PREFACE

This thesis is submitted for fulfillment of the PhD degree at the Norwegian University of Science and Technology. The work has been performed at the Ugelstad Laboratory at the Department of Chemical Engineering. I received my master degree in June 2003 at the Department of Chemical Engineering, NTNU in the field of sterically demanding bis (2-silylindenyl) zirconium (IV) dichlorides as polymerization catalyst precursors. The design of the catalyst ligand framework symmetry was successfully employed for the tailoring of tacticity and comonomer incorporation of homo- and copolymers. This study has been done within the scope of the Joint Industrial Program (JIP1) on Particle-Stabilized Emulsions/Heavy Crude Oils, financed by industry. The thesis consists of five papers or manuscripts, motivated with an introduction to the chemistry, production, and characterization of petroleum.

ACKNOWLEDGEMENTS
I would like to express my gratitude to Professor Johan Sjblom for excellent supervision and for all his effort in establishing a modern and well-equipped laboratory within the field of colloid and surface chemistry. The participating members of the JIP consortium from ABB, Aker Kvrner, British Petroleum, Champion Technologies, Chevron, ENI, Hydro, Mrsk, Petrobras, Shell, Statoil, Total, and Vetco have contributed considerably through their sponsorship and useful discussions. Especially, I want to thank Harald, Arild and Einar at Statoil R&D Centre and Robert at Hydro Research Centre for also generously displaying their laboratories and professional skills. Students and researchers at the Ugelstad Laboratory, other staff at the Department of Chemical Engineering and people in the Technical Division are gratefully acknowledged for creating a professional and social working environment in chemistry building 5. I will thank Jan Ole Sundli for his help in implementing the oscillating pendant drop instrument. His eyes lightened up when he got a chance to build a low budget amplifier from a radio tube. This time not for his home stereo system, but in the spirit of science even more fun! Stine, thanks for all your love and support!

LIST OF PAPERS Paper 1

A. Hannisdal, P.V. Hemmingsen, J. Sjblom, Group-Type Analysis of Heavy Crude Oils Using Vibrational Spectroscopy in Combination with Multivariate Analysis, Ind. Eng. Chem. Res., 2005, 44, 1349-1357.

Paper 2
A. Hannisdal, M-H. Ese, P.V. Hemmingsen, J. Sjblom, Particle-Stabilized Emulsions: Effect of Heavy Crude Oil Components Pre-Adsorbed onto Stabilizing Solids, Colloids and Surfaces A: Physicochemical and Engineering Aspects, 2006, 276, 45-58.

Paper 3
A. Hannisdal, R. Orr, J. Sjblom, Viscoelastic Properties of Crude Oil Components at Oil-Water Interfaces. 1: The Effect of Dilution. J. Dispersion Sci. Technol., 2007, 28 (1).

Paper 4
A. Hannisdal, R. Orr, J. Sjblom, Viscoelastic Properties of Crude Oil Components at Oil-Water Interfaces. 2: Comparing 30 crude oils. J. Dispersion Sci. Technol., 2007, 28 (3).

Paper 5
A. Hannisdal, P.V. Hemmingsen, A. Silset, J. Sjblom, Stability of Water/Crude Oil Systems Correlated to the Physicochemical Properties of the Oil Phase.

Additional Publications
P.V. Hemmingsen, A. Silset, A. Hannisdal, J. Sjblom, Emulsions of Heavy Crude Oils. 1: Influence of Viscosity, Temperature, and Dilution, J. Dispersion Sci. Technol., 2005, 26, 615-627. M. Fossen, P.V. Hemmingsen, A. Hannisdal, J. Sjblom, H. Kallevik, Solubility Parameters Based on IR and NIR Spectra: I. Correlation to Polar Solutes and Binary Systems, J. Dispersion Sci. Technol., 2005, 26, 227-241. M-H. Ese, C.M. Selsbak, A. Hannisdal, J. Sjblom, Emulsions Stabilized by Indigenous Reservoir Particles. Influence of Chemical Additives, J. Dispersion Sci. Technol., 2005, 26, 145-154. J. Sjblom, G. ye, W.R. Glomm, A. Hannisdal, M. Knag, . Brandal, M-H. Ese, P.V. Hemmingsen, T.E. Havre, H-J. Oschmann, H. Kallevik, Modern Characterization Techniques for Crude Oils, Their Emulsions and Functionalized Surfaces, Emulsions and Emulsion Stability, 2nd Edition, J. Sjblom, Editor, Taylor and Francis: New York, 2005. A. Hannisdal, A.C. Mller, E. Rytter, Sterically demanding bis(2-silylindenyl) zirconium(IV) dichlorides as polymerisation catalyst precursors. Macromolecular Symposia, 2004, 213, 79-88. A. Hannisdal, E. Rytter, J.A. Stvneng, A.C. Mller, R. Blom, O. Swang, K. Angermund, M. Graf, W. Thiel, T. Seraidaris, On the Non-Single-Site Character of Bis (2-DMS-indenyl) zirconium(IV) Dichloride/MAO and Bis (2-TMS-indenyl) zirconium (IV) Dichloride/MAO: Polymerisation Characteristics and Mechanistic Implications, Macromolecules, Submitted.

Conferences and Seminars

J. Sjblom, A. Hannisdal, M-H. Ese, Fluid Characterization and Multiphase Flow. (oral presentation by J.S) Norway North America Petroleum Research Workshop, 2005. J. Sjblom, A. Hannisdal, M-H. Ese, The Influence of Interfaces on Multiphase Flow. (oral presentation by J.S) SPE 2005 Series. Molecules on move. Physicochemical Effects on Multiphase Transport, 2005. A. Hannisdal, M-H. Ese, P. V. Hemmingsen, J. Sjblom, Particle-Stabilized Emulsions: Effect of Heavy Crude Oil Components Pre-Adsorbed onto Stabilizing Solids. (Poster) The 6th International Conference on Phase Behavior and Fouling, Amsterdam, 2005. A. Hannisdal, J. Sjblom, Emulsions of Water in Crude Oil. (oral presentation by A.H.) Statoils Seminar on Raw Materials, 2004. J. Sjblom, P. V. Hemmingsen, A. Hannisdal, A. Silset, Stability Mechanisms of Crude Oil Emulsions A Review. (oral presentation by P.V.H) The 7th International Conference on Phase Behavior and Fouling, North Carolina, USA, 2006.

INTRODUCTION
In the past, operators on the Norwegian Continental Shelf have been blessed with conventional or light (low density, easy flowing) crude oils. However to date, crude oil exploitation has advanced to the point that the era of large fields with a high quantity and quality is over. Since the possibility of finding highly productive formations is small the corresponding strategy is focused on an improved exploitation of the existing large fields and to tie in, in an efficient way, small fields. The most important thing when focusing on existing large fields is to increase the recovery rate. This can however be connected with typical flow assurance problems. At older fields the co-production of water is in most cases substantial. It is well known that many wells co-produce water to an extent of 50 to 70%. This production profile will cause problems with high water volumes in the separator and emulsions close to the inversion point. Emulsion stability mechanisms must be understood both from macroscopic and a microscopic points of view for successful operation. As oil producing areas become mature, the heavy oil reservoirs are attracting increasing attention. To date they have often been left undeveloped since the crude oils are difficult to process. The Norwegian oil company, Statoil, has a heavy oil portfolio of 5 fields: Zuata (8.5 oAPI, Venezuela), Alba (19 oAPI, UK), Linerle (16 oAPI, Norway), Falk (18 o API, Norway), and Bressay (10.8 oAPI, UK). Zuata and Alba are currently operating, while Linerle, Falk and Bressay are in the early phase of development. This change towards heavier crude oils has and will continue to make new demands with respect to both improved process technology and understanding of the crude oil systems. To tie in small fields with larger production streams means that the fluids that are to be mixed must be compatible. If one encounters precipitation of organic matter (for instance asphaltenes) the whole production line and the regularity will be endangered. In parallel with these efforts the trend to build up a sub-sea processing network will emerge. Also in this case the nature of the fluids will be decisive for the design and dimensions of the process unit. Hence fluid characterization will be a key technology for success in the future offshore processing. As our industrial partners presented a unified interest to study crude oils and their problematic emulsions, Professor Sjblom initiated a joint industrial program on the topic Particle-Stabilized Emulsions/Heavy Crude Oils. This thesis presents studies within the scope of the program.

The Ugelstad Laboratory was founded at the Norwegian University of Science and Technology (NTNU) in January 2002 to commemorate the late Professor John Ugelstad. The laboratory specializes in surfactant chemistry and its technical applications, emulsions and emulsion technology, preparation of polymers and polymer particles and their technical applications, plasma chemical modification of surfaces and silica-based chemistry. Applications include crude oil production and processing, pulp and paper, biomedicine, catalysis and materials science. Thanks to the interest and financial support from industry and the Research Council in Norway (NFR), the Ugelstad Laboratory has invested in modern instrumentation for petroleum and material science. The main purpose is to raise the national level of colloidal science.

CONTENTS
1. CRUDE OIL THE BLACK GOLD... 1 1.1 Chemical (Elemental) Composition 1 1.2 The Crude Oil System: SARA, and Petroleomics.. 1 1.3 Molecular Structure and Aggregation of Petroleum Asphaltenes.. 2 1.4 Light, Heavy and Extra Heavy Crude Oil... 6 References. 7 2. PETROLEUM PROCESSING AND FLOW ASSURANCE.. 8 2.1 Recovery. 8 2.2 Quality of the Well Product ... 10 2.3 Transportation. 13 2.4 Refining.. 15 2.5 Problems in Operation Steps.. 17 References. 19 3. EMULSION STABILITY AND STABILITY MECHANISMS. 20 3.1 Emulsion Type and Phase-Inversion.. 21 3.2 General Criteria for Stabilization Surface Forces 23 3.3 Flocculation in Emulsions... 24 3.4 Mechanism of Coalescence. 25 3.5 Interfacial Elasticity, Viscosity, and Diffusivity. 26 3.6 Steric Stabilization of the Interface. 27 3.7 Raising an Issue on the Stabilization of Crude Oil-Water Systems 28 References. 29 4. GENERAL ASPECTS OF THE CHARACTERIZATION OF CRUDE OILS... 31 4.1 From Live to Dead Oil.... 31 4.2 From Conventional Oils to Heavy Oils...31 4.3 Aspects of Emulsification... 32 4.4 Choosing Parameters and Correct Experimental Conditions.. 33 4.5 Statistical Analysis Feature or Artifact?.................................................. 34 References. 35 5. CHARACTERIZATION TECHNIQUES 36 5.1 Fractionation of Maltenes by HPLC... 36 5.2 Infrared Spectroscopy. 37 5.3 Near-Infrared Spectroscopy 39 5.4 Interfacial Tension.. 41 5.5 Dilational Relaxation.. 42 5.6 Electrically Induced Droplet Flocculation and Coalescence.. 43 5.7 QCM-D... 45 5.9 Contact Angle Measurements. 48 References. 50 6. MAIN RESULTS. 51

Paper 1.. 51 Paper 2.. 52 Paper 3.. 54 Paper 4.. 56 Paper 5.. 58 CONCLUDING REMARKS 60 PAPERS

1. CRUDE OIL THE BLACK GOLD

Step right up. In this bottle is a remedy of wonderful efficacy. Its curative powers are calculated to remove pain and alleviate human suffering and disease. Distilled 400 feet below the earth's surface, this remarkable liquid is Mother Nature's bounteous gift of healing. It is my pleasure - no, it is my duty - to bring this soothing restorative, this blessed ointment, this modern-day balm of Gilead to the public. Crude oil may not be the panacea that snake oil claimed to be. But for 20th century industrialized nations, it has proved to be more than good medicine [1].

1.1 Chemical (Elemental) Composition

Geologists generally agree that crude oil was formed over millions of years from the remains of tiny aquatic plants and animals that lived in ancient seas. There may be bits of brontosaurus thrown in for good measure, but the petroleum owes its existence largely to one-celled marine organisms. It is the particular crude oils geological history that is most important in determining its characteristics, which is why crude oils formed in similar marine deposits at different continents can resemble each other [1]. Regions characterized by different marine deposits, pressures and temperatures may however produce crude oil with a great variety in appearance, from honey-colored, greenish to black, light or heavy, waxy or not. When it comes to elemental composition, the proportions of the elements in petroleum surprisingly vary over narrow limits, taking into account the great diversity in physical properties [2]: C H N 83.0-87.0 10.0-14.0 0.1-2.0 wt.% wt.% wt.% O 0.05-1.5 S 0.05-6.0 Metals (Ni, V) < 0.1 wt.% wt.% wt.%

Despite the consistency in elemental composition, the crude oil system is extremely complex, including thousands of different compounds and physical properties which cannot be predicted by the elemental composition only. The precise chemical composition of crude oil components especially that of the heavy end of the system, is still speculative.

1.2 The Crude Oil System: SARA, and Petroleomics.

Due to the chemical complexity of crude oils, it has been normal practice to divide the crude oil system into fractions with respect to the solubility of the compounds, rather than to their elemental composition. The SARA fractions are such a family, composed of saturates, aromatics, resins and asphaltenes. Briefly, saturates are defined as the saturated hydrocarbons ranging from straight-chained paraffins to cycloparaffins (naphthenes) while the aromatic fraction includes those hydrocarbons containing one or more aromatic nuclei which may be substituted with naphthenes or paraffins. Asphaltenes are defined as the solubility class of crude oil that precipitates in the presence of aliphatic solvents while the resin fraction is defined as the fraction soluble

in light alkanes, but insoluble in liquid propane. Asphaltenes and resins are the most aromatic and most polar compounds of petroleum with the greatest amount of sulphur, nitrogen, and oxygen [2]. Asphaltenes have in particular been studied extensively because of their negative impact on several processes in petroleum production. However, until recently, the molecular structure of these crude oil compounds has been a mystery. The lack of proper understanding of asphaltenes has limited the predictive capabilities of petroleum science and resulted in phenomenological approaches. During recent years, the people at Florida State University have used the worlds highest resolution mass spectrometer, an Electrospray Ionization Fourier Transform Ion Cyclotron Resonance Mass Spectrometer (ESI-FT-ICR-MS) to reveal that petroleum crude oils contain more than 20000 compounds with distinct elemental composition (CcHhNnOoSs) [3]. Ultrahigh-resolution techniques open up for new possibilities in the area of crude oil characterization, a concept denoted as Petroleomics. Petrolemics is referred to as the next grand challenge for chemical analysis [3] and promises to revolutionize petroleum science. Fortunately, the Petroleomics concept has also proved to gradually reveal some secrets of the structure of asphaltenes. The next section is dedicated to recent advances in the understanding of the structure of petroleum asphaltenes. High resolution characterization techniques have also solved the mystery of naphthenate deposition (Figure 2.7 c) in crude oil processing. Krane and colleagues at NTNU [4] have recently reported on the exact molecular structure of the octaterpene tetracarboxylic acids and their geochemical origin.

1.3 Molecular Structure and Aggregation of Petroleum Asphaltenes

As indicated by the title of this section, the recent advances of asphaltene chemistry will be presented from a perspective which is similar to that presented by Mullins at the Annual SPE conference in Dallas, 2005 [5]. For the devoted reader, O.C Mullins, E. Y. Sheu, A. Hammimi, and A. G. Marshall will by September 2006 report on the topic: Asphaltenes, Heavy Oils, and Petroleomics (ISBN: 0-387-31734-1). Despite the tremendous strides that have been made for decades in the petroleum research community, the general perception of asphaltene structure is gradually converging. The wide range of reported molecular weights of asphaltene constituents, from a few hundreds to several millions, lead to speculation about self-association of asphaltene molecules [6]. Self-association and precipitation of asphaltene particles in concentrated paraffinic systems or as a result of pressure reduction of unsaturated live oils have been well documented. Structural studies have now made it clear that the extreme variation in reported molecular weight of asphaltenes is partly due to studies of aggregated systems, the molecular weight is closer to 1000 g/mol than several thousands or millions. Different characterization techniques have concluded on the chemical moieties in asphaltenes. Asphaltene carbon is roughly half (or slightly less) aromatic and half (or slightly more) saturate (13C NMR). Mean carbon chain length in saturated substituents is on order 4 (IR). Moreover, 13C NMR and carbon x-ray Raman spectroscopy (XRRS) indicate that asphaltene aromatic carbon is pericyclic ([5] and references herein). Several techniques have converged on the size of the fused ring systems, indicating that chromophores in crude oil asphaltenes typically have 4-10 fused

rings [5, 7]. X-ray absorption near edge structure (XANES) studies have characterized the sulfur and nitrogen chemical moieties, showing that asphaltene sulfur is mostly thiophenic and sulfidic with occasionally high sulfoxide concentrations. The asphaltene sulfur is more aromatic in more mature oils. Asphaltene nitrogen is in aromatic groups in pyrrolic and to a lesser extent pyridinic structures. Oxygen is largely in carbonyl functions and in OH, sometimes as carboxylic acid ([5] and references herein). Infrared spectroscopy (IR) and NMR have demonstrated that petroleum asphaltene hydrogen resides mostly on saturated carbon [7].

(a)

(b) (e)

(c)

(d)

Figure 1.1. Conformations of the C80 tetraacids responsible for naphthenate deposition problems according to Lutnaes et al. [4].

Largely based on compositional data, an archipelago model was proposed for the structure of asphaltenes. Figure 1.2 (a) shows the model with isolated aromatic ring systems cross-linked by alkane substituents. However, as discussed by Mullins [5], this model is in conflict with a molecular weight of about 750 g/mol. The archipelago model

with asphaltene molecular weight of 750 g/mol would result in small ring systems, thus giving asphaltenes a colorless appearance. The archipelago model with large ring systems would require much higher molecular weights, inconsistent with many recent studies. The molecular weight debate is possibly equivalent to determining whether asphaltenes are monomeric or whether they are polymeric [7]. Recent findings point in the direction that asphaltenes are monomeric with typically one or two chromophores (fused ring systems) per molecule. Figure 1.2 (b) shows some hypothetical asphaltene molecular structures that according to Mullins [5] are consistent with recent findings. Asphaltene properties are believed to vary continuously within the asphaltene solubility fraction and toward the resin fraction. There is no large discontinuity in asphaltene properties [7]. Primary aggregation threshold (critical nanoaggregate concentration, CNAC) in toluene has been reported at much lower concentrations (~60 mg/L) than previously investigated [5].

(a)

(b)

Figure 1.2. (a) A hypothetical asphaltene molecule according to Strausz [8] (Archipelago model). (b) Four hypothetical asphaltene molecules according to Mullins [5].

1.4 Light, Heavy and Extra Heavy Crude Oil

As the terminology indicates, the character of the crude oil product is usually defined with respect to its density. In the early years of petroleum industry, density was the principal specification for petroleum and refinery products [2]. Although the density gives no specific information about the chemical composition of the product, there are generally certain properties that directly or indirectly correlate to the density of the oil. The API (American Petroleum Institute) gravity indicates a transition from conventional to heavy oils at 20o API (0.934 kg/m3 at 60oF/15.6oC). Light oils have API gravity greater than 20, whereas extra heavy oils and bitumen have API gravity less than 10 (heavier than water). Heavy crude oils, which often result from bacterial oxidation of conventional oils, have different physical and chemical properties than conventional oils, generally greater viscosity, greater amount of asphaltenes, heavy metals, sulphur and nitrogen. The increased asphaltene content may give an impression of greater risk of asphaltene deposition problems during production. However, with the increased asphaltene content, also resin components increase in amount. The ratio of resins to asphaltenes is often more important than the absolute asphaltene content. In fact, severe asphaltene precipitation can often be encountered in reservoirs with very low total asphaltene amount whereas certain heavy oils may not show any appreciable deposition. The state of asphaltenes in crude oil is a function of every other component in the system. In asphaltenes, there is a balance of opposing effects; -bond stacking (via van der Waals force) reducing solubility vs. steric repulsion of alkane chains in the asphaltene molecule, increasing solubility [7]. Moreover, association of resin molecules increases the stability of the asphaltene unit (complexed or not). Molecules within the resin fraction are typically composed of highly polar end groups, which often contain heteroatoms such as O, S, and N as well as long, non-polar paraffinic groups. The attraction towards asphaltenes is a result of both hydrogen bonding through the heteroatoms and dipole-dipole interactions arising from the high polarities of the resins and asphaltenes. The paraffinic component of the resin molecule acts as a tail making the transition to the relatively non-polar bulk of the oil. Thus, the stability of petroleum can be represented by a phase system in which the asphaltene constituents, the aromatic fraction (including resins) and the saturate fraction are in a delicately balanced harmony [6]. Any alteration of this balance (composition, pressure, temperature etc.) may completely change properties of the crude oil system, i.e. rheology or deposition tendency. As mentioned previously, there is evidence that the structural aspects of certain components within resins differ very little from those of the corresponding asphaltene fraction, the main difference being the proportion of aromatic carbon within each fraction. Components within the same solubility fraction may also have very different chemical and structural features.

References
[1] [2] [3] [4] [5] [6] [7] [8] http://www.chevron.com/products/learning_center/crude J. G. Speight, The Chemistry and Technology of Petroleum. 3 ed. 1998: Marcel Dekker, Inc. 875. A. G. Marshall, R. P. Rodgers, Acc. Chem. Res. 37 (2004) 53. B. F. Lutnaes, . Brandal, J. Sjblom, J. Krane, Organic & Biomolecular Chemistry 4 (2006) 616 O. C. Mullins, SPE Annual Technical Conference and Exhibition, 2005, Dallas, Texas, U.S.A. J. G. Speight, Oil and Gas Science and Technology - Rev.IFP 59 (2004) 467. S. Badre, C. C. Goncalves, K. Noringa, G. Gustavson, O. C. Mullins, Fuel 85 (2006) 1 O. P. Strausz, M. T. W., E. M. Lown, Fuel 71 (1992) 1355.

2. PETROLEUM PROCESSING AND FLOW ASSURANCE

The image of James Dean dripping with black oil from his Texas gusher in the 1956 movie "Giant" may have been compelling, but it's not descriptive of today's oil producers. For one thing, the days when a gusher signaled a big discovery are long gone. Since the 1930s, oil producers have used blowout preventers to stop gushers. In addition, not all crude oils behave in the Hollywood manner. Some flow about as well as cold peanut butter [1]. This chapter is a very brief introduction to the processes in conventional and heavy oil production, in particular those used when recovering oil from offshore reservoirs. Recovery from oil sand resources (mining) has not been addressed here even though several aspects of the operation are similar to those encountered in conventional and heavy oil industry. Neither is the technology in discovering petroleum resources or the different exploration techniques (based on magnetism, gravity, sound waves, drilling) treated here. The wells have been completed and prepared for production of oil. With this starting point, crude oil recovery, water-oil separation, transportation, and refining are presented in Section 2.1-2.4. Section 2.5 deals with problems encountered in operation and some trends of todays petroleum production.

2.1 Recovery
Reservoir management is a formidable task and can only be done successfully with information of fluid characteristics, geological rock properties, reservoir geometry, reservoir depths etc. The course of the process is often divided into the three groups: primary, secondary and enhanced oil recovery based on the necessity of chemical, thermal or physical treatment methods in order to maintain efficiency of the recovery. The primary oil recovery largely depends on the natural driving energy of the reservoir available for moving oil from the reservoir to the wellbore. There are basically six driving mechanisms that provide the natural energy necessary for oil recovery: (1) rock and liquid expansion drive, (2) depletion drive, (3) gas cap drive, (4) water drive, (5) gravity drainage drive, and (6) combination drive. The gas cap drive and the water drive mechanisms are showed in Figure 2.1. Ahmed [2] has carefully described each mechanism. Although no reservoirs are identical in all aspects, they can be grouped according to the primary recovery mechanism by which they produce. It has been observed that each drive mechanism has certain typical characteristics in terms of ultimate recovery, pressure decline rate, gas-oil ratio, and water production [2]. Successful reservoir management involves early recognition of the type of drive mechanism present. In most cases, the natural driving mechanism is a relatively inefficient process and results in a low overall oil recovery. The lack of sufficient natural drive in most reservoirs has led to the practice of supplementing the natural reservoir energy by introducing some form of artificial drive. Secondary oil recovery refers to the additional recovery which results from injections of water or gas to maintain the depleted reservoir pressure and sweep the crude oil from the injection wells toward the producing well. Waterflooding is schematically presented in Figure 2.2 (b). A wide variety of injection-production well arrangements can be used

in injection projects even though Figure 2.2 (b) only shows a single injection well and a single production well. The result of secondary recovery depends on both the mechanism by which the injected fluid/gas displaces the crude oil (displacement efficiency) and the volume of the reservoir that the fluid/gas enters (sweep efficiency) [3]. Usually, the selected secondary recovery process follows the primary recovery, but it can also be conducted concurrently with the primary recovery.

(a)

Sea floor or land

(b)

Figure 2.1. Primary oil recovery mechanisms from Speight [3] (a) water drive and (b) gas cap drive.

Various methods of enhanced oil recovery (EOR) are essentially designed to recover oil, commonly described as residual oil, left in the reservoir after both primary and secondary recovery methods have been exploited to their respective economic limits. Figure 2.2 shows schematically how oil can be left behind in the reservoir by trappingand bypassing-phenomena. EOR processes use thermal, chemical, or fluid phase behavior effects to alter surface forces, fluid viscosity or the formulation in which the crude oil exists. EOR methods include polymer flooding, surfactant flooding, alkaline flooding, miscible fluid displacement, steam drive injection, cyclic steam injection, in

situ combustion, and many more. An introduction to the different methods is beyond the scope of this section, but has been reviewed by Speight [3]. Principles known from polymer, surfactant, and microemulsion science are put to practical use to enhance the recovery.

Oil (a) Oil

Oil

Oil

Water

Water

INITIAL CONDITION Production Well Oil Zone (b)

DURING DISPLACEMENT

AFTER DISPLACEMENT Bypassed Residual Oil

Water-Swept Zone Injection Well

Water-Swept Zone

Figure 2.2. Oil trapping and bypassing by waterflooding is presented similarly as by Speight [3]. (a) Trapping of oil in pores. Water flows faster through the narrow pore because its diameter increases the effect of capillary forces. Water will reach a common outlet before all the oil is displaced from the upper larger pore, which may be closed off, leaving behind residual oil. (b) Aerial view of a waterflooding process where water, injected through the injection well(s) displaces oil toward the production well. However, bypassed oil will result in limited recovery.

2.2 Quality of the Well Product

Unfortunately, the well product is rarely pure oil, but a complex mixture of oil, water, gases, inorganic material, production chemicals etc. The co-production of salt water is in particular problematic both due to separation problems and wastewater management. In a typical oilfield, well-streams are tied together and sent to a platform for separation. An example is the Troll West field on the Norwegian Continental Shelf which consists of 2 platforms, 27 clusters, and 106 wells. The platforms perform water/oil/gas separation and transport oil (more than 400 000 barrels/day) and gas with export quality to Mongstad and Kollsnes (via Troll A), respectively.

10

As discussed in Section 1, heavy oil is a chemically complex hydrocarbon system with indigenous surfactants, whereas the co-produced water often contains chlorides together with dissolved salts. Suspended solids can be present in both phases. Due to the indigenous surfactants, the water-oil system is very susceptible to droplet break-up processes. High shear regions, for example over the wellhead choke or in other process units can generate finely dispersed emulsions which might become extremely difficult to resolve completely. Such emulsions can concentrate over time as an emulsion band in separators and reduce the flux between water an oil phases, thus reducing the efficiency of the oil-water separation. Section 2.5 highlights some of the problems which need to be overcome during the separation of petroleum emulsions. Since this section in a way is ahead of its time, only some features of the processes which improve the quality of the well product will be presented here. The separation installation on a platform can vary significantly depending on the properties of the produced oil and water phases, but generally include gravity separators, electrostatic coalescers, hydrocyclones, pumps, heat-exchangers, sand management units, degassers, units for produced water treatment etc. Figure 2.3 shows the flowsheet of a separation installation on a platform, for example the Troll B platform on the Troll West field. This is not a heavy oil producing field. The flowsheet is simplified, but can still show the general separation methodology. Desander units have not been included. These are usually present either upstream of the first stage separator or as jet-systems at the bottom of the separators. In the latter case, nozzles whirl up sand and pump it with water for reinjection. Gas management may involve simple gas-liquid separation units upstream the first stage separator in addition to outlets from gravity separators and hydrocyclones. As shown in Figure 2.3, the water from the separators is sent via hydrocyclones to the clean side of the degassing vessel. The reject from hydrocyclones (about 4% oil) and the small amount of water from the electrocoalescer is sent to the oil side of the degassing vessel. Residual oil is skimmed off the water phase. Efficient operation of the plant requires constant monitoring of the performance of different installation units. Temporary shutdown of the process, due to separation problems or fouling of individual process units, involves huge financial losses and must be avoided. As the chemical and physical properties of well products may change considerably over time, compatibility of products from different wells is constantly evaluated to avoid potential fouling of the process. Compatibility of oil streams within the separation installation must also be carefully considered to avoid separation/deposition problems. Aspects of the separation will also be discussed in Section 2.5 with potential operational problems in mind.

11

1. stage separator

H-E

2. stage separator

Electrostatic coalescer

Flocculant Hydrocyclone Flocculation tank Hydrocyclone

Flocculant N2 gass Water treatment system Discharge water directly to sea or to reinjection Pump

Centrifuge

Degassing vessel

Clean Oily To sea To sea

Figure 2.3. Simplified flowsheet of the oil-water separation methodology on a platform like Troll B. Slightly modified compared to that presented by Larsen [4]. Read the comments to the flowsheet in Section 2.2.

12

2.3 Transportation
Crude oils are usually transported from the producing fields to the refinery by pipeline. Also, because of the high capital investment in offshore facilities, oil companies often decide to produce satellite fields from subsea completions tied back to existing facilities. Tiebacks are usually 10 to 20 miles long, but can be much longer. Transportation of multiphase systems over huge distances requires flow assurance considerations. A typical example is the Troika field in the Gulf of Mexico, located 150 miles offshore Louisiana under 2700 ft (820 m) of water. Troika was developed using a compact eightslot subsea manifold, tied back to the platform 14 miles (22.5 km) away. This required the longest multiphase subsea tie-back system in the Gulf of Mexico. The 25 cm in diameter production pipe is made of carbon steel and inhibited from corrosion by means of chemical injection into the subsea manifold. The annulus between the 25 cm and 61 cm pipes is pressurized with nitrogen to approximately 100 bars at 4.4C to prevent collapse of the carrier pipe in deep water. These pipelines are insulated to minimize paraffinic deposition and to provide reaction time for hydrate prevention, following an unplanned shut-down [5]. The breathtaking pipeline system crossing Alaska is another example of successful transport of crude oil. The Alyeska pipeline has transported about 15 billion barrels (2385 milliard liters) of oil over a distance of 800 miles (1287 km). For flow assurance, the pipe is heated to 50oC [6]. As this wasnt enough, Russia is currently planning to build the longest pipeline in the world, the East-Siberian Pacific Ocean Pipeline, more than three times the length of the Alyeska pipeline (2565 miles) [7].
(a) (b)

14 miles

2700 ft

Subsea manifold

Wellheads

Figure 2.4. (a) Troika well in the Gulf of Mexico at 2700 ft (820 m) water depth. The manifold is tied back to Shells Bullwinkle platform 14 miles (22.5 km) away [5]. (b) The famous Trans Alaska Pipeline System was finished in 1977.

For heavy crude oils special solutions must be applied to meet the viscosity specifications for pipeline transportation. According to Gateau et al. [8] the viscosity of petroleum systems should be restricted to below 400 mPas, whereas the viscosities of heavy crude oils may be as high as a few thousand to million of centipoises at reservoir temperatures. As correctly explained by Sanire et al. [9], based on their study of the structural properties of heavy crude oil [10], the origin of the high viscosity of heavy

13

crude oils is the structural formation (overlapping morphology) of asphaltenes, which are in great amount in these crude oils. Obviously, something has to be done in order to facilitate transportation of these systems. Sanire et al. [9] give an overview of the flow assurance methodology used for heavy oils. We will shortly define these transportation methods here. Section 2.5 deals with problems encountered in several stages of operation including deposition and fouling during transportation. The flow assurance methods for heavy oil either attempt to reduce the viscosity of the transported system (heating, dilution, inverting the emulsion to o/w type, or upgrading) or to reduce the friction between the transported system and the pipe wall (core annular flow). Viscosity has an exponential dependence on temperature or dilution, which makes these methods effective for heavy crude oil transport. However, there are certain drawbacks when using these flow assurance methods. Heating is costly in remote locations where energy is rare and it may introduce issues with regard to the design of the pipeline. Increased temperature can also result in greater corrosion rates of the pipe. Dilution implies greater pipeline capacity and depends on a secure access to diluent (naphtha, condensate, etc.). If the diluent is recycled, a large capital investment in additional pipelines is inevitably. Examples of the use of heat and dilution for assisted transport of crude oil are respectively the Alyeska pipeline (Figure 2.4) and the Orinoco projects in Venezuela. Inversion of viscous w/o emulsions can be done by controlling the water cut and adding/activating a surfactant which stabilizes o/w type emulsions. The major application of heavy oil aqueous emulsion is the commercial ORIMULSION, a bitumen emulsion consisting of 30% water and 70% extra heavy oil [11]. The ORIMULSION is sold directly as a fuel for electrical power plants [9]. Dehkissia et al. [12] have studied upgrading of heavy crude oil at the Doda oil field in Chad as an alternative to other flow assurance methods. They have demonstrated how a non-catalytic hydrovisbreaking technique which requires neither pre-distillation (topping) or post-deasphalting units could be used to meet the viscosity specifications for pipeline transport. One concern in pumping heavy oils over large distances is the higher pipeline drag which translates to higher energy requirements to run pumps. In addition to adding drag-reducing agents, the Core Annular Flow method has been used to reduce pipeline frictional losses. Oil and water is transported in a certain flow regime where the oil phase is in the centre of the pipe and water is flowing near the pipe wall. Water is injected in injection stations and lubricates the pipe wall [13].

14

2.4 Refining
A refinery is a complex network of integrated unit operations for the purpose of making different hydrocarbon products from crude oil. According to Speight [3], refinery processes can be divided into three major types: (1) separation, (2) conversion, (3) and finishing including reforming. Figure 2.6 shows the relative placement of different processes within the refinery. Cracking (hydro- and catalytic-) uses heat and pressure to crack heavy molecules into lighter ones. A cracking unit consists of one or more tall, bullet-shaped reactor and a network of furnaces, heat exchangers and other vessels. Visbreaking corresponds to a technique where the viscosity of heavy feedstock is reduced by controlled thermal decomposition. Coking is a process which uses heat and moderate pressure to turn non-distillable fractions (residua) into lighter products and a coal-like substance that is used as industrial fuel. Even though the distillation process from Figure 2.6 appears to be the first step in crude oil refining, pretreatment processes are used to remove hydrogen sulfide or chloride which are responsible for the corrosive attack and catalyst poisoning of refinery units. Desalting removes chlorides from the coproduced salt water with washing and subsequent emulsion separation in settling tanks, electrostatic coalescers, or chromatographic units (Figure 2.5).

Figure 2.5. The electrostatic dehydration and desalting technology from Aker Kvrner Process Systems provides state of the art crude oil treatment at oilfields and refineries.

15

Figure 2.6. The scheme shows the relative placement of unit operations in a refinery, from Speight [3]. See the text for a short description of the different operations.

16

2.5 Problems in Operation Steps

The problems encountered in production of petroleum are highly dependent on the properties of the processed fluids and the reservoir rock. In some parts of the world, operators face challenges due to high wax content, in other parts due to asphaltenes, sulfur, co-produced reservoir material etc. Problems can be found in all stages of operation. Problems can be due to rheological properties of the fluids, deposition and plugging, erosion, corrosion, catalyst poisoning, emulsion formation etc. The list is endless. Some aspects related to emulsion formation and deposition are treated here. Emulsions of crude oil and water can be encountered in reservoirs, at the well bore, at the platform, in transportation systems and refineries. Generally, these emulsions have to be broken to reach specified values of product quality, both for oil and produced water. Exporting oil not fulfilling the quality criteria (<0.5% w/o), results in large economic penalties. Due to the high viscosity of the continuous oil phase and the low density difference between oil and water phases, sedimentation of water drops in heavy crude oil is extremely time consuming. Conventional solutions for heavy oil applications include one, or often a combination of: (1) increasing the residence time in separators, (2) application of heat, and (3) increasing the drop size and distribution. Due to the squared dependence of sedimentation velocity and drop size (Stokes law, simplified system), any drop growth upstream of a separator will result in a dramatic improvement in the performance of gravity separators. Drop growth (or flocculation) can be achieved with the use of chemical additives like flocculants, electrostatic coalescers (separators), or in-line electrostatic fields. Electrostatic coalescers are the current workhorses of heavy oil production trains, but are usually utilized after second stage separators due to the possibility of short-circuiting the electric field when too much free water is present. However, Vetco Aibel has, as the first company in the world, come up with electrostatic modules to be used in first stage separators. The Vessel Internal Electrostatic Coalescer (VIEC) installation on the Troll C platform successfully improved the water and oil quality and has later proved its efficiency also for heavier crude oil-water systems [6]. Both Aker Kvrner (Compact Electrostatic Coalescer, CEC e.g. on Glitne) and Vetco Aibel (Low Water Content Coalescer, LOWACC) offer state of the art electrostatic separation systems to offshore industry. Gas can be another problem for heavy oil separation. The problems associated with gas production are foams, mousses and increased inlet velocity to the first stage separator nozzle, thus leading to high turbulence and droplet break-up. Section 1.3 dealt with the instability of crude oil systems, not of their emulsions with water, but of components present in the hydrocarbon phase. Obviously, precipitation and/or deposition of organic or inorganic material at pipe walls, in reservoirs or in equipment like heat exchangers, hydrocyclones etc. can cause serious problems with respect to flow conditions and efficiency of the process. Type of solids encountered during petroleum production can be: waxes, asphaltenes, hydrates, diamondoids, scales, calcium naphthenate, sulfur, and many more. The temperature in deep water, usually near 40F (4.4C), can create flow problems in risers and export pipelines. Some crude oils contain waxes which gel when cooled or which deposit on the cold pipe wall surface, gradually choking off the flow through the pipeline. Figure 2.7 (b) shows wax depositions from the Troika pipe (Figure 2.4 a) after pigging the pipeline. Pigging

17

devices fit the diameter of the pipe and scrape the pipe wall as they are pumped through the pipe. Other oils contain asphaltenes, which can destabilize due to large pressure drops, temperature changes, shear (turbulent flow), solution carbon dioxide (CO2) or other changes in oil composition (commingling of incompatible fluids from different wells). Crude oils that are susceptible to pressure-induced asphaltene precipitation are generally highly undersaturated; that is, the subsurface formation pressure is much higher than the bubble point. The crude oil can experience a large pressure drop without evolving gas. Once gas evolves, the light alkane fraction of the liquid phase is reduced, thus increasing the solvating power of the oil, thereby stabilizing asphaltenes. Figure 2.7 (d) shows asphaltene deposition in a pipeline. Particles, even in small amounts, can also cause huge problems for separation of water-in-oil emulsions as they gradually concentrate in the interfacial layer in separators. Speight [14] has reviewed the effect of asphaltenes and resin constituents on recovery and refining processes.

(a)

(b)

(c)

(d)

Figure 2.7. Deposition and fouling: (a) scale (BaCO3) deposition, (b) wax collected from pigging of the Troika field pipe [15], (c) hydrocyclone blocked with calcium naphthenate (C80-tetra acid) [16], (d) asphaltene deposit in a pipeline [17].

18

References
[1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] http://www.chevron.com/products/learning_center/crude/ T. Ahmed, Reservoir Engineering Handbook. 2nd ed. 2001, Houston, Texas: Gulf Publishing Company. J. G. Speight, The Chemistry and Technology of Petroleum. 3 ed. 1998, Marcel Dekker, Inc. 875. I. . Larsen, Presentation at the meeting for KmB Electrocoalescence II (2006) http://www.offshore-technology.com http://www.abb.com N. Thacker, The Longest Oil Pipe, Vladivostok Times. 2006. P. Gateau, I. Hnaut, L. Barr, J. F. Argillier, Oil and Gas Science and Technology-Rev. IFP 59 (2004) 503. A. Sanire, I. Hnaut, J-F. Argillier, Oil and Gas Science and Technology Rev.IFP 59 (2004) 455. I. Hnaut, L. Barr, J-F. Argillier, F. Brucy, R. Bouchard, SPE 65020. 2001: Houston. D. Langevin, S. Poteau, I. Hnaut, J-F. Argillier, Oil and Gas Science and Technology-Rev. IFP 59 (2004) 511. S. Dehkissia, F. Larachi, D. Rodrigue, E. Chornet, Energy & Fuels 18 (2004) 1156. A. Bensakhria, Y. Peysson, G. Antonini, Oil and Gas Science and Technology Rev.IFP 59 (2004) 523. J. G. Speight, Oil and Gas Science and Technology - Rev.IFP 59 (2004) 479. T. Y. Makogon, T. L. Johnson, K. F. Angel, The 4th International Conference on Petroleum Phase Behaviour and Fouling, Trondheim. (2003) S. J. Ubbels, M. Turner, The 6th International Conference on Petroleum Phase Behaviour and Fouling, Amsterdam (2005) O. C. Mullins, SPE Annual Technical Conference and Exhibition. 2005. Dallas, Texas, U.S.A.

19

3. EMULSION STABILITY AND STABILITY MECHANISMS

In petroleum industry, water-in-oil (w/o) or oil-in-water (o/w) emulsions can lead to enormous financial losses if not treated correctly. Knowing the particular system and the possible stability mechanisms is thus a necessity for proper processing and flow assurance. This chapter is a very brief introduction to the topic emulsion stability. The emphasis is on the factors affecting the type and resulting stability of emulsions, in particular those encountered in petroleum industry. Since the field of petroleum emulsions is still a vast area, the topics covered are selective rather than comprehensive. The aspect of emulsification is not the subject of this section, but will be discussed later in Section 4.3. Thus, interfaces are in general considered to be already enriched with some type of interfacially active components. First, in Section 3.1, the factors that determine the type of emulsion (w/o or o/w) will be treated. Then in Section 3.2 to 3.4 the criteria for emulsion stabilization and the course of events, finally leading to destabilization, are considered. Some material properties affecting film thinning and rupture are presented in Section 3.5 and 3.6. Finally, and based on the preceding sections, the stabilization of heavy crude oil/water systems is discussed. Several aspects of emulsion stabilization will still not be encountered in this thesis, either due to low relevance for petroleum emulsions or due to insufficient knowledge to the particular topic. Then it is good to know that it exist some excellent books on the stabilization of emulsions. If you havent already got them for Christmas, the Encyclopedic Handbook of Emulsion Technology [1] or Emulsions and Emulsion Stability [2], edited by Sjblom, are both good starting points. In the first chapter of the latter book, Dukhin, Sjblom, and Sther present a thorough review on the dynamic behavior of emulsions [3].

20

3.1 Emulsion Type and Phase-Inversion

Whether an emulsion is of the oil-in-water (o/w) or water-in-oil (w/o) type depends on a number of variables. Still, the very general formulation by Bancroft [4] (1913): A hydrophile colloid will tend to make water the dispersing phase while a hydrophobe colloid will tend to make water the disperse phase, can be used as a qualitative rule. It was Griffin who in 1949 introduced the first quantitative tool for predicting the type of emulsion formed from the surfactant hypofile-lipofile balance, HLB. The HLB concept assigns a number to a surfactant from its molecular composition. Low HLB-value of the surfactant tends to give w/o emulsions, whereas high HLB-value favors o/w emulsions. However, it has become clear that there are major problems with the HLB concept since the HLB take no account for the effective hypofile-lipofile balance of a surfactant in situ adsorbed at an oil-water interface [5]. Factors like temperature, electrolyte concentration, oil type and chain length, and cosurfactant concentration can all modify the geometry of the surfactant at an interface, thus changing the curvature of the surfactant layer. There has been increased understanding of the effect of such factors by studying macroemulsion concepts from microemulsion science: phase diagrams, critical packing parameter (CPP) considerations etc. [6]. Salager et al. [7] have reported on the topic: Heavy hydrocarbon emulsions making use of the state of the art in formulation engineering [7]. Binks and Lumsdon [8] have extended the concept of HLB to systems stabilized by inorganic particles. Just as the water or oil-liking tendency of a surfactant is quantified in terms of the HLB number, so can that of spherical particles be described in terms of its wettability via contact angle measurements. By increasing the hydrophobicity of the particles at fixed volume fraction of oil and water, it is possible to invert emulsions from o/w to w/o type. This inversion is known as transitional phase inversion and is demonstrated systematically in Paper 2 for silica particles modified with crude oil components, silanes or siloxanes. To be fair, this concept is directly based on Bancrofts statement from 1913. Binks and Lumsdon have also showed that transitional phase inversion is possible in emulsion systems of two types of silica particles, one hydrophilic and one hydrophobic [9]. Tambe and Sharma [10] have demonstrated transitional phase inversion in systems of calcium carbonate, barium sulfate, bentonite and silica modified to different extent by addition of stearic acid. Another unique feature of particle-stabilized emulsions is the possibility of inverting a system stabilized by only one type of particles simply by changing the oil to water ratio [11]. Binks and co-workers have reported this type of inversion, known as catastrophic phase inversion, for silica systems [12-15]. Catastrophic phase inversion took place at a particular volume fraction of the dispersed phase depending on the hydrophobicity of the particles and in which phase the particles were initially dispersed [12, 13]. We have demonstrated catastrophic phase inversion of silica particles coated with asphaltenes and resins in Paper 2 and studied droplet size distribution and stability of the emulsions. Figure 3.1 shows catastrophic phase inversion of emulsions stabilized by particles of intermediate wettability. The particles were coated with resins (for reference to Paper 2: 7000R particles). Stability to coalescence was reduced when approaching the inversion point. Kralchevsky et al. [16] have recently done a theoretical study on the thermodynamics of particle-stabilized emulsions. They have predicted the occurrence of

21

catastrophic phase inversion in particle-stabilized emulsions from the work of formation of the w/o and o/w emulsion type.
100 90 80 % Disperse Phase Resolved 70 60 50 40 30 20 10 0 20 stable o/w unstable o/w w/o

stable w/o

30

40

50 Oil (vol.%)

60

70

80

Figure 3.1. Catastrophic phase inversion of emulsions stabilized by 0.5% (w/v) particles with intermediate wettability (7200R). The ordinate axis represents the amount of disperse phase resolved after gravitational induced coalescence (1580 x g, 10 min).

Obviously, the unique phenomenon of catastrophic phase inversion in particle-stabilized emulsions could possibly be used in several industrial applications. Emulsions, which type and lifetime are easily tunable by the volume fraction of the two phases, can offer several advantages. Properties of the complete system (rheology, conductivity, properties due to additives dissolved in one phase etc.) can easily be altered by crossing the inversion point. Biodegradable particles or particles which permanently loose their stabilizing properties or even their solid state by changing process parameters, open up for endless possibilities. Catastrophic phase inversion is not possible in traditional surfactant-stabilized systems.

22

3.2 General Criteria for Stabilization Surface Forces

In general there are three coupled subprocesses that will influence the rate of breakdown processes in emulsions. These are aggregation, coalescence and floc fragmentation. Ostwald ripening (long term effect) is caused by molecular diffusion from small to large droplets. Coupled with aggregation and fragmentation, Ostwald ripening is a separate topic and will not be considered here. Figure 3.2 describes schematically two drops of an emulsion and the course of events which finally result in destabilization. The droplets are influenced by forces which reduce their distance and cause collision. When droplets come close, surface forces play an important role for the outcome of the collision. Theory of surface forces is described in all monographs and reviews concerning emulsion stability and can be reviewed there (Van der Waals interaction, electrical double layer theory (DLVO), hydrophobic interaction etc.) [17]. For low interfacial coverage of the stabilizing units, this collision of emulsion drops (Figure 3.2, ab) is usually terminated with their coalescence (bc). Depending on the interfacial conditions the zone of contact may be deformed, resulting in a liquid film which is thinned through lateral liquid displacement and ruptured at a critical thickness [18]. In emulsion systems with small drops (radii in the order of microns) the high capillary pressure causes drop approach without significant change of their spherical shapes [19]. In such case there is no film in the contact zone and the stability usually only depends on the surface-to-surface distance between the two drops. In both cases, due to thermal or mechanical fluctuations which lead to corrugations of the film surface, the film may rupture. However, if the characteristic time of coalescence is high compared to that of the fragmentation process, the drops may be separated (ba). In other words, flocculation can be reversible or irreversible depending on the characteristic time of the subprocesses. The flocculation process is described in detail in Section 3.3. If the barrier to coalescence is very high and the disintegration process is unlikely, state (d) may represent an energetically stable condition. Stability (instability) of an emulsion is caused by the coupling of coalescence and flocculation, which follow a multiplication principle rather than an additivity principle. This means that the total result of the application of a stabilizer (destabilizer) depends very much on both flocculation and coalescence processes [3].
(d) (a) (b)

(c)

Figure 3.2. Possible consequences from collision between two emulsion drops according to Danov et al. [18]. Two drops approach each other under the action of a driving force (ab) and can, after reaching a certain threshold distance, coalesce directly (bc) or via drop deformation (bd) and film thinning (dc).

23

3.3 Flocculation in Emulsions

Flocculation is the process in which emulsion drops aggregate, without rupture of the stabilizing layer at the interface. Flocculation of emulsions may occur under conditions when the van der Waals attractive energy exceeds the repulsive energy and can be weak or strong, depending on the strength of inter-drop forces. The rate of flocculation can be estimated from the product of a frequency factor (how often drops encounter each other) and the probability factor (how long they stay in contact). The first depends on the driving force for droplet movement whereas the latter depends on the interaction energy of the droplets [5]. The driving forces for flocculation can be (1) body forces, such as gravity and centrifugation causing creaming or sedimentation, depending on whether the mass density of the drops is smaller or greater than that of the continuous phase. Since drops of different size move with different velocities, drops are also subjected to aggregation during creaming/sedimentation. Gravitational sedimentation is used industrially in for example hydrocyclones or traditional settling tanks (2) Brownian forces or (3) thermocapillary migration (temperature gradients) may dominate the gravitational body force for very small droplets, less than 1 m [18]. With special applicability for industrial electrocoalescence of water in oil emulsions, electric fields can be used to assist flocculation of small water drops into larger ones that will settle more quickly under normal gravity conditions. Electrostatic effects arise from the very different properties of oil and water, water having dielectric permittivity and conductivity much higher than that of oil. Other surface forces which will influence the interaction energy are van der Waals attraction and steric/osmotic repulsion of drops from particles or adsorbed polymers with flexible chains protruding from the drop interface. Figure 3.3 shows steric stabilization of water drops with hydrophobic particles. The bridging effect schematized in Figure 3.3(a) for low interfacial coverage of particles may result in extremely stable flocculates if the contact angle is above 90 degrees.
w o w w w

> 90o
s

(a)

(b)

Figure 3.3. Sketch of hydrophobic particles which (a) at low interfacial coverage are bridging two water drops and result in irreversible flocculation. (b) At high coverage of the interface, one particle will not have contact with more than one drop which may result in a reversible flocculation.

24

3.4 Mechanism of Coalescence

Coalescence is the process in which two or more emulsion drops fuse together to form a single larger drop, and is irreversible as Figure 3.2 indicates. As already mentioned, for large drops approaching each other (no background electric field), the interfaces interact and begin to deform. A plane parallel thin film is formed, which rate of thinning may be the main factor determining the overall stability of the emulsion. The film thinning mechanism is strongly dependent on bulk properties (etc. viscosity) in addition to surface forces. Figure 3.4 shows two identical drops and the convective flux of the continuous phase causing film thinning. The interaction of the two drops across the film leads to the appearance of an additional disjoining pressure (sum of surface forces [20]) inside the film. Positive pressure corresponds to repulsion between the two film interfaces. The disjoining pressure is the major thermodynamic stabilizing factor against drop coalescence as described in more detail by Danov et al. [18]. However, the existence of a stable equilibrium state does not guarantee that a draining liquid film can safely reach this state. Experiments have shown that film surfaces have a wavy appearance rather than a plane-parallel appearance, and that the amplitude of these surface waves increases with the radius (r) of the film segment. These surface waves are believed to be the reason why calculations of the velocity of thinning for plane-parallel films are up to several orders of magnitude smaller than what observed experimentally [5]. The hydrodynamic instabilities, accompanying the drainage of the film, can rupture the film before it reaches its thermodynamic equilibrium state. There are however several kinetic stabilizing factors which suppress the hydrodynamic instabilities and increase the lifetime of the film. The next sections will deal with the effect of interface elasticity, viscosity and diffusivity.

Js

J H r convective flux

Figure 3.4. Sketch of two approaching drops separated by a distance H. Convective flow of liquid from the gap between the drops drags surfactants away from the contact region and cause surfactant fluxes: J and Js, from bulk and along the interface, respectively.

25

3.5 Interfacial Elasticity, Viscosity, and Diffusivity.

The interfacial mobility can be evaluated using several models, depending on whether the surfactant is initially located in the dispersed phase or the continuous phase. We will only deal with systems where the stabilizing unit is soluble in the continuous phase of the emulsion system. A drop collision, as depicted in Figure 3.4, causes a convective flux of continuous phase away from the contact zone. This may result in a tangential drag of surfactant molecules along the fluid flow. The appearance of gradients of surfactant concentration (interfacial tension) is then opposed by the dynamics of surfactant molecules by interfacial diffusion, bulk diffusion, Gibbs elasticity, or interfacial viscosity. Due to complex geometric considerations of the contact zone; the characteristics of the adsorbing unit; the mass transfer mechanism and rates; possibility for interfacial reactions or reorganization; and considerations of the interfacial coverage, the attempts to seek mathematical equations for the interfacial mobility are confusing. Describing the lifetime of emulsions can be even more confusing. Alternatively, some aspects of the interfacial mobility and its effect on the rate of film thinning have been stated here: (1) Two spherical drops with complete interfacial coverage have tangentially immobile interfaces and behave hydrodynamically as rigid spheres. (2) The mechanism of kinetic stability in coarse emulsions and fine emulsions is completely different due to the strong droplet deformation and flattening in a coarse emulsion. The deformation is not important for emulsions with very small droplets partly due to the small dimensions of the inter droplet film. (3) The Gibbs elasticity (EG) tends to reduce the gradients in adsorption (surface tension) by a tangential movement of surfactants from the menisci of the interface. This transfer (Marangoni effect) is coupled with a flux of continuous phase into the film center which reduces the rate of film thinning. (4) EG may damp the fluctuation capillary waves. (5) The Gibbs elasticity can only slow down the drainage, but cannot arrest it. The equilibrium state of the film is entirely determined by the repulsive part of the disjoining pressure, rather than by the Gibbs elasticity [21]. (6) The interfacial viscosity also slows down the drainage of the continuous phase because of dissipation of a part of the kinetic energy of the flow within the surfactant adsorption layer [18]. Interfacial viscosity includes both shear and dilational viscosities. (7) The bulk (J) and interface (Js) diffusion fluxes tend to reduce the interfacial tension gradients and restore the uniformity of the adsorption layer. High fluxes may accelerate the rate of film thinning [18]. (8) The location of the interfacially active components has a dramatic effect on the rate of film thinning [18]. The Bancroft rule generally defines the emulsion type, which for petroleum emulsions usually are of the w/o type. If an emulsion breaker is going to be

26

added in order to modify the rheological properties of the interfaces, its initial location will be important for the performance. (9) Film instability in electric fields has been studied much less than in gravitational fields. Certain observations point to different instability mechanisms than for traditional systems.

3.6 Steric Stabilization of the Interface

The presence of solids at interfaces may give rise to repulsive surface forces which thermodynamically stabilize the emulsion. As concluded by Aveyard et al. [11] and Binks [8], many of the properties of solids in stabilizing emulsion interfaces can be attributed to the very large free energy of adsorption for particles of intermediate wettability (partially wetted by both oil and water phases). This irreversible adsorption leads to extreme stability for certain emulsions and is in contrast to the behavior of surfactant molecules which are usually in rapid dynamic equilibrium between the oilwater interface and the bulk phases. A simple drop contraction experiment (Figure 3.5) can demonstrate the irreversible nature of particle adsorption. A water/toluene interface was covered with hydrophobic particles. When the interfacial area was reduced by withdrawing water from the capillary, the particles ran out of area and created a typical crumpled interface. The interface did not relax back to its original shape, but remained crumpled for a long time. According to the steric stabilization mechanism, coalescence requires the solid particles to be removed from the drop-drop contact region. Free energy considerations suggest that lateral displacement of the particles is most likely, since forcing droplets into either phases from the interface require extreme energies [10, 22, 23]. The steric stabilization effect for particles has already been pictured in Figure 3.3 where drop contact is prevented by a physical barrier, the particles. Other aspects of the steric stabilization of emulsion drops by particles have been discussed in the Introduction of Paper 2.
capillary toluene

water

Hydrophobic particles at the interface

Figure 3.5. The interface crumples like a paper bag when the drop volume is reduced. A crumpled interface which remains crumpled over time is a good sign of irreversible adsorbed material.

27

3.7 Raising an Issue on the Stabilization of Crude Oil-Water Systems

The emulsifying components of crude oil emulsions have been identified. However, the non-predictability and wild variation in the interfacial activity of these agents have frustrated many researchers. The important stabilization mechanisms have probably also been identified, but the difficulty still lies in the assessment of the relative level of influence that individual mechanisms and variables have over the final stability. For heavy crude oils with great viscosity, the hydrodynamic effects of droplet approach and collision are clearly important. Drag forces, film-thinning forces and impact velocity of water droplets are all strongly dependent on the viscosity of the continuous phase. For the coalescence step the steric component of the disjoining pressure seems to be of particular importance for the stability to film destruction. As an example, Wu et al [24] have concluded on the interfacial forces responsible for stability of w/o or o/w emulsions. Wu used an optical microscopic technique involving observation and analysis of droplet collision trajectories and shear-induced coalescence. For the w/o emulsions, steric repulsion was the main contributor to the emulsion stability. For the o/w emulsions, all interactions were exclusively DVLO-type forces and the electrostatic force played a major role in stabilizing the emulsion. From the authors point of view, the model presenting elastic interfaces tightly packed with uniform sheets of asphaltene molecules has got too much attention as the dominating stability mechanism in real petroleum emulsions. Formation of network-like films of crude oil components is usually fulfilled in model systems of low stability to asphaltene precipitation or with already precipitated asphaltenes. With the emerging knowledge of the chemistry and aggregation structure of asphaltenes, it is clear that several model studies of petroleum emulsions have been carried out with oil phases where asphaltenes were precipitated. By reviewing the phenomena of particlestabilization and adsorption/desorption at fluid interfaces, there are clearly several resemblances with what observed in model crude oil systems. The question that arises is whether these model systems correctly imitate real w/o petroleum systems, or not. In addition to the steric stabilization of emulsions, the efficiency of immediate stabilization of interfaces is important. Following droplet-break up the characteristic time of adsorption of stabilizing components must be fast in order to avoid immediate recoalescence of water droplets. The droplet size resulting from a given mixing process will have great importance for the overall stability of the emulsion through the surface energy of the droplets. Both the kinematics of droplets and the coalescence of interfaces will be completely different in a system if a background electric field is present as in industrial electrocoalescence. Several mechanisms which are present in gravimetrically-induced destabilization of emulsions are probably still present, but their relative level of influence is completely different due to electric forces.

28

References
[1] [2] [3] J. Sjblom, ed. Encyclopedic Handbook of Emulsion Technology, ed. J. Sjblom. 2001, Marcel Dekker, Inc.: Trondheim. J. Sjblom, ed. Emulsions and Emulsion Stability, 2nd ed. Surfactant Science Series. Vol. 132. 2005, Taylor and Francis: New York. S. Dukhin, J. Sjblom, . Sther, An Experimental and Theorethical Approach to the Dynamic Behavior of Emulsions, in Emulsions and Emulsion Stability, J. Sjblom, Editor. 2005, Taylor and Francis: New York. W. D. Bancroft, Journal of Physical Chemistry 17 (1913) 501. B. P. Binks, Emulsions-Recent Advances in Understanding, in Modern Aspects of Emulsion Science, B. P. Binks, Editor. 1998, The Royal Society of Chemistry: Cambridge. K. Holmberg, B. Jnsson, B. Kronberg, B. Lindman, Surfactants and Polymers in Aqueous Solution. 2nd ed. 2003: John Wiley & Sons Ltd. J.-L. Salager, M. I. Briceo, C. L. Bracho, Surface Forces and Emulsion Stability, in Encyclopedic Handbook of Emulsion Technology, J. Sjblom, Editor. 2001, Marcel Dekker, Inc.: Trondheim. p. 455. B. P. Binks, Current Opinion in Colloid and Interface Science 7 (2002) 21. B. P. Binks, S. O. Lumsdon, Langmuir 16 (2000) 3748. D. E. Tambe, M. M. Sharma, Journal of Colloid and Interface Science 157 (1993) 244. R. Aveyard, B. P. Binks, J. H. Clint, Advances in Colloid and Interface Science 100 (2003) 503. B. P. Binks, S. O. Lumsdon, Langmuir 16 (2000) 8622. B. P. Binks, S. O. Lumsdon, Physical Chemistry Chemical Physics 2 (2000) 2959. B. P. Binks, S. O. Lumsdon, Langmuir 16 (2000) 2539. B. P. Binks, C. P. Whitby, Langmuir 20 (2004) 1130. P. A. Kralchevsky, I. B. Ivanov, K. P. Ananthapadmanabhan, A. Lips, Langmuir 21 (2005) 50. P. M. Claesson, E. Blomberg, E. Poptoshev, Surface Forces and Emulsion Stability, in Encyclopedic Handbook of Emulsion Technology, J. Sjblom, Editor. 2001, Marcel Dekker, Inc.: Trondheim. p. 305. K. D. Danov, P. A. Kralchevsky, B. Ivanov, Dynamic Processes in SurfactantStabilized Emulsions, in Encyclopedic Handbook of Emulsion Technology, J. Sjblom, Editor. 2001, Marcel Dekker, Inc.: Trondheim. p. 621. D. S. Valkovska, K. D. Danov, I. B. Ivanov, Colloids and Surfaces A: Physicochemical and Engineering Aspects 175 (2000) 179. A. W. Angle, Chemical Demulsification of Stable Crude Oil and Bitumen Emulsions in Petroleum Recovery-a Review, in Encyclopedic Handbook of Emulsion Technology, J. Sjblom, Editor. 2001, Marcel Dekker, Inc.: Trondheim. p. 541. I. B. Ivanov, K. D. Danov, K. P. Ananthapadmanabhan, A. Lips, Advances in Colloid and Interface Science 114-115 (2005) 61. V. B. Menon, D. T. Wasan, Colloids and Surfaces 23 (1987) 353. D. E. Tambe, M. M. Sharma, Advances in Colloid and Interface Science 52 (1994) 1.

[4] [5]

[6] [7]

[8] [9] [10] [11] [12] [13] [14] [15] [16] [17]

[18]

[19] [20]

[21] [22] [23]

29

[24]

X. Wu, I. Laroche, J. Masliyah, J. Czarnecki, T. Dabros, Colloids and Surfaces A-Physicochemical and Engineering Aspects 174 (2000) 133.

30

4. GENERAL ASPECTS OF THE CHARACTERIZATION OF CRUDE OILS 4.1 From Live to Dead Oil
In its natural condition, i.e. high temperature and high pressure, the crude oil is referred to as live. The recovered system is generally a mixture of liquid and gaseous hydrocarbons and water. When this multiphase system is brought to the surface (platform) it experiences numerous depressurization steps and extreme shear forces. Pressure depletion alone can destabilize asphaltenes which can foul and clog oil production equipment at the well surface, in the borehole and even in the subsurface formation. Such deposits are often found at tubing locations corresponding to the saturation pressures of the crude oil. These observations are consistent with various experimental studies showing that the bulk of asphaltene precipitation from undersaturated crude oils occurs close to bubble-point pressures. The phenomenon is rationalized in terms of the reduced solubility of asphaltenes in the lower density crude created by depressurization. The composition is constant, but the volumes occupied by the C6- components are increasing more rapidly than those of the C7+, lesscompressible components of the oil. That is, the difference in molar volume between the asphaltene and the bulk oil is at maximum at the bubble point leading to a minimum in asphaltene solubility [1]. Below the bubble point, the volatile hydrocarbons evaporate from the liquid as a gas phase thereby changing the molar volume of the liquid phase and probably reestablishing some of its lost asphaltene solubility. Due to the cost of sampling and analyzing live crude oils and the increased risk associated with handling high pressure equipment (especially pressurized with gas) it is normal practice to work with dead crude oils. It is no surprise that the properties of the depressurized crude oil are different from that of the live oil. Chemical composition, the state of the colloidal system, and the viscosity of the oil are some properties that will be altered by flashing off the gas. Thus, laboratory tests performed on dead crude oil samples cannot directly imitate the properties at real conditions. Still, phenomenological studies can safely be undertaken in the laboratory setting, taking into account the restrictions that have been made. Dead samples can be re-pressurized with natural gas to also take into account the influence of a separate gas phase for the stability of multiphase systems.

4.2 From Conventional Oils to Heavy Oils

From a characterization point or view, moving from conventional to heavy oils introduce certain difficulties. Due to strong absorption in UV-VIS-NIR region, techniques which rely on transparency have no use for w/o emulsions of heavy crude oils. Thus, droplet size determination of w/o emulsion cannot be done with microscopy. Moreover, due to great absorption in the near-infrared region, the transmittance through very heavy oil is usually to low for normal detectors to operate properly even with a path length of only 1 mm. Heavy crude oils are also characterized by their great viscosity which introduce inertial effects and invalidate drop shape analyses when a drop is dilated (as in dilational relaxation experiments). The great viscosity of the

31

continuous phase of w/o emulsions and the low density difference between the two phases significantly slow down sedimentation of water droplets in oil, which makes it difficult to differentiate between stability of samples even in a high centrifugal field (centrifuge). In other words, the transition from analyzing conventional oil to heavy oil involves some complications.

4.3 Aspects of Emulsification

Most macro-emulsions require the input of considerable amount of energy for their formation and can only be stable in the kinetic sense. The process of emulsification during petroleum recovery and transportation is usually an undesired product of high shear and turbulence near the wellbore, over valves or in other process equipment. It is accompanied with the creation of new drops which size distribution depends on (1) fragmentation of water drops into smaller drops and (2) collision and coalescence of drops into larger ones [2]. The break-up of droplets depends strongly on the type and intensity of the flow which is determined by the rotational speed and geometry of the impeller, geometry of the vessel and material properties of the continuous phase (viscosity and mass density). Contrary, interfacial forces try to maintain a spherical shape of the droplet. If interfacially active components are present, their rate of adsorption should be fast in order to give sufficiently high coverage of the oil/water interface during the short period between two drop collisions [3]. Consequently, an important property of a surface active unit is its time of adsorption/desorption which may vary significantly depending on structural and chemical characteristics of the adsorbing unit. The break-up mechanisms and the corresponding equations [4] may be tested experimentally by knowing the design and power of the used emulsification apparatus, the material properties and the produced droplet sizes. In Paper 2 we studied the drop formation in particle-stabilized w/o and o/w emulsions. A very interesting phenomenon with respect to the drop formation process was identified. In experiments where the volume of the disperse phase was altered between 0.2 to 0.8 vol/vol, the total area of interface remained constant. Emulsions with great volume of the disperse phase gave large droplets, whereas systems with low volume of the disperse phase gave small droplets. All droplet size distributions were of a lognormal function. Thus, the total interfacial area was ultimately determined by the amount of particles present to stabilize such interfaces, which in our case was constant, 0.3 % wt./vol (Table 4, Paper 2). These findings point to a balance between break-up and coalescence processes which is usually found for non-stabilized emulsions. This seems reasonable, taking into account the low stability of a heptane/toluene-water interface without any stabilizing particles. In Paper 5, droplet size and distribution of 30 water-in-crude oil emulsions were measured with pulse field gradient spin echo nuclear magnetic resonance (PFGSENMR). In the laboratory as well as in the field, keeping track of the produced drop size distribution is of outmost importance as these properties to a certain extent define both the stability of the emulsion and the mechanism of droplet formation. Due to the non-

32

transparency of water-in-crude oils, such information is rarely reported with emulsion stability results. Thus the PFGSE-NMR technique opens up for several possibilities as it can quantify the amount, size distribution and type (w/o or o/w) of undiluted petroleum emulsion systems [5].

4.4 Choosing Parameters and Correct Experimental Conditions.

All the factors described in Section 3 (Emulsion Stability and Stability Mechanisms) and Section 2.5 (Problems in Operation Steps), which favor high emulsion stability are generally fulfilled for the petroleum fluids associated with our data matrix of 30 oils. Naturally occurring surfactants are present in high concentrations, solids and wax may be present, the w/o interfacial tension is low, and the bulk viscosity is high. To be able to understand the overall stability of water-in-oil emulsions the response parameter, emulsion stability, may be divided into stability to flocculation and stability to coalescence. Both phenomena should have their origin in certain physicochemical properties. We have investigated the stability to electrocoalescence and to gravityinduced separation. Paper 5 presents the physicochemical properties, bulk and interfacial properties, which were selected to get a greater understanding of the stability of w/o emulsion systems. These were compositional properties (SARA, water, solids), spectroscopic features in infrared and near-infrared region, stability of the colloidal system (pentane titration monitored with NIR), density, dynamic viscosity, molecular weight, acidity (TAN), interfacial tension (near-equilibrium), interfacial relaxation (dilation), and droplet size (PFGSE-NMR). On the way to this superior objective, predicting stability of undiluted crude oil-water emulsions, it is of outmost importance that the experimental design is thought-through. Even though the handling of viscous and non-transparent crude oil samples may introduce difficulties in the way that certain experiments are performed, there should be very good reasons for not using undiluted samples. Moreover, for certain experiments it must be decided on the relevant property to be used as a parameter. For example, the near-equilibrium value has been chosen as the relevant property describing the tension of the water-oil interface. Dynamic experiments have shown that surfactant adsorption from concentrated crude oils is extremely fast. We expect the time of adsorption to be fast in the turbulent flow characterizing the Ultra Turrax (mixing device). Moran et al. [6] have observed constant interfacial tension of emulsified bitumen droplets, which imply that near-equilibrium values were reached during the emulsification process.

33

4.5 Statistical Analysis Feature or Artifact?

Multivariate modeling of the stability of water-in-crude oil emulsions is not only a formidable task due to the effort of measuring and reproducing an abundant number of experiments. Also from a multivariate point of view, building a prediction model with highly correlating variables is very difficult. For petroleum systems we expect a high degree of covariance between certain material properties. We must therefore be sure that what we model is a real feature, not an artifact. With this in mind we have decided to allow for a careful investigation of the data matrix prior to performing the multivariate analysis. This time for accumulation of knowledge (Paper 5, Figure 1) is necessary to be able to model the results properly. Paper 5 present the recapitulation which will be the basis for statistical analyses and further work. The next paragraph from Esbesen [7] should justify this delay: While correlation is a statistical concept for linear relationships it is basically a neutral, phenomenological measure. Cause and effect deal with the interpretation of the deterministic relationships. Many people think of cause and effect when they use the term correlation. This is wrong. One should most specifically use all application and domain-specific knowledge when interpreting correlated variables with the aim to determine cause and effect. For example, a statistical survey amongst a number of small Danish rural towns showed that the number of babies born in the towns could be correlated fairly well to the number of storks found in the area (squared correlation of 0.75 no less!). However few people believe that storks bring babies nowadays! Examples of this confusion between descriptive statistics and causal interpretation are common in science and technology be aware [7].

34

References
[1] [2] J. X. Wang, J. S. Buckley, N. E. Burke, Society of Petroleum Engineers Production and Facilities (2004) 152. S. Dukhin, J. Sjblom, . Sther, An Experimental and Theorethical Approach to the Dynamic Behavior of Emulsions, in Emulsions and Emulsion Stability, J. Sjblom, Editor. 2005, Taylor and Francis: New York. K. D. Danov, P. A. Kralchevsky, B. Ivanov, Dynamic Processes in SurfactantStabilized Emulsions, in Encyclopedic Handbook of Emulsion Technology, J. Sjblom, Editor. 2001, Marcel Dekker Inc., Trondheim. 621. F. Groeneweg, F. van Dieren, W. G. M. Agterof, Colloids and Surfaces A: Physicochemical and Engineering Aspects 91 (1994) 207. A. A. Pea, G. J. Hirasaki, Nmr Characterization of Emulsions, in Emulsions and Emulsion Stability, J. Sjblom, Editor. 2005, Taylor and Francis: New York. K. Moran, A. Yeung, J. Czarnecki, J. Masliyah, Colloids and Surfaces A: Physicochemical and Engineering Aspects 174 (2000) 147. K. H. Esbesen, Multivariate Data Analysis - in Practice. 2002, lborg: Camo Process AS.

[3]

[4] [5] [6] [7]

35

5. CHARACTERIZATION TECHNIQUES
This section highlights some of the characterization techniques utilized in this thesis. Some characterization techniques have already been thoroughly described in one of the following papers. In such cases, the reader will be guided to the paper which describes the technique. To make this chapter interesting, each sub-section has been put together with the following in mind: (1) principle of measurement, (2) suitability for crude oil systems, and (3) limitations and potential for improvement. Unpublished work and figures will be given to show the power or the limitations of the characterization technique.

5.1 Fractionation of Maltenes by HPLC

High Performance Liquid Chromatography, HPLC, is an analytical technique for separation and determination of organic and inorganic solutes in any sample, especially biological, pharmaceutical, food, environmental, and industrial. In a normal-phase liquid chromatographic process a non-polar mobile phase permeates through a porous, solid stationary phase (polar) usually in the form of small uniform particles, packed into a cylindrical column. The sample is injected into the mobile phase, travels through the column and is retained by the stationary phase mainly depending on polarity. Adsorption interactions between sample components, the mobile phase and the stationary phase can be manipulated by the choice of mobile and stationary phases and flow conditions [1]. We put together an automated HPLC system from different modules and columns and made a method for fractionation of crude oil maltenes into saturate, aromatic, and resin fractions. The switching-column system and the fractionation method are described in detail in Paper 1, Section 2.2 (b). It is well known that a conventional crude oil system contain several thousand of different compounds [2]. Thus, it may sometimes feel meaningless to perform a group type fractionation of a crude oil into four groups when only certain small subfractions may be responsible for the phenomenon under investigation (increased emulsion stability, deposition etc.) Even though it is very difficult and occasionally almost impossible to identify single components in petroleum, it is not at all impossible to significantly improve the HPLC fractionation method by concentrating on the heavier fractions of crude oil. Of particular interest is a HPLC separation scheme which is applicable for identification of the molecular components within the asphaltene fraction. Used in combination with ultraviolet spectroscopy, HPLC can be used to determine the degree of condensation of polycyclic aromatic ring systems [3-10]. This technique has provided strong indications of the ring-size distribution of polycyclic aromatic systems in petroleum asphaltenes [8, 10-15].

36

5.2 Infrared Spectroscopy

Infrared spectroscopy is an excellent method for providing detailed information about the chemical constitution of organic materials. Whether the material being examined is a single compound or a mixture, infrared absorption spectroscopy offers valuable information about the hydrocarbon skeleton and about the functional groups in petroleum [10]. In combination with an attenuated total reflectance (ATR) unit (Figure 5.1 a), a drop of oil or milligrams of a solid sample are sufficient for a chemical analysis. Infrared radiation hits the sample which is supported by a 4 mm2 diamond crystal. The beam is then reflected once and sent to a detector. Infrared radiation in the range from about 10,000100 cm-1 is adsorbed and converted by an organic molecule into energy of molecular vibration. Spectra appear as bands rather than as lines because a single vibrational energy change is accompanied by a number of rotational energy changes [16]. It is with these vibrational-rotational bands, particularly those occurring between 4000 and 600 cm-1, that we have matched features of unknown samples with features of known chemical compounds. Although the IR spectrum is characteristic of the entire molecule, certain groups of atoms give rise to bands at or near the same wavenumbers regardless of the structure of the rest of the molecule. There are two types of molecular vibrations: stretching and bending. The various stretching and bending modes for a CH2 group in a larger hydrocarbon molecule are shown in Figure 5.1 (b). The Figure also shows notations for the vibrations as well as the characteristic appearance of the absorption bands. Only those vibrations that result in a rhythmical change in the dipole moment of the molecule are detectable in the IR.

Asymmetrical stretching (asCH2) ~2926 cm-1

In-plane bending or scissoring (sCH2) ~1465 cm-1

In-plane bending or rocking ( CH2) ~720 cm-1

Symmetrical stretching (sCH2) ~2853 cm-1 (a)

Out-of-plane bending or wagging ( CH2) 1350-1150 cm-1 (b)

Out-of-plane bending or twisting ( CH2) 1350-1150 cm-1

Figure 5.1. (a) Attenuated Total Reflectance unit (Golden Gate, Specac). (b) Vibrational modes for a CH2 group. X and indicate movement perpendicular to the plane of the page [16].

37

Due to the small amount of sample needed and the good resolution of acidic and aromatic absorption bands, IR is especially suitable for verification of any group type fractionation method. Breakthrough of aromatics to the saturate fraction or delayed cutoff between the aromatic and resin fraction can be seen as aromatic or acidic absorption bands, respectively. The spectral features of saturates, aromatics, resins and asphaltenes are discussed in Paper 1, Section 3.3 with reference to Figure 4, which shows the absorption in the fingerprint area. Here we will show the power of infrared spectroscopy with an example of a peak-to-peak analysis of a hexane-insoluble fraction of a crude oil from the Gulf of Mexico. If free of inorganic particles, this fraction should correspond to the asphaltene fraction. However, as Figure 5.2 shows, the hexane-insoluble fraction shows some unexpected features. It turned out that wax has co-precipitated with the asphaltenes. The very typical signature of wax is identified from a shift of the scissoring band of methylene (sCH2) resulting in a sharp shoulder at 1473 cm-1 and from the methylene rocking vibration (CH2) at 720 cm-1 in which all of the methylene groups rock in phase. This latter band appears as a doublet in solid samples. The wax signature disappears when the sample is washed with hot hexane. The signal-to-noise ratio in IR is generally determined by the detector used and the interference from atmospheric components like H2O and CO2. The spectra in Figure 5.2 have not been smoothened or manipulated in any way.

Attenuated Total Reflectance

hexane (cold) insoluble

hexane (hot) insoluble

1800 1700 1600 1500 1400 1300 1200 1100 1000 Wavenumber (1/cm)

900

800

700

Figure 5.2. Infrared spectra of two hexane-insoluble fractions of a crude oil filtrated with cold and hot hexane.

38

5.3 Near-Infrared Spectroscopy

The near-infrared region is found between the visible and middle-infrared regions (MIR) of the electromagnetic spectrum from 780 to 2500 nm (12820 to 4000 cm1). The absorption spectrum of crude oil in the NIR range consists of overtones and combinations of the fundamental molecular vibration bands (C-H, N-H, S-H and O-H) found in the infrared range. Features of the vibrational absorption are presented in section 3.2 (a) in Paper 1 and will not be treated here. However, the most characteristic signature of heavy crude oils in NIR is not vibrational absorption, but electronic absorption. Electronic -* and n-* transitions within aromatic ring systems in the asphaltene and resin fraction cause a strong, structureless absorption edge as showed in Figure 5.3. The experiment was performed in transmission mode by depositing asphaltenes onto a glass slide from a toluene solution. The spectrum of a crude oil is shown for comparison. The spectral location of the electronic absorption edge varies over a wide range, from the near-infrared for heavy oils and asphaltenes to the UV for gas condensates [17], but is in all cases exponentially dependent on the wavenumbers (cm-1). Figure 5.3 shows the asphaltene sample plotted also on a logarithmic ordinate axis. NIR spectroscopy is an excellent method for predicting the amount of heavy constituents (asphaltenes and resins) in crude oils.

4000

5000

6000

7000

8000

9000

10000

11000

Wavenumber (cm-1)

Figure 5.3. The structureless electronic absorption edge of pure asphaltenes (deposited onto a glass slide) and the typical NIR features of an undiluted heavy crude oil. The Urbach tail of heavy aromatic compounds in oil has an exponential dependence on the wavenumbers.

In addition to molecular absorption, the NIR spectrum is dependent upon several physical parameters, where the most prominent is scattering from particles. When the particle size increases, the amount of light scattered by the sample is changed. This is

39

LOG(Apparent Absorbance)

Apparent Absorbance

reflected in the NIR spectra as an upwards shift of the baseline. For slightly lossy dielectric spheres in the Rayleigh limit (r/ 0.05, where is the wavelength of incident light and r is the particle radius), the scattering and absorption processes contribute separately to the light extinction, represented by the particle cross-section ().

total = scattering + absorption

(5.1)

The ratio of scattering to absorption scales with r3, indicating the importance of particle size on the total light extinction. The relation between optical density (OD), light intensity (I), particle diameter (N) and particle cross-section (total) is given as:
OD = log( I0 ) = 0.434 N total I

(5.2)

I0 and I are the intensities of incident and transmitted light. The effect of multiple scattering is not accounted for in this equation. Details on light scattering and absorption in the near-infrared region can be found in the literature [18, 19]. The experimental setup used for NIR measurements needs some comments. Considerable scattering and strong absorption from heavy crude oils can easily result in distorted spectra as very little light reaches the detector (low transmittance). As a result of this, the heavy crude oils could not be analyzed in the sample compartment of the Bruker Optics MPA due to a minimum path length of 5 mm of the sample tubes. We have used a constant path length of 0.16x2 mm. In Paper 1, we used a backscatter probe in combination with a roughened copper plate and in Paper 5, in combination with a gold reference background. We simply made a device to fix the backscatter probe, the gold reference (Bruker Optics), and the two glass slides which were separated by a 0.16 mm Teflon spacer. The setup has proved to give reproducible spectra and ensures easy sample injection and cleaning.
Probe end: Illuminator Pick up NIR probe

Crude oil sample Teflon spacer (0.16 mm) and glass slides

Piston to Ag background

Figure 5.4. Picture of the NIR accessory. A backscatter probe is facing the sample and a gold background (reference from Bruker Optics). The crude oil is pressed between two glass slides with a Teflon sheet (0.16 mm) ensuring a constant path length. No dilution is needed.

40

5.4 Interfacial Tension

There are several different methods for determining the interfacial tension between water and oil. All have their own characteristic time and range of application and their own limitations. The pendant drop method used in combination with axisymmetric drop shape analysis (ADSA) is no exception. A balance of the gravity force () with the pressure difference across the interface, given by the Laplace equation: P = (1/R1+1/R2), provides the tension of the interface . The radii of curvature, R1 and R2, are determined from geometrical considerations/correlations whereas the density difference between the two phases, , is measured with an independent method. A detailed explanation to the geometrical correlations can be found in Sjblom [1]. The drop profile close to the supporting capillary is not used for the ADSA since the drop shape may be slightly distorted by the wetting properties of the capillary. The pendant drop technique is especially suitable for monitoring surfactant adsorption from the dynamic interfacial tension response. However the experiments must be performed with some caution with respect to the following three aspects: (1) Keeping the drop volume constant is essential. If the drop volume decreases over time due to a leaking syringe, the interface will be compressed. As opposed to small surfactants which are in constant equilibrium between the bulk and the interface, irreversibly adsorbed compounds will not leave the interface due to the increased interfacial pressure. (2) It is essential to calibrate the instrument in order to correctly take into account the gravity forces on the pendant drop. Calibration is usually performed prior to the measurement with a geometric shape with known size. However refraction of light is different in the continuous phase than in air. It is strongly recommended to calibrate the instrument with the capillary (determine the size with a slide caliper or with microscopy) which supports the investigated drop. (3) The total access to interfacially active components is more important than the concentration of the components in the solution. If the interfacially active components are present in the surrounding liquid, the amount of sample may have importance. If the interfacially active components are present in the drop, the size of the drop does matter. The interfacial access to active components (per unit interfacial area) will be greater in a big drop than in a small drop. To complicate it even more, it has been shown that the drop shape (the deviation from a perfect spherical shape) results in different sub-layer concentration at different positions on the interface. Sample in the capillary may also have importance. Despite certain difficulties in measuring the tension for crude oil water interfaces, the pendant drop technique can potentially provide several system properties which cannot easily be identified with other techniques: presence of certain production chemicals, transfer of components (e.g. acids) across the interface at high pH, and the kinetics of interfacial relaxation.

41

5.5 Dilational Relaxation

Dilational relaxation experiments were performed in the work leading to Paper 3 and Paper 4. The principles and the mathematical framework have been thoroughly presented in Paper 3 whereas the interpretation of results was discussed in Paper 4. Limitations of the measurements are also discussed here. Still, the instrumental setup needs a preliminary remark: dilational relaxation is not only dependent on interfacial properties but also mass transport phenomena from the bulk phases. The intrinsic properties of the interface can only be measured in an experiment where mass transport is suppressed.

Radio tube

Flexible tip Piezo actuator

Amplifier

Syringe

Figure 5.5. Picture of the dilational relaxation accessory.

The dilational relaxation accessory was changed prior to the study leading to Paper 4 since a more robust setup was needed for the viscous oil samples. Figure 5.5 shows the individual parts. The piezoelectric actuator was connected mechanically to a syringe and driven by a waveform generator (Agilent 33250A) with a homebuilt amplifier. The amplifier was made from a radio tube by Jan Ole Sundli at the electrical support desk at NTNU. The piezoelectric actuator (PI-844.6) was purchased from Physik Instrumente with regional office in Sweden. The actuator has a push/pull force capacity of 3000/700 N and a resolution of 9 which makes it useful for our application. A flexible tip (P176.6) was mounted onto the syringe piston to protect the actuator against accidental bending forces. Syringes were purchased from SGE (SGE gastight, 25 mL) through Teknolab in Norway. Other syringes have been tested but were not robust or leaked (Teflon tip). The mounting device was machined in aluminum by the people at the workshop for fine mechanics at the NTNU. A method was written in Matlab 7.0 to: (1) divide the raw data into subsections (windows); (2) perform a Fast Fourier Transformation algorithm within each window; (3) unwrap the phase angle between area and interfacial tension waveforms; (4) calculate the rheological parameters; (5) plot the results. Each step of the method is described in the method script.

42

5.6 Electrically Induced Droplet Flocculation and Coalescence

The Critical Electric Field (CEF) cell was developed in connection with the Flucha II program at the Ugelstad Laboratory, a fluid characterization program on conventional crude oils and condensates. From a macroscopic point of view, the principle is simple: a stable w/o emulsion is placed between two brass plates with a volume determined by the thickness of a Teflon spacer. The brass plates are connected to a power supply (Agilent Model 6634B) which can stepwise increase the voltage and measure the current through the emulsion. When a direct current (DC) electric field is put up over the two brass plates, water droplets align to ultimately form chains connecting the two electrodes. The short-circuiting of the field can be seen as a sudden increase of the current through the sample which indicates that the electric field has broken the emulsion. The corresponding magnitude of the electric field is the CEF value.

Plexiglass

Sample compartment Teflon spacer

<100 V

Brass plate (a) (b) (c) 50 mm

Figure 5.6. Critical Electric Field cell (right); stable emulsion drops (a); stable drops aligned between the electrodes (b); broken emulsion (c). Pictures are not from the current CEF cell.

The real advantage of the CEF setup is the possibility of quantifying the stability of very stable emulsions which are often encountered when working with crude oil-water systems. When gradually moving towards heavier crude oil systems we early recognized the importance of hydrodynamic forces when breaking such w/o emulsions. Paper 5 discusses the contribution from flocculation and film breaking to the measured CEF value. Although the CEF cell has been of great value, we realize that from an electrical point of view, the design of the cell is far from optimal. We will therefore in collaboration with our friends at the Department of Electric Power Technology, SINTEF Energy Research make a new version of the CEF cell which also takes into account the following: (1) Microscopic observation of the emulsion. In combination with the high speed nearinfrared camera, phenomena which have never been studied before can be observed in real crude oil systems. These include single droplet features, droplet-droplet interactions and droplet-electrode interactions. Droplet-electrode deposition will change the magnitude of the field. Figure 5.7 shows a quality test of the NIR camera.

43

Figure 5.7. Water drop in 20 mm (path length) crude oil (0.8 wt.% asphaltenes, 11.5 wt.% resins) (Electrocoalescence II project, FLIR Phoenix, InGaAs detector, 900-1700 nm, 320 x 256 resolution, 345 frames per second with maximum resolution)

(2) The electrical field lines between the two brass plates are probably significantly distorted by one or more of the following design features: sharp edges at the end of each plate; two holes (sample injection) in the upper brass plate. The ideal infinite flat plates are somewhat difficult to realize in a lab of reasonable dimensions. Finite sized plates produce a uniform field at the middle of the plate, but the high field at the edges creates a problem. The setup can be designed more cleverly by curving the electrodes to have either uniform field strength (Rogowski profiles) or maximum field strength at a certain position. (3) By increasing the sensitivity of the system other phenomena than the short-circuiting of the field should be detectable. Any charged object will have a corresponding charge in the electrode. Movement or coalescence of drops should therefore be detectable and can possibly be verified with microscopic observation.

44

5.7 QCM-D
The dissipative quartz crystal microbalance (QCM-D) technique was used in Paper 2 to study asphaltene and resin adsorption on silica, and explained theoretically in section 2.6 of that paper. Since we did not present a scheme of the device, we will do so here. An example of asphaltene adsorption/desorption will show the value of the QCM technique in surfactant systems.

(a)

(b)

(c)

Figure 5.8. The heart of the QCM is the AT-cut quartz crystal (1.4 cm in diameter) sandwiched between two gold electrodes. (a) The crystal is given a mechanical lateral oscillation by an oscillating potential difference. The circuit can be disconnected to follow the dissipation of a freely oscillating crystal (b) The larger gold electrode is facing the solution investigated, and the area where the adsorption is measured is slightly larger than (c) the smaller counter electrode [1].

A QCM device consists of a thin quartz disc sandwiched between a pair of electrodes. The principle is shown in Figure 5.8. Due to the piezoelectric properties of quartz, it is possible to excite the crystal to oscillate laterally (1-2 nm) by applying an AC voltage across its electrodes. The resonant frequency (f) of the crystal depends on the total oscillating mass, including solvent coupled to the oscillation. When a film is attached to the sensor crystal the frequency decreases. Provided that the film is thin and rigid, the decrease in frequency is proportional to the mass of the film and the QCM operates as a very sensitive balance. The mass (m) of the adhering layer is calculated using the Sauerbrey relation [20].
m = C f n

(5.3)

C = 17.7 ng cm-2s for a 5 MHz quartz crystal and n is the overtone number. The odd numbered harmonics can also be recorded, and in the case of a rigid and thin adsorbate, they will give the equivalent response as the fundamental frequency. However, in many

45

situations the adsorbed film is not rigid and the Sauerbrey relation becomes invalid. A film that is "soft" (viscoelastic) will not fully couple to the oscillation of the crystal and the crystal's oscillation will be dampened. The dissipation (D) of the crystal's oscillation is a measure of the film viscoelasticity. Dissipation is defined as:
D= Elost 2 Estored

(5.4)

, where Elost and Estored are the energies lost (dissipated) and stored during one oscillation cycle. The more viscous the adsorbed layer is, the more energy is lost from the oscillation. The dissipation of the crystal is measured by suddenly disconnecting the oscillator and recording the response of a freely oscillating crystal that has been vibrating at its resonance frequency. Figure 5.9 shows the dynamic frequency and dissipation of a silica coated crystal successively exposed to different hydrocarbon solutions: toluene, 1.0 wt % asphaltenes in toluene, 10 ppm dodecylbenzenesulfonic acid (DBSA). The study was performed as a continuation of Paper 2, where we demonstrated that the stabilizing efficiency of silica particles was significantly enhanced by adsorption of asphaltenes. The objective was to find a chemical which could potentially remove already adsorbed asphaltenes from the silica surface. We will highlight some of the features in Figure 5.9. (Start) The frequency (f) and the dissipation (D) were set to zero. (5 min) Toluene was injected to prove that the system did not respond to a surfactant-free solution. D and f remained unaltered, as expected. (10 min) Asphaltenes adsorbed fast, in high amount, and as a relatively rigid layer when exposed to the silica surface. The higher order overtones were progressively more sensitive to the region close to the silica surface and showed increasing responses to the adsorption (f(7) > f(5) > f(3)). (15 min) The asphaltene solution was injected again to show that the surface was more or less saturated. (20 min) When rinsed twice in toluene, the resonance frequency increased and the dissipation factor decreased, suggesting that weakly bound asphaltenes desorbed. The adsorbed layer was rigid. A considerable amount of asphaltenes (5.6 mg/m2) was irreversibly adsorbed and could not be removed with toluene. (30 min) The oil soluble DBSA solution was injected twice, resulting in a significant increase of the resonance frequency. Asphaltenes were removed from the silica surface by DBSA. (40 min) The surface was rinsed twice with toluene.

46

50 40 30 20
Frequency (f, Hz)

Asphaltenes DBSA 10 ppm

7.5
n=3 n=5 n=7
Toluene

6 4.5 3 1.5
Dissipation (D, E-6) E-6)

Toluene

10
Toluene

0 -10 -20 -30 -40 -50 0

10 20 30 40

0 -1.5 -3 -4.5 -6 -7.5 50

Time (min)

Figure 5.9. QCM-D measurement of the adsorption/desorption behavior of asphaltenes (1.0 wt % in toluene) and DBSA (10 ppm in toluene) onto a silica surface. The change in frequency (f) and dissipation (D) was monitored for the 3rd, 5th and 7th overtone of the fundamental resonance frequency of the crystal. The solid/water/toluene contact angle of the crystal, measured into the water phase, was determined at start, after asphaltene adsorption and at the end.

47

5.9 Contact Angle Measurements

In Paper 2, we performed contact angle measurements in order to determine the wettability of silica surfaces modified with silanes, siloxanes, and crude oil components (asphaltenes and resins). Here, we will briefly present the very general theory of contact angle determination and then address some limitations of the method, when used to predict partitioning of particles at liquid interfaces. Contact angle experiments are in general performed in order to determine the wetting characteristics of a solid, by a liquid. The contact angle is defined by the Young equation, which is a force balance on the three-phase contact line. Gravity forces are neglected for small droplets. Figure 5.10 (a) presents the Young equation and the surface forces on the solid/phase1/phase2 contact line.
12
(a)

(b)

s1
s2 = s1 + 12cos (c) emulsion 1-in-2

s2

emulsion 2-in-1 2

1 2 1

Figure 5.10 (a) Surface forces (Young equation) acting on the solid/phase1/phase2 contact line determine the wettability of the solid. The phases may be insoluble liquids or a liquid and a gas. (b) Wettability of reservoir sandstone from Buckman [21]. Water droplets on hydrophobic clay (left) and hydrophilic quartz (right) (c) Particles are initially dispersed in phase 2. The wetting character of the particles determines the type of emulsion formed.

The wettability of oil-bearing rocks and its effect on imbibition processes significantly impact the amount of oil that can be recovered from an oil reservoir and the rate at which the oil is produced. Thus, it is of huge practical and economical interest to understand and control the changes in reservoir wettability brought about by the action of exploration and by production fluids such as acidic mud, scale inhibitors, surfactants and other chemical treatments. Figure 5.10 (b) shows an ESEM (Environmental Scanning Electron Microscopy) micrograph of oil and water wetting of a typical reservoir sandstone containing quartz grains (right) and the clay: illite (left) within the pore spaces. Surface wettability has also proven to be an important property for particlestabilized emulsions as it determines the partition of the particles across the water/oil interface. The interfacial partitioning defines the type of emulsion formed (Figure 5.10 c) and in many cases the stability of the system. The supereminent objective is thus to be able to predict the partitioning of a particle across a real water/oil interface. For large

48

particles, microscopic techniques may provide the true contact angle. However for micrometer sized particles, which are of particular relevance for real emulsion systems, we have to rely on other experimental techniques. The far most used approach is to compress a disk of the particles and then measure the contact angle of a sessile water or oil drop. We can generally live with effects caused by minor surface roughness (hysteresis etc.). However major experimental errors occur due to capillary forces in the disk. When the particle disk is immersed in the oil phase, oil will fill the voids of the porous disk. When the water drop is placed at the solid surface it will be partially repelled by the oil in the disk and give too high . Similar effects are seen if the disk is first contacted by water or if the measurements are performed in air. Thus the true contact angle of a single particle cannot be found by this method. It will still be possible to rank particles according to their hydrophobicity as we did with unmodified and modified silica particles in Paper 2. An example is showed in Figure 5.11. The influence of asphaltene and resin adsorption on the wettability of the hydrophilic silica was dramatic. In real systems, the change in particle wettability due to adsorption of resins and asphaltenes from crude oil is the primary reason for the accumulation of particles at the oil-water interface.
air water

Pure silica: = 14 o

Resin coated: = 73 o

Asph+ R coated: = 84 o

Figure 5.11. Surface modification of silica particles with crude oil components dramatically changed the wetting properties and the stabilizing effect of these particles in water-oil emulsions. The reported contact angles are the solid/water/air contact angle measured into the water phase.

49

References
[1] J. Sjblom, G. ye, W. R. Glomm, A. Hannisdal, M. Knag, . Brandal, M.-H. Ese, P. V. Hemmingsen, T. E. Havre, H.-J. Oschmann, H. Kallevik, Modern Characterization Techniques for Crude Oils, Their Emulsions and Functionalized Surfaces, in Emulsions and Emulsion Stability, J. Sjblom, Editor. 2005, Taylor and Francis: New York. A. G. Marshall, R. P. Rodgers, Acc. Chem. Res. 37 (2004) 53. T. V. Alfredson, Journal of Chromatography 218 (1981) 715. C. Bollet, J. C. Escalier, C. Souteyrand, M. Caude, R. Rosset, Journal of Chromatography 206 (1981) 289. J. Chmielowiec, J. E. Beshai, A. E. George, Fuel 59 (1980) 838. J. M. Colin, G. Vion, Journal of Chromatography 280 (1983) 152. S. Coulombe, H. Sawatzky, Fuel 65 (1986) 552. G. Felix, C. Bertrand, F. Vangastel, Chromatographia 20 (1985) 155. A. E. George, J. E. Beshai, Fuel 62 (1983) 345. J. G. Speight, The Chemistry and Technology of Petroleum. 3 ed. 1998: Marcel Dekker, Inc. 875. M. L. Lee, M. S. Novotny, K. D. Bartle, Analytical Chemistry of Polycyclic Aromatic Compounds, ed. A. P. Inc. 1981, New York. A. Bjorseth, Handbook of Polycyclic Aromatic Hydrocarbons, ed. M. D. Inc. 1983, New York. J. C. Monin, R. O. Pelet, in Advances in Organic Geochemistry, Bjoroey, Editor. 1983, John Wiley and Sons Inc.: New York. S. D. Killops, J. W. Readman, Organic Geochemistry 8 (1985) 247. T. Yokota, F. Scriven, D. S. Montgomery, O. P. Strausz, Fuel 65 (1986) 1142. R. M. Silverstein, Spectroscopic Identification of Organic Compounds. 6 ed. 1998, New York: John Wiley & Sons, Inc. O. Mullins, S. Mitra-Kirtley, Y. Zhu, Applied Spectroscopy 46 (1992) 1405. O. C. Mullins, Analytical Chemistry 62 (1990) 508. O. C. Mullins, N. B. Joshi, H. Groenzin, T. Daigle, C. Crowell, M. T. Joseph, A. Jamaluddin, Applied Spectroscopy 54 (2000) 624. P. Ekholm, E. Blomberg, P. Claesson, I. H. Auflem, J. Sjblom, A. Kornfeldt, Journal of Colloid and Interface Science 247 (2002) 342. J. O. Buckman, Wettability Studies of Petroleum Reservoir Rocks, in ESEM Application Note, Department of Petroleum Engineering, Heriot-Watt University: Edinburgh.

[2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21]

50

6. MAIN RESULTS Paper 1: Group-Type Analysis of Heavy Crude Oils Using Vibrational Spectroscopy in Combination with Multivariate Analysis.
The focus of this paper was to investigate the ability of near-infrared (NIR) and infrared (IR) spectroscopy in combination with partial least-squares regression to predict the amount of SARA components in heavy and particle rich crude oils. The response variables were SARA composition determined experimentally with a semi-preparative fractionation procedure, involving high-performance liquid chromatography (HPLC).

Figure 6.1. Correlation between modelled and measured amount of resins from partial least squares calibration (stars) or prediction (circles) models of spectral features in the NIR (left) and IR (right) region.

Twenty crude oils were analyzed especially for vibrational features in the NIR and IR region. Pre-processing of the spectra was necessary in order to remove scattering contributions from particles and residual water, giving a baseline shift especially in the NIR region. Parallel experiments showed good reproducibility for the spectroscopic measurements and the fractionation method. NIR spectroscopy also proved to perform well for the prediction of SARA components with prediction variances (RMSEP) of 2.82 wt.% (S), 1.47 wt.% (A), 1.46 wt.% (R), and 0.44 wt.% (asphaltenes). The prediction vs. measured amount of resins is presented graphically in Figure 6.1. The resin and saturate fractions were also predicted excellently from IR data. However adequate models for the aromatic and asphaltene contents were not obtained. We hypothesized that the similarity in the vibrational features from aromatic carbon of asphaltenes and aromatics caused this. We also know that small amount of solids, with especially strong vibrational signatures could have complicated model development.

51

Paper 2: Particle-Stabilized Emulsions: Effect of Heavy Crude Oil Components Pre- Adsorbed onto Stabilizing Solids.
The presence of inorganic particles at oil-water interfaces can cause severe operational problems if emulsions are too stable to be destabilized by traditional methods. Even if the initial bulk concentration of inorganic particles is low, particles tend to accumulate in process equipment over time and build up sludge layers of oil/water/particles. Such layers require aggressive treatment methods or removal. When exposed to crude oil, inorganic particles may be modified by the adsorption of heavy crude oil components like resins and asphaltenes. The coating of these components will completely modify the stabilization profile compared to that obtained with no coating. Herein a study of model emulsions stabilized by means of silica nanoparticles coated with asphaltenes and resins is presented. The stabilization efficiency of the particles was explained from a thorough characterization of their surface properties, including spectroscopy, contact angle measurements () and zeta potential measurements. The stabilization efficiency was greatly enhanced by adsorption of crude oil components onto very hydrophilic or very hydrophobic silica. Unmodified silica particles (w/s/air=14o) did not act as good stabilizers due to their very hydrophilic character whereas silica particles coated with resins (w/s/air=73o, Figure 6.2 top) or asphaltenes (w/s/air=84o, lower) gave emulsions of large droplets which were very stable to coalescence (induced by a high centrifugal field). However, when the particles were suspended in the water phase prior to emulsification, emulsion stability was significantly reduced. We demonstrated catastrophic and transitional phase inversion, brought about by simply controlling the volume fraction of the disperse phase and the degree of surface modification, respectively. The stability of the emulsions decreased progressively with increasing volume fraction of the disperse phase, in line with increased drop size. Droplet size distributions of stable emulsions revealed that the total interfacial area of a system was directly determined by the amount of particles present. Thus, the total interface area remained constant when changing the volume fraction of the disperse phase. Such observation is consistent with a mixing mechanism characterized by both droplet fragmentation and coalescence processes.

52

100 Particles suspended in water Particles suspended in oil 80 % Disperse Phase Resolved

60 o/w emulsions 40

20

0 20

30

40

50 Oil (vol.%)

60

70

80

100 Particles suspended in water Particles suspended in oil 80 % Disperse Phase Resolved o/w emulsions 60

40

o/w w/o

20

0 20

30

40

50 Oil (vol.%)

60

70

80

Figure 6.2. Particle-stabilized emulsions of silica (Aerosil 200) coated with resins (top) and asphaltenes (lower). The ordinate axes show the amount of disperse phase resolved after 10 min centrifugation at 1580 x g.

53

Paper 3: Viscoelastic Properties of Crude Oil Components at Oil-Water Interfaces. 1: The Effect of Dilution.
Dilational relaxation properties of surface active components are probably of great importance during droplet fragmentation and coalescence. The oscillating pendant drop method was used to study diluted crude oil-water interfaces and the effect of altering aromaticity of the diluent and the concentration of crude oil. A careful calibration of the axisymmetric drop shape analysis was performed with respect to the influence of inertial forces. We defined a capillary number as the ratio of viscous and interfacial forces of clean o/w interfaces. The influence of viscous forces could only be safely neglected below a capillary number indicated by an arrow in Figure 6.3 (top). Increasing capillary number by: (1) greater viscosity difference between the two phases, (2) faster frequency of oscillation, (3) greater volume amplitude, (4) smaller capillary diameter, or (5) lower interfacial tension invalidated the axisymmetric drop shape analysis due to inertial effects. The storage E and loss E moduli of one of the crude oil/water systems is presented in Figure 6.3 (lower). At a perturbation frequency of 0.1 Hz, the equilibrium storage and loss moduli passed through distinct maxima as a function of bulk concentration. The apparently low viscoelasticity of the interfaces in systems with high bulk concentration was at least partly caused by high diffusion flux of interfacially active components from bulk. A direct relation between the measured interfacial relaxation parameters and the overall emulsion stability was not identified. The overall stability to gravity-induced separation seemed to be dominated by the aggregation state of asphaltenes and the hydrodynamic resistance to sedimentation in concentrated systems.

54

0.12

0.10

0.08

a / 0

0.06

0.04

0.02

0.01

0.02

0.03 0.04 Capillary Number

0.05

0.06

0.07

40 35 30 Modulus (mN/m) 25 20 15 10 5 0 -6 10

E' 0%H E'' 0% H E' 70% H E'' 70% H E' 100% H E'' 100% H

10

-5

10

-4

10 Oil conc. (v/v)

-3

10

-2

10

-1

10

Figure 6.3. Dilational relaxation experiments. The relative interfacial tension amplitude is presented as a function of the capillary number for clean o/w interfaces. The inset shows values for Ca < 0.01 (left). Storage and loss moduli (0.1 Hz) of diluted crude oilwater interfaces as a function of concentration and aromaticity of the diluent (right).

55

Paper 4: Viscoelastic Properties of Crude Oil Components at Oil-Water Interfaces. 2: Comparison of 30 Oils.
With good reasons, our industry partners advertise for characterization of undiluted crude oils and oil-water interfaces. The importance of working with undiluted crude oils instead of model systems when dilational properties of real oil-water systems are going to be reproduced in the laboratory setting has been discussed. For such studies, molecular exchange mechanisms and the aggregation of asphaltenes are too dependent on concentration to justify the use of model compounds, i.e. fractionated asphaltenes diluted in a solvent. However for highly viscous crude oils, dilution may be necessary to avoid inertial effects which invalidate the drop shape analysis. Thirty crude oils were diluted in 30 vol% toluene and analyzed with respect to the viscoelastic response to a sinusoidal modulation in the frequency range from 0.01 to 1 Hz. Diluted and undiluted systems were compared. Moreover, long-time dynamic interfacial tension experiments of static drops of undiluted crude oil in water were performed. Dynamic interfacial tension experiments showed that the o/w interfacial tension decayed logarithmically with time to near-equilibrium values in the range from 8 (5) to 24 mN/m during 12 hours of ageing. As expected in the low frequency range (0.01-1 Hz), molecular exchange from bulk strongly affected the measured dilational parameters. For this reason the systems which exhibited particularly low magnitude of the dilational modulus were of the heaviest crude oils in the sample set, whereas the systems with greatest dilational modulus were among the lightest crude oils (Figure 6.4, top). The frequency dependence of the dilational modulus increased with its magnitude as expected for diffusion-controlled relaxation of soluble films. The characteristic time of relaxation was less that 10 sec. as indicated by the maximum or increasing loss modulus in Figure 6.4 (lower) for the undiluted crude oils. Overall, the undiluted crude oil-water interfaces had similar relaxation characteristics as the diluted samples except for slightly reduced magnitude of the dilational modulus.

56

35 Diluted oils 30 Undiluted oils 24-27oAPI 24 25 |E (1)| (mN/m) 20 9 9 19 17 17 23 23 24

20 22 11 11 7 26 20 26 30 30

15 10

12-19oAPI

7 19

5 0 0.0 2.0 4.0 6.0 d |E |/d log (mN/mHz) 8.0 10.0

8 7 6 Loss Modulus E '' (mN/m) 5 4 3 2 1 0 0.01 24 23 17 14 26 2 5 20 7 30

0.1 Frequency (Hz)

Figure 6.4. Dilational relaxation properties of diluted (0.7 vol/vol in toluene) and undiluted crude oil-water interfaces. The dilational modulus is plotted as a function of its frequency dependence (top). The loss modulus and its frequency dependence are plotted for undiluted crude oil-water interfaces (lower).

57

Paper 5: Stability of Water/Crude Oil Systems Correlated to the Physicochemical Properties of the Oil Phase
A characterization of thirty crude oils was performed to investigate the relative level of influence that individual parameters have over the overall stability of w/o emulsions. The crude oils were analyzed with respect to bulk and interfacial properties and the characteristics of their w/o emulsions. As expected, a strong covariance between several physicochemical properties was identified. The characteristic time for droplet approach was predicted from a simplified force balance which modeled drag forces by Stokes law and dielectrophoretic forces by a point-dipole approximation. Atten (J. Electrostatics, 30, 1993) had previously presented expressions for the characteristic time of droplet approach in a constant electrical d.c. field (equation 1). The comparison of the experimental time for destabilization with the theoretical time of droplet approach is showed in Figure 6.5 (left) for a water-in-heavy oil emulsion with droplet size of 6 m. Given enough time, the water-in-heavy oil emulsions could be destabilized even at very low electric field magnitude (0.4 kV/cm). The time for destabilization seemed to follow the exp (Eo)-2 relationship as predicted by equation 1.

theo

8 = 15 E02

5/ 3 1 6w

(1)

80000

10 x theo
60000

Short circuiting at E0 = 0.4 kV/cm after 14 hours

1000 750 exp 500 250 0 0.004 kV/cm s -1

40000

20000

exp (s)

0 0.0 1.0 2.0 E0 (kV/cm) 3.0 4.0

50

100 150 theo

200

250

Figure 6.5 Time for destabilization compared to the theoretical time for droplet approach in an electric d.c. field. Destabilization of a water-in-heavy crude oil emulsion with constant electric field magnitude (left). Destabilization of 30 water-in-crude oil emulsions at a field rate of 0.004 kV/cm s-1(right).

58

The expression for the characteristic time was modified for a constantly increasing field magnitude (equation 2). Figure 6.5 (right) shows the experimental time for destabilization compared to the theoretical time for 30 water-in-crude oil emulsions at one of the studied field rates (0.004 kV/cm s-1). Surprisingly, the simple model captured the difference between the 30 crude oils reasonably well also at other field rates (showed in paper).

theo

8 dE = 0 5 dt
1/ 3

2 / 3

5/ 3 1 6w

1/ 3

(2)

It was proposed that the destabilization of static water-in-heavy crude oil emulsions in an electric field was predominantly retarded by the viscosity of the oil phase. When droplets approach each other in an inhomogeneous electric field the field magnitude increases greatly. Strong dielectrophoretic forces disintegrate the films and result in coalescence. The relative contribution from film stability to the overall emulsion stability may therefore be low.

59

CONCLUDING REMARKS
As apparent from the introduction to the production of petroleum and the topics of the five attached papers, the recurring theme in this thesis is the role of colloid and surface chemistry. As concluded previously by PhD candidates at the Ugelstad Laboratory, the proper understanding of this discipline is essential for successful operation and development within the petroleum technology. Paper 1 presented a fractionation method for crude oil and the advantage of spectroscopic methods for analysis of crude oils and crude oil fractions. Paper 2 showed the sophisticated phenomena of particle (colloids) stabilized interfaces and the dependence of the functionality of the particle surfaces for stability efficiency and phase inversion. Paper 3 and 4 converged in a study of concentrated crude oil-water interfaces. The oscillating pendant drop technique provides important information with respect to the relaxation of interfaces. However, in the low frequency range, the main advantage with this method is the assessment of the characteristic time of mass transfer from the bulk and not the intrinsic properties of the interface: i.e. interfacial elasticity and interfacial viscosity. Relaxation of interfaces is important in droplet break-up and coalescence processes. A very important contribution to the understanding of the stability mechanism of w/o emulsions could be realized with dilational rheology in the mid- to high frequency range (1-1000 Hz). Such approach would provide both the characteristic adsorption time of immediate adsorption from concentrated systems and the intrinsic properties of the interfacial film. Paper 5 presented physicochemical properties of 30 crude oils. As expected, the covariance between several physicochemical properties was strong. In particular it was proposed that the destabilization of water-in-heavy crude oil emulsions in electrical fields was predominantly dependent on hydrodynamic effects of droplet kinematics and not the barrier to coalescence. Above a certain magnitude of the electrical field, disintegration of the interfaces of colliding droplets may reduce the relative influence of the coalescence step compared to destabilization with gravitational forces. I strongly recommend continuing the work within the sample set presented in Paper 5. Especially droplet break-up mechanism and emulsion stability to gravitationally induced separation should be studied at lower mixing energies. The long-term stability of the water-in heavy oil emulsions may be assessed with rheological experiments.

60

Paper 1

Ind. Eng. Chem. Res. 2005, 44, 1349-1357

1349

Group-Type Analysis of Heavy Crude Oils Using Vibrational Spectroscopy in Combination with Multivariate Analysis
Andreas Hannisdal,* Pl V. Hemmingsen, and Johan Sjoblom 1
Ugelstad Laboratory, Department of Chemical Engineering, Norwegian University of Science and Technology (NTNU), N-7491 Trondheim, Norway

Twenty heavy and/or particle-rich crude oils have been quantitatively fractionated into saturates, aromatics, resins, and asphaltenes (SARA) by asphaltene precipitation in n-hexane and highperformance liquid chromatography (HPLC). The newly developed and fully automated HPLC method has a sample capacity corresponding to 0.6 g of heavy crude oil. The crude oils have been characterized by vibrational spectroscopy in the near-infrared (NIR) and infrared (IR) regions. Principal component analyses (PCA) of the data sets from IR and NIR were performed so that exploratory data analyses could be conducted. Partial least-squares (PLS) regression models were built for each SARA component from IR and NIR data to predict the amounts of SARA components. These models successfully fitted the experimental data from NIR analyses and showed good predictive ability for the crude oil composition. The regression models from IR data were not modeled properly for aromatics and asphaltenes but were modeled excellently for saturate and resin components. For SARA determination, NIR spectroscopy appears to be a favorable alternative to the more time-consuming fractionation method.
1. Introduction The overall character of the feedstocks entering refineries has changed to such an extent that the difference can be measured by a decrease of several points on the API (American Petroleum Institute) gravity scale.1 This change has and will continue to make new demands with respect to both improved process technology and understanding of the crude oil system. With the necessity of processing heavy oil, there has been the recognition that knowledge of the constituents of these higher-boiling feedstocks is of great importance. Indeed, the problems encountered in processing the heavier feedstocks can be equated to the chemical character and the amount of complex, higherboiling constituents.1 Heavy oil is a complex mixture of hydrocarbons containing a small amount of heteroatoms (N, O, and S). The classification of heavy oil is in some respect indistinct but has usually been restricted to the more viscous part of conventional petroleum, having an API gravity of less than 20. Because of its complex nature, it is not possible to determine its individual molecular constituents, and compositional studies are usually done by fractionation into predefined chemical families. The SARA group type fractionation separates the crude oil into the following classes: saturates, aromatics, resins, and asphaltenes. This fractionation method has found great utility in combination with highperformance liquid chromatography (HPLC).2-6 However, because this technique is time-consuming and requires expensive laboratory equipment, some attempts have been made recently to find alternative analytical options. Various applications of vibrational spectroscopy have been introduced with success.7-10 In this context, it is particularly noteworthy that the work done by Aske and co-workers11 has shown the ability of
* To whom correspondence should be addressed. Fax: +47 73 59 40 80. E-mail: [email protected].

near-infrared (NIR) and infrared (IR) spectroscopy in combination with partial least-squares regression to predict SARA components in lighter crude oils and condensates. In this study, we have extended the range of application from light crude oils and condensates to heavy and particle-rich crude oils. This has introduced additional aspects with respect to both sample fractionation and spectroscopic analysis. We have developed a fully automated preparative HPLC procedure with a sample capacity corresponding to 0.6 g of heavy crude oil. The large sample capacity makes it possible to perform further analyses on the individual SARA fractions. 2. Experimental Section 2.1. Materials. Crude oil samples at ambient temperature and pressure were received from exploration sites on the Norwegian Continental Shelf and sites located in Brazil, France, the South China Sea, the Atlantic Ocean, and the Gulf of Mexico. From these, 20 crude oils were chosen to form a sample set of oils differing in physical properties from conventional petroleum to heavy or particle-rich oil. Some samples contained a lot of water (up to 50 wt %), which had to be removed prior to analyses. Other samples with lower water content were kept in their original state because of very stable water-in-oil systems. To reduce the effect of oxidation, 1 L of each of the crude oils was stored in UV-protective containers under nitrogen atmosphere.12 The separation of crude oils into SARA fractions is well accepted and will not be presented in detail here. Speight1 provides a thorough introduction to the chemistry and structure of crude oils. In short, saturates are defined as the saturated hydrocarbons ranging from straight-chain paraffins to cycloparaffins (naphthenes), whereas the aromatic fraction includes those hydrocarbons containing one or more aromatic nuclei that might be substituted with naphthenes or paraffins. Asphaltenes are defined as the solubility class of crude oils that

10.1021/ie0401354 CCC: $30.25 2005 American Chemical Society Published on Web 01/19/2005

1350

Ind. Eng. Chem. Res., Vol. 44, No. 5, 2005

Figure 2. Scheme of the flow channel selection valves. Valve A controls the flow direction through columns (normal/backflush), and valve B includes or excludes the amino precolumn.

Figure 1. Schematic representation of the HPLC system with main modules.

precipitate in the presence of aliphatic solvents (here, n-hexane), whereas the resin fraction is defined as the fraction soluble in light alkanes but insoluble in liquid propane. Asphaltenes and resins are known as large, polar, polynuclear molecules consisting of condensed aromatic rings and heteroatoms such as sulfur, nitrogen, and oxygen. Asphaltenes and resins have gained increased interest with the knowledge of the large effect of these constituents in particular on the overall performance of heavy crude oils. 2.2. Group-Type Fractionation. (a) Precipitation of Asphaltenes. The precipitation yield of asphaltenes from a crude oil in liquid hydrocarbons has been shown to be dependent on both contact time and the degree of dilution.1 Precipitation is done in three parallel trials, each of 4 g of crude oil diluted in 160 mL of n-hexane, left for mixing overnight and filtered through a 0.45m membrane filter (Millipore HVLP). Asphaltene content is determined gravimetrically after solvent removal. (b) HPLC Method: Fractionation of Maltenes. The high-performance liquid chromatography (HPLC) system used in this study, shown schematically in Figure 1, was built from the following modules from Shimadzu Inc.: helium degassing unit (DGU-10B), subcontroller (FCV-130AL), preparative pump (LC-8A), automatic sample injector (SIL-10AP), two high-pressure flow channel selection valves (FCV-12AH), dualmode UV-vis detector (SPD-10), refractive index detector (RID-10A), fraction collector (FRC-10A), system controller (SCL-10A). The columns were purchased from Phenomenex: unbonded silica 15-m 21.2 250 mm, amino 10-m 21.2 50 mm. Dichloromethane (99.8%) and n-hexane (95%) were used as mobile phases. Preparative separations of organic samples favor the use of unmodified silica as the column packing material.13 However, experiments on a preparative silica column showed that dichloromethane did not have the necessary solvent strength to elute the most polar components in crude oils. Instead, an amino column, weaker in column strength, was used in the multidimensional column setup shown in Figure 2. The amino precolumn and the silica preparative column could be reloaded individually. The new setup reduced the time of analysis from 70 to 26 min. A sample of 5 mL of asphaltene-free crude oil in n-hexane (corresponding to 0.6 g of crude oil) is injected into the 20 mL/min flow (isocratic) of filtered and degassed n-hexane. Crude oil components travel through the columns and are retained mainly depending on

Figure 3. Typical chromatogram from preparative HPLC analysis. The solid line represents the RI signal, and the broken line represents the UV chromatogram at 254 nm. Peaks are cut for better visualization.

polarity. Saturates, having no retention on the columns, are collected from refractive index (RI) signals (3-5 min). After 15 min, all aromatics have left the amino precolumn, but some yellow-colored aromatics have not yet eluted from the silica column. These are collected from the preparative column by a dichloromethane backflush (<23 min), whereas the resin fraction is desorbed from the amino precolumn by a dichloromethane backflush 23-26 min after sample injection. This procedure provides the chromatogram in Figure 3 where the three fractions are well resolved. The SAR weight fractions are determined gravimetrically after controlled evaporation of solvents in N2 atmosphere. To evaluate experimental errors related to sample preparation and fractionation, the SARA procedure was repeated for five crude oils, representative of the sample set, several weeks after the original measurements. Because standard deviations are not additive, the experimental error cannot be estimated from the individual standard deviations of the five samples. However, the variances are additive, and average standard deviations (SDev) can be estimated by calculating the pooled average for all samples from the replicate variances

SDev )

j (yij - yi)2 n(J - 1)i)1 j)1

(1)

where j is the replicate number of sample i. (c) H2O Determination. The 20 crude oils and the non-asphaltenic HPLC samples, diluted in n-hexane, were analyzed for water by a Karl Fisher titration method (Metrohm 756 KF Coulometer). The coulometric

Ind. Eng. Chem. Res., Vol. 44, No. 5, 2005 1351

titration can be summarized by the following overall equation

H2O + I2 + [RNH]SO3CH3 + 2RN T [RNH]SO4CH3 + 2[RNH]I

The titration end point is indicated voltametrically, with a Pt electrode, by the presence of free iodine. 2.3. Vibrational Spectroscopic Measurements. (a) NIR Spectroscopy. As the intensity of the first overtones is generally an order of magnitude less than that of the fundamental vibrations, path lengths are usually much longer in the near-infrared (NIR) region. The advantages of these lower intensities include the fact that nonlinearities due to strong absorptions are less likely to occur. Analyses of asphaltene-rich heavy crude oils have still shown that special requirements in instrumental setup are necessary. To cope with considerable scattering and resulting nonlinearity in the low-frequency range of the NIR region, a diffuse reflectance/absorbance unit with an apparent path length of only 0.32 mm was used. A 0.16-mm-thick Teflon sheet with a 10-mm hole was placed on a glass plate with a rough copper plate background. Crude oils were heated to 60 C to obtain representative samples and transferred to the glass plate to cool to room temperature before analysis. A backscatter probe with a 5-mm-outer-diameter sapphire window was mounted onto the Teflon plate to ensure a constant apparent path length of 0.32 mm through the crude oil sample. The 20 crude oils were analyzed in the wavenumber range 10000-4000 cm-1 (1000-2500 nm) with 64 scans and a spectral resolution of 2 cm-1, using this diffuse reflectance/absorbance accessory connected to a Bruker MPA FT-NIR spectrometer. The resultant NIR spectra were Fourier transformed with a Blackman-Harris three-term apodization function and a zero-filling factor of 2. All data acquisitions and Fourier transformations were performed using the OPUS software, version 4.2 (Bruker Optics). (b) IR Spectroscopy. The IR spectra were recorded in the range between 600 and 4000 cm-1 with a Tensor 27 FT-IR spectrometer (Bruker Optics) and a MKII Golden Gate ATR unit (Specac) in single reflection configuration. The instrument is fitted with a Michelson interferometer and a N2-cooled MCT detector (Mercury Cadmium Telluride). One drop of the essential oil was placed onto the diamond ATR crystal (only 4 mm2) and analyzed. To improve the signal-to-noise ratio, 256 scans were recorded at a resolution of 4 cm-1 and averaged. The resultant IR spectra were Fourier transformed with a Blackman-Harris three-term apodization function and a zero-filling factor of 2. In addition to the spectra of original crude oil samples, we have access to spectra of individual SARA fractions (undiluted). Some of these will be applied to understand the individual contributions to the overall vibrational absorption of a crude oil. All data acquisitions and Fourier transformations were performed using the OPUS software, version 4.2 (Bruker Optics). (c) Chemometrics and Data Preprocessing. Because of the diversity of chemical substances both in a particular crude oil system and in each of the predefined SARA classes, the resulting IR and NIR spectra will most often show a large number of features that can be exploited to the most effect only with the help of chemometric analytical tools. It might still be necessary

to perform preprocessing of the data prior to multivariate analysis. There are a number of possible methods of data reduction and preprocessing. However, these transformations should only be performed with problemspecific justification. Sharp peaks characterize vibrational features in the IR. However, because the crude oil system is very complex with hundreds of different chemical structures, some overlap of adjacent bands will occur. Minor baseline shifts are also observed. To eliminate these effects, first-order derivatives of the FT-IR spectra were computed using the Savitzky-Golay algorithm with sevenpoint smoothing, fitted to a second-order polynomial. Savitzky-Golay differentiation computes derivates using a polynomial approximation (by least-squares) of a given number of adjacent data points in a spectrum. Peaks in the NIR spectra are not nearly as distinct as those observed in the fingerprint region of the IR spectra. As discussed later, the specific crude oil set also shows typical baseline shifts and decreasing spectral intensity at lower frequencies (1100-1600 nm) from light scattering effects and electronic transitions within asphaltene aggregates, respectively.14 To maximize the contributions from vibrational absorption and minimize effects caused by physical phenomena, the second-order derivatives of the original FT-NIR spectra were computed using the Savitzky-Golay algorithm, fitted to a second-order polynomial with a 21-point window. When performing multivariate analysis on a set of data from spectroscopic measurements, it is common practice to use all variables available. The belief is that the statistical method used [such as partial leastsquares (PLS) and principal component analysis (PCA)] will extract from the data those variables that are most important and discard irrelevant information. Statistical theory shows that this is incorrect. In particular, the principle of parsimony15 states that a simple model (one with fewer variables) will tend to be better at predicting a new, previously unseen data set.16 Variable reductions were done on the basis of the jack-knifing ranking system developed by Martens and explained thoroughly by Martens and co-workers.17 In short, the jack-knifing principle used for model variables identifies those variables that appear to give particularly unstable model parameter estimates or that show little or no predictive ability. (d) Calibration and Prediction. Development of appropriate chemometric models was carried out with Unscrambler, version 8.0, software from CAMO. Principal component analyses (PCA) were performed according to the NIPALS18 algorithm so that exploratory data analyses could be conducted. PCA is a multivariate statistical technique whereby the information carried by the original variables (in this case, spectral frequencies) is projected onto a smaller number of uncorrelated variables called principal components (PCs).18,19 By plotting the principal components, one can view the interrelationships between different variables and detect and interpret sample patterns, similarities, or differences. Partial least-squares (PLS) regressions18,20 were performed on each of the SARA classes with predictive purposes. PLS is a method for relating the variations in one response variable (y variable, in this case, mass percent of a solubility class) to the variations of several predictors (x variables, NIR or IR frequencies). (e) Validation and Error Estimates. Full crossvalidation represents an efficient way of utilizing a set

1352

Ind. Eng. Chem. Res., Vol. 44, No. 5, 2005

Table 1. Experimental Compositions of Crude Oils Separated into SARA and Water Fractions crude oil no. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 5 8 11 15 18 SDev
a

origin Brazil Brazil Brazil China Mexico North Sea West Africa Brazil North Sea North Sea North Sea North Sea China West Africa Brazil Brazil North Sea Brazil North Sea France Mexico Brazil North Sea Brazil Brazil

saturates (wt %) 33.5 38.8 30.5 36.0 43.0 26.5 46.9 33.6 45.0 45.7 43.6 33.0 32.0 53.1 33.4 37.3 41.5 26.0 44.3 23.1 43.3 34.9 44.4 34.8 26.1 1.2

aromatics (wt %) 46.5 36.6 43.2 30.8 29.7 38.8 37.9 38.1 32.9 38.6 32.9 42.8 32.5 30.9 38.6 42.6 33.3 41.1 26.3 47.5 29.5 36.2 30.2 41.2 40.7

resins (wt %)

asphaltenes (wt %) meana SDeva 1.2 1.9 5.0 3.1 1.2 2.8 1.7 12.9 1.2 0.8 0.6 3.9 4.3 1.0 4.6 3.8 0.2 10.2 0.2 5.4 1.2 12.3 0.6 4.6 10.6 0.2
b

water (wt %) 0.8 0.9 0.2 9.6 0.3 12.6 0.1 0.1 0.2 0.1 0.4 0.5 1.2 0.2 0.1 0.3 0.1 0.1 0.5 4.2 0.3 0.1 0.4 0.1 0.1

yield (wt %) 100.8 90.9 99.9 101.5 84.5 101.3 100.6 101.7 89.2 96.6 84.8 91.8 98.6 93.3 94.8 98.1 82.0 99.4 79.4 91.7 82.4 99.5 82.5 99.6 99.3

Calibration Set 18.8 12.6 20.9 22.1 10.4 20.5 14.0 16.8 9.8 11.5 7.2 11.6 28.5 8.2 18.2 14.1 7.0 21.9 8.1 11.5 Replicates 8.1 16.0 6.9 18.9 21.9

0.1 0.1 0.3 0.1 0.1 0.1 0.0 0.2 0.1 0.1 0.0 0.0 0.2 0.0 0.2 0.4 0.0 0.3 0.0 0.1 0.1 0.2 0.0 0.2 0.3 0.02-0.6c

Average Experimental Errorb 0.8 0.8

Asphaltene mean values and standard deviations are estimated from three parallel precipitations. Average standard deviations are calculated from the variance of replicates and the corresponding samples in the calibration set (eq 1). c Heteroscedastic error distribution. Increasing uncertainties with increasing water concentration.

of rather few samples, as the same samples are used for both model calibration and validation. A sample is left out from the calibration set and predicted with the calibration model built from the remaining samples. The process is repeated for each sample until all samples have been left out once. Root-mean-square error of prediction (RMSEP) is defined as the square root of the average of squared differences between predicted (y) and measured (y) response values of the validation objects.18 RMSEP can be interpreted as the average prediction error, expressed in the same unit as the original response values. Rootmean-square error of calibration (RMSEC) is the corresponding measure for the model fit, calculated from the calibration objects only.

RMSEP )

(yi - yi)2 ni)1

(2)

Because the RMSEP represents a fair estimate of the average prediction error, it is natural to compare this error with the experimental standard deviation. Replicate spectroscopic measurements were performed on five crude oils for visual inspection of replicate variance compared to intersample variance. 3. Results and Discussions 3.1. Crude Oil Composition. Table 1 reports the separation results of crude oils into SARA and water fractions. Some crude oil samples (nos. 4, 6, 20) are high in water content. However, water determination of the filtrates from asphaltene precipitation showed that the

water content was reduced to less than 0.02 wt % (of the crude oil). Water is mainly removed during filtration, also supported by the observation of some water droplets on the filter paper. Asphaltene contents are reported with mean values and standard deviations of three parallel filtrations. It is believed that the observed insufficient sample recovery from the lighter crude oils (low resin content) is attributed to evaporation losses during solvent removal. GC analyses performed by Radke and co-workers5 showed that losses of compounds in the carbon number range up to 14 were found predominantly from the saturate fraction and to some extent from the aromatic fraction. Saturate and aromatic mass fractions were normalized to a total yield of 100%. Cut points between HPLC fractions are chosen from detector signals and validated through systematic IR analyses of all SARA fractions in the sample set (nondiluted fractions). The fingerprint region of crude oil no. 1 fractions in Figure 4 supports our specific definition of the maltenes as the saturate fraction is free of aromatics and the aromatic fraction is free of polar N, S, or O compounds. Similar conclusions are drawn from spectra of the 19 remaining crude oils in the sample set. The average experimental errors based on five replicates are reported as average standard deviations. Average errors for saturates and aromatics are within reasonable limits. Determination of asphaltene content is associated with an average standard deviation of 0.2 wt %, which is also satisfactory. The individual standard deviation of each crude oil (three parallel trials) might,

Ind. Eng. Chem. Res., Vol. 44, No. 5, 2005 1353

Figure 4. IR spectra of SARA fractions from crude oil no. 1 are shown together with a spectrum of the original crude oil. The broken line is a theoretical spectrum of crude oil no.1 calculated from a linear combination (weighted with the experimental composition) of the individual SARA spectra.

Figure 5. NIR spectra of crude oil samples 1, 4, 6, 15, 17, and 18. Baseline shifts from light scattering, -* and n-* transitions (1100-1600 nm) from asphaltenes, and typical O-H combinations (1850-2100 nm) for water containing crude oils are identified. Diff(18) is the second-order-differentiated spectrum of crude oil no. 18 in arbitrary ordinate values.

however, indicate a slightly heteroscedastic error distribution (smaller deviation for low response values). The error of water determination will most certainly be heteroscedastic and increase with increasing amount of water to be determined. However, the maximum standard deviation of 0.6 wt % is acceptable. 3.2. NIR Spectroscopy. (a) Spectral Features. Even though the features in the NIR spectra are broad and highly overlapping, they contain some relevant vibrational features. The most prominent bands are in the region from 1695 to 1765 nm, which corresponds to the first overtone of C-H bands from the fundamental vibrations in the 3000-2800 cm-1 region and the CH3/ CH2 bending bands around 1400 cm-1 of the infrared region.10,9 Furthermore, the weak second overtone of C-H and the C-H combinations appear at 1152-1225 nm and in the region from 1360 to 1440 nm, respectively. Aromatic functionalities are apparent at 1143 nm (C-H stretch second overtone), 1417 nm, 1446 nm (C-H combinations, 1685 nm (C-H stretch first overtone), and 2140-2190 nm (C-H, CdC combinations). In addition to direct evidence of the vibrational structure, NIR spectroscopy also provides information about other properties of the system such as particle size and flocculation behavior through contributions from elastic scattering and electronic transitions. As seen from Figure 5, the baseline in the 1100-1600-nm range shows an exponential decay pattern for asphaltene-rich crude oils. This feature is taken into account by strong absorption into the visible region from electronic -* and n-* transitions of asphaltene aggregates.10,14 Scattering can also significantly contribute to light extinction in the NIR spectrum. Sizes of inorganic particles, wax, or emulsified water can be considerably larger than those of asphaltenes and result in scattering and baseline displacements such as those observed in Figure 5. The wavelength dependence of scattering depends on particle sizes. Chemometric tools in addition to proper pretreatment of NIR spectra are necessary to extract the relevant information from vibrational absorption. Second-order differentiation turned out to be the ideal pretreatment to enhance features from vibrational absorption. Figure 5 shows the preprocessed spectrum of crude oil no. 18. (b) Principal Component Analysis. To explore the potential of regression model development, principal

Figure 6. Score plot from principal component analysis of secondorder derived FT-NIR spectra. Replicates are included. Explained variance: PC1, 77%; PC2, 11%.

component analysis was performed on the second-orderdifferentiated NIR spectra. Replicates were included in the model to get a visual impression of the replicate variation in its relation to the overall variance. To avoid getting an over-optimistic interpretation of the model, replicates were cross-validated with the corresponding samples from the calibration set. From the score plot of the first two principal components in Figure 6, it can be seen that the replicate variation is quite small compared to the average intersample variation, indicating good reproducibility. The first two PCs account for 88% of the variation in NIR spectra, which is promising for PLS model development. With spectroscopic data, the loading plot of the first few PCs can be useful in interpreting which effect is being modeled. Principal component 1 explains carbon-hydrogen vibrational absorption (not shown), whereas the second and fourth PCs model crude oil water and asphaltene content. Figure 7 shows the loading plot of the second PC with typical absorption bands from water at 1400 and 1900 nm and the fourth PC modeled by electronic transitions of asphaltene aggregates. Crude oils 4 and 6 are positioned as potential outliers in the score plot but are kept and justified as extreme end members of the

1354

Ind. Eng. Chem. Res., Vol. 44, No. 5, 2005

Figure 7. Loading plot of the second and the fourth principal components from PCA of second-order-differentiated FT-NIR spectra.

calibration set, rather than outliers. These two samples are known to be high in water content. There are reasons to believe that a regression model built from second-order-differentiated NIR spectra will be able to handle these samples. (c) PLS Regression. As the major objective of multivariate analysis was prediction, individual PLS regression models were built for each SARA component. The model quality parameters are listed in Table 2 for the optimal number of principal components. The number of principal components that are significant for predicting the SARA values are determined by crossvalidation and monitored with RMSEC and RMSEP. As an additional PC is introduced, the RMSEC will generally decrease because the PCs are found in such a way that the residuals are minimized. However, improved model fit will not necessarily result in a model with enhanced predictive ability. If the introduced PC carries noise or spectral regions not related to the predicted property, the model would fail to predict new crude oils with optimal accuracy, indicated by increased RMSEP. The optimal number of components is selected from the point where RMSEP shows a local minimum or levels off. Much of the predictive ability of the PLS regression models is introduced already with the first principal component. Asphaltenes shows optimal regression with as few as two PCs, whereas the predictive ability of the resin fraction continues to improve until the regression model consists of six PCs. This seems plausible, as the spectral features from asphaltenes are clearer than from resins in the NIR results. As already mentioned, the spectral contribution from asphaltenes is mainly from the very typical electronic -* and n-* transitions of asphaltene aggregates. Spectral contributions from the various functional groups in resins will naturally be more diverse. The correlation of predicted to experimentally determined amounts of resins is presented in Figure 8. Both calibration and validation objects are included. From the RMSEP values in Table 2, with correlation ranges taken into account, the predictive abilities of the SARA components are quite similar (slightly poorer for saturates). Because the experimental error associated with asphaltene determination (SDev) ) 0.2 wt %, Table 1) is considerably lower than the RMSEC (0.72 wt %) and RMSEP (0.89 wt %), it is obvious that the regression introduces some model error. Remembering the typical features of asphaltenes in NIR spectra,

second-order differentiation is associated with the danger of losing some spectral information from these compounds. Therefore, a PLS regression was built from first-order-differentiated (Savitzky-Golay, 21 points) spectra. The quality parameters for the new asphaltene regression model are included in Table 2. Compared to the original regression, RMSEC and RMSEP are reduced by a factor of 2 with respect to the new model (to 0.38 and 0.44 wt %). The model from first-orderdifferentiated spectra uses only one principal component as the water-containing crude oils are removed as outliers. Overall, RMSEPs are comparable to the model fit reported as RMSECs. As expected, some additional errors are introduced during model development from experimental SARA results. However, these model errors are acceptable. Spectral regions selected for correlation model developments are specified in Table 2. Further variable reductions inside the reported ranges are performed on the basis of the jack-knifing principle. 3.3. IR Spectroscopy. (a) Spectral Features. The most useful information from IR spectra of crude oils is in the 3200-2700 and 1800-650 cm-1 regions. Spectral information within these regions has been used to build regression models. The fingerprint area of crude oil no. 1 is introduced in Figure 4. To trace the individual contributions from each SARA component, IR spectra of the components are included. Generally, for the saturate fraction, CH3 groups have characteristic C-H stretching bands at 2962 and 2872 cm-1, whereas CH2 groups have bands at 2924 and 2853 cm-1. The bending vibration of methyl appears at 1378 and 1450 cm-1 while the vibration of methylene appears at 1467 cm-1.21 All of these bands are seen for the saturate fraction of crude oil no. 1 in Figure 4. The band resulting from the methylene rocking vibration, in which all methylene groups rock in phase, appears near 721 cm-1 for straight-chain alkanes of seven or more C atoms and at some higher frequencies for shorter chains.21 Aromatic C-H stretching bands occur in the 31003000 cm-1 region, whereas the in-plane and out-of-plane bending bands appear in the 1300-1000 cm-1 and 900675 cm-1 regions, respectively.1,22 For the aromatic fraction in Figure 4, the group of peaks at about 870, 814, and 740 cm-1 are believed to be due to substituted aromatic ring structures. These latter bands are the most prominent and most informative for aromatics. In addition, the aromatic fraction of crude oil no. 1 shows the typical absorption at 1605 cm-1 from carbon-carbon stretching within the ring. As for the aromatic fraction of crude oil no. 1, the aromatic bands at 900-675 and 1605 cm-1 are evident for resins and asphaltenes, and the broad and strong absorption between 1330 and 1100 cm-1 has been assigned to sulfur-oxygen functions.1 From Figure 4, the aromaticity of the asphaltenes is significantly greater than that of the maltenes fraction. Both the resin and asphaltene fractions show absorption between 1020 and 1030 cm-1, which is typical for aromatic ethers and sulfoxide functionalities. The resin fraction typically contains acids and phenols and will be featured by O-H and CdO stretching vibrations at 3700-3100 cm-1 (not shown) and 1710-1680 cm-1, respectively. The CdO stretching vibration in aliphatic acids (naphthenic acids) appears at 1705 cm-1, as for crude no. 1 resins.

Ind. Eng. Chem. Res., Vol. 44, No. 5, 2005 1355

Table 2. Quality Parameters for the Prediction of Crude Oil Composition Using NIR Spectroscopy in Combination with PLS Regression group type saturates aromatics resins asphaltenes asphaltenesd prediction range (wt %) 25.9-57.3 30.1-53.1 7.0-28.5 0.2-12.9 0.2-12.9 PCsa 4 3 6 2 1 x-expl (%)b 98.7 93.7 99.5 94.3 99.8 y-expl (%)b 96.7 96.0 98.9 95.1 98.8 RMSEC (wt %)c 1.88 1.10 0.64 0.72 0.38 RMSEP (wt %)c 2.82 1.47 1.46 0.89 0.44 sample outliers 5, 10 20, 1 20 4, 6, 20, 13 spectral range (nm) C-H vibl bands + ArCH, ArCC e 1120-1400 1110-1152

a Number of principal components used in the PLS regression model. b Explained calibration variance in x (NIR spectra) and y (experimental values). c Root-mean-square error of calibration/prediction. d First-order-differentiated spectra. e Because the regression model uses a large number of spectral frequencies, these are not listed.

Figure 8. Correlation of resins from the PLS model of the NIR spectra to measured values. Stars represent samples that were used to calibrate the model, and open circles indicate unknown samples that were predicted with full cross-validation. The solid line is a linear least-squares regression calculated from the calibration set, and the broken line is the corresponding regression calculated from validation samples. RMSEP, 1.46 wt % with six PCs.

Figure 9. Score plot from principal component analysis of firstorder-differentiated FT-IR spectra. Replicates are included. Explained variance: PC1, 79%; PC2, 17%.

From the above discussion, there should not be any doubt about the value of a peak-by-peak analysis in the infrared region, especially for the separate SARA fractions. Spectra of original crude oils also seem to contain important information even though the signal-to-noise ratio is reduced for some spectral ranges compared to the spectra of individual SARA fractions. To explore the potential of quantitative determination and regression from IR spectroscopy, a principal component analysis of the first-order-differentiated set of crude oil spectra was performed. (b) Principal Component Analysis. The score plot in first and second principal components is shown in Figure 9. The crude oils with high water content are positioned as extreme values. These samples could be included in the PLS regression models, but they might result in problems as the crude oil components are expected to be modeled with the same frequencies as those frequencies where water exhibits strong absorption (3700-2800 cm-1, 1800-1500 cm-1). Samples 4, 6, and 20 were removed from the calibration sample set. Visual inspection of the score plot indicates that the reproducibility of the IR analysis is very good, as replicate variations are small compared to intersample variations. The explained variance in the first two PCs is also very high (96%). (c) PLS Regression. Although the distinctive bands of individual SARA fractions have been used for qualitative structural analysis, their additive contributions prevent the use of specific bands for accurate quantita-

tive determination of aromatics and asphaltenes from the crude oil data set. Aromatics and asphaltenes are more or less characterized by similar vibrational features. We have reasons to believe that this will cause problems for the regression. The aromatics will be modeled properly as long as the asphaltene content is low. In this case, the contribution from asphaltenes to the total aromatic absorption will be very small. However, for asphaltene-rich crude oils with strong aromatic vibrational absorption, the contribution is significant and will affect the model. This conclusion is supported by the observation that the PLS regression model of aromatics is not stable against samples with high asphaltene content. Aromatics are not modeled well until these samples are removed. In addition, IR analysis of separate asphaltene fractions showed that the vibrational characteristics of the 20 asphaltenes are different. This intersample variation will complicate the regression of asphaltenes from IR spectra. Despite the aromatic structure of resins, these are still modeled properly because of the typical acidic band at 1705 cm-1. The spectral range from 1550 to 1750 cm-1 was used for model development. The correlation of predicted to measured amounts of resins is plotted in Figure 10. Application of the regression model from the IR spectra resulted in better predictive ability (RMSEP ) 1.32 wt %, three PCs) than the model from the NIR spectra (RMSEP ) 1.46 wt %; six PCs) and required fewer principal components to accommodate the spectral features of the data set. Quality parameters are listed in Table 3. Saturates are modeled by the response from only six variables (wavenumbers) from typical carbon-hydrogen stretching and bending vibrations in the 2880-2960 and

1356

Ind. Eng. Chem. Res., Vol. 44, No. 5, 2005

Table 3. Quality Parameters for the Prediction of Crude Oil Composition Using FT-IR Spectroscopy in Combination with PLS Regression group type saturates resins
a

prediction range (wt %) 26.2-57.3 7.0-28.5

PCsa 3 3

x-expl (%)b 99.3 98.5

y-expl (%)b 98.2 96.6

RMSEC (wt %)c 1.29 1.06

RMSEP (wt %)c 1.84 1.32

sample outliers 3 3

spectral range (cm-1) 1382-1470 2880-2960 1550-1750

Number of principal components used in the PLS regression model. b Explained calibration variance in x (IR spectra) and y (experimental values). c Root-mean-square error of calibration/prediction.

Model development from IR data resulted in inadequate model fits for the aromatic and asphaltene contents. We expect the diversity of asphaltenes within the sample set, indicated by IR analyses of pure asphaltenes, to be the reason. Because aromatics are modeled by more or less the same spectral frequencies as asphaltenes, this crude oil fraction is also modeled badly. Saturates and resins are predicted excellently from IR data, with RMSEPs of 1.84 and 1.32 wt %, respectively. The saturate content is predicted with the absorption at only six spectral frequencies, all of them in the regions typical for saturates. Three principal components capture 99.3% of the total variance in the six wavenumbers and explain 98.2% of the total variance in saturate content within the data set. Acknowledgment
Figure 10. Correlation of resins from the PLS model of the IR spectra to measured values. Stars represent samples that were used to calibrate the model and open circles indicate unknown samples that were predicted with full cross-validation. The solid line is a linear least-squares regression calculated from the calibration set, and the broken line is the corresponding regression calculated from validation samples. RMSEP, 1.32 wt % with three PCs.

The technology program Flucha III financed by industry is gratefully acknowledged for financial support. The Norwegian Research Council (NFR) financed the HPLC and IR instruments. Literature Cited
(1) Speight, J. G. The Chemistry and Technology of Petroleum, 3rd ed.; Marcel Dekker: New York, 1998. (2) Ali, M. A.; Nofal, W. A. Application of High-Performance Liquid Chromatography for Hydrocarbon Group Type Analysis of Crude Oils. Fuel Sci. Technol. Int. 1994, 12, 21. (3) Fan, T. G.; Buckley, J. S. Rapid and accurate SARA analysis of medium gravity crude oils. Energy Fuels 2002, 16, 1571. (4) Hammami, A.; Ferworn, K. A.; Nighswander, J. A.; Overa, S.; Stange, E. Asphaltenic crude oil characterization: An experimental investigation of the effect of resins on the stability of asphaltenes. Pet. Sci. Technol. 1998, 16, 227. (5) Radke, M.; Willsch, H.; Welte, D. H. Preparative Hydrocarbon Group Type Determination by Automated Medium-Pressure Liquid Chromatography. Anal. Chem. 1980, 52, 406. (6) Suatoni, J. C.; Swab, R. E. Preparative Hydrocarbon Compound Type Analysis by High-Performance Liquid Chromatography. J. Chromatogr. Sci. 1976, 14, 535. (7) Chung, H.; Choi, H. J.; Ku, M. S. Rapid identification of petroleum products by near-infrared spectroscopy. Bull. Korean Chem. Soc. 1999, 20, 1021. (8) Blanco, M.; Maspoch, S.; Villarroya, I.; Peralta, X.; Gonzalez, J. M.; Torres, J. Determination of physico-chemical parameters for bitumens using near infrared spectroscopy. Anal. Chim. Acta 2001, 434, 133. (9) Chung, H.; Ku, M. S.; Lee, J. S. Comparison of near-infrared and mid-infrared spectroscopy for the determination of distillation property of kerosene. Vib. Spectrosc. 1999, 20, 155. (10) Chung, H.; Ku, M. S. Comparison of near-infrared, infrared, and Raman spectroscopy for the analysis of heavy petroleum products. Appl. Spectrosc. 2000, 54, 239. (11) Aske, N.; Kallevik, H.; Sjoblom, J. Determination of Saturate, Aromatic, Resin, and Asphaltenic (SARA) Components in Crude Oils by Means of Infrared and Near-Infrared Spectroscopy. Energy Fuels 2001, 15, 1304. (12) Boukir, A. Subfractionation, characterization and photooxidation of crude oil resins. Chemosphere 2001, 43, 279. (13) Snyder, L. R.; Kirkland, J. J.; Glajch, J. L. Practical HPLC Method Development, 2nd ed.; John Wiley and Sons: New York, 1997.

1382-1470 cm-1 regions. This shows that IR spectra of crude oils contain considerable information about saturates concentrated in a small number of wavenumbers. Three principal components capture 99.3% of the total variance in the six wavenumbers and explain 98.2% of the total variance in saturate content within the data set. Both saturate and resin PLS models separate sample no. 3 from the regression as an outlier. We have no indications that this sample is substantially different from the other samples in the sample set. Reproductions of the IR analysis and SARA fractionation should indicate whether data for this sample are affected by experimental errors. 4. Conclusions From the reported study, near-infrared spectroscopy appears to be a favorable alternative to the more timeconsuming fractionation method for the determination of heavy crude oil composition. A previous study performed by Aske and co-workers11 concluded that NIR spectroscopy performed well for similar models of lighter crude oils and condensates. With this report, NIR spectroscopy has proven to perform well in a broad range for the prediction of SARA components. In this study, the predictive ability of the regression models are reported as RMSEP values of 2.82 wt % (S), 1.47 wt % (A), 1.46 wt % (R), and 0.44 wt % (A). These uncertainties more or less agree with the corresponding calibration uncertainties reported as RMSEC values and the experimental uncertainties of the reference method (HPLC) reported as average standard deviations.

Ind. Eng. Chem. Res., Vol. 44, No. 5, 2005 1357

(14) Mullins, O. C. Asphaltenes in Crude Oil: Absorbers and/ or Scatterers in the Near-Infrared Region? Anal. Chem. 1990, 62, 508. (15) Seasholtz, M. B.; Kowalski, B. The parsimony principle applied to multivariate calibration. Anal. Chim. Acta 1993, 277, 165. (16) Shaw, A. D.; Winson, M. K.; Woodward, A. M.; McGovern, A. C.; Davey, H. M.; Kaderbhai, N.; Broadhurst, D.; Gilbert, R. J.; Taylor, J.; Timmins, E. M.; Goodacre, R.; Kell, D. B.; Alsberg, B. K.; Rowland, J. J. Rapid Analysis of High-Dimensional Bioprocesses Using Multivariate Spectroscopies and Advanced Chemometrics. Adv. Biochem. Eng./Biotechnol. 2000, 66. (17) Martens, H.; Hoy, M.; Westad, F.; Folkenberg, D.; Martens, M. Analysis of designed experiments by stabilised PLS Regression and jack-knifing. Chemom. Intell. Lab. Syst. 2001, 58, 151. (18) Esbesen, K. H. Multivariate Data AnalysessIn Practice, 5th ed.; CAMO, Aalborg University: Esbjerg, Denmark, 2001. (19) Wold, S.; Esbensen, K.; Geladi, P. Principal Component Analysis. Chemom. Intell. Lab. Syst. 1987, 2, 37. (20) Geladi, P.; Kowalski, R. B. Partial Least Squares Regression: A Tutorial. Anal. Chem. 1986, 58, 1. (21) Silverstein, R. M. Spectroscopic Identification of Organic Compounds, 6th ed.; John Wiley & Sons: New York, 1998. (22) Sastry, M. I. S.; Chopra, A.; Sarpal, A. S.; Jain, S. K.; Srivastava, S. P.; Bhatnagar, A. K. Determination of physicochemical properties and carbon-type analysis of base oils using mid-IR spectroscopy and partial least-squares regression analysis. Energy Fuels 1998, 12, 304.

Received for review April 29, 2004 Revised manuscript received November 15, 2004 Accepted November 24, 2004 IE0401354

Paper 2

Colloids and Surfaces A: Physicochem. Eng. Aspects 276 (2006) 4558

Particle-stabilized emulsions: Effect of heavy crude oil components pre-adsorbed onto stabilizing solids
Andreas Hannisdal , Marit-Helen Ese, P l V. Hemmingsen, Johan Sj blom a o
Ugelstad Laboratory, Department of Chemical Engineering, Norwegian University of Science and Technology (NTNU), N-7491 Trondheim, Norway Received 5 February 2005; received in revised form 5 October 2005; accepted 8 October 2005 Available online 21 November 2005

Abstract Model emulsions stabilized by means of silica nanoparticles have been investigated. The effect of the modication of the particle surface has been monitored using infrared (IR) spectroscopy, contact angle measurements and zeta potential measurements. It is shown how coating of the nanoparticles with asphaltenes and resins will modify the stabilization prole compared to that obtained with no coating. The stabilization efciency was greatly enhanced by adsorption of crude oil components onto very hydrophilic or very hydrophobic silica. Possible stabilization mechanisms have been discussed. We have demonstrated catastrophic phase inversion of emulsions stabilized by particles with intermediate wetting properties, induced by simply increasing the volume of the disperse phase. In all cases, the stability to gravitational induced separation (coalescence) passes through a minimum approaching inversion in line with a maximum in drop size of the disperse phase. Transitional phase inversion from o/w to w/o emulsion type can be achieved by modifying of the hydrophilicity of the particles, either by silylation or by a controlled coating with heavy components from crude oil. The relevance of this study for real petroleum systems is discussed. 2005 Elsevier B.V. All rights reserved.
Keywords: Adsorption; QCM-D; Asphaltenes; Resins; Silica particles; Wettability; Emulsion stability; Inversion

1. Introduction Every day the complex petroleum industry faces the challenge of resolving emulsions. The process of recovering crude oil from reservoirs to production facilities requires techniques that can introduce pressure drops and large shear energies, resulting in increased interfacial area between the oil and water phases. At these interfaces, inorganic colloids will accompany organic particles and interfacially active molecules from the crude and build up mechanically strong, viscoelastic lms. The build up of such a protective layer will restrain the process of droplet coalescence, and reduce the rate of emulsion destabilization. While previous studies have focused mainly on the stabilizing effect of asphaltenes and asphaltene complexes, it has been shown that emulsions containing inorganic particles can be even more stable than those stabilized by asphaltenes only [14]. With the acknowledgement of the important role of solid particles in stabilizing petroleum emulsions, it has been done a systematic

Corresponding author. Fax: +47 73 59 40 80. E-mail address: [email protected] (A. Hannisdal).

approach in mapping the important parameters for good stabilization efciency in well-dened particle-stabilized systems. In particular, silica particles have been studied extensively because their properties resemble those of natural inorganic colloids like clays. Aveyard et al. [5] have reviewed the studies, which have contributed to the understanding of solid-stabilized emulsions. These studies have allowed for some general conclusions to be drawn on the important parameters for the effectiveness of particles in stabilizing emulsions. Factors that have to be considered are particle-size and shape [6], concentration [712], wettability of the particles, pH and electrolyte concentration in the aqueous phase, and other inter-particle interactions. We will address some of these parameters with results from literature and leave other parameters with only a reference. As concluded by Aveyard et al. [5] and Binks [13], many of the properties of solids in stabilizing emulsion interfaces can be attributed to the very large free energy of adsorption for particles of intermediate wettability (partially wetted by both oil and water phases). This effectively irreversible adsorption leads to extreme stability for certain emulsions and is in contrast to the behavior of surfactant molecules which are usually in rapid dynamic equilibrium between the oil/water interface

0927-7757/$ see front matter 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.colsurfa.2005.10.011

46

A. Hannisdal et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 276 (2006) 4558

and the bulk phases. Since Finkle et al. [14] in 1923 discovered the correlation between the wettability of particles and their ability to stabilize emulsions there has been a large number of studies dealing with wettability alteration and the effect on emulsion stability [7,11,12,1523]. Binks and Lumsdon [20] did an investigation into the inuence of the wettability of spherical, nanometer-sized silica particles on the type and stability of water-toluene emulsions. The silica surface was silylated to various extents by reaction with dimethyldichlorosilane, thus reducing the number of silanol (SiOH) groups at the particle surface end making the hydrophilic silica more hydrophobic. As expected, emulsions stabilized by either very hydrophilic or very hydrophobic particles were unstable to coalescence whereas particles with intermediate hydrophobicity, in which particles were held strongly at interfaces, were stable to coalescence indenitely. In addition, this study demonstrated a type of inversion of emulsions, known as transitional phase inversion, brought about by changing the silanol content of the particles. Just as the water or oil-liking tendency of a surfactant is quantied in terms of the HLB number, so can that of spherical particles be described in terms of its wettability via contact angle measurements [13]. By increasing the hydrophobicity of the particles at xed volume fraction of oil and water, it was possible to invert emulsions from o/w to w/o type. Binks and Lumsdon have also showed that transitional phase inversion is possible in emulsion systems of two types of silica particles, one hydrophilic and one hydrophobic [22]. Tambe and Sharma [12] have demonstrated transitional phase inversion in systems of calcium carbonate, barium sulfate, bentonite and silica modied to different extent by addition of stearic acid. Another unique feature of particle-stabilized emulsions is the possibility of inverting a system stabilized by only one type of particles simply by changing the oil to water ratio [5]. Binks et al. have reported this type of inversion, known as catastrophic phase inversion, for silica systems [7,9,20,24]. Catastrophic phase inversion took place at a particular volume fraction of the dispersed phase depending on the hydrophobicity of the particles and in which phase the particles were initially dispersed [20,24]. Kralchevsky et al. [23] have recently done a theoretical study on the thermodynamics of particle-stabilized emulsions. They have predicted the existence of catastrophic phase inversion in particle-stabilized emulsions from the work of formation (w()) of the normal and reverse emulsion type. These results will be discussed later in comparison to results from this study. Early work led to the formulation of a basic stabilization mechanism, involving steric hindrance to dropletdroplet coalescence provided by interfacially adsorbed colloids (closely packed) extending into the continuous phase of the emulsion. The importance of particles to change the rheology of the interface was also recognized. More recent studies have conrmed the signicance of steric effects and inter-particle interactions and also proven the mechanism of particle stabilization to be very sophisticated. Kralchevsky et al. [23] have explained the stability of particle-stabilized emulsions from the interfacial bending energy ( Wb ()) and dilatation energy ( Wdil ) for both non-closely packed and closely packed particle monolayers. In the case of non-closely packed particle monolayer, the

bending energy leads to an effective attraction between two colliding emulsion drops, which are, however, unable to occulate because of the predominant repulsive effect of the interfacial dilatation. For closely packed particles and < 90 (measured into the continuous phase), both Wb and Wdil are positive and will lead to repulsion between two colliding drops, thus stabilizing the emulsion. In contrast, for > 90 , energy considerations predicted the appearance of spontaneous occulation. The so-called particle bridging effect were discussed. In the contact region between two emulsion drops, it may happen that one particle is adsorbed simultaneously at the two lm interfaces. Depending of the wetting characteristic of the particles the phenomena can lead to stable occulates or coalescence. Bridging of uid interfaces by particles has been observed directly by Vignati et al. [25] and by Stancik et al. [26,27]. In agreement with free energy considerations, bridged particles were able to generate signicant adhesion between the two uid interfaces, which should result in additional emulsion stability in real systems. Friberg [28] has calculated the stabilizing properties of bi-layers of particles. Whereas earlier studies claimed that the formation of a close-packed arrangement of silica particles at the interface was necessary in order to obtain stable emulsions, Midmore [29] showed that very stable emulsions could be formed with only 29% coverage of the interface with silica spheres. Later, Vignati et al. [25] studied oil-in-water emulsions stabilized by monodisperse, silica colloids with a uorescent core. With this clever approach they could keep track of the location of the colloids. Interestingly, they were able to make stable emulsions even with very low droplet surface coverage (5%). No straightforward relation was found between the degree of droplet surface coverage and macroscopic emulsion stability. Vignati et al. also discussed the role of surface dynamics on the stabilization of poorly covered droplets as they observed diffusion constants of the particles at the oil/water interface, comparable to that in the bulk. Tarimala et al. [30] have also observed surprisingly fast Brownian motion of solid particles at the two-dimensional oil/water interface. According to the steric stabilization mechanism, coalescence requires the solid particles to be removed from the dropdrop contact region. Free energy considerations suggest that lateral displacement of the particles is most likely, since forcing droplets into either phases from the interface require extreme energies [12,31,32]. Tambe and Sharma [33,34] early showed how particles could modify the rheological properties of the interfacial region. At sufcient high surface coverage, colloidladen interfaces exhibited viscoelastic properties. They stated that the viscoelastic interface affected emulsion stability both by retarding the rate of lm drainage between coalescing emulsion droplets and by increasing the energy required to displace particles from the contact region between droplets. Recently, these interfacial dynamic features have been observed directly when two colloidladen interfaces are brought into contact with each other under a microscope. Beautiful pictures representing lm drainage, the Marangoni effect, trapped particles and particle morphology have been published [26,27,30,35]. If charged, particles can also impart a degree of electrostatic repulsion, which may affect the approach of particle-laden uid

A. Hannisdal et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 276 (2006) 4558

47

interfaces and the occulation-tendency of particles both in the bulk and at the interface [6,12]. The charge on the particles and their extent of occulation can be modied by pH control and addition of electrolytes to the aqueous phase of the emulsion [6,8,10,11,19,31,36]. If silica particles assemble to form a network extending through the coherent phase, bulk viscosity will increase and add stability to the emulsion as the rate of occulation and lm thinning is reduced [7,3739]. At different stages prior to, and during crude oil production, the inorganic material from reservoirs will be exposed to asphaltenes and resins. These components may adsorb onto initially hydrophilic surfaces, primary through their polar functional groups, and inuence the wettability and charge of the particles and their afnity for the oil/water interface [4042]. Yan and Masliyah have studied the partitioning of clay particles with different wettability at the oil/water interface [18,43]. They found a minimum in emulsion droplet size for asphaltene-treated particles with intermediate wettability, indicating that these particles were the most effective stabilizers for the oil-in-water emulsions. It was also demonstrated how the emulsions could be destabilized by the addition of fresh oil [44] or by adjusting the pH of the aqueous phase [19]. Sullivan and Kilpatrick [1] studied the effect of size, concentration, and type of inorganic solid particles on crude oil emulsion stability. They showed that hydrophilic particles, protected from asphaltene adsorption by a preadsorbed water lm, inverted crude oil emulsions from w/o to o/w whereas the same particles added stability to w/o emulsions when dried and added to the asphaltene containing oil phase. Similarly, Gu et al. [11] tuned emulsion type and stability of a toluene-diluted bitumen/water system by adding hydrophilic clay to the water phase (variable pH). Adsorption of asphaltenes and resins onto mineral surfaces has been studied extensively [4554]. In general, Langmuir type adsorption isotherms are reported for asphaltene adsorption from toluene solution at low bulk concentrations with multiple steps or other deviations from the simple model at higher concentrations. This has been explained by surface phase reorientation, multilayer formation or aggregation. Acevedo et al. [45,47,48] have reported formation of asphaltene multi-layers or aggregates at silica surfaces from concentrated asphaltene/toluene solutions. The tendency towards multilayer formation of asphaltenes has been related to the history of deposition problems in their original oils. It has also been shown that desorption of asphaltenes from a silica surface at atmospheric condition is slow enough to be neglected [45]. Whereas the typical adsorption studies in literature are done in combination with spectroscopic techniques, Ekholm et al. [55] used the dissipative quartz crystal microbalance technique (QCM-D) to study adsorption of asphaltenes and resins onto hydrophilic gold surfaces. They found that asphaltenes form a rigid layer on the surface at moderate concentrations. The adsorption at increased concentrations was more extensive than that observed for non-associating polymers, indicating aggregate adsorption. Resins were found to form a compact monolayer when adsorbed from a heptane solution. However, the amount of resins adsorbed was signicantly reduced with increasing toluene content of the solvent. Desorption studies showed that resins did not desorb the already adsorbed

asphaltenes, neither did they adsorb onto the asphaltene-covered surface. In this work, we have studied stabilizing properties of four different types of commercially available silica nanoparticles. Two of these products are hydrophobic; one is extremely hydrophilic, and one is expected to be wet to an intermediate extent by both oil and water. These dry particles have been modied with asphaltenes and resins and the effect on emulsion stability, type and droplet size of the emulsion has been investigated. We will relate the performance of these particles in model systems to their surface properties as monitored by vibrational spectroscopy, contact angle measurements and electrophoretic mobility. To address some important questions related to the adsorption of asphaltenes and resins onto silica, we have performed a preliminary adsorption study using the QCM-D technique with a silica surface. This study is aimed at developing a better understanding of how indigenous inorganic particles interact with heavy crude oil compounds to form stable emulsions. 2. Experimental 2.1. Extraction of asphaltenes and resins from crude oil Asphaltenes were precipitated from a Brazilian heavy crude oil (1213 American Petroleum Institute gravity) in a 1:40 excess of n-hexane and ltrated two times through a 0.45 m Millipore lter. Asphaltenes were washed carefully with nhexane to remove any resins from the asphaltene precipitate. The crude oil has been analyzed with respect to SARA components, density, viscosity, acidity and water content. Table 1 lists characteristics of the crude oil. Details about the SARA fractionation are reported elsewhere [56]. In short, asphaltenes were precipitated in n-hexane (1:40) and the remaining maltenes were separated using amino- and silica chromatography in a fully automated HPLC method with a sample capacity corresponding to 0.6 g of heavy crude oil. The SARA weight fractions were determined gravimetrically after solvent removal and analyzed with IR spectroscopy. In order to extract resins, maltenes from 250 ml oil were shaken with 2.5 kg of silica (4063 m, MERCK) for 2 days on a shaker (IKA HS 501) until the solvent was colorless, indicating that resins and aromatics had adsorbed
Table 1 Characterization of Brazilian crude oil Parameter SARA fractionation Saturates (wt.%) Aromatics (wt.%) Resins (wt.%) Asphaltenes (wt.%) Emulsied watera (wt.%) Density (20 C)b Viscosity (40 C) Acidity (TAN)
a b

26.2 41.5 21.9 10.2 0.1 1213 API 0.98 g/cm3 2497 cP 3.73 mg KOH/mg

KarlFisher titration. AP PAAR density meter DMA 48.

48

A. Hannisdal et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 276 (2006) 4558

Fig. 1. Fingerprint region of IR. The solid line shows the vibrational features of the extracted resins while the broken line represents the resins from the semipreparative HPLC fractionation.

onto silica. The silica slurry was transferred to a glass column (8 cm diameter) and allowed to settle. Toluene was used as mobile phase to elute aromatics. We collected samples from the chromatographic process and analyzed for aromaticity with UV and for typical vibrational features from resins with IR spectroscopy. Once the aromatics were removed, resins were eluted with a more polar solvent (40:30:30 acetone:toluene:methylene dichloride) as proposed by Spiecker and co-workers [57]. The resin-solvent mixture was ltrated through a 0.45 m Millipore lter to remove any silica nes and rotary evaporated until dry. Resins were diluted in toluene, transferred to a non-transparent ask and dried at 60 C under nitrogen until completely dry. A total of 50 g of resins were extracted from the crude oil. IR spectroscopy was performed in order to compare the extracted resins with the resins from the automated semi-preparative method. Spectra are shown in Fig. 1. 2.2. Coating of silica particles Aerosil particles from Degussa are fumed silica, a synthetic amorphous silicon dioxide manufactured by continuous ame hydrolysis of silicon tetrachloride. Siloxane and silanol groups can be found on the surface of unmodied Aerosil particles, where silanol groups account for the hydrophilic behavior. Reactions of the silanol groups with various organosilanes and siloxanes can chemically modify the surface of the partiTable 2 Chemistry and specication of silica particles Product id Aerosil Aerosil Aerosil Aerosil
a

cles. We used four different products in this study: Aerosil 200, Aerosil 7200, Aerosil 202 and Aerosil 972. Aerosil 200 is an unmodied silica while Aerosil 7200 is a structure modied (3-methacryl-oxypropyl-trimethoxysilane) product based on Aerosil 200. The hydrophobic silica products, Aerosil 202 and Aerosil 972 are fumed silica after-treated with polydimethylsiloxane and dimethyldichlorosilane, respectively. Table 2 summarizes the properties of the four silica products. Note that the fumed silica products do not appear as individual particles in aqueous solution, but occulate to aggregates or agglomerates, which can be several hundred nanometers in size. This common feature of fumed silica is due to siloxane (Si O Si) linkages between primary particles. The difference between fused silica (Aerosil) and silica produced by alkaline hydrolysis of sodium silicate solutions via nucleation and growth is discussed by Binks and Lumsdon [8]. Particles were coated with extracted resin- and asphaltene-fractions from toluene in 24 h batch adsorption experiments. The particles were coated using 1 wt.% asphaltene, 1 wt.% resin and 0.5 wt.% asphaltene + 0.5 wt.% resin mixtures. The mixtures were sonicated for 10 min before particles were added (25 g particles/L). Coated particles were centrifuged and washed thoroughly with toluene to remove any asphaltenes or resins not adsorbed onto the particles. 2.3. Wetting characteristics of nescontact angle measurement When working with particle-stabilized emulsions it is of great importance to be able to estimate the three phase contact angle () the particles makes with the interface as this property determine both the type of emulsion formed, and the degree of steric stabilization of two approaching interfaces. However, for sub-micron sized particles indirect methods have to be used for determining the partitioning of the solids at the oil/water interface. Some of these techniques are reviewed by Binks [13]. We have measured the three phase contact angle of both pure and coated solids with the compressed disk method. We report contact angles, measured into the water phase, of disks in air and immersed in oil. Pictures were taken continuously with a CAM 200 Optical Contact Angle Meter from KSV Instruments (Helsinki, Finland) to record contact angles before any penetration of water into the pellet had started. This is especially important when analyzing hydrophilic surfaces in air. The reported contact angle is the mean value of left and right contact angles with replicates included. We also performed some qualitative powder immersion experiments [20], where the particles

After-treated with Hydrophilic Hydrophilic Hydrophobic Hydrophobic 3-Methacryl-oxypropyl-trimethoxysilane Polydimethyl-siloxane Dimethyl-dichlorosilane

Specic surface area (BET) (m2 /g) 200 25 150 25 100 20 110 20

Tapped density (g/l)a 50 230 60 50

Primary particle size (nm) 12 14 16

200 7200 202 972

Volume of product is measured after a standard vibration condition is applied (ISO 787/11).

A. Hannisdal et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 276 (2006) 4558

49

were placed on a water/air interface. We then carefully replaced the air phase with oil and monitored the partitioning of the particles (visually, brown particles) after 24 h. Some particles had entered the water phase, some were adsorbed at the interface whereas some could easily be suspended in the oil phase with gentle stirring in the oil phase. The concentration of particles in each phase was not quantied. 2.4. FT-IR spectroscopy of coated particles Infrared (IR) spectra of the pure and coated silica products were recorded in the range between 600 and 4000 cm1 with a Tensor 27 FT-IR spectrometer (Bruker Optics) equipped with a MKII Golden Gate diamond ATR unit (Specac) in single reection conguration. The instrument is tted with a Michelson interferometer and a N2 -cooled MCT detector (Mercury Cadmium Telluride). To improve the signal-to-noise ratio 32 scans were recorded at a resolution of 1 cm1 and averaged. All the spectra were normalized with respect to the absorption band at 10501080 cm1 , assigned to Si O Si bonds, which form the basic network in the silica products [58]. The spectra are dominated by these and will only show weak evidence of crude oil components at the silica surface. We have used the spectral range between 2800 and 3000 cm1 with characteristic absorption bands from C H stretching in methyl and methylene groups (2962 cm1 vas CH3 , 2872 cm1 vs CH3 , 2924 cm1 vas CH2 , 2853 cm1 vs CH2 ) [59] as a measure of hydrocarbons at the silica surface. 2.5. Electrophoretic measurements Zeta potentials () of the solid nes were determined as a function of pH on a Malvern 3000HS Zetasizer connected to a titrator (Malvern Instruments, UK). Particles were suspended in aqueous NaCl (102 M) solution and sonicated for 15 min. Large particle aggregates were allowed to settle and removed prior to the measurement. NaOH (0.5 M) and HCl (0.5 M) solutions were used to adjust pH. The electrophoretic mobility was measured for each pH level in the range from 2 to 10 by laser Doppler velocimetry and the zeta potentials were determined by the Smoluchowski approximation [60]. The zeta potential reported at each pH level is the average value of three measurements. Attempts to measure the electrophoretic mobility of oil-wet particles were not done as a high degree of occulation was expected. 2.6. Adsorption study-quartz crystal microbalance A QCM device consists of a thin quartz disc sandwiched between a pair of electrodes. Due to the piezoelectric properties of quartz, it is possible to excite the crystal to oscillation by applying an AC voltage across its electrodes. The resonant frequency (f) of the crystal depends on the total oscillating mass, including solvent coupled to the oscillation. When a lm is attached to the sensor crystal the frequency decreases. Provided that the lm is thin and rigid, the decrease in frequency is proportional to the mass of the lm and the QCM operates as a

very sensitive balance. The mass ( m) of the adhering layer is calculated using the Sauerbrey relation [55]. m= C f n (2)

where C is 17.7 ng Hz1 cm2 for a 5 MHz quartz crystal and n = 1, 3, 5, 7 is the overtone number. In many situations the adsorbed lm is not rigid and the Sauerbrey relation becomes invalid. A lm that is soft (viscoelastic) will not fully couple to the oscillation of the crystal and the crystals oscillation will be dampened. The dissipation (D) of the crystals oscillation is a measure of the lms viscoelasticity. Dissipation is dened as: D= Elost 2Estored (3)

where Elost is the energy lost (dissipated) during one oscillation cycle and Estored is the total energy stored in the oscillator. The dissipation of the crystal is measured by suddenly disconnecting the oscillator and recording the response of a freely oscillating crystal that has been vibrating at its resonance frequency. By simultaneously measuring the frequency of oscillation and the energy dissipation at several overtone frequencies, we can evaluate the adsorbed amount, adsorption kinetics and rigidity of the adsorbed lm. The dissipative quartz crystal microbalance device from Q-Sense (Gothenburg, Sweden) has been used to study the adsorption behavior of asphaltenes and resins from toluene solutions. We have used a sensor crystal with a silica surface. The instrument has been described in detail by Rodahl et al. [61]. The third, fth and seventh overtones of the fundamental resonance frequency were used to study the adsorption processes. Crystals were treated with a Piranha solution (H2 SO4 :H2 O2 , 3:1), washed in milli-Q water and dried with N2 prior to the measurement. The silica surface was exposed to asphaltene and resin solutions (identical to those used to coat silica particles), followed by several rinses with toluene. 2.7. Emulsion stability measurements and drop size determination Measurements were done on model systems of water (milliQ, 3.5 wt.% NaCl) and oil (heptane:toluene, 70:30 v/v) always adding to a total volume of 12 ml. If not otherwise specied water and oil will refer to these solutions. To disperse possible aggregates, particles were sonicated for 5 min in either water or oil. The other phase was added to the system and dispersed with an Ultra Turrax (IKA, T18 with 10 mm head) rst at 18,000 rpm for 1 min, then at 22,000 rpm for 2 min. The samples were immediately transferred to graded centrifuge tubes. The emulsion stabilities are reported as the amount of dispersed phase resolved after centrifugation at 3000 rpm (1580 g) for up to 30 min. We explored the effect of: (1) Changing the initial location of the particles (dispersed in oil or water) (2) Varying the ratio of the two phases (2080 vol.% oil) (3) The amount of particles added to the system (0.0010.01 g particles/ml oil and water, later referred to as 0.11.0%

50

A. Hannisdal et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 276 (2006) 4558

(w/v). Notice that the amount of particles in the system is held constant when changing the oil to water ratio.) The stabilizing efciency of both pure and coated particles was investigated. We identied the type of emulsion formed (water or oil continuous) with a test similar to that explained schematically by Ese and Kilpatrick [62] where a drop of the emulsion was added to oil and water. Water continuous emulsions were easily dispersed in water (as small oil droplets) but remained as a big emulsion drop (o/w/o) when added to oil. The opposite was observed for oil continuous emulsions. For stable emulsion systems, the Sauter mean diameter, d32 , is reported to quantify the average size of dispersed droplets. A sample of the emulsion phase from a freshly prepared system was carefully diluted in its continuous phase at a ratio of 1/40. A Nikon Eclipse ME 600 digital video microscope (DVM) was used to acquire 50 pictures of the droplets in the continuous phase and the images were processed with Image Pro Plus 5.0 to give d32 and the volume distribution of drops. 3. Results and discussion 3.1. Adsorption study In order to address a few important issues related to the adsorption process and the state of adsorbed asphaltenes and resins, we performed a preliminary study using the QCM-D technique. The silica crystal was calibrated with toluene, and exposed to asphaltene and/or resin solutions identical to those used to coat silica particles. Fig. 2 shows the adsorption from toluene-solutions with asphaltenes and resins. The shift in resonance frequency (f) along the left axis is a measure of the adsorbed mass, including bound solvent. The dissipation factor (D) along the right axis shows how fast the energy from the vibrating crystal is dissipated to the surrounding liquid. The higher order overtones (n = 3, 5, 7) are more sensitive to the region close to the surface than the fundamental frequency (n = 1) [63]. Fig. 2a shows an immediate drop in the resonance frequency and an increase in dissipation when the crystal is exposed to the asphaltene solution at 10 min. This behavior suggests a very rapid adsorption of asphaltenes onto the silica surface. The asphaltene solution was exchanged with an identical solution at 15 min without any signicant response in frequency or dissipation, indicating more or less saturation of the silica surface. When rinsed with toluene at 20 min, the resonance frequency increased and the dissipation factor decreased, suggesting that weakly bound asphaltenes desorbed, and were washed off during rinsing. However, a considerable amount of asphaltenes was retained at the surface. By using the Sauerbrey relation with the frequency change of the third overtone after the nal rinse with toluene (30 Hz), we obtain an adsorbed mass of 5.6 0.4 mg/m2 (three replicates). The fth and seventh overtone shear waves are not as sensitive to the outer part of the adsorbed layer and shows, as expected, slightly smaller frequency shifts. The dissipation shifts in Fig. 2a are rather small, taking into account the signicant amount of

Fig. 2. QCM-D measurement of the adsorption behavior of toluene solutions with (a) 1.0 wt.% asphaltenes and (b) 1.0 wt.% resins, onto a silica surface. The solution is introduced at 10 and 15 min followed by rinsing with toluene at 20 and 25 min. The change in frequency (f) and dissipation (D) is monitored for the 3rd, 5th and 7th overtone of the fundamental resonance frequency of the crystal.

irreversible adsorbed asphaltenes. We can see how the three normalized frequency overtones follow each other when the adsorbed layer is rinsed with toluene (time >20 min). A dissipation shift in the third overtone of only 1.5 106 at 30 min conrms that the remaining asphaltene lm is rigid. Similar experiment with a toluene solution of asphaltenes and resins (0.5 wt.% asphaltenes + 0.5 wt.% resins) showed that the frequency and dissipation responses were in the same ranges as those observed for asphaltenes in Fig. 2a. The irreversibly adsorbed mass (6.4 0.5 mg/m2 , 3 replicates) is rigidly attached to the silica surface. As discussed by Ekholm et al. [55], this is considerably more than expected for non-associating polymers (13 mg/m2 ), which may indicate multilayer or aggregate formation at the surface. Acevedo et al. [64] have reported aggregation of asphaltenes in toluene solution at concentrations above 50 mg/L and indications of multilayer formation in even more dilute solutions [45]. Whereas asphaltene and asphaltene + resin systems show more or less the same adsorption behavior as the pure asphaltenes, the resin system stands out as completely different. First of all, the shift in resonance frequency of the crystal,

A. Hannisdal et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 276 (2006) 4558 Table 3 Prescence of hydrocarbons at the silica particles Aerosil 200 7200 972 202 Pure 0.000 0.009 0.003 0.003 R 0.079 0.033 0.044 0.007 A 0.122 0.104 0.057 0.036 R+A 0.125 0.101 0.010 0.028

51

Relative absorbance at 2924 cm1 .

when being exposed to resins, is dramatically smaller than when asphaltenes contribute to the adsorption. Moreover, the amount of resins left on the silica surface after rinsing is very small. Earlier studies have shown smaller adsorption rate constants for resins compared to those for asphaltenes [65]. Ekholm et al. [55] concluded with insignicant resin adsorption onto a hydrophilic gold surface from toluene. The results shown in Fig. 2b can conrm a low degree of resin adsorption at the timescale of the experiment. It would be too ambitious to draw conclusions about the lm properties of the adsorbed layer given the weak responses in Fig. 2b. However, it should be emphasized that small amounts of resins adsorbed onto silica surfaces may still have great inuence on the partitioning of solids at oil/water interfaces, and can make a difference in the stabilization of such particle-stabilized emulsions. 3.2. Characterization of particles We have characterized the pure and coated silica particles to be used for emulsion stability with infrared spectroscopy and contact angle measurements. Visual inspection of the resin-coated silica particles clearly showed that the initially hydrophilic silica (Aerosil 200 and 7200) adsorbed considerably more resins than the hydrophobic silica (Aerosil 972 and 202). The resin-coated 200 product seemed to have adsorbed most resins based on its dark brown color. All particles that had been exposed to asphaltene/toluene solutions for 24 h were dark. By analyzing the toluene solutions for remaining asphaltenes with near-infrared (NIR) spectroscopy, we could conclude that the initially hydrophilic particles had adsorbed considerably more asphaltenes than the hydrophobic particles. These ndings were conrmed with IR spectroscopy of the coated solids as summarized in Table 3. The asymmetric stretching vibration from methylene groups at 2924 cm1 was used as a marker for the presence of hydrocarbons at the silica surface. Fig. 3 shows an example of the vibrational features of the unmodied and coated 200 products. It is seen from Table 3 that the initially hydrophilic silicates (200 and 7200) adsorb signicantly more asphaltenes than the hydrophobic silica particles. The hydrophobic silica products, 972 and 202, are surface-modied with dimethyldichlorosilane and polydimethylsiloxane, respectively. These additives occupy a number of the active (silanol) sites on the silica, leaving fewer available sites for asphaltene adsorption. All four products exhibited less adsorption of resins than of asphaltenes. The two hydrophilic silica products adsorbed more or less the same amount of hydrocarbons from asphaltene and asphaltene + resin solutions. Initially hydrophobic products

Fig. 3. Hydrocarbon content of pure and coated Aerosil 200 (hydrophilic) particles was measured by FTIR spectroscopy.

adsorbed more hydrocarbons from asphaltene solutions than from asphaltene/resin solutions. Results from contact angle measurements in air and in oil are presented in Fig. 4. The contact angles at the watersolidoil contact line will generally be lower than those in the real systems. For this reason the watersolidoil contact angle is usually reported, often without any further comments, as a representative parameter for the partitioning of individual particles at the oil/water interface. However, the determination of the true contact angle by the compressed disk method is, by no means, straightforward if possible at all. Binks and Lumsdon [20] and Yan and Masliyah [19] have raised a concern about measurements of the true contact angle. It is clear that the watersolidoil

Fig. 4. Contact angle measurements of particles pressed into disks. The open circles represent the water/solid/air contact angle and whereas the lled squares represent the water/solid/oil contact angle. Both contact angles were measured into the water phase. Contact angles, which exceeded 170 were set to 170 due to high uncertainty in the method for these particular measurements.

52

A. Hannisdal et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 276 (2006) 4558

contact angles reported in Fig. 4 is generally higher than the watersolidair contact angles but follows the same trends. When the particle disk is immersed in the oil phase, oil will ll the voids of the porous disk due to capillary forces. When the water drop is placed at the disk it will be partially repelled by the oil in the disk and give too high . This is even more prominent if the disk is rst immersed in water and then contacted by an oil drop due to the protecting water lm. Yan and Masliyah [19] made the same observations when performing contact angle measurements of clays. They showed that, when the clays were rst immersed into oil, the contact angle of the clays was much higher than that when the clays were rst contacted with oil. It will still be possible to rank the products by their hydrophobicity from the results presented in Fig. 4. The inuence of asphaltene and resin adsorption on the wettability of the hydrophilic silica is dramatic. In real systems, the change in particle wettability due to adsorption of resins and asphaltenes from crude oil is the primary reason for the accumulation of particles at the oil/water interface. The 7200 particles show similar alteration of wettability upon resin and asphaltene adsorption as the 200 particles. Considering that hydrophobic particles adsorb less than hydrophilic particles, and that contact angles of resins and asphaltenes have been measured to 101 and 108 in air, it is not surprising to see only minor wettability changes for 972 and 202 particles. So, in summary, very hydrophilic particles like the unmodied 200 or very hydrophobic particles like the 972 and 202 products will most certainly prefer the water and oil phase, respectively. On the other hand, the true partitioning of the remaining particles at the water/oil interface is not obvious from the contact angle results. We performed simple qualitative powder immersion experiments, where the particles were placed on a water/air interface. We then carefully replaced the air phase with oil and visually observed the partitioning of the colored particles (aggregates) between the two phases after 24 h. Some of the 200R particles had left the interface to the water phase and some resided at the interface with a greater portion in the water phase indicating that these particles had a true contact angle less than 90 . The 7200 seemed to be equally partitioned between the two phases whereas the 200A, 200R + A, 7200R, 7200A and 7200R + A preferred the oil phase. These particles could easily be suspended in the oil phase by gently mixing the oil phase with a spatula, which indicated that these particles had a true contact angle larger than 90 . We know that electrostatic forces are generally believed to not play a signicant role in the stabilization of w/o emulsions due to the low dielectric constant of the continuous oil phase. However, for o/w emulsions, electrostatic repulsion from overlapping double layers can signicantly reduce the rate of coalescence if the double layer thickness is large. Increasing ionic strength (salinity of water phase) compresses the electrical double layer, and reduces the electrophoretic mobility of the particles [66]. Consequently, the inuence of electrostatics as a stabilization mechanism will be reduced in systems with high concentration of counter-ions. We have measured the zeta potential of the particles in 0.01 M NaCl solutions, whereas emulsion stability testing was done using aqueous solutions with 0.6 M (3.5 wt.%) NaCl.

Due to experimental limitations (electrode polarization), electrophoretic mobility of the particles could not be measured in 3.5 wt.% NaCl solution. We do not know exactly to what extent electrostatic repulsion will inuence the stability of the particlestabilized emulsions studied here. However, Binks and Lumsdon [8] have investigated the effect of pH and NaCl addition on the stability to occulation of Aerosil 200 particles and on the stability of toluene-in-water emulsions prepared from them. At pH 2, 4 and 5 no occulation occurred with addition of up to 5 M NaCl. At pH 6 and above, occulation was accompanied by a sharp increase in turbidity. The concentration of salt at the onset decreased with an increase in pH, whereas the maximum turbidity increased with pH possibly indicating a large occulation size, which forms when fully charged particles are screened. At pH 6 and pH 7 the critical coagulation concentration of NaCl was reported to be 1.2 and 0.5 M, respectively. As already mentioned, electrostatic repulsion between oil drops covered with particles may not be signicant if the double layer thickness of such colloids is too small to prevent droplet coalescence. However, electrostatic effects may affect the arrangement of particles at the interface. If crude oil components occupy silanol sites and neutralize the surface charge, there is a total contraction of the double layer around the negatively charged particles. Hence, the effective size of the nes is reduced, thus facilitating a closer packing of particles at the oil/water interface [6]. Changing the inter-particle interactions will affect the dynamic properties of the particles as well as the rheological properties of the interface. Fig. 5a shows the zeta potential () of pure and coated 200 particles as function of pH. The pH at which the zeta potential is zero denes the iso-electric point (IEP) of the material. In all cases the silicates show an increasing negative zeta potential when pH is raised above the IEP. At high pH, the silica surface shows a plateau, at which all silanol groups are deprotonated. At neutral pH the particles carry a net negative charge. The 200 particles exhibit an IEP at pH 2 while the coated particles show an IEP at pH 4.55.5. As discussed previously, resins and asphaltenes can occupy silanol sites and neutralize the charge of the silica surface. Surface charges from resins and asphaltenes arise from the dissociation of a large number of different acid functionalities and the protonation of different basic functional groups [67]. Zeta potentials of pure and resin-coated 7200 particles are shown as a function of pH in aqueous solution in Fig. 5b. We know that the 7200 product is a structure-modied product based on unmodied silica (Aerosil 200) where the number of silanol groups is reduced during the treatment to about 30% of the initial value. The pure 7200 particle surface is more positively charged than the pure 200 surface over the entire pH range studied. At high pH, remaining silanol groups are dissociated. The -prole of resin-coated particles is shown in Fig. 8. At low pH, basic functional groups protonate and give rise to positive charge. Resin-coated particles show similar -plateau as the uncoated particles in alkaline solution. However, in the pH range from 4 to 8, the particles are more negatively charged upon adsorption of resins. Due to complex surface chemistry, unknown stoichiometry, and steric effects, we will not account for this behavior.

A. Hannisdal et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 276 (2006) 4558

53

Fig. 5. Zeta potential of: (a) pure and coated Aerosil 200 particles and (b) pure and coated Aerosil 7200 particles.

Fig. 6. Separation of emulsions stabilized by: (a) 0.5% (w/v) 200R particles and (b) 0.5 (w/v) 200A particles. The left ordinate axis represents the amount of disperse phase resolved after gravitational induced coalescence (1580 g, 10 min). The Sauter mean diameter d32 is represented along the right ordinate axis for stable emulsion systems in a.

3.3. Performance The efciency of solids to stabilize water and oil emulsion systems was determined from the amount of dispersed phase resolved after centrifugation at 3000 rpm (1580 g) for up to 30 min. To simplify the representation, we will only report the amount of dispersed phase resolved after 10 min of centrifugation. We ensured that the particles (pure or coated) were the only stabilizers in the emulsion system, meaning that asphaltenes or resins did not exist in either oil or water solutions. The coated particles were washed repeatedly with toluene before they were used and the model oil (heptane:toluene, 70:30) showed no traces of asphaltenes or resins. Unmodied silica particles (200) did not act as good stabilizers due to the very hydrophilic character and separated over time even at 1 g. Fig. 6 shows that the stabilization efciency was greatly enhanced by adsorption of crude oil components. When crude oil components adsorb onto hydrophilic silica, the number of silanol sites on the surface is reduced as shown in Fig. 5a. Unlike the 200 particles, the resin-coated particles are interfacially active and will be partitioned at the oil/water inter-

face with a greater portion extending through the water phase. The hydrophilic nature of the 200R particles is reected by the o/w-type emulsions formed even at high volume of the oil phase (80 vol.%), which ts well with the observations from the powder immersion test. Moreover, Fig. 6a shows the importance of the initial location of the 200R particles. The particles stabilized the emulsion to some degree at 1 g when suspended in water but the emulsion separated completely within the rst 10 min of centrifugation. These particles performed much better as stabilizers when rst suspended in the oil phase. The partitioning of particles at the interface is most certainly different depending on the initial location of the particles, which can be seen in conjunction with observations from contact angle measurements. Aveyard et al. [5] have discussed similar observations with contact angle hysteresis and particleuid interactions. Menon and Wasan [41] and Yan and Masliyah [19] have studied water co-adsorption onto clay particles. The phenomenon seems to be very much related to studies of reservoir wettability [68]. Depending on the hydrophilicity of the reservoir, water will adsorb and protect the mineral surface from the oil phase. Rupture of such protective water lms has proven to be very difcult and involve slow kinetics. Both resin coated (Fig. 6a)

54

A. Hannisdal et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 276 (2006) 4558

and asphaltene coated 200 particles (Fig. 6b) produce stable emulsions of very large drop sizes (up to 100 m) when rst suspended in oil. This is a typical feature of particle-stabilized emulsions, which is not observed for surfactant-stabilized emulsions [5]. As seen in Fig. 6a the droplet size in the systems were particles were rst immersed in oil increase progressively with increasing volume of the disperse phase and decreasing emulsion stability. The 200 particles coated with asphaltenes (200A) are generally the most efcient in stabilizing emulsions and produced extremely stable o/w systems with gradually increasing drop size (from 20 m at 20 vol.% oil to 57 m at 60 vol.% oil) up to a point between 60 and 70 vol.% oil, where the emulsions invert to w/o-type. The catastrophic phase inversion, brought about by increasing the volume of the disperse phase has previously been shown to be dependent on the wettability [20] and the initial location [24] of the particles. The 200 particles coated with resins + asphaltenes (200R + A) show more or less similar performance as the 200A particles, with slightly less stable emulsions and phase inversion between 60 and 70 vol.% oil. The inuence of coating on the stabilizing efciency can be summarized as: (200A) > (200A+R) 200R > 200. Catastrophic phase inversion is also seen for the 7200 products in Fig. 7. Aerosil 7200 is a structure modied (3-methacryl-oxypropyl-trimethoxysilane) product based on Aerosil 200. We know from contact angle measurements and powder immersion tests that the 7200 particles have intermediate wetting properties. As seen for the 200A particles, the initial location of the particles seems to affect the point of inversion whereas the emulsion stability is more or less the same irrespective of which phase the particles are suspended in. The stability to gravitational induced separation (coalescence) passes through a minimum approaching inversion in line with a maximum in drop size of the disperse phase. Since microscopy indicated unspherical droplets and some degree of aggregation of the particles in the continuous phase, we have not calculated the Sauter mean diameters for these series. It should be mentioned that Binks and Lumsdon [7,20] have reported a maximum in stability to sedimentation/creaming at the point of inversion. However, the stability experiments by Binks and Lumsdon were done by sequential addition of oil or water where the oil phase contained 2 wt.% hydrophobic particles. The particle concentration in the emulsion was not held constant as in our experiments. In another study by Binks and Whitby [9], oil was added continuously to an aqueous phase containing 0.7 wt.% hydrophobic silica particles. The system underwent catastrophic phase inversion with maximum drop size at inversion. Type of coating plays a minor role for the stabilizing efciency of the 7200 solids. All particles show similar stability proles with minimal stability approaching inversion and a shift of the inversion point by 1020 vol.% oil when the particles are rst contacted with the water phase. The stability prole for the asphaltene coated particles (7200A) is shown in Fig. 7b whereas the proles for 7200R and 7200A + R are not presented. From results presented in Figs. 68 it is apparent that transitional phase inversions can be induced by changing the wettability of the particles, either by modifying the surface with silanes or siloxanes or by coating the particles with crude oil components. Thus, one way to

Fig. 7. Separation of emulsions stabilized by: (a) 0.5% (w/v) 7200 particles and (b) 0.5 (w/v) 7200A particles. Catastrophic phase inversion was induced by increasing the volume of the disperse phase.

affect transitional inversion is to simply decrease the silanol content on the particles as done systematically by Binks and Lumsdon [20] for hydrophilic silica particles silylated to different extents by reaction with dimethyldichlorosilane. By coating the 7200 particles with asphaltenes (7200A), hydrophobicity increased and the preferred emulsion type changed from o/w to w/o. The point of phase inversion is shifted to the left in Fig. 7. All 972 particles (pure and coated) are very hydrophobic and we did not succeed to suspend the particles in water. Addition of more particles, from 0.1% (w/v) in Fig. 8a to 0.3% (w/v) in Fig. 8b caused a signicant increase in emulsion stability of the w/o emulsions and resulted in smaller drops as more particles were available to stabilize the interfaces. The stabilizing effectiveness of the particles can be presented in the following order: (972A) (972A + R) > (972R) (972). Sauter mean diameters of the water drops in the emulsions in Fig. 8b are presented in Table 4. Generally, and unlike that for surfactant stabilized systems, these coarse emulsions were very stable to coalescence even when exposed to a high gravitational eld (1580 g for 10 min). The diameter increased progressively with increasing volume of the disperse water phase in line with the decrease

A. Hannisdal et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 276 (2006) 4558

55

Fig. 8. Separation of w/o emulsions stabilized by: (a) 0.1% (w/v) and (b) 0.3 (w/v) 972-particles. The Sauter mean diameters for the systems are reported in Table 4.

in emulsion stability, whereas different type of coating did not affect the mean diameter to the same extent. As mentioned in the introduction, Vignati et al. [25] have succeeded in making stable emulsions with particle coverage of only 5%. It was therefore, in our interest to do some kind of evaluation of the total droplet surface and compare it to the amount of particles present to potentially cover the interface. It is generally believed
Table 4 Sauter mean diameter (d3,2 ) of w/o emulsions stabilized by 0.3% (w/v) particles Oil (vol.%) 972 d32 30 40 50 60 70 80 74 63 54 42 32 85 70 58 48 36 2 26 21 13 17 12 972R d32 71 60 47 45 29

that, for a given input of mechanical energy during emulsication, the initial drop size (before any further coalescence) is dependent on the amount of the two phases present, the tension of the interface and the mass density of the continuous phase [23]. After agitation, drop growth will proceed either by coalescence or by Ostwald ripening, thus reducing the total interfacial area, until the particle coverage is sufcient to stabilize drops against further coalescence. The diameter at which all particles are accommodated at the interface is often referred to as the limited coalescence diameter [5,9,10,19,31,69]. The Sauter mean diameter and the volume of the disperse phase in Table 4 provide the necessary information to estimate the total interfacial area of the water/oil interface for the systems stabilized by 972/972R/972A particles. Doing this interestingly show that the total interfacial areas of the 16 different systems are more or less the same (mean = 0.56 m2 , S.D. = 0.07). Microscopy showed no evidence of excess particle aggregates in the continuous phase. It seems like the particle amount, which is equal in all these cases, determines the limiting water/oil interfacial area irrespectively of the ratio between oil and water. For all droplet size results presented in this study, the cumulative volume distributions were tted successfully to a lognormal distribution function (regression residuals <1). We have not observed any multimodal droplet size distributions. Table 4 shows the distribution parameters for the emulsions stabilized by 972 particles. We can see that the size distributions were systematically broadened (increasing 2 ) proportionally with increasing drop size, in contrast to what would be expected for systems where droplet growth had narrowed the drop size distribution. Similar results have been found throughout this study, for both hydrophilic and hydrophobic silica particles. From these observations it seems like fragmentation and coalescence (collision) during emulsication differs for systems with high or low volume concentration of the disperse phase. After emulsication, drops may grow in size until the surface concentration of particles prevents further growth. Results show that these processes lead to different polydispersity of the system with high or low volume concentration of the disperse phase. The measurements were performed within the 10 rst minutes after making the emulsion, which is generally too short time to expect any signicant Ostwald ripening of the drops. We expect any droplet growth to be caused by coalescence, which is rapid initially and ceases to become very slow. The systems seem to be very stable to coalescence only minutes after emulsication. We will not take a decision on whether the

972A 86 67 51 49 32 2 31 22 14 13 8 d32 90 60 56 48 44 30 99 70 60 55 48 33 2 29 21 16 17 12 8

The volume distribution was tted to lognormal distribution function with reported mean () and standard deviation ( 2 ).

56

A. Hannisdal et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 276 (2006) 4558

Fig. 9. Separation of w/o emulsions stabilized by 0.3% (w/v) 202A particles. The Sauter mean diameter d32 is represented along the right ordinate axis.

Fig. 10. Irregular shape of a 500 m drop covered with particles.

resulting interfacial area is that where all particles are accommodated in a hexagonal closed packed arrangement or not. Since the aggregation state of the silica particles at the interface is unknown it is difcult to say whether the interface is completely covered with particles or not. However, by assuming partitioning of primary particles at the interface, the amount of particles used in our experiment will generally be in large excess to that needed for full coverage of the interface. Since microscopy showed no evidence of particle aggregates in the continuous phase, the silica particles will most likely be present as agglomerates with sizes of several hundred nanometers, which reduces the effective packing of these particles at the interface. Binks and Whitby [9] have done a systematic study of the parameters affecting emulsication of silica-stabilized emulsions. We expect extremely hydrophobic particles to perform poorly as stabilizers since these particles will have low activity for the water/oil interface. Emulsions stabilized by the 202 particles and the resin coated particles (202R) resulted in complete emulsion destabilization under the high centrifugal eld even with particle concentrations up to 1.0% (w/v). These emulsions were stable for at least 30 min at 1 g, but the emulsion stabilized by 202 particles was destabilized within the rst minute of centrifugation at 800 rpm (113 g). Coating the 202 particles with asphaltenes (202A) or asphaltenes + resins (202A + R) dramatically increased emulsion stability and formed very stable emulsions at 0.3% (w/v) particle concentration. Increasing the concentration of particles from 0.1% (w/v) to 0.3% (w/v) signicantly improved stability of these systems. We can see from Fig. 9 that the stability of the 202A emulsions decreased progressively with increasing volume of the disperse water phase, in line with increasing mean drop size of the disperse water phase. The 202A + R particles showed only slightly lower performance as stabilizers. The total interfacial area of the different systems are the same (mean = 0.62 m2 , S.D. = 0.07). The stabilization effectiveness of 202 particles can be summarized as: (202A) > (202A + R) (202R) > (202). All 202-products produced w/o emulsion type even at high

water phase volumes, reecting the hydrophobic nature of the particles. We have in some cases observed stable irregular emulsion droplets like the 500 m droplet presented in Fig. 10 (also smaller drops), which seems to be highly covered with particles. With an oscillating pendant drop measurement similar to that reported by Tambe and Sharma [32], we have proven that such interface will possess very high elasticity which will add stability to the interface (not presented here). If particles are in large excess of that required for accommodation at the water/oil interface, increased viscosity of the continuous phase due to particle network formation (H-bond linkages) can also add a certain stability to the emulsion system [5]. As apparent from the previous section, we have not studied how pH or different salt concentrations will affect the performance of the particles in the emulsions. Both properties will be important in determining the electrostatic forces and particle coagulation in the systems. We did our stability experiments with aqueous phase pH 6.7 and salt concentration of 0.6 M NaCl. For similar studies, we will recommend to measure the degree of particle aggregation in bulk with some kind of light scattering instrument. Microscopic drop size determination may prove aggregation of larger aggregates, but not provide quantitative results. 3.4. Relevance to petroleum industry Wettability alteration of reservoir rock has been extensively studied by geologists, and will not be discussed here even though the surface afnity of heavy crude oil components is a key factor for oil recovery. The presence of inorganic particles at oil/water interfaces, in process equipment like settlers or electrocoalescers, can cause severe operational problems if emulsions are too stable to be destabilized by traditional methods. Even if the initial concentration of inorganic particles is low, particles tend to accumulate in process equipment over time and build up sludge layers of oil/water/particles [70]. Such layers require aggressive treatment methods or removal. Oil trapped in the sludge layer is lost and will constitute an economical loss

A. Hannisdal et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 276 (2006) 4558

57

as well as a waste problem. With knowledge about the relevant stabilizing mechanisms as well as information of the production history (pressure cycles, energy input, and chemical additives) of the oil/brine/particle system, particle-stabilized emulsions can be treated or prevented [71]. A correct choice of chemicals may displace particles from the interface, or prevent interfacial partitioning altogether. Gu et al. [11] have discussed the possibility to use particles for viscosity control in transportation of crude oil systems. 4. Conclusions We have coated dry silica nanoparticles with heavy crude oil components and investigated the performance of these solids as stabilizers in model oil/water emulsion systems. Adsorption studies, using the QCM-D technique, have shown multilayer or aggregate formation of asphaltenes at the silica surface. Asphaltenes are irreversibly adsorbed as a rigid lm. The adsorbed amount from resin solution is dramatically smaller than when asphaltenes contribute to the adsorption. The effect of the surface modication has been monitored using infrared spectroscopy, contact angle measurements and zeta potential measurements. Initially hydrophilic particles show signicant adsorption of crude oil components leading to wettability alteration and surface charge effects. Hydrophobic particles adsorb smaller amounts of asphaltenes and resins, and show only minor wettability changes. It is shown how coating of the nanoparticles with asphaltenes and resins will modify the stabilization prole compared to that obtained with no coating. The stabilization efciency was greatly enhanced by adsorption of crude oil components onto very hydrophilic or very hydrophobic silica. Generally, the performance of the particles as stabilizers seems to be strongly dependent on their wettability. We have demonstrated catastrophic phase inversion of emulsions stabilized by particles with intermediate wetting properties, induced by simply increasing the volume of the disperse phase. In all cases, the stability to gravitational induced separation (coalescence) passes through a minimum approaching inversion in line with a maximum in drop size of the disperse phase. The initial location of the particles is important for the stability of the system and the type of emulsion formed. Transitional phase inversion from o/w to w/o emulsion type can be achieved by modifying the hydrophilicity of the particles, either by silylation or by a controlled coating with heavy components from crude oil. Generally, the emulsions consist of a lognormal distribution of droplets centered at a large mean diameter (up to 50 m). The emulsions are still very stable, which is typical for particle-stabilized emulsions. Acknowledgement The technology program Flucha III, Particle-stabilized Emulsions/Heavy Crude Oils, nanced by industry and the Norwegian Research Council (Strategisk omstillings program) are gratefully acknowledged for nancial support. We are also grateful to Degussa for providing us with the samples of the silica particles.

References
[1] A.P. Sullivan, P.K. Kilpatrick, Ind. Eng. Chem. Res. 41 (2002) 3389. [2] O.V. Gafonova, H.W. Yarranton, J. Colloid Interface Sci. 241 (2001) 469. [3] Z. Yan, J.A.W. Elliott, J.H. Masliyah, J. Colloid Interface Sci. 220 (1999) 329. [4] D.M. Sztukowski, H.W. Yarranton, J. Colloid Interface Sci. 285 (2005) 821. [5] R. Aveyard, B.P. Binks, J.H. Clint, Adv. Colloid Interface Sci. 100 (2003) 503. [6] B.P. Binks, S.O. Lumsdon, Langmuir 17 (2001) 4540. [7] B.P. Binks, S.O. Lumsdon, Langmuir 16 (2000) 2539. [8] B.P. Binks, S.O. Lumsdon, Phys. Chem. Chem. Phys. 1 (1999) 3007. [9] B.P. Binks, C.P. Whitby, Langmuir 20 (2004) 1130. [10] N.P. Ashby, B.P. Binks, Phys. Chem. Chem. Phys. 2 (2000) 5640. [11] G. Gu, Z. Zhou, Z. Xu, J.H. Masliyah, Colloids Surf. A: Physicochem. Eng. Aspects 215 (2003) 141. [12] D.E. Tambe, M.M. Sharma, J. Colloid Interface Sci. 157 (1993) 244. [13] B.P. Binks, Curr. Opin. Colloid Interface Sci. 7 (2002) 21. [14] P. Finkle, H.D. Draper, J.H. Hildebrand, J. Am. Chem. Soc. 45 (1923) 2780. [15] J.H. Schulman, J. Leja, Trans. Faraday Soc. 50 (1954) 598. [16] J. Mizrahi, E. Barnea, Br. Chem. Eng. 15 (1970) 497. [17] A. Gelot, W. Freisen, H.A. Hamza, Colloids Surf. A: Physicochem. Eng. Aspects 12 (1984) 271. [18] N. Yan, J.H. Masliyah, Colloids Surf. A: Physicochem. Eng. Aspects 96 (1995) 229. [19] N. Yan, J.H. Masliyah, J. Colloid Interface Sci. 181 (1996) 20. [20] B.P. Binks, S.O. Lumsdon, Langmuir 16 (2000) 8622. [21] B.P. Binks, C.P. Whitby, Colloids Surf. A: Physicochem. Eng. Aspects 253 (2005) 105. [22] B.P. Binks, S.O. Lumsdon, Langmuir 16 (2000) 3748. [23] P.A. Kralchevsky, I.B. Ivanov, K.P. Ananthapadmanabhan, A. Lips, Langmuir 21 (2005) 50. [24] B.P. Binks, S.O. Lumsdon, Phys. Chem. Chem. Phys. 2 (2000) 2959. [25] E. Vignati, R. Piazza, T.P. Lockhart, Langmuir 19 (2003) 6650. [26] E.J. Stancik, M. Kouhkan, G.G. Fuller, Langmuir 20 (2004) 90. [27] E.J. Stancik, G.G. Fuller, Langmuir 20 (2004) 4805. [28] S.E. Friberg, J. Dispersion Sci. Technol. 26 (2005) 647. [29] B.R. Midmore, Colloids Surf. A: Physicochem. Eng. Aspects 132 (1998) 257. [30] S. Tarimala, S.R. Ranabothu, J.P. Vernetti, L.L. Dai, Langmuir 20 (2004) 5171. [31] V.B. Menon, D.T. Wasan, Colloids Surf. 23 (1987) 353. [32] D.E. Tambe, M.M. Sharma, Adv. Colloid Interface Sci. 52 (1994) 1. [33] D.E. Tambe, M.M. Sharma, J. Colloid Interface Sci. 162 (1994) 1. [34] D.E. Tambe, M.M. Sharma, J. Colloid Interface Sci. 171 (1995) 456. [35] K.P. Velikov, F. Durst, O.D. Velev, Langmuir 14 (1998) 1148. [36] T.R. Briggs, J. Ind. Eng. Chem. 13 (1921) 1008. [37] J. Thieme, S. Abend, G. Lagaly, Colloid Polym. Sci. 277 (1999) 257. [38] S. Chen, G. ye, J. Sj blom, J. Dispersion Sci. Technol. 26 (2005) 791. o [39] S. Chen, G. ye, J. Sj blom, J. Dispersion Sci. Technol. 26 (2005) 495. o [40] V.B. Menon, D.T. Wasan, Colloids Surf. 19 (1986) 89. [41] V.B. Menon, D.T. Wasan, Colloids Surf. 19 (1986) 107. [42] Y. Yan, J.H. Masliyah, Colloids Surf. A: Physicochem. Eng. Aspects 75 (1993) 123. [43] N. Yan, J.H. Masliyah, J. Colloid Interface Sci. 168 (1994) 386. [44] N. Yan, J.H. Masliyah, Colloids Surf. A: Physicochem. Eng. Aspects 96 (1995) 243. [45] S. Acevedo, M.A. Ranaudo, C. Garca, J. Castillo, A. Fern ndez, M. a Caetano, S. Goncalvez, Colloids Surf. A: Physicochem. Eng. Aspects 166 (2000) 145. [46] S. Acevedo, J. Castillo, A. Fernandez, S. Goncalves, M.A. Ranaudo, Energy Fuels 12 (1998) 386. [47] S. Acevedo, M.A. Ranaudo, G. Escobar, L. Gutierrez, P. Ortega, Fuel 74 (1995) 595.

58

A. Hannisdal et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 276 (2006) 4558 [60] S. Kokal, T. Tang, L. Schramm, S. Sayegh, Colloids Surf. A: Physicochem. Eng. Aspects 94 (1995). [61] M. Rodahl, R. H ok, A. Krozer, B. Kasemo, P. Breszinsky, Rev. Sci. o Instruments 66 (1995) 3924. [62] M.-H. Ese, P.K. Kilpatrick, J. Dispersion Sci. Technol. 25 (2004) 253. [63] J.J.R. St lgren, P.M. Claesson, T. W rnheim, Adv. Colloid. Interface a a Sci. 8990 (2001) 383. [64] S. Acevedo, M.A. Ranaudo, J.C. Pereira, J. Castillo, A. Fern ndez, P. a P rez, M. Caetano, Fuel 78 (1999) 997. e [65] G. Gonz lez, A. Middea, J. Dispersion Sci. Technol. 8 (1987) 525. a [66] B.J. Kirby, E.F. Hasselbrink, Electrophoresis 25 (2004) 187. [67] H. Parra-Barraza, D. Hern ndez-Montiel, J. Lizardi, J. Hern ndez, R.H. a a Urbina, M.A. Valdez, Fuel 82 (2003) 869. [68] E.M. Freer, T. Svitova, C.J. Radke, J. Petrol. Sci. Eng. 39 (2003) 137. [69] R.M. Wiley, J. Colloid Sci. 9 (1954) 427. [70] P.J. Breen, D.T. Wasan, Y.H. Kim, A.D. Nikolov, C.S. Shetty, Emulsions and emulsion stability, in: J. Sj blom (Ed.), Emulsions and Emulsion o Stability, Marcel Dekker, Inc., New York, 1996, p. 237. [71] M.-H. Ese, C.M. Selsbak, A. Hannisdal, J. Sj blom, J. Dispersion Sci. o Technol. 26 (2005) 145.

[48] S. Acevedo, M.A. Ranaudo, C. Garca, J. Castillo, A. Fern ndez, Energy a Fuels 17 (2003) 257. [49] K.R. Dean, J.L. McAtee Jr., Appl. Clay Sci. 1 (1986) 313. [50] G. Gonz lez, A. Middea, Colloids Surf. 33 (1988) 217. a [51] Y.W. Jeon, S.H. Yi, S.J. Choi, Fuel Sci. Technol. Int. 13 (1995) 195. [52] A.W. Marczewski, M. Szymula, Colloids Surf. A: Physicochem. Eng. Aspects 208 (2002) 259. [53] T. Pernyeszi, A. Patzko, O. Berkesi, I. Dekany, Colloids Surf. A: Physicochem. Eng. Aspects 137 (1998) 373. [54] M. Szymula, A.W. Marczewski, Appl. Surf. Sci. 196 (2002) 301. [55] P. Ekholm, E. Blomberg, P. Claesson, I.H. Auem, J. Sj blom, A. Korno feldt, J. Colloid Interface Sci. 247 (2002) 342. [56] A. Hannisdal, P.V. Hemmingsen, J. Sj blom, Ind. Eng. Chem. Res. 44 o (2005) 1349. [57] P.M. Spiecker, K.L. Gawrys, C.B. Trail, P.K. Kilpatrick, Colloids Surf. A: Physicochem. Eng. Aspects 220 (2003). [58] A. Jada, A.A. Chaou, Fuel 81 (2002) 1669. [59] R.M. Silverstein, Spectroscopic Identication of Organic Compounds, sixth ed., John Wiley & Sons, Inc., New York, 1998.

Paper 3

Viscoelastic Properties of Crude Oil Components at Oil-Water Interfaces. 1. The Effect of Dilution.
Andreas Hannisdal a,*, Robert Orr b, and Johan Sjblom a
Ugelstad Laboratory, Department of Chemical Engineering, Norwegian University of Science and Technology (NTNU), N-7491 Trondheim, Norway. b Norsk Hydro ASA, Research Park, N-3915 Porsgrunn, Norway.
a

The dilational viscoelastic properties of diluted crude oil-water interfaces have been studied using the oscillation drop method. The study focuses on the effect of altering aromaticity of the solvent and the concentration of the crude oil on the viscoelastic response of the crude oil-water interface. Dynamic interfacial tension experiments (pendant drop), asphaltene aggregation state experiments (nearinfrared spectroscopy) and emulsion stability experiments (bottle test) have complemented the studies of interfacial rheological parameters in order to understand the mechanisms of film formation and emulsion stabilization. The overall w/o-emulsion stability seemed to be determined by the aggregation state of asphaltenes in bulk and the reduced sedimentation rate of water droplets in concentrated systems. The storage E and loss E moduli of the crude oil/water systems were determined with dilation rheology. At a perturbation frequency = 0.1 Hz, the equilibrium storage and loss moduli passed through distinct maxima as a function of bulk concentration. The apparently low viscoelasticity of the interfaces in systems with high bulk concentration was probably caused by high diffusion flux of interfacially active components from bulk and was not entirely due to interactions within the adsorbed layers. Dilation experiments at other perturbation frequencies confirmed this phenomenon. Results have been discussed in connection with recent findings regarding the stabilization of water-in-oil emulsions. Keywords: Crude oil; Dilution; Asphaltene aggregation; Kinetics of film formation; Dilational viscoelasticity; Emulsion stability

INTRODUCTION
As mentioned by Saniere et al. [1], the heavy, extra-heavy oils and bitumen represent a huge amount of hydrocarbon resources but yet only a little part of the worldwide oil production. However, with the combination of an increase of world energy demand and the decline of conventional oils, heavy oil and oil sand are expected to become more important hydrocarbon resources in the future. Greene et al. [2] have done a quantitative analysis of the transition from conventional to unconventional oil resources from an optimists perspective. Heavy crudes which often result from bacterial oxidation of conventional oils have different physical and chemical properties than conventional oils, generally: greater viscosity, greater amount of asphaltenes, heavy metals, sulphur and nitrogen. This leads to new technological challenges in all phases of operation: recovery, transportation and refining [1]. Due to their very high viscosity, heavy crude oils cannot be transported with conventional pipelines and acquire additional treatment. An efficient flow assurance method is to dilute the heavy oil with a less viscous hydrocarbon phase such as a condensate, naphtha, kerosene or light crudes (other approaches are heating, making

oil-in-water emulsions, upgrading). Dilution with a light hydrocarbon is also necessary in oil sand industry before froth treatment of bitumen (removing water and solids) due to the similar density of water and bitumen [3]. Chlorides and solids in water may lead to serious corrosion and fouling problems in downstream upgrading units and must be removed [4]. Adding a diluent reduces the oil phase density and viscosity which enhances the sedimentation rate of water droplets. Both in heavy oil and oil sand industry, formation of stable water-in-oil emulsions obviously cause serious operation problems. It is of great interest to identify the crude oil properties (bulk and interfacial) contributing to the overall stability of these emulsion systems. As already mentioned, the bulk properties of the oil phase are important factors at least for the rate of flocculation of water droplets [5]. However, research has made it clear that these systems also have certain interfacial properties which make water droplets extremely stable against coalescence. Many of these properties have been assigned to the natural surfactants in the crude oil, asphaltenes and resins, which typically are in high amounts in heavy oil and bitumen. Asphaltenes are defined as the crude oil components which are

soluble in toluene but precipitate in the presence of aliphatic solvents, whereas the resin fraction is defined as the fraction soluble in light alkanes but insoluble in liquid propane [6]. Asphaltenes and resins are characterized by large, polar, polynuclear molecules consisting of condensed aromatic ring systems and heteroatoms like sulphur, nitrogen and oxygen. The significance of the solubility state of asphaltenes in determining the emulsion stability of crude oil/water systems has been studied extensively. It is generally agreed upon that asphaltenes tend to self-associate and precipitate in unfavorable solvent conditions which may lead to particle-stabilization of water/oil interfaces or deposition in process units. McLean and Kilpatrick [7, 8] have presented molecular models to explain how asphaltenes respond to changes in solution conditions by varying the amount of oil components that contribute to asphaltene solubility. Another feature of asphaltenes is their strong adsorption at oilwater interfaces at certain conditions. If an aged oil drop is retracted, a rigid, wrinkled skin is clearly visible. After the initial retraction, the drop slowly relaxes back to a shape similar to that prior to retraction [9]. The phenomenon is related to the irreversible adsorption of material at the interface. Due to the extreme energy of displacement of particles from oil/water interfaces [10-12], the particles are practically irreversibly adsorbed and run out of area when the drop is contracted. Skin formation has been directly correlated to interfacial elasticity and given much attention as one of the dominating mechanisms for stabilization of asphaltene rich crude oils [13, 14]. However, reliance only on the appearance of rigid interfacial films may be misleading since clearly interfacial elasticity is evident even when skins are not visible [15]. The two-dimensional dilational rheology of crude oil-water interfaces, studied with the oscillating pendant drop technique, can provide quantitative information of the dilational viscoelasticity of the interfaces. With this technique, the interfacial tension of an oil drop in water is determined from the shape of the drop according to the Young-Laplace equation. The interfacial rheological parameters can be calculated from the tension-response to a dilation of the drop. Interfacial rheology of crude oil systems has been studied previously [9, 15-21]. Some of these studies will be discussed later with the results from our work. With strong relevance to this study, Freer et al. [22] recently raised caution for reporting interfacial rheological properties using the oscillating drop technique due to the effect of viscous forces on the drop shape. In the case of a static drop, analysis of the drop shape by Young-Laplace equation is exact as the only forces acting on the drop are gravity and interfacial tension. However, for oscillating drop experiments also viscous forces can influence the shape of the drop. Freer et al. showed that for clean fluid/fluid interfaces, the measured interfacial tension oscillates sinusoidally with the same frequency as the modulation of the drop volume. When the Capillary number Ca, defined as the ratio of viscous to capillary forces, was beyond a

certain value viscous forces distorted the drop shape and led to false interfacial tension results.
Ca = V 0a2

(1)

In Equation 1, is the difference in viscosity between the two phases, is the oscillation frequency, V is the amplitude of volume oscillation, 0 is the equilibrium interfacial tension and a is the radius of the capillary. To avoid instrumental problems due to the viscous nature of heavy oil, it is common practice to dilute the sample or to extract the asphaltenes and the resins from the crude oil and work with their solutions in a solvent. Because asphaltenes and resins are the main surface active components of crude oil, we may expect that these model solutions mimic correctly the interfacial behavior. However, diluting the sample is not at all straightforward and may change the bulk and interfacial properties completely. Several studies on bitumen have concluded that there is a critical dilution ratio: diluent/bitumen (D/B) which coincides with precipitation of asphaltenes, at which many of the system properties change dramatically [23]. Below this critical dilution (high bitumen concentration) the oil/water interface is flexible, whereas above this critical dilution the interface is rigid and shows the typical crumpling effect. Wu [4] has isolated and characterized the interfacial material both below and above the critical concentration. The flexible interfacial film was composed of a mixture of asphaltene and carboxylic salts with a combined H/C ratio of 1.32, while the rigid interfacial film was composed of asphaltenes alone with H/C ratio of 1.13. It was proposed that the presence of sodium naphtenates at high bitumen concentration might hinder the conformational change of asphaltenes at the interface, thus preventing rigid film formation. Czarnecki and Moran [23] and Wu and Czarnecki [3] have presented models aiming at explaining the mechanism of water-in-oil emulsion stabilization based on results from the isolation procedure presented by Wu [4]. Goual et al. [24] have studied adsorption of bituminous components at oil-water interfaces with the QCM (Quartz Crystal Microbalance) technique. For Athabasca bitumen in heptane/toluene (1/1 vol. ratio) they observed two adsorption regimes depending on the bitumen concentration. At low bitumen concentration a slow, non-steady-state and irreversible adsorption took place whereas at high concentration, a steadystate adsorption with limited reversibility resulted in quick adsorption saturation. The threshold concentration between the two adsorption regimes was related to the critical dilution ratio characterizing the rigid-to-flexible-transition of the interfacial film. These and other studies of the state of asphaltenes, their adsorption, and the composition of the interfacial material have significantly contributed to a better understanding of the stabilization mechanism of water droplets in diluted bitumen or crude oil.

Let us consider a study where crude oils (depressurized) with different characteristics are analyzed and evaluated with respect to the overall emulsion stability and the mechanisms involved. Oscillating drop tensiometry is used to determine the rheological parameters of the oil/water interface and its contribution to the stability of the emulsion. The question that arises is: At which experimental conditions can we imitate the undiluted heavy crude oils and justify a comparison between different crude oil samples? This study focuses on the effect of altering aromaticity of the solvent and the crude oil concentration in the bulk phase. Viscoelastic parameters have been measured with the oscillating drop technique and related to the asphaltene aggregation state in bulk (NIR spectroscopy) in addition to the overall stability against coalescence in the emulsion (bottle-test). The kinetics of film formation was studied with dynamic interfacial tension measurements. This study will have relevance for decisions taken during production and transportation (crude oil compatibility) of heavy oils and to the ongoing studies of the stabilization mechanisms of water/crude oil and water/dilutedbitumen systems. EXPERIMENTAL Materials We have used three Brazilian crude oils with typical physicochemical properties for heavy crude oils: high viscosity, large amount of asphaltenes and resins, high viscosity and high molecular weight. The crude oils have been analyzed with respect to SARA components, density, acidity, viscosity and molecular weight. Table 1 lists characteristics of the crude oils. API (American Petroleum Institute) gravity ranges from 13 to 20. Sample A is a relatively acidic crude oil with as much as 10.2 wt.% of asphaltenes and a viscosity of more than 21000 mPas at 20oC. Neither of the samples contain inorganic particles, wax or water (< 0.3 wt.%). Sampling had been done in a way to avoid any contamination of production chemicals. The crude oils were diluted to different extents (crude oil concentration: Coil from 10-5 to 1 v/v) in heptane and toluene mixtures ranging from pure heptane to pure toluene. The crude oil was diluted by adding toluene, then heptane while continuously stirring the sample to avoid local concentration gradients of heptane leading to asphaltene precipitation. The diluted sample was sonicated for 5 minutes and left to age for at least 24 hours before using it. We used milli-Q water with 3.5 wt.% NaCl and pH of 6.5-7.0 throughout this study. The following viscosity standards were used to demonstrate the effect of bulk viscosity when measuring interfacial rheology with axisymmetric drop shape analysis: 9.4, 49.6, 104, 500, 950 mPas at 25oC (Brookfield) and 1276 mPas at 25oC (Paragon Scientific Ltd.). Near-Infrared Spectroscopy The near-infrared region is found between the visible and middle-infrared regions of the electromagnetic spectrum (780 2500 nm, 12820 - 4000 cm-1). The absorption spectra in this region consist of overtones and combinations of the fundamental

molecular vibration bands which are primarily due to stretches of hydrogen in C-H, N-H, S-H and O-H bonds. In addition to molecular absorption, the NIR spectra are dependent upon several physical parameters, where the most prominent is scattering from particles. As the particle size increases, more light is scattered by the sample which is reflected in the NIR spectrum as an upwards shift of the baseline. Details on light scattering in the near-infrared region can be found in literature [26, 27]. We have used a one-characteristic- wavelength (1600 nm) model to detect asphaltene aggregation in the diluted oil samples. The diluted crude oil samples were analyzed 24 hours after preparation in transmission mode (5 mm) by using a Bruker MPA FT-NIR spectrometer with a TE-InGaAs detector. All data acquisitions and Fourier transformations were performed with the OPUS software, version 4.2 (Bruker Optics). Table 1. Physicochemical properties of the three Brazilian crude oils. Property Unit A B C Saturates a (wt.%) 26.2 38.2 30.6 Aromatics (wt.%) 41.5 43.6 43.3 Resins (wt.%) 21.9 14.1 20.9 Asphaltenes (wt.%) 10.2 3.8 5.0 (g/ml) 20oC 0.980 0.935 0.940 Density b Acidity (TAN) (mg KOH/g) 3.73 0.60 1.29 284 447 Viscosity (mPas) 20oC 21400 199 309 (mPas) 25oC 12000 MW c (g/mol) 434 335 343 a Asphaltene precipitation in n-hexane followed by HPLC fractionation of maltenes. Details about the SARA method are reported elsewhere [25] b AP PAAR DMA 48 c Determined from the freezing point depression of benzene

Emulsion Stability Measurements We did emulsion stability measurements for the three crude oils and their dilutions (Coil from 10-3 to 1 v/v) in heptane and toluene mixtures with a total of 180 bottle tests. Water (milli-Q, 3.5 wt.% NaCl, 40 vol.%) was added to the oil phase and dispersed with an Ultra Turrax (IKA, T18 with a 10 mm head) first at 18000 rpm for 1 minute, then at 22000 rpm for two minutes. The water-in-oil emulsion (12 ml) was immediately after preparation transferred to a graded centrifuge tube. Due to the extreme stability of these systems, the stability to coalescence was quantified as the amount of water phase resolved after centrifugation at 3000 rpm (1580 x g) in steps from 2 minutes up to 50 minutes. Drop size or drop size distributions were not determined, but we recognize that factors like interfacial tension, bulk viscosity and mass density of the continuous phase will affect the droplet size after the fragmentation process. The energy input and the volume fraction of water were fixed parameters.

Dynamic Interfacial Tension Measurements The drop tensiometer (CAM 200 KSV Instruments, Finland) was used to measure the interfacial tension by analyzing the axial symmetric shape (Laplacian profile) of a rising oil drop in water. All the measurements were performed at controlled temperature (25 oC). We created an oil drop from a needle connected to a syringe (1mL). The syringe was equipped with a valve to close off the sample volume in the needle from that in the syringe. This is very important in order to keep a constant volume of the drop during long-time experiments. If sample climbs the Teflon plunger, the drop volume will decrease slightly. Due to the irreversible nature of asphaltene adsorption and film formation, a reduction of the drop volume will lead to increased surface pressure and significant measurement errors. We recorded images continuously during 24 hours after creating the oil/water interface, the first 17 minutes at a rate of 1 image/sec then at 2 images/minutes. We performed dynamic interfacial tension measurements for diluted oil samples in the range: Coil =10-5-1 v/v in mixtures of heptane and toluene (total 180 experiments).
Piezoelectric Actuator Amplifier Signal Generator Oscilloscope Light Source Camera

Here a is the tension amplitude and 0 is the equilibrium interfacial tension. The phase angle can be estimated by a phase comparison between the variation in area and the response in interfacial tension. The Gibbs interfacial dilational modulus E describes the response of an interface to local compression and expansion, and is defined as the interfacial tension increment per unit fractional change in the interfacial area.
E= d d ln A

(4)

When analyzing the relaxation processes at or near the interface between oil and water, we expect the rheological properties of the interface to be viscoelastic rather than pure elastic. The interfacial dilational modulus is a complex function of the angular frequency of oscillation and has been defined in literature according to Equation 5 [28].
E* = Ed + i d = E '+ iE "

(5)

Cuvette Temp. controlled

Computer & Monitor

Fig.1. Shematic representation of the CAM 200 tensiometer from KSV, modified with a piezoelectric pump mechanically connected to a syringe and a signal generator. The oscilloscope was used to monitor the signal sent to the piezo-pump.

Interfacial Rheology Determination of the rheological properties of an interface can be done with dilational stress where the interfacial resistance to changes in area is monitored. Measurements are made by simultaneously analyzing the interfacial tension and surface area of a drop, whose volume is periodically perturbed. We chose a sinusoidal variation of the surface area of the drop as this function is particularly easy to analyze mathematically because the Fourier transform is an obvious fit.
A = A A0 = Aa sin(t )

Ed is known as the interfacial dilational elasticity whereas d is presented in literature as the interfacial viscosity. The complex dilational modulus E* may be written as the sum of two contributions. An elastic component accounting for the recoverable energy stored in the system (the storage modulus E) and a viscous component accounting for the energy lost through relaxation processes (the loss modulus E). The storage modulus is in phase with the modulation and the loss modulus is 90 degrees out of phase. There have been discussions on the partitioning of the complex modulus into interfacial elasticity and viscosity. These quantities may not correctly define the elasticity and the viscosity of the interface and should only be treated as apparent quantities. We will still report the storage E and loss E moduli from the dilation experiments and treat them as process parameters. By using the Gibbs definition of elasticity (Equation 4) with insertion of the variation in drop area (Equation 2) and the corresponding variation in interfacial tension (Equation 3), the complex interfacial dilational modulus E* can be expressed as:
E* =

sin(t ) cos + a cos(t ) sin = = a ln A A / A0 Aa sin(t ) / A0

cos( t ) sin =| E | cos + i | E | sin = cos + Aa / A0 sin(t )

(6)

(2)

Aa is the area amplitude and A0 is the equilibrium area. The axisymmetric drop shape analysis provided the best fit of the Young-Laplace equation to the drop profile.
= 0 = a sin(t + ) = a sin(t )cos + a cos( t ) sin (3)

In our experiments, a fresh oil droplet (12 L) was formed upwards from the tip of a U-bent needle (1.55 mm) immersed in an optical glass cuvette (Hllma) with saline water (3.5 wt.% NaCl). The volume of the oil drop was modulated (Aa/A0 < 10%) by a piezoelectric actuator (Physik Instrumente PI-840.6) connected mechanically to the syringe (SGE gastight, 25 mL) and to a signal generator (Agilent 33250A). Due to the 4 V limit of the signal generator, we built in a homemade amplifier from a radio tube to provide the piezoelectric positioner with a

Apparent absorbance at 1600 nm

voltage of up to 100 V. A scheme of the instrumental setup is shown in Figure 1. We oscillated the oil drop continuously at a frequency of 0.1 Hz for 2.5 hours while recording images of the drop every second. With the analyzed movie in hand, there are several approaches for the determination of rheological parameters. We wrote a script (Matlab) to calculate rheological parameters from the interfacial tension and area waveforms using Fast-Fourier-Transformations (FFT, 512-points). The motivation for this was that the frequency component corresponding to the modulation could easily be picked out from the spectrum with frequency domain filtering. To keep the time information while transforming to the frequency domain, we subdivided the signal in windows consisting of 9 cycles. The duration of each subdivision is a tradeoff between smoothness (more cycles result in less variance) and time resolution (less cycles gives quicker response). The (n) and the (n+ 1) window were set to overlap with 3 cycles. The differential area and corresponding differential tension were picked out as the frequency responses and were used to calculate the interfacial dilational modulus.
E* = FT (t ) FT(A / A0 )(t )

5 100% H

Apparent absorbance at 1600 nm

95% H 90% H 85% H

80% H

70% H

20% H
0 0 0.1 0.2 0.3 Oil conc. (v/v) 0.4 0.5

100% H

1.5

95% H

(7)

90% H
1

The phase angle was unwrapped from the waveforms and the storage (E) and loss (E) moduli were calculated directly according to Equation 5 and 6. By repeating this procedure for every subdivisions of the waveform we could plot (t), E(t), and E(t) as in Figure 2. The figure shows interfacial tension and rheological parameters for a sample of crude oil B diluted in heptane (Coil = 2x10-4 v/v).
40
40

85% H
0.5

80% H 70% H 20% H

0 0
2

0.1

0.2 0.3 Oil conc. (v/v)

0.4

0.5

100% H

C
Apparent absorbance at 1600 nm
1.5

E'

30

35

90% H
1

Moduli (mN/m)

20
IFT

30

IFT (mN/m)

85% H
0.5

80% H

10
E''

25

10-70% H 0 0
0.1 0.2 0.3 Oil conc. (v/v)

0.4 0.4

0.5 0.5

0 0

0.5

1.0
Time(h)

1.5

2.0

20 2.5

Fig. 2. Results from an interfacial rheology measurement of a diluted sample of crude oil B: Coil = 0.02 v/v, H = 70 vol.%, = 0.1 Hz. Each point represents a Fourier transformation analysis of the area and interfacial tension waveform in a 90 second window.

Fig. 3. Figure a, b, and c show asphaltene stability results of crude oil A, B, and C respectively. The crude oils were diluted to different extents in heptane-toluene mixtures (example: 10 ml of the sample at Coil = 0.1 v/v and 90% H consists of 1 ml crude oil, 8.1 ml heptane and 0.9 ml toluene). The combination of light absorbed and scattered by the sample gives an apparent absorbance at 1600 nm in the NIR range.

RESULTS AND DISCUSSION

Asphaltene Aggregation Induced by Dilution We have used a one-characteristic-wavelength (1600 nm) model to detect asphaltene aggregation in the diluted oil samples with near-infrared spectroscopy. The apparent absorbance at 1600 nm is frequently used as a measure of the contribution from scattering because vibrational features are believed to be minimal at this wavelength. However, the electronic transitions from condensed aromatic ring systems in asphaltenes give rise to a structureless absorption band extending into the NIR range. Thus, Beers law suggests a linear relationship of the absorbance at 1600 nm and the crude oil concentration in a diluted system. Any deviation from this linear relationship between absorbance and crude oil concentration is a result of light scattering from asphaltene particles formed by creating an unfavorable solubility medium. Figure 3 a, b and c show precipitation of asphaltenes from crude oil A, B and C brought about by diluting the crude oil systems with solvent mixtures of low aromaticity. Generally, the first sign of precipitated asphaltenes occurs at 0.3 volume fraction of crude oil. Above this concentration, asphaltenes seem to be solvated by resins and aromatics in the crude oil sample and the absorbance at 1600 nm follows the Beers law (valid up to Coil = 1 v/v although presented only up to 0.5). Moreover, solvent mixtures with low aromaticity precipitate more or larger asphaltene colloids and reach maximum precipitation at higher oil concentrations than solvent mixtures with high aromaticity. Crude oil A, being the oil with more than 10 wt.% asphaltenes, shows considerable more scattering than crude oil B and C at unfavorable solvent conditions. Changes in the state of asphaltenes at different oil concentrations and aromaticity of the solvent should affect the stability of emulsions prepared with these samples and water.
100

100

B
80

0% H 20% H 70% H 80% H 100% H

Water resolved (vol%)

60

40

20

0 0
100

0.2

0.4 0.6 Oil conc. (v/v)

0.8

C
80

0% H 20% H 70% H 80% H 100% H

Water resolved (vol%)

60

40

20

0 0

0.2

0.4 0.6 Oil conc. (v/v)

0.8

A
80
Water resolved (vol%)

0% H 20% H 70% H 80% H 100% H

Emulsion Stability Emulsion stability was determined from the amount of water phase resolved after centrifugation at 3000 rpm (1580 x g) for up to 50 min. To simplify the representation, we will only report graphically the amount of water resolved after 10 min of centrifugation. Figure 4 a, b and c show emulsion stability results for the three crude oils diluted to different extents in heptane-toluene mixtures ranging from pure toluene (0 vol.% H) to pure heptane (100 vol.% H). The abscissa axis shows the crude oil concentration whereas the ordinate axis shows the volume percent of water separated as a free water phase. Starting with sample A in Figure 4a it is clear that there are several properties that influence the stability of these diluted systems. At the lowest oil concentration (0.001 v/v) all systems are relatively unstable and separate completely within the first 10 minutes of centrifugation. Increasing the crude oil concentration up to 0.2 results in more stable emulsions for systems with low aromaticity, whereas systems with high aromaticity reach a maximum stability at lower oil concentration. Up to this point the stability of these emulsions correlates exactly to the state of

60

40

20

0 0

0.2

0.4 0.6 Oil conc. (v/v)

0.8

Fig. 4. Emulsion stability results of crude oil A, B, and C diluted to different extents in heptane-toluene mixtures. The ordinate axis represents the amount of water separated as a free phase after centrifugation for 10 minutes at 3000 rpm (1580 x g).

the asphaltenes as presented in Figure 3a. As expected, asphaltene colloids contribute to increased stability of the emulsions through steric stabilization and probably also by changing the rheology of the water/oil interfaces. If the state of the asphaltenes was the only property determining the stability of the emulsion systems, we would expect the stability to decrease progressively with increasing oil concentration above 0.1-0.2 v/v in line with increased solubility of asphaltenes. However, at oil concentrations above 0.3 the viscosity of the continuous oil phase is considerably increased. Crude oil A is a heavy crude oil with viscosity of 21400 mPas and density of 0.980 g/ml at 20oC. The increased viscosity will significantly reduce the rate of sedimentation of water droplets and the rate of film thinning during coalescence [5]. Resin-stabilized asphaltenes may also provide flocculation-stability to water droplets at high oil concentration according to the model by Goual et al. [24]. We have observed reduced sedimentation rates for concentrated systems as opposed to more dilute systems where water droplets were deposited at the bottom of the centrifugation tube during the first 2 minutes of centrifugation. For oil concentrations above 0.7 v/v water droplets were still dispersed in the viscous oil phase even after 50 minutes of centrifugation. The stability profiles are dominated by two regimes: one at Coil < 0.3 v/v where the stability of these systems seems to be dependent on the colloidal state of asphaltenes, and another regime above this concentration where reduced sedimentation rate contributes to the stability of the systems. After 50 minutes of centrifugation all systems at concentrations above 0.7v/v were stable, whereas the systems diluted in solvent mixtures with low aromaticity were the only stable at concentrations below 0.2 v/v. All other systems were completely separated. Similar features can be observed in Figure 4b for crude oil B. Compared to crude oil A, this crude oil is less viscous and has lower amount of asphaltenes. Thus after 10 minutes of centrifugation, only samples made with low aromatic solvent mixtures are stable in the range below 0.1 v/v crude oil. Reduced rate of sedimentation (due to viscosity or resins) contributes to the emulsion stability at concentrations above 0.7. After 50 minutes of centrifugation, only those dilutions with concentration 0.7 v/v or larger were stable. All other emulsion systems were completely separated. The 19 API crude oil C is slightly heavier than crude oil B and has a viscosity of 447 mPas at 20oC. The crude oil contains 5.0 wt.% of asphaltenes and 20.9 wt.% of resins. Figure 4c shows that the diluted samples of crude oil C are remarkable stable, even more stable than the samples from crude oil A (stability for Coil > 0.5 can not be differentiated). After 50 minutes of centrifugation all samples except the most diluted systems (Coil < 0.1 v/v) with 0% and 20% heptane are more or less stable. The stability to coalescence seems to increase progressively with increased crude oil concentration and decreased aromaticity of the solvent. Although not evident from the stability profiles after 10 minutes of centrifugation, emulsion stability reaches a maximum for 80% H and decrease slightly for 90% and 100% heptane dilutions.

We can add to the information of physicochemical properties of crude oil C in Table 2 that we received this oil as a stable system of water in oil with as much as 40 vol.% percent water. There are apparently components in this oil which very efficiently stabilize water/oil interfaces.
45
2E-5

40

35
IFT (mN/m)

1E-4 2E-4

30
0.001

25

20

0.02

15

0.1- 0.3

10

10

10

10

Time (s)

Fig. 5. Dynamic interfacial tension profiles for crude oil C for concentration from 2*10-5 to 0.3 v/v in a heptane-toluene mixture (80 vol.% H). Fits are according to a multi-exponential decay function.

Kinetics of Film Formation The relaxation of the water/oil interface has been studied in function of the bulk concentration, the aging time and the aromaticity of the system. We will summarize our findings with reference to Figure 5 which displays the dynamic interfacial tension isotherms (t) corresponding to different bulk concentration (Coil) of crude oil C in a heptane/toluene mixture (80 vol.% H). First of all, the relaxation of oil-water interfaces in the presence of asphaltenes has previously [9, 15, 17, 19, 29] as in Figure 5 been shown to be extremely slow. Due to the extremely long time needed for reorganization and progressive building of layers in these systems, conclusions have frequently been based on non-equilibrium conditions. Jeribi et al. [29] have discussed this and referred to studies of cmc (critical micelle concentration) determination of asphaltene-in-toluene systems which are far from thermodynamic equilibrium. Even after the big discussion on the possibility of micelle formation, critical micelle concentrations of asphaltenes are still reported from tension values after only minutes of interfacial aging. For example using the interfacial tension values at 1000 seconds in Figure 5, the apparent cmc-profile ( vs. log C) has no relation at all to the onset of aggregation in the bulk. In fact, as apparent from Figure 5, the limiting interfacial tension seems to be more or less independent on the bulk concentration. Similar features for asphaltene-in-toluene systems have been reported by Moran

et al. [30], Yang and Czarnecki [31] and Jeribi et al. [29]. We know from NIR experiments that asphaltenes tend to aggregate in some of these systems. However interfacial tension experiments cannot be used to determine the onset of this phenomenon in the concentration range from 10-5 to 1 v/v in our system. Moreover, crude oil samples diluted to different aromaticity seem to approach more or less similar equilibrium interfacial tension values. This may suggest that the measured interfacial tension after long times is not that of the oil/water interface but rather that of the film/water interface (Laplacian drops have been obtained for all experiments). The equilibrium interfacial tension is obtained only after several days. We have reviewed the general theory of relaxation processes at liquid-liquid interfaces hoping to understand the molecular mechanisms by which tension is lowered in Figure 5. Successive or overlapping relaxation steps may be involved in the equilibration process: diffusion, adsorption barriers, structural reorganization at the interface, desorption, and mass transfer to the opposite phase. To our knowledge the complete relaxation of the oil/water interface cannot be described quantitatively by one equation alone. However, by assuming certain processes to be at equilibrium with respect to the rate determining mechanism, the relaxation can be described with reference to kinetic theories of the rate determining process. When the rate determining adsorption process is diffusion controlled, a plot of versus t0.5 should give a linear relationship according to the equation of Ward and Torda [32] for the initial step of the adsorption process.

Interfacial Rheology of Viscous Samples We have performed oscillation experiments of viscosity standards to validate our experimental setup and find a Capillary Number (Equation 1) where we can safely neglect the influence of viscous forces. The viscosity standards ranged in viscosity from 9.4 to 1276 mPas at 25oC and showed constant interfacial tension over time, indicating clean water/oil interfaces. However, when oscillated at high frequency, high volume amplitude or when supported by narrow needles, the viscous forces (bulk) distorted the drop and resulted in apparent interfacial viscoelasticity. Figure 6 shows the result of a number of experiments where parameters in Equation 1 were varied. The inset in Figure 6 shows experimental points with Ca < 0.01. The results support the work by Freer et al. [22] and confirm that viscous forces are the origin of the apparent viscoelasticity and the linear relationship between /0 and Ca. Noise in the measured interfacial tension for a static drop has been determined to be approximately a/0 = 0.0030, which means that the contribution from viscous forces must be above 0.0030 to be distinguished from experimental noise. This corresponds to Ca > 0.0021 (arrow in figure 6) or a viscosity of 60 mPas given Equation 1 and V = 1 L, = 0.1 Hz, 0 = 20 mN/m, water = 1 mPas, and a = 1.55/2 mm. Thus, to analyze crude oil samples with viscosity higher than 60 mPas the capillary size must be increased or the volume amplitude reduced.
0.12

0.10

0 (t ) C Dt

(8)
0.08

The parameters C and D represent the bulk concentration and monomer diffusion coefficient of the surfactant, and 0 is the interfacial tension of the pure interface. For the results presented in Figure 5, Equation 8 may be fulfilled for very low crude oil concentrations, but could not be used at higher concentrations where the initial diffusion process was finished during drop formation. Jeribi et al. [29], Freer and Radke [9], and Sztukowski and Yarranton [16] have all estimated characteristic times of diffusion (seconds) of asphaltenes which are much smaller than the experimental time of relaxation to equilibrium (days). It is very likely that the rate determining process causing the tension lowering in Figure 5 is not bulk diffusion but rather a reorganization of the interfacial layer. Relaxation of asphaltene films has been compared to the relaxation of other macromolecular layers due to the similarities in the interfacial tension profiles during film formation. There have been several attempts in explaining the reorganization process [17, 19, 29, 3335]. Due to the complicity of the relaxation it is without the scope of this study to explain theoretically the macromolecular reorganization in interfacial layers. However, we think that the understanding of crude oil/water interfaces could be increased by modeling the interfacial tension relaxation.

a / 0

0.06

0.04

0.02

0.01

0.02

0.03 0.04 Capillary Number

0.05

0.06

0.07

Fig. 6. Influence of bulk viscous forces on the relative interfacial tension amplitude measured with oscillation drop technique. The Capillary Number (equation 1) was altered by changing the viscosity, capillary size, frequency and volume amplitude of the experiment. The inset shows values for Ca < 0.01. The regression line has been forced to intercept both axes at zero.

Viscoelastic Properties of the Interface The structure formation of crude oil components within the adsorption layers has been studied with two-dimensional

Modulus (mN/m)

rheology. The results have been interpreted on the basis of previous studies [9, 15, 16, 18, 20, 21, 28, 34, 36] and the features of interfacial relaxation, studied with dynamic tension experiments (Figure 5). Figure 7 a, b, and c show the variation of the near-equilibrium (2.5 h) storage E and loss E moduli as a function of the bulk crude oil concentration and the aromaticity of the solvent. The measurements were carried out with a constant applied frequency = 0.1 Hz. We have monitored the rheological properties during the aging process. Except for some experiments at low Coil (indicated by arrows), both moduli achieved more or less constant values during the time of the experiment (2.5 h). For comparison Goual et al. [24] studied the adsorption of bitumen components at o/w interfaces. They found that the adsorption kinetics followed two regimes depending on the bitumen concentration: one steady-state regime at high bitumen concentrations and one non-steady-state regime at low concentrations where a continuous interfacial build-up resulted re in multilayer adsorption. Generally, Figure 7 shows similar rheological features for crude oil A, B, and C. (1) The module E is superior to the E for all experiments, indicating solid-like properties of the interface. (2) The interfaces were apparently most viscoelastic in systems with low aromaticity, as also observed for interfaces of only asphaltenes [18]. (3) A maximum in E occurs at low oil concentrations which is directly correlated to observations of crumpled interfaces [37]. The maximum is shifted to higher oil concentrations in aromatic solvents. (4) A further increase in bulk concentration significantly decreases the storage modulus E.
35

40 35 30 Modulus (mN/m) 25 20 15 10 5 0 -6 10

E' 0%H E'' 0% H E' 70% H E'' 70% H E' 100% H E'' 100% H

10

-5

10

-4

10 Oil conc. (v/v)

-3

10

-2

10

-1

10

40 35 30 25 20 15 10

E' 0% H E'' 0% H E' 80% H E'' 80% H

A
30 25 Modulus (mN/m) 20 15 10 5 0 -6 10

E' 0% H E'' 0% H E' 80% H E'' 80% H

5 0 -6 10

10

-5

10

-4

10 Oil conc. (v/v)

-3

10

-2

10

-1

10

10

-5

10

-4

10 Oil conc. (v/v)

-3

10

-2

10

-1

10

Fig. 7. Figure a, b, and c show the near-equilibrium (2.5 h) storage E and loss E moduli as a function of the bulk crude oil concentration and the aromaticity of the solvent for crude oil A, B, and C respectively. The crude oils were diluted to different extents in heptane-toluene mixtures as indicated in the legend (vol.% H). The measurements were carried out with a constant applied frequency = 0.1 Hz.

For a one-component system with limited adsorption an increase in the bulk concentration can affect the interfacial dilational viscoelasticity in two ways: by increasing the interfacial concentration and by increasing the rate of diffusion of interfacially active molecules diffusing from the bulk to a newly created interfacial segment [16, 38]. At low bulk concentrations the diffusion flux of molecules from bulk to the interface is low (Equation 8). An increase in interfacial concentration is accompanied by an increase in tension gradients along the interface, which results in higher storage modulus. Higher interfacial concentration may also increase the number of relaxation mechanisms (e.g. reorganization) which will affect the interfacial loss module E. However at some point the interface reaches saturation and a further increase in bulk concentration can only increase the diffusion flux of molecules to the interface, thereby decreasing the interfacial tension gradients that result from interface deformation. E and E decreases significantly.

Our diluted crude oil systems contain asphaltene and resin molecules (constant ratio) which are both known to be interfacially active. They both have the ability to decrease the interfacial tension gradients during dilation, according to the theory of increased diffusion flux at higher concentrations. We are faced with a rather complex system with at least two groups of interfacially active components with different adsorption characteristics. In reality these two solubility classes may include thousands of different chemical compounds. Czarnecki and Moran [23] have presented a model on the stabilization mechanism of w/o interfaces based on competition between subfractions within asphaltenes and resins. The competition for the interface is assumed to be determined by the difference in adsorption rate and reversibility rather than by differences in adsorption energies. Asphaltenes are known to adsorb at the o/w interface through a slow, irreversible process whereas resins adsorb quickly, reaching equilibrium adsorption. Czarnecki and Moran [23] have explained the dramatic difference in the composition of the interfacial material in dilute and concentrated systems with blockage of the interface by compounds within the resin fraction. Decreasing phase angle = tan-1(E/E) with decreasing concentration Coil in Figure 7 may indicate that the solid-like behavior of the adsorption layer becomes more pronounced. We cannot safely rule out the possibility that compounds within the resin fraction make up a significant amount of the interfacial material at high crude oil concentrations. However, a maximum in the apparent elasticity at low concentrations (Coil ~ 10-4 v/v in Figure 7a) has also been reported for solvent mixtures with asphaltenes by Sztukowski and Yarranton [16] or for other single component systems [21, 34, 36] where diminished apparent viscoelasticity at high concentration cannot be attributed to competition from a second component (resin). We recognize that only a sub-fraction of the asphaltenes may be responsible for the stability of the interface. Thus, the influence of increased diffusion flux at high bulk concentrations will certainly contribute to the dilational properties of the crude-oil water interfaces. Sztukowski and Yarranton [16] and Freer and Radke [9] have attempted to model the viscoelasticity of asphaltene films. Solvent properties have been given much attention in determining asphaltene adsorption [13]. Increasing Coil in a diluted system with low aromaticity will also increase the aromaticity of the system and the solubility of asphaltenes in bulk since the crude oil properties dominate the solution. Generally, the difference in E for systems diluted with solvents of high and low aromaticity was smaller at high Coil that at low Coil. We have shown that the aggregation state of asphaltenes is not significantly influenced by the aromaticity of the solvent at high crude oil concentrations (Coil > 0.3 v/v in Figure 3). It is tempting to explain the increased storage modulus, E, with the sudden precipitation of asphaltenes at the critical dilution. However, we do not know if the bulk state of asphaltenes will affect the interfacial composition or viscoelasticity.

Altogether we believe that both increased diffusion flux and a change in composition of the interfacial material may be responsible for the decreasing E at high bulk concentrations. The interface has apparently been liquefied.
40
E' 10(-4)

35 30
Modulus (mN/m)

E'' 10(-4) E' 0.02 E'' 0.02

25 20 15 10 5 0

10

-3

10 Frequency (Hz)

-2

10

-1

10

Fig. 8. The frequency response of E() and E() for two systems of crude oil B (80% H) (Coil = 10-4 and 0.02 v/v). Measurements were performed after aging the interface for 2.5 hours.

The rheological parameters of the system have been compared using a dilation frequency (0.1 Hz) allowing certain relaxation processes to take place. We have enlarged the applied frequency range for two systems of crude oil B (80% H) in order to prove that the conclusions drawn from the one-characteristic-frequency model (Figure 7) are valid for a broader range. Figure 8 displays the variation of E() and E() as a function of the frequency for one low concentration (Coil = 10-4 v/v) and one high concentration (Coil = 0.02 v/v) system. Measurements were performed after aging the interface for 2.5 hours. To prove that the interfacial layer did not change appreciably during the experiment, we performed measurements at = 0.1 Hz as the first and last measurement. The rheological parameters of the interface did not change significantly during the course of the experiment. First, the storage modulus at low Coil was greater than at higher concentration as expected from the measurements at = 0.1 Hz. At low Coil the response to the interface dilation was mainly elastic in the frequency range from 10-3.25 to 1 Hz. The characteristic frequency of various relaxation processes occurring at and near the interface (diffusion and reorganization) was smaller than the experimental frequencies, giving the interface the character of an insoluble film. At low experimental frequencies some relaxation mechanisms contributed to the viscous component whereas at high frequencies these were too slow. At higher Coil (0.02 v/v) the response to interface dilation was still mainly elastic. However the characteristic time of relaxation mechanisms were now orders of magnitude smaller

than for the low bulk concentration experiment. The interface was apparently much less solid-like. It is clear that the character of the aged film is dependent on the access to interfacially active components from the bulk. Freer et al. [9] cleverly demonstrated this phenomenon for an asphaltene film at the interface between toluene and water. The aged interface was characterized with interfacial rheology both before and after washout of bulk components. When asphaltenes in the bulk solution surrounding the aged water droplet were removed, the interface appeared as much more solid-like with characteristic time an order of magnitude larger than that observed prior to washout. Moreover the response in storage modulus E was significantly narrowed and the broad relaxation (also apparent in E) at higher frequencies disappeared. How Can We Imitate a Real w/o Emulsion System? The previous discussion and results have clearly showed the influence of bulk and interfacial concentration of crude oil components for the rheological parameters of oil/water interfaces. In addition to changing the solubility of asphaltenes, alteration of the crude oil concentration with a solvent will significantly affect the characteristics of the interface. We recognize the difficulty in separating the parameters contributing to increased stability of the emulsion systems as the amount and state of heavy crude oil components will influence several interfacial properties (steric and rheological) simultaneously. Moreover, the association state of heavy crude oil components may not be identical in bulk and at interfaces [13]. As the main objective is to understand the contributions from different physico-chemical properties to the overall stability of crude-oilwater systems, it is of great importance to perform the measurements at experimental conditions as close as possible to those where emulsion stability is measured. Our opinion is that in order to separate the rheological properties of the interface from other properties of the system, it is necessary to measure interfacial rheology on simple systems (single interfaces). Whether real conditions are best imitated by dilation, elongation or shear experiments will not be treated here. Neither will we conclude on the importance of interfacial viscoelasticity for determining the stability of oil/water-interfaces. With the results reported previously and from this study we know about the effect of aging time, bulk concentration, interfacial concentration, bulk viscosity [22], experimental frequency [9, 16], and pH of the water phase [17] when characterizing the rheological properties of heavy oil-water interfaces. We leave it to following studies to decide which experimental conditions to be used in order to mimic a certain emulsion system.

properties. The following main conclusions could be formulated from our study: (1) The overall emulsion stability seemed to be strongly dependent on the aggregation state of asphaltenes in the bulk solution and the reduced sedimentation of water droplets in concentrated systems. We have not determined the individual effect of interfacial rheology. (2) The relaxation of the oil/water interfaces seem to occur under diffusion control at low surface coverage (short time), but under mixed (diffusionkinetic) control at large surface coverage (long periods of time). This phenomenon is probably related to conformational changes within the adsorption layer, but need further studies. To our knowledge, cmc of asphaltenes cannot be determined from dynamic interfacial tension experiments (pendant drop) as the equilibrium interfacial tension appears to be more or less independent on bulk concentration of asphaltenes. Conclusions should not be based on nonequilibrium systems. (3) Measurements of interfacial rheology with dilation are limited by inertia of the system. (4) The dilational moduli depend on the ratio of the enforced dilation frequency to the characteristic frequencies of the relaxation processes (diffusion, structural reorganization etc.). In all the rheological experiments the applied frequency = 0.1 Hz was superior to the characteristic frequency of the ratedetermining relaxation process. (5) Reduced bulk concentration gives apparently more solidlike character (E > E) of the film. We have showed that this phenomenon is related at least partly to the reduced diffusion flux of molecules from the bulk to the interface. Therefore, as the property does not originate exclusively from interactions within the adsorption layer, it is not the true viscoelastic character of the interface, but rather a property of the system.

ACKNOWLEDGEMENT
The technology program Flucha III, Particle-Stabilized Emulsions/Heavy Crude Oils, financed by industry and the Norwegian Research Council are gratefully acknowledged for financial support.

REFERENCES
[1] [2] [3] [4] A. Sanire, I. Hnaut, J-F. Argillier, Oil and Gas Science and Technology - Rev.IFP 59 (2004) 455. D. L. Greene, J. L. Hopson, J. Li, Energy Policy (In Press) X. A. Wu, J. Czarnecki, Energy and Fuels 19 (2005) 1353. X. Wu, Energy and Fuels 17 (2003) 179.

CONCLUSIONS
The dynamic interfacial tension and dilational rheology experiments have been applied to study the adsorption layers of crude oil components at the oil-water interface. Asphaltene aggregation state experiments (NIR) and emulsion stability experiments have complemented the studies of interfacial

[5]

[6] [7] [8] [9] [10] [11] [12] [13]

[14] [15] [16] [17] [18]

[19] [20]

[21] [22] [23] [24] [25] [26] [27]

[28] [29]

P. V. Hemmingsen, A. Silset, A. Hannisdal, J. Sjblom, Journal of Dispersion Science and Engineering 26 (2005) 615. J. G. Speight, The Chemistry and Technology of Petroleum. 3 ed. 1998: Marcel Dekker, Inc. 875. J. D. McLean, P. K. Kilpatrick, Journal of Colloid and Interface Science 196 (1997) 23. J. D. McLean, P. K. Kilpatrick, Journal of Colloid and Interface Science 189 (1997) 242. E. M. Freer, C. J. Radke, Journal of Adhesion 80 (2004) 481. V. B. Menon, D. T. Wasan, Colloids and Surfaces 23 (1987) 353. D. E. Tambe, M. M. Sharma, Journal of Colloid and Interface Science 157 (1993) 244. B. P. Binks, Current Opinion in Colloid and Interface Science 7 (2002) 21. P. K. Kilpatrick, P. M. Speicker, Asphaltene Emulsions, in Encyclopedic Handbook of Emulsion Technology, J. Sjblom, Editor. 2001, Marcel Dekker, Inc.: Trondheim. p. 707. H. W. Yarranton, H. Hussein, J. H. Masliyah, Journal of Colloid and Interface Science 228 (2000) 52. E. M. Freer, T. Svitova, C. J. Radke, 39 (2003) D. Sztukowski, H. W. Yarranton, Langmuir 21 (2005) 11651. S. Poteau, J-F. Argillier, D. Langevin, Energy and Fuels 19 (2005) 1337. N. Aske, R. Orr, J. Sjoblom, H. Kallevik, G. Oye, Journal of Dispersion Science and Technology 25 (2004) 263. F. Bauget, D. Langevin, R. Lenormand, Journal of Colloid and Interface Science 239 (2001) 501. T. Sun, L. Zhang, Y. Wang, S. Zhao, B. Peng, M. Li, J. Yu, Journal of Colloid and Interface Science 255 (2002) 241. Y. Wang, L. Zhang, T. Sun, S. Zhao, J. Yu, Journal of Colloid and Interface Science 270 (2004) 163. E. M. Freer, H. Wong, C. J. Radke, Journal of Colloid and Interface Science 282 (2005) 128. J. Czarnecki, K. Moran, Energy and Fuels 19 (2005) 2074. L. Goual, G. Horvth-Szab, H. Masliyah, Z. H. Xu, Langmuir 21 (2005) 8278. A. Hannisdal, P. V. Hemmingsen, J. Sjblom, Industrial & Engineering Chemistry Research 44 (2005) 1349. O. C. Mullins, Analytical Chemistry 62 (1990) 508. O. C. Mullins, N. B. Joshi, H. Groenzin, T. Daigle, C. Crowell, M. T. Joseph, A. Jamaluddin, Applied Spectroscopy 54 (2000) 624. R. Myrvold, F. K. Hansen, Journal of Colloid and Interface Science 207 (1998) 97. M. Jeribi, B. Almir-Assad, D. Langevin, I. Hnaut, J-F. Argillier, Journal of Colloid and Interface Science 256 (2002) 268.

[30] [31]

[32] [33] [34]

[35] [36]

[37]

[38]

K. Moran, A. Yeung, X. TYang, J. Czarnecki, CIChE Conference, Vancouver, BC (2002) X. Yang, J. Czarnecki, Colloids and Surfaces A: Physicochemical and Engineering Aspects 211 (2002) 213. A. F. H. Ward, L. Torda, Journal of Chemical Physics 14 (1946) 453. G. Serrien, G. Geeraerts, L. Ghosh, P. Joos, Colloids and Surfaces 68 (1992) 219. V. G. Babak, J. Desbrires, V. E. Tikhonov, Colloids and Surfaces A: Physicochemical and Engineering Aspects 255 (2005) 119. C. J. Beverung, C. J. Radke, H. W. Blanch, Biophysical Chemistry 81 (1999) 59. E. H. Lucassen-Reynders, A. Cagna, J. Lucassen, Colloids and Surfaces A: Physicochemical and Engineering Aspects 186 (2001) 63. C. Tsamantakis, J. Masliyah, A. Yeung, T. Gentzis, Journal of Colloid and Interface Science Article in Press (2005) J. Lucassen, M. Van Den Tempel, Chemical Engineering Science 27 (1972) 1283.

Paper 4

Viscoelastic Properties of Crude Oil Components at Oil-Water Interfaces. 2: Comparison of 30 Oils.

Andreas Hannisdal a,*, Robert Orr b, and Johan Sjblom a
Ugelstad Laboratory, Department of Chemical Engineering, Norwegian University of Science and Technology (NTNU), N-7491 Trondheim, Norway. b Norsk Hydro ASA, Research Park, N-3915 Porsgrunn, Norway.
a

The dilational properties of diluted (0.7 vol/vol in toluene) and undiluted crude oil-water interfaces have been studied using the oscillation drop method with the objective of understanding the properties contributing to the overall stability of crude oil emulsions. The importance of working with undiluted crude oils instead of model systems when dilational properties of real oil-water systems are going to be reproduced in the laboratory setting has been discussed. For such studies, molecular exchange mechanisms and the aggregation state of asphaltenes are too dependent on concentration to justify the use of model compounds, i.e. fractionated asphaltenes diluted in a solvent. As expected in the low frequency range (0.01-1 Hz), molecular exchange from the bulk oil phase strongly affected the measured dilational parameters. For the diluted crude oils, the frequency dependence of the dilational modulus increased with its magnitude. The systems which exhibited particularly low magnitude of the dilational modulus were of the heaviest crude oils in the sample set, whereas the systems with greatest dilational modulus were among the lightest crude oils. The overall characteristic time of relaxation of the crude oil-water interfaces was in the range below 10 sec. The undiluted crude oil-water interfaces had similar interfacial properties as the diluted samples except for slightly reduced magnitude of the dilational modulus. The crude oil-water interfaces appeared to be soluble, but some observations pointed to intrinsic rheological properties of the interfaces. Intrinsic elasticity and viscosity of the films should be studied outside the range used here at low ( ~ 0 Hz) and high ( 500 Hz) frequencies, respectively. Keywords Crude oil; Kinetics of film formation; Dilational viscoelasticity; Molecular exchange

INTRODUCTION
Although the stabilization of petroleum emulsions is not completely understood, it is well recognized that the formation of viscoelastic films at the interface between crude oil and water plays an important role in the stabilization mechanism. Indigenous crude oil components, in particular from the heavy end of petroleum (asphaltenes and resins), have high affinity for the water-oil interface and build up a film material of high mechanical strength. For general surfactant stabilized systems, theoretical models of colliding interfaces have taken into account the mobility of surfactants (along the interface and from the sub-phases) and explained the phenomena present during film-thinning and rupture [1-5]. The importance of dilational viscoelastic properties of the interface is confirmed by these studies. However, in the chemically and structurally complex petroleum systems, the composition and structure of the interfacial films is not yet fully grasped. Phenomenological studies have concluded on the viscoelastic properties of watermodel oil interfaces [6-14]. The components which are believed to be most essential for the film rheology mechanism is extracted from the oil and diluted in a hydrocarbon, generally a mixture of toluene and an alkane. Based on this, it is

fundamentally important to question whether these systems are representative for the constituents in the film material and whether the contribution from the bulk phases is correctly mimicked. We acknowledge the difficulties in reproducing the real scenario in a laboratory setting. Reproducing high pressure, high viscosity, choosing reasonable dilation frequencies and mass transport from bulk phases is a formidable task. Moreover it is well known that any dilution or extraction of crude oil fractions will disturb the delicate balance between components within the oil [15]. The introduction of the Petroleomics concept has solved certain aspects with regard to the molecular structure and chemical composition of asphaltenes [16-19]. These are per definition the most polar compounds in crude oil and have a tendency to self-associate and precipitate in unfavorable conditions. Asphaltenes self-assemble in toluene solution above a critical concentration [16] and form even larger agglomerates which precipitate in toluene/alkane mixtures. Such particleformation will not only completely change the stabilization mechanism of the emulsion but also the rheological properties of the interface. Rheological characteristics of diluted heavy crude oil-water interfaces have been reported previously [20]. Until an understanding of the degree of asphaltene aggregation in crude oil is obtained (it may be very different for crude oils), the

rheological properties of the water-oil interface should be characterized for undiluted crude oils if possible. These are after all the samples that we want to characterize with respect to other physicochemical properties and their stabilizing properties for w/o emulsion systems. The overall objective is to be able to understand which physicochemical properties are the most important for the stability of these emulsions. 30 heavy or particle-rich crude oils have been characterized with respect to bulk and interfacial properties. Here we will report on the interfacial properties, interfacial tension and interfacial rheological parameters. Dilational viscoelastic parameters have been measured with an oscillating pendant drop technique in combination with axisymmetric drop shape analysis. Due to the viscous nature of the heavy crude oils, some of these had to be diluted to avoid inertial problems which will affect the drop shape analysis. Other oils could be analyzed without dilution. The advantages and limitations of the experimental setup and alternative measurement methods have been considered. Some examples of reported studies are given here. The rheological properties of crude oil-water interfaces and model oil-water interfaces have been compared. Moreover, we have discussed the practical signification of the measured parameters and its relevance for the stability of petroleum emulsions.

from inertial phenomena (a-b). When the drop was left at constant volume it gradually relaxed back to a Laplacian shape (b-c). Inertial phenomena were also obvious when the drop volume was reduced (c-d). Table 1. Physicochemical properties of the crude oils.
Property Saturates Aromatics Resins C6 insolubles a Toluene insolubles b Density c Acidity (TAN) Viscosity Unit Max Min Median Average (wt.%) 63.7 25.9 42.1 43.2 (wt.%) 53.1 28.6 39.3 39.1 (wt.%) 28.5 6.2 13.2 13.6 (wt.%) 12.9 0.1 2.2 2.9 (wt.%) 0.18 0.01 0.02 0.03 (g/ml) 15oC 0.996 0.857 0.929 0.925 (mg KOH/g) 7.5 ~0 1.3 1.8 (mPas) 25oC 190000 7.7 108.0 8060 (mPas) 55oC 4590 2.9 34.7 285 MW d (g/mol) 535 204 333 333 a Asphaltene precipitation in n-hexane (1g:40ml), HPLC separation of maltenes according to Hannisdal et al. [21]. b ASTM D4807-88, 80oC c AP PAAR DMA 48 d Determined from the freezing point depression of benzene

EXPERIMENTAL
Crude Oil Characteristics and Sample Preparation The 30 crude oils were received from 9 different oil companies with production sites on the Norwegian Continental Shelf, in the South China Sea, the Gulf of Mexico, United Kingdom, France, Brazil, West-Africa and Alaska. All crude oils were sampled in a way that minimized contamination by production chemicals. About half of the oils were per definition heavy (< 20oAPI) whereas the lightest sample was a 33.6oAPI crude oil. Other characteristics are reported in Table 1. Some samples were expected to be waxy and showed Bingham plastic behavior below 25oC, typical for wax crystallization [22]. Except for two crude oils, the crude oils were generally stable with respect to asphaltene flocculation (titration with pentane monitored with light scattering in the near-infrared region). At least one of these two samples (oil no. 30) was produced from very deep water with asphaltene content of 3.3 wt.% and resin content of 7.6 wt.%. When emulsified with water, the system formed a network of asphaltene particles and water (visible by microscopy) which behaved as a gel (oscillation bulk rheology). Both crude oils were light (~ 30oAPI). In Table 1, the viscosities of the crude oils are reported at 25oC. The high viscosity of the samples made it impossible to measure interfacial dilation viscoelastic parameters due to inertia [23] of the systems. The working range for our dilational drop instrument has been reported previously [20]. With a conservative approximation, the limit where inertial effects could safely be neglected was 60 mPas at a dilation frequency of 0.1 Hz and a capillary diameter of 1.55 mm. An extremely exaggerated example of inertia is given in Figure 1 of the heaviest oil in the sample set. When the viscous oil drop was expanded from a capillary, non Laplacian drop profiles resulted

To reduce the viscosity difference between the water and oil phases, the rheological experiments were performed at 55oC. The temperature increase obviously caused a significant reduction in oil viscosity. However for 17 of the 30 oils the viscosity was still too high to perform a dilational study at greater frequencies than 0.1 Hz. With this in mind the crude oils were diluted with 30 vol.% solvent. For reasons to become clear later, toluene was used as diluent. We know for certain that the delicate balance of compounds in the crude oil system may be altered. With reference in previous studies on the effect of dilution [20], we are still confident that the difference in interfacial characteristics will be measurable. Differences between diluted and undiluted samples will be compared. The diluted crude oils were stored for several weeks prior to being analyzed. The diluted crude oil densities were measured at 55oC to correctly account for buoyancy forces of the pendant oil drop. Milli-Q quality water, buffered to pH 6.5 and with a salinity of 3.5 wt.% NaCl, was used for all measurements (1.013 g/ml at 55oC) .

(a)

(b)

(c)

(d)

Fig. 1. Dilation of a crude oil drop with extreme viscosity from a capillary in water. Inertial effects complicate the experiments.

Dynamic Interfacial Tension Measurements Static Drop The drop tensiometer (CAM 200 KSV Instruments, Finland) was used to measure the interfacial tension by analyzing the axial symmetric shape (Laplacian profile) of a rising oil drop in water (3.5 wt.% NaCl). All the measurements were performed at controlled temperature (25oC). An oil drop were created from a needle connected to a syringe (1mL). The syringe was equipped

with a valve to close off the sample volume in the needle from that in the syringe. This is very important in order to keep a constant volume of the drop during long-time experiments. If sample climbs the Teflon plunger, the drop volume will decrease slightly. Due to the irreversible nature of asphaltene adsorption and film formation, a reduction of the drop volume will lead to increased surface pressure and significant measurement errors. Images were recorded continuously at a rate of 2 images per minute during 12 hours after creating the oil-water interface. Reproduced experiments (103 experiments in total) indicated the mean variance of the interfacial tension determination. Interfacial Dilational Rheology The instrumental setup was similar to that reported previously [20]. Shortly, a drop is modulated with a piezoelectric actuator, connected mechanically to a syringe (SGE gastight, 25 mL) and driven by a waveform generator (Agilent 33250A) with a homebuilt amplifier. The piezoelectric actuator (Physik Instrumente PI-844.6) has a push/pull force capacity of 3000/700 N and a resolution of 9 which makes it useful for our application. A flexible tip (P-176.6) was used to protect the actuator from accidental bending forces. All rheological measurements were performed at a controlled temperature of 55oC. The oil drop was aged without dilation for 1.5 hour prior to being modulated in a sinusoidal waveform in the frequency range between 0.011 Hz. Lower oscillation frequencies are accessible given that the interface has reached near-equilibrium. Higher oscillation frequencies are easily accessible from the instrumental setup, but introduce inertial problems for these viscous systems. In this frequency range, a satisfactory description of the dilational moduli may be realized by using only some discrete frequencies, here log = {-2, -1, -0.5, 1} Hz. The magnitude of the dilational elasticity modulus |E|=A0d/dlnA, the phase angle () between area (A) and interfacial tension (), and the in-phase (E=|E|cos) and out-ofphase (E=|E|sin) moduli have been reported for these frequencies. A0 is the equilibrium area of the drop and E and E are also referred to as storage and loss moduli, respectively. The oscillation amplitude was changed to verify that the measurements were in the linear regime. The parameters characterize the elastic and viscous properties of the interface as well as molecular exchange processes between the interface and the bulk. Even for the diluted samples (viscosities less than 7.7 mPas) we were at high frequencies dangerously close to the condition where inertial effects started to affect the measured viscoelastic parameters. We have therefore, at all times carefully kept track of the capillary number [20, 23] and increased the capillary size or reduced the modulation amplitude when necessary. The drop tensiometer (CAM 200 KSV Instruments, Finland) was used to capture images of the drop (20 images per cycle) and determine the interfacial tension and area from an iterative axisymmetric drop shape analysis. A 3.2 GHz Intel Pentium processor calculates the parameters at a rate of 4 images/sec (can be run in parallel with other operations). The viscoelastic parameters were estimated with a Fourier Transformation in Matlab 7.0 as described previously [20].

30

25

20 IFT (mN/m)

15

10 30 5

0 2 10

10

10 Time (sec)

10

Fig. 2. Dynamic interfacial tension of some selected crude oilwater interfaces at 25oC. Crude oil 30 shows peculiar interfacial tension profile.

RESULTS AND DISCUSSION

Interfacial Relaxation Dynamic interfacial tension measurements were performed in order to determine the close-to-equilibrium interfacial tension of the water/oil interface. Dynamic tensiometry has been used previously for crude oil systems to understand the adsorption mechanisms and kinetics [6, 12, 24-27]. To our best knowledge, the processes responsible for the interfacial tension decay have not been completely understood yet. However certain characteristics of the decay function have been identified. The relaxation is extremely slow. Equilibrium values are only reached after several days or weeks in a static experiment. After an induction period, the relaxation seems to deviate from a diffusion controlled adsorption mechanisms. Interfacial reactions, i.e. reorganization of molecules, are believed to be responsible for this feature. Such reorganization of macromolecules is well known in food science from adsorption of proteins [5, 28-30] or in polymer science from adsorption of polyoxyethylenated surfactants [31]. For nearly all static drop experiments, the relaxation seemed to follow a logarithmical decay function. Since the immediate relaxation could not be measured, the diffusion-controlled adsorption process may already be finished for these concentrated solutions when the experiments were started. Dynamic interfacial tension experiments for more dilute crude oils have also captured the diffusion process [20]. Figure 2 shows the dynamic interfacial tension for some selected crude oils on a semi-logarithmic scale. Due to some variation in the aging time when an experiment was started, the interfacial tension results have been reported from 100 seconds. The first striking effect is the linear relationship between the interfacial

tension and logarithmic time. Moreover the slope is very similar for these selected oils. Not all reproductions gave the same slope as showed in Figure 2. Some oils gave slightly different slopes, the extreme viscous oils relaxed to Laplace shape during the first 30 minutes due to inertial effects, and one oil showed a minimum in the IFT profile. We do not know if the decay functions appear to be similar just due to the fact that the major interfacial tension reduction is already finished. We still think the observation is noteworthy. The oil with a minimum (oil no. 30) had very low interfacial tension for crude oils, possibly due to acids. We have seen similar dynamic interfacial tension profiles previously for crude oil systems with water soluble acids (dissociated at high pH). Poteau et al. [10] have reported on the influence of pH on the interfacial properties of crude oil-water systems. Overall the near equilibrium (12 hours) interfacial tension results gave heteroscedastic distributed variances with a mean variance for all experiments of 3 mN/m (the worst sample variance was 12 mN/m). This corresponds to a mean standard deviation of 2 mN/m. Figure 3 shows the near equilibrium (12 h) interfacial tension results from static experiments with those from dilation studies of the corresponding crude oils, undiluted (triangles) or diluted to 70 vol.% in toluene (circles). Sample reproducibility is indicated with error bars of one standard deviation at each side of the average value. The interfacial tension for undiluted crude oils is in the range between 8 and 24 mN/m at 25oC (sample 30 excluded).
30 0.7 vol/vol 25 Undiluted

properties in the laboratory scale. However, unlike its shear counterpart, interfacial viscoelasticity does not have exact definition in literature [32]. The particular model used for interpretation of the data obtained from dilational experiments may therefore influence the estimated viscoelastic properties. Lucassen and van den Tempel (LVDT) [33, 34] introduced the complex elasticity with a real (Gr) and imaginary (Gi) part which were later interpreted as the interfacial elasticity and viscosity moduli, respectively. However, Ivanov et al. [32] stated that, based on theoretical approaches, the only correct definitions of interfacial elasticity and viscosity (apparent, because of its dependence on bulk diffusion) for an oscillating plane interface (big drop) were the Maxwells equations. The Maxwell model describes the process of deformation by an elastic and viscous element connected in series (The LVDT approach is however consistent with Voights model where the elements are connected in parallel). According to Ivanov the LVDT parameters can be used for data processing, but have no parallelism in real interfacial properties. Although the theory of reduced film drainage (Marangoni effect and Boussinesq effect) [3, 4] and damping of interfacial corrugations by the viscoelastic property of the interfacial film is well studied, the exact prediction of such phenomena with rheological parameters from dilation experiments should be done with caution. Fruhner et. al. [35] compared the dilational rheology data with measurements of foam stability in the 1-500 Hz range and concluded on the extraordinary importance of the surface intrinsic dilational viscosity on the foam stability. According to Ivanov et al. [32] any attempts to correlate Gi with lifetime of emulsions or foams based on the idea of reduced film drainage have only some parallelism due to the fact that both Gi and the disjointing pressure are caused by surfactant adsorption. Morover, Ivanov states that on closer examination, the processes (film drainage and interfacial dilation) are quite different and have, apart from their identical origin, nothing else in common. For petroleum systems the formation of a network-like film of asphaltene/resin complexes has been identified as an important stabilization mechanism and directly correlated to the elastic modulus from dilational rheology [7, 11, 13, 36]. The fact that such network-formation is also accompanied with the presence of sterically demanding units has been disregarded by some authors. For protein or asphaltene films the steric component of the disjointing pressure has to be taken into account. Previous studies (Hannisdal et al. [37] and references herein) have showed that particle-stabilized emulsions can have extreme stability against coalescence due to the energy of attachment of the stabilizing units (intermediate hydrophobicity like asphaltenes) at the water-oil interface. Thus, to simply rely on an experimental method that is not able to distinguish between rheological and steric phenomena when proving the correlation of the viscoelastic moduli to emulsion stability is not sufficient. With this in mind it appears even more important to work on undiluted systems where the solubility state of asphaltenes is not disturbed by dilution or by the fractionation method. For some crude oils, asphaltenes are present as particles and may be

IFT, 55oC, 1.5h (mN/m)

20

15

10

0 0 5 10 15 20 IFT, 25oC,12h (mN/ m) 25 30

Fig. 3. Near-equilibrium interfacial tension values of crude oilwater interfaces at 25oC against those for both diluted (0.7 vol/vol oil in toluene) and undiluted samples at 55oC. Error bars represent one standard deviation at each side of the average value.

Correlating E and E to the Lifetime of Emulsions A Confusing Task! Due to the importance of dilational viscoelastic properties of interfaces during film thinning of coalescing droplets or bubbles, it is highly desirable to be able to estimate these particular

correctly imitated by fractionated asphaltenes in a toluene/alkane mixture. In addition, the access to interfacially active components in the bulk phase is completely changed in a model system. If such modification is not done deliberately, the effect should not be disregarded. The molecular exchange of interfacially active components (not necessary those that make up the film material) from the bulk towards the interface will have significance for the measured viscoelastic moduli (through the characteristic time of bulk diffusion) and must be considered if the measured parameters should be attributable to real systems. There are still unresolved issues with regard to the mobility of interfacially active components in the contact zone between two colliding water droplets. It has been shown that the interfacial properties of two colliding droplets are highly dependent on whether the interfacially active components are present in the drops or in the surrounding medium [38]. The perturbation methods can with some caution be used for qualitative comparison of different interfaces. The main advantage of such methods is that they can provide information on the kinetics of interfacial reactions and its effect on interfacial adsorption and mobility. For such information to be accessible, the experimental setup should offer a wide range of perturbation frequencies. However, for viscous crude oils inertial effects kick in at higher frequencies and invalidate the measured rheological parameters. Bouriat and coworkers [8, 11] have recently applied the time-temperature superposition (TTS) principle to dilational rheology of crude oil-water interfaces. The principle is based on the temperature dependent relaxation of chain segments of for example polymers. The dilational rheology data at various temperatures were brought together to a master curve by applying a horizontal (log) shift factor. Thus, rheological properties could be estimated over a frequency range which is otherwise inaccessible to the range of the experimental measurement. Wantke and coworkers [39-42] have characterized several surfactant systems in the mid frequency range(3-500 Hz) by also taking into account inertial effects. Such approach has great advantage when trying to determine intrinsic properties of interfaces since bulk diffusion effects can be suppressed at higher modulation frequencies. Equilibrium interfacial properties when are they reached in real processes? To be able to relate the measured interfacial parameters to the stability of emulsions, initially in laboratory scale experiments then in real processes, the previous question has to be answered. Goual et al. [43] have done direct measurement of adsorption of bituminous components at oilwater interfaces with the Quartz Crystal Microbalance technique. They concluded that a slow, non-steady state absorption took place above a critical dilution with heptane/toluene 50/50 where asphaltenes began to form particles. Below this critical dilution, adsorption was quick (minute) and steady state. Adsorption in the high dilution range was up to several times larger in mass than below critical dilution. Thus, any interfacial relaxation exceeding the characteristic adsorption time observed by Goual et al. [43] must be related to other phenomena. Results in Figure 2 prove that the interfacial tension relaxation exceeds that of the

adsorption process (Figure 2). Adsorption may be even faster in turbulent systems. Still, we have observed that the rheological character, E(0.1 Hz), of interfaces between water and oil reach more or less steady values during the first hours of aging when asphaltenes are well solubilized in the crude oil medium [20]. Slow and non-steady state increase of E has previously been observed in diluted systems where asphaltenes are present as aggregates or particles [20]. It should be mentioned that Speicker and Kilpatrick [7] have reported gradually increasing film elasticity over 24 hours in systems where asphaltenes are believed to be soluble in the toluene/heptane (55/45) mixture. They also observed solid like properties of the horizontal films. Any surfactant adsorption which lowers the interfacial tension of the oil-water interface will have a viscoelastic response to dilation in the frequency range corresponding to the characteristic time of the surfactant. Thus viscoelastic properties of the water/oil interface may also be important for the stability of such interfaces even when the reorganization processes are not completed. Thus interfacial rheological properties have also importance for applications (certain stages in petroleum industry) where the emulsion is separated minutes after its formation.
35

30 25 |E | (mN/m)

20 30 15

10 5

0 0.01

0.1 Frequency (Hz)

Fig. 4. Magnitude of the dilational modulus and frequency dependence of 30 diluted (0.7 vol/vol oil in toluene) crude oils.

Dilational Characteristics of Diluted Crude Oil-Water Interfaces Dilational rheology of model oil systems, in particular of asphaltenes has been studied extensively. Generally, irreversible adsorption of asphaltenes from toluene/alkane solutions has been reported, either based on high dilational elasticity or the observation of crumpled interfaces when aged films are compressed [9, 13, 24, 36, 44]. Goual et al. [43] verified the irreversibility of asphaltene adsorption from diluted bitumen solutions with a QCM technique. Adsorption at the water-oil interface was found to be irreversible above a critical dilution where asphaltenes were precipitated and partially reversible below the critical dilution where asphaltenes were well solvated.

Although the irreversibility of the adsorption may also result from reorganization within the film, thus being not directly dependent on the state of asphaltenes in the bulk, it is obvious that the aggregation state of asphaltenes in bulk have great influence on the film properties. Asphaltene molecules are soluble in toluene (low concentrations), but have great tendency to self-associate in mixtures of alkane and toluene. Aske et al. [45] compared the viscoelastic properties of 20 different conventional crude oils diluted to 0.002, 0.01, and 0.02 vol/vol oil in a mixture of toluene and heptane (50/50). As reported by the authors, the dilution of the crude oils resulted in precipitation of asphaltene particles. Such systems cannot correctly mimic the viscoelastic properties of the undiluted crude oil-water interfaces. To be sure to not enhance the tendency of asphaltene aggregation or precipitation, the 30 crude oils were diluted in toluene (0.7 vol/vol oil). Figure 4 shows the frequency dependence of the near-equilibrium (1.5 h) dilational modulus of the 30 diluted crude oils. Each experiment was performed 3 times. The apparent logarithmic frequency dependence of the moduli is only a result of the narrow range of frequencies studied with respect to the complete profile. The moduli are in the range from 5-22 mN/m at = 0.1 Hz which is comparable to previous studies of diluted crude oils [20]. Except for sample 30, the frequency dependence generally increases with the dilational modulus so that crude oils with high |E| have a steeper slope than those with low |E|. This appears even more clearly when the frequency dependence of the dilational elasticity modulus, d|E|/dlog, is plotted with the dilational elasticity at 1 Hz as in Figure 5.
35 Diluted oils 30 Undiluted oils 24-27oAPI 24 25 |E (1)| (mN/m) 20 9 9 19 17 17 23 23 24

be discussed. First of all, the crude oil sample no. 30 again proves to be different from the rest of the crude oils. The crude oils in the top-right corner of Figure 5 are light crude oils (2427oAPI) whereas the 8 samples in the lower left corner turn out to be among the heaviest samples of the sample set (12-19oAPI). The dilational modulus depends on diffusional exchange from bulk following interfacial extension. Thus it seems reasonable to assume that the great amount of interfacially active components (asphaltenes and resins) in the heavy crude oils may suppress the measured dilational modulus.
50 45 40 Phase Angle (Degrees) 35 30 25 20 15 10 5 12-19oAPI samples 0 0.01 0.1 Frequency (Hz) 1 30

Fig. 6. Phase angle of the dilational modulus of diluted (0.7 vol/vol in toluene) crude oil-water interfaces.

20 22 11 11 7 26 20 26 30 30

15 10

12-19 API

7 19

5 0 0.0 2.0 4.0 6.0 d |E |/d log (mN/mHz) 8.0 10.0

Figure 5. Magnitude of the dilational modulus at 1 Hz for diluted (0.7 vol/vol oil in toluene) and undiluted crude oils as a function of their frequency dependence.

For the moment the results for the undiluted crude oils will not

Dicharry et al. [8] and Bouriat et al. [11] have attempted to explain the crude oil-water interfaces with formation of a twodimensional network exhibiting a glass transition zone. Components within the heavy end of the crude oil have been reported to be responsible for the formation of the critical gel. Dicharry et al. [8] related the slope (n) of the linear master curve (brought together with the time-temperature superposition principle), E ~ n to the phase angle, = 180n/2, and the strength of the gel. The measured frequency responses (Figure 4) are in the same range as that reported by Dicharry, but do not lead to agreement between the theoretical and measured phase angles. Neither are the measured phase angles independent on the frequency of oscillation as those reported by Dicharry et al. [8]. It seems like the preparation of the model oils by Dicharry et al., i.e. dilution in cyclohexane, was responsible for extensive aggregation of heavy components (asphaltenes) in bulk thus causing a laterally immobile particle-rich interface. The phase angle and its frequency dependence are presented in Figure 6. The interfacial tension response of sample no. 30 lags behind about 40 degrees with respect to the area oscillation. At frequency 0.01 the phase angle even exceeds the limit of 45 degrees for LVDT model (pure diffusional relaxation) which

may indicate that there is an intrinsic viscous effect (nonequilibrium state) or an adsorption barrier present. This is the sample which had precipitated asphaltenes when received. However, since we have not identified the origin of the very low interfacial tension of this sample (Figure 3) we do not know what is causing the observed effect. Without going into detail about the response of each particular sample, the crude oils with great amount of interfacially active components generally show slightly greater frequency dependence (steeper decay function) than the lighter oils. The corresponding loss moduli (E) are presented in Figure 7. In our experiments, E contains all relaxation processes, i.e., intrinsic interfacial viscosity and diffusion effects. The loss moduli generally increase with frequency or show a maximum in the frequency range from 0.01-1 Hz. The characteristic time of relaxation (multiple processes may be present) is thus in the time interval studied or faster. Since the storage moduli more or less resemble that of the dilational elasticity moduli, they are not included here.
10 9 8 Loss Modulus E '' (mN/m) 7 6 5 4 3 2 1 0 0.01 24 23 17 30

of these results as also intrinsic interfacial properties could be identified. From previous studies [9, 13] and the observation of irreversible crude oil-water films (next section) we expect that real crude oil systems also possess such rheological character. Only then, results from dilational rheology could be compared to those from shear rheology [7] where the interfacial area is constant (no exchange).

Dilational Characteristics of Crude Oil-Water Interfaces By returning to Figure 5, the interfacial dilational modulus at 1 Hz is plotted as a function of its frequency dependence for undiluted crude oils as for the diluted samples. These two parameters characterize the dilational modulus in the studied frequency range. Based on the concentration dependence of the dilational modulus, the viscoelastic parameters were expected to be slightly diminished with the increasing concentration of interfacially active components in the undiluted samples. Of the 13 undiluted samples, 11 samples showed slightly reduced frequency dependence and E(1). The dilational character of sample no. 30 was once again different from the other undiluted crude oils, apparently more viscous (E(0.01)>E(0.01)). The phase angle and the resulting storage and loss moduli of the undiluted samples were more or less similar to that of the diluted samples as demonstrated in Figure 8 for E(). The characteristic time of relaxation was less than 10 sec. Although the measured rheological parameters were affected by diffusion and cannot directly describe the intrinsic interfacial properties, it seemed like these interfaces were much more soluble than the solid-like films observed in asphaltene-in-toluene/heptane (ratio<1) systems. Still, certain films showed some degree of crumpling when excessively compressed after the experiments were completed.
8 7 6 Loss Modulus E '' (mN/m) 5 4 3 2 1 0 0.01 24 23 17 14 26 2 5 20 7 30

0.1 Frequency (Hz)

Fig. 7. Loss modulus of diluted crude oil-water interfaces (0.7 vol/vol oil in toluene).

Clearly, the measured dilational properties are strongly affected by molecular exchange from the bulk, as they should. Freer and Radke [13] demonstrated this phenomenon for an asphaltene film at the interface between toluene and water by washing out the interfacially active components from the bulk, thus eliminating molecular exchange. When asphaltenes in the bulk solution surrounding the aged water droplet were removed, the interface appeared as much more solid-like with characteristic time an order of magnitude larger than that observed prior to washout. Moreover the response in storage modulus E was significantly narrowed and the broad relaxation (also apparent in E) at higher frequencies disappeared. An experimental setup similar to that developed by Wantke et al. [41] (now, with accessible frequencies up to 1000 Hz) would increase the value

0.1 Frequency (Hz)

Fig. 8. Loss modulus of undiluted crude oil-water interfaces.

CONCLUSIONS
Interfacial properties of diluted (0.7 vol/vol in toluene) and undiluted crude oil-water interfaces have been reported. The tension of the 30 crude oil-water interfaces decayed logarithmically during the first 12 hours of aging to values in the range between 8 and 24 mN/m at 25oC (pH 6.5). The oscillating drop method was used to study dilational properties of the interfaces. As expected in the low frequency range, molecular exchange from the bulk oil phase affected the measured dilational parameters. For the diluted crude oils, the frequency dependence of the dilational modulus increased with its magnitude as expected for diffusion-controlled relaxation of soluble films. For this reason, the systems which exhibited particularly low magnitude of the dilational modulus were of the heaviest crude oils in the sample set. Consequently, the systems with greatest dilational modulus were among the lightest crude oils. The overall characteristic time of relaxation of the crude oil-water interfaces was in the range below 10 sec. The undiluted crude oil-water interfaces had similar interfacial properties as the diluted samples except for slightly reduced magnitude of the dilational modulus. Even though the crude oil-water interfaces behaved at least partially as soluble, some observations pointed to intrinsic rheological properties of the interfaces. Some samples showed great phase angles () which may indicate a kinetic barrier to adsorption. Moreover, certain aged crude oilwater interfaces crumpled slightly when extensively compressed, a sign of irreversible adsorption. Due to inertial effects, the measurements were restricted to the low frequency range of oscillations (0.01-1 Hz) and could not verify these phenomena in a frequency range where the effect of molecular exchange mechanisms are suppressed. An experimental setup which also makes the medium frequency range (1-500 Hz) accessible could significantly increase the value of studies of concentrated crude oil systems.

[5] [6]

[7] [8] [9] [10] [11] [12] [13] [14] [15] [16]

[17] [18] [19] [20] [21]

ACKNOWLEDGEMENT
The technology program Flucha III, Particle-Stabilized Emulsions/Heavy Crude Oils, financed by industry and the Norwegian Research Council are gratefully acknowledged for financial support. Jan Ole Sundli enthusiastically made the amplifier from a radio tube. We cannot thank him enough for his extra effort.

[22]

[23] [24]

[25]

REFERENCES
[1] [2] D. S. Valkovska, K. D. Danov, I. B. Ivanov, Colloids and Surfaces 156 (1999) 547. D. S. Valkovska, K. D. Danov, I. B. Ivanov, Colloids and Surfaces A: Physicochemical and Engineering Aspects 175 (2000) 179. D. S. Valkovska, K. D. Danov, I. B. Ivanov, Advances in Colloid and Interface Science 96 (2002) 101. K. D. Danov, D. S. Valkovska, I. B. Ivanov, Journal of Colloid and Interface Science 211 (1999) 291.

[26] [27] [28]

[3] [4]

[29]

M. A. Bos, T. van Vliet, Advances in Colloid and Interface Science 91 (2001) 437. G. Horvth-Szab, J. H. Masliyah, J. A. W. Elliott, H. W. Yarranton, J. Czarnecki, Journal of Colloid and Interface Science 283 (2005) 5. P. M. Speicker, P. K. Kilpatrick, Langmuir 20 (2004) 4022. C. Dicharry, D. Arla, A. Sinquin, A. Graciaa, P. Bouriat, Journal of Colloid and Interface Science (In Press) D. Sztukowski, H. W. Yarranton, Langmuir 21 (2005) 11651. S. Poteau, J-F. Argillier, D. Langevin, Energy and Fuels 19 (2005) 1337. P. Bouriat, N. E. Kerri, A. Graciaa, J. Lachaise, Langmuir 20 (2004) 7459. F. Bauget, D. Langevin, R. Lenormand, Journal of Colloid and Interface Science 239 (2001) 501. E. M. Freer, C. J. Radke, Journal of Adhesion 80 (2004) 481. Y. Wang, L. Zhang, T. Sun, S. Zhao, J. Yu, Journal of Colloid and Interface Science 270 (2004) 163. J. G. Speight, Oil and Gas Science and Technology Rev.IFP 59 (2004) 467. O. C. Mullins. Molecular Structure and Aggregation of Asphaltenes and Petroleomics. in SPE Annual Technical Conference and Exhibition. 2005. Dallas, Texas, U.S.A.: Society of Petroleum Engineers. S. Badre, C. C. Goncalves, K. Noringa, G. Gustavson, O. C. Mullins, Fuel 85 (2006) 1. A. G. Marshall, R. P. Rodgers, Acc. Chem. Res. 37 (2004) 53. Y. Ruiz-Morales, Journal of Physical Chemistry 108 (2004) 10873. A. Hannisdal, R. Orr, J. Sjblom, Journal of Dispersion Science and Technology 28 (2007) A. Hannisdal, P. V. Hemmingsen, J. Sjblom, Industrial & Engineering Chemistry Research 44 (2005) 1349. P. V. Hemmingsen, A. Silset, A. Hannisdal, J. Sjblom, Journal of Dispersion Science and Technology 26 (2005) 615. E. M. Freer, H. Wong, C. J. Radke, Journal of Colloid and Interface Science 282 (2005) 128. M. Jeribi, B. Almir-Assad, D. Langevin, I. Hnaut, J-F. Argillier, Journal of Colloid and Interface Science 256 (2002) 268. E. Y. Sheu, M. M. De Tar, D. A. Storm, Fuel 71 (1992) 1277. E. Y. Sheu, D. A. Storm, M. B. Shields, Fuel 74 (1995) 1475. Y. Xu, Energy and Fuels 9 (1995) 148. J. Maldonado-Valderrama, V. B. Fainerman, E. Aksenenko, M. J. Galvez-Ruiz, M. A. Cabrerizo-Vilchez, R. Miller, Colloids and Surfaces A-Physicochemical and Engineering Aspects 261 (2005) 85. C. J. Beverung, C. J. Radke, H. W. Blanch, Biophysical Chemistry 81 (1999) 59.

[30] [31]

[32]

[33] [34] [35]

[36]

[37]

T. D. Gurkov, S. C. Russev, K. D. Danov, B. Ivanov, B. Campbell, Langmuir 19 (2003) 7362. L. Liggier, M. Ferrari, A. Massa, F. Ravera, Colloids and Surfaces A: Physicochemical and Engineering Aspects 156 (1999) 455. I. B. Ivanov, K. D. Danov, K. P. Ananthapadmanabhan, A. Lips, Advances in Colloid and Interface Science 114-115 (2005) 61. J. Lucassen, M. van den Tempel, Shemical Engineering Science 27 (1972) 1283. J. Lucassen, M. van den Tempel, Journal of Colloid and Interface Science 41 (1972) 491. H. Fruhner, K.-D. Wantke, K. Lunkenheimer, Colloids and Surfaces A: Physicochemical and Engineering Aspects 162 (1999) 193. C. Tsamantakis, J. Masliyah, A. Yeung, T. Gentzis, Journal of Colloid and Interface Science Article in Press (2005) A. Hannisdal, M.-H. Ese, P. V. Hemmingsen, J. Sjblom, Colloids and Surfaces A: Physicochemical and Engineering Aspects 276 (2006) 45.

[38]

[39]

[40] [41] [42]

[43] [44] [45]

K. D. Danov, P. A. Kralchevsky, B. Ivanov, Dynamic Processes in Surfactant-Stabilized Emulsions, in Encyclopedic Handbook of Emulsion Technology, J. Sjblom, Editor. 2001, Marcel Dekker, Inc.: Trondheim. p. 621. K.-D. Wantke, J. rtegren, H. Fruhner, A. Andersen, H. Motschmann, Colloids and Surfaces A: Physicochemical and Engineering Aspects 261 (2005) 75. K.-D. Wantke, H. Fruhner, Journal of Colloid and Interface Science 237 (2001) 185. K.-D. Wantke, H. Fruhner, J. Fang, K. Lunkenheimer, Journal of Colloid and Interface Science 208 (1998) 34. K.-D. Wantke, H. Fruhner, J. rtegren, Colloids and Surfaces A: Physicochemical and Engineering Aspects 221 (2003) 185. L. Goual, G. Horvth-Szab, H. Masliyah, Z. H. Xu, Langmuir 21 (2005) 8278. X. A. Wu, J. Czarnecki, Energy and Fuels 19 (2005) 1353. N. Aske, R. Orr, J. Sjblom, H. Kallevik, G. ye, Journal of Dispersion Science and Technology 25 (2004) 263.

Paper 5

Stability of Water/Crude Oil Systems Correlated to the Physicochemical Properties of the Oil Phase
Andreas Hannisdal*, Pl V. Hemmingsen, Anne Silset, and Johan Sjblom Ugelstad Laboratory, Department of Chemical Engineering, Norwegian University of Science and Technology (NTNU), N-7491 Trondheim, Norway. A characterization of thirty crude oils has been performed to determine the relative level of influence that individual parameters have over the overall stability of w/o emulsions. The crude oils have been analyzed with respect to bulk and interfacial properties and the characteristics of their w/o emulsions. The parameters include compositional properties, acidity, spectroscopic signatures in the infrared and near-infrared region, density, viscosity, molecular weight, interfacial tension, dilational relaxation, droplet size distribution and stability to gravitationally and electrically induced separation. As expected, a strong covariance between several physicochemical properties was found. Near-infrared spectroscopy proved to be an effective tool for crude oil analysis. In particular, we have showed the importance of the hydrodynamic resistance to electrically-induced separation (static) in heavy crude oil-water emulsions. A rough estimate of the drag forces and dielectrophoretic forces seemed to capture the difference between the 30 crude oils. Given enough time, water-in-heavy oil emulsions could be destabilized even at very low electric field magnitudes (d.c.). When droplets approach each other in an inhomogeneous electric field, strong dielectrophoretic forces disintegrate the films and result in coalescence. The relative contribution from film stability to the overall emulsion stability may therefore be very different in a gravitational field compared to that in an electrical field. Keywords: Crude oil; Characterization; NIR; PFGSE-NMR; Interface; Emulsion stability; Electrocoalescence; Viscosity.

INTRODUCTION
In petroleum industry, stable water-in-oil (w/o) emulsions can lead to enormous financial losses if not treated properly or prevented. Thus, successful operation can only be achieved by carefully controlling properties that will influence the tendency for emulsion formation and subsequently the destabilization of such systems. The current research in the area of petroleum emulsions is vast and involves different disciplines. The dynamic behavior of emulsions is studied to understand the macroscopic criteria for effective separation methodology and design [1, 2]. Electrostatic phenomena of droplet interaction aim to describe the efficiency of industrial electrocoalescers [3-7]. The chemistry and structural aspects of crude oil components, in particular asphaltenes and resins [8-10], wax [11] and naphtenates [12] have been studied extensively due to their

influence on both bulk and interfacial properties. Interfacial film characterization [13-16] and compositional analysis [17-20] can explain the forces [21] associated with coalescence of w/o interfaces. Inorganic particles and their partitioning at interfaces is studied due to their effect on stability and phase inversion phenomena in petroleum emulsions with coproduced solids (etc. reservoir material) [22-26]. The important stabilization mechanisms for water-in-crude oil emulsions have probably been identified, but the difficulty still lies in the assessment of the relative levels of influence that individual variables have over the final stability. Moreover, the wild variation in properties and performance of different crude oils or crude oil fractions has frustrated many researchers. From our point of view, this requires a study of several crude oil systems with different origin and geological history. The overall objective must be to understand the system properties which in particular have importance for the stability of w/o systems and relate these properties to different stages of the destabilization process. Multivariate analyses can be an effective tool when several parameters contribute simultaneously to the same phenomenon. However, from a statistical point of view, building a prediction model with highly correlating variables is very difficult. For petroleum systems a high degree of covariance between certain material properties is expected. With this in mind we have decided to allow for a careful investigation of the data matrix. The process of accumulating knowledge (Figure 1) is necessary to point out the direction for further work and to enter the modeling stage with the proper domain-specific information. Earlier and elsewhere, we have reported on the group-type fractionation method and the ability of near-infrared and infrared spectroscopy to predict the SARA components in crude oil [27]. Interfacial properties of the crude oil-water interfaces and some preliminary studies of stability to electrically induced destabilization were presented by Hannisdal et al. [28] and Hemmingsen et al. [29], respectively. Since then, we have been working further in this direction and wish to give a full account on the data obtained from bulk and interfacial characterization. Herein, a characterization of 30 crude oils is presented with respect to bulk characteristics, interfacial properties and the stability of w/o emulsions. The focus is on results and discussions rather than a complete experimental and theoretical description of principles of characterization techniques.

EXPERIMENTAL Materials
Thirty dead crude oils were received from 9 different oil companies with production sites on the Norwegian Continental Shelf, in the South China Sea, the Gulf of Mexico, United Kingdom, France, Brazil, West-Africa and Alaska. About half of the oils were per definition heavy (< 20oAPI) whereas the lightest sample was a 33.6oAPI crude oil. Some samples were originally highly undersaturated from deep water reservoirs, whereas other samples had been recovered from shallow depth as dead oil. Some general characteristics

are reported in Table 1. Reflecting the great variety in crude oil properties, the operators face different challenges with respect to flow assurance and quality of the well product due to wax, co-produced solids, asphaltene precipitation or great viscosity of the oil phase. Fifteen of the samples were expected to be waxy from their Bingham plastic behavior below 25oC, typical for wax crystallization [29]. All crude oils were sampled in a way that minimized contamination by production chemicals. Some samples were received with considerable amount of co-produced water which was removed by gravitationally-induced separation. However, for some viscous samples the residual water could not be removed easily and the w/o systems were analyzed as received. For experiments involving a separate water phase, Milli-Q quality water with pH 6.0-6.5 and 3.5 wt.% NaCl was used.

Experimental Design
The study was performed according to the experimental design presented in Figure 1. Thirty crude oils were analyzed with respect to bulk and interfacial properties which were expected to be of importance for the overall stability of w/o emulsions. Group-type fractionation of the crude oils was done according to Hannisdal et al. [27], involving precipitation of n-hexane insolubles and semi-preparative liquid chromatography (HPLC) of the remaining oil. Toluene insolubles and residual water were quantified according to ASTM D4807-88 and by Karl Fisher titration, respectively. Spectroscopic investigations within the infrared and near-infrared regions were used for compositional analyses [27]. The amount and property of heavy hydrocarbons and solids are expected to have great importance for the stability of w/o emulsions as well as physical properties of the oil phase [30-32]. Acidic functionalities have great importance for interfacial activity of crude oil components and were identified by titration and spectroscopy. Due to the strong signatures of electronic absorption and scattering, NIR spectroscopy was used to predict physical properties of the crude oils and the tendency for asphaltene precipitation (continuous titration with pentane). Bulk properties have been measured, including density (20-50oC), molecular weight, and dynamic viscosity (0-60oC). Dynamic interfacial tension (12 hours) and dilational relaxation properties are expected to have importance for droplet formation and coalescence processes and have been published recently [28]. Water-in crude oil emulsions were created by heavy mixing with an Ultra Turrax. The resulting droplet size distribution was determined by pulse field gradient spin echo (PFGSE) NMR [29]. The destabilization of emulsions induced by gravity or an electric field was studied. The ultimate objective was to decouple the flocculation and coalescence steps of the emulsion destabilization process, if necessary with multivariate statistical analyses. As a base case for emulsion preparation and analyses the following conditions were used: 20 ml total sample, w = 0.3 vol/vol water (3.5 wt.% NaCl, pH 6.0-6.5), 40oC, mixing with Ultra Turrax (IKA T25, S25-N-18G head) for 2 min at 22000 rpm, 30 mm tube diameter.

Instrumentation
High performance liquid chromatography (HPLC) was performed on an instrument from Shimadzu Inc. The residual water amount was quantified with a Karl Fisher titrator (Metrohm 756 KF Coulometer). The IR spectra were recorded with a Tensor 27 FT-IR spectrometer (Bruker Optics) with a MKII Golden Gate ATR unit (Specac) in single reflection configuration. NIR spectra of crude oils (0.32 mm path length) were recorded with a diffuse reflectance-absorbance accessory connected to a Bruker MPA FT-NIR spectrometer. HPLC, Karl Fisher, IR and NIR spectroscopy were performed as carefully described by Hannisdal et al. [27] with one exception. For NIR spectroscopy, a gold reference (Bruker Optics) was used instead of the copper plate background. Asphaltene precipitation onset experiments were performed by continuous titrating crude oil with pentane in a stirred sample container with a transflectance (2*1mm path length) probe connected to the Bruker MPA spectrometer. Acidity of the crude oils was determined by potentiometric titration according to ASTM D664 (Metrohm, 808 Titrando, 2 parallels) and validated from the acidic signature (C=O, ~1705 cm-1) in the infrared region. Density was measured at two independent laboratories in the temperature range from 2060oC with an Anton Paar density meter (DMA 48, 2 parallels) and a Mettler Toledo density meter (Densito 30PX ). Viscosities were measured from 60oC to 0oC with shear rate scans from 0.1 to 1000 Hz on a Physica rheometer (MCR 100) with a cylindrical cell according to Hemmingsen et al. [29]. The reported viscosities are the Newtonian fit to the shear rate scans. Molecular weight of the crude oils was measured with a cryometer for benzene (Type 13 KAutocal) with two parallels. Interfacial properties was measured with a CAM 200 (KSV, Finland) pendant drop instrument modified according to Hannisdal et al [28] with a piezoelectric actuator for dilation relaxation experiments. The PFGSE-NMR experiments were performed on a Bruker Minispec mq NMR Analyzer 20 MHz equipped with a diffusion probe as described elsewhere by Hemmingsen et al. [29]. The critical electric field (CEF) cell is a parallel plate coalescer with brass electrodes separated by a 0.25 mm Teflon sheet. The CEF cell has been presented previously [29, 33].

RESULTS Crude Oil Composition

Table 1 summarizes the compositional analysis according to the method presented by Hannisdal et al. [27]. Shortly, n-hexane (1g: 40ml) insolubles were precipitated from the oil in 3 parallel filtrations with a reproducibility indicated by an average standard deviation of 0.2 wt.%. The maltenes were fractionated with silica/amino HPLC and their amount determined gravimetrically with average standard deviations of 1.2 wt.% (S), 0.8 wt.% (A), 0.8 wt.% (R). As discussed by Hemmingsen et al. [29] the dominating uncertainty of the method lies in the insufficient sample recovery of the lightest crude oil, which may be as low as 80 wt.% due to evaporation of light components during solvent removal. However, the most interesting components are within the heavy end of the crude oil. A peak-by-peak

identification in the IR region reveals characteristics of the SARA fractions. The n-hexane insoluble fraction was most aromatic followed by the resin fraction, the aromatic fraction and the whole crude oil. Acidic functional groups were predominantly found within the resin fraction. Acidity of the crude oils (Total Acid Number, TAN) from titration experiments was compared with the height of the C=O vibrational band at 1705 cm-1 for carboxylic acids. The average standard deviation for TAN determination for all samples was 0.05 mg KOH/g oil. With two exceptions, the acidity from the two methods correlated well with a squared correlation of 0.982 as shown in Figure 2. The error bars represent one sample standard deviation in every direction. Repeated measurements verified the results for sample no. 30, whereas the variance of the TAN measurements on the extremely viscous sample 13 (with 4.3 wt.% residual water) was rather large. On closer examination, sample 30 had very strong signature for dimeric carboxylic C=O stretch (1705 cm-1), C-O stretch of carboxylic moieties (1013 cm-1) and stretching absorption of hydrogen bonded hydroxyl groups (~3300 cm-1) in the resin and n-hexane insoluble fraction. Strong acidic signatures within the asphaltene fraction are untypical. Intermolecular hydrogen bonding is considered important for asphaltene aggregation. As apparent from the statistical parameters in Table 1, the sample set was biased with respect to the amount of n-hexane insolubles. Four samples had more than 5 wt.% n-hexane insolubles, whereas 9 samples had less than 1 wt.%. The amount of toluene insolubles was determined from 3 parallel filtrations of about 50 g crude oil according to ASTM D480788, but at 80oC. Surprisingly, the amount of solids was very low with a maximum of less than 0.2 wt.%. However, since previous studies [22, 23, 32, 34-39] have shown effective particle-stabilization, even at very low concentrations, solids should not be disregarded.

Stability of the Crude Oil System

The solubility state of asphaltenes is of great importance in petroleum processing due to increased deposition tendency and stabilization efficiency of flocculated asphaltenes [30, 31]. Figure 3 shows the apparent absorbance at 1600 nm (NIR) during pentane titration of certain crude oils in the sample set. Since equilibrium solubility can only be realized after long time, the titration experiments were performed over 12 hours, with stirring. The linear decay in absorbance for oil no. 26 and 29 is a result of continuous dilution of the crude oil sample. However, at a certain critical dilution, asphaltenes started to flocculate as a result of insufficient solubility of the crude oil medium. Flocculated asphaltene aggregates scattered light and increased the optical density of the system. Crude oil no. 30 showed a completely different behavior with immediate increase in optical density at 1600 nm following dilution. Asphaltenes were already precipitated when the oil was received from the operator as verified microscopically in Figure 4 (left). This particular crude oil (~30oAPI) was produced from very deep water with asphaltene content of 3.3 wt.% and resin content of 7.6 wt.%. When emulsified with water, the system apparently formed a network of asphaltene particles and water (visible by microscopy in Figure 4, right) which behaved as a gel (oscillation bulk rheology). Except for crude oil no. 30 and one more

sample, the crude oils in the sample set were generally very stable to asphaltene aggregation with onset values in the volume fraction range 0.260.60 vol/vol corresponding to concentrations in the range 0.250.54 g oil/ml. The stability to asphaltene aggregation increased progressively (with a non negligible dispersion) with the mass density of the crude oil so that heavy crude oils were more stable than light oils.

Near-Infrared Spectroscopy
The applicability of NIR spectroscopy in petroleum industry can be highlighted with previous studies of asphaltene precipitation from live crude oil [40, 41], compositional analysis of gas and condensates at elevated pressures and temperatures [42, 43], gas-oil ratio of live oils [44], compositional analyses of conventional and heavy oil at ambient conditions [27, 45], and of the dispersive effects of amphiphiles and napthenic acids on asphaltene aggregates [46]. The NIR range contains valuable information due to vibrational features, electronic absorption and scattering. NIR can be used online at high temperatures and pressures as in the borehole. In addition to the asphaltene stability experiments presented in the previous section, NIR spectroscopy was used to capture characteristic physicochemical features of the crude oils. Due to very low transmittance through the heaviest samples, an experimental setup with only 2x0.16 mm (reflectance) path length was used. As apparent from Figure 5 the NIR spectra of the 30 crude oils (average of 3 parallels) show overtone and combination bands of hydrocarbon molecular vibration predominantly in the 1100-1250, 1350-1550, 16001750, 2100-2200 nm ranges. These vibrational features can be very informative in certain applications, but will not be discussed further here. We are primarily concerned with the broad, increasing absorption profile at low wavelengths. Previous studies have shown that absorption, rather than scattering from dispersed asphaltenes produces this typical signature of crude oils [47]. However, scattering contributions from residual water, crystallized wax, precipitated asphaltenes and inorganic solids causes an upwards baseline shift for several crude oils, most apparent for sample 13, 20, 4 and 6 which contained up to 12.6 wt.% residual water. The reproducibility of the NIR experiments was generally poorer for these samples compared to non-scattering samples which proved to be very reproducible. Microscopic observation of crude oil samples pressed between two glass slides verified the observations from NIR. Without exception, the 7 samples with water content exceeding 1 wt.% were easily identified. Moreover, precipitated asphaltenes were found in two oils in line with the asphaltene stability experiments (Figure 3 and 4). Interestingly, in the crude oils which showed greatest Bingham plastic yield stress [29] rod-shaped particles with birefringent properties (cross-polarized light) were observed. To even further verify the origin of the baseline shift, the crude oils which seemed to not only display electronic absorption, but also scattering were heated and centrifuged repeatedly. Figure 6 shows the resulting spectra, once again with 3 parallel experiments. First of all, the reproducibility was satisfactory. The strong signature from water (1450, 1950nm) was almost completely disappeared for all crude oil except crude oil no. 6. Now, the origin of the exponential

feature can be verified. In the spirit of Mullins [48] the average spectra from Figure 6 is plotted on a logarithmic ordinate axis against wavenumbers in Figure 7. The similarity in the spectra is much more striking in this figure due to identical slopes in the electronic absorption edge. Mullins has discussed the signification of this observation which also holds for the 30 crude oils in this sample set. The origin of the exponential feature is electronic excitation from aromatic chromophores in asphaltenes which also for this sample set seems to follow the Urbach phenomenon where the tail in the optical absorption coefficient = o exp(h/Eo) can be explained by the photon energy (h) and the Urbach width Eo. Crude oil no. 15 exhibited a magnitude of the Urbach width of 2317 cm-1 which is within the range (one st.dev.) reported for crude oils by Mullins [48]: 2162 260 cm-1.

Bulk Properties
Density, molecular weight, and dynamic viscosity characterize the bulk properties of the crude oils as schematized in Figure 1. Density was determined in the temperature range from 20-60oC by two independent laboratories with a squared correlation of 0.987 for the density at 40oC. Molecular weight was quantified with a variance of 123 g/mol corresponding to an average standard deviation of 11 g/mol. Dynamic viscosity was measured in the shear rate range from 0.1 to 1000 Hz in 5oC (or 10oC) temperature steps from 60 to 0oC as reported previously by Hemmingsen et al. [29] for 27 of the crude oils in the sample set. Generally, some crude oils showed pseudoplastic flow character at low temperatures. As already mentioned, half of the 30 samples were Bingham plastic at low temperatures, typical for wax-containing crude oils. Statistics for the distribution of density, molecular weight and Newtonian viscosity at 40oC are presented in Table 1. It is seen from the median and average of density and molecular weight that the sample set is evenly distributed with respect to these variables. The distribution of viscosity is biased with more samples having low viscosity.

Dynamic Interfacial Tension and Dilational Relaxation

Characterization of the crude oil-water interfaces has been presented recently [28]. Shortly, dynamic interfacial tension experiments of a static oil droplet in water (~ planar interface) showed rapid initial relaxation of the interface followed by a slow relaxation taking a logarithmic decay function. The near-equilibrium interfacial tension value (12 h) of undiluted crude oils is plotted with the interfacial tension from dilational relaxation experiments of diluted crude oils (0.7 vol/vol in toluene) and undiluted crude oils in Figure 8. Error bars represent one standard deviation in each direction from the average value. The average standard deviation of more than 100 static drop experiments was reported to be 2 mN/m. Dilational relaxation experiments were performed with sinusoidal modulation in the frequency range between 0.01 and 1 Hz (aged for 1.5 h, 55oC). Due to inertial effects the heaviest crude oils could not be analyzed without being diluted. As expected in the low

frequency range (0.01-1 Hz), molecular exchange from the bulk oil phase strongly affected the measured dilational parameters. For the 30 diluted crude oils, the frequency dependence of the dilational modulus, |E|, increased with its magnitude. As apparent from Figure 9, the systems which exhibited particularly low magnitude of the dilational modulus were of the heaviest crude oils in the sample set, whereas the systems with greatest dilational modulus were among the lightest crude oils. The overall characteristic time of relaxation of the crude oil-water interfaces was in the range below 10 sec. The undiluted crude oil-water interfaces had similar interfacial properties as the diluted samples except for slightly reduced magnitude of the dilational modulus. The dilational relaxation results need a comment. Due to some previous studies, there is a general misconception of a direct correlation between a low frequency (e.g. 0.1 Hz) storage modulus (E=|E|cos() also incorrectly called interfacial elasticity) and the stability of petroleum emulsions. Unfortunately, the reality is much more complicated than that as discussed by Hannisdal et al. [28]. Dilational relaxation parameters characterize all processes that contribute to relaxation, including molecular exchange from bulk. Such parameters may provide important information with respect to the characteristic time of relaxation processes, i.e. diffusional exchange, when studied in a proper frequency range. However, if not studied in a frequency range where the effect of bulk diffusion is suppressed, these parameters cannot explain intrinsic properties of the interface (interfacial elasticity and viscosity). Generally, the crude oil/water interfaces appeared to be soluble.

Droplet Size PFGSE NMR

The size of droplets in petroleum emulsions is detrimental for the stability of the system. Therefore an identification of the drop size distribution must be undertaken when studying stability of such emulsions even though the non-transparency of petroleum emulsions complicates the analysis. Even more desirable is an understanding of how the droplet size depends on physicochemical properties of the phases and the process conditions. In general the size will be determined by the balance between droplet break-up and coalescence. To achieve the very high shearing stress or the very intense velocity fluctuations needed to deform and break up small droplets, considerable amount of energy has to be dissipated in the liquid. Interfacially active components reduce interfacial tension, thereby facilitation break-up and prevent immediate recoalescence through different film stability mechanisms, given that the characteristic time of adsorption is fast. Droplet size distribution of the concentrated water-in-crude oil emulsions was determined with PFGSE-NMR as described carefully by Hemmingsen et al. [29]. The methodology followed that by Balinov et al. [49]. A summary of the overall results, which will be compared to the dynamic behavior of water droplets in electric fields, are presented in Table 2. The volume distribution was fitted to a lognormal distribution function with mean and standard deviation reported in Table 2. The lognormal approximation was generally very good. For the majority of the emulsions at w = 0.3 vol/vol, droplet size distributions could be characterized by a lognormal distribution function centered at about 6 x 10-6 m with

standard deviation (2) of 0.16 x 10-6 m. From the experiments done here, it cannot be concluded whether a steady state drop size was reached during the emulsification process. Previous studies with the same mixing geometry have shown that the droplet size was reduced with increased mixing speed. Moreover, experiments at w = 0.3 and w = 0.1 (15 samples) showed that the size was barely dependent on the volume fraction of the dispersed phase, which may indicate that the process under consideration was a one-droplet event (break-up). The predominant mechanism for droplet break-up (viscosity or inertia dominated [50]) will be a topic for further investigations and will not be concluded on here. Certain light crude oils in the sample set (low amount of stabilizers and low viscosity) proved to be different from the remaining samples in that they produced emulsions with larger drop sizes. As an example, the distribution statistics of sample no. 24 are reported in Table 2 immediately after droplet formation, 10 minutes after and 20 min after. Clearly, the mean drop size of emulsion no. 24 was larger than the average emulsion system. Moreover, the mean and standard deviation of the distribution function increased progressively during the first half hour after emulsification probably due to both sedimentation and coalescence. This was verified from the experimentally determined water cut in the lower part (~2 cm) of the sample tube where drop sizes were measured. Due to sedimentation, the water cut exceeded the total (theoretical) water cut which was 0.3 vol/vol. The emulsion was not stable over time and separated during the first two hours at 1 x g. The observation supports the conclusions from dilational relaxation experiments where it was stated that the observed dilational characteristics were largely affected by mass transfer mechanisms from the bulk oil phase. Crude oil no. 24 was the sample with greatest dilational modulus (Figure 9). It is clear that this crude oil sample, even when undiluted, did not offer sufficient access to interfacially active compounds through bulk diffusion. Similar conclusions could be done for crude oil no. 23 which also showed great magnitude of the dilational modulus. Depending on the drop break-up mechanism these crude oils may require more energy for droplet break-up due to low viscosity. However, other crude oils with low viscosity, but more surface active components produced smaller drop sizes. As apparent from Table 1, the crude oils in the sample set were extremely different especially with respect to viscosity. Depending on the droplet break-up mechanism, viscosity of the continuous phase may be an important parameter. Increasing viscosity also means decreasing Reynolds number, thus giving less intense turbulence. For reasons to become clear later when the stability of the emulsions is discussed, the droplet size of the most viscous sample (oil no. 13, (40oC) = 23400 mPas) is included in Table 2. The most viscous crude oils produced w/o emulsions with droplet size within the range of the average water-in-crude oil system.

Stability of Emulsions
The stability to separation was tested in a high gravitational field (13000 x g, 25oC). Emulsions were mixed at 40oC according to the base case. Generally, the sedimentation rate of water droplets reflected the viscosity of the continuous phase. The lightest w/o

emulsions were completely separated into a concentrated emulsion phase and excess oil during the first ten minutes of centrifugation, whereas the most viscous crude oils showed no sedimentation at all, even after 40 minutes of centrifugation. More interesting was the comparison of the stability of the sedimented emulsion phase. Whereas some emulsions coalesced immediately after sedimentation, certain emulsions remained stable. In particular, the two crude oils with precipitated asphaltene particles (no. 30, 5) and two of the crude oils which showed Bingham plastic flow behavior (wax) (no. 9, 26) produced w/o emulsions with stability to coalescence. These crude oils were all among the less viscous crude oils in the sample set. It seems reasonable that particles enhanced the stability to coalescence in these emulsions. The emulsion stability to electrically-induced separation was tested with a parallel plate capacitor with a homogeneous background d.c field increasing linearly in magnitude with time. Due to the voltage range (0-100 V) of the instrumental setup, the maximum field was 4.0 kV/cm. The instrumental setup has been presented previously [29]. The applied electrical field induces polarization of the droplets (same magnitude, but opposite polarity on each hemisphere) due to the mismatch in dielectric constants of the dispersed and the continuous phases. The surface charge distribution preserves zero net charge. Thus, in a homogeneous field, the net force on a droplet is zero. However, when subjected to an inhomogeneous field the droplet will experience a stronger field at one side than at the other which results in a net force acting in the direction of the field gradient, a phenomenon called dielectrophoresis [51]. The inhomogeneous electric field may for instance be set up by a nearby dielectric droplet. When two uncharged, but polarized drops get close to each other, the drops experience each others inhomogeneous field and may either attract or repel each other depending on their relative position in the external field (critical angle 54.7o). The total dielectrophoretic force induces a torque tending to align the drop pair with the applied electric field and an attraction force when drops move within the critical angle relative to the background field [6]. Attractive interaction and alignment of stable droplets can result in chains developing and ultimately bridging the gap between the electrodes. The critical electric field, CEF, is defined as the field strength where the capacitor is shortcircuited. Figure 10 shows the CEF for w/o emulsions of the 30 crude oils as a function of the rate at which the electrical field was increased. Each point represents the average of 2-5 measurements of the same emulsion with heteroscedastic variance in the 0-0.25 kV/cm range, greatest for most stable emulsions. Making new emulsions gave little additional error contribution with exception of the heaviest crude oils which were generally more difficult to emulsify due to extreme viscosity. The CEF values increased progressively with increasing field rate which pointed in the direction of a viscous contribution to the observed short-circuiting. Moreover, w/o emulsions of heavy crude oils were generally more stable than emulsions of lighter crude oils. Some viscous emulsions could not be destabilized and were assigned the maximum CEF value: 4.0 kV/cm. For the lightest crude oils, droplet motion due to gravitational forces could not safely be neglected during the experimental time (especially for low field rates). The drop size distribution of such emulsions changed

with time as apparent from the PFGSE-NMR experiments (Table 2). The continuous phase provided a hydrodynamic resistance to droplet motion through its viscosity. Thus, with increasing field rate, droplets were given less time to respond to the electric field and the stability of the emulsion appeared to be greater. The contribution from film stability was at its maximum for experiments at low field rates. The results in Figure 10 suggested that a viscous o/w emulsion could be separated with a constant electric field considerably lower in strength than the observed CEF value with increasing field magnitude. With this hypothesis, one of the most stable emulsions (no. 6, oil = 2270 mPas, 40oC) was exposed to a continuous electric field magnitude in the 0.4-4.0 kV/cm range. Indeed, even though very slowly at the lowest field magnitude, all tests resulted in short-circuiting of the electric field. Even though water-in-heavy crude oil emulsions are expected to have a considerable barrier to drop coalescence through the film strength, such interfaces seemed to provide limited stability to electrical-induced disintegration. Atten [5] came to the same conclusion when the performance of a new type of coaxial cylindrical electrocoalescer (a.c field) was evaluated. Shear-induced droplet collisions with a coalescence probability close to 1 for a certain electric field magnitude could explain the effectiveness of such industrial electrocoalescers which operates with residence time of only seconds.

MODELS AND DISCUSSIONS Correlated Properties and Their True Origin

Asphaltenes self-associate in solution giving rise to more or less extended aggregates which have a profound influence on the viscosity of asphaltenic material. The rheological behavior of asphaltenes is an important property for many industrial applications including roadway construction or petroleum flow assurance. Especially the strongly nonlinear viscosity increase with concentration for concentrated solutions and the non-Arrhenius dependence of viscosity with temperature have been the topic of previous studies. The exponential dependence of asphaltene concentration on viscosity can be explained by hydrodynamic interactions in contrast to Einsteins viscosity model for spherical, monodisperse, and non-interacting particles. In fact, it has been observed that the solution viscosity increases more than exponentially at high particle concentrations. Lin et al. [52] have reviewed previous studies of the rheological behavior of asphaltenes and presented a new suspension viscosity model which takes into account the shape and thermodynamic properties of asphaltene association. Figure 11 shows the dynamic viscosity (40oC), density (40oC), molecular weight and amount of heavy constituents of the 30 crude oils in the sample set. The strong covariance between physicochemical properties can be explained by their identical origin, the properties and amounts of heavy crude oil components. Due to the interfacial activity of such components, interfacial properties like tension and dilational relaxation will also be strongly influenced. For this reason, the interpretation of deterministic relationships can be very difficult for crude oil systems, also from a multivariate statistical point of view.

Predicting Physicochemical Properties from Near-Infrared Spectroscopy

The electronic adsorption edge of the crude oils, asphaltenes and resins all obey the Urbach phenomenon with slightly different Urbach width due to differences in the chromophore population distribution [53]. Thus, it should be possible to relate the amount of asphaltenes to the location and slope of the electronic absorption edge. Asphaltenes also anti-correlate to some degree with the magnitude of the overtone at 1725 nm due to their low H content. The asphaltene content has previously been successfully modeled with NIR spectra [27]. As an initial approach, plotting the optical density (OD) difference between two frequencies which do not exhibit considerable vibrational signatures defines the amount of aromatic chromophores in the crude oil. By reporting the OD difference instead of the OD at one single wavelength, contributions to a constant baseline increment (large scatterers) are removed. Figure 12 (left) shows the amount of asphaltenes as a function of the OD difference. The five samples: 4, 6, 13, 20, 30 appeared as outliers. By comparing the OD difference with the optical density at a single wavelength, the reason for this became clear (Figure 12, right). The five samples had clearly scattering contributions which affected the slope of the NIR profile. Sample 30 had precipitated asphaltene particles whereas sample 4, 6, 13 and 20 had residual water droplets. The remaining samples followed the Urbach phenomenon with respect to absorption in the NIR region. The results presented here, and previous applications of on-line NIR spectroscopy (see Results section) show the power of NIR spectroscopy in determining the state of the crude oil system and other physicochemical properties (density, viscosity, GOR, total aromatics ASTM D 5186 etc.) which will have a severe effect on the efficiency of petroleum operations (compatibility of commingling streams, water and oil quality, flow assurance).

Modeling the Stability of w/o Emulsions

Sedimentation stability of concentrated (w = 0.3) w/o emulsions has been treated theoretically and experimentally by Dukhin et al. [1]. Although the simple gravitational separation experiments showed a clear influence of the viscosity and density of the continuous phase, the dynamic behavior of polydisperse emulsions is generally very complicated. However, given certain system properties (viscosity, density, droplet size distribution, Re number etc.) primitive models may be used to predict the sedimentation rate. Here, the results from electrical-induced separation will be discussed in more detail. As mentioned previously, the experiments with electric fields revealed a pronounced dependence of viscous forces to the observed CEF values. Chiesa et al. [51] performed theoretical and experimental investigations of the role of viscosity on electrocoalescence phenomena. The viscosity of the continuous phase directly influences the drag force and the film-thinning force and indirectly the coalescence rest time due to lower impact velocity of droplets in a viscous medium. The drag force was corrected for internal circulation induced in the droplet. Chiesa et al. [48] modeled the electric forces acting on the water droplets as a function of their separation with expressions for the surface charge distribution by Davis

[54]. For a system of two droplets, analytical expressions can account for the droplet kinematics in an electrical field. However, a multibody system cannot be explained analytically. As an initial approach, the hydrodynamic resistance can be explained by Stokes law, whereas the point-dipole approximation can account for electric forces acting on spherical droplets at large drop separations (h/r >> 1) lying on similar field line. By also assuming a monodisperse emulsion and an initial repartition of the drops on the vertices of a cubic lattice, the characteristic time of droplet approach can be explained according to Atten [5] as:

theo

8 = 15 E02

5/ 3 1 6w

(1)

By taking into account that the electric field magnitude was increased linearly with time at a rate, dE0/dt, the characteristic time of droplet approach takes the form:

theo

8 dE = 0 5 dt
1/ 3

2 / 3

5/ 3 1 6w

1/ 3

(2)

The characteristic time is modeled by the viscosity () and the permittivity ( ~ 30) of the continuous phase and the water cut (w = 0.3). Although very simplified, this approximation could possibly capture the most underlying effects, the drag on droplets (point-dipoles) in an electric field. The corresponding experimental time, exp, followed directly from the CEF value and the field rate, dE0/dt, as presented in Figure 10. The linear dependence of exp and (dE0/dt)-2/3 indicated some success in the approach (not showed here). Figure 13 shows the experimental vs. theoretical time in seconds for 4 different field rates. The results from experiments at field rate 0.02 kV/cm are not shown, but have exactly the same features as for 0.04 kV/cm. In every subplot, the maximum value of the ordinate axis represents the limit of the experimental setup (E0 4kV/cm). Therefore, the emulsions which were too stable to be separated hold this maximum time. The first apparent discrepancy between theoretical and experimental time is the greater magnitude of the experimental values, about 10 times greater than the theoretical. However, this observation should not give rise to concern. The primitive model does not account for the final approach of droplets, i.e. film thinning forces, neither concentrated and polydisperse systems. Compared to the velocity of droplets in very dilute emulsions as predicted by Stokes law, concentrated emulsions are generally characterized by a reduced velocity below the Stokes rate [55]. A simple test of a model w/o emulsion (surfactant-free, 1000 mPas) showed also characteristic time much greater than expected from equation 2. Therefore, the large experimental times could not be explained by a significant contribution from a barrier to the coalescence step as may be anticipated from classical film stabilization mechanisms in

petroleum emulsions. In fact, a very low electric field magnitude was sufficient to destabilize the w/o interface. The time to break the emulsion of crude oil no. 6 at constant electric field (0.4-4.0 kV/cm) magnitude followed exactly the same trend as experiments performed at increasing field strength (Figure 14). In addition to an underestimation of the characteristic time by a factor of 10, the relationship between the characteristic time and E02 was remarkable. Still, as apparent from the experiments at lower field rates (0.002 - 0.008 kV/cm) in Figure 10, the most viscous emulsions did not follow the same trend as the remaining samples. These water-in-heavy oil emulsions were destabilized sooner than expected from the behavior of the remaining samples. Due to extreme viscosity of the oil phase, the emulsions were generally more difficult to emulsify than the less viscous samples. Broad droplet size distribution could have explained the electrical disruption from the disintegration of a sufficient number of big drops, independently of the evolution of the emulsion. However, PFGSE-NMR results (Table 2) proved that the droplet size distribution of the most viscous samples were not different from the average sample. Due to great amount of heavy constituents, the viscous crude oils may have different electrochemical properties than the lighter crude oils. Equation 1 and 2 only account for the permittivity of the continuous phase which was set to a constant: 30 (permittivity in vacuum). We acknowledge that the electrochemical properties of the crude oil medium may differ considerably among the 30 crude oils in the sample set. This will be a task for a future study. Even though the primitive model explains the results much better than expected, a hypothesis for the discrepancy of the most viscous samples is now presented. The primitive model in equation 2 describes to a certain extent, through a simplified force balance, the dielectrophoretic effect and the resulting drag force on spherical water droplets. Even by assuming monodispersity of the emulsion, perfect spherical and noncharged droplets, no contribution from interfacial dynamics of surfactants, and insignificant contribution from thermal [7] or gravitational forces, equation 2 can only model droplets separated by a distance much greater than their radius. When droplets approach each other the point-dipole approximation does not correctly account for the dielectrophoretic phenomena and film thinning forces start to be effective. At zero electric field, filmthinning forces [56] (F(dh/dt)r2/h) completely decelerates the motion of two colliding droplets (velocity approaches zero). In an electric field, the last stage of the droplet approach is defined by the magnitude of electric and film-thinning forces. Chiesa et al. [51] presented experimental results of the dynamics of droplets in an electric field, compared to numerical predictions with the proper mathematical framework. With increasing field magnitude, the magnitude of the dipole forces dominated over the viscosity dependent forces and the droplets accelerated upon collision. The experimental time for shortcircuiting is then dependent on the final approach which may be slow for low background field magnitudes (E0) where film-thinning forces dominate, but fast for high background field magnitudes where dipole-forces dominate. As E0 was high at drop impact in the heavy

oil emulsions, the relative time required for the film-thinning process may have been reduced.

CONCLUSIONS
From the results presented here and previous applications, NIR spectroscopy seems to be an efficient tool for offshore hydrocarbon analysis. NIR successfully captured components which contributed to scattering of the incident light, i.e asphaltene particles and water droplets. The strong signature of electronic absorption from aromatic moieties (Eo~2317 cm-1) could predict the amount of asphaltenes in the crude oils. As expected, a strong covariance between several physicochemical properties was observed. Interfacial characterization proved that the w/o interfaces were predominantly of soluble character with near-equilibrium interfacial tension in the 5-24 mN/m range at 25oC. Droplet size determination with PGFSE-NMR revealed that the emulsification process generally resulted in narrow drop size distributions centered at about 6 micrometer diameters. The influence of changing the water cut from w = 0.3 to w = 0.1 seemed to be small. Some light crude oils produced w/o emulsions with considerably larger and less stable droplets, assumable due to insufficient access to stabilizing components during the emulsification process. The stability to gravitationally-induced separation showed strong dependence on viscosity of the continuous phase and indicated stability to coalescence of certain particle rich samples (asphaltenes and wax). However, the experiments did not provide sufficient information about the viscous samples due to extreme stability to sedimentation. We will work further in the direction of rheological assessment of long-term stability (flocculation and coalescence) of these emulsions. To a large extent the results from critical electric field experiments could be explained by a simplified model taking drag forces and dielectrophoretic forces into account. The drag force was modeled with Stokes law whereas dielectrophoretic forces were approximated with a point-dipole at the center of the water droplets. Given enough time, water-in-heavy oil emulsions could be destabilized even at very low electric field magnitude (d.c.). When droplets approach each other in an inhomogeneous electric field, strong dielectrophoretic forces disintegrate the films and result in coalescence. The relative contribution from film stability to the overall emulsion stability may therefore be very different in a gravitational field compared to that in an electrical field.

ACKNOWLEDGEMENTS
We thank the joint industrial program: Flucha III, Particle-Stabilized Emulsions/Heavy Crude Oils, and its industrial sponsors for financial support. A.H. gratefully acknowledges support from the Electrocoalescence II project by NFR (The Norwegian Research Council) and industry. People at Statoil R&D centre in Trondheim analyzed the crude oils with

respect to acidity (TAN), density, and molecular weight - cheers! We want to thank Pierre Atten (LEMD, CNRS, France) and Gunnar Berg (Sintef Energy Research, Electric Power Technology, Trondheim) for helpful discussions on the interpretation of electrocoalescence results.

REFERNCES
[1] S. Dukhin, J. Sjblom, . Sther, An Experimental and Theorethical Approach to the Dynamic Behavior of Emulsions, in Emulsions and Emulsion Stability, J. Sjblom, Editor. 2005, Taylor and Francis: New York. J. D. Friedemann, Design Criteria and Methodology for Modern Oil-Water Separation Systems, in Emulsions and Emulsion Stability, J. Sjblom, Editor. 2005, Taylor and Francis: New York. J. S. Eow, M. Ghadiri, Colloids and Surfaces A-Physicochemical and Engineering Aspects 215 (2003) 101. J. S. Eow, M. Ghadiri, A. O. Sharif, T. J. Williams, Chemical Engineering Journal 84 (2001) 173. P. Atten, Journal of Electrostatics 30 (1993) 259. L. E. Lundgaard, G. Berg, S. Ingebrigtsen, P. Atten, Electrocoalescence for OilWater Separation: Fundamental Aspects, in Emulsions and Emulsion Stability, J. Sjblom, Editor. 2005, Taylor and Francis: New York. D. J. Klingenberg, F. van Swoi, C. F. Zukoski, Journal of Chemical Physics 91 (1989) 7888. O. C. Mullins. Molecular Structure and Aggregation of Asphaltenes and Petroleomics. in SPE Annual Technical Conference and Exhibition. 2005. Dallas, Texas, U.S.A.: Society of Petroleum Engineers. A. G. Marshall, R. P. Rodgers, Acc. Chem. Res. 37 (2004) 53. S. Badre, C. C. Goncalves, K. Noringa, G. Gustavson, O. C. Mullins, Fuel 85 (2006) 1. E. E. Johnsen, H. P. Rnningsen, Journal of Petroleum Science and Engineering 38 (2003) 23. B. F. Lutnaes, . Brandal, J. Sjblom, J. Krane, Organic & Biomolecular Chemistry 4 (2006) 616 E. M. Freer, C. J. Radke, Journal of Adhesion 80 (2004) 481. I. B. Ivanov, K. D. Danov, K. P. Ananthapadmanabhan, A. Lips, Advances in Colloid and Interface Science 114-115 (2005) 61. D. Platikanov, D. Exerowa, Symmetric Thin Liquid Films with Fluid Interfaces, in Emulsions and Emulsion Stability, J. Sjblom, Editor. 2005, Taylor and Francis: New York. K. D. Danov, P. A. Kralchevsky, B. Ivanov, Dynamic Processes in SurfactantStabilized Emulsions, in Encyclopedic Handbook of Emulsion Technology, J. Sjblom, Editor. 2001, Marcel Dekker, Inc.: Trondheim. p. 621. L. Goual, G. Horvth-Szab, H. Masliyah, Z. H. Xu, Langmuir 21 (2005) 8278.

[2]

[3] [4] [5] [6]

[7] [8]

[9] [10] [11] [12] [13] [14] [15]

[16]

[17]

[18] [19] [20] [21]

[22] [23] [24] [25] [26] [27] [28] [29] [30] [31] [32] [33]

[34] [35] [36] [37] [38] [39] [40] [41] [42]

J. Czarnecki, K. Moran, Energy and Fuels 19 (2005) 2074. X. Wu, Energy and Fuels 17 (2003) 179. G. Gu, L. Zhang, X. A. Wu, Z. Xu, J. Masliyah, Energy and Fuels 20 (2006) 673. J.-L. Salager, M. I. Briceo, C. L. Bracho, Surface Forces and Emulsion Stability, in Encyclopedic Handbook of Emulsion Technology, J. Sjblom, Editor. 2001, Marcel Dekker, Inc.: Trondheim. p. 455. B. P. Binks, S. O. Lumsdon, Langmuir 16 (2000) 2539. B. P. Binks, C. P. Whitby, Langmuir 20 (2004) 1130. B. P. Binks, S. O. Lumsdon, Langmuir 16 (2000) 3748. B. P. Binks, Current Opinion in Colloid and Interface Science 7 (2002) 21. P. A. Kralchevsky, I. B. Ivanov, K. P. Ananthapadmanabhan, A. Lips, Langmuir 21 (2005) 50. A. Hannisdal, P. V. Hemmingsen, J. Sjblom, Industrial & Engineering Chemistry Research 44 (2005) 1349. A. Hannisdal, R. Orr, J. Sjblom, Journal of Dispersion Science and Technology 28 (2007) P. V. Hemmingsen, A. Silset, A. Hannisdal, J. Sjblom, Journal of Dispersion Science and Technology 26 (2005) 615. P. M. Spiecker, K. L. Gawrys, P. K. Kilpatrick, Journal of Colloid and Interface Science 267 (2003) 178. P. M. Spiecker, K. L. Gawrys, C. B. Trail, P. K. Kilpatrick, Colloids and Surfaces aPhysicochemical and Engineering Aspects 220 (2003) 9. A. Hannisdal, M.-H. Ese, P. V. Hemmingsen, J. Sjblom, Colloids and Surfaces A: Physicochemical and Engineering Aspects 276 (2006) 45. J. Sjblom, G. ye, W. R. Glomm, A. Hannisdal, M. Knag, . Brandal, M.-H. Ese, P. V. Hemmingsen, T. E. Havre, H.-J. Oschmann, H. Kallevik, Modern Characterization Techniques for Crude Oils, Their Emulsions and Functionalized Surfaces, in Emulsions and Emulsion Stability, J. Sjblom, Editor. 2005, Taylor and Francis: New York. E. Vignati, R. Piazza, T. P. Lockhart, Langmuir 19 (2003) 6650. A. P. Sullivan, P. K. Kilpatrick, Industrial & Engineering Chemistry Research 41 (2002) 3389. N. P. Ashby, B. P. Binks, Physical Chemistry Chemical Physics 2 (2000) 5640. B. P. Binks, S. O. Lumsdon, Physical Chemistry Chemical Physics 1 (1999) 3007. G. Gu, Z. Zhou, Z. Xu, J. H. Masliyah, Colloids and Surfaces A: Physicochemical and Engineering Aspects 215 (2003) 141. D. E. Tambe, M. M. Sharma, Journal of Colloid and Interface Science 157 (1993) 244. N. Aske, H. Kallevik, E. E. Johnsen, J. Sjblom, Energy & Fuels 16 (2002) 1287. N. B. Joshi, O. C. Mullins, A. Jamaluddin, J. Creek, J. McFadden, Energy and Fuels 15 (2001) 979. G. Fujisawa, M. A. van Agthoven, F. Jenet, P. A. Rabbito, O. C. Mullins, Applied Spectroscopy 56 (2002) 1615.

[43] [44] [45] [46] [47] [48]

[49] [50] [51] [52]

[53] [54] [55] [56]

M. A. van Agthoven, G. Fujisawa, P. A. Rabbito, O. C. Mullins, Applied Spectroscopy 56 (2002) 593. O. C. Mullins, T. Daigle, C. Crowell, H. Groenzin, N. B. Joshi, Applied Spectroscopy 55 (2001) 197. N. Aske, H. Kallevik, J. Sjblom, Energy & Fuels 15 (2001) 1304. I. H. Auflem, T. E. Havre, J. Sjblom, Colloid and Polymer Science 280 (2002) 695. O. C. Mullins, Analytical Chemistry 62 (1990) 508. O. C. Mullins, Optical Interrogation of Aromatic Moieties in Crude Oils and Asphaltenes, in Structures and Dynamics of Asphaltenes, O. C. Mullins, E. Y. Sheu, Editors. 1998, Plenum Press: New York. B. Balinov, O. Urdahl, O. Soderman, J. Sjblom, Colloids and Surfaces APhysicochemical and Engineering Aspects 82 (1994) 173. F. Groeneweg, F. van Dieren, W. G. M. Agterof, Colloids and Surfaces A: Physicochemical and Engineering Aspects 91 (1994) 207. M. Chiesa, S. Ingebrigtsen, J. A. Melheim, P. V. Hemmingsen, E. B. Hansen, . Hestad, Separation and Purification Technology xxx (2006) xxx. M.-S. Lin, J. M. Chaffin, R. R. Davison, C. J. Glover, J. A. Bullin, A New Suspension Viscosity Model and Its Application to Asphaltene Association Thermodynamics and Structures, in Structures and Dynamics of Asphaltenes, O. C. Mullins, E. Y. Sheu, Editors. 1998, Plenum Press: New York. O. Mullins, S. Mitra-Kirtley, Y. Zhu, Applied Spectroscopy 46 (1992) 1405. M. H. Davis, Rand. Corp. Memorandum RM-3860-PR (1964) T. Tadros, Advances in Colloid and Interface Science 108-109 (2004) 227. R. H. Davis, Phys. Fluids A 1 (1989) 77.

TABLES
Table 1. Physicochemical properties of the crude oils. Property Unit Max Min Median Average Saturates (wt.%) 63.7 25.9 42.1 43.2 Aromatics (wt.%) 53.1 28.6 39.3 39.1 Resins (wt.%) 28.5 6.2 13.2 13.6 C6 insolubles a (wt.%) 12.9 0.1 2.2 2.9 b Toluene insolubles (wt.%) 0.18 0.01 0.02 0.03 Density c (g/ml) 40oC 0.976 0.840 0.912 0.909 MW d (g/mol) 535 204 333 333 Acidity (TAN) (mg KOH/g) 7.5 ~0 1.3 1.8 Viscosity (mPas) 40oC 23400 4.5 59.2 1262 a Asphaltene precipitation in n-hexane (1g:40ml), HPLC separation of maltenes according to Hannisdal et al. [27]. b ASTM D4807-88, 80oC c AP PAAR DMA 48 d Determined from the freezing point depression of benzene.

Table 2. Droplet size from PFGSE-NMR experiments of w/o emulsions. The volume distribution was fitted to a lognormal distribution function with reported mean diameter () and standard deviation (2). Sample w (theo) (10-6 m) 2 (10-6 m) w(meas.)b Average a 0.3 6.2 2.2 0.16 0.01 Average 0.1 6.1 2.1 0.17 0.03 no. 24, t = 0 min 0.3 12.5 0.19 0.39 no. 24, t = 10 min 0.3 17.0 0.37 0.40 no. 24, t = 20 min 0.3 18.2 0.37 0.43 no. 13 0.3 6.9 0.14 a The reported sample variation is represented with two standard deviation in each direction. b The measured water cut in the lower part of the sample cylinder deviated from the theoretical water cut due to sedimentation of water droplets.

FIGURES
Experimental Design - Stability of 30 w/o Emulsions Composition
SARA Water Solids Acidity

Spectroscopy
Infrared Near-Infrared

Bulk Properties
Density Molecular Weight Rheology

Accumulation of knowledge

Modeling

Interfacial Prop.
Interfacial Tension Dilational Relaxation

Stability
Droplet Size Electric Field Gravity

Fig. 1. Experimental design of the current study.

0.030

0.025

0.020 A(1705 cm-1)

0.015

30 13

0.010

0.005

0.000 0 1 2 3 4 5 6 7 8 TAN (mg KOH/g)

Fig. 2. Comparison of crude oil acidity from potentiometric titration (ASTM D664: Total acid number) and from the signature of carboxylic acids in the infrared region.

2.0

1.6 Optical Density at 1600 nm

1.2

0.8 Flocculation onset 0.4

Start of titration

26 29 30

0.0 0.2 0.4 0.6 0.8 Crude oil in pentane (v/v) 1

Fig. 3. Asphaltene stability experiments by continuously adding pentane (12 h) and monitoring the optical density (incl. scattering) with near-infrared spectroscopy.

Fig. 4. Microscopic investigation of crude oil no. 30 (left) and its w/o emulsion (right). Asphaltenes were already precipitated when the crude oil was received from the operator.

1 0.9 0.8 0.7 4 13 6 4 13

Optical Density

0.6 0.5 0.4 0.3 0.2 0.1 20

0 1100 1200 1300 1400 1500 1600 1700 1800 1900 2000 2100 2200 Wavelength (nm)

Fig. 5. Average (3 parallels) NIR spectra of 30 crude oil systems show strong absorption from asphaltenes in addition to scattering and vibration (1950 nm) from residual water.
1 0.9 0.8 0.7

Optical Density

0.6 0.5 0.4 0.3 0.2 0.1 0 1100 1200 1300 1400 1500 1600 1700 1800 1900 2000 2100 2200 Wavelength (nm)

Fig. 6. NIR spectra of 30 centrifuged crude oils (3 parallels of each oil).

10

Optical Density

Eo

7c 31 =2

-1

10

5000

6000

7000

8000

9000

10000

11000

12000

Wavenumber (cm1)

Fig. 7. Same spectra as in Figure 6, but in a semi-logarithmic representation as a function of wavenumbers (cm-1=107/nm). The crude oils generally follow the Urbach phenomenon with respect to electronic absorption with Eo ~ 2317 cm-1.
30 0.7 vol/vol 25 Undiluted

IFT, 55oC, 1.5h (mN/m)

20

15

10

30

0 0 5 10 15 20 IFT, 25oC,12h (mN/m) 25 30

Fig. 8. Figure from Hannisdal et al. [28]. Near-equilibrium interfacial tension values of crude oil-water interfaces at 25oC against those for both diluted (0.7 vol/vol oil in toluene) and undiluted samples at 55oC. Error bars represent one standard deviation at each side of the average value.

35 Diluted oils 30 Undiluted oils 24-27oAPI 24 25 |E (1)| (mN/m) 20 9 9 19 17 17 23 23 24

20 22 11 11 7 26 20 26 30 30

15 10

12-19 API

7 19

5 0 0.0 2.0 4.0 6.0 d |E |/d log (mN/mHz) 8.0 10.0

Fig. 9. Figure from Hannisdal et al. [28]. Magnitude of the dilational modulus at 1 Hz for diluted (0.7 vol/vol oil in toluene) and undiluted crude oils as a function of their frequency dependence.

4.5 4.0 limit 3.5 3.0 CEF (kV/cm) 2.5 2.0 1.5 1.0 0.5 0.0 0.00 6 limit

0.01

0.02 dE/dt (kV/cm s -1)

0.03

0.04

Fig. 10. The figure shows electrocoalescence of 30 water-in-crude oil emulsions (w = 0.3 vol/vol) with a background d.c. field. The ordinate axis represents the maximum electric field magnitude without short-circuiting (CEF) as a function of the rate by which the electrical field magnitude was increased. The samples which are assigned the maximum field magnitude (4 kV/cm) could not be destabilized.

10

56

12

18

24

R+Asp (wt.%) 30 36

42

48

54

10

p As R+

. vs

n De

y sit

60 1.00

0.95

Viscosity (mPas)

M
2

10

10

vs

ns De

ity

0.85

0.80

10 100

150

200

250

300 350 MW (g/mol)

400

450

500

0.75 550

Fig. 11. The plot reveals the strong covariance between physicochemical properties of the crude oils (one sample was removed).
0.7 0.6 0.5 0.4 0.3 0.2 0.1 6 0.0 0 20 30 13 4 0.7 0.6 0.5 0.4 0.3 0.2 0.1 30 0.1 0.2

OD(1300 nm)-OD(1600)

Ele

ctr

on ic

ab s

or pti on
13 4 20 6 0.3 0.4 0.5 OD(1600 nm) 0.6 0.7

20

0.0 40 60 80 100 120 140 0 Asphaltenes (g/L)

Fig. 12. Predicted amount of asphaltenes from near-infrared spectroscopy (left). Scattering contribution in sample 4, 6, 13, 20 and 30 is identified by comparing the optical density between 1300 and 1600 nm with the optical density at 1600 nm (right). These samples could not be predicted.

Density (g/ml)

10

vs

isc .V

it os

y
0.90

2000 1500 exp 1000 500 0 0 500 375 exp 100 200 theo 300 400 0.002 kV/cm s -1

1000 750 exp 500 250 0 0 100 75 exp 50 0.04 kV/cm s -1 25 0 0 50 theo 100 150 0 10 20 50 100 150 theo 200 250 0.004 kV/cm s -1

250 0.008 kV/cm s -1 125 0

theo

30

40

50

Fig. 13. The characteristic time of destabilization (exp) of emulsions in the d.c field is compared to the theoretical (exp) value of droplet approach at four different field rates. The experimental time is given by the measured CEF value and dEo/dt whereas the theoretical time is predicted according to equation 2.

70000 60000

10 x theo

Short circuiting at E 0 = 0.4 kV/cm after 14 hours

50000 40000

exp (s)

30000 20000

10000 0 0.0

theo

0.5

1.0

1.5

2.0

2.5

3.0

3.5

4.0

E 0 (kV/cm)

Fig. 14. The experimental and theoretical times for emulsion no. 6 under the influence of a constant background electric field with magnitude from 0.4 to 4.0 kV/cm.