A Mechanistic Study of the Oxidative Reaction of Hydrogen- Terminated Si(111) Surfaces with Liquid Methanol

Transcription

1 Article Subscriber access provided by Caltech Library A Mechanistic Study of the Oxidative Reaction of Hydrogen- Terminated () Surfaces with Liquid Methanol Noah T. Plymale, Mita Dasog, Bruce S. Brunschwig, and Nathan S. Lewis J. Phys. Chem. C, Just Accepted Manuscript DOI:./acs.jpcc.b Publication Date (Web): Jan Downloaded from on January, Just Accepted Just Accepted manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides Just Accepted as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. Just Accepted manuscripts appear in full in PDF format accompanied by an HTML abstract. Just Accepted manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI ). Just Accepted is an optional service offered to authors. Therefore, the Just Accepted Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the Just Accepted Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these Just Accepted manuscripts. The Journal of Physical Chemistry C is published by the American Chemical Society. xteenth Street N.W., Washington, DC 0 Published by American Chemical Society. Copyright American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

2 Page of The Journal of Physical Chemistry 0 A Mechanistic Study of the Oxidative Reaction of Hydrogen-Terminated () Surfaces with Liquid Methanol Noah T. Plymale, Mita Dasog, Bruce S. Brunschwig, and Nathan S. Lewis*,,, Division of Chemistry and Chemical Engineering, Beckman Institute, and Kavli Nanoscience Institute, California Institute of Technology, Pasadena, California, United States ABSTRACT H () surfaces have been reacted with liquid methanol (CH OH) in the absence or presence of a series of oxidants and/or illumination. Oxidant-activated methoxylation of H () surfaces was observed in the dark after exposure to CH OH solutions that contained the one-electron oxidants acetylferrocenium, ferrocenium, or, -dimethylferrocenium. The oxidant-activated reactivity toward CH OH of intrinsic and n-type H () surfaces increased upon exposure to ambient light. The results suggest that oxidant-activated methoxylation requires that two conditions be met: () the position of the quasi-fermi levels must energetically favor oxidation of the H () surface and () the position of the quasi-fermi levels must energetically favor reduction of an oxidant in solution. Consistently, illuminated n-type H () surfaces underwent methoxylation under applied external bias more rapidly and at more negative potentials than p-type H () surfaces. The results under potentiostatic control indicate that only conditions that favor oxidation of the H () surface need be met, with charge balance at the surface maintained by current flow at the back of the electrode. The results are described by a mechanistic framework that analyzes the positions of the quasi-fermi levels relative to the energy levels relevant for each system.

3 The Journal of Physical Chemistry Page of 0 I. INTRODUCTION The device properties of can be manipulated through control over the structure and chemical composition of crystalline surfaces. - The reactivity of hydrogen-terminated () (H ()) surfaces toward organic nucleophiles, including alkenes, - alkynes, - amines, - thiols and disulfides, - Grignards, - and alcohols, - has been widely exploited to impart desirable functionality to the interface. These surface reactions have been used to control the interface between and metals, - metal oxides, - polymers, - and redox assemblies. - Many nucleophiles react with H () surfaces to yield unexpected products 0- compared to analogous reactions with molecular silanes. For example, the reaction of H () surfaces with alcohols has no analogous molecular counterpart., The unique reactivity of H () surfaces with alcohols, including CH OH, has been exploited as a versatile method to impart a desired functionality to the surface via the robust O bond, without formation of a thick insulating silicon oxide layer on the surface. For example, n-/ch OH junctions have yielded high open-circuit voltages ( mv) and high device efficiencies ( %) - in regenerative photoelectrochemical cells, with the device performance correlated with low surface recombination velocities - as well as the favorable band-edge positions of the methoxylated surface. The methoxy termination can moreover be converted to F-termination or OH-, termination, both surfaces. of which are synthetically difficult to achieve otherwise on () Small molecule silanes, such as tris(trimethylsilyl)silane (TTMSS), are generally good models for radical reactions of H () surfaces, because the H bond strength is comparable for both systems. However, in the presence of CH OH, TTMSS decomposes by bond cleavage., milarly, the bonds on H-terminated nanoporous surfaces,

4 Page of The Journal of Physical Chemistry 0 which typically have wider band gaps than planar, cleave by thermal reaction with alcohols., In contrast, the H () surface primarily undergoes a substitution reaction with alcohols, resulting in OR functionality on the surface. Scanning-tunneling microscopy (STM) data have demonstrated that the H () surface undergoes concomitant methoxylation and microscopic etching after min in warm ( C) CH OH, but, as surface sites are methoxylated, the methoxylated sites become less likely to undergo etching. The electronic structure of crystalline differs significantly from that of small molecule silanes or nanoporous, and is expected to play an important role in reactions that involve charge-transfer with the surface. The general understanding of the alkoxylation mechanism has been focused primarily on a qualitative discussion that implicates both the narrow band gap and the much larger number of electronic states of bulk than of analogous molecular silanes. The previously proposed mechanisms and net chemical reactions of the methoxylation of H () surfaces (reproduced in Scheme and Scheme ) provide a description of the chemical transformations that take place during surface alkoxylation, but the kinetic formalism that governs the oxidative methoxylation of H () surfaces is not clearly defined. Elucidating the mechanistic framework that governs the reaction of H () surfaces with CH OH in the presence of a series of oxidants thus offers an opportunity to determine the role that the electronic structure plays in oxidative reactions at surfaces. Scheme. Methoxylation of H () Surfaces in the Absence of a Molecular Oxidant. H δ- H δ+ δ+ H δ- CH O Coulombic Attraction Net: H H H O CH Electron H Rearrangement H CH H O Structural Rearrangement H + CH OH OCH + H H O CH

5 The Journal of Physical Chemistry Page of 0 Scheme. Oxidant-Activated Methoxylation of H (). H H Herein we characterize the reaction of H () surfaces with CH OH in the absence or presence of an oxidant and in the absence or presence of visible-light illumination. We have elucidated the dependence of the methoxylation reaction on the formal potential of the oxidant. Additionally, we have delineated the roles of the electronic surface states and bulk band structure in the electron-transfer processes that are involved in the methoxylation reaction. A general mechanism by which the reaction of H () surfaces with CH OH takes place under the conditions studied is proposed in the context of a charge transfer kinetics framework at the semiconductor interface. II. EXPERIMENAL H CH O A A H H H H CH CH H + H A A O H H O II.A. Materials and Methods. Water (. MΩ cm resistivity) was obtained from a Barnstead E-Pure system. Aqueous ammonium fluoride (NH F(aq), % semiconductor grade, Transene Co. Inc., Danvers MA) was deaerated by bubbling with Ar(g) (.%, Air Liquide) for h prior to use. Methanol (CH OH, anhydrous,.%, gma-aldrich) was used as received. Float-zone-grown, phosphorus-doped n-() intrinsic wafers (University Wafer, Boston, MA) were double-side polished, ± µm thick, oriented within 0. of the () crystal plane, and had a resistivity of.0 kω cm. Czochralski-grown, phosphorus-doped n-() wafers (Virginia Semiconductor, Fredericksburg, VA) were double-side polished, ± µm thick, oriented within 0. of the () crystal plane, and had a resistivity of.0 Ω cm. Czochralski-grown, boron-doped p-() wafers (Addison Engineering, San Jose, CA) were H H O Net: H + CH OH + A OCH + H + + A CH H + H O CH

6 Page of The Journal of Physical Chemistry 0 double-side polished, 0 ± µm thick, oriented within 0. of the () crystal plane, and had a resistivity of 0. Ω cm. The dopant densities (N d ) were determined from the resistivities for each substrate. For intrinsic (), N d =. cm ; for n-type (), N d =. cm ; and for p- type (), N d =. cm. The positions of the n-type and p-type () bulk Fermi levels (E F,n,b and E F,p,b, respectively) were determined using eq and eq, respectively, for each of the dopant concentrations used. E F,n,b = E cb + k B T ln N d () N C E F,p,b = E vb k B T ln N d () E cb and E vb are the energy positions of the conduction band minimum (.0 ev vs E Vac ) and valence band maximum (. ev vs E Vac ), respectively, k B is the Boltzmann constant, T is the temperature in Kelvin ( K), q is the absolute value of the elementary charge, N C is the effective density of states of the conduction band (. cm ) and N V is the effective density of states of the valence band (.0 cm ). The bulk Fermi level of intrinsic () (E F,i,b ), which was lightly n-type, was determined using eq. II.A.. Preparation and Purification of the Oxidants., -dimethylferrocenium (Me Cp Fe + BF, bis(methylcyclopentadienly)iron(iii) tetrafluoroborate), octamethylferrocenium (Me Cp Fe + BF, bis(tetramethylcyclopentadienyl)iron(iii) tetrafluoroborate), and decamethylferrocenium (Cp* Fe + BF, bis(pentamethylcyclopentadienyl)iron(iii) tetrafluoroborate) were prepared by chemical oxidation of the neutral metallocenes -, -dimethylferrocene (Me Cp Fe, bis(methylcyclopentadienyl)iron(ii), %, gma-aldrich), octamethylferrocene (Me Cp Fe, bis(tetramethylcyclopentadienyl)iron(ii), %, Strem Chemical), and N V

7 The Journal of Physical Chemistry Page of 0 decamethylferrocene (Cp* Fe, bis(pentamethylcyclopentadienyl)iron(ii), %, Strem Chemical), respectively. Ferrocenium (Cp Fe + BF, bis(cyclopentadienyl)iron(iii) tetrafluoroborate, technical grade, gma-aldrich) Me Cp Fe + BF, Me Cp Fe + BF, and Cp* Fe + BF were purified by recrystallization from diethyl ether (inhibitor-free,.% gma-aldrich) and CH CN (anhydrous,.% gma-aldrich). Methyl viologen (MV + (Cl ),, -dimethyl-, - bipyridinium dichloride hydrate, %, gma-aldrich) and cobaltocenium (Cp Co + PF, bis(cyclopentadienyl)cobalt(iii) hexafluorophosphate, %, gma-aldrich) were recrystallized from ethanol (anhydrous,.% gma-aldrich) prior to use. Acetylferrocenium ((CpCOCH )CpFe +, (acetylcyclopentadienyl)cyclopentadienyliron(iii)) was generated in CH OH containing.0 M LiClO (battery grade, gma-aldrich) from acetylferrocene ((CpCOCH )CpFe, acetylcyclopentadienyl)cyclopentadienyliron(ii), gma-aldrich, purified by sublimation) by controlled-potential electrolysis at a Pt mesh working electrode with a Pt mesh counter electrode located in a compartment separated from the working electrode by a Vycor frit. The electrolysis was performed at +0. V vs a Pt pseudo-reference electrode, and sufficient current was passed to generate.0 mm (CpCOCH )CpFe + from mm (CpCOCH )CpFe. The (CpCOCH )CpFe + was used within min of its generation. II.A.. Preparation of H () Surfaces. wafers were scored and broken to the desired size using a diamond- or carbide-tipped scribe. The surfaces were rinsed sequentially with water, methanol (.%, BDH), acetone (.%, BDH), methanol, and water. The samples were oxidized in a piranha solution (: v/v of % H O (aq) (EMD): M H SO (aq) (EMD)) for min at ± C. The wafers were removed, rinsed with copious amounts of water, and immersed in buffered hydrofluoric acid(aq) (BHF, semiconductor grade, Transene Co. Inc.) for s. The BHF solution was drained and the wafers were rinsed with H O. Anisotropic

8 Page of The Journal of Physical Chemistry 0 etching was then performed in an Ar(g)-purged solution of % NH F(aq) for. min for wafers having a 0. miscut angle, and for.0 min for wafers having a 0. miscut angle. The wafers were removed, rinsed with H O, and dried under a stream of Ar(g). II.A.. Methoxylation of H () Surfaces. The H () wafers were transferred to a N (g)-purged glovebox (< ppm O (g)) and immersed in either neat CH OH or in a solution of CH OH that contained.0 mm of (CpCOCH )CpFe +, Cp Fe + BF, Me Cp Fe + BF, Me Cp Fe + BF, Cp* Fe + BF, MV + Cl, or Cp Co + PF. Reactions in the dark were performed in test tubes wrapped in black vinyl electrical tape with the top covered in Al foil, whereas reactions in the light were performed under ambient illumination. Upon completion of the reaction, the wafers were removed from the CH OH solution and rinsed sequentially with CH OH and tetrahydrofuran (THF, anhydrous,.%, gma-aldrich). The THF was allowed to evaporate and the samples were removed from the glovebox. Prior to analysis, the wafers were rinsed briefly with H O and dried under a stream of Ar(g). II.A.. Potentiostatic Methoxylation of H (). samples were cleaned using a piranha solution for min at ± C and etched in BHF for s prior to electrode fabrication. Ohmic contacts to the back of n-() (.0 Ω cm) electrodes were made by using a diamondtipped scribe to apply Ga-In eutectic (% Ga, % In by weight). Ohmic contacts to the back of p-() (0. Ω cm) electrodes were made by electron-beam evaporation (Denton Vacuum) of 0 nm of Al to the rear face of the electrode, followed by a min anneal under forming gas (% H (g) in N (g)) at 0 C. Using high-purity conductive Ag paint (SPI supplies, West Chester, PA), the working electrodes were adhered to a coil of tinned Cu wire that had been threaded through / outer-diameter Pyrex tubing. Loctite epoxy was used to insulate the rear face of the electrode from the electrolyte as well as to immobilize the wafer such that the electrode

9 The Journal of Physical Chemistry Page of 0 face was perpendicular to the length of the tubing. The electrode areas (0. 0. cm ) were measured by analysis of scanned images of the electrodes using ImageJ software. Immediately prior to performing the electrochemical experiments, the electrodes were immersed in BHF for s, rinsed with water, and dried thoroughly. The electrodes were then immersed in % NH F(aq), with wafers having a 0. miscut angle immersed for. min whereas wafers that had a 0. miscut angle were immersed for.0 min. The electrodes were rinsed with water and dried thoroughly prior to performing electrochemical measurements. Electrochemical measurements were performed in a four-port, cylindrical, flat-bottomed, borosilicate glass cell that contained CH OH with.0 M LiClO. Electrochemical measurements were collected using a Solartron model potentiostat operated by CorrWare software v..c. Current density vs potential (J-E) measurements were collected using a -electrode setup with a working electrode, a saturated calomel reference electrode (SCE, CH instruments, Inc., Austin TX), and a Pt mesh counter electrode. J-E data were collected by sweeping the potential from negative to positive using a scan rate of 0 mv s and a sampling rate of mv per data point, with active stirring of the electrolyte. Measurements were performed in the dark as well as under mw cm of illumination intensity provided by a 0 W ELH-type W-halogen lamp. The illumination intensity was determined by use of a calibrated photodiode (Thor Laboratories, Newton, NJ). II.A.. Determination of the Formal Potentials of the Oxidants. The formal potentials of the oxidant species (.0 mm in CH OH) were determined by cyclic voltammetry vs a Cp Fe +/0 internal reference using Pt wire working and reference electrodes with a Pt mesh counter electrode. The formal potential of the Cp Fe +/0 couple was measured vs a SCE reference (E (Cp Fe +/0 ) = 0. V vs SCE), and this potential was used to convert from V vs Cp Fe +/0 to

10 Page of The Journal of Physical Chemistry 0 V vs SCE. The supporting electrolyte was.0 M LiClO for (CpCOCH )CpFe +/0, Cp Fe +/0, Me Cp Fe +/0, Me Cp Fe +/0, Cp* Fe +/0, and Cp Co +/0 and was 0. M tetrabutylammonium hexafluorophosphate (TBAPF, gma-aldrich) for MV +/+. Nernstian solution potentials (E(A/A )) and formal reduction potentials (E (A/A )) were converted to absolute electron energies using a value of. ev vs E vac for the absolute electrochemical potential of the standard hydrogen electrode (SHE) at C. -0 Because the redox potential of SCE is +0. V vs SHE, E(A/A ) and E (A/A ) in V vs SCE can be readily converted to solution energies (E(A/A )) and formal reduction energies (E (A/A )) in ev vs E Vac using eqs and, respectively. ( ) vs E Vac = E ( A/A ) vs SCE E A/A E ʹ A/A ( ). ev () ( ). ev () ( ) vs E Vac = E ʹ( A/A ) vs SCE II.B. Instrumentation. II.B.. Transmission Infrared Spectroscopy. Transmission infrared spectroscopy (TIRS) was performed using a Thermo Scientific Nicolet 00 optical spectrometer equipped with an electronically temperature-controlled EverGlo mid-ir source, a KBr beamsplitter, a deuterated L-alanine-doped triglycine sulfate (DLaTGS) detector, and a N (g) purge. - The samples, cut to ~.. cm, were mounted using a custom attachment such that the angle between the path of the beam and the surface normal was (Brewster s angle for ). The reported spectra are averages of 00 consecutive scans collected at cm resolution. Spectra were referenced to a spectrum of a wafer that was H-terminated or that was covered with silicon oxide (O x ), with reference spectra collected separately for each sample. Data were collected and processed using OMNIC software v.... The baseline was flattened and residual water peaks were subtracted to produce the reported spectral data.

11 The Journal of Physical Chemistry Page of 0 II.B.. X-ray Photoelectron Spectroscopy. X-ray photoelectron spectroscopic (XPS) data were collected using a Kratos AXIS Ultra spectrometer equipped with a monochromatic Al Kα X-ray source at. ev, a hybrid electrostatic and magnetic lens system, and a delay-line detector. - Photoelectrons from a 00 µm 0 µm area were ejected at 0 with respect to the sample surface. Survey and high-resolution spectra were collected with an analyzer pass energy of 0 ev and ev, respectively. The chamber base pressure was ~ Torr. Calibration of the energy scale and work function was performed using clean Au, Ag, and Cu samples. The data were collected using Vision Manager software v... revision. II.C. Analysis of TIRS and XPS Data. II.C.. Calculation of the Fractional Methoxy Monolayer Coverage from TIRS Data. The fractional monolayer (ML) coverage of H () sites remaining after reaction in CH OH (θ H ) was determined using eq θ H = A ν ( H), f A ν ( H),i () where A ν( H),f and A ν( H),i are the final and initial area, respectively, under the H stretching peak. Assuming that all atop sites are terminated by H or OCH groups, the fractional ML coverage of CH O () sites (θ OCH ) was determined using eq θ OCH = θ H () II.C.. Calculation of the Fractional Methoxy Monolayer Coverage from XPS Data. High-resolution C s and p XPS data were analyzed using CasaXPS v.... C s spectra were fitted using a Voigt GL() line function, which consisted of 0% Gaussian and % Lorentzian character. Bulk p data were fitted using a line function of the form LA(a, b, n), where a and b define the asymmetry of the line shape and n defines the Gaussian width. LA(.,

12 Page of The Journal of Physical Chemistry 0., 0) was used in this work. Contributions from O x were fitted using the GL() line shape. Spectra were analyzed using a linear background. A substrate-overlayer model was used to determine the thickness of the methoxy overlayer, d OCH, by XPS: I C O I SF SF C O ρ ρ C O = d OCH λ C O sinϕ e e d OCH λ sinϕ I C O and I are the areas under the photoemission peaks arising from C bound to O and from bulk, respectively, SF C O and SF are the sensitivity factors for the C s (0.) and p (0.) photoemission signals, respectively, ρ C O and ρ are the density of the hydrocarbon overlayer (.0 g cm ) and (. g cm ), respectively, λ C O and λ are the attenuation lengths of C s photoelectrons (. nm) and p photoelectrons (.0 nm), respectively, moving through a hydrocarbon overlayer, and φ is the angle between the photoelectron emission vector and the sample surface normal (0 for this work). The thickness d OCH was determined using an iterative calculation process, and θ OCH was calculated by dividing the calculated d OCH by the estimated thickness of ML of methoxy groups (0. nm). III. RESULTS III.A. Reaction of H () with CH OH Solutions in the Dark. Figure presents time dependence for the dark reaction of H () surfaces with CH OH in the absence or presence of Cp Fe +. The reaction of H () with CH OH proceeded slowly in the absence of electron acceptors in solution, but increased substantially upon addition of Cp Fe +. This observation agrees with previous results that describe increased reaction rates for the methoxylation of H () in CH OH solutions that contained Cp Fe + or Me Cp Fe +,, -. ()

13 The Journal of Physical Chemistry Page of 0 After min of reaction, a difference in the area of the H stretching peak, ν( H), was observed in the presence of Cp Fe + relative to the reaction in the absence of the deliberately added oxidant. A greater difference was observed for a min reaction time, but measurable subsurface silicon oxide (O x ) was consistently observed after min in the presence of Cp Fe +. Hence, a min reaction time was chosen as a standard point of comparison in this work. q -H (ML).0 Neat CH OH mm Cp Fe Reaction Time (min) Figure. Time dependence of the reaction of H () with neat CH OH (black squares) and CH OH containing.0 mm Cp Fe + (red circles) in the dark. θ H was determined from TIRS measurements using eq. The blue triangle represents the assumed.0 ML H on surfaces that have not been reacted with CH OH. The reaction of H () surfaces with liquid CH OH was performed for min in solutions that contaianed.0 mm of the oxidants indicated in Figure as well as in neat CH OH. The experimentally determined formal potentials, E (A/A ), for each oxidant are indicated relative to the energy positions of the valence-band maximum, the conduction-band minimum, and the Fermi levels of the planar samples used in this work. The formal potentials of the oxidants span the range from below the valence-band maximum to above the

14 Page of The Journal of Physical Chemistry 0 conduction-band minimum. Note that because only the oxidized species was added to the CH OH reaction solutions for all reagents except (CpCOCH )CpFe +, the Nernstian solution potentials were thus shifted positive of the measured formal potentals. Energy (ev vs E Vac ) E cb = -.0 ev E F,n,b = -. ev E F,i,b = -. ev -. E F,p,b = -.0 ev E vb = -. ev E (Cp Co +/0 ) = -.0 V E (MV +/+ ) = -0.0 V E (Cp* Fe +/0 ) = -0. V E (Me Cp Fe +/0 ) = -0.0 V E (Me Cp Fe +/0 ) = +0. V E (Cp Fe +/0 ) = +0. V E ((CpCOCH )CpFe +/0 ) = +0. V Figure. Energy diagram showing the relative energy positions vs the vacuum level (E Vac ) of the valence-band maximum (E vb ), conduction-band minimum (E cb ), and the calculated bulk Fermi levels (eq and eq ) of the intrinsic () (E F,i,b ), n-type () (E F,n,b ), and p- type () (E F,p,b ) used in this work. The measured formal potentials (E ) vs SCE of each oxidant are indicated relative to the band positions, where eq was used to convert potentials in V vs SCE to energies in ev vs E Vac Potential (V vs SCE) Figure shows TIRS data for H () surfaces after min of exposure to CH OH in the presence or absence of the indicated oxidants. The peak positions, shifts, and assignments have been described previously. The ν( H) peak (Figure a) broadened and reduced in intensity with increasing formal potential of the oxidant. For the reactions using (CpCOCH )CpFe +, Cp Fe +, and Me Cp Fe + as oxidants, the ν( H) peak was significantly less intense and broader than for reactions performed using weaker oxidants, indicating increased loss of H on the surface. The increased activity of these three oxidants was additionally

15 The Journal of Physical Chemistry Page of 0 supported by large relative areas observed under the C H stretching, ν(c H) CH, peaks (Figure b), as well as the large areas observed under the C H symmetric bending, δ s (C H) CH, peak in addition to the peak that is composed of an O C stretching mode coupled with a CH rocking mode, ν(o C) + ρ(ch ) (Figure c). A broad and low-intensity shoulder on the low energy side of the ν(o C) + ρ(ch ) peak was occasionally observed, and corresponds to transverse optical O stretching, ν( O ) TO. Absorbance x - a Neat Cp Co + MV + Cp* Fe + Me Cp Fe + Me Cp Fe + Cp Fe + (CpCOCH )CpFe Wavenumber (cm - ) Figure. TIRS data for intrinsic H () surfaces after treatment in neat CH OH or CH OH containing.0 mm of oxidant in the dark for min. The H stretching peak (a), the C H stretching peaks (b), and the C H bending and O C stretching coupled with the CH rocking peaks (c) are shown with the oxidant indicated above each spectrum. The spectra are offset vertically for clarity. The symbols ν, δ, and ρ denote stretching, bending, and rocking motions, respectively. n(-h) Absorbance x Neat Cp Co + MV + Cp* Fe + Me Cp Fe + Me Cp Fe + Cp Fe + n a (C-H) CH n a (C-H) CH (CpCOCH )CpFe + n s (C-H) CH Wavenumber (cm - ) The transition from the broad, low-intensity ν( H) peak observed for reactions performed with Me Cp Fe + to the sharp peak observed for reactions performed with Me Cp Fe + suggests a shift in the behavior of the reaction. The ν( H) peak generally sharpened gradually b Absorbance x Neat Cp Co + MV + Cp* Fe + Me Cp Fe + Me Cp Fe + Cp Fe + d s (C-H) CH (CpCOCH )CpFe + n(o-c) + r(ch ) n(-o-) TO c Wavenumber (cm - )

16 Page of The Journal of Physical Chemistry 0 as E (A/A ) decreased beyond E (Me Cp Fe +/0 ). The overall peak shape for reactions performed with Cp Co + was comparable to reactions performed in the absence of oxidant. The intensity and width of the modes associated with the grafting of OCH groups to the H () surface presented in Figure b and Figure c were reduced for Me Cp Fe + and Cp* Fe + compared with the peak areas observed for more oxidizing species. These spectral features were further decreased in intensity for reactions performed in solutions that contained MV +, and reactions performed in neat CH OH or in CH OH that contained Cp Co + yielded nearly undetectable ν(c H) CH, δ s (C H) CH, or ν(o C) + ρ(ch ) signals. These results indicate that the thermodynamic formal potential of oxidants with E (A/A ) E (Me Cp Fe +/0 ) is not sufficient to efficiently promote oxidant-activated addition of CH OH to H (). Figure a presents values of θ OCH determined from TIRS measurements as a function of the oxidizing conditions used for the reaction of H () with CH OH for intrinsic, n-type, and p-type H () samples. The data presented in Figure a show a relatively constant θ OCH for (CpCOCH )CpFe +, Cp Fe +, and Me Cp Fe + as the oxidant. A substantial decrease in θ OCH was observed for reactions performed in CH OH solutions that contained oxidants having E (A/A ) E (Me Cp Fe +/0 ). This transition, which occurred for E (A/A ) between 0. V and 0.0 V vs SCE, is marked in Figure by a vertical black dotted line. The general trend indicated that (CpCOCH )CpFe +, Cp Fe +, and Me Cp Fe + were capable of serving as electron acceptors in the reaction of H () with CH OH, while Me Cp Fe +, Cp* Fe +, MV +, and Cp Co + were not effective oxidants in driving this reaction.

17 The Journal of Physical Chemistry Page of 0! -OCH (ML)! -OCH (ML) a b +0. (CpCOCH )CpFe + E (A/A ) (V vs SCE) +0. Cp Fe Me Cp Fe Me Cp Fe + intrinsic () n-type () p-type () Figure. Correlation between θ OCH and the oxidizing conditions used in the reaction of H () surfaces with CH OH in the dark. Reactions were performed for min in neat CH OH or CH OH containing.0 mm of the oxidant species indicated. The upper x-axis of panel a gives the oxidant used for each reaction in addition to the formal potential for each oxidant. Panel a gives θ OCH determined by TIRS measurements using eqs and, and panel b gives θ OCH determined by XPS measurements using eq. The orange and green dotted lines are averages of θ OCH for all samples left and right of the black dotted line, respectively. Error bars represent statistical variation across multiple samples, and data points with no error bars represent single measurements. The error for single measurements in panel a can be estimated as ±0. ML based on the average standard deviation across all samples. -0. Cp* Fe + Oxidant -0.0 MV Cp Co + N/A Neat CH OH intrinsic () n-type () p-type ()

18 Page of The Journal of Physical Chemistry 0 Figure b presents values of θ OCH determined from XPS measurements as a function of the oxidizing conditions used. For all sample types, the values of θ OCH determined by XPS were substantially larger than those determined using TIRS measurements. For reference, the C s core level XP spectra for representative H () samples after min exposure to CH OH solutions is shown in Figure S (Supporting Information). Adventitious hydrocarbon species, the majority of which are likely unbound CH OH or THF, contributed to the XPS signal ascribable to C bound to O, artifically increasing the value of θ OCH determined by XPS. The the results in Figure b nevertheless show a large and relatively stable θ OCH for reactions performed in CH OH solutions that contained (CpCOCH )CpFe +, Cp Fe +, or Me Cp Fe + as oxidants, with a marked decrease in θ OCH for reactions performed in CH OH solutions that contained Me Cp Fe +. Compared with results for TIRS measurements, for oxidants with E (A/A ) 0.0 V vs SCE, the overall trend showed that θ OCH decreased more gradually as the formal potential of the oxidant in solution became less oxidizing. Figure S (Supporting Information) presents the area under the ν(c H) CH peaks (0 00 cm ) as a function of the oxidizing conditions used. These data parallel the XPS data presented in Figure b, with the largest peak areas observed for reactions performed in CH OH solutions that contained (CpCOCH )CpFe +, Cp Fe +, and Me Cp Fe + as oxidants. Oxidants with E (A/A ) 0.0 V vs SCE yielded a decreased ν(c H) CH peak area compared with the three most oxidizing species. The ν(c H) CH peak area, like the C s photoemission signal, is sensitive to adventitious hydrocarbons, making absolute comparisons difficult. Nonetheless, the data indicate a trend showing a gradual decrease in the ν(c H) CH peak area as E (A/A ) decreased, in agreement with the data presented in Figure.

19 The Journal of Physical Chemistry Page of 0 Figure S (Supporting Information) presents the area under the δ s (C H) CH and complex ν(c O) + ρ(ch ) peaks (0 0 cm ) as a function of the oxidizing conditions used. The results show a clear difference between oxidants having E (A/A ) 0. V vs SCE and oxidants with E (A/A ) 0.0 V vs SCE. Specifically, a substantial decrease in peak area was apparent between reactions performed in CH OH solutions with Me Cp Fe + relative to reactions in CH OH solutions with Me Cp Fe +. The data in Figure S closely parallel the observed trend in Figure a. The TIRS peak area analysis data presented in Figure a and Figure S indicate that the oxidant-activated reaction of H () surfaces with CH OH is not facilitated effectively by oxidants having E (A/A ) E (Me Cp Fe +/0 ). Additionally, the data indicate that, in the dark, the reaction of H () samples with CH OH under the specified conditions is independent of the dopant type and dopant density of the () substrate. III.B. Reaction of H () with CH OH Solutions in the Presence of Light. Figure shows θ OCH determined by TIRS (eqs and ) as a function of the oxidizing conditions used for the reaction of H () with CH OH in the dark or under ambient light. For intrinsic or n- type H () reacted in ambient light with CH OH that contained Me Cp Fe +, Cp Fe + or Me Cp Fe +, θ OCH increased relative to the same reaction in the dark. The values for θ OCH increased for these three oxidants, but the values for Me Cp Fe + were nevertheless smaller than that for Cp Fe + or Me Cp Fe +. For solutions that contained Cp* Fe +, MV +, or Cp Co +, intrinsic or n-type H () samples exhibited θ OCH that was close to the value observed in the dark. The results for p-type H () surfaces in the light were not statistically different from the results observed in the dark.

20 Page of The Journal of Physical Chemistry 0! q -OCH (ML)! -OCH (ML)! -OCH (ML) a b c +0. CpFe + CpFe + CpFe + E (A/A ) (V vs SCE) +0. MeCpFe + MeCpFe + MeCpFe MeCpFe + MeCpFe + MeCpFe Cp*Fe + MV + Oxidant CpCo + Neat CHOH intrinsic () dark intrinsic () light Cp*Fe MV + Oxidant -.0 CpCo + N/A Neat CHOH n-type () dark n-type () light Cp*Fe + MV + Oxidant CpCo + Neat CHOH p-type () dark p-type () light Figure. Correlation between θ OCH and the oxidizing conditions used in the reaction of (a) intrinsic, (b) n-type, and (c) p-type H () surfaces with CH OH in the absence (striped) and presence (solid) of ambient light. Reactions were performed for min in neat CH OH or CH OH containing.0 mm of the oxidant specified at the top of each plot. Experimentally determined formal potentials, E (A/A ), for each oxidant are given above panel a. The values for θ OCH

21 The Journal of Physical Chemistry Page of 0 were determined by TIRS measurements using eqs and. Error bars represent statistical variation across multiple samples, and data points with no error bars represent single measurements. The error for single measurements in panels b and c can be estimated as ±0. ML based on the average standard deviation across all samples. The intrinsic samples, so called because of their low dopant density, were very lightly n- doped (Figure ), implying that the valence-band states are fully occupied in the absence of illumination. The reactivity observed herein indicates that samples having Fermi levels situated positive of (above) the middle of the band gap, that is, intrinsic or n-type samples, exhibited an increased rate of methoxylation of H () surfaces in the presence of illumination relative to the rate of methoxylation in the dark. III.C. Potentiostatic Reaction of H () with CH OH. The reaction of H () with CH OH was investigated under applied external bias in the dark as well as under illumination. Figure presents J-E data for n-type and p-type H () surfaces in contact with CH OH solutions that contained.0 M LiClO as the supporting electrolyte. In the dark, the methoxylation of n-type H () was observed as a gradual increase in current starting near 0. V vs SCE, with a peak observed at 0.0 V vs SCE. A gradual decline in current was observed past the peak current, suggesting that the initial rise in current arose from methoxylation of H () surfaces. Under illumination, n-type H () surfaces showed much higher current density, in addition to a sharp peak centered at 0. V vs SCE that is consistently assigned to the methoxylation of H () surfaces. At more positive potentials, a rapid rise in current density was observed beginning at 0 V vs SCE, and was indicative of subsurface oxidation of the () substrate.

22 Page of The Journal of Physical Chemistry 0 Current Density (μa (µa cm - ) Current Density (µa (μa cm - - ) ) a E (MV +/+ ) b Figure. J-E behavior of (a) n-type and (b) p-type H () samples in contact with CH OH solutions containing.0 M LiClO as supporting electrolyte. Data collected in the dark (solid black) and under mw cm of simulated solar illumination (dashed orange) are shown. The formal potentials of the oxidants used in this work are indicated by the vertical blue lines at the top of panel b. E (Cp*Fe +/0 ) E (Me Cp Fe +/0 ) E (MeCpFe +/0 ) E (Cp Fe +/0 ) n-()-h Dark n-()-h Light Potential (V vs SCE) p-()-h Dark p-()-h Light For p-type H (), a gradual increase in current density that is assigned to methoxylation was observed near 0. V vs SCE. A distinct peak current due to the methoxylation reaction was not observed, as the onset of the subsurface oxidation of the H () substrate was observed near +0. V vs SCE. The results indicate that the methoxylation

23 The Journal of Physical Chemistry Page of 0 of p-type H () samples was not sensitive to illumination, as the dark and light curves in Figure b overlapped substantially. A second sweep of the n-type and p-type electrodes in Figure (not shown) was flat through the region ascribed to potentiostatic methoxylation. Table summarizes the quantification of the current passed for the region assigned to the methoxylation of H () surfaces. The quantification of θ OCH assumed that electrons were passed for the reaction of each surface site with CH OH. The θ OCH resulting from potentiostatic methoxylation of n-type H () surfaces in the dark was comparable to θ OCH obtained from potentiostatic methoxylation of p-type H () surfaces in the dark as well as under illumination. In contrast, illuminated n-type H () surfaces showed significantly greater anodic current that resulted in higher θ OCH compared with samples that were methoxylated in the dark. The potential at which half of the current ascribed to the methoxylation reaction had been passed (E / ) is specified in Table. For multiple samples, E / was the same within error for n- type samples in the dark and p-type samples in the dark or under illumination. For n-type samples under illumination, E / was shifted by 0. ± 0.0 V compared with n-type samples in the dark, indicating the presence of a photovoltage at the n- interface that produced higher anodic current densities and consequently shifted E / to more negative potentials.

24 Page of The Journal of Physical Chemistry 0 Table. Summary of the Quantification for the Potentiostatic Methoxylation of H () Surfaces. Dopant Illumination Potential Range θ OCH (ML) b Potential at Half Type Quantified (V) a θ OCH (V) c n-type Dark 0. to ± ± 0.0 n-type mw cm 0. to 0 0. ± ± 0.0 p-type Dark 0. to ± ± 0.0 p-type mw cm 0. to ± ± 0.0 a Potential vs SCE. b Quantified based on anodic current passed assuming electrons per surface site that reacts with CH OH. The density of surface sites on the () surface is. cm, requiring. C cm to methoxylate all surface sites. c Potential vs SCE at which half the charge attributed to the methoxylation of the surface was passed (E / ). The values of θ OCH and E / were quantified without use of a baseline to subtract current resulting from subsurface O x formation. To account for this, the regions quantified were selected to include a fraction of current from O x formation and excluded a roughly equivalent amount of current from the methoxylation reaction where applicable. IV. DISCUSSION IV.A. Conditions that Lead to Methoxylation of the H () Surface. The data reported in this work support previous reports that collectively indicate the presence of two reaction pathways, with mutually different relative rates of electron transfer, for the methoxylation of H (). Both reaction pathways involve a net loss of electrons per methoxylated H () surface site, with the electrons producing H in Scheme or reducing

25 The Journal of Physical Chemistry Page of 0 two equivalents of oxidant in Scheme. When a molecular oxidant with a sufficiently positive E (A/A ) is present in solution, as indicated in Scheme, the surface is more readily oxidized and reacts rapidly with methanol. This section of the discussion develops the kinetic model for charge transfer at the H () interface that accounts for these two charge-transfer reaction pathways. Figure depicts representative band structures for n-type and p-type surfaces in the dark and under illumination. We postulate that the methoxylation of H () surfaces can be broken into two half-reactions. The first half-reaction is the oxidation of the surface by the bulk. The oxidation energy for the H () surface (E( +/0 )) is estimated using the value of E / determined by potentiostatic methoxylation (Table ). As E / measured in the dark falls near 0 V vs SCE, E( +/0 ) is estimated using eq to lie within the band gap, at ~. ev. The electrons that comprise this electronic state are likely associated with a surface resonance that has been identified on the H () surface. This surface resonance has been estimated to lie ~0. ev below E vb in vacuum, - but hydrogen bonding between CH OH and the H () surface is predicted to increase the electron density at the surface, shifting the electrons in the surface resonance more positive in energy. Oxidation of the surface resonance completes the first half of the proposed pathway, leading to methoxylation of the H () surface.

26 Page of The Journal of Physical Chemistry 0 n-type p-type Energy (ev) (+) ( ) Energy (ev) (+) ( ) E cb E F,n E F,p a k E vb,a vb k vb,c E cb E F,n E F,p c Dark E vb k cb,c k cb,a A A E( +/0 ) k cb,c k cb,a k vb,a k vb,c Figure. Schematic representation of charge transfer across a semiconductor/liquid interface for oxidant-activated methoxylation. The energy positions for the semiconductor valence band (E vb ), conduction band (E cb ), electron quasi-fermi level (E F,n ), hole quasi-fermi level (E F,p ), the oxidation energy of the surface (E( +/0 )), and solution energy (E(A/A )) are indicated. The rate constants for cathodic and anodic charge transfer to and from the valence band are k vb,c and k vb,a, respectively, and the rate constants for cathodic and anodic charge transfer from and to the conduction band are k cb,c and k cb,a, respectively. Solid and dashed arrows indicate cathodic and anodic charge transfer, respectively. Panels a and c give the band structure for n-type and p-type, contacts, respectively, in the dark, and panels b and d show the same contacts in a and c, respectively, under illumination. E F,n E(A/A ) E(A/A ) A A A A b E F,n E(A/A ) E(A/A ) E( +/0 ) A A Light d E cb E F,p k E vb vb,a k vb,c E cb k cb,c k cb,a k cb,c k cb,a E F,p k vb,a E vb k vb,c A A E( +/0 ) A A A A E( +/0 ) A A

27 The Journal of Physical Chemistry Page of 0 Conditions that lead to oxidation of the surface can be described using a charge-transfer kinetics model. The theoretically expected ratios of the cathodic to anodic current density, which describe the direction and magnitude of current flow between surface resonance (E( +/0 )) and the valence band or conduction band (R vb,sr or R cb,sr, respectively) are given in eqs and : R vb,sr = e R cb,sr = e ( ) k B T E F,p E +/0 ( ) k B T E F,n E +/0 where E F,p and E F,n are the hole and electron quasi-fermi levels, respectively. The derivation of eqs and is given in the Supporting Information. When E F,p < E( +/0 ) and E F,n < E( +/0 ), R vb,sr < and R cb,sr <, respectively, and oxidation of the surface resonance dominates the charge-transfer process. Conversely, when E F,p > E( +/0 ) and E F,n > E( +/0 ), R vb,sr > and R cb,sr >, respectively, and according to eqs and, the surface is not readily oxidized by the bulk. The second half of the methoxylation reaction is the reduction of a species in solution by electrons in the bulk, maintaining charge neutrality in the. When a well-defined solution energy (E(A/A )) is produced by a molecular oxidant in solution, the cathodic to anodic current density ratios that describe the direction and magnitude of current flow between the valence band or conduction band (R vb,sol or R cb,sol, respectively) and a solution with energy E(A/A ) are given by eqs and : R vb,sol = e R cb,sol = e ( ) k B T E F,p E A/A ( ) k B T E F,n E A/A () () () ()

28 Page of The Journal of Physical Chemistry 0 Eqs and are derived in the Supporting Information. For E F,p < E(A/A ) and E F,n < E(A/A ), R vb,sol < and R cb,sol <, respectively, and species in solution are not readily reduced by the. Conversely, when E F,p > E(A/A ) and E F,n > E(A/A ), R vb,sol > and R cb,sol >, respectively, cathodic current flow to the solution readily occurs, and the oxidants in solution are reduced to complete the second half of the methoxylation reaction. In this framework, oxidant-activated methoxylation is obtained when the surface resonance is readily oxidized by the bulk (R vb,sr < and/or R cb,sr < ) and concomitantly when the oxidants in solution are readily reduced by the bulk (R vb,sol > and/or R cb,sol > ). When one or both of these conditions is not met, the methoxylation reaction is postulated to proceed according to Scheme. When no oxidant was present in solution, or when the oxidant in solution was not sufficiently oxidizing, the reaction of H () surfaces with CH OH occurred slowly, and the reaction rate was not responsive to changes in the dopant type or concentration or the presence of illumination. Because changes in the semiconductor bulk Fermi level had no effect on the rate of methoxylation in these cases, the rate of oxidation of the surface resonance by the bulk (eqs and ) appears to have minimal influence on the rate of methoxylation. Instead, for these cases, the rate of methoxylation appears to be limited by the reduction of the H and O H bonds to produce H. The observations are consistent with expectations that the reduction of H and O H bonds to form H is kinetically slower than reduction of one-electron molecular oxidants that provide sufficiently oxidizing solution potentials to allow for facile methoxylation of the H () surface. The methoxylation pathway depicted in Scheme is postulated to occur under all conditions used herein, and, when the oxidizing environment favors oxidant-activated methoxylation (Scheme ), both methoxylation pathways occur simultaneously.

29 The Journal of Physical Chemistry Page of 0 IV.B. Oxidant-Activated Methoxylation in the Dark. In the dark, intrinsic, n-type, and p-type samples exhibited similar methoxylation behavior, with solutions containing (CpCOCH )CpFe +, Cp Fe +, and Me Cp Fe + yielding increased rates of reaction compared with solutions that contained weaker oxidants. For all oxidants except (CpCOCH )CpFe +/0, only the oxidized form was added to the solution, giving a Nernstian solution potential that was shifted positive of the E (A/A ) value given in Figure. Assuming that the reduced species were present as contaminants with an oxidized/reduced ratio of ~0:, the Nernstian solution potentials would then be shifted by ~0. V relative to the formal reduction potential. In the case of (CpCOCH )CpFe +/0, the oxidized/reduced ratio was :, yielding a Nernstian solution potential E((CpCOCH )CpFe +/0 ) = +0. V vs SCE. For oxidants that produce E(A/A ) sufficiently positive of E vb, as indicated for n-type and p-type samples in Figure a and c, respectively, the semiconductor dopant atoms can ionize and shift the quasi-fermi level positions at the surface, yielding E F,p = E F,n = E(A/A ) after equilibrium is reached. Under these conditions, the oxidant-activated pathway for methoxylation is inactive. Although the quasi-fermi levels in this case can independently reach equilibrium with the solution energy, equilibrium between the bulk and the surface resonance is not necessarily achieved. Despite this, when E F,p = E F,n = E(A/A ), there is no propensity for the to reduce oxidants in solution (eqs and ), and electrons are forced down the pathway that produces H (Scheme ). As E(A/A ) shifts more negative, approaching E vb, the dopants become less effective at equilibrating E F,p and E F,n with the solution energy, and E F,p and E F,n begin to fall positive of E(A/A ). With E F,p > E(A/A ) and E F,n > E(A/A ), eqs and indicate that cathodic current flows from the to the solution. For E(A/A ) near E vb, E F,p and E F,n are moved negative of

30 Page of The Journal of Physical Chemistry 0 E( +/0 ), and oxidation of the surface resonance occurs by eqs and. Under this situation, with E(A/A ) near or negative of E vb, the conditions for the oxidant-activated methoxylation of the H () surface are satisfied. This framework therefore accounts for the observation of oxidant-activated methoxylation occurring for solutions containing (CpCOCH )CpFe +, Cp Fe +, and Me Cp Fe +, which effect E(A/A ) near E vb. IV.C. Oxidant-Activated Methoxylation in the Light. Examples of illuminated n-type and p-type interfaces are depicted in Figure b and d, respectively. Illumination of n-type interfaces effectively shifts E F,n positive as the electron concentration at the surface increases. This positive shift in E F,n yields R cb,sol >, and cathodic current from the conduction band can complete the reduction half of the methoxylation reaction. For weak oxidants with E(A/A ) positive of E( +/0 ), E F,p at n-type interfaces will not be capable of oxidizing the surface resonance, and the oxidant-activated methoxylation reaction is inactive even under illumination. When the solution potential is capable of shifting E F,p negative of E( +/0 ), the oxidant-activated mechanism can become active at n-type interfaces under illumination. For example, solutions containing Me Cp Fe + yielded an increased rate of methoxylation at illuminated n-type surfaces, while solutions containing Cp *Fe + showed no obvious sensitivity to light. Illumination of p-type interfaces results in an increase in the hole concentration at the surface, shifting E F,p negative, as shown in Figure d. This decrease in E F,p allows for oxidation of the surface resonance for oxidants that effect E(A/A ) positive of E( +/0 ) by eq. However, illumination of p-type samples has a negligible effect on the position of E F,n relative to E(A/A ), and the rate of reduction of the oxidants in solution by the bulk is unchanged by eq. Thus, the reduction half-reaction required for oxidant-activated methoxylation is not affected by illumination at p-type interfaces.