ARTICLE

Received 29 Nov 2013 | Accepted 3 Jul 2014 | Published 7 Aug 2014 DOI: 10.1038/ncomms5576

Edge and confinement effects allow in situ

measurement of size and thickness
of liquid-exfoliated nanosheets
Claudia Backes1,2, Ronan J. Smith1,2, Niall McEvoy1,3, Nina C. Berner1,3, David McCloskey1,2, Hannah C. Nerl1,2,3,
Arlene O’Neill1,2, Paul J. King1,2, Tom Higgins1,2, Damien Hanlon1,2, Nils Scheuschner4, Janina Maultzsch4,
Lothar Houben5, Georg S. Duesberg1,3, John F. Donegan1,2, Valeria Nicolosi1,2,3 & Jonathan N. Coleman1,2

Two-dimensional nanomaterials such as MoS2 are of great interest both because of their
novel physical properties and their applications potential. Liquid exfoliation, an important
production method, is limited by our inability to quickly and easily measure nanosheet size,
thickness or concentration. Here we demonstrate a method to simultaneously determine
mean values of these properties from an optical extinction spectrum measured on a liquid
dispersion of MoS2 nanosheets. The concentration measurement is based on the
size-independence of the low-wavelength extinction coefficient, while the size and thickness
measurements rely on the effect of edges and quantum confinement on the optical spectra.
The resultant controllability of concentration, size and thickness facilitates the preparation of
dispersions with pre-determined properties such as high monolayer-content, leading to first
measurement of A-exciton MoS2 luminescence in liquid suspensions. These techniques are
general and can be applied to a range of two-dimensional materials including WS2, MoSe2
and WSe2.

1 Centre for Research on Adaptive Nanostructures and Nanodevices (CRANN), Trinity College Dublin, Dublin 2, Ireland. 2 School of Physics, Trinity College

Dublin, Dublin 2, Ireland. 3 School of Chemistry, Trinity College Dublin, Dublin 2, Ireland. 4 Institut für Festkörperphysik, Technische Universität Berlin,
Hardenbergstrae 36, 10623 Berlin, Germany. 5 Ernst Ruska Center for Microscopy and Spectroscopy with Electrons, Research Center Jülich, 52425 Jülich,
Germany. Correspondence and requests for materials should be addressed to J.N.C. (email: colemaj@tcd.ie).

NATURE COMMUNICATIONS | 5:4576 | DOI: 10.1038/ncomms5576 | www.nature.com/naturecommunications 1

& 2014 Macmillan Publishers Limited. All rights reserved.
ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms5576

O
ver the last few years the study of two-dimensional (2D) microscopic analysis, we generate quantitative relationships
nanomaterials has become an important area of between spectral properties and nanosheet dimensions. Quantify-
nanoscience1–4. The palette of 2D materials currently ing these effects allows an extinction spectrum to be used to
under study has expanded from graphene3,5, to include transition obtain not only the nanosheet concentration but also the
metal dichalcogenides (TMDs) such as MoS2 and WSe2, layered nanosheet length and thickness. The ability to control size and
transition metal oxides such as MnO2 and TiTaO5 and a host of thickness is very useful, facilitating for example the preparation of
other interesting structures such as GaS, germanane and Bi2Te3 monolayer-rich dispersions. Critically, we find this methodology
(refs 1,2,4,6) These materials are diverse, including metals, to be general to a range of 2D materials.
semiconductors, insulators or superconductors1,4,7, and display
a range of interesting properties from thickness-dependent
bandgaps4 to catalytic activity1. These properties make them Results
useful for both fundamental studies and applications in areas as Size selection of MoS2 nanosheets. To produce dispersions of
diverse as optoelectronics, electrochemistry and medicine. MoS2 nanosheets with varying thickness and lateral size dis-
While 2D materials can be grown directly8, they are generally tributions from the same stock sample, we used a simple cen-
obtained by exfoliation of layered crystals. This has long been trifugation technique based on band sedimentation (BS, Fig. 1a).
possible by mechanical exfoliation, leading to samples of very A stock dispersion of the surfactant-exfoliated29 nanomaterial
high quality, but at very low throughput9–11. When larger was layered on top of a race layer of higher density and subjected
production rates are required, chemical methods12–15, often to a short (typically 10–40 min) centrifugation using a normal
based on ion intercalation, have been used to exfoliate layered benchtop centrifuge. The centrifugation leads to the spreading of
crystals in liquids2. However, even though these methods result in the material through the vial such that nanosheet mass increases
large quantities of monolayers, they are time consuming, going from top to bottom39. Fractionation then allows the
involving harsh chemical treatments, which must be carried out collection of size-selected samples. Similar techniques have
in an inert atmosphere. previously been described in literature to sort carbon
More recently, a much simpler method, liquid phase exfolia- nanotubes40,41, graphene oxide nanosheets42 and nano-
tion (LPE), has been reported16–27. This method involves the particles43 by their lateral dimensions. While the higher density
sonication18 or shearing27 of layered crystals in certain solvents or of the race layer is normally achieved using a density gradient
solutions of surfactants or polymers. It results in the exfoliation of medium, we use a combination of water and deuterated water to
the layered crystal to give large numbers of 2D nanosheets, which minimize the additive content and avoid potential interactions
are stabilized by interaction with the liquid. The resultant between gradient medium and MoS2.
dispersions can be easily processed into films, coatings or After the BS centrifugation, B6 mm deep fractions were
composites: systems which are ideal for applications in a range collected from the vial, where the smallest, slowest sedimenting
of areas from batteries28,29 to photodetectors30 to reinforced species are found in the top fraction (F1). We measured
materials.21,25,31 This method is general and has been used to give extinction spectra for each fraction. To distinguish it from the
exfoliated dispersions of graphene16,17,20, BN,18,23,25 TMDs such true absorbance, Abs, we will refer to the extinction as Ext, where
as MoS2 and WS218,19,24,26,29,32, transition metal oxides such as T ¼ 10  Ext (T is the optical transmittance). The normalized
MnO2 (ref. 29) and MoO3 (ref. 33), as well as more exotic extinction spectra of a typical set of fractions (Fig. 1b) display the
structures such as Ti3C2F2 nanosheets34 or functionalized-layered excitonic transitions typical of MoS2 (ref. 7). However, we find
double hydroxides35. In fact, we expect this method to be that they show very distinct and clear changes that obviously
applicable to any layered compound, where the layers are bound reflect the different size distributions of the MoS2 nanosheets in
predominately by van der Waals interactions. the dispersion (Supplementary Fig. 1). The extinction-weighted
While LPE is a versatile and useful technique, it has some sum of the individual extinction spectra of the fractions matched
notable disadvantages. First, optical characterization of nanosheet well to the stock dispersion (Supplementary Fig. 2) confirming
dispersions is hampered by the presence of a scattering that the changes in the extinction spectra indeed reflect the
background, which depends sensitively on the size of the separation process.
dispersed nanosheets18,24. As such, it is not clear how to define In Figure 1b, we see very pronounced spectral changes from
an intrinsic extinction coefficient, making even concentration fraction to fraction. These changes were quantified by plotting the
measurements challenging. ratio of extinction at the B-exciton peak (605 nm) to that at the
Moreover, LPE gives nanosheets, which are polydisperse both local minimum at 345 nm, ExtB/Ext345 (Fig. 1c). We observe a
in thickness and lateral size. While lateral size separation is considerable change with a fourfold increase in this ratio
possible24,36, separation by thickness is more challenging. In occurring from F1 to F7. Since no such observation has been
addition, measurement of nanosheet size is tedious, requiring made when exploring the extinction spectra of films of restacked
time-consuming statistical microscopic analysis. While chemically exfoliated MoS2 of different nanosheet thicknesses14,
approximate in situ measurements of lateral size can be made this strongly suggests that the spectral changes shown in Fig. 1b,c
optically37, no in situ method for nanosheet thickness are related to changes in nanosheet lateral size. This implies that
measurement exists. To measure thickness, it is necessary to the extinction spectra contain quantitative information on the
deposit nanosheets on surfaces, avoiding aggregation and use lateral dimensions of the nanosheets.
statistical Raman or atomic force microscopy (AFM)38. This can In addition to the obvious spectral changes, it is clear that the
be very time consuming if a mean over many nanosheets is position of the A-exciton, lA, varied from fraction to fraction.
required. Ultimately, a simple method to separate nanosheets by Because the extinction spectra for MoS2 dispersions include a
size and thickness would be very useful. More importantly, a fast contribution from a size-dependent scattering background18, we
in situ method to measure both lateral size and thickness of found lA using the second derivative of the extinction spectra
nanosheets is urgently required. (Supplementary Fig. 3). The resultant peak position is plotted in
Here we address these points. We demonstrate size-selection Fig. 1d, showing a clear trend versus fraction number (Fig. 1d).
techniques which give fractions with optical extinction spectra This is consistent with red-shifting of the peak for thicker
which vary systematically with both nanosheet size and thickness. nanosheets (that is, in higher fractions) such that lA increases as
By combining the spectroscopic analysis with statistical the number of layers per nanosheet, N, increases. Such behaviour

2 NATURE COMMUNICATIONS | 5:4576 | DOI: 10.1038/ncomms5576 | www.nature.com/naturecommunications

& 2014 Macmillan Publishers Limited. All rights reserved.
NATURE COMMUNICATIONS | DOI: 10.1038/ncomms5576 ARTICLE

a b 2
F1 F1 F1
Large and thick
F2

Sedimentation velocity
F2 F2
F3
F3 F3

Ext (a.u.)
B A F4
F4 F4 F5
1
F5 F5 F6
F7
F6 F6 Small
and thin
F7 F7
0
300 450 600 750 900
Wavelength (nm)
c d 680
1.2

675
ExtB/Ext345

0.8

A (nm)
670

0.4
665
Nanosheet size Nanosheet size
0.0 660
1 2 3 4 5 6 7 1 2 3 4 5 6 7
Fraction Fraction

Figure 1 | MoS2 band sedimentation and size-dependent extinction spectra. (a) Experimental setup for band sedimentation centrifugation involving
layering a nanomaterial stock dispersion on top of a race layer. Centrifugation leads to spreading of the material throughout the vial and separation of the
nanosheets according to their sedimentation velocities and so their size. Fractions as indicated are withdrawn from top to bottom. (b) Extinction spectra of
the fractions normalized to the local minimum at 345 nm. The positions of the A- and B-excitons are marked. (c) Ratio of extinction at B-exciton to that at
345 nm, ExtB/Ext345 plotted versus fraction number. (d) Peak position (wavelength) of the A-exciton, lA, as a function of fraction number. We propose that
lA and ExtB/Ext345 can be used as metrics for flake thickness and length, respectively.

has been observed for a number of TMDs and is due to indicates that, even though scattering is always present,
confinement effects14,44,45. This implies that, in addition to lateral information encoded in an absorption spectrum can be
size information, the extinction spectra contain quantitative effectively extracted from the extinction spectrum.
information on the thickness of the nanosheets.
The effect of nanosheet length on optical spectra. To test
Differentiating absorbance from scattering. However, before whether the extinction spectrum contains useful information
any quantitative information can be extracted from the extinction about nanosheet length, we characterized a number of fractions
spectra, it is important to realize that such spectra include con- using transmission electron microscopy (TEM). Representative
tributions from both absorbance and scattering. In fact, scattering TEM images of dispersions which are expected to contain large
is known to be significant for dispersions of nanosheets18, and small nanosheets are depicted in Fig. 3a–d. These images
especially those with large lateral size24. To assess the effects of clearly show the expected size differences. To quantify these
scattering, we prepared a set of MoS2 dispersions with a range of effects, we performed statistical analysis of nanosheet lengths for
flake sizes. The concentration was measured by filtration and both band-sedimented dispersions and dispersions prepared by
weighing to allow the calculation of extinction (as well as conventional centrifugation at different rates (Supplementary
absorption and scattering) coefficients (Supplementary Fig. 4). Figs 5–7). We note that BS has the clear advantage of producing
Extinction spectra were measured and converted to extinction much narrower distributions (Supplementary Fig. 7). However,
coefficient (e) spectra as shown in Fig. 2a (N.B. Ext ¼ eCl, where C homogenous centrifugation results in larger volumes and so lar-
is the concentration and l is the cell length). We used an ger quantities of separated materials. We found that the fractions
integrating sphere to differentiate the contributions of absorbance described above contained nanosheets with mean lengths which
and scattering, leading to the absorption (a) and scattering (s) varied from B60 (F1) to 360 nm (F6), although considerably
coefficient spectra shown in Fig. 2b,c (N.B. e(l) ¼ a(l) þ s(l)). It larger flakes (mean length up to 600 nm) can be made by
is clear from these spectra that scattering makes a significant adjusting BS conditions (that is, centrifuging at 2k r.p.m., 10 min).
contribution to the extinction spectra. For l4700 nm, spl  n, Before discussing the possibility of developing a length metric,
with n roughly in the range 1–4 as suggested previously18,24,46. it is worth investigating how the extinction, absorbance and
However, for lo700 nm, s(l) is broadly similar in shape to a(l). scattering coefficients depend on nanosheet length. Interestingly,
This is important because, as a result, a(l) and e(l) are also the data in Fig. 2a suggests that when l ¼ 345 nm, e is relatively
similar in shape. This means that any metrics extracted from an invariant with fraction number and so with L. In Fig. 3e, we plot
extinction spectrum will reflect the information contained in the e, a and s (all at l ¼ 345 nm) versus L. We find e at l ¼ 345 nm to
absorption spectrum. To test this, we extracted lA from both be dominated by absorption and to be invariant with length with
extinction and absorbance spectra, plotting one versus another a mean of /eSE69 ml mg  1 cm  1 (Supplementary Fig. 8). This
for a range of fractions as shown in Fig. 2d. We also plotted ExtB/ allows extinction spectra to be used to estimate the nanosheet
Ext345 versus AbsB/Abs345 in Fig. 2e. For all but the largest concentration for dispersions of any flake length (for Lo300 nm
fractions (that is, F6 and F7), we find good scaling between the at least). Previously, this was not possible because the influence of
values extracted from extinction and absorption spectra. This the size-dependent scattering background was unknown18,24.

NATURE COMMUNICATIONS | 5:4576 | DOI: 10.1038/ncomms5576 | www.nature.com/naturecommunications 3

& 2014 Macmillan Publishers Limited. All rights reserved.
ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms5576

To assess the possibility of using ExtB/Ext345 as a metric for increase in ExtB/Ext345 with increasing L, indicating that this ratio
length, we plot this parameter versus mean nanosheet length, L, can be used as a metric for L. The ratio AbsB/Abs345, calculated
for a range of size-selected dispersions in Fig. 3f. We see a clear from the true absorbance spectra follows the same trend in line
with the data in Fig. 2e.
We propose that these size-dependent spectral changes reflect
the effect of flake edges on the local electronic structure, density
100 F1 of states and hence local optical absorption coefficient of the
F3 nanosheets47. This effect may be due to structural relaxations or
 (ml mg–1 cm–1)

80
F4 the misalignment of stacked monolayer edges. In any case, the
60 F6 observed absorption spectrum will be the sum of edge and centre
contributions. Because the population ratio of edge to centre
40 atoms increases with decreasing nanosheet size, we expect the
spectral shape to be size dependent. We can model this behaviour
20 by treating the flakes as consisting of edge (E) and centre (C)
Extinction regions (Fig. 3f inset), each of which has a well-defined
0 absorption coefficient (that is, aE and aC). Modelling the
nanosheets as arbitrary 2D shapes with long dimension, L, and
aspect ratio, k (k ¼ length/width, Supplementary Note 1), we can
100 F1 take the area-weighted sum of centre and edge absorption
F3 coefficients to derive an expression for the average effective
 (ml mg–1 cm–1)

80
F4 nanosheet extinction coefficient:
60 F6  
2x Da
a  aC 1 þ ðk þ 1Þ ð1Þ
40 L aC
where Da ¼ aE  aC and x is the width of the edge region.
20
Absorbance
To test this, we plot the absorbance coefficient at both the
0 A- and B-exciton versus the inverse nanosheet length in Fig. 3g.
We observed good linearity as predicted by equation (1) so long
as L460 nm, below which saturation occurs, consistent with
50 Scattering F1 small-flake absorbance being dominated by edges. The fit implies
F3 values of aC of 31±1 and 38±2 ml mg  1 cm  1 for the A-
 (ml mg–1 cm–1)

40 F4 and B-excitons, respectively, while saturation implies equivalent

F6 values for aE of 11±1 and 17±1 ml mg  1 cm  1. Because TEM
30 analysis shows kE2 for all nanosheet lengths (Supplementary
Fig. 5), we can use the fits, coupled with equation (1), to estimate
20
x ¼ 12±7 nm. This value is larger than would be expected for
10 effects solely due to electronic edge states, suggesting more
complex phenomena to be at work. However, it is worth noting
0 that scanning tunnelling microscopy measurements on MoS2 on
300 400 500 600 700 800 graphite have shown a distinct edge region at least 5 nm thick48.
Wavelength (nm) Nevertheless, the effects described here may not be universal but
perhaps specific to nanosheets surrounded by a weakly
interacting environment (for example, surfactant micelle or
678 1.2 F6 solvent solvation shell).
F6
Extinction A (nm)

To confirm the presence of a distinct edge region, X-ray

photoelectron spectroscopy (XPS) was performed on a range of
ExtB/Ext345

0.8 films prepared from nanosheets of different sizes. No systematic

670
changes in the Mo core-level spectra (Supplementary Fig. 9) were
0.4 detected. However, in addition to the expected 2H component of
the S2p core-level spectra, as long as we investigate freshly
662 F1 F1 prepared samples, we observe an extra component that cannot be
0.0 identified with known morphologies, for example 1 T MoS2
660 668 676 0.0 0.4 0.8 1.2 (Fig. 3h) (also Supplementary Fig. 10). Critically, the intensity of
Abs A (nm) AbsB/Abs345 this additional component increases with decreasing flake length
allowing us to associate it with the edge region. Then, the fraction
Figure 2 | Extinction, absorbance and scattering spectra. (a) Wavelength
of the S2p signal associated with this component will equal the
dependence of optical extinction coefficient, e, for fractions F1, F3, F4 and
fraction of S atoms associated with the edge region; NS,E/NS,T.
F6. NB e is defined by T ¼ 10  eCl where T is the transmittance, C is the
This ratio is plotted as a function of flake length in Fig. 3i. The
MoS2 concentration and l the cell length. e can be split into contributions
fraction of S atoms within a distance x from the edge is given by
from absorption, a, and scattering, s, such that e ¼ a þ s. (b,c) Wavelength
  
dependence of (b) optical absorption coefficient, a, and (c) optical NS;E 2x 2x
scattering coefficient, s. (d) A-exciton position measured from the ¼ 1þk 1 ð2Þ
NS;T L L
extinction spectra (Fig. 2a) plotted versus that measured from the
absorption spectra (Fig. 2b) for fractions F1–F6. (e) ExtB/Ext345 (measured This expression describes the data in Fig. 3i rather well, giving a
from the extinction spectra, Fig. 2a) plotted versus AbsB/Abs345 (measured value of x ¼ 2.7±0.5 nm. We note that this value for x is lower
from the absorption spectra, Fig. 2b) for fractions F1–F6. than the one quoted above, suggesting that XPS probes edge

4 NATURE COMMUNICATIONS | 5:4576 | DOI: 10.1038/ncomms5576 | www.nature.com/naturecommunications

& 2014 Macmillan Publishers Limited. All rights reserved.
NATURE COMMUNICATIONS | DOI: 10.1038/ncomms5576 ARTICLE

properties which are distinct to those seen by absorbance ratio for a number of spectra, measured at a range of positions on
measurements. three different nanosheets, as a function of distance from
To further illustrate differences in electronic structure between nanosheet edge as shown in Fig. 3m. We took care to choose
edge and centre of MoS2 nanosheets, we have performed scanning regions of uniform thickness to exclude effects from thickness
TEM (STEM) imaging (Fig. 3j–k) and low-loss STEM electron variations in the EELS. This data clearly shows I2.8 eV/I6.4 eV to fall
energy-loss spectroscopy (EELS, Supplementary Figs 11 and 12). off with increasing distance from the edge before saturating for
Moving inward from the nanosheet edge (Fig. 3k), the EEL spectra collected at a distance of 48 nm from the flake edge. This
spectra (Fig. 3l) exhibited gradual changes in the energy-loss suggests the presence of a distinct edge region with thickness of
regions between 2 and 10 eV. These changes reflect differences in xB8 nm in good agreement with the optical measurements
electronic structure near the flake edge compared with the centre. (taking into account limitations to the spatial resolution).
We examined these changes via the ratio of EELS intensities at However, this result has to be interpreted with caution, since in
two energies, for example, 2.8 and 6.4 eV, I2.8 eV/I6.4 eV. These this region of the low-loss spectra the exact origin of the peaks
specific positions were chosen, as they show the largest changes in remains uncertain.
the spectra with distance from the edge. We have calculated this The strong effect of nanosheet edges on electronic properties is
also consistent with recent reports, which show MoS2 in the
vicinity of edges48,49 or grain boundaries50 to have electronic
a F1 b F1 c F6 d F6 properties distinct from the central regions. Incidentally, this
suggests that this approach could be used to measure grain size in
chemical vapour deposition-grown MoS2.

Length metric based on nanosheet edge effects. We can use our

model to understand the size-dependent changes to the shape of
absorption spectra. We represent the spectral shape by the ratio of
absorbance intensities at two distinct wavelengths, l1 and l2.
2H MoS2 S 2p
100 345 nm Using equation (1), we can express this as
 Edge S
(ml mg–1 cm–1)


20 nm Absðl1 Þ aC ðl1 ÞL þ 2xðk þ 1ÞDaðl1 Þ
, , 

¼ ð3Þ
Absðl2 Þ aC ðl2 ÞL þ 2xðk þ 1ÞDaðl2 Þ

10 55 nm Because of the linear relationship between ExtB/Ext345 and AbsB/
Abs345 (Fig. 2e), we can re-write this as
Intensity

0 100 200 300 400

Length, L (nm)
114 nm
ExtB aC ðlB ÞL þ 2xðk þ 1ÞDaðlB Þ
1.6 ¼ ð4Þ
Ext345 aC ð345 nmÞL þ 2xðk þ 1ÞDað345 nmÞ
1.2 x E
c
0.8 Equation (4) can be used to fit the data in Fig. 3f, showing very
170 nm
Ratio

0.4 Figure 3 | Length metric and edge effects. (a,b) Representative TEM
ExtB/Ext345
AbsB/Abs345 207 nm images of slowly sedimenting MoS2 nanosheets (small flakes, F1, (a) scale
0.2
20 100 800 166 164 162 160 158
bar, 100 nm, (b) scale bar, 10 nm). (c,d) Representative TEM images of
Length, L (nm) Binding energy (eV) faster sedimenting MoS2 nanosheets (large flakes, F6, (c) scale bar, 1 mm,
(d) scale bar, 100 nm). (e) Extinction (e), absorption (a) and scattering (s)
40 100 coefficients at 345 nm plotted versus nanosheet length. The line represents
Ratio, NS,E/NS,T (%)
 (ml mg–1 cm–1)

30 /eS ¼ 69 ml mg  1 cm  1. (f) Ratio of extinction at B-exciton to that at

345 nm, ExtB/Ext345, plotted versus the lateral dimensions of the flake
B-exciton
20 10 (expressed as mean length, L, obtained by statistical TEM analysis). The
10 line is a fit to equation (4). Also shown is the equivalent data taken from the
A-exciton x= 2.7 nm absorbance spectra (AbsB/Abs345). The inset schematically shows the
0 1
0.00 0.01 0.02 0.03 0.04 10 100 400
division of a nanosheet into edge and central regions. (g) Absorption
Inverse length, 1/L (nm–1) Length, L (nm) coefficient, a, at both A- and B-excitons plotted versus inverse nanosheet
length. The dashed lines are fits to equation (1). (h) Fitted XPS S2p core-
level spectra of filtered MoS2 dispersions with different mean flake lengths.
Edge The additional component at higher binding energies is attributed to edge S.
2 nm (i) Fraction of S2p signal associated with edge S as a function of flake
4 nm
length. This fraction is equal to the ratio of edge S to total S atoms. The line
9 nm represents a fit to equation (2) and yields the width of the edge region to be
x ¼ 2.7±0.5 nm. (j) Scanning transmission electron microscopy (STEM)
image of MoS2 nanosheets (scale bar, 20 nm). (k) STEM image of the
region marked by the box in j showing the edge region of a typical MoS2
2.8 6.4
nanosheet (scale bar, 5 nm). The boxes indicate the regions over which EEL
Centre 2.5 x~8 nm
Ratio I2.8 eV/I6.4 ev
Intensity (a.u.)

spectra were acquired and analyzed. (l) Summed EEL spectra, normalized to
2.0
the zero-loss peak, of the regions marked by the boxes in k (same data set
9 nm
4 nm 1.5 as Supplementary Fig. S12c). The EEL spectra exhibited gradual changes
2 nm
when moving from the edge to the centre. (m) Plot of EELS intensity ratio at
Edge Edge 1.0
2.8 eV to that at 6.4 eV (as indicated in l) as a function of distance from the
2 4 6 8 10 0 4 8 12 16 edge. Data from three different regions are shown. The intensity ratio
Energy loss (eV) Distance from edge (nm) saturates at B8 nm from the edge.

NATURE COMMUNICATIONS | 5:4576 | DOI: 10.1038/ncomms5576 | www.nature.com/naturecommunications 5

& 2014 Macmillan Publishers Limited. All rights reserved.
ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms5576

good agreement, at least in the size range of 70–350 nm. Once the both surfactant-tip and surface-tip interactions contribute to the
fit parameters are known, equation (4) can be rearranged to give a apparent height of the nanosheets, making N assessment difficult.
quantitative relationship between L and ExtB/Ext345: To resolve this, we utilized the fact that incompletely exfoliated
nanosheets often display terraces separated by steps associated
3:5ExtB =Ext345  0:14 with flake edges (Fig. 4d inset). By analyzing the apparent AFM
LðmmÞ¼ ð5Þ
11:5  ExtB =Ext345 height on a large number of steps over many flakes27 and plotting
the step height in ascending order (Fig. 4d), it is clear that the step
By assessing the root-mean-square relative difference between height is always a multiple of 1.9 nm. Even though this is larger
data and fit we estimate the error in L to be o10%. This is than expected for monolayer MoS2, the absence of smaller step
important and it allows ExtB/Ext345 to be used to assess flake heights implies surfactant-exfoliated MoS2 to display apparent
length for MoS2 suspended in liquids simply by measuring the monolayer heights of 1.9 nm (cf 1.1 nm for chemically exfoliated
extinction spectrum. It works effectively for nanosheets in the size MoS2)14. Furthermore, the thinnest flakes we have observed have
range 70oLo350 nm, which is almost exactly the range of an apparent height of 3 nm, consistent with a monolayer height of
nanosheets typically produced by LPE. For larger nanosheets, a 1.9 nm plus B1 nm surfactant under the flakes
different metric, based on the scattering exponent n can be used Unequivocal evidence that the nanosheets with an apparent
(Supplementary Figs 13–16 and Supplementary Note 2). height of 3 nm are monolayered MoS2 can be obtained using
photoluminescence (PL) and Raman spectroscopy, which allows
Nanosheet thickness metric based on confinement effects. identification of mono-, bi- and tri-layer MoS2 (refs 38,44). For
While a spectroscopic method to determine nanosheet length is this, we relocated the sample area characterized by AFM
undoubtedly important, even more useful would be an in situ (Fig. 4a,b) to a Raman microscope to optically characterize those
method to assess thickness. As discussed above, it is clear that the flakes with known AFM thickness. Raman mapping, plotted as
position of the A-exciton peak in the absorbance spectrum has intensity of the A1g mode in Fig. 4e, shows the presence of many
the potential to provide such a metric14,44. However, the number MoS2 flakes on the substrate in these regions. Critically, some of
of layers per flake, N, must be measured before the relationship these flakes also display PL (Fig. 4f). The observed PL (and
between peak position and N can be quantified. Raman spectra) from flakes with a height of B3 nm was always
To do this, we deposited nanosheets from a number of consistent with monolayered MoS2 (Fig. 4c,g and Supplementary
fractions onto Si/SiO2 wafers for characterization by AFM Figs 20–22). In contrast, slightly thicker flakes (for example,
(Fig. 4a,b). In this way, we measured mean thicknesses for a 5–7 nm in height), showed weaker PL with a clear shoulder at
large number of nanosheets across a number of fractions (Fig. 4c 620 nm, consistent with bi- and few-layered species38,51–53.
and Supplementary Figs 17–19). This combination of AFM, Raman and PL spectroscopy clearly
However, conversion of the measured height into the actual shows that surfactant-exfoliated monolayers display an apparent
number of monolayers per flake, N, is challenging due to height of 3 nm with every additional monolayer contributing an
adsorbed surfactant on both flakes and substrate. This means extra 1.9 nm. This in turn allows the conversion of the mean AFM

1.2
Intensity (a.u.)

0.8

0.4

0.0
600 650 700
Raman PL Wavelength (nm)

30
10 13.3 Ext metric
Number of layers, N

10 Ext/AFM
11.4
Step height (nm)

Abundance (%)

Abs/AFM 51
8
Height (nm)

9.5 20 Mak et al.

PL
6 7.6
4 5.7
10
3.8
2 1.9 nm 1
1.9
0 0.0 0
0.0 0.1 0.2 0.3 0.4 0 15 30 45 60 1 2 3 4 5 6 7 8 9 650 660 670 680
Position (µm) Sample number Layer number, N A (nm)

Figure 4 | Thickness metric. (a) Representative overview AFM images of a surfactant-exfoliated MoS2 dispersion deposited on Si/SiO2 wafers (scale bar,
1 mm). (b) Zoomed-in AFM images of the regions indicated in a. (c) Height profiles of the nanosheets in (b). (d) Heights of steps observed on deposited
MoS2 nanosheets such as that displayed in the inset (scale bar, 250 nm). The step height is found to be a multiple of 1.9 nm. (e) Raman (integrated A1g
phonon area, scale bar, 1 mm) and (f) PL maps (excitation 532 nm, integrated area from 640–700 nm, scale bar 1 mm) of the region shown in a.
(g) Photoluminescence spectra of the areas marked by circles (colour coded). The blue circle and spectrum, respectively, is associated with a flake of
apparent height 3 nm (Fig. 4c). Inset: zoom-in of PL from multi-layered flakes marked by red and green circles associated with flakes of apparent thickness
5–7 nm (Fig. 4c). (h) Example (fraction F2) of a layer number histogram after conversion of AFM height to number of layers. (i) Plot of the mean number of
layers as obtained from AFM thickness analysis versus the wavelength associated with the A-exciton measured from both extinction and absorbance
spectra. The vertical error bars represent the standard error of the AFM height distribution. The red filled diamond represents free-standing
micromechanically cleaved MoS2 monolayer, where lA was obtained from the PL peak position. The experimental findings are consistent with literature
data on the photoluminescence of free-standing MoS2 extracted from Mak et al.51 NB profiles in c are colour coded with the lines in b.

6 NATURE COMMUNICATIONS | 5:4576 | DOI: 10.1038/ncomms5576 | www.nature.com/naturecommunications

& 2014 Macmillan Publishers Limited. All rights reserved.
NATURE COMMUNICATIONS | DOI: 10.1038/ncomms5576 ARTICLE

height obtained by statistical analysis of a given dispersion drop- 1.5 0.3

cast on Si/SiO2 wafers to the actual number of layers. This is

Length, L (mm)
ExtB / Ext345 nm
depicted in Fig. 4h for a BS fraction, which optical measurement 0.2
showed to have a low mean nanosheet thickness (that is, F2 with 1.0
A-exciton position 666 nm1.86 eV).
0.1
These results allow us to use AFM to measure the mean value 0.5
of N in different fractions and so quantify the relationship
between the mean number of monolayers per nanosheet, N, and 0.0 0.05
the measured A-exciton position, lA. We plot this data in Fig 4i 660 670 680 1 2 4 6 8 10
for lA taken both from extinction and absorption spectra. In both A (nm) Thickness, N
cases, the data are very similar, showing a well-defined relation-
ship between N and lA. The data point for the monolayer was 0 0.5
obtained from determining the A-exciton energy from the PL of

Overpotential (V)
120 mV per dec

J (mA cm–2)
free-standing micromechanically cleaved MoS2 (Supplementary 0.4 L=190 nm
–1
Fig. 23, in absence of surfactant). We can test these results by
comparing them to literature data51 for A-exciton position for 0.3
mono-, bi- and tri-layer mechanically exfoliated MoS2, finding –2 75 mV per dec

very close agreement. This suggests any solvatochromic effect54 0.2 L=60 nm
L=190 nm L=60 nm
associated with the presence of adsorbed sodium cholate to be –3
small as confirmed by additional experiments (Supplementary –0.5 –0.3 –0.1 0.2 1 2
Fig. 24). V v RHE (V) J (mA cm–2)
The data in Fig. 4i clearly show an exponential relationship 600 7
between N and lA. Fitting the data where lA is taken from the Excitation em=650 nm
6 Emission ex=435 nm
extinction spectrum gives 550

Intensity (a.u.)
Ram an - OH 5
A
ex (nm)
C
N¼2:310 e 36  54;888=lA
ð6Þ 500 4
3 B
with lA in nanometre. This expression can be used to find the 450 2
mean number of monolayers per nanosheet directly from the 400
1
extinction spectrum of a MoS2 dispersion. However, we note that 0
for NB10, the extinction and absorbance data begin to diverge, as 620 640 660 680 400 500 600 700
scattering becomes more important. Thus, while equation (5) is em (nm)  (nm)
only accurate for No10, this is not a serious limitation as LPE
typically gives dispersions with No10. In addition, because the Figure 5 | Use of the metric for applications and fundamental studies.
properties of MoS2 become bulk like as N approaches 10 (refs 38), (a,b) Map of extinction metrics for length and thickness for a number of
knowledge of the exact thickness in this regime is unnecessary. fractions obtained by band sedimentation under varying centrifugation
It is important to demonstrate that these metrics are indeed conditions from various stock dispersions. (a) Raw data from extinction
independent of each other. We do this by analyzing dispersions spectroscopy and (b) converted to actual mean length and thickness
prepared in a way which allows thickness and length to be by using the calibration equations obtained from spectroscopy and
varied independently. Such analysis (Supplementary Note 3; microscopy. (c) Linear sweep voltammograms (5 mVs  1) for films of
Supplementary Figs 25 and 26) shows that ExtB/Ext345 and lA are length-selected MoS2 on indium tin oxide (ITO) substrates, comparing
not intrinsically linked (although there may be a relationship the electrocatalytic response towards hydrogen-evolution reaction.
between mean nanosheet thickness and length, which is due to Measured potentials were subjected to iR correction (see Supplementary
the exfoliation mechanism, see below). In addition, we have Information). Supporting electrolyte was 0.5 M H2SO4. (d) Overpotential
shown (Supplementary Fig. 27) that the metrics work well even at versus current density plots showing Tafel slopes expressed as mV per
the highest dispersion concentrations. This is an important result, decade (mV per dec). (e) Photoluminescence excitation contour plot of
which makes it possible to determine both mean length and a monolayer-rich MoS2-SC dispersion showing the expected A-exciton
number of layers for solution-processed MoS2 nanosheets using fluorescence at 651 nm. (f) Excitation spectrum of the emission signal
two metrics extracted from a simple extinction spectrum, which is collected at 650 nm and emission spectrum excited at 435 nm. A, B and
measureable in any laboratory. C-excitonic features are clearly identified. The Raman signal of O–H was
subtracted from the excitation spectrum.

Applying length and thickness metrics. We can use these measurements of MoS2-sodium cholate (SC) (Supplementary
metrics to assess the dimensions of flakes in a wide range of Fig. 28). Importantly, Fig. 5b shows that dispersions can be
dispersions produced under a range of conditions including produced with mean values of N close to 1.
simple homogenous centrifugation at different rates as well as The ability to both control and easily measure nanosheet size
samples that were fractionated by BS. Shown in Fig. 5a, is a graph and thickness will be extremely useful, facilitating the processing
of extinction ratio ExtB/Ext345 plotted versus lA for 450 dis- of materials for both applications and basic research. We
persions. It is clear that these data all fall on the same master demonstrate the application of these techniques in both of these
curve. We can use equations (5 and 6) to transform Fig. 5a into a areas. First, we demonstrate the utility of accurate size control and
graph of L versus N (Fig. 5b). This curve is interesting, showing a measurement by preparing hydrogen-evolution catalysts from
clear correlation between lateral size and thickness. This is not solution-exfoliated MoS2. The sites responsible for the catalysis of
because the metrics are correlated. Rather, it is due to the greater hydrogen evolution by MoS2 are known to reside on the flake
interlayer binding energy associated with larger flakes: they are edges55, making small flakes desirable and illustrating the need
harder to exfoliate and so are thicker on average. Such a corre- for size control. We prepared two dispersions using distinct
lation has previously been observed for liquid-exfoliated MoO3 centrifugation protocols. The measured extinction spectra were
nanosheets33 and confirmed in this work on flake by flake used to determine the mean lateral size of nanosheets (190 and

NATURE COMMUNICATIONS | 5:4576 | DOI: 10.1038/ncomms5576 | www.nature.com/naturecommunications 7

& 2014 Macmillan Publishers Limited. All rights reserved.
ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms5576

60 nm) as well as the concentration of dispersed material. lex ¼ 651 nm) is depicted in Fig. 5f, clearly revealing signatures of
Vacuum filtration was used to produce thin films the B and C-excitonic transitions. We note that while the
(30 mg cm  2), which were transferred onto indium tin oxide- measurement of contour maps and excitation spectra is virtually
coated glass substrates and characterized for hydrogen-evolution impossible using micromechanically cleaved MoS2, it is
catalysis (see methods and Supplementary Note 1). A clear size straightforward in liquid once size selection and measurement
effect is seen (Fig. 5c,d), with considerably lower onset potential techniques are known.
and Tafel slope55 observed for the smaller flakes compared with
larger ones. This demonstrates the potential of our method to
Extension of metrics to other nanosheet types. The nanosheet
facilitate optimized sample preparation for applications of
nanosheets. extinction metrics described in this paper rely on fundamental
principles that both flake edge and degree of exfoliation influence
Furthermore, the ability to produce dispersions with controlled
nanosheet dimensions allows the preparation of samples for more the electronic structure. This implies that this is a general method
that can be used for a wide range of liquid-exfoliated nanosheets.
fundamental studies. To demonstrate this, we used size selection
coupled with analysis via the thickness metric to produce To demonstrate this, we show the same changes to extinction
spectra for band-sedimented, surfactant-exfoliated dispersions of
monolayer-rich (that is, with low average N) MoS2 dispersions,
which allow the study of solution-phase PL. Shown in Fig. 5e is a WS2 (Fig. 6), MoSe2 and WSe2 (Supplementary Figs 32–33).
Almost identical behaviour is observed in all materials although
PL-excitation contour plot for such a dispersion, which shows the
direct band gap luminescence associated with monolayer MoS2 the extinction ratio must be calculated using appropriate wave-
lengths: for WS2, we use ExtA/Ext295, where A represents the
(lem ¼ 651 nm). The emission spectrum (lex ¼ 435 nm) shows a
narrow peak (full-width half-maximum ¼ 23 nm) that can be A-exciton position (for example, Fig. 6a–c). The ability to identify
well-exfoliated WS2 allowed us to prepare dispersions containing
fitted to a single Lorentzian, consistent with PL from mono-
layered MoS2. The PL spectrum is size independent with no very thin nanosheets (that is, with low lA). A combination of
Raman and PL spectroscopy, coupled with AFM (Fig. 6d–g)
apparent edge effect other than reduction of PL intensity for very
small nanosheets (Supplementary Note 4 and Supplementary confirmed such dispersions to be rich in monolayers. Compared
with MoS2, the WS2 PL (normalized to the Raman 2LA and
Figs 29 and 30). Fluorescence from few and multi-layered species
in the dispersion (with a mean thickness of two layers) cannot be nearby modes in Fig. 6h) appears significantly enhanced. In
addition, unlike MoS2, the PL associated with WS2 edges was red
resolved due to the significantly lower quantum efficiency51. The
excitation spectrum (after subtraction of Raman mode of water, shifted and more intense than that from the nanosheet centre
(Fig. 6h). This suggests that the band gap of the edge region in
liquid-exfoliated WS2 is smaller than that of the centre, in
agreement with previous results56. Most importantly, neither very
1.5 small (B50 nm), nor reasonably sized (B300 nm) few-layer
F1
F2 species show a notable PL signal further supporting that the
Ext (a.u.)

1.0 F3
F4 8 nanosheets with a nominal AFM height of 3 nm are monolayered
Height (nm)

A F5 nanosheets.
F6 6
0.5 F7 4
2
0.0 0 Discussion
300 450 600 750 0.0 0.2 0.4 In conclusion, we have introduced a simple and versatile
Wavelength (nm) Length (µm)
centrifugation technique based on BS to rapidly sort liquid-
Raman PL suspended TMDs according to their mass using a normal
630 benchtop centrifuge. This process gave a range of fractions,
which had significantly different extinction spectra. Specifically,
A (nm)

626 both the energy of the A-exciton and the relative intensity of the
B-excitonic transition varied systematically with fraction number
(that is, with nanosheet size). These parameters are related to
622
nanosheet thickness and lateral size, respectively. Statistical
0 1 2 3 4 5 6 7 8 measurements of nanosheet length and thickness, made using
Fraction TEM and AFM, allowed us to quantify these relationships. Such
5 quantification gives access to the determination of mean flake
ML center
ExtA / Ext295 nm

Norm. intensity

0.6 4 ML edge length and thickness in one single extinction measurement. In

3 addition, we find the extinction coefficient at 345 nm to be widely
0.4 independent of nanosheet thickness and length, allowing the
2
extinction spectrum to be used to measure concentration.
0.2 1
The variations in the MoS2 absorbance and extinction spectra
0 with nanosheet size described above are due to the effects of both
1 2 3 4 5 6 7 575 600 625 650
Fraction Wavelength (nm)
confinement and nanosheet edges on the electronic structure of
the material. Since these phenomena are general to all 2D
Figure 6 | Metric for WS2. (a) Extinction spectra, normalized to 295 nm, of materials, the extinction metrics described here can be applied
surfactant exfoliated WS2 fractions after band sedimentation at 2.5k r.p.m. to a range of nanosheet types as we demonstrate for WS2, WSe2
for 10 min. The A-exciton is labelled. (b) A-exciton wavelength plotted as a and MoSe2.
function of fraction number. (c) Extinction peak ratio, ExtA/Ext295, plotted We believe this to be an important result for a number of
versus fraction number. (d) Typical AFM images of deposited WS2 reasons. On the one hand, the availability of a simple spectro-
nanosheets (scale bar, 1 mm) and (e) height profile of the nanosheet at the scopic method to measure both thickness and lateral size will be
top right (scale bar, 250 nm). (f) Raman and (g) PL intensity map (532 nm extremely important and will greatly simplify the basic char-
excitation) of the same area (scale bar, 1 mm). (h) Representative PL spectra acterization of dispersions of liquid-exfoliated nanosheets. In
of the nanosheet displayed in e at the centre and in the edge region. addition, these results will facilitate the preparation of liquid-

8 NATURE COMMUNICATIONS | 5:4576 | DOI: 10.1038/ncomms5576 | www.nature.com/naturecommunications

& 2014 Macmillan Publishers Limited. All rights reserved.
NATURE COMMUNICATIONS | DOI: 10.1038/ncomms5576 ARTICLE

suspended nanosheets with well-defined length and thickness were acquired over similar regions. For the plot of EELS intensity ratio, I2.8 eV/
distributions. This will be extremely important for applications I6.4 eV, in Fig. 3m, spectra in three different regions of interest perpendicular to the
flake edge were averaged parallel to the edge. Three sets of EELS maps were
where nanosheet size is crucial, as well as for understanding analyzed in total: set 1:150 spectra acquired over 4.17 nm  1 nm (direction per-
fundamental physical properties of layered inorganic materials as pendicular to the edge of the flake) were summed for each spectrum that was
a function of size (for example, by fluorescence spectroscopy in analyzed; set 2:282 spectra acquired over 14.67 nm  0.47 nm (direction perpen-
solution). Furthermore, this work highlights the influence the dicular to the edge of the flake) were summed for each spectrum that was analyzed;
edge has on the electronic and optical properties of the nanosheet set 3:66 spectra acquired over 8.34 nm  0.51 nm (direction perpendicular to the
edge of the flake) were summed for each spectrum that was analyzed. The spectra
as a whole. were then normalized to the intensity of the zero-loss peak. Subsequently, the
ratio at 2.8 eV in the EELS to that at 6.4 eV as a function of distance from the edge
was plotted. Electrochemical characterization consisted of linear sweep voltam-
Methods metry (2mV s  1) and electrochemical impedance spectroscopy using a three-
Sample preparation. Details are presented in the Supplementary Methods. In
electrode electrochemical cell and Gamry Reference 3,000 potentiostat. The
brief, TMD dispersions (100 ml) were prepared by probe sonicating (Sonics VX-
measurements were performed in 0.5 M H2SO4 solution using a three-electrode
750) the powder (Sigma Aldrich, typical 20 g l  1) in an aqueous SC with the
electrochemical cell, with a Ag/AgCl (3 M NaCl) reference electrode and graphite
constant weight ratio of SC/TMD ¼ 0.3. The resultant raw dispersion was left to
rod counter electrode.
settle overnight and subjected to a pre-centrifugation step (Hettich Mikro 220R
centrifuge, fixed angle rotor 1016, 750 r.p.m., 60 g) to remove unexfoliated material.
Prior to BS, 1 ml of the stock dispersion of the nanomaterial in aqueous SC was References
layered on top of the race layer containing 5 ml of deuterated water at the bottom 1. Chhowalla, M. et al. The chemistry of two-dimensional layered transition metal
and 5 ml of 1:1 mixture of deuterated water and water (equal surfactant con- dichalcogenide nanosheets. Nat. Chem. 5, 263–275 (2013).
centrations). After centrifugation in a Heraeus Megafuge 16 benchtop centrifuge 2. Nicolosi, V., Chhowalla, M., Kanatzidis, M. G., Strano, M. S. & Coleman, J. N.
equipped with a 3,655 swinging bucket rotor, typically (unless otherwise noted) at Liquid exfoliation of layered materials. Science 340, 6139 (2013).
5k r.p.m. (4,695 g) for 10 min, the seven fractions were collected from the initial 3. Novoselov, K. S. et al. A roadmap for graphene. Nature 490, 192–200 (2012).
11 ml liquid from top to bottom to obtain samples with standard size and thickness 4. Wang, Q. H., Kalantar-Zadeh, K., Kis, A., Coleman, J. N. & Strano, M. S.
distributions. To obtain larger flakes, conditions were modified: large and thick Electronics and optoelectronics of two-dimensional transition metal
flakes were enriched in a dispersion by first centrifuging the stock dispersion at dichalcogenides. Nat. Nanotechnol. 7, 699–712 (2012).
1k r.p.m. (107 g, Hettich Mikro 220R centrifuge, fixed angle rotor 1016). The 5. Geim, A. K. Graphene: status and prospects. Science 324, 1530–1534 (2009).
supernatant was then subjected to a second centrifugation at 2k r.p.m. (425 g). The 6. Huang, X., Zeng, Z. Y. & Zhang, H. Metal dichalcogenide nanosheets:
sediment after this centrifugation was collected for BS at 2k r.p.m. for 10 min preparation, properties and applications. Chem. Soc. Rev. 42, 1934–1946 (2013).
(1,135 g). To obtain the monolayer-rich dispersion for PLE measurements, the 7. Wilson, J. A. & Yoffe, A. D. Transition metal dichalcogenides discussion
stock dispersion was centrifuged at 5k r.p.m. (2,663 g, 2 h, Hettich Mikro 220R and interpretation of observed optical, electrical and structural properties. Adv.
centrifuge, fixed angle rotor 1016), the sediment was discarded, the supernatant Phys. 18, 193–335 (1969).
was again centrifuged at 15k r.p.m. (22,500 g, 2 h, Hettich Mikro 220R centrifuge,
8. Zhan, Y. J., Liu, Z., Najmaei, S., Ajayan, P. M. & Lou, J. Large-area vapor-phase
fixed angle rotor 1195-A). After the second centrifugation step, the sediment was
growth and characterization of MoS2 atomic layers on a SiO2 substrate. Small 8,
reagitated in 0.1 mg ml  1 SC (reducing the initial volume by 50% to yield a high
966–971 (2012).
concentration) and subjected to PL measurements. For XPS and electrochemical
9. Frindt, R. F. single crystals of MoS2 several molecular layers thick. J. Appl. Phys.
measurements, the dispersions were vacuum-filtered using porous cellulose filter
37, 1928–1929 (1966).
membranes. For Raman, PL and AFM, dispersions were drop-casted on Si/SiO2
wafers. 10. Novoselov, K. S. et al. Two-dimensional atomic crystals. Proc. Natl. Acad. Sci.
USA 102, 10451–10453 (2005).
11. Radisavljevic, B., Radenovic, A., Brivio, J., Giacometti, V. & Kis, A. Single-layer
Characterization. Optical extinction was measured on a Varian Cary 5,000 in MoS2 transistors. Nat. Nanotechnol. 6, 147–150 (2011).
quartz cuvettes with a pathlength of 0.4 cm. To distinguish between contributions 12. Murphy, D. W. & Hull, G. W. Monodispersed tantalum disulfide and
from scattering and absorbance to the extinction spectra, dispersions were mea- adsorption complexes with cations. J. Chem. Phys. 62, 973–978 (1975).
sured in an integrating sphere using a home-built sample holder to place the 13. Joensen, P., Frindt, R. F. & Morrison, S. R. Single-Layer MoS2. Mater. Res. Bull.
cuvette in the centre of the sphere of a Perkin Elmer Lamda 650 spectrometer (NB 21, 457–461 (1986).
cuvettes need to be transparent to all sides). The absorbance spectrum is obtained 14. Eda, G. et al. Photoluminescence from chemically exfoliated MoS2. Nano Lett.
from the measurement inside the sphere. A second measurement on each dis- 11, 5111–5116 (2011).
persion was performed outside the sphere to obtain the extinction spectrum. This 15. Zeng, Z. Y. et al. Single-Layer semiconducting nanosheets: high-yield
allows calculation of the scattering spectrum (extinction-absorbance). Bright field preparation and device fabrication. Angew. Chem. Int. Ed. 50, 11093–11097
TEM imaging was performed using a JEOL 2100, operated at 200 kV, while high- (2011).
resolution TEM was conducted on a FEI Titan TEM (300 kV) on holey carbon 16. Blake, P. et al. Graphene-based liquid crystal device. Nano Lett. 8, 1704–1708
grids (400 mesh). Tapping-mode AFM was carried out on a Veeco Nanoscope-IIIa (2008).
system (Digital Instruments). Raman and PL spectroscopy on substrate was per- 17. Bourlinos, A. B., Georgakilas, V., Zboril, R., Steriotis, T. A. & Stubos, A. K.
formed using a WITec alpha 300 with 532 nm excitation laser in air under ambient Liquid-phase exfoliation of graphite towards solubilized graphenes. Small 5,
conditions. The PL in solution was acquired on a Fluorolog-3 spectrometer (Horiba 1841–1845 (2009).
Scientific) with a thermoelectrically cooled R928P photomultiplier tube detector. 18. Coleman, J. N. et al. Two-dimensional nanosheets produced by liquid
The samples were excited with a 450 W Xe lamp with a double monochromator in exfoliation of layered materials. Science 331, 568–571 (2011).
excitation (600 grooves per mm, 500 nm blaze grating). XPS was performed under 19. Cunningham, G. et al. Solvent exfoliation of transition metal dichalcogenides:
ultra-high vacuum conditions (o5  10  10 mbar), using monochromated Al Ka dispersability of exfoliated nanosheets varies only weakly between compounds.
X-rays (1,486.6 eV) from an Omicron XM1000 MkII X-ray source and an Omicron
ACS Nano 6, 3468–3480 (2012).
EA125 energy analyzer. An Omicron CN10 electron flood gun was used for charge
20. Hernandez, Y. et al. High-yield production of graphene by liquid-phase
compensation and the binding energy scale was referenced to the adventitious
exfoliation of graphite. Nat. Nanotechnol. 3, 563–568 (2008).
carbon 1s core level at 284.8 eV. Mo 3d and S 2p core-level regions were recorded
at an analyzer pass energy of 15 eV and with slit widths of 6 mm (entry) and 3 mm 21. Khan, U. et al. Polymer reinforcement using liquid-exfoliated boron nitride
x 10 mm (exit), resulting in an instrumental resolution of 0.48 eV. After subtraction nanosheets. Nanoscale 5, 581–587 (2013).
of a Shirley background, the core-level spectra were fitted with Gaussian-Lor- 22. Khan, U., O’Neill, A., Lotya, M., De, S. & Coleman, J. N. High-concentration
entzian line shapes and using Marquardt’s algorithm. STEM imaging and STEM- solvent exfoliation of graphene. Small 6, 864–871 (2010).
EELS investigations were carried out using an FEI Titan 60–300 Ultimate micro- 23. Lin, Y., Williams, T. V. & Connell, J. W. Soluble, exfoliated hexagonal boron
scope (‘PICO’) operated at 80 keV, equipped with a high-brightness electron gun, a nitride nanosheets. J. Phys. Chem. Lett. 1, 277–283 (2010).
Cs probe corrector, and a Gatan Quantum postcolumn energy filter system (Ernst 24. O’Neill, A., Khan, U. & Coleman, J. N. Preparation of high concentration
Ruska-Centre Juelich). The suspended MoS2 flakes were dropped onto lacey car- dispersions of exfoliated MoS2 with increased flake size. Chem. Mater. 24,
bon-coated copper TEM grids that were then heated up to 120°C overnight under 2414–2421 (2012).
vacuum to minimize contamination of the sample. EELS spectra were acquired 25. Zhi, C. Y., Bando, Y., Tang, C. C., Kuwahara, H. & Golberg, D. Large-scale
with a convergence angle of 24 mrad, a collection angle of B28 mrad, a dispersion fabrication of boron nitride nanosheets and their utilization in polymeric
of 0.01 eV/channel and an energy resolution over vacuum of 0.1 eV. STEM-EELS composites with improved thermal and mechanical properties. Adv. Mater. 21,
maps were acquired with 0.01 s acquisition time per spectra per pixel. Subse- 2889–2893 (2009).
quently, the spectra were aligned and calibrated using the zero-loss peak. Individual 26. Zhou, K. G., Mao, N. N., Wang, H. X., Peng, Y. & Zhang, H. L. A mixed-solvent
spectra (1,152) were summed to obtain each of the spectra shown in Fig. 3l. strategy for efficient exfoliation of inorganic graphene analogues. Angew. Chem.
The spectra shown here are representative of a set of 20 STEM-EELS maps that Int. Ed. 50, 10839–10842 (2011).

NATURE COMMUNICATIONS | 5:4576 | DOI: 10.1038/ncomms5576 | www.nature.com/naturecommunications 9

& 2014 Macmillan Publishers Limited. All rights reserved.
ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms5576

27. Paton, K. R. et al. Scalable production of large quantities of defect-free few-layer 48. Zhang, C., Johnson, A., Hsu, C.-L., Li, L.-J. & Shih, C.-K. Direct imaging of
graphene by shear exfoliation in liquids. Nat. Mater. 13, 624–630 (2014). band profile in single layer MoS2 on graphite: quasiparticle energy gap, metallic
28. Wang, J. Z. et al. Development of MoS2-CNT composite thin film from layered edge states, and edge band bending. Nano Lett. 14, 2443–2447 (2014).
MoS2 for lithium batteries. Adv. Energy Mater. 3, 798–805 (2013). 49. Yin, X. et al. Edge nonlinear optics on a MoS2 atomic monolayer. Science 344,
29. Smith, R. J. et al. Large-scale exfoliation of inorganic layered compounds in 488–490 (2014).
aqueous surfactant solutions. Adv. Mater. 23, 3944–3948 (2011). 50. van der Zande, A. M. et al. Grains and grain boundaries in highly crystalline
30. Cunningham, G. et al. Photoconductivity of solution-processed MoS2 films. monolayer molybdenum disulphide. Nat. Mater. 12, 554–561 (2013).
J. Mater. Chem. C 1, 6899–6904 (2013). 51. Mak, K. F., Lee, C., Hone, J., Shan, J. & Heinz, T. F. Atomically thin MoS2. a
31. May, P., Khan, U. & Coleman, J. N. Reinforcement of metal with liquid- new direct-gap semiconductor. Phys. Rev. Lett. 105, 136805 (2010).
exfoliated inorganic nano-platelets. Appl. Phys. Lett. 103, 163106 (2013). 52. Splendiani, A. et al. Emerging photoluminescence in monolayer MoS2. Nano
32. May, P., Khan, U., Hughes, J. M. & Coleman, J. N. Role of solubility Lett. 10, 1271–1275 (2010).
parameters in understanding the steric stabilization of exfoliated 53. Li, H. et al. From bulk to monolayer MoS2: evolution of raman scattering. Adv.
two-dimensional nanosheets by adsorbed polymers. J. Phys. Chem. C 116, Funct. Mater. 22, 1385–1390 (2012).
11393–11400 (2012). 54. Mao, N., Chen, Y., Liu, D., Zhang, J. & Xie, L. Solvatochromic effect on the
33. Hanlon, D. et al. Liquid exfoliation of molybdenum trioxide. Chem. Mater. 26, photoluminescence of MoS2 monolayers. Small 9, 1312–1315 (2013).
1751–1763 (2013). 55. Jaramillo, T. F. et al. Identification of active edge sites for electrochemical H-2
34. Naguib, M. et al. Two-dimensional transition metal carbides. ACS Nano 6, evolution from MoS2 nanocatalysts. Science 317, 100–102 (2007).
1322–1331 (2012). 56. Gutiérrez, H. R. et al. Extraordinary room-temperature photoluminescence in
35. Xu, Y., Kominami, K., Ishikawa, Y. & Feng, Q. Layered hydroxide nickel triangular WS2 monolayers. Nano Lett. 13, 3447–3454 (2012).
benzoates: hydrothermal synthesis, structure characterization, and exfoliation
reaction. J. Colloid. Interface. Sci. 386, 107–113 (2012).
36. Khan, U. et al. Size selection of dispersed, exfoliated graphene flakes by Acknowledgements
controlled centrifugation. Carbon 50, 470–475 (2012). The research leading to these results has received funding from the European Union
37. Lotya, M., Rakovich, A., Donegan, J. F. & Coleman, J. N. Measuring the lateral Seventh Framework Programme under grant agreement n604391 Graphene Flagship. In
size of liquid-exfoliated nanosheets with dynamic light scattering. addition, J.N.C. acknowledges the European Research Council grant SEMANTICS and
Nanotechnology 24, 265703 (2013). the SFI grant number 11/PI/1087 for financial support. C.B. acknowledges the German
38. Lee, C. et al. Anomalous lattice vibrations of single- and few-layer MoS2. ACS research foundation DFG (BA 4856/1-1). V.N. and H.C.N. acknowledge ERC 2DNa-
Nano 4, 2695–2700 (2010). noCaps, SFI PIYRA and FP7 MoWSeS. N.S. and J.M. acknowledge support from DFG
39. Svedberg, T., Pederson, K. O. & Bauer, J. H. The Ultracentrifuge (Oxford SPP 1459 (MA 4079/7-1) and European Research Council (grant number 259286).
University Press, 1940). G.S.D. and N.C.B. acknowledge SFI for PI_10/IN.1/I3030. N.M. thanks FP7-2010-PPP
40. Fagan, J. A., Becker, M. L., Chun, J. & Hobbie, E. K. Length fractionation of Green Cars Electrograph (266391).
carbon nanotubes using centrifugation. Adv. Mater. 20, 1609–1613 (2008).
41. Fagan, J. A. et al. Centrifugal length separation of carbon nanotubes. Langmuir Author contributions
24, 13880–13889 (2008). C.B., N.M., N.C.B., N.S., J.M., G.S.D., D.M. and J.F.D. performed spectroscopy, C.B.,
42. Sun, X., Luo, D., Liu, J. & Evans, D. G. Monodisperse chemically modified R.J.S., H.C.N., A.O., D.H., L.H. and V.N. performed microscopy, P.J.K. and T.H. fabri-
graphene obtained by density gradient ultracentrifugal rate separation. ACS cated and tested devices, C.B. and J.N.C. planned the experiments and wrote the paper.
Nano 4, 3381–3389 (2010).
43. Bonaccorso, F., Zerbetto, M., Ferrari, A. C. & Amendola, V. Sorting
nanoparticles by centrifugal fields in clean media. J. Phys. Chem. C 117, Additional information
13217–13229 (2013). Supplementary Information accompanies this paper at http://www.nature.com/
44. Mak, K. F., Lee, C., Hone, J., Shan, J. & Heinz, T. F. Atomically thin MoS2: a naturecommunications
new direct-gap semiconductor. Phys. Rev. Lett. 105, 136805 (2010).
45. Zhao, W. J. et al. Evolution of electronic structure in atomically thin sheets of Competing financial interests: The authors declare no competing financial interests.
WS2 and WSe2. ACS Nano 7, 791–797 (2013). Reprints and permission information is available online at http://npg.nature.com/
46. He, G. S., Qin, H. Y. & Zheng, Q. Rayleigh, Mie, and Tyndall scatterings of reprintsandpermissions/
polystyrene microspheres in water: wavelength, size, and angle dependences.
J. Appl. Phys. 105, 023110 (2009). How to cite this article: Backes, C. et al. Edge and confinement effects allow in situ
47. Dolui, K., Das Pemmaraju, C. & Sanvito, S. Electric field effects on armchair measurement of size and thickness of liquid-exfoliated nanosheets. Nat. Commun. 5:4576
MoS2 nanoribbons. ACS Nano 6, 4823–4834 (2012). doi: 10.1038/ncomms5576 (2014).

10 NATURE COMMUNICATIONS | 5:4576 | DOI: 10.1038/ncomms5576 | www.nature.com/naturecommunications

& 2014 Macmillan Publishers Limited. All rights reserved.