Removal of Ions From Produced Water Using Powder River Basin Coal

Although becoming less attractive as an energy source, coal has signicant potential for other, more sustainable uses including water treatment. In this study, we present a simple approach to treat water that was produced during oil production and contained a total dissolved solids (TDS) content of over 150 g/L using Powder River Basin (PRB) coal. PRB coal used as packing material in a ow-through column effectively removed 60-80% of the cations and anions simultaneously. Additionally, 71-92% of the total organic carbon in the produced water was removed as was all of the total suspended solids. The removal mechanisms of both cations and anions were investigated. Cations were removed by ion exchange with protons from oxygen-containing functional groups such as carboxylic and phenolic hydroxyl groups. Anions, mainly Cl -1 , appeared to be removed through either the formation of resonance structures as a result of delocalization of electrons within coal molecules or through ion-π interactions. We propose that coal is a “pseudo-amphoteric” exchange material that can remove cations and anions simultaneously by exchanging ions with both ionized and non-ionized acids that are ubiquitous in coal structure or resonance effect.


Introduction
Global population growth and climate change have made sustainable water supply a challenge to all human beings. Hoekstra estimated that 2 billion people are living in areas with severe water scarcity (Hoekstra 2014). The same author updated the number to 4 billion in a later model (Mekonnen and Hoekstra 2016). Yet "produced water" is produced in large quantities concurrently with oil production, gas production, and other unconventional oil and gas production including coalbed methane and shale oil and gas extraction (Stoll et al. 2015;Veil et al. 2004;Vengosh et al. 2014). The water-to-oil ratio for oil production is about 3:1 (SPE). The oil and gas industries produce about 14.5 billion m 3 /yr (250 million barrels/day) of produced water, of which 40% is discharged into the environment (Fakhru'l-Razi et al. 2009). Produced water accounts for more than 80% of the liquid waste (Igunnu and Chen 2012) produced during oil and gas production.
Although a signi cant potential source of water, produced water contains organic and inorganic compounds. Speci cally, dissolved and dispersed oils (BTEX, PAHs, and phenols), grease, heavy metals, radionuclides, treating chemicals, formation solids, salts, dissolved gases, scale products, waxes, microorganisms, and dissolved oxygen are commonly found in produced water. Furthermore, the treatment and disposal of produced water are costly. The expense for treating this potential resource is estimated at $40 billion per year (SPE) while the disposal cost on a unit basis could be as high as $4/barrel, depending on the quality of the produced water and the methods that are used (Duraisamy et al. 2013).
Nevertheless, due to the need for water and the desire to limit environmental damage due to discharge, treatment of produced water for potable and irrigation uses has become an option (Qi et al. 2021).
Conventionally, produced water is disposed by direct discharge or reinjection into disposal wells (Duraisamy et al. 2013;Jiménez et al. 2018), thus providing no bene cial use of the water. Physical treatment processes include adsorption, sand lters, hydrocyclones, evaporation, dissolved air precipitation, C-TOUR, freeze-thaw evaporation, devaporation, electrodialysis/elesctodialysis reversal, gas otation, and macro-porous polymer extraction. Chemical treatment processes include chemical precipitation, chemical oxidation, electrochemical processes, photocatalytic treatment, in situ chemical oxidation (ISCO), room temperature ionic liquid, and demulsi er (Dickhout et  However, membrane fouling and high capital and operations costs remain a big hurdle for these membrane ltration technologies (Duraisamy et al. 2013;Stoll et al. 2015). Hackney and Wiesner estimated the cost for treating average quality produced water to remove most of solids, organic, and inorganic components with unit processes to be $8.06/m 3 adjusted to the in ation rate of 2019 (Hackney and Wiesner 1996) and increasing up to $35.00/m 3 for produced water having high concentrations of organic constituents, total suspended solids (TSS), and total dissolved solids (TDS), with TDS removal dominating the cost.
Another fossil fuel, coal, may have potential for cost-effectively treating produced water from other fossil fuel production processes. Coal accounts for over 88% of the worlds' fossil fuels (Wang et al. 2019), is widely available and found in about 100 countries all over the world (Andruleit et al. 2016) and is predicted to outlast other hydrocarbon resources by hundreds of years. Furthermore, the conventional uses of coal for power generation, steel making, and chemical feedstock production (Falbe et al. 1982) are becoming less attractive because these uses produce pollutants and emit more CO 2 than any other energy sources (Huang et al. 2017). Coal is now being examined as an inexpensive raw material for  (Song et al. 2006). The preparation required carbonization of the coal at an elevated temperature for up to 900 ºC. Sulfonated or ammoniated coal were prepared with chemical reactions to add additional functional groups to enhance the ion exchange capacity of coal (Nachod 2012).
These uses require processing of coal in some manner prior to use. Yet coal is a complex material that may have inherent properties suitable for treating produced water directly, without extensive processing of the coal. The objectives of this work were (a) to test the hypothesis that native subbituminous coal from Wyoming's Powder River Basin can effectively treat produced water by removing extremely high concentrations of TDS as well as organic carbon, and suspended solids simultaneously, and (b) to evaluate the ion exchange and sorption mechanisms by which cations and anions are removed. To the best of our knowledge, this is the rst work to investigate the treatment of produced water with native coal directly for simultaneous removal of suspended solids, organic, and inorganic components.

Coal and Produced Water
Powder River Basin coal from Wyodak was provided by Black Hills Corporation. Coal was milled by Wyoming Analytical Lab (Laramie, WY, USA). The fraction of coal with particle sizes between 40-60 mesh was used for column preparation. The raw coal contained 16.00% moisture, 8.18% ash on a moisture-free basis, and 48.73% volatile matter on a dry and ash-free basis. The elemental composition of the coal was 78.87% C, 3.72% H, 1.01% N, and 15.93% O on a dry and ash-free basis (Liu et al. 2018).
The pore properties of the coal were BET speci c surface area (N 2 at 77 K) 2.598 m 2 /g, DFT pore diameter 0.844 nm, average pore diameter 13.64 nm, and total pore volume 0.018 cm 3

Experimental Design and Operation
All experiments were conducted by passing produced water through a column packed with coal. A 500mm long, 25-mm ID Kontes chromatography column equipped with a PTFE stopcock plug (Kimble Chase, Rockwood, TN, USA) was packed with ground coal (30 g). A pinch of cotton was used as a strainer right before the stopcock plug to strain the coal particles. The coal was washed with 100-mL of deionized (DI) water that was passed through the packed coal by gravity. The nal height of packed coal was approximately 65 mm.
Filtration experiments were conducted to examine the removal of suspended solids, organic, and inorganic compounds from produced water. An aliquot of 30 mL produced water was ltered (by gravity unless otherwise stated) through the column packed with coal. The ltrate was collected and designated as the sample to be analyzed for that cycle. Then the column was washed with 30 mL of 1% HCl (wt/vol, 37% TraceMetal Grade, Fisher Scienti c, Pittsburgh, PA, USA), followed by a DI water rinse until the pH of the wash was circumneutral. After washing, another 30-mL aliquot of produced water was ltered through the same coal column. This ltration-washing cycle was repeated 10 times.
To investigate the impact of functional groups on the water treatment, the coal was extracted with solvents including NMP (N-Methyl-2-pyrrolidone), (THF Tetrahydrofuran), and methanol before being placed in the column. The ground coal was soaked in solvent at a ratio of 1 to 3 (coal wt/solvent wt) at ambient temperature for one week. The coal-solvent mixture was then ltered through a lter paper (Whatman qualitative, Grade 1, GE, Pittsburg, PA) and the coal was washed with that solvent until the ltrate contained no color. The coal was recovered and dried at 80 ºC. The dried, extracted coal was then used in the ltration apparatus as described for the sequential ltration experiments.

Analytical Measurements
Conductivity, TDS and pH were measured by an H280G pH, conductivity and DO meter (Hach, Longmont, CO, USA). The alkalinity measurement was done with an alkalinity test kit AL-AP MG-L (Hach, Longmont, CO, USA). TOC was measured for both raw and treated produced water. TOC was analyzed with a Shimadzu TOC analyzer (TOC V CSN, Shimadzu Corporation, Kyoto, Japan). The original water sample was also analyzed with a Varian Cary Eclipse Fluorescence Spectrophotometer (Agilent, Walnut Creek, California, USA) for 3D-EEM under emission 3D mode. The scanning setup is as follows: emission wavelengths (Em) 290-590 nm with 2 nm increments; excitation wavelengths (Ex) 225-450 nm with 2.5 nm increments, slit bandwidths 5 nm for both emission and excitation; and scan rate 1200 nm/min were effectively removed by ltration through the coal column (Fig. 1, Supplementary Table 1). Ten ltration-regeneration cycles were conducted and showed 60-80% removal of individual ions with an average removal of 73%. Interestingly, the percent overall removal of all ions was similar, ranging between 64% and 76%, suggesting there was no speci c removal selectivity of these ions. An average of about 0.12 g of TDS was removed per g coal for each ltration-regeneration cycle, although there is residual water in the pores of the packed column after washing. There may be dilution effects from the residual water that need further investigation.
The TOC concentration of the original produced water was 42 mg/L. During ltration through the coal column, 71-92% of the TOC was removed. The removal e ciency of the initial cycle (Cycle-1) was the lowest (71%), whereas the mean value of the remaining 9 cycles was 90%. The removal capacity of organic components was signi cantly improved after coal regenerations. The TSS was decreased from 233 mg/L in the raw produced water to a negligible level (below detection) in the ltrate. Figure 2 shows SEM and EDS images of the coal samples directly after ltration and also after regeneration with the acid wash. The EDS analysis (Fig. 2b) clearly shows adsorption/deposition of major inorganic ions on the surface of coal particles. After regeneration with the acid wash, most of the cations were removed (Fig. 2d), thus con rming how the treatment cycles worked. Ions were retained in the coal through ltration and subsequently removed by acid wash, thereby regenerating the coal medium for further ltration cycles. After regeneration with the acid wash, Cl dominated the surface of coal (Fig. 2d) because diluted HCl solution was used as the washing agent.

Results of 13 C NMR analysis
The raw coal, washed coal before ltration (Before ltration), and coal after ltration (After Filtration) were subjected to NMR analysis to identify functional groups and investigate the changes of these functional groups. As shown in Fig. 3, many oxygen-containing functional groups, including alcohol, carboxyl, carbonyl, phenol, ester, and ether were identi ed in all samples with different intensity (Kim et al. 2013). Other non-oxygen-containing functional groups including aliphatic and aromatic C-H groups such as methyl, methylene, and methyne were also identi ed. This is in line with other studies suggesting low rank coal contains hydroxyl-, methoxy-, and/or methyl-substituted benzene rings, carboxylic acids, and aliphatic linkers line -CH In general, the intensity of the three samples decreased in the order Before ltration > raw coal > After ltration. Before ltration coal was raw coal that had been washed in preparation for use in the column. The washing process removed ions within the coal. In contrast, the After ltration sample contained ions that were removed from the produced water and were bound/adsorbed onto the coal. Because the electron distribution of 13 C can be affected by factors such as binding partners, the binding of ions can signi cantly reduce the response in NMR analysis. This may explain the intensity differences among coal samples of before ltration, raw and after ltration. In particular, the differences in intensity of the phenolic and carboxylic functional groups were prominent, suggesting their involvement in the removal of ions from the produced water.

Effects of Solvent Extraction on Coal and Solvents
To investigate the mechanisms of ion removal by coal ltration, PRB coal was solvent-extracted with tetrahydrofuran (THF), methanol, or N-methyl-2-pyrrodidone (NMP). The total ion removal capacity of the coal (as determined from a ltration cycle without subsequent acid washing) was signi cantly decreased by solvent extraction, to 26%, 23%, and 16% removal for THF, methanol, and NMP extractions, respectively.
Solid coal, both before and after extraction, and liquid solvent after extraction were examined by FTIR analysis. Although there are di culties in using FTIR to analyze heterogeneous materials like coal with respect to sample preparation, band assignments, and baseline correction (Solomon and Carangelo 1982), a number of functional groups and changes due to solvent extraction were identi ed in this study. Figure 4 shows the FTIR spectra of coal and liquid from solvent extraction. The peaks were identi ed according to the literature (Sigma-Aldrich ; Xie 2015). The coal structure was modi ed by solvent extraction to different extents (Fig. 4a). Speci cally, coal extracted with NMP exhibited the greatest structural changes with respect to functional groups, followed by methanol and THF. This is consistent with the ion removal capacity where NMP extraction showed the greatest decrease in capacity, followed by methanol and then THF which showed the smallest decrease in ion removal capacity.
The liquid solvents after extraction contained compounds with oxygen-and nitrogen-containing functional groups (Fig. 4b), consistent with the changes in the solid coal functional groups. The intensity of N-H stretching for solvent-extracted coal was signi cantly reduced, whereas peaks of N-H stretching existed in all the liquid extracts. The intensity of the N-H peaks of the coal followed the same tendency of ion removal capacity, with less ion removal capacity corresponding to decreased intensity of the N-H peaks. This was also true for other functional groups such carboxyl and phenolic-hydroxyl groups, suggesting that these functional groups may be involved in ion removal. These results are most evident in the NMP treatment where the intensity of these peaks in the extracted coal were signi cantly reduced while the NMP after extraction had a strong presence of these peaks. NMP has been shown to facilitate the extraction of hydroxyl-containing moieties from bituminous coal (Sun et al. 2014). These results indicate that the extraction of these functional groups signi cantly impaired the ion removal capacity of the coal, largely related the removal capacity to the number of these oxygen-and nitrogen-containing moieties.

Composition of coal extracts
Volatile and small-molecular compounds in the extracts were characterized by GC/MS to investigate the impact of extractable components on the ion removal. As shown in Fig. 5, the group components detected in the extracts mainly included alkanes, alkenes, arenes, alcohols, phenols, ketones, carboxylic acids, and esters. The relative abundances of oxygen-containing compounds (i.e., alcohols, phenols, ketones, carboxylic acids, and esters) compared to all extracted compounds were 47.2%, 86.4%, and 82.0% for THF, methanol, and NMP extracts, respectively. Oxygen-containing functional groups, especially carboxyl and hydroxyl, in these compounds may interact with metal cations to form cation-bridging linkages (Liu et al. 2016; Mathews and Chaffee 2012), which are bene cial for ion removal. Extraction of these oxygen-containing compounds appears to have signi cantly reduced the ion removal capacity of the coal.
The detected compounds in the methanol and NMP extracts were dominated by esters with 76.6% and 52.9% relative abundance, respectively (Fig. 5), while the hydroxyl-containing compounds (like carboxylic acids, alcohols, and phenols) had a relatively low abundance. As polar solvents, methanol and NMP have proven to be effective for extracting hydroxyl-containing compounds from coals (Liu et al. 2016;Sun et al. 2014). Because GC/MS is only sensitive for volatile and less polar compounds, the GC/MS detectable compounds account for only a fraction of the compounds in the extracts. According to FTIR analysis, NMP extract may also contain many hydroxyl-containing compounds, especially carboxylic acids with low carbon numbers, which were not detected by GC/MS.

Proposed Mechanisms Of Ion Removal
The structure of a coal macromolecule may be visualized as a condensed aromatic carbon-atom lattice surrounded by a typical "fringe" formed by functional side groups. The left panel of Fig. 6 is a hypothetical model of coal structure (Malumbazo 2011). It is a heterogeneous mixture composed of a macromolecule network with varying degrees of cross-linking (Smith et al. 2013). Modi ed lignin, as well as cellulose and melanoidin-type materials, are considered to be the 'backbone' of this macromolecule network. The cross-linkage of lower rank coal, including subbituminous coal, is dominated by alkyl and aryl ether groups with oxygen functional groups. The chemical heterogeneity of coal decreases from low rank coal to high rank while the aromaticity increases, suggesting that lower rank coals (lignite and subbituminous) have more complex chemical structures than high-rank coals (bituminous coal and anthracite) because the low rank coal contains several distinct classes of constituents (Hofrichter and Fakoussa 2001;Wang et al. 2015). Carboxyl and hydroxyl groups, among others, are the main oxygencontaining functional groups in coal structure which are present in low-rank coal (Xie 2015). Phenolic hydroxyl groups are the main form, but some alcoholic hydroxyl groups also exist.
The removal of cations from produced water by coal ltration is proposed to be through ion exchange (Fig. 6a). The carboxyl and hydroxyl groups may act as ion exchanger sites by exchanging protons (H + ) for other cations (Na + , K + , Ca 2+ , Mg 2+ , etc.) allowing these cations to bind to the negatively charged hydroxyl and carboxyl functional groups. This hypothesis is supported by the pH measurements of the raw produced water before ltration (6.83) and the treated water (as low as 2.28), indicating that protons were released by the coal, presumably due to cation exchange. Additionally, NMR analysis of before and after ltration coal samples also showed signi cant intensity reduction of carboxyl and hydroxyl groups, further supporting the cation exchange hypothesis. The solvent extraction results provide addition support for the hypothesis. Solvent extraction removed carboxyl and hydroxyl functional groups, as illustrated by the FTIR analyses where the functional groups were removed from the coal and present in the extracts, resulting in signi cantly impaired ion removal capacity.
Proposed mechanisms for Cl − removal are less clear. One possible removal mechanism is adsorption of Cl − due to the delocalization of electrons within molecules to form resonance structures (Fig. 6b). This resonance effect or mesomeric effect or electron-donating effect occurs between a lone pair of electrons and a pi bond or two pi bonds next to each other (Dewick 2006). For example, the lone electron pair in the oxygen of the phenolic hydroxyl group may be donated to form a double bond and leaving a positive charge. The donation stabilizes the structure of the non-ionized acid such as phenols or derivatives. The donating effect passes the electrons on to the pi bonds along the ring to produce a negative charge of the para-or ortho-carbon in the same aryl ring. These charged molecules could then bond with both negative and positive ions in the produced water to remove them. The resonance effect could be positive or negative. In the positive resonance effect, -OH, -SH, -OR, and -SR can increase the electron density of the stabilizing ring while -NO 2 , -S = O, and -C = O could decrease the electron density of the stabilizing ring in the negative resonance effect (Dewick 2006). The removal of these functional groups, as occurred with solvent extraction, would reduce the ion removal capacity of coal.
Alternatively, the ion removal of any particular ion may not be attributed to any single functional group.
Other possible mechanisms to produce charges in coal include changes of electron density by binding alkali metal ions (which are dominant species in produced water) with the functional groups in aromatic structures and electrostatic induction (Xie 2015). In recent decades, ion-π interactions have been recognized and found to be widely exist as a form of general noncovalent bonding (Dougherty 1996;Ma and Dougherty 1997;Schottel et al. 2008). Ion-π interactions happen not only in aromatic systems, but are also well documented in other simple π systems such as ethylene and acetylene. Studies show that highly solvated cations can be sequestered by such binding force in aromatic-containing structures (Ma and Dougherty 1997), while anion-π interactions happen in electron de cient aromatic systems (Schottel et al. 2008). The anion-π interaction combines effects of electrostatic and anion-induced polarization, with the former correlated to permanent quadruple moment, Q zz and the latter to molecular polarizability (Quiñonero et al. 2004;Schottel et al. 2008). Aromatic molecules with lower absolute values of Q zz could bind to both anion and cation which might account for some of the ion removal in our system (Schottel et al. 2008). The ion-π interaction binding energy is estimated to be 20-50 kJ/mol which is energetically favorable and is comparable to the binding energy of hydrogen bonds. In the ion-π theory, the inductive effect, rather than the resonance effect, facilitates binding(Ma and Dougherty 1997; Schottel et al. 2008).
Although the mechanisms of anion removal remain hypothetical, based on the results presented herein, coal may be considered a "pseudo-amphoteric" exchanger that has the capability of removing both cations and anions simultaneously and effectively.

Conclusions
In this study, we have demonstrated that TOC, TSS, and especially TDS can be removed simultaneously from produced water using Power River Basin subbituminous coal as the treatment material. The material is stable with repetitive ltration-regeneration cycles. More importantly, the coal was shown to have a surprisingly high ion removal capability for produced water containing extremely high dissolved solids (> 150,000 mg/L or 15%). We propose that the PRB coal is a "pseudo-amphoteric" ion exchange material that could simultaneously remove both positive and negative ions. The analyses strongly support the hypothesis of ion exchange as the mechanism for cation removal whereas the mechanism for anion removal is proposed to be by adsorption through resonance structures induced by delocalization of electrons within coal molecules or through ion-π interactions with aromatic clusters. Further work is needed to con rm the anion removal mechanisms. These results may have important rami cations in the reclamation of produced water from energy production sector.

Declarations
Availability of data and materials The data supporting the ndings of this study are available within the article and its Supplementary Information les.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.