V), and according to eq 4, theflat band potential represents the photocurrent onset potential(i.e., for E > −0.63 V, i > 0 nA).We note it is generally observed,52 and specifically for bulk

TiO2 electrodes,36,43 that i−E data qualitatively follows theGartner model for sufficiently positive potentials (e.g., +0.5 Vversus Efb). For the dilute TiO2 nanorod-coated ITO electrodestudied herein, the Gartner model satisfactorily describes thephotocurrent dependences on E, as well as I0 for E > 0.25 Vversus Efb (i.e., ≥−0.3 V vs Ag/AgCl electrode here).Furthermore, it is worth discussing the photocurrent

dependences on time and potential over the potential regimewhere the data are not adequately described by the Gartnermodel. Beginning at −0.4 V and toward more negativepotentials, the three components, iinitial, iss, icath, show differentpotential dependences, and the scaling relation of i−(E)1/2 nolonger holds for iinitial and iss. Specifically, iss decreases rapidly tozero at a potential slightly more negative than −0.45 V; iinitialincreases to a magnitude greater than the photocurrentobserved at positive potentials, reaching a maximum at −0.6V before decreasing again with more negative potentials; andicath increases rapidly to an absolute magnitude larger than thatof the anodic photocurrent at positive potentials, reaching amaximum at −0.6 V before decreasing at more negativepotentials. Similar behaviors of iss, iinitial, and icath were alsoreported for bulk TiO2 electrodes previously;26,33 iss closelyfollows iinitial while icath is negligible over a range of morepositive potentials, and the three current components differ intheir potential dependences at some critical potential whenpolarizing the electrode at negative potentials.36,43 It is worthnoting that when iinitial and icath reach their maximal values at−0.6 V, iss = 0, in agreement with previous studies that usedchopped light to measure the photocurrent response.36,43,51,56

In some specific cases with TiO2 thin films46,57,58 and TiO2

The Journal of Physical Chemistry C Article

DOI: 10.1021/acs.jpcc.6b01848J. Phys. Chem. C 2016, 120, 20668−20676

20673

nanotube arrays,59 the peak feature was larger in magnitudethan the photocurrent at positive potentials, as we observedhere. In those studies,46,57−59 the peak in iinitial and icath has beenattributed to photoexcited electrons that originally populateenergy levels within the band gap (gap states). That bothphotocurrent components reach a maximum value atapproximately −0.6 V can be explained as follows. Uponillumination of TiO2, the electrons that occupy the gap statesare excited, besides those from the valence band, leading to anadditional instantaneous photocurrent at negative potentials. Apeak is observed in the plot of iinitial versus E presumably fortwo reasons: (i) if the energy distribution of the gap statefollows a Gaussian distribution,59 the additional electronsexcited from the gap states will reach a maximum as thepotential becomes more negative, and (ii) the collectionefficiency of the electron−hole pairs simultaneously decreasesas the potential is made increasingly more negative.Correspondingly, when the light is interrupted at negativepotentials, photogenerated electrons react to recombine withholes or intermediates that have accumulated at the surfaceduring illumination, as well as repopulate the band gap states,accounting for the peak in icath at the same potentials as iinitial.4.3. Super-Resolution Imaging of AR Oxidation

Reactions and Reaction Rate Dependences on Potentialand Light Power. Here we use quantitative single-moleculereaction imaging to establish AR oxidation rate scaling relationsversus E and I0 at fixed 50 nM AR concentration.Parts A, C, and E (left panels) of Figure 2 show positions of

product molecules (each orange dot represents a molecule)generated from single catalytic turnovers (i.e., single-moleculereaction imaging) on the surfaces of individual nanorods overthe potential range of −0.6 to +0.2 V. Product positions weremapped onto the nanorod structural contour determined viaSEM imaging (parts B, D, and F of Figure 2; see ref 21 fordetails of this mapping). Quantitative super-resolution activitymaps were then constructed by binning the product positionsinto 40 × 40 nm2 pixels in a two-dimensional histogram (partsA, C, and E (right panels) of Figure 2). These activity mapsindicate significant nanorod-to-nanorod activity heterogeneityas well as significant intrananorod activity heterogeneity alongthe nanorod length. These data motivate single-particle-levelstudies because they reveal nanoscale activity heterogeneitiesthat are averaged in ensemble-level experiments.We then examined the magnitudes of the single-nanorod and

nanorod-averaged AR oxidation rates. Figure 2G showsrepresentative AR oxidation rate (vAR) versus potential (vAR−E) data for 37 individual nanorods (red circles) and thenanorod-averaged rate from 37 nanorods (black squares).There is a large heterogeneity in the single nanorod rates. Theonset of AR oxidation occurs at about −0.5 V and increaseswith positive E, in qualitative agreement with iss. The averageproduct formation rate per nanorod at +0.2 V is about 32molecules s−1 or 64 holes s−1 nanorod−1 (note AR oxidation isoverall a two-electron oxidation reaction). Comparing themaximum single-nanorod hole-induced oxidation rate of AR(180 molecules s−1 nanorod−1 or 360 holes s−1 nanorod−1)with iss at +0.2 V (30 nA/1000 nanorods or 108 holes s−1

nanorod−1) indicates that AR contributes insignificantly to issand also validates our earlier assumption that AR oxidationreactions do not affect steady-state surface concentrations ofthe oxidizing species (i.e., ox) that react with AR.We then examined the scaling relations of vAR with E and I0.

Figure 2H shows the nanorod-averaged (vAR)4 data scales

linearly with E (i.e., vAR scales with (E)1/4) for E ≥ −0.3 V.These data also rule out the possibility that the observedphotoelectrocatalytic product formation is due to impurity P inthe electrolyte that diffuses to the surface of the TiO2 nanorodbecause the diffusion of P to the surface is not expected toexhibit (E)1/4 dependence. At fixed positive potentials (e.g.,+0.2 V), vAR scales linearly with the square root of I0 (Figure2I).

4.4. AR Is Oxidized via an Indirect Pathway (OH•).Equations 14−16 predict distinct scaling relations for vAR vs thelight intensity I0 or the applied potential E for various reactionmechanisms. For the direct mechanism, vAR should scale with I0and (E)1/2 (eq 14), which is inconsistent with the data in partsH and I of Figure 2. For the indirect mechanism via reactionwith surface adsorbed H2O2, vAR should be independent of theapplied potential and light power (eq 16), again inconsistentwith the data. Furthermore, vAR should increase with increasing(H2O2)s; however we did not observe an increase in vAR inensemble level photocatalysis measurements that contained 161mM H2O2 (Supporting Information Figure S2). On the otherhand, for the indirect mechanism via surface adsorbed OH•, vARis predicted to scale with (I0)

1/2 and (E)1/4 (eq 15). Weexperimentally observed that vAR scales with (I0)

1/2 and (E)1/4

at fixed 50 nM bulk [AR] (parts H and I of Figure 2).Therefore, quantitative single-molecule reaction imaging revealsthat photoelectrocatalytic AR oxidation by these TiO2 nanorods

Figure 2. (A, C, E) Single-molecule reaction imaging of photo-electrocatalysis. Scatter plot (left) and two-dimensional histograms in40 × 40 nm2 pixels (right) of all individual P molecules (orange dots)generated from h+-induced AR oxidation reactions over the potentialrange of −0.6 to +0.2 V at 60 W/cm2 375 nm laser illumination and 50nM AR on single TiO2 nanorods. White line is the nanorod structuralcontour from its SEM image (B, D, F). All scale bars are 200 nm. (G)Single-molecule AR oxidation rate (vAR) for 37 individual nanorods(red dots) and the average vAR over 37 nanorods (black squares). Errorbars represent SD. (H) Nanorod-averaged vAR data from (G) plottedas vAR

4 versus potential for E ≥ −0.3 V. Error bars represent errorpropagated SEM. (I) vAR per nanorod versus the square root of lightpower impinging on the nanorods (I0) in the presence of 50 nM bulk[AR] and +0.2 V. These data were obtained from 23 nanorodsdifferent from those in (G). The red dots indicate data from individualnanorods, and the black squares represent nanorod-averaged data(error bars represent SD).

The Journal of Physical Chemistry C Article

DOI: 10.1021/acs.jpcc.6b01848J. Phys. Chem. C 2016, 120, 20668−20676

20674

occurs via an indirect pathway and specifically via reacting withsurface adsorbed OH•. We note here that our conclusion of theindirect pathway applies to the low concentration range of ARwe studied; it remains to be determined if the indirect pathwayalso applies at very high AR concentrations.

5. CONCLUSION

We have used single-molecule fluorescence microscopy toimage quantitatively organic substrate oxidation reactions onthe surfaces of individual single-crystalline rutile TiO2 nanorodphotoanodes under photoelectrocatalytic water oxidationconditions in operando. The single-molecule approach allowedfor quantitation of amplex red oxidation reaction kinetics that isotherwise difficult to achieve in a conventional ensemble-levelexperiment. The scaling relations for the substrate oxidationrate versus applied potential at a fixed light intensity and versuslight intensity at a fixed applied potential, together with thephotocurrent scaling relations, were consistent with a kineticmodel whereby the substrate molecule was oxidized indirectlyby photogenerated surface adsorbed hydroxyl radical species(or terminal O•− species on the surface) rather than directly byphotogenerated holes.

■ ASSOCIATED CONTENT

*S Supporting InformationThe Supporting Information is available free of charge on theACS Publications website at DOI: 10.1021/acs.jpcc.6b01848.

Explanation of an alternative indirect oxidation mecha-nism, determination of surface coverage of TiO2

nanorods, surface recombination probability, and amplexred oxidation in the presence of H2O2 (PDF)

■ AUTHOR INFORMATION

Corresponding Author*Phone: (607) 254-8533. E-mail: pc252@cornell.edu.

Author ContributionsJ.B.S. and P.C. designed the research. J.B.S. performed research.J.B.S. and P.C. analyzed the data and wrote the manuscript.

NotesThe authors declare no competing financial interest.

■ ACKNOWLEDGMENTS

We thank Dr. Eric Choudhary for electron microscopymeasurements and Prof. H. D. Abruna for electrochemicalinstruments. The research is supported by the Department ofEnergy, Office of Science, Basic Energy Science (Award DE-FG02-10ER16199), and in part by Army Research Office(Grants 63767-CH and 65814-CH), National ScienceFoundation (Grant CBET-1263736; Grant CHE-1137217(J.B.S.)), and Petroleum Research Foundation (Grant 54289-ND7). This work used the Cornell Center for MaterialsResearch Shared Facilities, a NSF MRSEC program (GrantDMR-1120296), and the Cornell NanoScale Facility, a memberof the NSF National Nanotechnology Infrastructure Network(Grant ECCS-15420819).

■ ABBREVIATIONS

AR, amplex red; i, photocurrent; E, applied potential vs Ag/AgCl reference electrode

■ REFERENCES(1) Fox, M. A.; Dulay, M. T. Heterogeneous photocatalysis. Chem.Rev. 1993, 93, 341−357.(2) Frank, S. N.; Bard, A. J. Semiconductor electrodes. 12.Photoassisted oxidations and photoelectrosynthesis at polycrystallinetitanium dioxide electrodes. J. Am. Chem. Soc. 1977, 99, 4667−4675.(3) Hoffmann, M. R.; Martin, S. T.; Choi, W.; Bahnemann, D. W.Environmental applications of semiconductor photocatalysis. Chem.Rev. 1995, 95, 69−96.(4) Osterloh, F. E.; Parkinson, B. A. Recent developments in solarwater-splitting photocatalysis. MRS Bull. 2011, 36, 17−22.(5) Augugliaro, V.; Camera-Roda, G.; Loddo, V.; Palmisano, G.;Palmisano, L.; Soria, J.; Yurdakal, S. Heterogeneous photocatalysis andphotoelectrocatalysis: from unselective abatement of noxious speciesto selective production of high-value chemicals. J. Phys. Chem. Lett.2015, 6, 1968−1981.(6) Fujishima, A.; Rao, T. N.; Tryk, D. A. Titanium dioxidephotocatalysis. J. Photochem. Photobiol., C 2000, 1, 1−21.(7) Li, Y.; Wen, B.; Yu, C.; Chen, C.; Ji, H.; Ma, W.; Zhao, J. Pathwayof oxygen incorporation from O2 in TiO2 photocatalytic hydrox-ylation of aromatics: oxygen isotope labeling studies. Chem. - Eur. J.2012, 18, 2030−2039.(8) Turchi, C. S.; Ollis, D. F. Photocatalytic degradation of organicwater contaminants: Mechanisms involving hydroxyl radical attack. J.Catal. 1990, 122, 178−192.(9) Monllor-Satoca, D.; Gomez, R.; Gonzalez-Hidalgo, M.; Salvador,P. The “Direct−Indirect” model: An alternative kinetic approach inheterogeneous photocatalysis based on the degree of interaction ofdissolved pollutant species with the semiconductor surface. Catal.Today 2007, 129, 247−255.(10) Villarreal, T. L.; Gomez, R.; Neumann-Spallart, M.; Alonso-Vante, N.; Salvador, P. Semiconductor photooxidation of pollutantsdissolved in water: a kinetic model for distinguishing between directand indirect interfacial hole transfer. I. photoelectrochemical experi-ments with polycrystalline anatase electrodes under current doublingand absence of recombination. J. Phys. Chem. B 2004, 108, 15172−15181.(11) Peterson, M. W.; Turner, J. A.; Nozik, A. J. Mechanistic studiesof the photocatalytic behavior of titania: particles in a photo-electrochemical slurry cell and the relevance to photodetoxificationreactions. J. Phys. Chem. 1991, 95, 221−225.(12) Davydov, L.; Smirniotis, P. G. Quantification of the primaryprocesses in aqueous heterogeneous photocatalysis using single-stageoxidation reactions. J. Catal. 2000, 191, 105−115.(13) Upadhya, S.; Ollis, D. F. Simple photocatalysis model forphotoefficiency enhancement via controlled, periodic illumination. J.Phys. Chem. B 1997, 101, 2625−2631.(14) Prairie, M. R.; Evans, L. R.; Stange, B. M.; Martinez, S. L. Aninvestigation of titanium dioxide photocatalysis for the treatment ofwater contaminated with metals and organic chemicals. Environ. Sci.Technol. 1993, 27, 1776−1782.(15) Bideau, M.; Claudel, B.; Otterbein, M. Photocatalysis of formicacid oxidation by oxygen in an aqueous medium. J. Photochem. 1980,14, 291−302.(16) Kesselman, J. M.; Shreve, G. A.; Hoffmann, M. R.; Lewis, N. S.Flux-matching conditions at TiO2 photoelectrodes: is interfacialelectron transfer to O2 rate-limiting in the TiO2-catalyzed photo-chemical degradation of organics? J. Phys. Chem. 1994, 98, 13385−13395.(17) Mora-Sero, I.; Villarreal, T. L.; Bisquert, J.; Pitarch, A.; Gomez,R.; Salvador, P. Photoelectrochemical behavior of nanostructuredTiO2 thin-film electrodes in contact with aqueous electrolytescontaining dissolved pollutants: A model for distinguishing betweendirect and indirect interfacial hole transfer from photocurrentmeasurements. J. Phys. Chem. B 2005, 109, 3371−3380.(18) Morrison, S. R.; Freund, T. Chemical role of holes and electronsin ZnO photocatalysis. J. Chem. Phys. 1967, 47, 1543−1551.(19) Monllor-Satoca, D.; Borja, L.; Rodes, A.; Gomez, R.; Salvador,P. Photoelectrochemical behavior of nanostructured WO3 thin-film

The Journal of Physical Chemistry C Article

DOI: 10.1021/acs.jpcc.6b01848J. Phys. Chem. C 2016, 120, 20668−20676

20675

electrodes: the oxidation of formic acid. ChemPhysChem 2006, 7,2540−2551.(20) Waldner, G.; Go mez, R.; Neumann-Spallart, M. Usingphotoelectrochemical measurements for distinguishing between directand indirect hole transfer processes on anatase: Case of oxalic acid.Electrochim. Acta 2007, 52, 2634−2639.(21) Sambur, J. B.; Chen, T. Y.; Choudhary, E.; Chen, G.; Nissen, E.J.; Thomas, E. M.; Zou, N.; Chen, P. Sub-particle reaction andphotocurrent mapping to optimize catalyst-modified photoanodes.Nature 2016, 530, 77−80.(22) Liu, B.; Chen, H. M.; Liu, C.; Andrews, S. C.; Hahn, C.; Yang, P.Large-scale synthesis of transition-metal-doped TiO2 nanowires withcontrollable overpotential. J. Am. Chem. Soc. 2013, 135, 9995−9998.(23) Zhou, X.; Andoy, N. M.; Liu, G.; Choudhary, E.; Han, K.-S.;Shen, H.; Chen, P. Quantitative super-resolution imaging uncoversreactivity patterns on single nanocatalysts. Nat. Nanotechnol. 2012, 7,237−241.(24) Han, K. S.; Liu, G.; Zhou, X.; Medina, R. E.; Chen, P. How doesa single Pt nanocatalyst behave in two different reactions? a single-molecule study. Nano Lett. 2012, 12, 1253−1259.(25) Xu, W.; Jain, P. K.; Beberwyck, B. J.; Alivisatos, A. P. ProbingRedox Photocatalysis of Trapped Electrons and Holes on Single Sb-doped Titania Nanorod Surfaces. J. Am. Chem. Soc. 2012, 134, 3946−3949.(26) Zhang, Y.; Lucas, J. M.; Song, P.; Beberwyck, B.; Fu, Q.; Xu, W.;Alivisatos, A. P. Superresolution fluorescence mapping of single-nanoparticle catalysts reveals spatiotemporal variations in surfacereactivity. Proc. Natl. Acad. Sci. U. S. A. 2015, 112, 8959−8964.(27) Han, R.; Ha, J. W.; Xiao, C. X.; Pei, Y. C.; Qi, Z. Y.; Dong, B.;Bormann, N. L.; Huang, W. Y.; Fang, N. Geometry-assisted three-dimensional superlocalization imaging of single-molecule catalysis onmodular multilayer nanocatalysts. Angew. Chem., Int. Ed. 2014, 53,12865−12869.(28) Ha, J. W.; Ruberu, T. P. A.; Han, R.; Dong, B.; Vela, J.; Fang, N.Super-resolution mapping of photogenerated electron and holeseparation in single metal-semiconductor nanocatalysts. J. Am. Chem.Soc. 2014, 136, 1398−1408.(29) Mayer, K. M.; Shnipes, J.; Davis, D.; Walt, D. R. Catalytickinetics of single gold nanoparticles observed via optical microwellarrays. Nanotechnology 2015, 26, 055704.(30) Frank, A. J.; Kopidakis, N.; Lagemaat, J. v. d. Electrons innanostructured TiO2 solar cells: transport, recombination andphotovoltaic properties. Coord. Chem. Rev. 2004, 248, 1165−1179.(31) Gartner, W. W. Depletion-layer photoeffects in semiconductors.Phys. Rev. 1959, 116, 84−87.(32) Lindquist, S. E.; Finnstrom, B.; Tegner, L. Photoelectrochemicalproperties of polycrystalline TiO2 thin film electrodes on quartzsubstrates. J. Electrochem. Soc. 1983, 130, 351−358.(33) Fabrega, C.; Monllor-Satoca, D.; Ampudia, S.; Parra, A.; Andreu,T.; Morante, J. R. Tuning the fermi level and the kinetics of surfacestates of TiO2 nanorods by means of ammonia treatments. J. Phys.Chem. C 2013, 117, 20517−20524.(34) Hoang, S.; Guo, S.; Hahn, N. T.; Bard, A. J.; Mullins, C. B.Visible light driven photoelectrochemical water oxidation on nitrogen-modified TiO2 nanowires. Nano Lett. 2012, 12, 26−32.(35) Butler, M. A. Photoelectrolysis and physical properties of thesemiconducting electrode WO3. J. Appl. Phys. 1977, 48, 1914−1920.(36) Salvador, P. Kinetic approach to the photocurrent transients inwater photoelectrolysis at n-titanium dioxide electrodes. 1. Analysis ofthe ratio of the instantaneous to steady-state photocurrent. J. Phys.Chem. 1985, 89, 3863−3869.(37) Gutierrez, C.; Salvador, P. Mechanisms of CompetitivePhotoelectrochemical Oxidation of I− and H2O at n-TiO2 Electrodes:A Kinetic Approach. J. Electrochem. Soc. 1986, 133, 924−929.(38) Fermin, D. J.; Ponomarev, E. A.; Peter, L. M. Dynamics ofphoto-processes at the n-TiO2 aqueous electrolyte interface. InSymposium on Photoelectrochemistry; Electrochemical Society Proceed-ings, Vol. 97; Rajeshwar, K., Ed.; Electrochemical Society Inc:Pennington, NJ,1997; pp 62−71.

(39) Montoya, J. F.; Bahnemann, D. W.; Peral, J.; Salvador, P.Catalytic role of TiO2 terminal oxygen atoms in liquid-phasephotocatalytic reactions: oxidation of aromatic compounds inanhydrous acetonitrile. ChemPhysChem 2014, 15, 2311−2320.(40) Roy, B.; Ahrenkiel, S. P.; Fuierer, P. A. Controlling the size andmorphology of TiO2 powder by molten and solid salt synthesis. J. Am.Ceram. Soc. 2008, 91, 2455−2463.(41) Chen, T.-Y.; Santiago, A. G.; Jung, W.; Krzeminski, L.; Yang, F.;Martell, D. J.; Helmann, J. D.; Chen, P. Concentration- andchromosome-organization-dependent regulator unbinding from DNAfor transcription regulation in living cells. Nat. Commun. 2015, 6, 7445.(42) Le Formal, F.; Sivula, K.; Gra tzel, M. The transientphotocurrent and photovoltage behavior of a hematite photoanodeunder working conditions and the influence of surface treatments. J.Phys. Chem. C 2012, 116, 26707−26720.(43) Sagara, T.; Sukigara, M. Photogenerated Surface States at n-TiO2/Aqueous Solution Interface. J. Electrochem. Soc. 1988, 135, 363−367.(44) Abrantes, L. M.; Peter, L. M. Transient photocurrents at passiveiron electrodes. J. Electroanal. Chem. Interfacial Electrochem. 1983, 150,593−601.(45) Bockris, J. O. M.; Uosaki, K. The rate of the photo-electrochemical generation of hydrogen at p-type semiconductors. J.Electrochem. Soc. 1977, 124, 1348−1355.(46) Mollers, F.; Tolle, H. J.; Memming, R. On the origin of thephotocatalytic deposition of noble metals on TiO2. J. Electrochem. Soc.1974, 121, 1160−1167.(47) Peter, L. M.; Li, J.; Peat, R. Surface recombination atsemiconductor electrodes: Part I. Transient and steady-state photo-currents. J. Electroanal. Chem. Interfacial Electrochem. 1984, 165, 29−40.(48) Klahr, B. M.; Hamann, T. W. Current and voltage limitingprocesses in thin film hematite electrodes. J. Phys. Chem. C 2011, 115,8393−8399.(49) Klahr, B.; Gimenez, S.; Fabregat-Santiago, F.; Bisquert, J.;Hamann, T. W. Electrochemical and photoelectrochemical inves-tigation of water oxidation with hematite electrodes. Energy Environ.Sci. 2012, 5, 7626−7636.(50) Sivula, K. Metal oxide photoelectrodes for solar fuel production,surface traps, and catalysis. J. Phys. Chem. Lett. 2013, 4, 1624−1633.(51) Salvador, P.; Gutierrez, C. Analysis of the transient photo-current-time behaviour of a sintered n-SrTiO3 electrode in waterphotoelectrolysis. J. Electroanal. Chem. Interfacial Electrochem. 1984,160, 117−130.(52) Peter, L. M. Dynamic aspects of semiconductor photo-electrochemistry. Chem. Rev. 1990, 90, 753−769.(53) Kong, D.-S.; Wei, Y.-J.; Li, X.-X.; Zhang, Y.; Feng, Y.-Y.; Li, W.-J.pH dependent behavior and effects of photoinduced surface statesduring water photooxidation at TiO2/solution interface: studied bycapacitance measurements. J. Electrochem. Soc. 2014, 161, H144−H153.(54) Bolts, J. M.; Wrighton, M. S. Correlation of photocurrent-voltage curves with flat-band potential for stable photoelectrodes forthe photoelectrolysis of water. J. Phys. Chem. 1976, 80, 2641−2645.(55) Cooper, G.; Turner, J. A.; Nozik, A. J. Mott−Schottky plots andflatband potentials for single crystal rutile electrodes. J. Electrochem.Soc. 1982, 129, 1973−1977.(56) Iwanski, P.; Curran, J. S.; Gissler, W.; Memming, R. Thephotoelectrochemical behavior of ferric oxide in the presence of redoxreagents. J. Electrochem. Soc. 1981, 128, 2128−2133.(57) Laser, D.; Gottesfeld, S. Photocurrents induced by subbandgapillumination in a Ti-oxide film electrode. J. Electrochem. Soc. 1979, 126,475−478.(58) Preusser, S.; Stimming, U.; Tokunaga, S. The wavelength andpotential dependency of spatially resolved photoelectrochemicalmeasurements on TiO2. J. Electrochem. Soc. 1995, 142, 102−111.(59) Zhang, Q.; Celorrio, V.; Bradley, K.; Eisner, F.; Cherns, D.; Yan,W.; Fermín, D. J. Density of deep trap states in oriented TiO2nanotube arrays. J. Phys. Chem. C 2014, 118, 18207−18213.

The Journal of Physical Chemistry C Article

DOI: 10.1021/acs.jpcc.6b01848J. Phys. Chem. C 2016, 120, 20668−20676

20676