Water Quality Evolution-Guided Pilot-Scale Hardness Control of Saline Coalbed Methane Fracturing Flowback Fluids

Abstract

Hydraulic fracturing of the Benxi No. 8 coalbed methane reservoir produces saline coalbed methane fracturing flowback fluids whose chemistry changes during flowback, creating a moving treatment target for hardness control, scaling prevention and internal oilfield reuse. This study combined time-resolved gas and water monitoring from Well M, CO₂ pressurization experiments with carbonate bearing drill cuttings and pilot scale treatment to link water quality evolution with fit-for-purpose process design. During flowback, total mineralization increased from 34502.17 to 324236.66 mg/L, pH decreased from 6.56 to 4.17, and Cl^– and Ca²⁺ increased from 20121.26 to 199918.01 mg/L and from 3191.37 to 38216.28 mg/L, respectively. Mg²⁺ reached 6059.27 mg/L, while total organic carbon decreased from 5053.62 to about 1200 mg/L but remained persistent in late flowback. The enrichment of Cl^– and ΣNa⁺ + K⁺ indicated an increasing formation derived brine contribution, while pressurized tests confirmed that dissolved CO₂ promoted Ca²⁺ release from carbonate bearing solids. In pilot operation, alkaline oxidation removed iron ions to below the detection limit and shifted most Mg²⁺ removal upstream. CO₂ microbubble softening converted residual Ca²⁺ into crystalline CaCO₃, and crossflow ceramic membrane filtration removed the generated carbonate particles. The final effluents contained no detectable suspended solids or iron ions, with Ca²⁺ of 75.15 and 57.62 mg/L and Mg²⁺ of 12.16 and 31.30 mg/L in two treatment batches. The integrated process reduced hardness and scaling risk while retaining the dissolved salt matrix, providing a selective treatment route for internal oilfield reuse of saline coalbed methane fracturing flowback fluids.

1. Introduction

Coalbed methane (CBM) and coal related gas are now important parts of unconventional natural gas development in coal rich countries, including Russia, the United States, China, Canada and Australia. Commercial CBM production has been established in several of these regions, with the United States, Australia, China and Canada providing useful benchmarks for reservoir development and field operation. As development expands, CBM production is moving from shallow and relatively permeable coal seams toward deeper reservoirs that have lower permeability, stronger stress sensitivity and greater dependence on hydraulic fracturing. This shift is especially relevant in China, where many coal reservoirs have low permeability and deep seams are commonly compressed by high in situ stress, making stimulation and water management central to field development. , The Ordos Basin is one of China’s main regions for CBM and coal related gas exploration. The Carboniferous Benxi Formation has become a major target because its coal measures are widely distributed and have large resource potential. Basin scale estimates indicate that coal related gas resources in the Benxi Formation at burial depths greater than 2000 m exceed 12.33 × 10¹² m³. In the eastern Ordos Basin, structural and hydrological conditions further influence reservoir pressure, water residence time and gas preservation. Within this setting, the Benxi No. 8 coal seam is an important interval for CBM and coal related gas production. Field development of this reservoir generates coalbed methane fracturing flowback fluids (CBM-FFFs), making flowback water management a direct constraint on wellsite operations, water reuse, treatment costs and environmental risk.

CBM-FFFs are difficult to manage because their chemistry changes during flowback. Returned fracturing fluid, formation brine, coal fines, dissolved organic matter and continuing water-coal and water-rock reactions all contribute to the final water quality. − This problem is amplified in the Benxi No. 8 reservoir, where deep formation water in Ordos Basin is commonly saline and CaCl₂ rich, with calcium (Ca²⁺) and magnesium (Mg²⁺) forming a large fraction of the cation load. Under these conditions, salinity increase is coupled with hardness accumulation, carbonate scaling risk and higher treatment demand. Existing treatment options can remove selected contaminants, but their performance is limited when high-salinity, high-hardness and suspended solids (SS) occur at the same time. − Flocculation and sedimentation can reduce SS, oil and turbidity, but they do not address dissolved salts or persistent dissolved organic matter. , Advanced oxidation can degrade part of the organic fraction, but reagent demand, energy input and byproduct control restrict its use as a standalone step. , Biological treatment is sensitive to high-salinity, toxic additives and rapid water quality changes. , Membrane processes can produce higher quality effluent, and integrated membrane systems have been used for fracturing flowback fluids, yet fouling, concentrate management and pretreatment needs remain major barriers when salts, hardness ions and SS increase together. , Produced water may also contain recoverable dissolved constituents, although such resource pathways depend on reliable pretreatment and separation before further use.

For later stage CBM-FFFs, hardness control is a more practical treatment target than full desalination. Ca²⁺ and Mg²⁺ control carbonate scaling, influence membrane stability and determine whether treated water can be reused onsite. − Conventional hardness removal methods include lime or soda precipitation, ion exchange, nanofiltration, induced crystallization and electrochemical softening. − These methods can be effective, but they also introduce trade-offs such as chemical consumption, sludge production, secondary brine generation, fouling, concentrate disposal, cathode scaling and long-term stability concerns. For internal oilfield uses, including reinjection, drilling fluid preparation, fracturing fluid preparation and wellsite circulation, complete desalination is often more treatment than the reuse target requires. − A fit for purpose process should instead remove the components that most directly affect reuse. For saline CBM-FFFs, these components are SS, iron, hardness forming ions and scaling precursors, while most monovalent salts can remain in solution when they do not limit the intended application.

The processes responsible for salinity and hardness evolution in CBM-FFFs from the Benxi No. 8 reservoir are still not well constrained. This gap matters because a treatment process selected for early return water may become unsuitable when salinity, hardness, iron and residual organic matter increase during continued flowback. The motivation of this work was to link field scale water quality evolution with process selection for fit for purpose reuse. In this study, gas composition, time-resolved water chemistry and pilot scale treatment performance were combined to define a hardness control route for saline fracturing flowback fluids from Well M in the Benxi No. 8 reservoir. The study first tracks changes in gas composition, salinity, hardness, pH, iron, total organic carbon (TOC) and SS during flowback. It then uses pressurized CO₂ dissolution experiments with carbonate bearing drill cuttings to evaluate whether dissolved CO₂ can promote Ca²⁺ release and contribute to hardness accumulation. Finally, it tests a pilot scale process that integrates alkaline oxidation pretreatment, CO₂ microbubble softening and crossflow ceramic membrane filtration for later stage flowback fluids. This process was designed to stabilize the feed, remove iron and most Mg²⁺ during pretreatment, convert residual Ca²⁺ hardness into separable carbonate solids and remove the generated particles without deep desalination. The CaCO₃ rich precipitate may also have potential value in oilfield applications because acid soluble calcium carbonate (CaCO₃) is used as a bridging and filter cake forming material, and modified ultrafine CaCO₃ has been evaluated for enhanced oil recovery. This study links flowback water evolution to pilot scale hardness control and provides a treatment route for scale risk reduction and internal oilfield reuse of saline CBM-FFFs from the Benxi No. 8 reservoir.

2. Study Area

The Benxi No. 8 coal seam is located in Ordos Basin, one of the major regions for deep coalbed methane and coal related gas development in China. The Carboniferous Benxi Formation contains widely distributed coal measures and has become an important exploration target because of its large gas resource potential. Within this formation, the No. 8 coal seam is a key production interval because of its stable distribution, large burial depth and favorable gas enrichment conditions. In this study, the geological setting is considered mainly in terms of how it controls the generation, mixing and hydrochemistry of coalbed methane fracturing flowback fluids. Structural setting, hydrodynamic conditions and aquifer connectivity regulate fluid preservation, water residence time, pressure conditions and water rock interaction, which together define the hydrochemical baseline of the produced water.

2.1. Structural Geology

The location of the study area was mapped in the eastern Ordos Basin near Suide County (Figure ). The area lies near the transition from the central gentle slope to the eastern margin structural belt. The Ordos Basin has a relatively stable cratonic framework and is characterized by broad uplifts, slopes and depressions. This structural framework commonly produces low dip angles and weak deformation in coal bearing strata. Regional groundwater studies also show that deep groundwater circulation in Ordos Basin is controlled by structural pathways, hydraulic gradients and basin scale recharge patterns. These controls are relevant to coalbed methane reservoirs because structural stability favors gas preservation, while restricted hydrodynamic exchange can extend water residence time and strengthen water rock interaction.

1.

Location of Ordos Basin and the study area.

Local faults and fracture sets may occur in the Suide area, but they are generally short and discontinuous. They are therefore unlikely to form persistent vertical conduits across the Benxi No. 8 coal seam. At the reservoir scale, fluid communication is controlled mainly by the overall structural stability and by the sealing capacity of the roof and floor around the target seam. Recent work on high salinity coal reservoir water in Ordos Basin shows that long residence time, evaporation concentration, cation exchange and mineral dissolution can all contribute to highly mineralized formation water. This structural and hydrodynamic setting provides the geological basis for the high salinity and major ion composition observed in the flowback fluids.

2.2. Aquifers

Aquifers in the coal measure strata of the eastern Ordos Basin influence methane preservation and define the hydrochemical background of the flowback fluids. Their effect depends on hydraulic connectivity. Thin and laterally discontinuous water bearing layers may provide local water storage, but they are less likely to sustain continuous vertical leakage through the target coal seam. Broader studies of water gas rock interaction in coal measure systems show that formation water chemistry is commonly shaped by carbonate dissolution, evaporite dissolution, cation exchange and sulfate related reactions during basin evolution. These processes are directly relevant to the Benxi No. 8 reservoir because they determine the ionic composition of formation water before hydraulic fracturing.

During hydraulic fracturing and subsequent flowback, injected fluid enters the coal matrix and fracture network. This process can trigger imbibition, ion diffusion, mineral dissolution, secondary precipitation and ion exchange. Field and laboratory studies have shown that water rock interaction during hydraulic fracturing can substantially change the chemistry of returning water, especially through the release of major ions and hardness forming ions. Mineralized water can also affect coal wettability and coal fluid interaction, which may influence the transport of aqueous fluids in coal seams. The flowback chemistry in the study area should therefore be interpreted as the combined result of formation brine mixing, water coal interaction and mineral reactions during flowback.

Chloride (Cl^–) and divalent cations are useful indicators in this hydrochemical setting. Chloride traces the increasing contribution of formation derived brine, while Ca²⁺ and Mg²⁺ reflect hardness accumulation and scaling potential. Recent work on coalbed methane stimulation also shows that salinity gradients and fluid imbibition can regulate ion migration, gas water interaction and flowback behavior in deep coal seams. These processes provide the hydrochemical basis for identifying the main treatment constraints of the flowback fluids. They also explain why the pilot process in this study prioritizes iron removal, hardness control and scale prevention before ceramic membrane filtration. The mineralized water system and coal surface response support treating salinity, hardness and SS as coupled treatment constraints rather than isolated water quality parameters.

3. Materials and Methods

3.1. Reagents and Feedwater

Analytical grade reagents were used as received. Hydrochloric acid (HCl), sodium hydroxide (NaOH), sodium carbonate (Na₂CO₃), sodium bicarbonate (NaHCO₃), sodium chloride (NaCl), potassium chloride (KCl), magnesium chloride (MgCl₂), anhydrous calcium chloride (CaCl₂), CaCO₃, ethylenediaminetetraacetic acid (EDTA), disodium ethylenediaminetetraacetate (Na₂EDTA), ammonium chloride (NH₄Cl), ammonia solution (NH₃·H₂O), silver nitrate (AgNO₃), potassium chromate (K₂CrO₄), phenolphthalein (C₂₀H₁₄O₄), methyl orange (MO), Eriochrome Black T (EBT) and citric acid (C₆H₈O₇) were used for water quality analysis and volumetric titration. These reagents were used to determine Cl^–, Ca²⁺, Mg²⁺, total hardness, bicarbonate (HCO₃ ^–) and carbonate (CO₃ ^2–). Sodium hydroxide (NaOH) and sodium hypochlorite (NaClO) were used for alkaline oxidation pretreatment and NaOH was also used for pH control during Carbon dioxide (CO₂) microbubble softening. CO₂ with a purity of at least 99.99% was used for pressurization experiments and CO₂ microbubble softening (CMBS). A 9.95 M NaOH stock solution was prepared with ultrapure water and used for pH adjustment. Ultrapure water was used for reagent preparation, sample dilution and blank controls. Real CBM-FFFs were collected from the Second Gas Production Plant of Changqing Oilfield, PetroChina. The samples were obtained from field flowback operations in the Benxi No. 8 reservoir and were used as feedwater for water quality characterization and pilot scale treatment experiments.

3.2. Sampling and Characterization of Gas and CBM-FFFs

Gas and CBM-FFF samples were collected from Well M in the Benxi No. 8 coal seam during flowback. Time matched gas and water samples were used to track the coupled evolution of gas composition and aqueous chemistry. ,,, Gas sampling and compositional analysis followed the relevant Chinese standards for natural gas quality and gas composition analysis. The molecular composition of the gas samples was analyzed using gas chromatography (GC) on an Agilent 8890 system equipped with a Pora PLOT Q capillary column (30 m × 0.25 mm × 0.25 μm). Helium was used as the carrier gas. The oven temperature was held at 70 °C for 6 min, increased to 180 °C at 15 °C/min, and then held for 4 min. The relative error of the gas analysis was less than 0.5%.

The aqueous phase was characterized using inductively coupled plasma mass spectrometry (ICP MS), ion chromatography, volumetric titration and TOC analysis. Samples were diluted as needed before instrumental analysis to keep all analytes within validated calibration ranges. Sodium and potassium were quantified using ICP MS. For highly saline CBM-FFFs, Cl^–, calcium, magnesium and total hardness were measured by volumetric titration according to standard oilfield water analytical procedures. Sulfate was measured using ion chromatography with an AG14 AS14 column. The eluent contained 8 mM Na₂CO₃ and 1 mM NaHCO₃, and the flow rate was 1.2 mL/min. HCO₃ ^– and CO₃ ^2– were determined by acid titration. Total mineralization (TM) was calculated as the sum of measured dissolved inorganic ions in each sample, including ΣNa⁺ + K⁺, Ca²⁺, Mg²⁺, Fe²⁺, Fe³⁺, Cl^–, SO₄ ^2–, HCO₃ ^– and CO₃ ^2–. This calculation follows the oilfield water analytical convention used for reporting total dissolved mineral content, as specified in SY/T 5523–2016, Method for Analysis of Oilfield Water, and the values are reported in mg/L. TOC was measured using a TOC analyzer after acidification and purging to remove inorganic carbon. pH, conductivity, temperature, density, SS and iron speciation were measured using standard methods for oilfield water characterization. All concentrations are reported in mg/L unless otherwise stated.

3.3. Experimental Evaluation of CO₂-Induced Carbonate Dissolution in Simulated Formation Brine

A CO₂ pressurization experiment was used to test whether dissolved CO₂ can mobilize Ca²⁺ from carbonate bearing coal seam drill cuttings and thereby contribute to the hardness increase observed in CBM-FFFs during flowback. , The test was designed as a controlled batch experiment to isolate the carbonate dissolution pathway from the broader field process, where conservative ions are mainly supplied by formation brine inflow. The experimental workflow is shown in Figure .

2.

Experimental workflow for CO₂ induced carbonate dissolution in simulated formation water.

A NaCl based brine with a total dissolved solids concentration of 80,000 mg/L was prepared as the simulated formation water. This fixed salinity was selected to represent a high salinity brine matrix while allowing the effect of CO₂ pressure to be isolated. The brine contained no added Ca²⁺ or Mg²⁺, so any increase in dissolved Ca²⁺ during the experiment could be attributed mainly to CO₂ induced dissolution of carbonate minerals from the added drill cuttings. A 2000 mL aliquot of the simulated formation water was adjusted to approximately pH 7 with HCl and NaOH, transferred into a 2000 mL stainless steel high pressure reactor, and mixed with 10 g of drill cuttings collected from the horizontal section of Well M. The sealed reactor was pressurized with CO₂ to 7, 11, or 15 MPa, corresponding to the reservoir pressure range of the study area. The reactor was operated at 50 °C and 200 rpm for 2 h before sampling began. Over the next 7 d, a 50 mL sample of CO₂ saturated simulated formation water was collected every 24 h, and Ca²⁺ was measured immediately according to SY/T 5523–2016. After each sampling event, CO₂ was reintroduced for 1 h to restore the target pressure. The time dependent change in dissolved Ca²⁺ was used to quantify CO₂ induced carbonate dissolution under a fixed water volume, fixed salinity and fixed solid inventory. ,

3.4. Pilot Treatment Process for Later Stage CBM-FFFs

Field monitoring showed that the treatment target for CBM-FFFs changed during flowback. Early return water was managed onsite by removing coarse solids, after which the treated water could be remixed with fracturing fluid for field reuse. As flowback continued, salinity and hardness increased, and simple clarification was no longer sufficient. Later stage CBM-FFFs were therefore treated using a pilot scale process that integrated alkaline oxidation pretreatment, CMBS and crossflow ceramic membrane filtration.

The pilot treatment train was organized into three sequential functions: feed stabilization, hardness conversion and solid liquid separation (Figure a). Alkaline oxidation pretreatment buffered feed variation and removed iron ions, readily settleable SS, most Mg²⁺ and part of the organic load. CMBS converted dissolved Ca²⁺ and residual Mg²⁺ into carbonate precipitates under controlled CO₂ dosing and pH conditions. Crossflow ceramic membrane filtration then retained the generated CaCO₃ rich solids and residual SS, producing a clarified effluent without deep desalination. The same pilot system was assembled and operated under field conditions (Figure b).

3.

Pilot-scale treatment system for CBM-FFFs, showing (a) the process flowchart and (b) the field pilot equipment.

3.4.1. Alkaline Oxidation Pretreatment

Alkaline oxidation pretreatment was used to stabilize the feed before CMBS and ceramic membrane filtration. Raw CBM-FFFs were first equalized in a holding tank and then kept under quiescent conditions to allow readily settleable solids to deposit. NaOH at 30 wt % was added to increase the pH to approximately 12.0. This alkaline step promoted Mg²⁺ removal mainly through Mg­(OH)₂ precipitation and also destabilized part of the organic fraction in the water. For residual polyacrylamide based organics, alkaline hydrolysis of amide groups was considered the primary transformation pathway, while oxidative main chain scission was treated as a secondary degradation process. NaClO at 10 wt % was then added to oxidize dissolved Fe²⁺ to Fe³⁺ and to further degrade reactive organic residues. Under the combined alkaline and oxidative conditions, iron ions were removed mainly through oxidation, hydrolysis and precipitation, while part of the Mg²⁺, SS and organic matter was transferred into the settled phase. No additional coagulant or polymer was used in this step. Raw water and alkaline oxidation pretreatment effluent samples were collected for TOC analysis to evaluate organic carbon removal during alkaline oxidation pretreatment. TOC was measured using a TOC analyzer after inorganic carbon removal by acidification and purging. After gravity settling, the supernatant was transferred to the CMBS unit, and the SS were removed periodically.

3.4.2. CO₂ Microbubble Softening

The softening experiments were performed using the alkaline pretreated supernatant of FFFs 11, which represents the later stage CBM-FFFs (Table ). The supernatant was fed into the softening reactor for CMBS. This unit was used to remove hardness by converting dissolved divalent ions into carbonate precipitates. Because alkaline oxidation pretreatment had removed most Mg²⁺, CMBS mainly targeted residual Ca²⁺, while the remaining Mg²⁺ was further reduced during carbonate precipitation. CO₂ was introduced through a microbubble generation system, and NaOH was continuously dosed to maintain the reaction pH at 9 to 10. The effects of CO₂ dosage and pH on Ca²⁺ removal efficiency and residual Ca²⁺ concentration were evaluated. Bulk CO₂ bubbling and CO₂ microbubbling were compared under the same CO₂ dosage and pH conditions. After softening, the slurry was allowed to settle, and the supernatant was then introduced into the downstream ceramic membrane unit. This arrangement allowed hardness conversion, gravity settling and ceramic membrane polishing to operate as sequential steps. ,

3. Geochemical Parameters of CBM-FFFs from Well M in the Benxi No. 8 Reservoir .

water ID	date	pH	TM	TOC	ΣNa+ + K+	Ca2+	Mg2+	Fe2+	Fe3+	CO3 2–	HCO3 –	Cl–	SO4 2–
FFFs 1	17 Jan 2025	6.56	34502.17	5053.62	8638.96	3191.37	744.49	82.03	13.61	ND	1565.16	20121.26	145.28
FFFs 2	1 Feb 2025	6.34	71561.02	3553.51	19156.70	5402.78	1778.88	54.80	ND	ND	2056.37	43029.16	82.31
FFFs 3	19 Feb 2025	5.36	186685.99	2236.24	43950.93	23577.06	2372.05	136.48	27.58	ND	489.69	116054.01	78.20
FFFs 4	3 Mar 2025	5.52	191580.56	1863.56	44850.20	23577.06	2964.60	29.32	40.84	ND	1411.09	118645.71	61.74
FFFs 5	19 Mar 2025	5.35	210536.60	1238.75	44188.96	31933.74	2372.05	115.89	40.84	ND	765.80	131049.36	69.97
FFFs 6	25 Mar 2025	5.35	202365.88	1220.70	41108.53	31432.74	2675.92	40.84	13.61	ND	675.03	126360.82	57.62
FFFs 7	4 Apr 2025	5.34	201344.14	1240.62	39739.88	31432.74	3267.87	327.42	139.63	ND	30.51	126360.82	45.27
FFFs 8	24 Apr 2025	5.84	189574.73	1238.81	27604.81	38807.46	2964.60	82.10	13.68	ND	552.23	119512.94	37.04
FFFs 9	12 May 2025	6.00	187965.82	1218.83	37083.27	29949.78	2675.92	54.80	82.03	ND	337.14	117733.49	49.39
FFFs 10	26 Jun 2025	4.93	303125.31	1248.38	67224.64	38847.54	6059.27	871.96	0.00	ND	137.30	189931.10	53.50
FFFs 11	21 Jul 2025	4.88	285081.31	1212.86	62361.42	39478.80	3991.09	1009.84	487.29	ND	262.39	177464.97	25.52
FFFs 12	15 Aug 2025	4.48	277774.01	1229.48	69855.38	30049.98	3991.09	1322.95	ND	ND	79.33	172441.53	33.75
FFFs 13	20 Sep2025	4.17	324236.66	1215.32	79300.13	38216.28	3602.13	2401.55	ND	ND	762.75	199918.01	35.81

^aNote: Concentrations are given in mg/L unless otherwise stated. pH is dimensionless. TM, total mineralization; ND, not detected.

3.4.3. Characterization of Softening Precipitates

The precipitates formed during CMBS were collected from the softened slurry and characterized to determine their particle size distribution, morphology, thermal decomposition behavior and crystalline phase. The particle size distribution of the recovered CaCO₃-rich precipitates was determined using a laser particle size analyzer (Malvern AWM2000, Malvern, UK). , Scanning electron microscopy (SEM; CLARA, TESCAN, Czech Republic) was used to investigate the morphology and aggregation structure of the dried precipitates. , Thermal decomposition was analyzed using a thermogravimetric analyzer (TGA DSC 1, METTLER TOLEDO, Switzerland) under an N₂ atmosphere. The sample was heated from room temperature to 850 °C at a rate of 10 °C/min. The crystalline phase of the precipitates was identified using an X-ray polycrystalline diffractometer (D8 ADVANCE, Bruker, Germany). These analyses were used to verify whether hardness removal during CMBS was governed mainly by CaCO₃ precipitation and to evaluate whether the generated solids had a particle structure suitable for downstream ceramic membrane separation.

3.4.4. Crossflow Ceramic Membrane Filtration

Because the softened water still contained a high load of suspended carbonate particles, a membrane separation step was introduced after softening. Crossflow ceramic membrane filtration was selected as the final solid liquid separation unit. The softened suspension flowed tangentially across the membrane surface to limit particle accumulation and maintain hydraulic stability under high salinity conditions. Ceramic media were selected because of their high permeability, chemical robustness, and tolerance to carbonate rich particulate slurries. The nominal pore size of the membrane was 0.2 μm. Three ceramic membrane discs were used, each with an effective filtration area of 0.20 m², corresponding to a total effective membrane area of 0.60 m².

The membrane system was evaluated under both constant flux and constant pressure modes. In constant flux mode, target permeate fluxes of 566.67, 666.67, and 800 LMH were imposed, and the transmembrane pressure was monitored continuously to track resistance growth during operation. In constant pressure mode, transmembrane pressure (TMP) values of 0.08, 0.10, and 0.12 MPa were applied, and permeate flux was recorded continuously to evaluate flux decline under fixed hydraulic driving force. Physical cleaning and backwashing were applied periodically during operation, and the response of TMP or flux after cleaning was used to evaluate fouling reversibility. Pure water flux was also measured over a range of TMP values to determine intrinsic membrane permeability.

To link hydraulic performance with fouling morphology, membrane surface characteristics before use and after fouling were observed by scanning electron microscopy. The fouled membrane samples were collected from the constant flux test at 566.67 LMH and from the constant pressure test at a TMP of 0.1 MPa, corresponding to the operating conditions used for fouling model fitting. SEM observations were performed using a CLARA scanning electron microscope (TESCAN, Czech Republic). The images were used to identify whether the dominant resistance was associated with pore entrance coverage, surface deposition, or cake layer formation, and to provide morphological evidence for interpreting the complete blocking and cake filtration models.

3.5. Hydraulic Analysis and Fouling Model Fitting

Membrane performance was evaluated by cumulative permeate volume, permeate flux, TMP, and suspended solids concentration in the slurry. Flux was calculated from permeate volume, membrane area, and filtration time. Under constant flux operation, the increase in TMP with time was used to characterize resistance development. Under constant pressure operation, the decline in permeate flux with time was used to evaluate fouling evolution. Pure water permeability was obtained from the linear relationship between flux and TMP.

To interpret membrane fouling behavior, four classical fouling models were fitted to the experimental data, namely complete blocking, standard blocking, intermediate blocking, and cake filtration. Under constant flux operation, model fitting was carried out using the linearized forms based on TMP evolution. Under constant pressure operation, fitting was carried out using the linearized forms based on flux decline. Linear regression was used to determine the fitting equations and correlation coefficients, and the dominant fouling mechanism under each operating mode was identified from the best fitting model. This analysis was used to distinguish pore restriction from surface cake growth and to connect the membrane behavior with the particulate characteristics of the softened suspension.

4. Results and Discussion

4.1. Gas Characteristics of CBM

The gas composition of Well M was analyzed to provide the gas phase background for interpreting the coupled evolution of CBM-FFFs (Table ). CH₄ remained the dominant component throughout the monitoring period, ranging from 93.10 to 96.80 mol %. Total hydrocarbon content ranged from 93.62 to 97.10 mol %, confirming that the produced gas was consistently hydrocarbon rich. Minor hydrocarbon contents varied over time. C₂H₆ increased from 0.23 to 0.51 mol %, while C₃H₈ ranged from 0.0134 to 0.0413 mol %. These components remained low, but their increase shows that the gas stream did not remain compositionally static during flowback.

1. Time Series Gas Composition of CBM from Well M in the Benxi No. 8 Reservoir.

gas ID	Ddate	CH4 (mol %)	C2H6 (mol %)	C3H8 (mol %)	total hydrocarbon (mol %)	relative density	critical temperature (K)	He (mol %)	H2 (mol %)	O2 (mol %)	N2 (mol %)	CO (mol %)	CO2 (mol %)	density (kg/m3)	critical pressure (MPa)
M1	17 Jan 2025	96.80	0.23	0.0413	97.10	0.57	192.34	0.03	0.0309	ND	0.94	ND	1.88	0.69	4.64
M2	1 Feb 2025	95.45	0.35	0.0134	95.82	0.59	195.44	0.02	ND	ND	0.12	ND	4.02	0.71	4.71
M3	19 Feb 2025	95.74	0.35	0.0154	96.11	0.59	194.91	0.02	ND	ND	0.23	ND	3.62	0.71	4.70
M4	3 Mar 2025	96.06	0.36	0.0157	96.44	0.59	194.74	0.01	ND	ND	0.13	ND	3.40	0.71	4.69
M5	19 Mar 2025	96.36	0.39	0.0180	96.78	0.58	194.42	0.02	ND	ND	0.11	ND	3.07	0.70	4.68
M6	25 Mar 2025	96.42	0.42	0.0212	96.86	0.58	194.27	0.02	ND	ND	0.16	ND	2.93	0.70	4.68
M7	4 Apr 2025	96.02	0.41	0.0201	96.46	0.59	194.81	0.02	ND	ND	0.11	ND	3.39	0.71	4.69
M8	24 Apr 2025	95.62	0.42	0.0216	96.07	0.59	195.29	0.02	ND	ND	0.10	ND	3.79	0.71	4.70
M9	12 May 2025	95.76	0.47	0.0250	96.27	0.59	195.15	0.02	ND	ND	0.09	ND	3.60	0.71	4.70
M10	26 Jun 2025	95.62	0.45	0.0222	96.11	0.59	195.28	0.02	ND	ND	0.10	ND	3.75	0.71	4.70
M11	21 Jul 2025	95.81	0.46	0.0251	96.31	0.59	195.04	0.02	ND	ND	0.12	ND	3.53	0.71	4.70
M12	15 Aug 2025	93.1045	0.4779	0.0267	93.62	0.60	192.71	0.02	ND	ND	3.15	ND	3.19	0.72	4.65
M13	20 Sep 2025	93.2635	0.5166	0.0309	93.88	0.61	195.65	0.02	ND	ND	1.35	ND	4.73	0.73	4.71

Non hydrocarbon gases also changed during production. CO₂ ranged from 1.88 to 4.73 mol % in Well M, and N₂ increased to 3.15 mol % on 15 Aug 2025 before remaining elevated at 1.35 mol % on 20 Sep 2025 (Table ). The late stage decrease in CH₄, together with the increase in N₂ and CO₂, indicates greater gas compositional heterogeneity during later flowback. This change did not alter the methane dominated nature of the gas, but it is relevant to water chemistry interpretation. The measurable CO₂ in the produced gas supports the need to examine CO₂ related acidification and carbonate dissolution as contributors to Ca²⁺ release and hardness accumulation. ,

The single well data were compared with representative gas quality data from the Benxi No. 8 reservoir (Table ). At the reservoir scale, CH₄ ranged from 90.29 to 96.80 mol %, total hydrocarbon content ranged from 91.37 to 97.29 mol %, N₂ ranged from 0.06 to 3.15 mol %, and CO₂ ranged from 1.88 to 8.51 mol %. The Well M data fall within these ranges, indicating that the observed temporal variation is consistent with the broader gas composition window of the Benxi No. 8 reservoir. CBM from this reservoir was therefore methane dominated, but the time series from Well M still provides a gas phase constraint for the following water quality analysis, where increasing salinity, hardness, acidity and Fe²⁺ mobilization occurred during the same production period. ,

2. Representative Gas-Quality Characteristics of the Benxi No. 8 Reservoir.

parameter	representative range	unit
CH4	90.29–96.80	mol %
C2H6	0.23–2.59	mol %
C3H8	0.01–0.33	mol %
iC4H10	0.0017–0.0842	mol %
nC4H10	ND–0.0554	mol %
iC5H10	0.0005–0.0313	mol %
nC5H12	ND–0.1413	mol %
neoC5H12	ND–0.0048	mol %
C6+	ND–0.0446	mol %
total hydrocarbon	91.37–97.29	mol %
relative density	0.57–0.64	-
density	0.69–0.77	kg/m3
critical temperature	192.34–201.26	K
He	0.0185–0.0353	mol %
H2	ND–0.0309	mol %
O2	ND	mol %
N2	0.0682–3.1597	mol %
CO	ND	mol %
CO2	1.88–8.51	mol %
higher heating value	34.14–37.11	MJ/m3
lower heating value	30.76–33.46	MJ/m3
critical pressure	4.64–4.83	MPa

4.2. Water Quality Evolution of CBM-FFFs during Flowback

The water quality evolution of CBM-FFFs from Well M was evaluated together with the production history, gas composition, coal fine release and reservoir scale water chemistry. The production response defined the hydraulic background of the returned fluids (Figure ). The coal fine data provided the SS background for treatment design (Figure ). The time-resolved water chemistry of Well M captured the geochemical transition during flowback (Table ), and the representative water quality range of the Benxi No. 8 reservoir defined the broader hydrochemical envelope (Table ). These data sets show that the treatment challenge developed from increasing salinity and hardness-forming ions, elevated iron and SS, decreasing pH, and persistent TOC.

4.

Time series of tubing pressure, casing pressure, gas rate and liquid rate for Well M during CBM flowback.

5.

Particle size distribution of coal fines in CBM-FFFs.

4. Representative Water Quality Characteristics of CBM-FFFs from the Benxi No. 8 Reservoir.

parameter	representative range or characteristic	unit
pH	3.97–6.34	-
TOC	101.78–6530.57	mg/L
density	1.0013–1.1914	g/cm3
ΣNa+ + K+	9700.32–57008.77	mg/L
Ca2+	3092.67–48643.10	mg/L
Mg2+	297.80–8826.32	mg/L
Fe2+	ND–3328.40	mg/L
Fe3+	ND–636.83	mg/L
Cl–	19121.42–257522.52	mg/L
SO4 2–	11.74–356.17	mg/L
CO3 2–	ND	mg/L
HCO3 –	30.68–2055.76	mg/L
OH–	ND	mg/L
TM	28675.91–359879.48	mg/L

The production history of Well M shows why the source contribution of the returned water changed with time (Figure ). The well first entered a liquid unloading stage, during which the daily liquid rate reached 21.6 m³/d and both casing and tubing pressures declined rapidly. This stage was dominated by the removal of retained fracturing fluid and near wellbore water. During the following shut in period, casing pressure recovered to about 16.9 MPa, indicating partial pressure recovery in the reservoir system. After the well was reopened, pressure fluctuation and intermittent liquid surges showed that drainage remained unstable. Around 13 Sep 2024, tubing pressure fell to 4.35 MPa while casing pressure remained at 15.1 MPa, indicating strong drawdown within the production string. After the wellhead separation system was introduced around 13 May 2025, casing and tubing pressures stabilized near 12.44 and 8.21 MPa, and the liquid rate was mainly controlled within 0.7 to 2.0 m³/d. This production sequence suggests that the returned water shifted from retained fracturing fluid and near wellbore water toward a larger contribution from formation brine, mobilized fines and gas water reaction products.

The single well water chemistry shows three stages of flowback evolution (Table ). The first stage occurred from 17 Jan to 1 Feb 2025 and represents an early, relatively dilute return. Total mineralization (TM) increased from 34502.17 to 71561.02 mg/L. Cl^– increased from 20121.26 to 43029.16 mg/L, ΣNa⁺ + K⁺ increased from 8638.96 to 19156.70 mg/L, Ca²⁺ increased from 3191.37 to 5402.78 mg/L and Mg²⁺ increased from 744.49 to 1778.88 mg/L. pH remained near neutral, ranging from 6.34 to 6.56. TOC showed the opposite trend, decreasing from 5053.62 to 3553.51 mg/L. This early TOC decline is consistent with the return, dilution and transformation of residual organic additives from the fracturing fluid, while the dissolved salt load had only begun to increase. At this stage, the water still fell near the lower end of the reservoir water quality envelope for salinity and hardness (Table ).

A sharp geochemical transition occurred by 19 Feb 2025 (Table ). TM increased to 186685.99 mg/L, about 5.41 times the value on 17 Jan. Cl^– increased to 116054.01 mg/L, about 5.77 times the initial value. Ca²⁺ increased to 23577.06 mg/L, about 7.39 times the initial value, while Mg²⁺ increased to 2372.05 mg/L. TOC decreased further to 2236.24 mg/L. From 19 Feb to 12 May 2025, the water remained in a high salinity window. TM stayed between 186685.99 and 210536.60 mg/L, and Cl^– ranged from 117733.49 to 131049.36 mg/L during this period. Ca²⁺ remained high, ranging from 23577.06 to 38807.46 mg/L, and Mg²⁺ ranged from 2372.05 to 3267.87 mg/L. TOC decreased from 2236.24 to 1218.83 mg/L and then approached a stable background near 1200 mg/L. This transition marks the shift from injected fluid influenced return water to water with a stronger formation brine contribution, as indicated by the parallel enrichment of Cl^–, ΣNa⁺ + K⁺, Ca²⁺ and Mg²⁺. The parallel enrichment of Cl^–, ΣNa⁺ + K⁺, Ca²⁺ and Mg²⁺ shows that salinity increase and hardness accumulation occurred together. The declining TOC follows a different control and is more consistent with residual fracturing fluid organics being returned, diluted or transformed during flowback.

The third stage occurred from 26 Jun to 20 Sep 2025 and represents the most chemically aggressive water quality window (Table ). TM increased to 277774.01 to 324236.66 mg/L, and Cl^– increased to 172441.53 to 199918.01 mg/L. ΣNa⁺ + K⁺ rose to 62361.42 to 79300.13 mg/L. Ca²⁺ remained high, ranging from 30049.98 to 39478.80 mg/L, and Mg²⁺ increased to 3602.13 to 6059.27 mg/L. During the same period, pH decreased from 4.93 to 4.17, and Fe²⁺ increased sharply from 871.96 to 2401.55 mg/L. TOC remained within a narrow range of 1212.86–1248.38 mg/L. This stage produced a saline, low pH, iron ions rich and hardness rich brine with persistent TOC. The increase in Fe²⁺ is important for treatment because it raises NaClO and NaOH demand during pretreatment and can promote secondary precipitation and membrane fouling if it enters the softening unit. Persistent TOC also matters because residual organic matter can increase oxidant demand, enhance particle adhesion and reduce the hydraulic stability of ceramic membrane filtration.

The reservoir scale data confirm that this single well evolution occurred within a broader saline and hardness rich water system (Table ). CBM-FFFs from the Benxi No. 8 reservoir were weakly acidic to acidic, with pH ranging from 3.97 to 6.34. TM ranged from 28675.91 to 359879.48 mg/L. Cl^– was the dominant anion, ranging from 19121.42 to 257522.52 mg/L. The main cations were ΣNa⁺ + K⁺, Ca²⁺ and Mg²⁺, with ranges of 9700.32 to 57008.77 mg/L, 3092.67 to 48643.10 mg/L and 297.80 to 8826.32 mg/L, respectively. TOC ranged from 101.78 to 6530.57 mg/L. CO₃ ^2– and OH^– were not detected, and HCO₃ ^– remained low at 30.68 to 2055.76 mg/L. This water chemistry indicates low carbonate alkalinity, high ionic strength, high-hardness and variable TOC. Any increase in formation water contribution during flowback would therefore increase salinity and scaling risk, while residual TOC would add an additional constraint on oxidation, settling and membrane operation.

The gas composition of Well M provides a useful constraint on the source and reaction controls behind the water chemistry transition (Table ). The produced gas remained CH₄ dominated, but the late stage showed lower CH₄ and higher non hydrocarbon gas contents. CO₂ reached 4.7304 mol % in the final gas sample, while N₂ increased to 3.15 mol % in August and remained elevated at 1.35 mol % in September. This gas phase change occurred during the same period when TM, Cl^–, ΣNa⁺ + K⁺, Ca²⁺, Mg²⁺ and Fe²⁺ were high in the water phase (Table ). The increase in Cl^– and ΣNa⁺ + K⁺ is consistent with a stronger formation brine contribution because these ions mainly trace saline water mixing. The parallel increase in Ca²⁺ and Mg²⁺ shows that brine inflow was accompanied by hardness accumulation. The presence of CO₂ in the gas and the decrease in pH suggest that CO₂ related acidification may also have promoted carbonate mineral dissolution and enhanced the release of Ca²⁺ and Mg²⁺. , The TOC trend follows a separate pattern. TOC was highest during early return and then decreased to a persistent late stage background, suggesting that residual fracturing fluid derived organics or their transformation products remained present after the inorganic ion chemistry became strongly influenced by formation derived brine.

The coal fine data show that the returned water also carried a solid phase that affected treatment design (Figure ). The particle size distribution was strongly right skewed. Most of the particle volume occurred in the tens of micrometres range, with a dominant peak around 80–100 μm. The cumulative curve increased rapidly once particle size exceeded several tens of micrometres, showing that the coarse fraction contributed most of the SS volume. The fine fraction below about 10 μm remained important because these particles can stay dispersed, sustain turbidity, adsorb dissolved components, provide nucleation sites for inorganic precipitation and enter the membrane filtration stage. This particle structure explains why gravity based pretreatment can reduce part of the solids load, while a finer separation step is still required after softening.

These water quality trends define the treatment logic used in the following sections. Early CBM-FFFs had lower salinity and hardness but higher TOC, making clarification and organic load reduction important. Middle and late stage CBM-FFFs showed formation brine contribution, hardness accumulation, acidification, Fe²⁺ mobilization, persistent TOC and SS release at the same time. A fixed treatment route would therefore be poorly matched to the changing feed. Alkaline oxidation pretreatment was needed to remove iron ions, reduce unstable SS and lower part of the organic load. CMBS was needed to convert Ca²⁺ and residual Mg²⁺ into separable carbonate precipitates. Crossflow ceramic membrane filtration was needed to remove both residual particles and the precipitates generated during hardness control. The linked production, chemistry and particle data provide the basis for the CO₂ dissolution experiment in Section and for the pilot treatment results in Section .

4.3. CO₂ Enhanced Carbonate Dissolution and Its Contribution to High Salinity and Hardness

The water chemistry of Well M showed a progressive increase in TM and hardness during flowback, accompanied by lower pH and measurable CO₂ in the produced gas. These field observations suggest that the increase in TM and hardness was controlled by more than one process. The enrichment of conservative ions, especially Cl^– and ΣNa⁺ + K⁺, mainly reflects increasing formation brine inflow. In parallel, CO₂ bearing water may have promoted carbonate dissolution and released additional Ca²⁺ and Mg²⁺ into the aqueous phase, thereby increasing hardness and ionic strength. A pressurized dissolution experiment was therefore conducted with a simulated formation brine that initially contained no detectable Ca²⁺ and a fixed mass of carbonate bearing coal seam drill cuttings. The increase in dissolved Ca²⁺ was used to evaluate the conversion of carbonate minerals in the drill cuttings into aqueous Ca²⁺ under CO₂ bearing conditions. The experiment was performed at one representative high salinity level of 80,000 mg/L NaCl. Salinity was not varied because the test was designed to isolate the effect of CO₂ pressure under a fixed brine matrix. The results therefore verify pressure dependent CO₂ induced carbonate dissolution but do not quantify how salinity itself changes carbonate dissolution kinetics.

Dissolved CO₂ promoted Ca²⁺ release from the carbonate bearing drill cuttings at 50 °C and near neutral initial pH (Figure ). At 7 MPa, Ca²⁺ increased rapidly during the early stage, reached 933.27 mg/L, and then remained at 881.03 mg/L on day 7. When CO₂ pressure increased to 11 MPa, the maximum Ca²⁺ concentration increased to 1295.09 mg/L, and the day 7 concentration remained high at 1267.50 mg/L. At 15 MPa, Ca²⁺ further increased to 1322.73 mg/L and remained at 1287.40 mg/L on day 7. The pressure response was stronger from 7 to 11 MPa than from 11 to 15 MPa. Based on day 7 values, Ca²⁺ increased by about 43.9% from 7 to 11 MPa, but only by about 1.6% from 11 to 15 MPa. This pattern shows that higher CO₂ pressure enhanced the conversion of solid carbonate phases into dissolved Ca²⁺, although the response became less pressure sensitive above 11 MPa under the tested batch conditions. ,

6.

CO₂ pressure dependent Ca²⁺ release from carbonate bearing drill cuttings in simulated formation water.

The weaker increase at higher pressure should be interpreted as a finite batch conversion response, not as a universal solubility limit for the reservoir system. The experiment used a fixed water volume, a fixed mass of drill cuttings and a closed reactor. The measured Ca²⁺ concentration therefore reflected the amount of carbonate dissolution that occurred within this limited water solid system. In field CBM-FFFs, Ca²⁺ concentrations can be higher because fresh formation brine continues to enter the wellbore, water rock interaction continues during flowback, the reactive mineral inventory is larger, and the contact time among water, gas and rock is longer. CO₂ enhanced carbonate dissolution therefore explains part of the hardness increase, while the increase in TM still requires formation brine inflow, especially for Cl^– and ΣNa⁺ + K⁺.

The dissolution experiment provides a mechanistic link to the field evolution described in Section . In the late flowback stage, Well M showed higher TM, higher Ca²⁺ and Mg²⁺, lower pH and measurable CO₂ in the gas phase. The pressurized test confirms that CO₂ bearing water can mobilize Ca²⁺ from carbonate minerals. This pathway increases hardness and adds to the ionic load of CBM-FFFs. It does not account for all dissolved ions in the returned water. Conservative ions such as Cl^– and ΣNa⁺ + K⁺ mainly trace formation brine mixing. The CO₂ dissolution pathway instead explains why hardness forming ions increased together with acidification during later production.

This pressure dependent dissolution behavior also affects treatment design. Under high CO₂ pressure, carbonate bearing solids can be converted more readily into dissolved Ca²⁺. During flowback, pressure decline, CO₂ degassing and pH adjustment may then shift the carbonate system toward supersaturation and increase CaCO₃ scaling risk. The results support a treatment strategy that targets hardness conversion and scale control, rather than clarification alone. This provides the basis for the CMBS stage, where dissolved Ca²⁺ is intentionally converted into separable carbonate precipitates under controlled operating conditions.

4.4. Pilot Treatment of Later Stage CBM-FFFs

4.4.1. Alkaline Oxidation Pretreatment for Feed Stabilization and Iron Removal

Alkaline oxidation pretreatment was used as the first control step because later stage CBM-FFFs contained high iron ions, low pH and unstable SS, as shown by the water quality evolution in Section . Direct feeding of this water into CMBS would increase NaOH demand, interfere with carbonate precipitation and carry additional solids into the ceramic membrane unit. The alkaline oxidation pretreatment step was therefore designed to stabilize the feed before hardness control. Its purpose was to remove iron ions, readily settleable SS, most Mg²⁺ and part of the TOC load while preserving the saline matrix.

The iron ions speciation and removal performance were evaluated during alkaline oxidation pretreatment operation (Figure a,b). Influent iron ions changed substantially with flowback time. Fe²⁺ ranged from 27.37 to 2408.53 mg/L, while Fe³⁺ ranged from below the detection limit to 488.69 mg/L. Total influent iron ions ranged from 54.73 to 2408.53 mg/L, corresponding to a more than 40 fold change in iron ions loading. The highest iron ions concentrations occurred in later samples, consistent with the low pH and Fe²⁺ rich water described in Section . Even with this feed variation, effluent total iron ions remained near the detection limit, with a maximum of 1.86 mg/L. Iron ions removal efficiency remained between 98.08% and 100.00% throughout operation (Figure e). This response shows that alkaline oxidation pretreatment provided a stable barrier against one of the most reactive and variable components of CBM-FFFs.

7.

Alkaline oxidation pretreatment performance and chemical dosage variation during CBM-FFFs treatment. (a) Influent Fe²⁺ and Fe³⁺ concentrations, (b) influent and effluent total iron ions concentrations, (c) NaOH dosage, (d) influent pH, (e) Fe removal efficiency, (f) 10 wt % NaClO dosage and (g) Influent and effluent TOC concentrations.

The chemical dosage profiles explain how this stability was achieved (Figure c,d,f). Influent pH ranged from 5.36 to 6.34 during the earlier period and then declined to 4.17 to 4.93 during the later period. This pH decrease coincided with higher Fe²⁺ loading. NaOH was added to raise pH to approximately 12.0, which supplied alkalinity for iron ions hydrolysis and precipitation. The same alkaline condition also shifted most Mg²⁺ removal to alkaline oxidation pretreatment through Mg­(OH)₂ precipitation. This design was important because the following CMBS unit was more suitable for CaCO₃ precipitation and less efficient for Mg²⁺ removal through carbonate precipitation alone. By removing most Mg²⁺ before CMBS, alkaline oxidation pretreatment reduced the Mg²⁺ burden entering the softening stage and allowed CMBS to focus mainly on residual Ca²⁺ conversion.

NaClO demand increased with the rise in Fe²⁺ loading (Figure f). The 10 wt % NaClO dosage increased from 0.32 to 2.72 L/m³ in the earlier period to 5.09 to 14.06 L/m³ in the later period. After day 161, when total influent iron ions exceeded 800 mg/L, NaClO demand increased sharply. This change shows that NaClO served two coupled functions during alkaline oxidation pretreatment. It oxidized Fe²⁺ to Fe³⁺, enabling hydrolysis, precipitation and settling, and it also degraded part of the reactive organic matter in the influent. The TOC trend supports this interpretation (Figure g). Influent TOC decreased over operation and approached a lower late stage level, while effluent TOC remained consistently below the influent value. This pattern indicates that alkaline oxidation and settling removed part of the organic load together with iron ions and SS.

Alkaline oxidation pretreatment therefore redefined the feed entering the downstream units. Iron was removed mainly through NaClO oxidation followed by alkaline hydrolysis and precipitation. Mg²⁺ was preferentially reduced under high pH conditions. Part of the TOC was degraded or transferred into the settled phase. Removing these iron ions rich, Mg bearing and organic bearing solids before CMBS reduced chemical competition, lowered the risk of sludge carryover and limited the fouling load entering ceramic membrane filtration. Residual iron ions and organic matter can consume reagents, provide nucleation sites, increase particle adhesion and reduce hydraulic reversibility during membrane operation. By reducing total iron ions to below 1.86 mg/L or lower, lowering TOC and shifting most Mg²⁺ removal upstream without adding coagulant or polymer, alkaline oxidation pretreatment converted a highly variable CBM-FFFs feed into a more stable influent for CMBS and downstream ceramic membrane filtration.

4.4.2. CO₂ Microbubble Softening

After alkaline oxidation pretreatment, the main treatment target shifted from iron ions removal to hardness control. The late stage CBM-FFFs still contained high Ca²⁺ under high salinity and low pH conditions, while most Mg²⁺ had already been removed during alkaline oxidation pretreatment. The CO₂ pressurization experiment also showed that CO₂ bearing water can mobilize Ca²⁺ from carbonate minerals. CMBS was therefore used as a controlled hardness conversion step, where residual Ca²⁺ was converted into separable CaCO₃ precipitates by regulating CO₂ dosage and pH.

Ca²⁺ removal increased strongly with CO₂ dosage, and CMBS performed much better than bulk CO₂ bubbling under the same operating conditions (Figure a). In the bulk bubbling system, Ca²⁺ removal increased slowly from 0.60% at 1 L/min to 40.82% at 60 L/min. In the CMBS system, Ca²⁺ removal increased from 1.88% at 1 L/min to 99.62% at 60 L/min. The difference between the two systems became larger as CO₂ dosage increased. At 20 L/min, CMBS removed 33.26% of Ca²⁺, compared with 14.78% for bulk bubbling. At 40 L/min, the corresponding values were 66.48% and 28.80%. At 60 L/min, CMBS almost exhausted dissolved Ca²⁺, while bulk bubbling still left most Ca²⁺ in solution. The residual Ca²⁺ concentration followed the same dosage controlled trend (Figure b). These data indicate that microbubble delivery increased the effective CO₂ dissolution and carbonate formation capacity of the system.

8.

Effect of CO₂ dosage on Ca²⁺ removal during bulk CO₂ bubbling and microbubbles. (a) Ca²⁺ removal efficiency and (b) residual Ca²⁺ concentration as a function of CO₂ dosage.

pH provided the second control on Ca²⁺ removal (Figure a). Under bulk bubbling, Ca²⁺ removal increased from 0.89% at pH 7.0 to 43.50% at pH 9.0 and reached only 50.03% at pH 9.5. Further pH increase did not produce a clear improvement, with removal remaining between 45.53% and 49.31% from pH 10.0 to 12.0. Under CMBS, Ca²⁺ removal increased much more sharply. It rose from 1.53% at pH 7.0 to 34.24% at pH 7.5, 78.34% at pH 8.0 and 84.62% at pH 8.5. Removal reached 99.89% at pH 9.0 and remained high from pH 9.5 to 12.0. The softening reaction was therefore operated mainly at pH 9 to 10, where high Ca²⁺ removal could be achieved without excessive alkalinity input.

9.

pH dependent Ca²⁺ removal and carbonate speciation during bulk CO₂ bubbling and CO₂ microbubbles. (a) Ca²⁺ removal efficiency and (b) HCO₃ ^– and CO₃ ^2– concentrations as a function of pH.

The carbonate speciation data explain this pH response (Figure b). In the bulk bubbling system, HCO₃ ^– increased from 22.35 mg/L at pH 7.0 to 56.21 mg/L at pH 8.0 and then decreased to 0.17 mg/L at pH 12.0. CO₃ ^2– remained low below pH 9.0 and increased mainly under stronger alkaline conditions. In the CMBS system, the dissolved inorganic carbon pool was much larger. HCO₃ ^– reached 112.54 mg/L at pH 8.0, and CO₃ ^2– increased to 12 mg/L at pH 9.0, 78 mg/L at pH 10.0, 232 mg/L at pH 11.0 and 298 mg/L at pH 12.0. Higher HCO₃ ^– and CO₃ ^2– availability under CMBS provides a stronger precipitation driving force for CaCO₃ formation, which accounts for the sharp increase in Ca²⁺ removal near pH 9.0.

The precipitates formed during CMBS had a particle structure suitable for downstream separation (Figure ). The particle size distribution was bimodal. A fine population appeared at about 0.7 to 1.0 μm and contributed about 1.0% to 1.2% of the particle volume. The dominant population appeared at about 8.5 to 10.0 μm, with a maximum volume fraction of about 8.8% to 9.1% (Figure a). Increasing pH caused only a slight shift of the dominant peak toward larger sizes, suggesting that pH mainly promoted particle growth and aggregation after nucleation. The SEM image shows irregular granular particles and compact aggregates, with fine particles attached to larger clusters (Figure b). This morphology is favorable for ceramic membrane retention, although the fine fraction explains why softening alone could not produce a clarified effluent.

10.

Particle size distribution and SEM morphology of CaCO₃-rich precipitates produced by CMBS, showing (a) particle size distributions of precipitates formed at different pH values and (b) SEM image of the recovered precipitates.

The mineral identity of the precipitates was confirmed by TGA-DTG (Figure a) and XRD (Figure b). The TGA curve changed little below about 650 °C, indicating limited adsorbed water or volatile matter. The main mass loss occurred between about 680 and 800 °C, with a DTG peak at about 755 to 765 °C. The total mass loss was about 45%, close to the theoretical CO₂ release from CaCO₃ decomposition. The XRD pattern showed a strong reflection near 29.4° 2θ and characteristic peaks near 39–49°, 57–61° and 64–66°, matching crystalline CaCO₃. No dominant impurity peak was observed. These results confirm that CMBS removed hardness mainly by converting dissolved Ca²⁺ into crystalline CaCO₃.

11.

TGA-DTG (a) and XRD (b) patterns of the precipitates formed by CMBS.

CMBS therefore served two process functions. It reduced residual Ca²⁺ to a low concentration through controlled carbonate precipitation, and it converted dissolved hardness into a particulate phase that could be separated in the following units. The highest Ca²⁺ removal reached 99.62% at a CO₂ dosage of 60 L/min and 99.89% at pH 9.0. Because residual hardness responded to CO₂ dosage and pH, the softening intensity can be adjusted for different field reuse targets. After softening, the slurry was settled, and the softened supernatant was introduced into the ceramic membrane unit for final clarification. This step remained necessary because the settled softening effluent still contained CaCO₃ rich particles, including a fine fraction near 1 μm, which could not be reliably removed by gravity settling alone.

4.4.3. Crossflow Ceramic Membrane Filtration of Softened CBM-FFFs

After CMBS and gravity settling, the softened supernatant still contained residual CaCO₃ rich particles and fine SS that could not be removed reliably by settling alone. Crossflow ceramic membrane filtration was used as the final clarification step to remove these particles while allowing the dissolved salt matrix to pass through. This design was consistent with the treatment target of this work, which was hardness control, scaling risk reduction and solids removal rather than deep desalination.

The hydraulic response of the ceramic membrane was evaluated under constant flux operation at 566.67, 666.67, and 800 LMH (Figure ). During the test, concentrate recirculation continuously increased SS in the slurry, creating an increasingly severe particle loading condition for the membrane. At 566.67 LMH, cumulative permeate volume increased from 5.67 to 986.00 L over 174 min, while SS increased from 3680.12 to 17486.93 mg/L. TMP increased to about 0.151 to 0.152 MPa within each filtration cycle and returned to about 0.050 MPa after physical cleaning. At 666.67 LMH, SS further increased from 17630.41 to 22275.89 mg/L, while TMP cycled between about 0.065 and 0.150 MPa. At 800 LMH, SS reached 23613.28 mg/L, and TMP still remained within about 0.075 to 0.151 MPa before cleaning. These results show that crossflow ceramic membrane filtration remained hydraulically stable as SS increased from 3680.12 to 23613.28 mg/L, demonstrating its ability to clarify CaCO₃ rich softened water under extremely high suspended solid loading.

12.

Operational performance and effluent quality of the crossflow ceramic membrane under constant-flux operation at 566.67, 666.67, and 800 LMH.

The imposed flux mainly controlled how quickly filtration resistance accumulated. At 566.67 LMH, TMP required about 57 to 58 min to rise from the post cleaning baseline to the upper operating level. At 666.67 LMH, this interval shortened to about 24 to 25 min. At 800 LMH, it decreased further to about 12 to 13 min. The membrane was still able to operate at 800 LMH, but the higher flux sharply increased the cleaning frequency. The more favorable operating window was therefore 566.67 to 666.67 LMH, where high throughput was maintained and pressure recovery after physical cleaning remained reproducible. This behavior shows that the ceramic membrane can tolerate CaCO₃ rich slurry at very high SS, provided that flux and cleaning frequency are matched to the particle loading.

The constant pressure tests further showed the effect of hydraulic driving force on flux decline (Figure ). At 0.08 MPa, the initial flux was 750.00 LMH and decreased to 400.00 LMH within the first cycle. After cleaning, the flux recovered to about 750.00 to 766.67 LMH and then followed a similar decline. At 0.10 MPa, the recovered flux reached 950.00 LMH, but the late cycle flux again decreased to about 436.67 to 441.67 LMH. At 0.12 MPa, the initial flux increased to 1000.00 LMH, while the late cycle flux still converged near 441.67 LMH. During the constant pressure test, SS increased from 3799.59 to 28896.35 mg/L. Increasing TMP from 0.08 to 0.12 MPa increased the initial flux, but it did not proportionally improve the sustainable late cycle flux. The convergence of late cycle flux under different TMP values indicates that particle deposition, rather than hydraulic pressure alone, controlled the limiting resistance.

13.

Permeate flux and effluent quality during constant-pressure crossflow ceramic membrane filtration at TMP values of 0.08, 0.10, and 0.12 MPa.

The pure water test confirmed that the hydraulic limitation came from slurry fouling rather than insufficient intrinsic membrane permeability (Figure ). Pure water flux increased linearly from 345 LMH at 0.01 MPa to 2313 LMH at 0.18 MPa. The fitted relationship was J = 11499TMP + 226.58, with R ² = 0.9916. Based on this relationship, the expected pure water fluxes at 0.08, 0.10, and 0.12 MPa were about 1147, 1377, and 1607 LMH, respectively. These values were much higher than the stabilized fluxes measured during softened water filtration. The constant flux tests also operated below the clean water flux envelope, yet TMP still increased within each cycle. The operating resistance was therefore governed by CaCO₃ rich particle deposition on or near the membrane surface.

14.

Pure water flux of crossflow ceramic membrane at different TMP.

Fouling model fitting was used to identify the dominant resistance pathway under each hydraulic mode. Under constant flux at 566.67 LMH, the complete blocking model gave the best fit, with TMP = 0.00080t + 0.08203 and R ² = 0.9927 (Figure ). The standard blocking and cake filtration models also showed high correlations, with R ² values of 0.9795 and 0.9700, while the intermediate blocking model gave R ² = 0.9584. This fitting order indicates that pore entrance coverage was the main resistance pathway under constant flux. The imposed flux maintained particle transport toward the membrane surface, so fine CaCO₃ particles were continuously driven toward membrane openings as TMP increased. SEM observations support this interpretation. The fouled membrane surface was covered by irregular carbonate crystals, fine particles and compact agglomerates, and some fine particles occupied surface voids and pore entrance regions (Figure b). This morphology is consistent with complete blocking dominated behavior, while the presence of larger aggregates explains why cake filtration also showed a high correlation.

15.

Linear fitting of CBM-FFFs fouling models under a constant flux of 566.67 LMH, where (a) represents complete blocking, (b) represents standard blocking, (c) represents intermediate blocking, and (d) represents cake filtration.

17.

SEM comparison of (a) pristine ceramic membrane, (b) fouled membrane under constant flux (566.67 LMH), and (c) fouled membrane under TMP of 0.1 MPa.

Under constant pressure at 0.1 MPa, the dominant fouling pathway shifted toward cake formation. The cake filtration model gave the best fit, with 1/J ² = 4.4752 × 10^–8 t + 2.6734 × 10^–6 and R ² = 0.9177 (Figure ). The intermediate blocking model gave R ² = 0.8898, followed by standard blocking with R ² = 0.8739 and complete blocking with R ² = 0.8567. In this mode, flux decreased as resistance developed, reducing the convective driving force toward pore entrances and favoring particle accumulation on the membrane surface. SEM images showed a more continuous particle layer composed of fine grains, blocky carbonate fragments and agglomerated crystals (Figure c). These deposits bridged the membrane surface and formed a heterogeneous porous cake, supporting the model result that constant pressure filtration was controlled mainly by external cake growth.

16.

Linear fitting of CBM-FFFs fouling models under TMP of 0.1 MPa, where (a) represents complete blocking, (b) represents standard blocking, (c) represents intermediate blocking, and (d) represents cake filtration.

The combined hydraulic, permeability, model fitting and SEM results show that crossflow ceramic membrane filtration is suitable for final clarification of softened CBM-FFFs. Constant flux operation maintained stable water production but required more frequent cleaning at higher flux. Constant pressure operation allowed flux decline and shifted the main resistance toward cake growth. In both modes, the membrane retained CaCO₃ rich particles under high SS loading, and physical cleaning restored much of the hydraulic performance. These results demonstrate that the ceramic membrane unit is not simply a polishing step. It is the key separation barrier that converts the softened, particle rich slurry into a clarified effluent while preserving the dissolved salt matrix.

4.5. Effluent Quality Across the Staged Treatment Process

The effluent quality of the staged pilot process was evaluated using two later stage CBM-FFF batches with different initial compositions (Table ). The visual sequence was consistent in both batches (Figure ). Raw water was yellow and turbid, reflecting the combined presence of SS, iron ions and organic matter. After alkaline oxidation pretreatment, the water became lighter but remained slightly turbid, which shows that alkaline oxidation pretreatment reduced the most reactive fraction without producing a fully clarified effluent. CO₂ microbubble softening then produced a milky suspension because dissolved hardness was converted into carbonate precipitates. After ceramic membrane filtration, the effluent became visually clear. This sequence captures the role of each unit: alkaline oxidation pretreatment stabilizes the feed, CMBS converts dissolved hardness into particulate carbonate solids and ceramic membrane filtration completes solid liquid separation.

5. Geochemical Characteristics of Treated CBM-FFFs from the Benxi No. 8 Reservoir.

parameter	raw water		alkaline oxidation pretreatment effluent		softening effluent		membrane filtration effluent
sample batch	batch 1	batch 2	batch 1	batch 2	batch 1	batch 2	batch 1	batch 2
SS mg/L	1620.20	1501.72	126.26	105.21	25624.58	25126.65	ND	ND
pH	4.17	4.51	12.03	12.05	9.42	9.35	9.30	9.32
conductivity (ms/cm)	486.16	361.25	491.52	366.72	473.30	355.28	473.31	355.29
Cl– (mg/L)	199938.00	180693.97	200287.89	181193.81	200187.92	181143.83	200237.91	181118.84
HCO3 – (mg/L)	762.75	94.58	ND	ND	48.82	61.02	115.22	48.82
CO3 2– (mg/L)	ND	ND	732.12	90.02	114.02	120.02	48.82	126.02
OH– (mg/L)	ND	ND	166.70	170.10	ND	ND	ND	ND
SO4 2– (mg/L)	36.01	114.21	37.04	111.12	37.04	103.30	32.93	98.78
Fe2+ (mg/L)	2408.53	1658.05	ND	ND	ND	ND	ND	ND
Fe3+ (mg/L)	ND	ND	ND	ND	ND	ND	ND	ND
Ca2+ (mg/L)	38201.25	29934.75	36773.40	27304.50	75.15	60.12	75.15	57.62
Mg2+ (mg/L)	3608.52	3912.39	75.97	60.78	15.19	30.39	12.16	31.30
ΣNa+ + K+ (mg/L)	79318.37	75532.56	88364.16	86420.87	129834.21	117513.17	129871.33	117517.48
TM (mg/L)	324273.43	291940.51	326437.28	295351.20	330312.36	299031.85	330393.50	298977.31
TOC (mg/L)	1260.32	1293.15	756.21	771.69	350.62	360.82	349.62	358.20

18.

Visual comparison of water samples collected from sequential stages of the pilot treatment process, with raw water, pretreatment effluent, CO₂ softening effluent and membrane effluent shown from left to right in each batch: (a) batch 1 and (b) batch 2.

Alkaline oxidation pretreatment reduced the unstable feed load before softening. SS decreased from 1620.20 to 126.26 mg/L in Batch 1 and from 1501.72 to 105.21 mg/L in Batch 2. Fe²⁺ decreased from 2408.53 and 1658.05 mg/L to ND in both batches, and Fe³⁺ remained ND after treatment. TOC decreased from 1260.32 to 756.21 mg/L in Batch 1 and from 1293.15 to 771.69 mg/L in Batch 2. These changes show that alkaline oxidation and settling removed iron ions, readily settleable SS and part of the organic load before the water entered the softening unit. Mg²⁺ also decreased sharply during alkaline oxidation pretreatment, from 3608.52 to 75.97 mg/L in Batch 1 and from 3912.39 to 60.78 mg/L in Batch 2, confirming that high pH alkaline oxidation pretreatment shifted most Mg²⁺ removal upstream. Ca²⁺ decreased only slightly, from 38201.25 to 36773.40 mg/L in batch 1 and from 29934.75 to 27304.50 mg/L in batch 2, leaving residual Ca²⁺ as the main hardness target for CMBS.

CMBS produced the main Ca²⁺ removal step. Ca²⁺ decreased from 36773.40 to 75.15 mg/L in batch 1 and from 27304.50 to 60.12 mg/L in batch 2. Mg²⁺ was further reduced to 15.19 and 30.39 mg/L, respectively. At the same time, SS increased sharply to 25624.58 and 25126.65 mg/L because hardness removal generated a large amount of carbonate rich suspended solids. This increase does not indicate poorer treatment performance. It reflects the intended conversion of dissolved hardness into a separable particulate phase. TOC also decreased further to 350.62 and 360.82 mg/L, suggesting that part of the remaining organic matter was removed with, or incorporated into, the precipitated solids.

Ceramic membrane filtration removed the particles generated during CMBS and produced the final clarified effluent. SS was ND in both final effluents. Fe²⁺ and Fe³⁺ also remained ND. Ca²⁺ was maintained at 75.15 mg/L in batch 1 and decreased slightly to 57.62 mg/L in batch 2, while Mg²⁺ was 12.16 and 31.30 mg/L, respectively. TOC changed only slightly across the membrane step, from 350.62 to 349.62 mg/L in batch 1 and from 360.82 to 358.20 mg/L in batch 2. This result confirms that the ceramic membrane mainly acted as a particle retention barrier rather than a dissolved organic carbon removal unit.

The dissolved salt matrix was largely retained through the process. TM remained high, changing from 324273.43 and 291940.51 mg/L in the raw waters to 330393.50 and 298977.31 mg/L in the final effluents. Cl^– also remained high, and ΣNa⁺ + K⁺ increased after alkaline dosing. These changes are consistent with the design goal of the process. The pilot train did not aim to desalinate CBM-FFFs. It targeted hardness control, iron ions removal, SS removal and partial TOC reduction while preserving most dissolved salts. The final effluent therefore remained saline but had much lower scaling potential and no detectable SS or iron ions, making it more suitable for internal oilfield reuse where deep desalination is not required. ,

5. Conclusions

This study linked flowback water evolution with pilot scale hardness control for saline CBM-FFFs from the Benxi No. 8 reservoir in Ordos Basin. The gas phase remained CH₄ dominated during flowback, with CH₄ ranging from 93.10 to 96.80 mol % in Well M. The aqueous phase changed much more strongly. TM increased from 34502.17 to 324236.66 mg/L, pH decreased from 6.56 to 4.17, and Cl^– increased from 20121.26 to 199918.01 mg/L. Ca²⁺ increased from 3191.37 to 38216.28 mg/L, while Mg²⁺ reached 6059.27 mg/L in the late stage. TOC showed a different trajectory. It decreased from 5053.62 mg/L in the early return water to about 1200 mg/L in the late stage, but it remained persistent after the inorganic ion chemistry became strongly influenced by formation derived brine. These changes show that later stage CBM-FFFs evolved into a saline, low pH, iron ions rich and hardness rich fluid with residual organic load.

The increase in TM and hardness was mainly associated with formation brine inflow and CO₂ related carbonate dissolution. The strong enrichment of Cl^– and ΣNa⁺ + K⁺ indicates a growing contribution from formation derived brine. The simultaneous increase in Ca²⁺ and Mg²⁺ shows that brine inflow was accompanied by hardness accumulation. The pressurized dissolution experiment confirmed that dissolved CO₂ can mobilize Ca²⁺ from carbonate bearing drill cuttings. Ca²⁺ reached 881.03 mg/L at 7 MPa, 1267.50 mg/L at 11 MPa and 1287.40 mg/L at 15 MPa on day 7. The response was stronger from 7 to 11 MPa than from 11 to 15 MPa, indicating that CO₂ pressure enhanced carbonate conversion under the tested batch conditions. This mechanism explains part of the hardness increase, while conservative ions still mainly reflect formation brine mixing.

The pilot process translated this water quality evolution into a selective treatment route. Alkaline oxidation pretreatment stabilized the feed by removing Fe²⁺/Fe³⁺, readily settleable SS, most Mg²⁺ and part of the TOC load before CMBS. Even when influent total iron ions varied from 54.73 to 2408.53 mg/L, effluent total iron ions remained at or below 1.86 mg/L, with iron ions removal of 98.08% to 100.00%. High pH pretreatment shifted most Mg²⁺ removal upstream, reducing the burden on the softening unit. CMBS then provided the main Ca²⁺ conversion step. Ca²⁺ removal reached 99.62% at a CO₂ dosage of 60 L/min and 99.89% at pH 9.0. The generated precipitates showed micrometre scale aggregation, a main thermal decomposition interval near 680 to 800 °C and XRD features of crystalline CaCO₃, confirming that hardness removal was governed mainly by carbonate precipitation.

Crossflow ceramic membrane filtration served as the final solid retention barrier after hardness conversion. The membrane clarified CaCO₃ rich softened water under high SS loading, with SS increasing from 3680.12 to 23613.28 mg/L during constant flux operation while TMP remained recoverable after physical cleaning. Fouling behavior depended on hydraulic mode. Complete blocking dominated under constant flux at 566.67 LMH, while cake filtration dominated under constant pressure at 0.1 MPa. The final process produced clear, low hardness effluent while retaining the saline matrix. In the final effluent, SS was not detected, Fe²⁺ and Fe³⁺ were not detected, Ca²⁺ decreased to 75.15 and 57.62 mg/L, and Mg²⁺ decreased to 12.16 and 31.30 mg/L in the two treatment batches. TM remained high, confirming that the process did not perform deep desalination. This treatment route is therefore suited to internal oilfield reuse scenarios where the main requirements are hardness control, scale risk reduction, iron ions removal and solids removal rather than complete salt removal.

The main limitations of this work are the single well time series, two pilot treatment batches, the fixed salinity used in the CO₂ dissolution experiment and the short duration of membrane operation. Future work should test CO₂ induced carbonate dissolution across broader salinity, temperature and pressure windows, quantify reaction kinetics and saturation control, evaluate long-term ceramic membrane fouling and chemical cleaning, and assess reuse compatibility, precipitate recovery and process economics under continuous field operation.

Y. M.: Conceptualization, supervision, project administration, funding acquisition, methodology, writing review and editing. R. L.: Methodology, investigation, formal analysis, data curation, visualization, writing original draft, writing review and editing. D. W.: Resources, field coordination, investigation, project administration, technical support. S. H.: Resources, field sampling, pilot scale operation, investigation, technical support. J. D.: Field sampling, pilot scale operation, data collection, technical support. W. T.: Field sampling, pilot scale operation, data collection, technical support.

The authors declare no competing financial interest.