You are on page 1of 3


Photoluminescence and band gap modulation in graphene oxide

Zhengtang Luo, Patrick M. Vora, Eugene J. Mele, A. T. Charlie Johnson, and
James M. Kikkawaa兲
Department of Physics and Astronomy, The University of Pennsylvania, 209 South 33rd Street,
Philadelphia, Pennsylvania 19104, USA
共Received 21 January 2009; accepted 21 February 2009; published online 19 March 2009兲
We report broadband visible photoluminescence from solid graphene oxide, and modifications of the
emission spectrum by progressive chemical reduction. The data suggest a gapping of the
two-dimensional electronic system by removal of ␲-electrons. We discuss possible gapping
mechanisms, and propose that a Kekule pattern of bond distortions may account for the observed
behavior. © 2009 American Institute of Physics. 关DOI: 10.1063/1.3098358兴

Single layer and bilayer graphene systems can exhibit a the excitation, and the reflection 共transmission兲 geometry for
remarkable diversity of phenomena, including observations s-GO corresponded to collection on the same 共opposite兲 side
of a room-temperature, unconventional quantum Hall of the film. PL spectra were excited by Xe lamp passed
effect,1–3 predictions of the quantum spin Hall effect,4 broken through a monochrometer, and additional filters were em-
spin5,6 or pseudospin7 symmetries, and finite size effects that ployed on excitation and collection to reject excitation scat-
can be used to control band structure5,6,8,9 and ter, second order grating effects, and leakage of Xe lamp
magnetism.5,6,9,10 Taken together, these properties suggest spikes. Spectra were spectrally corrected for detector effi-
that graphene is a promising platform for seamlessly ex- ciencies, and normalized by excitation power. All PL data
changing information between different degrees of freedom. shown here 共both maps and single spectra兲 are further nor-
An outstanding challenge in this regard is photonic integra- malized to a maximum value of unity and taken at 300 K.
tion and band gap manipulation. Several theoretical works Figure 1共a兲 compares PL for both l-GO and s-GO
predict that a direct gap in the visible would occur for suffi- samples. Both peaks in the visible with a long infrared emis-
ciently small graphene nanoribbons,5,6,8,9 but no observations sion tail. Differences in measurement geometry make quan-
of this finite size effect have been reported. Additionally, titative comparisons of the quantum yield impossible, but
modifications of the graphene sheet by oxidation can intro- generally little difference was seen in PL intensity. An inter-
duce direct gap behavior.11 esting question is whether energy relaxation and spectral dif-
Here we show that graphene, although intrinsically a fusion are qualitatively altered by aggregation. For isolated
zero-gap semimetal, may be oxidized in a manner that pro- flakes in l-GO, diffusion of free carriers or bound excitons
duces photoluminescence 共PL兲 for solid, drop-cast samples. should be confined to the two dimensional GO plane. How-
We find that despite the high surface area of graphene oxide ever, for s-GO, atomic force and optical microscopy, both
共GO兲 and in marked contrast to carbon nanotubes, the indicate films of layered GO flakes, which could give rise to
strength of PL from GO flakes does not differ significantly additional interflake relaxation pathways. If interlayer cou-
between aqueous and drop-cast samples. The resilience of pling is strong enough, the emission spectrum could redshift.
PL for solid GO samples is encouraging for technological s-GO indeed shows more PL spectral weight in the infrared,
applications, implying that GO may be a useful photonic but the redshift in the PL peak position is not a robust feature
material when incorporated in solid state devices. The large of the experiment and was inconsistent from sample to
sample perhaps due to variations in the oxidation density. In
observed gap creates the possibility for spatially modulating
addition to exciton diffusion, several plausible changes could
the band structure within a single graphene flake by local
control of the oxidation profile. Studies of progressive
chemical reduction show quenching of PL for both drop-cast
and aqueous samples, coordinated with changes in absorp- 1 (a) (b)
tion. These studies also find signatures of band gap manipu- l-GO l-GO
Normalized PL (arb. units)
Normalized PL (arb. units)

1 1
s-GO s-GO
lation, albeit with different character for solid and liquid

Aqueous dispersions of single layer GO with an average
area of ⬃100 ␮m2 were synthesized following a procedure
described elsewhere.12 Solid samples 共s-GO兲 were obtained
by drop-casting the concentrated GO solution resulting from
this procedure onto polished, low auto-fluorescence, s-GO
Suprasil-2 substrates and then baking at 95 ° C for 30 min. 0 0 0
600 800 1000 1200 400 500 600
Liquid samples 共l-GO兲 were held in quartz cuvettes, diluted a Wavelength (nm) Wavelength (nm)
hundredfold or more as necessary to adjust optical density.
PL for both l-GO and s-GO was collected at 90° degrees to FIG. 1. 共Color online兲 共a兲 Normalized PL spectra excited at 500 nm 共s-GO
taken in transmission兲. 共b兲 Absorption 共left axis, solid兲 and PL intensity
detected at 752 nm 共right axis, dashed兲, as a function of excitation
Electronic mail: wavelength.

0003-6951/2009/94共11兲/111909/3/$25.00 94, 111909-1 © 2009 American Institute of Physics

Downloaded 30 Oct 2010 to Redistribution subject to AIP license or copyright; see
111909-2 Luo et al. Appl. Phys. Lett. 94, 111909 共2009兲

1 (a) 1.0 0s (a)

Normalized PL (arb. units)

1 min
0 sec 0.8
8 min
Excitation Wavelength (nm)

18 min
500 0.6

30 sec 0.2

600 0.0
(c) 600 800 1000 1200 1400
Wavelength (nm)

120 sec 1.0 (b)

Normalized PL (arb. units)

800 1000 1200 1400
Emission Wavelength (nm) 0.6 700-750nm

FIG. 2. 共Color兲 Normalized PL excitation-emission maps for s-GO taken in 0.5 0 min
transmission during hydrazine vapor exposure. 3 min
22 min
0.2 62 min
0 60 120 121 min
also influence the peak position, including modifications of 0.0
t (min)

the dielectric environment, spectral reabsorption, and varia- 600 700 800 900 1000
tions in oxidation density. Wavelength (nm)

The GO absorption 关Fig. 1共b兲兴 increases with energy FIG. 3. 共Color online兲 Normalized PL spectra during hydrazine exposure for
from the near infrared to 3.1 eV, and provides an interesting 共a兲 s-GO in a reflection geometry and 共b兲 l-GO, excited at 488 nm. Data in
contrast to the nonmonotonic PL spectrum. We further note 共a兲 have been corrected for reabsorption as described in the text, whereas 共b兲
that PL excitation 共PLE兲 spectra do not mirror the absor- does not require a correction. 共Inset兲 Changes in integrated PL intensity I共t兲
bance increases at higher energies. To the contrary, as shown as a function of cumulative hydrazine exposure t for l-GO. Legend indicates
spectral regions over which I共t兲 was integrated.
in Fig. 1共b兲, as the excitation energy increases, PL intensity
detected at a fixed wavelength decreases. The latter indicates
the presence of nonradiative energy relaxation pathways, but tiple reflections are ignored. IPL shows a significant redshift
also calls into question the relevance of the absorbance spec- in the PL emission for increased exposure times 关Fig. 3共a兲兴,
trum to the emissive process. With these data in mind, one consistent with an increase in the disorder length scale upon
must consider the possibility that the electronic structure reduction.
within the flakes is heterogeneous, and that PL emission We also studied chemical reduction in l-GO, where re-
originates from absorbance into excited states whose total absorption effects could be brought to a negligible level.
absorptive cross-section is, nevertheless, only a small con- l-GO was reduced and held in suspension following the pro-
tributor to the total absorptive spectrum. Absolute values for cedure introduced by Li et al.,13 which may involve electro-
the emissive quantum yield would help to illuminate this static stabilization. During reduction, l-GO transmission re-
discussion, but accurate measurements of this quantity are mained ⬎80% relative to water for a 1 cm path length. To
problematic for ensembles of heterogeneous nanomaterials further minimize reabsorption, l-GO was excited within 1
and are beyond the scope of this work. Nevertheless we may mm of the collection window. Figure 3共b兲 shows the devel-
qualitatively say that the observed quantum yields appear to opment of a flatter spectral profile as a function of reduction
be far less than unity. time. Currently, differences between s-GO and l-GO reduc-
To study the relationship between oxidation density and tion studies are not understood, but the data do hold certain
the optical gap, we performed experiments to vary the oxi- features in common. In all cases, the relative intensity of PL
dation density and while monitoring changes in the PL spec- in the infrared increases. However, the absolute intensity of
trum. s-GO samples were placed in a covered Petri dish with PL was always seen to decrease with reduction everywhere
a container of hydrazine, heated to 50 ° C, and their PL maps in the spectrum, including the infrared 关Fig. 3共b兲 inset兴.
characterized in the transmission geometry. The process was A theoretical framework for interpreting these data is
repeated several times to progressively reduce the sample, only just emerging. It can be expected that oxidation pro-
resulting in a marked redshift 共Fig. 2兲. Modeling shows that duces a disruption of the ␲ network and can open a direct
changes in the absorptive spectrum 共not shown兲 could also electronic band gap for single sheet graphene in one of two
produce these spectral shifts through reabsorption of the ways. The first is a quantum confinement effect whereby the
emitted light. To better control these effects, we repeated this ␲-electron wave functions occupy a potential landscape with
study on a different sample in the reflection geometry, while strongly repulsive hard wall barriers at oxidized sites. In the
also continually monitoring changes in absorption. Penetra- infinite potential limit, a delocalized ␲-electron wave func-
tion of the exciting and emitting light was then estimated tion will develop nodes at each of these sites. The presence
using the relationship Imeas共1 − T1兲ln共T1T2兲 = IPL共1 or absence of a gap for a sample with many such oxidized
− T1T2兲ln T1, where T1 共T2兲 is the film transmission for the sites then depends on the spatial distribution of these nodes.
excitation 共emission兲 wavelength, Imeas and IPL are the raw For example, in graphene ribbons where the edges break the
and absorbance-corrected intensities, respectively, and mul- sublattice symmetry, the lateral confinement of the wave
Downloaded 30 Oct 2010 to Redistribution subject to AIP license or copyright; see
111909-3 Luo et al. Appl. Phys. Lett. 94, 111909 共2009兲

function produces a band gap at its charge neutrality level. In summary, we demonstrate that s-GO emits PL, and
Alternatively, for the special edge that preserves sublattice that GO PL can be altered by chemical reduction. Broad PL
symmetry 共a “zig-zag” ribbon兲 one finds instead a resonant suggests a dispersion of hard gaps, which may arise from
electronic state exactly at zero energy. These results general- bond alternation within the GO plane giving rise to interval-
ize to a disordered potential landscape where the nodes occur ley scattering. The loss of quantum yield during reduction in
in the interior of the sample. Pereira et al.14 modeled the our experiments suggests that some regions might remain
effect of lattice vacancies in a single valley picture, and heavily oxidized. In this case, restoration of electrical con-
found that a hard gap opens only in the special situation ductivity resembles a percolation problem and gives very
where there is complete sublattice asymmetry in the vacancy little information about the growth of these regions below the
distribution. For intermediate asymmetries where unequal percolation threshold. Future studies might use complemen-
fractions of the defects reside on different sublattices, they tary methods such as Raman scattering,21 infrared Landau
find a finite, but reduced, density of states 共a soft gap兲 in an level spectroscopy,22,23 or magnetic anisotropy24 to quantify
energy interval of order បvF冑n, where n is the vacancy den- the emergence of graphenelike regions in reduced GO. Op-
sity and vF is the Fermi velocity. In the limit of full sublattice tical anisotropy measurements24 may help to assess the im-
symmetry, they find a resonant state at the charge neutrality portance of charge transfer transitions, which we have ex-
point. Usually the observation of PL implies a hard gap, cluded from the discussion here and should have markedly
since a soft gap permits nonradiative energy relaxation un- different optical polarization anisotropies when compared to
less a peculiar bottleneck exists. Moreover, despite a lack of transitions native to the two-dimensional GO plane.
consensus as to the structural motif in GO,15–17 there is not
P.V. and J.K. supported by NSF MRSEC under Grant
yet any observation or calculation suggesting a complete
No. DMR05–20020, Z.L. and A.J. supported by the JSTO
sublattice asymmetry in the oxidation profile.
DTRA and ARO under Grant No. W911NF-06–1-0462, and
A second gapping mechanism arises when one considers,
E.M. supported by DOE under Grant No. DE-FG02-
in addition, the effects of intervalley scattering from the short ER45118.
range potential of the oxidized carbons. Intervalley scattering
produces a coherent superposition of Bloch waves near the K 1
K. Novoselov, A. Geim, S. Morozov, D. Jiang, M. Katsnelson, I. Grig-
and K⬘ points of the Brillouin zone, giving rise to a 冑3 orieva, S. Dubonos, and A. Firsov, Nature 共London兲 438, 197 共2005兲.
⫻ 冑3 modulation of the charge density, which has been im-
K. Novoselov, E. McCann, S. Morozov, V. Fal’ko, M. Katnelson, U.
Zeitler, D. Jiang, F. Schedin, and A. Geim, Nat. Phys. 2, 177 共2006兲.
aged by scanning tunnel microscope.18 For a bond-centered 3
Y. Zhang, Y. Tan, H. Stormer, and P. Kim, Nature 共London兲 438, 201
scattering potential, this describes a modulation 共alternation兲 共2005兲.
of the bond charge density and a self-consistent potential 4
C. L. Kane and E. J. Mele, Phys. Rev. Lett. 95, 226801 共2005兲.
with this spatial symmetry. The amplitude of this modulation K. Nomura and A. H. MacDonald, Phys. Rev. Lett. 96, 256602 共2006兲.
is determined by the scattering strength, direct and exchange Y.-W. Son, M. Cohen, and S. Louie, Nature 共London兲 444, 347 共2006兲.
M. Katsnelson, K. Novoselov, and A. Geim, Nat. Phys. 2, 620 共2006兲.
electron-electron interactions, and the electron-phonon cou- 8
L. Brey and H. Fertig, Phys. Rev. B 73, 235411 共2006兲.
pling. The latter is likely to produce some degree of bond 9
K. Kusakabe and M. Maruyama, Phys. Rev. B 67, 092406 共2003兲.
alternation in the graphitic regions near an oxidized site, N. M. R. Peres, F. Guinea, and A. H. C. Neto, Phys. Rev. B 73, 125411
which might even template further oxidation in such a way 共2006兲.
X. Sun, Z. Liu, K. Welsher, J. T. Robinson, A. Goodwin, S. Zaric, and H.
as to reinforce this pattern of distortions. In this way a Dai, Nano Res. 1, 203 共2008兲.
“Kekule pattern” emerges naturally in the electronic poten- 12
Z. Luo, Y. Lu, L. A. Somers, and A. T. C. Johnson, J. Am. Chem. Soc.
tial, and provides a spatially modulated intervalley gap pa- 131, 898 共2009兲.
rameter of the form envisioned by Hou et al.19 Note that
D. Li, M. B. Mueller, S. Gilje, R. B. Kaner, and G. G. Wallace, Nat.
bond alternation in conjugated polymers such as polyphene- Nanotechnol. 3, 101 共2008兲.
V. M. Pereira, J. M. B. L. dos Santos, and A. H. C. Neto, Phys. Rev. B 77,
lyne vinylene gives rise to gaps of a similar energy scale.20 115109 共2008兲.
Within a context of bond disorder induced energy gaps, 15
W. Cai, R. D. Piner, F. J. Stadermann, S. Park, M. A. Shaibat, Y. Ishii, D.
one regards the GO plane as a landscape containing a wide Yang, A. Velamakanni, S. J. An, M. Stoller, J. An, D. Chen, and R. S.
range of local band gap minima. This notion is consistent Ruoff, Science 321, 1815 共2008兲.
H. He, J. Klinowski, M. Forster, and A. Lerf, Chem. Phys. Lett. 287, 53
with the very broad range of observed emission energies. If 共1998兲.
the reduction procedure for l-GO does indeed prevent aggre- 17
T. Szabo, O. Berkesi, P. Forgo, K. Josepovits, Y. Sanakis, D. Petridis, and
gation, as claimed in Ref. 13, then the observation of PL I. Dekany, Chem. Mater. 18, 2740 共2006兲.
quenching across the entire spectrum for the liquid sample P. Mallet, F. Varchon, C. Naud, L. Magaud, C. Berger, and J.-Y. Veuillen,
Phys. Rev. B 76, 041403 共2007兲.
indicates that interflake or substrate contact is not vital for 19
C.-Y. Hou, C. Chamon, and C. Mudry, Phys. Rev. Lett. 98, 186809
quenching of PL. Neither theoretical framework discussed 共2007兲.
above accounts for a loss of quantum yield unless there is J. H. Burroughes, D. D. C. Bradley, A. R. Brown, R. N. Marks, K.
significant inhomogeneity of the oxidation profile within the Mackay, R. H. Friend, P. L. Burns, and A. B. Holmes, Nature 共London兲
347, 539 共1990兲.
GO plane. If one imagines that chemical reduction results in 21
A. C. Ferrari, J. C. Meyer, V. Scardaci, C. Casiraghi, M. Lazzeri, F. Mauri,
the nucleation and/or growth of regions where GO is fully S. Piscanec, D. Jiang, K. S. Novoselov, S. Roth, and A. K. Geim, Phys.
reduced to graphene, then migration of nonequilibrium car- Rev. Lett. 97, 187401 共2006兲.
riers to these zero gap regions could provide an efficient M. Dresselhaus and G. Dresselhaus, Adv. Phys. 30, 139 共1981兲.
route for nonradiative recombination. In this picture, there is P. Plochocka, C. Faugeras, M. Orlita, M. L. Sadowski, G. Martinez, M.
Potemski, M. O. Goerbig, J.-N. Fuchs, C. Berger, and W. A. de Heer,
no simple relationship between the oxidation density and the Phys. Rev. Lett. 100, 087401 共2008兲.
length scale between oxidation sites within gapped, oxidized 24
O. N. Torrens, D. E. Milkie, H. Y. Ban, M. Zheng, G. B. Onoa, T. D.
regions. Gierke, and J. M. Kikkawa, J. Am. Chem. Soc. 129, 252 共2007兲.

Downloaded 30 Oct 2010 to Redistribution subject to AIP license or copyright; see