Aggregation States of Asphaltenes - American Chemical Society

Aug 19, 2009 - Bernard - Lyon 1, Sciences Analytiques, CNRS UMR 5180, Domaine scientifique de la Doua, ESCP. E-Lyon, 69622 Villeurbanne, France...
18 downloads 0 Views 2MB Size
16266

J. Phys. Chem. C 2009, 113, 16266–16276

Aggregation States of Asphaltenes: Evidence of Two Chemical Behaviors by 1H Diffusion-Ordered Spectroscopy Nuclear Magnetic Resonance Emmanuelle Durand,†,‡ Martin Clemancey,‡ Jean-Marc Lancelin,‡ Jan Verstraete,† Didier Espinat,† and Anne-Agathe Quoineaud†,* Institut Franc¸ais du Pe´trole (IFP), BP 3, 69360 Solaize, France, and UniVersite´ de Lyon, UniVersite´ Claude Bernard - Lyon 1, Sciences Analytiques, CNRS UMR 5180, Domaine scientifique de la Doua, ESCP E-Lyon, 69622 Villeurbanne, France ReceiVed: March 3, 2009; ReVised Manuscript ReceiVed: July 10, 2009

The mean molecular weight and particle size of asphaltene aggregates extracted from Buzurgan feedstock have been evaluated by measuring the diffusivity by means of 1H diffusion-ordered spectroscopy (DOSY) nuclear magnetic resonance (NMR). This technique, recently applied in the petroleum industry, appeared as a key tool to investigate the behavior of asphaltenes diluted in toluene over a wide range of concentrations (from 0.01 to 15 wt %). The results show that the diffusivity is highly dependent upon solute concentration. Indeed, interactions occurring in dilute systems differ from interactions occurring in the semi-dilute regime. In the dilute regime (below 0.25 wt %), physical characterization of the nanoaggregates detected by 1H DOSY NMR could be achieved. An average molecular weight of roughly 6900 g mol-1 (taken at the highest diffusion peak and given as a polystyrene equivalent) with a range of 1500-85000 g mol-1 was obtained, while a mean radius of 15.6 Å was determined from the solute diffusivity at infinite dilution [D∞asp ∼ (2.4 ( 0.1) × 10-10 m2 s-1 for Buzurgan asphaltenes]. Average masses (Mn and Mw) were also calculated from DOSY NMR data and compared to results obtained from size exclusion chromatography (SEC) analyses. We also used DOSY NMR techniques to investigate the molecular dynamics of asphaltenes. A concentration of 0.25 wt % was found to represent the onset of the aggregation process. We believe DOSY NMR allowed us to observe the beginning of aggregation in phase transition. The asphaltene system is very polydisperse. At low concentrations, it is a polydisperse diluted system, but increasing the solute concentration induces a distinction of different aggregates, presenting various sizes in a macroscopic homogeneous phase. When the distinction occurs, there is a zone that is richer in high molecular weight aggregates of asphaltenes, and as a result, a second zone is impoverished in high molecular weight aggregates. In a microscopic point of view, there is a difference in nano or macroaggregate concentration, whereas in a macroscopic point of view, an average density is observed. For the first time, a clear separation between two families of aggregates of asphaltenes is presented in the diffusion dimension for concentrations higher than 3 wt %. 1H DOSY spectra and diffusion profiles confirm these results. The key point of this study resides in the detection and presentation of two classes of aggregates of asphaltenes achieved for concentrated solutions, without any assumption concerning the composition of the mixture. 1. Introduction New energy supplies need to be developed to face the growing worldwide energy demand. Heavy crude oils have attracted attention because they represent an important energy reserve. They are becoming increasingly more important in terms of exploration and upgrading. Heavy oil fractions can be upgraded through various processes, among which is hydroprocessing. The main problem to be concerned with during hydroprocessing is the presence of asphaltenes. Asphaltenes are macromolecules composed of highly polyaromatic clusters surrounded and interconnected by alkyl chains. They also contain some heteroatoms (N, S, and O) and metals in trace amounts (Ni and V) in proportions depending upon the origin of the sample.1-3 Vanadium and nickel are the most abundant metals and are often present as chelated porphyrin complexes, * To whom correspondence should be addressed. Telephone: +33-478-02-29-63. Fax: +33-4-78-02-27-45. E-mail: A-Agathe.QUOINEAUD@ ifp.fr. † Institut Franc¸ais du Pe´trole (IFP). ‡ Universite´ de Lyon.

which are known to be responsible for catalyst deactivation during hydrotreatment processes.4 Asphaltenes are defined as a solubility class of components; they stand for the portion of crude oil that is insoluble in light n-alkane (e.g., n-pentane or n-heptane), while being completely miscible with toluene. Over the years, a great deal of attention has been given to asphaltenes. They represent the most enigmatic components,5 but little is precisely known. Even basic scientific data such as molecular weight, size, shape, and molecular structure still represent big issues for asphaltene characterizations and are not yet clearly established. The complexity of the study is due to the nature of asphaltenes themselves; they are composed of polydisperse entities in terms of molecular weight and chemical composition. Asphaltenes are known to self-associate and aggregate leading to severe problems during transport and upgrading.6-9 They clog the porous media of reservoir rocks10 and are as a result responsible for a reduction in oil flow. Their stability depends upon several factors, including temperature, pressure, and concentration.11-13 Asphaltenes form nanoaggregates at very low concentration (c < 0.2 g L-1)14,15 and can

10.1021/jp901954b CCC: $40.75  2009 American Chemical Society Published on Web 08/19/2009

Aggregation States of Asphaltenes further cluster into larger species or even flocculate when concentration increases. The influence of chemical composition is also not well-defined. Many attempts have been made to relate their reactivity to their chemical structure. As a result, it is imperative to obtain a better understanding of these species to develop new catalysts and novel conversion processes. Asphaltene characterization is still a challenge because they are composed of many heavy molecules. Their structure can vary significantly, depending on their origin and method of recovery. Many analytical techniques are commonly used to analyze asphaltenes: size exclusion chromatography (SEC), laser desorption mass spectrometry (LDMS), matrix-assisted laser desorption ionization (MALDI), small-angle X-ray scattering (SAXS), small-angle neutron scattering (SANS), calorimetric measurements, vapor pressure osmometry (VPO), etc.3,6,13,16-25 However, it is difficult to establish their physicochemical properties by conventional techniques26 as they tend to form aggregates, depending upon their concentration and solvent employed. Nuclear magnetic resonance (NMR) is widely used to analyze petroleum samples.27-30 Carbon aromaticity and average carbon parameters31 can be measured. Nevertheless, the proton and carbon spectra hardly enable the establishment of a chemical structure because the onedimensional petroleum spectra are crowded and overlapped because of the presence of numerous aromatic, naphthenic, and aliphatic functions. Two different structural models have been proposed for asphaltenes. According to the continental32 model also called the “island model” or the “like your hand model”, asphaltenes are shaped as a condensed aromatic molecule composed of a single aromatic core substituted by some alkyl chains. However, the “archipelago1,19,33-35 model” represents asphaltenes as several small fused ring systems interconnected by some alkyl chains and thioether bridges. The development of new NMR techniques such as diffusionordered 2D-NMR spectroscopy (DOSY),36 which stands for a high-resolution version of the pulsed field gradient (PFG) sequences, have been used for complex and heterogeneous mixtures such as biological samples37-39 and polymers,40,41 for measuring protein-ligand interactions42 or even complexes,43,44 and more recently for petroleum fractions.45-47 We have published46 recently that DOSY is a suitable tool to analyze asphaltenes. This new technique is expected to provide valuable information concerning the physicochemical properties of asphaltenes because it is sensitive to molecular weight and structure. In a recent work, Lisitza et al.47 used DOSY NMR techniques to investigate the molecular dynamics of the initial state of asphaltenes aggregation. They studied asphaltenes at various low concentrations ranging from 0.05 to 2.1 g L-1 in toluene and showed a kink in the diffusion measurements near 0.2 g L-1. From the diffusion coefficient for concentrations below 0.2 g L-1 (D ) 2.9 × 10-10 m2 s-1), the authors extracted an average monomer radius of 1.2 nm assuming a spherical particle. Structural and dynamic information (molecular size and aggregation states) of complex mixtures can be extracted from the measurement of self-diffusion. Many techniques can be used to measure diffusion coefficients [pulsed field gradient stimulated echo NMR (PFGSE NMR) and fluorescence correlation spectroscopy]48,49 and to study size and shape (SAXS and SANS).2,50-53 DOSY presents a huge advantage by giving a physical and chemical characterization in one experiment. Some authors26,49,54 have already used PFGSE NMR sequences to analyze asphaltenes, but they have neither shown nor used the bidimensional spectrum of the DOSY sequence. The second

J. Phys. Chem. C, Vol. 113, No. 36, 2009 16267 TABLE 1: Detailed Characterization of Buzurgan Feedstock and Asphaltenes Buzurgan feedstock

property C7 asphaltenes content (NF T60-115) Conradson carbon content d15.4 viscosity at 98 °C SARA analysis saturated aromatics resins asphaltenes

Buzurgan asphaltenes

asphaltenes elemental analysis carbon hydrogen nitrogen oxygen sulfur nickel vanadium

value 11.8 wt % 22.0 wt % 1.037 2670 cSt 9.9 wt % 37.1 wt % 35.3 wt % 15.3 wt % value 81.22 wt % 7.21 wt % 1.09 wt % 0.87 wt % 8.04 wt % 251 ppm 801 ppm

dimension offers a rich diffusion spectrum, from which data on the complex composition of the mixture can be obtained. We showed the first DOSY spectrum of an asphaltene sample46 to demonstrate the potential of this type of sequence to obtain physicochemical data. In the current work, we report 1H DOSY NMR experiments on unconverted asphaltenes from Buzurgan feedstock diluted in toluene-d8 and show the influence of sample concentration on the diffusion coefficient. Results demonstrate that intermolecular interactions between solvent and solutes are very dependent upon asphaltene concentration. Asphaltene 1H-DOSY spectra and diffusion profiles demonstrate the separation achieved by the two main classes of aggregates of asphaltenes above a given concentration. Analyses of polystyrene samples were used as a calibration for the molecular weight determination. 2. Experimental Methods 2.1. Polystyrene Samples. Polystyrenes (PS) (162, 370, 580, 970, 1310, 2970, 5030, 19880, 70950, 120000, and 316500) from Polymer Laboratories (Varian, Inc.) were analyzed as a molecular weight calibration methodology. All polymers were used without any further purification. 2.2. Petroleum Samples. Distillation is the method commonly used to fractionate petroleum samples according to hydrocarbon volatility. The heaviest fraction that remains after distillation (ASTM D5236 and D2892) is called the vacuum residue. The components belonging to this residue fraction typically have a boiling point higher than 550 °C. In the present study, a Buzurgan vacuum residue feedstock obtained from a crude oil vacuum distillation was used. This feedstock is representative of vacuum residues from Middle East crude oils. Asphaltenes were prepared by precipitation in an excess of n-heptane at 80 °C, according to a standard analytical procedure derived from the NF T60-115 method. Table 1 presents the Buzurgan feed characteristics and elemental analyses of asphaltenes. Mass balances were performed to ensure complete solvent removal. The asphaltenes content amounted to 15.3 wt %. 2.3. DOSY NMR Measurements. 2.3.1. NMR Parameters. Asphaltenes were diluted in perdeuterated toluene (99.8% D) supplied by Eurisotop. A wide

16268

J. Phys. Chem. C, Vol. 113, No. 36, 2009

Figure 1. Calibration curve relating molecular mass (MM) and relative self-diffusion of polystyrenes (PS) analyzed at 1 wt % in toluene-d8 (solvent stands for the internal reference).

range of concentrations (from 0.01 to 15 wt %) in toluene-d8 were prepared and added to 5 mm NMR tubes. DOSY NMR experiments were performed on a Unity-Varian 600 spectrometer operating at a 1H frequency of 600 MHz equipped with a 5 mm 1H-X reverse z-axis gradient probe capable of generating 60 G cm-1 field strengths. The Doneshot55 sequence was used to measure solutes and solvent self-diffusion with an R of 0.2. This sequence will be detailed in a following paragraph. The gradient pulse strength g was varied in 50 linear steps from 0 to 60 G cm-1 to obtain 95% signal attenuation. Gradients were calibrated against the HOD diffusion constant at 25 °C (D2O, 99.9%, D ) 19.0 × 10-10 m2 s-1)56,57 as recommended by Price.57 The calibration was confirmed by measuring adenosine triphosphate (ATP), sodium dodecyl sulfate (SDS), and glucose diffusion coefficients mixed in D2O at 22 °C.39 These molecules, heavier than n-heptane or n-dodecane (commonly used as references for gradients calibration), will better reflect heavy petroleum molecules than the n-alkane solvents. Effective temperature was checked against methanol first and then ethylene glycol, and the linearity of the gradients was checked against sodium dodecyl sulfate. According to the authors, the Doneshot sequence (Figure 1, dotted line; ref 55) presents a potential advantage where static and radio frequency field inhomogeneities are significant. In the Doneshot sequence using a stimulated echo, the gradient pulse duration (δ) and the diffusion delay (∆) were kept constant, ranging between 1.5 and 4 ms for δ and between 0.2 and 0.35 s for ∆. Spectra were acquired at 20 °C with a 90° pulse duration of 6.38 µs and a relaxation delay of 30 s. 2.3.2. Software AWailable for DOSY Processing. There are two different modes of processing DOSY data. In univariate processing, the signal attenuation is analyzed separately at each frequency.58 It includes mono and bi exponential fittings and the spline model (SPLMOD), continuous distributions of exponentials (CONTIN), and maximum entropy (MaxEnt). Nevertheless, multivariate methods deal with the covariance between the signals belonging to a specific component and resolution of the entire spectrum of each component in the mixture. Direct exponential curve resolution algorithm (DECRA), component resolved (CORE), and multivariate curve resolution (MCR) belong to this type of processing. Published DOSY spectra generally use either mono- or multiexponential fittings. One of the main advantages of the fitting method is the application of least-squares fitting leading to

Durand et al. simple algorithms and as a result to fast processing. However, such process techniques cannot be applied to highly complex mixtures like ours because they require an assumption for the number of components. In the case of SPLMOD which is the successor to DISCRETE,59 the user has to specify the maximum number of discrete components present in the mixture. In case of a doubt on the nature or composition of the solution, an analysis method handling continuous distribution should be performed.60 Our samples are composed of thousands of unknown polydisperse constituents. CONTIN61,62 could be used to process the DOSY NMR spectra of polydisperse samples. CONTIN has been the first software used for the analysis of DOSY in the case of polydisperse samples.39 The user has only to specify a threshold value. However, no information on the composition of the mixture is required, which represents a great advantage in our case where no information on the nature of the sample is known. Another key point is that CONTIN can be used both for mono and polydisperse samples. Nevertheless, this software shows a lack of resolution to enable a separation of the different species presenting similar sizes.60,63 MaxEnt generally competes with CONTIN for the analysis of polydisperse samples as no information on the composition of the sample is required. The maximum entropy (MaxEnt)63 Laplace inversion is an implementation of the NPK64 sotware developed by Delsuc. This module gives access to the whole diffusion spectrum without any knowledge of the number of species present in the sample and is therefore well-adapted for polydisperse mixtures. The goal is to build the inverse Laplace transform by maximizing the entropy of spectral distribution and to compare it with the real spectrum. According to its authors, MaxEnt appears to be more efficient than CONTIN for weak components analysis.63 It was reported that CONTIN showed some inadequacy in finding the correct number of components because of oversmoothing. According to the authors,63 this software provides a high quality spectrum over a wide range of experimental data ensuring resolution, accuracy, and reliability of results. MaxEnt seems to be more appropriate than CONTIN for analysis of heavy polydisperse complex mixtures, where no assumptions are required except possibly a threshold level. The problem of resolving exponential components can be limited, if we managed to find isolated NMR peaks for individual species. The ideal case is to have enough resolution to separate the different peaks. However, in the case of real samples, overlapping frequently occurs. In such cases, multivariate methods may be useful as they exploit all available signal covariance in order to resolve the spectra of the components of the mixture.58 The most common multivariate methods used for DOSY analysis are DECRA58,65 and CORE.65 Multivariate analysis of DOSY data are not appropriate for the analysis of DOSY as these algorithms require a certain amount of user input from which the number of components the algorithm should fit the experimental data. Asphaltenes are complex mixtures for which few things are precisely known. Indeed, asphaltenes are composed of thousands of compounds, but an exact number cannot be given. 2.3.3. Software Used. NMR data were processed with NMRnotebook (from NMRtec) with a DOSY module (MaxEnt algorythm). Asphaltenes are composed of thousands of polydisperse heavy compounds. The number of components is not precisely known. We selected MaxEnt because no prior knowledge on the composition of the mixtures is required for polydisperse sample analysis. MaxEnt is also incorporated into the NNB software developed in Java and OpenGL66 with an

Aggregation States of Asphaltenes easy interface for the user. As a result, MaxEnt presently seems to be the compromise to reach our objectives. Other softwares have already proved to be powerful, but they are not adapted for our real petroleum complex mixtures. The reported diffusion coefficients are taken at the highest intensity distribution peak. 2.4. Size Exclusion Chromatography (SEC). SEC was performed on a Waters 150CV+ system, using a refractive index detector. The system was controlled using a Millenium chromatography manager. Columns packed with polystyrene-divinylbenzene supports (PS-DVB, Polymer Laboratories) were chosen. Corresponding porosities are 100, 1000, and 10000 Å, and column characteristics are as follows: packing particle size, dp ) 5 µm; column length, L ) 300 mm; and internal diameter, 8 mm. Ten monodisperse polystyrene standards with masses ranging from 162 to 120000 g mol-1 were used to perform the molecular weight calibration. The Buzurgan asphaltene sample was injected at a concentration of 5 g L-1 in tetrahydrofuran (THF) with a volume of 50 µL. The temperature was fixed at 40 °C, and the flow rate was fixed at 0.7 mL min-1. The SEC data provided in this work is given as Mw ) (∑NiMi2)/(∑NiMi) and Mn ) (∑NiMi)/(Ni), where Ni represents the number of molecules with a molecular weight of Mi (given in units of g mol-1). The choice of these operating conditions has been described elsewhere.67 2.5. 1H and 13C Nuclear Magnetic Resonance (1H and 13C NMR). Asphaltene structural parameters were measured by proton 1H and carbon 13C NMR experiments performed on an Avance 300 MHz Bruker spectrometer, using a 10 mm BBO 1 H/X/D NMR probe. A total of 100 mg of asphaltenes were diluted in 3 mL of CDCl3 to obtain a homogeneous solution. The chemical shifts were referenced using CDCl3 as the solvent. The attached proton test (APT) series was carried out to identify and quantify the proportion of atoms of carbon as a function of the number of protons in their neighborhood. Two 13C NMR experiments, based on the scalar coupling between protons and carbons, were carried out in order to obtain some data on paraffinic, naphthenic, and aromatic carbon atoms. 13C NMR direct acquisition spectra were carried out with a 60° flip angle at a radio frequency pulse of 20 kHz. These experiments provided quantification of saturated and unsaturated C atoms. Moreover, spin-echo experiments were set to quantify aromatic and aliphatic species separately. 3. Theoretical Basis 3.1. DOSY NMR Theory. Diffusion ordered spectroscopy (DOSY)36,60 is a method devised by Morris and Johnson,36 which is a high-resolution version of the pulsed field gradient stimulated echo nuclear magnetic resonance (PFGSE NMR) sequence of Stejskal and Tanner68 based on the nuclear spin-echo concept of Hahn69 and Carr and Purcell.70 DOSY NMR experiments allow analytical separation and identification of the mixture constituents. It aims at measuring the mobility of each component present in a mixture. Diffusion coefficients are sensitive to size (and molecular weight), shape, and aggregation status. The spectrum obtained from such an experiment is a bidimensional spectrum in which the first dimension (x-axis) is a conventional NMR spectrum (chemical shift), whereas the diffusion coefficients are presented along the y-axis. One of the sequences available for Varian Inova spectrometers is the Doneshot sequence.55 On the basis of a PFGSE sequence, it is described by its authors as a powerful technique for the analysis of mixtures. The Doneshot sequence presents several advantages compared to the bipolar pulse pair

J. Phys. Chem. C, Vol. 113, No. 36, 2009 16269 stimulated echo (BPPSTE), from which it derives. First of all, diffusion measurements can be performed in just one shot and with only one transient per gradient strength without any phase cycling, which enables a quite short time of analysis. The sequence comprises extra balancing gradient pulses in addition to asymmetric bipolar field gradient pulse pairs in order to select a unique coherence transfer pathway and to minimize at the same time the eddy current effect. Some gradients have been added at the beginning and end of the diffusion delay to keep the deuterium lock signal focused and to minimize field frequency lock disturbance. It aims at maintaining the system constant. A thorough description of the sequence is given by Pelta et al.55 In practice, while increasing gradient strengths, the intensity of the signal is attenuated because of diffusion according to an exponential function. Thus, the diffusion coefficient (D) is related to the signal amplitude (I), while applying gradient (g) in the z direction and to the signal amplitude at zero gradient (I0) in the same direction through eq 155 derived from the Stejskal-Tanner equation

I ) I0e-Dγ g δ [∆+δ(R -2)/6+τ(R -1)/2] 2 2 2

2

2

(1)

where δ and ∆ are the width of the field gradient pulses and the diffusion delay, respectively; γ is the magnetogyric ratio (2.675 × 108 rad T-1 s-1 for 1H); R is the imbalance factor; and τ is the time between the midpoints of the individual gradient pulses in one diffusion-encoding period. Translational self-diffusion coefficients (D) of the different compounds of the mixture are extracted from this equation. 3.2. Physical Information Extracted from D. The StokesEinstein equation is often used to extract the hydrodynamic radius RH of the solute from its diffusion coefficient D

D)

kBT 6πηRH

(2)

where kB is the Boltzmann constant, T is the absolute temperature, and η is the solvent viscosity. However, this equation was established for spherical particles that are much larger than the solvent molecules (RH > 5Rsolv), which is not obviously the case here. Thus, the shape and size of the solute must also be taken into account. The modified Stokes-Einstein equation71 is therefore

D)

kBT c(Rsolv, RH)fs(a, b)πηRH

(3)

First of all, the shape and the size of the solute should be taken into account through fs in which the ratio between the major (a) and the minor (b) semiaxis of the ellipsoid is considered. The expression for fs varies depending on whether prolate (cigarlike) or oblate (disk-like) ellipsoid molecules71,72 are considered and is given by eqs 4 and 5, respectively

16270

J. Phys. Chem. C, Vol. 113, No. 36, 2009

Durand et al.

For prolate ellipsoid molecules fs )

 () b 1+ 1-( )  b a ( a ) × ln b

(

2/3

b 1a

2

2

(a)

)

(4)

For oblate ellipsoid molecules b -1 (  a) f ) ( ab ) arctan( ab ) - 1 2

s

2/3

(5)

2

The difficulty with petroleum samples resides in the number of components: thousands of heavy hydrocarbons with different shapes. An average spherical particle is therefore assumed, and fs is set equal to 1. The simplification is obviously idealized, but establishing an exact fs value is impossible given the complexity of the system. It has also been proposed to correct for the actual difference in size of the solute and solvent molecules. Hence, the modified Stokes-Einstein equation contains a correction factor c, which depends upon the ratio of the radius of the solvent (Rsolv) to that of the solute RH,71,73 which is equal to

c)

6 1 + 0.695 × (Rsolv/RH)2.234

(6)

The second approach consists in determining a molecular weight from diffusion measurements. A well-known relation41,74 relating the diffusion coefficient D to the molecular weight M has been established according to the Flory scaling law (Rg ∝ M-R, where Rg is the radius of gyration) valid in the dilute regime and to the Stokes-Einstein equation (for a spherical particle, Rg ∝ RH):

D ) KM-R, R > 0

(7)

where K is a constant dependent upon the nature of the molecule and R a coefficient, termed shape factor, which depends upon the shape of the particle.75 From a theoretical point of view, R should be 0.33 for a spherical particle, whereas a value of 0.5 for R would reflect a random coil oligomer in a θ solvent or a flat disk. In the case of a good solvent, R should be equal to 0.6.76 For this method, the use of an internal reference provides several benefits. In our case, the solvent was chosen as internal reference. Diffusion analysis can be done on the solvent and solute in a single experiment. It improves the accuracy without additional experimental time.77 A relative diffusivity as called by Jones et al.78 is defined as the ratio of the diffusion coefficients of the solute and solvent.

Drel )

Dsolute Dsolvent

(8)

This approach allows reducing the impact of viscosity or of any temperature variation and is hence thought to provide more robust data.

4. Results and Discussion 4.1. Molecular Mass Calibration to Polystyrene. In order to estimate an average molecular weight for asphaltene samples, a polystyrene (PS) calibration curve was established. As a result, in the following sections, molecular weights from DOSY measurements will be given as equivalent polystyrene and are therefore comparable to SEC data. Accurate 1H DOSY NMR experiments were performed on PS polymers, covering a large range of molecular weights (ranging from 162 to 316500 g mol-1). In the present study, the polymers were studied at a concentration of 1 wt % in toluene-d8 to get a good signal-tonoise ratio and to be in a diluted state in order to apply Stokes-Einstein and Flory laws. Figure 1 shows a relative diffusivity-molecular mass calibration curve in a doublelogarithmic plot generated from experimental results. As expected, the data obtained for PS fall on a straight line

(

Dsolute Msolute ) 10-0.144 × Dsolvent Msolvent

)

-0.42(0.02

(9)

where D is the self-diffusion coefficient and M is the molecular weight. In our experiments, R is found to be 0.42 ( 0.02. According to the Stokes-Einstein equation, R is supposed to be 0.33 for a spherical particle, whereas a value of 0.5 for R would reflect a linear particle in a θ solvent and obtaining a value between 0.33 and 0.5 is not surprising. This value reflects a soft interpenetrable sphere as defined by Pyun and Fixman.79 This equation is subsequently used as a calibration curve in order to determine an average molecular weight of the asphaltenes studied. Such scaling laws have been recently observed for asphaltenes for other measured parameters:50Rg ) 0.43Mw0.45 and [η] ) 0.049Mw0.41. Thus, assuming a shape factor of 0.42 for asphaltenes is adequate. It is also in agreement with some previous work.80 4.2. Asphaltenes Chemical Structure According to 13C and 1 H NMR Data. The characterization of average molecular features has been performed using a 13C NMR experiment (direct observation) and a series of attached Proton Test (APT) NMR81,82 and spin-echoes. The direct 13C method is used to quantify the different types of carbon present in the sample. It is possible to determine the total amount of aromatic (Caro) and aliphatic (Cali) carbon species by integration of the 160-100 ppm and 70-0 ppm chemical shift areas, respectively. The quantity of unsubstituted carbons CH (CHaro) and quaternary carbon (Cq,aro), either condensed or substituted, present in the aromatic portion is calculated using the integration of the 160-100 ppm region of the 13C spin-echo and 13C. Furthermore, the APT NMR method based on the scalar coupling between a carbon and a proton was performed to determine the relative percentages of aromatic, naphthenic, and paraffinic carbons. This sequence allows us to distinguish carbon close to an even (quaternary carbons and methylene groups) or an odd (methyne and methyl groups) proton number. APT, corresponding to 1/2JCH with JCH ) 160 Hz is characteristic of protonated aromatic carbons. The percentages of CH3, CH2, CH, and quaternary carbon (Cq) present in the aliphatic components are measured using integration of the 70-0 ppm region of the 13 C spin-echo and 13C APT NMR, corresponding to 1/2JCH and 1/JCH with JCH ) 125 Hz, which is typical of sp3 carbons. Using the amount of total aromatic carbon and detailed percentages of CH and Cq, it is possible to calculate the substitution index. The substitution index Is is defined as

Aggregation States of Asphaltenes

J. Phys. Chem. C, Vol. 113, No. 36, 2009 16271

TABLE 2: Buzurgan Asphaltenes 13C NMR Data type of carbons

percentage (%)

Cali CH3 CH2 CH Cquat Caro CH Cq Cq,sub Cq,cond Is Icp Icc Cali/Caro

46.8 26.3 59.0 14.1 0.0 53.2 21.4 78.6 35.7 42.9 0.6 5.1 12.8 0.88%

Figure 2. 10 wt % Buzurgan asphaltenes proton spectrum in toluened8 at 20 °C.

the ratio between the substituted and substitutable aromatic carbons.

Is )

Cq,sub Caro - Cq,cond

(10)

It is also possible to calculate the condensation index Ic, which represents the average number of condensed aromatic rings. However, there are two different condensation indices Ic: the peri-condensated index Icp and the cata-condensated index Icc defined as follow Figure 3. Buzurgan asphaltene 1H-DOSY spectrum analyzed at 0.1 wt % in toluene-d8 at 20 °C.

Cq,cond 1+2× Caro Icp ) Cq,cond 1-3× 2Caro

(11)

Cq,cond Caro Icc ) Cq,cond 1-2× Caro

(12)

1+2×

where Cq,cond stands for the amount of quaternary condensed aromatic carbons, Cq,sub corresponds to the amount of quaternary aromatic carbons that are substituted, and Caro corresponds to the total aromatic carbons. The main NMR data of the asphaltenes are presented in Table 2. The first thing to note is that the Buzurgan asphaltenes contain more aromatic than aliphatic carbons. If we focus on the aromatics first, we see that most of the aromatic carbons are substituted or condensed. However, 13C NMR also enables the characterization of aliphatic species. There are either many alkyl chains (which would be consistent with Cq,sub and Is) or the alkyl chains are highly substituted (high CH3 and CH percentage). Moreover, we suppose that asphaltenes are composed of a mix of peri- and cata-condensed forms. Taking this hypothesis into account, the number of aromatic rings in an asphaltene entity should be between 5 (pure peri-condensated form) and 13 (pure cata-condensed form). Furthermore, looking at the 1H spectrum (Figure 2) of an asphaltene sample analyzed at 10 wt % in toluene, it is possible to distinguish asphaltene signals from solvent signals. In fact,

the sharpest peaks are from toluene, whereas the broad spectral feature around 1-2 ppm is from the aliphatic protons of asphaltenes. There are few aromatic protons as expected for asphaltene molecules whose aromatic cores are highly substituted. As mentioned above, asphaltene peaks are very broad, leading to a reduction of their intensities, especially for aromatic signals which cannot be differentiated from toluene signals. 4.3. Concentration Effect on Asphaltene Behavior. As shown in Figure 2, the proton spectrum of an asphaltene sample is crowded because of the overlapping signals from the solvent and solute but also because of the complexity of the sample. DOSY spectra allow a better resolution in the diffusion dimension, enabling separation between the solvent and oil molecules. The 1H-DOSY spectrum of a Buzurgan asphaltene analyzed at two different concentrations, 0.1 and 10 wt % in toluene-d8, are shown in Figures 3 and 4. At first, it is possible to separate the toluene signal from those of asphaltenes. Some residual heptane signals have also been detected for the highest concentration and are present at very low concentration. The presence of heptane is due to the asphaltene extraction method used (derived from the NF T60-115 method). If we focus on the asphaltene signals, a molecular weight dispersion can be noticed and is more pronounced for the 10 wt %. It shows that at higher concentration asphaltenes are very polydispersed. Theoretically, it should be possible to isolate all components according to their mobility, but asphaltenes are composed of too many components. However, for this type of asphaltene, two different classes of aggregates have been detected above 3 wt %. Figure 4 shows

16272

J. Phys. Chem. C, Vol. 113, No. 36, 2009

Durand et al.

Figure 6. Relative diffusivity of Buzurgan asphaltenes in toluene-d8 as a function of solute concentration. Figure 4. Buzurgan asphaltene 1H-DOSY spectrum analyzed at 10 wt % in toluene-d8 at 20 °C.

Figure 5. Distribution of diffusion profiles extracted from the 1HDOSY spectra of Buzurgan asphaltenes analyzed at 0.1 (amplified 10 times), 3, and 10 wt % in toluene-d8.

the DOSY spectrum when the separation between two main classes of asphaltene aggregates is achieved. No DOSY spectrum showing two different species of asphaltenes aggregates has been found in the open literature. We also present the projection of the diffusion dimension of DOSY spectra taken on the aliphatic peaks (taken between 0 and 5 ppm) of asphaltenes analyzed at 0.1, 3, and 10 wt % to confirm the results (Figure 5). Intensities correspond to arbitrary units and are not normalized. The smallest peak belonging to the 0.1 wt % sample has been amplified 10 times compared to the other samples. Only one peak is observed at this concentration, whereas at 3 wt % the distribution peak is composed of two shoulders attributed to nanoaggregates (Figure 5, right) and macroaggregates (Figure 5, left), and at 10 wt % two distinct peaks are detected. The broad peak observed at 3 wt % is split into two different peaks, with a clear separation at 10 wt %. It is obvious that the distribution of the asphaltene diffusion coefficient is very wide, which implies a wide range of different aggregates. The DOSY spectrum presented in Figure 4 as well as the diffusion distribution (Figure 5) show two well-separated zones of petroleum species besides the solvent signal. These two figures illustrate the key advantage of 1H-DOSY: it has been possible to isolate two different asphaltene classes that were overlapped in the 1H-proton spectrum, without making any assumption on the sample composition. This last point is very important as the composition of asphaltene molecules is not yet clearly established.

The relative diffusivity (Dsample/Dsolvent) of Buzurgan asphaltenes as a function of solute concentration is presented in Figure 6. The samples were diluted in a solution of toluene-d8 and analyzed at 20 °C. A wide range of concentrations has been covered (0.01-15 wt %) and is presented here. Average diffusion coefficients have been obtained on aliphatic peaks ranging from 0.7 to 5 ppm for the asphaltene samples and on aromatic peaks for the solvent (toluene) as shown on the proton spectrum (Figure 2). Both diffusion coefficients (solvent and sample) strongly depend upon asphaltene concentration.83 We decided to show the relative diffusivity (with the solvent as the internal reference) to avoid viscosity and temperature dependence. The relative diffusivities remain constant until the concentration reaches 0.25 wt % corresponding to a dilute state. In dilute solutions, the entities are isolated enough not to be influenced by other molecules. In this range of concentrations, solute-solvent interactions are predominant and determine the limiting diffusion coefficient [D∞tol ∼ (19.57 ( 0.05) × 10-10 m2 s-1 for toluene, D∞asp ∼ (2.4 ( 0.1) × 10-10 m2 s-1 for Buzurgan asphaltenes]. The value obtained for D∞tol agrees well with the diffusion coefficient of pure toluene-d8. As the concentration increases, obstruction effects strongly modify the diffusing properties of solvent molecules. This can dramatically influence the solvent diffusion capacity. The measured self-diffusion coefficient of the solute depends strongly and nonlinearly upon all types of intermolecular interactions:84 attractive solvent-solute interactions and attractive solute-solute interactions. In fact, there is a continuous decrease in diffusion coefficient beyond 0.25 wt %, which is attributed to a change of regime. The range of concentrations of 0-0.25 wt % stands for a dilute regime, whereas beyond 0.25 wt % there is a semidilute regime. This behavior illustrates that intermolecular interactions strongly depend upon solute concentration. A mean average diffusion of the asphaltene molecules was measured. For example, the sample at 0.01 wt % presented only a wide distribution of diffusion constants from (0.8 ( 0.1) to (4.5 ( 0.1) × 10-10 m2 s-1. These values are in good agreement with what Freed et al.85 obtained [D ranging from (0.2 ( 0.1) to (5 ( 0.1) × 10-10 m2 s-1].85 In the dilute regime, no drop of the diffusion coefficient was observed as shown by Lisitza et al.47 (Figure 4, ref 47). Two reasons may explain this observation. First, we assume that this is because we worked just above the critical nanoaggregate concentration (CNAC),14 which is supposed to be at approximately 0.010-0.020 wt %.14,86-88 These values were determined from experiments based on

Aggregation States of Asphaltenes ultrasonic sound velocity measurements, ac electrical conductivity, dc electrical conductivity, and NMR. Second, this difference may be due to the nature of the asphaltenes themselves. Asphaltenes are indeed known to present differences, depending on the origin of the sample. Moreover, Freed et al.85 presented two different types of species for dilute samples in toluene-d8 over all of their studied concentration range (c < 0.15 wt %). The authors used the terms of fast and slow components, but it was the fast and slow components of the double exponential fit of the diffusion decay. However, in their latest article,47 they only showed one type of species. More recently, Kawashima et al.54 also presented the existence of two types of species at concentrations of 0.1 (∼0.007 wt %) and 1 g L-1 (∼0.07 wt %) in CDCl3 but not for the highest concentrations studied (10 and 30 g L-1, which correspond to 0.7 and 2 wt %, respectively). In their paper,54 these authors used a two-state model as a simple model to consider the distribution observed. They approximated the results by a biexponential fit. In our studies, we did not detect two types of asphaltenes in the dilute regime, but then the solvent used and concentration levels investigated differ, which may explain why the observations are different. Even though in our studies only a continuous distribution of asphaltene aggregates was detected in the dilute regime, two different classes of asphaltene aggregates, termed nanoaggregates and macroaggregates, were detected beyond 3 wt %. As mentioned previously, although some authors have found two different asphaltene types for dilute samples (c < 0.1 wt %),54 this has never been shown for concentrated solutions. The innovative contribution of our work resides in the detection of two different types of asphaltenes, without making any assumption on the composition of the mixture. The presence of two families of asphaltene aggregates for Buzurgan solutions, one diffusing quickly and one diffusing more slowly, suggests two asphaltene behaviors, depending on their molecular affinity. For high concentration, the self-affinity of heavy species increases, which causes a further aggregation process. As a consequence, the local concentration of toluene in the mixture varies. It decreases in the neighborhood of heavy species favoring aggregation and causing a decrease in the asphaltene diffusion coefficient, while the concentration of solvent increases nearby nanoaggregates leading to an increase in the asphaltene diffusion coefficient. In fact, in concentrated solutions, more attractive interactions are favored with large species, resulting in an increase in the friction of the solute,84 leading to an aggregation process. As a result, if some regions are becoming richer in macroaggregates of asphaltenes, other regions are less concentrated in macroaggregates, where smaller entities or nanoaggregates remain free and can diffuse faster. Starting from a liquid mixture with a rather homogeneous spatial distribution of asphaltene aggregates and toluene, the increase in asphaltene concentration is responsible for the distinction observed between smaller and larger aggregates.89 In a microscopic point of view, there is a difference in nano or macroaggregate concentration, whereas in a macroscopic point of view, an average density is observed. Solvent molecules are thought to be ejected out of large asphaltene aggregates. Thus, the concentration of toluene around the small aggregates becomes locally higher, which induces an increase in their mobility as illustrated in Figure 7. 1 H DOSY experiments allowed the identification of an aggregation point as compared to the aggregation of clays.90 Water diffusivity D was measured by using T2 transverse NMR relaxation times, which are sensitive to the volume-to-surface ratio in Na smectite suspensions. The authors observed a decrease in the diffusivity with an increasing smectite fraction

J. Phys. Chem. C, Vol. 113, No. 36, 2009 16273

Figure 7. Schematic representation of the transition of asphaltene aggregates when their concentration increases.

(up to 50 wt %) and an increase in diffusivity with salt addition (NaCl or CaCl2 aqueous solutions) at a fixed clay fraction. The increase was explained by aggregation of clay particles when high salinities are reached. They also detected bimodal distributions on CaCl2 mixtures. Cousin et al.89 have shown that solid samples were microscopically phase separated, while analyzing the behavior of a mixture of discotic nanoparticles of laponite and spherical magnetic nanoparticles of maghemite. The samples were made of dense connected domains of laponite nanoparticles surrounding liquid pockets of maghemite nanoparticles. Moreover, other techniques have also suggested the possibility to have some heterogeneity in asphaltene densities. In fact, SAXS data of Safaniya asphaltenes in toluene (6 wt %)51 showed an upturn of the X-ray scattered intensities at very low Q-values (Figures 16 and 38 in ref 51). According to the authors, this observation can be explained by dense heterogeneity in the medium. Similar observations were presented by Sirota and Lin91 mentioning separation of phases inside the solution. More recently, Headen et al.92 also highlighted by SANS the coexistence of large aggregates with smaller ones. However, the calculated size of the aggregates is very high (500 nm). On the basis of all of this, we believe that the DOSY NMR experiments allow detection of the beginning of the aggregation process. Asphaltene systems are also known to be very polydisperse. This has been demonstrated by fractionating the asphaltenes by ultracentrifugation50 and membrane filtration.93,94 All fractions obtained remained stable, and no recombination was observed. At low concentrations, it is a polydisperse diluted system, but an increase in the solute concentration induces a phase transition, which is somehow similar to polydisperse polymer systems near the cloud point.95 Two phases appear. The first one is richer in high molecular weight polymers, and the second one is consequently impoverished. This concept has been applied to asphaltenes in a recent paper.96 An illustration of this transition induced by concentration is nicely illustrated by Kuznicki et al.97 The asphaltene aggregation process is clearly shown for archipelago and continental models favored by π-π interactions of the aromatic rings. This is illustrated by Figure 7, but we are fully convinced that if the concentration was further increased, more macroaggregates would be generated, although they would not be necessarily observed by NMR due to the fast NMR relaxation process. As previously discussed, it would be difficult to observe very large aggregates (500 nm) as shown by Headen et al.92 Finally, all of these representations are not so far from that proposed in 1940 by Pfeiffer et Saal,98 suggesting asphaltene aggregation in pure crude oil to gel formation. Our experiments, performed over a wide range of concentrations, are very useful to study in which domain average molecular weight can be determined. It is possible to extract physical properties in the dilute regime (C < 0.25 wt %).

16274

J. Phys. Chem. C, Vol. 113, No. 36, 2009

TABLE 3: Average Masses Mn, Mw, and Mpa of Buzurgan Asphaltenes SEC -1

Mn (g mol ) Mw (g mol-1) Mp (g mol-1) a

1630 ( 270 5050 ( 920 8160 ( 1080

DOSY NMR 5890 ( 1420 11170 ( 2690 6900 ( 1660

Masses are at the highest distribution peak.

4.4. Molecular Weight and Radius Determination from Diffusion Measurements in the Dilute Regime. In the dilute regime (C < 0.25 wt %), the hydrodynamic radius and average molecular weight of the nanoaggregates may be inferred from the mean diffusion coefficient at infinite dilution. We found that the mean diffusion coefficient of nanoaggregates of Buzurgan asphaltenes in toluene is about (2.4 ( 0.1) × 10-10 m2 s-1 at 20 °C and infinite dilution. This value is in agreement with previous works.48,49,54 According to the modified Stokes-Einstein equation (eq 3 with fs ) 1, average spherical particle assumed), this average diffusion coefficient corresponds to a mean hydrodynamic radius of about 15.6 ( 1.5 Å. This result is in reasonable agreement with time-resolved fluorescence depolarization (TRFD),99 fluorescence correlation spectroscopy (FCS),48 and NMR results.47 On the basis of the established PS calibration curve (Figure 1), an average molecular weight can be estimated from the ratio of the diffusion coefficients of the solute and solvent at infinite dilution [D∞tol ∼ (19.44 ( 0.05) × 10-10 m2 s-1 for toluene, D∞asp ∼ (2.4 ( 0.1) × 10-10 m2 s-1 for Buzurgan nanoaggregates of asphaltenes]. This leads to a mean molecular weight of approximately 6900 g mol-1 (as polystyrene equivalent) at the highest distribution peak (Mp) found for the Buzurgan asphaltenes analyzed at 0.01 wt % in toluene-d8 at 20 °C. The observed diffusion spectrum also reflects the molecular weight distribution, which is due to sample heterogeneity in terms of the nature of the molecules present in the sample. From the observed range for the asphaltene diffusion coefficient at infinite dilution, which varies between (0.8 ( 0.1) × 10-10 m2 s-1 and (4.5 ( 0.1) × 10-10 m2 s-1, an average molecular weight range of roughly 1500-85000 g mol-1 is found for the Buzurgan asphaltene nanoaggregates at 0.01 wt % in toluene-d8 at 20 °C. Average masses (Mn, Mw, and Mp) calculated by DOSY NMR and obtained by SEC are summarized in Table 3. As outlined above, both results are given as polystyrene equivalent and can therefore be compared to each other keeping in mind that the solvents used are different. Assuming an average molecular weight of monomers ranging from 500 to 2000 depending on the experimental conditions, on the sample and on the analytical technique used, we find the average molecular weights reported in Table 3 represent the molecular weight of nanaoagregates composed of 3-15 asphaltenic molecules. The results obtained by DOSY are higher than those obtained by SEC. This is not surprising as asphaltenes were dissolved in toluene for DOSY measurements, whereas THF was used for SEC analysis. Indeed, it is proven that average molecular weight estimated by some analytical techniques strongly depends on the nature of the solvent. Espinat et al.51 emphasized that the average molecular weight was much larger in toluene than in THF. Hence, the difference in solvent may explain the difference of estimated average molecular weights obtained from SEC and DOSY data. Moreover, adsorption of compounds onto the column may occur during SEC analysis, leading to underestimated apparent molecular masses. SEC is a technique based on hydrodynamic volumes of molecules and not on mass. Last but not least, the distribution profile obtained by SEC is bimodal,31 while there is only one diffusion peak for asphaltenes analyzed by DOSY

Durand et al. TABLE 4: Influence of Shape Factor r on Calculated Molecular Weight R 0.33 0.42 0.5 0.55

shape associated

rigid impermeable sphere polystyrene equivalent random coil polymer large ellipsoid extending the major axis and maintaining the minor axis constant 0.6 large ellipsoid extending the major axis and maintaining the minor axis constant 0.65 disc-like 0.7 disc-like 1 rod-shaped solute, which increases in length while maintaining a constant cross-sectional area

relative molecular mass (Mp) associated (g mol-1) 21860 6900 3500 2540 1940 1540 1270 590

NMR at 0.01 wt %. The lowest concentration accessible by NMR is probably higher than the asphaltenes concentration achieved during SEC measurements. The bimodal distribution is probably due to the distribution of asphaltenes between macroaggregates (high molecular weight) and nanoaggregates (low molecular weight). An increase in the concentration would lead to a widening peak shifting toward the high molecular weights.67 This is what is observed in DOSY NMR: a unique distribution peak. We have no real explanation of this discrepancy. Both techniques are sensitive to hydrodynamic volume, but it was suggested that a steric exclusion mechanism may foster dissociation effects of asphaltene aggregates when they break into pores of the polystyrenes gels where concentration significantly decreases. Additional complementary works will be worth persuing to gain insight. Finally, even though SAXS and SANS results give molecular weights that are roughly a factor of 10 higher than our values,50,53,100 the mean molecular weight estimated in this work is higher than those recently published by several authors.18,32 Two hypotheses can explain the differences. First, as already stated, at the concentrations investigated in our study, we are above the CNAC (which is supposed to be at approximately 0.010-0.020 wt %).14,88,101 It has been shown by absorption and fluorescence spectroscopy102 that dimerization may occur at concentrations of 0.005 wt %, depending upon the origin of the sample. As a result, in our solutions, even at the lowest concentrations, aggregates are mixed in with single molecules. So, the molecular weight determined in this paper by 1H-DOSY (polystyrene equivalent) does not represent the molecular weight of the asphaltenes monomer but that of more or less large-sized aggregates of asphaltenes. The second hypothesis concerns the shape factor of polystyrenes, which may not be well-adapted for asphaltenes. In the literature, Baltus103 reported various values of shape factors, depending upon the origin of the sample and the solvent used. The author summarizes many studies, presenting values ranging from 0.5 to 1.0, and linked them to the probable shape of the molecule. We want to highlight the importance of the values of the shape factor for determination of the molecular weight from diffusion coefficients. Table 4 shows the influence of R on the calculated average molecular weight. The calculated molecular weights are only relative values and should only be used to show the impact of R. The higher R is the lower the molecular weight. Thus taking a higher value for R would lead to a lower molecular weight than the one presented above (polystyrene equivalent). It is difficult to extract absolute molecular weights from diffusion coefficients as long as a power

Aggregation States of Asphaltenes law has not been clearly established for these types of asphaltenes (Buzurgan) in toluene at 20 °C. For this reason, our results are relative values and are given as equivalent polystyrenes. 5. Conclusions The goal of this study was to investigate Buzurgan asphaltene behavior by NMR techniques. Diffusivity measurements were used as a means to evaluate the size and average molecular weights of nanoaggregates of asphaltenes. Such information is crucial in the development of new catalysts and novel upgrading processes. The solute size determined from 1H DOSY NMR was found to be 15.6 Å. Moreover, from these experiments, an average molecular weight of 6900 g mol-1 (given as PS equivalent and taken at the highest distribution diffusion peak, Mp) was obtained, with a range of roughly 1500-85000 g mol-1. Even at the lowest concentration analyzed, the asphaltenes are very polydisperse. We highlighted the importance of the value of the shape factor for the calculation of the absolute molecular weights, which could then be compared with results obtained with other analytical techniques such as MS. Here, relative molecular weights against PS (Mn and Mw) were given and compared with SEC data. DOSY appears as a promising global approach for the physicochemical characterization of complex samples. We also explored the molecular dynamics of asphaltenes by studying the influence of the asphaltene concentration on their mobility in solution. In the dilute regime, the diffusion coefficients remain constant as only solute-solvent interactions limit the translational self-diffusion, whereas more intermolecular interactions start to be involved above 0.25 wt %. From this observation, we defined 0.25 wt % as the onset of the aggregation process. In the dilute regime, asphaltenes form a polydisperse diluted system, whereas an increase in the concentration of these macromolecules induces a phase transition. The first phase is rich in macroaggregates of asphaltenes, resulting in an impoverishment in large aggregates for the second phase. For the first time, a clear separation in the diffusion dimension between two classes of asphaltene aggregates was observed from concentrations of 3 wt % onward, reflecting the beginning of the flocculation. We showed the 1H DOSY spectra and diffusion profile to confirm these unexpected results. The originality of this study resides in the detection and presentation of two classes of asphaltene aggregates, without making any working hypothesis concerning the composition of the mixture. Further work on asphaltenes from different sources is still in progress and will be presented in a subsequent paper. Acknowledgment. The authors acknowledge M.A. Delsuc from the University of Strasbourg and H. Vandamme from ESPCI for their advice regarding this work and C. Ferreira from IFP for providing the SEC data. References and Notes (1) Speight, J. G. Asphaltene Constituents. In The Chemistry and Technology of Petroleum; CRC Press: Boca Raton, FL, 2007; pp 315344. (2) Gawrys, K. L.; Blankenship, G. A.; Kilpatrick, P. K. Energy Fuels 2006, 20, 705–714. (3) Spiecker, P. M.; Gawrys, K. L.; Kilpatrick, P. K. J. Colloid Interface Sci. 2003, 267, 178–193. (4) Mitra-Kirtley, S.; Mullins, O. C.; Van Elp, J.; George, S. J.; Chen, J.; Cramer, S. P. J. Am. Chem. Soc. 1993, 115, 252–258. (5) Schneider, M. H.; Andrews, A. B.; Mitra-Kirtley, S.; Mullins, O. C. Energy Fuels 2007, 21, 2875–2882. (6) Merdrignac, I.; Espinat, D. Oil Gas Sci. Technol. 2007, 62, 7–32. (7) Ancheyta, J.; Centeno, G.; Trejo, F.; Marroquin, G. Energy Fuels 2003, 17, 1233–1238. (8) Havre, T. E.; Sjo¨blom, J. Colloids Surf., A 2003, 228, 131–142.

J. Phys. Chem. C, Vol. 113, No. 36, 2009 16275 (9) Zhao, B.; Zhang, X.; Shaw, J. M. Energy Fuels 2008, 22, 1747– 1758. (10) Yudin, I. K.; Anisimov, M. A. Dynamic Light Scattering Monitoring of Asphaltenes Aggregation in Crude Oils and Hydrocarbons Solutions. In Asphaltenes, HeaVy Oils, and Petroleomics; Springer: New York, 2007; pp 439-468. (11) Espinat, D.; Fenistein, D.; Barre, L.; Frot, D.; Briolant, Y. Energy Fuels 2004, 18, 1243–1249. (12) Buenrostro-Gonzalez, E.; Lira-Galeana, C.; Gil-Villegas, A.; Wu, J. Z. AIChE J. 2004, 50, 2552–2570. (13) Shaw, J. M.; Zou, X. Phase BehaVior of HeaVy Oils. In Asphaltenes, HeaVy Oils, and Petroleomics; Springer: New York, 2007; pp 489-510. (14) Andreatta, G.; Bostrom, N.; Mullins, O. C. Langmuir 2005, 21, 2728–2736. (15) Tanaka, R.; Sato, S.; Takanohashi, T.; Hunt, J. E.; Winans, R. E. Energy Fuels 2004, 18, 1405–1413. (16) Yen, T. F.; Erdman, J. G.; Pollack, S. S. Anal. Chem. 1961, 33, 1587–1594. (17) Groenzin, H.; Mullins, O. C. Energy Fuels 2000, 14, 677–684. (18) Klein, G. C.; Kim, S.; Rodgers, R. P.; Marshall, A. G.; Yen, A. Energy Fuels 2006, 20, 1973–1979. (19) Gawrys, K. L.; Blankenship, G. A.; Kilpatrick, P. K. Langmuir 2006, 22, 4487–4497. (20) Herod, A. A.; Bartle, K. D.; Kandiyoti, R. Energy Fuels 2007, 21, 2176–2203. (21) Qian, K.; Edwards, K. E.; Siskin, M.; Olmstead, W. N.; Mennito, A. S.; Dechert, G. J.; Hoosain, N. E. Energy Fuels 2007, 21, 1042–1047. (22) Gawrys, K. L.; Kilpatrick, P. K. Instrum. Sci. Technol. 2004, 32, 247–253. (23) Meyer, V.; Pilliez, J.; Habas, J. P.; Montel, F.; Creux, P. Energy Fuels 2008, 22, 3154–3159. (24) Trejo, F.; Ancheyta, J.; Morgan, T. J.; Herod, A. A.; Kandiyoti, R. Energy Fuels 2007, 21, 2121–2128. (25) Verruto, V. J.; Kilpatrick, P. K. Energy Fuels 2007, 21, 1217– 1225. (26) Norinaga, K.; Wargardalam, V. J.; Takasugi, S.; Iino, M.; Matsukawa, S. Energy Fuels 2001, 15, 1317–1318. (27) Dickinson, E. M. Fuel 1980, 59, 290–294. (28) Morgan, T. J.; George, A.; Davis, D. B.; Herod, A. A.; Kandiyoti, R. Energy Fuels 2008, 22, 1824–1835. (29) Desando, M. A.; Lahajnar, G.; Ripmeester, J. A.; Ivan, Z. Fuel 1999, 78, 31–45. (30) Nielsen, K. E.; Dittmer, J.; Malmendal, A.; Nielsen, N. C. Energy Fuels 2008, 22, 4070–4076. (31) Merdrignac, I.; Quoineaud, A. A.; Gauthier, T. Energy Fuels 2006, 20, 2028–2036. (32) Ruiz-Morales, Y.; Mullins, O. C. Energy Fuels 2007, 21, 256– 265. (33) Murgich, J.; Abanero, J. A.; Strausz, O. P. Energy Fuels 1999, 13, 278–286. (34) Strausz, O. P.; Mojelsky, T. W.; Lown, E. M. Fuel 1992, 71, 1355– 1363. (35) Gray, M. R. Energy Fuels 2003, 17, 1566–1569. (36) Morris, K. F.; Johnson, C. S. J. Am. Chem. Soc. 1992, 114, 3139– 3141. (37) Barjat, H.; Morris, G. A.; Smart, S.; Swanson, A. G.; Williams, S. C. R. J. Magn. Reson., Ser. B 1995, 108, 170–172. (38) Hinton, D. P.; Johnson, C. S. J. Phys. Chem. 1993, 97, 9064– 9072. (39) Morris, K. F.; Johnson, C. S. J. Am. Chem. Soc. 1993, 115, 4291– 4299. (40) Chen, A.; Wu, D.; Johnson, C. S. J. Am. Chem. Soc. 1995, 117, 7965–7970. (41) Jerschow, A.; Muller, N. Macromolecules 1998, 31, 6573–6578. (42) Lucas, L. H.; Larive, C. K. Concepts Magn. Reson., Part A 2004, 20A, 24–41. (43) Ambrus, A.; Friedrich, K.; Somogyi, A. Anal. Biochem. 2006, 352, 286–295. (44) Viel, S.; Mannina, L.; Segre, A. Tetrahedron Lett. 2002, 43, 2515– 2519. (45) Kapur, G. S.; Findeisen, M.; Berger, S. Fuel 2000, 79, 1347– 1351. (46) Durand, E.; Clemancey, M.; Quoineaud, A. A.; Verstraete, J.; Espinat, D.; Lancelin, J. M. Energy Fuels 2008, 22, 2604–2610. (47) Lisitza, N. V.; Freed, D. E.; Sen, P. N.; Song, Y. Q. Energy Fuels 2009, 23, 1189–1193. (48) Andrews, A. B.; Guerra, R. E.; Mullins, O. C.; Sen, P. N. J. Phys. Chem. A 2006, 110, 8093–8097. (49) Ostlund, J. A.; Andersen, S. I.; Nyden, M. Fuel 2001, 80, 1529– 1533. (50) Barre, L.; Simon, S.; Palermo, T. Langmuir 2008, 24, 3709–3717.

16276

J. Phys. Chem. C, Vol. 113, No. 36, 2009

(51) Espinat, D.; Rosenberg, E.; Scarsella, M.; Barre, L.; Fenistein, D.; Broseta, D. Colloidal Structural Evolution from Stable to Flocculated State of Asphaltene Solutions and Heavy Crudes. In Structures and Dynamics of Asphaltenes; Plenum Press: New York, 1998; pp 145-202. (52) Tanaka, R.; Hunt, J. E.; Winans, R. E.; Thiyagarajan, P.; Sato, S.; Takanohashi, T. Energy Fuels 2003, 17, 127–134. (53) Gawrys, K. L.; Kilpatrick, P. K. J. Colloid Interface Sci. 2005, 288, 325–334. (54) Kawashima, H.; Takanohashi, T.; Iino, M.; Matsukawa, S. Energy Fuels 2008, 22, 3989–3993. (55) Pelta, M. D.; Morris, G. A.; Stchedroff, M. J.; Hammond, S. J. Magn. Reson. Chem. 2002, 40, S147–S152. (56) Antalek, B. Concepts Magn. Reson. 2002, 14, 225–258. (57) Price, W. S. Concepts Magn. Reson. 1998, 10, 197–237. (58) Nilsson, M.; Morris, G. A. Magn. Reson. Chem. 2007, 45, 656– 660. (59) Zhu, Y. K.; Gregory, R. B. Nucl. Instrum. Methods Phys. Res., Sect. A 1989, 284, 443–451. (60) Johnson, C. S. Prog. Nucl. Magn. Reson. Spectrosc. 1999, 34, 203–256. (61) Provencher, S. W. Comput. Phys. Commun. 1982, 27, 213–227. (62) Provencher, S. W. Comput. Phys. Commun. 1982, 27, 229–242. (63) Delsuc, M. A.; Malliavin, T. E. Anal. Chem. 1998, 70, 2146– 2148. (64) Tramesel, D.; Catherinot, V.; Delsuc, M. A. J. Magn. Reson. 2007, 188, 56–67. (65) Nilsson, M.; Morris, G. A. Anal. Chem. 2008, 80, 3777–3782. (66) Delsuc, M. A.; Tramesel, D. Comptes Rendus Chimie 2006, 9, 364–373. (67) Merdrignac, I.; Truchy, C.; Robert, E.; Guibard, I.; Kressmann, S. P. Pet. Sci.Technol. 2004, 22, 1003–1022. (68) Stejskal, E. O.; Tanner, J. E. J. Chem. Phys. 1965, 42, 288–292. (69) Hahn, E. L. Phys. ReV. 1950, 80, 580. (70) Carr, H. Y.; Purcell, E. M. Phys. ReV. 1954, 94, 630. (71) Macchioni, A.; Ciancaleoni, G.; Zuccacia, C.; Zuccacia, D. Chem. Soc. ReV. 2008, 37, 479–489. (72) Perrin, F. J. Phys. Radium. 1936, 7, 1–11. (73) Chen, H. C.; Chen, S. H. J. Phys. Chem. 1984, 88, 5118–5121. (74) Hakansson, B.; Nyden, M.; Soderman, O. Colloid Polym. Sci. 2000, 278, 399–405. (75) Auge´, S.; Schmit, P. O.; Crutchfield, C. A.; Islam, M. T.; Harris, D. J.; Durand, E.; Clemancey, M.; Quoineaud, A. A.; Lancelin, J. M.; Prigent, Y.; Taulelle, F.; Delsuc, M. A. J. Phys. Chem. B 2009, 113, 1914– 1918. (76) Cosgrove, T.; Griffiths, P. C. Polymer 1995, 36, 3335–3342. (77) Crutchfield, C. A.; Harris, D. J. J. Magn. Reson. 2007, 185, 179– 182. (78) Jones, J. A.; Wilkins, D. K.; Smith, L. J.; Dobson, C. M. J. Biomol. NMR 1997, 10, 199–203.

Durand et al. (79) Pyun, C. W.; Fixman, M. J. Chem. Phys. 1964, 41, 937–944. (80) Nortz, R. L.; Baltus, R. E.; Rahimi, P. Ind. Eng. Chem. Res. 1990, 29, 1968–1976. (81) Patt, S. L.; Shoolery, J. N. J. Magn. Reson. 1982, 46, 535–539. (82) Jakobsen, H. J.; Sorensen, O. W.; Brey, W. S.; Kanyha, P. J. Magn. Reson. 1982, 48, 328–335. (83) Ostlund, J. A.; Nyden, M.; Auflem, I. H.; Sjoblom, J. Energy Fuels 2003, 17, 113–119. (84) Biswas, R.; Bhattacharyya, S.; Bagchi, B. J. Phys. Chem. B 1998, 102, 3252–3256. (85) Freed, D. E.; Lisitza, N. V.; Sen, P. N.; Song, Y. Q. Molecular Composition and Dynamics of Oils. In Asphaltenes, HeaVy Oils, and Petroleomics; Springer: New York, 2007; pp 279-299. (86) Freed, D. E.; Lisitza, N. V.; Sen, P. N.; Song, Y. Q. Magn. Reson. Imaging 2007, 25, 544. (87) Sheu, E.; Long, Y.; Hamza, H. Asphaltene Self-Association and Precipitation in Solvents-AC conductivity Measurements. In Asphaltenes, HeaVy Oils, and Petroleomics; Springer: New York, 2007; pp 259-278. (88) Zeng, H.; Song, Y. Q.; Johnson, D. L.; Mullins, O. C. Energy Fuels 2009, ASAP. (89) Cousin, F.; Cabuil, V.; Grillo, I.; Levitz, P. Langmuir 2008, 24, 11422–11430. (90) Guichet, X.; Fleury, M.; Kohler, E. J. Colloid Interface Sci. 2008, 327, 84–93. (91) Sirota, E. B.; Lin, M. Y. Energy Fuels 2007, 21, 2809–2815. (92) Headen, T. F.; Boek, E. S.; Stellbrink, J.; Scheven, U. M. Langmuir 2009, 25, 422–428. (93) Marques, J.; Merdrignac, I.; Baudot, A.; Barre, L.; Guillaume, D.; Espinat, D.; Brunet, S. Oil Gas Sci. Technol. 2008, 63, 139–149. (94) Zhao, B.; Shaw, J. M. Energy Fuels 2007, 21, 2795–2804. (95) Behme, S.; Sadowski, G.; Arlt, W. Fluid Phase Equilib. 1999, 158, 869–877. (96) Manshad, A. K.; Edalat, M. Energy Fuels 2008, 22, 2678–2686. (97) Kuznicki, T.; Masliyah, J. H.; Bhattacharjee, S. Energy Fuels 2008, 22, 2379–2389. (98) Pfeiffer, J. P.; Saal, R. N. J. J. Phys. Chem. 1940, 44, 139–149. (99) Groenzin, H.; Mullins, O. C. J. Phys. Chem. A 1999, 103, 11237– 11245. (100) Fenistein, D.; Barre, L.; Broseta, D.; Espinat, D.; Livet, A.; Roux, J. N.; Scarsella, M. Langmuir 1998, 14, 1013–1020. (101) Andersen, S. I.; del Rio, J. M.; Khvostitchenko, D.; Shakir, S.; Lira-Galeana, C. Langmuir 2001, 17, 307–313. (102) Goncalves, S.; Castillo, J.; Fernandez, A.; Hung, J. Fuel 2004, 83, 1823–1828. (103) Baltus, R. E. Characterization of Asphaltenes and Heavy Oils using Hydrodynamic Property Measurements. In Structures and Dynamics of Asphaltenes; Plenum Press: New York, 1998; pp 303-335.

JP901954B