Fuel 86 (2007) 1855–1864

www.fuelfirst.com

Characterization of Algerian Hassi-Messaoud asphaltene

structure using Raman spectrometry and X-ray diffraction
a,b
Y. Bouhadda , D. Bormann b, E. Sheu c,*
, D. Bendedouch d, A. Krallafa d, M. Daaou d

a
Faculté des Sciences, T.C.T, Université Hassiba Ben-Bouali, Chlef, Algeria
b
Centre de Recherches sur les Matériaux à Hautes Températures CNRS, CRMHT UPR 4212, Université d’Orléans, 1D,
Avenue de la Recherche Scientifique, 45071 Orleans Cedex 2, France
c
Vanton Research Laboratory, Inc., 7 Olde Creek Place, Lafayette, CA 94549, USA
d
Faculté des Sciences, Laboratoire de Chimie Physique Macromoléculaire, Université d’Es-Senia, Oran, Algeria

Received 12 July 2006; received in revised form 28 November 2006; accepted 4 December 2006
Available online 8 January 2007

Abstract

Asphaltenes derived from Algerian oil wells were characterized by Raman spectrometry to investigate their molecular structures in
solid-state. The aromatic sheet diameter (La) was estimated using the integrated intensities of the G and the D1 mode, along with the
Tuinstra equation [Tuinstra F, Koenig JL. J Chem Phys 1970;53:1126]. The values obtained for three Algerian Hassi-Messaoud asphal-
tene samples are on the same order as the literature values and also consistent with the X-ray diffraction results. The Raman spectra
corresponding to the G and the D bands can be fitted with Gaussian, Lorentzian, and Gaussian/Lorentzian hybrid functions in a
self-consistent manner. A three peak fitting procedure found that the Gaussian/Lorentzian hybrid function with G being Gaussian
and D Lorentzian is the best combination. Incorporation of the X-ray data on the height of the crystallite, Lc, gives rise to an estimate
of eight asphaltene molecules in each asphaltene aggregate.
Ó 2007 Elsevier Ltd. All rights reserved.

Keywords: Asphaltene; Raman spectrometry; X-ray diffraction; Aromatic sheet diameter; Fitting functions

1. Introduction side chains attached to it. In addition, the aromatic core

Asphaltenes are a fraction of petroleum liquids insoluble
in n-heptane but soluble in some aromatic solvents such as
toluene, pyridine, or benzene [1]. Since the definition is
based on solubility, asphaltenes represent a class of com-
pounds (a mixture of compounds of different structures)
with similar solubility property rather than a pure sub-
stance with a well-defined molecular structure. This com-
plicates the selection of characterization techniques and
choice of processes for handling asphaltene containing
materials. Despite of this microscopic complexity, asphal-
tene molecules have been shown to generally comprise a
poly-condensed aromatic (PCA) core with hydrocarbon
*
Corresponding author. Tel.: +1 925 878 5142; fax: +1 925 947 1978.
E-mail address: ericsheu@vantonlab.com (E. Sheu).
may include heteroatoms, such as nitrogen, oxygen, sulfur
and metals [2]. These heteroatoms are likely relevant to
their high propensity of self-aggregation, flocculation and
precipitation [3].
These micro phase separations often aggravate into
macro phase separations and hinder the production flow,
storage stability, transportation liability, and refining. To
avoid such potential operational disaster, it is vital to inter-
actively optimize operational parameters during produc-
tion, to control asphaltene flocculation, precipitation, or
their corresponding kinetics. Controlling these parameters
requires thorough understanding of the molecular inter-
plays among different species in asphaltenic materials. It
is thus important to study the relevance between the
asphaltene molecular structures and their impact on the
bulk physical properties.

0016-2361/$ - see front matter Ó 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.fuel.2006.12.006
1856 Y. Bouhadda et al. / Fuel 86 (2007) 1855–1864
The aromatic moiety in asphaltenic molecules, through
p–p and dipole interactions, was thought to be one of the
dominant contributors to asphaltene self-association. The
aromaticity changes during both upstream production or
down stream refining processing [4]. Due to this, a simple
and reliable method for characterization of the asphaltene
PCA contents and the molecular structure can serve as a
gateway toward controlling asphaltene self-association
led flocculation and precipitation during field operations
[5]. On the research end, it can possibly lead to elucidation
of the structure-driven thermodynamic and transport prop-
erties under natural geological conditions, which are diffi-
cult to simulate in a laboratory [6].
The number of aromatic hexagons in a graphene or a
carbonaceous material (equivalent to the aromatic sheet
diameter) has been a subject of interests and much work
has been explored using various techniques – NMR [7],
XANES [8], and AFM [9]. On the asphaltene structure,
experimental evidence had shown that an average aromatic
sheet comprises of 4–10 benzene-condensed rings [5]. This
was a breakthrough but with tremendous amount of work
and cumbersome data interpretation processes. In looking
for a field friendly technique or any potential techniques
with different probes that can either confirm or invalidate
this important result, Raman spectroscopy stands out,
from both experimental convenience and data interpreta-
tion point of view. This technique has been applied for
characterization of various carbonaceous materials and
fundamental data analysis schemes are readily available
[10].
Incorporation of X-ray diffraction (XRD) with Raman
spectroscopy has been the most frequently used ‘‘twin-
technique’’ for characterizing aromatic containing carbo-
naceous materials [10–12]. They characterize the aromatic
portion of the molecules via different mechanisms. XRD
evaluates the interplays between the aromatic sheets of
carbonaceous materials (or asphaltene molecules) [13].
Through Bragg diffraction law, it provides dimension of
the molecules stacking. The (10) band, which normally
appears around 43° for carbonaceous materials, is related
to the diameter of the basal area of the molecule (La).
Raman spectrometry on the other hand provides evidence
of the molecular bond vibrations that are regulated by the
intra and inter molecular interactions. When applied to
aromatic compounds, Raman scattering carries informa-
tion about the crystallinity (or the crystalline domain
dimension) of the system and is sensitive to the edge of
an irregular structure [14,15]. Because of its sensitivity
to the edge structure in 2-D, the local PCA structure
can be extracted [16]. In this work we adopted the Tuin-
stra and Koenig equation [17] to estimate the extent of
the aromatic structure using the integrated intensities of
the so-called G band (near 1580 cm1) and the D1 band
(near 1350 cm1) [17]. The diameter so obtained is an esti-
mate of the diameter of the fused benzene condensed
rings, from which the number of rings can be roughly
calculated.
Raman spectrometry is popular for studying commer-
cial graphite-like or diamond-like polycyclic aromatic
hydrocarbon (PAH) compounds, as well as carbon blacks
that are commonly used for high-tech industry and envi-
ronmental products [11–15]. This technique, however,
was rarely used for characterizing asphaltenes, even though
it should be a potential candidate for such applications. To
the best of our knowledge, there was only one attempt in
using Raman scattering for asphaltene characterization.
Bergmann et al. [18] applied X-ray Raman to study asphal-
tene, which allowed them to evaluate the molecular inside
of this complex system. However, a synchrotron based X-
ray source is needed to perform such an experiment, which
is not practical for field applications. Lacking Raman
works for asphaltene characterization might be attributed
to the fact that in graphite systems, the existence of D
bands was only unambiguously identified and proved lately
[19].
In an attempt to apply this potential technique to
asphaltene characterization, we set a rather focused goal
here. We employed Raman spectrometry as an alternative
to characterize the average molecular dimension of asphal-
tene and used it in conjunction with X-ray diffraction mea-
surements to confirm its applicability. Algerian asphaltenes
samples obtained from the Hassi-Messaoud field were used
for testing this technique. The goal was to estimate the
PCA diameter with accuracy to the first order of approxi-
mation. The analyses used here are (1) fitting the spectra
using Gaussian, Lorentzian, or Gaussian/Lorentzian
hybrid functions to represent G and/or D bands and
extract the integrated intensities for each band (2) calculate
the integrated intensity ratio of the G and the D1 bands,
and subsequently use of the Tuinstra and Koenig equation
to estimate the molecule dimensions.
2. Experiments
2.1. Asphaltene sample preparation
Four asphaltene samples were investigated by Raman
spectroscopy. The asphaltene sample, DP45, was from a
deposit of the Hassi-Messaoud oil field in Algeria. A solu-
tion containing 1 g of deposit and 40 ml of heptane was
prepared. The solution was stirred at 60 °C for 2 h, fol-
lowed by standing at ambient temperature overnight. The
insoluble part was filtered using a 0.45 lm diameter pore
membrane. Following the filtration, the insoluble fraction
was dissolved in enough amount of hot toluene (50 °C)
and re-filtered with a 0.45 lm diameter pore membrane.
The filtered solution was then dried using a Rotavapor.
The retrieved asphaltene powder was further air-dried in
an autoclave at 140 °C for one week to evaporate the trace
amount of toluene. The other three samples were prepared
in a similar way. HH1 asphaltene was directly derived from
the oil in one of the Hassi-Messaoud wells, RSV asphaltene
was from a refinery residue, and HM-3077 was from a
deposit of the Hassi-Messaoud oil field at a 3077-m depth
 Y. Bouhadda et al. / Fuel 86 (2007) 1855–1864
in Algeria. Similar to DP45 sample, a solution containing
1 g of the respective sample and 40 ml of heptane was pre-
pared and left to stand over night under continuous stirring
at room temperature. The solution was then filtered with a
filter paper (Whatman #5). The insoluble fraction on the
paper was washed by heptane until colorless. The insoluble
matter was then dissolved in toluene and filtered to remove
fractions that are not toluene soluble. The soluble fraction
was distillated to evaporate toluene. Elemental analysis
was performed on all samples using a Thermoquest CE
instrument.
2.2. Raman experiments
For Raman measurements, the sample was milled in a
mortar to reduce the particle sizes before being used. The
micro-Raman experiments were conducted on a DILOR
XY 800 Raman spectrophotometer equipped with a
charged coupled device (CCD) camera as the detector.
The laser wavelength for excitation was 514.5 nm from a
Spectra Physics laser tube (Model 2018). All spectra were
recorded in the micro-confocal retro-diffusion configura-
tion for the wave number range from 1000 to 1800 cm1
at room temperature. This covers the entire first-order
region (1000–1800 cm1) [17]. Prior to the measurements,
a small quantity of mortared sample was put on a slide
and the surface was manually flattened under the micro-
scope. A visually acceptable flat area was used as the focal
point for laser impact/excitation. Once a flat surface was
chosen, series of spectra acquisitions in this area were per-
formed in order to explore the response of the sample and
to optimize the experimental conditions. In order to
improve the statistics, several spectra were collected from
different points of the flat area.
2.3. X-ray diffraction measurement
X-ray diffraction (XRD) analyses were carried out with
a Rigaku Miniflex instrument using 30 kV voltages and
15 mA current. The CuKa (1.5405 Å) was selected as the
monochromic energy radiation. A NaI detector (a scintilla-
tion detector) was used to collect the diffraction signals.
The diffraction angles (2h) was scanned from 5° to 60° at
a 1°/min scan rate with a 0.05° step size. Only DP45 sample
was measured by XRD. The sample was mortared into fine
powders prior to measurements and all measurements were
performed at room temperature.
3. Theory
3.1. Raman
Raman spectrum of a carbonaceous material can be
divided into the first- and the second-order regions
[20,21]. In the first-order regions (1100–1800 cm1) a hex-
agonal graphite crystal with the space group symmetry
D46h exhibits a Raman spectrum with zone center modes
1857
that can be decomposed into the following irreducible
representations:
C ¼ A2u þ 2B2g þ E1u þ 2E2g : ð1Þ
Theoretically, only the E2g mode is Raman active, provided
the sample is a perfect graphite crystal. This mode corre-
sponds to the stretching vibration of the sp2 carbon atoms
in the aromatic hexagonal sheet as well as those sp2 atoms
in the chains. Because the aromatic bond involves very high
energy, this mode occurs at a rather high frequency
(1580 cm1) and is generally referred to as the G mode.
The stretching vibration of the chain sp2 atoms can perturb
the vibration frequency, leading to shift or shoulder crea-
tion of the peak. Since it is for sp2 stretching, the relevant
structures are mostly two-dimensional structures.
The natural poorly organized carbonaceous materials
and most commercial graphite (microcrystalline graphite)
show other bands in addition to the G mode. Among them
are modes at 1150, 1350, 1500, and 1620 cm1. The
1150 cm1 mode only occurs when the carbonaceous mol-
ecules are spatially poorly organized. Flame soot and fos-
sils matter are the representative materials. The
1350 cm1 mode is referred to as the D1 mode. It arises
from a change in the selection rules between the K and
the M points of the Brillouin zone [20–22]. Physically, the
D1 peak arises from the in-plane defects and the heteroa-
toms that initiate breathing vibration modes of the sp2
atoms in the rings [23]. Another word, it reflects the bound-
ary of a crystalline-like structure, or in our case, the bound-
ary of an asphaltene molecule. As a perfect graphite
crystal, which has no 1350 peak, breaks into more and
more pieces of small crystalline structures, the number of
defected boundary lines increase, so do the 1350 band
intensity, and is often in a linear fashion. In this regard,
it becomes possible to estimate the ‘‘size’’ of this crystalline
structure, since it is seen by the Raman spectrum as a
‘‘large molecule’’ rather than a crystalline structure. A pos-
sible mechanism leading to the appearance of D1 mode can
be as follows. An electron with momentum h~ k near the K
point resonantly excited from the p to the p* band by an
incident photon. The electron is then inelastically scattered,
by emission of a phonon with a momentum h~ q, to a point
near the M point with a momentum hð~ k þ~qÞ in the p*
band. This is a resonant process since it connects two real
electronic states. The electron is scattered back to ~ k by a
defect or edge in disordered graphite and in this case the
backscattering is elastic and non-resonant. Finally, the
electrons–holes recombine at ~ k position, giving rise to a
scattered photon, which is not resonant with the p–p* tran-
sition. This mechanism is known as a double resonant pro-
cess or double resonant Raman scattering (DRRS) [24].
The 1500 cm1 mode (the D3 band) only occurs as a
very wide band in poorly crystallized carbonaceous materi-
als. It is due to the defect outside the plane of the aromatic
layers, such as tetrahedral carbons [15]. Appearance of this
band can be an indication of vertical ordering, which,
in the case of asphaltene can mean disorder molecular
1858 Y. Bouhadda et al. / Fuel 86 (2007) 1855–1864
stacking. Finally, the 1620 cm1 mode (D2) always occurs
as a shoulder to the G mode, whenever the D1 mode exists.
Its physical attribute is not fully understood. Some
evidence suggested that it is attributed to the interstitial
insertion of the graphene moieties or to the stacking defor-
mations induced by the intercalated heteroatoms or metals
[25]. Another word, it is related to the PAC surface vectors
but perpendicular to the crystalline boundary.
The above descriptions of the modes in the first-order
region can only be adopted for qualitative analysis. A sim-
ple yet quantitative description linking the observed modes
and the molecular structure was proposed by Tuinstra and
Koenig [17] through the intensity ratio between the G
mode and the D1 mode. It has been largely adopted as
an empirical relation. Tuinstra and Koenig showed that
the integrated intensities (areas) of the D1 and G bands
(respectively ID and IG) are sensitive to the microcrystalline
planar crystal size La according to
IG
La ðnmÞ ¼ 4:4 ð2Þ
ID
Since the focus of this work was to evaluate the molecular
(or the aggregate) structure of asphaltene, Eq. (2) was the
sole equation used for estimating La. However, La so eval-
uated would be jeopardized, if the integrated G and D1
intensities were not accurately determined, or if the G is
not in the right range (i.e. 1570–1610 cm1) where Eq. (2)
holds [26]. Thus, accurate determination of G and D inten-
sities is required.
Curve fitting of the Raman spectra is often used to
extract more reliable G and D band intensities. Most com-
mercial data analysis software provides such an option.
There are several models and functions commonly used
for curve fitting. In this study, we first applied a three-peak
fitting (D1, D2 and G) with three functional forms for the
peaks, namely the Gaussian, Lorenzian, and the combina-
tion of them (Gaussian/Lorenzian). Due to the shoulder
incompatibility in the three-peak fitting, we also attempted
a nine-peak fitting, which, although gave much more satis-
factory fitting quality, and to some extend provides sup-
porting information for the overall structural argument,
it may pose risk of ambiguity. The three peak data were
used for our analysis.
3.2. X-ray diffraction
The aromatic sheet structure of an asphaltene can also
be characterized by X-rays diffraction patterns of the
(0 0 2), and the (1 0) peaks. In fact, one can estimate the
diameter of the aromatic planar sheet La and the height
of the crystallite in the c-axis direction, Lc, using the Bragg
and the Warren equations, from which the following
parameters can be extracted:
k
Inter-sheet distance d: d ¼ ð3Þ
2  sin h
0:9k
Height of the crystallite Lc : Lc ¼ ð4Þ
x  cos h0 0 2
1:84k
Diameter of the sheet La : La ¼ ð5Þ
x  cos h1 0
Lc
Average number of sheet per crystallite n: n ¼ þ1
d0 0 2
ð6Þ
where x is the width of the band at half maximum, k is the
X-ray wavelength and h the diffraction angle. Details about
this technique and analysis for asphaltene can be found
from pioneer works of Yen [13]. Eqs. (3)–(6) are applied
to non-deformed circular shaped aromatic crystalline.
For non-uniform inter-sheet distance and dislocated in-
ter-sheet stacking other equations are more suitable [10].
In this work, we adopted these equations assuming non-
uniformity of the inter-sheet distance is not significant.
4. Results and analysis
Analysis of the Raman data was performed in detail
for the DP45 sample, from which an optimal condition
and analysis scheme was established for the other three
samples.
Table 1 shows the results of the elemental analysis (an
average of two runs) for the DP45 asphaltene samples.
Nearly half of the deposit is composed of asphaltenic com-
pounds, indicating its potential impact on the production
and control of the deposits in the transportation pipelines
and refining streams. The C/H ratios are close to unity
for both the deposit and asphaltene, suggesting substantial
amount of aromatics in this sample.
As mentioned in Section 2, multiple Raman spectra were
collected for the same sample at different locations across a
prepared sample flat area. This allows us to examine the
homogeneity of the sample and also enabled us to obtain
a realistic average spectrum. Fig. 1 shows a snapshot of a
chosen point of data collection and its dimension in lm.
Some speckles are observed, however, the laser focal
dimension is approximately 200 lm length scale, much
greater than an average asphaltene molecule, which means
that the response functions are an ensemble average over
many molecules. The collected spectra should thus be rep-
resentative for asphaltenes molecules.
For all the selected irradiated locations, the uncorrected
Raman spectra (the background arises from incoherent
vibration of the atoms) are visibly similar to each other
as the ones depicted in Fig. 2. The D1 and G modes can
easily be identified while the D2 mode is less appealing
and, as expected, is more of a shoulder of the G mode.
In addition to these expected modes, the D1 mode shows
a shoulder just below 1300 cm1. The spectra collected
Table 1
Elemental analysis of DP45
Fractions C H S N H/C C7 insoluble
Deposit 86.58 7.54 0.5 0.57 1.03
Asphaltene 83.82 7.4 0.49 0.44 1.05 66%
 Y. Bouhadda et al. / Fuel 86 (2007) 1855–1864 1859

1.0

Intensity (Arb. Unit)

0.8

0.6

0.4

1000 1200 1400 1600 1800

Wave number (cm-1)
Fig. 3. Uncorrected average spectrum of all measured spectra.

Fig. 1. Microphotography of the irradiated surface. 1.0

Intensity (Arb. Unit)

0.8
1500
0.6
Intensity (counts)

0.4
1000
0.2

500 0.0

1000 1200 1400 1600 1800

Wave number (cm-1)
0
Fig. 4. Baseline corrected spectrum fitted with Gaussian function.
1000 1200 1400 1600 1800
Wave number (cm-1)
Fig. 2. Raman baseline uncorrected spectra from different irradiated
points on the same sample. Their similarity indicates good homogeneity of 1.0
the sample at the experiment scale.
Intensity (Arb. Unit)

0.8

from different locations in the flat area are nearly identical, 0.6
confirming the homogeneity of the sample. The collected
0.4
DP45 spectra appear to be similar to the spectra of a
poorly organized non-crystal carbon black. 0.2
The sharp G mode peak suggests that there exist at least
some short-range order of the aromatic sheets in the solid 0.0
DP45 sample, much similar to the one seen in commercial 1000
1000 1200
1200 1400
1400 1600
1600 1800
1800
carbon black. The resemblance, however, does not auto- Wave number (cm-1)
matically mean that DP45 asphaltene molecules are
stacked as perfect as a graphite-like crystal structure. It Fig. 5. Baseline corrected spectrum fitted with Lorentzian function.
however does enable us to adopt similar analysis, specifi-
cally, the Tuinstra and Koenig equation as described in
the theory section (see Eq. (2)), as long as the G mode is for estimate of the molecular dimension using the Tuinstra
within the range it applies. Since G mode observed is near and Koenig equation [17]. The fittings were performed
1600 cm1 the Tuinstra and Koenig equation is applicable. using the conventional three peaks fitting procedure. The
Fig. 3 shows the average uncorrected spectrum of the starting peak positions chosen were 1350, 1580 and
spectra in Fig. 2. This average spectrum was fitted with 1600 cm1, approximately corresponding to D1, G, and
three sets of functions – Lorentzian, Gaussian, and D2 modes [15,17,23]. Upon correction of the incoherent
Lorentzian/Gaussian hybrid functions [15,23]. Their fitting vibrations, the fluorescence contribution was removed
quality varies. Figs. 4–6 show these fittings from which the and three peaks were resolved in spite of the presence of
G and D1 integrated intensities, IG and ID, were extracted shoulders, which may be the results of other peaks.
1860
1.0
0.8
Intensity (Arb. Unit)
0.6
0.4
0.2
0.0
1000 1200 1400 1600
Wave number (cm-1)
Fig. 6. Baseline corrected DP45 spectrum fitted with Gaussian/Lorentzian
hybrid function using three peaks.
The extracted fitting parameters are given in Table 2,
along with the residual mean square deviation RMSD val-
ues and the asphaltene structural parameters derived from
Eq. (2). The La values obtained (11–18 Å depending on the
models), using the integrated intensity ratio between G and
D1 bands, are on the same order as those published in lit-
erature [1,7,9,13,27]. The value is also comparable to our
earlier publication using surface tension and viscosity for
another Algerian asphaltene (9.5 Å) [28]. The Gaussian/
Lorentzian function appears to give better fitting quality
even though the (La) values obtained are more or less con-
sistent, from the first order of approximation point of view.
To further interpret the extracted parameter values,
some theoretical consideration is needed. First, we realized
that the G band position (see Table 2) extracted is
1574 cm1, nearly identical to the reported Raman spec-
trum for the perfect single graphite (G peak at 1575 cm1
[17]). It is obvious that DP45 is not a single crystal because
it is an asphaltene, generally comprised of over a million of
components. In addition, both D1 and D2 bands were
observed, directly indicating existence of both crystalline
boundary defects and inhomogeneity along the intravalley
line perpendicular to the crystalline boundary.
According to Mapelli et al. [29], the G band frequency
position should decrease upon increasing carbon sheet
diameter. The DP45 G band position seems to suggest that
it contains perfect 2-D honey-combs like carbon sheet
structures, at least locally. However, Jawahiri et al. [23]
argues that G mode can reside at higher frequency, if there
is high local ordering. This seems to suggest that our sys-
tem, which contains non-perfect polynuclear aromatic
structures, as seen in the C/H ratio in the elemental analy-
sis, should have relative low local ordering, in order for the
Table 2
Fitting parameters extracted from the average DP45 Raman spectrum using three peaks
Fitting function D1 band (cm1)
Lorentzian 1353
Gaussian 1352
Gaussian/Lorentzian 1352
Gaussian character 0.54
Y. Bouhadda et al. / Fuel 86 (2007) 1855–1864
G band to somehow coincide with a single crystal graphite
G band position. It is not possible to differentiate the con-
tribution of a small aromatic sheet from that of local order-
ing using Raman technique alone. But, we tend to believe
that the local low order is the cause because of the fact that
asphaltene is known to be polynuclear aromatic. In this
case, the engagement of the D bands and their intensities
become important information to identify the true local
structures.
In addition to the G band, the D1 band is clearly seen on
1800
the DP45 spectrum. The extracted D1 band position is
about 1350 cm1 and the peak is broad yet pronounced,
indicating substantial amount of edge defects, as normally
depicted in a DRRS driven D1 band. Along with the D1
band was the D2 band with a comparable intensity. As
mentioned earlier, D1 band is a reflection of the edge defect
and the intensity is dependent on the total surface bound-
ary of the edge. On the other hand, D2 is the result of an
intravalley process connecting the points that belong to
the same circle around the K point of a hexagonal Brillouin
zone structure, as suggested by Cançado et al. [30] although
its physical attribution is not all that clear.
Taking the attribution of Cançado et al. [30], the conser-
vation of momentum can be satisfied by only the vectors
that are perpendicular to the edge. This is to say that the
D2 intensity only relates to the defects along the vectors
that are perpendicular to the boundary, but not the bound-
ary structure itself. In this regard, estimate of the crystalline
dimension using Tuinstra and Koenig equation should only
involve the D1 intensity. If the high local ordering plays a
role in D2 intensity, it would be an indication that the
aromatic cores of several asphaltenes stack into a graph-
ite-like structure, at least locally, with the D2 band possibly
representing the edge structure of the molecule adjacent to
the reference one, and with a edge line perpendicular
to the boundary of the reference aromatic sheet. In which
case, the combination of small aromatic sheets and high
local orderings may be the cause of the G band being on
the single crystal G band frequency and co-exist with D1
and D2 bands. If one takes this argument for granted, the
aromatic sheet stacking not only exists, the inter-sheet spac-
ing should be similar to graphite. This cannot be quantified
by the Raman spectra, but can be determined by the X-ray
diffraction technique.
An interesting point in this simple fitting analysis that
supports the high local stacking order is the Gaussian char-
acter of the de-convoluted bands on the Lorentzian/Gauss-
ian fitting. It appears that G and D2 bands are better
G band (cm1) D2 band (cm1) RMSD Aromatic sheet diameter La (Å)
1574 1603 0.00117 13.61
1582 1597 0.000466 17.83
1574 1597 0.000250 11.85
1 0.92
 Y. Bouhadda et al. / Fuel 86 (2007) 1855–1864 1861

described by the Gaussian function, suggesting much

longer coherence lifetime than the amplitude correlation 1.0
time, a typical feature of solid-like phase. A high local
0.8

Intensity (Arb. Unit)

ordering will naturally result in solid-like characteristics.
On the argument of small aromatic sheets with boundary 0.6
defects, it may be interpreted through the D1 model. The
D1 mode shows Lorentzian characteristics rather than 0.4
Gaussian, meaning that the incoherence sets in rapidly
0.2
and that the dephasing is the dominant energy loss channel,
similar to gases. This disordering phenomenon at the D1 0.0
band suggests small sheet aromatic core with defect bound-
1000 1200 1400 1600 1800
ary (in Raman language). Combining Gaussian G and D2,
Wave number (cm-1)
and Lorentzian D1 leads to a pictorial local structure as
follows. Asphaltene PCA cores stack to form short-range Fig. 7. Initial position of the nine considered peaks.
orders structure with short horizontal dimensions
(12 Å, assuming the Lorentzian/Gaussian hybrid model)
with imperfect vertical stacking. Because of this, there are
appreciable amount of peripheral aromatics that contribute 1.0

Intensity (Arb. Unit)

to gas-like incoherent breathing vibrations that are
reflected in the D1 mode while the intravalley double reso-
nance process that is perpendicular to the edge results in 0.6
the D2 band.
With all the arguments given above, the D1 band inter-
pretation given here should still be taken with caution, as
0.2
the D1 band is known to vary with laser wavelength, its
power, the incident radiation angle, and obviously the
1000 1200 1400 1600 1800
morphology of the material (powder or crystal) [31,32].
Cançado et al. [30] showed, with Raman, AFM and Wave number (cm-1)
TEM results that the structural difference at the edge Fig. 8. Baseline corrected spectrum fitted with Gaussian/Lorentzian
(armchair or zigzag) coupled with the crystallite size could function using nine peaks.
exhibit several D1 bands at different Raman shifts. This
may be why there are shoulders observable to the left of
the main D1 peak. Table 3
As mentioned earlier, the three-peak analysis results in Fitting parameters extracted for DP45 average spectrum using nine peaks
poor fitting quality near the shoulders. In order to improve Band position Full width half maximum Gaussian Band
the fitting quality and to some extend investigate possible (cm1) (cm1) character
structural relevance to the observed Raman shoulders, we 1192 110 0.98 D4
adopted the analysis process of Sadesky et al. [15]. In his 1277 82 1 n.a.
1330 57 0.26 n.a.
work, he demonstrated that unlike a perfect or turbostatic
1358 80 1 D1
carbon graphite, soot and related carbonaceous materials 1435 93 1 n.a.
(matter containing a weak amount of hetero-atoms) are 1527 95 1 D3
much less defined in structures and thus require more peaks 1581 43 1 G
to fit the Raman spectra. Adopting this concept and used a 1591 100 0.61 n.a.
1605 40 0.58 D2
converging root mean square value (RMSD) value for
goodness of fit, we noted that the RMSD value would dras- n.a.: not assigned.
tically decreases with additional fitting peaks up to more
than 20. It is highly questionable that this fitting procedure
and the overly simplified mathematical criterion (RMSD) It was found that nine peaks are adequate to fit accu-
can unambiguously extract the real bands. Moreover, rately the experimental data including all the shoulders.
RMSD value is sensitive to the background signal as In addition to the D1, D2, and the G bands, the D3 model
pointed out by Sadezky [15] who recommended caution near 1527 cm1 was clearly seen and the D4 band near
in exclusive use of RMSD as a sole criterion. We thus 1192 cm1 is also obvious. D4 was proposed to be a peak
decided to use nine peaks and visual assessment to judge representing the sp2–sp3 bonds or C–C and C@C stretching
the goodness of fit, particularly near the shoulders. Fig. 7 vibration of polyene-like structure [33]. Since there have
shows the initial assigned peak positions. Fig. 8 shows been no previously published works of Raman spectro-
the fit with Gaussian/Lorentzian hybrid functions. The scopy for graphite or carbon black using so many peaks
extracted parameters are given in Table 3. in analysis, the interpretation of the other peaks (1277,
1862 Y. Bouhadda et al. / Fuel 86 (2007) 1855–1864

1330, 1435, and 1791 cm1), if they are real, is still lacking.
Nevertheless, computational works conducted by several 1.0
groups [29–31] showed that the theoretical Raman spectra 0.8

Intensity (Arb. Unit)

of polycyclic aromatic hydrocarbon might contain several
peaks whose appearance may not all be evident in the 0.6
experimental spectra because of their relatively weak inten-
0.4
sities. In addition, the co-existence of D3 and D4 with D1
and D2 in the nine-peak analysis can serve as supporting 0.2
information to the structural arrangement we proposed
0.0
for the small dimension asphaltene aggregate. This is
because the D4 supports the existence of sp2–sp3 bonds 10 20 30 40 50 60
vibration, which may be the result of the vibration from Diffraction Angle 2θ
the aliphatic–aromatic intercalation structure in the
Fig. 9. X-ray diffractogram of Hassi-Messaoud asphaltene. The peak with
stacked asphaltene molecules. star is a signal contributed by the sample holder.
Question regarding the simultaneous occurrence of two
D bands (D1 and D2) and G bands and their contribution
to the accuracy in estimating the (La) value is still often Table 4
raised. In this context Rigolio et al. [34] argued that when Structural parameters deduced from X-ray diffraction for DP45
an aromatic sheet is large enough (multiple rings) it is likely La (Å) d0 0 2 (Å) dc (Å) Lc (Å)
accompanied with a high aromaticity. This, in Raman lan- 17.58 3.56 4.35 28.24
guage, corresponds to a decrease of D1 intensity, a shift of
G band towards low frequencies, a decrease of the FWHM,
and to the splitting of the D1 mode into two peaks, similar Mapelli et al. [29] and Jawahiri et al. [23] that the Raman
to what we observed here. G model frequency would decrease with increasing PCA
In order to focus on the goal set in this study, we simply sheets dimension (La) and high local ordering, as observed
argue that asphaltene aromatic sheet has a size smaller than in our asphaltene case.
perfect graphite sheet, but large enough to require a com- Detail analysis of the DP45 Raman data suggests that a
bination of D1, D2, (and possibly D3 and D4) and G, in structural view of stacked asphaltene molecule in the solid
order to integrate the picture together. However, estimate state is likely correct and that a three peak analysis for the
of the La value requires only the D1 intensity because it Raman spectrum is likely adequate if one only looks for the
is the one that are edge dependent. Additionally, D3 and first order of accuracy. As such, we applied the same anal-
D4 can only be obtained through the nine-peak analysis ysis for DP45 to the rest of the samples and estimated the
and that this analysis may pose ambiguity issue, the signif- La values. The RSV sample could only be fitted with four
icance of these two bands is not discussed in detail here. In peaks due to its very different spectral characteristics.
fact, only the three-peak analyses were used for asphaltene Tables 5–8 show the elemental analysis results and the
structural argument. extracted parameter values for all the studied samples
Having analyzed the Raman data and the proposal of whereas Figs. 10–12 show the corresponding fitted spectra.
the aggregated imperfect asphaltene stacking structure, it The La values from G and D1 band analysis were 10.44,
is still necessary to quantify the inter-stacking space. It,
in many ways, evaluates the legitimacy of the aromatic Table 5
sheet dimensions obtained from the Raman analysis and Elemental analysis of HH1, RSV and HM-3077 samples
the local high ordering argument. X-ray diffraction experi- Fractions C H S N H/C
ments were conducted for these purposes. Fig. 9 shows the
HH1 84.75 6.51 0.49 0.61 0.91
X-ray diffraction spectrum from the same sample. The two RSV 84.27 7.52 0.42 0.73 1.06
sharp peaks at 36° and 43° are from the sample holder. HM-3077 87 7.2 0.25 1 0.98
From Fig. 9, the detected (0 0 2), (1 0) and the c bands were
used to calculate the structural parameters (d, La, Lc, and
the number in each asphaltene of stack) using Eqs. (2)–
(6). The La value (=17.58 Å) obtained from X-ray is consis- Table 6
tent with the result derived from the three-peak Raman Raman spectrum analysis (Gaussian/Lorentzian) for HH1 using three
analysis. peaks and IG/ID1 (La = 10.44 Å)
X-ray diffraction also provides the Lc value and the Band Intensity Full width half Gaussian Integrated
inter-sheet spacing, d, from which the average aggregation position maximum (cm1) character intensity
number can be estimated. The result for DP 45 was given in (cm1)
Table 4. It has an average of eight aromatic sheets, a com- 1331.34 5.25 179.61 0.34 2316.37
fortable number reported previously [35]. It also, to some 1566.23 1.87 121.84 1 549.64
1596.63 5.09 41.67 0.95 518.13
extend, indirectly confirms the arguments proposed by
 Y. Bouhadda et al. / Fuel 86 (2007) 1855–1864 1863

Table 7 HM-3077 sample

Raman spectrum analysis (Gaussian/Lorentzian) for RSV using four
peaks and IG/ID1 (La = 9.87 Å)

Intensity (Arb. Unit)

Band Intensity Full width half Gaussian Integrated
position maximum (cm1) character intensity
(cm1)
1135.04 149.91 105.8 1 36012.09
1345.38 729.51 169.26 0.7 297197.6
1516.32 254.74 109.36 1 66681.16
1594.54 707.77 70.14 0 139385.4

1000 1200 1400 1600 1800

180
Table 8
Raman spectrum analysis (Gaussian/Lorentzian) for HM-3077 using three Wave number (cm-1)
peaks and IG/ID1 (La = 16.90 Å)
Fig. 12. Baseline corrected HM-3077 spectrum fitted with Gaussian/
Band Intensity Full width half Gaussian Integrated Lorentzian hybrid function using three peaks.
position maximum (cm1) character intensity
(cm1)
1359.18 1.32 199.75 0.23 883.38 focused on evaluating a new technique, Raman spectros-
1537.2 0.39 94.83 1 122.33 copy, for field application rather than trying to accurately
1599.32 1.46 66.3 0.66 339.46 determine the asphaltene structures, which are rather com-
plex and inhomogeneous. For this focused goal, the
Raman performs well and should be considered a viable
9.90 and 16.89 Å for HH1, RSV, and HM-3077 samples, technique for this purpose.
all well within the estimated range and are consistent with
other literature values. As mentioned earlier, this work 5. Conclusion

Combination of Raman and X-ray spectrometry suc-

HH1 sample
cessfully characterizes the Hassi-Messaoud asphaltene
IIntensity (Arb. Unit)

molecular structures and their local ordering in solid state.

The estimated asphaltene molecular sheet dimension (hor-
izontal), La, was about in the range of 11–17 Å for the sam-
ples used in this study when using a three-peak method to
analyze the Raman spectra. The X-ray study of the DP45
sample gave 17.5 Å for La and 28 Å for Lc, which charac-
terizes the vertical dimension of the aggregates. The La val-
ues from both techniques are comparable. Using Lc and the
1000 1200 1400 1600 1800 inter-sheet spacing obtained from the X-ray spectrum, an
Wave number (cm-1) average of eight molecular stacking was estimated. The
intent of using Raman spectroscopy for asphaltene charac-
Fig. 10. Baseline corrected HH1 spectrum fitted with Gaussian/Lorentz- terization in field applications is well justified, at least to
ian hybrid function using three peaks. the first order of approximation. For a system as complex
as asphaltene, the results obtained here is significant. How-
ever, it should be substantiated in future investigation to
RSV sample
legitimize this technique for complex systems. Computa-
Intensity (Arb. Unit)

tional studies to simulating Raman spectra could be a lead,

when it is used in conjunction with Raman. With this, it is
possible to realistically characterize molecular structures
(or even the chemical structures) as complex as asphaltenes
[36,37].

Acknowledgements

1000 1200 1400 1600 1800

Funding of this work by le Ministère de l’Enseignement
Supérieure et de la Recherche Scientifique is gratefully
Wave number (cm-1) acknowledged. Authors would like to thank Mrs. Fergoug
Fig. 11. Baseline corrected RSV spectrum fitted with Gaussian/Lorentz- Teffaha for fruitful discussion; to A. Boubguira and Sekkal
ian hybrid function using three peaks. of Sonatrach Algerian Oil Company for providing the
1864 Y. Bouhadda et al. / Fuel 86 (2007) 1855–1864
samples and to Mr. D. Huchette, Mr. R. Desfeux and Mr.
M. Belalia for their technical assistance.
References
[1] Mullins OC, Sheu EY, editorsStructures and dynamics of asphalt-
enes. New York: Plenum Press; 1998.
[2] Speight JG. In: Yen TF, Chilingarian GV, editors. Asphaltenes and
asphalts, developments in petroleum science. Amsterdam: Elsvier;
1994.
[3] Anisimov MA, Yudin IK, Nikitin V, Nikolaenko G, Chermoustan A,
Thoulhoat H, et al. J Phys Chem 1995;99:9576.
[4] Trejo F, Ancheyta J, Centeno G, Marroquin G. Catal Today
2005;109:178–84.
[5] Groenzin H, Mullins OC. E&F 2000;14:677.
[6] Sheu EY. E&F 2002;16:74–82.
[7] Christopher J, Sarpal AS, Kapur GS, Krishna A, Tyagi BR, Jain MC,
et al. AK Bhatnagar Fuel 1996;75(8):999–1008.
[8] Mullins OC. In: Sheu EY, Mulins OC, editors. Asphaltenes,
fundamentals and applications. New York: Plenum Press; 1995
[Chapter 2].
[9] Toulhoat H, Prayer C, Rouquet G. Colloids and Surfaces A:
Physicochemical and Engineering Aspects 1994;91:267–83.
[10] Ungar T, Gubicza J, Ribarik G, Pantea G, Zerda TW. Carbon
2002;40:929–37.
[11] Fauteux C, Pegna J. Appl Phys A 2004;78:883–8.
[12] Darmstadt H, Roy G, Kaliaguine S, Xu G, Auger F, Tuel A, et al.
Carbon 2000;38:1279–87.
[13] Yen TF, Erdman JG, Pollak SS. Anal Chem 1961;33:1587–94.
[14] Dresselhaus MS, Dresselhaus G. Topics in applied physics, 51. Ber-
lin: Spring-Verlag; 1982. p. 3–57.
[15] Sadezky A, Muckenhuber H, Grothe H, Niessner R, Poschel U.
Carbon 2005;43(8):1731–42.
[16] Nestler K, Dietrich D, Witke K, Rößler R, Marx G. J Mol Struct
2003;661–662:357–62.
[17] Tuinstra F, Koenig JL. J Chem Phys 1970;53:1126.
[18] Bergmann U, Groezin H, Mullins OC, Glatzel P, Fetzer J, Cramer
SP. Chem Phys Lett 2003;369:184–91.
[19] Thomsen C, Reich S. Phys Rev Lett 2000;85:5214.
[20] Nemanich RJ, Solin SA. Phys Rev B 1979;20:392.
[21] Ferrari AC, Robertson J. Phys Rev B 2000;61:14095.
[22] Matthews MJ, Pimenta A, Dresselhaus G, Dresselhaus MS, Endo M.
Phys Rev B 1999;59:6585.
[23] Jawhari TA, Roid J, Casado J. Carbon 1995;33:1561–5.
[24] Tan P, An L, Liu L, Guo Z, Czerw R, Carroll DL, et al. Phys Rev B
2002;66:245410.
[25] Dresselhaus MS, Dresselhaus GS. Adv Phys 1981;30:290.
[26] Ferrari AC, Robertson J. Phil Trans R Soc Lond A
2004;362:2477–512.
[27] Andersen SI, Jensen JO, Speight JG. E&F 2005;19(6):2371–7.
[28] Bouhadda Y, Sheu EY, Bendedouch D, Krallafa A. E&F
2000;14:845–53.
[29] Mapelli C, Castiglioni C, Meroni E, Zerbi G. J Mol Struct
1999;480:615–20.
[30] Cançado LG, Pimenta MA, Neves BRA, Dnatas MSS, Jorio A. Phys
Rev Lett 2004;93:247401.
[31] Kawashima Y. Phys Rev B 2002;66:104109.
[32] Huang SM, Sun Z, Lu YF, Hong MH. Appl Phys A 2002;74:
515.
[33] Dippel B, Heintzenberg J. J Aerosol Sci 1999;30(suppl. 1):907.
[34] Rigolio M, Castiglioni C, Zerbi G, Negri F. J Mol Struct
2001;563:79–87.
[35] Tanaka R, Sato E, Hunt JE, Winans RE, Sato S, Takanohashi T.
Energy Fuels 2003;18:1118–25.
[36] Nestler K, Dietrich D, Witke K, Robler R, Marx G. J Mol Struct
2003;661:357–62.
[37] Maultzsch J, Reich S, Thomsen C. Phys Rev B 2004;70:155403.