Two-dimensional atomic crystals

K. S. Novoselov*, D. Jiang*, F. Schedin*, T. J. Booth*, V. V. Khotkevich*, S. V. Morozov†, and A. K. Geim*‡

*Centre for Mesoscience and Nanotechnology and School of Physics and Astronomy, University of Manchester, Manchester M13 9PL, United Kingdom;
and †Institute for Microelectronics Technology, Chernogolovka 142432, Russia

Edited by T. Maurice Rice, Swiss Federal Institute of Technology, Zurich, Switzerland, and approved June 7, 2005 (received for review April 6, 2005)

We report free-standing atomic crystals that are strictly 2D and can wafer (Fig. 1d), because even a monolayer adds up sufficiently
be viewed as individual atomic planes pulled out of bulk crystals or to the optical path of reflected light so that the interference color
as unrolled single-wall nanotubes. By using micromechanical cleav- changes with respect to the one of an empty substrate (phase
age, we have prepared and studied a variety of 2D crystals contrast). The whole procedure takes literally half an hour to
including single layers of boron nitride, graphite, several dichal- implement and identify probable 2D crystallites. Their further
cogenides, and complex oxides. These atomically thin sheets (es- analysis was done by atomic force microscopy (AFM), for which
sentially gigantic 2D molecules unprotected from the immediate single-layer crystals were selected as those exhibiting an appar-
environment) are stable under ambient conditions, exhibit high ent (12) thickness of approximately the interlayer distance in the
crystal quality, and are continuous on a macroscopic scale. corresponding 3D crystals.
Despite its simplicity, the described cleavage technique has
graphene 兩 layered material several nonobvious features that are instructive to analyze,
because it also allows one to understand why 2D crystals were

D imensionality is one of the most defining material param-

eters; the same chemical compound can exhibit dramatically
different properties depending on whether it is arranged in a 0D,
not discovered earlier (e.g., see refs 9–11, 13, and 14, in which
mechanically cleaved graphitic flakes 10–100 layers thick were
reported). First, monolayers are in a great minority among
1D, 2D, or 3D crystal structure. Although quasi-0D [e.g., cage accompanying thicker flakes. Second, unlike nanotubes, 2D
molecules (1)], quasi-1D [e.g., nanotubes (2–4)], and, of course, crystals have no clear signatures in transmission electron mi-
3D crystalline objects are well documented, dimensionality two croscopy (6–8). Third, monolayers are completely transparent to
is conspicuously absent among experimentally known crystals. visible light and cannot be seen in an optical microscope on most
On the other hand, there are many layered materials with strong substrates (e.g., on glass or metals). Fourth, AFM is currently the
in-plane bonds and weak, van der Waals-like coupling between only method that allows definitive identification of single-layer
layers. Because of this layered structure, it has long been crystals, but it has a very low throughput (especially for the case
tempting to try splitting such materials into individual atomic of the high-resolution imaging required), and in practice it would
layers, although it remained unclear whether free-standing be impossible to find cleaved 2D crystallites by scanning surfaces
atomic layers could exist in principle [thin films become ther- at random. Finally, as mentioned earlier, it was not obvious that
modynamically unstable (decompose or segregate) below a isolated atomic planes could survive without their parent crystals
certain thickness, typically, of many dozens layers]. Thus far, [for example, mechanically cleaved quasi-1D NbSe3 crystallites
most efforts have focused on chemical exfoliation of strongly ⬇100 nm in diameter were found to deteriorate rapidly (16)].
layered materials and, in particular, of stage I intercalated With the benefit of hindsight, the critical step that allowed us to
graphite (5). During exfoliation, monolayers at some moment find 2D crystallites is the discovered possibility of their tentative
must separate from each other. However, no 2D crystals have identification in an optical microscope when they are placed on
ever been isolated from the resulting slurries, possibly because top of an oxidized Si wafer.
single layers appear only as a transient state and involve detach- Representative samples of several 2D materials (namely, of
ments over microscopic regions. Indeed, the latest studies of BN, MoS2, NbSe2, Bi2Sr2CaCu2Ox, and graphite) obtained and
chemically exfoliated graphite have shown that its sediments identified by the procedures described above were investigated
consist of restacked and scrolled multilayer sheets rather than further by scanning tunneling, scanning electron, and high-
individual monolayers (6–8). An alternative approach has been resolution transmission electron microscopy (HRTEM). Fig. 2
the use of mechanical cleavage (9–14). The earlier reports shows examples of the obtained atomic-resolution images. These
described mechanically cleaved flakes consisting of tens and studies§ confirmed that the prepared 2D crystallites remained
hundreds of layers, but the recently renewed interest in thin monocrystalline under ambient conditions and no degradation
graphitic films led to flaky materials with a thickness of just a few was noticed over periods of many weeks. Within experimental
graphene layers (12–15). Now we have extended the approach to resolution, the crystal structure of isolated layers remained the
its ultimate limit: We have isolated individual crystal planes from same as for stacked layers within 3D crystals. Note that 2D
a large variety of strongly layered materials and shown that the Bi2Sr2CaCu2Ox showed a superstructure with a unidirectional
resulting 2D crystals exhibit high crystal quality and macroscopic modulation period of ⬇28 Å, which is similar to the superstruc-
continuity.
PHYSICS

Materials and Methods This paper was submitted directly (Track II) to the PNAS office.
Fig. 1 shows several examples of cleaved samples and illustrates Abbreviations: AFM, atomic force microscopy; HRTEM, high-resolution transmission elec-
that they are only one atomic layer thick but nearly macroscopic tron microscopy.
laterally. To extract such 2D crystallites, we used a simple but ‡To whom correspondence may be addressed. E-mail: [email protected].
effective procedure. A fresh surface of a layered crystal was §In the case of HRTEM studies [we used an FEI (Eindhoven, The Netherlands) Tecnai F30], the
rubbed against another surface (virtually any solid surface is cleaved material was deposited directly on holey carbon films, which made the described
suitable), which left a variety of flakes attached to it (the rubbing route of preliminary identification of 2D crystallites in an optical microscope impossible.
To find them on top of holey carbon among thicker flakes, a different procedure was
process can be described as similar to ‘‘drawing by chalk on a
developed. First, we used scanning electron microscopy imaging at low acceleration
blackboard’’). Unexpectedly, among the resulting flakes we voltages (FEI Sirion at 500 V). Then, the flakes that were found most transparent in
always found single layers. Their preliminary identification amid scanning electron microscopy were studied by AFM (i.e., directly on top of holey carbon)
thicker flakes and other residue was done in an optical micro- to define their thickness and select single-layer crystals.
scope. 2D crystallites become visible on top of an oxidized Si © 2005 by The National Academy of Sciences of the USA

www.pnas.org兾cgi兾doi兾10.1073兾pnas.0502848102 PNAS 兩 July 26, 2005 兩 vol. 102 兩 no. 30 兩 10451–10453

Fig. 1. 2D crystal matter. Single-layer crystallites of NbSe2 (a), graphite (b),
Bi2Sr2CaCu2Ox (c), and MoS2 (d) visualized by AFM (a and b), by scanning
Fig. 2. Atomic-resolution images of 2D materials. (a) Unfiltered scanning
electron microscopy (c), and in an optical microscope (d). (All scale bars: 1 ␮m.)
tunneling microscopy image of the crystal lattice in the NbSe2 monolayer on top
The 2D crystallites are on top of an oxidized Si wafer (300 nm of thermal SiO2)
of an oxidized Si wafer. Note that for the scanning tunneling microscopy mea-
(a, b, and d) and on top of a holey carbon film (c). Note that 2D crystallites were
surements, an Au film was deposited around 2D crystallites to provide an elec-
often raised by an extra few angstroms above the supporting surface, prob-
trical contact. (b) HRTEM images of the 2D Bi2Sr2CaCu2Ox crystal shown in Fig. 1c.
ably because of a layer of absorbed water. In such cases, the pleated and
(c) HRTEM image of a double-layer MoS2. This image is shown to make a con-
folded regions seen on many AFM images and having the differential height
nection between our approach based on AFM identification of 2D crystals and the
matching the interlayer distance in the corresponding 3D crystals help to
traditional HRTEM approach used for quasi-1D crystals (all nanotubes were first
distinguish between double-layer crystals and true single sheets such as those
found by using HRTEM, where dark lines indicating the nanotube’s walls parallel
shown here.
to the electron beam are easily visible). No similar signature exists for 2D crystals
(see refs. 6 – 8), and we also found it difficult to align 2D samples exactly parallel
to the electron beam. However, for two-layer crystals, their thickness is easily
ture observed in thinned samples of bulk Bi2Sr2CaCu2Ox pre- identifiable not only in AFM but also in HRTEM because of folded regions seen as
pared for HRTEM (17). two dark lines (in the case of c, the separation is ⬇6.5 Å, in agreement with the
interlayer distance in bulk MoS2). We occasionally observed short dark lines
Results and Discussion (compare with ref. 8) that might be folded monolayers, but no independent
We also investigated electrical conductivity of the selected five proof for this (e.g., by simultaneous AFM studies) has been obtained yet.
2D materials to assess their microscopic quality and macroscopic
continuity. This was done by using field-effect-transistor-like
devices such as the one shown in Fig. 3 Inset (devices were between 0.5 and 3 cm2兾Vs for different samples, in agreement
prepared by electron-beam lithography). 2D Bi2Sr2CaCu2Ox and with mobilities for the corresponding 3D crystals at room
BN were found to be highly insulating, and no induced conduc- temperature. Both 2D dichalcogenides were found to be electron
tivity was detected even at gate electric fields as high as 0.3 V兾nm
(i.e., close to the electrical breakdown of SiO2), which probably
indicates that band gaps in these 2D materials are larger than in
SiO2. We also tried annealing single-layer Bi2Sr2CaCu2Ox in
oxygen, but the crystals always remained insulating.
On the contrary, 2D graphite (graphene) and both 2D dichal-
cogenides were found to be metallic and exhibited a pronounced
electric field effect (Fig. 3). Their carrier mobilities were deter-
mined as ␮ ⫽ ␴(Vg)兾en(Vg), where e is the electron charge and
n ⬀ Vg is the carrier concentration induced by gate voltage Vg
(n ⬇ 7.2 ⫻ 1010 cm⫺2兾V for 300-nm SiO2). As seen in Fig. 3, ␴
was proportional to Vg over large intervals of n, showing that ␮
is independent of carrier concentration. Furthermore, by ex-
trapolating the experimental dependences ␴(Vg) to zero ␴, we
could determine initial (Vg ⫽ 0) concentrations of charge
carriers and their type. Graphene behaved rather similarly to
few-layer graphitic samples reported in ref. 12 and is either a
Fig. 3. Electric field effect in single-atomic-sheet crystals. Changes in electrical
shallow-gap semiconductor or a small-overlap semimetal, in conductivity ␴ of 2D NbSe2, 2D MoS2, and graphene as a function of gate voltage
which positive and negative gate voltages induce 2D electrons are shown (300 K). (Inset) Our typical devices used for such measurements: It is an
and holes, respectively, in concentrations up to ⬇1013 cm⫺2. optical image (in white light) of 2D NbSe2 on top of an oxidized Si wafer (used as
Graphene exhibited typical values of ␮ between 2,000 and 5,000 a gate electrode) with a set of Au contacts. The crystal is seen as a bluer region in
cm2兾Vs. For 2D NbSe2 and MoS2, we measured mobilities the center. (Scale bar: 5 ␮m.)

10452 兩 www.pnas.org兾cgi兾doi兾10.1073兾pnas.0502848102 Novoselov et al.

conductors with n ⬇ 1012 to 1013 cm⫺2. Detailed studies of their macroscopic continuity and such high quality that their carrier
conductivities as a function of temperature and Vg revealed that mobilities remain almost unaffected. The found class of 2D
2D MoS2 was a heavily doped semiconductor with an activation crystals offers a wide choice of new materials parameters for
gap of ⱖ0.6 eV, whereas NbSe2 was a semimetal. The found possible applications and promises a wealth of new phenomena
electron concentration in 2D NbSe2 is two orders of magnitude usually abundant in 2D systems. We believe that, once investi-
smaller than carrier concentrations per monolayer in 3D NbSe2, gated and understood, 2D crystals can also be grown in large
which indicates significant changes in the energy spectrum of sizes required for industrial applications, matching the progress
NbSe2 from a normal metal in 3D to a semimetal in 2D. achieved recently for the case of single-wall nanotubes (18).
In conclusion, we have demonstrated the existence of 2D
atomic crystals that can be prepared by cleavage from most We thank P. B. Kenway for help with transmission electron microscope
strongly layered materials. It is most unexpected, if not coun- studies. This work was supported by the Engineering and Physical
terintuitive, that isolated 2D crystals can be stable at room Sciences Research Council (United Kingdom). K.S.N. acknowledges The
temperature and in air, leaving aside the fact that they maintain Leverhulme Trust for financial support.

1. Kroto, H. W., Fischer, J. E. & Cox, D. E., eds. (1993) The Fullerenes (Pergamon, 12. Novoselov, K. S., Geim, A. K., Morozov, S. V., Jiang, D., Zhang, Y., Dubonos,
Oxford). S. V., Grigorieva, I. V. & Firsov, A. A. (2004) Science 306, 666–669.
2. Iijima, S. (1991) Nature 354, 56–58. 13. Zhang, Y., Small, J. P., Amori, M. E. S. & Kim, P. (2005) Appl. Phys. Lett. 86, 073104.
3. Chopra, N. G., Luyken, R. J., Cherrey, K., Crespi, V. H., Cohen, M. L., Louie, 14. Bunch, J. S., Yaish, Y., Brink, M., Bolotin, K. & McEuen, P. L. (2005) Nano
S. G. & Zettl, A. (1995) Science 269, 966–967. Lett. 5, 287–290.
4. Tenne, R., Margulis, L., Genut, M. & Hodes, G. (1992) Nature 360, 444–446. 15. Berger, C., Song, Z., Li, T., Li, X., Ogbazghi, A. Y., Feng, R., Dai, Z.,
5. Dresselhaus, M. S. & Dresselhaus, G. (2002) Adv. Phys. 51, 1–186. Marchenkov, A. N., Conrad, E. H., First P. N. & de Meer, W. A. (2004) J. Phys.
6. Shioyama, H. (2001) J. Mater. Sci. Lett. 20, 499–500. Chem. B 108, 19912–19916.
7. Viculis, L. M., Mack, J. J. & Kaner, R. B. (2003) Science 299, 1361. 16. Slot, E., Holst, M. A., van der Zant, H. S. J. & Zaitzev-Zotov, S. V. (2004) Phys.
8. Horiuchi, S., Gotou, T., Fujiwara, M., Asaka, T., Yokosawa, T. & Matsui, Y. Rev. Lett. 93, 176602.
(2004) Appl. Phys. Lett. 84, 2403–2405. 17. Shaw, T. M., Shivashankar, S. A., La Placa, S. J., Cuomo, J. J., McGuire, T. R.,
9. Ebbesen, T. W. & Hiura, H. (1995) Adv. Mater. 7, 582–586. Roy, R. A., Kelleher, K. H. & Yee, D. S. (1988) Phys. Rev. B Condens. Matter.
10. Ohashi, Y., Hironaka, T., Kubo, T. & Shiiki, K. (1997) Tanso 180, 37, 9856–9859.
235–238. 18. Zheng, L. X., O’Connell, M. J., Doorn, S. K., Liao, X. Z., Zhao, Y. H., Akhadov,
11. Lu, X., Huang, H., Nemchuk, N. & Ruoff, R. S. (1999) Appl. Phys. Lett. 75, E. A., Hoffbauer, M. A., Roop, B. J., Jia, Q. X., Dye, R. C., et al. (2004) Nat.
193–195. Mater. 3, 673–676.

PHYSICS

Novoselov et al. PNAS 兩 July 26, 2005 兩 vol. 102 兩 no. 30 兩 10453