**An Experimental and Numerical Study of CO2–Brine-Synthetic Sandstone Interactions under High-Pressure (P)–Temperature (T) Reservoir Conditions**

#### **Zhichao Yu 1,2, Siyu Yang 1,3,\*, Keyu Liu 4, Qingong Zhuo 1,2 and Leilei Yang <sup>5</sup>**


Received: 28 June 2019; Accepted: 12 August 2019; Published: 15 August 2019

**Abstract:** The interaction between CO2 and rock during the process of CO2 capture and storage was investigated via reactions of CO2, formation water, and synthetic sandstone cores in a stainless-steel reactor under high pressure and temperature. Numerical modelling was also undertaken, with results consistent with experimental outcomes. Both methods indicate that carbonates such as calcite and dolomite readily dissolve, whereas silicates such as quartz, K-feldspar, and albite do not. Core porosity did not change significantly after CO2 injection. No new minerals associated with CO2 injection were observed experimentally, although some quartz and kaolinite precipitated in the numerical modelling. Mineral dissolution is the dominant reaction at the beginning of CO2 injection. Results of experiments have verified the numerical outcomes, with experimentally derived kinetic parameters making the numerical modelling more reliable. The combination of experimental simulations and numerical modelling provides new insights into CO2 dissolution mechanisms in high-pressure/temperature reservoirs and improves understanding of geochemical reactions in CO2-brine-rock systems, with particular relevance to CO2 entry of the reservoir.

**Keywords:** CO2 sequestration; physical simulation; Numerical modelling; dissolution; precipitation

#### **1. Introduction**

Carbon dioxide emissions from fossil-fuel combustion are projected to increase from 13 Gt yr−<sup>1</sup> in 2010 to 20–24 Gt yr−<sup>1</sup> in 2050 [1]. CO2 capture and storage (CCS) technologies beneficially affect the lifecycle of greenhouse gases emitted from fossil-fuel power plants [2,3], with CCS expected to account for up to 19% of global CO2 emission reductions by 2050, making it the most significant technology worldwide in this area [4]. Suitable geological formations for CCS include depleted oil and gas reservoirs, un-mineable coal seams, salt caverns, and deep saline aquifers [5,6]. After CO2 injection, the initial physico-chemical equilibrium between saline formation fluid and reservoir rocks can be disturbed by the triggering of reactions between CO2, fluid (brine), and reservoir rock [2]. Such interactions could lead to the dissolution of carbonates, feldspars, and clay cement in the aquifers [7,8]. In the absence of dynamic forces, such mineral dissolution could increase porosity and permeability by etching new pore spaces or widening narrow pore channels, temporarily increasing injectivity [9,10]. However, while sequestration of CO2 in carbonate minerals can contribute to long-term storage security [11], rapid mineral dissolution, especially of carbonates, could corrode caprocks, wellbores, and fault seals, potentially leading to migration of CO2 into overlying formations. Study of CO2-fluid-rock interactions is thus crucial for us to understand the physico-chemical processes involved.

Laboratory experiments can reveal the mineralogical and chemical changes resulting from CO2-brine-rock interactions, how they impact the lithological porosity and permeability of the geological sequence, and the effects on CCS potential [12–15]. However, experiments are limited to short-term effects of CO2 injection, whereas CCS is a long-term geochemical issue. Numerical modelling or simulation is useful for longer-tern studies. Several reactive geochemical transport models have been developed to simulate CCS, including NUFT [16], PFLOTRAN [17], CMG-GEM [18], STOMP [19], and TOUGHREACT [20,21]. The TOUGHREACT program has been widely used in studying geological CO2 sequestration [22–26]. However, simulations are less reliable without the availability of parameters derived from laboratory studies, so a combination of physical experiments and numerical simulation is the optimal choice for investigating the geochemical effects following CO2 injection.

In this study, both laboratory experiments (physical simulation) and numerical modelling were used to study geochemical interactions between CO2-induced fluids and reservoir rock during CCS. In the physical simulation, synthetic cores with composition consistent with geological samples were used to avoid interference from other geological factors such as sedimentary processes and diagenesis. The numerical simulation involved the same conditions of sample compositions, temperature, pressure, and fluid composition, with the two simulation types being mutually authenticating. Both numerical and physical simulations were used to document the process of short-term geochemical interactions after CO2 injection. A consistency of results would indicate the reliability of the simulations, with outcomes expected to be similar to those pertaining to actual geological conditions.

#### **2. Samples and Methods**

#### *2.1. Sample Descriptions*

Six synthetic sandstones were prepared for the physical simulation, with mineralogical compositions similar to sandstones of the Cretaceous Bashijiqike Formation (K1bs) of the Kuqa Depression, Tarim Basin, and western China. In order to identify mineralogical compositions of K1bs sandstones, the sandstone samples were prepared in thin sections and examined petrographically by point counting 300 to 400 points per section. In addition, these sandstones were also measured using quantitative X-ray diffraction analysis (D/max2500, Rigaku, Tokyo, Japan), which can provide quantitative mineralogical results within ±0.1 weight percentage (wt. %). The detail analysis processes can be found in Yu et al. (2012) [15]. The analytical results indicated that K1bs sandstones are fine- to medium-grained lithic sandstones with particle sizes of 0.25~0.5 mm, comprising mainly quartz (average ~37.5 wt. %), plagioclase (~20.8 wt. %), K-feldspar (~23.3 wt. %), calcite (~9.5 wt. %), dolomite (~7.4 wt. %), and kaolinite (~1.5 wt. %) (Table 1). According to Yu et al. (2015) [27], the K1bs reservoir sandstones were at the stage of mesogenetic diagenetic phase. Then we used the fine- to medium-grained mineral powders (particle size of 0.25~0.5 mm), having the above-mentioned mineralogical compositions, to reconstruct the six synthetic cores under the condition of mesogenetic diagenesis.

**Table 1.** The mineral composition of synthetic core samples.


#### *2.2. Physical Experimental Conditions*

The experimental condition is outlined as the following: (1) 48.45 MPa back-pressure (pore fluid pressure), (2) 60 MPa confining pressure, (3) 150 ◦C reaction autoclave temperature (formation temperature). The injection solutions were prepared by dissolving NaCl in deionized water saturated with CO2 at 150 ◦C and 48.45 MPa, similar to actual K1bs conditions. The injection solutions had a salinity of 14,182 mg L<sup>−</sup>1, approximating K1bs formation water. Here we only used the NaCl solution as the injection fluids and did not employ the imitate reservoir brines, because an amount of divalent cations, such as Ca2<sup>+</sup> and Fe2<sup>+</sup>, were present in the reservoir bines. After CO2 induced fluid injection into the autoclaves, some carbonates will precipitate and affect experimental results. Thus, pure NaCl solution, having a similar salinity with K1bs formation water, would be the most appropriate.

Under the experimental condition (P = 48.45 MPa and T = 150 ◦C), the injection solution was saturated with CO2. For the solution with a salinity of 14,000 mg L−1, the solubility of CO−<sup>2</sup> was 1.5451 mol/Kg, according to the CO−<sup>2</sup> solubility in bine of Duan and Sun (2003) [28]. During the experiment, the injected Vbrine (brine volume), and the volume of CO2 injected into the cylinder was VCO2 . Based on the equation of sate (EOS ) for gas, PV = ZnRT, where Z is the compressibility, n is the mole number of CO2 (nCO2 ) in the injection solution, R is gas constant, and T is temperature, we can obtain the volume of CO2 (VCO2soluble) dissolved in the injection solutions under the experimental condition (P = 48.45 MPa and T = 150 ◦C). Thus we can calculate the volume of the CO2 gas cap (VCO2gc) in the intermediate container. The derivation is as follows:

$$\mathbf{V\_{CO\_2gc}} = \mathbf{V\_{CO\_2}} - \mathbf{V\_{CO\_2solable}} \tag{1}$$

$$\text{V}\_{\text{CO}\_2} = 1030 - \text{V}\_{\text{keroserne}} - \text{V}\_{\text{brine}} \tag{2}$$

$$\text{V}\_{\text{CO}\_2\text{solubble}} = \text{Zn}\_{\text{CO}\_2} \text{RT/P} \tag{3}$$

Based on the above, it is possible to calculate the volume of CO2 in the gas cap of the intermediate container, which was ca.190.56 mL. Therefore the brine was CO2 saturated throughout the experiments.

#### *2.3. Experimental Apparatus*

The physical simulation experiment was conducted at the Key Laboratory of Basin Structure and Hydrocarbon Accumulation of the China National Petroleum Corporation, Beijing, China. A reservoir diagenesis modelling system with six identical reaction autoclaves was employed (Figure 1). The system includes six modules: heating furnace, pressure system, fluid-injection system, sampling system, control panel, and auxiliary system. In addition, a corrosion-resistant HP/HT CFR-50-100 cylinder (1030 mL) from TEMC, USA was used as an intermediate container for storing the CO2-bearing experimental solution. The six reaction autoclaves (Huaxing Company, Nan tong, Jiangxi Province, China) have a working pressure of 165 MPa and temperature of 300 ◦C. The pressure and fluid injection systems are controlled by the injection syringe pump and a back-pressure regulator. The 100DX syringe pump (Teledyne ISCO, Lincoln NE, USA) was used to control the fluid injection system, which consists of two separate systems (A and B), each of which has a capacity of 103 mL (Figure 1). It is capable of injecting at rates of 0.001~60 mL/min, with a precision of 0.5% of set point. The pump can handle pressure from 0.1 to 68.97 MPa. The advantage of this pump is its capability of continuous injection of any fluids including supercritical CO2. The pore-fluid pressure was controlled by the back-pressure regulator (DBRP-005, Honeywell, USA), which has a high precision and operating pressure range up to 51.72 MPa. All experimental parameters including the injection pressure, pore fluid pressure, and temperature were monitored.

**Figure 1.** Schematic diagram of CO2-formation water-rock physical experiment.

#### *2.4. Physical Simulation Workflow and Analysis*

The experiment was undertaken in two steps: preparation of the synthetic core, and geochemical reaction between the core and CO2 fluids. During the first step, the selected mineral powder (particle size 0.25~0.5 mm) was blended with distilled water and placed in six columnar autoclaves (diameter 3.0 cm, length 11 cm, volume 77.7 cm3). The six core samples (# 1 to # 6) were used in the experiment over 5 days under P/T conditions equivalent to mesogenetic diagenesis (Figure 2). The syringe-pump injection system injected synthetic formation water saturated with CO2 into the six synthetic core pores at 150 ◦C and 48.45 MPa, after which temperature and pressure were kept constant for 4 d (# 1), 7 d (# 3), 10 d (# 4), 13 d (# 5), and 16 d (# 6), while # 2 was used as a blank.

During the experiment, the temperature and pressure of each autoclave were monitored automatically by the control system. After reaction, core and fluid samples were analyzed for ion contents, mineralogical changes, and porosity. The producing fluid was measured for its pH values using an Orion4 STAR acidity meter from Thermo within 6 h of each sampling. The ionic compositions of the water were analyzed after being spiked with 1 mol/L HCl in order to avoid carbonates precipitation, and measured using an OPTIMA 7300DX ICP-OES (Inductively Coupled Plasma–Optical Emission Spectrometry) with an analytical precision of 10−3~10−9. Mineralogical changes were examined using a JSM6700F scanning electron microscope from JEOL with EDS (Energy Dispersive Spectrometer) from INCA software (Oxford Company, Oxford, England). The porosity changes were analyzed using visual porosity estimation, which is an image analysis technique. Firstly, core samples were impregnated with blue epoxy and then polished and made into casting thin sections. Then, combined high-resolution images of these thin sections were taken under the optical petrographic microscope; the image analysis software can delineate different types of porosity and calculate the percentages of these porosities in the thin sections with an accuracy of up to 0.01%.

**Figure 2.** Synthetic core samples made by the physical experiment.

#### *2.5. Numerical Simulation*

The program TOUGHREACT was used in the numerical simulations. This program is a non-isothermal, multiphase reactive transport simulation code that was used here to simulate fluid-rock interactions [21]. The kinetic data used during the simulation are shown in Table 2.


**Table 2.** List of minerals considered and parameters for calculating the kinetic rate constants.

Note that: (1) All rate constants are listed for dissolution; (2) A is specific surface area, k25 is kinetic constant at 25 ◦C, Ea is activation energy, and n is the power term (Equation (A1) in Appendix A); (3) The power terms n for acid mechanisms are with respect to H+. Data from Palandri and Kharaka (2004) [29].

According to the columnar autoclaves employed by the physical simulation, three identical cubic grids with volumes of 77.7 cm<sup>3</sup> were used to construct the model (Figure 3). The upper and lower grids were used as boundary cells, while the middle grid was the objective model grid for simulating the processes of injection and sampling. The numerical model simulated six autoclave reactions, corresponding to the laboratory experiment, with the same mineralogical cores, temperature, pressure, and pore fluids. We used the simulation duration to mimic the six numbered autoclaves. The entire simulation ran for 16 days with intermittent sampling on day 0, 4, 7, 10, 13, and 16, corresponding

to the physical simulation. At the start of simulation (Day zero), the numerical model had an initial mineralogical composition and visual porosity, which corresponded to Autoclave # 2. In the same way, Day 4 corresponded to Autoclave # 1, Day 7 corresponded to Autoclave # 3, Day 10 corresponded to Autoclave # 4, Day 13 corresponded to Autoclave # 5, and Day 16 corresponded to Autoclave # 6. Accordingly, these results of different simulation duration from the numerical models can be used for comparison with the results from the physical simulations. The boundary cell here is an "inactive" element, whose thermodynamic conditions do not change at all from fluid or heat exchange with finite-size blocks (numerical model cell) in the flow domain. The boundary cell can confine geochemical interactions that only occur in the numerical model, which makes the results more reasonable.

**Figure 3.** Schematic diagram of CO2-formation water-rock numerical simulation.

#### **3. Results**

#### *3.1. Changes in Fluid Chemistry*

Results of physical and numerical analyses of reaction products are summarized in Table 3. Significant changes in solution chemistry were observed in both sets of experiments (Figure 4). In the physical simulation, the pH continued to increase during the 16 d of the experiment, from 5.86 to 6.44 (Figure 4). In the numerical simulation, the pH first decreased to ~2.8 within 12 d, then increased to 4.6 over the next 4 d (Figure 4).

Fluid Si and K contents show similar changes in both simulations (Figure 4), with concentrations continuing to increase with reaction time (Figure 4). Fluid Ca and Mg concentrations increased with reaction time in the physical simulation, but were more constant in the numerical simulation (Figure 4). The Al content exhibited a distinct trend (Figure 4), reaching maximum values after 7 d and 10 d for the physical and numerical simulations, respectively, and then decreasing during further reaction. Absolute value of ion concentration differed between the simulations, with the numerical simulation set generally being higher, not including pH and Al (Figure 4).


**Table 3.** Chemical composition of outlet solutions.


**Table 3.** *Cont.*

**Figure 4.** Changes of pH and typical ion concentrations over the physical and numerical simulations.

#### *3.2. Changes in Mineral Morphology during the Physical Simulation*

Scanning electron microscope (SEM) analyses of core samples before and after physical simulations showed that minerals such as quartz, K-feldspar, albite, and dolomite dissolved after CO2 injection, with feldspar and dolomite showing pronounced dissolution and quartz weak dissolution. Before the experiment, mineral surfaces of quartz grains were generally smooth with terraced growth patterns (Figure 5A), with dissolution effects and corrosion pits being evident afterwards (Figure 5B). Initially, the albite surface was relatively flat and exhibited no obvious dissolution, but dissolution pits and fissures along cleavage surfaces were evident after the experiment (Figure 5C,D). K-feldspar was partially dissolved after the experiment, with the formation of corrosion pits (Figure 5E,F). The dissolution of K-feldspar was stronger than that of quartz and weaker than that of albite. Carbonates exhibited stronger dissolution than silicates, with entire dolomite particles being dissolved into a cloud-like phase and showing a paste-like flow structure (Figure 5G,H). Calcite was not observed after the experiment, indicating that it was completely dissolved.

**Figure 5.** Scanning electron photomicrographs of pre-and post-experimental cores. (**A**) Quartz before the experiment; (**B**) Quartz after the experiment; (**C**) detrital albite before the experiment; (**D**). detrital albite after the experiment; (**E**) K-feldspar before the experiment; (**F**) K-feldspar after the experiment; (**G**) dolomite before the experiment; (**H**) dolomite after the experiment. Q—quartz; Ab—detrital albite; Kf—K-feldspar; Do—dolomite.

#### *3.3. Changes in Porosity*

Surface porosity of the synthetic cores remained relatively constant at 12.64% during the physical simulation, with no variation being observed (perhaps limited by the analytical method). Similarly, porosity changes were not evident in the numerical simulation (Figure 6), with porosity being constant up to 8 d of reaction, then increasing with carbonate dissolution to only 12.646% over the next 8 d (Figure 6).

**Figure 6.** Porosity changes over time in the numerical simulation.

#### **4. Discussion**

#### *4.1. Mineral Dissolution and Precipitation*

Feldspars and carbonates are known to be easily corroded by acidic fluids during CO2 injection[2,30,31], as confirmed by numerical simulations [32–34], in situ, real-time field monitoring [35,36], and natural analogies [37,38]. Changes in fluid ion contents and SEM core observations in the physical simulation confirm that feldspar and carbonate were altered by CO2 injection. This is consistent with the numerical simulation, which also indicated dissolution of feldspars and carbonates (Figure 7). Both simulations indicate that Si and K, and Ca and Mg exhibit similar trends with ongoing reaction (Figure 4). Statistical analysis of Si, K, Ca, and Mg data using SPSS (Statistical Program for Social Sciences) software indicates correlation coefficients >0.5 (Table 4). Ion contents are thus likely controlled by a common reaction mechanism, as follows.

**Table 4.** Correlation coefficient matrix of the outlet solution ions.


The main mechanism controlling these reactions involves the formation of H2CO3 from dissolved CO2, causing the formation water to become acidic (Equation (4)), with reducing pH. Reactions between the acidic fluid and core minerals, especially carbonates and feldspars (Equations (5)–(8), below), buffer formation-water pH, causing an increase in pH of fluid produced during the experiments [39]. This process is described by the following equations:

$$\rm{HCO}\_{2} + \rm{H}\_{2}\rm{O} \rightarrow \rm{H}^{+} + \rm{HCO}\_{3}^{-} \tag{4}$$

$$\text{CaCO}\_3\text{ (calcite)} + \text{H}^+ \rightarrow \text{Ca}^{2+} + \text{HCO}\_3^- \tag{5}$$

$$\text{CaMg (CO}\_3\text{)}\_2\text{ (dolomite)} + 2\text{H}^+ \rightarrow \text{Ca}^{2+} + \text{Mg}^{2+} + 2\text{HCO}\_3^-\tag{6}$$

$$2\text{KAlSi}\_3\text{O}\_8\text{ (K-feldspar)} + 2\text{H}^+ + 9\text{H}\_2\text{O} \rightarrow \text{Al}\_2\text{Si}\_2\text{O}\_5\text{(OH)}\_4\text{ (kaolinite)} + 2\text{K}^+ + 4\text{H}\_4\text{SiO}\_4\text{(aq)}\tag{7}$$

$$
\Delta \mathbf{G}^0 = 18 \text{ KJ mol}^{-1}, \Delta \mathbf{S}^0 = 73 \text{ J mol}^{-1}
$$

NaAlSi3O8 (albite) + CO2 + H2O → NaAlCO3 (OH) <sup>2</sup> (dasownite) + 3SiO2 (chalcedony) (8)

$$
\Delta \mathbf{G}^0 = -132 \text{ KJ mol}^{-1}, \\
\Delta \mathbf{S}^0 = -101 \text{J mol}^{-1}
$$

where ΔG0 is the Gibbs free-energy change and ΔS<sup>0</sup> is the entropy change.

**Figure 7.** Mineral changes over time in the numerical simulation.

The degree of dissolution of albite is significantly greater (by a factor of ~2) than that of K-feldspar (Figure 7), possibly due to differences in their ΔG0 and ΔS<sup>0</sup> values. Albite ΔG<sup>0</sup> and ΔS0 values are both negative, with albite therefore needing little energy to dissolve, whereas K-feldspar values are positive, with more energy input needed for dissolution. The numerical simulation indicated that some kaolinite (up to 0.25 mol m−3) and quartz (up to 19.8 <sup>×</sup> 10−<sup>6</sup> mol m−3) precipitated after 4 d of reaction (Figure 7), although quartz precipitation can obviously be ignored. Equations (7) and (8) indicate that kaolinite precipitation restrains the reactions, leading to reduced K-feldspar dissolution. This is consistent with the results of other studies [15,40].

The precipitation of carbonate minerals is common during CO2-induced reactions [41], and our physical and numerical results indicate that the concentrations of carbonate minerals, calcite, and dolomite all decreased significantly during reaction. In particular, dolomite was almost completely dissolved, with no carbonate minerals remaining after the experiments. This is consistent with previous experimental findings [36,37,42]. However, the numerical simulations indicate that calcite and dolomite have similar dissolution tendencies (Figure 8), whereas calcite was completely dissolved in the physical simulations. We infer that under actual geological conditions, CO2 fluids react first with the most reactive minerals until they are exhausted before reacting with other minerals. In contrast, in the numerical simulations the reactions followed normal geochemical dynamic processes associated with the different minerals. Carbonate minerals did not precipitate during reaction (but produced minor amounts of kaolinite and quartz) because under the experimental conditions the reaction liquid was unsaturated with carbonates (Figure 8). Similarly, results akin to the above-mentioned calculations have also been presented by Ketzer et al. (2009) [43] and Tutolo et al. (2015) [44]. Quartz dissolution began after 5 d, and it precipitated later (Figure 8), but this reaction was very weak and is ignored here. Kaolinite was the predominant precipitated mineral (Figures 7 and 8), consistent with the results of Yu et al. (2012) [15]. However, carbonate precipitation is usually observed in CO2-formation-water–rock autoclave experiments conducted in closed systems over extended periods. For example, in an experiment using Triassic Sherwood Sandstone and sea water, Pearce et al. (1996) [37] observed calcite precipitation on the sample surface in an autoclave reaction under reservoir P/T conditions after almost eight months.

**Figure 8.** Saturation indices of carbonate minerals vs. reaction time in the numerical simulation.

Equations (4)–(6) indicate that calcite is the main reaction product, with some dolomite dissolving rapidly in the CO2-saturated formation water at the beginning of the experiments. However, the silicate minerals (mainly detrital albite and K-feldspar) also gradually become unstable and start dissolving. Precipitation of clay minerals such as kaolinite occurs under acidic conditions during the reaction

process. Details of the reaction process are indicated by Equations (7) and (8). These reactions lead to a rapid increase in pH of liquid produced during the initial stage, but with pH gradually reaching a stable equilibrium value, as also observed by Bowker and Shuler (1991) [35].

#### *4.2. Porosity Changes*

No obvious porosity changes were observed in the synthetic core after the physical simulation, with plane porosity being constant (within measurement uncertainty) at 12.64%. This was also observed in the numerical simulation (Figure 6) where porosity only increased from 12.64% to 12.646% (Figure 6). Minor changes in mineral contents after CO2 injection lead to minor changes in core porosity, as confirmed by changes in ion content in the physical simulation and other mineral changes in the numerical simulation (Figures 4 and 7). However, there were variations in porosity in the numerical simulations, where after six days of reaction the dissolution of minerals was very weak and porosity did not change noticeably, but over the following four days mineral dissolution increased with marked changes in porosity (Figures 4 and 7). Especially, a notable changes happened in porosity (Figure 6). This is due to the remarkable changes in the mineral dissolution (Figure 7). After nine days, the dissolution of feldspars and carbonates reached their peaks, indicating that the dissolution volume induced by the CO2-fluid injection increased to its maximum. A large number of newly added pore spaces lead to the porosity increase. By Day 10, minerals such as kaolinite and quartz began to precipitate, with porosity becoming less variable (Figures 4, 6 and 7). Overall, porosity varied little, indicating limited dissolution and precipitation during short-term CO2 injection.

The lack of reduction in porosity is common in CO2-induced reactions in sandstone [7,14,40,45], with a reduction of permeability being the dominant result of short-term CO2 injection. The precipitation of kaolinite, solid-phase materials, and clay particles released by the dissolution of carbonate cement may account for the non-reduction of porosity and the reduction of permeability. Shiraki and Dunn (2000) [40] considered that the precipitation of kaolinite crystals in pores is the main reason for the reduction of permeability after CO2 displacement reactions, while Luquot et al. (2012) [14] considered that newly formed minerals of amorphous carbon cause the reduction in permeability. Our results also indicate that precipitation of new minerals is related to the non-reduction of porosity. In both the physical and numerical simulations, the concentration of Al increased over the first six days before decreasing over the following 10 days. In the numerical simulation, the precipitation of kaolinite occurred after six days of reaction, with this requiring large amounts of Al (Equation (7)). While minor kaolinite was precipitated during the reaction, core porosity remained almost unchanged, for two possible reasons: (1) the dissolution of minerals was very weak in short-term CO2-induced reactions, with few changes occurring in feldspars and carbonates after CO2 injection (<1% mol m−<sup>3</sup> variation); and (2) the precipitation of minerals was limited. Kaolinite content varied by a few percent, while changes in quartz content were negligible, with porosity being unchanged during such weak reactions.

The physical simulation was an autoclave experiment with the inlet connected to an injection pump (an open system), and with the outlet being a closed system opened only during sampling at the end of the experiment. The reaction system was therefore a semi-closed system. Under conditions of deep burial in semi-closed space, dissolution of carbonates rarely occurs or is very weak [46]. Regarding the volumes of water required to increase porosity through calcite or dolomite dissolution, the problem is essentially the inverse of the effect on porosity loss in limestones of calcite cementation caused by dissolved calcium carbonate from external sources [47–50]. For example, to increase the porosity of a 100 m thick limestone bed by 1%, 1 m3 of calcite must be dissolved for each m<sup>2</sup> of bedding surface. For pore water that is undersaturated by 100 ppm, ~27,000 volumes of water are required to dissolve one volume of calcite. Increasing the porosity by 1% in 100 m thick limestone thus requires 27,000 m3 of water per square meter of surface. Even if the limestone was underlain by 5 km of sediments in which an average porosity loss of 10% of total rock volume occurred, the pore water released from the underlying sediments would not exceed 500 m<sup>3</sup> m−<sup>2</sup> [46], which, in an actual geological reservoir, would not be sufficient to dissolve the carbonates. In our experiment, the autoclave volume was 77.7 cm3, and it was

impossible to provide sufficient water for carbonate dissolution. However, it is certain that dissolution and precipitation are very weak at the beginning of CO2 fluid-rock interactions, with our physical and numerical simulations confirming that only limited geochemical reactions, including dissolution and precipitation, occur during short-term CO2 injections, with no sharp variations in core porosity or permeability. Similar results were also found by Tutolo et al. (2015), which confirmed that only very weak geochemical reactions could happen during the reaction of CO2 and feldspar-rich sandstone [51]. For long-term CO2 injections, however, dissolution and precipitation are the dominant geochemical processes occurring between CO2-induced fluids and sandstones [51–53]. Our study of short-term geochemical interactions in a semi-closed system therefore showed no remarkable changes in the porosity of cores.

#### **5. Conclusions**


**Author Contributions:** Z.Y. collected and analyzed the data and wrote the original draft. S.Y. reviewed and edited the paper. K.L. conducted the English editing work of this paper. Q.Z. contributed to the constructive discussions. L.Y. contributed to the numerical simulations.

**Funding:** This research and APC were both funded by the National Hydrocarbon Accumulation, Distribution and Favorable Areas Evaluation in Foreland Thrust Belts and Complex Tectonic Zones (No. 2016ZX05003-002).

**Acknowledgments:** We also thank the Key Laboratory of Basin Structure and Hydrocarbon Accumulation for allowing us to carry out laboratory experiments and access its rock characterization facilities.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **Appendix A. Kinetic Rate Law for Mineral Dissolution and Precipitation**

The general rate expression used in TOUGHREACT is taken from Lasaga et al. (1994) [54]:

$$\mathbf{r}\_{\rm n} = \pm \mathbf{k}\_{\rm n} \mathbf{A}\_{\rm n} \left| 1 - \left( \frac{\mathbf{Q}\_{\rm n}}{\mathbf{K}\_{\rm n}} \right)^{\partial} \right|^{\eta} \tag{A1}$$

where n denotes the kinetic mineral index, positive values of rn indicate dissolution, while negative values indicate precipitation; kn is the rate constant (moles per unit mineral surface area and unit time) and is temperature dependent; An is the specific reactive surface area per kg H2O; Kn is the equilibrium constant for the mineral-water reaction for the destruction of one mole of mineral n; and Qn is the reaction quotient. The parameters θ and η must be determined from experiments. However, they are usually, but not always, set to 1.

For many minerals, the kinetic rate constant k can be summed from three mechanisms (Palandri and Kharaka, 2004) [29]:

$$\begin{split} \mathbf{k} = \mathbf{k}\_{25}^{\mathrm{nu}} \exp\left[ -\frac{\mathrm{E}\_{\mathrm{a}}^{\mathrm{nu}} \left( \frac{1}{\mathsf{T}} - \frac{1}{298.15} \right)}{\mathsf{R}} \right] + \mathbf{k}\_{25}^{\mathrm{H}} \exp\left[ -\frac{\mathrm{E}\_{\mathrm{a}}^{\mathrm{H}} \left( \frac{1}{\mathsf{T}} - \frac{1}{298.15} \right)}{\mathsf{R}} \right] \mathbf{a}\_{\mathrm{H}}^{\mathrm{nu}} \\ \quad + \mathbf{k}\_{25}^{\mathrm{OH}} \exp\left[ -\frac{\mathrm{E}\_{\mathrm{a}}^{\mathrm{OH}} \left( \frac{1}{\mathsf{T}} - \frac{1}{298.15} \right)}{\mathsf{R}} \right] \mathbf{a}\_{\mathrm{OH}}^{\mathrm{n}\_{\mathrm{OH}}} \end{split} \tag{A2}$$

where superscripts or subscripts nu, H, and OH indicate neutral, acidic, and alkaline mechanisms, respectively; Ea is the activation energy; k25 is the rate constant at 25 ◦C; R is gas constant; T is the absolute temperature; a is the activity of the species; and n is an exponent (constant).

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **Carbon Spheres as CO2 Sorbents**

#### **P. Staciwa 1, U. Narkiewicz 1,\*, D. Sibera 1, D. Moszy ´nski 1, R. J. Wróbel <sup>1</sup> and R. D. Cormia <sup>2</sup>**


Received: 27 June 2019; Accepted: 12 August 2019; Published: 15 August 2019

**Abstract:** Microporous nanocarbon spheres were prepared by using a microwave assisted solvothermal method. To improve the carbon dioxide adsorption properties, potassium oxalate monohydrate and ethylene diamine (EDA) were employed, and the influence of carbonization temperature on adsorption properties was investigated. For nanocarbon spheres containing not only activator, but also EDA, an increase in the carbonization temperature from 600 ◦C to 800 ◦C resulted in an increase of the specific surface area of nearly 300% (from 439 to 1614 m2/g) and an increase of the CO2 adsorption at 0 ◦C and 1 bar (from 3.51 to 6.21 mmol/g).

**Keywords:** carbon nanospheres; nanocarbon spheres; carbon dioxide uptake; EDA

#### **1. Introduction**

The global economy requires a great amount of energy, which is produced primarily by the combustion of fossil fuels. Carbon dioxide emissions are a significant negative side-effect of this activity. Transportation also requires a great amount of petroleum and is responsible for significant emissions of greenhouse gases [1,2]. The cumulative emission of CO2 strongly contributes to climate change and is the greatest single contributor to the greenhouse effect [3,4]. The average concentration of CO2 in the earth's atmosphere in 2018 was 407 ppm, which is about 40% higher than in the preindustrial age [5]. The effort to develop technologies that will reduce CO2 emissions is very important for both the global economy and the environment.

Recently, methods of CO2 capture from flue gas have been based on absorption into liquids (e.g., amines [6] or methanol [7]). These technologies are energy intensive and not environmentally sound. Solid sorbents offer an alternative solution. There are a number of criteria that must be met for a successful sorbent material, namely: high selectivity and adsorption capacity for CO2, fast adsorption/desorption kinetics, efficient regeneration of sorbents, and low cost [8].

In recent years, a number of materials have been investigated as solid state adsorbents for CO2, such as: zeolites [9], silica [10], porous polymer materials [11], metal organic frameworks [12], and carbon materials [13–16]. The most efficient for CO2 adsorption are carbon materials, which exhibit a high surface area, large porous volume, chemical stability, affinity for carbon dioxide, low cost, and the possibility of modification with heteroatoms [17]. The weak side of carbon sorbents is their poor selectivity.

The application of carbon materials for CO2 uptake has been widely investigated. There are many sources of carbon that can be used for the production of activated carbon: polymers, biomass, or resins. Some examples are shown in Table 1. Potassium compounds, namely potassium hydroxide or potassium oxalate, are most often used as chemical activators. Special attention should be paid to resins. Gradual growth of the polymer chain allows incorporating modificators into the carbon matrix. Homogeneously distributed activator can improve not only the surface area of the material

(impregnation), but its whole volume. A resorcinol–formaldehyde resin mixture could be a suitable carbon source for CO2 adsorption.


**Table 1.** Comparison of CO2 uptakes on various carbon adsorbents.

According to the results presented in Table 1, resins are very promising as a carbon source to produce solid sorbents for carbon dioxide capture.

Among nanocarbon materials, spherical structures have been widely studied. The most popular method to obtain porous nanocarbon spheres is the method of Stöber, using resins as a carbon source. The application of this method to produce carbon spheres was described in the work of Liu et al. [26], where the source of carbon was a resorcinol–formaldehyde resin. Thanks to the development of the Stöber method, researchers discovered a simple method to produce polymer beads. The product was in the form of spherical regular particles. Since then, the phenol derivatives were widely used as a carbon source [27,28]. In work by Zhao and co-workers [28], using 3-aminophenol as a precursor, highly monodisperse material were obtained. They also proved that changing different parameters allowed for tuning spherical size in a very broad range.

To enhance the surface area and porous volume, various processes of activation are employed, with two primary methods to activate carbon materials. First, physical activation carried out through carbonization in the presence of proper gases [29]. Second, chemical activation is induced by the addition of a strong base, i.e., potassium oxalate [19], potassium hydroxide [20,30,31], and potassium carbonate [32]. In the work of Choma et al. [18], chemical activation with potassium oxalate resulted in a large increase in the surface area of carbon materials (from 680 m2/g to 1490 m2/g) and an increase of CO2 uptake from 3.03 mmol/g to 7.67 mmol/g in 0 ◦C at 1 atm. This example showed how modification with potassium oxalate can significantly enhance specific surface area and CO2 adsorption of carbon spheres.

In order to improve synthesis conditions of carbon nanospheres, microwave assisted solvothermal reactor has been used [33,34]. Performing the reaction in a common autoclave takes a significant amount of time, often several hours, while the reaction in microwave assisted solvothermal reactor is very fast, about 15 min. The temperature gradient using a microwave in the reactor volume is very low and can be negligible. The microwave's influence on the behavior of polar solvents in the reaction is significant, and volume nucleation points are created rapidly.

In this work, the influence of the concentration of activator, potassium oxalate, carbonization temperature, and influence of ethylene diamine (EDA) on the physical properties and adsorption of carbon dioxide were investigated.

Ludwinowicz and Jaroniec [19] performed a simple one-pot synthesis of carbon spheres and obtained very good CO2 adsorption values. In this work, simple autoclave was replaced by microwave assisted solvothermal reactor. The use of such a reactor enabled a significantly shortened reaction time. Heating with microwaves avoids a variance in the temperature profile in the reactor volume, no local overheating, and the products obtained are of very good quality, with uniform shape and size of the produced particles.

Previous research obtained in this research community has been promising [33,34]. In the present work, we describe more in depth research on the influence of modificator and carbonization temperature on surface area, porosity, and carbon dioxide adsorption.

#### **2. Experimental**

#### *2.1. Sample Preparation*

The samples were synthesized as follows: First, an aqueous–alcohol solution consisting of 60 mL distilled water and 24 mL ethanol was prepared by mixing at an ambient temperature. Subsequently, 0.60 g of resorcinol and 0.30 mL of ammonium hydroxide (25 wt.%) was added to the mixture under continuous stirring for 10 min. After dissolving of the resorcinol, the proper amount of potassium oxalate was added and the mixture was stirred for 30 min. The weight ratio potassium: carbon was 7:1 and 9:1. For samples modified with EDA, 0.3 mL of EDA was added. Next, 0.9 mL of 37 wt.% formaldehyde was dropped into the solution and kept under magnetic stirring for 24 h. Afterwards, the solution was treated in a solvothermal microwave reactor ERTEC MAGNUM II (pressure 2 MPa, time—15 min). The resulting materials were dried at 80 ◦C for 48 h. The carbonization of the carbon nanospheres was performed in argon atmosphere at 350 ◦C for 2 h with a heating rate of 1 ◦C/min, then the temperature was raised to 600 ◦C, 700 ◦C, or 800 ◦C with the same heating rate and also for 2 h. The materials obtained were washed two times with 200 mL of distilled water and dried at 80 ◦C for 48 h.

#### *2.2. Characterization*

The morphology of the produced samples was determined using a Hitachi SU8020 Ultra-High Resolution Field Emission Scanning Electron Microscope (FE-SEM).

The density of the materials was determined using a helium pycnometer Micro-Ultrapyc 1200e.

The chemical composition of the samples' surface was studied by X-ray Photoelectron Spectroscopy (XPS). The measurements were conducted using Mg *Ka* (hν = 1253.6 eV) radiation in a Prevac (Poland) system equipped with a Scienta (Sweden) SES 2002 electron energy analyzer operating with constant transmission energy (*Ep* <sup>=</sup> 50 eV). The analysis chamber was evacuated to a pressure below 1·10−<sup>9</sup> mbar. A powdered sample of the material was placed on a stainless steel sample holder.

Thermal stability of the produced materials was investigated using Thermal Gravimetric Analysis (TGA). The thermogravimetric measurements were carried out with the use of STA 449 C thermobalance (Netzsch Company, Germany). Approximately 10 mg of the sample was heated at 10 ◦C/min to 950 ◦C under air atmosphere.

To determine textural properties of the carbon spheres, the low temperature physical adsorption of nitrogen was carried out at −196 ◦C using the Quadrasorb volumetric apparatus (Quantachrome Instruments). Carbon dioxide uptake was gathered at temperature 0 ◦C and 25 ◦C using the same apparatus.

#### **3. Results and Discussion**

#### *3.1. Samples' Morphology*

The morphology of the carbon spheres was studied using Scanning Electron Microscopy (SEM). The SEM images of the samples carbonized in 700 ◦C are shown in Figure 1. For the material without modification (Figure 1a), small, monodisperse spheres were obtained. The average diameter of the carbon spheres was determined by SEM to be about 400 nm. For the material prepared by the addition of the activator (potassium oxalate, Figure 1b), two classes of spheres were observed: smaller spheres, about 500 nm in diameter, and larger, about 2–3 μm. The large difference in the diameter of the spheres was the result of the addition of potassium oxalate. The resorcinol–formaldehyde spheres were influenced by the oxalate moieties, and thus larger spheres were formed. However, there was a fraction of the smaller spheres, where oxalate moieties were likely less present. The higher concentration of

potassium oxalate (Figure 1c) resulted in higher saturation of the solution. The spheres containing more potassium oxalate were larger. Nonetheless, there was a large amount of small spheres, which did not contain oxalate moieties. Thus, a large amount of carbon material was not modified.

(**d**) RF EDA 700 (**e**) RF 7/1 EDA 700

**Figure 1.** Scanning Electron Microscopy (SEM) images of the spheres: (**a**) without modification; (**b**) with activator concentration 7/1; (**c**) with activator concentration 9/1; (**d**) with ethylene diamine (EDA) modification; (**e**) with activator concentration 7/1 and EDA modification.

For the material modified with EDA only (Figure 1d), the monodispersity of the spheres was kept, however larger spheres (diameter ca. 800 nm) were formed. The larger diameter and the monodispersity of the spheres are evidence that EDA was well dispersed in the whole volume, and so, all carbon spheres contained EDA. In the case of the material modified with both EDA and potassium

oxalate (Figure 1e), the influence of both modifiers can be noticed. The average diameter of the spheres was larger and the spheres were more uniform. EDA provided better dispersion of potassium oxalate in the reaction volume, thus potassium ions were present in a higher amount of the spheres. In the end, a much bigger specific surface area value was reached.

The size distribution of the produced particles was evaluated from the SEM images using the ImageJ software tool and is illustrated in Figure 2a–e. The quantity of spheres taken into account was 50 for every kind of the sample.

The results for the reference sample are given in Figure 2a. RF 700 exhibited the highest monodispersity among all the samples. The diameter of the spheres was about 600 nm.

**Figure 2.** *Cont*.

**Figure 2.** *Cont*.

**Figure 2.** (**a**) Size distribution of the reference sample. (**b**) Size distributions of the samples with lower activator content (7/1). (**c**) Size distribution of the samples with higher content of the activator (9/1). (**d**) Size distribution of the samples modified with EDA only. (**e**) Size distribution of the samples modified with EDA and activator.

The modification of carbon materials with the lower content of potassium oxalate resulted in higher variation in the spheres' size distribution. A considerable amount of produced spheres had a diameter from 300 to 1000 nm, as shown in Figure 2b. With the increase of the carbonization temperature, the formation of larger spheres (about 2000 nm) was observed. For the material carbonized at 600 ◦C, the majority of spheres were about 500 nm, whereas on the other hand for the sample carbonized at 800 ◦C, this value shifted to about 700 nm.

Comparing the samples with different amounts of activator, the strong influence of the activator concentration on the spheres' size was noticed. Higher activator content in the samples resulted in the widest size distribution (Figure 2c). Moreover, the large spheres of diameter over 2000 nm were formed.

The size distribution of the samples modified with EDA only is presented in Figure 2d. Compared to the samples modified with potassium oxalate, the highest monodispersity of the spheres was gained. Nonetheless, increasing the carbonization temperature caused the distribution to be broader.

As can be seen in Figure 2e, the application of both the modificators limited the production of the large spheres (over 2000 nm). Unlike previous distributions, by increasing the carbonization temperature, the shift of distribution towards smaller spheres was noticed.

#### *3.2. Surface Chemistry*

The surface composition of materials was analyzed by X-ray Photoelectron Spectroscopy (XPS). The survey spectra acquired for all analyzed samples are shown in Figure 3. The evaluation of the elemental composition of the surface of all samples is presented in Table 2. In all samples, carbon and oxygen was present, and potassium was observed in samples prepared with potassium oxalate.

**Figure 3.** X-ray Photoelectron Spectroscopy (XPS) survey spectra.


**Table 2.** Elemental composition of the surface of the samples.

The highest carbon content was observed for the pure carbon material (RF\_700); oxygen constitutes only 3% of the surface atoms. The surface of the samples prepared with EDA only also contained a relatively small number of oxygen atoms (approximately 4%), however those surfaces also contained about 4% of nitrogen atoms. The presence of potassium in the internal structure of the material is associated with an increased concentration of oxygen atoms. The more potassium observed in the material, the higher the concentration of oxygen observed, as residual potassium atoms were bound with oxygen. In general, when the carbonization temperature was increased to 800 ◦C, this resulted in a lower oxygen concentration than that observed for samples carbonized at 600 ◦C. There is noticeably more potassium retained on the surface of the samples modified by both potassium oxalate and EDA in comparison to materials modified by potassium oxalate only. A possible explanation for this is that a reaction of potassium with amine groups occurred.

The analysis of high-resolution XPS data brings a more detailed view of the chemistry of the surface of the studied materials. In Figure 4, the spectral region of binding energy between 280 eV and 300 eV is displayed for two samples of carbon spheres obtained with the weight ratio potassium: carbon of 9:1 (RF\_9\_1\_600 and RF\_9\_1\_800). This region contains the spectrum components originating from C 1s and K 2p electrons. The peak maximum of K 2p3/<sup>2</sup> is located at 293 eV and it is accompanied with a K 2p1/<sup>2</sup> spin-orbit component at 295 eV. The peak maximum of XPS C 1s spectrum is located at 284.4 eV. This energy is characteristic for highly graphitized carbon materials. However, a distinctive shoulder at about 288 eV is present in the spectrum for both samples, though more prominent on the sample carbonized at 600 ◦C. This position is usually ascribed to the general group of carbon moieties containing O–C=O bindings. The intensity of the spectra is normalized in respect to the intensity of the main peak of carbon. It can be pointed out that the relative intensity of lines corresponding with potassium atoms as well as carbon atoms in O–C=O bindings decreases in comparison to C–C bonds, reflected by XPS C 1s peak at 284.4 eV. This shows that increased carbonization temperature results in a partial depletion of potassium atoms from the surface as well as a decomposition of a part of C–O bonds. This corresponds well with the quantitative analysis described above. Similar observations are also valid for samples with a lower potassium:carbon ratio.

**Figure 4.** X-ray photoelectron spectrum of C 1s and K 2p regions for samples prepared with potassium oxalate with a potassium:carbon ratio of 9:1.

Slightly different behaviour of the surface species is observed for samples prepared with EDA. In Figure 5, the spectral region of the binding energy between 280 eV and 300 eV is displayed for two samples of carbon spheres obtained with the weight ratio potassium:carbon of 7:1 with the addition of EDA (RF\_7\_1\_EDA600 and RF\_7\_1\_EDA800). The position of the K 2p peaks is identical to samples without EDA admixture, indicating that the chemical state of potassium atoms is not changed by EDA presence during the preparation stage. However, the peak maximum of the C 1s line for sample RF\_7\_1\_EDA600 is located at 284.6 eV, which is characteristic for C–C bonding in aliphatic sp<sup>3</sup> bonds or non-graphitic amorphous carbon. For the sample carbonized at 800 ◦C, the respective peak maximum of C 1s line is shifted to 284.4 eV. Similar to the samples prepared without EDA, this peak position is assigned to C–C bonds in graphitized carbon material. It is worth noting that the relative intensity of K 2p lines for 600 ◦C and 800 ◦C of carbonization is only slightly changed.

**Figure 5.** X-ray photoelectron spectrum of C 1s and K 2p regions for samples prepared with potassium oxalate with a potassium:carbon ratio of 7:1 with the addition of EDA.

#### *3.3. Thermogravimetric Analysis*

In order to investigate the thermal stability of the samples, thermogravimetric measurements were performed. The results of the TGA studies are shown in Figure 6 and Table 3. The reference sample began to lose mass at about 364 ◦C. This can be explained by the decomposition of the functional groups. Further, the decomposition of the carbon matrix occurred.

**Figure 6.** Results of thermogravimetric studies (heating in air).

Compared with the non-modified sample RF 700, the addition of the EDA alone did not affect the thermal stability of the sample. On the contrary, the addition of potassium oxalate led to a significant decrease of the thermal stability. All samples containing potassium oxalate were characterized by a lower decomposition temperature. Potassium ions are attracted to polar water molecules, thus the addition of potassium to the carbon matrix resulted in higher hydrophilicity of the material. The mass loss began at 100–150 ◦C because of the removal of water molecules. Due to mobile energized potassium ions, the depleted carbon matrix is less resistant to thermal decomposition (start of decomposition was detected at about 180 ◦C).

The thermal stability of the sample with the addition of both modificators was similar to that modified with potassium oxalate only.


**Table 3.** Results of the thermogravimetric studies.

#### *3.4. Adsorption Studies*

According to the low-temperature nitrogen adsorption–desorption studies, for samples modified with oxalate only, the increase of carbonization temperature resulted in a higher volume of adsorbed nitrogen. The opposite effect was observed for samples modified with EDA only. The addition of both modificators gave a similar result to the use of oxalate only.

Some examples of low-temperature nitrogen adsorption isotherms are shown in Figure 7.

**Figure 7.** Low-temperature nitrogen adsorption isotherms of carbon spheres modified with potassium oxalate and/or with EDA.

The isotherms are of type Ia [35], characteristic for microporous materials, however a slight increase of adsorbed nitrogen volume at the highest P/P0 can be attributed to the presence of macropores (type II). Spheres, modified with EDA only, adsorbed the lowest nitrogen volume. Modification with potassium oxalate resulted in higher nitrogen adsorption, slightly increasing with dopant concentration. The highest amount of nitrogen was adsorbed in the sample modified with both potassium oxalate and EDA.

Physico-chemical properties of the samples were measured, and the results are shown in Table 4. In almost all cases, except samples modified with EDA only, an increase in carbonization temperature resulted in an increase of the samples' density, specific surface area, total pore volume, and CO2

adsorption. An unusual increase in density, simultaneously with an increase in surface area and porosity can be explained by the decomposition of modificators and removal of gaseous decomposition products. The same phenomenon was reported for activated carbon produced from palm shell and modified with potassium carbonate [36] or phosphoric acid [37].

An extremely high increase in specific surface area and CO2 adsorption was observed for the samples modified with both potassium oxalate and EDA. In contrast, samples without potassium oxalate carbonized at higher temperatures did not exhibit larger surface area, and a lower amount of CO2 was adsorbed (because of a lower microporosity). However, a higher concentration of the activator did not improve the specific surface area. Due to the higher saturation of the mixture, bigger spheres were formed, but oxalate moieties were not well dispersed within the volume of the sample (as shown before in SEM images).


**Table 4.** Physico-chemical properties of the obtained samples.

For adsorption of carbon dioxide, the presence of the micropores below 1 nm is considered to be most important, and the pore size distribution in this area was calculated from CO2 adsorption at 0 ◦C and is shown in Figure 8.

**Figure 8.** PSD (Pore Size Distribution) calculated from CO2 adsorption at 0 ◦C for the samples modified with EDA only.

In the research paper [21], doping of the carbon spheres with EDA was described. Increasing the EDA ratio (from 0.2 mL to 0.8 mL for 0.4 g of resorcinol) led to an improvement of specific surface area and CO2 uptake at 25 ◦C. The work of Sibera et al. [34] also reported a positive effect of a higher concentration of EDA as a modificator, improving the CO2 uptake. In the present paper, more detailed studies on the influence of carbonization temperature on samples modified with EDA were performed.

The samples modified with EDA showed much lower surface areas and CO2 adsorption than the reference sample RF 700. A higher carbonization temperature resulted in lower surface area (Table 4) and lower total pore volume, but an increase of the CO2 uptake was observed. This observation can be explained by higher micropore volume, below 0.4 nm for the sample RF EDA 800 (Figure 8). At elevated temperatures, carbon spheres have a tendency to aggregate, thus the effective surface area decreased. Density measurements proved the increase of density (from 1.59 g/cm3 for RF EDA 600 to 1.72 g/cm<sup>3</sup> for RF EDA 800). In contrast, an increase in the carbonization temperature increased the volume of the pores below 0.4 nm (Figure 8) and consequently the CO2 adsorption capacity.

Significant differences were noticed for the samples modified with potassium oxalate. In the paper [19], Ludwinowicz and Jaroniec applied three potassium oxalate concentrations, with a K:C ratio of 3:1, 5:1, and 7:1. The growth in surface area (460 m2/g for pure material and 2130 m2/g for the highest concentration potassium oxalate) and CO2 adsorption (2.8 mmol/g for pure material and 6.6 mmol/g for the highest concentration potassium oxalate) was observed. In order to investigate the influence of the activator concentration on the physico-chemical properties of the spheres, we employed a higher concentration of potassium oxalate monohydrate (weight ratio K:C = 9:1). The specific surface area values were similar to the values for samples with a lower activator concentration. A significant difference in CO2 uptake at 0 ◦C and 25 ◦C was observed for sample RF 9/1 carbonized at 800 ◦C.

The microporosity of these samples carbonized in 700 ◦C is given in Figure 9. The sample RF 700, compared to the samples modified with potassium oxalate, had the lowest specific surface area. This was caused by a lack of energized potassium ions to interact with the carbon matrix and a lack of carbon dioxide released in the result of decomposition of potassium oxalate, creating porosity. For the sample RF 7/1 700, modified with the lower amount of activator, a significant increase of the microporosity in the range of width from 0.3 to 0.7 nm was observed. Application of the higher concentration of the activator in the sample RF 9/1 700 improved the specific surface area, but the lower amount of adsorbed CO2 was noticed, which was in agreement with the lower volume of pores below 0.7 nm, as shown in Figure 9. The highest values of the specific surface area and CO2 adsorption were obtained for the samples modified with potassium oxalate and EDA simultaneously. Energized potassium ions penetrated the nanocarbon material, but on the other hand, EDA improved the basicity of the material and distribution of the oxalate moieties throughout the nanocarbon sphere. For the sample RF 7/1 + EDA 700, the value of the specific surface area was twice as high as the sample modified with potassium oxalate only, but the CO2 adsorption was only slightly higher. The microporosities of both samples with a diameter of 0.7 nm were comparable. The sample RF 7/1 + EDA 700 had a higher volume of pores from 0.7 to 0.9 nm, however this feature did not improve the CO2 adsorption significantly. Despite the higher value of the specific surface area of the sample carbonized in 800 ◦C (500 m2), the CO2 uptake at 0 ◦C was only slightly better, however at 25 ◦C, a decrease of the adsorbed value for the sample RF 7/1 + EDA 700 was observed.

The adsorption capacity values of all samples are presented in Figure 10. The samples carbonized in 600 ◦C were more resistant to a decrease in CO2 adsorption at the higher adsorption temperature. Mostly, the increase of carbonization temperature led to higher surface area and CO2 adsorption, but a significant decrease of the adsorbed values of CO2 at 25 ◦C compared to 0 ◦C was observed.

**Figure 9.** PSD calculated from CO2 adsorption at 0 ◦C for the samples carbonized at 700 ◦C.

**Figure 10.** CO2 adsorption values of the tested samples.

#### **4. Discussion**

The modificators played a double role in this reaction system: first creating more porosity and second, giving a basic chemical character to the produced material.

In our previous paper [33], chemical activation of carbon spheres using a similar amount of potassium oxalate monohydrate was achieved. In the case of the materials prepared with potassium oxalate monohydrate, two activation mechanisms can be distinguished. First, potassium ions penetrate the carbon material and a high porosity material was formed [38]. The effect of a higher carbonization temperature resulted in more energetic potassium ions migrating into the nanocarbon spheres. Second, decomposition of potassium oxalate monohydrate at about 570 ◦C resulted in the release of CO2, which would help remove pyrolyzed volatile products from the carbon matrix and could also prevent an aggressive pore widening process, leading to better microporosity [39,40]. Potassium oxalate decomposes to release carbon dioxide and to form potassium carbonate. The latter decomposed above

700 ◦C, also with the release of carbon dioxide, and then, an increase of carbonization temperature from 700 to 800 ◦C resulted in an increase in specific surface area by over 100%, from 645 to 1331 m2/g (Table 4). Nonetheless, the CO2 adsorption values were only slightly enhanced (from 5.15 to 5.52 mmol/g at 0 ◦C).

EDA also decomposes at elevated temperatures, with the release of ammonia, carbon dioxide, carbon monoxide, nitrogen oxides, and/or volatile amines. However, the release of the gaseous decomposition products did not result in an increase of the porosity of the material. Thus, the use of EDA alone did not change the physical properties of the material.

The best results were obtained when both activators were applied. We posit that EDA reacted with potassium oxalate, forming the chelates, which improved the homogeneous distribution of potassium within the sample volume. This can be explained by the trapping of migrating potassium ions by amine groups. The ability of EDA to form chelates with metals ("amino acid metals") is well known. Half of EDA produced by the Dow Chemical Company [41] is used as a chelating agent, forming complexes with certain metal ions to prevent the ions from interfering with processing or to promote buffering, concentration, separation, or transport.

Carbonization at high temperatures caused decomposition of both modificators and of the formed chelates, nevertheless some potassium remained in the samples and had a positive effect on the adsorption properties towards carbon dioxide, increasing surface basicity.

Potassium can form carbides with carbon. According to the literature [42], there is a possibility of the formation of the following potassium carbides: KC8, KC16, KC24, KC32, KC48, and KC60. These carbides have graphite-like lattices in which the metal atoms are situated between the layers of carbon atoms. The metal atoms are located at the centers of the carbon hexagons.

Relevance of the presence of pores below 1 nm in the matter of CO2 adsorption has been widely documented [43,44]. Presser and coworkers [45] claimed that under atmospheric pressure, the contribution of the pores below 0.8 nm to the CO2 adsorption was the most significant. Pore size distributions of the modified samples showed that use of chemical activator was necessary, i.e., potassium oxalate monohydrate was responsible for the creation of micropores beneath 0.4 nm.

Activation with potassium led to the creation of irregularities on the surface of carbon materials [46]. Gadkaree and Jaroniec [47] investigated pore structure development in carbon materials produced from resins. They fabricated two types of carbon honeycomb structures: standard type A, which involved phenolic resin as the liquid precursor, and type B, which involved the same phenolic resin but containing cobalt acetate dissolved at 1 wt.%. Both kinds of samples were carbonized in nitrogen at a high temperature and then activated in CO2. For samples of type A, only micropores (no mesopores at all) were formed and their volume increased as a result of the deepening of pores formed during carbonization. No pore broadening was observed for these samples. The introduction of a metallic catalyst (cobalt) in the precursor resin changed the pore structure dramatically (samples B). The pore structure on carbonization remained the same as that of the carbon without the catalyst (samples A). The difference between type A and B appeared upon activation. A bimodal pore size distribution was observed for samples B. Despite micropores of the same size range as in samples A, large meso and macropores were formed in samples B. In both kinds of samples, A and B, the micropore volume increase took place because of pore deepening, rather than pore broadening.

According to Casso et al. [48], the adsorption in pores depended on the applied pressure. With an increase of adsorption pressure, bigger and bigger pores govern the adsorption, nonetheless researchers state that at the atmospheric pressure, adsorption is contributed to by the narrow micropores (below 0.6 nm) with high adsorption potential. To use the whole adsorption potential of bigger micropores, higher fugacity of adsorbent is required. Hence, the higher pressure must be applied.

Investigation of Chen et al. [49] showed that the adsorption of CO2 molecules in the 0.3 nm slit pores, due to the similar size of CO2 molecules, was very poor. In contrast, the highest CO2 adsorption was noticed in the 0.4 nm pores. Furthermore, strong decrease of the stabilization energy of CO2 molecules in pores larger than 0.4 nm was noticed. With an increase of pore size, the interactions between CO2 molecules and carbon pores were more and more weak.

To summarize, the presence of micropores below 0.7 nm is one of the essential traits for a good CO2 adsorbent. Comparing the samples modified with potassium oxalate, the 7:1 weight ratio is an optimum one value, and there is no benefit to employ more.

According to the analysis of the obtained results, the adsorption properties of nanocarbon materials towards carbon dioxide increased with both increasing specific surface area and porosity (Figures 11 and 12).

**Figure 11.** Relation between specific surface area and CO2 uptake at 0 ◦C.

**Figure 12.** Relation between the total pore volume of the samples and CO2 uptake at 0 ◦C.

The presence of potassium on the surface of the samples had a positive effect on the CO2 adsorption (Figure 13), however the presence of surface oxygen had no apparent effect on adsorption. Surprisingly, the presence of surface nitrogen decreased the ability to adsorb carbon dioxide (Figure 14).

**Figure 13.** Relation between the surface concentration of potassium and CO2 uptake at 0 ◦C.

**Figure 14.** Relation between the surface concentration of nitrogen and CO2 uptake at 0 ◦C.

#### **5. Conclusions**

Highly porous nanocarbon materials for CO2 adsorption were produced through a novel synthesis method using a microwave assisted solvothermal reactor and varying the concentration of key reactants and modifactors. Replacement of an autoclave by a microwave assisted solvothermal reactor resulted in a significant shortening of reaction time (from several hours to minutes) and very good quality of the obtained product (uniform shape and narrow size distribution).

Using potassium oxalate monohydrate as an activator agent resulted in a high volume of micropores in the material, which are responsible for CO2 adsorption at atmospheric pressure. EDA by itself did not improve the physicochemical properties of the carbon material, as shown in CO2 uptake. However, use of both modificators led to the formation of a highly microporous material exhibiting both large specific surface areas and high CO2 uptake. It is thus concluded that surface morphology (microporosity) and surface chemistry, especially an amine promotor, lead to the best CO2

adsorption profile. Further, the differences between low temperature CO2 adsorption (0 ◦C), where physisorption dominates, and higher temperature adsorption, where chemisorption is more dominant, highlights the importance of surface chemical engineering of nanocarbon materials and, additionally, the importance of surface analysis in process development and optimization using both carbonization and modificators.

**Author Contributions:** Investigation, P.S., D.S., D.M. and R.J.W.; Methodology, R.D.C.; Supervision, U.N.; Writing—original draft, P.S. and U.N.; Writing—review & editing, R.D.C.

**Funding:** This research was partially funded by the Polish-Norwegian Research Programme operated by the National Centre for Research and Development under the Norwegian Financial Mechanism 2009–2014 in the frame of Project Contract No. Pol-Nor/237761/98.

**Conflicts of Interest:** The authors declare no conflict of interest

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).
