ARTICLE

Rearrangement of π‑Electron Network

and Switching of Edge-Localized π
State in Reduced Graphene Oxide
Shintaro Fujii* and Toshiaki Enoki*

Department of Chemistry, Tokyo Institute of Technology, 2-12-1 Ookayama, Meguro-ku, Tokyo 152-8551, Japan

ABSTRACT Introduced defects can modulate the intrinsic electronic structure

Downloaded via INTL IBERIAN NANOTECHNOLOGY LAB INL 21701 on May 9, 2023 at 22:30:10 (UTC).

of graphene, causing a drastic switch in its electronic and magnetic properties, in

See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.

which defect-induced localized π states near the Fermi level play an important
role. Accordingly, considerable effort has been directed toward detailed char-
acterization of the defect-induced state; however, identification of the chemical
nature of the defect-induced state remains a challenge. Here, we demonstrate a
method for reliable identification of the localized π states of oxidized vacancy edges in reduced graphene oxide. Depending on the dynamic changes in the
oxygen-binding modes, i.e., between carbonyl and ether forms in the vacancy edges, the π-electron network near the edges can rearrange, leading to
drastic on
off switching of the localized π state. This switching can be manipulated via scanning-probe-induced local mechanical force. This study
provides fundamental guidance toward understanding how oxidized defect structures contribute to the unique electronic state of graphene oxide and its
potential future applications in electronic devices.

KEYWORDS: oxidized graphene . defect . electronic state . scanning probe microscopy . density functional theory

G
raphene, which comprises a single belonging to one of sublattices and both of
atomic sheet of sp2-hybridized car- the paired sublattices in graphene bipartite
bon atoms, is a key building block lattice, respectively.3
5,7,19
21 The zigzag-
for future generations of carbon-based elec- shaped edges support edge-localized non-
tronic devices because of the tunability bonding π state (edge state) located close
of its electronic and magnetic properties to the Fermi energy, which originates from
through nanofabrication and chemical mod- breaking of the local sublattice symmetry
ification.1
8 Among techniques of tuning at the zigzag-edge-sites of the bipartite
properties, the introduction of defects into lattice,7,19
21 while the armchair-shaped edges
√
graphene sheet is of particular importance, cause standing wave states with 3a 
√
and extensive experimental efforts have 3a periodicities (a = graphene lattice
focused on exploiting the defect-induced constant) near the edge regions, which
electronic states to functionalize graphene. come from the AB sublattice symmetry
The most widely discussed defect structures and can be explained by intervalley scatter-
are linear defects of graphene edges and ing of conduction electrons at the armchair
point defects in graphene plane.8 Recent edge boundary.7,20
24 The edge state and
advances in structural and electronic char- standing wave in zigzag and armchair
acterization of the graphene edges,2,7,9,10 edges, respectively, are the source of the
combined with development of their nano- potential functionality in graphene. Namely,
fabrication processes based on (sono-)- the edge state, which has a large local * Address correspondence to
chemical,11,12 lithographic,13,14 and bottom- density of states at the Fermi energy with [email protected],
[email protected].
up synthetic15
18 methods, have revealed strong spin polarization, is the origin of
that electronic properties of the graphene electronic, chemical and magnetic activities Received for review September 22, 2013
edges are crucially dependent on the atom- in graphene edges,25 while the standing and accepted November 9, 2013.
ic-scale-edge shapes. There are two typical wave is responsible for the stability of
Published online November 09, 2013
graphene edge shapes that are referred to edges. Beyond the issue of the typical 10.1021/nn404937z
as zigzag and armchair shapes, in which edge-shape dependent electronic proper-
qthe edges are terminated by carbon atoms ties, the effects of edge-terminated chemical C 2013 American Chemical Society

FUJII AND ENOKI VOL. 7 ’ NO. 12 ’ 11190–11199 ’ 2013 11190

www.acsnano.org
 ARTICLE
functional groups26
29 as well as reconstructed30 and Herein, we present the structural and electronic
unreconstructed31,32 chiral edge-shapes other than the characterization of the point defects in defective gra-
typical ones on the electronic properties have been phene, which is prepared by chemical and thermal
recently studied by means of scanning tunneling reduction of graphene oxide (GO).44 High-resolution
microcopy and transmission electron microscopy, in noncontact atomic force microscopy (NC-AFM) and
which atomic details at the edges, in fact, significantly conductive-AFM observations are utilized to investi-
alter the atomic-scale electronic properties. gate the low-energy π states (defect edge state) of
Point defects possess edge-shape and edge-chem- atomic-scale defects in reduced GO and to correlate
istry dependence on the electronic structures similar to the results with realistic oxidized defect structures with
that of extended linear edges. From experimental the aid of density functional calculations. Finally, we
aspects, isolated point defects in graphene sheet have demonstrate the dynamic behavior of the defect
attracted much attention possibly due to their easier structures in which the defect-induced low-energy π
experimental accessibility, in which atomically sized state switches on and off upon rearrangement of the
defects are introduced by standard industrial irradia- π-electron network, which is caused by a reversible
tion technology. The introduction of defect structures, change in the binding modes of the oxygen atoms in
such as monatomic vacancies, adsorbed hydrogen the oxidized defect.
defects,33
35 and substitutional heteroatoms36 into
graphene breaks sublattice symmetry in the graphene RESULTS AND DISCUSSION
bipartite lattice and is predicted to induce defect- Chemical Reduction of Oxidized Graphene Sheets. Figure
localized nonbonding π states (defect edge state) in 1a shows an NC-AFM image of a single-layer GO sheet
the vicinity of the Fermi level. Throughout this paper, with a lateral size of ∼500 nm  500 nm. From exten-
the terms of “defect edge state” and “defect-localized sive spectroscopic characterizations, it is known that
edge state” are used to describe the defect-localized COOH, OH, and CdO groups are present at the edge,
nonbonding π states around atomically sized defects while the basal plane is covered with mostly epoxide
in graphene plane, while the term “edge state” is used and hydroxyl groups, as shown in the structural
to refer to the localized nonbonding π states at the model45 in Figure 1b. To remove the oxygen-containing
zigzag-shaped graphene edge.20,21 Recent scann- functional groups, the GO sheets were reduced by
ing probe microscopy (STM) studies have confirmed exposing them, on Au(111), to hydrazine vapor46
the existence of the defect edge states at mono- followed by annealing at ∼800 °C for 8 h under an
vacancy defects37
40 and monosubstitutional nitrogen ultrahigh vacuum. Previous X-ray photoelectron spec-
defect.41 Together with the emergence of defect edge troscopy, Fourier-transform infrared spectroscopy, and
states, the monatomically sized point defects feature thermogravimetry characterizations, density func-
standing wave states around the defect sites,37
40 tional theory (DFT) simulations have demonstrated
which can originate from the short-range impurity that GO is reduced through a combination of (i) ring-
potential localized at the defect sites that is a source opening of the epoxide groups on the basal plane to
of intervalley as well intravalley electron scattering.42 form hydroxyl groups by chemical reduction and (ii)
While the amount of research dealing with atomic- dehydroxylation by moderate heat treatment.44
scale characterization of the linear defects with a Desorption of OH radicals on the basal plane restores
variety of the edge-shapes and chemistry at the edges the π-electron network, while desorption at the edges
has increased,26
32 a majority of the atomic-scale yields σ radicals. The COOH groups at the edges are
characterization on the point defects has been focused removed by thermal reduction at 100
150 °C.44,47
on several model systems such as pristine and/or However, the CdO moieties at the edges are much
reconstructed single vacancy defects,37
40 a mono- more stable and can only be reduced at temperatures
hydrogen-adsorbed defect,43 and a substitutional as high as 1730 °C.44,47 Therefore, carbonyl groups at
defect.41 This is most probably due to the difficulty in the edges or defective edge sites (i.e., atomic-scale
introducing atomic-scale defects in a controlled man- vacancy edges in the basal plane) are rarely eliminated
ner. Introduction of chemical functional groups into using the current reduction process. Figure 1c and d
the point defects as a form of chemically modified show current mapping images of GO sheets before
vacancy defects and chemical-functional-group- and after chemical/thermal reduction, respectively.
adsorbed defects (i.e., on plane impurity defects) in- A significant increase in the vertical electronic con-
evitably induces large out-of-plane deformation of ductivity and recovery of the π-electron network are
graphene plane around the defects, which would lead apparent after reduction. Although reduction causes
to formation of several metastable structures. This considerable loss of oxygen-containing functional
inherent variability of defect structures in defective groups, such as epoxide and hydroxyl groups, from
graphene has hampered detailed electronic character- the GO surface, oxidized defect structures remain in
ization associated with the atomic-scale chemistry of the basal plane. In the following section, we focus
the point defects. on atomic-scale characterization of the modulated

FUJII AND ENOKI VOL. 7 ’ NO. 12 ’ 11190–11199 ’ 2013 11191

www.acsnano.org
 ARTICLE
Figure 1. NC-AFM and conductive-AFM images of the GO sheets before and after reduction. (a) NC-AFM image of single-layer
GO sheets on freshly cleaved highly oriented pyrolytic graphite (frequency shift =
10 Hz). (b) Structural model of GO,
wherein the basal plane is modified by hydroxyl and epoxide functional groups. (c) Current-mapping image of GO on Au(111)
before reduction recorded at a sample bias voltage (Vs) of 10 mV. Significant spatial variation in the vertical conductivity is
apparent at the nanometer scale. The darker regions with almost no current signals are attributable to oxidized domains,
while the patched conductive regions are nonoxidized graphene domains. (d) Current-mapping image of GO on Au(111) after
chemical and thermal reduction (Vs = 2 mV). Dotted circles are guide for eye to indicate some of the atomically sized defects
with higher LDOS. (e) Current versus bias voltage curves recorded on reduced graphene regions with clear honeycomb π-state
patterns in the reduced GO sheet (
10 mV < Vs < þ10 mV). (f
h) Current-mapping images of the π states of reduced GO taken
at regions (f), (g), and (h) in image (d) as indicated by dotted square of 1.5 nm 1.5 nm: (f) Honeycomb pattern that
√ is typical
√
for pristine graphene, (g) graphene sublattice pattern with a  a periodicity, and (h) superperiodic patterns with 3a  3a
periodicity (a = graphene lattice constant). Graphene honeycomb model (solid hexagons) and observed periodic patterns
(orange dots) are schematically superimposed on each image. Image size is 1.5 nm  1.5 nm.

electronic structure in the presence of the remaining We believe that, in a similar manner to pristine graphene,
oxidized defects. interaction between reduced GO and Au(111) is weak
High-Resolution Imaging of Atomic-Scale Defects in Reduced enough and hole-doping effect is negligibly small.
Graphene Oxide. In the small bias regime where the In the reduced GO, bright protrusions, which are
tunneling current is linearly dependent on the bias characterized by enhanced LDOS distributions, are dis-
(Figure 1e), the tunneling current signal is proportional tributed in defect regions as can be seen in Figure 1d.
to the local density of states (LDOS) near the Fermi level This demonstrates the presence of quasilocalized π
(Ef), and thus basically reflects the π states of the states (defect edge state). The defect edge states
reduced GO sheet. In the presence of metal substrate, exhibit a spatial distribution of 1
2 nm, which can be
Ef in reduced GO can be shifted from its original ascribed to atomically sized defect structures. Similar
position because of the doping effect at the interface spatial distribution have been typically reported for
between graphene and substrate.48
50 Previous STM defect-localized states of the monovacancy defects in
studies have revealed that Dirac point is downshifted graphite, in which the atomic sized defects show
0.35
0.4 eV upon interaction between exfoliated pris- several nm-sized enhanced LDOS distribution with
tine graphene and polycrystalline gold,48 while no triangular shaped 3-fold symmetry.37,38,40 Figure 1f
h
appreciable doping exists on chemical-vapor-depos- shows high-resolution imaging of the electronic states
ited graphene on Au(111).49 A recent DFT study has (Vs = 2 mV) of the reduced GO sheets, which are taken
demonstrated that graphene on Au(111) is either in the region encircled with (f), (g), and (h) in Figure 1d.
undoped or at most slightly p-type doped.50 Figure 1g and h reveal that the LDOS distributions

FUJII AND ENOKI VOL. 7 ’ NO. 12 ’ 11190–11199 ’ 2013 11192

www.acsnano.org
 ARTICLE
differ noticeably from the honeycomb pattern (Figure 1f)
that is typical for pristine graphene. The results of the
LDOS distributions are summarized with (i) quasiloca-
lized π states (defect edge states) with a  a periodicity
√
(Figure 1g) and (ii) superperiodic pattern with 3a 
√
3a periodicities (a = graphene lattice constant)
(Figure 1h). These experimental findings could be
intuitively explained by the coexistence of zigzag and
armchair geometries in the observed defects. The
quasilocalized π states originate from the broken
symmetry of the pseudospin (sublattice) in the gra-
phene bipartite lattice.7,22,35 Nearest-neighbor tight
Figure 2. Reversible on
off switching of a defect-localized
binding models for the electronic states of vacancies state. (a) High-resolution images of the π state of the ON
in graphene planes have demonstrated that if the state in the defects of reduced GO (Vs = 2 mV). Triangular
distribution of π vacancies between the two sublattices pattern is a guide for the eye and is superimposed on the
image. (b) Magnified image of the vicinity of the defect
is uneven, zero-energy modes (i.e., the quasilocalized indicated by the dotted square in (a) in which a  a
defect edge states) will appear.51 Topological lattice periodicity is shown schematically. (c) Averaged cross-sec-
defects such as adsorbed defects on the graphene tional profiles from the defect site to the bulk region along
the typical graphene crystallographic direction, as indicated
lattice also break the local sublattice symmetry by by the arrows in the ON and OFF state images ((a) and (d),
removing a π electron from the graphene-π-electron respectively). The dotted line for the ON state indicates
network. DFT studies have demonstrated that single exponential curve fitting with a decay length of 0.3 nm. (d)
High-resolution images of the π state of the OFF state (Vs = 2
impurity defects such as “monovacancy”, “adsorbed mV). (e) Magnified image √ of the√ area around the dotted
hydrogen”, and “adsorbed hydroxyl” defects, which square region in (d); 3a  3a periodicity is super-
have broken sublattice symmetry, induce quasiloca- imposed on the image.
lized zero-energy states at nearest neighbor carbon
sites around the defect.33
35,52 The quasilocalized GO during repeated imaging, indicating that the geo-
zero-energy-states decay into bulk with a finite ampli- metry of the π-electron network around the defect can
tude at sites in one of the graphene sublattices, which be reconstructed. About ∼10% of the atomic-scale
manifests a  a periodic patterns. In contrast, such defects undergo a dramatic switch of the enhanced
zero-energy states vanish for an “adsorbed epoxide” LDOS at Ef. Higher resolution imaging shows that the
defect,52 which removes two π electrons from a pair of “ON” state in Figure 2a and b exhibits enhanced LDOS
neighboring A and B carbon-sublattice sites in gra- distribution with a triangular pattern at the defect,
phene plane (it should also be noted that the zero- which decays into the bulk with an exponential decay
mode state necessarily appears in the sublattice- length of ∼0.3 nm and a  a periodicity (Figure 2c),
imbalanced system, while it does not always vanish while the “OFF” state in Figure 2d and e is characterized
in a sublattice-balanced system). In analogy with the by a complete disappearance of the enhanced LDOS
single-impurity defects, oxidized defect structures in (i.e., bright spots) and evolution of the extended super-
√ √
the reduced GO can break the local sublattice sym- periodic patterns with 3a  3a periodicity near the
metry and cause unique edge-localized π states. defect region (Figure 2c).
√ √
The superperiodic patterns with 3a  3a peri- To investigate the dynamic on
off switching be-
odicities (Figure 1h) arise from a standing wave forma- havior, a series of LDOS images around the defect were
tion due to intervalley scattering of conduction elec- recorded at 17 min intervals, during which reversible
trons at defect sites (e.g., see refs 23, 24) or defect- switching between the ON and OFF states occurs
induced electronic gap formation due to electron
several times (Figure 3). Among the 28 images re-
lattice interactions (i.e., charge-density wave formation corded, (2), (12), (24), and (27) (Figure 3) show abrupt
in the π states near Ef) (e.g., see refs 22, 53). Similar switching events induced by the raster scanning of the
modulated LDOS with the superperiodic patterns has AFM tip. This scan-induced switching occurs in both
been commonly identified around atomically sized the ON-to-OFF and OFF-to-ON directions. Halfway
defects in graphene materials.37
40,49 through the upward raster imaging in image (27), the
A more detailed characterization of the electronic OFF state suddenly switches to the ON state (Figure 4a)
structures in the two different kinds of defect-induced when the AFM tip scans just across the defect site, as
states is presented with DFT simulations in the later shown in the scan profiles (Figure 4b). The AFM-tip-
section. In the following section, we focus on dynamic induced local mechanical force on the defect induces
behavior of the edge-localized π state. structural deformation around the defect, which could
Reversible Switching of the Defect-Localized State in Reduced eventually modify the bond configurations.
Graphene Oxide. Interestingly, we observed drastic on
Concurrent with the scan-induced direct switching,
off switching of the defect-localized π states in reduced thermally activated stochastic switching occurs, which

FUJII AND ENOKI VOL. 7 ’ NO. 12 ’ 11190–11199 ’ 2013 11193

www.acsnano.org
 ARTICLE
Figure 3. Switching behavior of the defect-localized state. A series of high-resolution images of the π state at the oxidized
defect, which is subject to on
off switching of the defect-localized state (Vs = 2 mV). Image size is 7.5 nm  7.5 nm. Each image
is recorded by raster scanning the AFM tip at 17 min intervals. The fast scan direction is from left to right (horizontal), and the
slow scan direction (vertical) is represented by a white arrow in each image. In images (2), (12), (24), and (27), scan-induced
switching behaviors are observed when the AFM tip is scanned across the defect site. Consequently, the LDOS patterns
suddenly change halfway through the imaging of the defect site. The ON state switched to the OFF state in images (2) and (24)
(i.e., the bright protrusion disappeared halfway through the imaging of the defect site), and the OFF state switched to the ON
state in images (12) and (27) (i.e., the bright protrusion appeared halfway through the imaging of the defect site).

causes ∼80% of the observed states without the scan- switching in the edge-localized state with the reversi-
induced direct switching event to be in the ON state ble configurational changes (see Supporting Informa-
(Figure 4c). It should be noted that a reversible struc- tion SI-1). The calculated total energy differences
tural change in an AFM-tip apex (i.e., change in the between the two configurations in (2), (3), (6), and (7)
electronic state of the tip) may explain the switching of are approximately þ4 eV (E2i
E2ii),54 þ2.5 eV (E3i
E3ii),
the appearance of the LDOS at the defect site. Gen-
0.1 eV (E6i
E6ii), and
0.3 eV (E7i
E7ii), respectively,
erally, tip structural changes switch the LDOS patterns which indicates that the reversible configurational
of not only the defect site but also the whole sample switch in (2) and (3) is unlikely because of the larger
surface area, which does not occur in this case. Thus the energy differences between the two configurations.
switching of the LDOS patterns that are observed only During the configurational switch from (6i) to (6ii)
at the defect sites is attributable to not just a structural (Figure 5), two carbonyl bonds at the vacancy edges
change of the tip apex but a structural change of the convert into two ether moieties in a concerted manner
defects. (Figure 6a and b). Consequently, the hybridization of
As a result of the chemical/thermal reduction pro- the two oxygen atoms changes from sp2 to sp3, which
cess studied here, oxidized structures in the basal leads to significant rearrangement of the π-electron
plane most likely comprise atomic-scale vacancy edges network around the vacancy-edge sites. In contrast to
modified with σ dangling bonds and oxygen atoms, as the sp3-hybridized oxygen atom in the ether groups,
discussed in the previous section. To identify the the sp2-hybridized oxygen atom in the carbonyl group
optimal defect structure for reversibly switchable bond has a free pz orbital, which can participate into the
configurations, which lead to the evolution and disap- graphene π-electron network.
pearance of the edge-localized state, DFT calculations On the basis of DFT-LDA using linear and quadratic
within the local density approximation (LDA) were synchronous transit methods, as implemented in
performed using graphene supercell models with CASTEP code, the energy barrier of switching from
mono- and dicarbon vacancies modified by oxygen (6i) to (6ii) is estimated to be 0.8 eV; however, no
atoms. The initial atomic configurations were gen- reaction path was evident for the transition between
erated with the oxygen atoms on the top and/or bottom (7i) and (7ii) (see Supporting Information SI-2). During
of the vacancy-edge sites in the graphene supercell.54 switching between (7i) and (7ii), the steric interactions
The atomic positions were then relaxed to the closest of the four carbonyl groups cause large energetic
local minimum of the LDA total energy. Among the destabilization and impede the transition between
metastable configurations found in the structural (7i) and (7ii) (see Supporting Information SI-2).
relaxation (Figure 5), the interconvertible forms of While other scenarios that involve a wide variety of
(2i
2ii), (3i
3ii), (6i
6ii), and (7i
7ii) exhibit on
off chemical functional groups other than oxygen at the

FUJII AND ENOKI VOL. 7 ’ NO. 12 ’ 11190–11199 ’ 2013 11194

www.acsnano.org
 ARTICLE
Figure 5. DFT-relaxed structures of oxidized vacancy edges
in graphene supercells. (1
3) Oxidized monovacancy edges
with 1
3 oxygen atoms, and (4
7) oxidized divacancy
Figure 4. Scan-induced on
off switching of the localized edges with 1
4 oxygen atoms. Oxygen atoms are denoted
state and time evolution of the switching behavior. with red circles. For models 2, 3, 6, and 7, there are two
(a) Magnified high-resolution images of the π state in image structures with the same stoichiometry, which are labeled
(27) (Figure 3); the slow scan direction is indicated by an (i) and (ii). On the basis of DFT calculations, the presence and
arrow. The OFF state suddenly switched to the ON state absence of the edge-localized state are labeled by “ON” and
when the AFM tip scanned across the defect site. “OFF” on the right corner in each model, respectively.
(b) Averaged cross-sectional profiles along the dotted arrow
in (a) before and after switching. The switching event is
characterized by an abrupt decrease in current and is
indicated by an arrow in the current profile. The direction
of the profiles is denoted by the dotted arrow. (c) Switching
behavior as a function of time (image number in Figure 3).
Images (
5) to (0) are not shown in Figure 3. The filled
arrows denote scan-induced switching events (i.e., when the
localized state suddenly turns on or off as the AFM tip scans
across the defect site), while the dotted arrows indicate
stochastic switching events. Beside the scan-induced
switching events, stochastic switching process (e.g., see
image numbers from (4) to (5), from (
4) to (
3), from
(10) to (11), from (19) to (20), from (20) to (21) in c) are also
observed, in which the state switches on and off without
external mechanical perturbation induced by AFM-tip-scan-
ning. Among the total of 34, 5 images show the scan-
induced switching. Among the remaining 29 images with-
out the switching events, 23 images (i.e., ∼80%) exhibit the
ON state. Figure 6. (a) and (b) Top perspective of the two configura-
tions of (6), i.e., (6i) and (6ii), with and without an edge-
localized π state, respectively. Two carbonyl bonds at the
defect edges undergoing reversible structural changes vacancy edges of (6i) can convert into two ether moieties
are possible, only switching of the bonding patterns of (6ii) in a concerted manner. Electron movement is indicated
by single-headed arrows. Oxygen atoms are represented by
oxygen between carbonyl (sp2-hybridized) and ether red circles. (c) and (d) Simplified π-site models of (6i) and
(sp3-hybridized) forms rearranges the π-electron net- (6ii). Black and white balls represent the π-sites belonging
work without considerable energetic destabilization to A(B) and B(A) sublattices, respectively. The sp3-hybri-
dized oxygen site in the ether group and the sp3-hybridized
due to breakage of chemical bonds (see Supporting two carbon sites without free pz orbitals cannot act as
Information SI-3). Taking into account the on
off effective π-site-vacancies. The sp2-hybridized oxygen site
switching of the edge localized states along with the in the carbonyl group with a pz orbital can act as an effective
π-site. The white balls singly bonded in the defect periphery
reversible structural changes, the scenario can be in (c) are oxygen atoms belonging to a sublattice same as
limited to switching behaviors associated with con- that of carbon atoms marked with white balls.
certed reactions where the bonding configurations of
two oxygen atoms at the defect edges convert from graphene π-electron network,55 (ii) the carbonyl
sp2- to sp3-hybridization within the mono- and diva- groups add an additional π-site to the graphene
cancy edge models studied here (see Supporting π-system around the defect edges,29,55 and (iii) the
Information SI-3). epoxide group removes the neighboring π-sites of the
The oxidation-induced rearrangement of the graphene lattice.52 On the basis of these simple guide-
π-electron network (i.e., the local sublattice symmetry) lines, the π-electron networks in the defect structures
could be intuitively modeled by considering that (i) the of (6i) (ON state) and (6ii) (OFF state) can be re-
π electrons in the ether group do not participate in the presented, as shown in Figure 6c and d, respectively.

FUJII AND ENOKI VOL. 7 ’ NO. 12 ’ 11190–11199 ’ 2013 11195

www.acsnano.org
 ARTICLE
Figure 8. (a
d) Simulated current-mapping images of the
(a, b) ON and (c, d) OFF states superimposed on their
structural models (Vs = þ0.3 V). The ON and OFF states
correspond to (6i) and (6ii), respectively. In (b) and (d), the
image contrast is enhanced. The√triangular√ pattern, a  a
periodicity of the ON state, and 3a  3a periodicity of
Figure 7. DFT-calculated band structures and DOSs of the the OFF state are schematically shown in the images. (e)
ON and OFF states of configurations (6i) and (6ii), respec- Averaged cross-sectional profiles of the simulated images
tively. (a) and (b) Band structures of the (a) ON and (b) OFF along the same directions as the experimental profiles
states along the high symmetry points (Γ-K-M-Γ). The (Figure 2c). The directions are denoted by arrows in the
dotted lines correspond to the band structure of pristine ON- and OFF-state images. The STM intensities are normal-
graphene obtained from a similar calculation method. (c) ized to the maximum value of the ON state. The dotted line
DOSs of the ON and OFF states. for the ON state indicates exponential curve fitting. The
exponential decay length is estimated to be 0.35 nm, which
The benzenoid graph theory35 predicts that the zero- is in good agreement with the experimental value of
energy state (edge state) appears when the sublattice ∼0.3 nm (see main text).
sites in a π-electron network are imbalanced (i.e.,
ΔN ¼6 0 where ΔN = |NA
NB|, where NA and NB are the carbon
carbon bond in the epoxy ring (Figure 6b),
numbers of π-sites in sublattices A and B, respectively). in which the bond length is deviated 7% (i.e., 0.153 nm)
In a crude approximation, ΔN can be estimated by from the equilibrium length (i.e., 0.142 nm), whereas
counting the sublattice sites per unit cell in Figure 6c the deviations are less than 4% for remaining carbon

and d. A simple analysis of the sublattice imbalance carbon bonds at the periphery of defects (Figure 6a
demonstrates that ΔN is nonzero for the ON state [(6i)] and b).
and zero for the OFF state [(6ii)], which indicates that DFT-simulated current-mapping images within the
the on
off switching behavior can be understood in Tersoff
Hamann approximation56,57 indicate that (i)
terms of breaking the sublattice symmetry (i.e., a the LDOS at Ef (Figure 7c) for the ON state is strongly
sublattice imbalance). A similar explanation is valid localized at the vacancy edge sites, which decay into
for the on
off switching in (2), (3), and (7). For a more the bulk with a finite amplitude at one of the graphene-
in-depth discussion, the significant out-of-plane defor- sublattice sites (Figure 8a and b) and (ii) the edge-
mations of the carbonyl groups in the oxidized edges localized π state is quenched for the OFF state, in which
of (2), (3), (6i), and (7) and the relevant contribution of extended π state propagates around the defect region
√ √
the oxygen π-electrons into the graphene π-system to form a superperiodic pattern of 3a  3a peri-
should be carefully considered. odicity (Figure 8c and d). A closer look at the simulated
The calculated band structures and DOSs for the images reveals that the decaying π state distribution
two configurations (Figure 6) are shown in Figure 7: with a  a periodicity for the ON state and the
√ √
The ON state exhibits a sharp DOS peak at Ef that extended π state with 3a  3a periodicity for the
originates from nearly flat bands of the defect-localized OFF state correlate well with the experimentally ob-
edge state near Ef (Figure7a and c), whereas the OFF served patterns (Figure 2). The triangular patterns with
state does not have such state (Figure 7b and c). The asymmetric localized state distributions in the ON state
ON state is energetically more stable than the OFF state (i.e., at one of the three vertices of the triangular
(see Figure S2, Supporting Information), which is con- pattern that has a localized state with a much lower
sistent with the preferential emergence of the ON state amplitude) are in qualitative agreement with the ex-
during the stochastic switching process (Figure 4c). perimental data, while the complete disappearance of
The higher stability of the ON state could be derived the edge-localized state in the OFF state is evident
from the energetic destabilization of the OFF state that from both theoretical and experimental results.
occurs because of the structural distortion induced by
the on-plane ether group (i.e., the epoxide). In the LDA- CONCLUSION
relaxed structures (Figure 6a and b), the large strain In summary, we successfully reduced GO on Au(111)
can be recognized as an exceptionally elongated and characterized it by scanning probe microscopy.

FUJII AND ENOKI VOL. 7 ’ NO. 12 ’ 11190–11199 ’ 2013 11196

www.acsnano.org
 ARTICLE
The current-mapping images reveal that the LDOS of the binding mode of the C
O bonds (i.e., either sp2-
the reduced GO is considerably different from that of hybridized carbonyl or sp3-hybridized ether bonds) in
pristine graphene. The LDOS distributions are charac- the oxidized vacancy defect. Consequently, the evolu-
terized by (i) extended superperiodic patterns with tion and disappearance of the defect-localized π state
√ √
3a  3a periodicity as a result of intervalley (i.e., local sublattice symmetry) can be modulated
scattering of conduction electrons and (ii) a defect- through the C
O binding mode, which can be con-
localized edge state that decays into the bulk with a  trolled by tip-induced external mechanical perturba-
a periodicity originating from breaking the symmetry tions, as demonstrated in the AFM manipulation
of the pseudospin in the graphene bipartite lattice. experiment. This study reveals that precise control
Atomic-scale oxidized vacancy edge defects feature of the oxidized structures is crucial for the emer-
on
off switching of the defect-localized π state, which gence of the unique defect-localized π state in
is explained by a tip-induced reversible change in the reduced GO samples, which could be an impor-
oxidized structures, as determined by DFT calculations. tant aspect of future graphene-based electronic/
The π-electron network around the defect depends on magnetic devices.

METHODS No. 25790002) from the Ministry of Education, Culture, Sports,

Science and Technology of Japan.
Sample Preparation. GO samples were prepared from natural
graphite (Kojundo Chemical Laboratory) using the conventional Supporting Information Available: (SI-1) LDA-calculated
Hummers method58 and drop-cast onto freshly cleaved highly band structures for the models (2i), (2ii), (3i), (3ii), (7i), and (7ii).
ordered pyrolytic graphite or Au(111) substrates. The Au(111) (SI-2) Estimation of the energy barrier between the ON and OFF
substrates were prepared via the thermal evaporation of Au states. (SI-3) Proposed oxidized defect models. Figures S1
S9.
onto mica. The GO sheets on Au(111) underwent hydrazine This material is available free of charge via the Internet at http://
reduction through exposure to hydrazine monohydrate (Kanto pubs.acs.org.
Chemical Co., Inc.) vapor. The resultant hydrazine-exposed GO
sheets were then thermally reduced by heating on Au(111) at
800 °C for 8 h under UHV. REFERENCES AND NOTES
AFM Characterization. All AFM measurements were performed 1. Bekyarova, E.; Sarkar, S.; Wang, F.; Itkis, M. E.; Kalinina, I.;
under UHV (∼5  10
8 Pa) at room temperature using a Tian, X.; Haddon, R. C. Effect of Covalent Chemistry on the
commercially available AFM system (VT beam-deflection AFM, Electronic Structure and Properties of Carbon Nanotubes
Omicron). NC-AFM was performed using a phase-locked loop and Graphene. Acc. Chem. Res. 2013, 46, 65–76.
unit (Easy PLL-plus, Nanosurf) for frequency modulation 2. Jia, X.; Campos-Delgado, J.; Terrones, M.; Meunier, V.;
feedback.59 NC-AFM images were acquired using a silicon Dresselhaus, M. S. Graphene Edges: a Review of Their
cantilever (NSC11, Micromash) or titanium
platinum-coated Fabrication and Characterization. Nanoscale 2011, 3,
silicon cantilever (NSC11/Ti
Pt). The cantilevers were oscillated 86–95.
at a constant peak-to-peak amplitude of ∼20 nm at their 3. Loh, K. P.; Bao, Q.; Ang, P. K.; Yang, J. The Chemistry of
resonant frequencies. The resonant frequencies and force con- Graphene. J. Mater. Chem. 2010, 20, 2277–2289.
stants of the cantilevers were 350
450 kHz and 40
50 N/m, 4. Rao, C. N. R.; Sood, A. K.; Subrahmanyam, K. S.; Govindaraj,
respectively. Current mapping was performed using a platinum- A. Graphene: The New Two-Dimensional Nanomaterial.
coated silicon cantilever (CONTPt, Nanosensors) with a spring Angew. Chem., Int. Ed. 2009, 48, 7752–7777.
constant of 0.2 N/m. During imaging, a force of ∼10 nN and a 5. Müllen, K.; Rabe, J. P. Nanographenes as Active Compo-
bias voltage of (0.002
0.150 V were maintained. Within this nents of Single-Molecule Electronics and How a Scanning
bias range, no significant change in the contrast of the current- Tunneling Microscope Puts Them to Work. Acc. Chem. Res.
mapping image was observed. Typically, simultaneously 2008, 41, 511–520.
obtained topographic images exhibited no atomic contrast; 6. Ruoff, R. Calling All Chemists. Nat. Nanotechnol. 2008, 3,
therefore, only tunneling-current images are presented here. 10–11.
DFT Calculations. DFT calculations were performed within the 7. Enoki, T. Role of Edges in the Electronic and Magnetic
LDA using the Perdew
Zunger exchange correlation60 scheme, Structures of Nanographene. Phys. Scr., T 2012, 146,
as implemented in the PWSCF code of the Quantum-ESPRESSO 014008
1–14.
package.61 Rectangular graphene supercells containing 72 8. Banhart, F.; Kotakoski, J.; Krasheninnikov, A. V. Structural
carbon atoms (1.476 nm  1.278 nm) and hexagonal supercells Defects in Graphene. ACS Nano 2011, 5, 26–41.
containing 162 carbon atoms (9a  9a; a = 0.246 nm) were used 9. Zhang, X.; Xin, J.; Ding, F. The Edges of Graphene. Nano-
for modeling of the oxidized structures in Figure 5 and Figures 7 scale 2013, 5, 2556–2569.
and 8, respectively. The defect structures were made by (i) 10. Kobayashi, Y.; Fukui, K.; Enoki, T.; Kusakabe, K.; Kaburagi, Y.
removing one or two carbon atoms from pristine graphene and Observation of Zigzag and Armchair Edges of Graphite
(ii) placing oxygen atoms on the vacancy edges.54 For the Using Scanning Tunneling Microscopy and Spectroscopy.
ultrasoft pseudopotentials adopted here, the cutoff energy for Phys. Rev. B: Condens. Matter Mater. Phys. 2005, 71, 193406.
the expansion of the wave functions was set at 25 Ry for 11. Kosynkin, D. V.; Higginbotham, A. L.; Sinitskii, A.; Lomeda,
calculating the optimized geometries. All atoms were fully J. R.; Dimiev, A.; Price, B. K.; Tour, J. M. Longitudinal
relaxed until the forces were less than 10
3 Ry/a.u. For simulations Unzipping of Carbon Nanotubes to Form Graphene Nano-
of the current-mapping images, the wave function cutoff energy ribbons. Nature 2009, 458, 872–876.
was set at 60 Ry. Brillouin-zone integration was performed using a 12. Tao, C.; Jiao, L.; Yazyev, O. V.; Chen, Y.-C.; Feng, J.; Zhang, X.;
uniform 8  8  1 Monkhorst-type k-point grid.62 Capaz, R. B.; Tour, J. M.; Zettl, A.; Louie, S. G.; et al. Spatially
Conflict of Interest: The authors declare no competing Resolving Edge States of Chiral Graphene Nanoribbons.
financial interest. Nat. Phys. 2011, 7, 616–620.
13. Jiao, L.; Zhang, L.; Wang, X.; Diankov, G.; Dai, H. Narrow
Acknowledgment. This work was supported by Grants-in- Graphene Nanoribbons from Carbon Nanotubes. Nature
Aid for Scientific Research (No. 20001006, No. 23750150, and 2009, 458, 877–880.

FUJII AND ENOKI VOL. 7 ’ NO. 12 ’ 11190–11199 ’ 2013 11197

www.acsnano.org
 ARTICLE
€ Meyer, J. C.; Erni, R.; Rossell, M. D.; Kisielowski, C.;
14. Girit, C-. O.; 34. Boukhvalov, D. W.; Katsnelson, M. I.; Lichtenstein, A. I.
Yang, L.; Park, C.-H.; Crommie, M. F.; Cohen, M. L.; Louie, Hydrogen on Graphene: Electronic Structure, Total Energy,
S. G.; et al. Graphene at the Edge: Stability and Dynamics. Structural Distortions and Magnetism from First-Principles
Science 2009, 323, 1705–1708. Calculations. Phys. Rev. B: Condens. Matter Mater. Phys.
15. Chen, L.; Hernandez, Y.; Feng, X.; Müllen, K. From Nano- 2008, 77, 035427.
graphene and Graphene Nanoribbons to Graphene 35. Yazyev, O. V. Emergence of Magnetism in Graphene Materi-
Sheets: Chemical Synthesis. Angew. Chem., Int. Ed. 2012, als and Nanostructures. Rep. Prog. Phys. 2010, 73, 056501.
51, 7640–7654. 36. Hou, Z.; Wang, X.; Ikeda, T.; Terakura, K.; Oshima, M.;
16. Huang, H.; Wei, D.; Sun, J.; Wong, S. L.; Feng, Y. P.; Neto, Kakimoto, M. Electronic Structure of N-Doped Graphene
A. H. C.; Wee, A. T. S. Spatially Resolved Electronic Struc- with Native Point Defects. Phys. Rev. B: Condens. Matter
tures of Atomically Precise Armchair Graphene Nanorib- Mater. Phys. 2013, 87, 165401.
bons. Sci. Rep. 2013, 2, 983. 37. Kelly, K. F.; Sarkar, D.; Hale, G. D.; Oldenburg, S. J.; Halas, N. J.
17. Konishi, A.; Hirao, Y.; Matsumoto, K.; Kurata, H.; Kishi, R.; Threefold Electron Scattering on Graphite Observed with
Shigeta, Y.; Nakano, M.; Tokunaga, K.; Kamada, K.; Kubo, T. C60-Adsorbed STM Tips. Science 1996, 273, 1371–1373.
Synthesis and Characterization of Quarteranthene: Eluci- 38. Ruffieux, P.; Melle-Franco, M.; Gröning, O.; Bielmann, M.;
dating the Characteristics of the Edge State of Graphene Zerbetto, F.; Gröning, P. Charge-Density Oscillation on
Nanoribbons at the Molecular Level. J. Am. Chem. Soc. Graphite Induced by the Interference of Electron Waves.
2013, 135, 1430–1437. Phys. Rev. B: Condens. Matter Mater. Phys. 2005, 71, 153403.
18. Morita, Y.; Suzuki, S.; Sato, K.; Takui, T. Synthetic Organic 39. Rutter, G. M.; Crain, J. N.; Guisinger, N. P.; Li, T.; First, P. N.;
Spin Chemistry for Structurally Well-Defined Open-Shell Stroscio, J. A. Scattering and Interference in Epitaxial
Graphene Fragments. Nat. Chem. 2011, 3, 197–204. Graphene. Science 2007, 317, 219–222.
19. Sasaki, K.; Saito, R. Pseudospin and Deformation-Induced 40. Ugeda, M. M.; Brihuega, I.; Guinea, F.; Gómez-Rodríguez,
Gauge Field in Graphene. Prog. Theor. Phys. Suppl. 2008, J. M. Missing Atom as a Source of Carbon Magnetism. Phys.
176, 253–278. Rev. Lett. 2010, 104, 096804.
20. Fujita, M.; Wakabayashi, K.; Nakada, K.; Kusakabe, K. Pecu- 41. Zhao, L.; He, R.; Rim, K. T.; Schiros, T.; Kim, K. S.; Zhou, H.;
liar Localized State at Zigzag Graphite Edge. J. Phys. Soc. Gutiérrez, C.; Chockalingam, S. P.; Arguello, C. J.; Pálová, L.;
Jpn. 1996, 65, 1920–1923. et al. Visualizing Individual Nitrogen Dopants in Monolayer
21. Nakada, K.; Fujita, M.; Dresselhaus, G.; Dresselhaus, M. S. Graphene. Science 2011, 333, 999–1003.
Edge State in Graphene Rribbons: Nanometer Size Effect 42. Matis, B. R.; Houston, B. H.; Baldwin, J. W. Low-Energy
and Edge Shape Dependence. Phys. Rev. B: Condens. Resonant Scattering from Hydrogen Impurities in Gra-
Matter Mater. Phys. 1996, 54, 17954–17961. phene. Phys. Rev. B: Condens. Matter Mater. Phys. 2013,
22. Fujii, S.; Enoki, T. Nanographene and Graphene Edges: 88, 085441.
Electronic Structure and Nanofabrication. Acc. Chem. Res. 43. Balog, R.; Jørgensen, B.; Wells, J.; Lægsgaard, E.; Hofmann,
2013, 46, 2202–2210. P.; Besenbacher, F.; Hornekær, L. Atomic Hydrogen Adsor-
23. Yang, H.; Mayne, A. J.; Boucherit, M.; Comtet, G.; Dujardin, bate Structures on Graphene. J. Am. Chem. Soc. 2009, 131,
G.; Kuk, Y. Quantum Interference Channeling at Graphene 8744–8745.
Edges. Nano Lett. 2010, 10, 943–947. 44. Pei, S.; Cheng, H.-M. The Reduction of Graphene Oxide.
24. Park, C.; Yang, H.; Mayne, A. J.; Dujardin, G.; Seo, S.; Kuk, Y.; Carbon 2012, 50, 3210–3228.
Ihm, J.; Kim, G. Formation of Unconventional Standing 45. Lerf, A.; He, H.; Forster, M.; Klinowski, J. Structure of
Waves at Graphene Edges by Valley Mixing and Pseudos- Graphite Oxide Revisited. J. Phys. Chem. B 1998, 102,
pin Rotation. Proc. Natl. Acad. Sci. U. S. A. 2011, 108, 18622– 4477–4482.
18625. 46. Watcharotone, S.; Dikin, D. A.; Stankovich, S.; Piner, R.; Jung,
25. Enoki, T.; Takai, K. The Edge State of Nanographene and I.; Dommett, G. H. B.; Evmenenko, G.; Wu, S. E.; Chen, S. F.;
the Magnetism of the Edge-State Spins. Solid State Com- Liu, C-Pu.; et al. Graphene-Silica Composite Thin Films as
mun. 2009, 149, 1144–1150. Transparent Conductors. Nano Lett. 2007, 7, 1888–1892.
26. Talirz, L.; Söde, H.; Cai, J.; Ruffieux, P.; Blankenburg, S.; 47. Gao, X.; Jang, J.; Nagase, S. Hydrazine and Thermal Reduc-
Jafaar, R.; Berger, R.; Feng, X.; Müllen, K.; Passerone, D.; et al. tion of Graphene Oxide: Reaction Mechanisms, Product
Termini of Bottom-Up Fabricated Graphene Nanoribbons. Structures, and Reaction Design. J. Phys. Chem. C 2010,
J. Am. Chem. Soc. 2013, 135, 2060–2063. 114, 832–842.
27. Kobayashi, Y.; Fukui, K.; Enoki, T.; Kusakabe, K. Edge State 48. Klusek, Z.; Dabrowski, P.; Kowalczyk, P.; Kozlowski, W.;
on Hydrogen-Terminated Graphite Edges Investigated by Olejniczak, W.; Blake, P.; Szybowicz, M.; Runka, T. Graphene
Scanning Tunneling Microscopy. Phys. Rev. B: Condens. on Gold: Electron Density of States Studies by Scanning
Matter Mater. Phys. 2006, 73, 125415. Tunneling Spectroscopy. Appl. Phys. Lett. 2009, 95,
28. Ziatdinov, M.; Fujii, S.; Kusakabe, K.; Kiguchi, M.; Mori, T.; 113114.
Enoki, T. Visualization of Electronic States on Atomically 49. Martínez-Galera, A. J.; Brihuega, I.; Gómez-Rodríguez, J. M.
Smooth Graphitic Edges with Different Types of Hydrogen Ethylene Irradiation: A New Route to Grow Graphene on
Termination. Phys. Rev. B: Condens. Matter Mater. Phys. Low Reactivity Metals. Nano Lett. 2011, 11, 3576–3580.
2013, 87, 115427. 50. Szawi nska, J.; Dabrowski, P.; Zasada, I. Doping of Graphene
29. Ohtsuka, M.; Fujii, S.; Kiguchi, M.; Enoki, T. Electronic State by a Au(111) Substrate: Calculation Strategy within the
of Oxidized Nanographene Edge with Atomically Sharp Local Density Approximation and a Semiempirical Van Der
Zigzag Boundaries. ACS Nano 2013, 7, 6868–6874. Waals Approach. Phys. Rev. B: Condens. Matter Mater. Phys.
30. Pan, M.; Girao, E. C.; Jia, X.; Bhaviripudi, S.; Li, Q.; Kong, J.; 2010, 83, 245429.
Meunier, V.; Dresselhaus, M. S. Topographic and Spectro- 51. Pereira, V. M.; dos Santos, J. M. B. L.; Neto, A. H. C. Modeling
scopic Characterization of Electronic Edge States in CVD Disorder in Graphene. Phys. Rev. B: Condens. Matter Mater.
Grown Graphene Nanoribbons. Nano Lett. 2012, 12, 1928– Phys. 2008, 77, 115109.
1933. 52. Yan, J.-A.; Chou, M. Y. Oxidation Functional Groups on
31. Zhang, X.; Yazyev, O. V.; Feng, J.; Xie, L.; Tao, C.; Chen, Y.-C.; Graphene: Structural and Electronic Properties. Phys. Rev.
Jiao, L.; Pedramrazi, Z.; Zettl, A.; Louie, S. G.; et al. Experi- B: Condens. Matter Mater. Phys. 2010, 82, 125403.
mentally Engineering the Edge Termination of Graphene 53. Wassmann, T.; Seitsonen, A. P.; Saitta, A. M.; Lazzeri, M.;
Nanoribbons. ACS Nano 2013, 7, 198–202. Mauri, F. Clar's Theory, π-Electron Distribution, and Geo-
32. Suenaga, K.; Koshino, M. Atom-by-Atom Spectroscopy at metry of Graphene Nanoribbons. J. Am. Chem. Soc. 2010,
Graphene Edge. Nature 2010, 468, 1088–1090. 132, 3440–3451.
33. Yazyev, O. V.; Helm, L. Defect-Induced Magnetism in 54. Kaloni, T. P.; Cheng, Y. C.; Faccio, R. Oxidation of Mono-
Graphene. Phys. Rev. B: Condens. Matter Mater. Phys. vacancies in Graphene by Oxygen Molecules. J. Mater.
2007, 75, 125408. Chem. 2011, 21, 18284–18288.

FUJII AND ENOKI VOL. 7 ’ NO. 12 ’ 11190–11199 ’ 2013 11198

www.acsnano.org
 ARTICLE
55. Lee, G.; Cho, K. Electronic Structures of Zigzag Graphene
Nanoribbons with Edge Hydrogenation and Oxidation.
Phys. Rev. B: Condens. Matter Mater. Phys. 2009, 79, 165440.
56. Tersoff, J.; Hamann, D. R. Theory and Application for the
Scanning Tunneling Microscope. Phys. Rev. Lett. 1983, 50,
1998–2001.
57. Tersoff, J.; Hamann, D. R. Theory of the Scanning Tunneling
Microscope. Phys. Rev. B: Condens. Matter Mater. Phys.
1985, 31, 805–813.
58. Hummers, W. S.; Offeman, R. E. Preparation of Graphite
Oxide. J. Am. Chem. Soc. 1958, 80, 1339.
59. Fujii, S.; Enoki, T. Cutting of Oxidized Graphene into Nano-
Sized Pieces. J. Am. Chem. Soc. 2010, 132, 10034–10041.
60. Vanderbilt, D. Soft Self-Consistent Pseudopotentials in a
Generalized Eigenvalue Formalism. Phys. Rev. B: Condens.
Matter Mater. Phys. 1990, 41, 7892–7895.
61. Giannozzi, P.; Baroni, S.; Bonini, N.; Calandra, M.; Car, R.;
Cavazzoni, C.; Ceresoli, D.; Chiarotti, G. L.; Cococcioni, M.;
Dabo, I.; et al. QUANTUM ESPRESSO: A Modular and Open-
Source Software Project for Quantum Simulations of
Materials. J. Phys.: Condens. Matter 2009, 21, 395502.
62. Monkhorst, H. J. Special Points for Brillouin-Zone Integra-
tions. Phys. Rev. B: Solid State 1976, 13, 5188–5192.

FUJII AND ENOKI VOL. 7 ’ NO. 12 ’ 11190–11199 ’ 2013 11199

www.acsnano.org