Solvent Extraction for Heavy Oil from Carbonate AsphaltsExperiment and Molecular Dynamics Study

Abstract

We have systematically studied the effects of solvents such as benzene, toluene, isooctane, and cyclooctane on the selective separation of four heavy oil components (saturates, aromatics, resins, asphaltenes), aiming to clarify the molecular mechanism of solvent extraction as an oil–solid separation method in heavy oil extraction. However, the time-dependent extraction kinetics and the detachment behavior of different saturates, aromatics, resins, and asphaltenes (SARA) fractions from mineral surfaces under solvent action remain insufficiently understood. In this work, solvent extraction experiments were combined with molecular dynamics simulations to analyze solvent-dependent extraction behavior from the perspectives of SARA-selective recovery, molecular diffusion, solvent–solute interaction, and oil–solid interfacial detachment. In terms of experiments, we used SARA component analysis, contact angle measurement, and oil–solid interaction force test to comprehensively characterize the extraction efficiency and interface behavior. In addition, we used Materials Studio 2022 software and COMPASS II force field to simulate molecular dynamics. By analyzing the evolution of molecular conformation, mean-square displacement (MSD), radial distribution function (RDF), and concentration distribution, the diffusion and interaction kinetics of molecular scale are revealed. Based on comprehensive experimental and simulation results, we found that four solvents showed remarkable selectivity: toluene was the best solvent for asphaltenes, while isooctane excelled in separating saturates. The MSD analysis revealed the diffusion order of “saturates > aromatics > asphaltenes > resins”, and the RDF results indicate structure-dependent solvent–solute interactions among different SARA fractions. On this basis, we propose that the solvent extraction behavior can be interpreted by a combined mechanism involving molecular structure matching, solvent–solute interactions, diffusion behavior of SARA fractions, and oil–solid interfacial detachment. This mechanism, together with the established correlative interpretation of solvent-heavy oil component interaction, provides a mechanistic reference for the reasonable design of efficient heavy oil extraction process. Compared with previous studies that mainly focused on solvent screening, extraction yield, or process optimization, this work provides an integrated experimental-simulation interpretation of solvent-dependent extraction behavior by linking SARA-selective recovery, extraction kinetics, oil–solid interfacial detachment, and MD-derived diffusion/RDF evidence in carbonate asphalt systems.

1. Introduction

In the past oil extraction and processing activities, the leakage and discharge of thick oil often seriously polluted the soil. Because heavy oil components remain in soil for a long time, they tend to disrupt the soil structure and hinder plant growth, which poses a threat to both the ecological environment and human health. At the same time, as global energy demand continues to grow and conventional light crude oil reserves are increasingly depleted, , unconventional oil resources including heavy oil, oil sands, and shale oil are gradually becoming important components of the global energy supply, with their strategic significance being increasingly prominent. At present, global unconventional oil reserves exceed 10 trillion barrels, of which heavy oil and oil sands account for more than 60%. These resources are primarily found in Canada, Venezuela, Indonesia, and other countries, becoming the most potential energy type in the next few decades. Heavy oil is mainly composed of four types of components: saturates, aromatics, resins, and asphaltenes (SARA). Due to its complex composition, high viscosity, and poor fluidity, heavy oil often forms a strong adhesion between sandstone, mudstone, and other reservoir rocks. As a result, heavy oil mineral separation has become a major technical challenge in the production and utilization of heavy oil. , Inefficient separation not only results in low recovery rates but also triggers a series of downstream issues such as equipment coking and catalyst deactivation. These problems severely hinder the large-scale development and clean utilization of heavy oil resources. − Teng et al. reviewed the application of different treatment processes for oily sludge and pointed out that efficient separation and remediation technologies are essential for reducing the environmental impact of petroleum-contaminated solids. More recently, Yan et al. critically summarized methods for improving the biodegradability of oily sludge, emphasizing that enhancing pollutant removal and subsequent biological treatment remains an important direction for oily sludge remediation. To improve oil–solid separation efficiency, it is necessary to break down interfacial forces between heavy oil and solids (including van der Waals forces, hydrogen bonds, and π–π interactions) and reduce the viscosity of heavy oil to enhance its fluidity.

In the past, the industrial methods most widely adopted for oil–solid separation included thermal alkaline washing, thermal cracking, and solvent extraction. − Thermal alkaline washing utilizes high-temperature alkaline lye to react with acidic constituents in heavy oil, producing in situ surfactants to lower oil–solid interfacial tension. However, it suffers from high energy consumption, equipment corrosion, and low separation efficiency. − Although thermal cracking can decompose heavy oil molecules into lighter fractions under high-temperature and high-pressure conditions, it imposes strict requirements for equipment and is prone to produce byproducts. − In contrast, solvent extraction gradually becomes research hotspots due to its mild reaction conditions, high separation efficiency, and excellent selectivity, attracting extensive research attention. − For example, Li et al. studied the nonaqueous solvent extraction technology used for enhanced separation and explored its application potential in heavy oil recovery, laying the foundation for optimizing the solvent system. Song et al. studied the kinetic process of using CO₂ bubbles to separate residual solvents from solid waste, which provided insights for solvent recovery during the extraction process. Ogunkunle et al. assessed the performance of solvent-assisted low-salinity waterflooding in cyclic injection mode for Alaska heavy oil recovery, verifying the feasibility of combined processes. Zhai et al. explored the potential of a new water-soluble agent for enhancing heavy oil recovery through pore-scale investigations, enriching the types of solvents used in extraction. Other relevant studies ,, also supplemented research on process parameters and solvent screening, promoting the development of this technology.

Although solvent extraction and its intensification methods had advantages such as high extraction rate of heavy oil and wide applicability to oil sands, they still had the following disadvantages: (1) most organic solvents had the characteristics of low boiling point, high volatility, and high toxicity, so their use had certain risks and toxicity; (2) it was necessary to separate organic solvents from heavy oil by distillation, which consumed a large amount of energy; (3) organic solvents remained in the extracted heavy oil, reducing the recovery rate of organic solvents; and (4) there were residual organic solvents in solid sand particles. The above shortcomings were the main reasons limiting the application of the solvent extraction method and required in-depth research. Additionally, traditional solvent extraction technologies exhibited poor selectivity for heavy oil’s SARA components, , and solvent recovery efficiency was relatively low, leading to increased process costs.

Several representative studies have provided useful references for solvent-based bitumen extractions. Wu and Dabros showed that light-hydrocarbon solvent extraction could obtain bitumen recovery and product quality comparable to conventional hot-water extraction, although the recovery of residual solvent from mineral solids remained necessary. Nikakhtari et al. then emphasized that solvent selection should not be judged only by bitumen recovery but should also consider residual solvent and fine-solid entrainment. On this basis, Pal et al. investigated mixed-solvent systems and suggested that extraction performance depends on the balance among solvent polarity, volatility, and bitumen solubility. From a kinetic perspective, Yuan et al. demonstrated that mass-transfer behavior plays an important role in determining extraction time and process design. Lin et al. further summarized the development of nonaqueous extraction and pointed out that solvent recovery, fine-solid control, and process economics are still major barriers to industrial application. For Indonesian oil sands, Cui et al. reported that toluene gave favorable extraction performance after optimization of solvent ratio, temperature, stirring speed, and extraction time. These representative studies are summarized in Table .

1. Representative Studies on Solvent-Based Bitumen/Oil-Sand Extraction and Related Separation Technologies.

representative study	research focus	innovation	research significance	main conclusion
Wu and Dabros	light-hydrocarbon solvent extraction of Athabasca oil sands, followed by centrifugal or pressure filtration	introduced a nonaqueous route to replace hot-water extraction and combined it with vacuum recovery of residual solvent from filter cake	provided a route to reduce water use, tailings generation, and energy demand in oil-sand extraction	under suitable solvent conditions, bitumen recovery and product quality were comparable to conventional hot-water extraction followed by naphtha treatment, and solvent recovery was relatively easy
Nikakhtari et al.	screened aromatic, naphthenic, bio-based, and mixed solvents by comparing recovery, residual solvent, and fine-solid content	established solvent-screening criteria beyond bitumen solubility alone	showed that high recovery should be balanced with low solvent residue and low fine-solid entrainment	many solvents achieved approximately 95% recovery; cyclohexane showed a favorable combination of recovery, residue control, and product cleanliness
Pal et al.	evaluated solvent mixtures for nonaqueous extraction of Alberta oil sands	extended solvent screening from single solvents to mixture design, emphasizing the balance among polarity, volatility, and bitumen solubility	provided guidance for reducing solvent consumption, tailing residue, and fine-solid entrainment	properly designed solvent mixtures improved extraction and solid–liquid separation, but recovery, volatility loss, and fine-solid entrainment required simultaneous optimization
Yuan et al.	investigated mass-transfer kinetics for solvent-based bitumen extraction from oil sands	developed an improved pseudo-first-order kinetic model incorporating changes in mass-transfer area	supported reactor design, extraction-time optimization, and scale-up analysis	a petroleum ether/dichloromethane system showed strong performance, and the model described bitumen dissolution and transport well
Lin et al.	reviewed nonaqueous extraction of mineable oil sands, including solvent selection, solvent recovery, fines removal, and engineering challenges	integrated fundamental studies, bench tests, and pilot progress in nonaqueous extraction	provided a core overview for understanding the development and remaining bottlenecks of nonaqueous extraction	nonaqueous extraction can reduce water consumption and tailings problems, but commercialization is still limited by solvent recovery, fine-solid entrainment, and process economics
Lin et al.	evaluated an aqueous/nonaqueous hybrid process in which a small amount of solvent was used before low-temperature water-based separation	advanced the process from laboratory work to continuous pilot testing at about 100 kg/h	connected conventional hot-water extraction with fully nonaqueous extraction and suggested a transitional industrial route	low-temperature hybrid extraction obtained recovery and froth quality comparable to or better than high-temperature water extraction, and solvent loss was related to unrecovered bitumen
Khammar and Xu	built a theoretical model for the flow of solvent-diluted bitumen from oil-sand filter cake during centrifugal filtration	used a variable-saturation porous-media model to describe filtration and desaturation stages	provided theoretical support for centrifuge separation and nonaqueous extraction equipment design	bitumen extraction involved filtration and desaturation stages, and centrifugal force and solvent ratio significantly affected liquid recovery
Khammar and Xu	developed experimental methods to determine hydraulic properties of solvent-diluted bitumen flow through oil-sand cake	used online pressure and temperature measurements to infer filtrate volume and flow rate without expensive high-speed imaging	supported continuous centrifugal filtration scale-up and process-parameter optimization	the method determined saturated hydraulic conductivity and unsaturated hydraulic parameters and predicted extraction efficiency under different centrifugal forces
Khammar	used a conical filtering centrifuge for continuous separation of toluene-diluted bitumen from mineral particles	moved batch separation toward a continuous filtering-centrifuge concept	addressed the key solid–liquid separation step needed for nonaqueous extraction industrialization	conical filtering centrifuges were suitable for oil-sand mineral systems with relatively coarse particles and could serve as candidate continuous separation equipment
Khammar et al.	studied centrifugal recovery of solvent-diluted bitumen from different oil-sand mineral cakes	monitored liquid-volume changes during centrifugation using light-transmission profiles and fitted porous-media flow parameters	solved the practical issue of efficiently removing extracted bitumen from mineral solids	high-grade oil-sand cakes gave higher liquid recovery, and cake properties strongly affected the efficiency of large-scale continuous centrifugation
Osacky et al.	examined the mineralogy and surface chemistry of Alberta oil sands and their effects on nonaqueous extraction	explained extraction differences from mineralogical and surface-chemical perspectives	showed that nonaqueous extraction is not a simple dissolution process because mineral surfaces also control bitumen release and fines migration	clay minerals, fines, and surface wettability were important factors affecting recovery and product cleanliness
Liu et al.	studied the stabilization mechanism of bitumen-coated fine solids in organic media after nonaqueous extraction	combined AFM and QCM-D to measure solid–solid forces and bitumen adsorption behavior	explained why fine solids are easily entrained in nonaqueous extraction products	in good solvents, extended bitumen chains can generate steric repulsion and stabilize fine solids; changing solvent composition affects adsorption and aggregation
Osacky et al.	investigated interactions between smectite/bentonite and bitumen in nonaqueous extraction	linked specific clay properties, including pore structure, particle size, swelling, and cation-exchange capacity, with extraction performance	helped explain why different ores show different recoveries under the same solvent conditions	pore volume, particle size, and swelling behavior were key factors controlling bitumen retention and solvent extraction efficiency
Liu et al.	used AFM to characterize heterogeneous bitumen adsorption on fine-solid surfaces after solvent-based extraction	demonstrated nonuniform bitumen distribution and distinguished hydrophilic mineral regions from hydrophobic bitumen regions using force curves	provided microscopic evidence for fine-particle stability and the difficulty of fines removal	fine solids showed chemical and mechanical heterogeneity; adsorbed bitumen regions enhanced hydrophobic interactions and promoted fines transfer into product bitumen
Zhang et al.	reviewed approaches for removing fine mineral solids from nonaqueous extraction bitumen	classified post-treatment removal and in-process reduction of fines entrainment	addressed the product-quality bottleneck in nonaqueous extraction commercialization	polymer additives, antisolvents, water droplets, magnetic particles, electric/magnetic fields, ionic liquids, and switchable-wettability solvents are promising, but cost and scale-up stability remain challenges
Pulati et al.	used choline chloride/urea deep eutectic ionic-liquid analogues with naphtha for Alberta and Utah oil sands	introduced relatively low-cost and more environmentally benign DES-like ionic-liquid analogues for oil-sand extraction	provided a green-solvent option to replace conventional ionic liquids and hot-water extraction	ChCl/urea reduced bitumen-sand adhesion and formed clear phase separation; simulated counter-current extraction achieved more than 90% bitumen recovery
Sui et al.	studied ethyl acetate/n-heptane binary solvent assisted by [Emim][BF4] ionic liquid	combined binary-solvent ratio optimization with ionic-liquid interfacial regulation	showed that ionic liquids can act not only as solvents but also as agents promoting bitumen detachment from mineral surfaces	an ethyl acetate:n-heptane ratio of 3:6 performed well; adding ionic liquid increased recovery from about 83% to about 93%
Sun et al.	enhanced dichloromethane extraction of oil-sand bitumen using [Emim][BF4]	constructed a multiphase separation system including sand/clay, ionic-liquid-solvent mixed phase, and concentrated bitumen phase	targeted fine-clay entrainment, solvent residue, and distillation energy consumption	[Emim][BF4] enabled bitumen recovery of about 92%, reduced clay fines in the product, and allowed recycling of the ionic liquid and dichloromethane
Berton et al.	designed fatty-acid/amine protic ionic liquids for room-temperature nonaqueous extraction of Alberta oil sands	designed ionic liquids based on lipophilic/hydrophilic balance and interfacial interactions rather than simply screening commercial ionic liquids	demonstrated the possibility of low-energy extraction without conventional volatile solvents and with low product solids	[HN888][Oleate] achieved nearly complete bitumen extraction under suitable conditions with product solids below 1%, although ionic-liquid recovery still required optimization
Joshi and Kundu	reviewed ionic-liquid-promoted bitumen extraction from oil sands	summarized experimental studies, computational simulations, ionic-liquid selection principles, and future directions	provided a core review for ionic-liquid applications in oil-sand extraction	ionic liquids can promote bitumen liberation and reduce fine-solid entrainment, but most studies remain focused on limited imidazolium or task-specific ionic liquids, and more modeling and scale-up studies are needed
Holland et al.	used a switchable hydrophilicity solvent, CyNMe2, for extracting bitumen from low-grade, high-fines oil sands	used CO2 to switch the solvent between hydrophobic and hydrophilic states, avoiding distillation recovery	provided a new strategy for solving the high-energy distillation problem in solvent extraction	bitumen removal reached 94–97%, solids retained low residual bitumen, and the solvent could be recovered by carbonated-water washing without conventional distillation
Liu et al.	designed pH-responsive natural deep eutectic solvents based on fatty acids and ethyl lactate for petroleum-hydrocarbon extraction from oil sands	combined pH responsiveness with natural deep eutectic solvents to enable reversible hydrophilic/hydrophobic tuning after extraction	provided a recent reference for green and recyclable solvent design	the HA/EL system showed strong performance and high recyclability; compared with single solvents, NADES improved hydrocarbon stripping and subsequent separation
Yuan et al.	used ChCl/urea, ChCl/ethylene glycol, and ChCl/propanedioic acid deep eutectic solvents to enhance petroleum-ether extraction	demonstrated that DES can function as an intermediate separation layer and reduce clay-fines migration into bitumen	clarified a phase-separation mechanism for organic-solvent + DES synergistic extraction	adding ChCl-based DES improved bitumen recovery and mainly reduced bitumen-sand adhesion while blocking fine-particle entrainment
Wang et al.	investigated liquid dimethyl ether extraction of preheated oil sands	combined local oil sands preheating with room-temperature liquid DME extraction to reduce the overall heating energy	provided a green, easily recoverable, and low-energy nonaqueous extraction route	the optimum bitumen recovery was about 89.8%; preheating oil sands to 40 °C followed by room-temperature DME treatment maintained recovery above 86% and reduced energy consumption
Cui et al.	applied solvent extraction to Indonesian oil sands and optimized solvent ratio, temperature, stirring speed, and extraction time	extended solvent extraction to oil-wet Indonesian oil sands and combined process optimization with bitumen composition/structure analysis	expanded oil-sand solvent-extraction research beyond Alberta oil sands	toluene performed well; suitable extraction was achieved at V(solvent)/m(oil sands) = 3:1, 40 °C, 300 rpm, and 30 min

As summarized in Table , previous studies have made important contributions to nonaqueous bitumen and heavy oil extraction, including solvent screening, solvent-mixture design, extraction kinetics, solvent recovery, solid–liquid separation, mineral surface effects, fine-solid removal, ionic-liquid/deep-eutectic-solvent-assisted extraction, and Indonesian oil-sand extraction. These studies provide important process and mechanistic foundations. However, most prior work mainly focused on total bitumen recovery, solvent selection, process optimization, solvent recovery, or fine-solid control, while SARA-selective extraction, oil–solid interfacial detachment, molecular diffusion, and solvent–solute association were often discussed separately. The present work does not aim to establish a new predictive thermodynamic descriptor beyond classical solubility concepts. Instead, it provides an integrated experimental-MD interpretation of solvent-dependent extraction behavior by combining SARA analysis, extraction kinetic fitting, contact-angle measurement, oil–solid interaction-force measurement, and MD-derived mean-square displacement (MSD)/radial distribution function (RDF)/concentration-distribution results for carbonate asphalt systems.

The root causes of these limitations lie in two aspects. For a long time, existing research has mostly focused on evaluating macroscopic extraction efficiency. Limited attention has been paid to the thermodynamic and kinetic mechanisms in the extraction process. There is also a lack of systematic analysis of key parameterssuch as extraction rate, equilibrium time, solubility parameters, and interaction energybetween heavy oil components and different solvents. At the molecular level, the significant differences in structure, polarity, and molecular weight among SARA components lead to different patterns and intensities of their interactions with solvent molecules. However, how these microscopic interactions determine macroscopic extraction behavior remains unclear. As a result, a combined interpretation that connects component selectivity, molecular diffusion, and oil–solid interfacial detachment remains insufficient.

Although classical solubility concepts, such as the “like dissolves like” principle and solvent–solute compatibility, provide a useful basis for understanding heavy oil extraction, they do not fully explain the extraction behavior of heavy oil from solid-containing systems. In addition to solvent solubility, the extraction process is also influenced by the diffusion behavior of SARA components and by oil–solid interfacial interactions on mineral surfaces. Therefore, in this study, solvent extraction experiments and molecular dynamics simulations were combined to analyze the extraction behavior of heavy oil from the perspectives of recovery performance, component diffusion, molecular interaction, and interfacial detachment.

In response to the above problems, this study investigates heavy oil solvent extraction from both experimental and molecular-simulation perspectives. The main work is as follows:

• (1)The extraction behavior of heavy oil in different solvents was examined by considering both the final recovery and the change in recovery with extraction time. The experimental data were fitted using quasi-first-order and quasi-second-order kinetic models to describe the apparent extraction kinetics and equilibrium behavior. SARA component analysis, contact-angle measurement, and oil–solid interaction-force measurement were also conducted to evaluate the component distribution and interfacial behavior after solvent treatment.
• (2)Molecular dynamics simulations were carried out using Materials Studio 2022 and the COMPASS II force field to assist in understanding the interaction between solvents and different heavy oil fractions. The diffusion behavior, radial distribution characteristics, solvent–solute association, and spatial distribution of SARA in different solvents were analyzed at the molecular level.
• (3)Based on these experimental and simulation results, the solvent extraction behavior is discussed as a combined process involving heavy oil dissolution, SARA component diffusion, solvent–solute association, and oil detachment from mineral surfaces. This interpretation goes beyond a simple “like dissolves like” explanation by also considering diffusion and oil–solid interfacial effects. The results provide a mechanistic reference for understanding solvent-dependent extraction behavior in carbonate asphalt systems and for further improving solvent-assisted heavy oil extraction.

2. Materials and Methods

2.1. Materials and Reagents

The heavy oil samples in this study were collected from the weathered zone of Buton Island Indonesia. Analysis results show that the samples contain 0.12 wt % water, 27.63 wt % heavy oil, and 72.25 wt % solid components. All organic solvents employed were of analytical grade, such as toluene, benzene, isooctane, cyclooctane, and trichloroethylene.

The solvents used in this study exhibit different levels of toxicity and volatility, which should be considered in terms of green chemistry and industrial applications. The relevant acute toxicity and volatility data are summarized in Table . These data indicate that although aromatic solvents such as benzene and toluene show higher extraction efficiency, their higher volatility and potential health risks require careful handling and efficient solvent recovery.

2. The Property of the Four Solvents .

solvent	mouse acute toxicity	boiling point (°C)	vapor pressure	volatility, toxicity
benzene	oral LD50 = 4700 mg/kg	80	21.9 kPa (38 °C)	high volatility, moderate toxicity
toluene	inhalation LC50 = 5320 ppm (7 h)	110–111	30.88 hPa (21 °C)	moderate volatility and toxicity
isooctane	no mouse LD50; rat oral LD50 = 2500 mg/kg	98–99	51 mbar (20 °C)	high volatility, limited mouse toxicity data
cyclooctane	not fully studied	152	5.5 hPa (20 °C)	lowest volatility, limited toxicity data

^aNote (1): smaller LD₅₀/LC₅₀ values indicate higher acute toxicity. (2) Lower boiling points and higher vapor pressures generally indicate higher volatility. (3) Vapor pressure values were measured at different temperatures; therefore, volatility comparisons should consider both the boiling point and vapor pressure. (4) On the basis of these data, the studied solvent system still presents challenges in terms of toxicity and volatility, especially for benzene and toluene. Therefore, solvent recovery and exposure control are necessary for practical applications consistent with green chemistry considerations.

2.2. Experimental Section

2.2.1. Optimization of Extraction Conditions

To determine the optimal operating conditions for solvent extraction, the effects of temperature, stirring speed, and liquid-to-solid (L/S) ratio on heavy oil recovery were systematically investigated by using a single-factor optimization method. During each test, only one operating parameter was varied, while all other conditions were kept constant. The parameter value giving the highest heavy oil recovery was then adopted in the subsequent optimization step, following the same strategy reported in previous solvent extraction studies.

In a typical experiment, 5 g of oil sand and 30 mL of solvent were mixed with the selected solvent in a flask and stirred in a thermostatic water bath. After extraction, the mixture was transferred to a centrifuge tube and centrifuged at 7000 rpm for 5 min to separate the solvent phase containing dissolved heavy oil from the solid residue. The supernatant was collected, and the extraction was repeated until the solvent phase became nearly colorless. The collected extract was then subjected to solvent removal, and the recovered heavy oil was used to calculate the extraction yield.

For the optimization of temperature, the extraction experiments were carried out at 25, 35, 45, and 55 °C while keeping the stirring speed and L/S ratio constant. After the optimal temperature was identified, the effect of stirring speed was evaluated at 100, 200, 300, 400, 500, 600, 700, and 800 rpm under the selected temperature and a fixed L/S ratio. Subsequently, the effect of the L/S ratio was examined at 1:1, 2:1, 3:1, 4:1, 5:1, 6:1, and 7:1 (v/w) using the previously optimized temperature and stirring speed. All extraction experiments were carried out in duplicate, and the average heavy oil recovery was used for evaluation. The optimal process conditions were obtained.

The samples were stored in a dry oven at 10 °C for 24 h to remove moisture. Subsequently, approximately 5 g of the sample was weighed and placed into an Erlenmeyer flask, to which 30 mL of solvent (toluene, benzene, isooctane, cyclooctane) was added. The upper phase was transferred into a flask. After centrifugation, the heavy oil solvent in the centrifuge tube separated into upper and lower phases. The upper solution was pipetted into a previously weighed 200 mL Erlenmeyer flask. The remaining heavy oil solvent mass in the centrifuge tube was extracted three times until the solvent became colorless after extraction. The heavy oil solution was then recovered, and the solvent was removed at 80 °C and a vacuum of 0.098 MPa in a rotary evaporator. Finally, the flask was stored in a dry oven at 80 °C and 0.02 MPa vacuum for 6 h to cool the dried heavy oil sample to room temperature and weigh it (m ₁). The total mass of the heavy oil m ₀ was measured using the solvent extraction method. The heavy oil recovery was as in eq

$$\mathrm{h}\mathrm{e}\mathrm{a}\mathrm{v}\mathrm{y}\mathrm{o}\mathrm{i}\mathrm{l}\mathrm{r}\mathrm{e}\mathrm{c}\mathrm{o}\mathrm{v}\mathrm{e}\mathrm{r}\mathrm{y}=\frac{m_1}{m_0}\times 100\%$$ 1

2.2.2. Solvent Residue Experiment

The residual organic solvent in the extracted sands was determined using an ethanol extraction-gas chromatography method. A known mass of solvent-extracted tailing sample was quickly weighed using a BAS224S analytical balance (Sartorius AG, Germany) and transferred into a sealed glass bottle to minimize solvent volatilization. Then, 35 mL of ethanol was added as the extraction solvent, and the bottle was immediately sealed and shaken for 120 s to allow the residual organic solvent adsorbed on the mineral surface and trapped in the particle pores to transfer into the ethanol phase. After shaking, the mixture was allowed to stand for 30 min until solid–liquid separation was complete. Subsequently, two 2 mL aliquots of the supernatant were collected, mixed with a fixed amount of internal standard, and analyzed by gas chromatography. The remaining supernatant was collected and weighed for subsequent calculation. To ensure sufficient extraction of the residual solvent, a second extraction was conducted by adding another 35 mL of ethanol to the same solid residue, followed by the same shaking, standing, sampling, and GC analysis procedures. The solvent concentration in each ethanol extract was calculated using a pre-established calibration curve, and the total residual solvent content in the tailings was obtained from the sum of the solvent masses detected in the two extraction steps. Gas chromatographic analysis was performed using an AutoSystem XL gas chromatograph (PerkinElmer Inc., USA) equipped with an INNOWAX capillary column (Agilent Technologies, USA), and micropipettes used for liquid transfer were supplied by DLAB Scientific Co., Ltd. (Beijing, China). The residual solvent content was expressed as the mass of residual solvent per unit mass of tailings.

2.2.3. Saturates, Aromatics, Resins, and Asphaltenes Analysis

After the solvent extraction step was finished, the heavy oil was carried out by the SARA composition analysis. At the beginning, 1.0 g of thick oil was accurately weighed out and put into a 100 mL conical flask, with 50 mL of isooctane being added. We made the mixture undergo ultrasonic extraction that was carried out at 50 °C, with a 100 W power output, for a duration of 30 min. The extracted thing was added into the centrifuge tube, and then centrifugation was carried out at 7000 rpm for a time of 5 min. The supernatant liquid was, with great carefulness, poured into a 500 mL conical bottle, and the matter that cannot dissolve was left in the original centrifuge tube. A fitting quantity of fresh isooctane was added into the centrifuge tube, and again the extraction and centrifugation steps were carried out until the extract became without color. All extracts were added together, then the concentration was carried out under 50 °C and 0.08 MPa until the volume became about 5 mL. The remainder that was obtained after centrifugation was dried by vacuum under 80 °C and 0.08 MPa for 4 h. When the cooling was completed until room temperature was reached, the final mass of the residue was weighed; this mass stands for the content of asphaltene. 150 g of neutral aluminum oxide was weighed, which was activated by baking at 500 °C for 4 h and then soaked with isooctane for later use. First, a small pinch of absorbent cotton was plugged into the bottom of the chromatography column, and the plug valve was tightly closed. The mixed aluminum oxide and isooctane were shaken well and poured into the column. After waiting until the aluminum oxide at the bottom completely sank and the interface was stable, a layer of quartz sand was sprinkled on it, keeping the liquid surface that was almost 5 mm thick. The three-component samples that had been concentrated before were carefully added to the column, and then the column was flushed with isooctane, toluene, toluene and methanol 1 to 1, and trichloromethane. Those colorless saturated fractions, yellow aromatic fractions, and black resin fractions were collected with bottles. The collected materials were rotary-evaporated to remove the solvent and then placed in a vacuum drying chamber for 4 h under the conditions of 80 °C and 0.08 MPa. Finally, the saturates, aromatics, and resins obtained were weighed on the scale.

2.2.4. Analysis of Heavy Oil Internal Properties

The elemental analyzer (Thermo Scientific Flash Smart, Germany) measured the elemental composition of the batch of heavy oil samples obtained by solvent extraction, mainly to see what the hydrocarbon ratio of the heavy oil was. In order to further see what the extraction effect was, the viscosity of the extracted heavy oil was measured with a viscometer (MODEL DV-III+, Wells/Brookfield) to understand its flow characteristics; the viscosity of the heavy oil was controlled at a temperature of 60 °C, and one measurement was performed.

2.2.5. Contact Angle Measurement

In order to see whether the solvent extraction has changed the wettability of the sandstone surface, we measured the contact angle. The surface of sandstone is particularly important for recovering heavy oil; if the surface of sandstone can be more water-wet, then the oil recovery rate can usually be improved. After the extraction, we pressed the dried solid samples into thin pellets with a precision tablet press. Then, the deionized water was sprayed on a solid flat surface, and the contact angle was immediately measured with a contact angle meter (model SL200B, produced by Kino in Winslow, Arizona, USA).

2.2.6. Oil–Solid Interaction Force Measurement

We have used the atomic force microscope (AFM, the one produced by the German company Bruker in Karlsruhe) to measure the interaction force between heavy oil and solid particles. In the process of measurement, the speed of the probe was controlled at 2 μm/s. The heavy oil sample extracted from the solvent was first made into a toluene solution with a concentration of 0.5 wt % and then diluted to 0.01 wt %. CaCO₃ microspheres were used as model particles, firstly, because their size distribution can be very uniform, and secondly, their surface morphology is also regular enough to measure the results of reproducible oil–solid interactions. Before use, these microspheres were washed with acetone, ethanol, and distilled water in turn, so that a hydrophilic surface could be obtained; then, they were modified with octyl trimethoxysilane to give them a hydrophobic surface. These modified microspheres were soaked in a heavy oil–toluene solution with a concentration of 0.01 wt % for 4 h and then deposited on a glass slide. Although the main reservoir minerals are carbonate rocks, CaCO₃ particles are widely used as a simplified model substrate in order to study the mechanism of basic oil–solid interaction under controllable conditions.

2.3. Molecular Dynamics Simulation

2.3.1. Building Molecular Models

Using Materials Studio (2020) software, molecular dynamics simulation is carried out under the COMPASS II force field, which is widely used in the field of heavy oil molecular simulation. With the help of the SARA group component method, four heavy oil componentsSARAare constructed. This force field covers both bonded and nonbonded interactions, enabling analysis of the conformational behavior and intermolecular interaction characteristics of the heavy oil solvent systems. The structural formulas of the SARA components used in this study are depicted in Figure .

1.

Molecular modeling of SARA components: (a) saturate fraction represented by C₂₀H₄₂ (MW = 282 g/mol), (b) aromatic fraction modeled with C₄₅H₅₀S (MW = 635 g/mol), (c) resin fraction represented by C₅₀H₈₀S (MW = 713 g/mol), and (d) asphaltene fraction modeled with C₅₀H₄₈O₄ (MW = 713 g/mol). Color coding: sulfur (yellow), oxygen (red).

Saturates are linear alkanes with no branching, which facilitate fast diffusion in the simulation box. Aromatics are polycyclic aromatic hydrocarbons with short alkyl side chains, resulting in a moderate diffusion. Resin materials are big multicycle aromatic molecules that have branch chains and sulfur atoms; therefore, they bring bigger steric obstruction and lead to slower diffusion speed. Asphaltenes are constituted by polycyclic aromatic molecules, which have fewer side chains than resins; this therefore gives them diffusion speeds that are intermediate between those of aromatics and resins. The measured diffusion sequence (saturates > aromatics > asphaltenes > resins) hence reflects the particular molecular descriptions that are utilized in the simulations, instead of opposing the general petroleum chemistry tendencies.

These molecular models provide a representative framework for studying the mobility and interaction behavior of heavy oil fractions in solvent systems, and their structural characteristics directly influence the simulated diffusion and adsorption behaviors. Details of the molecular formulas, molecular weights, and structural features of the SARA models are summarized in Table .

3. SARA Molecular Models and Diffusion Trends.

fraction	molecular formula	MW (g/mol)	structural features	diffusion trend
saturates	C20H42	282	linear alkanes with no branching	fastest
aromatics	C45H50S	635	polycyclic aromatic hydrocarbons with short alkyl side chains	medium
resins	C50H80S	713	large polycyclic aromatic molecules containing branched chains and sulfur atoms	slowest
asphaltenes	C50H48O4	713	polycyclic aromatic molecules with few side chains and oxygen atoms	medium-slow

2.3.2. MD Simulation Process

The amorphous cell module in Materials Studio 2020 was used to construct the molecular systems by packing the four SARA molecular models and solvent molecules into a periodic simulation box according to the specified composition. Molecules were randomly distributed in the box to generate the initial configurations, and three-dimensional periodic boundary conditions (PBCs) were applied to eliminate surface effects and mimic an infinite system. The initial packing density was set to avoid unreasonable molecular overlaps. Under PBC, molecules exiting one side of the box re-enter from the opposite side, which may visually appear as if they “escape” the box.

We have employed the COMPASS II force field, and the initial geometry of our system has been energy-minimized through utilization of the Forcite module with 5000 iterations, for the purpose of removing high-energy configurations and reducing potential energy; hence, it ensures we obtain a stable starting structure. After the minimization operation, the system was slowly brought to an equilibrium state at 298 K through the use of a brief NVT simulation with a controlled temperature-increasing process, which was followed by extra equilibration for the stabilization of temperature and energy distributions before the production operation was carried out. In the process of this equilibration, the system density was continuously monitored and finally reached a stable value, which was kept during the whole 1500 ps production simulation that uses a 1 fs time step; hence, this confirms that the system reached a stable simulation state before the production run. The Berendsen temperature control apparatus was utilized for temperature regulation in all processes. For the analysis of the diffusion behavior of SARA components inside the solvent, the built-in MSD module of the Forcite analysis tool was utilized by us to obtain MSD curves for every component. This method permits quantitative measurement of diffusion speeds and the relative movement of different thick oil components; hence, it guarantees that the outcomes reflect a fully balanced system instead of false products from starting high-energy conditions or close arrangement.

2.3.3. Equilibration and Convergence Analysis

To verify that the simulated heavy oil solvent systems reached equilibrium within the selected simulation time, the evolution of density, temperature, and energy was monitored during the NVT simulations. As shown in Figure S1, the densities of the SARA component–solvent systems fluctuate in the initial stage and then gradually become stable without obvious long-term drift, indicating that the packing structures reached a relatively equilibrated state during the simulation.

The temperature change curves in Figure S2 display that the temperatures of all systems have fluctuation around the goal value of 298 K in the whole process of the simulation; hence, this suggests that the thermostat has effective maintenance of thermal stability. Besides this, the energy change curves in Figure S3 prove that the potential energy, kinetic energy, nonbond energy, and total energy all have an early-stage adjustment and then keep nearly steady in the later period of the simulation; hence, they do not have continuous drift.

Putting all of this together, the density, temperature, and energy profiles indicate that the systems reached stable states within the 1500 ps simulation time. Hence, the following analyses of MSD, diffusion coefficients, and RDFs were carried out upon equilibrated configurations, which give support to the reliability of the molecular dynamics results.

The density, temperature, and energy convergence results shown in Figures S1–S3 confirm that the subsequent MSD and RDF analyses were conducted on equilibrated systems. These results are used to support the reliability of the molecular dynamics analysis, rather than to claim a quantitative thermodynamic model or predictive descriptor.

3. Results and Discussion

3.1. Optimization of Extraction Parameters

In order to further give support to the practical importance of solvent extraction, and in order to provide quantitative proof for technique optimization, the influences of temperature, stirring rate, and liquid-to-solid (L/S) ratio upon heavy oil recovery have been systematically studied, and the corresponding outcomes are shown in Figure . These parameters directly give influence to the mass-transfer efficiency, solvent go through, and dissolve action of heavy oil compositions in the extraction process. Therefore, the optimization of these working conditions is a necessary thing for promoting the extraction effect and for guiding possible industrial use.

2.

Optimization of extraction parameters for heavy oil recovery: (a) effect of temperature; (b) effect of stirring speed; and (c) effect of liquid-to-solid ratio.

According to what Figure a shows, the recovery rate of heavy oil was increased gradually when the temperature went up from 25 °C to 45 °C, and the maximum recovery attained about 93 wt % at 45 °C. This ascension can be ascribed to the reduction of heavy oil stickiness under rising temperature, which strengthens the movement ability of heavy oil molecular groups and helps their spreading into the solvent phase. Besides this, a suitable degree of temperature rising can make intermolecular touching better and quicken the dissolving of the light and medium components. However, when we further increase the temperature to 55 °C, therefore the recovery has a slight decrease. This circumstance therefore may have a connection with the higher volatility that the solvent possesses when temperature is raised, which hence can make the effective quantity of solvent in the extraction system become lower. An over-high temperature can also break the equilibrium between dissolved and not dissolved heavy components, thus leading to a small decrease in recovery rate. These findings tell us that a not too high not too low temperature is good for extraction; thus, excessively high temperature not necessarily can promote the effect.

Just like what is displayed in Figure b, the speed of stirring also possesses a quite remarkable influence upon the recovery of heavy oil. The recovery increased from around 72 wt % when the rotational speed is 100 rpm to about 93 wt % when the rotational speed is 600 rpm. The increase of the stirring speed strengthens the mixing strength between solvent and oil sand, lowers the outer mass-transfer resistance, and facilitates the getting off of heavy oil from the solid surface. Especially, stronger stirring raises the renewal frequency of the liquid membrane that surrounds the solid particles; hence, it improves the solvent getting close to the adsorbed oil phase. Notwithstanding this fact, when the stirring rotational speed exceeded 600 rpm, the recovery degree exhibited a tiny reduction. This perhaps is because excessive stirring brings about unnecessary turbulent flow, which therefore can cut down the stability of phase separation after extraction and centrifugation. Hence, 600 rpm has been confirmed by us as the most suitable stirring speed under the current experimental conditions.

According to what is displayed in Figure c, the L/S ratio exerted a strong influence on extraction efficiency. The heavy oil recovery got a big rise from about 80 wt % when the L/S ratio is 1:1 to around 92 wt % when that ratio is 3:1, and after this, it kept on increasing in a slower way until it got to about 94 wt % when that ratio is 7:1. Under the condition of low L/S ratios, the total amount of solvent is not enough to completely dissolve and carry away the extracted oil components; hence, it leads to limited recovery. Along with the enhancement of the L/S ratio, the solvent can offer a bigger contact area and stronger dissolving ability; therefore, it strengthens the extraction of heavy oil out from the solid base body. However, when the L/S ratio turned into comparatively high, the enhancement in recovery became less obvious, hence indicating that the extraction process step by step got closer to saturation. This indicates that, although raising the solvent use amount can promote recovery rate, the advantage becomes tiny after one certain threshold, which needs to be considered in actual uses from both economic and process efficiency aspects.

Overall, the optimal extraction conditions determined in this study were 45 °C, 600 rpm, and an L/S ratio of 7:1, under which the heavy oil recovery reached approximately 93–94 wt %. Compared with the lower operating conditions tested, the recovery increased by about 11 percentage points with increasing temperature, about 21 percentage points with increasing stirring speed, and about 14 percentage points with increasing L/S ratio. These results provide direct quantitative evidence for the optimization of the solvent extraction process. More importantly, they demonstrate that the extraction efficiency is not governed only by solvent type but also strongly depends on the operating conditions that control solvent–oil contact, diffusion, and phase transfer. Therefore, the optimization of extraction parameters is essential for improving heavy oil recovery and for enhancing the industrial applicability of solvent-based extraction systems.

3.2. Heavy Oil Internal Analysis

Figure a shows the recovery rate of heavy oil obtained after multiple extraction cycles of the four solvents of benzene, toluene, isooctane, and cyclooctane. In general, with the increasing number of cycles, the recovery rates of all of these solvents decreased only slightly. Even after 15 cycles, they could still be maintained at a level of more than 82%, which shows that they have fairly good cycle stability. First, among these four solvents, toluene showed the highest extraction efficiency. At the beginning, its recovery rate reached 93.4%, and after the end of 15 cycles, its value was still close to 90%. The second point is that benzene also showed a relatively high recovery rate, which generally fluctuated within the range of 85–95%. On the other hand, compared with cyclooctane, their efficiency was slightly lower, but they could still maintain a stable performance during the entire cycle. The reason why toluene had such an excellent performance is probably because it has a structure of aromatic rings, which can enhance its interaction with those aromatic components in heavy oil.

3.

Heavy oil characteristics: (a) recovery with different recycle times; (b) SARA component contents; (c) C/H ratio; and (d) viscosity.

Figure b shows the specific situation of the SARA components of the part of the heavy oil obtained after many extraction operations, which uses four different solvents: benzene, toluene, isooctane, and cyclooctane. The values reported here actually represent the mass fraction of each SARA component in the total extracted heavy oil. We can clearly see that the proportion of the asphaltene component in the toluene solvent system will be found to be the highest, reaching 29.27 percentage points by weight, and the performance of the benzene system that follows is also quite close, reaching 29.06 wt % age points. The level, however, if it is changed to two solvents, isooctane or cyclooctane, the value obtained will be relatively slightly less. Let us discuss about the aromatics. In the cyclooctane system, its content is the highest, which can reach 28.07 wt % age points, and in other solvent systems, the measured value is also about the same level. The component of resins will account for about 22.5 percentage points of weight in benzene and toluene, but in the aliphatic solvent system, this value will be slightly lower. No matter which solvent is used, the content of the saturates is undoubtedly the lowest in the whole extracted heavy oil. Combining all these situations, it is not difficult to find that aromatic solvents with benzene ring structures such as toluene and benzene are obviously stronger for those heavier components, that is, we call resins and asphaltenes, and the affinity is obviously stronger; on the other hand, those of aliphatic solvents, their solubility to these heavy components is relatively weak, which can also reflect that the structural characteristics of different solvent molecules will actually affect the distribution of SARA components.

As shown in Figure c, the hydrocarbon ratio of the heavy oil recovered using four different solvents was analyzed. Through measurement, it is found that these values fall within the relatively narrow range of 8.48–8.73, which means that under different solvent systems, the overall composition of the heavy oil finally extracted varies only very slightly. Although these differences are indeed not large, there is still a relatively consistent law that can be recognized by us. The specific order of the size of this hydrocarbon ratio is as follows: first, the toluene system reaches 8.73; second, the benzene system, which is 8.62; third, the isooctane system, which is 8.56; and the last aspect is the cyclooctane system, which is measured at 8.48. Generally speaking, a higher hydrocarbon ratio value means that the extracted product contains a larger proportion of heavy components rich in carbon, which also shows that this solvent is more efficient when used to recover heavy fractions from oil sands. Among these systems, the system of toluene has the highest hydrocarbon ratio, which corresponds to the relatively good extraction efficiency that it has.

Figure d presents the viscosity of the heavy oil recovered using the four solvents under 60 °C. The results show a gradual increase in viscosity depending on the solvent used. The lowest viscosity was obtained for the heavy oil extracted with toluene, measuring 57.2 mPa s. The viscosity of the sample recovered with benzene was slightly higher at 63.9 mPa s. For isooctane, the viscosity further increased to 68.8 mPa s, while the highest value, 76.6 mPa s, was observed for cyclooctane.

To further evaluate the possible influence of residual solvent on bitumen mass determination, the residual solvent content in the extracted bitumen was additionally examined after drying under the same conditions (Figure ). The results showed that the residual solvent contents were 77.34 mg/g for benzene, 64.27 mg/g for toluene, 81.83 mg/g for isooctane, and 86.19 mg/g for cyclooctane. These results indicate that the extent of solvent retention differed among solvent systems, and therefore, the influence of residual solvent should be considered when interpreting the extraction yield data.

4.

Residual solvent content in extracted bitumen after drying.

3.3. Solid Property

For oil-sand systems, heavy oil recovery is governed not only by solvent–oil compatibility but also by the ability of the solvent to weaken oil–solid adhesion and modify interfacial wettability. Therefore, the extraction process should be understood as a combined dissolution–detachment process rather than a simple solubility-controlled process.

The contact angle reflects the wettability of the solid surface after solvent extraction. As shown in Figure , there are significant differences in the contact angles of the solid surfaces after extraction with the four solvents. Herein, toluene has the smallest contact angle (35.7°), indicating that the solid surface exhibits the best wettability after toluene extraction. The contact angle for benzene was slightly larger at 41.5°, while those for isooctane and cyclooctane were 45.2° and 48.3°, respectively. The wettability of the solid surface decreased gradually in the order of toluene > benzene > cyclooctane > isooctane. The wettability of the solid surface after extraction with aromatic solvents was significantly greater than that after extraction with aliphatic solvents.

5.

The contact angles of water drop on the sand surface after solvents.

3.4. Heavy Oil–Solid Interaction

Figure a presents the zeta potential of the heavy oil–quartz sand system after treatment with four hydrocarbons: toluene, benzene, isooctane, and cyclooctane. Zeta potential is commonly used to characterize the electrostatic properties and dispersion behavior of colloidal systems. In the heavy oil phase, all samples exhibited negative potentials, ranging from −19.3 mV to −28.6 mV. From toluene to cyclooctane, the absolute value gradually increases, indicating that the surface charge of heavy oil particles becomes more negative under these conditions. In contrast, quartz sand particles showed a positive zeta potential, from toluene to cyclooctane, and the value increased from 8.6 mV to 21.7 mV.

6.

(a) The zeta potential of the heavy oil and sands; (b) interaction force measured on the CaCO₃ mineral surface.

Figure b shows the interaction forces measured between heavy oil and the CaCO₃ surface after treatment with four different hydrocarbon solvents. The measured values range from 2.26 mN/m to 4.63 mN/m. Among the four solvents, toluene exhibits the lowest interaction force, whereas cyclooctane shows the highest. This difference is likely associated with the molecular structures of the solvents and the intermolecular interactions occurring at the heavy oil–carbonate interface. From the perspective of oil–solid interaction, the magnitude of the interaction force reflects how strongly the heavy oil components adhere to the CaCO₃ surface. A lower interaction force indicates that the oil components are less tightly bound to the mineral surface and therefore can be more easily detached during solvent extraction. In contrast, a higher interaction force suggests stronger adhesion between the oil molecules and the carbonate surface, resulting in a more compact interfacial association and making the separation process more difficult. Overall, the results indicate that toluene is the most effective solvent for weakening the heavy oil–CaCO₃ interaction, while cyclooctane shows the weakest ability to reduce interfacial adhesion.

3.5. Solvent Extraction Dynamics

For the dynamic behavior of solvent extraction, this study adopts two widely used models for analysis, namely, the pseudo-first-order kinetic model and the pseudo-second-order kinetic model. Their mathematical expressions are given as follows:

Pseudo-first-order model

$$q_t=q_{\mathrm{e}}(1-e^{-k_{1}t})$$

In the formula: q _t is the extraction amount (mg/g) at time t (min); q _e is the equilibrium extraction capacity (mg/g); k is the pseudo-first-order rate constant (min^–1); and t is the extraction time (min).

Pseudo-second-order model

$$q_t=\frac{k_2{q}_{\mathrm{e}}^{2}t}{1+k_2{q}_{\mathrm{e}}t}$$

In the formula, k ₂ is the pseudo-second-order rate constant (g/(mg min)), and the meaning of the rest of the parameters is the same as that of the pseudo-first-order model. The pseudo-first-order and pseudo-second-order models are used as empirical formulas to describe the kinetic process of thick oil extraction. Although these two models are commonly used in adsorption research, they can also be used as practical tools to characterize the extraction rate behavior.

Figure gives the kinetic fitting results of the change of the recovery rate of thick oil (in mass percentage wt %) with the extraction time (min) under the action of four solvents, each of which corresponds to a subchart. In each subchart, the black squares are the data of the recovery rate of thick oil measured by the experiment, and the red curve and blue curve represent the fitting results of the pseudo-first-order and pseudo-second-order kinetic models, respectively. For all solvents, the recovery rate of thick oil shows a consistent law of change: it rises rapidly in the initial stage (within the first 10–15 min)mainly due to the rapid occupation of sufficient surface interaction sites on the extraction interface by solute molecules. As extraction proceeded, the growth rate slowed gradually; eventually, the recovery approached 100% and stabilized, which corresponded to the saturation of available interaction regions.

7.

Kinetic fitting curves of heavy oil recovery in solvents (a) benzene, (b) toluene, (c) isooctane, and (d) cyclooctane.

It is worth noting that compared with other solvents, the curve of isooctane (subfigure c) rises more steeply in the initial stage, which is consistent with the higher initial extraction rate reflected in its larger rate constant. In contrast, the recovery curve of cyclooctane (subfigure d) is relatively flat in the early rise, which is also consistent with its smaller rate constant. The experimental data points are highly consistent with the two model curves (this is especially obvious for cyclooctane), which further verifies that these two kinetic models have a good fitting effect, as has been analyzed in the solvent extraction kinetic part.

Table summarizes the fitting parameters of four solvents (benzene, toluene, isooctane, and cyclooctane) under two models, specifically including equilibrium extraction rate (q _e), rate constant (k or k ²), and correlation coefficient (R ²); the latter is used to measure the degree of coincidence between the model and the experimental data.

4. Fitting Parameters of Pseudo-First- and Pseudo-Second-Order Models for Different Solvent Extraction.

models	parameters	benzene	toluene	isooctane	cyclooctane
pseudo-first-order model	q e (mg/g)	0.9191	0.9413	0.8974	0.8936
	k (min–1)	0.1112	0.1381	0.1473	0.0942
	R 2	0.9758	0.9760	0.9857	0.9885
pseudo-second-order model	q e (mg/g)	1.1832	1.1633	1.0951	1.1641
	k (g/(mg min))	0.0855	0.1174	0.1373	0.0735
	R 2	0.9617	0.9545	0.9653	0.9885

The equilibrium extraction amount q _e of four solvents calculated by the pseudo-first-order model is between 0.8936 and 0.9413 mg/g. Among them, the q _e value of toluene is the highest (0.9413 mg/g), indicating that its affinity for the target solute is relatively strong; the q _e value of cyclooctane is the lowest (0.8936 mg/g). Judging from the rate constant k, isooctane is the largest (0.1473 min^–1), followed by toluene (0.1381 min^–1), benzene (0.1112 min^–1), and cyclooctane (0.0942 min^–1). A larger k value means that the initial extraction rate is faster because the extraction interface provides sufficient surface action sites for solute binding in the early stage. The R ² values of the four solvents all exceed 0.97 (between 0.9758 and 0.9885), indicating that the pseudo-first-order model can effectively describe the extraction kinetics process, which suggests that its rate-limiting step may be the initial physical adsorption.

The q _e value (1.0951–1.1832 mg/g) obtained by the pseudo-second-order model is higher than that of the pseudo-first-order model, because the former assumes that chemical adsorption or multilayer adsorption is the dominant mechanism. In terms of q _e value, benzene is the highest in this model (1.1832 mg/g) and isooctane is the lowest (1.0951 mg/g). Looking at the rate constant k ₂, its change trend is basically the same as k: isooctane (0.1373 g/(mg min)) > toluene (0.1174 g/(mg min)) > benzene (0.0855 g/(mg min)) > cyclooctane (0.0735 g/(mg min)), which once again proves that the extraction rate of isooctane is the fastest and cyclooctane being the slowest. Judging from the R ² value, the four solvents are between 0.9545 and 0.9885, and the fitting effect is not badcyclooctane has the highest R ² in both models (0.9885), indicating that its extraction process follows two kinetic mechanisms at the same time. As for benzene and toluene, their R ² values are slightly lower under the pseudo-second-order model (0.9617 and 0.9545, respectively), which means that the effect of physical adsorption may be stronger than chemical adsorption.

The whole extraction process is usually divided into two stages: the rapid initial stage and the subsequent slow balance stage. In the rapid stage, the solute molecules quickly occupy the available surface action sites, resulting in a rapid increase in the q _t value; as the surface action sites gradually saturate, the extraction rate begins to slow down until q _t stabilizes near q _e. Based on the pseudo-first-order rate constant, the time required to reach 90% of q _e can be estimated. The calculation formula is t _90% ≈ 2.303/k. The t _90% of isooctane ≈2.303/0.1473 ≈ 15.6 min, which means that it is the fastest to reach equilibrium; in contrast, cyclooctane takes longer due to the lower rate constant. The time for benzene and toluene to reach the near-equilibrium state is about 20.7 and 16.7 min, respectively. The above results show that the solvent type has a significant impact on the extraction rate and equilibrium timein terms of kinetic efficiency, branched-chain alkanes (isooctane) are better than aromatic hydrocarbons (benzene, toluene) and cycloalkanes (cyclooctane).

Overall, both pseudo-first-order and pseudo-second-order models can better describe the solvent extraction process, among which the pseudo-first-order model has a slightly better fitting effect on most solvents. Isooctane has the fastest extraction rate and the shortest equilibrium time, while benzene has the highest equilibrium extraction amount. Cyclooctane, by contrast, had the slowest kinetics and lowest q _e in the pseudo-first-order model. These differences were related to solvent properties such as polarity, molecular structure, and solute solubility, which affected the mass-transfer rate and interfacial interaction between the solvent and solute.

3.6. Molecular Dynamics Simulation

Before analyzing the MSD and RDF results, the equilibration of the simulated heavy oil solvent systems was verified by monitoring the convergence of density, temperature, and energy, as described in Section and shown in Figures S1–S3 in the Supporting Information.

3.6.1. Saturates, Aromatics, Resins, and Asphaltene Component Conformations of Solvents

The conformational plots in the MD simulations clearly demonstrate molecular structures, spatial configurations, and diffusion states. On the basis of the optimized molecular model, periodic boxes were constructed in Amorphous Cell with 10 molecules of the four-component system and 100 solvent molecules. Molecular dynamics simulations were performed on the successfully constructed boxes for 1500 ps to obtain the optimized model. Six conformational plots were selected from frames 1, 300, 600, 900, 1200, and 1500 for the visualization. Finally, the conformational plots under four different solvents were combined, compared, and analyzed.

As shown in Figure , the image data indicated that the simulation time for these conformational diagrams was 1500 ps. As the simulation progressed, the diffusion effect of the four-component heavy oil molecules gradually intensified, with the diffusion order being saturates > aromatics > asphaltenes > resins. At the start of the simulation, the four-component molecules aggregated with the solvent molecules. As the simulation advanced, the molecules gradually diffused into the solvent. By 1200 ps, molecular diffusion in the system stabilized. By 1500 ps, the diffusion difference between the four-component molecules and the solvent molecules became negligible.

8.

Diffusion frame number plot of SARA components in solvents (A) benzene, (B) toluene, (C) isooctane, (D) cyclooctane: (a) saturates, (b) aromatics, (c) resins, and (d) asphaltenes (yellow = solvents; black = SARA components).

3.6.2. Mean-Square Displacement of Solvent and Saturates, Aromatics, Resins, and Asphaltene Components

In MD simulations, atoms immediately entered a state of perpetual Brownian motion upon detachment from their initial coordinates. Over time, the atomic coordinates and velocity vectors in the three-dimensional space underwent continuous updates, exhibiting highly dynamic stochastic distribution characteristics. To systematically characterize the atomic diffusion behavior in phase space, MSD was introduced as a core physical quantity. By calculating the mathematical expectation of the displacement of each atom in a specific time interval, this method can quantitatively analyze the intensity of thermal motion.

Figure shows the results of MSD spectrum analysis, in which the horizontal coordinates represent the simulation time (0–1500 ps), and the vertical coordinates represent the MSD value (Ų) that reflects the molecular migration capacity. In the benzene–heavy oil system (Figure A), the diffusion efficiency of the four components shows the same law: saturation component > aromatic component > asphalt > colloid. Figure A­(a) shows that the MSD curve of the saturated part rises the steepest, reaching about 280 Ų at 1500 psthanks to its smaller molecular weight (282 g/mol) and simple straight-chain/branched-chain alkane structure, so that their interaction with van der Waals with benzene is more weak, and the spatial resistance is also the smallest. The diffusion rate of aromatic molecule (molecular weight 635 g/mol) ranks second, with an MSD value of 240 Ų at 1500 ps. This is because the aromatic cyclic energy of benzene forms a strong π–π accumulation effect with the aromatic structure in heavy oil, which enhances solubility, but the slightly larger molecular size makes its diffusion rate slightly lower than that of saturated fraction. The MSD value of collagen and asphalt (both molecular weight 713 g/mol) is significantly lowercollagen only reaches 120 Ų. This is because they have a complex macromolecular structure containing heteroatoms (S, O, N), which is prone to aggregation, while the nonpolar hexadal ring of benzene can only act with it through the weak van der Waals force, which limits its dispersion ability. Figure A­(b) reveals an interesting phenomenon: in the first 600 ps, the diffusion rate of aromatic particles is even faster than that of the saturationthis is because the π–π accumulation action rapidly improves the migration ability of aromatic molecules until the influence of its larger molecular size gradually appears; after 800 ps, all curves tend to stabilize and show a steady linear upward trend, indicating that the system has reached dynamic equilibrium. The small fluctuations in the early stage (0–800 ps) reflect the process of structural adjustment of benzene molecules in the gap of components of heavy oil.

9.

Time-dependent curves of the MSD of SARA components and solvents: (A) benzene; (B) toluene; (C) isooctane; (D) cyclooctane; (a) SARA components; (b) solvents (unit: Å).

Similar to benzene, the toluene (Figure B) system also maintains the diffusion order of saturates > aromatics > asphaltenes > resins, but due to the existence of methyl (−CH₃) substituents, there are obvious differences. Figure B­(a) shows that the diffusion rate of aromatics in toluene (MSD value is 260 Ų at 1500 ps) is faster than that in benzenethe methyl group enhances the solvent polarity, thus strengthening the π–π stacking with aromatic components in heavy oil and promoting their dispersion. In contrast, the diffusion of resins and asphaltenes in toluene is slightly slower (110 Ų for resins and 150 Ų for asphaltenes), because the methyl group introduces steric hindrance, making it more difficult to penetrate into the dense clusters formed by these macromolecular components. The time required for the toluene–heavy oil system to reach stability is shorter than that of the benzene system. Figure B­(b) shows that each curve transitions smoothly to linear growth by 800 ps; the diffusion of aromatics can still be observed to be faster than that of saturates within the first 600 ps, but the stronger interactions induced by the methyl group accelerate the system’s transition to equilibrium.

In contrast to aromatic solvents, isooctane, a branched alkane (Figure C), induced dramatic changes in diffusion dynamics. Figure C­(a) shows that saturates now diffused far faster than other components, with an MSD of 500 Ų at 1500 ps, nearly double that in benzene or toluene. Isooctane’s flexible branched structure easily slipped into the gaps between saturates (also alkane-based) and pulled them apart via van der Waals forces, driving exceptional dispersion. In contrast to aromatics, resins, and asphaltenes, their movement is almost stagnant, and the MSD value is as low as 60–120 Ų. The nonpolarity characteristics of isooctane are difficult to disassemble the aggregate structure formed by polar groups in these componentsespecially asphaltenes. Its layered polycyclic aromatic hydrocarbon structure makes it densely stacked between molecules, so it shows a nearly horizontal diffusion curve in Figure C­(a). Figure C­(b) shows that the diffusion rate of isooctane itself is very fast, and its curve is highly coincident with that of saturates, indicating that the two have good miscibility. In contrast, the curves of other components lag far behind, showing obvious phase separation characteristics. The isooctane system quickly stabilized, and there were no obvious fluctuations in the early stagebecause the spatial adaptability of its branched chain structure made it able to quickly form a stable interaction with saturates.

Cyclooctane (Figure D) shows a different diffusion law from other solventsthe diffusion speed of aromatics is faster than that of saturates, and the overall order is aromatics > saturates > asphaltenes > resins. Cyclooctane has a flexible ship-shaped configuration, and this structural feature affects the way it interacts with the components of heavy oil. It can be seen from Figure D­(a) that the diffusion area at 1500 ps is about 230 Ų, slightly higher than the 200 Ų of saturates, while the diffusion of resins and asphaltenes is much slower, only about 70 Ų and 60 Ų, respectively, reflecting the stronger aggregation tendency of these components. Figure D­(b) shows the diffusion curve of cyclooctane, in which the curves of aromatics and saturates cross around 500 ps, indicating that aromatics gradually become the component with stronger mobility in the system. Compared with isooctane, the diffusion curve of cyclooctane fluctuates slightly more in the early stage, suggesting that its molecules have undergone a brief structural adjustment period when interacting with different components. After about 800 ps, the curves tend to be smooth, indicating that the system has entered a relatively stable diffusion state.

3.6.3. Radial Distribution Function of Heavy Oil Extracted by Solvents

To clarify the molecular interactions underlying the diffusion behavior observed in MSD analysis, RDFs were calculated for solvent molecules, SARA components, and solvent–SARA interactions. In the plots, “Total (Solvent)” denotes the intramolecular distribution between solvent molecules, “Total (Solute, SARA)” represents the intramolecular distribution between heavy oil molecules, and “Total (Solvent–SARA)” (blue line) corresponds to the intermolecular RDF between solvent and SARA molecules, reflecting solvent–solute interactions. Although the blue line appears smooth due to averaging over the entire simulation box and production time, its peak positions and intensities vary among solvents because of differences in the molecular structure and interaction type. This explanation provides a clear basis for interpreting the solvent-specific RDF features presented in the following figures.

In molecular dynamics simulation, the RDF is the core tool for analyzing the distribution of micromolecules in the system. The function is represented by g(r). By quantifying the ratio of the probability density of other particles at a given distance r from a particle and the corresponding probability density in the ideal gas state, it accurately describes the spatial distribution law of microscopic ions, which provides an important basis for an in-depth disclosure of the interaction mechanism and structural characteristics between particles in the system. The g(r) curve intuitively shows the degree of aggregation of particles at different spatial distances; the peak value corresponds to the strong aggregation tendency area, and the valley value represents the position with the lowest probability of particle occurrence. This accurate portrayal of spatial distribution provides key evidence for the study of the microstructure and dynamic behavior of materials.

Figure A shows the RDF analysis results of benzene and SARA components. For the saturate subsystem (Figure A­(a)), the Total (Sat) curve shows a sharp peak in the 2–4 Å range, indicating strong self-aggregation between saturate molecules. In the aromatics subsystem (Figure A­(b)), the Total (Aro) and Total (Ben) curves show characteristic peaks at 3–5 Å, suggesting ordered molecular packing associated with aromatic structures. For the resin system (Figure A­(c)), the Total (Re) curve has a small peak around 2.8 Å, which may be related to the heteroatom-containing structures in resins, while the Total (Ben) curve shows a relatively broad distribution, indicating weaker structural ordering of benzene in this subsystem. In the asphaltene system (Figure A­(d)), the Total (Asp) curve maintains a high value even beyond 5 Å, which can be attributed to the ordered layered stacking of its polycyclic aromatic structure.

10.

The RDF curves of SARA components in solvents (A) benzene; (B) toluene; (C) isooctane; (D) cyclooctane; (a) saturates; (b) aromatics; (c) resins; (d) asphaltenes (unit: Å).

Figure B shows the RDF of the toluene–heavy oil system. The methyl group of toluene has a subtle influence on molecular packing. For the saturate subsystem (Figure B­(a)), the Total (Sat) curve shows a sharp peak at 2–3 Å, while the Total (Tol) curve shows a small peak near 2.5 Å, which may be associated with short-range contacts involving the methyl group. In the aromatic subsystem (Figure B­(b)), the Total (Aro) and Total (Tol) curves both maintain sharp peaks at 3–5 Å, indicating ordered aromatic packing related to π–π stacking. For the resin system (Figure B­(c)), the characteristic Total (Re) peak at around 2.8 Å still exists, which may be related to heteroatom-containing structures in resins. In the asphaltene system (Figure B­(d)), the Total (Asp) curve still maintains a high value over a longer distance, reflecting the ordered layered stacking of polycyclic aromatic structures.

Figure C focuses on the isooctane–heavy oil system, where the branched-chain structure of isooctane alters the overall molecular packing behavior. In the saturate subsystem (Figure C­(a)), the Total (Sat) and Total (Iso) curves both exhibit broad peaks in the 4–6 Å range, indicating relatively loose and extended packing dominated by long-range van der Waals interactions, with branched isooctane molecules occupying intermolecular voids. In the aromatic subsystem (Figure C­(b)), the Total (Aro) curve maintains a sharp peak at 3–5 Å, reflecting intrinsic aromatic stacking, while the Total (Iso) curve shows a broader and lower-intensity distribution, suggesting weaker structural ordering of isooctane. For the resin system (Figure C­(c)), the characteristic Total (Re) peak at around 2.8 Å remains, whereas the Total (Iso) curve lacks distinct features, indicating limited organization of isooctane around resin molecules. In the asphaltene system (Figure C­(d)), the Total (Asp) curve retains a long-range high value due to layered stacking of polycyclic aromatic structures, while the Total (Iso) curve remains relatively flat, further confirming the weak association of isooctane within the asphaltene environment.

Figure D shows the RDF of the cyclooctane–heavy oil system. Unlike aromatic solvents, cyclooctane does not possess a planar ring structure but adopts a flexible boat-shaped conformation, which influences its interaction with heavy oil components. In the saturate subsystem (Figure D­(a)), the Total (Sat) curve exhibits clear peaks in the 2–4 Å range, while the Total (Cyc) peaks are mainly concentrated around 2–3 Å, indicating short-range molecular packing; overall, the distributions suggest that cyclooctane is less compatible with the ordered stacking of saturate chains. In the aromatic subsystem (Figure D­(b)), the Total (Aro) curve shows peaks in the 3–5 Å range, while the Total (Cyc) curve partially overlaps in this region, indicating some degree of spatial proximity, although distinct from the ordered stacking typical of aromatic systems. For the resin system (Figure D­(c)), the characteristic Total (Re) peak at around 2.8 Å remains, whereas the Total (Cyc) curve is relatively weak, suggesting limited structural organization of cyclooctane in this subsystem. In the asphaltene system (Figure D­(d)), the Total (Asp) curve maintains a high value over long distances, reflecting the aggregation behavior of asphaltenes, while the Total (Cyc) curve shows only weak features, indicating relatively weak association within the asphaltene environment.

The present work extends the analysis from macroscopic extraction behavior to molecular-scale solvent–solute interactions. By combining diffusion analysis and RDF results, the interaction differences between aromatic/aliphatic solvents and the four SARA fractions can be directly compared, providing a more detailed molecular interpretation of solvent selectivity.

3.6.4. Concentration Profile of Saturates, Aromatics, Resins, and Asphaltene Components and Solvents

Figures and show the interface enrichment behavior of SARA components and solvent molecules on C(100), C(010), and C(001) crystal surfaces. In the distance range of 0–35 Å, the relative concentration of all components fluctuates within a narrow range of 0.3–1.3. The concentration distribution curve along the direction of C(100), C(010), and C(001) shows significant component-dependent differences. Among the four solvents, aromatic solvents benzene and toluene show stronger affinity for aromatics and asphaltenes. This behavior is mainly due to the π–π interaction between the aromatic ring of the solvent and the polycyclic aromatic hydrocarbon structure of heavy oil molecules, which promotes the enrichment of these components near the interface. Compared with benzene, toluene has a more obvious enrichment effect on aromatics and resins: the methyl group on the toluene ring increases the electron cloud density and strengthens its interaction with the heteroatomic groups in the aromatic structures and resins, thus enhancing the solvation and dispersion effects of these components. In contrast, aliphatic solvents isooctane and cyclooctane mainly interact with saturates through van der Waals forces. The distribution of saturates is the most uniform in isooctaneits branched-chain alkyl structure has good compatibility with the long hydrocarbon chains of saturate molecules, so that solvent molecules can penetrate into the saturate molecular chains and promote more uniform dispersion. The dispersion ability of cyclooctane to resins and asphaltenes is weak: cyclooctane adopts a nonplanar boat configuration, and this geometric structure introduces steric hindrance, which reduces its structural compatibility with heavy oil molecules. Consequently, cyclooctane penetrates less effectively into dense aggregates such as resins and asphaltenes, producing a more heterogeneous distribution near the interface. Cyclooctane and isooctane exhibit lower capability to disperse heavy oil components compared with the aromatic solvents benzene and toluene.

11.

Relative concentration curve of SARA components on different crystal faces: (A) benzene; (B) toluene; (C) isooctane; (D) cyclooctane; (a) saturates; (b) aromatics; (c) resins; and (d) asphaltenes (unit: Å).

12.

Relative concentration curves of solvent molecules on different crystal surfaces in various solvents: (A) benzene; (B) toluene; (C) isooctane; (D) cyclooctane; (a) saturates; (b) aromatics; (c) resins; and (d) asphaltenes (unit: Å).

3.7. The Mechanism of Solvent Extraction

Figure shows a schematic diagram of the heavy oil extraction mechanism. The solvent extraction process of heavy oil was essentially the result of the solvent’s interaction with the oil–solid interface, as well as the molecular interactions between the solvent and SARA components. Specifically, the solvent preferentially adsorbed on the oil–solid interface, altered the wettability of the solid surface, and reduced the adhesion between the oil and the solid, thereby weakening the attachment of heavy oil to the solid matrix. Subsequently, the solvent molecules diffused into the oil phase and, based on their molecular structure and compatibility with the SARA components, formed stable solvated systems through structure-dependent intermolecular interactions. Eventually, these solvated SARA components separated from the solid matrix and migrated along with the solvent phase, achieving the separation of heavy oil from the solid. The self-structural characteristics of the solvent determine its ability to interact with different SARA components.

13.

Mechanism of solvent extraction.

Different solvents show different affinities for SARA fractions due to differences in molecular structure, functional groups, and nonbonded interaction patterns. For aromatic solvents, the benzene ring structure can form π–π interaction with aromatic components and asphaltenes, and toluene with a methyl group can also produce weak polar effects, making the treatment effect on asphaltenes more prominent. In saturated hydrocarbon solvents, the branched-chain alkyl structure of isooctane forms a strong van der Waals force with the alkane chains of saturates, while the cyclic structure of cyclooctane can have a certain impact on saturates and some aromatic components with low molecular weight. As for the resin component, due to its unique polarity and spatial structure, it only interacts significantly with aromatic solvents containing weakly polar substituents.

Experimental data corroborated this pattern: oils extracted using different solvents had distinct C/H ratios, indicating variations in the composition of the SARA components in the extracts. Molecular dynamics simulations further indicated structure-dependent nonbonded associations between solvent molecules and SARA components, as reflected by the MSD, RDF, and concentration distribution results, supporting the observed solvent selectivity.

In this study, benzene, toluene, isooctane, and cyclooctane are selected as representatives of aromatic and aliphatic solvents, respectively, to explore the extraction behavior of heavy oil components. It should be noted that some of the solvents are certainly toxic and volatile, which may raise concerns about their environmental sustainability. In this work, they are mainly used to explore the interaction mechanism between solvents and components at the molecular level. In practical applications, solvent consumption and potential environmental impacts can be effectively reduced by establishing solvent recovery and recycling processes. Future work should focus on reducing the IL dosage, improving closed-loop recycling efficiency, and conducting technoeconomic and life-cycle assessments.

It should be emphasized that the proposed mechanism is not intended to demonstrate nonadditive synergy or cooperative thermodynamic stabilization. Rather, the extraction behavior is interpreted as the combined result of solvent–component compatibility, molecular mobility, solvent–solute association, and oil–solid interfacial weakening. Therefore, the term “combined mechanism” is used here to describe the coexistence and consistency of these factors, rather than a thermodynamic synergistic effect. In this sense, the present work does not replace the classical “like dissolves like” principle, but rather extends it by incorporating kinetic and interfacial factors that are particularly important in heavy oil extraction from mineral-containing systems.

Therefore, the innovative contribution of this work lies in bridging the measurable extraction performance with component-level selectivity and molecular descriptors. By combining SARA analysis, kinetic fitting, contact angle/interaction force measurements, MSD/RDF analyses, and concentration profiles, the proposed mechanism explains why toluene favors asphaltene-rich fractions, whereas isooctane shows faster kinetic extraction of lighter saturate-rich fractions.

4. Conclusion

The present research has carried out an analysis on the extraction behavior of four kinds of solvents (benzene, toluene, isooctane, and cyclooctane) for heavy oil by means of experiments and MD simulations. The experimental outcome shows that the kind of solvent possesses a quite notable impact upon the recovery efficiency of heavy oil. In the four kinds of solvents, toluene has displayed the highest extraction speed, keeping an extraction efficiency of about 90% after 15 cycles of use. This indicates that solvents that possess aromatic ring structures possess stronger dissolving ability for heavy oil constituent parts.

The further analysis of SARA fractions has indicated that different solvents display selective dissolution behavior for different heavy oil components. Aromatic kind solvents like toluene and benzene have displayed higher affinity to resins and asphaltenes; hence, they can dissolve these fractions more effectively, therefore, aliphatic solvents thus have displayed stronger preference to saturates. On the other hand, the contact angle and the force measurements of the oil–solid interaction have shown that aromatic solvents have a stronger effect in decreasing the adhesion between heavy oil and the surface of mineral, hence helping the separation of the oil phase from the solid surface.

Molecular dynamics simulations further revealed the microscopic behavior during the extraction process. The diffusion order of heavy oil components in the solvents was found to be saturates > aromatics > asphaltenes > resins. Meanwhile, aromatic solvents could interact strongly with heavy oil molecules through intermolecular π–π stacking, which promoted the dispersion and solvation of resins and asphaltenes.

To make the summary, aromatic kinds of solvents have a better effect than aliphatic kinds of solvents when they are used to extract oil sands. Nevertheless, in the use of industry, extra factors that contain solvent poison, volatility, and matching degree with process equipment must also be carefully thought about. Especially speaking, the high-efficiency recovery of solvents is an essential thing for the lowering of environmental dangers when we carry out large-scale application works. In addition, the potential corrosion properties or matching degrees of materials of the solvent system must be assessed when the industrial processes are being designed. Therefore, future research works should hence place emphasis on the preparation of more environmentally friendly solvents, mixed solvent systems, and optimized recycling methods to attain a superior balance between extraction effect and operation security. In addition, these obtained results indicate that extraction of heavy oil from oil sands is not only controlled by solvent solubility but also by diffusion behaviors of SARA components and by interactions between oil and solid interfaces. Hence, the current research gives a more overall explanation of solvent extraction action in systems that have minerals inside the heavy oil.

Overall, this study should be regarded as a combined experimental and molecular dynamics analysis of solvent extraction behavior, rather than as the development of a new predictive thermodynamic theory. Its main contribution is to connect SARA-selective recovery, extraction kinetics, solvent-dependent molecular diffusion, RDF-based solvent–solute association, and oil–solid interfacial detachment in carbonate asphalt systems. This interpretation supplements the conventional “like dissolves like” understanding by showing that diffusion behavior and interfacial detachment also influence solvent-dependent heavy oil recovery from mineral-containing systems.

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsomega.6c01207.

• Density evolution of SARA components in four heavy oil solvent

• systems during NVT simulations (Figure S1); temperature balance analysis of different simulation systems (Figure S2); and energy balance analysis of different simulation systems (Figure S3) (PDF)

The authors declare no competing financial interest.