The Mechanism of Hydrogen Gas Evolution on GaAs Cathodes Elucidated by In Situ Infrared Spectroscopy

2948

J. Phys. Chem. B 1999, 103, 2948-2962

The Mechanism of Hydrogen Gas Evolution on GaAs Cathodes Elucidated by In Situ

Infrared Spectroscopy

B. H. Erne´,* F. Ozanam, and J.-N. Chazalviel

Laboratoire P.M.C. (U.M.R. 7643 du C.N.R.S.), EÄ cole Polytechnique, 91128 Palaiseau, France

ReceiVed: December 16, 1998

The microscopic mechanism of hydrogen gas evolution on GaAs(100) cathodes is elucidated by studying the

reaction intermediate, adsorbed hydrogen, using in situ infrared spectroscopy in real time. The spectra show

that hydrogen adsorbs cathodically at arsenic sites only and replaces As-OH groups present in the anodic

range. Absolute submonolayer coverages by adsorbed hydrogen and variations in surface dipole potential

due to hydrogen adsorption are both determined quantitatively. Molecular hydrogen is formed upon proton

reduction at As-H sites. The microscopic reaction mechanism at GaAs is found to be pH-independent; the

reason is that rate-limiting electron transfer is not to solution species but to As-H surface sites. A simple

mechanistic model accounts for the infrared spectroscopic results as well as for the electrochemistry, explaining

both the approximately linear potential dependence of hydrogen surface coverage and the hysteresis loop in

cyclic voltammograms.

1. Introduction

gas evolution requires two conduction band electrons per H₂

molecule,¹²and some transient photocurrents had been ascribed

to oxidation of surface hydrogen.¹³Electroluminescence mea-

surements by Uosaki et al.¹⁴gave strong indirect evidence for

the existence of adsorbed hydrogen as an intermediate in the

hydrogen evolution reaction, but direct proof had not yet been

reported. By in situ spectroscopic ellipsometry, we demonstrated

that surface hydrogen destabilizes the surface, a small percentage

of the cathodic current leading to the formation of metallic

gallium as soon as the potential is more negative than the Ga⁰/

Ga³⁺redox potential.¹¹The formation of metallic gallium, whose

presence is easily detected by in situ infrared spectroscopy too,

will generally be avoided in this paper by not going more

negative than that potential. Under such conditions, the surface

is thought to be rough and arsenic rich, since gallium is

preferentially dissolved.¹¹In the present paper, which expands

upon a recent brief report,¹⁵we focus on hydrogen adsorption

and hydrogen gas evolution at such surfaces.

Hydrogen adsorbed at the surface of solids plays an important

role in different fields of surface and materials science. It

diffuses into the materials, causing embrittlement and, in the

case of semiconductors, lowering conductivity by orders of

magnitude.^1-4It affects surface electrical properties, for instance

leading to record low surface recombination rates of photocar-

riers at silicon surfaces.⁵At the surface of heterogeneous

catalysts, it is a reaction intermediate in various commercially

important reactions. Here, we examine its role in the cathodic

production of hydrogen gas, a promising fuel for the chemical

storage of energy. In this reaction, two electrons are transferred

from a metal or semiconductor electrode to two aqueously

solvated protons H⁺_(aq), yielding a hydrogen gas molecule H_2(g)

;

a hydrogen atom is temporarily adsorbed after transfer of the

first electron. The fraction of the surface covered by adsorbed

hydrogen is determined kinetically, by the relative rates of

adsorption and desorption. With n-type gallium arsenide, we

present the first quantitative and dynamic study of hydrogen

adsorption at a solid-liquid interface by in situ infrared

spectroscopy in real time.

The in situ spectroscopic detection of hydrogen adsorbed at

electrode surfaces is based on the resonant vibrational absorption

of hydrogen-to-surface bonds. Despite considerable experimental

difficulties, a few authors reported infrared absorption by

hydrogen adsorbed on platinum cathodes, although some

controversy exists concerning peak assignments.^6-10Surface

hydrogen is more conveniently detected at attenuated-total-

internal-reflection prisms of semiconducting electrodes such as

GaAs,¹¹and this is done here. Signals are weak even in this

geometry, infrared absorption of the solution being stronger than

that of a submonolayer amount of surface hydrogen by more

than 3 orders of magnitude.

The cathodic hydrogen gas evolution reaction has been used

since the beginning of the 20th century as a model system for

the study of electrode kinetics. This is due in part to the

simplicity of the reaction: with metal electrodes, even indirect

information such as the slope of current-potential curves can

be interpreted as an indication for a precise mechanism.¹⁶This

and many other multiple-step reactions in electrochemistry are

assumed to proceed via surface intermediates, but rarely have

such intermediates been observed by direct methods. Here, with

n-GaAs electrodes, we detect cathodic surface hydrogen spec-

troscopically, and we determine its concentration quantitatively,

the first time this is done for any electrode material using in

situ infrared spectroscopy. The effect of hydrogen coverage on

interfacial potential distribution is a key element for the

understanding of the current-potential characteristics and will

be investigated using an original transient flatband potential

measurement technique. After some experimental details, the

results will be presented. They include time-resolved potential-

dependent measurements of surface coverage by hydrogen and

surface free-electron concentration. Then, a highly detailed

Until now, little was known concerning hydrogen adsorption

on GaAs electrodes. It had been verified that cathodic hydrogen

* Corresponding author.

Published on Web 03/27/1999

Hydrogen Gas Evolution on GaAs Cathodes

J. Phys. Chem. B, Vol. 103, No. 15, 1999 2949

picture is drawn of the microscopic reaction mechanism of

hydrogen gas evolution on GaAs cathodes. A reaction model

is developed, and it is tested by comparing simulated electro-

chemical behavior with experiment.

2. Experimental Section

The experiments were performed on 100 -oriented n-GaAs

single crystals grown by the liquid encapsulated Czochralski

method and purchased from MCP Wafer Technology, U.K. (N_D

) 8 × 10¹⁵cm^-3) and from the Institute of Electronic Materials

Technology, Poland (N_D) 4 × 10¹⁵and 7 × 10¹⁶cm^-3). Two-

side polished 500 µm thick wafers were used to make multiple-

total-internal-reflection prisms. Ohmic contacts were obtained

by evaporating gold (+10% germanium) under reduced pressure

(10^-3Pa) while the sample was being cooled from 400 °C down

to room temperature. The surface was pretreated with concen-

trated HCl in order to remove oxides and according to a recipe

by Aspnes and Studna¹⁷for obtaining ellipsometrically clean

GaAs surfaces. On the basis of our previous results, the GaAs

surface is probably OH-covered after such a treatment.¹⁵

The electrochemical cell is described in ref 18, with the

vertically oriented prismatic GaAs electrode (45° angles)

exposing an area of 0.8 cm²to a circulating aqueous solution

of HCl or NaOH. The cell includes a platinum counter electrode

and an Ag/AgCl in saturated KCl reference electrode. The

circulation rate was sufficiently high that no hydrogen bubbles

were detected at the surface during the experiments reported

here (when a bubble is present, this causes a strong decrease in

infrared absorbance by the electrolyte solution). When needed,

illumination was achieved using a white light source. The

experiments were carried out at room temperature on a short

time scale, so that diffusion of hydrogen into GaAs can be

neglected.

Figure 1. (a) Total infrared absorbance of the n-GaAs/6 M HCl_(aq)

interface (reference: the n-GaAs/N₂interface); (b) change in infrared

absorbance when the applied potential is changed from -0.3 to -0.8

V vs Ag/AgCl (average of 1500 cycles; for each cycle, potential was

set to -0.8 V during 4 s and to -0.3 V during 8 s; spectra recorded

under white light illumination with an intensity corresponding to an

anodic photocurrent density of 100 µA cm^-2).

3. Results

The results are presented in four parts. First, the adsorption

site of hydrogen at the surface of GaAs cathodes is characterized.

Second, a coulometric titration of surface hydrogen is performed

in order to determine the surface concentration of adsorbed

hydrogen quantitatively. Third, the transient capacitance mea-

surement technique is introduced. And fourth, surface hydrogen

coverage and interfacial capacitance are measured in real time

during cyclic potential scans.

3.1. Spectroscopic Characterization of Surface Hydrogen.

Figure 1a shows the total infrared absorbance of the n-GaAs/

HCl_(aq)interface. The spectrum is completely dominated by the

absorbance of the solution probed by the evanescent wave

(probing depth ≈ 0.5 µm at 1000 cm^-1). For instance, O-H

stretch vibrations are seen in the 3000-3500 cm^-1range and

H-O-H bending modes around 1650 cm^-1. To reveal the

presence of surface species, measurements are performed in the

potential difference mode. In this way, constant contributions

to the absorbance are eliminated.

Figure 1b presents the change in interfacial infrared absor-

bance as a result of going from the potential where no current

is measured in the steady state (-0.3 V vs Ag/AgCl) to a

potential where hydrogen evolution occurs (-0.8 V vs Ag/

AgCl). Absorbance increases at 2050 cm^-1and decreases in

the 2500-3700 cm^-1range. Comparison with the literature

indicates that the absorbance at 2050 cm^-1corresponds to As-H

bonds only; no Ga-H absorption is found, even when the

potential is set to a value where metallic gallium is present at

the surface. In contrast, spectra of hydrogen adsorbed on GaAs-

(100) from the gas phase show absorption both by As-H and

Ga-H.^20,21The nearly symmetric line shape suggests that

monohydride species are the dominant species. The broadness

of the signal may be accounted for in terms of interaction

between As-H and water (see ref 22 for the analogous case of

The aqueous oxygen concentration was kept minimal except

when otherwise indicated. The solution was purged using a flow

of nitrogen from which traces of oxygen were removed by

bubbling in a chromous chloride solution (CrCl₂). The solution

was purged in a separate compartment before it was brought

into contact with GaAs, to limit GaAs surface damage due to

reaction with residual oxygen. During the electrochemical

measurements, a nitrogen flow continued to purge the electrolyte

solution and simultaneously ensured solution circulation. The

counter electrode, at which oxygen is produced during cathodic

polarization of GaAs, was placed in another separate compart-

ment, connected to the rest of the cell via a salt bridge.

Infrared absorbance was measured with a Bomem MB 100

Fourier transform infrared spectrometer set at a nominal

resolution of 4 cm^-1. The absorbance at wave number σ is

defined as (1/N) ln[I^σ(U₀)/I^σ(U)], where N ≈ 10 is the number

of useful reflections at the electrochemical interface, I^σ(U) is

the light intensity at wave number σ reaching the detector at

potential U, and I^σ(U₀) the same but under reference conditions

at potential U₀. Spectra were recorded in the dark, unless

otherwise specified. During dynamic measurements of surface

hydrogen concentration, infrared absorbance was recorded

continuously; the integration time is therefore given by the

distance between two experimental points in a figure.

Changes in flatband potential were monitored by interrupting

the cyclic potential scans or other experiments every 100 ms

for a 1 ms period and stepping the potential to various positive

potentials, at which the differential capacitance of the semi-

conductor space charge layer was measured (for details on the

setup, see ref 19).

2950 J. Phys. Chem. B, Vol. 103, No. 15, 1999

Erne´ et al.

Figure 2. Evolution of As-H coverage at the n-GaAs surface in 1 M

HCl, after stepping potential from -1 to +0.5 V vs Ag/AgCl in the

dark, for two different concentrations of dissolved oxygen.

Si-H) and/or by the existence of distinct As-H sites at the

surface. We ascribe the absorbance decrease in the 2500-3700

cm^-1range to loss of surface OH-groups, present at the surface

in the anodic range and replaced by adsorbed hydrogen in the

cathodic range. The signal might arise either from surface OH-

groups or from H₂O molecules; however, changes in water

absorption by O-H stretch vibrations would have occurred over

a narrower range and would have given rise to a stronger signal

in the 1650 cm^-1range (compare with Figure 1a). Furthermore,

the separate peak at 3600 cm^-1evokes absorption by isolated

OH-groups not hydrogen bonded to other OH-groups, which

leads to the ascription of the broad signal to AsO-H. Several

weak vibrational features are also detected near 1450, 1700,

and 2900 cm^-1; these are artifacts ascribed to imperfect

cancellation of O-ring seal absorption. Absorption by As-O

bonds falls outside our spectral window. Peak assignments were

verified by measurements in deuterated solutions, in which the

Figure 3. Changes in (a) current density and (b) As-H absorbance

of illuminated n-GaAs electrodes in 1 M HCl upon stepping potential

from -0.3 to -0.8 V vs Ag/AgCl and back (white light illumination

with an intensity corresponding to the maximum anodic current density

seen in the figure).

x

signals shift to wave numbers lower by a factor ≈ 2, due to

the higher reduced mass of the oscillators.¹¹Moreover, the

As-H bonds were found to absorb p-polarized light more

strongly than s-polarized light,¹¹confirming that they are at the

surface rather than inside the solid. Absorbance by As-H is

found at the same wave numbers in acidic and alkaline solutions.

3.2. Absolute Surface Concentration of Hydrogen. Quali-

tatively, we already showed in another paper that the surface

concentration of hydrogen adsorbed on GaAs cathodes depends

more or less linearly on the applied potential.¹¹Measurements

of the anodic charge required to convert a cathodic GaAs surface

into an anodic surface suggest that the absolute surface

coverages are of the order of a monolayer.^13,23,24Here, in situ

infrared spectroscopy and anodic charge measurements are

combined in order to determine absolute hydrogen surface

concentrations quantitatively.

Figure 4. Anodic charge required to remove hydrogen from n-GaAs

surfaces in 1 M HCl or 1 M NaOH as a function of the change in

As-H absorbance (different symbols refer to different electrodes).

These coulometric measurements provide an absolute calibration of

the As-H surface concentration (right-hand scale).

their oxidation by photogenerated holes is kinetically even more

favorable than recombination with electrons, hence a transient

photoanodic current is observed at this potential due to oxidation

of surface hydrogen. The anodic charge measured by integrating

the transient photocurrent is proportional to the initial As-H

absorbance (Figure 4). In HCl solutions, polarization was kept

brief (a few seconds) at potentials U < -0.8 V vs Ag/AgCl (U

< -1.6 V in NaOH solutions), to limit the formation of metallic

gallium, which represents a few percent of the cathodic current

at those potentials.¹¹The proportionality is strong evidence that

the anodic charge is used almost exclusively for the oxidation

of surface hydrogen. This is no longer the case when polarization

in the stability range of metallic gallium is carried out for longer

periods, because the anodic charge required to oxidize the

surface then includes a sizable contribution due to the oxidation

of metallic gallium. The same proportionality constant is found

in acidic and alkaline solutions with different GaAs samples

and light intensities. On the assumption, suggested by Figure

1b, that all As-H is oxidized to As-OH, requiring two holes

per As-H, integrated absorbances by surface As-H can be

converted to absolute surface coverage using Figure 4. The

integrated absorbance when all 6.26 × 10¹⁴cm^-2arsenic atoms

in a (100) plane are bound to a hydrogen atom is apparently

1.08 × 10^-2cm^-1per reflection. This value seems quite

reasonable when compared with the absorbance of surface

hydrogen in other systems under conditions where surface

coverage appears to be maximal.^20,21,27-29

A coulometric titration of the surface hydrogen observed by

in situ infrared spectroscopy can be carried out using anodic

current, provided that the dissolved oxygen concentration is

sufficiently low. Indeed, time-resolved measurements reveal that

the disappearance of cathodically adsorbed hydrogen at positive

potentials is due only to anodic reaction with holes (negligible

in the dark, when the current density j was ≈20 nA cm^-2for

N_D≈ 10¹⁶cm^-3) and chemical reaction with dissolved oxidizing

agents (e.g., oxygen, see Figure 2). After measurement of the

As-H absorbance under cathodic conditions, the potential was

changed to the steady-state rest potential under illumination.

Just after the jump, a photocurrent appeared and decreased to

zero as surface hydrogen was removed (Figure 3). It is well-

known that no electrochemical dissolution of n-GaAs occurs at

this potential in indifferent electrolyte solution; a photogenerated

hole is much more likely to recombine with an electron than to

oxidize a GaAs bond under those conditions.^25,26Our measure-

ments show that when As-H bonds are present at the surface,

Hydrogen Gas Evolution on GaAs Cathodes

J. Phys. Chem. B, Vol. 103, No. 15, 1999 2951

Simultaneous infrared and capacitance measurements were

performed only on samples with a dopant density of 7 × 10¹⁶

cm^-3. In situ infrared measurements are more conveniently

carried out on less doped samples, because free electrons absorb

infrared light over the entire spectral range,³³and the transmitted

light intensity decreases exponentially with dopant density.

However, at low dopant densities, capacitance measurements

are seriously affected by deep electronic centers in the band

gap generally present at a concentration of ≈10¹⁶cm^-3, so-

called EL2 centers.^1,30,34,35Hysteresis effects occur, because

whereas these levels gain electrons easily when the Fermi level

is sufficiently high, the reverse process, electron loss, is

extremely slow, occurring on a time scale of seconds at room

temperature. At a dopant density of 7 × 10¹⁶cm^-3, these

undesired effects are negligible and infrared transmittance is

not yet prohibitively low.

Figure 5. Transient capacitance of the n-GaAs/1 M HCl interface in

the dark plotted according to the Mott-Schottky relation (eq 1) for

different stationary potentials U_stat. Every 100 ms, the potential was

switched from U_statto U for a 1 ms period during which the differential

capacitance was measured; U was scanned at 50 mV s^-1(N_D≈ 7 ×

10¹⁶cm^-3). The extrapolated flatband potential U_fbis seen to depend

3.4. Dynamic Potential Dependence of j, θ_H, and U_fb.

Simultaneous dynamic measurements of the current density j,

the hydrogen surface coverage θ_H, and the flatband potential

U_fbwere carried out on n-GaAs electrodes under illumination

and in the dark.

on U_stat

.

3.3. The Transient Capacitance Measurement Technique.

In an earlier paper,¹¹it was concluded that hydrogen adsorption

on GaAs cathodes probably causes a change of the surface

dipole potential. In the present work, to quantify such changes,

in situ infrared measurements are performed simultaneously with

transient measurements of the interfacial capacitance. This

section illustrates how the transient capacitance measurement

technique¹⁹allows the rapid determination (in a millisecond or

less) of the flatband potential corresponding to a short-lived

surface chemical composition.

Under illumination, the hydrogen surface coverage was

monitored during cyclic potential scans in 1 M HCl solution at

two different scan rates. Current density is shown in Figure 6a,

hydrogen surface coverage in Figure 6b, and flatband potential

in Figure 6c. The 0.5 mV s^-1scan approximates steady-state

conditions, while the 50 mV s^-1scan illustrates the system

dynamics.

The n-GaAs electrochemical interface is a rectifying contact,

much like a Schottky diode. Under forward bias, cathodic

current density due to hydrogen gas evolution changes expo-

nentially with the applied potential, while hydrogen surface

coverage changes more or less linearly. Under reverse bias, the

anodic current density is very low in the dark and corresponds

to one hole per absorbed photon under illumination.^25,26In the

steady state, anodic current causes dissolution of the semicon-

ductor,^25,26while in the 50 mV s^-1scan in Figure 6, as long as

As-H is present, it can be shown using the coulometric titration

in Figure 4 that all the anodic current is consumed for the

oxidation of surface hydrogen.

The electrode capacitance under depletion conditions is

determined by the capacitance C_scof the semiconductor space

charge layer. Therefore, its dependence on the applied potential

U obeys the Mott-Schottky relation:³⁰

C_sc^-2) 2 (eꢀꢀ₀N_D)^-1(U - U_fb- k_BT/e)

(1)

where e is the positive elementary charge, ꢀ the dielectric

constant of GaAs,³⁰ꢀ₀the permittivity of vacuum, U_fbthe

flatband potential, k_Bthe Boltzmann constant, and T the absolute

temperature. Changes of the surface dipole potential are revealed

by changes of the flatband potential value obtained by extrapo-

lating a plot of C^2-versus U to C^-2) 0.

One important difference with a Schottky diode is that the

applied potential drops partly across the semiconductor space

charge layer and partly on the electrolyte side of the inter-

face.^25,26In Figure 6c, changes in potential drop on the

electrolyte side of the interface are revealed by changes in the

flatband potential. In the reverse bias range, U_fbis about -0.85

V in the dark and somewhat more positive under illumination,

due to the positive charge of dissolution intermediates.^32,36-38

Under forward bias, U_fbshifts negatively, in agreement with

steady-state measurements by Uhlendorf et al.³⁹Our dynamic

measurements clearly indicate that the position of U_fbis

correlated to the hydrogen surface coverage. U_fband hydrogen

surface coverage exhibit approximately the same hysteresis,

about 200 mV during the 0.5 mV s^-1scan and significantly

more during the 50 mV s^-1scan. During the negative scans,

first the positive surface charge disappears, and then, an

additional negative U_fbshift of about 50 mV occurs per 10¹⁴

cm^-2As-H bonds. During the positive scans, first As-H

disappears, after which positive surface charge reappears, both

causing positive shift of U_fb. This is the first direct evidence in

electrochemistry that a change from OH- to H-surface coverage

causes a change in the surface dipole potential.

Figure 5 presents the Mott-Schottky plots for n-GaAs(100)

electrodes in 1 M HCl having reached a steady-state chemical

surface composition at +1, 0, and -0.8 V vs Ag/AgCl. Every

100 ms, the stationary potential U_statwas left for a period of 1

ms during which the differential capacitance was measured at

potentials scanned in the anodic range. The transient Mott-

Schottky plots are the same for stationary potentials of +1 and

0 V, indicating that the surface dipole potential does not change.

When the Mott-Schottky plot is measured in a more classical

way, i.e., the stationary potential and the potential at which the

capacitance measurements are carried out are the same, points

are obtained which fall on the same curve. In contrast, the

flatband potential corresponding to an electrode polarized at

-0.8 V is significantly more negative, in agreement with rapidly

measured Mott-Schottky plots reported by others.^31,32The

transient capacitance technique makes it possible to monitor

flatband potential changes in real time during electrochemical

surface treatments. For example, the measurement of cyclic

voltammograms can be interrupted every 100 ms in order to

check the flatband potential by a 1 ms measurement of the

capacitance at a potential where the Mott-Schottky plot is

linear.

In the dark, the relationship between hydrogen surface

coverage and surface dipole potential was investigated further.

2952 J. Phys. Chem. B, Vol. 103, No. 15, 1999

Erne´ et al.

Figure 7. Changes in (a) current density, (b) As-H absorbance (surface

concentration according to calibration in Figure 4, reference potential

+0.5 V vs Ag/AgCl), and (c) flatband potential (determined as in Figure

5) of n-GaAs electrodes in the dark in 1 M HCl upon stepping potential

from +0.5 to -0.7 V vs Ag/AgCl and back. Notice the close correlation

between the variations in As-H absorbance and U_fb.

similar at pH 0 and pH 14, although at a same current density,

θ_Hvalues are higher at pH 14 by a factor of about two. At

potentials where hydrogen evolution occurs, θ_Hand U_fbwere

found to be practically the same in the dark and under

illumination.

The scans presented by full lines in Figure 8 were not

measured on freshly polished electrodes but rather after two or

three preliminary scans. This pretreatment ensured that subse-

quent consecutive scans were practically the same, so that

irreversible changes occurring during a single scan could be

neglected. The first scans on a polished electrode give lower

current densities at potentials where hydrogen evolution occurs;

however, the potential dependences of θ_Hand U_fbare practically

the same (see the dotted lines in Figure 8).

The oxygen concentration has a significant effect on the

current density at which As-H first appears in the dark during

potential scans in the negative direction. In an oxygen-saturated

solution, no As-H is observed before j ≈ -100 µA cm^-2, while

As-H first appears at much lower j in purged solutions, at j ≈

-1 µA cm^-2in Figure 8. At potentials positive of those at which

As-H is observed, the current depends on the solution circula-

tion rate, as expected for the diffusion-limited reduction of

aqueous oxidants. An oxygen concentration of 2 µM can be

calculated from a diffusion-limited current density of j ≈ -1

µA cm^-2(4 electrons per O₂molecule, 200 µm diffusion layer

thickness, diffusion coefficient by Kolthoff and Miller⁴⁰), versus

about 1 mM for oxygen-saturated solutions.^41,42Oxygen can

be fully removed from solution by adding an excess concentra-

tion of reductor, e.g., 0.3 M Cr²⁺in 1 M HCl. With that solution,

no cathodic current is observed before -0.36 V, but hydrogen

gas evolution is difficult to study, since Cr³⁺reduction competes

with that of H⁺.

Figure 6. (a) Current density, (b) surface coverage by hydrogen

(determined by infrared measurements, according to the calibration in

Figure 4 and for a reference potential of +0.5 V vs Ag/AgCl under

steady-state conditions), and (c) flatband potential (determined as in

Figure 5) of illuminated n-GaAs electrodes in 1 M HCl monitored

during cyclic potential scans at two different scan rates (same light

intensity as for Figure 3).

First, the effect of hydrogen adsorption on the surface dipole

potential was corroborated by the following experiment. The

surface was hydrogenated in the cathodic range, after which

θ_Hand U_fbwere monitored during chemical dehydrogenation

by residual oxygen in the anodic range (Figure 7). U_fbrequired

approximately the same amount of time to reach a new steady

state as θ_Hdid to drop to zero, demonstrating that the two are

correlated. The process was more rapid when the electrolyte

solution was circulated than when it was not, because circulation

accelerated chemical oxidation of surface hydrogen by increas-

ing the rate at which dissolved oxygen was brought to the

surface. The current density is stationary at +50 nA cm^-2soon

after +0.50 V is applied. The disappearance of As-H is purely

chemical and does not involve electron injection, because the

disappearance of 6 × 10¹⁴cm^-2As-H bonds during the first

100 s would have corresponded to a much higher average current

density of about +1 µA cm^-2

.

In the dark, current density, θ_H, and U_fbwere also monitored

during cyclic potential scans at pH 0 and pH 14 (Figure 8).

The current density and θ_Hcurves at pH 14 are shifted

negatively by about 800 mV with respect to those at pH 0. In

the reverse bias range, the flatband potential is shifted by the

same amount. The current-potential curves and the changes in

θ_Hand U_fbwhich occur during cathodic polarization are very

After sufficient cathodic aging, scans negative of -0.8 V at

pH 0 are significantly affected by the appearance of metallic

gallium,¹¹correlated to a positive shift of the flatband potential

(Figure 9). When enough surface metallic gallium is present,

the current no longer shows hysteresis between potential scans

Hydrogen Gas Evolution on GaAs Cathodes

J. Phys. Chem. B, Vol. 103, No. 15, 1999 2953

[H₂PO₄^-] concentration ratio at pH 0 and a constant [HPO₄^2-]/

[PO₄^3-] ratio at pH 14 during current-potential scans, it can

be concluded that the pH at the surface is always the same as

that in the bulk of the electrolyte solution during our experi-

ments, within less than 0.1 pH unit.

4. Discussion

On the basis of the presented experimental results, a detailed

picture is drawn of the microscopic mechanism of hydrogen

gas evolution on GaAs cathodes. First, the structure and

chemical composition of the surface are discussed, as well as

the different origins of variations of the flatband potential. The

most important electrochemical reaction steps are reviewed, and

the rate-limiting step is identified. Finally, all the acquired

information is combined in order to explain the dynamic

behavior of GaAs electrodes.

4.1. Surface Structure and Chemical Composition. Dif-

ferential in situ infrared spectra (e.g., Figure 1b) indicate that

GaAs(100) electrode surfaces are OH-covered in the anodic

range and that a large fraction of the OH-groups is replaced by

adsorbed hydrogen in the cathodic range. Surface hydrogen is

bound at arsenic sites only. Since replacing AsH by AsOH

involves two electrons and replacing AsH by As^•only one

electron, the equal results of the coulometric titrations at pH 0

and pH 14 (Figure 4) indicate that the anodic surface is OH-

terminated in the entire pH range.

At the most negative potentials, the absolute surface hydrogen

coverage exceeds 6.26 × 10¹⁴cm^-2, the density of arsenic atoms

at an unreconstructed GaAs(100) surface. While this might

suggest the presence of arsenic dihydride, it is just as likely

due to surface (micro)roughness, resulting from cathodic and

anodic decomposition. We demonstrated elsewhere that a minor

percentage of the cathodic current is due to decomposition

reactions which cause surface roughening.¹¹Considering the fact

that three hydrogen atoms per arsenic atom yield arsine gas,

surface roughness is obvious after prolonged cathodic polariza-

tion, when hydrogen surface coverage corresponds to more than

two monolayers.¹¹Surface coverage by hydrogen was not

observed to stabilize at a maximum value at the most negative

potentials due to experimental limitations (metallic gallium

deposition and hydrogen bubble formation). In conclusion, our

data indicate that AsH is the dominant surface species during

hydrogen gas evolution, but it is likely that some residual

coverage by AsOH is maintained even at the most negative

potentials in this study.

Figure 8. (a) Current density, (b) surface coverage by hydrogen

according to the calibration in Figure 4 and for reference potentials of

+0.5 V vs Ag/AgCl at pH 0 and -0.3 V vs Ag/AgCl at pH 14 under

steady-state conditions, and (c) flatband potential (determined as in

Figure 5) of cathodically pretreated n-GaAs electrodes in the dark at

pH 0 (1 M HCl) and pH 14 (1 M NaOH), and monitored during cyclic

potential scans at 1 mV s^-1. The dotted lines indicate measurements at

pH 0 before cathodic pretreatment.

Under vacuum conditions, surface reconstruction occurs at

GaAs(100),²⁰because this lowers the energy by eliminating

dangling bonds. Our results explain why, in contrast, GaAs-

(100) surfaces at the electrolyte interface can be found unre-

constructed, as was shown first by Koinuma and Uosaki⁴⁶using

atomically resolved in situ atomic force microscopy (AFM):

before hydrogen is adsorbed, dangling bonds are already

saturated with adsorbed OH groups. Our results suggest that

the plane seen by AFM consists only of arsenic atoms. For

further illustration, see the didactic picture drawn by Allongue

and Cachet⁴⁷of the first atomic planes of GaAs(100) terminated

by a (1×1) arsenic face.

Figure 9. Flatband potential of n-GaAs electrodes in the dark in 1 M

HCl monitored during cyclic potential scans at 1 mV s^-1to potentials

where metallic gallium is formed (U < -0.8 V vs Ag/AgCl¹¹).

in negative and in positive directions in the stability range of

metallic gallium.¹¹

It was verified by the infrared-active local pH probe

technique⁴³that surface pH remained constant during the

current-potential scans in pH 0 and pH 14 solutions. This

technique uses in situ infrared spectroscopy and relies on the

fact that the evanescent wave penetrates only about 0.5 µm into

the solution;⁴⁴this is a very small distance compared to the

diffusion layer thickness of about 200 µm calculated under our

circulation conditions from a diffusion-limited current density

It is perhaps somewhat surprising that only As-H bonds are

observed instead of Ga-H, considering that surface stability

against (electro)chemical oxidation is known to increase in the

sequence (111)As < (100) < (111)Ga.^25,26,48-50The stability

of Ga(111) against oxidation is presumably related to a lack of

unpaired electrons at this face.^51,52Since roughening processes

promote a surface terminated by the most stable facets, one

of -4 mA cm^-2in 0.01 M H⁺

and from the diffusion

Phosphates were used as the infrared

(aq)

45

coefficient of H⁺

.

(aq)

absorbing pH probes. From an observed constant [H₃PO₄]/

2954 J. Phys. Chem. B, Vol. 103, No. 15, 1999

Erne´ et al.

would therefore expect that the bare microrough anodic surface

is terminated by Ga(111) facets. The situation is different for

cathodic surfaces, however. As-H bonds are much more stable

than Ga-H bonds,^53,54so that it seems plausible that cathodic

polarization promotes arsenic-terminated surfaces. Bonding to

OH-groups could also favor an arsenic-terminated surface:

judging from the fact that all aqueous arsenic species are

oxygenated while Ga³⁺is stable at sufficiently low pH,^53,54

As-O bonds are more stable than Ga-O bonds.

The higher surface coverages by hydrogen in alkaline

solutions suggest higher surface roughness than in acidic

solutions. Relatively low stability of GaAs surfaces in alkaline

solutions was reported before.⁵⁵One pH-dependent change in

surface chemistry could be related to the solubility of oxidized

arsenic. Elemental arsenic is thermodynamically stable in

deaerated solutions at any pH, but residual oxygen oxidizes

arsenic, and surface oxides on elemental arsenic are orders of

magnitude more soluble at pH 14 than at pH 0.⁵⁴This would

explain observations that (photo)anodic treatment yields a porous

arsenic layer in acidic solution,^11,56-59whereas elemental arsenic

is removed in alkaline solution.^56,60-63The observation of an

Figure 10. Flatband potential plotted against surface hydrogen

coverage (same data as Figure 8).

Shifts in flatband potential as a result of pH changes are then

-

due to changes in the surface coverage θ_AsOby the negatively

charged AsO^-. A contribution -eN_sθ_AsOto the surface charge

-

density Q_s, where N_sis the total density of surface sites, affects

the potential drop across the Helmholtz layer by ∆φ_H(and

therefore also U_fb) according to

46

atomically flat (100) face by AFM in 10 mM H₂SO₄

demonstrates that GaAs surfaces in the dark are relatively stable

in oxygen-free solutions. In oxygenated solutions of 1 to 12 M

HCl, the chemical dissolution rate of GaAs is what one would

expect on the basis of oxidation by dissolved oxygen.⁶³

∆φ_H) Q_s/C_H

(5)

where C_His the capacitance of the Helmholtz layer. The effect

of pH on U_fbcan be calculated on the assumption that eq 4

represents a thermodynamic equilibrium and that the activities

of surface species are proportional to their surface coverages

(see also refs 69,70):

4.2. Origins of the Shifts in Flatband Potential. A change

∆φ in the potential applied to a GaAs electrode is distributed

over the semiconductor space charge layer (∆φ_sc), the semi-

conductor surface, and the electrolyte side of the interface.

Surface contributions may include those of surface states,

specifically adsorbed ionic species, and dipolar species attached

to the surface. Following Gerischer et al.⁶⁴we call ∆φ_dipthe

contribution associated with attached neutral dipoles, and we

include in the Helmholtz term ∆φ_Hthe other surface and

electrolyte contributions:

∆φ_H) -eN_sθ_AsO_-/C_H

k_BT

e

k_BT -∆φ_HC_H

) -

ln(10) pH +

ln

-

(

)

e

eN_s

k_BT

e

ln(θ_AsOH) + constant (6)

∆φ ) ∆φ_sc+ ∆φ_dip+ ∆φ_H

(2)

Closer examination of the various terms shows that variations

of ∆φ_Hare dominated by the first term of the right-hand

member: -∆φ_Hvaries by 60 mV [≈ (k_BT/e) ln(10)] per pH

unit at room temperature. Surface coverage by AsO^-adjusts

itself to a small value proportional to the pH.

In this work, the semiconductor space charge layer always

remains depleted, even in the forward bias range, due to changes

in potential drop on the solution side of the interface (∆φ_dip

+

∆φ_H). Those changes are observed as shifts in the extrapolated

flatband potential in capacitance measurements (e.g., see Figure

5), shifts which reveal changes in ∆φ_scat constant potential:

Charged electrochemical reaction intermediates can also cause

a shift in U_fbto an extent given by eq 5. It is well-known that

during photoanodic dissolution of n-GaAs, positively charged

surface intermediates cause a positive shift of U_fb.^32,36-38With

some materials, charged electrochemical reaction intermediates

are even responsible for the pH dependence of U_fb.^71,72Here,

surface coverage by negatively charged intermediates in the

hydrogen evolution reaction is very low, because they are rapidly

neutralized by protons, as will be discussed later. Evidence that

surface negative charge can be neglected is given in Figure 7,

in which the U_fbtransient at open circuit appears to be correlated

only to hydrogen surface coveragesno rapid part is present due

to the disappearance of surface charge. The rate at which j attains

the stationary reverse bias value (Figure 7a) suggests that

cathodic surface charge is at most a small fraction of a

microcoulomb per square centimeter.

∆φ_sc) U - U_fb

(3)

The strongest effect on U_fbobserved here is that of pH. Under

reverse bias, U_fbis about 800 mV more negative at pH 14 than

at pH 0 (Figure 8), in agreement with others who found that

the flatband potential of GaAs electrodes shifts by 60 mV per

pH unit.^65,66The classical explanation is in terms of a pH-

dependent adsorption/desorption equilibrium.^26,67,68It has been

convincingly demonstrated for GaAs that the equilibrium is

chemical, i.e., it does not involve electron exchange: at fixed

pH, the band positions are identical for p- and for n-type

electrodes under severe depletion, when no majority carriers

are present at the surface.^67,68Here, the equilibrium probably

involves the dissociation of surface hydroxyl groups bound to

arsenic atoms of GaAs:

Figure 7 indicates that surface hydrogen can maintain U_fbat

more negative values than when no surface hydrogen is present

even when almost no current is flowing. This can be explained

in terms of the earlier conclusion that As-H dipoles replace

As-OH dipoles, causing a change of the potential drop across

the layer of surface dipoles. Figure 10 gives the flatband

AsOH a AsO^-+ H⁺

(4)

(aq)

Hydrogen Gas Evolution on GaAs Cathodes

J. Phys. Chem. B, Vol. 103, No. 15, 1999 2955

Figure 12. Reaction scheme for changes in surface coverage of GaAs

cathodes. Expressions for the rates of the reaction steps are given in

italics, as well as numbers referring to the corresponding equations in

the text.

Figure 11. Spatial distribution of the electrostatic potential φ(x) at

the n-GaAs/aqueous solution interface. Changes ∆φ in the potential

difference between semiconductor (φ_F) and solution (φ_aq) fall partially

across the semiconductor space-charge layer (∆φ_sc), the adsorbed

dipole layer (∆φ_dip), and the Helmholtz layer (∆φ_H).

while anodic hydroxylation requires two holes

k₃

As-H + h⁺98 As^•+ H⁺

(9)

(aq)

potential as a function of hydrogen surface coverage. The same

data were presented as a function of the applied potential in

Figure 8, in which both U_fband θ_Hshowed considerable

hysteresis. The lack of hysteresis in Figure 10 demonstrates that

one hydrogen surface coverage corresponds to only one potential

drop across the surface dipole layer. No other U_fbshift is

observed than that correlated to hydrogen surface coverage. The

change in dipolar potential drop φ_dipis not completely linear

with the dipole surface coverages, and several partial explana-

tions can be proposed for this. First, interfacial charge in AsO^-

changes with AsOH coverage, especially when θ_AsOHbecomes

small (eq 6). Second, interactions between As-OH groups and

water or neighboring adsorbed species are likely to be different

when As-OH coverage is complete or when it is partial. The

negative sign of the U_fbshift due to the replacement of OH-

by H- groups is consistent with a higher electronegativity of

surface OH-groups compared to H-groups: As^+δ(OH)^-δis

replaced by As^-δ′(H)^+δ′. On the assumption that the capacitance

k₄

9

As^•+ h⁺+ H₂O

8 As-OH + H⁺

(10)

(aq)

The factors k_iare positive reaction rate constants. That the

oxidation of As-H or As^•does not occur by electron injection

into the conduction band is supported by two observations: no

electron injection was measured when a (partially) hydrogen-

covered cathode was polarized in the anodic range in the dark

(Figure 7), and the highest anodic current measured during the

coulometric titrations with photogenerated holes was equal to

that in the photocurrent plateau, for which electron injection is

known to be completely negligible²⁶(compare Figures 3a and

6a). It will be assumed that the hole capture steps are irreversible

too.

The reduction of a solvated proton, yielding an adsorbed

hydrogen atom H_ads, is the first step in the mechanism of

hydrogen gas evolution at a cathode. Gerischer et al.¹²deter-

mined that the formation of a hydrogen molecule requires two

conduction-band electrons on GaAs. For a given electrode

material and given experimental conditions, an important

question is whether the second step proceeds by surface diffusion

to another adsorbed hydrogen atom (11a) or by a second charge-

transfer step at the same site (11b):¹⁶

of the dipole layer is about 10^-4F cm^{-2 16}

, an effective charge

variation [δ + δ′] of about 0.2 e per site can be estimated.

Adsorbed OH and H groups have a similar influence on the

flatband potential of germanium electrodes, discussed exten-

sively by Gerischer, Memming, and others,^64,73although surface

charges may also play a significant role in that system.⁷⁴

The different parts of the GaAs/solution interface over which

changes in applied potential are distributed are summarized in

Figure 11, inspired from a similar figure by Gerischer et al.⁶⁵

for germanium electrodes. The layer with surface dipoles results

in a potential drop ∆φ_dipdue to the orientation of the permanent

dipoles As^+δ(OH)^-δ, thereby affecting the barrier for electron

transfer from GaAs. On the other hand, the negative charge on

dissociated OH groups, compensated by a positive charge in

solution, largely determines the potential drop ∆φ_Hacross the

Helmholtz layer.

4.3. Inventory of Reaction Steps. Adsorption and desorption

compete continuously at every surface site. For an understanding

of hydrogen adsorption at the electrochemical interface, it is

therefore crucial to examine the reaction pathway.

The switch from As-OH to As-H or back requires two

charge carriers, implying the presence of a short-lived As^•radical

intermediate. Cathodic hydrogenation requires two electrons,

H_ads+ H_adsf H_2(g)

(11a)

(11b)

H_ads+ H⁺_(aq)+ e^-f H_2(g)

For example, the hydrogen evolution route includes reaction

step 11a on platinum electrodes and reaction step 11b on gold

electrodes.¹⁶For n-GaAs, the stability of adsorbed hydrogen in

the dark in the anodic range (Figures 2 and 7) is direct evidence

against there being mobile adsorbed hydrogen atoms which

combine to form hydrogen molecules. Therefore, hydrogen

evolution proceeds via reaction 11b. The mobility of adsorbed

hydrogen is negligible, in line with the covalent character of

the As-H bond. Therefore, in the cathodic range, a hydrogen

molecule is produced by two consecutive electron-transfer steps

at the same site, i.e., reaction step 8 followed by

k₅

As-H + e^-+ H⁺_(aq)98 As^•+ H_2(g)

(12)

k₁

As-OH + e^-+ H⁺_(aq)98 As^•+ H₂O

(7)

(8)

An overview of the reaction steps is given in Figure 12. The

same scheme is valid at pH 0 and pH 14, as will be discussed

in the next section.

k₂

As^•+ e^-+ H⁺_(aq)98 As-H

2956 J. Phys. Chem. B, Vol. 103, No. 15, 1999

Erne´ et al.

Figure 14. Same as Figure 13, but at pH 14.

Figure 13. (a) Current density and (b) hydrogen surface coverage

plotted against the semiconductor bandbending at pH 0 (same data as

Figure 8).

Since it was already concluded that the hydrogenation of

As-H is rate limiting, it can now be concluded that the rate-

limiting step is electron transfer to As-H sites. This seems very

plausible, since electron transfer is expected to be much faster

to very reactive As^•surface radical sites than to much more

stable As-H sites, and the electric field in the Helmholtz layer

is very favorable for the arrival of protons to neutralize

transferred negative charge (Figure 11). An interesting parallel

can be drawn with the anodic dissolution of semiconductors:

proposed mechanisms also often involve reaction steps in which

rate determining charge transfer is alternated with rapid chemical

neutralization of the transferred charge.^26,75

4.4. Identification of the Rate-Limiting Step. Cathodic

hydrogen gas evolution on GaAs occurs according to reactions

8 and 12. Considering the expected high reactivity of surface

radical As^•species, coverage by As^•is probably very small.

This conclusion is supported by two arguments on the basis of

our experimental results. First, the direct correlation of U_fbwith

θ_H(Figure 10) suggests that the surface composition has only

one degree of freedom. Since replacing AsH by As^•or AsOH

should have two significantly different effects on the surface

dipole potential, one concludes that adsorbed H is always

replaced by the same species, adsorbed OH. As a result, most

surface sites must be either OH- or H-terminated, and coverage

by As^•is very low. A second argument is the observed high

coverage by hydrogen at the most negative potentials. It is

concluded that k₅, k₂, meaning that the hydrogenation of

As-H is rate limiting for hydrogen evolution (see Figure 12).

This conclusion is consistent with the mechanistic view of a

simplified two step mechanism: As f As-H f As + H₂. Near

unity surface coverage by As-H stands as a clear-cut proof

that the second step is rate limiting.

Most authors make the distinction between H⁺reduction

(aq)

at pH 0 and H₂O reduction at pH 14.^12,16This is because the

overall reaction is

2H⁺_(aq)+ 2e^-f H₂v

in acidic solutions and

2H₂O + 2e^-f H₂v + 2OH^-

(14)

(15)

(aq)

in alkaline solutions. However, since electron transfer is rate-

limiting here, the source of protons is of no importance for the

The hydrogenation of adsorbed hydrogen involves the transfer

of an electron and that of a proton. All experimental evidences

indicate that the two transfers are not simultaneous. Electron

transfer depends on the concentration n_sof electrons at the

surface and thus on (U - U_fb) ) ∆φ_sc:³⁰

reaction rate. It makes no difference to assume that H⁺is

(aq)

reduced over the entire pH range and that at pH 14, H⁺

(aq)

originates from the water dissociation equilibrium,

H₂O a H⁺_(aq)+ OH^-

(16)

_fb)/k_BT

n_s) N_De^-e(U-U

(13)

(aq)

In contrast, ion transfer depends on the potential drop on the

solution side of the interface, affected by variations of U_fb

directly () ∆φ_dip+ ∆φ_H), not (U - U_fb). Figures 13 and 14

replot the data of Figure 8 as j and θ_Hversus (U - U_fb).

Comparison shows that when the surface electron concentration

is the same, the current density is practically the same at pH 0

and at pH 14. Moreover, in both figures, j shows much less

hysteresis than in the j-U curves in Figure 8a. This indicates

that current density depends essentially on the electron con-

centration at the surface. If proton transfer over the Helmholtz

layer had been rate limiting, then, under conditions of constant

pH (fulfilled in Figures 8, 13, and 14), current density during

the forward and the reverse scan would not have been the same

for a same value of n_s, but rather for a same value of θ_H.

It is often assumed that rates of electrochemical charge-

transfer steps are proportional to the surface concentrations of

reduced or oxidized surface sites and to the surface concentration

of the electrons or holes. On this basis, taking reaction 12 as

rate limiting, the total current density could be expressed as

-2k₅N_sθ_Hn_se, where N_sis the total density of surface sites. If

this were true, the hysteresis in j vs (U - U_fb) between forward

and reverse scans (Figures 13a and 14a) would be much larger,

because θ_Hvs (U - U_fb) shows significant hysteresis. The very

small hysteresis indicates that the majority of the current is due

to charge transfer at a minority of the As-H sites. This is in

agreement with the observation that, at a same potential, current

density increases considerably during the first current-potential

cycles while θ_Hand U_fbremain approximately unchanged. For

Hydrogen Gas Evolution on GaAs Cathodes

J. Phys. Chem. B, Vol. 103, No. 15, 1999 2957

a same surface electron concentration, surface roughening

increases current density by increasing the number of active

sites through which most of the current passes.

In most of the present study, metallic surface gallium resulting

from cathodic decomposition of GaAs was avoided by main-

taining the applied potential outside the thermodynamic stability

range of metallic gallium. For metallized semiconductor elec-

trodes, current densities are often limited by thermionic emis-

sion, i.e., all electrons which reach the semiconductor surface

are transferred.^30,39,76The current density is then given by:

reversible one-electron redox systems, mediation by surface

species can greatly enhance rates of electron transfer at

semiconductor electrodes, as was argued by Vanmaekel-

bergh.^79,80

4.5. Dynamic Potential Dependence of θ_Hand j. It is now

possible to examine whether the dynamic potential dependence

of hydrogen surface coverage and current density can be

explained on the basis of the microscopic mechanism described

above. The origin of the hystereses in θ_Hand j during cyclic

potential scans and the approximately linear dependence of θ_H

on the applied potential are of special interest.

j ≈ A*T²exp(-e∆φ_sc/k_BT)

(17)

It was found that the hydrogen gas evolution current density

is directly proportional to the concentration n_sof free electrons

at the surface. Furthermore, at constant pH and given potential,

changes in n_sare essentially due to changes in hydrogen surface

coverage, through ∆φ_dipand U_fb(see eqs 2, 3, and 13).

Hysteresis in j during cyclic potential scans is thus directly due

to hysteresis in θ_H. An explanation of the hysteresis in θ_H

therefore simultaneously accounts for the hysteresis in j.

where A* is ≈ 8 A cm^-2K^-2for GaAs.^30,77In Figures 13 and

14, no metallic gallium is present, and current density during

hydrogen gas evolution is lower than the thermionic emission

limit, but not to a very large extent. The much lower current

density obtained for a polished electrode not yet cathodically

pretreated supports the conclusion that a majority of the current

flows through a minority of very reactive surface sites at the

microrough surface.

θ_Hresults from a balance between reaction steps which adsorb

and steps which desorb hydrogen. It is often assumed that

hydrogen surface coverages of cathodes are determined by the

relative rate constants of the two steps of the hydrogen gas

evolution reaction.¹⁶However, with GaAs, the θ_Hdependence

on potential is clearly not merely due to a balance between steps

8 and 12, since the ratio of the rate constants of those steps

(k₂/k₅) does not depend on potentialsno activation occurs across

the Helmholtz layer during rate-limiting electron transfer to As^•

and AsH surface sites. Steps creating or removing adsorbed OH-

groups are crucial in determining θ_H, and a general origin for

hysteresis is the fact that an increase in θ_Hrequires different

reactants than a decrease in θ_H: an increase in θ_Hrequires

electrons from the conduction band, whereas a decrease in θ_H

requires oxidants to convert As^•into AsOH, else As^•is

immediately rehydrogenated (Figure 12).

At potentials where oxygen is reduced at a rate below the

oxygen diffusion limit, the negative current density significantly

exceeds the calculated thermionic emission limit. This is

surprising, since in the thermionic emission limit, all electrons

which reach the surface are transferred and a higher current

density would seem impossible. Due to its negative sign, the

current must nevertheless be an electron current. It could at first

be surmised that holes are injected and that they recombine with

electrons within the semiconductor at a short distance from the

surface, where the electron density in the conduction band is

much higher. However, measurements on p-GaAs electrodes

under reverse bias indicate the absence of hole injection (|j| <

0.05 µA cm^-2) under conditions where diffusion-limited O₂

reduction attains -100 µA cm^-2on n-type GaAs. An alternative

explanation is that the barrier height for electron transfer is not

homogeneous over the entire surface. Since the electron transfer

rate increases exponentially with decreasing barrier height, a

majority of the current could pass through a minority of the

sites, while the measured e(U - U_fb) represents an average

barrier height determined essentially by the majority of the sites.

The flatband potential of GaAs electrodes is known to depend

on the exposed crystallographic face,^24,25so that the barrier

height could be different for morphologically different sites at

the microrough surface. Moreover, either one or two OH- or

H-groups can be attached to a surface arsenic atom, depending

on the site, which should have a marked influence on the local

barrier height. It should however be kept in mind that charge-

transfer rates depend not only on the barrier height inside the

semiconductor, but also on the energy overlap between the

semiconductor bands and the surface species.

On the basis of our experimental results, two comments can

be made regarding theories which other authors have assumed

to be valid for hydrogen gas evolution on GaAs. First, it should

now be clear that it is incorrect to assume that the charge-transfer

steps are activated across the Helmholtz layer,^23,24because the

electrons are transferred to surface sites, without passing through

the Helmholtz layer. Second, hydrogen gas evolution on GaAs

cathodes is not suitable to test the validity of theories putting

an upper limit on rates of direct charge transfer to dissolved

electrolyte species, in contrast to the suggestion by Nozik and

Memming:⁷⁸even when the cathodic hydrogen gas evolution

rate is limited by thermionic emission, hydrogen evolution is

bound to be a two-step process with an adsorbed hydrogen

intermediate. And it should not be forgotten that even for

The hydrogenation mechanism was addressed in detail in the

previous sections. The inverse process, hydroxylation, requires

holes or, in the absence of an anodic current, it can occur by

chemical reaction with oxygen (Figure 2). The current density

corresponding to the flow of oxygen toward the surface will be

referred to as j_oxand the current density of electrons transferred

from the conduction band as j_e. When |j_e| > |j_ox|, all oxygen

which reaches the surface is reduced cathodically. Nevertheless,

since cathodic oxygen reduction at GaAs occurs partially via

the valence band,⁸¹a hole current density of the order of j_oxis

injected under cathodic conditions, which may contribute to the

oxidation of surface species.

During negative potential scans, hydrogen gas evolution

begins at a somewhat more negative potential than the H₂/H⁺

redox potential [E°(H₂/H⁺) ) -0.20 V vs Ag/AgCl(saturated

KCl)]: AsH appears when |j_e| > |j_ox|. θ_Hincreases more or

less linearly with U during the negative scan, in such a way

that a large part of changes in applied potential falls across the

Helmholtz layer. When the scan direction is reversed, θ_Hdoes

not start to decrease immediately. In fact, closer inspection of

Figure 8 reveals that θ_Hdoes not start to decrease before current

density has dropped back to about j_ox, i.e., when the electron

flow is no longer sufficient to reduce all the oxidant molecules

which reach the surface from solution. This is even clearer in

Figure 15, in which θ_His plotted against j. With deaerated

solutions, the transition current density j_oxis of the order of 1

µA cm^-2(Figure 15a,b), while with oxygen-saturated solutions,

it is of the order of 100 µA cm^-2(Figure 15c). The apparent

condition that θ_Hwill not decrease before |j_e| < |j_ox| describes

2958 J. Phys. Chem. B, Vol. 103, No. 15, 1999

Erne´ et al.

postulated that the radical intermediate in the hydrogen evolution

reaction is a very efficient recombination center.^84,85

Despite the conclusion that k₄. k₃, a majority of holes might

nevertheless be consumed in the oxidation of AsH instead of

As^•during cathodic hydrogen gas evolution. The reason is that

the reaction rates depend not only on the rate constants but also

on the surface coverages. If (k₂/k₅) . (k₄/k₃), both (k₄. k₃)

•

and k₃θ_H. k₄θ_Ascan be valid simultaneously, because (θ_H/

•

θ

_As) ≈ (k₂/k₅) when hydrogen gas evolution is the dominating

reaction. In this way, only a minute fraction of the holes is

available for the oxidation of As^•to AsOH during reverse scans,

so that, except at extremely low scan rates, θ_Hdoes not decrease

during reverse scans before |j_e| < |j_ox|.

In the steady state, the oxidation of As^•to AsOH must be

balanced by the reduction of AsOH sites. On the assumption

that the current density due to the latter process is given by

-ek₁N_sθ_OHn_s, one of the steady-state conditions is

k₄θ_As

^•j_ox) ek₁N_sθ_OHn_s

(18)

k₃θ_H

The dependence of U_fbon θ_His approximated by a linear relation

(Figure 10):

U_fb(θ_H) ) U_fb(θ_H) 0) + â θ_H

(19)

•

where â < 0. With eq 13 and (θ_H/θ_As) ≈ (k₂/k₅), the last two

equations lead to the following expression for θ_H:

k_BT ek₁N_sN_D(1 - θ_H)

ln

1

â

1

â

θ_H≈ U - U_fb(θ_H) 0) -

âe

(20)

Figure 15. Hydrogen surface coverage of n-GaAs electrodes as a

function of the current density on a logarithmic scale for deaerated

solutions of (a) 1 M HCl and (b) 1M NaOH (same data as Figure 8)

and (c) for an oxygen-saturated solution of 1 M HCl.

(k₄k₅/k₂k₃)j_ox

The general origin of the linear potential dependence of θ_H

during negative scans is clear: a large part of changes in applied

potential falls across the surface dipole layer, while U_fbshifts

more or less linearly with θ_H(eq 19).

The dynamic potential dependence of hydrogen surface

coverage under illumination or in the dark was simulated

numerically on the basis of the reaction scheme in Figure 12.

The method is detailed in the Appendix. Resulting j-U and

θ_H-U curves are presented in Figures 16 and 17. The simulated

behavior is in general agreement with the experiments (compare

the hysteresis at low potential scan rates, for which it was

already remarked elsewhere¹¹that hysteresis becomes essentially

independent of the potential scan rate. At high potential scan

rates (Figure 6), the additional hysteresis occurs because the

highest rate at which adsorbed hydrogen can be removed is equal

to j_ox. Although Figures 13b and 14b evoke a Frumkin isotherm

for adsorbed species showing strong attractive interaction,^82,83

the pivotal role of j_ox(Figure 15) clearly demonstrates that the

origin of the present hystereses is kinetic, not thermodynamic.

An explanation must therefore be sought in the reaction

mechanism.

with Figures 6 and 8, respectively). Under illumination (j_ox

_photo), a photocurrent equal to j_oxflows anodically, and it is

compensated by the electron current j_eat the open circuit

)

j

The magnitudes of the rate constants in the reaction scheme

in Figure 12 are now examined. It was already concluded that

k₂. k₅sAsH hydrogenation is the rate-limiting step in

hydrogen gas evolution. Since θ_Hand θ_OHare more or less of

the same order of magnitude while hydrogen evolution is the

dominant reaction, it can also be concluded that k₅. k₁. The

hole current is distributed over the oxidation of AsH and that

of As^•. The rapid scan in the positive direction in Figure 6

indicates that when an AsH surface is oxidized to AsOH, the

surface does not go through an intermediate situation where it

is mainly covered with As^•, because AsH and U_fbwould be

less well correlated. Hence, when an AsH site is oxidized

anodically to As^•, it is immediately further oxidized to AsOH,

so that k₄. k₃in the reaction scheme. The conclusions that k₂

. k₅and k₄. k₃agree with the expectation that a surface

radical like As^•is much more reactive than AsOH and AsH

sites. The high reactivity of As^•could be part of the explanation

for the high surface recombination rates of photogenerated

electron-hole pairs at GaAs electrodes³²(see the j ≈ 0 plateau

in Figure 6). For germanium, Memming and Neumann also

potential. In the dark (j_ox) j_O), j_ecompensates j_oxwhen all

2

oxygen which arrives at the surface is reduced cathodically. In

terms of j_eand j_ox, the behaviors under illumination and in the

dark are very much alike. Hysteresis of the current-potential

curves is the largest during rapid scans, and some hysteresis

remains even during very slow scans. Simulated coverage by

As^•is negligible with respect to θ_Hand θ_OH, so that θ_OH≈ 1 -

θ_H. When |j_e| equals |j_ox|, electrons convert AsOH into As^•and

holes As^•back into AsOH, i.e., As^•acts as a recombination

center. At more negative potentials, |j_e| exceeds |j_ox|, and excess

electrons lead to AsH and its hydrogenation: hydrogen gas is

evolved. During a slow scan, θ_Hbegins to increase sooner than

in a rapid scan, leading to a larger negative shift in flatband

potential and hence to a lower current density. θ_Hincreases as

potential becomes more negative, but it remains lower than the

value which would cause the entire change in applied potential

to drop across the surface dipole layer; otherwise, band bending

would no longer change and there would no longer be a driving

force for θ_Hto change either. The Tafel slope during scans in

the negative direction is very much dependent on the value for

Hydrogen Gas Evolution on GaAs Cathodes

J. Phys. Chem. B, Vol. 103, No. 15, 1999 2959

The simulation in Figure 17 reveals an effect which one could

easily have overlooked in Figure 8: between the potential where

AsH first appears during the negative scan and the potential

where AsH is no longer present during the positive scan, current

density is higher during the forward scan compared to the

reverse scan by a factor of 2. This is related to the fact that

cathodic oxygen reduction on GaAs occurs for one-half via the

conduction band and for one-half via the valence band. Two

electrons are transferred to each oxygen molecule, leading to

the injection of two holes. During the scans in negative direction,

the electron concentration at the surface is sufficiently high that

the injected holes recombine with electrons, so that four

electrons are measured per oxygen molecule. During the scans

in positive direction, the electron concentration at the surface

is much lower at the same potentials, due to the presence of

surface hydrogen and its effect on semiconductor bandbending,

so that the injected holes do not recombine with electrons but

lead to the dissolution of GaAs: only two electrons are measured

per oxygen molecule.

Differences between the simulations and the experiments are

in part due to experimental complications not taken into account

in the simulations. Irreversible surface changes were not taken

into account, nor was the role of a minority of extremely reactive

sites. The potential- and time-dependent coverage by positively

charged anodic dissolution intermediates could have been

included in Figure 16, leading to shifts of the flatband potential

and to a less abrupt change from zero current to the photocurrent

plateau. Chemical oxidation by oxygen was neglected in Figure

16, and its kinetics were oversimplified in Figure 17sdepending

on the kinetics, time-dependent changes of the local oxygen

concentration could affect the curves too. Despite all such

simplifications, the simulations reproduce most essential features

of the experiments. It can be concluded that the hysteresis and

other dynamic behavior observed cathodically under illumination

and in the dark can be quite well understood on the basis of the

simple reaction scheme in Figure 12.

Figure 16. Numerical simulations of the dynamic potential dependence

of (a) current density and (b) hydrogen surface coverage according to

the reaction scheme in Figure 12. Notice the similarity with Figure

6a,b. The photohole current density is given by j_ox) 7 × 10^-6A cm^-2

,

and the following values were chosen for the rate constants: k₁) 10^-13

cm³s^-1, k₂) 10^-7cm³s^-1, k₃) 10^-12cm³s^-1, k₄) 10^-11cm³s^-1, k₅

) 10^-12cm³s^-1, k₆) 10^-14cm³s^-1, k₇) 10^-2cm³s^-1, and k₈)

10^-10cm³s^-1. Other input parameters were U_fb(θ_H) 0) ) -0.85 V vs

Ag/AgCl, â ) -0.4 V per monolayer (eq 19), N_s) 10¹⁵cm^-2, T )

300 K, and N_D) 7 × 10¹⁶cm^-3

.

5. Conclusion

The microscopic mechanism of hydrogen gas evolution on

GaAs cathodes has been unraveled by in situ infrared spectros-

copy. Surface hydrogen is bound to arsenic atoms and replaces

surface OH-groups present under anodic conditions, causing a

change in the surface dipole potential. Electron transfer to As-H

sites is rate-determining for the hydrogen gas evolution reaction,

so that the microscopic reaction mechanism is pH-independent.

During potential scans in a negative direction, hydrogen surface

coverage increases linearly with potential in such a way that

changes in applied potential fall in large part across the dipole

layer. During positive scans, θ_Hremains constant until the

current density due to the transfer of electrons from the

conduction band no longer exceeds the flow of oxidants toward

the surface. Since a change in θ_Haffects the surface electron

concentration at constant potential, the hysteresis in θ_Hduring

potential cycles is the cause of the hysteresis in the current

density.

Figure 17. Same as Figure 16, but in the darksj_ox) 2 × 10^-6

A

cm^-2now represents the rate at which oxygen arrives at the surface,

and j_anodic) 5 × 10^-8A cm^-2is the reverse bias current thermally

generated inside the semiconductor depletion layer (scan rate ) 1 mV

s^-1). Notice the similarity with measurements at pH 0 shown in Figure

8a,b. For k₉and k_ch(eq 25), it is assumed that k₉) (10^-2cm³) × k_ch.

â in eq 19 but is also affected in a complex way by the rate

constants. During reverse scans, θ_Hat first remains constant,

because most holes are consumed in the oxidation of AsH to

As^•, which is rapidly rehydrogenated. Since changes in applied

potential fall almost exclusively across the semiconductor space

charge layer, the Tafel slope is (k_BT/e) ln(10). Only when the

flow of oxidants toward the surface is no longer compensated

by the flow of electrons does AsH decrease. The highest rate at

which AsH can disappear corresponds to j_ox.

The techniques applied here can also be applied to investigate

the electrochemical behavior of other materials. A recent

example is the in situ infrared spectroscopic study of germanium

electrodes which reveals how the dynamic current-potential

behavior is correlated to variations in the surface concentrations

of adsorbed hydroxyl groups, adsorbed hydrogen atoms, and

free charge carriers.⁷⁴The fundamental knowledge acquired on

n-GaAs electrodes in indifferent electrolyte solutions forms a

basis for investigations of more complex systems, such as GaAs

2960 J. Phys. Chem. B, Vol. 103, No. 15, 1999

Erne´ et al.

electrodes in the presence of hydrogen peroxide. The latter

chemical is commonly used in GaAs surface preparation

technology,^25,26,86,87although its electrochemical behavior is not

yet completely understoodsthe current density exhibits stable

periodic oscillations in the cathodic range.^88,89It will be shown

in another paper that such effects are strongly correlated to

variations in surface chemistry.⁹⁰

ing oxygen is assumed to oxidize AsH and As^•chemically (for

simplicity, one and the same rate constant k_chis taken for AsH

and As^•). The electron transfer rate to dissolved oxygen is taken

proportional to -ek₉n_sunder conditions where it is kinetically

limited, resulting in an electron transfer current density and a

hole current density of equal magnitude j_h:

j

k₉n_s

O₂

Acknowledgment. The authors thank D. Vanmaekelbergh,

J. J. Kelly, D. Paget, and F. Maroun for helpful discussions, F.

Maroun for help with the numerical fitting of infrared spectra,

J. J. Kelly for GaAs material, and the European Commission

for a TMR Research Training Grant (B.H.E.).

j )

(25)

( )

h

2

k₉n_s+ k_ch

In addition to j_hin the dark and the photogeneration current

density j_photounder illumination, another source of holes at the

surface is the small reverse bias current density j_anodicthermally

generated in the semiconductor space charge layer, fixed at 50

nA cm^-2in the simulations (in principle, it is proportional to

the square root of band bending³⁰). The concentration p_sof holes

at the surface is assumed to approximate a steady state:

APPENDIX Numerical Simulation of the

Current-Potential Curves

Details are given of the method by which current-potential

curves were simulated numerically (Figures 16 and 17).

Despite comments concerning the role of a minority of

extremely reactive sites, the rate of each reaction step is taken

to be proportional to the surface coverage of the species reduced

or oxidized and to the concentration n_sof free electrons or p_s

of free holes, respectively. The contributions of the steps to the

j_photo+ j_anodic+ j_h

p_s)

(26)

(k₃θ_H+ k₄θ_As_•+ k₆θ_OH+ k₈θ_ox)N_se

The concentration n_sof electrons at the surface is calculated

using eq 13, on the assumption that only a fraction of the

electrons reaching the surface are transferred. The electron

capture constants are chosen according to k₁, k₅, k₂and

(k₂/k₅) . (k₄/k₃).

•

current density are thus assumed to be -ek₁N_sθ_OHn_s, -ek₂N_sθ_Asn_s,

15

•

+ ek₃N_sθ_Hp_s, +ek₄N_sθ_Asp_s, and -ek₅N_sθ_Hn_s. N_sis set at 10

cm^-2. It is assumed that oxidations are due to a flow of oxidants

corresponding to a current density j_ox. The flatband potential

U_fb(θ_H) 0) in the anodic range is taken to be -0.85 V vs

Ag/AgCl, linear U_fbdependence on θ_His assumed (â ) -0.4

V per monolayer, eq 19), and charge by surface intermediates

is neglected.

•

Starting in the anodic range with θ_OH) 1, θ_As) 0, and θ_H

) 0, the applied potential is slowly changed at a chosen scan

rate in increments corresponding to time steps of 1 µs. After

each U increment, the resulting new values of p_sand n_sare

Under illumination, j_oxrepresents the photohole current

density j_photo. Under stationary conditions in the anodic range,

j_ox) 2ek₆N_sθ_OHp_sis taken to lead to dissolution of the

semiconductor:

calculated (eqs 26 and 13) and from that the changes in θ_H,

•

θ

_As, and θ_OH

:

∆θ_As_•) k₁θ_OHn_s+ k₃θ_Hp_s+ k₅θ_Hn_s- k₂θ_As_•n_s-

(

k₆p_s

AsOH + h⁺

8 ox.

(21)

(22)

(23)

θ_As

•

k₄θ_As_•p_s- (j - 2j_h)

∆t (27)

O₂

)

k₇n_s

θ_As_•+ θ_H

ox. + e^-

8 AsOH

k₈p_s

ox. + h⁺

8 AsOH + dissolution products

∆θ_H) k₂θ_As_•n_s- k₅θ_Hn_s- k₃θ_Hp_s- (j

-

O₂

(

θ_H

θ_As_•+ θ_H

Equation 22 is included in order to simulate surface recombina-

tion of photogenerated holes with electrons at photoanodic

dissolution intermediates. A value of 10^-14cm³s^-1is taken

for k₆, corresponding to the value found by Lingier et al.⁹¹for

p-GaAs anodes in aqueous solutions; k₃and k₄were chosen

according to k₆, k₃, k₄[k₆, k₃, because all anodic current

goes toward the oxidation of surface hydrogen as long as AsH

is present (Figures 4 and 6)]. k₇and k₈were chosen according

to k₇θ_ox. k₁θ_OHand k₈. k₆, so that coverage by “ox.” always

remains very small and the surface is essentially OH-covered

in the anodic range. Surface coverage θ_oxby dissolution

intermediate “ox.” is assumed to be in a steady state:

2j_h)

∆t (28)

∆θ_OH) -∆θ_As_•- ∆θ_H

(29)

From the new value of θ_H, the new value of U_fbis calculated

(eq 19). Finally, the total (electron) current density is determined:

j ) j_photo+ j_anodic- j_h- eN_s(k₁θ_OHn_s+ k₂θ_As_•n_s+ k₅θ_Hn_s+

k₇θ_oxn_s) (30)

The j_photoand j_anodicterms are included because they correspond

to the generation of electron-hole pairs. The electrons move

toward the inside of the semiconductor, due to the electric field,

and are measured in the external circuit if the holes do not

recombine with electrons at the surface. The j_hterm has a

negative sign, because it corresponds to the transfer of electrons

from the semiconductor to dissolved oxygen. The last four terms

are negative also, since they correspond to the transfer of

electrons from the semiconductor to surface sites.

k₆p_s

θ_ox

)

θ_OH

(24)

k₇n_s+ k₈p_s

In the dark, j_ox) j_Ois taken to correspond to a potential-

2

independent rate at which oxygen arrives at the surface. Part of

the oxygen is reduced cathodically, for one half by taking

electrons from the conduction band and for one half by injecting

holes into the valence band (“current doubling”⁸¹). The remain-

Hydrogen Gas Evolution on GaAs Cathodes

J. Phys. Chem. B, Vol. 103, No. 15, 1999 2961

(30) Sze, S. M. Physics of Semiconductor DeVices; Wiley: New York,

List of Frequently Used Symbols

1981.

U

applied potential (V vs Ag/AgCl)

flatband potential (V vs Ag/AgCl)

(31) Schro¨der, K.; Memming, R. Ber. Bunsen-Ges. Phys. Chem. 1985,

89, 385.

(32) Li, J.; Peter, L. M. J. Electroanal. Chem. 1986, 199, 1.

(33) Pankove, J. I. Optical Processes in Semiconductors; Prentice Hall:

Englewood Cliffs, 1971.

U_fb

U_stat

stationary potential during transient capacitance mea-

surements (V vs Ag/AgCl)

total current density (A cm^-2

current density corresponding to the flow of oxidants

toward the surface (A cm^-2

current density of electrons transferred from the

conduction band (A cm^-2

the positive elementary charge (C)

)

(34) Chantre, A.; Vincent, G.; Bois, D. Phys. ReV. B 1981, 23, 5335.

(35) Hofmann, D. M.; Meyer, B. K.; Lohse, F.; Spaeth, J.-M. Phys.

ReV. Lett. 1984, 53, 1187.

j

j_ox

)

(36) Allongue, P.; Cachet, H.; Horowitz, G. J. Electrochem. Soc. 1983,

130, 2352.

j_e

(37) Vanmaekelbergh, D.; Gomes, W. P.; Cardon, F. Ber. Bunsen-Ges.

Phys. Chem. 1985, 89, 994.

(38) Allongue, P.; Blonkowski, S.; Lincot, D. J. Electroanal. Chem.

1991, 300, 261.

)

e

k_B

the Boltzmann constant (eV K^-1

)

(39) Uhlendorf, I.; Reineke-Koch, R.; Memming, R. Ber. Bunsen-Ges.

Phys. Chem. 1995, 99, 1082.

(40) Kolthoff, I. M.; Miller, C. S. J. Am. Chem. Soc. 1941, 63, 1013.

(41) Geffeken, G. Z. Phys. Chem. 1904, 49, 257.

(42) The Chemical Rubber Co. Handbook of Chemistry and Physics,

78th ed.; 1997.

(43) Maroun, F.; Ozanam, F.; Chazalviel, J.-N. J. Electroanal. Chem.

1997, 435, 225.

(44) Chabal, Y. J. Surf. Sci. Rep. 1988, 8, 211.

T

absolute temperature (K)

θ_H, θ_OH, ...

N_s

k_i

fractional surface coverage by H, OH, ...

total concentration of surface sites (cm^-2

)

positive reaction rate constant of step i (cm³s^-1

dopant density (cm^-3

concentration of free electrons at the surface (cm^-3

concentration of free holes at the surface (cm^-3

)

N_D

n_s

)

p_s

)

(45) Landolt-Bo¨rnstein, Numerical Data and Functional Relationships

in Science and Technology; Springer-Verlag: Berlin, 1960; Vol. 2-7, p

259.

(46) Koinuma, M.; Uosaki, K. J. Electroanal. Chem. 1996, 409, 45.

(47) Allongue, P.; Cachet, H. Electrochim. Acta 1988, 33, 79.

(48) Yamamoto, A.; Yano, S. J. Electrochem. Soc. 1975, 122, 260.

(49) Tarui, Y.; Komiya, Y.; Harada, Y. J. Electrochem. Soc. 1971, 118,

118.

(50) Uosaki, K.; Koinuma, M.; Kondo, T.; Ye, S.; Yagi, I.; Noguchi,

H.; Tamura, K.; Takeshita, K.; Matsushita, T. J. Electroanal. Chem. 1997,

429, 12.

(51) Gatos, H. C.; Lavine, M. C. J. Electrochem. Soc. 1960, 107, 427.

(52) Gatos, H. C.; Lavine, M. C. J. Electrochem. Soc. 1960, 107, 433.

(53) Van Muylder, J.; Pourbaix, M. In Atlas d′EÄ quilibres EÄ lectro-

chimiques; Pourbaix, M., Ed; Gauthier-Villars: Paris, 1963.

(54) Vanleugenhaghe, C.; de Zoubov, N.; Pourbaix, M. In Atlas

d’EÄ quilibres EÄ lectrochimiques; Pourbaix, M., Ed.; Gauthier-Villars: Paris,

1963.

(55) Batchelor, R. A.; Hamnett, A.; Peat, R.; Peter, L. M. J. Appl. Phys.

1991, 70, 266.

(56) Solomun, T.; McIntyre, R.; Richtering, W.; Gerischer, H. Surf. Sci.

1986, 169, 414.

(57) Woodall, J. M.; Oelhafen, P.; Jackson, T. N.; Freeouf, J. L.; Pettit,

G. D. J. Vac. Sci. Technol. B 1983, 1, 795.

(58) Lingier, S.; Gomes, W. P. Ber. Bunsen-Ges. Phys. Chem. 1991,

95, 170.

(59) Frese, K. W., Jr.; Morrison, S. R. Appl. Surf. Sci. 1981, 8, 266.

(60) Doligue, R. In NouVeau Traite´ de Chimie Mine´rale; Pascal, P., Ed.;

Masson: Paris, 1958; Vol. 11, p 59.

References and Notes

(1) Pearton, S. J.; Corbett, J. W.; Shi, T. S. Appl. Phys. A 1987, 43,

(2) Chevallier, J.; Aucouturier, M. Annu. ReV. Mater. Sci. 1988, 18,

(3) Chevallier, J.; Clerjaud, B.; Pajot, B. In Hydrogen In Semiconduc-

153.

219.

tors, Semiconductors and Semimetals; Pankove, J. I., Johnson, N. M., Eds.;

Academic: New York, 1990; Vol. 34, Chapter 13.

(4) Rahbi, R.; Pajot, B.; Chevallier, J.; Marbeuf, A.; Logan, R. C.;

Gavand, M. J. Appl. Phys. 1993, 73, 1723.

(5) Yablonovitch, E.; Allara, D. L.; Chang, C. C.; Gmitter, T.; Bright,

T. B. Phys. ReV. Lett. 1986, 57, 249.

(6) Nichols, R. J.; Bewick, A. J. Electroanal. Chem. 1988, 243, 445.

(7) Ogasawara, H.; Ito, M. Chem. Phys. Lett. 1994, 221, 213.

(8) Peremans, A.; Tadjeddine, A. Phys. ReV. Lett. 1994, 73, 3010.

(9) Peremans, A.; Tadjeddine, A. J. Electroanal. Chem. 1996, 409,

115.

(10) Ren, B.; Huang, Q. J.; Cai, W. B.; Mao, B. W.; Liu, F. M.; Tian,

Z. Q. J. Electroanal. Chem. 1996, 415, 175.

(11) Erne´, B. H.; Stchakovsky, M.; Ozanam, F.; Chazalviel, J.-N. J.

Electrochem. Soc. 1998, 145, 447.

(12) Gerischer, H.; Mu¨ller, N.; Haas, O. J. Electroanal. Chem. 1981,

119, 41.

(13) Gerischer, H.; Mattes, I. Z. Phys. Chem. Neue Folge 1966, 49, 112.

(14) Uosaki, K.; Kita, H. J. Am. Chem. Soc. 1986, 108, 4294.

(15) Erne´, B. H.; Ozanam, F.; Chazalviel, J.-N. Phys. ReV. Lett. 1998,

80, 4337.

(16) Bockris, J. O’M.; Reddy, A. K. N. Modern Electrochemistry;

Plenum: New York, 1970; Vol. 2.

(17) Aspnes, D. E.; Studna, A. A. Phys. ReV. B 1983, 27, 985.

(18) da Fonseca, C.; Ozanam, F.; Chazalviel, J.-N. Surf. Sci. 1996, 365,

1.

(19) Chazalviel, J.-N. Electrochim. Acta 1992, 37, 865.

(20) Gee, P. E.; Hicks, R. F. J. Vac. Sci. Technol. A 1992, 10, 892.

(21) Qi, H.; Gee, P. E.; Hicks, R. F. Phys. ReV. Lett. 1994, 72, 250.

(22) Venkateswara Rao, A.; Ozanam, F.; Chazalviel, J.-N. J. Electro-

chem. Soc. 1991, 138, 153.

(23) Brummer, K. D. N. J. Electrochem. Soc. 1967, 114, 1274.

(24) Frese, K. W., Jr. Studies of Physical and Theoretical Chemistry,

Vol. 55: Semiconductor Electrodes; Finklea, H. O., Ed.; Elsevier: Am-

sterdam, 1988.

(25) Notten, P. H. L.; van den Meerakker, J. E. A. M.; Kelly, J. J. Etching

of III-V Semiconductors, An Electrochemical Approach; Elsevier Advanced

Technology: Oxford, 1991.

(61) Chang, C. C.; Citrin, P. H.; Schwartz, B. J. Vac. Sci. Technol. 1977,

14, 943.

(62) Lu, Z. H.; Lagarde, C.; Sacher, E.; Currie, J. F.; Yelon, A. J. Vac.

Sci. Technol. A 1989, 7, 646.

(63) Hsieh, H. F.; Shih, H. C. J. Electrochem. Soc. 1991, 138, 1965.

(64) Gerischer, H.; Hoffmann-Pe´rez, M.; Mindt, W. Ber. Bunsen-Ges.

Phys. Chem. 1965, 69, 130.

(65) Lafle`re, W. H.; Cardon, F.; Gomes, W. P. Surf. Sci. 1974, 44, 541.

(66) Horowitz, G.; Allongue, P.; Cachet, H. J. Electrochem. Soc. 1984,

131, 2563.

(67) Benard, D. J.; Handler, P. Surf. Sci. 1973, 40, 141.

(68) Morrison, S. R. Electrochemistry at Oxidized Metal Electrodes;

Plenum: New York, 1980.

(69) Lohmann, F. Ber. Bunsen-Ges. Phys. Chem. 1966, 70, 428.

(70) Gomes, W. P.; Cardon, F. Prog. in Surf. Sci. 1982, 12, 155.

(71) Lincot, D.; Vedel, J. J. Electroanal. Chem. 1987, 220, 179.

(72) Lincot, D.; Vedel, J. J. Phys. Chem. 1988, 92, 4103.

(73) Memming, R.; Neumann, G. J. Electroanal. Chem. 1969, 21, 295.

(74) Maroun, F.; Ozanam, F.; Chazalviel, J.-N. To be published.

(75) Vanmaekelbergh, D.; Gomes, W. P. J. Phys. Chem. 1990, 94, 1571.

(76) Allongue, P. In Modern Aspects of Electrochemistry, Vol. 23;

Conway, B. E., Bockris, J. O’M., White, R. E., Eds.; Plenum: New York,

1992; Chapter 4.

(26) Gomes, W. P.; Goossens, H. H. In AdVances in Electrochemical

Science and Engineering; Gerischer, H., Tobias, C. W., Eds.; VCH:

Weinheim, 1994; Vol. 3.

(27) Rappich, J.; Lewerenz, H. J.; Gerischer, H. J. Electrochem. Soc.

1993, 140, L187.

(77) Crowell, C. R. Solid-State Electron. 1965, 8, 395.

(78) Nozik, A. J.; Memming, R. J. Phys. Chem. 1996, 100, 13061.

(79) Vanmaekelbergh, D. Electrochim. Acta 1997, 42, 1121.

(80) Vanmaekelbergh, D. Electrochim. Acta 1997, 42, 1135.

(28) Maroun, F.; Ozanam, F.; Chazalviel, J.-N. Chem. Phys. Lett. 1998,

292, 493.

(29) Higashi, G. S.; Chabal, Y. J.; Trucks, G. W.; Raghavachari, K.

Appl. Phys. Lett. 1990, 56, 656.

2962 J. Phys. Chem. B, Vol. 103, No. 15, 1999

Erne´ et al.

(81) Li, J.; Peat, R.; Peter, L. M. J. Electroanal. Chem. 1986, 200, 333.

(82) Frumkin, A. Z. Phys. Chem. 1925, 116, 466.

(83) Bard, A. J.; Faulkner, L. R. Electrochemical MethodssFundamentals

and Applications; Wiley: New York, 1980.

(88) Koper, M. T. M.; Meulenkamp, E. A.; Vanmaekelbergh, D. J. Phys.

Chem. 1993, 97, 7337.

(89) Koper, M. T. M.; Vanmaekelbergh, D. J. Phys. Chem. 1995, 99,

3687.

(84) Memming, R.; Neumann, G. Surf. Sci. 1968, 10, 1.

(85) Memming, R.; Neumann, G. J. Electroanal. Chem. 1969, 21, 295.

(86) Tuck, B. J. Mater. Sci. 1975, 10, 321.

(90) Erne´, B. H.; Stchakovsky, M.; Vanmaekelbergh, D.; Ozanam, F.;

Chazalviel, J.-N. To be published.

(91) Lingier, S.; Gomes, W. P.; Cardon, F. Ber. Bunsen-Ges. Phys.

Chem. 1989, 93, 2.

(87) Kern, W. RCA ReV. 1978, 39, 278.

Article Doi

The Mechanism of Hydrogen Gas Evolution on GaAs Cathodes Elucidated by In Situ Infrared Spectroscopy

DOI: 10.1021/jp984765t

Source and publish data:

Authors:

Article abstract of DOI:10.1021/jp984765t

Full text of DOI:10.1021/jp984765t

Products guided by the article

R&D Labs maybe for 1333-74-0

Relevant to this article

Hot Product