Evolution of Crude Oil Properties Under Supercritical CO₂ and Its Implications for Field-Scale Enhanced Recovery in the Mabei Shale Oil Reservoir

Abstract

Supercritical CO₂ injection can mitigate depletion-induced deterioration of shale oil, but the relationship between laboratory-scale fluid-property changes and field-scale recovery remains insufficiently understood. This study integrates time-lapse produced-oil characterization, high-pressure PVT experiments, whole-hydrocarbon gas chromatography, and compositional reservoir simulation for the Mabei shale oil reservoir. From October 2023 to October 2025, the viscosity of dehydrated and degassed produced oil at 80 °C increased from 20.93 to 56.8 mPa·s, accompanied by depletion of light hydrocarbons and enrichment of heavy components, indicating progressive compositional deterioration during depletion. At 106.75 °C and 65.07 MPa, increasing the added CO₂/oil molar ratio from 0 to 80% reduced live-oil viscosity from 9.7841 to 3.8470 mPa·s and density from 0.8326 to 0.7897 g/cm³. Most viscosity, density, and oil-phase compositional changes occurred within the first 6 h in the closed, continuously stirred PVT cell. CO₂ contact preferentially transferred C3–C7 hydrocarbons out of the analyzed oil-rich phase, resulting in relative enrichment of C15–C37 components in the residual oil. The history-matched compositional model showed that CO₂ improved reservoir pressure maintenance and expanded low-viscosity regions primarily along fracture-connected flow paths, although the incremental response became limited between the 20 and 25% cases. Regional compositional analysis further demonstrated that strongly CO₂-contacted residual oil became heavier, whereas the produced oil was enriched in mobilized light and intermediate hydrocarbons. These results establish a consistent laboratory-to-field interpretation of CO₂-induced compositional redistribution and provide a basis for optimizing CO₂-assisted development in deeply buried shale oil reservoirs. CO₂-induced asphaltene precipitation/deposition and the associated permeability impairment were not measured or represented in the numerical model. Therefore, the reported recovery response reflects CO₂–oil property and transport effects in the absence of solid-phase formation damage.

1. Introduction

Shale oil reservoirs are characterized by low matrix permeability, complex pore structures, and strong dependence on hydraulic fractures. Production is commonly dominated by fracture-connected flow during the early stage, whereas long-term productivity increasingly depends on slow matrix-to-fracture fluid transfer and is accompanied by rapid pressure depletion [1,2]. Cyclic gas injection has therefore been investigated as a means of replenishing reservoir energy and mobilizing oil that cannot be effectively recovered through primary depletion [3].

Among the available injection gases, CO₂ is particularly attractive because of its high solubility in crude oil and its potential for simultaneous enhanced oil recovery and geological storage. Dissolved CO₂ can swell the oil phase and reduce its viscosity, while compositional exchange between CO₂-rich and oil-rich phases can mobilize light and intermediate hydrocarbons [4,5]. The magnitude of these effects depends strongly on pressure, temperature, crude oil composition, and the saturation or miscibility state of the fluid [6]. Reservoir-specific PVT characterization is therefore essential for distinguishing viscosity reduction caused primarily by CO₂ dissolution from that associated with compositional redistribution.

CO₂ transport in shale reservoirs differs substantially from that in conventional flooding systems. Injected CO₂ first occupies hydraulic and natural fractures and subsequently enters the low-permeability matrix through pressure-driven flow and molecular diffusion [7]. Consequently, the effective CO₂–oil contact volume and recovery response depend on injection pressure, soaking time, matrix permeability, reservoir heterogeneity, and fracture geometry [8,9]. Laboratory equilibration times obtained in a closed and continuously stirred PVT cell therefore characterize fluid-property stabilization within a confined volume and cannot be directly interpreted as field-scale soaking times.

In addition to viscosity and density changes, CO₂ injection can redistribute hydrocarbons between fluid phases. Laboratory and core-scale studies have shown that CO₂ preferentially mobilizes light and intermediate fractions, potentially leaving the contacted residual oil relatively enriched in heavier components while producing a lighter hydrocarbon stream [10,11]. This distinction is important because residual-oil composition and produced-oil composition describe different fluid populations and should not be interpreted as contradictory observations. CO₂–crude oil systems may also exhibit complex liquid–liquid–vapor behavior, particularly when compositional changes destabilize the heavy fraction [12]. CO₂-induced asphaltene precipitation may affect phase behavior, formation damage, and flow assurance, but its identification requires dedicated onset pressure, SARA, or solid-phase measurements and an appropriate thermodynamic model [13]. These phenomena were not directly quantified in the present study and are therefore considered when defining the scope and limitations of the phase-behavior interpretation. Direct pore-scale and coreflood studies have shown that CO₂-induced asphaltene deposition can plug pore throats and reduce permeability [14,15]. In the present work, no SARA analysis, asphaltene onset test, or deposited solid mass balance was available, and the compositional model did not include a solid-asphaltene phase. Incorporating precipitation and deposition would therefore require unconstrained parameters and could imply a level of predictive capability unsupported by the available data. Accordingly, these effects were excluded from both the experimental interpretation and field-scale simulation; where CO₂ destabilizes asphaltenes, the predicted conductivity and incremental recovery may be higher than those attainable in the reservoir.

Previous studies have generally investigated shale oil production decline, laboratory CO₂–oil interactions, hydrocarbon compositional changes, or reservoir-scale optimization separately [16,17,18]. As a result, the connection among depletion-induced oil deterioration, CO₂-induced live-oil property changes, phase-specific compositional redistribution, and field-scale production response remains insufficiently established. This limitation is particularly relevant to deeply buried lacustrine shale oil reservoirs under high-temperature and high-pressure conditions.

The Mabei shale oil reservoir in the Junggar Basin provides a representative case for investigating these coupled processes. In this study, produced-oil samples collected from the same well at two development stages were used to characterize depletion-induced changes in dead-oil viscosity and whole-hydrocarbon composition. Recombined live oil was then investigated at reservoir temperature and pressure using high-pressure PVT measurements, viscosity and density tests, and whole-hydrocarbon gas chromatography. Finally, a history-matched compositional reservoir model was used to evaluate pressure maintenance, spatial viscosity evolution, regional residual-oil composition, produced-oil composition, and production response under different added CO₂/oil molar ratios. The objectives were to (1) quantify depletion-induced changes in produced-oil properties; (2) determine the effects of the added CO₂/oil molar ratio and contact time on live-oil viscosity, density, and oil-phase composition; and (3) establish a consistent laboratory-to-field interpretation of CO₂-induced compositional redistribution.

2. Experimental Materials and Methods

2.1. Materials

The experimental fluids consisted of dehydrated and degassed crude oil, recombination gas, and CO₂ with a purity of 99.9%. The crude oil sample used for live-oil recombination was collected from the lower sweet-spot interval of the Fengcheng Formation in Well M at a true vertical depth of 4666.0–4691.6 m. After standing for 3 d, the upper oil phase was separated, dehydrated, and degassed. Recombination gas was prepared according to the measured formation-gas composition of the target interval.

Recombined live oil was prepared by charging the dehydrated and degassed crude oil and the prepared recombination gas into the reservoir fluid recombination apparatus at a target gas–oil ratio of 65 m³/m³. The system was heated to 106.75 °C, pressurized to 65.07 MPa, and continuously stirred until the pressure and fluid properties stabilized. The measured saturation pressure of the recombined live oil was 47.8 MPa at 106.75 °C. Because the experimental pressure was higher than the saturation pressure, the initial recombined oil was undersaturated and contained no free hydrocarbon gas phase before CO₂ addition. The principal PVT properties are summarized in

Table 1. PVT properties of the recombined live oil.

Sampling Depth (m)	Reservoir Temperature (°C)	Reservoir Pressure (MPa)	Gas-Oil Ratio (m3/m3)	Saturation Pressure/MPa	Oil Viscosity (mPa·s)	Oil Density (g/cm3)
4666.0–4691.6	106.75	65.07	65	47.8	9.78	0.832

. The whole-hydrocarbon compositions of crude oil under different states are shown in Figure 1.

Two produced-oil samples were collected from Well M to characterize depletion-related property evolution: an early-stage sample from October 2023 and a later-stage sample from October 2025. Both samples were dehydrated and degassed using the same procedure before viscosity and whole-hydrocarbon analyses. These samples were used only for the comparison of surface dead-oil properties and were not assumed to directly represent live-oil viscosity under reservoir conditions.

2.2. Instrument

High-pressure phase behavior and live-oil preparation were conducted using a PVT CELL 240/1500FV analyzer (Sanchez Technologies, Frépillon, France). Viscosity was measured using an RBV 1000 high-temperature and high-pressure falling-ball viscometer (Vinci Technologies, Nanterre, France), and whole-hydrocarbon composition was analyzed using an Agilent 7890 gas chromatograph (Agilent Technologies Co., Ltd., Beijing, China). CO₂ was quantified in a calibrated intermediate vessel at 5 MPa and subsequently displaced into the PVT cell using an ISCO metering pump (Teledyne ISCO, Lincoln, NE, USA). The experimental setup is shown in Figure 2.

2.3. Experimental Protocol and Procedures

Two groups of experiments were conducted. First, the viscosities of the October 2023 and October 2025 dehydrated and degassed produced-oil samples were measured at 50, 80, and 100 °C under atmospheric pressure to characterize depletion-related dead-oil property evolution. Second, high-pressure CO₂–oil experiments were performed using independently prepared recombined live-oil samples at 106.75 °C and 65.07 MPa to evaluate the effects of the added CO₂/oil molar ratio and contact time. The experimental matrices are summarized in

Table 2. Experimental scheme for dead oil measurements.

No.	Experimental Objective	Oil Sample	Temperature/°C	Pressure /MPa	Added CO2/Oil Molar Ratio/%	CO2–Oil Contact Time/h
1	Viscosity–temperature relationship of surface dead oil	degassed crude oil	50	0.1	/	/
2			80	0.1	/	/
3			100	0.1	/	/

and

Table 3. Experimental scheme for recombined live oil.

No.	Experimental Objective	Oil Sample	Temperature/°C	Pressure /MPa	Added CO2/Oil Molar Ratio/%	CO2–Oil Contact Time/h
1	Effect of added CO2/oil molar ratio	Recombined live oil	106.75	65.07	0	/
2			106.75	65.07	10	6
3			106.75	65.07	20	6
4			106.75	65.07	40	6
5			106.75	65.07	60	6
6			106.75	65.07	80	6
7	Effect of contact time		106.75	65.07	80	12
8			106.75	65.07	80	24

.

The reported added CO₂/oil molar ratio (%) is defined as the molar ratio of added CO₂ to the initial recombined oil:

$${\mathrm{R}}_{\mathrm{C}{\mathrm{O}}_2/\mathrm{o}\mathrm{i}\mathrm{l}}={\displaystyle \frac{n_{C{O}_2}}{n_{oil}}}\times 100\%$$

where $n_{C{O}_2}$ is the amount of injected CO₂, and $n_{oil}$ is the estimated mole amount of the initial recombined oil. The average molecular weight of the oil was estimated as 198.6 g/mol from the measured whole-hydrocarbon composition. With an oil density of 0.832 g/cm³, 100 mL oil corresponded to approximately 0.419 mol. The CO₂ amounts required for the designed ratio of 0, 10, 20, 40, 60, and 80 mol CO₂ per 100 mol oil were then calculated accordingly.

For each added CO₂/oil molar ratio, a fresh 100 mL recombined live-oil sample was charged into the evacuated PVT cell. The sample was heated to 106.75 °C and maintained for 10 h, after which the pressure was adjusted to 65.07 MPa. The initial live-oil viscosity and density were measured before CO₂ injection. The calculated amount of CO₂ was then injected using the metering pump, and the cell was continuously stirred to promote mass transfer. Independent oil samples were used for all added CO₂/oil molar ratios to avoid cumulative compositional changes caused by sequential CO₂ addition.

For the added CO₂/oil molar ratio experiments, the fluid was maintained under continuous stirring for 6 h before measurement. Additional tests at the highest added CO₂/oil molar ratio were conducted for 12 and 24 h to evaluate the effect of contact time. The 0% reference denotes the original recombined live oil before CO₂ addition. After the prescribed contact period, the system was considered to have approached laboratory-scale quasi-equilibrium when temperature and pressure remained within ±0.5 °C and ±0.1 MPa and consecutive viscosity measurements differed by less than 5%. Viscosity and density were measured for the equilibrated oil-rich phase. Samples withdrawn for chromatography likewise represented the oil-rich residual phase rather than the total CO₂–oil mixture. Each viscosity measurement was repeated three times using the same equilibrated sample, and the arithmetic mean was reported. The error bars in the corresponding viscosity plots represent one standard deviation. The relative standard deviation was below 5% for all viscosity measurements.

2.4. Laboratory-to-Field Scaling and Numerical Simulation

A field-scale compositional reservoir model was constructed based on the geological, fluid, well, and hydraulic-fracture characteristics of Well M. The model dimensions were 1600 m × 300 m × 40 m and comprised 107,800 grid blocks. Local grid refinement was applied around the horizontal well and hydraulic fractures to improve the representation of fracture-controlled flow and CO₂–oil contact.

The 1400 m horizontal well was intersected by six transverse hydraulic fractures. The main reservoir, grid, fracture, and operating parameters are summarized in

Table 4. Reservoir model and CO₂ huff-n-puff parameters.

Model Parameter	Value	Model Parameter	Value
Model dimensions	1600 × 300 × 40 m	Fracture permeability	3000 mD
Total grid blocks	107,800	Matrix permeability	0.01 mD
Horizontal section length	1400 m	Matrix porosity	4%
Number of hydraulic fractures	6	Initial oil saturation	75%
Fracture spacing	200 m	Initial reservoir pressure	65.07 MPa
Fracture half-length	100 m	Reservoir temperature	106.75 °C
Fracture height	30 m	Initial oil formation volume factor	1.15
Fracture width	3 mm	Added CO2/oil molar ratio	0, 10, 15, 20, and 25%

. The initial reservoir pressure and temperature were set to 65.07 MPa and 106.75 °C, respectively. CO₂ huff-n-puff consisted of 90 days of injection and 20 days of soaking, followed by 15 years of production. Five cases with added CO₂/oil molar ratios of 0, 10, 15, 20, and 25% were simulated.

The compositional fluid model was developed using CMG WinProp. Based on the whole-hydrocarbon composition, the reservoir fluid was represented by nine pseudo-components: CO₂, N₂, C1, C2–C3, C4–C6, C7–C13, C14–C21, C22–C36, and C37+. The Peng–Robinson equation of state was used to describe the phase behavior of the recombined live oil and CO₂–oil mixtures. No solid asphaltene phase, deposition kinetics, pore-blocking function, or permeability reduction module was activated. This modeling choice was made because the available measurements constrain fluid-phase PVT behavior and hydrocarbon composition, but not asphaltene onset, transport, deposition, or formation damage parameters.

The fluid model was constrained by the measured saturation pressure, solution gas–oil ratio, oil formation volume factor, viscosity, and density. CO₂ swelling calculations were then conducted for mixtures corresponding to added CO₂/oil molar ratios of 0–80% at 106.75 °C and 65.07 MPa. As shown in Figure 3, the calculated oil viscosity agreed well with the measured values under different added CO₂/oil molar ratios. Similarly, Figure 4 shows that the calculated oil density also reproduced the measured trend satisfactorily. These results indicate that the fluid model adequately represented the CO₂-induced viscosity reduction and density change under reservoir conditions.

Before CO₂ prediction, the base model without CO₂ injection was calibrated against the first 280 days of production history. The measured bottom-hole flowing-pressure schedule was imposed as the production constraint, while daily oil rate and water cut were used as the history-matching targets. As shown in Figure 5, the simulated oil rate reproduced the main variations and overall magnitude of the historical production data. Figure 6 shows that the simulated water cut also captured the general trend of the field data. The pressure, saturation, and component distributions at the end of the history-matching period were subsequently used as the common initial state for all CO₂ huff-n-puff prediction cases.

3. Results and Discussion

3.1. Evolution of Crude Oil Mobility During Depletion Development

During the 280 d history period, bottom-hole pressure declined from approximately 30 to 22 MPa, while oil rate generally remained between 30 and 60 m³/d and water cut increased from approximately 30% to nearly 50% (Figure 7). Although the well maintained appreciable early-stage production, time-lapse fluid analyses revealed progressive compositional deterioration. As shown in Figure 8, the dead-oil viscosity at 80 °C increased from 20.93 mPa·s in October 2023 to 56.8 mPa·s in October 2025. As shown in Figure 9, the C3–C13 fraction decreased by 10.2%, while the heavy fraction increased by 46.8%, and both $\Sigma C_{13-}/\Sigma C_{14+}$ and $\Sigma C_{21-}/\Sigma C_{22+}$ decreased. These trends indicate preferential production of light and intermediate hydrocarbons during depletion.

Because both samples were dehydrated and degassed using the same procedure, the dead-oil measurements provide a consistent indicator of compositional evolution. They do not directly represent in situ viscosity or an equivalent loss of reservoir mobility, which is also controlled by dissolved gas, pressure, relative permeability, saturation, and fracture–matrix connectivity.

3.2. Viscosity and Density Evolution After CO₂ Addition

Figure 10 shows the changes in live-oil viscosity and density with increasing added CO₂/oil molar ratio at 106.75 °C and 65.07 MPa. The initial recombined live oil had a saturation pressure of 47.8 MPa and was therefore undersaturated before CO₂ addition. As the added CO₂/oil molar ratio increased from 0 to 80%, viscosity decreased from 9.7841 to 3.8470 mPa·s, corresponding to a reduction of 60.7%, while density decreased from 0.8326 to 0.7897 g/cm³. The relative standard deviation of the replicate viscosity measurements remained below 5%.

The viscosity response can be divided empirically into three regions. Between 0 and 20%, the sharp decrease was mainly associated with CO₂ dissolution into the initially undersaturated oil, resulting in swelling, dilution, and weaker intermolecular interactions. Between 20 and 40%, the marginal viscosity reduction became smaller, indicating a gradual decline in the effectiveness of additional CO₂ dissolution. Between 40 and 80%, viscosity continued to decrease, but the response likely reflected the combined effects of further dissolution, dilution, and compositional redistribution. Because the experiments did not independently quantify dissolved CO₂ and transferred hydrocarbons in each phase, the individual contributions of these mechanisms cannot be separated quantitatively.

No distinct macroscopic phase boundary was observed through the PVT cell window under the tested conditions, and the fluid appeared visually homogeneous at the observation scale. This observation supports the use of an oil-rich measurement phase but does not exclude microscopic phase splitting or CO₂-rich domains. The phase interpretation is therefore limited to macroscopic visual observation and the measured bulk properties.

Figure 11 shows that most changes in viscosity and density occurred during the first 6 h of CO₂–oil contact. At the highest tested added CO₂/oil molar ratio, viscosity decreased from 9.78 to 3.85 mPa·s during 0–6 h, whereas only minor additional changes occurred between 6 and 24 h. A similar stabilization was observed for density and oil-phase composition, indicating that the confined, continuously stirred fluid approached a reproducible dynamic quasi-equilibrium after approximately 6 h.

This 6 h period represents laboratory-scale mixing and equilibration within the PVT cell and should not be interpreted as a reservoir soaking time. In a shale reservoir, CO₂ transport from fractures into the matrix is controlled by diffusion distance, effective diffusivity, adsorption, confinement, and fracture–matrix connectivity. A characteristic diffusion time may be expressed as $t_d\approx L^2/D_{\mathrm{e}\mathrm{f}\mathrm{f}}$, but $D_{\mathrm{e}\mathrm{f}\mathrm{f}}$ was not measured in this study; therefore, no direct conversion from the laboratory contact time to field soaking duration was attempted.

3.3. Whole-Hydrocarbon Compositional Response to CO₂ Contact

Figure 12 presents the oil-rich phase sampled after CO₂ contact rather than the total closed system or a produced-fluid stream. Hydrocarbons transferred into a CO₂-rich phase remained in the cell but were not included in the analyzed oil sample.

With increasing added CO₂/oil molar ratio, the C3–C7 fraction in the analyzed oil-rich phase decreased by 16.5%, including reductions of 79.9% and 37.1% in C3 and C4, respectively. In contrast, the relative abundance of C15–C37 increased by 31.5%, and both $\Sigma C_{13-}/\Sigma C_{14+}$ and $\Sigma C_{21-}/\Sigma C_{22+}$ decreased. The corresponding light-to-heavy hydrocarbon ratios are summarized in

Table 5. Light/heavy hydrocarbon ratio of whole-oil composition under different added CO₂/oil molar ratios.

	0%	10%	20%	40%	60%	80%
$\Sigma C_{13-}$$/\Sigma C_{14+}$	0.96	0.93	0.89	0.87	0.86	0.82
$\Sigma C_{21-}/\Sigma C_{22+}$	4.33	4.18	3.88	3.77	3.62	3.48

. These trends indicate preferential transfer of low-carbon-number hydrocarbons out of the analyzed oil-rich phase and relative enrichment of intermediate and heavy components in the residual oil. Because the CO₂-rich phase was not separately sampled and analyzed, the term “extraction” is used here in a phase-transfer sense rather than as a complete component mass balance.

The changes in whole-hydrocarbon composition under different reaction times are shown in Figure 13. The compositional response with contact time was consistent with the viscosity and density results. Most of the decrease in C3–C7 and the relative increase in C15–C37 occurred during the first 6 h, whereas the oil-phase composition changed only slightly between 6 and 24 h. This consistency supports the interpretation that the tested PVT system approached a dynamic quasi-equilibrium after approximately 6 h.

3.4. Field-Scale Pressure, Oil-Mobility and Produced-Oil Compositional Responses

Figure 14 shows that the no-CO₂ case developed an expanding pressure-depletion zone around the horizontal well. Increasing the added CO₂/oil molar ratio progressively reduced the extent of this low-pressure region, with the 20 and 25% cases maintaining the highest pressure levels after 3 years. The limited difference between these two cases indicates diminishing pressure-maintenance benefits at a high added CO₂/oil molar ratio.

Figure 15 shows that CO₂-induced viscosity reduction was concentrated along the horizontal well and hydraulic fracture network. Low-viscosity zones expanded with increasing added CO₂/oil molar ratio, whereas regions remote from dominant fracture-controlled flow paths retained relatively high-viscosity. The limited difference between the 20 and 25% cases indicates that field-scale improvement was constrained by accessible sweep volume rather than by fluid property response alone. The simulated expansion of low-viscosity regions and the associated production benefit therefore do not account for local permeability loss caused by asphaltene deposition. If precipitation occurs in strongly contacted fracture matrix pathways, effective conductivity and incremental recovery could be lower than predicted; targeted CO₂ corefloods, solid detection measurements, and post-flood permeability tests are required to quantify this effect for the Mabei fluid–rock system.

Figure 16 shows the composition of the produced oil stream rather than that of the oil remaining in the reservoir. After 1 year, increasing the simulated added CO₂/oil molar ratio from 0 to 20% increased the C7–C13 and C14–C21 fractions from 29.0 and 41.8% to 31.8 and 44.2%, respectively, while decreasing C22+ from 27.2 to 21.3%. The additional change at 25% was limited. The laboratory and simulation results describe different fluid populations. The PVT analysis characterizes the residual oil-rich phase after interphase transfer, whereas the simulation reports hydrocarbons transported to the production well. Preferential mobilization of light and intermediate components can therefore make the produced stream lighter while leaving contacted residual oil relatively heavier.

Table 6. Simulated compositions of the initial oil, regional residual oils, and produced oil after 3 years of CO₂-assisted production (25%).

Region	C4–C6	C7–C13	C14–C21	C22–C36	C37+
Initial oil	2.2	30.2	43.8	21.0	2.8
CO2 contacted residual oil	0.4	16.7	29.1	40.5	13.3
Weakly contacted residual oil	0.7	22.3	34.4	34.8	7.8
Produced oil	2.0	28.4	38.6	26.8	4.2

compares the compositions of the initial oil, residual oil in regions with different degrees of CO₂ contact, and the produced oil after 3 years of production. In the CO₂-contacted region, the C22–C36 and C37+ fractions increased from 21.0 and 2.8% in the initial oil to 40.5 and 13.3%, respectively, whereas the combined C4–C21 fraction decreased from 76.2 to 46.2%. This indicates that strong CO₂–oil interaction preferentially mobilized light and intermediate hydrocarbons, leaving the local residual oil relatively enriched in heavy components. The weakly contacted residual oil showed an intermediate composition, with C22+ accounting for 42.6%, confirming that the extent of compositional redistribution depended on the degree of CO₂ contact.

In contrast, the produced oil contained 69.0% C4–C21 and only 31.0% C22+ and was therefore considerably lighter than the CO₂-contacted residual oil, for which C22+ reached 53.8%. This difference shows that the light and intermediate hydrocarbons mobilized from CO₂-contacted regions were preferentially transported through fracture-connected flow paths to the production well. This spatially heterogeneous CO₂–oil contact and the associated compositional redistribution are schematically illustrated in Figure 17. The heavying of residual oil in strongly contacted regions and the relative lightening of produced oil are complementary consequences of CO₂-induced compositional redistribution rather than contradictory results.

4. Conclusions

• Between October 2023 and October 2025, dead-oil viscosity at 80 °C increased from 20.93 to 56.8 mPa·s, accompanied by depletion of light hydrocarbons and enrichment of heavy components. These measurements indicate compositional deterioration but do not represent an equivalent loss of in situ mobility.
• At 106.75 °C and 65.07 MPa, increasing the added CO₂/oil molar ratio from 0 to 80% reduced live-oil viscosity from 9.7841 to 3.8470 mPa·s and density from 0.8326 to 0.7897 g/cm³. The response was dominated by dissolution, swelling, and dilution with low CO₂ addition, with compositional redistribution becoming more important with higher addition.
• CO₂ contact caused preferential transfer of C3–C7 hydrocarbons out of the analyzed oil-rich phase and relative enrichment of C15–C37 components. Most property changes occurred within the first 6 h in the stirred PVT cell; this period represents laboratory equilibration rather than reservoir soaking time.
• Reservoir simulation showed that CO₂ maintained pressure and expanded low-viscosity regions mainly along the horizontal well and hydraulic fracture network, while regions outside the effective contact volume retained relatively high viscosity. The difference between the 20 and 25% cases was limited, indicating diminishing field-scale benefits at high added CO₂/oil molar ratio. After 3 years, strongly CO₂-contacted residual oil contained 53.8% C22+, whereas the produced oil contained 31.0% C22+. Thus, local residual-oil heavying and produced-oil lightening are complementary consequences of the preferential mobilization and fracture-controlled transport of light and intermediate hydrocarbons. CO₂-induced asphaltene precipitation/deposition was outside the measured and modeled scope; thus, the field-scale gains reported here do not include possible organic solid plugging and should be reassessed when reservoir-specific asphaltene onset and permeability damage data become available.