**1. Introduction**

While abundant, shale formations are unique and serve as unconventional hosts for hydrocarbons [1]. In recent times, a shale revolution has redefined the energy equation in the world [2]. Although the hydrocarbon content of these reservoirs was known for long, the economic developments of these reservoirs became possible only by the coupling of multi-stage hydraulic fracturing with horizontal drilling [3–5]. This coupling overcomes the ultra-low permeability nature of the shale by providing highly conductive pathways connecting the natural fractures. However, the current recovery factors are less than 10%, even with the most efficient completion schemes [6,7].

Enhanced oil recovery techniques have been widely-used in conventional reservoirs [8,9]. Some provide pressure maintenance and others improve the hydrocarbon mobility by tuning the interfacial and physical properties of the reservoir fluids [10]. In contrary to conventional reservoirs, the interconnectivity between shale wells does not always warrant a field-wide design of EOR operations [11,12]. Alternatively, single-well approaches have been proposed [13,14]. Zhang et al. [15] numerically evaluated the efficiency of cyclic methane injection considering both the molecular diffusion and nano-confinement effects. Their results sugges<sup>t</sup> implementing a cyclic injection strategy during the early stages of production. On the other hand, Meng et al. [16] evaluated the efficiency of carbon dioxide cyclic injection, and they experimentally observed more than 30% increase in the recovery. Assef and Pereira Almao [17] further economically optimized the cyclic gas injection operations. Currently, the Hu ff and Pu ff technique is a commonly-used technique that involves injecting the EOR fluid into the well, shutting down the well for a soaking time and resuming the production from the well [18,19]. While the EOR fluid could be gas, water or surfactant, Sheng and Chen showed that gas flooding yields better performance compared to water flooding [20].

The optimization of the gas injection has been the focus of several studies [21–23]. For instance, Ho ffman [24] studied the feasibility of various gases for injection in Bakken and reported better performance for the miscible cases. In addition, they encouraged the implementation of EOR since significant oil could be recovered regardless of the gas type. Sheng [25] favored lean gases to enhance the liquid hydrocarbon from shale condensate reservoirs. In addition, Fragoso et al. [26] proposed a holistic approach to develop fields with multiple fluid-type windows, like Eagle Ford and Duvernay, where the gas production is used to enhance the liquid production from the liquid and condensate windows.

Experiments and lab studies have been providing numerous insights about the potential of shale EOR [27,28]. Tovar et al. [29] showed promising results for CO2-EOR (18–55% recovery factors). In addition, Gamadi et al. [30] observed up to 85% improvement in the oil recovery from Eagle Ford samples using cyclic CO2 injection. On the other hand, Yu and Sheng [31] studied the N2 performance in Eagle Ford core and found better results for a longer flooding time and a higher injection pressure. Nguyen et al. [32] used microfluidic experiments to probe the CO2 and N2 performance and attributed the superior CO2 performance to its miscibility characteristics. Alharthy et al. [23] found that both CO2 and Natural gas liquids, C1-C4+, have a similar e fficiency when extracting oil through countercurrent flow from the matrix instead of displacing oil. Hawthorne et al. [33] identified the molecular di ffusion of CO2 as the main mechanism for oil recovery in Bakken samples.

Following the promising lab results, field pilots of CO2-EOR have been conducted in Bakken and Eagle Ford [34]. Liu et al. [35] designed a case study to evaluate the potential of CO2-EOR in the Bakken formation where promising results were observed. Pankaj et al. [36] devised an Eagle Ford case study to investigate the CO2-EOR potential where they refuted the need for infill wells and reported an extra 9% increase in the recovery factor. Kerr et al. [37] developed an Eagle Ford case study to engineer single- and multi-well CO2-EOR techniques and reasonable agreemen<sup>t</sup> with the field results was reported. Although promising results were observed for Eagle Ford, fruitless results were reported for Bakken [28]. Rassenfoss [38] attributed these contradictory results to the reservoir containment of each field. For CO2-EOR to work, su fficient and sustainable contact between CO2 and the matrix should be achievable. While Eagle Ford pilots provided the containment required for CO2 to work, the fractured nature of the Bakken formation obstructed the containment. However, the fractured nature of the Bakken formation allowed multi-well EOR techniques to be implemented. Todd and Evans analyzed Hu ff and Pu ff and continuous injection pilots in Bakken, and found that continuous injection was favored for CO2-EOR operations [34].

Molecular simulations have been widely used to probe the gas injection in shale at the molecular scale [39–41]. For instance, Wu et al., studied the displacement mechanisms of CO2, N2 and CH4 [42]. They attributed the slow breakthrough of CO2 to the superior adsorption characteristics. By contrast, N2 exhibits a fast breakthrough and a wide front. Wang et al., confirmed the adsorption selectivity of kerogen pores to CO2 and CH4 to range from 2.53 up to 7.25 [43]. Sun et al., studied the di ffusion of methane and CO2 in kerogen and observed that the di ffusion of dissolved molecules was smaller than that of those adsorbed, which were smaller than the bulk [44]. Liu et al., studied the oil flow displacement by CO2 in silica nano channels and recommended a small injection rate to assure that the miscible front is developed [45].

Zhou et al. found that the pressure drawdown is e fficient in extracting the lighter hydrocarbons and the CO2 is more e fficient is stripping the heavier hydrocarbons from the middle of the pore [46]. Zhang et al., reported the CO2 behavior in organic and inorganic pores [47]. They observed that C2 and C3 remain adsorbed during the primary production regardless of the pore type compared to C1. On the other hand, CO2 injection disrobes the hydrocarbon from the pore surface, which enhances the extraction of heavier components.

Fluid behavior and phase properties deviate from the bulk behavior under confinement [48–50]. Consequently, confinement a ffects the e fficiency of the gases to extract liquid hydrocarbons and requires revisiting both the design and the operation to account for the confinement e ffects. We used molecular dynamics to evaluate the performance of various gases for enhancing the hydrocarbon recovery from shale formations. We organized the rest of this article as follows: the methodology section briefly presents the modeling approach and the simulation details, and the results section discusses the impact of confinement and operational parameters on the performance of di fferent gases. The main findings are summarized in the conclusions section.
