The Raman spectrum of grapheme

In this lecture we will look at the Raman spectrum of Graphene, and learn
about the origins of the different peaks that appear in the Raman spectrum.
The most significant peaks in the Raman spectrum of pristine graphene are the
G band which appears at 1580 inverse centimeter and the 2D band which
appears at around 2700 inverse centimeter.
This is also sometimes known as the G prime band.
If the graphene is disordered, two additional peaks appear, the D band, at
around 1350 inverse centimeter, and the D prime band which appears at about
1620 inverse centimeter.
The positions of some of these peaks are dependent on the exciting laser
energy as we will see later. So the values mentioned here are only indicative.
We will now discuss the origin of these peaks in the next slide.
First, we will take a close look at the origin of the peaks in mono layer
graphene. The G band is associated with a doubly degenerate TO and LO
phonon mode at the Brillouin zone center, or the gamma point. The G band
arises from a normal first order Raman scattering process in graphene.
On the other hand, the D and the 2D bands originate from a second order
process involving one TO phonon, and one defect in the case of the D band,
and two TO phonons near the K point for the 2D band.
In the case of the D band, the two scattering processes consist of one elastic
scattering by a defect of the crystal and one inelastic scattering by emitting or
absorbing a phonon.
When we say an inelastic event it means that there is no change in the energy
involved, which means that there is no phonon involved. Similarly, an elastic
scattering event means that there is an energy change and, consequently, a
phonon involved.
This double resonance process is called an intervalley process because it
connects points on non-equivalent K and K prime points in the first Brillouin
zone of graphene.
On the other hand, the double resonant process that forms the D prime band is
an intravallley process since it connects two points around the K point or the K
prime point alone.
As mentioned before, the 2D peak is a two phonon band, which is allowed in
the second order of Raman spectra of graphene hing without any kind of

disorder or defects. In the case of the 2D band, the two scattering processes
are both inelastic events, and two phonons are involved.
In the case of bilayer graphene, or with Bernal, or ABAB stacking, the 2D peak
looks very different.
It needs to be fit with four components rather than the single component shape
of the 2D peak in monolayer graphene.
As a result, the 2D peak of bilayer graphene looks wider than that of monolayer
graphene, and is also asymmetric. Whereas, the 2D peak of monolayer
graphene is perfectly symmetric.
In the case where the two layers are not ABAB stacked, the 2D peak is broad,
but not asymmetric, and we will discuss this later.
In Bernal-stacked bilayer graphene, the electronic and phonon bands split into
two components.
We have seen the electronic band structure of bilayer graphene previously,
showing two valence bands and two conduction bands.
For these two phonon components, the motion of the carbon atoms in the
upper and lower layers are in phase with respect to each other for one
component, and out of phase with respect to each other for the other
component.
For the case of AB stacked bilayer graphene, the double resonance conditions
are more complex than in the case of monolayer graphene where there is only
one double process contributing to the 2D band. In bilayer graphene, the
resulting four different scattering processes give rise to four Raman peaks In
the 2D peak.
This discussion can be extended now to three layer graphene, four layer
graphene, and so on, until the case of multilayer graphene, or graphite.
In this optical image, we see a grapheme flake which has different regions, or
different thicknesses, of graphene. And we obtained the Raman spectrum from
each of these regions, and compared their 2D peaks.
It can be shown that the number of allowed components for the 2D peak of
trilayered graphene is 15.
Fortunately, from an experimental standpoint, many of the energies of these
peaks lie very close to each other and, often, the 2D peak of trilayer graphene
can be effectively fitted with as few as five components.

The Raman spectrum of grapheme

This discussion can now be extended to three layer graphene, four layer
graphene, and so on, until the case of multi-layer graphene, or graphite.
In this optical image we see a grapheme flake that has different regions of
different thicknesses. And we obtain the Raman spectrum from each of these
regions and compare their 2D peaks.
It can be shown that the number of allowed components of the 2D peak of trilayer graphene is 15. Fortunately, from an experimental standpoint, many of
the energies of these peaks lie very close to each other. And often the 2D peak
of tri-layer graphene can effectively be fitted with as few as five components.
The 2D band shape, a number of components continues to evolve this way,
until we reach the case of bulk graphite. The 2D band of graphite can be fit
with just two peaks as seen here.
We see that it is possible to use the shape of the 2D peak to uniquely identify
monolayer graphene, AB stacked bilayer graphene, AB-stacked trilayer
graphene, and graphite. It is not straightforward to distinguish unambiguously
between 4-layer graphene, 5-layer graphene, etc.
And finally, we consider the case of bilayer graphene which is not AB-stacked,
which we have previously referred to as rotated bilayer graphene, and also,
other rotated multilayer graphene structures. Monolayer graphene has a
symmetric 2D peak, with well defined full width at half maximum of about 23
inverse centimeters, which is a measure of the phonon lifetime.
Non-AB-stacked bilayer and multi-layer graphenes show a deceptively
symmetric 2D peak, unlike the asymmetric 2D peak of AB-stacked bilayer
graphene. However, this is much broader than the 2D peak of monolayer
graphene.
Therefore, a symmetric single component 2D peak alone should not be used as
validation of monolayer graphene. And attention must be paid to the width of
this peak before drawing conclusions.
Finally, modern Raman spectrometers often contain an automated sample
stage which can move in the x and y directions, exposing a grid of points on
the sample to the incoming laser in a pre-programmed sequence. The
spectrometer will capture the Raman spectrum at each point. And once the
entire region is mapped, it is possible to generate spatial plots of the positions,
intensities, and widths of various Raman peaks of graphene.
By this process one can generate what are known as a Raman maps of various
peak characteristics, which often correlate with various anatomical features of

the graphene samples. Here you see a graphene sample with different
numbers of layers, and the corresponding Raman maps of the intensity of the G
peak, the intensity of the D peak, and the full width at half maximum of the 2D
peak.
It is possible to identify regions of correlation, and such maps can provide
valuable snapshots of the quality and uniformity of graphene samples. One
particular aspect to note in this map is the D peak intensity that occurs at the
edge of a graphene flake. This is because the edge of a flake is essentially an
elongated defect.
When characterizing graphene samples where the flake size is comparable to
or less than the laser diameter, which is usually about one micron, there will
always be a strong D peak rising from the flake edges. And care should be
taken that this is not interpreted as a defective graphene structure, but rather
as a measure of the flake size distribution.
Analysis of graphene Raman spectra

Now that we know the origins of the various peaks in the Raman spectrum of
graphene, we will look at how the properties of these peaks. Such as their
position, intensity and full width at half maximum will change as a result of
external factors.
External perturbations need to be taken into account when analyzing the
Raman spectrum of graphene.
And consequently, the Raman spectrum yields invaluable information about the
state of the graphene material. In the double resonance process, we have
encountered so far, both in the case of the intervalley and intravalley
scattering.
We have set of two coupled quantities, the wave vectors of the electrons and
phonons and their dispersion relations.
But changing the laser excitation energy, different points in the momentum
space of the electrons and phonon dispersion are probed.
And in this way, it is possible to map the phonon or electron dispersion
relations.
This figure shows the Raman spectra of the 2D and the G star bands of monolayer graphene for 1.92 electron volts, 2.18 electron volts, 2.41 electon volts,
2.54 electron volts, and 2.71 electron volt, laser excitation energies.

The next graph shows the dependence of the frequency of the 2D and G* on
the laser excitation energy. The double resonance Raman peaks of grapheme
are therefore said to be dispersive, since their position varies as a function of
the excitation energy.
Since the electron dispersion is linear, the higher the incident energy, the
higher the phonon wave vector. And the higher the phonon energy. The 2D
peak has a typical dispersion of about 100 centimeters squared inverse per
electron volt.
The d peak is also similarly dispersive with a dispersion of about 50 inverse
centimeter per electron volt. And we will encounter this again later.
The 2D peak of bi-layer graphene, as we have seen before, comprises of four
components.
Each of these four components is similarly dispersive, and
the overall 2D peak of bi-layer grapheme also shifts with the excitation energy.
The coupling between the phonons and electron/hole pairs re-normalizes
the phonon energy of graphene, and the re-normalization depends on
the electron/hole concentrations.
The effect is strong near the Gamma and K points in the Brillouin zone. The E
to G phonon mode of graphene, which gives rise to the raman G band is also
known to exhibit a very strong coupling with the electron hold pairs.
This effect can be measured directly by Raman spectroscopy, as it probes the G
band frequency as a function of the Fermi energy. The Raman spectra of
monolayer grapheme seen here, shows how the modification of the Fermi
energy induces changes in the frequency and the width of the G band.
The G band frequency of monolayer graphene increases and the G band line
width decreases, when the Fermi energy is moved away from the dirac point.
Due to doping in either electrons or holes which corresponds to positive or
negative changes in the Fermi level.
It should be obvious, therefore, that while it is possible to judge whether a
graphene sample is doped or not. The extent of doping by observing the
position and the width of the G peak, it is not possible to distinguish whether
the doping is p type or n type.
Additional peaks need to be observed in order to make this determination.

In order to distinguish between electron and hole doping, we have to look at

the 2D peaks of graphing. For electron doping, the position of the 2D peak does
not change much until a high degree of doping, about three into ten to the 13th
per square centimeter.
At high gate voltages, there is a significant softening of the peak.
For hole doping the position of the 2D peak increases steadily similar to the
case of the G peak. An additional effect of doping can be observed in the
intensity of the G and 2D peaks.
Generally, absolute intensities of the G and 2D peak are hard to interpret
quantitatively. However, the ratio of intensities of the G and 2D peak is a good
metric. Normally, in the case of monolayer graphing, which is undoped and free
of defects and other perturbations.
The 2D peak intensity is about three times the G peak intensity. This is
generally regarded as the surest sign of high quality graphene material. This
next graph shows the dependence of the ratio of the G, and 2D peak intensities
as function of doping level.
The intensity of G and 2D modes decreases with increase doping level.
However, the dependents of the 2D mode on doping is much stronger than that
of the G mode.
Hence, the intensity of 2D over the intensity of G ratio is a strong function of
doping level.
However, once again, this measure is unable to distinguish between p-type and
n-type doping.
In most cases of graphene samples, the graphene sheet is supported on
a dielectric or metallic substrate, especially in electronic applications.
This substrate has historically been a Silicon Oxide coated Silicon wafer, which
is also popular due to the choice, due to the ease of visualizing graphing on
such substrates.
When with appropriate oxide thickness as we have seen earlier.
An undesired consequence of any substrate and particularly the Silicone
Dioxide substrate, is the unintentional and non uniform doping of the graphing
by interaction with the substrate.
Evidence for this and quantification for the doping by means of Raman's
spectroscopy will be discussed in this slide.

This was first noticed as strong variations in the Raman spectrum of several
graphene samples deposited on similar Oxidized Silicon substrates.
It was confirmed through great gated electronic transport experiments that the
graphene on oxidized silicon substrate is usually unintentionally doped by the
substrate.
Which can reach levels of up to 10 to the 13 per square centimeter under
ambient conditions. And this can change from sample to sample. And it is this
inhomogeniously distributed doping which is responsible for the changes in the
Raman spectrum within the same sample.
This can be attributed to absorbed water on the surface, or trapped water
between the silicon and the graphene, or the charged impurities in the
substrate. As we have seen in the previous slide, as the extent of doping
increases.
The Raman G peak becomes sharper and up shifted in wave numbers. This can
be captured jointly as shown in this next figure. The filled black circles
represent unintentionally doped samples on a silicon dioxide substrate. While
the open white circles here, show what happens when a grapheme sample is
intentionally and controllably doped by gating, which follows the same trend.
We can also confirm that while the 2D peak is dispersive, because of double
resonance process, the G peak is not dispersive. And finally, combining the
information of the 2D peak position and the G peak position.
We can distinguish between p-type and n-type doping as mentioned in the
previous slide, and is presented here in a different manner. The stars on the
lower left indicate the positions of the 2D and G peak in a totally undoped
sample. The dopes graphing sample follows two trend lines in this sort of a
plot.
One for p-doping and one for n-doping. Next, we consider the effects of
increased defects and disorder in the grapheme lattice on the Raman
Spectrum.
When moving from perfect graphene to a defective graphene, both the full
width, and half maximum of the G peak, and the position of the G peak will
increase. Due to the relaxation of the Raman selection rule.
The D and D prime peaks activated by disorder, and their intensity increases
with the amount of disorder. This is seen in this figure, which shows the Raman
spectrum of UV disordered

graphene compared to pristine graphene. Other effects of disorder include a

decrease in the ratio of the 2D peak to the G peak intensity and the broadening
of the 2D peak.
As mentioned previously, the D peak is also a double resonance process. And
like all double resonance processes is also dispersive.
The dispersion of the 2D peak is about 50 centimeter inverse per electron volt
(eV), compared to about 100 centimeter inverse per electron vote for the 2D
peak.
Generally, doping in defects both occur in a graphene sample. Increased
doping causes the D peak position to increase, and so does increase disorder.
However, disorder tends to broaden the G peak while doping tends to sharpen
it.
These two effects are always in conflict with each other.
In this figure, if we assume that bulk graphite does not suffer.
The substrate doping effects that mono-layer graphing is vulnerable to, then
the green dots represent the trend for the existence of the disorder in the
complete absence of doping.
The black and red dots are from the graph in the previous slide, and correspond
to the existence of doping without disorder.
Most as prepared of processed grapheme samples tend to fall somewhere in
the region of the blue dots and the blue squares.
Where both doping and disorder interplay. When a graphene sample lies in this
chart, it's a good indication of the electronic and structural quality of the
graphene material.
The most defect and doping-free graphene sample is usually suspended or
free-standing grapheme sample shown here by stars.
Characterized by low G-peak position of about 1580 centimeter inverse and
narrow full-width to half maximum of less than 20 inverse centimeter.
Finally, we will look at the effect of strain on the Raman peak of graphene. In
particular, we will look at the effect of uniaxial strain, where the in-plane strain
is applied on the graphene, is significantly greater or focused along one
direction. Which does not need to align with any crystallographic axis.

In order to controllably and reproducibly induce the strain, graphene layers

prepared by micromechanical cleavage of graphite, were deposited on a
flexible substrate.
The substrate is then subjected to a fourth point or two point bending as shown
here.
Under such a bending condition, the top surface of the flexible sub straight is
under tensile strain while the bottom surface is under compressive strain.
There's a strain transfer from the top surface of the bend support onto the
graphene, which is adhering strongly to it. In this fashion, one can can apply
uniaxial tensile strain of up to 1.3% to the graphene monolayer.
Returning to the E2G phonon responsible for the G mode in the Raman
spectrum of graphene. We learned before that this phonon mode is doubly
degenerate, as evident from the phonon dispersion in graphene. It was
observed that upon tensile strain loading, the double degenerate E2 mode
splits into two components.
One polarized along the strain and the other perpendicular. This split in the
Raman G peak into two sub-bands, known as the G+ and G-, is analogous to
that induced by curvature in carbon nanotubes. These sub-bands red shift with
increasing strain While their splitting increases as well.
Their relative intensities vary with polarization, allowing us to probe the
sample's crystallographic orientation with respect to the strain.
It was also observed that the 2D band downshifts, but is not split for such small
strains.

Raman spectra of graphene derivatives and other 2-D materials

In the final section of this lecture, we will look at how Raman spectrum can be
used to characterize not only graphene, but also derivatives of graphene and
other two-dimensional materials beyond graphene.
First, we will look at the two derivatives of graphene, the fluorinated version
called fluorographene and the hydrogenated version called graphane. In both
cases partial or complete conversion of graphene is possible, meaning that
either some or all of the carbon atoms in a graphene sheet will have a
hydrogen or a fluorine atom attached to it. In the case of complete
hydrogenation or fluorination, the structure of the resultant derivative can be
seen here.

Each carbon atom goes from sp2 hybridization to sp3 hybridization. And the
hydrogen or fluorine atom or adjacent carbon atoms are located on opposite
sides of the graphene sheet resulting in a two-dimensional arrangement of
tetrahedral carbon atoms, which resembles a two-dimensional plane of
diamond.
Because the carbon atoms are now tetrahedral corordinated, the carbon atoms
are no longer in a single plane, but distorted into a shape where the A and B
sub-lattices occupied their own parallel plane.
The process of conversion of graphene into fluorographene or graphane can be
followed by Raman spectroscopy, and the signatures of partial and complete
conversion can be observed.
Here you will see the step-by-step conversion of graphene to fluorographene
followed by Raman spectroscopy.
First, we can see that the prominent D peak emerges. This indicates the
appearance of atomic scale defects. As the fluorination time increases, the
double resonance 2D peak disappears, whereas the G and D peak intensities
remain approximately same. With increasing fluorination time of up to a few
days, all the graphene features gradually completely disappear. In contrast,
when we hydrogenate graphene, one can observe the Raman spectrum even
for fully hydrogenated graphene.
And the 2D band always remains strong, even when the graphene has been
completely hydrogenated. The complete disappearance of the Raman spectrum
of fluorographene can be attributed to the fact that the band gap of
fluorographene is about 3.5 eV, which is greater in energy than the green light
used to obtain the Raman spectrum. And the sample becomes completely
transparent to the excitation laser since no laser energy can be absorbed to
initiate the scattering process.
The hydrogenation of graphene, as can be seen here, is a partially reversible
process. And an annealing step can be used to reverse the hydrogenation to
result in graphene. And the Raman spectrum of the annealed graphane sample
will resemble the grapheme sample prior to conversion. And finally we will look
at how Raman spectroscopy can be applied to other two-dimensional materials
beyond graphene. In this case, we consider a transition metal dichalcogenide,
molybdenum disulfide.
Monolayer, bilayer, and trilayer and thicker layers of molybdenum disulfide can
be seen here in this optical micrograph of a micromechanically exfoliated
MoS2.

Similar to what you have encountered previously for micromechanically

exfoliated graphene. The number of layers is determined by atomic force
microscopy. MoS2 consists not of one atomic plane repeatedly stacked, as in
the case of graphene and graphite, but rather one atomic layer comprised of
three atomic planes which are stacked.
Each layer has a central plane of molybdenum atoms coordinated to four
sulphur atoms arranged in two staggered atomic planes, one above and one
below. In this next figure we see the representative Raman spectrum of singleand few-layer MoS2. 2D crystals of MoS2 are characterized By the E12G and
the EA1G modes' mere 400 centimeter inverse.
We see that the E12G vibrations softens or red shifts, while the A1G vibration
stiffens or blue shifts with increasing sample thickness. For films of four or more
layers, the frequency of both modes will converge on the bulk values.
This can be used to identify the number of MoS2 layers, just as we used the
shape of the 2D peak to identify the number of graphene layers.
One can also obtain spatial Raman maps for MoS2 of the Raman frequency, for
example, of the E12G, and the A1G modes, clearly showing these opposing
shifts as a function of layer thickness. The shifts can be correlated to the
phonons involved in each of the two peaks. In-plane vibrations in the case of
the E12G and the out of plane vibrations in the case of the A1G phonons.